Discovery of a dual Ras and ARF6 inhibitor from a GPCR endocytosis screen

Abstract

Internalization and intracellular trafficking of G protein-coupled receptors (GPCRs) play pivotal roles in cell responsiveness. Dysregulation in receptor trafficking can lead to aberrant signaling and cell behavior. Here, using an endosomal BRET-based assay in a high-throughput screen with the prototypical GPCR angiotensin II type 1 receptor (AT1R), we sought to identify receptor trafficking inhibitors from a library of ~115,000 small molecules. We identified a novel dual Ras and ARF6 inhibitor, which we named Rasarfin, that blocks agonist-mediated internalization of AT1R and other GPCRs. Rasarfin also potently inhibits agonist-induced ERK1/2 signaling by GPCRs, and MAPK and Akt signaling by EGFR, as well as prevents cancer cell proliferation. In silico modeling and in vitro studies reveal a unique binding modality of Rasarfin within the SOS-binding domain of Ras. Our findings unveil a class of dual small G protein inhibitors for receptor trafficking and signaling, useful for the inhibition of oncogenic cellular responses.

Fig. 1. High-throughput screening identifies compound 21 as an inhibitor of GPCR internalization.

a Flowchart outlining the steps of the screening and selection of hits to the lead compound. b HTS results of 40 active hits on AT1R internalization using trafficking sensors. BRET responses were quantified as percent AngII-promoted BRET compared to DMSO and are presented as individual values, n = 2 biologically independent experiments. c Structure of the selected compound 21. d Effects of 21 (50 μM) on AT1R (closed red squares and line), B2R (closed green triangles and line), and β₂AR (closed orange triangles and line) internalization into endosomes. BRET responses were quantified as percent ligand-promoted BRET compared to DMSO. The mean values of the ligand-promoted BRET responses (BRET_ligand-BRET_basal) in DMSO were 0.260, 0.549, and 1.061 for AT1R, B2R and β₂AR, respectively. Data are presented as the means values ± SEM, n = 4 biologically independent experiments performed in triplicate. e Confocal microscopy images of YFP-tagged ATIR internalization, repeated independently three times with similar results. Scale bar = 10 μm. Bottom micrographs are 5× enlargements of the boxed areas. f BRET recording of the recruitment of β-arrestin1 and β-arrestin2 to AT1R in the absence (DMSO, black triangles and dotted line and black circles and solid line, respectively) or presence of 21 (50 μM, red triangles and squares, respectively, and lines). BRET responses were quantified as AngII-promoted BRET. Data are presented as mean values ± SEM, at least n = 3 biologically independent experiments performed in triplicate. Source data are provided as a Source Data file. g Confocal microscopy images of YFP-tagged β-arrestin2 internalization, repeated independently three times with similar results. Scale bar = 10 μm. Bottom micrographs are 5× enlargements of the boxed areas.

Fig. 2. Compound 21 targets ARF6 to inhibit GPCR internalization.

a Effects of 21 (50 μM, closed red squares and line) and Barbadin, a β-arrestin2/AP-2 inhibitor (100 μM; closed cyan triangles and line), on the binding of β-arrestin2 to AP-2. BRET responses were quantified as AngII-promoted BRET and are presented as mean values ± SEM, n = 3 biologically independent experiments performed in triplicate, *p < 0.05, ***p < 0.005, ****p < 0.0001, two-way ANOVA corrected with Dunnett’s test. b Coomassie of GST and GST-GGA3-PBD proteins and western blots of AT1R-mediated HA-ARF6 activation as assessed by glutathione beads pull-downs. Right panel is the quantification of AT1R-mediated ARF6 activation in the absence (DMSO, open black bars) or presence of 21 (50 μM, open red bars) calculated as the amount of ARF6-GTP over total ARF6 and are presented as the mean values ± SEM, n = 4 biologically independent experiments, ****p < 0.0001, one-way ANOVA with Bonferroni correction. c BRET kinetics of AT1R-mediated ARF activation in the absence (DMSO, closed black circles and line) or presence of 21 (50 μM, closed red squares and line). Data were quantified as AngII-promoted BRET and are represented as the mean values ± SEM, at least n = 3 biologically independent experiments performed in triplicate, **p < 0.01, two-tailed unpaired Student’s t-test. d BRET recordings of AT1R internalization represented as the removal of AT1R from the PM. Cells were transfected with an empty vector (mock, closed black circles and line), HA-ARF6-WT (closed blue squares and line), or HA-ARF6-T27N (closed green triangles and line). Data were quantified as AngII-promoted BRET and are presented as the mean values ± SEM, at least n = 3 biologically experiments performed in triplicate, *p < 0.05, **p < 0.01, ***p < 0.005, two-way ANOVA corrected with Dunnett’s test. Western blots of HA and β-actin in the inset are used as controls of protein expression. e BRET recordings of AT1R removal from the PM in cells transfected with ARF6-T27N (closed green triangles and line) or empty vector (closed black circles and line) and incubated or not with 21 (50 μM, closed red squares and line). Data were quantified as AngII-promoted BRET and are presented as the mean values ± SEM, at least n = 3 biologically independent experiments performed in triplicate, *p < 0.05, **p < 0.01, ****p < 0.0001, two-way ANOVA corrected with Dunnett’s test. Source data are provided as a Source Data file.

Fig. 3. Compound 21 inhibits receptor signaling.

a Western blots of ERK1/2 phosphorylation kinetics in cells mediated by AT1R, B2R, β₂AR, and EGFR in absence (DMSO, black open bars) or presence of 21 (50 μM, red open bars). Data were quantified as p-ERK1/2 over ERK1/2, normalized to basal (0 min), and compared to DMSO. Data are presented as mean values ± SEM, n = 4 biologically independent experiments, *p < 0.05, **p < 0.01, ****p < 0.0001, two-way ANOVA with Bonferroni correction. b–d Western blots and quantification of AT1R- and EGFR-mediated ERK1/2, and EGFR-mediated Akt activation in the absence (DMSO) or presence of 21 (at indicated concentrations). Data were quantified as AngII- (closed red squares and line) and EGF-mediated (closed blue triangles and line) c p-ERK1/2 over ERK1/2, and EGF-mediated (closed red circles and line) d p-Akt over Akt, respectively, and normalized as fold over basal, percent compared to DMSO (dotted lines). Data are presented as mean values ± SEM, n = 3 biologically independent experiments. Source data are provided as a Source Data file.

Fig. 4. Compound 21 targets Ras.

a Coomassie of GST and GST-Raf1-RBD proteins and western blots of AT1R-mediated and b EGFR-mediated Ras activation in absence (DMSO, open black bars) or presence of 21 (50 μM, open red bars) as assessed by GST- and GST-Raf1-RBD-coupled to glutathione beads pull-downs. Ras activation was calculated as the amount of Ras-GTP over the total Ras detected and are presented as mean values ± SEM, n = 5 for AT1R (a) and n = 3 for EGFR (b) biologically independent experiments, **p < 0.001, one-way ANOVA with Bonferroni correction. c BRET kinetics of AT1R-mediated Ras activation in absence (DMSO, closed black circles and line) or presence of 21 (50 μM, closed red squares and line). Data were quantified as AngII-promoted BRET and are presented as mean values ± SEM, n = 4 biologically independent experiments performed in triplicate, ***p < 0.005, two-tailed unpaired Student’s t-test. d BRET recording of the activation of Rho (purple squares), Rac (orange diamonds), ARF (green triangles and line) and Ras (blue triangles and line) by AT1R in the presence of 21 (50 μM). BRET responses were quantified as AngII-promoted BRET, percent compared to DMSO (dotted line) and are presented as mean values ± SEM, n = 3 (Rho and ARF) and n = 4 (Rac and Ras) biologically independent experiments performed in triplicate. e, f In vitro kinetics of mant-GTP loading onto H-Ras. Purified H-Ras nucleotide exchange was induced using e purified SOS1 or f 40 mM EDTA in the presence of DMSO (black dots and lines) or different concentrations of Rasarfin (as indicated with respective colors). The relative fluorescence unit (RFU) was measured every 30 s for 30 min and quantified as the delta RFU (RFU post-addition minus the 5 averaged RFU pre-addition, per condition). Data are presented as mean values ± SEM, at least n = 4 biologically independent experiments. Source data are provided as a Source Data file.

Fig. 5. In silico studies reveal interactions between Rasarfin and Ras.

a Shown are the 8 best-scored docking poses of Rasarfin in licorice representation (orange) bound in the groove between switch I and switch II on Ras (protein in white cartoon, binding groove surface in silver). b–c Unbiased simulations of Rasarfin association to Ras. b Shown is the initial placement of Rasarfin molecules (orange licorice) around Ras to study the spontaneous association of the ligand to Ras (white depiction). c Spontaneous association of Rasarfin to Ras approximated using a volumetric map of average ligand occupancy at 0.15 threshold (cyan transparent surface). The binding mode of Rasarfin obtained through docking is overlayed for comparison (orange licorice).

Fig. 6. Estimation of the stability of Rasarfin and compound 21.4 binding modes.

a Root Mean Square Deviation (RMSD) (y axis) for Rasarfin (orange) and 21.4 (red) in respect to their initial position were analyzed in three separate MD runs. RMSD values higher than 6 correspond to simulation frames in which the ligand fully unbinds. The ligand position was monitored across 3 separate replicates of 4 μs (total 12 μs). b Comparison of predicted binding mode of Rasarfin (orange licorice) and compound 21.4 (red licorice). The molecular surface of Ras is represented in white and reveals the presence of a structural cavity in the vicinity of the predicted binding modes (black circle). In the predicted binding mode of Rasarfin, the structural cavity is occupied by the chlorine atom (in green), which is missing in compound 21.4. c Overlay of Rasarfin binding into the Ras-SOS binding interface (PDB:1BKD). Rasarfin is displayed in orange licorice, SOS in yellow, and Ras in red ribbons. The surface occupied by Rasarfin is shown in transparent gray. Residues on Ras and SOS are highlighted in respective colors and labeled accordingly. Rasarfin is binding in the cavity which is normally occupied by residues His911 and Lys939 of SOS. Those residues are shown superimposed with Rasarfin, displaying His911 overlaying with the furan of the benzofuran moiety (AR1), and the Lysine elongated in the binding groove like Rasarfin, partially hidden as they overlay perfectly.

Fig. 7. In silico assessment of Rasarfin interactions with Ras.

a 2D structure of Rasarfin labeled with established pharmacophoric features interacting with Ras residues. AR1-2 aromatic feature, H1-H4 hydrophobic feature, HBA1-2 hydrogen bond-acceptor, HBD hydrogen bond donor, XBD halogen bond donor. b Stick representation of Rasarfin embedded in the binding pocket of Ras (transparent gray) with interacting Ras residues labeled. Yellow spheres correspond to H1-H4; green arrow shows HBD of Rasarfin interacting with Ile55 backbone carbonyl. c Plot of interaction frequencies between residues of Ras (x-axis) and the pharmacophoric features of Rasarfin (y-axis). Color coding according to interaction frequency.

Fig. 8. Functional selectivity of Rasarfin analogs.

a BRET recording of AT1R internalization into endosomes in the absence (DMSO, open black bar) or presence of 50 μM Rasarfin (open red bar) or compounds 21.1–21.8 (open black bars). BRET responses were quantified as AngII-promoted BRET response, percent compared to DMSO (red dotted line). Data are presented as mean values ± SEM, at least n = 4 biologically independent experiments performed in triplicate, *p = 0.0234, **p = 0.0013, ****p < 0.0001, one-way ANOVA with Dunnett’s test. b Quantification of AT1R-mediated ERK1/2 activation in cells the absence (DMSO, open black bar) or presence of 50 μM Rasarfin (open red bar) or compounds 21.1–21.8 (open black bars) and representative western blots. Data are quantified as p-ERK1/2 over ERK1/2 and normalized as fold over basal. Data are presented as mean values ± SEM, n = 5 biologically independent experiments, ****p < 0.0001, one-way ANOVA with Dunnett’s test. c In vitro kinetics of mant-GTP loading into H-Ras. Purified H-Ras was activated using purified SOS1 and in the presence of DMSO, Rasarfin, 21.4, 21.7 or 21.8 (50 μM). The relative fluorescence unit (RFU) was measured every 30 s for 30 min and quantified as the delta RFU (RFU post-addition minus the 5 averaged RFU pre-addition, per condition). Data are presented as mean values ± SEM, n = 3 independent experiments. Source data are provided as a Source Data file.