Combined Inhibition of Gα_q and MEK Enhances Therapeutic Efficacy in Uveal Melanoma Free

Uveal melanoma is the most common intraocular malignancy with approximately 3,000 new cases per year in the United States (1, 2). Metastatic uveal melanoma has a median survival of less than 6 months and a 5-year survival rate of approximately 15% (3). Unlike cutaneous melanoma, where there has been significant progress in targeted therapy and immunotherapy, there remains no effective systemic therapeutic option for advanced uveal melanoma.

Uveal melanoma is characterized by aberrant activation of the heterotrimeric G-protein alpha-q (Gα_q) pathway, which canonically activates phospholipase Cβ (PLCB) and downstream effectors, including inositol 1,4,5-triphosphate (IP₃), diacylglycerol, and protein kinase C (PKC; ref. 4). Approximately 90% of uveal melanomas harbor activating mutations in two homologous α subunits of Gα_q GNAQ and GNA11; the remaining cases harbor activating mutations in the upstream G protein–coupled receptor (GPCR) cysteinyl leukotriene receptor 2 (CYSLTR2; refs. 5–7) and downstream target PLCB4 (8). These mutations are not exclusive to uveal melanoma, as they have been identified in a majority of blue nevi, leptomeningeal melanocytic neoplasms, hepatic small vessel neoplasms, Sturge–Weber syndrome, and in a small subset of cutaneous and mucosal melanomas (7, 9–15). The identification of similar genetic aberrations among these diseases demonstrates the significance of this pathway and the need for effective therapeutic options against it.

The MAPK pathway is an important downstream signaling output of Gα_q; prior studies highlighted the high levels of MAPK activity in the absence of canonical drivers found in cutaneous melanoma (6, 7, 16). Despite this, single-agent MEK inhibitors (MEKi) have failed to provide clinical benefit (1, 17, 18). Recent mechanistic studies identified the RAS guanine exchange factor, RasGRP3, as an essential intermediary between Gα_q and RAS/MAPK signaling (19, 20). Inhibition of RasGRP3 reduced ERK signaling and cell growth further implicating the dependence of MAPK signaling in uveal melanoma (19). Promising preclinical studies have pointed to the combination of a pan-PKC inhibitor with MEKi as a potential therapeutic approach. However, clinical applications of these compounds have been hampered by toxicity from pan-PKC inhibition and therefore the lack of therapeutic window (1, 17, 18, 21, 22). Direct inhibition of mutant Gα_q signaling could potentially circumvent such toxicity and provide an effective therapeutic window and conceivably durable response. The recent development of allele-specific RAS inhibitors revealed mutations in KRAS, another small GTPase, can continue to cycle between the GTP-GDP states, albeit at reduced rates compared with wild-type (WT; ref. 23). In fact, KRAS^G12C-specific inhibitors, similar to YM-254890 (YM), target the GDP-bound state and are efficacious in inhibiting KRAS^G12C-driven cancers (24–26). These studies have shown that direct targeting of a mutant GTPase is not only possible, but highly efficacious in KRAS^G12C-driven cancers.

YM is a naturally occurring cyclic depsipeptide that inhibits platelet aggregation by perturbing Gα_q-mediated Ca²⁺ mobilization (27, 28). YM is an allosteric inhibitor that binds to the hydrophobic cleft between two interdomain linkers of Gα_q, stabilizing the inactive GDP-bound form by hindering the flexibility of the linkers (29). YM inhibition of Gα_q prevents canonical nucleotide exchange; without binding GTP, the protein is unable to activate downstream signaling partners. Studies have shown YM can act as a therapeutic agent for inhibition of platelet aggregation and hypocalcemia (30–32). However, initial studies using overexpression systems indicated YM was unable to inhibit mutant Gα_q^Q209L (27). The lack of inhibition was hypothesized to be due to the reduced GTPase activity of the Gα_q^Q209L mutant preferentially locking the mutant protein in the GTP-bound state similar to the paralogous Gα_s^Q227L mutant (33). Three recent studies have highlighted the use of FR900359 (FR), an analog of YM, to directly target Q209-mutant Gα_q. These studies demonstrated FR's ability to inhibit signaling and growth in Q209-mutant uveal melanoma cell lines, and efficacy against one human xenograft model (34–36). These reports illustrate direct inhibition of Q209 mutant Gα_q as a potential therapeutic avenue for uveal melanoma. However, both YM and FR are potent inhibitors of physiologically active WT Gα_q signaling (27–29, 37), which raises concerns of therapeutic window of YM and FR in targeting mutant Gα_q in uveal melanoma. Thus, discovery of novel synergistic combinations that can potentiate therapeutic efficacy and broaden the therapeutic window of YM or FR is imperative. Furthermore, there remains a need for a comprehensive understanding of direct Gα_q inhibition across the mutational landscape of uveal melanoma.

Here, we identified a novel GNAQ mutational hotspot, G48, and showed exogenous expression of GNAQ^G48V, similar to the canonical GNAQ mutants, transformed melanocytes. We demonstrated YM effectively inhibited cellular growth and downstream Gα_q signaling in four uveal melanoma mutants, GNAQ (G48V, R183Q, Q209L) and CYSLTR2 (L129Q). We further confirmed the efficacy of YM in a series of in vivo experiments with human uveal melanoma cell lines harboring Gα_q^Q209 mutations. Furthermore, transcriptomic and synergistic analysis revealed that together, YM and MEKi led to enhanced and sustained inhibition of MAPK signaling to significantly decrease tumor growth. These results suggest direct Gα_q inhibition provides an effective therapeutic strategy in uveal melanoma and synergizes with MEKi to further increase therapeutic potential.

MSK-IMPACT testing for patients with uveal melanoma was ordered by the treating physician who signed informed consent to research protocol (ClinicalTrials.gov, NCT01775072). All animal studies were performed in accordance to Memorial Sloan Kettering Cancer Center (MSKCC, New York, NY) Institutional Animal Care and Use Committee (IACUC) 11-12-029.

YM was purchased from Wako Pure Chemical Industries (catalog no. 257-00631). MEK162 (binimetinib) was purchased from Array BioPharma Inc. Trametinib was purchased from Selleck Chemicals (catalog no. S2673).

For the MSKCC (New York, NY) uveal melanoma cohort, all patients provided informed consent for tissue procurement and mutational testing, and the study was approved by the institutional review board (ClinicalTrials.gov, NCT01775072). Patients had a confirmed diagnosis of uveal melanoma. A total of 124 specimens from 116 patients were analyzed including 47 specimens from primary uveal melanoma and 77 from metastasis (56 liver, 6 lymph node, 5 lung, 10 other). In the 7 patients with multiple samples, the Gα_q mutational status were identical. Tumor samples were sequenced using MSK-IMPACTv3, v5, or v6 that captures 341, 410, and 468 genes, respectively. The MSK-IMPACT panel includes GNAQ and GNA11 in all versions and CYSLTR2 was added to v6.

For integrative analysis of published cohorts, level 2 whole-exome mutational data were downloaded from the NIH The Cancer Genome Atlas (TCGA) server. Processed whole-exome sequencing data from the University of Duisburg-Essen (UNI-UDE; n = 22) and MD Anderson Cancer Center/Massachusetts Eye and Ear Infirmary (MDACC/MEEI; n = 52) uveal melanoma cohorts and processed whole-genome sequencing data from the Cancer Research UK (CRUK; n = 12) and QIMR (n = 28) cohorts were extracted from the supplementary tables of relevant publications (20–24). For downstream analysis, we merged the data for three duplicate samples present in both the CRUK and TCGA databases and removed three samples from the QIMR database that lacked any somatic mutations, leaving 188 samples. The OncoPrint was generated using MSKCC cBioPortal.

For transient transfection experiments in HEK-293T, we used a synthetic gene for codon optimized CYSLTR2 (5) and WT cDNA for GNAQ (cDNA.org). We performed site-directed mutagenesis to generate mutants Q209L, R183Q, and G48V using QuickChange (Agilent Technologies) site-directed mutagenesis. HEK-293T cells were transiently transfected with the plasmids encoding for GNAQ WT and mutants subcloned into pcDNA3.1(+) using Lipofectamine according to the manufacturer's instructions. Briefly, 7,000 HEK-293T cells were transfected in low-volume 384-well plates with 11 ng of total DNA/well for 24 hours.

For stable expression in melan-a cells, we employed human cDNAs for CYSLTR2 and GNAQ obtained from Origene, mutagenized using QuickChange and cloned into the retroviral vector MSCV-Puro (Addgene plasmid #68469; ref. 38). NRAS^Q61K (Addgene plasmid #49404) and MEKDD (Addgene plasmid #15268; ref. 39) constructs for rescue experiments were purchased from Addgene. Retroviral production was performed through transfection of retroviral expression vector with the pCl-Ampho packaging vector with X-tremeGENE 9 (Sigma-Aldrich). QuikChange primers were used to introduce the various mutant constructs. All constructs were confirmed by sequencing. Site directed mutagenesis primers are listed in Supplementary Table S1.

Mutation-dependent melan-a cells were generated by transducing cells with cDNAs in MSCV-PURO (retrovirus) and then selecting cells with puromycin (1 μg/mL) for 2 days. 12-O-tetradecanoyl-phorbol-13-acetate (TPA; Sigma-Aldrich) was then withdrawn from cultures. Cells with activating mutations (CysLT₂R^L129Q, GNAQ^{G48V/R183Q/Q209L}, KRAS^G12V, BRAF^V600E, and NRAS^Q61K) continued to proliferate over several passages (>2 weeks). Vector and WT controls were unable to maintain proliferation or pigmentation after a few passages without TPA. All cells were cultured in media containing 10% FBS, l-glutamine (2 nmol/L), penicillin (100 U/mL), and streptomycin (100 μg/mL). Melan-a cells provided by D. Bennett, St. George's Hospital, University of London (London, UK) were cultured in RPMI supplemented with 200 nmol/L TPA unless otherwise noted (40). HEK-293T (from ATCC) and A375 cells were cultured in DMEM and uveal melanoma cells were cultured in RPMI. All cells tested negative for Mycoplasma. Human uveal melanoma cell lines Mel202, OMM1.3, and OMM1 were described previously (19) and MP41 was purchased from ATCC (CRL-3297).

Cell growth and dose–response curves for YM were assayed using CellTiter-Glo 2.0 (Promega) and readout on a Glomax Luminometer (Promega). For growth assays, melan-a cells were seeded in the absence of TPA, 1,000 cells/well in a 96-well plate, and then counted on days 1, 3, and 6. Growth assays were baselined to day 1 readings. For dose–response curves melan-a and human uveal melanoma cells were seeded 1,000–3,500 cells per well in a 96-well plate and treated 24 hours later with YM for 5 days. Data were baselined to vehicle. Growth assays and dose–response curves were analyzed and IC₅₀ values calculated using GraphPad Prism 7.0 software. Data shown are representative of at least three independent experiments with at least three technical replicates.

Cells were seeded in 12- or 6-well plates, media and drug were replenished every 2 days. Cells were washed twice with ice cold 1× PBS, fixed with ice-cold 100% methanol, and stained with 0.5% crystal violet (Sigma) solution in 25% ethanol. Plates were imaged with a GELCOUNT (Oxford Optronix).

Synergy assays were performed in 96-well plates (1,000–3,000 cells/well) for 5 days with trametinib and YM, readout by CellTiter-Glo 2.0. Viability was baselined to vehicle (100%), input into Combenefit software to obtain highest single agent (HSA) synergy and viability plots (41). Data shown are representative of at least three independent experiments with three technical replicates.

For melan-a allograft studies, 2.0 × 10⁶ cells were resuspended in 100 μL of 1:1 mix of RPMI media and Matrigel (BD Biosciences) and subcutaneously injected into 6- to 8-week-old C57BL/6J mice (Jackson Laboratory). For Mel202 xenograft studies, 3.0 × 10⁶ cells were injected into 6- to 8-week-old CB17-SCID mice (Taconic). Tumor sizes were measured twice a week with calipers starting from 3 weeks after graft and were calculated using the following formula: tumor volume = (π × D × d × h)/6, whereby D, d, and h refers to long diameter, short diameter, and height of the tumor, respectively. Treatment began at a tumor size of approximately 100 mm³. Mice were treated with YM (daily intraperitoneal injection), binimetinib (twice daily oral gavage), a combination of both drugs, or a matched vehicle (0.5% DMSO in 1× PBS and 1% carboxymethyl cellulose + 0.5% Tween 80 in ddH₂O, or both, respectively). Mice were treated for 5 days and then given 2 days for recovery for the length of the experiment. Mice were euthanized in response but not limited to the following: tumor ulceration, tumors located too close to the trunk of the mice to impede movement and blood flow, and tumor burden. For all experiments, mice were grafted double flank, except Fig. 6A (single flank).

OMM1.3 cells were transduced with pBMN (CMV-copGFP-Luc2-Puro) for luciferase expression (Addgene plasmid # 80389; ref. 42). OMM1.3 xenografts were setup the same as Mel202 grafts described above. Treatment was started 3 days after grafting, following the methods described above. Mice were injected with 15 mg/mL d-luciferin potassium salt (GoldBio) in 1× PBS and imaged 15 minutes later using Xenogen IVIS Spectrum. The raw photon flux was calculated using Living Image 4.4 Software and baselined to day 1 readings.

Cell lysates were harvested after indicated treatments as described previously (5, 19). For Western blots of OMM1.3 tumors, mice were treated once (every day) with vehicle, YM, binimetinib, or combo and then scarified 4 hours later. For Western blots of Mel202 tumors, mice were treated with vehicle, YM (every day), binimetinib (twice a day), or combo for a week and then scarified 4 hours after the last treatment. Tumor lysates were generated as described previously (19). Primary antibodies were incubated at a 1:1,000 dilution, unless noted otherwise. Western blot antibodies are listed in Supplementary Table S2 and raw blots are shown in Supplementary Fig. S7.

All tissues were fixed at 4°C overnight in 4% paraformaldehyde. Tissue processing, embedding, sectioning, and hematoxylin and eosin (H&E) staining were performed by Histoserv. Staining was done as described previously (19). IHC antibodies and dilutions are listed in Supplementary Table S3.

RNA was extracted from cells using TRIzol (Invitrogen) following the manufacturer's standard TRIzol extraction protocol. A total of 2 μg of RNA from each sample was reverse transcribed to cDNA using the High Capacity cDNA Reverse Transcription Kit (Applied Biosystems). Power SYBR Green PCR Master Mix (Applied Biosystems) was then used for PCR on a QuantStudio 6 Flex System (Applied Biosystems). Expression was normalized to ribosomal protein RPL27. Relative expression of mRNA was plotted as 2^–ΔΔC_t and each experiment was performed in triplicate and repeated in at least three independent experiments. qRT-PCR primers are listed in Supplementary Table S4.

TCGA uveal melanoma and skin cutaneous melanoma (SKCM) mutational and RNA sequencing (RNA-seq) data were downloaded from cBioPortal (www.cbioportal.org) using the PanCancer Atlas Version. The six SKCM samples with GNAQ or GNA11 Q209L mutations are TCGA-RP-A6K9-06, TCGA-RP-A690-06, TCGA-ER- A3ET-06, TCGA-ER-A3ES-06, TCGA-ER-A2NF-06, and TCGA-ER-A2NF-01. All RNA-seq was performed by MSKCC genomics core facility using poly-A capture. The libraries were sequenced on an Illumina HiSeq-2500 platform with 50 bp paired end to obtain a minimum yield of 40 million reads per sample. The sequence data for all melan-a cells were mapped to the mouse reference genome (GRCm38), whereas, Mel202 and OMM1.3 were mapped to the human reference genome (GRCh38) using STAR v2.330 (43). Gene counts were quantified using STAR to Ensembl gene annotations GRCm38.91 and GRCh38.90 for mouse and human samples, respectively. Counts were r-log transformed using DESeq2 (44). Hierarchical clustering and heatmaps of WT melan-a cells grown in TPA and GNAQ^Q209L and KRAS^G12V transduced melan-a cells grown without TPA was performed on r-log transformed genes with SD>1.5 using Partek Genomics Suite. k-means clustering and heatmaps of drug-treated melan-a and human uveal melanoma cells were performed on r-log transformed, Z-scored (mean = 0, SD = 1), and then partition clustered (k-means) were performed using Partek Genomics Suite. Sum Z-scores were calculated by adding transformed gene expression data from the specified gene signatures. Gene sets are listed in Supplementary Table S5. Cluster and gene set (Hallmark) enrichment analysis (GSEA) were obtained from https://www.gsea-msigdb.org/gsea/index.jsp and are shown in Supplementary Tables S6–S11. Raw and processed data are deposited in Gene Expression Omnibus accession numbers: GSE152705 and GSE160112.

IP₁ concentrations in transfected HEK-293T cells, stably transduced melan-a cells, and uveal melanoma cells were measured using a competitive homogenous time-resolved fluorescence assay (CisBio). Briefly, 7,000 transiently transfected HEK-293T cells, 5,000 melan-a cells, and 5,000 uveal melanoma cells were seeded in low-volume 384-well plates in 7 μL media for 24 hours prior to YM treatment. Cells were then treated with various concentrations of YM at 37°C for 3 or 24 hours. One hour and 45 minutes prior lysis, cells were supplemented with 1× Stimulation Buffer provided by the manufacturer (HEPES 10 mmol/L, CaCl₂ 1 mmol/L, MgCl₂ 0.5 mmol/L, KCl 4.2 mmol/L, NaCl 146 mmol/L, glucose 5.5 mmol/L, LiCl 50 mmol/L, pH 7.4) with 0.2% BSA and 50 mmol/L of LiCl (to prevent IP₁ degradation). Following incubation, cells were lysed by addition of 3 μL/well of d2-labeled IP₁ analog as the fluorescence acceptor and the terbium cryptate–labeled anti-IP1 mAb as the fluorescence donor, diluted in the kit-supplied lysis buffer. The plates were incubated overnight at room temperature and time-resolved fluorescence signals were read using the BioTek Synergy NEO plate reader (BioTek Instruments) at 620 and 665 nm. Results were calculated as a 665 nm/620 nm signal ratio, and IP₁ concentrations were interpolated from a standard curve prepared using the supplied IP₁ calibrator. Results are shown as IP₁ (nmol/L). Dose–response curves and bar graphs were analyzed using GraphPad Prism 7.0 software.

All statistical comparisons were done with Graphpad Prism 7.0 software and used a two-tailed Student t test for comparison between groups. Data are shown as the mean ± SEM (unless otherwise noted) from triplicate samples from at least three independent experiments. P less than 0.05 was used to designate significance. Significant differences between groups are indicated by ns, P > 0.05; *, P < 0.05; **, P < 0.005; ***, P < 0.0005; ****, P < 0.0001.

We examined the Gα_q pathway mutations from a series of 116 consecutive patients with uveal melanoma seen at MSKCC (New York, NY) and who had undergone clinical sequencing on the MSK-IMPACT platform (45). Similar to previously reported cohorts, 107 of 116 patients harbored mutually exclusive activating mutations in either GNAQ or GNA11 and 2 patients had a CYSLTR2^L129Q mutation (Fig. 1A; ref. 5). For GNAQ, in addition to previously known recurrent mutations at amino acids Q209 and R183, one sample harbored a novel G48L mutation. To further evaluate the GNAQ^G48 mutation, we next integrated 188 patients with uveal melanoma from five published whole-exome or whole-genome uveal melanoma cohorts (8, 46–49). We identified 2 patients with GNAQ^G48L and 1 patient with a GNAQ^G48V mutation, indicating G48 as a third mutational hotspot in GNAQ (Fig. 1B). The codon G48 resides in the phosphate-binding loop (P-loop) of Gα_q and is paralogous to codon G12 in RAS GTPases, which is frequently mutated in various cancers (23). Furthermore, the paralogous G47V mutation in GNAS exhibits constitutive activity (33). Structural studies of active Gα_q illustrate G48 of the P-loop, R183 of switch I, and Q209 of switch II are in spatial proximity adjacent to the nucleotide-binding pocket (Fig. 1C; ref. 50), indicating mutations in G48, like Q209 and R183 would hinder GTPase activity.

To determine the role of distinct uveal melanoma driver mutations in a genetically defined context, we stably expressed human cDNAs encoding CYSLTR2^L129Q, GNAQ^G48V, GNAQ^R183Q, GNAQ^Q209L, WT controls, as well as BRAF^V600E, KRAS^G12V, and NRAS^Q61K in melan-a cells (Supplementary Fig. S1A and S1B; ref. 40). BRAF^V600E, KRAS^G12V, and NRAS^Q61K hyperactivate the RAS-RAF-MEK-ERK kinase (MAPK) signaling pathway, distinct from the GPCR-Gα_q-PLCβ signaling axis. Melan-a cells are immortalized mouse melanocytes that require media supplemented with TPA, a diacylglycerol analog, for continued proliferation and are characterized by pigmentation and melanocytic morphology (5, 40). After TPA withdrawal, melan-a cells expressing empty vector, CYSLTR2^WT or GNAQ^WT lost pigmentation and eventually growth arrested (Fig. 1D and E; Supplementary Fig. S1C). However, cells expressing activating mutations of GNAQ and CYSLTR2 exhibited TPA-independent growth and enhanced melanocytic features (e.g., dark pigmentation, enlarged melanosomes). Interestingly, expression of BRAF^V600E, KRAS^G12V, and NRAS^Q61K that hyperactivate MAPK signaling also conveyed TPA independence but failed to maintain pigmentation (Fig. 1D and E; Supplementary Fig. S1C and S1D).

We next examined the ability of Gα_q to stimulate PLCβ and generate IP₃, by measuring accumulation of the IP₃ degradation product IP₁. Melan-a cells expressing CYSLTR2^L129Q and all three GNAQ mutations exhibited enhanced IP₁ accumulation, whereas the CYSLTR2^WT-, GNAQ^WT-, BRAF^V600E-, KRAS^G12V-, and NRAS^Q61K-expressing cells did not, indicating that this cellular system faithfully recapitulates the distinct signaling pathways driven by CysLT₂R and Gα_q oncoproteins (Fig. 1F; Supplementary Fig. S1E). Western blot analysis of melan-a cells following TPA withdrawal showed that uveal melanoma–associated CysLT₂R and Gα_q oncoproteins maintained expression of melanocyte markers MITF, c-KIT, TRP2/DCT, and RASGRP3 (uveal melanoma specific), whereas WT CYSLTR2^WT, GNAQ^WT, KRAS^G12V, and NRAS^Q61K did not (Fig. 1G; Supplementary Fig. S1F). In TCGA datasets, we found these genes (MITF, KIT, DCT, and RASGRP3) were expressed higher in uveal melanoma and Gα_q-mutated SKCM compared with RAS/RAF-mutated SKCM (Supplementary Fig. S1G), suggesting that the engineered melan-a cells capture important oncogene-specific biology of human melanomas (5, 19).

To determine the transcriptome response to expression of activated Gα_q and RAS/RAF pathway in melan-a cells, we performed RNA-seq in WT melan-a cells grown in TPA, melan-a–expressing GNAQ^Q209L, and melan-a–expressing KRAS^G12V cells after TPA withdrawal. Unsupervised hierarchical clustering showed WT cells in TPA were more similar with GNAQ^Q209L cells than KRAS^G12V cells, consistent with direct Gα_q signaling to PLCB to generate diacylglycerol and IP₃ (Supplementary Fig. S1H). To determine pathways activated in these melan-a lines, we performed enrichment analysis of gene ontology, Hallmark, and KEGG gene sets as well as custom gene sets defined by genes differentially expressed between human Gα_q versus BRAF/NRAS/KRAS-mutated melanoma from TCGA uveal and cutaneous melanoma datasets (46, 51) and by genes differentially expressed in genetically engineered mouse melanoma in Bap1^KO;Gna11^Q209L versus Bap1^KO;BRAF^V600E mice (19). We found that GNAQ^Q209L melan-a cells had increased expression of genes in human Gα_q-mutated melanomas and mouse Bap1^KO;Gna11^Q209L tumors while KRAS^G12V melan-a cells had increased expression of genes in human RAS/RAF-mutated tumors and mouse Bap1^KO;Braf^V600E tumors (Supplementary Fig. S1I). In addition, gene sets including calcium-mediated signaling and melanogenesis were strongly enriched in GNAQ^Q209L cells, whereas MAPK signaling was enriched in KRAS^G12V cells (Supplementary Fig. S1I; ref. 19). These data indicate Gα_q activation generates a distinct oncogenic phenotype that maintains melanocyte lineage specification, and is consistent with previous observations in genetically engineered murine models of uveal melanoma (19, 52). This system of engineered mutant oncoprotein-transformed melanocytes allows for context-relevant systematic evaluation of Gα_q-pathway activating mutations that are not available in human cancer cell lines or patient-derived xenograft uveal melanoma models.

We next determined the ability of YM to inhibit the activity of distinct uveal melanoma mutations. Because YM is thought to stabilize the GDP-bound state of Gα_q^WT, we hypothesized CysLT₂R^L129Q would be particularly sensitive to YM. First, we tested the ability of YM to inhibit Gα_q signaling in a panel of HEK-293T cells transfected with oncoprotein cDNA constructs and assayed the dose response of IP₁ accumulation 24 hours following treatment of YM. In this system, CysLT₂R^L129Q and Gα_q^G48V appeared to be most sensitive to YM with subnanomolar potency (Supplementary Fig. S2A). Gα_q^R183Q was approximately 20-fold less sensitive to YM compared with CysLT₂R^L129Q and Gα_q^G48V, whereas Gα_q^Q209L was the least sensitive to YM and its signaling was incompletely inhibited within 24-hour YM treatment, consistent with previous reports (Supplementary Fig. S2A; ref. 27).

Transient transfection experiments in HEK-293T cells result in nonphysiologic overexpression of oncoproteins and can lead to misleading observations. In the more physiologic context of Gα_q pathway mutant-transformed melan-a cells, when we assayed IP₁ after 3 hours of treatment, rapid reduction of IP₁ was observed with uveal melanoma oncoproteins except Gα_q^Q209L, which achieved only half-maximal inhibition (Fig. 2A). Interestingly, 24-hour treatment with YM completely inhibited IP₁ accumulation across all uveal melanoma–activating mutants, including Gα_q^Q209L. CysLT₂R^L129Q and Gα_q^G48V were most sensitive to YM, whereas Gα_q^R183Q and Gα_q^Q209L mutants were approximately 10-fold less sensitive (Fig. 2A). These data suggest activating Gα_q mutants, while prominently GTP bound, undergo GTP hydrolysis, albeit presumably at decreased levels compared with WT, allowing for YM to bind in the Gα_q GDP-state, reminiscent of covalent KRAS^G12C inhibitors (24–26, 29).

We next assayed the effect of YM on downstream Gα_q signaling, including RASGRP3 phosphorylation by PKC, MAPK signaling by CRAF, MEK, and ERK phosphorylation and cyclin D1 expression that integrates signaling to promote cell-cycle progression. YM inhibited these downstream signaling targets, with slower kinetics in Gα_q^R183Q- and Gα_q^Q209L-expressing cells compared with CysLT₂R^L129Q- and Gα_q^G48V-expressing cells (Fig. 2B), consistent with the IP₁ biochemical assay. Because TPA is a diacylglycerol analog and acts downstream of Gα_q, supplementing it in the media restored downstream signaling targets 24 hours after YM treatment (Fig. 2B). YM was ineffective in inhibiting MAPK signaling in BRAF^V600E-, KRAS^G12V-, and NRAS^Q61K-expressing cells (Supplementary Fig. S2B), confirming the specificity of YM to Gα_q inhibition. Expressing KRAS^G12V in CysLT₂R^L129Q and Gα_q^Q209L melan-a cells rescued YM-mediated MAPK inhibition (Supplementary Fig. S2E). We next examined whether YM could effectively inhibit TPA-independent cell growth in the engineered melan-a cells. Consistent with IP₁ accumulation, all uveal melanoma–activating mutants were highly sensitive to YM. CysLT₂R^L129Q and Gα_q^G48V cells exhibited low-nanomolar sensitivity to YM, and the Gα_q^R183Q- and Gα_q^Q209L-expressing cells were modestly less sensitive, whereas the BRAF^V600E-, KRAS^G12V-, and NRAS^Q61K-expressing cells were insensitive (Fig. 2C; Supplementary Fig. S2C). TPA supplementation or KRAS^G12V expression rescued YM-mediated growth inhibition in CysLT₂R^L129Q-expressing cells (Supplementary Fig. S2D).

We performed pilot pharmacokinetic studies suggesting YM has a short serum half-life in mice (data not shown). However, given its potency, YM may engage and inhibit Gα_q more durably. We thus performed a pharmacodynamic study examining downstream signaling in CysLT₂R^L129Q cells allografted into C57BL/6J mice. After a single dose of YM given at 7.5 mg/kg intraperitoneally, we observed inhibition of downstream signaling in CysLT₂R^L129Q tumors as early as 30 minutes (Fig. 2D). By 24 hours, a sustained decreased in ERK phosphorylation and cyclin D was observed despite rebound of CRAF phosphorylation. We observed a paradoxical increase in RASGRP3 phosphorylation levels which may indicate feedback activation in this model.

Given these data, we proceeded to dose YM daily and assessed efficacy. We observed YM significantly inhibited tumor growth of CysLT₂R^L129Q or Gα_q^Q209L allografts (Fig. 2E and F). Importantly, YM treatment did not cause adverse effects such as weight loss (Supplementary Fig. S2F and S2G). These data suggest YM is a highly selective inhibitor against Gα_q signaling and can effectively inhibit proliferation both in vitro and in vivo.

Given the sensitivity of all three activating hotspot mutations to YM, including the most common GNAQ^Q209L mutation, in the melan-a system, we proceeded to study the effect of YM on a panel of human uveal melanoma cell lines that harbor GNAQ^Q209L/P or GNA11^Q209L mutations and A375 (BRAF^V600E) cutaneous melanoma cell line as control. Uveal melanoma cells exhibited high basal activity of IP₁ accumulation, which was inhibited upon YM treatment at 24 hours (Fig. 3A; Supplementary Fig. S3A). Analysis of downstream signaling showed RASGRP3, MAPK output, and cyclin D1 proteins were inhibited in uveal melanoma cells, but not in A375 cells, upon treatment with YM (Fig. 3B; Supplementary Fig. S3B). Expressing NRAS^Q61K or MEK^DD in uveal melanoma cells completely rescued YM-mediated MAPK inhibition (Supplementary Fig. S3G). In OMM1.3 and Mel270 cells, treatment with YM rapidly inhibited the MAPK pathway while suppression of the intermediary target, p-RASGRP3, required prolonged drug exposure. In contrast, YM treatment in Mel202 cells showed potent inhibition of p-RASGRP3 at early time points, but the MAPK pathway was only mildly inhibited; p-ERK1/2 rebounded by 24 hours. These results indicate uveal melanoma cells may have different wiring downstream of Gα_q. Regardless, YM potently and effectively inhibited cell viability in Gα_q-mutant uveal melanoma cell lines but not A375 cells (Fig. 3C; Supplementary Fig. S3C and S3D). TPA supplementation, NRAS^Q61K, or MEK^DD expression rescued YM-mediated growth inhibition in uveal melanoma cells (Supplementary Fig. S3C, S3E, and S3F).

We next used a uveal melanoma xenograft model to test the efficacy of YM in vivo. In mice with xenografts of OMM1.3 cells, YM inhibited tumor formation compared with vehicle and no adverse weight loss was observed (Fig. 3D; Supplementary Fig. S3H). Taken together, these data indicate that Gα_q inhibition is effective against human uveal melanoma xenografts in vivo, in agreement with previously reported data using FR (53).

To gain further mechanistic insight, we compared the transcriptional response with YM with that of trametinib treatment for 4 or 24 hours in the CysLT₂R^L129Q melan-a cells. Perturbed genes were partitioned into six clusters by k-means clustering algorithm (Fig. 4A). Genes downregulated by either YM or trametinib behaved similarly to one another (clusters 1, 2, 3), whereas upregulated genes were distinct (clusters 4, 5, 6). We then computed the top five GSEA HALLMARK gene set overlaps with each cluster to better understand the impact of treatment on these clusters (Supplementary Fig. S4A). YM, similar to trametinib, effectively inhibited the cell-cycle progression (E2F targets) of CysLT₂R^L129Q cells at 24 hours (cluster 1; Fig. 4B). However, trametinib more potently suppressed KRAS and inflammatory signaling than YM (cluster 2; Fig. 4B; Supplementary Fig. S4A). To further understand the suppression of KRAS genes, we used a 52-gene set comprised of genes rapidly downregulated by MEK inhibition in melanoma cells (54), PRATILAS_MAPK signature. This gene signature was downregulated by both YM and trametinib at 4 hours, but by 24 hours, genes rebounded with YM treatment while trametinib further suppressed this gene signature (Fig. 4C). Specifically, this rebound pattern was seen in MAPK signaling output genes DUSP6 and SPRY2/4 (Fig. 4D). This suggests that YM does not durably suppress MAPK signaling at 24 hours, potentially limiting therapeutic effect.

This prompted us to ask whether YM could be used in combination with MEKi to prevent potential MAPK rebound and induction. In our melan-a cellular system, we found significant synergy in an HSA assay between YM and trametinib in cells dependent on CYSLTR2 or GNAQ mutants but not in melan-a cells still dependent on TPA (Fig. 4E; Supplementary Fig. S4B). Long-term treatment of YM plus trametinib greatly reduced proliferation of CysLT₂R^L129Q cells at 26 days compared with single-agent treatments (Fig. 4F). In melan-a cells with TPA, only trametinib showed long-term effects on proliferation. Combination treatment in CysLT₂R^L129Q melan-a cells led to rapid MAPK signaling inhibition that was sustained for 72 hours, whereas p-ERK had returned to basal levels in single treatment groups (Fig. 4G). To further investigate these findings, we used uveal melanoma cell lines to assess the effect of YM and MEKi on viability, transcriptomics, and signaling. We performed HSA synergy assays on four uveal melanoma cell lines and found significant synergy between YM and trametinib (Fig. 5A; Supplementary Fig. S5A). Growth assays in these uveal melanoma cell lines showed variable sensitivity to either YM or MEKi, but combination of the two drugs lead to fewer colonies than single agents in all uveal melanoma cells (Fig. 5B). However, BRAF^V600E-mutant A375 cells did not show synergy and only showed sensitivity to MEKi and combination with YM did not lead to a decrease in colonies (Supplementary Fig. S5B and S5C).

To better understand the synergistic effect of the combination, we performed RNA-seq in two human uveal melanoma cell lines treated with vehicle, single-agent YM or trametinib, or a combination of both for 24 hours (Supplementary Fig. S5D and S5E). Similar to the melan-a cells, we observed trametinib suppressed the PRATALIS_MAPK gene signature with a greater effect than YM (Fig. 5C). Combination resulted in more dramatic gene expression changes when compared with single agent or vehicle and also led to greater inhibition of MAPK, cell cycle, and MYC gene sets, with the exception of cell cycle in Mel202 cells (Fig. 5C; Supplementary Fig. S5D and S5E). Specifically, combination led to greater suppression of MAPK signaling output genes DUSP6 and SPRY2/4 (Fig. 5D). This transcriptomic data indicate that YM and trametinib have very similar effects on gene expression; however, combination of the drugs further enhances the effects on gene expression.

We next determined the effects of YM, trametinib, and combination treatment on signaling. In Mel202 cells, YM treatment alone inhibited RASGRP3, CRAF, and MEK phosphorylation at all time points but downstream ERK and cyclin D inhibition was modest (Fig. 5E). Trametinib treatment alone inhibited ERK phosphorylation at 24 hours with rebound by 48 hours, and combination treatment sustained robust inhibition at all time points. Similarly, in OMM1.3 and MP41 cells, YM single agent was very effective at inhibiting RASGRP3, CRAF, and MEK phosphorylation but ERK phosphorylation rebounded by 72 hours (Fig. 5F and G). Combination treatment led to sustained inhibition of ERK phosphorylation and cyclin D. Expression of KRAS^G12V in OMM1.3 cells was sufficient to rescue YM-mediated p-ERK inhibition, but not trametinib- or combination-mediated p-ERK inhibition (Supplementary Fig. S5F). These data suggest that Gα_q and MEK inhibition are more effective against uveal melanoma cell growth and signaling than single agents, indicating that combination therapy may have an increased clinical benefit.

We tested the combination of YM with an FDA-approved MEKi binimetinib in vivo. Using the Gα_q^Q209L melan-a allograft model, we tested the combination on immunocompetent C57BL/6J mice treated over the course of 20 days. Mice treated with YM or binimetinib alone had a significant inhibition in tumor growth, with about a 2-fold increase compared with day 1 size (Fig. 6A). However, combination treatment led to significant tumor reduction compared with vehicle or single agents (Fig. 6A). Similarly, we found that the combination greatly reduced tumor growth in the CysLT₂R^L129 allograft model. (Fig. 6B). No treatment group resulted in weight loss (Supplementary Fig. S6A and S6B). These experiments show that together, YM and MEKi further reduced tumor burden in allograft models driven by two different activating mutations found in patients.

In the OMM1.3 luciferase xenograft model, single-agent treatments did inhibit growth, but combination treatment decreased tumor luminescence by 1,000-fold, and we were unable to identify the grafts at the end of treatment (Fig. 6C; Supplementary Fig. S6C). OMM1.3 tumors from mice treated for 4 hours showed that combination potently inhibited both PKC and MAPK signaling (Fig. 6D). Single-agent binimetinib inhibited p-ERK but increased p-MEK, typical of MAPK feedback (55), whereas YM treatment potently inhibited RASGRP3 and MAPK signaling despite minimal changes in ERK phosphorylation (Fig. 6D). IHC of these OMM1.3 tumors showed no change in Ki67 staining, most likely due to the short treatment time, but did show a dramatic decrease in p-S6R staining (Supplementary Fig. S6E). We also tested this combination in Mel202 xenografts, which exhibit much slower in vivo growth. YM and binimetinib both showed a tumor reduction of about 30% whereas combination led to a greater than 75% average tumor reduction (Fig. 6E; Supplementary Fig. S6D). Tumors from Mel202 xenografted mice, treated for 1 week, displayed low basal levels of activated RASGRP3, CRAF, and MEK but did have measurable p-ERK levels (Fig. 6F). YM treatment decreased p-RASGRP3 and downstream ERK signaling, while p-ERK remained unchanged, whereas binimetinib potently inhibited p-ERK and downstream targets but caused rebound in p-RASGRP3, p-CRAF, and p-MEK. Combination treatment suppressed the reactivation of p-MEK and p-CRAF compared with single-agent binimetinib and lead to more potent inhibition of p-ERK and p-P90RSK. IHC of these Mel202 tumors show residual Ki67 and p-S6R staining and no significant p-ERK staining in combination-treated mice (Fig. 6G), which corroborates the Western blot analysis. These in vivo data indicate that combination treatment in mice is feasible, highly active, and more efficacious than single-agent YM or MEKi.

Advances in targeted therapy have shown that direct inhibition of mutated oncoproteins, such as EGFR, KIT, KRAS, and BRAF can be highly efficacious, due to the high selectivity against the oncoprotein over WT. Inhibition of downstream signaling is more challenging and often limited by on-target toxicity and narrow therapeutic window. Uveal melanoma is molecularly defined by mutational activation of the Gα_q pathway and harbor low mutational burden suggesting effective direct Gα_q inhibition could have high efficacy. Compared with kinases, rational design of drugs against mutant GTPases, such as Gα_q and RAS, have been challenging for a number of reasons: activating mutations in GTPases are enzymatically impaired whereas those in kinases are enzymatically hyperactive; GTPases activate downstream effectors through protein–protein interactions whereas kinases activate effectors through enzymatic modification; and the abundance of intracellular GTP also makes it difficult to design high-affinity inhibitors that can directly access and bind to the GTP-binding pocket (56). Therefore, GTPases cannot be targeted through inhibition of enzymatic activity but require allosteric drugs that affect conformation or effector binding. Nature has evolved strategies that target human GTPases, including brefeldin A that inhibits ARF, pertussis toxin inhibits Gα_i, and YM from bacteria and FR from the plant Ardisia crenata inhibit Gα_q. In each case, the inhibitor functions through stabilization of the GDP-bound inactive or transitional states, suggesting that active cycling is critical for inhibitor function.

The varying susceptibility to YM across our panel of Gα_q mutants underscores the biochemical differences of these mutants. These novel findings have strong parallels to mutant RAS proteins. Structurally, RAS and Gα_q share homologous mutational “hotspots”: Q61/Q209 and G12/G48, respectively. Interestingly, codon Q61 mutations in KRAS are thought to be more active than those at codon G12, with higher rates of intrinsic nucleotide exchange and a greater reduction of intrinsic hydrolysis (23). This is consistent with our observations of increased IP₁ output for Gα_q^Q209L and could also explain the differences in YM sensitivity we see between Gα_q mutants. The GDP-bound state of Gα_q^Q209L may be short-lived compared with Gα_q^G48V or WT Gα_q in CysLT₂R^L129Q cells, making it less susceptible to inhibition by YM. While YM effectively inhibits the panel of mutants both in vitro and in vivo, these biochemical differences require foresight clinically.

As observed in RAS-mutant tumors, allele-specific inhibitors show limited efficacy as monotherapies and will require combinations with other inhibitors (23). As oncogenic mutations in uveal melanoma activate RAS proteins through RasGRP3, we hypothesize resistance mechanisms to direct inhibitors of Gα_q could occur through reactivation of the MAPK pathway. In fact, compared with trametinib, we observed YM fail to fully suppress the MAPK gene signature. Therefore, the combination strategy of YM plus MEKi could prevent reactivation and durably suppress MAPK signaling. This strategy may also circumvent some of the disadvantages of targeting WT Gα_q, which allows for lower doses of YM and MEKi and broadens the therapeutic window for maximal clinical benefit.

Our work and recent studies highlight the possibility of directly targeting oncogenic Gα_q signaling characteristic of uveal melanoma and the importance of pursuing this strategy further (34–36). One important limitation of this study and others is the lack of testing Gα_q inhibition in liver metastasis models of uveal melanoma. Although these models are technically difficult and typically require intrasplenic injection of cells, it is important to understand Gα_q inhibition efficacy in the liver tumor microenvironment. Recently, complete synthesis of YM as well as FR, and novel analogs of each have been reported (37, 57–60) and a large toolbox of Gα_q inhibitors will likely soon be available.

In summary, we showed cells harboring activating mutations at one of three residues in GNAQ, as well as WT Gα_q driven by CYSLTR2^L129Q were exquisitely sensitive to YM treatment. Transcriptomic and synergy analysis revealed combination of YM with MEKi provides an efficacious and durable response. Our work demonstrates that combination of YM and MEKi leads to enhanced reduction in tumor growth and signaling in uveal melanoma, making it an ideal treatment strategy to pursue clinically.

T.D. Hitchman reports grants from NIH/NCI during the conduct of the study. A.R. Moore reports personal fees from Genentech Inc. outside the submitted work, and is a stockholder of Revolution Medicines, AstraZeneca, and Sanofi. A.N. Shoushtari reports grants and personal fees from Bristol Myers Squibb and Immunocore; personal fees from Castle Biosciences; and grants from Xcovery, Novartis, and Pfizer outside the submitted work. J.H. Francis reports other from NCI, Cancer Center Support during the conduct of the study. B.S. Taylor reports grants from Genentech, Inc, as well as personal fees from Boehringer Ingelheim and Loxo Oncology at Lilly outside the submitted work. P. Chi reports grants and personal fees from Deciphera; personal fees from Exelixis and Zai Lab; and grants from Array/Pfizer, NIH/NCI, FDA/OPD, and NTAP/Bloomberg Fundation outside the submitted work. Y. Chen reports other from ORIC Pharmaceuticals during the conduct of the study; in addition, Y. Chen has a patent for methods and compositions for treating activated G_q mutant cancers or melanocytic malignancies pending. No disclosures were reported by the other authors.

T.D. Hitchman: Conceptualization, formal analysis, supervision, funding acquisition, validation, investigation, visualization, methodology, writing-original draft, project administration, writing-review and editing. G. Bayshtok: Validation, investigation, methodology. E. Ceraudo: Validation, investigation, methodology. A.R. Moore: Methodology, writing-review and editing. C. Lee: Investigation, methodology. R. Jia: Investigation, methodology. N. Wang: Resources, investigation, methodology. M.R. Pachai: Resources, methodology. A.N. Shoushtari: Resources, investigation, methodology. J.H. Francis: Resources, methodology. Y. Guan: Data curation, software, formal analysis, methodology. J. Chen: Data curation, software, formal analysis, methodology. M.T. Chang: Data curation, software, formal analysis. B.S. Taylor: Conceptualization, resources, data curation, software, formal analysis, supervision, funding acquisition, project administration. T.P. Sakmar: Conceptualization, resources, supervision, funding acquisition, project administration, writing-review and editing. T. Huber: Conceptualization, resources, data curation, software, formal analysis, supervision, funding acquisition, investigation, visualization, methodology, project administration, writing-review and editing. P. Chi: Conceptualization, resources, supervision, funding acquisition, investigation, project administration, writing-review and editing. Y. Chen: Conceptualization, resources, data curation, software, formal analysis, supervision, funding acquisition, investigation, visualization, methodology, project administration, writing-review and editing.

This work was supported by the Ruth L. Kirschstein National Research Service Award (NRSA) Individual Predoctoral Fellowship from the NCI (1F31CA236030-01A1, to T.D. Hitchman), MSKCC Support Grant/Core Grant (P30 CA008748), and grants from the NCI (K08CA140946, to Y. Chen; R01CA193837, to Y. Chen; P50CA092629, to Y. Chen; P50CA140146, to P. Chi; K08CA151660, to P. Chi; DP2 CA174499, to P. Chi), U.S. DOD (W81XWH-10-1-0197, to P. Chi), Prostate Cancer Foundation (to Y. Chen), Starr Cancer Consortium (to Y. Chen and P. Chi), Geoffrey Beene Cancer Research Center (to Y. Chen and P. Chi), Gerstner Family Foundation (to Y. Chen), Bressler Scholars Fund (to Y. Chen), and Cycle for Survival (to Y. Chen). B.S. Taylor is supported by R01 CA204749 and R01 CA245069. E. Ceraudo was supported by the Francois Wallace Monahan Fellowship and the Thomas Haines Fellowship. We thank the Robertson Therapeutic Development Fund for partial support.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.