Abstract PO-024: Tumor-specific cell populations in clear cell renal carcinoma associated with clinical outcome identified using single-cell protein activity inference

Abstract

Primary Clear Cell Renal Carcinoma (ccRCC) is a highly heterogenous disease with a variable disease course post-surgery, and ccRCC tumor micro-environment has thus far not been characterized at the single-cell level with the whole-transcriptome resolution enabled by single-cell RNA Sequencing (scRNASeq). To elucidate the cellular and transcriptional mechanisms driving disease recurrence, we performed scRNASeq on both hematopoietic and non-hematopoietic populations from tumor and tumor-adjacent tissue from primary resections of treatment-naïve ccRCC patients (n=11), thus producing a complete atlas of the ccRCC tumor microenvironment. Furthermore, in order to mitigate gene dropout inherent to scRNASeq and commonly preventing detection of >80% of genes, we leveraged the VIPER algorithm, which quantitates protein activity, to infer the proteomic regulators of cell state. Clustering at the gene expression level enabled construction of lineage-specific gene regulatory networks applying ARACNe, the Algorithm for Reconstruction of Accurate Cellular Networks, from which protein activity of upstream regulatory molecules could be inferred by VIPER. Comparison with protein-level expression data from antibody-based, high-parameter spectral flow cytometry in the same patients shows that VIPER-measured protein activity systematically abrogated gene dropout effects on a repertoire of >6,000 regulatory proteins, comparing favorably with antibody-based measurements. This helped comprehensively characterize the individual cellular sub-populations comprising the ccRCC tumor and peri-tumor microenvironment, as well as their specific master regulators and candidate cell-cell interactions, revealing several populations undetectable by gene expression analysis. Specifically, we uncovered a novel tumor-specific, macrophage subpopulation characterized by significant upregulation of TREM2, APOE, and C1Q, as validated by spatially-resolved, quantitative multispectral immunofluorescence (qmIF). These tumor-specific macrophages were statistically significantly over-represented in a separate validation cohort of patients who recurred following surgery (n=4) compared to patients who did not recur (n=4). This study highlights the substantial resolution increase afforded by VIPER and identifies TREM2/APOE/C1Q-positive macrophage infiltration as a potential biomarker for ccRCC recurrence and a candidate target for intervention. Furthermore, it provides a highly-generalizable methodology to study the role of rare subpopulations by leveraging VIPER on scRNASeq data. Citation Format: Aleksandar Obradovic, Nivedita Showdhury, Casey Ager, Vinson Wang, Lukas Vlahos, Xinzheng V. Guo, David H. Aggen, James McKiernan, Andrea Califano, Charles G. Drake. Tumor-specific cell populations in clear cell renal carcinoma associated with clinical outcome identified using single-cell protein activity inference [abstract]. In: Proceedings of the AACR Virtual Special Conference on Tumor Heterogeneity: From Single Cells to Clinical Impact; 2020 Sep 17-18. Philadelphia (PA): AACR; Cancer Res 2020;80(21 Suppl):Abstract nr PO-024.

BACH2 inhibition reverses β cell failure in type 2 diabetes models

“…To capture T2D-specific cellular states, we used single-cell RNA sequencing (scRNA-Seq) profiles from 4 nondiabetic controls (ND) and 6 T2D donors (Supplemental Table 1; supplemental material available online with this article; https://doi.org/10.1172/JCI153876DS1) to build islet-specific transcription regulatory networks and to identify the transcription factors (TFs) and cofactors (co-TFs) that control their differential gene expression signatures (Figure 1A). To measure differential TF activity in individual cells, we first used activity of a protein is measured from the differential expression of 50 or more of its transcriptional targets, akin to a highly multiplexed gene reporter assay, thus allowing reliable protein activity assessment even from low-depth profiles (18,22).…”

“…Statistically significant superiority of protein activity versus gene-expression-based clustering was recently demonstrated (21), including at the single-cell level (22). Thus, to further cluster islet cells, including both ND and T2D cells, we performed unsupervised hierarchical clustering analysis using iterative clustering (iterClust) algorithms (23) with the full, metaVI-PER-inferred protein activity matrix.…”