G a s directly drives PDZ-RhoGEF signaling to Cdc42

G a proteins promote dynamic adjustments of cell shape directed by actin-cytoskeleton reorganization via their respec-tive RhoGEF effectors. For example, G a 13 binding to the RGS-homology (RH) domains of several RH-RhoGEFs allosterically activates these proteins, causing them to expose their catalytic Dbl-homology (DH)/pleckstrin-homology (PH) regions, which triggers downstream signals. However, whether additional G a proteins might directly regulate the RH-RhoGEFs was not known. To explore this question, we first examined the morphological effects of expressing shortened RH-RhoGEF DH/PH constructs of p115RhoGEF/ARHGEF1, PDZ-RhoGEF (PRG)/ ARHGEF11, and LARG/ARHGEF12. As expected, the three constructs promoted cell contraction and activated RhoA, known to be downstream of G a 13 . Intriguingly, PRG DH/PH also induced filopodia-like cell protrusions and activated Cdc42. This pathway was stimulated by constitutively active G a s (G a s Q227L), which enabled endogenous PRG to gain (46). The cells were serum-starved for 16 h before experiments and all done 48 h after transfec-tion. GST fusion proteins and interactors detected by pulldown (13, cell lysates pulldowns analyzed Western blotting

Migrating cells follow extracellular cues that guide dynamic protrusions and contractions (1,2). At the plasma membrane, phosphoinositides and signaling proteins allosterically activate Rho guanine nucleotide exchange factors (RhoGEFs) exposing their catalytic DH/PH modules, composed of Dbl-homology and Pleckstrin-homology domains in tandem (3)(4)(5). RhoGEFs stimulate their cognate GTPases to exchange GDP for GTP, orchestrating cytoskeleton remodeling pathways (6,7). RhoA promotes the assembly of stress fibers and contractile actomyosin structures, whereas Rac and Cdc42 lead the extension of actin-driven plasma membrane protrusions known as lamellipodia and filopodia, respectively (8). Although these Rho GTPases exhibit contrasting effects, they can be alternatively activated at edges of moving cells (9). Therefore, fine-tuning mechanisms are likely involved.

RhoGEFs
Consistent with a functional effect, Ga s -QL stimulated PRG-DH/PH to activate Cdc42 (Fig. 2G, top panel and graph) and remained bound to the PRG-DH/PH construct pulled down with nucleotide-free Cdc42 (Fig. 2, B, D, and G, middle panel). Furthermore, the interaction of PRG DH/PH with Cdc42-T17N was further stimulated by Ga s -QL (Fig. 2H). . Ga s -Q227L binds PRG DH/PH enabling this prototypical RhoA-specific GEF to directly activate Cdc42. A, hypothetic model postulating Ga subunits as potential regulators of PRG DH/PH catalytic module. B and C, the effect of GTPase-deficient Ga subunits on the interaction of EGFP-PRG-DH/PH-CAAX with Cdc42-G15A (B) and RhoA-G17A (C) was analyzed by pulldown (PD) using lysates from HEK293T cells transfected with HAtagged Ga s , Ga i , Ga q , or Ga 13 QL mutants and EGFP-PRG-DH/PH-CAAX. The graph in B represents the means 6 S.E. (n = 3). ***, p , 0.001; n.s., no significance, one-way ANOVA followed Tukey. D, to address whether Ga s -QL detected in the PRG-DH/PH·Cdc42-G15A pulldown was part of a ternary complex, pulldown experiments were done in the presence or absence of PRG-DH/PH. The graph represents the means 6 S.E. (n = 3). **, p , 0.0001, t test. E, the potential interaction between active Ga subunits and PRG DH/PH was analyzed in HEK293T cells transfected with GST-PRG-DH/PH and HA-tagged GTPase-deficient Ga subunits subjected to pulldown assays. F, interaction between Ga s -QL and the catalytic domain of the three RH-Rho-GEFs was assayed by pulldown using HEK293T cells transfected with HA-Ga s -QL and GST-p115-DH/PH, GST-PRG-DH/PH, or GST-LARG-DH/PH. GST and HA-Ga 13 -QL served as negative controls. G, the effect of Ga s -QL on the activation of Cdc42 by PRG-DH/PH was assessed by pulldown using lysates of transfected HEK293T cells. The graph represents the means 6 S.E. (n = 3). **, p = 0.01, t test. Representative blots show the fraction of active Cdc42 (top panel) and the active fraction of PRG-DH/PH with affinity for Cdc42-G15A (middle panel). H, the effect of Ga s -QL on the interaction between PRG-DH/PH and Cdc42-T17N was analyzed by pulldown using lysates of transfected HEK293T cells. The graph represents the means 6 S.E. (n = 3). *, p = 0.01, t test. I and J, the effect of Ga s -QL on full-length PRG affinity for Cdc42 was analyzed in HEK293T cells that were transfected with full-length AU1-PRG (I) without or with HA-Ga s -QL or only with HA-Ga s -QL to address its effect on endogenous PRG (J). The active fraction of full-length PRG with affinity for Cdc42-G15A was isolated by pulldown and revealed by immunoblotting with anti-PRG antibodies. The graphs represents the means 6 S.E. (n = 3). *, p = 0.01 in H and 0.04 in I, t test.

Ga s -Q227L drives full-length PRG to interact with Cdc42
To address whether full-length PRG is sensitive to be driven by Ga s -QL to gain affinity for Cdc42, we used lysates from transfected HEK293T cells. We found that Ga s -QL stimulated full-length PRG, either transfected (Fig. 2I) or endogenous (Fig.  2J), to bind Cdc42. Furthermore, Ga s -QL remained bound to PRG pulled down with nucleotide-free Cdc42. Expression of transfected and endogenous proteins was confirmed in total cell lysates (Fig. 2, C-J, TCL).

Ga s -Q227L interaction interface at PRG involves the DH and PH domains and the linker region joining them
To characterize how Ga s -Q227L binds PRG-DH/PH guiding this prototypic RhoA-specific GEF to interact with Cdc42, we cotransfected Ga s -Q227L with different PRG constructs spanning the DH/PH module, fused to GST (Fig. 3A), and addressed by pulldown their potential interaction. As shown in Fig. 3B, Ga s -QL interacted with the three PRG-DH/PH constructs that had in common the linker region that joins the DH and PH domains. When PRG-DH, PRG-linker, and PRG-PH were used as independent constructs, the three of them interacted with Ga s -QL, indicating that PRG-linker is critical to strengthen the interaction (Fig. 3C). Consistent with previous results, Ga 13 -QL served as negative control (Fig. 3B). Because the PRG-linker sequence is conserved among the three RH-RhoGEFs (Fig. 3D) but lacks homology with any other protein, we used an EGFP-tagged PRG-linker construct as a potential inhibitor of Ga s -dependent PRG·Cdc42 interaction. Based on previous results and the structure of the PRG-DH/PH·RhoA complex (Fig. 3E) (24), we postulated that Ga s -Q227L interacts with PRG-DH/PH, forming a complex with affinity for Cdc42 (Fig. 3F). Then we tested the potential inhibitory effect of PRG-linker on the interaction between Ga s -Q227L and PRG (Fig. 3G). As predicted, the PRG-linker construct not only inhibited Ga s -QL·PRG-DH/PH interaction (Fig. 3H) but also interfered on the effect of Ga s -QL to drive full-length PRG to be pulled down as an active Cdc42-GEF (Fig. 3I); EGFP served as negative control.

Agonist-dependent stimulation of G s -coupled receptors drives PRG to gain affinity for Cdc42
To investigate whether G s -coupled receptors stimulate PRG to acquire affinity for Cdc42, we first used PAE and HT29 cells as models of endogenous prostaglandin-dependent G s signaling. As an initial readout of agonist-driven G s -dependent effect on PRG, we stimulated PAE and HT29 cells with PGE2 and butaprost, respectively, and assessed PRG recruitment to membrane fractions. In both cases, PRG exhibited a significant timedependent association to membrane fractions (Fig. 4, A and B, respectively). We then used COS7 cells expressing Gs-DREADDs to test the effect of endogenous G s on PRG·dc42 interaction. In these cells, clozapine N-oxide (CNO), the agonist of G s -DREADDs, enabled PRG to bind nucleotide-free Cdc42 in a time-dependent manner (Fig. 4C). This effect was elicited by G s -coupled, but not by G i -or G q -coupled DREADDs (Fig. 4D), and was inhibited by the PRG-linker peptide (Fig. 4E), which also interfered on CREB phosphorylation (Fig. 4F). Endogenous G s -coupled endothelial EP2 receptors signaling to cAMP/PKA pathway was also inhibited by the PRG-linker construct, as indicated by a decrease on butaprost-dependent phosphorylation of PKA substrates (Fig. 4G). (25)(26)(27). They are activated by Ga 12/13 proteins, which bind the RH domains unleashing the catalytic DH/PH region, known as specific for RhoA (14,23). Here we demonstrate that active Ga s directly constrains the PRG DH/PH catalytic module to activate Cdc42, whereas its effect on RhoA is unaltered. Although future work using purified proteins is guaranteed, our results suggest that RhoGEF DH/PH domains can be allosterically controlled to expand their specificity.

RH-RhoGEFs link heterotrimeric G proteins to Rho GTPases
Direct activation of RhoGEF DH/PH domains by active Ga subunits of heterotrimeric G proteins has been described. Specifically, Ga q stimulates p63RhoGEF and TRIO (21,28). Physiological control of this system is lost by GNAQ mutation, causing the Ga q ·TRIO signaling system to drive uveal melanoma progression (28,29). Similarly, mutant GNAS is a driving oncogene in neuroendocrine cancers (30); however, a pathological link to Rho GTPases has not been established. Our results are reminiscent of the regulation of p63RhoGEF and TRIO by Ga q (21,22,28) but differ in the fact that Ga s expands PRG specificity directing the DH/PH module to gain affinity for Cdc42 without an apparent effect on RhoA. We speculate that Ga s pulls the PRG DH/PH module to accommodate Cdc42. Consistent with this possibility, conserved residues that directly bind the GTPase are more distant in intersectin-1, a Cdc42-specific GEF compared with the PRG DH/PH module in complex with RhoA (31)(32)(33).
Consistent with their reported effects (25,27,32,(34)(35)(36), RH-RhoGEF DH/PH catalytic modules strongly activated RhoA and promoted the assembly of actin stress fibers and cell contraction. These results confirmed that DH/PH constructs maintain catalysis and specificity (36)(37)(38)(39)(40)(41)(42)(43)(44). However, we found that PRG DH/PH also exhibited a previously unrecognized ability to stimulate Cdc42 and filopodia formation. Thus, we addressed the possibility that PRG directly activates Cdc42. We used pulldown assays to isolate active RhoGEFs based on their affinity for nucleotide-free GTPases (20,45) and revealed that Ga s stimulates PRG to gain affinity for Cdc42, pointing to a direct effect (attenuated by PKA; Fig. S1). We demonstrated that GTPase-deficient Ga s binds the DH/PH module. The linker region joining these domains strengthen their interaction with active Ga s . Our evidence arguing for a functional relevance of this interaction derives from the inhibitory effect of the PRG-linker construct, which prevented PRG response to . Agonist-dependent stimulation of G s -coupled receptors enables PRG to bind Cdc42. A and B, membrane recruitment of endogenous PRG promoted by G s -coupled GPCR signaling was assessed in PAE (A) and HT29 (B) cells stimulated with 1 mM PGE 2 or butaprost, respectively. PRG in membrane fractions was revealed by Western blotting. GLUT1 and AKT1 were used as membrane and cytosolic markers, respectively. The graphs represent the means 6 S.E. (n = 3, *, p , 0.05 in A; and n = 4, **, p , 0.01 in B; t test). C, time course of PRG·Cdc42 interaction was assessed in COS7 expressing G s -DREADD receptors. The cells were stimulated with 1 mM CNO and subjected to Cdc42-G15A pulldown. The graph represents the means 6 S.E. (n = 3). *, p , 0.05, one-way ANOVA followed Tukey. D, the effect of different endogenous heterotrimeric G proteins on PRG affinity for Cdc42 was studied in COS7 cells transfected with AU1-PRG and G s -, G i -, or G q -DREADDs. The cells were stimulated with CNO for 15 min and subjected to Cdc42-G15A pulldown assays. The graph represents the means 6 S.E. (n = 3). *, p , 0.05, one-way ANOVA followed Tukey. E, effect of the PRG-linker construct on agonist-stimulated interaction between PRG and Cdc42 was assessed in COS7 cells transfected with Gs-DREADD, AU1-PRG, and EGFP-PRG-linker or EGFP. The cells were stimulated with CNO for 15 and 30 min and subjected to Cdc42-G15A pulldown. The graph represents the means 6 S.E. (n = 3). *, p = 0.0342; **, p = 0.0056, t test. F, effect of PRG-linker on agonist-dependent phosphorylation of CREB was assessed in COS7 cells expressing G s -DREADDs and stimulated with CNO. The graph represents the means 6 S.E. (n = 5). **, p , 0.01, t test. G, agonist-dependent phosphorylation of PKA substrates was assessed using PAE cells expressing EGFP or EGFP-PRG-linker and stimulated with butaprost. Lysates from EGFP-PKA-Ca2transfected cells served as control to detect PKA substrates. The graph represents the means 6 S.E. (n = 4). ***, p = 0.0009; ****, p , 0.0001; n.s., no significance, one-way ANOVA followed Tukey. H, model depicts the canonical G 13 -PRG signaling axis to Rho and the emerging GPCR-Ga s -PRG-Cdc42 pathway based on the current findings. In cells, both systems putatively guide dynamic adjustments on actin-cytoskeleton reorganization.
agonist-dependent stimulation of Gs-DREADDs and to GTPase-deficient Ga s coexpression. Our results not only indicate that Ga s guides PRG to bind Cdc42 but also suggest that this effector competes with other Ga s -dependent effectors. The G s /PKA pathway activates Rho GTPases and regulates cytoskeletal dynamics at multiple levels. Recent evidence documented a role for PKA R1a subunit as a cAMP-dependent activator of P-REX1, a RacGEF (46), whereas kinase activity of PKA is linked to cytoskeletal dynamics at cell edges and is reciprocally regulated during cell migration (47)(48)(49). Our findings showing that Ga s activates a PRG/Cdc42 pathway expand the mechanisms of G s signaling to Rho GTPases.
Although further experiments are needed to define the spatiotemporal conditions in which PRG is guided to activate Cdc42, our current model (Fig. 4H) illustrates the potential of G s -coupled receptors to activate this pathway. Our work raises new questions and research avenues on how G s and G 13 signaling pathways are integrated to fine-tune Cdc42 activity in the context of strong RhoA activation to regulate cytoskeletal dynamics and set the basis to further investigate how the Gs/ PRG/Cdc42 pathway guides polarized cell migration and its potential pathological implications, particularly in cancers in which mutant GNAS is a driving oncogene.

Experimental procedures
Plasmids and cDNA constructs RH-RhoGEF DH/PH catalytic modules and PRG DH-PH fragments were amplified by PCR and cloned into pCEFL-EGFP-CAAX, pCEFL-EGFP, and pCEFL-GST. Primer sequences are available upon request. Other constructs have been previously described (13,46).

Membrane and cytoplasmic fractionation of PAE and HT29 cells
Serum-starved PAE and HT29 cells, grown in 10-cm Petri dishes, were stimulated with 1 mM prostaglandin E2 or butaprost, as indicated in Fig. 4. The cells were washed with cold PBS, scraped into 1 ml of cold PBS containing protease and phosphatase inhibitors, and subjected to three freeze/thaw cycles. The lysates were centrifuged at low speed (1,400 rpm for 10 min at 4°C). Supernatants were centrifuged at 13,000 rpm for 10 min at 4°C. Cytosol-enriched supernatants were prepared with Laemmli buffer. The pellets were washed once with cold PBS, centrifuged again, incubated with 250 ml of lysis buffer containing 1% Triton X-100 for 20 min, and centrifuged at 13,000 rpm for 10 min at 4°C. Supernatants containing solubilized membranes were prepared with Laemmli buffer. PRG was analyzed by Western blotting, together with GLUT1 and AKT1, as membrane and cytosol markers, respectively.
Cytoskeletal effects of RH-RhoGEF DH/PH constructs PAE cells were seeded at low density on gelatin-coated coverslips. Transfected cells were starved for 16 h with serum-free medium. Subsequently, the cells were fixed in 4% paraformaldehyde in PBS for 20 min, washed twice with PBS, and prepared for conventional phalloidin staining. The cell images were visualized in a Leica confocal laser scanning microscope TCS SP8 using a 633 1.4 oil immersion objective. The images were analyzed with FIJI-ImageJ software. The cells were counted as having filopodia-like structures when they had at least nine of these finger-like protrusions containing F-actin (19).

Statistical analysis
The data are presented as means 6 S.E. of at least three independent experiments. Densitometric quantitation of Western blots was done with ImageJ. Active proteins and interactions in pulldowns were normalized respect to total proteins and pulldown efficiency. Statistical analysis was performed using Sigma Plot 11.0, and graphs were prepared with Prism software V8.0. Statistical tests are indicated at the figure legends.

Data availability
All the described data are contained within this article.