Thermal stability of Escherichia coli ribonuclease HI and its active site mutants in the presence and absence of the Mg2+ ion. Proposal of a novel catalytic role for Glu48.

Escherichia coli ribonuclease HI, which requires divalent cations (Mg2+ or Mn2+) for activity, was thermostabilized by 2.6-3.0 kcal/mol in the presence of the Mg2+, Mn2+, or Ca2+ ion, probably because the negative charge repulsion around the active site was canceled upon the binding of these metal ions. The dissociation constants were determined to be 0.71 mM for Mg2+, 0.035 mM for Mn2+, and 0.16 mM for Ca2+. Likewise, various active site mutants at Asp10, Glu48, Asp70, or Asp134 were thermostabilized by 0.4-3.0 kcal/mol in the presence of the Mg2+ ion, suggesting that this ion binds to these mutant proteins as well. The dissociation constants of Mg2+ were determined to be 9.8 mM for D10N, 1.1 mM for E48Q, 18.8 mM for D70N, and 1.8 mM for D134N. Thus, the mutation of Asp10 or Asp70 to Asn considerably impairs the Mg2+ binding, whereas the mutation of Glu48 to Gln or Asp134 to Asn does not. Comparison of the thermal stability of the mutant proteins with that of the wild-type protein in the absence of the Mg2+ ion suggests that the negative charge repulsion between Asp10 and Asp70 is responsible for the binding of the metal cofactor. Glu48 may be required to anchor a water molecule, which functions as a general acid.

Ribonuclease HI from Escherichia coli endonucleolytically hydrolyzes the RNA strand of a DNA/RNA hybrid at the P-O3Ј bond in the presence of an Mg 2ϩ ion, which is the most preferable metal cofactor for the enzyme and can be replaced only by an Mn 2ϩ ion (Berkower et al., 1973). The optimum pH for the enzymatic activity is around pH 8 -9. The enzyme is structurally homologous to the RNase H domain of human immunodeficiency virus 1 (HIV-1) 1 reverse transcriptase (Hostomsky et al., 1993). A similar folding topology has been observed in the structures of other enzymes as well, such as the RuvC Holliday junction resolvase (Ariyoshi et al., 1994), the catalytic domains of the integrases from HIV-1 (Dyda et al., 1994), avian sarcoma virus (Bujacz et al., 1996), and the bacteriophage Mu transposase core (Rice and Mizuuchi, 1995), which are functionally unrelated and have poor sequence similarity. A characteristic common to all of these enzymes is that they require an Mg 2ϩ or Mn 2ϩ ion for activity and that a cluster of the acidic amino acid residues, typically 3-4 in number, provides the metal ion binding sites. However, none of the catalytic mechanisms of these enzymes is fully understood.
For the catalytic mechanism of E. coli RNase HI (EC 3.1.26.4), two alternative mechanisms have been proposed. They are a general acid-base mechanism , Oda et al., 1993, Katayanagi et al., 1993b, Kashiwagi et al., 1996, which had previously been proposed as a carboxylate-hydroxyl relay mechanism, and a two-metal ion mechanism (Yang et al., 1990). These mechanisms differ from each other in the number of metal ions required for activity and in the activation mechanism of the hydroxyl ion, which attacks the phosphate group for the RNA cleavage. According to the general acid-base mechanism, one metal ion, instead of two, is required for activity, and the attacking hydroxyl ion is activated by an amino acid residue, instead of the metal ion. This discrepancy in the catalytic mechanism mainly results from the observations of different numbers of enzyme-bound metal ions in the crystallographic studies. The RNase H domain of HIV-1 contains two metal ions (Davies et al., 1991), whereas those of E. coli RNase HI (Katayanagi et al., 1993b) and the catalytic domain of avian sarcoma virus integrase (Bujacz et al., 1996) contain only one metal ion. Although further studies will be required to determine the catalytic mechanism of the enzyme, the data accumulated thus far suggest that the general acidbase mechanism is more likely than the two-metal ion mechanism. For example, crystallographic (Katayanagi et al., 1993b), NMR , and kinetic (Huang & Cowan, 1994) studies suggest that only one metal ion binds to the substratefree enzyme. A kinetic study using inert transition-metal complexes suggests that the RNA hydrolysis does not proceed through a nucleophilic attack by a hydroxyl ion activated by the metal cofactor, but rather through the stabilization of a transient intermediate of an outer sphere complex with the metal cofactor (Jou & Cowan, 1991). A kinetic study using a synthetic substrate modified at the cleavage site also suggests that a metal ion interacts with the 2Ј-hydroxyl group of the RNA, instead of the phosphate group, to form an outer sphere complex (Uchiyama et al., 1994). Recently, bacteriophage T4 RNase H was shown to have two metal ion binding sites (Mueser et al., 1996). However, this protein is structurally and functionally greatly diverged from E. coli RNase HI.
Site-directed mutagenesis experiments revealed that Asp 10 , Glu 48 , Asp 70 , His 124 , and Asp 134 are involved in the catalytic function (Kanaya et al., 1990a;Oda et al., 1993;Haruki et al., 1994). The mutation of Asp 10 to Asn or Ala, Glu 48 to Gln or Ala, Asp 70 to Asn or Ala, or Asp 134 to Ala almost completely abolishes the enzymatic activity. In contrast, the mutation of Asp 134 to Asn does not seriously affect the enzymatic activity. The mutation of Asp 10 to Glu, Glu 48 to Asp, or Asp 70 to Glu does not completely abolish the enzymatic activity, but dramatically reduces it. Structural determinations of the mutant proteins with the Asp 10 3 Asn, Glu 48 3 Gln, or Asp 70 3 Asn mutation suggested that the formation of hydrogen bond networks freezes the catalytic residue or prevents the binding of the Mg 2ϩ ion and thereby inactivates the enzyme (Katayanagi et al., 1993a). It was later shown that such hydrogen bond networks are not formed in the Asp 134 to Asn mutant (Kashiwagi et al., 1996). The mutation of His 124 to other residues dramatically reduces the catalytic activity without seriously affecting the substrate binding (Oda et al., 1993). Based on these observations, as well as the result that Asp 10 and Glu 48 are involved in the metal binding (Katayanagi et al., 1993b), a carboxylatehydroxyl relay mechanism was proposed , Oda et al., 1993, Katayanagi et al., 1993b. In this mechanism, two water molecules act as a general acid and base. Asp 70 functions as a proton acceptor and facilitates the attack of the water molecule, which acts as a general base, on the phosphodiester substrate. His 124 functions as a proton pump and enhances the catalytic efficiency by removing a proton from Asp 70 (Oda et al., 1993). However, the recent determination of the crystal structures of the mutant proteins, D134H, D134N, and D134A, in which Asp 134 is replaced by His, Asn, and Ala, respectively, allowed us to revise this mechanism so that His 124 , instead of Asp 70 , acts as the proton acceptor (Kashiwagi et al., 1996). The mutant proteins D134H and D134N retain high enzymatic activity (60 -90% of the wild-type enzyme), whereas D134A is almost fully inactive (Haruki et al., 1994). Interestingly, the overall structures of these mutant proteins are identical to that of the wild-type protein, and the positions of the ␦-polar atoms at residue 134 in the wild-type, D134H, and D134N proteins coincide well with one another. It is therefore likely that this ␦-polar atom is required to hold the attacking water molecule. The basis for the proposal of this new catalytic mechanism is the location of the imidazole group of His 124 . It is located near the ␦-polar atom at residue 134 so that they share the attacking water molecule. The mutations of His 124 dramatically reduce the catalytic activity, but do not abolish it, because Asp 134 can act as a proton acceptor and can substitute for His 124 (Kashiwagi et al., 1996).
Like the active site residues of E. coli RNase HI, charged residues are often clustered in the active sites of enzymes. Therefore, it seems likely that electrostatic strain caused by the unfavorable interactions causes a local instability in the active sites of the enzymes. In fact, many mutations that alleviate such electrostatic repulsion have been shown to increase the conformational stabilities of proteins such as staphylococcal nuclease (Hibler et al., 1987), pig citrate synthase (Zhi et al., 1991), cellular retinoic acid-binding protein (Zhang et al., 1992), barnase (Meiering et al., 1992), and T4 lysozyme (Shoichet et al., 1995). In addition, a good correlation between an increase in the conformational stability and a decrease in the enzymatic activity has been observed for these active site mutants. These results suggest that the local instability caused by the electrostatic strain makes the active site conformationally flexible, and thereby makes the enzyme functional. This hypothesis is supported by observations that the active site mutants of retinoic acid-binding protein are more thermostable than the wild-type protein in the absence of ligand (retinoic acid), but are as thermostable as the protein-ligand complex (Zhang et al., 1992), and that some of the crystal structures of the active site mutants of T4 lysozyme are similar to that of the enzyme in an enzyme-product complex (Shoichet et al., 1995).
NMR determination of the pK a values of all of the carboxyl groups in E. coli RNase HI, in the absence and presence of the Mg 2ϩ ion, indicated that there are strong electrostatic interactions among the four carboxylates in the active site, especially between Asp 10 and Asp 70 , which can be canceled when the Mg 2ϩ ion binds to the protein (Oda et al., 1994). Therefore, it would be informative to analyze the conformational stability of this protein and its active site mutants in the presence and absence of the metal cofactor at an alkaline pH at which the enzyme is functional. If these proteins were stabilized in the presence of the metal cofactors, it would be possible to determine the number of metal ions bound to the protein and the dissociation constant by the method that was used to analyze the binding of anions or cations to RNase T 1 (Pace and Grimsley, 1988). Since the role of each active site residue in the catalytic function of the enzyme is not fully understood, such studies would provide important information.
We report that binding of the metal ion or elimination of the carboxyl ion at residue 10, 70, or 134 by mutation, or both, increased the thermal stability of E. coli RNase HI by up to ϳ3 kcal/mol at pH 9, probably due to the cancellation of the negative charge repulsion around the active site. We also report that, in contrast to the corresponding mutations of the other three acidic active site residues, the mutation of Glu 48 to Gln does not seriously affect either the protein stability or the binding of the Mg 2ϩ ion. Nevertheless, this mutation almost fully inactivates the enzyme. We therefore propose an alternative catalytic mechanism for the enzyme in which Glu 48 anchors a water molecule that functions as a general acid.
Mutations-Two plasmids, pJAL10H and pJAL10S, used for the overproduction of the mutant proteins D10H and D10S, in which Asp 10 is replaced by His and Ser, were constructed by replacing the rnhA gene in the plasmid pJAL600 with the mutated gene, in which the codon for Asp 10 was altered from GAT to CAT (His) or TCT (Ser). Since the plasmid pJAL600 has a unique BglII site in the sequence encompassing amino acid residues 6 -8 and a unique SalI site about 50 base pairs downstream of the termination codon of the rnhA gene, these mutated rnhA genes were constructed by polymerase chain reaction, as described previously , by using BglII site-containing 5Ј-mutagenic primers and a SalI site-containing 3Ј-primer. These plasmids were confirmed by DNA sequence determination (Sanger et al., 1977) and were used to transform E. coli HB101 to construct overproducing strains. Cells were grown in Luria Bertani medium (Miller, 1972) containing 100 mg/liter ampicillin. Overproduction and purification of the mutant proteins D10H and D10S were carried out as described previously .
Enzymatic Activity-The RNase H activity was determined at 30°C, in 10 mM Tris-HCl (pH 8.0) containing 10 mM MgCl 2 , 50 mM NaCl, 1 mM 2-mercaptoethanol, and 10 g/ml bovine serum albumin by measuring the radioactivity of the acid-soluble digestion product from the 3 Hlabeled M13 DNA/RNA hybrid .
Protein Concentration-The protein concentrations were determined by UV absorption, using the absorption coefficient A 280 0.1% of 2.0 (Kanaya et al., 1990b).
Thermal Denaturation-Thermal denaturation curves and the temperature of the midpoint of the transition (T m ) were determined as described previously (Kimura et al., 1992) by monitoring the change in the CD value at 220 nm, using a J-600 spectropolarimeter from Japan Spectroscopic Co. Proteins were dissolved in 10 mM glycine HCl buffer (pH 3.0) or in 50 mM glycine NaOH buffer (pH 9.0) containing 20% glycerin and 2.8 M urea. The difference in the free energy change for unfolding between the mutant and wild-type proteins (⌬⌬G) was calculated as ⌬G(mutant) Ϫ ⌬G(wild-type). Since the ⌬G(wild-type) value is zero at the T m of the wild-type protein, the ⌬⌬G value at this temperature (⌬⌬G m ) is equal to the ⌬G m (mutant) value, which is calculated by . ⌬G m , ⌬H m , and ⌬S m are the free energy, enthalpy, and entropy changes of unfolding at the T m of the wild-type protein, respectively, and ⌬C p is the change in the heat capacity at constant pressure. The ⌬H m and ⌬C p values of 89.3 kcal/mol at 47°C and 1.4 kcal/mol/K, which were previously determined for the wild-type protein (Yamasaki et al., 1995), were used in this analysis.
Binding Analysis of the Metal Ions-The thermal denaturation curves and the parameters characterizing the thermal denaturation of the proteins were determined at pH 9.0, as mentioned above, in the presence of various concentrations of the salts. On the assumption that the unfolding equilibrium of the protein in the presence of ligand follows a two-state mechanism represented by N ⅐ L i D ϩ ⌬n ⅐ L, where ⌬n represents the difference in the number of metal ions (L) bound to the protein in the native (N) and denatured (D) states, the ⌬n value was calculated by the equation: ⌬n ϭ Ϫ d⌬G m /d(ln a L ) ϫ 1/RT (Record et al., 1978;Pace & Grimsley, 1988), where ⌬G m is the free energy change for unfolding at the T m of the protein in the absence of the metal ion and a L is the activity of the metal ion (Lide, 1994). Since the relationship between ⌬⌬G, a L , and the dissociation constant K d is given by ⌬⌬G ϭ ⌬G(L) Ϫ ⌬G(L ϭ 0) ϭ (⌬n) ⅐ RT ⅐ ln(1 ϩ [a L ]/K d ), according to Schellman (1975), the K d values were calculated from curve fitting of the ⌬⌬G versus a L data on the basis of a least square analysis.

Stabilization by Cation and Anion
Binding-In order to analyze the interaction between E. coli RNase HI and various ions by monitoring the changes in the protein stability, the thermal denaturation curves of the protein were measured at pH 9.0 in the presence of 2.8 M urea, 20% glycerol, and various concentrations of MgCl 2 , MnCl 2 , CaCl 2 , or NaCl. In these salts, the Cl Ϫ ion is the common anion. The Mg 2ϩ and Mn 2ϩ ions are the metal cofactors for the enzyme. The Ca 2ϩ and Na ϩ ions represent divalent and monovalent cations that cannot substitute for the Mg 2ϩ ion as a metal cofactor. The thermal unfolding of the protein is reversible under these conditions. It is noted that the protein precipitated at pH 9.0 upon thermal denaturation in the absence of 20% glycerol and 2.8 M urea. Therefore, these additives are required to prevent the aggregation of the protein upon thermal denaturation.
The thermal denaturation curves of the protein in the presence of various concentrations of MgCl 2 (Fig. 1A) and NaCl (Fig. 1B) indicate that the thermal stability of the protein was dependent on the salt concentration and was increased as the salt concentration was increased. Similar results were obtained for CaCl 2 and MnCl 2 as well (data not shown). The plots of the T m versus the salt concentration (Fig. 2) show that the effects on the protein stability greatly varied for the different salts.
However, they also show that the protein stability did not increase beyond a constant level, at which the protein is stabilized by approximately 10°C in T m or 3 kcal/mol in ⌬G m , in the presence of low concentrations of these salts except for NaCl. The protein stability was not considerably increased in the presence of a low concentration of NaCl but was increased beyond this level in the presence of a high concentration of NaCl. These results strongly suggest that the mechanism by which the protein is stabilized in the presence of MgCl 2 , MnCl 2 , or CaCl 2 is different from that in the presence of NaCl and that the protein stability increased in the presence of these salts due to the binding of the divalent cations to the active site of the protein.
The additivity of the stabilization effects of MgCl 2 and NaCl was analyzed by measuring the thermal denaturation curve of the protein in the presence of both of these salts. The protein was stabilized by either 8.6 or 16.6°C in T m in the presence of either 50 mM MgCl 2 or 2.5 M NaCl, respectively. If the stabilization effect of NaCl resulted from the binding of the Na ϩ ion to the Mg 2ϩ binding site of the protein, no additivity would be observed. However, the protein stability in the presence of both of these salts was much higher than that in the presence of either one of them (Fig. 3). The stabilization effect of the Mg 2ϩ ion in the presence of 2.5 M NaCl (5.8°C in T m ) was slightly lower than, but comparable with, that in the absence of NaCl (8.6°C in T m ). These results strongly suggest that the stabilization of the protein in the presence of the high concentration of NaCl is not primarily due to the binding of the Na ϩ ion to the Mg 2ϩ binding site but mainly results from the weak binding of the chloride ion to the protein and/or an increase in the internal hydrophobic interactions of the protein. It seems unlikely that there is an additional cation binding site in the protein because E. coli RNase HI is a basic protein with a pI value of 9.0 (Kanaya et al., 1989) and is rich in basic amino acid residues.
The number of metal ions bound to the protein (⌬n) was calculated from the least-squares line of the plot of ⌬⌬G versus ln a L (activity of the metal ion) data (Fig. 4) as 0.84 Ϯ 0.04 for Mg 2ϩ , 1.10 Ϯ 0.10 for Mn 2ϩ , and 0.95 Ϯ 0.32 for Ca 2ϩ . These results support the previous observation (Katayanagi et al., 1993b;Oda et al., 1991) that only one metal ion binds to the active site of the substrate-free enzyme. The dissociation constants were also calculated by using the equation of Schellman (1975) by assuming the ⌬n value as 1.0 so that the calculated data fit the experimental data well, as shown in Fig. 5. They were 0.71 mM for Mg 2ϩ , 0.035 mM for Mn 2ϩ , and 0.16 mM for Ca 2ϩ in the units of the ion activities of these salts. Thus, among these divalent cations, the Mn 2ϩ ion binds most strongly to the protein.
Thermal Stabilities of the Mutant Proteins-The pK a values of the carboxyl groups in the active site of E. coli RNase HI have been reported to be 6.1 for Asp 10 , 4.4 for Glu 48 , 2.6 for Asp 70 , and 4.1 for Asp 134 (Oda et al., 1994). Therefore, the repulsive forces among these carboxyl groups must be the strongest at pH values above 6.1, whereas they must be weakened at pH values lower than 6.1 and almost completely abolished at a pH below 2.6. To determine whether the acidic active site residues contribute equally to the negative charge repulsion in the active site at alkaline pHs, the thermal stabilities of 13 active site mutants (five for Asp 10 , three for Glu 48 and Asp 70 , and two for Asp 134 ) were analyzed at both pH 3.0 and 9.0 in the absence of the metal cofactor. It has previously been reported that the thermal unfolding of the protein is fully reversible at pH 3.0 in the absence of denaturants . If the mutation increased the protein stability at pH 9.0, without seriously affecting it at pH 3.0, the stabilization of the protein would mainly result from the elimination of the electrostatic repulsion around the active site. Alternatively, if the mutation equally increased the protein stability at either pH, the stabilization of the protein would occur for other reasons. The results are summarized in Table I. All the Asp 10 mutant proteins, except for D10N, were more stable than the wild-type protein at both pH 9.0 and 3.0. However, the increment in the protein stability at pH 9.0 (1.96 -4.05 kcal/mol in ⌬G) was much larger than that at pH 3.0 (0.44 -2.42 kcal/mol in ⌬G) for these mutant proteins, except for D10E, in which the number of carboxyl groups in the active site is unchanged. Thus, the ionization of Asp 10 was shown to be responsible for the negative charge repulsion around the active site. Likewise, the ionizations of Asp 70 and Asp 134 were each shown to be responsible for the negative charge repulsion around the active site. In contrast, all the Glu 48 mutant proteins were nearly as stable as the wild-type protein, at both pH 3.0 and 9.0, indicating that the ionization of this residue is not responsible for any negative charge repulsion around the active site.
Binding  (Table I), indicating that the Mg 2ϩ ion also binds to these mutant proteins. It is noted that all of the mutant proteins, except for D10A, D134N, and D134A, were nearly as stable as the wild-type protein under these conditions. As typical examples, in Fig. 6 the thermal denaturation curves of the mutant proteins, D10N, E48Q, D70N, and D134N, are compared with that of the wild-type protein in the absence or presence of the Mg 2ϩ ion. These results suggest that the negative charge repulsion in the active site contributes to the destabilization of the protein by approximately 3 kcal/mol in ⌬G, and can be canceled by the binding of the metal cofactor, the elimination of the carboxyl groups in the active site by mutation, or by both. Both Asp 134 mutant proteins were more stable than the wild-type protein even in the presence of the Mg 2ϩ ion, probably because this residue is affected by long-range, nonspecific negative charge repulsion, rather than the specific repulsion in the active site. In the vicinity of Asp 134 , Glu 131 and Glu 135 are clustered in addition to other active site residues. The thermal stabilities of the mutant proteins D10N, E48Q, D70N, and D134N were analyzed in the presence of various concentrations of Mg 2ϩ ion to determine the Mg 2ϩ ion dissociation constants. These were calculated by employing the ⌬n value of 1.0 as 9.8 mM for D10N, 1.07 mM for E48Q, 18.8 mM for D70N, and 1.82 mM for D134N in the units of the ion activity of Ϫ T m(no salt) , were determined from thermal denaturation curves, such as those shown in Fig. 1, in the presence of MnCl 2 (Ⅺ), CaCl 2 (Ç), MgCl 2 (q), or NaCl (E). The data are commonly expressed in panels A-C, which differ from one another only in the abscissa scale (salt concentration). The plot for CaCl 2 was removed from panel A, and the plots for MnCl 2 and CaCl 2 were removed from panel C, for clarification.
FIG. 4. Plot of ⌬⌬G as a function of ln a L . The number of metal ions (⌬n) bound to the protein in the native state was calculated by the gradient of the least squares fitting lines. ⌬⌬G is the difference in the free energy change for unfolding at the T m of the protein in the absence and presence of the metal ion at pH 9.0, and a L is the mean ion activity of the salt. Symbols are the same as in Fig. 2.

FIG. 5. Plot of ⌬⌬G as a function of the concentration of added salt.
The ⌬⌬G values were calculated from the data in Fig. 2, as described under "Experimental Procedures." Symbols are the same as in Fig. 2. The solid lines represent the best fit to the experimental data, calculated according to the equation given by Schellman (1975), by using dissociation constants of 0.035 mM for Mn 2ϩ , 0.16 mM for Ca 2ϩ , and 0.71 mM for Mg 2ϩ and the ⌬n value of 1.0. Panels A and B differ with each other only in the abscissa scale.
the Mg 2ϩ ion. For each mutant protein, the plots of ⌬⌬G versus the concentration of the Mg 2ϩ ion, in which the ⌬⌬G values were calculated using these K d values, fit well with the experimental data (Fig. 7). These results suggest that the binding of the Mg 2ϩ ion to the protein is not seriously affected by the elimination of the negative charge at either position 48 or 134 but is considerably impaired by the elimination of the negative charge at either position 10 or 70 (most seriously at position 70). Conservation of the metal binding site in the mutant proteins, in which the negative charge is eliminated at position 48, is further supported by the observation that the wild-type, E48Q, and E48A proteins were almost equally stabilized by 8.6 -10.3°C in T m by the addition of 0.2 M MgCl 2 .
Enzymatic Activity-The mutant proteins D10H and D10S were constructed for these experiments. Like the enzymatic activities of the mutant proteins D10N and D10A, the enzymatic activities of these mutant proteins were below the background level of 0.1%, which may reflect the RNase H activities of the wild-type enzyme and/or other enzymes that were purified together with the mutant proteins. To examine whether the loss of the enzymatic activity of the active site mutants reflects weaker binding of the Mg 2ϩ ion, the enzymatic activities of the mutant proteins D10N, E48Q, and D70N were ana-lyzed in the presence of concentrations of Mg 2ϩ ion greater than 10 mM. However, very little enzymatic activity was observed for these mutant proteins, even when the concentration of MgCl 2 was increased to 100 mM. Since the elimination of the negative charge at either position 10 or 70 considerably weakens the binding of the Mg 2ϩ ion to the protein, it seems likely that the Mg 2ϩ ion cannot bind to the precise site, at which it is functional as a metal cofactor, in the mutant proteins D10N and D70N. Therefore, the negative charges at positions 10 and 70 are probably required to direct the Mg 2ϩ ion to the functional position. In contrast, the negative charge at position 48 is probably required for the catalytic function because the elimination of this negative charge does not seriously affect the binding of the Mg 2ϩ ion to the protein.

TABLE I Parameters characterizing the thermal unfolding of E. coli RNase Hl and its active site mutants
Thermal denaturation curves of the proteins were determined in 50 mM glycine NaOH buffer (pH 9.0) containing 2.8 M urea and 20% glycerol, either in the absence or presence of 0.2 M MgCl 2 as well as in 10 mM glycine HCl buffer (pH 3.0) in the absence of MgCl 2 . T m is the temperature of the midpoint of the thermal denaturation transition. ⌬T m is the change of the T m of the mutant proteins relative to that of the wild-type protein. ⌬⌬G m is the change of the free energy change of unfolding of the mutant proteins relative to that of the wild-type protein at the T m of the wild-type protein, which was calculated as described under "Experimental Procedures." ⌬H m is the enthalpy change of unfolding at T m , which was calculated by van't Hoff analysis. ⌬T m (Mg) and ⌬⌬G m (Mg) are the changes of the T m and ⌬G m of the proteins, which were determined at pH 9.0 in the presence of MgCl 2 relative to those determined at pH 9.0 in the absence of MgCl 2 . Errors, which represent the 67% confidence limits, are within Ϯ0.8°C in T m . Protein pH 9.0 pH 9.0 (ϩ0.  6. Thermal denaturation curves of the wild-type and mutant proteins. Thermal denaturation curves were determined at pH 9.0 in the absence (A) and presence (B) of 0.2 M MgCl 2 , as described in the legend for Fig. 1. E, wild-type; ϩ, D10N; q, E48Q; Ç, D70N; å, D134N.
FIG. 7. Plot of ⌬⌬G as a function of the concentration of added MgCl 2 . The ⌬⌬G values were calculated from the ⌬T m values, which were determined from the thermal denaturation curves such as those shown in Fig. 6, as described under "Experimental Procedures." Symbols are the same as in Fig. 6. The lines represent the best fit to the experimental data calculated according to the equation given by Schellman (1975) by using the ⌬n value of 1.0 and Mg 2ϩ dissociation constants of 9.8 mM for D10N, 1.07 mM for E48Q, 18.8 mM for D70N, and 1.82 mM for D134N. The ⌬⌬G value is the difference between the ⌬G value of the wild-type or the mutant protein either in the presence or absence of the Mg 2ϩ ion and the ⌬G value of the wild-type protein in the absence of the Mg 2ϩ ion.

DISCUSSION
Stability-Activity Relationship-Amino acid residues that are involved in catalytic function or substrate binding are not necessarily optimized for stability. In fact, in a number of enzymes that have been examined, catalytic residue mutations that almost fully abolish the enzymatic activity frequently increase the protein stability, probably due to the elimination of unfavorable electrostatic interactions. Examples of such mutations are Glu 11 to Phe, Met, and Ala and Asp 20 to Asn, Thr, and Ser for T4 lysozyme (Shoichet et al., 1995); Lys 27 to Ala and His 102 to Ala for barnase (Meiering et al., 1992); and His 274 to Gly and Arg and Asp 375 to Gly, Asn, and Gln for pig citrate synthase (Zhi et al., 1991). When E. coli RNase HI was examined for this type of a stability-activity relationship by sitedirected mutagenesis experiments, among the mutations at the four acidic active site residues, those at Asp 10 , Asp 70 , and Asp 134 , which almost fully inactivate the enzyme, increase the protein stability by 1.08 -4.05 kcal/mol in ⌬G at pH 9.0 in the absence of the Mg 2ϩ ion. In contrast, the mutation of Glu 48 to Gln or Ala, which almost fully inactivates the enzyme, does not seriously affect the protein stability. Since these mutant proteins with different types of amino acid substitutions could not be distinguished from each other in terms of the protein stability, it seems unlikely that the increase in the stability, due to the elimination of the carboxyl group at this position, is fully canceled by the decrease in the stability, due to the introduction of a given residue. In fact, crystallographic studies have previously shown that the conformation of the protein is not seriously changed by the mutation of Glu 48 to Gln (Katayanagi et al., 1993a). Furthermore, the observation that the carboxyl group of Glu 48 has the normal pK a value of 4.4 (Oda et al., 1994), suggests that this carboxylate is not affected either by specific electrostatic interactions in the active site or by longrange nonspecific electrostatic interactions.
Metal Binding Site-It has been reported that, unlike the Ca 2ϩ binding site, the Mg 2ϩ binding site is partially buried in a shallow cleft of the protein molecule and uses solvent oxygens to coordinate one hemisphere of the bound metal ion (Needham et al., 1993). This site, therefore, can easily vary its size to accommodate various sizes of metal ions and, thereby, exhibits less selectivity for metal ions. In the current study, we showed that Mg 2ϩ , Ca 2ϩ , and Mn 2ϩ ions bind to E. coli RNase HI with dissociation constants of 0.7 mM, 0.16 mM, and 0.035 mM, respectively. This indicates that the Mg 2ϩ ion binds most weakly and the Mn 2ϩ ion binds most tightly to the protein. These differences reflect the differences in the concentrations of these ions required for optimal enzymatic activity. The optimal concentrations of these ions for enzymatic activity were previously reported to be 2 to 4 mM for Mg 2ϩ and Ͻ0.2 mM for Mn 2ϩ (Berkower et al., 1973). However, the enzymatic activity of E. coli RNase HI in the presence of 0.2 mM MnCl 2 is much less (7.5%) than that in the presence of 4 mM MgCl 2 (Berkower et al., 1973). In addition, little enzymatic activity is observed in the presence of 0.1-10 mM CaCl 2.
2 Thus, the metal binding site of E. coli RNase HI is highly selective with respect to the enzymatic function, although it does not provide a constrained ion cavity. It remains to be determined why the enzyme prefers Mg 2ϩ for activity.
Conformational Change of Glu 48 upon Mg 2ϩ Binding-The determination of the crystal structure of E. coli RNase HI complexed with the Mg 2ϩ ion previously showed that the carboxyl group of Glu 48 directly coordinates with the Mg 2ϩ ion (Katayanagi et al., 1993b). The conformations of the active site residues in E. coli RNase HI complexed with the Mg 2ϩ ion are shown superimposed upon those in the Mg 2ϩ -free enzyme in Fig. 8. Upon the binding of the Mg 2ϩ ion, the side chain of Glu 48 moves by 1.87 Å so that the distance between the carboxyl group and the Mg 2ϩ ion is reduced to 2.4 Å. In contrast, the distances between the other carboxyl groups and the Mg 2ϩ ion remain almost constant. The distances from the Mg 2ϩ ion are 2.1 Å to Asp 10 , 4.4 Å to Asp 70 , and 5.4 Å to Asp 134 . Based on these results, it has been proposed that Glu 48 is not involved in the catalytic function but is responsible for the Mg 2ϩ binding. However, the present studies of the thermal stabilities of the E. coli Rnase HI active site mutants do not support this proposal because they suggest that the binding of the Mg 2ϩ ion is not seriously affected by the elimination of the carboxyl group at residue 48. It seems likely, therefore, that the carboxyl group of Glu 48 is not responsible for Mg 2ϩ binding but moves toward the bound Mg 2ϩ ion to coordinate with it, probably by an electrostatic attraction. The observations that the pK a values of Asp 10 and Asp 70 shift from 6.1 to 4.2 and from 2.6 to 3.4, respectively, upon the binding of the Mg 2ϩ ion to the protein, whereas those of Glu 48 and Asp 134 do not (Oda et al., 1994) support our proposal that Glu 48 is not involved in Mg 2ϩ binding. In addition, the findings that Asp 64 and Asp 121 , in the catalytic domain of avian sarcoma virus integrase (Bujacz et al., 1996), and Asp 443 and Asp 498 , in the RNase H domain of HIV-1 reverse transcriptase (Davies et al., 1991) coordinate a metal cofactor support our proposal because these residues correspond to Asp 10 and Asp 70 in E. coli RNase HI, respectively. Then, the question arises as to the role of Glu 48 in the catalytic function of the enzyme. Why does the mutation of this residue to Gln or Ala almost fully abolish the enzymatic activity?
Role of Glu 48 in the Catalytic Function-According to the latest model for the catalytic mechanism of E. coli RNase HI, His 124 accepts a proton from the water molecule (water molecule B), which acts as a general base; Asp 134 assists it by correctly positioning the water molecule B; Asp 70 anchors the water molecule (water molecule A), which acts as a general acid; and Asp 10 and Glu 48 coordinate with the Mg 2ϩ ion (Kashiwagi et al., 1996). However, our results strongly suggest that repulsive forces between the negatively charged side chains of Asp 10 and Asp 70 are responsible for the binding of the Mg 2ϩ ion to the correct position in the active site of the enzyme. Therefore, Glu 48 , instead of Asp 70 , may be required to anchor 2 S. Kanaya, personal communication.  Katayanagi et al. (1990) (unfilled) and Yang et al. (1990) (filled gray) and those in the enzyme complexed with the Mg 2ϩ ion (Katayanagi et al., 1993b) (filled dark) are superimposed. The imidazole groups of His 124 in the crystal structures of Katayanagi et al. (1990Katayanagi et al. ( , 1993b are not shown because they are not located close enough to interact with other active site residues. The coordinate bonds for the Mg 2ϩ ion are represented by dashed lines. These three crystal structures of Katayanagi et al. (1993b), Yang et al. (1990), and the enzyme-Mg 2ϩ ion complex are deposited in the Protein Data Bank as 2RN2, 1RNH, and 1RDD, respectively. the water molecule A. When the water molecule A supplies a proton to the 3Ј-oxyanion group of the RNA product, it is converted into the hydroxyl ion. This hydroxyl ion should be promptly removed to facilitate the next cleavage reaction cycle. Since the acidic nature of Glu 48 is responsible for the enzymatic activity and because Glu 48 is located close to the water molecule A in a model for the enzyme-substrate complex (Iwai et al., 1995), this residue may be required to eject the hydroxyl ion from the active site due to the negative charge repulsion. We propose that the negatively charged side chain of Glu 48 moves to the functional position upon Mg 2ϩ binding, in which it anchors the water molecule A, and facilitates the replacement of a negatively charged hydroxyl ion with a neutral water molecule during the catalytic cycle (Fig. 9).
Recently, the double mutant protein D10R/E48R, in which Asp 10 and Glu 48 are both replaced by Arg, was shown to retain 87% of the enzymatic activity of the wild-type protein in the absence of the Mg 2ϩ ion (Casareno et al., 1995). This mutant protein was constructed to examine whether mutations of the acidic amino acid residues that are responsible for the Mg 2ϩ binding, to Arg or Lys, provide sufficient positive charge density in the active site to mimic the role of the divalent cation. This result shows that Glu 48 is not related to the catalytic function, and it does not support our proposal. However, the optimal pH of the enzymatic activitiy of the mutant protein greatly shifted to lower values (ϳpH 5.5) as compared with that of the wild-type protein. At an acidic pH, the hydroxyl ion, which is produced when the 3Ј-oxyanion group of the RNA product extracts a proton from the water molecule A, is probably easily replaced by a water molecule without the assistance of an acidic amino acid residue such as Glu 48 . Nevertheless, studies on the catalytic function of the mutant protein D10R/ E48R would provide useful information about the catalytic mechanism of the wild-type enzyme. It would be informative to examine whether the shift in the pH optimum of the enzymatic activitiy of the mutant protein reflects the shift in the pK a value of His 124 .
Stability of D10A-The mutant protein D10A is exceptionally more stable than the wild-type protein at both pH 3.0 and 9.0. The T m value of this mutant protein is the highest among those of the E. coli RNase HI variants with single amino acid substitutions constructed thus far (Ota et al., 1995). The near ultraviolet circular dichroism spectrum of this mutant protein is similar to that of the wild-type protein, 3 indicating that the tertiary structure of the protein is not markedly changed by the mutation. The ratio of the aqueous surface area in the folded state to that in the unfolded state was calculated for Asp 10 as 13.4% (Ooi et al., 1987;Ooi and Oobatake, 1991), indicating that this residue is almost fully buried inside the protein molecule. Since other mutant proteins in which Asp 10 is replaced by polar residues are less stable than D10A, the dramatic thermostability of the mutant protein D10A may be due to an increase in the hydrophobic interactions. Further studies, such as crystallographic analyses, will be necessary to understand the stabilization mechanism of this mutant protein. FIG. 9. A proposed catalytic mechanism of E. coli RNase HI. The possible involvement of Glu 48 in a general acid-base mechanism for the hydrolysis of the P-O3Ј bond of RNA by E. coli RNase HI is represented schematically. Water B and Water A represent the water molecules that act as a general base and a general acid, respectively. Glu 48 coordinates with the Mg 2ϩ ion according to the crystal structure of the enzyme complexed with the Mg 2ϩ ion (Katayanagi et al., 1993b). However, this coordinate bond is not shown in this figure because our results suggest that this residue is not responsible for the Mg 2ϩ binding but is required to anchor the water A molecule. The interaction between Asp 70 and Mg 2ϩ represents an electrostatic bond and not a coordinate bond. His 124 activates the water B molecule by extracting a proton from it, and Asp 134 holds this water molecule. The Mg 2ϩ ion interacts with the 2Ј-hydroxyl group of the substrate, instead of the phosphate group, to stabilize a transient intermediate by the formation of an outer sphere complex (Jou and Cowan, 1991;Uchiyama et al., 1994).