Sequential Scmultiomics of In Vivo CAR-T Cells Allows Characterization of Transcriptional Differences between Patients, and Identifies IL10 As a Potential Mechanism of Resistance to CAR-T Cells in MM

Background: CAR-T cells have revolutionized cancer immunotherapy, representing a promising option for R/R Multiple Myeloma (MM). Despite high remission rates observed after BCMA CAR-T therapy, many patients still relapse and knowledge of the molecular mechanisms governing CAR-T cell function is very limited. To shed some light on specific transcriptomic programs activated after CAR-T cell administration, we interrogated longitudinal samples of CAR-T cells collected from patients enrolled in CARTBCMA-HCB-01 (NCT04309981) and academic clinical trial in patients with RRMM (Oliver-Caldes A, et al. Lancet Oncol, 2023).

Methodology: We characterized 50.805 CAR-T cells from 11 different samples collected from 3 patients, including final infusion products (IP) and CAR-T cells isolated from bone marrow (BM) and peripheral blood (PB), at one and three months after infusion. Single-cell RNA and TCR sequencing was performed using Chromium Single-Cell Immune Profiling solution from 10x Genomics, which allows simultaneous analysis of gene expression and paired T-cell receptors. Gene Regulatory Network (GRN) analysis was performed using SimiC, a novel machine learning method developed by our group, that infers regulatory dissimilarities from single cell data (Peng J, et al. Commun Biol., 2022)

Results: scRNA-seq revealed that although CAR-T cells from IP presented similar profiles, with highly proliferative CD4⁺ and CD8⁺ memory cells, CAR-T cells remaining after infusion were mainly non-proliferating CD8⁺ cells, with effector/effector-memory phenotypes. Interestingly, transcriptomic profile of CAR-T cells differed among patients with different degrees of response. Partial responders presented increased presence of terminally differentiated effector cells with an exhausted signature, while complete responders presented CAR-T cells in transition to central or effector memory phenotype. Additionally, we found that BM and PB CAR-T cells presented different phenotypes, potentially due to abrogated regulatory mechanisms. Specifically, CAR-T cells infiltrating BM presented increased expression of cytotoxic and exhaustion markers compared to their PB counterparts. GRN analysis with SimiC identified several regulons (PRDM1, ARID4B) with increased activity in CAR-T cells from BM, which could be responsible for these differences. PRDM1 has already been associated with CAR-T cell exhaustion and its depletion promotes TCF7-dependent CAR-T cell stemness and proliferation. ARID4B, a chromatin remodeler, could be acting as an epigenetic regulator of CAR-T cell function. Importantly, combination of scTCR-seq and scRNA-seq allowed the identification of a hyperexpanded CAR-T clone, with immunosuppressor features, mainly present in the BM of a patient with partial response. Deeper characterization showed that this clone had higher expression of cytotoxic and activation markers, with increased expression of IL10. Further analysis with SimiC showed association of IL10 with transcription factors related to exhausted CD8⁺ T cells, like CREM, BHLHE40 or PRDM1, also implicated in the production of IL10 in Treg. Additional in vitrostudies suggested that subsequent activation of endogenous TCR after CAR-T cell activation led to IL10 production, and functional validations corroborated that IL10 reduces CAR-T cell functionality.

Conclusions: Overall, our analysis combining scRNA-seq/scTCR-seq with novel machine learning models, allowed us to characterize key transcriptional differences observed between patients, infusion products and in vivo infused CAR-T, as well as between CAR-T according to their location (PB vs. BM). Importantly, our results identify IL10 as a regulatory mechanism promoting CAR-T cell dysfunction, representing a potential target to be modulated for the development of improved CAR-T therapies for MM.