Presence of Tertiary Lymphoid Structures and Exhausted Tissue-Resident T Cells Determines Clinical Response to PD-1 Blockade in Renal Cell Carcinoma

Abstract

We describe a paradigm wherein combined high TLS and low tissue-resident exhausted CD8+ T cells are required for optimal response to PD-1 blockade in RCC. This analysis identifies key determinants of response to PD-1 blockade in advanced RCC and suggests avenues for future immune modulation through rational combination therapy strategies.

Figure 1:

Overview of translational analysis of the HCRN GU16–260 trial (A) Schematic overview of translational study for HCRN GU16–260 trial. If different from patient number, sample number is specified in parentheses. For RNA-seq, 22 patients had matched primary and metastatic samples. For scRNA and scTCR-seq, 2 patients had paired pre- and post-treatment samples. For WES and RNA-seq, pre-treatment or archival (within 1 year) tissue was used for analysis. For scRNA-seq and scTCR-seq, fresh RCC tumor tissue was locally cryopreserved and centrally processed prior to or at resistance to nivo monotherapy for eligible patients. “Created with BioRender.com” (B) Significant recurrently mutated genes in RCC based on somatic mutation calling on WES data. Top tracks indicate histological, clinical, and demographic characteristics. Right histogram indicates frequency of mutation. (C-D) UMAP representation shows well distributed cell types by (C) treatment progression and (D) IMDC risk. (E) UMAP representation of 67,403 high quality cells from scRNA-seq analysis revealed transcriptionally diverse immune and non-immune populations recovered following cryopreservation. Erythrocytes not visualized See also Fig S1 and Tables S1–S5.

Figure 2:

Tertiary lymphoid structures are enriched in patients responsive to nivolumab monotherapy (A) Differential gene expression (DGE) analysis using bulk RNA-seq data reveals an upregulation of B cell-related genes in patients responsive to nivolumab monotherapy (CR/PR, n = 31 ) compared to non-responders (PD, n = 30). Thresholds indicate FDR-adjusted p-value (q-value) less than or equal to 0.05 and log2-fold change greater than 1.9. (B) GSEA normalized enrichment scores for the top 25 ontology gene sets significantly enriched (Q<0.05, FDR-adjusted p-value following phenotype permutation-based two-sided test with 1000 permutations) in patients with complete/partial response to nivolumab compared with progressive disease patients. Highlighted pathways indicate 15/25 (60%) of the most upregulated pathways were B-cell related. (C) RCC tumors exhibited a wide range of TLS signature scores, with validation of high and low TLS gene expression signature by immunofluorescence. Depicts 4 out of 19 representative images corresponding to the two highest and lowest samples by TLS score. (D) TLS signature score was higher in responders (CR/PR, n = 31) than non-responders (PD, n = 30). Boxplots include centerline median and hinges indicating first and third quartiles with lower and upper whiskers extending to at most 1.5× the IQR past first and third quartile, respectively. (two-sided Wilcoxon rank sum). (E) A high (≥median) TLS signature score was associated with improved progression free survival (two-sided univariate cox regression). See also Fig S2 and Tables S2–S9.

Figure 3:

Molecular subtypes in advanced clear cell renal cell carcinoma. A random forest model trained on (Motzer et al., Cancer Cell, 2020(26)) was used to classify clear cell tumors into one of six molecular clusters using bulk RNA-seq data. (A) Distribution of molecular clusters by IMDC clinical risk category. Notably, in this cohort, IMDC favorable risk tumors were not enriched for angiogenic subtypes (clusters 1 and 2) (B) Number of samples per cluster and the proportion of tumors with sarcomatoid differentiation. (C) Overall response rate by molecular cluster showed highest overall response rates in Angiogenic/Stromal cluster 1 (66.66%) and T-effector cluster 4 (63.64%) (D) Kaplan-Meier survival curves of progression-free survival in patients receiving nivolumab monotherapy (two-sided univariate cox regression) (E) Summary of clinical outcomes by cluster showed maximum median PFS and landmark 1-year PFS in cluster 4 (Teff/Proliferative) See also Fig S3–S4 and Table S9.

Figure 4:

Sarcomatoid differentiation is associated with distinct cell phenotypes in RCC non-immune and myeloid compartments (A) UMAP representation of 10,219 non-immune cells revealed tumor, fibroblast, and endothelial populations. Erythrocytes not visualized. (B) Non-immune populations were not associated with changes in either progression-free survival (two-sided univariate cox regression, p > 0.05) or objective response (two-sided Wilcoxon rank-sum test, p > 0.05) (C) Violin and boxplots comparing gene expression module scores for tumor cells in sarcomatoid (n = 3 samples) and non-sarcomatoid (n = 16 samples) RCC, as previously defined in (Bi et al., Cancer Cell, 2021(14)). Tumor cells from non-sarcomatoid RCC (n = 4,696 cells) were enriched for the TP1 signature, while tumor cells from sarcomatoid RCC (n= 2,347 cells) were enriched for TP2. Boxplots include centerline median and hinges indicating first and third quartiles with lower and upper whiskers extending to at most 1.5× the IQR past first and third quartile, respectively. The violin component is indicative of kernel density and encompasses all tumor cells of indicated pathology. (D) Violin and boxplots comparing module scores of gene sets previously found to be enriched in sarcomatoid/rhabdoid tumors confirmed upregulation in sarcomatoid compared to non-sarcomatoid tumor cells. Boxplots include centerline median and hinges indicating first and third quartiles with lower and upper whiskers extending to at most 1.5× the IQR past first and third quartile, respectively. The violin component is indicative of kernel density and encompasses all tumor cells of indicated pathology. (E) UMAP representation of 3,477 myeloid cells revealed distinct macrophage, monocyte, and dendritic cell populations (plasmacytoid dendritic cells not visualized). (F) Myeloid populations were not associated with changes in either progression-free survival (two-sided univariate cox regression, p >0.05) or objective response (two-sided Wilcoxon rank-sum test, p > 0.05) (G) Violin and boxplots comparing module scores of IL10 pathway gene sets showed elevated IL10 pathway expression in macrophages from sarcomatoid tumors (n = 461 cells) compared to non-sarcomatoid tumors (n = 1228 cells). Boxplots include centerline median and hinges indicating first and third quartiles with lower and upper whiskers extending to at most 1.5× the IQR past first and third quartile, respectively. The violin component is indicative of kernel density and encompasses all macrophages of indicated pathology. See also Fig S5 and Tables S10–S12.

Figure 5:

ZNF683+ SLAMF7+ Exhausted CD8+ T cells are enriched in nivo-resistant tumors (A) UMAP representation of 28,268 CD3+ T cells. (B) ZNF683+ SLAMF7+ Exhausted CD8+ T cell population was associated with changes in both progression-free survival (two-sided univariate cox regression, p < 0.05) and objective response (two-sided Wilcoxon rank-sum test, p < 0.05). (C) Volcano plot of differentially expressed genes between ZNF683+ SLAMF7+ Exhausted CD8+ T cells and all other CD3+ T cell clusters. Thresholds indicate FDR-adjusted p-value less than or equal to 2 ×10⁻³⁹ and log fold change greater than 1. (D) The proportion of ZNF683+ SLAMF7+ Exhausted CD8+ T cells was higher for patients with progressive disease (n = 4) compared to patients with complete or partial response (n = 6) for scRNA-seq samples with at least 100 T cells, n =13. (two-sided Wilcoxon rank-sum test). (E) A high (≥median) proportion of ZNF683+ SLAMF7+ Exhausted CD8+ T cells was associated with worse progression free survival (two-sided univariate cox regression). (F) Diagram summarizing T cell functional assay to evaluate SLAMF7 biology. (G-H) Cytokine and granzyme B production by (G) CD8+ and (H) CD4+ T cells overexpressing SLAMF7-GFP or GFP control, co-cultured with SLAMF7-GFP or GFP control K562 cells (n= 5). Each paired point represents the median of one healthy donor, with 1–3 technical replicates per donor (student’s paired t test). See also Fig S6–S9 and Tables S11, S13, and S15.

Figure 6:

Integration of ZNF683+ SLAMF7+ Exhausted CD8+ T cell specific and TLS gene expression signatures in relation to clinical outcomes in patients receiving nivolumab monotherapy (A) ZNF683+ SLAMF7+ Exhausted CD8+ T cell gene expression signature is increased in progressive disease in independent samples from the HCRN cohort. For (non-overlapping) bulk RNA-seq tumor samples in the HCRN GU16–260 cohort (n=81), ZNF683+ SLAMF7+ Exhausted CD8+ T cell GES was increased in patients with progressive disease (n=28) compared to patients with complete or partial response (n=27) (two-sided Wilcoxon rank-sum test) and (B) High ZNF683+ SLAMF7+ Exhausted CD8+ T cell gene expression signature (≥median) was associated with worse progression-free survival in independent samples from the HCRN cohort (bulk RNA-seq; two-sided univariate cox regression). (C-F) Bulk RNA-seq analysis of an independent cohort (CheckMate-025) investigating nivolumab vs everolimus. (C) In nivo-treated patients, ZNF683+ SLAMF7+ Exhausted CD8+ T cell GES score was higher in patients with progressive disease (n =69) compared to patients with complete or partial response (n =39) (two-sided Wilcoxon rank sum). (D) A high (≥median) ZNF683+ SLAMF7+ Exhausted CD8+ T cell GES score was associated with worse progression free survival (two-sided univariate cox regression). (E-F) In everolimus-treated patients, the ZNF683+ SLAMF7+ Exhausted CD8+ T cell signature was not associated with significant differences in (E) objective response or (F) progression-free survival. (G) Patients with both high TLS and low ZNF683+ SLAMF7+ Exhausted CD8+ T cell GES scores (median split) had improved progression free survival relative to all other patients (two-sided univariate cox regression) (H) Summary of results, depicting tumors with high TLS and low ZNF683+ SLAMF7+ Exhausted CD8+ T cells as having the best clinical outcomes with anti-PD-1 monotherapy. “Created with BioRender.com” See also Figs. S10–12 and Table S17.