2026 ASH®
Annual Meeting & Exposition December 12-15, 2026 ▪ New Orleans, Louisiana

Paper Details

Abstracts, authors, affiliations, and disclosures.

Monday, December 14, 2026
02:45 p.m. - 04:15 p.m.
MCCNO - La Nouvelle A-B
Technology and Procedures, Omics Technologies, Bioinformatics, Computational Biology, Profiling, Treatment Considerations, Biological Therapies, Bispecific Antibody Therapy, (CAR)-T Cell Therapies, Artificial Intelligence (AI), Deep Learning, Machine Learning, Natural Language Processing, Biological Processes, Pathogenesis, Genomics, Hematopoiesis, Immune Mechanism, Molecular Biology, Diseases, Lymphoid Malignancies, Plasma Cell Disorders, Research, Fundamental Science, Translational Research, Clinical Research, Real-World Evidence, Study Population, Human, Adult, Elderly
PAPER

935 A bone marrow single cell atlas defines conserved immune aging and remodeling from precursor disease to relapsed multiple myeloma.

Joshua Rivera, Ph.D., Yiyang Fan, B.S.2, Anil Aktas-Samur, PhD3, Srikanth Talluri, PhD4, Bahar Dakiki Korucu, MD5, Kenneth Anderson, MD6, Giovanni Parmigianni, PhD7, Nikhil Munshi, MD8, Mehmet Kemal Samur, PhD9
1 Dana-Farber Cancer Institute, Boston, Massachusetts, United States, Dana-Farber Cancer Institute, Boston, Massachusetts, United States, Harvard T.H. Chan School of Public Health, Boston, Massachusetts, United States
2 Fred Hutch Cancer Center, Seattle, Washington, United States
3 Dana-Farber Cancer Institute, Boston, Massachusetts, United States, Dana-Farber Cancer Institute, Boston, Massachusetts, United States
4 Dana-Farber Cancer Institute, Boston, Massachusetts, United States
5 Dana-Farber Cancer Institute, Boston, Massachusetts, United States, Dana-Farber Cancer Institute, Boston, Massachusetts, United States
6 Dana-Farber Cancer Institute, Boston, Massachusetts, United States
7 Dana-Farber Cancer Institute, Boston, Massachusetts, United States, Harvard T.H. Chan School of Public Health, Boston, Massachusetts, United States
8 Dana-Farber Cancer Institute, Boston, Massachusetts, United States, Dana-Farber Cancer Institute, Boston, Massachusetts, United States, Harvard T.H. Chan School of Public Health, Boston, Massachusetts, United States
9 Dana-Farber Cancer Institute, Boston, Massachusetts, United States, Dana-Farber Cancer Institute, Boston, Massachusetts, United States, Harvard T.H. Chan School of Public Health, Boston, Massachusetts, United States

Introduction
The bone marrow immune microenvironment is altered during progression of multiple myeloma, yet the timing, extent, and mechanisms are not fully understood. Existing studies are limited to individual cohorts or underpowered stage representation and do not provide a reference across healthy bone marrow, MGUS, SMM, NDMM, relapse, immunotherapy, and remission.

Methods
We integrated a single cell atlas of more than 3 million bone marrow immune cells from healthy donors (NBM=222), precursor plasma cell disorders (MGUS=17, SMM=42), and multiple myeloma states (NDMM=326, RRMM=81, MM-immunotherapy=48, MM-remission=56). Clinical age, stage, survival, and outcomes were incorporated. We evaluated transcriptomic dysregulation, cell representation, and transcription factor programs driving the SMM phenotype. We developed lineage resolved bone marrow immune aging models from healthy donors and evaluated predicted immune age across disease stages after adjustment for chronological age. We developed an autoencoder and positive-unlabeled (PU) classifier to identify MM associated CD4+ T cell states by deviation from healthy latent expression space.

Results
Nine immune cell states were sufficiently represented across NBM, SMM, and NDMM for cross-dataset analysis. Leave-one-dataset-out (LODO) modeling identified SMM and NDMM enriched genes ranging from 50 to 192 genes per subpopulation, together with ~70 SMM specific genes whose changes were significantly stronger than in NDMM. The largest SMM enriched programs were observed in memory B cells and GZMK CD8+ T cells, with evidence across CD4+ and CD8+ T cell, NK-cell, and monocyte compartments. LODO DESeq2 modeling provided cross-method support across all nine cell states. SMM immune cell types shared enriched genes including NFKB1, NFKBIZ, REL, BHLHE40, TGFB1, CSRNP1, and PELI1. Integrated regulatory analysis identified candidate regulators, including FOXO3, JUN, ATF6, FOXP3, and TP53.

Compared with healthy bone marrow, SMM showed disease-associated transcriptional programs across lymphoid, NK-cell, and MHCII monocyte compartments, including features significantly stronger than in NDMM, indicating that changes associated with MM progression occur during precursor stages. However, cell state representation changed further across disease stages, including increases in GZMB and GZMK CD8+ T cells, memory CD4+ T cells, IFN-γ cells, and other activated lymphoid populations, while myeloid, progenitor, and stromal populations broadly decreased.

Bone marrow immune aging models revealed lineage specific trajectories. CD8+ T cell predicted immune age was already higher than expected during precursor disease and further elevated in symptomatic and relapsed MM. CD4+ T cell predicted immune age remained comparable to healthy bone marrow during precursor disease but became higher during symptomatic disease. B cell predicted immune age was lower than expected during precursor disease and higher during symptomatic and relapsed MM, with the strongest effects in early B cell populations. In NDMM, high composite immune age was associated with inferior overall survival (p=0.0037), with significant associations for B cell (p=0.016) and CD8+ T cell immune age (p=0.021). Longitudinal profiling during immunotherapy revealed dynamic remodeling of immune age states. Autoencoder analysis identified CD4+ T cell changes not captured by cell frequency alone. At relapse myeloma diagnosis, more than half the CD4+ T cells were classified as MM associated by deviation from healthy latent expression space despite no change in overall CD4 T cell frequency, while only 30-40% retained healthy like expression state.

Conclusion
This bone marrow single cell atlas defines immune remodeling across multiple myeloma initiation and progression. Conserved transcriptional and regulatory changes are already present at SMM stage, indicating the features of the MM associated immune microenvironment emerge before symptomatic disease. SMM also contained proinflammatory and cytotoxic lymphoid features that were not fully represented in NDMM. Lineage resolved immune aging and deep learning models identified clinically relevant changes not captured by cell frequency alone, while higher immune age at diagnosis was associated with adverse outcomes. Together, these findings show that immune remodeling begins early, progresses with disease, and may contain features relevant for SMM risk stratification and early intervention before MM develops.

Joshua Rivera, Ph.D.: No relevant conflicts of interest to declare.