Transcriptional stratification reveals distinct molecular subtypes with treatment-specific outcomes in multiple myeloma patients from the GEM2017FIT trial

Abstract

Abstract INTRODUCTION Multiple myeloma (MM) is a genetically and clinically heterogeneous plasma cell malignancy that remains incurable despite advances in therapeutic strategies. While several transcriptomic classifiers have been proposed, the integration of transcriptional subtypes with genomic alterations and treatment response remains poorly understood. In the randomized phase 3 GEM2017FIT trial, patients received either induction with 18 cycles of VMP followed by Rd (control) or KRd ± Daratumumab (D-KRd/KRd). We performed integrative bulk multiomic analyses to investigate whether transcriptional profiling at diagnosis can stratify patients by outcome. METHODS Baseline bone marrow aspirates from 109 newly diagnosed MM patients were analyzed. DNA and RNA was co-isolated from isolated CD138+ plasma cells for bulk RNA sequencing and whole-exome sequencing (WES). Paired tumor-normal WES was performed in 54 cases. Non-negative matrix factorization (NMF) of transcriptomic data identified five discrete transcriptional subgroups (A–E). Groups were characterized by mutational profiles, structural variants (SVs), copy number alterations (CNAs), and pathway enrichment analyses. Clinical outcomes, including progression-free survival (PFS) and overall survival (OS), were analyzed with transcriptional groups and treatment regimens. RESULTS We analyzed 109 MM patients selected by their best achieved response during the treatment. 54 patients achieved complete response and negative MRD; 12 patients achieved complete response with positive MRD; 25 patients achieved partial response; and 18 patients were primary refractory. The five transcriptional subgroups identified (A–E) demonstrated distinct molecular characteristics suggesting that transcriptional stratification captures unique and clinically relevant biological dimensions. Each transcriptional group exhibited a distinct genomic and functional landscape. Groups A and C shared transcriptional hallmarks related to energy metabolism, including oxidative phosphorylation and mitochondrial activity. Group A was further characterized by hyperdiploidy and NOTCH1 mutations, whereas no defining genomic biomarker was identified for Group C. Notably, patients in these two groups demonstrated exceptional clinical outcomes regardless of treatment modality, with only one relapse observed among a combined total of 31 patients, suggesting the presence of a favorable intrinsic transcriptional program. Group E, defined by the presence of ANKRD26 and BRAF mutations, was enriched not only for energy-related metabolic pathways but also for protein biosynthesis and translational machinery. This group exhibited a differential response to treatment: patients receiving D-KRd achieved 100% PFS, while those treated with KRd alone showed no PFS benefit compared to control arm VMP-Rd after 56 months of follow-up, highlighting a strong predictive interaction between this transcriptional subtype and response to Daratumumab. In contrast, Groups B and D were enriched for oncogenic signaling and immune-related pathways, with BRAF and PRPF8 mutations characterizing Group B, and del(13q) and CREBBP mutations prominent in Group D. Importantly, patients within these groups achieved superior PFS and OS when treated with KRd without Daratumumab, indicating a differential benefit profile where the addition of anti-CD38 therapy may not confer further advantage. Collectively, these findings demonstrate that the transcriptional landscape of MM at diagnosis provides prognostic and predictive information beyond cytogenetics or mutational status alone. The transcriptional subtypes not only delineate distinct biological behaviors but also predict differential therapeutic responses, supporting their integration into treatment decision-making frameworks. CONCLUSIONS Transcriptomic profiling at diagnosis reveals distinct molecular subtypes of MM with differential genomic features, biological pathway activation, and treatment responses. These findings underscore the prognostic and predictive utility of transcriptional classification, advocate for the superiority of KRd-based regimens over VMP+Rd and highlight the selective benefit of Daratumumab in specific subtypes. Prospective integration of transcriptomic data may refine patient stratification and guide personalized therapy in MM. Ongoing analyses of APOBEC signatures and mutational signatures aim to elucidate the underlying drivers of therapy resistance and disease progression.