Association between vascular patterns of clear cell renal cell carcinoma and immune cell infiltration and therapeutic response.

Abstract

575 Background: Clear cell renal cell carcinoma (ccRCC) is a highly vascularized tumor with histological heterogenous appearance. We explored whether vascular pattern (VP) are predictive to response of therapy. Methods: We defined three categories of VP in the publicly available TCGA cohort: high-branching (HB), low-branching (LB) and sinusoid. VP-based gene signatures were generated by integrating the transcriptomes of matched ccRCC samples. We developed a learning-based (AI) algorithm for the classification of VP based on CD31 immunohistochemistry and analyzed histology specimens from patients treated with tyrosine kinase inhibitors +/- immunotherapy (TKI) (University Clinic Dresden, retrospective cohort n=38, and NivoSwitch trial NCT: NCT03013946 , n=38 respectively) as well as the publicly available transcriptome datasets from two phase III clinical trials (JAVELIN Renal 101; IMmotion 151). Multiplex immunofluorescence (CODEX) was used for spatial mapping. Outcome measures employed KM-plots and log-rank analyses. Results: We identified a trajectory from a HB to LB vascular phenotype, paralleled by a decline in the expression of proximal tubule cell lineage traits. Applying the VP gene signatures to the transcriptome datasets from JR101 and IM151 we found that the progression-free survival (PFS) benefit from adding immunotherapy to anti-angiogenic therapy (IMmotion151: Bevacizumab; JAVELIN Renal 101: Axitinib) was limited to low-branching ccRCC (IM151: HR (95% CI): 0.64 (0.51-0.80), JR101: HR (95% CI): 0.45 (0.33-0.61, both p < 0.001), respectively) and linked to an immune cell infiltrated microenvironment. Patients with a HB ccRCC demonstrated prolonged PFS in the Dresden TKI cohort (p = 0.026) and in the NivoSwitch trial (p = 0.01). CODEX analysis using the SPACEc pipeline revealed profound differences in cellular neighborhoods, characterized by dense immune cell infiltration of ccRCC with LB compared to HB vascular patterns. Using 30 ccRCC patient-derived organoid in air-liquid interface (ALI) cultures, we confirmed the association between LB pattern and higher T cell infiltration, resulting in reduced viability of LB ccRCC organoids under immune-stimulating conditions. Conclusions: Vascular pattern can reliably predict therapy response in advanced ccRCC. Tumors with high branching phenotype respond better to anti-angiogenic TKI therapy. Low branching pattern was associated with improved response to immunotherapy.