Paper Details

ABSTRACT

Background: High-risk multiple myeloma (HRMM) is a major driver of early disease progression and myeloma-related mortality. Current risk models rely largely on static cytogenetic abnormalities, which incompletely capture prognosis and disease biology. We hypothesized that HRMM is defined by convergence on a shared malignant transcriptional state rather than by any single genomic lesion.

Methods: We analyzed single-cell multiomic data (scRNA-seq, scATAC-seq, and WGS) from 92 CD138-selected bone marrow samples from healthy donors (n=8), MGUS (n=9), smoldering MM (SMM n=19), newly diagnosed MM (NDMM n=41), and relapsed/refractory MM (RRMM n=23). After modality-specific quality control and singlet filtering, data were integrated using MultiVI, yielding 639,300 cells for Leiden clustering. Clusters were annotated using lineage markers, disease stage, and genomic features. Gene set enrichment, differential expression, and risk scoring were derived from RNA-seq data. Clinical significance was assessed in 598 CoMMpass patients.

Results: Analysis of 5,000 highly variable genes identified 19 transcriptionally distinct cell clusters. Healthy cells were enriched in cluster 11, MGUS in clusters 7, 10, and 17, and NDMM predominantly in clusters 0-6 and 14, whereas RRMM cells were enriched in clusters 9, 12, 13, 15, and 18. MZB1 was broadly expressed across plasma-cell clusters, while BACH2 was largely restricted to healthy cells. Most SMM cells clustered with NDMM, supporting biological continuity between these stages. Cells from distinct molecular subtypes did not fully segregate, suggesting shared malignant cell states across diverse genomic backgrounds.

Four clusters were enriched for KI67-positive cells, with the strongest proliferative signal in cluster 9 and additional proliferative populations in clusters 12, 13, and 18. Cluster 9 was composed almost exclusively of NDMM and RRMM cells (70% RRMM, 29% NDMM). This cluster was enriched for high KI67/PHF19 co-expression, UAMS-PR and GEP70 high-risk signatures, and adverse lesions including 1q gain and 1p loss, indicating convergence of genomic and transcriptional features of aggressive disease into a shared HR malignant state.

This HR state was characterized by enrichment of E2F, MYC, MTORC, and mitotic programs, consistent with a proliferative and chromatin-deregulated phenotype. We identified a 6-gene transcriptional core associated with this state that was prognostic and significantly increased at relapse by linear mixed-effects modeling. Co-expression of PHF19 and KI67 (log2CPM ≥4), present in ~11% of patients, identified a group with markedly inferior outcomes compared with the remaining 89% of patients (median PFS 14.7 vs 36.3 months), with a median OS 29.3 months (p<0.0001), while the remainder of the cohort did not reach median OS. Similarly, a 50-gene signature robustly stratified outcomes, with patients in the top 15% showing inferior survival compared with the remaining 85% (median PFS 18.0 vs 35.6 months), with a median OS of 36.8 months (p<0.0001). The remainder of the cohort again did not reach median OS.

Among the top upregulated genes in the high-risk cluster were oncogenes associated with aggressive non-hematologic malignancies, including ATAD2, an epigenetic reader and transcriptional coactivator near MYC at 8q24. More than 90% of patients contributing to the high-risk cluster harbored MYC translocations, prompting analysis of the broader 8q24 locus. Across 14 MM cell lines, MYC expression correlated strongly with genes in the extended MYC locus including ATAD2 (r=0.84) and PVT1 (r=0.80), but less strongly with NSMCE2 (r=0.22). In patient samples, however, MYC-ATAD2 correlation was weaker, suggesting that this high-risk program may reflect broader regulatory remodeling at 8q24.

Conclusions: These results align with a model in which HRMM evolves over time through structural and epigenetic activation of MYC-associated regulatory circuits, combined with disruption of RB/E2F-mediated cell-cycle checkpoints. Broadly, the data indicate that HRMM represents the emergence of a MYC/E2F-driven malignant plasma-cell state—characterized by cell-cycle re-entry, epigenetic plasticity, stress adaptation, and remodeling of the 8q24 regulatory landscape rather than a static molecular subtype. Together, these findings provide a biologically grounded framework for refined risk stratification and support convergent high-risk circuitry as a potential therapeutic target.