Longitudinal Transcriptomic Analysis of CAR-T Cells at Single-Cell Level Allows the Identification of Molecular Mechanisms Promoting Long-Term CAR-T Cell Persistence in MM

Background: CAR-T therapies targeting BCMA are a promising option for Multiple Myeloma (MM) patients. However, a significant number of patients still relapse and the limited persistence of CAR-T cells after infusion has been associated with lack of long-term responses. To shed light on the potential mechanisms driving CAR-T cell persistence we performed longitudinal transcriptomic analysis, at single cell level, of CAR-T cells from a patient enrolled in CARTBCMA-HCB-01 clinical trial (NCT04309981) (Oliver-Caldes A, et al. Lancet Oncol, 2023), that presented detectable CAR-T cells more than one year after CAR-T cell administration.

Methodology: Single-cell transcriptomic and TCR analysis was performed on 42.719 CAR-T cells isolated from infusion product (IP) and paired bone marrow (BM) and peripheral blood (PB) samples at 1, 3, 6, 12 and 18 months after infusion (11 samples), using the Chromium Single-Cell Immune Profiling solution from 10x Genomics. Gene Regulatory Network (GRN) analysis was performed using SimiC, a novel machine learning method developed by our group, that infers regulatory dissimilarities form single cell data (Peng J, et al. Commun Biol., 2022).

Results: We observed that only 4 out of the 30 patients included in the first pilot academic clinical trial evaluating ARI0002h presented detectable CAR-T cells 12 months after treatment (Oliver-Caldes A, et al. Lancet Oncol, 2023). We performed a deeper analysis of one patient still presenting high levels of CAR-T cells 18 months after treatment that finally ended up relapsing at month 23. scRNA-seq analysis revealed that post-infusion CAR-T cells were mainly composed by effector/effector-memory CD8⁺ cells, with a polyclonal TCR repertoire. Interestingly, this patient suffered a respiratory tract infection with CMV reactivation during the first month after infusion that was also reflected on CAR-T cells features. Thus, circulating CAR-T cells at M1 expressed high levels of genes involved in IFN response, with an equivalent anti-viral response in CAR-T cells obtained from BM at M3. However, phenotypes of CAR-T cells from both PB and BM samples converged at M6, with a prevailing CAR-T cell population (termed ‘CD8⁺ GZMK’) presenting cytotoxic features and expressing key markers related to T cell memory and survival that could explain the prolonged CAR-T cell persistence. However, at M12 we observed changes on CAR-T cell populations, with increased presence of CAR-T cells with more cytotoxic features, enrichment of effector and activation signatures, and expressing lower levels of memory markers (‘CD8⁺ GZMB’ cluster), indicating CAR-T cell reactivation. Finally, applying GRNA analysis with SimiC we identified several regulons (FOS and NR4A2), presenting increased activity at M12 that could represent potential drivers of CAR-T reactivation in response to tumor relapse. Further analysis at month 18 are being performed to elucidate mechanisms of tumoral resistance in the presence of CAR-T cells.

Conclusions: Our analysis clearly showed that most CAR-T cells after administration presented expression of key markers related to T cell memory and survival without clonal expansion, suggesting that memory induction is crucial for long-term persistence. Interestingly, prior to tumor relapse, we observed reactivation of CAR-T cell with the enrichment of effector and activation signatures, and GRN analysis identified FOS and NR4A2 regulons as potential drivers involved in CAR-T cell reactivation. Overall, scRNA-seq/scTCR-seq coupled with novel machine learning models provides relevant mechanistic insights that could uncover useful targets to be modulated for the development of long-term persistent CAR-T cells.