br and most promoters Fig e Fig
and most promoters (Fig. 6e, Fig. S9c). These trends were consistent with H3K9me3 binding patterns reported previously [22,23]. Further, migrated populations showed greater loss of H3K9me3 over open chromatin regions. Analysis of H3K9me3 distribution over gene bodies showed a depletion of H3K9me3 over transcription start sites (Fig. S9a, b), and an increase in level along the gene body, which has been pre-viously reported  and suggested to prevent aberrant transcription initiation within the gene body. Next, to understand links between changes in chromatin and the transcriptional state in migrated popu-lations, results of the H3K9me3 ChIP-seq were compared against RNA-seq gene expression data. No direct correlation was observed between H3K9me3 occupancy and differential gene expression in migrated and non migrated populations (Fig. S9d). However, a correlation was noted
between changes in this methylation and the absolute level of mRNA expression. Specifically, while promoters of low expression genes (0–25 percentile expression) did not show significant changes in H3K9me3, promoters of high expression genes (75–100 percentile expression) showed low H3K9me3, and this mark is further depleted in migrated populations broadly across these promoters (p < 10-4) (Fig. 6f, Fig. S9e). Thus, migrated populations of cervical cancer cells show enlarged nuclei and a broad loss of H3K9me3, especially over promoter elements of highly expressed genes.
We observed that a SUV39H1 knockdown enhances migration in
Fig. 5. Migrated populations of cervical cancer cells show SUV39H1-linked migratory transcriptional signatures. a. GO Biological Process terms corresponding to genes upregulated and downregulated in migrated populations (across replicates, n = 3). b. Plots depicting mRNA abundance (y axis depicts normalised read count) of representative differentially expressed genes SNAIL2 and CASP10 in migrated and non migrated populations, in a single replicate (n = 1 of 3). c, d. Expression of core EMT transcription factors and downstream markers/effectors across replicates (n = 3), FPKM values depicted as a heat map. e. Putative transcription factors active in migrated populations, inferred using overlap between genes upregulated in migrated populations and known UCSC ENCODE transcription factor Pimozide (TFBS). f. Overlap of migrated population upregulated genes with set of genes upregulated in TCGA SUV39H1-low tumours.
Fig. 6. Migrated populations of cervical cancer cells show a global reduction of H3K9me3, the histone mark associated with SUV39H1. a. H3K9me3 abundance in migrated populations of SiHa cells using western blotting (n = 3). b-d. H3K9me3 abundance (left panels) and nuclear size (right panels) assessed by immuno-fluorescence. b. Representative confocal images. Scale bar, 35 µm, c. Quantitation of H3K9me3 staining intensity and nuclear size in individual cells range from a single representative experiment (n = 1 of 4 and 3, respectively). Bars depict median and interquartile range. d. Comparison of median H3K9me3 fluorescence intensity and nuclear area in migrated and non migrated populations across replicates (n = 4 and 3, respectively). e. Enrichment of H3K9me3 ChIP reads against input across repetitive element (left) and gene element (right) classes (n = 1 of 2 replicates). f. Distribution of H3K9me3 ChIP reads over promoters of low expressed (0–25 percentile) and high expressed (75–100 percentile) genes (n = 1 of 2 replicates), ***, p < 0.0001, Wilcoxon signed rank test.
cervical cancer cells (Fig. 1d, Fig. S1e), SUV39H1-low cervical tumours and SUV39H1low cell populations within tumours show features of migration (Fig. 2e, Fig. 3), and that migrated populations show SUV39H1-linked transcriptional changes and a broad loss of H3K9me3 (Fig. 5a–d, f, Fig. 6a, b, f). Our observations indicate an anti-migratory role for SUV39H1 in cervical cancers, in contrast with pro-migratory roles reported in colon, breast and prostate cancer settings [25–27]. Reinforcing this complexity, across a panel of 21 cancers, SUV39H1 expression showed a negative correlation with survival only in cervical cancer (Table S1, Fig. 2a–c, Fig. S2b, c). SUV39H1 has been suggested in oncogenic as well as tumour suppressor roles across cancers, de-pending on the context [25,26,28–30]. This is likely because SUV39H1, like most chromatin regulators, does not possess sequence specificity - the genomic loci to which it is recruited to depends on the transcription