Whole Genome Sequencing Reveals the Extent of Structural Variants in Multiple Myeloma and Identifies Recurrent Mutational Hotspots within the Non-Coding Regions

Introduction. Genetic analysis of Multiple Myeloma (MM) has focused on the coding region, comprising 2% of the genome, leaving large regions uncharacterized where important driver genes and mutational mechanisms remain to be identified. Recent advances have allowed phasing of sequencing data permitting better reconstruction of the genome combined with an ability to characterize structural variants (SVs), the sites at which they occur, and the underlying mechanisms leading to their generation. This genome-wide approach aids reconstructution of the genome to define regulatory regions and characterizing the mechanisms that underlie the malignant transformation of normal plasma cells to a clone with the features of aggressive high-risk clinical disease states. We have used 10X Genomics Chromium Whole Genome Sequencing (WGS) to study the genome of a series of high risk MM with a focus on the non-coding regions, structural events and mutational mechanisms.

Methods. We analyzed 49 pairs of tumor and germline control myeloma samples using WGS. The samples consisted of high risk groups, t(4;14), t(14;16), and t(14;20), along with hyperdiploid samples. Gene expression data, risk status and clinical parameters were available for all samples. WGS was performed to a median depth of 45x (range 31-78x) with median molecule length of 27.9 kb and phase block N50 of 750.1 kb. Sequences were aligned to hg38 using Longranger (v2.1.0).

Somatic SVs were calculated by statistically comparing shared barcodes across identified breakpoints between tumor and germline samples. Somatic mutations were called using Strelka and Mutect, mutational signatures determined by NMF, and kataegis was determined by analyzing the distance between SNVs. SNVs were categorized according to genomic location and function. Hotspots of mutations surrounding promoters were determined by examining regions 2 kb upstream and 200 bp downstream of transcription start sites.

Results . Mutational spectrum - Mutational analysis showed there was a median of 35891 (12.1/Mb) total mutations, 19405 (16.6/Mb) mutations in genes (mostly intronic), and 16787 (9.3/Mb) intergenic mutations per sample.Across all samples, hotspots of mutation in non-coding regions were detected around the transcriptional start sites of BCL6 (n=30), CXCR4 (n=13), and MAX (n=9). The mutation hotspots were enriched in CpG islands, particularly in bases comprising the CpG dinucleotide, indicating they may affect gene expression through changes in methylation or altered transcription factor binding.

Mutational mechanisms - Samples with an APOBEC mutational signature have more exonic mutations and this trend was seen regardless of the functional class of mutation (nonsilent (P= 7.4x10^-6), UTR (P=7.4x10^-6), intergenic (P=2.6x10^-6) and silent (P=2.2x10^-6)). Localized sites of hypermutation, kataegis, were detected in 14/49 samples, were associated with a higher number of SVs (P<0.001), and were often seen in association with highly complex chromothripsis-like patterns, involving multiple chromosomes with up to 35 breakpoints. Kataegis was present in 33-50% of each of the cytogenetic subgroups, except for t(14;20) (n=10) where it was not detected.

Structural variants - Samples had a median of 22 (range 1-146) somatic SVs with a median of 6 (range 0 - 29) inter-chromosomal and 14 (range 1 - 142) intra-chromosomal rearrangements. After adjusting for chromosome size, chromosomes 8, 14, 16, 20 and 22 had significantly more inter-chromosomal rearrangements due to physiological and commonly known translocation events. Conversely, chromosomes 10, 15 and 18 had significantly fewer inter-chromosomal SVs. MYC rearrangements were present in 49% of this dataset and were often complex, with evidence of 'genome jumping' where up to 8 translocation events involving 6 chromosomes were detected.

Conclusion. Using WGS we have shown there is significant genetic variation in the non-coding region of the MM genome. We identified significant mutation in non-coding regions of key hematological genes, including BCL6 (hypermutated in DLBCL), CXCR4 (mutated in Waldenströms macroglobulinemia) and MAX (interacts with MYC), indicating that the mutational driver spectrum extends beyond the coding regions in MM. Complex SVs were seen, involving multiple chromosomes, resulting from kataegis and 'genome jumping', indicating that genome instability is a key feature of high risk MM.