Biochemical and NMR characterization of the interactions of Vav2-SH2 domain with lipids 1 and the EphA2 juxtamembrane region on membrane.

11 Vav2 is a ubiquitous guanine nucleotide exchange factor (GEF) for Rho family GTPases that is 12 involved in regulating a wide range of biological processes. It interacts with several tyrosine- 13 phosphorylated cell surface receptors, including the Eph family receptors, through its SH2 domain. 14 The interaction of Vav2 with EphA2 is crucial for EphA2-mediated tumor angiogenesis. Here we 15 show that Vav2-SH2 domain is a lipid-binding module that can recognize PI(4,5)P2 and PI(3,4,5)P3 16 lipids weakly but specifically. The specific lipid-binding site in Vav2-SH2 domain was identified 17 by NMR chemical shift perturbation experiments using the head groups of PI(4,5)P2 and 18 PI(3,4,5)P3, both of which bind to Vav2-SH2 with millimolar binding affinities. In addition, the 19 interaction between Vav2-SH2 and the phosphorylated juxtamembrane region (JM) of EphA2 20 (Y594 phosphorylated) was investigated using NMR techniques. Furthermore, by using a nickel- 21 lipid containing peptide-based nanodiscs with other types of liposome was observed. These results suggest that Vav2-SH2 has a binding preference for PI(4,5)P2 and PI(3,4,5)P3. EphA2 juxtamembrane region by biochemical and NMR methods Our that Vav2-SH2 can bind to PI(4,5)P2 and PI(3,4,5)P3 lipids weakly but specifically. NMR titrations demonstrated that Vav2-SH2 recognized the head groups of these two lipids with millimolar binding affinities through a lipid binding site separated from its pY-pocket. In addition, NMR titrations also revealed the structural basis of Vav2-SH2 for binding to Y594-phosphorylated EphA2. In order to further explore the effect of membranes on this interaction, we developed a strategy in which the histidine-tagged Y594-phosphorylated EphA2 peptide was attached to peptide-based nanodiscs using nickel chelating lipids. Our ITC analysis demonstrated that membranes increase the binding affinity of Vav2 to EphA2 by ~2 fold. Taken together, this study provides a way to study the protein- protein interaction on complex membrane environment and also a point of view for future work in the significance of membrane environment in membrane-spanning protein-protein interactions.


Introduction 25
The Src-homology 2 (SH2) domains are prototypical protein interaction modules that specifically 26 recognize phosphotyrosine (pY) containing motif [1]. They are found in diverse cell-signaling 27 proteins and play crucial roles in cellular signal transduction. The human genome encodes 121 SH2 28 domains in 111 different proteins [1]. SH2 domains typically contain approximately 100 amino acid 29 residues and share a common fold composed of a central antiparallel β-sheet flanked by two α-30 helices [2]. They specifically recognize pY and a few residues C-terminal to pY using a conserved 31 positively charged pY-binding pocket and a more variable secondary binding site, respectively [3]. 32 Downloaded from https://portlandpress.com/biochemj/article-pdf/doi/10.1042/BCJ20200300/892739/bcj-2020-0300.pdf by guest on 09 September 2020 Biochemical Journal. This is an Accepted Manuscript. You are encouraged to use the Version of Record that, when published, will replace this version. The most up-to-date-version is available at https://doi.org/10.1042/BCJ20200300 of Vav members different from any other Ras superfamily GEF. Through its SH2 domain, Vav2 is 48 able to interact with several tyrosine-phosphorylated transmembrane receptors or coreceptors, 49 including Eph family receptors [10, 12,13], platelet-derived growth factor receptor (PDGFR) [14], 50 epidermal growth factor receptor (EGFR) [15] and CD19 [16]. Through these SH2-mediated 51 interactions, Vav2 is translocated to the plasma membrane and then mediates different extracellular 52 signals to intracellular responses. EphA2 is a member of the Eph family receptors. The interaction 53 of Vav2 with EphA2 is crucial for EphA2-mediated tumor angiogenesis [13]. Two tyrosine 54 phosphorylation sites in the JM region of EphA2, Y588 and Y594, have been shown to be involved 55 in Vav2-SH2 binding. However, until now, the detailed molecular mechanism of this binding has 56 not been determined. In addition to mediate protein-protein interactions, Vav2-SH2 domain has been 57 recently shown to be also capable of binding PM-mimetic vesicles [6]. However, the lipid 58 recognition specificity of Vav2-SH2 domain and the underlying molecular mechanisms are unclear. 59 Moreover, it is also unknown whether the lipid membrane environment has any effect on the 60 interaction of Vav2-SH2 with its membrane-anchored partners. 61 To study membrane-associated events in vitro, a suitable membrane mimetic is needed. 62 Detergent micelles and liposomes are commonly used as membrane mimetics. However, micelles 63 lack a flat lipid bilayer. In addition, the detergents often have deteriorating effects on the structure 64 and activity of a protein. Liposomes contain a lipid bilayer, but they are too large for solution NMR 65 studies. Nanodisc is a soluble nanoscale membrane mimic. It is composed of a planar phospholipid 66 bilayer surrounded by two copies of amphipathic membrane scaffold proteins (MSPs) [17,18]. Due 67 Downloaded from https://portlandpress.com/biochemj/article-pdf/doi/10.1042/BCJ20200300/892739/bcj-2020-0300.pdf by guest on 09 September 2020 Biochemical Journal. This is an Accepted Manuscript. You are encouraged to use the Version of Record that, when published, will replace this version. The most up-to-date-version is available at https://doi.org/10.1042/BCJ20200300 5 until the mixture became clear, followed by several cycles of freeze and thaw between -80 º C and 136 room temperature to make the nanodiscs more homogeneous. The resulting solution containing 137 nanodiscs was then purified by SEC on a Superdex 200 10/300 GL column. 138

Dynamic light scattering (DLS) 139
DLS were performed using a Malvern Zetasizer nano ZS (Malvern, UK) instrument. The 140 temperature was set to 25°C and refractive index was assumed to be equal to that of Tris/NaCl. 141 Samples were analyzed in triplicate. The size and size distributions of nanodiscs were calculated 142 through analyses of autocorrelation functions using Zetasizer software. 143

Nuclear Magnetic Resonance (NMR) experiments 148
All NMR experiments were carried out at 293 K on a Bruker Avance 600 MHz NMR spectrometer 149 equipped with a CryoProbe. The NMR data were processed with NMRPipe [29] and analyzed using 150 mM NaCl, PH 7.4) was loaded into the syringe. The free EphA2-JM-pY594 peptide (12 μM) and 166 the peptide tethered on different nanodiscs, as indicated in Figure 5, were prepared with the same 167 buffer and loaded into the sample cell, respectively. The titration protocol consisted of a single initial 168 Downloaded from https://portlandpress.com/biochemj/article-pdf/doi/10.1042/BCJ20200300/892739/bcj-2020-0300.pdf by guest on 09 September 2020 Biochemical Journal. This is an Accepted Manuscript. You are encouraged to use the Version of Record that, when published, will replace this version. The most up-to-date-version is available at https://doi.org/10.1042/BCJ20200300 injection of 0.4 µL, followed by 19 injections of 2 µL the Vav2-SH2 protein into the sample cell. 169 The heat of dilution was determined by titration of the protein into buffer E under the same condition 170 and subtracted from the corresponding raw experimental data. Thermodynamic data were analyzed 171 with a single-site binding model using MicroCal PEAQ-ITC Analysis Software provided by the 172 manufacturer. 173 protein was recombinantly expressed and purified to electrophoretic homogeneity. In the liposome-181 binding assay ( Figure 1A), the purified Vav2-SH2 was pre-incubated with liposomes. After 182 ultracentrifugation, the supernatant and pellets were analysed by SDS-PAGE. As shown in Figure  183 1B, only in the presence of PI(4,5)P2-or PI(3,4,5)P3-containing liposomes, a small amount of the 184

Results and Discussion
Vav2-SH2 protein could be detected in the pellet fraction, indicating that Vav2-SH2 can bind directly, 185 albeit weakly to these two liposomes. The purified GST, which was used as a negative control did 186 not exhibit any binding to PI(4,5)P2-or PI(3,4,5)P3-containing liposomes. No detectable interaction 187 of Vav2-SH2 with other types of liposome was observed. These results suggest that Vav2-SH2 has 188 a binding preference for PI(4,5)P2 and PI(3,4,5)P3. 189

NMR analysis of the specific binding of Vav2-SH2 to diC4-PI(4,5)P2 and diC4-PI(3,4,5)P3. 190
Recent studies indicate that the molecular location and morphology of lipid-binding sites in SH2 191 domains are highly variable [6, 32, 33]. To investigate the specific lipid recognition mechanism of 192 Vav2-SH2, we took advantage of the previously published Vav2-SH2 solution structure and 193 backbone amide assignments and performed NMR titration experiments using diC4-PI(4,5)P2 and 194 diC4-PI(3,4,5)P3, the water-soluble analogs of PI(4,5)P2 and PI(3,4,5)P3, respectively. As a control, 195 a short-chain phosphatidylserine (diC6-PS) was also included. 2D 1 H-15 N HSQC spectra of Vav2-196 SH2 at a series of protein to lipid ratios were collected, respectively. The titration with diC6-PS did that from the liposome-binding assays. In the case of diC4-PI(4,5)P2 binding, the significantly 203 perturbed residues (with CSPs above mean value plus one standard deviation) are Q682, A702, 204 A704, R706, F707, I720, K721, V722, V723 ( Figure 2B-D). Among them, residues Q682, A702, 205 I720, K721 and V722 displayed fast exchange on the NMR timescale, while residues A704, R706, 206 F707 and V723 also exhibited decreased intensities upon titration, suggesting fast to intermediate 207 chemical exchange (Table S1). Besides, the less perturbed residues Q681 and A708 also revealed 208 intensity reductions. Structural mapping showed that all these residues, except Q681 and Q682, 209 formed a contiguous surface on Vav2-SH2, indicating of the specific binding site for diC4-PI(4,5)P2. 210 Titration of diC4-PI(3,4,5)P3 into Vav2-SH2 induced a similar pattern of chemical shift perturbation 211 and intensities reduction ( Figure S2 and Table S1), suggesting both the lipids bind to the same region 212 of the protein. Besides, the revealed lipid-binding site in Vav2-SH2 was adjacent to its highly 213 cationic pY-binding pocket ( Figure 2D and Figure S3). However, the invariable R698 residue 214 locating in the pY-pocket was less perturbed upon diC4-PI(4,5)P2 or diC4-PI(3,4,5)P3 binding, 215 suggesting that the pY-pocket may not be involved in lipid-binding. Interestingly, the lipid-binding 216 site in Vav2-SH2 is composed of cationic residues as well as hydrophobic and aromatic residues,   corresponding to the JM region of EphA2 (residues 559-607 with Y594 phosphorylated) was 240 synthesized. NMR titration experiment was then carried out. 1 H-15 N HSQC spectra were recorded 241 for 15 N-labeled Vav2-SH2 domain before and after addition of EphA2-JM-pY594 peptide to 242 different molar ratios of protein/peptide. As shown in Figure S4A, the peptide induced large changes 243 in the 1 H-15 N HSQC spectrum of the Vav2-SH2, indicating a direct binding. Assignment of the 244 signals in the bound form was made by performing a series of triple resonance NMR experiments 245 using 15 N, 13 C-labeled Vav2-SH2 in the presence of unlabeled peptide with a molar ratio of 3:1 246 peptide: protein ( Figure S4B). The peptide binding caused a global decrease in signal intensities. 247 The relative resonance peak heights measured from Vav2-SH2 in the presence of peptide (3:1), 248 represented as a percentage of the corresponding resonance peak heights from free Vav2-SH2 are 249 plotted as a function of Vav2-SH2 residue number ( Figure S4C). The average relative peak height 250 is 42.8%. In addition to intensity reductions, CSPs for many residues were observed ( Figure S4A  and βD' (V723). Among them, the pY-binding pocket residues R680, R698 and R700 interact 259 electrostatically with the oxygen atoms of pY, as seen in the complex structures of Vav2-SH2 with 260 its pY ligands [28,37]. Other residues that displayed slow exchange are located in BG loop (S755-261 Q758), EF loop (I731, E733, A734, F737) and βE (I729 and H730). These may be involved in the 262 contact with the residues adjacent to pY594 in the EphA2 peptide. As seen in the SH2-pY complexes, 263 the pY+3 pocket is formed by loops EF and BG that is used to recognize the pY+3 residues in their 264 pY-ligands and to determinate the specificity of an SH2 domain. Overall, it could be clearly seen 265 that the region with significant CSPs contained a central pY-pocket along with a larger binding 266 interface that could accommodate the extended peptide ( Figure 4B). The results indicates that Vav2-267 SH2 binds to the phosphorylated JM region of EphA2 in a canonical SH2-ligand recognition manner. 268 Under physiological condition, the recognition of the phosphorylated EphA2 JM region by 269 Vav2-SH2 takes place on the plasma membrane interface. Therefore, we further studied the 270 influences of membrane environment on this interaction. To better mimic the physiological state of 271 the peptide EphA2-JM-pY594, we adopted a strategy of tethering this histidine-tagged peptide to 272 Downloaded from https://portlandpress.com/biochemj/article-pdf/doi/10.1042/BCJ20200300/892739/bcj-2020-0300.pdf by guest on 09 September 2020 Biochemical Journal. This is an Accepted Manuscript. You are encouraged to use the Version of Record that, when published, will replace this version. The most up-to-date-version is available at https://doi.org/10.1042/BCJ20200300 the membrane mimic containing nickel chelating lipid. Herein, the peptide-based nanodiscs were 273 chosen as the membrane mimetics, which were assembled using a 22-residue peptide derived from 274 apolipoprotein A-I (Apo A-I) and neutral lipid DMPC [22]. The assembled nanodiscs (lipid:peptide 275 ratio of 1:1 w/w for all preparations) were purified by size exclusion chromatography (SEC). The 276 single peak of nanodiscs was eluted at 13.3 ml ( Figure S5A). Transmission electron microscopy 277 (TEM) images showed the formation of disc shaped, monodispersed particles ( Figure S5B). The 278 diameter of peptide nanodiscs was determined as about 8.7 nm by dynamic light scattering (DLS) 279 analysis ( Figure S5C). To prepare the nickel-chelating nanodiscs, which is hereafter referred to as 280 Ni-NTA-nanodiscs, 2% DGS-NTA(Ni) was incorporated into the peptide nanodiscs. The efficient 281 tethering of the histidine-tagged EphA2-JM-pY594 peptide to Ni-NTA-nanodiscs was confirmed 282 by ITC assay. As shown in Figure S6, the peptide forms stable complex with Ni-NTA-nanodiscs, 283 with a binding KD of about 1.4 μM. Again, the interaction between Vav2-SH2 and membrane-284 tethered EphA2-JM-pY594 were characterized by NMR. The 2D 1 H-15 N HSQC spectrum of Vav2-285 SH2 in the presence of membrane-tethered EphA2-JM-pY594 (SH2/peptide 1:3) was recorded and 286 compared with that mixed with the free peptide at the same SH2/peptide molar ratio. As shown in 287 Figure 4C, the membrane-tethered peptide and free peptide induced nearly identical CSPs in the 288 Vav2-SH2 spectrum, indicating that the protein binds to the free and membrane-tethered peptide 289 with the same binding site. The spectrum of Vav2-SH2 binding to membrane-tethered peptide 290 exhibited an overall reduction in signal intensities compared with that binding to the free peptide 291 ( Figure S7A, B). An average relative peak height of 62.9% was observed. The global reduction in 292 signal intensities was likely caused by the large molecular weight of membrane-tethered peptide. 293 Interestingly, significant intensity reduction for several residues was observed ( Figure S7C), 294 including A670, H690, K718, D726, H730, D738, L740, E742, H750, S751, E754 and K764. We 295 guess this may due to the effect of membrane environment. 296

Lipid membrane environment increases the binding affinity of Vav2-SH2 to Y594-297 phosphorylated EphA2 298
To further quantitatively analyze the binding of Vav2-SH2 to the peptide EphA2-JM-pY594 with 299 or without membrane tethering, ITC experiments were then carried out. The results are shown in 300 Figure 5 and Table S1. In all cases, Vav2-SH2 binds to the peptide with a binding stoichiometry of 301 1:1 and the binding is strongly enthalpically driven. The binding KD of Vav2-SH2 with the free 302 peptide was determined to ~3.6 μM ( Figure 5A). Interestingly, the KD of Vav2-SH2 to the 303 membrane-tethered peptide was measured as ~1.8 μM ( Figure 5B), which is about 2-fold lower than 304 that to the free peptide. As a control, we also measured the binding KD of Vav2-SH2 to the free 305 peptide in the presence of Ni-NTA-free nanodiscs (100% DMPC). The presence of Ni-NTA-free 306 nanodiscs did not show any significant effect on the affinity of Vav2-SH2 to the free peptide ( Figure  307 Downloaded from https://portlandpress.com/biochemj/article-pdf/doi/10.1042/BCJ20200300/892739/bcj-2020-0300.pdf by guest on 09 September 2020 Biochemical Journal. This is an Accepted Manuscript. You are encouraged to use the Version of Record that, when published, will replace this version. The most up-to-date-version is available at https://doi.org/10.1042/BCJ20200300 5C). Taken together, these results clearly demonstrated that membrane-tethering enhances the 308 affinity of the Vav2-SH2 domain to the EphA2-JM-pY594 peptide, revealing a critical regulatory 309 role of membrane in this protein-protein interaction. A likely explanation is that the membrane-310 tethered EphA2-JM-pY594 peptide may adopt a conformation that binds stronger to Vav2-SH2 than 311 its water-soluble conformation. SH2 domains are known to bind pY-containing peptides with 312 variable affinity and a significant degree of promiscuity [38]. However, exquisite protein interaction 313 specificity is required for high-fidelity pY signaling in vivo. This membrane enhanced protein-314 protein interaction maybe of great regulatory significance for cell signal transduction. 315 Finally, given that Vav2-SH2 can bind weakly to PI(4,5)P2 or PI(3,4,5)P3, we next probe 316 whether the specific lipid binding of Vav2-SH2 has any effect on its pY-binding using this Ni-NTA-317 nanodiscs system. PI(4,5)P2 or PI(3,4,5)P3 was then introduced into the Ni-NTA-nanodiscs with a 318 mole ratio of 5%. The KD values of Vav2-SH2 binding to the peptide tethered to the PI(4,5)P2 or 319 PI(3,4,5)P3 containing Ni-NTA-nanodiscs were similar to that of peptide tethered to Ni-NTA-320 nanodiscs ( Figure 5D, E), indicating that PI(4,5)P2 or PI(3,4,5)P3 did not affect the binding affinity 321 of Vav2-SH2 to the membrane-tethered pY-peptide. In other words, the SH2-pY recognition is 322 independent of PI(4,5)P2 or PI(3,4,5)P3 lipid binding. This may due to the fact that the lipid-binding 323 pocket of Vav2-SH2 is separated from the pY binding site ( Figure 2D and Figure 5F). Similar

Conclusions 336
In present work, we have characterized the binding of Vav2-SH2 domain with lipids and the Y594 337 phosphorylated EphA2 juxtamembrane region by biochemical and NMR methods. Our data shows 338 that Vav2-SH2 can bind to PI(4,5)P2 and PI(3,4,5)P3 lipids weakly but specifically. NMR titrations 339 demonstrated that Vav2-SH2 recognized the head groups of these two lipids with millimolar binding 340 affinities through a lipid binding site separated from its pY-pocket. In addition, NMR titrations also 341 revealed the structural basis of Vav2-SH2 for binding to Y594-phosphorylated EphA2. In order to 342 further explore the effect of membranes on this interaction, we developed a strategy in which the 343 histidine-tagged Y594-phosphorylated EphA2 peptide was attached to peptide-based nanodiscs 344 using nickel chelating lipids. Our ITC analysis demonstrated that membranes increase the binding 345 affinity of Vav2 to EphA2 by ~2 fold. Taken together, this study provides a way to study the protein-346 protein interaction on complex membrane environment and also a point of view for future work in 347 the significance of membrane environment in membrane-spanning protein-protein interactions.  Vav2-SH2 (20 μg) was mixed with different liposomes (640 μg), as indicated. The GST protein was 470 used as negative control. After centrifugation, the pellet (P) and supernatant (S) were analysed by 471 SDS/PAGE and Coomassie Blue-staining. 472