A Non-canonical BCOR-PRC1.1 Complex Represses Differentiation Programs in Human ESCs

Abstract

Polycomb group proteins regulate self-renewal and differentiation in many stem cell systems. When assembled into two canonical complexes, PRC1 and PRC2, they sequentially deposit H3K27me3 and H2AK119ub histone marks and establish repressive chromatin, referred to as Polycomb domains. Non-canonical PRC1 complexes retain RING1/RNF2 E3-ubiquitin ligases but have unique sets of accessory subunits. How these non-canonical complexes recognize and regulate their gene targets remains poorly understood. Here, we show that the BCL6 co-repressor (BCOR), a member of the PRC1.1 complex, is critical for maintaining primed pluripotency in human embryonic stem cells (ESCs). BCOR depletion leads to the erosion of Polycomb domains at key developmental loci and the initiation of differentiation along endoderm and mesoderm lineages. The C terminus of BCOR regulates the assembly and targeting of the PRC1.1 complex, while the N terminus contributes to BCOR-PRC1.1 repressor function. Our findings advance understanding of Polycomb targeting and repression in ESCs and could apply broadly across developmental systems.

Keywords: BCOR; human embryonic stem cells; pluripotency; polycomb repressive complexes.

Figure 1. BCOR is required for the maintenance of the pluripotent state in hESCs

(A) Western blotting of BCOR and the core pluripotency factors OCT4, NANOG and SOX2 during a nine-day RA differentiation time course. (B) KD efficiencies of the three independent shRNAs used to validate the differentiation phenotype in shBCOR-hESCs. shRNA 3 was used throughout this study. RT-qPCR* (top) and Western blotting (bottom) were performed on day 10 cultures. (C) Competitive proliferation assay in H1, H7, and H9 hESCs confirms the self-renewal defect induced by BCOR shRNAs. (D–F) Apoptosis, cell cycle and morphological analyses in day 10 shBCOR-hESCs. (G) Levels of the core pluripotency factors were assessed by RT-qPCR* (left) and western blotting (right) in day 10 shBCOR-hESCs. (H) Endoderm (FOXA2) and mesoderm (T) precursors were visualized by immunofluorescence in day 10 cultures. (I) Two-step targeting strategy used to derive BCOR rescue clones. The clone used in this study is shown at the bottom. The sgRNA sequence is in blue, the PAM sequence in red. BCOR is an X-linked gene, thus only one allele was targeted in H1 (XY) cells. (J) FLAG-BCOR expression in BCOR^R-hESCs after Dox removal. (K) Morphology of BCOR^R-hESCs 96 hours after Dox removal. (L) Expression of differentiation markers in RNA-Seq data from a six-day time course Dox removal experiment. *Data are represented as mean ±SEM of biological triplicates. See also Figure S1.

Figure 2. BCOR represses targets in association with the PRC1.1 complex

(A) BCOR-interacting proteins were immunoprecipitated from the inducible FLAG-BCOR hESC line and identified by mass spectrometry. FLAG-BCOR protein is indicated by an asterisk. (B) Mass spectrometry data was validated by IP-western blotting. SUZ12, CBX2 and PHC1 are shown as negative controls. (C) Nuclear extracts from BCOR^R-hESCs maintained with (+) and without (−) Dox for six days were fractionated on a sucrose gradient and analyzed by western blotting. Molecular weight markers (top) were fractionated in the same run. (D) ChIP-Seq data for BCOR, and other PRC1 and PRC2 components and marks associated with DNA accessibility and transcription were used as parameters for k-means clustering which separated all annotated TSS regions into three binding groups. Shown are boxplots (top) and heatmaps (bottom) of Log₂-transformed ChIP-Seq read counts within a 20 kb TSS-centered window. ChIP-Seq datasets for BCOR, KDM2B, RNF2, PCGF1, RYBP, H2AUb and K36me2 were generated in this study. CGI, SUZ12, CBX2, DNase I hypersensitivity and H3K27me3, H3K4me3 and H3K79me2 datasets were downloaded from the ENCODE (Consortium, 2012). POL2-Ser5p/7p datasets were from previous studies (Estaras et al., 2015). (E) Pairwise correlation coefficients between parameters shown in (D) were clustered to identify similarities in binding profiles within the strong and weak binding groups. (F) Boxplots of Log₂-transformed FPKM+1 values (RNA-seq) in shBCOR- and Control-hESCs. P-values are from a paired t-test. (G) Metaplots of indicated parameters for upregulated, downregulated and non-responsive genes. The binding profiles were normalized to the maximum value across all categories and binding group. P-values from a two-sided unpaired t-test are shown in Figure S2H. See also Table S1, Figures S2–4.

Figure 3. BCOR regulates the recruitment of the PRC1.1 complex, the maintenance of the H3K27me3 and the canonical PRC complexes at BCOR-responsive targets

(A) Enrichment levels of PRC1.1 components and histone marks were quantified by ChIP-qPCRin Control- and shBCOR-hESCs ten days after shRNA transduction*. Both activated and non-responsive BCOR targets were analyzed. Four non-bound (NB) regions are shown as controls. Enrichment is shown as ChIP to Input ratio normalized to the average enrichment at the non-bound regions. (B) Time-course analyses of PRC1.1 components and histone marks in BCOR^R-hESCs following Dox removal*. Enrichment is shown as ChIP to Input ratio normalized to the enrichment at day 0. The solid line represents enrichment at the bound region, the dotted line is the enrichment at the non-bound region. EOMES and T represent genes activated in shBCOR-hESCs, EN2 and NEUROD1 represent non-responsive gene group. (C) Enrichment levels of cPRC1 subunit CBX2 and ncPRC1 subunit RYBP were assessed by ChIP-qPCR in BCOR^R-hESCs maintained with and without Dox for six days*. (D) The working model for the recruitment/repression by the BCOR-PRC1.1 complex, integrating our results thus far with the published data on the PRC1 and PRC2 the complex. *One of the three independent experiments is shown. Data are represented as mean ± SEM of technical triplicates. See also Figures S3 and S4.

Figure 4. KDM2B and PRC2 are dispensable for the recruitment of the BCOR-PRC1.1 complex to a subset of critical differentiation genes and their repression in hESCs

(A) Western blotting with an antibody that recognizes an epitope located after the CXXC domain (aa726-817) confirmed the absence of both the long and the short KDM2B isoforms in KDM2B-KO hESCs. (B) KDM2B is depleted at BCOR targets in KDM2B-KO cells*. (C) Summary of BCOR ChIP-Seq analyses in WT and KDM2B-KO cells. A fold change (FC) of read counts within a 2 kb TSS-centered window between KDM2B-KO and WT cells was used to classify promoter regions into three groups. BCOR binding was considered retained if the FC was greater than 0.75, partially retained if the FC was between 0.25 and 0.75 and lost if the FC was less than 0.25. (D) Examples of BCOR-PRC1.1 targets that retained (EOMES), partially retained (NODAL) and lost (ZIC1) BCOR binding in KDM2B-KO hESCs. (E) Heatmaps of BCOR ChIP-Seq read counts for tagrets in the strong binding group in WT and KDM2B-KO hESCs. Promoters are sorted in increasing order of their FC values. (F) Gene expression summary in shBCOR- and KDM2B-KO hESCs. (G) Metaplots of ChIP-Seq parameters for genes with differential response to either KDM2B or BCOR depletion. Strong binding group is shown. Asterisks denote statistically significant (p<0.05) parameters. Full summary is shown in Figure S5H. (H) Western blotting shows greatly reduced H3K27me3 levels in WT and KDM2B-KO cells treated with 4 μM of EZH2 inhibitor GSK126 for seven days. (I) Cellular morphologies of WT and KDM2B-KO hESCs treated with GSK126 and KDM2B-KO cells transduced with BCOR shRNA. (J) Enrichment levels of PRC1.1 components and marks were quantified by ChIP-qPCR* at targets that retain BCOR upon KDM2B KO. WT and KDM2B-KO hESCs maintained with and without GSK126 for 7 days. (K) Expression of pluripotency markers (top) and BCOR targets (bottom) was analyzed in WT hESCs and three KDM2B-KO hESC lines maintained with and without GSK126 for 7 days. shBCOR-KDM2B-KO hESCs were also included. Analyses were carried out in biological triplicates. * One of three independent experiments is shown. Data are represented as mean ± SEM of technical triplicates. See also Figures S4 and S5.

Figure 5. C-terminus of BCOR targets the BCOR-PRC1.1 complex to chromatin

(A) BCOR mutants with deletions of indicated protein domains used in this study. (B) Co-IPs of BCOR-interacting proteins were performed using anti-FLAG affinity beads and analyzed by western blotting. Dox was added to the culture medium two days prior to experiments to induce the expression of mutant proteins. (C) The recruitment of BCOR deletion mutants to targets was quantified by ChIP-qPCR. Mutant hESC lines were transduced with Control and BCOR shRNAs and maintained in the presence of Dox for ten days. The enrichment is shown as a ChIP to Input ratio. Heatmap enrichments were normalized to the enrichment in BCOR-A cells transduced with Control shRNA. Enrichment at EOMES locus is shown next to the heatmaps. One of the three independent experiments is shown. Data are represented as mean ± SEM of technical triplicates. (D) The levels of RNF2 and CBX2 were examined in BCOR deletion mutants that exhibit loss of recruitment and in BCOR^CT-hESCs that do not have recruitment defect. The data is processed and visualized as in (C). (E) Cellular morphologies of WT and BCOR deletion mutant hESCs transduced with BCOR shRNA and maintained with or without Dox for ten days. (F) Expression of BCOR targets in BCOR mutant lines transduced with BCOR and Control shRNAs. Analyses were carried out in biological triplicates.

Figure 6. RNF2 is required for the repression of PRC1.1 targets and BCOR recruitment to chromatin

(A) RNF2 and H2AK119ub levels in RNF2^R-hESCs are comparable to those of WT hESCs and are rapidly depleted after Dox removal. (B) Cellular morphologies of RNF2^R-hESCs cultured with or without Dox for six days. (C) PRC1.1 targets from the strong binding group are upregulated upon RING1/RNF2 depletion in hESCs. The Venn diagram (left) and the heatmap (right) represent differential expression in RNA-Seq time courses in RNF2^R-hESCs and BCOR^R-hESCs, and RNA-Seq data KDM2B-KO hESCs. (D) Time-course analyses of PRC1.1 components and marks at EOMES and T loci were performed in RNF2^R-hESCs following Dox removal. The enrichment is shown as a ChIP to Input ratio normalized to the enrichment at day 0.. The solid line represents enrichment at the bound region, the dotted line is the enrichment at the non-bound regions. Data are represented as mean ± SEM of technical triplicates. (E) Heatmaps of BCOR and KDM2B ChIP-seq read counts for promoters in the strong binding group in RNF2^R-hESCs maintained with and without Dox for six days. Promoters are sorted in increasing order of their BCOR-FC values. Shown on the right is gene expression in RNF2^R-hESCs in the absence of Dox. (F) Genomic coverage data tracks for ChIP-seq and RNA-Seq experiments in RNF2^R-hESCs at BCOR-responsive (GATA4) and KDM2B-responsive (ZIC4-ZIC1) targets. (G) Morphologies of RNF2^R-hESCs reconstituted with WT RNF2 (RNF2^WT) and catalytically impaired RNF2 mutant (RNF2^I53A). Cells were grown with and without Dox for six days. (H) Recruitment of BCOR to targets was quantified by ChIP-qPCR in RNF2^R-hESCs reconstituted with the RNF2^WT and RNF2^I53A constructs. Anti-FLAG-RNF2 and H2AK119ub ChIPs were performed to quantify the binding of the RNF2 mutants and the level of ubiquitination. The enrichment is shown as ChIP to Input ratio normalized to the average enrichment at the non-bound regions. One of the three independent experiments is shown. Data are represented as mean ± SEM of technical triplicates. See also Figures S4 and S6.

Figure 7. HSPD1-interacting domain of BCOR is required for repression

(A) Morphologies of WT, BCOR^CT and BCOR^ΔHSPD1 hESC lines transduced with BCOR and non-targeting Control shRNAs. Following transduction, cells were maintained with or without Dox for ten days. The BCOR^ΔHSPD1 mutant shown here lacks the 227-729aa region. Similar data was obtained with the mutant carrying two smaller 227-327aa and 629-729aa deletions (data not shown). (B) Co-IP and western blotting confirmed the loss of interaction between BCOR^ΔHSPD1 and HSPD1 in BCOR^ΔHSPD1-hESCs. (C) Expression of BCOR targets in WT, BCOR-A, BCOR^CT, BCOR^ΔHSPD1 mutant hESC lines transduced with BCOR shRNAs. The data from biological triplicates is shown as fold change relative to -Dox WT hESCs transduced with Control shRNA. (D) Schematic diagram of the TetO-GAL4-LUC luciferase reporter. (E) Firefly luciferase activity in TetO-GAL4-LUC hESCs transduced with the indicated rTetR fusion proteins and co-transfected with GAL4-VP16 transactivator and Renilla luciferase vectors. Shown are the ratios of Firefly to Renilla luciferase signal normalized to the ratio in rTetR-transduced cells. Data are represented as mean ± SEM of biological quadruplicates. (F) The binding of rTetR fusion proteins and the deposition of H2AK119ub and H3K27me3 at the TetO array was confirmed by ChIP-qPCR. The enrichment is shown as a ChIP to Input ratio normalized to the maximum across all constructs. One of the three independent experiments is shown. Data are represented as mean ± SEM of technical triplicates. See also Figure S7.