Synovial Sarcoma Chromatin Dynamics Reveal a Continuum in SS18:SSX Reprograming

Abstract

Synovial sarcoma (SyS) is an aggressive soft-tissue malignancy characterized by a pathognomonic chromosomal translocation leading to the formation of the SS18::SSX fusion oncoprotein. SS18::SSX associates with mammalian BAF complexes suggesting deregulation of chromatin architecture as the oncogenic driver in this tumour type. To examine the epigenomic state of SyS we performed comprehensive multi-omics analysis on 52 primary pre-treatment human SyS tumours. Our analysis revealed a continuum of epigenomic states across the cohort at fusion target genes independent of rare somatic genetic lesions. We identify cell-of-origin signatures defined by enhancer states and reveal unexpected relationships between H2AK119Ub1 and active marks. The number of bivalent promoters, dually marked by the repressive H3K27me3 and activating H3K4me3 marks, has strong prognostic value and outperforms tumor grade in predicting patient outcome. Finally, we identify SyS defining epigenomic features including H3K4me3 expansion associated with striking promoter DNA hypomethylation in which SyS displays the lowest mean methylation level of any sarcoma subtype. We explore these distinctive features as potential vulnerabilities in SyS and identify H3K4me3 inhibition as a promising therapeutic strategy.

Figure 1.. Whole genome landscape of Synovial sarcoma.

(a) Schematic sample overview. (b) Upper;. Barplot showing the number of amino acid altering mutations per sample; Lower; Barplot showing the number of non-coding SNVs and indels per sample. (c) Whole genome copy number frequency plot based on CNV calls from WGBS data showing the proportion of samples harboring either gains (red) or losses (blue) across autosomes. (d) Kaplan–Meier (KM) survival analysis of the fraction of genome altered (FGA) groups. (e) Dendrogram showing unsupervised hierarchical clustering of the top 15% most variably expressed genes (distance = 1 - spearman correlation, ward.D2 clustering method). Abbreviations: DoD, dead of disease; Met, metastasis; Bi, biphasic; Mo, monophasic; PD, poorly differentiated; A, axial; DE, distal extremity; PE, proximal extremity; BP, bootstrap probability.

Figure 2.. Distinct H3K27ac marker enhancer groups.

(a) Unsupervised hierarchal clustering of the top 1 % most variable 500bp bins of genome wide H3K27ac signal (distance = 1 - spearman correlation, ward.D2 clustering method). (b) Log10(signal intensity) of group 1 specific enhancer regions – log10(signal intensity) of group 2 specific enhancer regions. (c) Fraction of the group-specific peaks plotted against the distance to the transcription start site of genes (TSS) for the upper (Q1) and lower (Q4) quartiles of the distribution. (d) Fold change difference of transcription factor family motifs enriched at either the distal or proximal group specific enhancers. Gene Ontology enrichment analysis of differentially expressed genes between the (e) proximal and (f) distal group. Difference in expression between the distal and proximal groups of stem and epithelial markers including (g) LGR5, (h) LGR6 and (i) SOX2. ** indicates P-value < 0.01, *** indicates P-value < 0.001 for a Welch two-sample t test. Abbreviations: BP, bootstrap probability.

Figure 3.. H2AK119Ub1 and variable target gene activation.

Fraction of genome wide H2AK119Ub1 occupancy overlapping other histone marks in (a) a non-SyS samples, (b) SyS tumors, (c) the SYO1 cell line treated with either siRNA against SSX2 (siSSX2) or scramble control (siCtrl), (d) HIC⁺ driven mouse tumor (hSS2), adult HIC⁺ MSCs (Ctrl) or embryonic MSCs (E12.5). (e) Fraction of protein coding promoters that are marked by H2AK119Ub1 and are either SS18::SSX targets or not. Promoters are then further subdivided into H3K4me3 (H2AK119Ub1 and H3K4me3 without H3K27me3), bivalent (H2AK119Ub1 and H3K4me3 and H3K27me3) or H3K27me3 (H2AK119Ub1 and H3K27me3 without H3K4me3). (f) The number of protein coding promoters marked by H3K4me3 (H3K4me3 without H3K27me3), bivalent (H3K4me3 and H3K27me3) or H3K27me3 (H3K27me3 without H3K4me3). Promoters are then further subdivided into being marked by H2AK119Ub1 or not. (g) Gene expression values of signature synovial sarcoma genes ordered by variance. (h) ChIP-seq signal enrichment and gene expression tracks for five SyS primary tumors at the SOX2 and PAX3 locus. H2AK119Ub1 (dark blue), H3K27me3 (brown), H3K4me3 (red) and gene expression (grey) are displayed. (i) Visual representation of peak calls for H3K4me3 (red), H3K27me3 (brown) or bivalent (blue) regions around the SOX2, PAX3, ZIC2 and ZIC5 genes. The amplitude represents the number of samples that have a called peak in that region. ** indicates P-value < 0.01, *** indicates P-value < 0.001 for a Welch two-sample t test.

Figure 4.. Bivalency is a prognostic marker for synovial sarcoma.

(a) Binary heatmap over bivalently marked promoters (rows) in synovial sarcoma samples (columns) ordered by the total number of marked promoters. (b) Kaplan–Meier (KM) analysis of metastasis-free survival (MFS) in bivalency high (bivH) and bivalency low (bivL) groups. Statistically significant difference between the curves was calculated using the log-rank test. (c) Volcano plot displaying the number of differentially expressed genes between the upper and lower quartiles of the bivH and bivL groups. Gene Ontology enrichment analysis of genes found to be (d) upregulated or (e) downregulated in the bivH quartile. (f) Spearman correlation between the number of bivalently marked promoters and the number of promoters marked by H3K4me3 only. (g) Box plot showing the gene expression values for promoters being either bivalently marked or not. (h) KM analysis of MFS in the validation set, comparing the difference in survival between tumors predicted to be bivH or bivL. Statistically significant difference between the curves was calculated using the log-rank test. *** indicates P-value < 0.001 for a Welch two-sample t test. Abbreviations: DoD, dead of disease; Met, metastasis; Bi, biphasic; Mo, monophasic; PD, poorly differentiated; A, axial; DE, distal extremity; PE, proximal extremity.

Figure 5.. Synovial tumors harbor broad H3K4me3 domains

(a) Unsupervised hierarchal clustering (distance = 1 - spearman correlation, ward.D2 clustering method) of H3K4me3 in primary SyS (pink) and other normal and diseased tissue from the International Human Epigenome Consortium (IHEC). Heat is the proportion of promoter regions marked by H3K4me3 peaks. Box plots showing (b) the genomic occupancy of H3K4me3, (c) mean peak width of H3K4me3 at promoter regions and (d) fractional methylation of promoters in SyS and other normal and diseased tissue from IHEC. (e) Mean methylation (beta values) of promoter regions in SyS and other sarcoma subtypes. (f) Gene Ontology analysis of genes whose promoters are overlap differentially hypomethylated regions (DMRs) in SyS compared to other sarcoma subtypes. * indicates P-value < 0.05 using pairwise Wilcoxon signed-rank test.

Figure 6.. CGI hypomethylation associated with cell of origin and elevated H3K4me3 provides a potential therapeutic vulnerability.

(a) Fractional methylation of SyS and a collection of IHEC tissues in CGI shores. (b) Fractional methylation of genome wide, promoters, CGIs and CGI shores in the hSS2, Ctrl and E12.5 mouse samples. (c) Cell viability assay in SyS (HYSSII and SYO1) and Osteosarcoma (U2OS) cells upon 7-day treatment with the WDR5 inhibitor OICR-9429. (d) Cell competition assay performed in the SyS lines (HSSYII, SYO1, and Yamato) and Osteosarcoma (KHOS) cells transduced with an empty sgRNA as control or with guides targeting WDR5. (e) Proposed summative model. * indicates P-value < 0.05 using pairwise Wilcoxon signed-rank test.