LETTER doi:10.

1038/nature09531

A genome-wide RNAi screen reveals determinants of

human embryonic stem cell identity
Na-Yu Chia1,2*, Yun-Shen Chan1,3*, Bo Feng1*, Xinyi Lu1,4, Yuriy L. Orlov5, Dimitri Moreau6, Pankaj Kumar6, Lin Yang1,
Jianming Jiang1, Mei-Sheng Lau1, Mikael Huss5, Boon-Seng Soh7, Petra Kraus7, Pin Li7, Thomas Lufkin7, Bing Lim7,8,
Neil D. Clarke5,9, Frederic Bard6,9 & Huck-Hui Ng1,2,3,4,9

The derivation of human ES cells (hESCs) from human blastocysts
represents one of the milestones in stem cell biology1. The full
potential of hESCs in research and clinical applications requires a
detailed understanding of the genetic network that governs the
unique properties of hESCs. Here, we report a genome-wide RNA
interference screen to identify genes which regulate self-renewal and
pluripotency properties in hESCs. Interestingly, functionally dis-
tinct complexes involved in transcriptional regulation and chro-
matin remodelling are among the factors identified in the screen.
To understand the roles of these potential regulators of hESCs, we
studied transcription factor PRDM14 to gain new insights into its
functional roles in the regulation of pluripotency. We showed that
PRDM14 regulates directly the expression of key pluripotency gene
POU5F1 through its proximal enhancer. Genome-wide location
profiling experiments revealed that PRDM14 colocalized exten-
sively with other key transcription factors such as OCT4, NANOG
and SOX2, indicating that PRDM14 is integrated into the core
transcriptional regulatory network. More importantly, in a gain-
of-function assay, we showed that PRDM14 is able to enhance the
efficiency of reprogramming of human fibroblasts in conjunction
with OCT4, SOX2 and KLF4. Altogether, our study uncovers a
wealth of novel hESC regulators wherein PRDM14 exemplifies a
key transcription factor required for the maintenance of hESC iden-
tity and the reacquisition of pluripotency in human somatic cells.
Embryonic stem cells (ESCs) were first derived in mouse blastocysts
as cell lines that can undergo extensive self-renewal in vitro and have
the ability to undergo multi-lineage differentiation, also defined as
pluripotency2,3. Although human ESCs (hESCs) share many similar-
ities with mouse ESCs (mESCs), there are intriguing differences
between these ESCs4–6. These differences could be due to species-
specific differences in embryonic development. Alternatively, the
ESCs could be derived from cells originating from different develop-
mental stages. Consistent with this idea is the identification of
post-implantation murine epiblast-derived stem cells which show
characteristics of hESCs7,8. Hence, a comprehensive elucidation of
regulators that specify the hESC identity is essential to understand
the differences between pluripotent stem cells.
To define the genetic network that governs hESC identity, we carried
out a whole genome RNA interference (RNAi) screen (Supplementary
Fig. 1). A green fluorescent protein (GFP) reporter construct driven by
the POU5F1 upstream regulatory region was introduced into H1 hESC
to generate a stable hESC line and we used GFP fluorescence as a proxy
for the undifferentiated state of hESCs (Fig. 1a and Supplementary
Fig. 2). Using this reporter cell line, we screened a siRNA library
targeting 21,121 human genes. The screen was carried out in duplicate
and the cells were imaged for GFP and Hoechst fluorescence (Fig. 1a,
1
Gene Regulation Laboratory, Genome Institute of Singapore, 60 Biopolis Street, Singapore 138672. 2School of Biological Sciences, Nanyang Technological University, 60 Nanyang Drive, Singapore
637551. 3Graduate School for Integrative Sciences & Engineering, National University of Singapore, 28 Medical Drive, Singapore 117456. 4Dept of Biological Sciences, National University of Singapore, 14
Science Drive 4, Singapore 117543. 5Computational and Systems Biology group, Genome Institute of Singapore, 60 Biopolis Street, Singapore 138672. 6Institute of Molecular and Cell Biology, 61 Biopolis
Drive, Proteos, Singapore 138673. 7Stem Cell and Developmental Biology, Genome Institute of Singapore, 60 Biopolis Street, Singapore 138672. 8Center for Life Sciences, Harvard Medical School, 330
Brookline Avenue, Boston, Massachusetts 02115, USA. 9Department of Biochemistry, National University of Singapore, 8 Medical Drive, Yong Loo Lin School of Medicine, Singapore 117597.
*These authors contributed equally to this work.
b). The mean of the z-score for GFP fluorescence reduction (Fav) and
nuclei number reduction (Nav) were calculated. From the Fav candidate
list (Supplementary Table 1), it is reassuring that POU5F1, which is
essential for the maintenance of ESCs, was ranked first. In addition to
that, we identified several regulators implicated in mESCs, including
HCFC1, TCL1A, ZSCAN10, ZIC3, NANOG and ZNF143 to be among
the top 5% of the gene list (Supplementary Fig. 3a). Genes that could
affect survival and proliferation of hESCs could also be obtained by
quantifying the nuclei number (Supplementary Table 2 and Sup-
plementary Fig. 4).
We set a cutoff for genes with Fav z-score .2 to be considered
potential candidates, and 566 genes were obtained (Fig. 1c). Gene
ontology (GO) analysis of the 566 genes showed an enrichment for
transcription factors and translation factors (Supplementary Fig. 3b).
Reactome analysis also revealed enrichment of pathways involved in
transcription and translation (Fig. 2a and Supplementary Table 3)9.
Through the STRING database, we obtained protein–protein interac-
tions for 263 out of the 566 genes (Supplementary Fig. 3c) and among
them, we identified components of the INO80 chromatin remodelling
complex10, the mediator complex11, the COP9 signalosome12, the TAF
complex13, the eukaryotic initiation factor complex14, and spliceosome
complex15 (Fig. 2b and Supplementary Fig. 3d, e). Hence, genes coding
for proteins in known biochemical complexes which have not been
previously implicated as important for hESCs were identified.
Next, we performed a secondary validation screen for 200 of the 566
candidates from the primary screen. The pooled siRNAs for each gene
were deconvoluted into four individual siRNAs. Candidates were con-
sidered positive if they were scored by at least two siRNAs. To further
enhance the confidence of the hit genes, we adopted a multiparametric
approach where the importance of each gene in the maintenance of
hESCs was assessed by different stemness markers (OCT4 and
NANOG) for different hESCs cell lines (Supplementary Table 4).
For POU5F1–GFP hESCs, the validation rate based on the reduction
of GFP reporter, OCT4 and NANOG expression were 86.1%, 87.6%
and 63.4%, respectively, and 127 genes were validated by all three
markers. For HES2 hESCs, 86 common genes were obtained based
on OCT4 and NANOG expression and the validation rate was 75.2%
and 43.6%, respectively. Repeating with HES3 hESCs, we yielded 124
common genes with a validation rate of 64.9% and 72.3% for OCT4
and NANOG, respectively (Supplementary Fig. 5a, b and Supplemen-
tary Table 5). The higher validation rate for H1 POU5F1–GFP hESCs,
compared to other hESC lines, corroborated the fact that this cell line
was used for the primary screen. We identified 93 genes that down-
regulated OCT4 expression and 54 genes that downregulated NANOG
expression in the three different hESC lines (Supplementary Fig. 5c
and Supplementary Tables 6 and 7). In addition, we also observed a

3 1 6 | N AT U R E | VO L 4 6 8 | 1 1 NO V E M B E R 2 0 1 0
©2010 Macmillan Publishers Limited. All rights reserved
 LETTER RESEARCH

a Pluripotency can be reinstated in somatic cells through the intro-

GFP-positive duction of defined transcription factors such as OCT4, SOX2, KLF4
POU5F1 GFP
reporter and c-Myc 16,17. Hence, we used reprogramming as a gain-of-function
hESC line
promoter
1 kb
assay to assess the role of PRDM14, NFRKB and YAP1 in the acquisi-
tion of pluripotency. PRDM14 is a transcriptional regulator with a
Primary screen
potential enzymatic PR/SET domain whereas NFRKB is a component
of the INO80 chromatin remodelling complex10. YAP1 served as a
Reverse
transfection negative control. From the secondary validation screen, knockdown
on 67 of PRDM14 and NFRKB, but not YAP1, reduced OCT4 expression in
384-well
plates
the three hESCs. We transduced human fibroblasts with retroviruses
printed with containing expression constructs for OCT4 (O), SOX2 (S), KLF4 (K)
siRNAs
and c-MYC (M) to generate human induced pluripotent stem cell
(hiPSC) colonies that resembled hESCs. Interestingly, the number of
Secondary screen
hiPSC colonies was increased by 3.5-fold when we co-introduced
with individual siRNAs for 200 genes PRDM14 (P) along with OSKM (Fig. 3a). NFRKB (N) enhanced
b reprogramming by twofold whereas YAP1 did not show a positive
Non-targeting effect on reprogramming. Furthermore, knockdown of PRDM14 or
siRNA
POU5F1 siRNA
GFP siRNA
NFRKB affected the reprogramming process as fewer hiPSC colonies
were obtained (Fig. 3b). The OSKMP and OSKMN hiPSCs showed
similar characteristics and gene expression profile as hESCs (Sup-
Non-targeting plementary Fig. 7). In the presence of PRDM14, an early induction
siRNA
POU5F1 siRNA of genes associated with initiation and maturation stages of repro-
GFP siRNA gramming was observed, indicating that PRDM14 can accelerate these
stages of the reprogramming process (Supplementary Fig. 8)18. Next,
we showed that PRDM14 and NFRKB can enhance OSK-mediated
c reprogramming by 7- and 3.5-fold, respectively (Supplementary Fig.
10
9a–h). Interestingly, we also found that hiPSCs could be generated
with OSMP and OSMN, indicating that PRDM14 and NFRKB can
replace KLF4 (Supplementary Fig. 9i, j). However, no hiPSCs were
Non-targeting siRNA obtained in the absence of OCT4 or SOX2. Taken together, we demon-
5
POU5F1 siRNA strated that PRDM14 and NFRKB have important functions in the
GFP siRNA acquisition of pluripotency in human cells. This also highlights the
z score

2
Candidates with
z-score > 2
0 The role of PRDM14 in regulating the pluripotency of hESCs was
siRNAs against
other genes
–5
0 5,000 10,000 15,000 20,000
siRNAs
Figure 1 | Genome-wide screen for regulators that maintain hESC identity.
a, Schematic representation of siRNA screen. H1 hESC line with a GFP reporter
gene driven by the POU5F1 promoter was used for the screen. For the primary
screen, 21,121 genes were analysed. For secondary screen, individual siRNAs
for 200 high-confidence candidates from the primary screen were selected for
further validation using the reporter cells as well as two other hESC lines HES2
and HES3. b, Montage representing the images for GFP fluorescence (green)
and Hoechst staining (blue) for a typical 384-well plate. The hESCs were
imaged 4 days after transfection. The negative control siRNA (non-targeting)
and the positive controls siRNA (GFP and POU5F1) were added to the
designated wells as indicated. c, Dot plot of the genome-wide screen result. The
y-axis represents the average z-scores for the GFP reduction for each targeted
gene. Controls are represented by the black (non-targeting siRNA), red
(POU5F1 siRNA) and green (GFP siRNA) dots. Genes with z-score .2,
highlighted in blue, are potential candidates required for the maintenance of
hESC identity. The rest of the genes are indicated in orange.
positive correlation between the stemness markers (Supplementary
Fig. 5d). We conducted further counter-screens using a hESC line
harbouring a constitutively active EF1a (also known as EEF1A1)–
GFP reporter and also assessed the level of control proteins b-actin
and GAPDH (Supplementary Table 8). The results filtered out a minor
fraction of the candidates that could be affecting global transcription or
translation processes (Supplementary Fig. 6). Results from the secondary
validation assay were used to select genes for further functional assays
for pluripotency.
©2010 Macmillan Publishers Limited. All rights reserved
value of the candidates that we identified in our screen.
studied in greater detail as it was ranked highly in the primary screen.
Furthermore, PRDM14 is highly expressed in a variety of hESCs19 and
is a target of the core transcription factors in hESCs20. PRDM14 has
also been implicated to regulate self-renewal of hESCs since its deple-
tion induced expression of differentiation marker genes and altered the
cellular morphology21. We repeated the knockdown of PRDM14 with
four shRNA constructs in non-reporter hESCs. Depletion of PRDM14
resulted in a reduction in the expression of hESC-associated genes
in three different hESC lines (H1, H9 and HES3) (Supplementary
Figs 10a–c, 11 and 12). On the other hand, proteins that are expressed
in differentiated cell-types were upregulated upon PRDM14 depletion
(Supplementary Fig. 10d), indicating differentiation of hESCs.
Importantly, the phenotypes induced by PRDM14 shRNA could be
rescued by co-expression of RNAi-immune PRDM14 cDNA, indi-
cating that the knockdown effect is specific (Supplementary Fig. 13).
In mice, Prdm14 is essential for the establishment of the germ cell
lineage, but not required for early embryonic development22. The
knockdown of Prdm14 in mouse ESCs yielded no observable pheno-
typic change and no reduction in Pou5f1 and Sox2 (Supplementary
Fig. 14a, b). Mouse epiblast-derived stem cells (EpiSCs) are also pluri-
potent and have features of hESCs7,8, but they are deficient in Prdm14
(Supplementary Fig. 14c, d). Thus, Prdm14 is required only for the
maintenance of hESCs, but not mouse EpiSCs. It is noteworthy that
other members of the Prdm family are expressed in mouse EpiSCs and
these could be of functional importance (Supplementary Fig. 15).
To investigate further the function of PRDM14, we used chromatin
immunoprecipitation coupled with sequencing (ChIP-seq) to map the
PRDM14 binding loci (Supplementary Table 9). After validation of the
ChIP-seq data set (Supplementary Fig. 16), we derived the consensus
sequence motifs (Fig. 4a) by using the de novo motif-discovery algo-
rithm, CisFinder23. Interestingly, PRDM14 colocalizes with OCT4,
SOX2, NANOG and p300 (Fig. 4b). Co-motif analysis of the PRDM14
1 1 NO V E M B E R 2 0 1 0 | VO L 4 6 8 | N AT U R E | 3 1 7
 RESEARCH LETTER

a Carbohydrate Transmembrane
Mediator complex Translation initiation
metabolism transport
Splicing complex POL II transcription tRNA metabolism

Cell cycle Cell migration Protein metabolism Hormone biosynthesis Lipid metabolism
Number of hits per reaction

b INO80 complex Mediator complex TAF complex

MED1 TAF1
INO80D

INO80B INO80E MED26 MED28 MED12 TAF11 TAF4

TAF12
NFRKB MED13L MED17
MED24 ERCC3 MED13 TAF2
MCRS1 ACTR8 TAF10 TAF5
INO80 CDK8
CCNC TAF7
YY1 RUVBL2 MED14 TAF9 TAF6
MED16 MED19
RUVBL1 TAF8
MED15

COP9 signalosome Eukaryotic initiation Spliceosome complex

factor complex
COPS1 SF3A1
EIF2S1
EIF2S2 SF3B1 SF3A2
EIF2B1
COPS8 COPS2 RPP14 SNRPD3
CREBBP EIF2S3
SF3B2 CDK2AP2 SF3A3
EIF2B2 ADRB2 ROBLD3
COPS6 ADRA2A
COPS3 CDKN2A
SF3B3 S100A9 U2AF1
EIF2B3 EIF2B5

COPS5 EIF2B4 SF3B4 U2AF2

COPS4 SUB1

Figure 2 | Pathway analyses. a, Reactome analysis. The 566 genes (identifiers) (1 matching identifier) to red (12 matching identifiers). b, Components of the
were analysed using the web resource Reactome to determine the events INO80 chromatin remodelling complex, mediator complex, TAF complex,
(reactions and/or pathways) that were statistically over-represented. Twelve COP9 signalosome, eukaryotic initiation complex and spliceosome complex
categories with P-value , 0.05 were over-represented. Single events were with z-score . 2 are indicated in red.
coloured according to the number of matching identifiers from blue

binding loci showed a significant enrichment of a joint Sox2-Oct4 motif, region 4 (CR4)24–26. Using electrophoretic mobility shift assay (EMSA),
confirming the co-occurrence of PRDM14, OCT4 and SOX2 sites we confirmed that the CR2 sequence which contains the PRDM14 motif
(Supplementary Fig. 17). Our ChIP-seq analysis identified 2,755 genes is indeed bound directly by both recombinant and native PRDM14
that were bound by PRDM14 (Supplementary Table 10). Interestingly, proteins (Supplementary Fig. 18a, b). Similar to the differential activity
we observed a PRDM14 peak at the POU5F1 upstream regulatory of the proximal and distal enhancers in mouse EpiSCs, the CR2 reporter
region (Fig. 4c). This region is known to contain a proximal enhancer is more active than the CR4 reporter in hESCs7,8,24. Depletion of
with conserved region 2 (CR2) and a distal enhancer with conserved PRDM14 led to a downregulation of CR2 enhancer activity (Sup-
plementary Fig. 18c). Mutation of PRDM14 site at CR2 could also
reduce its activity (Supplementary Fig. 18e). Using ChIP assay, we
(Number of hiPSC colonies
(Number of hiPSC colonies

a b
confirmed that PRDM14 binds to CR2 but not CR4 (Supplementary
compared to control)
compared to control)

4 125 Fig. 18d). Furthermore, PRDM14 could activate a reporter construct

100
Percentage
Fold change

3
75 containing the CR2 sequence in 293T cells (Fig. 4d). These data indicate
2
50 that PRDM14 is directly activating the CR2 enhancer.
1 25
0 0
We further mapped the functional domains of PRDM14 with con-
structs expressing different fragments of the protein (Supplementary
1

2
l
k

P1
KB
14

tro
oc

A
DM

YA

on

RN

RN

RN

RN
FR
M

C
N

sh

sh

sh

sh
PR

Figure 3 | PRDM14 and NFRKB can enhance reprogramming of human

14

14

KB

KB
DM

DM

OSKM +
FR

FR

fibroblasts to iPSCs. a, Graph depicts fold change of the number of hiPSC

N

N
PR

PR
colonies compared to control)
colonies compared to control)

(Number of TRA-1-60-positive
(Number of TRA-1-60-positive

colonies generated from PRDM14, NFRKB or YAP1 in conjunction with

OCT4, SOX2, KLF4 and c-MYC (OSKM) with respect to the control (OSKM)
4 125
(upper panel) and fold change of the number of TRA-1-60-positive colonies
Fold change

Percentage

3 100
75 (lower panel). Each column represents the average of three replicates.
2
50 b, PRDM14 is required for reprogramming of human somatic cell. Retroviruses
1 25 harbouring PRDM14 shRNA or NFRKB shRNA were cotransduced with the
0 0 four reprogramming factors. Two independent shRNAs were used for the
1

2
l

depletion of either PRDM14 or NFRKB. The number of hESC-like iPSC (upper

tro
k

P1
KB
14

A
oc

on

RN

RN

RN

RN
M

YA
FR
M

panel) or TRA-1-60-positive (lower panel) colonies was counted 4 weeks after

D

sh

sh

sh

sh
N
PR

14

14

KB

KB

infection. All values are means 6 s.e.m. from three independent experiments
DM

DM

OSKM +
FR

FR

(n 5 3).
N

N
PR

PR

3 1 8 | N AT U R E | VO L 4 6 8 | 1 1 NO V E M B E R 2 0 1 0
©2010 Macmillan Publishers Limited. All rights reserved
 LETTER RESEARCH

a b RNAi screens in murine ESCs27–30 have uncovered many important

2.0 –1.0 1.0 regulators. Despite these efforts, little is known about the key players in
hESCs. Given the substantial differences between the two, it is of
1.5 interest to investigate the key genetic components that are specific to
SOX2
1.0
hESCs. Through a genome-wide RNAi screen, we identified novel
p300
NANOG
factors that are essential for the maintenance of hESCs and characterized
0.5 PRDM14 PRDM14 in detail as one of the hitherto unknown critical regulators of
OCT4 hESCs. Previous work on a Prdm14 knockout mouse model showed that
0
KLF4 Prdm14 is critical for the establishment of the germ cell lineage22. Hence,
1
3
5
7
9
11
13
15
17

MYC our study links a germ cell factor in mouse to the regulation of a key
Position

MYC
KLF4
OCT4
PRDM14
NANOG
p300
SOX2
pluripotency gene in hESCs.
We revealed the involvement of several components of the different
c d
functionally distinct complexes (Fig. 2b) in the maintenance of hESCs.
POU5F1 3 × CR2
It is particularly intriguing to find that chromatin remodelling complex
50 Luciferase (INO80 complex), transcriptional regulatory complexes (mediator
chr6:31235655–31254671 PRDM14
complex and TAF complex) and signalling complex (COP9 signalo-
Minimal
1
90
promoter some) are implicated in hESC biology. Future work will enable us to
OCT4 define more specifically their individual function and the mechanisms
through which these regulators operate collectively in hESCs. Our
Relative luciferase

1 10
339
NANOG 8 study shows that these regulators may hold the potential in advancing
activity

6 the methodology and understanding the mechanisms of human so-

1 4 matic cell reprogramming.
50
CTCF 2
0
1 METHODS SUMMARY
l

14
tro

50
O

Control POU5F1–GFP reporter hESC-line was generated for the genome-wide RNAi
DM
on

AN
C

PR
N

1
Figure 4 | PRDM14 is integrated into the core hESC transcriptional
regulatory network. a, PRDM14 motif is predicted using the de novo motif-
discovery algorithm CisFinder. b, PRDM14 shows co-binding with OCT4,
SOX2, NANOG and co-activator p300. Colours in the heat map reflect the
colocalization frequency of each transcription factor (the descending frequency
of localization ranges from white to yellow and to red). c, ChIP-seq binding
profile of PRDM14, OCT4, NANOG and CTCF at the POU5F1 locus. Control
ChIP-seq library was obtained from sequencing of input DNA. d, Three copies
of CR2 (conserved region 2) consensus motif were inserted in tandem before
the minimal promoter of the reporter construct. NANOG or PRDM14
expression construct was cotransfected with the reporter construct into 293T
cells and luciferase activity was normalized against the control vector. All values
are means 6 s.e.m. from three independent experiments (n 5 3).
Fig. 19a). The putative DNA-binding domain and the amino-terminal
region are required for transcriptional activation (Supplementary
Fig. 19b). Importantly, we showed that these domains are required
for the acquisition of pluripotency using the reprogramming assay
(Supplementary Fig. 19c). The DNA binding activity resides within
the carboxy-terminal zinc finger region as deleting this region
abolished direct interaction with DNA (Supplementary Fig. 19d).
Five of the six zinc fingers are required for DNA binding as well as
transcriptional activity (Supplementary Fig. 19e, f).
The positive regulation of POU5F1 expression by PRDM14 is un-
expected as previous studies implicate PRDM14 as a transcriptional
repressor21,22. To identify the genes which are regulated by PRDM14,
we performed expression profiling upon PRDM14 depletion. We
found that 358 of the 2,645 PRDM14-bound genes (13.5%) were
downregulated and 638 of the PRDM14-bound genes (24.1%) were
induced (Supplementary Figs 20a–d and 21 and Supplementary Tables
11 and 12). This finding indicates that PRDM14 can have both positive
and negative roles on transcription. An OCT4 position weight matrix
was enriched at the PRDM14 sites and PRDM14 ChIP-seq peaks
were also observed to co-occur with OCT4, SOX2 and NANOG at
PRDM14-regulated genes (Supplementary Fig. 22). Altogether, our
genome-wide analyses showed that the target genes of PRDM14 are
involved in diverse cellular processes (Supplementary Fig. 20e).
Identification of new molecules required to maintain the identity of
ESCs is critical to the understanding of the mechanisms that govern
self-renewal and pluripotency. Both small-scale and genome-wide
screen. The primary, secondary and counter screens were carried out in a 384-
well plate format. Gene Ontology was performed with Panther classification
(www.pantherdb.org). Pathway analyses were carried out using Reactome
(www.reactome.org) and STRING (http://string.embl.de/). Individual shRNAs
for each gene were designed using the WI siRNA selection program (http://jur-
a.wi.mit.edu/bioc/siRNAext/). ChIP-seq data was generated using Illumina GA
single-read sequencing and peak calling was carried out with MACS. For expres-
sion profiling experiments, RNA was amplified and hybridized to an Illumina
microarray (HumanRef-8 v3.0 Expression BeadChips). pMXs retroviral plasmids
that carry cDNAs of human OCT4, SOX2, KLF4, c-MYC, PRDM14, NFRKB
and YAP1 were used for reprogramming. Teratomas were generated 8 weeks
after subcutaneous injection of hESCs or hiPSCs into SCID mice. Immuno-
fluorescence staining, immunohistochemistry, ChIP, RT–PCR (PCR with reverse
transcription ), western blotting, co-immunoprecipitation and EMSA were per-
formed using standard protocols.
Full Methods and any associated references are available in the online version of
the paper at www.nature.com/nature.
Received 11 January; accepted 28 September 2010.
Published online 17 October 2010.
1. Thomson, J. A. et al. Embryonic stem cell lines derived from human blastocysts.
Science 282, 1145–1147 (1998).
2. Evans, M. J. & Kaufman, M. H. Establishment in culture of pluripotential cells from
mouse embryos. Nature 292, 154–156 (1981).
3. Martin, G. R. Isolation of a pluripotent cell line from early mouse embryos cultured
in medium conditioned by teratocarcinoma stem cells. Proc. Natl Acad. Sci. USA
78, 7634–7638 (1981).
4. Yu, J. & Thomson, J. A. Pluripotent stem cell lines. Genes Dev. 22, 1987–1997
(2008).
5. Van Hoof, D. et al. A quest for human and mouse embryonic stem cell-specific
proteins. Mol. Cell. Proteomics 5, 1261–1273 (2006).
6. Pera, M. F. & Trounson, A. O. Human embryonic stem cells: prospects for
development. Development 131, 5515–5525 (2004).
7. Brons, I. G. et al. Derivation of pluripotent epiblast stem cells from mammalian
embryos. Nature 448, 191–195 (2007).
8. Tesar, P. J. et al. New cell lines from mouse epiblast share defining features with
human embryonic stem cells. Nature 448, 196–199 (2007).
9. Joshi-Tope, G. et al. Reactome: a knowledgebase of biological pathways. Nucleic
Acids Res. 33, D428–D432 (2005).
10. Conaway, R. C. & Conaway, J. W. The INO80 chromatin remodeling complex in
transcription, replication and repair. Trends Biochem. Sci. 34, 71–77 (2009).
11. Casamassimi, A. & Napoli, C. Mediator complexes and eukaryotic transcription
regulation: an overview. Biochimie 89, 1439–1446 (2007).
12. Chamovitz, D. A. Revisiting the COP9 signalosome as a transcriptional regulator.
EMBO Rep. 10, 352–358 (2009).
13. Albright, S. R. & Tjian, R. TAFs revisited: more data reveal new twists and confirm
old ideas. Gene 242, 1–13 (2000).

1 1 NO V E M B E R 2 0 1 0 | VO L 4 6 8 | N AT U R E | 3 1 9
©2010 Macmillan Publishers Limited. All rights reserved
 RESEARCH LETTER
14. Jackson, R. J., Hellen, C. U. & Pestova, T. V. The mechanism of eukaryotic translation
initiation and principles of its regulation. Nature Rev. Mol. Cell Biol. 11, 113–127
(2010).
15. Rino, J. & Carmo-Fonseca, M. The spliceosome: a self-organized macromolecular
machine in the nucleus? Trends Cell Biol. 19, 375–384 (2009).
16. Takahashi, K. et al. Induction of pluripotent stem cells from adult human
fibroblasts by defined factors. Cell 131, 861–872 (2007).
17. Park, I. H. et al. Reprogramming of human somatic cells to pluripotency with
defined factors. Nature 451, 141–146 (2008).
18. Samavarchi-Tehrani, P. et al. Functional genomics reveals a BMP-driven
mesenchymal-to-epithelial transition in the initiation of somatic cell
reprogramming. Cell Stem Cell 7, 64–77 (2010).
19. Assou, S. et al. A meta-analysis of human embryonic stem cells transcriptome
integrated into a web-based expression atlas. Stem Cells 25, 961–973 (2007).
20. Boyer, L. A. et al. Core transcriptional regulatory circuitry in human embryonic
stem cells. Cell 122, 947–956 (2005).
21. Tsuneyoshi, N. et al. PRDM14 suppresses expression of differentiation marker
genes in human embryonic stem cells. Biochem. Biophys. Res. Commun. 367,
899–905 (2008).
22. Yamaji, M. et al. Critical function of Prdm14 for the establishment of the germ cell
lineage in mice. Nature Genet. 40, 1016–1022 (2008).
23. Sharov, A. A. & Ko, M. S. Exhaustive search for over-represented DNA sequence
motifs with CisFinder. DNA Res. 16, 261–273 (2009).
24. Hanna, J. et al. Metastable pluripotent states in NOD-mouse-derived ESCs. Cell
Stem Cell 4, 513–524 (2009).
25. Nordhoff, V. et al. Comparative analysis of human, bovine, and murine Oct-4
upstream promoter sequences. Mamm. Genome 12, 309–317 (2001).
26. Yeom, Y. I. et al. Germline regulatory element of Oct-4 specific for the totipotent
cycle of embryonal cells. Development 122, 881–894 (1996).
27. Ivanova, N. et al. Dissecting self-renewal in stem cells with RNA interference. Nature
442, 533–538 (2006).
28. Fazzio, T. G., Huff, J. T. & Panning, B. An RNAi screen of chromatin proteins identifies
Tip60-p400asaregulatorofembryonicstemcellidentity.Cell134,162–174(2008).
3 2 0 | N AT U R E | VO L 4 6 8 | 1 1 NO V E M B E R 2 0 1 0
©2010 Macmillan Publishers Limited. All rights reserved
29. Ding, L. et al. A genome-scale RNAi screen for Oct4 modulators defines a role of
the Paf1 complex for embryonic stem cell identity. Cell Stem Cell 4, 403–415
(2009).
30. Hu, G. et al. A genome-wide RNAi screen identifies a new transcriptional module
required for self-renewal. Genes Dev. 23, 837–848 (2009).
Supplementary Information is linked to the online version of the paper at
www.nature.com/nature.
Acknowledgements We are grateful to the Biomedical Research Council (BMRC),
Agency for Science, Technology and Research (A*STAR) and Singapore Stem Cell
Consortium for funding. We are grateful to K. Kuay, L.-P. Yaw, C.-K. Tong and C.-W.
Chang for technical assistance. We acknowledge V. Cacheux-Rataboul for karyotype
analysis and the GTB group for sequencing. We are grateful to A. Surani, P. Tesar and
R. Mckay for gift of EpiSCs and Q. Yu for plasmids. We thank A. Colman, A. Hutchins and
T. Huber for comments on the manuscript.
Author Contributions N.-Y.C. conducted the genetic screen, generated the POU5F1–
GFP line and performed the secondary screens. ChIP experiments and EMSA were
conducted by Y.-S.C. and M.-S.L. Reprogramming experiments were done by B.F. and
L.Y. Luciferase experiments and target validations were carried out by X.L.
Bioinformatics analyses were performed by Y.L.O., P.K., M.H. and N.D.C.; D.M. printed
the siRNA plates. P.K. and T.L. supported the in vivo mouse work. B.-S.S. and P.L.
generated the EF1–GFP reporter cells. H.-H.N., F.B. and N.-Y.C. wrote the manuscript
with contributions from Y.-S.C., B.F., B.L. and J.J.; N.-Y.C., H.-H.N. and F.B. designed the
experiments.
Author Information Microarray and ChIP-seq data are deposited at the Gene
Expression Omnibus (http://www.ncbi.nlm.nih.gov/geo) under accession numbers
GSE22792, GSE22795 and GSE22767. Reprints and permissions information is
available at www.nature.com/reprints. The authors declare no competing financial
interests. Readers are welcome to comment on the online version of this article at
www.nature.com/nature. Correspondence and requests for materials should be
addressed to H.-H.N. (nghh@gis.a-star.edu.sg) or F.B. (fbard@imcb.a-star.edu.sg).

METHODS
Cell culture and generation of POU5F1–GFP reporter cell-line. The hESC lines
H1 (WA-01, passage 28), H9 (WA-09, passage 26), HES2 (ES-02, passage 79),
HES3 (ES-03, passage 97) and H1 POU5F1–GFP reporter cells (passage 56) were
used for this study1. They were cultured feeder-free on Matrigel (BD)1. Condition
medium used for culturing hESCs contained 20% KO serum replacement, 1 mM
L-glutamine, 1% non-essential amino acids and 0.1 mM 2-mercaptoethanol and
an additional 8 ng ml21 of basic fibroblast growth factor (Invitrogen) supplemen-
ted to the hESC unconditioned medium. Medium was changed daily. The hESCs
were subcultured with 1 mg ml21 collagenase IV (Gibco) every 5–7 days.
A 3,064-base pairs upstream region of human POU5F1 gene was cloned
upstream of a GFP reporter gene, in replacement of the cytomegalovirus promoter
of the N-EGFP plasmid with Geneticin (Gibco) drug selection marker. The
POU5F1–GFP construct (2 mg) was transfected into H1 hESCs using 6 ml of
FuGENE (Roche). Drug resistant colonies appeared after 2 weeks of drug selection.
Transfection, staining and imaging in 384-well plates. Plates (384-well, Grenier)
were coated with 10 ml of matrigel for 30 min at 37 uC before removing the excess
Matrigel. Pooled or individual siRNAs (5 ml of 500 nM; siGenome, Dharmacon)
were printed on the plates and frozen at 220 uC before use. During reverse transfec-
tion, a master mix of 0.05 ml of Dharmafect1 (Dharmacon) transfection reagent and
4.95 ml of OptiMEM (Invitrogen) mix was added to each well of the siRNA plates
and incubated for 20 min. Subsequently, 3,000 cells in 40 ml of conditioned medium
with 10 mM Rock inhibitor (Calbiochem) were seeded in each well. Reagents and
cells were dispensed onto the plate using a multidrop (Thermo Scientific).
Using the H1 POU5F1–GFP reporter cell-line, we carried out a pilot experiment
to determine the optimized conditions (amount of transfection reagent and cell
density) for reverse transfection. To enable high transfection efficiency, the cells
were dissociated into single cells and reverse-transfected with the siRNA/transfec-
tion lipid complexes. A Z9 factor of more than 0.5 was obtained from this pilot
screen, indicating a robust dynamic range between the positive (GFP siRNA) and
negative (non-targeting siRNA) controls for a high throughput screen.
For the genome-wide screen, the whole genome Dharmacon SMARTpooled
siRNA library targeting 21,121 human genes was printed on 67 Matrigel-coated
384-well plates where each well contained a mixture of four siRNAs targeting a single
gene. On each plate, we included the negative controls (non-targeting siRNA) and
positive controls (GFP siRNA and POU5F1 siRNA) in the designated wells. After
4 days of transfection, the medium was replaced with 30 ml of 4% paraformaldehyde
(Sigma). The cells were fixed for 15 min before washing with PBS. Hoechst 3342
(1:10,000, Invitrogen) in 0.1%Triton-X/1% BSA was added to each well and stained
for 30 min. The cells were then washed once with PBS and covered in 30 ml PBS.
©2010 Macmillan Publishers Limited. All rights reserved
LETTER RESEARCH
Cells were imaged with an IXU ultra confocal microscope (Research
Instruments) at 320 magnification and four pictures were taken per well.
Integrated fluorescent intensity and nuclei number were quantified using
MetaXpress Image Acquisition and Analysis software V1.7. Z9 factor was calcu-
lated for the primary screen based on the formula Z9 5 1 2 3(sp 1 sn)/(mp 2 mn)
where sp is the standard deviation of the positive control, sn is the standard
deviation of the negative control, mp is the mean of the positive control and mn is
the mean of the negative controls. z-score was calculated using the formula
z 5 (X 2 m)/s.d. where m is the mean of the negative controls and s.d. is the
standard deviation of the whole population. X is the sample value calculated based
on the integrated fluorescent intensity/ number of cells. The Z9 factor for the entire
screen was 0.76.
Retroviral production and human iPSC induction. pMXs retroviral plasmids
that carry cDNAs of human OCT4, SOX2, KLF4 and c-MYC genes were obtained
from Addgene (plasmids 17217, 17218, 17219 and 17220, S. Yamanaka)16. cDNAs
of human PRDM14, NFRKB and YAP1 were cloned into pMX vector for retrovirus-
mediated overexpression. Retroviruses were packaged using Pantropic Retroviral
Expression System (Clontech) and concentrated with centrifugal filter devices
(Millipore). MRC-5 cells obtained from ATCC were cultured in 15% FBS/
DMEM. Confluent MRC-5 cells were split into 24 wells 1 day ahead and then
transduced with equal amount of the retroviruses stock in presence of 4 mg ml21
polybrene (Sigma). After 24 h, the cells were changed to fresh 15% FBS/DMEM
medium, and then split from 1 well of a 24-well plate into 2 wells of a 6-well plate
with pre-seeded CF-1 feeders the next day. The cultures were then maintained in
human ESC culture medium and fed every 2 days. To expand and characterize
hiPSCs, each emerged hESC-like colony was mechanically dissociated to small
clumps and transferred into one 6-well plate with CF-1 feeder.
Reporter assays. A minimal POU5F1 proximal promoter region (350 bp) was
cloned into the pGL3 basic vector (Promega), driving the luciferase gene via the
cloning site BglII/NcoI. The CR2 and CR4 fragments (550 and 500 bp, respec-
tively) were cloned into the PGL3-POU5F1 proximal promoter vector down-
stream of the luciferase gene via the cloning site BamHI/SalI. For the cloning of
reporter vector used to test the functional domains of PRDM14, three copies of
30-bp CR2 consensus motif were synthesized and cloned into the XhoI/BglII site
in front of the minimal promoter of pGL4.23 vector (Promega) in tandem. H1,
HES2 and HES3 hESCs were transfected with the reporter constructs using
FuGENE (Roche) and E14 mESCs and 293T cells using Lipofectamine 2000
(Invitrogen). Cells were harvested 48–60 h after transfection and the luciferase
activities were quantified using the Dual-luciferase Reporter Assay System
(Promega).