Identifying antitumor T cell receptors for cellular immunotherapy

Division Of Basic Sciences - Nci

Investigators

Linked publications, trials & patents

Abstract

Following my recent appointment as a Stadtman tenure-track Investigator at the National Cancer Institute, Center for Cancer Research, I have now acquired key reagents and the necessary equipment and technology for carrying out my research. We have now begun acquiring data for multiple studies understanding the phenotypic states of antitumor T cells from cancer patients. 1) We have initiated a collaborative study with Dr. Diego Chowell (Mount Sinai School of Medicine, NY) to design a TCR-prediction machine learning model that integrates both single cell transcriptomic states along with TCR-aminoacid sequences (transforming them into a large language model) to design an advanced antitumor TCR prediction model for cell therapy. In preliminary analyses, we applied the TCR-language model trained on unpaired TCR- TCR aminoacid sequence data on multiple patients' single cell transcriptome and t cell receptor data (scRNA-scTCR) from Surgery Branch (without regard for specificity). Two-dimensional projection of multiple CD8+ TCRs targeting the same tumor neoantigen self-assembled into sequence-specific clusters, even though their CDR3 sequences differed between each TCR. The TCR-language model was also able to discriminate TCR-sequence between individuals with different HLA genotypes suggesting that the model could indirectly infer TCR-HLA usage. These promising preliminary data suggest that a TCR protein language model can learn and discriminate TCRs based on their sequences and can be applied to our paired scRNA-scTCR data to train on TCR- specificities. We are currently training the model to accurately integrate both transcriptomic and TCR-sequences into an integrated multimodal TCR-prediction model for experimental evaluation. 2) We have acquired data evaluating the importance of the coreceptor (CD8 or CD4) in genetically-engineered TCR-transduced T cells in cell therapies. We have currently evaluated more than 30 neoantigen-specific HLA class I- and class II-restricted TCRs for their ability to recognize and function in their opposite class T cells (i.e. HLA-I restricted TCRs in CD4 T cells, and HLA-II-restricted TCRs in CD8 T cells respectively). We observed a spectrum of CD8-coreceptor dependence in neoantigen-specific TCRs that closely correlated with anti-tumor efficacy. CD8-coreceptor dependence was inversely correlated with published neoantigen-specific TCR affinities7 in tetramer assays (Spearman r = -0.96, p = 0.0028). TCRs targeting the KRAS G12V A*11:01-restricted neoantigen demonstrated a similar spectrum of coreceptor dependence. Two of these new TCRs appeared to outperform the existing TCR (utilized in a phase 2 clinical trial) with respect to tumor recognition and tetramer evaluation of coreceptor dependence. These data were reported in the 2024 SITC Immunotherapy conference (Dinerman et al, SITC 2024 : https://doi.org/10.1136/jitc-2024-SITC2024.0367). This study is currently being prepared for submission as a peer-reviewed research article to a journal in the upcoming months (Dinerman et al, in preparation for submission). 3) We have initiated a collaborative study with Dr. Malcolm Sim (Oxford University, UK) trying to understand the biochemical features of effective antitumor TCRs and their impact on the phenotypic states of T cells in vivo in humans. In preliminary studies we have characterized the TCR-affinities of TCRs specific against KRAS G12V restricted by A*11:01 (described above). We have determined that the TCR with the highest affinity is also the most coreceptor independent and has the highest efficacy against patient-derived tumor material. We are currently performing these analysis on additional neoantigen-specific TCRs that we have identified in my group and within the Surgery Branch to determine which biochemical peptide-HLA or TCR properties are useful for stratifying the efficacy of antitumor TCRs for cell therapy.