Shared immune features correspond to high-risk multiple myeloma across multiple human subtypes and murine models

Abstract Background Novel quadruplet, bispecific and CAR-T therapies have extended survival in myeloma, nonetheless about 20 % of patients still relapse within 12–24 months. The 2025 IMWG consensus genomic staging (CGS) recognizes four high-risk (HR) categories: TP53 loss, co-occurring IgH translocations with 1q gain or 1p loss, double-hit chromosome 1, and increased high β₂-microglobulin with normal creatinine. However, it remains unclear how different genetic lesions remodel the tumor micro-environment (TME) or why a minority of CGS-standard-risk (SR) tumors relapse early (“functional HR”). Here we integrate single-cell multi-omics from newly diagnosed patients with immunocompetent mouse models to define, for the first time, lesion-specific immune programs that drive aggressive disease. Methods We profiled fresh bone-marrow aspirates from 159 newly diagnosed patients (77 CGS-HR, 82 CGS-SR) by 10x Genomics 5′ single-cell RNA-seq (435,000 cells). Copy-number profiles were inferred with inferCNV and somatic SNVs with SComatic. A weighted-nearest-neighbour (WNN) approach integrated RNA, CNV and SNV to delineate tumor clones. TME composition was quantified per patient and tested by Dirichlet regression, negative-binomial GLM, Propeller and scComp. Ligand–receptor (LIG-R) interactions were ranked by LIANA. To model HR versus SR in vivo we used Vk*MYC (Vk14451O) as SR and two aggressive HR lines, VQ-D1 and VQ-D2 with Nras Q61R mutations (Wen et al., 2021) and profiled their TMEs by high-dimensional flow cytometry. Results WNN partitioned patient tumors into a median of eight clones (range 3–35). Clone number itself was not prognostic, but a single-clone fraction ≥19 % doubled the hazard of death (HR 2.3, p = 0.044), indicating that clonal sweeps, rather than diversity, influence outcome. Single-cell cytogenetics re-classified 9 % of chart-review SR cases as HR and improved the progression-free-survival separation (p < 0.01). In HR patients the TME showed a depletion of naïve B-cell precursors and NK cells, with expansion of activated/exhausted CD4+ and CD8+ T cell subsets, indicating common immune stress and altered hematopoiesis. LIG-R analysis revealed a monocyte to T cell Alarmin axis in HR tumors, S100A8 → CD69 from CD14+ Monocytes to CD4+ TNF+ T effectors. We also observed lesion-specific immune changes. In TP53-deleted cases, we observed an increase of CD163/206+ M2 macrophages and CD14+ monocytes. LIG-R analysis showed that M2-derived APOE engaged LRP1 and IL-10 → IL-10R signaling on monocytes, reinforcing suppression. IgH translocation + 1q-gain exhibited expanded T regs and CD16+ monocytes with depletion of TNF+ CD8 effectors. Dominant LIG-R pairs were TGFB1 → TGFBR1 (T reg → CD8+ T eff) and HMGB1 → HAVCR2/TIM-3 (CD16+ Monocyte → CD8+ T eff), explaining TIM-3-mediated exhaustion. Additional lesion-specific changes included NK depletion in t(14;20) with 1q gain cases, dendritic cells recruitment in t(14;20) with 1p del, and NK expansion in β₂-microglobulin-high. Analysis of functional HR confirmed the functional impact of these programs. At 24 months, 32/40 early progressors were CGS-HR, whereas eight CGS-SR tumors that relapsed carried the same activated-T/suppressive-myeloid signature. Both VQ-D1 and D2 mouse models led to faster cell death than Vk*MYC, however VQ-D2 was more aggressive and showed tropism to lymph nodes. VQ-D1/D2 tumors mirrored three human HR findings: (i) activated yet exhausted CD4/CD8 T cells (Ki-67, PD-1, TIM-3, TIGIT), (ii) reduced B-cell activation, and (iii) increased monocytes/neutrophils with macrophage re-programming. In the highly aggressive VQ-D2 murine models we also identified numerous changes in expression of adhesion molecules and chemokine receptors, including over-expressed CXCR3/4, matching chemokine pathways uncovered in the human LIG-R map. Conclusions We report for the first time distinct modules in the TME that correspond to specific HR subtypes (both by 2025 IMWG classification as well as functional high risk), with concordance between patients and mouse models. Future work will further dissect the functional role of these components and determine potential interventions for prospective testing in clinical trials to improve outcomes for high-risk myeloma patients.