Systems Synthetic Biology System Models, User-Oriented Specifica

SYSTEMS BIOLOGY - THEORY, TECHNIQUES AND APPLICATIONS
SYSTEMS SYNTHETIC BIOLOGY

SYSTEM MODELS, USER-ORIENTED
SPECIFICATIONS, AND APPLICATIONS
No part of this digital document may be reproduced, stored in a retrieval system or transmitted in any form or
by any means. The publisher has taken reasonable care in the preparation of this digital document, but makes no
expressed or implied warranty of any kind and assumes no responsibility for any errors or omissions. No
liability is assumed for incidental or consequential damages in connection with or arising out of information
contained herein. This digital document is sold with the clear understanding that the publisher is not engaged in
rendering legal, medical or any other professional services.
SYSTEMS BIOLOGY - THEORY, TECHNIQUES
AND APPLICATIONS
Additional books in this series can be found on Nova’s website

under the Series tab.
Additional e-books in this series can be found on Nova’s website

under the eBook tab.
SYSTEMS BIOLOGY - THEORY, TECHNIQUES AND APPLICATIONS
SYSTEMS SYNTHETIC BIOLOGY

SYSTEM MODELS, USER-ORIENTED
SPECIFICATIONS, AND APPLICATIONS
BOR-SEN CHEN
AND
CHIH-YUAN HSU
New York
Copyright © 2017 by Nova Science Publishers, Inc.
All rights reserved. No part of this book may be reproduced, stored in a retrieval system or
transmitted in any form or by any means: electronic, electrostatic, magnetic, tape, mechanical
photocopying, recording or otherwise without the written permission of the Publisher.
We have partnered with Copyright Clearance Center to make it easy for you to obtain permissions
to reuse content from this publication. Simply navigate to this publication’s page on Nova’s
website and locate the “Get Permission” button below the title description. This button is linked
directly to the title’s permission page on copyright.com. Alternatively, you can visit
copyright.com and search by title, ISBN, or ISSN.
For further questions about using the service on copyright.com, please contact:
Copyright Clearance Center
Phone: +1-(978) 750-8400 Fax: +1-(978) 750-4470 E-mail: info@copyright.com.
NOTICE TO THE READER

The Publisher has taken reasonable care in the preparation of this book, but makes no expressed or
implied warranty of any kind and assumes no responsibility for any errors or omissions. No
liability is assumed for incidental or consequential damages in connection with or arising out of
information contained in this book. The Publisher shall not be liable for any special,
consequential, or exemplary damages resulting, in whole or in part, from the readers’ use of, or
reliance upon, this material. Any parts of this book based on government reports are so indicated
and copyright is claimed for those parts to the extent applicable to compilations of such works.
Independent verification should be sought for any data, advice or recommendations contained in
this book. In addition, no responsibility is assumed by the publisher for any injury and/or damage
to persons or property arising from any methods, products, instructions, ideas or otherwise
contained in this publication.
This publication is designed to provide accurate and authoritative information with regard to the
subject matter covered herein. It is sold with the clear understanding that the Publisher is not
engaged in rendering legal or any other professional services. If legal or any other expert
assistance is required, the services of a competent person should be sought. FROM A
DECLARATION OF PARTICIPANTS JOINTLY ADOPTED BY A COMMITTEE OF THE
AMERICAN BAR ASSOCIATION AND A COMMITTEE OF PUBLISHERS.
Additional color graphics may be available in the e-book version of this book.
Library of Congress Cataloging-in-Publication Data

ISBN: (eBook)
Published by Nova Science Publishers, Inc. † New York

CONTENTS
Preface ix
Chapter 1 Introduction 1
Chapter 2 Systematic Biological Filter Design with a Desired I/O Filtering
Response Based on Promoter-RBS Libraries 7
2.1. Introduction 7
2.2. Construction of Promoter-RBS Library 9
2.3. Systematic Design Methodology of the Synthetic Biological
Filter Based on Promoter-RBS Libraries 10
2.4. Biological Filter Design Based on Promoter-RBS Libraries
and Gene Circuit Topology 19
2.5. Discussion 23
2.6. Conclusion 26
2.7. Appendix 26
References 33
Chapter 3 Systematic Design Methodology for Robust Genetic Transistors 37
3.1. Introduction 37
3.2. Construction of the Promoter-RBS Libraries for Genetic
Transistors 39
3.4. Synthetic Genetic Transistor Design Examples
Based on Promoter-RBS Libraries 46
3.5. Discussion 50
3.6. Conclusion 51
3.7. Appendix 52
References 56
Chapter 4 Systematic Design of a Quorum Sensing-Based Biosensor
for Enhanced Detection of Metal Ion in Escherichia Coli 61
4.2. Methods 63
4.3. Results 69
4.4. Conclusion 73
4.5. Materials and Methods 74
vi Contents
4.6. Appendix 75
References 78
Chapter 5 Systematic Design of a Metal Ion Biosensor: A Multi-Objective
Optimization Approach 83
5.3. Results 95
5.4. Discussion 97
5.5. Conclusion 98
5.6. Appendix 98
References 103
Chapter 6 Systematic Approach to Escherichia Coli Cell Population
Control Using a Genetic Lysis Circuit 107
6.2. Methods 109
6.3. Results 115
6.4. Discussion 117
6.5. Conclusion 118
6.6. Appendix 119
References 119
Chapter 7 Engineering Bacteria to Search for Specific Concentrations
of Molecules by a Systematic Synthetic Biology Design Method 123
7.3. Results 132
7.4. Discussion 150
7.5. Conclusion 151
7.6. Appendix 151
References 167
Chapter 8 Construction of Promoter-RBS Libraries for
the Cyanobacterium Synechococcus SP. PCC 7942
and Their Applications to Systematic Synthetic Circuit
Design to Match User-Oriented Specifications 169
8.2. Methods 171
8.3. Results 178
8.4. Discussion 180
8.5. Conclusion 181
References 182
Chapter 9 A Robust Design of Quorum Sensing Symbiotic Ecosystems
Controlled by Small Molecule Regulators through Cell-Cell
Communication 185
9.2. Methods 187
Contents vii
9.3. Results 195

9.4. Discussion 198
9.5. Conclusion 200
9.6. Appendix 200
References 201
Chapter 10 Systematic Design of a Multicellular Molecular Communication
System with Desired Detection Capability, Transduction
Ability, and System Sensitivity 205
10.2. Methods 207
10.3. Discussion 228
10.4. Conclusion 229
10.5. Appendix. Dynamic Equation of Catalysis Process 230
References 231
About the Author 235
Index 237
PREFACE
A major goal of synthetic biology is to engineer artificial gene circuits with some desired
behaviors or products. One useful analogy to conceptualize the part of synthetic biology is the
electronic engineering hierarchy. At the bottom of the hierarchy of synthetic biology are
DNA, RNA, and proteins, analogous to the physical layer of transistors, capacitors, and
resistors in analog electronics. The next layer, the device layer, comprises biochemical
reactions that regulate the flow of information and manipulate biochemical process,
equivalent to engineered logic gates to perform computations. At the module layer, synthetic
biologists use a diverse library of biological devices to assemble complex pathways that
function like the integrated circuits (ICs) for filter, amplifier and computer.
However, these biological devices and modules are not independent biological objects
but with biological context and within cellular environments. In this book, the synthetic
genetic circuits are modeled by nonlinear stochastic systems in vivo, i.e., the biological
genetic systems suffer from intrinsic random fluctuation due to random genetic variations and
random environmental disturbance due to changes in the cellular context in host cells.
Therefore, our design purpose is to engineer a genetic circuit to achieve a desired behavior or
product despite intrinsic random fluctuation and random environmental disturbance.
For the systematic design of synthetic genetic circuits, we first construct several
promoter-RBS component libraries according to the promoter-RBS regulatory strengths based
on the regulatory model of promoter-RBS components and the experimental data. After
constructing promoter-RBS component libraries, based on system dynamic model we could
design a synthetic genetic circuit to achieve some prescribed design (user-oriented)
specifications by selecting an adequate set of promoter-RBS components from the
corresponding promoter-RBS libraries.
In this book, based on stochastic system model and promoter-RBS libraries, we design
biological filter for signal detection, biological transistor for signal amplification, biosensor
for signal processing, genetic lysis circuit for population control, genetic transmitter and
receiver for cell-cell communication to meet the prescribed design (user-oriented)
specifications through the proposed library-based searching method. Finally, these synthetic
genetic circuits are implemented by experiments to confirm their design performances. We
find these synthetic genetic circuits based on the proposed synthetic method could fit these
design specifications so that we could save much trial times by the conventional methods. In
future, system designers should cooperate with synthetic biologists to produce more complex
genetic circuit. In this situation, the development of systematic design methods is an
important topic for synthetic biology. Therefore, the proposed systematic biology methods in
this book have much potential applications to systems synthetic biology in future.
Chapter 1
INTRODUCTION
Synthetic biology is a new engineering discipline, which employs the concepts of

systems biology for genetic circuit design. Synthetic biologists utilize mathematical models to
describe biochemical reaction systems and then provide genetic engineering techniques to
implement the designed biological gene circuits in living cells with desired functions for a
variety of applications (Benner and Sismour 2005, Andrianantoandro et al., 2006).
Considering accessibility and easy manipulation, it is appealing to use micro-organisms, such
as Escherichia coli (E. coli) or Saccharomyces cerevisiae (S. cerevisiae), as a new technology
platform to solve several important synthetic biological problems. For examples, the biofuel
production is a crucial issue for utilizing as alternative energy to replace fuel sources, and
antibiotic therapy is another important strategy to effectively address medical treatment
problems (Withers and Keasling 2007, Lee et al., 2008, Lu and Collins 2009). In the future,
the biological factory built inside micro-organisms is expected to have many applications in
our lives.
In the past decade, several studies have shown that biological gene circuits can be
engineered to act as a repressilator, toggle switch, biosensor and bacterial filter in E. coli
(Elowitz and Leibler 2000, Gardner, Cantor, and Collins 2000, Looger et al., 2003, Stricker
et al., 2008, Basu et al., 2005, Entus, Aufderheide, and Sauro 2007, Sohka et al., 2009, Tuttle
et al.,). A repressilator causes stable oscillations of protein concentrations. Toggle switch can
be used as a biological switch or a memory device in micro-organisms. Biosensor and
bacterial filter can detect the concentrations of specific molecule signals. The biological
circuit mentioned above can achieve the desired outcome with low cost and less complexity,
but the functional scalability for further utilization may not be taken into account. In recent
years, some synthetic biologists started to use biological components such as promoters,
ribosomal binding sites (RBSs), repressive genes and reporter genes to assemble smaller
biological modules with different simple logical operations, i.e., NOT gate, AND gate, OR
gate and NOR gate, similar to the logic circuit design in digital integrated circuits (ICs)
(Huang et al., 2001). In addition, these functional networks of biological logic gates can be
repeatedly reused as a module to build more complex computational functions, such as
biological NAND gate, XOR gate, multiplexer and half adder (Wang et al., 2011, Tamsir,
Tabor, and Voigt 2011, Regot et al., 2011, Chuang and Lin 2014).
Furthermore, it is possible to design and achieve more complex integrations of bio-
synthetic circuits such as bacterial band-pass filter and genetic transistor (Sohka et al., 2009,
2 Bor-Sen Chen and Chih-Yuan Hsu
Regot et al., 2011, Lee et al., 2013). When the synthetic system becomes more and more
complex, new design strategies to handle such complexity become crucial for the next-
generation bio-synthetic circuit design and implementation. Novel computational and
systematic technologies are necessary to quickly and reliably engineer complex biological
systems, just like the approach of very-large scale integration(VLSI) design in the present
electronic circuit systems (Endy 2005). With the engineering principles of abstraction,
hierarchy and modularity, next generation higher-order gene networks and computational
gene circuits can be constructed (Lu, Khalil, and Collins 2009, Purnick and Weiss 2009, Beal
et al., 2012, Huynh et al., 2013). For this reason, synthetic biology researchers started to build
the biological component libraries where activities of promoters are quantitatively measured
in Biobrick (Kelly et al., 2009, Ellis, Wang, and Collins 2009). After the activities of different
promoter components in the library have been properly quantified and redefined by the
amount of fluorescence measured by flow cytometer, researchers employ a new library-based
search method through the genetic algorithm (GA) or other programing to efficiently select
the most appropriate biological components from the corresponding libraries to achieve the
optimal reference trajectory tracking (Wu, Lee, and Chen 2011, Huynh et al., 2012, Rodrigo,
Carrera, and Jaramillo 2011).
In Chapter 2, a robust biology filter design with a desired I/O filtering response is
introudced based on well characterized promoter-RBS libraries and a cascade gene circuit
topology. In the field of synthetic biology, biological filter serves as powerful detector or
sensor to sensor to sense different molecular signals and produces a specific output response
only if the concentration of input molecular signal higher or lower than a specified threshold.
In this chapter, the systematic design procechure of robust biological filters is devided into
three steps: (i) several well-characterized promoter-RBS libraries are estahlished as shown in
Chapter 2 for biological filters. (ii) The topology of synthetic biological filter is decomposed
into three cascade gene regulatory module. (iii) Based on the proposed systematic method, a
robust biological filter is engineered by searching the most adequate set of promoter-RBS
components from the corresponding libraries in Chapter 2 to achieve the specified I/O filter
response.
Synthetic genetic transistors are vital for signal amplification and switching in genetic
circuits. In Chapter 3, a systematic design methodology for robust genetic transistors is
introduced for practical applications based on I/O s specifications via promoter-RBS libraries.
Three kinds of promoter-RBS libraries are constructed for systematic genetic circuit design
using the identified kinetic stength of their promoter-RBS components. According to the
dynamic model of genetic transistors, a GA-based searching algorithm is developed to search
for a set of promoter-RBS components and adequate concentrations of inducers to achieve the
prescribed I/O characteristics of a genetic transistor. We also demonstrate the applicability of
the proposed design methodology to genetic transistors through experimental design in vivo.
In Chapter 4, a QS-based amplifier is designed to aid in the detection of metal ions. A
model is presented for calculating the concentration of reporter protein from the selection of
promoter-RBS components. Given a desired detection response, we propose a genetic
algorithm-based method to select the components. Experimental results show that selected
components were able to produce the desired detection response, and that the QS-based
amplifier made ion detection much more sensitive. In Chapter 5, a multi-objective (MO)
H2/H∞ performance criterion is derived for design specifications of a metal ion biosensor to
Introduction 3
achieve the H2 optimal matching of a desired input/output (I/O) response and simultaneous H
∞ optimal filtering of intrinsic parameter fluctuations and external cellular noise. According
to the two design specifications, the analysis and design of a metal ion biosensor with optimal
I/O response matching and optimal noise filtering ability then can be achieved by solving the
multi-objective problem under a set of LMIs. Moreover, a multi-objective evolutionary
algorithm (MOEA)-based library search method is employed to find adequate components
from corresponding libraries to solve LMI-constrained MO H2/H∞ design problems.
In Chapter 6, the aim is to control cell population density by controlling a lysis gene. For
this, as a first ingredient, a constitutive promoter-RBS region is used to control
transcription/translation of a certain gene. The expression product of this gene (its activity)
can be controlled by external inducers. The concentration of these inducers depends on
population-density. A second ingredient in the system is a promoter-RBS region of the lysis
gene, which is activated or repressed by the expression product of gene with the constitutive
promoter-RBS region. To make the design more easily, the lytic ability of the constructed
genetic lysis circuit is described by the relationship between the promoter-RBS components
and inducer concentrations in a steady state model. According to user-oriented specifications,
the genetic lysis circuit can be constructed via selecting adequate promoter-RBS components
in combination with a feasible range of inducer concentrations. In general, it requires long
computation times when component libraries become large. Hence, a genetic algorithm (GA)
search method is proposed to save time in evaluating and selecting promoter-RBS
components.
In Chapter 7, a synthetic biology approach is used to manipulate bacterial motility to
search for specific concentration of molecules. The CheY concentration is altered through the
quorum sensing system to regulate tumbling frequency so as to alter the drift velocity. The
study demonstrates that one can have varying drift velocities as shown using agar plate and
micro-fluidic systems, which can be further correlated to the concentration of AHL. The
effect of different brake component were identified by experimentation and supported by a
mathematical model. Based on this mathematical model one can choose appropriate brake
component from a library and effectively develop a genetic circuit. This genetic circuit can be
used to achieve necessary behavioral changes in bacteria.
In Chapter 8, a promoter-RBS library is constructed in the cyanobacterium
Synechococcous sp. PCC 7942 (hereafter PCC 7942). A mathematical model is used to obtain
a set of constitutive promoter-RBS component library involving two promoters - psbA1 and
J23101 and six RBS regions namely RBS1-3 and B0031, B0032 and B0034. The kinetic
strengths of the various promoter-RBS components are estimated by determining the eYFP
fluorescence intensity driven by the respective promoter-RBS unit. To test their promoter-
RBS component library, a synthetic gene circuit is designed for cell population control in
PCC 7942. The gene circuit design contains a bacteriorhodopsin gene to function as a light
driven proton pump and the ictB gene to control the cell density. These two genes are
controlled by two different sets of promoter-RBS units chosen based on the mathematical
model. Compare the values obtained in silico with experimental data and observe an
experimental error ranging between 1.7% and 3% only.
Chapter 9 examines the problem of engineering the behavior of complex microbial
ecosystems. There is increased interest in implementing multicellular microbial networks in a
variety of applications, but predicting and controlling the behaviors of such networks remains
a significant challenge. Towards addressing this problem, this chapter develops a

mathematical model and optimization scheme for programming a two species microbial
ecosystem. The model incorporates biological and external noise. The biological system
consists of two cell types which communicate via diffusible signals. These signals regulate
the production of an antibiotic resistance gene, creating a mutualistic interaction when the
antibiotic is present in the environment. To demonstrate the ability to tune the emergent
behavior of this system, the model adjusts both the antibiotic concentration in the
environment and the RBS region of the two signal synthases. Using a genetic algorithm, these
three parameters are adjusted to meet the design criteria, namely a specified steady-state ratio
of the two cell types.
REFERENCES
Andrianantoandro, Ernesto, Subhayu Basu, David K. Karig, and Ron Weiss. 2006. “Synthetic
biology: new engineering rules for an emerging discipline.” Molecular Systems Biology 2
(1).
Basu, Subhayu, Yoram Gerchman, Cynthia H. Collins, Frances H. Arnold, and Ron Weiss.
2005. “A synthetic multicellular system for programmed pattern formation.” Nature 434
(7037):1130-1134. doi: http://www.nature.com/nature/journal/v434/n7037/suppinfo/
nature03461_S1.html.
Beal, Jacob, Ron Weiss, Douglas Densmore, Aaron Adler, Evan Appleton, Jonathan Babb,
Swapnil Bhatia, Noah Davidsohn, Traci Haddock, Joseph Loyall, Richard Schantz,
Viktor Vasilev, and Fusun Yaman. 2012. “An End-to-End Workflow for Engineering of
Biological Networks from High-Level Specifications.” ACS Synthetic Biology 1 (8):317-
331. doi: 10.1021/sb300030d.
Benner, Steven A., and A. Michael Sismour. 2005. “Synthetic biology.” Nat Rev Genet 6
(7):533-543.
Chuang, C. H., and C. L. Lin. 2014. “A Novel Synthesizing Genetic Logic Circuit: Frequency
Multiplier.” IEEE/ACM Transactions on Computational Biology and Bioinformatics 11
(4):702-713. doi: 10.1109/TCBB.2014.2316814.
Ellis, Tom, Xiao Wang, and James J. Collins. 2009. “Diversity-based, model-guided
construction of synthetic gene networks with predicted functions.” Nat Biotech 27
(5):465-471. doi: http://www.nature.com/nbt/journal/v27/n5/suppinfo/nbt.1536_S1.html.
Elowitz, Michael B., and Stanislas Leibler. 2000. “A synthetic oscillatory network of
transcriptional regulators.” Nature 403 (6767):335-338.
Endy, Drew. 2005. “Foundations for engineering biology.” Nature 438 (7067):449-453.
Entus, Robert, Brian Aufderheide, and Herbert M. Sauro. 2007. “Design and implementation
of three incoherent feed-forward motif based biological concentration sensors.” Systems
and Synthetic Biology 1 (3):119-128. doi: 10.1007/s11693-007-9008-6.
Gardner, Timothy S., Charles R. Cantor, and James J. Collins. 2000. “Construction of a
genetic toggle switch in Escherichia coli.” Nature 403 (6767):339-342. doi:
http://www.nature.com/nature/journal/v403/n6767/suppinfo/403339a0_S1.html.
Introduction 5
Huang, Yu, Xiangfeng Duan, Yi Cui, Lincoln J. Lauhon, Kyoung-Ha Kim, and Charles M.
Lieber. 2001. “Logic Gates and Computation from Assembled Nanowire Building
Blocks.” Science 294 (5545):1313.
Huynh, Linh, John Kececioglu, Matthias Köppe, and Ilias Tagkopoulos. 2012. “Automatic
Design of Synthetic Gene Circuits through Mixed Integer Non-linear Programming.”
PLoS ONE 7 (4):e35529. doi: 10.1371/journal.pone.0035529.
Huynh, Linh, Athanasios Tsoukalas, Matthias Köppe, and Ilias Tagkopoulos. 2013.
“SBROME: A Scalable Optimization and Module Matching Framework for Automated
Biosystems Design.” ACS Synthetic Biology 2 (5):263-273. doi: 10.1021/sb300095m.
Kelly, Jason R., Adam J. Rubin, Joseph H. Davis, Caroline M. Ajo-Franklin, John Cumbers,
Michael J. Czar, Kim de Mora, Aaron L. Glieberman, Dileep D. Monie, and Drew Endy.
2009. “Measuring the activity of BioBrick promoters using an in vivo reference
standard.” Journal of Biological Engineering 3 (1):1-13. doi: 10.1186/1754-1611-3-4.
Lee, Sung Kuk, Howard Chou, Timothy S. Ham, Taek Soon Lee, and Jay D. Keasling. 2008.
“Metabolic engineering of microorganisms for biofuels production: from bugs to
synthetic biology to fuels.” Current Opinion in Biotechnology 19 (6):556-563. doi:
http://dx.doi.org/10.1016/j.copbio.2008.10.014.
Lee, Yi-Ying, Chih-Yuan Hsu, Ling-Jiun Lin, Chih-Chun Chang, Hsiao-Chun Cheng, Tsung-
Hsien Yeh, Rei-Hsing Hu, Che Lin, Zhen Xie, and Bor-Sen Chen. 2013. “Systematic
design methodology for robust genetic transistors based on I/O specifications via
promoter-RBS libraries.” BMC systems biology 7 (1):1.
Looger, Loren L., Mary A. Dwyer, James J. Smith, and Homme W. Hellinga. 2003.
“Computational design of receptor and sensor proteins with novel functions.” Nature
423 (6936):185-190. doi: http://www.nature.com/nature/journal/v423/n6936/suppinfo/
Lu, Timothy K., and James J. Collins. 2009. “Engineered bacteriophage targeting gene
networks as adjuvants for antibiotic therapy.” Proceedings of the National Academy of
Sciences 106 (12):4629-4634.
Lu, Timothy K., Ahmad S. Khalil, and James J. Collins. 2009. “Next-generation synthetic
gene networks.” Nat Biotech 27 (12):1139-1150.
Purnick, Priscilla E. M., and Ron Weiss. 2009. “The second wave of synthetic biology: from
modules to systems.” Nat Rev Mol Cell Biol 10 (6):410-422.
Regot, Sergi, Javier Macia, Nuria Conde, Kentaro Furukawa, Jimmy Kjellen, Tom Peeters,
Stefan Hohmann, Eulalia de Nadal, Francesc Posas, and Ricard Sole. 2011. “Distributed
biological computation with multicellular engineered networks.” Nature 469 (7329):207-
211. doi: http://www.nature.com/nature/journal/v469/n7329/abs/nature09679.html
#supplementary-information.
Rodrigo, Guillermo, Javier Carrera, and Alfonso Jaramillo. 2011. “Computational design of
synthetic regulatory networks from a genetic library to characterize the designability of
dynamical behaviors.” Nucleic Acids Research 39 (20):e138. doi: 10.1093/nar/gkr616.
Sohka, Takayuki, Richard A. Heins, Ryan M. Phelan, Jennifer M. Greisler, Craig A.
Townsend, and Marc Ostermeier. 2009. “An externally tunable bacterial band-pass
filter.” Proceedings of the National Academy of Sciences 106 (25):10135-10140. doi:
10.1073/pnas.0901246106.
Stricker, Jesse, Scott Cookson, Matthew R. Bennett, William H. Mather, Lev S. Tsimring,
and Jeff Hasty. 2008. “A fast, robust and tunable synthetic gene oscillator.” Nature
456 (7221):516-519. doi: http://www.nature.com/nature/journal/v456/n7221/suppinfo/
Tamsir, Alvin, Jeffrey J. Tabor, and Christopher A. Voigt. 2011. “Robust multicellular
computing using genetically encoded NOR gates and chemical /`wires/'.” Nature
469 (7329):212-215. doi: http://www.nature.com/nature/journal/v469/n7329/abs/nature
09565.html#supplementary-information.
Tuttle, Lisa M., Howard Salis, Jonathan Tomshine, and Yiannis N. Kaznessis. “Model-Driven
Designs of an Oscillating Gene Network.” Biophysical Journal 89 (6):3873-3883. doi:
10.1529/biophysj.105.064204.
Wang, Baojun, Richard I. Kitney, Nicolas Joly, and Martin Buck. 2011. “Engineering
modular and orthogonal genetic logic gates for robust digital-like synthetic biology.”
Nat Commun 2:508. doi: http://www.nature.com/ncomms/journal/v2/n10/suppinfo/
ncomms1516_S1.html.
Withers, Sydnor T., and Jay D. Keasling. 2007. “Biosynthesis and engineering of isoprenoid
small molecules.” Applied Microbiology and Biotechnology 73 (5):980-990. doi:
10.1007/s00253-006-0593-1.
Wu, Chih-Hung, Hsiao-Ching Lee, and Bor-Sen Chen. 2011. “Robust synthetic gene network
design via library-based search method.” Bioinformatics 27 (19):2700-2706. doi:
10.1093/bioinformatics/btr465.
Chapter 2
SYSTEMATIC BIOLOGICAL FILTER DESIGN

WITH A DESIRED I/O FILTERING RESPONSE
BASED ON PROMOTER-RBS LIBRARIES
ABSTRACT
In this chapter, robust biological filters with an external control to match a desired
input/output (I/O) filtering response are engineered based on the well-characterized
promoter-RBS libraries and a cascade gene circuit topology. In the field of synthetic
biology, the biological filter system serves as a powerful detector or sensor to sense
different molecular signals and produces a specific output response only if the
concentration of the input molecular signal is higher or lower than a specified threshold.
The proposed systematic design method of robust biological filters is summarized into
three steps. Firstly, several well-characterized promoter-RBS libraries are established for
biological filter by identifying and collecting the quantitative and qualitative
characteristics of their promoter-RBS components via nonlinear parameter estimation
method. The topology of synthetic biological filter is decomposed into three cascade gene
regulatory modules, and an appropriate promoter-RBS library is selected for each module
to achieve the desired I/O specification of a biological filter. Finally, based on the
proposed systematic method, a robust externally tunable biological filter is engineered by
searching the promoter-RBS component libraries and a control inducer concentration
library to achieve the optimal reference match for the specified I/O filtering response.
2.1. INTRODUCTION
In this chpater, we focus on the systematic biological filter design scheme, specifically,
how to engineer a synthetic gene circuit to act as an externally tunable filter in Escherichia
coli (Sohka et al., 2009c, Sohka, Heins, and Ostermeier 2009c). Although biological filter or
detector is a common research topic, a systematic design approach is still lacking. Since the
effective filtering response range of a biological filter is limited and only certain molecular
signals can be detected, the biological filter is often employed in pattern formation (Sohka
et al., 2009c, Sohka, Heins, and Ostermeier 2009c). But it has no further applications in
environmental detection and regulation of substance concentrations in micro-organisms. For
practicability, applicability, and scalability, a genetic regulation network is reconstructed
based on the component libraries and gene circuit topology. A robust externally tunable
biological filter with desired input/output (I/O) filtering response can be implemented using
the procedure described below.
In order to build large-scale gene circuits of higher complexity, a bottom-up approach is
needed. Therefore, efforts to establish a standard library of biological components are very
important. Similar to the field of electrical engineering, the large and complex ICs are
assembled by standard electrical elements, namely, resistors, inductors, capacitors, and
transistors. In the field of synthetic biology, the quantitative and qualitative characteristics of
each standard biological component need to be identified and collected to build different
component libraries, that is, promoter libraries based on promoter activities are provided to
engineer gene circuits (Wang et al., 2011, Ellis, Wang, and Collins 2009b, Wu, Lee, and
Chen 2011b, Wu, Zhang, and Chen 2011). The standard biological components used in this
study are taken from BioBricks. With these quantitative parameters, synthetic biologists can
first compute, simulate, and analyze the overall gene circuit system in silico based on
mathematical dynamic models before actual genetic circuit assembly. Thus, the first step of
our design procedure is to systematically build our own libraries of promoters fused to
ribosomal binding sites (RBS) by using BioBrick components (promoter-RBS) according to
fluorescence data.
Next, our design goal is to engineer a robust biological filter with an external control
through appropriate selection of the topology of the interconnection of biological components
that are available in our constructed promoter-RBS library. Notably, the biological filter
differs from the frequency filter for signal filtering used in the field of electrical engineering.
The biological filter system serves as a powerful detector or sensor for the concentration of
specific molecular signals and shows the prescribed I/O response behavior depending on the
concentration of specific molecular signals. In order to realize the functionality of the
biological filter system, a cascade gene circuit topology is presented. The gene circuit
topology of a biological filter is separated into three cascade gene regulation modules.
Subsequently, an appropriate promoter-RBS library is selected for each module to achieve the
desired I/O specification of a biological filter. In addition, the biological filter has an
additional control function to tune the system I/O filtering response more precisely by
adjusting the concentration of the external control inducer. Using the design example of
biological filters that transform a concentration gradient of input molecular signals into a
desired I/O filtering response, we show that the I/O filtering response of biological filters can
be qualitatively and quantitatively tuned by our proposed cascade gene circuit topology and
design.
Finally, the proposed robust design for biological filter is optimized via a mixed library-
based search method. The synthetic biological filter can robustly and optimally match the
desired I/O filtering response by selecting an appropriate set of promoter-RBS components
and a specified control inducer concentration. This study provides an important systematic
design approach for the development of next-generation synthetic biological circuits, from
biological component library construction to biological circuit assembly. When more
promoter-RBS libraries are built in the future, the biological filter can be systematically
designed and improved to detect various molecular signals with different concentrations and a
higher accuracy. The designed biological filter can be used alone as a powerful detector of
particular molecular signals and can also produce different response behaviors depending on
the concentration of specific molecular signals. Biological filter with greater complexity may
Systematic Biological Filter Design with a Desired I/O Filtering Response … 9
be realized by rewiring the connectivity of various biological filter modules, that is, a band-
pass filter can be attained by combining low-pass and high-pass filters (Kampf et al., 2012).
With the concept of modularity, the designed biological filter can be further applied as a
module in other regulatory networks, such as biological metabolic regulation (Bulter et al.,
2004, Farmer and Liao 2000).
The main focuses of this chapter are on: (1) Three kinds of promoter-RBS libraries are
established based on promoter-RBS kinetic strengths. (2) A novel synthetic gene circuit with
three promoter-RBS components is reconstructed to function as a biological filter based on a
cascade gene circuit topology. According to the design specifications, a robust externally
tunable biological filter is engineered by searching an adequate set of three promoter-RBS
components and a control inducer concentration from the corresponding libraries to achieve
the optimal reference matching for the specified I/O filtering response in spite of the
parameter fluctuations or external molecular noises. (3) A mixed library-based search method
is employed to find the most adequate set of promoter-RBS components and inducer
concentration via genetic algorithm. Further, the external tunability of these synthetic
biological filters that shows the potential in practical applications is also verified.
2.2. CONSTRUCTION OF PROMOTER-RBS LIBRARY

The exchangeable protocol of BioBrick-related data has been developed for a
standardized, extensive, scalable, and machine-processable interface (Canton, Labno, and
Endy 2008b). However, only few BioBrick parts are well-characterized quantitatively. In fact,
these more well-characterized BioBrick parts have been efficiently used in the synthetic gene
circuit through systematic design. In order to facilitate the design of the synthetic gene circuit,
the well-characterized BioBrick parts need to be extended. In particular, since gene
expression is both regulated by the promoter and RBS, we consider the promoter and RBS as
a promoter-RBS component. Subsequently, the strength of each promoter-RBS component
should be systematically identified and constructed as the index in the promoter-RBS
libraries. We provide a stochastic model to capture the dynamic behaviors to identify the
strengths of the promoter-RBS components under the influence of molecular noise.
Subsequently, information regarding the strengths of promoter-RBS components can be
constructed as the indices of promoter-RBS libraries in Tables 2.1-2.4. The detailed
construction procedures of promoter-RBS libraries are described in Appendix. In the
following sections, we will design a biological filter with the desired I/O filtering response by
selecting an appropriate set of promoter-RBS components from the constructed promoter-
RBS libraries in Tables 2.1-2.4.
2.3. SYSTEMATIC DESIGN METHODOLOGY OF THE SYNTHETIC

BIOLOGICAL FILTER BASED ON PROMOTER-RBS LIBRARIES
A Cascade Gene Circuit Topology for Biological Filter Design
After promoter-RBS libraries are built upon the parameter identification of promoter-
RBS strengths, a set of biological components from the promoter-RBS libraries, i.e., Tables
2.1-2.4 in Appendix, and their corresponding regulatory genes are selected to engineer a
biological filter on the basis of the cascade gene circuit topology in Figure 2.1. The synthetic
gene circuit functions as a biological filter with inducers I1 and I2 to refine the activities of
corresponding promoter-RBS components, repressor-regulated promoter-RBS component and
activator-regulated promoter-RBS component, which are all selected from the corresponding
component libraries in Tables 2.1-2.4, respectively. When one of the inducers is taken as
input to the synthetic cascade gene circuit, the threshold of the biological I/O filtering
response can be tuned externally by changing the concentration of another inducer in Figure
2.1. The threshold is the point that must be exceeded to begin producing a given effect or to
elicit a response (see Figure 2.2(B) and 2.2(C)).
Figure 2.1. A synthetic biological filter design based on a cascade gene circuit topology. A synthetic
cascade circuit with three promoter-RBS components is designed to function as a biological filter. The
first node x1 has been fixed by a constitutive promoter-RBS component. The second node x2 and third
node x3 can be engineered by an activator-regulated promoter-RBS component or a repressor-regulated
promoter-RBS component, respectively. When different kinds of promoter-RBS components and
regulatory genes are selected from the corresponding libraries, they will activate (+ sign) or repress
(− sign) the connected node. Inducer I1 and I2 are used to tune the activities of the corresponding
promoter-RBS components for fine tuning of biological filter.
(A) I1 I2
reporter
fluorescence
protein
xr1 xr2 reporter G

RBS RBS RBS
gene gene gene
(B) (C)
Gss Gss
low-pass threshold I1 high-pass threshold I2
Figure 2.2. A synthetic biological filter with promoter-RBS components selected from the constitutive
and repressor-regulated promoter-RBS libraries, and the I/O filtering response of the biological low-
pass and high-pass filters. (A) For convenience of explaining the operating mechanism of the biological
filter, a synthetic biological filter is assembled by selecting a constitutive promoter-RBS component C1i,
and two different repressor-regulated promoter-RBS components R2i and R3i from their corresponding
libraries in Tables 2.1-2.4, to regulate their corresponding regulatory genes and reporter gene. (B) The
I/O filtering response of the low-pass filter for input inducer I1 and output fluorescence Gss at steady
state. (C) The I/O filtering response of the high-pass filter for input inducer I2 and output fluorescence
Gss at steady state.
For the convenience of description and explanation, a synthetic biological filter circuit is
assembled by selecting a set of biological components and genes, namely, a constitutive
promoter-RBS component C1i from Table 2.1, two different kinds of repressor-regulated
promoter-RBS components R2i and R3i from Table 2.2 or 2.3, two regulated repressor genes,
and an appropriate reporter gene, as shown in Figure 2.2(A). First, if the inducer I1 and the
fluorescence expression of the reporter protein are taken as input and output, respectively, the
low-pass I/O filtering response can be obtained. The biological low-pass filter would be
maintained at high expression when input concentration I1 is low, and would be maintained at
low expression when input concentration I1 is high, as shown in Figure 2.2(B). In the
proposed method, the inducer concentration I2 in Figure 2.2(A) is an additional design
feature, which can be adjusted to precisely and externally tune the threshold of the biological
low-pass filter. In the biological filter design, the threshold is the point at which the system
I/O response transfers from high to low or from low to high in Figures 2.2(B) and 2.2(C),
respectively. In addition to the external control of inducer concentration I2, promoter-RBS
components with different strengths can be selected to enable distinct thresholds based on
promoter-RBS libraries. This flexible design feature is novel in synthetic biological filters.
For the convenience of illustrating the operating mechanism of the biological low-pass
filter in Figure 2.2(A), the reporter gene is under the control of repressor xr2, which is
repressed by repressor xr1. When the input inducer I1 is absent, a continuous production of
repressor xr1 completely inhibits the generation of repressor xr2. Therefore, the output
fluorescence expression without inhibition would remain at a high level. When the input
inducer concentration I1 is gradually increased, the repressor activity of repressor xr1 is
reduced, because inducer I1 binds with repressor xr1 to prevent free repressor xr1 from binding
with the corresponding promoter-RBS component R2i. As a result, repressor xr2 is generated,
thus inhibiting the production of reporter protein. Therefore, the output fluorescence G is
inhibited and converted from a high level to low level, with a low-pass I/O filtering response
as shown in Figure 2.2(B). Notably, the concentration of another inducer I2 functions as a
buffer to adjust the I/O filtering response of the biological low-pass filter to the input inducer
I1. Fluorescence G is not inhibited by low concentrations of repressor xr2, since inducer I2
binds to repressor xr2 to prevent free repressor xr2 from binding to the corresponding
promoter-RBS component R3i. Hence, the threshold for the biological low-pass filter is
pushed forward when the input inducer concentration I1 is increased gradually. This is why
the threshold of the biological low-pass filter can be precisely tuned by changing the control
inducer concentration I2 externally.
The operation mechanism of the biological high-pass filter is also shown in Figure
2.2(A), where the fluorescence expression of reporter gene is under the control of the
repressor xr2. Here, I2 is the input inducer instead. When the input inducer I2 is absent, the
repressor xr2 completely binds to the corresponding promoter-RBS component R3i. Therefore,
the production of immature reporter protein and output fluorescence expression G would be
inhibited. When the input inducer concentration I2 is increases, the repressor activity of
repressor xr2 decreases, because inducer I2 binds to the repressor xr2 to prevent free repressor
xr2 from binding to the corresponding promoter-RBS component R3i. Thus, fluorescence
expression G is no longer inhibited and is converted from a low level to a high level, with a
high-pass filtering response as shown in Figure 2.2(C). Notably, the inducer concentration I1
functions as a regulatory mechanism to regulate the production of repressor concentration xr2
of the biological high-pass filter. The inducer concentration I1 could regulate the gene
expression of repressor xr2 because I1 can bind to the repressor xr1. The regulation of
biological high-pass filter is driven by the constitutive promoter-RBS component C1i, which
prevents the free repressor xr1 from binding to the corresponding promoter-RBS component
R2i. When the concentration of control inducer I1 increases, the production of repressor xr2
then increases, thus inhibiting the output fluorescence expression. Therefore, the threshold for
the biological high-pass filter would be pushed forward when the control inducer
concentration I1 is increased gradually. This is why the threshold of the biological high-pass
filter can be precisely tuned by changing the control inducer concentration of I1 externally.
Dynamic Model of the Biological Filter Based on Genetic Circuit Topology
The cascade gene circuit topology is depicted in Figure 2.1, from which the dynamic
model of the synthetic biological filter can be introduced for systematic design. The
concentrations of two different regulatory proteins are denoted by x1 and x2, and the
concentrations of the immature and mature reporter proteins are denoted by x3 and G,
respectively. In the following discussion, the concentration of mature reporter protein is
assumed to be equal to that measured by ELISA. The dynamic model of the biological filter
in Figure 2.1 is given as follows:
 x1 (t )  P( S1u , S1l ,0,0)  (    x1 ) x1 (t )  1 (t )


 x2 (t )  P( S2u , S2l , x1 , I1 )  (    x2 ) x2 (t )  2 (t )
 (2.1)
 x3 (t )  P( S3u , S3l , x2 , I 2 )  (    x3 ) x3 (t )  3 (t )

 G (t )  mx3  (    G )G (t )  4 (t )
where γx1 and γx2 are the degradation rates of two different regulatory proteins x1 and x2,
respectively; γx3 and γG are the degradation rates of the immature and mature reporter
proteins, respectively; μ and m denote the dilution rate due to cell growth and the maturation
rate of the reporter protein, respectively; ω1(t), ω2(t), and ω3(t) denote the molecular noises in
the transcription and translation processes, respectively; ω4(t) denotes the molecular noise in
mature protein expression; {S1u, S1l}, {S2u, S2l}, and {S3u, S3l} denote the strengths of three
different promoter-RBS components; I1 and I2 denote the inducer concentrations for
regulating the activities of x1 and x2, respectively. On the basis of the actual synthetic gene
circuit and the promoter-RBS regulation function P(Su, Sl, TF, I) with maximum and
minimum promoter-RBS strengths Su and Sl, respectively, the corresponding transcription
factor concentration TF and related inducer concentration I in (2.1) may be substituted with
the constitutive promoter-RBS regulation function for the constitutive promoter-RBS
component Ci. Similarly, it may be substituted with the repressor-regulated promoter-RBS
regulation function for the repressor-regulated promoter-RBS component Ri, or with the
activator-regulated promoter-RBS regulation function for the activator-regulated promoter-
RBS component Ai (see Appendix). The transcription factor concentration may be equal to
zero for the constitutive promoter-RBS component Ci. Alternatively, it may be equal to the
repressor concentration xr for the repressor-regulated promoter-RBS component Ri, and
activator concentration xa for the activator-regulated promoter-RBS component Ai.
To achieve the desired I/O specification, a synthetic biological filter is assembled by
selecting a set of appropriate promoter-RBS libraries, namely, a constitutive promoter-RBS
component C1i from Table 2.1, two different kinds of repressor-regulated promoter-RBS
components R2i and R3i from Table 2.2 or Table 2.3, two regulated repressor genes, and an
appropriate reporter gene, as shown in Figure 2.2(A). To emphasize that repressor-regulated
promoter-RBS are selected for the circuit topology, the regulatory protein concentrations x1
and x2 in (2.1) are replaced by two different repressor concentrations denoted by xr1 and xr2,
respectively. The first promoter-RBS regulation function P(S1u, P1l, 0, 0) in (2.1) is
substituted with the constitutive promoter-RBS regulation function. The second and third
promoter-RBS regulation functions P(S2u, P2l, x1, I1) and P(S3u, P3l, x2, I2), respectively, in
(2.1) are substituted with the repressor-regulated promoter-RBS regulation function. On the
basis of the gene circuit topology in Figure 2.2(A), the dynamic model for the biological filter
in (2.1), which has a set of promoter-RBS components C1i, R2i, and R3i selected from the
promoter-RBS libraries, is rewritten as follows:
 xr1 (t )  Pc ( S1u ,i ,0,0,0)  (    xr 1 ) xr1 (t )  1 (t )


 xr 2 (t )  Pr ( S2u ,i , S2l ,i , xr1 , I1 )  (    xr 2 ) xr 2 (t )  2 (t )

 x3 (t )  Pr ( S3u ,i , S3l ,i , xr 2 , I 2 )  (    x3 ) x3 (t )  3 (t )

 G (t )  mx3 (t )  (    G )G (t )  4 (t )
{S1u ,i }  Libconst , S2u ,i , S2l ,i   Librepressor1 , S3u ,i , S3l ,i   Librepressor 2 (2.2)
repressor1  TetR, LacI  , repressor 2  TetR, LacI 
in which
Pc ( S1u ,i ,0,0,0)  S1u ,i

S 2u ,i  S 2l ,i xr1
Pr ( S2u ,i , S2l ,i , xr1 , I1 )  S2l ,i  nr 2
, xr*1 ( xr1 , I1 )  nI 1
 x* ( x , I )   I 
1   r1 r1 1  1  1 
 Kr 2   KI 
 1
S 3u , i  S 3l , i xr 2
Pr ( S3u ,i , S3l ,i , xr 2 , I 2 )  S3l ,i  nr 3
, xr*2 ( xr 2 , I 2 )  nI 2
 x* ( x , I )   I 
1   r2 r2 2  1  2 
 Kr3   KI
 2 
where γxr1 and γxr2 are the degradation rates of repressors xr1 and xr2, respectively;
{S1u,i} Libconst denotes the set of promoter-RBS strengths S1u,i of the constitutive promoter-
RBS components C1i, which are selected from Libconst. {S2u,i, S2l,i} Librepressor1 denotes the set
of the maximum and minimum promoter-RBS strengths S2u,i and S2l,i, respectively of the
repressor-regulated promoter-RBS components R2i in Librepressor1. {S3u,i, S3l,i} Librepressor2
denotes the set of the maximum and minimum promoter-RBS strengths S3u,i and S3l,i,
respectively, of the repressor-regulated promoter-RBS components R3i in Librepressor2. Kr2 and
nr2 denote the binding affinity and the binding cooperativity between repressor xr1 and
promoter-RBS component R2i, respectively, in the promoter-RBS library Librepressor1. Kr3 and
nr3 defined in the corresponding promoter-RBS library Librepressor2 denote the binding affinity
and the binding cooperativity between repressor xr2 and promoter-RBS component R3i,
respectively. KI1 and KI2 denote the dissociation rate between inducers I1, I2 and repressors
xr1, xr2, respectively. nI1 denotes the binding cooperativity between repressor xr1 and inducer I1
; nI2 denotes the binding cooperativity between repressor xr2 and inducer I2 . Therefore, the
concentrations of inducers I1 and I2 could affect the repressor activities x*r1(xr1, I1) and
x*r2(xr2, I2), respectively, thereby affecting the strengths of the promoter regulation functions
Pr(S2u,i, S2l,i, xr1, I1) and Pr(S3u,i, S3l,i, xr2, I2) in (2.2), respectively. Eventually, the expressions
of repressor xr2 and reporter protein GFP are affected, as shown in Figure 2.2(A). Meanwhile,
the expression of repressor xr1 is not affected by the concentrations of inducers I1 and I2
because the repressor xr1 is driven by a constitutive promoter-RBS component C1i without
external regulation.
To describe the I/O filtering response of the biological filter, the dynamic model is
transformed to a steady-state model by assuming that the system dynamics in (2.2) are equal
to zero. The steady-state concentrations of the repressors, as well as immature and mature
reporter proteins are denoted by xr1ss, xr2ss, x3ss, and Gss, respectively. The steady-state model
of the biological filter system with promoters C1i, R2i, and R3i selected from the corresponding
promoter-RBS libraries Libconst, Librepressor1, and Librepressor2 respectively, is derived as follows:
 Pc ( S1u ,i , 0, 0, 0)
 xr1ss   1
    xr 1
 P (S , S , x , I )
 xr 2 ss  r 2u ,i 2l ,i r1 1  2
    xr 2

 Pr ( S3u ,i , S3l ,i , xr 2 , I 2 ) (2.3)
 x3 ss   3
   x3

 m
Gss  x  4
    G 3ss
{S1u ,i }  Libconst , S2u ,i , S2l ,i   Librepressor1 , S3u ,i , S3l ,i   Librepressor 2
where ω1, ω2 and ω3 denote the molecular noises in the transcription and translation processes
of gene expression at steady-state, respectively; ω4 denotes the molecular noises in mature
protein expression. The steady-state protein expression of the biological filter system can be
obtained from the steady-state model (2.3) with the promoter regulation functions Pc(S1u,i,
0,0,0), Pr(S2u,i, S2l,i, xr1, I1), and Pr(S3u,i, S3l,i, xr2, I2); and repressor activities x*r1(xr1, I1) and
x*r2(xr2, I2) in (2.2), respectively.
Next, the biological low-pass I/O filtering response is analyzed on the basis of the steady-
state model (2.3) of the cascade gene circuit topology in Figure 2.2(A). When the inducer I1 is
absent, the repressor activity x*r1(xr1, I1) is equal to the repressor concentration xr1. Since the
steady-state repressor concentration xr1ss regulated by the constitutive promoter-RBS
component C1i is a constant in the steady-state model (2.3), the promoter-RBS regulation
function Pr(S2u,i, S2l,i, xr1, I1) is close to S2l,i, the minimum promoter-RBS strength of the
corresponding repressor-regulated promoter-RBS component R2i in the library Librepressor1.
The repressor activity x*r2(xr2, I2) is close to zero because the steady-state concentration of
repressor xr2 regulated by repressor-regulated promoter-RBS component R2i is maintained at a
low level. Thus, the promoter-RBS regulation function Pr(S3u,i, S3l,i, xr2, I2) is close to S3l,i, the
maximum promoter-RBS strength. Finally, the steady-state concentration of immature
reporter protein x3ss is maintained at a high level and the steady-state GFP fluorescence
expression Gss in (2.3) is also maintained at a high level. As the inducer concentration I1
gradually increases, the repressor activity x*r1(xr1, I1) decreases. The promoter-RBS regulation
function Pr(S2u,i, S2l,i, xr1, I1) increases to a level greater than S2l,i, and the steady-state
concentration of repressor xr2 in (2.3) gradually increases. Thus, the repressor activity x*r2(xr2,
I2) is increased at a fixed concentration of inducer I2, and the value of the promoter-RBS
regulation function Pr(S3u,i, S3l,i, xr2, I2) decreases to a level lower than S3u,i. Therefore, the
steady-state concentration of the immature reporter protein x3ss gradually decreases, and the
steady-state fluorescence expression Gss in (2.3) also decreases. The repressor activity x*r2(xr2,
I2) can be regulated by the inducer concentration I2. For example, a higher repressor
concentration xr2 regulated by the inducer I1 is needed to achieve the same repressor activity
x*r2(xr2, I2) when the inducer concentration I2 is increased. This is why the threshold of the
biological low-pass filter can be precisely tuned by the concentration of external inducer I2.
Meanwhile, the biological high-pass I/O filtering response is also analyzed based on the
steady-state model (2.3) of the cascade gene circuit topology in Figure 2.2(A). When the
inducer I2 is absent, repressor activity x*r2(xr2, I2) is equal to the repressor concentration xr2.
The promoter-RBS regulation function Pr(S3u,i, S3l,i, xr2, I2) would be close to S3l,i, the
minimum promoter-RBS strength of the promoter-RBS component R3i in the library
Librepressor2. Hence, the steady-state concentration of immature reporter protein x3ss and
fluorescence Gss would be maintained at a low level in (2.3). As the inducer concentration I2
is gradually increased, the repressor activity x*r2(xr2, I2) is decreased. The promoter-RBS
regulation function Pr(S3u,i, S3l,i, xr2, I2) would be increased to a level greater than S3l,i.
Therefore, the steady-state concentration of immature reporter protein x3ss and steady-state
fluorescence expression Gss would gradually increase until the promoter-RBS regulation
function Pr(S3u,i, S3l,i, xr2, I2) reaches the maximum strength S3u,i. The repressor activity
x*r1(xr1, I1) can be regulated by the concentration of inducer I1 and then influence the
promoter-RBS regulation function Pr(S2u,i, S2l,i, xr1, I1), thereby affecting the steady state
concentration of repressor xr2 in (2.3). For example, when the concentration of inducer I1 is
increased, the expression of repressor xr2 is increased. To achieve the same repressor activity
x*r2(xr2, I2), a higher inducer concentration I2 is needed to bind increased repressor
concentration xr2. This is why the threshold of the biological high-pass filter can be precisely
tuned by the concentration of external inducer I1.
Formulation of Design Specifications for Biological Filters
Our design purpose is to engineer a synthetic biological filter by selecting an adequate set
of promoter-RBS components from the corresponding libraries and an appropriate control
inducer concentration to achieve the optimal reference match for the desired I/O filtering
response. When the steady-state model of the biological filter is represented in (2.3), a desired
I/O filtering response of fluorescence expression with different input concentrations needs to
be prescribed for the reference matching of biological filtering response. The desired I/O
filtering responses for the reference matching of biological low-pass filter and high-pass
filter, Gr1(I1) and Gr2(I2), respectively, can be specified as follows:
1
Gr1 ( I1 )  Gmax (2.4)
2
I 
12   1 
 I lc 
I2
I hc
Gr 2 ( I 2 )  Gmax (2.5)
2
 I 
12   2 
 I hc 
where Gmax denotes the maximum fluorescence expression of the biological filter. With an
adequate set of promoter-RBS components to be selected from the corresponding promoter-
RBS libraries, we seek to achieve a match for Gr1(I1) or Gr2(I2).
If u is denoted as the input signal to the filter system, for a desired biological low-pass
I/O response in (2.4), the input signal u is the concentration of inducer I1 based on the
biological filter circuit in Figure 2.2(A). For a desired biological high-pass amplitude
response in (2.5), the input signal u is the concentration of inducer I2 based on the biological
filter circuit in Figure 2.2(A). Ilc denotes the threshold specified for the biological low-pass
filtering response and Ihc denotes the threshold specified for the biological high-pass filtering
response. For the desired I/O filtering response of the biological low-pass filter in (2.4), a low
concentration of inducer I1, such that I1 << Ilc, leads to high fluorescence levels close to the
maximum expression; a high concentration of inducer I1, such that I1 >> Ilc, leads to low
fluorescence levels close to zero. On the other hand, for the desired I/O filtering response of
the biological high-pass filter in (2.5), a low concentration level of inducer I2, specifically, I2
<< Ihc, leads to low fluorescence levels; and a high concentration level of inducer I2,
specifically, I2 >> Ihc, leads to high fluorescence levels. When the input signal u is equal to
the low-pass threshold Ilc or high-pass threshold Ihc, the fluorescence expression of the genetic
filter reach times 1/ the maximum fluorescence level Gmax, i.e., Gr1(Ilc) = 1/ Gmax and
Gr2(Ihc) = 1/ Gmax in Figure 2.2. On the basis of the gene circuit topology and the steady-
state model of the biological filter, a set of promoter-RBS components and an appropriate
regulatory inducer concentration can be selected from the corresponding libraries to match the
specified I/O filtering responses Gr1(I1) and Gr2(I2) in (2.4) and (2.5) for the biological low-
pass and high-pass filters, respectively.
Biological Filter Design by the Reference Matching via a Mixed Library-

Based Searching Method
After the desired I/O filtering response is specified as described above, a library-based
search method (Wu, Lee, and Chen 2011b) is employed to efficiently select the most adequate
set of promoter-RBS components from the corresponding promoter-RBS libraries to achieve
the desired I/O filtering response for the biological filter design. Furthermore, a concentration
of regulatory inducer is taken into consideration as another inducer concentration library,
resulting in a mixed library-based search method. After the steady-state model of the
biological filter is established in (2.3), our ultimate goal is to design a biological filter that
matches the specified I/O filtering response by selecting an adequate inducer concentration to
be tuned externally and an appropriate set of promoter-RBS components from the
corresponding promoter-RBS libraries in Tables 2.1-2.4.
In addition, the standard deviations of stochastic parameter fluctuations need to be
considered in the modeling as biological components are inherently uncertain in vivo because
of gene-expression noise, cell growth, DNA mutations, evolution, and parameter estimation
errors. The standard deviations of stochastic parameter fluctuation and environmental
molecular disturbance can be considered as quantitative specifications of system robustness
for the biological filter. Here, we summarize our proposed design procedure for an externally
tunable biological filter via a mixed library-based search method with the following four
design specifications:
1. Specify a desired I/O filtering response of the biological filter with a desired
maximum fluorescence Gmax, and a specified threshold Ilc for low-pass response or Ihc
for high-pass response in (2.4) and (2.5), respectively.
2. Provide the well-characterized promoter-RBS libraries listed in Tables 2.1-2.4 and
inducer concentrations.
3. Specify the parameter uncertainties to be tolerated by the biological filter in the host
cell, namely, the standard deviations σi of environmental disturbances in the
transcription and translation processes ωi; standard deviations ∆Su,i and ∆Sl,i of the
maximum and minimum promoter-RBS strengths Su,i and Sl,i, respectively; and
standard deviations ∆γxi, ∆γx3, ∆γG, ∆m, and ∆μ of the regulatory protein degradation
rate γxi, immature and mature reporter protein degradation rates γx3 and γG,
maturation rate m, and dilution rate μ due to cell growth, respectively.
4. Define an explicit cost function:
J ( fi )  E  (Ge ( fi , Ii )  Gri ( Ii ))2 dIi (2.6)
where fi denotes the combination {pR1, pR2, pR3, I} of promoter-RBS set {pR1, pR2, pR3} with
a control inducer concentration I; Ge(fi,Ii) and Gri(Ii) are the estimated and desired steady-state
fluorescence levels under the specific combination fi with input concentration u, respectively;
pR1, pR2, and pR3 represent the first, second, and third promoter-RBS components selected
from three types of promoter-RBS libraries, based on gene circuit topology; and I denotes an
appropriate control inducer. If i = 1, the synthetic gene circuit functions as a low-pass filter
and takes the inducer I1 and I2 as the input molecular signal u and control molecular signal I,
respectively. On the other hand, if i = 2, the synthetic gene circuit functions as a high-pass
filter and takes the inducers I2 and I1 as the input molecular signal u and control molecular
signal I, respectively.
If the cost function in (2.6) can be minimized by selecting the most adequate set
fi = {pR1*, pR2*, pR3*, I} from their corresponding libraries under the design specifications 1-
*
4, then the I/O filtering response Ge(fi*,Ii) of the biological filter can be optimally matched
with the specified I/O filtering response Gri(Ii) under the intrinsic parameter fluctuations and
environmental disturbances on the host cell.
Current optimization design methods for synthetic gene circuits rely on exhaustive
searching, which is a time-consuming process. When the scale of libraries is gradually
expanded, the combinatorial optimization to select the optimal promoter-RBS components to
satisfy the four design specifications requires a prohibitively large running time. This
computational complexity issue would become even more severe if the promoter-RBS
libraries are further mixed with a library of control inducer concentrations. The regulatory
inducer concentration is an external adjustable library, and therefore increases the complexity
when combined with the promoter-RBS set. In this study, a genetic algorithm (GA)-based
library search method is introduced to efficiently select the optimal combination fi* from the
mixed libraries. This is done to minimize J(fi) in (2.6) and to achieve the desired filtering
response under the four design specifications. GA is a stochastic optimization algorithm,
which is inspired by natural selection and genetic evolution mechanisms. It has been proven
to be efficient at solving constrained optimization problems with multiple local extrema in
many cases (Grefenstette 1986, Goldberg 1989, Chen, Cheng, and Lee 1995).
Based on the mixed library-based search method for the four design specifications, the
optimal combination set fi* of a promoter-RBS set and a regulatory inducer concentration can
be selected from the mixed libraries. Despite intrinsic parameter fluctuations and
environmental disturbances in design specification 3, the biological filter can be designed
with a desired I/O filtering response in design specification 1 by minimizing the cost function
J(fi) in design specification 4. The biological filter can robustly and optimally achieve the
desired I/O response by using the proposed mixed library-based search method with the GA
to efficiently select the optimal combination set fi* = {pR1*, pR2*, pR3*, I} to satisfy the four
design specifications 1-4.
2.4. BIOLOGICAL FILTER DESIGN BASED ON PROMOTER-RBS

LIBRARIES AND GENE CIRCUIT TOPOLOGY
In this section, the genetic circuit design of a biological filter is engineered for
demonstration according to the previously proposed design procedure. Afterward, a selected
set of promoter-RBS components is assembled through gene engineering experiments with an
external control inducer concentration to validate the design result.
Figure 2.3. A design example of the biological low-pass filter. The biological low-pass filter can be
divided into three cascade modules. Each module has both a promoter-RBS component selected from
the corresponding promoter-RBS libraries and a specified gene downstream of the promoter. The first
module contains a constitutive promoter-RBS component C1i and a repressor LacI gene. The second
module contains a LacI-regulated promoter-RBS component R2i and a repressor TetR gene, and the
third module contains a TetR-regulated promoter-RBS component R3i and a reporter protein GFP gene.
Promoter-RBS components C1i, R2i and R3i are to be selected from the promoter-RBS libraries in Tables
2.1-2.4 to minimize the cost function in (2.9) to optimally meet the desired low-pass filtering response
as shown in Figure 2.4.
Design Example of Biological Low-Pass Filter
Consider the design of biological low-pass filter. The desired low-pass I/O response
Gr1(I1) of the biological low-pass filter in Figure 2.3, with a specified threshold Ik = 0.01mM
and maximum fluorescence level Gmax = 2500(a.u.), is prescribed in the following for
biological low-pass filter matching in (2.4):
2500
Gr1 ( I1 )  (2.7)
2
 I 
1  1 
2
 0.01 
Note that the standard deviations of parameter fluctuations that are supposed to be
tolerated in vivo are given by
3000 3000
2500 2500
Fluorescence (a.u.)
Fluorescence (a.u.)
2000 2000
1500 1500
1000 1000
500 500
0 0
-2 -1 -2 -1
10 10 10 10
IPTG (mM) IPTG (mM)
(a) (b)
Figure 2.4. The simulation and experiment results of two biological low-pass filters. (a) The most
adequate set {C6, RL3, RT3, 0 ng/ml} is selected to match the desired I/O response of the biological low-
pass filter, with a specified threshold Ilc = 0.01 mM and maximum fluorescence level Gmax = 2500 (a.u.).
(b) The most adequate set {C6, RL3, RT3, 1 ng/ml} is selected to match the desired I/O response of the
low-pass filter, with a specified threshold Ilc = 0.015 mM and maximum fluorescence level Gmax = 2500
(a.u.). The green points are the experimental results. The gray solid lines are the desired I/O responses,
respectively.
S1u ,i  0.05S1u ,i
(S2u ,i , S2l ,i )  (0.05S2u ,i ,0.05S2l ,i )
(S3u ,i , S3l ,i )  (0.05S3u ,i ,0.05S3l ,i )
(2.8)
 xrT  0.05 xrT ,  xrL  0.05 xrL
 x3  0.05 x3 ,  G  0.05 G
m  0.05m,   0.05
and the environmental disturbances ωi of the transcription and translation processes are
independent Gaussian white noises with zero means and unit variances. In order to efficiently
solve the constrained optimal matching design problem of the filter via the proposed mixed
library-based search method, a GA-based library search method is employed to search a set
{pR1, pR2, pR3, I2} from corresponding libraries in Tables 2.1-2.4 to minimize the following
cost function:
J ( f 1* )  min E  (Ge ( f1 , I1 )  Gr1 ( I1 ))2 dI1 (2.9)

f1 { pR1 , pR2 , pR3 , I 2 }
and I1 [0.005,0.75]. Then, the most adequate promoter-RBS component from the
corresponding library and aTc concentration are found to be {pR1*, pR2*, pR3*, I2*} = {C6, RL3,
RT3, 0ng/ml}. Similarly, when the desired I/O response of the low-pass filter, with a specified
threshold Ik = 0.015mM and maximum fluorescence level Gmax = 2500(a.u.), is prescribed, the
most adequate promoter-RBS component and aTc concentration are found to be {pR1*, pR2*,
pR3*, I2*} = {C6, RL3, RT3, 1ng/ml}. It demonstrates that the threshold of the biological low-
pass filter can be externally tuned by changing the concentration of the control inducer aTc.
The experiment and desired filtering response for both low-pass filters are all shown in Figure
2.4. Clearly, the fluorescence level of the biological low-pass filter can robustly and match
the desired low-pass I/O response despite the intrinsic parameter fluctuations and
environmental disturbances.
Figure 2.5. A design example of the biological high-pass filter. The biological high-pass filter can be
divided into three cascade modules. The first module contains both a constitutive promoter-RBS
component C1i and a repressor TetR gene. The second module contains a TetR-regulated promoter-RBS
component R2i and a repressor LacI gene, and the third module contains a LacI-regulated promoter-RBS
component R3i and a reporter protein GFP gene. Promoter-RBS components C1i, R2i and R3i are to be
selected from the promoter-RBS libraries in Tables 2.1-2.4 to minimize the cost function in (2.11) to
optimally meet the desired low-pass filtering response as shown in Figure 2.6.
Design Example of Biological High-Pass Filter
Consider the design of biological high-pass filter. The desired high-pass I/O response
Gr2(I2) of the biological high-pass filter in Figure 2.5, with a specified threshold
Ihc = 0.025mM and maximum fluorescence level Gmax = 535(a.u.), is prescribed in the
following for biological high-pass filter matching in (2.5):
I2
535 
Gr 2 ( I 2 )  0.025 (2.10)
2
 I 
1  2 
2
 0.025 
tolerated in vivo are the same as in (2.8) and the environmental disturbances ωi of the
transcription and translation processes are independent Gaussian white noises with zero
means and unit variances. In order to efficiently solve the constrained optimal matching
design problem of the biological filter via the proposed mixed library-based search method, a
GA-based library search method is employed to search a set {pR1, pR2, pR3, I1} from
corresponding libraries in Tables 2.1-2.4 to minimize the following cost function:
J ( f 2* )  min E  (Ge ( f 2 , I 2 )  Gr 2 ( I 2 ))2 dI 2 (2.11)

f2 { pR1 , pR2 , pR3 , I1 }
700
700
600
600
Fluorescence (a.u.)
Fluorescence (a.u.)
500 500
400 400
300 300
200 200
100 100
-2 -1 -2 -1
10 10 10 10
IPTG (mM) IPTG (mM)
(a) (b)
Figure 2.6. The simulation and experiment results of two biological high-pass filters. (a) The most
adequate set {C6, RT3, RL3, 0 ng/ml} is selected to match the desired I/O response of the biological high-
pass filter, with a specified threshold Ihc = 0.025 mM and maximum fluorescence level Gmax = 535 (a.u.).
(b) The most adequate set {C6, RT3, RL3, 1 ng/ml} is selected to match the desired I/O response of the
high-pass filter, with a specified threshold Ihc = 0.03 mM and maximum fluorescence level Gmax = 535
(a.u.). The green points are the experimental results. The gray solid lines are the desired I/O responses,
respectively.
and I2 [0.005,0.75]. Then, the most adequate promoter-RBS component from the
corresponding library and aTc concentration are found to be {pR1*, pR2*, pR3*, I1*} = {C6, RT3,
RL3, 0ng/ml}. Similarly, when the desired I/O response of the high-pass filter, with a specified
threshold Ihc = 0.03mM and maximum fluorescence level Gmax = 535(a.u.), is prescribed, the
most adequate promoter-RBS component and aTc concentration are found to be {pR1*, pR2*,
pR3*, I1*} = {C6, RT3, RL3, 1ng/ml}. It demonstrates that the threshold of the biological high-
pass filter can be externally tuned by changing the concentration of the control inducer aTc.
The experiment and desired filtering response for both high-pass filters are all shown in
Figure 2.6. Clearly, the fluorescence level of the biological high-pass filter can robustly and
match the desired high-pass I/O response despite the intrinsic parameter fluctuations and
2.5. DISCUSSION
Synthetic biologists utilize mathematical models to describe biochemical reactions of
genetic circuits and then implement the biological gene circuits by genetic engineering
techniques. The mathematical models of promoter activities were developed by considering
the binding reactions of TFs to DNA in the thermodynamic equilibrium (Kinkhabwala and
Guet 2008, Bintu et al., 2005). A mathematical model for predicting the relative mRNA
translation rate with different RBS sequences was also proposed (Salis, Mirsky, and Voigt
2009). In our study of biological filter design, the effects of promoter and RBS are merged
together to discuss the transcription and translation regulatory strength simultaneously
because the promoter activity may be affected by the RBS fused to the downstream of the
promoter (Mutalik et al., 2013).
When the mathematical models of protein expressions are properly constructed, the main
challenge in synthetic gene circuit design is to select the appropriate genetic components to
assemble the regulatory network and produce the desired performance reliably. There is not a
suitable and general screening method for us to correctly pick up the proper components for
our design purposes because the characteristics of biological components have not been
quantified appropriately. Recently research has shown that the studies of promoter libraries
have significant progress by quantitative measurements of reporter protein. Certain promoter
libraries have been gradually established by measuring the activity of promoters in Biobrick
(Kelly et al., 2009, Ellis, Wang, and Collins 2009a), but these characterized libraries still have
compatibility issues for synthetic biological filter design, i.e., the same background of
experimental operation is required when we refer the experimental data and parameter from
the references (Kelly et al., 2009, Ellis, Wang, and Collins 2009a, Wu, Lee, and Chen 2011a,
Huynh et al., 2012, Wang et al., 2011). Hence, the refinement and standardization of synthetic
biological components need to be established (Canton, Labno, and Endy 2008a, Arkin 2008).
In fact, standardized components with quantitative and qualitative descriptions are widely
used in many engineering disciplines. An engineer is allowed to quickly determine whether
the component might be employed to meet the requirements of a system by the standardized
designs and components (Canton, Labno, and Endy 2008a). Here, the promoter and RBS are
considered as a promoter-RBS component to be identified together similar to reference
(Wang et al., 2011) and with the regulatory protein and reporter protein genes, the externally
tunable biological filter can be implemented. Therefore, the establishment of the promoter-
RBS libraries is very important for engineering a synthetic biological filter circuit with
external tunable control.
As for the biological filter implementation, the biological filter is different from the
general filter for frequency filtering in the field of electrical engineering. In the biological
engineering, the biological filter is regarded as a powerful detector or a sensor with response
to the specific concentration of the particular molecular signal. Moreover, from the past
studies of biological filter, a systematic design approach and external tunable technique for
wide range adjustment of the threshold for the biological filter still need to be developed. The
past studies of biological filter are often employed in pattern formation (Sohka et al., 2009a,
Sohka, Heins, and Ostermeier 2009a) since the effective response range of the filter is limited
and only certain molecular signals can be detected. For practicability, applicability and
scalability, a novel synthetic transcriptional cascade is reconstructed to function as a
biological filter. In line with the plug and play features of synthetic circuits, different
promoter-RBS components corresponding to different regulatory molecules can be replaced
for different applications under the same gene circuit topology structure of the biological
filter. Therefore, the biological filters for detecting different molecular signals can be
engineered by substituting with different promoter-RBS components and different regulatory
molecules, i.e., different inducers, activators and repressors. And the same type of promoter-
RBS components with different strengths can also be substituted to obtain a wide range
adjustment of the threshold. In addition, the proposed biological filter has an additional
control scheme for adjusting the filtering threshold by tuning the concentration of control
inducer externally. By using the mixed library-based search method to achieve the optimal
I/O response matching design, synthetic biologists can select suitable promoter-RBS
components and a proper control inducer concentration for the biological filter to meet the
user-defined design specifications.
After both of the low-pass filter and high-pass filter are engineered, a band-pass filter can
further be constructed by superposing a low-pass filter on a high-pass filter in a manner that
the threshold of high-pass filter is lower than the threshold of low-pass filter (Sohka et al.,
2009b, Sohka, Heins, and Ostermeier 2009b, Kämpf et al., 2012). Following the conjunction-
type operation, the outputs of biological low-pass and high-pass filters are rewired. The
conjunction-type operation is similar with an AND gate in biological logic gates (Wang et al.,
2011, Tamsir, Tabor, and Voigt 2010, Regot et al., 2010). The outcome was produced only in
the presence of both of the two inputs, just like the feature of AND gate, i.e., output is true if
and only if both two inputs are true. Therefore, a biological band-pass filter can be
constructed subsequently by rewiring the biological low-pass and high-pass filter modules
according to fundamental computational operations. As long as the biological low-pass and
high-pass filters with the same input molecular signal but with different specified thresholds
are engineered individually by the above-mentioned design procedure, then we can introduce
an AND-type operation rewiring in the conjunction-type configuration to produce a desired
band-pass I/O response.
A key next step in synthetic biology is to integrate simple biological modules into higher-
order systems. Our designed biological filters can be expanded into a biological switchboard
that independently controls the expression of multiple genes in parallel. A switchboard is as
an assembly of switches that is useful for controlling and linking electrical circuits. A genetic
switchboard as an assembly of orthogonal genetic switches is useful for controlling and
linking biological circuits and pathways (Callura, Cantor, and Collins 2012). With the more
different types of promoter-RBS components in response to more different molecular signals
and more different types of reporters, we can reconstruct the biological switchboard by
integrating several designed biological filters. Because the internal adjustability by

substituting the promoter-RBS components with different strengths and the external tuning by
the control inducer concentration for the threshold of our designed biological filter, the
reconstructed biological switchboard will have the same tunability for each genetic switch in
the overall switchboard. For example, we can first employ the same four environmentally
sensitive promoters, PluxI, PLlexO, PLfurO and PMgrB mentioned in (Callura, Cantor, and Collins
2012), fused to appropriate RBSs in our study to build new types of promoter-RBS libraries.
Then four externally tunable biological filters are engineered individually with different input
molecule signals, AHL, MMC, Iron Chelator, and Mg2+ and four different reporters GFP,
mCherry, β-galactosidase (LacZ), and firefly luciferase, respectively. In addition, each
biological filter has another control inducer to tune the threshold externally.
Biological networks that regulate many gene expressions in parallel lend themselves to a
variety of biotechnological applications and particularly to metabolic engineering. Synthetic
biologist has made effort to develop biological components for metabolic engineering, i.e.,
biosynthetic pathways and enzyme scaffolds (Keasling 2012, Boyle and Silver 2011, Lynch
and Gill 2011). From above discussions, the externally tunable biological switchboard is a
well-defined biological module that possesses more flexibility in aid of different kinds of
metabolic engineering strategies. Therefore, in the future, an externally tunable metabolism
switchboard that controls carbon flux can be further constructed through three E. coli glucose-
utilization pathways, the Embden-Meyerhof (EMP), Entner–Doudoroff (EDP), and pentose
phosphate (PPP) pathways, via the regulation of four different genes, pgi, zwf, edd, and gnd,
respectively (Callura, Cantor, and Collins 2012).rem ipsum dolor sit amet, consectetuer
adipiscing elit. Maecenas porttitor congue massa. Fusce posuere, magna sed pulvinar
ultricies, purus lectus malesuada libero, sit amet commodo magna eros quis urna.
A variety of applications in living cells could be produced by integrating engineered
synthetic biological subsystems, i.e., multi-input sensing, sophisticated information
processing, and precisely regulated actuation. Unlike the above-mentioned multi-input and
multi-output structure of externally tunable biological switchboard for different metabolic
engineering strategies, here we provide a multi-input and single-output system that integrates
several designed biological filters and a fundamental computational operation to function as a
biological classifier in aid of anticancer therapies. A scalable synthetic regulatory circuit
called cell-type “classifier” can sense expression levels of a customizable set of endogenous
molecule signals such as microRNAs and trigger a cellular response only if the expression
levels match a prescribed profile of interest (Xie et al., 2011). As a scheme of externally
tunable HeLa cancer cell classifier, multiple cancer-specific biomarkers can be sensed and
integrated in this system to provide precise control of therapeutic agent hBax protein. Two
HeLa-high markers, miR-21 and miR-17, are taken as inputs to the biological high-pass
filters. Three HeLa-low markers, miR-141, miR-142(3p) and miR-146a, are taken as inputs to
the biological low-pass filters. Then the outputs of individual biological filters can be rewired
by the conjunction-type operation. Hence, the HeLa cells can be identified to trigger
apoptosis through the externally tunable HeLa cancer cell classifier.
2.6. CONCLUSION
In this chapter, a systematic design approach is developed to engineer robust externally
tunable biological filters rapidly and reliably. For the design procedure of the biological filter,
we first established three kinds of promoter-RBS libraries based on promoter-RBS kinetic
strengths and mathematical model of gene transcriptional and translational expression. For
practicability, applicability and scalability, a novel synthetic gene circuit with three promoter-
RBS components is reconstructed to function as a biological filter based on a cascade gene
circuit topology as shown in Figure 2.1. According to four design specifications 1-4, a robust
externally tunable biological filter is engineered by searching an adequate set of three
promoter-RBS components and a control inducer concentration from the corresponding
libraries to achieve the optimal reference matching for the specified I/O filtering response in
spite of the parameter fluctuations or external molecular noises. Notably, a mixed library-
based search method is employed to efficiently find the most adequate set of promoter-RBS
components and inducer concentration via genetic algorithm. Further, the external tunability
of these synthetic biological filters that shows the potential in practical applications is also
verified in silico. The computational simulations and practical experiments are also given to
illustrate the design procedures and confirm the performance of the proposed systematic
design approach of the biological filter. For biological applications of the biological filter in
synthetic biotechnology, a key next step in the future is to integrate simple biological filter
modules into higher-order systems. An externally tunable biological band-pass filter can be
constructed by rewiring a biological low-pass and high-pass filter modules via the
conjunction-type operations. Furthermore, the proposed externally tunable biological filters
can be expanded into an externally tunable biological switchboard to independently control
the expression of multiple genes in parallel and could be practically employed to produce an
externally tunable metabolism switchboard in metabolic engineering. Another important
application is an expansion to several designed biological filters with a conjunction-type
operation to function as a biological classifier to aid anticancer therapies.
2.7. APPENDIX
Three kinds of promoter-RBS libraries are constructed according to the characteristic
indexes of components, namely, constitutive, repressor-regulated, and activator-regulated
promoter-RBS libraries. The anhydrotetracycline (ATc)-responsive LibTetR and the isopropyl-
β-D-thiogalactopyranoside (IPTG)-responsive library LibLacI are considered in the repressor-
regulated promoter-RBS library Librepressor; the 3-oxy-hexanoyl-homoserine-lactone-
responsive library LibLuxR is considered in the activator-regulated promoter-RBS library
Libactivator.
In general, transcription of a gene gives rise to mRNA, which is subsequently translated
into a protein. The synthesis of mRNA and protein is counterbalanced by dilution and
degradation of the gene products. These processes determine the net accumulation of mRNA
and protein in the cell. The reporter protein is usually used to produce fluorescence, which
can be measured by ELISA. The expression of the reporter gene roughly follows the same
stages as those of the protein; however, the maturation time may be different. The response of
the reporter protein to light excitation is dependent on post-translational modifications,

notably, folding of the protein to the fluorescent form (Tsien 1998). This maturation process
gives rise to an additional reaction step from reporter protein to fluorescence, hence leading to
some difference between protein and reporter protein.
Here, the effects of translation and translation are integrated as a combined reaction
because the half-lives of mRNA are shorter than the corresponding half-lives of proteins. Let
us consider a protein that is produced by the promoter-RBS component c, degraded by
protease, and diluted by cell growth. Denoting the concentration of the protein as x, the
protein expression arising from a promoter-RBS component c can be described as
x(t )  P(Su , Sl , TF , I )  (   x ) x(t ) (2.12)
where x(t) denotes the concentration of protein; P(Su, Sl, TF, I) denotes the promoter-RBS
regulation function with the maximum and minimum promoter-RBS strengths Su and Sl,
respectively, from the corresponding libraries; TF denotes the related transcription factor
concentration; I denotes the related inducer concentration. The dilution rate due to cell growth
and the degradation rate of the corresponding protein are denoted by μ and γx, respectively.
On the other hand, the protein expression of a reporter protein produced from a promoter-
RBS component c is described as
x(t )  P(Su , Sl , TF , I )  (m     i ) x(t )

(2.13)
g (t )  mx(t )  (    g ) g (t )
where x(t) and g(t) denote the concentrations of immature and mature reporter protein,
respectively; γi and γg denote the degradation rates of immature and mature reporter protein,
respectively; m denotes the maturation rate of the reporter protein. In general, the
concentration of mature reporter protein is represented by fluorescence, which can be
measured by ELISA.
When a promoter-RBS component c is placed at the upstream of the protein, protein
production cannot be measured. However, when the promoter-RBS component c is placed at
the upstream of the reporter protein, we can measure the fluorescence by ELISA. Hence, we
can observe the different protein expressions when a promoter-RBS component c is placed at
the upstream of the protein or reporter protein, which can be modeled by (2.12) and (2.13),
respectively. Therefore, from the viewpoint of systems biology, the dynamic characterization
of the promoter-RBS component is important and feasible. We can measure fluorescence by
ELISA to characterize the promoter-RBS component, and subsequently obtain the promoter-
RBS strength P of a promoter-RBS component c from the fluorescent data and dynamic
model in (2.13). Thereafter, the characteristic indexes of promoter-RBS libraries can be
constructed by collecting the promoter-RBS strengths of promoter-RBS components. Note
that the strength of the promoter-RBS component c for regulating the protein or reporter
protein is the same. Hence, we can also estimate the regulation of the protein if we can
identify the promoter-RBS strength from the fluorescent data measured using the reporter
protein. According to this assumption, the engineering of a synthetic gene circuit with desired
behaviors becomes possible if we can select promoter-RBS components with adequate
promoter-RBS strengths.
Construction of the Constitutive Promoter-RBS Library
Before identifying the promoter-RBS strength by measuring fluorescence, two important

factors should be considered. First, the newly synthesized fluorescent protein must undergo a
series of self-modifications in order to become fluorescent (Tsien 1998). The appearance of
fluorescence reportedly lags about 3.5 hours behind the actual synthesis of the protein (using
the wild-type green fluorescent protein (GFP) from the jellyfish Aequorea victoria) (Heim,
Cubitt, and Tsien 1995, Albano et al., 1996). Second, the growth rate of the cell is an
important factor. The faster that cells divide, the faster is the dilution of GFP (Leveau and
Lindow 2001). Therefore, growth rate should be also considered as a parameter in the
stochastic model for promoter-RBS strength identification. The dilution rate μ due to cell
growth should be considered initially. Since the cells grow in the same media and condition in
the experiments, the μ can be considered as the same value. Synthetic biologists can estimate
the dilution rate μ in their experiments using different media and experimental conditions.
The dilution rate of a cell can be estimated from the cell density by means of the classical
formula:
d ln s ds 1
  (2.14)
dt dt s
where s denotes the O.D. 600 value in the experiment. Here, according to the experimental
data, the dilution rate is 0.011946 ± 0.000198 min−1 at 95% confidence.
On the basis of the above two factors, that is, maturation and dilution rates, and the
resistance of the constitutive promoter-RBS component Ci, to any TF, its downstream gene
can be expressed continually at the maximum expression of the component. The stochastic
model for protein expression from a constitutive promoter-RBS component (Kelly et al.,
2009, Leveau and Lindow 2001, Wang, Errede, and Elston 2008, de Jong et al., 2010) is
described as
x  t   Pc (Su ,0,0,0)      i  x  t   1  t 
(2.15)
g  t   mx  t       g  g  t   2  t 
where x(t) and g(t) denote the concentrations of immature and mature reporter protein,
respectively; γi and γg are the degradation rates of immature and mature reporter protein,
respectively; Pc denotes the promoter-RBS strength; m denotes the maturation rate of reporter
protein; ω1(t) and ω2(t) denote environmental noises in the immature and mature reporter
proteins, respectively. Further, the promoter-RBS regulation function of constitutive
promoter-RBS components Ci from the constitutive promoter-RBS library Libconst can be
represented as
Pc ( Su ,i ,0,0,0)  Su ,i
(2.16)
{Su ,i }  Libconst
Realistically, the mRNA transcripts of the protein, which initiate production of non-
fluorescent reporter protein x(t), subsequently mature into fluorescence g(t) that is measured
by ELISA. In general, GFP is typically used as the reporter protein. According to the
literature, the maturation rate m can be calculated as ln2 divided by the time constant of GFP
maturation, which has been determined as 2.0 h for wide-type GFP (wtGFP) and 0.45 h for
faster-folding S65T mutants such as EGFP (Heim, Cubitt, and Tsien 1995). The model in
(2.15) is considered to be effective while transcription and translation are integrated as a
combined reaction because mRNA half-lives are shorter than the corresponding protein half-
lives (typically, the half-life of mRNA of E. coli is 6.8 min (Selinger et al., 2003)). This
consideration is similar to the results of previous studies (Alper et al., 2005), despite the fact
that there are a variety of both promoter and RBS strengths that could be combined to allow
synthetic biologists to achieve a range of transcription and translation strengths.
The detailed procedure used to construct the constitutive promoter-RBS library Libconst is
divided into four steps: (I) Selecting of a promoter and a RBS to construct a promoter-RBS
component Ci, and then placing them upstream of the reporter protein; (II) measuring the
time-profile fluorescence of the selected promoter-RBS component Ci; (III) estimating the
dilution rate μ by (2.14), obtaining the remaining parameters m, γi and γg from the literature
(Canton, Labno, and Endy 2008a, Kelly et al., 2009, Leveau and Lindow 2001, Heim, Cubitt,
and Tsien 1995), and then estimating the promoter-RBS strength Su,i from (2.15) by using
identification techniques; (IV) constructing the indexes of the constitutive promoter-RBS
library Libconst, including the promoter-RBS component Ci and the identified promoter-RBS
strength Su,i.
Here, we select three promoters, namely, BBa_J23101, BBa_J23105, and BBa_J23106;
and three RBSs, namely, BBa_B0031, BBa_B0032, and BBa_B0034 to construct promoter-
RBS components. Afterward, the promoter-RBS strengths for each promoter-RBS component
can be identified by the least-squares method (Johansson 1993). Hence, the indexes of
constitutive promoter-RBS library Libconst are constructed and based on the indexes of its
promoter-RBS components listed in Table 2.1.
Table 2.1. The constitutive promoter-RBS library Libconst
Index BioBrick component Su

C1 J23101-B0031 94.520
C2 J23101-B0032 65.892
C3 J23101-B0034 237.000
C4 J23105-B0031 14.493
C5 J23105-B0032 6.512
C6 J23105-B0034 28.586
C7 J23106-B0031 22.011
C8 J23106-B0032 12.193
C9 J23106-B0034 58.781
Construction of the Repressor-Regulated Promoter-RBS Library
The repressor-regulated promoter-RBS component produces the maximum and minimum

gene expressions with and without saturation concentrations of the inducer, respectively.
Therefore, the concentration of inducer should be considered in the characterization of the
repressor-regulated promoter-RBS component, which can be described by a simple
representative scheme. The repressor is produced by the constitutive promoter-RBS
component, which can be selected from the constitutive promoter-RBS library Libconst, and
then the inducer is added to regulate the concentration of repressor. The repressor, regulated
by the inducer, binds to the repressor-regulated promoter-RBS component Ri to generate
fluorescence.
The relationship between the concentrations of the repressor and inducer, which
characterizes the repressor-regulated promoter-RBS component, is difficult to determine, and
requires data on the relationship between the concentrations of inducer and fluorescence in
addition to time-profile data. When the synthetic gene circuit reaches the steady state,
different concentrations of inducer produce different fluorescence patterns of reporter protein.
Hence, the dose-dependent measurement of inducer and fluorescence is also important for
estimating the strength of promoter-RBS components when constructing promoter-RBS
libraries.
For the construction of repressor-regulated promoter-RBS libraries Librepressor, we
summarize the construction procedure of Librepressor. Suppose the repressor-regulated
promoter-RBS component Ri has maximum and minimum promoter-RBS strengths Su,i and
Sl,i, respectively. The construction of repressor-regulated promoter-RBS libraries Librepressor
can be subsequently divided into two kinds of gene circuit topology. First, the repressor-
regulated promoter-RBS component Ri can be constructed by following the representative
scheme, where the constitutive promoter-RBS component is replaced by the repressor-
regulated promoter-RBS component. Therefore, the maximum promoter-RBS strength Su,i of
Ri can be identified using the identification technique in (2.15). In fact, each repressor-
regulated promoter-RBS component produces the maximum gene expression with the inducer
at saturation concentration. The maximum gene expression should be the same as that
produced with the promoter-RBS strength Su,i. Second, different concentrations of inducer can
drive different fluorescence patterns, implying that different strengths of repressor-regulated
promoter-RBS components correspond to different concentrations of inducer. In this case, the
constitutive promoter selected from the constitutive promoter-RBS library Libconst is used to
produce the stable and constitutive repressor xr, which is regulated by the inducer Ir. In a
repressor-regulated promoter-RBS component, xr represses the gene expression under the
control of a repressor-regulated promoter. When the inducer Ir is added, it binds to the
repressor xr to influence gene expression under the control of the repressor-regulated
promoter. When Ir is removed, xr becomes active, thereby repressing gene expression again.
The repressor activity xr*(xr,Ir) can be described as follows:
xr
xr* ( xr , I r )  (2.17)
I
1 r
K Ir
where xr denotes the total repressor concentration, including free and inducer-bound
repressor; Ir denotes the inducer concentration for regulating the repressor activity xr*(xr,Ir);
KI denotes the inducer-repressor dissociation rate.
The promoter-RBS regulation function of repressor-regulated promoter-RBS component
Ri dependent on the repressor activity xr*(xr,Ir) can then be represented as follows:
S u , i  Sl , i
Pr ( Su ,i , Sl ,i , xr , I r )  Sl ,i  nr
 x* ( x , I ) 
1  r r r  (2.18)
 Kr 
S u ,i , Sl ,i   Librepressor
where Kr and nr denote the repressor-promoter binding affinity and binding cooperativity
between the regulatory protein and repressor-regulated promoter-RBS component,
respectively; Su,i and Sl,i denote the maximum and minimum repressor-regulated promoter-
RBS strengths from the corresponding promoter-RBS library Librepressor, respectively. In this
manner, the characteristics of repressor-regulated promoter-RBS components, including the
maximum and minimum promoter-RBS strengths as well as the binding affinities, are
collected as characteristic indexes of promoter-RBS components to construct the repressor-
regulated promoter-RBS libraries, Librepressor.
Here, two repressor-regulated promoter-RBS libraries, LibTetR and LibLacI, are constructed
by the characteristic indexes of repressor-regulated promoter-RBS components that contain
the promoter-RBS strengths of the ATc-responsive promoter Ptet and the IPTG-responsive
promoter Plac combined with different RBSs. The identification results of the promoter-RBS
libraries LibTetR and LibLacI are shown in Table 2.2 and Table 2.3, respectively. In the future,
synthetic biologists can construct more repressor-regulated promoter-RBS libraries following
procedures similar to those described here.
Table 2.2. The repressor-regulated promoter-RBS library LibTetR
Index BioBrick component Su Sl Parameters

RT1 R0040-B0031 11.965 3.031 Katc = 6.7 ng/ml
RT2 R0040-B0032 8.740 3.381 KTetR = 45 nM
RT3 R0040-B0034 162.557 4.837 nTetR = 2
Table 2.3. The repressor-regulated promoter-RBS library LibLacI

RL1 R0010-B0031 3.859 2.039 KIPTG = 0.008 mM
RL2 R0010-B0032 2.586 2.529 KLacI = 27.484 nM
RL3 R0010-B0034 33.934 1.495 nLacI = 2
Construction of the Activator-Regulated Promoter-RBS Library
In general, the promoters that can be activated are also used to engineer synthetic gene
circuits. By the same manner, an activated promoter, regulated by the inducer, always has the
maximum and minimum expression with and without saturation concentrations of the
inducers, respectively. Hence, for the construction of activator-regulated promoter-RBS
libraries, the concentration of inducer should also be considered when characterizing the
activator-regulated promoter-RBS component. The activator is produced by the constitutive
promoter-RBS component, which can be selected from the constitutive promoter-RBS library
Libconst. The inducer is added to form the complex with the activator, and it then binds to the
activator-regulated promoter-RBS component to influence the gene expression.
It is worthy to note that the construction of activator-regulated promoter-RBS libraries
differs from the construction of repressor-regulated promoter-RBS libraries. The key point
when measuring activator-regulated promoter-RBS components is that there is no
fluorescence production when no inducer is added. Hence, the measurement method should
be modified to aid the construction of activator-regulated promoter-RBS libraries. Suppose
the activator-regulated promoter-RBS component Ai has the maximum and minimum
promoter-RBS strengths Su,i and Sl,i with and without saturated inducer-bound activator
binding, respectively. Therefore, the procedure for constructing the activator-regulated
promoter-RBS library Libactivator can be divided into two steps. First, we take specific
concentrations of inducers to measure fluorescence and obtain time-profile data of
fluorescence. Second, different concentrations of inducer are used to drive different
fluorescence patterns, showing different strengths of activator-regulated promoter-RBS
components corresponding to different concentrations of the inducer. In this case, the
constitutive promoter selected from the constitutive promoter-RBS library Libconst is used to
produce the stable and constitutive activator xa, which is regulated by the inducer Ia. In an
activator-regulated promoter-RBS component, xa activates gene expression under the control
of an activator-regulated promoter. The concentration of inducer-bound activator, which is
known as the activator activity xa*(xa,Ia) and regulated by external inducer Ia through an
inducer binding reaction, can be described as follows:
xa  I a
xa* ( xa , I a )  (2.19)
I a  K Ia
where xa denotes the total activator concentration including free and inducer-bound activator;
Ia denotes inducer concentration for regulating the activator activity xa*(xa,Ia); KIa denotes the
inducer-activator dissociation rate. Therefore, the regulatory strength of activator-regulated
promoter-RBS component Ai dependent on the activator activity xa*(xa,Ia) can be represented
by a promoter-RBS regulation function as follows (Alon 2007):
( Su ,i  Sl ,i )   xa* ( xa , I a ) 
na
Pa ( Su ,i , Sl ,i , xa , I a )  Sl ,i 
 x (x , I )
na
*
a a a  K a na (2.20)
S u ,i , Sl ,i   Libactivator
where Ka and na denote the activator–promoter binding affinity and binding cooperativity
between the regulatory protein and activator-regulated promoter-RBS component,
respectively; Su,i and Sl,i denote the maximum and minimum activator-regulated promoter-
RBS strengths from the corresponding promoter-RBS library Libactivator, respectively. In this
manner, the characteristics of activator-regulated promoter-RBS components including the
maximum and minimum promoter-RBS strengths and binding affinity are collected to
construct the indexes of activator-regulated promoter-RBS libraries.
Here, the promoter-RBS library LibLuxR of activator-regulated promoter-RBS
components, in which the index of the library contains the promoter-RBS strengths of the N-
acyl homoserine lactone-responsive promoter Plux combined with different RBSs, are
constructed. The identification results of promoter-RBS library LibLuxR are shown
in Table 2.4.
Table 2.4. The activator-regulated promoter-RBS library

AL1 R0062-B0032 360.761 7.970 KAHL = 4.541 nM
AL2 R0062-B0034 460.353 8.978 KLuxR = 55.964nM
nLuxR = 1
REFERENCES
Albano, C. R., L. RandersEichhorn, Q. Chang, W. E. Bentley, and G. Rao. 1996.
“Quantitative measurement of green fluorescent protein expression.” Biotechnology
Techniques 10 (12):953-958.
Alon, U. 2007. An Introduction to Systems Biology: Design Principles of Biological Circuits.
London: Chapman & Hall/CRC.
Alper, H., C. Fischer, E. Nevoigt, and G. Stephanopoulos. 2005. “Tuning genetic control
through promoter engineering.” Proceedings of the National Academy of Sciences of the
United States of America 102 (36):12678-12683.
Arkin, A. 2008. “Setting the standard in synthetic biology.” Nat Biotechnol 26 (7):771-773.
Bintu, L., N. E. Buchler, H. G. Garcia, U. Gerland, T. Hwa, J. Kondev, and R. Phillips. 2005.
“Transcriptional regulation by the numbers: models.” Current Opinion in Genetics &
Development 15 (2):116-124.
Boyle, P.M., and P.A. Silver. 2011. “Parts plus pipes: Synthetic biology approaches to
metabolic engineering.” Metab Eng.
Bulter, Thomas, Sun-Gu Lee, Wilson WaiChun Wong, Eileen Fung, Michael R. Connor, and
James C. Liao. 2004. “Design of artificial cell–cell communication using gene and
metabolic networks.” Proceedings of the National Academy of Sciences of the United
States of America 101 (8):2299-2304.
Callura, J.M., C.R. Cantor, and J.J. Collins. 2012. “Genetic switchboard for synthetic biology
applications.” Proceedings of the National Academy of Sciences 109 (15):5850-5855.
Canton, B., A. Labno, and D. Endy. 2008a. “Refinement and standardization of synthetic
biological parts and devices.” Nature biotechnology 26 (7):787-794.
Canton, Barry, Anna Labno, and Drew Endy. 2008b. “Refinement and standardization of
synthetic biological parts and devices.” Nat Biotech 26 (7):787-793.
Chen, B.S., Y.M. Cheng, and C.H. Lee. 1995. “A genetic approach to mixed H2/H∞ optimal
PID control.” IEEE Control Systems Magazine 15 (5):51-60.
de Jong, H., C. Ranquet, D. Ropers, C. Pinel, and J. Geiselmann. 2010. “Experimental and
computational validation of models of fluorescent and luminescent reporter genes in
bacteria.” BMC Syst Biol 4:55.
Ellis, T., X. Wang, and J. J. Collins. 2009a. “Diversity-based, model-guided construction of
synthetic gene networks with predicted functions.” Nat Biotechnol 27 (5):465-71. doi:
10.1038/nbt.1536.
Ellis, Tom, Xiao Wang, and James J. Collins. 2009b. “Diversity-based, model-guided
construction of synthetic gene networks with predicted functions.” Nat Biotech 27
(5):465-471.
William, R., and James C. Liao. 2000. “Improving lycopene production in Escherichia coli by
engineering metabolic control.” Nat Biotech 18 (5):533-537.
Goldberg, D.E. 1989. Genetic algorithms in search, optimization, and machine learning:
Addison-wesley.
Grefenstette, J.J. 1986. “Optimization of control parameters for genetic algorithms.” IEEE
Trans. Systems, Man and Cybernetics 16 (1):122-128.
Heim, R., A. B. Cubitt, and R. Y. Tsien. 1995. “Improved green fluorescence.” Nature 373
(6516):663-4. doi: 10.1038/373663b0.
Huynh, L., J. Kececioglu, M. Koppe, and I. Tagkopoulos. 2012. “Automatic design of
synthetic gene circuits through mixed integer non-linear programming.” PLoS One 7
(4):e35529.
Johansson, Rolf. 1993. System modeling & identification. New Jersey: Prentice-Hall
International.
Kämpf, M.M., R. Engesser, M. Busacker, M. Hörner, M. Karlsson, M.D. Zurbriggen, M.
Fussenegger, J. Timmer, and W. Weber. 2012. “Rewiring and Dosing of Systems
Modules as Design Approach for Synthetic Mammalian Signaling Networks.” Molecular
BioSystems.
Kampf, Michael M., Raphael Engesser, Moritz Busacker, Maximilian Horner, Maria
Karlsson, Matias D. Zurbriggen, Martin Fussenegger, Jens Timmer, and Wilfried Weber.
2012. “Rewiring and dosing of systems modules as a design approach for synthetic
mammalian signaling networks.” Molecular BioSystems 8 (6):1824-1832.
Keasling, J.D. 2012. “Synthetic biology and the development of tools for metabolic
engineering.” Metab Eng.
Kelly, J. R., A. J. Rubin, J. H. Davis, C. M. Ajo-Franklin, J. Cumbers, M. J. Czar, K. de
Mora, A. L. Glieberman, D. D. Monie, and D. Endy. 2009. “Measuring the activity of
BioBrick promoters using an in vivo reference standard.” J Biol Eng 3:4.
Kinkhabwala, A., and C.C. Guet. 2008. “Uncovering cis regulatory codes using synthetic
promoter shuffling.” PloS one 3 (4):e2030.
Leveau, J. H., and S. E. Lindow. 2001. “Predictive and interpretive simulation of green
fluorescent protein expression in reporter bacteria.” J Bacteriol 183 (23):6752-62. doi:
10.1128/JB.183.23.6752-6762.2001.
Lynch, S.A., and R.T. Gill. 2011. “Synthetic Biology: New strategies for directing design.”
Metab Eng.
Mutalik, V. K., J. C. Guimaraes, G. Cambray, C. Lam, M. J. Christoffersen, Q. A. Mai, A. B.
Tran, M. Paull, J. D. Keasling, A. P. Arkin, and D. Endy. 2013. “Precise and reliable
gene expression via standard transcription and translation initiation elements.” Nature
Methods 10 (4):354-+.
Regot, S., J. Macia, N. Conde, K. Furukawa, J. Kjellén, T. Peeters, S. Hohmann, E. De Nadal,
F. Posas, and R. Solé. 2010. “Distributed biological computation with multicellular
engineered networks.” Nature 469 (7329):207-211.
Salis, H.M., E.A. Mirsky, and C.A. Voigt. 2009. “Automated design of synthetic ribosome
binding sites to control protein expression.” Nat Biotechnol 27 (10):946-950.
Selinger, D. W., R. M. Saxena, K. J. Cheung, G. M. Church, and C. Rosenow. 2003. “Global
RNA half-life analysis in Escherichia coli reveals positional patterns of transcript
degradation.” Genome research 13 (2):216-23.
Sohka, T., R. A. Heins, and M. Ostermeier. 2009a. “Morphogen-defined patterning of
Escherichia coli enabled by an externally tunable band-pass filter.” J Biol Eng 3:10. doi:
10.1186/1754-1611-3-10.
Sohka, T., R. A. Heins, R. M. Phelan, J. M. Greisler, C. A. Townsend, and M. Ostermeier.
2009a. “An externally tunable bacterial band-pass filter.” Proc Natl Acad Sci U S A 106
(25):10135-40.
Sohka, T., R.A. Heins, and M. Ostermeier. 2009b. “Morphogen-defined patterning of
Escherichia coli enabled by an externally tunable band-pass filter.” J Biol Eng 3 (1):10.
Sohka, T., R.A. Heins, R.M. Phelan, J.M. Greisler, C.A. Townsend, and M. Ostermeier.
2009b. “An externally tunable bacterial band-pass filter.” Proceedings of the National
Academy of Sciences 106 (25):10135.
Sohka, Takayuki, Richard A. Heins, and Marc Ostermeier. 2009c. “Morphogen-defined
patterning of Escherichia coli enabled by an externally tunable band-pass filter.” Journal
of Biological Engineering 3 (1):1-9.
Townsend, and Marc Ostermeier. 2009c. “An externally tunable bacterial band-pass
filter.” Proceedings of the National Academy of Sciences 106 (25):10135-10140.
Tamsir, A., J.J. Tabor, and C.A. Voigt. 2010. “Robust multicellular computing using
genetically encoded NOR gates and chemical/wires/'.” Nature 469 (7329):212-215.
Tsien, R. Y. 1998. “The green fluorescent protein.” Annual Review of Biochemistry 67:509-
544.
Wang, B., R.I. Kitney, N. Joly, and M. Buck. 2011. “Engineering modular and orthogonal
genetic logic gates for robust digital-like synthetic biology.” Nat Commun 2:508.
Wang, X., B. Errede, and T. C. Elston. 2008. “Mathematical analysis and quantification of
fluorescent proteins as transcriptional reporters.” Biophys J 94 (6):2017-26.
Wu, C. H., H. C. Lee, and B. S. Chen. 2011a. “Robust synthetic gene network design via
library-based search method.” Bioinformatics 27 (19):2700-6.
Wu, Chih-Hung, Hsiao-Ching Lee, and Bor-Sen Chen. 2011b. “Robust synthetic gene
network design via library-based search method.” Bioinformatics 27 (19):2700-2706.
Wu, Chih-Hung, Weihei Zhang, and Bor-Sen Chen. 2011. “Multiobjective H2/H∞ synthetic
gene network design based on promoter libraries.” Mathematical Biosciences 233
(2):111-125.
Xie, Z., L. Wroblewska, L. Prochazka, R. Weiss, and Y. Benenson. 2011. “Multi-input
RNAi-based logic circuit for identification of specific cancer cells.” Science's STKE 333
(6047):1307.
Chapter 3
SYSTEMATIC DESIGN METHODOLOGY

FOR ROBUST GENETIC TRANSISTORS
ABSTRACT
Synthetic genetic transistors are vital for signal amplification and switching in gene
circuits. However, it is still problematic to efficiently select the adequate promoters,
ribosome binding sides (RBSs) and inducer concentrations to construct a genetic
transistor with the desired linear amplification or switching in the Input/Output
characteristics for practical applications. Three kinds of promoter-RBS libraries, i.e., a
constitutive promoter-RBS library, a repressor-regulated promoter-RBS library and an
activator-regulated promoter-RBS library, are constructed for systematic gene circuit
design using the identified kinetic strengths of their promoter-RBS components.
According to the dynamic model of genetic transistors, a design methodology for genetic
transistors via a genetic algorithm-based searching algorithm is developed to search for a
set of promoter-RBS components and adequate concentrations of inducers to achieve the
prescribed I/O characteristics of a genetic transistor. Furthermore, according to design
specifications for different types of genetic transistors, a look-up table is built for genetic
transistor design, from which we could easily select an adequate set of promoter-RBS
components and adequate concentrations of external inducers for a specific genetic
transistor. This systematic design method will reduce the time spent using trial-and-error
methods in the experimental procedure for a genetic transistor with a desired I/O
characteristic. We demonstrate the applicability of the proposed design methodology to
genetic transistors that have desirable linear amplification or switching by employing
promoter-RBS library searching.
3.1. INTRODUCTION
Synthetic biology aims to perform various specific functions in organisms by inserting a
designed gene network. In the past, synthetic biology could be classified as having two broad
purposes. The first was to create artificial life from natural biology using the synthetic
methods. The other was to assemble some functional components using interchangeable
natural components which are nonexistent in natural biology (Benner and Sismour 2005). A
lot of the recent literature focuses on performing electronic circuit behaviors in organisms
using genetic devices such as toggle switches (Gardner, Cantor, and Collins 2000, Kobayashi
et al., 2004, Hasty, McMillen, and Collins 2002), oscillators (Elowitz and Leibler 2000,
Stricker et al., 2008, Danino et al., 2010, Mondragón-Palomino et al., 2011, Shetty, Endy, and
Knight 2008, Mutalik et al., 2013, Lou et al., 2012), pulse generators (Chang, Lin, and
Jennawasin 2013, Wang and Chen 2005), logic gates (Wang, Jing, and Chen 2005, Wang,
Chen, and Aihara 2006, Basu et al., 2004, Voigt 2006), and filters (Anderson, Voigt, and
Arkin 2007, Tamsir, Tabor, and Voigt 2011, Wang et al., 2011). Synthetic biologists also
design various types of genetic circuits with different functionalities by employing genetic
devices to solve useful tasks, such as biosensor decisions or edge detection (Khalil and
Collins 2010, Sohka et al., 2009).
In many industries, such as the electronics and manufacturing industries, the
characterization and standardization of components and the institution of specifications are
already key elements in the production line. In the synthetic biology, the expression of a
specific protein needs a promoter, a ribosome binding side (RBS), a protein coding sequence
and a terminator, which are DNA fragments. The Registry of Biological Standard Parts
(http://www.partsregistry.org), formed by MIT, shows many standard BioBricks, and these
standard biological components provide synthetic biologists with a quick and standardized
way of constructing gene circuits. Also, BioFAB provides some sorts of biobricks (see
http://biofab.org/data), enabling the rapid design and prototyping of genetic constructs.
However, in the past, the strength of a promoter and an RBS, which are the main components
of transcription and translation, were defined according to their relative strength with other
promoters and RBSs. Now, however, the promoter-RBS strength can be quantified by
measuring the fluorescence of proteins whose coding gene is constructed at the downstream
of the promoter-RBS component.
Based on the kinetic strengths of promoter-RBS components, promoter-RBS libraries are
constructed for gene circuit design. In our gene circuit design, promoter-RBS libraries are
built based on kinetic parameters of the dynamic gene regulation, which are identified by the
nonlinear least squares method based on experimental data. We formulate the design
specifications of desired gene circuits in advance and choose an adequate set of promoter-
RBS components from the promoter-RBS libraries based on the characterized, standardized
and quantified components. For our promoter-RBS libraries, we select three kinds of
promoters, i.e., constitutive promoters, repressor-regulated promoters and activator-regulated
promoters, to combine with RBSs as the promoter-RBS components. Each of these is
constructed using the green fluorescent protein in Escherichia coli. and characterized to allow
for the construction of the following three types of promoter-RBS libraries: constitutive
promoter-RBS libraries, repressor-regulated promoter-RBS libraries and activator-regulated
promoter-RBS libraries.
To date, several researchers have demonstrated that synthetic gene circuits have the
functionality of amplification or switching (Kampf et al., 2012, Basu et al., 2005, Tabor et al.,
, Tabor, Levskaya, and Voigt 2011, Callura, Cantor, and Collins 2012). These gene circuits
can amplify the input signal or switch the output signal as it exceeds a specific threshold
level. Often shown in these genetic amplifiers is the use of a two-stage cascade of promoters
to achieve the function of amplification. However, these circuits only amplify a low level of
input signal or low concentration of inducer at the first stage while the second stage consists
of the promoter-RBS activity being fixed. On the other hand, switching circuits switch the
output signal using an external inducer. When the inducer is externally increased, the circuit
Systematic Design Methodology for Robust Genetic Transistors 39
is on, and vice versa. Nevertheless, at present, the on-state, or high level, of switching has not
been clearly defined.
In this chapter, we demonstrate that a simple repressive gene circuit can work like an
electrical transistor as an amplifier or a switch. The amplification gains or switch levels of the
genetic transistor are regulated by the concentrations of inducer and different combinations of
promoters with RBSs. For the convenience of measurement and application, reporter genes
are constructed as the measurable input and output. We show that the I/O characteristics of
the repressive gene circuit regulated by inducer concentration can be effectively predicted by
adequate selection of promoter-RBS components from our libraries. Thus, based on these
promoter-RBS libraries, a look-up table is built to quickly select adequate promoter-RBS
components for the design of genetic transistors with different design specifications.
In the following sections, we first construct the promoter-RBS libraries based on the
promoter-RBS strength through the dynamic regulatory model of promoter-RBS components.
Then, we describe the I/O characteristics of a genetic transistor with different kinetic
strengths of promoter-RBS components in the promoter-RBS libraries and different
concentrations of inducers. Finally, a look-up table (or genetic transistor library) is
constructed for genetic transistor design requiring prescribed I/O characteristics, which is
used by searching the most appropriate sets of promoter-RBS components and concentrations
of inducers via the genetic algorithm (GA).
3.2. CONSTRUCTION OF THE PROMOTER-RBS LIBRARIES

FOR GENETIC TRANSISTORS
Here, we introduce the characterization and standardization of promoter-RBS libraries

and employ a dynamic mathematical model to construct the promoter-RBS libraries
according to the identified kinetic strengths of promoter-RBS components, populated via
experimental data.
Promoter-RBS Libraries Based on the Identified Kinetic Strengths

of Promoter-RBS Components
In a systematic design procedure, the characterization and standardization of components

are important preparatory tasks before practical design process. These can save designers a
significant amount of time and avoid unnecessary trial-and-error attempts. In the field of
synthetic biology, a particular technique is developed to create standard interchangeable
biological components called BioBricks (Gianna De and Davies 2003, Nistala et al., 2010).
These allow the synthetic biologists to focus on the design of more complex genetic circuits
rather than the basic construction of the gene components.
BioBricks are DNA fragments with specific functions, and include promoters, ribosome
binding sites (RBS), repressors, activators, reporters and terminators. In the database, there
are only a few BioBrick components that are well-characterized. The well-characterized
BioBrick components are conducive to the systematic design of synthetic genetic circuits. In
order to facilitate the design of synthetic gene circuits, wider libraries of well-characterized
BioBricks need to be constructed.
In our promoter-RBS libraries, the library indexes are the kinetic strengths of promoter
and RBS, which are considered together as a promoter-RBS component because the gene
expression is regulated by a promoter-RBS component. The kinetic strength of a promoter-
RBS component can be systematically identified by a stochastic model which simulates the
dynamic behavior of promoter-RBS components under some external molecular or
environmental noises. In order to identify the kinetic strength of a promoter-RBS component,
the green fluorescence gene is embedded into the downstream of the promoter-RBS
component. By measuring the fluorescence dynamic time profile and using the nonlinear least
squares method (Martineau, Stout, and Towe 2010), we identify the kinetic strengths of the
promoter-RBS components to be used as the indexes of promoter-RBS libraries.
The construction procedure of the promoter-RBS libraries can be generally divided into
four steps (Deans, Cantor, and Collins): (i) choose the required promoter-RBS components,
(ii) select the suitable reporter protein and growth conditions, (iii) measure the time-profile
data of the dynamic behavior, and (iv) construct the dynamic regulatory model for identifying
kinetic strengths of promoter-RBS components to be used as library indexes according to the
nonlinear least squares method. In the first step, some promoters can be regulated by specific
transcription factors, and different combinations of promoters with RBSs give different
kinetic strengths of promoter-RBS components, which increase the diversity of the libraries.
In order to rapidly obtain a variety of kinetic strengths of promoter-RBS components, a
mutation technique was used to create different kinetic strengths of promoters and RBSs to
increase the varieties of promoter-RBS components through the mutation of a specific region
on promoters or RBSs (Basu et al., 2004, Deans, Cantor, and Collins). In the second and third
step, since different reporter proteins, such as the green fluorescent protein or red fluorescent
protein have different degradation rates, the measurement times may differ. Further, the cell
growth conditions have an effect on the results of the measurement. Biological component
can be characterized at different cellular growth phases, under different culture conditions, or
at different resolutions. In our experiment, the GFP is selected as the reporter protein and the
time profiles of fluorescence are measured by the microplate reader. In the final step, a
mathematical dynamic model is built to describe the time profile of protein expression. Using
the protein expression time profile measurements, the nonlinear least squares method is
employed to identify the kinetic strengths of promoter-RBS components to be used as the
library indexes with the mathematical model. For the systematic design of genetic transistors,
we construct three kinds of promoter-RBS libraries, i.e., constitutive, repressor-regulated and
activator-regulated promoter-RBS libraries. The promoter-RBS components in promoter-RBS
libraries and all BioBrick components used in this study are listed in Tables 2.1, 2.2 and 2.3,
respectively. The detailed construction procedures of constitutive, repressor-regulated and
activator-regulated promoter-RBS libraries are described in Chapter 2.
3.3. CONSTRUCTION AND DESIGN OF GENETIC TRANSISTOR

After the introduction of regulatory functions of promoter-RBS components and the
construction of the promoter-RBS libraries, we design a synthetic genetic circuit, similar to a
transistor, with prescribed I/O characteristics of amplification or switching through the

external inducer.
Input device Ouput device
input signal =
= output signal
(a)
Output signal measure device
Input signal measure device
(b)
Figure 3.1. The representation of synthetic genetic transistor circuit. (a) A genetic transistor. (b) A
genetic transistor with measurement circuit. The input signal of the genetic transistor is measured by
RFP reporter and the output signal of the genetic transistor is measured by GFP reporter.
Construction of Genetic Transistor
A genetic transistor is shown in Figure 3.1(a). The transistor is constructed to obtain the
output protein concentration xprotein of the transistor for amplification or switching behavior.
The genetic transistor consists of the repressor-regulated promoter-RBS component c3 and a
repressor coding gene. The input repressor xrepressor2 to the genetic transistor is controlled by
the repressor-regulated promoter-RBS component c2, which is regulated by the corresponding
repressor. The input repressor xrepressor2 will form the complex and restrict the production of
output protein xprotein by binding the corresponding repressor-regulated promoter-RBS
component c3 to decrease its kinetic strength. However, when the inducer is added, this
inducer will bind input repressor xrepressor2 and prevent it from binding to the repressor-
regulated promoter-RBS component c3. Then, both the kinetic strength of repressor-regulated
promoter-RBS component c3 and the production of output protein xprotein will increase. The
dynamic model of a genetic transistor is described as follows:
 
xrepressor 2  c2 , t   prepressor PM ,c2 , Pm ,c2 ,0,0      repressor 2  xrepressor 2  c2 , t 
(3.1)
 
x protein  c3 , t   prepressor PM ,c3 , Pm,c3 , xrepressor 2 , I 2      protein  x protein  c3 , t 
where xrepressor2 and xprotein denote the concentrations of input repressor2 and the output protein
of the genetic transistor, respectively, and γprotein denotes the degradation rate of the protein.
However, the protein concentration is difficult to directly measure and quantify. To
determine characteristics of the synthetic genetic transistor, a genetic transistor with
measurement circuit is constructed as shown in Figure 3.1(b). In Figure 3.1(b), we construct
an additional repressor-regulated promoter-RBS component c2 so that the input reporter
protein x1 can be measured by input fluorescence g1 and the output reporter protein x2 can be
measured by output fluorescence g2. Note that RFP is used to measure input while GFP is
used to measure output. Additionally, for the convenience of the input regulation, we
construct an input signal generation device with the concentration of inducer I1 to control the
input g1 of the genetic transistor circuit. Then, the dynamic model of a synthetic genetic
transistor circuit with I/O measure devices under environmental disturbances is described by
the following set of equations:

 
 xrepressor1  c1 , t   pconst Pc      repressor1  xrepressor1  c1 , t   v1 (t )
1
x
 repressor 2 2
 
 c , t   prepressor PM ,c2 , Pm,c2 , xrepressor1 , I1      repressor 2  xrepressor 2  c2 , t   v2 (t )
x c ,t   p
 1 2    
repressor PM , c2 , Pm , c2 , xrepressor1 , I1  m1     im , x1 x1  c2 , t   v3 (t )

 
 g1 (c2 , t )  m1  x1  c2 , t      m, x1 g1 (c2 , t )  v4 (t )

   
 x2  c3 , t   prepressor PM ,c3 , Pm,c3 , xrepressor 2 , I 2  m2     im, x2 x2  c3 , t   v5 (t )

 
 g 2 (c3 , t )  m2  x2  c3 , t      m, x2 g 2  c3 , t   v6 (t )

, c1  Libconst , c2 and c3  Librepressor
(3.2)
where m1 and m2 denote the maturation rates of reporter 1 x1 and reporter 2 x2, respectively,
and vi(t), i = 1,2,…,6 denote the noises.
To explore the I/O characteristics of a synthetic genetic transistor with the function of
amplification or switching, the steady state model of (3.2) is given by

 
 xrepressor1  c1   pconst Pc     repressor1   vs
1 1
x  
 c , I   prepressor PM ,c2 , Pm,c2 , xrepressor1 , I1     repressor 2   vs2
 repressor 2 2 1
x c , I   p
 1 2 1 
repressor PM , c2 , Pm , c2 , xrepressor 1 , I1  
m1     im , x1  vs3


 g1 (c2 , I1 )  m1  x1  c2     m, x1  vs4  (3.3)


 x2  c3 , I1 , I 2   prepressor PM ,c3 , Pm ,c3 , xrepressor 2 , I 2  
m2     im , x2  vs5


 g 2  c3 , I1 , I 2   m2  x2  c3     m, x2  vs6
 
, c1  Libconst , c2 and c3  Librepressor
From (3.3), if m1≈m2, γm,x1≈γm,x2 and γim,x1≈γim,x2, then the I/O characteristic can be
regarded as input/output = xrepressor2/xprotein≈x1/x2≈g1/g2, i.e., we could use the x1/x2 or g1/g2
ratio to replace the I/O characteristic of the synthetic genetic transistor. Further, the I/O
characteristic can be controlled and regulated by the selection of promoter-RBS components
c3 and inducer concentration I2. Therefore, we need to define the I/O characteristic of
synthetic genetic transistor circuits to design a genetic transistor with the desired I/O
characteristic. This is done as follows:
yss (c3 , I 2 , g1 (c2 , I1 ))  g2 (c3 , I1 , I 2 ) , g1   g1e , g1n  (a.u.) (3.4)
where yss(c3,I2,g1) denotes the I/O response of the synthetic genetic transistor circuit between
input signal g1 and output signal g2, and g1e and g1n denote the lower bound and upper bound
of g1. a.u. stands for arbitrary unit.
In general, genetic components are inherently uncertain in the biological system as a
result of gene expression noises in transcription or translation processes. Hence, we model the
uncertain kinetic strengths of promoter-RBS components, degradation rate of proteins and
transcription/ translation rates as stochastic processes in the following model:
Pc1  Pc1  Pc1 n1  t  , PM ,c2  PM ,c2  PM ,c2 n2  t  , Pm,c2  Pm,c2  Pm,c2 n2  t  ,
PM ,c3  PM ,c3  PM ,c3 n3  t  , Pm,c3  Pm,c3  Pm,c3 n3  t  ,
 repressor1   repressor1   repressor1n1  t  ,  repressor 2   repressor 2   repressor 2 n2  t  ,
(3.5)
 im, x   im, x   im, x n2  t  ,  m, x   m, x   m, x n2  t  ,
1 1 1 1 1 1
 im, x   im, x   im, x n3  t  ,  m, x   m, x   m, x n3  t  , m1  m1  m1n2  t  ,

2 2 2 2 2 2
m2  m2  m2 n3  t  ,      n1  t 
where ∆Pc1, ∆PM,c2, ∆Pm,c2, ∆PM,c3, ∆Pm,c3, ∆γrepressor1, ∆γrepressor2, ∆γim,x1, ∆γm,x1, ∆γim,x2, ∆γm,x2,
∆m1, ∆m2 and ∆μ denote the standard deviations of stochastic parameters to be tolerated and
could be specified before design and ni(t), i = 1,2,3 denote Gaussian noises with zero mean
and unit variance. Therefore, ∆Pc1, ∆PM,c2, ∆Pm,c2, ∆PM,c3, ∆Pm,c3, ∆γrepressor1, ∆γrepressor2,
∆γim,x1, ∆γm,x1, ∆γim,x2, ∆γm,x2, ∆m1, ∆m2 and ∆μ denote the deterministic parts of parameter
variations and ni(t), i = 1,2,3 denote different random fluctuation sources. For robust design of
the genetic transistor circuit, these parameter fluctuations in (3.5) will henceforth be
considered in the design procedure so that the synthetic genetic transistor can tolerate these
kinds of parameter fluctuations in vivo.
With fixed concentration of inducer I2, we expect that the input signal g1/ output signal g2
(I/O) characteristics of the synthetic genetic transistor in (3.4) would be similar to the voltage
I/O characteristics of the electronic transistor shown in Figure 3.5(b) in Appendix. When the
inducer concentration I1 increases, the kinetic strength of promoter-RBS component c2
increases along with the fluorescence of the input signal g1, which means that the repressor
concentration xrepressor2 increases. Due to the fixed concentration of inducer I2, the redundant
repressors xrepressor2, which are not bound by the inducer I2, will repress the promoter-RBS
component c3, and the fluorescence of output signal g2 will decrease. Therefore, the I/O
characteristic of the synthetic genetic transistor is similar to Figure 3.5(b). Additionally, from
Figure 3.5(b), we see that if input signal is in the operation range of linear amplification, the
input signal would be inversely amplified.
Now, consider the alternative viewpoint, i.e., the voltage I/O characteristics of an
electronic transistor. When R2/R1 increases, the reverse amplification gain will become large
and the operation region of linear amplification will narrow as shown in (3.17)-(3.19) and
Figure 3.6. In the synthetic genetic transistor, we expect that when the concentration of
inducer I2 changes as per the R2/R1 ratio in (3.18)-(3.19), the I/O characteristics would be
similar to the voltage I/O characteristics of electronic transistor in Figure 3.6. Due to different
concentrations of inducer I2, the effect of the inducer on the input repressor can vary. When
the inducer concentration I2 decreases, the I/O characteristics would sharpen, so the reverse
amplification gain becomes large in the operation region of linear amplification.
Finally, when R2/R1 is large enough in (3.18)-(3.19), the operation region of linear
amplification will become too narrow and result in a sharp change in this region.
Correspondingly with a synthetic genetic transistor, when the inducer concentration I2 is low
enough, the input signal g1 will produce a small variation, and the output signal g2 will have
an acute change like a switch. Therefore, according to the analysis above, we could obtain
varying reverse amplification gains and switch levels by changing the concentration of
inducer I2.
Systematic Design of a Genetic Transistor Based on Design Specification
According to the above analysis in Figure 3.1(b), we can obtain different reverse
amplification gains or switch behaviors via regulation of different concentrations of inducer
I2. Additionally, due to the output signal g2 being under the controlled by promoter-RBS
component c3, we could change the output range by selecting different repressor-regulated
promoter-RBS components c3 from the repressor-regulated promoter-RBS libraries. In this
way, we can control the I/O characteristics of a synthetic genetic transistor to obtain different
reverse amplification gains or switch levels by choosing different concentrations of inducer I2
and selecting different repressor-regulated promoter-RBS components c3 from the repressor-
regulated promoter-RBS libraries.
In Figure 3.1(b), the input signal generation device consists of a constitutive promoter-
RBS component c1, and a repressor-regulated promoter-RBS component c2 and an inducer I1.
The constitutive promoter-RBS component c1 is selected to produce the input repressor
continually. For convenience of design, the repressor-regulated promoter-RBS component c2

is selected from the corresponding promoter-RBS library to have sufficient kinetic strength to
obtain an adequate maximum regulation range of input signal regulated by inducer I1.
However, the operation region of linear amplification is still limited in the amplifier design of
genetic transistor, and the input signal range might not be fully contained in the operation
region of linear amplification. Therefore, the input signal range should be considered in
relation to the design purpose. In the procedure of amplifier design, the input operation range
g1 [g1,l, g1,u] can be set by transforming the inducer concentration I1 into the input
fluorescence according to (3.2) or (3.3) as follows
Input operation range: I1 [I1,l , I1,u ]  g1 [g1,l , g1,u ] (a.u.) (3.6)
where I1 and g1 denote the inducer concentration and input fluorescence, respectively, I1,l and
I1,u denote the lower and upper bound of inducer concentrations, respectively, and g1,l and g1,u
denote the lower and upper bound of input fluorescences respectively.
Note that, in the future, when the promoter-RBS libraries are large enough, the promoter-
RBS components c1 and c2 can be designed and selected to match the input operation range.
However, due to the limited size of our promoter-RBS libraries and for the convenience of
design, we will select the repressor-regulated promoter-RBS component c2 from the
corresponding promoter-RBS library.
From the above analysis, the design purpose of an amplifier will lead to the selection of a
suitable repressor-regulated promoter-RBS component c3 from the repressor-regulated
promoter-RBS libraries and concentration of inducer I2, i.e., {c3, I2}, so that the I/O
characteristics of the synthetic genetic transistor in (3.4) in a specific input range g1 [g1,l,
g1,u] can match the desired I/O response similar to (3.18), i.e.,
yd  g1   gain  ( g1  g1,l )  g2,u , g1   g1l , g1u  (a.u.) (3.7)
where g1,l and g2,l denote the lower bound of input fluorescence g1 and upper bound of output
fluorescence g2, respectively, and gain denotes the amplification gain of the genetic transistor.
On the other hand, the switching behavior will occur when the input signal has a small
variation (see Figure 3.6), i.e., a high level signal can be switched into a low level signal and
vice versa. In the switching behavior of synthetic genetic transistor, each promoter-RBS
component has its own basal level. Thus, when the input signal increases, the output signal
will rapidly decrease to the basal level. Therefore, the desired I/O response of a switch is
described as follows:
 H s  Ls 
yd  g1   Ls  , g1   g1l , g1u  (a.u.) (3.8)
1   g1 gt 
2
where Hs and Ls denote the high level and low level of switching, respectively, and gt denotes
the transition point of input fluorescence. Moreover, the input signal range of I/O
characteristics of the switch can be set by (3.6).
Finally, for matching the desired I/O response of an amplifier or switch, the genetic
algorithm (GA) is employed to select an adequate repressor-regulated promoter-RBS
component c3 in the repressor-regulated promoter-RBS libraries and the concentration of
inducer I2 to minimize the following cost function (Wu, Lee, and Chen 2011),
respectively, i.e.,
min J  c3 , I 2 
c3 Librepressor , I 2 [ I 2,l , I 2,u ]
g1,u (3.9)
 min E ( yss (c3 , I 2 , g1 )  yd ( g1 ))2 dg1
c3 Librepressor , I 2 [ I 2,l , I 2,u ] g1,l
To summarize the above design procedure of a biological amplifier and switch, a genetic
transistor design procedure of by the promoter-RBS library searching method using GA is
proposed as follows (Wu, Lee, and Chen 2011):
1. Construct the genetic transistor circuit such as in Figure 3.1(a).

2. Build the dynamic and steady state mathematical model in (3.2) and (3.3),
respectively.
3. Provide the design specification of amplifier with the desired I/O response as in (3.6)
and (3.7) or switch with the desired I/O response as in (3.6) and (3.8).
4. Provide the standard deviations of parameter fluctuations and environmental
disturbances to be tolerated in vivo in (3.5).
5. Minimize the cost function J(c3,I2) in (3.9) by selecting an optimal set via GA.
Based on the design procedure of a genetic transistor using the promoter-RBS library
searching method with GA, the promoter-RBS component c3 is selected from the
corresponding repressor-regulated promoter-RBS library and the inducer concentration I2 is
selected within [I2,l,I2,u], while the cost function is calculated in each iteration of the selection
process. Then, GA would select the most adequate promoter-RBS component c3 from the
corresponding repressor-regulated promoter-RBS library and inducer concentration I2
[I2,l,I2,u] to minimize the cost function.
3.4. SYNTHETIC GENETIC TRANSISTOR DESIGN EXAMPLES

BASED ON PROMOTER-RBS LIBRARIES
We have presented the construction and design procedure of a synthetic genetic
transistor. In this section, the synthetic genetic transistor is designed and simulated to verify
the I/O characteristics of amplification and switching. Subsequently, based on our promoter-
RBS libraries, the amplification gain in a specific input operation range and switching level
are designed by employing GA to select the most adequate promoter-RBS components and
inducer concentrations. Finally, to support future application of this method, a look-up table
for genetic transistors is built for different genetic transistor design specifications.
Amplifier Design Example of Synthetic Genetic Transistor
Consider the amplifier design of the synthetic genetic transistor. Firstly, to obtain the I/O
characteristics of amplifier, promoter-RBS components {c1,c2} = {J6,L3} are selected to
obtain the maximum input operation range. The dynamic model and the steady state model
have been described in (3.2) and (3.3). The input operation range and desired I/O response of
genetic transistor are specified as follows:
Input range: I1 [1.5*102 ,4*102 ] (mM)  g1 [298,431] (a.u.) (3.10)
and
yd  g1   2  g1  1286 (3.11)
where -2 is the desired amplification gain as shown in Figure 3.2. Note that the standard
deviations of parameter fluctuations that are supposed to be tolerated in vivo are given by
1000
900
800
Output Fluorescence (a.u.)
700
600
500
400
300
200
100
0
340 360 380 400 420 440 460
Input Fluorescence (a.u.)
Figure 3.2. The amplifier design example of synthetic genetic transistor. For amplifier design example
of synthetic genetic transistor, the prescribed amplification gain = -2 within the range g1 [298,431] is
computed by (3.11) (red line). The most adequate promoter-RBS component c3 and aTc concentration
IaTc to fit the prescribed amplification gain are searched as {c3,IaTc} = {T3,0 ng/ml} by minimizing
JA(c3,IaTc) in (3.13) from the corresponding promoter-RBS library LibTetR and concentration range of
inducer IaTc. The green points are the experimental results based on {c3,IaTc}, and the error bars are the
standard deviations. The green line is the estimation of I/O response of the synthetic genetic transistor
based on experimental data, with the estimated amplification gain = -1.978. Obviously, the
amplification gain of the designed genetic transistor could match the desired amplification gain quite
well.
  
Pc1  0.05Pc1 , PM ,ci , Pm,ci  0.05PM ,ci ,0.05Pm ,ci , i  2,3 
 LacI  0.05 LacI ,  TetR  0.05 TetR
 im, x1  0.05 im, x1 ,  m, x1  0.05 m, x1 (3.12)
 im, x2  0.05 im, x2 ,  m, x2  0.05 m, x2
m1  0.05m1 , m2  0.05m2 ,   0.05
and the environmental disturbances vi(t) are independent Gaussian noises with zero mean and
unit variance. Finally, GA is employed to search a set {c3,IaTc} from corresponding libraries
to minimize the following cost function:
min J A  c3 , I aTc 
c3 LibTet , I 2 [1.5*102 ,4*102 ]
431 (3.13)
 min E ( yss (c3 , I aTc , g1 )  yd ( g1 ))2 dg1
c3 LibTet , I 2 [1.5*102 ,4*102 ] 298
Then, the most adequate promoter-RBS component from the corresponding library and
aTc concentration are found to be {c3,IaTc} = {T3,0 ng/ml}. The estimation of I/O response of
genetic transistor based on experimental results is shown in Figure 3.2, with experimental
details summarized in Appendix. The I/O characteristics of genetic transistor can match the
desired I/O response in a workable input range g1 [298,431] under the intrinsic fluctuations
and environmental disturbances.
Switch Design Example of Synthetic Genetic Transistor
Consider the switch design of the synthetic genetic transistor. The switch design
procedure is similar to the amplifier design procedure of a synthetic genetic transistor. Firstly,
to obtain the complete I/O characteristics of switching, promoter-RBS components {c1,c2} =
{J6,L3} are selected to obtain the maximum input operation range. The dynamic model and
the steady state model have been described in (3.2) and (3.3), respectively. The input
operation range and desired I/O switch response are specified as follows:
Input range: I1 [2*103 ,10] (mM)  g1 [103,614] (a.u.) (3.14)
and
 3249.7  Ls 
yd  g1   Ls  (3.15)
1   g1 150.4 
2
where Ls denotes the low level of switching or basal level of promoter-RBS component c3.
Note that the standard deviations of parameter fluctuations that are supposed to be tolerated in
vivo and from environmental disturbances are the same as in (3.12). Finally, GA is employed
to search a set {c3,IaTc} from corresponding libraries to minimize the following cost function:
min J S  c3 , I aTc 
c3 LibTet , I 2 [2*103 ,10]
614 (3.16)
 min 3
E ( yss (c3 , I aTc , g1 )  yd ( g1 ))2 dg1
c3 LibTet , I 2 [2*10 ,10] 103
2500
2000
Output Fluorescence (a.u.)
1500
1000
500
0
100 150 200 250 300 350 400 450 500 550 600 650
Input Fluorescence (a.u.)
Figure 3.3. The switch design example of synthetic genetic transistor. For switch design example of
synthetic genetic transistor, a desired I/O switch response is computed by (3.15) as shown in red line.
The most adequate promoter-RBS component c3 and aTc concentration IaTc to fit the desired I/O switch
response are searched as {c3,IaTc} = {T3,0 ng/ml} by minimizing JS(c3,IaTc) in (3.16) from the
corresponding promoter-RBS library LibTet and concentration range of inducer IaTc. The green points are
the experimental results, and the error bars are the standard deviations. The green line is the estimated
I/O switch response based on experimental data. Obviously, the I/O response of the designed genetic
switch could match the desired I/O switch response quite well.
Then, the most adequate promoter-RBS component from the corresponding library and
aTc concentrations are found to be {c3,IaTc} = {T3,0 ng/ml}. The estimation of I/O response of
genetic transistor based on experimental results is shown in Figure 3.3, with experimental
details summarized in Appendix. Clearly, the switching I/O characteristics of synthetic
genetic transistor can match the desired I/O response under the intrinsic fluctuations and
According to the above examples, the amplification or switching I/O characteristics of a
synthetic genetic transistor with different design specifications can be achieved by selecting
the most adequate promoter-RBS component c3 and inducer concentration using the proposed
library-based searching method. However, not just the promoter-RBS component c3 LibTet
can be selected to achieve the amplification or switching design specification of the synthetic
genetic transistor circuit, but also other promoter-RBS components, i.e., LibLac, can be
selected to achieve the desired I/O response. However, for various design specifications, more
promoter-RBS libraries are needed to achieve these design specifications.
Table 3.1. The look-up table with different gain specifications for synthetic genetic
transistors. Given the desired amplification gains and their input signal ranges, we
could select adequate promoter-RBS component and inducer concentration from the
table to achieve the minimum matching error in (3.9)
Amplifier gain specifications of synthetic genetic transistor

Input range (a.u.) Gain Promoter-RBS component Inducer concentration
120~180 -10.00 T3 0 ng/ml
150~225 -7.50 T3 1 ng/ml
LibTet
260~460 -2.00 T3 0 ng/ml

400~550 -1.00 T3 1 ng/ml
460~560 -0.75 T3 0 ng/ml
575~620 -0.50 T3 1 ng/ml
LibLac
40~140 -0.15 L1 0 mM
40~140 -2.50 L3 0 mM
For the convenience of synthetic genetic transistor design for synthetic biologists, one
look-up table has been built for the various design specifications as shown in Table 3.1 via
selecting adequate promoter–RBS components from the corresponding libraries and adequate
inducer concentration to achieve the optimal matching in (3.9). Based on various
amplification gains in some specific operation range, the synthetic genetic transistors can be
designed by first checking the look-up table. In future, more promoter-RBS components and
inducer concentrations for different I/O characteristics of synthetic genetic transistors can be
accumulated to build much larger look-up tables to match a lot of design specifications. From
this look-up table, based on the desired design specifications, we can select the adequate
promoter-RBS components and inducer concentrations to synthesize the genetic transistors
with desired I/O responses. Thus, less time will be spent on the design procedure as a
designer will be able to easily construct transistors with the desired I/O characteristics.
3.5. DISCUSSION
One major aim of synthetic biology is to construct a gene circuit with the desired
functionality of an organism. Recently, promoter libraries and promoter-RBS libraries have
been built to simulate the in vivo behavior of a gene circuit (Gianna De and Davies 2003,
Johansson 1993, Mondragón-Palomino et al., 2011). By identifying the kinetic strengths of
promoter-RBS components, the protein expressions in the gene circuit can be estimated and
predicted. However, in the process of constructing promoter-RBS library, the identified
kinetic parameters in the promoter-RBS library can be affected by several conditions,
including the medium, copy number of plasmid, terminator and so on. Therefore, for the
extensive application of promoter-RBS libraries, the construction conditions of promoter-
RBS libraries need to be unified and standardized. This will allow standardized promoter-
RBS libraries, similar to electronic component libraries, which can be easily used and
expanded by other gene circuit designers.
In this chapter, by the promoter-RBS libraries we established, a genetic transistor has
been constructed and implemented in silico. Additionally, the synthetic genetic transistor can
perform amplification and switching like an electronic transistor according to its I/O
characteristics. The I/O characteristics of the synthetic genetic transistor circuit are simulated
by a mathematic model with random parameter fluctuation to guarantee the robustness of the
design in vivo. The design specification of amplification or switching in the genetic transistor
can be achieved by the library-searching method using GA. By optimally matching the
desired I/O response of amplification or switching, the most adequate set of promoter-RBS
component and inducer concentration {c3,IaTc} can be selected to construct a genetic transistor
with the desired design specifications. The library-searching method using GA is introduced
to reduce the number of trial-and-error attempts, as well as the searching time in libraries
when the libraries have a large number of components. Furthermore, for the convenience of
synthetic genetic transistor design for synthetic biologists, one look-up table has been built
for the various design specifications as shown in Table 3.1. From this look-up table, based on
the desired design specifications, we can select the adequate promoter-RBS components and
inducer concentrations to synthesize the genetic transistors with desired I/O responses. Thus,
less time will be spent on the design procedure as a designer will be able to easily construct
transistors with the desired I/O characteristics.
For applications of the genetic transistor, the various biological components need to be
characterized and standardized. By using characterized and standardized genetic components,
the design specification of a genetic transistor can be set and the look-up tables can be used to
support the genetic circuit design. The genetic transistor described here has a number of
potential applications. The amplifier can be used to amplify the oscillation signal reversely
and linearly. Based on the designed oscillatory genetic circuits (Stricker et al., 2008, Tuttle
et al., , Purcell et al., 2011, Rey et al., 2005, Kelly et al., 2009, Mutalik et al., 2013, Lou et al.,
2012) in oscillatory metabolic pathways (Di Capite, Ng, and Parekh , Taylor et al., 2008, Lin,
Liu, and Chuang 2009), an adequate genetic transistor selected from the look-up tables
according to the oscillation range and desired amplification gain can be inserted into these
circuits directly to amplify the oscillatory signal. In this way, the original genetic circuits do
not need to be redesigned. On the other hand, the switch can be used to detect some signals
and act like a detector or biosensor (Kampf et al., 2012, Callura, Cantor, and Collins 2012,
Canton, Labno, and Endy 2008, Cuero, Lilly, and McKay 2012). When the input signal
changes, the output signal will switch to the other state and make the downstream circuit
respond to the signal change. Also, the switch of a genetic transistor can work as logic gates
as in an electronic transistor (Wang, Jing, and Chen 2005, Wang, Chen, and Aihara 2006,
Basu et al., 2004). With different combinations of genetic transistors, different logic gates can
be constructed.
3.6. CONCLUSION
In this chapter, three kinds of libraries, i.e., a constitutive promoter-RBS library,
repressor-regulated promoter-RBS library and activator-regulated promoter-RBS library,
were established for constructing synthetic gene circuits with the desired transistor
amplification or switching function. The amplification gain and switching level of a genetic
transistor could be calibrated by selecting adequate promoter-RBS components and inducer
concentrations from the corresponding libraries. For the measurement of I/O response, we
could embed an additional repressor-regulated promoter-RBS component with reporter
protein at the input terminal to measure the input signal while replacing the output protein
with a reporter protein to measure the output signal. Further, for the convenience of input
regulation, an external circuit was constructed to control the input signal using the
concentration of inducer. Based on the desired I/O response in relation to amplification or
switching of a genetic transistor, the GA-based searching algorithm was introduced to search
for the most appropriate set of promoter-RBS components and inducer concentration from the
corresponding promoter-RBS libraries to achieve prescribed I/O characteristics in the genetic
transistor. In the simulation results for this study, we demonstrated that the genetic transistor
designed here has the prescribed function of amplification or switching. By the library-
searching method using GA, different design specifications of amplifier or switch could be
achieved the most appropriate set of promoter-RBS components from the corresponding
promoter-RBS libraries and inducer concentration within a feasible region. Finally, a look-up
table was built for genetic transistor design with different genetic transistor design
specifications. Using this table, we could easily select an adequate set of promoter-RBS
components and inducer concentration to construct the desired genetic transistor. This
innovation saves much time in trial and error attempts in the iterative experimental procedure.
3.7. APPENDIX
The Operation of an Electronic Transistor
A transistor is a semiconductor device used for different functions in electronic circuits,

such as signal amplification and switching. It has the advantages of low volume, high
efficiency, a long life-span and high speed switching. Two major types of transistor are the
bipolar junction transistor (BJT) and metal-oxide-semiconductor field-effect transistor
(MOSFET). The functions of these two types of transistors are similar. The electronic signal
can be amplified linearly or switched by controlling the bias voltage. Considering the BJT for
example, a three-terminal npn BJT has three terminal connections which are called the
emitter, base and collector, as shown in Figure 3.4(a). The complete npn BJT circuit in the
common-emitter configuration is shown in Figure 3.5(b), in which the electronic signal from
base terminal iB can be amplified linearly or switched through BJT by controlling the voltage
between the collector and emitter terminals VCE in the linear region larger than 0.3V or
saturation region smaller than 0.3V. When VCE is controlled in the linear region, iB is
amplified linearly. The signal output from collector terminal iC is almost equal to β∙iB and the
variation in input voltage Vin is also amplified reversely and linearly, where the parameter β is
the common-emitter current gain within the range of 50 to 300. On the other hand, when VCE
is controlled in the saturation region, the output signal will be switched such that a high level
signal will be switched into a low level signal and vice versa. The mathematical models to
describe the behaviors of an npn BJT circuit in the common-emitter configuration are shown
as follows.
Collector  C 
ic
R2
Base  B Vo
R1
VCE+ VCC
Emitter  E 
iB VBE
+ -
Vin
+ -
-
(a) (b)
Figure 3.4. The npn bipolar junction transistor: (a) The npn BJT circuit symbol. BJT contains three
terminal connections, i.e., the so- called Base, Collector and Emitter. (b) The npn BJT circuit in the
common-emitter configuration.
In the cut-off region: Vin  VBE ( on)
Vo  VCC (3.17)
In the forward-active region (linear region): Vin  VBE ( on)
Vo   
R2
R1
Vin  VBE (on)   VCC (3.18)
In the saturation region: Vin 

R1
V  V   VBE (on)
  R2 CC CE ( sat )
Vo  VCE ( sat ) (3.19)
where VBE(on) and VCC denote the turn-on and bias voltages of the npn BJT circuit,
respectively, and VCE(sat) denotes the voltage between collector and emitter in saturation; R1
and R2 are the values of the resistance. The current-voltage and voltage I/O characteristics of
the npn BJT circuit in the common-emitter configuration are shown in Figures 3.5(a) and
3.6(b), respectively, and the voltage I/O characteristics of electronic transistor is shown in
Figure 3.6.
(a)
(b)
Figure 3.5. The current-voltage characteristic and voltage I/O characteristic of BJT in the common-
emitter configuration: The BJT circuit is shown in Figure 3.4(b), and its characteristics are simulated by
the PSpice with a standard 2N3904 transistor from PSpice library. (a) The current-voltage
characteristic: The base current iB changes from 20μA to 100μA. (b) The voltage I/O characteristic: Set
R1 = 5kΩ, R2 = 150kΩ, VCC = 5V, VBE(on) = 0.7V, VCE(sat) = 0.2V and the input voltage Vin from
0 V to 3 V. For Vin ≤0.7V, the transistor is cut off as shown; for 0.7 V Vin 1.9V , the transistor is in the
region of the linear amplification as shown in (3.18); and for Vin ≥1.9V, the transistor is in the
saturation region as shown in (3.19).
Figure 3.6. The voltage I/O characteristics of the common-emitter circuit for different R2/R1 ratio: The
circuit is shown in Figure 3.4(b), and simulated by the PSpice with a standard 2N3904 transistor from
PSpice library. The voltage I/O characteristics are simulated by changing the R2/R1 ratio. According to
(3.18), when the R2/R1 ratio becomes large, the voltage I/O characteristic is much sharper as the
amplifier in linear region. And when the R2/R1 ratio is large enough, the voltage I/O characteristic will
be like a switch.
Materials and Methods
Reagents
All restriction enzymes and DNA ligation kit were purchased from New England
Biolabs. The chemicals used in this study were ordered from Sigma-Aldrich.
Oligonucleotides were ordered from Integrated DNA Technologies.
Bacteria strains and Medium

E. coli DH5α cells from Yeastern biotech (ECOS) were used for the constructing
procedure. E. coli MG1655 ∆recA cells were used for the fluorescence measurable
experiments. M9 working medium (34 g/L Na2HPO4, 15 g/L K2HPO4, 2.5 g/L NaCl, 5g/L
NH4Cl, 10 g/L casamino acids, 2 μM vitamin B1, 2 mM MgSO4, 0.2% Glucose) with proper
antibiotics was used for E. coli cultivation at 37°C at 200 r.p.m..
Plasmid construction
The plasmids used in this study were from Biobricks and constructed by the standard
iGEM assembly and the DNA devices, including promoters, RBS, report gene, and
terminator, were listed in Table 1. All the DNA parts were constructed into the backbone
pSB3K3 (20~30 copies per cell).
Fluorescence Measurement
For fluorescence determination, cell containing the genetic circuit were inoculated in M9
working medium with 50 mg/mL Kanamycin. After 14~16 hours incubation, the culture was
diluted 500-fold and further incubated until the OD600 reached 0.1. Then different
concentrations of IPTG (Isopropyl β-D-1-thiogalactopyranoside) and aTc
(Anhydrotetracycline) were added into the medium. And the fluorescence intensity of the
culture was subsequently measured by the microplate reader (BioTek, Synergy™ H1, GFP
settings were 490/530 nm for excitation and emission).
REFERENCES
Anderson, J Christopher, Christopher A Voigt, and Adam P Arkin. 2007. “Environmental
signal integration by a modular AND gate.” Molecular Systems Biology 3 (1). doi:
10.1038/msb4100173.
Basu, Subhayu, Rishabh Mehreja, Stephan Thiberge, Ming-Tang Chen, and Ron Weiss. 2004.
“Spatiotemporal control of gene expression with pulse-generating networks.”
Proceedings of the National Academy of Sciences of the United States of America 101
(17):6355-6360. doi: 10.1073/pnas.0307571101.
Benner, Steven A., and A. Michael Sismour. 2005. “Synthetic biology.” Nat Rev Genet 6
(7):533-543.
Callura, Jarred M., Charles R. Cantor, and James J. Collins. 2012. “Genetic switchboard for
synthetic biology applications.” Proceedings of the National Academy of Sciences 109
(15):5850-5855. doi: 10.1073/pnas.1203808109.
Canton, Barry, Anna Labno, and Drew Endy. 2008. “Refinement and standardization of
synthetic biological parts and devices.” Nat Biotech 26 (7):787-793. doi:
http://www.nature.com/nbt/journal/v26/n7/suppinfo/nbt1413_S1.html.
Chang, Yen-Chang, Chun-Liang Lin, and Tanagorn Jennawasin. 2013. “Design of Synthetic
Genetic Oscillators Using Evolutionary Optimization.” Evolutionary Bioinformatics
Online 9:137-150. doi: 10.4137/EBO.S11225.
Cuero, Raul, J. Lilly, and David S. McKay. 2012. “Constructed molecular sensor to enhance
metal detection by bacterial ribosomal switch–ion channel protein interaction.” Journal
of Biotechnology 158 (1–2):1-7. doi: http://dx.doi.org/10.1016/j.jbiotec.2012.01.011.
Danino, Tal, Octavio Mondragon-Palomino, Lev Tsimring, and Jeff Hasty. 2010. “A
synchronized quorum of genetic clocks.” Nature 463 (7279):326-330. doi:
http://www.nature.com/nature/journal/v463/n7279/suppinfo/nature08753_S1.html.
Deans, Tara L., Charles R. Cantor, and James J. Collins. “A Tunable Genetic Switch Based
on RNAi and Repressor Proteins for Regulating Gene Expression in Mammalian Cells.”
Cell 130 (2):363-372. doi: 10.1016/j.cell.2007.05.045.
Di Capite, Joseph, Siaw Wei Ng, and Anant B. Parekh. “Decoding of Cytoplasmic
Ca<sup>2+</sup> Oscillations through the Spatial Signature Drives Gene Expression.”
Current Biology 19 (10):853-858. doi: 10.1016/j.cub.2009.03.063.
Elowitz, Michael B., and Stanislas Leibler. 2000. “A synthetic oscillatory network of
transcriptional regulators.” Nature 403 (6767):335-338.
Gardner, Timothy S., Charles R. Cantor, and James J. Collins. 2000. “Construction of a
genetic toggle switch in Escherichia coli.” Nature 403 (6767):339-342. doi:
http://www.nature.com/nature/journal/v403/n6767/suppinfo/403339a0_S1.html.
Gianna De, Rubertis, and S. W. Davies. 2003. “A genetic circuit amplifier: design and
simulation.” IEEE Transactions on NanoBioscience 2 (4):239-246. doi: 10.1109/
TNB.2003.820283.
Hasty, Jeff, David McMillen, and J. J. Collins. 2002. “Engineered gene circuits.” Nature 420
(6912):224-230.
Johansson, Rolf. 1993. System modeling and identification: Prentice-hall.
Kampf, Michael M., Raphael Engesser, Moritz Busacker, Maximilian Horner, Maria
Karlsson, Matias D. Zurbriggen, Martin Fussenegger, Jens Timmer, and Wilfried Weber.
2012. “Rewiring and dosing of systems modules as a design approach for synthetic
mammalian signaling networks.” Molecular BioSystems 8 (6):1824-1832. doi:
10.1039/C2MB05509K.
Kelly, Jason R., Adam J. Rubin, Joseph H. Davis, Caroline M. Ajo-Franklin, John Cumbers,
Michael J. Czar, Kim de Mora, Aaron L. Glieberman, Dileep D. Monie, and Drew Endy.
2009. “Measuring the activity of BioBrick promoters using an in vivo reference
standard.” Journal of Biological Engineering 3 (1):1-13. doi: 10.1186/1754-1611-3-4.
Khalil, Ahmad S, and James J Collins. 2010. “Synthetic biology: applications come of age.”
Nature Reviews Genetics 11 (5):367-379.
Kobayashi, Hideki, Mads Kærn, Michihiro Araki, Kristy Chung, Timothy S. Gardner,
Charles R. Cantor, and James J. Collins. 2004. “Programmable cells: Interfacing natural
and engineered gene networks.” Proceedings of the National Academy of Sciences of the
United States of America 101 (22):8414-8419. doi: 10.1073/pnas.0402940101.
Lin, Chun-Liang, Yuan-Wei Liu, and Chia-Hua Chuang. 2009. “Control design for signal
transduction networks.” Bioinformatics and biology insights 3:1.
Lou, Chunbo, Brynne Stanton, Ying-Ja Chen, Brian Munsky, and Christopher A. Voigt.
2012. “Ribozyme-based insulator parts buffer synthetic circuits from genetic context.”
Nat Biotech 30 (11):1137-1142. doi: 10.1038/nbt.2401
http://www.nature.com/nbt/journal/v30/n11/abs/nbt.2401.html#supplementary-information.
Martineau, R. L., V. Stout, and B. C. Towe. 2010. “Optical tracking of a stress-responsive
gene amplifier applied to cell-based biosensing and the study of synthetic architectures.”
Biosensors and Bioelectronics 25 (8):1881-1888. doi: http://dx.doi.org/10.1016/
j.bios.2009.12.036.
Mondragón-Palomino, Octavio, Tal Danino, Jangir Selimkhanov, Lev Tsimring, and Jeff
Hasty. 2011. “Entrainment of a Population of Synthetic Genetic Oscillators.” Science 333
(6047):1315-1319. doi: 10.1126/science.1205369.
Mutalik, Vivek K., Joao C. Guimaraes, Guillaume Cambray, Quynh-Anh Mai, Marc Juul
Christoffersen, Lance Martin, Ayumi Yu, Colin Lam, Cesar Rodriguez, Gaymon Bennett,
Jay D. Keasling, Drew Endy, and Adam P. Arkin. 2013. “Quantitative estimation of
activity and quality for collections of functional genetic elements.” Nat Meth 10 (4):347-
353. doi: http://www.nature.com/nmeth/journal/v10/n4/abs/nmeth.2403.html#
supplementary-information.
Nistala, Goutam J., Kang Wu, Christopher V. Rao, and Kaustubh D. Bhalerao. 2010. “A
modular positive feedback-based gene amplifier.” Journal of Biological Engineering 4
(1):1-8. doi: 10.1186/1754-1611-4-4.
Purcell, Oliver, Mario di Bernardo, Claire S. Grierson, and Nigel J. Savery. 2011. “A Multi-
Functional Synthetic Gene Network: A Frequency Multiplier, Oscillator and Switch.”
PLoS ONE 6 (2):e16140. doi: 10.1371/journal.pone.0016140.
Rey, Osvaldo, Steven H. Young, Jingzhen Yuan, Lee Slice, and Enrique Rozengurt. 2005.
“Amino Acid-stimulated Ca2+ Oscillations Produced by the Ca2+-sensing Receptor Are
Mediated by a Phospholipase C/Inositol 1,4,5-Trisphosphate-independent Pathway That
Requires G12, Rho, Filamin-A, and the Actin Cytoskeleton.” Journal of Biological
Chemistry 280 (24):22875-22882. doi: 10.1074/jbc.M503455200.
Shetty, Reshma P., Drew Endy, and Thomas F. Knight. 2008. “Engineering BioBrick vectors
from BioBrick parts.” Journal of Biological Engineering 2 (1):1-12. doi: 10.1186/1754-
1611-2-5.
Townsend, and Marc Ostermeier. 2009. “An externally tunable bacterial band-pass
filter.” Proceedings of the National Academy of Sciences 106 (25):10135-10140. doi:
10.1073/pnas.0901246106.
Stricker, Jesse, Scott Cookson, Matthew R. Bennett, William H. Mather, Lev S. Tsimring,
and Jeff Hasty. 2008. “A fast, robust and tunable synthetic gene oscillator.” Nature 456
Tabor, Jeffrey J., Anselm Levskaya, and Christopher A. Voigt. 2011. “Multichromatic
Control of Gene Expression in Escherichia coli.” Journal of Molecular Biology 405
(2):315-324. doi: http://dx.doi.org/10.1016/j.jmb.2010.10.038.
Tabor, Jeffrey J., Howard M. Salis, Zachary Booth Simpson, Aaron A. Chevalier, Anselm
Levskaya, Edward M. Marcotte, Christopher A. Voigt, and Andrew D. Ellington. “A
Synthetic Genetic Edge Detection Program.” Cell 137 (7):1272-1281. doi:
10.1016/j.cell.2009.04.048.
Tamsir, Alvin, Jeffrey J. Tabor, and Christopher A. Voigt. 2011. “Robust multicellular
computing using genetically encoded NOR gates and chemical /`wires/'.” Nature 469
(7329):212-215. doi: http://www.nature.com/nature/journal/v469/n7329/abs/nature
09565.html#supplementary-information.
Taylor, S. R., R. Gunawan, L. R. Petzold, and III F. J. Doyle. 2008. “Sensitivity Measures for
Oscillating Systems: Application to Mammalian Circadian Gene Network.” IEEE
Transactions on Automatic Control 53 (Special Issue):177-188. doi:
10.1109/TAC.2007.911364.
Tuttle, Lisa M., Howard Salis, Jonathan Tomshine, and Yiannis N. Kaznessis. “Model-Driven
Designs of an Oscillating Gene Network.” Biophysical Journal 89 (6):3873-3883. doi:
10.1529/biophysj.105.064204.
Voigt, Christopher A. 2006. “Genetic parts to program bacteria.” Current Opinion in

Biotechnology 17 (5):548-557. doi: http://dx.doi.org/10.1016/j.copbio.2006.09.001.
Wang, Baojun, Richard I. Kitney, Nicolas Joly, and Martin Buck. 2011. “Engineering
modular and orthogonal genetic logic gates for robust digital-like synthetic biology.” Nat
Commun 2:508. doi: http://www.nature.com/ncomms/journal/v2/n10/suppinfo/
ncomms1516_S1.html.
Wang, Ruiqi, and Luonan Chen. 2005. “Synchronizing Genetic Oscillators by Signaling
Molecules.” Journal of Biological Rhythms 20 (3):257-269. doi: 10.1177/
0748730405275653.
Wang, Ruiqi, Luonan Chen, and Kazuyuki Aihara. 2006. “Synchronizing a multicellular
system by external input: an artificial control strategy.” Bioinformatics 22 (14):1775-
1781. doi: 10.1093/bioinformatics/btl182.
Wang, Ruiqi, Zhujun Jing, and Luonan Chen. 2005. “Modelling periodic oscillation in gene
regulatory networks by cyclic feedback systems.” Bulletin of Mathematical Biology 67
(2):339-367. doi: 10.1016/j.bulm.2004.07.005.
Chapter 4
SYSTEMATIC DESIGN OF A QUORUM SENSING-BASED

BIOSENSOR FOR ENHANCED DETECTION OF METAL
ION IN ESCHERICHIA COLI
ABSTRACT
With recent industrial expansion, heavy metal and other pollutants increasingly
contaminate our living surroundings. The non-degradability of heavy metals may lead to
accumulation in food chains and the resulting toxicity would cause damage in organisms.
Therefore, detection techniques has gradually received attention. In this chapter, a
quorum sensing (QS)-based amplifier is introduced to improve the detection performance
of metal ion biosensor. The design utilizes diffusible signal molecules, which can freely
pass through the cell membrane into the environment to communicate with others. Via
the cell-cell communication process, bacteria cooperate to display synchronous behavior,
even if only a minority of the cells detect the metal ion. In order to make the design
easier, the metal ion detection ability of the engineered biosensor is described in a steady
state model. According to user-oriented specifications, the design can be constructed via
selecting adequate promoter-RBS components from the corresponding libraries, with the
help of a genetic algorithm (GA)-based design method. The experimental results verify
that the quorum sensing-based biosensor can efficiently enhance the detection
performance of metal ion.
4.1. INTRODUCTION
Metal ions play important roles in biological metabolism pathways and affect cellular
processes via a wide variety of reactions. Organisms require metal ions involved in, such as
bacterial respiration, electron transport, and peroxide reduction (Achtman and Suerbaum
2001). The uptake and efflux systems in biology regulate ion homeostasis to avoid lack or
excess of metal ions. With recent industrial expansion, there has been either an excess of
metal ions in wastewater (Ngah and Hanafiah 2008, Rezvani-Boroujeni et al., 2015, Singha
and Guleria 2014, Teodosiu et al., 2014, Mulligan, Yong, and Gibbs 2001) or a discharge of
other contaminants (Migaszewski and Galuszka 2015) into nature intentionally or carelessly.
According to their effects and toxicity in cells, metal ions can be classified into essential ions,
trace ions, and heavy metals. The stronger redox ability of heavy metals will compete with
essential ions and trace ions in redox actions, leading to superoxide or hydroxyl radicals
(Hartwig 1995). Superoxide and hydroxyl radicals may damage lipids and proteins (Davies
2005, Raha and Robinson 2000), disrupt DNA oxidation (Hartwig 1995), and even cause cell
death (Hartwig 1995). Furthermore, heavy metals are hard to remove via metabolism and can
accumulate in organs causing permanent damage. Hence, detection techniques has gradually
received great attention. In this study, we introduced a synthetic genetic circuit that can
enhance the detection performance of metal ion biosensors and has the potential to be applied
for environmental detection and environmental bioremediation.
With recent progress in research on synthetic biology (Chang, Lin, and Jennawasin 2013,
Danino et al., 2010, Khalil and Collins 2010, Silva-Rocha and de Lorenzo 2014, Sohka et al.,
2009, Stricker et al., 2008, Wang et al., 2011, Cameron, Bashor, and Collins 2014, Kotula
et al., 2014), it offers an alternative means for metal ion detection, via the help of specific
promoter elements, such as PpbrA found in R. metallidurans (Taghavi et al., 1997, Hobman,
Julian, and Brown 2012) or PcusC and PpcoE derived from E. coli (Munson et al., 2000). In
order to improve detection performance, a metal ion-induced promoter connects with bacteria
quorum sensing system. Quorum sensing is a mechanism, which regulates gene expression
for many functions through cell-cell communication (Brint and Ohman 1995, Darch et al.,
2012, Miller and Bassler 2001, Passador et al., 1993, Pearson et al., 1995). Bacteria produce
diffusible signal molecules, generally known as autoinducers (AIs) (Kaplan and Greenberg
1985, Dunlap and Kuo 1992, Fuqua, Winans, and Greenberg 1994, Ruby and Mcfallngai
1992, Ruby and Nealson 1976), and release them into the environment to communicate with
others. Through this process, bacteria cooperate to display synchronous behavior. The most
extensively studied quorum sensing system is the autoinduction of luminescence in the
marine bacteria, Vibrio fischeri, which is symbiotic with squids and some marine fishes
(Nealson and Hastings 1979, Ruby and Mcfallngai 1992, Ruby and Nealson 1976). The
luminescence here is produced by the lux operon acquired from Vibrio fischeri, of which the
LuxR protein is the transcriptional activator of luminescence, and the LuxI protein
synthesizes the specific N-acylated homoserine lactone (AHL) (Engebrecht and Silverman
1984, Hanzelka and Greenberg 1996, Val and Cronan 1998). After the LuxR protein binds
AHL to form a complex, the complex can bind the target promoter sequence and activate
downstream gene transcription, even if only some of the bacteria can detect the metal ion. It is
helpful to enhance the detection ability by signal amplification.
To make the design of the QS-based metal ion biosensor easier, a mathematical model is
introduced to describe the dynamic and steady state regulatory behavior, which is related to
transcriptional and translational processes (Chen and Wang 2014). A promoter allows RNA-
polymerase molecules to latch onto a DNA strand and initialize transcription of a downstream
gene into mRNA, and a RBS allows ribosome to bind and translate mRNA. A promoter
combined with a RBS thus in this study is viewed as a component, of which characterizations
can be identified with a reporter protein by measuring the fluorescence intensity. One can
therefore use kinetic parameters as component libraries to design the QS-based metal ion
detection circuit. According to user-oriented specifications, the design thus can be constructed
via selecting adequate promoter-RBS components within a feasible range of metal ion
concentrations. In general, a long computation time is required when the component libraries
become large. Hence, a genetic algorithm-based design method proposed here provides a
simple and useful tool to save time in evaluating and selecting components. So in summary,
Systematic Design of a Quorum Sensing-Based Biosensor … 63
the study provides a systematic design method for the development of next-generation
synthetic biology, from component library construction to gene circuit assembly. When the
libraries are more complete, more precise detection can be achieved.
The main focuses of this chapter are fourfold: (a) A QS-based amplifier is introduced into
biosensor to enhance the metal ion detection ability. (b) Based on promoter-RBS kinetic
strengths, we establish three kinds of component libraries. (c) According to user-oriented
specifications, the quorum sensing-based metal ion biosensor can be constructed via
evaluating and selecting adequate components from the corresponding promoter-RBS
libraries to achieve a desired metal ion detection ability. (d) The proposed GA-based
searching method could provide synthetic biologists with a useful tool to design metal ion
biosensor.
4.2. METHODS
Construction of QS-Based Metal Ion Biosensor
With the help of recent research on synthetic biology, it offers an alternative means for
metal ions detection, using specific promoter elements derived from microorganisms. For
example, PcusC and PpcoE are copper-inducible promoters regulated by CusRS, which are a
part of the E. coli chromosome cus system and pco system, respectively (Bondarczuk and
Piotrowska-Seget 2013). The cus system (cusCFBA) found in E. coli is a tetrapartite system
endowing resistance to copper ion and is involved in periplasmic copper detoxification
(Franke et al., 2003). The pco system (pcoABCDRSE) is discovered in bacteria that survived
in copper-rich environments and it is verified that the plasmid-encoded operons in this system
is responsible for resistance to copper toxicity (Rouch and Brown 1997, Lee et al., 2002,
Tetaz and Luke 1983). Another example of metal ion-regulated promoter is lead-resistance
promoter PpbrA (Taghavi et al., 1997). The promoter PpbrA is acquired from Ralstonia
metallidurans strain CH34 isolated from lead-contaminants soil, containing at least seven
determinants encoding resistances to toxic heavy metals (Taghavi et al., 1997, Hobman,
Julian, and Brown 2012). Combining functions involved in uptake, efflux and accumulation
of Pb(II) ion, the lead resistance operon pbr (pbrABCDRT) is regulated by PbrR, which
belongs to the MerR family of metal regulatory proteins (Borremans et al., 2001,
Brown et al., 2003).
To improve the detection performance, in Figure 4.1, a metal ion-induced promoter-RBS
Mi connects with the LuxI protein coding sequence, allowing the LuxI protein to be translated
by the activation of the metal ion-induced promoter. The luxI component, derived from lux
operon in V. fischeri, is involved in the production of the LuxI protein, which catalyzes S-
adenosylmethionine and acyl-acyl carrier protein into a specific AHL (referred to as HSL) as
signal molecules (Hanzelka and Greenberg 1996, Val and Cronan 1998). HSLs can be
classified into several types according to acyl group length (C4–C18). Here, the type of HSL
synthesized by LuxI protein is C6-HSL. Because the lux operon is an exogenous DNA
sequence for Escherichia coli (E. coli), it is required to supply LuxR protein for the activation
of promoter Plux. The LuxR protein coding sequence is connected with a constitutive
promoter-RBS component Ci. When sufficient amounts of the LuxR protein is produced using
a constitutive promoter-RBS component with the presence of C6-HSL, C6-HSL can then bind
LuxR protein to form the LuxR complex. The complex targets the cognate QS-dependent
promoter-RBS component Ak and activates the transcription of the green fluorescent protein
(GFP) coding sequence, even if only some of the bacteria can detect the metal ion. It is
helpful to enhance the detection ability by signal amplification.
Cell 1
Metal ion
Cell N
xI xR G
xG
RBS luxI RBS luxR RBS gfp
Metal ion-induced Constitutive QS-dependent

promoter-RBS Mi promoter-RBS Cj promoter-RBS Ak
QS-based amplifier
Cell i
Figure 4.1. QS-based metal ion biosensor. A metal ion-induced promoter-RBS connects with a quorum
sensing system found in the marine bacteria, V. fischeri. Via the transmission of signal molecules, the
output signal will be produced significantly by the activation of the QS-dependent promoter.
Mathematical Model of QS-Based Metal Ion Biosensor
To construct the dynamic model, promoter-RBS regulation must be introduced. The

promoter-RBS regulation function is first defined by P(Pu, Pl, TF, I), in which Pu and Pl
denote the maximum and minimum promoter-RBS strengths, respectively, TF denotes
transcription factor concentration, and I is inducer concentration. This function describes the
biochemical aspect of the transcription and translation process. The dynamic model of the
metal ion biosensor with QS-based amplifier in Figure 4.1 thus is described as follows. Note
that the QS-based metal ion biosensor is divided into three stages.
The first stage involves a metal ion-induced promoter-RBS Mi for producing autoinducer
synthase LuxI, which catalyzes S-adenosylmethionine and acyl-acyl carrier protein into a
specific AHL. The dynamics is described as the equation below (Alon 2007):
xE (t )  PM ( Pu ,i , Pl ,i , xS , I M )  (d  rE )  xE (t ) (4.1)
in which
Pu ,i  Pl ,i
PM ( Pu ,i , Pl ,i , xS , I M )  Pu ,i  nSI
 K SI 
1  
 xSI ( xS , I M ) 
xS
xSI ( xS , I M ) 
K 
1  M 
 IM 
where xE denotes the concentration of autoinducer synthase. IM is the concentration of metal

ion; xS is the total concentration of metal ion-dependent regulatory protein. xSI denotes the
complex of xI and IM. i denotes the ith metal ion-induced promoter-RBS component in Table
4.1 in Appendix. PM(Pu,i,Pl,i,xS,IM) denotes the promoter-RBS regulation activity of the metal
ion-induced promoter-RBS components. Pu,i and Pl,i denote the maximum and minimum
promoter-RBS strengths of the ith metal ion-induced promoter-RBS component in Table 4.1.
rE denotes the degradation rate for autoinducter synthase. d is the dilution rate due to cell
growth. KSI and nSI denote the binding affinity and binding cooperativity between the complex
xSI and the corresponding promoter-RBS component, respectively. KM is the dissociation rate
between the metal ion IM and the metal regulatory protein xS. Then, the dynamics for the
catalysis of S-adenosylmethionine and acyl-acyl carrier protein into a specific autoinducer is
given by
xI (t )  axE (t )  (d  rI )  xI (t ) (4.2)
where xI is the concentration of autoinducer. a is autoinducer synthesis rate; rI denotes the

degradation rate for autoinducter.
The second stage involves a constitutive promoter-RBS part Ci for producing protein,
LuxR. The change in concentration of regulatory protein LuxR with respect to time is given
by (Alon 2007)
xR (t )  PC ( Pu , j ,0,0,0)  (d  rR )  xR (t ) (4.3)
and
PC ( Pu , j ,0,0,0)  Pu , j
where xR denotes the concentration of transcriptional activator protein, LuxR. j is the jth
constitutive promoter-RBS component in Table 4.2. PC(Pu,i,0,0,0) is the regulation activity of
the constitutive promoter-RBS components. Pu,j is the promoter-RBS strength of the jth
constitutive promoter-RBS component in Table 4.2. rR denotes the degradation rate for
transcriptional activator protein.
The third stage contains a QS-dependent promoter-RBS part Ak for driving the expression
of the immature reporter protein. When complex of xR and xI accumulates, it activates the QS-
dependent promoter and enhances the expression of reporter gene. Thus, the dynamics is
governed by the equation as below (Alon 2007):
xG (t )  PA ( Pu ,k , Pl ,k , xR , xI )  (d  rG )  xG (t ) (4.4)
and
Pu , k  Pl , k
PA ( Pu , k , Pl , k , xR , xI )  Pu , k  nRI
 K RI 
1  
 RI R I 
x ( x , x )
xR
xRI ( xR , xI ) 
 KI 
1  
 xI 
where xG is the concentration of immature reporter protein. xRI represents the complex of xR
and xI. k is the kth QS-dependent promoter-RBS component in Table 4.3. PA(Pu,k,Pl,k,xR,xI) is
the activity of the QS-dependent promoter-RBS components. Pu,k and Pl,k are the maximum
and minimum promoter-RBS strengths of the kth QS-dependent promoter-RBS component in
Table 4.3. rG is the degradation rate for immature reporter protein. KRI and nRI are the binding
affinity and binding cooperativity between the complex xRI and the promoter-RBS part,
respectively. KI is the dissociation rate between the autoinducer xI and the transcriptional
activator protein xR. Finally, the dynamics for the maturation of xG into the mature reporter
protein G is given by
G(t )  mxG  (d  r )G(t ) (4.5)
where m is the maturation rate for the reporter protein and r is the degradation rate for mature
reporter protein.
The dynamic model for the QS-based metal ion biosensor is then transformed into
steady-state model by assuming that the dynamics in (4.1)-(4.5) are equal to zero. The steady-
state model of the QS-based metal ion biosensor with promoter-RBS parts Mi, Cj, and Ak
being selected from the corresponding libraries in Appendix, respectively, is derived as
follows:
 PM ( Pu ,i , Pl ,i , xS , I M ) a
 xISS  
 d  rE d  rI
 PC ( Pu , j ,0,0,0)
 xRSS 
 d  rR
 (4.6)
 PA ( Pu , k , Pl , k , xR , xI )
 xGSS  d  rG

 m
GSS  xG
 d r
where xISS, xRSS, xGSS and GSS are the steady-state concentrations of autoinducter, transcriptional
activator protein, and immature reporter protein, as well as steady state mature reporter
protein, respectively. The steady-state expression of the QS-based metal ion biosensor can be
obtained from the steady-state model in (4.6) with the regulation functions PM(Pu,i,Pl,i,xS,IM),
PC(Pu,i,0,0,0) as well as PA(Pu,k,Pl,k,xR,xI), respectively.
In general, a synthetic genetic circuit in vivo also suffers from extrinsic environmental
cellular noise, such as transmitted noise from upstream and global noise affecting all cells
(Chen and Wang 2006). Hence, the equation in (4.6) should be modified as follows:
 PM ( Pu ,i , Pl ,i , xS , I M ) a
 xISS    v1
 d  rE d  rI
 PC ( Pu , j ,0,0,0)
 xRSS   v2
 d  rR
 (4.7)
 PA ( Pu , k , Pl , k , xR , xI )
 xGSS  d  rG
 v3

 m
GSS  xG  v4
 d r
where the Gaussian random noise vp, p = 1, 2, 3, with a zero mean and variance of σp2,
denotes cellular noise for both the transcriptional and translational gene expression processes
at the steady state. v4 denotes cellular noise in mature protein expression at the steady state.
Besides, biological components are inherently uncertain in a molecular biological system
(Chen and Wang 2006). For example, the intrinsic kinetic parameters of the components
including the processes of transcription and translation, the degradation rates of regulatory
proteins, dilution rates of the cells, and the maturation rate for the reporter proteins, are all
stochastically uncertain in vivo, as a result of gene expression noise from biochemical
processes, thermal fluctuations, DNA mutation, and evolution. These perturbations are
defined as follows:
Pu ,i  Pu ,i  Pu ,i n1 (t ), Pl ,i  Pl ,i  Pl ,i n1 (t ),
Pu , j  Pu , j  Pu , j n1 (t ),
Pu , k  Pu , k  Pu , k n1 (t ), Pl , k  Pl , k  Pl , k n1 (t ),
rE  rE  rE n1 (t ),
rR  rR  rR n1 (t ), (4.8)
rG  rG  rG n1 (t ),
m  m  mn2 (t ),
r  r  rn2 (t ),
d  d  dn3 (t )
where ΔPu,i, ΔPl,i, ΔPu,j, ΔPu,k, ΔPl,k, ΔrE, ΔrR, ΔrG, Δr, Δm, and Δd are the standard
deviations of the corresponding stochastic parameters, and nq(t), q = 1,2,3, denote Gaussian
noise, which have zero mean and unit variance, and account for random fluctuation sources.
If the kinetic parameters in the steady state model in (4.7) are replaced by the parameter
perturbations shown in (4.8) for robust design of the gene circuit, then the QS-based metal ion
biosensor can tolerate these fluctuations in vivo, i.e., a design that accounts for (4.8) should be
able to tolerate the parameter fluctuations.
Design Specifications for QS-based Metal Ion Biosensor
The purpose of our design is to construct a QS-based metal ion biosensor by selecting a
set of suitable components from the corresponding libraries to achieve optimal tracking of a
desired I/O response within a feasible range of metal ion concentrations. To achieve this, the
following design specifications are needed:
 Desired I/O response Gref(IM) of the synthetic genetic QS-based metal ion biosensor
 A feasible range of metal ion concentrations
 Standard derivations of cellular disturbances and parameter fluctuations to be
tolerated in vivo
 A cost function between the desired reference steady state fluorescence intensity Gref
and the steady state fluorescence intensity GSS in (4.7) within a specified range of
metal ion concentrations is given to be minimized as follows:
J (S )  E   Gss (S , I M )  Gref ( I M )  dI M
2
(4.9)
where S denotes the set of promoter-RBS components Mi, Cj, and Ak to be selected from the
corresponding libraries.
If the cost function in (4.9) is minimized by choosing the most appropriate set of
components under design specifications, the fluorescence intensity of a engineered metal ion
biosensor will match the specified steady state fluorescence intensity Gref(IM) under parameter
fluctuations and environmental disturbances optimally. Although the cost function J(S) can be
minimized by traditional conventional search methods, it will require long computation times
and several trial-and-error experimentations when component libraries become large. Thus, a
more effective and efficient genetic algorithm (GA)-based searching method is proposed here
to save time in evaluating and selecting adequate promoter-RBS components.
Design Procedure for the QS-Based Metal Ion Biosensor
The design procedure for a QS-based metal ion biosensor is summarized as follows:
1. Provide user-defined design specifications Gref(IM) for the quorum sensing-based

metal ion biosensor.
2. Select an initial set S of promoter-RBS components.
3. Calculate the cost value J(S) in (4.9) for S. The relationship between the fitness value
F(S) and the cost value J(S) is inversely proportional to (4.10):
1
F (S )  (4.10)
J (S )
The fitness value plays a key role in natural selection for selecting a set S of promoter-
RBS components to maximize the fitness value F(S), or equivalently to minimize the cost
function J(S) in(4.9). For example:
1
max F ( S )  (4.11)
S min J ( S )
S
4. Create an offspring set S, using GA operators such as reproduction, crossover, and

mutation.
 Make copies of possible solutions, on basis of their fitness.
 Swap values between two possible solutions.
 Randomly alter the value in a possible solution.
5. Calculate the cost value of the new set S obtained by natural selection. Stop when the
design goal is achieved or an acceptable solution is obtained. Otherwise, create the
next generation and return to step 3.
4.3. RESULTS
For the convenience of description and explanation, as shown in Figure 4.1, a QS-based
metal ion biosensor is assembled by selecting a set of promoter-RBS components, namely, a
metal ion-induced promoter-RBS component Mi, a constitutive promoter-RBS component Cj,
and a QS-dependent promoter-RBS component Ak. The design specifications are as follows:
 The desired fluorescence intensity to different metal ion concentrations as shown in

Figure 4.2 is described as follows:
5000
Gref ( I M )  65  2
(4.12)
 101 
1  
 IM 
 Input operation range: 10-3 to 100 μM

 The standard deviations of parameter fluctuations that are supposed to be tolerated in
vivo are given by
Pu ,i  0.05Pu ,i , Pl ,i  0.05Pl ,i , Pu , j  0.05 Pu , j ,

Pu , k  0.05Pu , k , Pl , k  0.05Pl , k , rE  0.05rE ,
(4.13)
rR  0.05rR , rG  0.05rG , r  0.05r ,
m  0.05m, d  0.05d ,
as well as the external environmental noise vp, p = 1,2,3, for transcription and translation
processes, and noise v4 for mature reporter protein expression, are all Gaussian, with zero
mean and unit variance.
 The cost function is given by
G (S , I M )  Gref ( I M )  dI M
100
J (S )  E 
2
3 ss (4.14)
10
Fig. 4.2. Results for the QS-based metal ion biosensor. By minimizing the cost function for the metal
ion biosensor in Fig. 4.1, an adequate set S  (M1 , C6 , A3 ) is selected from the corresponding libraries
in Tables 4.1- 4.3. The green points are the experimental results (mean of three trials) by
S  (M1 , C6 , A3 ) . The gray solid line is the desired I/O response in (4.12)
In order to efficiently solve the constrained optimal matching design problem of the
metal ion biosensor, a GA-based library search method is employed to search a set S from
corresponding libraries in Appendix to minimize the cost function (4.14). The adequate
promoter-RBS components from the corresponding libraries are found to be M1, C6 and A3.
The desired response is shown in Figure 4.2, with the fluorescence intensity values taken
from Figure 4.3 under different Cu(II) ion concentrations. Clearly, the metal ion biosensor
can match the desired response despite the parameter fluctuations and environmental
disturbances. Besides, the detection performance is better than one without quorum sensing-
based amplifier in Figure 4.4 at the same copper ion concentrations (see Figure 4.5 and Figure
4.6). In particular, the QS-based metal ion biosensor increases the sensitivity to Cu(II) ion by
roughly one order of magnitude and the dynamic range by roughly fourfold. It confirms the
proposed QS-based metal ion biosensor is useful for enhancing the detection ability.
-3 -2
Cu(II) = 10  M Cu(II) = 10  M
7000 7000
6000 6000
5000 5000
GFP/OD600 (a.u.)
GFP/OD600 (a.u.)
4000 4000
3000 3000
2000 2000
1000 1000
0 0
0 50 100 150 200 250 300 350 400 450 500 0 50 100 150 200 250 300 350 400 450 500
Time (min) Time (min)

-1
Cu(II) = 5x10  M
-2 Cu(II) = 10  M
7000 7000
6000 6000
5000 5000
GFP/OD600 (a.u.)
4000
GFP/OD600 (a.u.)
4000
3000 3000
2000 2000
1000 1000
0 0
0 50 100 150 200 250 300 350 400 450 500 0 50 100 150 200 250 300 350 400 450 500

-1
Cu(II) = 5x10  M 0
Cu(II) = 10  M
7000
7000
6000 6000
5000 5000
GFP/OD600 (a.u.)
4000
GFP/OD600 (a.u.)
4000
3000 3000
2000 2000
1000 1000
0 0
0 50 100 150 200 250 300 350 400 450 500 0 50 100 150 200 250 300 350 400 450 500
Figure 4.3. Time profiles for QS-based metal ion biosensor under different concentrations of Cu(II) ion
with measurement being taken every 15 min. The green points are the experimental results by S = (M1,
C6, A3).
Cell 1
Cu(II)
Cell N
RBS gfp
Metal ion-induced
promoter-RBS
Cell i
Figure 4.4. Copper ion biosensor without quorum sensing-based amplifier, using metal ion-induced
promoter-RBS component M1 selected from the library in Appendix.
7000
6000
5000
GFP/OD600 (a.u.)
4000
3000
2000
1000
0
-3 -2 -1 0
10 10 10 10
Cu(II) ( M)
Figure 4.5. The results of a Cu(II) ion biosensor without QS-based amplifier. The green points are the
experimental results by M1 from the library in Appendix.
Figure 4.6. Comparison of Cu(II) ion biosensor with (top) and without (bottom) QS-based amplifier.
The Cu(II) concentrations from left to right are 0, 10-3, 10-2, 5x10-2, 10-1, 5x10-1, and 1 μM.
4.4. CONCLUSION
In this chapter, a QS-based biosensor is constructed to improve the performance of metal
ion detection. Based on four design specifications, the QS-based biosensor can be constructed
via selecting adequate promoter-RBS components in combination with a feasible range of
metal ion concentrations. The proposed GA-based searching method could provide synthetic
biologists with a useful tool to save time in selecting adequate promoter-RBS components to
meet the specified detection performance. The experimental results verify that the quorum
sensing-based biosensor can efficiently enhance the detection performance of metal ion. In
future, the mechanism of metal ions detection can be used to environmental bioremediation
(Wang and Chen 2009). Bioremediation, which utilizes the ability of biosorption (e.g.,
protein) to bind the chemicals or ions with biomass or biopolymers, is referred to as the
promising approach for environmental treatment. Proteins for biosorption are extensively
studied for their specific binding regions and are engineered as biosorbent because of their
higher adsorption capacity, low cost, and reusability (Wang and Chen 2009, He and Chen
2014). Research indicates that biosorbents contain a variety of functional sites including
carboxyl, phosphate, amino groups, etc., and biology materials, like bacteria, yeast, and fungi
have gained attention for their removal and recovery abilities (Wang and Chen 2009, He and
Chen 2014). Ravikumar et al., used a recombinant strain based on the specific interaction of
Cu(II) with metal binding peptide, which displayed selective adsorption of Cu(II) from
aqueous solutions (Ravikumar et al., 2012). Thus, the QS-based biosensor can be combined
with the Cu(II) bioadsorption system to promote bioremediation (see Figure 4.7).
Cell 1
Cu(II)
Cell N
C6-HSL
RBS luxI RBS luxR RBS OmpC CBP
Metal ion-induced Constitutive QS-dependent

promoter-RBS promoter-RBS promoter-RBS
QS-based amplifier
Cell i
Figure 4.7. A copper sensing and adsorbing system in E. coli.
4.5. MATERIALS AND METHODS
Biosensor Reagents
All restriction enzymes and DNA ligation kit are purchased from New England Biolabs.
The chemicals used here are from Sigma-Aldrich. The oligonucleotides are from Integrated
DNA Technologies.
Bacteria Strains and Medium
Escherichia coli strain DH5α cells from Yeastern biotech (ECOS) are used for the
construction of the procedure and for the fluorescence measurement experiments. An M9
working medium (34 g/L Na2HPO4, 15 g/L K2HPO4, 2.5 g/L NaCl, 5 g/L NH4Cl, 2 g/L
casamino acids, 2 μM vitamin B1, 2 mM MgSO4, 0.2% glucose) with proper antibiotics is
used for E. coli cultivation at 37 °C and 200 r.p.m..
Plasmid Construction
The DNA parts used in this study are selected from BioBrick or synthesized by MDBio
Biotech Co. Ltd. All DNA parts are assembled into the backbone pSB3K3 (20–30 copies per
cell), via the BioBrick standard assembly method.
Fluorescence Measurement
The cells containing the genetic circuit are inoculated in the M9 working medium with 50
mg/mL Kanamycin. The culture is incubated for approximately 14–16 hours, after which it is
diluted 250-fold and further incubated until the OD600 reaches 0.1. Different concentrations
of CuSO4 are then added to the medium. The fluorescence intensity of the culture is
measured subsequently by the microplate reader (BioTek, Synergy™ H1, GFP settings are
490/510 nm for excitation and emission).
4.6. APPENDIX
To construct a component library, promoter-RBS regulation must be introduced first. The
regulation function here is defined by P(Pu, Pl, TF, I), in which Pu and Pl denote the
maximum and minimum promoter-RBS strengths, respectively, TF is transcription factor
concentration, and I denotes inducer concentration. Thus, the expression of a reporter protein
produced from a promoter-RBS component can be described as
 xG (t )  P( Pu , Pl , TF , I )  (d  rG ) xG (t )
 (4.15)
G(t )  mxG  (d  r )G(t )
where xG is the concentration of immature reporter protein; G is the concentration of mature

reporter protein. d represents the dilution rate due to cell growth. m denotes the maturation
rate for the reporter protein. rG and r are the degradation rates for immature reporter protein
and mature reporter protein, respectively. Besides, the details of the regulation functions for
metal ion-induced promoter-RBS component Mi, constitutive promoter-RBS component Ci
and QS-dependent promoter-RBS component Ak can be found as follows.
Regulation Function for Metal Ion-Induced Promoter-RBS Component
A metal-ion induced promoter-RBS component Mi is regulated by regulatory protein xS,

which can bind with metal ion IM to influence downstream gene expression. The
concentration of metal ion-bound regulatory protein can be described as follows:
xS
xSI ( xS , I M )  (4.16)
K 
1  M 
 IM 
where xS denotes the total metal ion-dependent regulatory protein including free and metal
ion-bound regulatory protein; IM denotes metal ion concentration; KM is the dissociation rate
between the metal ion IM and the metal regulatory protein xS. Therefore, the regulation
activity of the metal ion-induced promoter-RBS component Mi can be represented by a
promoter-RBS regulation function as follows:
Pu ,i  Pl ,i
(4.17)
 K SI 
1  
 xSI ( xS , I M ) 
where i denotes the ith metal ion-induced promoter-RBS component. Pu,i and Pl,i are the
maximum and minimum promoter-RBS strengths of the ith metal ion-induced promoter-RBS
component. KSI and nSI denote the binding affinity and binding cooperativity between the
complex xSI and the corresponding promoter-RBS component, respectively. The metal ion-
induced promoter-RBS components here are constructed by two copper-induced promoters,
namely, PcusC and PpcoE and a RBS, namely, B0034.
Regulation Function for Consontitutive Promoter-RBS Component
A constitutive promoter-RBS part Ci can express downstream gene constitutively, with

no influence by any transcription factor. Therefore, the regulation activity of the constitutive
promoter-RBS component Ci can be represented by a promoter-RBS regulation function as
follows:
PC ( Pu , j ,0,0,0)  Pu , j (4.18)
where j denotes the jth constitutive promoter-RBS component from the corresponding library.
Pu,j denotes the promoter-RBS strength of the jth constitutive promoter-RBS component. The
constitutive promoter-RBS components here are constructed by three constitutive promoters,
namely, J23101, J23105, and J23106, as well as three RBSs, namely, B0031, B0032 and
B0034.
Regulation Function for QS-Dependent Promoter-RBS Component
A QS-dependent promoter-RBS component Ak is regulated by activator protein xR, which

can bind with autoinducer xI to influence downstream gene expression. The concentration of
inducer-bound activator can be described as follows:
xR
xRI ( xR , xI )  (4.19)
K 
1  I 
 xI 
where xR denotes the total concentration of transcriptional activator protein; xI denotes the
concentration of autoinducter. KI is the dissociation rate between the autoinducer xI and the
transcriptional activator protein xR. Therefore, the regulation activity of the QS-dependent
promoter-RBS component Ak can be represented by a promoter-RBS regulation function as
follows:
Pu , k  Pl , k
(4.20)
 K RI 
1  
 xRI ( xR , xI ) 
where k is the kth QS-dependent promoter-RBS component. Pu,k and Pl,k are the maximum and
minimum promoter-RBS strengths of the kth QS-dependent promoter-RBS component. KRI
and nRI are the binding affinity and binding cooperativity between the complex xRI and the
promoter-RBS part, respectively. The QS-dependent promoter-RBS components here are
constructed by a QS-dependent promoter, namely, R0062 and three RBSs, namely, B0031,
B0032 and B0034.
The dynamic model is transformed into steady-state model by assuming that the
dynamics in (4.15) are equal to zero. The steady-state model is derived as follows:
 P( Pu , Pl , TF , I )
 xGSS (t ) 
 d  rG
 (4.21)
G (t )  m x


SS
d r
G
where xGSS and GSS are the steady-state concentrations of immature reporter protein and
mature reporter protein, respectively. Then based on the steady-state model in (4.21), the
identification results are used to construct libraries and listed in Tables 4.1, 4.2 and 4.3,
respectively.
Table 4.1. Metal ion-induced promoter-RBS component library in E. coli strain DH5α
Index Component Pu Pl
M1 PcusC-B0034 100.630 4.000 KSI = 81 nM
KM = 0.3 μM
nSI = 1
M2 PpcoE-B0034 98.237 1.903 KSI = 154nM
KM = 0.3 μM
nSI = 1
Table 4.2. Constitutive promoter-RBS component library in E. coli strain DH5α
Index Component Pu
C1 J23101-B0031 48.203
C2 J23101-B0032 22.263
C3 J23101-B0034 145.947
C4 J23105-B0031 5.178
C5 J23105-B0032 3.449
C6 J23105-B0034 14.342
C7 J23106-B0031 8.580
C8 J23106-B0032 5.813
C9 J23106-B0034 28.874
Table 4.3. QS-dependent promoter-RBS component library in E. coli strain DH5α
Index Component Pu Pl
A1 R0062-B0031 362.902 1.100 KRI = 95 nM
A2 R0062-B0032 238.132 1.000 KI = 3.8nM
A3 R0062-B0034 848.678 1.740 nRI = 1
REFERENCES
Achtman, Mark, and Sebastian Suerbaum. 2001. Helicobacter pylori: molecular and cellular
biology. Wymondham, Norfolk, England: Horizon Scientific Press.
Alon, Uri. 2007. An introduction to systems biology: design principles of biological circuits,
Chapman & Hall/CRC mathematical and computational biology series. Boca Raton, FL:
Chapman & Hall/CRC.
Bondarczuk, K., and Z. Piotrowska-Seget. 2013. “Molecular basis of active copper resistance
mechanisms in Gram-negative bacteria.” Cell Biology and Toxicology 29 (6):397-405.
doi: DOI 10.1007/s10565-013-9262-1.
Borremans, B., J. L. Hobman, A. Provoost, N. L. Brown, and D. van Der Lelie. 2001.
“Cloning and functional analysis of the pbr lead resistance determinant of Ralstonia
metallidurans CH34.” J Bacteriol 183 (19):5651-8. doi: 10.1128/JB.183.19.5651-
5658.2001.
Brint, J. M., and D. E. Ohman. 1995. “Synthesis of multiple exoproducts in Pseudomonas
aeruginosa is under the control of RhlR-RhlI, another set of regulators in strain PAO1
with homology to the autoinducer-responsive LuxR-LuxI family.” J Bacteriol 177
(24):7155-63.
Brown, N. L., J. V. Stoyanov, S. P. Kidd, and J. L. Hobman. 2003. “The MerR family of
transcriptional regulators.” Fems Microbiology Reviews 27 (2-3):145-163. doi: Doi
10.1016/S0168-6445(03)00051-2.
Cameron, D. E., C. J. Bashor, and J. J. Collins. 2014. “A brief history of synthetic biology.”
Nature Reviews Microbiology 12 (5):381-390. doi: DOI 10.1038/nrmicro3239.
Chang, Y. C., C. L. Lin, and T. Jennawasin. 2013. “Design of Synthetic Genetic Oscillators
Using Evolutionary Optimization.” Evolutionary Bioinformatics 9.
Chen, B. S., and Y. C. Wang. 2006. “On the attenuation and amplification of molecular noise
in genetic regulatory networks.” BMC Bioinformatics 7. doi: 10.1186/1471-2105-7-52.
Chen, Bor-Sen, and Yu-Chao Wang. 2014. Synthetic gene network : modeling, analysis, and
robust design methods. Boca Raton: CRC Press,Taylor & Francis Group.
Danino, T., O. Mondragon-Palomino, L. Tsimring, and J. Hasty. 2010. “A synchronized
quorum of genetic clocks.” Nature 463 (7279):326-330.
Darch, S. E., S. A. West, K. Winzer, and S. P. Diggle. 2012. “Density-dependent fitness
benefits in quorum-sensing bacterial populations.” Proc Natl Acad Sci U S A 109
(21):8259-63. doi: 10.1073/pnas.1118131109.
Davies, M. J. 2005. “The oxidative environment and protein damage.” Biochimica Et
Biophysica Acta-Proteins and Proteomics 1703 (2):93-109. doi: DOI
10.1016/j.bbapap.2004.08.007.
Dunlap, P. V., and A. Kuo. 1992. “Cell Density-Dependent Modulation of the Vibrio-Fischeri
Luminescence System in the Absence of Autoinducer and Luxr Protein.” Journal of
Bacteriology 174 (8):2440-2448.
Engebrecht, J., and M. Silverman. 1984. “Identification of Genes and Gene-Products
Necessary for Bacterial Bioluminescence.” Proceedings of the National Academy of
Sciences of the United States of America-Biological Sciences 81 (13):4154-4158. doi:
DOI 10.1073/pnas.81.13.4154.
Franke, S., G. Grass, C. Rensing, and D. H. Nies. 2003. “Molecular analysis of the copper-
transporting efflux system CusCFBA of Escherichia coli.” Journal of Bacteriology 185
(13):3804-3812. doi: Doi 10.1128/Jb.185.13.3804-3812.2003.
Fuqua, W. C., S. C. Winans, and E. P. Greenberg. 1994. “Quorum Sensing in Bacteria - the
Luxr-Luxi Family of Cell Density-Responsive Transcriptional Regulators.” Journal of
Bacteriology 176 (2):269-275.
Hanzelka, B. L., and E. P. Greenberg. 1996. “Quorum sensing in Vibrio fischeri: evidence
that S-adenosylmethionine is the amino acid substrate for autoinducer synthesis.” J
Bacteriol 178 (17):5291-4.
Hartwig, A. 1995. “Current Aspects in Metal Genotoxicity.” Biometals 8 (1):3-11.
He, J., and J. P. Chen. 2014. “A comprehensive review on biosorption of heavy metals by
algal biomass: materials, performances, chemistry, and modeling simulation tools.”
Bioresour Technol 160:67-78. doi: 10.1016/j.biortech.2014.01.068.
Hobman, J. L., D. J. Julian, and N. L. Brown. 2012. “Cysteine coordination of Pb(II) is
involved in the PbrR-dependent activation of the lead-resistance promoter, PpbrA, from
Cupriavidus metallidurans CH34.” Bmc Microbiology 12. doi: Artn 109. Doi
10.1186/1471-2180-12-109.
Kaplan, H. B., and E. P. Greenberg. 1985. “Diffusion of Autoinducer Is Involved in
Regulation of the Vibrio-Fischeri Luminescence System.” Journal of Bacteriology 163
(3):1210-1214.
Khalil, A. S., and J. J. Collins. 2010. “Synthetic biology: applications come of age.” Nature
Reviews Genetics 11 (5):367-379. doi: Doi 10.1038/Nrg2775.
Kotula, J. W., S. J. Kerns, L. A. Shaket, L. Siraj, J. J. Collins, J. C. Way, and P. A. Silver.
2014. “Programmable bacteria detect and record an environmental signal in the
mammalian gut.” Proceedings of the National Academy of Sciences of the United States
of America 111 (13):4838-4843. doi: DOI 10.1073/pnas.1321321111.
Lee, S. M., G. Grass, C. Rensing, S. R. Barrett, C. J. D. Yates, J. V. Stoyanov, and N. L.

Brown. 2002. “The Pco proteins are involved in periplasmic copper handling in
Escherichia coli.” Biochemical and Biophysical Research Communications 295 (3):616-
620. doi: Pii S0006-291x(02)00726-X. Doi 10.1016/S0006-291x(02)00726-X.
Migaszewski, Z. M., and A. Galuszka. 2015. “The Characteristics, Occurrence, and
Geochemical Behavior of Rare Earth Elements in the Environment: A Review.” Critical
Reviews in Environmental Science and Technology 45 (5):429-471. doi: Doi
10.1080/10643389.2013.866622.
Miller, M. B., and B. L. Bassler. 2001. “Quorum sensing in bacteria.” Annual Review of
Microbiology 55:165-199. doi: DOI 10.1146/annurev.micro.55.1.165.
Mulligan, C. N., R. N. Yong, and B. F. Gibbs. 2001. “Remediation technologies for metal-
contaminated soils and groundwater: an evaluation.” Engineering Geology 60 (1-4):193-
207. doi: Doi 10.1016/S0013-7952(00)00101-0.
Munson, G. P., D. L. Lam, F. W. Outten, and T. V. O'Halloran. 2000. “Identification of a
copper-responsive two-component system on the chromosome of Escherichia coli K-12.”
J Bacteriol 182 (20):5864-71.
Nealson, K. H., and J. W. Hastings. 1979. “Bacterial bioluminescence: its control and
ecological significance.” Microbiol Rev 43 (4):496-518.
Ngah, W. S. W., and M. A. K. M. Hanafiah. 2008. “Removal of heavy metal ions from
wastewater by chemically modified plant wastes as adsorbents: A review.” Bioresource
Technology 99 (10):3935-3948. doi: DOI 10.1016/j.biortech.2007.06.011.
Passador, L., J. M. Cook, M. J. Gambello, L. Rust, and B. H. Iglewski. 1993. “Expression of
Pseudomonas-Aeruginosa Virulence Genes Requires Cell-to-Cell Communication.”
Science 260 (5111):1127-1130. doi: DOI 10.1126/science.8493556.
Pearson, J. P., L. Passador, B. H. Iglewski, and E. P. Greenberg. 1995. “A second N-
acylhomoserine lactone signal produced by Pseudomonas aeruginosa.” Proc Natl Acad
Sci U S A 92 (5):1490-4.
Raha, S., and B. H. Robinson. 2000. “Mitochondria, oxygen free radicals, disease and
ageing.” Trends in Biochemical Sciences 25 (10):502-508. doi: Doi 10.1016/S0968-
0004(00)01674-1.
Ravikumar, S., I. Ganesh, I. K. Yoo, and S. H. Hong. 2012. “Construction of a bacterial
biosensor for zinc and copper and its application to the development of multifunctional
heavy metal adsorption bacteria.” Process Biochemistry 47 (5):758-765. doi: DOI
10.1016/j.procbio.2012.02.007.
Rezvani-Boroujeni, A., M. Javanbakht, M. Karimi, C. Shahrjerdi, and B. Akbari-adergani.
2015. “Immoblization of Thiol-Functionalized Nanosilica on the Surface of Poly(ether
sulfone) Membranes for the Removal of Heavy-Metal Ions from Industrial Wastewater
Samples.” Industrial & Engineering Chemistry Research 54 (1):502-513. doi: DOI
10.1021/ie504106y.
Rouch, D. A., and N. L. Brown. 1997. “Copper-inducible transcriptional regulation at two
promoters in the Escherichia coli copper resistance determinant pco.” Microbiology-Uk
143:1191-1202.
Ruby, E. G., and M. J. Mcfallngai. 1992. “A Squid That Glows in the Night - Development of
an Animal-Bacterial Mutualism.” Journal of Bacteriology 174 (15):4865-4870.
Ruby, E. G., and K. H. Nealson. 1976. “Symbiotic association of Photobacterium fischeri

with the marine luminous fish Monocentris japonica; a model of symbiosis based on
bacterial studies.” Biol Bull 151 (3):574-86.
Silva-Rocha, R., and V. de Lorenzo. 2014. “Engineering Multicellular Logic in Bacteria with
Metabolic Wires.” Acs Synthetic Biology 3 (4):204-209.
Singha, A. S., and A. Guleria. 2014. “Chemical modification of cellulosic biopolymer and its
use in removal of heavy metal ions from wastewater.” International Journal of Biological
Macromolecules 67:409-417. doi: DOI 10.1016/j.ijbiomac.2014.03.046.
2009. “An externally tunable bacterial band-pass filter.” Proceedings of the National
Academy of Sciences of the United States of America 106 (25):10135-10140.
Stricker, J., S. Cookson, M. R. Bennett, W. H. Mather, L. S. Tsimring, and J. Hasty. 2008. “A
fast, robust and tunable synthetic gene oscillator.” Nature 456 (7221):516-U39.
Taghavi, S., M. Mergeay, D. Nies, and D. vanderLelie. 1997. “Alcaligenes eutrophus as a
model system for bacterial interactions with heavy metals in the environment.” Research
in Microbiology 148 (6):536-551. doi: Doi 10.1016/S0923-2508(97)88361-1.
Teodosiu, C., R. Wenkert, L. Tofan, and C. Paduraru. 2014. “Advances in preconcentration/
removal of environmentally relevant heavy metal ions from water and wastewater by
sorbents based on polyurethane foam.” Reviews in Chemical Engineering 30 (4):403-420.
doi: DOI 10.1515/revce-2013-0036.
Tetaz, T. J., and R. K. Luke. 1983. “Plasmid-controlled resistance to copper in Escherichia
coli.” J Bacteriol 154 (3):1263-8.
Val, D. L., and J. E. Cronan. 1998. “In vivo evidence that S-adenosylmethionine and fatty
acid synthesis intermediates are the substrates for the LuxI family of autoinducer
synthases.” Journal of Bacteriology 180 (10):2644-2651.
Wang, B. J., R. I. Kitney, N. Joly, and M. Buck. 2011. “Engineering modular and orthogonal
genetic logic gates for robust digital-like synthetic biology.” Nature Communications 2.
Wang, J., and C. Chen. 2009. “Biosorbents for heavy metals removal and their future.”
Biotechnol Adv 27 (2):195-226. doi: 10.1016/j.biotechadv.2008.11.002.
Chapter 5
SYSTEMATIC DESIGN OF A METAL ION BIOSENSOR:

A MULTI-OBJECTIVE OPTIMIZATION APPROACH
ABSTRACT
With the recent industrial expansion, heavy metals and other pollutants have
increasingly contaminated our living surroundings. Heavy metals, being non-degradable,
tend to accumulate in the food chain, resulting in potentially damaging toxicity to
organisms. Thus, techniques to detect metal ions have gradually begun to receive
attention. Recent progress in research on synthetic biology offers an alternative means for
metal ion detection via the help of promoter elements derived from microorganisms. To
make the design easier, it is necessary to develop a systemic design method for
evaluating and selecting adequate components to achieve a desired detection
performance. A multi-objective (MO) H2/H∞ performance criterion is derived here for
design specifications of a metal ion biosensor to achieve the H2 optimal matching of a
desired input/output (I/O) response and simultaneous H ∞ optimal filtering of intrinsic
parameter fluctuations and external cellular noise. According to the two design
specifications, a Takagi-Sugeno (T-S) fuzzy model is employed to interpolate several
local linear stochastic systems to approximate the nonlinear stochastic metal ion
biosensor system so that the multi-objective H2/H∞ design of the metal ion biosensor can
be solved by an associated linear matrix inequality (LMI)-constrained multi-objective
(MO) design problem. The analysis and design of a metal ion biosensor with optimal I/O
response matching and optimal noise filtering ability then can be achieved by solving the
multi-objective problem under a set of LMIs. Moreover, a multi-objective evolutionary
algorithm (MOEA)-based library search method is employed to find adequate
components from corresponding libraries to solve LMI-constrained MO H2/H∞ design
problems. It is a useful tool for the design of metal ion biosensors, particularly regarding
the tradeoffs between the design factors under consideration.
5.1. INTRODUCTION
Metal ion pollutants are commonly found in soil, water, and crops. With the recent
industrial expansion, wastewater containing heavy metal increasingly contaminates our living
surroundings (Ngah and Hanafiah 2008, Rezvani-Boroujeni et al., 2015, Singha and Guleria
2014, Teodosiu et al., 2014). Furthermore, non-degradable heavy metals may accumulate in
food chains, and the resulting toxicity damages organisms (Hartwig 1995, Davies 2005, Raha
and Robinson 2000). Hence, detection techniques have gradually begun to receive attention.
Recent progress in research on synthetic biology offers an alternative means for metal ion
detection via the help of promoter elements, such as PcusC and PpcoE derived from E. coli
(Munson et al., 2000) or PpbrA acquired from R.metallidurans (Taghavi et al., 1997, Hobman,
Julian, and Brown 2012). To make the design of detectors easier, it is necessary to develop a
method to evaluate and select adequate components for achieving a desired detection
performance.
In recent years, large numbers of genetic tools and engineering approaches have been and
still are being developed for metal ion biosensors. Synthetic biologists are forced to find
interchangeable parts, such as promoters, ribosome binding sites (RBSs), and regulatory
sequences, that can be validated as construction units and assemble devices. The ability to
quickly and reliably engineer biological systems from libraries of standard interchangeable
parts is one trademark of modern technology (Canton, Labno, and Endy 2008, Endy 2005,
Hasty, McMillen, and Collins 2002, Kaern, Blake, and Collins 2003, Cameron, Bashor, and
Collins 2014). Thus, to build a metal ion biosensor for a specified purpose, one may need a
systematic design process that begins with the specification, which states the desired goal and
technical details. Based on the specification, the biosensor is then represented by a block
diagram which consists of functional units of the system. At later stage the design is evaluated
and verified its feasibility via computational simulations and experimental validations until
the configuration and combination of biological parts reach suitable performance [16].
Although a great deal has been accomplished in a short time, engineering a metal ion
biosensor to produce a desired behavior still remains an acute problem, due to the
uncertainties and fluctuations at the molecular level (Bajic and Poyatos 2012, Fraser et al.,
2004, Blake et al., 2003, Elowitz et al., 2002, Murphy, Balazsi, and Collins 2007, Balazsi,
Murphy, and Collins 2007, Zhang, Chen, and Chen 2012, Hooshangi and Weiss 2006,
Pedraza and van Oudenaarden 2005, Chen and Wang 2006, Chen and Chang 2008, Thattai
and van Oudenaarden 2001).
Recently, applying the analysis of nonlinear stochastic molecular systems to evaluate the
flexibility of combinations of biological parts has been a subject of considerable interest. A
multi-objective H2/H∞ performance criterion is derived here for the design specifications of a
metal ion biosensor to achieve the H2 optimal tracking of a desired I/O response and H∞
optimal attenuation of parameter fluctuations and cellular noise simultaneously. Based on the
design specifications, the optimal design of the biosensor can be solved by an associated
Hamilton Jacobi inequality (HJI)-constrained optimization problem, which cannot be easily
achieved by present analytical or numerical methods. In order to simplify the analysis and
design of a nonlinear stochastic metal ion biosensor with multi-objective H2/H∞ performance,
a Takagi-Sugeno (T-S) fuzzy model is employed here to interpolate several local linear
stochastic systems to approximate the nonlinear stochastic metal ion biosensor system. This
allows the HJI-based design problem to be replaced by a linear matrix inequality (LMI)-based
design problem. Thus, the multi-objective H2/H∞ I/O response matching design of a
synthetic biosensor then can be achieved by solving a LMIs-constrained multi-objective
optimization problem.
However, there are tradeoffs between the H2 and H∞ performances. As natural selection
is an important mechanism in defining traits best suited to environmental change in the face
of evolutionary trade-offs (Ayala 2007), one question that arises is whether a similar strategy
Systematic Design of a Metal Ion Biosensor 85
could be adopted for multi-objective design problems. Inspired by biological evolution events,
such as mutation, crossover, and selection, a multi-objective evolutionary algorithm (MOEA)
is a method to determine non-dominated Pareto optimal solutions (Back, Hammel, and
Schwefel 1997, Deb et al., 2002). In particular, MOEA is useful when considering the
tradeoffs between design factors under consideration in multi-objective H2/H∞ design
problems. Consequently, according to the criterion required for the user-oriented
specifications, the design can be constructed by selecting adequate components with the help
of a multi-objective evolutionary algorithm (MOEA)-based searching method. In summary,
this chapter provides a systematic design method for developing next-generation synthetic
biology, from biological component selection to genetic circuit assembly. When the
component libraries are more complete, more precise detection for metal ion can be achieved.
Figure 5.1. A metal ion biosensor. The metal ion biosensor is assembled by selecting a set of promoter-
RBS components from the corresponding component libraries in Chapter 4, namely, a metal ion-
induced promoter-RBS component Mi from the component library in Table 4.1, a constitutive
promoter-RBS component Cj from the component library in Table 4.2, and a QS-dependent promoter-
RBS component Ak from the component library in Table 4.3.
The main focuses of this chapter are fourfold. (a) A nonlinear stochastic system is
introduced to model a metal ion biosensor with intrinsic parameter fluctuations and extrinsic
molecule noise. (b) A multi-objective (MO) H2/H∞ I/O response matching performance
criterion is derived to fit the design specification for a metal ion biosensor, which achieves the
H2 optimal tracking of a desired I/O response and H∞ optimal robust attenuation of
parameter fluctuations and cellular noise simultaneously. (c) By solving a LMIs-constrained
optimization problem, a metal ion biosensor can be constructed, which achieves the H2/H∞
multi-objective design by selecting adequate components from existing libraries. (d) The
proposed MOEA-based search method provides synthetic biologists with a useful tool for the
design of metal ion biosensors, particularly in the face of tradeoffs between the design factors
considered in next-generation synthetic biology.

For the convenience of description and explanation, as shown in Figure 5.1, the metal ion
biosensor is assembled by selecting a set of promoter-RBS components from the
corresponding component libraries in Chapter 4. The assembly included a metal ion-induced
promoter-RBS component Mi from the component library in Table 4.1, a constitutive
promoter-RBS component Cj from the component library in Table 4.2, and a quorum sensing
(QS)-dependent promoter-RBS component Ak from the component library in Table 4.3. The
metal ion-induced promoter-RBS Mi connects downstream with the LuxI coding sequence,
the production of which synthesizes a specific N-acylated homoserine lactone (AHL) as a
signal molecule (Hanzelka and Greenberg 1996, Val and Cronan 1998, Pearson et al., 1995).
The LuxR coding sequence is connected to the constitutive promoter-RBS component Ci.
When a sufficient amount of the LuxR protein is produced in the presence of AHL, AHL
binds to the LuxR protein to form a complex (Val and Cronan 1998, Engebrecht and
Silverman 1984, Nealson and Hastings 1979). The complex targets the cognate QS-dependent
promoter-RBS component Ak and thereby activates transcription of the green fluorescent
protein (GFP) coding sequence.
The dynamic model of the metal ion biosensor in Figure 5.1 can then be described as
follows:
 xE (t )  PM ( Pu ,i , Pl ,i , xS , I M )  (d  rE )  xE (t )

 xI (t )  axE (t )  (d  rI )  xI (t )
 (5.1)
 xR (t )  PC ( Pu , j ,0,0,0)  (d  rR )  xR (t )

 xG (t )  PA ( Pu , k , Pl ,k , xR , xI )  (d  rG )  xG (t )
G (t )  mx  (d  r )  G (t )
 G O
in which
PC ( Pu , j , 0, 0, 0)  Pu , j
Pu ,i  Pl ,i
 K SI 
1  
 xSI ( xS , I M ) 
Pu , k  Pl , k
 K RI 
1  
 xRI ( xR , xI ) 
xS xR
xSI ( xS , I M )  , xRI ( xR , xI ) 
 KM   KI 
1   1  
 IM   xI 
where xE, xI, xR, and xG denote the concentrations of autoinducer synthase, autoinducer,
transcriptional activator protein, and immature reporter protein, respectively, and G denotes
the intensity of GFP fluorescence. IM is the concentration of metal ions and xS is the total
concentration of the metal ion-dependent regulatory protein. xSI denotes the complex of xI and
IM, while xRI represents the complex of xR and xI. PM(Pu,i,Pl,i xS,IM), PC(Pu,i,0,0,0), and
PA(Pu,k,Pl,k,xR,xI) are the activities of the metal ion-induced promoter-RBS component, the
constitutive promoter-RBS component, and the QS-dependent promoter-RBS component,
respectively. Pu,i and Pl,i are the maximum and minimum promoter-RBS strengths of the ith
metal ion-induced promoter-RBS component in Table 4.1; Pu,j is the promoter-RBS strength
of the jth constitutive promoter-RBS component in Table 4.2, and Pu,k and Pl,k are the
maximum and minimum promoter-RBS strengths of the kth QS-dependent promoter-RBS
component in Table 4.3. rE denotes the degradation rate for autoinducer synthase, rI denotes
the degradation rate for the autoinducer itself, rR denotes the degradation rate for the
transcriptional activator protein, rG denotes the degradation rate for the immature reporter
protein, and rO denotes the degradation rate for the mature reporter protein. d is the dilution
rate due to cell growth. a is the autoinducer synthesis rate. m is the maturation rate for the
reporter protein. KSI and nSI denote the binding affinity and binding cooperativity between the
xSI complex and the corresponding promoter-RBS component, respectively. KM is the
dissociation rate between the metal ion IM and the metal regulatory protein xS. KRI and nRI are
the binding affinity and binding cooperativity between the xRI complex and the promoter-RBS
part, respectively. KI is the dissociation rate between the autoinducer xI and the transcriptional
activator protein xR.
However, biological components are inherently uncertain in a molecular biological
system. For example, the kinetic parameters of the components, including the processes of
transcription and translation, the degradation rates of regulatory proteins, dilution rates of the
cells, and the maturation rates for the reporter proteins, are all stochastically uncertain in vivo
as a result of gene expression noise from biochemical processes, thermal fluctuations, DNA
mutation, and evolution. Additionally, a synthetic gene circuit in vivo also suffers from
environmental molecular noise. Therefore, the equations in (5.1) should be modified as
follows:
 xE (t )   PM ( Pu ,i , Pl ,i , xS , I M )  (d  rE )  xE (t ) 
   
 xI (t )   axE (t )  (d  rI )  xI (t ) 
 xR (t )    PC ( Pu , j , 0, 0, 0)  (d  rR )  xR (t ) 
   
 xG (t )   PA ( Pu , k , Pl , k , xR , xI )  (d  rG )  xG (t ) 
 G (t )   mxG  (d  rO )  G (t ) 
   
(5.2)
 PM (Pu ,i , Pl ,i , xS , I M )  (d  rE )  xE (t )   v1 (t ) 
   
 axE (t )  (d  rI )  xI (t )   v2 (t ) 
  PC (Pu , j , 0, 0, 0)  (d  rR )  xR (t )  n(t )   v3 (t ) 
   
 PA (Pu , k , Pl , k , xR , xI )  (d  rG )  xG (t )   v4 (t ) 
 mxG  (d  rO )  G (t )   v (t ) 
   5 
where ΔPu,i, ΔPl,i, ΔPu,j, ΔPu,k, ΔPl,k, ΔrE, ΔrI, ΔrR, ΔrG, ΔrO, Δa, Δm, and Δd are the standard
deviations of the corresponding stochastic parameters and n(t) is Gaussian noise, which has a
mean of zero and unit variance, and accounts for sources of random fluctuation. The Gaussian
noise parameters vp, p = 1, 2, 3, 4, with a zero mean and variance of σp2, are molecular noise
for both the transcriptional and translational gene expression processes. v5 denotes molecular
noise in mature protein expression.
Consequently, the whole QS-based metal ion biosensor is expressed by (5.2), which can
also be represented by the more generalized nonlinear ordinary differential equation:
x(t )  f ( x(t ), S , I M )  f w ( x(t ), S , I M )n(t )  Hv(t )

(5.3)
y(t , S )  Cx(t )
where x(t) represents the state vector of the QS-based metal ion biosensor. y(t,S) is the output
vector. v(t) is extrinsic molecular noise from the environment. S = (Mi,Cj,Ak) is the set of
promoter-RBS components selected from the corresponding component libraries in Tables
4.1, 4.2, and 4.3. f(x(t),S,IM) is a smooth nonlinear function that characterizes the behavior of
the QS-based metal ion biosensor. fw(x(t),S,IM)n(t) is the intrinsic parameter fluctuations of
the QS-based metal ion biosensor. H denotes the noise-coupling matrix. C is the output
matrix. For the convenience of analysis and design of the QS-based metal ion biosensor
inserted into host cells, the nonlinear stochastic differential equation of metal ion biosensor in
(5.3) can be represented by the following Ito’s stochastic differential equation:
dx(t )   f ( x(t ), S , I M )  Hv(t )  dt  f w ( x(t ), S , I M )dw(t )

(5.4)
y(t , S )  Cx(t )
where w(t) is a standard Wiener process or Brownian motion with d (t) = n(t)dt to represent
the random parameter fluctuations of the synthetic gene circuit. In general, x(t) in (5.4) is
dependent on IM, i.e., the solution of (5.4) can be represented by x(t, IM). If the output y(t,S) is
the last state of metal ion biosensor, then C = [0,0,0,0,1].
The purpose of our design is to construct a metal ion biosensor by selecting a set of
suitable components from the corresponding libraries to achieve optimal matching of a
desired I/O response and minimize the effect of external disturbance and noise simultaneously
within a feasible range of metal ion concentrations, i.e., to achieve optimal H2 matching and
optimal H∞ disturbance filtering simultaneously. To achieve this, the following design
specifications are needed:
 A reference model with the desired I/O response to be matched by the metal ion
biosensor in (5.4) is given as follows:
dxr (t )   Ar xr (t )  r (t , I M )  dt
(5.5)
yr (t )  Cr xr (t )
where xr(t) is the desired reference state, yr(t) is the output vector of the desired reference
model, r(t,IM) represents a desired steady state trajectory for x(t), Ar is a matrix to be specified
for the transient behavior of xr(t), and Cr is the output matrix of the desired reference model.
In general, C = Cr. At the steady state, xr(t) = -Ar-1r(t,IM). If we set Ar = -I, then at the steady
state, xr(t,IM) = r(t,IM). Therefore, if the desired steady state xr(t,IM) of the metal ion biosensor
in (5.4) is set as r(t,IM) and we could select an adequate set S of components from the
corresponding libraries so that the stochastic dynamic equation (5.4) of the metal ion
biosensor could match the desired reference model in (5.5), i.e., at the steady state, xr(t) = -Ar-
1r(t,I ) and the I/O response is given by y = -C A -1r(t,I ).
M r r r M
 Standard derivations of molecular noise and parameter fluctuations in (5.2) to be

tolerated in vivo are specified in order to guarantee the robust design of the metal ion
biosensor.
 H2 design performance between the engineered biosensor output y in (5.4) and the
desired reference output yr is given by:
J 2 ( S )  E   yr (t , S )  y (t )  Q  yr (t , S )  y (t )  dt
T
(5.6)
= E  y (t , S )T Qy (t , S )dt
where Q is the weighting matrix and is the output of the following augmented system:
 
dx (t )  f ( x (t ), S , I M )  Hv (t ) dt  f w ( x (t ), S , I M )dw(t )
(5.7)
y (t , S )  Cx (t )
and
 x(t )   y (t )   v(t )  C 0   H 0
x (t )    , y (t )    , v (t )   , C   , H =  ,
 xr (t )   yr (t )   r (t )   0 Cr   0 I
 f (t , S , I M )   f w (t , S , I M )   Q Q 
f (t , S , I M )    , f w (t , S , I M )   , Q   
 Ar   0   Q Q 
Since the reference signal r(t) is treated as an uncertain external input by the designer, it
account for sources of noise.
 H∞ filtering performance to attenuate the effect of (t) on matching error is given as

follows:
E   yr (t , S )  y (t )  Q  yr (t , S )  y (t )  dt
T
J  (S ) 
E  v (t )T v (t )dt
(5.8)
E  y (t , S )T Qy (t , S )dt

Thus, if the H2 matching performance and H∞ filtering performance in (5.6) and (5.8) are
minimized simultaneously by choosing an appropriate set of components from the
corresponding libraries in Chapter 4, the engineered metal ion biosensor will then optimally
match the specified I/O response and optimally filter parameter fluctuations and
environmental disturbances simultaneously, i.e., to select a component set S from component
libraries in Chapter 4 to solve the following simultaneous minimization problem:
min  J 2 (S ), J  (S )  (5.9)
S
where J2(S) and J∞(S) are defined in (5.6) and (5.8), respectively. To make the design easier,
an indirect method is proposed by simultaneously minimizing the upper bounds of J2(S) and
J∞(S), i.e., the multi-objective problem in (5.9) is transformed to a suboptimal problem as
follows:
 ,    min  ,  
* *
S
(5.10)
subject to
 y (t, S ) Qy (t, S )dt  

T
J 2 (S )=E (5.11)
E  y (t , S )T Qy (t , S )dt
J  (S )   (5.12)
where α and β are the upper bounds of H2 and H∞ performances, respectively.

Remark 1: H2 performance in (5.6) can be considered as the penalty of the quadratic
matching error under the assumption (t) and α in (5.11) denotes the upper bound of H2
performance under the assumption (t) .
Remark 2: The inequality in (5.12) means that the effect of extrinsic molecular noise on
the matching error is less than β from an average energy point of view. Because the statistics
of extrinsic molecular noise may be unavailable or uncertain, it is very difficult to obtain the
noise filtering ability β* for all possible extrinsic noise (t) directly and only the upper bound
β of the noise-filtering ability β* can be given in (5.12) at first. Similarly, the upper bound α
of α* is also given. We will then decrease the upper bound (α, β) to as small a value as
possible to approach the lower bound (α*, β*), i.e., to get (α*, β*) by minimizing (α, β)
indirectly.
Remark 3: If the extrinsic environmental molecular noise (t) is deterministic, then the
expectation on (t) in (5.11) and (5.12) should be disregarded.
Remark 4: If the initial condition (0) is considered, then the noise-filtering upper bound
in (5.12) should be modified as follows:
 
E y (t , S )T Qy (t , S )dt  EV ( x (0))   E v (t )T v (t )dt (5.13)
for some Lyapunov function V( (0)), i.e., the energy due to the initial condition (0) should
be considered in the effect of noise (Chen and Zhang 2004, Zhang, Chen, and Tseng 2005).
Based on the multi-objective H2/H∞ design criterion, we obtain the following result for
the QS-based metal ion biosensor design.
Proposition 5.1: The multi-objective I/O matching problem in (5.10)–(5.12) is
equivalent to how to select components Mi, Cj, and Ak of the metal ion biosensor from the
corresponding component libraries in Appendix A to solve the following HJI-constrained
multi-objective problem:
 ,   
* *
min  ,  
M i ,C j , Ak
(5.14)
subject to
1  2V ( x )
x T C T QCx  f w ( x , S , I M )T fw (x , S, IM )
2 x 2
T
 V ( x ) 
  fw ( x , S, IM )  0
 x 
1  2V ( x )
x T C T QCx  f w ( x , S , I M )T fw (x , S, IM )
2 x 2
T T
 V ( x )  1  V ( x )  T  V ( x ) 
  f w ( x , S , I M )+  x  HH  x   0
 x  4    
with V( (t))>0 and EV( (0))<α, i.e., the I/O response of an engineered metal ion biosensor
will optimally match the specified I/O response of the reference model and optimally filter
intrinsic fluctuations and external disturbances simultaneously.
Proof: See Appendix A.

It is still very difficult to solve the constrained multi-objective minimization in (5.14) to
simultaneously achieve the optimal matching of the specified reference output in (5.5) and
optimal filtering of parameter fluctuations and environmental disturbances. Recently, the
fuzzy dynamic model has been widely used to interpolate local dynamic models to efficiently
approximate a nonlinear dynamic system (Takagi and Sugeno 1985, Chen, Tseng, and Uang
1999). Hence, in this situation, we employ the T-S fuzzy model to interpolate several linear
systems at different local operation points to efficiently and globally approximate the
augmented nonlinear system in (5.7) so that the design procedure for a multi-objective
optimal design of a synthetic metal ion biosensor can be simplified.
In this chapter, the T-S fuzzy method is employed to simplify the analysis and design
procedure for the QS-based metal ion biosensor under intrinsic parameter fluctuations and
extrinsic environmental molecular noise. The T-S fuzzy model is described by fuzzy if-then
rules. The pth rule of the fuzzy model for the augmented system in (5.7) is proposed in the
following form (Chen and Zhang 2004, Chen, Tseng, and Uang 1999, Takagi and Sugeno
1985):
Rule p : If z1 (t ) is Fp1 and z2 (t ) is Fp 2 and  and z g (t) is Fpg

then dx (t )   Ap x (t )  Hv (t )  dt  Awp x (t )dw(t ) (5.15)
y (t )  Cx (t )
for p = 1,2,…,L, where zg is the element of premise variables of the pth augmented system, i.e.,
z = [z1,…,zg]T, Fpg is the fuzzy set, and are the fuzzy system matrices, L is the number
of if-then rules, and g is the number of premise variables. The physical meaning of the fuzzy
rule p is that if the premise variables z1(t),…,zg(t) are with the fuzzy sets Fp1,…,Fpg, then the
augmented system in (5.7) can be represented by interpolating the linearized system in (5.15)
via the fuzzy basis. The fuzzy dynamics in (5.15) are denoted as follows (Chen, Tseng, and
Uang 1999, Tseng and Chen 2001, Tseng, Chen, and Uang 2001):

dx (t )    p ( z )  Ap x (t )  Hv (t )  dt  Awp x (t )dw(t ) 
L
p 1 (5.16)
y (t )  Cx (t )
in which
A 0 A 0
Ap   p , Awp   wp 
 0 Ar   0 0
where , Fpq(zq) is the grade of membership of

zq (t) in Fpq or the possibility function of zq (t) in Fpq, and μp is the fuzzy basis function for k =
1,2,…,L. The denominator in the above fuzzy basis function is only for
normalization so that the total sum of the fuzzy basis is . The physical
meaning of (5.16) is that the fuzzy stochastic system interpolates L local linear stochastic
systems through the nonlinear basis μp(z) to approximate the nonlinear stochastic system in
(5.7).
Remark 5: In (Takagi and Sugeno 1985), Takagi and Sugeno proposed a systematic
method to build a T-S fuzzy model for nonlinear function approximation by a system
identification tool, i.e., the local system matrices Ap and Awp in (5.16) can be identified by the
least square estimation method. Conversely, many studies have proved that the T-S fuzzy
model can approximate a continuous function to any degree of accuracy. However, there is
still some fuzzy approximation error in (5.16). In the design, for simplicity, the fuzzy
approximation error can be merged into the external noise.
After investigating the approximation of the nonlinear stochastic QS-based metal ion
biosensor by the fuzzy interpolation method, in order to avoid solving the nonlinear
constrained simultaneous optimization problem in (5.14) for the multi-objective design
problem of a QS-based metal ion biosensor under intrinsic parameter fluctuations and
extrinsic molecular noise, the measurement procedure for the matching and filtering abilities
of a QS-based metal ion biosensor could also be simplified by the fuzzy approximation
method. We then get the following result.
Proposition 5.2: Based on the T-S fuzzy model in (5.16), the H2/H∞ I/O response
matching problem in Proposition 1 becomes how to select promoter-RBS components Mi, Cj,
and Ak from the corresponding component libraries in Chapter 4 to solve the following multi-
objective problem:
 ,   
* *
min  ,  
M i ,C j , Ak
(5.17)
subject to
x (0)T Px (0)    0 (5.18)
C T QC  PAp  ApT P  AwpT PAwp  0 (5.19)
 C T QC  PAp  ApT P  AwpT PAwp PH 

 0 (5.20)
 HT P - 

for p = 1,2,…,L. Based on the optimal selection of these promoter-RBS components, the I/O
response of an engineered metal ion biosensor will achieve the optimal matching for the I/O
response of the specified reference model and the optimal filtering of parameter fluctuations
and environmental disturbances simultaneously.
Proof: See Appendix B.
Thus, the multi-objective H2/H∞ optimal I/O response design of the QS-based metal ion
biosensor obtained by solving the HJI-constrained multi-objective optimization problem in
(5.14) could be replaced by solving the following LMI-constrained multi-objective
optimizations:
 ,   
* *
min
( M i ,C j , Ak )
 ,  
(5.21)
subject to P  0 and LMIs in (5.18)-(5.20)
where Ω is the feasible set of promoter-RBS libraries in Chapter 4.

Remark 6: In this study, the fuzzy approximation method in (5.15) or (5.16) is only
employed to simplify the analysis and design procedure via solving P>0 for LMIs in (5.20)
instead of solving V( (t))>0 for HJIs directly. Further, based on the fuzzy interpolation of
local linear systems, i.e., replacing ( (t),S,IM) and ( (t),S,IM) by the fuzzy approximations
in (5.16), V( ) = TP is employed in in Proposition 5.2 to solve the HJI in Proposition 5.1.
The HJI in Proposition 5.1 is replaced with a set of LMIs in Proposition 5.2 and we only need
to solve P>0 for LMIs to guarantee the output of an engineered QS-based metal ion biosensor
that will optimally match the specified reference I/O response in (5.5) and optimally filter
parameter fluctuations and cellular disturbances simultaneously.
Remark 7: In general, it is very difficult to directly solve the LMI-constrained multi-
objective optimization in (5.21) for a synthetic gene circuit. In this study, a MOEA-based
library searching method is proposed to solve the LMI-based multi-objective I/O response-
matching problem in (5.21) for a metal ion biosensor in sequel. In general, no unique solution
exists such that α and β in (5.21) are minimized simultaneously. Therefore, more effort is
needed for the multi-objective optimization problem in (5.21) to seek a set of Pareto optimal
solutions, from which the designer can select the preferred option.
However, a problem remains with the tradeoff between H2 and H∞ performance. In light
of evolutionary trade-offs, the mechanism of natural selection produces traits best-suited for
adapting to environmental change. A similar strategy can be adapted for the multi-objective
design problem in (5.21). Inspired by biological evolution, a MOEA is a population-based
method to determine Pareto optimal solutions that are non-dominated. Compared with the
weighted sum method, MOEA is useful for considering multi-objective design problems, in
particular for assessing tradeoffs between design factors. Thus, before discussing the design
procedure of the multi-objective I/O response-matching problem in (5.21), some properties
regarding the Pareto optimal solutions are given as follows:
Definition 1: (Dominance) Consider two solutions (Mi1, Cj1, Ak1) and (Mi2, Cj2, Ak2) in Ω
for two objective values (α1, β1) and (α2, β2) subject to the LMIs in (5.18)–(5.20), respectively.
(α1, β1) is said to dominate (α2, β2), if α1 β1 and α2 β2.
Definition 2: (Pareto optimal solution) A solution (Mi*, Cj*, Ak*) is the Pareto optimal
solution of the multi-objective optimization problem in (5.21) with respect to Ω if another
feasible solution does not exist (Mi , Cj , Ak ) such that objective values (α , β )
dominate (α*, β*).
Definition 3: (Pareto front) The Pareto front for the optimization problem in (5.21) is
defined as Γ {(α*, β*)|(Mi*, Cj*, Ak*). This is the Pareto optimal solution of the optimization
problem in (5.21) and (α*, β*) is generated by (Mi*, Cj*, Ak*) subject to the LMIs in (5.18)–
(5.20)}.
The design procedure for a QS-based metal ion biosensor is then summarized as follows:
i. Provide user-defined design specifications as a desired reference model in (5.5)

for the quorum sensing-based metal ion biosensor.
ii. Select an initial set S of promoter-RBS components from corresponding libraries,
each of which can be satisfied with the LMIs in (5.18)–(5.20) with P>0.
iii. Sort the current set S into different fronts by Pareto dominance ranking and
assign a crowding distance to each of them.
iv. Create an offspring set S using MOEA operators, such as reproduction, crossover,
and mutation.
v. Calculate the objective values of the new set S obtained by natural selection.
Stop when the Pareto front is achieved or an acceptable solution is obtained.
Otherwise, create the next generation and return to step 3.
Remark 8: In addition to the design of a QS-based metal ion biosensor, the proposed
method can be applied to the design of synthetic gene regulatory networks with any kind of
dynamic behavior.
5.3. RESULTS
The design procedure begins by representing the nonlinear stochastic augmented system
of a metal ion biosensor and the desired reference model in (5.7) by the Takagi-Sugeno (T-S)
fuzzy model in (5.16) using the interpolation of linear stochastic systems. In particular, at
steady state, the desired fluorescence intensity of the metal ion biosensor to different metal
ion concentrations is described as follows:
5000
Gref ( I M )  65  (5.22)
1  101 I M 
2
According to (5.5), at steady state, our design goal for the steady state in (5.5) is xr(t) = –
Ar–1r(t,IM) and thereby yr = –CrAr–1r(t,IM). In order to let the steady state yr in (5.5) match
Gref(IM) in (5.22), if we select the followings for the reference model in (5.5)
Ar  -I , Cr   0,0,0,0,1 , r (t , I M )   0,0,0,0, G( I M ) 
T
(5.23)
then yr = Gref(IM) at the steady state of the reference model in (5.5). We suppose the quorum
sensing-based metal ion biosensor suffers from intrinsic parameter fluctuations, with zero
mean and unit variance, as well as the external environmental noises v1, v2, v3, and v4 for the
transcription and translation processes, and noise v5 for mature reporter protein expression,
are all Gaussian, with zero mean and unit variance. In order to then efficiently achieve the
desired I/O response matching design problem of the metal ion biosensor under intrinsic
parameter fluctuations and external disturbances, the multi-objective H2/H∞ matching design
in (5.9) is applied to the design problem. Based on the design procedure, a MOEA-based
library search method is employed to search a set S from corresponding libraries in Chapter 4
to minimize the objective values in (5.21) subject to P>0 and the LMIs in (5.18)–(5.20). From
the Pareto front in Figure 5.2, there are six Pareto solutions. The one with the red cross that
makes a compromise between the optimal H2 solution and H∞ solution is selected for the
multi-objective H2/ H∞ I/O response of the metal ion biosensor. In this design case, the
components from the corresponding libraries are found to be M1, C3, and A3. The desired
response is shown in Figure 5.3, with the fluorescence intensity values under different Cu(II)
ion concentrations. At steady state, the metal ion biosensor can match the desired I/O
response in (5.22), despite the parameter fluctuations and environmental disturbances.
Figure 5.2. Pareto front obtained by solving the multi-objective problem in (5.21) through the proposed
MOEA-based library search method from Tables 4.1–4.3.
Figure 5.3. The resulting metal ion biosensor. The adequate set S = (M1, C6, A3) is selected from the
corresponding libraries in Supplementary Chapter 4. The green points are the experimental results
(mean of three trials) by S = (M1, C6, A3). The gray solid line is the desired I/O response in (5.22).
5.4. DISCUSSION
With recent industrial expansion, heavy metal and other pollutants increasingly
contaminate our living surroundings (Rezvani-Boroujeni et al., 2015, Singha and Guleria
2014, Teodosiu et al., 2014). Heavy metals are non-degradable and may accumulate in food
chains, where the resulting toxicity can damage organisms (Davies 2005, Raha and Robinson
2000). Therefore, heavy metal detection techniques have gradually begun to receive attention.
In order to more easily design a QS-based metal ion biosensor, a multi-objective H2/H∞
performance criterion is derived to infer a sufficient condition required for user-oriented
specifications using a direct method by minimizing the upper bound of H2 and H∞
performance simultaneously. Based on the multi-objective design criterion, a metal ion
biosensor can then be designed by solving an associated HJI-constrained optimization
problem. However, the HJI-constrained optimization problem is difficult to solve directly by
any analytical or numerical method because of the complexity of nonlinear dynamics.
Therefore, a Takagi-Sugeno (T-S) fuzzy model is employed here to solve the HJI easily and
indirectly. The T-S fuzzy model has been widely applied to approximate nonlinear systems by
interpolating several local linearized stochastic systems. By using a T-S fuzzy model and
choosing an appropriate Lyapunov function, the HJI-constrained multi-objective optimization
problem in (5.14) for solving the H2/H∞ I/O response matching of a nonlinear stochastic metal
ion biosensor is reduced to an equivalent LMI-constrained multi-objective optimization
problem in (5.21), which can be solved efficiently by an MOEA algorithm with the help of
MATLAB’s LMI toolbox. Thus, according to the LMI-constrained criterion, the multi-
objective metal ion biosensor design can be constructed by evaluating and selecting adequate
promoter-RBS components from corresponding libraries within a feasible range of metal ion
concentrations.
However, because the multi-objective design problem has no unique solution, a problem
remains in dealing with the tradeoff between H2 and H∞ performance. In light of natural
selection on traits best-suited for environmental change being an important mechanism for
determining evolutionary trade-offs, a similar strategy seems to be adaptable for the multi-
objective design problem. Inspired by biological evolution, the MOEA is a population-based
method to determine non-dominated Pareto optimal solutions. Unlike the necessity for
complicated computations in conventional design strategies, only simple operators (e.g.,
selection, crossover, and mutation) and some simple calculations are required for the iterative
selection of adequate components. Therefore, MOEAs are useful when considering design
problems, in particular for assessing tradeoffs between the design factors under consideration.
Consequently, according to the user-specified criteria, this method may offer possible design
guidelines for selecting adequate components for a QS-based metal ion biosensor from the
corresponding libraries. When the component libraries are more complete, a more precise
detection performance of metal ion biosensor can be achieved. In fact, in addition to the QS-
based metal ion biosensor, the proposed method can be applied to the design of synthetic gene
regulatory networks with any kind of dynamic behavior.
5.5. CONCLUSION
In this chapter, a nonlinear stochastic system is introduced to model a synthetic metal ion
biosensor with intrinsic parameter fluctuations and extrinsic molecule noise. A multi-
objective H2/H∞ I/O response matching performance criterion is derived here for the design
specifications of the metal ion biosensor in order to simultaneously achieve the optimal H2
matching of the desired I/O behavior and the optimal H∞ filtering of parameter fluctuations
and cellular noise. An indirect method is proposed to solve the multi-objective H2/H∞ I/O
response matching design by minimizing their upper bounds simultaneously. Further, based
on a fuzzy interpolation technique, the HJI-constrained multi-objective design problem for the
metal ion biosensor is transferred to a more simple LMI-constrained multi-objective design
problem. According to the LMI-constrained multi-objective design criterion, a metal ion
biosensor can be constructed with a desired I/O response by evaluating and selecting adequate
components from the corresponding promoter-RBS libraries. The proposed MOEA-based
search method provides synthetic biologists with a useful tool for the design of gene circuits,
particularly in regards to tradeoffs between the design factors under consideration. The
experimental results verify that the design can optimally match the specified reference I/O
response and can optimally filter parameter fluctuations and environmental disturbances
simultaneously.
5.6. APPENDIX
Appendix A: Proof of Proposition 5.1
First, in the H2 design case with (t) , consider the following equality
 T dV ( x ) 
 
tf tf
E  y T Qydt  EV  x (0)   EV x (t f )  E   y Qy  dt dt (5.24)
0 0  
By Ito formula, we get
T
 V ( x ) 
dV ( x )  
 x 
  f (x, S, I M )dt  f w ( x , S , I M )dw 
(5.25)
1  2V ( x )
 f w ( x , S , I M )T f w ( x , S , I M )dt
2 x 2
Substituting (5.24) into (5.25) and by the fact Ed = 0, EV( (tf)) 0, we get
tf
E 
0
y T Qydt
tf  T T  V ( x ) 
T
 EV  x (0)   E   x C QCx    fw (x , S , IM ) (5.26)
0   x 

1  2V ( x ) 
 f w ( x , S , I M )T f w ( x , S , I M )  dt
2 x 2

Then, we get the following inequality
tf
E 
0
y T Qydt
tf  T T  V ( x ) 
T
 EV  x (0)   E   x C QCx    fw (x , S , IM ) (5.27)
0   x 

1  2V ( x ) 
 f w ( x , S , I M )T f w ( x , S , I M )  dt
2 x 2

By the first inequality in (5.14), we get
tf
E 
0
y T Qydt  EV  x (0)  (5.28)
By the fact
EV  x (0)    (5.29)
we can conclude
J 2 (S )   (5.30)
That is, if the inequalities (5.14) and (5.29) hold, then J2(S) is bounded by . Similarly,
following (5.24) and by Ito formula, we get
 f (x , S , I 
T
 V ( x ) 
dV ( x )  
 x 
 M 
)  Hv dt  f w ( x , S , I M )dw
(5.31)
1  2V ( x )
 f w ( x , S , I M )T f w ( x , S , I M )dt
2 x 2
By the fact
T T
 V ( x )  1  V ( x )  T  V ( x ) 
 x  Hv  4  x  HH  x    v v
T
(5.32)
     
Then, we get the following inequality
tf
E 
0
y T Qydt
tf  T T  V ( x ) 
T
 EV  x (0)   E 0
 x C QCx  

  x 
 fw ( x , S , I M )  v T v (5.33)
1  V ( x ) 
T
T  V ( x )  1 T  V (x )
2 
+  x  HH  x  2 w  f ( x , S , I ) f w ( x , S , I M )  dt
4    
M
x 2 

By the second inequality in (5.14) with V( (t))>0 and V( (0)) = 0, we get
tf tf
E 
0
y T Qydt  E  
0
v T vdt (5.34)
Then we obtain
tf
J  (S ) 
E  0
y T Qydt
 (5.35)
tf
E v T vdt
0
From (5.30) and (5.34), then the multi-objective problem in (5.10)-(5.12) becomes the
multi-objective problem in (5.14).
Appendix B: Proof of Proposition 5.2
First, we use the following fuzzy interpolation system

dx (t )    p ( z )  Ap x (t )  Hv (t )  dt  Awp x (t )dw(t ) 
L
p 1 (5.36)
y (t )  Cx (t )
in (5.16) to replace (5.7). Following the proof of Proposition 5.1 in Appendix A and by the
fact that EV( (tf)) 0, we get the following result
tf
E 
0
y T Qydt
 L   V ( x ) T

tf
 EV  x (0)   E   x T C T QCx   p ( z)  
  x  p
A x (t ) (5.37)
0 
 p 1 
1  2V ( x )  
 x (t )T AwpT Awp x (t )  dt 
2 x 2 
 
T
If we choose V( ) = P , then
V ( x )  2V ( x )
 2 Px ,  2P (5.38)
x x 2
then we get the following result from (5.37)
tf
E 
0
y T Qydt

L

tf
 Ex (0)T Px (0)  E x T C T QC  PAp  ApT P (5.39)
0
p 1

tf
 AwpT PAwp xdt  E 0
v T vdt
If
C T QC  PAp  ApT P  AwpT PAwp  0 (5.40)
then we get
tf
E 
0
y T Qydt  Ex (0)T Px (0) (5.41)
By the fact
Ex (0)T Px (0)   (5.42)
we can conclude
J 2 (S )   (5.43)
That is, if the inequalities (5.41) and (5.43) hold, then J2(S) is bounded by . Similarly,
by the following inequality
tf
E 
0
y T Qydt
 L   V ( x ) T

tf
 EV  x (0)   E   x T C T QCx   p (z)  
  x  p
A x (t ) (5.44)
0 
 p 1 
 V ( x ) 
T
1 T  V (x )
2  
+  Hv  x (t )T
A Awp x (t )  dt 
 x  2
wp
x 2  
 
and by the fact
T T
 V ( x )  1  V ( x )  T  V ( x ) 
 x  Hv  4  x  HH  x    v v
T
(5.45)
     
we get the following inequality
tf
E 
0
y T Qydt

L

tf
 Ex (0)T Px (0)  E x T C T QC  PAp  ApT P (5.46)
0
p 1
1  tf
+

PHH T P  AwpT PAwp  xdt  E 

0
v T vdt
If
1
C T QC  PAp  ApT P + PHH T P  AwpT PAwp  0 (5.47)

then we get
tf tf
E 
0
y T Qydt  Ex (0)T Px (0)  E 
0
v T vdt (5.48)
If (0) = 0, we get
tf tf
E 
0
y T Qydt  E  
0
v T vdt (5.49)
Then we obtain
tf
J  (S ) 

E
0
y T Qydt
 (5.50)
tf
E v T vdt
0
By Schur complement, the inequalities in (5.40), (5.42) and (5.47) are equivalent to the
LMIs in (5.18), (5.19), and (5.20), respectively, i.e., if the LMIs in (5.18), (5.19), and (5.20)
have a common solution P>0, then the output of an engineered multi-objective H2/H∞ QS-
based metal ion biosensor will optimally match the specified reference output and optimally
filter parameter fluctuations and cellular noise.
REFERENCES
Ayala, F. J. 2007. “Darwin”s greatest discovery: Design without designer.” Proceedings of
the National Academy of Sciences of the United States of America 104:8567-8573. doi:
DOI 10.1073/pnas.0701072104.
Back, T., U. Hammel, and H. P. Schwefel. 1997. “Evolutionary computation: comments on
the history and current state.” Evolutionary Computation, IEEE Transactions on 1 (1):3-
17. doi: 10.1109/4235.585888.
Bajic, D., and J. F. Poyatos. 2012. “Balancing noise and plasticity in eukaryotic gene
expression.” Bmc Genomics 13. doi: Artn 343. Doi 10.1186/1471-2164-13-343.
Balazsi, G., K. F. Murphy, and J. J. Collins. 2007. “Combinatorial promoter design for
engineering noisy gene expression.” Febs Journal 274:258-258.
Blake, W. J., M. Kaern, C. R. Cantor, and J. J. Collins. 2003. “Noise in eukaryotic gene
expression.” Nature 422 (6932):633-637. doi: DOI 10.1038/nature01546.
Nature Reviews Microbiology 12 (5):381-390. doi: DOI 10.1038/nrmicro3239.
Canton, B., A. Labno, and D. Endy. 2008. “Refinement and standardization of synthetic
biological parts and devices.” Nature Biotechnology 26 (7):787-793. doi: DOI
10.1038/nbt1413.
Chen, B. S., and Y. T. Chang. 2008. “A systematic molecular circuit design method for gene
networks under biochemical time delays and molecular noises.” Bmc Systems Biology 2.
doi: Artn 103. Doi 10.1186/1752-0509-2-103.
Chen, B. S., C. S. Tseng, and H. J. Uang. 1999. “Robustness design of nonlinear dynamic
systems via fuzzy linear control.” Ieee Transactions on Fuzzy Systems 7 (5):571-585.
Chen, B. S., and Y. C. Wang. 2006. “On the attenuation and amplification of molecular noise
in genetic regulatory networks.” BMC Bioinformatics 7. doi: 10.1186/1471-2105-7-52.
Chen, B. S., and W. H. Zhang. 2004. “Stochastic H(2)/H(infinity) control with state-
dependent noise.” Ieee Transactions on Automatic Control 49 (1):45-57. doi: Doi
10.1109/Tac.2003.821400.
Davies, M. J. 2005. “The oxidative environment and protein damage.” Biochimica Et
Biophysica Acta-Proteins and Proteomics 1703 (2):93-109. doi: DOI
10.1016/j.bbapap.2004.08.007.
Deb, K., A. Pratap, S. Agarwal, and T. Meyarivan. 2002. “A fast and elitist multiobjective
genetic algorithm: NSGA-II.” Evolutionary Computation, IEEE Transactions on 6
(2):182-197. doi: 10.1109/4235.996017.
Elowitz, M. B., A. J. Levine, E. D. Siggia, and P. S. Swain. 2002. “Stochastic gene
expression in a single cell.” Science 297 (5584):1183-1186. doi: DOI
10.1126/science.1070919.
Endy, D. 2005. “Foundations for engineering biology.” Nature 438 (7067):449-453. doi: DOI
10.1038/nature04342.
Engebrecht, J., and M. Silverman. 1984. “Identification of Genes and Gene-Products
Necessary for Bacterial Bioluminescence.” Proceedings of the National Academy of
Sciences of the United States of America-Biological Sciences 81 (13):4154-4158. doi:
DOI 10.1073/pnas.81.13.4154.
Fraser, H. B., A. E. Hirsh, G. Giaever, J. Kumm, and M. B. Eisen. 2004. “Noise minimization
in eukaryotic gene expression.” Plos Biology 2 (6):834-838. doi: ARTN e137. DOI
10.1371/journal.pbio.0020137.
Hanzelka, B. L., and E. P. Greenberg. 1996. “Quorum sensing in Vibrio fischeri: Evidence
that S-adenosylmethionine is the amino acid substrate for autoinducer synthesis.” Journal
of Bacteriology 178 (17):5291-5294.
Hartwig, A. 1995. “Current Aspects in Metal Genotoxicity.” Biometals 8 (1):3-11.
Hasty, J., D. McMillen, and J. J. Collins. 2002. “Engineered gene circuits.” Nature 420
(6912):224-230. doi: DOI 10.1038/nature01257.
Hobman, J. L., D. J. Julian, and N. L. Brown. 2012. “Cysteine coordination of Pb(II) is
involved in the PbrR-dependent activation of the lead-resistance promoter, PpbrA, from
Cupriavidus metallidurans CH34.” Bmc Microbiology 12. doi: Artn 109. Doi
10.1186/1471-2180-12-109.
Hooshangi, S., and R. Weiss. 2006. “The effect of negative feedback on noise propagation in
transcriptional gene networks.” Chaos 16 (2). doi: Artn 026108. Doi 10.1063/1.2208927.
Kaern, M., W. J. Blake, and J. J. Collins. 2003. “The engineering of gene regulatory
networks.” Annual Review of Biomedical Engineering 5:179-206. doi: DOI
10.1146/annurev.bioeng.5.040202.121553.
Journal of Bacteriology 182 (20):5864-5871. doi: Doi 10.1128/Jb.182.20.5864-
5871.2000.
Murphy, K. F., G. Balazsi, and J. J. Collins. 2007. “Combinatorial promoter design for
engineering noisy gene expression.” Proceedings of the National Academy of Sciences of
the United States of America 104 (31):12726-12731. doi: DOI 10.1073/pnas.0608451104.
Nealson, K. H., and J. W. Hastings. 1979. “Bacterial Bioluminescence - Its Control and
Ecological Significance.” Microbiological Reviews 43 (4):496-518.
Ngah, W. S. W., and M. A. K. M. Hanafiah. 2008. “Removal of heavy metal ions from
wastewater by chemically modified plant wastes as adsorbents: A review.” Bioresource
Technology 99 (10):3935-3948. doi: DOI 10.1016/j.biortech.2007.06.011.
Pearson, J. P., L. Passador, B. H. Iglewski, and E. P. Greenberg. 1995. “A 2nd N-
Acylhomoserine Lactone Signal Produced by Pseudomonas-Aeruginosa.” Proceedings of
the National Academy of Sciences of the United States of America 92 (5):1490-1494. doi:
DOI 10.1073/pnas.92.5.1490.
Pedraza, J. M., and A. van Oudenaarden. 2005. “Noise propagation in gene networks.”
Raha, S., and B. H. Robinson. 2000. “Mitochondria, oxygen free radicals, disease and
ageing.” Trends in Biochemical Sciences 25 (10):502-508. doi: Doi 10.1016/S0968-
0004(00)01674-1.
Rezvani-Boroujeni, A., M. Javanbakht, M. Karimi, C. Shahrjerdi, and B. Akbari-adergani.
2015. “Immoblization of Thiol-Functionalized Nanosilica on the Surface of Poly(ether
sulfone) Membranes for the Removal of Heavy-Metal Ions from Industrial Wastewater
Samples.” Industrial & Engineering Chemistry Research 54 (1):502-513. doi: DOI
10.1021/ie504106y.
Singha, A. S., and A. Guleria. 2014. “Chemical modification of cellulosic biopolymer and its
use in removal of heavy metal ions from wastewater.” International Journal of Biological
Macromolecules 67:409-417. doi: DOI 10.1016/j.ijbiomac.2014.03.046.
Taghavi, S., M. Mergeay, D. Nies, and D. vanderLelie. 1997. “Alcaligenes eutrophus as a
in Microbiology 148 (6):536-551. doi: Doi 10.1016/S0923-2508(97)88361-1.
Takagi, T., and M. Sugeno. 1985. “Fuzzy Identification of Systems and Its Applications to
Modeling and Control.” Ieee Transactions on Systems Man and Cybernetics 15 (1):116-
132.
Teodosiu, C., R. Wenkert, L. Tofan, and C. Paduraru. 2014. “Advances in
preconcentration/removal of environmentally relevant heavy metal ions from water and
wastewater by sorbents based on polyurethane foam.” Reviews in Chemical Engineering
30 (4):403-420. doi: DOI 10.1515/revce-2013-0036.
Thattai, M., and A. van Oudenaarden. 2001. “Intrinsic noise in gene regulatory networks.”
(15):8614-8619. doi: DOI 10.1073/pnas.151588598.
Tseng, C. S., and B. S. Chen. 2001. “H-infinity decentralized fuzzy model reference tracking
control design for nonlinear interconnected systems.” Ieee Transactions on Fuzzy
Systems 9 (6):795-809.
Tseng, C. S., B. S. Chen, and H. J. Uang. 2001. “Fuzzy tracking control design for nonlinear
dynamic systems via T-S fuzzy model.” Ieee Transactions on Fuzzy Systems 9 (3):381-
392.
Zhang, H., Y. L. Chen, and Y. Chen. 2012. “Noise Propagation in Gene Regulation Networks
Involving Interlinked Positive and Negative Feedback Loops.” Plos One 7 (12). doi:
ARTN e51840. DOI 10.1371/journal.pone.0051840.
Zhang, W. H., B. S. Chen, and C. S. Tseng. 2005. “Robust H-infinity filtering for nonlinear
stochastic systems.” Ieee Transactions on Signal Processing 53 (2):589-598. doi: Doi
10.1109/Tsp.2004.840724.
Chapter 6
SYSTEMATIC APPROACH TO ESCHERICHIA

COLI CELL POPULATION CONTROL
USING A GENETIC LYSIS CIRCUIT
ABSTRACT
Cell population control allows for the maintenance of a specific cell population
density. In this chapter, we use lysis gene BBa_K117000 from the Registry of Standard
Biological Parts, formed by MIT, to lyse Escherichia coli (E. coli). The lysis gene is
regulated by a synthetic genetic lysis circuit, using an inducer-regulated promoter-RBS
component. To make the design more easily, it is necessary to provide a systematic
approach for a genetic lysis circuit to achieve control of cell population density. Firstly,
the lytic ability of the constructed genetic lysis circuit is described by the relationship
between the promoter-RBS components and inducer concentration in a steady state
model. Then, three types of promoter-RBS libraries are established. Finally, according to
design specifications, a systematic design approach is proposed to provide synthetic
biologists with a prescribed I/O response by selecting proper promoter-RBS component
set in combination with suitable inducer concentrations, within a feasible range. This
chapter provides an important systematic design method for the development of next-
generation synthetic gene circuits, from component library construction to genetic circuit
assembly. In future, when libraries are more complete, more precise cell density control
can be achieved.
6.1. INTRODUCTION
Cell population control is a method of regulating cell density to maintain a self-sustaining
population. Self-regulating mechanisms of cell population control have existed in nature for a
long time. For example, the spore-forming bacterium Bacillus subtilis delays sporulation
under nutrient-limited conditions by killing non-sporulating siblings and feeding on the dead
cells to support spore formation (Ellermeier et al., 2006). Colicin-producing bacteria produce
bacteriocins to destroy nearby competitors through amino-acid starvation or DNA damage
(Gardner, West, and Buckling 2004). Recently, a programmed control of cell population is
proposed, using an endogenous genetic regulatory circuit (You et al., 2004). However, to
make genetic lysis circuit design more easily, it is necessary to provide a systematic approach
to achieve control of cell population density.
Naturally occurring lytic systems have the ability to trigger host cell lysis with specific
proteins under certain circumstances, such as in the presence of antibiotics or competitors, or
under conditions of amino-acid starvation. Some species of bacteria and bacteriophage have
lytic protein that induces cell lysis. For example, expression of cloned T4 phage genes e or t
can be used to the disrupt E. coli cells (Morita et al., 2001, Pasotti et al., 2011). Lysis by the
T4 phage usually requires two gene products, transcribed by genes e and t, respectively. Gene
e encodes a lysozyme, Gpe; gene t encodes a holin, Gpt. The expression of gene e weakens E.
coli cell wall but does not lead to cell disruption, while the expression of gene t enables to
induce cell lysis. Another example of a lytic protein is colicin E7 (ColE7), which is encoded
by E. coli. Expression of ColE7 is regulated by an SOS response operon. The SOS response
operon is composed of the ceaE7, ceiE7 and celE7 genes, the products of which are ColE7,
ImE7 and LysE7, respectively (Chak, Kuo, and James 1991). ColE7 can be neutralized
through the formation of a protein complex with the immunity protein ImE7. The ColE7–
ImE7 complex is exported from the colicin-producing cells by the lysis protein LysE7, which
induces lysis of the host cell through break down of the cell membrane (Cavard et al., 1987,
CHAK and JAMES 1984, Inukai et al., 1984, PUGSLEY and COLE 1987, Cavard 1991,
Ellis, Wang, and Collins 2009).
Lysis protein is able to activate the outer membrane phospholipase A (OMPLA), which
allows colicin to cross the cell envelope and enter into the medium. Colicins are plasmid-
encoded bacteriocins produced by E. coli that have antibiotic-like activity against closely
related bacteria (Cascales et al., 2007, Ellis, Wang, and Collins 2009). Release of lysis protein
can therefore control the population density of E. coli. In this study, lysis protein expression
level in the genetic lysis circuit relies on tunable genetic components and inducer
concentrations. The degree of cell lysis could therefore be fine-tuned by changing external
inducer signals. Different inducer concentrations could result in different fates of the cells, for
example severe death, modulated death, or slow growth. To make the design more easily, the
lytic ability of the constructed genetic lysis circuit is described by the relationship between
the promoter-RBS components and inducer concentrations in a steady state model. According
to user-oriented specifications, the genetic lysis circuit can be constructed via selecting
adequate promoter-RBS components in combination with a feasible range of inducer
concentrations. In general, it requires long computation times when component libraries
become large. Hence, a genetic algorithm (GA) search method is proposed to save time in
evaluating and selecting promoter-RBS components.
This chapter provides an important systematic design method for the development of
next-generation synthetic biology, from component library construction to genetic circuit
assembly. When libraries are more complete, more precise cell density control for genetic
lysis circuit can be achieved. The mechanism of cell lysis can be used to release useful
macromolecules that cannot pass through cell membranes (Morita et al., 2001, Lo et al.,
2013). For example, an operon encoding bcsA, bcsB, bcsC, and bcsD is required for bacterial
cellulose synthesis (bcs) in Acetobacter xylinum (Ross, Mayer, and Benziman 1991). The
expression of the operon can transform redundant glucose into cellulose to maintain intestinal
peristalsis. In addition, an appropriate amount of cellulose has a proven role in preventing
obesity. Cellulose, however, is a macromolecule that normally cannot pass through the cell
membrane. A genetic lysis circuit can be combined with a bacterial cellulose synthesis system
Systematic Approach to Escherichia Coli Cell Population Control … 109
to promote cellulose release. In future, genetic lysis circuits may apply to development in, for
example, drug discovery, metabolic control, and therapeutic treatment, with the help of the
proposed design methodology.
The main focuses of this chapter are threefold: (1) Based on promoter-RBS kinetic
strengths, we establish three kinds of promoter-RBS libraries. (2) Inducible promoter-RBS
components are used to construct genetic lysis circuits with different lytic abilities for cell
population control. (3) The proposed GA-based systematic searching method could provide
synthetic biologists with a useful tool to design synthetic genetic lysis circuits.
6.2. METHODS
Construction of Genetic Lysis Circuit
Two type of synthetic genetic lysis circuits are shown in Figure 6.1(a) and (b), with
different external inducers I1 and I2. In Figure 6.1(a), a constitutive promoter continuously
produces the repressor protein xr1. The protein xr1 represses the downstream promoter and
reduces expression of the lysis gene. The inducer I1 can bind to the repressor protein and
prevent binding to the promoter operon, enhancing expression of the lysis gene. In Figure
6.1(b), a constitutive promoter continuously produces the activator protein xa1. The protein xa1,
however, needs to form a complex with the inducer I2. The complex constitutes a quorum
sensing mechanism. When this complex accumulates, it activates the downstream promoter
and enhances expression of the lysis gene.
(a)
(b)
Figure 6.1. Genetic lysis circuit for controlling cell population density. (a) Inducible repressor-regulated
circuit with lysis gene in E. coli. (b) Inducible activator-regulated circuit with lysis gene in E. coli.
Dynamic Model of Genetic Lysis Circuit
Before we introduce the dynamic model of a genetic lysis circuit, a normal equation for
cell growth, without the lysis protein, is as follows:
 N (t ) 
N (t )  k  N (t ) 1   (6.1)
 N max 
where N and Nmax denote cell population density (O.D. 600) and maximum cell population
density, respectively, and k denotes the dilution rate due to cell growth. Since our study
utilizes the lysis protein to regulate cell population density, the relationship between cell
density and the concentration of the lysis protein can be described in the following equation:
 N (t ) 
N (t )  k  N (t ) 1     N  N (t )  xl (t ) (6.2)
 N max 
where x1 and γN denote the concentration of the lysis protein and the lysis rate of the lysis
protein, respectively. Under normal growth conditions, cells do not suffer from the toxicity of
the lysis protein. We used the endogenous gene (BBa_K117000) encoded in E. coli to control
cell population density. The lysis protein can activate OMPLA to cause cell lysis and reduce
cell density (Cascales et al., 2007, Ellis, Wang, and Collins 2009). From equation (6.2), it can
be seen that the concentration of lysis protein x1 is the negative regulator of cell population
density. Because lysis ability is related to the concentration of lysis protein x1 and the toxicity
γN of the lysis protein, we construct a dynamic model of the genetic lysis circuit to regulate
the concentration of lysis protein x1, and thus lysis ability.
To construct the dynamic model, promoter-RBS regulation must be introduced. We first
define the promoter-RBS regulation function P(Pu, P1, TF, I), in which Pu and P1 denote the
maximum and minimum promoter-RBS strengths, respectively, TF denotes transcription
factor concentration, and I denotes inducer concentration. This function describes the
biochemical aspect of the transcription and translation process. The details of the function
(the mathematical model) can be found in Appendix. The dynamic model of the genetic lysis
circuit with the repressor-regulated promoter-RBS component in Figure 6.1(a) thus is
described as follows:
 xr1 (t )  Pc1 ( Pu ,i ,0,0,0)  ( x  k )  xr1 (t )  1 (t ) , i  1,..., C

 r1
 xl 2 (t )  Pr 2 ( Pu , j , Pl , j , xr1 , I1 )  ( xl 2  k )  xl 2 (t )  2 (t ) , j  1,..., R

 N (t )  k  N (t ) 1  N (t ) N max    N  N (t )  xl 2 (t )  3 (t ) (6.3)
Pu , j  Pl , j xr1
Pr 2 ( Pu , j , Pl , j , xr1 , I1 )  Pl , j  , xr*1 ( xr1 , I1 ) 
 x* ( x , I ) 
nr 2
 I 
1   r1 r1 1  1  1 
 KI
 Kr 2   1 
where (i, j) denote the ith constitutive promoter-RBS component and the jth repressor-
regulated promoter-RBS component, respectively. Pc1(Pu,j,0,0,0) denotes the promoter-RBS

regulation activity of the first stage of the constitutive promoter-RBS components. Pu,i
denotes the maximum promoter-RBS strength of the ith constitutive promoter-RBS
component. Pr2(Pu,j,Pi,j,xr1,I1) denotes the promoter-RBS regulation activity of the second
stage of the repressor-regulated components. Pu,j and Pi,j denote the maximum and minimum
promoter-RBS strengths of the jth repressor-regulated promoter-RBS component, respectively.
I1 denotes the concentration, and xr1 denotes the concentration of the repressor. Kr2 and nr2
denote the binding affinity and binding cooperativity between the repressor xr1* and the
corresponding promoter-RBS component in the second stage, respectively. KI1 denotes the
dissociation rate between the inducer I1 and the repressor xr1. γN is the lysis rate of the lysis
protein. ω1(t) and ω2(t) denote cellular noise in the transcriptional and translational processes,
respectively. Finally, ω3(t) denotes cellular noise in the cell population density.
From equation (6.3), the change in concentration of repressor protein xr1 is due to the
difference between the protein generation rate Pc1(Pu,j,0,0,0) and the protein degradation rate
γxr1, in combination with the dilution rate k. The repressor protein xr1 binds to the second stage
of repressor-regulated components and reduces the regulation activity Pr2(Pu,j,Pi,j,xr1,I1). The
inducer I1 can remove the inhibitory effect of the repressor and increase the regulation activity
Pr2(Pu,j,Pi,j,xr1,I1). Similarly, the change in concentration of lysis protein xl2 is due to the
difference between the protein generation rate Pr2(Pu,j,Pi,j,xr1,I1) and the protein degradation
rate γxl2, in combination with the dilution rate k. Cell population density is regulated by the
concentration of lysis protein xl2. Lytic ability is therefore controlled by the regulation activity
Pr2(Pu,j,Pi,j,xr1,I1), and the four regulated factors Pu,j, Pi,j, xr1, and I1 can be used to control cell
population density by selecting appropriate promoter-RBS components and inducer
concentrations.
The only difference between Figures 6.1(a) and (b) is the replacement of a repressor-
regulated promoter-RBS component with an activator-regulated promoter-RBS component.
The dynamic model of the genetic lysis circuit in Figure 6.1(b) can be described as follows:
 xa1 (t )  Pc1 ( Pu ,i ,0,0,0)  ( x  k )  xa1 (t )  1 (t ) , i  1,..., C

 a1
 xl 2 (t )  Pa 2 ( Pu , m , Pl , m , xa1 , I 2 )  ( xl 2  k )  xl 2 (t )  2 (t ) , m  1,..., A

 N (t )  k  N (t ) 1  N (t ) N max    N  N (t )  xl 2 (t )  3 (t ) (6.4)
Pu , m  Pl , m xa1
Pa 2 ( Pu , m , Pl , m , xa1 , I 2 )  Pl , m  , xa*1 ( xa1 , I 2 ) 
 K 
na 2
 KI 
1   * a2  1  2 
 xa1 ( xa , I 2 )   I2 
where (i, m) denote the ith constitutive promoter-RBS component and the mth activator-
regulated promoter-RBS component, respectively. I2 denotes the concentration of the inducer,
xa1 denotes the concentration of the activator protein, and Ka2 and na2 denote the binding
affinity and the binding cooperativity between activator xa1 and the corresponding promoter-
RBS component in the second stage, respectively. Finally, KI2 denotes the dissociation rate
between inducer I2 and activator xa1.
The dynamic models for the genetic lysis circuit are then transformed into steady-state
models by assuming the derivatives of the dynamic models in (6.3) and (6.4) are equal to zero.
The steady-state concentrations xr1SS, xa1SS, and xl2SS of the repressor protein, activator protein,
and lysis protein, respectively, as well as the steady state cell population density NSS are
obtained as follows:
 Pc1 ( Pu ,i ,0,0,0)
 xr1ss   v1 , i  1,..., C
  xr 1  k
 Pr 2 ( Pu , j , Pl , j , xr1ss , I1 )
 xl 2 ss   v2 , j  1,..., R (6.5)
  xl 2  k

 N ss  N max 1   N xl 2 ss   v3
  k 
for the genetic lysis circuit with repressor-regulated promoter-RBS component in Figure
6.1(a), and
 Pc1 ( Pu ,i ,0,0,0)
 xa1ss   v1 , i  1,..., C
  xa1  k
 Pa 2 ( Pu , m , Pl , m , xa1ss , I 2 )
 xl 2 ss   v2 , m  1,..., A (6.6)
  xl 2  k

 N ss  N max 1   N xl 2 ss   v3
  k 
for the genetic lysis circuit with activator-regulated promoter RBS component in Figure
6.1(b). In these equations, the Gaussian noise parameter vi, i = 1, 2, with a zero mean and
variance of σi2, denotes cellular noise for both the transcriptional and translational gene
expression processes in the steady state. v3 denotes cellular noise in cell population density in
the steady state. In (6.5) and (6.6), the steady state population density NSS changes depending
on inducer concentration, constitutive promoter-RBS components, repressor-regulated
components, and activator-regulated promoter-RBS components, i.e., i = 1,…,C, j = 1,…,R or
i = 1,…,C, m = 1,…,A. In this study, we select values for these components from promoter-
RBS libraries (see Appendix) to achieve a specified I/O reponse with the genetic lysis circuit.
The details of the construction procedure for promoter-RBS component libraries can see in
Appendix.
In general, biological components are inherently uncertain in a molecular biological
system. Hence, parameter uncertainties in equations (6.5) and (6.6) must be taken into
consideration. For example, the kinetic parameters of the promoter-RBS components
including the processes of transcription and translation, the degradation rates of regulatory
proteins, dilution rates of the cells, and the lysis rates of the lysis proteins, are all
stochastically uncertain in vivo, as a result of gene expression noise from biochemical
processes, thermal fluctuations, DNA mutation, parameter estimation errors, and evolution
(Alon 2007). These are defined as follows:
Pu ,i  Pu ,i  Pu ,i n1  t  ,
Pu , j  Pu , j  Pu , j n2  t  , Pl , j  Pl , j  Pl , j n2  t  ,
Pu , m  Pu , m  Pu ,m n2  t  , Pl ,m  Pl ,m  Pl ,m n2  t  ,
 xr1   xr1   xr1n1  t  ,  xa1   xa1   xa1n1  t  , (6.7)
 xl 2   xl 2   xl 2 n2  t  ,
k  k  kn3  t  ,
 N   N   N n3  t 
where ΔPu,i, ΔPu,j, ΔPl,j, ΔPu,m, ΔPl,m, Δγxr1, Δγxa1, Δγxl2, ΔγN, and Δk denote the standard
deviations of the corresponding stochastic parameters, and ni(t), i = 1,2,3 denote Gaussian
noise, have zero mean and unit variance, and account for random fluctuation sources. Thus,
ΔPu,i, ΔPu,j, ΔPl,j, ΔPu,m, ΔPl,m, Δγxr1, Δγxa1, Δγxl2, ΔγN, and Δk denote the deterministic aspects
of parameter variation, and ni(t), i = 1,2,3 denote the random fluctuation sources. The kinetic
parameters in the steady state model in equations (6.5) and (6.6) are replaced by the
parameter perturbations shown in (6.7) for robust design of the genetic circuit. These
parameter fluctuations must be considered in the design procedure so that the synthetic
genetic circuit can tolerate fluctuations in vivo (Chen et al., 2011).
Design Specifications for the Genetic Lysis Circuit
The purpose of our design is to construct a genetic lysis circuit by selecting a set of
suitable promoter-RBS components from the corresponding libraries in combination with a
feasible range of inducer concentrations, to achieve optimal tracking of a desired I/O response.
To achieve this, the following design specifications are needed:
 Desired I/O response Nref(I) of the genetic lysis circuit.

 Well-characterized promoter-RBS component libraries and a feasible range of
inducer concentrations.
 Standard derivations of parameter fluctuations and environmental disturbances to be
tolerated in vivo.
 A cost function between the steady state cell population density NSS in equations (6.5)
and (6.6) and the desired reference steady state cell population density Nref as follows:
J (S , I )  E   N ss (S , I )  N ref ( I )  dI
2
(6.8)
where S denotes the set of promoter-RBS components i and j (or i and m) selected from the
corresponding component libraries, i.e., S = (i, j) for constitutive promoter-RBS components
and repressor-regulated promoter-RBS components in (6.5), and S = (i, m) for constitutive
promoter-RBS components and activator-regulated promoter-RBS components in (6.6). The
inducer concentration I for I1 is S = (i, j) and I2 is S = (i, m).
If the cost function in equation (6.8) is minimized by choosing the most appropriate set of
components in combination with a feasible range of inducer concentrations under design
specifications, the cell population density of a engineered genetic lysis circuit will optimally
match the specified steady state cell population density Nref(I) under intrinsic parameter
fluctuations and environmental disturbances. Although the cost function can be minimized by
traditional conventional search methods, the combination of promoter-RBS components with
inducer concentrations to minimize J(S, I) will require long computation times as well as
trial-and-error experimentation when component libraries become large. Hence, the more
effective and efficient genetic algorithm (GA) search method is proposed to save time in
evaluating and selecting promoter-RBS components. Since genetic algorithm searches for the
optimal solution, on basis of the maximum fitness, which is inversely proportional to the
minimum error in (6.8), we need to define the fitness function as follows
1
F (S , I )  (6.9)
J (S , I )
Therefore,
1
max F ( S , I )  (6.10)
(S ,I ) min J ( S , I )
(S ,I )
Summary of Design Procedure for the Genetic Lysis Circuit
1. Choose a species for cell population density control. In this study, we used E. coli.
2. Provide user-defined design specifications for the genetic lysis circuit.
3. Select an initial set S of promoter-RBS components and inducer concentrations.
4. Calculate the fitness function F(S, I) in equation (6.9) for each set S of promoter-RBS
components and inducer concentrations.
5. Create an offspring set S using GA operators such as reproduction, crossover, and
mutation.
 Make copies of possible solutions, on basis of their fitness.
 Swap values between two possible solutions.
 Randomly alter the value in a possible solution.
6. Calculate the cost value of the new set S obtained by natural selection. Stop when the
design goal is achieved or an acceptable solution is obtained. Otherwise, create the
next generation and return to step 5.
6.3. RESULTS
Genetic Lysis Circuit Design Example for Cell Population Density Control
In Silico and Verification Via Experiment In Vivo
For the convenience of description and explanation, as shown in Figure 6.2, a genetic
lysis circuit is assembled by selecting a set of promoter-RBS components, namely, a
constitutive promoter-RBS component Ci from Table 6.1 in Appendix and an activator-
regulated promoter-RBS component Am from Table 6.3 in Appendix. The lysis gene is
embedded downstream of the activator-regulated promoter-RBS component. The genetic lysis
circuit is divided into two stages. The first stage involves a constitutive promoter-RBS
component Ci for producing regulatory protein, LuxR. The second stage involves an
activator-regulated promoter-RBS component Am for driving the expression of the lysis
protein. The external inducer AHL is used to regulate the lysis activity. The desired
population response Nref(I) to different inducer concentrations described as follows:
0.6
N ref ( I )  0.1  (6.11)
1 2 I 2
tolerated in vivo are given by
Pu ,i  0.05Pu ,i , Pu ,m  0.05Pu ,m , Pl ,m  0.05Pl ,m ,  xa1  0.05 xa1 ,

(6.12)
 xl 2  0.05 xl 2 , k  0.05k ,  N  0.05 N
as well as the environmental noise parameters v1 and v2, for transcription and translation
processes, and v3 for cell population density, are all Gaussian, with zero mean and unit
variance. In order to efficiently solve the constrained optimal matching design problem of the
genetic lysis circuit, a GA-based library search method is employed to search a set S from
corresponding libraries in Appendix to minimize the following cost function:
J (S , I )  E   N ss (S , I )  Nref ( I )  dI , I  0.1, 10 (nM)

2
(6.13)
Then, to minimize the cost function (6.13), the adequate promoter-RBS components from
the corresponding libraries are found to be C6 and AL2. The desired response is shown in
Figure 6.3, with the experimental steady state O.D. 600 values taken from Figure 6.4 under
different inducer concentrations at 240 min. Clearly, the cell population density of the genetic
lysis circuit can robustly match the desired I/O response despite the intrinsic parameter
fluctuations and environmental disturbances. If the desired cell density, for example, is 0.5,
the suitable AHL concentration of 0.5nM can be taken, based on the prescribed I/O response.
The experimental result of 0.487 is observed, with percentage error of 2.6%, confirming that
the prescribed cell population density is well controlled by the proposed lysis circuit.
Similarly, when the desired cell density is 0.3, the suitable AHL concentration of 1nM can be
taken, based on the prescribed I/O response. The experimental result of 0.311 is observed,
with percentage error of 3.6%, again indicating that the prescribed cell population density is
well controlled by the proposed genetic circuit.
Figure 6.2. Inducible LuxR-regulated circuit with lysis gene in E. coli. A constitutive promoter
continuously produces the activator protein LuxR. The protein LuxR needs to form a complex with the
inducer AHL. The LuxR-AHL complex constitutes a quorum sensing mechanism. It activates the
downstream promoter and enhances expression of the lysis gene when this complex accumulates.
0.8
0.7
Cell Population Density (OD600)
0.6
0.5
0.4
0.3
0.2
0.1
0
-1 0 1
10 10 10
AHL Concentration (nM)
Figure 6.3. The simulation and experiment results of a synthetic genetic lysis circuit. By minimizing the
cost function in (6.13) for the genetic lysis circuit in Figure 6.2, the adequate set S = (C6, A12) is selected
from the corresponding libraries in Appendix. The green points are the experimental results by S =
(C6,A12). The gray solid line is the desired I/O response.
AHL = 0.1nM AHL = 0.2nM

0.8 0.8
0.7 0.7

0.6 0.6
0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0.1 0.1
0 0
0 50 100 150 200 0 50 100 150 200
(a) (b)
0.8 0.8
0.7 0.7

0.6 0.6
0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0.1 0.1
0 0
0 50 100 150 200 0 50 100 150 200
(c) (d)
0.8 0.8
0.7 0.7
0.6 0.6
0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0.1 0.1
0 0
0 50 100 150 200 0 50 100 150 200
(e) (f)
Figure 6.4. Population density time profiles under different inducer concentrations of AHL. The green
points are the experimental results.
6.4. DISCUSSION
In this chapter, we focus on the design of synthetic genetic lysis circuit to achieve density
control of cell population. A mathematical model is introduced to describe the dynamic and
steady state regulatory behavior of the genetic lysis circuit. From the steady state
mathematical models in (6.5) or (6.6), we find that if we want to control cell population
density, we need to control lysis ability, which is related to transcriptional and translational
processes. The promoter allows the RNA-polymerase molecules to latch onto a DNA strand
and initialize the transcription of a downstream gene into mRNA, and the RBS allows the
ribosome to bind and translate the mRNA. Hence, in this study, the promoter combined with
the RBS is viewed as a genetic component for regulating lytic ability.
For convenient measurement of promoter-RBS components, we use GFP as a reporter to
identify characterizations of promoter-RBS components. We can therefore use the parameters
identified in Appendix as component libraries to design the genetic lysis circuit. We also
provide the design specifications for genetic lysis circuit for use by synthetic biologists. With
a systematic approach, we can control the desired cell population density by selecting
appropriate promoter-RBS component set in combination with a feasible range of inducer
concentrations. In this study, for validation of the proposed design method, we perform some
experiments with proper promoter-RBS components and inducer concentrations. The
experimental results show that the simulation can robustly predict actual cell population
density.
The precise control of cell population density with protein release is useful in medicine to
cure disease (Saeidi et al., 2011), because changing medicine dosages may not achieve the
desired effects. Additionally, evidence from systems biology indicates that apoptosis is
involved in many pathways causing cell death (Elmore 2007, Green 2011). Each pathway
depends on expression levels and can trigger cell death. For example, we embed a genetic
lysis circuit into E. coli. Through the control of different expression levels of repressor-
regulated or activator-regulated promoter-RBS components, a simple genetic lysis circuit can
regulate cell death. This bottom-up design approach can potentially be extended to other
complicated processes involving programmed cell death, which could help us to understand
systematic phenomena that have always existed in nature.
The construction procedure for cell population density control is very important for
engineering a more complex synthetic genetic lysis circuit. We characterize the genetic circuit
and provide the desired cell population density beforehand. Using GA, we search for an
appropriate set of promoter-RBS components in combinations with a feasible range of
inducer concentrations to achieve the desired cell population density. The GA provides a
useful tool in the construction of genetic lysis circuit for control of cell population density. In
future, when libraries are more complete, more effective and efficient design methods for
synthetic genetic lysis circuits may aid developments in drug discovery, metabolic control,
and therapeutic treatment.
6.5. CONCLUSION
In this chapter, we engineer a genetic lysis circuit to control cell population density.
Inducible promoter-RBS components are used to construct genetic lysis circuits with different
lytic abilities for cell population control. Moreover, we provide four design specifications for
cell population density control, allowing designers to select proper promoter-RBS
components from libraries for the synthetic genetic lysis circuits. The design problem can be
transformed by selecting proper promoter-RBS components in combination with a feasible
range of inducer concentrations to achieve a desired I/O response. The proposed GA-based
systematic searching methodology could provide synthetic biologists with a useful tool to
design synthetic genetic lysis circuits. From the experimental results, we find that the data are
close to the prescribed cell densities. Therefore, we believe that the proposed user-oriented
design method for cell population density control will provide a useful guide in the rapidly
growing field of synthetic biology.
6.6. APPENDIX
Table 6.1. The constitutive promoter-RBS library
Index BioBrick component Pu

C1 J23101-B0031 94.520
C2 J23101-B0032 65.892
C3 J23101-B0034 237.000
C4 J23105-B0031 14.493
C5 J23105-B0032 6.512
C6 J23105-B0034 28.586
C7 J23106-B0031 22.011
C8 J23106-B0032 12.193
C9 J23106-B0034 58.781
Table 6.2. The repressor-regulated promoter-RBS library
Index BioBrick component Pu Pl Parameters

RT1 R0040-B0031 11.965 3.031 Katc = 6 ng/ml
RT2 R0040-B0032 8.740 3.381 KTetR = 45 nM
RT3 R0040-B0034 162.557 4.837 nTetR = 2
Table 6.3. The activator-regulated promoter-RBS library
Index BioBrick component Pu Pl Parameters

AL1 R0062-B0032 360.761 7.970 KAHL = 4nM
AL2 R0062-B0034 460.353 8.978 KLuxR = 55nM
nLuxR = 1
REFERENCES
Alon, Uri. 2007. An introduction to systems biology: design principles of biological circuits.
Vol. 10: Chapman & Hall/CRC.
Cascales, Eric, Susan K Buchanan, Denis Duché, Colin Kleanthous, Roland Lloubes,
Kathleen Postle, Margaret Riley, Stephen Slatin, and Daniele Cavard. 2007. “Colicin
biology.” Microbiology and Molecular Biology Reviews 71 (1):158-229.
Cavard, D, D Baty, SP Howard, HM Verheij, and C Lazdunski. 1987. “Lipoprotein nature of

the colicin A lysis protein: effect of amino acid substitutions at the site of modification
and processing.” Journal of bacteriology 169 (5):2187-2194.
Cavard, DANItLE. 1991. “Synthesis and functioning of the colicin E1 lysis protein:
comparison with the colicin A lysis protein.” Journal of bacteriology 173 (1):191-196.
Chak, Kin-F, White-S Kuo, and R James. 1991. “Cloning and characterization of the ColE7
plasmid.” Journal of general microbiology 137 (1):91-100.
CHAK, KIN-FU, and RICHARD JAMES. 1984. “Localization and characterization of a gene
on the ColE3-CA38 plasmid that confers immunity to colicin E8.” Journal of general
microbiology 130 (3):701-710.
Chen, Bor-Sen, Chia-Hung Chang, Yu-Chao Wang, Chih-Hung Wu, and Hsiao-Ching Lee.
2011. “Robust model matching design methodology for a stochastic synthetic gene
network.” Mathematical Biosciences 230 (1):23-36. doi: http://dx.doi.org/
10.1016/j.mbs.2010.12.007.
Ellermeier, Craig D, Errett C Hobbs, Jose E Gonzalez-Pastor, and Richard Losick. 2006. “A
three-protein signaling pathway governing immunity to a bacterial cannibalism toxin.”
Cell 124 (3):549-559.
Ellis, T., X. Wang, and J. J. Collins. 2009. “Diversity-based, model-guided construction of
synthetic gene networks with predicted functions.” Nat Biotechnol 27 (5):465-71. doi:
10.1038/nbt.1536.
Elmore, Susan. 2007. “Apoptosis: a review of programmed cell death.” Toxicologic
pathology 35 (4):495-516.
Gardner, Andy, Stuart A West, and Angus Buckling. 2004. “Bacteriocins, spite and
virulence.” Proceedings of the Royal Society of London. Series B: Biological Sciences
271 (1547):1529-1535.
Green, Douglas R. 2011. Means to an end: apoptosis and other cell death mechanisms: Cold
Spring Harbor Laboratory Press Cold Spring Harbor, NY.
Inukai, M, J Ghrayeb, K Nakamura, and M Inouye. 1984. “Apolipoprotein, an intermediate in
the processing of the major lipoprotein of the Escherichia coli outer membrane.” Journal
of Biological Chemistry 259 (2):757-760.
Lo, Tat-Ming, Mui Hua Tan, In-Young Hwang, and Matthew Wook Chang. 2013.
“Designing a synthetic genetic circuit that enables cell density-dependent auto-regulatory
lysis for macromolecule release.” Chemical Engineering Science.
Morita, Masatomo, Kazuhiro Asami, Yasunori Tanji, and Hajime Unno. 2001. “Programmed
Escherichia coli cell lysis by expression of cloned T4 phage lysis genes.” Biotechnology
progress 17 (3):573-576.
Pasotti, Lorenzo, Susanna Zucca, Manuel Lupotto, M Cusella De Angelis, and Paolo Magni.
2011. “Characterization of a synthetic bacterial self-destruction device for programmed
cell death and for recombinant proteins release.” Journal of biological engineering 5 (8).
PUGSLEY, ANTHONY P, and STEWART T COLE. 1987. “An unmodified form of the
ColE2 lysis protein, an envelope lipoprotein, retains reduced ability to promote colicin E2
release and lysis of producing cells.” Journal of general microbiology 133 (9):2411-2420.
Ross, Peter, Raphael Mayer, and Moshe Benziman. 1991. “Cellulose biosynthesis and
function in bacteria.” Microbiological Reviews 55 (1):35-58.
Saeidi, N., C.K. Wong, T.M. Lo, H.X. Nguyen, H. Ling, S.S.J. Leong, C.L. Poh, and M.W.
Chang. 2011. “Engineering microbes to sense and eradicate Pseudomonas aeruginosa, a
human pathogen.” Molecular systems biology 7 (1).
You, Lingchong, Robert Sidney Cox, Ron Weiss, and Frances H Arnold. 2004. “Programmed
population control by cell–cell communication and regulated killing.” Nature 428
(6985):868-871.
Chapter 7
ENGINEERING BACTERIA TO SEARCH

FOR SPECIFIC CONCENTRATIONS OF MOLECULES
BY A SYSTEMATIC SYNTHETIC BIOLOGY
DESIGN METHOD
ABSTRACT
Bacteria navigate environments full of various chemicals to seek favorable places for
survival by controlling the flagella’s rotation using a complicated signal transduction
pathway. By influencing the pathway, bacteria can be engineered to search for specific
molecules, which has great potential for application to biomedicine and bioremediation.
In this chapter, genetic circuits were constructed to make bacteria search for a specific
molecule at particular concentrations in their environment through a synthetic biology
method. In addition, by replacing the “brake component” in the synthetic circuit with
some specific sensitivities, the bacteria can be engineered to locate areas containing
specific concentrations of the molecule. Measured by the swarm assay qualitatively and
microfluidic techniques quantitatively, the characteristics of each “brake component”
were identified and represented by a mathematical model. Furthermore, we established
another mathematical model to anticipate the characteristics of the “brake component”.
Based on this model, an abundant component library can be established to provide
adequate component selection for different searching conditions without identifying all
components individually. Finally, a systematic design procedure was proposed.
Following this systematic procedure, one can design a genetic circuit for bacteria to
rapidly search for and locate different concentrations of particular molecules by selecting
the most adequate “brake component” in the library. Moreover, following simple
procedures, one can also establish an exclusive component library suitable for other
cultivated environments, promoter systems, or bacterial strains.
7.1. INTRODUCTION
Bacteria navigate to suitable living environments with the help of flagellum propulsion
controlled by a sophisticated signal sensing and transduction pathway (Sourjik and Wingreen
2012). Employing techniques from synthetic biology enables scientists to utilize such signal
transduction pathways to endow bacteria with novel moving behavior in response to specific
stimuli. Based on this concept, we are able to make bacteria automatically search for and
locate a particular environment to perform various missions. For example, bacteria could be
engineered to seek pollutants in the environment and destroy them (Sinha, Reyes, and
Gallivan 2010). Alternatively, bacteria could also be designed to locate diseased tissue, thus
can be employed to synthesize therapeutics or invade cancer cells (Anderson et al., 2006). It
is the searching mechanism that may make pollutant dissolution or drug delivery much more
specific and efficient. Here, we use a synthetic biology-based method to create bacterial
populations, which can rapidly search for and locate different concentrations of particular
molecules.
Escherichia coli is chosen to perform the genetic circuits we designed in this chapter.
Like many bacteria, E. coli swim randomly and switch between two states: smooth runs and
tumbles, which usually cause a change in direction. Smooth runs are maintained through
counterclockwise flagellum rotation until one or more flagellum rotates clockwise and makes
the E. coli tumble. The switching frequency between the two states is influenced by the
chemotactic system. As E. coli move toward gradients of favorable molecules for growth, the
frequency of tumbling will decline as bacteria approach the source. Conversely, if E. coli
deviate from gradients of favorable molecules, tumbling frequency increases to improve the
chances bacteria find a favorable environment.
E. coli detect chemotactic molecules, such as amino acids and sugars, by six
transmembrane chemoreceptor proteins. The chemoreceptors control activity downstream of
the CheA protein, which autophosphorylates the CheY protein. Phosphorylated CheY protein
(CheY-P) binds to the flagellar switch protein FliM and causes the flagellum to rotate
clockwise, which results in tumbling of the bacteria. A phosphatase protein, CheZ, is able to
dephosphorylate CheY-P, restoring the flagellum to a counterclockwise rotation and enabling
the bacteria resume smooth swimming (Sourjik and Wingreen 2012, Hazelbauer, Falke, and
Parkinson 2008, Tindall et al., 2008). When binding chemotactic molecules to
chemoreceptors, the activity of CheA protein is squelched and thereby CheY protein
predominates. In this form, CheY protein fails to interact with FliM protein and causes the
flagellum to rotate counterclockwise, which subjects the bacteria to smooth swimming. The
above-mentioned signal transduction pathway decides how E. coli respond when facing
chemical signals in the environment. Therefore, interfering with these transduction pathways
could provide an effective method for controlling E. coli movement behaviors.
In recent years, some groups have succeeded in reprogramming bacteria to recognize and
trace specific chemicals through protein engineering, RNA engineering, and synthetic biology
(Hwang et al., 2014). Goulian et al., reengineered a wild type Tar receptor, which responds to
aspartate, by directed evolution and enabled the Tar receptor to respond differently to
phenylacetic acid (PAA) (Mishler et al., 2010, Goldberg et al.,). By transforming a gene
encoding penicillin acylase, an enzyme hydrolyzes phenylacetylglycine (PAG) to PAA,
enabling E. coli to ascend PAG gradients. The advantage of direct reprogramming of
chemoreceptor proteins is that preserving the sophisticated chemotactic transduction pathway
retains the essential merits of the chemotactic system, such as quick response times and
adaptions. However, the disadvantage of protein engineering is that only minor modifications
can be made based on the existing protein structure. When handling chemicals with extensive
changes in structure compared to the native chemotactic molecules, an applicable ligand
specificity is hard to achieve.
Engineering Bacteria to Search for Specific Concentrations of Molecules … 125
Gallivan et al., chose an RNA engineering approach (Derr, Boder, and Goulian 2006),
using a strain lacking the CheZ gene. Absence of CheZ allows CheY-P to accumulate, which
in turn causes the bacteria to keep tumbling rather than swim effectively. The authors coupled
a designed riboswitch to respond to theophylline with a ribosome binding site (RBS) and a
CheZ encoding gene. Without theophylline, the three-dimensional structure of the riboswitch
will shelter the RBS and prevent transcription of the downstream CheZ gene. Once the
riboswitch combines with theophylline, the structural transformation will open the RBS to
ribosome binding so that the following transcription of CheZ will restore mobility to the
bacteria. That is, under this design, the engineered bacteria are able to distribute freely over
the region containing theophylline unless they move away from theophylline, which
constrains their moving ability. The advantage of this approach is that the components that
recognize a specific molecule are RNA aptamers that can be selected in vitro. Thus, the
design of the aptamer is no longer limited by the scaffold of the receptor protein and can be
used to detect new molecules. However, the disadvantage is that this approach relies on CheZ
protein production after activation, rather than direct signal transduction by the modified
receptor protein. Therefore, response time will be longer compared to the former protein
engineering method. Furthermore, only if the molecules of interest pass through the cell
membrane will the riboswitch be activated. That is, the sensitivity will be reduced in response
to detecting molecules, when part of them may fail to penetrate the membrane.
Cirino et al., studied communication-induced mobility on bacteria by a synthetic biology
method (Topp and Gallivan 2007). By giving different synthetic circuits, the authors created
two E. coli populations, sender cells and receiver cells, such that the former, which
constitutively produces acyl homoserine lactone (AHL), is able to restore the moving ability
of the latter. The mechanism is that receiver cells with a MotB gene knockout lose their
flagellum rotating ability. When receiver cells are close to sender cells that release AHL, the
AHL molecule will initiate the lux system, a quorum sensing system from Vibrio fischeri, in
the receiver cell's synthetic circuit. Initiating the lux system leads to expression of the MotB
gene located behind the lux promoter in the synthetic circuits and thus restores the receiver
cells' moving ability. The advantage of using synthetic biology is the increasing homogeneous
or heterogeneous detection systems that can be chosen as this field evolves, such as the fdhF
promoter which responds to hypoxia in E. coli (Anderson et al., 2006) and the PpbrA
promoter which responds to lead in Ralstonia metallidurans (Weiss et al., 2008), to name a
few. Similar to RNA engineering approaches, the disadvantage of synthetic biology is its long
response time.
In this chapter, a systematic synthetic biology-based method is designed to achieve the
searching function. A CheY double mutant D13K Y106W (CheY**) (Taghavi et al., 1997,
Dyer et al., 2004), which is resistant to phosphorylation and can convert into its active
conformation to mimic phosphorylated CheY, is chosen in this method to continuously
induce tumbling. CheY** production is controlled by the lux quorum sensing system, which
will be activated as AHL is detected. Therefore, the E. coli population will disperse
throughout the whole area until they run into the region containing AHL and start tumbling
instead.
With previous methods (Derr, Boder, and Goulian 2006, Topp and Gallivan 2007),
bacteria were deprived of their moving ability until specific molecules were detected,
adopting a stop-then-trace strategy. However, if the molecules of interest disperse in the area
away from the bacteria at the beginning, the bacteria rarely arrive at the designated area. In
this situation, the searching effect can be achieved only if the bacteria have saturated the
environment, which requires a considerable amount of bacteria. On the contrary, the synthetic
circuits in our design are based on a search-then-locate strategy to solve this searching
problem. The ability to explore environmental molecules enables the bacteria to actively and
efficiently search for specific molecules. This may reduce the amount of bacteria required for
searching. In addition, our design has two additional advantages. First, the method depends
on producing CheY** after stimulation. CheY** is able to directly deprive bacteria of
movement ability. That is, the method can be applied to most E. coli strains without
modification, whereas some genes, such as CheZ or MotB, must be knocked out in advance
under other methods (Derr, Boder, and Goulian 2006, Topp and Gallivan 2007). Second, E.
coli strains can be modified to meet specific requirements without losing the synthetic
circuits’ function. For instance, CheY or CheA genes can be knocked out to enable smooth
movement. The knocked out genes also block the chemotaxis transduction pathway, which
provides a chemotaxis-free experimental environment, neglecting the influence of
chemotactic molecules. Additionally, the mediums for measuring bacterial movement are no
longer limited to chemotactic molecule-free ones. These modifications cannot be adopted by
previous synthetic biology methods, where the lack of CheA or CheY causes E. coli to fail to
tumble in absence of stimulation, which breaks the fundamental stop-then-trace strategy.
However, the potential drawback of our design is that bacteria may stop around the source of
the molecule instead of invading it precisely owing to the over sensitivity of the circuits.
To solve this problem, we improved our circuit design through several modifications,
such as altering the order of genes in the pLux operon, replacing RBSs by weaker ones,
inserting a terminator in front of the RBS, and choosing CheY with different mutant levels to
design a library of “brake components” with various braking strengths. Putting different
“brake components” behind the promoter enables the circuits to respond to molecules of
interest with corresponding sensitivities. That is, different “brake components” can be chosen
to handle different concentrations of molecules. Stronger components are chosen to construct
more sensitive circuits when finding molecules with low concentration. Weaker components
are chosen to deal with molecules with high concentrations in order to prevent E. coli from
being excessively sensitive, which makes them stop too far away from the molecular source.
To verify and define how the brake components influence mobility requires applying
approaches to measure diffusion rate of E. coli populations. In this chapter, two tools,
swimming plate assays and microfluidic devices, were used to observe bacteria mobility.
Swimming plate assays can simply show how E. coli populations respond to chemicals by
inoculating the population on a semisolid agar plate. While the bacteria population diffuses
outward and forms concentric bands, molecules are dropped near the bacteria population to
simulate a condition representing a scenario where the bacteria population has approached the
source molecules of interest. The outcome of the response and distribution of the entire
population can be easily observed. The microfluidic device is used as a further accurate
measure of the bacteria population’s ability to diffuse. Utilizing the design from Stocker
et al., (Scharf, Fahrner, and Berg 1998, Ahmed, Shimizu, and Stocker 2010), we made the
microfluidic device restrict the distribution of the bacteria population into a specific area. As
the experiment begins, bacteria are allowed to swim freely from their original location. We
can then calculate diffusion ability by measuring the resulting population distribution over
time.
Using a microfluidic device, we tested the effect of “brake components” on population

diffusion rate at different AHL concentrations. For each “brake component”, a characteristic
equation is given to determine the relationship between diffusion rate and the activation level
of such components. The activation level is relatively quantified by measuring the expression
level of the green fluorescent protein (GFP) in front of the components. Through the
characterization of the components, we anticipate that these components can be used in other
promoter systems besides the quorum sensing system used in this study. Further, by simply
measuring the promoter systems’ GFP expression level in advance, one can combine the
promoter system with the appropriate components chosen in the library to design adequate
genetic circuits suitable for searching particular molecules at a specific concentration.
Figure 7.1. The basic design scheme of the synthetic circuit. Figure 7.1 shows the basic design scheme
of the synthetic circuit used in this study. For figures throughout this study, arrows that turn right
denote promoters, almonds denote RBS, arrows denote genes of functional proteins, and octagons
denote terminators. If there is no additional explanation, genes used in this study are from the MIT
Registry of Standard Biological Parts (database available online at http://parts.igem.org/Catalog). In the
figure, LuxR protein, expressed constitutively by a weak promoter-RBS pair (J23105-B0034), is
responsible for inducing expression of CheY** and GFP at pLux when AHL is combined. CheY** can
cause the bacteria to tumble continuously while GFP exhibits green fluorescence, which aids
measurement.
Construction and Verification of the Basic Circuit
In this study, synthetic genetic circuits are designed to reengineer E. coli to search for
AHL. Here, several experiments were carried out to demonstrate that components used in the
circuits worked successfully. The performances of the complete circuits were also tested to
show that components work normally when integrated together in complex circuits.
The basic circuit structure (named B1chey) is shown in Figure 7.1. The LuxR protein,
expressed by a weak constitutive promoter (J23105), is responsible for inducing expression of
GFP and CheY** at pLux when AHL is added. Expression of GFP enables us to quantify the
protein expression levels induced by the pLux promoter. The cheY gene is amplified from E.
coli MG1655 genomic DNA and modified at two positions by PCR technique (primers are
shown in supplementary Table 7.4), which yielded the double mutant CheY**.
Figure 7.2. Verifying the performance of the synthetic circuit in Figure 7.1 in response to AHL. The
searching ability induced by the circuit was tested by swarm assays. Without AHL, E. coli with the
synthetic circuit in Figure 7.1 distributed evenly across the semi-solid LB agar plate in (b), as seen for
E. coli with the control circuit in (a). While one microliter of 10 -3 M and 10-4 M of AHL were added on
the upper plate in (c) and (d), respectively, the E. coli stopped around the AHL sources and started
producing GFP. The relationship between the efficiencies of the plux promoter and the concentration of
AHL can be measured by the expression level of GFP. The result is given in (e).
In our design, the Chey** gene plays a central role in affecting E. coli mobility. We
tested the performance of the Chey** gene using the swarm assay. E. coli MG1655 is
transformed with circuit B1chey shown in Figure 7.2 (a). The information of all E. coli strains
used in this work is described in Supplementary C–a. Circuits with constitutive expression of
GFP (J05I13504) were sent into E. coli MG1655 as a control. E. coli were cultivated in an LB
medium and diluted to O.D. 0.05 before being dropped onto the semi-solid LB agar plate.
After being cultivated for 16 hours, one microliter of 10-3 M, 10-4 M, or 0 M AHL was
dropped beside the E. coli colonies. The colonies were then cultivated overnight. The result is
shown in Figure 7.2(a–d). E. coli with control circuits evenly diffused over the semi-solid LB
agar plate. Without AHL, E. coli with circuit B1chey also distributed evenly. As AHL was
added, the E. coli colonies diffused outward until they detected AHL and stopped around the
AHL sources. The higher the concentration of the AHL sources, the wider the area where E.
coli stopped moving forward. In line with our design, once E. coli stopped near the AHL
sources, the E. coli population started accumulating and producing GFP. As a result, the
genetic synthetic circuit is proven to make the E. coli population locate a specific area and
execute the task they were engineered to perform.
For the same circuit, to quantify the expression level of the lux system in response to
different AHL concentrations, an Elisa Reader measured GFP for 12 hours after adding AHL.
The result is shown in Figure 7.2 (e). After AHL was added, the expression of GFP gradually
rose with time and achieved a steady state after 4 hours. With incremental AHL
concentrations, GFP expression level also increased, as expected. The expression starts rising
obviously as the AHL concentration exceeds 10-10 M and gradually saturates as the
concentration reaches 5 × 10-8 M. Therefore, the most sensitive concentration range of the
system to AHL is between 10-10 M and 5 × 10-8 M.
Figure 7.3. Improving the basic circuit by constructing a series of “brake components.” A basic “brake
component” library in (b) was constructed to endow the standard circuit in (c) with various sensitivities
to AHL. The structure of a “brake component” is shown in (a). A terminator and an RBS are used to
regulate the expression level of the downstream CheY** protein by influencing the transcription and
translation efficiency of CheY**, respectively. By choosing different terminators or RBSs, the “brake
component” library in (b) can be constructed. The brake component with the part (terminator, RBS, or
CheY**) is represented by O) while X represents without the part. Finally, the standard circuit in (c) is
constructed by selecting specific brake components in response to different situations. This circuit is
composed of three parts: (i) conRFP expresses RFP constitutively to facilitate the observation of the
bacterial distribution; (ii) LuxTemplate detects AHL and induces downstream gene expression; and (iii)
the aforementioned “brake component.”
Improvement of the Circuits’ Performance by Constructing the “Brake

Components” Library
In the previous section, the distribution of E. coli populations in response to different

AHL concentrations was demonstrated by observing diffusion on semi-solid agar plates. The
result shows that E. coli stopped around the AHL sources and started expressing GFP as
expected. However, E. coli could hardly invade the center of the sources. The area that
constrained E. coli from entering further was affected by the concentration of the sources. The
higher the concentration, the wider the area that develops. This phenomenon is due to the fact
that the circuit responds intensively when the AHL concentrations rise above certain values
and deprive the E. coli of almost their entire movement ability. Therefore, there is a specific
threshold of AHL concentration above which E. coli’s production of TetR (tetracycline

repressor) or CheY** will saturate. Once the E. coli population meets the threshold isopleth,
the population stops and forms a border area shown in Figure 7.2. It is the hyperactivity of the
circuits that makes our designs too sensitive to deal with high concentration cases.
To make our circuit less sensitive, one can choose to modify the CheY** structure. If
CheY** were less active, E. coli might retain its basal movement ability and keep moving in
the area with an AHL concentration above the threshold. However, to modify a protein
properly is a difficult task since it is difficult to predict the effect of various mutations. Here,
we chose the following synthetic biology method to solve this problem.
First, circuits are modified by three ways to yield six combinations of components as
shown in Figure 7.3(a–b). First, the Chey** gene is placed at the last position of the lux
operon.
Second, the Chey**-related RBS is replaced by two weaker ones: B0031, whose
efficiency when compared with B0034 is 0.07; and B0032, whose efficiency when compared
with B0034 is 0.3. The data above were acquired from MIT Registry of Standard Biological
Parts (Ahmed and Stocker 2008).
Third, three sequences with different terminating abilities are placed in front of the
Chey**-related RBS: None; B0014 with 60.4% termination efficiency; and B0015 with
98.4% termination efficiency. The data above were acquired from MIT Registry of Standard
Biological Part.
In addition, a “con” component, which lacks a terminator, RBS, and CheY** gene, was
constructed as a control. The components were then combined with an RFP expressing
circuits (conRFP) and a Lux-promoter-activated circuit (LuxTemplate) containing a GFP
gene, which served as an index for the transcription and translation level. The complete
circuits are shown in Figure 7.3(c), which is now ready for observing bacterial population
distribution under a fluorescent microscope and measuring the GFP expression level by an
Elisa Reader.
Verification of the Function of the Components by Swarm Assays
In this section, we simply tested the modified circuits in Figure 7.3 (c) by the swarm
assays. Three of the components, XB32, B14B32, and B15B32, which contain different
terminators, were chosen to demonstrate the circuits’ performance. We anticipate that
different amounts of CheY** gene can be transcribed according to the termination efficiency
of the terminators. A strong terminator, B0015, can block most of the transcription, which
restricts the amount of the CheY** protein, unless considerable AHL is detected. Thus,
component B15B32 is expected to have the weakest sensitivity. Conversely, the components
B14B32 or XB32, with weak or no terminator, show high sensitivity.
We verified the aforementioned components by swarm assays using E. coli strain RP437,
which is widely used in chemotaxis and swimming studies. E. coli and AHL were dropped on
the agar spanning a fixed distance. After sufficient time for diffusing, the results are shown in
Figure 7.4. In the figure, we can see that each component has its ideal concentration range for
molecule detecting. Taking XB32 for example, the bacteria population can appropriately
surround a drop of 10-6 M AHL. However, in the case that the concentration is lowered to 10-7
M, the bacteria can hardly detect the source of the AHL. In another case, while the
concentration is increased to 10-5 M, the bacteria population intends to stop too far away from
the molecule source. Similarly, both the components B14B32 and B15B32 have the same
property. The appropriate concentration for molecule detection is 10-5 M for B14B32,
whereas the appropriate concentration for molecule detection is between 10-4 M and 10-5 M
for B15B32. This corresponds to our expectation that sensitivity increases with decreasing
termination efficiency. As a result, we confirm that our design is effective to engineer bacteria
to search for different concentrations of AHL.
Figure 7.4. Verification of the brake components by swarm assays. Swarm assays were performed to
verify the function of different brake components. Standard circuits containing three brake components,
XB32, B14B32, and B15B32, were sent into E. coli strain RP437. After being cultivated in M9 medium,
E. coli were dropped at the left end of the line on the semi-solid agar made with M9 medium.
Simultaneously, a 1.5 microliter with a specific concentration of AHL was dropped at the right end of
the line on the agar (4 cm away from the E. coli). The distribution of the E. coli population was
observed after 20 hours of cultivation.
7.3. RESULTS
Measurement and Identification of the “Brake Components”
In the previous section, we verified the circuit design method in this study and improved
it by several modifications. Further, we have proven that the modifications are effective by
swarm assays. However, it is worth noting that the swarm assays can only verify the circuits’
property qualitatively. The bacteria population was placed on semi-solid agar and allowed to
search for the AHL source, which was organized by dropping a particular amount of AHL on
the agar. The spatiotemporal distribution of the AHL is undetectable and unpredictable.
Hence, although we know the initial amount of the source and can observe the distance
between the center of the source and the place where the population stopped, we cannot tell
the exact concentration which stops the population from diffusing. Therefore, the microfluidic
technique is applied in this study to measure diffusion rate of the bacterial population under
different AHL concentrations.
Here, the experimental methods to measure the diffusion rate include the mathematical
principle, fabrication of the microfluidic device, experimental procedure, and data processing.
Following these methods for the circuit in Figure 7.3(c), we observed the relationship
between AHL concentrations and the corresponding population diffusion rate, which
generated the characteristics of each “brake component.” To compare components more
objectively, we adopted a relative quantification method. The expression levels of the
components were represented by the expression levels of the upstream GFP. A sigmoid-like
function was thus constructed to simulate the relationship between GFP expression and
population diffusion rate. This function can provide useful information, such as effective
region and saturation point of each component, which we can use to compare components in
the library and select appropriate circuit design. Further, different mutant levels of CheY
protein were also tested. We anticipate that the mutant levels may make small adjustments to
the components and enrich our components library.
Methods for Measuring Bacteria Population Diffusion Rate
In this section, we describe the method for measuring diffusion rate of the bacterial
population by a microfluidic device. There are several advantages to measuring diffusion rate
using a microfluidic technique: (i) channels with micrometer-scale are fabricated rapidly and
precisely; (ii) more accurate flow can be produced; and (iii) devices can be made from
PDMS, which is transparent, such that cells can be directly observed by a microscope (Scharf,
Fahrner, and Berg 1998). By utilizing the design from Stocker et al., we constructed a
microfluidic device capable of locating bacterial populations in a particular area. By
observing bacterial distribution spatiotemporally, we calculated the diffusion rate of the
population. In the following paragraphs, the aforementioned sections containing the
mathematical principles, fabrication of the microfluidic device, experimental procedure, and
data processing will be introduced in order.
So far, a number of mathematical models describing bacteria movement have been

established. At the population scale, the most widely used model is that of Keller and Segel
(Lin and Levchenko 2012), which provides an expression for the flux (J) of bacteria:
(7.1)
where B(x,t) denotes the concentration of bacteria, is the random motility coefficient that
indicates the diffusivity of the population owing to their random walk behavior, and is the
chemotaxis velocity resulting from their preferred movement toward attractants or away from
repellents. Substituting the above equation into the total cell number density conservation
equation (Keller and Segel 1971) yields the bacterial transport equation:
(7.2)
In this study, we characterized components using E. coli strain RP5232, which lacks the
CheY gene. This strain was chosen for two reasons. First, the phosphorylated CheY protein
can influence the rotating direction of the flagella and make E. coli tumble. Without CheY, E.
coli that swim smoothly can be easily observed to switch between their original and tumbling
state induced by the “brake components”. Second, in the absence of the CheY gene, the
chemotactic pathway is blocked so that no chemotactic behavior will affect the measurement
of the diffusion rate. Therefore, since , Eq. (7.2) will reduce to the diffusion equation:
(7.3)
Solving the partial differential equation yields:
(7.4)
Based on Eq. (7.4), we established a one-dimensional band of E. coli using the

microfluidic device. By measuring the concentration (B) of E. coli at different positions (x),
we compute by fitting the measurement to Eq. (7.4). Calculating at
different time points (t), we can obtain the diffusion rate by linear regression.
Next, to demonstrate the experimental procedure, we measure E. coli strain RP5232
containing circuit “conRFP-LuxTemplate-con” without AHL being added as an example.
First, the microfluidic device in Figure 7.5(a) was constructed. A and B in Figure 7.5(a) are
the inlets for medium and bacteria, respectively. The flows of bacteria and medium converge
at the end of the micro-injector and outflow from outlet C. As liquids were pushed into inlets
A and B under a stable pressure, liquid from inlet B was sandwiched in the middle of the
channel shown in Figure 7.5(b) (displayed by red and blue coloring). When the pressure was
turned off, both the flow stopped and the band of E. coli population started to expand to both
sides. To record the distribution of the population, red fluorescence protein produced by E.
coli was captured by a fluorescence microscope. We took bursts of 25 pictures over 5 seconds
interspaced with 30 second intervals without imaging.
Finally, coupling these 25 pictures gives the trajectory of E. coli over 5 sec, as shown in
Figure 7.5(c). Distribution (B) of E. coli at different positions (x) was computed by summing
up the trajectory of pixels, which was x micrometers from the center axis and normalized by
the total number of pixels of the trajectory. Distribution (B) was then fitted by a genetic
algorithm (GA) to Eq. (7.4), which yielded values of . The fitting results are shown
in Figure 7.5(d). Blue and red lines denote the experimental results and their fitting results,
respectively. Diffusion rate was acquired by handling values of over time by
linear regression shown in Figure 7.5(e). In this case, .
Figure 7.5. The methodology for measuring bacterial population diffusion rates. Figure (a) shows the
pattern of the microfluidic device. In the figure, A and B are the input of the medium and bacteria,
respectively, and C is the output of the microfluidic device. The outer flow will restrict the distribution
of the inner flow as shown in (b) (displayed by red and blue coloring). Once both flows stop, the E. coli
in the inner flow starts to diffuse outward. The trajectory of the E. coli at different time points can be
recorded by microscope in (c). The distribution of the bacteria population over time can also be
calculated in (d). The bacterial population diffusion rate is obtained by handling the standard deviation
of the population distribution over time by linear regression, as shown in (e). In this case, the diffusion
rate is .
Measurement of the “Brake Components”
Based on the methods, the effect of circuits with different “brake components” shown in
Figure 7.3(a) on bacterial population diffusion rates were measured. For each component,
several AHL concentrations were added to test the performance under different conditions
and components’ sensitivities. Each result is shown in Supplementary Figure 7.13 and
summarized in Figure 7.6.
In Figure 7.6, the “con” circuit, whose “brake component” is blank, is the control of the
experiment. Notably, without the production of the CheY** protein, the diffusion rate drops
to the basal level at high concentrations of AHL (exceeding 10-7 M). This suggests that the
mass expression of GFP may reduce bacteria mobility. The six components in the library
show a similar tendency toward stopping the E. coli completely as AHL concentration
increases. As expected, the six components succeed in displaying different sensitivities
toward AHL. By comparing their performance, we conclude that the sensitivities of the
components mainly depend on their terminator in front of the RBS and Chey**. Components
without a terminator (XB31, XB32) present the highest sensitivity, components with a weak
terminator (B14B31, B14B32) present normal sensitivity, and components with a strong
terminator (B15B31, B15B32) present the lowest sensitivity.
RBSs also provide minor modifications to the components. However, according to this
figure, strength does not differ between the two RBSs. In cases without a terminator (XB31,
XB32) and with a strong terminator (B15B31, B15B32), components containing B32 present
lower sensitivities. While in another cases (B14B31, B14B32), components containing B32
present higher sensitivities. The difficulty in distinguishing the effect derived from different
RBSs can be explained by inconsistencies between component activation levels. In this study,
GFP was put in front of the “brake components” to quantify expression between components.
Thus the GFP expression levels denote the components’ activation level. Supplementary
Figure 7.15(a) shows the GFP expression levels of the control circuits during the diffusion
experiment. The concentration region of AHL to which the circuits are most sensitive is
between 10-10 M and 10-8 M. That is, in this region, any fluctuation to the promoter system
tends to be significantly amplified. To demonstrate this phenomenon, we show in
Supplementary Figure 7.15 (b) the GFP expression level of all components during the
experiment. On the condition that no AHL was added or that AHL reached saturation (10-8
M), variation between components was negligible. However, GFP expression levels varied
obviously when AHL concentrations (10-10, 10-9) were near the most sensitive region. For
both cases, the highest value is three times more than the lowest one. The reasons for the
variation might be due to intrinsic factors, such as bacterial individual differences or different
secondary structures of mRNA derived from different components that may influence
translation efficiency, or extrinsic factors like pipetting errors when preparing extremely low
concentrations of AHL.
Briefly speaking, while comparing performance among different components, the
activation levels of components tend to be inconsistent, especially when AHL concentrations
are near the most sensitive region. Moreover, this region (10-10 M ~ 10-8 M) covers the range
where many components perform their function significantly, which makes comparison
difficult. To solve this problem, we compare and define each component directly through
their activation levels, instead of AHL concentrations. The method and results will be
discussed in the next section.
Figure 7.6. Comparing the effect of each brake component in response to different AHL concentrations.
All the brake components in the component library in Figure 7.3 (b) were combined with a standard
circuit and sent into E. coli strain RP5232 to measure their effect on population diffusion rate by the
microfluidic device. Each line represents how a specific component influences the diffusion of bacteria
in response to different AHL concentrations.
Constructing a Mathematical Model to Identify the “Brake Components”
In this section, we establish a mathematical model for describing the relationship between
component activation levels and their corresponding bacterial population diffusion rates. By
characterizing by activation levels, we can directly compare components by the well-
measured GFP expression levels, as well as utilize components in other promoter systems
without re-characterization. Through measuring the promoters’ GFP expression levels in
advance, we can estimate the consequent performance of the circuits when selecting different
components. We expect that this may contribute to circuit design substantially.
The experimental data suggested that, for every component, the curves of diffusion rates
remained high and smooth when the activation levels were below specific values, above
which the rates decreased abruptly. The main differences among components were when and
how the diffusion rates fell. To describe this desired diffusion rate phenomenon, we set up a
sigmoid-like mathematical model:
(7.5)
where and denote diffusion rate and nominal diffusion rate when no AHL was added,
respectively, CAL is the abbreviation of components’ activation levels, which are represented
by the GFP expression levels, a indicates the decreasing degree of the curve, while b indicates
the activation levels required for the diffusion rate to fall to half of its initial value . Both a
and b can be calculated by fitting the experimental data to the model by genetic algorithm.
Figure 7.7. Representation of components characterized by Eq. (7.5). The results of fitting the
experimental data of each brake component to Eq. (7.5) are shown in Supplementary Figure 7.14 and
summarized in (a). The x-axis of the figure represents the GFP expression level of the circuit, which
can be viewed as the components’ activation levels (CAL) in Eq. (7.5). Based on Eq. (7.5), some
criteria of the components can be defined to facilitate the comparison among components. In (b), the
“Effective Region” indicates a range of CAL upon which the diffusion rate of the population is less
than 70% of the initial diffusion rate. The “Saturation Point” denotes the CAL that makes the diffusion
rate of the population equal to 10% of the initial diffusion rate.
Fitting the experimental data of each component to the model yielded the results shown
in Figure 7.14. To facilitate comparison, we normalized by and summarized this in
Figure 7.7(a). First, in Figure7.7 (a), the curve of the control circuits drops significantly after
the GFP expression level exceeds a value of approximately (A.U.), beneath which
the curves of all components have already decreased to their basal levels. This indicates that
the decreased diffusion rate is mainly due to component expression rather than mass
expression of GFP protein. Further, similar to Figure 7.6, the six components can be
obviously divided into three categories according to their terminators. As GFP expression
levels rise, components without a terminator (XB31, XB32) cause diffusion rates to drop first,
followed by components with a weak (B14B31, B14B32) and then strong (B15B31, B15B32)
terminator. This is consistent with results in the previous section. Moreover, in the previous
section, the difference between the two RBSs was not obvious. Here, by comparing the fitted
value “b” in Figure 7.7(b), we find in the cases with fixed terminators, components with RBS
B31 always have a smaller “b”, which indicates that lower component activation levels are
required to meet half their diffusion rates. Thus, components with B31 are slightly more
sensitive than those with B32.
Table 7.1. Information of the identified “brake components”
Components a b Effective region Saturation point

con 6.38×10-6 1.69×106 CAL>1.55×106 2.03×106
XB31 2.67×10-4 1.01×104 CAL>6.92×103 1.83×104
XB32 4.12×10-4 1.06×104 CAL>8.58×103 1.60×104
B14B31 1.02×10-5 2.84×105 CAL>2.01×105 5.00×105
B14B32 6.37×10-6 3.06×105 CAL>1.73×105 6.51×105
B15B31 4.75×10-5 1.12×106 CAL>1.10×106 1.16×106
B15B32 1.62×10-5 1.23×106 CAL>1.17×106 1.36×106
XB32CheY 1.50×10-4 4.12×104 CAL>3.48 ×104 5.51 ×104
B14B32CheY 5.00×10-4 1.06×106 CAL>1.05 ×106 1.11×106
XB32CheYD13K 4.42×10-4 7.04×103 CAL>5.13×103 1.20×104
B14B32CheYD13K 1.22×10-5 2.16×105 CAL>1.46×105 3.96×105
Characterizing components provides some useful information for circuits design. To

facilitate the selection of the proper components, we define some criteria, shown in Figure
7.7(b), which contribute to comparison among components. “Effective region” denotes the
ranges of activation levels in which the diffusion rates are less than 70% of . This suggests
that, for a component, the diffusion rate of bacteria will not be affected until the component’s
activation level enters its effective region. “Saturation point” denotes the values of activation
levels upon which the diffusion rates drop to one-tenth of their initial diffusion rates . Once
the component’s activation level exceeds the saturation point, the increment will make no
difference to diffusion rate.
These criteria provide some considerations for circuit design. For example, it is worth
noticing that when dealing with a promoter that has a tendency to leak, components whose
effective regions cover the leakage level of such promoter systems are not suitable for this
situation. Otherwise, the components will be activated consistently, irrespective of whether or
not the promoter is induced. In another case, when a component is chosen in order to make
the bacteria trace a specific concentration of molecule, the most appropriate option among the
components library is the one whose saturation point is slightly lower than the expression
level of the promoter under such concentrations. Thus, the bacteria will express adequate
levels of the brake component and stop as they move into the area containing the expected
concentration.
Finally, we list the information of each component, including the fitted value (a, b) and
the above-mentioned criteria in Table 7.1. In addition, the application of the table and the
mathematical model to circuits design will be discussed in future research.
Enrichment of the “Brake Components” Library by Mutations
(CheY**) was chosen in this study to induce tumbling in E. coli. In the

absence of CheA, a kinase to phosphorylate CheY, CheY** is still active without
phosphorylation. Even when CheZ, a protein known to accelerate the dephosphorylation of
CheY-P, was expressed in high levels, Chez failed to respond to CheY** [12]. Studies
showed that when CheY** binds , a peptide derived from the flagellar motor switch
FliM, CheY** has a conformation similar to -activated wild type CheY, which is a
stable analog of the phosphorylation-induced conformation. In the absence of FliM, CheY**
has a conformation similar to unphosphorylated wild type CheY. In short, CheY** is suitable
to convert between its active and inactive conformation in the absence of phosphorylation or
dephosphorylation (Taghavi et al., 1997).
Recent studies have reported that there are other mutations that show individual effects.
For example, similar to CheY**, is active in the absence of
phosphorylation. and are hyperactive when they are phosphorylated
(Rivero et al., 1989, Schuster et al., 1998). Compared with the previous two mutations,
has lower activity, which causes a decreased tumble signaling, whereas
, , , and show no activity and cause
smooth Swarm (Schuster et al., 1998). Introducing the various mutations of CheY, the “brake
components” library can be efficiently enriched by simply carrying out the mutations based
on the existing circuits.
Here, we expanded two more versions of components XB32 by adopting different mutant
levels. Utilizing wild type CheY and we could create two new components named
XB32CheY and XB32CheYD13K, respectively. We tested the characteristics of these
components by previous methods. The results are shown in Table 7.1. The wild type version
of XB32 has a larger activation level “b” than the others. This suggests, as expected, that it
makes the component less sensitive. Surprisingly, XB32CheYD13K shows the smallest
activation level “b”, which indicates that the version is even more sensitive than
the double mutant version, CheY**. Therefore, all the components in the library can be
altered to a more sensitive ( ) or less sensitive (wild type) version through this
method.
Simulation of the Proposed Genetic Circuit
The purpose of this section is to simulate the operation of the genetic circuits. Two kinds
of model will be discussed in this section. First, we will construct the model that could
establish the relationship between AHL concentration and the corresponding protein
expression level to show component activation levels (CAL) of the “brake components”.
Coupling this model with identifying components in Eq. (7.5), we are able to simulate the
performance of the complete circuits, i.e., to simulate the bacterial population diffusion rate at
all AHL concentrations. This method is not restricted to the lux promoter system in this study.
When the components are transplanted to other promoter systems, we can simulate the
complete circuits’ performance at all concentrations of the new stimulus by identifying the
new promoter systems. Combining the identification of the components with the model, we
can simulate the diffusion rate at all AHL concentrations. The simulated results are then
compared with the experimental ones.
However, for each component, how the parts (terminators, RBSs and different CheY
mutations) influence the parameters in Eq. (7.5) (i.e., “a” and “b”) has not yet been discussed.
The second model is established to describe the relationship between the components
activation levels (CAL) and the bacterial population diffusion rate. By highlighting the
relationship between the factors and parameters, differences in parameters can be explained
between components. More importantly, the characteristic of those uncharacterized
components can be anticipated. That is, we are able to construct new components based on
existing parts and simulate their properties without measuring each of them. For example, we
introduced the wild type CheY and . The number of the components in the library
can be tripled by replacing the CheY** in the existing components with the two new ones.
The characteristics of the new components in the library can be anticipated by simulation.
Combined with the model, enrichment of the library is beneficial for selecting the ideal
components for detecting specific AHL concentrations when designing the circuits.
Dynamic Model of the Synthetic Circuit
In this section, we aim to establish the relationship between AHL concentration and the
corresponding protein production generated by the lux promoter system. Considering the two
factors that may affect transcription level, the amounts of transcription factor luxR protein
and luxR-AHL complex, we construct the following dynamic equations to describe the
performance of the genetic circuits in Figure 7.3(c):
(7.6)
where and denote the amount of luxR and GFP respectively, which are
affected by the following factors: and denote the transcription and translation
strength of the indexed promoter-RBS pair; and is a constitutive value, whereas
varies according to the amount of and , which is the concentration of inducer
AHL. The strength of the promoter-RBS pairs ( , ), as well as the degradation
( ) and dilution rate (D), owing to cell division, control the amount of protein per
cell. Further, the strength of can be described by the third equation in Eq. (7.6).
and denote the basal and maximum expression level of the promoter-RBS pair.
and are the binding affinity and binding co-operativity between the promoter and
the luxR-AHL complex , which is influenced by the amount of luxR, AHL, and the
dissociation rate between luxR and AHL.
During the experiment, if a sufficient time is provided for bacteria to respond to AHL
(about four hours), the expression level of the protein will reach steady state in which the
production rate of the protein is equal to its degradation and dilution rate. Thus, the above-
mentioned dynamic model can be simplified to its steady state form as follows:
(7.7)
With the steady state equations, through measuring the amount of the proteins and
referring to previous studies for some parameters ( ), we can identify the
remaining parameters ( ) by a genetic algorithm and
accomplish the dynamic model. The results of the identification and simulation are shown in
the following section.
Identification of the Diffusion Rate of Bacterial Population at All

AHL Concentrations
To identify the unknown parameters ( ) in Eq.

(7.7), two experiments were performed to acquire the protein expression level derived from
the two promoter-RBS pairs, J23105B0034 and PluxB0034.
First, to identify the strength of the constitutive promoter-RBS pair J23105B0034,
synthetic circuits shown in Figure 7.8(a) were constructed and sent into E. coli strain RP5232.
To facilitate the measurement of protein production, the luxR protein was replaced by a GFP
protein, which can be simply measured by an Elisa Reader. Under this condition, the
degradation rate in the first equation of Eq. (7.7) must be altered, which yields
. As a result, simply measuring the GFP expression level at
steady state gives the strength of the promoter-RBS pair.
Second, to identify the parameters related to the AHL-activated promoter-RBS pair
PluxB0034, the synthetic circuit in Figure 7.3(c) with the “con” component was selected to
represent all circuits in the figure. We assume that the components behind the GFP sequence
will make no difference to the expression level of the GFP protein. Hence, for all the circuits
in the figure, the equation and its parameters describing the PluxB0034 system can be unified.
To identify the parameters, E. coli strain RP5232 was transformed by circuits with the “con”
components. Different AHL concentrations ( ) were added to the bacterial culture and the
corresponding GFP expression levels ( ) at the steady state were measured. Further,
was calculated according to the first equation in Eq. (7.7) whose was obtained
from the first experiment. Based on the above information, the remaining parameters
( ) can be estimated according to Eq. (7.7).
To verify whether the model and the identified parameters corresponded to the
experimental results, we next compare the simulated and experimental results, shown in
Figure 7.8(b). The figure shows that the simulation accurately captures GFP expression in
response to different AHL concentrations. In addition, we show the experimental results of all
the components in Figure 7.3(c) in blue stars, which show that the simulation also succeeds in
representing them.
Fig 7.8. Simulating “stimulus vs. CAL” and “stimulus vs. diffusion rate” based on the dynamic model.
In (a), the circuit was used to identify the strength of the constitutive promoter-RBS pair J23105B0034.
The experimental data of the standard circuit with the “con” component was used to identify the
parameters in Eq. (7.7). In (b), the simulated equation, in Eq. (7.7), is represented by a black
line, whereas the experimental result of the representative “con” component is represented by red stars.
For the rest of the brake components in Figure 7.3 (c), their experimental data are collectively
represented by blue stars. In (c), the simulated (using the unified parameters identified from the “con”
component) and experimental results for each brake component are represented by a blue line and red
star, respectively. Based on the experimental results, the simulation can be improved (purple lines) by
identifying the respective parameters for each component.
Finally, we combine the simulation with the characteristics of the components, which
links the amount of AHL and the consequent diffusion ability. The simulations in Figure
7.8(c) have unified these parameters to approximate the performance of each circuit and can
be improved by identifying their own parameters, shown by purple lines. The improvement
contributes to reducing the intrinsic factors that cause differences in GFP expression levels
between circuits. That is, through identifying the promoter-RBS system, we can estimate the
performance of the synthetic circuits at the design stage. Simulations can be further improved
after constructing synthetic circuits.
Constructing a Mathematical Model to Simulate the Characteristic of

the “Brake Components” Based on Components’ Composition
In the previous section, we constructed a mathematical model to describe the relationship

between AHL concentration and protein production, which is expressed by the lux promoter
system. Here, we further aim to establish the relationship between the above mentioned
protein production, which also indicates the component’s activation levels (CAL), and the
bacterial population diffusion rate.
The sigmoid-like equation in Eq. (7.5) can appropriately depict the behavior of the
bacterial population diffusion rate in response to CAL. Therefore, we established the model
based on this existing equation. In the case of the component XB32 without a terminator in
front of RBS and CheY**, Eq. (7.5) can be viewed as the direct relationship between the
relative amount of the CheY** protein and the bacterial population diffusion rate. Compared
with XB32, other components can alter the transcription and translation efficiency of the
CheY** protein, while the relationship between the amount of CheY** and bacterial
diffusion rate remains the same. For example, combining terminator B0014 in front of XB32
makes B14B32. Therefore, the terminator can block a certain ratio of CheY** production.
Thus, Eq. (7.5) for B14B32 can be directly represented by the modified XB32’s equation
whose CAL is divided by a specific value, which indicates the termination efficiency of
terminator B014. Based on this concept, we modified Eq. (7.5) by considering the effect of
each part of the component (in the following, “part” includes terminators, RBSs, and CheY
mutants that compose components). The relative diffusion rate for each brake
component can be expressed as follows:
(7.8)
1. The result of a terminator can block a specific ratio of protein production. After
adding a terminator, production is reduced to 1/m of the initial amount (Figure
7.9(b)).
2. The strength of an RBS can be represented by n times the strength of the RBS B0032
(Figure 7.9(c)).
3. denotes CAL (the relative amount of the CheY** protein) needed to halve the
bacterial population diffusion rate. For different CheY types (wild type or mutants),
the CAL needed to halve the diffusion rate can be represented by p × (Figure
7.9(d)).
4. The parameters m, n, and p for a component are only decided by the chosen
terminator, RBS, and CheY version, respectively. The effects of the terminators,
RBSs, and CheY mutants are mutually independent (Figure 7.9(e)).
That is, replacing a terminator (Figure 7.9(b)) or an RBS (Figure 7.9(c)) considers how to
alter the scale of the x-axis of the “CAL vs. diffusion rate” curve of the standard circuit XB32
(Figure 7.9(a)), whereas replacing CheY** with another CheY mutant (Figure 7.9(d)) shifts
the “CAL vs. diffusion rate” curve of the standard circuit XB32 (Figure 7.9(a)) horizontally.
Most importantly, if the effects of all parts are mutually independent, the characteristic curve
of any component can be anticipated according to its terminator, RBS and CheY
mutant by Eq. (7.8).
Further, by comparing components that differ in only one part, the parameter of that part
can be obtained. Depending on this concept, the identified components in the library were
compared with each other to obtain the parameters of all parts used in this study. The results
are given in Table 7.2 as follows.
Table 7.2. Comparison of components that differ in only one part
Comparison Result average

XB32 : B14B32
XB31 : B14B31
XB32CheYD13K : B14B32CheYD13K
XB32CheY : B14B32CheY
XB32 : B15B32
XB31 : B15B31
XB32 : XB31
B14B32 : B14B31
B15B32 : B15B31
XB32 : XB32CheY 3.82
B14B32: B14B32CheY 3.49 3.66
XB32 : XB32CheYD13K 0.66
B14B32: B14B32CheYD13K 0.70 0.68
For terminators B0014 and B0015, the m parameters for these terminators are consistent.
All the m parameters for B0014 and B0015 are close to 28.5 and 111, respectively. In
addition, for RBS B0031, all parameters n are also around 1.08. This shows that the model to
describe the operation of terminators and RBSs is persuasive and we can validate assumptions
1 and 2 in Figure 7.9.
To verify that our model is suitable for other CheY versions, two circuits were further
constructed. The component B14B32 was modified by changing its CheY** to two other
CheY versions (wild type and ), which yielded the components B14B32CheY and
B14B32CheYD13K, respectively. These components were measured and compared to
B14B32. The results are consistent with those in previous. is the most sensitive,
whereas wild type CheY is the least sensitive. Further, even if each CheY version was put
with different components (XB32 or B14B32), the p values are quite consistent, which
validate our assumption 3. The p values for wild type CheY and are around 3.66
and 0.68, respectively.
Figure 7.9. Assumptions for establishing part-based mathematical model. Component XB32 is defined
as the standard component, whose composition and characteristic equation are given in (a).
Assumptions 1, 2, and 3 suppose that the effect of changing a terminator, RBS or CheY mutant, can be
regarded as modifying the initial characteristic equation in (a) by parameter m, n, and p shown in (b),
(c), and (d), respectively. The modified equation can be viewed as to alter the x-axis’ scale ((b) and (c))
of the initial figure or shift the initial characteristic curve horizontally (d). Assumption 4 supposes that
the effects of terminator, RBS and CheY mutant, are mutually independent. If so, every component can
be represented by Eq. (7.8) in (e) which is derived from the characteristic equation of the standard
component XB32 in (a).
From the above comparison, each part of a component (terminator, RBS, or CheY
mutants) was placed in at least two different contexts. The results show that the m, n, and p
values remain constant, supporting assumption 4 that these values are mutually independent.
Figure 7.10. Comparison between the simulated characteristics of the components and the experimental
results. The simulated characteristics of the components based on Eq. (7.8) are shown by blue lines,
whereas the experimental results of the corresponded components are shown by green lines.
Finally, in Figure 7.10, we simulate the characteristic of all components (blue lines) and
compare them with experimental data (green lines). The results show that the model can
approximately simulate the experimental results, except for components B15B31 and
B15B32. The reason is that a values for these components deviate from . In general,
indicates how the CheY** protein influences the diffusion rate as the production of
CheY** increases. A deviated a shows that the diffusion rate could change as the production
of CheY** protein remains the same, which means there are other factors that influence the
diffusion rate. Figure 7.7(a) shows that the diffusion rate decreases intensively at high GFP
production. Thus, we consider the mass production of GFP protein that causes the diffusion
rate to drop faster than expected.
Table 7.3. Members and parameters of the “parts library”
Classes Parts Parameters

Terminator None m=1
Terminator B0014 m = 28.5
Terminator B0015 m = 111
RBS B0032 n=1
RBS B0031 n = 1.08
CheY CheY** p=1
CheY CheY wild type p = 3.66
CheY CheY D13K p = 0.68
In conclusion, we confirm that Eq. (7.8) can properly depict the characteristic of the
circuit depending on the composed parts. From the aforementioned comparison, the parts
used in this study can be organized into a “parts library” in Table 7.3. Depending on this parts
library, a 3 × 2 × 3 components library can be constructed without being individually verified
by experiments. Further, the components library can be expanded efficiently by adding new
members in the parts library.
Design Strategy for the Synthetic Circuit
In this section, we propose a design procedure to construct synthetic circuits for searching
for a specific concentration of stimulus by another promoter. The flow chart is shown in
Figure 7.11.
At the beginning, one can choose to construct a new library for the new promoter (the
right route) or use the existing library and model provided (the left route). In this study,
components were identified according to component activation level (CAL). Despite changes
in transcription efficiency due to the new promoter or environment, the relationship between
CAL and the diffusion rate is likely to remain the same. Therefore, we can assume the
identification of the components is still suitable for different situations.
If one chooses to use the existing library, the first task is to identify the parameters of the
new promoter. We recommend that one can identify parameters using the same RBS-GFP as
we did. This can avoid the coupling effect in an operon (Levin-Karp et al., 2013), and the
difference in RBS at the first position may affect the efficiency of RBS at the second position.
The parameter identification of the promoter provides sufficient information for establishing
the “stimulus vs. CAL” model. Combined with the model in previous, the adequate
component can be selected by choosing the proper parts in the “parts library” to satisfy the
following minimization problem:
(7.9)
where denotes the CAL induced by the desired concentration (I) of the stimulus,
which is obtained by the “stimulus vs. CAL” model in Eq. (7.7). Rearranging Eq. (7.8) yields
, which represents the CAL at the saturation point (shown in Figure 7.7(b)) of the
component. By solving the above minimization problem, once the bacteria enter an area
containing the desired stimulus concentration, the chosen brake component will attain its
saturation point and deprive the bacteria of their moving ability (reduced to 10% of their
initial diffusion rate).
Indeed, the effective region (shown in Figure 7.7(b)) of the chosen component should not
cover the basal leakage of the promoter, which can be expressed as follows:
(7.10)
Figure 7.11. Flow chart for designing a synthetic circuit that can search for a specific stimulus of a
specific concentration. This flow chart introduces the recommended procedure for designing a synthetic
circuit used to search for a specific stimulus of a particular concentration. According to the flow chart, a
design example is illustrated.
Therefore, we need to solve the constrained optimization problem in Eq. (7.9) subject to
Eq. (7.10) from the parts library in Table 7.3 for the synthetic circuit. Rearranging Eq. (7.8)
yields which represents the CAL at the cut-off of the effective region of the
component. denotes the basal leakage of the promoter, which is obtained by
with the stimulus concentration . If the component fails to meet the above condition, it
will be activated consistently even if no stimulus is detected.
To demonstrate the above design procedure, we take the case of designing a synthetic
circuit to search for 10-9 M AHL as an example. First, the characteristics of the plux promoter
system have to be identified by experiment to establish the mathematical model for the
promoter system. The result is given in Figure 7.15(a). Based on this information, for all the
components, the bacterial population diffusion rate at all AHL concentrations can be
simulated, as shown in Figure 7.8(c). Further, the adequate component can be composed by
choosing the proper parts of the “parts library” in Table 7.3 to satisfy the minimization
problem in Eq. (7.9). After solving the constrained optimization problem in (7.9) and (7.10),
the component B14B32CheYD13K is the most effective component for the synthetic circuit.
For B14B32CheYD13K, , which is close to 1.7×105,
and . Therefore, the component
B14B32CheYD13K is a good choice for constructing the synthetic circuit. Finally, the
characteristic of the circuit were measured and verified in different AHL concentrations. The
result is given in Figure 7.12 to confirm that bacteria almost stop as the concentration of AHL
rises to 10-9 M.
Figure 7.12. The simulated and experimental result of the component chosen to search for 10 -9 M AHL.
Following the recommended design procedure, by solving the constrained optimization Eq. (7.9) and
Eq. (7.10), a standard circuit with the component B14B32CheYD13K was constructed to search for 10 -9
M AHL. The simulated characteristic of the circuit is shown by a blue line, whereas the experimental
result is shown by red stars.
After selecting the proper components, one may construct the complete circuits and
observe the relationship between stimulus concentration and the bacterial population diffusion
rate by microfluidic device quantitatively or simply by microscope qualitatively. If the
circuits operate as expected, the genetic circuits that are able to search for a specific
concentration of stimulus can be constructed. However, we cannot guarantee that the
components we identified operate the same in different environments or bacteria. In these
cases, a new components library has to be established.
Utilizing the model, one can construct a new component library efficiently with the
following procedure. First, set up the standard circuit and component (ex: conRFP +
LuxTemplate + XB32). Second, construct the “parts library” by measuring m, n, and p of
each terminator, RBS, and CheY versions. Finally, simulate “CAL vs. diffusion rate” by the
model and construct the components to ensure m, n, and p are consistent. If not, one can try
another standard circuit composition or choose other parts. Following these steps, the library
can be established efficiently. For example, one can construct a component library containing
64 components using 4 terminators, 4 RBSs, and 4 CheY mutants. Further, by the comparison
method, only 10 components (1 standard component, 3 for measuring terminators, 3 for
measuring RBSs, and 3 for measuring CheY mutants) must be identified to obtain all m, n,
and p values. Once the library is constructed, one can design the synthetic circuits following
the above mentioned procedure (the left route).
7.4. DISCUSSION
Compared with previous studies, we improved the searching efficiency by adopting a
search-then-locate strategy rather than the stop-then-trace strategy (Derr, Boder, and Goulian
2006, Topp and Gallivan 2007). Proven by the swarm assays and microfluidic devices, E. coli
show sufficient mobility to navigate the environment until they detect adequate amounts of
AHL. Further, we modified our circuits design to create components with various
sensitivities. These components can be chosen to make bacteria search for specific
concentrations of molecules. This solved the problem that the search-then-locate strategy has
a limited searching range. That is, if the molecule is too dense, the bacterial population will
stop far away from the source. By choosing the appropriate components for systematic
circuits, we can not only solve this limited searching range problem, but also provide other
applications. For example, the bacteria can be controlled so as to not approach specific areas
or other bacterial populations by choosing considerably different sensitivity components.
The components were verified qualitatively by swarm assays and quantitatively by
microfluidic devices. The swarm assay was originally adopted for quantifying the diffusion
rate in this chapter. However, the chemotaxis phenomenon of the strain RP437 complicates
the diffusion calculation. As a result, we selected the strain RP5232, a non-chemotactic
smooth swimming strain, for the swarm assay. Nevertheless, the strain was immobile in the
semi-solid agar because smooth swimming makes E. coli susceptible to becoming trapped in
agar. Previous studies have shown that bacteria spread effectively in agar when tumbles are
frequent. However, the bacteria that tumble incessantly also fail to spread effectively (Zhu
et al., 1996). The swarm assay is not suitable for measuring diffusion rate in this study due to
diffusion rate lacking a clear relationship with tumbling in agar, where tumbling is the main
factor influencing diffusion.
As a result, a microfluidic device is used to measure diffusion rate. RP5232 is selected to
carry the synthetic circuits because of its non-chemotactic properties. Without being
influenced by the chemotaxis phenomenon, the main factor that influences bacterial
population diffusion is the amount of CheY** protein. To simplify the model, the model in
Eq. (7.5) is directly connected with the relative amount of CheY** (CAL) and the
corresponding diffusion rate. The complicated relationships among the amount of CheY**,
tumbling frequency, and diffusion rate are neglected because they have already been
discussed in previous studies (Chen and Hsu 2012, Lee et al., 2013, Hsu et al., 2014, Hsu
et al., 2015). Moreover, the simplified model is still confirmed by experimental result.
Based on this model, the efficiencies of all parts (terminators, RBS, wild type CheY, and
CheY mutant) used in the components were calculated. Compared to previous studies, the
efficiencies of these parts are not consistent. For example, the terminator B0014 and B0015 in
a previous study have a termination efficiency of 60.4% and 98.4%, respectively. However,
the efficiencies (calculated by 1-1/m) of B0014 and B0015 are 96.5% and 99.1% in this
study. Another previous study (Ahmed and Stocker 2008) shows that RBS B0032 is stronger
than B0031; in this study, B0031 is slightly stronger than B0031 ( ),
suggesting the efficiency of these parts depends significantly on the context. That is, the
identification of the parts should be measured and compared with complete circuits, as we
did. Identifying each part separately is not recommended for establishing the “parts library” in
this study.
7.5. CONCLUSION
In this chapter, we engineered synthetic circuits for bacteria to search for specific
molecules using a systematic synthetic biology method. Under the control of the synthetic
circuits, the bacteria can navigate the environment until they detect specific molecules and
stop at a particular concentration in the environment. Particularly, components library and
mathematical models were constructed to facilitate the systematic design of synthetic circuits
that prompt bacteria to search for a specific concentration of molecule. These engineered
bacteria may be applied to many situations, such as searching out and destroying herbicides
(Sinha, Reyes, and Gallivan 2010) or invading cancer cells (Anderson et al., 2006). Further,
we anticipate that controlling the spatial distribution of the bacteria population is beneficial
for creating bacteria consortia, which can cooperate efficiently. Well-cooperating and
engineered consortia have the potential for carrying out complex functions normally
unachievable for single strains. We hope that our systematic design method can expand the
field of synthetic biology and boost tangible advances in synthetic biotechnology.
7.6. APPENDIX
Appendix A: Supplement Tables
Table 7.4. Primers used in this study
Gene or Mutant Primer sequences

Forward:
cheY TATGAATTCGCGGCCGCTTCTAGAATGGCGGATAAAGAA
Reverse:
ATACTGCAGCGGCCGCTACTAGTTCACATGCCCAGTTT
Forward:
D13K TTTTTGGTTGTGGATAAATTTTCCACCATGCGA
Reverse:
TCGCATGGTGGAAAATTTATCCACAACCAAAAA
Forward:
Y106W CGGGGGCCAGTGGCTGGGTGGTGAAGCCATTTA
Reverse:
TAAATGGCTTCACCACCCAGCCACTGGCCCCCG
Appendix B: Supplement Figures
Figure 7.13. Results of the bacteria population diffusion rate measured by the
microfluidic device. Figure 7.13 shows all the bacterial diffusion rate experiments in this
study. Several AHL concentrations were tested for each component. The calculated values of
over time are shown in the figure to obtain the diffusion rate by linear regression.
For each component, the average of the diffusion rate at each AHL concentration is
summarized in the last figure and compared with the result of the “con” component. All the
components tested in this study are listed.
A. con
B. XB31
C. XB32
D. B14B31
E. B14B32
F. B15B31
G. B15B32
H. XB32CheY
I. B14B32CheY
J. XB32CheY D13K
K. B14B32CheY D13K
Standard circuit +”con” component

Standard circuit +”XB31” component

Standard circuit +”XB32” component
Standard circuit +”B14B31” component

Standard circuit + “B14B32” component


Standard circuit + “XB32CheY” component

Standard circuit + “B14B32CheY” component

Standard circuit + “XB32CheY D13K” component

Standard circuit + “B14B32CheY D13K” component

Figure 7.14. Identification of the parameters in Eq. (7.5) of each brake component. For each
component, the parameters a and b are identified from the experimental data by genetic algorithm. The
results are shown in the figure. Red lines indicate the identified results, whereas blue stars are the
experimental data.
“con”
“XB31” “XB32”
“B14B31” “B14B32”
“B15B31” “B15B32”
“XB32CheY” “B14B32CheY”
“XB32CheYD13K” “B14B32CheYD13K”
Figure 7.15. GFP expression levels of each brake component. Figure (a) is the GFP expression level of
the standard circuit with the control component “con” at different AHL concentrations. In (b), the
results of all components are summarized in four figures classified by the added AHL concentration.
(a)
(b)
Figure 7.16. Selecting individuals with high moving ability on semi-solid agar by swarm assays. Figure
(a) shows the results of the swarm assay without bacterial selection. By collecting the bacteria from the
peripheral of the colony, individuals with higher moving ability on semi-solid agar plate are obtained.
Figure (b) is the results of the swarm assays using the chosen bacteria.
(a)
(b)
Appendix C: Materials and Methods
a. Strains and growth media
Escherichia coli strain RP437 was used as a wild type in this study. RP5232 is a
derivative of RP437, which shows a deletion of the CheY gene. Both RP437 and RP5232
were kindly provided by J. S. Parkinson. Escherichia coli strain MG1655 was used as the
source of some genes used in this study. Details of these E. coli strains are listed as follows:
MG1655 F- lambda-ilvG-rfb-50 rph-1

RP437 thr(Am)-1 leuB6 his-4 metF(Am)159 eda-50 rpsL136 [thi-1 ara-14 lacY1
mtl-1 xyl-5 tonA31 tsx-78]
RP5232 (cheY) Δm60-21 thr(Am)-1 leuB6 his-4 metF(Am)159 eda-50 rpsL136 [thi-
1 ara-14 lacY1 mtl-1 xyl-5 tonA31 tsx-78]
Two mediums, Luria-Bertani broth (LB) and M9 minimal medium, were used for
cultivation of E. coli in this study. LB was prepared by dissolving 20 g LB powder (BD) in 1
L water supplemented with 2 mM kanamycin. M9 minimal medium was prepared by
dissolving 11.28 g M9 minimal salt 5X (BD) and 2 g casamino acid (BD) powder in 1 L
water supplemented with 2 mM kanamycin, 0.2% glucose, 2 uM thiamine HCl, and 2 mM M
magnesium sulfate.
b. Genetic circuit construction
All circuit constructions were assembled following the Biobrick assembly method
(http://parts.igem.org/Help:Assembly/3A_Assembly) and most of the functional parts were
obtained from the Registry of Standard Biological Parts (http://parts.igem.org). The Biobrick
sequences including promoters, ribosome binding sites, and terminators we used are listed in
Table B. The CheY gene was obtained from E. coli strain MG1544 genomic DNA. Two
mutants were modified by PCR. All the primers are listed in Table A. The Biobrick Vector
pSB1K3, a high copy number plasmid, was used to characterize all the genetic circuits in this
study.
c. Swarm assays
Two mediums, LB and M9, were used to prepare the 0.25% semi-solid agar plate. Plates
made by LB medium were used for simply verification of the basic circuit, while plates made
by the M9 medium were used to compare brake components.
For the assay with its plate made by LB medium, E. coli (MG1655) were cultivated in LB
medium at 37ºC overnight and diluted to O.D. 0.05 before being dropped on the semi-solid
LB agar plate. After being cultivated for 16 hours, one microliter of 10-3 M, 10-4 M, or 0 M
AHL was dropped beside the E. coli colonies. The colonies were then cultivated overnight
and observed.
For the assay with its plate made by M9 medium, E. coli (RP437) were cultivated in M9
medium at 30ºC overnight and diluted (1:100) to fresh M9 minimal medium. After being
cultivated for 4 hours, the medium was centrifuged at 3000 rpm for 5 minutes. The pellet was
re-suspended in fresh M9 minimal medium to an OD around 0.2 and dropped on the semi-
solid M9 agar plate. Simultaneously, 1.5 microliter of specific concentration of AHL was
dropped 4 centimeter away from the bacteria as shown in Figure 7.4. The colonies were
cultivated for 20 hours before being observed.
Notably, the bacteria used in the swarm assay needed to be selected previously owing to a
greater variability observed between individuals shown in Supplementary Figure 7.16(a). By
picking the bacteria from the periphery of the colonies, individuals with better moving ability
on the agar were obtained. As shown in Supplementary Figure 7.16(b), the selected
individuals not only showed better movement ability on the agar, but also showed little
variability among each other. Further, by testing the GFP expression level in response to
different AHL concentrations, we demonstrate that there was no significant difference in
circuit performance between individuals before and after the screening (data not shown). As a
result, E. coli was chosen previously in all the swarm assays.
d. Fabrication of the microchannel
The devices were fabricated using standard soft-lithography techniques. First, negative
photoresist (SU-8-2050) of 50 um depth was laid on a 3-inch silicon wafer by spin coating at
3000 rpm twice (8 sec and 45 sec). The wafer was baked twice, at 65 ºC for 3 minutes and at
95 ºC for 6 minutes. The wafer was then covered with the mask, a transparency film with the
pattern shown in Figure 7.5(a), and exposed to ultraviolet light using the mask aligner (Karl
sussMJB3) for 65 seconds to polymerize the exposed region of photoresist. The wafer with
molded pattern was then baked twice, 65 ºC for 1 minute and 95 ºC for 8 minutes, and
washed by the developer solution SU-8 for 6 minutes. Unexposed photoresist was washed
away, which gave an opposite pattern of Figure 7.5(a) with embossed channels.
Polydimethylsiloxane (PDMS, RTV-615, Momentive, Inc., solution A: solution B = 10:1,
degassed for 1 hour in a vacuum) was laid on the wafer and hardened by baking at 80ºC for 1
hour to produce a replica with concave channels. A thin layer of silane was laid in advance on
the wafer by vaporization to assist the separation of the wafer and the hardened PDMS to
avoid the wafer cleaving when peeling from the PDMS. Before sealing the PDMS against a
glass microscope slide, holes were poked at the inlets and outlets positions for tubing. Finally,
both PDMS and the glass microscope slide were put in the plasma cleaner (Harrick Plasma
Cleaner PDC-32, 18W, 1 torr) for 45 seconds and combined together, which formed a
covalent bond that sealed both of their surfaces.
e. Diffusion experiment procedure and GFP expression assay
E. coli was cultivated in M9 minimal medium at 30 ºC on an orbital shaker (250 rpm)

overnight and diluted (1:100) into a M9 minimal medium containing different concentrations
of chemicals of interest. After being cultivated for 4 hours, the medium was centrifuged at
3000 rpm for 5 minutes. The pellet was re-suspended in chemical-added M9 minimal medium
to an OD around 0.15 to 0.3 (depending on the fluorescent expression level, the weaker the
fluorescent presented, the higher the density needed). E. coli was then cultivated for 20
minutes again to rewarm the medium and reduce the influence of centrifugation. Just before
and after the diffusing rate experiment, the OD and GFP expression levels of the E. coli
population were measured by an Elisa Reader (BioTek, Synergy H1 Multi-Mode Reader;
excitation: 490 nm; emission: 530 nm; gain 50 and 100).
Before the measurement, channels of the microfluidic device were soaked in M9 minimal
medium containing 10 mg/mL bovine serum albumin for 30 minutes, which can reduce the
adhering of E. coli to PDMS during the experiment. The channels were then washed by
chemical-added M9 minimal medium to set up the experimental environment. A M9 minimal
medium with added chemicals and E. coli were pushed into the channel from inlet A and B
respectively by a pressure of 4 psi controlled by a computer-controlled gas valve. As the
pressure shut down, the flow stopped in seconds. Bacteria were observed by a fluorescent
microscope using 10× objective. Images were taken (ANDOR Technology, Zyla sCMDS,
2560 × 2160 pixels, 2 bytes per pixel) for every 30 seconds in 5 minutes. Each image
contained 25 continuous frames whose exposure is 200 milliseconds. For every concentration,
three repeats of E. coli from different colonies were tested in 45 minutes.
f. Data acquisition and processing
Frames at each time point were filtered to remove salt and pepper noise from the frames.
To obtain bacterial trajectories, each frame was subtracted from the following one to remove
the background and obtain a cleaner image. The distribution of the trajectories was then
calculated by counting the pixel numbers at each region, which was at different distances
from the center axis. Each region was parallel to the center axis and covered a 50-pixel-wide
area. For each sample, the center axis was calculated in advance from the first frame of the
first time point. The distribution line B(x,t) was fitted to Eq. (7.4) using gene algorism for
1000 runs, which yielded the value of . Diffusion rate was acquired by handling
values of over time by linear regression.
REFERENCES
Ahmed, Tanvir, Thomas S Shimizu, and Roman Stocker. 2010. “Microfluidics for bacterial
chemotaxis.” Integrative Biology 2 (11-12):604-629.
Ahmed, Tanvir, and Roman Stocker. 2008. “Experimental verification of the behavioral
foundation of bacterial transport parameters using microfluidics.” Biophysical Journal 95
(9):4481-4493.
Anderson, J Christopher, Elizabeth J Clarke, Adam P Arkin, and Christopher A Voigt. 2006.
“Environmentally controlled invasion of cancer cells by engineered bacteria.” Journal of
molecular biology 355 (4):619-627.
Chen, Bor-Sen, and Chih-Yuan Hsu. 2012. “Robust synchronization control scheme of a
population of nonlinear stochastic synthetic genetic oscillators under intrinsic and
extrinsic molecular noise via quorum sensing.” BMC systems biology 6 (1):1.
Derr, Paige, Eric Boder, and Mark Goulian. 2006. “Changing the specificity of a bacterial
chemoreceptor.” Journal of molecular biology 355 (5):923-932.
Dyer, Collin M., Michael L. Quillin, Andres Campos, Justine Lu, Megan M. McEvoy,
Andrew C. Hausrath, Edwin M. Westbrook, Philip Matsumura, Brian W. Matthews, and
Frederick W. Dahlquist. 2004. “Structure of the Constitutively Active Double Mutant
CheYD13K Y106W Alone and in Complex with a FliM Peptide.” Journal of Molecular
Biology 342 (4):1325-1335. doi: http://dx.doi.org/10.1016/j.jmb.2004.07.084.
Goldberg, Shalom D, Paige Derr, William F DeGrado, and Mark Goulian. “Engineering
single-and multi-cell chemotaxis pathway in E. coli.” Molecular Systems Biology 5.
Hazelbauer, Gerald L, Joseph J Falke, and John S Parkinson. 2008. “Bacterial
chemoreceptors: high-performance signaling in networked arrays.” Trends in
biochemical sciences 33 (1):9-19.
Hsu, Chih-Yuan, Zhen-Ming Pan, Rei-Hsing Hu, Chih-Chun Chang, Hsiao-Chun Cheng, Che
Lin, and Bor-Sen Chen. 2015. “Systematic biological filter design with a desired I/O
filtering response based on promoter-RBS libraries.” IEEE/ACM Transactions on
Computational Biology and Bioinformatics 12 (3):712-725.
Hsu, Chih-Yuan, Tsu-Chun Yu, Ling-Jiun Lin, Rei-Hsing Hu, and Bor-Sen Chen. 2014.
“Systematic approach to Escherichia coli cell population control using a genetic lysis
circuit.” BMC systems biology 8 (Suppl 5):S7.
Hwang, In Young, Mui Hua Tan, Elvin Koh, Chun Loong Ho, Chueh Loo Poh, and Matthew
Wook Chang. 2014. “Reprogramming Microbes to Be Pathogen-Seeking Killers.” ACS
Synthetic Biology 3 (4):228-237. doi: 10.1021/sb400077j.
Keller, Evelyn F, and Lee A Segel. 1971. “Model for chemotaxis.” Journal of theoretical
biology 30 (2):225-234.
Lee, Yi-Ying, Chih-Yuan Hsu, Ling-Jiun Lin, Chih-Chun Chang, Hsiao-Chun Cheng, Tsung-
Hsien Yeh, Rei-Hsing Hu, Che Lin, Zhen Xie, and Bor-Sen Chen. 2013. “Systematic
design methodology for robust genetic transistors based on I/O specifications via
promoter-RBS libraries.” BMC systems biology 7 (1):1.
Levin-Karp, Ayelet, Uri Barenholz, Tasneem Bareia, Michal Dayagi, Lior Zelcbuch, Niv
Antonovsky, Elad Noor, and Ron Milo. 2013. “Quantifying translational coupling in E.
coli synthetic operons using RBS modulation and fluorescent reporters.” ACS synthetic
biology 2 (6):327-336.
Lin, Benjamin, and Andre Levchenko. 2012. “Microfluidic technologies for studying
synthetic circuits.” Current opinion in chemical biology 16 (3):307-317.
Mishler, Dennis M, Shana Topp, Colleen MK Reynoso, and Justin P Gallivan. 2010.
“Engineering bacteria to recognize and follow small molecules.” Current opinion in
biotechnology 21 (5):653-656.
Rivero, Mercedes A, Robert T Tranquillo, Helen M Buettner, and Douglas A Lauffenburger.
1989. “Transport models for chemotactic cell populations based on individual cell
behavior.” Chemical engineering science 44 (12):2881-2897.
Scharf, Birgit E, Karen A Fahrner, and Howard C Berg. 1998. “CheZ has no effect on
flagellar motors activated by CheY13DK106YW.” Journal of bacteriology 180
(19):5123-5128.
Schuster, Martin, Walid N Abouhamad, Ruth E Silversmith, and Robert B Bourret. 1998.
“Chemotactic response regulator mutant CheY95IV exhibits enhanced binding to the
flagellar switch and phosphorylation‐dependent constitutive signalling.” Molecular
microbiology 27 (5):1065-1075.
Sinha, Joy, Samuel J Reyes, and Justin P Gallivan. 2010. “Reprogramming bacteria to seek
and destroy an herbicide.” Nature chemical biology 6 (6):464-470.
Sourjik, Victor, and Ned S Wingreen. 2012. “Responding to chemical gradients: bacterial
chemotaxis.” Current opinion in cell biology 24 (2):262-268.
Taghavi, S, M Mergeay, D Nies, and D Van der Lelie. 1997. “Alcaligenes eutrophus as a
in microbiology 148 (6):536-551.
Tindall, Marcus J, SL Porter, PK Maini, G Gaglia, and Judith P Armitage. 2008. “Overview
of mathematical approaches used to model bacterial chemotaxis I: the single cell.”
Bulletin of mathematical biology 70 (6):1525-1569.
Topp, Shana, and Justin P Gallivan. 2007. “Guiding bacteria with small molecules and
RNA.” Journal of the American Chemical Society 129 (21):6807-6811.
Weiss, Lucien E, Jonathan P Badalamenti, Lane J Weaver, Anthony R Tascone, Paul S
Weiss, Tom L Richard, and Patrick C Cirino. 2008. “Engineering motility as a
phenotypic response to LuxI/R‐dependent quorum sensing in Escherichia coli.”
Biotechnology and bioengineering 100 (6):1251-1255.
Zhu, Xiangyang, Charles D Amsler, Karl Volz, and Philip Matsumura. 1996. “Tyrosine 106
of CheY plays an important role in chemotaxis signal transduction in Escherichia coli.”
Journal of bacteriology 178 (14):4208-4215.
Chapter 8
CONSTRUCTION OF PROMOTER-RBS
LIBRARIES FOR THE CYANOBACTERIUM
SYNECHOCOCCUS SP. PCC 7942
AND THEIR APPLICATIONS TO SYSTEMATIC
SYNTHETIC CIRCUIT DESIGN TO MATCH
USER-ORIENTED SPECIFICATIONS
ABSTRACT
In the temperate to tropical oceans, Synechococcus is one the most notable
constituents of freshwater phytoplankton. The Synechococcus sp. PCC 7942 purely
depends on the generation of photosynthetically-derived energy for growth. Because
most algae are naturally transformable, they can provide an efficient way to produce
genetically engineered circuits. To make the design more easily, it is necessary to provide
a systematic approach for a synthetic genetic circuit to achieve user-oriented
specifications. A promoter-RBS library is set up in cyanobacterium Synechococcus sp.
strain PCC 7942. The library is based on promoter-RBS strengths found through
experimental data and allowed for the design of various applications in cyanobacteria. On
the basis of the library, a design procedure is developed to select an adequate set of
components to achieve a robust synthetic genetic circuit with the desired behaviors,
despite the effects of intrinsic variations and external noise. It is helpful to reduce the
number of trial-and-error experiments in selecting the adequate component set for genetic
circuits in cyanobacteria. Consequently, the proposed library-based search method can
help a user select an adequate set of promoter-RBS parts to realize various applications in
the rapidly developing synthetic biology field using blue-green algae.
8.1. INTRODUCTION
In the temperate to tropical oceans, Synechococcus is one the most notable constituents of
freshwater phytoplankton. The Synechococcus sp. PCC 7942 purely depends on the
generation of photosynthetically-derived energy for growth. Because most algae are naturally
transformable, they can provide an efficient way to produce genetically engineered circuits.
The Synechococcus sp. PCC 7942 has been generally used as a model organism in
biochemical, physiological, and genetic studies of valuable cellular processes, such as carbon-
concentrating mechanisms (Chen et al., 2012, Maeda et al., 2000, Omata et al., 2001),
responses to iron (Durham et al., 2003, Sandstrom et al., 2002), and nitrate metabolism
(Sakamoto and Bryant 1999, Lee, Vazquez-Bermudez, and de Marsac 1999, Vazquez-
Bermudez et al., 2002), as well as adaptation to variations in light intensity (Schwarz and
Grossman 1998) and ambient temperatures (Sane et al., 2002, Sarcina, Tobin, and Mullineaux
2001). Cyanobacteria are also useful model organisms for studying plant photosynthesis, and
their potential is enhanced by the available genomic sequences and genetic tools (Koksharova
and Wolk 2002, Abe et al., 2014, Huang and Lindblad 2013).
Large numbers of genetic tools and engineering approaches have been and still are being
developed for prokaryotes, such as cyanobacteria. Recently, synthetic scientists are forced to
find interchangeable parts that can be validated as construction units and assemble devices. A
registry of Standard Biological Parts started at MIT maintains and contains over 3400
BioBrick standard parts (http://parts.igem.org/Main_Page). The ability to quickly and reliably
engineer biological systems from libraries of standard interchangeable parts is one trademark
of modern technology (Canton, Labno, and Endy 2008). On the mechanical level, individual
BioBricks (as the MIT group calls the parts) can be fabricated and stored separately, then later
assembled together to compose larger pieces of DNA sequence (Gibbs 2004). On a functional
level, each element sends and accepts standard biochemical signals. However, engineering a
synthetic gene circuit to produce a desired behavior still remains an acute problem, due to the
views available to the conventional design strategies for assembling devices usually by
intuition alone (Ellis, Wang, and Collins 2009).
In order to employ these BioBrick parts to systematically engineer a gene circuit,
synthetic biologists need more suitable component libraries. In general, protein levels are
commonly controlled by the modulation of transcription and translation initiation rates.
Transcription is executed by the RNA polymerase holoenzyme, comprised of a core enzyme
and a sigma (σ) factor, which confers the promoter selectivity (Espah Borujeni,
Channarasappa, and Salis 2014, Imamura and Asayama 2009, Camsund and Lindblad 2014).
Transcription rates are changed by the varied kinetic strength of promoters and by modulating
the degree of repression with inducers, while translation rates are controlled by changing the
kinetic intensity and accessibility of ribosome binding sites (RBSs) (Balzer et al., 2013,
Vellanoweth and Rabinowitz 1992). Since the promoter and RBS both affect protein
expression, the production of protein is determined by the promoter and RBS as a promoter-
RBS unit, where the kinetic strength of promoter-RBS component is identified with
experimental data. One can therefore use kinetic parameters as component libraries to design
genetic circuits with desired behavior in cyanobacteria.
On the basis of the libraries, a genetic circuit for cell population control is given to
illustrate the proposed design method. The design contains a light-driven proton pump and
ictB gene to control cyanobacteria cell density. The light-driven proton pump can enhance
cyanobacteria survival in adverse conditions. In fact, the BR protein could influence ATP
concentration; thus, cell density will increase during NaCl stress. Besides, the ictB gene also
enhances cell growth of the Synechococcus sp. PCC 7942 thereby to control cell population
density. To make the design easier, a detailed mathematical model of the synthetic genetic
circuit is described. According to user-oriented specifications, the design with a desired
behavior is constructed by selecting adequate components from corresponding libraries. In
Construction of Promoter-RBS Libraries … 171
general, a long computation time is required when the component libraries become large.
Hence, the genetic algorithm (GA)-based design method proposed here provides a simple and
useful tool that saves time in evaluating and selecting components.
The main focuses of this chapter are threefold: (a) On the basis of the promoter-RBS
kinetic strengths, we establish a component library. (b) The synthetic genetic circuit for cell
population control can be constructed according to user-oriented specifications by evaluating
and selecting adequate components from the corresponding promoter-RBS library to achieve
a desired cell density. (c) The proposed genetic algorithm (GA)-based method could provide
synthetic biologists with a useful tool to design a variety of applications in cyanobacteria.
8.2. METHODS
Construction of Promoter-RBS Library
The fundamental framework of a synthetic genetic circuit consists of a promoter, a

ribosome binding site (RBS), a functional gene, and a terminator. Since the promoter and
RBS both affect protein expression, the production of protein is determined by the promoter
and RBS as a promoter-RBS unit, where we identify the kinetic strength of promoter-RBS
components with experimental data. Consider the protein biosynthesis of the cyanobacteria
genetic circuit, which is regulated by the promoter-RBS, diluted due to the cell growth, and
degraded due to protease. By denoting the concentration of the reporter gene as x(t), the
protein expression produced by promoter-RBS can then be defined as follows:
x(t )  P( PM , Pm , TF , I )  (   )  x(t ) (8.1)
whereγand μ denote the protein degradation rate and protein dilution rate with respect to the
protein, respectively, and P denotes the promoter-RBS kinetic strength of the promoter-RBS
component. The general formula of the regulation function can be written as P(PM,Pm,TF,I)
with the minimum and maximum promoter-RBS strengths Pm and PM , corresponding
transcription factor concentration TF, and the related inducer concentration I.
A constitutive promoter-RBS part Ci can express downstream gene constitutively, with
no influence by any transcription factor, with the construction shown in Figure 8.1. Therefore,
the regulation activity of the constitutive promoter-RBS component Ci can be represented by
a promoter-RBS regulation function as follows:
PC ( PM ,i ,0,0,0)  PM ,i (8.2)
where i denotes the ith constitutive promoter-RBS component from the corresponding library.
PM,i denotes the promoter-RBS strength of the ith constitutive promoter-RBS component in
Table 8.1. The constitutive promoter-RBS components here are constructed by two
constitutive promoters, namely, J23101 and psbAI as well as six RBSs, namely, RBS1, RBS2,
RBS3, B0031, B0032 and B0034. Then based on the steady-state model in (8.1), the
identification results are used to construct library and listed in Table 8.1, from which some
constitutive promoter-RBS components are to be selected to design a synthetic genetic circuit

for population density control of cyanobacteria.
Figure 8.1. A representative scheme for characterizing constitutive promoter-RBS components. The
reporter gene placed downstream of the promoter-RBS component is used to measure the fluorescence
of various promoter-RBS components.
Table 8.1. Constitutive promoter-RBS component library Libconst
Index Component PM,i

C1 psbAI-RBS1 4.6913
C4 psbAI-B0032 0.9557
C5 psbAI-B0033 0.6766
C6 psbAI-B0034 6.4481
C7 J23101-RBS1 12.0136
C8 J23101-RBS2 8.3429
C9 J23101-RBS3 14.6317
C10 J23101-B0032 8.6261
C11 J23101-B0033 2.8068
C12 J23101-B0034 21.4532
For constructing the constitutive promoter-RBS library, the detailed library building
process is divided into four steps in Figure 8.2: (I) Select a promoter and an RBS to construct
a promoter-RBS part Ci, and place them upstream of the EYFP; (II) measure the fluorescence
of the selected promoter-RBS part Ci; (III) estimate the dilution rate μ by = μD, where D
enotes the value of O.D.730 in the experiment and then estimate the promoter-RBS intensity
PM,i using the system identification technique; and (IV) construct the kinetic intensities of the
constitutive promoter-RBS library as shown in Table 8.1, which includes the promoter-RBS
part Ci and its identified promoter-RBS intensity PM,i.
Figure 8.2. The procedure for construction of the constitutive promoter-RBS library.
Construction of a Gene Circuit for Cell Population Control in Synechococcus

sp. PCC 7942
On the basis of the libraries, a genetic circuit for cell population control is given to
illustrate the proposed design method. The design contains a light-driven proton pump and
ictB gene to control cyanobacteria cell density as shown in Figure 8.3.
Bacteriorhodopsin (BR) is the other photosynthetic system in nature and is a
transmembrane protein found in the organism Halobacterium salinarum (Chu, Yen, and El-
Sayed 2010). It acts as a proton pump and catches solar energy to move protons across the
membrane and out of the cell. Hence, the BR-induced proton motive force is light-activated
and could drive adenosine triphosphate (ATP) synthesis as protons reenter the cell through
the ATP synthase complex. ATP, an energy-bearing molecule found in all living cells, is a
nucleoside triphosphate used as a coenzyme, and is often called the “molecular unit of
currency” of intracellular energy transfer. The resulting proton gradient is converted into
energy, such that ATP can be used as a power source in the cell to drive various mechanical
and chemical activities.
Additionally, ictB, a HCO3- transporter, was overexpressed to increase carbon
accumulation within the cell, which could enhance cell growth and photosynthesis rate (Chen
et al., 2012, Lieman-Hurwitz et al., 2003). In PCC 7942, ictB is involved in the accumulation
of inorganic carbon. Mutants with an impaired ictB gene demand a high CO2 concentration
for converting light energy into chemical energy. Another study of the high affinity
bicarbonate consumption system of Synechococcus elongatus PCC 7942 found the system
was rapidly induced only in the presence of Na+. Therefore, there was a correlation between
the appearance of the ictB-encoded protein and the capability of rapid induction (Chu, Yen,
and El-Sayed 2010). Although the role of ictB uptake in cyanobacteria is not totally
understood, past research has shown the ictB gene shows potential for stimulating crop yield.
Mathematical Model
Since this chapter utilizes the regulatory protein to control cell population density of
cyanobacteria, the relationship between cell population density and the concentration of
regulatory protein can be defined as in the following equation:
 N (t ) 
N (t )    N (t ) 1     N  N (t )  x p (t ) (8.3)
 Nm 
where N(t) and Nm denote the cell population density (O.D. 730) and the maximum cell
population density, respectively,; xp(t) and γN denote the concentration of regulatory protein
and the rate of control protein, respectively, and μ denotes the dilution rate due to cell growth.
The dynamic model of a cell population control genetic circuit with the two constitutive
promoter-RBS components shown in Figure 8.3 then is described by the following:
 x1 (t )  Pc ( PM ,i ,0,0,0)    1  1   x1 (t )


 x2 (t )  Pc ( PM , j ,0,0,0)    2  2   x2 (t )
 (8.4)
 N (t )   N  N (t ) 1  N (t ) N m    N  N (t )  ( x1 (t )  x2 (t ))

PM ,i , PM , j   Libconst
where x1(t), x2(t) are concentrations of two different regulatory proteins for BR and IctB,
respectively; γ1 and γ2are degradation rate of the different regulatory proteins, μ1 and μ2
denote the dilution rate of the different regulatory proteins, and the promoter-RBS strengths
PM,i and PM,j of the constitutive promoter-RBS components can be selected from Libconst in
Table 8.1 to achieve a desired population control in cyanobacteria..
After constructing the dynamic model of the genetic circuit for cell population control,
for the purpose of achieving the desired steady state cell population density, the dynamic
model is transformed into the steady-state models by assuming the derivatives of dynamic
models in (8.4) equal zero. The steady state models of the synthetic genetic circuit in Figure
8.3 will be given as follows:
 Pc ( PM ,i ,0,0,0)
 x1ss 
  1  1

 Pc ( PM , j ,0,0,0)
 x2 ss  (8.5)
  2  2
  
 N ss  N m 1   N x1ss   N x2 ss 
  N N 
where x1ss and x2ss denote concentrations of the two different regulatory proteins, at the
steady-state, respectively and Nss denotes cell population density at the steady state.
In general, the genetic circuit for cell population control in vivo also suffers from
extrinsic environmental noise. Thus, the equation in (8.5) should be modified as follows:
 Pc ( PM ,i ,0,0,0)
 x1ss   v1
  1  1

 Pc ( PM , j ,0,0,0)
 x2 ss   v2 (8.6)
  2  2
  
 N ss  N m 1   N x1ss   N x2 ss   v3
  N N 
where the zero mean Gaussian noises vp, p = 1,2 with variances σp2denote cellular noise in the
transcriptional and translational processes of gene expression, and v3 denotes cellular noise in
the cell population density.
Besides, the dynamic model should consider the stochastic parameter fluctuation because
biological parts are inherently uncertain in vivo due to gene expression noise, DNA mutations,
cell growth, molecular noise, and kinetic parameter estimation errors. Thus, the parameter
uncertainties in (8.6) must be considered. For example, (a) the kinetic parameters of the
promoter-RBS component in the process of transcription and translation, (b) degradation rates
of regulatory function genes, (c) dilution rate from cell growth, and (d) the rate of the control
proteins under stochastic uncertainty are as follows:
PM ,i  PM ,i  PM ,i n1 (t ), PM , j  PM , j  PM , j n1 (t )
 1   1   1n1  t  ,  2   2   2 n2  t 
1  1  1n1  t  , 2  2  2 n2  t  (8.7)
 N   N   N n3  t 
 N   N   N n3  t 
where ΔPM,i, ΔPM,j, Δγ1, Δγ2, ΔγN, and Δμ denote the standard deviations of the corresponding
stochastic parameters and ni(t), i = 1,2,3 are Gaussian noise with zero mean and unit variance
to take advantage of the different random fluctuation sources. Thus, ΔPM,i, ΔPM,j, Δγ1, Δγ2,
ΔγN, and Δμ denote the kinetic parameter variations and ni(t), i = 1,2,3 denote the random
fluctuation sources. For the robust design of a population control genetic circuit, the
parameter fluctuations in (8.7) substitute for the kinetic parameters at steady state in (8.6) so
that the synthetic genetic circuit can tolerate these kinds of kinetic parameter fluctuations in
the cyanobacterium.
Figure 8.3. Design schematic of the genetic circuit for population control in the cyanobacterium
Synechococcus sp. PCC 7942. Two constitutive promoter-RBS parts are to be selected from library in
Table 8.1 to achieve a desired population density.
Design Specifications
The purpose is to design a population control genetic circuit in Figure 8.3 by selecting a
set of suitable promoter-RBS component S = (Ci, Cj) from the library in Table 8.1 to achieve
the optimal reference tracking of cell population density. Thus, the design specifications are
given as follows.
 Specify a desired reference Nref of cell population density.

 Provide the well-characterized promoter-RBS component library listed in Table 8.1

to be selected.
 Specify the standard derivations of parameter uncertainties in (8.7) and the standard
deviations of environmental disturbances vp in (8.6) to be tolerated by the synthetic
genetic circuit in vivo.
 The cost function of mean square error between cell population density, at the steady
state, Nss and the desired reference cell population density, at steady state, Nref, is
indicated as follows:
J (S )  E ( N ss (S )  Nref )2 (8.8)
where S = (Ci, Cj) denote the combination of promoter-RBS components to be selected from
their corresponding promoter-RBS libraries.
If the cost function in (8.8) can be minimized by picking the set of promoter-RBS
components that are suitable under the design specifications, then the desired cell population
density can optimally match the specified steady state cell population density Nref under the
intrinsic parameter fluctuations and environmental disturbances in the host cell. In this study,
a GA-based library search method is provided to efficiently select the most adequate
combination from the mixed libraries and more efficiently evaluate the kinetic parameters in
the following procedure.
Design Procedure Genetic Circuit for Cell Population Control
Although exhaustive searching can minimize the cost function, this is a time-consuming
process. Finding a feasible combination of promoter-RBS components to minimize J(S) in
equation (8.8) requires long computation times, as well as trial-and-error experimentation
when component libraries are large. In this study, a genetic algorithm-based library search
method is applied to evaluate and select the most adequate promoter-RBS component set in
an efficient and effective way while searching large libraries. Thus, the synthetic genetic
circuit can robustly and optimally achieve the desired population density by using a library-
based search method combined with GA to efficiently choose the optimal set of promoter-
RBS components to satisfyy the four design specifications. The design procedure is
summarized as follows:
1. Determine the desired cell population density Nref of the cyanobacterium

Synechococcus sp. PCC 7942 to be achieved in this study.
2. Build well-characterized promoter-RBS libraries for a population control genetic
circuit of PCC 7942 by identifying EYFP values from experimental data at steady
state.
3. Select an initial combination S of promoter-RBS components.
4. The cost value J(S) for each combination S of promoter-RBS components from the
corresponding promoter-RBS libraries is evaluated. The relationship between the
fitness value F(S) and the cost value J(S) is given as follows:
1
F (S )  (8.9)
J (S )
The fitness value plays an important role in searching for the combination of promoter-
RBS components to maximize the fitness value F(S) or equivalently minimize the cost
function J(S) in equation (8.9):
1
max F ( S )  (8.10)
S min, J ( S )
S
1. Create offspring combination S by GA operators (e.g., mutation, selection, and
crossover).
2. Calculate the fitness or cost value of the new combinations S in promoter-RBS
libraries. Stop when the design target is achieved or an acceptable solution is
obtained, otherwise create the next generation and return to step 4.
8.3. RESULTS
In the section, we will demonstrate that the desired control density of cell population in
cyanobacteria can be achieved from the cell population control genetic circuit using the
proposed library-based search method. Suppose three desired steady state cell population
densities O.D. 730 are prescribed as 0.9, 0.8, and 0.75, respectively. Suppose random
parameter variations 5% to be tolerated as follows:
PM ,i  0.05  PM ,i , PM , j  0.05  PM , j ,  1  0.05   1 ,  2  0.05   2 ,

(8.11)
1  0.0  51 , 2  0.05  2 ,  N  0.05   N ,  N  0.05   N
i.e., the standard deviations of parameter fluctuations in (8.7) are allowed to be 5% of their
nominal values. Finally, it is necessary to select the most adequate set of promoter-RBS
components S from the corresponding promoter-RBS libraries to minimize the cost function
of the genetic circuit in (8.8) and approach the desired cell population density.
According to the mathematical predictions, three combinations of promoter-RBS
components are chosen to verify whether the intensities of the promoter-RBS parts would
affect the cell population as expected. For the purpose of controlling the desired cell
population, the prescribed cell population density and the most suitable sets of promoter-RBS
parts from the promoter-RBS libraries are shown in Table 8.2, in which the simulation results
are also drawn for reference and experimental data of the population control synthetic gene
circuits are also given in Figure 8.4 to compare the variances of cell population densities in
silico and in vivo. The prescribed cell population densities are 0.9, 0.8, and 0.75 and the
population densities with the corresponding promoter-RBS components are 0.923, 0.814, and
0.776, respectively. The percentage of error between the simulation results and experimental
results in Table 8.2 is defined by the absolute value of the difference between the simulation
result and the experimental result divided by the experimental result. The light-driven proton
pump can enhance cyanobacteria survival in adverse conditions. The stronger constitutive
promoter-RBS components could produce more BR protein to influence ATP concentration,
thus cell population density will increase during NaCl stress. Moreover, the ictB gene also
enhances cell growth of the Synechococcus sp. PCC 7942. This identified characterization of
a promoter-RBS component is more suitable for the design of population control genetic
circuits to achieve the desired cell population density.
Table 8.2. The experimental results to validate the simulation results
Ci Cj O.D. 730 in silico O.D. 730 in vivo Percentage Error

T1 C6 x 0.776 0.763 1.7%
T2 C6 C6 0.814 0.789 3.1%
T3 C6 C12 0.923 0.895 3.0%
(a)
(b)
(c)
Figure 8.4. The experimental results to validate the simulation results by the proposed systematic
design method with the most adequate promoter-RBS components from Table 8.1. (a) psbAIB34BR (b)
psbAIB34BR-psbAIB34ictB (c) psbAIB34BR-J01B34ictB.
8.4. DISCUSSION
Algae are developing into one of the most promising sources of long-term food,
sustainable biomass, and carbon dioxide fixation. First, algae can double their numbers every
few hours, can be harvested daily, and have the potential to produce a volume of biomass and
biofuel many times greater than our most productive crops. Second, these high algal biofuel
yields and biomasses could provide a large portion of future global transport fuel
requirements (Ullah et al., 2014). Third, autotrophic production of renewable chemicals is an
attractive approach because autotrophic algae convert solar energy into biomass via
photosynthesis. Therefore, reliance on fossil fuels would be reduced while recycling the
greenhouse gas CO2 (Shen and Liao 2012, Davis, Aden, and Pienkos 2011). Finally, algae
cultivation uses both land and water sources that are in many cases unsuitable for traditional
agriculture. As such, algae are more environmentally beneficial than terrestrial crops. One of
the key advantages of using algae as a feedstock for biofuels is that they can be used to
produce many different types of fuel. Whether it is biodiesel, biojet fuel, green gasoline, or
ethanol, algae have the ability to meet our transportation fuel needs (Hannon et al., 2010).
Current research on promoter libraries has made significant progress in quantitative
measurements. Based on knowledge on the effects of promoter architecture on mutation-
selection techniques and transcriptional activity (Hammer, Mijakovic, and Jensen 2006, Alper
et al., 2005, Cox, Surette, and Elowitz 2007, Gertz, Siggia, and Cohen 2009, Segal and
Widom 2009), promoter-RBS libraries can be extensively built. Synthetic biologists utilize
mathematical models to describe the biochemical reactions of genetic circuits and then
implement the biological gene circuits into algae by genetic engineering techniques. The
models of promoter activity are developed by considering the binding reactions of TFs to
DNA in thermodynamic equilibrium. A mathematical model for predicting the relative

mRNA translation rate with different RBS sequences has also been proposed (Kinkhabwala
and Guet 2008, Salis, Mirsky, and Voigt 2009). However, the promoter activity would still be
influenced by the RBS placed downstream of the promoter. This is also the reason why
promoter and RBS libraries cannot be built independently.
In this study, the building process is provided for promoter-RBS library, i.e., constitutive
promoter-RBS library for systematic genetic circuit design in blue-green algae. The
application of synthetic biology requires well-characterized components to precisely control
gene expression. By constructing and identifying promoter-RBS components, extensive
promoter-RBS libraries can be created to conveniently design and construct complex gene
circuits. These efforts, coupled with in silico system modeling, will serve to fast-track
synthetic biology. For building reliable models of promoter-RBS libraries, the construction
method should be provided to acquire meaningful and reasonable data. For the synthetic
genetic circuit and biological implementation, it is possible to create and refer to standard
biological parts, such as adequate promoters, RBSs, function genes, and reporter proteins.
Thus, synthetic biologists can increase the efficiency of gene circuit design through well-
characterized promoter-RBS libraries.
The proposed library-based search method can reduce the number of trial-and-error
experiments in selecting the most adequate promoter-RBS component set for synthetic gene
circuits for population density control. The performance of population control gene circuits
can then be predictably implemented by a systematic method before trial-and-error
experiments. When the mathematical models of protein expressions are properly constructed,
it enables the design of more complex synthetic circuits to conveniently meet design
specifications. By adopting the systematic method and promoter-RBS libraries for the design
of systematic genetic circuits in cyanobacteria, the biological parts with the desired
characteristics can be efficiently assembled into the synthetic gene circuit in cyanobacteria to
meet design specifications with experimental validation in vivo.
8.5. CONCLUSION
The building process is provided for the promoter-RBS library, which includes the
constitutive promoter-RBS library in cyanobacteria. Notably, a mixed library-based search
method is employed to efficiently find the most adequate set of promoter-RBS parts to
achieve the desired behavior of the synthetic genetic circuit in cyanobacteria via GA. By
developing promoter-RBS libraries and testing the designed population control genetic
circuits in PCC 7942, the proposed systematic design method based on a GA search algorithm
enables scientists to accelerate the engineering of other synthetic gene circuits in PCC 7942
and achieve desired and robust behaviors in cyanobacteria. The synthetic gene systems, which
combine dynamic regulatory models in silico and experiments in vivo, will provide systematic
approaches to a wide range of applications, such as biomass, isobutanol production, CO2
fixation, and ethanol production in cyanobacteria.
The systematic method to select the most suitable promoter-RBS set from libraries to
tolerate kinetic parameter variations, to filter the environmental noise, and to meet the desired
steady state concentration will help synthetic biologists to simplify the design procedure of
synthetic gene circuits with desired robust behavior and accelerate progress of using synthetic
biology methods with cyanobacteria. Due to the high design efficiency, larger amounts of
trail-and-error experimental processes can be avoided in selecting an adequate promoter-RBS
component to achieve the prescribed design specifications, i.e., the desired steady state and
the intrinsic fluctuations and extrinsic noise to be tolerated. With rapid advances in modern
synthesis technologies, the proposed GA-searching systematic method from promoter-RBS
libraries, which have well-defined characteristics of promoters for standard promoter-RBS
parts, could be widely applied for a more rapid and systematic design of synthetic gene
circuits in cyanobacteria in the future.
REFERENCES
Abe, K., K. Miyake, M. Nakamura, K. Kojima, S. Ferri, K. Ikebukuro, and K. Sode. 2014.
“Engineering of a green-light inducible gene expression system in Synechocystis sp
PCC6803.” Microbial Biotechnology 7 (2):177-183. doi: 10.1111/1751-7915.12098.
Alper, H., C. Fischer, E. Nevoigt, and G. Stephanopoulos. 2005. “Tuning genetic control
through promoter engineering.” Proceedings of the National Academy of Sciences of the
United States of America 102 (36):12678-12683. doi: 10.1073/pnas.0504604102.
Balzer, S., V. Kucharova, J. Megerle, R. Lale, T. Brautaset, and S. Valla. 2013. “A
comparative analysis of the properties of regulated promoter systems commonly used for
recombinant gene expression in Escherichia coli.” Microbial Cell Factories 12. doi: Artn
26 10.1186/1475-2859-12-26.
Camsund, D., and P. Lindblad. 2014. “Engineered transcriptional systems for cyanobacterial
biotechnology.” Front Bioeng Biotechnol 2:40. doi: 10.3389/fbioe.2014.00040.
Canton, B., A. Labno, and D. Endy. 2008. “Refinement and standardization of synthetic
biological parts and devices.” Nature Biotechnology 26 (7):787-793. doi:
10.1038/nbt1413.
Chen, P. H., H. L. Liu, Y. J. Chen, H. Cheng, W. L. Lin, C. H. Yeh, and C. H. Chang. 2012.
“Enhancing CO2 bio-mitigation by genetic engineering of cyanobacteria.” Energy &
Environmental Science 5 (8):8318-8327. doi: 10.1039/c2ee21124f.
Chu, L. K., C. W. Yen, and M. A. El-Sayed. 2010. “On the Mechanism of the Plasmonic
Field Enhancement of the Solar-to-Electric Energy Conversion by the Other
Photosynthetic System in Nature (Bacteriorhodopsin): Kinetic and Spectroscopic Study.”
Journal of Physical Chemistry C 114 (36):15358-15363. doi: 10.1021/jp105468x.
Cox, R. S., M. G. Surette, and M. B. Elowitz. 2007. “Programming gene expression with
combinatorial promoters.” Molecular Systems Biology 3. doi: ARTN 145
10.1038/msb4100187.
Davis, R., A. Aden, and P. T. Pienkos. 2011. “Techno-economic analysis of autotrophic
microalgae for fuel production.” Applied Energy 88 (10):3524-3531. doi:
10.1016/j.apenergy.2011.04.018.
Durham, K. A., D. Porta, R. M. L. McKay, and G. S. Bullerjahn. 2003. “Expression of the
iron-responsive irpA gene from the cyanobacterium Synechococcus sp strain PCC 7942.”
Archives of Microbiology 179 (2):131-134. doi: 10.1007/s00203-002-0503-6.
Ellis, Tom, Xiao Wang, and James J. Collins. 2009. “DIVERSITY-BASED, MODEL-
GUIDED CONSTRUCTION OF SYNTHETIC GENE NETWORKS WITH
PREDICTED FUNCTIONS.” Nature biotechnology 27 (5):465-471. doi:
10.1038/nbt.1536.
Espah Borujeni, Amin, Anirudh S. Channarasappa, and Howard M. Salis. 2014. “Translation
rate is controlled by coupled trade-offs between site accessibility, selective RNA
unfolding and sliding at upstream standby sites.” Nucleic Acids Research 42 (4):2646-
2659. doi: 10.1093/nar/gkt1139.
Gertz, J., E. D. Siggia, and B. A. Cohen. 2009. “Analysis of combinatorial cis-regulation in
synthetic and genomic promoters.” Nature 457 (7226):215-U113. doi:
10.1038/nature07521.
Gibbs, W. W. 2004. “Synthetic life.” Scientific American 290 (5):74-81.
Hammer, K., I. Mijakovic, and P. R. Jensen. 2006. “Synthetic promoter libraries - tuning of
gene expression.” Trends in Biotechnology 24 (2):53-55. doi:
10.1016/j.tibtech.2005.12.003.
Hannon, M., J. Gimpel, M. Tran, B. Rasala, and S. Mayfield. 2010. “Biofuels from algae:
challenges and potential.” Biofuels 1 (5):763-784.
Huang, H. H., and P. Lindblad. 2013. “Wide-dynamic-range promoters engineered for
cyanobacteria.” Journal of Biological Engineering 7 (1). doi: Artn 10 10.1186/1754-
1611-7-10.
Imamura, S., and M. Asayama. 2009. “Sigma factors for cyanobacterial transcription.” Gene
Regul Syst Bio 3:65-87.
Kinkhabwala, A., and C. C. Guet. 2008. “Uncovering cis Regulatory Codes Using Synthetic
Promoter Shuffling.” Plos One 3 (4). doi: ARTN e2030 10.1371/journal.pone.0002030.
Koksharova, O. A., and C. P. Wolk. 2002. “Genetic tools for cyanobacteria.” Applied
Microbiology and Biotechnology 58 (2):123-137. doi: 10.1007/s00253-001-0864-9.
Lee, H. M., M. F. Vazquez-Bermudez, and N. T. de Marsac. 1999. “The global nitrogen
regulator NtcA regulates transcription of the signal transducer PII (GlnB) and influences
its phosphorylation level in response to nitrogen and carbon supplies in the
Cyanobacterium synechococcus sp. strain PCC 7942.” J Bacteriol 181 (9):2697-702.
Lieman-Hurwitz, J., S. Rachmilevitch, R. Mittler, Y. Marcus, and A. Kaplan. 2003.
“Enhanced photosynthesis and growth of transgenic plants that express ictB, a gene
involved in HCO3- accumulation in cyanobacteria.” Plant Biotechnology Journal 1
(1):43-50. doi: DOI 10.1046/j.1467-7652.2003.00003.x.
Maeda, S., G. D. Price, M. R. Badger, C. Enomoto, and T. Omata. 2000. “Bicarbonate
binding activity of the CmpA protein of the cyanobacterium Synechococcus sp strain
PCC 7942 involved in active transport of bicarbonate.” Journal of Biological Chemistry
275 (27):20551-20555. doi: DOI 10.1074/jbc.M003034200.
Omata, T., S. Gohta, Y. Takahashi, Y. Harano, and S. Maeda. 2001. “Involvement of a CbbR
homolog in low CO2-induced activation of the bicarbonate transporter operon in
cyanobacteria.” Journal of Bacteriology 183 (6):1891-1898. doi: Doi
10.1128/Jb.183.6.1891-1898.2001.
Sakamoto, Toshio, and Donald A. Bryant. 1999. “Nitrate Transport and Not Photoinhibition
Limits Growth of the Freshwater Cyanobacterium Synechococcus Species PCC 6301 at
Low Temperature.” Plant Physiology 119 (2):785-794.
Salis, H. M., E. A. Mirsky, and C. A. Voigt. 2009. “Automated design of synthetic ribosome
binding sites to control protein expression.” Nature Biotechnology 27 (10):946-U112.
doi: 10.1038/nbt.1568.
Sandstrom, S., A. G. Ivanov, Y. I. Park, G. Oquist, and P. Gustafsson. 2002. “Iron stress
responses in the cyanobacterium Synechococcus sp PCC7942.” Physiologia Plantarum
116 (2):255-263. doi: DOI 10.1034/j.1399-3054.2002.1160216.x.
Sane, P. V., A. G. Ivanov, D. Sveshnikov, N. P. Huner, and G. Oquist. 2002. “A transient
exchange of the photosystem II reaction center protein D1:1 with D1:2 during low
temperature stress of Synechococcus sp. PCC 7942 in the light lowers the redox potential
of QB.” J Biol Chem 277 (36):32739-45. doi: 10.1074/jbc.M200444200.
Sarcina, M., M. J. Tobin, and C. W. Mullineaux. 2001. “Diffusion of phycobilisomes on the
thylakoid membranes of the cyanobacterium Synechococcus 7942 - Effects of
phycobilisome size, temperature, and membrane lipid composition.” Journal of
Biological Chemistry 276 (50):46830-46834. doi: 10.1074/jbc.M107111200.
Schwarz, R., and A. R. Grossman. 1998. “A response regulator of cyanobacteria integrates
diverse environmental signals and is critical for survival under extreme conditions.”
(18):11008-11013. doi: DOI 10.1073/pnas.95.18.11008.
Segal, E., and J. Widom. 2009. “From DNA sequence to transcriptional behaviour: a
quantitative approach.” Nature Reviews Genetics 10 (7):443-456. doi: 10.1038/nrg2591.
Shen, C. R., and J. C. Liao. 2012. “Photosynthetic production of 2-methyl-1-butanol from
CO2 in cyanobacterium Synechococcus elongatus PCC7942 and characterization of the
native acetohydroxyacid synthase.” Energy & Environmental Science 5 (11):9574-9583.
doi: 10.1039/c2ee23148d.
Ullah, K., M. Ahmad, Sofia, V. K. Sharma, P. M. Lu, A. Harvey, M. Zafar, S. Sultana, and C.
N. Anyanwu. 2014. “Algal biomass as a global source of transport fuels: Overview and
development perspectives.” Progress in Natural Science-Materials International 24
(4):329-339. doi: 10.1016/j.pnsc.2014.06.008.
Vazquez-Bermudez, M. F., J. Paz-Yepes, A. Herrero, and E. Flores. 2002. “The NtcA-
activated amt1 gene encodes a permease required for uptake of low concentrations of
ammonium in the cyanobacterium Synechococcus sp PCC 7942.” Microbiology-Sgm
148:861-869.
Vellanoweth, R. L., and J. C. Rabinowitz. 1992. “The influence of ribosome-binding-site
elements on translational efficiency in Bacillus subtilis and Escherichia coli in vivo.” Mol
Microbiol 6 (9):1105-14.
Chapter 9
A ROBUST DESIGN OF QUORUM SENSING SYMBIOTIC

ECOSYSTEMS CONTROLLED BY SMALL MOLECULE
REGULATORS THROUGH CELL-CELL
COMMUNICATION
ABSTRACT
Microbial ecosystems play key roles in maintaining biosphere and supporting life.
Research has highlighted various interactions between the environment and microbial
ecosystems, but the complexity and structure of natural microbial ecosystems still needs
investigating. In this chapter, a controllable symbiotic ecosystem is constructed to
provide different levels of mutualistic behaviors via cell-cell communication. With a
detailed mathematical model of the synthetic symbiotic ecosystem, it can interpret and
reveal the natural functions and complex interactions between the gene regulatory
networks and environmental effects. According to the model of symbiosis of ecosystem,
a design procedure is developed to select an adequate set of components from
corresponding libraries to achieve a robust synthetic symbiotic ecosystem with the
desired mutualistic behaviors, despite the effects of intrinsic variations and external
noise. The controllable symbiotic ecosystem proposed here thus may act as a framework
for constructing more complicated microbial ecosystems through cell-cell communication
in the fast growing field of synthetic biology.
9.1. INTRODUCTION
Microbial ecosystems play key roles in maintaining biosphere and supporting life (Pace
1997). Most natural microbes live with each other, compete for natural resources, share for
their metabolism, and form a dynamically complex microbial ecosystem (Hibbing et al.,
2010b, Klitgord and Segrè 2011). To reveal about natural ecosystem functions and behaviors,
synthetic biology research has often constructed genetic circuits realized in multicellular
systems (Basu et al., 2005, You et al., 2004). This field uses established artificial synthetic
microbial ecosystems with genetic backgrounds and cellular interactions to achieve their
goals. In the past few years, Balagaddé et al., constructed a synthetic Escherichia coli (E. coli)
predator-prey ecosystem with cell-cell communication to regulate gene expression and
survival between individuals (Balagaddé et al., 2008). Hu et al., proposed a synthetic

symbiotic ecosystem where two bacteria populations mutually benefited from each other and
engineered a population in which the dynamics were influenced by environmental factors (Hu
et al., 2010). Mao et al., mimicked the growth of bacteria in a resource-limited environment
by constructing a population including two species, and searched for optimal strategies in
bacterial competition (Mao, Blanchard, and Lu 2014). Researchers aim to analyze how
microbial ecosystem operate in nature and thereby engineer synthetic circuits that can mimic
the interactions and behaviors between species in microbial ecosystems.
Of the aforementioned natural behaviors, symbiosis, an interaction between two or more
different species, is a common relationship offering additional dynamics and diversity in
which participants gain advantages in finite communities. In nature, most species exist in
complex and dynamically changing environments full of interactions; they do not live in
isolation. There are several types of symbiosis, including mutualism, commensalism,
parasitism, etc., with each contributing to different complex dynamics and forming variable,
evolving microbial ecosystems (Little et al., 2008, Waters and Bassler 2005). To discuss these
complex biological behaviors, a quorum sensing symbiotic ecosystem is developed here
whose members are embedded with different regulated circuits in order to display several
mutualistic behaviors beneficial to population growth. Through a combination of two quorum
sensing systems, E. coli could compete and coexist with their neighbors, sharing nutrients,
resources, and living space (Hibbing et al., 2010a). Bacteria benefit from each other by
expressing small quorum sensing molecules to induce expression of resistance proteins,
which can protect bacteria from environmental pressures (Hu et al., 2010). The synthetic
circuits thus provide valuable information for analyzing and engineering the complex
composition and structure of symbiotic microbial ecosystems.
In this study, the major construction of the circuits consists of two types of quorum
sensing systems, the Las system and the Rhl system. LasI and RhlI, two kinds of synthase
enzymes, synthesize the autoinducers, i.e., N-3-(oxododecanoyl)- homoserine lactone(PAI-1)
and N-(butyryl)- homoserine lactone(PAI-2), which are referred to as N-acylated homoserine
lactone (AHL) (Pearson, Pesci, and Iglewski 1997). In the Las quorum sensing system, the
autoinducer synthase LasI synthesizes C4-HSL, which stimulates the LasR-encoded
transcriptional activator (Gambello, Kaye, and Iglewski 1993, Toder et al., 1994, Toder,
Gambello, and Iglewski 1991). The second quorum sensing system is the Rhl system, which
consists of the RhlR-encoded putative transcriptional activator, RhlR. When these cells
communicate with each other via intercellular signals, it can induce expression of resistance
proteins to protect bacteria from environmental pressures. Chloramphenicol is an antibiotic
that treats a number of bacterial infections (Izard 2001, Jardetzky 1963, Shaw, Bentley, and
Sands 1970, Gambello, Kaye, and Iglewski 1993), and here it is used as an environment
factor to inhibit cell growth. To make the design easier, a detailed mathematical model of the
synthetic symbiotic ecosystem is described. Based on user-oriented specifications, the design
with a desired mutualistic behavior is constructed by selecting adequate components from
corresponding libraries, with the help of a proposed genetic algorithm (GA)-based design
method.
The main focuses of this chapter are threefold: (a) A quorum sensing symbiotic
ecosystem is developed here to display mutualistic behaviors beneficial to population growth.
(b) The quorum sensing symbiotic ecosystem can be constructed according to user-oriented
specifications by evaluating and selecting adequate components from the corresponding
A Robust Design of Quorum Sensing Symbiotic Ecosystems … 187
libraries to achieve a desired mutualistic behavior. (c) The proposed GA-based design method
could provide synthetic biologists with a useful tool to design complicated symbiotic
ecosystems through cell-cell communication.
9.2. METHODS
Construction of Symbiotic Ecosystem
The structures of the synthetic antibiotic resistance circuits are shown in Figure 9.1,
which represents two different bacteria (ER and EG) controlled by the Rhl and Las systems,
respectively. PRhl in ER is a promoter activated by RhlR in cooperation with C4-HSL. After
the signal receptor RhlR binds with C4-HSL produced by RhlI, the promoter PRhl can be
activated and the CamR protein can be produced. CamR is an important component, allowing
bacteria to survive after inhibiting the environmental factor chloramphenicol.ER will keep
growing due to chloramphenicol resistance unless ER loses the benefits from EG. Like ER,
EG also has the same structure to allow a similar function in opposing antibiotic inhibition.
When the signal receptor LasR combines with the 3OC12-HSL produced by LasI, the
promoter PLas activates and EG produces the CamR protein to block the environmental
factor chloramphenicol for EG to sustain their growth. Coding sequences for red fluorescent
protein (RFP) and green fluorescent protein (GFP) represent the reporter proteins of ER and
EG, respectively. With this combination of quorum sensing systems, a synthetic symbiotic
ecosystem is constructed to investigate the increased complexity and dynamics of symbiotic
ecosystems in nature.
Remark 1: The coupling of the two cell populations (ER and EG) generates a synthetic
symbiotic ecosystem if the major architecture of quorum sensing systems and several
interactive components are included. The communication between ER and EG is the primary
design principal that allows them to survive in this strict environment. Without this interplay,
both ER and EG will struggle to survive in a medium enriched with the antibiotic
chloramphenicol. Here, the synthetic genetic circuits is introduced, which are beneficial to
both ER and EG, resulting in ER and EG coexisting in this symbiotic system. In the circuits,
two kinds of AHL, C4-HSL and 3OC12-HSL are used, as our small regulator factors and the
antibiotic chloramphenicol as our environmental factor. Furthermore, we design switchable
circuits that could provide several promoter-RBS compositions with different strengths and
form different levels of mutualistic behaviors in which ER and EG can benefit from each
other. Without benefiting each other, ER and EG cannot survive in an environment full of
chloramphenicol. After small signal molecules induce the promoters, downstream genes start
to produce the protein CamR, which confers resistance to chloramphenicol. Through the
design, switchable symbiosis circuits are developed, which are closer to the complex
mutualistic dynamics in nature.
Figure 9.1. Schematic of the synthetic symbiotic ecosystem in this study. Figure 9.1 shows the genetic
circuits of the synthetic symbiotic ecosystem in this study. The system can be divided into two
populations, ER and EG. These two engineered Escherichia coli populations benefit from each other
through two quorum sensing systems, resulting in ER and EG coexisting in this symbiotic system. The
regulator factors AHL 3OC12-HSL and C4-HSL pass through cell membranes, allowing ER and EG to
communicate with each other in order to survive. The changeable promoter-RBS components can
present several biological behaviors according to the kinetic strength of the promoter-RBS components.
The antibiotic resistance gene CamR is introduced to the symbiotic ecosystem, allowing the bacteria to
survive by inhibiting the antibiotic chloramphenicol.
Remark 2: In this study, the small molecules C4-HSL and 3OC12-HSL are used, both
found in the human pathogen Pseudomonas aeruginosa, as regulator factors in the symbiotic
ecosystem due to their ability to pass in and out of bacterial membranes. The purpose of
applying these autoinducers is to construct a communication mechanism between ER and EG
cells, involving two different quorum sensing promoters to express downstream functional
components.
Mathematical Models of Symbiotic Ecosystem
In this study, a promoter combined with RBS is viewed as a component. To construct the
dynamic model, promoter-RBS regulation must be introduced first. The promoter-RBS
regulation function is defined by P(PM, Pm, TF, I), where PM and Pm denote the maximum and
minimum promoter-RBS strengths, respectively. TF is transcription factor concentration, and
I denotes inducer concentration. Thus, the dynamic model of the antibiotic resistance circuit
in the ER cells, as shown in Figure 9.1, can be described with the following equations:
 
 xRhlR (t )  Pc  PM ,0,0,0    x  k  xRhlR (t )
 RhlR
x
 1CamR
 
(t )  PRhl ( PM , Rhl , Pm, Rhl , xRhlR , xC 4 )   xCamR  k  x1CamR (t )
  N t   N2 t   (9.1)
 N1  t   k  N1  t   1  1    N 1  N1 (t )   I Cam  Cam  x1CamR (t ) 
  N max 

 xC12 (t )  N1 (t )  Pc ( PM ,i ,0,0,0)  ( xC12  k )  xC12
in which
( PM , Rhl  Pm, Rhl )

PRhl ( PM , Rhl , Pm, Rhl , xRhlR , xC 4 )  Pm, Rhl  nRhl
 K Rhl 
1   * 
 x RhlR ( xRhlR , xC 4 ) 
xRhlR
x*RhlR ( xRhlR , xC 4 ) 
K 
1   C4 
 xC 4 
where PRhl(PM,Rhl,Pm,Rhl,xRhlR,xC4) denotes the promoter-RBS regulation activity of the second

stage of the activator-regulated promoter-RBS component of the Rhl system; PM,Rhl and Pm,Rhl
respectively denote the maximum and minimum kinetic strength of promoter PRhl combined
with RBS B0032; Pc(PM,0,0,0) denotes the promoter-RBS regulation activity of the first stage
of the constitutive promoter-RBS component with promoter J23105 and RBS B0034; xC4
denotes the concentration of the small molecule C4-HSL; xRhlR is the concentration of the
RhlR protein; x1CamR, which is influenced by xC4 and xRhlR, denotes the concentration of the
chloramphenicol resistance protein produced by ER; N1 denotes the cell population density
(OD600) of ER, which is controlled by protein production of x1CamR; ηCam denotes the effect
constant of the chloramphenicol resistance protein on chloramphenicol Icam; xC12 is the
concentration of the small molecule 3OC12-HSL; γxrhlR represents the protein degradation rate
of the transcriptional activator RhlR; γN1 denotes the antibiotic inhibition constant of
chloramphenicol for ER; γxCamR denotes the degradation rate of the chloramphenicol
resistance protein; KC4 denotes the dissociation rate between the inducer C4-HSL and the
transcriptional activator protein x*RhlR; KRhlR and nRhlR denote the binding affinity and the
binding cooperativity between transcriptional activator protein x*RhlR and the corresponding
promoter-RBS component in the second stage, respectively; γxC12 denotes the degradation rate
of the autoinducer 3OC12-HSL.
Similarly, the dynamic equation of antibiotic resistance circuits in EG cells as shown in
Figure 9.1 can be described as:

 xLasR (t )  Pc  PM ,0,0,0    x  k  xLasR (t )
 LasR

x
 2CamR
 
(t )  PLas ( PM , Las , Pm, Las , xLasR , xC12 )   xCamR  k  x2CamR (t )
  N t   N2 t   (9.2)
 N 2  t   k  N 2  t   1  1    N 2  N 2 (t )   I Cam  Cam  x2CamR (t ) 
  N max 

 xC 4 (t )  N 2  t   Pc ( PM , j ,0,0,0)  ( xC 4  k )  xC 4
in which
( PM , Las  Pm, Las )

PLas ( PM , Las , Pm, Las , xLasR , xC12 )  Pm, Las  nLas
 K Las 
1  * 
 x LasR ( xLasR , xC12 ) 
xLasR
x*LasR ( xLasR , xC12 ) 
K 
1   C12 
 xC12 
where PLas(PM,Las,Pm,Las,xLasR,xC12) denotes the promoter-RBS regulation activity at the second

stage of the activator-regulated promoter-RBS component in the Las system; PM,Las and Pm,Las
respectively denote the maximum and minimum kinetic strengths of the PLas promoter
combined with RBS BBa_B0034; Pc(PM,0,0,0) denotes the promoter-RBS regulation activity
at the first stage of the constitutive promoter-RBS component with promoter BBa_J23105
and RBS BBa_B0034; xC12 denotes the concentration of the small molecule 3OC12-HSL;
xLasR denotes the concentration of the LasR protein; x2CamR, which is influenced by x12 and
xLasR, denotes the concentration of chloramphenicol resistance protein produced by EG; N2
denotes the cell population density (OD600), which is controlled by protein production of
x2CamR; xC4 denotes the concentration of the small molecule C4-HSL; ηCam denotes the effect
constant of the chloramphenicol resistance protein to chloramphenicol; γxLasR denotes the
degradation rate of the transcriptional activator LasR protein; γN2 denotes the antibiotic
inhibition constant of chloramphenicol for EG; γxCamR denotes the degradation rate of the
chloramphenicol resistance protein; KC12 denotes the dissociation rate between the inducer
3OC12-HSL and the transcriptional activator protein x*LasR; KLasR and nLasR denote the binding
affinity and the binding cooperativity between transcriptional activator protein x*LasR and the
corresponding promoter-RBS component at the second stage, respectively; γxC4 denotes the
degradation rate of the autoinducer C4-HSL.
Equation (9.1) and equation (9.2) describe the dynamic models of population density of
ER and EG, respectively. As shown in Figure 9.1, ER and EG have similar circuit structures.
Thus, their dynamic models are also similar. In equation (9.1), the population density of ER
N1(t) is affected by the concentration of the chloramphenicol resistance protein x1CamR(t)
regulated by the LuxR-type transcriptional activator RhlR and the autoinducer C4-HSL. In
equation (9.2), the population density of EG N2(t) is also influenced by the concentration of
the chloramphenicol resistance protein x2CamR(t) regulated by LasR and AHL 3OC12-HSL.
Most importantly, C4-HSL and 3OC12-HSL play similar roles of cell-cell communication,
passing through the membrane to activate the corresponding promoter for the purpose of life
extension. The autoinducers C4-HSL and 3OC12-HSL are engineered as beneficial regulators
to assist the growth of ER and EG in the symbiotic ecosystem. In addition, Pc(PM,i,0,0,0) is
the regulation function of the constitutive promoter-RBS components, representing the kinetic
strength of the jth constitutive promoter-RBS component. The candidates of constitutive
promoter-RBS components are provided in Table 9.1 (see Appendix) to present several
possible mutualistic behaviors according to the kinetic strength of the components.
After constructing the dynamic models of the synthetic symbiotic ecosystem, for the
purpose of achieving the desired steady state cell population density, the dynamic models are
transformed into the steady-state models by assuming the derivatives of dynamic models in
(9.1) and (9.2) equal zero. The steady state models of the synthetic symbiotic ecosystem in
Figure 9.1 will be given as follows:
 Pc  PM ,0,0,0 
 xRhlRss 
  xRhlR  k
 P (P ,P ,x ,x )
 x1CamRss  Rhl M , Rhl m, Rhl RhlRss C 4
  xCamR  k
ER :  (9.3)
   N1 
 N1ss  N max  1  k   I Cam  Cam  x1CamRss    N 2 ss
  
 N1ss  PC ( PM ,i ,0,0,0)
 xC12 ss 
  xC12  k
and
 Pc  PM ,0,0,0 
 xLasRss 
  xLasR  k
 P (P ,P ,x ,x )
 x2CamRss  Las M , Las m , Las LasRss C12
EG:  (9.4)
   N2 
 N 2 ss  N max  1  k   I Cam  Cam  x2CamRss    N1ss
  
 N 2 ss  Pc ( PM , j ,0,0,0)
 xC 4 ss 
  xC 4  k
where xLasRss, x1CamRss, xC12ss, xLasRss, x2CamRss, and xC4ss denote the concentrations of RhlR
protein, chloramphenicol resistance protein produced by ER, small molecule 3OC12-HSL,
LasR protein, chloramphenicol resistance protein produced by EG, and small molecule C4-
HSL, at steady state, respectively; N1ss and N2ss denote cell population densities (OD600) of
ER and EG, at steady state, respectively.
In general, a synthetic symbiotic ecosystem in vivo also suffers from extrinsic
environmental noise. Thus, the equations in (9.3) and (9.4) should be modified as follows:
 Pc  PM ,0,0,0 
 xRhlRss   v1
  xRhlR  k
 P (P ,P ,x ,x )
 x1CamRss  Rhl M , Rhl m, Rhl RhlRss C 4  v2
ER:  (9.5)
   N1 
 N1ss  N max  1  k   I Cam  Cam  x1CamRss    N 2 ss  v3
  
 N1ss  PC ( PM ,i ,0,0,0)
 xC12 ss   v4
  xC12  k
and
 Pc  PM ,0,0,0 
 xLasRss   v5
  xLasR  k
 P (P ,P ,x ,x )
 x2CamRss  Las M , Las m , Las LasRss C12  v6
EG:  (9.6)
   N2 
 N 2 ss  N max  1  k   I Cam  Cam  x2CamRss    N1ss  v7
  
 N 2 ss  Pc ( PM , j ,0,0,0)
 xC 4 ss   v8
  xC 4  k
where the Gaussian random noise vp (p = 1,…,8) with a zero mean and variance of σp2
denotes cellular noise. Furthermore, biological components in a molecular biological system
are inherently uncertain (Chen and Wu 2009, Chen et al., 2011). These parameter
perturbations are defined as follows:
PM ,i  PM ,i  PM ,i n1  t  ,
PM , Rhl  PM , Rhl  PM , Rhl n2  t  , Pm, Rhl  Pm, Rhl  Pm, Rhl n2  t  ,
PM , Las  PM , Las  PM , Las n3  t  , Pm , Las  Pm , Las  Pm , Las n3  t  ,
xC 12
  xC 12   xC 12 n1  t  ,  xC 4   xC 4   xC 4 n1  t  ,
(9.7)
xRhlR
  xRhlR   xRhlR n2  t  ,  xLasR   xLasR   xLasR n2  t  ,
xCamR
  xCamR   xCamR n3  t  ,
k  k  kn4  t  , Cam  Cam  Cam n3  t  ,
 N 1   N 1   N 1n4  t  ,  N 2   N 2   N 2 n4  t 
where ΔPM,i, ΔPM,Rhl, ΔPm,Rhl, ΔPM,Las, ΔPm,Las, ΔγxC12, ΔγxC4, ΔγxRhlR, ΔγxLasR, ΔγxCamR, Δk,
ΔηCam, ΔγN1, and ΔγN2 denote the standard deviations of the stochastic parameters and np(t), p
= 1,2,3,4 are Gaussian noise with a mean of zero and unit variance exploiting the random
fluctuation sources. Hence, ΔPM,i, ΔPM,Rhl, ΔPm,Rhl, ΔPM,Las, ΔPm,Las, ΔγxC12, ΔγxC4, ΔγxRhlR,
ΔγxLasR, ΔγxCamR, Δk, ΔηCam, ΔγN1, and ΔγN2 denote the deterministic parts of parameters
variations and np(t), p = 1,2,3,4 denote different random fluctuation sources.
Remark 3: (i) The models in equations (9.5) and (9.6) provide a way to predict the
behavior of a symbiotic ecosystem based at the steady state on its functional components and
well-defined kinetic parameters. These equations can describe, predict, and stimulate
mutualistic behavior in the symbiotic ecosystem at the steady state according to the
corresponding regulatory components. Based on these equations, adequate regulatory
components can be selected to design a symbiotic ecosystem with a desired mutualistic
behavior at the steady state. (ii) If a symbiotic ecosystem could tolerate the parameter
perturbations, then it is called a robust symbiotic ecosystem. In the design, the covariance of
vp in (9.5) and (9.6) and ΔPi, Δγi, Δk and Δηi in (9.7) are prescribed beforehand so that the
designed symbiotic ecosystem could tolerate environmental noise vi and parameter
perturbations in (9.7).
Design Specifications for Robust Symbiotic Ecosystem
The design purpose of the study is to engineer a synthetic genetic circuit that controls a
symbiotic ecosystem by selecting a suitable set of promoter-RBS components from existing
libraries in combination with a proper external antibiotic chloramphenicol concentration to
achieve the optimal reference match for the desired mutualistic behaviors. The steady-state
models of the symbiotic ecosystem are represented by equations (9.5) and (9.6), which
requires describing the desired mutualistic behavior with some promoter-RBS components for
reference matching. In order to achieve the aforementioned design purpose, a library-based
search method is applied to efficiently select the most suitable set of promoter-RBS
components from the corresponding libraries (Wu, Lee, and Chen 2011). The design
procedure is provided here, with the following user-oriented specifications:
 Specify a desired reference mutualistic behavior, which will be represented by the

proportion of population densities in the synthetic symbiotic ecosystem.
 Provide the well-characterized promoter-RBS component library listed in Table 9.1
(see Appendix) to be selected.
 Specify the standard derivations of parameter uncertainties in (9.7) and the standard
deviations of environmental disturbances vp in (9.5) and (9.6) to be tolerated by the
synthetic symbiosis ecosystem in vivo, namely, the standard deviations σp of
environmental disturbances of the transcription and translation processes vp; standard
deviations ΔPM,i of the constitutive promoter-RBS strengths PM,i; standard deviations
ΔPM,Rhl, ΔPm,Rhl, ΔPM,Las, and ΔPm,Las of the maximum and minimum activator
promoter-RBS strengths PM,Rhl, Pm,Rhl, PM,Las, and Pm,Las; standard deviations ΔγxC12,
ΔγxC4, ΔγxRhlR, ΔγxLasR, and ΔγxCamR of autoinducer degradation rates γxC12 and γxC4;
regulatory protein degradation rates γxRhlR and γxLasR; chloramphenicol resistance
protein degradation rate γxCamR; and standard deviations Δk, ΔηCam, ΔγN1, and ΔγN2 of
dilution rate k due to cell growth, the effect constant of the chloramphenicol
resistance protein ηCam, and the antibiotic inhibition constants γN1 and γN2,
respectively.
 The cost function is specified as follows:

J (Ci , C j , I cam )  E 0.5  ( N1ss (Ci , I cam )  N1ref )  0.5  ( N 2 ss (C j , I cam )  N 2 ref )
2 2
 (9.8)
where (Ci,Cj,Icam) denotes the combination set of constitutive promoter-RBS components Ci

and Cj from the corresponding libraries with the external antibiotic chloramphenicol
concentration Icam. N1ss(Ci,Icam) and N2ss(Cj,Icam) denote the steady-state population densities
of ER and EG, respectively. N1ref and N2ref denote the desired reference steady-state cell
population densities of ER and EG, respectively, i.e., the desired mutualistic behavior of the
symbiotic ecosystem. The cost function combines an ER and an EG part by the weighted
mean square error method, which is the well-known weighted sum method for multi-objective
optimization (Marler and Arora 2010).
If the cost function in equation (9.8) is minimized by selecting the most adequate set of
promoter-RBS components from the corresponding libraries and the most adequate
concentration Icam under the design specifications (i)–(iv), then the population density of the
synthetic symbiotic ecosystem will be as close as possible to the specified reference steady-
state cell population density of the symbiotic ecosystem under the intrinsic parameter
fluctuations and environmental disturbances.
Specifications for Robust Symbiotic Ecosystem
Although exhaustive searching can minimize the cost function, this is a time-consuming
process. Finding a feasible combination of promoter-RBS components to
minimize J(Ci,Cj,Icam) in equation (9.8) requires long computation times, as well as trial-and-
error experimentation when component libraries are large. In this study, a genetic algorithm-
based library search method is applied to evaluate and select the most adequate promoter-
RBS component set and the concentration of chloramphenicol in an efficient and effective
way while searching large libraries. Therefore, the synthetic symbiotic ecosystem can
robustly and optimally achieve the desired population density of the symbiotic ecosystem by
using a library-based search method combined with GA to efficiently choose the optimal set
of promoter-RBS components and a concentration of chloramphenicol to satisfy the four
design specifications. The design procedure is summarized as follows:
1. Provide user-defined design specifications N1ss(Ci,Icam) and N2ss(Cj,Icam) for the

symbiotic ecosystem.
2. Select an initial set (Ci,Cj) of promoter-RBS components and an initial concentration
Icam of chloramphenicol.
3. Calculate the cost value J(Ci,Cj,Icam) in (9.8) for (Ci,Cj,Icam).
4. Create an offspring set (Ci,Cj,Icam) by using GA operators such as reproduction,
crossover, and mutation
5. Calculate the cost value of the new set (Ci,Cj,Icam) obtained by natural selection. Stop
when the design goal is achieved or an acceptable solution is obtained. Otherwise,
create the next generation and return to step 3.
9.3. RESULTS
Based on the structure of the synthetic symbiotic ecosystem in Figure 9.1, the set of
constitutive promoter-RBS components Ci, Cj, and the concentration Icam of chloramphenicol
will be selected for the antibiotic resistance circuits from the corresponding promoter-RBS
component libraries in Table 9.1. In the symbiotic ecosystem, the chloramphenicol resistance
protein is embedded downstream of the activator-regulated promoter-RBS components of
both ER and EG. The genetic circuits are divided into two stages. The first stage is the
constitutive promoter-RBS components for producing the small molecules 3OC12-HSL and
C4-HSL, which pass through the membrane to enable cell-cell communication between ER
and EG. The second stage is the activator-regulated promoter-RBS components for driving
the expression of the chloramphenicol protein to resist the antibiotic effects. The antibiotic
chloramphenicol is used to regulate the mutualistic behaviors.
For achieving the desired mutualistic behaviors represented by the proportions of cell
population densities, the design specifications are provided as follows:
 The desired steady-state proportion of cell population densities are prescribed as 1:1,
1:3, and 2:1, respectively.
 The index of constitutive promoter-RBS component i and j is from 1 to 21. The
feasible range of antibiotic chloramphenicol concentration is from 0 M to 0.1 M.
 The environmental noises v1, v2, v4, v5, v6, and v8 for transcription and translation
processes, and the noise v3 and v7 for cell population density are all zero mean
Gaussian noise with unit variance and the 5% parameter fluctuations are prescribed
to tolerate, i.e.,
PM ,i  0.05PM ,i ,
PM , Rhl  0.05PM , Rhl , Pm , Rhl  0.05Pm , Rhl ,
PM , Las  0.05PM , Las , Pm, Las  0.05Pm, Las ,
 xC 12  0.05 xC 12 ,  xC 4  0.05 xC 4 ,
(9.9)
 xRhlR  0.05 xRhlR ,  xLasR  0.05 xLasR ,
 xCamR  0.05 xCamR ,
k  0.05k , Cam  0.05Cam ,
 N 1  0.05 N 1 ,  N 2  0.05 N 2
 The cost function of the synthetic symbiotic ecosystem in (9.8) is to be minimized by

selecting an adequate promoter-RBS component set (Ci,Cj) from the corresponding
libraries in Table 9.1 and a concentration of chloramphenicol Icam within [0, 0.1] M
to achieve the desired steady-state cell population density.
According to the above design procedures, the GA search method is applied to search the
adequate set of promoter-RBS components Ci, Cj, and suitable antibiotic chloramphenicol
concentration Icam to minimize J(Ci,Cj,Icam) in equation (9.8). The simulation results with the
experimental data are compared to verify the robust design results of the proposed systematic
approach for a symbiotic ecosystem as follows.
In example 1, the simulation and experimental population density ratios are 0.9993 and
0.9663, respectively. The percentage of error is 3.42% showing that the prescribed cell
population density ratio is well represented by the experimental data. If one wants to achieve
the desired mutualistic population density ratio of 1, it can be chosen the set of constitutive
promoter-RBS components Ci as J23118-B0034 and Cj as J23118-B0034 from the library
with chloramphenicol concentration of 0.0025 M. The comparison between simulation results
and experimental data is shown in Figure 9.2.
Figure 9.2. The comparison of simulation results and experimental data to verify an adequate set of
constitutive promoter-RBS components and chloramphenicol concentration. The simulation results and
experimental data in proportion to cell population densities are 1:1, where Ci is J23118-B0034 and Cj is
J23118-B0034 with 0.0025 M chloramphenicol concentration.
0.3258, respectively. The percentage error of 6.35% shows that the prescribed cell population
density ratio is also represented quite well by the experimental data. If the desired mutualistic
population density ratio is described as 0.33, it can then be chosen the set of constitutive
promoter-RBS components Ci as J23101-B0034 and Cj as J23109-B0032 from the library
with the chloramphenicol concentration as 0.01 M. The comparison between simulation
results and experimental data is shown in Figure 9.3.
1.6465, respectively. The percentage error of 22.11% shows that the prescribed cell
population density ratio is not well represented by the experimental data. The reason is
probably because EG shows greater competence than ER when cultured in the same medium.
For equal antibiotic concentration environments, ER tends to require more maintenance than
EG to survive, which partially indicates that ER is less tolerant to antibiotic chloramphenicol.
Although the result is not an absolute match to the desired population density ratio, it
provides a closer option in our design. Therefore, if one want to achieve a mutualistic
population density ratio of 2, it may be chosen the set of constitutive promoter-RBS
components Ci as J23118-B0034 and Cj as J23101-B0034 from the library with the
chloramphenicol concentration as 0.005 M to arrive closer to the desired mutualistic
population density. The comparison of the simulation results with experimental data is shown
in Figure 9.4.
Based on the above design results and discussion, the mutualistic behaviors can be
controlled by changing different sets of components along with the adequate antibiotic
chloramphenicol concentration. Although EG shows a greater competence than ER in the
symbiotic relationship, this competence discrimination is correlated to the strength of the
antibiotics and may sometimes make the ecosystem more stable. When the number of
component libraries increases, the design of synthetic symbiotic ecosystem makes it easier to
achieve the desired mutualistic population densities by selecting a more adequate set of
components and antibiotic chloramphenicol concentration.
9.4. DISCUSSION
In this chapter, a controllable quorum sensing symbiotic ecosystem is designed to mimic
complex mutualistic relationships in natural ecosystems. The design principles mainly focus
on cell-cell communication to achieve mutualistic behavior by two quorum sensing systems.
Changeable promoter-RBS components are also provided to regulate the production of
autoinducers, which play communicator roles between different cells. To more deeply discuss
the relationship between microorganisms, a synthetic microbial ecosystem is designed,
constructed, and simulated to discuss mutualistic behaviors through cell-cell communication
between E. coli populations. The components have different kinetic strengths, resulting in the
production of small molecule regulation factors that can activate the function of
corresponding quorum sensing promoters. With dynamic models of gene expression and well-
defined kinetic parameters, it can be chosen a variety of promoter-RBS component sets and
estimate a priori their kinetic parameters from the corresponding responses by mathematical
analysis. The experimental results have shown that mutualism indeed occurs between ER and
EG to sustain two E. coli populations. As anticipated, the promoter-RBS components can
provide different and exact levels of mutualistic behaviors. As a result, using the functional
components and well-defined kinetic parameters, one can apply a systematic way of
predicting and verifying the behavior of organisms and system responses in a standardized
design procedure.
By combining promoter-RBS components with quorum sensing systems, E. coli
population density can be controlled more accurately to present the desired symbiotic
behavior as per natural ecosystems. Cell-cell communication is created between two

populations to mimic the complex mutualistic relationships in natural microbial ecosystems
via two different quorum sensing systems, the Las system and the Rhl system, which include
LasI/LasR and RhlI/RhlR, respectively (de Kievit et al., 2002, Miller and Bassler 2001, Pesci
et al., 1997). These two quorum sensing systems were first found in the human pathogen
Pseudomonas aeruginosa in relation to the production of several poisonous factors (Pesci et
al., 1997). These systems communicate with each other by the small signal molecule N-Acyl
homoserine lactone (AHL), which is produced by the LuxI-type autoinducer synthase. After
the N-Acyl homoserine lactone (AHL) concentration reaches a threshold, these autoinducers
activate a LuxR-type transcriptional activator to induce specially appointed genes.
Furthermore, quorum sensing systems also auto-regulate LuxI-type autoinducer synthase,
producing a generation of more N-Acyl homeserine lactone (AHL). Thus the mechanism
which regulates gene expression in a cell density-dependent manner is called quorum sensing
(Pearson, Pesci, and Iglewski 1997). Additionally, quorum sensing synchronizes behaviors of
the bacteria, creating behaviors that vary with population density and species composition.
This synchronicity results in noise reduction, allowing us to construct more robust circuits
and systems (Koseska et al., 2009).
In this study, an antibiotic is used as an environmental pressure to restrain cell growth.
The higher the antibiotic concentration, the lower the cell population density. Antibiotics are
widely used for treating and preventing bacterial infections by either killing bacteria or
preventing them from reproducing. The antibiotic chloramphenicol is chosen here as our
environmental factor to control the cell density of bacteria. Chloramphenicol is originally
derived from the bacterium Streptomyces venezuelae (AKAGAWA, OKANISHI, and
UMEZAWA 1975, Izard 2001). It is regarded as a prototypical broad-spectrum antibiotic,
which can be used effectively against a wide range of gram-positive bacteria and gram-
negative bacteria. Mechanistically, chloramphenicol inhibits bacterial protein synthesis
without affecting many other microbial metabolic processes (Jardetzky 1963). It prevents the
elongation of protein chains by blocking peptidyl transferase activity of bacterial ribosomes,
which may prevent the formation of the peptide bone. In addition to chloramphenicol,
antibiotic resistance, i.e., chloramphenicol resistance (CamR), is also used as a survival factor
in the study to maintain the bacteria populations. Chloramphenicol resistance is a commonly
found genetic component. The chloramphenicol acetyltransferase (CAT) enzymes can
inactivate the antibiotic chloramphenicol. Mechanistically, CATs provide chloramphenicol
resistance to S. epidermidis by preventing chloramphenicol from binding to and interacting
with ribosomes (Shaw, Bentley, and Sands 1970).
This study shares some similarities with previous research elucidating the interspecies
interaction between complex natural ecosystems and population dynamics, such as the
predator–prey ecosystem by Balagaddé et al., and the environment sensitive synthetic
microbial ecosystem by Hu’s group. However, the purpose of our design is to systematically
engineer a controllable symbiotic ecosystem in a strict environment by a mathematical
modularized approach under intrinsic parameter uncertainties and external disturbances. The
robust ability to resist and tolerate the above uncertainties or disturbances is also an important
design aspect in complex microbial ecosystems (Banning et al., 2008, Berga, Székely, and
Langenheder 2012, Chen and Wu 2009, Williams and Lenton 2008). Our study provides a
library-based GA search method to select the most suitable kinetic parameters for genetic
circuits in our symbiotic ecosystem to engineer the desired mutualistic behavior. Based on a
bottom-up design approach, this can be extended to a more complicated process of

programmed mutualistic behavior in the future to precisely explain the mutualistic
phenomenon that always exists in natural microbial ecosystems through cell-cell
communication. In conclusion, the well-defined experimental results and detailed analyses of
mathematical models for this controllable quorum sensing symbiotic ecosystem provide a
design method and information about mutualistic functions of natural microbial ecosystems
and complicated interactions between genetic components and system behavior through cell-
cell communication in the symbiotic ecosystem.
9.5. CONCLUSION
In this chapter, a controllable symbiotic ecosystem is constructed, which uses cell-cell
communication to generate different levels of mutualistic behaviors. The two synthetic
antibiotic resistance circuits proposed (ER and EG) combined two quorum sensing systems,
promoter-RBS regulated components, and antibiotic resistance genes. These circuits provide
valuable information when discussing complex compositions and structures of symbiotic
microbial ecosystems. With the well-characterized kinetic components in Biobrick and by
engineering dynamic equations of protein expression, one can predict a priori the system
behavior of our symbiotic ecosystem. The results also showed that the identification of kinetic
parameters for the promoter-RBS component library was a reliable method and that the GA
searching method was a dependable tool to select the most adequate promoter-RBS
components for the robust design of genetic circuits to achieve desired mutualistic behaviors,
which could be specified beforehand. Three adequate sets of promoter-RBS components are
selected from the corresponding library for practical verification of three different mutualistic
behaviors specified beforehand. The experimental results have demonstrated that the
proposed deign method provided a systematic way of predicting the behavior of organisms
and system response, from which a synthetic symbiotic ecosystem with a more desirable
mutualistic behavior could be designed accordingly. Therefore, the proposed controllable
symbiotic ecosystem may act as a framework for constructing more complicated microbial
ecosystems in the fast growing field of synthetic biology.
9.6. APPENDIX
Table 9.1. Constitutive promoter-RBS component library
Index Component PM
C1 J23101-B0031 22.3983
C2 J23101-B0032 15.8371
C3 J23101-B0034 58.745
C4 J23105-B0031 2.1678
C5 J23105-B0032 1.6382
C6 J23105-B0034 6.2221
C7 J23106-B0031 4.9324
Index Component PM
C8 J23106-B0032 3.0406
C9 J23106-B0034 13.5193
C10 J23109-B0031 0.3619
C11 J23109-B0032 0.3058
C12 J23109-B0034 0.3952
C13 J23110-B0031 9.9948
C14 J23110-B0032 6.4898
C15 J23110-B0034 24.1504
C16 J23114-B0031 0.7089
C17 J23114-B0032 0.6195
C18 J23114-B0034 1.3814
C19 J23118-B0031 16.7616
C20 J23118-B0032 10.5129
C21 J23118-B0034 44.8676
REFERENCES
AKAGAWA, H., M. OKANISHI, and H. UMEZAWA. 1975. “A Plasmid Involved in
Chloramphenicol Production in Streptomyces venezuelae: Evidence from Genetic
Mapping.” Journal of General Microbiology 90 (2):336-346. doi: 10.1099/00221287-90-
2-336.
Balagaddé, Frederick K, Hao Song, Jun Ozaki, Cynthia H Collins, Matthew Barnet, Frances
H Arnold, Stephen R Quake, and Lingchong You. 2008. A synthetic Escherichia coli
predator–prey ecosystem. Vol. 4. Journal Article.
Banning, Natasha, Daniel Murphy, Natasha Banning, and Daniel Murphy. 2008. “Effect of
heat-induced disturbance on microbial biomass and activity in forest soil and the
relationship between disturbance effects and microbial community structure.” Applied
Soil Ecology 40 (1):109-119. doi: 10.1016/j.apsoil.2008.03.011.
(7037):1130-1134. doi:
http://www.nature.com/nature/journal/v434/n7037/suppinfo/nature03461_S1.html.
Berga, Mercè, Anna J. Székely, and Silke Langenheder. 2012. “Effects of Disturbance
Intensity and Frequency on Bacterial Community Composition and Function.” PLoS
ONE 7 (5):e36959. doi: 10.1371/journal.pone.0036959.
Chen, Bor-Sen, Chia-Hung Chang, Yu-Chao Wang, Chih-Hung Wu, and Hsiao-Ching Lee.
2011. “Robust model matching design methodology for a stochastic synthetic gene
network.” Mathematical Biosciences 230 (1):23-36. doi: http://dx.doi.org/
10.1016/j.mbs.2010.12.007.
Chen, Bor-Sen, and Chih-Hung Wu. 2009. “A systematic design method for robust synthetic
biology to satisfy design specifications.” BMC Systems Biology 3 (1):66.
de Kievit, Teresa R., Yoshio Kakai, J. Kristen Register, Everett C. Pesci, and Barbara H.
Iglewski. 2002. “Role of the Pseudomonas aeruginosa las and rhl quorum-sensing
systems in rhlI regulation.” FEMS Microbiology Letters 212 (1):101-106. doi:
http://dx.doi.org/10.1016/S0378-1097(02)00735-8.
Gambello, M. J., S. Kaye, and B. H. Iglewski. 1993. “LasR of Pseudomonas aeruginosa is a
transcriptional activator of the alkaline protease gene (apr) and an enhancer of exotoxin A
expression.” Infection and Immunity 61 (4):1180-1184.
Hibbing, M. E., C. Fuqua, M. R. Parsek, and S. B. Peterson. 2010a. “Bacterial competition:
surviving and thriving in the microbial jungle.” Nat Rev Microbiol 8 (1):15-25. doi:
10.1038/nrmicro2259.
Hibbing, Michael E., Clay Fuqua, Matthew R. Parsek, and S. Brook Peterson. 2010b.
“Bacterial competition: surviving and thriving in the microbial jungle.” Nature reviews.
Microbiology 8 (1):15-25. doi: 10.1038/nrmicro2259.
Hu, B., J. Du, R. Y. Zou, and Y. J. Yuan. 2010. “An environment-sensitive synthetic
microbial ecosystem.” PLoS One 5 (5):e10619. doi: 10.1371/journal.pone.0010619.
Izard, Tina. 2001. “Structural basis for chloramphenicol tolerance in Streptomyces
venezuelae by chloramphenicol phosphotransferase activity.” Protein Science : A
Publication of the Protein Society 10 (8):1508-1513.
Jardetzky, Oleg. 1963. “Studies on the Mechanism of Action of Chloramphenicol: I. THE
CONFORMATION OF CHLORAMPHENICOL IN SOLUTION.” Journal of Biological
Chemistry 238 (7):2498-2508.
Klitgord, Niels, and Daniel Segrè. 2011. “Ecosystems biology of microbial metabolism.”
Current Opinion in Biotechnology 22 (4):541-546. doi: http://dx.doi.org/10.1016/
j.copbio.2011.04.018.
Koseska, A., A. Zaikin, J. Kurths, and J. Garcia-Ojalvo. 2009. “Timing cellular decision
making under noise via cell-cell communication.” PLoS One 4 (3):e4872. doi:
10.1371/journal.pone.0004872.
Little, Ainslie E.F., Courtney J. Robinson, S. Brook Peterson, Kenneth F. Raffa, and Jo
Handelsman. 2008. “Rules of Engagement: Interspecies Interactions that Regulate
Microbial Communities.” Annual Review of Microbiology 62 (1):375-401. doi:
doi:10.1146/annurev.micro.030608.101423.
Mao, Junwen, Andrew E. Blanchard, and Ting Lu. 2014. “Slow and Steady Wins the Race: A
Bacterial Exploitative Competition Strategy in Fluctuating Environments.” ACS Synthetic
Biology. doi: 10.1021/sb4002008.
Marler, R. Timothy, and JasbirS Arora. 2010. “The weighted sum method for multi-objective
optimization: new insights.” Structural and Multidisciplinary Optimization 41 (6):853-
862. doi: 10.1007/s00158-009-0460-7.
Miller, Melissa B., and Bonnie L. Bassler. 2001. “QUORUM SENSING IN BACTERIA.”
Annual Review of Microbiology 55 (1):165-199. doi:
doi:10.1146/annurev.micro.55.1.165.
Pace, N. R. 1997. “A molecular view of microbial diversity and the biosphere.” Science 276
(5313):734-40.
Pearson, J. P., E. C. Pesci, and B. H. Iglewski. 1997. “Roles of Pseudomonas aeruginosa las
and rhl quorum-sensing systems in control of elastase and rhamnolipid biosynthesis
genes.” Journal of Bacteriology 179 (18):5756-5767.
Pesci, E. C., J. P. Pearson, P. C. Seed, and B. H. Iglewski. 1997. “Regulation of las and rhl
quorum sensing in Pseudomonas aeruginosa.” Journal of Bacteriology 179 (10):3127-
3132.
Shaw, W. V., D. W. Bentley, and L. Sands. 1970. “Mechanism of Chloramphenicol
Resistance in Staphylococcus epidermidis.” Journal of Bacteriology 104 (3):1095-1105.
Toder, D. S., S. J. Ferrell, J. L. Nezezon, L. Rust, and B. H. Iglewski. 1994. “lasA and lasB
genes of Pseudomonas aeruginosa: analysis of transcription and gene product activity.”
Infection and Immunity 62 (4):1320-1327.
Toder, D. S., M. J. Gambello, and B. H. Iglewski. 1991. “Pseudomonas aeruginosa LasA: a
second elastase under the transcriptional control of lasR.” Molecular Microbiology 5
(8):2003-2010. doi: 10.1111/j.1365-2958.1991.tb00822.x.
Waters, Christopher M., and Bonnie L. Bassler. 2005. “QUORUM SENSING: Cell-to-Cell
Communication in Bacteria.” Annual Review of Cell and Developmental Biology 21
(1):319-346. doi: doi:10.1146/annurev.cellbio.21.012704.131001.
Williams, Hywel T. P., and Timothy M. Lenton. 2008. “Environmental regulation in a
network of simulated microbial ecosystems.” Proceedings of the National Academy of
Sciences of the United States of America 105 (30):10432-10437. doi:
10.1073/pnas.0800244105.
You, Lingchong, Robert Sidney Cox, Ron Weiss, and Frances H. Arnold. 2004.
“Programmed population control by cell-cell communication and regulated killing.”
Nature 428 (6985):868-871.
Chapter 10
SYSTEMATIC DESIGN OF A MULTICELLULAR

MOLECULAR COMMUNICATION SYSTEM WITH
DESIRED DETECTION CAPABILITY, TRANSDUCTION
ABILITY, AND SYSTEM SENSITIVITY
ABSTRACT
The limited affordability of exogenous components in single-celled organisms has
created a design problem with regards to coordination between single-celled organisms.
This is an ongoing challenge in synthetic biology because synthetic biologists require
communication systems which are capable of performing multi-functional behaviors. In
this study, we created a design scheme for a multicellular molecular communication
system (MMCS) which consisted of sender and receiver cells. In addition, we
demonstrated that our design scheme enabled the MMCS to meet system performance
requirements. First, we created three well-characterized promoter–ribosome binding site
libraries for genetic circuits. These were then implemented in sender and receiver cells.
The MMCS included three important system functions: detection capability, transduction
ability, and system sensitivity. These system functions allowed the MMCS to select the
appropriate components for the sender and receiver cells from the libraries. They also
enabled the MMCS to attain prescribed levels under conditions which included intrinsic
random fluctuations and channel noise. Finally, we report several design examples with
in silico and in vitro experimental results to illustrate and to justify the MMCS design
scheme.
10.1. INTRODUCTION
Synthetic biology has received a lot of attention in recent years. This is because synthetic
biology enables researchers to both modify an existing biological system and also to
understand whole biological systems at a systematic level. From an engineering perspective,
the purpose of synthetic biology is to develop various specific cellular functions within a
biological system using systematic design methods (Endy 2005). A systematic design method
uses various biological components to achieve a specified cellular function. Using such
design methods, many synthetic biologists have been able to design and analyze a variety of
biological components and assembled these components into a genetic circuit to achieve a
specified function. For example, toggle switches (Gardner, Cantor, and Collins 2000),
oscillators (Chang, Lin, and Jennawasin 2013, Elowitz and Leibler 2000, Stricker et al., 2008),
logic evaluators (Rinaudo et al., 2007), filters (Sohka, Heins, and Ostermeier 2009, Sohka et
al., 2009), and cell–cell communicators (Mikami et al., 2015, Pai et al., 2009).
The next-generation gene network (Cameron, Bashor, and Collins 2014, Khalil and
Collins 2010, Lu, Khalil, and Collins 2009) is an emerging area of research in synthetic
biology. It is used to design more complex biological systems, such as modular genetic logic
(Wang et al., 2011), multicellular logic (Silva-Rocha and de Lorenzo 2014), and
computational circuits (Chang, Lin, and Jennawasin 2013, Regot et al., 2011, Tamsir, Tabor,
and Voigt 2011). However, to construct and design more complex biological systems,
synthetic biologists encounter several problems. The most important issue is that it is still
difficult to satisfy all of the designer's behavioral requirements using conventional methods.
The second important issue is that it is not easy to institute all of the specific functions
desired within a single-celled organism. It is difficult to create a complex genetic circuit with
a few different components, due to the limitation of the number of exogenous components in
a single-celled organism. As a result, it is more appealing for synthetic biologists to
coordinate several simple genetic circuits in different single-celled individuals to perform like
a complex genetic circuit via a cell-to-cell communication system.
Recently, there have been several studies of a multicellular systems which have used the
genetic elements of quorum sensing to engineer several cell-to-cell communications (Bassler
1999, Basu et al., 2005). Cell-to-cell communication uses positive and negative regulation of
gene expression to obtain a pulse-generator within a multicellular synthetic system (Basu et
al., 2004). The receiver cells are engineered with the control element. The quorum-sensing
system can control the density of cell populations through luxI and luxR genes by tuning the
autoinducer signal (You et al., 2004). Furthermore, there are additional synthetic circuits
which can engineer cell-to-cell communication (McDaniel and Weiss 2005, Weiss and Knight
2001). However, not all of these systems are compatible with some of the design
specifications desired from a systematic method.
Here we proposed a robust synthetic gene circuit design method for the multicellular
molecular communication system (MMCS) which operates using a biosensor to promoter–
RBS libraries communication system. Our system is superior to the method of using different
single-celled organisms in a community because it can be implemented with different
functions and can be coordinated to perform a systematic behaviors through cell-to-cell
communication where a single-celled organism cannot (Cameron, Bashor, and Collins 2014,
Khalil and Collins 2010, Lu, Khalil, and Collins 2009).
In this chapter, we used a biosensor as the sensing device to detect the input signal for the
MMCS. We focused on metal ion detection using a genetic copper biosensor circuit, taken
from the Escherichia coli chromosome Cus system (Bondarczuk and Piotrowska-Seget 2013).
The Cus system found in E. coli consists of four proteins (CusC, CusB, CusA, and CusF),
which mediate resistance to copper and silver by cation efflux (Franke et al., 2003). CusC is
an outer membrane protein that undergoes specific acquisition of tolerance under high Cu2+
and Ag+ concentrations (Macomber and Imlay 2009). We chose the promoter PcusC and its
associated regulatory system CusSR, as our Cu2+-induced system. In this procedure, the
CusSR regulatory system was used to detect exogenous Cu2+ via membrane-associated sensor
Systematic Design of A Multicellular Molecular Communication System … 207
kinase. In addition, PcusC,, was also used as the binding site for CusR because it is a CusR-
dependent promoter (Ravikumar et al., 2011, Munson et al., 2000).
For our MMCS we used a quorum-sensing mechanism as the channel in which the signal
was delivered and received. Quorum sensing refers to the regulation of gene expression with
different functions in response to cell-to-cell communication. The quorum-sensing process
allows bacteria to produce and release small signal molecules such as autoinducers.
The most widely studied quorum-sensing system is the autoinduction of luminescence in
the marine bacteria Vibrio fischeri (Mcfallngai and Ruby 1991). Several studies have
analyzed this particular quorum-sensing system. Some investigations have shown that high
cell density can be achieved by using higher concentrations of the diffusible molecules
(Dunlap and Kuo 1992, Ruby and Nealson 1976). For example, through a series of
experiments, Darch et al., showed that quorum-sensing systems can benefit cell density
(Darch et al., 2012).
Acylated homoserine lactone (AHL), also known as N-3-homoserine lactone (HSL) can
be classified according to acyl group length (C4-C18) (Hanzelka and Greenberg 1996, Val
and Cronan 1998). Here, the type of HSL we employed was C6-HSL. The quorum-sensing
systems we used in the present study were LuxI and LuxR. The LuxI protein catalyzes the
production of AHL and the LuxR protein binds to AHL to form an AHL–LuxR complex. The
AHL–LuxR complex is able to bind to the target promoter Plux and activate downstream
transcription. There are other species of quorum-sensing systems that have similar functions
to the LuxI and LuxR proteins (Miller and Bassler 2001). These include the human pathogen
Pseudomonas aeruginosa, which has two pairs of quorum-sensing systems: LasI/LasR and
RhlI/RhlR. LasI and RhlI are used to synthesize two different types of autoinducers, whereas
LasR and RhlR act as the cognate transcriptional activators (Brint and Ohman 1995, Passador
et al., 1993, Pearson et al., 1995).
In summary, by utilizing available libraries, well-defined biological parts can be
assembled into a desired genetic circuit to carry out a specific function. Here, we used a
promoter–RBS library searching method to select adequate promoter–RBS components.
These were then used to engineer a multicellular genetic circuit to achieve a robust cellular
function coordinated by cell-to-cell communication. The promoter–RBS libraries can be used
to solve global optimization problems by selecting adequate promoter–RBS components to
achieve the design specifications of robust multicellular synthetic gene networks via a genetic
algorithm (GA) (Chen and Goldberg 2005). Thus, the most suitable promoter–RBS can be
easily and efficiently selected from the corresponding libraries to achieve the desired system
performance of the MMCS.
10.2. METHODS
Design Method of the Multicellular Molecular Communication System
In order to build a molecular communication system of high complexity, we used the

model of the multicellular molecular communication system (MMCS) depicted in the block
diagram in Figure 10.1. This model includes three major modules: the sense process, the
diffusion process, and the response process.
Figure 10.1. Block diagram of the multicellular molecular communication system (MMCS). Our
MMCS consisted of three major modules and each module contained a specific genetic circuit to fulfill
its own purposes. (1) The module of the sense process consists of sender cells with their biosensors, (2)
the module of the diffusion process consists of molecules passing through channels, and (3) the module
of the response process consists of receiver cells with their applications.
Sense Process
Before constructing a well-designed communication system, we needed the available

input signal to be able to solve a genetic engineering problem. Moreover, an input signal
needed to be transferred from the outside into a molecular system via some kind of sense
device. In this study, we chose a Cu2+-induced biosensor native to E. coli to sense Cu2+. Using
this input, the designer can design a genetic circuit for the sender cell to respond with the
appropriate action. The detection capability D of the sense process in Figure 10.1 is defined
as follows:
Transmitted signal intensity

D
Input signal intensity (10.1)
We then needed to select an adequate component to enable the sender cell to have the
desired detection capability from the library.
Diffusion Process
MMCS is a design strategy in which a sender emits signals that are diffused to an
intended receiver. Here we defined the signal as the diffusible small signal molecule, AHL,
which is produced by sender cell. Small signal molecules have an advantage in that they
diffuse quickly after being released by a transmitter. After the sender cell transmitted the
small signal molecule, we also needed a channel to transfer small signal molecules from
sender to receiver. Thus, we defined the channel as the quorum-sensing mechanism, which is
native in bacteria. The quorum-sensing process allows bacteria to detect signals from other
bacteria and to change behavior on a large scale in response to the changes in the environment
(Waters and Bassler 2005). The quorum-sensing systems we used here were LuxI and LuxR
proteins.
In order to ensure the receiver accurately attains the sender's message, signals should not
be wrongly identified in the medium during the transmitting process. Fortunately, the
quorum-sensing mechanism has specificity. The sender cell has the LuxI protein to catalyze
the substrates S-adenosylmethionine (SAM), and acyl-acyl carrier protein (acyl-ACP), into an
autoinducer - AHL. Furthermore, the receiver contains the LuxR protein which binds to AHL
to form the AHL–LuxR complex. This complex is then able to bind to the target promoter
sequence and activate the initial response of the receiver. Using this molecular binding
process, the signal can be effectively transmitted to the receiver cell.
Response Process
As a result of the diffusion process, the small signal molecules could successfully pass
through the channel to the next stage of the communication system. This is where the
intended targets were preparing to receive the signal and produce the appropriate actions.
Here, we defined the reception place of the signal as the receiver cell. The designer can create
a genetic circuit for the receiver cell to respond with the appropriate actions. In order to
achieve such transduction capabilities in the MMCS, we defined the transduction ability (T in
Figure 10.1) as follows:
Output signal intensity

T
Input signal intensity (10.2)
In addition, in the MMCSs, there are many intrinsic random fluctuations as well as
channel noise. In order to design a robust MMCS despite such intrinsic random fluctuations
and channel noise, we defined the sensitivity S of MMCS (represented in Figure 10.1) as the
following change ratio:
 Output signal intensity

S
 Input signal intensity (10.3)
In summary, we set up several promoter–RBS component libraries for the systematic

genetic circuit design of multicellular molecular communication. We identified three specific
functions required to design a multicellular molecular communication system. These were: an
effective detection ability in the sender cells, an effective transduction ability from input to
output, and an effective system robustness against intrinsic random fluctuations and channel
noise. For the MMCS to achieve the desired detection capability in (10.1), desired
transduction ability in (10.2), and a desired sensitivity in (10.3) we needed to select the
adequate promoter–RBS components from the corresponding libraries.
Construction of Synthetic Gene Network through Multicellular

Molecular Communication System
First, we characterized and standardized the promoter–RBS libraries of the genetic

circuits to be introduced into the MMCS. Next, we employed a dynamic mathematical model
to construct the components of the promoter–RBS libraries by identifying their kinetic
strengths using experimental data. Finally, we used the library searching method to choose
the most appropriate promoter–RBS components from the corresponding libraries to construct
a predictable multicellular synthetic gene network for cell-to-cell molecular communication.
Construction of Promoter–RBS Libraries
Designers usually choose components to construct genetic circuits from the registry of
biological standard parts created by MIT (Kelly et al., 2009). Such fundamental synthetic
genetic circuits usually consists of four kinds of standard parts. These are a promoter, a
ribosome binding site (RBS), a reporter gene and a terminator. When the genetic circuit is
operating, the promoter is the element which initiates the transcription of DNA. The reporter
gene can be a functional coding sequence for the characterization and generation of the
desired specific behavior. When translation begins, the RBS affects the translation rate at
which ribosomes are recruited to the mRNA. A design schematic depicting the basic function
of a genetic circuit is shown in Figure 10.2, with an example of green fluorescent protein
(GFP) expression. Reporter genes can produce reporter proteins in response to the expression
of a specific inducer. They can also be replaced by a specific gene to achieve a prescribed
behavior. Here, the GFP coding sequence (BBa_E0040) is used as the reporter protein
downstream of the promoter and RBS. Furthermore, by inserting a powerful terminator
(BBa_B0015) at the end, one can stop the transcription of DNA with almost 100% efficiency.
Sender and receiver cells are essential for the performance of the MMCS in cell-to-cell
communication. Because of this, we first constructed the sender cell with the copper
biosensor circuit. Figure 10.3 shows a schematic of a sender cell with tunable components. To
measure the promoter–RBS components, we used a GFP sequence. We also used BBa_E0040
to represent the luxI sequence and BBa_C0061 in the sender cell. An RBS was inserted
between the copper-induced promoter and the GFP sequence. Different RBS components
have different RNA translation efficiencies. Using the described method, a designer can
specify their desired level of translation efficiency. If the designer chooses the strongest RBS
(BBa_B0034), the GFP will be the strongest. On the other hand, if the designer chooses the
weakest RBS (BBa_B0034), the GFP will be the weakest.
Second, we constructed the genetic circuit for the receiver cell. Figure 10.4 shows a
schematic of a receiver cell with two kinds of tunable promoter–RBS components. These are
the constitutive promoter–RBS component and the activator regulated promoter–RBS
component. The genetic circuit consisted of lux operon derived from a V. fischeri quorum-
sensing system. The component luxR sequence (BBa_C0062) was inserted between the RBS
and a terminator to produce the LuxR protein. The promoter Plux (BBa_R0062) was inserted
after the first stage of the constitutive promoter–RBS component sequence.
Figure 10.2. Design scheme of GFP protein expression in the fundamental synthetic genetic circuit.
This figure shows the scheme of the fundamental synthetic genetic circuit, which consists of a promoter
(yellow arrow) with an RBS (blue ellipse), a green fluorescent protein coding sequence, BBa_E0040
(green arrow) and a terminator, BBa_B0015 (red octagon) from the BioBrick.
Figure 10.3. Design scheme of a synthetic genetic circuit in a sender cell with a tunable RBS
component. This figure shows a schematic of a sender cell with a Cu 2+-induced promoter–RBS
component, and a promoter PcusC, which is native in E. coli.
In order to facilitate the design of the synthetic gene circuit, well-characterized library
components needed to be constructed based on their kinetic strengths. Because gene
expression is both regulated by the promoter and RBS, we considered the promoter and RBS
as a promoter–RBS component. The kinetic strength of each promoter–RBS component
needed to be systematically identified and constructed within the promoter–RBS libraries for
them to able to be efficiently used in the synthetic gene circuit. As a result, we included three
kinds of RBS with promoter PcusC in our library, i.e., B  B1 , B2 , B3 as shown in Table
10.1. Further, we included 24 kinds of constitutive promoter–RBS components in our library,
i.e., C  C1 , , C24  , and 3 kinds of activator regulated promoter–RBS components, i.e.,
R  {R1 , R2 , R3} , see Table 10.2 and Table 10.3 respectively.
Kinetic strengths of promoter–RBS components were estimated by the least square
parameter estimation method based on a dynamic model described in the next section using
experimental data. This ensured that the designer can choose different promoter–RBS
components from the corresponding libraries. In addition, using the library-based searching
scheme, the synthetic molecular communication system can achieve the desired
communication system performance in detection capability in (10.1), transduction ability in
(10.2) and sensitivity in (10.3).
Table 10.1. Kinetic parameters of Cu2+-induced promoter–RBS library in E. coli DH5α
Symbol Promoter-RBS component PM ,ion Pm,ion Other kinetic

parameters
B1 PcusC-B0034 59.21404802 0.550287858 Kcus = 30.78179516
KCu2+ = 2×10-4
ncus = 0.589634725
xcusRS = 100.0301323
B2 PcusC-B0031 2.930570745 0.481929389 Kcus = 5.241765425
KCu2+ = 2.98255×10-4
ncus = 0.555913758
xcusRS = 100.0106996
B3 PcusC-B0032 2.5 0.582821058 Kcus = 10.58588213
KCu2+ = 9.62501×10-4
ncus = 0.637978871
xcusRS = 100.0000214
Table 10.2. Kinetic parameters of constitutive promoter–RBS library in E. coli DH5α
Symbol BioBrick part kinetic strength

C1 J23101-B0031 2.9727
C2 J23101-B0032 7.8578
C3 J23101-B0034 15.5418
C4 J23105-B0031 0.1994
C5 J23105-B0032 1.1591
C6 J23105-B0034 4.9142
C7 J23106-B0031 0.5730
C8 J23106-B0032 1.7395
C9 J23106-B0034 10.1788
C10 J23109-B0031 0.0859
C11 J23109-B0032 0.7293
C12 J23109-B0034 0.7135
C13 J23110-B0031 0.9624

C14 J23110-B0032 2.8423
C15 J23110-B0034 19.4795
C16 J23111-B0031
C17 J23111-B0032 23.4814
C18 J23111-B0034
Symbol BioBrick part kinetic strength
C19 J23114-B0031 0.0686
C20 J23114-B0032 0.7219
C21 J23114-B0034 1.2134
C22 J23118-B0031 1.3636
C23 J23118-B0032 5.5411
C24 J23118-B0034 38.4107
Table 10.3. Kinetic parameters of the activator-regulated promoter–RBS library,

Plux in E. coli DH5α
Symbol BioBrick part PM ,Lux Pm,Lux Other kinetic parameters
R1 R0062-B0031 48.78657 1.18476 KLux = 6.84573

KAHL = 7.3×10-8
nLux = 1.675
R2 R0062-B0032 40.58938 0.705558 KLux = 6.6898
KAHL = 9.6×10-8
nLux = 1.48795
R3 R0062-B0034 60.00649 1.027211 KLux = 10.00759
KAHL = 2.21×10-8
nLux = 2.939956
Dynamic Model of Copper Biosensor Circuit in Sender Cells
To choose the most appropriate promoter–RBS components, we needed to construct a

mathematical model capable of identifying the kinetic strength of a given promoter–RBS
component. The library-based design method has attracted much attention in recent years
(Lee et al., 2013, Wu, Lee, and Chen 2011, Wu, Zhang, and Chen 2011). Here we used three
types of component libraries. These were the copper-induced promoter–RBS library, the
constitutive promoter–RBS library, and the activator-regulated promoter–RBS library.
To estimate the kinetic strengths of promoter–RBS libraries, we used fluorescent proteins.
The characteristics of fluorescent proteins enable designers to measure the fluorescence
intensity of different promoter–RBS components. The following equation shows the dynamic
model of general protein expression:
xP (t )  Rp  PM , p , Pm, p , TFp , I   (d P  DP ) xP (t )  w(t )

(10.4)
where xP denotes the concentration of protein P and Rp  PM , p , Pm, p , TFp , I  denotes the gene
regulation function of gene p. PM , p is the maximum promoter–RBS strength and Pm, p is the
minimum promoter–RBS strength. TFp is the corresponding transcription factor
concentration of gene p, and I is the related inducer concentration. d P is the degradation rate
of protein P, and DP is the dilution rate of protein P. The gene circuit topology of the
molecular communication system is depicted in Figure 10.5, which also shows the dynamic
model of the synthetic molecular communication system.
First, we constructed a dynamic model of the sender cell. If we consider the promoter
PcusC to be represented by the copper biosensor shown in Figure 10.3, the equation (10.4)
can be modified to create the dynamic model of the sender cell with copper biosensor circuit
in the following manner:

xLuxI (t )  RcusC PM ,cusC , Pm,cusC , xCusR , I Cu 2 
(d LuxI  DLuxI ) xLuxI (t )  w1 (t )
(10.5)
in which
 
RcusC PM ,cusC , Pm,cusC , xCusR , I Cu 2 
PM ,cusC  Pm,cusC
Pm,cusC  nCusR
 K CusR 
1  * 
 
 xCusR xCusR , I 2
Cu  

*
xCusR  
xCusR , I Cu 2 
xCusR
K 2
1  Cu
I Cu 2

and in which xLuxI (t ) is the LuxI protein concentration. RcusC PM ,cusC , Pm,cusC , xCusR , I Cu2  is
the corresponding regulation function determined by the characteristics of promoter PcusC.
PM ,cusC is the maximum promoter–RBS strength of the promoter PcusC and Pm,cusC is the
minimum promoter–RBS strength of the promoter PcusC. The copper-induced promoter
PcusC is regulated by the regulatory protein. xCusR is the amount of CusR protein in CusSR
regulatory system. This is considered a constant because it is an endogenous chromosome in
E. coli. I Cu2 is the Cu2+ concentration in the experimental medium. dLuxI is the protein
degradation rate, and DLuxI is the protein dilution rate.
In order for the intracellular concentration to be equal to the concentration of copper we
add, the variation of the promoter PcusC activity should be considered. There are several
*
other kinetic parameters, such as K Cu 2 , K CusR , nCusR , xCusR , and w1 (t ) . First, K Cu 2 is the
dissociation rate between the copper inducer and the regulatory protein, KCusR is the binding
affinity between the regulatory protein and the cognate promoter sequence. Moreover, nCusR is
the binding cooperativity between the activated cusSR complex protein and the cognate
*
promoter–RBS components. Typically, the value of xCusR is the amount of inactivated
*
regulatory protein. For an activated regulatory protein, xCusR transcription downstream of
the promoter can be regulated. Finally, w1 (t ) is the level of environmental noise. This
includes instrument calibration, the experimental conditions, and the cellular noise in both the
transcriptional and translational processes used for imitating the cell process in vivo.
The dynamics for the catalysis of SAM and acyl-ACP into a specific AHL by LuxI
protein is provided by the following equation (see Appendix A for details):
K2
I AHL (t )  xLuxI (t )  (d AHL  DAHL ) I AHL (t )
1  Km (10.6)
where I AHL (t ) is the concentration of autoinducer AHL, and denotes the reaction constant in
the catalytic process of LuxI protein and AHL, respectively. dAHL denotes the degradation
rate of AHL, and DAHL denotes the dilution rate for autoinducer AHL.
Our constructed dynamic model of a sender cell with a copper biosensor circuit can be
used to identify and select the kinetic parameters of promoter–RBS components from the
library. The identified kinetic parameters can then be used to engineer a genetic circuit with
predictable behavior in the sender cell
Dynamic Models of Genetic Circuits in Receiver Cells
After the genetic circuit design of the sender cell was completed, we needed to design the
receiver cell to be able to select the promoter–RBS component with the most appropriate
kinetic strength. To estimate the kinetic strengths of the promoter–RBS libraries for the
receiver cell, we used fluorescent proteins to quantify the kinetic strengths of promoter–RBS
libraries. The GFP produced by the receiver cell is the reporter protein in Figure 10.4.
Dynamic models for the genetic circuits of receiver cells with different promoter–RBS
components are given as follows:
xLuxR (t )  RC  PC , 0, 0, 0 
(d LuxR  DLuxR ) xLuxR (t )  w2 (t ) (10.7)
xGFP (t )  Rlux  PM ,lux , Pm,lux , xLuxR , I AHL 

(d GFP  DGFP ) xGFP (t )  w3 (t ) (10.8)
in which
RC  PC , 0, 0, 0   PC
PM ,lux  Pm,lux
Rlux  PM ,lux , Pm,lux , xLuxR , I AHL   Pm,lux  nLuxR
 K LuxR 
1  * 
 xLuxR  xLuxR , I AHL  
xLuxR
*
xLuxR  xLuxR , I AHL   K
1  AHL
I AHL
Figure 10.4. Design scheme of the synthetic genetic circuit in the receiver cell. This figure shows the
scheme of the receiver cell with two kinds of promoter–RBS components: the constitutive promoter–
RBS component and the activator-regulated promoter–RBS component. The genetic circuit consists of
the lux operon derived from a V. fischeri quorum-sensing system.
The dynamic model of GFP intensity is valid only when the activator-regulated function
Rlux  PM ,lux , Pm,lux , xLuxR , I AHL  is activated, where PM ,lux is the maximum strength of Plux,
and Pm,lux is the minimum strength of Plux. xLuxR is the amount of transcription factor of
the promoter Plux, and I AHL is the inducer AHL concentration. K LuxR is the binding affinity
between the transcriptional activator and the cognate DNA sequence. K AHL is the
dissociation rate between LuxR protein and the inducer AHL. nLuxR is the binding
cooperativity between the activated LuxR complex protein and the target DNA. Finally,
w1 (t ) is the environmental and cellular noise, w2 (t ) is cellular noise in the transcriptional
process, and w3 (t ) is the cellular noise in the translational process.

Dynamic Models of the Multicellular Molecular Communication System
The dynamic models and processes of the MMCS are described as follows. For the
simplification of the design procedure of the MMCS, the steady state model was used. The
steady state model of the sender cell in (10.5) is given as follows:
xLuxISS 

RcusC PM ,cusC , Pm,cusC , xCusR , I Cu 2  
d LuxI  DLuxI
1
(10.9)
where xLuxISS denotes the LuxI protein concentration corresponding to the copper biosensor
systems at the steady state and 1 denotes the environmental and cellular noise. The
maximum strength and minimum strength of Cu2+-induced promoters, PM ,cusC and Pm,cusC , are
defined as the highest LuxI protein concentration when induced, and the basal LuxI protein
concentration when uninduced, respectively.
To make the steady state of the dynamic model fit the corresponding measurement, we
identified the parameter values in the dynamic equation using experimental data with reliable
processes. First, we identified the minimum strength required for the copper-induced
promoter by adding a zero inducer concentration into the regulation function as a basal LuxI
protein concentration:
RcusC  PM ,cusC , Pm,cusC , xCusR , 0   Pm,cusC

(10.10)
By including the regulation function without the inducer being added into the steady state
model, we could identify the kinetic strength. This was done by minimizing the error of the
GFP intensity between the simulation value in silico and the experimental measurement in
vivo. Next, other kinetic parameters were identified using a GA to search for the optimal
results by mimicking the process of natural selection. Before employing the GA searching
method, a fitness function needed to be set up to search for the suitable kinetic parameters
required at the steady state. In addition, the kinetic parameters of the constitutive promoter–
RBS components needed to be identified by the same procedure. We used the dynamic
equation to transform the constitutive promoter–RBS components into the steady state and
identified the corresponding kinetic strengths using GA searching, as shown in Table 10.2.
In this study, we used only the strongest RBS, B1 , to construct the promoter–RBS
component library in MMCS. The reason for this is that BBa_B0034 has a superior detection
capability compared with the other two RBSs. With the well-identified kinetic parameters,
and let the GFP intensity at the steady state correspond to the concentrations of copper ions to
identify the kinetic parameters for the corresponding libraries.
Table 3 lists the kinetic parameter strengths for the activator-regulated promoter–RBS
component library of circuit Plux. For the various system design specifications in (10.1)–
(10.3) of the proposed molecular communication system, we selected different kinetic
strengths of the circuit P{lux} with three strengths of RBSs. These wereBBa_B0031,
BBa_B0032, and BBa_B0034, taken from the activator-regulated promoter–RBS libraries in
Table 10.3. The maximum kinetic strength of the promoter–RBS component changed with the
distinct strengths of the RBSs, where the maximum kinetic strength indicates the maximum
GFP intensity when fully induced. In theory, RBSs can only affect the maximum kinetic
strength because they are involved in the translation efficiency. However, based on
experimental measurements, we also found that different RBSs result in a variety of other
kinetic parameters in dynamic models, as shown in Table 10.3.
Similarly, we could rewrite the dynamics of (10.7) and (10.8) at steady state for a
molecular communication system with a copper biosensor, as follows:

 xLuxISS 

RcusC PM ,cusC , Pm ,cusC , xCusR , I Cu 2+  
d LuxI  DLuxI
1

 1 K2
 I AHL  xLuxISS  2
 SS
d AHL  DAHL 1  K m

x RC ( PM ,C , 0, 0, 0)
  3
 LuxR SS d LuxR  DLuxR

 Rlux  PM ,lux , Pm ,lux , xLuxR , I AHL 
 xGFPSS   4
 d GFP  DGFP
(10.11)
where the Gaussian random noise  p , p  1, 2,3 , with a mean of zero and a variance of
 p2 , denotes cellular noise for both the transcriptional and translational gene expression
processes at the steady state, and 4 denotes cellular noise in mature protein expression at
the steady state.
Biological components are inherently uncertain in molecular biological systems. The
intrinsic kinetic parameters of the components of such systems include the processes of
transcription and translation, the degradation rates of regulatory proteins, dilution rates of the
cells, and the maturation rate for the reporter proteins. These are all stochastically uncertain in
vivo as a result of gene expression noise from biochemical processes, thermal fluctuations,
DNA mutation, and evolution. These perturbations are defined as follows:
PM ,cusC  PM ,cusC  PM ,cusC n1 (t ),

Pm ,cusC  Pm ,cusC  Pm ,cusC n1 (t ),
PC  PC  PC n1 (t ),
PM ,lux  PM ,lux  PM ,lux n1 (t ),
Pm ,lux  Pm ,lux  Pm ,lux n1 (t ),
DLuxI  DLuxI  DLuxI n1 (t ),
DLuxR  DLuxR  DLuxR n1 (t ),
DGFP  DGFP  DGFP n1 (t ) (10.12)
where PM ,cusC , Pm,cusC , PC , PM ,lux , Pm,lux , DLuxI , DLuxR , DGFP , m , D ,
and d are the standard deviations of the corresponding stochastic parameters, and nq (t ) ,
q  1, 2,3 denote Gaussian noise, which has a mean of zero and unit variance, to account for
sources of random fluctuations. If the kinetic parameters in the steady state model are
considered to be the parameter perturbations shown in (10.12) for a robust design of the gene
circuit, then the MMCS can tolerate these fluctuations in vivo. That is, a design that accounts
for random parameter perturbations (10.12) may be able to tolerate such parameter
fluctuations.
Figure 10.5. Design scheme of a synthetic genetic circuit in a multicellular molecular communication
system. This figure shows the schematic of a receiver cell with two kinds of promoter–RBS
components: a constitutive promoter–RBS component and an activator regulated promoter–RBS
component. The genetic circuit consists of the lux operon derived from the V. fischeri quorum-sensing
system.
Using a least-squares parameter identification method, we could effectively identify the

kinetic strengths of the promoter–RBS components from the corresponding experimental data.
After establishing promoter–RBS libraries in Table 10.1, Table 10.2, and Table 10.3 using a
GA, we searched for a suitable set of kinetic parameters for a synthetic genetic circuit from
the corresponding libraries. We did this by minimizing the square error between the simulated
result and the desired system performance so that the MMCS achieved optimal detection
capability, transduction ability, and sensitivity.
Design Specifications for the Multicellular Molecular

Communication System
The promoter–RBS libraries were built upon the parameter identiﬁcation of promoter–
RBS strengths, as shown in Table 10.1, Table 10.2, and Table 10.3. The purpose of our
design was to construct a robust MMCS. We did this by selecting a set of suitable
components from the corresponding libraries to achieve optimal desired detection capability,
transduction ability, and sensitivity within a feasible range of input concentrations. To
achieve the desired performance, the following design specifications were needed:
1. The desired detection capability Dref , transduction ability Tref , and sensitivity S ref
of the MMCS to be within the feasible range of metal ion concentrations, i.e.,
I Cu2 [ I1 , I 2 ]
.
2. Standard deviations of cellular disturbances and parameter fluctuations to be
tolerated in vivo. PM ,cusC , Pm,cusC , PC , PM ,lux , Pm,lux , DLuxI , DLuxR ,
and DGFP denote the standard deviations of corresponding parameter fluctuations
in (10.12).
3. Promoter–RBS components to be selected from the well-characterized promoter–
RBS libraries listed in Table 10.1, Table 10.2, and Table 10.3.
4. An explicit cost function to be defined within a specified range of metal ion
concentrations to achieve the system performance ( Dref , Tref , S ref ) as follows:
J ( L)  E  W1  Dsim ( L, I )  Dref 

2 I 2
I1 
W2 Tsim ( L, I )  Tref  W3  Ssim ( L, I )  S ref   dI

2 2
 (10.13)
Detection capability Dsim , transduction ability Tsim , and sensitivity Ssim in the feasible
I Cu2 in (1)–(3) are given in the following equations, respectively
I AHLSS ( L)
Dsim ( L, I Cu 2+ ) 
I Cu 2+
(10.14)
xGFPSS ( L)
Tsim ( L, I Cu 2+ ) 
I Cu 2+
(10.15)
xGFPSS ( L)
Ssim ( L, I Cu 2+ ) 
I Cu 2+
(10.16)
where
Wi denotes the weightings, 0  W1  1,0  W2  1,0  W3  1 and W1  W2  W3  1 ;
L {( Bi , C j , Ak ) | Bi  B, C j C, Ak  A} denotes
the set of promoter–RBS components
to be selected from the corresponding libraries in Table 10.1, Table 10.2, and Table 10.3,
respectively. Dsim ( L, I Cu2 ) and Dref represent the simulated and desired detection
capability. Tsim ( L, I Cu2 ) and Tref represent the simulated and desired transduction ability.
Ssim ( L, I Cu2 ) and S ref represent the simulated and desired sensitivity. I AHLSS ( L) denotes
the steady state AHL intensity of the sender cell and xGFPSS ( L) denotes the steady state
fluorescence intensity of the receiver cell.
Since expectation E is considered for random intrinsic parameter fluctuations and
environmental noise, a Monte Carlo simulation should be performed to account for the
minimization problem in (10.13). If the cost function in (10.13) can be minimized by
choosing the most appropriate set of promoter–RBS components under systematic design
specifications, the designed MMCS performance will optimally match the desired detection
Dref , transduction ability Tref and sensitivity S ref under parameter fluctuations and
environmental disturbances. As the scale of the component libraries is gradually expanded, a
more effective and efficient GA-based searching method may be adopted here to save time in
trial-and-error experimentation.
Design Procedure for the Robust Multicellular Molecular

Communication System
Our design target was to engineer a robust MMCS capable of achieving a desired
detection capability in (10.1), transduction ability in (10.2), and sensitivity in (10.3). We
provided a design procedure to minimize J ( L) in (10.13) to achieve this in the MMCS
using a promoter–RBS library searching method based on a GA:
1. Provide user-defined specifications of Dref , Tref , and S ref for the MMCS.
2. Select an initial set of promoter–RBS components L .
3. Calculate the cost value J ( L) for each set of the components of L . The
relationship between the fitness value F ( L) and the cost value J ( L) is inversely
proportional as follows:
1
F ( L) 
J ( L) (10.17)
The fitness value plays a key role in natural selection for identifying a set of promoter–
RBS components. The fitness value L from the corresponding libraries in Table 10.1, Table
10.2, and Table 10.3 may be used to maximize the fitness value F ( L) , or equivalently to
minimize the cost function J ( L) :
1
max L F ( L) 
min L J ( L) (10.18)
1. Create an offspring set of promoter–RBSs, using GA operators such as reproduction,

crossover, and mutation.
2. Calculate the cost value for the new set of promoter–RBS components obtained by
natural selection. Stop when the design goal is achieved or an acceptable solution is
obtained. Otherwise, create the next generation and return to step 4.
Design Examples of the Multicellular Molecular Communication System

In Silico and Verification of Experimental Results In Vivo
In this section, the MMCS was engineered for demonstration according to the previously
proposed design procedures. Here, nine design examples for a molecular communication
system (MCS) were given by selecting their adequate biological components from the
promoter–RBS libraries to achieve their desired system performance.
For a robust system design, the standard deviations of parameter fluctuations in vivo are
supposed to be 5% of the parameter values. In addition, the environmental disturbances i

w
of the transcription and translation processes are supposed to be independent Gaussian white
noise with means of zero and unit variances.
Table 10.4. Design specifications and their optimal solutions for a library-based search
method in silico, and verifications via experiments in vivo for MMCS with different
strengths of RBS components
MMCS# Transduction ability Sensitivity Detection ability

Tref Tsim Texp S ref Ssim Sexp Dref Dsim Dexp
MMCS1 500 504.6 495.7 400 402.1 352.47 200 217.6 206.8
MMCS2 150 168.7 145.65 150 152.66 129.61 200 217.6 206.8
MMCS3 70 69.36 69.35 60 57.19 50.43 200 217.6 206.8
MMCS4 35 33.77 27.87 30 23.909 18.01 200 217.6 206.8
MMCS5 350 325.7 359.33 300 274.49 338.45 200 217.6 206.8
MMCS6 350 382.4 432.30 150 169.8 170.70 200 217.6 206.8
MMCS7 350 366.6 358.07 50 48.2 40.79 200 217.6 206.8
MMCS8 300 302.72 308.3 200 210.15 192.56 200 217.6 206.8
MMCS9 250 237.7 259.12 200 201 245 200 217.6 206.8
Design Specification 1 (MMCS 1)
In this example the designer wants a molecular communication system with a high
transduction ability and high sensitivity. The desired transduction ability of Tref  500 ,
sensitivity of Sref  400 , and detection capability of Dref  200 , based on our design
procedure, gave the following:
1. The desired transduction ability Tref  500 , sensitivity Sref  400 , and detection
capability Dref  200 are within the feasible range
I 21 2[ I , I ]  [0.1 M,10 M]
of Cu .
2. The standard derivations of cellular disturbances and parameter fluctuations to be
tolerated in vivo are given by:
PM ,cusC  0.05PM ,cusC , Pm,cusC  0.05Pm,cusC ,

PM ,lux  0.05PM ,lux , Pm,lux  0.05Pm,lux ,
PM ,C  0.05PM ,C , DGFP  0.05DGFP ,
DLuxI  0.05DLuxI , DLuxR  0.05DLuxR (10.19)
In addition, the external environmental noise w p , p  1, 2,3 for the transcription and
translation processes, and noise w4 for mature reporter protein expression are all Gaussian.
The promoter–RBS components from the well-characterized promoter–RBS libraries
listed in Table 10.1, Table 10.2, and Table 10.3 were selected to minimize the following cost
function within the metal ion concentration range [0.1 M,10 M] .
min L J ( L)
1
10  M
 min L E   Dsim ( L, I )  200 
2


0.1 M 3
1 1 2
 Tsim ( L, I )  500    Ssim ( L, I )  400   dI
2
3 3 
where the detection capability, transduction ability, and sensitivity are within the feasible
I Cu2as denoted in (10.14)–(10.16).
Using the promoter–RBS library search, we identified a set of adequate promoter–RBS
components from Table 10.1, Table 10.2, and Table 10.3 to be L  ( B1 , C6 , R3 ) . The results
from a Monte Carlo simulation with an average of 100 runs (as shown in Table 4) robustly
matched the desired design specifications despite intrinsic parameter fluctuations and
environmental disturbances. The simulation found the transduction ability to be Tsim  504.6 ,
the sensitivity to be Ssim  402.1 , and the detection capability to be Dsim  217.6 .
Furthermore, the design specification for MMCS 1 could be verified with an experiment in
vivo. The experimental results found a transduction ability of Texp  495.7 , sensitivity of
Sexp  352.47 , and detection capability of Dexp  206.8 , as shown in Table 10.4. Finally,
the design specifications of MMCS 1 were simulated in silico and verified via
experimentation
in vivo with different copper input concentrations as shown in Figure 10.6a.
The MMCSs were modeled using a similar design procedure as above for the following 8
design specifications, which had different desired transduction abilities, sensitivities, and
detection capabilities. We found adequate promoter–RBS sets for the synthetic genetic circuit
for all MMCSs, and these are summarized in the following sections.
In this example, the designer wants a molecular communication system with a medium
transduction ability and a medium sensitivity result, i.e., the desired transduction
ability Tref  150 , sensitivity Sref  150 , and detection capability Dref  200 .
For the proposed design procedure, we identified an adequate set of promoter–RBS
components from the promoter–RBS libraries in Table 10.1, Table 10.2, and Table 10.3 and
to be L  ( B1 , C12 , R3 ) . The simulation found a transduction ability Tsim  168.7 ,
sensitivity Ssim  152.66 , and detection capability Dsim  217.6 . The results from the
Monte Carlo simulation with an average of 100 runs (as shown in Table 10.4) robustly
matched the desired design specifications despite intrinsic parameter fluctuations and
environmental disturbances. Furthermore, the design specifications for MMCS 2 could be
verified with an experiment in vivo. The experimental results determined a transduction
ability Texp  145.65 , sensitivity Sexp  129.61 , and detection capability Dexp  206.8 ,
as shown in Table 10.4. Finally, the design specifications of MMCS 2 were simulated in
silico and verified via in vivo experiments with different copper input concentrations as
shown in Figure 10.6b.
Figure 10.6. Multicellular molecular communication systems (MMCS 1 to 9) in E. coli DH5α with the
adequate set of promoter–RBS components from GA searching. The simulation results from an average
of 100 runs of Monte Carlo simulations of constructed dynamic equations (solid line) match the
experimental data (circle) with error bars under the adequate sets of promoter–RBS components
determined by GA searching.
In this example, the designer wants a molecular communication system with a low
transduction ability and a low sensitivity as follows: the desired transduction ability
of Tref  70 , sensitivity of Sref  60 , and detection capability of Dref  200 .
Using the proposed design procedure, we identified an adequate set of promoter–RBS
components from the promoter–RBS libraries in Table 10.1, Table 10.2, and Table 10.3 to be
L  ( B1 , C12 , R1 ) . The simulation found a transduction ability Tsim  69.36 , sensitivity Ssim  57.19 ,
and detection capability Dsim  217.6 . The results from the Monte Carlo simulation with an
average of 100 runs (as shown in Table 10.4) robustly matched the desired design
specifications despite intrinsic parameter fluctuations and environmental disturbances.
Furthermore, the design specification for MMCS 3 could be verified experimentally in vivo.
The experimental results found a transduction ability of Texp  69.35 , sensitivity of Sexp  50.43 , and
detection capability of Dexp  206.8 as shown in Table 10.4. Finally, the design specifications of
the MMCS 3 were simulated in silico and verified via experimentation in vivo with different
copper input concentrations, as shown in Figure 10.6c.
In this example, the designer wants a molecular communication system with a lower
transduction ability and a lower sensitivity (half of design specification 3), such that the
desired transduction ability Tref  35 , sensitivity Sref  30 , and detection capability Dref  200 .
Using the proposed design procedure, we identified an adequate set of promoter–RBS
components from the promoter–RBS libraries in Table 10.1, Table 10.2, and Table 10.3 to be
L  ( B1 , C12 , R2 ) . The simulation found a transduction ability of Tsim  33.77 , sensitivity of
Ssim  23.9 , and detection capability of Dsim  217.6 . The results from the Monte Carlo
simulation with an average of 100 runs (as shown in Table 10.4) robustly matched the desired
design specifications, despite the intrinsic parameter fluctuations and environmental
disturbances. Furthermore, the design specification for MMCS 4 were verified with an
experiment in vivo. The experimental results found a transduction ability of Texp  27.87 ,
sensitivity of Sexp  18.01 , and detection capability of Dexp  206.8 , as shown in Table 10.4.
Finally, the design specifications of the MMCS 4 were simulated in silico and verified via
experimentation in vivo with different copper input concentrations, as shown in Figure 10.6d.
In the following three design cases, the desired transduction ability and detection
capabilities remain the same, but the sensitivity is varied to be strong, medium, and weak,
respectively.
In this example the designer wants a desired sensitivity Sref  300 , transduction
ability Tref  350 , and detection capability Dref  200 . Using the proposed design procedure, we
identified an adequate set of promoter–RBS components from the promoter–RBS libraries in
Table 10.1, Table 10.2, and Table 10.3 to be L  ( B1, C6 , R1 ) . The simulation found a
transduction ability of Tsim  325.7 , sensitivity Ssim  274.49 , and detection capability Dsim  217.6
with the promoter–RBS component. The results from the Monte Carlo simulation with an
average of 100 runs (as shown in Table 4) robustly matched the desired design specifications,
despite intrinsic parameter fluctuations and environmental disturbances. Furthermore, the
design specification for the MMCS 5 could be verified experimentally in vivo. The
experimental results found that the transduction ability Texp  359.33 , sensitivity Sexp  338.45 , and
detection capability Dexp  206.8 , as shown in Table 10.4. Finally, the design specifications of
the MMCS 5 were simulated in silico and verified via in vivo experiments with different
copper input concentrations, as shown in Figure 10.6e.
In this example, the desired medium sensitivity was Sref  150 , or half of the original
design specification, with the desired transduction ability Tref  350 , and the desired detection
capability Dref  200 . Using the proposed design procedure, we identified an adequate set of
promoter–RBS components from the promoter–RBS libraries in Table 10.1, Table 10.2, and
Table 10.3 to be L  ( B1 , C3 , R1 ) . The simulation determined the transduction ability to be
Tsim  382.4 , sensitivity to be Ssim  169.8 , and detection capability to be Dsim  217.6 . The results from
the Monte Carlo simulation with an average of 100 runs (as shown in Table 10.4) robustly
matched the desired design specifications, despite intrinsic parameter fluctuations and
environmental disturbances. The experimental results found a transduction ability of
Texp  432.30 , sensitivity of Sexp  170.70 , and detection capability of Dexp  206.8 , as shown in
Table 10.4. Finally, the design specifications of MMCS 6 were simulated in silico and
verified via experimentation in vivo with different copper input concentrations as shown in
Figure 10.6f.
In this example, the designer wants a low sensitivity Sref  50 , one-third of the original
design specification in example 6, a desired transduction ability Tref  350 , and a desired
detection capability Dref  200 . Using the proposed library-based searching method, we
identified an adequate set of promoter–RBS components from the promoter–RBS libraries in
Table 10.1, Table 10.2, and Table 10.3 to be L  ( B1 , C3 , R3 ) . The simulation found that
transduction ability Tsim  366.6 , sensitivity Ssim  48.2 , and detection capability Dsim  217.6 . The
results from the Monte Carlo simulation with an average of 100 runs (as shown in Table 10.4)
robustly matched the desired design specifications, despite intrinsic parameter fluctuations
and environmental disturbances. The experimental results found a transduction ability
Texp  358.07 , sensitivity Sexp  40.79 , and detection capability Dexp  206.8 , as shown in Table
10.4. Finally, the design specifications of MMCS 7 were simulated in silico and verified via
experimentation in vivo with different copper input concentrations as shown in Figure 10.6g.
In the following two design cases, the desired sensitivity and detection capability were
kept the same, but transduction abilities differed.
In this example, the MMCS had a desired transduction ability Tref  300 , desired
sensitivity Sref  200 , and detection capability Dref  200 . Using the proposed library-based
searching method, we identified an adequate set of promoter–RBS components from the
promoter–RBS libraries in Table 10.1, Table 10.2, and Table 10.3 to be L  ( B1 , C3 , R2 ) . The
simulation found the transduction ability to be Tsim  302.72 , sensitivity to be Ssim  210.15 , and
detection capability to be Dsim  217.6 . The results from the Monte Carlo simulation with an
average of 100 runs (as shown in Table 10.4) robustly matched the desired design
specifications, despite intrinsic parameter fluctuations and environmental disturbances. The
experimental results found a transduction ability Texp  308.3 , sensitivity Sexp  192.56 , and
detection capability Dexp  206.8 (Table 10.4). Finally, the design specifications of MMCS 8
were simulated in silico and verified via experimentation in vivo with different copper input
concentrations as shown in Figure 10.6h.
In this example, the MMCS had a desired sensitivity Sref  200 , transduction
ability Tref  250 , and detection capability Dref  200 . Using the proposed library-based
searching method, we identified an adequate set of promoter–RBS components from the
promoter–RBS libraries in Table 10.1, Table 10.2, and Table 10.3 to be L  ( B1 , C6 , R2 ) . The
simulation found a transduction ability of Tsim  237.7 , sensitivity of Ssim  201 , and detection
capability of Dsim  217.6 . The results from the Monte Carlo simulation with an average of
100 runs (as shown in Table 4) robustly matched the desired design specifications, despite
intrinsic parameter fluctuations and environmental disturbances. The experimental results
found a transduction ability of Texp  259.12 , sensitivity of Sexp  245 , and detection capability of
Dexp  206.8 , as shown in Table 10.4. Finally, the design specifications of MMCS 9 were
simulated in silico and verified via experimentation in vivo with different copper input
concentrations as shown in Figure 10.6i.
10.3. DISCUSSION
In this chapter, we created and examined a systematic design method for a synthetic gene
circuit of a MMCS. We classified the molecular communication system into three major
modules. First, we used a copper induced promoter as the biosensor device to sense input data
from metal ions. The promoter PcusC encoded CusSR regulatory system was able to sense
copper directly via a membrane-associated sensor kinase. This is unlike other biosensor
devices, which require an additional chemical material to complete a sense operation. Second,
the sender cell emitted the diffusible small signal molecule AHL, which was synthesized by
LuxI proteins during the diffusion process. Third, the small signal molecules successfully
passed through channels to arrive at the designated receiver cell. This receiver cell contained
LuxR proteins, which bonded with the AHL molecules to form the AHL-LuxR complex. The
AHL-LuxR complex was then able to bind with the target promoter Plux and activate the
transcription of the downstream genetic circuit or green fluorescent protein.
To make the design of MMCS more efficient and robust, we constructed a mathematical
model of the genetic circuit. The model characterized the genetic circuit and described
dynamic and steady state regulatory behaviors with intrinsic parameter variations and
environmental noise, related to transcriptional and translational processes. We considered the
promoter and RBS together because the effects of transcription and translation regulation are
closely related. The promoter enabled the RNA polymerase molecules to latch onto a DNA
strand and initialize the transcription of a downstream gene into mRNA. In addition, the RBS
enabled the ribosome to bind to and translate the mRNA. For this reason we combined the
promoter with the RBS as a genetic design component for the synthetic genetic circuit of the
MMCS.
There was one primary challenge with regards to the process of constructing and
identifying the stochastic mathematical models of genetic circuit protein expressions with
random parameter fluctuations and disturbances. This challenge was to design a synthetic
molecular communication system with the ability to select suitable genetic components. Such
suitable components enable designers to achieve their desired system performance in
detection capability, transduction ability, and sensitivity within a feasible range of input
concentrations. In turn, this serves to guarantee system performance despite intrinsic
parametric variation and environmental disturbances.
In order to achieve user-defined specifications, we provided three promoter–RBS
libraries, which were searched using a GA search method to identify the appropriate set of
promoter-RBS components. This enabled designers to achieve several desired levels of
system performance for the MMCSs. Experimental findings showed that simulation results
could robustly predict system behavior and match the desired specifications of the MMCS.
We also constructed a design procedure for the MMCS to efficiently fulfill the desired
specifications using the proposed library-based search method via a GA. The future creation
of more complete promoter–RBS libraries will thus enable the addition of alternative types of
promoter–RBS components, which can then be searched for use in MMCS. This in turn will
give designers a greater diversity of design specifications.
10.4. CONCLUSION
In this chapter we provided design specifications for the systematic design of a MMCS,
which defined not only desired behaviors but also desired detection abilities, transduction
capabilities, and sensitivities. Simulation results supported the reliability of the MMCS
system performance using the proposed design method.
Furthermore, using our library-based search method we were able to efficiently select the
most appropriate components from corresponding promoter–RBS libraries to design a MMCS
with desired specifications. As a result of this systematic method, the MMCS was able to
achieve a reliable and robust synthetic gene design.
The synthetic gene network design coordinated the behavior of synthetic genetic circuits
distributed in different single cells through the MMCS. This not only allows designers to
construct complex genetic circuits beyond those possible with single-celled organisms, but
also to achieve coordinated synthetic behavior in molecular communities. Learning how to
engineer the robust synthetic gene design of MMCS is important for improving the
investigation of biological parts in modules and for the understanding of predictable synthetic
gene networks.
The key to the next generation in synthetic biology is to integrate simple biological
modules into higher-order systems. Fortunately, our work can independently control the high-
order systems of multiple genes in parallel through the MMCS. Moreover, we can reconstruct
such biological circuits by integrating several designed biological gene circuits. We believe
that this type of molecular communication system will be essential for the coordination of
multicellular systems for further practical applications. These may include environmental
monitoring and control, environmental remediation, biosensors, and biotechnology.
10.5. APPENDIX. DYNAMIC EQUATION OF CATALYSIS PROCESS

We constructed the dynamic models of participants in the catalysis process in Figure 10.7
as follows:
d[ ES ]
 K1[ E ][ S ]  K 1[ ES ]  K 2 [ ES ]
dt (10.20)
d[ P]
 K 2 [ ES ]
dt (10.21)
where [ E ] , [ S ] , and [ P] denote the concentration of enzymes, substances, and production,

respectively, in the catalysis process described in Figure 10.7. [ ES ] denotes the concentration
of the complex that is formed between the enzyme E and the substance in the transition state.
Figure 10.7. Catalysis process. E denotes the enzyme and S denotes the substance which will be
transformed into another particle in the catalysis process; ES denotes the complex which is formed by
enzyme E and the substances in the transition state of the catalysis process; P denotes the final
production in whole catalysis process.
As the catalysis reaction reaches the steady state, it is assumed that the dynamics of [ ES ]
d[ ES ]
will not change because generation equals consumption. That is,  0 . Then, we get
dt
the concentration of the complex at the steady state,
K1 1
[ ES ]  [ E ][ S ]  [ E ][ S ]
K 1  K 2 K m (10.22)
K 1  K 2
K 
where m K1 . In addition, according to the law of conservation of mass, we have
[ E]total  [ E]  [ ES ] .
The equation of the total amount of enzyme must be the sum of
individual enzymes and the enzyme forming the complex. Next, we took the law of
conservation of mass (10.22) into (10.21), and assumed the concentration of the substance
was a constant α. (10.21). This can be rewritten as:
d[ P] K 2
 [ E ]total  [ ES ] [S ]
dt K m
 K2  d[ P]
 [ E ]total 
K m K m dt
And we can get
d[ P] K2
 [ E ]total
dt 1  K m (10.23)
K
K  m
where m  . Finally, we considered that the products may be inactivated as time
passes by, so we added a term involving the degradation of the product in the equation. Then,
we derived the equation describing the relationship between the amount of the enzyme and
concentration of the product in (10.24).
d[ P] K2
 [ E ]total  d P [ P]
dt 1  K m (10.24)
REFERENCES
Bassler, B. L. 1999. “How bacteria talk to each other: regulation of gene expression by
quorum sensing.” Current Opinion in Microbiology 2 (6):582-587. doi: Doi
10.1016/S1369-5274(99)00025-9.
Basu, S., Y. Gerchman, C. H. Collins, F. H. Arnold, and R. Weiss. 2005. “A synthetic

multicellular system for programmed pattern formation.” Nature 434 (7037):1130-1134.
doi: 10.1038/nature03461.
Basu, S., R. Mehreja, S. Thiberge, M. T. Chen, and R. Weiss. 2004. “Spatiotemporal control
of gene expression with pulse-generating networks.” Proceedings of the National
Academy of Sciences of the United States of America 101 (17):6355-6360. doi:
10.1073/pnas.0307571101.
Bondarczuk, K., and Z. Piotrowska-Seget. 2013. “Molecular basis of active copper resistance
mechanisms in Gram-negative bacteria.” Cell Biol Toxicol 29 (6):397-405. doi:
10.1007/s10565-013-9262-1.
Brint, J. M., and D. E. Ohman. 1995. “Synthesis of Multiple Exoproducts in Pseudomonas-
Aeruginosa Is under the Control of Rhlr-Rhli, Another Set of Regulators in Strain Pao1
with Homology to the Autoinducer-Responsive Luxr-Luxi Family.” Journal of
Bacteriology 177 (24):7155-7163.
Nature Reviews Microbiology 12 (5):381-390. doi: 10.1038/nrmicro3239.
Chang, Y. C., C. L. Lin, and T. Jennawasin. 2013. “Design of Synthetic Genetic Oscillators
Using Evolutionary Optimization.” Evolutionary Bioinformatics 9. doi:
10.4137/Ebo.S11225.
Chen, Y. P., and D. E. Goldberg. 2005. “Convergence time for the linkage learning genetic
algorithm.” Evol Comput 13 (3):279-302. doi: 10.1162/1063656054794806.
Darch, S. E., S. A. West, K. Winzer, and S. P. Diggle. 2012. “Density-dependent fitness
benefits in quorum-sensing bacterial populations.” Proceedings of the National Academy
of Sciences of the United States of America 109 (21):8259-8263. doi:
10.1073/pnas.1118131109.
Dunlap, P. V., and A. Kuo. 1992. “Cell Density-Dependent Modulation of the Vibrio-Fischeri
Luminescence System in the Absence of Autoinducer and Luxr Protein.” Journal of
Bacteriology 174 (8):2440-2448.
Elowitz, M. B., and S. Leibler. 2000. “A synthetic oscillatory network of transcriptional
regulators.” Nature 403 (6767):335-338. doi: Doi 10.1038/35002125.
Endy, D. 2005. “Foundations for engineering biology.” Nature 438 (7067):449-453. doi:
10.1038/nature04342.
Franke, S., G. Grass, C. Rensing, and D. H. Nies. 2003. “Molecular analysis of the copper-
transporting efflux system CusCFBA of Escherichia coli.” Journal of Bacteriology 185
(13):3804-3812. doi: 10.1128/Jb.185.13.3804-3812.2003.
Gardner, T. S., C. R. Cantor, and J. J. Collins. 2000. “Construction of a genetic toggle switch
in Escherichia coli.” Nature 403 (6767):339-342.
Hanzelka, B. L., and E. P. Greenberg. 1996. “Quorum sensing in Vibrio fischeri: Evidence
that S-adenosylmethionine is the amino acid substrate for autoinducer synthesis.” Journal
of Bacteriology 178 (17):5291-5294.
Kelly, J. R., A. J. Rubin, J. H. Davis, C. M. Ajo-Franklin, J. Cumbers, M. J. Czar, K. de
Mora, A. L. Glieberman, D. D. Monie, and D. Endy. 2009. “Measuring the activity of
BioBrick promoters using an in vivo reference standard.” J Biol Eng 3:4. doi:
10.1186/1754-1611-3-4.
Khalil, A. S., and J. J. Collins. 2010. “Synthetic biology: applications come of age.” Nature
Reviews Genetics 11 (5):367-379. doi: 10.1038/nrg2775.
Lee, Y. Y., C. Y. Hsu, L. J. Lin, C. C. Chang, H. C. Cheng, T. H. Yeh, R. H. Hu, C. Lin, Z.

Xie, and B. S. Chen. 2013. “Systematic design methodology for robust genetic transistors
based on I/O specifications via promoter-RBS libraries.” Bmc Systems Biology 7. doi:
Artn 109 10.1186/1752-0509-7-109.
Lu, T. K., A. S. Khalil, and J. J. Collins. 2009. “Next-generation synthetic gene networks.”
Nature Biotechnology 27 (12):1139-1150. doi: 10.1038/nbt.1591.
Macomber, L., and J. A. Imlay. 2009. “The iron-sulfur clusters of dehydratases are primary
intracellular targets of copper toxicity.” Proceedings of the National Academy of Sciences
of the United States of America 106 (20):8344-8349. doi: 10.1073/pnas.0812808106.
McDaniel, R., and R. Weiss. 2005. “Advances in synthetic biology: on the path from
prototypes to applications.” Current Opinion in Biotechnology 16 (4):476-483. doi:
10.1016/j.copbio.2005.07.002.
Mcfallngai, M. J., and E. G. Ruby. 1991. “Symbiont Recognition and Subsequent
Morphogenesis as Early Events in an Animal-Bacterial Mutualism.” Science 254
(5037):1491-1493. doi: DOI 10.1126/science.1962208.
Mikami, Y., K. Yamamoto, Y. Akiyama, M. Kobayashi, E. Watanabe, N. Watanabe, M.
Asano, N. Shimizu, and K. Komiyama. 2015. “Osteogenic Gene Transcription Is
Regulated via Gap Junction-Mediated Cell-Cell Communication.” Stem Cells and
Development 24 (2):214-227. doi: 10.1089/scd.2014.0060.
Miller, M. B., and B. L. Bassler. 2001. “Quorum sensing in bacteria.” Annual Review of
Microbiology 55:165-199. doi: DOI 10.1146/annurev.micro.55.1.165.
Journal of Bacteriology 182 (20):5864-5871. doi: Doi 10.1128/Jb.182.20.5864-
5871.2000.
Pai, A., Y. Tanouchi, C. H. Collins, and L. You. 2009. “Engineering multicellular systems by
cell-cell communication.” Curr Opin Biotechnol 20 (4):461-70. doi:
10.1016/j.copbio.2009.08.006.
Passador, L., J. M. Cook, M. J. Gambello, L. Rust, and B. H. Iglewski. 1993. “Expression of
Pseudomonas-Aeruginosa Virulence Genes Requires Cell-to-Cell Communication.”
Pearson, J. P., L. Passador, B. H. Iglewski, and E. P. Greenberg. 1995. “A second N-
acylhomoserine lactone signal produced by Pseudomonas aeruginosa.” Proc Natl Acad
Sci U S A 92 (5):1490-4.
Ravikumar, S., I. K. Yoo, S. Y. Lee, and S. H. Hong. 2011. “Construction of Copper
Removing Bacteria Through the Integration of Two-Component System and Cell Surface
Display.” Applied Biochemistry and Biotechnology 165 (7-8):1674-1681. doi:
10.1007/s12010-011-9386-9.
Regot, S., J. Macia, N. Conde, K. Furukawa, J. Kjellen, T. Peeters, S. Hohmann, E. de Nadal,
F. Posas, and R. Sole. 2011. “Distributed biological computation with multicellular
engineered networks.” Nature 469 (7329):207-211. doi: 10.1038/nature09679.
Rinaudo, K., L. Bleris, R. Maddamsetti, S. Subramanian, R. Weiss, and Y. Benenson. 2007.
“A universal RNAi-based logic evaluator that operates in mammalian cells.” Nature
Biotechnology 25 (7):795-801. doi: 10.1038/nbt1307.
Ruby, E. G., and K. H. Nealson. 1976. “Symbiotic association of Photobacterium fischeri

with the marine luminous fish Monocentris japonica; a model of symbiosis based on
bacterial studies.” Biol Bull 151 (3):574-86.
Silva-Rocha, R., and V. de Lorenzo. 2014. “Engineering Multicellular Logic in Bacteria with
Metabolic Wires.” Acs Synthetic Biology 3 (4):204-209. doi: 10.1021/sb400064y.
Sohka, T., R. A. Heins, and M. Ostermeier. 2009. “Morphogen-defined patterning of
Escherichia coli enabled by an externally tunable band-pass filter.” J Biol Eng 3:10. doi:
10.1186/1754-1611-3-10.
2009. “An externally tunable bacterial band-pass filter.” Proceedings of the National
Academy of Sciences of the United States of America 106 (25):10135-10140. doi:
10.1073/pnas.0901246106.
Stricker, J., S. Cookson, M. R. Bennett, W. H. Mather, L. S. Tsimring, and J. Hasty. 2008. “A
fast, robust and tunable synthetic gene oscillator.” Nature 456 (7221):516-U39. doi:
10.1038/nature07389.
Tamsir, A., J. J. Tabor, and C. A. Voigt. 2011. “Robust multicellular computing using
genetically encoded NOR gates and chemical 'wires'.” Nature 469 (7329):212-215. doi:
10.1038/nature09565.
Wang, B. J., R. I. Kitney, N. Joly, and M. Buck. 2011. “Engineering modular and orthogonal
genetic logic gates for robust digital-like synthetic biology.” Nature Communications 2.
doi: ARTN 508 10.1038/ncomms1516.
Waters, C. M., and B. L. Bassler. 2005. “Quorum sensing: cell-to-cell communication in
bacteria.” Annu Rev Cell Dev Biol 21:319-46. doi: 10.1146/
annurev.cellbio.21.012704.131001.
Weiss, Ron, and ThomasF Knight, Jr. 2001. “Engineered communications for microbial
robotics.” In DNA Computing, edited by Anne Condon and Grzegorz Rozenberg, 1-16.
Springer Berlin Heidelberg.
Wu, C. H., H. C. Lee, and B. S. Chen. 2011. “Robust synthetic gene network design via
library-based search method.” Bioinformatics 27 (19):2700-2706. doi:
Wu, C. H., W. H. Zhang, and B. S. Chen. 2011. “Multiobjective H-2/H-infinity synthetic
gene network design based on promoter libraries.” Mathematical Biosciences 233
(2):111-125. doi: 10.1016/j.mbs.2011.07.001.
You, L. C., R. S. Cox, R. Weiss, and F. H. Arnold. 2004. “Programmed population control by
cell-cell communication and regulated killing.” Nature 428 (6985):868-871. doi:
10.1038/nature02491.
ABOUT THE AUTHOR
Prof. Chen received the Ph.D degree from the University of Southern California in 1982.
He was a professor at National Tsing Hua University and became a distinguished chair
professor in 2014. He is a life fellow of IEEE. He has published about 300 journal papers in
control, signal processing, communication, systems and synthetic biology. Dr. Hsu received
his Ph.D degree form National Tsing Hua University in 2015.
INDEX
bacteriorhodopsin (BR), 3, 168, 171, 172, 176, 180

A binding cooperativity, 14, 31, 33, 65, 66, 75, 76, 85,
109, 187, 188, 213, 214
activator–promoter binding affinity, 33
BioBrick parts, 9, 58, 168
activator-regulated promoter-RBS component, 10,
BioBrick standard assembly method, 74
13, 32, 33, 109, 110, 111, 113, 116, 187, 188,
BioBrick-related data, 9
193, 208, 210, 214, 216, 217
BioBricks, 8, 38, 39, 55, 168
activator-regulated promoter-RBS library, 26, 32, 33,
biofuels, 5, 178, 181
37, 38, 40, 51, 117
biological band-pass filter, 24, 26
acyl-acyl carrier protein (acyl-ACP), 63, 64, 65, 207,
biological component libraries, 2
213
biological filter, ix, 2, 7, 8, 9, 10, 11, 13, 14, 15, 16,
acylated homoserine lactone (AHL), 3, 25, 62, 63,
17, 18, 19, 22, 23, 24, 25, 26, 165
64, 84, 113, 114, 115, 123, 125, 126, 127, 128,
biological filter design, 7, 10, 11, 17, 23, 165
129, 130, 131, 133, 134, 137, 138, 139, 140, 141,
biological gene circuits, 1, 23, 178, 228
146, 147, 148, 150, 161, 163, 184, 185, 186, 188,
biological high-pass filter, 12, 16, 17, 21, 22, 23, 25
197, 205, 206, 207, 213, 214, 219, 227
biological low-pass filter, 11, 16, 17, 19, 20, 21, 25
adenosine triphosphate (ATP), 168, 172, 176
biological low-pass I/O filtering response, 15
AHL-activated promoter-RBS pair, 139
biological switchboard, 24, 25, 26
algae, 167, 178, 179, 181
bioremediation, 62, 72, 121
amplifier design example of synthetic genetic
biosensor, v, vi, ix, 1, 3, 38, 51, 61, 63, 64, 66, 68,
transistor, 47
71, 72, 73, 79, 81, 82, 83, 84, 86, 87, 88, 89, 90,
antibiotic chloramphenicol, 185, 186, 191, 192, 193,
91, 92, 93, 94, 95, 96, 204, 206, 208, 211, 212,
196, 197
213, 215, 216, 227
antibiotic chloramphenicol concentration, 191, 192,
blue-green algae, 167, 179
193, 196
brake component, 3, 121, 124, 125, 127, 129, 130,
antibiotic resistance gene, 4, 186, 198
131, 132, 133, 134, 135, 136, 137, 140, 141, 145,
applicability, 2, 7, 24, 26, 37
159, 161, 163
autoinducer, 64, 65, 66, 75, 76, 77, 78, 80, 85, 102,
BR-induced proton motive force, 171
103, 184, 187, 188, 191, 197, 204, 207, 213, 230,
231, 232
C
B C6-HSL, 63, 205
cascade gene circuit topology, 2, 7, 8, 9, 10, 12, 15,
bacteria population, 124, 128, 130, 132, 149, 150,
16, 26
184, 197
catalysis process, 228, 229
bacterial band-pass filter, 1, 5, 35, 58, 80, 232
cell population control, 3, 105, 107, 116, 165, 168,
bacterial filter, 1
169, 171, 172, 173, 176
bacterial motility, 3
238 Index
cell population density (OD600), 3, 56, 74, 105, 106, corresponding libraries, 2, 3, 9, 10, 11, 16, 17, 18,
107, 108, 109, 110, 111, 112, 113, 115, 116, 168, 21, 22, 26, 27, 48, 50, 52, 61, 66, 68, 71, 81, 86,
172, 173, 174, 175, 176, 187, 188, 189, 192, 193, 87, 88, 92, 93, 94, 95, 111, 113, 114, 168, 183,
194, 196, 197 184, 185, 191, 192, 193, 205, 207, 208, 210, 216,
cell population density control, 112, 116 218, 219, 220
cell-cell communication, ix, 61, 62, 183, 185, 188, Cu(II) bioadsorption, 72
193, 196, 197, 198, 200, 201, 231, 233 CusSR regulatory system, 204, 212, 227
cell-cell communication process, 61 cyanobacteria, 167, 168, 169, 170, 171, 172, 176,
cell-to-cell communication, 204, 205, 208, 232 179, 180, 181, 182
cell-to-cell communication system, 204 cyanobacterium, vi, 3, 167, 174, 175, 180, 181, 182
cell-type classifier, 25 cyanobacterium Synechococcous sp. PCC 7942, 3
cellular noise, 3, 67, 81, 82, 83, 96, 101, 109, 110,
173, 190, 213, 215, 216
chemotactic molecules, 122, 124 D
chemotaxis phenomenon, 148
data acquisition and processing, 164
chloramphenicol acetyltransferase (CAT) enzymes,
design of synthetic genetic lysis circuit, 115
197
design problem of the genetic lysis circuit, 113
chloramphenicol resistance, 185, 187, 188, 189, 191,
design procedure, 8, 17, 19, 24, 26, 39, 44, 46, 48,
193, 197
50, 51, 68, 89, 91, 92, 93, 111, 121, 145, 146,
commensalism, 184
147, 167, 175, 179, 183, 191, 192, 193, 196, 215,
communication mechanism between ER and EG
219, 220, 221, 222, 224, 225, 227
cells, 185, 186, 193
design procedure for a QS-based metal ion
complex microbial ecosystems, 3, 197
biosensor, 68, 92
computational circuits, 204
design procedure of a biological amplifier, 46
constitutive promoter-RBS component, 3, 10, 11, 12,
design procedure of a synthetic genetic transistor, 46,
13, 14, 15, 19, 21, 28, 30, 32, 44, 64, 65, 69, 74,
48
75, 77, 83, 84, 85, 108, 109, 110, 111, 113, 169,
design schematic of the genetic circuit for population
170, 172, 176, 187, 188, 189, 192, 193, 194, 195,
control, 174
196, 198, 208, 210, 214, 215, 217
design scheme of a synthetic genetic circuit, 209,
constitutive promoter-RBS component library, 3, 77,
217
170, 198
design scheme of GFP protein expression, 209
constitutive promoter–RBS component sequence,
design scheme of the synthetic genetic circuit, 214
208
design specifications, ix, 2, 9, 18, 19, 24, 26, 37, 38,
constitutive promoter-RBS components, 14, 28, 65,
39, 46, 49, 50, 51, 52, 68, 69, 72, 81, 82, 83, 86,
75, 109, 110, 111, 169, 170, 172, 176, 189, 192,
92, 96, 105, 111, 112, 116, 174, 175, 179, 180,
193, 194, 195, 196
192, 193, 199, 204, 205, 216, 218, 219, 221, 222,
constitutive promoter-RBS library, 28, 29, 30, 32,
223, 224, 225, 226, 228
37, 38, 51, 117, 170, 171, 179
design specifications and their optimal solutions, 221
constitutive promoter-RBS library in cyanobacteria,
design specifications for the genetic lysis circuit, 112
179
desired input/output (I/O) response, 3, 81
construction of genetic lysis circuit, 116
detection abilities, 228
controllable quorum sensing symbiotic ecosystem,
detection capability, 203, 206, 207, 210, 215, 218,
196, 198
219, 221, 222, 223, 224, 225, 226, 227
controllable symbiotic ecosystem, 183, 197, 198
diffusion experiment procedure and GFP expression
copper biosensor, 204, 208, 212, 213, 215, 216
assay, 164
copper biosensor circuit, 204, 208, 212, 213
diffusion process, 206, 207, 227
copper ion biosensor, 71
diffusion rate, 124, 125, 130, 131, 132, 133, 134,
copper-induced promoter PcusC, 212
135, 136, 137, 138, 140, 141, 142, 144, 145, 146,
copper-rich environments, 63
147, 148, 150, 164
Index 239
dynamic equation of antibiotic resistance circuits in

EG cells, 187
G
dynamic equations, 138, 198, 223
GA-based searching method, 2, 21, 22, 52, 63, 71,
dynamic model, ix, 2, 8, 12, 13, 14, 27, 37, 40, 42,
72, 107, 113, 116, 175, 180, 219
47, 48, 64, 66, 76, 84, 89, 108, 109, 139, 140,
gene circuit topology, 2, 7, 8, 9, 10, 12, 13, 15, 16,
172, 173, 186, 188, 189, 196, 210, 211, 212, 213,
17, 18, 24, 26, 30, 212
214, 215, 216, 228
genetic algorithm (GA)-based method, 2, 3, 4, 9, 18,
dynamic model for the biological filter, 13
26, 34, 37, 39, 46, 61, 62, 68, 102, 106, 112, 132,
dynamic model of a cell population control genetic
134, 139, 159, 169, 175, 184, 192, 205, 230
circuit, 172
genetic circuit construction, 162
dynamic model of a synthetic genetic transistor
genetic copper biosensor circuit, 204
circuit, 42
genetic lysis circuit, ix, 3, 105, 106, 107, 108, 109,
dynamic model of general protein expression, 211
110, 111, 112, 113, 114, 115, 116, 165
dynamic model of the antibiotic resistance circuit in
genetic regulation network, 7
the ER cells, 187
genetic switchboard, 24, 33, 56
dynamic model of the genetic lysis circuit, 108, 109
genetic transistor, 1, 2, 5, 37, 39, 40, 41, 42, 43, 44,
dynamic model of the sender cell, 212
45, 46, 47, 48, 49, 50, 51, 52, 165, 231
dynamic models for the genetic circuits of receiver
genetic transistor design specifications, 46, 52
cells, 213
genetic transistors with desired I/O responses, 50, 51
dynamic models of population density of ER and
gram-negative bacteria, 77, 197, 230
EG, 188
gram-positive bacteria, 197
green fluorescent protein (GFP) coding sequence, 14,
E 15, 19, 21, 25, 28, 29, 40, 41, 42, 56, 64, 74, 84,
85, 116, 125, 126, 127, 128, 130, 133, 134, 135,
electronic transistor, 44, 51, 53 138, 139, 140, 144, 145, 161, 163, 164, 185, 208,
ELISA, 13, 26, 27, 29, 127, 128, 139, 164 209, 213, 214, 215, 216
environmental bioremediation, 62, 72
environmental detection, 7, 62
environmental molecular noise, 85, 88, 89
H
environmental noise, 28, 40, 70, 93, 113, 173, 179,
H2 optimal tracking, 82, 83
189, 191, 193, 213, 219, 221, 227
H2/H∞ multi-objective design, 83
ER and EG, 185, 186, 188, 189, 192, 193, 196, 198
H∞ filtering performance, 87, 88
Escherichia coli (E. coli), 1, 4, 7, 25, 29, 34, 35, 38,
H∞ optimal robust attenuation, 83
55, 57, 58, 62, 63, 73, 76, 77, 78, 79, 80, 82, 102,
Halobacterium salinarum, 171
105, 106, 107, 108, 112, 114, 116, 118, 122, 123,
Hamilton Jacobi inequality (HJI)-constrained
124, 125, 126, 127, 128, 129, 131, 132, 133, 134,
optimization problem, 82, 89, 91, 95, 96
137, 139, 148, 162, 163, 164, 165, 166, 180, 182,
high-pass filter, 9, 11, 12, 16, 17, 18, 21, 22, 24, 25,
183, 184, 186, 196, 199, 204, 206, 209, 210, 211,
26
212, 223, 231, 232
high-pass I/O filtering response, 16
ethanol production in cyanobacteria, 179
HJI-constrained multi-objective problem, 89
externally tunable biological band-pass filter, 26
human pathogen Pseudomonas aeruginosa, 186,
externally tunable biological filter, 7, 8, 9, 18, 23, 25,
197, 205
26
F I
I/O characteristics of a synthetic genetic transistor,

fabrication of the microchannel, 163
42, 44, 49
fitness function, 112, 215
I/O filtering response, 2, 7, 8, 9, 10, 11, 12, 14, 15,
fluorescence measurement, 73
16, 17, 18, 19, 26, 165
four design specifications, 18, 19, 26, 72, 116, 175,
I/O response of an amplifier, 46
192
I/O response of the synthetic genetic transistor
fuzzy interpolation method, 90
circuit, 43
240 Index
identification of the parameters, 159 lysis ability, 108, 115

inducer, 3, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, lysis gene, 3, 105, 107, 113, 114, 118
19, 21, 23, 24, 25, 26, 27, 30, 31, 32, 37, 38, 39, lysis protein, 106, 108, 109, 110, 113, 118
41, 42, 43, 44, 45, 46, 47, 49, 50, 51, 52, 64, 74,
75, 105, 106, 107, 108, 109, 110, 111, 112, 113,
114, 115, 116, 138, 169, 187, 188, 208, 212, 213, M
214, 215
mature reporter protein, 12, 13, 15, 18, 27, 28, 66,
inducer-bound activator, 32, 75
67, 70, 74, 76, 85, 93, 221
inducer-regulated promoter-RBS component, 105
maximum and minimum kinetic strengths, 188
inducible activator-regulated circuit, 107
maximum and minimum promoter-RBS strengths,
inducible repressor-regulated circuit, 107
13, 14, 18, 27, 30, 31, 32, 33, 64, 65, 66, 74, 75,
input/output (I/O) filtering response, 3, 7, 8, 11, 17,
76, 85, 108, 109, 187
19, 20, 21, 22, 24, 43, 45, 46, 47, 48, 49, 50, 51,
maximum fitness, 112
52, 68, 81, 82, 83, 86, 87, 88, 89, 91, 92, 93, 94,
metal ion biosensors, 2, 61, 62, 63, 64, 66, 67, 68,
95, 96, 105, 111, 113, 114, 116
69, 70, 71, 81, 82, 83, 84, 86, 87, 88, 89, 90, 91,
interspecies interaction, 197
92, 93, 94, 95, 96, 101
intrinsic random fluctuations and channel noise, 203,
metal ion pollutants, 81
207
metal ion-induced promoter-RBS component, 65, 69,
ion-induced promoter, 62, 63, 64, 65, 69, 71, 74, 75,
71, 74, 75, 76, 83, 84, 85
76, 83, 84, 85
metal ion-induced promoter-RBS component library,
ion-induced promoter-RBS component, 65, 69, 71,
76
74, 75, 76, 83, 84, 85
methodology for measuring bacterial population
diffusion rates, 132
K microbial ecosystems, 4, 183, 184, 196, 197, 198,
200, 201
kinetic parameters of constitutive promoter–RBS microfluidic device, 124, 125, 130, 131, 132, 134,
library, 210 147, 148, 150, 164
kinetic parameters of the activator-regulated minimum and maximum promoter-RBS strengths,
promoter–RBS library, 211 169
mixed library-based search method, 8, 9, 17, 18, 19,
21, 22, 24, 26, 179
L MMCS design scheme, 203
MMCS system performance, 228
lead-contaminants soil, 63
model of symbiosis of ecosystem, 183
least-squares parameter identification method, 218
MOEA-based library search method, 3, 81, 83, 92,
library-based GA search method, 51, 52, 197
93, 94, 95, 96
library-based search method, 2, 6, 8, 9, 17, 18, 19,
multicellular logic, 204
21, 22, 24, 26, 35, 59, 167, 175, 176, 179, 191,
multicellular molecular communication system
192, 201, 221, 228, 233
(MMCS), 203, 204, 205, 206, 207, 208, 215, 217,
light-driven proton pump, 168, 171, 176
218, 219, 220, 221, 222, 223, 224, 225, 226, 227,
LMI-based multi-objective I/O response-matching
228
problem, 92
multicellular synthetic system, 204
LMI-constrained MO H2/H∞ design problems., 3, 81
multi-objective (MO) H2/H∞ performance criterion,
LMI-constrained multi-objective optimizations, 91
82, 95
LMIs-constrained multi-objective optimization
multi-objective design problem, 83, 90, 92, 95, 96
problem, 82
multi-objective evolutionary algorithm (MOEA)-
look-up table, 37, 39, 46, 50, 51, 52
based library search method, 3, 81, 83, 92, 93, 94,
low-pass filter, 11, 12, 16, 17, 18, 19, 20, 21, 24, 25
95, 96
lux operon, 62, 63, 128, 208, 214, 217
multi-objective H2/H∞ design, 81, 83, 89
luxI component, 63
multi-objective H2/H∞ I/O response matching
LuxI-type autoinducer synthase, 197
design, 82, 96
LuxR protein, 62, 63, 84, 125, 205, 206, 207, 208,
multi-objective H2/H∞ performance, 82, 95
214, 227
multi-objective H2/H∞ performance criterion, 82, 95
Index 241
multi-objective optimal design, 89 population density control, 112, 116, 117, 170, 179
multi-objective problem, 3, 81, 88, 89, 91, 94, 98 Population density time profiles, 115
mutualism, 79, 184, 196, 231 practicability, 7, 24, 26
mutualistic behavior, 183, 184, 185, 189, 191, 192, predator-prey ecosystem, 183, 197, 199
193, 196, 197, 198 primers, 126, 149, 162
mutualistic behaviors through cell-cell procedure for construction of the constitutive
communication, 196 promoter-RBS library, 171
mutualistic dynamics, 185 programmed mutualistic behavior, 198
mutualistic interaction, 4 promoter PcusC, 204, 209, 210, 212, 227
promoter-RBS component library, ix, 2, 3, 5, 7, 8, 9,
10, 11, 13, 14, 15, 17, 18, 19, 21, 23, 25, 26, 27,
N 28, 29, 30, 31, 32, 33, 37, 38, 39, 40, 44, 45, 46,
47, 49, 50, 51, 52, 63, 76, 77, 91, 96, 105, 107,
N-3-homoserine lactone (HSL), 63, 184, 185, 186,
110, 117, 165, 167, 169, 170, 171, 175, 176, 178,
187, 188, 189, 193, 205
179, 180, 191, 198, 205, 207, 208, 209, 210, 211,
next-generation gene network, 204
213, 214, 215, 216, 217, 218, 219, 220, 222, 223,
nonlinear least squares method, 38, 40
224, 225, 226, 228, 231
nonlinear stochastic molecular systems, 82
promoter-RBS kinetic strengths, 9, 26, 63, 107, 169
nonlinear stochastic QS-based metal ion biosensor,
promoter-RBS library, 3, 7, 8, 14, 26, 28, 29, 30, 31,
90
32, 33, 37, 45, 46, 47, 49, 50, 51, 117, 167, 169,
171, 179
O promoter-RBS library searching method, 46
promoter-RBS regulation activity, 65, 109, 187, 188
optimal filtering of parameter fluctuations and promoter-RBS regulation function, 13, 15, 16, 27,
environmental disturbances, 89, 91 28, 31, 32, 64, 75, 76, 108, 169, 187
optimal H∞ filtering, 96 promoter-RBS strengths, 10, 13, 14, 18, 27, 29, 30,
orthogonal genetic switches, 24 31, 32, 33, 64, 65, 66, 74, 75, 76, 85, 108, 109,
167, 169, 172, 187, 191
proposed design procedure for an externally tunable
P biological filter, 18
Pseudomonas aeruginosa, 77, 79, 119, 186, 197,
parameter fluctuations, 3, 9, 17, 18, 19, 20, 21, 22, 200, 201, 205, 232
23, 26, 44, 46, 47, 48, 68, 69, 71, 81, 82, 83, 86, pulse-generator, 204
87, 88, 89, 90, 91, 93, 96, 101, 111, 112, 113,
174, 175, 176, 192, 193, 217, 218, 219, 220, 221,
222, 223, 224, 225, 226, 227 Q
parameter perturbations, 68, 111, 190, 191, 217
parameter variations, 43, 174, 176, 179, 227 QS-based amplifier, 2, 63, 64, 71, 72
parasitism, 184 QS-based biosensor, 72
pareto front, 92, 93, 94 QS-based metal ion biosensor, 62, 64, 66, 67, 68, 69,
pareto optimal solution, 83, 92, 95 70, 71, 86, 89, 90, 91, 92, 93, 95, 101
part-based mathematical model, 143 QS-based metal ion detection circuit, 62
parts library, 144, 145, 146, 147, 149 QS-dependent promoter-RBS component, 64, 66, 69,
pattern formation, 4, 7, 24, 56, 199, 230 74, 75, 76, 77, 83, 84, 85
PCC 7942, vi, 3, 167, 168, 171, 172, 174, 175, 177, QS-dependent promoter-RBS component library, 77
179, 180, 181, 182 QS-dependent promoter-RBS part, 65
PcusC, 62, 63, 75, 76, 82, 205, 209, 210, 212, 227 quorum, v, vi, 3, 56, 61, 62, 63, 64, 68, 71, 72, 78,
perturbations, 67, 68, 111, 190, 191, 216, 217 79, 84, 92, 93, 102, 107, 114, 123, 125, 165, 166,
photosynthesis, 168, 172, 178, 181 183, 184, 185, 186, 196, 198, 200, 201, 204, 205,
photosynthetic system, 171 206, 207, 208, 214, 217, 230, 231, 232
plant photosynthesis, 168 quorum sensing mechanism, 107, 114
Plasmid Construction, 55, 74 quorum sensing symbiotic ecosystem, 184, 196, 198
population control gene circuits, 179 quorum sensing system, 3, 62, 64, 123, 125, 184,
population control genetic circuits in PCC 7942, 179 185, 186, 196, 198
242 Index
quorum sensing-based amplifier, 71 Self-regulating mechanisms, 105

quorum sensing-based biosensor, 61, 72 sense process, 206
quorum-sensing mechanism, 205, 206, 207 sensing (QS)-based amplifier, 61
quorum-sensing system, 200, 204, 205, 206, 208, sensitivities, 121, 124, 127, 133, 148, 222, 228
214, 217 sensitivity S of MMCS, 207
sigmoid-like mathematical model, 134
simulation and experimental population density
R ratios, 194, 196
small molecules C4-HSL and 3OC12-HSL, 186
reference model, 86, 89, 91, 92, 93
specified I/O filtering response, 7, 9, 17, 18, 26
regulation activity, 65, 75, 76, 109, 169, 187, 188
specified I/O response, 88, 89
regulatory system CusSR, 204
specified range of metal ion concentrations, 68, 218
reporter gene, 1, 11, 12, 13, 26, 34, 39, 66, 169, 170,
specified reference I/O response, 91, 96
208
spore-forming bacterium Bacillus subtilis, 105
reporter protein, 2, 11, 12, 13, 14, 15, 16, 18, 19, 21,
steady state cell population, 110, 111, 112, 173, 175,
23, 26, 27, 28, 29, 30, 40, 42, 52, 62, 65, 66, 67,
176, 189
70, 74, 76, 85, 93, 179, 185, 208, 213, 216, 221
steady state model of the sender cell, 215
repressilator, 1
steady state models of the synthetic genetic circuit,
repressor, 10, 11, 12, 13, 14, 15, 16, 19, 21, 26, 30,
173
31, 32, 37, 38, 40, 41, 42, 44, 45, 46, 51, 57, 107,
steady state models of the synthetic symbiotic
108, 109, 110, 111, 116, 117, 128
ecosystem, 189
repressor-regulated promoter-RBS library, 11, 26,
steady-state model, 14, 15, 16, 17, 66, 67, 76, 109,
30, 31, 32, 37, 38, 44, 45, 46, 51, 117
169, 173, 189, 191
repressor-regulated promoter-RBS library LibLacI, 31
stop-then-trace strategy, 123, 148
repressor-regulated promoter-RBS library LibTetR, 31
strength of the constitutive promoter-RBS pair, 139,
reprogramming bacteria, 122, 166
140
reprogramming of chemoreceptor proteins, 122
Streptomyces venezuelae, 197, 199, 200
response process, 206
stronger constitutive promoter-RBS components,
ribosomal binding sites (RBS), v, vi, ix, 1, 2, 3, 4, 5,
176
7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 21,
swarm assays, 126, 128, 129, 130, 148, 161, 163
22, 23, 24, 26, 27, 28, 29, 30, 31, 32, 33, 37, 38,
switch design example of synthetic genetic
39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51,
transistor, 49
52, 55, 61, 62, 63, 64, 65, 66, 68, 69, 71, 72, 74,
switchable symbiosis circuits, 185
75, 76, 77, 83, 84,85, 86, 91, 92, 95, 96, 105, 106,
switchboard, 24, 25, 26, 33, 56
107, 108, 109, 110, 111, 112, 113, 116, 123, 124,
switching I/O characteristics, 49
125, 127, 128, 133, 135, 138, 139, 140, 141, 142,
symbiosis, 80, 183, 184, 185, 191, 232
143, 144, 145, 147, 148, 165, 167, 168, 169, 170,
symbiotic ecosystem, 183, 184, 185, 186, 189, 191,
172, 173, 174, 175, 176, 178, 179, 185, 186, 187,
192, 193, 194, 196, 197, 198
188, 189, 191, 192, 193, 194, 195, 196, 198, 204,
symbiotic microbial ecosystems, 184, 198
205, 207, 208, 209, 210, 211, 212, 213, 214, 215,
synchronicity, 197
216, 217, 218, 219, 220, 221, 222, 223, 224, 225,
Synechococcus, vi, 167, 168, 171, 172, 174, 175,
226, 227, 228, 231
177, 180, 181, 182
robust biological filter, 2, 7, 8
synthase LuxI, 64
robust multicellular synthetic gene networks, 205
synthetic antibiotic resistance circuits, 185, 198
robust synthetic genetic circuit, 167
synthetic biological filter, 2, 7, 8, 9, 10, 11, 12, 13,
robust synthetic symbiotic ecosystem, 183
16, 23, 26
synthetic biology, ix, 1, 2, 3, 4, 5, 6, 7, 8, 24, 33, 34,
S 35, 37, 38, 39, 50, 56, 57, 59, 62, 63, 77, 78, 80,
81, 82, 83, 84, 101, 106, 117, 121, 122, 123, 124,
S-adenosylmethionine (SAM), 63, 64, 65, 78, 80, 128, 149, 165, 167, 179, 180, 183, 198, 199, 203,
102, 103, 207, 213, 231, 232 204, 228, 230, 231, 232, 235
scalability, 1, 7, 24, 26 synthetic biology method, 121, 123, 124, 128, 149,
search-then-locate strategy, 124, 148 180
Index 243
synthetic gene circuit of a MMCS, 227 the repressor-regulated promoter-RBS component,

synthetic gene network design, 6, 35, 36, 59, 201, 13, 14, 30, 41, 45, 108
228, 233 the stop-then-trace strategy, 148
synthetic genetic circuit, ix, 39, 40, 62, 67, 111, 118, toggle switch, 1, 4, 37, 57, 204, 231
125, 167, 168, 169, 170, 173, 174, 175, 179, 185, transcriptional activator LasR protein, 188
191, 208, 209, 214, 217, 218, 222, 227, 228 transduction ability, 203, 207, 210, 218, 219, 221,
synthetic genetic circuit in cyanobacteria, 179 222, 224, 225, 226, 227
synthetic genetic lysis circuit, 105, 107, 114, 115, transduction capabilities, 207, 228
116 tunable biological filter, 7, 8, 9, 18, 23, 25, 26
synthetic genetic transistor circuit, 41, 42, 43, 49, 51 tunable biological switchboard, 25, 26
synthetic genetic transistors, 2, 37, 50 two cell populations (ER and EG), 185
synthetic metal ion biosensor, 89, 96 two quorum sensing systems, 184, 186, 196, 197,
synthetic symbiotic ecosystem, 183, 184, 185, 186, 198
189, 191, 192, 193, 196, 198 two repressor-regulated promoter-RBS libraries, 31
system I/O filtering response, 8
system sensitivity, 203
systematic biological filter design scheme, 7 U
systematic design method, ix, 2, 5, 7, 37, 63, 83, 105,
user-defined specifications, 220, 227
106, 149, 165, 178, 179, 199, 203, 227, 231
user-oriented design method, 117
systematic design of a MMCS, 228
user-oriented specifications, 3, 61, 62, 63, 83, 95,
systematic design of synthetic gene circuits in
106, 167, 168, 169, 184, 191
cyanobacteria, 180
systematic genetic circuit design, 2, 179, 207
systematic genetic circuits in cyanobacteria, 179 V
systematic synthetic biology-based method, 123
V. fischeri quorum-sensing system, 208, 214, 217
variances of cell population densities, 176
T
verification of the brake components by swarm
assays, 129
Takagi-Sugeno (T-S) fuzzy model, 81, 82, 89, 90,
verifying the performance of the synthetic circuit,
91, 93, 95, 103
126
the desired I/O filtering response, 8, 9, 16, 17
Vibrio fischeri, 62, 78, 102, 123, 205, 231
the npn bipolar junction transistor, 53

Systems Synthetic Biology System Models, User-Oriented Specifica

Загружено:

Сведения о документе

Авторское право

Доступные форматы

Поделиться этим документом

Поделиться или встроить документ

Параметры публикации

Этот документ был вам полезен?

Это неприемлемый материал?

Авторское право:

Доступные форматы

Systems Synthetic Biology System Models, User-Oriented Specifica

Загружено:

Авторское право:

Доступные форматы

SYSTEMS BIOLOGY - THEORY, TECHNIQUES AND APPLICATIONS

SYSTEMS SYNTHETIC BIOLOGY

Additional books in this series can be found on Nova’s website

Additional e-books in this series can be found on Nova’s website

SYSTEMS SYNTHETIC BIOLOGY

NOTICE TO THE READER

Library of Congress Cataloging-in-Publication Data

Published by Nova Science Publishers, Inc. † New York

9.3. Results 195

Synthetic biology is a new engineering discipline, which employs the concepts of

a significant challenge. Towards addressing this problem, this chapter develops a

SYSTEMATIC BIOLOGICAL FILTER DESIGN

2.2. CONSTRUCTION OF PROMOTER-RBS LIBRARY

2.3. SYSTEMATIC DESIGN METHODOLOGY OF THE SYNTHETIC

A Cascade Gene Circuit Topology for Biological Filter Design

xr1 xr2 reporter G

low-pass threshold I1 high-pass threshold I2

Dynamic Model of the Biological Filter Based on Genetic Circuit Topology

 x1 (t )  P( S1u , S1l ,0,0)  (    x1 ) x1 (t )  1 (t )

 xr1 (t )  Pc ( S1u ,i ,0,0,0)  (    xr 1 ) xr1 (t )  1 (t )

Pc ( S1u ,i ,0,0,0)  S1u ,i

Formulation of Design Specifications for Biological Filters

Biological Filter Design by the Reference Matching via a Mixed Library-

J ( fi )  E  (Ge ( fi , Ii )  Gri ( Ii ))2 dIi (2.6)

2.4. BIOLOGICAL FILTER DESIGN BASED ON PROMOTER-RBS

Design Example of Biological Low-Pass Filter

J ( f 1* )  min E  (Ge ( f1 , I1 )  Gr1 ( I1 ))2 dI1 (2.9)

Design Example of Biological High-Pass Filter

J ( f 2* )  min E  (Ge ( f 2 , I 2 )  Gr 2 ( I 2 ))2 dI 2 (2.11)

integrating several designed biological filters. Because the internal adjustability by

the reporter protein to light excitation is dependent on post-translational modifications,

x(t )  P(Su , Sl , TF , I )  (   x ) x(t ) (2.12)

x(t )  P(Su , Sl , TF , I )  (m     i ) x(t )

Construction of the Constitutive Promoter-RBS Library

Before identifying the promoter-RBS strength by measuring fluorescence, two important

Table 2.1. The constitutive promoter-RBS library Libconst

Index BioBrick component Su

Construction of the Repressor-Regulated Promoter-RBS Library

The repressor-regulated promoter-RBS component produces the maximum and minimum

Table 2.2. The repressor-regulated promoter-RBS library LibTetR

Index BioBrick component Su Sl Parameters

Table 2.3. The repressor-regulated promoter-RBS library LibLacI

Index BioBrick component Su Sl Parameters

Construction of the Activator-Regulated Promoter-RBS Library

Table 2.4. The activator-regulated promoter-RBS library

Index BioBrick component Su Sl Parameters

SYSTEMATIC DESIGN METHODOLOGY

3.2. CONSTRUCTION OF THE PROMOTER-RBS LIBRARIES

Here, we introduce the characterization and standardization of promoter-RBS libraries

Promoter-RBS Libraries Based on the Identified Kinetic Strengths

In a systematic design procedure, the characterization and standardization of components

3.3. CONSTRUCTION AND DESIGN OF GENETIC TRANSISTOR

transistor, with prescribed I/O characteristics of amplification or switching through the

Input device Ouput device

Input signal measure device

Construction of Genetic Transistor

yss (c3 , I 2 , g1 (c2 , I1 ))  g2 (c3 , I1 , I 2 ) , g1   g1e , g1n  (a.u.) (3.4)

 im, x   im, x   im, x n3  t  ,  m, x   m, x   m, x n3  t  , m1  m1  m1n2  t  ,

Systematic Design of a Genetic Transistor Based on Design Specification

continually. For convenience of design, the repressor-regulated promoter-RBS component c2

Input operation range: I1 [I1,l , I1,u ]  g1 [g1,l , g1,u ] (a.u.) (3.6)

yd  g1   gain  ( g1  g1,l )  g2,u , g1   g1l , g1u  (a.u.) (3.7)

1. Construct the genetic transistor circuit such as in Figure 3.1(a).

3.4. SYNTHETIC GENETIC TRANSISTOR DESIGN EXAMPLES

Input range: I1 [1.5102 ,4102 ] (mM)  g1 [298,431] (a.u.) (3.10)