Integrative Oncogenomic Analysis Combining Whole Genome, Transcriptome and Epigenome Identifies Altered Chromatin Accessibility Landscape in Multiple Myeloma

Multiple myeloma (MM) is a plasma cell malignancy characterized by clinical and genomic heterogeneity. Genomic investigations have identified few recurrently mutated protein-coding genes and various aberrations at the transcriptomic level. However, these changes do not accurately reflect the disease behavior. With accumulation of data in large patient population, it is becoming clear that analysis integrating various genomic correlates will be required to understand chromatin architecture that drives the disease behavior.

To gain further insight into MM biology, we integrated whole-genome sequencing, transcriptome sequencing, and epigenome profile using various assays (including the assay for transposase-accessible chromatin sequencing [ATAC-seq] to detect open chromatin regions, multiplexed, ChIP-seq and indexed T7 ChIP-seq [Mint-ChIP] to identify active enhancers and super-enhancers) to characterize the chromatin state of MM. We have analyzed 33 samples including primary MM cells from 18 newly diagnosed and 5 relapsed patients, 5 normal plasma cells from healthy donors and 5 Myeloma cell lines.

We first investigated the open chromatin landscape in myeloma cells, identifying over one million open chromatin regions (OCR) in our cohort, with an average of over 60,000 regions per sample. Despite a high heterogeneity between myeloma samples, unsupervised cluster analysis clearly separates primary myeloma samples, normal plasma cells and MM cell lines suggesting pathogenetic relevance. Although there is clear link between transcribed genes and OCR, we observe a significant number of OCR in the intergenic region with an as yet unclear biological significance. Furthermore, ATAC-seq data predicted several copy number abnormalities including the distinction of hyperdiploid from hypodiploid samples. We next characterized enhancers and super-enhancers in myeloma cells and their assigned genes using H3K27ac ChIP-Seq to identify important genes in MM biology. Despite heterogeneity between myeloma samples and myeloma cell lines, we could identify common SE-assigned genes as compared with normal plasma cells. These genes correspond to both known myeloma specific dependencies such as MYC, IRF4, CCND1, and TNSFR17 (BCMA) as well as new therapeutic targets that are currently under investigation.

Next, we used WGS data from 10 samples matching with ATAC and RNA sequencing to evaluate the mutational landscape of open chromatin regions and the consequences of the mutational processes targeting myeloma epigenome. Of 53,798 single nucleotide substitutions (SNV) (average of 5380 per sample), 2.7% of SNVs involve OCR (total of 1470 SNV within ATAC peaks, median - 104 mutations involving OCR per sample). We evaluated the presence of mutations within OCR targeting known binding motifs for transcription factors (TFs). Out of 1470 mutations affecting OCR, we identified 547 (37%) SNV targeting a known binding motif for transcription factors such as Sp1, NFkB1 or YY1. Using RNA seq data, we filtered the mutations affecting expressed TFs and assigned genes and observed that the presence of these mutations could significantly impact the assigned gene expression level. Indeed, these mutations generate either a de novo binding motif or alter the binding motif of a given TFs and are potentially responsible for TF hi-jacking to modulate the expression of number of genes important for MM biology. Finally, using non-negative matrix factorization (NMF), we evaluated the mutational signatures affecting the OCR and observed that a single predominant mutational signature affects open chromatin regions, as opposed to other processes such as APOBEC, which targets other regions of the genome.

In conclusion, global analysis of primary patient myeloma cells combining whole genome, transcriptome as well as epigenomic characterization reveals dysregulated chromatin landscape in myeloma that may be significantly influenced by specific mutational processes. Such integrated oncogenomic analysis provides an early view on perturbed genomic landscape with a view to understand the biology and specific targetable dependencies.