## Abstract

The glycosyl hydrolases present a large family of enzymes that are of great significance for industry. Consequently, there is considerable interest in engineering the enzymes in this family for optimal performance under a range of very diverse conditions. Until recently, tailoring glycosyl hydrolases for specific industrial processes mainly involved stability engineering, but lately there has also been considerable interest in engineering their pH–activity profiles. We mutated four neutral residues (N190, F290, N326 and Q360) in the chimeric Bacillus Ba2 α-amylase to both charged and neutral amino acids. The results show that the pH–activity profile of the Ba2 α-amylase can be changed by inserting charged residues close to the active site. The changes in the pH–activity profile for these neutral → charged mutations do not, however, correlate with the predictions from calculations of the p Ka values of the active site residues. More surprisingly, the neutral → neutral mutations change the pH–activity profile as much as the neutral → charged mutations. From these results, it is concluded that factors other than electrostatics, presumably the dynamic aspects of the active site, are important for the shape of the pH–activity profiles of the α-amylases.

## Introduction

α-Amylases are used in several industrial processes such as starch liquefaction, laundering, dye removal and feed pre-processing (Guzman-Maldonado and Paredes-Lopez, 1995). Several of these processes take place at pH values which are very different from those where the α-amylases perform optimally (Nielsen and Borchert, 2000) and consequently there is great interest in changing the pH–performance profile of the α-amylases (Nielsen et al., 1999a; Shaw et al., 1999) and related enzymes (Fang and Ford, 1998; Wind et al., 1998). The engineering of pH–activity profiles for these enzymes has proven particularly difficult and in the present paper we present experiments and calculations that suggest several reasons for the difficulties encountered when engineering the pH–activity profiles of α-amylases and related enzymes.

### α-Amylases

The α-amylases consist of three domains called A, B and C. Domain A is a TIM-barrel [(α/β)8-barrel], which is interrupted by an irregular β-domain (domain B) inserted between the third β-strand and the third α-helix of the TIM-barrel. Domain C is a Greek key motif which is located approximately on the opposite side of the TIM-barrel with respect to domain B. The active site is situated in a cleft at the interface between domains A and B.

The active site consists of a large number of charged groups, among which are three acids essential to catalytic activity. Two of these, Asp231 and Glu261 (numbering according to the BLA sequence), are believed to be the two catalytic groups. Asp231 is the catalytic nucleophile, while evidence has been presented for Glu261 being the catalytic hydrogen donor (McCarter and Withers, 1996; Uitdehaag et al., 1999). The third essential acid (Asp328) is believed to assist catalysis by hydrogen bonding to the substrate and by elevating the p Ka of Glu261 (Klein et al., 1992; Knegtel et al., 1995; Strokopytov et al., 1995; Uitdehaag et al., 1999).

The catalytic reaction is believed to consist of three steps (Sinnot, 1990; McCarter and Withers, 1994; Davies and Henrissat, 1995; McCarter and Withers, 1996) (see Figure 1). Step one is the protonation of the glycosidic oxygen by the proton donor (Glu261). This is followed by a nucleophilic attack on the C1 of the sugar residue in position –1 by Asp231 [nomenclature as described by Davies et al. (Davies et al., 1997)]. After the aglycone part of the substrate has left, a water molecule is activated, presumably by the now deprotonated Glu261. This water molecule hydrolyses the covalent bond between the nucleophile oxygen (of Asp231) and the C1 of the sugar residue in position –1, thus completing the catalytic cycle.

α-Amylase catalysis is thought to be limited by the protonation of the nucleophile at low pH and by the deprotonation of the hydrogen donor at high pH (Qian et al., 1994; Strokopytov et al., 1995; Fang and Ford, 1998). The rate-limiting step at intermediate pH values is not known. This step has to be largely independent of pH since many α-amylases have a pH–activity profile with a flat top'. Binding of the substrate and release of the product are expected to be independent of pH and it has therefore been speculated that the rate-limiting step is either substrate binding or product release at intermediate pH values.

### Changing the pH–activity profile

If we accept the assumption that α-amylase catalysis is limited at low pH by protonation of the nucleophile (Asp231) and at high pH by deprotonation of the hydrogen donor (Glu261), then the pH–activity profile is determined by the pKa values of the these two active site groups (Kyte, 1995). We consider two types of pH–activity profiles: kcat profiles and the kcat/Km profiles. The kcat profile of an enzyme is determined by the pKa values of the active site groups in the substrate bound form of the enzyme (Kyte, 1995).

Since the substrate is present in high concentrations in the majority of industrial processes where α-amylases are used, it is obviously the kcat profile and not the kcat/Km profile that limits the activity of the enzyme. In order to change the kcat profile, we therefore have to identify the factors that determine the pKa values of the active site residues when the substrate is bound. The protein environment certainly is important and although not much is known about the effect of the substrate, it is undoubtedly also important, since shifts in α-amylase pH–activity profiles have been measured when changing the substrate (Keating et al., 1998). Also, elegant experiments with Bacillus circulans xylanase have shown that the pKa value of the catalytic hydrogen donor cycles' during the catalytic reaction in order to fulfil its dual role as hydrogen donor and hydrogen acceptor (McIntosh et al., 1996).

Changing the substrate is unfortunately not an option when trying to engineer the pH–performance profile of an α-amylase for a specific industrial process and we are therefore left with the option of changing the enzyme (in our case by site-directed mutagenesis) and in that way changing the pKa values of the active site groups and thereby the kcat profile.

### Changing pKa values

The pKa value of a residue depends on the free energy difference between the neutral and the charged states of the residue in the protein. The free energy difference between these two states is influenced both by desolvation effects and by the charges and dipoles in the protein and the substrate. Desolvation effects and the interaction with dipoles are short-ranged and therefore result mainly from interactions with residues in the immediate environment. Mutations that aim at changing the pKa value of a titratable group by changing these energies should therefore be placed in the vicinity of the titratable group.

Mutations that introduce or remove unit charges can be placed further away from the titratable group, because the interaction energy between a titratable group and a unit charge (which may itself be a part of another titratable group) is less dependent on distance.

### Identifying the determinants of α-amylase pH–activity profiles

Previously (Nielsen et al., 1999a), we constructed 15 mutants in and around the active site of Bacillus licheniformis α-amylase (BLA), in an attempt to change its pH–activity profile. The mutations in the active site were conservative in nature and were an attempt to change the pH–activity profile of the enzyme by changing the hydrogen-bond interactions and the solvent accessibility of the active site residues. The mutations outside the active site aimed at changing the pH–activity profile by introducing or removing a charge and in this way perturbing the pKa values of the active site acids.

The mutations in the active site were found to be highly deleterious to the activity of the enzyme, whereas most of the mutations further away from the active site did not change the activity of the enzyme significantly, but nevertheless they produced some changes in the pH–activity profile. Unfortunately, the changes in the pH–activity profiles did not correlate with the predictions from electrostatic potential calculations, prompting us to suggest that effects other than electrostatics were important for determining the pH–activity profile.

In this paper, we describe how these other effects' are indeed important for determining the pH–activity profile. We designed mutations at four positions around the active site of a chimeric Bacillus α-amylase (Ba2). This enzyme was chosen as a model system because high-resolution structures are available for both the apo and holo forms of this enzyme (Brzozowski et al., 2000). We mutated the original neutral residues at the four positions to both neutral and charged residues and examined the effects on the pH–activity profile.

The results indicate that point mutations that are likely to change the dynamics of the active site can change the pH–activity profile. The effects on the pH–activity profile of a neutral → neutral mutation are so large that effects caused by neutral → charged mutations in this study are also likely to be dominated by the associated differences in active site dynamics.

## Materials and methods

### Site-directed mutagenesis

The MegaPrimer' method for site-directed mutagenesis (Sarkar and Sommer, 1990) was used to construct a DNA fragment carrying the mutation. Mutagenic primers were designed so as to introduce or remove a unique site in the gene encoding the hybrid α-amylase (Ba2). The mutant DNA fragments were inserted into a Bacillus expression plasmid in the context of the Bacillus licheniformis α-amylase promoter, signal sequence and transcriptional terminator. The resulting mutant plasmids were transformed into Bacillus subtilis. Mutant colonies were identified by endonuclease digests of colony polymerase chain reaction (PCR) fragments and confirmed by DNA sequencing throughout the region covered by the mutagenic primer.

### Fermentation

Fermentation was carried out at 37°C for 5 days in shake flasks containing a complex medium mainly consisting of potato flour, barley flour and sucrose soy meal (Bang et al., 1999; Beier et al., 2000).

### Protein purification

The supernatant of the fermentation mixture was isolated by flocculation and centrifugation. During ultrafiltration the buffer was changed to 20 mM sodium acetate, pH 5.5. The protein was subsequently applied to an SP-Sepharose Hi-Load column (Pharmacia). A dialysis step was used to change the buffer to 20 mM H3BO3 + 5 mM KCl, pH 9.6. The final step of the purification procedure consisted of anion-exchange chromatography on a Mono-Q Hi-Load column (Pharmacia). All protein preparations were at least 95% pure as shown by SDS–PAGE.

### Activity assays

The activity as a function of pH was measured over the range pH 4.0–10.5, using the Phadebas α-amylase test kit (Pharmacia). Measurements were carried out in duplicate at 30°C in 50 mM Britton–Robinson buffer (50 mM H3PO4 + 50 mM CH3COOH + 50 mM H3BO3) containing 0.1 mM CaCl2. The Phadebas α-amylase test kit is based on the release of blue colour from the substrate (blue-coloured starch) upon cleavage. Hydrolysis is stopped by adding 1/6 volume of 1 M NaOH to the reaction mixture. After removal of the unhydrolysed blue starch by filtration, the amount of hydrolysed substrate is proportional to the absorbance at 620 nm.

Since the substrate is insoluble (and added in large quantities), the activity measurements obtained by this method can be regarded as kcat for insoluble starch.

### Stability assays

Stability was measured as the residual activity after incubation at 30°C for 15 min. Measurements were carried out at pH 4.5, 7.0 and 9.0 and were performed in duplicate. None of the mutants showed any detectable differences from the wild-type stability.

### pKa calculations

pKa calculations were performed with the WHAT IF pKa calculation routines (Nielsen and Vriend, 2001). The routines apply the hydrogen-bond optimization procedure of Hooft et al. (Hooft et al., 1996) in order to model each of the protein protonation states accurately. DelPhi II (Nicholls and Honig, 1991) was used to calculate electrostatic energies. The parameters for DelPhi II were set as described previously (Nielsen and Vriend, 2001), namely protein dielectric constant, 16 (for residues that participate in crystal contacts, have an alternative accessible rotamer or an average B-factor of >20) or 8 (all other residues); ionic strength, 0.160 M; solvent probe radius, 1.4 Å; and solvent dielectric constant, 80. The calculations were performed without water molecules, because the inclusion of water molecules was found to decrease the accuracy of the pKa calculations for a test set of nine proteins (Nielsen and Vriend, 2001).

A model of Ba2 in complex with malto-nonaose was constructed from the Ba2 α-amylases–acarbose X-ray structure (Brzozowski et al., 2000). The few changes that were necessary in order to convert the inhibitor to a substrate were made using Quanta. The CHARMm 22 parameter set was used as the source of charges and radii for the substrate in the pKa calculations.

### Preparation of mutant structures

Mutant structures for use in the pKa calculations were designed using the WHAT IF position-specific rotamer libraries (Chinea et al., 1995). Mutant structures were inspected visually.

## Results

We constructed 12 point mutations to examine the determinants of the pH–activity profile for Ba2. The activity of all mutant amylases was within one order of magnitude of the activity of the wild-type at pH 7.0, with three variants having higher activity than the wild-type. The stability assays showed that the pH-dependent stability of the mutants was indistinguishable from that of the wild-type. The results are summarized in Table I.

Mutations were clustered at four positions, namely N190, F290, N326 and Q360 (Figure 3). The sites for the mutations were chosen reasonably close to the active site in order for the mutations to have a significant effect on the pKa values of the active site acids (Table II). Also, we required that the wild-type residue at the position should be neutral. This was necessary in order to be able to construct neutral → neutral mutations. Visual inspection and WHAT IF's (Vriend, 1990) structure analysis module furthermore indicated that mutations could be made at these positions without perturbing the local structure significantly. The apparent kcat profiles for the wild-type and the mutants were measured with the Phadebas assay. These kcat profiles will be referred to as pH–activity profiles' in the following.

### F290K/A/E

F290 is situated in the α-helix that lies between β-strand 6 and α-helix 6 of the TIM-barrel. This in-between' α-helix is not a part of the classical TIM-barrel. Phe290 sits next to His289, which hydrogen bonds to Ser337. Ser337 is positioned in the loop that covers the active site and its mutation to glycine results in a protein with 2% of wild-type activity at pH 7.0, thus hinting at an important role played by the hydrogen bond between Ser337 and His289.

### N326L/A/D

The pH–activity profile of N326L has lost the characteristic peak around pH 5.5 and has become almost completely flat. Furthermore, the basic limb of the pH–activity profile is shifted slightly towards more basic pH values.

### pKa calculations

#### Wild-type pKa calculations

pKa calculations were performed for both the apo and holo forms of the enzyme. In the apo form, Glu261 and Asp328 form a tightly coupled system that titrates roughly as one group (hence the large oscillations in the titration curves for these two residues) with a pKa value of ~10, whereas Asp231 titrates with an apparent pKa of ~2. This corresponds fairly well with the established view of the α-amylase catalytic mechanism with Glu261 being the hydrogen donor and Asp231 being the nucleophile. In the calculations with the substrate, however, the titration curve for Asp231 becomes biphasic with apparent pKa values of ~4 and 7, whereas Glu261 remains negatively charged over the entire pH range. Asp328 is predicted to have a pKa value of ~8. The calculations for the holo form thus suggest that Asp328 is the hydrogen donor. It is, however, not clear how reliable the calculations for the holo form are as the coordinates for the holo form complex are derived from an enzyme–inhibitor complex rather than from an enzyme–substrate complex.

### pKa calculations for the mutants

The pKa calculations for the mutants were carried out with the holo structure of the enzyme to give the pKa shifts for the active site acids when the substrate is bound. The perturbations in the titration curves for the active site residues were fairly small for all mutations of Asn190 and Phe290. This is in perfect agreement with the experimental results for the mutations at position 190, but does not correlate with the results for the mutations at position 290. Changes were observed for the mutations of Asn326 and Gln360. These changes are described in more detail below.

### N326D/A/L

The calculations for N326A show an upward shift in the pKa value of His327 and a downward shift in the pKa value of Asp328. The calculations for N326L show a slight upward shift in the pKa value of Asp328 and an even smaller upward shift in the pKa value of Glu261. The N326D pKa calculations predict that the pKa values of Asp231, His327, Asp328 and Glu261 increase.

### Q306A/E/K

The pKa calculations for the mutants in position 360 predict that the pKa value of Asp231 becomes elevated upon all three mutations. In the case of Q360A and Q360K, the shift is very large, whereas a smaller shift is seen for Q360E. The pKa values of His327 and Asp328 are predicted to become slightly lower for Q360A. The calculations for Q360K predict that the pKa values for His327 and Asp328 increase.

### Correlation of the calculated pKa shifts with the experimentally observed pH–activity profile shifts

The magnitude and direction of the pH–activity profile shifts for the Ba2 mutants are not reproduced by the pKa calculations. This is most clearly seen when comparing the calculations with the experimental results for F290A and N326A. F290A gives the largest shift in the pH–activity profile, but the pKa calculations produce insignificant changes in the active site pKa values for this mutation. In the case of N326A, the pKa calculations show a significant perturbation of the pKa values of Asp328 and His327, but the pH–activity profile for this mutant is almost identical with that of the wild-type.

The calculations for the mutations that introduce charges also do not correlate well with the experimental pH–activity profile shifts and this strongly suggests that long-range electrostatics are less important for the pH–activity profile of the α-amylases than previously thought.

## Discussion

The most pronounced effect on the pH–activity profile for a neutral → neutral mutation is seen for F290A. Phenylalanine and alanine have different sizes and the large effect is seen for this mutation is therefore not surprising. The imidazole ring of His289 packs against the aromatic ring of Phe290 in the wild-type structure and removing the aromatic ring of Phe 290 (by the F290A mutation) makes His289 more solvent accessible. The higher solvent accessibility changes the pKa of His289 slightly (by 0.1 unit in the pKa calculations) and in this way the F290A mutation changes the electrostatic field in the protein. This change in the electric field cannot, however, be responsible for the observed change in the pH–activity profile, since both the F290E and F290K mutations should change the electrostatic field even more than F290A. They should thus produce pH–activity profiles that are even more different from the wild-type profile than the pH–activity profile of F290A. This is, however, clearly not the case and the changes in the pH–activity profile caused by F290A therefore have to be ascribed to effects other than electrostatics as outlined below.

### The His289–Ser337 hydrogen bond

His289 hydrogen bonds to Ser337 and the change in the pKa, and possibly in the dynamics of His289, is likely to affect the strength of this hydrogen-bond. Ser337 is situated in a loop that covers the active site and it is therefore likely that a change in the Ser337–His289 hydrogen bond could affect the dynamics of the active site. The hydrogen bond between His289 and Ser337 is known to be important, since the mutation of Ser337 to glycine produces an enzyme with a 50-fold reduction activity (J.E.Nielsen, G.Vriend and T.V.Borchert, unpublished results).

The reason for the discrepancy between the pKa calculations and the experiments for the mutants F290K and F290E now becomes clear, since these mutations are also likely to change the solvent accessibility and the dynamics of His289. The pH–activity profile shifts resulting from these two mutations are therefore the combined effects of the charge and the change in the mobility of His289 that are induced by the mutations. The pKa calculations do not attempt to model such effects and are therefore unable to reproduce the pH–activity profile shifts.

### Asn326

Takase constructed the N326D mutation (Takase, 1993) in the Bacillus stearothermophilus α-amylase and reported the same downward shift in the pH–activity profile as we found. This shift is not readily explainable and is possibly due to changes in the mobility. The neutral → neutral mutations in this position (N326A and N326L) are both likely to influence the active site dynamics, but only N326L shows a significant change in the pH–activity profile.

### Gln360

The mutations at position 360 are remarkable in that Q360E has a large impact on the pH–activity profile, whereas the mutations Q360K and Q360A, which are expected to change the dynamics of the enzyme more than Q360E, have wild-type pH–activity profiles. This suggests that the effect of Q360E is purely electrostatic, although it is difficult to understand why Q360K does not have an equally large effect on the pH–activity profile.

### Effect of introducing a point mutation on a tightly coupled system of titratable groups

In constructing the point mutations in BLA and BA2 we have silently assumed that the effect of inserting a titratable group near the active site could be predicted using the well-established rules that are summarized in Figure 2. The pKa calculations for Q360K show that this is not always the case, as the pKa of Asp328 is calculated to increase when Gln360 is mutated to a lysine. This is clearly not what would be expected and shows that counterintuitive effects can indeed be achieved when a system of tightly coupled titratable groups is perturbed. The phenomenon is illustrated in Figure 6, where the rules of Figure 2 are shown to break down for an Asp–Glu system much like the two catalytic acids in the α-amylases. It is therefore essential in many cases to use pKa calculations to predict the effect of charged point mutations on the pKa values in the active site.

### Conclusion

We have shown that significant changes in the activity of pH–activity profiles of a Bacillus α-amylase can indeed be achieved using site-directed mutagenesis. The shifts in the pH–activity profiles for the mutants did not agree with the calculated changes in the active site electrostatics.

We speculate that changes in the dynamics of the active site residues are at least as important for the pH–activity profile as the changes in the active site electrostatics caused by the introduction of a charged residue. This is corroborated by the mutations F290A and N326L, which change the pH–activity profile without changing the net charge on the molecule. This implies that the pKa values of the active residues can be changed significantly by altering the dynamics of the active site and thus suggests an alternative approach to the engineering of pH–activity profiles.

A detailed explanation of the effects of the neutral → neutral mutations is difficult, as we do not know the motions that are important for the catalytic activity of the α-amylases. It is likely that very accurate and long molecular dynamics (MD) simulations could provide insights into this, but in view of the present day MD force fields and computer speeds, this is not a feasible solution. The pKa calculation method that we used does not model heavy atom mobility and the pKa shifts induced by mobility changes can therefore not be reproduced in the calculations. Furthermore, we may well have underestimated the value of the dielectric constant in the active site in our calculations. Employing a higher dielectric constant in the pKa calculations would reduce the magnitude of the calculated shifts in the titration curves, but it is unlikely that it would give a better qualitative agreement with the experimental data.

The pKa calculations also show that the α-amylase active site is a strongly connected system of titratable groups. The example with a strongly coupled Asp–Glu system given in Figure 6 shows that the effect of inserting a titratable group cannot always be predicted by using the simple scheme of Figure 2. If the circumstances are right, one can observe the exact opposite of what was predicted. We speculate that the shifts in the wrong' direction seen for several of the mutants in this and in the previous study (Nielsen et al., 1999a) could be partly due to such effects. It seems more likely, though, that the change in the pH–activity profiles result primarily from the change in the active site dynamics.

The findings of this study stress the point that dynamics are an essential part of every enzyme and that rational engineering of enzyme activity is more likely to succeed if a detailed description of the enzyme mobility is available when designing the point mutations.

Table I.

Specific activity of the Ba2 point mutations at pH 7.0 as measured by the Phadebas assay

Mutant Relative activity (%)
N190D  35.72
N190K  38.0
N190G  38.0
F290E 125
F290K  75.3
F290A 152
N326D  43.4
N326A  49.0
N326L  20.0
Q360E 242
Q360K  61.0
Q360A  70.0
Mutant Relative activity (%)
N190D  35.72
N190K  38.0
N190G  38.0
F290E 125
F290K  75.3
F290A 152
N326D  43.4
N326A  49.0
N326L  20.0
Q360E 242
Q360K  61.0
Q360A  70.0
Table II.

Distance from mutated residues to the active site acids

Residue/atom Distance to Asp231 Cγ (Å) Distance to Glu261 Cδ (Å) Distance to Asp328 Cγ (Å)
Asn190 Cγ 17.5 15.6 17.2
Phe290 Cγ 18.3 13.0 12.4
Asn326 Cγ 13.1 12.6 7.1
Gln360 Cδ 13.2 11.0 11.3
Residue/atom Distance to Asp231 Cγ (Å) Distance to Glu261 Cδ (Å) Distance to Asp328 Cγ (Å)
Asn190 Cγ 17.5 15.6 17.2
Phe290 Cγ 18.3 13.0 12.4
Asn326 Cγ 13.1 12.6 7.1
Gln360 Cδ 13.2 11.0 11.3
Fig. 1.

Catalytic mechanism of the retaining glycosyl hydrolases. (I) Protonation of the glycosidic oxygen by the hydrogen donor (Glu261) and attack on the glucose C1 by the nucleophile (Asp231). Departure of the reducing end of the substrate. (II) Activation of a water molecule, cleavage of C1–Asp231 covalent bond. (III) Regeneration of the initial protonation states.

Fig. 1.

Catalytic mechanism of the retaining glycosyl hydrolases. (I) Protonation of the glycosidic oxygen by the hydrogen donor (Glu261) and attack on the glucose C1 by the nucleophile (Asp231). Departure of the reducing end of the substrate. (II) Activation of a water molecule, cleavage of C1–Asp231 covalent bond. (III) Regeneration of the initial protonation states.

Fig. 2.

Environmental effects on the pKa of titratable residues. The effect of placing a titratable group in a negative, positive and hydrophobic environment is summarized.

Fig. 2.

Environmental effects on the pKa of titratable residues. The effect of placing a titratable group in a negative, positive and hydrophobic environment is summarized.

Fig. 3.

The position of the point mutations in Ba2. Domain A, green; domain B, cyan; domain C, light grey. The calcium ions are shown as red spheres and the sodium ion is shown as an orange sphere. The nonaose substrate used in the pKa calculations is shown in yellow. The active site acids Asp231, Glu261 and Asp328 are shown in red.

Fig. 3.

The position of the point mutations in Ba2. Domain A, green; domain B, cyan; domain C, light grey. The calcium ions are shown as red spheres and the sodium ion is shown as an orange sphere. The nonaose substrate used in the pKa calculations is shown in yellow. The active site acids Asp231, Glu261 and Asp328 are shown in red.

Fig. 4.

pH–activity profiles for the Ba2 wt and the mutants. (a) wt, N190K, N190G and N190D; (b) wt, F290E, F290K and F290A; (c) wt, N326L, N326A and N326D; (d) wt, Q360K, Q360A and Q360E.

Fig. 4.

pH–activity profiles for the Ba2 wt and the mutants. (a) wt, N190K, N190G and N190D; (b) wt, F290E, F290K and F290A; (c) wt, N326L, N326A and N326D; (d) wt, Q360K, Q360A and Q360E.

Fig. 5.

Calculated titration curves of the three active site acids Asp231, Glu261 and Asp328 in the apo-form and in the holo-form of the enzyme. Asp231, red; Glu261, magenta; Asp328, green.

Fig. 5.

Calculated titration curves of the three active site acids Asp231, Glu261 and Asp328 in the apo-form and in the holo-form of the enzyme. Asp231, red; Glu261, magenta; Asp328, green.

The authors thank Vibeke Holbo for help with the construction and purification of mutant proteins, Jytte Piil for providing purified wild-type protein and Rebecca Wade, Barry Honig and Lawrence McIntosh for stimulating discussions.

## References

1999
) Cloning and characterization of an endo-beta-1,3(4)glucanase and an aspartic protease from Phaffia rhodozyma CBS 6938.
Appl. Microbiol Biotechnol.
,
51
,
215
–222.
Beier,L., Svendsen,A., Andersen,C., Frandsen,T.P., Borchert,T.V. and Cherry,J.R. (
2000
) Conversion of the maltogenic alpha-amylase Novamyl into a CGTase.
Protein Eng.
,
13
,
509
–513.
Brzozowski,A.M., Lawson,D.M., Turkenburg,J.P., Bisgaard-Frantzen,H., Svendsen,A., Borchert,T.V., Dauter,Z., Wilson,K.S. and Davies,G.J. (
2000
) Structural analysis of a chimeric bacterial α-amylase. High resolution analysis of native and ligand complexes.
Biochemistry
,
39
,
9099
–9107.
Chinea,G., Padron,G., Hooft,R.W.W., Sander,C. and Vriend,G. (
1995
) The use of position-specific rotamers in model building by homology.
Proteins
,
23
,
415
–421.
Davies,G.J. and Henrissat,H. (
1995
) Structures and mechanisms of glycosyl hydrolases.
Structure
,
3
,
853
–859.
Davies,G.J., Wilson,K.S. and Henrissat,B. (
1997
) Nomenclature for sugar-binding subsites in glycosyl hydrolases.
Biochem. J.
,
321
,
557
–559.
Fang,T.Y. and Ford,C. (
1998
) Protein engineering of Aspergillus awamori glucoamylase to increase its pH optimum.
Protein Eng.
,
11
,
383
–388.
Guzman-Maldano,H. and Paredes-Lopez,O. (
1995
) Amylolytic enzymes and products derived from starch; a review.
Crit. Rev. Food. Nutr.
,
36
,
373
–403.
Hooft,R.W., Sander,C. and Vriend,G. (
1996
). Positioning hydrogen atoms by optimising hydrogen-bond networks in protein structures.
Proteins
,
26
,
363
–376.
Keating,L., Kelly,C. and Fogarty,W. (
1998
). Mechanism of action and the substrate-dependent pH maximum shift of the α-amylase of Bacillus coagulans.
Carbohydr. Res.
,
309
,
311
–318.
Klein,C., Hollender,J., Bender,H. and Schulz,G.E. (
1992
) Catalytic center of cyclodextrin glycosyltransferase derived from X-ray structure analysis combined with site-directed mutagenesis.
Biochemistry
,
31
,
8740
–8746.
Knegtel,R.M, Strokopytov,B., Penninga,D., Faber,O.G., Rozeboom,H.J., Kalk,K.H., Dijkhuizen,L. and Dijkstra,B.W. (
1995
) Crystallographic studies of the interaction of cyclodextrin glycosyltransferase from Bacillus circulans strain 251 with natural substrates and products.
J. Biol. Chem.
,
270
,
29256
–29264.
Kyte,J. (
1995
) Mechanism in Protein Chemistry. Garland Publishing, New York, pp. 258–283.
McCarter,J.D. and Withers,S.G. (
1994
) Mechanisms of enzymatic glycoside hydrolysis.
Curr. Opin. Struct. Biol.
,
4
,
885
–892.
McCarter,J.D. and Withers,S.G. (
1996
) Unequivocal identification of Asp-214 as the catalytic nucleophile of Saccharomyces cerevisiae α-glucosidase using 5-fluoro glycosyl fluorides.
J. Biol. Chem.
,
271
,
6889
–6894.
McIntosh,L.P., Hand,G., Johnson,P.E., Joshi,M.D., Korner,M., Plesniak,L.A., Ziser,L., Wakarchuk,W.W. and Withers,S.G. (
1996
) The pKa of the general acid/base carboxyl group of a glycosidase cycles during catalysis: a 13C-NMR study of Bacillus circulans xylanase.
Biochemistry
,
35
,
9958
–9966.
Nicholls,A. and Honig,B. (
1991
) A rapid finite difference algorithm, utilizing successive over-relaxation to solve the Poisson–Boltzmann equation.
J. Comput. Chem.
,
12
,
435
–445.
Nielsen,J.E. and Borchert,T.V. (
2000
) Protein engineering of bacterial α-amylases.
Biochim. Biophys. Acta
,
1543
,
253
–274.
Nielsen,J.E. and Vriend,G. (
2001
) Optimising the hydrogen-bond network in Poisson–Boltzmann equation based pKa calculations.
Proteins
,
43
,
403
–412.
Nielsen,J.E., Beier,L., Otzen,D., Borchert,T.V., Frantzen,H.B., Andersen,K.V. and Svendsen,A. (
1999
) Electrostatics in the active site of an α-amylase.
Eur. J. Biochem.
,
264
,
816
–824.
Qian,M., Haser,R., Buisson,G., Duee,E. and Payan,F. (
1994
) The active centre of a mammalian α-amylase. Structure of the complex of a pancreatic α-amylase with a carbohydrate inhibitor refined to 2.2-Å resolution.
Biochemistry
,
33
,
6284
–6294.
Sarkar,G. and Sommer,S.S. (
1990
). The megaprimer' method of site-directed mutagenesis.
Biotechniques
,
8
,
404
–407.
Shaw,A., Bott,R. and Day,A.G. (
1999
) Protein engineering of α-amylase for low pH performance.
Curr. Opin. Biotechnol.
,
10
,
349
–352.
Sinnot,M.L. (
1990
) Catalytic mechanisms of enzymic glycosyl transfer.
Chem. Rev.
,
90
,
1171
–1202.
Strokopytov,B., Penninga,D., Rozeboom,H.J., Kalk,K.H., Dijkhuizen,L. and Dijkstra,B.W. (
1995
) X-ray structure of cyclodextrin glycosyltransferase complexed with acarbose. Implications for the catalytic mechanism of glycosidases.
Biochemistry
,
34
,
2234
–2240.
Takase,K. (
1993
) Effect of mutation of an amino acid residue near the catalytic site on the activity of Bacillus stearothermophilus alpha-amylase.
Eur. J. Biochem.
,
211
,
899
–902.
Uitdehaag,J.C., Mosi,R., Kalk,K.H., van der Veen,B.A., Dijkhuizen,L., Withers,S.G. and Dijkstra,B.W. (
1999
) X-ray structures along the reaction pathway of cyclodextrin glycosyltransferase elucidate catalysis in the α-amylase family.
Nature Struct. Biol.
,
6
,
432
–436.
Vriend,G. (
1990
) WHAT IF: a molecular modelling and drug design program.
J. Mol. Graphics
,
8
,
52
–56.
Wind,R.D., Uitdehaag,J.C., Buitelaar,R.M., Dijkstra,B.W. and Dijkhuizen,L. (
1998
) Engineering of cyclodextrin product specificity and pH optima of the thermostable cyclodextrin glycosyltransferase from Thermoanaerobacterium thermosulfurigenes EM1.
J. Biol. Chem.
,
273
,
5771
–5779.