AFFILIATIONS
1 Beth Israel Deaconess Medical Center, Boston, Massachusetts, United States, Broad Institute of MIT & Harvard, Cambridge, Massachusetts, United States, Harvard Medical School, Boston, Massachusetts, United States
2 Beth Israel Deaconess Medical Center, Boston, Massachusetts, United States, Broad Institute of MIT & Harvard, Cambridge, Massachusetts, United States, Harvard Medical School, Boston, Massachusetts, United States
3 Multiple Myeloma Research Foundation, Norwalk, Connecticut, United States
4 Beth Israel Deaconess Medical Center, Boston, Massachusetts, United States, Broad Institute of MIT & Harvard, Cambridge, Massachusetts, United States, Harvard Medical School, Boston, Massachusetts, United States
5 Multiple Myeloma Research Foundation, Norwalk, Connecticut, United States
6 Emory University, Atlanta, Georgia, United States
7 Icahn School of Medicine at Mount Sinai, New York, New York, United States
8 Washington University in Saint Louis, Saint Louis, Missouri, United States
9 Emory University, Atlanta, Georgia, United States
10 Beth Israel Deaconess Medical Center, Boston, Massachusetts, United States, Broad Institute of MIT & Harvard, Cambridge, Massachusetts, United States, Harvard Medical School, Boston, Massachusetts, United States
11 Mayo Clinic, Rochester, Minnesota, United States
12 Icahn School of Medicine at Mount Sinai, New York, New York, United States
13 Washington University in Saint Louis, Saint Louis, Missouri, United States
14 Emory University, Atlanta, Georgia, United States
15 Multiple Myeloma Research Foundation, Norwalk, Connecticut, United States
16 Beth Israel Deaconess Medical Center, Boston, Massachusetts, United States, Broad Institute of MIT & Harvard, Cambridge, Massachusetts, United States, Harvard Medical School, Boston, Massachusetts, United States
17 Mayo Clinic, Rochester, Minnesota, United States
18 Icahn School of Medicine at Mount Sinai, New York, New York, United States
19 Washington University in Saint Louis, Saint Louis, Missouri, United States
20 Emory University, Atlanta, Georgia, United States
21 Multiple Myeloma Research Foundation, Norwalk, Connecticut, United States
22 Beth Israel Deaconess Medical Center, Boston, Massachusetts, United States, Broad Institute of MIT & Harvard, Cambridge, Massachusetts, United States, Harvard Medical School, Boston, Massachusetts, United States
ABSTRACT
Introduction: Multiple myeloma (MM) progression and treatment response are influenced by the bone marrow (BM) tumor immune microenvironment (TIME). Our previous single-cell (sc) transcriptomic study identified immune states associated with rapid progression and improved outcome stratification beyond established clinical features. Here, we extend these findings at larger scale and longitudinal depth, generating the largest sc immune proteomic atlas in MM to define immune correlates of survival, progression, treatment, and response.
Methods: We analyzed 974 BM samples from 673 patients enrolled in the multicenter MMRF CoMMpass study (NCT01454297), including 604 baseline samples and longitudinal samples collected from 215 patients for up to 3.5 years after diagnosis. Samples were profiled across 5 academic centers using a harmonized 41-antibody cytometry by time-of-flight (CyTOF) panel. Following quality control, ~280 million cells were retained for analysis. Treatment exposures were grouped according to receipt of autologous stem cell transplantation (ASCT) and maintenance regimen.
To enable atlas-scale integration while preserving common and rare populations, we performed site-stratified sampling and trained CytoVI, a deep generative model for antibody-based sc data, on 15 million cells. The learned latent space was inferred across the complete dataset, and expert-curated cell-state annotations were propagated using multinomial logistic-regression label transfer. Thirteen major immune compartments and >30 T-cell states were resolved. Manual annotation was performed across all identified cell populations. Baseline immune abundances were evaluated in relation to overall survival (OS), progression-free survival (PFS), and rapid progression using clinically adjusted survival and comparative analyses. Serial samples were used to define immune remodeling associated with ASCT and clinical response.
Results: Established clinical factors, including ISS stage, ECOG performance status, creatinine, ASCT, and maintenance therapy, showed the expected associations with OS and PFS, validating the clinical performance of the cohort. At baseline, higher apparently normal plasma-cell (CD19+CD38+CD56-) abundance was the strongest immune correlate of improved OS and remained significant after adjustment for age, sex, collection site, and ASCT (HR 0.67, p=0.006). Higher B-cell abundance showed a concordant favorable association (HR 0.77, p=0.024), whereas greater CD8 central-memory T-cell abundance was nominally associated with improved OS (HR 0.81, p=0.021). Higher γδ T-cell abundance was associated with inferior PFS (HR 1.16, p=0.003), driven primarily by terminally differentiated CD45RA+GZMB+CD57+ (HR 1.13, p=0.036) and CXCR3+CCR4+ γδ T-cell populations (HR 1.13, p=0.029). Patients progressing within 18 months had higher baseline malignant plasma-cell abundance and increased CD4 naïve and central-memory states than patients progression-free for ≥4 years (OR 1.39, p=0.031; OR 1.38, p=0.040). Longitudinally, ASCT was associated with recovery of apparently normal plasma cells (+0.047, q=0.033), contraction of naïve (−0.156, q=3.6×10?²4) and central-memory (−0.095, q=1.4×10??) T-cell states, and expansion of CD8 effector (+0.114, q=1.4×10?¹7) and highly cytotoxic (+0.126, q=7.1×10?8) populations. Patients achieving complete response showed a similar trajectory, with declining malignant plasma cells (−0.046, q=0.030), recovery of apparently normal plasma cells (+0.110, q=3.1×10?4), contraction of CD4 naïve (−0.164, q=3.9×10?¹4) and central-memory (−0.089, q=8.7×10?4) states, and expansion of highly cytotoxic CD8 effector cells (+0.123, q=6.2×10?4). Patients with stable or progressive disease showed trends toward persistent malignant plasma-cell dominance, increased CD4 central-memory cells, and reduced cytotoxic CD8 states.
Conclusions: This 280-million-cell proteomic atlas identifies apparently normal plasma-cell abundance at diagnosis as an independent cross-site correlate of improved OS and implicates terminally differentiated γδ T-cell states in inferior PFS. ASCT and complete response were associated with convergent remodeling from naïve and central-memory states toward cytotoxic CD8 effectors, a transition not observed in patients with stable or progressive disease. These findings establish a high-resolution framework for integrating immune states with clinical features to improve MM prognostication and longitudinal response monitoring.