Thermodynamic characterization of two homologous protein complexes: Associations of the semaphorin receptor plexin-B1 RhoGTPase binding domain with Rnd1 and active Rac1

Abstract

Plexin receptors function in response to semaphorin guidance cues in a variety of developmental processes involving cell motility. Interactions with Rho, as well as Ras family small GTPases are critical events in the cell signaling mechanism. We have recently determined the structure of a cytoplasmic domain (RBD) of plexin-B1 and mapped its binding interface with several Rho-GTPases, Rac1, Rnd1, and RhoD. All three GTPases associate with a similar region of this plexin domain, but show different functional behavior in cells. To understand whether thermodynamic properties of the GTPase–RBD interaction contribute to such different behavior, we have examined the interaction at different temperatures, buffer, and pH conditions. Although the binding affinity of both Rnd1 and Rac1 with the plexin-B1 RBD is similar, the detailed thermodynamic properties of the interactions are considerably different. These data suggest that on Rac1 binding to the plexin-B1 RBD, the proteins become more rigid in the complex. By contrast, Rnd1 binding is consistent with unchanged or slightly increased flexibility in one or both proteins. Both GTPases show an appreciable reduction in affinity for the dimeric plexin-B1 RBD indicating that GTPase binding is not cooperative with dimer formation, but that a partial steric hindrance destabilizes the dimer. However, a reduced affinity binding mode to a disulphide stabilized model for the dimeric RBD is also possible. Consistent with cellular studies, the interaction thermodynamics imply that further levels of regulation involving additional binding partners and/or regions outside of the RhoGTPase binding domain are required for receptor activation.

Introduction

Recently, large-scale proteomics-based approaches are elucidating the interactomes of proteins in cell signaling. This analysis is so far qualitative in most of the screens. At the structural level of characterization, a quantitative understanding of protein–protein interactions is equally difficult: even if high resolution structures of the unbound proteins are available, current protein docking algorithms and free energy calculations are not yet sufficiently accurate to predict binding affinities and specificities for most systems. The experimental characterization of protein–protein interaction thermodynamics, such as the determination of the enthalpy of binding (ΔH), entropy of interaction (ΔS), and specific heat (ΔC_p), thus, continues to provide critical information about the nature of the interacting surfaces. Furthermore, these thermodynamic parameters also indicate the conformational changes that accompany protein–protein interactions in most cell signaling events.^1–3 Using site specific mutagenesis, minimal binding interfaces or hotspot amino acids of the interactive protein can also be mapped.^4–9

Several themes are emerging for the systems that have been studied extensively, such as Ras GTPase–effector protein interactions: here, the change of enthalpy, entropy, and heat capacity was shown to be similar for the association of the same effector with different Ras family members, whereas, the thermodynamic contributions are different when the same Ras protein interacts with different effectors.¹⁰ In addition, protein–protein interactions are often cooperative or competitive, for example, the binding of Cdc42 is coupled to a folding–unfolding equilibrium in an effector protein, WASP.¹¹ Also, a monomer–dimer equilibrium of one binding partner can be coupled to the functional behavior of the other protein. The latter is the case for the guanine nucleotide exchange factor, Cool-2. As a monomer Cool-2 can activate both Rac1 and Cdc42, whereas the dimer specifically binds to Rac1, effecting signaling through this GTPase.¹²

Plexin-B1 is a transmembrane receptor, which is known to regulate several cellular processes including axonal guidance in the developing nervous system.^13–15 Upon activation by the ligand, semaphorin, plexin initiates a variety of signaling processes, which involve several small GTPases of the Ras and Rho families (R-Ras, Rac1, Rnd1, and RhoD).^16,17 The GTPases are known to be important signaling components in the plexin system that regulate cytoskeletal dynamics and cell adhesion.¹⁸ Plexins are unique amongst transmembrane receptors because several cytoplasmic regions interact directly with small GTPases.

Specifically, plexins possess a domain with homology to GTPase activating proteins (GAPs).^19,20 As part of their activation, plexin-A1 and -B1 show GAP activity toward R-Ras in the presence of Rnd1. R-Ras deactivation in turn inhibits integrin-mediated cellular adhesion.^19–21 Although active Rac1 is thought to be required to localize the receptor to the membrane and to initiate conformational changes in plexin,^22–24 it does not appear to be sufficient to activate plexin's GAP function itself (Negishi, Personal Communication).^20,25 Instead, a second Rho family GTPase, Rnd1 is needed to bind in the same location in a proposed stepwise activation process.²¹ Similarly, Rac1 has no effect on the interaction between plexin-B1 and an exchange factor, PDZ-Rho GEF that binds to the C-terminus of plexin-B1, whereas association of the receptor with Rnd1 appears to enhance the interaction.^25,26 These observations suggest that Rnd1 causes a different set of conformational changes, compared with Rac1 binding. However, the biophysical bases of these proposed mechanisms are not yet clear.

We have recently begun to characterize the structural and dynamical features of small GTPase, Rnd1, and Rac1 interaction with the Rho GTPase binding domain (RBD) of the human plexin-B1 receptor.^27–29 The interaction has been mapped to a similar surface of the RBD.²⁸ In this study, we have investigated the Rho GTPase–plexin RBD interactions quantitatively in different solution conditions to obtain a thermodynamic insight into the nature of the interactions. The dissociation constant, K_d, measured for Rnd1 and active Rac1 with the receptor domain was found to be in the 6–10 μM (±1 μM) range, suggesting a simple model, which involves the plexin-RBD as an effector domain, undergoing a common conformational change for receptor activation, such as the destabilization of a dimeric form.²⁸ However, as mentioned above, there is also evidence that the mode of interaction of active Rac1 and Rnd1 with plexin is different. Recently, the structure of a dimeric plexin-B1 RBD: Rnd1 complex has been deposited in the protein databank, whereas the plexin-B1 complex with Rac1 has not yet been crystallized. It is likely that such differences are manifest in the detailed thermodynamics of the interaction of plexin-B1 RBD with the two different GTPases.

Using isothermal titration calorimetry (ITC), we find that the detailed thermodynamics of the interaction are remarkably different, even though similar binding surfaces are involved on the side of plexin. Furthermore, we previously suggested that a principal function of the binding event is the conformational change that affects a dimerization loop, possibly altering the association between cytoplasmic domains.²⁷ Here, we examine GTPase binding to the dimeric RBD protein, as well as to a model of the RBD in which two cysteine residues have been added and the dimeric form has been stabilized by disulphide bond formation. The results not only show that binding affinities are sensitive to dimerization but also suggest that factors outside the RBD are likely to play a considerable role in the signaling mechanism of plexin-B1.

Results

Binding thermodynamics and their temperature dependence

The interaction of Rac1 and Rnd1 with the Plexin-B1 RhoGTPase binding domain (RBD) is exothermic and depends on the temperature and pH of the medium. Table I shows the binding constant (K_a = 1/K_d), free energy change (ΔG), entropy change (ΔS), and enthalpy change (ΔH) obtained from the ITC experiments (typical binding isotherms are shown in Fig. 1). The binding affinity and free energy are similar for both the Rac1–RBD and Rnd1–RBD interactions [∼K_d 6 μM (±1 μM)], though their binding enthalpy and entropy are quite different. The interaction between Rac1 and the RBD is driven by a favorable enthalpy change overcoming an unfavorable entropy change. By contrast, the Rnd1 interaction with the RBD is driven by both a favorable enthalpy and a slightly favorable entropy change. The enthalpy change that accompanies Rac1/Rnd1 binding to the RBD shows a considerable dependence on temperature (Supporting Information Fig. S1a); from moderately exothermic at 20°C to highly exothermic at 30°C. ΔH is negative and linearly decreases with temperature (Supporting Information Fig. S1b). However, the change of free energy (ΔG∼ −7.1 kcal/mol) is found to be nearly temperature independent, consistent with an often observed enthalpy–entropy compensatory effect. The change of entropy (TΔS) is slightly positive for Rnd1–RBD complex formation and decreases upon increasing temperature, whereas TΔS is negative for Rac1–RBD complex formation, again further decreasing with increasing temperature.

Table I.

Thermodynamic Parameters for the Association of Rac1/Rnd1 with the Plexin-B1 RhoGTPase Binding Domain (RBD) in Phosphate Buffer, pH 7.0

Complex	Temp. (K)	Ka (M−1) × 105	Kd (μM)	ΔH (kcal/mol)	ΔG (kcal/mol)	TΔS (kcal/mol)	ΔCp (kcal/mol/K)
Rac1-Plexin RBD	293	2.4 (0.12)	4.1 (0.2)	−8.6 (0.4)	−7.34 (0.03)	−1.3 (0.4)	−0.59 (0.02)
	298	1.6 (0.14)	6.1 (0.5)	−12.1 (0.2)	−7.10 (0.05)	−5.0 (0.1)
	303	1.5 (0.2)	6.6 (0.8)	−14.5 (0.2)	−7.06 (0.08)	−7.4 (0.1)
Rnd1-Plexin RBD	293	2.8 (0.3)	3.5 (0.4)	−4.6 (0.3)	−7.43 (0.07)	+2.9 (0.2)	−0.22 (0.01)
	298	1.8 (0.15)	5.5 (0.4)	−6.7 (0.3)	−7.17 (0.05)	+0.5 (0.2)
	303	1.8 (0.13)	5.5 (0.4)	−6.8 (0.2)	−7.17 (0.04)	+0.4 (0.1)

Maximum error is shown in brackets.

Figure 1.

Typical binding isotherm for GTPase association with the human plexin-B1 RhoGTpase binding domain (RBD) from isothermal titration calorimetry (ITC) in phosphate buffer pH 7.0 with 50 mM sodium chloride, 4 mM MgCl₂, and 1 mM TCEP at 25°C, (a) Rac1-RBD, (b) Rnd1-RBD.

The change of specific heat (ΔC_p) is determined from the slope of the linear dependence of ΔH with temperature. It is found that the Rac1–RBD complex exhibits a more negative heat capacity upon association as compared with the Rnd1–RBD complex (−0.59 and −0.22 kcal/mol K, respectively). In the thermodynamic analysis of protein folding and binding, the change of heat capacity is empirically associated with the change in accessible surface area due to protein dehydration. The negative ΔC_p indicates that both GTPase–RBD complexes burry more hydrophobic than hydrophilic surface, suggesting that hydrophobic amino acids play a major role in forming protein–protein contacts.

Effect of pH

Both Rac1 and Rnd1 binding are affected by pH of the buffer. At lower pH, the proteins have an increased tendency to form complexes, whereas the association constant decreases upon increasing the pH of the medium. This is accompanied by a corresponding change in enthalpy, showing that at a pH lower and higher than 7.0, the interaction is slightly less exothermic for both Rac1 and Rnd1 (Table II, Supporting Information Fig. S2). Importantly, no GTPase–RBD interaction could be detected by calorimetry at a pH of 8, suggesting that histidine residues are likely to be important (see below).

Table II.

pH Dependence Thermodynamic Parameters for the Association of Rac1/Rnd1 with the RBD

Complex	pH	Ka (M−1) × 105	Kd (μM)	ΔH (kcal/mol)	ΔG kcal/mol	TΔS (kcal/mol)
Rac1-Plexin	6.0	4.4 (0.45)	2.2 (0.2)	−10.8 (0.2)	−7.7 (0.05)	−3.1 (0.14)
	6.4	1.9 (0.2)	5.2 (0.5)	−11.5 (0.3)	−7.2 (0.06)	−4.3 (0.24)
	7.0	1.6 (0.14)	6.2 (0.5)	−12.1 (0.2)	−7.1 (0.01)	−5.0 (0.17)
	7.4	1.5 (0.2)	6.6 (0.8)	−10.5 (0.3)	−7.1 (0.06)	−3.4 (0.22)
	8.0	–	–	–	–	–
Rnd1-Plexin RBD	6.0	3.4 (0.27)	2.9 (0.2)	−4.6 (0.1)	−7.5 (0.05)	+2.9 (0.05)
	6.4	3.2 (0.35)	3.1 (0.3)	−7.0 (0.15)	−7.5 (0.06)	+0.5 (0.08)
	7.0	1.8 (0.15)	5.5 (0.4)	−7.1 (0.3)	−7.2 (0.05)	+0.1 (0.06)
	7.4	1.8 (0.2)	5.4 (0.6)	−4.5 (0.6)	−7.2 (0.07)	+3.7 (0.6)
	8.0	–	–	–	–	–

Experiments were carried out in 50 mM phosphate buffer containing 4 mM MgCl₂, 1 mM TCEP, 50 mM sodium chloride, at different pH. – shows no measurable interaction, and maximum error is shown in brackets.

Effect of salt and buffer ionization

The extent of electrostatic contacts in the GTPase–RBD interaction was tested in ITC experiments at an elevated salt concentration of 250 or 500 mM NaCl. We found that at ∼40 μM protein, GTPase–RBD interactions are only modestly if at all affected (Table III). Within the uncertainty of the measurement, the proteins show an increased tendency to form complexes at higher salt concentration (results with higher protein concentrations are discussed below). ITC experiments were also carried out in buffers, which posses a low and high enthalpy of ionization. Specifically, phosphate and tris-HCl buffers at 25°C and pH 7.0, have enthalpies of ionization of 1 and 11 kcal mol⁻¹, respectively.³⁰ The binding enthalpy of association is found to be −12.5 and −6.5 kcal mol⁻¹ for the RBD–Rac1 and −7.1 and −5.5 kcal mol⁻¹ for RBD–Rnd1 in phosphate and tris-HCl buffer, respectively, at pH 7. This difference typically arises because of a proton transfer on complex formation. Furthermore, this difference decreases in acidic medium. The number of protons involved can be calculated from the apparent binding enthalpy (ΔH^app), which is sum of the enthalpy of association and ionization of the buffer.³⁰

Table III.

Effect of salt on the association of Rac1/Rnd1 with the RBD

Complex	Salt	Ka (M−1) × 105	Kd (μM)	ΔH (kcal/mol)	ΔG (kcal/mol)	TΔS (kcal/mol)
Rac1-RBD	Low	1.6 (0.2)	6.2 (0.6)	−12.1 (0.2)	−7.1 (0.07)	−5.0 (0.07)
	High	2.0 (0.3)	5.0 (0.6)	−10.5 (0.2)	−7.2 (0.07)	−3.3 (0.11)
Rnd1-RBD	Low	1.8 (0.15)	5.5 (0.4)	−7.1 (0.3)	−7.2 (0.05)	+0.1 (0.1)
	High	2.8 (0.3)	3.6 (0.3)	−7.2 (0.2)	−7.4 (0.06)	+0.2 (0.1)

Experiments were carried out in 50 mM phosphate buffer, pH 7, containing 4 mM MgCl₂, 1 mM TCEP, sodium chloride [0 mM (low) and 250 mM (high)]. Maximum error is shown in brackets.

(1)

(2)

0.6 protons are transferred in RBD–Rac1 and 0.3 in RBD–Rnd1 complex formation at pH 7.0. As ΔH in tris–HCl is less exothermic compared with phosphate buffer, protons are taken up during complex formation. Partial numbers of protons arise because a His group with pKa of ∼7.0 in the free protein, for example, would present a 1:1 population of protonated (+1 charge) and deprotonated (neutral) forms. The difference to the free form would then be measurable as a partial change in protonation status. From the pH analysis of the binding affinity (Table II), it is shown that at lower pH, the interaction is stronger, consistent with this idea. Residues that may be responsible for this behavior in the plexin RBD–GTPase interaction are discussed below.

Plexin-B1 RBD monomer–dimer equilibrium and the effect of GTPase binding

Earlier, we have found that the RBD dimerizes at moderate protein concentrations (∼25 μM at 4°C),²⁷ and we presented gel filtration data showing that GTPase binding appeared to disrupt RBD dimerization.²⁸ The dimerization affinity decreases with increasing temperature (K_d of ∼40 μM at 30°C; Supporting Information Table S1 and Fig. S3a). Above 100 μM, the RBD predominantly exists in the dimeric state (Supporting Information Fig. S3b) and the question arises, how the GTPases interact with this state of the protein, given that the dimerization region and the GTPase interaction surface are adjacent to one another in the structure and may partially overlap.²⁸

To examine this issue more closely, we recorded ¹H-¹⁵N HSQC-TROSY NMR spectra at different concentrations of the RBD in which all carbon-attached hydrogens are perdeuterated, yielding high-quality spectra for the 35–37 kDa complexes. Spectra were recorded with RBD: GTPase ratios of 1:0, 1:1, 1:2, 1:3, and 1:4. The RBD concentration was 30 or 150 μM to show protein that is largely monomeric and dimeric, respectively, at 25°C. Representative regions of the corresponding spectra of unbound RBD are plotted in Figure 2(a,b), showing that resonances belonging to the monomer–dimer interface are diminished in intensity [res. 1820–1824, 1830 and 1836–1837 as well as res. 1775–1778]. Several are exchange broadened completely [residues 1827–1835]. Upon GTPase binding, the intensity of resonances that arise from the structured parts of the RBD generally decreases and many amides at or near the binding region exhibit significant change in chemical shift as reported.²⁸ However, we find that several resonances from residues near the dimerization interface gain in intensity or reappear. In case of both Rac1 and Rnd1 binding [Fig. 2(d,f), respectively], resonances for amides 1820, 1822, 1824, 1828, 1830–32, 1838 as well as 1775–1778 are clearly visible at a concentration ratio of 150 μM RBD to 450 μM GTPase. We monitored the intensity of resonances also at an even higher GTPase concentration and in regions outside the dimerization and GTPase binding regions (Supporting Information Fig. S4). It is apparent that intensities of most resonances (except, in fact, those at the GTPase binding interface) dramatically decrease with increasing concentration of GTPase (>300 μM), especially in the case of Rac1. Because most of the protein is affected, such interactions are likely to be nonspecific. By contrast, such a decrease is not seen at a protein concentration at 30 μM RBD to 450 μM GTPase [Fig. 2(c,e)]. This suggests that a fraction of higher order oligomers may form at high RBD concentrations in presence of the RBD–GTPase complex.

Figure 2.

¹H-¹⁵N HSQC-TROSY spectra of ²H,¹⁵N Plexin-B1 RBD at a concentration of (a) 30 μM and (b) 150 μM recorded at 800 MHz. Several resonances that arise from amides at the dimerization interface are labeled. (c) and (d) the RBD at 30 μM and 150 μM with 450 μM Rac1.Q61L and (e) and (f) RBD at 30 μM and 150 μM with 450 μM Rnd1, respectively.

Effect of dimerization on GTPase binding thermodynamics

To examine whether GTPase binding and RBD dimerization oppose each other thermodynamically, we monitored the interaction at a higher concentration of Plexin-B1 RBD (placing 100 μM in the ITC cell—a concentration at which the protein is mostly dimeric). The binding affinity of both Rac1 and Rnd1 for the RBD is reduced [from K_d ∼6 to ∼14 μM (±1 μM)] suggesting that dimerization partially opposes GTPase binding. The thermodynamic parameters (ΔH and ΔS) of Rnd1 binding become more similar to those of Rac1. It is apparent that these changes present a common trend at high protein concentration, such as the formation of higher order oligomers (see NMR data above). The latter, nonspecific association process obscures the monomer–dimer thermodynamics, and thus, high-protein concentration studies have limited value.

To test the opposition of GTPase binding to dimerization at lower protein concentration, the dimeric form of the RBD needed to be stabilized. We generated a model for this dimeric protein by covalently crosslinking the protein at cysteine residues that have been introduced into the dimerization loop (L1829C, L1833C). The sites were chosen based on the crystal structure of the RBD dimer, where residues L1829 and L1833 are 4.5 Å from one another, a distance required for disulphide bond formation.³¹ Computational modeling also shows that the thiol sidechains have chiral angles favorable for a covalent linking. The thiol linkages are obtained through hydrogen peroxide treatment, and the extent of dimerization is verified by a number of tests [Fig. 3(a,b) and materials and methods]. It is found that the dimeric form of the RBD appreciably decreases binding affinity for both GTPases [K_d > 11 μM (±1 μM)] (Table IV). By contrast to the data at high RBD concentration, the covalently linked dimer behaves similarly to the monomer RBD on binding to Rnd1, except for a slight further increase in positive entropy change, which appears to account for the weaker binding affinity. These results show that GTPase binding does not cooperate but weakly opposes dimerization.

Figure 3.

(a) UV absorbance at 280 nm for gel filtration of plexin-B1 RBD mutant (L1829CL1833C) before (^____) and after reaction with hydrogen peroxide (----) showing monomer and monomer/dimer respectively. (b) Absorption spectra of HPDP reacted with monomer mutant RBD (^____) and purified dimer RBD (----) showing free thiol groups (by formation of pyridine-2-thione λmax, 343 nm).

Table IV.

Thermodynamic Parameters for the Association of Rac1/Rnd1 with di-cys RBD (1829C1833C) Monomer and Covalently Attached Dimer

PlexinB1 (1829C1833C)	GTPase	Ka (105)(M−1)	Kd (μM)	ΔH (kcal/mol)	ΔG (kcal/mol)	TΔS (kcal/mol)
Monomer	Rac1	2.0 (0.2)	5.0 (0.4)	−13.5 (0.7)	−7.2 (0.05)	−6.0 (0.07)
	Rnd1	1.8 (0.1)	6.0 (0.5)	−6.9 (0.3)	−7.2 (0.04)	+0.3 (0.1)
Covalently bound (constitutive) Dimer	Rac1	0.8 (0.09)	12 (1.3)	−14.5 (0.3)	−6.69 (0.04)	−7.8 (0.3)
	Rnd1	0.9 (0.1)	11 (1.2)	−6.6 (0.1)	−6.76 (0.06)	+1.2 (0.04)

Maximum error is shown in brackets.

Discussion

The Rho GTPase binding domain (RBD) of plexin-B1 is remarkable in that it interacts with several significantly diverse small GTPases, including active Rac1 and Rnd1. Rac1 is involved in lamellipodia formation in cells, whereas Rnd1 expression inhibits formation of stress fibers and integrin-based focal adhesions.³² The GTPase structures are very similar: the core regions of Rnd1 and Rac1.GMPPNP superimpose to 1.05 Å and the region typically interacting with effectors residues 5–85 in Rac1 is 45% sequence identical to the corresponding region in Rnd1. Nevertheless, there is considerable difference in the sequence and structural surface detail (Fig. 4 and Supporting Information Fig. S5) and in GTPase function.²⁴ Rac1 has more charged residues in the switch II region (2 for Rnd1; 4 for Rac1). Rnd1 has a larger number of nonpolar hydrophobic residues (85 for Rnd1; 79 for Rac1), also in the switch I and II regions where Rnd1 has one extra hydrophobic residue in each region with respect to Rac1. Although modeling of a common possible interaction surface based on the electrostatic and nonpolar complementation of surfaces of the unbound proteins is a speculative exercise, it is likely that differences in the character of such and nearby surfaces will feature in the detailed thermodynamics of GTPase–plexin complexes formation, and thus have functional consequences, motivating the current study.

Figure 4.

Schematic representation of the mainchain fold of Rac1, plexin-B1 RBD, and Rnd1 [PDB ids. 1HM1, 2JPH, and 2CLS, respectively] showing the regions involved in the GTPase-plexin interaction in case of the RBD and Switch I and II in case of the GTPases. Below, electrostatic accessible surface representation in the same orientation reveals a hydrophobic ridge (white line) with complementary positive charge on the side of plexin and negative charge on the side of the GTPases. (Binding surfaces are facing the viewer. To generate putative complexes the RBD is rotated about axis in page indicated and superimposed). This picture implies a possible common interaction surface on the side of the GTPases, although such suggestion is speculative because the charge distribution is highly sensitive to the conformation of sidechains and local conformational changes that are likely in the proteins upon binding. The proteins are rendered with pymol using the APBS tool.

Similar association characteristics

Both GTPases (Rac1 and Rnd1) have a similar binding affinity (K_d) and free energy of association (ΔG) with the Plexin-B1 RBD. The heat capacity change is negative for both interactions, suggesting that the majority of the contacts involve nonpolar residues. The dominance of the hydrophobic contribution for binding is also evident by the modest effect of ionic strength on the plexin RBD–GTPase interaction at moderate protein concentrations (40 μM). Electrostatic interactions between molecules decrease in presence of ionic species, due to a screening of charged residues. As shown for complex formation of the small GTPases Ras and Rap with their effectors, the binding affinity diminishes up to 2-fold when the amount of sodium chloride in the buffer is increased from 0 to 100 mM.¹⁰ In case of the Rac1- and the Rnd1 complex with the RBD, the interaction is only moderately sensitive to salt concentration and the binding affinity slightly increases showing that the interaction is hydrophobic rather than electrostatic in nature. This implies that the surfaces for the Rac1-plexin-B1 interaction are either poorly matched in their electrostatics or are largely hydrophobic in nature (see below).

The thermodynamic analysis provides a remarkable detail on the similarities and differences of the plexin RBD Rac1 or Rnd1 complex formation. From measurements in different buffers, we find that both association processes involve a partial proton uptake event at pH 7.0. Histidines are obvious candidates for this effect. For example, previously, it has been reported that Asp38 in the switch I region of Rac1/Cdc42 interacts with a Histidine of an effector protein containing a CRIB motif.^33,34 The Ras–Raf complex formation was also studied in different buffers and revealed a partial proton release.¹⁰ Recently, we have mapped the surfaces of the Plexin-B1 RBD and of the Rac1 GTPase^27,29 that are involved in complex formation using NMR spectroscopy. None of the Histidine residues in the Rac1 or Rnd1 GTPase is close to the interaction surface with the plexin-B1 RBD; however, the RBD has three Histidine residues with His1814 and His1838 located at the GTPase binding interface. NMR studies of the free RBD domain and in complex with Rac1 at different pH values revealed that His1814 does not titrate with pH in the latter (Hota, Chugha, and Buck, In preparation). Mutation of His1814 to Ala considerably reduces the binding affinity for both GTPases, Rac1 [∼K_d 10 μM(±1 μM)], and Rnd1 [∼K_d 12 μM(±1 μM)] and is thus required to be protonated for high affinity complex formation (Supporting Information Fig. S6). However, mutation of His1838 to Ser also reduces binding affinity by virtue of its location at the binding interface.^27,28 There is no good test for the need for Histidine protonation apart from pH itself.

Differences in the thermodynamic contributions for complex formation

Despite the similarities just outlined, our thermodynamic characterization reveals that the association is quite different in terms of enthalpy and entropy changes (ΔH, TΔS). The Rac1–RBD complex formation exhibits a highly exothermic reaction with unfavorable entropy (i.e. entropy of the system decreases). On the other hand, the Rnd1–Plexin-B1 RBD complexation is associated with slightly favorable entropy (i.e. entropy of the system increases or close to zero). A priori it is not known for a protein–protein interaction whether the changes in entropy reflect changes in the flexibility of the polypeptide chain or whether reorganization of solvent plays a dominant role. However, recently we have reported an NMR dynamics analysis for Rac1–RBD complex formation, comparing protein internal mainchain dynamics on the ps-ns timescale between the free and the bound proteins.²⁹ It could be shown that while there is no net change in RBD dynamics, significant parts of the Rac1 polypeptide chain become more rigid upon binding. Remarkably, several of these regions are distant from the interaction surface with the RBD (Rac1 insert helix 3 and helix 6), suggesting an allosteric effect.²⁹ Although mainchain fluctuations are only one contributor to the overall entropy change, this NMR analysis of changes in the protein dynamics, and therefore entropy, are consistent with the global change seen in ITC measurements for Rac1 binding. NMR measurements have not yet been made for Rnd1, but the NMR spectra of the RBD suggest that the perturbations to its structure and dynamics on binding to Rnd1 are limited.

In principle, the entropy gain in case of the Rnd1–RBD interaction could be due to a more hydrophobic interaction, compared to Rac1. This is indicated considering the sequence difference between the two GTPases in the switch I and II regions (see above, Supporting Information Fig. S5) and would release a greater number of water molecules that were partially ordered around hydrophobic groups in the unbound RBD and Rnd1 to the bulk solvent on complex formation. However, ΔC_p, derived from the temperature dependence, suggests that complex formation with Rac1, rather than with Rnd1 is associated with a greater change in hydrophobic surface area. This raises the possibility that it is the Rnd1 that becomes more disordered on binding, that is, an amount of hydrophobic surface area that is exposed compensates the common area that is buried on complex formation.

Substantial conformational and dynamical changes in Rnd1 are not implausible because this GTPase is not a highly stable protein. Although the thermal unfolding is irreversible after the GTP nucleotide is displaced from the protein, the approximate T_m of this transition is ∼49°C as opposed to ∼66°C for Rac1 (Hota and Buck, Unpublished data). This difference in global stability may play a role in the overall character of the interaction of Rac1 and Rnd1 with plexin B1. Although plexin is an effector protein in both cases, Rnd1 may be poised to undergo a more substantial conformational change: The bound GTPase will then be able to interact with other regions of the plexin cytoplasmic domain, such as the segments which are homologous to a Ras GAP protein and which surround the RBD. The bound GTPase may also interact with other regulatatory and adaptor proteins, which remain to be discovered. It is intriguing, in this context that studies reporting the activation of plexin-A1 and -B1 GAP function towards R-Ras have been carried out with Rnd1 and not Rac1. From the published results and discussions it is not clear whether Rac1 also can activate the GAP function of the receptor, but this appears to be unlikely (M. Negishi, person. Communication). However, a sequential model involving the binding of both GTPases has been proposed by Toyofuku et al.³⁵ This model and our observations emphasize the differences between the Rac1–plexin and Rnd1–plexin interaction.

Thermodynamics of dimerization and GTPase binding

We have shown previously that the RhoGTPase interaction surface of plexin-B1 RBD and the dimerization interface are adjacent to one another and may partially overlap. The observation led to the proposal of a model for receptor activation in which the GTPase binds to the RBD as an effector for the receptor, leading to a disruption of the dimer or at least to a considerable conformational change. The model is consistent with the experimental data showing that RBD dimerization is at least 5–8-fold weaker than the affinity for the GTPases for a monomeric protein. Furthermore, gel filtration and NMR spectra (the latter shown in detail in this report) reinforce this conclusion. Recently, an X-ray crystal structure has been deposited in the protein databank showing the plexin-B1 RBD in complex with Rnd1. Surprisingly, the Rnd1 is bound to the dimeric form of the RBD in the crystal (in a Rnd1-RBD-RBD-Rnd1 arrangement), suggesting that the GTPases can also bind to the RBD interface when the adjacent and partially overlapping dimerization loop is not flexible.

In principle, GTPase binding and dimerization could either reinforce or oppose each other, or there could be no effect at all. The observation of dimer disruption in solution (NMR and gel filtration) is made during titrations in which an excess of GTPase is added. This may push the equilibrium to the GTPase: RBD complex by mass action. Although not necessarily unphysiological, this does not completely demonstrate that the binding to monomer is thermodynamically preferred over binding to the dimer. Clearly, the dimeric form has some considerable affinity: if dimerization and GTPase were opposed to one another, it should be possible to construct a thermodynamic cycle in which the free energy of GTPase binding to dimer would be diminished considerably by the free energy needed to disrupt the dimer. An experimental disadvantage is that high protein concentrations are needed to ensure that the RBD is in the dimeric state, and we have shown in this report that such high protein concentrations alter the thermodynamics of the interaction. Thus, we designed a protein that has been stabilized in the dimeric form by the formation of interdomain loop disulphides. This model allowed us to obtain thermodynamics data on GTPase biding to the RBD dimer at low concentration. The results show that this ultrastable dimeric RBD also binds Rac1 and Rnd1 but with a considerably lower affinity than the monomeric form (Table IV). It is likely that the dimerized RBD is more compatible with the crystal lattice when bound to Rnd1. There has been no success to date in crystallizing the monomeric RBD:Rnd1 complex, or indeed any complex of the RBD (monomer or dimer) with Rac1 (Tong and Park, Personal Communication). Again this suggests that the nature of the Rac1- and Rnd1–RBD interactions is somewhat different.

Functional implications

The character of the GTPase PlexinB1 RBD interaction is quite different to the thermodynamics and kinetics of Ras–effector complex formation, which has been studied extensively. Here, polar interactions are significant and an electrostatic steering is thought to play an important role, leading to the rapid formation of an encounter complex.¹⁰ In plexin, the role of GTPases is to cause a conformational change in the cytoplasmic region of plexin and hydrophobic interaction play an important role. This may allow a more persistent signaling, or in this case, receptor activation.

Our thermodynamic analysis of plexin-B1 RBD complex formation with active Rac1 and Rnd1 is consistent with the model we recently proposed involving a destabilization of a dimeric inactive form of the receptor. Intriguingly, the more detailed study reported here shows that dimerization and GTPase binding are not completely opposed to one another and the difference in their binding affinities differ by less than an order of magnitude. This may ensure that the interactions are easily reversible and, as seen in other systems,¹⁵ suggests that further levels of regulation, such as the binding of additional proteins or additional interactions are required. That the dimeric RBD also has considerable affinity for the GTPases, as shown here with a model for the constitutive dimer, could also be important for targeting of the inactive receptor to the plasma membrane. Both Rac1 and Rnd1 contribute to this localization process.^22,26

The thermodynamic characterization of the complexes point to a difference in the binding of Rac1 and Rnd1, indicating that the latter GTPase becomes in part more flexible, while the former becomes more rigid (Fig. 5). Rigidification of parts of the mainchain of Rac1 on binding to the plexin-B1 RBD, even of regions far removed from the binding interface, has been confirmed by an NMR relaxation study.²⁹ This change in protein dynamics may be sufficient to generate new binding affinities, as the role of dynamics in protein allostery is becoming established.^36,37 A similarly extensive NMR study is needed to reveal the changes in protein conformation and dynamics that occur in Rnd1 upon binding to the plexin-RBD in solution. However, our findings and those of others regarding GTPase specificity³⁸ already point to regions of the GTPases, other than the switch I and II regions that interact with the plexin RBD. An example of a small Rho GTPase–protein interaction that does not involve the switch I and II regions, the RhoE/Rnd3 complex formation with Rock-I kinase, has been reported recently.³⁹ Such allosterically regulated surfaces are likely to be critical for the next level of contacts with other regions of the plexin-B1 cytoplasmic portion of the receptor or with other binding partners, bringing about the differential effects of Rac1 and Rnd1 on GAP activity and on PDZ-RhoGEF binding of the full receptor.

Figure 5.

Schematic representation of the thermodynamic differences of Rac1/Rnd1 GTPase association with plexin suggesting a conformational change in the Rnd1 GTPase.

Materials and Methods

Protein expression and purification

The Rho GTPase binding domain (RBD) of human plexin-B1 (res. 1742–1862) and the Rho GTPases, Rac1, and Rnd1 were expressed in E. coli BL21 cells and purified following previous protocols.²⁷ Rac1 was obtained at higher yield in a constitutively active form (RacQ61L/C178S/C-term truncated after K184).²⁹ The data for full length and GMPPNP-activated Rac1 were similar to that reported here. All proteins were further purified on a Superdex-75 size exclusion column. Proteins for ITC experiments were extensively washed with phosphate and TRIS buffer at the desired pH, containing 50 mM sodium chloride, 4 mM MgCl₂, and 1 mM TCEP.

Preparation of disulphide-linked plexin RBD

RBD dimers were stabilized by the formation of intermolecular disulphide crosslinks in the dimerization loop. The location of disulphides was determined according to our recent published crystal structure,²⁸ suggesting that mutations L1829C and L1833C (in a no-cys background) can be designed to form covalent bonds across a short dimerization β-sheet. The disulphide bonds were formed by overnight incubation of the mutant RBD with hydrogen peroxide (1:1 ratio) at 50 μM concentration.⁴⁰ Monomeric and dimeric proteins were separated using a Superdex-75 size exclusion column, and the extent of disulphide bond formation was confirmed using the chemical reagent, BIOTIN HPDP. In the reaction with free thiol groups, the release of pyridine-2-thione is measured by its absorbance at 343 nm.⁴¹ The possibility exists that the covalently linked RBD dimer has less binding affinity toward the GTPases because of (a) the Cysteine mutations and (b) strain or conformational changes that are introduced on disulphide bond formation. Several pieces of evidence argue against these scenarios. To counter (a), ITC shows that the monomeric mutant RBD protein binds the GTPases as well as the wild type monomeric protein. To counter (b), the binding affinity of the constitutive dimer is actually similar to (and even better than) binding of higher concentration RBD with GTPase. NMR spectra of the covalent unbound dimer again do not show significant chemical shift perturbation for residues near the GTPase binding site (data not shown).

Calorimetric measurements

Interaction studies of plexin-B1 RBD and Rac1/Rnd1 were carried out using an isothermal titration microcalorimeter (VP-ITC, MicroCal). Both the proteins were exchanged with identical buffer prior to the experiment (see above). One binding partner (e.g. Rac1/Rnd1) was placed into a sensitive temperature controlled cell (volume 1.43 mL) and the second protein (e.g. plexin-B1) was placed into a syringe immersed into the cell (capacity 0.295 mL). The concentration of Rac1/Rnd1 was 40 μM and the concentration of plexin-B1 was set 15 to 20 times higher than Rac1/Rnd1. The data were analyzed using Origin software giving the stoichiometry (N), binary equilibrium binding constant (K_a = 1/K_d which is equal to [GTPase. Plexin]/[GTPase] × [Plexin]) and enthalpy of binding (ΔH). The free energy of binding (ΔG) is determined from the well known relation, ΔG = −2.303 RT log K_a and the entropy of binding is determined using Gibbs-Helmholtz equation, ΔG = ΔH–TΔS. The experimental error in ΔH is less than 2% corresponding to error of 5 to 10% in K_a. The stoichiometric ratio N varied from 0.7 to 1.0 upon data fitting. For the pH study, the phosphate buffer was titrated over a pH range from 6.0 to 8.0. For temperature dependence, ITC was performed from 10 to 30°C. The change of the specific heat is determined from the plot of ΔH versus temperature and using equation, ΔC_p = [Δ(ΔH)/ΔT]. Plexin-B1 RBD was injected at volumes of 5 to 10 μL and at intervals of 3 to 4 min.

To determine the fraction of plexin-B1 monomer, titration experiments were carried out in the above buffer at different temperatures, 10, 20, and 30°C and the fraction of monomer, F_m, was calculated using: F_m = (ΔH_d–ΔH_i)/(ΔH_d–ΔH_m), where ΔH_d: heat effect due to dimer injection; ΔH_m: heat effect due to monomer; ΔH_i: heat effect at ith concentration.⁴² The disulphide stabilized dimeric form of the RBD was studied with GTPases in the same phosphate buffer without TCEP.

NMR spectroscopy and analysis

NMR experiments were carried out at 298 K on a Bruker Avance 800 MHz spectrometer equipped with cryoprobe. ¹⁵N, ²H-labeled, NMR samples were prepared for the RBD.⁴³ The NMR sample buffer consisted of 50 mM NaCl, 4 mM MgCl₂, 4 mM DTT in 50 mM phosphate buffer, pH 6.8, 90% H₂O/10% D₂O. ¹H-¹⁵N HSQC-TROSY spectra were recorded of mixtures of the plexin-B1 RBD and unlabeled Rac1 or Rnd1 protein. The total protein concentration varied from 0.03 to 0.75 mM. All data were processed with NMRPipe,⁴⁴ and peak intensities were read out using NMRDraw. Assignment of chemical shifts was reported previously for the monomeric, dimeric, and GTPase-bound RBD.^27,28,45