DNA methylation subtypes guiding prognostic assessment and linking to responses the DNA methyltransferase inhibitor SGI-110 in urothelial carcinoma

Abstract

Background: At present, the extent and clinical relevance of epigenetic differences between upper tract urothelial carcinoma (UTUC) and urothelial carcinoma of the bladder (UCB) remain largely unknown. Here, we conducted a study to describe the global DNA methylation landscape of UTUC and UCB and to address the prognostic value of DNA methylation subtype and responses to the DNA methyltransferase inhibitor SGI-110 in urothelial carcinoma (UC).

Methods: Using whole-genome bisulfite sequencing (n = 49 samples), we analyzed epigenomic features and profiles of UTUC (n = 36) and UCB (n = 9). Next, we characterized potential links between DNA methylation, gene expression (n = 9 samples), and clinical outcomes. Then, we integrated an independent UTUC cohort (Fujii et al., n = 86) and UCB cohort (TCGA, n = 411) to validate the prognostic significance. Furthermore, we performed an integrative analysis of genome-wide DNA methylation and gene expression in two UC cell lines following transient DNA methyltransferase inhibitor SGI-110 treatment to identify potential epigenetic driver events that contribute to drug efficacy.

Results: We showed that UTUC and UCB have very similar DNA methylation profiles. Unsupervised DNA methylation classification identified two epi-clusters, Methy-High and Methy-Low, associated with distinct muscle-invasive statuses and patient outcomes. Methy-High samples were hypermethylated, immune-infiltrated, and enriched for exhausted T cells, with poor clinical outcome. SGI-110 inhibited the migration and invasion of Methy-High UC cell lines (UMUC-3 and T24) by upregulating multiple antitumor immune pathways.

Conclusions: DNA methylation subtypes pave the way for predicting patient prognosis in UC. Our results provide mechanistic rationale for evaluating SGI-110 in treating UC patients in the clinic.

Keywords: DNA methylation subtype; DNA methyltransferase inhibitor; Prognosis; Upper tract urothelial carcinoma; Urothelial carcinoma of the bladder; Whole-genome bisulfite sequencing.

Fig. 1

Flow diagram of the study. A training cohort of 45 UC samples and two validation cohorts were used for classification and prognosis analysis. RNA-seq data of 9 UTUC patients revealed the transcriptional features of the two subtypes. WGBS and RNA-seq were performed on SGI-110-treated cell lines

Fig. 2

WGBS analysis reveals that the DNA methylation profiles of UTUC and UCB are similar. A Global changes in average 5mC levels in different genomic elements determined by WGBS (the promoter is defined as ±1000 bp of the TSS) in UCB (n = 9), UTUC (n = 36), and paired adjacent urothelium specimens (n = 4). B Distribution of the average methylation levels of all genome-wide 10-kb bins in paired UTUC tumors and adjacent urothelium (n = 4). C Density plot of average DNA methylation within all genome-wide 10-kb bins in UCB (n = 9) and UTUC (n = 36) tumor samples. D Box plot of the correlation of the average methylation level within all genome-wide 10-kb bins within and between the two groups. E Graphical representation of the dynamic 5mC pattern of a region from chromosome 1. Two normal urothelium samples, UTUC tumor tissues and UCB tumor tissues were randomly selected

Fig. 3

Association between DNA methylation subtype and clinicopathological tumor and patient features. A Unsupervised clustering of the top 20,000 most variable DNA methylation haplotype blocks (MHBs) in UC showing two epi-clusters: Methy-C1 (n=19; 42.2%) and Methy-C2 (n=26; 57.8%). These clusters featured component 1 and component 2 MHBs, respectively. B Heatmap of differentially methylated MHBs between the Methy-C1 and Methy-C2 subtypes. Methy-C2 presented frequent hypermethylation compared to Methy-C1, and Methy-C1 was redesignated Methy-Low while Methy-C2 was redesignated Methy-High accordingly. C Kaplan–Meier survival curves showing that the DNA methylation subtypes can predict both overall survival and progression-free survival for UC patients. P-values were calculated by the log-rank test. n, number of cases. D, E The bar graph shows the association between the two subtypes of UC and clinicopathologic features. F Kaplan–Meier survival curves showed that the DNA methylation subtypes could predict both disease-specific survival and progression-free survival for the Japanese cohort. G Kaplan–Meier survival curves showed that the DNA methylation subtypes could predict overall survival for the TCGA muscle-invasive UCB cohort. H Forest plots displaying the results of multivariate Cox analysis of demographic variables predicting OS in TCGA UCB patients. CI, confidence interval OS, overall survival

Fig. 4

Methy-High UCs predominantly exhibit a basal expression pattern with high immune and stromal scores. A Heatmap of differentially expressed genes between Methy-High (n = 4) and Methy-Low (n = 5) UTUC patients. The top 30 differentially expressed genes are highlighted. B Differences in 50 hallmark pathway activities scored with GSVA software between Methy-High and Methy-Low UTUC patients. The t values calculated by a linear model are shown. C GSEA showing that EMT and hypoxia signaling are enriched in Methy-High UTUC patients compared to Methy-Low UTUC patients. D, E Immune and stromal gene expression scores in Methy-High and Methy-Low UC samples. P-values were calculated by two-tailed Student’s t test. F Expression profiles of the indicated gene pathways of biological relevance implicated in the TCGA UCB cohort in Methy-High and Methy-Low UC samples. EMT, epithelial-mesenchymal transition

Fig. 5

Methy-High UCs have a T cell inflamed and immunosuppressive environment. A The heatmap shows the anticancer immunity activity scores in seven steps across the cancer immunity cycle in UC patients with different methylation subtypes. The activity scores were calculated by TIP software. B, C Violin diagram analyzing the activity status of Step 4 and Step 6. Step 4: trafficking of immune cells to tumors; Step 6: recognition of cancer cells by T cells. P-values were calculated by the Wilcoxon rank test. D, E, F Heatmap displaying the population abundance of tissue-infiltrating immune and stromal cell populations in UC patients with different methylation subtypes. The proportion of infiltrating cells was evaluated by MCPcounter. P-values were calculated by the Wilcoxon rank test. G Expression profiles indicating differences in the anticancer immune response of the two methylation subtypes. APM, antigen presentation machinery, F-TBRS, fibroblast TGFβ response signature. H Immunosuppressive scores in Methy-High and Methy-Low UCB samples. P-values were calculated by two-tailed Student’s t test

Fig. 6

Guadecitabine (SGI-110), a DNA methyltransferase inhibitor, showed therapeutic effects in T24 and UMUC-3 cells. A, B Cell migration capacity with or without SGI-110 treatment was evaluated through wound healing assays. Values are the mean ± SD of eight independent experiments. C, D Cell invasion capacity with or without SGI-110 treatment was evaluated through Transwell assays. The bar chart on the right represents the number of cells passing through the compartment. Data are presented as the means ± SEM. n = 3. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001

Fig. 7

Integrative analysis identifies SGI-110 target genes in T24 and UMUC-3 cells. A Summary of GSEA immune-related gene sets upregulated by SGI-110 in T24 and UMUC-3 cells. The “immune” sector is broken down further into specific pathways characterized as part of the interferon response, cytokines/chemokines, antigen presentation, inflammatory, and cancer testis antigen (CTA) categories. B, C Volcano plot of gene expression data obtained by RNA-seq in SGI-110-treated T24 and UMUC-3 cells compared to untreated cells. SGI-110 upregulated cytokine/chemokine genes are highlighted. D, E Changes in the expression of dsRNA defense genes after treatment with SGI-110. F, G The proportions of the read counts from endogenous retroviral long terminal repeats (LTRs), long interspersed nuclear elements (LINEs), and short interspersed nuclear elements (SINEs). H Heatmap of cancer testis antigen (CTA) expression in SGI-110-treated T24 and UMUC-3 cells compared to untreated cells. Gene expression was calculated as transcripts per million (TPM) values