24TGF-Beta alters CD8+ T cell phenotype and drives resistance to immune checkpoint inhibitors (ICI) in renal cell carcinoma (RCC)

Abstract Background The most successful treatments for advanced RCC have been ICI-based combination therapies. However, the vast majority of patients with advanced RCC ultimately have disease progression despite ICI treatment, thus necessitating further analysis of the immunological differences in patients that respond to or are resistant to ICI therapy. CD8+ T cells display tremendous phenotypic diversity and play a critical role in anti-tumor immunity. However, it is unclear what drives different CD8+ T cell phenotypes in the RCC tumor microenvironment (TME), and how individual phenotypes impact ICI response and resistance. Methods 70 tumor samples from 63 RCC patients were collected, either before (n = 48) or after (n = 22) therapies (VEGFi, n = 9; ICI monotherapy, n = 20; ICI + ICI, n = 17; ICI + VEGFi, n = 9; others, n = 15). 11 samples were also collected from patients without tumors. RCC variants included 59 clear cell and 11 non-clear cell samples. Of these samples, 18 were labeled as clinical benefit (CB, partial or complete response as the best response) and 11 as no-clinical benefit (NCB, progressive disease as the best response). Single-cell RNA sequencing (10x Genomics) was performed on these samples in order to generate a transcriptome of the RCC tumor microenvironment (TME). Graph-based clustering was performed in order to identify cell type populations, which were then annotated using known lineage genes. Non-negative matrix factorization (NMF) was used to identify gene programs within the exhausted CD8+ T cell (Tex) population. The NicheNet algorithm was utilized to predict ligand-target interactions between macrophage cells and the Tex population. Results Within the CD8+ T cell population, Tex cells were identified through elevated expression levels of TOX, PDCD1 (PD-1), and HAVCR2 (TIM-3). NMF generated 4 gene programs within Tex cells, which expressed markers relating to immediate early genes, exhaustion/activation, tissue residency, and stress response respectively. The NMF program corresponding to tissue residency within Tex cells was associated with resistance to ICI-based therapy (P = .05); the stress program was increased in ICI response (P .002). NicheNet identified TGF-beta (TGFB1), produced by macrophages in non-responding tumors, as the ligand with the highest regulatory potential that generated the “resistant” tissue residency program in Tex cells. It specifically identified CD69 and IL7R as notable downstream genes, which were shown to be associated with NCB through differential gene expression (P .001). Notably, the expression of the TGF-beta receptor (TGFBR2) was significantly higher (P = .01) in Tex cells from nonresponsive tumors. To further explore the effect of TGF-beta on the tissue resident program, we sorted naïve CD8 T cells from healthy donor peripheral blood mononuclear cells, and cultured them with stimulation under hypoxia (1% O2) or normoxia, and with or without TGFb. CD69/CD103-double positive tissue-resident-like T cells were found to be significantly higher with TGFb under hypoxia (P = .02). Further, in hypoxic conditions (mimicking the RCC TME), TGF-beta also led to an increase in PD-1 expression on CD8+ T cells (P = .012). Conclusions Through single-cell RNA-seq analysis, we identify an ICI-resistance circuit whereby tumor-associated macrophages produce TGF-beta, which then leads to a tissue residency gene program in Tex cells associated with non-response to immunotherapy. Experimental validation studies demonstrated that TGF-beta and hypoxia are sufficient for the induction of a tissue residency program and PD-1 expression on CD8+ T cells. Overall, this study provides a framework for using scRNA-seq to identify mechanisms of ICI resistance in RCC, and nominates the TGF-beta axis as a potentially targetable pathway to improve CD8 T cell-mediated anti-tumor immunity.