Fig 2: ARID1A promotes STAG1 expression. a ARID1A ChIP-seq and input tracks of the STAG1 gene locus in RMG1 cells. b Validation of ARID1A binding to the STAG1 promoter by ChIP-qPCR in parental and ARID1A knockout RMG1 cells. c, d Validation of STAG1 downregulation by ARID1A knockout at the mRNA levels determined by qRT-PCR analysis (c) and at the protein levels determined by immunoblot (d) in RMG1 cells. Expression of other cohesin subunits STAG2 and SMC1 was used negative controls. e, f Expression of STAG1 at the mRNA levels determined by qRT-PCR analysis (e) and at the protein levels by immunoblot (f) in the indicated clear cell ovarian cancer cell lines or primary cultures highlighted in red. g Immunoblot of cohesin subunits Stag1 and Stag2 in wild-type controls, P53−/−/Pten−/− and P53−/−/Pten−/−/Arid1a−/− mouse bladder organoid cultures. h Correlation analysis between the expression of ARID1A and STAG1 in tumor microarray (TMA) of clear cell ovarian carcinomas determined by immunohistochemical staining. n = 3 independent experiments unless otherwise stated. Data represent mean ± s.e.m. P values were calculated using a two-tailed t test except for 4e by multilevel mixed-effects models, and for 4 h by Pearson correlation analysis. Relative intensities of immunoblot bands were quantified underneath

Fig 3: Ectopic STAG1 rescues the telomere damage and mitotic defects in ARID1A-inactivated cells. a Immunoblot validation of STAG1 knockdown in RMG1 cells. b, c Representative images (b) and quantification (c) of telomere DNA damage in RMG1 shRNA vector control and STAG1 knockdown cells determined by telomere FISH and γH2AX co-staining. d Quantification of distance between distal ends of sister chromatids enriched by colcemid treatment from the indicated RMG1 cells. e–g Schematics of STAG1 wild-type and mutant that lacks nuclear localization sequence (e), and validation of ectopic STAG1 expression by immunoblot (f) or immunofluorescence (g) in ARID1A knockout RMG1 cells. h–j Quantification of distance between distal ends of sister chromatids (h), and percentage of anaphase bridge (i) and lagging chromosome (j) positive-mitotic cells in the indicated parental, ARID1A knockout, and ARID1A knockout RMG1 cells rescued with wild-type or mutant STAG1, respectively. k, l Co-staining of telomere FISH and γH2AX (k) and quantification of telomeric DNA damage (l) in mitotic parental, ARID1A knockout, and ARID1A knockout RMG1 cells rescued with wild-type or mutant STAG1. m RMG1 cells expressing shSTAG1 or ARID1A knockout RMG1 cells rescued with wild-type or mutant STAG1 were subjected to time-lapse video microscopic analysis. Cell nuclei were visualized by staining for DNA using siR-DNA. Time is expressed as minutes: seconds. Arrows point to examples of lagging chromosomes. n = 3 independent experiments unless otherwise stated. Data represent mean ± s.e.m. Scale bar = 10 μm. P values were calculated using a two-tailed t test. Relative intensities of immunoblot bands were quantified underneath

Fig 4: ARID1A inactivation is selective against the survival of cells during mitosis. a, b Representative images (a) and quantification of colony formation efficiency (b) of colonies formed by single parental or ARID1A knockout RMG1 cells at the indicated G1 or G2/M phases of the cell cycle sorted by flow cytometry based on Hoechst 33342 staining. c, d Representative images (c) and quantification of colony formation efficiency (d) of colonies formed by single parental or ARID1A knockout RMG1 cells at the indicated synchronized G1 or G2/M phases of the cell cycle. e, f Representative images (e) and quantification of colony formation efficiency (f) of colonies formed by single ARID1A knockout RMG1 cells rescued with wild-type or mutant STAG1 at the indicated G1 or G2/M phases of the cell cycle sorted by flow cytometry based on Hoechst 33342 staining. g, h Expression of ARID1A and apoptosis markers cleaved caspase 3 or cleaved PAPR p85 in a panel of clear cell ovarian cancer cell lines (g) or endometrial cancer cell lines (h), respectively. i, j Images of orthotopic tumors formed by parental and ARID1A knockout RMG1 cells (i) and the sizes of the tumors formed were quantified (j). k, l Integrated density analysis of colonies formed by single cell G1 phase RMG1 parental and ARID1A knockout cells (k) or ARID1A knockout RMG1 cells rescued by wild-type or mutant STAG1 (l). m Compared with ARID1A wild-type tumors, ARID1A-mutated tumors exhibit a significant less copy number variations in the indicated cancer types in the TCGA datasets. n = 3 independent experiments unless otherwise stated. Data represent mean ± s.e.m. P values were calculated using a two-tailed t test except in 6 m by multilevel mixed-effects models. Relative intensities of immunoblot bands were quantified underneath

Fig 5: Cohesin enrichment at human pericentromeres but not a-satellite HOR arraysA. Linear binding profile of three cohesin subunits - RAD21, SA1, SA2 and the chromatin organizer CTCF across the length of chromosome 18, as determined by paired-end ChIP-seq in CHM13 cells and aligned to the T2T-CHM13-v1.0 reference. Centromere haplotype (CenHap) of chromosome 18 is denoted in red. Tracks with vertical gray bars denote peaks called by macs2. Data is average representation from 3 biological replicates. B. Overlap of binding sites between RAD21, SA1 and SA2 in the CHM13 genome. C. Zoomed in view of Centromere 18 CenHap with 10 Mb flanking regions (Cen Adjacent) on either side. CENP-A ChIP peaks are enriched within the D18Z1 a-satellite HOR array (red). Cohesin subunits lack binding sites within the asatellite HOR array but are enriched outside of the HOR array in the pericentromeric domains within the CenHap. D. Peak densities (number of peaks per Mb) of CENP-A, SA1, SA2 and RAD21 within active a-satellite HOR array (asHOR), centromere haplotype (CenHap), 2 Mb regions flanking CenHap (CenAdj) and genomic regions outside of the centromere haplotype (nonCen) of chromosome 18 in CHM13 cells. E. Metagenome analysis of peak densities (log2) of CENP-A, SA1, SA2 and RAD21 within the above stated regions, across all chromosomes. F. Proportion of mapped reads for RAD21, SA1 and SA2 in non-centromeric regions (nonCen), active a-satellite HOR arrays (asHOR) and centromere haplotype (CenHap) regions of the CHM13 genome. G. Peak densities (log2) of CTCF in the above stated regions of the CHM13 genome mirroring peak densities of cohesin factors.