Distinct molecular profiles and shared drug vulnerabilities in pancreatic metastases of renal cell carcinoma

Abstract

Clear-cell renal cell carcinoma (ccRCC) is the most common origin of pancreatic metastases (PM). Distinct genomic aberrations, favorable prognosis, and clinical observations on high angiogenesis, and succeeding tyrosine kinase inhibitor (TKI) sensitivity have been reported in PM-ccRCC. However, no functional or single-cell studies have been conducted thus far. We recruited five PM-ccRCC patients and investigated the genomic, single-cell transcriptomic, and drug sensitivity profiles of their patient-derived cells (PDCs). The PM depicted both expected and novel genomic alterations. Further, the transcriptomics differed from both primary and metastatic ccRCC, with upregulations of the PI3K/mTOR and - supporting the clinical observations - angiogenesis pathways. Data integration at pathway level showed that transcriptomics explained drug sensitivities the best. Accordingly, PM-ccRCC PDCs shared sensitivity to many PI3K/mTOR inhibitors. Altogether, we show distinct genomic and transcriptomic signatures in PM-ccRCC, highlight the superiority of transcriptomics in interpreting drug sensitivities, and encourage the use of TKIs and PI3K/mTOR inhibitors in PM-ccRCC.

Fig. 1. Clinical information, patient history, and flowchart of study protocol.

a Schematic of disease progression in individual patients, anatomical location (right versus left) of the primary tumor, and study protocol. PM-ccRCC samples (n = 5) were processed in the laboratory for cell culture in 2D and 3D, drug sensitivity screening, exome sequencing, scRNA-seq, and multiplexed immunohistochemical (mIHC) analysis of metastatic tissue formalin-fixed paraffin embedded (FFPE) sections. b Computer tomography (CT) scan of the five patients showing pancreatic masses (red arrowheads). Scale bars 5 cm. c Hematoxylin and eosin staining of the PM-ccRCC tumors (n = 5). Scale bars 50 µm.

Fig. 2. CcRCC markers identified in mIHC of the PM-ccRCC are preserved in 2D and 3D PDCs.

a PAX8 expression in mIHC staining of tissue (PAX8 red, DNA blue) and multiplexed IF (mIF) stainings of 2D PDC cultures (PAX8 red, DNA blue), and PAX8 expression in scRNA-seq of the 3D PDCs (tumor cells encircled in pt 5). b CAIX expression in tissue mIHC (CAIX, green;DNA/nuclei, blue) and mIF of 2D PDC cultures (CAIX, green;DNA/nuclei, blue) and CAIX expression in scRNA-seq of the 3D PDCs. Percentage of PAX8 or CAIX-positive cells in scRNA-seq from each 3D PDC culture is indicated (tumor cells encircled in pt 5). Note that besides showing the positive cells, the scRNA-seq plots also indicate the intensity of the PAX8 and CAIX expression (grey to red or green, respectively). For scRNA clusters, please see Fig. 5. White scale bars in mIHC images; 50 µm, and in mIF; 100 µm, yellow scale bars 25 µm.

Fig. 3. All patients exhibit known ccRCC somatic mutations and/or CNAs.

a Somatic mutations and (b) relative CN in the tumor compared to the matched normal of genes i) recurrently mutated in our cohort, ii) harboring functionally and/or clinically relevant mutations and iii) otherwise known to be commonly mutated in ccRCC based on COSMIC and prior literature. Prevalence of mutation as reported in COSMIC is shown within brackets. Mutations occurring in our PM-ccRCC tissue samples with a statistically higher frequency than expected are marked with an asterisk (*). c Patient-specific fraction of each cytoband affected by CN losses (dashed line) and gains (solid line) in chromosomes 3, 9, 14 and 15. In red: previously reported most common metastatic losses are marked with an asterisk (*) and losses enriched in our PM-ccRCC tissue samples with two asterisks (**). d CNAs affecting more than 50% of a given cytoband in more than two of our five PM-ccRCC tumors. Red indicates gains and blue the losses in particular chromosomes; color intensity correlates with the fraction of the cytoband affected.

Fig. 4. The five PM-ccRCC patients showed both individual as well as shared drug vulnerabilities.

a DSS shown for the drugs with sensitivities (DSS ≥ 6 in any PDC model). The arrows indicate the drugs with highest common sensitivities across all patient PDCs; asterisks indicate approved status and red asterisks indicate ESMO recommendations for ccRCC treatment^,. b Dose response curves, DSS, and EC50 for vinorelbine (mitotic inhibitor), omipalisib (mTOR inhibitor), and napabucasin (CSC inhibitor). Blue dots represent observed viability (percentages) from which the dose response curves (blue lines) are plotted. Inh. = inhibitor.

Fig. 5. PM-ccRCC PDCs and primary ccRCC tissue samples share similar tumor cell clusters but differ in expression levels of PM-ccRCC related signature genes as well as in deregulation of HALLMARK pathways.

a UMAP visualization of different cell clusters of PM-ccRCC (3D PDC cultures, pts 1-5) and primary ccRCC tissue samples (pts 6-9) with 15 distinct cell clusters of which 7 were tumor cells, 2 fibroblasts, 1 endothelial and the remaining clusters were immune cells, including T- and NK-, myeloid and dendritic cells. b UMAP and (c) diagram of the top differentially expressed genes (DEGs) between PM- (pts 1-5) and primary (pts 6-9) ccRCC tumor cells. *p < 0.05, **p < 0.001, ***p < 0.0001. Figure b. also shows expression of genes (APOL, SAA1, SAA2 and MET) previously published to be associated with metastatic ccRCC (Alchahin et al.). The expression of these genes is not consistent in the five PM-ccRCC PDCs. d Pathway analysis showing normalized enrichment scores (NES) based on Gene Set Enrichment Analysis (GSEA) of comparing PM-ccRCC tumor cells (n = 5) to primary ccRCC tumor cells (n = 4).

Fig. 6. Patient-specific drug responses are best explained by DE at pathway-level.

a Gene targets of effective drugs (DSS ≥ 6) show varying degrees of patient-specific deregulations (DE in each patient’s tumor vs stromal cells, somatic mutation, or CNA). Overall, 11–50% of patient-specific drug responses can be explained at the gene-level by overlapping alterations (percentages per data type shown below the heatmap). b Pathway-level analysis of patient-specific drug responses and deregulations (DE in respective tumor vs stroma cells, somatic mutation, or CNA) increases the overall explainability of patient-specific drug effectiveness to 25–66%. c In 4/5 patients, DE HALLMARK pathways showed a higher fraction of effective drug (max DSS ≥ 6) targets than those not DE. d The association between pathway-level DE and drug response was statistically significant (one-sided paired Wilcoxon test p = 0.062).

Fig. 7. Pathway-level overview accentuates patient-specific drug responses and deregulations (DE in each patient’s tumor vs stromal cells, somatic mutations, and CNAs).

a The panel illustrates for each patient the detailed multi-omics data of those pathways that harbor three or more genes targeted by our drug library (number of genes targeted shown inside the innermost circle of pt 1). Somatic CN ratios of individual genes (outermost circle), DE log2 fold change in scRNA-seq between each patient’s tumor and stromal cells, the maximum DSS across all genes of a pathway as well as the maximum somatic variant allele frequency (VAF) across all genes of a given pathway (innermost circle) is shown. DE pathways effectively targeted by the DSRT (DSS ≥ 6) are marked (boxes for pt 1, asterisks for pts 2–5). b Heatmap displaying number of approved drugs listed in the TTD that target genes within each of the pathways differentially expressed within each patient. Triangles mark pathways targeted by 0 drugs (red) or 1-2 drugs (blue) in our DSRT drug library.