Sherry Bhalla, David Melnekoff, BS, MS, Jonathan J Keats, PhD, Kenan Onel, MD, PhD, Deepu Madduri, MD, Joshua Richter, M.D., Shambavi Richard, MD, Ajai Chari, MD, Hearn Jay Cho, MD, PhD, Joel Dudley, PhD, Sundar Jagannath, MD,, Alessandro Lagana, PhD, and Samir S. Parekh, MD
Significant progress has been made in the past 20 years in dissecting the molecular heterogeneity of Multiple Myeloma (MM). Despite such progress, however, current classification systems based on FISH or karyotyping do not fully capture and represent the diversity observed in the patient population and response to treatment still varies significantly within the same cytogenetic groups. Next generation sequencing technologies enable high resolution characterization of DNA mutations, copy number alterations and gene expression in patient tumors. Patient Similarity Networks (PSN) have emerged as a powerful tool to integrate multiple platforms of genetic information across a patient population. In a PSN, patients are represented as nodes and connected with one another based on similar genomic and transcriptomic profiles. We present MM-PSN, the first patient similarity network of MM, generated based on multi-omics data from 655 patients enrolled in MMRF CoMMpass. We identified twelve prognostic subgroups and potential vulnerabilities in each subgroup based on data from large CRISPR/Cas9 screens in MM cell lines and subgroup-specific therapeutic options by applying an RNA-based (connectivity map) drug repurposing approach. Findings from this study have been included in a manuscript submitted to bioRxiv (Bhalla et al, doi: https://doi.org/10.1101/2020.06.02.129767).
In MM-PSN, we integrated 5 data types: gene expression, gene fusions, copy number alterations (CNA), somatic mutations and chromosomal translocations with clinical data from each patient. Spectral clustering of MM-PSN revealed 3 main patient groups and 12 subgroups of highly similar patients, each characterized by specific patterns of genetic and molecular features (Fig. 1A). Group 1 included n=357 patients (54.5%) mainly enriched for hyperdiploidy (HD), MYC translocations and NRAS mutations and was comprised of 4 subgroups. Group 2 included 166 patients (25.3%) and was overall enriched for the MMSET (tMMSET) and MAF (tMAF) translocations. Group 3 included n=132 patients (20.15%), was enriched for the CCND1 translocation (tCCND1) and was comprised of 3 subgroups (Fig. 1B).
Survival analysis revealed several novel prognostic findings that improve on current cytogenetic risk classification systems. For example, while HD patients are considered to have better prognosis, our analysis revealed that concurrent gain(1q) identifies a subgroup of HD patients at higher risk (1c) for relapse (Fig. 1C). Similarly, the MMSET translocation t(4;14) is currently considered to confer poor prognosis and identifies high-risk in R-ISS. Our findings demonstrate significant heterogeneity within this group, where patients in subgroup 2e tMMSET+gain(1q) had the poorest prognosis in terms of both progression-free (PFS) and overall survival (OS) in the entire cohort, while those in subgroup 2a (tMMSET alone) had a significant better prognosis, comparable to that of the HD subgroups (Fig. 1D). This finding has important implications, as it shows that tMMSET does not always confer poor prognosis and that only the fraction of patients with concurrent gain(1q) are at high-risk. Additionally, our model revealed a protective effect conferred by gain(15q), whose presence determined a significantly longer PFS and OS, even after adjusting for gain(1q).
Further analysis of pathways enriched by gene expression in subgroups revealed specific enrichment for interleukin signaling in group 1, growth, proliferation and differentiation pathways in group 2 and inflammation and other immune-related pathways in group 3. We have further identified therapeutic vulnerabilities and treatment options that may be important in a subgroup specific manner, for example CDK6 in the tMAF subgroup 2b, IGF1R in the tMMSET+gain(1q) subgroup 2e and CCND2 across several subgroups of the HD and tMMSET/tMAF groups. Drug repurposing analysis identified novel potential therapeutic options in a subgroup-specific manner, such as XPO1 inhibitors in HD subgroup 1b and MAPK inhibitors in tMMSET+gain(1q) subgroup 2e.
In conclusion, MM-PSN integrates multi-omics data from MM patients and reveals greater molecular and clinical heterogeneity than current cytogenetic classifications. Ongoing research is focused on a deeper investigation of the molecular mechanisms driving each subgroup and further validation of treatment options identified for specific subgroups.
VIEW THE PUBLICATION