Structure of the metastatic factor P-Rex1 reveals a two-layered autoinhibitory mechanism

Abstract

P-Rex (PI(3,4,5)P₃-dependent Rac exchanger) guanine nucleotide exchange factors potently activate Rho GTPases. P-Rex guanine nucleotide exchange factors are autoinhibited, synergistically activated by Gβγ and PI(3,4,5)P₃ binding and dysregulated in cancer. Here, we use X-ray crystallography, cryogenic electron microscopy and crosslinking mass spectrometry to determine the structural basis of human P-Rex1 autoinhibition. P-Rex1 has a bipartite structure of N- and C-terminal modules connected by a C-terminal four-helix bundle that binds the N-terminal Pleckstrin homology (PH) domain. In the N-terminal module, the Dbl homology (DH) domain catalytic surface is occluded by the compact arrangement of the DH-PH-DEP1 domains. Structural analysis reveals a remarkable conformational transition to release autoinhibition, requiring a 126° opening of the DH domain hinge helix. The off-axis position of Gβγ and PI(3,4,5)P₃ binding sites further suggests a counter-rotation of the P-Rex1 halves by 90° facilitates PH domain uncoupling from the four-helix bundle, releasing the autoinhibited DH domain to drive Rho GTPase signaling.

Fig. 1. Crystal structure of autoinhibited P-Rex1 DH-PH-DEP1.

a, P-Rex1 domain layout. b, P-Rex1 GEF activity increases upon truncation of the C-terminal domains. Activity of indicated P-Rex1 variants monitored at 100 nM using mant-GDP activity assay. For timecourse graphs, symbols show mean and error bars show s.d. of n = 3 independent experiments conducted in duplicate. For bar graphs, symbols show rate constant (k_obs) from independent experiments, bars show mean and error bars show s.d. (n = 3). ***P < 0.001 versus full-length (P = 0.0082 for DH-PH-DEP1 and P < 0.0001 for DH-PH); ^^^ P = 0.0005 versus DH-PH-DEP1; repeated measures one-way ANOVA with Tukeyʼs multiple comparisons test. Numerical data for graphs in b are available as source data. c, Domain layout of the ^ΔN40DH-PH-DEP1^T4L structure highlighting the placement of T4L in the β₃-β₄ PH domain loop. d, Crystal structure of the P-Rex1 DH-PH-DEP1 (residues 41–305–T4L–323–502) domains in a closed conformation highlighting the DH domain hinge helix and DEP1 latch regions. The placement of T4L is indicated. However, the domain was too flexible to be accurately placed or built into the electron density maps. Instead, comparison of ion-exchange profiles indicates that T4L permitted the purification of a homogenous protein preparation (Extended Data Fig. 2a–d). e, Rotated view of d. f,g, Comparison of the autoinhibited P-Rex1 DH-PH-DEP1 structure (f) with the active P-Rex1 DH-PH:Rac1 (ref. ) complex (PDB 4YON) (g). Upon transition from the autoinhibited to active states, the DH domain hinge helix opens by around 126° flipping the DH domain away from the PH domain to expose the Rac1 binding site. For clarity, the DEP1 domain (light gray and yellow) is modeled onto the active P-Rex1 DH-PH:Rac1 (ref. ) structure (PDB 4YON) to illustrate its position. h, Comparison of the DH hinge helix in the open and closed conformations. Source data

Fig. 2. Structural basis of DH domain autoinhibition.

a, Structure of DH-PH-DEP1 with the DH domain hinge region and the DEP1 latch region indicated. b, Close-up of the DH domain hinge region (rotated relative to a) highlighting the central Thr240 buried at the hinge point. Side chains of select interfacing residues are shown as stick and hydrogen bonds as dotted lines. c, Close-up of the DEP1 latch interface (rotated relative to a) highlighting the positioning of the DH domain in a cleft formed by the DEP1 domain and the PH-DEP1 linker helix. Side chains of select interfacing residues are shown as sticks and hydrogen bonds as dotted lines. d–e, GEF activity assay of the DH-PH-DEP1 domain demonstrating structure-guided hyperactivating DH-PH-DEP1 mutants. Activity of indicated P-Rex1 variants monitored at 20 nM using mant-GDP activity assay. Combination of DH hinge and DEP1 latch mutants provide an additive increase in P-Rex1 activity over either mutation alone. For timecourse graphs (d), symbols show mean and error bars show s.d. of n = 3 independent experiments conducted in duplicate. For bar graphs (e), symbols show rate constant (k_obs) from independent experiments, bars show mean and error bars show s.d. (n = 3). *P < 0.05 and **P < 0.01 versus DH-PH-DEP1 (P = 0.0071 for L177E, P = 0.0115 for L178E, P = 0.0494 for T240N, P = 0.0079 for T240K, and P = 0.0219 for L178E/T240N), repeated measures one-way ANOVA with Dunnett’s multiple comparisons test. Numerical data for graphs in d and e are available as source data. Source data

Fig. 3. Cryo-EM structure of full-length autoinhibited P-Rex1.

a, Cryo-EM map and structure of full-length P-Rex1. Map density for T4L is omitted for clarity (Extended Data Fig. 8) b, Inset, 2D classification from an equivalent molecular view for reference. c, Multiple indicated P-Rex1 views with domains highlighted. d, BS³ crosslink constraints (FDR = 6.8 %) from wild-type full-length P-Rex1 mapped onto the pipes-and-planks depiction of the full-length autoinhibited P-Rex1 model indicates numerous close range spatial constraints that are consistent with the modeled domain topology shown in a–c. Satisfied crosslinks are colored blue (<30 Å between lysine Cβ atoms), while crosslinks that exceed the maximum allowable distance (indicative of conformational flexibility) are in red (>30 Å between lysine Cβ atoms). Lysine Cβ atoms are shown as blue spheres. e, Circos plot of BS³ crosslinking mass spectrometry of full-length wild-type P-Rex1. Dashed line (black) indicates position of an IDL (located on the tip of the 4HB) removed in the cryo-EM construct.

Fig. 4. Model of synergistic P-Rex1 activation by Gβγ and PI(3,4,5)P₃ at the plasma membrane.

a, Cytoplasmic P-Rex1 is autoinhibited under basal conditions. A hinge and latch mechanism locks the DH domain in a closed conformation where the PH domain sterically occludes the Rac1 binding site. In the autoinhibited P-Rex1 structure, the known membrane-binding regions are off-axis by around 90°. b, Concurrent binding of PI(3,4,5)P₃ (ref. ) and Gβγ at the membrane requires counter-rotation of the N-terminal (DH-PH-DEP1) and C-terminal (DEP2-PDZ1/2-IP4P) P-Rex1 halves to align the membrane-binding regions along a single plane. c, We hypothesize that the counter-rotation mechanism provides the conformational change required to release the DEP1 latch from the DH domain to activate P-Rex1. Structural analysis indicates that rotation-induced PH domain movement toward the membrane causes the PH-DEP1 linker helix to clash with the DH domain and likely triggers DEP1 latch release (Extended Data Fig. 10).

Extended Data Fig. 1. P-Rex1 signalling, variant purification, and activity analysis.

a. Schematic of P-Rex1 activation. P-Rex1 co-ordinates signalling from GPCRs and RTKs. The key GPCR and RTK effectors, Gβγ and PI(3,4,5)P₃, respectively, bind to and activate P-Rex1 (solid arrows). Bidirectional cross-talk between GPCRs and RTKs may provide additional pathways to P-Rex1 activation (dotted arrows). For example, Gβγ subunits can generate PI(3,4,5)P₃ via activation of PI3K; GPCRs can transactivate RTKs to increase PI(3,4,5)P₃; and RTKs can activate G proteins to release Gβγ subunits^,. b. Coomassie-stained SDS-PAGE analysis of purified full-length P-Rex1, P-Rex1 DH-PH-DEP1 (residues 1-502), and P-Rex1 DH-PH (residues 1-404). c-e. P-Rex1 GEF activity increases upon truncation of the C-terminal domains. Activity of indicated P-Rex1 variants monitored at 20 nM and 100 nM using mant-GDP activity assay. Symbols show mean and error bars show S.D. (n = 3). f. Rate constant (k_obs) of mant-GDP activity determined from (c-e). Symbols show k_obs from independent experiments, bars show mean and error bars show S.D. ^ p = 0.0165 and ^^^ p < 0.0001 versus 20 nM, two-way ANOVA with Šídák’s multiple comparisons test. Numerical data for graphs in c-f are available as source data. Source data

Extended Data Fig. 2. Purification of P-Rex1 DH-PH-DEP1^T4L.

a. MonoS ion-exchange chromatography profile of purified P-Rex1 DH-PH-DEP1^T4L. b. Coomassie-stained SDS-PAGE analysis of fractions indicated in (a). Representative SDS-PAGE from two independent purifications. c. Comparative MonoS ion-exchange chromatography profile of purified P-Rex1 DH-PH-DEP1. d. Coomassie-stained SDS-PAGE analysis of fractions indicated in (c). Representative SDS-PAGE from three independent purifications. Profiles suggest that T4L insertion promoted the purification of a more homogenous protein preparation. e. Ion-exchange and f. size exclusion chromatography profiles of the ^ΔN40P-Rex1 DH-PH-DEP1^T4L protein utilised to obtain the final 3.2 Å crystal structure. g. Coomassie-stained SDS-PAGE analysis of ^ΔN40DH-PH-DEP1^T4L purification (1) nickel elution, (2) overnight (o/n) TEV digest, (3) pellet after o/n TEV and phosphatase treatment, (4) supernatant after o/n TEV and phosphatase digest, (5-6) main-peak MonoS fractions, (7-9) main-peak size exclusion chromatography fractions. Representative SDS-PAGE from three independent purifications.

Extended Data Fig. 3. Electron density maps of P-Rex1 ^ΔN40DH-PH-DEP1^T4L.

a-d. 2Fo-Fc maps of the P-Rex1 DH-PH-DEP1 structure contoured between 1-1.5σ. Inset regions highlighting the density in the b. DEP1 domain, c. the DH hinge helix, and d. the PH domain. Electron density for T4L was diffuse, preventing the confident placement and refinement of a T4L model. As such, no contacts between T4L and P-Rex1 (either within the P-Rex1 chain or across the crystal lattice) were observed that could affect the conformation of the crystal structure. The T4L domain appears to be positioned within a solvent cavity in the crystal, enabling a high degree of mobility. Regardless, T4L was essential for crystal formation.

Extended Data Fig. 4. Cross-linking mass spectrometry analysis of P-Rex1.

BS³ cross-linking constraints and Circos plots of cross-linking mass spectrometry are shown for a. the isolated DH-PH domains, b. the DH-PH^T4L domains, c. the DH-PH-DEP1 domains, d. the DH-PH-DEP1^T4L domains, e. full-length wild-type (WT) P-Rex1, and f. the full-length ^ΔN40P-Rex1^{T4L, Δloop} construct utilised for cryo-EM studies. Loop refers to residues 1119-1211. Lysine Cβ atoms are shown as blue spheres with blues lines indicating a compatible constraint (<30 Å between lysine Cβ atoms) and red lines indicating an incompatible constraint (>30 Å between lysine Cβ atoms). Constraints are mapped onto active (left) or autoinhibited (right) models of the DH-PH-DEP1 domains. Cross-links are in excellent agreement with the closed conformation observed in our DH-PH-DEP1 crystal structure. These data indicate that the closed conformation is stably populated in solution in the presence or absence of T4L. Conversely, cross-links frequently exceeded the allowable constraint distance when modelled on the active conformation. Full-length P-Rex1 displays a similar cross-linking pattern in the absence or presence of T4L (e-f). Interestingly, incompatible constraints (red lines) cluster across the DEP1–DEP2 linker region in full-length P-Rex1 indicating potential conformational flexibility across this interface. Dashed line (black) indicates position of an intrinsically disordered loop (IDL) and the FDR is indicated for each dataset.

Extended Data Fig. 5. GEF activity of T4L P-Rex1 insertion variants or its domains and validation of DH-PH-DEP1 structure via mutagenesis.

a. GEF activity of P-Rex1^T4L increases upon truncation of the C-terminal domains in a comparable pattern to wild-type P-Rex1. Activity of ^ΔN40P-Rex1^{T4L,Δ1119-1211}, DH-PH-DEP1^T4L, and DH-PH^T4L variants monitored at 100 nM using mant-GDP activity assay. For timecourse graphs, symbols show mean and error bars show S.D. of n = 3 independent experiments conducted in duplicate. For bar graphs, symbols show rate constant (k_obs) from independent experiments, bars show mean and error bars show S.D. (n = 3). * p < 0.05 versus full-length^T4L (p = 0.0392 for DH-PH-DEP1^T4L and p = 0.0141 for DH-PH^T4L); ^^ p = 0.0097 versus DH-PH-DEP1^T4L; repeated measures one-way ANOVA with Tukey’s multiple comparisons test. b. Coomassie-stained SDS-PAGE analysis of the indicated P-Rex1^T4L variants. c. Plot of relative α-helix propensity of each residue in the DH domain hinge helix for P-Rex1 and P-Rex2. A loss of α-helical propensity is observed in conserved residues surrounding the hinge-point (Ile237). d. Alignment of the DH domain hinge α₆-helix in the autoinhibited (closed) structure and the active (open) conformation (PDB 4YON). Structure in ball and stick format with the Cα atoms as spheres. e-f. Coomassie-stained SDS-PAGE analysis of the indicated P-Rex1 DH-PH-DEP1 mutants. g-h. GEF activity of the indicated DH-PH-DEP1 mutants. Activity of indicated P-Rex1 variants monitored at 20 nM using mant-GDP activity assay. For timecourse graphs, symbols show mean and error bars show S.D. of n = 3 independent experiments conducted in duplicate. For bar graphs, symbols show rate constant (k_obs) from independent experiments, bars show mean and error bars show S.D. (n = 3). * p < 0.05, ** p < 0.01, and *** p < 0.001 versus DH-PH-DEP1 (p = 0.0431 for M401A/M408A, p = 0.0063 for T240K, p = 0.0398 for T240S, p = 0.0355 for L177A/L178A and p = 0.0009 for T240K), repeated measures one-way ANOVA with Dunnett’s multiple comparisons test. Numerical data for graphs in a, g and h are available as source data. Source data

Extended Data Fig. 6. Cryo-EM image processing workflow and strategy.

Initial data collection by Arctica and pre-processing indicated quality particles with density for the N-terminal module. An ab-initio volume and 2D class averages were used to bootstrap 3D classification and particle picking, respectively. A topaz model was trained using the top 100 micrographs ranked by number of quality particles. This was subsequently used to pick roughly 0.9 million particles. Template matching from re-projected templates of the ab-initio volume was performed yielding 6.5 million particles. Multiple rounds of 2D classification of this combined set yielded 0.9 million particles with good secondary structure. Two rounds of 3D classification with incrementally higher angular sampling was performed, and a single class was selected for refinement in RELION with SIDESPLITTER. Local particle motion and CTF were further refined by Bayesian polishing and CTF refinement. Two modules of P-Rex1 showed notable flexibility and were therefore separated for local refinements. The final maps were sharped using DeepEMhancer (1.0) to suppress effects of anisotropy and visualise high-resolution features.

Extended Data Fig. 7. Cryo-EM reconstruction summary and validation.

a. Warp-denoised K3 micrograph of vitrified ^ΔN40P-Rex1^{T4L,Δ1119-1211}. Representative micrograph from a dataset of 9597 movies. b. RELION 2D class averages (no alignments) after particle polishing. c. Comparison of 2D class averages between autoinhibited P-Rex1 (without T4L, ^ΔN40P-Rex1^{Δ1119-1211, Δ990-1007}), P-Rex1 (with T4L, ^ΔN40P-Rex1^{T4L,Δ1119-1211}), and a single class from the previously published P-Rex1:Gβγ dataset (EMPIAR 10285) showing all three constructs form an equivalent closed domain conformation. d. Final composite reconstruction coloured according to local resolution as estimated by windowed FSC with 0.5 threshold criterion. e. Local resolution histogram showing distribution for both localised refinements and the consensus. f. Angular distribution of final reconstruction and corresponding views (left). g. Gold-standard Fourier Shell Correlation curves of independent half-maps (masked, unmasked, noise substitution corrected, and phase randomised) for both the C-terminal and N-terminal localised reconstructions. h. As in (g) for the full consensus reconstruction.

Extended Data Fig. 8. Density to model agreement.

a. Three rotated views of the cryo-EM reconstruction of ^ΔN40P-Rex1^{T4L,Δ1119-1211}. Position of T4L is shown in purple. b. Select regions of high-resolution highlighting clear sidechain density c. Agreement between P-Rex1 model and cryo-EM density for each P-Rex1 domain. Regions of poor density due to anisotropy are shown with two views. Arrow indicates direction of the projection axis of the preferred orientation.

Extended Data Fig. 9. ConSurf and electrostatic surface analysis of full-length P-Rex1.

a. Three 90° rotated views (surface rendering) of P-Rex1 coloured according to surface conservation. b. Equivalent views coloured according to surface charge. c. Pipes-and-planks depiction in equivalent orientations as in (a). d. Split interface of PH and 4HB domain interaction reveals a strongly conserved interface. e. Focused view of the IP4P conserved pseudo-active site. f. Focused view of the PH domain PI(3,4,5)P₃ binding site. Highly conserved surface regions are observed on the DH, DEP1 and PH domains. These correlate with positively charged surface regions that may collectively mediate interactions with the negatively charged inner leaflet of the plasma membrane, as shown for the isolated P-Rex1 PH domain.

Extended Data Fig. 10. Proposed molecular mechanism of P-Rex1 synergistic activation by Gβγ and PI(3,4,5)P_3.

a. Autoinhibited P-Rex1 may transiently interact with the inner leaflet of the plasma membrane via DEP1, DEP2, PH or the IP4P positively charged surfaces. b. Collectively, binding of Gβγ and DEP1, DEP2 and IP4P requires the counter-rotation of the N-terminal module (DH-PH-DEP1) relative to the C-terminal module (DEP2-PDZ1/2-IP4P:Gβγ). Movement of the PH domain toward the membrane is essential to unlock the autoinhibited DH domain. c. Counter-rotation of N- and C-modules disengages the 4HB and PH interaction, freeing the PH domain to bind PI(3,4,5)P₃. PI(3,4,5)P₃ binding causes the DH domain latch to unlock, thereby releasing the DH domain to bind Rac1. Autoinhibited P-Rex1 is modelled by the full-length cryo-EM reconstruction and AlphaFold (4HB regions). The cryo-EM reconstruction of Gβγ:P-Rex1 DEP2-PDZ1/2-IP4P allows placement of Gβγ:DEP2-PDZ1/2-IP4P against the bilayer (revealing offset planes defined by N- and C-modules). The crystal structure of PH: PI(3,4,5)P₃, as well as functional data of DEP1/2 membrane binding regions, guided the placement of the AlphaFold model of PH-DEP (kinked V-conformation) against the membrane. Lastly, superposition of the crystal structure of DH-PH:Rac1 onto the membrane-anchored PH domain provides a model of the full active state. The Rac1 lipidation site provides additional short-range distance constraints on the placement of DH-PH:Rac1.