IntroductionThe centromere is a specialised region of the chromosome that is responsible for the attachment of spindle fibres via the kinetochore during mitosis, ensuring that each daughter cell receives an equal and identical set of chromosomes. Kinetochore assembly is a highly regulated process that builds upon the constitutive centromere associated network (CCAN). The CCAN bridges the CENP-A-containing centromeric chromatin with the outer kinetochore components. CENP-B contributes to this assembly by bridging CENP-A and the CCAN subunit CENP-C. Human centromeric DNA is made up of repetitive arrays of alpha satellite (α-Sat) repeats forming higher order repeats (HORs)1. Many α-Sat repeats contain a motif termed the B-box, which is bound by CENP-B. CENP-B is important for creating three dimensional structures in the centromere that promote appropriate chromosome segregation2.Centromeres are flanked by pericentromeres or transition regions, which are made up of a wider range of repetitive sequences in human cells, including α-Sat and non-α-Sat repeats, transposable elements, and duplications1. H3K9me3 and other repressive marks are enriched in pericentromeres and the repetitive elements are mostly silenced3. However, pericentromeric regions also contain non-repetitive elements, including protein coding genes, some of which are expressed1. Therefore, while largely heterochromatic, there are interspersed regions of open chromatin in pericentromeric regions and transition arms.The repetitive nature of centromeric and pericentromeric sequences facilitates topological organisation, but this comes at a cost, since these repetitive sequences are vulnerable to inappropriate rearrangements4,5. Moreover, recent work demonstrated that cells use centromere-associated DNA breaks to help specify functional centromeres6, which increases vulnerability if these are processed or repaired inappropriately. Pathological centromere fragility contributes to human disease, such as cancer4. Cells must therefore balance the use of these specialised and repetitive structures with mechanisms that protect genome integrity. How this is achieved is not yet fully understood.PBRM1 (or BAF180) is a subunit of the PBAF chromatin remodelling complex, one of three mammalian SWI/SNF remodelling complexes. PBRM1 is frequently mutated in cancer, and evidence supports the idea that PBRM1 can function as a tumour suppressor7. Loss of function mutations are particularly prevalent in clear cell renal cell carcinoma (ccRCC), but they are also found across a range of other cancer types8. A critical question is what the fundamental functions of PBRM1 are, which, when lost, contribute to the development or evolution of cancer.Here, we identify PBRM1 as a factor that prevents centromere fragility. We show that cells lacking PBRM1 have lower levels of centromere- and pericentromere-associated proteins and have altered patterns of organisation of these structures in cells. PBRM1 and the SMARCA4 subunit of PBAF are physically present at these regions, and the SMARCA4 binding pattern changes when PBRM1 is absent. Patterns of histone H3K9 methylation in centromeres and pericentromeres also change in cells lacking PBRM1. Furthermore, PBRM1 loss leads to mitotic defects and creates a dependence on the spindle assembly checkpoint, revealing a potential therapeutic vulnerability. Importantly, we find that even without external perturbations, PBRM1 loss leads to significant centromere fragility, highlighting a previously unrecognised role in centromere protection.ResultsAnalysis of isogenic PBRM1 knockout (KO) cell lines identifies misregulation of centromere- and pericentromere-associated proteinsTo identify core functions of PBRM1, we generated a panel of 17 clonal cell lines with CRISPR-Cas9 engineered loss of function mutations in PBRM1 across five different cell line backgrounds, including both cancer-derived and immortalised non-cancerous parental cell lines (Fig. 1a and ref. 9). The growth rate, cell cycle profile, and morphological changes in the knockout (KO) cells were analysed relative to the parental lines (Fig. 1b-f, and Supplementary Fig. 1). We found none of these features were substantially altered in any of the cell lines other than a modestly reduced growth rate when PBRM1 is lost (Fig. 1d and Supplementary Fig. 1).Fig. 1: Analysis of isogenic PBRM1 knockout (KO) cell lines identifies centromere associated protein misregulation as a common feature.a Workflow for generation of PBRM1 knockouts in a panel of cell lines. Number of independent clones validated for each cell line is indicated at the bottom. b–f Characterisation of PBRM1 knockouts using the hTERT-RPE1 cell line as an example. b Western blotting of whole cell lysates from parental and PBRM1 KO cells for PBRM1. α-tubulin is used as a loading control. c Scaled abundances of PBRM1 in proteomic analyses of whole cell protein extracts. Points correspond to independent biological replicates (n = 2). d Proliferation of RPE1 parental and two PBRM1 KO clones, measured using phase contrast Incucyte images (n = 2). e Cell cycle distribution of RPE1 parental and two PBRM1 KO clones measured using flow cytometry. n = 4, mean ± SEM, data were non-significant (ns) based on a 2way ANOVA using Dunnett’s multiple comparisons test. f Immunofluorescence images of nuclear morphology in RPE1 parental and PBRM1 KO clones. Scale bar corresponds to 40 µm. g–i Protein abundances in PBRM1 knockouts compared to parental cells in hTERT-RPE1 (g), 1BR3-hTERT (h), and U2OS (i) cell lines, detected using LC-MS of whole cell protein extracts. The mean log2 fold change (Log2FC) of protein abundance in PBRM1 knockouts versus parental cells is plotted against the -log p value, calculated using two-sided one-sample t-test. PBRM1 is highlighted in red, while centromere- & pericentromere-associated proteins are highlighted in purple. j Schematic outlining regions of the centromere and pericentric heterochromatin, including the kinetochore in mitosis. k Median Log2FC of annotated centromere- & pericentromere-associated proteins in RPE1 PBRM1 knockouts compared to parental cells. l Transcript levels of annotated centromere- & pericentromere-associated genes corresponding to the proteins in k detected using RNA-seq. Median Log2FC of annotated genes transcribing centromere- & pericentromere-associated proteins in RPE1 PBRM1 knockouts was plotted compared to parental cells. Points in k and l correspond to individual knockout clones from two independent biological replicates. Source data are provided as a source data file.Full size imageTo characterise the molecular profiles of the cells, we performed mass spectrometry on all PBRM1 KO clones and the corresponding parental lines. Pathway analysis showed alterations in chromatin organisation, DNA repair and recombination, and innate immune signalling were apparent across cell lines (Supplementary Fig. 2a and b). While apoptosis was identified as an altered pathway, we found no difference in the percentage of apoptotic cells when PBRM1 is deficient (Supplementary Fig. 2c). Previously, we found that PBRM1 is important for mediating sister chromatid cohesion at centromeres10, raising the possibility that PBRM1 is important for chromatin structure and organisation at or near centromeres. We therefore interrogated the proteomic datasets for centromere- and pericentromere-associated proteins, including CENP-A interacting proteins, the constitutive centromere-associated network (CCAN) complex, the outer kinetochore, the chromosomal passenger complex (CPC), pericentromeric heterochromatin (PHC) proteins, and other annotated pericentromere associated proteins. Individual protein levels were modestly altered in the PBRM1 KO cells, but looking across the pathways, a trend towards lower protein levels was apparent in the PBRM1 KO cells when compared with their isogenic parental cell lines (Fig. 1g-k, and Supplementary Fig. 2d, e and 3a). In contrast, transcript levels of these genes were not consistently downregulated in the PBRM1 KO cells, indicating that the protein level changes were not primarily being driven by misregulation of gene expression (Fig. 1l, and Supplementary Fig. 3b, c). This is consistent with previous work showing that there is a poor correlation between changes in the transcriptome and the proteome of SWI/SNF-deficient cells11, and suggests that SWI/SNF activity often influences protein stability through mechanisms other than direct regulation of gene expression. Moreover, downregulation was specific to centromere-associated complexes; by contrast, no consistent changes were apparent when centrosome proteins were analysed (Supplementary Fig. 3f, g).We further investigated available proteomic datasets in the cancer cell line encyclopedia (CCLE) to explore whether downregulated centromere proteins are a general feature of PBRM1 loss. In the absence of isogenic comparisons, we ranked cell lines according to PBRM1 protein levels. Notably, we found a significant correlation between PBRM1 protein levels and centromere- and pericentromere-associated proteins (Supplementary Fig. 3d, e). These data suggest that lower levels of centromere- and pericentromere-associated proteins is a common feature of cells lacking PBRM1 expression.Loss of PBRM1 results in altered organization of centromeres and pericentromeric heterochromatinWe next set out to understand whether the decreased level of centromere- and pericentromere-associated proteins had any functional consequence on their organisation. We first looked at whether there were any detectable changes in centromere structure using FISH probes against centromere α-Sat repeats in chromosome 2 or 10. We found a modest trend towards increased area in the KO clones compared with the parental RPE1 or 1BR3 cells (Fig. 2a–c and Supplementary Fig. 4). This pattern was previously observed in CENP-A-depleted cells12 raising the possibility that PBRM1 loss leads to CENP-A deficiency. Since CENP-A was not present in our proteomic dataset, we interrogated CENP-A by immunofluorescence (IF). However, when CENP-A signal area and shape (eccentricity) were measured, we found no clear difference in the PBRM1 KO cells (Supplementary Fig. 5a-c), suggesting that CENP-A deficiency is not driving the changes in α-Sat signal in the PBRM1 KO cells. Since α-Sat repeats are also present outside of CENP-A-containing chromatin, one possible explanation for the slight changes in FISH signals is a failure of these regions to form appropriate folded structures in the KO cells, leading to an increased volume of α-Sat repeat-containing chromatin in the KO cell nuclei.Fig. 2: PBRM1 KO cells display increased H3K9me3 intensity around centromeres.a, b Quantification of the area of individual foci in cells stained for α-satellite centromeric regions in (a) chromosome 2 and (b) chromosome 10, using FISH probes. n = 3, coloured points represent median of biological replicates, line = median of 3 replicates, data were normalised to median area of parental cells and analysed using two-sided t-test of experiment medians, *p = 0.0338 **p = 0.0061. c Representative images of α-satellite FISH of chromosomes 2 and 10 in RPE1 parental and PBRM1 knockouts. Scale bars correspond to 20 µm; or 5 µm in inset images. d Quantification of H3K9me3 signal per nucleus normalised to median intensity of parental nuclei. Grey points correspond to individual nuclei. n = 4, coloured points represent median of biological replicates, line = median of 4 replicates. At least 325 nuclei were analysed per condition and data were analysed using two-sided t-test of experiment medians, *p = 0.0360. e Representative images showing H3K9me3 and CENP-A signal in RPE1 parental and PBRM1 KO cells. Scale bars correspond to 20μm. f Schematic describing the method of quantifying H3K9me3 signal around centromeres, which were defined by the presence of CENP-A. Briefly, a Cell Profiler pipeline was used to identify nuclei and CENP-A foci within each nucleus. CENP-A signal was expanded to a total distance of 4 µm from the centre of each CENP-A focus, and divided into 8 rings. H3K9me3 intensity was measured in each ring. g Boxplot indicating H3K9me3 intensity at increasing distances from the centre of CENP-A foci in RPE1 parental and PBRM1 KOs. Boxes contain the 25th to 75th percentiles with a line at median, and whiskers extend to the largest value no further than 1.5 times the inter-quartile range, n = 3. CENP-A foci in at least 290 nuclei quantified per condition and data were analysed using two-sided t-test of experiment medians corrected for multiple comparisons using the Holm-Sidak method (threshold = 0.05), *p = 0.0485 and *p = 0.0272 for 0.5 µm and 2.5 µm distances, respectively. Source data are provided as a source data file.Full size imageWe used IF to interrogate CENP-B and NDC80 patterns to further understand the impact of PBRM1 loss on centromere-associated structures. While there was no difference in the CENP-B signal area, we found a modest, but reproducible decrease in eccentricity scores in the PBRM1 KO cells (Supplementary Fig. 5d–f), suggesting that the three-dimensional organisation of the region is impacted by the loss of PBRM1. NDC80 patterns were also subtly altered in PBRM1 knockout cells, with an increase in both area and eccentricity (Supplementary Fig. 5g–i). Interestingly, we also noticed an increase in the distance between mitotic sister chromatid-associated NDC80 signals in the PBRM1 KO cells (Supplementary Fig. 5j), consistent with a change in the three-dimensional organisation of pericentromeric regions.We next examined whether there were any microscopically detectable changes in pericentromeric heterochromatin. We performed immunofluorescence using an antibody against H3K9me3, a marker of heterochromatin, and found that the intensity of this signal is subtly but consistently increased in the absence of PBRM1 (Fig. 2d, e), suggesting altered organisation or prevalence of heterochromatin in these cells. To see whether there were changes specifically in pericentromeric heterochromatin, we used an antibody against CENP-A to identify centromeres and then quantified the surrounding H3K9me3 signal (Fig. 2f). Interestingly, we find that the pattern of H3K9me3 intensity changes in the PBRM1 KO clones (Fig. 2g). The signal was lower in the shell closest to CENP-A, but higher in outer shells in the PBRM1 KO cells, again suggesting that PBRM1 helps to organise chromatin in regions surrounding centromeres.PBRM1 and SMARCA4 are present at pericentromeres, and PBRM1 loss leads to changes in SMARCA4 associationOne possible mechanism by which PBRM1 is regulating the structure and organisation of chromatin and associated proteins in centromeres and pericentromeric regions is through direct remodelling activity in these regions. There is some evidence to support this possibility. PBAF was reported to associate with kinetochores of mitotic chromosomes13, and SWI/SNF subunits have been identified in protein-interaction studies of centromere-associated proteins, including CENP-C, INCENP, and Bub114,15,16. However, in contrast to our understanding of PBAF binding patterns in euchromatic regions of the genome, little is known about the specific binding pattern of PBAF in repetitive regions.We therefore set out to determine whether PBAF associates with centromeric or pericentromeric chromatin (including HORs, Hsat (Human satellite) repeats, transition (ct) arms; Fig. 3a) and gain a comprehensive view of PBAF localisation patterns. To do this, we performed CUT&RUN sequencing in both low and high salt conditions to ensure the capture of heterochromatic regions17. We mapped PBRM1 and SMARCA4 (BRG1), one of two catalytic subunits of PBAF, using both IgG and a SMARCA4 KO cell line (Fig. 3a and Supplementary Figs. 6 and 7b) as negative controls, and CENP-B as a positive control. Because of the repetitive nature of the region, we analysed the data in two ways: peak calling of uniquely mapping reads, and a k-mer analysis of reads mapping to these regions, thus allowing analysis of reads mapping to multiple locations (Fig. 3a and Supplementary Fig. 6, described below).Fig. 3: SMARCA4 is present at centromeres, and this pattern changes in PBRM1 KOs.a Simplified flowchart describing the mapping strategy to centromeric and pericentromeric sequences (detailed version in Supplementary Fig. 6). b Representative genome tracks displaying coverage of reads from CENP-B (purple), PBRM1 (blue), SMARCA4 (green) and IgG control (grey) CUT&RUN sequencing in RPE1 parental cells across the centromere. c Venn diagram indicating the overlap of significantly enriched SMARCA4 and PBRM1 peaks in centromeric and pericentromeric regions in RPE1 parental cells. d Stacked colour bar representing the genomic distribution of enriched PBRM1 and SMARCA4 peaks, categorised by feature, within centromeric and pericentromeric regions. e CUT&RUN and ChIP-seq signal heatmaps (PBRM1 – top, blue; SMARCA4 – bottom, green) in the indicated cell lines +/−5kb from the centre of RPE1 PBRM1 or SMARCA4 CUT&RUN peaks in the centromere and pericentromere, versus the IgG or input control (grey) of each experiment. Peaks are ordered by signal of the left-most heatmap, i.e. CUT&RUN peaks. An average signal plot is shown at the top of each heatmap. f Representative genome tracks displaying mapping locations of enriched k-mers across the centromere and pericentromere from analysis of CENP-B (purple), PBRM1 (blue), and SMARCA4 (green) CUT&RUN sequencing in RPE1 parental cells. g Venn diagram indicating the overlap of enriched k-mers (FC > 2) in centromeric and pericentromeric regions, in SMARCA4 and PBRM1 in RPE1 parental cells. h CENP-B motif detection following motif analysis of CENP-B-enriched k-mers compared to a shuffled control, where CENP-B k-mer sequences were shuffled maintaining 3-mer frequencies. i Percentage of enriched k-mers that map to specific regions in the centromere and pericentromere in each dataset, including SMARCA4 k-mers enriched in SMARCA4 KO cells (negative control).Full size imageWe first aligned uniquely mapping reads using the reported telomere-to-telomere (T2T) CHM13 genome1,18. CENP-B localised primarily to HORs, as expected, with some additional sites of enrichment in flanking regions (Fig. 3b and Supplementary Fig. 7a). We found enrichment of SMARCA4 and PBRM1 primarily in pericentromere sequences when compared with negative controls, and most SMARCA4 and PBRM1 peaks were located in the transition (ct) arms (Fig. 3b and Supplementary Fig. 7a). As expected, many of the PBRM1 and SMARCA4 peaks overlapped, but non-overlapping peaks were also identified (Fig. 3c, Supplementary Fig. 7c, e and f). This likely reflects the combinatorial flexibility of the complexes (i.e. PBAF can contain SMARCA2 instead of SMARCA4, and SMARCA4 is found in other SWI/SNF complexes).When analysing the peaks found in centromeres and pericentromeric regions, we find that the peak distribution profile of PBRM1 has a stronger association with promoters when compared with SMARCA4 (Fig. 3d). This profile is similar to that of the genome-wide distributions of both PBRM1 and SMARCA4 in our dataset (Supplementary Fig. 7d), and is consistent with previous studies showing that, in euchromatin, PBRM1-containing PBAF complexes are enriched at both promoters and enhancers, whereas BAF complexes are more often found at enhancers19,20. To determine whether pericentromere-association is a conserved feature of PBAF, we interrogated datasets in which PBRM1 and SMARCA4 were mapped21,22, and strikingly, we find that the association of PBRM1 and SMARCA4 in these regions is apparent in other cell line models (Fig. 3e).Because of the repetitive nature of these regions, we also analysed centromere- and pericentromere-associated reads that mapped to multiple locations, and were therefore unable to be precisely mapped, by performing a k-mer analysis1,23. This analysis identifies 51-mer sequences that are significantly enriched in the mapping datasets (see Fig. 3a and Supplementary Fig. 6 for workflow) relative to the IgG control. CENP-B was analysed relative to IgG as a positive control.In order to look for binding patterns and proximity to other features, enriched centromere- and pericentromere-associated k-mers were mapped back onto the T2T genome. The CENP-B-associated k-mers map primarily to the active higher order repeats (HORs) where CENP-B is known to bind (Fig. 3f, i, and Supplementary Fig. 7g, h), and motif analysis of the CENP-B associated k-mers identified the CENP-B box (Fig. 3h), providing support for the utility of this approach.Similar to peak enrichment, we found that the majority of PBRM1- and SMARCA4-associated k-mers are located within the transition arms, but a small proportion map to repeats (Fig. 3f, i Supplementary Fig. 7g, h), which may indicate some association across these regions. Again, we find both overlapping as well as PBRM1- and SMARCA4-specific k-mers (Fig. 3g and Supplementary Fig. 7i).We additionally mapped SMARCA4 in PBRM1 KO cells to interrogate changes to PBAF binding patterns when PBRM1 is absent. A subset of SMARCA4-enriched peaks and SMARCA4-enriched k-mers are lost when PBRM1 is absent (Fig. 4a and Supplementary Fig. 8a), suggesting that PBRM1 is important for targeting SMARCA4 and PBAF to these locations. Interestingly, a considerable number of SMARCA4-enriched peaks and SMARCA4-enriched k-mers are gained when PBRM1 is deficient (Fig. 4a, b and Supplementary Figs. 8 and 9a), suggesting that PBRM1 loss leads to aberrant SMARCA4 binding at sites not normally bound by SMARCA4. Together, these data indicate that PBAF binds to specific sites in chromatin flanking centromeres, and this binding pattern is disrupted when PBRM1 is deficient.Fig. 4: SWI/SNF enrichment at a subset of centromeric chromatin marks is altered in a PBRM1-dependent manner.a Venn diagram indicating the overlap of significantly enriched peaks (top, orange) or k-mers (bottom, red) in the centromere and pericentromere, of SMARCA4 and PBRM1 in RPE1 parental cells, and SMARCA4 in PBRM1 knockout (KO) cells. Venn diagrams show enriched peaks or k-mers found in at least two of three independent biological replicates, versus their IgG controls. The colour corresponds to the total number of enriched peaks or k-mers in each region of the Venn diagram (count). b Representative genome tracks displaying coverage of reads from PBRM1 (blue), SMARCA4 (green), and IgG control (grey) CUT&RUN sequencing in RPE1 parental cells and SMARCA4 in PBRM1 knockout cells (light green), showing an example of peaks gained (left) or lost (right) in the PBRM1 knockout cells. One representative independent biological replicate is shown, with boxes underneath representing peaks that were called as significantly enriched (q