The Sam domain of the lipid phosphatase Ship2 adopts a common model to interact with Arap3-Sam and EphA2-Sam

Background Sterile alpha motif (Sam) domains are small protein modules that can be involved in homotypic or heterotypic associations and exhibit different functions. Previous studies have demonstrated that the Sam domain of the lipid phosphatase Ship2 can hetero-dimerize with the Sam domain of the PI3K effector protein Arap3. Results Here, we determine the NMR solution structure of Arap3-Sam and implement a multidisciplinary approach consisting of NMR spectroscopy, ITC (Isothermal Titration Calorimetry), mutagenesis and molecular modeling studies to analyze the interaction between Ship2-Sam and Arap3-Sam. This work reveals that Arap3-Sam may associate with Ship2-Sam by adopting a binding mode common to other Sam domains. This binding mode is identical to what we have very recently observed for the association between Ship2-Sam and the Sam domain from the Ephrin A2 receptor. Conclusion Our studies further clarify the structural features that are relevant for Sam-Sam interactions involving Ship2 and give additional hints that could be used for the identification of new molecules able to selectively inhibit Sam-Sam associations.


Background
Arap3 is a protein involved in phospatidylinositol 3 kinase (PI3K) signaling pathways linked to regulation of the actin cytoskeleton, cell spreading and the formation of lamellipodia [1,2]. It works as a GTP-ase activator protein (GAP) for the small G-proteins Arf and Rho [2]. Previous studies have reported that Arap3 binds Ship2 (Src homology 2 domain-containing phosphoinositide-5-phosphatase 2) by forming a hetero-dimer via the sterile alpha motif domains (Sam) of both proteins [3]. The dissociation constant for this complex is about 100 nM as determined by previous Isothermal titration Calorimetry (ITC) studies [3].
Here, we report on structural and binding studies of Arap3-Sam and Ship2-Sam. First, we determined the NMR solution structure of the Sam domain of Arap3 and characterized its interaction site for Ship2-Sam. Furthermore, we also established the binding site of Ship2-Sam for Arap3-Sam. Based on our observations, we speculate that the Ship2-Sam/Arap3-Sam complex may adopt the ML (Mid Loop)/EH (End Helix) interaction model that is common to many Sam/Sam associations [7][8][9]. For example, we recently reported on a similar binding mode involving Ship2-Sam and the Sam domain from the Ephrin A2 receptor (EphA2-Sam) [10], a Sam-Sam association that is important to regulate receptor endocytosis [11]. In agreement with these data, our NMR displacement experiments confirmed that Arap3-Sam and EphA2-Sam compete for a common binding site on the surface of Ship2-Sam.
Our studies further clarify the structural features that are important for Sam-Sam interactions involving Ship2 and give additional hints useful for identifying new molecules able to selectively inhibit Sam-Sam associations.

NMR Solution Structure of Arap3-Sam
The aggregation state of Arap3-Sam in solution was analyzed by means of analytical ultracentrifugation studies and backbone 15 N R1 and R2 nuclear spin relaxation rates measurements.
Arap3-Sam has a rotational correlation time τc estimated by the R2/R1 average value of 7.7 ± 0.7 ns at a protein concentration of 150 μM which increases only slightly to 8.2 ± 0.4 ns at the concentration of 1.4 mM. The Arap3-Sam τc value is indeed similar to those reported for other Sam domains, including Ship2, which only weakly self-associate [10,12,13]. On the contrary, the τc of Ship2-Sam bound to Arap3-Sam is 11 ± 1 ns, this higher value reflects the increase of the molecular weight caused by the Sam-Sam association and points towards the formation of a dimer [10,13,14].
Analytical ultracentrifugation measurements show that Arap3-Sam is a monomer in solution, in fact the experimentally measured molecular weight (10.84 kDa) is in perfect agreement with the expected molecular weight. Moreover, no concentration dependent changes can be noticed in [ 1 H, 15 Table 1. Distance and angle constraints are well satisfied in the ensemble of structures (Table 1), and the conformers con-verge well ( Figure 1, left panel) as demonstrated by the low root-mean-square deviation (rmsd) values evaluated for the residues of the core domain (Table 1).
To identify the binding interface of Arap3-Sam for Ship2-Sam, titration experiments were carried out with 15 N labeled Arap3-Sam and unlabeled Ship2-Sam ( Figure 3A). The residues of Arap3 that were greatly perturbed by the interaction were assessed by analyzing the normalized chemical shift deviations ( Figure 3B). The largest variations were observed at the α 5 helix and the adjacent α1α2 and α 4α 5 loop regions ( Figure 3B, C, D).
The chemical shift mapping results were further supported by isothermal titration calorimetry studies with the Ship2-peptide Ac-EGLVHNGWDDLEFLSDITEEDL-NH2 (Shiptide) that we have previously identified [10]. This peptide encompasses a region of Ship2-Sam (amino acids from 43 to 64) highly affected by chemical shifts variations upon binding to Arap3-Sam ( Figure 2C, left panel). ITC measurements proved that the Shiptide could interact with Arap3-Sam with a dissociation constant K d of 40 ± 7 μM, a single binding site model (binding stoichiometry n = 0.8 ± 0.1), binding enthalpy ΔH = -2602 ± 922 cal/mol and entropy change ΔS = 11 ± 3 cal/(mol K) ( Figure 2D and Additional File 1).
Molecular modelling studies were also performed with the software Haddock 1.3 [18] to generate a model of the Ship2-Sam/Arap3-Sam complex (see Materials and Methods section for details on the docking procedure). The best Haddock solution (i.e., the one with the lowest Haddock score) is shown in Figure 4. The dimer interface is mainly stabilized by electrostatic interactions in between acidic residues of Ship2-Sam and basic residues of Arap3-Sam ( Figure 4). This model has been further confirmed by mutagenesis studies. To this end, an Arap3-Sam triple mutant was designed with the positively charged residues H37, R77 and R80 replaced by negatively charged Asp residues. This mutant fails to bind Ship2-Sam with high affinity as evaluated by chemical shift perturbation studies (See Additional File 2).

NMR displacement experiment
To assess if Arap3-Sam could compete with EphA2-Sam for the same binding site of Ship2-Sam, we performed 2D [ 1 H, 15

Discussion
Sam domains are small protein modules that act in several biological events and can form homotypic or heterotypic associations [20,21]. The lipid phosphatase Ship2 [5,6] contains a Sam domain at its C-terminus. Binding partners of this Sam domain have been only recently reported and consist of the Sam domain from the EphA2 receptor [11] and the Sam domain from the PI3K effector protein Arap3 [3]. The NMR solution structure of Ship2-Sam (pdb code: 2K4P) has been recently determined in our laboratory [10]. We have also investigated the interaction between Ship2-Sam and EphA2-Sam by means of ITC and NMR chemical shift perturbation studies [10]. ITC data Mapping the Ship2-Sam binding interface for Arap3-Sam The solution structure of Arap3-Sam consists of a small five helix bundle and represents a classical Sam-domain fold (Figure 1). In fact, Arap3-Sam presents highest sequence homology with the Sam domain of Arap2 (pdb code: 1X40, Riken Structural Genomics Initiative) as shown by a blastp [22] search of the Arap3-Sam primary sequence versus the Protein Data Bank (pdb) database [23]. Sequence similarities with Ship2-Sam (pdb code: 2K4P [10]) and EphA2-Sam (pdb code: 2E8N, Riken Structural Genomics Initiative) are also relatively high (49% and 58% respectively) (Figure 4, upper panel). Previous studies have already reported that like Ship2-Sam, Arap3-Sam does not have a strong propensity to self-associate and prefers to be involved in heterotypic interactions [3]. Our 15 N R1 and R2 relaxation measurements together with analytical ultracentrifugation studies, further validate these findings. Chemical shift mapping studies indicate that the interaction surface of Ship2-Sam for Arap3-Sam is mainly made up of the central regions of the protein (Figure 2). This is the same area that we identified as responsible for the binding of Ship2-Sam to EphA2-Sam (Figure 4). The Shiptide, a 22 residue long Ship2-Sam peptide, representing the minimal Ship2-Sam region capable of binding to EphA2-Sam, retains some ability to bind Arap3-Sam as shown by ITC ( Figure 2D). In the case of Arap3-Sam/Shiptide interaction, enthalpic contributions are responsible for ~44% of the free energy of binding (ΔH = -2.6 kcal/ mol, ΔG = -5.9 kcal/mol) whereas for EphA2-Sam/Shiptide interaction [10] the enthalpy contributes only ~22% to the free energy of binding (ΔH = -1.4 kcal/mol, ΔG = -6.5 kcal/mol). Thus, in the Arap3-Sam/Shiptide complex hydrogen bonding and electrostatic interactions are more predominant.

Docking Studies
In addition, NMR displacement experiments clearly show that Arap3-Sam and EphA2-Sam compete for the same binding site on the surface of Ship2-Sam (Additional File 3).
The binding area of Arap3-Sam for Ship2-Sam is primarily made up of the C-terminal α 5 helix and adjacent loop regions ( Figure 3). Again, the location of this binding site closely resembles the interaction surface of EphA2-Sam for Ship2-Sam. From these binding data, we conclude that Ship2-Sam and Arap3-Sam most likely interact by using the Mid-Loop (ML)/End-Helix (EH) Model that is common among Sam-Sam associations [8,24,25] and where Ship2-Sam and Arap3-Sam are providing the Mid-Loop and End-Helix interfaces respectively ( Figure 4). The same interaction mode has been previously proposed by us for the interaction between Ship2-Sam and EphA2-Sam.
A model of the Ship2-Sam/Arap3-Sam complex was generated with the software Haddock 1.3 [18] by using chemical shift perturbation data ( Figure 4). The best scoring solution represents well the ML/EH topology that is present in other experimental structure of Sam-Sam complexes [8,25].
The binding site of Ship2-Sam for Arap3-Sam contains many negatively charged residues, while the EH interface of Arap3-Sam includes several positively charged amino acids. As a consequence, this model appears largely stabilized by electrostatic interactions (Figure 4). In fact, by destroying some of these interactions through simultaneous mutation of the positively charged Arap3-Sam residues H37, R77 and R80 to aspartic acids, the binding to Ship2-Sam is abolished or at least highly attenuated as shown by NMR binding data (Additional File 2).
A very similar interaction model has been obtained by us for the Ship2-Sam/EphA2-Sam association, by means of docking procedures [10] (Figure 4). The pattern of interactions at the dimer interface is analogous in the two complexes ( Figure 4). It is worth noting that in the Ship2-Sam/ EphA2-Sam model the EphA2-Sam residue Y81 may form a stacking π-π interaction with the Ship2-Sam residue F55 that can be replaced by the cation-π interaction between the Arap3-Sam residue R80 and F55 of Ship2 in the complex Ship2-Sam/Arap3-Sam. Furthermore, in the Ship2-Sam/Arap3-Sam complex H61 of Arap3-Sam could provide an additional electrostatic interaction with D63 of Ship2-Sam that is not permitted in the Ship2-Sam/EphA2-Sam dimer (Figure 4). These observations may reflect the relatively stronger binding observed between Arap3-Sam and Ship2-Sam (K d = ~0.1 μM [3]) compared to the binding of EphA2-Sam to Ship2-Sam (K d = 0.75 ± 0.12 μM) as well as the better Haddock score for the Arap3-Sam/ Ship2-Sam model (Figure 4, lower panel).

Conclusion
We have described the 3D solution structure of Arap3-Sam and reported on binding studies with Ship2-Sam. Our work leads us to hypothesize that the interaction mode of these two Sam domains is best described by the canonical Mid-Loop/End-Helix model [8,[24][25][26] in which Ship2-Sam and Arap3-Sam are providing the Mid-Loop and End-Helix interfaces respectively. A similar model has been recently suggested by us for the interaction in between Ship2-Sam and EphA2-Sam [10]. Our studies also show that Arap3-Sam competes with EphA2-Sam for binding to Ship2-Sam. Together with binding studies on the Shiptide peptide, our results provide a framework onto which one could envision designing novel molecular probes able to selectively interfere with either the Ship2-Sam/EphA2-Sam or with the Ship2-Sam/Arap3-Sam associations.

Protein expression
Ship2-Sam and EphA2-Sam were expressed as previously reported [10]. A synthetic genes construct containing residues from 1 to 80 of human Arap3 (UniprotKB/TrEMBL code: Q8WWN8), encompassing the Sam domain (residues from 4 to 68), was purchased from Celtek (Nashville, TN). Genes were cloned into the PET15b plasmid and transformed using BL21-Gold (DE3) competent cells (Stratagene).
Unlabeled proteins were expressed at 37°C in LB medium. Protein over-expression was induced at OD 600 = 0.6 for 4 hours with isopropyl β-D-thiogalactopyranoside (IPTG) at 1 mM concentration. Expression of 15 N/ 13 C double labeled and 15 N labeled proteins was carried out in M9 minimal medium containing 2 g/l of 13 C-Glucose and/or 0.5 g/l of 15 NH 4 Cl. 10% fractional 13 C labeling for stereo-specific assignments of Leu-CH 3 δ1,2 /Val-CH 3 γ1,2 methyl groups [27] was obtained by adding 3.6 g of 12 Cglucose (natural abundance) and 0.4 g of 13 C-glucose to the M9 medium.
After dissolving the pellet in the following buffer: 50 mM Tris (pH = 8), 500 mM NaCl, 5 mM imidazole, cell were harvested by sonication. The protein was purified on a nickel column (His-trap TM FF, 5 ml, Amersham) by using an AKTA prime plus FPLC system; the elution buffer consisted of 50 mM Tris (pH = 8), 500 mM NaCl, 200 mM imidazole. To avoid non-specific interactions with the Shiptide during ITC experiments, the His-tag tail of Arap3-Sam was cut away from the protein by incubating it overnight at 4°C with thrombin. The thrombin was then removed with a benzamidine column (FF (HS), 1 ml, Amersham).
The protein concentration was estimated by a nanodrop ND-1000 spectrometer.

Resonance assignments of Arap3-Sam
Experiments for resonance assignments were recorded at 25°C on a Bruker Avance 600 MHz or a 700 MHz Bruker AvanceIII spectrometers both equipped with TCI cryoprobes. NMR samples consisted of 15  ITC experiments were repeated twice to evaluate the reproducibility of the data. Data were fit to a standard one binding site model using Origin as supplied by Microcal.

Analytical ultracentrifugation
A Beckman ProteomeLab™ Optima XL-I analytical ultracentrifuge was used to carry out sedimentation equilibrium analysis. Three runs were performed by using samples with protein concentrations of 1 mg/ml, 0.33 mg/ml, and 0.11 mg/ml, respectively. Data were collected at the angular velocity of 30,000 rpm and at 20°C. Data were analyzed with the software HeteroAnalysis (James L. Cole; http://www.biotech.uconn.edu/auf/).

Docking studies
The program Haddock 1.3 [18] was used to generate a model of the Ship2-Sam/Arap3-Sam complex. The NMR structures number one of both Ship2-Sam (pdb code: 2K4P, [10]) and Arap3-Sam (pdb code: 2KG5) were implemented for these studies. Chemical shift perturbation data were exploited to produce ambiguous interaction restraints (AIR). Residues H47, N48, W50, D51, D52, E54, F55, S57, D58, I59, T60, E61, E62, E66, Q70 of Ship2-Sam were set as active. For Arap3-Sam, residues H37, S70, T72, G73, K76, R77, R80, Q83 were considered active, whereas H61, E62, E63, K65, Q66 were set as passive. Residues for the AIR restraints were chosen among the ones with the greatest chemical shift perturbation because they either show high solvent exposure (> 30% as evaluated with MOLMOL [33]) or because they could provide interactions at the interface as shown in experimental structures of Sam-Sam complexes. The limit for the AIR restraints was kept to the default value of 2 Å. During the rigid body energy minimization stage, 2000 structures were calculated; in the second iteration a semi-flexible simulated annealing of the best 200 solutions was performed, finally a refinement in water was carried out. Segments 48-66 and 70-80 of Ship2-Sam and Arap3-Sam respectively, were set as semi-flexible and movements of their side-chains were permitted during the semi-rigid body docking protocol. Besides, residues of the C-terminal tail of Arap3-Sam (88-100) were set completely flexible during the whole docking calculation.