Experimental Characterization of the Interaction between the N-Terminal SH3 Domain of Crkl and C3G

Crkl is a protein involved in the onset of several cancer pathologies that exerts its function only through its protein–protein interaction domains, a SH2 domain and two SH3 domains. SH3 domains are small protein interaction modules that mediate the binding and recognition of proline-rich sequences. One of the main physiological interactors of Crkl is C3G (also known as RAPGEF1), an interaction with key implications in regulating cellular growth and differentiation, cell morphogenesis and adhesion processes. Thus, understanding the interaction between Crkl and C3G is fundamental to gaining information about the molecular determinants of the several cancer pathologies in which these proteins are involved. In this paper, through a combination of fast kinetics at different experimental conditions and site-directed mutagenesis, we characterize the binding reaction between the N-SH3 domain of Crkl and a peptide mimicking a specific portion of C3G. Our results show a clear effect of pH on the stability of the complex, due to the protonation of negatively charged residues in the binding pocket of N-SH3. Our results are discussed under the light of previous work on SH3 domains.


Introduction
Crkl is a ubiquitously expressed 39 kDa adapter protein, member of the protooncogene CRK family, that mediates and regulates the linking of several signaling proteins. It was originally discovered in cells from patients with chronic myelogenous leukemia, and its overexpression is correlated with the onset of a number of cancer diseases (recently reviewed in [1]). Crkl plays an essential role in regulating several physiological pathways linked to cytoskeletal changes and cell migration and possesses a prominent role in the onset of human cancers, such as, for example, chronic myelogenous leukemia [2]. Intriguingly, Crkl possesses no catalytic or transcriptional activity and exerts its functions through its protein-protein interaction modules that compose the entire protein, i.e., one N-terminal SH2 domain followed by two SH3 domains (namely N-SH3 and C-SH3).
SH3 domains are small protein interaction modules composed of a five strands βsandwich and a 3 10 helix, that typically mediate the binding and recognition of proline-rich sequences, in particular those characterized by the P-X-X-P consensus (X being any amino acid) [3,4], although atypical binding sequences interacting with SH3 have been identified [5,6]. Importantly, the P-X-X-P motif can be arranged into two opposite orientations, defined by the formation of a salt bridge between a positively and a negatively charged residue(s) on the SH3 binding surface [7]. Crk and Crkl display an overlapping list of cellular interactors (such as for example SOS, EPS15 and C3G) [8], which may be explained by their ability to recognize and bind similar consensus sequences P-X-L-P-X-K (Proline-X-Leucine-Proline-X-Lysine) [9,10].
C3G (also known as RAPGEF1, Rap guanine nucleotide exchange factor 1) is the first guanine nucleotide exchange factor discovered to interact with the SH3 domain of

The Recognition Event of C3G by the N-SH3 Domain Is Electrostatically Driven
In order to characterize the binding reaction between the N-SH3 domain of Crkl and C3G, we conducted kinetic binding experiments with a stopped-flow apparatus, by rapidly mixing the N-SH3 domain versus the C3G 277-296 peptide, the latter covalently linked with a dansyl group at its N-terminus (Dansyl-VVDNSPPPALPPKKRQSAPS). This modification allowed us to monitor the binding reaction by following the change of Förster resonance energy transfer (FRET) signal upon binding, which generates from the two naturally present tryptophan residues in the N-SH3 domain in position 164 and 165 (donor) and the dansyl group linked to the peptide (acceptor).
To characterize the binding between N-SH3 and the C3G 277-296 peptide, we resorted to carry out pre-steady state rapid mixing stopped-flow experiments. In these types of experiments, a common practice lies in performing the kinetic analysis under the so-called pseudo-first-order conditions, i.e., a condition in which one of the two reactants is held at a much higher concentration than the other. In practice, in many cases, it is extremely difficult to achieve such conditions [19]. This is particularly true in cases in which the observed rate constants are so high to approach the experimental limitations of the stopped-flow apparatus. Accordingly, as described below, the analysis of the binding data must be performed by applying the analytical solution of the bimolecular binding transition [19,20].
We rapidly mixed a constant concentration of C3G 277-296 peptide (0.5 µM) versus increasing concentrations of N-SH3 (ranging from 0.5 to 5 µM) and the observed rate constants (k obs ) obtained at different ionic strength conditions (buffer Hepes 50 mM, pH 7.0, 0.15 M, 0.3 M, 0.5 M and 1 M NaCl) were plotted as function of concentration of N-SH3 and fitted with a linear equation (Figure 1). By following previously derived equations [19,20], the dependence of k obs as function of [N-SH3] was fitted with Equation (1), taking into account the non-pseudo-first order conditions [19], allowing us to calculate the microscopic association rate constant (k on ), the microscopic dissociation rate constant (k off ) of the binding reaction and the equilibrium dissociation rate constant K D , such as k off /k on . To increase the reliability of the calculated K D , we resorted to directly obtaining a k off value through displacement experiments (Table 1), in which a preincubated complex of N-SH3 domain and dansylated C3G 277-296 peptide (at the concentration of 0.5 and 2 µM, respectively) were rapidly mixed versus a high excess of nondansylated peptide (ranging from 20 to 40 µM). In agreement with the theory in [21], the observed rate constants were insensitive to displacer concentration. In all binding and displacement experiments conducted, traces were satisfactorily fitted with a single-exponential equation.
into account the non-pseudo-first order conditions [19], allowing us to calculate the microscopic association rate constant (kon), the microscopic dissociation rate constant (koff) of the binding reaction and the equilibrium dissociation rate constant KD, such as koff/kon. To increase the reliability of the calculated KD, we resorted to directly obtaining a koff value through displacement experiments (Table 1), in which a preincubated complex of N-SH3 domain and dansylated C3G277-296 peptide (at the concentration of 0.5 and 2 µM, respectively) were rapidly mixed versus a high excess of nondansylated peptide (ranging from 20 to 40 µ M). In agreement with the theory in [21], the observed rate constants were insensitive to displacer concentration. In all binding and displacement experiments conducted, traces were satisfactorily fitted with a single-exponential equation.
It should be noticed that the observed rate constants reported in Figure 1 are at the limit of the experimental detection by stopped-flow experiments. Consequently, all the binding experiments reported in this work were measured at 10 °C, in order to slow down the apparent transitions. Hence, it is important to note that the experimental conditions significantly deviate from the physiological conditions, and their interpretation should be mainly taken with comparative purposes. Changing the ionic strength of the solution is the simplest way to modulate an electrostatically driven binding reaction. It is generally known that shielding the electrostatic It should be noticed that the observed rate constants reported in Figure 1 are at the limit of the experimental detection by stopped-flow experiments. Consequently, all the binding experiments reported in this work were measured at 10 • C, in order to slow down the apparent transitions. Hence, it is important to note that the experimental conditions significantly deviate from the physiological conditions, and their interpretation should be mainly taken with comparative purposes.
Changing the ionic strength of the solution is the simplest way to modulate an electrostatically driven binding reaction. It is generally known that shielding the electrostatic attraction of diffusion controlled binding reactions leads to the opposite effects on the microscopic association and dissociation rate constants, with k on rapidly decreasing upon increasing ionic strength [22]. In in vitro experiments, the interaction between two proteins is the result of a random collision forming an encounter complex, stabilized in a final com-plex after one (or more) transition state(s). In such scenarios, the early events of recognition between charged residues of the two proteins are diffusion controlled and affected by ions dissolved in solution. Then, the binding follows, with desolvation of polar and charged residues at the interface between the two proteins, the bound complex remaining mostly unaffected by increasing concentration of salt in solution. Accordingly, in the case of N-SH3 domain binding with C3G 277-296 , the inspection of Figure 1 and the analysis of kinetic data at different ionic strength conditions (reported in Table 1) display a pronounced effect of salt concentration on the k on .

Protonation of Negatively Charged Residues Abolishes Binding
To further investigate the role of charges in the binding reaction between N-SH3 domain and C3G, we resorted to performing kinetic binding experiment at different pH conditions, at a range of pH between 5.0 and 9.0. The observed rate constants obtained at different pH conditions were fitted with Equation (1) (Figure 2, left), and the calculated kinetic parameters k on and k off (obtained from separated displacement experiments) are reported in Table 1.
The dependence of k on and k off as function of pH, reported in Figure 2, right, clearly shows that the affinity between the two interacting molecules decreases with decreasing pH. Unfortunately, the very high value of k off prevents any reliable analysis using a Henderson-Hasselbalch equation. In fact, in the case of the N-SH3, we could not obtain a sigmoidal profile, due to the impossibility to measure binding at pH < 5.0. Although the beginning of a transition is clearly visible, an accurate fit would require more data points for pH < 5.0. On the basis of the available data, it may be concluded that acidic pH conditions cause a dramatic increase in the k off , with the k on being less affected, determining a pronounced destabilization of the complex.
The analysis of these data provides us with important information about the binding mechanism of the N-SH3 domain with C3G. In light of what was previously shown for the ionic strength dependence, kinetic data obtained at different pH highlight a double role for salt bridges formation in both the early recognition events and the stabilization of the complex. Moreover, since SH3 domains have evolved to recognize and bind polyproline sequences, and proline binding occurs mainly through C-H·π interactions with aromatic residues [23,24], the formation of canonical salt bridges may give a substantial contribution in optimizing the recognition and binding of the substrate, improving the specificity in the crowded intracellular environment. Interestingly, the comparison of the primary structures of Crk and Crkl highlights the conservation of three negatively charged residues (D147, E149 and D150 on Crk, D138, E140 and D141 on Crkl) physically located at the binding interface of the proteins (Figure 3). Based on this evidence and on previous structural work on Crk [18], all together our results suggest that D138, E140 and D141 residues of the N-SH3 domain of Crkl may be responsible for a salt bridge formation with a positively charged residue on C3G. Table 1. Kinetic parameters at different ionic strength conditions and different pH (in presence of 0.5 M NaCl) obtained from linear fitting of data reported in Figures 1 and 2. The NaCl concentrations reported in the left column were added to buffer Hepes 50 mM pH 7.0.

Determining the Role of D138, E140 and D141 by Site-Directed Mutagenesis
In an effort to analyze the mechanistic role of the residues D138, E140 and D141 in the binding reaction with C3G, we performed site-directed mutagenesis and generated the D138A, E140A and D141A variants of the N-SH3 domain. At first, to monitor the effect of these mutations on the stability of the domain, we performed (un)folding kinetic experiments in buffer Hepes 50 mM pH 7.0 at 25 °C. The dependence of the logarithm of the observed rate constants (kobs) obtained at different [GdnHCl] (chevron plot) for the wild-

Determining the Role of D138, E140 and D141 by Site-Directed Mutagenesis
In an effort to analyze the mechanistic role of the residues D138, E140 and D141 in the binding reaction with C3G, we performed site-directed mutagenesis and generated the D138A, E140A and D141A variants of the N-SH3 domain. At first, to monitor the effect of these mutations on the stability of the domain, we performed (un)folding kinetic experiments in buffer Hepes 50 mM pH 7.0 at 25 • C. The dependence of the logarithm of the observed rate constants (k obs ) obtained at different [GdnHCl] (chevron plot) for the wildtype, and the three variants are reported in Figure 4. All the chevron plots were globally fitted by sharing kinetic m-values [25] with Equation (2) describing a two-state scenario, suggesting the absence of intermediate(s) populating along the reaction pathway [26,27]. Importantly, none of the three mutations cause a disruption of the native state, albeit D141A mutation appears mildly destabilizing compared to D138A and E140A.  Then, we employed D138A, E140A and D141A variants in kinetic binding experiments with C3G, and we explored the effect of increasing ionic strength on the binding reaction of these variants. The results obtained are reported in Figure 5, and the calculated kinetic data are listed in Table 2. The inspection of Figure 5 and analysis of kinetic data highlight the D138A variant to be affected by increasing salt concentrations, while E140A shows no evident effects. A comparison of kinetic data obtained in the absence of NaCl shows that whilst the microscopic association rate constants calculated for D138A and E140A variants is comparable with the one obtained for the wt, an increase in koff is appreciable. All these aspects suggest a key role of E140 in the recognition event, with D138A being involved mainly in the late events. This scenario is further confirmed by the evidence of a rapidly decreasing kon upon increasing ionic strength of the solution for D138A variant, with the electrostatic charges carried by D138 residue not being involved in the recognition of a positively charged residue on C3G. On the other hand, the overall absence of effect of salt dependence on binding kinetics upon removal of E140 negative charge suggests the interactions formed by this residue acting as a prominent determinant of the early events of the binding reaction between the N-SH3 domain and C3G.
It is of particular interest to discuss the effect of D141A mutation. When employed in kinetic binding experiments, this variant did not return any measurable trace describing a change in FRET signal upon mixing, possibly due to an overall destabilization of the complex and/or sub-millisecond kinetics that could not be resolved by the stopped-flow. To further investigate this aspect and obtain a quantitative measure of the binding affinity of D141A variant with C3G, we resorted to conduct equilibrium binding experiments. Ex- Then, we employed D138A, E140A and D141A variants in kinetic binding experiments with C3G, and we explored the effect of increasing ionic strength on the binding reaction of these variants. The results obtained are reported in Figure 5, and the calculated kinetic data are listed in Table 2. The inspection of Figure 5 and analysis of kinetic data highlight the D138A variant to be affected by increasing salt concentrations, while E140A shows no evident effects. A comparison of kinetic data obtained in the absence of NaCl shows that whilst the microscopic association rate constants calculated for D138A and E140A variants is comparable with the one obtained for the wt, an increase in k off is appreciable. All these aspects suggest a key role of E140 in the recognition event, with D138A being involved mainly in the late events. This scenario is further confirmed by the evidence of a rapidly decreasing k on upon increasing ionic strength of the solution for D138A variant, with the electrostatic charges carried by D138 residue not being involved in the recognition of a positively charged residue on C3G. On the other hand, the overall absence of effect of salt dependence on binding kinetics upon removal of E140 negative charge suggests the interactions formed by this residue acting as a prominent determinant of the early events of the binding reaction between the N-SH3 domain and C3G.
interactions occurring downhill the main energy barrier of the reaction. The increase in kon which must occur in order to maintain a relatively high affinity suggests that the absence of the negative charge in position 141 strongly improves the early recognition event of C3G. Overall, our data support a scenario in which, aside from the nonspecific ionic interactions occurring between N-SH3 and C3G, an additional specific step occurs, driven at least in part by D141 residue, contributing to the final stabilization of the formed complex.  It is of particular interest to discuss the effect of D141A mutation. When employed in kinetic binding experiments, this variant did not return any measurable trace describing a change in FRET signal upon mixing, possibly due to an overall destabilization of the complex and/or sub-millisecond kinetics that could not be resolved by the stopped-flow. To further investigate this aspect and obtain a quantitative measure of the binding affinity of D141A variant with C3G, we resorted to conduct equilibrium binding experiments. Experiments were conducted by exciting samples containing a constant concentration (1 µM) of N-SH3 D141A at 280 nm and following the progressive quenching of two tryptophan residues (in position 164 and 165) fluorescence emission at increasing concentrations of dansyl-C3G 277-296 . The dependence of fluorescence measured at 350 nm as function of dansyl-C3G concentration is reported in Figure 5. Interestingly, the fitting of data with a hyperbolic equation returned a K D = 1.07 ± 0.05 µM, demonstrating that D141A variant is capable of binding. On the other hand, although this value reflects a relatively high affinity in the low µM range, it appears to be 10-fold higher than what was measured about wt from kinetic data (k off /k on = K D = 0.09 ± 0.05 µM). Such a decrease in binding affinity may be at the basis of our impossibility to time-resolve the binding reaction with a stopped-flow apparatus, with the reaction possibly occurring in the dead time of the instrument.
To further investigate this aspect, we resorted to performing displacement experiments, targeted to the direct calculation of microscopic dissociation rate constant k off . We challenged a preincubated complex of D141A NSH3 domain (2 µM) and dansyl-C3G 277-296 (10 µM) versus a high excess of nondansylated C3G (50 µM), and we could not measure any change in fluorescence emission, suggesting the k off being too high to be measured at the stopped-flow. To test this and to slow down the diffusion of molecules, we repeated the experiment increasing the viscosity of the solution adding 20% w/v sucrose to the buffer Hepes 50 mM pH 7.0. As expected, the increase in solution viscosity allowed us to obtain a displacement trace (shown in Figure 5 bottom right panel). Although the trace was satisfactorily fitted with a single-exponential equation, the k off calculated is 800 ± 30 s −1 , a value that is far beyond the resolution capability of the stopped-flow, a consistent part of the reaction occurring in the dead time of the instrument. Based on the calculated affinity of D141A variant for C3G in the absence of sucrose, this dramatic increase in k off might be accompanied by a strong increase in k on .
It is of particular interest to compare the effect of the three site-directed mutants described in this work. In fact, whilst a specific role in the early and late events of the binding reaction could be determined for D138 and E140 residues, in the case of D141, both microscopic association and dissociation rate constants were affected upon mutation. Hence, whereas D138 plays a key role in the stabilization of the formed complex and E140 is mainly involved in the early recognition events of the binding reaction, the D141 sidechain appears to play a key role in both events, in particular in the formation of electrostatic interactions occurring downhill the main energy barrier of the reaction. The increase in k on which must occur in order to maintain a relatively high affinity suggests that the absence of the negative charge in position 141 strongly improves the early recognition event of C3G. Overall, our data support a scenario in which, aside from the nonspecific ionic interactions occurring between N-SH3 and C3G, an additional specific step occurs, driven at least in part by D141 residue, contributing to the final stabilization of the formed complex.

Determinants of N-SH3 Domain Binding Selectivity
SH3 domains are widespread protein-protein interaction domains. Their main biochemical property relies in the recognition of proline rich sequences, generally identified with the P-X-X-P consensus. Given their fundamental importance in many physiological and molecular pathways in the eucaryotic cell and their role in several human pathologies, SH3 domain gained a strong attention of the scientific community since their discovery, and many of them have been characterized in their mechanisms of interaction with their ligands [4,[28][29][30][31][32]. The general structural properties of the recognition of ligands by SH3 domains are well established [30]. However, understanding the molecular determinants of specificity of SH3 domains in general is a difficult task to address, given the large amount of atypical consensus sequences that have been discovered [28].
Our group previously described in detail the mechanism of interaction of the C-SH3 domain of Grb2 with Gab2 [31,32], which is regulated by a complex allosteric mechanism. An analysis of the structure of C-SH3:Gab2 complex (PDB: 2vwf) highlights the presence of negatively charged residues in the binding pocket in direct contact with basic residues of Gab2. Importantly, the topological distribution of these negative charges appears conserved between the C-SH3 domain of Grb2 and the N-SH3 domain of Crk. Ionic strength and pH dependence analysis of the binding reaction, together with mutational analysis presented in this paper, highlight a prominent role of D138, E140 and D141 residue of Crkl in the binding of C3G 277-296 . Our data show E140 being mainly involved in the early events of recognition of a positively charged residue carried by C3G and D138 with a stabilizing effect on the formed complex. Although we could not resolve kinetics for D141A variant, the data obtained from equilibrium binding experiments and displacement experiments show an effect on binding affinity. In analogy to what was previously found for the N-SH3 domain of Crk [18] and in light of the structural and sequence alignment (Figure 3) our data suggest that D138, E140 and D141 residues may coordinate K289 residue of C3G through salt bridges formation. This electrostatic attraction driving this fundamental proteinprotein interaction may represent a key aspect of a conserved mechanism of binding in the Crk family, although on the other hand, it raises questions about how promiscuity is avoided in the intracellular milieu. Future work based on structural determination of the N-SH3:C3G complex and on extensive site-directed mutagenesis would allow us to characterize the selectivity determinants of the N-SH3 domain of Crkl, determine which specific positive residue on C3G plays a role in the recognition, and pinpoint possible longrange allosteric regulation of the binding (described also for other small protein-protein interaction modules, such as PDZ domains [33,34]) occurring simultaneously and/or finely tuning the binding interface of the domain.

Conclusions
Achieving a deep understanding of the interaction occurring between Crkl protein and its ligands is of fundamental importance to gaining useful information about the molecular basis of several physiological pathways and human pathologies in which this protein is involved. The employment of rigorous kinetic characterization of the binding reaction at different experimental conditions, together with site-directed mutagenesis, allowed us to describe in detail the roles of electrostatic forces occurring between the N-SH3 domain and a peptide mimicking one of its physiological partners, C3G. Importantly, whilst the SH3 domains are generally thought to recognize their ligands via the P-X-X-P recognition motif, our data exemplify the existence of a negatively charged stretch in the SH3, which is critical in determining the affinity between the interacting molecules. In this view, our study complements and enriches the structural knowledge on this important protein system, by providing a mechanistic insight on the role of these residues, as probed by the effect of sitedirected mutagenesis on the association and dissociation rate constant, respectively. Our data represent a first step for future structural and extensive mutational characterization of this protein system.

Site-Directed Mutagenesis
The construct encoding the N-SH3 domain of Crkl was subcloned in a pET28b+ plasmid vector. The constructs encoding D138A, E140A and D141A were obtained through site-directed mutagenesis using the QuikChange Lightning Site-Directed Mutagenesis kit (Agilent technologies) according to the manufacturer's instructions. All the mutations were confirmed by DNA sequencing.

Protein Expression and Purification
The expression of all the His-tagged constructs was performed in E. coli cells, strain BL21. Bacterial cells were grown in LB medium, with 30 µg/mL of kanamycin, at 37 • C until OD 600 = 0.7−0.8 and then induced with 0.5 mM IPTG. The cultures were grown at 37 • C for three hours after induction, kept at 25 • C overnight and then collected by centrifugation. Purification was performed resuspending the pellet in 50 mM TrisHCl, 0.3 M NaCl, pH 7.5 buffer with the addition of antiprotease tablet (Complete EDTA-free, Roche), and then sonicated and centrifuged. The soluble fraction from bacterial lysate was loaded onto a nickel-charged His-Trap chelating HP (GE Healthcare) column equilibrated with 50 mM TrisHCl, 0.3 M NaCl and pH 7.5. Protein was then eluted with a gradient from 0 to 0.5 M imidazole by using an ÄKTA-prime system. Fractions containing the protein were collected, and the imidazole was removed using a HiTrap Desalting column (GE Healthcare), with the protein purified in the final buffer of TrisHCl 50 mM, NaCl 0.3 M, pH 7.5. The purity of the proteins was analyzed through SDS-page.
Peptides mimicking the portion of C3G ranging from residue 277 to 296 (sequence VVDNSPPPALPPKKRQSAPS) in their dansylated and nondansylated variants were purchased from GenScript Biotech.

Stopped-Flow (un)folding Experiments
Kinetic (un)folding experiments were performed on an Applied Photophysics Pi-star 180 stopped-flow apparatus, monitoring the change of fluorescence emission, exciting the sample at 280 nm and recording the fluorescence emission by using a 320 nm cutoff glass filter. In all experiments, performed at 25 • C in buffer 50 mM Hepes pH 7.5, refolding and unfolding were initiated by an 11-fold dilution of the denatured or the native protein with the appropriate buffer (0 M and 6 M Guanidine HCl). For each denaturant concentration, at least five individual traces were averaged, and the final protein concentration was 1.5 µM. The fluorescence time courses obtained was satisfactorily fitted by using a singleexponential equation. The chevron plots obtained were fitted using an equation describing a two-state folding mechanism.

Stopped-Flow Kinetic Binding and Displacement Experiments
Kinetic binding experiments were performed on a single-mixing SX-18 stopped-flow instrument (Applied Photophysics), by mixing a constant concentration (0.5 µM) of C3G dansylated versus increasing concentrations of N-SH3 at 10 • C. Ionic strength dependence of wt N-SH3 was performed with concentrations ranging from 1 to 5 µM for 0 M NaCl condition and from 0.

Equilibrium Binding Experiment of D141A
Equilibrium experiment on D141A (fixed concentration at 1 µM) was carried out on a Fluoromax single-photon counting spectrofluorometer (Jobin-Yvon, Newark, NJ, USA), by mixing the construct with increasing dansyl-C3G concentrations. Experiments were performed at 10 • C, using a quartz cuvette with a path length of 1 cm, in 50 mM Hepes pH 7.0 and measuring the change in fluorescence of the naturally present tryptophan residues in position 164 and 165 at increasing concentration of dansyl-C3G. The excitation wavelength was 280 nm, and fluorescence spectra were recorded between 300 and 400 nm.