Investigation of T cell phenotypes associated with response or resistance to immune checkpoint inhibitors (ICI) through single-cell analysis of renal cell carcinoma (RCC).

4515 Background: RCC is notable for a high CD8+ T cell infiltration despite its modest tumor mutational load. However, CD8+ T cell infiltration does not correlate with ICI response, highlighting the need to understand cellular composition and phenotype. We conducted a comprehensive dissection of the tumor microenvironment (TME) using pre- and post-ICI treatment samples to identify specific T-cell populations associated with ICI treatment efficacy in RCC. Methods: A total of 70 tumor samples (n = 59 clear cell; n = 11 non-clear cell) from 63 patients with RCC were collected before (n = 48) and/or after (n = 22) systemic therapies (VEGFi, n = 9; mono-ICI, n = 20; ICI + ICI, n = 17; ICI + VEGFi, n = 9; others, n = 15). This cohort contained 12 paired samples on pre and post from 5 patients, and 58 unpaired samples. Responders (R) were defined as complete and partial responses (n = 22), and non-responders (NR) as disease progression (n = 33) according to the best response based on RECIST. We performed single-cell RNA-sequence (scRNA-seq) on all samples and established a transcriptomics atlas in RCC. We utilized established gene expression signatures to interrogate cellular composition and functional states for samples from ICI-treated patients. We used non-negative matrix factorization (NMF) to identify gene programs, offering superior feature preservation and interpretability. Results: 443,337 high-quality viable cells were annotated to lymphoid, myeloid, tumor, endothelial, or fibroblast compartments, capturing the RCC TME landscape. Among CD8+ T cells, we observed significant heterogeneity, particularly in exhausted T cells (Tex) expressing PD-1 and TIM-3. Tex in NR showed enrichment for tissue-residency and innate-like genes and gene programs, exemplified by significant upregulation of ZNF683 (p = 0.031) and ITGAE (p = 0.0041). In contrast, Tex in R exhibited a marked upregulation of heat shock protein genes, such as HSP1B (p < 2.22E-16) and DNAJB1 (p < 2.22E-16), highlighting a distinct genomic profile. Notably, through NMF analysis, Tex in R showed a significantly higher stress response program and terminal exhaustion program than in NR at baseline and after ICI treatment. Further analysis through gene signature scoring showed an association between Tex in R and enhanced IFN and chemokine activities, stress response, and terminal differentiation post-ICI. Conclusions: Our single-cell transcriptomic analysis uncovered the relationship between Tex with active stress responses and ICI efficacy, additional suggesting T cell revival with ICI-exposure. This study identifies the specific Tex characteristics associated with ICI responsiveness, highlighting scRNA-seq as a scientific strategy for deep correlative analysis in large patient cohorts, and emphasizing the need for further investigation into the unique intricacies of the RCC TME.