Directed evolution of a bacterial α-amylase: Toward enhanced pH-performance and higher specific activity

Abstract

α-Amylases, in particular, microbial α-amylases, are widely used in industrial processes such as starch liquefaction and pulp processes, and more recently in detergency. Due to the need for α-amylases with high specific activity and activity at alkaline pH, which are critical parameters, for example, for the use in detergents, we have enhanced the α-amylase from Bacillus amyloliquefaciens (BAA). The genes coding for the wild-type BAA and the mutants BAA S201N and BAA N297D were subjected to error-prone PCR and gene shuffling. For the screening of mutants we developed a novel, reliable assay suitable for high throughput screening based on the Phadebas assay. One mutant (BAA 42) has an optimal activity at pH 7, corresponding to a shift of one pH unit compared to the wild type. BAA 42 is active over a broader pH range than the wild type, resulting in a 5-fold higher activity at pH 10. In addition, the activity in periplasmic extracts and the specific activity increased 4- and 1.5-fold, respectively. Another mutant (BAA 29) possesses a wild-type-like pH profile but possesses a 40-fold higher activity in periplasmic extracts and a 9-fold higher specific activity. The comparison of the amino acid sequences of these two mutants with other homologous microbial α-amylases revealed the mutation of the highly conserved residues W194R, S197P, and A230V. In addition, three further mutations were found K406R, N414S, and E356D, the latter being present in other bacterial α-amylases.

The main natural substrate of α-amylases (EC 3.2.1.1. 1.4–α-d-glucan glucanohydrolases) is starch, which is cleaved into branched and unbranched oligosaccharides. The industrial interest in α-amylases is based on their application in sugar producing processes and more recently in detergency, where α-amylases with high thermostability and/or high activity are required. The α-amylase from Bacillus amyloliquefaciens (BAA) is a liquefying α-amylase with a temperature optimum of 50°C–70°C and a pH optimum of 6. The BAA is stable up to 50°C and shows 12% residual activity after incubation at pH 12 (Granum 1979). The industrial applications of BAA include starch liquefaction and detergency (Norman 1982; Kottwitz et al. 1999). Other liquefying α-amylases frequently used in industrial processes are the α-amylases derived from Bacillus licheniformis (BLA) and Bacillus stearothermophilus (BStA), which share an amino acid sequence similarity of 87% and 73% with BAA.

The use of α-amylases in industrial processes requires that they be adapted to the prevailing process conditions. Protein engineering techniques have been applied to the BLA to improve its thermal stability by rational protein engineering (Svensson and Sogaard 1992; Svensson 1994; Declerck et al. 1995). Recently, the tolerance of the BLA toward low pH was enhanced by directed evolution (Shaw et al. 1999). Extensive studies were carried out to identify determinants for the pH-profile of BLA (Nielsen et al. 1999, 2001). Comparatively little work has been done on improving the BAA, and has focused instead on the thermostability of the BLA. Thermostability determinants were identified by the construction of BAAxBLA hybrids (Conrad et al. 1995) or by deletion of two amino acids of the BAA (Suzuki et al. 1989). More examples for protein engineering of bacterial α-amylases can be found in the review of Nielsen and Borchert (2000).

The structure of α-amylase is highly conserved and consists of the three domains A, B, and C. Domain A, which is an (α/β)₈-barrel (TIM-barrel), includes the N terminus and the active site. Domain B shows the highest variability within several structures (Svensson 1994), and is almost exclusively formed by β-strands. Domain C, which forms a Greek key motif is located on the other side of the TIM-barrel and contains the C terminus.

Directed evolution is a powerful tool to improve the properties of proteins (Kuchner and Arnold 1997) and usually comprises the creation of a pool of mutated genes and the subsequent screening for improved gene products. Improved mutants can be subjected to consecutive rounds of mutagenesis and screening. We used error-prone PCR to randomly introduce mutations to the gene coding for the BAA in combination with gene shuffling (Stemmer 1994a,b) for the recombination of mutations. To detect amylase activity, we developed a high throughput screening protocol based on the activity of the α-amylase toward an insoluble dye-conjugated starch polymer. To screen simultaneously for improved activity at high pH and improved specific activity, two-dimensional screening at two different pH values was carried out to reduce the appearance of expression mutants and other false positives, a problem frequently seen in directed evolution experiments (Bornscheuer et al. 1999; Zhao et al. 1999).

Because screening at different pH values requires enzyme solutions with low buffer capacity, care has to be taken with the sample preparation. Escherichia coli has no mechanism for the active secretion of heterologously expressed proteins into the media. Therefore, disruption of the cells is needed for the release of expression products into the medium. The BAA expressed in E. coli is secreted into the periplasmic space (Pretorius et al. 1988), allowing the isolation of the BAA from the periplasm by a cold osmotic shock. Unfortunately, existing protocols require multiple pipetting and incubation steps (Neu and Heppel 1965), and therefore, lower the throughput. We decided to facilitate the secretion of the BAA into the culture medium by the coexpression of the Bacteriocin release protein (BRP; van der Wal et al. 1995a^,b). The BRP stimulates phospholipase C, which in turn hydrolyses the phospholipids present in the inner and outer cell membrane, thereby generating permeable regions. The secretion of an α-amylase from E. coli by BRP coexpression has been successfully demonstrated (Yu and San 1992).

Results

Expression of the BAA and the coexpression of BRP in E. coli

The BAA wild-type gene was cloned into the NdeI-PstI-site of the plasmid pG-PFE, to give pG-BAA (Fig. 1 ▶). The expression of BAA in E. coli XL1-Blue under control of the rhamnose promoter proved to be constitutive, because uninduced cultures showed substantial amounts of α-amylase activity. Nevertheless, the coexpression of the Bacteriocin Release Protein (BRP) from plasmid pJL3 for the secretion of the BAA increased the α-amylase activity in the culture medium and was accompanied by a strong increase of the α-amylase activity in the periplasmic fraction (Fig. 2 ▶).

Figure 1.

Map of plasmid pGBAA for expression of BAA in E. coli. Expression is controlled by the rhamnose inducible promoter P_Rh. The plasmid contains an ampicillin resistance gene Amp^R, and a pUC18 origin of replication.

Figure 2.

Expression of BAA in E. coli XL1-Blue. For induced cultures, 0.2% rhamnose was added at OD₆₀₀ = 0.4. For both rhamnose-induced and -uninduced cultures, the expression of BRP was induced with 20 μM IPTG 1 h after OD₆₀₀ = 0.4 was reached. Cultivation was stopped 3 h after IPTG induction, periplasmic extracts were prepared, and the α-amylase activity concentration in the supernatant and the periplasmic extracts was determined.

Assay evaluation

The microtiter plate modified Phadebas HTS assay for the determination of α-amylase activity was evaluated by measuring different dilutions of the commercial α-amylase BAN 240L in a 96-well microtiter plate with eight replicas for each activity. Between 0 and 1000 U/L a linear correlation between absorption and activity was found with R² = 0.9995. The linearity was remarkably high for activities up to 2000 U/L (Fig. 3 ▶). The relative standard deviation ranged from 0.8% to 5.6%. The average relative error lay between 1.7% and 11.8% (Table 1).

Figure 3.

Calibration curve of the modified Phadebas HTS assay. The linear range lies between 0 U/L and 1000 U/L. The error bars show the maximal errors of eight replicates within one row of a microtiter plate.

Table 1.

The modified Phadebas assay

Activity concentration/(U/l)	A(620)	Standard deviation	Avg. error	Avg. rel. error/%
0	0.000	0.008	0.012	6.8
8	0.012	0.013	0.022	11.8
19	0.020	0.008	0.014	7.3
49	0.055	0.004	0.006	2.8
91	0.082	0.003	0.004	1.7
195	0.178	0.013	0.022	6.3
403	0.392	0.013	0.022	3.8
605	0.604	0.014	0.021	2.7
1021	1.007	0.056	0.091	7.7

The standard curve was obtained by measuring the α amylase activity of a dilution row of BAN 240L at pH 7.

Two-dimensional assay setup

Because one of the desired parameters was improved activity at alkaline pH, we elected to screen at two different pH values: pH 7 and pH 10. Clones were expected to be located within defined regions of the two-dimensional plot according to their activity properties (Fig. 4 ▶): Wild-type clones or wild-type-like clones should lie along an axis with the slope of the wild-type ratio of pH 7 activity/pH 10 activity. Variants with improved activity at pH 10 should possess a lower pH 7/pH 10 activity ratio, and therefore should lie along an axis with a lower slope. Variants with the second desired parameter improved, that is, specific activity, should show higher activity than the wild type but the same pH 7/pH 10 ratio and therefore, should be located farther from the origin. In this area variants with improved expression/procession properties should be found as well. Variants showing improvements in both activity at alkaline pH and specific activity should show a lower slope and be located farther from the origin.

Figure 4.

The two-dimensional screening system. α-Amylase activity was measured at pH 10 and pH 7. Clones with wild-type-like pH activity profile should be located between the parallels, whereas clones with improved activity at alkaline pH should be shifted rightward. Clones with improved activity should be located farther away from the origin.

To determine the background of the assay, a population of 96 α-amylase negative clones (E. coli XL1-Blue cotransformed with the plasmids pG-PFE coding for an esterase from Pseudomonas fluorescence and pJL3 encoding BRP to allow secretion) was induced and screened at pH 7 and pH 10 as described in the Materials and Methods section. This construct with a gene coding for a nonamylolytic enzyme was chosen to simulate a comparable expression stress. The absorption values at pH 7 and pH 10 were plotted against each other resulting in a background of about 0.15 absorption units in each dimension. To determine the wild-type region, 96 colonies of E. coli transformed with the plasmids for BAA and BRP expression (E. coli; pGBAA WT, pJL3) were induced and screened. Measured values were distributed between absorption values for activity at from pH 7 to activity at pH 10. Based on the knowledge that for the wild-type enzyme the activity at pH 10 drops to 12% compared to pH 7 it was anticipated that the data points would lie along a curve with a gradient of 1/0.12. In fact, it was found that the data set could be bound by two curves with the equations A(620. pH 7) = 8 × A(620. pH 7) and A(620. pH 7) = 8 × A(620. pH 7)-0.8 (Fig. 5 ▶) (note that 8 = 1/0,125).

Figure 5.

Screening of the DNA-shuffling library: (diamonds) DNA-shuffling library, (squares) best hits from the same library, (triangles) best hits from the error-prone library. The two mutants BAA 29 and BAA 42 are annotated.

Starting points for directed evolution

In addition to the wild-type BAA we used the two point mutants BAA S201N and BAA N297D as starting points for directed evolution. Both mutants were constructed by site-directed mutagenesis of the BAA gene (Bessler et al. 2000). The activity of BAA S201N at pH 10 and pH 11 is increased by 16% and 50%. While the activity of BAA N297D at pH 10 is comparable to the wild type, the activity at pH 11 is increased by 50%.

Construction and screening of the error-prone library

A random library of BAA mutants was created under the conditions described resulting in the production of 7200 clones. Screening for α-amylase active colonies was performed on agar plates containing 1% starch and staining for activity with iodine. The inactivation rate was found to be 30%. DNA sequencing of 40,000 base pairs of randomly picked clones revealed a mutation rate of 0.6/1000 corresponding to 1 mutation per gene for the 1545-bp gene. Screening for α-amylase activity at pH 10 and pH 7 using the modified Phadebas HTS assay identified 26 clones with improved activity compared to the wild type. DNA-sequencing proved that 16 clones carried at least one nonsilent mutation while 10 of these were shown to be wild type or possessing silent mutations. This demonstrates that the two-dimensional screening method does not completely prevent false positive mutants being found, and therefore, subsequent rescreening is advisable.

Construction and screening of the DNA-shuffling library

The BAA genes of these 16 clones were subjected to DNA-shuffling resulting in a total of 10,000 clones. Nine hundred sixty active clones were found by activity staining as described, corresponding to a deactivation rate of 90%. Sequencing of eight clones revealed a comparably high mutation rate of 3.5 mutations per gene, which is hold responsible for the high deactivation rate. Screening of these 960 clones (Fig. 5 ▶) and subsequent rescreening as well as screening of retransformed colonies yielded the mutants BAA 42 and BAA 29, which were found to have improved activity at pH 10 (BAA 42) and an improved overall activity (BAA 29 and BAA 42). Both mutants were sequenced completely and further characterized.

Characterization of the mutants: Biochemical analysis

Both BAA 42 and BAA 29 show an enhanced activity in periplasmic fractions due to a higher protein concentration and higher specific activity (Table 2). The activity concentration in periplasmic fractions of BAA 42 is increased by a factor of about 3.6. In the periplasmic fractions of BAA 29, the activity concentration is approximately 41.2 times higher. Although the specific activity of BAA 42 is enhanced 1.5-fold, the specific activity of BAA 29 is more than 9.3 times higher compared to the wild type. To determine if the mutants show higher expression rates the approximate α-amylase concentration in periplasmic extracts was calculated from the ratio of activity concentration divided by the specific activity. For BAA 42, the value is 2.4-fold higher than for the wild type, for BAA 29 even 4.5-fold.

Table 2.

Activity and specific activity of BAA, BAA 42, and BAA 29

	Activity concentration (periplasm) [U/L]	Specific activity [U/mg]	Protein concentration (periplasm) [mg/L]
BAA WT	1420 ± 38 (1.00 ± 0.03)	15.1 ± 1.1 (1.00 ± 0.07)	93.8 (1.0)
BAA 29	59484 ± 1602 (41.89 ± 1.13)	140.9 ± 9.5 (9.27 ± 0.63)	422.2 (4.5)
BAA 42	5035 ± 95 (3.55 ± 0.06)	22.0 ± 1.2 (1.46 ± 0.08)	229.0 (2.4)

The values in parentheses give the relative activities and concentrations in comparison to the wild type.

The pH activity profile (Fig. 6 ▶) of BAA 42 shows an optimum at pH 7 compared to the wild type, with an activity optimum at pH 6. The relative activities of BAA 42 at pH 9 and 10 are about 1.5 and 5.7 times higher than those of the wild type, resulting in a broader pH activity profile for BAA 42. In contrast, the pH activity profile of BAA 29 is very similar to the wild-type profile with a pH optimum of pH 6 and only a marginally lower relative activity at alkaline pH.

Figure 6.

pH activity profiles. Activity measurement was carried out using the following 50-mM buffers: pH 4–6 acetate, pH 7–8 Tris/HCl, pH 9–10 glycine NaOH, pH 11 carbonate. (squares) WT, (circles) BAA 29, (×) BAA 42. The relative activity is the ratio of the catalytic activity at a certain pH and the maximum activity of each enzyme.

Characterization of the mutants: Sequence analysis

DNA sequencing of BAA 42 revealed five mutations (L13P, W194R, S197P, E356D, and N414S), which lead to a change in the amino acid sequence. Three of these mutations, W194R, E356D, and N414S were inherited from the error-prone PCR-derived clones BAA 18, BAA 19, and BAA 1, respectively (Fig. 7 ▶). The mutation L13P and the mutation S197P were acquired during the process of recombination. The mutation L13P is located in the signal peptide of the α-amylase. The mutations W194R, S197P, E356D, and N414S lie within the mature α-amylase. Comparison with the amino acid sequences of other bacterial α-amylases, namely the α-amylases of Bacillus licheniformis (BLA), Bacillus megaterium (BMA), Bacillus sp. KSM-1376 (LAMY), Bacillus sp. #707 (S707), Bacillus stearothermophilus (BstA), and Bacillus sp. TS-23 (TS-23), showed (Fig. 8 ▶) that the mutations W194R and S197P are located in a highly conserved region. In fact, W194 is present in every sequence of the compared α-amylases. In the case of S197, only BMA possesses a glycine at this position. Both mutations are located on a solvent accessible loop in domain B of the BLA structure (Fig. 9 ▶).

Figure 7.

Genealogic tree of BAA 29 and BAA 42. New mutations of each round are annotated by an asterisk.

Figure 8.

Parts of a ClustalX sequence alignment of BAA, BAA 29, and BAA 42 with the α-amylases from B. licheniformis (Sibakov and Palva 1984; Stephens et al. 1984), B. megaterium (BMA; Metz et al. 1988), B. sp KSM-1376 (LAMY; Igarashi et al. 1998), B. sp #707 (S707; Tsukamoto et al. 1988), B. stearothermophilus (BStA; Nakajima et al. 1985), B. sp. TS-23 (TS-23; Lin et al. 1998). Mutated amino acids are printed in bold face.

Figure 9.

Mutations in the structure of BAA. The structure was yielded by homology modeling based on the structure of BLA (1bli) using the SwissModel server (http://www.expasy.ch/swissmod).

At position 356, a glutamate residue is present in the sequences of BAA and BMA, whereas in all other amylase sequences this position is occupied by an aspartate. Therefore, the mutation E356D represents a mutation toward a more highly conserved amino acid. The amino acid 356 is located in the core of domain A of the α-amylase structure, not being accessible to the solvent. At position 414, the asparagine is replaced by serine in mutants BAA42 and BAA 29. In the sequences of BLA, BMA, LAMY, S707, BstA, and TS-23, a conserved lysine is found in this position, which is located on the surface of one of the TIM-barrel forming α-helices of domain A.

In BAA 29, six amino acids were replaced: L13P, V32A, A230V, N297D, K406R, and N414S. From these, V32A, K406R, and N414S were inherited from the clones BAA 13, BAA 4, and BAA 1. N297D was adopted from BAA N297D over clone BAA 3. The mutations L13P and A230V were generated during the shuffling process. In addition to the L13P mutation in the signal peptide, mutation V32A is located at the N terminus of the mature protein. V32A converts the VNG-sequence into ANG, which is known to be processed by the E. coli signal peptidase with higher probability according to the von Heijne rule (von Heijne 1985).

In the mature α-amylase, BAA 29 contains the mutations A230V, N297D, K406R, and N414S, with the latter also being present in BAA 42. A230V is located on a β-strand in the B domain above the active site of the α-amylase, on the surface of the structure. Amino acid comparisons show that A230 is conserved in the amino acid sequences of the α-amylases BAA, BLA, BMA, LAMY, S707, BStA, and TS-23. Mutation K406R is also located in domain A, on the top of a loop, and is accessible to the solvent. Mutation N297D, which is located in an α-helix of domain A, was found to stabilize the local structure by means of a salt bridge (Bessler et al. 2000).

Discussion

To our knowledge, this article is the first on the successful improvement of the alkali activity of an α-amylase by directed evolution. Although some efforts have been done on changing the pH profile of the BLA by site-directed mutagenesis (Nielsen et al. 1999, 2001), only one example is known in the literature, where the activity of the BLA under acidic conditions has been increased (Shaw et al. 1999). The specific activity of the BLA has so far only been improved by site-directed mutagenesis (Nielsen et al. 1999, 2001).

We could achieve a high α-amylase activity in the periplasm and the effective secretion of the BAA in the medium by coexpression of BRP in E. coli XL1-Blue cells. This method also sped up sample preparation from 90 min to about 60 min, reducing the needed pipetting steps from four to one, and thus increasing the throughput. Together with the Phadebas assay modified for the use with microtiter plates, we developed a versatile assay system for α-amylases with a very low error and standard deviation. The variability of the experimental conditions allows the adaptation to a multitude of parameters. For example, by altering the reaction temperature, screening can be performed to search for thermophilic or cold-active α-amylases. The stability of the substrate also allows screening in the presence of detergents and other chemicals present in industrial applications.

The finding of false positives, that is, the identification of wild-type genes as positive clones, in screening is a common problem. This is mainly due to the complexity of both the expression system and the enzyme containing solution. It is known in the literature that although cultivated in the same microtiter plate under the same conditions, expression can differ significantly from well to well (Zhao et al. 1999).

On the other hand, by setting up our assay in two dimensions and defining a sharp region for wild-type BAA, we sucessfully identified positive mutants. The method of measuring several parameters and correlating the results has been successfully applied to other screening methods before (Ness et al. 1999; Morawski et al. 2000; Wintrode et al. 2000). Nevertheless, it seems essential to carry out rescreening of retransformed colonies in order to validate results.

A mutation rate of one per gene accompanied by an inactivation rate of 30% was sufficient to successfully generate improved mutants by error-prone PCR, although current theoretical investigations suggest that higher deactivation rates and therefore higher mutation rates provide better access to a higher diversity of improved mutants (Miura and Sonigo 2001).

The DNA shuffling of the BAA mutants obtained by error-prone PCR resulted in two mutants with markedly altered properties. However, the inactivation rate of almost 90% required the subsequent selection of active clones to keep the screening costs low. In the case of amylases, this can be easily done by prescreening on starch agar plates with KI.

The library size of 7200 clones for the error-prone PCR library appears to be smaller than the size of other libraries described in literature, which typically ranges from 10,000 to 30,000 (Cohen et al. 2001). Nevertheless, successful examples using libraries with similar size have been reported. For example, the stabilization of the horseradish peroxidase by directed evolution was carried out screening libraries of 8000 clones per generation or less (Morawski et al. 2001). Similarly, for DNA shuffling small libraries were also successfully screened for improved recombinants. For example, after shuffling of 26 different subtilisin genes, from a library of 654 active clones chimeras with improved properties were found (Ness et al. 1999).

Both mutant BAA 42 and BAA 29 carry mutations that were inherited from clones created by error-prone PCR as well as mutations created during DNA shuffling. Mutant L13P shows that mutations are introduced during any DNA shuffling step as it is not found in any error-prone clone, but in both mutants. Therefore, it had to be introduced during an early step of the shuffling process.

It is quite difficult to predict the influence of every single mutation found in BAA 29 and BAA 42. Although not proven, it appears likely that the amino acid substitutions found in both mutants (L13P, N414S), are not responsible for a change in the pH activity profile by themselves. Hence, the mutations W194R, S197P, and E356D of BAA 42 seem to be responsible for the change in the alkaliphilicity. They may lead in combination with N414S to the observed change in the pH activity profile. Furthermore, because both mutant proteins are found in higher concentrations in the periplasm than the wild type, L13P seems to improve the accumulation of the enzyme in the periplasmic space.

Looking at the structure of BAA, it is surprising that the mutations were found in domains A and B but not in domain C. It seems that domain C is more vulnerable toward negative mutations than domain A or B.

None of the mutations found in BAA 42 or BAA 29 have been previously described to have an influence on the pH activity profile or upon the specific activity (Nielsen et al. 1999, 2001).

Conclusions

The methods used in this article show the possibility of an efficient trimming of the BAA toward multiple parameters required for industrial applications. The assay used in this experiment allows for screening several parameters simultaneously, for example, thermal activity and thermal stability as well as solvent and salt-dependent parameters. In addition, it can be applied to every α-amylase containing solution within a pH range from pH 4 to pH 12. We were able to alter the pH activity profile resulting in a shift of the pH optimum by one pH unit to pH 7. In addition, we improved the specific activity of the BAA by the factor 1.5 (BAA 42) and 9.3 (BAA 29) as well as the expression and processing of the BAA. The comparison of the amino acid sequences of the mutants with other homologous α-amylases showed that mutation in highly conserved regions does not necessarily lead to inactivation. By combining mutagenesis and recombination with a multidimensional assay, several parameters can be improved in a few steps. Further rounds of mutagenesis and recombination can be used to further improve the BAA.

Materials and methods

Materials

All chemicals used were of reagent grade or higher. As long as not otherwise specified, chemicals were received from Sigma Chemie. Oligonucleotides were obtained from Sigma-Ark. Restriction enzymes, T4-DNA ligase and desoxynucleotides were purchased from MBI Fermentas. MAR5N50 filter microtiter plates and Centricons were obtained from Millipore.

Cloning of the BAA gene in pGPFE

The BAA gene in the cloning vector pGEM-Teasy was a gift from Henkel KGaA. Plasmid pG-PFE encoding an esterase from P. fluorescence under control of the rhamnose promoter was generously provided by Erik Henke (2001). A 1700-bp fragment containing the BAA gene (1545 bp) was cloned after BamHI-SpeI-digestion into the expression vector pCYTEX P1 cut with the same enzymes to give pCYTBAA using standard procedures (Sambrook et al. 1989). After introduction of a NdeI-site at the start codon of the BAA gene by site directed mutagenesis (Dalbadie-McFarland et al. 1982) using Pfu-Polymerase (Stratagene) and the primers 5′-GAGAGGGAGAGGACATATGATTCCAAAACG-3′ and 5′-CGTTTTGGAATCATATGTCCTCTCCCTCTC-3′, the resulting plasmid pCYTBAA-NDE was NdeI-PstI-digested and the BAA-containing DNA fragment was ligated into the NdeI-PstI site of pG-PFE giving pG-BAA. The correct orientation of the BAA gene and the correct sequence were determined by DNA sequencing.

Random mutagenesis and recombination

Libraries of BAA mutants were constructed using error-prone PCR and plasmid pG-BAA as described (Zhao et al. 1999): 10 μL 10× mutagenic buffer (70 mM MgCl₂, 500 mM KCl, 100 mM Tris pH 8, 0.1 (w/v) gelatin), 10 μL 10× mutagenic dNTP-mix (2 mM dGTP and dATP, 20 mM dCTP and dTTP), and 50 pmole of each primer BAF 5′-CTTAAGAAGGAGATATACATATG-3′ and BAR 5′-GCCAAAACAGAAGCTTGGCTGCAG-3′, 5–15 ng of template, 60 μM MnCl₂, and 5 U Taq polymerase (Qiagen) in a total volume of 100 μL. Thermal cycling parameters were: 1 min, 95°C; 30 sec, 95°C, 45 sec, 55°C, 30 sec, 72°C (30 cycles); 2 min, 72°C. The amplified product was purified using the QIaQuick PCR-Purification kit (Qiagen), digested with NdeI and PstI and ligated into the NdeI-PstI site of the dephosphorylated vector pG-PFE. For expression of the BAA variants, E. coli XL1-blue competent cells (Stratagene) were at first transformed with pJL3 (MoBi-Tech), a plasmid encoding the Bacteriocin release protein (BRP) gene using the protocols suggested by the manufacturer to give E. coli XL1-blue (pJL3). Subsequently, competent E. coli XL1-blue (pJL3) cells were transformed with the variants of the pG-BAA plasmid.

DNA shuffling

For DNA shuffling, the plasmid variants of pG-BAA from the error-prone PCR were isolated from positive clones and the BAA genes were amplified by using the primers 5′-gcaaaaacaggaag gcaaaatgccg-3′ and 5′-cctccgggccgttgcttcgcaacg-3′ in a PCR reaction containing 5–10 ng of template, 3 mM MgCl₂, 50 pmole of each primer, 0.2 mM of each dNTP, and 2.5 U of Taq polymerase in a total volume of 100 μL of the 1× buffer supplied by the manufacturer. The temperature program was: 240 sec, 95°C; 60 sec, 95°C, 60 sec, 55°C, 60 sec, 72°C (25 cycles); 240 sec, 72°C. The products were purified with the Wizard PCR Preps-Kit (Promega) and subjected individually to partial DNaseI digestion for 7 min at 15°C using 0.01 mU DNaseI (Boehringer, now Roche Diagnostics) in 25 mM Tris/HCl (pH 7.4) in the presence of 5 mM MnCl₂ in a total volume of 50 μL. Fragments of 50–200 bp of the different templates were extracted from 1% agarose gels using the QIaExII kit (Qiagen), and mixed in equimolar amounts. For the reassembly, a PCR-reaction without primers containing 10 μL of fragment mix, 3 mM MgCl₂, and 2.5 U of Taq polymerase in a total volume of 50 μL was performed. PCR program: 90 sec, 95°C; 30 sec, 95°C; 30 sec, 60°C; 30 sec, 55°C; 60 sec + 5 sec/cycle (35 cycles), 72°C; 300 sec, 72°C. The full-length product was amplified from the reassembly mixture using the primers BAF and BAR under the same conditions as the first PCR and cloned into pG-PFE as described above. Clones producing active α-amylase were prescreened by plating on LB-Amp-agar plates containing 1% of soluble starch, and staining with a saturated aqueous solution of I₂/KI for 10 sec. Positive clones were identified by the formation of a clear halo against the violet background.

Microtiter plate assay

Colonies from LB-agar plates with ampicillin and chloramphenicol were isolated by a picking robot (Biorobotics) into 96-well microtiter plates containing 150 μL LB supplemented with 100 μg/mL ampicillin and 30 μg/mL chloramphenicol (LB_{amp, cm}) and were grown at 37°C for 16 h. After this period, 50 μL 50% sterile glycerol were added to each well and the microtiter plates were stored at −80°C. For expression cultivation, microtiter plates containing 200 μL LB_{amp, cm} supplemented with 0.2% rhamnose and 5 μM IPTG were inoculated from the glycerol stocks and cultivated for 42 h at 37°C in a fully equilibrated incubator. One freeze thaw cycle (−196°C, 5 min; 37°C, 60 min; 0°C, 15 min) followed by centrifugation (910 × g, 15 min, 15°C), yielded the enzyme containing supernatant.

Screening for increased alkali activity was carried out by measuring the activity at pH 10 and pH 7. For measurements at pH 10, 4 Phadebas tablets (Pharmacia and Upjohn) were suspended in 30-mL double-distilled water and filtered over a fritted glass filter (pore size 0 or 1). The retentate was resuspended in 32 mL 0.1 M glycine/NaOH buffer (pH 10). For measurements at standard pH (pH 7), 4 Phadebas tablets were suspended in 32-mL double-distilled water. One hunderd fifty microliters of the well-mixed suspension of the appropriate pH were added to each well of a MAR5N50 filter microtiter plate by means of an eight-channel multipette. Fifty microliters of enzyme solution were added to each well using a Biomek 2000 (Beckman Coulter) pipetting robot. After incubation for 15 min at 37 °C in a fully equilibrated incubator and centrifugation at 910 xg for 15 min at 15°C, the absorption of the flow through was measured at a wavelength of 620 nm in a Fluorstar microtiter plate reader (BMG Labtechnologies).

The standard curve at pH 7 was obtained by measuring the α-amylase activity of different dilutions of the commercial α-amylase BAN240L, which was generously provided by Novo Nordisk.

Cuvette assay

α-Amylase activity was routinely measured using a modified protocol of the Phadebas assay. One tablet was suspended in 10-mL double-distilled water. Five hundred microliters of the resulting suspension were transferred to a microfuge tube, 50 μl of the appropriately diluted enzyme (<1000 U/L) were added and incubated for 15 min at 37°C. The reaction was stopped with 150 μL 0.5 M NaOH and the remaining substrate was removed by centrifugation (20,000 × g, 15 min). Five hundred microliters of the supernatant were mixed with 500 μL of double-distilled water in a microcuvette before the absorption (620 nm) was measured. pH activity profiles from pH 4 to pH 11 were performed using the corresponding 50 mM buffer (pH 4–6 acetate, pH 7–8 Tris/HCl, pH 9–10 glycine/NaOH, and pH 11 carbonate) and appropriately diluted α-amylase containing solution (<1000 U/L).

Isolation from periplasm and protein purification

For the isolation of the recombinant α-amylase, 50 ml LB_{amp, cm} medium were inoculated 1:100 from an overnight culture and incubated at 37°C and 200 rpm until OD₆₀₀ = 0.4. Protein expression was induced with 0.2% rhamnose. Cells were harvested 3 h after induction by centrifugation (3200 × g, 15 min, 15°C) and the pellets were subjected to a cold osmotic shock to release the proteins from the periplasm (Neu and Heppel 1965).

The purification was done by adsorption of the α-amylase to starch (Candussio et al. 1990), and each step was carried out at 4°C (including solutions and instruments). Periplasmic fractions were mixed 1:10 with 0.1 M glycine/NaOH buffer pH 9. After the precipitate was removed by centrifugation, the supernatant was brought to a final concentration of 0.1 M Tris/HCl (pH 7), 20 mM CaCl₂, and 20% saturation of (NH₄)₂SO₄. Soluble potato starch was added to a final concentration of 3% and adsorption was performed in an overhead shaker (30 rpm, 360°) for 3 h. The pellet was removed by centrifugation (3300 × −g, 15 min, 4°C) and the supernatant was treated with 3% soluble potato starch for another 1 h. The pellets were combined and washed with one-half of the initial volume of a 0.1 M Tris/HCl buffer pH 7.5 containing 20% saturated (NH₄)₂SO₄ and 1 M NaCl. Desorption from the starch was achieved by treating the pellets with one-half of the initial volume of a Tris/HCl buffer (0.1 M, pH 7.5) containing 3 M NaCl and 0.1 M maltose for 90 min. After centrifugation (3300 × g, 15 min, 4°C) the pellet was again treated using the same buffer. The supernatants were combined, concentrated 10-fold using a Centricon (30-kD cutoff) and desalted using a PD-10 column (Amersham Pharmacia).

The homogeneity of the enzyme solution was checked by SDS-PAGE according to Laemmli (1970). For storage, 0.025% Brij 35 and 20 mM CaCl₂ were added to the enzyme preparation. Protein concentration was determined using a commercial BCA-assay (Pierce Chemical Company) using BSA as standard.

Computer modeling methods

The theoretical structure of BAA was obtained by homology modeling from the SwissModel server (Peitsch et al. 1995, 1996; Peitsch 1996; Guex and Peitsch 1997) with the structure of BLA (1bli.pdb) as a template, which was received from the Protein Databank (PDB; Bernstein et al. 1977). Amino acid mutations were inserted to the structure using the mutate tool of the SwissPDB-Viewer, followed by the side chain reconstruction for neighbored amino acids and energy minimization.

Abbreviations

• BAA, α-amylase from Bacillus amyloliquefaciens

• BLA, α-amylase from Bacillus licheniformis

• BstA, α-amylase from Bacillus stearothermophilus

• LAMY, α-amylase from Bacillus sp. KSM-1376

• S707, α-amylase from Bacillus strain #707

• TS-23, α-amylase from Bacillus sp. TS-23

• BRP, Bacteriocin release protein