26 Applying single cell RNA sequencing of sarcomatoid renal cell carcinoma for the development of a novel transcriptomic biomarker to predict immunotherapy response

Abstract Background Renal cell carcinoma with sarcomatoid features (sRCC) is a unique kidney cancer subtype associated with aggressive biology and poor clinical outcomes. Intriguingly, sRCC has recently exhibited preferential responsiveness to immune checkpoint blockade (ICB) therapies. Within the RCC population, however, there currently is a paucity of biomarkers predictive for ICB response, representing a critical unmet clinical need for optimal therapy selection. Therefore, we sought to derive a transcriptomic signature from sRCC samples encapsulating this paradoxical hyper-aggression and ICB-responsiveness to identify RCC patients most likely to benefit from ICB irrespective of sarcomatoid feature presence. Methods Nephrectomy specimens from patients with RCC were processed for single cell RNA sequencing (scRNAseq). Clustering was performed and annotated tumor cells were computationally isolated. Tumor cell counts were aggregated and differential expression between sRCC and non-sRCC cases was performed. Genes significantly upregulated in sRCC were next filtered against differential gene expression data of sRCC vs non-sRCC from three clinical RCC datasets (TCGA, CheckMate, 8 clear cell RCC) were analyzed by scRNAseq, including 5,386 tumor cells. Differential gene expression of tumor clones revealed 20 genes significantly enriched in sRCC relative to ccRCC. Filtering against public differential gene expression data resulted in 10 genes included within the final Sarcomatoid Signature (SS). SS expression scores for clinical specimens were calculated by single-sample gene set enrichment analysis. SS scores were significantly enriched in sRCC patient tumors across TCGA KIRC, CheckMate, and IMmotion151 datasets (p0.001 for all cohorts). Within TCGA KIRC, SS scores were significantly increased in patients with nuclear grade 4 (p0.001) and metastatic (p0.001) disease, while stratification by median SS score revealed worsened overall (HR=2.19, p0.001) and disease-free (HR=2.08, p0.001) survival among SS-high patients. Amongst the CheckMate cohort (including patient samples from the CheckMate-010 0.001) irrespective of treatment arm. However, SS-high patients experienced improved objective response rates with the ICB-containing atezolizumab/bevacizumab regimen compared to sunitinib (p0.001) and relative to SS-low patients receiving atezolizumab/bevacizumab (p=0.028) (Figure 1). Within the SS-high population, patients experienced prolonged PFS with atezolizumab/bevacizumab (HR=0.70, p=0.003) relative to sunitinib whereas no PFS difference between arms was seen in the SS-low population (HR=1.01, p0.99). Conclusions This novel SS derived from scRNAseq demonstrates negative prognostic yet positive predictive value for ICB response. In a domain currently void of efficacious biomarkers, the SS, upon prospective validation, may assist in optimal therapy selection for patients with RCC.