Table 1.
Tumors surveyed for DE analysis.
TCGA cancer type and subclassification, if applicable, for solid tumors with distinct histological classification are shown, along with the number of replicates and GPCRs with increased or decreased expression for each type of tumor.
Fig 1.
Heatmaps of GPCR expression and DE in solid tumors.
(A) For all 45 tumor subtypes, a heatmap showing the log2 fold-change of GPCR expression in tumors compared to normal tissue (positive values indicate higher expression in tumors), with hierarchical clustering of GPCR genes to reveal patterns of DE. (B) The number of GPCRs that show significant (FDR < 0.05) changes in expression compared to normal tissue among tumor types tested with large numbers of replicates (Table 1) and that correspond to the most lethal types of cancer. (C–D) The GPCRs that most frequently (i.e., in most tumor types) show increases (C) or decreases (D) in expression among the 45 tumor subtypes. DE data for GPCRs in all analyzed tumor types can be found in S2 Table, sheets 6–8. BRCA, breast cancer; COAD, colon adenocarcinoma; DE, differential expression; FDR, false discovery rate; GPCR, G protein-coupled receptor; Her2, Human Epidermal Growth Factor Receptor-2; IDC, Infiltrating Ductal Carcinoma; LSQC, lung squamous cell carcinoma; LUAD, lung adenocarcinoma; NOS, Not Otherwise Specified; PDAC, pancreatic ductal adenocarcinoma; PRAD, Prostate Adenocarcinoma.
Fig 2.
Solid tumors have large numbers of differentially expressed GPCRs compared to normal tissue.
(A–B) The 30 highest expressed GPCRs in PDAC (A) or primary SKCM (B) and their corresponding expression in normal pancreatic (A) or skin (B) tissue. Expression data in TPM and CPM for all tumor types and normal tissue can be found at https://insellab.github.io/gpcr_tcga_exp. (C–D) The 30 GPCRs with the highest fold-increase in expression in tumors compared to normal tissue for (C) PDAC and (D) primary SKCM, sorted by fold-increase in tumors compared to normal tissue. Data on GPCR DE used in these plots can be found in S2 Table, sheets 6–8. (E–F) For two highly expressed GPCRs in (A–D) as examples, the median expression of (E) GPRC5A and (F) GPR143 in all tumor types tested and corresponding normal tissue, normalized in CPM, allowing for comparison between tissue/tumor types. A lookup file that enables generation of similar plots (as well as upper and lower quartiles of expression) for any GPCR can be found at https://insellab.github.io/gpcr_tcga_exp. Plots for GPCR expression can be generated using the spreadsheet for visualization of expression provided at https://insellab.github.io/gpcr_tcga_exp. CPM, Counts Per Million; DE, differential expression; GPCR, G protein-coupled receptor; PDAC, pancreatic ductal adenocarcinoma; SKCM, skin cutaneous melanoma; TPM, Transcripts Per Million.
Fig 3.
Increased expression of certain GPCRs in PDAC is more frequent than that of KRAS and TP53 mutations; is independent of tumor grade, pathological T, and patient sex; and includes GPCRs that couple to each of the 4 major G protein families.
(A) The expression of GPRC5A in all PDAC samples and normal pancreas tissues analyzed. (B) Frequency of 2-fold increase and percent of TCGA-PDAC samples with higher maximal expression compared to normal pancreas of the indicated GPCRs with comparison of the frequency of mutations of KRAS and TP53, the most frequent somatic, nonsilent mutations in PDAC tumors in TCGA. (C) Changes in GPCR expression in PDAC alter the GPCR repertoire that couple to different G proteins. (D–F) Tumor grade (D), pathological T (E), and patient sex (F) do not impact on GPCR expression. The 30 highest expressed GPCRs in PDAC tumors are shown for each case; no statistically significant differences occur between the groups. The numerical values used to generate panels A–F can be found at https://insellab.github.io/data. CPM, Counts Per Million; GPCR, G protein-coupled receptor; GTEx, Gene Tissue Expression Project; PDAC, pancreatic ductal adenocarcinoma; TCGA, The Cancer Genome Atlas; TPM, Transcripts Per Million.
Fig 4.
Grouping GPCRs and tumor types based on patterns of GPCR expression.
(A) A phylogenetic tree showing the hierarchical clustering of GPCRs from Fig 1A reveals subsets of GPCRs that are either high or low expressed in solid tumors. (B) Hierarchical clustering of types of solid tumors, based on their expression (in CPM) of GPCRs, reveals clusters of tumor types, characterized by expression of particular GPCRs. CPM, Counts Per Million; GPCR, G protein-coupled receptor; TPM, Transcripts Per Million.
Fig 5.
Hierarchical clustering of types of solid tumors, based on their expression (in CPM) of GPCRs, reveals clusters of tumor types with expression of various GPCRs at characteristic levels of expression.
BRCA, breast cancer; CPM, Counts Per Million; GPCR, G protein-coupled receptor; TPM, Transcripts Per Million.
Fig 6.
GPCR expression correlates with cancer-related pathways and predicts survival in PDAC.
(A) Network analysis via STRING [14] of the genes positively correlated with expression of GPCRs highly expressed in PDAC, with FDR < 0.001, revealing distinct clusters of genes associated with specific cancer-related pathways. A high-resolution version of the network is available on the accompanying website. (B) Based on GSEA preranked analysis [15], KEGG [16] gene sets with positive enrichment among genes most significantly positively or negatively correlated with expression of the identified subset of overexpressed GPCRs. (C) Analysis via GO [17,18] and Enrichr [19] of the genes positively correlated in (A), with FDR < 0.001, indicates enrichment for cellular compartments associated with exosomes, microvesicles and others; data shown are for results from Enrichr using the Jensen compartment database [20]. (D) Leading-edge analysis of GSEA results from (B), showing numerous cancer-related genes, including KRAS, which are common to multiple enriched KEGG genes sets. (E) GPCRs in PDAC show a positive, statistically significant association (FDR < 0.001) between their expression and that of markers for various cell types as shown. The top 10 GPCRs associated with each cell type are highlighted, along with the strength/significance of these correlations. (F, G) Kaplan-Meier curves for the impact of combined, normalized expression of subsets of GPCRs on PDAC patients. Total number of patients = 141; 59 patients were censored due to inadequate follow-up. (F) Impact of highly expressed GPCRs on median survival: 652 days (if below median expression) and 470 days (if above median expression) and for (G) impact of highly expressed immune-related GPCRs on median survival: 460 days (below median expression) and 603 days (above median expression). Numerical values used to generate panels B, C, D, F, and G can be found at https://insellab.github.io/data. FDR, false discovery rate; GO, gene ontology; GPCR, G protein-coupled receptor; GSEA, gene set enrichment analysis; KEGG, Kyoto Encyclopedia of Genes and Genomes; PDAC, pancreatic ductal adenocarcinoma.
Fig 7.
GPCR expression and presence of driver mutations and the similarity in GPCR expression of primary tumors, metastases, and cancer cells derived from the tumors.
(A) Correlation of median expression of GPCRs in TP53 mutant and PI3K mutant HR-positive BRCA IDC tumors. (B) Median expression of the 25 highest expressed GPCRs in TP53-mutated tumors compared to expression in PI3K mutant HR-positive BRCA IDC tumors. (C) Fold-changes of GPCRs in TP53 mutant and PI3K mutation HR-positive BRCA IDC tumors compared to normal breast tissue. (D) Fold-changes of GPCRs in Stage 1 and Stage 3 HR-positive BRCA IDC tumors over normal breast tissue. (E) Expression of the 25 highest expressed GPCRs and (inset) correlation of median GPCR expression between primary and metastatic SKCM. (F) Gene expression of primary and distant metastatic SKCM tumors cluster differently and have large numbers of differentially expressed genes (S2 Table). (G) Median expression of highest expressed GPCRs in PDAC tumors compared to cancer cells, including those analyzed via methods in [37] (CCLE [38], n = 33; Genentech [39], n = 16; Witkiewicz and colleagues [40], n = 72). (H) Median expression of highest expressed GPCRs in CCLE PDAC cell lines compared to other cell lines (as in panel G), primary cells and PDAC tumors. (I) Median expression of highest expressed GPCRs in SKCM tumors compared to cancer cells, including those analyzed via methods in [37] (CCLE [38], n = 45; Genentech [39], n = 44; Müller and colleagues [41], n = 29). (J) Median expression of highest expressed GPCRs in CCLE SKCM cell lines compared to other cell lines (as in panel I), primary cells and SKCM tumors. Numerical values used to generate panels A–E and G–J can be found at https://insellab.github.io/data. MDS files for panel F and for all other tumor types can be found at https://insellab.github.io/mds_plots. BRCA, breast cancer; CCLE, Cancer Cell Line Encyclopedia; CPM, Counts Per Million; FC, fold-change; FPKM, Fragments Per Kilobase of exon, per Million reads; GPCR, G protein-coupled receptor; HR, hormone receptor; IDC, infiltrating ductal carcinoma; MDS, Multidimensional Scaling; PDAC, pancreatic ductal adenocarcinoma; SKCM, skin cutaneous melanoma; TCGA, The Cancer Genome Atlas; TPM, Transcripts Per Million.
Fig 8.
Frequent overexpression of GPCRs in tumors does not correlate with frequency of mutation.
(A) The number of tumors with GPCR mutations and of tumor types/subtypes in which the same GPCR is overexpressed for the 10 most frequently mutated GPCRs in TCGA tumors surveyed and the 10 most frequently overexpressed GPCRs. (B) The frequency of overexpression compared to the frequency of mutations. Numerical values used to generate panels A and B can be found at https://insellab.github.io/data. GPCR, G protein-coupled receptor; TCGA, The Cancer Genome Atlas.
Fig 9.
Solid tumor-expressed GPCRs with potential as drug targets.
(A) The number of GPCRs that are targets for approved drugs and have increased expression in 1–3, 4–9, or ≥10 tumor subtypes. (B) The linkage to G proteins of the 77 GPCRs targeted by approved drugs and with increased expression in at least 1 tumor subtype. Note: multiple GPCRs couple to more than one G protein. (C) The number of GPCRs targeted by approved drugs that show increased expression in lung, colon, pancreatic, breast, and prostate cancers, the leading causes of cancer deaths in the US. (D) Overrepresentation of GPCRs among genes with >4-fold elevated expression (FDR < 0.05) for the indicated tumor types/subtypes with p-value calculated via Fischer’s exact test. (E) The magnitude of overrepresentation (relative enrichment) of GPCRs corresponding to the p-value in panel H. Numerical values used to generate panel C can be found in S2 Table, in sheets on DE of druggable GPCRs. Numerical values for panels D and E can be found at https://insellab.github.io/data. ACC, Adrenocortical Cancer; Ad, Adenocarcinoma; BRCA, breast cancer; CESC, Cervical Cancer; COAD, colon adenocarcinoma; DE, differential expression; ESCA, esophageal cancer; FDR, false discovery rate; GPCR, G protein-coupled receptor; Her2, Human Epidermal Growth Factor Receptor-2; IDC, infiltrating ductal carcinoma; KIRC, kidney clear cell carcinoma; KIRP, kidney papillary cell carcinoma; LIHC, liver hepatocellular carcinoma; LSQC, lung squamous cell carcinoma; LUAD, lung adenocarcinoma; NOS, Not Otherwise Specified; OV, ovarian cancer; PDAC, pancreatic ductal carcinoma; PRAD,; SKCM, skin cutaneous melanoma; STAD, stomach adenocarcinoma; TGCT, testicular cancer; THCA, thyroid cancer; UCS, Uterine Carcinosarcoma.
Fig 10.
GPCR expression is associated with survival in many types of solid tumors.
(A–C) Kaplan-Meier survival curves in the indicated tumor types, for weighted, combined expression of the indicated GPCRs. (D) In the indicated tumor types, the number of GPCRs with significant (p < 0.05) association with survival. (E–F) As examples, the impact on mean survival of GPCRs in (E) ESCA adenocarcinoma and (F) SKCM (distant metastases). Numerical values used to generate panels A–C can be found at https://insellab.github.io/data. Numerical values for panels D–F can be also found at https://insellab.github.io/data; in addition, the analysis results, metadata, GPCR expression values, and R-code for generating these survival data can be found at https://insellab.github.io/analyses, under the survival analysis section. ACC, Adrenocortical Cancer; BLCA, bladder cancer; BRCA, breast cancer; CESC, Cervical Cancer; COAD, colon adenocarcinoma; ESCA, esophageal cancer; GPCR, G protein-coupled receptor; KIRC, kidney clear cell carcinoma; KIRP, kidney papillary cell carcinoma; LIHC, liver hepatocellular carcinoma; LSQC, lung squamous cell carcinoma; LUAD, lung adenocarcinoma; NOS, Not Otherwise Specified; OV, ovarian cancer; PDAC, pancreatic ductal adenocarcinoma; PRAD, Prostate Adenocarcinoma; SKCM, skin cutaneous melanoma; THCA, thyroid cancer.
Fig 11.
Somatic nonsilent mutations of GPCRs.
(A) Frequency of GPCR mutations in the TCGA cohort (n = 5,103). Inset: 20 most frequently mutated GPCRs. (B) The average number of all genes (red) and GPCRs (blue) with somatic, nonsilent mutations per tumor genome for the TCGA tumor types surveyed. (C) The number of mutated GPCRs per tumor for several types of solid tumors in TCGA (black), the proportion of samples in each tumor type with at least one mutated GPCR (orange), and proportion with nonsilent mutations for the most commonly mutated GPCR (gray) for each tumor type. (D) GPCR mutation frequency is linearly related to Nmut in SKCM. (E) Probability of GPCR mutation as Nmut increases in SKCM for the 10 most frequently mutated GPCRs. (F) Normalized probability of GPR98 mutation as Nmut increases in SKCM, and the same for several cancers with high mutational burden and combined for BLCA, LUAD, LSQC, COAD, and SKCM. (G) The 20 most frequently mutated GPCRs in primary SKCM, with frequency of mutation and median (and upper and lower quartile) expression in TPM. (H) Expression (CPM) of the 5,000 most abundant genes in SKCM correlate closely in primary SKCM tumors that have or lack GPR98 mutations. (I) MutSig 2CV version 3.1 (gdac.broadinstitute.org) scores in SKCM obtained from https://gdac.broadinstitute.org/, showing the q-values for the significance of mutation scores for each annotated coding gene (blue); GPCR (black) and for those GPCRs, the number of mutation events (red) among the SKCM cohort. Numerical values used to generate all panels of this figure can be found at https://insellab.github.io/data. BLCA, bladder cancer; COAD, colon adenocarcinoma; CPM, Counts Per Million; GPCR, G protein-coupled receptor; LSQC, lung squamous cell carcinoma; LUAD, lung adenocarcinoma; Nmut, Number of Mutations per Genome; SKCM, skin cutaneous melanoma; TCGA, The Cancer Genome Atlas; TPM, Transcripts Per Million.
Fig 12.
Mutated GPCRs show high density of mutations but do not impact survival.
(A–B) Number of somatic (silent and nonsilent) mutations per unit of gene length (total length of all exons per genes) for (A) all genes and (B) genes >10 kb in length in SKCM. (C–D) Survival analysis in metastatic SKCM for (C) GPR112 and (D) GPR98 mutations, with p-values calculated using the Peto-Peto modification of the Gehan-Wilcoxon test. Numerical values used to generate all panels of this figure can be found at https://insellab.github.io/data. GPCR, G protein-coupled receptor; SKCM, skin cutaneous melanoma.
Fig 13.
CNVs of GPCRs in solid tumors.
(A) The number of solid tumors with CNV for each GPCR across all TCGA samples (plotted in descending order of frequency for high-level amplification and homozygous deletions; see text for definition of high- and low-level amplification). (B) The same as in (A) for low-level amplification and single-copy deletions. (C) In SKCM tumors (n = 367), the distribution of high-level amplification of GPCRs (n = 390 genes) compared to that of all protein coding genes (n = 24,776). (D) The total number of homozygous deletions, single-copy deletions, and low-level and high-level amplifications for all GPCRs combined in the 7,545 TCGA tumors surveyed for CNV. (E) The most frequently amplified GPCR for each TCGA tumor type and the proportion of samples with high and low-level amplification for that GPCR. (F) Proportion of samples of various tumors types with high and low-level amplification of GPR160, the most frequently amplified GPCR overall). (G) OV samples with and without high-level amplification of GPR160, with the median for each group indicated. The difference between groups was not statistically significant). (H) The risk ratio of elevated GPR160 expression (above the median value for OV) when GPR160 also shows high-level amplification; amplification of GPR160 increases the likelihood that GPR160 expression is elevated. (I) For the 30 GPCRs in OV with the highest fold-increase in expression relative to normal ovarian tissue (with FDR < 0.05 and median expression in OV > 1 TPM), the corresponding number (of 579) OV tumors with high-level amplification of those same GPCRs. (J) The same as (I) but comparing fold-increase against low-level amplification. Numerical values used to generate all panels of this figure can be found at https://insellab.github.io/data. CNA, copy number amplification; CNV, copy number variation; GPCR, G protein-coupled receptor; OV, ovarian cancer; SKCM, skin cutaneous melanoma; TCGA, The Cancer Genome Atlas; TPM, Transcripts Per Million.
Table 2.
Software used and availability of data.