The presence of genomic alterations in chromatin modifiers is pervasive in cancer
We tested the prevalence and distribution of genomic alterations in 57 genes selected from literature [8, 9] coding for chromatin modifiers in 158 non-redundant studies encompassing 31254 tumor samples from the TCGA study and others, as well as 28 genes in 10945 tumor samples from the MSK-IMPACT study [10] (Additional File 1, Suppl. Figures 1 and 2) deposited in the cBioportal platform [11, 12]. Our preliminary analysis revealed that chromatin remodelers were the most frequently altered epigenetic category of regulators, followed by histone-modifying, and DNA modifying enzymes (Fig. 1; Additional File 1, Suppl. Figures 3 and 4), with mutations (missense and truncating) being the most common genomic alterations (Additional File 1, Suppl. Figure 3C). As a proof of principle, we have detected readily known aberrations such as enrichment of PBRM1 in renal cancer, and enrichment of ATRX with IDH1 in gliomas [13] (Fig. 1). Particular types of cancers (e.g lung cancer, bladder cancer, diffuse glioma, uterine cancer, skin cancer) had a higher incidence of genomically altered chromatin modifiers in contrast to others (e.g rhabdoid cancer, thyroid cancer, peripheral nervous system cancers, Wilms tumors) (Fig. 1; Additional File 1, Suppl. Figure 2).
Genomic alterations in chromatin modifiers are mutually exclusive in cancer
To get a better insight of the mutual relationships between genomically altered chromatin modifiers, we wanted to study their finer distribution in available cancer maps taken from patients and cancer cell lines (Fig. 2; Additional File 1, Suppl. Figures 5 and 6). The vast majority of altered chromatin modifiers followed a trend of mutual exclusivity in the TCGA datasets, as well as MSK-IMPACT and cancer cell lines as visualized by oncoprint analysis, and the lack of or very weak correlation (Fig. 2; Additional File 1, Suppl. Figures 7–9; Additional File 2). Contrary to the general trend of mutual exclusivity, we noted a couple of exceptions of co-occurrence that can be readily visualized by oncoprint, and were moderately correlated (Fig. 2; Additional File 1, Suppl. Figures 7–9; Additional File 2 and 4), for instance, HISTH1B-HIST1H3B-DAXX (Log2 Odds ratio > 3, P-value < 0.001); CHD3-PHF23 (CNAs, Spearman ρ = 0.77, Log2 Odds ratio > 3, P-value < 0.001), EZH2-KMT2C (CNAs, Spearman ρ = 0.68, Log2 Odds ratio > 3, P-value < 0.001); BRD4-DNMT1-SMARCA4 (Log2 Odds ratio > 3, P-value < 0.001); PBRM1-SETD2 (CNAs, Spearman ρ = 0.52, Log2 Odds ratio > 3, P-value < 0.001); ATRX-IDH1 (Mutations, Spearman ρ = 0.34, Log2 Odds ratio > 3, P-value < 0.001), as well as KMT2C/KMT2D with chromatin remodelers and histone acetyltransferases (Additional File 1, Suppl. Figures 7–11; Additional File 4). These co-occurring alterations in chromatin modifiers were even more visible in selected cancer types (Additional File 1, Suppl. Figures 12–16), such as deletion/amplifications of SMARCA4 and DNMT1 in lung cancer (Additional File 1, Suppl. Figure 12), amplifications of SMARCA4, DNMT1 and BRD4 in gliomas/glioblastomas and uterine cancers (Additional File 1, Suppl. Figures 14 and 16). In the same vein, we observed a pervasive overlap of alterations in histone-modifying genes such as KMT2C, KMT2D, KDM6A, CREBBP, EP300 with chromatin remodelers and other chromatin modifiers in bladder cancer (Additional File 1, Suppl. Figure 13), and deletions of HDAC2 and ARID2 in melanoma (Additional File 1, Suppl. Figure 15).
Genomic alterations in selected chromatin modifiers are associated with higher TMB and an improved response to checkpoint immunotherapy
Next, we sought to investigate the relationships between genomically altered chromatin modifiers and other molecular and clinical features, such as tumor mutational burden (TMB), and patient survival. Interestingly, cancers with high mutation rates (Additional File 1, Suppl. Figure 17A and B), such as endometrial cancer, colorectal cancer, skin cancer, glioma, lung cancer, and bladder cancer were the types of cancers with a high prevalence of genomically altered chromatin modifiers (Fig. 1; Additional File 1, Suppl. Figure 2). To delve deeper into this relationship, we analyzed the TMB levels per chromatin modifier in comparison to canonical TMB markers such as the presence of DNA mismatch repair (MMR) mutations (MLH1, MSH2, MSH6, PMS2) and POLE or POLD1 mutations. These results demonstrated higher TMB levels in cancers harboring alterations of certain chromatin modifiers (e.g. DNMT3B, SMARCD1 and other) in comparison to cancers harboring mutations in canonical oncogenes/tumor-suppressors, the MMR machinery or POLE/POLD1 (Fig. 3A; Additional File 1, Suppl. Figure 18A). Next, we wanted to ascertain if the co-occurrence of mutations in the MMR machinery or POLE/POLD1 together with alterations in chromatin modifiers would have an effect on TMB. Indeed, the co-occurrence of mutations in the MMR machinery or POLE/POLD1 together with alterations in selected (high TMB CMs, medium TMB CMs, and low TMB CMs) chromatin modifiers had a cumulative effect, and the highest level of TMB (Fig. 3B; P < 0.0001, Dunn’s test). To further refine this observation, we selected datasets harboring only MMR or POLE/POLD1 mutations respectively, mutations only in high TMB CMs or combination of both. These analyses validated our previous observation that the presence of mutations in both chromatin modifiers and the MMR or POLE/POLD1 machinery leads to a cumulative effect and the highest level of TMB (Fig. 3C and D, P < 0.0001, Dunn’s test). Since high TMB levels have been associated with better response to immunotherapy with checkpoint inhibitors [14, 15], to assess the clinical relevance of our observations, we reanalyzed the data of non-small lung cell carcinoma patients treated with checkpoint inhibitors from the MSKCC cohort [16]. Patients harboring mutations in chromatin modifiers, which are associated with high TMB (Additional File 1, Suppl. Figure 18A) had a better durable effect compared to controls of all MSK genes or oncogenes/tumor-suppressors (P value, 0.028, and 0.029, respectively, Fisher’s exact test) (Fig. 3E; Additional File 1, Suppl. Figure 18B).
The presence of genomic alterations in chromatin modifiers have an effect on survival
Finally, we wanted to evaluate the effect on the survival of patients harboring alterations in a chromatin modifier. We performed logistic regression analysis on TCGA data, and noticed that patients with mutations in the genes SMARCD1, DNMT3L, TET3, and KMT2A and amplifications in EP300, KMT2B, KDM5C, DNMT3L, HDAC4, SMARCA4, ARID2 and HDAC9 led to a worse survival prognosis, while mutations in NSD1, CHD6, ARID1A, IDH1, IDH2 and amplifications in HDAC2, IDH1 had a better survival prognosis (Additional File 1, Suppl. Figure 19; Additional File 5).