Induction of c-Jun during the naïve to primed transition

To understand how the naïve state transitions to the primed one, we cultured mouse ESCs in naïve conditions containing LIF and two inhibitors (2i) CHIR99021 (WNT signal pathway inhibitor) and PD0325901 (ERK inhibitor) [13]. Naïve ESCs were differentiated to primed EpiSCs (converted cells are referred to as EpiSC-like cells) by using a protocol involving basal medium containing four small molecules: Activin A, bFGF, XAV939 and IWR [27, 28]. Morphologically, the embryo-derived (EpiSCs) or induced EpiSC-like cells are nearly identical (Fig. 1A, right panels and 1B). Principal component analysis (PCA) on time-course RNA-seq data during the naïve to primed transition (NPT), showed a clear trajectory from ESCs to EpiSCs (Fig. 1B). Naïve-specific pluripotent genes, such as Nanog, Esrrb, Dppa5a were rapidly downregulated, conversely, the primed-specific marker genes Lin28a, Fgf5, T were activated only in the late stage of NPT (Fig. 1C). Interestingly, at both the RNA and protein level, we show that c-Jun was expressed as early as D2 (Day) during NPT, and highly expressed in EpiSC-like cells (Fig. 1D, E).

To study the role of c-Jun in the NPT, we designed two sgRNAs that target two flanking regions of the single exon of c-Jun (Additional file 1: Fig. S1A). We transfected those two sgRNAs into ESCs to knock out c-Jun. As c-Jun is not expressed in ESCs, we converted them to the primed state and confirmed the knockout by Western blot (Fig. 1F). We next looked at RNA-seq of the c-JunKO EpiSC-like cells, and found that the transcriptomes were nearly identical: EpiSC-like versus EpiSCs, R = 0.959; and c-JunKO EpiSC-like versus EpiSCs, R = 0.947 (Additional file 1: Fig. S1B, C). As expected, c-JunKO EpiSC-like cells express normal levels of both pluripotent (Pou5f1, Nanog) and EpiSC-specific (Fgf4) genes that remained expressed at similar levels when c-Jun was knocked out (Additional file 1: Fig. S1B). c-JunKO EpiSC-like cells also clustered by PCA closely with normal EpiSCs and EpiSC-like cells (Fig. 1B). In the ATAC-seq data, chromatin accessibility of c-JunKO EpiSC-like cells was similar with WT EpiSC-like cells and EpiSC (Additional file 1: Fig. S1D). Together these results indicate that c-Jun is not required for NPT nor primed state despite of being unregulated robustly.

c-Jun knockout promotes primed to naive transition

Whilst c-Jun is not required for the NPT, we next explored if c-Jun is needed for the reverse, primed to naïve transition (PNT). To explore this, we exploited the ability of exogenously expressed Klf4 to drive the PNT and established a Klf4-transgenic mEpiSC line (kEpiSCs), that bears two transgenes: An Oct4-GFP reporter system where GFP is only expressed only in the naïve state [29], and Klf4 under the control of a CAG promoter that converts EpiSCs to ESCs [23] (Additional file 1: Fig. S2A). When Klf4 was expressed, however, kEpiSCs remained in the primed state if they remained in primed culture medium, and there was no significant difference in the expression of Fgf5, Pou5f1(Oct4), Nanog, Esrrb and Sox2 (Additional file 1: Fig. S2B, C). This suggests that Klf4 expression alone is insufficient to bring about the PNT. We then evaluated the PNT efficiency in three culture media: 15%FBS + 2i + LIF (serum + 2i + LIF medium, primed ESC culture medium), N2B27 + 2i + LIF (naïve mouse ESCs culture medium), and iCD1 (a high-efficiency somatic reprogramming medium) [30]. Comparatively, GFP + cells emerged earlier in iCD1 medium (1.8% at day 3), compared to 0.01/0.04% for the other media (Additional file 1: Fig. S2D, E). Consequently, we used iCD1 for the subsequent PNT experiments. iCD1 was also more efficient in activating the expression of pluripotent marker genes such as Sox2 and Esrrb which are known to remodel the chromatin structure during PNT [20, 30] (Additional file 1: Fig. S2F). PCA of the RNA-seq time course during PNT showed that the cells appeared to traverse an alternative path back to naïve ESCs, and did not retrace the NPT (Fig. 1B). Additionally, there was a spike in c-Jun expression in the PNT day 3 cells (Fig. 1D), which is likely due to cell heterogeneity as the PNT is an inefficient process. c-Jun expression is up-regulated at day 3 but is ultimately silenced in the resulting reverted iPSCs (Fig. 1D). These results suggest that the iCD1 medium is suitable for the conversion of primed EpiSCs to naïve ESCs.

Fig. 1

c-Jun inhibits primed to naïve transition. A. Schematic for the derive mouse naive ESC, primed EpiSC and induce naïve ESC differentiation into EpiSC-like cells (naive to primed transition, NPT) under EpiSC culture medium (see methods). And induce primed to naive transition (PNT) (see methods). D indicates the day of the naive to primed transition or primed to naive transition. B. Principal Component Analysis (PCA) of the gene expression (RNA-seq) during naive to primed transition (NPT) and primed to naive transition (PNT). C. Heatmap of naive and primed marker genes expressed during NPT. D. RNA-seq data show the c-Jun expression levels in ESC, a typical naive to primed transition and primed to naive transition time course, EpiSC-like cells, EpiSC and iPSCs. D = day. E. Western blot of the levels of c-Jun protein in ESC, a typical naive to primed transition time course, and EpiSC. D = day. F. Western blot of c-Jun protein in WT ESC, EpiSC-like, c-JunKO ESC and c-JunKO EpiSC-like cells. G. Flow cytometry analysis of the Oct4-GFP (OG2) reporter at primed to naïve transition (PNT) day 3. D = day. Scale bar = 100 μm. H. Effect of knock out of c-Jun in day 3 primed to naive transition. The number of GFP positive colonies was counted on day 3. D = day. Data from 6 biological replicates from 2 independent experiments and shown as the mean ± SEM. *** indicates P value < 0.001, t-test analysis between the c-Jun knock out EpiSClike and WT EpiSC-like cells.

To explore the impact of c-Jun on the PNT, we next performed PNT experiments on the c-JunKO EpiSC-like cells in iCD1 medium and measured the proportion of GFP positive cells at day 3 (D3) as a readout of PNT efficiency. Interestingly, we found that knock out of c-Jun significantly promoted the conversion of primed EpiSCs into naïve iPSCs, as measured by the number of GFP positive cells (2.7% versus 13.7% in the c-JunKO cells; Fig. 1G). Counting the resultant GFP + colonies indicated the conversion efficiency of c-JunKO EpiSC-like cells to ESCs was ~ 4 times higher than WT EpiSC-like cells on day 3 (Fig. 1G, H). Overall, these results suggest that c-Jun acts as a one-way valve to enable the PNT. In ESCs, c-Jun is not expressed and is not required for the NPT or EpiSCs, yet when the cells went through the PNT the loss of c-Jun drastically increased the efficiency of this conversion.

c-Jun N-terminal kinase inhibitor promotes the primed to naïve transition

Loss of c-Jun could dramatically improve the efficiency of the PNT. We next explored if small-molecule inhibition of upstream factors of c-Jun could also promote the PNT. We used a JNK inhibitor (JNKi, SP600125), which can indirectly inhibit c-Jun activity, to validate the negative function of c-Jun in PNT (Fig. 2A). Indeed, our results showed that the PNT was improved by JNKi in a dose-dependent manner, reaching almost 10% GFP + cells and ~ 1200 Oct4-GFP-positive colonies from 12000 starting cells on day 3 (D3) (Fig. 2B, C). A similar dose-dependent effect of JNKi could be seen when using two further independent lines of EpiSCs (B1 and B2) (Fig. 2D). The resulting iPSCs had normal naive morphology (Fig. 2E), karyotype (Fig. 2F), and were capable of generating chimeras with germline transmissions (Fig. 2G). These results suggest that inhibition of JNK phenocopies c-Jun KO in PNT.

Fig. 2

c-Jun N-terminal Kinase inhibitor facilitates primed to naïve Transition. A. Work model shows JNK inhibitor (SP600125) in primed to naive transition (PNT). B, Flow cytometry analysis represent reprogramming efficiency increased with a higher dose of JNKi. C, D. A dose-response graph for JNKi in promoting primed to naive transition. This experiment was performed with three different EpiSC cell lines called EpiSC B4 (this cell line was named EpiSC in the text, and used for this study), EpiSC B1, EpiSC B2. JNKi was used with 3 doses for 3 days during primed to naive transition (PNT) and scored for GFP positive colonies. Data are from 6 biological replicates in 2 independent experiments and shown as the mean ± SEM. *** indicates P value < 0.001, one-way ANOVA with Sidak correction between the DMSO and the dosing groups. E. Phase and GFP images of iPSC. Scale bar = 100 μm. F. Karyotypes of iPSC. G. Chimeric mice from iPSC.

Meanwhile, we found knock out other AP-1 family members such as Fosl2 also enhanced iPSCs formation (Additional file 1: Fig. S2G). On the contrary, PNT process was repressed when c-Jun was highly expressed in primed EpiSC (Additional file 1: Fig. S2H, I). Together, these results indicate that c-Jun is induced, but not needed for the naïve to primed transition or in EpiSCs. However, c-Jun functions as a barrier to the primed to naïve transition.

c-Jun binds to closed naïve enhancers in EpiSCs

To explore the mechanism of how c-Jun impairs the PNT, we began by mapping the chromatin accessibility landscapes of both naïve ESCs and primed EpiSCs using ATAC-seq [31, 32]. 63135 peaks remained open in both cell types, but there was a surprisingly large number of ESC-specific (53010 peaks) or EpiSC-specific peaks (53591 peaks) (Fig. 3A). DNA motif enrichment analysis indicated that ESCs and EpiSCs had distinct transcription factor (TF) motifs associated with open chromatin in both states (Fig. 3B). As expected, OCT4 was common between the two cell types, including the compound OCT4-SOX2 motif that is instrumental in controlling the pluripotent state [33]. However, SOX, and especially KLF motifs, were specific to ESCs, whilst BACH, ATF and especially AP-1 family motifs were enriched in the EpiSCs (Fig. 3B). Indeed, the motif enrichment in naïve ESCs closely mirrored known regulatory programs, for example, TFCP2L1 and KLF motifs were specific to naïve, and when Tfcp2l1 and Klf5 are overexpressed they can drive EpiSCs to naïve ESCs in a PNT [34, 35]. Intriguingly, considering that c-Jun has no apparent function in EpiSCs, the AP-1 family of TF motifs (FOSL1, FOSL2, c-JUN) were enriched in primed-specific open chromatin (Fig. 3B). This enrichment of AP-1 family motifs led us to hypothesize that despite their apparent lack of function in EpiSCs, they may nonetheless be important in controlling the primed state.

Fig. 3

c-Jun bind to closed naïve enhancers in EpiSCs. A. ATAC-seq datasets from mouse EpiSC and ESC presented as open or closed state as described [9]. Top scale indicates normalized tag count in color shades. Note: primed for loci open in EpiSC; naive for loci open in ESC; Bothopen as open loci in both ESC and EpiSC. B. TF motifs enriched at least 1.5-fold higher than background and P value less than 0.01 for each category of ATAC-seq peaks defined in panel A. *P value < 10–20 from HOMER. C. Heatmap shows naive and primed loci closed and open during naive to prime transition in vitro at different time points. D indicates day of the differentiation process. Naive and primed ATAC-seq peaks defined in A. And c-Jun ChIP-seq data was performed in EpiSC, which divide naive and primed peaks into four groups. D. Pie chart presenting the number of peaks c-Jun occupied in primed and naive loci. E. Motif analysis showing pluripotent transcription factors that enriched in C3 peaks (peaks defined in C). F. Heatmap of sequence read density for ATAC-seq and H3K27ac, H3K27me3, H3K4me3 data in ESC and EpiSC signal on C3 peaks and pluripotency transcription factors Esrrb, Klf4, Nr5a2, Sox2 and Tfcp2l1 binding profile in C3 peaks. Sequenced reads ranked by mean signal strength. Windows are centered on the ATAC-seq pea summit. Each row of the heatmap is a genomic locus. The ChIP-seq data of H3K27me3 and H3K4me3 in ESCs were taken from GSE80280 [51], the ChIP-seq data of H3K27ac in ESCs was taken from GSE98404 [9] and the ChIP-seq data of H3K27me3, H3K27ac, H3K4me3 were taken from GSE57407 [52]. The ChIP-seq data of Sox2 and Klf4 were taken from GSE90893 [8], the ChIP-seq data of Esrrb and Tfcp2l1 were taken from GSE11431 [53] and the ChIP-seq data of Nr5a2 was taken from GSE1901 [54]. G. Violin plots for the normalized ATAC-seq tag density for all peaks in C3 and C4 groups. Data were converted to a Z score to the emphasis change. Data were converted to a Z score based on the row-wise SD for each peak. P value was calculated by the Mann-Whitney U test. * means P < 0.0005.

To explore this, we performed ChIP-seq for c-Jun in EpiSCs (Additional file 1: Fig. S3A) and analyzed the binding sites with GREAT [36]. GREAT analysis reveals that c-Jun was bound to genes involved in normal cellular activity, such as apoptotic signaling, mitochondrial membrane organization, and adherens junction organization (Additional file 1: Fig. S3B). Unexpectedly, c-Jun was enriched at naïve-specific genes that are silenced in EpiSCs, for example, c-Jun was bound close to the genes, Dppa5a, Zfp42, and Esrrb (Additional file 1: Fig. S3C, D). Indeed, when we compared c-Jun binding sites to chromatin-accessible dynamics during NPT, we identified four groups: naïve closed but primed open with c-Jun bound (C1, 3521 peaks); naïve closed but primed open without c-Jun bound (C2, 50079 peaks); naïve open but primed closed and with c-Jun bound (C3, 2464 peaks); and naïve open but closed in primed, and without c-Jun (C4, 50546 peaks) (Fig. 3C, D). The most interesting group was C3, which had a distinctive epigenetic pattern: these loci were open in ESCs and were marked by enhancer-associated histone modifications H3K4me3, H3K27ac (Fig. 3F). However, in EpiSCs, these same loci are bound by c-Jun, their chromatin is closed and they have lost associated enhancer histone marks (Fig. 3C–F). This suggests those loci function as naïve-specific enhancers, that are silenced during NPT (Fig. 3C–F). Indeed, C3 loci are rapidly closed during the NPT as early as day 2 (Fig. 3C). Gene ontology analysis agreed with this designation and assigned the genes in C3 as related to transcription, cell proliferation and somatic stem cell population maintenance (Additional file 1: Fig. S3E). When we performed motif discovery on the C3 loci, motifs for many naïve-specific transcription factors (TFs): Klf4, Esrrb, Sox2, Nr5a2, and Tfcp2l1 were enriched (Fig. 3E). Analysis of ChIP-seq for these naïve-specific TFs in ESCs indicated that the C3 group loci are bound by all five of these TFs in ESCs (Fig. 3F). These results indicate that C3 loci are naive-specific enhancers near naïve-specific genes, that are bound by c-Jun in EpiSCs and are silenced.

We next looked if the loss of c-Jun altered chromatin at C3 in the c-JunKO EpiSC-like cells. Surprisingly, although consistent with the phenotype that c-Jun has minimal effects on EpiSCs, c-Jun deficiency had only a minimal impact on C3 loci closure during NPT (Fig. 1B). There was however a significant (if modest) increase in chromatin accessibility in the c-JunKO EpiSC-like cells (red dash line, Fig. 3G). This suggests that the C3 loci may be slightly more prone to opening, which may help explain why the loss of c-Jun promotes the PNT.

c-Jun inhibits naïve loci opening during primed to naïve transition

To directly test the function of c-Jun in the primed to naïve transition (PNT), we analyzed the transcriptome and found that c-Jun deficiency promotes the up-regulation of 3746 genes during the PNT, amongst which 1688 genes are highly expressed in ESCs, and activated in c-Jun knock out cells at PNT day 3 (D3) (Additional file 1: Fig. S4A, B). Gene ontology analysis showed that those genes are involved in transcription, histone modification and stem cell population maintenance, and inner cell mass cell proliferation, and includes naïve-specific or embryonic-specific genes such as Dppa5a, Zfp42, Klf4, Gdf3, Kdm3a, Kdm5b, Sap30, Sall4 (Additional file 1: Fig. S4C, D). When we looked at chromatin accessibility dynamics, we found that c-Jun deficiency also promoted the opening of naïve-related loci, especially amongst the C3 group (Fig. 4A). For example, enhancers close to the Dppa5a and Zfp42 genes undergo considerably faster opening in the c-JunKO cells, compared to the wild type cells at day 3 (D3) (Fig. 4B). These changes in chromatin accessibility were accompanied by similarly rapid inductions of gene expression (Fig. 4B).

Fig. 4

c-Jun represses PNT by maintain the closed naïve enhancers. A. Violin plots for the normalized ATAC-seq tag density for all peaks within the indicated primed (C1 and C2) and naive (C3 and C4) groups in WT EpiSC-like, c-Jun knock out EpiSC-like, PNT-D3 WT EpiSC-like, PNT-D3 c-Jun knock out EpiSC-like and ESC. D = day. Data were converted to a Z score based on the row-wise SD for each peak. B. Selected genome views for the ATAC-seq data and c-Jun ChIP-seq data in EpiSC, WT EpiSC-like, c-Jun knock out EpiSC-like, PNT-D3 WT EpiSC-like, PNT-D3 c-Jun knock out EpiSC-like and ESC. D = day. Genome views: Dppa5a (chr9:78,363,039–78,377,161), Zfp42 (chr8:43,293,267–43,309,209). The RNAseq expression values for the respective genes is shown in the bar plot below the genome view. RNA-seq data are from 2 biological replicates, expression units are in normalized tag counts. C. Violin plots for the normalized ATAC-seq tag count for the ATAC-seq data for C3 peaks, for EpiSC, PNT-JNKi-D3, PNT + JNKi-D3 and ESC. Data were converted to a Z score based on the row-wise SD for each peak. P value was from a Mann–Whitney U test. * means P < 0.0005. D. Scatterplot showing the difference between transcriptional profiles with or without JNKi at day 3 (D3) in PNT. JNKi promotes naive marker genes Nanog, Esrrb, Tfcp2l1 and Klf2 activation during PNT

As we have shown that JNKi could promote PNT efficiency (Fig. 2), we also performed ATAC-seq in the JNKi-treated cells. The ATAC-seq data shown that JNKi also accelerate the opening of the C3 peaks (Fig. 4C). Consistently, JNKi activates the expression of Esrrb, Tfcp2l1, Klf2, Nanog in advance of the untreated PNT cells (Fig. 4D). These results indicate that loss of c-Jun or inhibition of JNK leads to a more rapid opening of chromatin at naïve-specific enhancers, and especially at the C3 group of genomic loci.

Esrrb promotes PNT by opening c-Jun locked naïve loci

Our data suggest that c-Jun is required to suppress a class of naive-specific enhancers in EpiSCs, and the loss of c-Jun leads to priming of those loci and more rapid opening and gene expression in the PNT. To explore this further, we looked at the C3 loci, which were also enriched with the naïve-specific TFs Klf4, Esrrb, Nr5a2 and Tfcp2l1 (Fig. 3E, F). These four TFs have been implicated in earlier reports to mediate the PNT [20, 23, 24, 34]. Thus, we hypothesized that the loss of c-Jun primed the C3 group of loci and allowed easier access for these naïve-specific TFs. We co-transfected PiggyBac (integration-capable) plasmids bearing one of Esrrb, Klf4, Klf2, Nr5a1, Sox2 or Tfcp2l1 into EpiSCs, and then performed PNT. PNT was quantified by flow cytometry of the number of GFP + cells on day 3 (D3). Of the factors we transfected, Esrrb, Klf4, Klf2 and Nr5a1 could improve PNT, albeit to varying degrees of efficiency (Additional file 1: Fig. S5A, B). Interestingly, Esrrb could accelerate the PNT, and up to 15% of cells were Oct4-GFP-positive by day 3 (D3) (Additional file 1: Fig. S5B). Conversely, transfection of Sox2 and Tfcp2l1 could not improve the PNT (Additional file 1: Fig. S5B). To help explain the apparent differences in efficiency between the transgenes, we generated ATAC-seq data for all 6 factors during the PNT, and looked at the chromatin accessibility of C3 peaks at day 3 (PNT-D3). We found that, in agreement with the PNT efficiencies, Esrrb promoted the opening of C3 peaks more efficiently, while the others transgenes were less capable or were inefficient (Additional file 1: Fig. S5C). Overall, these results suggest that the loss of c-Jun primes naïve loci for opening, which leads to accelerated chromatin changes in combination with naïve-specific TFs (Additional file 1: Fig. S5D).

c-Jun regulates chromatin accessibility in EpiSCs

As we have shown c-Jun is not needed for the NPT, but it impedes the PNT and chromatin is more accessible in EpiSC-like cells when c-Jun was knocked out (Figs. 3G, 4A). This suggests that c-Jun might regulate EpiSC cell fate by locking naïve-specific enhancers. To study the mechanistic role of c-Jun in EpiSCs, we analyzed c-Jun ChIP-seq data in EpiSCs and found that the distribution of c-Jun appears to mainly localize at intragenic (39%) or intergenic (37%) regions at with only 16% at promoter regions. With a combined ~ 76% outside of coding regions [10], we hypothesized that c-Jun may participate in the regulation of chromatin dynamics in EpiSCs. However, the genome-wide pattern of chromatin accessibility was correlated between c-JunKO EpiSC-like cells, WT EpiSC-like cells and EpiSC (Additional file 1: Fig. S1D). When we looked at the accessibility of naïve and primed peaks in the c-JunKO EpiSC-like cells and WT EpiSC-like cells (peaks defined in Fig. 3A), we found the accessibility increased in both primed and naive peaks when c-Jun was knocked out (Fig. 5A, B). Among the C1 ~ C4 peaks, C1 and C3 were bound by c-Jun and the chromatin accessibility was elevated in the c-Jun knockout. This was especially evident in the C3 peaks which we previously identified as naïve-specific and bound by c-Jun (Figs. 3C, 5B, top-right). Meanwhile, C2 and C4 loci, which were not bound by c-Jun (Fig. 3C), also shown elevated levels of accessible chromatin in c-JunKO EpiSC-like cells (Fig. 5A, B). These data suggest that c-Jun is regulating chromatin at a specific subset of naïve-specific loci, and the loss of c-Jun leads to incomplete chromatin closure.

Fig. 5

c-Jun regulates chromatin accessibility in EpiSCs. A. Heatmap of sequence read density for ATAC-seq data in EpiSC-like and c- Jun knock out EpiSC-like signal on both open, primed open and naive open category defined in Fig. 3A, ranked by mean signal strength. Windows are centered on the ATAC-seq peak summit. Each row of the heatmap is a genomic locus. B. Pileup of ATAC-seq read density at the naive and primed peaks in EpiSClike and c-Jun Knock out EpiSC-like. C. Boxplot of the expression level for all genes with a TSS within 10 kb of an ATAC-seq peak in C1 ~ C4 loci. Data were converted to a log2 value for each gene. D. Volcano plot show the expression of the genes related to C3 peaks, mapping the high expressed genes in EpiSC-like (left) and high expressed genes in c- JunKO EpiSC-like (right)

We next looked at the impact of the loss of c-Jun and changes in chromatin on gene expression. At the transcriptome level, we overlapped the genes annotated within 10 kb of a loci in C1 ~ C4 loci and shown similar expression pattern in EpiSC, EpiSC-like cells and c-JunKO EpiSC-like cells (Fig. 5C and Additional file 1: Fig. S1B). When detailed analysis the gene related to C3 locus, less genes were activated in c-Jun knockout EpiSC-like cells, such as Cdh1, E2f7 and Fat1 were up-regulated when c-Jun defect (Fig. 5D). These results indicate that c-Jun is required to maintain closed chromatin at a class of naïve-specific genes. Loss of c-Jun induces leaky expression of these genes which helps potentiate the EpiSCs for the PNT.

Collectively, our data indicate that c-Jun plays a stabilizing role in locking the silenced state of naïve loci in the primed state, and this helps block the primed to naïve transition and maintain EpiSCs cell identity (Fig. 6).

Fig. 6

c-Jun functions as a lock to maintain closed naïve enhancers and represses the PNT. A model depicting the chromatin accessibility dynamics in naive to primed transition (NPT) and primed to naive transition (PNT). During NPT naïve enhancers were closed and bound by c-Jun. c-Jun functions as a lock to keep the naive enhancers in a closed state. However, if c-Jun is removed, the lock is loosened and the closed naive enhancers are primed and ready to recruit key naive-specific TFs to reactivate the naive gene regulatory network during the primed EpiSC to naive ESC transition