Competition between PRC2.1 and 2.2 subcomplexes regulates PRC2 chromatin occupancy in human stem cells

Abstract

Polycomb repressive complex 2 (PRC2) silences expression of developmental transcription factors in pluripotent stem cells by methylating lysine 27 on histone H3. Two mutually exclusive subcomplexes, PRC2.1 and PRC2.2, are defined by the set of accessory proteins bound to the core PRC2 subunits. Here we introduce separation-of-function mutations into the SUZ12 subunit of PRC2 to drive it into a PRC2.1 or 2.2 subcomplex in human induced pluripotent stem cells (iPSCs). We find that PRC2.2 occupies polycomb target genes at low levels and that homeobox transcription factors are upregulated when this complex is exclusively present. In contrast with previous studies, we find that chromatin occupancy of PRC2 increases drastically when it is forced to form PRC2.1. Additionally, several cancer-associated mutations also coerce formation of PRC2.1. We suggest that PRC2 chromatin occupancy can be altered in the context of disease or development by tuning the ratio of PRC2.1 to PRC2.2.

Keywords: CRISPR; EZH2; H3K27me3; PRC2; SUZ12; cancer; epigenetics; homeobox; polycomb; stem cells.

Figure 1.. Small Amino Acid Changes in SUZ12 Cleanly Separate PRC2.1 and PRC2.2 Subcomplexes

(A) The SUZ12 locus was genome edited by transfecting iPSCs with a CRISPR-Cas9 plasmid targeting SUZ12 and a homology-directed repair plasmid (HDR). Puromycin selection gave polyclonal cells subsequently used for co-immunoprecipitation (IP), ChIP, and immunofluorescence (IF). (B) Genome-editing strategy to introduce SUZ12 cDNA containing wild-type or mutant sequences and a C-terminal 3xFlag tag into the SUZ12 locus. A Cas9-sgRNA complex targeted exon 2 of SUZ12, and the SUZ12 cDNA consisting of either WT, single-mutant (PEmut, A2J2mut1, A2J2mut2), or double-mutant (PEmut+A2J2mut1 or PEmut+A2J2mut2) was incorporated into the SUZ12 locus while retaining intron 1. Puro_R is the puromycin cassette located downstream of the edited gene. (C) A 995 bp PCR product amplified using genomic DNA of polyclonal cell lines indicated genome-editing. PCR was performed using primers F1 and R1 (panel B). (D) Western blots showing that SUZ12 expression and H3K27me3 levels were comparable between polyclonal cell lines containing either wild-type or mutant SUZ12. Flag(SUZ12) was only detectable in genome-edited cells. Cell lysates were loaded at two concentrations and histone H3 was used as a loading control. (E) Western blots detecting Flag(SUZ12), EZH2, JARID2, and MTF2 in input lysate and after αFlag IP. All mutations are within SUZ12 (see Methods). A2J2mut1 and A2J2mut2 disrupt binding to JARID2, while PEmut disrupts binding to PCL protein MTF2 (red dashed boxes). (F) Mass spectrometry was performed on lysate from genome-edited polyclonal cells after a two-step IP, first using an αFlag antibody and then using an αSUZ12 antibody after Flag peptide elution. Heatmap shows iBAQ mass-spec values normalized to SUZ12 (mean iBAQ values from N=2 biological replicates). Red dashed boxes indicate the set of accessory subunits that are disrupted by either PEmut or A2J2mut2. (G) Cartoon illustrating how mutations within SUZ12 give rise to either a PRC2.1 or a PRC2.2 complex. See also Figure S1 and Table S1.

Figure 2.. The Interaction with either PRC2.1 or PRC2.2 Accessory Subunits Is Sufficient to Maintain iPSC Pluripotency

(A) hiPSCs lose stem cell colony morphology when SUZ12 is deleted (SUZ12 KO) or the interaction with both sets of PRC2 accessory subunits is disrupted (PEmut+A2J2mut). Representative brightfield microscopy images on genome-edited iPSCs containing the indicated single or double mutations in SUZ12. (B) hiPSCs retain SOX2 expression in A2J2mut1, PEmut, or A2J2mut2 cells but lose SOX2 expression when the mutations are combined. DAPI stain for nuclei or IF using αFlag (for genome-edited SUZ12) and αSOX2 antibodies. (C) Quantification of large fields of view using CellProfiler. The y-axis indicates the percentage of cells that contain both Flag and SOX2 signal over the total number of Flag-positive cells. Individual data points are shown as well as mean ± SD from N ≥ 3 unbiased fields of view and > 10,000 cells per image. p-value < 0.001 is indicated by *** as determined by an unpaired two-tailed t-test. See also Figure S2.

Figure 3.. PRC2 Accessory Proteins Determine its Chromatin Occupancy

(A, B, C) Flag(SUZ12) ChIP-seq using an αFlag antibody. Each immunoprecipitation was performed with 75 μg of chromatin and sequenced reads were normalized to counts per million (CPM). (A) ChIP-seq signal over the PRDM12 locus after αFlag IP on lysate from genome-edited polyclonal cell lines. Square brackets, scale for genome-browser traces [CPM]; note scale changes. (B) The mean αFlag ChIP-seq signal centered on a consensus αFlag peak set containing 3008 peaks. Each polyclonal cell line is indicated by a different color and the unique replicates are either a solid or dashed line. (C) Heatmaps of input DNA and individual αFlag ChIP-seq replicates from each polyclonal cell line centered on the consensus αFlag peak set. (D, E, F) Equivalent to panels A-C, except SUZ12 ChIP-seq used an antibody to the protein itself. In panel E, the ChIP-seq signal was centered on a consensus αSUZ12 peak set containing 2876 peaks. (G, H, I) Equivalent to panels A-C, except αH3K27me3 immunoprecipitation on 50 μg chromatin from genome-edited polyclonal cell lines. In panel H, the mean αH3K27me3 ChIP-seq signal from wild-type or A2J2mut2 polyclonal cells was centered on a consensus αH3K27me3 peak set containing 5893 peaks. [In all cases, reads were normalized to CPM and placed into genomic bins of 50 bp for genome-browser traces, metapeak analysis and heatmaps. For heatmaps, each row represents the same peak across all experiments ordered according to mean coverage intensity.] See also Figure S3.

Figure 4.. Altered PRC2 Chromatin Occupancy Occurs Specifically on Polycomb Target Genes

(A) αFlag ChIP-seq data from A2J2mut2 and WT polyclonal cells plotted as a MA plot. Red, genes with a significant fold change relative to WT, p-value ≤ 0.005. (B) αFlag ChIP-seq data from PEmut and WT polyclonal cells plotted as a MA plot. Blue, genes with a significant fold change relative to WT, p-value ≤ 0.005. (C) Scatterplot from the αFlag ChIP-seq data in panels A and B, plotting fold change of each mutant compared to WT. Purple, genes with significantly changed occupancy in both datasets. Dashed line, best fit linear regression (slope = −0.11) indicating that genes which gain occupancy in A2J2mut2 cells lose occupancy in PEmut cells. (D) Venn diagram showing the number of genes that significantly gain or lose Flag(SUZ12) occupancy in the A2J2mut2 and PEmut datasets, respectively. Overlapping genes are indicated in the intersection of the two ellipses. (E) DAVID functional annotation of the 234 genes shared between the two datasets. [In all cases, fold change and base mean expression were determined by counting reads over all 60,609 annotated genes in gencode hg38 from N=2 biological replicates (experiments performed on separate populations of cells). DESeq2 was used to calculate fold change, mean occupancy, and p-values.] See also Figure S4.

Figure 5.. PRC2.1 Is Required for Homeobox Gene Silencing

(A) Clonal WT and PEmut cell lines were selected from the polyclonal population and PCR screened. The polyclonal cells were used for ChIP-seq, as previously described, while the clonal cells were used for RNA-seq. (B) RNA-seq performed on two WT and two PEmut clones comparing the expression of the polycomb-regulated genes IRX5 and BARX1. Reads were normalized to RPKM. (C) MA (ratio intensity) plot of RNA-seq comparing relative gene expression between PEmut and WT SUZ12 clones. The expression of blue genes changed significantly between PEmut and WT with p-value ≤ 10⁻⁸. The inset bar graph shows the number of significantly down and upregulated genes within the p-value cutoff. (D) Volcano plot showing the log₂(fold change) and -log₁₀(p-value) of RNA-seq performed on PEmut and WT SUZ12 clones. Blue genes are considered significant with a p-value ≤ 10⁻⁸ and a log₂(fold change) ≤ −1 or ≥ 1. Grey dashed lines indicate the p-value and fold change cutoffs. (E) DAVID functional annotation performed on the set of significantly up- or down-regulated genes using the high stringency setting. (F) Genome browser traces of αFlag ChIP-seq replicates from polyclonal cells over IRX5 and BARX1. [DEseq2 was used to calculate p-values, fold change, and base mean expression from biological duplicates of each of two clones (N=4 for both WT and PEmut). 3 upregulated genes (EBF2, TBX15, and MDGA2) are not shown because they were off the y-axis scale with p-values < 10⁻⁵⁵.] See also Figure S5.

Figure 6.. PRC2.1 has a Higher Affinity for DNA than PRC2.2, and PRC2.1-containing hiPSCs Gain Flag(SUZ12) Foci

(A) EMSA showing the binding of purified PRC2 subcomplexes to dsDNA probes of various lengths containing a portion of the LHX6 promoter sequence. (B) K_d^apparent values and their associated 95% confidence intervals (in brackets) of PRC2 subcomplexes for dsDNA probes containing a portion of the LHX6 promoter or an AT-rich sequence (N=3 independent experiments performed on separate days. Two different protein purifications of each complex were included.) (C) The number and intensity of Flag(SUZ12) foci increase in A2J2mut2 cells and decrease in PEmut cells. DAPI stain for nuclei or IF using αFlag (for genome-edited SUZ12) and αH3K27me3 antibodies. White arrowheads in WT cells indicate Flag(SUZ12) foci that overlap with H3K27me3 foci. Images were taken under identical conditions and scaled equivalently. (D) Quantification of the number of Flag(SUZ12) foci per cell. (E) Quantification of the intensities of individual Flag(SUZ12) foci. (D,E) The median is shown as a thick dashed line and the upper and lower bounds separate individual quartiles. p-values < 0.005 are indicated by ** and p-values < 0.001 are indicated by ***, determined by an unpaired two-tailed t-test. In (D), the number of cells in each condition was N_WT=239, N_PEmut=144, N_A2J2mut2=337. In (E), the number of Flag(SUZ12) foci was N_WT=2184, N_PEmut=842, N_A2J2mut2=8504. Cells without any detectable Flag bodies were not used in this analysis. See also Figure S6.

Figure 7.. A Single Amino Acid Mutation in SUZ12 Is Sufficient to Disrupt the Balance between PRC2.1 and 2.2

(A) Schematic showing generation of R103 mutant cells. New WT and A2J2mut2 polyclonal cells were genome edited alongside to ensure all cell types had equivalent passage numbers. (B) Western blots detecting Flag(SUZ12), JARID2, and MTF2 in input lysate and after αFlag IP. SUZ12 mutations (A2J2mut2, R103P, and R103Q) disrupt binding to JARID2 and enhance binding to MTF2. Cell lysates were loaded at two concentrations. Parent indicates IP on cell lysate from parental hiPSCs loaded at the higher concentration. (C) Quantification of the amount of JARID2 present in each lane. Western blot signals for JARID2 were normalized to the amount of Flag(SUZ12) for each loading concentration and plotted as fraction of signal relative to WT. (Data points represent the mean while error bars show the range from N=2 biological replicates. Each biological replicate was performed on a separate population of cells and carried out on a different day.) (D) Quantification of the amount of MTF2 present in each lane. Data quantification and representation described in panel (C). (E) αFlag ChIP-qPCR from polyclonal Flag(SUZ12) cells with the indicated SUZ12 mutations. Enrichment (y-axis) is percent pull-down relative to input normalized to WT for each biological replicate (raw data shown in Figure S7). (Mean ± SD from N=3 biological replicates. Each biological replicate was performed on a separate population of cells and carried out on a different day. p-values < 0.05 are indicated by *.) (F) DAPI stain for nuclei or IF using αFlag (for genome-edited SUZ12) and αH3K27me3 antibodies. Images were taken under identical conditions and scaled equivalently. (G) Quantification of the number of Flag(SUZ12) foci per cell. (H) Quantification of the intensities of individual Flag(SUZ12) foci. (G,H) The median is shown as a thick dashed line and the upper and lower bounds separate individual quartiles. p-values < 0.001 are indicated by ***, determined by an unpaired two-tailed t-test. In (G), the number of cells in each condition was N_WT=334, N_A2J2mut2=508, N_R103Q=476, N_R103P=598. In (H), the number of Flag(SUZ12) foci was N_WT=6426, N_A2J2mut2=13,207, N_R103Q=11,692, N_R103P=17,137. Cells without any detectable Flag bodies were not used in this analysis. See also Figure S7.