Fig 2: Mathematical model predicts potential for major differences in downstream activation for GNAQ and CYSLTR2 mutations.a Pie chart of mutations in the pathway involving CYSLTR2/GNAQ/GNA11/PLCB4 for 80 UM patients reported in previous studies53,54. b Schematic of oncogenic Gαq and oncogenic CysLT2R signaling in the model. The schematized G-protein alpha subunits (red circles) can represent either (or both) wild-type and mutant forms. c Sobol total sensitivity index for the difference in active TRIO and PLCβ between the GNAQ Q209L and CYSLTR2 L129Q mutant settings based on simulation of the mathematical model. See the supplement for details on the parameters varied and ranges. d Schematic highlighting two diverging behaviors suggested from the extended simulations of the model across parameter space: (1) that CysLT2R mutant networks are less able to activate TRIO/YAP/TAZ than Gαq/11 mutant networks, or (2) that CysLT2R mutant networks are less able to activate PLCβ/PKC/ERK than Gαq/11 mutant networks. Which situation occurs will depend on where the actual system is within parameter space, and could be determined by experimentally measuring parameters and/or by experimentally measuring cellular phenotypes. The bar charts on the right display representative sample cases of the two diverging behaviors; fold change in activation relative to the WT case is used as a measure of pathway activation within these bar charts.

Fig 3: Experimental determination that oncogenic CYSLTR2 is impaired at activating the FAK/YAP/TAZ pathway.a Gene set enrichment analysis (GSEA) of differentially expressed genes for GNAQ Q209L and CYSLTR2 L129Q transfected 293T cells, each compared with mock transfected cells. b GSEA of differentially expressed genes between the GNAQ Q209L and CYSLTR2 L129Q transfected conditions for the same cells as in (a). c Total cell lysates were probed by western blot for phospho-ERK, phospho-YAP, phospho-FAK, Gαq, and GAPDH. Blots are representative of n = 3 experimental replicates. d Schematic of the experimentally observed differences in the relative strength of downstream activation for the Gαq Q209L and CysLT2R L129Q mutations, which identifies which of the two possible diverging behaviors described in Fig. 4d is experimentally demonstrated to occur here. The schematized G-protein alpha subunits (red circles) can represent either (or both) wild-type and mutant forms.

Fig 4: Experimentally informed re-evaluation of parameter sets and parameter ranges.a Normalized distributions of parameter values: the bias of the Gαq Q209L mutant binding effectors (denoted by kat2mut/kat1mut) and the GAP-stimulated hydrolysis of Gαq-bound GTP by PLCβ (denoted here as the fold over basal hydrolysis khyd2/khyd) for simulations that qualitatively reproduced the experimental activation patterns (shown in pink) and simulations that did not (shown in cyan). A total of n = 11,200 parameter sets were simulated (Fig. 4c and the supplement). b Plotted are the receiver operating characteristic (ROC) curves for using either: (1) the bias of the Gαq Q209L mutant binding effectors (denoted by kat2mut/wt/kat1mu/wt) with a threshold between kat2mut/wt/kat1mu/wt ≥1 and kat2mut/wt/kat1mu/wt ≥14, or (2) the rate constant for GAP-stimulated hydrolysis of Gαq-bound GTP by PLCβ (denoted here as the fold over basal hydrolysis khyd2/khyd) with a threshold between khyd2/khyd <1 and khyd2/khyd <100, as a binary classifier to identify whether the model simulated at various parameter sets either matched or did not match the experiment. It is worth noting that khyd2/khyd was set to roughly 770 in Fig. 2 based on the available data. c Active TRIO and PLCβ levels resulting from simulation of the mathematical model for the WT, GNAQ Q209L, and CYSLTR2 L129Q settings over a range of GAP-stimulated hydrolysis of Gαq-bound GTP by PLCβ (denoted here as the fold over basal hydrolysis khyd2/khyd). The pink region denotes the parameter range for which the simulations qualitatively reproduced the experimental activation patterns and the cyan region denotes the parameter range for which the simulations did not qualitatively reproduce the experimental activation patterns. Based on the experimental observations in Fig. 5 and the analysis in (a, b) above, the bias of Q209L TRIO/PLCB binding was set to kat2mut/wt/kat1mu/wt = 4 for these simulations.

Fig 5: Extension of the model to Gαq/11 inhibition.a Schematic of extensions to the model presented in Fig. 1b to include a Gαq/11-inhibitor. The schematized G-protein alpha subunits (red circles) can represent either (or both) wild-type and mutant forms. b Model simulations of Gαq/11-inhibitor dose responses on modeled GNAQ/11 WT, heterozygous R183C, and heterozygous Q209L genotypes.