Whsc1 links pluripotency exit with mesendoderm specification

Abstract

How pluripotent stem cells differentiate into the main germ layers is a key question of developmental biology. Here we show that the chromatin-related factor Whsc1 (Nsd2, MMSET) has a dual role in pluripotency exit and germ layer specification of embryonic stem cells (ESCs). Upon induction of differentiation, a proportion of Whsc1-depleted ESCs remain entrapped in a pluripotent state and fail to form mesendoderm, although they are still capable of generating neuroectoderm. These functions of Whsc1 are independent of its methyltransferase activity. Whsc1 binds to enhancers of the mesendodermal regulators Gata4, T (Brachyury), Gata6 and Foxa2 together with Brd4, and activates the genes’ expression. Depleting each of these regulators also delays pluripotency exit, suggesting that they mediate the effects observed with Whsc1. Our data indicate that Whsc1 links silencing of the pluripotency regulatory network with activation of mesendoderm lineages.

Introduction

Embryonic stem cells (ESCs) are an excellent model to study the generation of the three main germ layers, endoderm, mesoderm and ectoderm ¹. The first event during the transition of ESCs into more differentiated cells is the exit from pluripotency, which includes silencing of the core pluripotency transcription factors (TFs) Oct4, Sox2 and Nanog ^2–4. This process can either be externally initiated by deprivation of self-renewal signals ⁵ or by exposure of the cells to differentiation-inducing cues ⁶. Subsequently, germ layer-instructive TFs become up-regulated and activate markers that define the various lineages ⁴. These factors comprise Brachyury (T) and Gata4 for mesoderm, Gata6, Gata4 and Foxa2 for endoderm and Pax6 and Sox1 for neuroectoderm ^{3, 4, 6, 7}. Screens with ESC reporter lines and haploid ESCs have identified several regulators required for pluripotency exit, including several TFs and RNA binding proteins, all of which were shown to affect members of the pluripotency network ^8–13.

Besides TFs and noncoding RNAs, chromatin-related factors (CRFs) play an important role in ESC differentiation, which involves a progressive transition from a relatively open chromatin state to a more compact one ¹⁴. Several CRFs are broadly involved in ESC differentiation, such as Polycomb group proteins that are required for the repression of pluripotency associated and lineage inappropriate genes ¹⁵. Others act in a more restricted fashion, as exemplified by the requirement of Setd2 for endodermal ¹⁶, Mel18 for mesodermal ¹⁷ and Zrf1 for neuroectodermal differentiation ¹⁸.

Here we report that Whsc1 is required for efficient pluripotency exit and mesendoderm specification, independent of its methyltransferase activity. These effects can be explained by its ability to activate several mesendoderm regulators.

Results

Whsc1 is required for efficient exit from pluripotency

To search for CRFs that play a role in pluripotency exit and subsequent ESC differentiation we performed an in-silico screen by analysing datasets that record transcriptome changes during the differentiation of mouse ESCs into embryoid bodies (EBs). We compiled a list of 653 genes encoding histone writers and readers, histone variants, nucleosome positioning proteins and a selection of long non-coding RNAs (Supplementary Table 1 and Supplementary Fig. 1a). Three candidates consistently scored among the most up-regulated genes. These were Cbx4, encoding a Polycomb group protein that has been shown to orchestrate ESC differentiation ¹⁹, L3mbtl3, a putative Polycomb group family member ²⁰ and Whsc1, encoding a SET-domain methyltransferase which in humans is associated with Wolf-Hirschhorn syndrome ²¹.

To study whether these candidates are required for pluripotency exit we tested the effects of knockdowns on the differentiation of a pluripotency reporter ESC line. This line, which contains a destabilized version of GFP knocked into the Rex1 locus (Rex1GFPd2) ⁸, was induced to differentiate by transfer into medium containing N2 and B27 supplements, Activin A and foetal bovine serum (FBS) ^{22, 23}. Upon induction, the reporter line became GFP negative within 48–72h (Fig. 1a–c) confirming that the loss of Rex1 expression is a highly sensitive readout for exit from the pluripotent state ^{24, 25}. Concomitantly, the cells down-regulated pluripotency associated genes and up-regulated mesendoderm and neuroectoderm specific genes (Supplementary Fig. 1b).

Figure 1. Whsc1 is involved in the exit from pluripotency of mESCs.

a, Experimental procedure to assess exit from pluripotency using Rex1GFPd2 ESCs. b, Representative FACS plots of GFP (Rex1) expression in cells expressing scrambled shRNA (shScr) as a control or shWhsc1 knockdown constructs (shWhsc1.2 and shWhsc1.4) at 0 hr and 72 hrs after induction. Three independent experiments were performed with similar results. c, Representative images of ESCs expressing shScr and shWhsc1.4 48 hrs after induction of exit from pluripotency. Three independent experiments were performed with similar results. Scale bar: 500 μm d, Expression of Whsc1 quantified by RT-qPCR (left panel) and Western blot (right panel) in control (shScr) and Whsc1 knockdown ESCs. Tubulin was used as loading control. For left panel, data represent mean±s.d. from n=3 independent experiments and p-values were calculated by two-tailed unpaired t-test. For right panel, three independent experiments were performed with similar results and scanned images of unprocessed blots are shown in Supplementary Fig. 8. e, Expression of pluripotency genes 72 hrs after induction of exit from pluripotency. Data represent mean±s.d. from n=3 independent experiments and p-values were calculated by two-tailed unpaired t-test.

Two independent small hairpin RNA (shRNA) lentiviral constructs were used to knock down the expression of each of the three candidate genes. After induction of differentiation, cells expressing scrambled shRNA (shScr) became GFP negative within 48–72h post induction (p.i.), as did cells expressing shCbx4 and shL3mbtl3 (Supplementary Fig. 1c). In contrast, 25–35% of ESCs transduced with the shWhsc1 constructs retained GFP expression (Fig. 1b, c). Notably, both Whsc1 knockdown constructs strongly reduced the expression of both Whsc1 mRNA and the ~180 and ~100kDa protein isoforms (Fig. 1d). Interestingly, shWhsc1 ESCs retained elevated expression levels of the pluripotency genes Rex1, Pou5f1, Nanog, Lin28a and Esrrb (Fig. 1e). In addition, Whsc1-depleted cells, when re-plated after induced differentiation, yielded large numbers of colonies positive for alkaline phosphatase (AP, an indicator of pluripotency) even after 2 consecutive rounds of treatment, whereas no such colonies could be recovered from control cells as early as after one round (Supplementary Fig. 1d). Depletion of Whsc1 in ESCs neither resulted in an up-regulation of pluripotency factors nor in an alteration of the cells’ growth kinetics (Supplementary Fig. 1e–g), ruling out more indirect effects on pluripotency exit.

In summary, Whsc1 depletion impaired the down-regulation of pluripotency markers, indicating a requirement for efficient pluripotency exit.

Depletion of Whsc1 impairs induction of mesendoderm differentiation in vitro and in vivo

To evaluate the effect of Whsc1 depletion on the differentiation capacity of ESCs, we examined embryoid bodies (EBs) derived from knockdown ESCs (Fig. 2a). Whsc1-depleted EBs were smaller than those from control ESCs (Fig. 2b), exhibited lower levels of Whsc1 mRNA and protein while retaining pluripotency factor expression (Fig. 2c, d). Importantly, Whsc1-depleted EBs also showed a strongly reduced expression of the mesendoderm regulator genes Gata4, T, Gata6, Foxa2, Sox17 and Flk1, while neuroectodermal gene expression (Sox1, Pax6 and Nes) remained unaffected (Fig. 2c). In addition, when shWhsc1 EBs were dissociated and re-plated onto mouse embryo fibroblasts (MEFs) in medium with 2i and LIF they formed large numbers of AP positive colonies, whereas control cells yielded few such colonies (Fig. 2e), suggesting that shWhsc1 EBs retained pluripotent cells. Moreover, EBs derived from shWhsc1 Epiblast Stem Cells (EpiSCs) also retained Pou5f1 expression and showed reduced developmental potential towards the mesendoderm lineage (Supplementary Fig. 2a).

Figure 2. Whsc1 is required for mesendoderm differentiation.

a, Schematics of EB differentiation assay. b, Images of Day 6 EBs derived from control and Whsc1 depleted Rex1GFPd2 cells (left panel) and quantification of EB size (right panel). Data in the right panel represent mean±s.d. from shScr EBs (n=21) and shWhsc1.4 EBs (n=22) respectively. The p-value was calculated by two-tailed Mann-Whitney test. Scale Bar: 500 μm. c, Expression of Whsc1, pluripotency, mesendoderm and ectoderm markers in Day 6 EBs quantified by RT-qPCR from control and Whsc1 depleted Rex1GFPd2 cells. Data represent mean±s.d. from n=3 independent experiments and p-values were calculated by two-tailed unpaired t-test. d, Levels of Whsc1 and pluripotency factors in Day 6 EBs evaluated by Western Blot from control and Whsc1-depleted Rex1GFPd2 cells. Tubulin was used as loading control. Two independent experiments were performed with similar results. Scanned images of unprocessed blots are shown in Supplementary Fig. 8. e, Schematics of EB re-plating assay and images of alkaline phosphatase (AP) stained colonies obtained from control and Whsc1-depleted EBs. Scale bar: 1mm; Right panel: quantification of AP positive colonies. Data represent mean±s.d. from n=3 independent experiments and p-values were calculated by two-tailed unpaired t-test.

To probe their differentiation potential in vivo, we generated teratomas in immunodeficient mice. Control cell-derived teratomas contained tissues from all three germ layers, including muscle, cartilage, gut epithelium and neuro-epithelium. In contrast, Whsc1-depleted teratomas derived were mainly composed of neuro-epithelium and poorly differentiated cells (Supplementary Fig. 2b). Accordingly, shWhsc1 teratomas retained pluripotency and neuroectoderm marker expression while showing reduced levels of mesendoderm markers (Supplementary Fig. 2c).

Since both EBs and teratomas comprise a complex mixture of cells, we next tested the effects of Whsc1 knockdown on the directed differentiation of ESCs towards mesoderm, endoderm and ectoderm. Whsc1-depleted cells generated significantly fewer beating cardiomyocyte-containing colonies than control cells (Supplementary Fig. 3a, Supplementary Movie 1, 2) and a reduced expression of the cardiac regulators Mef2c and Nkx2.5 at 10 days p.i., while exhibiting increased Rex1 and Pou5f1 expression (Supplementary Fig. 3b, c). To test the effects of shWhsc1 on definitive endoderm differentiation, Eomes-GFP reporter ESCs ²⁶ were transferred into medium containing Activin A, Fgf4, heparin, PI-103 and CHIR ²⁷. Seven days later about 30% of control cells became GFP positive, whereas only 5% of Whsc1-depleted cells did so (Supplementary Fig. 3d, e). In line with these findings, shWhsc1-expressing cells showed impaired up-regulation of the endoderm markers Cxcr4, Cldn6, Foxa2, Gata6 and Sox17 and inefficient down-regulation of the pluripotency regulators Rex1, Pou5f1 and Esrrb (Supplementary Fig. 3f, g). Finally, we derived neural progenitor cells ²⁸ from control and Whsc1 knockdown ESCs and found no significant differences in the expression of the neural markers Pax6, Sox1 and Nes nor that of Pou5f1 and Esrrb (Supplementary Fig. 3h).

Taken together, the results obtained with embryoid bodies, teratomas and directed germ layer induction experiments confirm that the delay of pluripotency exit induced by Whsc1-depletion is coupled to an impairment of mesendoderm differentiation.

The catalytic SET-domain of Whsc1 is dispensable for efficient pluripotency exit and mesendoderm differentiation

Whsc1 requires the catalytic SET-domain on the C-terminus to di-methylate the lysine 36 on histone 3 (Supplementary Fig. 4a) ^{29, 30}. This function has been linked to the normal foetal heart and cartilage development in a murine model in which the SET-domain of Whsc1 was excised ³¹. Surprisingly, however, ESCs derived from these mice (hereafter referred to as ΔSET) behaved like WT ESCs in our assays (Supplementary Fig. 4b, c). In contrast, expression of shWhsc1.4 which targets the 5’ end of Whsc1 mRNA in ΔSET ESCs caused a delay in pluripotency exit, and EBs derived from these cells also showed the retention of pluripotency gene expression and the selective reduction of mesendoderm marker expression (Supplementary Fig. 4b, c). As the Whsc1 gene produces two major transcripts encoding a full-length protein of ~180kD and a short isoform of ~100kD that corresponds to the N-terminal portion (Fig. 3a), we examined Whsc1 expression by Western blot. We found that ΔSET ESCs also express two proteins: the larger protein corresponds to a C-terminal truncated form of ~130kD and the 100kD isoform (Supplementary Fig. 4, d). These results raise the possibility that in the absence of catalytic SET domain, the N-terminus of Whsc1 plays a role in both the induction of pluripotency exit and mesendoderm specification.

Figure 3. Ablation of Whsc1 results in delayed down-regulation of pluripotency genes and impairment of mesendoderm formation.

a, Schematics of strategy for generating ESCs with a complete Whsc1 knockout using CRISPR-Cas9 approach. Top panel: organization of Whsc1 that can generate two transcript isoforms (long and short). The targets of the gRNAs (exon1 and exon15) are indicated by arrows_; Middle panel: genomic sequences of wild-type Whsc1 (Whsc1 +/+) spanning exon1 and exon 15 (exon1 and exon15 are indicated as red boxes); Bottom panel: sequence and chromatograph of the Whsc1 knockout (Whsc1−/−). b, Western blot of Whsc1 and pluripotency factors, Sox2, Oct4 and Nanog in Whsc1 +/+ and two Whsc1 −/− ESC clones. Tubulin was used as loading control. Two independent experiments were performed with similar results. Scanned images of unprocessed blots are shown in Supplementary Fig. 8. c, Expression kinetics of pluripotency genes in Whsc1 +/+ and Whsc1 −/− cells were monitored by RT-qPCR after transfer into N2B27 medium containing Activin A and FBS. Data represent mean±s.d. from n=3 independent experiments and p-values were calculated by two-tailed unpaired t-test. d, Expression kinetics of mesendodermal, ectodermal and pluripotency genes were evaluated by RT-qPCR during EB formation using Whsc1 +/+ or Whsc1 −/− ESCs (n=3). Data represent mean±s.d. from n=3 independent experiments and p-values were calculated by two-tailed unpaired t-test. e, Volcano plot showing differentially expressed genes in Whsc1 −/− ESCs compared to Whsc1 +/+ ESCs. All genes with RPKM>1 were plotted and those up-regulated in Whsc1 −/− cells were indicated as red spots (Log2FC>1 and p<0.05). Genes down-regulated in Whsc1 −/− cells were indicated as light blue spots (Log2FC<−1 and p<0.05). Pluripotency genes, Pou5f1, Sox2, Nanog, Esrrb and Rex1 were indicated as black spots.

To explore this hypothesis, we used a CRISPR-Cas9 approach to generate an ESC line with a complete Whsc1 knockout, by targeting exon1 and exon15. This led to the deletion of a ca. 40Kb fragment encoding both the long and short isoforms of Whsc1 (Fig. 3a) and also eliminated multiple AUGs that could serve as potential start codons. Whsc1−/− ESCs lacked both protein isoforms (Fig. 3b), expressed unaltered levels of Sox2, Oct4 and Nanog (Fig. 3b) and showed a delayed down-regulation of Rex1, Pou5f1 and Nanog after differentiation induction (Fig. 3c). EB derivatives were impaired in the up-regulation of mesendoderm but not of neuroectoderm genes and exhibited a delayed down-regulation of Rex1, Pou5f1 and Nanog (Fig. 3d).

To test the possibility that the observed phenotype of Whsc1−/− cells could be explained by alterations already pre-existing in the ESCs caused by the knockout, we performed an RNA-seq analysis and found no significant expression changes of pluripotency genes Pou5f1, Sox2, Nanog, Esrrb and Rex1 compared to Whsc1+/+ ESCs. Moreover, only 71 genes were found to be differentially expressed in Whsc1−/− cells (Fig. 3e). Of note, this list did not include factors described to dismantle the pluripotency network, such as Flcn, Tcf3, Zfp706, Foxd3, Mettl3 and Pum1 ^{8–13, 32} (Supplementary Table 2).

In sum, our data show that the SET domain is dispensable for the ability of Whsc1 to facilitate rapid pluripotency exit and to induce mesendoderm specification.

The N-terminus of Whsc1 is sufficient to induce pluripotency exit and mesendoderm differentiation

In an attempt to show that the N-terminal domain of Whsc1 is sufficient for the biological properties of Whsc1−/− ESCs we performed rescue experiments. For this, we engineered Flag tagged constructs of full length human WHSC1 (FL), two N-terminal fragments of different lengths (Nter-1 and 2) and a C-terminal portion containing the SET-domain (Cter) (Fig. 4a). Whsc1−/− cells expressing these constructs were subjected to the pluripotency exit and germ layer formation assays; as expected, FL-WHSC1 rescued the cell’s phenotype in both assays (Fig. 4b, c). Remarkably, expression of either N-terminal construct (Nter-1 and 2) in Whsc1 −/− cells was sufficient to rescue the cells’ phenotype, showing rapid down-regulation of Rex1, Pou5f1 and Nanog after induction of differentiation. In contrast, the C-terminal construct containing the SET domain failed to rescue (Fig. 4b). In line with these findings, expression of either N-terminal construct was also sufficient to restore mesendoderm marker up-regulation and pluripotency gene down-regulation during EB formation, while the C-terminal construct showed no such effect (Fig. 4c). Since Whsc1 has been described to methylate H3K36 and to alter EZH2 binding in multiple myeloma cells ^{29, 30}, we compared Whsc1+/+ and Whsc1−/− ESCs for the relevant histone marks. However, we found no significant differences in the global levels of H3K36me2, H3K36me3 and the EZH2-associated H3K27me3 mark (Supplementary Fig. 4e).

Figure 4. N-terminus of Whsc1 is sufficient to rescue pluripotency exit and mesendoderm differentiation in Whsc1 −/− cells.

a, Schematics of full length and truncated human WHSC1 constructs tagged by 3XFlag. b, Expression kinetics (RT-qPCR) of pluripotency genes in Whsc1 +/+ and in Whsc1 −/− cells infected with WT, Nter-1, Nter-2 or Cter WHSC1 constructs after transfer into N2B27 medium containing Activin A and FBS. Data represent mean±s.d. from n=3 independent experiments. c, Expression of mesendodermal, ectodermal and pluripotency genes was measured by RT-qPCR in Day 6 EBs derived from Whsc1 +/+ and Whsc1 −/− ESCs infected with WT, Nter-1, Nter-2 or Cter WHSC1 constructs. Data represent mean±s.d. from n=3 independent experiments and p-values were calculated by two-tailed unpaired t-test.

Our findings show that the N-terminus of Whsc1 is sufficient to facilitate rapid exit from pluripotency and induce mesendoderm formation. They also suggest that the methyltransferase activity of Whsc1 acts in a cell context-dependent manner.

Whsc1 mediates enhancer activation of mesendoderm regulators

Our observations raise the possibility that Whsc1 directly controls the expression of transcription factors that specify mesodermal and endodermal lineages. We therefore asked whether Whsc1 binds to and activates mesendodermal specific enhancers. We focused on genomic loci of the mesendoderm-instructive TF genes Gata4, T, Gata6, and Foxa2. Putative enhancers of the neuroectodermal TF genes Pax6 and Sox1 were included as controls. To identify candidate regulatory regions of these genes, we mapped the distribution of the enhancer mark H3K27ac in cells either positive for the mesendodermal markers Eomes or Flk1, in mesendodermal progenitors induced by Activin A, and in neural progenitor cells. In addition, as it has been shown that promoter-enhancer interactions occur mainly within the same Topologically Associating Domains (TADs) ^{33, 34}, we focused on regions within the same TAD. These analyses revealed the presence of several putative enhancers selectively marked by H3K27ac in either mesendodermal or neuroectodermal cells (Fig. 5a and Supplementary Fig. 5a). To explore whether Whsc1 binds to these regions, we performed ChIP analyses using day 6 (D6) EBs. This showed that Whsc1 is significantly enriched at 8 out of 26 putative enhancers tested within the T, Gata6, Foxa2 and Gata4 loci and that this enrichment is reduced in Whsc1−/− EBs (Fig. 5b and Supplementary Fig. 5b). Notably, no significant Whsc1 binding was observed in 12 regions specifically marked by H3K27ac in the Pax6 and Sox1 loci in ectodermal cells (Supplementary Fig. 5b). To further explore whether these Whsc1-bound regulatory regions are required for target gene regulation, we performed UMI-4C experiments using promoter regions of T, Gata6, Foxa2 and Gata4 as viewpoints in ESCs and D6 EBs, comparing Whsc1+/+ and −/− cells. We found that most Whsc1-bound putative mesendodermal enhancers appear to interact with target gene promoter regions in Whsc1+/+ D6 EBs but not in ESCs, and that these contacts are lost/decreased in Whsc1−/− D6 EBs (Fig. 5c and Supplementary Fig. 5c). These observations are consistent with the notion that the relevant regions are involved in Whsc1 mediated target gene regulation.

Figure 5. Whsc1 controls the enhancer activity of mesendoderm transcription factors.

a, Top panel: HiC contact maps at 10kb resolution from ESCs around T (Left) and Gata6 (right) loci (GSE96611). TADs are marked by solid black lines and subTADs are indicated by dashed black lines. The regions analysed using H3K27ac ChIP-seq data are indicated as black boxes; Bottom panel: H3K27ac ChIP-seq profiles on the loci of T (left) and Gata6 (right) in neural progenitor cells (NPC) (GSE35496), Eomes+ mesendodermal progenitors (GSE103262), Activin A-induced mesendodermal precursors (MP) (GSE38596) and Flk1+ mesendodermal progenitors (GSE47082). The black arrows below each panel correspond to putative regulatory regions up or downstream of the respective gene. b, ChIP-qPCR quantification of Whsc1 occupancy on the same enhancers as shown in Fig. 5a above in D6 EBs from Whsc1 +/+ and Whsc1 −/− cells, with numbers indicating distance to the TSS in kb. Data represent mean±s.d. from n=3 independent experiments and p-values were calculated by two-tailed unpaired t-test. c, Replicated UMI-4C profiles for baits located on the T (Chr17:843375) (left) and Gata6 (Chr18:11053621) (right) promoters assayed in Whsc1 +/+ and Whsc1 −/− ESCs and D6 EBs. Top panel, average contact profiles generated from the average of two independent biological replicates; bottom panel: average contact fold change D6 EBs versus D0 ESCs from two independent biological replicates. d, H3K27ac enrichments on putative enhancers of T (left) and Gata6 (right) were quantified by ChIP-qPCR in Day 6 Whsc1 +/+ and Whsc1 −/− EBs. Data represent mean±s.d. from n=3 independent experiments and p-values were calculated by two-tailed unpaired t-test.

To test whether Whsc1 binding is associated with enhancer activity, we monitored H3K27ac enrichment at Whsc1-bound regions in Whsc1+/+ and −/− D6 EBs within the T, Gata6, Foxa2 and Gata4 loci. Regions not bound by Whsc1 served as controls. All 10 Whsc1-bound regions showed a significant decrease of H3K27ac decoration in the Whsc1−/− cells. In contrast, in the same cells 14 out of 16 regions not bound by Whsc1 and all 12 sites within the Sox1 and Pax6 regions tested showed no significant H3K27ac change (Fig. 5d and Supplementary Fig. 5d). Similar results were obtained for the H3K4me2 mark (Supplementary Fig. 5e). In addition, the 10 Whsc1-bound regions showed no significant binding of Whsc1 in ESCs (Supplementary Fig. 5f) and their H3K27ac enrichment in ESCs was lower than in D6 EBs (Supplementary Fig. 5f).

Taken together, these data show that Whsc1 binding is associated with active enhancers of target genes.

Whsc1 interacts with Brd4 to facilitate activation of mesendoderm enhancers

The bromodomain protein Brd4 has been reported to be required for enhancer activation and gene expression ³⁵ and to interact with Whsc1 ³⁶. We therefore tested whether Brd4 interacts with Whsc1 at the enhancer regions described above (Fig. 5b and Supplementary Fig. 5b). We found that in D6 Whsc1+/+ EBs Brd4 indeed binds to the enhancers of T, Gata6, Foxa2 and Gata4 as well as to those of Sox1 and Pax6. Importantly, Brd4 binding was significantly decreased in Whsc1−/− EBs at the mesendoderm enhancers but not at the ectoderm enhancers (Fig. 6a). In addition, EBs treated during their formation with the Brd4 inhibitor JQ1 ³⁷ showed reduced expression of the mesendodermal regulator genes Gata4, T, Gata6 and Foxa2, compared to DMSO treated control EBs (Fig. 6b). We also observed an increase of neuroectoderm markers Sox1, Pax6 and Nestin in JQ1-treated EBs, in line with the reported induction of neural differentiation after Brd4 depletion in ESCs ³⁸ (Fig. 6c). Moreover, both Whsc1 isoforms were found to co-immunoprecipitate with Brd4 in D6 EB extracts (Fig. 6d), consistent with the notion that the N-terminus of Whsc1 is sufficient for this interaction.

Figure 6. Whsc1 interacts with Brd4 and Gata6 on enhancers of mesendoderm transcription factors.

a, ChIP-qPCR quantification of Brd4 occupancy on Whsc1-bound putative enhancers of T, Gata6, Gata4 and Foxa2 and control regions (Sox1 and Pax6) in Day 6 Whsc1 +/+ and Whsc1 −/− EBs. Data represent mean±s.d. from n=3 independent experiments and p-values were calculated by two-tailed unpaired t-test. b-c, Expression of mesendodermal markers (b) and ectodermal markers (c) quantified by RT-qPCR in Day 3 control or JQ1-treated EBs. Data represent mean±s.d. from n=3 independent experiments and p-values were calculated by two-tailed unpaired t-test. d, Western blots of endogenous Whsc1 (~180 and ~100kDa isoforms) and Brd4 co-immunoprecipitated from total protein extracts of Day 6 EBs. Inputs correspond to 10% of total extract. Two independent experiments were performed with similar results. Scanned images of unprocessed blots are shown in Supplementary Fig. 8. e, Western blots of endogenous Whsc1 (~180 and ~100kDa isoforms) and Gata6 co-immunoprecipitated from total protein extracts of Day 6 EBs. Inputs correspond to 10% of total extract. Two independent experiments were performed with similar results. Scanned images of unprocessed blots are shown in Supplementary Fig. 8. f. ChIP-qPCR quantification of Whsc1 occupancy on Whsc1-bound putative enhancers of T, Gata6, Gata4 and Foxa2 in Day 6 Gata6 +/+ and Gata6 −/− EBs. Data represent mean±s.d. from n=3 independent experiments and p-values were calculated by two-tailed unpaired t-test.

Together, our results show that Whsc1 interacts with Brd4, and that this interaction favours the activation of mesendoderm enhancers but not of ectoderm enhancers.

Gata6 recruits Whsc1 to a subset of mesendodermal enhancers

We next asked whether mesendoderm specific regulators are also able to interact with Whsc1 and recruit the factor to their target sites. As a model, we chose to study Gata6 and first performed co-immunoprecipitation experiments. Indeed, both long and short Whsc1 isoforms co-immunoprecipitate with Gata6 in D6 EBs, suggesting that the SET domain, not contained in the 100kDa isoform, is dispensable for this interaction (Fig. 6e). To probe the second question, we generated Gata6 knockout ESCs by CRISPR-Cas9 mediated genome editing. These cells retain an ESC-like phenotype when grown in LIF- and 2i- containing medium. The Gata6−/− ESCs were then used to generate EBs and to perform Whsc1 ChIP experiments in D4 EBs. We tested all 10 previously identified Whsc1-bound enhancer regions of T, Gata4, Gata6 and Foxa2 and found significantly decreased binding of Whsc1 at 6 of these regions (Fig. 6f).

These findings suggest that the mesendodermal lineage specifier Gata6 is capable of recruiting Whsc1 to a subset of mesendoderm specific enhancers.

Depletion of mesendoderm TFs in ESCs causes a delay in pluripotency exit

The finding that Whsc1 is associated with the activation of Gata4, T, Gata6, and Foxa2 during ESC differentiation raised the possibility that the encoded mesendoderm instructive TFs are downstream mediators of Whsc1 in pluripotency exit and germ layer specification. To test this hypothesis, we transfected Rex1GFPd2 ESCs with siRNAs against Gata4, T, Foxa2 and Gata6, respectively, resulting in significant downregulation of the genes (Supplementary Fig. 6a). The cells were then induced to differentiate by transfer into N2B27 medium containing Activin A and FBS. While control siRNA cells showed a complete loss of GFP expression at 48 hours p.i., a significant proportion of cells transfected with siRNAs against Gata6, T, Gata4 and Foxa2 remained GFP positive (Supplementary Fig. 6b). To further study the role of these factors in pluripotency exit, we tested Gata6 and Foxa2 knockout ESCs generated by CRISPR-Cas9-mediated genome editing. Monitoring gene expression at different times after induction of differentiation revealed that, as for Whsc1−/− cells, Gata6−/− and Foxa2−/− cells showed a delayed down-regulation of the pluripotency genes Oct4, Rex1 and Nanog (Fig. 7a). The retention of Oct4 and Nanog expression in Gata6 and Foxa2 knockout cells was also confirmed at the protein level in cells induced to differentiate for 72 hours (Fig. 7b). Moreover, when these cells were induced to generate EBs, they showed a delayed down-regulation of pluripotency markers similar to that found with Whsc1−/− EBs (Fig. 7c)

Figure 7. Mesendoderm transcription factors are required for efficient pluripotency exit.

a, Expression kinetics of pluripotency genes in Whsc1 −/−, Gata6 KO, Foxa2 KO and control ESCs (WT) monitored by RT-qPCR after transfer into N2B27 medium containing Activin A and FBS Data represent mean±s.d. from n=3 independent experiments. b, Western blot analysis of Oct4 and Nanog in cells described in (a). Tubulin was used as loading control. Two independent experiments were performed with similar results. Scanned images of unprocessed blots are shown in Supplementary Fig. 8. c, Expression kinetics of pluripotency genes evaluated by RT-qPCR during EB formation using Whsc1 −/−, Gata6 KO and Foxa2 KO and control ESCs (WT). Data represent mean±s.d. from n=3 independent experiments.

In conclusion, our data have shown that the depletion or knockout in ESCs of the mesendoderm regulators Gata6, T, Gata4 and Foxa2 downstream of Whsc1 delays pluripotency exit.

Discussion

Here we describe that the chromatin-related factor Whsc1 has a role in both pluripotency exit and germ layer specification of ESCs. Depletion of Whsc1 entraps the cells in a pluripotent state and inhibits their specification towards mesoderm and endoderm lineages, without impairing neuro-ectoderm formation. Mechanistically, Whsc1 binds to and activates enhancers of mesendoderm lineage-instructive TFs during lineage specification, independently of its methyltransferase activity mediated by the SET-domain. Moreover, depletion of downstream mesendodermal regulators also delayed pluripotency exit, explaining, at least in part, the effects observed after Whsc1 ablation.

Several factors previously described to be required for pluripotency exit have been shown to dismantle directly the pluripotency-associated regulatory network. Hence, folliculin together with its interaction partners Fnip1/2 sequesters the pluripotency associated TF Tfe3 in the cytoplasm ¹³; the TFs Tcf3, Foxd3 and Zfp706 repress pluripotency gene expression ^8–11; and the RNA binding proteins Mettl3 and Pum1 destabilize pluripotency gene transcripts ^{11, 12, 32}. In contrast, we have found that Whsc1 does not act primarily on the pluripotency network, but controls the up-regulation of mesendodermal regulators, including Gata4, T, Gata6 and Foxa2, which on their own also control pluripotency exit. Our observations suggest that the regulatory networks controlling pluripotency and germ layer specification are intimately linked. Indeed, several studies have shown that during murine gastrulation in the primitive streak the pluripotency factors Nanog, Oct4 and Sox2 are co-expressed with the mesendoderm lineage factors T ³⁹, Gata6 ⁴⁰ and Foxa2 ⁴¹. Moreover, it has been shown that pluripotency TFs are crucial for the up-regulation of germ layer genes upon differentiation, as differentiation fails when pluripotency TFs are acutely ablated ^{7, 42–44}. In addition, one study also suggests that during in vitro differentiation of ESCs GATA6 represses pluripotency factor expression ⁴⁵.

Our findings support the notion that Whsc1 is recruited to chromatin by lineage regulators. Thus, co-immunoprecipitation experiments showed that Whsc1 interacts with Gata6 and that it loses its ability to bind to most of the mesendodermal enhancers in Gata6 deficient embryoid bodies. It is possible that other sites bound by Whsc1 additional regulators are operative, such as has been described for Nkx2.5 in cardiac cells ³¹. Our study also showed an interaction between Whsc1 and the transcriptional co-activator Brd4 in EBs, supporting earlier studies ³⁶. Moreover, EBs lacking Whsc1 showed a reduced binding of Brd4 specifically at mesendodermal enhancers suggesting that Whsc1 facilitates the recruitment of Brd4 to these enhancers, leading to their activation. The finding that the BRD4 inhibition by BET inhibitor JQ1 impairs the upregulation of mesendoderm regulators while dramatically enhancing the expression of ectodermal regulators supports previous observations in ESCs ³⁸. Together, our data suggest that Whsc1 enables Brd4 to act as a coactivator of mesendodermal enhancers.

In certain forms of cancer, including multiple myeloma and acute lymphoblastic leukaemia, WHSC1 has been found to be either overexpressed or hyper-activated, resulting in an increased methylation of H3K36 on promoters of oncogenes that drive the disease ^{29, 30, 46}. This is mediated by the methyltransferase activity of the SET-domain. In contrast, a short isoform of NSD3 (another member of the Nsd family), lacking the SET-domain, can act together with Brd4 on enhancers of genes that drive acute myeloid leukemia ⁴⁷. Interestingly, the PWWP domain responsible for this interaction is highly conserved between members of the Nsd family, and in Whsc1 it is located within the N-terminus that interacts with Brd4 ⁴⁸.

Wolf-Hirschhorn syndrome patients with WHSC1 deletions often exhibit symptoms characteristic of midline defects, including craniofacial malformations and heart defects ⁴⁹. These features have been attributed to the SET-domain of the protein, since mice lacking this domain of Whsc1 largely reflect this phenotype, exhibiting aberrant craniofacial, cardiac and cartilage structures ^{31, 50}. It will now be interesting to elucidate whether the function of Whsc1 described here is also involved in gene regulation during embryo development, cardiogenesis and cancer.