The Contribution of Conformational Adjustments and Long-range Electrostatic Forces to the CD2/CD58 Interaction*

The CD2 family of cell-surface glycoproteins are structurally related members of the Ig superfamily (1, 2). Expressed mainly on hematopoietic cells, they modulate immune responses by homotypic or heterotypic interactions with other members of the CD2 family. CD2 binds CD58 (in humans) or CD48 (in rodents), and these interactions enhance T cell recognition of antigen on antigen-presenting or target cells. This enhancement is thought to involve both adhesion and signaling mechanisms (1, 3–7).

CD2 and its interaction with its ligands have been studied intensively and have emerged as an important paradigm for understanding the molecular basis of cell-cell recognition (1, 8, 9). CD2 and its ligands have structurally related ectodomains comprised of two Ig domains, with the membrane-distal domains involved in ligand binding (1, 8). The interaction of human CD2 with CD58 is characterized by low affinity (K_d ∼10 μm at 37 °C), which is the result of a very fast dissociation rate constant (k_off > 4s^–1) (10). Structural studies of the individual proteins and site-directed mutagenesis locate the binding sites on the equivalent GFCC′C″ β-sheets of CD2 and CD58 and reveal them to be highly charged (11–13). Solution of the crystal structure of the complex between the human CD2 and CD58 ligand binding domains provides a detailed view of the binding interface (14). This is relatively small (buried surface area ∼1160 Ų) and has poor surface-shape complementarity, consistent with the low affinity (14). Comparison of the structure of the complex with the structure of unbound CD2 (15–17) and CD58 (12, 13) reveals significant differences, particularly in the case of CD2. The most prominent differences are in the C′C″ and FG loops of both molecules (14, 16). In addition NMR analysis shows that the CD58 binding site on unbound CD2 is highly flexible, with most of the movement occurring in the C′C″ and FG loops (16, 17). Taken together, these data suggest CD2 binding to CD58 is accompanied by conformational adjustment and stabilization of a flexible interface. Although these conformational changes provide an explanation for the low affinity of the CD2/CD58 interaction, they appear to be inconsistent with its relatively fast k_on (10).

To further investigate these putative conformational changes and the discrepancy between the structural and kinetic data, we undertook a detailed thermodynamic and kinetic analysis of the CD2/CD58 interaction. We show here that the interaction is enthalpically driven and accompanied by unfavorable entropic changes, consistent with stabilization of a flexible binding interface. We also show that, despite having a highly charged binding interface, long-range electrostatic interactions have no net effect on the CD2/CD58 interaction.

EXPERIMENTAL PROCEDURES

Proteins—Soluble forms of CD2 and CD58 were prepared and purified as described previously (18). These comprised the full ectodomains with C-terminal oligohistidine tags. The C termini of the encoded CD2 and CD58 were SCPEKHHHHHH and TCIPSSHHHHHH, respectively.

Surface Plasmon Resonance—These studies were performed on a BIAcore 2000 (BIAcore AB) (19). Unless otherwise stated, experiments were performed at 25 °C using HBS buffer (10 mm HEPES (pH 7.4), 150 mm NaCl, 1 mm CaCl₂, and 1 mm MgCl₂) at a flow rate of 10 μl·min^–1. Human CD2 was directly coupled to research grade CM5 sensor chips (BIAcore AB) using the amine coupling kit (BIAcore) as described previously (10). Kinetics measurements were performed at a flow rate of 50 μl·min^–1 and confirmed at three different immobilization levels of CD2 to rule out mass transport artifacts.

Affinity, kinetic, and thermodynamic properties were determined as described (20). Equilibrium thermodynamic parameters were obtained by measuring the affinity over a range of temperatures (5–37 °C) and fitting the nonlinear form of the van't Hoff equation to these data (21), ΔG =ΔH_T0 – TΔS + ΔC_p (T – T₀) – TΔC_pln(T/T₀), where T is the temperature (in degrees Kelvin), T₀ is an arbitrary reference temperature (e.g. 298.15 K), ΔG is the free energy of binding at the standard state (all components at 1 mol·liter^–1), ΔH_T0 is the enthalpy change at T₀ (kcal·mol^–1), ΔC_p is the heat capacity change (kcal·mol^–1K^–1) at constant pressure, and ΔS is the entropy change at the standard state.

ΔG was calculated from the affinity constant (K_D) using the equation ΔG = RTln(K_D/C), where R is 1.987 cal·mol^–1K^–1, K_D is expressed in mol/liter, and C is the standard state concentration (1 m).

The activation enthalpy of dissociation (Δ^‡H_d) was determined by measuring k_off over a range of temperature (10–30 °C) and plotting ln(k_off/T) against 1/T, the slope of which equals –Δ^‡H_d/R (20). Δ^‡H_a was calculated from the relationship Δ^‡H_a =Δ^‡H_d +ΔH.

In experiments varying the ionic strength a series of HBS stocks were prepared with the indicated NaCl or KF concentration. CD58 samples were diluted in the appropriate HBS buffer, and the same HBS buffer was used as the running buffer.

Isothermal Titration Calorimetry—ITC⁵ measurements were performed at 25 °C using a MicroCal VP-ITC unit (MicroCal Inc.) as described previously (22). Samples were dialyzed extensively into HBS. CD58 at 245 μm was injected in 10-μl aliquots into a cell containing 1345 μl of CD2 (at 35 μm), and the heat release was measured. The heat of dilution was obtained by injection of CD58 into HBS and subtracted prior to data analysis. The titration data were fitted by nonlinear curve fitting using Origin software, supplied with the instrument, to obtain the K_D, stoichiometry, and ΔH. TΔS was determined by the relationship TΔS =ΔH – RTln(K_D/C).

RESULTS AND DISCUSSION

Thermodynamics of Human CD2 Binding to CD58—To investigate whether CD2 binding to CD58 is accompanied by energetically significant conformational changes, we undertook a detailed thermodynamic analysis. The binding of CD58 to CD2 was analyzed by surface plasmon resonance. The affinity constant (K_D) was measured by equilibrium binding analysis (10). The binding free energy of an interaction (ΔG), which can be calculated from the affinity constant (see “Experimental Procedures”), is comprised of enthalpic (ΔH) and entropic (–TΔS) components (ΔG =ΔH – TΔS).

The relative contribution of enthalpic and entropic components can be determined by measuring the dependence of ΔG on temperature, a procedure termed van't Hoff analysis (see “Experimental Procedures”). For protein/protein interactions, ΔH and TΔS typically vary with temperature, and this variation is measured as the change in heat capacity (at constant pressure) or ΔC_p. The binding energy of the CD2/CD58 interaction was measured over the temperature range of 5 to 37 °C and ΔH, –TΔS, and ΔC_p were determined by fitting the nonlinear form of the van't Hoff equation to the data (Fig. 1A). From several different determinations the ΔH at 25 °C was determined to be –11.5 ± 0.2 kcal·mol^–1 (mean ± S.E., n = 3), which is highly favorable. The corresponding entropic component (–TΔS) was +4.4 ± 0.2 kcal·mol^–1, which is unfavorable (Table 1).

ΔC_p determined by van't Hoff analysis was –118 cal·mol^–1 K^–1 (Table 1). This falls at the high end of values typically reported for protein/protein interactions, which range from –1000 to 0 cal·mol^–1 K^–1, with an average value of approximately –300 cal·mol^–1 K^–1 (23). A negative ΔC_p, which is typical of protein/protein interactions, is thought to be the result of the tendency of water to form an ordered “shell” adjacent to nonpolar surfaces which “melts” at higher temperatures (24). Upon binding, the burial of nonpolar surfaces disrupts this shell, ejecting the water into free solution, with favorable entropic and unfavorable enthalpic effects. Increasing the temperature melts the shell so that these effects are gradually lost. The relatively high ΔC_p measured for the CD2/CD58 interaction is consistent with the fact that the binding interfaces are highly polar and that, as a result, a comparatively small amount of nonpolar surface is buried upon binding.

It is possible to estimate ΔC_p from structural data using the empirically determined relationship where ΔA_np is the change in the buried nonpolar surface area and ΔA_p is the change in the buried polar surface area. Using an implementation of the Lee and Richards algorithm developed by Hubbard (25–27), ΔA_np and ΔA_p were estimated from the crystal structure of the CD2/CD58 complex to be 660 and 690 Ų, respectively. Using these values, the calculated ΔC_p is –116 cal·mol^–1 K^–1, which is in good agreement with the experimental value obtained from van't Hoff analysis (Table 2).

Several studies have reported discrepancies between ΔH measured indirectly by van't Hoff analysis (ΔH_VH) and ΔH measured directly by calorimetry (ΔH_cal) (28–30). We therefore measured ΔH directly using ITC (Fig. 2). The ΔH and –TΔS thus measured were similar to those determined by van't Hoff analysis (Table 1). It has been suggested that differences between ΔH_cal and ΔH_VH indicate the presence of linked equilibria that contribute to ΔH_cal but not ΔH_VH. Recent studies dispute this, however, arguing that ΔH_cal should equal ΔH_VH and that differences are more likely to be the result of experimental artifacts (31–33).

When conformational changes are required for binding, this may result in a high activation enthalpy of association (Δ^‡H_a), which is a measure of the net number of bonds that need to be broken in order to form the transition state complex (34). Because the binding kinetics were too fast to measure the Δ^‡H_a directly, we determined the Δ^‡H_d and calculated the Δ^‡H_a from the relationship Δ^‡H_a =ΔH+TΔ^‡H_d. We determined Δ^‡H_d by measuring the k_off over a range of temperatures (10–30 °C) and plotting ln(k_off/T) against 1/T (Fig. 1B), the slope of which equals –Δ^‡H_d/R (20). Using this approach the Δ^‡H_a thus estimated was 5.3 ± 0.8 kcal·mol^–1, which is relatively small (35). Although this does not support conformational change, it does not rule it out either, because it is possible that new bonds, such as long-range electrostatic interactions, are formed in the transition state that compensate for the bonds that are broken (36). The preponderance of charged residues in the binding interface supports this possibility, leading us to test (see below) whether long-range electrostatic interactions accelerate binding.

Our results show that, at physiological temperatures, the CD2/CD58 interaction is enthalpically driven and accompanied by unfavorable entropic changes. This contrasts with most protein/protein interactions (23), which are accompanied by favorable entropic and enthalpic changes (Fig. 3). It is similar to what has been observed with protein/protein interactions that are known to be accompanied by conformational adjustments and a reduction in conformational flexibility, such T cell receptor/ligand (Fig. 3) and gp120/CD4 (37) interactions. However the unfavorable entropic change is relatively modest in comparison with these examples, and it is possible that it arises from sources other than changes in conformational flexibility. For example, it may arise purely from solvent effects, such as the trapping of water molecules in the binding interface. We therefore investigated the source of the unfavorable entropic change.

Entropic changes accompanying protein/protein interactions are the sum of favorable changes in solvent (water) entropy and unfavorable changes in protein entropy. Changes in solvent entropy arise mainly from the burial of nonpolar surfaces (i.e. the hydrophobic effect). This arises from the movement of water adjacent to hydrophobic surfaces, where it adopts an organized, shell-like structure, into free solution, where it is more disorganized. Changes in protein entropy arise from loss of rotational and translational freedom and conformational flexibility. To investigate the extent to which unfavorable entropic changes arise from changes in conformational flexibility, we used the approach suggested by Spolar and Record (38) to dissect total entropy change (ΔS) into three main components, where ΔS_HE is the entropy change associated with the hydrophobic effect, ΔS_RT is the change associated with loss of translational and rotational freedom of the interacting proteins, and ΔS_other is the change arising from other sources, including a reduction in conformational flexibility upon binding. ΔS_HE arises from the burial of nonpolar surface area; it can be calculated from structural data using the empirically determined relationship where A_np is the buried nonpolar surface area (in Ų). As noted previously, A_np is estimated to be 660 Ų for the CD2/CD58 interaction.

Based on empirical measurements Spolar and Record (38) estimated ΔS_RT to be –50 cal·mol^–1K^–1 for a 1:1 protein/protein interaction. Because for protein/protein interactions ΔS is temperature-dependent, there exists a temperature (T_s) where ΔS = 0. This can be calculated using the relationship where T is an arbitrary temperature (e.g. 298.15 K) and ΔS_T is the entropy change at that temperature. Using the ΔS_T and ΔC_p values determined above, T_s was calculated to be 261 K for the CD2/CD58 interaction. ΔS_HE at T_s was calculated using Equation 3 to be 83 cal·mol^–lK^–1.

ΔS_other at T_s can thus be calculated from the relationship If one assumes that ΔS_RT is temperature-independent over this temperature range, as suggested by Spolar and Record (38), then ΔS_other is calculated to be –33 cal·mol^–1 K^–1 (Table 2). This negative ΔS_other value indicates that binding is accompanied by a reduction in conformational entropy. Spolar and Record (38) show that the number of residues () undergoing the transition from a flexible to a folded or rigid state can be estimated from the empirical relationship suggesting that ∼6 or more residues are converted from a flexible to a stable state upon CD2 binding to CD58.

Recently it has been argued that a better estimate of ΔS_RT for protein/protein association would be the cratic entropy (39), which is –8 cal·mol^–1 K^–1 (Rln1/55.5) for a bimolecular interaction in water in the standard state. Substituting this value into Equation 5 would give –74.6 cal·mol^–1 K^–1 for ΔS_other and substituting this in Equation 6 gives a value of 13 for (Table 2). Thus dissection of the binding entropy suggests that CD2 binding to CD58 is accompanied by a reduction in conformational flexibility involving 6–13 residues. Although these changes might occur anywhere in the protein complex, structural studies support the notion that this reduction in conformational flexibility involves the BC, C′C″, and FG loops of CD2 (16, 17) and the C′C″ and FG loops of CD58 (14). As noted above, CD2 binding to CD58 is accompanied by and driven by a large, favorable enthalpy change. Such favorable enthalpy changes may arise from an increased number of contacts forming at the interface and/or from the trapping of solvent within the binding interface. The binding interface of the CD2/CD58 complex is ∼1150 Ų (14). Although this is at the low end of the range for protein/protein interactions (23, 41), there appear to be numerous contacts; 10 salt bridges and five hydrogen bonds were proposed based on analysis of the crystal structure (14). Site-directed mutagenesis studies of the CD2/CD58 interaction suggest that a centrally positioned CD2 tyrosine and surrounding charged residues make major contributions to the binding energy (42–45). Interestingly, the CD2/CD58 binding interface exhibits poor surface shape complementarity (14), suggesting that water molecules bridge the binding surfaces and raising the possibility that trapped water contributes to the favorable enthalpy changes.

Effect of Electrostatic Interactions on CD2 Binding to CD58—Proteins typically associate in solution with a rate constant of 10⁵-10⁶ m^–1s^–1 (36, 46). This rate may decrease if binding requires conformational adjustments (37, 47). Given the evidence from structural and thermodynamic studies that CD2 binding to CD58 is accompanied by conformational change, it is notable that the association rate constant of the CD2/CD58 interaction is not particularly slow (k_on ≥ 4 × 10⁵ m^–1s^–1) (10). The association may be accelerated, sometimes dramatically (48), by favorable long-range electrostatic interactions. The mechanism is believed to involve accelerated collisions and/or steering proteins into the correct orientation for binding (36, 46, 49). Given that the CD2/CD58 binding interface is highly charged (14), we investigated the contribution of long-range electrostatic interactions to binding by examining the effect of varying solution ionic strength.

We initially varied the ionic strength by changing the NaCl concentration. Increasing the NaCl concentration from 150 to 1500 mm resulted in an ∼6-fold decrease in affinity (Fig. 4A), consistent with a previous report (43). This effect was primarily a result of an effect on the k_on (Fig. 4B), consistent with screening of long-range electrostatic interactions. Recent studies have shown that electrostatic screening results in a linear relationship between lnk_on and 1/(1 + κa), where the latter is proportional to the ionic strength (36). In the case of the CD2/CD58 interaction, this relationship was clearly not linear (Fig. 4C), suggesting that the effect of NaCl was not solely the result of electrostatic screening. One possible explanation was that the sodium and/or chloride ions were disrupting the structure of CD2 and/or CD58 either by direct interactions with the proteins or through effects on solvent structure. Because such effects are dependent on the salt, we used a different salt (KF) to vary the ionic strength. In contrast to the effect of NaCl, the affinity and kinetics of CD2 binding to CD58 showed no significant change when the concentration of KF was varied from 150 to 1500 mm (Fig. 4). It follows that the effect of varying NaCl on binding is not the result of changes in ionic strength and must result from another mechanism. Interestingly, the rat CD2/CD48 interaction was unaffected by varying the NaCl concentration (50), implying that rat CD2 and/or CD48 are resistant to this effect. Given that the surface charge distributions on CD2/ligand binding surfaces are not conserved across species and that the ligand binding surfaces of the human CD2/ligand pair are more highly charged, it is not surprising that they exhibit differential sensitivity to NaCl (12, 14).

The absence of an effect when varying KF concentration demonstrates that long-range electrostatic interactions have no net effect on the CD2/CD58 interaction. This is somewhat unexpected given that the binding interfaces in the complex are highly charged with excellent charge complementarity (14). However, it remains possible that there are some favorable electrostatic interactions that are balanced by unfavorable electrostatic interactions. To investigate this further we used site-directed mutagenesis to disrupt a subset of electrostatic interactions. We had recently generated a number of human CD2 mutants in which single charged residues in the binding interface were mutated to alanine (18). Because it was not possible directly to measure k_on, we calculated it from the K_D and k_off (Table 3). Although mutation of CD2 residues Lys⁴¹, Lys⁵¹, and Lys⁹¹ to alanine had no significant effect on the k_on, mutation of Asp³¹ to alanine led to a 5-fold decrease in the k_on value from 4 × 10⁵ to 0.8 × 10⁵ m^–1s^–1 (Table 3). Interestingly, Asp³¹ forms a salt bridge with the CD58 residue Arg⁴⁴ in the bound complex (14). Taken together, these data suggest that a favorable electrostatic interaction involving Asp³¹ and Arg⁴⁴ accelerates CD2 binding to CD58 but that other, unfavorable interactions inhibit association. Interestingly, there is also evidence that the rat CD2/CD48 interaction is also characterized by balancing unfavorable and favorable electrostatic interactions; whereas this interaction is unaffected by changes in salt concentration, mutation to alanine of individual charged residues in the binding interface resulted in both decreases and increases in affinity (50).

Given the highly charged nature of the CD2/CD58 binding interface, our finding that long-range electrostatic interactions do not accelerate binding is unexpected and in striking contrast to the barnase/barstar interaction, where electrostatic attractions between highly charged surfaces accelerate binding by several orders of magnitude (48). One difference between the CD2/CD58 and barnase/barstar interfaces is that the latter is less heterogeneous; barnase is predominantly positively charged, and barstar is uniformly negatively charged. In contrast, the CD2 and CD58 binding surfaces each carry an intricate mixture of negative and positive charges (12, 14). We suggest that the presence of heterogeneous charges on each surface results in a complex electrostatic potential energy landscape with a balancing of favorable and unfavorable long-range electrostatic interactions. Given that it has no effect on binding kinetics, what is the functional significance of charged nature of the CD2/CD58 binding interface? We have shown previously, in an analysis of the rat CD2/CD48 interaction, that charged residues that contribute little to the binding affinity may nevertheless contribute to binding specificity by imposing a requirement for charge complementarity on the other surface (50). The reason proposed for their small contribution to affinity is that, in order for them to form favorable interactions (e.g. salt bridges) with the ligand, existing favorable electrostatic interactions (e.g. with water, salt, or adjacent residues) need to be broken. The molecular interactions that mediate transient cell/cell interactions need to be low affinity to facilitate detachment. We suggest that the CD2/CD58 binding interface is charged because electrostatic complementarity enables the interaction to be both weak and specific.

Conclusion—In this report we have shown that the CD2/CD58 interaction is accompanied by an unfavorable entropy change. Dissection of the entropy change using the approach of Spolar and Record (38) suggests that the unfavorable entropy change results in part from a reduction in conformational flexibility. Taken together with previous structural studies, these findings suggest that energetically significant structural rearrangements accompany binding. Despite the highly charged binding interface, we have shown that long-range electrostatic interactions have no net favorable effect on CD2 binding to CD58, probably because of the presence of balancing favorable and unfavorable electrostatic interactions.