Structural and thermodynamic responses of mutations at a Ca2+ binding site engineered into human lysozyme.

Structural determinants of Ca2+ binding sites within proteins typically comprise several acidic residues in appropriate juxtaposition. Three residues (Ala-83, Gln-86, and Ala-92) in human lysozyme are characteristically mutated to Lys, Asp, and Asp, respectively, in natural Ca2+ binding lysozymes and alpha-lactalbumins. The effects of these mutations on the stability and Ca2+ binding properties of human lysozyme were investigated using calorimetry and were interpreted with crystal structures. The double mutant, in which Glu-86 and Ala-92 were replaced with Asp, clearly showed Ca2+ binding affinity, whereas neither point mutant showed Ca2+ affinity, indicating that both residues are essential. The further mutation of Ala-83 --> Lys did not affect the Ca2+ binding of the double mutant. The point mutations Ala-83 --> Lys and Glu-86 --> Asp did not affect the stability, whereas the mutation Ala-92 --> Asp was about 1.3 kcal/mol less stable. Structural analyses showed that both Asp-86 and Lys-83 were exposed to solvent. Side chains of Asp-86 and Asp-91 were rotated in opposite directions about chi1 angle, as if to reduce the electrostatic repulsion. The charged amino acids at the Ca2+ binding site did not significantly affect stability of the protein, possibly because of the local conformational change of the side chains.

Calcium binding proteins are known to take part in several important functions in biological systems (1). Ca 2ϩ binding sometimes accompanies conformational changes, which are considered to be responsible for biological functions such as signal transduction and the formation of macromolecular complexes. Ca 2ϩ binding sites usually consist of several acidic residues chelating to bound Ca 2ϩ in a pentagonal bipyramidal manner (2). The close locations of acidic residues in a Ca 2ϩ binding site should negatively affect protein stability due to charge repulsion between the acidic residues. How does the introduction of charged residues within a calcium binding site affect the stability and Ca 2ϩ binding behavior? There are some reports describing the effect of mutations within Ca 2ϩ binding sites on the stability of the protein (3,4). Our approach in this paper is to introduce the minimum perturbation by amino acid replacement and determine the effect of these mutations on the stability and the Ca 2ϩ binding function using calorimetry. Furthermore, these effects are interpreted in terms of the structural information from x-ray crystallography. For this purpose, we chose human lysozyme because lysozymes are known to be a suitable model for studying protein function and stability (3,(5)(6)(7). More than 90 sequences of chicken type lysozymes and ␣-lactalbumins are in the sequence data base. Among these lysozymes, several have been found to have Ca 2ϩ binding ability (8 -11). Three residues (Ala-83, Gln-86, and Ala-92) in human lysozyme are usually mutated to Lys, Asp, and Asp, respectively, in natural Ca 2ϩ binding lysozymes and ␣-lactalbumins as shown in Fig. 1. We have already found that only the double mutation (Gln-86 3 Asp and Ala-92 3 Asp) in human lysozyme resulted in the formation of a Ca 2ϩ binding site (5). The precise analyses of the stability (3,5,6) and the Ca 2ϩ affinity (7) using calorimetry have also been performed, and the high resolution structural data of the wild type (12) and Ca 2ϩ binding mutant (13) lysozymes from x-ray crystallography are available. Here we show the effect of the subsequent mutations (Gln-86 3 Asp, Ala-92 3 Asp, and Ala-83 3 Lys) on the Ca 2ϩ binding properties, conformational stabilities, and tertiary structures of these mutants. It was found that both aspartic acids (Asp-86 and Asp-86) are essential for Ca 2ϩ binding. The introduction of an aspartic acid at the Ala-92 position resulted in destabilization of lysozyme, whereas the introduction of an aspartic acid at the Gln-86 position or the introduction of positive charge the Ala-83 position did not affect the stability. The observed instability with the Ala-92 3 Asp mutation is proposed to result from the difference in the hydration effect.

EXPERIMENTAL PROCEDURES
Materials-Klenow fragment of DNA polymerase I and restriction enzymes were purchased from Boehringer Mannheim (Mannheim, Germany) and Takara Shuzo (Kyoto, Japan). T4 DNA ligase was from New England Biolabs, Inc. (Beverly, MA). Human lysozyme was purchased from Green Cross Corp. (Tokyo, Japan). ␣-Lactalbumin and Micrococcus lysodeikticus were from Sigma. Glycol chitin was kindly provided by Prof. T. Imoto. Other chemicals were from Wako Pure Chemical Industries Ltd. (Osaka, Japan). Enzyme reactions were carried out under the conditions recommended by the suppliers.
DNA-Oligonucleotides were synthesized using a model 380A DNA Synthesizer (Applied Biosystems, Foster City, CA) and purified by high pressure liquid chromatography on a TSK gel ODS-120T (Toyo-Soda, Tokyo, Japan).
Sequencing and Plasmid Construction-To confirm the mutations, a Takara M13 sequencing kit (Takara Shuzo, Japan) was used for sequencing by the dideoxy method (17). The mutant human lysozyme gene in pERI8811 was subsequently replaced by the XhoI-SmaI fragment of the mutated gene.
Enzymatic Analyses of the Lysozymes-The lytic activities of the native and the holo-Q86D/A92D lysozymes were determined by the lysis of M. lysodeikticus cells (0.2 mg/ml) according to the procedure of Kikuchi et al. (18). Activities of the lysozymes using glycol chitin as a substrate were measured in 0.1 M acetate buffer (pH 5.5) at 40°C as described previously (19).
Measurement of Thermal Stability of the Mutant Human Lysozyme-Thermal stability of the wild type and mutant human lysozymes were determined by differential scanning calorimetry using a DASM4 microcalorimeter (20, 21) as described previously (3). The scan rate was 1.0 K/min, and the lysozyme concentrations used were 1.5 mg/ml. Sample solutions were prepared by dissolving the lysozymes in 0.05 M sodium acetate buffer at pH 4.5 with and without 10 mM CaCl 2 . The pH of the sample solution was confirmed before and after each measurement. Calorimetric (⌬H cal ) and van't Hoff enthalpies (⌬H vH ) were calculated by the method of Kidokoro and Wada (22).
Measurement of Ca 2ϩ Affinity to the Mutant Human Lysozymes-The affinity of the Ca 2ϩ to mutant lysozymes was determined by using a Micro Cal OMEGA titration calorimeter (23). 2 mg/ml of the protein solution was prepared in 0.05 M sodium acetate buffer (pH 5.5), and 1.7 ml of this solution was injected to the cell of the calorimeter. The titration was performed with the same buffer containing 5 mM CaCl 2 at 30°C. The calorimetric enthalpies, the binding constants, and the number of bound Ca 2ϩ ions were calculated using the computer program ORIGIN (MicroCal Inc.).
X-ray Crystallography-Diffraction quality crystals of Q86A and A92D were grown from 30 mM phosphate buffer (pH 6.0) containing 20 mg/ml protein, 2.5 M NaCl, and 1.5 mM CaCl 2 . After 7-10 days, the crystals grew up 0.4 mm on an edge. The crystal of A83K/Q86D/A92D was obtained in the same buffer (30 mM phosphate containing 2.5 M NaCl and 1.5 mM CaCl 2 ). About 1 year later, a crystal having the size of 0.8 ϫ 0.2 mm of A83K/Q86D/A92D lysozyme was obtained. X-ray diffraction data collection and data processing were performed as described previously (13,24,25). The structural refinements were performed using the program TNT (26). The final refinement parameters are summarized in Table I. The accessible surface area calculation based on the crystal structure was performed by the program insight II (MSI) with a water radius of 1.4 Å.

Enzymatic Activities and Ca 2ϩ Affinities of the Mutant Lysozymes-
The lytic activities of the mutant lysozymes measured at pH 6.2 and 25°C are summarized in Table II. In the presence of Ca 2ϩ , the lytic activity of all mutant lysozymes is lower than that of the wild type. The lytic activity of the mutant lysozymes having Ca 2ϩ binding affinity (Q86D/A92D and A83K/Q86D/A92D) is approximately 70% of that of the wild type. In the absence of Ca 2ϩ (in the presence of 5 mM EDTA), the lytic activity of all mutant lysozymes is almost the same as that of the wild type. Because the net charge of the protein affects the lytic activity, the enzymatic activity against glycol chitin was investigated in the presence of 10 mM CaCl 2 and pH 5.5. It was found that both Q86D and A92D mutant lysozymes are approximately 80% as active as the wild type, whereas activity of the Ca 2ϩ binding mutants, Q86D/A92D and A83K/ Q86D/A92D, lysozymes were 140 and 120% larger than that of the wild type enzyme, respectively.
The numbers of bound Ca 2ϩ ions (n), association constants (K a ), and enthalpy change (⌬H a ) for the binding to the mutant human lysozymes are summarized in Table III. No Ca 2ϩ binding to the wild type, Q86A, and A92D proteins was detected under these conditions. On the other hand, it was found that one Ca 2ϩ ion bound to the Q86D/A92D and A83K/Q86D/A92D mutants with a binding constant (K a ) of 3.9 ϫ 10 5 M Ϫ1 and enthalpy change (⌬H a ) of 2.2 kcal/mol. These values are similar to those (K a ϭ 1.9 ϫ 10 5 M Ϫ1 , ⌬H a ϭ 1.6 kcal/mol) in Q86D/ A92D lysozyme reported previously (3,7).
Thermal Stabilities of the Mutant Human Lysozymes-Thermal stabilities of the wild type and the mutant human lysozymes (A83K/Q86D/A92D, Q86D, and A92D) were measured in the absence and presence of 10 mM CaCl 2 using differential scanning calorimetry. The pH 4.5 condition was used because a high degree of reversibility of the unfolding of the mutant human lysozyme (more than 90%) can be obtained. Ca 2ϩ binding can also be observed by the shift of denaturation temperature (T m ) under these conditions. For comparison, previous data on the thermal stability of the wild type, apo, and holo Q86D/A92D lysozymes (3) are also reported. The calorimetric enthalpies (⌬H cal ), van't Hoff enthalpies (⌬H vH ), and the heat capacity changes in the denaturation were obtained directly from the analysis of these curves (summarized in Table IV). The unfolding temperatures (T d ) of A83K/Q86D/A92D, Q86D, and A92D mutant lysozymes were determined to be 77.0, 80.3, and 77.0°C, respectively. The A83K/Q86D/A92D and A92D mutant lysozymes were about 3°C less stable than the wild type (T d ϭ 80.3°C) in the absence of Ca 2ϩ . In the presence of 10   (Table IV). These values are comparable with the values of the wild type and apo-Q86D/A92D lysozymes at the same temperature. In addition, the ⌬H cal values of the mutants in the presence of 10 mM CaCl 2 at their T d were found to be 145.1, 134.0, and 131.5 kcal/mol, respectively. The ratios of ⌬H cal /⌬H vH for every mutant was about 0.95, which is also similar to that of the wild type and Q86D/A92D lysozymes (3), indicating two state denaturation (22). The heat capacity changes (⌬C p ) in the denaturation of the mutants were also similar to that of the wild type and Q86D/A92D lysozyme reported previously (3,5).
Three-dimensional Structures of the Mutant Human Lysozymes-Both the Q86D and A92D mutants were crystallized in space group of P2 1 2 1 2 1 , the same as that of the wild type, apo-, and holo-Q86D/A92D lysozyme reported previously (13). Three-dimensional structures of the Q86D and A92D mutants were determined at 1.8 A and 1.9 A resolution by x-ray crystallography. The structures were refined to R-factor of 16.1 and 16.4%, respectively. The overall structures of Q86D and A92D were quite similar to that of the wild type and apo-and holo-Q86D/A92D lysozyme. The root mean square deviations be-tween the wild type and the Q86D and A92D mutants for the main chain C␣ atoms were 0.278 and 0.309 Å, respectively.
The A83K/Q86D/A92D mutant lysozyme was crystallized in space group P2 1 . The tertiary structure of A83K/Q86D/A92D was determined by the molecular replacement method with the program X-sight (MSI) using the structure of holo-Q86D/A92D lysozyme (13) as the initial search model. Four molecules (chains A-D) were found in the asymmetric unit. The structure of A83K/Q86D/A92D lysozyme was refined to an R-value of 17.2%. Each molecule in the asymmetric unit was similar to that of the wild type (root mean square deviation, Յ0.3 A) and holo-Q86D/A92D mutant (root mean square deviation, Յ0.4 A) lysozymes.
The structures in the vicinity of the residues at positions 86, 91, and 92 in Q86D, A92D, and A83K/Q86D/A92D mutant lysozymes are shown in Fig. 2. No Ca 2ϩ bound in this region was observed in the Q86D and A92D mutants. In the structure of the Q86D mutant lysozyme, the conformation of the side chain of Asp-86 is rotated around the 1 angle toward solvent, which is similar to that seen in the structure of apo-Q86D/ A92D lysozyme. The water molecules occupied the same position relative to the bound Ca 2ϩ . The water configuration is quite similar to that of the wild type (Fig. 2a). In the structure of A92D mutant lysozyme, there is no significant conformational change observed in the side chain of this region in comparison with that of the wild type. However, the network of water molecules in the structure of the A92D mutant is quite different from that of the wild type and is rather similar to the structure of apo-Q86D/A92D lysozyme (Fig. 2b). The water molecule seen in the structure of A92D mutant lysozyme is considered to be a sodium ion, because at least five interactions within a 3-Å distance are observed. The changes in the structure of the Q86D and A92D lysozymes are consistent with the results from isothermal titration calorimetry, which show that no Ca 2ϩ binds to the Q86D and A92D mutant lysozymes. In the structure of the A83K/Q86D/A92D mutant lysozyme, one Ca 2ϩ ion was found at the same position as that of the Q86D/A92D mutant lysozyme. Three side chain oxygens from Asp-86, Asp-91, and Asp-92 chelate to the bound Ca 2ϩ as seen in the structure of the Q86D/A92D mutant lysozyme (Fig. 2c). With the other two water molecules, as well as two main chain

Mutation of an Engineered Ca 2ϩ Binding Site 34313
carbonyl groups from Lys-83 and Asn-88, a total of seven oxygens are observed to chelate Ca 2ϩ . From the structural analysis of these mutants with the data from the previous analysis (13), the structural perturbations due to the mutations are found to be different as summarized in Fig. 3. If both aspartic acids are present, the side chains changed their conformation to chelate Ca 2ϩ . The presence of either aspartic acid reflected the conformation of the Ca 2ϩ binding site without Ca 2ϩ . The average thermal factors (B-factor) of the main chain atoms versus each residue for the wild type, Q86D, and A92D lysozymes are similar to those of the wild type including the mutation sites (data not shown). These findings are also similar to that of apo-and holo-Q86D/A92D as reported previously (13). DISCUSSION Several acidic residues accompanying some basic residues are characteristically associated to make a Ca 2ϩ binding site in Ca 2ϩ binding proteins. Acidic residues are found to directly chelate to Ca 2ϩ in a pentagonal bipyramidal manner (2,13), and some basic residues located in the vicinity of Ca 2ϩ binding sites appear to provide a counter ion charge to the chelating residues. Three chelating aspartic acids, Asp-86, Asp-91, and Asp-92 and the positively charged Lys-83 are conserved in Ca 2ϩ binding lysozymes and ␣-lactalbumins. Because Asp-91 already exists in wild type human lysozyme (Fig. 1), the effect of the three mutations, Gln-86 3 Asp, Ala-92 3 Asp, and Ala-83 3 Lys on the stability and affinity of Ca 2ϩ was investigated using calorimetry. In Table V, the effect of the mutations on the stability is summarized, in which ⌬⌬G at 80°C was calculated according to Becktel and Shellman (27) by assuming that ⌬S and ⌬C p values of Ca 2ϩ bound and unbound mutants are the same as those of the holo-Q86D/A92D and the wild type lysozymes, respectively, as reported previously (3). In the presence of 10 mM Ca 2ϩ , the mutant A83K/Q86D/A92D showed about 3.6 kcal/mol stabilization, which is similar to that of Q86D/A92D lysozyme. Other mutants, Q86D and A92D lysozymes, did not show any stabilization in the presence of Ca 2ϩ . According to Schellman (28,29), the existence of ligand binding sites, such as a Ca 2ϩ binding site, should improve the stability of the protein in the presence of Ca 2ϩ . Therefore, the addition of 10 mM CaCl 2 , which is about 100 times in excess of the protein concentration, should result in stabilization of the protein. In both Q86D and A92D mutant lysozymes, however, no stabilization was observed in the presence of Ca 2ϩ (unlike the mutants Q86D/A92D and A83K/Q86D/A92D). This indi-cates that both the Q86D and A92D mutants do not have Ca 2ϩ affinity, which is consistent with the results of the binding experiments as listed in Table III. This also indicates that both acidic residues Asp-86 and Asp-92 are essential for Ca 2ϩ binding in the mutant human lysozyme. In the absence of Ca 2ϩ , we can evaluate the effect of the mutation on the stability of lysozyme. The stability of Q86D lysozyme was almost the same as that of the wild type lysozyme, indicating no affect of the mutation from Gln-86 to Asp. The A92D lysozyme was about 1.3 kcal/mol less stable than the wild type, which is similar to the apo-Q86D/A92D and apo-A83K/Q86D/A92D lysozymes, indicating that the mutation Ala-83 to Lys did not affect the stability but the mutation Ala-92 to Asp resulted in 1.3 kcal/ mol destabilization.
To interpret the stability in terms of the structural information, high resolution structural data were obtained from x-ray crystallography. Although these aspartic side chains are located within 5 Å distance to chelate Ca 2ϩ (holo-Q86D/A92D and holo-A83K/Q86D/A92D lysozymes in Figs. 2c and 3), the introduction of aspartic acid at position 86 resulted in a shift of the side chain conformation toward the outside of the site as if to reduce the charge repulsion to Asp-91 (Q86D lysozyme in Fig. 3). The charged groups of Asp-86, Asp-92, and Lys-83 are now located 7.5, 6.6, and 10 Å apart from Asp-91 in the wild type lysozyme, the mutations Gln-86 3 Asp or Ala-83 3 Lys did not affect the stability. It indicates that a charged cluster formed when constructing a Ca 2ϩ binding site does not always decrease the protein stability. Only a small rotation of the side chain may be enough to reduce the repulsion between the charges because of the solvent shielding. It has been reported that an ion pair located on the surface of a protein does not always affect the stability of the protein (30). On the other hand, a destabilization was observed for the mutation Ala-92 3 Asp. One explanation for this destabilization is the difference in hydration between the Ala and Asp side chains upon unfolding. The tertiary structure of the mutant A92D indicated that the side chain of Asp-92 is almost buried and the side chain conformation was quite similar to those of the apo-and holo-Q86D/A92D lysozyme (Fig. 3). According to Oobatake and Ooi (31,32), the difference in hydration free energy between alanine and aspartate is calculated to be about 2.5 kcal/mol at 80°C. Because Asp-92 in the mutant and Ala-92 in the wild type structures are almost buried with the accessible surface areas of these amino acids calculated to be less than 3 Å 2 , the difference in hydration is considered to be the major part of the instability (1-2 kcal/mol in Table V) observed in the mutants having Asp-92. The smaller than expected effect on stability may be caused by the weak binding of Na ϩ ion to Asp-92 in the native state as seen in the crystal structures of the A92D and apo-Q86D/A92D lysozyme.  (3). b ⌬⌬G values were extrapolated to 80°C. The parameters, ⌬S ϭ 0.390 cal/mol and ⌬C p ϭ 1.54 cal/mol, for Ca 2ϩ bound mutants (holo-A83K/ Q86D/A92D in 10 mM CaCl 2 ) were assumed to be the same as those of the holo-Q86D/A92D-lysozyme (3), and the parameters, ⌬S ϭ 0.377 cal/mol and ⌬C p ϭ 0.97 cal/mol for Ca 2ϩ unbound mutants (Q86D, A92D, apo-Q86D/A92D and apo-A83K/Q86D/A92D) were assumed to be the same as those of the wild type lysozyme (3).