Previous work suggested that PIP₃ binding to the PH domain activates P-Rex1 purely through an allosteric mechanism (Cash et al., 2016). Thus, we anticipated that binding of the soluble headgroup of PIP₃, IP₄, to full-length P-Rex1 could also lead to conformational changes characteristic of the activated state. To test this, we analyzed P-Rex1 in the presence and absence of IP₄ using HDX-MS. We observed strong protection from deuterium incorporation in the PIP₃-binding site on the PH domain in the presence of IP₄ (Figure 1A and B). However, we also observed strong protection in other regions of the protein: namely on the surface of the PH domain, particularly in the β5/β6 loop, and in several regions within an extension of the C-terminal IP4P domain that was not visualized in the P-Rex1–Gβγ complex (Figure 1A and C; Cash et al., 2019). We speculated that these diverse regions form more stable long-range interactions in the presence of IP₄.

Based on our HDX-MS data, we hypothesized that IP₄ could inhibit activity of full-length P-Rex1. Using an in vitro GEF activity assay on soluble Cdc42 in the presence of liposomes, we observed that IP₄ inhibits PIP₃-mediated activation of P-Rex1 with an IC₅₀ value of 1.4 μM (Figure 1D). Competition was not observed with Ins(1,4,5)P₃, indicating that inhibition is dependent on the 3-phosphate, which is critical for PIP₃ binding to the P-Rex1 PH domain (Cash et al., 2016). However, IP₄ did not affect the activity of the P-Rex1 DH/PH or DH/PH-DEP1 fragments (Figure 1—figure supplement 1A and B). Collectively, these results indicated that IP₄ inhibits P-Rex1 allosterically and that this inhibition is dependent on long-range interactions between the regions shown to be protected by IP₄ in the Gβγ-binding scaffold (DEP2-PDZ1-PDZ2-IP4P) and in the DH/PH-DEP1 module. Negatively charged liposomes (containing PC/PS), including those that also contain PIP₃, inhibit the GEF activity of the DH/PH-DEP1 and DH/PH fragments (Figure 1—figure supplement 1C). Because full-length P-Rex1 is not affected by PC/PS liposomes, this suggests that the observed inhibition represents a non-productive interaction of the DH/PH-DEP1 and DH/PH fragments with negatively charged surfaces in our assay. The lack of activation of DH/PH-DEP1 by PIP₃ prevents us from testing in this assay whether IP₄ can inhibit via direct competition with PIP₃ at the PH domain.

To understand the molecular basis of IP₄-mediated stabilization and inhibition, we analyzed full-length P-Rex1 with and without IP₄ using cryo-EM SPA (Figure 2—figure supplement 1A). Initial datasets were collected using a Glacios transmission electron microscope and then processed through 2D classification (Figure 2—figure supplement 1B–D). In both datasets, most classes showed only the Gβγ-binding core of P-Rex1 (Figure 2—figure supplement 1C and D). A few classes contained particles with additional mass to the side of the core, close to the PDZ or DEP domains (Figure 2—figure supplement 1C and D, orange boxes). This mass could represent either the N-terminal domains or another P-Rex1 particle in proximity. Only in the sample containing IP₄ could we observe class averages with additional mass next to the core opposite the side that binds Gβγ (Figure 2—figure supplement 1D, green boxes). Based on its size, location, our HDX-MS data (Figure 1), and low-resolution maps of P-Rex1 generated in a previous study (Cash et al., 2019), this mass most likely corresponded to the N-terminal DH/PH-DEP1 domains interacting with an elongated subdomain extending from the IP4P domain.

We next collected much larger datasets on the P-Rex1·IP₄ complex using a Krios transmission electron microscope and determined the structure of this complex at an average resolution of 4.1 Å (Figure 2A and B, Table 1, and Figure 2—figure supplement 2). Similar to the P-Rex1–Gβγ complex (Cash et al., 2019), this sample exhibited a preferred orientation on grids, necessitating the addition of data collected on a tilted sample (Figure 2—figure supplement 2A and C). The resulting 3D reconstruction clearly shows the Gβγ-binding core, composed of DEP2, PDZ1, PDZ2, and the majority of the IP4P domain, from which there are two extensions of density that contact one another to form a loop-like structure (Figure 2A, Figure 2—figure supplement 2D). One extension corresponds to a large insertion in the IP4P domain that contains IP₄-stabilized regions (Figure 1) and that was disordered in the P-Rex1–Gβγ complex (Cash et al., 2019). The ordered elements of the insertion form a long 4-helix bundle (4HB) most similar in fold to focal adhesion targeting (FAT) domains (Hayashi et al., 2002), which are found in other peripheral membrane proteins involved in cell adhesion and migration. The other extension corresponds to the DH/PH-DEP1 domains. Based on its distinct shape, the PH domain (Cash et al., 2016) was fit into density along the side of the 4HB. Here, IP₄ is observed bound to the PH domain PIP₃-binding site (Figure 2B–D). The position of the DH domain was also obvious, but individual helices of the DH domain were lower in resolution and more dynamic relative to the rest of the structure (Figure 2—figure supplement 2D). The relative positions of the DH and PH domains mandate a severe bend in the helix connecting the DH and PH domains (Figure 2A and B), resulting in a jack-knifed configuration of the DH/PH module that blocks access to the GTPase binding site on the DH domain. The remaining mass, immediately adjacent to the end of the DH domain opposite its N-terminus, corresponds to DEP1. Using the required connectivity of its N- and C-termini to the PH and DEP2 domains, respectively, DEP1 was docked in a manner that complemented residues on the DH domain (Figure 3A and B). Weak density corresponding to a long five-turn extension of the αC helix of the PH domain connects to the N-terminus of DEP1, but its C-terminal connection to DEP2 is disordered, likely explaining lower local resolution in the DEP1 region (Figure 2—figure supplement 2D). Overall, the conformation of the P-Rex1 Gβγ-binding core is essentially the same as in the P-Rex1–Gβγ complex (RMSD deviation of 1.1 Å for 701 Cα atoms).

The contact between the PH domain and the 4HB is primarily mediated via the β1 and β2 strands and β5/β6 loop of the PH domain (Figure 2C and D). In all previous crystal structures including the P-Rex1 PH domain, this same surface formed extensive protein–protein lattice contacts (Cash et al., 2016; Lucato et al., 2015). The residues directly involved in the interface are among the most strongly protected in the presence of IP₄ as measured by HDX-MS (Figure 1A–C, Figure 4A). The PH domain β3/β4 loop, which we previously showed to be a nonspecific anionic membrane-binding loop, remains unstructured and is situated near a loop at the tip of the 4HB that is also unstructured (residues 1109–1209, Figure 2C) and contains known phosphorylation sites, some of which regulate activity (Barber et al., 2012). At the interface, surface hydrophobic residues Leu279 and Ile286 on the PH β1 and β2 strands and Tyr353 on the β5/β6 loop interact with a surface of 4HB including Tyr1096 and His1224 (Figure 2D). Charge complementarity is formed between Lys302 in the PH domain and Asp1216 in 4HB, and between the 1-phosphate of IP₄ and Lys1217. Otherwise, IP₄ does not make direct contact with the 4HB. However, because different PH domain ligands uniquely perturb the conformation of these regions in the PH domain (Cash et al., 2016), IP₄ could indirectly stabilize this interface by trapping the β1 and β2 strands and β5/β6 loop in a conformation with higher affinity for 4HB.

The jack-knife in the helix between the DH and PH domains is required to allow the PH domain to interact with the 4HB and is stabilized by the DEP1 domain docking to the DH domain (Figure 2B). DH domain residues Leu173, Leu177, and Leu178 form a hydrophobic interface with DEP1 residues Ile409, Ile457, Leu466, and Ala469 (Figure 3A). Leu173 and Leu177 were previously noted to be conspicuously exposed in structures of the DH/PH tandem (Cash et al., 2016). Thus, the DEP1 domain stabilizes an inactive DH/PH tandem that is further stabilized by interaction with the 4HB of the IP4P domain.

While conducting these studies, the structure of human P-Rex1 in the absence of IP₄ was reported (PDB entry 7SYF) (Chang et al., 2022), allowing a comparison between the IP₄ and IP₄-free states of autoinhibited P-Rex1 (Figure 2—figure supplement 3). Overall, the domain organization is very similar, but there is an ~3° rotation of the Gβγ-binding core in the IP₄ complex relative to the PH–4HB interface such that DEP1 and DEP2 move closer together. It is possible that the binding of IP₄ at this interface drives this conformational change. A caveat is that the protein used for the 7SYF structure contained a T4 lysozyme domain inserted into the β3/β4 loop of the PH domain. Although this domain was not visible in the reconstruction, its proximity to the IP₄ binding site may influence the global conformation of P-Rex1. The similarities between P-Rex1 ± IP₄ also suggest that the IP₄-binding site in the PH domain is freely accessible in the autoinhibited state. PIP₃ would, however, not have access because the 4HB domain would block binding to a membrane surface.

To test the contribution of the observed DH–DEP1 interface to autoinhibition, the interface was disrupted by site-directed mutagenesis in the context of the DH/PH-DEP1 fragment. This fragment is five- to tenfold less active (depending on assay conditions) than the DH/PH tandem alone (Figure 1—figure supplement 1C; Ravala et al., 2020), confirming a specific role for DEP1 in autoinhibition (Figure 3C). Single-point mutations in the interface profoundly affected GEF activity on soluble Cdc42. The L173A variant had an ~2.5-fold higher activity, whereas L177E and L178E exhibited four- to fivefold higher activity, similar to the activity of DH/PH without the DEP1 domain. In the DEP1 domain, I409A and L466A mutations resulted in approximately two- and fourfold higher activity, respectively. The α1/α2 loop (residues 77–90) of the DH domain, although ordered in previous structures of DH/PH bound to GTPases (Cash et al., 2016; Lucato et al., 2015), is disordered in our structure (Figure 3B). However, modeling the DH domain from previous DH/PH crystal structures suggests that Lys89 and Arg78 would be close enough to form a bipartite ionic interaction with Glu456 in the DEP1 domain. Consistent with this hypothesis, the E456K variant was ~2.5-fold more active. Surprisingly, one of the intended disruptive mutations, A170K, instead inhibited GEF activity by ~50% (Figure 3C). Based on the available structures, Lys170 in the DH domain could form complementary electrostatic interactions with Glu411 along with additional non-polar interactions with the backbone and side chains of residues 409, 411, and 412 in the DEP1 domain, thereby stabilizing the DH–DEP1 interface (Figure 3B).

Because disruption of the DH–DEP1 interface led to increased activity, we predicted that these variants would have a more extended DH/PH module that is likely more dynamic. To test this, wild-type (WT) DH/PH-DEP1 and its variants were assessed using a ThermoFluor assay to determine their melting temperatures (T_m). Indeed, variants with increased activity also had lower T_m values (Figure 3D, Table 2). Conversely, A170K, which was less active than WT, showed a higher T_m. We also analyzed these variants using size-exclusion chromatography coupled to small-angle X-ray scattering (SEC-SAXS) (Figure 3—figure supplement 2A and B). In previous SAXS analyses, we observed that DH/PH-DEP1 exhibited a more compact state with a smaller conformational landscape in solution relative to DH/PH (Ravala et al., 2020). We hypothesized that mutations which disrupt the DH–DEP1 interface would likewise lead to more elongated ensembles. Compared to WT DH/PH-DEP1, which had a radius of gyration (R_g) of 30 ± 0.3 Å, the L177E variant (the most active DH/PH-DEP1 variant tested) had an R_g of 31 ± 0.2 Å, suggestive of expansion (Figure 3E, Table 3). Variants A170K and I409A had R_g values similar to that of WT. Kratky plots indicated that all the samples had heterogeneous conformations (Figure 3—figure supplement 2C). The shapes of the P(r) functions were similar for all variants except a longer tail for L177E and I409A, consistent with a higher proportion of extended conformations in solution (Figure 3—figure supplement 2D).

Because the samples exhibited a high degree of heterogeneity in solution, the conformational distribution of these variants was assessed using the ensemble optimization method (EOM) (Figure 3E and F, Figure 3—figure supplement 2A; Tria et al., 2015). The resulting ensemble for WT shows predominant conformations with R_g values ~28 Å and a small fraction of extended conformations with R_g ~39 Å (Figure 3E). The selected ensemble for A170K had R_g values similar to WT (~29 Å); however, the A170K peak is broader in comparison, suggesting conformational heterogeneity and structural changes. The L177E variant exhibited a larger shift to higher R_g, with an average R_g ~30 Å, and a second significant population with R_g >32 Å (Figure 3E). Similar to the distribution of R_g values, the D_max function distribution shows that the L177E variant had the most extended conformation of the variants tested (Figure 3F).

One of our initial goals in this project was to determine a high-resolution structure of the autoinhibited DH/PH-DEP1 core by X-ray crystallography. To this end, we started with the DH/PH-DEP1 A170K variant, which was more inhibited than wild-type but still dynamic, and then introduced S235C/M244C and K207C/E251C double mutants to completely constrain the hinge in the α6-αN helix via disulfide bond formation in a redox-sensitive manner. Single cysteine variants K207C and M244C were generated as controls. The S235C/M244C variant performed as expected, decreasing the activity of the A170K variant to nearly background in the oxidized but not the reduced state (Figure 3—figure supplement 1). However, the M244C single mutant exhibited similar effects, suggesting that it forms disulfide bonds with cysteine(s) other than S235C. Indeed, the side chains of Cys200 and Cys234 are very close to that of M244C. The K207C/E251C mutant was similar to S235C/M244C under oxidized conditions, but ~15-fold more active (similar to WT DH/PH levels, see Figure 3C) under reducing conditions. The K270C variant, on the other hand, exhibited higher activity than A170K on its own under oxidizing conditions, but similar activity to all the variants except K207C/E251C when reduced. These results suggest that K207C/E251C in a reduced state and K270C in an oxidized state favor a configuration where the DEP1 domain is less able to engage the DH domain and maintain the kinked state. The mechanism for this is not known. Regardless, these data show that perturbation of contacts between the kinked segments of the α6-αN helix can have profound consequences on the activity of the DH/PH-DEP1 core.

To evaluate the roles of the P-Rex1 DH–DEP1 and PH–4HB interfaces in living cells, we utilized SRE luciferase-gene reporter assays as a read out of full-length P-Rex1 activity in HEK293T cells. Mutations at the DH–DEP1 interface had a strong effect on activity, with L177E and L466E (Figure 3A) exhibiting approximately tenfold and fourfold higher activity relative to WT, respectively (Figure 5A). Perturbation of the PH–4HB interface (Figure 2D) also increased activity in that Y1096A was approximately fivefold more active than WT (Figure 5A). Mutation of other residues in the PH–4HB interface (Figure 2D) also increased activity, although to a lesser extent.

The most affected variants (L177E, L466E, and Y1096A) were then tested for their effects on cell migration in response to chemokine gradients (Figure 5B and C, Figure 5—figure supplement 1A and B). For this, the endogenous P-Rex1 gene in HeLa cells was first knocked out by CRISPR-Cas9 (Figure 5—figure supplement 1C) and the resulting cells were transfected with various P-Rex1 constructs (Figure 5—figure supplement 1D). Cell migration was then evaluated in the presence or absence of CXCL12 (upstream of P-Rex1 and Gβγ signaling) and epidermal growth factor (EGF). CXCL12-induced chemotaxis was dependent on the expression of P-Rex1 (Figure 5B and C, Figure 5—figure supplement 1), and all three variants caused a significantly larger number of cells to migrate. However, EGF-induced chemotaxis in HeLa cells, which is not dependent on P-Rex1, was unaffected by P-Rex1 expression. These data support that the DH–DEP1 and PH–4HB interfaces of P-Rex1 mediate autoinhibition that specifically modulates chemokine-induced cell migration.

Given the known binding site for Gβγ (Cash et al., 2019), the position of membrane-binding loops such as the β1/β2 loop of the DEP1 domain (Ravala et al., 2020), and the position of the 4HB domain, it does not seem that the PH domain would be able to interact with PIP₃ in a cell membrane while P-Rex1 is in its autoinhibited conformation. To better understand the molecular consequences of PIP₃ binding, HDX-MS measurements were taken on P-Rex1 in the presence of liposomes ± PIP₃. Without PIP₃, the most notable changes in P-Rex1 in response to liposomes were increases in protection of the lipid binding elements of both DEP domains (primarily their β1/β2 loops; Figure 6—figure supplement 1A), which suggests that they directly interact with lipid bilayers. Indeed, the isolated DEP1 can independently bind negatively charged liposomes (Ravala et al., 2020). Within the PH domain, there was deprotection of the αC helix, but there was no deprotection at the PH–4HB interface, suggesting that it remained intact. In contrast, membranes with PIP₃ caused deprotection of the entire DH domain, the α6-αN linker and the helices at the DH–DEP1 interface (Figure 6A, Figure 6—figure supplement 1B). Although the core of the PH domain showed, as expected, an increase in protection in the presence of PIP₃, the structural elements contacting 4HB became deprotected, as did the PH-binding regions of 4HB. Collectively, these data are consistent with loosening of interdomain contacts and unraveling of at least some fraction of P-Rex1 onto the surface of the liposome.

Full-length human P-Rex1, Cdc42, and DH/PH-DEP1 expression constructs were described previously (Cash et al., 2019; Cash et al., 2016; Ravala et al., 2020). Mutations in DH/PH-DEP1 were created using QuikChange (QIAGEN) and confirmed by DNA sequencing. Mutations in the pCEFL-HA-HaloTag-P-Rex1 WT construct were created by QuikChange II site-directed mutagenesis (Agilent 200523). All constructs were confirmed by sequencing and expression was tested by immunoblot.

Full-length P-Rex1 was transiently expressed in Freestyle 293-F cells and purified as discussed previously (Cash et al., 2019). Briefly, 48 hr after transfection, the cells were harvested and lysed with Cell Lytic M (Sigma). After ultracentrifugation to remove the insoluble fraction, the protein was purified using glutathione agarose resin (Gold Biotechnology Inc). The protein was subjected to TEV cleavage to remove the GST tag and then further purified using a Mono Q 5/50 GL anion exchange column (GE Healthcare Life Sciences). Finally, the protein was purified over an affinity column generated by conjugating human Rac1 to Affi-Gel 10 resin, although for Krios cryo-EM and HDX-MS experiments, this step was omitted.

P-Rex1 DH/PH-DEP1 proteins were expressed and purified as described previously (Ravala et al., 2020). Briefly, His-tagged protein was expressed in Escherichia coli BL21(DE3) cells, which were then lysed using an Avestin Emulsiflex-C3 high-pressure homogenizer. The cell lysate was clarified with high-speed centrifugation, the supernatant was collected, and protein was purified using Ni-NTA resin. The protein was subjected to TEV cleavage to remove the tag. The protein was then purified using a HiTrap SP sepharose column, concentrated, and subjected to size-exclusion chromatography on a Superdex S75 column (GE Healthcare) column. Cdc42 was produced in an unprenylated form in E. coli and purified as previously described (Cash et al., 2016).

HDX-MS experiments were performed as previously described (Cash et al., 2019). Briefly, samples were mixed with D₂O buffer to initiate the HDX reaction and, at various time points, the reaction was quenched with ice cold quench buffer and the samples immediately frozen on dry ice. Samples were thawed at 4°C and subjected to enzymatic digestion on an immobilized pepsin column followed by LC separation and MS analysis. Data were analyzed using HDExaminer (Sierra Analytics, LLC, Modesto, CA). Each sample was analyzed twice by HDX-MS, and the data shown in the graphs of Figure 1B and C represent the average of these experiments. Each image in Figure 4—source data 1 and Figure 6—source data 1 shows one experiment (rainbow plots) or a difference analysis from those experiments (red to blue plots). Only one of the two sets of experiments performed for each condition (±liposomes or ± IP₄) are shown. For each peptide, the average of all five time points was calculated and used to plot the difference data onto the coordinates. Coordinates are colored using a range of –20% (darkest blue, protection) to 20% (darkest red, deprotection). P-Rex1 was used at a concentration of 1.7 mg/ml. IP₄ (Cayman Chemical) was added at a concentration of 100 μM. For experiments with liposomes, liposomes were added at a molar ratio of 1 P-Rex1 to 4000 total lipids. Liposomes were composed of 80:80:1 POPC:POPS:PIP₃ and prepared as previously described (Cash et al., 2019).

Proteins were evaluated for their GEF activity using a fluorescence-based assay (Cash et al., 2019). Briefly, N-methyl-anthraniloyl-GDP (mant-GDP)-loaded soluble Cdc42 was used as a substrate GTPase (2 µM) in a buffer containing 20 mM HEPES pH 8, 100 mM NaCl, 0.5 mM MgCl₂, 100 µM GTP, and reactions were started by addition of P-Rex1 (100 nM). The loss of fluorescence was measured over time at 10 s intervals on a Flexstation 3 plate reader for 40 min. The data was fit to the one-phase exponential decay model in GraphPad Prism with the span (Y_o-plateau) shared among samples. For IP₄ competition curves, GEF assays were carried out in the presence of liposomes containing 2.5 µM PIP₃, as indicated, and 200 µM each of POPC and POPS, prepared as described previously (Cash et al., 2019).

For cryo-EM sample preparation, P-Rex1 was used at a final concentration of 3 μM and n-dodecyl-β-d-maltoside (DDM) was added to a final concentration of 0.08 mM. For samples with IP₄, a final concentration of 40 μM IP₄ was added. A sample of 4 μl was applied to a glow-discharged Quantifoil (1.2/1.3) 300-mesh grid, which was then blotted with filter paper and plunge-frozen into liquid ethane cooled with liquid nitrogen using a Vitrobot Mark IV (Thermo Fisher Scientific) set to 4°C, 100% humidity, 4 s blot, and a force of 10. Micrographs were collected either using Leginon (Suloway et al., 2005) on a Glacios transmission electron microscope (Thermo Fisher Scientific) operating at 200 keV and a K2 Summit direct electron detector (Gatan, Inc) in counting mode (0.98 Å/pixel) at a nominal magnification of ×45,000 or using EPU (Thermo Fisher Scientific) on a Titan Krios transmission electron microscope (Thermo Fisher Scientific) operating at 300 keV and a K3 direct electron detector (Gatan, Inc) in counting mode (1.054 Å/pixel) at a nominal magnification of ×81,000. On the Krios, datasets were collected on both untilted and 30° tilted grids (Table 1).

To overcome the severe preferred orientation problem of our sample on grids, we collected data on 0° and 30° tilted samples on a Krios electron microscope and processed these datasets separately up through 2D classification (Table 1). For each dataset, micrograph assessment, particle picking, and contrast transfer function estimation were performed using Warp (Tegunov and Cramer, 2019). Particle stacks were taken into CryoSPARC (Punjani et al., 2020; Punjani et al., 2017) and extensively cleaned using 2D classification. A final merged particle stack was used for ab initio reconstruction into one class followed by non-uniform refinement to obtain a map at an overall 4.1 Å resolution for the P-Rex1·IP₄ complex. For Glacios datasets, data were processed only through 2D classification.

Initial model building relied on docking existing atomic models for the DH and PH domains of P-Rex1 (PDB entries 5FL1 and 5D3Y) and the Gβγ-binding scaffold (PDB entry 6PCV). 5D3Y was used for the PH domain because the maps were consistent with IP₄ bound to the PIP₃ site of the PH domain. The DEP1 domain was placed using a non-domain swapped atomic model derived from PDB entry 6VSK. The linker between the PH and DEP1 domains and the 4HB domain was built by hand. A Dali search (Holm and Laakso, 2016) using backbone helices of 4HB revealed its topology to be similar to FAT domains, which was then used to adjust the register of each of its four helices. When Alphafold2 (Jumper et al., 2021) became available, it was used to further adjust the modeling of the 4HB domain and associated structural elements in the IP4P domain. Finally, when the cryo-EM structure of P-Rex1 (PDB entry 7SYF) and the crystal structure of the DH/PH-DEP1 module (PDB entry 7RX9) became available, they were used to confirm less certain regions. Final rounds of real space refinement iterating with manual building were performed in Phenix (Adams et al., 2019). Final structure statistics are given in Table 1, and the structure and associated maps were deposited as PDB entry 8TUA and EMDB entry EMD-41621. Raw data were deposited as EMPIAR entry EMPIAR-11967.

UCSF ChimeraX (Pettersen et al., 2021) was used to make figures showing cryo-EM maps. PyMOL (The PyMOL Molecular Graphics System, version 2.5.5, Schrödinger, LLC) was used to create all other structure images.

For in-solution characterization of DH/PH-DEP1 and its variants, SEC-SAXS was conducted at the BioCAT beamline (Sector18) at the Advanced Photon Source, Argonne National Laboratory, using an AKTA Pure FPLC and a Pilatus3 X 1M detector. Purified proteins were injected onto a Superdex 200 Increase 10/300 column at a final concentration of 3–5 mg/ml in 20 mM HEPES, pH 7, 300 mM NaCl, 2% glycerol, and elution from this column flowed into SAXS flow cell for X-ray scattering. Data were collected every 1 s with 0.5 s exposure times at room temperature using 12 keV X-rays (1.033 Å wavelength) and a 3.67 m sample-to-detector distance. The achievable q range for this experimental setup was (0.0043–0.3546 Å).

The scattering data were processed using BioXTAS RAW 1.6.3 software (Hopkins et al., 2017; Figure 3—figure supplement 2A) and used to determine the forward scattering I(0) and the radius of gyration, R_g via Guinier analysis (Figure 3—figure supplement 2B). The Kratky plot showed proteins to be flexible (Figure 3—figure supplement 2C), leading to unsuccessful rigid modeling efforts. However, use of the EOM within ATSAS/3.0.5-2 aided in generating ensembles representing distinct conformational states of the DH/PH-DEP1 fragment and its variants A170K, L177E, and I409A in solution at equilibrium (Ravala et al., 2020). For EOM, the models were generated using crystallographic coordinates from their respective crystal structures: PDB 5FI1 and PDB 6VSK. EOM generated 50,000 possible profiles for the full pool using default settings and native-like structures. From these profiles, a sub-ensemble that matches the experimental scattering data is selected by a genetic algorithm run 10 times using default settings to verify the stability of the results (results from 1 run are shown in Figure 3E and F). The pair distance distribution function P(r), which provides maximum particle dimension of each protein, was calculated using GNOM (DI, 1992) from the ATSAS 2.8.4 package (Franke et al., 2017; Figure 3—figure supplement 2D).

For quantitative analyses of the flexibility of the selected ensembles, R_flex and R_σ metrics were derived from EOM data. WT DH/PH-DEP1 and A170K show R_flex values smaller than those of the pools, which indicates that these proteins do not exhibit a fully flexible conformation. L177E and I409A variants exhibit R_flex values close to those of the pool, suggestive of being highly flexible. Since Guinier analyses and the normalized residual fits of the proteins show that SAXS data quality is good (Figure 3—figure supplement 2B), the R_σ value >1 is due to flexibility in the protein, consistent with the EOM analysis (Figure 3E and F, Figure 3—figure supplement 2A). Results of the SAXS analysis are presented in Table 3 in accordance with the revised guidelines for publishing SAXS data (Trewhella et al., 2017). The SAXS data are deposited in SASBD (https://www.sasbdb.org/) with accession codes SASDUF2 for DH/PH-DEP1 WT, SASDUG2 for DH/PH-DEP1 A170K, SASDUH2 for DH/PH-DEP1 L177E, and SASDUJ2 for DH/PH-DEP1 I409A.

ThermoFluor experiments were performed on a QuantStudio 5 Real-Time PCR system in duplicate with n = 3 independent experiments. Purified DH/PH-DEP1 and its variants were incubated at 1 mg/ml in a buffer containing 20 mM HEPES pH 7.0, 300 mM NaCl, and 2 mM DTT with 2.5× Sypro Orange dye in a final volume of 10 μl in a 384-well PCR plate. Fluorescence was monitored as a function of temperature, and T_m was determined by fitting the fluorescence data to a sigmoidal curve and calculating the inflection point in GraphPad Prism.

HEK293T cells seeded in 12-well plates coated with poly-d-lysine were transfected with 500 ng of empty vector pHTN HaloTag CMV-neo (Promega G7721) or pCEFL-HaloTag-P-Rex1 constructs and co-transfected with 500 ng of SRE-firefly luciferase and 50 ng of Renilla luciferase plasmids. Thirty-six hours after transfection, the cells were serum-starved overnight and then luminescence signal was measured using Dual-Glo assay system (Promega E2920) according to the manufacturer’s instructions. Firefly-luminescence reads were normalized with Renilla-luminescence signal and adjusted to the negative control.

HeLa cells were lentiviral transduced with pLentiCRISPRv2 P-Rex1 guide RNA3 - AGGCATTCCTGCATCGCATC (GenScript SC1678). Forty-eight hours after transduction, HeLa cells were selected with puromycin (3 μg/ml) for 7 days. P-Rex1 KO was confirmed by western blot. Chemotactic migration was measured by trans-well assays (Thermo Scientific 140629). Inserts of 24-well plates were pre-treated with fibronectin (50 μg/ml) for 3 hr at 37°C. Subsequently, 5 × 10⁴ HeLa cells prepared in serum-free DMEM were plated on the inserts at the upper chamber. Human CXCL12 (50 ng/ml) and EGF (50 ng/ml) (Sigma-Aldrich SRP3276 and SRP6253) were prepared in serum-free DMEM and used as chemoattractant in the lower chamber. Serum-free DMEM was used as negative control. The plates with the inserts were incubated in a humidified atmosphere at 37°C and 5% CO₂ for 6 hr. After incubation, the cells at the upper surface of the membrane insert were carefully removed and the cells attached to the lower surface were gently washed with PBS and then fixed with 4% paraformaldehyde for 15 min. After fixation, the cells were gently washed with PBS followed by staining with 0.5% crystal violet for 20 min. Excess of crystal violet was removed by gentle PBS washes. Migrating cells were imaged using an inverted microscope. Quantification of particles corresponding to migrating cells was performed with FIJI software.

Protein samples prepared in Laemmli buffer were separated with SDS-PAGE using 4–12% gradient gels followed by transfer to PVDF membranes. The membranes were blocked using 5% non-fat milk in TBS-0.05% Tween20 (TBST) and incubated overnight at 4°C with primary antibodies against P-Rex1 and GAPDH (Cell Signaling Technology #13168 and #5174, respectively). The membranes were washed three times with TBST and then incubated with secondary antibodies in blocking solution for 2 hr at room temperature. After washing three times with TBST, the reactive bands were visualized using ECL detection reagents and CL-X-posure films.

GEF assays described in this study were performed with n ≥ 3 replicates, with statistical significance determined using one-way ANOVA test with a post hoc Dunnett’s test for multiple comparisons. Luciferase assays were conducted in triplicate, with 5 measurements for each variant, whereas chemotaxis experiments were from at least three independent measurements. In both cases, significance was determined using multiple comparison ANOVA followed by Šidák statistic test.