Paper Details

ABSTRACT

Background:
High-risk multiple myeloma (MM) is currently defined by genomic lesions, yet lesion status alone does not explain divergent outcomes or early relapse. A central unresolved question is whether the marrow microenvironment modifies high-risk biology and exposes therapeutic vulnerabilities. We hypothesized that high-risk MM becomes clinically penetrant when adverse malignant populations occupy sufficient tumor space and are coupled to remodeled marrow ecosystems.

Methods:
We profiled 159 newly diagnosed MM patients from 3 centers by 10x 5' scRNA-seq of whole bone marrow or CD138-depleted marrow. 2025 IMWG criteria classified 77 patients as high-risk (HR) and 82 as standard-risk (SR). TME composition was modeled using compositional models and graph-neighborhood analysis. Early relapse was defined as progression within 18 months. Malignant populations were defined by integrating inferred CNV, somatic SNV, and expression features. Complementary in vivo studies compared SR Vk*MYC tumors with aggressive HR VQ-D1/VQ-D2 NrasQ61R models by flow cytometry across BM, spleen and LN plus surfaceome profiling.

Results:
Composite HR disease showed reproducible but heterogeneous TME remodeling, including increased CD4 T cells, early B-lineage depletion and bidirectional cytotoxic-cell changes (FDR=0.01). Lesion-level analysis resolved distinct ecosystems. TP53-altered disease showed NK/inflammatory-CD4 enrichment with B-lineage and cytotoxic-like/effector-memory neighborhood depletion. Residual cells in the depleted cytotoxic-like compartment were IFN-high (beta=0.159, p<0.01) and proliferation-high (beta=0.071, p<0.01), defining an IFN-active but depleted cytotoxic compartment. In contrast, the combination of 1q gain with 1p deletion defined a myeloid damage-response ecosystem, with expansion of CD16/scavenger and hypoxia/CTSS monocytes, enrichment of KLRB1+ ICOS/OX40 CD4 memory-like cells, and loss of NK, B-lineage and progenitor-associated states.

Baseline marrow states also marked early relapse. HR early relapsers showed developing-B loss with inflammatory CD4/Treg enrichment. Sample-level models also identified a CD34/HSC-like signal. Focused re-clustering of HSCs revealed stress/IFN/adhesion-like HSPC states as the source of this association, and these stress-remodeled HSPC states were strongly enriched in SR early relapsers (RR=4.58, p<10^-4).

Tumor-population occupancy linked adverse genomics to outcome. The cellular burden of 1q gain, rather than binary detection alone, refined prognosis. A higher fraction of gain1q-positive malignant cells predicted shorter PFS after adjustment for co-occurring high-risk lesions, including adverse translocations and del17p (HR=1.73, FDR=0.016), supporting tumor occupancy as a quantitative measure of biological penetrance.

Among 5 immunotherapy targets, GPRC5D was the only one significantly increased in early-relapse tumors, with higher mean expression (beta=0.204, FDR=0.026) and larger GPRC5D-positive tumor fraction (OR=1.82, p=0.012), nominating a clinically actionable GPRC5D-high phenotype at diagnosis. Lesion-resolved analysis also identified BCMA-low/FcRH5-high del17p/del1p populations. We also observed marked upregulation of CDC42BPA in the t(4;14)+1q subtype. CDC42BPA encodes the druggable kinase MRCK. In prior work, we showed that CDC42BPA knockdown suppressed NSD2, NF-κB, and IRF4 signaling, increased apoptosis, and reduced myeloma-cell viability, supporting MRCK as a candidate vulnerability in this HR subtype (Laganà et al, 2018).

Murine models supported the principle that aggressive tumor genetics impose distinct immune ecosystems. Compared with Vk*MYC, VQ-D1/VQ-D2 produced faster disease, with VQ-D2 showing lymph-node tropism and upregulation of CD62L, CXCR3, CXCR4 and LFA-1. At similar tumor burdens, HR tumors induced activated Ki67+ antigen-experienced effector T cells with PD1/TIGIT/LAG3 co-expression, granulocyte/monocyte infiltration, and reduced B-cell numbers/maturation; these changes varied by organ site (BM, spleen and LN), suggesting immune responses and suppressive mechanisms may be organ-specific.

Conclusions:
Human and murine profiling support a biological-penetrance model in which HR MM becomes aggressive through the combination of lesion identity, malignant-population occupancy, and remodeled marrow ecosystems. These data nominate actionable and testable axes, including a GPRC5D-high early-relapse phenotype and a t(4;14)+1q CDC42BPA/MRCK kinase program.