A Systematic and Comprehensive Combinatorial Approach to Simultaneously Improve the Activity, Reaction Specificity, and Thermal Stability of p-Hydroxybenzoate Hydroxylase*

We have simultaneously improved the activity, reaction specificity, and thermal stability of p-hydroxybenzoate hydroxylase by means of systematic and comprehensive combinatorial mutagenesis starting from available single mutations. Introduction of random mutations at the positions of four cysteine and eight methionine residues provided 216 single mutants as stably expressed forms in Escherichia coli host cells. Four characteristics, hydroxylase activity toward p-hydroxybenzoate (main activity), protocatechuate-dependent NADPH oxidase activity (sub-activity), ratio of sub-activity to main activity (reaction specificity), and thermal stability, of the purified mutants were determined. To improve the above characteristics for diagnostic use of the enzyme, 11 single mutations (C152V, C211I, C332A, M52V, M52Q, M110L, M110I, M213G, M213L, M276Q, and M349A) were selected for further combinatorial mutagenesis. All possible combinations of the mutations provided 18 variants with double mutations and further combinatorial mutagenesis provided 6 variants with triple mutations and 9 variants with quadruple mutations with the simultaneously improved four properties.

We have simultaneously improved the activity, reaction specificity, and thermal stability of p-hydroxybenzoate hydroxylase by means of systematic and comprehensive combinatorial mutagenesis starting from available single mutations. Introduction of random mutations at the positions of four cysteine and eight methionine residues provided 216 single mutants as stably expressed forms in Escherichia coli host cells. Four characteristics, hydroxylase activity toward p-hydroxybenzoate (main activity), protocatechuate-dependent NADPH oxidase activity (sub-activity), ratio of sub-activity to main activity (reaction specificity), and thermal stability, of the purified mutants were determined. To improve the above characteristics for diagnostic use of the enzyme, 11 single mutations (C152V, C211I, C332A, M52V, M52Q, M110L, M110I, M213G, M213L, M276Q, and M349A) were selected for further combinatorial mutagenesis. All possible combinations of the mutations provided 18 variants with double mutations and further combinatorial mutagenesis provided 6 variants with triple mutations and 9 variants with quadruple mutations with the simultaneously improved four properties.
As recently reported, our evolutional design method, namely, "quasi-additive adaptive walking with mutant data base" (QAW), 2 provides a systematic and comprehensive combinatorial scheme for obtaining an improved protein with a single desired property (1). Because the previous work focused on a single property (enzyme activity), whether or not our QAW method could be extended to improve simultaneously multiple properties of any enzyme should be examined for generalization and availability as to industrial uses of our method. In a test study, we used p-hydroxybenzoate hydroxylase (PHBH) as a target enzyme and all Cys and Met residues as mutation sites, similar to the previous study (1).
Among the 20 naturally occurring amino acids, only Cys and Met residues contain a sulfur atom in their side chains. Because of the high reactivity of sulfur in sulfhydryl and thiomethyl groups, proteins are easily oxidized and show molecular heter-ogeneity because of the formation of inter-and intra-molecular disulfide linkages, and changes in isoelectric points caused by the formation of methionine sulfoxide. Therefore, sulfur-free or reduced proteins should be resistant to oxidative damage and could be used in highly oxidizing environments such as those found in pollution treatment facilities.
PHBH, which is a FAD-containing monooxygenase, catalyzes the hydroxylation of p-hydroxybenzoate to protocatechuate (main activity) ( Fig. 1A) (2,3) and plays an important role in degrading various aromatic compounds in the environment (4). The enzyme also shows protocatechuate-dependent NADPH oxidase activity (sub-activity), with the mechanism being unknown ( Fig. 1B) (5). The genes (pobA) encoding PHBH from various microorganisms have been sequenced (6 -35). The crystal structure and function of PHBH have provided information on the biological evolution of PHBH-like structure and flavin-containing aromatic compound oxygenases (36 -39). PHBH has also been a subject of dissection and reconstruction for the development of general catalysts with oxygen and electronic transfer.
From a clinical viewpoint, various assay methods for serum cholinesterase activity for the diagnosis of a liver function have been proposed. One popular method includes determination of p-hydroxybenzoate, the reaction product of cholinesterase, using PHBH (40). However, the sub-activity of the enzyme reduces the accuracy of the method. Thus, raising of the reaction specificity by simultaneously raising the main activity and reducing the sub-activity are desired as well as protein stability, as usually desired for protein engineering.
Here we report an attempt at simultaneous improvement of these properties of PHBH from Pseudomonas fluorescens NBRC 14160 by means of combinatorial searches starting from a Cys and Met single mutant pool. A number of improved mutants were effectively obtained after seven combinatorial searches, four with one selection parameter, two with two selection parameters, and one with four selection parameters. In addition, comparison of various combinatorial conditions was carried out.

EXPERIMENTAL PROCEDURES
Materials-The PCR, DNA extraction, DNA purification, DNA sequencing, and nickel-nitrilotriacetic acid protein purification kits were purchased from Qiagen; the protein extraction kit was from Novagen; and S-Gal/LB Agar Blend was purchased from Sigma. Sodium p-hydroxybenzoate was from Fluka, and other aromatic carboxylic acid compounds were from Wako Pure Chemical Industries. The following PCR primers were obtained from Jbios.
Overexpression-To isolate the gene encoding PHBH (pobA), chromosomal DNA from P. fluorescens NBRC 14160 (DDBJ/ DAD accession numberBAB20910) was used as the template (20). To generate a 1297-bp fragment containing the respective pobA sequence with an EcoT22I-restriction site at the N terminus and a His 6 tag and EcoRI restriction site at the C terminus, primers SD/EcoT22I/pobA-N and pobA/H6/EcoRI-C were used. A 1238-bp fragment was amplified by PCR with 35 cycles of melting at 94°C for 60 s, annealing at 55°C for 60 s, and comprising extension at 72°C for 90 s. The amplified DNA fragment was purified and then used as the template for the following PCR. To obtain a 1278-bp fragment containing the P35 and SD regions upstream of pobA, primers P35/SD/ EcoT22I-N and pobA/H6/EcoRI-C were used. PCR and purification of the resulting fragment were performed as described above. To generate a 1307-bp fragment containing a PstI restriction site upstream of the P35 region, PCR was carried out with the 1278-bp fragment as the template and primers PstI/ P35-N and pobA/H6/EcoRI-C. The 1307-bp fragment was purified, digested with restriction enzymes (PstI and EcoRI), and then repurified. The digested fragment was ligated into similarly digested pUC19, transformed into Escherichia coli JM109 cells by means of heat shock, and then plated on LB agar plates containing ampicillin and p-hydroxybenzoate. Transformants exhibiting the formation of a violet color because of the production of protocatechuate were isolated as single colonies. DNA was isolated from the resulting transformants and sequenced using a Big Dye sequencing kit. The plasmid containing pobA with the three restriction sites (PstI, EcoT22I, and EcoRI), two regions (P35 and SD), and His 6 tag was designated as pPHBH.
Mutant Libraries-To create libraries of single mutants as to various Cys and Met positions, the pPHBH plasmid was used as a PCR template. To produce a fragment of C152X-A, two primers (PstI/P35-N and C152X-C) were used, and to produce a fragment of C152X-B, two other primers (C152X-N and pobA/ H6/EcoRI-C) were used. Amplified DNA fragments C152X-A and C152X-B were purified and used as templates for a primerless PCR. After 10 cycles of the same condition, two primers (P35/SD/EcoT22I-N and pobA/H6/EcoRI-C) were added for an additional 35 cycles to enhance amplification under the complete 1238-bp fragment. The 1238-bp fragment containing C152X mutations was purified and digested with restriction enzymes (EcoT22I and EcoRI). For functional screening, the digested fragment was ligated into a vector (pPHBH digested at EcoT22I and EcoRI sites). Transformation, overexpression, plasmid extraction, and identification of C152X mutations were carried out as described above. To obtain libraries of double variants, for example, M98V/M110S, the M98V single mutant, as a template, two primers (P35/SD/EcoT22I-N and M110S-C) for M98V/M110S-A, and two primers (M110S-N and pobA/H6/EcoRI-C) for M98V/M110S-B, respectively, were used to generate M98V/M110S.
Characterization of Enzymes-E. coli JM109 cells were transformed with the plasmid containing mutations in pobA, and a single colony was used to inoculate 200 ml of LB medium containing ampicillin. After overnight incubation, the cells were harvested by centrifugation, and the cell pellet was lysed with phosphate buffer. The cell-free supernatant was purified using His 6 tag and a nickel-nitrilotriacetic acid protein purification kit (20). PHBH activity was monitored by following the rate of p-hydroxybenzoate-dependent oxidation of NADPH at 340 nm at 25°C using as the assay mixture the following: 100 mM HEPES/NaOH buffer (pH 8.0) containing 0.2 mM sodium p-hydroxybenzoate (or an other aromatic compound), 0.2 mM NADPH, 2 M FAD, and a suitable amount of enzyme solution in a total volume of 2 ml (20). Thermal stability was calculated by measuring the remaining activity after the purified enzyme in 100 mM HEPES/NaOH buffer (pH 8.0) was left standing at 50°C for 30 min and then cooled.

RESULTS
A Strategy for a Combinatorial Method-In an attempt at simultaneous improvement of enzymatic properties using the QAW concept, a comprehensive combination with a mutant data base (1), we devised a strategy of combinatorial mutations starting from available single mutations. The combinatorial method for generating mutants used in this study does not use the combination of only the single best mutant at various positions. Rather, our method has the following key features: (i) mutants (if possible, three or more) are selected from a single and multiple mutant data base using various selection parameters; (ii) all combinations of the selected mutants are generated to construct a secondary mutant data base; and (iii) iterative selection, combination, and construction of subsequent mutant data bases are carried out until multiple mutants with the desired properties are obtained. Such combinatorial mutations as to various residues using multiple selection parameters can be referred to as multidimensional or matrix combination. It is also notable that the quadruple mutants were not obtained by chance from a random mutant library. It should be possible to reproducibly reconstruct any of the variants with improved properties, because all of the combinations were carried out based on the single and multiple mutant data bases.  mutants were examined as to three properties as follows: hydroxylase activity toward p-hydroxybenzoate (main activity; units/mg), protocatechuate-dependent NADPH oxidase activity (sub-activity; units/mg), and thermal stability (%). Subsequently, the ratio of sub-activity to main activity (reaction specificity; %) was calculated based on the main activity and sub-activity. Using these four properties, a data base of single mutants as to Cys and Met sites was constructed (supplemental Table 1). Fig. 3 shows the distribution of the 216 single mutants with each of the four properties. As shown in Fig. 3 (A-D), there was a tendency that the substitution of Cys and Met residues with aliphatic amino acids caused the main activity and sub-activity values of the single variants to increase; however, their replacement with charged ones caused them to decrease. As shown in Fig. 3 (E-H), the mean and standard variation (shown in parentheses) values for the main activity, sub-activity, reaction specificity, and thermal stability for the 216 single mutants were similar to the wild type values: 14.7 (2.3) units/mg, 0.734 (0.126) units/mg, 5.0% (0.7%), and 31% (7%), respectively. In particular, only 28% of the single mutants exhibited reaction specificity values lower than that of the wild type, indicating that the distortion values of distribution for reaction specificity were rather high.

Simultaneous Improvement of Properties of PHBH
A detailed study of the single mutants revealed a loose relationship between the main activity and sub-activity with a correlation coefficient of 0.648 (see supplemental Fig. 7 (A1)). As compared with the wild type, 50% of the 216 single mutants exhibited both lower main activity and lower sub-activity, 22% both higher main activity and higher sub-activity, 23% lower main activity and higher sub-activity, and only 5% higher main activity and lower sub-activity. This tendency suggested the difficulty of simultaneous alterations with increasing main activity and decreasing sub-activity.
Combination Using One Selection Parameter-Combination of single mutants was carried out using only one selection parameter, main activity. First, eight single mutants with the highest main activity (M98V, M110S, M346G, M346A, M346I, M346E, M349A, and M349P) were selected from among the single mutants (Fig. 4a). All possible combinations of the selected eight single mutants were performed, resulting in 11 active and stable double variants (Fig. 4b). Next, four double variants with the highest main activity were selected and comprehensively combined. This combination provided three active and stable triple variants ( In addition, all of the triple variants exhibited not only higher main activity but also higher thermal stability than the wild type (Fig. 4c). On the other hand, the reaction specificity values of the multiple mutants were not significantly different from that of the wild type. A combinatorial experiment using only one selection parameter (sub-activity, reaction specificity, or thermal stability) was performed (data not shown). The experimental results indicated the same tendency, i.e. only the property used as the selection parameter was improved.
Combination Using Two Selection Parameters-From the results of combination using one selection parameter, it was hypothesized that the combination of mutations causing both higher main activity and higher thermal stability should allow simultaneous and efficient improvement of two properties (main activity and thermal stability). Based on such a hypothesis, first, 27 single mutants were selected that showed both higher main activity and higher thermal stability as compared with the wild type. Next, from among the selected 27 single mutants, 4 single mutants with the highest main activity (M110S, M346G, M346E, and M349A) and 4 single mutants with the highest thermal stability (M52V, M98L, M213G and M276Q) were selected, respectively (Fig. 4d). Comprehensive combination of these selected 8 single mutations generated 15 active and stable double variants (Fig. 4e). From these double variants, the two with the highest main activity and the two with the highest thermal stability were selected, respectively. The selected four double variants were comprehensively combined, resulting in five active and stable triple and two quadruple variants (Fig. 4, f and g). Because all replacement positions in the two quadruple variants (M98L/M110S/M276Q/M346G and M98L/M110S/M276Q/M346E) were the same, no further combinations were possible. The maximal main activity and thermal stability values for the selected single mutants and the double, triple, and quadruple variants were 21.3 units/mg and 43% (M346G), 21.5 units/mg and 47% (M110S/M346G), and 23.5 units/mg and 51% (M98L/M110S/M276Q/M346G), respectively. In contrast, neither sub-activity nor reaction specificity was reduced, implying that only the two properties used as selection parameters were improved.
Alternatively, to improve the reaction specificity through simultaneous enhancement of main activity and sub-activity, the combination by using two selection parameters (main activity and sub-activity) was performed. Twelve single mutants with both higher main activity and lower sub-activity than the wild type were selected (C152V, C211I, C332A, M52V, M52Q, M110L, M110I, M110P, M213G, M213L, M276Q, and M349A) (Fig. 4h). The selected 12 single mutants were comprehensively combined, which generated 15 active and stable double variants showing both higher main activity and lower sub-activity than the wild type (Fig. 4i). From among the 15 double variants, the 3 with the highest main activity and the 3 with the lowest subactivity were selected, respectively. These six double variants were comprehensively combined, which produced one active and stable triple and eight quadruple variants (Fig. 4 (j and k)). These results obviously indicated that both the main activity and subactivity of multiple variants were simultaneously improved, and thus the reaction specificity was efficiently reduced.
Combination Using All Four Selection Parameters-The two results of combinatorial experiments using two selection parameters led the combination by using all four selection parameters (main activity, sub-activity, reaction specificity, FIGURE 4. Alteration of the four properties of the selected single mutants and double, triple, and quadruple variants obtained through the combinations. a-c, histograms of the properties of the selected single mutants and double and triple variants obtained on combination using one selection parameter (main activity), respectively. d-g, histograms of the properties of the selected single mutants, and double, triple, and quadruple variants obtained on combination using two selection parameters (main activity and thermal stability), respectively. h-k, histograms of the properties of the selected single mutants and double, triple, and quadruple variants obtained on combination using two selection parameters (main activity and sub-activity), respectively. l-o, histograms of the properties of the selected single mutants and double, triple, and quadruple variants obtained on combination using four selection parameters (main activity, sub-activity, reaction specificity, and thermal stability), respectively. Red, yellow, green, and blue show main activity, sub-activity, reaction specificity, and thermal stability, respectively. Black shows the wild type. and thermal stability). Eleven single mutants were selected (C152V, C211I, C332A, M52V, M52Q, M110L, M110I, M213G, M213L, M276Q, and M349A), which exhibited higher main activity, lower sub-activity, and higher thermal stability than the wild type (Fig. 4l). These selected single mutants were comprehensively combined, resulting in 18 active and stable double variants (Fig. 4m). From among the 18 double variants, the 3 with the lowest reaction specificity and the 3 with the highest thermal stability were chosen, respectively. Comprehensive combination of the six double variants produced six active and stable triple and nine quadruple variants (Fig. 4 (n and o)).  Table 2). The minimal sub-activity value (0.357 units/ mg) and the maximal thermal stability value (57%) were observed for the quadruple variants (M52Q/ M110L/C211I/M276Q and M52Q/ C152V/M213G/C332A), respectively (see supplemental Table 2). In particular, the quadruple variant (M52Q/ C211I/M213G/C332A) showed not only simultaneous but also well balanced improvement of the four properties as follows: main activity, sub-activity, reaction specificity, and thermal stability values of 131, 54, 40, and 171% as compared with those of the wild type, respectively (see supplemental Table 2).
Alteration of Catalytic Properties of Variants-The catalytic properties (K m values for p-hydroxybenzoate, K m values for NADPH, k cat values, and K d values against protocatechuate) of the single mutants and the double, triple, and quadruple variants selected through combination using four selection parameters were investigated (see supplemental Table 2). Fig. 5B shows the distribution of three properties (K m values for p-hydroxybenzoate, K m values for NADPH, and k cat values) of the variants in three-dimensional space. The tendency was observed that with the first combination both K m values for p-hydroxybenzoate and NADPH decreased but on the other hand, with the second combination, k cat values rather increased (see supplemental Fig. 8). These results suggested that the simultaneous alterations decrease in K m for p-hydroxybenzoate, decrease in K m for NADPH, and increase in k cat led to efficient improvement of the main activity. The quadruple variant (M52Q/C211I/M213G/C332A) exhibited increases in k cat /K m for p-hydroxybenzoate of 1.6fold and for NADPH of 1.5-fold, respectively. In addition, Fig.  5C shows the distribution of three properties (K d values against protocatechuate, sub-activity, and reaction specificity) of the variants. It was observed that K d increased, sub-activity decreased, and the reaction specificity decreased simultaneously. This tendency suggested that increasing K d led to a reduction in reaction specificity (see supplemental Fig. 9).
In addition, K d values against aromatic compounds such as benzoate, monohydroxybenzoates, and dihydroxybenzoates of the wild type and quadruple variant (M52Q/C211I/M213G/ C332A) were measured (see supplemental Table 3). It was observed that only K d values against p-hydroxybenzoate and 2,4-dihydroxybenzoate decreased.
Comparison of Naturally Evolving Substitutions and Variants Used for the Combination-To examine the prevalence of amino acids at the positions equivalent to the residues in PHBH, naturally occurring PHBH species were retrieved from the DDBJ (6 -35). Fig. 6 shows that among 38 naturally substituted examples (i.e. naturally occurring single mutations) as to 12 positions, the amino acids are 11 (Ala, Val, Leu, Ile, Ser, Thr, Gln, Phe, Tyr, His, and Arg) and that the most abundant amino acids are aliphatic or small/polar ones, except for Tyr-110.
As shown in Fig. 3 (E-H), the variation of each of the four properties among the 216 single mutants constructed in this study was wider than that of the naturally occurring single mutations. For these naturally occurring single mutants, the maximal and minimal values, respectively, were 19.0 and 10.4 units/mg for main activity, 1.07 and 0.42 units/mg for sub-activity, 6.0 and 3.3% for reaction specificity, and 47 and 11% for thermal stability. In comparison, for the 216 single mutants, the maximal and minimal values, respectively, were 21.3 and 7.9 units/mg for main activity, 1.21 and 0.42 units/mg for sub-activity, 7.5 and 3.3% for reaction specificity, and 50 and 2% for thermal stability, respectively. For example, there are only three naturally occurring single mutations at the Met-346 position, and these three single mutants (M346T, M346L, and M346F) exhibited lower main activity than the wild type. However, a single mutant (M346G) created in this study exhibited the highest main activity value (134% of the wild type) among the 216 single mutants (see supplemental Table 1).
In the 11 single mutants selected on combination using four selection parameters, there were only 5 naturally occurring single mutations (C211I, C332A, M52V, M110I, and M213L), and only 2 naturally occurring single mutations (C211I and C332A) were observed in the quadruple variants. These results suggested that it would be difficult to predict the effects of replacements from the results of phylogenetic and structural analyses of naturally occurring PHBHs.

Protein Design by Means of an Adaptive Walk Search in
Sequence Space-To demonstrate the applicability of our QAW strategy, an evolutional design approach, for simultaneous improvement of particular properties, PHBH was used as a model protein in this study to improve the characteristics in terms of main activity, sub-activity, reaction specificity, and thermal stability, using a diminished amino acid component such as Cys-and Met-reduced (41)(42)(43). The merit of the QAW strategy is the great reduction in the number (no more than 10 4 ) of mutants to be tested in sequence space, and therefore the method can be, in principal, applied to improve any particular characteristics (1).
In our experiments, the total number of variants tested was only 249, which resulted in eight quadruple variants with improved characteristics, where the size of the sequence space to be considered is 18 12 ϭ 1.2 ϫ 10 15 for constructing a complete library with mutations at four Cys and eight Met residues (44), i.e. an ϳ5 ϫ 10 12 -fold reduction in the number to be tested was attained in this study. Usually, iterative selection and combination can greatly reduce the redundancy of the sequence space (45)(46)(47)(48)(49). Based on the properties of all the purified single mutant proteins as to each Cys and Met site and assuming strict additivity, the number of mutants expected to be more active than the wild type in terms of the main activity value is estimated to be 8.9 ϫ 10 7 (ϭ 4 (number of mutants with higher main activity than the wild type at Cys-152) ϫ 2 (at Cys-158), ϫ 4 (at Cys-211), ϫ 3 (at Cys-332), ϫ 6 (at Met-52), ϫ 3 (at Met-65), ϫ 5 (at Met-98), ϫ 6 (at Met-110), ϫ 7 (at Met-213), ϫ 5 (at Met-276), ϫ 7 (at Met-346), and ϫ 7 (at Met 349)). The success rate is only 0.000007%. Similarly, the number of mutants expected to be improved as compared with the wild type in terms of the four characteristics (main activity, sub-activity, reaction specificity and thermal stability) is estimated to be only 863 (success rate Ͻ10 Ϫ10 % on the basis of the four characteris- tics), and worse it is impossible to investigate 8.9 ϫ 10 7 variants (on the basis of the main activity). These estimations are consistent with the following results. When we attempted to obtain a Cys-and Met-reduced PHBH with increased main activity, reduced sub-activity, reduced reaction specificity, and increased thermal stability as compared with those of the wild type by creating mutant libraries by random mutagenesis, before starting the experiments described in this paper, such variants could not been obtained. As mentioned above, the QAW strategy should be significantly potent and effective in cases where it is inefficient to construct mutant pools and to screen them by means of directed evolution strategies (1).
Through combination with four selection parameters, a single mutant (M110I) exhibited the highest main activity value among the selected single mutants, and only two double variants (M52Q/M110I and M110I/C152V) could be generated. However, M110I/C152V showed lower main activity than M110I although M52Q/M110I indicated higher main activity than M110I (see supplemental Table 2). Similarly, although a single mutant (M52V) exhibited the highest thermal stability among the selected single mutants, a double variant (M52V/ C211I), which was the only generated variant containing the M52V-mutation, exhibited lower thermal stability than M52V (see supplemental Table 2). These results indicated obviously that combinations of the best mutations as to each site does not necessarily generate the best multiple variants.
To assess the interactions among individual single mutations observed in the quadruple mutant, M52Q/C211I/M213G/ C332A, the four properties of four single mutants (M52Q, C211I, M213G, and C332A), four double mutants (M52Q/ C211I, M52Q/M213G, C211I/M213G, and M213G/C332A), and one quadruple mutant (M52Q/C211I/M213G/C332A) were compared with each other (see supplemental Table 2). The main activity values of the single mutants decreased in the following order: C332A Ͼ M52Q Ͼ M213G Ͼ C211I. On the other hand, the order of the main activity values for the double mutants containing M213G was M213G/C332A Ͼ C211I/ M213G Ͼ M52Q/M213G (i.e. C332A Ͼ C211I Ͼ M52Q). In addition, the thermal stability values of the single mutants decreased in the following order: M213G Ͼ M52Q Ͼ C211I Ͼ C332A. However, the order of the thermal stability values for the double mutants containing M213G was M213G/C332A Ͼ M52Q/M213G Ͼ C211I/M213G (i.e. C332A Ͼ M52Q Ͼ C211I). The orders vary so that there is quasi-additivity of enzymatic properties (1). Hence, it should be emphasized that the QAW method is able to lead to more improved enzymes than those obtained by using the replacements giving the best improvement in the single mutants, where the simple additive condition is not applicable.
When we tried to construct a Cys-and Met-free or reduced PHBH with improved characteristics compared with the wild type based solely on structural information on each amino acid, what was produced? If one makes rational replacements of the sulfur-containing amino acids with non-sulfur amino acids, Cys residues would be generally replaced by Ala or Ser, with the same approximate shape and molecular weight, and Met residues would be substituted with Leu, Ile, or Lys residues having a similar shape to Met. As far as the four Cys residues at sites 152, 158, 211, and 332 of PHBH are concerned, van Berkel and co-workers (50,51) previously replaced these Cys residues by Ser to have similar kinetic and stable properties to those of the wild type PHBH. In this study, Val and Ala replaced the Cys residues at the 152-, 211-, and 332-sites of PHBH, respectively; Gln/Val, Leu/Ile, Gly/Leu, and Gln/Ala replaced the Met residues at the 52-, 110-, 213-, 276-, and 349-position, respectively (see supplemental Fig. 8). These residues are rather different from those expected to be chosen as to the similarity of the side of chain shape.
van Berkel et al. (50) reported that replacement of Cys at the 158-position with Ser caused a remarkable decrease (Ͼ10-fold) in the K m value for FAD, suggesting that mutations at the Cys-158 position would not help improve the properties of PHBH. Indeed, this is well consistent with the fact that C158X single mutants were not selected in this work because no single mutations at the Cys-158 position caused improvement of any of the properties. In other words, through iterative selection and combination, single mutants and multiple variants with either insignificant or negative changes were excluded. Finally, among the 12 Cys and Met sites, the replacement positions numbered 8 (Cys-152, Cys-211, Cys-332, Met-52, Met-110, Met-213, Met-276, and Met-349) and seven (Cys-152, Cys-211, Cys-332, Met-52, Met-110, Met-213, and Met-276), respectively, in the selected single mutants and the obtained quadruple variants through combination using four selection parameters. Regarding useful systematic improvement of a target enzyme, the combinatorial method makes it possible to carry out effective adaptive walking while reducing the sequence space that must be searched (1).
In addition, the effects of the selection parameters on improvement of the properties of multiple variants were discussed. Among the quadruple variants obtained on combination using two selection parameters (main activity and subactivity) and four selection parameters, the maximal or minimal characteristic values were as follow: highest main activity, 19.3 and 21.3 units/mg; lowest sub-activity, 0.357 and 0.357 units/ mg; lowest reaction specificity, 2.1 and 1.9%; and highest thermal stability, 48 and 57%, respectively (Fig. 4). This showed that increasing the number of selection parameters would result in improvement for better enzymatic properties. In other words, diversifying the combination by increasing the selection parameters would lead to an unexpected effect on properties in a limited sequence space (52,53).
Improvement of Flavoprotein Aromatic Compound Hydroxylase-In this study, by applying our QAW methods, an improvement of the properties of PHBH was demonstrated, although we did not know the manner of how replacements at these Cys and/or Met sites can influence such properties. All Cys and Met residues in the wild type PHBH are distant from the substrate-or NADPH-binding sites (Fig. 2). From a general evolutionary viewpoint, it is noteworthy that the substitutions of amino acids leading to the enhanced properties are not necessarily dependent on the amino acid residues involved in the catalytic activity and/or structural conformation. Indeed, it was recently reported that many protein functions are not confined to a small set of amino acids but rather can be affected by residues far away from the enzyme active site (54). These results suggest that the QAW method must be useful for searching for mutants with more improved functions with poor structural information on the target enzyme (1).
PHBH is an archetypal enzyme for type of NADPH/NADHdependent flavoprotein aromatic compound hydroxylases, which include, for example, salicylate hydroxylase, m-hydroxybenzoate hydroxylase, 6-hydroxynicotinic acid monooxygenase, phenol hydroxylase, and kynurenine 3-monooxygenase (39). These flavoprotein aromatic hydroxylases can act on various xenocompounds, and there is growing interest in the catabolic versatility of aromatic compound-degrading microorganisms growing in extreme environments (55)(56)(57)(58)(59). To exploit such enzymes for industrial processes such as the utilization of biosensors, production of fine chemical products, and application to bioremediation, considerable effort is being made (60 -62). Among these subjects, of importance are those mutations that broaden or alter the substrate specificity through several evolutionary strategies (63,64). Hence, the construction of single and multiple mutant data bases of PHBH by means of multidimensional combination on the basis of the QAW strategy would provide useful information in terms of simultaneous improvement of substrate/reaction specificity and other characteristics (65,66).