Comprehensive Comparison of Early Relapse and End-Stage Relapsed Refractory Multiple Myeloma

Introduction: Treatments incorporating autologous transplantation (ASCT), proteasome inhibitors (PIs), and immunomodulatory drugs (IMIDs) induce deep and durable remissions in most multiple myeloma (MM) patients, resulting in prolonged survival. Yet, patients who suffer from early relapse within 2 years of treatment initiation or become refractory to PIs and IMIDs still have a dismal outcome. The mutational landscape in early relapsed MM (ERMM) and relapsed/refractory MM (RRMM) has been comprehensively described. Several aberrations are associated with these two types of high-risk disease but little is known about the biological difference between them. To this end, we have comparatively and sequentially analyzed whole genome and RNA sequencing data from ERMM and RRMM patients.

Methods: We included 32 patients who had relapsed within 2 years of first-line therapy (ERMM group, 30 after ASCT). Samples were collected at first relapse. Paired baseline samples were available from 17 patients. For the RRMM group, we included 43 patients with a median of 5 prior lines of therapy (range 2 – 13; 88% with ASCT), who had relapsed after PIs and IMiDs. For 22 of them consecutive tumor samples were available. Sequencing data were pre-processed using in-house pipelines. Mutations, indels, translocations, and copy number variants were called using Platypus, SOPHIA and ACESeq. Mutational signatures were identified with MMSig and subclonal structures with SciClone. Differential gene expression was assessed using DESeq, gene set enrichment analysis was performed with hypeR, and gene fusions were detected with Arriba.

Results: Nonsynonymous mutations occurred more frequently in RRMM (median=180) compared to ERMM at first relapse (median=62, p<0.001). While TP53 mutations were more often seen in ERMM (31% of cases vs. 21% in RRMM), NRAS mutations were enriched in RRMM (37% vs. 22%). Bi-allelic inactivation of TP53, RB1, or CDKN2C was more frequent in ERMM (44% vs. 30% in RRMM). In 11/14 ERMM patients these events were already present at baseline. Genes associated with sensitivity to PARP inhibition, homologous recombination deficiency, HECT, Pi3K and NOTCH signaling were more often mutated in RRMM (p<0.05). While mutations associated with PI-resistance were equally common in both groups (~20%), IMID-resistance mutations were more common in RRMM (23% vs 9%). We observed a median number of 60 and 48 fusion genes in ERMM and RRMM, respectively. Fusions involving B2M, TXNDC5, PVT1 and MYC were more frequent in ERMM, while SPINK family and MAGEC1 fusions were more common in RRMM. Analysis of mutational signatures revealed a major impact of signature MM1 (associated with melphalan-exposure), in 66% of RRMM patients. In contrast, only 22% of ERMM samples showed this signature (p<0.001). Signature 3 (defective homologous recombination-based DNA damage repair) was rarely detectable in ERMM (4/32) but one of the major signatures in RRMM (16/43). Analyzing expression profiles, we found upregulated genes in ERMM that were enriched for epithelial-mesenchymal transition, hypoxia, glycolysis and KRAS/IL6-JAK-STAT3 signaling. For RRMM we found no significantly enriched gene set. Yet, 50 upregulated genes were ribosomal protein pseudogenes. Longitudinally, we mainly observed branching evolution in ERMM and RRMM. Major changes in the clonal substructure with new dominant clones were seen in 65% and 55% of ERMM and RRMM, respectively. No changes (“stable” evolution) were rare in both ERMM (3/17), and RRMM (4/22).

Conclusions: According to our results ERMM and RRMM are biologically distinct entities of MM. While ERMM is characterized by inactivation of tumor suppressors and upregulation of gene sets associated with hypoxia and glycolysis, RRMM shows mutations in multiple gene networks, upregulation of ribosomal protein pseudogenes with unknown function and a signature linked to defective DNA repair, suggesting multifactorial mechanisms that lead from first relapse to end-stage relapsed refractory disease. Comparing paired samples, we did not observe major difference in evolution patterns between ERMM and RRMM. Yet, the low prevalence of the melphalan MM1 signature in ERMM suggests selection of pre-existing clones in this entity. In contrast, single tumor cells exposed to melphalan are often the precursors of clones dominating at the RRMM stage, indicating that first-line ASCT has a long-term effect on MM evolution.