Keywords: SRGAP2, X-ray crystallography, F-BAR domain, human evolution, structural biology

F-BARx Crystallography

We solved the F-BARx crystal structure by applying a variation of an exhaustive molecular symmetry search approach that took advantage of the 2-fold anti-parallel F-BAR conserved symmetry (Sporny et al. 2016). This allowed solution of a structure (refined to 2.7 Å resolution) using a remote search model [FBP17, PDB 2EFL (Shimada et al. 2007)] with only 19% sequence identity, which represented 13% (as a poly-ala model) of the asymmetric unit contents. This molecular symmetry-constrained systematic search approach was vital for structure solution as it enabled the resolution of characteristic problems common for coiled coil proteins (Dauter 2015). The exhaustive search was followed by electron density modification procedures and cycles of model refinement and re-building (table 1). The quality of the resulting electron density has allowed us to ensure correct assignment of all amino acid side chains.

Crystal Structure of F-BARx

The crystal structure reveals that the F-BARx (residues 1–484,supplementary fig. S1,Supplementary Material online) is composed of a noncanonical F-BAR (residues 1–355) domain followed by a coiled-coil domain extension (residues 356–484), designated here as the F-BAR extension (Fx). The F-BARx is the largest BAR superfamily member characterized thus far. The tip-to-tip distance of the F-BARx dimer is approximately 275 Å [electron density in the tip regions is weak due to high B-factor values (supplementary fig. S2B,Supplementary Material online), and cannot be defined accurately], reaching a width of 55 Å. The buried surface interface (BSI) between the two protomers is 8370 Ų. In comparison, the F-BAR domain of FBP17 has 288 residues, a dimeric tip-to-tip distance of 222 Å, a width of 36 Å, and a BSI of 4647 Ų.

Regardless of the size difference, the same structural components that form the F-BAR domains of FBP17, CIP4, and FCHo2, are also found in SRGAP2A. These include three primary helices (α1, α2, and α3) and one short helix (α4) (fig. 2A andB andsupplementary figs. S1 and S2,Supplementary Material online). In SRGAP2A, the carboxy-half of α1 from each protomer, together with approximately one third of each α2 and α3, form a six-helix bundle at the center of the dimer, in which α1 and α3 are engaged in homotypic interactions with their dimeric counterpart.

Two arms protrude from the central helix bundle; each includes the long amino-half of α1, and a coiled-coil composed by the remaining two-thirds of α2 and α3.

The membrane-binding N-surface is composed by α1 and most of α2, together with small portions of α3 and α6. On the opposing side to the N-surface, the short α4 extends directly from α3 and further continues as a long random-coil towards the tip of the second protomer.

A Homer-EVH1 binding motif,³³⁹PPMKF, is located on the α4–α5 connecting loop (supplementary figs. S1 and S3,Supplementary Material online). It was recently demonstrated that through this motif, SRGAP2A directly interacts with the Homer protein in excitatory synapses (Fossati et al. 2016). However, our structural data reveals that because the Phe343 side-chain is buried in a cleft formed between the three main α1, 2, and 3 helices of the reciprocal protomer, it can actually not be engaged with EVH1 domain binding, unless a considerable conformational change takes place in this region. On the Homer side, this assessment is based on the crystal structure and analysis of the Homer EVH1 domain complexed with a peptide from mGluR (TPPSPF) (PDB 1DDV) (Beneken et al. 2000).

The Fx, a 72 Å long coiled-coil composed by α5 and α6, extends from one SRGAP2A protomer and packs against the α2–α3 arm of the reciprocal protomer in the SRGAP2A dimer (fig. 2 andsupplementary fig. S2,Supplementary Material online). The Fx is arranged in such a way that α5 is predominantly surface exposed, and has only minor direct interactions with the F-BAR, while α6 is mostly buried between α5 and the α2–α3 coiled-coil arm. Similar coiled-coil extension with reminiscent packing is also observed in the crystal structure of the Fes-kinase F-BAR domain (PDB 4DYL).

SRGAP2A Has an Extended Dimerization Interface

Like other BAR-superfamily members, the isolated F-BAR domain (F-BARA, residues 1–355) of SRGAP2A is dimeric, as can be demonstrated by size-exclusion chromatography (SEC) (supplementary fig. S4,Supplementary Material online). The F-BARx crystal structure reveals an elaborated dimerization interface that involves the Fx in addition to the F-BAR.

Furthermore, we found that the RhoGAP-SH3 domains are also likely to be involved in dimerization (fig. 2C andD). With the F-BARx crystal structure at hand, we could fit it into the low resolution Small Angle X-ray Scattering (SAXS) structure of the full-length SRGAP4 that we determined recently (Guez-Haddad et al. 2015) (fig. 2D). Adjustment of the F-BARx arms angle by 10° and docking of RhoGAP (the homologous RhoGAP of Beta2-chimaerin, PDB 1XA6;Canagarajah et al. 2004) and SH3 (from our recent work, PDB 4RUG;Guez-Haddad et al. 2015) crystal structures into the densities next to the tips of the F-BARx arms results in a very good fit, using the fit-in-map application in Chimera (Yang et al. 2012), with real-time mean correlation = 9.0438e−05, SD = 0.00066913, RMS = 0.00067521. Domain fitting was also guided by distance constraints between the carboxy end of the Fx, which is visible in the crystal structure, and the amino terminus of the RhoGAP model. The resulting domain arrangement implies that the RhoGAP and SH3 domains interact directly with the F-BAR arms of the reciprocal protomers.

To confirm this observation, we measured the binding affinity of an isolated RhoGAP-SH3 to a fluorescently labeled F-BAR and F-BARx using microscale thermophoresis (MST) (fig. 2C). Our measurements show that the RhoGAP-SH3 domains specifically bind the F-BAR domain withK_D = 3 µM and the F-BARx withK_D = 3.5 µM. This shows that the RhoGAP-SH3 directly interacts with the F-BAR and that the Fx does not contribute to the binding affinity. These results are also consistent with our recent binding measurements of Robo1 to SRGAP2A (Guez-Haddad et al. 2015). Therein, we discovered that the F-BARx, RhoGAP, and SH3 domains cooperatively participate in Robo1 binding, further supporting direct interactions between the three domains.

In conclusion, the dimerization interface in SRGAP2A extends beyond the F-BARs, and includes the Fx and probably the RhoGAP-SH3 domains of one SRGAP2 protomer that wrap around the F-BAR domain of the reciprocal SRGAP2A (fig. 2D).

SRGAP2A Inverse Curvature

As mentioned earlier, it is typical for F-BAR proteins to associate with membrane invaginations, however, in concordance with prior studies, we find that the F-BAR-containing SRGAP2A associates with membrane protrusions (fig. 3B andC) (Frost et al. 2008;Guerrier et al. 2009;Yamazaki et al. 2013;Fritz et al. 2015). In order to explain how this might be, it is necessary to look at the structures of the SRGAP2A F-BAR domain in comparison to that of other F-BAR proteins. In general, the six-helix core region of F-BAR domains is essentially flat, and the overall curved shape of F-BAR dimers is generated by the direction and angle of the α2–α3 arms relative to the central region. In FBP17, CIP4, Syndapins, and FCHo2 the α2-α3 arms are bent towards the direction of the N-surface, thereby generating a concave membrane binding surface. Unlike these, our structural data for SRGAP2A reveal that the α2–α3 arms are bent in the opposite direction, away from the N-surface, resulting in a convex membrane binding surface (fig. 3A). Structure-based sequence comparison between the F-BAR domain of SRGAP2A and its closest available F-BAR homolog, FBP17 (PDB 2ELF, 19% sequence identity), revealed a 25-residue long helical insertion (residues¹²⁷RFV…LQD¹⁵¹) into α2 at the point that links the central region and the α2–α3 arm. Since the bases of the α2–α3 arms are held with α1 in the six-helix core region, the 25 aa insertion would push the α2–α3 arm to point away from the insertion. Indeed, when we removed the 25-residue helix from the F-BARx of SRGAP2A and expressed the truncated construct in COS-7 cells, no membrane protrusions were observed (supplementary fig. S5,Supplementary Material online). Taken together, such inverse geometry of the membrane-binding surface readily explains how, contrary to other F-BAR proteins, SRGAPs associate with membrane protrusions rather than with invaginations (fig. 3).

Membrane Binding Is Driven and Maintained by Electrostatic Interactions

In most BAR superfamily domains, membrane association is driven by electrostatic interactions between negatively charged membrane surfaces and positively charged residues on the membrane-binding surface of the protein (Mim and Unger 2012,Kessels and Qualmann 2015). For the SRGAP2A F-BARx, the convex N-surface is predominantly electropositive, while the opposite concave surface is electronegative (fig. 4A). We identified 46 lysine and arginine side chains that contribute to the electropositive potential of the dimeric N-surface (23 on each protomer). These positively charged residues are spread throughout the N-surface, with higher concentrations around the six-helix center and the two tip regions (fig. 4A andsupplementary fig. S1,Supplementary Material online).

The Fx does not encroach upon the membrane-binding N-surface. In contrast, it constitutes a lateral addition that generates a transverse curvature, in addition to the more conspicuous longitudinal F-BARx curvature. To evaluate whether, and to what extent, the Fx contributes to membrane binding we compared liposome cosedimentation with F-BARA (residues 1–355,fig. 1B) to that of the F-BARx (residues 1–484). The results show that the presence of the Fx increases liposome binding by ∼25% (fig. 4B andsupplementary fig. S6,Supplementary Material online). This result demonstrates that although the F-BAR is sufficient for membrane binding, the Fx strengthens membrane-association for SRGAP2A.

In order to experimentally test whether membrane binding is indeed mediated by electrostatic interactions, we first measured SRGAP2A F-BARx binding to liposomes under different conditions of phosphate concentration. In line with electrostatic interactions, binding was weaker as phosphate concentrations in the reaction buffer increased. About 140 mM phosphate almost completely dissociated F-BARx from liposomes after their prior association (fig. 4C). These experiments show that the F-BARx membrane association is primarily driven and maintained by electrostatic interactions. To confirm that the convex N-surface is the membrane-binding surface, we substituted five arginine and lysine residues with glutamates. The resulting RKKKR 54,55,234,235,238 EEEEE (designated F-BARx-R5E) contains charge alterations of three residues on each of the tips, and four in the six-helix center of the dimer (fig. 4A andsupplementary fig. S1,Supplementary Material online). Liposome cosedimentation assays demonstrated that membrane binding of F-BARx-R5E was abolished (fig. 4B andsupplementary fig. S6,Supplementary Material online), as was its ability to form membrane protrusions when expressed in COS-7 cells (fig. 4D).

Taken together, these experiments verify that the principle of membrane binding through the N-surface is retained in SRGAP2A, and demonstrate that membrane binding is a prerequisite for membrane scaffolding.

SRGAP2C Is Insoluble and Inflicts Insolubility on Coexpressed SRGAP2A

With the SRGAP2A F-BARx crystal structure at hand, we could readily investigate the molecular mechanism by which SRGAP2C antagonizes the activity of SRGAP2A.

SRGAP2C roughly spans the F-BARx of SRGAP2A, with three differences (fig. 1). First, SRGAP2C is missing exon 10 onwards. The F-BARx crystal structure now reveals that this leaves a carboxy truncation that disturbs the Fx fold. Second, it has a seven-residue (VRECYGF) carboxy-terminal addition, translated from intron 9. Third, it has five arginine alterations (R73H, R108W, R205C, R235H, and R250Q), all located on the F-BAR domain and none on the Fx. The first two modifications were inflicted simultaneously by the partial duplication ofSRGAP2A (∼3.4 Ma) (Dennis et al. 2012;Dennis and Eichler 2016) and were therefore present in the primal forms:P-SRGAP2B andP-SRGAP2C (fig. 1A). Because each of the contemporary forms (SRGAP2B andSRGAP2C) have different nonsynonymous substitutions, we conclude that all their substitutions (two on SRGAP2B and five on SRGAP2C) were introduced afterSRGAP2C duplication fromSRGAP2B (∼2.4 Ma). As the contemporary F-BARx region in the human SRGAP2A sequence is 100% identical to those of chimpanzees (supplementary fig. S1,Supplementary Material online), bonobos and orangutans, we also conclude that SRGAP2A was unchanged at least in the past 6 My of human evolution (fig. 1A).

We first expressed SRGAP2C alone in COS-7 and Sf9 cells. SRGAP2C failed to induce membrane protrusions in COS-7 cells, displaying a granulated distribution in the cells' cytoplasm, and unlike SRGAP2A, was insoluble in the Sf9 culture (fig. 5A–C). We re-created P-SRGAP2C, which is an SRGAP2C that lacks the five arginine substitutions and is identical to P-SRGAP2B (fig. 1B andsupplementary fig. S1,Supplementary Material online). P-SRGAP2C was also insoluble in Sf9 culture (fig. 5D). Consistently with our findings, Polleux and colleagues showed that an F-BARx that is missing exon 10, but lacks the five arginine substitutions does not generate membrane protrusions in COS-7 cells (Guerrier et al. 2009).

Analysis of the F-BARx crystal structure explains the reason for SRGAP2C and P-SRGAP2C insolubility. Since the part of helix 6 that is encoded by exon 10 supports the Fx structure, its absence in SRGAP2C (and P-SRGAP2C) would inflict Fx misfolding and insolubility due to exposure of significant hydrophobic patches (fig. 5E).

The insolubility of SRGAP2C and P-SRGAP2C is in sharp contrast to the full solubility of SRGAP2A (fig. 5). However, when coexpressed with either SRGAP2C or P-SRGAP2C, 60% and 40% of the SRGAP2A are found in the insoluble fraction, respectively (fig. 6A–C). It seems that SRGAP2C and P-SRGAP2C interact with SRGAP2A to pull it out of solubility.

We next investigated the properties of these interactions. Guided by the propencity of the SRGAP2A F-BAR domain to form dimers, the underlying assumption is that native SRGAP2C protomers associate to form hetero-dimers with SRGAP2A protomers (Charrier et al. 2012;Dennis et al. 2012). Indeed, Polleux and coworkers coexpressed SRGAP2A and SRGAP2C in HEK293T cells and demonstrated that the two proteins interact, observed through coprecipitation (Charrier et al. 2012), suggesting direct interactions between the two proteins, but does not necessarily prove hetero-dimerization. Notably, the experiment by Charrier et al. does not contradict our finding that both SRGAP2C and the SRGAP2A:SRGAP2C complex are insoluble, since they used RIPA buffer containing denaturing ionic detergents (SDS and sodium-deoxycholate), capable of solubilizing the SRGAP2A:SRGAP2C complex.

The SRGAP2A:SRGAP2C Hetero-Dimerization Conundrum

Considering SRGAP2C's smaller dimeric contact interface and size (fig. 2D), the mechanism of hetero-dimerization that outcompetes the SRGAP2A homo-dimerization is not straightforward. We sought to investigate whether SRGAP2C interaction with SRGAP2A involves hetero-dimerization, and to further characterize the biochemical properties of SRGAP2C and the SRGAP2A:SRGAP2C complex. Because SRGAP2C and the complex that it forms with SRGAP2A are insoluble and therefore cannot be thoroughly investigated, we truncated the residual Fx of SRGAP2C, leaving a soluble F-BAR portion (residues 1–355, designated F-BARC,fig. 1B). This construct differs from its F-BARA counterpart by the five SRGAP2C arginine substitutions. We then coexpressed F-BARA and F-BARC (separately) with SRGAP2A (a construct spanning residues 1–799, the F-BARx-RhoGAP-SH3 domains,fig. 1B) inE.coli, and purified the SRGAP2A:F-BARA and SRGAP2A:F-BARC complexes using metal-chelate and size exclusion chromatography (supplementary fig. S7,Supplementary Material online). Strikingly, SEC-MALS (SEC in-line with multi-angle light scattering) analysis (fig. 6D–F) showed that while the SRGAP2A:F-BARA complexes tend to dissociate during elution (fig. 6E), F-BARC forms stable hetero-dimers with SRGAP2A (fig. 6F), that even withstand stringent ion exchange MonoQ chromatography conditions (supplementary fig. S7E,Supplementary Material online). Only coexpressed proteins form hetero-dimers, while proteins that were expressed and purified separately, could not exchange protomers when mixed, as evident from SEC analysis (supplementary fig. S7F,Supplementary Material online), indicating that pre-formed SRGAP2A homo-dimers dissociate very weakly, if at all.

These results are consistent with our observation that P-SRGAP2C (which does not have the five arginine substitutions, and is represented by F-BARA) is less effective than SRGAP2C (represented by F-BARC) in insolubilizing coexpressed SRGAP2A (fig. 6A–C).

In conclusion, the five arginine substitutions that were introduced to SRGAP2C in the later phase of its evolution facilitate tighter hetero-dimerization between F-BARC and SRGAP2A. In the case of the full-length SRGAP2C, these substitutions allow more effective insolubilization of coexpressed SRGAP2A.

SRGAP2C Substitutions Are Destabilizing, and Facilitate Tight Hetero-Dimerization

We considered three possible mechanisms by which the arginine substitutions of SRGAP2C may promote hetero-dimerization with SRGAP2A: (1) creation of new, direct interactions with SRGAP2A, (2) elimination of repulsive forces that may exist in an SRGAP2A homo-dimer, and (3) indirect effects.

Analysis of the F-BARx crystal structure does not support the possibility that the substitutions are engaged in new hetero-dimeric contacts with SRGAP2A. However, because we do not have an SRGAP2A:SRGAP2C crystal structure (the complex is insoluble) we cannot rule out this possibility. The second mechanism, that is, eliminating repulsive forces, appears more plausible. As homo-dimerization may involve overcoming same-charge repulsive forces, elimination of positive charges (interestingly, all five substitutions are of arginine residues) from SRGAP2C would promote hetero-dimerization. The third possibility, indirect effects, concerns altered structural properties of SRGAP2C itself. To investigate the structural changes that the arginine substitutions may inflict, we compared the observed solubilities and stabilities of F-BARC and F-BARA after expressing them individually inE.coli (fig. 7A andB). While F-BARA is mostly found in the soluble fraction, F-BARC is mostly (but not entirely) insoluble. Nevertheless, once isolated to homogeneity, F-BARA and F-BARC show similar SEC elution and CD spectra profiles (supplementary fig. S4,Supplementary Material online andfig. 7C, respectively), demonstrating that both have the same general fold and dimeric arrangement.

We next measured the thermal stability of purified F-BARC and F-BARA and found that they have profoundly different denaturation profiles (fig. 7D). The F-BARC denaturation slope is wider and is considerably less thermo-stable than F-BARA, with denaturation temperature values of 32 and 38 °C, respectively (fig. 7E). The wider denaturation slope of F-BARC demonstrate a loss of cooperativity in the F-BARC structure (Kuwajima 1989), in which the F-BARC retains an F-BARA-like fold, but with a looser packing of the three long helices in the tertiary structure.

We next introduced each one of the five arginine substitutions separately into F-BARA, and found that the greatest destabilization effect is inflicted by the R108W mutation (fig. 7F). We next showed that R108K, R108L, and R108S substitutions did not have the same effect, which implies that the destabilizing effect of R108W is due not only to loss of arginine interactions, but is rather inflicted by a tryptophan-specific contribution (supplementary fig. S8,Supplementary Material online).

Taken together, these results reveal a fundamental biochemical property of SRGAP2C: structural instability of the F-BAR fold. This property was acquired at the final stage of SRGAP2C evolution, with the mutagenesis of R108 to tryptophan, and is not present in P-SRGAP2B and P-SRGAP2C.

SRGAP2A:F-BARC Hetero-Dimers Are Also Defective in Membrane Binding and Robo1 Association, but Not in RhoGAP Activity

Up to this point, we showed that SRGAP2C hetero-dimerizes with SRGAP2A, and that once formed, these hetero-dimers do not dissociate and are insoluble, which explains SRGAP2A inactivation by SRGAP2C. Next, we investigated whether additional properties of SRGAP2A (critical for its cell biological functions) are modulated by hetero-dimer formation with SRGAP2C. Since SRGAP2C and SRGAP2A:SRGAP2C hetero-dimers are insoluble, and cannot be isolated for biochemical investigations, SRGAP2A:F-BARC heterodimers were used for measurements of membrane binding and scaffolding, RhoGAP catalysis activity and Robo1-CC3 binding, as these are soluble.

Liposome cosedimentation experiments showed that F-BARC and SRGAP2A:F-BARC hetero-dimers have a reduced affinity for membranes in comparison to F-BARA, F-BARx, and SRGAP2A (fig. 4B andsupplementary fig. S6,Supplementary Material online). In addition, while F-BARA shows some membrane scaffolding activity in COS-7 cells, F-BARC does not (supplementary fig. S9A,Supplementary Material online).

We considered that the SRGAP2C arginine substitutions, in particular those engaged in direct electrostatic contacts with the negatively charged membrane, inflict weaker SRGAP2C membrane binding. Mapping the five substituted arginine residues of SRGAP2C onto the F-BARx crystal structure (supplementary fig. S10,Supplementary Material online) reveals that two of them: R73 and R235 point at the membrane binding surface. The third arginine residue (R250) is positioned on the opposite surface, the fourth (R108) is packed against the molecule's side, and the fifth (R205) is located at the arm tip region, where electron density is not clearly visible. The SRGAP2C substitutions R73H and R235H deprive the membrane binding surface of two of its positive charges, and are therefore likely to reduce the protein's affinity for negatively charged membranes. We introduced SRGAP2C's R73H and R235H substitutions to the F-BARx construct (termed F-BARx-R2H) and compared the mutant's membrane binding to that of F-BARx-WT. F-BARx-R2H showed mildly reduced binding to liposomes (fig. 4B andsupplementary fig. S6,Supplementary Material online) and reduced protrusion formation in COS-7 cells (supplementary fig. S9B,Supplementary Material online).

We next measured and compared the binding affinity of Robo1 for SRGAP2A and the SRGAP2:F-BARC hetero-dimer. Previously, we have found that the SRGAP2A SH3 domain is necessary, but not sufficient for an effective Robo1–SRGAP2A interaction, and that the addition of the RhoGAP and F-BARx domains change the binding kinetics and greatly strengthen Robo1–SRGAP2A association (Guez-Haddad et al. 2015). Specifically, measuredK_D values for Robo1–CC3 association with SH3, RhoGAP–SH3, and F-BARx–RhoGAP–SH3 were 13, 4.5, and 0.6 µM, respectively. We concluded that the F-BARx, RhoGAP, and SH3 domains form a composite surface that binds the Robo1-CC3 better than the isolated SH3 domain. Surface plasmon resonance (SPR) studies of the SRGAP2A:F-BARC hetero-dimer interactions with Robo1-CC3, revealed a kinetic profile significantly different to that of SRGAP2A. The Robo1 interaction was significantly weaker and the obtained kinetic profile merely reached that of a RhoGAP-SH3 fragment, with a calculatedK_D value of 5.6 µM (supplementary fig. S11A,Supplementary Material online). These results indicate that the composite F-BARx–RhoGAP–SH3 Robo1 binding surface in the SRGAP2A:SRGAP2C hetero-dimer is disturbed, thus hampering potential to be recruited to Robo1 signaling sites.

In contrast, the enzymatic GAP activity (Rac1 GTP hydrolysis rate, evaluated using the single-turnover GTPase assay) of an isolated SRGAP2A RhoGAP domain was similar to that of SRGAP2A:F-BARC hetero-dimers (supplementary fig. S11B,Supplementary Material online).

Taken together, these experiments show that arginine substitutions in SRGAP2C reduce its affinity for negatively charged membranes, and that a soluble truncation of SRGAP2C (F-BARC) is incapable of membrane scaffolding in COS7 cells. A soluble version of the SRGAP2A:SRGAP2C hetero-dimer (SRGAP2A:F-BARC) also shows weaker affinity for membranes. Furthermore, this SRGAP2A:F-BARC hetero-dimer has a weaker affinity for Robo1, which reflects a defective F-BARx–RhoGAP–SH3 domain arrangement, disturbing the ligand-binding mechanism, and thereby effectively antagonizing SRGAP2A functions.

A New BAR Domain Fold

Our work reveals a new BAR domain fold: the “inverse F-BAR”, which explains how SRGAP2A interacts with membranes in protruding sub-cellular regions as opposed to invaginated endocytic sites, the typical interaction for canonical F-BAR proteins (fig. 3).

The crystal structure of the extended F-BAR domain (F-BARx) of SRGAP2A also reveals domain-swapped coiled-coil F-BAR extensions (Fx) that pack on the lateral sides of the reciprocal F-BAR protomers. Direct MST binding measurements and fitting of crystal structures into the 3D SAXS envelope of full-length SRGAP4 shows that the RhoGAP-SH3 domains are also likely to participate in dimeric interactions, revealing a large SRGAP2A dimerization interface that includes the F-BAR, Fx, and RhoGAP-SH3 domains (fig. 2).

SRGAP2C Hetero-Dimerization with and Inactivation of SRGAP2A

We show that the antagonism of SRGAP2A by SRGAP2C has two separate, yet equally imperative components. One is the hetero-dimerization of the two proteins, and the second is the inactivation of the hetero-dimerized SRGAP2A.

We demonstrate that unlike the soluble SRGAP2A, SRGAP2C is insoluble and that consequently, SRGAP2A:SRGAP2C hetero-dimers are also insoluble. SRGAP2C’s insolubility is inflicted by a carboxy-truncation in the Fx that most likely results in Fx misfolding. As an insoluble complex, SRGAP2A:SRGAP2C hetero-dimers aggregate, unable to reach designated recruitment sites, an impediment that by itself renders the SRGAP2A:SRGAP2C complex inactive. Nonetheless, by engineering a soluble version of SRGAP2C (F-BARC) we could, to some extent, probe the biochemical properties of the SRGAP2A:F-BARC heterodimers and show that these hetero-dimers are additionally impaired in other aspects besides solubility: they have weaker membrane- and Robo1-binding affinities.

SRGAP2A–SRGAP2C hetero-dimerization poses a thermodynamic puzzle. Evidence for the contribution of the Fx and RhoGAP–SH3 domains to dimerization, raises the question of how does SRGAP2C, that lacks the RhoGAP–SH3 domains and has an impaired Fx, form hetero-dimers while competing with the larger and seemingly favorable SRGAP2A homo-dimers? We show that unlike the stable SRGAP2A:F-BARC hetero-dimers, dimers between SRGAP2A and F-BARA are not stable and dissociate into separate SRGAP2A and F-BARA populations during SEC elution (fig. 6 andsupplementary fig. S7,Supplementary Material online). Evidently, F-BARC and F-BARA behave very differently in respect to association with SRGAP2A. We found that this difference stems from the five arginine substitutions present in SRGAP2C. R108W, which has a dramatic effect on the thermal stability of the F-BAR in SRGAP2C, causes loosening of structural cooperativity and a decrease in the denaturation temperature of the F-BAR (fig. 7). We suggest that the stabilization of SRGAP2C in the SRGAP2A:SRGAP2C hetero-dimers brings a thermodynamic gain that allows SRGAP2C, despite its smaller dimerization surface, to compete with thermodynamically favorable SRGAP2A homo-dimerization.

Were the Primal P-SRGAP2B/P-SRGAP2C Biochemically Different to Modern SRGAP2C, and If So, How?

The five arginine substitutions in modern SRGAP2C (including R108W) were acquired in the final stage of SRGAP2C evolution (fig. 1). Thus, we propose that the primal forms of SRGAP2C, if they were indeed active during the period of their existence, were less effective in SRGAP2A hetero-dimerization and therefore in antagonizing SRGAP2A functions.

Throughout an estimated 1 My (∼3.4–2.4 Ma),SRGAP2B andSRGAP2C existed in their primal forms. The amino acid sequences of these partially duplicatedSRGAP2A copies were identical to each other, free from nonsynonymous substitutions (fig. 1A). Later, each acquired a different set of nonsynonymous mutations, giving rise to the modern forms:SRGAP2B andSRGAP2C (Dennis et al. 2012). An interesting question is whether the primal copies differed from the modern ones. In this work, we show that the biochemical properties of P-SRGAP2B/P-SRGAP2C were different from those of modern SRGAP2C. Although both SRGAP2C and P-SRGAP2B/P-SRGAP2C appear insoluble on isolation, the primal forms are less effective in bringing coexpressed SRGAP2A out of solubility (fig. 6A–C). As a result, more SRGAP2A homo-dimers were able to form and signal effectively, while presence of coexpressed modern SRGAP2C brought about stronger attenuation of SRGAP2A functions.

The final question is: could it be possible that SRGAP2C mutants designed to make stronger heterodimeric contacts with SRGAP2A, would further attenuate activity, and what might be the physiological effects of this.

Design and Cloning of SRGAP Constructs

The human SRGAP2A full-length cDNA clone (KIAA0456) was purchased from ImaGenes GmbH, and its internal BamHI digestion site was mutated by polymerase cycling assembly. SRGAP2C was cloned from a human cDNA library, and P-SRGAP2C was prepared using the SRGAP2A cDNA template for PCR amplification (SRGAP2A 1–452) with reverse primer that adds the carboxy-extension⁴⁵³VRECYGF⁴⁵⁹ (primal sequences derived fromDennis et al. 2012).

Mutants were generated by PCR with appropriate primers introducing the mutations. Multiple mutations were introduced by multiple rounds of mutagenesis using templates already carrying some of the desired mutations. All constructs generated by PCR were checked by sequencing.

ForE.coli expression, SRGAP2 constructs were digested with BamHI and XhoI and ligated into a modified pHis-parallel2 vector (Novagen), with an amino-terminal His-Tag, TRX fusion protein, and TEV cleavage sequence. The RhoGAP-SH3-C’ term (SRGAP2A 485–1071) construct was digested with BamHI and XhoI and ligated into a His-tag deleted pET28 vector (Novagen). The F-BARx–RhoGAP–SH3 (SRGAP2A 1–799) construct was digested with BamHI and NotI and ligated into a modified pRSFDuet-1 vector (Novagen), with the N-terminal His-tag replaced by a SUMO fusion protein.

For Sf9 expression, full length SRGAP2A, SRGAP2C, and P-SRGAP2C cDNA was digested with BamHI and XhoI and ligated into the pFastBac-HTB vector (Invitrogen) with a carboxy-terminal His–Tag. The pFastBac Dual vector (Invitrogen) was used to coexpress SRGAP2A (no tag, ligated at the EcoRI and SalI sites) under the polyhedrin promoter, with either SRGAP2C or P-SRGAP2C (both His-Tagged, ligated at the XhoI and NheI sites) under the p10 promoter.

For COS-7 expression, SRGAP2constructs were GFP-fusions encoded by pEGFP.

Expression and Purification

For bacterial expression, all constructs were expressed inE. coli BL21 Tuner strain (Novagen), also expressing the RIL Codon Plus plasmid (Barak and Opatowsky 2013). Transformed cells were grown for 3–4 h at 37 °C in 2xYT media containing 100 μg/mL ampicillin and 34–50 μg/mL chloramphenicol. Protein expression was induced with 200 μM IPTG over a 16 h period at 16 °C. Cells were harvested and frozen prior to lysis and centrifugation. Forin vitro assays, F-BARx-RhoGAP-SH3, F-BARx, F-BARA, F-BARC, and coexpressed protein constructs were purified by the following steps: supernatant was loaded onto a pre-equilibrated Ni-chelate column (HisTrap, GE Healthcare) with buffer A [50 mM Phosphate buffer, pH 8, 400 mM NaCl, 5% glycerol, 5 mM β-MercaptoEthanol (βME)], washed and eluted with a buffer B gradient (50 mM Phosphate buffer, pH 8, 400 mM NaCl, 5% glycerol, 500 mM imidazole, 5 mM βME). Protein-containing fractions were pooled, incubated with TEV protease (1:50 v/v) and dialyzed O/N at 4 °C against buffer A. The protein was then passed through a Ni-chelate column and further isolated by SEC (HiLoad 26/60 Superdex 200 column, GE Healthcare), pre-equilibrated with buffer C (120 mM NaCl, 50 mM phosphate buffer pH 8, 1 mM DTT). Proteins were concentrated using Vivaspin 20 (Sartorius). When required, SUMO and TRX fusions were removed by TEV and/or SENP proteolysis. For quality control, purified proteins were analyzed by SEC and circular dichroism.

Baculoviruses ware generated using the Bac-to-Bac® system (Invitrogen). Proteins were expressed in baculovirus-infected Sf9 cells grown in 50ml ESF 921 culture medium (Expression systems) using a 2 L roller bottle (Greiner Bio-One). P3 infected Sf9 cells were grown at 27 °C and harvested after 72 h by centrifugation (200 × g, 10 min), then frozen in liquid N₂. Cell pellets were thawed and re-suspended in 10 ml buffer D (50 mM Phosphate buffer, pH 7.5, 200 mM NaCl, 5 mM βME, 5% glycerol (w/v), 0.5 mM EDTA). About 2 mM PMSF was added after lysis using a dounce homogenizer. Cell lysates were clarified by centrifugation at 20,000 × g for 20 min and the supernatant treated as above for Ni-chelate column purification using buffer D for equilibration and buffer E for a gradient elution [50 mM Phosphate buffer, pH 7.5, 200 mM NaCl, 5 mM βME, 5% glycerol (w/v), 0.5M imidazole]. Eluted fractions were analyzed by SDS-PAGE and immunoblotting with anti-His (Santa Cruz) and anti-SRGAP2A (abcam, ab124958) antibodies.

Cell Culture, Transfection, and Staining

HEK293 cells and COS-7 cells were maintained in 10 ml DMEM containing 2 mMl-glutamine, 10% (v/v) fetal bovine serum and gentamycin at 37 °C and 5% CO₂.

HEK293 cells were seeded in six-well plates and transfected with TurboFect according to manufacturer’s instructions (Thermo Scientific) 12 h later. Twenty-four hours after transfection, cells were harvested, resuspended in homogenization buffer [5 mM HEPES pH7.4, 0.32 M sucrose, 1 mM EDTA supplemented with Complete® protease inhibitor (Roche)], and homogenized by passing through a syringe and pottering (12 strokes). The lysates were centrifuged for 10 min at 1000 × g at 4 °C to remove cell debris and nuclei. The supernatant was then spun at 11,700 × g for 20 min at 4 °C to obtain a crude membrane fraction. The membrane fraction was then resuspended in homogenization buffer and again spun to yield a washed membrane fraction as a pellet. Aliquots were analyzed by anti-GFP immunoblotting.

COS-7 cells were seeded in 24-well plates and transfected ∼8 h later using TurboFect (Thermo Scientific). After 18 h, the cells were fixed with 4% PFA for 7 min and processed for immunofluorescence microscopy (Schneider et al. 2014), with Phalloidin AlexaFluor®568 (Molecular Probes).

Light Microscopy and Quantitative Assessment of COS7 Cell Membrane Topology

Confocal images were recorded using a Zeiss AxioObserver.Z1 microscope equipped with an ApoTome and AxioCam MRm CCD camera (Zeiss). Digital images were acquired as a z-series (0.2–0.3 µm intervals) by AxioVision Software (Vs40 4.8.2.0). Transfected cells with perimeters available for quantitative analysis of membrane topology were imaged in systematic sweeps across the coverslips using a Plan-Apochromat 63x/1.4oil objective. Quantitative image processing was done by ImageJ 1.46r software using maximum intensity projections of confocal image stacks at high magnification. The perimeter of each cell was outlined (segmented line tool) and the length determined (ROI manager). The number of protrusions from the cell perimeter (irrespective of their varying individual length) was determined and expressed per µm cell perimeter. In most cases, 15–27 cells were analyzed per condition in 3–4 assays reaching sumn-numbers of cells evaluated ranging from 54 to 87. The analyses of the srGAP2 F-BARdel25 mutant in comparison to wild-type and GFP control include only two assays and 32–34 cells per condition. Data is presented as mean + SEM. Statistical significance was calculated using Prism 5.03 software (GraphPad). Nonquantitative image processing was done by Adobe Photoshop.

Docking of Crystal Structures and Molecular Graphics

Docking of crystal structures into SAXS densities was performed manually using COOT (Emsley et al. 2010) and UCSF Chimera (Yang et al. 2012). Images were produced using Pymol (http://www.pymol.org) and UCSF Chimera.

1. Barak R, Opatowsky Y.. 2013. Expression, derivatization, crystallization and experimental phasing of an extracellular segment of the human Robo1 receptor.Acta Crystallogr Sect F Struct Biol Cryst Commun. 69:771–775. [DOI] [PMC free article] [PubMed] [Google Scholar]
2. Beneken J, Tu JC, Xiao B, Nuriya M, Yuan JP, Worley PF, Leahy DJ.. 2000. Structure of the Homer EVH1 domain-peptide complex reveals a new twist in polyproline recognition. Neuron 26:143–154. [DOI] [PubMed] [Google Scholar]
3. Bianchi S, Stimpson CD, Bauernfeind AL, Schapiro SJ, Baze WB, McArthur MJ, Bronson E, Hopkins WD, Semendeferi K, Jacobs B, et al. 2013. Dendritic morphology of pyramidal neurons in the chimpanzee neocortex: regional specializations and comparison to humans. Cereb Cortex. 23:2429–2436. [DOI] [PMC free article] [PubMed] [Google Scholar]
4. Canagarajah B, Leskow FC, Ho JY, Mischak H, Saidi LF, Kazanietz MG, Hurley JH.. 2004. Structural mechanism for lipid activation of the Rac-specific GAP, beta2-chimaerin. Cell 119:407–418. [DOI] [PubMed] [Google Scholar]
5. Charrier C, Joshi K, Coutinho-Budd J, Kim JE, Lambert N, de Marchena J, Jin WL, Vanderhaeghen P, Ghosh A, Sassa T, Polleux F.. 2012. Inhibition of SRGAP2 function by its human-specific paralogs induces neoteny during spine maturation. Cell 149:923–935. [DOI] [PMC free article] [PubMed] [Google Scholar]
6. Coutinho-Budd J, Ghukasyan V, Zylka MJ, Polleux F.. 2012. The F-BAR domains from srGAP1, srGAP2 and srGAP3 regulate membrane deformation differently. J Cell Sci. 125:3390–3401. [DOI] [PMC free article] [PubMed] [Google Scholar]
7. Dauter Z.2015. Solving coiled-coil protein structures. IUCr J. 2:164–165. [DOI] [PMC free article] [PubMed] [Google Scholar]
8. Dennis MY, Eichler EE.. 2016. Human adaptation and evolution by segmental duplication. Curr Opin Genet Dev. 41:44–52. [DOI] [PMC free article] [PubMed] [Google Scholar]
9. Dennis MY, Nuttle X, Sudmant PH, Antonacci F, Graves TA, Nefedov M, Rosenfeld JA, Sajjadian S, Malig M, Kotkiewicz H, et al. 2012. Evolution of human-specific neural SRGAP2 genes by incomplete segmental duplication. Cell 149:912–922. [DOI] [PMC free article] [PubMed] [Google Scholar]
10. Elston GN, Benavides-Piccione R, DeFelipe J.. 2001. The pyramidal cell in cognition: a comparative study in human and monkey. J Neurosci. 21:RC163.. [DOI] [PMC free article] [PubMed] [Google Scholar]
11. Emsley P, Lohkamp B, Scott WG, Cowtan K.. 2010. Features and development of Coot. Acta Crystallogr D Biol Crystallogr. 66:486–501. [DOI] [PMC free article] [PubMed] [Google Scholar]
12. Endris V, Haussmann L, Buss E, Bacon C, Bartsch D, Rappold G.. 2011. SrGAP3 interacts with lamellipodin at the cell membrane and regulates Rac-dependent cellular protrusions. J Cell Sci. 124:3941–3955. [DOI] [PubMed] [Google Scholar]
13. Fossati M, Pizzarelli R, Schmidt ER, Kupferman JV, Stroebel D, Polleux F, Charrier C.. 2016. SRGAP2 and its human-specific paralog co-regulate the development of excitatory and inhibitory synapses. Neuron 91:356–369. [DOI] [PMC free article] [PubMed] [Google Scholar]
14. Fritz RD, Menshykau D, Martin K, Reimann A, Pontelli V, Pertz O.. 2015. SrGAP2-dependent integration of membrane geometry and slit-Robo-repulsive cues regulates fibroblast contact inhibition of locomotion. Dev Cell. 35:78–92. [DOI] [PubMed] [Google Scholar]
15. Frost A, Perera R, Roux A, Spasov K, Destaing O, Egelman EH, De Camilli P, Unger VM.. 2008. Structural basis of membrane invagination by F-BAR domains. Cell 132:807–817. [DOI] [PMC free article] [PubMed] [Google Scholar]
16. Gilbert SF, Barresi MJF.. 2016. Developmental biology, Chapter 26. 11th ed. Sunderland (MA): Sinauer Associates, Inc.
17. Guerrier S, Coutinho-Budd J, Sassa T, Gresset A, Jordan NV, Chen K, Jin WL, Frost A, Polleux F.. 2009. The F-BAR domain of srGAP2 induces membrane protrusions required for neuronal migration and morphogenesis. Cell 138:990–1004. [DOI] [PMC free article] [PubMed] [Google Scholar]
18. Guez-Haddad J, Sporny M, Sasson Y, Gevorkyan-Airapetov L, Lahav-Mankovski N, Margulies D, Radzimanowski J, Opatowsky Y.. 2015. The neuronal migration factor srGAP2 achieves specificity in ligand binding through a two-component molecular mechanism. Structure 23:1989–2000. [DOI] [PubMed] [Google Scholar]
19. Harmand S, Lewis JE, Feibel CS, Lepre CJ, Prat S, Lenoble A, Boes X, Quinn RL, Brenet M, Arroyo A, et al. 2015. 3.3-million-year-old stone tools from Lomekwi 3, West Turkana, Kenya. Nature 521:310–315. [DOI] [PubMed] [Google Scholar]
20. Kessels MM, Qualmann B.. 2015. Different functional modes of BAR domain proteins in formation and plasticity of mammalian postsynapses. J Cell Sci. 128:3177–3185. [DOI] [PubMed] [Google Scholar]
21. Kuwajima K.1989. The molten globule state as a clue for understanding the folding and cooperativity of globular-protein structure. Proteins 6:87–103. [DOI] [PubMed] [Google Scholar]
22. Mim C, Unger VM.. 2012. Membrane curvature and its generation by BAR proteins. Trends Biochem Sci. 37:526–533. [DOI] [PMC free article] [PubMed] [Google Scholar]
23. Pertz OC, Wang Y, Yang F, Wang W, Gay LJ, Gristenko MA, Clauss TR, Anderson DJ, Liu T, Auberry KJ, et al. 2008. Spatial mapping of the neurite and soma proteomes reveals a functional Cdc42/Rac regulatory network. Proc Natl Acad Sci USA. 105:1931–1936. [DOI] [PMC free article] [PubMed] [Google Scholar]
24. Schneider K, Seemann E, Liebmann L, Ahuja R, Koch D, Westermann M, Hubner CA, Kessels MM, Qualmann B.. 2014. ProSAP1 and membrane nanodomain-associated syndapin I promote postsynapse formation and function. J Cell Biol. 205:197–215. [DOI] [PMC free article] [PubMed] [Google Scholar]
25. Shimada A, Niwa H, Tsujita K, Suetsugu S, Nitta K, Hanawa-Suetsugu K, Akasaka R, Nishino Y, Toyama M, Chen L, et al. 2007. Curved EFC/F-BAR-domain dimers are joined end to end into a filament for membrane invagination in endocytosis. Cell 129:761–772. [DOI] [PubMed] [Google Scholar]
26. Sporny M, Guez-Haddad J, Waterman DG, Isupov MN, Opatowsky Y.. 2016. Molecular symmetry-constrained systematic search approach to structure solution of the coiled-coil SRGAP2 F-BARx domain. Acta Crystallogr D Struct Biol. 72:1241–1253. [DOI] [PubMed] [Google Scholar]
27. Sudmant PH, Kitzman JO, Antonacci F, Alkan C, Malig M, Tsalenko A, Sampas N, Bruhn L, Shendure J, Genomes P, Eichler EE.. 2010. Diversity of human copy number variation and multicopy genes. Science 330:641–646. [DOI] [PMC free article] [PubMed] [Google Scholar]
28. Wong K, Ren XR, Huang YZ, Xie Y, Liu G, Saito H, Tang H, Wen L, Brady-Kalnay SM, Mei L, et al. 2001. Signal transduction in neuronal migration: roles of GTPase activating proteins and the small GTPase Cdc42 in the Slit-Robo pathway. Cell 107:209–221. [DOI] [PubMed] [Google Scholar]
29. Yamazaki D, Itoh T, Miki H, Takenawa T.. 2013. srGAP1 regulates lamellipodial dynamics and cell migratory behavior by modulating Rac1 activity. Mol Biol Cell. 24:3393–3405. [DOI] [PMC free article] [PubMed] [Google Scholar]
30. Yang Z, Lasker K, Schneidman-Duhovny D, Webb B, Huang CC, Pettersen EF, Goddard TD, Meng EC, Sali A, Ferrin TE.. 2012. UCSF Chimera, MODELLER, and IMP: an integrated modeling system. J Struct Biol. 179:269–278. [DOI] [PMC free article] [PubMed] [Google Scholar]
31. Zhang YE, Long M.. 2014. New genes contribute to genetic and phenotypic novelties in human evolution. Curr Opin Genet Dev. 29:90–96. [DOI] [PMC free article] [PubMed] [Google Scholar]

Supplementary Materials