Abstract

Introduction

Epigenetic alterations are prevalent in ALL and have been implicated as determinants of chemotherapy resistance and relapse. Epigenetics modulating drugs, such as DNA methyltransferase and HDAC inhibitors, are of interest as a potential means of enhancing chemotherapy sensitivity and improving clinical outcomes. Interpretation of clinical responses in trials of epigenetic drugs has been hampered by lack of precision in assessing the pharmacodynamic effects of the drugs. In the context of a clinical trial testing the epigenetic modulators decitabine (10-15 mg/m2) and vorinostat (180 mg/m2) (D/V) given for one week prior to reinduction with multiagent chemotherapy in young (age 1 - 21 years) patients with relapsed ALL, we assessed genome-wide cytosine methylation with single base-pair resolution using WGBS.

Methods

DNA was isolated from leukemia specimens (marrow or blood) obtained before and after 1 week of treatment with D/V from N=10 patients (20 paired samples). For N=6 of these patients (12 paired samples), RNA was also isolated. WGBS at 30X coverage and RNA-seq were performed on the DNA and RNA samples, respectively. We used a custom WGBS analysis pipeline to trim reads (TrimGalore), align to hg19 genome (Bismark/bowtie2), remove PCR duplicates and repeats, call B allele frequency at SNPs, and call methylation states at CpG sites. We used rsem, STAR and EBSeq for RNA-seq alignment and differential gene expression calls between pre- and post- treatment samples. 16 million CpGs were segmented into 1.3 million segments based on pre/post D/V methylation change (HaarSeg/R). We plotted 5 quintiles of gene expression (N=22,522 genes), from most expressed (Q1) to unexpressed (Q5), and linked to associated promoter CpG segment methylation (deepTools, Fig.1). Gene set enrichment analysis (GSEA/MSigDB) was performed on genes ranked in order by greatest promoter hypomethylation, highest significance of differential gene expression, and highest gene log2 fold overexpression post- treatment to create an "epigenetically activated" ranked gene list.

Results

WGBS showed case-specific areas of leukemia-associated allelic imbalance in each pair of samples, confirming that pre/post DV methylation differences are specific to leukemic blast population. D/V treatment reduced CpG segment methylation by 12.4% on average (range: 5.1-20.7%). Drug-induced hypomethylation was equally distributed (not specific to exon, intron, intergenic, promoter, or enhancer regions). In the 6 cases with paired RNA-seq data, we found a strong correlation between promoter methylation and gene expression in pre-treatment samples, and this correlation was maintained in the hypomethylated post-treatment samples. One of these cases, with global methylation decreased by 12.9%, is shown in Fig.1. GSEA on these 6 cases combined showed that 75/4169 gene sets (including targets of PRC2 and TP53, H3K27ME3- and H3K4ME3- bound, FDR<25%, p<0.01) were overrepresented in the "epigenetically activated" ranked gene list. In 4/6 cases we detected hypomethylated segments at the CD19 promoter and an associated 1.05- to 5.13- fold CD19 upregulation.

Conclusions

With high resolution, we observed base-pair specific reduction in CpG methylation after epigenetic modulation with decitabine and vorinostat across the genome in all cases, although the pharmacodynamic effect varied by as much as 15.6% between patients. The drug-induced activation of PRC2 targets suggests that epigenetic treatment of relapsed ALL may induce differentiation, and the activation of TP53 targets suggests that epigenetic treatment can reactivate tumor suppressor genes to suppress cell proliferation. Furthermore, epigenetic activation of CD19 expression suggests that epigenetic pre-treatment can potentially sensitize cells to CD19-targeted immunotherapies. Integrated WGBS/RNA-seq represents a powerful tool for pharmacodynamic evaluation of epigenetic therapy.

Figure 1. D/V treatment reduced average CpG segment methylation by 12.9% in case T0300. Methylation of gene promoters, defined as ±3kb from the transcription start site (TSS), showed strong correlation with expression of 22,522 genes ranked and split into quintiles in order of upregulated (Q1) to downregulated (Q4-Q5) (A, B). Hypomethylation was equally distributed across promoter regions and gene bodies (TSS-TES, scaled transcription start-end sites) (C, D).

Disclosures

Burke:AMGEN: Speakers Bureau; JAZZ: Speakers Bureau; Shire: Speakers Bureau.

Author notes

*

Asterisk with author names denotes non-ASH members.

This icon denotes a clinically relevant abstract