PD01-12: Pan-Cancer Characterization of the Tumor Microenvironment Reveals Stroma and Immune Enriched Subtype with distinct molecular, clinical, and prognostic features in different genitourinary cancers

Mark Farha*, Brittney Cotta, Srinivas Nallandhighal, Kevin Hu, Ann Arbor, MI, Steven Goldenthal, Columbus, OH, Aaron Udager, Simpa Salami, Ann Arbor, MI

Introduction: Transcriptomic data driven tumor microenvironment (TME) deconvolution tools have enabled robust analyses correlating TME subtypes with prognosis and therapeutic response. Here, we performed a pan-cancer analysis of tumors included in the TCGA to assess the impact of the TME on oncological outcome.

Methods: Data was downloaded from the TIMEx portal. TIMEx utilizes scRNA-seq signatures from 16 solid cancers to classify 37 cell types and activation states, including malignant, benign tissue, immune, and stromal cells. Unsupervised hierarchical clustering was performed to divide the TIMEx deconvoluted TCGA data (31 cancers, n=9848) into three clusters. Prevalence of each cancer type was calculated for each of the clusters. Progression free survival (PFS) was assessed by cluster for the entire cohort and for genitourinary (GU) cancers in a sub-analysis – Bladder (BLCA), Prostate (PRAD), Kidney [Clear Cell (KIRC), Papillary (KIRP), and Chromophobe (KICH)]. Tumor infiltrating lymphocyte (TIL) architecture data was collected from Saltz et al. (2018) in which “computational staining” was performed on the TCGA pathology slide archive.

Results: Three clusters with similar fractions of malignant and native tissue cells, but different immune and stromal enrichment were identified: CL1ImmuneLow/StromaLow, CL2ImmuneLow/StromaMod, and CL3ImmuneHi/StromaHi. While BLCA was heterogeneous, the majority of PRAD tumors were in CL1 (65%, p<2.2x10-16). The majority of KIRC tumors were in CL3 (80%, p<2.2x10-16), while KICH were mostly made up of CL1 (52%, p<2.2x10-16), and most KIRP were in CL2 (62%, p=2.2x10-13). Overall across cancers, CL3 had a longer PFS compared to grouped CL1 and CL2 combined (100.5 mo (95% CI 83.9-NR) vs. 85.7 mo (95% CI 74.4-101 mo), p=0.016). In a survival analysis of GU cancers, CL3 was associated with a longer PFS in BLCA (p=0.03) but shorter PFS in KIRC (p=0.05). Notably, BLCA CL3 patients had fewer FGFR mutations (p=0.007), brisk TIL spatial pattern (p=1.5x10-5), and more Basal/Squamous tumors by the Kamoun classifier (p=3.6x10-8) compared to CL1 and CL2. KIRC CL3 patients on the other hand, showed no difference in detection of VHL, PBRM1, SETD2, or TCEB1 alterations compared to CL1 and CL2.

Conclusions: Our analyses reveal the heterogeneous composition and prognostic implications of the TME across various cancers. Further studies are needed to explore the capacity of this classification scheme to guide prognostication and therapeutic decisions.

Source of Funding: A.M.U., S.S.S. are supported by the University of Michigan Health System-Peking University Health Science Center (UMHS-PUHSC) Joint Institute (JI). S.S.S. is supported by the Robert Wood Johnson Foundation as part of the Harold Amos Medical Faculty Development Program (AMFDP). S.S.S. is supported in part by the Urology Care Foundation Rising Stars in Urology Research Award Program and Astellas, Inc. S.S.S. is supported by P30CA046592. A.M.U. is supported by the Department of Defense (W81XWH-19-1-0407) and National Institutes of Health (U01 CA232931and R37 CA222829). A.M.U. and S.S.S. are supported by the National Institutes of Health (P50 CA186786).