Structural and Biochemical Characterization of the Catalytic Core of the Metastatic Factor P-Rex1 and Its Regulation by PtdIns(3,4,5)P3  Jennifer N. Cash, Ellen M. Davis, John J.G. Tesmer  Structure  Volume 24, Issue 5, Pages 730-740 (May 2016) DOI: 10.1016/j.str.2016.02.022 Copyright © 2016 Elsevier Ltd Terms and Conditions

Structure 2016 24, 730-740DOI: (10.1016/j.str.2016.02.022) Copyright © 2016 Elsevier Ltd Terms and Conditions

Figure 1 Interactions of the P-Rex1 PH Domain (A) Primary structure of P-Rex1. (B) Ins(1,3,4,5)P4 (IP4), a soluble analog of PIP3, binds in a pocket formed at one end of the PH domain, suggesting that the membrane plane would in this view be found along the top of the panel. The adjacent basic β3/β4 loop is disordered in all of our crystal structures. (C) The binding pocket is lined with basic residues that tightly coordinate the 3- and 4-position inositol phosphates. A 3σ omit map for Ins(1,3,4,5)P4 is shown as a blue wire cage. See also Figures S1, S5, and S6. Structure 2016 24, 730-740DOI: (10.1016/j.str.2016.02.022) Copyright © 2016 Elsevier Ltd Terms and Conditions

Figure 2 DSF Reveals PH Domain Residues Important for Binding and Specificity for Ins(1,3,4,5)P4 (A) The independent PH domain exhibits a melting temperature of 47.4°C, and Ins(1,3,4,5)P4 induces a ΔTm of +11°C. None of the mutants exhibited basal Tm values that were appreciably different from WT (data not shown), suggesting that each variant was properly folded. Shown are the combined data from three experiments performed in triplicate. Errors bars indicate SD. (B) WT P-Rex1 PH domain exhibits a stronger interaction with the headgroup of PIP3 than with other inositol phosphate headgroups. Shown are the combined data from three experiments performed in triplicate. Error bars indicate SEM. (C) The P-Rex1 PH domain exhibits preference for binding Ins(1,3,4,5)P4 over other inositol phosphate headgroups. DSF was performed on WT P-Rex1 PH domain mixed with either 0.05 mM Ins(1,3,4,5)P4, 1.25 mM other headgroup, or a mixture of the two, as indicated. Shown are the combined data from two experiments performed in triplicate. Error bars indicate SEM. See also Figure S2. Structure 2016 24, 730-740DOI: (10.1016/j.str.2016.02.022) Copyright © 2016 Elsevier Ltd Terms and Conditions

Figure 3 P-Rex1 Binding PIP3 in Cells Is Highly Correlated with Activity (A) Mutations that reduce ΔTm by Ins(1,3,4,5)P4 binding also reduce activity of P-Rex1. Luciferase-gene reporter assays were performed in HEK293T cells with essentially full-length P-Rex1 constructs containing the same point mutations tested using DSF on the independent PH domain (Figure 2). Shown is the average of the 20 ng DNA data points from at least two experiments, performed in triplicate. See also Figure S3. (B–E) Reducing PIP3 accumulation reduces P-Rex1 (B and D) but not LARG (C and E) activity. Luciferase-gene reporter assays were performed in HEK293T cells by titrating in DNA for WT P-Rex1 or a LARG construct. In (B) and (C), cells were treated with 1 μM wortmannin to inhibit PI3K activity and PIP3 accumulation. Residual apparent activity is likely due to accumulation of luciferase before treatment with wortmannin. In (D) and (E) PTEN DNA was titrated into cells to reduce PIP3 levels. Shown are the combined data from at least three experiments performed in triplicate. ∗∗p < 0.005, ∗∗∗p < 0.0005. Error bars indicate SEM. Structure 2016 24, 730-740DOI: (10.1016/j.str.2016.02.022) Copyright © 2016 Elsevier Ltd Terms and Conditions

Figure 4 Determinants of Membrane Localization of Full-Length P-Rex1 EGFP-tagged P-Rex1 variants were expressed in HEK293T cells, and membrane- and cytosol-associated proteins were separated. EGFP-tagged protein in each fraction was quantified by fluorescence and normalized to total EGFP-tagged protein expressed. Data shown represent the average of at least three experiments. ∗p < 0.05, ∗∗p < 0.005. Error bars indicate SEM. (A) The relative amount of each variant associated with the membrane was compared with that of WT, which was set to 100%. EGFP (2.3% of the total was associated with the membrane) and a LARG construct engineered to strongly associate with the membrane via binding PtdIns(4,5)P2 (65% of the total was associated with the membrane) were used as controls. For WT P-Rex1, 18% of the total localized to the cell membrane, similar to what had been found previously (23% of total) in a related cell-based assay (Barber et al., 2007). Reduction in PIP3 accumulation by (B) treatment with wortmannin or (C) co-expression of PTEN has no effect on localization of P-Rex1 to the membrane fraction. ∗p < 0.05, ∗∗p < 0.005. (D) The β3/β4 loop of P-Rex1 is a non-specific membrane-anchoring element. Deleting the β3/β4 loop from P-Rex1 significantly reduces membrane localization, whereas adding this sequence to the β3/β4 loop of P-Rex2 significantly increases membrane localization. Structure 2016 24, 730-740DOI: (10.1016/j.str.2016.02.022) Copyright © 2016 Elsevier Ltd Terms and Conditions

Figure 5 Overview and Comparison of P-Rex1 DH/PH-GTPase Complexes (A) Cdc42 and Rac1 bind to the DH domain in a canonical fashion and form no contacts with the PH domain. See also Figures S5 and S6. (B) A large degree of flexibility in the α6-αN linker that joins the DH and PH domains is observed among the five unique DH/PH-GTPase structures reported in this paper. To generate this figure, we aligned the DH domains using Coot. The Cα positions of Ala283 (β1/β2 loop) and Lys366 (β6/β7 loop) in the P-Rex1 PH domain are shown as spheres to emphasize the different positions of the PH domain in each of the unique chains. The highlighted structure is that of the DH/PH-Cdc42 complex, and the others are the four unique DH/PH-Rac1 complexes. See also Figure S4. (C) RhoGEFs that act with specificity toward multiple GTPases may do so through adaptive interfaces formed between the DH α5 helix and variable α4/α5 loop with the specificity patch of the GTPase. The side chain of RhoA Glu54 is shown in gray to illustrate one of the clashes that might occur with the α4/α5 loop of the DH domain and thereby disfavor RhoA binding. See also Figure S7. Structure 2016 24, 730-740DOI: (10.1016/j.str.2016.02.022) Copyright © 2016 Elsevier Ltd Terms and Conditions

Figure 6 P-Rex1 Activation by PIP3 (A) Model of the catalytic core of activated P-Rex1 at the membrane. The C terminus of the bound GTPase, the bound PIP3 molecule, and the β3/β4 loop (blue plus signs), which all contain membrane-interacting elements (Figure 4), are modeled adjacent to the putative membrane surface at the top of the panel. Dashed lines indicate flexible protein or lipid components. (B) Model of activation by PIP3. Our data indicate that PIP3 binding is required for P-Rex1 activity in HEK293T cells but not for membrane localization. We propose that, in its basal state, P-Rex1 is in equilibrium with the membrane, with binding favored by the basic β3/β4 loop (shown in blue). P-Rex1 association with the membrane becomes stabilized through some yet undetermined process, perhaps by binding Gβγ. Subsequently, PIP3 binding to the PH domain induces a conformational change that frees the DH domain to bind a Rho GTPase. We speculate that, in its inactive state, the DH/PH core is in a kinked conformation as observed in inactive Asef and collybistin (Murayama et al., 2007; Xiang et al., 2006). The question mark indicates uncertainty in the location of the Gβγ-binding site, although some evidence points toward the DH/PH core (Hill et al., 2005). The dashed red line indicates a flexible lipid component. Structure 2016 24, 730-740DOI: (10.1016/j.str.2016.02.022) Copyright © 2016 Elsevier Ltd Terms and Conditions