Identification of the missing pluripotency mediator downstream of leukaemia inhibitory factor

1Early mammalian embryos are characterized by the presence of a regulative population of cells each with the capability of giving rise to all somatic lineages and to germ cells. This property, pluripotency, first emerges in a naïve form in the epiblast in the pre-implantation blastocyst. When exposed to an appropriate environment ex vivo, mouse naïve epiblast cells can be expanded indefinitely as embryonic stem (ES) cells. The cytokine leukaemia inhibitory factor (LIF) potently promotes ES cell self-renewal, and is routinely used in the derivation and culture of mouse ES cells. Binding of LIF to the gp130/LIF-R complex leads to activation of JAK kinases that, in turn, phosphorylate the transcription factor Stat3. Phosphorylated Stat3 dimerizes, enters the nucleus, and activates the expression of target genes. Artificial activation of Stat3 is sufficient to sustain ES cell self-renewal in the absence of LIF, whereas antagonism of Stat3 leads to differentiation. These findings indicate that Stat3 is the key mediator of LIF action in ES cells. However, LIF activates via JAK other signalling pathways that have also been proposed to play a role in ES cell maintenance.

2LIF/Stat3 signalling also plays a key role in the conversion of primordial germ cells into pluripotent EG cells and facilitates transcription factor directed reprogramming. Hyperactivation of Stat3 potently enhances reprogramming of somatic cells into induced pluripotent stem (iPS) cells, while blockade of the LIF/Stat3 pathway abolishes iPS cell generation. Furthermore, post-implantation epiblast stem cells, EpiSCs, can be reprogrammed to a naïve pluripotent state simply by transient activation of Stat3. These actions are separable from the self-renewal effect of LIF/Stat3 on established pluripotent stem cells.

3Despite the central role ascribed to Stat3 in mouse pluripotency, its downstream effectors are incompletely described. LIF does not directly regulate core pluripotency factors Oct4, Sox2, Nanog, or Esrrb. Efforts to delineate the transcriptional programme stimulated by LIF have identified transcription factor targets of Stat3, such as Klf4, Pim1, and Gbx2. However, none of these factors are indispensable for LIF responsiveness, nor can their forced expression fully recapitulate LIF activity. Notably, Klf4 is one of the four canonical Yamanaka factors that direct somatic cell reprogramming, but LIF is required in addition. These observations suggest either, functional redundancy and additive effects between multiple Stat3 targets, or alternatively the existence of a pivotal unidentified target.

4We previously showed that two selective small molecule inhibitors (2i) of Gsk3 and Mek kinases eliminate ES cell differentiation and can sustain self-renewal in the absence of LIF. Furthermore, 2i allows derivation and expansion of Stat3−/− ES cells. Here, we exploited these null cells in a refined search for critical effectors of the LIF/Stat3 pathway.

5Stat3 null ES cells can be derived and expanded when differentiation stimuli are blocked using 2i. We confirmed the identity and pluripotency of these cells by chimaera formation after blastocyst injection (Figure 1A). Consistent with this, when we examined the expression of genes associated with either pluripotency or germ layer specification we found no major differences between Stat3 null and wild-type cells maintained in 2i (Figure 1B). Furthermore, null cells did not exhibit any overt sign of spontaneous differentiation or appreciable cell death (Figure 1C) and were able to generate undifferentiated colonies at clonal density with efficiency equal to wild-type cells (Figure 1D). We therefore conclude that deletion of Stat3 does not impair ES cell self-renewal efficiency in 2i. In other culture conditions, however, the mutant cells cannot self-renew because they are non-responsive to LIF, indicating that activation of Stat3 cannot be substituted by alternative mediators.

6Nonetheless, LIF is known to activate PI3 kinase and Erk signalling in addition to Stat3. It is also known that LIF enhances self-renewal efficiency in all ES cell culture conditions, including 2i. We therefore tested whether LIF has any additive effect on self-renewal of Stat3 null cells. As previously observed, LIF increased the colony-forming efficiency of wild-type cells (Figure 2A). However, Stat3 null cells showed no response, further verifying the primary role of Stat3 in mediating the contribution of LIF to ES cell self-renewal.

7The existence and developmental potency of Stat3 null ES cells highlights the regulative nature of the naïve pluripotency network. This flexibility creates the opportunity for manipulating the extrinsic environment to delineate the functional contributions of individual components. Accordingly, we exploited these mutant ES cells to define genes that are directly induced by activation of Stat3 rather than other signals downstream of LIF receptor. We exposed wild-type and Stat3 null cells to LIF for 1 h and prepared RNA for transcriptome analysis by deep sequencing. The short period of LIF stimulation is expected to enrich for primary transcriptional targets. We found that 188 genes were induced in Stat3 wild-type cells (Figure 2B, orange), and among these only 5 were induced in Stat3 null cells (Figure 2B, green). This indicates that the majority of genes acutely responsive to LIF require Stat3 for induction. We then used published Stat3 ChIP-seq (chromatin immunoprecipitation followed by massively parallel sequencing) data to generate a list of genes (see Materials and methods) bound, and thus potentially directly regulated, by Stat3 (top panel of Figure 2B, purple). This yielded 3935 unique genes, representing ∼17% of all annotated genes. Significantly, a high proportion of genes induced by LIF in Stat3 wild-type cells were also bound by Stat3 (38.8%, P-value=3.4 × 10−14). In contrast, Stat3-bound genes were significantly underrepresented among those induced by LIF in Stat3 null cells (1.9%, P-value=0.0094). We performed a similar analysis on genes downregulated after LIF stimulation (bottom panel of Figure 2B), and found that Stat3-bound genes were underrepresented (11.2%, P-value=0.003). These results are consistent with the characterized role of Stat3 as a transcriptional activator.

8Accordingly, we focussed on genes induced by LIF in wild-type cells (Figure 2C and D), and selected those involved in the regulation of transcription, such as transcription factors or chromatin modifiers. We performed gene expression analysis by qRT-PCR on a set of 13 genes and validated 6 to be induced after LIF stimulation (Figure 2D). Induction of the remaining seven genes could not be confirmed because they are expressed at low levels and thus were difficult to quantify reliably by qRT-PCR. Among these six genes are Stat3 itself and other genes previously identified as Stat3 direct targets. Interestingly, however, the most highly expressed transcription factor in this group, Tfcp2l1 (also known as Crtr-1), has not previously been linked to LIF signalling. We therefore tested whether Tcfp2l1 is indeed directly responsive to LIF by stimulating cells in the presence of the protein synthesis inhibitor cycloheximide (Figure 2F). Induction was unaffected, confirming that Tfcp2l1 is a primary target of LIF signal transduction.

9We then tested dependence on LIF for sustained expression. For this, we cultured ES cells under serum-free conditions in the presence of LIF and the Mek inhibitor PD0325901 (PD), then withdrew LIF and added a JAK kinase inhibitor (JAKi) to block LIF signalling completely. We analysed the expression of the six Stat3 targets over time and found that, with the exception of Gbx2, all genes were downregulated upon removal of the LIF signal (Figure 2G). Expression of Klf4, Pim1, Prr13, Stat3, and Tfcp2l1 therefore directly correlates with LIF stimulation, whereas Gbx2 expression is responsive to LIF (Figure 2D and E) but can be maintained by additional mechanisms.

10LIF, in combination with serum, maintains ES cell self-renewal. Therefore, omitting Stat3 itself, we tested whether the identified targets could confer self-renewal in serum without LIF. We used Rex1GFPd2 reporter cells to monitor uncommitted status and transfected these with piggyBac expression vectors. Following hygromycin selection for two passages, we tested the response to LIF withdrawal in the presence of serum. Empty vector (PB-Vector)-transfected cells rapidly downregulated the GFP reporter and differentiated (Figure 3A; Supplementary Figure S1B). This was unchanged in Pim1 and Prr13 transfectants. Klf4 and Gbx2 expression maintained a proportion of ES cells in a morphologically undifferentiated state for several passages, and this was reflected in persistent expression of the Rex1-GFP reporter. However, Tfcp2l1 transfectants showed a stronger suppression of differentiation and maintenance of Rex1GFP, comparable to continuous culture in LIF (Figure 3A, compare top-left and bottom-right flow profiles). We also examined self-renewal capacity at the single-cell level by colony forming assay (Figure 3B). Gbx2 and Klf4 transfectants produced some undifferentiated colonies in the absence of LIF, consistent with previous reports, but many more colonies were obtained from Tfcp2l1 transfectants.

11Klf4 and Gbx2 have previously been identified as factors promoting self-renewal downstream of LIF. However, the preceding data suggest that Tfcp2l1 has more potent self-renewal activity. LIF also supports self-renewal in serum-free conditions when combined with Mek inhibition (PD). We found that Tfcp2l1 expression conferred long-term self-renewal capacity under serum-free conditions in the presence of PD without LIF (Figure 3C, see also Figure 4). Thus, overexpression of Tfcp2l1 largely recapitulates LIF stimulation in different culture conditions.

12We then tested which target genes are required for LIF-responsive ES cell self-renewal by knockdown using a doxycycline (DOX)-inducible shRNA system (see Materials and methods). For each gene, we used two independent shRNA constructs with knockdown efficiencies >60% (Supplementary Figure S1C). When cultured in serum-containing media in the presence of LIF, ES cells transfected with a control shRNA (shLacZ) gave rise to the same number of colonies in the presence or absence of DOX (Figure 3D, first bar). Knockdown of Klf4 or Gbx2 had no effect in agreement with previous reports showing that suppression of either of these factors does not markedly impair ES cell self-renewal. In contrast, Tfcp2l1 knockdown caused a marked decrease in the number of ES cell colonies (Figure 3D). Similar results were obtained in LIF/PD (Figure 3E). Furthermore, Tfcp2l1 depletion in bulk culture in presence of LIF precipitated differentiation (Figure 3F, middle and bottom panels, see also Supplementary Figure S1F). Consistent with this, gene expression analysis after 4 days of shTcfcp2l1 induction showed a reduction in mRNA levels of key pluripotency factors, including Nanog, Oct3/4, Sox2, and Esrrb (Figure 3G). Significantly, however, Tfcp2l1 knockdown was well tolerated by ES cells cultured in 2i+LIF (Supplementary Figure S1D and E), confirming that the phenotype is specific to LIF-stimulated self-renewal. These results suggest that Tfcp2l1 exerts a unique function downstream of Stat3 that is necessary for a full self-renewal response to LIF (Figure 3I).

13Many naïve pluripotency factors such as Klf4 exhibit mosaic expression in ES cells cultured in LIF and serum. In contrast, we found that Tfcp2l1 protein is widely expressed, with >90% of Oct3/4 expressing cells being Tfcp2l1 positive (Figure 3H and data not shown). This broad expression may underlie the capacity of LIF to sustain self-renewal efficiently in serum.

14As shown above, Tfcp2l1 overexpression can maintain self-renewal of ES cells in the absence of LIF (Figure 3A–C). In those experiments, however, Tfcp2l1 transgene mRNA may be up to three-fold higher than endogenous levels (Supplementary Figure S1B) which, given that endogenous Tfcp2l1 transcript is very abundant in ES cells (Figure 2D), raises the possibility of neomorphic effects. We therefore used low-dose Hygromycin selection to obtain ES cell transfectants in which Tfcp2l1 transgene expression was restricted at close to LIF-stimulated endogenous levels (Figure 4A), as measured by qRT–PCR and immunoblotting (Figure 4B). This constrained expression of Tfcp2l1 was fully sufficient to sustain self-renewal after LIF withdrawal, as shown by undifferentiated morphology (Figure 4C), expression of pluripotency markers (Figure 4D and F), and capacity to form AP-positive colonies (Figure 4E).

15We then tested whether the effect of Tfcp2l1 on self-renewal in these cells was reversible by excising the floxed Tfcp2l1 transgene after expansion in the absence of LIF for seven passages (Figure 4A; Supplementary Figure S1G). After excision, cells differentiated unless maintained in the presence of LIF (Figure 4C–E). We carried out blastocyst injection with the excised cells to assay their developmental identify and potency. As a rigorous test, we injected single cells. We obtained chimaeras with an efficiency of ∼30%, similar to the parental ES cell line (Figure 4G, table). Mid-gestation chimaeras exhibited widespread contribution to all germ layers (Figure 4G, top panels). Single Tfcp2l1 excised cells could also give rise to healthy term chimaeras (Figure 4H). We conclude that constitutive expression of Tfcp2l1 is sufficient to replace LIF and sustains ES cell self-renewal without transformation or impairment of somatic differentiation potential. Collectively, these results indicate that Tfcp2l1 is the major effector of self-renewal downstream of Stat3.

16Next, we examined the epistatic relationship between Stat3 and Tfcp2l1. We transfected Stat3 null cells with expression vectors for Klf4 or Tfcp2l1. After transfection, cells were cultured in serum-free medium containing PD and the selection agent hygromycin (Figure 5A). PB-Vector cells rapidly differentiated or died, as did PB-Klf4 cells and no stable transfectants were obtained (data not shown). In contrast, Tfcp2l1 expressing cells expanded efficiently and could be maintained over several passages. They displayed undifferentiated morphology (Figure 5B) and were able to self-renew at clonal density (Figure 5C). We characterized the Stat3 null cells expressing Tfcp2l1 by qRT–PCR and found that they expressed Oct3/4, Sox2, Rex1, and Esrrb at levels similar to wild-type ES cells in 2i. Reduced Klf4 expression was observed, consistent with the absence of LIF/Stat3 signalling (Figure 5D). These data suggest that Tfcp2l1, but not Klf4, can replace Stat3 in ES cell self-renewal.

17We then consulted the ES cell ChIP-seq compendium (http://bioinformatics.cscr.cam.ac.uk/ES_Cell_ChIP-seq_compendium.html) to interrogate available ChIP-seq data sets. Unsupervised hierarchical clustering showed that Stat3 and Klf4 genome-wide binding profiles are closely related (Figure 5E). In contrast, Tfcp2l1 has a distinct binding profile and clusters with neither the ‘core’ factors nor Stat3 (Figure 5E). This is consistent with Tfcp2l1 acting downstream of Stat3 and additive to the ‘core’ factors to sustain ES cell self-renewal (Figure 5B).

18LIF signalling through Stat3 plays a key role in induction of pluripotency. Activation of LIF/Stat3 is required for iPS cell formation, and is sufficient to reprogram post-implantation EpiSCs to naïve pluripotency. It is not known, however, whether the target genes that sustain self-renewal are the same as those that mediate reprogramming. To identify Stat3 targets active during reprogramming, we used embryo-derived EpiSCs that stably express a chimaeric gp130 receptor, GY118F, which elicits hyperactivation of endogenous Stat3 in response to granulocyte colony stimulating factor (GCSF). Exposure of these EpiSCs to GCSF for 48 h in 2i is sufficient to induce formation of a modest number of reprogrammed colonies (Figure 6B, see also ). Importantly, exposing EpiSCs to LIF is not effective, suggesting that activation of Stat3 is limiting for induction of key target gene(s) to a level sufficient to drive reprogramming. We examined expression of a panel of candidate genes in EpiSCs exposed to either 2i+LIF or 2i+GCSF for 48 h in comparison with EpiSCs cultured in FGF2 and Activin A (F/A). We also included ES cells cultured in 2i+LIF as a reference. We chose 30 Stat3 targets, based on the present and previous studies, and determined their expression by qRT–PCR. We searched for genes that were induced in EpiSCs and, more specifically, were upregulated to higher levels in 2i/GCSF compared to 2i+LIF (Figure 6C, compare the second to the first column in the heatmap). Interestingly, we found that Klf4 and Myc, two reprogramming factors previously reported as Stat3 targets, were not induced by GCSF in EpiSCs, whereas four transcription factors, Klf5, Gbx2, Pim1, and Tfcp2l1, were upregulated to levels comparable to, or higher than, their expression in ES cells (Figure 6C, blue box in the heatmap and histograms). We tested whether induction of these four Stat3 targets was required for reprogramming by knockdown. We performed siRNA transfection followed by GCSF stimulation and scored the number of Oct4-GFP-positive colonies after 4 days in 2i+LIF (Figure 6D, top). Stat3 knockdown resulted in a ∼10-fold reduction, confirming that Stat3 mediates induction of pluripotency downstream of the chimaeric GCFS/gp130 receptor. We found that knockdown of Gbx2 and Klf5 had no effect, and knockdown of Pim1 resulted in a ∼2-fold decrease in colony number. However, Tfcp2l1 ablation substantially reduced the number of colonies recovered. This suggests that Tfcp2l1 plays a critical role in induction of naïve pluripotency downstream of Stat3.

19We then tested whether Tfcp2l1 expression is sufficient to convert EpiSCs into naïve pluripotency. We found that Oct4-GFP reporter EpiSCs (O4GIP) stably transfected with a Tfcp2l1 expression vector efficiently converted into naïve pluripotency when exposed to 2i (Figure 6E). Importantly, addition of LIF resulted in only a modest further increase in the number of colonies, suggesting that forced expression of Tfcp2l1 could largely recapitulate the contribution of LIF to EpiSC reprogramming. The reprogrammed colonies obtained could be passaged in 2i+LIF and gave rise to stable iPS cell lines (Figure 6F). Gene expression analysis confirmed activation of the naïve markers Tbx3 and Esrrb and downregulation of the EpiSC markers Fgf5 and Sox17 (Figure 6G).

20EpiSCs are variable and one line (GOF18) derived on feeders in the presence of FGF2 has been reported to convert spontaneously to naïve pluripotency when exposed to 2i+LIF. We found that GOF18 cells retained this ability even after prolonged culture under feeder-free conditions in the presence of FGF2 and Activin A. The conversion frequency in 2i alone was extremely low, but was greatly increased by addition of LIF (Supplementary Figure S2A). Moreover, we found that addition of LIF for only the first 48 h gave rise to the same number of colonies as continuous exposure (Supplementary Figure S2A, compare second and third bar). Therefore, we measured the expression of Stat3 targets by qPCR in GOF18 cells exposed to either 2i or 2i+LIF for 48 h. Both Gbx2 and Tfcp2l1 were expressed in these EpiSCs in 2i, but significantly only Tfcp2l1 was further upregulated by LIF (Supplementary Figure S2B). We transiently transfected GOF18 cells with a Tfcp2l1 expression vector and 24 h after transfection replated them in 2i without LIF. After 6 days, we scored the number of Oct4-GFP-positive colonies. We observed an ∼10-fold increase in colony formation after Tfcp2l1 transfection (Supplementary Figure S2C, left), indicating that a transient increase in Tfcp2l1 expression is sufficient to drive the reprogramming process. The iPS cells generated after Tfcp2l1 expression could be expanded in 2i+LIF over multiple passages, showing undifferentiated morphology, stable expression of the Oct4-GFP reporter (Supplementary Figure S2C, right panels) and of naïve pluripotency markers, and a concomitant reduction in EpiSC markers (Supplementary Figure S2D).

21We analysed data on a panel of EpiSC lines and found no significant differences in expression of LIF/Stat3 pathway components between lines that convert spontaneously and lines that do not (Supplementary Figure S2E, compare red and blue bars). We also found that both O4GIP and GOF18 cells express Tfcp2l1 mRNA at similar very low levels (Supplementary Figure S2F). Moreover, Tfcp2l1 protein 1 could not be detected by immunofluorescence in GOF18 cells, excluding the possibility of a rare sub-population of expressing cells (Supplementary Figure S2G). We conclude that spontaneous conversion into naïve pluripotency does not correlate with increased Tfcp2l1 expression in the EpiSC state, but rather with the ability to activate Tfcp2l1.

22Finally, to confirm functional naïve pluripotent status after Tfcp2l1-mediated reprogramming, we transfected O4GIP-derived iPS cells with a constitutive GFP expression vector, excised the Tfcp2l1 transgene by Cre plasmid transfection, and carried out blastocyst injection. Two out of six transferred embryos showed widespread GFP expression at mid-gestation (Figure 6H). Taken together, these results suggest that Tfcp2l1 plays a major role downstream of LIF/Stat3 in the conversion of EpiSCs into authentic naïve pluripotency.

23The LIF/Stat3 pathway plays crucial facultative roles in mouse ES cell self-renewal, maintenance of pluripotent epiblast during embryonic diapause, generation of EG cells (Leitch et al, 2013), and molecular induction of pluripotency. How LIF input is integrated with the pluripotency gene regulatory network has remained unclear, however. cMyc has been proposed as a mediator, but neither cMyc nor NMyc are appreciably induced by LIF in ES cells. Klf4 has been shown to be a direct target of Stat3, but is neither required nor sufficient to account for the potent effects of LIF stimulation. A significant role for alternative modes of LIF signalling has also been invoked, either through PI3 kinase or by JAK modification of chromatin.

24To resolve this issue, we first confirmed that Stat3 null cells have functional ES cell identity and potency yet show no self-renewal response to LIF. We then sought the missing target(s) of Stat3 and identified the transcription factor Tfcp2l1. Tfcp2l1, also known as Crtr1, has previously been described in ES cells, but its relationship to LIF/Stat3 has gone unnoticed. Tfcp2l1 is regulated directly by activated Stat3 and induced to high levels. Gain- and loss-of-function perturbations established that Tfcp2l1 is both necessary and largely sufficient to mediate the effects of LIF on ES cell self-renewal in either serum or serum-free culture conditions, in contrast to previously identified targets, Klf4, Gbx2, and Pim1. Strikingly, forcing expression of Tfcp2l1 at close to endogenous levels was sufficient to phenocopy LIF stimulation. Finally, we showed that Tcfcp2l1 is epistatic to Stat3 in sustaining ES cell self-renewal (Figure 5). From these results, we conclude that Tfcp2l1 is the major mediator of ES cell self-renewal downstream of LIF/Stat3, and that it exerts functions that cannot be compensated by other LIF targets. We previously noted that Tfcp2l1 is also induced by GSK3 inhibition. It may therefore constitute the postulated point of intersection between these two mechanisms for supporting ES cell self-renewal. Notably, however, while Tfcp2l1 appears sufficient to replace LIF/Stat3, it cannot reproduce the full effect of GSK3 inhibition (Figure 3A in ).

25We also investigated mediators of LIF/Stat3 action during reprogramming. Klf4 is a canonical reprogramming factor. However, LIF is required for efficient reprogramming of both somatic cells and EpiSCs even when Klf4 is overexpressed, suggesting a requirement for other LIF/Stat3 target(s). Analysis of the initial phase of EpiSC reprogramming did not show induction of known reprogramming factors such as Klf4 and cMyc, consistent with previous observations. In contrast, Gbx2, Klf5, Pim1, and Tfcp2l1 were induced (Figure 6C). Transient expression of Tfcp2l1 was sufficient to convert EpiSCs into naïve pluripotency without LIF stimulation, while Tfcp2l1 knockdown caused a dramatic reduction in reprogramming efficiency (Figure 6D). Importantly, when Tfcp2l1 is knocked down in ES cells in 2i+LIF there was no significant compromise to self-renewal (Supplementary Figure S1D and E). This suggests that Tfcp2l1 has a specific role in resetting of naive pluripotency in addition to its action in self-renewal.

26Tfcp2l1 is a member of the CP2 family of transcription factors, which are generally considered to be activators of transcription. This has been demonstrated for Tcfp2l1 in ES cells, although in some contexts it may act as a repressor. Interestingly, genome location data indicate that Tbx3, Esrrb, and Nanog are bound by Tfcp2l1 (data from and ES cell ChIP-seq Compendium) and upon Tfcp2l1 knockdown we noted downregulation of the three genes (Figure 3G). Similar effects of Tfcp2l1 depletion are also apparent in data from a recent microarray study (; Supplementary Figure S3A and B). Furthermore, we observed that upon overexpression of Tfcp2l1, Tbx3 and Nanog are significantly upregulated (Figure 6G; Supplementary Figures S2D, S3A and B and data not shown). These observations suggest that Tfcp2l1 sustains ES cell self-renewal in large part through transcriptional activation of Nanog and Tbx3, both of which have been proposed to lie genetically downstream of LIF but neither of which are direct targets.

27Tfcp2l1 is expressed in the inner cell mass of the mouse blastocyst and is downregulated shortly after implantation. It is rapidly downregulated when ES cells start to differentiate and is absent from mouse EpiSCs (Figure 6) in common with other naïve pluripotency markers such as Esrrb, Rex1, and Klf2/4. Human pluripotent ES cells are considered to be in a primed pluripotent state closer to mouse EpiSCs than ES cells. Tfcp2l1 is detected in the human ICM, but downregulated during derivation of human ES cells. Human ES cells have been reported to convert into a naïve-like state by overexpression of Klf2, Klf4, and Oct3/4. Interestingly, inspection of the microarray data from that study reveals that Tfcp2l1 is highly upregulated. It is possible therefore that Tfcp2l1 may play a role in generating and stabilizing a human naïve pluripotent state.

28LIF is a stimulus that confers self-renewal and can promote emergence of the naïve pluripotent gene regulatory network, but is not itself integral to pluripotent identity. Consistent with this, Tfcp2l1 genome occupancy does not cluster with that of canonical pluripotency factors Oct4, Sox2, and Nanog. Instead, it binds to a largely distinct, and broad, set of genes. Significantly, however, these targets include key naïve pluripotency factors Nanog and Tbx3 previously shown to support ES cell self-renewal. Furthermore, the Tfcp2l1 promoter is bound by Oct4, Sox2, Nanog, Esrrb, and Klf4 (ES cell ChIP-seq compendium), and is targeted by the pluripotency repressor Tcf3, indicating that Tfcp2l1 is hard-wired into the pluripotency gene regulatory network. We conclude that Tfcp2l1 is a previously overlooked transcriptional regulator that plays a pivotal role in the capability to generate, derive, and propagate naïve pluripotent stem cells. Although mutation of Tfcp2l1 is reported to be compatible with mouse embryo development, the characterized gene trap mutation may not create a null allele since residual full-length transcript is detectable. It will therefore be of interest to determine the effects of complete deletion of Tfcp2l1 on the emergence of pluripotency in the blastocyst, and epiblast maintenance during diapause when the Stat3 pathway is indispensable.

29RNA was extracted using the TRIzol method (Invitrogen) followed by treatment with TURBO DNase (Ambion). Ribosomal RNA was depleted using RiboMinus (Invitrogen), and the remaining RNA was sheared by ultrasonication on a Covaris S2 for 90 s with the following parameter settings: Duty Cycle=10, Cycles per Burst=200, Intensity=5. Fragmented RNA was reverse-transcribed with SuperScript III (Invitrogen) at 50°C for 2 h using random hexamer and oligo-dT primers (10:1) in the presence of 6 μg/ml actinomycin D to inhibit the generation of second-strand products. Second-strand cDNA was synthesized by DNA Polymerase I for 2 h at 16°C in the presence of RNase H and with dUTPs substituted for dTTPs. End repair of double-strand cDNAs was carried out with T4 DNA polymerase and T4 polynucleotide kinase (New England Biolabs). Blunt-end, 3′-phosphorylated products were 3′-adenylated by exo-Klenow fragment in the presence of dATPs and ligated to sequencing adapters (Illumina) by T4 DNA ligase (New England Biolabs) at 20°C for 30 min. Following adapter ligation, the second strand of the library constructs was digested with uracil DNA glycosylase (UDG) and apurinic/apyrimidinic endonuclease 1 (APE 1) for 30 min at 37°C. PCR amplification of first-strand library constructs was carried out with Phusion DNA polymerase (Finnzymes) for 15 cycles. Purification of reaction products between each step was performed with Ampure XP paramagnetic beads (Beckman Coulter). The molarity and size distribution of the libraries was assessed by DNA 1000 microfluidic chips on the Agilent 2100 Bioanalyzer. Sequencing was performed on the Illumina GAIIx yielding 35–40 M 105 bp reads per sample. RNA-seq data are available in the ArrayExpress repository under accession E-MTAB-1796.

30Reads were aligned to the mouse reference genome (build mm10) using GSNAP version of 2012-05-24. The aligner was provided with known splice sites based on Ensembl version 68, but enabled to identify novel splice sites. A maximum of 10 mismatches were allowed for read alignment. Gene counts were calculated using the htseq-count utility (version 0.5.3p3, http://www-huber.embl.de/users/anders/HTSeq/doc/count.html) and used as an input for differential gene expression analysis with DESeq version 1.10.1. Genes with a P-value of <0.05 (Benjamini–Hochberg adjusted) were selected for further analysis.

31We obtained ChIP-seq data for Stat3 and a GFP control from a previous study. Reads were mapped to the mouse genome (build mm10) using bowtie version 0.12.8, discarding non-unique alignments. Peak detection was performed using MACS version 1.4.1. Each peak was associated with up to two genes as follows: the most proximal gene on each strand was identified by examining 50 kb upstream and downstream of the peak. Peaks were assigned to intersect genes unless an overlapping gene was found on the opposite strand. For the heatmap in Figure 5E a binary peak matrix was generated as described, analysed by unsupervised hierarchical clustering using Pearson’s correlation coefficients, and displayed using the heatmap function in R. The ChIP-seq data used for target gene intersection in Figure 5F are Esrrb—GSE11431 (GSM288355) and TFCP2L1—GSE11431. All raw and processed data used in this study are available at http://bioinformatics.cscr.cam.ac.uk/ES_Cell_ChIP-seq_compendium.html

32ES cells were cultured without feeders on plastic coated with 0.1% gelatine (Sigma, cat. G1890) and replated every 3 days at a split ratio of 1 in 10 following dissociation with Accutase (PAA, cat. L11-007). Cells were cultured either in the GMEM (Sigma, cat. G5154) supplemented with 10% FCS (Sigma, cat. F7524), 100 mM 2-mercaptoethanol (Sigma, cat. M7522), 1 × MEM non-essential amino acids (Invitrogen, cat. 1140-036), 2 mM L-glutamine, 1 mM sodium pyruvate (both from Invitrogen), and 100 units/ml LIF, or in the serum-free media N2B27 (NDiff N2B27 base medium, Stem Cell Sciences Ltd, cat. SCS-SF-NB-02) supplemented, as indicated, with small-molecule inhibitors PD (1 μM, PD0325901) and CH (3 μM, CHIR99021) and LIF prepared in-house. Colony forming assays were carried out by plating 60 ES cells/cm2 on plates coated with laminin (Sigma, cat. L2020). Plates were fixed and stained for alkaline phosphatase (Sigma, cat. 86 R-1KT) according to the manufacturer’s protocol. Plates were scanned using a CellCelector (Aviso) and scored manually. EpiSCs were cultured as previously described. O4GIP (OEC-2 line) was described in. GOF18 was described in (. For DNA transfection, we used Lipofectamine 2000 (Invitrogen, cat. 11668030) and performed reverse transfection. For one well of a 6-well plate (10 cm2), we used 6 μl of transfection reagent, 2 μg of plasmid DNA, and 300 000 cells in 2 ml of N2B27 medium. The medium was changed after overnight incubation. The piggyBac vectors used for overexpression followed by transgene excision have been described in. The transgene is flanked by loxP sites, allowing efficient excision by transfection with Cre.

33Total RNA was isolated using the RNeasy kit (Qiagen) and 500 ng used for cDNA synthesis using SuperScript III (Invitrogen) and oligo-dT primers. Quantitative real-time PCR was carried out with SYBR green detection. Primers are detailed in Supplementary Table S2. Technical replicates were carried out for all reactions.

34siRNAs were transfected at a final concentration of 40 nM using Dharmafect 1 (Dharmacon, cat. T-2001-01), following the protocol for reverse transfection. For a 12-well plate (4 cm2), we used 2 μl of transfection reagent, 2 μl of 20 μM siRNA solution, and 30 000 ES cells in 1 ml of N2B27 medium. The medium was changed after overnight incubation. siRNAs were purchased from Qiagen (Flexitube GeneSolution, see Supplementary Table S3).

35To generate a Dox-inducible shRNA system, we inserted into a piggyBac vector the murine pre-miR-155 hairpin (from the BLOCK-IT Pol2 miR RNAi system; Invitrogen) downstream of a third-generation Tet-responsive promoter (TRE3G). Sequences targeting the gene of interest (Supplementary Table S4) were cloned into the miR-155 hairpin by homologous recombination using the Infusion HD cloning kit (Clontech). The third-generation TET activator (TET3G) was cloned in a piggyBac vector under control of a constitutive CAG promoter. Cells were co-transfected with pBase helper plasmid and the two piggyBac vectors containing both the TRE3G-shRNA and the TET3G, and cultured in the presence of selection agent for 7 days prior to experimental assays.

36After treatment with Accutase, live ES cells were resuspended in PBS with 3% FCS and 0.05 nM ToPro-3 (Invitrogen) was added at a concentration of 0.05 nM to detect dead cells. Flow cytometry analyses were performed using a Dako Cytomation CyAn ADP high-performance cytometer with Summit software.

37Cells were fixed for 10 min in 4% PFA at Room Temperature (RT), permeabilized for 5 min in PBS+0.2% Triton X at RT, and blocked for 30 min in PBS+3% donkey serum at RT. Cells were incubated overnight at 4°C with the primary antibodies (anti-Tcfcp2l1: rabbit IgG, Abcam 123354, 1:400; anti-Klf4: goat IgG, R&D Systems AF3158, 1:400; anti-Oct4: mouse IgG, Santa Cruz C-10, used at a 1:300 dilution). After washing in PBS, the cells were incubated with Alexa Fluor secondary antibodies (Invitrogen, 1:500 in PBS+3%serum), for 30 min at RT. After DAPI staining, images were acquired using a Zeiss AxioObserver D1 microscope. Automated single-cell image quantification was performed using CellProfile (Broad Institute—).

38Immunoblotting was performed as previously described in. Antibodies used are rabbit polyclonal anti-Tfcp2l1 (Abcam, AB123354, 1:500 dilution in 1% milk) and mouse monoclonal anti-GAPDH (Sigma-Aldrich, G8795, 1:1000 in 5% milk).