A first-in-primate In Vivo CRISPR/Cas9 screen in a rhesus macaque model identifies drivers of CAR-T expansion, tissue tropism and persistence

Abstract Background: Chimeric antigen receptor T-cell therapy (CAR-T) has revolutionized treatment for B-cell malignancies, yet >50% of patients ultimately relapse, highlighting the need to overcome barriers to durable responses. While CAR-T dysfunction is often driven by T cell-intrinsic inhibitory pathways, how these pathways regulate in vivo proliferation, tissue infiltration, and persistence remains poorly understood. To identify key regulators, we performed the first in vivo CRISPR/Cas9 loss-of-function screen in a non-human primate (NHP) model of B cell-directed CAR-T therapy. This screen enabled high-resolution, cross-tissue dissection of the dominant regulators of CAR-T efficacy in an immunocompetent model that recapitulates human CAR-T biology. This competitive screen uncovered key gene targets that differentially constrain CAR-T expansion, persistence, and trafficking across tissues, offering new opportunities for fine-tuning of CAR-T function. Methods: NHP T-cells were virally transduced to express CD20.BBz CAR and a CRISPR/Cas9 library targeting 1,502 T-cell relevant genes and positive/negative controls, followed by Cas9 electroporation. 6x10⁶ autologous edited CAR-Ts/kg (>1,000x guide coverage) were infused into two NHPs after lymphodepletion chemotherapy. Guide enrichment was analyzed in sorted CD4⁺ and CD8⁺ CAR-Ts from peripheral blood (PB), bone marrow (BM), lymph node (LN), and CSF across CAR-T expansion (days 8-9), contraction (d16-26), and late persistence (d70-90). Top candidates were functionally validated using in vitro cytotoxicity and proliferation assays, and in vivo testing in the NHP CD20 CAR-T model. Results: The two in vivo screen CAR-T recipients demonstrated robust concordance in guide distribution. The infusion products showed expected distribution of positive and negative controls with an ROC AUC >0.8, and importantly, similar distribution of targets to the plasmid library, emphasizing the ability to identify targets that were selected in vivo and not during the in vitro manufacturing process. CAR-T expansion was comparable to non-genetically modified CAR-T controls (58.2+/-13.9% vs 42.3+/-4.7% of total T-cells) and recipients experienced comparable low-grade CRS and ICANS to control recipients. This in vivo screen uncovered a dynamic regulatory network comprising 221 high-confidence gene targets (FDR<0.05) consistently enriched across both animals. These targets stratified into distinct categories: a core set of 15 genes shared across multiple timepoints and compartments, likely representing potent universal regulators of CAR-T cell function; a late-emerging subset enriched during contraction and late time points (95 genes) likely involved in persistence; and compartment-specific hits (23 in PB, 33 in CSF, 9 in BM, 11 in LN) potentially driving tissue-resident activity and infiltration. Among these hits, genes involved in immune synapse formation and T activation were preferentially enriched in PB (MEF2D, CD2, PTPN2, SLAMF6); regulators of cytokine signaling dominated in BM (IFNAR1, IL2RA, IL2RB); cell trafficking and cytoskeletal modulator genes were enriched in LN and CSF (PLXNA4, CTNNAL1, TNS2). Among the top hits consistently enriched across timepoints and tissues were TET2, PI3Kδ, PTPN2, and MEF2D, representing key universal regulators of CAR-T cell efficacy. Functional validation confirmed that deletion of these targets significantly enhanced CAR-T proliferation, increased secretion of pro-inflammatory cytokines (IFN-γ, TNF-α, IL-2), and improved leukemia-targeted cytotoxicity in vitro. In vivo, PTPN2-KO CD20 CAR-T cells demonstrated superior dose-dependent expansion (82.4% vs. 42.3% CAR+ T cells) and more rapid B-cell aplasia compared to WT counterparts, supporting their enhanced therapeutic potency. Notably, the screen also identified enrichment of guides targeting canonical pathways involved in cytokine signaling and T cell activation/effector function (IL2RA, IL2RB, IFNGR, GZMB, and CD2), underscoring the need to fine-tune, rather than simply amplify, CAR-T activation to maximize efficacy while avoiding overstimulation-induced dysfunction. Conclusions: This study pioneers the first in vivo CRISPR-Cas9 screen in an immunocompetent NHP CAR-T model, generating a comprehensive atlas of genetic regulators that shape CAR-T expansion, trafficking to sites of disease, and persistence within the complex in vivo microenvironment, building a roadmap for rational next-generation CAR-T engineering.