Biosensors and Bioelectronics 22 (2007) 1756–1763

A recombinant Fab fragment-based electrochemical immunosensor

for the determination of testosterone in bovine urine
Huihui Lu a , Mark P. Kreuzer b,∗ , Kristiina Takkinen c , George G. Guilbault a
a Sensor Development Group, Analytical & Biological Chemistry Research Facility, UCC, Cork, Ireland
bInstitut de Ciències Fotòniques (ICFO), Mediterranean Technology Park, 08860 Castelldefels, Spain
c VTT Biotechnology, P.O. Box 1500, FIN-02044 VTT, Espoo, Finland

Received 30 March 2006; received in revised form 4 August 2006; accepted 9 August 2006
Available online 7 September 2006

Abstract
This work describes the development of an electrochemical, recombinant Fab fragment-based immunosensor for the detection of testosterone
in bovine urine. The sensor comprised of a testosterone conjugate on the surface of screen-printed electrodes, and recognition followed by an
anti-testosterone Fab fragment. The use of an IgG-horseradish peroxidase conjugate determined the degree of competition. Chronoamperometry at a
potential of +100 mV, was chosen to reductively measure the product of the catalysis of 3,3 ,5,5 -tetramethylbenzidine catalysis. ELISA was primarily
used to investigate the assay system, prior to transferring to SPEs. The final Fab-based sensor exhibited the linear range of 300–40,000 pg/ml with
limit of detection of 90 ± 13 pg/ml. Furthermore, the developed Fab sensor allowed for the determination of testosterone in bovine urine directly
after dilution, omitting the necessity of extraction and hydrolysis. Comparison of administrated bovine urine samples between the developed Fab
sensor and GC–MS data showed quantitative or semi-quantitative results and enabled identification of suspicious samples for further extensive
analysis by established analytical techniques. With simple sample preparation, low limit of detection, and good repeatability, the proposed method
can offer alternative advantages as a primary screening tool for meat quality control.
© 2006 Elsevier B.V. All rights reserved.

Keywords: Testosterone; Screen-printed electrode; Recombinant Fab; Immunoassay; Bovine urine

1. Introduction
Testosterone is an anabolic steroid often used for illegal
growth promotion in livestock farming and thereby causing
potential health risks to consumers. The use of steroids for
growth promoting purposes in animals is banned in the Euro-
pean Union for food safety reasons (Directive 88/146/EEC). The
determination of residues is required in animals and in fresh meat
as given by the EU Directive (86/849/EEC).
There are various analytical methods used for the detection of
testosterone in biological fluids including, HPLC (Li et al., 2002;
Navajas et al., 1995), GC–MS (Samuels et al., 1998; Saudan et
al., 2004) and LC–MS offers a simplified, specific and sensitive
alternative to GC–MS (Leinonen et al., 2004). Whilst these tech-
niques have been used as quantitative and confirmatory methods,
they do need expensive instrumentation, specialized personnel
∗ Corresponding author. Tel.: +34 93 553 40 32; fax: +34 93 553 40 00.
E-mail address: mark.kreuzer@icfo.es (M.P. Kreuzer).
and suffer from considerable time delays between sampling and
obtaining results. These disadvantages limit their routine use and
restrict assaying to extensive laboratory environments.
Immunological techniques are based on the molecular recog-
nition of antigens by antibodies to form a stable complex which
are capable of detecting low level concentrations of analytes in
biological matrices. Various assays have been developed to mea-
sure testosterone in plasma (Rajkowski et al., 1989), serum (Dhar
and Ali, 1992), and human urine (Al-Dujaili, 2006). The cumber-
some handling procedures involved, hinder their use as on-site
detection systems. By comparison, immunosensors combining
sensitivity of the antibody–antigen interaction with other advan-
tages, including cost-efficient, speed of analysis and portable
screening, allows for fast, decentralised measurement away from
the lab environment. They are widely applied in the area of
human and animal health and in the food sector (D’Orazio, 2003;
Mello and Kubota, 2002). Electrochemical-based screen-printed
electrodes (SPE) provide an approach to develop cost-effective
devices through disposable means and have been reported to
measure steroids, like progesterone in cow’s milk (Pemberton

0956-5663/$ – see front matter © 2006 Elsevier B.V. All rights reserved.
doi:10.1016/j.bios.2006.08.002
 H. Lu et al. / Biosensors and Bioelectronics 22 (2007) 1756–1763
et al., 1998; Kreuzer et al., 2004) and estradiol in human serum
(Pemberton et al., 2005).
Current immunoassays capable of detecting steroid hor-
mones are still performed using polyclonal and monoclonal
antibodies. It has been suggested that the main limitation of
immunosensor technology is related to the antibodies and their
properties (Hock, 1997). Steroid hormones are small, rigid,
hydrophobic molecules with only a few functional groups capa-
ble of specific interaction with antibodies. Furthermore, they
are poorly immunogenic in mice and rats; two species from
which monoclonal antibodies are usually generated, making it
difficult and expensive to produce. The supply of good poly-
clonal antibodies of uniform quality is not easy and requires
continuous immunization of many laboratory animals (Killard
et al., 1995). Antibody engineering provides excellent tools to
tailor the properties of antibodies in respect to affinity, speci-
ficity and performance for different applications. The possibility
to express soluble, active antibody fragments, Fab and scFv,
in E. coli (Better et al., 1988; Bird et al., 1988; Huston et al.,
1988) allows cost-effective production, feasible engineering of
the binding properties and fusions with peptide tags enhanc-
ing purification, detection and immobilisation. Furthermore, the
antibody phage display library technology has opened com-
pletely new possibilities to produce novel binding specificities
avoiding the limitations inherent in the mammalian immune
response, as reviewed in Hoogenboom (2005). A number of
recent publications describe the development of recombinant
antibody fragments to steroid analytics by antibody engineer-
ing techniques (Hemminki et al., 1998; Valjakka et al., 2002).
To our knowledge, there is little information of the application
of a recombinant Fab fragment to detection of testosterone in
electrochemical immunosensor area.
2. Materials and methods
2.1. Chemicals and reagents
Testosterone, 19-nortestosterone, boldenone, methyltestos-
terone, methylboldenone, eticholanolone, epieticholanolone,
epitestosterone, 4-androsten-3␤,17␤-diol, 4-androsten-3␣,17␤-
diol, 2␣-, 6␤-, 16␣-, and 16␤-hydroxytestosterone were
purchased from Steraloids Inc. (RI, USA). Bovine serum
albumin (BSA), testosterone 3-(o-carboxymethyl)oxime: BSA
conjugate, anti-mouse IgG peroxide conjugate, 3,3 ,5,5 -
tetramethylbenzidine dihydrochloride (TMB) were purchased
from Sigma–Aldrich (Dublin, Ireland). The inks used in screen-
printing (Electrodag® B-0851, 423-SS and 451-SS) were pur-
chase from Acheson (Plymouth, UK). All other solvents and
reagents were of analytical grade and were purchased from
Sigma–Aldrich (Dublin, Ireland).
The anti-testosterone Fab fragment (clone 220) was iso-
lated from a phage display library constructed from mice
immunized with testosterone 3-(o-carboxymethyl)oxime: BSA
(Sigma–Aldrich). The phage library construction, selection and
clone analysis was essentially done as described in Pulli et al.
(2005). The Fab clone 220 was isolated by selecting the library
for four rounds with the immunogen, testosterone-CMO-BSA,
1757
immobilised on microtiter plate wells. For large-scale, soluble
production of Fab 220, the Fab expression unit with six histidine
residues at the C-terminus of the heavy chain was cloned into
the pKKtac expression vector (Takkinen et al., 1991). The Fab
220 was produced in E. coli strain RV308 (ATCC 31608) by
high cell density fermentation and purified by immunoaffinity
and protein G chromatography as described in Nevanen et al.
(2001).
2.2. Apparatus
Ninety-six-well flat-bottomed polystyrene microtiter plates
were purchased from NUNC (Dublin, Ireland). Bio-Tek Instru-
ments (VT, USA) supplied the microplate reader (model EL
311). Incubations at elevated temperatures were carried out in
a thermostat oven supplied by Heraeus Instruments (Hanau,
Germany). Electrodes were prepared using a DEK (model
247) screen-printer (Dorset, UK). Chronoamperometric mea-
surements were performed using a potentiostat (␮AUTOLAB)
with GPES software (ECO-CHEMIE, Netherlands).
2.3. Urine sample preparation
Urine samples of heifers injected intramuscularly with testos-
terone (200 mg per animal) were provided by Dr. M. Crowe
(Faculty of Veterinary Medicine, National University of Ire-
land, Dublin). The study was performed in compliance with
protocols approved by the Ethics Committee, University Col-
lege Dublin, the Cruelty to Animals Act (Ireland, 1876), and
the European Union Directive 86/609/EC. An aliquot (1 ml) of
the urine specimen of interest was defrosted and centrifuged at
3000 × g for 3 min. The supernatant was collected, aliquoted
(20 ␮l) and frozen at −20 ◦ C until assayed to avoid repeated
freeze–thaw cycles of samples.
2.4. ELISA protocol
Ninety-six-well ELISA plates were coated with 50 ␮l of the
appropriate dilution of testosterone–BSA conjugate in phos-
phate buffered saline (PBS, pH 7.4) at 37 ◦ C for 45 min. After
washing three times with washing buffer (PBS containing 0.05%
(v/v) Tween 20), 200 ␮l per well of blocking buffer (PBS con-
taining 1% (w/v) BSA) was added and plates were then incu-
bated at 37 ◦ C for 45 min or at 4 ◦ C overnight followed by wash-
ing as described above. Thereafter, an optimised, one-step incu-
bation protocol was used for the remaining immunoreagents. A
volume of 25 ␮l of each different concentration (0–1000 ng/ml)
of testosterone standards, 25 ␮l of the suitable dilution of the Fab
fragment and 25 ␮l enzyme labelled anti-species (␣-IgG–HRP)
were all added at the same time into the wells and incubated at
37 ◦ C for 15 min. After a further washing step, 100 ␮l per well
of substrate solution containing 0.1 mg/ml TMB and 0.2% (v/v)
hydrogen peroxide in 0.05 M phosphate citrate buffer (pH 5,
containing 0.1 M potassium chloride) were added and the wells,
left to develop colour for 30 min at room temperature and subse-
quently stopped by the addition of 50 ␮l of 2 M sulphuric acid.
Absorbance was read at 450 nm. Standard-dose response curves
1758 H. Lu et al. / Biosensors and Bioelectronics 22 (2007) 1756–1763
were constructed by plotting the absorbance as a function of
testosterone concentration.
2.5. Electrochemical immunosensor protocol
The electrode design and layer development were prepared
in advance by the systematic and controlled deposition of the
varying layers (conducting, isolating and working) onto a PVC
sheet. After printing the last path, the electrodes were cured at
90 ◦ C for 1 h. The electrodes consisted of a silver conducting
ink coated by an insulating layer and a carbon working area of
0.16 cm2 . Immunoassays were performed on this working area.
The SPE were cut from the PVC sheet into 45 mm × 15 mm
long strips. The electrodes were modified as follows: a 5 ␮l
drop of appropriately diluted coating testosterone–BSA conju-
gate in PBS buffer was spread onto the surface of the working
electrode and incubated for 30 min at 37 ◦ C. The electrodes
were then washed with washing buffer followed by distilled
water to remove unbound coating conjugate and excess water
was allowed to evaporate at room temperature. The electrode
surface was then blocked by the addition of 100 ␮l blocking
buffer and incubated for 45 min at 37 ◦ C. For the competition
step, the electrodes were immersed in 10 ␮l of a 1:1 mixture
of analyte and suitable dilution of Fab fragment. For urine
analysis, a frozen aliquot was defrosted and diluted in PBS
(1/10, v/v). Dilutions of testosterone were then made with this
modified urine stock. After incubation and a washing step,
the electrodes were covered with 10 ␮l IgG–HRP conjugate
and incubation for 30 min at 37 ◦ C followed. The electrodes
were washed for a final time with distilled water and mea-
sured using chronoamperometry at +100 mV by adding TMB
substrate solution (0.24 mM TMB in phosphate citrate buffer,
pH 5, with 1.32 mM H2 O2 ) over all three electrodes (work-
ing, reference and auxiliary electrodes) in a non-hydrodynamic
manner.
All electrochemical experiments were performed at room
temperature. The parameters for chronoamperometry detec-
tion were as follows—pre-treatment: first conditioning potential
200 mV, duration 20 s, equilibrium time 5 s; measurement: inter-
val time 0.1 s, standby potential 200 mV, number of potential
steps 1; potential 100 mV; duration 10 s. All data points were
carried out in duplicate and each value was the mean value of
two readings. The measured current was recorded and plotted
against testosterone concentration to give a calibration plot. Cali-
bration curves were freshly prepared prior to each assay by using
10 serial dilutions from standard solutions with concentrations
in the range of 0–1000 ng/ml.
3. Results and discussion
3.1. ELISA development
The objective of this research was to develop a disposable
electrochemical immunosensor using a recombinant Fab frag-
ment for the determination of testosterone in bovine urine. The
first step in attaining such a system was the development of an
assay capable of incorporation into an immunosensor format.
ELISA was primarily performed in 96-well plates to investigate
assay performance due to its simple and quick determination of
protein characteristics.
3.1.1. Optimisation of ELISA parameters
Optimum working concentrations of both coating conju-
gate and anti-testosterone Fab fragment were obtained by
performing chessboard titrations, whereby serial dilutions of
both variables were carried out in duplicate (data not shown).
Optimum concentrations of coating testosterone–BSA conju-
gate was found to be 2.5 ␮g/ml with recombinant Fab frag-
ment around 0.31 ␮g/ml corresponding to 80% of the maxi-
mum signal. Subsequent ELISA assays were carried out under
these conditions incorporating a 1/600 dilution of IgG–HRP.
This was found to catalyse the turnover of substrate to prod-
uct efficiently and within a short time suitable for ELISA
measurements.
3.1.2. Competitive immunoassay using ELISA
The competition study was carried out setting up two proto-
cols: in the first protocol (one step), analyte, Fab and IgG–HRP
were mixed together directly in the microtiter well where the
coating conjugate had been immobilised. In the second one (two
steps), analyte and Fab fragment were added into the microtiter
wells followed by the addition of IgG–HRP as a subsequent step.
The competitive curves showed there was no difference between
two different protocols. Thus, the first protocol was used in
subsequent assays because shorter assay time could be used.
Different incubation times (15, 30, 45, 60 min) of the competi-
tion one-step were studied and the optimum time was determined
as 15 min at 37 ◦ C.
A competitive assay was established between immobilised
testosterone–BSA conjugate and analyte for Fab fragment.
Blank wells and controls were incorporated for measurement
of maximum signal (zero analyte) and for non-specific binding
(zero Fab). Fig. 2a shows averaged standard curves by ELISA.
Data is presented as the average of nine separate assays while
within each assay, each data point was the average of triplicate
measurements (n = 3, N = 9). Calibration curve statistics are pre-
sented in Table 1. Repeated analysis confirmed these findings.
At this point it was evident that the developed recombinant Fab
could be used as a recognition element, with suitable sensitivity
for the development of the electrochemical sensor and no further
effort was placed on the ELISA system. Complete optimisation
and analysis would be performed using the screen-printed elec-
trode set-up.
Table 1
Best-fit parameters for repeated analysis for testosterone using screen-printed
electrodes and ELISA
Parameters ELISA (n = 3, N = 9) SPE (n = 2, N = 3)
Hillslope −1.1 −1.0
EC50 (ng/ml) 2.8 ± 0.3 2.7 ± 0.5
LOD (pg/ml) 540 ± 193 90 ± 13
R2 0.989 ± 0.007 0.970 ± 0.018
R.S.D.a (%) 5 10
a
Average R.S.D. across all data points of the curve.
 H. Lu et al. / Biosensors and Bioelectronics 22 (2007) 1756–1763
3.2. Fab-based immunosensor using screen-printed
electrodes and chronoamperometric detection
In biological fluids, ascorbic acid, uric acid and xanthine
are common electrochemical interferences. In a previous study
(Jamal et al., 2005), ascorbic acid was shown to be the major
electrochemical interference using p-aminophenol as a model
system. Chronoamperometry was the most sensitive method
among a range of electrochemical techniques tested (differ-
ential pulse voltammetry, square wave voltammetry, linear
sweep voltammetry, chronoamperometry) in the detection of
p-aminophenol on a glassy carbon electrode. Therefore, the
electrochemical technique chosen in this study was chronoam-
perometry performed at a potential of +100 mV where the TMB
product undergoes reduction. The SPEs are produced in batches
of 30. Within a batch the typical error recorded for measuring
standard redox reactions was about 1% or less. Once the SPEs
have bioreagents immobilised, error accounts for ca. 8 ± 2%
depending on both the handling/manipulation (operator error)
and also the sample matrix. This value is quoted for buffered
systems. When more complex matrices are used, error is gener-
ally higher with errors between 10 and 15% noted.
3.2.1. Optimisation of electrochemical parameters
Due to the differences in surface area of the screen-printed
electrode (0.16 cm2 ) compared to the conventional microtiter
plate (0.28 cm2 ) and the binding capacity of both surfaces,
optimisation of reagents was re-evaluated. As for the ELISA
study, assay parameters such as the amount of proteins (coat-
ing testosterone–BSA conjugate and Fab fragment) and enzyme
labelled anti-species, incubation time of each step were re-
evaluated and optimised again.
Testosterone–BSA is passively adsorbed to the surface
of the SPE. From Fig. 1a, the currents noted when using
testosterone–BSA coating concentrations at 5, 10, 20 or
40 ␮g/ml were obviously higher than 0 ␮g/ml testosterone–BSA
conjugate coating (non-specific binding control). In the latter
case, the SPEs were subsequently blocked in PBS buffer (1%
BSA), Fab fragment added at various concentrations and the sig-
nals determined that an average of ca. 300 nA was attributed to
background and/or non-specific signal. Fig. 1a and b represents
Fig. 1. (a) Optimisation of coating testosterone–BSA conjugate and Fab concentrations in the presence IgG–HRP (1/500) on SPEs (n = 2). Coating concentration
(␮g/ml): , 40; , 20; , 10; ×, 5; *, 0. (b) Competitive immunoassay on SPEs using different concentrations of Fab (␮g/ml): , 4; , 1; 䊉, 0.5; , 0.25 (n = 2).
Fixed amount of coating testosterone–BSA conjugate (20 ␮g/ml) and IgG–HRP (1/500 dilution) were used. Chronoamperometric detection at +100 mV.
1759
the results obtained with the different concentrations of proteins
(coating and Fab fragment). The effect of increasing the coating
testosterone–BSA conjugate and Fab fragment (Fig. 1a) was
to increase the overall signal. However, higher concentrations
did not yield significantly larger signals and above 20 ␮g/ml of
testosterone–BSA, no increase in current was observed. Gener-
ally speaking, as a starting point, we tend to use a combination of
reagents that attains a current of ca. 1 ␮A. This yields a sufficient
sensitivity and spread of data points from a maximum signal of
1 ␮A to a background typically found at 10–15% or 150 nA. If
sensitivity is subsequently a problem, dilutions of reagents can
be adjusted to alter the precision, sensitivity and quantitative
range of the assay (Ekin, 1993). Theory indicates that decreas-
ing concentrations of antibody will afford greater sensitivity as it
will require less analyte to reach a constant reference level, such
as the EC50 . This approach, however, is limited by the precision
achieved with increasing dilution (Ruth, 2001). This theory was
applied to Fab fragment and from Fig. 1b it can be seen that lower
Fab fragment concentrations shift the standard curves to the
left, decreasing the limit of detection and the EC50 favourably.
Cost may also be a factor in favour of using lower concentra-
tions of proteins. But also this approach decreased the absolute
signal and thus the Hillslope, an important parameter when
determining assay sensitivity and detection. Therefore optimum
compromised concentration of coating testosterone–BSA con-
jugate was found to be 20 ␮g/ml with recombinant Fab fragment
of 0.5 ␮g/ml.
Different incubation time (15, 30, 45, 60 min) of coating con-
jugate, Fab fragment and enzyme labelled anti-species on SPE
were also studied and optimum incubation time for each step
was found to be 30, 45 and 30 min at 37 ◦ C, respectively (data
not shown).
3.2.2. Competitive immunoassay on screen-printed
electrodes
A competition study using one-step and two-step incubation
protocols was also investigated for the electrochemical system
and the results showed that competitive one-step incubation did
not give as high signal as two-step incubation as in ELISA case
previously determined. This difference presumably reflects the
relatively high protein-binding capacity of commercially treated
1760 H. Lu et al. / Biosensors and Bioelectronics 22 (2007) 1756–1763
Fig. 2. Competitive assays for testosterone detection using Fab fragment as the molecular recognition element (a) ELISA data (n = 3, N = 9) and (b) chronoamperometric
detection at +100 mV using SPEs (n = 2, N = 3). Maximum value represents zero analyte (testosterone).
microtiter plates compared to the carbon working surface of the
SPEs. Thus, two incubation steps were used for the electrochem-
ical system.
Fig. 2b shows competitive immunoassay curve for testos-
terone detection on SPEs using the optimised parameters found
previously. It can be seen that on decreasing the analyte con-
centration, the signal approaches the maximum control for the
competitive immunoassay. For this assay, the linear range can
be observed between 300 and 40,000 pg/ml. The list of param-
eters for repeated analysis for testosterone on SPEs, which
were also obtained in the ELISA studies are summarized in
Table 1. It was found that the EC50 parameter was very close to
that of the ELISA system. Regression values (R2 ) of the elec-
trochemical response curve showed that accuracy was good,
while error (average R.S.D. across all points of curve) was
still acceptable (10%). This can be attributed to the exces-
sive handling of the electrodes and the inability at present to
measure large numbers of electrodes simultaneously. These val-
ues are thought acceptable and showed the robustness of the
assays.
Non-specific binding of the sensor was determined in a num-
ber of ways. In Section 3.2.1, different Fab fragment concentra-
tions (0, 0.032, 0.16, 0.8, 4 and 20 ␮g/ml) at different coating
concentrations was compared (Fig. 1a). The current noted when
the Fab fragment concentration was zero was proportional to the
non-specific binding of the secondary label (IgG–HRP; 1:500
dilution). When the optimum protein conditions were subse-
quently used to give the competitive assay (Fig. 2b), the non-
specific binding or background was noted in the plateau at high
testosterone levels (>100 ng/ml). This value was recorded as ca.
150 nA.
Table 2
Immunoassays for testosterone detection (from literature) with associated limits of detection
Antibody used
Rabbit antiserum against testosterone 3CMO-BSA
Rabbit antiserum against testosterone 3CMO-BSA
Rabbit antiserum against testosterone 3CMO-BSA
Rabbit antiserum against testosterone 3CMO-BSA
Rabbit antiserum against testosterone 3CMO-BSA
␣-Testosterone antibody raised in sheep (Micropharm, London, UK)
In order to measure the non-specific binding of the secondary
label (at various dilutions) to testosterone–BSA coating conju-
gate (20 ␮g/ml) we noted a specific current of 900 nA and a
non-specific one of 143 nA (specific:non-specific ratio of 6.3)
for a 1:500 dilution. Background increased significantly when
lower dilutions of secondary label were used (710 nA) while
decreasing the specific:non-specific ratio to 4.7.
The sensitivity of the Fab sensor was determined by both
EC50 and LOD parameters. To calculate LOD, the following
parameters of EC50 , Hillslope and three times of standard devi-
ation of the maximum signal were used and values returned
were based on the average of three separate assays in triplicate
measurements. The Hillslope, which gives a measure of the sen-
sitivity of the assay, is measured by a tangent through the point
of inflection of the sigmoid (EC50 ). Values of Hillslope close
to 1 are considered optimum because of the inverse proportion-
ality of the signal to concentration. Determination of limit of
detection also uses this parameter and thus is an important char-
acteristic of any curve as a deviation from a value of 1 attributes
to a less sensitive assay based on the following equation:
 −1/k
Max − Min
LOD = EC50 −1
Max − Min − 3S.D.max
where EC50 is an effective concentration for 50% inhibition;
Max and Min are maximum and minimum signal; S.D.max is
standard deviation of maximum signal; k is Hillslope. The LOD
was found to be 90 ± 13 pg/ml based on this equation. Over
the past decade, numerous publications concerning testosterone
immunoassay measurement using various whole antibodies as
the molecular recognition element have been reported (Table 2).
Medium LOD (pg/ml) Reference
Human urine NA Hanquez et al. (1987)
Plasma 60–2000 Rajkowski et al. (1989)
Serum 100 Dhar and Ali (1992)
Culture medium 11.43 Illera et al. (1997)
Human serum 15 Shrivastav et al. (2003)
Buffer 12.2 Al-Dujaili (2006)
 H. Lu et al. / Biosensors and Bioelectronics 22 (2007) 1756–1763
Our results were comparable to values reported in the literature
and deemed to be ultra-sensitive.
3.3. Cross-reactivity study
In this work, the study was undertaken by adding various
cross-reactants instead of using androgen testosterone in an indi-
rect competitive format. The testosterone was done in parallel
which served as the standard. The cross-reactant response curve
is the curve parallel to that of the standard (testosterone). The
cross-reactivity was calculated based upon the EC50 values of
individual steroids.
In general, steroid metabolites are excreted in higher
concentrations and for prolonged periods compared to the
parent compound. 4-Androsten-3,17-diol, etiocholanolone
and epietiocholanolone are considered to be main metabolites
of testosterone in bovine urine following administration
(Samuels et al., 1998). Other possible major metabolites
are 2␣-hydroxytestosterone, 6␤-hydroxytestosterone, 16␣-
hydroxytestosterone, 16␤-hydroxytestosterone and estradiol
(Venturelli et al., 1995; Li et al., 2002; Sanwald et al., 1995).
Among the steroid compounds tested here, Fab fragment 220
recognized 2␣-hydroxytestosterone (74.9%), ␤-boldenone
(63.7%), 4-androsten-3␤,17␤-diol (36.4%) and 4-androsten-
3␣,17␤-diol (32.4%) very well, but 6␤-hydroxytestosterone
(11.2%) and epitestosterone (7.2%) slightly. Low cross-
reactivity with ␤-estradiol, epietiocholanolone, etiocholano-
lone, 16␣-hydroxytestosterone and 16␤-hydroxytestosterone
was observed (all <0.3%).
Cross-reactivity is a phenomenon inherent to all immunoas-
say. Cross-reactivity with other steroids can compromise the
specificity of immunoassay methods, but on the other hand gives
them an advantage as primary screening tool in steroid analysis
due to the ability of antibodies (or Fab fragment) to bind broad
classes of metabolites that contain a common substructure. As
has been shown, our methods offer the advantage to pick up
both testosterone and its major metabolites, such as 17-alkyl-
substituted anabolic steroids due to the high cross-reactivity
pattern with the Fab fragment used in this assay. But further
D-ring hydroxylated metabolites, such as epietiocholanolone,
etiocholanolone and 16␣- and 16␤-hydroxytestosterone would
not be detected in these assays, as the cross-reactivities for such
compounds were low.
3.4. Urine sample analysis
A comparison between calibration plots for testosterone stan-
dard solutions prepared in PBS and those prepared in bovine
urine gave clear evidence of matrix effect due to the presence of
electrochemical interferences. The effect of urine as a matrix,
decreased the absolute signal. In order to reduce the back-
grounds, urine samples were centrifuged and then supernatant
was collected followed by determining directly after dilution
(urine diluted 1/10 with PBS buffer). Further experiments were
performed in which the additions of serial dilutions of testos-
terone standard solution in 1/10 diluted urine were run in parallel
with PBS. As can be seen in Fig. 3, a 1/10 dilution of the
1761
Fig. 3. Comparison of competitive immunoassays in PBS () and in diluted
bovine urine () (1/10 in PBS) on SPEs using chronoamperometric detection
at +100 mV. I/I0 is the ratio of response I, to the maximum response when no
analyte is present, I0 .
urine in PBS reduced the urine matrix effect such that the curve
mimicked the standard buffer curve, which indicated minimal
interference of the urine matrix. The possibility of simplified
sample pre-treatment is important considering the possibility of
on-site measuring by the Fab sensor. Therefore in this study
bovine urine samples were determined directly after a 1/10 dilu-
tion in PBS buffer, minimising sample pre-treatment.
In order to demonstrate the precision (expressed as inter-
day variance % CV) and accuracy (expressed as % recovery)
of using electrochemical SPEs with urine samples, fortified
bovine urine samples of 1, 2, 5 and 10 ng/ml of testosterone were
prepared. Inter-day precision was assessed by measuring three
electrodes of each spiked level on three different days (n = 3,
N = 3). Accuracy was obtained by determining the recovery rate
by comparing percentage of concentrations found from the cali-
bration curve to nominal concentrations of the fortified samples.
For every spiked level (1, 2, 5 and 10 ng/ml) the recoveries found
were 107 ± 5%. In fact for the lower levels we found values of
102–107%. Only the 10 ng/ml level gave poorer recovery of
112%. This is likely due to the non-linear nature of the recov-
ery at this level. We are close to the limit of the linear range
here and thus higher error can be expected. All recoveries for
the spiked samples were higher than 100%. This could be that
natural bovine urine samples can contain minor, residual con-
centrations of testosterone, thus adding to the competition by the
analyte. In terms of the associated precision, within the spiked
values tested, we found all to be <8%. These values were consid-
ered acceptable, on account of the complexity of the biological
matrices and the excessive handling of the electrodes.
3.5. Application to urine samples of treated heifers and
comparison with GC–MS
Three adult heifers (458 ± 14 kg body weight) were assigned
for treatment with testosterone (200 mg per animal) by intra-
muscular injection. A pool of urine samples of three alternative
animals collected at 1, 3, 9, 48 and 144 h after administration
were used in this study. The results of two compared methods
are presented in Table 3. In order to compare the concentration
of metabolites detected by GC–MS with results obtained by the
1762 H. Lu et al. / Biosensors and Bioelectronics 22 (2007) 1756–1763
Table 3
Results of the comparative study of testosterone determination in bovine urine
by screen-printed electrodes and GC–MS
Time of urine SPEa (ng/ml) GC–MSb (ng/ml)
collection (h)
␣-Testosterone ␤-Testosterone
1 35.6 17 52
3 89.7 54 52
9 87.7 64 37
48 24.9 17 5.4
144 5.2 9.1 0.6
a Mean of triplicate measurements with R.S.D. value not higher than 15%.
Concentrations found from the developed calibration curves time 10-fold dilu-
tion factor.
b Analysed and supplied by European Union Community Reference Labora-
tory for Residues(CRL), National Institute of Public Health and the Environment
(RIVM), Netherlands.
Fab sensor, the peak concentrations (after 3 h, as determined
by ELISA analysis) and urine samples collected at 144 h after
treatment when the metabolites would ordinarily reach a steady
state were included. ␣- and ␤-isomers of testosterone concentra-
tions by GC–MS are presented (Lu et al., 2006). It was noticed
that there was some variation in the results between different
methods. This difference is not surprising in view of the varied
parameters (assay time, temperature and type of sample pre-
treatment) used for different techniques in different labs. Urine
samples were determined directly after centrifuge, followed by
a single step dilution by the Fab sensor in this study, while they
are used after hydrolysis step by GC–MS in RIVM lab. Urine
samples of treated heifers contain many endogenous steroids and
dietary components which may interfere with the detection of
target analytes. Antibody Fab fragment can recognize individ-
ual or several steroids isoforms. But overall the concentrations
of bovine urine samples obtained by the Fab sensor are in same
magnitude with those detected by GC–MS. The aim of screen-
ing is to find and select suspicious samples for further analysis
and our method which can provide semi-quantitative informa-
tion could be beneficial.
4. Conclusions
A disposable electrochemical sensor for the detection of
testosterone was developed using a screen-printed electrode sys-
tem, chronoamperometry and a recombinant anti-testosterone
Fab fragment for molecular recognition. In this context, the
results reported here suggest that Fab fragments generated by
the phage display library technology can be used in much the
same way in sensor applications as whole antibody molecules,
showing good affinity and specificity in measuring testosterone
and its cross-reactivity profile permits also detection of impor-
tant testosterone metabolites. The recombinant Fab fragments
can be produced and purified with high yields by cost-effective
processes. The purification yield of soluble, active Fab 220 from
4 l fermentation volume was 150 mg (data not shown). Fur-
thermore, antibody engineering techniques can be applied in
the future to refine the binding properties of Fab 220 towards
other important testosterone derivatives. The sensor exhibited
the linearity range between 300 and 40,000 pg/ml with limit of
detection of 90 pg/ml, which is comparable to values reported in
the literature and deemed to be ultra-sensitive. Hydrolysis and
extraction of urine sample, which is generally required prior
Total
to existing methods of analysis, is not a pre-requisite to detect
69
106
testosterone by our method. The simplified sample pre-treatment
101 is important for on-site monitoring. The quantitative or semi-
22.4 quantitative analysis obtained from this system can enable to
9.7 find and select suspicious samples for further extensive anal-
ysis by established analytical techniques. With simple sample
preparation, low limit of detection, and good repeatability, the
introduced method can offer alternative advantages as a primary
screening tool for meat quality control.
Acknowledgements
Financial support has been provided by the Higher Education
Authority of Ireland under the Program for Research in Third
Level Institutes. Pirkko Veijola-Bailey and Armi Boman (VTT
Biotechnology) are thanked for excellent technical assistance.
The authors thank Dr. R. Stephany (RIVM, Netherlands) for
comparison with GC–MS data and to Dr. M. Pravada for screen
printing electrodes.
References
Al-Dujaili, E.A.S., 2006. Clin. Chim. Acta 364, 172–179.
Better, Chang, C.P., Robinson, R.R., Horowitz, A.H., 1988. Science 240,
1041–1043.
Bird, R.E., Hardman, K.D., Jacobson, J.W., Johnson, S., Kaufmann, B.M., Lee,
S.-M., Lee, T., Pope, S.H., Riordan, G.S., Whitlow, M., 1988. Science 242,
423–426.
D’Orazio, P., 2003. Chin. Chim. Acta 334, 41–69.
Dhar, T.K., Ali, E., 1992. J. Immunol. Meth. 147, 167–172.
Ekin, R.P., 1993. Nucl. Med. Commun. 14, 721–735.
Hanquez, C., Urios, P., Desfosses, B., Samake, H., Lince, E., Rajkowski, K.M.,
Cittanova, N., 1987. Clin. Chim. Acta 164, 71–82.
Hemminki, A., Niemi, S., Hautoniemi, L., Söderlund, H., Takkinen, K., 1998.
J. Immunotechnol. 4, 59–69.
Hock, B., 1997. Anal. Chim. Acta 347, 177–186.
Hoogenboom, H.R., 2005. Nat. Biotechnol. 23, 1105–1116.
Huston, J.S., Levinson, D., Mudgett-Hunter, M., Tai, M.-S., Novotny, J., Mar-
golies, M.N., Ridge, R.J., Bruccoleri, R.E., Haber, E., Crea, R., Oppermann,
H., 1988. Proc. Natl. Acad. Sci. 85, 5879–5883.
Illera, J.C., Lorenzo, P.L., Silvan, G., Munro, C.J., Illera, M.J., Illera, M., 1997.
Theriogenology 47, 1375–1388.
Jamal, M., Crowe, M.A., Magner, E., 2005. Characterization of the composition
of bovine urine and its effect on the electrochemical analysis of the model
mediator, p-aminophenol. Anal. Chim. Acta. 554, 79–85.
Killard, A.J., Deasy, B., O’Kennedy, R., Smyth, M.R., 1995. Trends Anal. Chem.
14 (6), 257–265.
Kreuzer, M.P., McCarthy, R., Pravda, M., Guilbault, G.G., 2004. Anal. Lett. 37,
943–956.
Leinonen, A., Kuuranne, T., Kotiaho, T., Kostiainen, R., 2004. Steroids 69,
101–109.
Li, X.F., Ma, M., Cheng, A., Zheng, J., Tam, Y.K., 2002. Anal. Chim. Acta 457,
165–171.
Lu, H., Conneely, G., Crowe, M.A., Aherne, M., Pravda, M., Guilbault, G.G.,
2006. Anal. Chim. Acta 570, 116–123.
Mello, L.D., Kubota, L.T., 2002. Food Chem. 77, 237–256.
Navajas, R., Imaz, C., Carreras, D., Garcı́a, M., Pérez, M., Rodrı́guez, C.,
Rodrı́guez, A.F., Cortés, R., 1995. J. Chromatogr. B 673, 159–164.
 H. Lu et al. / Biosensors and Bioelectronics 22 (2007) 1756–1763
Nevanen, T.K., Söderholm, L., Kukkonen, K., Suortti, T., Teerinen, T., Linder,
M., Söderlund, H., Teeri, T.T., 2001. J. Chromatogr. A 925, 89–97.
Pemberton, R.M., Hart, J.P., Foulkes, J.A., 1998. Electrochim. Acta 43,
3567–3574.
Pemberton, R.M., Mottram, T.T., Hart, J.P., 2005. J. Biochem. Biophys. Meth.
63, 201–212.
Pulli, T., Höyhtyä, M., Söderlund, H., Takkinen, K., 2005. Anal. Chem. 77,
2637–2642.
Rajkowski, K.M., Hanquez, C., Bouzoumou, A., Cittanova, N., 1989. Clin.
Chim. Acta 183, 197–206.
Ruth, A., 2001. Best Pract. Res. Clin. Endocrinol. Metab. 15 (1), 1–16.
Samuels, T.P., Nedderman, A., Seymour, M.A., Houghton, E., 1998. Analyst
123, 2401–2404.
1763
Sanwald, P., Blankson, E.A., Duléry, B.D., Schoun, J., Huebert, N.D., Dow, J.,
1995. J. Chromatogr. B 672, 207–215.
Saudan, C., Baume, N., Mangin, P., Saugy, M., 2004. J. Chromatogr. B 810,
157–164.
Shrivastav, T.G., Basu, A., Kariya, K.P., 2003. One step enzyme linked
immunosorbent assay for direct estimation of serum testosterone. J.
Immunoassay Immunochem. 24 (2), 205–217.
Takkinen, K., Laukkanen, M.-L., Sizmann, D., Alfthan, K., Immonen, T., Vanne,
L., Kaartinen, M., Knowles, J.K.C., Teeri, T.T., 1991. Protein Eng. 4,
837–841.
Valjakka, J., Hemminki, A., Niemi, S., Söderlund, H., Takkinen, K., Bouvinen,
J., 2002. J. Biol. Chem. 277 (46), 44021–44027.
Venturelli, E., Cavalleri, A., Secreto, G., 1995. J. Chromatogr. B 671, 363–380.