High Risk Multiple Myeloma Demonstrates Marked Spatial Genomic Heterogeneity Between Focal Lesions and Random Bone Marrow; Implications for Targeted Therapy and Treatment Resistance

Introduction:

Recent next generation sequencing studies have defined the mutation spectrum in multiple myeloma (MM) and uncovered significant intra-clonal heterogeneity, showing that clinically relevant mutations are often only present in sub-clones. Longitudinal analyses demonstrated that tumor clones under therapeutic pressure behave in a “Darwinian” fashion, with shifting dominance of tumor clones over time. Recently, stratification of clonal substructures in distinct areas of the tumor bulk has been shown for multiple cancer types. So far, spatial genomic heterogeneity has not been systematically analyzed in MM. This stratification in space is becoming increasingly important as we begin to understand the contribution of Focal Lesions (FL) to tumor progression and emergence of drug resistance in MM. We have recently shown that high numbers of FL are associated with gene expression profiling (GEP) defined high risk (HR). A comparison of GEP data of 170 paired random bone marrow (RBM) and FL aspirates showed differences in risk signatures, supporting the concept of spatial clonal heterogeneity. In this study we have extended the analysis by performing whole exome sequencing (WES) and genotyping on paired RBM and FL in order to gain further insight into spatial clonal heterogeneity in MM and to find site-specific single nucleotide variant (SNV) spectra and copy number alterations (CNA), which contribute to disease progression and could form the basis of adaptation of the tumor to therapeutic pressure.

Materials and Methods:

We included 50 Total Therapy MM patients for whom paired CD138-enriched RBMA and FL samples were available. Leukapheresis products were used as controls. For WES we applied the Agilent qXT kit and a modified Agilent SureSelect Clinical Research Exome bait design additionally covering the immunoglobulin heavy chain locus and sequences located within 1Mb of the MYC locus. Paired-End sequencing to a minimum average coverage of 120x was performed on an Illumina HiSeq 2500. Sequencing data were aligned to the Ensembl GRCh37/hg19 human reference using BWA. Somatic variants were identified using MuTect. For detection of CNA we analyzed Illumina HumanOmni 2.5 bead chip data with GenomeStudio. Subclonal reconstruction was performed using PhyloWGS. Mutational signatures were investigated using SomaticSignatures. The GEP70 risk signature was calculated as described previously. Informed consent in accordance with the Declaration of Helsinki was obtained for all cases included in this study.

Results:

Analyzing RBM and FL WES data, we detected between 100 and 200 somatic SNVs in covered regions, with approximately 30% of them being non-synonymous, and less than 5% stop gained or splice site variants. A comparison of paired RBM and FL WES data showed different extents of spatial heterogeneity. Some pairs had very similar mutation profiles with up to 90% shared variants, whereas others demonstrated marked heterogeneity of point mutations. We did not detect differences in mutational signatures between RBM and FL using the ‘SomaticSignatures’ package. We found site-specific driver mutations with high variant allele frequencies, indicating replacement of other clones in these areas. For example we observed a clonal KRAS mutation exclusively in the RBM, whereas a NRAS variant was only identified in the paired FL. The same holds true for large-scale CNAs (>1 Mb). We identified a case in which the high risk CNAs gain(1q) and del(17p) were only detectable in the FL. Further examples for site-specific CNAs were a del(10q21) and a gain(4q13) detected in FLs only. As a prominent pattern, we observed outgrowth of sub-clonal RBM CNAs as clonal events in the FL. Based on mutation and CNA data we identified different forms of spatial evolution, including parallel, linear and branching patterns. Of note, a stratified analysis by GEP70-defined risk showed that a more pronounced spatial genomic heterogeneity of SNVs and CNAs was associated with HR disease. 

Conclusion:

We show that spatial heterogeneity in clonal substructure exists in MM and that it is more pronounced in HR. The existence of site-specific HR CNAs and driver mutations highlights the importance of heterogeneity analyses for targeted treatment strategies, thereby facilitating optimal personalized MM medicine.