Single-cell transcriptomics reveal mechanisms of efficacy and toxicity in anti-BCMA CAR-T cell therapies for multiple myeloma

Abstract

Abstract Introduction Anti-B-cell maturation antigen (BCMA) chimeric antigen receptor (CAR) T cell therapies, including FDA-approved ciltacabtagene autoleucel (cilta-cel) and idecabtagene vicleucel (ide-cel), have transformed treatment for adult relapsed or refractory multiple myeloma (rrMM). In real-world settings, both agents have demonstrated high initial response rates in heavily pre-treated patients. However, durable remissions are variable: up to 30-70% of patients relapse within 6-12 months. Emerging data suggest that cilta-cel mediates superior progression-free survival (PFS), but at increased toxicity. Meanwhile, the investigational agent anitocabtagene autoleucel (anito-cel) showed a 97% overall response rate with favorable toxicity in its phase 2 iMMagine-1 trial. Understanding the biological basis for differences in efficacy and toxicity across anti-BCMA CAR T cell therapies is key to improving outcomes. Methods We performed a retrospective study of patients treated with cilta-cel (n=23), ide-cel (n=22), or anito-cel (n=14) at Massachusetts General Hospital with biobanked samples available. Patients who developed movement and neurocognitive toxicities (MNTs) were intentionally enriched to study this less frequent complication. We collected, under informed consent, infusion product (IP; Day 0 [D0]) and CAR-sorted peripheral blood mononuclear cells (PBMCs) at day 14 (D14) post-infusion, along with additional CAR-sorted PBMC samples at toxicity onset in select cases, and performed single-cell RNA sequencing (scRNA-seq) with paired T cell receptor sequencing (scTCR-seq). Results Elevated %CAR+ and a HOPX+LAT- CAR+ T cell population at D14 are linked to MNTs in cilta-cel–treated patients We analyzed 173,762 CAR T cells from 109 samples across 59 patients. Among the patients selected who developed MNTs, five had received cilta-cel (three grade 3, two grade 1), and one had received ide-cel (grade 5). No anito-cel–treated patients developed MNTs, consistent with the iMMagine-1 trial data. Among the MNT patients, the median onset was 44 days (range, 20-126). All cases of MNTs demonstrated significantly elevated peripheral CAR+ T cell frequency compared to non-MNTs patients at D14 (%CAR+/T in MNTs: 93.1 ± 7.6% vs. others: 44.0 ± 30.3%, Welch's t-test P < 0.00001; measured by flow cytometry). We identified a distinct CAR+ T cell population in D14 cilta-cel MNT samples and used its top 100 enriched genes to develop an MNT signature score. This population, marked by high HOPX and absent LAT expression, showed features of tissue-resident, effector memory-like phenotype with potential neurotropic properties contributing to MNTs pathogenesis. The MNT signature score was significantly elevated in both CD4+ and CD8+ CAR T cells from cilta-cel MNT patients at D14 (P <0.05 for both; Welch's t-test), and remained high at MNT onset (n=4) and symptom resolution (n=1). In contrast, the score was consistently high in CD8+, but low in CD4+ T cells, from ide-cel and anito-cel patients, and showed no association with the ide-cel MNT case. These findings implicate CD4+ CAR T cells as potential key drivers of MNTs in cilta-cel–treated patients, and suggest the associated transcriptional program may persist beyond clinical resolution. Limited CAR+ T cell expansion at D14 linked to early relapse in cilta-cel–treated patients At D14, cilta-cel recipients with poor response showed a distinct T cell profile dominated by CD4 central memory (CM) or CD8 effector memory re-expressing CD45RA (EMRA) populations. These non-responder-associated subsets exhibited low to undetectable CAR transcript levels, indicating that insufficient expansion of effector CAR+ T cells associates with suboptimal response to cilta-cel. In contrast, no such phenotypic differences were observed in the initial infusion products (D0). We observed no link between D14 CAR T expansion and ide-cel response. Conclusions Our results show that elevated peripheral CAR+ T cell frequency and a HOPX+LAT- CAR+ T cell population at D14 and beyond are linked to MNTs in cilta-cel, offering a basis for investigating neurotoxicity mechanisms and opportunities for early detection. Additionally, limited effector CAR+ T cell expansion appears to drive poor cilta-cel outcomes, underscoring the importance of CAR T cell fitness and persistence. These findings highlight a potential therapeutic window for optimizing cilta-cel outcomes through early monitoring and potential opportunities for clinical interventions.