Tumor site-directed A1R expression enhances CAR T cell function and improves efficacy against solid tumors

Abstract

The efficacy of Chimeric Antigen Receptor T cells against solid tumors is limited by immunosuppressive factors in the tumor microenvironment including adenosine, which suppresses Chimeric Antigen Receptor T cells through activation of the A_2A receptor. To overcome this, Chimeric Antigen Receptor T cells are engineered to express A₁ receptor, a receptor that signals inversely to A_2A receptor. Using murine and human Chimeric Antigen Receptor T cells, constitutive A₁ receptor overexpression significantly enhances Chimeric Antigen Receptor T cell effector function albeit at the expense of Chimeric Antigen Receptor T cell persistence. Through a CRISPR/Cas9 homology directed repair "knock-in" approach we demonstrate that Chimeric Antigen Receptor T cells engineered to express A₁ receptor in a tumor-localized manner, enhances anti-tumor therapeutic efficacy. This is dependent on the transcription factor IRF8 and is transcriptionally unique when compared to A_2A receptor deletion. This data provides a novel approach for enhancing Chimeric Antigen Receptor T cell efficacy in solid tumors and provides proof of principle for site-directed expression of factors that promote effector T cell differentiation.

Fig. 1. A₁R expression enhances anti-tumor functionality and effector differentiation of mouse and human CAR T cells.

A Schematic of mouse or human adenosine receptors, A₁R and A₃R, tagged with mCherry/NGFR expression markers. B qRT-PCR cycle threshold normalized against Rpl32, n = 3 independent experiments. C cAMP response to ADO of mouse CAR T cells primed with FSK (1 μM), n = 2 independent experiments. D Cytokine production in mouse anti-HER2 CAR T cells after coculture with E0771-HER2 or 24JK-HER2 tumor cell lines for 16 h. E Cytokine production by human anti-Lewis Y CAR T cells after coculture with OVCAR-3 or MCF-7 tumor cell lines for 16 h in the presence or absence of an A₁R antagonist (DPCPX, 100 nM). F Cytokine production by mouse anti-HER2 CAR T cells after serial coculture with E0771-HER2 tumors every 24 h. G FACS plots and quantification of CAR T cells after 72-hour serial coculture. H GSEA plot of pathways from the mSigDB database based on RNA-seq of tumor stimulated A₁R or A₃R CAR T cells versus Control. I–K Human anti-Lewis Y A₁R CAR T cells stimulated with OVCAR-3 tumor cells for 16 h in the presence or absence of an A₁R inhibitor (100 nM), DPCPX. I FACS plots, J–K marker expression in CD8⁺/CD4⁺ CAR T cells. L Differentially expressed genes of OVCAR-3 stimulated human anti-Lewis Y CAR T cells. M GSEA of PID and KEGG datasets of stimulated hA₁R CAR T cells versus control. N EnrichR plots of differentially expressed genes in both mouse and human A₁R CAR T cells versus control for positively enriched genesets. P values were determined using the Fisher’s exact test/hypergeometric test based on the enrichR package. ****p < 0.0001, ***p < 0.001, **p < 0.01, *p < 0.05, mean ± SD (B–E, G, J, K). One-way ANOVA or (F) two-way ANOVA. D, F, G, H All mouse data is representative of at least 3 independent experiments, with the bulk RNA-seq experiment performed with 2 technical replicates. E, I–L All human data is representative of at least 3 independent experiments. Human bulk RNA-seq was performed with 3 technical replicates. H, M GSEA P-value estimation is based on an adaptive multi-level split Monte-Carlo scheme based off the fgsea package.

Fig. 2. Constitutive A₁R expression drives terminal differentiation of CAR T cells.

C57BL6-HER2 transgenic mice bearing orthotopic E0771-HER2 tumors were pre-treated with 4 Gy of total body irradiation prior to adoptive transfer of 2 doses of 10e⁶ anti-HER2 CAR T cells. 5 doses of 50,000 IU of IL-2 were injected i.p every 24 h. A Tumor growth curve, representative of 2 independent experiments with n = 4-6 mice per group. B CAR T cell counts in spleen and tumors from n = 5-6 mice except for non-treated group (n = 2 mice). Data represented as the mean ± SEM (A, B). C Quantification of %T_SCM (CD45RO^-CD45RA⁺CD27⁺CD62L⁺) in CD4⁺ and CD8⁺ NGFR⁺ human anti-Lewis Y CAR T cells shown as pooled data of means ± SEM from n = 4 independent healthy donors. D CD69 and LAG3 expression in human CAR T cells cultured with or without A₁R antagonist DPCPX (1 μM) in CD8⁺/CD4⁺ CAR T cells prior to tumor stimulation, representative of n = 4 donors. E Volcano plot of DEGs for unstimulated CAR T cells. F GSEA of DEGs enriched in A₁R CAR T cells using the KEGG and Sade-Feldman single-cell RNA-seq datasets (GSE120575). G Correlation of significant DEGs between tumor stimulated and unstimulated hA1R vs Ctrl CAR T cells. H CD8⁺ T cell meta-clusters from integrated atlas of single cell RNA-seq experiments across 21 tumor types and cells from 316 donors. Derived A₁R signature genes up or down-regulated are overlayed onto meta-clusters, I GSEA of up or down-regulated genes from each meta-cluster for unstimulated hA₁R CAR T cells vs control. J Quantification of gene expression in each meta-cluster GSEA for select core-enriched genes from the A₁R signature as defined by ref. . K Heatmap of significant DEGs from the A₁R signature upregulated or downregulated with a logFC≥1 or logFC≤1 respectively between tumor stimulated and unstimulated hA₁R vs Ctrl CAR T cells. ****p < 0.0001, ***p < 0.001, **p < 0.01, *p < 0.05. C paired two-sided t-test (B) One-way ANOVA or (A) two-way ANOVA. Human bulk RNA-seq was performed on a single donor with 3 technical replicates. F, I GSEA P-value estimation is based on an adaptive multi-level split Monte-Carlo scheme based off the fgsea package.

Fig. 3. CAR T cells engineered to express A₁R under the control of the NR4A2 promoter exhibit enhanced CAR T cell function without loss of memory precursors.

A Cytokine secretion of anti-Lewis Y CAR T cells that were pre-conditioned in culture with DPCPX (100 nM) prior to co-culture with OVCAR-3 tumors, representative with mean ± SD of triplicate cultures. B CRISPR/HDR template with homology arms for the NR4A2 gene, to knock-in payload transgene under the control of the NR4A2 promoter. C RNA-seq counts per million (CPM) of ADORA1 gene after serial coculture with OVCAR-3 tumors, represented as the mean ± SD (D) Cytokine secretion and E intracellular cytokine staining in CAR T cells after serial coculture with OVCAR-3 tumors, represented as the mean ± SD of triplicate cultures. F CD69 and LAG3 expression in CD8⁺/CD4⁺ CAR T cells without tumor stimulation. G Quantification of %T_SCM (CD45RO^-CD45RA⁺CD27⁺CD62L⁺) in CD4⁺/CD8⁺ human anti-Lewis Y CAR T cells shown as pooled data of means ± SEM from 3 donors. H Expression of markers following coculture with OVCAR-3 tumor cells, represented as the mean ± SD of triplicate cultures. I Cytokine secretion by anti-HER2 CAR T cells against MCF-7 breast cancer after serial coculture, represented as the mean ± SD of triplicate cultures. J–K Differential accessibility peaks from ATAC-sequencing in J unstimulated or (K) OVCAR-3 tumor stimulated human CAR T cells. L HOMER motif analyses of enriched motifs in NR4A2/hA₁R CAR T cells. M PCA plot for each round of coculture. N Quantificaiton of DEGs with each round of stimulation. O Magnitude of up or downregulated DEGs from the previously identified A₁R signature in NR4A2/hA₁R CAR T cells. Box-plot defined by box (1^st and 3^rd interquartile range), median-line, whiskers extending to points within 1.5x interquartile range. P–Q GSEA of gene-sets from mSigDB Hallmarks, KEGG and Sade-feldman gene signatures (GSE120575) for P hA₁R vs Control or (Q) hA₁R vs NR4A2KO CAR T cell groups after first stimulation with tumor cells. ****p < 0.0001, ***p < 0.001, **p < 0.01, *p < 0.05. A, F, G, K, O One-way ANOVA or (D, E, H, I, L, M, N) two-way ANOVA. C, J–Q Human bulk ATAC-seq and RNA-seq was performed in independent experiments, 3 technical replicates each.

Fig. 4. Tumor site specific expression of hA₁R enhances CAR T cell therapeutic efficacy against solid tumors in vivo.

NSG mice were injected in the fourth mammary fat pad or sub-cutaneously with 1.25 × 10⁶ HER-2^Low MDA-MB231 breast or 5 × 10⁶ LewisY⁺ OVCAR-3 ovarian tumor cells respectively. Once tumors were established (15-20 mm²), mice were then irradiated (1 Gy) and treated with one dose of 5 × 10⁶ anti-HER2 or 15 × 10⁶ anti-Lewis Y CAR T cells, respectively. Mice were supplemented with 50,000 IU of IL-2 on days 0-4 post treatment. A Tumor growth, data shown as means ± SEM of n = 5–6 mice per group (B) Tumor growth, data shown as means ± SEM of n = 7 mice per group and C individual tumor growth curves. D Waterfall plot of % change in tumor size on Day 28 post therapy. E–G Counts of CD8⁺ and CD4⁺ FLAG⁺ CAR T cells in E tumor and spleen at (F) D17 post therapy and G D28 post therapy. H Quantification of %T_SCM CAR T cells in spleen D17 post therapy. B–H Data represented as the mean ± SEM of n = 6 mice per group. I TSNE plots and unbiased clustering of tumor infiltrating CD8⁺FLAG⁺ CAR T cells analyzed by flow cytometry at day 17 post therapy. J Heatmap and K quantification of marker expression sorted by unbiased clustering in tumor at D17 post therapy. Box-plot defined by box (interquartile range, 1^st and 3^rd quartiles), median-line, whiskers extending to SD. L Quantification of memory subsets of CD8⁺ FLAG⁺ T_PEX (PD-1^INTTCF-1⁺CD62L^+/−), T_EFF (PD-1^INTTCF-1⁻) and T_EX (PD-1^HITCF-1⁻) in tumor and spleen at D17 post therapy. ****p < 0.0001, ***p < 0.001, **p < 0.01, *p < 0.05. E–H, L one-way anova or (A, B, K) two-way ANOVA. A, B Statistics shown are NR4A2/hA₁R versus NR4A2KO groups. Data representative of 2 (a, n = 5–6 mice per group) or 3 independent experiments (B, n = 7 mice per group).

Fig. 5. Enhanced effector function of A₁R expressing CAR T cells drives IRF8 dependent distinct transcriptional pathways compared to A_2AR deletion.

Bulk RNA-seq was performed on anti-Lewis Y CAR T cells stimulated for 72 h with OVCAR-3 tumors in the presence or absence of pan-adenosine receptor agonist, NECA (10 μM). A PCA plot, (B) DEGs for of CAR T cells stimulated with tumors in the absence of NECA (top). Venn Diagram showing overlap of up or down-regulated DEGs between NECA vs no NECA treated CAR T cells (bottom). C GSEA of A₁R gene signature for A_2ARKO and NR4A2/hA₁R CAR T cells versus NR4A2KO controls. D Network dendrogram and gene-coexpression modules identified using WGCNA of human CAR T cells after serial coculture with OVCAR-3 tumors over 72 h. E Eigengene score of individual groups at each round of stimulation for the red module, two-way ANOVA, n = 3 technical replicates per group, data represented as the mean ± SD. F Module-trait relationship of key gene signatures against modules eigengene scores. G Heatmap of 221 genes in the red module after 1 round of coculture. H GSEA of the genes identified from the red module using EnrichR. P values were determined using the Fisher’s exact test/ hypergeometric test based off the enrichR package. I Ranked connectivity scores for transcription factors within the red module (J) Network plot of IRF8 and 1st neighbors in the red module. K CPM counts of IRF8 at each stimulation timepoint, represented by mean ± SD of triplicates. L–N IRF8 knockout NR4A2/hA₁R CAR T cells were generated using CRISPR/Cas9 editing and stimulated with OVCAR-3 tumor cells overnight. L Cytokine secretion, M, N activation (CD69⁺CD62L^-) and memory (CD62L⁺CD69^-) phenotype. L–N Data shown as means ± SD of triplicate cultures and representative of 2 independent experiments with 2 healthy donors. ****p < 0.0001, ***p < 0.001, **p < 0.01, *p < 0.05. G, K two-way ANOVA or (L, N) one-way ANOVA. Human bulk RNA-seq was performed on a single donor with 3 technical replicates. C GSEA P-value estimation is based on an adaptive multi-level split Monte-Carlo scheme based off the fgsea package.