LETTERdoi:10.1038/nature12745

Derivation of novel human ground state naivepluripotent stem cellsOhad Gafni1*, Leehee Weinberger1*, Abed AlFatah Mansour1*, Yair S. Manor1*, Elad Chomsky1,2,3*, Dalit Ben-Yosef4,5,Yael Kalma4, Sergey Viukov1, Itay Maza1, Asaf Zviran1, Yoach Rais1, Zohar Shipony2,3, Zohar Mukamel2,3, Vladislav Krupalnik1,Mirie Zerbib1, Shay Geula1, Inbal Caspi1, Dan Schneir1, Tamar Shwartz4, Shlomit Gilad6, Daniela Amann-Zalcenstein6,Sima Benjamin6, Ido Amit7, Amos Tanay2,3, Rada Massarwa1, Noa Novershtern1 & Jacob H. Hanna1

Mouse embryonic stem (ES) cells are isolated from the inner cellmass of blastocysts, and can be preserved in vitro in a naive inner-cell-mass-like configuration by providing exogenous stimulationwith leukaemia inhibitory factor (LIF) and small molecule inhibitionof ERK1/ERK2 and GSK3b signalling (termed 2i/LIF conditions)1,2.Hallmarks of naive pluripotency include driving Oct4 (also known asPou5f1) transcription by its distal enhancer, retaining a pre-inactivationX chromosome state, and global reduction in DNA methylation andin H3K27me3 repressive chromatin mark deposition on develop-mental regulatory gene promoters3. Upon withdrawal of 2i/LIF, naivemouse ES cells can drift towards a primed pluripotent state resem-bling that of the post-implantation epiblast4. Although human EScells share several molecular features with naive mouse ES cells5,they also share a variety of epigenetic properties with primed mur-ine epiblast stem cells (EpiSCs)6,7. These include predominant use ofthe proximal enhancer element to maintain OCT4 expression, pro-nounced tendency for X chromosome inactivation in most femalehuman ES cells, increase in DNA methylation and prominent depos-ition of H3K27me3 and bivalent domain acquisition on lineageregulatory genes7. The feasibility of establishing human ground statenaive pluripotency in vitro with equivalent molecular and func-tional features to those characterized in mouse ES cells remains tobe defined1. Here we establish defined conditions that facilitate thederivation of genetically unmodified human naive pluripotent stemcells from already established primed human ES cells, from somaticcells through induced pluripotent stem (iPS) cell reprogramming ordirectly from blastocysts. The novel naive pluripotent cells validatedherein retain molecular characteristics and functional propertiesthat are highly similar to mouse naive ES cells, and distinct fromconventional primed human pluripotent cells. This includes com-petence in the generation of cross-species chimaeric mouse embryosthat underwent organogenesis following microinjection of humannaive iPS cells into mouse morulas. Collectively, our findings estab-lish new avenues for regenerative medicine, patient-specific iPScell disease modelling and the study of early human developmentin vitro and in vivo.

Naive 2i/LIF conditions are not sufficient to elicit a self-renewalresponse in human ES cells/human iPS cells and result in their differ-entiation7. Nevertheless, the proof of concept for the bi/metastabilitybetween naive and primed pluripotent states in rodents8, has raised thepossibility that the human genetic background might be less permissivefor allowing in vitro preservation of ground state-naive pluripotency incomparison to rodents1,7. Further, it is feasible that different exogenousfactor combinations, in addition to 2i/LIF, may allow in vitro stabiliza-tion of transgene-independent and indefinitely stable human naive pluri-potent cells1. To define such conditions, we used a ‘secondary’ human

C1.2 iPS cell line containing doxycycline (dox)-inducible lentiviraltransgenes encoding OCT4, SOX2, KLF4 and c-MYC (OSKM) repro-gramming factors, that was targeted with an OCT4–GFP–2A–PUROknock-in reporter construct (Fig. 1a)7. As previously established, exo-genous expression of reprogramming factors encoding transgenes bydox supplementation allows maintenance of cells that are morpholo-gically similar to mouse ES cells, while retaining approximately 60%OCT4–GFP1 cell fraction in 2i/LIF conditions (Extended Data Fig. 1a, b)7.This cell line was used to screen for components that, upon dox with-drawal, could stabilize C1.2 human iPS cells in 2i/LIF indefinitely withhomogenous OCT4–GFP1 expression (in nearly 100% of the cells)(Fig. 1a).

Whereas C1.2 cells rapidly differentiated in 2i/LIF only conditions,the combined action of 16 factor conditions (16F, divided into pool 1and pool 2 subgroups) attenuated the differentiation propensity andallowed retaining 32% of OCT4–GFP1 cells as measured at day 14after dox withdrawal (Extended Data Fig. 1b). This indicated that the16F combination contains factors that cooperatively promote humanpluripotency maintenance in 2i/LIF conditions. When pool 2 compo-nents were removed (Extended Data Fig. 1b), OCT4–GFP1 cell fractionincreased relative to 16F combination, indicating that pool 2 containedfactors that were negatively influencing GFP1 cell maintenance (ExtendedData Fig. 1b). We propose that FGFR and TGFR pathway inhibitionwas detrimental to growing human pluripotent cells in 2i/LIF condi-tions, as FGF2 and TGF-b cytokines have an evolutionary divergentfunction in promoting pluripotency in humans by inducing naivepluripotency KLF4 and NANOG transcription factor expression inhuman ES cells9, but not in murine EpiSCs, where they promote mur-ine pluripotency priming. Indeed, removing both TGFR and FGFRinhibitors recapitulated the phenotype obtained when removing doxand pool 2 components (Extended Data Fig. 1b). Moreover, pool 1components supplemented with exogenous FGF2 and TGF-b1 cyto-kines resulted in homogenous OCT4–GFP detection in .95% of cellsindependent of dox in 2i/LIF containing conditions (Extended DataFig. 1b). We then tested which of the pool 1 components were essential,and found that 2i/LIF, p38i, JNKi together with FGF2 and TGF-b1cytokine supplementation were essential to maintain exogenous trans-gene-independent GFP1 C1.2 clones (Extended Data Fig. 1c). Secondaryoptimization identified Rho-associated coiled-coil kinases (ROCK)10

and protein kinase C (PKC)11 inhibitors (included in pool 2, Fig. 1a) asoptional boosters of naive cell viability and growth (Extended DataFig. 1d–f), and resulted in optimized chemically defined conditionstermed NHSM (naive human stem cell medium) (Fig. 1b). NHSM con-ditions enabled expansion of karyotypically normal OCT4–GFP1 C1.2human iPS cells for over 50 passages independent of exogenous trans-gene activation (Fig. 1c). Remarkably, human pluripotent cells grown

*These authors contributed equally to this work.

1The Department of Molecular Genetics, Weizmann Institute of Science, Rehovot 76100, Israel. 2The Department of Biological Regulation, Weizmann Institute of Science, Rehovot 76100, Israel. 3TheDepartment of Computer Science and Applied Mathematics, Weizmann Institute of Science, Rehovot 76100, Israel. 4Wolfe PGD Stem Cell Lab, Racine IVF Unit, Lis Maternity Hospital, Tel Aviv SouraskyMedical Center, Tel-Aviv, Israel. 5The Department of Cell and Developmental Biology, Sackler Medical School, Tel-Aviv University, Israel. 6The Israel National Center for Personalized Medicine (INCPM),Weizmann Institute of Science, Rehovot 76100, Israel. 7The Department of Immunology, Weizmann Institute of Science, Rehovot 76100, Israel.

0 0 M O N T H 2 0 1 3 | V O L 0 0 0 | N A T U R E | 1

Macmillan Publishers Limited. All rights reserved©2013

in NHSM have domed-shaped colonies resembling murine naive cells;thus we refer to the selected cells as naive human pluripotent stem cells(hPSCs; includes human iPS cells and human ES cells), as we system-atically validate the evidence for this claim.

We examined whether NHSM conditions allow derivation of newhuman ES cell lines from human blastocysts. Human-blastocyst-derivedinner cell masses were plated in NHSM conditions, and successfullygenerated domed cell outgrowths after 6–8 days (Fig. 1d). We were ableto establish four newly derived naive human ES cell lines termed LIS1,LIS2, WIS1 and WIS2 (Fig. 1d and Extended Data Fig. 1g). Severalconventional (hereafter will be named ‘primed’) already establishedhuman ES cell lines (H1, H9, BGO1, WIBR1, WIBR2 and WIBR3) andhuman iPS cell lines (C1 and C2) were plated on gelatin/vitronectin-coated dishes in NHSM medium. After 4–8 days, dome-shaped col-onies with packed round cell morphology could be readily isolated andfurther expanded (Fig. 1e). Human fibroblast cells were reprogrammedto human iPS cells in NHSM following reprogramming factor trans-duction (data not shown). All human ES cell and iPS cell lines expandedin NHSM conditions were positive for pluripotent markers and formedmature teratomas in vivo (Extended Data Fig. 2). Human naive plur-ipotent lines maintained normal karyotype after extended passagingfollowing trypsinization and expansion on irradiated mouse embry-onic fibroblast (MEF) feeder cells or on vitronectin/gelatin-coatedplates (Extended Data Fig. 2a). The average doubling time was signifi-cantly reduced from 26 h for primed hPSCs down to 14 h for naivehPSCs (Extended Data Fig. 3a). Naive hPSCs displayed up to 35%single-cell cloning efficiency after trypsinization and sorting (withoutthe use of ROCK inhibitors), whereas primed hPSCs largely did notsurvive single-cell cloning (Extended Data Fig. 3b). In the presence ofROCK inhibitor, naive hPSCs had single-cell cloning efficiency of up to88%, whereas that of primed cells increased only up to 22% (ExtendedData Fig. 3b, c).

We next aimed to characterize epigenetic features in naive humaniPS cells and human ES cells established in NHSM conditions. Primedand naive human ES cell/human iPS cells were transfected with a luci-ferase reporter construct under the control of either the human distal

(DE) or the proximal (PE) enhancer sequences that control expressionof the OCT4 gene (also known as POU5F1). Consistent with previousreports, primed human ES cells showed preferential activation ofthe proximal OCT4 enhancer element as typically seen in murineEpiSCs6,7. Remarkably, predominant utilization of the OCT4 distalenhancer was detected in naive human ES cells and human iPS cells(Extended Data Fig. 3d). To further substantiate these findings, WIBR3human ES cells were stably transfected with full-length OCT4–GFP–2A–PURO, DPE–OCT4–GFP–2A–PURO or DDE–OCT4–GFP–2A–PURO engineered BAC reporter constructs (Extended Data Fig. 3e).The wild-type OCT4–GFP reporter was specifically active in bothnaive and primed conditions (Fig. 2a). The DPE–OCT4–GFP reporterwas predominantly active in naive growth conditions whereas theDDE–OCT4–GFP reporter was more active in primed pluripotent cells(Fig. 2a). We then analysed the frequency and properties of X inac-tivation state12 in several naive human ES cell/human iPS cell lines.Naive hPSCs captured in NHSM maintain a pre-X inactivation stateas evident by nearly complete lack of H3K27me3 nuclear foci anddown regulation of XIST transcription. The majority of primed hPSCsdemonstrated X inactivation as evident by the presence of H3K27me3nuclear foci and methylation of one of the XIST gene alleles (Fig. 2b andExtended Data Fig. 4a–d)13. Genome-wide mapping of H3K9me3 bychromatin immunoprecipitation, followed by deep sequencing (ChIP-seq) in female and male naive and primed human pluripotent cells,indicated a significant increase (P , 3.8 3 10–63) in this mark on the Xchromosome only in female primed pluripotent cells (Extended DataFig. 4c). XIST promoter alleles are demethylated in male and femalenaive ES cells (Extended Data Fig. 4d). Upon differentiation of femalenaive human ES cells/human iPS cells, inactivation of one of the Xchromosomes alleles becomes evident as the cells demonstrate H3K27me3clouds, upregulate XIST transcription simultaneously with methyla-tion of one of the promoters of XIST alleles (Fig. 2b and Extended DataFig. 4a–d).

Next we compared global gene expression patterns between naiveand primed human ES cells and human iPS cells. Unbiased clusteringof genome-wide expression profiles demonstrated that naive human

a

2i/LIF +

dox (OSKM)

Transgene independent

Naive human C1.2

OCT4–GFP+ hiPSCs

(no dox)

Transgene dox-dependent

naive human C1.2

+ OCT4–GFP–2A-Puro

hiPSC

???

dox withdrawal

2i/LIF

+ defined factors

ERK1/2i

GSK3βi

LIF

BMP4

IGF1

p38i

JNKi

Forskolin

FGFRi

TGFRi

Kenpaullone

PKCi

BayK8644

BixO1294

ROCKi

SCF

16 factors

Pool 1 Pool 2

b Essential components:

LIF (20 ng ml–1) TGFβ1 (1 ng ml–1)

FGF2 (8 ng ml–1) ERK1/2i (PD0325901 1 μM) GSK3βi (CHIR99021 3 μM)

JNKi (SP600125 10 μM)

p38i (SB203580 10 μM)

Optimizing components: ROCKi (Y-27632 5 μM) PKCi (Go6983 5 μM)

NHSM c Naive C1.2 hiPSC

(P51, in NHSM conditions)

d Naive hESC/hiPSC Naive hESC

+NHSM

WIS

1 h

ES

C

e

+NHSM

Primed/conventional human ESC or iPSC

BG

O1 h

ES

C

P0 (Day 6)

in NHSM

phase GFP

Primed C1.2 hiPSC

(P16)

Human blastocyst

P4 Day 4 in NHSM

P4 in NHSM

Phase OCT4–GFP

Figure 1 | Capturing the ground state of humannaive pluripotency. a, Schematic drawing of thestrategy used for calibrating conditions to isolatenaive transgene-independent human iPS cells(hiPSCs) in the presence of 2i/LIF. Inhibitor isabbreviated as ‘i’ in the dotted box. Scale bar,200mm. b, Components of optimized NHSMconditions. c, Representative large-field view ofOCT4–GFP1 human iPS cell colonies grown inNHSM (left) and, for comparison, primed/conventional iPS cells (right). Scale bars, left,200mm; right, 50mm. d, Human-blastocyst-derived inner cell masses were plated on feeder cellsin NHSM conditions. At day 6–8, the originaloutgrowth was trypsinized, and naive pluripotentcell lines were established in NHSM conditions(representative images of established line at theindicated passage (P) are shown). hESC, humanembryonic stem cell. Scale bars, left, 50mm; right,200mm. e, Epigenetic reversion of alreadyestablished primed/conventional human ES cells tonaive ground state. Representative images of BGO1human ES cell line are shown. Scale bars, left,200mm; right, 30mm.

RESEARCH LETTER

2 | N A T U R E | V O L 0 0 0 | 0 0 M O N T H 2 0 1 3

Macmillan Publishers Limited. All rights reserved©2013

ES cells and human iPS cells possess a distinct gene expression patternand clustered separately from primed human ES cells and human iPScells (Extended Data Fig. 5a). Multiple transcripts associated with naivepluripotency were significantly upregulated in naive cells, includingmembers of the NANOG and DUSP gene families (Extended DataFig. 5b)14. Importantly, naive pluripotent cells had profound down-regulation of transcripts associated with lineage-commitment genes,including ZIC1, SOX6 and SOX11 that were expressed at low, butappreciable, levels in primed human ES cells (Extended Data Fig. 5b)3.Functional annotation analysis of differentially expressed genes withGene Ontology (GO) revealed that genes downregulated in mouse andhuman naive pluripotency were significantly enriched for GO termslinked to developmental processes (Extended Data Fig. 5c). Furthermore,hierarchical clustering showed that human inner cell mass samples aretranscriptionally more similar to naive human ES cells than to primedcells (Extended Data Fig. 5d). Primed human ES cells demonstrateintermediate expression levels of MHC class I surface antigen in com-parison to somatic cells, whereas naive human ES cell/human iPS cellexpress only trace levels of MHC class I (Extended Data Fig. 6a)7.Moreover, although E-CADHERIN is expressed in primed humanES cells, the surface expression pattern becomes more prominent innaive human ES cell colonies (Extended Data Fig. 6b). Next we con-ducted an unbiased cross-species hierarchical clustering of the globallymeasured transcriptome to evaluate whether the human primed andnaive pluripotent cells, described herein, globally correspond to thoseestablished in mice. By applying an algorithmic cross-species geneexpression analysis on all 9,803 mouse–human orthologous genesfound in our gene expression data sets7,8, we found that whereasprimed human ES cells and human iPS cells clustered with murineEpiSCs, all naive human ES cells/human iPS cells clustered with naive

mouse ES cells/mouse iPS cells independent of the genetic backgroundor naive growth conditions used (Fig. 2c). Further, primed mouse andhuman pluripotent cells expression patterns were significantly moreheterogeneous in comparison with their naive counterparts, asreflected by the noise distributions of the two groups in both species(Extended Data Fig. 5e). Finally, nuclear localization of the transcrip-tion factor TFE3 was recently shown to be enhanced in naive mouse EScells, and compromised upon pluripotency priming15. Remarkably, asimilar nuclear enrichment pattern for TFE3 was evident in naivehuman ES cells, and relative enrichment was compromised in primedhuman ES cells (Fig. 2d, Extended Data Fig. 6c). Collectively, our datademonstrates that NHSM endows human pluripotent cells with defin-ing features of gene expression patterns typically seen in ground statenaive mouse ES cells3,8.

We moved on to map H3K4me3 and H3K27me3 chromatin marksby using ChIP-seq in mouse and human naive and primed pluripo-tent cells. Whereas distribution of both epigenetic markers over allgenes promoters and bodies in mouse ES cells showed a significant(P , 2 3 10237) decrease in naive conditions (Extended Data Fig. 4e),an even more notable decrease of both marks in mouse ES cells wasobserved over developmental genes (P , 1.6 3 10271; Fig. 3a, right)3.Similarly, there was a marked (P , 8.6 3 10261) reduction of H3K27me3in promoter and gene-body region over developmental genes (n 5 5,922) inhuman naive cells compared to primed (Fig. 3a, b). The reduction ofH3K27me3 mark to nearly background levels over developmentalgenes in human naive cells was also reflected in the number of geneswith bivalent marks near their TSS, which is more than 13-fold higherin primed (3,013 genes) compared to naive cells (226 genes). Notably,the naive pluripotency regulator, TBX3, acquired repressive H3K27me3marks in both primed human and mouse samples (Extended Data

b H3K27me3 OCT4

Prim

ed

E

Bs

Naiv

e

LIS2 hESC

Naive

NHSM

conditions

Primed

hESC medium

Differentiated

Even

t co

un

t

Merge with

DAPI

01

WIBR3 hESCWIBR3 hESC

WIBR2 hESCWIBR1 hESC

WIBR3 hESCWIBR3 re-primed

NOD mEpiSC

NOD mEpiSC

WIS

1 h

ES

CC

1 h

iPS

C

H1

hE

SC

WIB

R3

hE

SC

BG

O1

hE

SC

H9

hE

SC

WIS

2 h

ES

C

H9

hE

SC

BG

O1

hE

SC

WIB

R2

hE

SC

WIB

R1

hE

SC

C1

hiP

SC

WIB

R3

hE

SC

WIB

R3

re-p

rim

ed

NO

D m

ES

CN

OD

mE

SC

12

9 m

ES

C1

29

mE

SC

12

9 m

iPS

C

12

9 m

Ep

iSC

12

9 m

Ep

iSC

NO

D m

Ep

iSC

NO

D m

iPS

CN

OD

miP

SC

NO

D m

Ep

iSC

NO

D m

iPS

CN

OD

miP

SC

C1 hiPSC H1 hESCC2 hiPSC

BGO1 hESC H9 hESCWIS2 hESC

H9 hESCBGO1 hESC

C1 hiPSC

NOD mESCNOD mESC129 mESC129 mESC

129 miPSC

NOD miPSC

NOD miPSCNOD miPSC

c

01

0.8 1.0

–0.6 –0.4 –0.2 0 0.2

0.6 0.4

129 mEpiSC

NOD miPSC

129 mEpiSC

WIS1 hESC

Prim

ed

N

aiv

e

Primed Naive

d Naive WIBR3 hESCs

Primed WIBR3 hESCs

TF

E3

OC

T4

D

AP

IM

erg

e

GFP

WIBR3 hESC

ΔPE–OCT4–GFP

WIBR3 hESC

control

WIBR3 hESC

ΔDE–OCT4–GFP

a WIBR3 hESC

OCT4–GFP

MFI = 317 MFI = 317

MFI = 271 MFI = 271

MFI = 18 MFI = 18

MFI = 19 MFI = 19

MFI = 17 MFI = 17

MFI = 13 MFI = 13

MFI = 295 MFI = 295

MFI = 24 MFI = 24

MFI = 21 MFI = 21

MFI = 24 MFI = 24

MFI = 23 MFI = 23

MFI = 226 MFI = 226

MFI = 317

MFI = 271

MFI = 18

MFI = 19

MFI = 17

MFI = 13

MFI = 295

MFI = 24

MFI = 21

MFI = 24

MFI = 23

MFI = 226

C2

hiP

SC

WIB

R3

hE

SC

80%

72%

2%

Figure 2 | Naive human stem cells share definingmolecular features with mouse ES cells.a, WIBR3 human ES cells were stably transfectedwith the different engineered BAC reporterconstructs indicated. GFP levels were evaluated ingenetically matched cells by flow cytometry. MFI,mean fluorescence intensity. Average MFI values of3 biological replicates per sample are indicated.b, Representative confocal images obtained afterdouble immunostaining for OCT4 and H3K27me3on naive, primed and differentiated samples(obtained from naive cells). Average percentages ofH3K27me3 foci-positive nuclei out of 150–200cells per sample (arrowheads) are indicated. EBs,embryonic body differentiated cells. c, Unbiasedglobal transcriptional cross-species hierarchicalclustering of naive and primed pluripotent cellsfrom mice and humans. Correlation matrix of geneexpression was clustered using Spearmancorrelation and average linkage. Colour barindicates correlation strength. Each row/columnrepresents an independent cell line or clone.mEpiSCs, murine EpiSCs; mESC, mouse ES cell;miPSC, mouse iPS cell. NOD indicates non-obesediabetic ICR mice. d, Naive and primed human EScells were double immunostained for TFE3 andOCT4. Representative confocal images are shownfor WIBR3 human ES cells. Insets are enlargementsof the dashed boxes. Scale bars, 50mm.

LETTER RESEARCH

0 0 M O N T H 2 0 1 3 | V O L 0 0 0 | N A T U R E | 3

Macmillan Publishers Limited. All rights reserved©2013

Fig. 7b). H3K27me3 peaks undergo genomic redistribution and becomepreferentially depleted from promoters and gene-body regions, ratherthan from intergenic regions (Extended Data Fig. 7a, b). We nextmeasured H3K4me1 and H3K27ac marks using ChIP-seq, and globallymapped enhancers of class I and class II in naive and primed pluripo-tency states. Class I enhancers are characterized by the presence ofH3K4me1 and H3K27ac, and the absence of H3K4me3 and H3K27me3marks, and are associated with active expression of nearby genes16. Class IIenhancers, on the other hand, are characterized by the presence ofH3K4me1 and H3K27me3, and the absence of H3K27ac mark, and areassociated with being poised for activation (Extended Data Fig. 7c, e)16.Consistent with the reduction of lineage-priming molecular features innaive pluripotency, we observed a major reduction in the annotatedClass II enhancers in naive cell state, in both human (5.8-fold) andmouse (62.5-fold) (Extended Data Fig. 7c). Finally, reminiscent of recentreports on murine stem cells17, human naive ES cells/iPS cells expandedin NHSM conditions demonstrated a marked downregulation inDNMT3A, DNMT3B and DNMT3L de novo DNA methyltransferaseenzymes, but not in DNMT1 and TET enzymes (Fig. 3c and ExtendedData Fig. 6d). Sampling the DNA methylation states of human naiveand primed pluripotent cells by reduced representation bisulphitesequencing (RRBS) showed a significant reduction (paired sample t-testP value < 0) in CpG methylation in human naive WIBR3 ES cells afterexpansion in NHSM conditions for 17 days (Fig. 3d).

We next characterized and validated the distinct response patternsof naive and primed human pluripotent cells to different signallingstimuli or inhibitors, and compared them to those observed in rodentpluripotent cells (Extended Data Fig. 8a)7. Murine ES cells maintaintheir naive pluripotent state as measured by DPE–OCT4–GFP expres-sion and by all-ES cell mouse formation competence following tet-raploid embryo complementation assay, upon inhibition of either oneof the p38, JNK or ERK pathways (Extended Data Fig. 8a–d). In humannaive cells, simultaneous inhibition of all these pathways is essential formaintaining naive pluripotency (Extended Data Fig. 8a,f and ExtendedData Fig. 6e). Supplementation of TGF-b is necessary to inhibit thepro-neural differentiation effect of ERK inhibition in naive NHSMconditions (Extended Data Fig. 6f). Naive human ES cells are LIF/STAT3-dependent and when stably transfected with dominant-negative

Stat3-encoding transgenes, they rapidly differentiated and could not bemaintained in vitro with NHSM conditions (Extended Data Fig. 8a, g).Alternatively, human naive cells transgenic for a constitutively activeStat3 mutant (Stat3-CA) could be propagated in vitro in the absence ofexogenous LIF (Extended Data Fig. 8g). These findings highlight thatseveral signalling pathways distinctly regulate the two human pluripo-tent states, in a manner similar to that observed in mice3,7.

We next evaluated whether the altered biological features of humannaive PSCs directly endow them with unique or enhanced functionalcharacteristics. Naive H9 human ES cells were significantly more amen-able to gene targeting by homologous recombination with isogenic vectors,in comparison to H9-primed cells (Extended Data Fig. 9a, b). NHSMconditions were important for facilitating up to 100% efficiency inderiving naive human iPS cell formation from MBD3-depleted in vitro-differentiated secondary OKSM transgenic fibroblasts following doxinduction (Extended Data Fig. 9c)18. Consistent with changes in DNAmethylation regulation in naive human pluripotency (Fig. 3d), iPS cellsfrom fragile X chromosome patient specific fibroblasts were able toundergo FMR1 gene reactivation and demethylation after reprogram-ming under naive ground state conditions (Extended Data Fig. 9d, e)19.Finally, consistent with robust E-CADHERIN expression, enhancedsingle-cell survival and corresponding to an earlier developmental state,microinjection of GFP-labelled human naive iPS cells into mouse E2.5morulas showed robust integration in mouse ex vivo-developed innercell masses at E3.5, and higher integration than that of primed plur-ipotent cells (Extended Data Fig. 10a–c). The latter results are consistentwith previously reported near-absence of integration and chimaerismfollowing embryo micro-injection of primate primed ES cells20,21. Mouseblastocysts obtained after microinjection with human naive iPS cellswere implanted in pseudo-pregnant female mice and allowed to developfor 7–8 additional days in vivo, after which they were dissected andimaged in toto under confocal microscopy for GFP1 cell detection.Remarkably, we were able to obtain multiple (n 5 13) mouse embryoscorresponding to E8.5–E10.5 developmental stages that showed chimae-rism with naive human iPS-cell-derived GFP1 cells integrated in orga-nogenesis stages of embryonic development and in different locations,including craniofacial tissues and embryonic neural folds (representativeexamples in Fig. 4, Extended Data Fig. 10d, e and Supplementary Videos

a c

K2

7m

e3

TSS TES –3 Kb +3 Kb

50

40

30

20

10

TSS TES +3 Kb –3 Kb

Developmental genes

Primed PSCs

Average = 8.3 Variance = 16.2

P = 8.6 × 10–61

b Developmental genes

Naive C1 hiPSC

Primed C1 hiPSC

Naive LIS2 hESC

Primed LIS2 hESC

Naive BGO1 hESC

Naive C1 hiPSC Primed C1 hiPSC

Naive LIS2 hESC

Primed LIS2 hESC

Naive BGO1 hESC

Naive mESC

Primed mEpiSC

Naive mESC

Primed mEpiSC

K27me3

K4me3

K27me3

K4me3

Human

Mouse

HOXA1 HOTAIR HOXA9 FOXA2 GATA6 NKX2-5

Hoxa1 HoxA9 FoxA2 Gata6 Nkx2-5

10

60

Read

Den

sity

Human Mouse

Rela

tive

mR

NA

levels

Human naive Human primed

40

20

30

50

Read

den

sity

* *

*

d WIBR3 Naive – WIBR3 primed

Fre

qu

en

cy

10,000

30,000

–1.0 –0.5 0.0 +1.0 +0.5

n = 133868μ = –0.06

σ = 0.19

Human

n = 71476μ = –0.27

σ = 0.37

Naive mESC – primed mEpiSC

Mouse

Methylation (naive – primed)

70

40

70

70

70 70

40 40

40 40

20 20 20

20

Naive PSCs

Average = 20Variance = 63

P = 1.6 × 10–71

1.2

0.9

0.6

0.3

0

Fre

qu

en

cy

10,000

30,000

–1.0 –0.5 0.0 +1.0 +0.5

Methylation (naive – primed)

DNMT1

DNMT3

A

DNMT3

B

DNMT3

LTE

T1

Figure 3 | Ground state epigenetic landscape ofhuman naive pluripotency. a, Profiles ofH3K27me3 chromatin mark over developmentalgenes in human and mouse, represented asnormalized read-density. Blue, naive; red, primed.Human profiles indicate average and s.d. (errorbars) calculated over 5 different cell lines (C1, LIS2,WIBR3, WIBR3-MBD3mut and BGO1). Averagedifference between plots is indicated alongsidevariance and P-values (calculated with paired-sample t-test). b, Chromatin landscape of selected 5developmental regulatory genes. H3K27me3 andH3K4me3 marks are shown for naive (blue) andprimed (red) cell lines, in both human and mouse,showing high consistency between both species.c, Representative relative transcript levels in humanprimed and naive WIBR3 human ES cells after 48 hin NHSM conditions. *t-test P value , 0.01. Errorbars indicate s.d. (n 5 3). d, Histograms of thechange in methylation between primed and naivesamples in human and mouse. The histogramsdepict the distribution of the per-CpG difference inmethylation (naive methylation – primedmethylation), calculated for all differentiallymethylated CpGs residing in CpG-rich regions(.4% CpG content) and having a coverageof $ 310 in both analysed samples. n, the numberof covered, differentially methylated CpGs. m, s,the mean and standard deviation of thedistribution, respectively.

RESEARCH LETTER

4 | N A T U R E | V O L 0 0 0 | 0 0 M O N T H 2 0 1 3

Macmillan Publishers Limited. All rights reserved©2013

1 and 2). In-depth functional and developmental analysis of the in vivo-integrating human naive iPS-cell-derived cells in advanced cross-specieschimaeric mouse embryos is of great future scientific interest.

Our findings substantiate the concept of naive ground state pluri-potency in humans7,8, and indicate that its maintenance requires a uniquecombination of cytokines and small molecule inhibitors. The epigeneticchanges induced by NHSM conditions (Fig. 3) indicate that naive con-ditions are likely to resolve previously described technical phenotypesof epigenetic memory, lineage differentiation biases and aberrant repro-gramming in human iPSCs and ES cells1. Finally, defining a novel naivepluripotent state in humans that is stable and requires no genetic modifi-cations might be relevant for the molecular study of early lineage commit-ment, and for expanding the capabilities for using human iPS cells inregenerative medicine research and disease modelling in vitro and in vivothrough the generation of ‘humanized’ animal models.

METHODS SUMMARYFull details of stem cell lines, cell culture, plasmids, antibodies, embryo-micromanipulation and bioinformatics are provided in Methods.

Online Content Any additional Methods, Extended Data display items and SourceData are available in the online version of the paper; references unique to thesesections appear only in the online paper.

Received 19 May; accepted 10 October 2013.

Published online 30 October 2013.

1. Hanna, J. H., Saha, K. & Jaenisch, R. Pluripotency and cellular reprogramming:facts, hypotheses, unresolved issues. Cell 143, 508–525 (2010).

2. Ying, Q.-L. et al. The ground state of embryonic stem cell self-renewal. Nature 453,519–523 (2008).

3. Marks, H. et al. The transcriptional and epigenomic foundations of ground statepluripotency. Cell 149, 590–604 (2012).

4. Brons, I. G. M. et al. Derivation of pluripotent epiblast stem cells from mammalianembryos. Nature 448, 191–195 (2007).

5. Thomson, J. A. et al. Embryonic stem cell lines derived from human blastocysts.Science 282, 1145–1147 (1998).

6. Chia, N.-Y. et al. A genome-wide RNAi screen reveals determinants of humanembryonic stem cell identity. Nature 468, 316–320 (2010).

7. Hanna, J. et al. Human embryonic stem cells with biological and epigeneticcharacteristics similar to those of mouse ESCs. Proc. Natl Acad. Sci. USA 107,9222–9227 (2010).

8. Hanna, J. et al. Metastable pluripotent states in NOD-mouse-derived ESCs. CellStem Cell 4, 513–524 (2009).

9. Greber, B. et al. Conserved and divergent roles of FGF signaling in mouse epiblaststem cells and human embryonic stem cells. Cell Stem Cell 6, 215–226 (2010).

10. Watanabe, K. et al. A ROCK inhibitor permits survival of dissociated humanembryonic stem cells. Nature Biotechnol. 25, 681–686 (2007).

11. Rajendran,G.et al. Inhibitionofprotein kinaseCsignaling maintains rat embryonicstem cell pluripotency. J. Biol. Chem. 288, 24351–24362 (2013).

12. Okamoto, I. et al. Eutherian mammals use diverse strategies to initiateX-chromosome inactivation during development. Nature 472, 370–374(2011).

13. Mekhoubad, S. et al. Erosion of dosage compensation impacts human iPSCdisease modeling. Cell Stem Cell 10, 595–609 (2012).

14. Chappell, J., Sun, Y., Singh, A. & Dalton, S. MYC/MAX control ERK signaling andpluripotency by regulation of dual-specificity phosphatases 2 and 7. Genes Dev.27, 725–733 (2013).

15. Betschinger, J. et al. Exit from pluripotency is gated by intracellular redistributionof the bHLH transcription factor Tfe3. Cell 153, 335–347 (2013).

16. Rada-Iglesias, A.et al.Auniquechromatin signatureuncovers early developmentalenhancers in humans. Nature 470, 279–283 (2011).

17. Ficz, G. et al. Fgf signaling inhibition in ESCs drives rapid genome-widedemethylation to the epigenetic ground state of pluripotency. Cell Stem Cell 13,351–359 (2013).

18. Rais, Y. et al. Deterministic direct reprogramming of somatic cells to pluripotency.Nature 502, 65–70 (2013).

19. Urbach, A., Bar-Nur, O., Daley,G. Q. & Benvenisty, N. Differential modeling of fragileX syndrome by human embryonic stem cells and induced pluripotent stem cells.Cell Stem Cell 6, 407–411 (2010).

20. James, D., Noggle, S. A., Swigut, T. & Brivanlou, A. H. Contribution of humanembryonic stem cells to mouse blastocysts. Dev. Biol. 295, 90–102 (2006).

21. Tachibana, M., Sparman, M., Ramsey, C., Ma, H. & Lee, H. S. Generation of chimericrhesus monkeys. Cell 148, 285–295 (2012).

Supplementary Information is available in the online version of the paper.

Acknowledgements J.H.H. is supported by a gift from Ilana and Pascal Mantoux, andresearch grants from the European Research Council starting investigator grant(StG-2011-281906, for funding iPS cell experiments only), the Leona M. and HarryB. Helmsley Charitable Trust, the BIRAX (Britain Israel Research and AcademicExchange Partnership), The Sir Charles Clore Research Prize, the Israel ScienceFoundation (Bikura, ICORE (Israeli Centre of Research Excellence) and Regularresearchprogram), the IsraelCancerResearchFoundation, the ERANET E-Rarediseaseprogram, the Benoziyo Endowment fund, Fritz Thyssen Stiftung (used for human iPScell experimentsonly), EMBOyoung investigator program, the Alon Foundation scholaraward, a grant from E. A. and R. Drake, postdoctoral fellowships from ICRF and theWeizmann Dean fellowship award for A.A.M. We thank the embryologists of the RacineIn Vitro Fertilization Laboratory at the Lis Maternity Hospital (A. Carmon, T. Cohen andN.M.Raz) for their skilful assistance.We thank theWeizmann Institutemanagement forproviding financial and infrastructural support.

Author Contributions J.H.H. conceived the idea for this project, designed andconducted experiments, invented and optimized NHSM conditions, and wrote themanuscript with contributions from other authors. L.W. and A.A.M. conducted tissueculture optimization and molecular biology analysis. O.G. and M.Z. conductedmicroinjections. O.G. and R.M. conducted embryo imaging. E.C., Z.S., Z.M. and A.T.conducted RRBS analysis. Y.S.M. analysed gene expression data. N.N. and E.C.analyzed Chip-seq data. J.H.H., Y.K., T.S. and D.B.-Y. applied NHSM conditions onhuman blastocysts. O.G., I.M., I.A., S.Gi., D.A.-Z. and S.B. assisted in ChIP-seqexperiments. S.V. and J.H.H. engineered human stem cell lines. Y.R., V.K., I.C., D.S., S.Ge.assisted and participated in iPS cell related experimental assays. A.Z. conductedautomated imaging analysis.

Author Information Reprints and permissions information is available atwww.nature.com/reprints. The authors declare no competing financial interests.Readers are welcome to comment on the online version of the paper. Correspondenceand requests for materials should be addressed to J.H.H.(jacob.hanna@weizmann.ac.il), N.N. (noa.novershtern@weizmann.ac.il) orR.M. (rada.massarwa@weizmann.ac.il).

E10.5

GF

P C

ellT

racker

G

FP

R1 R2

Ho

ech

st

CellT

racker R1

R2

fba

sss

ss

ss

ss

ss

ss

flb

E10.5

GF

P C

ellT

racker

G

FP

R1 R2

Ho

ech

st

CellT

racker R1

R2

op

fba

ov

ss

ss

ss

ss

ss

ss

flb

Figure 4 | Robust generation of cross-species chimaeric humanized micefollowing naive human iPS cell microinjection into mouse morulas.Representative images showing widespread integration of GFP-labelled humannaive iPS-derived cells into different locations in the anterior part of an E10.5mouse embryo. Hoechst and CellTracker were used for counterstaining. Thefirst column shows the whole embryo (z-stack interval 30mm, 18 focal planes intotal). The second column shows a zoom in images focusing on the head region(white square R1) where the human iPS-derived cells (GFP-positive cells) arepointed out (arrowheads, z-stack interval 20mm, 11 focal planes in total). Thethird column shows the posterior part of the embryo (yellow square R2) whereno GFP-positive cells were detected (z-stack interval 20mm, 9 focal planes intotal). Squares in the first two images in the second column represent the areashown in the insets at the corner of each image. ov, optic vesicle; op, otic pit; fba,first branchial arch; flb, forelimb bud; s, somite. Scale bars, 50mm.

LETTER RESEARCH

0 0 M O N T H 2 0 1 3 | V O L 0 0 0 | N A T U R E | 5

Macmillan Publishers Limited. All rights reserved©2013

METHODSCapturing naive human pluripotent cells. The following conditions, termedNHSM (naive human stem cell medium) conditions (developed at theWeizmann Institute of Science and also referred to as WIS-NHSM conditions),were used to isolate, generate, derive and stabilize naive human pluripotent stemcells (iPS cells and ES cells) with the unique biological properties described in thisstudy. NHSM was generated by including: 475 ml knockout DMEM (Invitrogen),5 g AlbuMAXI (Invitrogen), 5 ml N2 supplement (Invitrogen), recombinanthuman insulin (Sigma, 12.5mg ml21 final concentration), 10mg of recombinanthuman LIF (Peprotech), 8 ng ml21 recombinant bFGF (Peprotech) and 1 ng ml21

recombinant TGF-b1 (Peprotech), 1 mM glutamine (Invitrogen), 1% nonessentialamino acids (Invitrogen), 0.1 mMb-mercaptoethanol (Sigma), Penicillin-Streptomycin(Invitrogen) and small molecule inhibitors: PD0325901 (1mM, ERK1/2i, AxonMedchem); CHIR99021 (3mM, GSK3bi, Axon Medchem); SP600125 (10mM,JNKi, TOCRIS) and SB203580 (10mM, p38i, TOCRIS). After further optimizationwe alternatively used SB202190 (5mM, p38i, Axon Medchem) or BIRB796 (2mM,p38i Axon Medchem) for enhanced p38 inhibition. Y-27632 (5mM, ROCKi AxonMedchem) and protein kinase C inhibitor Go6983 (5mM, PKCi, TOCRIS) can bepermanently used in NHSM conditions and result in positive metabolomics effectand reduction in background apoptosis levels in the cell cultures (Extended DataFig. 1f). Naive human ES cells/human iPS cells were grown on 0.2% gelatin 1

1 ng ml21 vitronectin-coated plates for at least 1 h at 37 uC. Media was replaced every24 h. Cells were passaged by single-cell trypsinization (0.05% trypsin 1 EDTA)every 3–4 days. Passage numbers of naive-human iPS cell/human ES cells indicatenumber of passages counted after induction or stabilization of the naive state (anddo not include previous passages when the cells were established and maintained inconventional/primed ES cell conditions). For transfection of mouse and humannaive iPS cells/ES cells, cells were collected with 0.05% trypsin-EDTA solution(Invitrogen), and cells resuspended in PBS were transfected with 75mg DNA con-structs (Gene Pulser Xcell System; Bio-Rad; 500 V, 25mF, 0.4-cm cuvettes).

While the cells described and analysed in this study were maintained asdescribed above, we note that naive human ES cells/iPS cells can be expandedon gelatin-coated plates together with mitomycin C-inactivated, or gamma irra-diated mouse ICR-DR4 embryonic fibroblast (MEF) feeder cells. When ROCKinhibitors are not continuously included in the NHSM growth conditions,enhanced single cell cloning efficiency can be obtained with NHSM supplementa-tion with ROCK pathway inhibitor (5mM, Y-27632, ROCKi Axon Medchem) for24 h before and after cell passaging. Further, alternative composition of NHSM ispossible by replacing Albumax and N2 mix components with 15–20% knockoutserum replacement (Invitrogen). Instead of adding 5 ml of ready N2 mix, indi-vidual components can be added to 500 ml media bottle at the indicated finalconcentration: (1) recombinant human insulin (Sigma), 12.5mg ml21 final concen-tration; (2) apo-transferrin (Sigma), 500mg ml21 final concentration; (3) progeste-rone (Sigma), 0.02mg ml21 final concentration; (4) putrescine (Sigma), 16mg ml21

final concentration; (5) sodium selenite (Sigma), add 5ml of 3 mM stock solutionper 500 ml NHSM. Supplementation of vitamin C (L-ascorbic acid 2-phosphate,Sigma, 50mg ml21 final concentration) or hypoxic growth conditions (5% O2)were tested and found compatible, but not essential for expanding OCT4–GFP1

naive human ES cells and iPS cells (data not shown).ES cell derivation from human blastocysts. The use of human pre-implantationembryos for ES cell derivation was performed in compliance with protocolsapproved by a Weizmann Institute and LIS hospital ESCRO committees, Lishospital Institutional review committee and Israeli National Ethics Committee(7/04-043) and following the acceptance of a written informed consent. Thecouples’ participation in the study was voluntary after signing informed consentforms and there was no monetary compensation for their embryo donation. Innercell masses were isolated mechanically by laser-assisted micromanipulation fromspare in vitro fertilized embryos, at day 6–7 following fertilization22. The intactinner cell mass clumps were placed on a feeder cell layer of irradiation treated DR4mouse embryonic fibroblasts and cultured in NHSM media. Initial outgrowths ofproliferating ES cells were evident by day 6, and were trypsinized into single cells,6–10 days following inner cell mass plating. The newly established cell lines werefurther propagated by trypsin and then either frozen or used for further analysis.Naive LIS1 and LIS2 are also catalogued as LIS38 and LIS39, respectively, in the LisMaternity Hospital Tel Aviv Sourasky medical centre registry (Tel-Aviv, Israel).Small molecule compounds and cytokines. Small molecules and cytokines weresupplemented as indicated at the following final concentration: JAK inhibitor(420099 JAKi, 0.6mM, CAL BIOCHEM), kenopaullone (KP, 5mM, Sigma Aldrich),PD0325901 (PD,1mM, Axon Medchem); CHIR99021 (CH,3mM,AXONMEDCHEM),Forskolin (FK, 10mM, TOCRIS), FGF-receptor inhibitors PD173074 (0.1mM,TOCRIS) and SU5401 (2 mM, TOCRIS); TGFR/ALK inhibitors A83-01 (1mM,Stemgent), PKC inhibitor Go6983 (PKCi, 5 mM, TOCRIS), ALK inhibitor (ALKi:SB431542, 2mM,TOCRIS); BixO1294 (1mM, Stemgent); BayK8644 (1mM, Stemgent);

SB203580 (5-10mM, p38i, TOCRIS); SBS202190 (5mM, p38i, TOCRIS), BIRB796(2mM Axon Medchem), SP600125 (10mM, JNKi, TOCRIS); recombinant humanBMP4 (5 ng ml21, Peprotech), recombinant human SCF (10 ng ml21, Peprotech),recombinant human IGF1 (10 ng ml21, Peprotech), Media with inhibitors wasreplaced every 24 h. For defining signalling characteristics of human and mousepluripotent cells (Extended Data Fig. 8a) we tested pluripotency maintenance bydifferent pluripotent cell lines cultivated in various media conditions. Cell popula-tions were equally divided and plated on gelatin/vitronectin-coated plates in theindicated growth medium in which these cell lines are normally maintained (mousenaive cells, N2B27-2i/LIF; naive human ES cells/human iPS cells, NHSM; murineEpiSCs and primed human ES cells/human iPS cells, KSR/FGF2/TGF-b1 condi-tions). 36 h later the wells were supplemented with the indicated inhibitors orgrowth factors. After 14 days (2 passages), wells were analysed by OCT4 immuno-staining of direct detection of OCT4–GFP pluripotency reporter expression, todetermine the relative percentage of undifferentiated pluripotent cells. Colonyformation is normalized to an internal control, as indicated by ‘‘Growth medium’’only on the far left column. When components already included in NHSM weresupplemented, this yielded a twofold increase in their relative concentration.Normalized percentages lower than 50% are defined as ‘‘sensitive’’ to the presenceof the supplemented inhibitor.Culture of conventional/primed human ES cells and iPS cells. The followingalready established conventional human ES cells and iPS cell lines were used(indicated passage number of the cell line taken for conversion into naive plur-ipotency is indicated in parentheses): human induced pluripotent stem cells C1(P11) and C2 (P9) human iPS cell lines and the human embryonic stem cell(human ES cell) lines BGO1 (P35) (National Institutes of Health ID code BG01;BresaGen), H1 (P40), H9 (P37), WIBR1 (P13), WIBR2 (P28), WIBR3 (P11).Human ES cells7,23 were maintained in 20% O2 conditions (unless indicated other-wise) on irradiated mouse embryonic fibroblast (MEF) feeder layers or gelatin/vitronectin-coated plates, in human ES cell medium: (425 ml knockout-DMEM orDMEM-F12, Invitrogen) supplemented with 15% knockout serum replacement(Invitrogen), 1 mM glutamine (Invitrogen), 1% nonessential amino acids (Invi-trogen), 0.1 mM b-mercaptoethanol (Sigma), and 8 ng ml21 bFGF (Peprotech)and 1 ng ml21 recombinant human TGF-b1 (Peprotech). Cultures were passagedevery 5–7 days either manually or by trypsinization (24 h pre and 24 h after addi-tion of ROCK inhibitor at 5–10mM concentration). For transfection of human iPScell and human ES cell lines, cells were cultured in Rho kinase (ROCK) inhibitor(Calbiochem) 24 h before electroporation. Primed human ES cell and iPS cellswere collected with 0.25% trypsin-EDTA solution (Invitrogen), and cells resus-pended in PBS were transfected with 75mg DNA constructs (Gene Pulser XcellSystem; Bio-Rad; 250 V, 500mF, 0.4-cm cuvettes). Cells were subsequently platedon MEF feeder layers (DR4 MEFs for puromycin selection) in human ES cellmedium supplemented with ROCK inhibitor for the first 24 h, and then antibioticselection was applied.Mouse naive and primed stem cell lines and cultivation. Murine naive V6.5 EScells (C57B6/129sJae) pluripotent cells were maintained and expanded in serum-free chemically defined N2B27-based media: 500 ml KO-DMEM (Invitrogen),5 ml N2 supplement (Invitrogen), 5 ml B27 supplement (Invitrogen), 1 mMglutamine (Invitrogen), 1% nonessential amino acids (Invitrogen), 0.1 mM b-mercaptoethanol (Sigma), penicillin/streptomycin (Invitrogen), 5 mg ml21 BSA(Sigma). Naive 2i/LIF conditions for murine PSCs included 5mg recombinanthuman LIF (Peprotech). Where indicated 2i was added: small-molecule inhibitorsCHIR99021 (CH, 3mM Axon Medchem) and PD0325901 (PD, 1mM TOCRIS).Murine naive ES cells and iPS cells were expanded on gelatin-coated plates, unlessindicated otherwise. Previously described Eras2/Y ES cells and mice were providedby S. Yamanaka24. Eras-KO ES cells were optimally expanded in N2B27 2i/LIF 1

5% FBS (Biological Industries). Primed 129Jae EpiSC line (derived from E6.5embryos) or C57BL6/129sJae EpiSC line (derived following in vitro priming ofV6.5 mouse ES cells) were expanded in N2B27 with 8 ng ml21 recombinanthuman bFGF (Peprotech Asia) and 20 ng ml21 recombinant human Activin(Peprotech Asia) and 1% KSR. Murine EpiSC were expanded on feeder-free gela-tin/vitronectin- or Matrigel-coated plates. Cell lines were routinely checked formycoplasma contaminations every month (Lonza, mycoalert kit), and all samplesanalysed in this study were not contaminated.Reprogramming of somatic cells and cell infection. Virus-containing super-natants of the different reprogramming viruses: STEMCCA-OKSM polycistronicvector (dox inducible and constitutive expression) was supplemented with theFUW-lox-M2rtTA virus (when necessary) and an equal volume of fresh culturemedium for infection. Human in vitro differentiated secondary fibroblast cellsharbouring doxycycline (dox) lentiviral vectors encoding Oct4, Sox2 and Klf4reprogramming factors and a constitutively active lentivirus encoding the reversetetracycline transactivator18,25, were grown in the presence of dox in NHSM con-ditions on vitronectin/gelatin-coated plates until initial iPS cell colonies were

RESEARCH LETTER

Macmillan Publishers Limited. All rights reserved©2013

observed and subcloned. Generation of BJ1 human iPS cells from human BJforeskin fibroblasts was conducted by OSKM and LIN28 messenger RNA trans-fection kit (Stemgent) according to manufacturer’s instructions, but in NHSMconditions applied starting from day 4 of the reprogramming process. Fragile Xchromosome male patient specific fibroblasts were obtained through Coriellrepository (GM05131 and GM07071). iPS cells were reprogrammed either inconventional or NHSM conditions as indicated in Extended Data Fig. 9d.Differentiation assays of human iPS cells and ES cells. For teratoma formationand analysis, naive human ES cells and iPS cells were collected by trypsinizationbefore injection. Cells were injected sub-cutaneously into 6–8-week-old male NSGmice (Jackson laboratories). Tumours generally developed within 4–6 weeks andanimals were euthanized before tumour size exceeded 1.5 cm in diameter. Allanimal studies were conducted according to the guideline and following approvalby the Weizmann Institute IACUC (approval 00960212-3).Mouse embryo micromanipulation and imaging. Pluripotent mouse ES cellswere injected into BDF2 diploid blastocysts, collected from hormone-primedBDF1 6-week-old females. Microinjection into E3.5 blastocysts placed in M16medium under mineral oil was done by a flat-tip microinjection pipette. A con-trolled number of 10–12 cells were injected into the blastocyst cavity. After injec-tion, blastocysts were returned to KSOM media (Invitrogen) and placed at 37 uCuntil transferred to recipient females. Ten to fifteen injected blastocysts weretransferred to each uterine horn of 2.5 days post coitum pseudo-pregnant females.4n tetraploid complementation assay was performed by fusing BDF2 embryos attwo-cell stage, and subsequently allowing the embryos to develop until the blas-tocyst stage at day 3.5, and were then used for mouse ES cell microinjection.Embryos were allowed to develop into full term. For human naive iPS cell injec-tion, C1 naive iPS cell was constitutively labelled by EGFP (by ZFN-mediatedtargeting of AAVS1 locus with CAGGS-EGFP knock-in reporter). Cells weretrypsinized and microinjected into E2.5 BDF2 diploid morulas (10–15 cells perembryo), and allowed to develop ex vivo until E3.5 (in KSOM supplementedmedium). Up to 20 injected blastocysts were transferred to each uterine horn of2.5 days post coitum pseudo-pregnant females. Embryos were dissected at theindicated time points and imaged for GFP1 cell localization. The latter experi-ments were approved by Weizmann Institute IACUC (00330111-2) and by theWeizmann Institute Bioethics and ESCRO committee. In toto confocal imaging ofchimaeric mouse embryos was conducted as previously described26. Briefly, embryoswere removed from the uterus into Tyrode’s solution, the decidua removed and theembryo carefully isolated with yolk sac intact. Embryos were moved into a dropletof culture media on a paper filter attached to a coverslip with vacuum grease.Starting farthest away from the embryo, the yolk sac was gently peeled open andlightly pressed onto the paper filter, which has adherence qualities, to anchor theembryo to the filter. To expose the tissue of interest, the embryo can be stretchedand lightly pressed onto the filter. After mounting the embryos, a glass-bottomeddish was placed over the embryos using vacuum grease drops for spacing. The dishwas inverted, filled with culture media and placed in the imaging chamber26. Noblinding or randomization was conducted when testing outcome of microinjectionexperiments. Mounted embryos were placed in a heat- and humidity-controlledchamber (37 uC; 5% O2, 5% CO2 and 90% N2 gas mixture) on a motorized stage ofan inverted Zeiss LSM700 confocal microscope. Images were acquired with a 320/0.8 M27 lens with 30.5–0.6 digital zoom-out or 310/0.3 Ph1 M27 lens with 30.5-0.6 digital zoom-out. For GFP detection, a 448-nm laser (30–50% power) was used.CellTracker (red, Molecular Probes, used at 20mM) was excited by a 555-nm laser(30% power). Hoechst (Sigma, used at 10mg ml21) was excited by a 405-nm laser(40% power). Images were acquired at 1,024 3 1,024 pixel resolution. The thick-ness of z-slices appears in the relevant figure legends. Post imaging processing wasdone using Zeiss Zendesk software (Carl Zeiss) and Adobe Photoshop CS4. Non-injected mouse embryos were routinely used as negative controls for signal detection.BAC recombineering and TALEN gene editing. TALEN-expressing plasmidswere designed with a help of TALEN targeter 2.0 and cloned using GoldenGateTALEN kit 2.0 purchased from Addgene according to the published protocol. Fortargeting G we have used NN-type repeat. 107 ES cells/iPS cells were electropo-rated with 30mg of previously described OCT4–GFP–2A–PURO knock-in donorplasmid (provided by R. Jaenisch through Addgene) and 10mg of each of theTALEN expressing plasmids and grown in the presence of antibiotic selection.Resistant clones were isolated and genomic DNA was extracted for Southern blotand PCR analysis. OCT4–GFP–2A–PURO, DPE-OCT4–GFP–2A–PURO, DDE-OCT4–GFP–2A–PURO constructs were made from BAC clone containinghuman OCT4 gene locus by using Red/ET recombination (Gene Bridges). Briefly,GFP–2A–PURO cassette was inserted into translation start site of OCT4, and thengenomic region from 8 kb upstream to 10 kb downstream of TSS was subclonedinto pBS vector. Proximal (PE) and distal (DE) elements were determined byhomology with respective elements of mouse Oct4 described in ref. 27, and basedon results obtained with luciferase test assay7,8. DPE andDDE deletions were made

in wild-type construct by exchanging these regions with kanamycin selectioncassette that was subsequently removed by restriction. Targeting vectors werelinearized with SspI and electroporated into different primed and naive pluripotentcells lines as indicated. For making mouse DPE–OCT4–GFP reporter we havemodified the previously described mGOF18DPE construct27, by pRed/ET recom-bination. The pgk-gb2-Neo cassette (GeneBridges) was inserted into the 39 end ofcloned Oct4 genomic fragment (9.7 kb downstream of Oct4 TSS). The inserted neocassette has the same orientation as Oct4 gene. The construct was linearized withSspI and electroporated into mouse V6.5 ES cells and used for experiments inExtended Data Fig. 8b–e. For gene targeting of human OCT4 and COL1A loci,targeting vectors and strategy were implemented in genetically matched H9 naiveand primed human ES cells as detailed in Extended Data Fig. 9a, b. Homology armswere amplified by PCR from H9 human ES cell genomic DNA enabling examina-tion of isogenic DNA vector targeting efficiency.DNA constructs and plasmid cloning. The pCAG-IresPuro or pCAG-flox-DsRed-IRES- Puro vectors7 encoded the following inserts (which were clonedby either cohesive or blunt-end ligations in XhoI–NotI sites): Stat3-CA mutant(Addgene 13373) (A662C and N664C mutations; by substituting cysteine residuesfor A662 and N664 of the Stat3 molecule, disulphide bonds may form betweenStat3 monomers and render the molecule capable of dimerizing without a phos-phate on Y705); Stat3-Y705F (dominant-negative allele, Addgene 8709). HumanOCT4 enhancer sequences (the Oct4DE- and Oct4PE-SV40-luciferase (Luc) con-structs) were cloned into the pGL3-Promoter Vector (Promega) with the follow-ing primers: 59 hOCT4PE KpnI: 59-GGTACCGGATACTCAGGCCAGGCCCAGAAA-39; 39 hOCT4PE XhoI: 59-CTCGAGTCCACAGACCTCTGGCACT-39;59 hOCT4DE KpnI: 59-GGTACCCATTGAGTCCAAATCCTCTTTACTAGGTG-39; 39 hOCT4DE XhoI: 59-CTCGAGCTGAGGCTCATGCTGCTGG-39. Reporterconstructs were used to determine the regulation pattern of Oct4 expression andwere electroporated into 0.5 3 106–3 3 106 cells along with the pRL-TK (Renilla)vector for normalization. Assays were performed 48 h later using the Dual-GloLuciferase Assay System (Promega). The basal activity of the empty luciferasevector was set as 1.0. Recombinant vitronectin was produced and purified in houseafter obtaining expression construct through Addgene (vector 30225).Reverse transcription and quantitative PCR analysis. Total RNA was isolatedusing TRIzol (Invitrogen). 1mg of DNase-I-treated RNA was reverse transcribedusing a first strand synthesis kit (Invitrogen) and ultimately re-suspended in 100mlof water. Quantitative PCR analysis was performed in triplicate using 1/50 of thereverse transcription reaction in a Viia7 platform (Applied Biosystems). Error barsindicate standard deviation of triplicate measurements for each measurement.RT–PCR primers used herein are: XIST-Forward: AGGGAGCAGTTTGCCCTACT; XIST-Reverse: CACATGCAGCGTGGTATCTT; OCT4-Forward: AGTGATTCTCCTGCCTCAGC; OCT4-Reverse: CTTCTGCTTCAGGAGCTTGG; SOX1-Forward: GGAATGGGAGGACAGGATTT; SOX1-Reverse: AACAGCCGGAGCAGAAGATA; PAX6-Forward: AAGGATGTTGAACGGGCAGA; PAX6-Reverse:TCCGTTGGAACTGATGGAGT; HPRT-Forward: TGACACTGGCAAAACAATGCA; HPRT-Reverse: GGTCCTTTTCACCAGCAAGCT; EOMES-Forward:CGCCACCAAACTGAGATGAT; EOMES-Reverse: CACATTGTAGTGGGCAGTGG; CDX2-Forward: CAGTCGCTACATCACCATCC; CDX2-Reverse: TTTCCTCTCCTTTGCTCTGC; HAND1-Forward: AACTCAAGAAGGCGGATGG;HAND1-Reverse: CGGTGCGTCCTTTAATCCT; GSC-Forward: CGCCTCGGCTACAACAACTA; GSC-Reverse: CGCCTCGGC TACAACAACTA; ID1-Forward:AAACGTGCTGCTCTACGACA; ID1-Reverse: TAGTCGATGACGTGCTGGAG;ID3-Forward: CTACAGCGCGTCATCGACTA; ID3-Reverse: TCGTTGGAGATGACAAGTTCC; ZIC1-Forward: GCGCTCCGAGAATTTAAAGA; ZIC1-Reverse:GTCGCTGCTGTTAGCGAAG; NANOG-Forward: GATTTGTGGGCCTGAAGAAA; NANOG-Reverse: CAGATCCATGGAGGAAGGAA; MIXL1-Forward:AGCTGCTGGAGCTCGTCTT; MIXL1-Reverse: CGCCTGTTCTGGAACCATAC;DNMT3A-Forward: AGTACGACGACGACGGCTA; DNMT3A-CACACTCCACGCAAAAGCAC; DNMT3B-Forward: AGGGAAGACTCGATCCTCGTC; DNMT3B-Reverse: GTGTGTAGCTTAGCAGACTGG; DNMT3L-Forward: TGAACAAGGAAGACCTGGACG; DNMT3L-Reverse: CAGTGCCTGCTCCTTATGGCT.Immunofluorescence staining. Cells were grown for 2 days on glass cover slips(13 mm 1.5H; Marienfeld, 0117530) fixed with 4% paraformaldehyde/phosphatebuffer for 15 min at room temperature, washed three times with PBS, and per-meabilized in PBS/0.1% Triton for 10 min. Cells were blocked with blocking solu-tion (2% normal donkey serum, 0.1% BSA in PBS/0.05% Tween) and incubatedwith primary antibody diluted in blocking solution overnight at 4 uC (antibodies inthis study have all been validated in the literature and by ourselves). Cells were thenwashed three times with PBS/0.05% Tween, incubated with secondary antibodiesfor 1 h at room temperature, washed in PBS/0.05% Tween, counterstained withDAPI, mounted with Shandon Immu-Mount (Thermo Scientific), and imaged. Forstaining of H3K27me3 and nuclear proteins, cells were permeabilized in PBS/0.5%Triton for 30 or 10 min, respective, and 0.1% Triton was included in the blocking

LETTER RESEARCH

Macmillan Publishers Limited. All rights reserved©2013

solution. All comparative experiments were done simultaneously. For MHC class Istaining of human cells, we used anti MHC class I antibody (BE Pharmingen) andcells were analysed on FACS ARIA III analyser and sorter system. The followingantibodies were use at the indicated dilutions: mouse anti-TRA-1-60 (Abcamab16288, 1:500), mouse anti-TRA-1-81 (Abcam ab16289, 1:500), mouse anti-SSEA1 (MC480) (Abcam ab16285, 1:100), mouse anti-SSEA4 (MC813) (Abcamab16287, 1:50), rat anti-SSEA3 (MC631) (Abcam ab16286, 1:50), rabbit anti-Nanog (Bethyl A300-397A, 1:400), rabbit anti-Nanog (Bethyl A300-397A,1:200), rabbit anti-Oct3/4 (H134 clone) (Santa Cruz SC9081, 1:400), mouseanti-Oct4 (C-10 clone) (Santa Cruz SC5279, 1:200), mouse anti-E-cadherin(Abcam ab1416-500, 1:100), rabbit anti-Sox2 (Millipore AB5603, 1:500), rabbitanti-TFE3 (Sigma HPA023881, 1:200), rabbit anti-H3K27me3 (Millipore 07-449,1:4,000), sheep anti-DNMT3B (R&D AF7646, 1:100).Imaging, quantifications and statistical analysis. Images were acquired with D1inverted microscope (Carl Zeiss) equipped with DP73 camera (Olympus) or withZeiss LSM 700 inverted confocal microscope (Carl Zeiss) equipped with 405 nm,488 nm, 555 nm and 635 nm solid state lasers, using a 320 Plan-Apochromatobjective (numerical aperture 0.8). All images were acquired in sequential mode.For comparative analysis, all parameters during image acquisition were kept con-stant throughout each experiment. Images were processed with Zeiss Zenblue2011 software (Carl Zeiss) and Adobe Photoshop CS4. X chromosome inactiva-tion was assayed by presence of a condensed H3K27me3 staining focus per indi-vidual nucleus and quantified using ImageJ software (NIH).Automated and quantitative image analysis for TFE3 localization. Single-cellfluorescence intensity was analysed by profile function in Zeiss Zenblue 2011software. An in house developed automated segmentation protocol segmentsthe cell’s nucleus contained in each image. Nucleus was defined by intersectionby both DAPI and OCT4 fluorescent illumination by a previously describedprotocol18. The segmented cells are then processed individually in order to estim-ate the ratio between intra-nucleus TFE3 intensity and cytoplasmic TFE3 inten-sity. Major steps in the processing: (1) defining a bounding box containing thesegmented nucleus and 10 pixels margin around the nucleus. (2) Estimating theaverage TFE3 intensity within nucleus mask. (3) Estimating the average TFE3intensity within the bounding box but outside of the nucleus mask. Data weremeasured per sample from 200 cells obtained from at least four independent imagefields. Distribution of intensity ratios for all cells is presented using box-plot. Boxplot centres indicate the median value, and box edges indicate the 25th and 75thpercentiles. P-values of distribution differences indicated in the graph were esti-mated with paired sample t-test.Protein western blotting analysis. Whole-cell protein extracts were isolated fromhuman ES cells. Blots were incubated with the following antibodies in 3% BSA/TBST or PBST: pSTAT3 (9318; 1:1,000; Cell Signaling), STAT3 (C-20; 1:1,000;Santa Cruz), pb-Catenin (9561; 1:750; Cell Signaling),b-Catenin (610153; 1:2,000;BD Biosciences), HSP90a (CA1016; 1:5,000; Calbiochem), pJNK (9251; 1:500; CellSignaling), JNK (SC-571; 1:100; Santa Cruz), pP38 (9215; 1:100; Cell Signaling),P38 (9212; 1:1,000; Cell Signaling), pERK (E4; 1:100; Santa Cruz), ERK1,2 (C-14;1:100; Santa Cruz), KLF4 (AF3158; 1:200; R&D), OCT4 (H-134; 1:1,000; SantaCruz), Nanog (397A; 1:1,000; Bethyl). Secondary antibodies were horseradishperoxidase-linked goat anti-mouse, goat anti-rabbit and rabbit anti-goat (1:10,000;Jackson). Blots were developed using ECL (Thermo).DNA methylation analysis. DNA was proteinase-K-treated and extracted, and1mg of DNA was subjected to conversion using the Qiagen EpiTect Bisulphite Kit.Promoter regions of OCT4 were amplified using previously described primers7:XIST forward primer (used on bisulphite treated DNA): TAAATTTTAAATTAATTAAATTAT; XIST reverse primer (used on bisulphite treated DNA): TGTTTTAGAAAGAATTTTAAGTGTAGAGA. PCR products were cloned using thepCR2.1-TOPO vector and sequenced using the M13 forward primer. Primers andmethylation analysis approach for FMR1 promoter were previously described18,28.Reduced representation bisulphite sequencing libraries for genome widesequencing profile of DNA methylation. RRBS libraries were generated asdescribed previously18. Briefly, DNA was isolated from snap-frozen cell pelletsusing the Quick-gDNA mini prep kit (Zymo). Primed WIBR3 human ES celland naive WIBR3 human ES cell (after 17 day reversion in NHSM) conditionswere used for this analysis. For mouse pluripotent cell DNA methylation analysis,we used mouse 129 EpiSCs and naive ES cells in 2i/LIF for 90 days. Isolated DNAwas then subjected to MspI digestion (NEB), followed by end repair using T4PNK/T4 DNA polymerase mix (NEB), A-tailing using Klenow fragment (39R59

exo-) (NEB), size selection for fragments shorter than 500 bp using SPRI beads(Beckman Coulter) and ligation into a plasmid using quick T4 DNA ligase (NEB).Plasmids were treated with sodium bisulphite using the EZ DNA Methylation-Gold kit (Zymo) and the product was PCR amplified using GoTaq Hot Start DNApolymerase (Promega). The PCR products were A-tailed using Klenow fragment,ligated to indexed Illumina adapters using quick T4 DNA ligase and PCR amplified

using GoTaq DNA polymerase. The libraries were then size-selected to 200–500 bpby extended gel electrophoresis using NuSieve 3:1 agarose (Lonza) and gel extrac-tion (Qiagen). Libraries were pooled and sequenced on an Illumina HiSeq 2500system. The sequencing reads were aligned to either the Human Genome Build 37(hg19) or the Mouse Genome Build 37 (mm9) using Bismark. Methylation levelswere calculated for all CpGs that were covered by 10 or more distinct sequencingreads in both the primed and naive samples. For each such CpG, its methylationlevel in the primed sample was deducted from the methylation level in the naivesample, and only CpGs with a non-zero difference were included in the plottedhistograms. Histograms of the per CpG difference in methylation were drawn forCpGs residing in CpG rich areas. CpGs were categorized as belonging to CpG richarea if the number of CpG dinucleotides found within a 500-bp window surround-ing them exceeded 20 (that is, 4% of all nucleotides). The statistical significance ofthe reduction in DNA methylation was assessed using a paired-samples t-test overall CpGs with a coverage of 310 or more.Gene expression data acquisition, processing and analysis. Total RNA wasisolated from indicated primed and naive (in NHSM with 5mM SB203580). Theconcentration of RNA was quantified and subjected to quality control on AgilentBioanalyzer. cDNA was fragmented, labelled and hybridized to Affymetrix HumanGene 1.0 ST GeneChip (Affymetrix). Transcripts levels were processed fromAffymetrix CEL files and CDF file (version V1.r3, which maps probes into 33,252probe sets) using Robust Multi-Array Average method, which corrects for non-biological sample variation using quantile normalization. All analyses were doneusing Matlab (Version R2012b) and its Bioinformatics toolbox, unless otherwisenoted. Data was further filtered to include probes that have at least one call higherthan 64 ( 5 26), and mapped to unique Entrez IDs using Affymetrix annotations(NetAffx Annotation Files version 33.1) and NCBI sites. Probes targeting the samegene are represented by their median, resulting in 13,894 genes. Microarray dataare available at the National Center for Biotechnology Information Gene Expres-sion Omnibus database under the series accession number GSE46872. Additionalhuman ES cell samples data from previously described Affymetrix GeneChipHuman Genome U133 Plus 2.0 Arrays which contain 54,675 probe sets (GSE21222;ref. 7) were passed through the same processing as previously described, resultingin 15,571 genes. The two gene lists were then intersected into one list containing12,071 genes. The batch effect resulting from two different platforms was correctedby normalizing each data set to a sample of the same genotype (WIBR3, referencesamples excluded from further analyses), allowing the two data sets to be united forfurther analysis. The samples were hierarchically clustered using average linkageand either Spearman or Pearson correlation as a distance matrix as noted, withsimilar results. To compare our samples to inner-cell-mass-stage cells, humanpre-implantation data from ref. 29 were included. These data, from AffymetrixHuGene 1.0 st microarrays described before, were passed through the same pro-cessing as previously described, resulting in 16,953 genes before and 12,062 genesafter intersection with the previous gene list. Batch effect was corrected by nor-malization of the new data to the mean values of its ES cell samples, making itconsistent with the normalization of our data. Differentially expressed genes betweennaive and primed samples were found using two-sample two-sided t-test, com-bined with a twofold change expression threshold. Differentially expressed geneswere checked for functional enrichment using the online tool GOrilla (http://cbl-gorilla.cs.technion.ac.il/), with the entire gene list used as background. A falsediscovery rate (FDR)-corrected P-value threshold of 0.05 was used.

Cross-species gene expression analysis was conducted on human arrays describedabove and previously described mouse ES cell and EpiSC gene expression data setson an Agilent 4 3 44 k array platform (GSE15603; described in ref. 8) containing45,018 probes. Mouse data were processed as described above, resulting in 17,885unique genes. Human–mouse orthology was downloaded from MGI (http://www.informatics.jax.org, on April 2013) containing 17,772 pairs of orthologous genes.Of these, 9,803 were mapped to our expression data. The expression values frommouse and human were transformed separately into relative abundance values:For each gene, the relative abundance value is the expression value divided by themean of expression values within the same gene across samples in the same species.The resulting expression matrix was subjected to hierarchical clustering (Spearmancorrelation, average linkage), as was the Spearman correlation matrix of the samples.Coefficient of variance, a measure for noise defined as the standard deviationdivided by the mean, was calculated separately for each of the two species andfor the naive and primed states, using all the samples shown in Fig. 2c. Thestatistical significance between the distributions of coefficient of variance for thetwo states was calculated using one-tail t-test. Outliers were not shown in the plot,but were included in the statistical tests.Chromatin immunoprecipitation and sequencing library preparation. Chromatinimmunoprecipitation followed by deep sequencing (ChIP-seq) was measured forH3K4me3, H3K27me3, H3K4me1, H3K27ac and H3K9me3 in mouse and humanpluripotent cells (ES cells, EpiSCs and/or iPS cells) expanded in fetal bovine free

RESEARCH LETTER

Macmillan Publishers Limited. All rights reserved©2013

and feeder free naive or primed/conventional growth conditions (FBS and feederfree expansion, on gelatin/vitronectin-coated plates). Approximately 40 3 106

cells were cross-linked in formaldehyde (1% final concentration, 10 min at roomtemperature), and then quenched with glycine (5 min at room temperature). Fixedcells were lysed in 50 mM HEPES KOH pH 7.5, 140 mM NaCl, 1 mM EDTA, 10%glycerol, 0.5% NP-40 alternative, 0.25% Triton supplemented with protease inhib-itor at 4 uC (Roche), centrifuged at 950g for 10 min and resuspended in 0.2% SDS,10 mM EDTA, 140 mM NaCl and 10 mM Tris-HCL. Cells were then fragmentedwith a Branson Sonifier (model S-450D) at 24 uC to size ranges between 200 and800 bp, and precipitated by centrifugation. 10mg of each antibody was pre-boundby incubating with Protein-G Dynabeads (Invitrogen100-07D) in blocking buffer(PBS supplemented with 0.5% TWEEN and 0.5% BSA) for 2 h at room temper-ature. Beads were added to the chromatin lysate, and then incubated overnight.Samples were washed 5 times with RIPA buffer, twice with RIPA buffer supple-mented with 500 mM NaCl, twice with LiCl buffer (10 mM TE, 250 mM LiCl, 0.5%NP-40, 0.5% DOC), once with TE (10 mM Tris-HCl pH 8.0, 1 mM EDTA), andthen eluted in 0.5% SDS, 300 mM NaCl, 5 mM EDTA, 10 mM Tris Hcl pH 8.0at 65 uC. Eluate was treated sequentially with RNase A (Roche) for 30 min andproteinase K (NEB) for 2 h, and then incubated at 65 uC for 4 h. DNA was purifiedwith The Agencourt AMPure XP system (Beckman Coulter Genomics, A63881).Libraries of cross reversed ChIP DNA samples were prepared according to a modi-fied version of the Illumina Genomic DNA protocol, as described previously25.Briefly, ChIP DNA was ligated to Illumina adaptors and subjected to 14 cycles ofPCR amplification. Amplified products between 200 and 800 bp were purified byusing magnetic bead size selection. 5 picomoles of DNA library was then applied toeach lane of the flow cell and sequenced on Illumina Hiseq2000 sequencer accord-ing to standard Illumina protocols. The following antibodies were used for chro-matin-IP experiments: control IgG (ChIP grade, ab46540, Abcam), anti-H3K4me3(ab8580, Abcam), anti-H3K27me3 (07-449, Millipore), anti-H3K4me1 (ab8895,Abcam), anti-H3K27acetyl (ab4729, Abcam), anti-H3K9me3 (ab8898, Abcam).Chromatin IP sequencing data analysis. The chromatin markers H3K27me3,H3K4me3, H3K4me1, H3K27Ac and H3K9me3 were measured in the differenthuman pluripotent cell lines: C1, WIBR3, LIS2 (naive and primed), BGO1 andWIBR3-MBD3mut (naive)18. WIS2 naive and primed hES cells were profiled for theH3K9me3 mark only. In addition, H3K27me3, H3K4me3, H3K4me1 and H3K27Acwere measured in mouse V6.5 naive and 129 primed EpiSCs. H3K27me3 andH3K4me3 measurements in naive mouse ES were previously published by ourgroup25. Each sample was accompanied by control sequencing experiment ofwhole-cell extract input. We used bowtie software version 1.0.0 to align humanreads to human reference genome hg19 (UCSC, February 2009) and mouse readsto mouse mm9 reference genome (UCSC, July 2007). We only considered readsthat were uniquely aligned to the genome with up to a single mismatch, taking thesingle best match of each read. To rule out sequencing depth bias, the alignedsequences were down-sampled such that all samples had the same number ofaligned reads. Human samples of the marks H3K4me3, H3K27me3, H3K4me1and H3K27Ac, as well as whole-cell extract, were down-sampled to include 3,750,000aligned reads. H3K9me3 samples were down-sampled to 5,900,000 aligned reads.Mouse samples of the marks H3K4me3, H3K27me3, H3K4me1 and H3K27Ac,were down-sampled to 3,420,000 aligned reads.

Chromatin profiles (Fig. 3a and Extended Data Fig. 4e) were calculated over allRefSeq genes (n 5 43,463), and over developmental genes in the following way: (1)read densities were calculated between 3 kb upstream to TSS, and 3 kb down-stream to TES. (2) Each gene was divided to 100 bins of identical size, and thesum of reads in each bin was calculated. (3) Average profile was calculated over allgenes, where the gene body, which is of changing size, was represented by 100quantiles. Genes of size less than 1 kb were filtered out. (4) Profiles of humansamples represent mean and s.d. (error bars) of primed and naive samples. Lastly,

developmental genes were selected if they have a GO annotation that is related todevelopment or differentiation. Using this criterion, we had 5,922 RefSeq humandevelopmental genes, and 420 RefSeq mouse developmental genes. Concreteexamples of genes (Fig. 3b and Extended Data Fig. 7b) were processed and visua-lized using IGV software version 2.0.

To measure the distribution of H3K9me3 accumulation in chromosome Xgenes (Extended Data Fig. 4c), RPKM (reads per kilobase per 5.9 million reads)was calculated for each gene (between 1 kb upstream to TSS and TES). P-valuesbetween distributions were calculated with one-tail paired-sample t-test.

Enhancers were detected following the guidelines set in ref. 16. Shortly, enhan-cers of type one are genomic intervals that contain H3K4me1 and H3K27Acmarks, do not contain H3K4me3 or H3K27me3 marks and are at least 500 bpaway from any TSS. Enhancers of type two are genomic intervals that containH3K4me1 and H3K27me3 marks, do not contain H3K27Ac mark, and are at least500 bp away from any TSS. To find those enhancers we first identified enrichedintervals of the marks above using MACS version 1.4.1. We used sequencing ofwhole cell extract as control in order to define a background model. Duplicatereads aligned to the exact same location are excluded by MACS default config-uration. Enriched intervals (peaks) that overlap by at least 1 bp were considered asoverlapping, and their union was defined as the enhancer interval, unless at least10% of it overlapped with any of the excluded marks (for example, H3K27me3 inthe case of type one enhancers) or with TSS. To calculate the number of enhancersin human samples (Extended Data Fig. 7c) we only considered enhancers that arecommon to 2 primed cell lines C1 and WIBR3, or common to 4 naive cells lines:C1, WIBR3, WIBR3-MBD3mut and BGO1. The expression level of genes assoc-iated with enhancers (Extended Data Fig. 7d) was calculated by mapping eachenhancer to the nearest gene (as long as it is at most 100 kb away), converting it toEntrez Gene and taking its expression value from the raw gene expression data setdescribed above.

To detect bivalent genes we identified enriched intervals of H3K4me3 andH3K27me3 using MACS. Enriched intervals were mapped to genes if they over-lapped a 3 kb interval before their transcription start sites (TSS) and after theirtranscription end site (TES). Genes marked as bivalent if they had both H3K27me3and H3K4me3 intervals. In human samples we required that the genes are bivalentin at least 2 cell lines. Distribution of H3K27me3 mark in genomic components(Extended Data Fig. 7a) was calculated by counting the number of H3K27me3peaks in each component: promoter (6 kb symmetric interval around TSS), genebody (between 3 kb downstream to TSS and TES), and intergenic (outside pro-moter and gene body). The peaks were divided to bins according to their height,calculated as RPKM.

22. Ben-Yosef, D. et al. Female sex bias in human embryonic stem cell lines. StemCells Dev. 21, 363–372 (2012).

23. Lengner, C. J. et al. Derivation of pre-X inactivation human embryonic stem cellsunder physiological oxygen concentrations. Cell 141, 872–883 (2010).

24. Takahashi, K., Mitsui, K. & Yamanaka, S. Role of ERas in promoting tumour-likeproperties in mouse embryonic stem cells. Nature 423, 541–545 (2003).

25. Mansour, A. A.et al. TheH3K27 demethylase Utx regulates somaticand germcellepigenetic reprogramming. Nature 488, 409–413 (2012).

26. Massarwa, R. & Niswander, L. In toto live imaging of mouse morphogenesis andnew insights into neural tube closure. Development 140, 226–236 (2013).

27. Yeom, Y. I. et al. Germline regulatory element of Oct-4 specific for the totipotentcycle of embryonal cells. Development 122, 881–894 (1996).

28. Pietrobono, R. et al. Quantitative analysis of DNA demethylation andtranscriptional reactivation of the FMR1 gene in fragile X cells treated with5-azadeoxycytidine. Nucleic Acids Res. 30, 3278–3285 (2002).

29. Vassena, R. et al. Waves of early transcriptional activation and pluripotencyprogram initiation during human preimplantation development. Development138, 3699–3709 (2011).

LETTER RESEARCH

Macmillan Publishers Limited. All rights reserved©2013

Extended Data Figure 1 | Defining and optimizing conditions for capturingtransgene-independent human naive pluripotent cells. a, qRT–PCR analysisfor relative expression of endogenous OCT4 pluripotency marker and SOX1neural marker. Relative values to those measured in the primed C1.2 iPS cellline are shown. In the presence of dox, the cells express the endogenouspluripotency OCT4 (consistent with previously described results and teratomaformation ability of these cells7); however, SOX1 is significantly upregulated,consistent with unstable pluripotency maintenance in these conditions7.NHSM conditions optimized at the Weizmann Institute of Science (alsoreferred to as WIS-NHSM conditions) maintain OCT4 expression and do notacquire aberrant SOX1 upregulation. Error bars indicate s.d. (n 5 3).*Student’s t-test P value , 0.05 relative to control. b, The percentage of OCT4–GFP1 colonies obtained in different combinations of indicated factors asmeasured at day 14. (‘2’ indicates without, ‘1’ indicates with.) Error barsindicate s.d. (n 5 3). *t-test P value, 0.01. c, Cells were expanded in the presenceof FGF2/TGF-b1 and pool 1 factors, and the essential factors required formaintaining OCT4–GFP1 cells in the absence of dox were determined by

screening for loss of GFP upon withdrawal of these components (green font).*t-test P value , 0.01. Error bars indicate s.d. (n 5 4). d, OCT4–GFP1 levels fornaive pluripotent cells expanded in NHSM on tissue culture plates coated withthe indicated components. Naive human iPS cells can be grown without feedercells on Matrigel-coated plates, vitronectin-coated plates or 0.2% gelatin 1

1 ng ml21 vitronectin-coated plates. Error bars indicate s.d. (n 5 3). *t-testP value , 0.01 relative to control. e, For further optimizing naive growthconditions, 1% Albumax can be substituted by using either 15% KSR or 1%chemically defined lipid concentrate (CDLC, Invitrogen). Error bars indicates.d. (n 5 3). *t-test P value , 0.01. This suggests the importance of lipidcomponents in maintaining human pluripotent cells in vitro. f, Phase images ofhuman naive ES cells in the indicated conditions for 4 passages (P).g, Representative images showing embryo-derived naive human ES cells inNHSM conditions (with or without feeder cells). Notably for all lines, naivehuman ES cells were transferred to primed conditions at passage 6, and cellswere passaged side by side for comparative analysis as indicated throughout thestudy.

RESEARCH LETTER

Macmillan Publishers Limited. All rights reserved©2013

Extended Data Figure 2 | Validation of human naive pluripotency.a, Karyotype analysis results indicating normal karyotypes of different naivehuman ES cells and human iPS cells. The passage number at which cells werecollected for karyotyping is indicated. b, The newly derived naive LIS1, LIS2,WIS1, and WIS2 human ES cell lines established from human inner cell mass inNHSM conditions were analysed, and showed a strong uniform staining for allindicated human pluripotency markers (OCT4, NANOG, SSEA3, SSEA4,TRA1-60, TRA1-81 and SOX2). Notably, SSEA1, which is specific for mouse

(both naive and primed stem cells) and not human pluripotent cells, is notexpressed on naive human ES cells. Similar results were validated for all humannaive PSCs derived in this study (data not shown). c, In vivo teratomadifferentiation of naive human ES cells and iPS cells. Naive pluripotent stem celllines were expanded in NHSM conditions for the indicated number of passages.All tested lines gave rise to well differentiated mature teratomas with cells fromthe three germ lineages: endoderm, ectoderm and mesoderm. P indicatespassage number in NHSM conditions at which cells were collected and injected.

LETTER RESEARCH

Macmillan Publishers Limited. All rights reserved©2013

Extended Data Figure 3 | Growth properties of human naive PSCs.a, Population doubling time of various mouse and human pluripotent stem celllines. After plating each cell line in replicates, cells were collected at days 2, 4 and6 and their growth rate was normalized by the number of cells collected andcounted at day 2. Error bars represent s.d. (n 5 3), and P values , 0.01represent t-test for the average difference between primed human ES cell/human iPS cell lines and naive human ES cell/human iPS cell lines. b, Single-cellcloning efficiency of different pluripotent stem cell lines as determined by thenumber of wells containing NANOG1 colonies 6 days after plating (with orwithout ROCK inhibitor as indicated in the figure). *t-test P values , 0.01 areindicated. Error bars indicate s.d. (n 5 3). c, Cross-species differences in ERASprotein expression influence remaining growth-property differences betweengenetically unmodified human and mouse naive PSCs. Single-cell cloningefficiency of different pluripotent stem cell lines as determined by the numberof wells containing Nanog1 colonies 6 days after plating (without using ROCKinhibitor). *t-test P value , 0.05 are indicated. Error bars indicate s.d. (n 5 3).These results highlight a relatively compromised single cell survival and cloning

efficiency of Eras knockout naive mouse ES cells, in comparison toconventional ES cells/iPS cells. Reconstitution of ERAS in naive human ES cellsrescues their relative deficiency in single cell colony formation assay.d, Predominant utilization of distal OCT4 enhancer element in mouse andhuman naive pluripotency. Evaluation of human OCT4 transcriptionalregulation by the activity of distal or proximal enhancer reporter gene (byluciferase reporter assay) in the indicated pluripotent cell lines. Baseline activitywas analysed by transfection with an empty vector. Predominant utilization ofdistal enhancer is significantly and dramatically upregulated in human ES cells/human iPS cells grown in naive NHSM conditions. Error bars indicate s.d.(n 5 3). *t-test P value , 0.01 between the matched naive and primed samples.e, Schematic diagram demonstrating the relative location of the evolutionaryconserved distal and proximal enhancer elements in the human OCT4 locus.Distal enhancer (DE) and proximal enhancer (PE) are highlighted in thehuman OCT4 locus, and were accordingly deleted in theDPE andDDE reporterconstructs used in this study. Insertion site of GFP-2A-Puromycin (Puro)resistance cassette by BAC recombineering is indicated.

RESEARCH LETTER

Macmillan Publishers Limited. All rights reserved©2013

Extended Data Figure 4 | Epigenetic properties of human naive groundstate pluripotency. a, Fraction of cells with nuclear H3K27me3 foci. Confocalimages, obtained after double immunostaining for OCT4 and H3K27me3 onnaive, primed and differentiated samples, were analysed and quantified for thefraction of cells with nuclear H3K27me3 foci that mark the inactive X allele.Average percentages of 150–200 individual cells counted per sample fromindependent frames are shown. Error bars indicate s.d. (n 5 10). *t-testP value , 0.01 when comparing naive pluripotent cells and their deriveddifferentiated cells. All 5 naive female pluripotent cells are predominantlydevoid of H3K27me3 foci. Male lines do not exhibit H3K27me3 foci/clouds inany of the states as expected. Differentiated fibroblasts from naive cell linesuniformly acquired H3K27me3 foci. In most (3 out of 5) primed human ES/human iPS cell lines tested, H3K27me3 foci (one per nucleus) became clearlydetectable in the majority of cells consistent with XaXi configuration. Twosubcloned primed cell lines (C2 human iPS cell and WIBR2 human ES cell)showed low H3K27me3 foci abundance, consistent with XaXa or XaXeconfigurations known to also be acquired in human primed cell lines13. b, qRT–PCR analysis indicates no/low expression levels of XIST in naive hPSCs, incomparison to female differentiated fibroblast cells that upregulate XISTexpression. Error bars indicate s.d. (n 5 3). c, Distributions of H3K9me3

RPKM (reads-per-kilobase-per-million reads) levels in chromosome X genes,measured in four different human cell lines. Boxes, 25th and 75th percentiles;horizontal lines, median; crosses, outliers. RPKM were measured for each geneon the region starting 1 kb upstream of the TSS and ending at the TES. LinesWIBR3, C1 and LIS2 are female and WIS2 is a male cell line. Distributions ofH3K9me3 in female primed cells are significantly higher compared to theirnaive counterparts, whereas in male cells they are the same. P values werecalculated with one-tail paired-sample t-test. NS, not significant. d, Bisulphitesequencing analysis of six CpG sites in single clones of an XIST promoteramplicon is shown. Note that in female human naive cells both alleles aredemethylated, whereas in primed cells one of the alleles becomes methylated.Methylation of OCT4 locus is shown for comparison. e, Global H3K27me3 andH3K4me3 deposition in naive and primed pluripotent cells. Profiles ofH3K27me3 and H3K4me3 chromatin modifications of all RefSeq genes inhuman (n 5 43,463) and mouse (n 5 30,480), naive (blue) and primed (red),represented as normalized read-density. Human profiles indicate average ands.d. (error-bars) calculated over 5 different cell lines (C1, LIS2, WIBR3, WIBR3-MBD3mut and BGO1). Average difference between plots is indicated alongsidevariance and P values (calculated with paired-sample t-test).

LETTER RESEARCH

Macmillan Publishers Limited. All rights reserved©2013

Extended Data Figure 5 | Distinct transcriptome for human naive ES cells/iPS cells. a, Genome-wide (n 5 12,071) transcriptional profile of the indicatedprimed and naive human ES cell/human iPS cell lines measured by Affymetrixmicroarrays and hierarchically clustered using Pearson correlation. Heat mapshows row-normalized expression levels (log2 scale) with red and green coloursrepresenting up and downregulated genes, respectively. b, Transcriptionalprofile of selected pluripotency and lineage-specific marker genes, their meanexpression ratio in primed human ES cells and naive human ES cells relative tothe median of all samples is presented. Values are shown in natural scale. Errorbars represent s.e.m of each gene. Asterisks denote statistically significantdifferentially expressed genes in which the false discovery rate was , 0.05between the naive and primed groups of samples. c, GO (gene ontology)categories significantly enriched for genes downregulated in naive compared toprimed human cell lines, with their (–log10) FDR-corrected P-value.

The respective mouse fold-enrichment values for the categories that are alsosignificant in mouse are indicated. d, Transcriptional comparison of in vitro-and in vivo-isolated human pluripotent cells. Hierarchical clustering of themean expression profile of differentially expressed genes between naive andprimed samples (FDR , 0.05), in the different groups (naive (this study),primed (this study), Belmonte’s primed ES cells, and human inner cell masses(Belmonte and colleagues29)), using Spearman correlation. Note that whileprimed ES cell samples previously derived by Belmonte and colleagues clusterwith the primed samples derived herein, human inner cell mass samples clusterwith the naive samples expanded in NHSM conditions. e, Box plot showing themedian and quartiles of the distributions of the coefficients of variance in geneexpression (s/m) calculated for each of the 9,803 homologous genes,throughout all the samples shown in Fig. 2c. The naive distributions aresignificantly smaller (one-tail t-test, P < 0) than the primed ones.

RESEARCH LETTER

Macmillan Publishers Limited. All rights reserved©2013

Extended Data Figure 6 | Transcriptional and molecular features of humanground state naive pluripotency. a, Histogram of surface expression of MHCclass I using FACS analysis on the indicated naive, primed and differentiatedcell lines. These results show that naive human ES cells and human iPS cellsexpress rudimentary or low levels of MHC class I on their cell surface, whereasprimed cells upregulate MHC class I expression to intermediate levels (incomparison to differentiated 293 cells that express high levels of MHC class I).These results are identical to patterns observed in murine naive and primedcells7. b, Representative confocal immunostaining images for the expression ofE-CADHERIN and OCT4 on genetically matched naive and primed human EScells. The samples were processed simultaneously and analysed under identicalconditions. Insets represent enlargements of boxed areas. WhereasE-CADHERIN is expressed in primed human ES cells, its expression becomeshomogenously distributed and more enhanced in naive human ES cellsexpanded in NHSM conditions. c, Box and whisker plots of nuclear/cytoplasmic TFE3 ratios in naive and primed mouse and human ES cells areshown. Quantitative unbiased imaging analysis for preferential nuclearlocalization was conducted on randomly selected 200 cells from independentimage frames per sample. Naive human ES cells showed distributions similar tothose in naive mouse ES cells, and the nuclear enrichment was lost in primed

human and mouse ES cells. *t-test P values , 1 3 102100. d, Representativeimmunostaining for OCT4 and DNMT3B in human primed and naivepluripotent cells, showing DNMT3B down regulation in the human naiveground state of pluripotency. This effect was highly dependent on efficient p38inhibition (data not shown). e, Representative western blot analysis for totaland phosphorylated levels of the indicated proteins in genetically matchedWIBR3 naive and primed human pluripotent cells (passage 25). The resultsindicate blocked and reduced activity for ERK1/2, p38 and JNK in naivepluripotency. Consistent with the presence of LIF in NHSM, phosphorylatedSTAT3 levels accumulate in naive human ES cells. Samples were loaded intriplicates for demonstrating consistency. Similar results were obtained whenanalysing naive and primed C1 human iPS cells and WIS2 human ES cells (datanot shown). SB202190 was used in NHSM conditions when p38 MAPKphosphorylation was tested. f, Naive WIS1 human ES cells expanded in NHSMconditions were expanded after omitting the indicated factors from NHSMconditions for 72 h. qRT–PCR of lineage commitment genes is shown. Theresults indicate that TGF-b1 blocks the SOX1 inductive effect induced bythe presence of ERKi in the medium. Error bars indicate s.d. (n 5 3).*t-test P value , 0.05.

LETTER RESEARCH

Macmillan Publishers Limited. All rights reserved©2013

Extended Data Figure 7 | Genome-wide redistribution of H3K27me3 inhuman naive pluripotency. a, Distribution of H3K27me3 peaks in differentgenomic components (promoter, gene-body and intergenic region), as afunction of the peak density (represented as RPKM, read per kilobase permillion reads), in human cell lines (WIBR3, C1 and BGO1). In addition to areduction in total number of peaks in naive conditions, the peaks distributedifferently, with a lower number of peaks in promoters and gene bodies,compared to primed conditions. b, Chromatin landscape examples of 5pluripotency genes. H3K27me3 and H3K4me3 marks are shown for naive(blue) and primed (red) cell lines, in both human and mouse, showing highconsistency between the organisms. c, Number of enhancers of class I (activetranscription) and class II (poised) in naive versus primed, human and mouse.For human, numbers correspond to common enhancers in lines WIBR3 andC1. The number of reads was normalized by down-sampling such that the

number of reads in all human samples is identical, and so is the number of readsin all mouse samples. A marked decrease in class II enhancers can be seen in thenaive cells of both human and mouse. d, Expression level distribution of genesassociated with class I and class II enhancers, present in either naive cells only(blue) or in primed cells only (red). Expression level distribution of all genes ispresented as control. Boxes represent 25th and 75th quantiles, and whiskersrepresent minimum and maximum values that are not outliers. Analysis showsthat primed state specific class II enhancers continue to be largely repressed inboth naive and primed states (despite the loss of H3K27me3 mark in naivepluripotency over their enhancer). e, Selected GO annotation enrichment(from top 25 categories obtained) for human primed specific class II (poised)enhancer marked genes. This analysis shows that class II enhancer associatedgenes are highly enriched for transcription factors (5.9 3 10277) anddevelopmental genes (5.5 3 10249).

RESEARCH LETTER

Macmillan Publishers Limited. All rights reserved©2013

Extended Data Figure 8 | Distinct signalling requirements for maintainingnaive pluripotency across species. a, Pluripotency maintenance by differentpluripotent cell lines cultivated in various media conditions. Cross-speciesfunctionally divergent pathways are highlighted in grey. b, Different definedgrowth conditions enable maintenance of mouse naive pluripotency. V6.5mouse ES cells carrying the naive pluripotency specific DPE–OCT4–GFPreporter were maintained in distinct chemically defined growth conditions asindicated in the panel labels. The analysis indicates that naive GFP1 mouse EScells can be maintained in LIF1GSK3bi, together with Erk1/2i, p38i or JNKi.c, Representative images of high-contribution chimaeras generated afterblastocyst microinjection of the indicated lines. Agouti coat colour indicateshigh-level chimaera contribution. d, Competence of different mouse naivegrowth conditions in tetraploid embryo complementation assay. V6.5 naivemouse ES cells were expanded in the indicated conditions for 12 passages, afterwhich they were tested for tetraploid complementation assay to form all-ES cellanimals. 135–160 mouse embryos were injected per condition tested. Theanalysis indicates unrestricted developmental potential is equivalently obtained

at the functional level when naive mouse ES cells are expanded with p38i, JNKior ERK1/2i (together with supplementation of LIF and GSK3bi in allconditions). e, Representative images of all-ES cell agouti coat coloured animalsobtained. f, Different human ES cells and human iPS cell lines carrying OCT4–GFP or DPE–OCT4 GFP reporter were used to evaluate different previouslypublished conditions (as indicated in the figure panel) for their ability to maintainpluripotency and specifically naive pluripotency (by DPE-Construct). OnlyNHSM conditions enabled maintenance of all cell lines tested while uniformlyactivating both OCT4–GFP and DPE-OCT4–GFP reporter. *Student t-test Pvalues , 0.01 between indicated compared samples/groups. Error bars indicates.d. (n 5 3). g, LIF/STAT3 signalling dependence by naive human ES cells/iPScells. Naive human ES cells/iPS cells carrying a constitutively active Stat3 mutant(Stat3-CA) remain pluripotent in the absence of exogenous LIF from NHSM,based on immunostaining for pluripotency marker expression. Error bars indicates.d. (n 5 3). These results show that naive human ES cells/iPS cells describedherein share defining features with rodent naive PSCs in their dependence onLIF/STAT3 signalling to remain pluripotent in NHSM conditions.

LETTER RESEARCH

Macmillan Publishers Limited. All rights reserved©2013

Extended Data Figure 9 | Distinct and enhanced functional properties ofhuman naive pluripotent cells. a, Targeting strategy by homologousrecombination of COL1A locus in H9 ES cells. Correct targeting efficiency indifferent H9 human naive and primed pluripotent cell experimental replicatesis shown. Correct targeting was scored after confirmation with both 59 and 39

Southern blot analysis on extracted DNA from analysed clones. *Indicatessignificant P value , 0.05. b, as in a, but for OCT4 locus. *Indicates significantP value , 0.05. c, NHSM conditions facilitate deterministic reprogramming ofMBD3-depleted human secondary fibroblasts. MBD3wt and MBD3mut iPS cellscarrying dox-inducible OKSM transgene were labelled with constitutivelyexpressed mCherry and targeted with an OCT4–GFP knock-in allele18. In vitrodifferentiated fibroblasts from the latter lines were reprogrammed as indicatedin the scheme either in primed or in naive growth conditions. Secondary iPS cell

reprogramming efficiencies of wild-type and MBD3-depleted (MBD3mut) cellswhen reprogramming in NHSM (blue scheme) or primed (red scheme)conditions are indicated. Error bars indicate s.d. (n 5 3). *t-test P value , 0.01.Deterministic up to 100% iPSC reprogramming efficiency of primaryhuman differentiated cells (directly obtained from human samples, rather thanbeing differentiated in vitro from pluripotent cells) remains to be determined.d, Relative transcript expression of FMR1 and OCT4 genes in cells isolatedfrom healthy or fragile X patients. Error bars indicate s.d. (n 5 3). Note thatnaive fragile X patient derived iPS cells reactivate FMR1 gene expression.e, Methylation analysis of FMR1 promoter region on DNA samples obtainedfrom wild-type and fragile X patient derived cells. ***Indicates student t-testP value , 0.01.

RESEARCH LETTER

Macmillan Publishers Limited. All rights reserved©2013

Extended Data Figure 10 | Chimaerism with human naive iPS-cell-derivedcells following mouse morula microinjection. a, C1 human naive iPS cellswere targeted with constitutively CAGGS promoter driven EGFP into thehuman AAVS1 locus via ZFN utilization. Subsequently, cells weremicroinjected into E2.5 mouse early morulas, and micro-manipulated embryoswere allowed to recover and develop to blastocyst stage ex vivo for an additional24 h. Images show specific GFP1 human cell survival and integration in mousepre-implantation embryos. b, Naive and primed GFP1 human iPS cell survivaland integration ex vivo, 24–36 h after microinjection into mouse morulas.*t-test P value , 0.01. Error bars indicate s.d. (n 5 3). c, Representativeconfocal analysis following immunostaining for GFP (green), OCT4 (red) andCDX2 (magenta) was done 24 h following mouse morula microinjections withGFP-labelled naive human iPS cells. Note surviving GFP1 cells (white arrow)that specifically integrate and stain positive for OCT4, but not CDX2 (no co-localization between GFP and CDX2 was observed). d, Subsequently, mouse

blastocysts were allowed to develop in vivo, following implantation in mice, foran additional 7–8 days (as indicated), before dissection and confocal analysis.Representative images showing robust integration of human iPS-derived cellsinto the neural folds of an E8.5 mouse embryo (upper panels). Notably, GFPwas not detected in the control non-injected mouse embryos (n 5 3, lowerpanels). z-stack interval 20mm; 17 focal-planes total (nf, neural folds). e, Naivehuman iPS-cell-derived cells are integrated into different locations at thecraniofacial region of an E10.5 mouse embryo. A series of different focal planesfrom Fig. 4 and Supplementary Videos 1 and 2, showing multiple human iPS-cell-derived cells integrated into different locations on the craniofacial region(arrowheads). First image in each row shows the maximum intensity projectionof all z-stacks. The distance between the different focal planes appears on theupper right corner of each one. fba, first branchial arch; op, otic pit; ov, opticvesicle.

LETTER RESEARCH

Macmillan Publishers Limited. All rights reserved©2013