Chromatin accessibility dynamics during JUN-induced PST
Taking advantages of the incompatibility between JUN and pluripotency, we constructed a robust PST system, JUNTetON ESC, in which 90% mESCs colonies (~ 90%) will exit from pluripotency within 8 h [15], providing us an ideal platform to investigate the underline mechanisms for cell-fate transition with time scale in hours. To evaluate the non-specific effect for Doxycycline (Dox) concentration, we test the response of WT ESC by 2 μg/ml Dox treatment. No significant change was observed in either the expression of pluripotent/AP-1 related genes (Additional file 1: Fig. S1A) or the morphology of ESCs (Additional file 1: Fig. S1B). In addition, knockdown JUN by shRNA in JUNTetON ESC, can partially rescue Dox induced PST (Additional file 1: Fig. S1C). Those data indicate that the non-specific effect for 2 μg/ml Dox treatment in JUN-induced PST system is extremely low.
To figure out the chromatin architecture dynamics during JUN-induced PST, we collected JUNTetON mESCs samples with Dox treatment for 0, 4, 8 and 12 h (Fig. 1A) for ATAC sequencing (ATAC-seq). The calling peaks, as shown previously [21], was divided into three basic categories: PO, permanently open during PST, which was further divided into PO-up (POU),PO-down (POD) and PO-no change(PON) subgroups according to the trends of chromatin accessibility dynamics; OC, open at 0 h but closed during PST, which was further divided into OC1(0–4 h), OC2(4–8 h) and OC3(8–12 h) subgroups according to different time window; CO, closed at 0 h but opened during PST, which was further divided CO1(0–4 h), CO2(4–8 h) and CO3(8–12 h) subgroups according to different time window (Fig. 1B). Based on the above catalogs, we show by heatmap (Fig. 1B) and line chart (Fig. 1C) that the chromatin state change dramatically by JUN induction within 12 h, e.g., 18,801 OC peaks, 45,416 CO peaks, 16,751 PO-up peaks and 21,976 PO-down peaks. Strikingly, the curve of CO1 peaks shows the largest increments (34,661) among all the subgroups (Fig. 1C). As 6 h exposure to Dox treatment leads to an irreversible pluripotency exit during JUN induced PST [15], we further investigate the chromatin accessibility dynamics (CAD) features for those subgroups in detail. Motif discovery for each subgroup indicates a gradual chromatin closing for loci occupied by pluripotent factors, e.g., ESRRB, TCFCP2L1, NR5A2, TCF3/4, KLF4/5, POU5F1, SOX2, OCT4-SOX2-TCF, while prominent chromatin opening for loci occupied by AP-1 factors, e.g., JUND, FOSL1/2, ATF1/2/3/4/7, BACH1/2, MAFA/MAFK (Additional file 2: Fig. S2A, B), demonstrating the closing of pluripotent chromatin and the opening of somatic chromatin in JUN-induced PST, consistent with the binary logic we proposed for cell fate transition in reprogramming [21]. We then show the top10 motifs enriched for each CAD subgroups. Significantly, motifs for AP-1 family factors are the dominant ones in all the CO1, CO2 and CO3 peaks, while motifs for SOXs/OCTs were obvious in all the OC1, OC2 and OC3 peaks (Fig. 1D and Additional file 2: Fig. S2B). Notably, motif for YY1 is present in OC1 and OC3, but not in OC2, indicating the difference for chromatin closing among those three stages (Additional file 2: Fig. S2B).
By combining RNA-seq and ATAC-seq data, we further investigate the correlation between gene expression pattern (GEP) and chromatin accessibility pattern (CAP) during PST. To achieve this, the differentially expressed genes are divided into three groups: Group 1, gradually upregulated,1719; Group 2, immediately upregulated,1250; Group 3, down regulated,1323 (Fig. 1E and Additional file 2: Fig. S2C). We extracted all the genes located in each CAP subgroups and made a mapping for the genes shared between each CAP subgroup and GEP subgroup (Fig. 1F, G). The mapping ratios are calculated to assess the correlation between GEP and CAP. Through simple statistical analysis, we show that, PON relative genes have similar ratios to GEP pattern; POU/CO1/CO2 relative genes have the most significant correlation with gene upregulation; POD/OC3/OC2 relative genes have the most significant correlation with gene down-regulation (Fig. 1F). In addition, genes in Group1/2 have higher mapping ratio to POU/CO1/2/3; Genes in Group3 have higher mapping ratio to POD/OC1/2/3 (Fig. 1G).
We further investigate the relationship between CAP and GEP in temporal dimension by analyzing specific genes in the time course RNA-seq and ATAC-seq data. We find a significant delay for the emergence of RNA peaks (8 h) to ATAC peaks (4 h) in the loci of somatic genes such as Anxa1, Fosl1 (Fig. 1H). These findings suggest a model in which chromatin open facilitating gene activation at the early stage of PST. In addition, the behaviors of ATAC peaks and RNA peaks in the loci of pluripotent relative genes such as Esrrb, Nanog, and Nr5a2 are quite synchronous, both disappearing at 12 h (Fig. 1H). These data indicate that JUN mediated chromatin opening initiates the transition of PST.
The relocation of SS18/BAFs during JUN-triggered PST.
Recently, using CRISPR/CAS9 genome-wide screening technology, we identified SS18/BAF complexes as critical epigenetic coagent for JUN-triggered PST [15]. To further explore the underline mechanisms, we performed ChIP-seq to detect the occupancy of JUN, SS18/BAFs in chromatin as well as the modification of H3K27ac during PST at 0 h and 8 h, respectively. The resulting omics data were combined to ATAC-seq data for further analysis (Fig. 2A). Generally, heatmaps for these omics data reveal a genome-wide consistency in pattern among chromatin accessibility state, JUN binding and H3K27ac modification (Fig. 2A, left three panel), consistent with the function of JUN protein in chromatin opening. Attractively, ChIP-seq signaling for SS18 and BRG1 show a significant switch from OC/POD peaks at 0 h to CO/POU peaks at 8 h (Fig. 2A, right two panel), indicating a dramatic relocation for SS18/BAFs during PST. Particularly, in loci relative to pluripotent genes, such as Esrrb, Nanog and Nr5a2, the loss of SS18/BRG1 binding is accompanied with chromatin closing, loss of H3K27ac as well as mRNA expression (Fig. 2B). In contrast, for loci related to somatic genes, such as Anxa1 and Fosl1, the emergency of SS18/BRG1 binding signaling is accompanied with chromatin opening, gain of H3K27ac, as well as mRNA expression (Fig. 2B). We then performed motif discovery for SS18/BRG1 binding sites at 0 h and 8 h, respectively. Top10 enriched motifs indicates an obvious translocation of SS18 from loci regulated by pluripotent factors, such as OCT4-SOX2-TCF-NANOG, ESRRB, NR5A2, SOX2 at 0 h to AP-1 family factors, such as FOSL2, JUN, FOSL1, ATF3, BATF, BACH1/2 at 8 h (Fig. 2C, D), the later one was largely overlapped to the motifs enriched in JUN binding loci at 8 h (Additional file 3: Fig. S3A). Quantitatively, SS18 undergoes a relocation from 3643 chromatin loci at 0 h to 9949 new chromatin loci at 8 h, remains only 1748 loci unchanged (Fig. 2E, left panel). Notably, JUN shares substantial co-occupancy with SS18 at 8 h (Fig. 2E, right panel), indicating a possible protein–protein interaction between JUN and SS18. We further investigate the chromatin accessibility dynamics (CAD) for loci occupied by SS18 (0 h, 8 h) and JUN(8 h), respectively, and showed by pie chart that, among loci occupied by SS18 at 0 h (5391), PON, 49.8%, indicating a constant open in chromatin state for those loci to 8 h; POD + OC,33.4% + 10.9%, indicating a majority close for those loci to 8 h (Fig. 2F, left panel). While for loci occupied by SS18 at 8 h (11,697), PON, 47.6%, POU + CO, 25.6% + 16.7%, represented a majority open for those loci to 8 h(Fig. 2F, middle panel). Those data indicate SS18 has a strong correlation with chromatin open. In addition, for loci occupied by JUN at 8 h (6154), PON (23.7%) + CO(34.3%) + POU(34.6%), 92.6% in sum (Fig. 2F, right panel), also suggests a significant correlation for JUN in chromatin open. Taken together, those data suggested a widely relocation for the binding of SS18/BAFs in chromatin at the early stage during JUN induced PST.
SS18/BAFs regulate PST through chromatin opening
Given the correlation among SS18/BAFs relocation, JUN binding pattern, chromatin accessibility dynamics, as well as the function of BAFs in chromatin remodeling, we hypothesize that SS18/BAFs may regulate PST by promoting the accessibility for JUN binding to its targets. To this end, we knocked down Brg1, the core ATPase of BAFs by shRNA and then performed ATAC-seq to detect the change of chromatin accessibility during PST. Heatmaps show a dramatic deficiency in CO/POU sites upon the absence of BRG1, while has little impact on the ones in OC/POD/PON groups at 8 h (Fig. 3A), suggesting the major function of BRG1 is chromatin opening. We further compared the chromatin accessibility state between shScr and shBrg1 at 8 h (Additional file 4: Fig. S4A). Brg1 knockdown leads 55,098 ATAC signaling down (ATD), 17,345 ATAC signaling up (ATU), as well as 60,086 ATAC signaling permanent (ATP) at 8 h. Motif discovery for ATD shows extremely significant for AP-1 family factors, such as FOSL1, JUNB, FOSL2 (Additional file 4: Fig. S4B), suggested the dominant control of these loci by AP1 family factors. Intriguingly, CTCF binding sites are dominant in ATP and ATU groups (Additional file 4: Fig. S4A, B). Meanwhile, we analysis the impact on transcriptome by Brg1 knockdown at 8 h during PST, and showed by heatmap that, 522 genes are failure to activation (Fig. 3B, up panel, and Additional file 4: Fig. S4C), 600 genes have no significant impact (Fig. 3B, middle panel, and Additional file 4: Fig. S4D), and 269 genes are failure to silence (Fig. 3B, down panel, and Additional file 4: Fig. S4E). We further examined the gene expression pattern for specific loci within different CAD groups upon Brg1 knockdown, and show that Srp14 and Hddc2, belongs to PON in CAD during PST has no change upon Brg1 knockdown in both chromatin accessibility state and gene expression (Fig. 3C, D, left panel); Tgfbi and Fosl1, JUN targeting genes, belongs to POU/CO1 in CAD during PST, present a failure in chromatin open, accompanied with recession in gene activation, upon Brg1 knockdown at 8 h (Fig. 3C, D, middle panel); Nanog, Klf4, Nodal, pluripotent genes, belongs to OC/POD group in CAD during PST, present a failure in chromatin close, accompanied with delay in gene repression (Fig. 3C, D, right panel).These data suggest that SS18/BAFs are critical for chromatin open in AP-1 binding loci at the early stage during JUN induced PST.
SS18/BAFs are not associated with JUN centric protein complexes
The massive and rapid relocation of SS18/BAFs from pluripotent loci to AP-1 binding loci raised a possibility that JUN centric protein complexes may mediate SS18/BAFs relocation by direct protein–protein interaction. To test this hypothesis, we performed IP-MS for JUN or SS18 to detect any possible interaction between JUN-centric protein complex to SS18/BAFs. As the probable protein–protein interaction would be more stable at later stage of PST, we collect JUNTetON ESC samples with DOX treatment for 0, 8 h and 24 h to detect the putative interaction (Fig. 4A). Venn plots showed the dynamics in composition for JUN centric or SS18-centric protein complexes by dox treatment with different time duration (Fig. 4B). Specifically, JUN centric protein complexes shared 34 members at 8 h and 24 h, while SS18 centric protein complexes shared 124 members at 0, 8 and 24 h (Fig. 4B), indicating a high stability for SS18/BAF complexes. We then focused on the interaction between SS18 and JUN. Scatterplots showed that neither JUN nor SS18 could pull down each other in IP-MS data (Fig. 4C). To find any possible protein–protein interaction(s) links JUN and SS18, we built SS18 centric and JUN centric network based on IP-MS data (Additional file 5: Fig. S5A, B). Here, we show that cBAF/ncBAF complexes, the super elongation complex (SEC) and nucleosome remodeling and deacetylase (NuRD) are shared in SS18 centric complex in 0, 8 and 24 h (Additional file 5: Fig. S5A), while AP-1 family members are shared in JUN centric network at 8 and 24 h (Additional file 5: Fig. S5A). Although no obvious interaction between AP-1 family members and subunits within BAF complexes was observed, three proteins, CEBPB, CEBPG and ZSAN4F are shared in both SS18 and JUN centric complexes at 8 h (Additional file 5: Fig. S5A, B), which may link JUN and SS18 during PST. To test this hypothesis, we knocked down those genes by shRNA (Additional file 5: Fig. S5C), and then tested the expression of JUN downstream genes, such as Anxa1, Runx1 or Pxdc1 by RT-PCR, no inhibitory effect was detected, as SS18 shRNA did during PST (Fig. 4D), suggesting function irrelevant for these three proteins to PST. Interestingly, when analysis time course IP-MS data for SS18 by heatmap (Additional file 5: Fig. S5D), we found NANOG and ESRRB, two key transcriptional factors, undergone a gradually separation from SS18/BAFs at 8 h and 24 h, respectively, suggesting a decoupling of SS18/BAFs for pluripotency maintain during PST (Additional file 5: Fig. S5D).
Together, these data suggest that the relocation of SS18/BAFs to target loci is unrelated to direct interaction between JUN centric complexes and SS18/BAFs.
SS18/BAFs recognize JUN induced H3K27ac signaling for relocation
Given no evidence support the relocation of SS18/BAFs mediated by direct interaction with JUN centric protein complexes, we turned to other mechanisms. BAF complex have been reported mainly co-localized with active histone modification such as H3K27ac, H3K4me1/3, etc. [22]. In addition, Polycomb complex dependent histone modification, such as H3K27me3, H2AK119ub have been shown incompatible with BAF complex [18]. To this end, we performed ChIP-seq for H3K4me1, H3K4me3, H3K9ac, H3K27me3 and H2AK119ub, together with the existing H3K27ac data to investigate the relationship between those histone modifications and SS18/BAFs during JUN induced PST. As shown by heatmap, the occupancy of SS18 in genomic have significant overlap with active histone modification, but barely no overlap with H3K27me3 and H2AK119ub (Additional file 6: Fig. S6A). Venn diagrams show the exact amount of co-occupancy sites between SS18 and those histone modifications at 0 h and 8 h, respectively (Additional file 6: Fig. S6B and S6C). Among all the active histone modification, H3K27ac have the largest overlap with SS18 in binding sites during JUN induced PST(Additional file 6: Fig. S6A–C), indicating a high correlation between H3K27ac and SS18, thus raised the hypothesis that the relocation of SS18/BAFs may regulated by recognizing of H3K27ac signaling during PST.
It was report that H3K27ac could be catalyzed by recruitment of CBP/P300 at JUN binding sites and recognized by proteins that contain bromodomain in BAF complex [23,24,25,26]. To this end, we test the above hypothesis by using two compounds, PFI-3, a BRG1 bromodomain (H3K27ac reader) inhibitor, and BI-9564, a BRD7/BRD9 bromodomain inhibitor, to block the BAF dependent H3K27ac recognition [25, 26]. Here, we show by RT-PCR that single or combined application of these two inhibitors, especially the later, could significantly block the activation of JUN target genes, such as Anxa1, Fosl2, Runx1 and Pxdc1 (Fig. 5A). We further performed RNA-seq to investigate the impact of the inhibitors on transcriptome at 8 h during PST. In detail,179 JUN upregulated genes were disturbed by the two inhibitors (Additional file 7: Fig. S7A and S7B). The disturbed genes are involved in the processes of cell-substrate adhesion, positive regulation of cell adhesion (Additional file 7: Fig. S7C), consistent with the morphologic change during JUN induced PST. We further performed ChIP-seq experiments to detect the location of SS18 upon the treatment of the two inhibitors at 8 h during PST, and show by heatmap that the binding of SS18 to new loci was reduced significantly (Fig. 5B). RT-PCR analysis further confirmed the repression of JUN target genes, such as Anxa1, Fosl2, Fosl1 and Runx1 by these two inhibitors (Fig. 5C). Furthermore, the JUN induced PST, measuring by cellular morphology and clonogenicity, was delay by the application of the two inhibitors (Fig. 5D, E). Taken together, these data suggest that H3K27ac links JUN to SS18/BAFs relocation in JUN induced PST.