Crystal Structure of Anti-Hen Egg White Lysozyme Antibody (HyHEL-10) Fv-Antigen Complex

In order to address the recognition mechanism of the fragments of antibody variable regions, termed Fv, toward their target antigen, an x-ray crystal structure of an anti-hen egg white lysozyme antibody (HyHEL-10) Fv fragment complexed with its cognate antigen, hen egg white lysozyme (HEL), was solved at 2.3 Å. The overall structure of the complex is similar to that reported in a previous article dealing with the Fab fragment-HEL complex (PDB ID code, 3HFM). However, the areas of Fv covered by HEL upon complex formation increased by about 100 Å2 in comparison with the Fab-HEL complex, and two local structural differences were observed in the heavy chain of the variable region (VH). In addition, small but significant local structural changes were observed in the antigen, HEL. The x-ray data permitted the identification of two water molecules between the VH and HEL and six water molecules retained in the interface between the antigen and the light chain complementarity determining regions (CDRs) 2 and 3 (CDR-L2 and CDR-L3). These water molecules bridge the antigen-antibody interface through hydrogen bond formation in the VL-HEL interface. Eleven water molecules were found to complete the imperfect VH-VL interface, suggesting that solvent molecules mediate the stabilization of interaction between variable regions. These results suggest that the unfavorable effect of deletion of constant regions on the antigen-antibody interaction is compensated by an increase in favorable interactions, including structural changes in the antigen-antibody interface and solvent-mediated hydrogen bond formation upon complex formation, which may lead to a minimum decreased affinity of the antibody Fv fragment toward its antigen.

The development of protein engineering of an antibody (e.g. its preparation without immunization, molecular design, or grafting enzymatic activity) (1,2) has stimulated the use of antibodies in a variety of medical and industrial applications. Recent advances in genetic engineering have made it possible to prepare antibody fragments (Fab and Fv) 1 using a bacterial expression system (3)(4)(5). This system has demonstrated a number of successful insights into the utilization of antibody molecules as biomaterials.
It is known that antibodies recognize their target antigens with their variable domains, and fragments of variable regions can also bind to their targets specifically. However, antibodies have constant domains both in heavy and light chains. Fab molecules, which are the fragments of antibody molecules, are composed of two polypeptide chains (light and heavy), each of which folds into two domains (variable and constant). Fv molecules, which are heterodimers consisting of light and heavy chain variable domains (VL and VH) of antibodies, can exhibit antigen binding specificity and affinity similar to that of Fab fragments (3)(4)(5). It is believed that the binding properties of Fv and Fab are identical to those of IgG. For a precise description of the molecular mechanism of antibody-proteinaceous antigen interactions, x-ray crystallographic studies of antigen-bound antibody fragments should be promising (reviewed in Refs. 6 -10).
We have focused on the interactions between hen egg white lysozyme (HEL) and its monoclonal antibody, HyHEL-10, the structural features of which have been analyzed by x-ray crystallography (11) in the Fab-HEL complex. The bacterial expression system of the HyHEL-10 Fv fragment has been established (12), and the interaction between the antigen and Fv has been investigated intensively using titration calorimetry in combination with site-directed mutagenesis (13)(14)(15). We have also reported that the association between VH and VL of the Fv is significantly stabilized in the presence of antigen HEL, and these structural characteristics can be applied to the novel enzyme-linked immunosorbent assay method for detection of the antigen (16). We also applied these characteristics to the selection method for a mutant Fv with a desired property from the library of mutant Fvs (17). The accurate structural information of a bacterially expressed Fv fragment of HyHEL-10 complexed with hen egg white lysozyme is crucial for a more precise discussion of molecular-recognition mechanisms and structural features of the Fv.
Here we report the x-ray crystal structure of the antibody Fv fragment of HyHEL-10, complexed with its antigen HEL, with a resolution of 2.3 Å. The high resolution of the structure permits a detailed description of antibody-combining sites, including water molecules retained in the interface. On the basis of the crystal structure of the HyHEL-10 Fv-HEL complex reported here, the antigen-recognition mechanism of the Fv fragment will be discussed.

MATERIALS AND METHODS
Preparation and Crystallization of Wild-type HyHEL-10 Fv-Preparation of the Fv fragment of anti hen egg white lysozyme was carried out according to the published method (12). The purified Fv fragment was concentrated to 0.45 mM using an ultrafiltration membrane Centriprep-10 (Amicon, Tokyo, Japan) performed at 3000 ϫ g. The antigen HEL, which was purchased from Seikagaku-kogyo Inc. (Tokyo, Japan), was dissolved in water at a concentration of 0.54 mM (1.2 times the molarity of the Fv solution). The HyHEL-10 Fv fragment-HEL complex to be crystallized was obtained by mixing an equal volume of the solution of the Fv fragment with the HEL solution. A hanging drop vapor diffusion method (18) was applied for the crystallization procedure at 20°C. The droplet size was 10 l, comprising the same volume of the complex and reservoir solution, which was 0.7 ml. The first screening of the crystallization conditions was attempted according to the sparse matrix sampling method (19). The crystals appeared under several conditions and were refined by varying the pH of the buffers and the concentration of the precipitants. The best crystals of the Fv-HEL complex were grown from 100 mM Hepes buffer (pH 7.6 -7.8), 9 -11% (w/v) polyethylene glycol 6000, and 7-9% (w/v) 2-methyl-2,4-pentanediol. They were elongated, bipyramidal crystals, 0.5 ϫ 0.5 ϫ 1.5 mm in size.
Data Collection and Structural Determination-The radiation was focused on a detector by the use of a two-bended-mirror system. All crystallographic data were collected on a diffractometer DIP-2000 (MAC Science, Yokohama, Japan) with the imaging plate of 200 mm as a detector using CuK␣ radiation generated at 50 kV and 90 mA, and processed with the HKL programs DENZO (20), SCALA (21), and AGROVATA (22) in the CCP4 suite (23). The crystal belongs to the tetragonal space group P4 1 2 1 2 or P4 3 2 1 2 with unit-cell dimensions of a ϭ b ϭ 57.2 Å and c ϭ 237.0 Å. Because the crystal is estimated to contain one complex per asymmetric unit, the values for the V m (24) and the solvent content were calculated to be 2.4 Å 3 ⅐Da Ϫ1 and 49%, respectively. Although this crystal diffracted x-rays up to a 1.7-Å resolution limit, all diffraction data within 2.3 Å were collected with a crystal imaging plate distance of 120 mm to obtain separated spots, because of the large cell constant for the c axis. The intensity data consisted of 18,301 independent reflections from a total of 146,870 measurements (redundancy factor, 8.0). The crystallographic data and the statistics of the data collection are summarized in Table I.
The structure of HEL-Fv was determined by a molecular replacement method with the program AMoRe (25) in the CCP4 package. The model coordinates that correspond to the HEL and Fv fragment were derived from the complex structure of HEL and the Fab fragment of HyHEL-10 (11) (PDB ID code, 3HFM). A rotation search using reflections having a resolution between 8.0 and 4.0 Å showed two major solutions, and both of them had the same scores of 7.3. These two solutions were subjected to a translation search, and one of the solu-tions showed a lower R-factor (0.375) than the other (0.394); the former solution was used for further refinement. The choice of the ambiguous space group was confirmed to be P4 1 2 1 2 by comparison with the result of the translation search at the enantiomorph space group P4 3 2 1 2, which indicated a higher R-factor (0.485).
Crystallographic Refinement-Refinements of the structure of the HEL-Fv complex were carried out with the programs XPLOR (26) and REFMAC (27). Reflection data chosen randomly from 5.0% of the observed data were used for the calculation of the free R-factor (28), although they were not used for the refinement calculations. The R-and free R-factors were monitored as the indexes of the accuracy of the model. The structure of the solution by the molecular-replacement method was refined as a rigid body using the reflections having a resolution between 8.0 and 3.0 Å, showing an R-factor of 0.380 with a free R-factor of 0.398. This model was then refined with XPLOR by a simulated annealing method with an initial temperature of 3000 K. The intensity data were expanded stepwise from a resolution limit of 3.0 -2.3 Å in the refinements. After several cycles of refinement, the R-factor was reduced to 0.244 with a free R-factor of 0.313. At this stage, the model was corrected manually by inspection of A weighted 2 m͉Fobs͉-D͉Fcalc͉ and m͉Fobs͉-D͉Fcalc͉ D maps (29). The graphic program O (30) was used for the adjustment of the molecular model. After several iterations of positional refinements with REFMAC and correction by hand, water molecules were introduced gradually. The final model has 2861 nonhydrogen atoms, corresponding to 114 residues from Asp-1 to Ala-114 of the VH chain, 107 residues from Asp-1 to Lys-107 of the VL chain, 129 residues from Lys-1 to Leu-129 of HEL, and 125 water molecules. The final R-and free R-factors were 0.175 and 0.235, respectively. The atomic parameters of the HEL-Fv complex were deposited in the PDB (ID code 1c08). Fig. 1 shows a A weighted 2 m͉Fobs͉-D͉Fcalc͉ electron density map of a part of the CDR, imposed on the refined model. A water molecule (W1) forms hydrogen bonds with the side chain of both the VL chain and HEL. In this region, the water molecule hydrates to -OHs of Tyr-96 and Ser-91 of VL (numbering of Kabat et al. (31)), and -OH of Tyr-20, -NH1 of Arg-21, and -OH of Ser-100 of HEL.

Quality of the Structure-
The torsion angles of the main chain atoms were analyzed in a Ramachandran plot (32) produced by the program PRO-CHECK (33) (data not shown). Most residues fell into the most-favored and the additional allowed regions. One residue fell in the disallowed region: Ala-51 in VL (designated LAla-51). This was well placed in density with a clear choice for its side chain and the preceding and following carbonyl O atoms. As a consequence of the and angles at LAla-51, the conformation of CDR-L2 is a class 3 ␥-turn (34). This conformation is also found in CDR-L2 of the D44.1-lysozyme complex (35), the NC10-neuraminidase complex (36), and the NC41-neuraminidase complex (37). The standard deviation of the coordinates is estimated to be 0.25 Å by a Luzzati Plot (38) (result not shown). The average value of the B-factor is 28.3 Å 2 for the protein atoms, 36.6 Å 2 for the oxygen atoms of the water molecules, and 28.2 Å 2 for all atoms in the complex, which is comparable with that of the intensity data derived from the Wilson plot (39).
Overall Structure and Water Molecules in the HyHEL-10 Fv-HEL Complex-The packing of the Fv-HEL complex in the crystal lattice was found to be one complex in an asymmetric unit. Some residues exist that have intermolecular contact within 4.0 Å of the adjacent molecules. These interactions involve neither of the residues located in the CDRs of the Fv fragment, nor any of the epitope residues in HEL. Fig. 2 shows a schematic model of the HyHEL-10 Fv-HEL complex. An assignment of the secondary structure was made with the program PROCHECK (33) using the criteria of Kabsch (40). Both the VH chain and the VL chain making up the Fv fragment are superimposed on the Fv portion of the Fab fragment, which was determined in the HyHEL-10 Fab-HEL complex (11). The VL chain can be related to the VH chain by a rotation angle of 170.8°around the pseudodyad axis between these two chains, which is almost the same as the Fab-fragment angle (170.7°). The location of the 125 water molecules that were identified in the complex is shown in Fig. 3. These molecules are located not only on the surface of the complex but also in the interface between the subunits making up the complex; six water molecules occupy the area where the VL chain interacts with HEL, whereas two are identified in the region between VH and HEL.
Six water molecules participating in the interactions between HyHEL-10-VL and HEL were found in the interface of the complex, and some of them formed hydrogen bond networks (Fig. 4). Almost all of the contacts of CDR-L2 and L3 with HEL are water-mediated interactions, suggesting that the imperfect complementarity between VL and HEL is improved by these solvent molecules. Only a few water molecules, or in some cases none, are found in the VH-VL interface of Fab-antigen complexes (9). However, 11 water molecules were also observed in the VH-VL interface, indicating that solvent molecules mediate the stabilization of variable domains.
Antibody-Antigen Contact- Table II shows the list of noncovalent interactions that occur between VH and HEL and between VL and HEL, taking into account the water-mediated hydrogen bonds among them. The list is based on the maximum number of van der Waals contacts and on the hydrogen bond distances for the atoms involved in the contacts as used by Sheriff et al. (41). These include interactions between not only the main chain and side chain atoms, but also the main chain and main chain atoms.
Comparison of Fv Structure with Fab-Both VH and VL of the Fv fragment were superimposed onto one of the Fab fragments by a least squares method, and the root mean square (r.m.s.) differences between the C␣ atoms of the Fv structures and those of the Fab were calculated. The VH and VL chains of the Fv fragment gave an r.m.s. difference of 0.55 and 0.48 against the Fab fragment, respectively. These results indicate that the overall structure of the HyHEL-10 Fv-HEL complex is similar to that of the HyHEL-10 Fab-HEL complex (3HFM (11)) (Table III). However, superimposition of the C␣ coordinates of Fv HyHEL-10 VH onto those of Fab HyHEL-10 VH yielded an r.m.s. difference of 0.55 for VH, and aligning the C␣ coordinates of the VL in Fv with those of the VL in Fab yielded an r.m.s. difference of 1.03 for VH. This r.m.s. difference may originate from the flexibility of the two domains in Fv due to a lack of constant domains (Table III).
The interfacial area of the HyHEL-10 Fv-HEL complex was calculated to be 1800 Å 2 , which is larger than that of the HyHEL-10 Fab-HEL complex by about 200 Å 2 . The interfacial area differences between Fab and Fv are summarized in Fig. 5. Almost all changes in the buried areas are within 11 Å 2 in the VL chain. On the other hand, the buried areas of some residues increased (Ser-31, Tyr-53, Ser-54, and Trp-95 in the VH chain) or decreased (Asp-32 in the VH chain) by more than 10 Å 2 . Correspondingly, the increase in the buried areas of some residues in HEL (Asp-18, Arg-73, Asn-77, Gln-93, and Lys-97) is greater than 10 Å 2 , and the number of residues with a decreased buried area is only two (Tyr-20 and Asp-101). These results clearly indicate that the side chains of the Fv residues (VH chain, for the most part) make more contact with the antigen in comparison with the Fab residues.
Although no large difference occurred in VL, two major conformational differences in the main chains in VH were observed. These residues are positioned at 31 and 99, both of which are located in the CDR (Fig. 5).
One conformational difference in the main chain in VH is located in CDR-H1, from Asp-27 to Ser-31 (Fig. 6A). In the case of Fab-HEL, Thr-30 and Ser-31 form hydrogen bonds and van der Waals interactions with Arg-73 of HEL. However, only the hydroxyl group of Ser-31 forms a weak hydrogen bond with a main chain atom of Arg-73 in the Fv-HEL complex (Table II). The average B factor of Arg-73 in the Fv-HEL complex suggests that the conformation of Arg-73 fluctuates in the complex. Asp-32 of VH forms a salt bridge with Lys-97 of HEL in both the Fab-HEL complex and the Fv-HEL complex despite relatively weak interactions due to solvent exposure. The distance between O␦1 of Asp-32 and N of Lys-97 in the Fv-HEL complex (2.6 Å) is closer than the distance in the Fab-HEL complex (3.6 Å), suggesting that a stronger salt bridge formed in the Fv-HEL complex than in the Fab-HEL complex.
Another structural difference in the main chain is located in CDR-H3 from Asp-96 to Asp-101 (numbering of Kabat et al. (31)) (Fig. 6B). In the Fab-HEL complex, Asp-96 interacts by means of the van der Waals force with the residues of VL, which seem to contribute to the VH:VL contact. However, in the Fv-HEL complex, Asp-96 forms a salt bridge with Lys-97 of HEL (2.4 Å). The C␣ atom of Asp-96 moves by about 2 Å toward the HEL, perhaps due to the salt bridge formation. In the Fab-HEL complex, one water molecule appears at the position at which the side chain of Asp-96 is located in the Fv-HEL complex (11), bridging Lys-97 of HEL with the CO of Trp-95 in CDR-H3. Asp-101 of VH forms a hydrogen bond with one water molecule (W137), and the water forms another hydrogen bond with Asn-77 of HEL, bridging the imperfect complementarity of the interface. Substitution of Asp-101 in VH with Ala resulted in a 100-fold reduction in affinity (42), whereas the mutation into Thr results in only a small reduction in affinity. Lavoie et al. 42 state that the hydrophilic residue is required at site 101 of VH.
Comparison of Antibody-bound Antigen HEL Structure with Antibody-free One-A least squares fit of the C␣ atoms of HyHEL-10 Fv-bound HEL with those of free HEL refined in its tetragonal form at a resolution of 1.6 Å (43, 44) gave a root mean square difference of 0.43 Å. As in the case of Fab, no gross conformational changes were observed in HEL upon complex formation. As shown above, however, the covered areas in the Fv-HEL complex increased in comparison with those in the Fab-HEL complex, and local conformational changes of side chains were observed in the regions around sites 18 -22, 45-48, 70 -71, 75-77, 82-84, 100 -104, and 119 -122 in HEL. These results indicate that not only direct contact residues (18 -22, 75-77, and 100 -104 in HEL) but also noncontact ones (sites 45-48, 70 -71, 82-84, and 119 -122 in HEL) caused a structural perturbation upon complexation (data not shown). The structural movement of sites 47, 101, and 102 was also observed in the Fab-HEL complex and other antibody-HEL complexes (9,45). Some of the conformational changes observed in the direct contact regions are summarized below.
Asn-19-Gly-22-A comparison of the Fv-HEL structure with that of Fab-HEL indicates one of the structural differences in the main chains occurs in this region. The side chain of Asn-19 is exposed to the solvent in the Fab-HEL complex, whereas it rotates and makes favorable contacts with the antibody CDR-H1 in the Fv-HEL. One water molecule (W116) mediates the contact between Asn-19 and Gly-22 of HEL and Asn-92 and Ser-93 of VL through hydrogen bond formation, resulting in closer contact of these antigenic regions with HyHEL-10 Fv (Fig. 7A).
Trp-62-In the Fab-HEL complex, the rotation of an indole ring of Trp-62 was observed upon binding to an antibody (11). However, the indole ring of Trp-62 in the Fv-HEL complex is only slightly attracted to the aromatic ring of Tyr-53 in VH (Fig. 7B). We previously described the enthalpic contribution from the rotation upon binding by investigating the interaction between the W62G mutant HEL and Fv (13). Thus, the results obtained from the thermodynamic analysis might reflect the favorable aromatic ring interaction between Trp-62 of HEL and Tyr-53 of VH.
Arg-73-Leu-75-Asn-77-As discussed above, no antibody residue makes favorable contact with the side chain of Arg-73 of HEL in Fv-HEL (Fig. 6B). In the Fv-HEL complex, the side chain of Asn-77 forms a hydrogen bond with Asp-101 of VH using a water molecule (W137). On the other hand, there is no confor- mational change around this region in Fab-bound lysozymes (11).
Asp-101-Gly-102-Asn-103-The C␤ atom of Asp-101 of HEL rotates in the complex, perhaps due to the location of Tyr-53 and Ser-54 in CDR-H2. Recognition of the main chain of the Asp-101-Gly-102-Asn-103 loop by CDR-H2 may lead to the conformational change around the loop (Fig. 7B). This local induced fitting was also observed in Fab-HEL. However, in Fv-HEL, the loop makes closer contact than in Fab-HEL, which includes a hydrogen bond between the side chain hydroxyl group of Ser-54 and the side chain carboxyl group of Asp-101. We demonstrated an enthalpic contribution (i.e. a favorable enthalpy change is offset by an entropy loss) resulting from the structural perturbation upon binding by investigating the interaction between D101G mutant HEL and Fv (13). Kam-Morgan et al. (46) reported the effect of systematic mutation into Asp-101 on the affinity and concluded that the volume of the side chain correlates well with the decrease in the affinity. Thus, the structural perturbation of Asp-101 is needed for the high affinity of HyHEL-10 toward HEL, although its side chain can be removed.
These results clearly indicate that small but significant local structural changes (i.e. induced fitting upon complex formation) occurred in the antigen, HEL.

DISCUSSION
Comparison of HyHEL-10 Fv-HEL Structure with the Fab-HEL-The overall structure of the complex is similar to one reported previously (11), dealing with the Fab fragment-HEL complex (PDB ID code, 3HFM). However, the area of Fv covered by HEL in the Fv-HEL complex increased by about 100 Å 2 in comparison with that in the Fab-HEL complex. In addition, two local structural differences were observed in VH, and small but significant local structural changes were observed in the antigen HEL. These structural changes must contribute significantly to the HyHEL-10-HEL interaction. The association constant of Fv for HEL was shown to be lower than that of Fab (47,15), suggesting that constant domains make a significant contribution to the stabilization of the complex. The Fv fragment lacks CH1 and CL fragments, which seems to impart a greater flexibility to the antibody fragment. Indeed, the association constant for the VH-VL interaction is relatively low (10 5ϳ6 M Ϫ1 ), and thus, the Fv fragment easily dissociates into VH and VL. Asp-96 in VH might be exposed to the solvent due to the low VH-VL association, leading to salt bridge formation upon complexation. Thus, the binding mechanism of the Fv fragment might be different from that of the Fab fragment. As shown by the interactions between antibody 4 -4-20 fragments (Fab, Fv, and scFv) and their antigen (48), it would be proper to conclude that the lack of constant domains leads to a decrease in affinity and structural diversity in CDRs in the HyHEL-10-HEL system. VL-VH domains might be more deformable in the absence of the CL-CH1 and stabilized under a coexistent antigen. We recently proposed an immunosorbent assay based on the mechanism of the Fv stabilization (12).
Role of Water Molecules in HyHEL-10 Fv-HEL Interaction-As noted in the previous report, bound water can play a significant role in the specificity and affinity of antibodies binding to an antigen (10,49). For instance, some solvent molecules improve complementarity by acting as fillers in places where the fit between the paratope (direct antigenrecognizing region of an antibody) and the epitope (region directly recognized by an antibody) is not precise (50). Water molecules that mediate hen lysozyme-HyHEL-10 Fv interaction are summarized in Table IV. Hydrogen bond network formation mediated by solvent molecules in the HyHEL-10 Fv-HEL interface exists only in the VL-HEL interface. This suggests that shape complementarity between VH and HEL is achieved primarily with the aid of only two solvents, whereas shape complementarity between VL and HEL is not.
One of the major regions recognized by HyHEL-10 VH is one edge of ␣-helix (Lys-97) and the following loop region (Ser-100-Asp-101-Gly-102), where high flexibility is observed in the crystal structure of the antibody-free lysozyme. The flexibility might permit the high degree of shape complementarity in the VH-HEL interface. However, a site-directed mutagenesis study FIG. 5. Comparison of interfacial areas of Fv-HEL complex with Fab-HEL. The buried surfaces of the Fv-HEL complex (this study) and of the Fab-HEL complex (11) were calculated by the program AREAIMOL in CCP4 (22) with a probe of 1.4-Å radius. The differences between the buried areas of the Fv-HEL complex and those of the Fab-HEL complex are shown here for the VL and VH domains of the antibody and for the HEL antigen. a, VL; b, VH; c, HEL. (14,15,51) showed the contribution to the Fv-HEL interaction from the water-mediated hydrogen bond formation in the VH-HEL interface to be relatively low, and the shape complementarity of the VH-HEL interface was demonstrated to be critical for high Fv affinity (14,51).
On the other hand, six water molecules participating in the interactions between HyHEL-10-VL and HEL were found in the interface of the complex (Fig. 4). Almost all of the contacts of CDR-L2 and CDR-L3 with HEL are water-mediated interactions, suggesting that the imperfect complementarity between CDR-L2, CDR-L3, and HEL is improved by these solvent molecules. The epitope of CDR-L2 and CDR-L3 is relatively rigid (i.e. with a relatively low B factor in free HEL), and thus, perfect complementarity cannot be achieved without the aid of the water molecules. Some site-directed mutation into the residues that have contact with the buried water molecules (W1 and W2) resulted in a significant loss of affinity (51,52). 2 Only a few water molecules, or in some cases none, are in the VH-VL interface of Fab-antigen complexes (9). In addition, it has been reported that the packing density in the VH-VL interface appears to be almost identical to those in the interior of the protein (53). However, 11 water molecules were also observed in the VH-VL interface, indicating that the stabilization of variable domains is mediated by solvent molecules. Because free HyHEL-10 Fv has not yet been determined, it is not certain that these are also retained in free-Fv. The association constant for the HyHEL-10 VH-VL interaction is lower than anti-HEL antibody D1.3 (16), suggesting a low complementarity between variable domains of HyHEL-10. Thus, stabilization of the HyHEL-10 Fv fragment by a coexistent antigen may also be mediated by water molecules.
Correlation of Thermodynamics of Antigen Binding by Hy-HEL-10 Fv with the Structure of the Complex-As reported previously (13), the interaction between HEL and HyHEL-10 Fv is enthalpy driven, i.e. a large negative enthalpy change (-⌬H, 92 kJ mol Ϫ1 ) is in part compensated by an unfavorable entropy change (-T⌬S, Ϫ42 kJ mol Ϫ1 ). It has been suggested that the major noncovalent forces contributing to the favorable 2 K. Tsumoto and I. Kumagai, unpublished results. decrease in enthalpy are van der Waals interactions and hydrogen bonding (54,55). Hydrogen bonds in an antigen-antibody complex, including side chain-side chain, side chain-main chain, and main chain-main chain interactions, must contribute to a reduction in enthalpy (56). In addition, high resolution x-ray crystallography permitted the precise location of water molecules retained in the HyHEL-10 Fv-HEL interface. These connect the antigen with antibody side chains through hydrogen bond formation (Fig. 7A). The number of hydrogen bonds is relatively high in comparison with the total number of direct contact hydrogen bonds in the interface. Thus, it can be concluded that one of the major contributors to the favorable enthalpy change is the hydrogen bonding mediated by water retained in the interface, despite the fact that an unfavorable entropy change is caused by the water molecules. Bhat et al. (49) have concluded that water plays a major enthalpic role in the binding of an antibody to an antigen.
Entropy loss in the interaction between HyHEL-10 Fv and HEL would in part originate from the water molecules retained in the interface. Following the method of Spolar and Record (57), the entropy change can be defined as the sum of the hydrophobic effect (⌬Shydr), the decrease in rotational and translational degrees of freedom (⌬Srt), and other effects (⌬Sothers).
⌬S ϭ ⌬Shydr ϩ ⌬Srt ϩ ⌬Sothers (Eq. 1) The entropy and heat capacity changes for the interactions between HyHEL-10 Fv and HEL are estimated to be Ϫ137 J mol Ϫ1 K Ϫ1 and Ϫ1.40 kJ mol Ϫ1 K Ϫ1 , respectively (13). ⌬Shydr is calculated to be 444 J mol Ϫ1 K Ϫ1 at 303 K (57); ⌬Srt is generally considered to be a constant Ϫ209 J mol Ϫ1 K Ϫ1 (57); and thus, from Equation 1, ⌬Sothers is Ϫ372 J mol Ϫ1 K Ϫ1 . ⌬Sothers would include system-specific contributions such as local folding and water uptake in the interface. Dunitz (58) has proposed the entropy of water tightly bound to the surface of a protein to be between 0 and 28 J mol Ϫ1 K Ϫ1 . Eight and 11 water molecules are completely buried in the Fv-HEL and VH-VL interfaces, respectively, as described above. Assuming the entropy loss (⌬Swater) introduced by the uptake of water to be almost the same as the average value (14 J mol Ϫ1 K Ϫ1 ), ⌬Swater is estimated to be 266 J mol Ϫ1 K Ϫ1 (ϭ 14 ϫ 19), and thus, the entropy change (⌬Sconf) derived from local folding (i.e. conformational changes) is estimated to be 106 J mol Ϫ1 K Ϫ1 (⌬Sothers Ϫ ⌬Swater ϭ ⌬Sconf). The entropy loss accompanying conformational changes has been proposed to be 23.4 J mol Ϫ1 K Ϫ1 (57); division of ⌬Sconf by Ϫ23.4 yields the number of residues involved in the conformational changes.  From Equation 2, the number of residues was estimated to be 4.5. Small but significant local structural changes were observed in the antigen HEL; therefore, no major conformational changes should occur in HyHEL-10 Fv. It should be emphasized that every water molecule retained and every structural change may not always accompany an equal entropic loss upon complex formation. Nonetheless, it is likely that a significant part of entropy loss originates from water uptake in the interface (59 -61).
Because all CDRs of HyHEL-10 Fv in its complex form are canonical (62), and framework regions are almost the same as other Fv or Fab fragments, one may assume that the structure of free-Fv is similar to that of the Fv-HEL complex. In fact, a recent report comparing antigen-free D1.3 Fv with antigenbound Fv suggests that no major conformational changes occur despite the change in the relative orientation of each domain (63). In a recent study of the affinity-matured, antibody-low molecule hapten complex, little structural change was observed in the antibody upon binding (64). In fact, the interaction between a hapten and its affinity-matured antibody can be defined as a "key and keyhole" interaction. On the other hand, structural changes were observed in some hapten-antibody interactions (65), DNA-or peptide-antibody interactions (66,67), and the idiotope-anti-idiotope complex (68). A precise analysis of the structure of antigen-free Fv, including crystallographic structures and dynamic structures using NMR spectroscopy, is required for further discussion of the mechanism of antigen-antibody interactions.
Conclusion-A high resolution crystallographic analysis of the HyHEL-10 Fv-HEL complex indicates that its overall structure is similar to that of the Fab-HEL complex (11). However, the Fv area covered by HEL increased in comparison with that of the Fab. Two major structural differences were observed, which may originate from the flexibility of two domains in Fv due to missing constant regions. Water molecules were found in both antigen-antibody and VH-VL interfaces, which supplement the imperfect complementarity of the interfaces. Some of these water molecules form a hydrogen bond network between the antigen and HyHEL-10 Fv. A significant contribution of water molecules to a favorable enthalpy change and unfavorable entropy change is proposed for the interaction. These results suggest that the effect of deletion of constant regions is compensated for by the increase in favorable interactions upon complex formation and that VL-VH domains may be more deformable in the absence of the constant regions.