Analysis of Mutational Signatures Suggest That Aid Has an Early and Driver Role in Multiple Myeloma

INTRODUCTION:

Multiple myeloma (MM) is a biologically and clinically heterogeneous disease. Different recurrent driver genomic events have been reported, but to date no unifying feature has been identified in MM evolution. The recent interest in signatures of mutational processes through analysis of whole-exome sequencing data has led to initial insights into what generates MM mutational repertoire (Bolli et al, Nat Com 2014). Here, taking advantage of the increased power provided by whole genome sequencing (WGS), we analyzed 22 paired samples from 11 patients first at the smoldering (SMM)/MGUS stage and subsequently at the time of progression to symptomatic MM to gain a deeper understanding of the full landscape of mutational processes operative in MM, especially during their evolution over time.

MATERIAL AND METHODS:

DNA from bone marrow CD138+ cells underwent WGS along with a matched normal sample using HiSeq X Ten machines (Illumina, Inc.). Mutational signature extraction was performed running non-negative matrix factorization (NMF) as previously described (Alexandrov et al, Nature 2013).

RESULTS:

We have observed and utilized a median number of 5780 (range 2599-7760) substitutions per patient at the asymptomatic stage and 5954 (ranges 2824-8227) at progression to MM stage to extract mutational signatures. NMF extracted 5 main signatures (http://cancer.sanger.ac.uk/cosmic/signatures). Specifically, APOBEC- (signature #2) and age-related signatures (signatures #1 and #5) accounted for 13% (1-21%) and 23% (3.2-40%) of all substitutions, respectively. In addition, we found two known signatures that were not implicated in MM so far: non-canonical AID (Signature #9), contributing to 28% of all substitutions (17-55%); and signature #8, accounting for 28% of all substitutions (13-45%) and pertaining to a yet unknown mutational process. Finally, the fifth signature did not match any of the previously described ones, representing a potential novel process which we defined as MM-1 (7%, range (1-16%).

Interestingly, we found a differential contribution of processes in non-coding and intronic regions where AID was more prevalent, while exonic regions where APOBEC and age signatures were more prevalent. In intronic regions we found widespread regions of kataegis (9/11 patients), reflective of localized hypermutation. In our patients, kataegis was associated with rearrangements in 60% of cases, and was present in both the SMM and MM sample in 84% of cases, suggestive of an early event during tumor development. Contrary to what is observed in solid cancers, APOBEC signature was only responsible for 25% of kataegis variants, vs 70% for AID, suggesting a causative role of aberrant AID activity in shaping the early mutational repertoire of neoplastic plasma cells. To confirm this, we looked at serial samples in our cohort. While the percent contribution of each signature varied in each patient, confirming genomic heterogeneity of MM, it did not change when paired SMM and MM samples from the same patient were compared. This shows that mutational processes required for the development of symptomatic MM act early, and have been already operative at the SMM stage. However, by clustering substitutions as clonal (early variants present at the time of tumor initiation) or subclonal (late variants arisen closer to the time of sampling) using a Bayesian hierarchical Dirichlet process (Bolli et al, Nature Comms 2014), we could analyze processes operative before SMM was diagnosed. NMF analysis of these clusters reported striking differences. Specifically, AID and age were the predominant mutational processes in early substitutions in all patients, contributing to a median of 35% (25-54%) and 30% (15-43%) of variants respectively. Conversely the contribution of AID was minimal among late substitutions (5%, 1-22%), where instead APOBEC, Signature #8 and MM-1 activity was prominent [19% (1-43), 38% (8-73%), 16% (2-50%) respectively].

CONCLUSION:

WGS data allowed the identification of mutational processes operative well before MM becomes clinically evident. Our observation that all samples have signs of aberrant AID at the time of tumor initiation supports a unifying model where MM precursors are initially transformed with the contribution of AID, providing a fertile ground for other later processes (i.e APOBEC and signature #8) to act and shape the final genomic landscape of overt multiple myeloma.