Tumor-Intrinsic and Microenvironmental Determinants of Impaired Antitumor Immunity in Chromophobe Renal Cell Carcinoma

Abstract

Introduction:

While immune checkpoint inhibition (ICI) has transformed the management of many advanced renal cell carcinomas (RCCs), the determinants of effective anti-tumor immunity for chromophobe RCC (ChRCC) and renal oncocytic tumors remain an unmet clinical and scientific need.

Methods:

Single-cell transcriptomic and T-cell receptor profiling was performed on tumor and adjacent normal tissue of patients with ChRCC and renal oncocytic neoplasms. Using machine learning, the cellular origin of renal oncocytic neoplasms was evaluated, with analysis of associated oncogenic pathways. Using immunohistochemistry, immune infiltration was analyzed in renal oncocytic neoplasms in comparison to clear-cell RCC (ccRCC). Immune checkpoint expression, clonal expansion, and tumor specificity were compared between ChRCC and ccRCC. Using the International Metastatic RCC Database Consortium (IMDC) dataset, clinical outcomes of patients with metastatic ChRCC treated with first-line systemic regimens were compared to patients with ccRCC.

Results:

We validated α-intercalated cells as the cellular origin of renal oncocytic neoplasms. We identified a downregulation of HLA class I molecules with enrichment of potentially targetable pathways including mTOR and ferroptosis in ChRCC. The tumor microenvironment of ChRCC showed markedly decreased immune infiltration, with a pronounced depletion in tumor-infiltrating CD8+ T-cells. ChRCC-infiltrating CD8+ T-cells demonstrated lower immune checkpoint expression, diminished clonal expansion, and decreased tumor specificity. Clinical analysis identified poor survival outcomes selectively among patients with metastatic ChRCC treated with immune-based therapies.

Conclusion:

Immunogenomic analysis of ChRCC revealed profound depletion of T-cells, with an immune phenotype marked by a lack of expression of immune checkpoint and poor tumor specificity, suggesting that the few T-cells in these tumor types are likely non-specific “bystanders”. This immune-cold environment hinders an effective response to immunotherapy and underscores the need for ChRCC-tailored treatments designed to improve tumor-specific T-cell infiltration into the microenvironment.

INTRODUCTION

Chromophobe renal cell carcinoma (ChRCC) is the second most common type of non-clear cell renal cell carcinoma (nccRCC), accounting for approximately 5% of all kidney cancer cases.^{1, 2} While metastatic disease is identified in a lower subset of patients with ChRCC (~7%) as compared to other histologies, patients who do develop advanced ChRCC unfortunately have poor clinical outcomes, as evidenced by a median overall survival of only 23.8 months.^{1, 3} ChRCC occurs sporadically, but also affects patients with two hereditary autosomal dominant genetic syndromes, namely Birt-Hogg-Dube (BHD) syndrome and tuberous sclerosis complex (TSC).^{4, 5} At the molecular level, hallmarks of ChRCC include multiple whole-chromosome losses (including frequent losses of chromosomes 1, 2, 6, 10, 13, 17, and 21) and a low tumor mutational burden.⁶ Based on early histopathological analyses and more recent explorations of bulk RNA-sequencing (RNA-seq) data, it has been hypothesized that ChRCC originates from intercalated cells of the distal tubule (DT) in the nephron.^{6, 7} This is opposed to ccRCC, where extensive investigations at the single-cell level support a proximal tubule (PT) origin, thereby implying distinct oncogenesis mechanisms in each tumor type.⁸

The definitive diagnosis of ChRCC can be challenging, in part due to shared morphologic and immunohistochemical characteristics with renal oncocytic neoplasms.^9–11 Notably, the most recent 2022 World Health Organization (WHO) classification of kidney cancer expanded the group of “oncocytic and chromophobe renal tumours” to include, in addition to ChRCC, other renal entities, some of which were previously classified as “RCC, not otherwise specified”.¹² Renal oncocytic neoplasms are benign tumors, and are comprised of diverse subtypes, the including renal oncocytoma (RO) tumors and low-grade oncocytic tumor (LOT).

LOT represents a novel entity which lies in the spectrum of oncocytoma- or chromophobe- like renal neoplasms, classified in a dedicated sub-category (“other oncocytic/chromophobe RCC”) in the 2022 WHO classification of kidney cancer.¹² Histologically, LOT is characterized by diffuse immunohistochemical (IHC) reactivity for CK7 but negative KIT reactivity.¹⁰ From a genomic standpoint, TSC1/2 mutations or activating mTOR mutations have been reported in LOT.¹⁰ ROs are predominantly (but not exclusively) benign renal tumors, characterized by the accumulation of defective mitochondria, and which can present with overlapping histopathological features with eosinophilic ChRCC.^{9, 13} Earlier genomic analyses identified full-chromosome losses (i.e., chromosomes 1, 14, 21) to a lower extent in RO as compared to ChRCC.^14–17

Recently, several phase II trials evaluated immune-checkpoint inhibitor (ICI)-based regimens – a standard of care in metastatic clear cell RCC (mccRCC) – across nccRCC subtypes.^18–20 Patients with advanced ChRCC experienced little or no benefit from these treatments. These studies typically include only small numbers of patients with ChRCC, limiting overall interpretation and leading to some conflicting results.^{19, 21} Several other therapeutic approaches for metastatic ChRCC have included vascular endothelial growth factor (VEGF) inhibitors (e.g., VEGF tyrosine kinase inhibitors [VEGF-TKI]) and mammalian target of rapamycin (mTOR) inhibitors (mTORi) monotherapy, with more recent efforts evaluating a combination of both regimens (e.g., lenvatinib and everolimus).^22–25 The optimal therapeutic regimen remains unknown, in large part owing to the poorly understood biology of this disease.

Prior efforts to characterize ChRCC and renal oncocytic tumors using bulk transcriptomic sequencing have advanced our understanding of the molecular biology of these neoplasms, but were limited in their ability to define the tumor-immune microenvironment and the distinct phenotypic states of immune cell populations.^{6, 14} Moreover, while ChRCC has been hypothesized to originate from the distal tubule and/or the collecting duct, its exact cellular origin and that of other renal oncocytic tumors has not been fully determined. Additionally, as preliminary evidence from recent studies appears to show a possible poor response to ICI regimens in patients with advanced ChRCC, such findings require further evaluation, specifically focusing on understanding the determinants of anti-tumor immunity in this disease. Finally, there is a concrete need to understand the properties of the cellular immune repertoire in ChRCC. Single-cell sequencing has emerged as a technology ideally suited to investigate complex and heterogeneous tumors at the cellular level, enabling to understand the transcriptional properties of tumor, epithelial and immune populations across many cancer types, notably ccRCC.^26–29 And while prior works aimed to leverage scRNA-seq to explore the transcriptomic profile of ChRCC, such efforts have been limited by a small sample size of a single ChRCC tumor, not enabling a complete exploration of the cellular immune and epithelial compartments.^{30, 31} Single-cell T-cell receptor sequencing (scTCR-seq) has also emerged as a relevant method to fully characterize T-cell repertoires, enabling the discovery of different states of clonal expansion in relation to immune-based therapies.

Herein, we sought to investigate the tumor-intrinsic and immune microenvironment characteristics of ChRCC and renal oncocytic neoplasms using single-cell transcriptomic and T-cell receptor profiling of ChRCC, LOT and RO tumors. Additionally, we aimed to comprehensively evaluate the clinical outcomes of patients with advanced ChRCC treated with various types of systemic anti-neoplastic therapies.

RESULTS

Single-cell profiling of renal chromophobe and renal oncocytic neoplasms

We collected fresh tumor specimens and adjacent non-tumor tissue from patients with pathologically confirmed renal chromophobe/oncocytic neoplasms, including ChRCC (n=3), LOT (n=1) and RO (n=1) (Figure 1A). ScRNA-seq and scTCR-seq was performed using the droplet-based 10x Genomics platform (Figure 1B). Following quality control and filtering, transcriptomic data were available for 46,817 cells (Figure 1C, Data S1, Methods), across ChRCC, LOT, RO and normal kidney samples (Figure S1A, S1B). Graph-based clustering analysis resulted in the identification of 23 cell types (Figure 1C, Figure S1C), with a representation of all sample types across immune and epithelial clusters (Figure S1D). Most of the analyzed tumor samples originated from the primary kidney tumor (n=4), with disease extent ranging from stage I for RO, stage III for LOT and stages II/III ChRCC, while one sample originated from a positive retroperitoneal lymph node for ChRCC (Figure 1D). At the time of sample collection, none of the patients had received systemic anti-neoplastic therapies.

Figure 1: Single-cell profiling and histopathological evaluation of renal oncocytic neoplasms.

(A) Overview of tumor malignant potential and hematoxylin and eosin (H&E) images of renal oncocytic neoplasms evaluated.

(B) Single-cell transcriptomic, single-cell TCR profiling, and immunohistochemistry of renal oncocytic neoplasms and adjacent tissue.

(C) Uniform manifold approximation and projection (UMAP) of malignant and non-malignant cells captured across all samples, colored by broad cell type. Granular cell types and states were discerned through iterative reprojection and unsupervised clustering of lymphoid, myeloid, and epithelial and tumor compartments, and merged into broader cell-type categories for this visualization.

(D) Summary of clinicopathological features across profiled patients with renal oncocytic neoplasms.

(E) Overview of copy-number alterations inferred from single-cell transcriptomic data across all evaluated tumor samples.

Abbreviations: CAFs: cancer-associated fibroblasts; cDC: classical dendritic cells; ChRCC: chromophobe renal cell carcinoma; DT-IC: distal tubule – intercalated cells; ECs: endothelial cells; IC: intercalated cells; ILC1 / trNK Cell: innate lymphoid cells 1 / tissue-resident natural killer cells; LOT: low-grade oncocytic tumor; Mem./Naïve B-Cell: memory/naïve B-cell; Mem./Naïve T-Cell: memory/naïve T-cell; NK Cell: natural killer cell; pDC: plasmacytoid dendritic cell; PT: proximal tubule; RO: renal oncocytoma.

Inferred copy number variation (CNV) from scRNA-seq data identified full-chromosome deletions in the three tumor ChRCC samples (i.e., CH1-T, CH103-T, and CH4-T), notably in chromosomes 1, 2, 6, 13, 17, and 21 (Figure 1E, Methods). This finding is consistent with the known genomic profile of ChRCC.^{6, 32} In contrast, LOT and RO cells did not harbor large CNV events, with the exception of deletions in chromosome 1, in concordance with prior reports.^15–17

ChRCC and renal oncocytic neoplasms share a common cell-of-origin

In an effort to elucidate the cellular origin of ChRCC, we first analyzed scRNA-seq data generated from matched normal samples in our cohort (n= 784 cells), identifying all major cell types of the healthy adult nephron across its different portions (i.e., PT cells, Loop of Henle (LOH)-DT cells, and collecting duct cell types: principal cells [PC], alpha-intercalated [ICA] cells, and beta-intercalated [ICB] cells) (Figures 2A and 2B). Through a previously developed and adopted machine-learning approach for the identification of the cell-of-origin of ccRCC (Methods), and using scRNA-seq data of epithelial cells from normal kidney samples in our cohort as a training set, we developed a model to predict the cellular identities of distinct kidney epithelial cells based on their transcriptomic profile, achieving significant accuracy in classifying cell types, with a high predicted similarity (>0.8 [range: 0–1]) across each cell type, and a low predicted similarity (< 0.5) across different cellular subtypes (Figure S2A). To identify the cellular origin of ChRCC, LOT and RO, the model was then tested on scRNA-seq data of tumor cells from each tumor type. In ChRCC, RO and LOT, tumor cells had the highest predicted similarity with ICA cells (Figure 2C). To further validate these findings, we obtained external scRNA-seq data of ChRCC cells as a second independent testing set, showing similar results with ICA cells having the most pronounced predicted similarity with ChRCC (Figure 2C).³¹

Figure 2: Single-cell transcriptomic analysis reveals the cell-of-origin of renal oncocytic neoplasms and identifies putative oncogenic alterations in ChRCC.

(A) Schematic of cell types of the adult healthy kidney.

(B) Uniform manifold approximation and projection (UMAP) of healthy kidney epithelial cell types from adjacent tissue and tumor cells across all analyzed samples.

(C) Heatmap of the predicted similarities of tumor cells from each sample from the local and external validation cohorts using a machine-learning model (N-binomial regression) trained on transcriptomic data (scRNA-seq) of the healthy kidney cell types.³¹

(D) Expression levels (normalized expression) of previously established canonical markers of ChRCC across kidney cell types of the adult healthy kidney.

(E) Differential gene expression of ChRCC versus ICA cells using single-cell transcriptomic data. Significantly downregulated genes (fold change < −1.5 and Q < 0.05) in ChRCC (versus ICA) are shown in blue, and significantly upregulated genes in ChRCC (versus ICA) are shown in red.

(F) Differential pathway expression (selected) of ChRCC versus ICA using single-cell transcriptomic data. Downregulated pathways in ChRCC (versus ICA) are shown in orange, and significantly upregulated pathways are shown in purple.

Abbreviations: Ag: antigen; ChRCC: chromophobe renal cell carcinoma; ER: endoplasmic reticulum; Ext. Valid.: external validation; ICA: alpha-intercalated cells; ICB: beta-intercalated cells; LOH-DT: loop of Henle – distal tubule; LOT: low-grade oncocytic tumor; PC: principal cells; PT: proximal tubule; RO: renal oncocytoma.

To further validate our findings, we used a second independent scRNA-seq dataset of normal adult kidney epithelial cells from the Kidney Precision Medicine Project (KPMP) atlas as a training set for a model (Methods) developed using a similar approach (Methods).³³ The KPMP scRNA-seq dataset is composed of 16,397 cells, encompassing nine epithelial cellular subtypes of the normal nephron (Figure S3A), with a range of 108 – 5,066 cells in each labeled cluster, enabling a more granular and representative annotation of epithelial cellular subtypes. The model trained on the KPMP scRNA-seq data was able to achieve a high accuracy in the prediction of normal epithelial kidney cellular subtypes (Figure S3B). Using the same testing set with the second model (i.e., KPMP-trained model), ICA cells were identified as the cellular subtype to have the highest predicted similarity with ChRCC, LOT, and RO cells (Figure S3C). Similar findings were also observed in relation to external scRNA-seq data of ChRCC cells with the second model (Figure S3C).

To further validate the cellular origin of ChRCC, we evaluated the expression of previously described ChRCC-specific markers (i.e., FOXI1, RHCG, KIT, CLCNKA, CLCNKB and RHBG) across cellular epithelial subtypes of the normal kidney, using scRNA-seq from (1) matched normal samples in our cohort (Figure 2D), and (2) normal human kidney scRNA-seq data from the KPMP atlas (Figure S3D).^34–38 Across both datasets, we found that ICA cells have the highest degree of expression of ChRCC-specific markers, further supporting their transcriptional similarity with ChRCC.

To better understand transcriptional changes related to the tumorigenesis of ChRCC, we performed a differential gene expression (DGE) analysis between ChRCC and its cell-of-origin (i.e., ICA cells) using scRNA-seq data. Top upregulated genes (in ChRCC versus ICA) included KLK1, NUPR1, FTL, and FTH1 (Figure 2E). KLK1 has been previously validated as a specific marker of ChRCC.³⁹ NUPR1 is a stress-inducible transcription factor shown to act as major driver of ferroptosis resistance.⁴⁰ FTL and FTH1 encode the two subunits of ferritin, a cytosolic iron storage protein. Similarly to NUPR1, both FTL and FTH1 have both been shown to inhibit ferroptosis. In cancer models, FTH1 was shown to promote resistance to ferroptosis, whereas degradation of FTL leads to its overactivation.^{41, 42} These findings are consistent with the previous identification of ferroptosis as a key molecular axis in the pathogenesis of ChRCC, with potential associated therapeutic vulnerabilities.^{43, 44} Top downregulated genes in ChRCC (versus ICA) cells included ADGRF5, HSPA1A, HSPA1B, HLA-A, HLA-B, HLA-C, HSPA8, HSPA5 and HSPA6 (Figure 2E). HLA-A, HLA-B and HLA-C are HLA class I genes, and downregulation is associated with resistance to immunotherapy.^{45, 46} HSPA5, HSPA6, HSPA8, HSPA1A and HSPA1B are part of the HSPA (HSP70) of heat shock proteins, and have been shown to suppress ferroptosis.^47–49 Similar results were also observed for DGE analysis between ChRCC and ICA cells when using KPMP scRNA-seq data as control (i.e., ICA) (Figure S2E). Differential pathway analysis (DPA) conducted on scRNA-seq data between ChRCC and ICA cells identified significant upregulation of ferroptosis pathways, p53 pathway, mTORC1 signaling and IL-15 signaling (Figure 2F). A top downregulated pathway was antigen processing and presentation (Figure 2F). DGE and DPA analyses between LOT and ICA cells yielded overall comparable results (Figures S2B and S2C). This highlights the enrichment of distinct oncogenic pathways in ChRCC, along with a downregulation in key pathways involved in tumor antigen processing.

Defining the immune microenvironment of renal oncocytic neoplasms

To assess the differences in the tumor microenvironment between ccRCC and renal oncocytic neoplasms, especially ChRCC, we employed a multimodal approach combining histopathological examination and scRNA-seq. Immunohistochemistry staining for CD45+ immune cells was performed on 5 ccRCC tumors and 5 oncocytic neoplasms (1 RO, 1 LOT and 3 ChRCC). All oncocytic tumors displayed markedly reduced immune infiltration as compared to the most common kidney cancer type, ccRCC (Wilcoxon p = 0.007) (Figures 3A and 3B).

Figure 3: Characterization of the immune landscape of renal oncocytic neoplasms at the single-cell level identifies major differences in immune cell infiltration and composition compared to ccRCC.

(A) H&E and CD45 immunohistochemistry (IHC) images of renal oncocytic neoplasms and ccRCC.

(B) Quantification of CD45+ cells from CD45 IHC images in renal oncocytic neoplasms versus ccRCC (cells/mm2) (two-sided Wilcoxon signed-rank test).

(C) UMAP of immune cells captured across all samples, colored by broad cell type. Granular cell types and states were discerned through iterative reprojection and unsupervised clustering of lymphoid and myeloid compartments, and merged into broader cell-type categories for this visualization.

(D) Composition of immune cells across analyzed samples (renal oncocytic neoplasms and normal adjacent kidney).

(E) Heatmap of immune canonical and functional marker expression in individual immune cell populations. Expression values are scaled between minimum and maximum for each gene across all clusters.

(F) UMAP of immune cells across renal oncocytic neoplasms samples (left panel) and normal adjacent kidney samples (right panel).

(G) Proportion of major immune cell types, as a percentage of all immune cells, derived from annotated single-cell transcriptomic data, in ChRCC versus ccRCC samples.²⁷

Abbreviations: cDC: classical dendritic cells; ccRCC: clear-cell renal cell carcinoma; ChRCC: chromophobe renal cell carcinoma; H&E: hematoxylin and eosin; ILC1 / TR NK Cell : innate lymphoid cells 1 / tissue-resident natural killer cells; LOT: low-grade oncocytic tumor; NK Cell: natural killer cell; Norm. Adj. Kidney: normal adjacent kidney; pDC: plasmacytoid dendritic cell; PT: proximal tubule; RO: renal oncocytoma, UMAP: uniform manifold approximation and projection

**: p-value < 0.01

We then examined the different immune subpopulations in oncocytic neoplasms, comparing them with normal adjacent tissue and ccRCC samples. To that end, we analyzed scRNA-seq data from 33,505 high-quality immune cells. Unsupervised clustering of lymphoid and myeloid cells revealed a broad representation of the major immune subpopulations across samples (Figures 3C, 3D, 3F). Furthermore, immune cell type annotations were validated through the expression of canonical biomarkers for each cell type (Figure 3E). The immune subpopulations identified were highly represented from ChRCC and renal oncocytic neoplasms, as opposed to the adjacent normal kidney (Figure 3E).

We then compared the distribution of immune cell types derived from annotated scRNA-seq data between ChRCC and ccRCC.²⁷ The tumor immune microenvironment in ccRCC (versus ChRCC) was enriched in CD8+ (44.6 vs 9.6%, respectively, Fisher’s exact p<0.001) and CD4+ T-cells (12.3 vs 3.2%, respectively, Fisher’s exact p<0.001), while ChRCC (versus ccRCC) displayed a higher relative proportion of B-lineage (20.6 vs 1.4%, respectively, Fisher’s exact p<0.001) and myeloid cells (34.7 vs 17.7%, respectively, Fisher’s exact p<0.001) (Figure 3G). These findings demonstrate a marked depletion of CD8+ T-cells in ChRCC and support a distinct composition of the tumor-infiltrating immune repertoire.

Characterization of the T-cell immune compartment and immune checkpoint expression in ChRCC

Given the poor infiltration and low representation of T-cells in ChRCC, it remained unclear whether T-cell populations in ChRCC could represent active mediators of anti-tumor immunity, particularly in the setting of ICI-based therapies. Therefore, we sought to characterize T-cell phenotypic features in ChRCC, notably in relation to immune checkpoint expression. Immunohistochemistry (IHC) evaluation staining for CD8 and PD-1 was performed on 4 ChRCC and 3 ccRCC samples. Based on IHC image analysis, we identified a markedly decreased CD8 T-cell infiltration along with decreased PD-1 expression in ChRCC compared to ccRCC (Figure 4A). To further quantify and compare PD-1 expression between ChRCC and ccRCC in the setting of CD8 T-cell depletion in ChRCC, we computed the density of PD-1 to CD8 IHC expression (density PD-1/CD8) across tumor samples. The density ratio of PD-1 to CD8 tended to be lower in ChRCC tumors compared to ccRCC (Wilcoxon p=0.11), suggesting that CD8+ T-cells in ChRCC display lower PD-1 expression (Figure 4B).

Figure 4: Combined immunohistochemistry, bulk and single-cell transcriptomic analyses reveal a markedly decreased expression of clinically actionable immune checkpoint molecules in ChRCC (as compared to ccRCC).

(A) CD8 and PD-1 immunohistochemistry (IHC) images of ChRCC and ccRCC.

(B) Density ratio of PD-1/CD8 based on quantification from IHC images in ChRCC versus ccRCC (two-sided Wilcoxon signed-rank test).

(C) Normalized expression of clinically relevant immune checkpoints in ChRCC versus ccRCC versus pRCC based on analysis of bulk transcriptomic data from The Cancer Genome Atlas (TCGA) (two-sided Wilcoxon signed-rank test).

(D) Expression of clinically relevant immune checkpoints at the single-cell level in CD8+ T-cells of ChRCC versus ccRCC samples (Braun et al., 2021), as well as adjacent normal kidney samples (two-sided Wilcoxon signed-rank test).

Abbreviations: Adj. Norm.: adjacent normal; ccRCC: clear-cell renal cell carcinoma; ChRCC: chromophobe renal cell carcinoma; KICH: kidney chromophobe; KIRC: kidney renal clear cell carcinoma; KIRP: kidney renal papillary cell carcinoma; TPM: transcript per million.

**: p-value < 0.01; ****: p-value < 0.0001; ns: non-significant.

To further evaluate these concepts, we assessed the expression of immune checkpoints in ChRCC, in relation to ccRCC. First, using bulk RNA data from The Cancer Genome Atlas (TCGA), we compared the expression of known clinically actionable inhibitory checkpoints (PD-1, CTLA-4, HAVCR2, LAG3 and TIGIT) between ChRCC, ccRCC and papillary RCC (pRCC), after normalizing for purity (Methods).^{6, 50, 51} Overall, ChRCC tumors exhibited significantly decreased expression of all these inhibitory checkpoint markers as compared to both ccRCC and pRCC (Figure 4C). Finally, we interrogated the expression of the same immune checkpoints using single-cell transcriptomic data from CD8+ T-cells derived from ChRCC and ccRCC tumors, with CD8+ T-cells from the adjacent normal kidney tissue in each tumor included as a control.²⁷ We confirmed that there was a significantly lower expression of immune checkpoints specifically in CD8+ T-cells from ChRCC as compared to ccRCC tumors (Figure 4D). Overall, these findings confirm a distinct immune phenotype among CD8+ T-cells in ChRCC, with a marked downregulation of clinically actionable immune checkpoints.

Determination of the T-cell repertoire and antigen specificity in ChRCC

While there was limited T-cell infiltration in ChRCC, determination of whether these T-cells demonstrated anti-tumor specificity could inform directed therapeutic approaches. Using scTCR profiling data from ChRCC and ccRCC samples, we determined the proportion of identified clonotypes in relation to their degree of expansion (i.e., singleton, doublet, and more expanded clonotypes).²⁷ The proportion of unexpanded T-cell clones (“singletons”), typically seen when T-cells do not encounter a cognate antigen, was significantly higher in ChRCC as compared to ccRCC (p=0.05) (Figure 5A). Additionally, a trend towards a lower proportion of expanded clonotypes was observed in ChRCC versus ccRCC (p=0.07) (Figure 5A). Similar results were also observed when comparing ChRCC, LOT and RO, versus ccRCC (Figure S4A). The decreased clonal expansion among T-cells isolated from ChRCC tumors was corroborated by evidence for a higher Shannon’s normalized entropy index (a measure of TCR diversity) in ChRCC compared to ccRCC (Wilcoxon’s p=0.017) (Figure 5B). The same trend was also observed with all oncocytic neoplasms compared to ccRCC (Figures S4A and S4C). These findings suggest a low degree of clonal expansion in ChRCC-infiltrating T-cells, with a high prevalence of unexpanded “singletons” across analyzed samples, overall suggestive of low tumor specificity.

Figure 5: Analysis of T-cell single-cell transcriptomic and TCR profiling data reveals a poor expansion and a low tumor-specificity of T-cell clonotypes in ChRCC.

(A) Proportion of singletons, doubletons, and more expanded clonotypes inferred from single-cell TCR profiling data in ChRCC versus ccRCC (two-sided Wilcoxon signed-rank test).

(B) Shannon normalized index from single-cell TCR profiling data in ChRCC versus ccRCC (two-sided Wilcoxon signed-rank test).

(C) UMAP of T-cell populations identified in renal oncocytic neoplasms.

(D) UMAP of T-cells across tumor samples (upper panel) and normal samples (lower panel).

(E) UMAP showing the mean scaled expression of the tumor-specific (upper panel) and viral-specific (lower panel) signatures across identified T-cells across tumor and normal samples.

(F) Heatmap of T-cell canonical and functional marker expression in individual T-cell populations. Expression values are scaled between minimum and maximum for each gene across all clusters.

(G) and (H), Expression of tumor-specific (G) and viral-specific (H) signatures using single-cell transcriptomic data of T-cells from ccRCC and renal oncocytic neoplasms.²⁷

Abbreviations: ccRCC: clear-cell renal cell carcinoma; ChRCC: chromophobe renal cell carcinoma; Foll. Help. T-Cell: follicular helper T-cell; LOT: low-grade oncocytic tumor; NK-like T-Cell: natural killer-like T-cell; RO: renal oncocytoma; UMAP: uniform manifold approximation and projection.

*: p-value < 0.05; ***: p-value < 0.001

To further evaluate the adaptive immune compartment in ChRCC and interrogate its tumor specificity, we next performed an analysis of the isolated T-cell compartment in ChRCC and oncocytic tumors (n=12,688 cells) using scRNA-seq, with a granular annotation of T-cell subtypes (Figures 5C and S4B). We identified a broad representation of different T-cell populations across oncocytic tumors and normal adjacent tissue samples (Figure 5C, 5D and S5A), confirmed through single-cell expression of canonical marker genes (Figure 5F). We then evaluated the expression of transcriptomic signatures for tumor- and viral-specificity of T-cells, previously validated based on T-cell antigenic specificities.^{52, 53} T-cells isolated from ChRCC and other renal oncocytic neoplasms had a markedly lower tumor-specific signature expression, contrasting with a higher expression of the viral-specific signature (Figure 5E).

We then focused on CD8+ T-cells separately and assessed signature expression in each renal oncocytic tumor and ccRCC. In comparison to ccRCC, ChRCC CD8+ T-cells exhibited significantly decreased expression of the tumor-specific signature (Wilcoxon p<0.001) (Figure 5G) and increased expression of the viral-specific signature (Wilcoxon p<0.001) (Figure 5H). We also investigated the expression of other validated transcriptomic signatures for CD8+ and CD4+ T-cell tumor specificity (neoTCR8 and neoTCR4 signatures, respectively), among T-cells in our cohort, identifying a similarly lower expression in ChRCC, as compared to ccRCC (Figures S5B, S5C, S5E and S5F). Finally, we performed a sensitivity analysis for the tumor-specific and neoTCR8 signatures where the expression of inhibitory immune checkpoints was not included (Figures S5D, S5G and S5H), confirming the robustness of these findings. Overall, these results show that tumor-associated CD8+ and CD4+ T-cells in renal oncocytic neoplasms lack tumor specificity, supporting the model that the T-cell infiltrate in these tumors acts as a bystander.

Clinical outcomes in patients with advanced ChRCC treated with immune-based regimens and other systemic anti-neoplastic therapies

Our analyses suggest that ChRCC largely immune excluded and that the limited T-cell infiltrates are not tumor-specific, which may provide the basis for the immunoresistant phenotype observed clinically. Using real-world data from the International metastatic RCC Database Consortium (IMDC), we identified a total of 229 patients with ChRCC, of which 31 were treated with standard of care ICI-based regimens (either dual ICI therapy, n = 12 or ICI + VEGF-targeted therapies [VEGF-TT], n = 19) in the first-line setting (Table S1), representing the largest cohort of patients with ChRCC to be described so far. A total of 8,931 patients with ccRCC were identified, of which 856 received ICI-based regimens (n = 503 and n = 353 for dual ICI and ICI + VEGF-TT, respectively) as a frontline therapy (Table S1).

With a median follow-up of 58.6 months, patients with metastatic ChRCC treated with ICI-based regimens exhibited a median overall survival of 24.7 (95% confidence interval [CI]: 16.0 – not reached [NR]) months, contrasting with 50.5 (95%CI: 42.5–67.4) months in patients with metastatic ccRCC (mccRCC; adjusted hazard ratio [aHR]: 2.80 [95%CI: 1.51–5.18]). Poorer survival outcomes were also observed for time-to-treatment-failure (TTF) among patients with ChRCC, with a median of 4.5 (95%: 2.4–16.0) months, as compared to 11.0 (95%CI: 9.8–13.6) months in patients with ccRCC (aHR: 2.23 [95%CI: 1.43–3.48]) (Figures 6A, 6B, S6A, S6B). Similarly, the objective response rate (ORR) among patients with metastatic ChRCC (mChRCC) treated with first-line ICI-based regimens was 12.0%, contrasting with an ORR of 47.1% among patients with mccRCC (aOR: 10.26 [95%CI: 2.95–64.84]) (Figure 6G). Notably, no complete response (CR) per RECIST criteria was observed among patients with mChRCC treated with frontline ICI-based regimens, as compared to 4.9% in patients with mccRCC.

Figure 6: Clinical outcomes of patients with metastatic ChRCC and ccRCC across first-line regimens based on data from the International Metastatic RCC Database Consortium (IMDC).

(A) Time to treatment failure (TTF) for patients with metastatic ChRCC versus ccRCC treated with first-line IO-based regimens.

(B) Overall survival (OS) for patients with metastatic ChRCC versus ccRCC treated with first-line IO-based regimens.

(C) TTF for patients with metastatic ChRCC versus ccRCC treated with first-line VEGF-TT.

(D) OS for patients with metastatic ChRCC versus ccRCC treated with first-line VEGF-TT.

(E) TTF for patients with metastatic ChRCC versus ccRCC treated with first-line mTOR inhibitors.

(F) OS for patients with metastatic ChRCC versus ccRCC treated with first-line mTOR inhibitors.

(G) Overall response (inner circle) and response type (outer circle) among patients with metastatic ChRCC versus ccRCC treated with IO-based regimens (upper panel), VEGF-TT (middle panel), or mTOR inhibitors (lower panel).

Abbreviations: ccRCC: clear-cell renal cell carcinoma; ChRCC: chromophobe renal cell carcinoma; CI: confidence interval; CR: complete response; IO Reg.: IO-based regimens; mTOR Inh.: mTOR inhibitor; OS: overall survival; PD: progressive disease; PR: partial response; SD: stable disease; TTF: time-to-treatment-failure; VEGF-TT: vascular endothelial growth factor targeted therapies.

No major differences in survival outcomes were observed among patients with mChRCC and mccRCC treated with first-line VEGF-TT, with a median OS of 23.1 (95%: 19.1–35.6) months for mChRCC versus 26.4 (95%CI: 25.5–27.7) in mccRCC (aHR: 1.25 [95%CI: 0.98–1.59]). Similar outcomes were also observed for TTF, which reached 7.3 (95%CI: 5.1–8.7) months in patients with mChRCC as compared to 8.3 (95%CI: 8.0–8.4) months in those with mccRCC (aHR: 1.25 [95%CI: 1.01–1.56]) (Figures 6C, 6D, S6C, S6D). Thus, patients with mChRCC appeared to have a selectively poor response to ICI-based therapies compared to mccRCC.

Importantly, our analysis of clinical outcomes among patients treated with first-line mTORi (i.e., everolimus or temserolimus) showed a higher OS among patients with mChRCC as compared to those with mccRCC (median OS: 41.3 [95%CI: 14.4-NR] versus 13.4 [95%CI: 10.9–15.3] months, respectively; aHR: 0.77 [95%CI: 0.48–1.25]). Similar results were also observed in relation to TTF with a median of 7.84 (95%CI: 5.29–16.6) months in patients with mChRCC as compared to 3.45 (95%CI: 2.99–3.98) months in those with mccRCC (aHR: 0.52 [95%CI: 0.33–0.82]) (Figures 6E, 6F, S6E, S6F).

Finally, comparison of survival outcomes among patients with mChRCC treated with different first-line regimens did not show any significant difference between patients treated with ICI-based regimens, VEGF-TT or mTORi for TTF (median TTF: 4.5 [95%CI: 2.4–16.0] versus 7.3 [95%CI: 5.1–9.7] versus 7.8 [95%CI: 5.3–16.6] months, respectively [log-rank p=0.3]; aHR [ICI versus VEGF-TT]: 1.35 [95%CI: 0.82–2.20] and aHR [mTORi versus VEGF-TT]: 0.73 [95%CI: 0.45–1.18]) and OS (median OS: 24.7 [95%CI: 16-NR] versus 23.1 [95%CI: 19.1–35.6] versus 43.1 [95%CI: 14.4-NR] months, respectively [log-rank p=0.7]; aHR [ICI versus VEGF-TT]: 0.9 [95%CI: 0.47–1.72] and aHR [mTORi versus VEGF-TT]: 0.87 [95%CI: 0.52–1.46]) (Figures S7A, S7B).

DISCUSSION

To further understand the biology of renal oncocytic tumors and the determinants of host immune responses, we conducted a comprehensive analysis of ChRCC, LOT, and RO using scRNA-seq and scTCR-seq approaches.^{26, 27} Building on prior work, we provide robust evidence that ICA cells of the collecting duct are the common cell-of-origin for ChRCC, LOT, and RO.^{6, 7, 34, 35}

By establishing a distinct cellular origin, we were able to use ICA cells as a transcriptional reference to provide a more granular understanding of the tumorigenesis in ChRCC. We identified a significant upregulation of genes like NUPR1, FTL, and FTH1 in ChRCC, which are associated with ferroptosis resistance, and downregulation of HSPA (HSP70) family genes, such as HSPA5, HSPA6, HSPA8, which play a central role in ferroptosis suppression.^{40–42, 47–49} Ferroptosis is a form of regulated cell death characterized by the production of iron-dependent reactive oxygen species, and its dysregulation is believed to contribute to tumorigenesis.^{54, 55} One method of inducing ferroptosis is by limiting the supply of glutathione. ChRCC cells rely on a cystine/glutamate antiporter to maintain exceptionally high levels of glutathione, up to 100 fold higher than matched normal kidney, making these tumor cells extremely sensitive to ferroptosis upon cysteine depletion.^{56, 57 58}These findings point to ferroptosis induction as a promising target for treating ChRCC tumor resistance and is already an active area of drug development in oncology.⁵⁸ In this setting, ferroptosis inducers (e.g. erastin, sulfasalazine, RAS-selective lethal 3 [RSL3]), which target the SLC7A11-GSH-GPX4 axis, have shown promising activity in pre-clinical studies.^{59, 60} More recently, proteolysis-targeting chimeras (PROTACs) and nanomaterials have also been explored in relation to ferroptosis induction.⁶¹

Beyond ferroptosis, genes involved in mTORC1 signaling were found to be upregulated in ChRCC, which may be related to the occurrence of ChRCC and oncocytomas in individuals with BHD and TSC.^62–65 Our analyses demonstrated that ChRCC have a better response to mTOR inhibition than ccRCC, although in most cases these responses are not sustained. Taken together, these data suggest that combinatorial therapeutic approaches that include mTOR inhibitors may have clinical benefit in ChRCC.

Previous studies have demonstrated that intraepithelial type 1 innate lymphoid cells (ILC1s) play a key role in restraining ChRCC progression, with IL-15 expression being a key determinant of ILC1s cytotoxic activities.³⁰ These data support IL-15 therapy as a potential ChRCC-specific treatment via ILC1-mediated tumor suppression.

We conducted numerous analyses designed to characterize the tumor-infiltrating immune cells, and identified a substantially lower density of CD45+ immune cells in ChRCC versus ccRCC, along with a markedly lower proportion of CD8+ and CD4+ T-cells, and a poor expression of targetable immune checkpoints among CD8+ T-cells, suggesting a poor response to immunotherapy. Additionally, we found that all three classic MHC class I genes (i.e., HLA-A, HLA-B and HLA-C) and antigen presentation and processing pathways are downregulated in ChRCC as compared to its cellular origin, helping to further explain the poor response to ICI-based regimens observed in patients with metastatic ChRCC.^{45, 46, 66–68}

Through scTCR-seq analysis, we found that the majority of T-cells in ChRCC were non-expanded, in contrast to ccRCC. Previously, baseline T-cell clonality and clone size have been shown to be associated with increased cytotoxicity and response to ICI-based regimens.^69–71 Furthermore, CD8+ T-cells in ChRCC showed a low expression of tumor-specific signatures and a high expression of viral-specific signatures.^{52, 53} Taken together, this suggests that the T-cells in ChRCC lack adequate expansion, a clinically targetable phenotype, and tumor-specificity to be effectively targeted with ICI-based therapies.

Using the IMDC database, we demonstrate that patients with metastatic ChRCC demonstrated substantially poorer survival outcomes in a selective manner when treated with ICI-based regimens, compared to patients with metastatic ccRCC, suggesting that the distinct tumor immunobiology of ChRCC contributes to its resistance to immune-based therapies.

This work has several limitations. While our study highlights the limited immune cell infiltration and the non-exhausted phenotype of T-cells in ChRCC, a limitation is the loss of spatial orientation of the tumor microenvironment as a result of droplet-based scRNA-seq analysis. Furthermore, despite characterizing three rare tumor types in this study, the sample size remains limited due to the low incidence of ChRCC and other oncocytic tumors, which may impact the generalizability of our results. While the use of different methodologies (i.e., IHC and bulk RNA-seq), external single-cell datasets and clinical exploration, help to strengthen our findings, further validation remains warranted in larger future efforts. In relation to our scTCR-seq analyses, only baseline attributes were captured, preventing further exploration of the dynamic changes that could potentially occur following ICI therapy. From a clinical standpoint, the retrospective nature of the clinical data may limit the strength of our conclusions. Additionally, selection bias might have influenced some of the clinical analyses, despite the use of multivariate models to adjust for confounding factors. This is, for instance, a possibility for the subgroup of patients with metastatic ccRCC who received mTOR inhibitors in the first-line setting, and who may have presented with poor prognostic features.

Our study confirms the cellular origin of ChRCC and identified several differentially expressed pathways, supporting the investigation of new targets for ChRCC therapy, including ferroptosis, mTORC1 signaling, and IL-15 signaling. Our work also provides the first in-depth look at the immune characteristics of ChRCC and renal oncocytic tumors, identifying key areas of immune dysfunction that provide a mechanistic basis for the poor responses to ICI therapies observed in ChRCC. By identifying these key axes of immune dysfunction, these data provide a foundation for the rational design of future immunotherapy strategies for ChRCC, including increasing tumor cell antigen presentation and expanding the repertoire of tumor-specific CD8+ T-cells that infiltrate the tumor microenvironment.

Context Summary.

Key Objective:

To determine the cellular origin, associated oncogenic pathways, and tumor-immune microenvironmental determinants of anti-tumor immunity for chromophobe renal cell carcinoma (ChRCC) and renal oncocytic neoplasms

Knowledge Generated:

ChRCC and renal oncocytic neoplasms appear to originate from α-intercalated cells of the normal kidney, and multiple features of the tumor-immune microenvironment suggest a poor response to immune checkpoint inhibition, including low immune infiltration, poor immune checkpoint expression on T-cells, and limited tumor specificity. Clinical characterization of patients with advanced ChRCC showed poor survival outcomes with immunotherapy-based regimens, compared to clear cell RCC.

Relevance (Necchi):

This comprehensive analysis of the biology of renal oncocytic tumors and the determinants of host immune responses provides an important focus on the rationale of currently available results and pinpoints suitable next avenues for clinical trials in these rare tumor entities.