Fig 2: GCN2 is non-specifically activated by RAF inhibitors.a Schematic of the LINCS KINOMEscan experimental workflow. This competitive pulldown assay is performed by combining phage-tagged GCN2 kinase domain with a known GCN2 inhibitor bound to agarose beads. A secondary compound is then added. Agarose beads are collected, phage is lysed, and its contents quantitatively sequenced to determine percent recovery. High percent recovery implies low interaction between GCN2 and the secondary compound. Low percent recovery implies high interaction between GCN2 and the secondary compound. b LINCS KINOMEscan assay data. Percent GCN2 recovery is plotted. All drugs tested at 10μM. c In vitro kinase assay of GCN2 and eIF2α phosphorylation in response to increasing concentrations of dabrafenib. d The noted drugs were tested in five cell lines: A375 (BRAFV600E melanoma), WM278 (BRAFV600E melanoma), SKMEL28 (BRAFV600E melanoma), MCF10A (non-transformed breast epithelial), and RPE-hTERT (non-transformed retinal pigment epithelial). Cells were treated for 6h, lysed, and probed for the indicated protein species. Histone H3 is used as a nuclear loading control. e Immunofluorescence images of ATF4 in WT (left) and GCN2 knockout (KO, right) HAP-1 cells in response to peak activating doses of dabrafenib (2μM), encorafenib (4μM) and vemurafenib (32μM) for 6h. f Violins of single-cell ATF4 immunofluorescence intensity in response to titrations of the indicated drug in ISR kinase knockout HAP-1 cell lines as shown in (e).

Fig 3: Arginine is required for TGF-β-induced signaling and gene expression in human lung fibroblasts.(A) Western blot analysis of collagen 1 and α-smooth muscle actin protein expression and S6-kinase and GCN2 phosphorylation in HLFs cultured in DMEM that contains either 0.4mM arginine or no arginine. Cells were treated with TGF-β for the indicated intervals. (B) Western blot analysis of SMAD2/3 phosphorylation in HLFs cultured in DMEM that contains either 0.4mM arginine or no arginine. Cells were treated with TGF-β for the indicated intervals. (C) qPCR analysis of COL1A1, ACTA2, CTGF, and SERPINE1 mRNA expression in HLFs cultured in the DMEM that contains either 0.4mM arginine or no arginine. Cells were treated with TGF-β for 24 hours or left untreated. (D) Western blot analysis of collagen 1 and α-smooth muscle actin protein expression and S6-kinase and GCN2 phosphorylation in HLFs cultured in HPLM that contains either 0.11mM arginine or no arginine. Cells were treated with TGF-β for the indicated intervals. (E) Western blot analysis of SMAD2/3 phosphorylation in HLFs cultured in HPLM that contains either 0.11mM arginine or no arginine. Cells were treated with TGF-β for the indicated intervals. (F) qPCR analysis of COL1A1, ACTA2, CTGF, and SERPINE1 mRNA expression in HLFs cultured in the HPLM that contains either 0.11mM arginine or no arginine. Cells were treated with TGF-β for 24 hours or left untreated. *P<0.05, **P<0.01, ***P<0.001.

Fig 4: In silico kinase inhibitor screen predicts how RAF inhibitors interact with GCN2.a Top: computational docking models of dabrafenib, encorafenib, and vemurafenib bound to BRAFV600E protein, or sorafenib bound to wild-type CRAF. Bottom: the same small molecules’ predicted binding modes to GCN2. b Intermolecular interactions of the GCN2 active site with dabrafenib and vemurafenib. c Waterfall plot of hits from an in silico kinase inhibitor screen with the kinase domain of GCN2. Hits are ranked by relative MM/GBSA value. d Pie chart of the cancer drug targets of the top 200 hits (of which >30% are cancer drugs) in the in silico inhibitor screen. e Validation of hits from the in silico inhibitor screen and LINCS database. A375s were treated with the indicated drug for 6h. Immunofluorescence of single-cell nuclear ATF4 signal is shown as violin plots.

Fig 5: UV-induced SG formation is dependent on G3BP1. (A–D) G3BP1, G3BP2, and control PKR knockout cells were generated using CRISPR/Cas9. (A) PKR, G3BP1 and G3BP2 expression in CRISPR knockout cells and eIF2α phosphorylation in response to 20 mJ/cm2 UV light (UV) were analysed by western blotting. Staining for actin was used as loading control. (B) UV-induced SG formation (20 mJ/cm2) was analysed by immunofluorescence microscopy in parental U2OS cells (U2OS) and CRISPR knockout cell lines lacking G3BP1 (ΔG3BP1), G3BP2 (ΔG3BP2) or PKR (ΔPKR), and stained for G3BP1 (red) and TIAR (green). Nuclei were stained with Hoechst dye (blue). (C) UV-induced SG formation was quantified (mean±s.d.) from immunofluorescence microscopy staining represented in B (n=3). *P<0.05 (two-way ANOVA followed by Tukey's multiple comparisons test). (D) Selenite-induced SG formation was analysed by immunofluorescence microscopy in parental U2OS cells (U2OS) and CRISPR knockout cell lines lacking G3BP1 (ΔG3BP1) or G3BP2 (ΔG3BP2) stained for G3BP1 (red) and G3BP2 (green). Nuclei were stained with Hoechst dye (blue). (E,F) Expression of the EGFP–USP10(1-40) fusion protein or EGFP control was induced in U2OS cells stably transduced with lentiviral constructs encoding these proteins. After 48 h, cells were left untreated (−), or were treated with UV light (UV, 20 mJ/cm2) or sodium arsenite (As). (E) Phosphorylation status of GCN2 and eIF2α post-treatment was analysed by western blotting, as well as the expression of EGFP–USP10(1-40) fusion protein or EGFP control. Staining for actin was used as loading control. (F) SG formation was analysed by immunofluorescence microscopy staining for G3BP1 (red) and TIAR (blue) in these GFP-positive cells (green). Numbers indicate the percentage of cells with SGs quantified from at least three random fields of view containing >100 cells (mean values, n=2).