Spatially controlled enhanced anti-tumor efficacy via CAR T cells engineered with a CTLA-4-based autoregulated circuit

The widespread application of CAR T cell therapy to solid tumors is hindered by antigen expression heterogeneity, on-target off-tumor toxicity, and limited functional persistence of CAR T cells. Next generation engineering strategies aiming to overcome these hurdles rely on multiple antigen-targeting approaches that fail to simultaneously address all the limitations and/or the use of synthetic receptor transcriptional circuits, which suffer from slow activation kinetics. Here, we leveraged the unique trafficking properties of CTLA-4, regulated by a rapid internalization process, to create CTLA-4-based activation-inducible antigen receptors (CAVI-Rs). CAVI-Rs were exclusively expressed on T cells upon activation with remarkably fast on/off kinetics and self-sustainable expression. In in vitro and in vivo models of heterogeneous tumors, CAR+CAVI-R dual-engineered T cells demonstrated enhanced cytotoxicity and tumor-selective, spatially restricted functionality. Additionally, intermittent co-expression of the CAVI-R lead to improved T cell metabolic and transcriptional fitness, resulting in superior expansion, persistence and anti-tumor efficacy. The CAVI-R provides a novel, simple, safe, potent, and fully-human approach to expand CAR T cell applicability while improving anti-tumor outcomes. Overall design: Double transduced CAR+CCR or CAR+CAVI-R T cells were cultured in 6-well tissue culture plates (Greiner Bio-One), with irradiated (60Gy) MM.1s cells in an E:T ratio of 2:1 and were expanded by weekly stimulations. T cell samples were isolated before the first stimulation (day 0) and at day 14 of expansion (day 14). T cells were stained and sorted (BD FACSMelody™ Cell Sorter) based on double positivity for dsRed (BCMACAR-dsRed) and LNGFR (CCR/CAVI-R-LNGFR) with 100% purity. Samples were spun down at 1500 RPM and stored at -80oC with RNALater (ThermoFischer) for further processing. RNA isolation was performed with RNeasy Plus Mini kit (Qiagen). Libraries were sequenced with the Illumina SBS technology at a sequencing depth of at least 75 million per sample (Macrogen). Paired-end reads were trimmed using Trimmomatic44 0.39 before mapping to the human transcriptome (Gencode v42)45 using Salmon 1.9.046. Mapping results were imported in R 4.4.0 using the Tximeta 1.22.1 package47 and features were annotated using biomaRt 2.60.0. Differential expression analysis was performed using DESeq2 1.44.048 and the results were ranked by -log10(p-value) * sign(foldchange). Geneset Enrichment Analysis was performed on the ranked results using clusterProfiler 4.12.049 and plotted using the dotplot function from Enrichplot 1.24.0. Volcanoplot was produced using EnhancedVolcano 1.22.0. Heatmap was produced by centering the counts to 0 after a variance stabilizing transformation using vst function from DESeq2 for genes of interest and plotting them using ComplexHeatmap 2.20.050, 51.