Figure S8. Effect of in vivo administration of Compound A on myeloma xenotransplants in NSG mice. a. Schematic representation of the experimental details with a timeline of treatments and tumor measurements. b. Relative luciferase activity in MM1.S and MM1.S-Luc cells. c. Average TBW of mice from each treatment group. d. In vivo effect of Compound A on tumor volume of MM1.S-Luc cells in OT-1 mice. Shown is the average tumor volume (y-axis) vs. time from implantation (x-axis) for each treatment group. e. Tumor monitoring was performed using the IVIS spectrum imaging system where BLI was quantified as total flux (p/s) as a measure of total tumor burden (y-axis) relative to days post-implantation (x-axis). Total flux is a measure of the radiance (photons/sec) in each pixel summed or integrated over the ROI area (cm2) x 4π. Average radiance, a measure of luminescence, was calculated by summing the radiance from each pixel within the ROI and dividing it by the number of pixels or super pixels, resulting in units of photons/second/cm²/steradian.

Figure S1. Chemical structure of the eight hits, designated as Compounds A-H, detected in the high throughput screen. Also shown are the known HDAC6 inhibitors tubastatin-A, ACY-1215, ACY-738, and nexturastat A.

Figure S3. Time and dose-dependent effect of the top eights compounds (A-H) identified in the HTS on (a.) proteasome ChT-like activity and (b.) cell viability.

Figure S4. Effect of Compound A on murine H2Kb MHC-I-SIINFEKL presentation and B3Z tumor lysis. a. Effect of the eight hits on proteasomal ChT-like activity at increasing concentrations in E.G7-ova cells. Compounds that increased proteasome ChT-like activity >1.5-fold and maintained cell viability >50% are represented by red boxes. b. Effect of the eight hits on proteasomal ChT-like activity and SIINFEKL-MHC-I complex presentation in E.G7-ova cells. E.G7-Ova cells were pre-treated with the indicated compounds for 72 h after which they were stained with a monoclonal antibody to SIINFEKL-bound to H2Kb. Fluorescence was quantitated by flow cytometry. Red boxes indicate compounds that increased proteasomal activity >1.5-fold, and increased SIINFEKL-H2Kb presentation >2-fold. c. Effect of the eight hits that increased proteasomal ChT-like activity with SIINFEKL-MHC-I complex presentation on lysis of E.G7-Ova cells by the B3Z T-cells. E.G7-Ova cells were pretreated with compounds (1 uM) after which cells were washed and co-cultured with the H-2Kb-restricted B3Z T-cells genetically engineered to express a TCR that specifically recognizes the SIINFEKL-H2Kb complex. E.G7-Ova cells were co-cultured with B3Z cells at E:T ratios of 1:1, 2:1, and 3:1.

Background Effective TCR-directed cell therapies in Multiple Myeloma (MM) require discovery of shared, tumor-restricted antigens with matched TCRs. Cryptic peptides from untranslated regions, aberrant splicing, and retroelements are promising candidates; however, systematic mapping of their cognate TCR:antigen pairs in MM remains limited. Here, we apply high-throughput antigen prioritization with DNA barcoded (dCODE) MHC-I Dextramer technology to validate epitopes and map TCR:antigen pairs, including cryptic epitopes from untranslated regions. Methods We developed a discovery framework integrating whole exome sequencing, RNA-seq, ribosome profiling, and single-cell (sc) immunoprofiling to identify antigens that are transcribed, translated, and predicted to be antigenic. We applied this to bone marrow (BM) from 10 HLA-A*02:01 MM patient samples from the BMT CTN 1401 trial (NCT02728102). Peptides were prioritized using a machine learning model trained on validated antigens. Prioritized antigens were confirmed for plasma cell specificity via sc RNA-seq, and MHC-I presentation potential assessed using public immunopeptidomic datasets. Top antigens were validated in patient BM and matched with cognate TCRs using sc multiomics with dCODE MHC-I Dextramers. Results Of >2,000 cryptic and mutation-derived peptides, 63 cancer-restricted antigens were prioritized, including 6 shared retroelement-derived epitopes from L1, L2, and ERV1 subfamilies. The L2 element antigen was consistently detected across 10 patients, highlighting its potential as a broadly relevant target. Many additional antigens arose from non-canonical ORFs in plasma cell-associated genes, including BCMA and SLAMF7. dCODE MHC-I Dextramer profiling of patient BM samples revealed distinct phenotypic distributions: viral antigen-specific TCRs (influenza, CMV) were enriched in CD8 tissue-resident and terminal effector memory subsets, whereas cryptic tumor antigen-specific TCRs (ERV1/L2, SLAMF7, BCMA) were enriched in early effector-memory and activated states, suggesting a proliferative potential and the ability to generate terminal effectors with strong cytotoxic function. In this pilot, >6,000 viral-/cryptic antigen-TCR pairs were mapped, representing the largest dataset of its kind in MM, with additional patients ongoing. Conclusion This study provides the first single-cell antigen-TCR atlas of cryptic antigens in MM, uncovering shared, tumor-restricted epitopes from retroelements and non-canonical ORFs and their cognate TCRs. By integrating genomics-driven discovery with dCODE MHC-I Dextramer profiling, we demonstrate a scalable, clinically relevant pipeline linking cryptic antigens to functional TCR states. These findings establish cryptic antigens as a class of actionable targets and deliver a blueprint for vaccine- and T cell-based immunotherapies in MM.

Effect of transient abatacept exposure on MM.1S growth.

e19555 Background: Age at diagnosis has traditionally been considered a host-related factor in MM, influencing treatment tolerability and competing mortality. However, emerging biologic and clinical observations raise the possibility that age-enriched trial populations may differ in disease behavior and treatment responsiveness. We performed a trial-level comparative analysis to evaluate whether median age of enrolled populations is associated with treatment effects across contemporary frontline randomized MM trials. Methods: We systematically assembled a dataset of 18 randomized frontline multiple myeloma clinical trials reporting treatment outcomes and baseline demographic characteristics. Trial-level median age was extracted for each study. The primary analysis used inverse-variance weighted meta-regression to evaluate the association between median age and log(HR) for PFS. A prespecified sensitivity analysis additionally adjusted for baseline control-arm risk using PFS-12 as a proxy. Exploratory analyses evaluated the robustness of findings across selected trial characteristics. Results: Among the 18 identified frontline trials, 14 trials reported sufficient data for inclusion in the primary inverse-variance meta-regression, and 13 trials were included in the sensitivity analysis adjusting for baseline risk. Median age across included trials ranged from approximately 57 to 73 years, reflecting systematic age enrichment across trial populations. In inverse-variance weighted meta-regression, median age demonstrated a modest positive association with log(HR) for PFS, indicating that trials enrolling younger populations tended to show greater relative PFS benefit from experimental therapies compared with trials enrolling older populations. The age association remained directionally consistent after adjustment for baseline risk and in exploratory analyses accounting for trial characteristics, suggesting that differences in control-arm efficacy or selected trial features alone do not fully explain the observed pattern. However, the magnitude of the association was small and substantial heterogeneity was observed across trials. Conclusions: In this trial-level analysis of frontline randomized multiple myeloma studies, age-enriched trial populations demonstrated a modest association with observed PFS treatment effects, with younger-enriched trials tending to show greater benefit from treatment intensification. While exploratory, the consistency of the association across sensitivity analyses suggests that age-related differences in treatment responsiveness may contribute to variability in outcomes across trials. These findings remain hypothesis-generating and do not establish age as an independent prognostic or biologic determinant at the patient level.

Background Profound racial and ethnic disparities persist in multiple myeloma (MM) care, with Black and Hispanic patients significantly less likely to undergo autologous transplant or CAR T-cell therapy despite comparable disease biology and outcomes. As machine-learning (ML) algorithms increasingly inform clinical decision-making and real-world data analysis, there is growing concern that biased models may perpetuate or amplify these inequities. Serum monoclonal protein (M-protein) levels remain a cornerstone for assessing response, relapse, and cellular therapy eligibility, yet incomplete or inconsistent documentation in electronic records creates analytic gaps. We evaluated whether inclusion of race and ethnicity meaningfully improves ML prediction of M-protein values and whether race-agnostic models can preserve accuracy while minimizing bias. Methods We analyzed 619 longitudinal M-protein observations from MM patients. Laboratory and clinical variables were used to train tree-based regression models to predict quantitative M-protein values. Race and ethnicity were encoded as categorical features and selectively included or excluded to assess their incremental predictive contribution. Model performance was evaluated using root mean squared error (RMSE), coefficient of determination (R²), and permutation-based variable importance. Subgroup analyses examined error distributions across racial groups to detect hidden bias. Results Among 542 patients with available demographic data, 90% were Non-Hispanic White, 6% African American, 3% Hispanic or Latino/a, and 4% other groups—reflecting persistent underrepresentation of minority patients in real-world datasets. Excluding race and ethnicity from the feature set did not degrade model performance (RMSE 0.263 vs 0.2631; R² 0.744 vs 0.7445) and, in one variant, slightly improved accuracy. Race and ethnicity ranked among the least informative predictors, and no systematic increase in error was observed within minority subgroups. Conclusions Race and ethnicity were not significant contributors to ML-based prediction of M-protein levels in MM. Their exclusion maintained model accuracy and fairness across racial groups, supporting the feasibility of race-agnostic predictive frameworks. Given that structural barriers—rather than biologic differences—drive disparities in cellular therapy access, analytic approaches that limit reliance on demographic variables may help prevent reinforcement of systemic bias. Incorporating fairness-audited, race-neutral models into registry-based analytics can promote equitable deployment of AI tools in transplant and CAR T-cell populations.

PURPOSEProlonged cytopenias are a common complication after chimeric antigen receptor (CAR) T-cell therapy for multiple myeloma, increasing the risk of severe infection. Infusion of previously collected autologous stem cells may mitigate this risk, but the clinical and economic implications of proactive collection remain uncertain.METHODSWe developed an 8-year, monthly cycle Markov model simulating 10,000 patients undergoing CAR T therapy. Two strategies were compared: (1) no stem-cell boost and (2) availability of a boost for patients with prolonged cytopenias. Transition probabilities for neutropenia, infection, and infection-related mortality were derived from CARTITUDE-4 and published stem-cell boost reports. Costs included hospitalizations for severe infection and upfront stem-cell reserve collection. Deterministic and probabilistic sensitivity analyses were performed.RESULTSIn the base case, universal reserve collection reduced severe infections from approximately 650 to approximately 260 per 10,000 patients and averted approximately 50 infection-related deaths. However, average per-patient costs were higher in the boost arm (approximately $19,700 US dollars [USD] v $4,500 USD), reflecting a gross reserve collection cost of $17,918 USD per patient plus lower residual infection-related hospitalization costs. Survival outcomes were similar between arms, with relapse-related mortality dominating long-term outcomes. Sensitivity analyses confirmed robustness, with hospitalization cost and reserve collection cost identified as the most influential parameters.CONCLUSIONProactive stem-cell collection for CAR T recipients reduces infectious complications and modestly improves infection-related survival but remains economically unfavorable when applied universally. A risk-adapted approach targeting patients at highest risk of prolonged cytopenias may better balance clinical benefit with cost-effectiveness.

CD86 expression on patient MM cells prior to CAR T therapy.

Background The Center for International Blood and Marrow Transplant Research (CIBMTR) increasingly relies on external electronic health record (EHR) data to enhance the precision of transplant and cellular therapy outcomes. However, variability in how key laboratory biomarkers are recorded can undermine data integration and analytic validity. In multiple myeloma (MM), serum protein electrophoresis (SPEP) and monoclonal protein (M-protein) levels remain central to defining response and relapse, yet these data are frequently entered in unstructured text or with nonstandard codes. Such heterogeneity complicates automated data ingestion and linkage with CIBMTR response, progression, and survival endpoints. We evaluated the scope and impact of unstructured M-protein reporting across institutions contributing to a large real-world MM registry. Methods Laboratory records for 1,725 MM patients from 18 U.S. institutions participating in the HealthTree Foundation Registry were analyzed. Facilities were categorized as Academic, Community, or Integrated Network centers, and EHR systems were classified by platform (Epic, Cerner, VA). M-protein results were identified as either standardized (LOINC-coded) or unstructured (free text, custom identifiers, or narrative results). Structured data prevalence was compared across sites, EHRs, and socioeconomic strata based on ZIP-code median income. The impact of data structure on analytic completeness and equity was assessed. Results Only 55.1% of patients had at least one standardized M-protein result; within three months of diagnosis, this dropped to 20.9%. Structured coding rates varied widely across EHR platforms (Epic 47.8%, Cerner 21.2%, VA 23.5%) and were lowest among centers serving lower-income populations. Among unstructured entries, 62% used nonstandard local codes, and 28% contained M-protein values embedded in free-text narratives, rendering them incompatible with automated registry integration. These gaps disproportionately affected community and safety-net institutions, introducing potential socioeconomic bias into downstream analyses and AI models. Conclusions Inconsistent and unstructured laboratory reporting of M-protein values limits the reliability of real-world MM data and constrains linkage to CIBMTR transplant and cellular therapy outcomes. Establishing standardized, LOINC-based reporting practices and harmonized data pipelines across contributing centers is essential to ensure data fidelity, analytic reproducibility, and equitable representation within CIBMTR-integrated registries. Improved structure will enable more accurate modeling of response, relapse, and survival in the era of advanced cellular therapies.

Characterization of short duration of CD28 blockade in mice.

Practice variation in allogeneic hematopoietic cell transplantation may contribute to preventable toxicity, nonrelapse mortality, and excess cost. We evaluated outcomes occurring with the implementation of a program-wide standardization initiative in a medium-sized program. We conducted a single-center retrospective cohort study of consecutive patients undergoing first allogeneic hematopoietic cell transplantation in 2024 during implementation of a system-level redesign. The intervention emphasized high-reliability operations through concise standard operating procedures, formal quality management, standardized recipient and donor selection, a limited conditioning regimen formulary, fluid-sparing conditioning and graft infusion practices, post-transplant cyclophosphamide-based graft-versus-host disease prophylaxis, and an early-discharge outpatient care model. The primary endpoint was 1-year overall survival. Secondary endpoints included graft-versus-host disease-free, relapse-free survival, nonrelapse mortality, acute and chronic graft-versus-host disease, length of stay, and pharmaceutical cost. Ninety-four patients underwent a first allogeneic hematopoietic cell transplantation in 2024. One-year overall survival was 92.6%, graft-versus-host disease-free, relapse-free survival was 56%, and nonrelapse mortality was 1%. The cumulative incidence of grade II to IV acute graft-versus-host disease was 32%, with grade III to IV acute graft-versus-host disease in 5%, and chronic graft-versus-host disease requiring systemic therapy in 18% at 1 year. Median hospital length of stay was 19 days overall and 14 days among adults. Average pharmaceutical cost per patient decreased from $88,810 across the preceding 3 fiscal years to $39,019 under the redesigned care model. In this single-center cohort, implementation of a standardized, high-reliability allogeneic hematopoietic cell transplantation care model was associated with excellent 1-year survival, low nonrelapse mortality, and lower pharmaceutical cost. These findings support the potential value of disciplined operational standardization as a complement to contemporary transplant platforms.

Background Understanding the T cell clonotypic composition within the multiple myeloma (MM) microenvironment is critical to assess the impact of tumor-associated effector T cell populations on disease outcome. We hypothesized a significant overlap between expanded clonotypes in the bone marrow (BM) and peripheral blood (PB), allowing circulating BM-infiltrating lymphocytes to serve as a window into the MM-associated T cell response. BMT CTN 1401 is a phase II trial of MM patients post-autologous hematopoietic cell transplant (HCT) randomized to a personalized dendritic cell/MM fusion vaccine plus lenalidomide maintenance or maintenance alone. Prior studies showed expansion of MM-reactive T cells in vaccinated patients. Here, we integrated paired BM and PB TCR sequencing and gene expression profiling to characterize the clonal architecture and functional features of tumor-associated T cells and their relation to clinical outcome. Methods Single-cell RNA and TCR sequencing was performed on ∼160 paired BM and PB mononuclear cell samples from 40 BMT CTN 1401 patients, collected at enrollment and 1-year post-HCT. Patients were stratified by response: stringent complete response/complete response (sCR/CR, n=26), very good partial response (VGPR, n=9), partial response (PR, n=4), and progressive disease (PD, n=1). We analyzed 178,135 T cell clonotypes (BM: n=47,790; PB: n=130,345) focusing on clonal distribution, BM-PB overlap, and association with clinical outcome. Results At 1-year post-HCT, patients in sCR/CR had higher proportions of expanded PB effector T cells than PR patients. Substantial TCR sharing was observed between PB and BM, particularly among expanded clonotypes. >70% of highly expanded BM TCRβ clonotypes (frequency > 4) were also detected in PB, and >65% of these also expanded in PB (TCR frequency >1) at baseline, indicating a robust circulating pool of MM-associated T cells and coordinated clonal dynamics across BM and PB compartments. Similarly, >70% of expanded PB TCRβ clonotypes (frequency>4) at baseline overlapped with BM clones, suggesting that a large fraction of circulating expanded T cells is MM-associated. TCR overlap between BM and PB was significantly greater in sCR/CR compared to VGPR/PR at baseline and post-HCT, with the largest differences observed post-HCT. These results were consistent across CD8 and CD4 T cell subsets. BM samples had consistently a higher CD4/CD8 ratio than PB. Conclusion High overlap between PB and BM TCR reflects robust clonal expansion and is associated with sCR/CR, suggesting that circulating T cells may serve as surrogates for tumor-infiltrating T cell dynamics. Ongoing studies are mapping dominant clonotypes to tumor antigens and evaluating the impact of vaccination on the T cell repertoire.

Background Depth of response measured by minimal residual disease (MRD) is a critical prognostic factor in multiple myeloma (MM) and is increasingly used as a surrogate endpoint in both transplant and cellular therapy trials. Bone marrow (BM) MRD assessment using next-generation sequencing (NGS, ClonoSEQ) remains the gold standard but is limited by the invasiveness of sampling, spatial heterogeneity of disease, and the infrequency of testing—typically once per year—which can delay recognition of relapse. Circulating multiple myeloma cells (CMMCs), defined as CD138⁺CD38⁺CD19⁻CD45⁻ malignant plasma cells detectable in peripheral blood, have emerged as a minimally invasive biomarker reflecting systemic tumor burden. Data from bispecific antibody studies suggest CMMCs clear rapidly in deep responders and reappear at relapse, but their relationship to BM MRD in patients treated with CAR T-cell therapy or autologous stem cell transplant (ASCT) has not been systematically evaluated. Methods We prospectively studied 51 MM patients undergoing CAR T-cell therapy or ASCT at Roswell Park. Paired peripheral blood and BM aspirates were collected for MRD testing. BM MRD was quantified using ClonoSEQ with a sensitivity of 10⁻⁶. Peripheral blood (4 mL) was processed using the CELLSEARCH® platform following anti-CD138 ferrofluid enrichment and fluorescent staining (CD38-PE, CD19/CD45-APC, DAPI). CMMC positivity was defined as >2 cells/4 mL. Sensitivity, specificity, and concordance between CMMC and BM MRD were calculated, and Spearman correlation assessed quantitative relationships between CMMC counts and BM MRD burden. Results The median patient age was 66 years (range 45–77), and 39% harbored high-risk cytogenetic abnormalities. Most patients (65%) were receiving daratumumab-based triplet or quadruplet regimens, and over half (53%) were in VGPR or better at sampling. CMMCs were detected in 28 of 51 patients (54.9%); among these, 25 were also MRD⁺ in the BM. CMMC detection achieved 89.3% sensitivity, 91.3% specificity, and 90.2% overall concordance with BM MRD (Cohen's κ = 0.79). CMMC counts correlated strongly with quantitative MRD levels (ρ = 0.72, p < 0.001). Performance was comparable between CAR T (88.2%) and ASCT (90.9%) subgroups, supporting assay robustness across therapeutic contexts. Conclusions Enumeration of circulating myeloma cells using CELLSEARCH provides an accurate and reproducible liquid correlate of bone marrow MRD in patients receiving CAR T-cell therapy or ASCT. This minimally invasive approach may enable more frequent, real-time disease monitoring, facilitate early detection of MRD conversion, and inform risk-adapted strategies to optimize outcomes in the era of high-efficacy cellular therapies.

Figure S7. Effect of Compound A pretreatment in vitro on the growth of E.G7-Ova-GFP-Luc tumors in vivo. a. Schematic representation of the experimental details with a timeline of treatments and tumor measurements. b. Phase contrast and FITC images of E.G7-Ova and E.G7-Ova-GFP-Luc cells. Scale bar = 100µm c. Relative luciferase activity in E.G7-Ova and E.G7-Ova-GFP-Luc cells. d. TBW for individual mice treated with either vehicle or Compound A. e. Average total flux as a measure of tumor burden in mice injected with vehicle or Compound A pretreated E.G7-Ova-Luc-GFP cells. Tumor burden (y-axis) was quantified as total flux (p/s) and monitored post-implantation (x-axis). Total flux represents a measure of the radiance (photons/sec) in each pixel integrated over the ROI area (cm2) x 4π. Average radiance was calculated by summing the radiance from each pixel within the ROI and then dividing by the number of pixels, resulting in units of photons/second/cm²/steradian. f. BLI of E.G7-Ova-GFP-Luc tumors from vehicle and Compound A-treated cells. Negative control represents an OT-1 mouse that was not injected with E.G7-Ova-GFP-Luc cells nor treated. Scale bar = 2 cm g. Percent survival of mice following injection of E.G7-Ova-GFP-Luc cells pretreated with either vehicle or Compound A.