The affinity maturation of anti-4-hydroxy-3-nitrophenylacetyl mouse monoclonal antibody. A calorimetric study of the antigen-antibody interaction.

To understand the mechanism of affinity maturation, we examined the antigen-antibody interactions between 4-hydroxy-3-nitrophenylacetyl (NP) caproic acid and the Fab fragments of three anti-NP antibodies, N1G9, 3B44, and 3B62, by isothermal titration calorimetry. The analyses have revealed that all of these interactions are mainly driven by negative changes in enthalpy. The enthalpy changes decreased linearly with temperature in the range of 25-45°C, producing negative changes in heat capacity. On the basis of the dependence of binding constants on the sodium chloride concentration, we have shown that, during the affinity maturation of the anti-NP antibody, the electrostatic effect does not significantly contribute to the increase in the binding affinity. We have found that, as the logarithm of the binding constants increases during the affinity maturation of the anti-NP antibody, the magnitudes of the corresponding enthalpy, heat capacity, and unitary entropy changes increase almost linearly. On the basis of this correlation, we have concluded that, during the affinity maturation of the anti-NP antibody, a better surface complementarity is attained in the specific complex in order to obtain a higher binding affinity.

To understand the mechanism of affinity maturation, we examined the antigen-antibody interactions between 4-hydroxy-3-nitrophenylacetyl (NP) caproic acid and the Fab fragments of three anti-NP antibodies, N1G9, 3B44, and 3B62, by isothermal titration calorimetry. The analyses have revealed that all of these interactions are mainly driven by negative changes in enthalpy. The enthalpy changes decreased linearly with temperature in the range of 25-45°C, producing negative changes in heat capacity. On the basis of the dependence of binding constants on the sodium chloride concentration, we have shown that, during the affinity maturation of the anti-NP antibody, the electrostatic effect does not significantly contribute to the increase in the binding affinity. We have found that, as the logarithm of the binding constants increases during the affinity maturation of the anti-NP antibody, the magnitudes of the corresponding enthalpy, heat capacity, and unitary entropy changes increase almost linearly. On the basis of this correlation, we have concluded that, during the affinity maturation of the anti-NP antibody, a better surface complementarity is attained in the specific complex in order to obtain a higher binding affinity.
A series of anti-NP mouse monoclonal IgG antibodies used in the present study were produced by the immune response of C57BL/6 mice against NP coupled to T cell-dependent carrier, chicken ␥-globulin (9, 10). The variable regions of the primary response anti-NP antibodies show low affinity for NP and carry few, if any, somatic mutations (2,4), whereas those of the secondary response anti-NP antibodies usually exhibit increased affinity for NP and are somatically mutated (5). The secondary response antibodies are divided into two groups by carrying or lacking a somatic Trp 3 Leu exchange at position 33 in the variable region of the heavy chain (11,12). In the present study we compare N1G9, a primary response anti-NP antibody, with 3B44 and 3B62, which are secondary response anti-NP antibodies with and without Trp 3 Leu exchange, respectively. Thermodynamic aspect of antigen-antibody association is essential in order to understand the mechanism of the high affinity and specificity of antigen-antibody interaction. Thermodynamic parameters such as Gibbs free energy change, ⌬G, enthalpy change, ⌬H, entropy change, ⌬S, and heat capacity change, ⌬Cp, can provide useful information to identify fundamental forces involved in the antigen-antibody interaction. For instance, the magnitude of ⌬Cp is usually related to the contribution of the hydrophobic effect to molecular association (13)(14)(15)(16).
In the present study we show the ITC analyses of the antigen-antibody associations in affinity maturation. We examined the interactions between 4-hydroxy-3-nitrophenylacetyl caproic acid (NP-Cap) antigen and the Fab fragments of three anti-NP antibodies, N1G9, 3B44, and 3B62. On the basis of the obtained thermodynamic data, we have found that the binding constants, K a , for NP-Cap correlate with each of the ⌬H, ⌬Cp, and ⌬Su values. As the logarithm of the K a values increases in the course of affinity maturation, the magnitudes of the corresponding ⌬H, ⌬Cp, and ⌬Su values increase almost linearly. Although the interactions between a series of monoclonal antibodies and their same antigen have been investigated in several cases (35-37, 39, 43-45), the linear relationship between log K a and each of ⌬H, ⌬Cp, and ⌬Su shown in the present study has not yet been observed. On the basis of this correlation of the thermodynamic data along with the previously reported nuclear magnetic resonance (NMR) data (46), we will discuss the mechanism of the affinity maturation.

Chemicals-4-Hydroxy-3-nitrophenylacetic acid was purchased from
Sigma. NP-Cap was synthesized from 4-hydroxy-3-nitrophenylacetic acid and ⑀-amino-n-caproic acid. All other chemicals were of reagent grade and used without further purification.
Preparation of Fab Fragments-Anti-NP mouse monoclonal IgG antibodies, N1G9, 3B44, and 3B62, were purified from C57BL/6 mice hybridoma cell lines kindly provided by Professor K. Rajewsky, as described previously (46). The Fab fragments of these antibodies were prepared by papain digestion according to the procedure described previously (47). For brevity, the Fab fragments derived from N1G9, 3B44, and 3B62 will be designated as Fab(N1G9), Fab(3B44), and Fab(3B62), respectively.
Concentration Determination-The concentration of NP-Cap was determined at 430 nm with use of the molar absorption coefficient, 4230 M Ϫ1 cm Ϫ1 . The concentration 1 mg/ml of the Fab solution is equivalent to the absorbance at 280 nm, 1.73 (N1G9 and 3B62), and 1.57 (3B44).
ITC-Isothermal titration experiments were carried out on a Microcal OMEGA or MCS calorimeter interfaced with a microcomputer (18). The Fab solution was prepared by extensive dialysis against the experimental buffer, and the antigen was dissolved in the same dialysis buffer. The antigen solution was injected 20 times in 5-l increments and 3-min intervals into the Fab solution. The heat for each injection was subtracted by the heat of dilution of the injectant, which was measured by injecting the antigen solution into the dialysis buffer. Each corrected heat was divided by the moles of NP-Cap injected and analyzed with Microcal Origin software supplied by the manufacturer. Fig. 1a shows a typical ITC profile for the interaction between NP-Cap and Fab(3B62) at 30.0°C. An exothermic heat pulse was observed after each injection of NP-Cap into Fab(3B62). The magnitude of each peak decreased gradually with each new injection, and a small exothermic peak was still observed at a molar ratio (NP-Cap)/(Fab(3B62)) ϭ 2. The area of this exothermic peak was equivalent to the heat of dilution measured in a separate experiment by injection of NP-Cap into the buffer solution. The area under each peak was integrated, and the heat of dilution of NP-Cap was subtracted from the integrated values. The corrected heat was divided by the moles of NP-Cap injected, and the resulting values were plotted as a function of the molar ratio (NP-Cap)/(Fab(3B62)), as shown in Fig. 1b. The resultant titration plot was well fitted to a sigmoidal curve by using a nonlinear least-squares method. The binding constant, K a , and the enthalpy change, ⌬H, were obtained from the fitted curve. Further, the Gibbs free energy change, ⌬G, and the entropy change, ⌬S, were calculated from the equation, ⌬G ϭ ϪRT lnK a ϭ ⌬H Ϫ T⌬S. The thermodynamic parameters obtained for Fab(3B62) are the following:

Determination of Thermodynamic Parameters-
The magnitudes of the ⌬S and ⌬G values are dependent on the concentration units for the standard state. In order to obtain unitary entropy change, ⌬Su, and unitary Gibbs free energy change, ⌬Gu, which are independent of the concentration units chosen for the standard state since, in essence, solute concentrations are measured in mole fraction units, we used the following equations (48): The cratic contribution to the entropy change, R lnX M , is R ln(1/55.6) ϭ Ϫ7.98 cal mol Ϫ1 K Ϫ1 , where 55.6 M is the concentration of water in dilute aqueous solution (13). In the case of Fig. 1 ⌬Su and ⌬Gu are Ϫ27.5 Ϯ 0.5 cal mol Ϫ1 K Ϫ1 and Ϫ12.0 Ϯ 0.1 kcal mol Ϫ1 , respectively. The thermodynamic parameters for the interaction between NP-Cap and each of Fab(N1G9) and Fab(3B44) were obtained in the same way.
Temperature Dependence of the Antigen-Antibody Interaction-As a function of temperature between 25 and 45°C, we analyzed the interaction between NP-Cap and each of Fab(N1G9), Fab(3B44), and Fab(3B62) by ITC. The thermodynamic parameters for these interactions are summarized in Table I. The thermodynamic parameters, ⌬Gu, ⌬H, and ϪT⌬Su are plotted as a function of temperature in Fig. 2. For all of these interactions, both the ⌬H and ⌬Su values are negative and exhibit strong temperature dependencies. By contrast, the negative ⌬Gu values show only a weak dependence on temperature. Since the temperature dependencies of ⌬H and ϪT⌬Su have the opposite signs, their contributions to the temperature dependence of ⌬Gu are almost canceled out. The enthalpy-entropy compensation has been observed previously for other antigen-antibody interactions (41,44,45).
The associations between NP-Cap and the three anti-NP Fab are mainly driven by favorable negative changes in ⌬H. The negative ⌬H values decrease with increasing temperature and show linear dependence on temperature in the range of 25-45°C (Fig. 2). The ⌬Cp value for each interaction can be determined from the slope of the temperature dependence of ⌬H. The negative ⌬Cp values of Ϫ309 Ϯ 27, Ϫ363 Ϯ 12, and Ϫ415 Ϯ 12 cal mol Ϫ1 K Ϫ1 are observed for Fab(N1G9), Fab(3B44), and Fab(3B62), respectively (Table I). Fig. 3 shows that the K a values for NP-Cap correlate with each of the ⌬H, ⌬Cp, and ⌬Su values. As the logarithm of the K a values increases in the order of Fab(N1G9), Fab(3B44), and Fab(3B62), the magnitudes of the corresponding ⌬H, ⌬Cp, and ⌬Su values increase almost linearly.
Ionic Strength Dependence of the Antigen-Antibody Interaction-Existence of positively charged amino acid residues was suggested in the combining site of anti-NP Fab fragments (49). In order to analyze the electrostatic effect, the interaction between NP-Cap and each of Fab(N1G9), Fab(3B44), and Fab(3B62) was investigated by ITC as a function of the sodium

DISCUSSION
In the present study we have carried out the ITC analyses of the antigen-antibody associations in affinity maturation, in order to understand the mechanism of affinity maturation. The present results reveal that all of the associations between NP-Cap and the three Fab are mainly driven by favorable negative changes in ⌬H. Van der Waals interactions and hydrogen bondings are usually considered to be the major potential sources of the negative ⌬H values (40,43,50). Thus, we suggest that van der Waals interactions and hydrogen bondings play a fundamental role in the interactions between NP-Cap and the three Fab. Also, the increase in the magnitude of the negative ⌬H with the increase in log K a (Fig. 3a) suggests that, in the course of the affinity maturation, the increase in the van der Waals interactions and hydrogen bondings promotes the increase in the binding affinity of the anti-NP antibody.
The negative ⌬Cp values in Table I are within the range of Ϫ100 to Ϫ650 cal mol Ϫ1 K Ϫ1 , which were previously reported for various antigen-antibody associations (28,29,35,37,38,40,41,(43)(44)(45). In general, the negative ⌬Cp values for protein folding and protein-ligand association are proportional to the reduction in water-accessible nonpolar surface areas of the molecules, and related to the contribution of hydrophobic effect to molecular association (13)(14)(15)(16)51). In order to interpret our data quantitatively, the empirical method of Sturtevant (52) was used to estimate the hydrophobic and intramolecular vibrational contribution to ⌬Cp (Table II). For all the three Fab, the calculated hydrophobic contribution to ⌬Cp is larger than the calculated vibrational contribution. Therefore, we suggest that the observed negative change in ⌬Cp may primarily result from the hydrophobic effect, that is, the decrease in solvent exposure of both the aromatic antigen and the nonpolar groups in the binding site of the three Fab caused by the antigenantibody association. Furthermore, the increase in the magnitude of the negative ⌬Cp with the increase in log K a (Fig. 3b) suggests that, in the course of the affinity maturation, the increase in the hydrophobic effect contributes to the increase in the binding affinity of the anti-NP antibody. This is consistent with the previous NMR result that the binding site of NP-Cap is located in a similar position, but the combining site of Fab(3B62) with higher affinity for NP-Cap is composed of more Tyr residues than that of Fab(N1G9) with lower affinity for NP-Cap (46).
The hydrophobic effect, which drives the association of nonpolar surfaces of molecules by excluding water from the interface, would contribute to the positive change in ⌬Su. However, we observed the unfavorable negative ⌬Su values in the range of 25-45°C, as shown in Table I. Negative ⌬Su has been observed previously for other antigen-antibody interactions (29 -32, 35, 36, 39 -44). Consequently, we conclude that some other factors should counteract the hydrophobic effect and make larger contributions to the negative ⌬Su. Such effect to the negative ⌬Su can be produced by the following factors: 1) the constraint of intramolecular vibrational flexibility of Fab due to the antigen binding (52); 2) the reduction in the translational and overall rotational degrees of freedom upon the complex formation (53,54); and 3) the conformational freezing of the amino acid residues of Fab caused by the antigen binding (55). The previously reported estimate of the factor 2) was almost constant for different antigen-antibody complexes (T⌬S TR ϭ 7-11 kcal mol Ϫ1 , where ⌬S TR is an amount of translational and overall rotational entropy change) (53)(54)(55). The estimation of the factor 3 was previously reported in the interactions between lysozyme and a few anti-lysozyme monoclonal antibodies (55). We applied the empirical method of Sturtevant (52) in order to estimate the hydrophobic and intramolecular vibrational (described above as the factor 1) contribution to ⌬Su (Table II). For all three Fab, the sign of the calculated hydrophobic contribution is positive, but that of the calculated vibrational contribution is indeed negative.
The obtained K a values shown in Table I increase in the order of Fab(N1G9), Fab(3B44), and Fab(3B62), which is consistent with the previously reported results (2,4,5). As the logarithm of the sodium chloride concentration increases, the logarithm of the K a values of the three Fab for NP-Cap decrease linearly. The dependence of the K a values on the sodium chloride concentration is similar for the three Fab (Fig. 4). These results suggest that the electrostatic effect is involved in the antigen-antibody associations, but the proportion of the elec- trostatic effect to the NP-Cap binding is similar for the three Fab. We conclude that, in the course of the affinity maturation, the electrostatic effect does not significantly contribute to the increase in the binding affinity of the anti-NP antibody.
We have found a linear correlation between log K a and each of ⌬H, ⌬Cp, and ⌬Su, as shown in Fig. 3. As the logarithm of the K a values increases in the course of affinity maturation, the magnitudes of the corresponding ⌬H, ⌬Cp, and ⌬Su values increase almost linearly. Although the interactions between a series of monoclonal antibodies and their same antigen have been investigated in several cases (35-37, 39, 43-45), the linear relationship between log K a and each of ⌬H, ⌬Cp, and ⌬Su shown in the present study has not been observed yet.
This linear relation of log K a , ⌬H, and ⌬Cp (Fig. 3, a and b) FIG. 2. Thermodynamic parameters (⌬H (E), ⌬Gu (छ), and ؊T⌬Su (Ç)) for the associations between NP-Cap and each of Fab(N1G9), Fab(3B44), and Fab(3B62) as a function of temperature. All experiments were performed in 5 mM sodium phosphate buffer, 200 mM sodium chloride, pH 8.0, at the experimental temperature. The solid lines were obtained from a linear least-squares regression. ⌬Cp was calculated from the slope of the regression line of ⌬H (E).  implies that the surface complementarity of the anti-NP antibody with the antigen increases in the course of the affinity maturation, which may reflect increased van der Waals interactions and hydrogen bondings between specific functional groups, and increased hydrophobic interactions with more exclusion of water molecules from the interface. However, the constancy of the electrostatic effect in the course of the affinity maturation (Fig. 4) suggests that the number of electrostatic interactions involved in the interface remains unchanged with the increase in the surface complementarity. On the other hand, we observed more pronounced effect of unfavorable negative ⌬Su in the course of the affinity maturation (Fig. 3c). This apparently contradicting effect seems feasible, because more enhanced surface complementarity would make more restraint in the intramolecular vibrations of the complex.
We conclude that, in the course of the affinity maturation of the anti-NP antibody, a better surface complementarity is attained in the specific complex in order to obtain a higher binding affinity. The attainment of a better surface complementarity may be produced by an increase in the number of Tyr side chains in the antibody combining site.