Crystal structures of p120RasGAP N-terminal SH2 domain in its apo form and in complex with a p190RhoGAP phosphotyrosine peptide

The Rho and Ras pathways play vital roles in cell growth, division and motility. Cross-talk between the pathways amplifies their roles in cell proliferation and motility and its dysregulation is involved in disease pathogenesis. One important interaction for cross-talk occurs between p120RasGAP (RASA1), a GTPase activating protein (GAP) for Ras, and p190RhoGAP (p190RhoGAP-A, ARHGAP35), a GAP for Rho. The binding of these proteins is primarily mediated by two SH2 domains within p120RasGAP engaging phosphorylated tyrosines of p190RhoGAP, of which the best studied is pTyr-1105. To better understand the interaction between p120RasGAP and p190RhoGAP, we determined the 1.75 Å X-ray crystal structure of the N-terminal SH2 domain of p120RasGAP in the unliganded form, and its 1.6 Å co-crystal structure in complex with a synthesized phosphotyrosine peptide, EEENI(p-Tyr)SVPHDST, corresponding to residues 1100–1112 of p190RhoGAP. We find that the N-terminal SH2 domain of p120RhoGAP has the characteristic SH2 fold encompassing a central beta-sheet flanked by two alpha-helices, and that peptide binding stabilizes specific conformations of the βE-βF loop and arginine residues R212 and R231. Site-directed mutagenesis and native gel shifts confirm phosphotyrosine binding through the conserved FLVR motif arginine residue R207, and isothermal titration calorimetry finds a dissociation constant of 0.3 ± 0.1 μM between the phosphopeptide and SH2 domain. These results demonstrate that the major interaction between two important GAP proteins, p120RasGAP and p190RhoGAP, is mediated by a canonical SH2-pTyr interaction.


Introduction
The Ras pathway is involved cell proliferation, differentiation, migration and apoptosis [1], and the Rho GTPases are essential in cell adhesion, protrusion, polarity, migration and cell motility [2]. These pathways interact with one-another, in what is known as 'cross-talk' [3-6], PLOS  but the exact basis for Ras-Rho cross-talk is not fully understood, and represents a deficiency in the understanding of how these pathways function. One of the ways by which Rho and Ras pathways are thought to interact is by the direct binding of two GTPase-activating proteins (GAPs), p190RhoGAP and p120RasGAP (RASA1) [7-9]. p190RhoGAP is a multidomain RhoGAP, and the two isoforms (-A and -B; ARHGAP35 and ARHGAP5, respectively) are significant regulators of Rho signaling [7,10,11]. The proteins contain multiple domains, including N-terminal pseudoGTPase domains [12][13][14], and a C-terminal RhoGAP [15][16][17] (Fig 1A). Between these domains there is a flexible region that has been found to be tyrosine phosphorylated; either one or both of the phosphotyrosine sites at residues Tyr-1087 and Tyr-1105 are responsible for p190RhoGAP's recruitment to the plasma membrane where appropriate regulation of Rho cascades is achieved [9,[18][19][20][21]. Src, Abl2 and other tyrosine kinases have been shown to phosphorylate Tyr-1105 [7-9], and this site is the most frequently observed [22], and is critical for normal p190RhoGAP function, including its direct interaction with p120RasGAP at the plasma membrane [3, 18, 19]. Locations of the pTyr binding site, and the specificity determining +3 site are indicated. C) Overall structure of the Nterminal SH2 domain of p120RasGAP in complex with a peptide corresponding to residues 1100-1112 of p120RasGAP. p120RasGAP is shown in cartoon format. p190RhoGAP is shown in stick format. pTyr-1105 is indicated. 2F obs -F calc electron density for the p190RhoGAP peptide is shown at a contour level of 1σ in blue. Positions of peptide residues are indicated. D) Side view of C, highlighting positions and density of pTyr-1105 and Pro-1108. p120RasGAP is a multidomain regulator of Ras signaling, containing two SH2 domains at its N-terminus and multiple other domains including a RasGAP [23][24][25] (Fig 1A). Binding of p190RhoGAP is mediated by phosphotyrosine-SH2 domain interactions, and a major driver is thought to be between the N-terminal SH2 of p120RasGAP and pTyr-1105 of p190Rho-GAP [18,19]. p190RhoGAP recruitment has multiple effects, including inhibition of Rho signaling by locating p190RhoGAP to the plasma membrane [26][27][28][29][30], and activation of Ras signaling by suppression of p120RasGAP's RasGAP activity when it is bound to p190Rho-GAP [3,31].
p120RasGAP was the first GAP to be identified [24,[32][33][34], is a major downregulator of RAS signaling [35][36][37][38][39], and is extensively mutated in the RASopathy, capillary malformationarteriovenous malformation (CM-AVM) [40][41][42][43], however, the structural basis of its signaling is not well understood. Likewise, the interaction of p120RasGAP and p190RhoGAP was first identified in 1995 [7], but further molecular level studies have not been pursued. Therefore, in this study we determine the crystal structures of the N-terminal SH2 domain of p120RasGAP alone, and its complex with a phosphorylated peptide corresponding to residues around and including pTyr-1105 of p190RhoGAP. These crystal structures demonstrate a canonical SH2-pTyr interaction between the proteins and provides the first structural study defining the interaction of p190RhoGAP and p120RasGAP.

Expression and purification of WT and R207A mutant p120RhoGAP N-SH2 domain
cDNA of residues 174 through 280 of human p120RasGAP (UniProt ID: P20936) encoding the N-terminal SH2 domain tagged was subcloned into a modified pET28a vector with an Nterminal hexahistidine tag and TEV protease recognition sequence. Two native cysteines, C236 and C261, were mutated to serine to prevent disulfide bridge formation. Following transformation into Rosetta(DE3) cells, a 1L culture was induced overnight at OD 600 of 0.6-0.8 and 18˚C using 0.2 mM IPTG. Cells were harvested by centrifugation and resuspended in lysis buffer (50 mM HEPES pH 7.3, 500 mM NaCl). Lysis was facilitated by lysozyme addition and three freeze/thaw cycles, prior to sonication and DNAse I addition. Clarified lysate was prepared by centrifugation at 5000 x g at 4˚C and incubated with nickel beads (Ni-NTA Agarose, Qiagen) for 1 hr at 4˚C to capture hexahistidine-tagged protein. Following washing with 3 column volumes of wash buffer (50 mM HEPES pH 7.3, 500 mM NaCl, 20mM imidazole), TEV protease was added overnight at 4˚C to cleave the SH2 domain protein from the hexahistidine-tag. The next day, flow through containing tag-free SH2 domain protein was collected and the beads were washed with several column volumes of wash buffer to fully remove the cleaved protein from the beads. Size exclusion chromatography (Superdex 75, GE Healthcare) was performed on p120RasGAP N-SH2 in 20 mM Tris pH 7.4 and 150 mM NaCl. Yield was 20 mg/L culture. Protein was concentrated by centrifugal filter (Amicon Ultra, Millipore Sigma).
A mutant form, with R207A mutation, was introduced by QuikChange mutagenesis (Agilent) using forward and reverse primers (5'-tccgatcactctctgctataagataactgcca gacttccct-3', 5'-agggaagtctggcagttatcttatagcagagagtgatcgga-3'). Expression and purification were by identical methods as wild type protein. Yield for the mutant was 4 mg/L culture. Both wild type and mutant constructs elute from size exclusion chromatography as monodisperse peaks.

Peptide synthesis
A synthetic 13 amino acid peptide of sequence EEENI(p-Tyr)SVPHDST native to p190RhoGAP residues 1100 to 1112 phosphorylated at Tyr-1105, with N-terminal acetylation and C-terminal amidation, was commercially synthesized (GenScript) and re-suspended in sterile-filtered water.

Crystallization and data collection
The purified N-terminal SH2 domain of p120RasGAP was concentrated to 17.5 mg/mL and initial crystal screening conducted using Index HT and PEG Rx HT (Hampton Research). Hits were observed in precipitant conditions of 0.2 M ammonium acetate, 0.1 M Tris pH 8.0 and 16% w/v polyethylene glycol 10,000 and grid screening conducted using hanging drop vapor diffusion VDX plates with 1 μL:1 μL protein: reservoir solution ratio suspended over 500 μL reservoir solution. Optimized crystals grew against 0.2 M ammonium acetate, 0.1 M Tris pH 8.0 and 21% w/v PEG 10,000 at room temperature. For data collection, crystals were cryoprotected in precipitant solution containing 34% w/v glycerol and flash-frozen in liquid nitrogen.
For co-crystallization of wild-type protein with phosphopeptide, native PAGE was first conducted to determine saturating phosphopeptide concentration. The protein-peptide mixtures used throughout the screening and optimization process had a saturating 1.7:1 peptide:protein molar ratio. Initial screening was conducted using Index HT, PEG Rx HT and PEG Ion HT screens (Hampton Research). Initial crystal hits were observed in conditions containing 1.8 M ammonium sulfate, 0.1 M Bis-Tris pH 6.5, 2% PEG MME 550 yielding 2-dimensional needle cluster crystals. Following optimization, single crystals grew against 1.8 M sodium malonate, 0.1 M Bis-Tris pH 6.5, 2% PEG MME 550 at room temperature using hanging drop vapor diffusion VDX plates with a 1 μL:1 μL ratio of a pre-mixed protein-peptide solution to reservoir solution suspended over 500 μL reservoir solution. Crystals were cryo-protected in reservoir buffer containing 2.9 M sodium malonate and flash-frozen in liquid nitrogen.

Structure determination and refinement
X-ray diffraction data for both crystals were collected at the Northeastern Collaborative Access Team (NE-CAT) beamline 24-ID-C at Argonne National Laboratory and reduced using HKL2000 [44]. For the apo structure, data were processed to 1.75 Å resolution. Initial scaling and data quality assessment in Phenix Xtriage [45] supported a spacegroup of P3 2 12 and detected the presence of translational pseudosymmetry with an off-origin Patterson function peak at height of 71.5%. Molecular replacement in P3 2 12 by Phaser [46] using the C-terminal SH2 domain of phospholipase C-γ (PDB ID: 4K44) as a search model yielded a solution containing three copies with a translation function Z-score (TFZ) of 9.3; however, Phenix autobuild [47] and refinement stalled at R free values around 50%. Thus, Zanuda [48] was used to aid in correct spacegroup assignment of P3 2 with 6 copies of SH2 per asymmetric unit. Reflection data were then reprocessed in P3 2 in HKL2000 [44]. Molecular replacement was performed in Phaser [46] using the C-terminal SH2 domain of phospholipase C-γ (PDB ID: 4K44) as a search model, and six copies were placed with a TFZ score of 12.2. Phenix autobuild [47] built 580 residues. For the peptide-bound structure, Matthews analysis predicted one copy in the asymmetric unit. Molecular replacement by Phaser [46] using the C-terminal SH2 domain of phospholipase C-γ (PDB ID: 4K44) as a search model yielded a single solution with a translation function Z-score (TFZ) of 11.3. Phenix autobuild [47] built 92 residues of the p120RhoGAP N-SH2 domain and 8 residues of the p190RhoGAP peptide. Residues of the peptide were updated to the correct sequence, and manual building conducted using Coot [49]. For both structures, multiple rounds of manual model building using Coot [49] and refinement using Phenix [45] were conducted, with NCS applied in the apo structure refinement. The density for the p190RhoGAP peptide is clear, and MolProbity [50] found good geometry for the final models. The two introduced cysteine to serine mutations, C236S and C261S, lie within linker loops and do not impact secondary structure. The asymmetric unit of the apo structure is composed of three dimers which we attribute to crystal packing: chains A/B, C/D, and E/F. Final R and R free values for the apo structure are 18.5% and 21.2%, respectively. For the peptide-bound structure the final R and R free values are 22.0% and 26.3%, respectively (Table 1). Structural figures were generated using CCP4mg [51]. The PISA server was used for analysis of the protein-peptide interaction interface [52]. All crystallography software was compiled by SBGrid [53].

Native polyacrylamide gel electrophoresis (PAGE)
A serial dilution series of pTyr-1105 phosphopeptide was performed in dilution buffer (50 mM Tris pH 7.4), and mixed with wild-type or R207A mutant N-terminal SH2 domain protein for a final protein concentration of 0.08 mM and a phosphopeptide concentration ranging from 0.02 mM to 0.15 mM (peptide:protein ratio ranging from 0.25 to 1.9). The protein-peptide mixtures, or protein alone as negative control, were incubated on ice for 10 minutes and then centrifuged for 1 minute at 4˚C. 10 μL of 2X native protein sample buffer (62.5 mM Tris HCL pH 6.8, 40% glycerol, 0.01% bromophenol blue) was then added to all samples. 15

Isothermal titration calorimetry (ITC)
The N-SH2 protein and pTyr1105 peptide were prepared for ITC by overnight dialysis in a common buffer with a 20mM Tris 7.4, 150mM NaCl composition. Slide-A-Lyzer Dialysis Cassettes with a molecular weight cut-off of 3,500 Da were used for protein dialysis and Micro Float-A-Lyzer Dialysis devices with a 100-500 Da cut-off for peptide dialysis. Samples were retrieved from their cartridges, spun down for 10 minutes at 4˚C. Concentrations were measured using a Nanodrop spectrophotometer at 280 nm. To determine peptide concentration, a phosphotyrosine extinction coefficient of 458.6 M -1 cm -1 at pH 7.4 was used [54]. Two ITC experiments were performed using a Nano-ITC (TA Instruments). Both protein and peptide were degassed for 3 minutes prior to loading. 350 μL of protein were injected into the sample cell. The injection burette was loaded with 52 μL of peptide. 20 injections of 2.5 μL of peptide, each spanning 300 seconds, were performed per ITC run. Results were analyzed using the nano-ITC software, through fitting to an independent binding model. Observed values for two runs were: K a = 2.3 x10 -5 M -1 , N = 0.724, ΔH = -17.6 kcal/mol, K d = 0.43 μM, TΔS = -8.9 kcal/ mol, ΔG = -8.7 kcal/mol; and K a = 4.2 x10 -6 M -1 , N = 0.63, ΔH = -13.9 kcal/mol, K d = 0.24 μM, TΔS = -4.9 kcal/mol, ΔG = -9.0 kcal/mol.

Crystal structures of apo and peptide-bound p120RasGAP N-terminal SH2 domain
Direct interaction of the N-terminal SH2 domain of p120RasGAP and pTyr-1105 of p190Rho-GAP in large part mediates the direct interaction of these proteins [18,19]. The molecular p120RasGAP N-SH2 in complex with p190RhoGAP pTyr-1105 basis for the interaction has not, however, been described. We therefore determined the crystal structure of this SH2 domain alone and in complex with a peptide corresponding to the pTyr-1105 region. We began by expressing and purifying the N-terminal SH2 domain of p120Ras-GAP alone, and by obtaining apo crystals of this domain. These crystals diffracted well, and we p120RasGAP N-SH2 in complex with p190RhoGAP pTyr-1105 obtained a 1.75 Å dataset that contains 6 copies per asymmetric unit. Of the 6 copies, chains A and B show significantly lower B-factors and improved electron density compared to the other copies and we have based our analyses on these chains. We next obtained co-crystals of the N-terminal SH2 domain of p120RasGAP in complex with a synthesized peptide, EEENI (p-Tyr)SVPHDST, corresponding to residues 1100 to 1112 of p190RhoGAP. The co-crystals diffracted to 1.6 Å resolution and contain one dimer per asymmetric unit. Good electron density is observed for all residues of the peptide except Thr-1112. The crystal structures of the N-terminal p120RasGAP SH2 domain and the p120RasGAP-p190RhoGAP complex are the first for the N-terminal SH2 domain, and the first showing interaction of p190RhoGAP and p120RasGAP.
In canonical SH2 domains, peptide binding occurs perpendicular to the β-sheet, with the phosphotyrosine (position 0) binding site located between the β-sheet and helix αA. The specificity determining +3 position (in Src this is the isoleucine of the preferred -pY-E-E-I-binding site) is located on the opposite side of the β-sheet, flanked by helix αB, and by the βE-βF loop and the βG loop immediately C-terminal to helix αB (Fig 1B). The interaction between p120RasGAP SH2 domain and the p190RhoGAP pTyr-1105 peptide follows this convention and is supported by good electron density (Fig 1C and 1D).

Conformational sampling of the apo SH2 domain of p120RasGAP
In our crystal structure of apo p120RasGAP, we observe six copies per asymmetric unit arranged as dimers (chains A/B, C/D, E/F). Superposition of the crystallographic dimers reveals that there are two conformational classes, which can be defined by the conformation of the βG and βE-βF loops. Interestingly, these loops important for specificity determination at the +3 position of the phosphopeptide [60][61][62]. We observe that in one class the loops are open and poised for peptide binding, and in the other they are closer and seem to occlude peptide binding (Fig 2A). In both classes, the B-factors of βG loop are high, and in the occluded form this loop cannot be built in two of the three copies, however, when the bound phosphopeptide structure is superposed on these conformations the occluded state clashes with His-1109 at peptide position +4, but the poised conformation can accommodate binding (Fig 2B). We interpret this to indicate conformational flexibility and sampling of the specificity defining loops in the absence of phosphopeptide binding.

Structural basis of p120RasGAP N-SH2 interaction with p190RhoGAP pTyr-1105
The co-crystal structure reveals a broadly canonical interaction between p120RasGAP N-terminal SH2 domain and the p190RhoGAP pTyr-1105 peptide. The interaction buries 1290 Å 2 of total surface area and the peptide binds perpendicular to the central SH2 domain β-sheet, with pTyr-1105 at position 0 and Pro-1108 at position +3 inserted into the expected pockets ( Fig 1C). pTyr-1105 is coordinated by a salt-bridge to Arg-207, which is the conserved arginine of the FLVR motif [63], also commonly referred to as ArgβB5 in SH2 convention [56]. It is also coordinated by Arg-188 (ArgαA2) and Ser-209 (SerβB7) (Fig 3A). Interestingly, a threeresidue cation-π stack is also observed between the phenyl-ring of pTyr-1105, Arg-231 and p120RasGAP N-SH2 in complex with p190RhoGAP pTyr-1105 Arg-212 ( Fig 3B). In the apo structure, we do not observe a similar orientation of arginines, Arg-231 and Arg-212, we therefore infer the cation-π stack to be induced by peptide binding. Pro-1108 of p190RhoGAP is inserted into the specificity determining SH2 hydrophobic pocket between βG and βE-βF loops. The pocket is defined by residues Phe-230, Leu-262, Ile-241 and Tyr-256 ( Fig 3C). Compared to Src, this is a shallow +3 specificity site and is thus predisposed for specificity towards proline, as has been shown experimentally [61]. In the co-crystal structure, the βE-βF loop is stabilized by peptide binding which allows a hydrogen-bond to form between the carbonyl oxygen of Leu-262 and the backbone nitrogen of His-1109. The βE-βF loop displays lower relative B-factors compared to the apo structure ( Fig 2B) and we interpret this to indicate stabilization of the βE-βF loop upon peptide binding. Additional direct hydrogen bonds are also observed between the backbone carbonyl of Glu-1101 and Arg-211, between the backbone nitrogen of pTyr-1105 and the backbone carbonyl of His-229, and between Asp-1110 and both Tyr-256 and Ser-260 of p120RasGAP.
Overall the co-crystal structure indicates a conventional SH2-pTyr interaction, and the surface electrostatics of the SH2 domain support this observation, with a positively charged region encompassing the phosphotyrosine binding site, and a hydrophobic region at the +3 pocket ( Fig 3D).

Biochemical validation of the mode of p120RasGAP-p190RhoGAP interaction
To validate the interaction that we observe crystallographically, we conducted native PAGE. The N-terminal SH2 domain of p120RasGAP slowly enters a native gel, however, addition of saturating concentrations of p190RhoGAP pTyr-1105 peptide result in a significant increase in its mobility in the gel (Fig 4 and S1 Fig). We hypothesize this to be due to an overall change in the surface charge of the complex compared to the apo SH2 domain [64]. The key conserved residue in almost all SH2 domains that is responsible for phosphotyrosine binding is the FLVR motif arginine, and its mutation to alanine is often used to generate a non-functional SH2  [56]. We introduced an R207A mutation into the N-terminal SH2 domain of p120RasGAP and find that addition of p190RhoGAP pTyr-1105 peptide to the mutant protein fails to shift (Fig 4). We interpret this to validate the importance of Arg-207 for the p120RasGAP N-SH2 interaction with p190RhoGAP pTyr-1105.

Affinity measurements of the p120RasGAP-p190RhoGAP interaction
We conducted isothermal titration calorimetry to assess the interaction between the N-terminal SH2 domain of p120RasGAP and the p190RhoGAP pTyr-1105 peptide. We injected the p190RhoGAP pTyr-1105 peptide at a concentration of 78.5 μM into N-SH2 at a concentration of 15 μM. Averaged over two independent experiments we observe a K d of 0.3 ± 0.1 μM ( Fig  5A). We do not observe measurable heat when we inject pTyr-1105 peptide at a concentration of 120 μM into R207A mutated N-SH2 at a concentration of 15 μM (Fig 5B).

Conclusions
In this study, we determine the crystal structure of a key interaction between p190RhoGAP and p120RasGAP. For over two decades, extensive studies have been conducted on the interaction and consequences of binding [65], however, the molecular basis for the interaction has not, until now, been observed. The SH2 domains of p120RasGAP bind to pTyr-1105 of p190RhoGAP [3,18,19], and this is thought to be the most critical component of the interaction between the proteins-although frequently observed [22] there remains controversy about the importance of the nearby phosphorylation site in p190RhoGAP, pTyr-1087 [18]. Our structural study demonstrates that the interaction of the N-terminal SH2 domain with pTyr-1105 of p190RhoGAP is a canonical FLVR motif arginine mediated SH2-phosphotyrosine complex. Our affinity measurement for this interaction of 0.3±0.1 μM is similar to other SH2-pTyr interactions [61], and mutation of the FLVR motif arginine residue abolishes interaction. Based on Dali analysis, the structure is most similar to the C-terminal SH2 domain from phospholipase C-γ1 (Z-score 17.3, r.m.s.d. 1.4 Å over 97 residues) [66]. The N-terminal SH2 domain of p120RasGAP has a preference for proline at the pY+3 position [67], and the PLC-γ1 SH2-C prefers proline, valine and isoleucine [67]. Recently, however PLC-γ1 SH2-C was shown to be promiscuous in its binding specificity at this position [68] perhaps implying a similar diversity for other SH2 domains, including those of p120RasGAP. Overall, our crystal structures therefore demonstrate the mode of binding of a key interface between p190RhoGAP and p120RasGAP, and provide a molecular understanding of specificity determinants of the interaction.