Illuminating the Onco-GPCRome: Novel G Protein-Coupled Receptor-Driven Oncocrine Networks and Targets for Cancer Immunotherapy

G protein–coupled receptors (GPCRs) are the largest gene family of cell membrane– associated molecules


Introduction
The G protein coupled receptor (GPCR) family of proteins includes over 800 members and comprises ~4% of the encoded human genome, making it the largest gene family involved in signal transduction (1,2). Common to all GPCRs is the 7-transmembrane domain structure, which has an extracellular N-terminus and an intracellular C-2 terminus. The importance of the multiple biological roles GPCRs is reflected in the range of key physiological processes that they regulate, including vision, olfaction, neurotransmission, hormone and enzyme release, immune response, hemostasis, cardiac response and blood pressure regulation, epithelial cell renewal, stem cell fate decisions, tissue development, and homeostasis. In fact, dysfunction of GPCRs contributes to some of the most prevalent human diseases, which is reflected by the 475 currently approved drugs that target 108 unique GPCRs and representing 34% of all FDA-approved drugs (https://www.centerwatch.com/druginformation/fda-approved-drugs/) (3,4). While drugs for GPCRs represent ~34% of the global therapeutic drug market (3,4), only a handful of these are drugs for oncology; of the current FDAapproved anti-cancer drugs, only 8 of these target GPCRs, as described in detail below. GPCRs have been a longstanding topic of interest in the Journal of Biological Chemistry, and here we will expand on the impact of GPCRs in cancer biology. This review will summarize the current knowledge of how GPCRs are altered in cancer and how these aberrations can contribute to cancer initiation and progression. We also bring forth an emerging role of GPCRs as part of autocrine and paracrine signaling processes, which we refer collectively as oncocrine networks that drive tumor formation, growth and immune evasion. We also highlight the potential benefits of targeting GPCRs in the new era of precision cancer immunotherapies.

Historical perspective
The first evidence demonstrating a role for GPCRs in tumorigenesis came over 30 years ago in 1986 when studies illustrated that the GPCR encoded by Mas1 gene (MAS1) produced tumors in nude mice (5). This finding was largely underappreciated, likely because in contrast to most oncogenes discovered at the time, these receptors did not harbor activating mutations, similarly to the behavior of wild type 5HT1c receptors (HTR1C) that resulted in NIH3T3 cell transformation (6). Further work, however, revealed that wild-type GPCRs can become tumorigenic in a ligand-dependent fashion. This was best demonstrated in 1991 in studies depicting the oncogenic transforming ability of muscarinic acetylcholine receptors (mAChRs) in NIH3T3 cells only in combination with the agonist, carbachol, and exclusively for Gαq coupled mAChR subtypes (M1, M3, and M5, gene names CHRM1, CHRM3, and CHRM5, respectively) (7). With this, these studies brought to light the possibility of G protein-dependent oncogenic roles for GPCRs when activated by locally produced or circulating ligands, and raised the possibility that activating mutations in key conserved GPCR residues could result in transforming potential even without agonist stimulation.
These early studies introduced GPCRs as a new class of receptors capable of oncogenic transformation. Aligned with this possibility, mutational alteration of α1B adrenergic receptor (ADRA1B) can lead to transformation, providing an enhanced ability for tumor generation in nude mice (8). The identification of activating mutations in the thyrotropin receptor gene (TSHR) in hyperfunctioning thyroid adenomas provided the first evidence that mutant GPCRs can initiate a neoplastic disease (9). Downstream of the receptor, somatic mutations that impair the GTPase activity of Gαs conferred constitutive activation of adenylyl cyclase, leading to development of hyperfunctioning thyroid adenomas and pituitary tumors (10)(11)(12). Although these lesions are benign in nature, and hence often neglected in cancer biology, recent studies demonstrated similar activating mutations in the Gαs encoding gene (GNAS oncogene) in multiple cancer types, including pancreatic and colorectal cancer (13)(14)(15)). In addition, our systematic analysis of the transforming potential of G proteins revealed that the genes encoding the Gαq/11 (GNAQ and GNA11) and Gα12/13 (GNA12 and GNA13) G protein α subunits harbor transforming potential (16-18), thus contributing to the more recent discovery of multiple G proteindriven cancers (see below).
Remarkably, many human cancerassociated viruses utilize GPCR signaling for their life cycle. These include Kaposi sarcomaassociated herpesvirus (KSHV/HHV8) (19,20), human cytomegalovirus (HCMV) (21,22) and Epstein-Barr virus (EBV) (23), which encode receptors in their genomes that resemble human chemokine receptors, and deploy them to recruit immune cells and exploit the immune system for viral dissemination (reviewed in (24)). Specifically, the discovery that the GPCR encoded by KSHV/HHV8, often referred to as vGPCR or ORF74, initiates Kaposi's sarcomagenesis provided the first link between GPCRs and virallyassociated human malignancies (20,25). Viral GPCRs can signal through Gα proteins independent of ligand activation, and they take advantage of this "constitutive activation" to promote tumorigenesis and aid in tumor survival, growth and metastasis (24,26). The dawn of these studies opened a new door to establish the link between GPCRs and cancers.
Despite this large body of information, GPCRs were generally not thought to represent traditional "genetic drivers" in cancer, thus pursuing GPCRs in oncology was neglected for some time. In the past decade, however, studies bloomed linking GPCRs to many cancers and mechanisms of tumorigenesis, metastasis, and immune evasion. The goal of the comprehensive expression, mutation, and copy number alteration omics information presented in this review is to shed light on understudied GPCRs and G proteins in different cancers, and for leading experts in studying particular cancers, to direct more attention in considering GPCRs as potential therapeutic targets.

Canonical and non-canonical G protein and GPCR signaling
As a result of the use of alignment tools and gene ontology, 342 functional nonolfactory human GPCRs (1) have been documented and further classified into major receptor families based on sequence similarity and function (27,28) (http://gpcrdb.org/). To date, over 600 inactivating and almost 100 activating mutations in GPCRs have been identified, which are responsible for more than 30 different human diseases (29). Consequently, GPCRs have remained a longstanding interest as pharmacological targets.
GPCRs bind a wide variety of agonists, including ions, amines, purines, lipids, peptides, and proteins. Upon agonist binding, a conformational change is induced in the extracellular loops of the transmembrane region for ligand binding, and in the intracellular loops (primarily in the second, third, and fourth loops), which promotes receptor activation and G protein coupling (30)(31)(32). The basic signaling unit of a GPCR system includes five main components: the receptor, the trimeric αβγ G protein, an effector, RGSs (regulators of G protein signaling) that accelerate GTP hydrolysis and inactivate G proteins, and arrestins that control receptor fate and signal modulation (2). Once activated, the receptor binds the heterotrimeric G proteins, which promotes the release of GDP from the Gα subunit and the exchange for GTP and the functional dissociation of the GTP-bound α subunit from βγ dimers (2,31). Both parts remain attached to the plasma membrane, but free from the GPCR in order to interact with downstream signaling proteins.
A defining feature of GPCRs is the ability to activate one or multiple Gα proteins, which can be subdivided into 4 major families based on sequence similarity: Gαs, Gαi, Gαq/11, and Gα12/13 ( Figure 1). As previously reviewed (33,34), Gαs activates adenylyl cyclases to catalyze the conversion of ATP to cAMP, which is produced as a second messenger and activates protein kinase A (PKA) and in some cells guanine nucleotide exchange factors (GEFs) for the small GTPase RAP1. Members of the Gαi family primarily inhibit cAMP production, and activate a variety of phospholipases and phosphodiesterases, and promote the opening of several ion channels. The Gαq/11 family converts PIP 2 to DAG and IP 3 to activate PKC and elevate intracellular Ca 2+ levels. In a non-canonical fashion, Gαq/11 also stimulates Rho GEFs thereby stimulating Rho GTPases (35,36), while DAG activates Ras-GEFs (37). Gα12/13 signaling involves a family of RhoGEFs harboring an RGS domain by which they associate with active Gα12/13 and stimulate Rho GTPase (reviewed in (38)). In turn, as depicted in Figure 1, the coordinated activation of second messenger systems and Rho and Ras GTPases will result in the stimulation of multiple kinase cascades regulating key cellular functions. These include one or more members of the mitogen activated protein kinase (MAPK) (e.g., ERK1 and ERK2, JNK1-3, p38α-δ, and ERK5, AKT and mTOR, second messenger regulated kinases (e.g., PKA, PKC, PKD, PKG, and CAMKs) and phosphatases (e.g., calcineurin), and multiple kinases regulated by Rho (e.g., ROCK, LIMK, PKN, Citron kinase, PAKs and MLKs) and Ras (e.g., BRAF, ARAF, CRAF) GTPases, which in turn regulate nuclear events contributing to normal and malignant cell growth (reviewed in (33,34)). In addition, Gαs-coupled receptors activate and Gα12/13-, Gαi-, and Gαq/11-coupled receptors inhibit LATS1/2 kinases, which are key components of the recently described Hippo kinase cascade (39). LATS kinases phosphorylate and inhibit the transcription coactivator Yesassociated protein (YAP) and its related protein, TAZ, thereby causing their cytoplasmic retention and degradation (40). By inhibiting LATS1/2, Gαq-and Gα12/13-coupled GPCRs stimulate the ability of YAP/TAZ to promote the expression growth and anti-apoptotic genes (39). See below also exciting new information on how oncogenic Gαq proteins regulate the Hippo pathway and its therapeutic potential for Gαq-driven malignancies.
Once functionally dissociated from Gα protein, Gβγ dimers also play a central signaling role, first described in the context of ion channel regulation. For example, Gβγ can inhibit some voltage-activated Ca 2+ channels and activate G protein activated inwardly rectifying K channels (GIRKs) (41). In the context of cancer signaling, Gβγ dimers were initially shown to mediate the activation of ERK downstream from GPCRs linked to Gαi (42). We now know that Gβγ stimulates PLCβ, adenylyl cyclases, PI3Ks (primarily PI3Kγ, and PI3Kβ in cells lacking PI3Kγ expression) and GEFs stimulating the small GTPase Rac, such as PREX1 (33). By doing so, Gβγ signaling contributes to the prosurvival and migratory activity of many GPCRs, with emphasis on chemokine receptors involved in cancer metastasis, such as CXCR4 (43) (and see below).
Ultimately, the signaling pathways stimulated by each GPCR depends on its Gprotein coupling specificity, which can be distinct for each ligand (often referred to as biased agonism), the intensity and duration of receptor activation, and the level of expression of each G protein subunit and the repertoire of signaling molecules expressed in each cell type. The most proximal signaling pathways stimulated by each G protein subunit are summarized in Figure 1.
In addition to canonical signaling through heterotrimeric G proteins, some classes of GPCRs can initiate G protein-independent signal transduction. For example, some GPCRs also initiate intracellular signaling by engaging the scaffolding activity of β-arrestins, particularly for the activation of ERK and JNK3 (44). However, it is possible that G proteins may be required to initiate signal transduction, with β-arrestins playing a more important modulatory role in signal transmission, by shaping and fine-tuning dynamic GPCR responses (45).
G protein-independent signaling is well exemplified by the Frizzled (FZD) family of receptors. In this case, the FZD ligand, WNT, stimulates a signal transduction cascade that results in β-catenin activation through the protein disheveled (DVl), which plays a key role in embryonic development and cancer. WNT proteins bind FZD and a single-pass transmembrane molecule, LDL receptor related proteins 5 and 6 (LRP5/6), leading to the dimerization of the two receptors (46). The resulting conformational changes cause the phosphorylation of the cytoplasmic tail of LRP at multiple residues and the recruitment of GSK3β bound to scaffold protein Axin, while FZD associates with DVL. This complex formation prevents the persistent phosphorylation and consequent degradation of β-catenin bound to its degradation complex, which includes Axin, the tumor suppressor APC, the kinases GSK-3α/β and CK1, and the E3-ubiquitin ligase β-TrCP, thereby stabilizing β-catenin and promoting its nuclear signaling activity (46). The WNT/β-catenin pathway is frequently dysregulated in cancer, with particularly high incidence in colorectal cancer (47).
An interesting aspect of WNT signaling is that FZDs are persistently ubiquitinated and downregulated by the transmembrane proteins ZNRF3 and RNF43, and that this negative effect can be circumvented by secreted proteins of the Rspondin family that bind ZNRF3/RNF43 together with the GPCRs LRG4 and LGR5, suppressing ZNRF3/RNF43 function and leading to enhanced WNT signaling indirectly (46). However, as recently reviewed, FZD can also activate G proteins of the Gαi, Gαq, and Gα13 families, mediating many of the responses initiated by WNT exposure (48).
Another GPCR involved in development and cancer, particularly in basal cell carcinoma, is smoothened (SMO), which acts in the sonic hedgehog (SHH) pathway primarily by regulating the activity of the GLI transcription factor by a not fully understood mechanism in mammalian cells (49). Traditionally, this effect was considered to be G protein-independent, but GLI activation requires the inhibition of PKA, and growing evidence suggests that this aspect may require the activation of Gαi proteins or the inhibition of Gαs or its coupled receptors (reviewed in 47). How G protein signaling by FZD is coordinated in space and time with canonical β-catenin signaling, and how SMO regulates G protein-independent and G protein -regulated pathways to activate GLI and other signaling events in the context of cancer stemness and metastasis is an active area of current investigation (48)(49)(50). Its full elucidation may have important implications for the design of new pharmacological interventions in cancers that involve persistent G protein-independent and/ordependent WNT and SHH signaling.

The mutational landscape of G proteins and GPCRs in cancer
The Cancer Genome Atlas (TCGA) is a comprehensive, publicly available database launched by the National Institutes of Health (NIH) which includes large-scale genome sequencing analyses through multiple omics platforms for a variety of cancer types (51). In addition to this, the TCGA database also includes array-based DNA methylation sequencing for methylation profiling and reverse-phase protein array for large-scale protein expression profiling. These platforms can add a multidimensional view to the landscape of GPCRs and G proteins in cancer. Here, we built on our prior cancer genomewide study (13), performing an in depth omics analysis of the mutational landscape of 33 cohorts of cancer patients in TCGA by new bioinformatics approaches (Supplementary Table 1B).
The power of this analysis revealed that 20% of all human tumors sequenced contained mutations in genes encoding GPCRs. In particular, we used MutSig2CV, a now widely used computational biology tool that takes mutations discovered by DNA sequencing to illuminate genes that are statistically more frequently mutated relative to the background mutation rate of individual lesions (52). Many G proteins and GPCRs were found to be mutated. For visualizing the data, we used a very stringent criterion (MutSig2CV q-value < 0.25) to identify the most statistically significant mutated genes in each cancer type. An unexpected observation was that among all cancer cohorts, cancers arising in the gastrointestinal (GI) tract, including colon adenocarcinoma (COAD), stomach adenocarcinoma (STAD), and pancreatic adenocarcinoma (PAAD) displayed the highest number of significantly mutated GPCRs and G proteins (Figure 2, Supplementary Tables 2A  and 2B). This may be independent of the mutational burden of these tumors, which are lower than that of other typical highly mutated cancers such as melanoma, and lung cancer, for example (53). However, the phenotypic and biological outcome of these mutations remains largely unknown, and thus these findings provide a wealth of information for the development of hypothesis-driven approaches to investigate their cancer relevance.
In addition to our analysis of the most statistically significant mutated and genomically altered G proteins and GPCRs in cancer (q < 0.25), we have compiled the frequency of mutations of all G proteins and GPCR genes for each cancer type investigated in TCGA (Supplementary Table 6). We expect that this color-coded table will provide easy access and visualization of the cancers in which G proteins and GPCRs of interest are most frequently mutated. We generated this table using the more recent and robust Multi-Center Mutation Calling in Multiple Cancers (MC3) project TCGA PanCancer 2018 dataset (54). This database includes mutation-calling algorithms that account for variance and batch effects to enable more precise cross-tumor-type analyses (54). We have also provided a direct link for each gene to their corresponding page in cBioPortal Cancer Genomics portal (http://www.cbioportal.org/) for the visualization, analysis and download of mutational information.
The cBioPortal for Cancer Genomics is a web resource for dissecting and visualizing multidimensional cancer genomics data. These data includes information about somatic mutations, copy number alterations, mRNA expression, DNA methylation, and transcript and protein abundance from multiple cancer omics studies (55). Please note that the percentage of mutated samples may vary with our analysis, as cBioPortal analysis uses different instances of the TCGA PanCancer dataset from 2013-2018 (56). We encourage our colleagues to follow the corresponding links to gain easy access to: a) "Cancer Types Summary", in which all genomics alterations are displayed for all cancer types, b) "Mutations", which provides a visual representation of the most frequently mutated and altered residues and a downloadable list of samples that includes their corresponding protein change mutations, mutation type, and CNV type, c) "Survival", which shows the overall survival (length of time that the patients are alive) of cancer patients harboring genomic alterations versus those without (though, we recommend to perform this analysis for each particular cancer type of interest), and d), "Expression", which provides a graphical representation of the mRNA expression level of each sample in every cancer type, together with their mutational status.

Significantly mutated G Proteins in cancer:
Whereas the contribution of each GPCR mutation in cancer is still under evaluation, the recent discovery of hot spot mutations in G proteins as oncogenic drivers in multiple highly prevalent cancer types has accelerated tremendously the research in this field. Indeed, many G protein genes (GNAS, GNA11, GNAQ, and GNA13) are part of the current ~400 gene panels of cancer-associated genes sequenced routinely by clinical oncology services in many cancer centers, and by all large cancer genomic testing providers and institutional genomics cores. Among them, the summary of our MutSig2CV analysis revealed that GNAS is the most highly mutated G protein in human cancer (Supplementary Table S2B). From this analysis, GNAS is significantly mutated in colon adenocarcinoma (COAD) (6.19%), pancreatic adenocarcinoma (PAAD) (5.09%), and stomach adenocarcinoma (STAD) (7.52%). As described above, GNAS is a known oncogene that was first described in growth-hormone secreting pituitary adenomas and has since been found to be mutated in a number of neoplasms, predominantly at the codon 201 hotspot (13,57). Mutations occurring at arginine 201 of GNAS activate adenylate cyclase and lead to constitutive cAMP signaling by reducing the rate of GTP hydrolysis of the active GTP-bound Gαs, as well as by adopting an activelike conformation even when bound to GDP (13,58). In COAD, a synergistic effect with the MAPK pathway is likely, as GNAS is co-mutated with KRAS in a large portion of adenomas and carcinomas. Similarly, GNAS mutations are found in approximately 50% of low-grade appendiceal mucinous neoplasms (59), and highly prevalent in a subset of pancreatic tumors, including intraductal papillary mucinous neoplasms and adenocarcinomas (14). In this regard, recent mouse models revealed that GNAS and KRAS mutations are necessary and sufficient to initiate this particular subtype of pancreatic adenocarcinomas (60,61).
Emerging studies have begun to explain the functional impact of GNAS mutations. In 1991, GNAS mutations were discovered in McCune-Albright syndrome and pituitary tumors (62). In cancer, GNAS has been linked to proinflammatory functions, which could mimic the impact of chronic inflammation on tumor development. Gαs is well documented to mediate the effects of inflammatory mediators like cyclooxygenase 2 (COX2)-derived prostaglandins. Its inflammatory role in cancer is best shown in colon neoplasia where COX2-derived prostaglandin E2 (PGE2) enhances colon cancer progression via activation of PI3K and AKT and relieving the inhibitory phosphorylation of βcatenin as part of Gαs oncogenic signaling (63). Activating mutations in GNAS have also been found in gastric adenocarcinomas, leading to activation of the Wnt/β-catenin signaling pathway (64).
Mutations in GNAQ and GNA11 are most relevant in uveal melanoma (UVM) incidence, as 93% of patients harbor mutations in these genes by guest on March 24, 2020 http://www.jbc.org/ Downloaded from encoding constitutively active Gαq family members (65,66). All cancer mutations in Gαq or Gα11 occur at either glutamine 209 or, in a smaller proportion, arginine 183 (Q209 and R183, respectively; R183 is the identical position to R201 in Gαs) (65,66). Mutations affecting Q209 in GNAQ or GNA11 are present in most primary UVM lesions and their metastases (66). Mutated residues impair GTPase activity (diminish GTP hydrolysis), which ultimately leads to prolonged signaling. While initial studies supported a role of ERK signaling in UVM development, targeting this pathway did not improve the survival of UVM metastatic patients (67). Instead, our genome wide RNAi screens revealed that the non-canonical activation of RhoGEFs, specifically TRIO, by Gαq mediates UVM progression (68). Furthermore, we discovered that the activation of YAP, the most downstream target of the Hippo pathway, by the novel TRIO-RHO signaling arm is essential for UVM, thus identifying a druggable target downstream from mutated Gαq (68).
GNAQ mutations are also associated with a smaller proportion of skin cutaneous melanoma (SKCM), and have been recently described in vascular tumors, such as hemangiomas and angiosarcomas (15,69). GNAQ R183Q mutations are also specifically responsible for a frequent congenital neurocutaneous disorder characterized by port wine skin lesions that are vascular-derived, which is known as Sturge-Weber syndrome (70). Thus, mutations in GNAQ appear to be responsible for numerous disease conditions for which there are no current targeted therapeutic options.
Mutations in GNA13 have been characterized in both liquid and solid tumors, and are present in high frequency in bladder carcinoma (BLCA). In addition, recent genome-wide sequencing efforts have unveiled the presence of frequent mutations in GNA13 in lymphomas, specifically Burkitt's lymphoma and diffuse large B-cell lymphoma (DLBCL) (71)(72)(73). These mutations in GNA13 as well as in RhoA, a downstream target of Gα13, have been shown to be inhibitory in nature, suggesting a tumor suppressive role for Gα13 and RhoA in Burkitt's lymphoma and DLBCL (71). In this case, loss of function (LOF) mutations rather than gain of function (GOF) mutations underlie the oncogenic activity of GNA13, likely by disrupting the normal differentiation program of B cells (71). In contrast, wild type GNA13 overexpression has been implicated in many solid tumors, such as in gastric cancer (74), nasopharyngeal carcinoma (75), prostate cancer (76), and breast cancer (77). -Furthermore, GNA13 levels modulate drug resistance and tumor-initiation phenotypes in patient-derived head and neck squamous cell carcinoma (HNSCC) cells in vitro and in vivo (78). In this case, GNA13 or GNA12 overexpression may enhance the proliferative and pro-migratory function of multiple GPCRs that converge to activate these G protein α subunits. A causal role of excessive Gα12 signaling may be elucidated by a use of a recently developed Gα12coupled chemogenetic designer GPCR (DREADD) (79)(Inoue et al. Cell, in press)..

Mutations in Gβ subunits are infrequent
, and yet activating mutations in Gβ1 and Gβ2 (GNB1 and GNB2, respectively) have been identified in myeloid and B cell neoplasms, which act as an oncogenic driver and confer resistance to kinase inhibitors targeting typically mutated kinases in these malignancies, including BCR-ABL, BRAF, and JAK2 (80). Certainly, this information suggests that other Gβ subunit mutations may also harbor tumorigenic potential.

The mutated oncoGPCRome:
The most frequently mutated GPCRs in each cancer type are depicted in Figure 2, and listed in Supplementary Table S2A with the corresponding statistical significance (q-value) and frequency. As mentioned above, the high frequency of GPCR mutations specifically in tumors arising from the gastrointestinal track is intriguing as it likely reflects their ability to stimulate organ-specific growth promoting pathways in these cancers. While a discussion of each specific GPCR is beyond the goals of our review, we will discuss new emerging concepts and specific cases that may exemplify the challenges and opportunities for future exploration in this area, and its potential for drug discovery.
Whether mutations in GPCRs result in GOF or LOF, or represent passenger mutations with little impact on cancer progression, is in most cases still unknown. A complicating factor is that most GPCRs do not harbor hotspot mutations, meaning that mutations in each GPCR do not occur with high frequency in a single or limited numbers of codons, and in addition, each tumor exhibits a different repertoire of mutated GPCRs. To address this daunting question, we have recently developed new bioinformatics approaches analyzing GPCR mutations in the context of multiple sequence alignments (MSA) defining the conserved seven transmembrane (7TM) domain, as well as considering 3D structures and interaction partners (Raimondi, et al., Oncogene, in press). We have used this approach to model the most significantly mutated GPCRs (Supplementary Table S2A). Remarkably, visualization of the most mutated 7TM positions on a representative GPCR 3D structure revealed that most mutations occur in "hotspot structural motifs" rather than being randomly distributed (Figure 3 and Supplementary Figure S3). This includes frequent mutations in the DRY arginine motif, which is important for Class A GPCR activation as it is responsible for the intramolecular polar contacts that keep the receptor inactive until ligand binding (81). Other structural mutation hotspots are found at or nearby highly conserved GPCR regions, including the ligand and G protein binding sites, as well as the NPxxY and other conserved motives which regulate in an allosteric way receptor's activation (82). Collectively, this supports that most cancerassociated mutations in GPCRs occur in "structural hotspots", similar to other oncogenes and tumor suppressor genes, a property that could have not been predicted from the analysis of individual GPCRs.
While the functional impact of these alterations may need to be investigated for each GPCR, our recent computational analysis of cancer genomes indicates that most Gαi-linked GPCRs exhibit DRY mutations that are inhibitory in nature (inhibit function), which typically occur mutually exclusively with GNAS activating mutations (Raimondi et al., Oncogene, in press). This suggests the exciting possibility that mutations in Gαi-GPCRs may mimic GNAS mutants leading to higher cAMP activity to drive tumorigenesis (Raimondi et al., Oncogene, in press).
A particular challenge when analyzing the potential impact of cancer mutations is that longer genes exhibit a higher number of mutations, which would achieve statistical significance (MutSig2CV analysis) only when higher than the background mutation rate of individual lesions. This is well exemplified by GPR98, which is the most frequently mutated GPCR across all cancer types, and concomitantly, is the GPCR with the highest number of amino acids. GPR98 is an adhesion receptor, and its ligand and physiological functions are currently poorly understood. GPR98 mutations are known to cause febrile seizures and one form of Usher syndrome, the most common genetic cause of combined blindness and deafness (83,84). GPR98 has been shown to have significant association with glioblastoma (GBM) (85) and lymphoblastic leukemia (86), and the evaluation of the impact of GPR98 mutations in cancer warrants further investigation. The family of metabotropic glutamate GPCRs, GRM1-8, are also frequently mutated in many cancer cohorts. Mutations of GRM1, GRM5, and GRM3 have been shown in breast cancer and melanoma (87)(88)(89). In addition, their transforming potential and increased secretion of their ligand, glutamate, by the tumor microenvironment makes the GRM receptor family an intriguing area of study.
The analysis of the mutational landscape of GPCRs suggest that COAD harbors the highest incidence of significantly mutated receptors. Among them, thyroid stimulating hormone receptor (TSHR) was the most frequently mutated GPCR, involving ~14% of COAD patients. Mutations in the P2Y purinoceptor 13 (P2RY13) gene were the most statistically significant in this cancer type, and occurred in ~5% of COAD patients. P2RY13 encodes for a purine receptor and has been shown to be overexpressed in AML samples, but not involved in other nonhematologic malignancies (90). On a related note, mucosal biopsies from the colon of Crohn's disease and ulcerative colitis patients have shown abnormalities in P2RY13, which may suggest a role for the receptor in GI inflammatory diseases (91). The importance of TSHR activating mutations in human neoplasia was first demonstrated in thyroid adenomas (9), and are also found in some thyroid carcinomas. However, the roles of both TSHR and P2Y 13 in COAD remain largely unexplored.
Recently, analysis of hotspot mutations in oncogenes uncovered a mutation in cysteinyl leukotriene receptor 2 (CYSLTR2) in a uveal melanoma (UVM) cohort. This GOF mutation results in a L129Q substitution and leads to the Gαq coupled receptor to be constitutively active (92). This mutant protein is insensitive to leukotriene stimulation and constitutively activates Gαq, and can promote tumorigenesis in melanocytes in vivo (92). According to MutSig2CV analysis, CysLT 2 is the most frequently mutated GPCR (3.75%) in UVM.
Although representing a small fraction of all UVM cases, these mutations in CYSLTR2 are mutually exclusive with known drivers in UVM (GNA11 and GNAQ) (92). Therefore, CYSLTR2 mutations promote persistent Gαq activation substituting for GNA11 and GNAQ mutations to drive aberrant Gαq signaling in UVM. This receptor is also mutated in COAD at a distinct amino acid, and hence its consequences (GOF or LOF) are still unknown. Recently, small molecules have been discovered and utilized against wild-type CysLT 2 , but development of higher affinity molecules or antibodies that can stabilize the mutated receptor in its inactive state will be required to explore the therapeutic benefit of targeting CysLT 2 in UVM.
Our current analysis also identified many adhesion receptors and class A GPCRs that are mutated with high frequency in cancer. The former includes GPR98, BAI3, ADGRL1, CELSR1, GPR125, GPR110, GPR112, and GPR126, which can now be prioritized for their individual analysis. A recent comprehensive mutagenesis screen in ADGRL1 revealed that many cancerassociated mutations result in GOF alterations and persistent activity (93).
Among the typical class A GPCRs, some of the more frequently mutated genes are muscarinic receptors M2 and M3 (CHRM2 and CHRM3), multiple P2Y receptors, serotonin receptors (HTR1E, HTR1F, HTR2A, and HTR7), and adenosine receptors (ADORA3), among others, all of which could be activated by locally produced ligands as well. Notable mutated GPCRs also include the PAR2 receptor (F2RL1), which is often amplified and will be discussed below, as well as multiple orphan GPCRs whose coupling specificity and biological activity is still largely unknown. Given the emerging studies supporting the notion that aberrant GPCR activity leads to tumor initiation and progression, we expect that the emerging mutational information will guide new cancer relevant studies addressing each of these frequently mutated GPCRs. Given that many ligands of GPCRs may be produced in significantly higher amounts in the hypoxic, metabolic, and acidic tumor microenvironment, the tumorigenic synergism between ligand availability and activating mutations in receptors should also be explored.

Gene copy number alterations and G protein and GPCR expression in cancer:
In addition to mutations, alterations in gene expression and copy number of G protein and GPCR genes have been detected. Determining the contribution of such alterations to cancer initiation and progression remains a significant challenge, yet it may be critical both for discovery of driver oncogenic processes and for the development of targeted therapeutics. Indeed, aberrant expression of many wild type G proteins and GPCRs can contribute to cancer growth even if not mutated, often as part of oncocrine signaling networks (see below).
Somatic alterations are acquired at random during cell division, and some of these participate in tumorigenesis or tumor growth. Here, we used GISTIC (Genomic Identification of Significant Targets in Cancer), an algorithm that identifies genes targeted by somatic copy-number variations (CNVs) that may contribute to tumorigenesis by evaluating the frequency and amplitude of observed events (94). To illuminate the most relevant GPCR candidates in tumorigenesis, we also filtered the large list of CNVs for those that correlated with mRNA expression. Our analysis revealed that 28 out of 33 TCGA cancer cohorts included alterations of GPCR and G protein that are significantly correlated with mRNA expression of the corresponding genes (R > 0.33) (Supplementary Tables S4A,S4B, S5, and Figure 4).
Among the G proteins, copy number gain in GNA12 is remarkably significant in ovarian cancer (OV). This cancer type is characterized by few driver mutations and by the accumulation of high concentrations of LPA in ascites fluids, which may work through Gα12 to promote growth and metastasis (reviewed in (95)). Similarly, GNAI1 (encoding Gαi1) is significantly amplified in breast invasive carcinoma (BRCA), a cancer type in which many Gαi-coupled GPCRs, including CXCR4, are well established as metastatic drivers (see below). The significance of other genomic alterations in G proteins, including copy number gains in Gβ (GNAB1, GNAB2, GNAB3, and GNAB5) and Gγ (GNG4, GNG5, GNG7, GNG12, and GNGT1) subunits in multiple cancers likely reflect the broad signaling capacity of Gβγ-dimers (see Figure 1).

Testicular germ cell tumor (TGCT)
displayed the most genomic alterations in genes encoding GPCRs, which included mostly orphan, taste, and adhesion receptors. On the other hand, F2RL1, the gene encoding protease-activated receptor 2 (PAR2), was the most significantly altered gene in ovarian cancer (OV). PAR2 is a protease-activated receptor and is expressed in many organs. The ability of proteases to degrade extracellular matrices and to activate PAR receptors render them important in the facilitation of tumor growth and metastasis (96,97). Overexpression of F2RL1 has been linked to some of the most diagnosed cancers, including lung, breast, colon, and pancreatic cancers (96,98,99). Functionally, PAR2 has also been linked to cancer cell migration and stimulates vascular endothelial growth factor (VEGF) production for angiogenesis (100,101). Another unexpected observation was that most kidney cancers (KIPAN) exhibit highly significant copy number gains in genes for multiple chemokine receptors (CCR2, CCR5, CCR6, CCR9, CX3CR1 and CXCR6) and histamine receptors (HRH2), among others. The frizzled family of GPCRs and LPA receptors (in particular LPAR6) were also genetically altered in multiple cancer types. Overall, while gene copy gains and losses may reflect cancer-associated genomic instability, most cancers exhibit a very specific pattern of copy number variations in G protein and GPCR genes, whose biological relevance can now be examined.

G proteins and GPCRs as tumor suppressor genes?
An interesting observation of the pattern of genomic alterations is that many cancers lose one or both copies of specific G protein and GPCR genes. This raises the possibility that certain G protein/GPCRs may act as tumor suppressors rather than oncogenes. Indeed, as described above, GNA13 is significantly mutated in diffuse B cell lymphoma and Burkitt's lymphoma, and detailed experimental analysis revealed that in all cases this involves LOF mutations, resulting in the inability of B cells to undergo terminal differentiation and hence increasing their uncontrolled growth (71). Analysis of candidate Gα13-coupled GPCRs identified inactivating mutations in P2RY8 and LOF mutations in RHOA in these (71,102). These mutations appear to be mutually exclusive, supporting the notion that in these B cell malignancies P2RY8-GNA13-RHOA are part of a tumor suppressive axis.
Surprisingly, while conducting conditional gene knock out studies of Gαs in the skin, we observed that mice develop massive basal cell carcinomas (BCC) in only two weeks after Gnas excision (103).
This involved a widespread activation of the sonic hedgehog (SHH) pathway in Gnas-/-tumor lesions (103), thus phenocopying the effects of LOF mutations in PATCHED (PTCH) or GOF mutations in SMO, which are the best known BCC tumor suppressor and oncogenes, respectively. Activation of the SHH pathway is also typical of a subset of medulloblastomas, a childhood malignancy (104).
Remarkably, homozygous GNAS gene loss was identified in a group of SHH subtype medulloblastomas that does not harbor mutations in PTCH or SMO (104). In these particular cancers that express SHH pathway components, Gαs and its downstream target, PKA, act as tumor suppressors by preventing the activation of GLI transcription factors (103), while in most GI tissues, GNAS and PKA signaling act as tumor promoters. The former implies that certain yet to be identified Gαs-coupled receptors may exert a tumor suppressive function in BCC. The latter raises the possibility that in GI cancers, Gαi-coupled GPCRs may act as tumor suppressors and hence that their LOF mutations might be protumorigenic. This is aligned with the large number of Gαi-coupled GPCRs that are mutated in GI tumors, however whether they exhibit GOF or LOF mutations has not been yet tested formally. These particular predictions are of high clinical relevance, as overactive cAMP/PKA activity in many GI tumors could be counteracted therapeutically by stimulating locally expressed Gαi-GPCRs, while BCCs and SHH-subtype medulloblastomas may be treated by raising cAMP using phosphodiesterase inhibitors (as proposed in (104)) or by stimulating locally (or systemically) Gαs-GPCRs expressed in these tissues. These exciting possibilities will likely be explored in the near future.

Pan-cancer GPCRs expression
In addition to mutations, normal GPCRs can play a key role in cancer progression, and they can be targeted pharmacologically for therapeutic purposes. A typical problem when analyzing gene expression changes in cancer is that often both normal and cancerous tissues are heterogeneous, including multiple cell types. Hence, relative changes (fold changes and over-and underexpression) may reflect cellular heterogeneity more than the progression from a normal cell to its distinct cancer states. For example, comparison of GPCRs expressed in cutaneous melanoma with normal skin may grossly overestimate the relative changes in expression between normal and cancerous melanocytes, as the normal skin includes a very limited number of melanocytes. Moreover, though fold changes can provide useful information, this takes attention away from GPCRs that may exert important functions for cancer transformation through increased local ligand secretion or aberrant downstream signaling activity. A recent study has documented relative changes in GPCR expression in cancer (105). Instead, we focus here on illuminating absolute expression levels of each GPCR and provide visual representations in order to gauge absolute GPCR levels. Certainly, a limitation of this analysis is that the precise cells that express each GPCR within the tumors, such as cancer and tumor stromal cells (e.g., cancer associated fibroblasts, blood vessels, and immune infiltrating cells) will need to be established in future efforts, for example by the use of modern single cell sequencing approaches. Nonetheless, we expect that we can gain an unprecedented new perspective on GPCR expression patterns in human malignancies by utilizing information gained from this analysis.
Specifically, as shown in Figure 5, an intriguing area of study is the expression of orphan GPCRs in cancer. The endogenous ligands of more than 140 of these receptors remain unidentified and /or poorly understood, thus, their natural function is currently largely unknown (106). Nevertheless, according to our pan-cancer analysis, orphan GPCRs are differentially expressed across cancer types, and they may exert multiple functions during cancer progression (Supplementary Figure 1M). For example, since a decrease in extracellular pH is a major tumor promoting factor in the tumor microenvironment, an intriguing area of research is the group of proton-sensing GPCRs: GPR132, GPR65, GPR68, and GPR4, which are highly expressed in a large range of human cancers. Both GPR4 and TDAG8 (GPR65) have been shown to be overexpressed in many cancers and can cause malignant transformation of cells in vitro (107). GPR132 (also known as G2A) was previously shown to have tumor suppressor properties, as it prevents oncogenic transformations of pre-B cells by the BCR-ABL oncogene, similar to the role of GNA13 in these cell type (108). However, GPR132 has been shown to be highly transforming in fibroblasts (109). Thus, proton-sensing GPCRs may display tumor promoting or suppressive functions depending on the cancer cell of origin, and may also display pro-tumorigenic activity when activated in the tumor stroma (105). Interestingly, in our recent G protein coupling predictor trained by a large experimental dataset, orphan GPCRs tend to show a higher proportion of coupling toward Gα12/13 than other GPCR classes (79) further suggesting potential importance of orphan GPCRs in cancers that involve aberrant Gα12/13 signaling.
The leucine-rich repeat-containing (LGR) receptors LGRs 4-8 are known for their role in development, and bone formation and remodeling, but LGR4 and LGR5 are also upregulated in several cancer types (110). These receptors are expressed in multiple tissue resident stem cells, and their overexpression may reflect the expansion of this cellular compartment as well as the establishment of cancer stem cell niches (110). Overexpression of LGR4 and LGR5 in colon and ovarian tumors most likely enhances cell proliferation and metastasis (111,112). Interestingly, many class A orphan GPCRs are rarely expressed across cancer types. These include the MAS oncogene, which can explain the limitations in analyzing its role in human cancer in spite of its initial identification during transfection experiments several decades ago. Others are expressed in a single (e.g., GPR22 in PCPG) or few cancers (e.g., GPR17 and GPR37L1 that are expressed only in GBM and LGG), while others are expressed in most cancers, such as OPN3 and LGR4. These studies de-orphaning GPCRs and uncovering the function of additional overexpressed GPCRs may provide promising candidates for therapeutic intervention in cancer.
The pan-cancer expression of each GPCR class is depicted in Supplementary Figures 1A-N. We hope that this information will be useful for hypothesis generation in our large community of scientists working in the field of GPCRs in academia and industry. While this review will not provide a comprehensive analysis of each GPCR, a few concepts may be worth discussing. For example, expression of the purinergic P2Y 11 and adenosine A 2A receptors is widespread in all cancers, while GBM tumors express high levels of ADORA1, ADORA2, and ADORA3, all of which can be activated by adenosine in the tumor microenvironment. Multiple lipid receptors for S1P (S1P 1-3 ) and LPA (LPA 1 , LPA 2 , and LPA 6 ) are widely expressed as well. These receptors are intriguing because ligands for these receptors have been shown to accumulate in the tumor microenvironment (113,114). Conserved residues in these receptors also display a high mutational rate, which suggests that they may play vital roles in receptor signaling initiation, termination, and coupling specificity (13). This is also highly relevant for the 17 known GPCRs that specifically recognize intermediates or (by)products of cellular metabolism, which are often involved in nutrient sensing (115). These include receptors sensing amino acids and amino acid metabolites (GPR142, CasSR, GPR35, TAAR1, and FOPR1/2), bile acid (TGR5/GPBAR1), triglyceride metabolites (e.g. FFA1/GPR40, FFA4/GPR120, and GPR119), products of the intermediary metabolism and small carboxylic metabolites such as acetate and propionate (FFA2/GPR43 and FFA3/GPR41), butyrate (FFA2/GPR43, FFA3/GPR41 and HCA2/GPR109A), β-hydroxybutyrate (HCA2/GPR109A), β-hydroxyoctanate (HCA3/GPR109B), lactate (HCA1/GPR81), succinate (GPR91) and capric acid (GPR84) receptors, as well as gut microbiota-derived products (e.g., short-chain fatty acids, such as acetate, propionate, and butyrate) (Reviewed in (115)). These receptors are highly expressed in multiple organs of the digestive tract and immune cells (116), and they may be persistently activated in the tumor microenvironment due to the high metabolic rate that characterizes most solid tumors.
The EP 4 (PTGER4) and EP 2 (PTGER2) receptors for the typical inflammatory mediator PGE2 (see below) are also widely expressed, whereas EP 3 (PTGER3) is mainly expressed in kidney cancer (KICH). PGE2 plays a critical role in epithelial regeneration following tissue injury and cancer growth, which occurs via a PI3K/Akt and β-catenin pathways (63,117). COX2 overexpression and enhanced PGE2 production is most notable in colorectal cancer, and COX2 blockade can help explain the cancer chemopreventive activity of aspirin and other nonsteroidal anti-inflammatory drugs (NSAIDs) (118). However, direct roles for PGE2 in tumorigenesis have been demonstrated for many other human malignancies, including breast, lung, liver, and gastric cancers, among others. For example, in laboratory models of breast and gastric cancers, COX2 overexpression and alterations in Wnt signaling both led to increased tumorigenesis (119,120). Moreover, EP 3 has been shown to be involved in angiogenesis in lung cancer cell lines by increasing VEGF and metalloproteinase-9 (MMP-9) expression (121).
Among the class of GPCRs for proteins, (Supplementary Figure 1E) which includes chemokine receptors, CXCR4 is the most widely expressed. This may include many cancers that express CXCR4 under hypoxic conditions, as well as in blood vessels and immune cells (see below) (122)(123)(124). Other chemokine receptors that are highly expressed in immune cells (see below) were less well represented, suggesting a more limited impact of immune infiltrating cells to the overall mRNA expression patterns in our pancancer analysis. The analysis of GPCRs activated by peptides (Supplementary Figure 1F) show a clear widespread expression in genes for thrombin PAR1 (F2R) and PAR2 (F2RL1) receptors and endothelin receptors (EDNRB), the latter with particularly higher expression in skin (SKCM) and uveal (UM) melanomas. HRH1, encoding H 1 histamine receptor, is the most widely expressed aminergic GPCR (Supplementary Figure 1G), while M1 muscarinic receptors (CHRM1) and β1adrenergic receptors (ADRB1) are highly expressed in prostate cancer, the latter receptor being of unexpected importance for the most highly prevalent cancer among males (see below). Another interesting finding was the high level of expression of dopamine receptor 2 (DRD2) in a well-defined set of cancers, including glioblastoma (GBM), considering that a new family of antagonists for this receptor has exhibited encouraging anti-tumor activity in multiple cancer types (125,126).
Interestingly, from our analysis of Frizzled GPCRs, smoothened (SMO) is widely expressed in most cancer, beyond its initial main role in basal cell carcinoma (BCC). This might be due to SMO being expressed in cancer stromal cells that are present in most solid tumors (Supplementary Figure 1G) (127,128). There is also widespread expression of FZD 6 and a more cancer restricted expression of FZD 1 and FZD 4 (Supplementary Figure 1H).
Intriguingly, analysis of the sensory GPCRs revealed high level of expression of the taste receptor, TAS1R3, across most cancer types, which has not been previously investigated (Supplementary Figure 1J).
The adhesion GPCR family has mainly been studied in immunological and developmental functions, but they have recently been linked to cancer (Supplementary Figure 1M). For example, EMR2 (ADGRE2) is overexpressed in human breast cancer and increased nuclear expression of EMR2 is negatively correlated with tumor grade (129). Additionally, CD97 (ADGRE5) and GPR56 (ADGRG1) are the highest expressed adhesion GPCRs across all cancers, but they have only been studied in the context of melanoma, gastric, esophageal, and thyroid cancers (130)(131)(132). Additionally, GPR65 (TDAG8) and GPR133 (ADGRD1) have also been associated with human cancers and linked to tumor promotion (107,133), but the role of this highly expressed family of GPCRs in tumor initiation and metastasis is still not fully understood.
Overall, we expect that the emerging pancancer information on GPCRs expression will ignite new interest on their study in human malignancies.

GPCRs in metastasis and angiogenesis
Metastasis is one of the cancer hallmarks, in which tumor cells can acquire the ability to migrate and disseminate from the tumor to distant tissues. Cancer cells spread from the primary organ to secondary sites through lymphatic vessels and blood, and is the result of a sequential, highly organized, and organ-selective process. The precise mechanisms determining the directional migration and invasion of tumor cells into specific organs remain to be fully established, but chemokine receptors, all of which are GPCRs, have been the most popular place to look (134,135). Chemokines are small, cytokine-like proteins that induce directional migration for immune cells through interaction with GPCRs. Chemokines are secreted by multiple organs and act in a coordinated fashion with cell-surface proteins to direct homing of immune cells to specific anatomical sites (136,137). To serve a similar purpose, tumor cells can hijack chemokine receptor networks and migrate towards specific chemokines, facilitating metastasis to other organs, primarily the liver, lungs, brain, lymph nodes, and bone marrow (134). In addition, the tumor microenvironment includes chemokines that can enhance the motility and survival of cancer cells in an autocrine and paracrine fashion, a process that we refer as oncocrine signaling.
There are 23 distinct chemokine receptors in humans, and they are divided into four classes according to the type of chemokine they interact with (CC, CXC, CX3C, or XC) (135,138,139). CXCR4 and CCR7 represent the best studied chemokine receptors driving cancer metastasis, as they play active roles in tumor growth, invasion, angiogenesis, metastasis, and cancer relapse and therapeutic resistance (134,140,141). CXCL12, the chemokine for CXCR4, is highly expressed in multiple tissues, mainly lungs, liver, and bone marrow, and it is also secreted by tumor and stromal cells (134,140). CXCL12 expression levels are highest in these common sites of metastasis, which could recruit cancer cells to these distant organs. CCR7 binds the ligands, CCL21 and CCL19, and guides the migration of lymphocytes and dendritic cells (DC) to lymph nodes (142). Expression of CCR7 in tumor cells has emerged as an important predictor of lymph node metastasis and poor prognosis in cervical cancer, gastric carcinoma, and breast cancer, among others (134).
CCR7 and CXCR4 are the main receptors typically present on metastasizing cells, but there are other chemokine receptors that may dictate more organ specific metastasis. For example, the small intestine is an organ that expresses high levels of CCL25 physiologically in order to guide CCR9+ lymphocytes to this tissue. Since melanoma, breast cancer, and ovarian cancer express high levels of CCR9, this receptor may play a pivotal role in the preferential metastasis of these tumors to the small intestine (143)(144)(145). Additionally, malignant melanoma express high levels of CCR10, a receptor that guides leukocytes to the skin, and consequently CCR10 expression in melanoma may drive metastasis to the skin (146,147). CXCR3 and CXCR5 have both been also shown to play a role in lymph node metastasis (148,149).
With increasing nutrients and oxygen demands by the tumor cell, solid tumors produce angiogenic factors that promote the migration and proliferation of endothelial cells to form new vessels. Many of these factors exert their functions through GPCRs expressed on endothelial cells, including thrombin, prostaglandins, S1P, and many chemokines. In addition, chemokines like CCL2, CCL5, and CXCL8/IL-8, can recruit leukocytes and macrophages to the tumor site, which leads to production of VEGF and other angiogenic factors that contribute to the growth of tumor associated blood vessels (134,135). Production of inflammatory cytokines can also promote new vessel formation by elevating COX2 expression, and in turn prostaglandin E2 (PGE2) increase the expression of VEGF, CXCL8, and CXCL5 by tumor cells (150,151).
In the case of thrombin, this serine proteinase plays a vital role in regulating hemostasis by converting fibrinogen into fibrin to stimulate platelet aggregation and coagulation (152). Thrombin carries out its effects through the PAR family of receptors, which exhibit the unique property of harboring a tethered ligand within the receptor that becomes exposed upon cleavage of the N-terminal extracellular region by thrombin (153). Thrombin promotes angiogenesis by increasing metalloproteinases and decreasing the ability of endothelial cells to adhere to the extracellular matrix (154,155), while promoting the expression and activity of the VEGF receptor, VEGFR-2, through Gα13 and its target RhoA (156). On the other hand, S1P 1 stimulates endothelial cell proliferation, survival, and migration, and also regulates sprouting angiogenesis through crosstalk with VEGFR-2 and enhanced tissue hypoxia and VEGF production (157)(158)(159). The effects of S1P on angiogenesis largely depend on the GPCR it binds (S1P [1][2][3][4][5]. S1P stimulates angiogenesis mainly through S1P 1 and S1P 3 , and mediates endothelial cell migration and formation of capillary structures through Gαi (or more likely its associated Gβγ subunits) activation of the small GTPase Rac1 (160). In contrast, S1P 2 has been shown to be involved in the suppression of angiogenesis most likely through the inhibition of Rac and cell migration (161). Altogether, GPCRs participate in angiogenesis either by promoting the proliferation, migration, and sprouting of endothelial cells, or by the release of pro-angiogenic factors for new blood vessel formation, thereby increasing the blood supply to the growing tumors.

A key role for GPCRs in cancer immunology
In the last few years, cancer immunotherapy became one of the most exciting breakthroughs in cancer treatment. Recent revolutionary discoveries have highlighted the importance of the tumor microenvironment (TME) and its associated immune cells in cancer development and therapeutic resistance. Tumors can deploy multiple mechanisms to avoid immune recognition and an anti-tumor immune response, including the recruitment of myeloid-derived suppressor cells (MDSC) and conditioning of the surrounding microenvironment to become highly immune suppressive by expressing cytokines, such as interleukin (IL)-6 (IL-6), IL-10 and transforming growth factor beta (TGFβ) (162). This can lead to the accumulation of suppressive regulatory T cells (Tregs) and the polarization of macrophages toward an immune suppressive phenotype, which is often referred to as M2 or tumor associated macrophage (TAM) phenotype (163). A key emerging mechanism of tumor immunosuppression involves T cell exhaustion, whereby T cell reactivity is impaired due to activation of T cell checkpoints, including programmed death 1 (PD-1), by its ligand, PD-L1, that is expressed by macrophages and some cancer cells, restraining T cell activation (164)(165)(166). Together, these conditions contribute to the suppression of cytotoxic CD8+ T lymphocytes (CTLs) recruitment, survival, and function, and ultimately to the loss of an effective anti-tumor immune response. While the aberrant function and dysregulated expression of GPCRs is now beginning to be linked directly to the tumor itself, the role of GPCRs on immune cells infiltrating tumors is still not fully understood and grossly underappreciated. Given the diversity of GPCRs and the variety of GPCR families, current studies have only scratched the surface of delineating GPCRs on immune cells in cancer. The importance of studying GPCRs in the context of cancer immunology is reflected by the multiple roles that this receptor family plays in inflammation, orchestrating immune cell trafficking, and regulating the tumor microenvironment, as summarized in Figure 6. A crucial first step in anti-tumor immunity is the migration of cytotoxic cells recognizing tumor antigens to the tumor, and this is mediated largely by chemokine receptors.

GPCRs orchestrate immune cell migration and recruitment:
In the 60s, it became clear that chemoattractants can bind and act directly on lymphocytes and from there, GPCRs in the immune system became mostly known for their ability to steer cell migration towards chemokine gradients (167,168). Cytotoxic immune cells, including natural killer (NK) cells, and CD8 T cells that are specific for tumor antigens, are guided to the tumor where they secrete cytotoxic molecules to induce tumor cell death.
CXCR3 on CD8 T cells and NK cells binds the ligands CXCL9 and CXCLl0 to migrate into tumors (reviewed in (169)). Indeed, increased CXCL9 and CXLC10 and tumor-infiltrating CD8 T cells correlates with improved survival and decreased cancer metastasis (170,171). Furthermore, CXCL10, which is also known as interferon induced protein 10 (IP10), as well as CXCL9 and CXCL11 are known to be induced by interferon α, β, and γ, and are part of the 'interferon gene signature" that is often used to predict a favorable response to anti-PD-1 treatment (172). This provides direct evidence highlighting the importance of chemokine and chemokine receptors in the new era of immunotherapies.
The activation of these antigenexperienced cytotoxic CD8+ T cells is driven by DCs that capture cancer cell antigens on their major histocompatibility (MHC) I or MHCII molecules and present them to naïve T cells to drive effector T cell activation, bridging the innate and adaptive immune system (173). DCs are antigen presenting cells that traffic to draining lymph nodes for stimulation and activation of T cells, and their mobilization is largely driven by CCR7 and its ligands CCL19 and CCL21 (174). The endocytosis of apoptotic cells has been shown to induce CCR7 expression and subsequent migration of DCs, and CCR7 mediated activation of Rac1 and Rac2 may have redundant functions in migration to the lymph nodes, as shown by the absence of DC mobilization in Rac1 and Rac2 deficient mice (175). Dendritic cells play a crucial role in immunosurveillance for elimination of cancer cells In addition to DC mobilization and recruitment of cytotoxic immune cells to the tumor, chemokine receptors also participate in promoting tumorigenesis by mediating the recruitment of immunosuppressive immune cells, specifically Treg cells and MDSCs (169). Treg cells express CCR4 and migrate to the tumor in response to CCL22 produced by macrophages and tumor cells (176). Blockade of this receptor with a therapeutic monoclonal antibody (mogamulizumab) or small molecule inhibitors is under current evaluation for cancer treatment when combined with immune-checkpoint inhibitors (177). The hypoxic regions of the tumor microenvironment also generate CCL28, which can recruit CCR10-expressing Tregs (178). The immunosuppressive microenvironment of the bone marrow has been linked to high frequencies of Tregs. Tregs can be mobilized from the bone marrow into the periphery by granulocyte colonystimulating factor (G-CSF), which promotes the degradation of CXCLl2, a ligand for CXCR4 (179). CXCL12 can lead to higher numbers of Tregs in the bone marrow, promoting an immunosuppressive environment that favors the establishment of metastatic niches, which may help explain why many cancers often metastasize to the bone marrow.
Chemokine receptors also direct MDSCs to the tumor thus favoring an immunosuppressive, tumor-promoting environment. Monocytic MDSCs, including tumor associated macrophages (TAMs), are recruited to the tumor by CCL2, CXCL5, and CXCL12 acting on CCR2, CXCR2, and CXCR4 (180). MDSCs have been also shown to secrete ligands for CCR5 to direct the migration of CCR5+ Tregs, though various studies have also shown that CCR5 guides MDSCs themselves as well (181)(182)(183). Moreover, CXCL8 and CXCL1 are often secreted by most solid tumors, and by certain subsets of Tregs, and act on granulocytic MDSCs that express CXCR1 and CXCR2, including neutrophils, to promote their recruitment to tumors. Since neutrophils secrete tumor-and angiogenesis-promoting molecules, CXCL8 is generally thought to contribute to the immunosuppressive environment leading to tumor angiogenesis and progression (169,184).
In summary, a variety of chemokines dictate the recruitment of different immunosuppressive immune cells into the tumor microenvironment, and through these processes, tumors can evolve to avoid detection and destruction by the innate and adaptive immune system. This can, in turn, provide an opportunity to disable this immune evasive mechanisms by targeting chemokine receptors with an increasing repertoire of small molecule inhibitors and negative allosteric modulators (NAMs) (185), and/or with blocking antibodies. These therapeutic strategies are already in use or under clinical evaluation in multiple chronic inflammatory diseases, and their study in the context of cancer immunotherapy will likely represent one of the most exciting areas of future exploration in the GPCR targeting field.

Modulation of immunosuppressive GPCRs by the tumor microenvironment:
The immunosuppressive and hypoxic nature of the tumor microenvironment can also largely influence the function of cytotoxic immune cells and the success of cancer immunotherapies. A driving force behind the malignancy and morbidity of cancer is its ability to proliferate unrestrained, by creating an immunosuppressive environment favoring tumor growth. The nucleoside adenosine is a potent physiologic and pharmacologic regulator that is released from injured and necrotic cells by extracellular breakdown of adenosine triphosphate (ATP) by the action of the ectonucleotidases CD39 and CD73 (reviewed in (186)). Typical extracellular adenosine levels are low, but at injury sites with tissue breakdown and hypoxia, adenosine levels can rise from nano-molar to micro-molar concentrations. Extracellular adenosine can signal through four GPCRs: A 1 , A 2A , A 2B , and A 3 adenosine receptors (ADORA1, ADORA2A, ADORA2B, and ADORA3, respectively) (187). A 1 and A 3 receptors signal through Gαi and lead to decreased cAMP. Activation of A 2A and A 2B receptors, which are expressed on immune and endothelial cells, lead to signaling through Gαs proteins and A 2B can also signal through Gαq (188). Of the four adenosine receptors, A 2A receptor (encoded by the ADORA2A gene) is the predominantly expressed subtype in most immune cells. In general, stimulation of A 2A receptor provides an immunosuppressive signal in T cells (187), NK cells (189), DCs (190), and neutrophils (191). A 2A receptor stimulation interferes with trafficking of T cells and NK cells by desensitizing chemokine receptors and reducing levels of proinflammatory cytokines (192). Blocking the adenosine generating pathway has shown tumor regression in breast cancer, colorectal cancer, and melanoma (reviewed in (186,193)), and small molecule inhibitors of A 2A receptor as well as blocking antibodies anti-CD73 and anti-CD39 are under current evaluation for combination cancer immunotherapies (186,194). While these immunotherapies aim to boost immune cell activity in the immunosuppressive tumor microenvironment, it is also important to consider the effects of tumor-driven inflammation, largely driven by prostaglandins and prostaglandin receptors.

GPCRs link inflammation to cancer immune evasion:
Inflammation occurs as the immune system responds to infection and injury to beneficially remove offending factors and restore tissue structure and physiological function. However, with subsequent tissue injury, cells that have sustained DNA damage or mutagenic assault will continue to proliferate in microenvironments rich in inflammatory cells and growth/survival factors that support their growth. Prostaglandins are a group of physiologically active lipid compounds found in almost every tissue in humans and animals and they play a key role in the generation of an inflammatory response (195). They are enzymatically derived from arachidonic acid by the cyclooxygenase (COX) isoforms, COX1 and COX2, and are powerful vasodilators to (196). PGE2 is the most abundant prostaglandin produced in cancers, and the prostanoid receptor family, which are GPCRs, includes: E prostanoid receptor 1 (EP 1 , PTGER1), EP 2 (PTGER2), EP 3 (PTGER3), and EP 4 (PTGER4). Of these, EP 1 is coupled to Gαq, EP 3 is coupled to Gαi, and both EP 2 and EP 4 are coupled to Gαs (196). PGE2 binding to different EP receptors can regulate the function of many immune cell types, including macrophages, DCs, T cells, and B cells, as will be discussed here.
PGE2 produced by cancer cells has been linked to increased expression of FOXP3 in Treg cells, promoting the immune-suppressive activity of Tregs (197). In addition to Tregs, PGE2 has also been linked to increased recruitment of MDSCs (198), decreased CD8 T cell activation (199,200) and increased expression of inhibitory markers, like PD-1 (198,199,201). PGE2 alters the differentiation, maturation, and cytokine secretion of DCs by upregulating CD25 and indoleaminepyrrole 2,3-dioxygenase (IDO) and decreased expression of CD80, CD86, and MHCI maturation markers (202). Recently, nonsteroidal antiinflammatory drugs (NSAIDs) that block COX2 and/or COX1 and COX2 were found to have beneficial effects on reducing the risk of developing esophageal, stomach, skin, and breast cancers, in addition to their best established function in preventing colorectal cancer (203,204). Hence, EP receptors may represent exciting targets for cancer immune prevention and treatment.

New "oncocrine" GPCRs networks
As outlined above, GPCRs play multiple roles in cancer progression, dissemination, angiogenesis and immune evasion upon their activation by ligands produced by cancer cells or by the multiplicity of cells within the tumor stroma. This is often referred to as autocrine and paracrine signaling, respectively, and to angiocrine signaling when referring to the release of endothelial cell-derived factors, including GPCR ligands, during organ regeneration and homeostasis, and cancer progression (205). However, these terms may not reflect fully the complexity of the cell communication networks deployed by cancer cells and their tumor microenvironment. Indeed, seminal findings have linked GPCRs to the cross talk among multiple cell types to promote the survival and proliferation of cancer cells, instruct the growth of new blood vessels, evade immune surveillance, and metastasize to secondary organs. In turn, targeting GPCRs and their cancer-driving oncocrine networks may provide unique opportunities to disrupt intratumoral cell-cell communication, thereby initiating tumor collapse.
The concept that cell-cell communication circuits mediate GPCR-induced tumorigenesis was well established by our prior studies on a virally encoded receptor expressed by the Kaposi's Sarcoma (KS) virus (KSHV/HHV8). This receptor, known as KSHV vGPCR, is constitutively active and is alone sufficient to initiate angiosarcomas when expressed in endothelial cells and their progenitors (206). However, as for human KS, only few cells within the tumor express KSHV vGPCR, but instead vGPCR expressing cells secrete multiple factors, including VEGF, IL6, IL-8/CXCL8 and IL-1, thereby causing what was termed as "paracrine transformation" of the surrounding cells that express other KSHV genes (such as LANA) (25,207,208).
Perhaps one of the best described oncocrine systems is that initiated by neuropeptides that act as oncogenic autocrine growth factors, such as bombesin-like peptides, including gastrin-releasing peptide (GRP) and neuromedin B (NMB), which bind to bombesin receptors, NMBR and GRPR, respectively (209,210). Most small cell lung cancer (SCLC), prostate, and ovarian cancers express GRP and NMB mRNA, and bombesin-like peptide antagonists reduce the growth of their derived cell lines in vitro (211)(212)(213)(214). Moreover, SCLC and prostate carcinoma cell lines also secrete arginine vasopressin and express vasopressin receptors (encoded by AVPR1A, AVPR1B and AVPR2) and neurotensin receptors (encoded by NTSR1 and NTSR2). Multiple studies have demonstrated the involvement of the cholecystokinin B receptor (CCK 2 ) autocrine system in human pancreatic carcinoma. The mRNA for the receptor was detected in all pancreatic carcinomas and growth of their derived cell lines was inhibited by antagonists blocking CCK 2 (215,216). Most of the neuropeptide receptors act through Gαq proteins, highlighting again the transforming potential of Gαq-GPCRs when activated locally by their ligands.
Remarkably, recent evidence has also demonstrated that oncocrine signaling by GPCRs mediate the well-established link between nerve innervation and growth of many malignancies (217,218). For example, in PDAC, signaling through β2 adrenergic receptor, which can be stimulated by circulating adrenaline and noradrenaline under stress conditions, leads to increased nerve growth factor (NGF) secretion by pre-cancer and cancer cells (219). In this case, neurotrophins, such as NGF, signal through Trk receptors to promote neuron survival and axonogenesis and tumor innervation, initiating a paracrine circuit through the local release of noradrenaline by the nerve ends. This can act on β2 adrenergic receptors in PDAC cells to promote increased growth and therapy resistance (219). A similar nerve-cancer paracrine mechanism, now referred collectively as oncocrine, was recently described in stomach cancer (220). In this case, acetylcholine released from tuft cells and nerves induces NGF release from gastric epithelial cells, leading to enhanced innervation and subsequent release of acetylcholine by the nerves. Acetylcholine stimulates M3 muscarinic receptors (encoded by CHRM3) expressed on stomach precancer cells to promote stomach cancer progression (220). Another interesting example of oncocrine signaling is the autocrine activation of muscarinic receptors in prostate cancer. In prostate cancer tissues, there is simultaneously an increased expression of choline acetyltransferase, the enzyme catalyzing the synthesis and secretion of acetylcholine, and M3 muscarinic receptors (CHRM3) (221). Experimentally, activation of acetylcholine receptors increases prostate cancer proliferation and migration (222). In addition to prostate cancer, muscarinic receptors have also been implicated in SCLC and CRC stimulating cancer cell growth with acetylcholine as an autocrine growth factor (223,224).
Frizzled receptors may also play important oncocrine functions. For example, microglia actively recruit and support the growth of glioblastoma (GBM) and contribute to brain metastasis in a GPCR (WNT/FZD)-dependent manner (225). Together, these findings support a critical and underappreciated role of GPCRs as part of autocrine and paracrine ("oncocrine") mechanisms driving tumor progression, some of which can be readily targeted with existing approved drugs.

Opportunities to target the underexploited GPCR axis for precision cancer therapy
In this new era of precision medicine, we have unique opportunities to intervene pharmacologically to correct faulty signaling circuits while sparing the normal ones. In this context, GPCRs represent the most widely studied receptor family, but their oncogenic signaling pathways are still not fully understood. We hope that the present review highlighting the oncoGPCRome, including mutations, gene copy variations, and expression of GPCRs in each cancer type and cancer-associated immune cells will ignite new efforts targeting this receptor family in cancer. Indeed, GPCRs and G proteins are widely dysregulated in cancer, and yet underexploited in oncology. This is reflected by the fact that only 8 FDA-approved anti-cancer drugs target GPCRs directly (Drugs@FDA: FDA Approved Drug Products). These include: 1) Cabergoline, a small molecule dopamine 1 (D 1 ) receptor inhibitor, which is approved for hyperprolactinemic disorders, either idiopathic or due to pituitary adenomas. 2) Sonidegib and 3) Vismodegib that are small molecule antagonists of SMO and block Shh signaling, and are approved for the treatment of locally advanced BCC that has recurred following surgery or radiation therapy, or BCC patients who are not candidates for surgery or radiation therapy. Vismodegib is also approved for metastatic BCC. 4) Lanreotide, a long-acting analogue of somatostatin, which stimulates somatostatin receptors (SST [1][2][3][4][5], and is indicated for the long-term treatment of acromegalic patients who do not respond to surgery and/or radiotherapy. 5) Degarelix, a gonadotropinreleasing hormone (GnRH) receptor antagonist that reduces testosterone production, is indicated for treatment of patients with advanced prostate cancer. 6) Plerixafor, a small molecule CXCR4 antagonist, is used in combination with G-CSF to mobilize hematopoietic stem cells to the peripheral blood for autologous transplantation in patients with non-Hodgkin's lymphoma and multiple myeloma. 7) Evista (raloxifene), a selective estrogen receptor modulator that also targets GPER1 (G protein-coupled estrogen receptor 30, GPR30, reviewed in (226)) in post-menopausal women for breast cancer prevention, providing a valuable alternative target for estrogen receptornegative breast tumors and also a preventative method less toxic than the standard tamoxifen treatment regimen. 8) Mogamulizumab, a humanized monoclonal antibody targeting CCR4, was approved in August 2018 for mycosis fungoides and Sézary syndrome, two subtypes of cutaneous T-cell lymphoma (CTCL). In Japan, mogamulizumab is also approved for the treatment of relapsed or refractory CCR4+ T-cell leukemia/lymphoma (ATCLL) and CCR4+ cutaneous T cell lymphoma (CTCL). Although CTCL and ATCLL are rare non-melanoma skin cancers, this represents a breakthrough in GPCR targeting using biologics (monoclonal antibodies) rather than GPCR modulation solely based on small molecule inhibitors.
Building on these successes, multiple monoclonal antibodies against GPCRs and/or small molecule inhibitors are now under clinical exploration for multiple cancer indications, and we can expect that new drug discovery efforts may soon enable targeting GPCRs to 1) prevent cancer in at risk patients with premalignant lesions or genetic predisposition, 2) diminish therapy resistance, 3) increase the response to new revolutionary cancer immunotherapies, and 4) prevent tumor relapse in patients that have had a prior cancer, often referred to as cancer survivors.
Regarding the latter, it is estimated that there are currently at least 15 million cancer survivors in the US alone (227). The emerging findings that disruption of GPCR-mediated oncocrine communication between cancer cells, cancer and stromal cells, cancer and immune cells, and cancer and local innervation can prevent cancer progression and metastasis may further support the clinical exploration of repurposing approved drugs as an adjuvant treatment in cancer patients and cancer survivors. For example, a typical β blocker, propranolol, improves prognosis in skin melanoma (228) and decreases resistance to EGFR inhibitors in lung cancer (229). Based on these and additional findings, propranolol is currently under clinical evaluation as an adjuvant treatment in multiple cancer types.
This may just represent the tip of the iceberg, as there is ample experience in using GPCR modulating agents for prolonged periods of time for a myriad of human diseases. Thus, we can anticipate that harnessing the power of patient health "big data" using computational analytic platforms and large epidemiological datasets can provide extraordinary opportunities for GPCRbased treatments in cancer through drug repurposing of approved GPCR-targeting therapies.
Antibodies targeting GPCRs could have broader applications, such as for cancer imaging and early detection, as well as to deliver cytotoxic agents and/or radiosensitizers. For example, the exclusivity of FZD 10 expression in tumor cells over normal tissues was recently exploited to develop anti-FZD 10 antibodies for radioimmunotherapy as a therapeutic β-particle radiation target for synovial sarcoma (230). As more monoclonal antibodies targeting GPCRs are developed, radioimmunotherapy could be utilized to target and destroy cancer cells exclusively expressing these receptors.
Another exciting opportunity is to target GPCRs as part of combination therapies to prevent or reverse cancer therapy resistance. Specifically, a major drawback of the use of targeted cancer therapies is that the prolonged inhibition of key signalling hubs, such as EGFR, PI3K, BRAF, and mTOR, often leads to the activation of an intricate network of negative and positive feedback loops, which may initiate compensatory bypass mechanisms overcoming the growth-suppressive activity of targeted therapies that inhibit oncogenic pathways (231,232). As such, combinatorial inhibition of oncogene-effector and feedback pathways is a promising cancer treatment strategy (233). In this context, we can anticipate that GPCRs may provide unexplored opportunities for combination treatment options. For example, two large omics approaches using a) genome wide open reading frame expression approaches and b) CRISPR/Cas9 activation methods enhancing the expression of endogenous human genes, revealed that GPCRs, G protein, and/or their downstream targets are the most highly represented class of molecules conferring BRAF inhibitor resistance in melanoma cells (234,235). This information, combined with the expression analysis of GPCRs expressed in each cancer type, may facilitate the hypothesis-driven exploration of the role of GPCRs in resistance to targeted agents in melanoma and other malignancies.
Developing effective treatment options for cancers driven by constitutively active mutants of G proteins α-subunits is expected to be more challenging, as it may require the development of cell-permeable agents targeting these G proteins directly or interfering with their oncogenic signaling pathways. Regarding the former, a new generation of natural occurring cyclic depsipeptides, FR900359 (FR) and YM-254890 (YM), which block Gαq have shown encouraging activity in experimental UVM models (236)(237)(238). While providing an exciting proof of principle, likely toxicities associated with general Gαqblockade may limit its therapeutic use. Instead, the use of new computational pipelines have recently revealed that the kinase FAK (encoded by the PTK2 gene) is synthetic lethal with active GNAQ, which means that while FAK inhibition can be tolerated in most normal cells in adults, cancer cells driven by GNAQ depend on FAK for their survival and growth (239). The discovery of additional signaling vulnerabilities for GNAQ and GNAS oncogenes, for example by genome-wide synthetic lethality screens, may soon provide novel precision therapies for the multiple emerging G protein-driven human malignancies and disease conditions.
Overall, we are now witnessing a revolution in our understanding of cancer at the molecular and cellular level. In this context, targeting GPCRs and their oncogenic circuits as single agents or as part of new combination modalities may provide unprecedented opportunities for the development of novel cancer prevention strategies and targeted and immune therapies.   Agonist-activated GPCRs promote the dissociation of GDP bound to the α subunit of heterotrimeric G proteins and its replacement by GTP. Gα and Gβγ subunits can then activate numerous downstream effectors. The 16 human G protein α subunits can be divided into the four subfamilies, and a single GPCR can couple to one or more families of Gα subunits. Downstream effectors regulated by their targets include a variety of second messenger systems (red), GEFs (yellow), and Rho and Ras GTPases (green), which will result in the stimulation of multiple kinase cascades (blue) regulating key cellular functions. These include members of the mitogen activated protein kinase (MAPK), AKT and mTOR, second messenger regulated kinases and phosphatases, and multiple kinases regulated by Rho and Ras GTPases. In addition, Gαs-coupled receptors inhibit and Gα12/13-, Gαi-, and Gαq/11-coupled receptors activate the transcription coactivator YAP and its related protein, TAZ, the most downstream targets of the Hippo kinase cascade, as well as β-catenin and the sonic hedgehog (Shh) pathway, among others. Ultimately, these large number of effector molecules can have multiple effects in the cytosol and nucleus to regulate gene expression, cell metabolism, migration, proliferation, and survival by GPCRs, which can contribute to normal and malignant cell growth. See text for details.  To identify these, mutated receptor sequences were aligned (using PFAM 7tm_1 Hidden Markov Model) and Ballesteros/Weinstein numberings assigned (see Supplemental Table 3). Conserved functional motives are highlighted and labelled. From GISTIC analysis, the proportion of TCGA cohorts (sample number) with highly significant (GISTIC q-value < 0.05 and mRNA correlation > 0.333) CNVs in genes encoding GPCRs (black) and G proteins (red) are shown. The significant genes for each cohort are plotted outside of the cohort pie; cohorts are colored based on number of significant genes and amplification is denoted by a red highlight while deletion is denoted by a blue highlight.