Skip to main page content
U.S. flag

An official website of the United States government

Dot gov

The .gov means it’s official.
Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

Https

The site is secure.
The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2021 May 27;184(11):2988-3005.e16.
doi: 10.1016/j.cell.2021.04.038. Epub 2021 May 20.

Single-cell protein activity analysis identifies recurrence-associated renal tumor macrophages

Aleksandar Obradovic  1 , Nivedita Chowdhury  2 , Scott M Haake  3 , Casey Ager  4 , Vinson Wang  5 , Lukas Vlahos  6 , Xinzheng V Guo  4 , David H Aggen  4 , W Kimryn Rathmell  7 , Eric Jonasch  7 , Joyce E Johnson  7 , Marc Roth  7 , Kathryn E Beckermann  7 , Brian I Rini  7 , James McKiernan  8 , Andrea Califano  9 , Charles G Drake  10
Affiliations

Single-cell protein activity analysis identifies recurrence-associated renal tumor macrophages

Aleksandar Obradovic et al. Cell. .

Abstract

Clear cell renal carcinoma (ccRCC) is a heterogeneous disease with a variable post-surgical course. To assemble a comprehensive ccRCC tumor microenvironment (TME) atlas, we performed single-cell RNA sequencing (scRNA-seq) of hematopoietic and non-hematopoietic subpopulations from tumor and tumor-adjacent tissue of treatment-naive ccRCC resections. We leveraged the VIPER algorithm to quantitate single-cell protein activity and validated this approach by comparison to flow cytometry. The analysis identified key TME subpopulations, as well as their master regulators and candidate cell-cell interactions, revealing clinically relevant populations, undetectable by gene-expression analysis. Specifically, we uncovered a tumor-specific macrophage subpopulation characterized by upregulation of TREM2/APOE/C1Q, validated by spatially resolved, quantitative multispectral immunofluorescence. In a large clinical validation cohort, these markers were significantly enriched in tumors from patients who recurred following surgery. The study thus identifies TREM2/APOE/C1Q-positive macrophage infiltration as a potential prognostic biomarker for ccRCC recurrence, as well as a candidate therapeutic target.

Keywords: CD8 T cell; Treg; clear cell renal carcinoma; clustering; gene regulatory networks; immunotherapy; post-surgical recurrence; protein activity inference; single-cell RNA sequencing; tumor microenvironment; tumor-infiltrating macrophage.

PubMed Disclaimer

Conflict of interest statement

Declaration of interests C.G.D. is a co-inventor on patents licensed from JHU to BMS and Janssen; has served as a paid consultant to AZ Medimmune, BMS, Pfizer, Roche, Sanofi Aventis, Genentech, Merck, and Janssen; and has received sponsored research funding to his institution from BMS IIoN and Janssen. A.C. is founder, equity holder, consultant, and director of DarwinHealth Inc., which has licensed IP related to these algorithms from Columbia University. Columbia University is an equity holder in DarwinHealth Inc. B.I.R. has served as a paid consultant to BMS, Pfizer, GNE/Roche, Aveo, Synthorx, Compugen, Merck, Corvus, Surface Oncology, 3DMedicines, Arravive, Alkermes, Arrowhead, GSK, and Shionogi and holds stock in PTC therapeutics.

Figures

Figure 1:
Figure 1:. Deep Profiling of CD45+ Microenvironment by Gene Expression and Protein Activity Reveals Tumor-Specific Immune Populations
1A) UMAP Plots for single-cell gene expression pooled across CD45+ samples, clusters visualized and labelled by cell type. Bottom plot is split by Tumor vs Adjacent Normal label 1B) UMAP Plots for VIPER-Inferred protein activity pooled across CD45+ samples. Bottom plot is split by Tumor vs Adjacent Normal label 1C) Heatmap of top5 upregulated genes for each cluster by expression; each row represents a gene and each column represents a cell. Legend shows cluster identity with cell type inferred by SingleR and Tumor (red) or Adjacent Normal (blue) tissue source. 1D) Heatmap of top5 differentially upregulated proteins for each cluster by VIPER-inferred activity. Legend as in 1C. 1E) Bar plots of patient-by-patient cluster frequency in Tumor minus frequency in Adjacent Normal for each Gene Expression cluster, grouped by stage; values < 0 (blue) indicate higher frequency in Adjacent Normal and values > 0 (red) indicate higher frequency in Tumor. 1F) Bar plots of patient-by-patient cluster frequency in Tumor minus frequency in Adjacent Normal for each VIPER cluster, grouped by stage, as in 1E.
Figure 2:
Figure 2:. Known and Novel Tumor-Infiltrating Immune Population Markers Discovered from Single-Cell Transcriptomic and Inferred Proteomic Data
2A) Violin plots of VIPER-inferred Proteins upregulated in CD45+ cell subsets corresponding to Tregs (FOXP3, CTLA4), Exhausted CD8s (TOX2, LAG3, PD1, CTLA4), and Tumor-specific Macrophages (LILRB5, APOE). 2B) Violin plots of top transcriptional markers (C1Q, APOE, TREM2) specifically up-regulated in Tumor-Infiltrating Macrophages as compared to other cell populations as well as non-tumor macrophages.
Figure 3:
Figure 3:. Flow Cytometry is Better Recapitulated by Protein Activity Than by Gene Expression
3A) Representative Flow Cytometry Gating in Tumor and Adjacent Normal and frequency plots in tumor and adjacent normal for all manually gated populations. Populations of PD1+CD39+ exhausted CD8 cells, Tregs, and CD11B+CD163+ Macrophages are of higher frequency in Tumor than Adjacent Normal. Representative plots showing two distinct NK cell phenotypes and three monocytic sub-phenotypes, consistent with Figure 1B. 3B) UMAP projection, clustering, and heatmap by flow cytometry proteins profiled in CyTEK Lymphoid Panel. 3C) UMAP and clustering by scRNASeq gene expression subset to the proteins profiled in 3B, showing noise-induced decrease in clustering resolution. 3D) UMAP and clustering by scRNASeq VIPER inference subset to the proteins profiled in 3B. 3E) UMAP and clustering by flow cytometry proteins profiled in CyTEK myeloid panel. 3F) UMAP and clustering by scRNA-Seq gene expression, subset to the proteins profiled in 3E. 3G) UMAP and clustering by scRNA-Seq VIPER inference, subset to the proteins profiled in 3E.
Figure 4:
Figure 4:. Deep Profiling of CD45- Cells by Gene Expression and Protein Activity Distinguishes Tumor Cells From Normal Epithelium
4A) UMAP of single-cell gene expression pooled across all CD45- samples, clusters labelled by cell type. Bottom plot is split by Tumor vs Adjacent Normal label 4B) UMAP of VIPER-inferred protein activity pooled across all CD45- samples, clusters labelled by cell type. Bottom plot is split by Tumor vs Adjacent Normal. 4C) Heatmap of top5 differentially upregulated genes for each cluster by expression; each row represents a gene and each column represents a cell. Legend shows cluster identity with cell type inferred by SingleR and Tumor (red) or Adjacent Normal (blue). 4D) Heatmap of top5 differentially upregulated proteins for each cluster by VIPER-inferred activity. Legend as in 4C. 4E) Bar plots of patient-by-patient cluster frequency in Tumor minus frequency in Adjacent Normal for each Gene Expression cluster, grouped by stage; values < 0 (blue) indicate higher frequency in Adjacent Normal, values greater < 0 (red) indicate higher frequency in Tumor. 4F) Bar plots of patient-by-patient cluster frequency in Tumor minus frequency in Adjacent Normal for each VIPER cluster, grouped by stage, as in 4E.
Figure 5:
Figure 5:. Tumor Cell Labeling is Validated by Copy Number Inference and Tumor Marker Expression
5A) ViolinPlots of VIPER-Inferred Activity for ccRCC tumor markers PAX2, PAX8, and CA9. Plots grouped by CD45- cluster label revealing increased expression in epithelial cells. 5B) CNA Inference for all CD45- populations, using CD45+ cells as reference. Columns represent chromosomal regions and rows represent cells, grouped by Gene Expression cluster, with a subset of copy-number-normal epithelial cells highlighted in green. 5C) CNA Inference re-grouped by VIPER cluster. Epithelial cell clusters 1, 3 and 4 contain consistent chromosome 3p deletions characteristic of ccRCC, while Epithelial cluster 2, highlighted in green, is grossly Copy-Number normal. 5D) Table of known receptor-ligand interaction pairs in which ligand is significantly upregulated by Gene Expression in one cluster and receptor is significantly upregulated by VIPER in another. Subset to interactions inferred between Tumor cells and T-cells, or between APOE+/TREM2+/C1Q+ Tumor Macrophages and Tumor cells. 5E) Visualization of receptor-ligand interaction pairs shown in 5D.
Figure 6:
Figure 6:. Enrichment of Tumor-Specific Macrophage Markers Defined from Single-cell RNASeq in Bulk RNASeq Data is Associated with Shorter Time-to-Recurrence
6A) Gene Set Enrichment Analysis (GSEA) of tumor-specific macrophage marker proteins in VIPER-transformed bulkRNASeq data from 4 patients with post-surgical recurrence vs 4 without. Proteins ranked by fold-change in recurrence versus no recurrence, p-value computed by GSEA vs gene shuffling null model with 1000 permutations. Note enrichment in patients with recurrence (NES=4.08, p=4.5*10−5). 6B) Kaplan-Meier curve of sample-by-sample tumor-specific macrophage GSEA associated with time to recurrence, yellow line indicates patients with low enrichment, blue line indicates patients with high enrichment. Log-rank p-value = 0.0067. 6C) Heatmap of leading-edge protein set from 6A. 6D) Sample-by-sample tumor macrophage GSEA, annotated with each sample’s recurrence status and time to recurrence or total observation time. Proteins ranked by inferred activity. 6E) Macrophage signature GSEA in recurrence vs. no recurrence in validation cohort (N=157). 6F) Kaplan-Meier curve of sample-by-sample GSEA in association with time to recurrence in the validation cohort, log-rank p-value = 0.0029.
Figure 7:
Figure 7:. A Novel Population of C1Q/TREM2+ Macrophages are Tumor-Specific and Associated with shorter time-to-recurrence by Immunohistochemistry (IHC)
7A) Representative IHC images for each marker in tumor stroma vs adjacent normal. Note high C1Q/TREM2/APOE staining within CA9+ tumor as compared to tumor-adjacent (CA9-) regions. 7B) Odds ratios (OR) across samples of tumor-specific macrophage markers C1Q, TREM2, and APOE co-staining with CD68/CD163+ macrophage cells vs CD68/CD163- non-macrophage cells, note association of C1Q and TREM2 with macrophage markers. Dotted red line represents OR=1. Individual OR for C1Q and TREM2 co-staining with CD68/CD163 is statistically significant by Fisher’s exact test (p<0.01). 7C) Frequency by IHC of C1Q+ or TREM2+ macrophages in tumor stroma vs adjacent normal across the 11 patient samples profiled by scRNASeq. Enrichment in tumor compared to adjacent normal assessed by paired Wilcox test, *<0.05, **<0.01. 7D) Frequency of C1Q+TREM2+CD68/CD163+ macrophages in tumor vs adjacent normal, plotted by stage (pT1a vs pT3b). No C1Q+TREM2+CD68/CD163+ cells were present in adjacent normal. 7E) Frequency of C1Q+ or TREM2+ macrophages in tumor stroma of patients with or without post-surgical recurrence, from the cohort profiled by bulkRNASeq in Figure 6A–6D. Higher frequency in patients with recurrence assessed by unpaired Wilcox test, *<0.05. 7F) Kaplan-Meier plot of C1Q+CD68/CD163+ frequency in association with time to recurrence. Log-rank p-value = 0.0067, with sample-by-sample frequency binarized by log-rank maximization to >0.01 = “high” and <0.01 = “low.”
See this image and copyright information in PMC

Similar articles

See all similar articles

Cited by

See all "Cited by" articles

References

    1. Alvarez MJ, and Califano A. (2018). Darwin OncoTarget/OncoTreat: NY CLIA certified tests to identify effective drugs on an individual cancer patient basis from RNASeq profiles (Dpt of Pathology and Cell Biology Web Site, Columbia University).
    1. Alvarez MJ, Shen Y, Giorgi FM, Lachmann A, Ding BB, Ye BH, and Califano A. (2016). Functional characterization of somatic mutations in cancer using network-based inference of protein activity. Nat Genet 48, 838–847. - PMC - PubMed
    1. Alvarez MJ, Subramaniam PS, Tang LH, Grunn A, Aburi M, Rieckhof G, Komissarova EV, Hagan EA, Bodei L, Clemons PA, et al. (2018). A precision oncology approach to the pharmacological targeting of mechanistic dependencies in neuroendocrine tumors. Nat Genet 50, 979–989. - PMC - PubMed
    1. Aran D, Looney AP, Liu L, Wu E, Fong V, Hsu A, Chak S, Naikawadi RP, Wolters PJ, Abate AR, Butte AJ, Bhattacharya M (2019). “Reference-based analysis of lung single-cell sequencing reveals a transitional profibrotic macrophage.” Nat. Immunol, 20, 163–172. - PMC - PubMed
    1. Arik D, Can C, Dundar E, Kabukcuoglu S, and Pasaoglu O. (2017). Prognostic Significance of CD24 in Clear Cell Renal Cell Carcinoma. Pathol Oncol Res 23, 409–416. - PubMed

Publication types

MeSH terms