Active-Site Engineering of ω-Transaminase for Production of Unnatural Amino Acids Carrying a Side Chain Bulkier than an Ethyl Substituent

Abstract

ω-Transaminase (ω-TA) is a promising enzyme for use in the production of unnatural amino acids from keto acids using cheap amino donors such as isopropylamine. The small substrate-binding pocket of most ω-TAs permits entry of substituents no larger than an ethyl group, which presents a significant challenge to the preparation of structurally diverse unnatural amino acids. Here we report on the engineering of an (S)-selective ω-TA from Ochrobactrum anthropi (OATA) to reduce the steric constraint and thereby allow the small pocket to readily accept bulky substituents. On the basis of a docking model in which l-alanine was used as a ligand, nine active-site residues were selected for alanine scanning mutagenesis. Among the resulting variants, an L57A variant showed dramatic activity improvements in activity for α-keto acids and α-amino acids carrying substituents whose bulk is up to that of an n-butyl substituent (e.g., 48- and 56-fold increases in activity for 2-oxopentanoic acid and l-norvaline, respectively). An L57G mutation also relieved the steric constraint but did so much less than the L57A mutation did. In contrast, an L57V substitution failed to induce the improvements in activity for bulky substrates. Molecular modeling suggested that the alanine substitution of L57, located in a large pocket, induces an altered binding orientation of an α-carboxyl group and thereby provides more room to the small pocket. The synthetic utility of the L57A variant was demonstrated by carrying out the production of optically pure l- and d-norvaline (i.e., enantiomeric excess [ee] > 99%) by asymmetric amination of 2-oxopantanoic acid and kinetic resolution of racemic norvaline, respectively.

INTRODUCTION

Unnatural amino acids are widely used as essential chiral building blocks for diverse pharmaceutical drugs, agrochemicals, and chiral ligands (1–3). In contrast to natural amino acids, a fermentative method for the production of unnatural amino acids is not yet commercially available (4, 5). This has driven the development of various biocatalytic approaches to afford scalable processes for preparing enantiopure unnatural amino acids. These processes include kinetic resolution of racemic amino acids using acylase (6, 7), amidase (8, 9), hydantoinase (10, 11), and amino acid oxidase (12, 13) and asymmetric reductive amination of keto acids using dehydrogenase (14, 15) and transaminase (16, 17). Asymmetric amination is usually favored over kinetic resolution because the former affords a 100% yield without racemization of an unwanted enantiomer.

In contrast to a mandatory requirement for the supply and regeneration of an expensive external cofactor for the dehydrogenase reactions, transaminase catalyzes the transfer of an amino group from an amino donor to an acceptor using pyridoxal 5′-phosphate (PLP) as a prosthetic group (18). Nevertheless, industrial implementation of the transaminase-mediated synthesis of unnatural amino acids has lagged behind because of low equilibrium constants, usually close to unity, for the reactions between keto acids and amino acids. However, recent studies have demonstrated that the reductive amination of keto acids can be driven to completion without the thermodynamic limitation by employing cheap amines as a cosubstrate when the transfer of an amino group can be mediated by ω-transaminase (ω-TA), which utilizes primary amines as an amino donor (19–21). For example, the equilibrium constant for the amination of pyruvic acid by isopropylamine was reported to be 67 (22). Therefore, use of a 1.5 molar equivalent of isopropylamine relative to the concentration of pyruvic acid affords a 97% theoretical conversion of pyruvic acid to l- or d-alanine, depending on the stereoselectivity of ω-TA. Moreover, acetone (i.e., the deamination product of isopropylamine) is highly volatile, and the resulting equilibrium shift by facile evaporation can drive even thermodynamically unfavorable reactions to completion, as demonstrated elsewhere with the reductive amination of ketones (23–25). Another advantage of using isopropylamine as an amino donor is that most ω-TAs show low levels of activity for acetone (26), which minimizes enzyme inhibition by the ketone product. Taken together, ω-TA-catalyzed asymmetric amination of keto acids using isopropylamine is a promising strategy for the scalable production of unnatural amino acids.

One crucial bottleneck for synthesizing diverse unnatural amino acids using ω-TAs is the very narrow substrate specificity of the enzyme toward keto acids. It is generally accepted that both (R)- and (S)-selective ω-TAs possess two substrate-binding pockets consisting of a large pocket that is capable of dual recognition of hydrophobic and carboxyl groups and a small pocket that can accept substituents no larger than an ethyl group (27, 28). The canonical structural features of the active site of ω-TAs render a carboxyl group and a side chain of keto acid substrates bound to the large and small pockets, respectively. This leads all known ω-TAs to accept a limited range of α-keto acids, such as glyoxylic acid, pyruvic acid, and 2-oxobutyric acid. The only exception to the canonical substrate specificity to date is an (S)-selective ω-TA from Paracoccus denitrificans that can accommodate substituents whose bulk is up to that of an n-butyl substituent of α-keto acid in the small pocket (29). We demonstrated that a single point mutation in the small pocket (i.e., V153A) endowed the ω-TA with substantial activity toward even 2-oxooctanoic acid carrying an n-hexyl side chain (29). Despite the noncanonical steric constraint of the ω-TA from P. denitrificans, a low level of enzyme activity for isopropylamine renders a synthetic potential of the ω-TA less optimal for the practical amination of the bulky keto acids (21). This led us to set out to engineer the substrate specificity of the canonical ω-TAs showing a high level of activity for isopropylamine. To this end, we chose an (S)-selective ω-TA from Ochrobactrum anthropi (OATA), which showed 43% of the activity for isopropylamine relative to its activity for (S)-α-methylbenzylamine [(S)-α-MBA] (21).

To date, there have been three reports dealing with engineering of the canonical steric constraint in the small pocket of ω-TAs. In the first example, an (R)-selective ω-TA from an Arthrobacter sp. was successfully engineered to create a variant that can accept bulky arylalkyl ketones as an amino acceptor (23). In our previous study, the substrate specificity of the engineered variant, harboring 27 amino acid substitutions, was examined using structurally diverse α-keto acids (21). In spite of the remarkable improvements in activity for bulky arylalkyl ketones, the canonical steric constraint of a parental enzyme toward α-keto acids was not significantly altered in the engineered variant; e.g., the activity for 2-oxopentanoic acid was only 2% of that for pyruvic acid. This result suggests that engineering of the steric constraint should be guided in the context of a substrate type, especially substituents bound to the large pocket (i.e., hydrophobic versus carboxyl). The second example was carried out with an (S)-selective ω-TA from Vibrio fluvialis to improve the activity toward a β-keto ester [i.e., (R)-ethyl-5-methyl 3-oxooctanoate] (30). The resulting variant, carrying 8 amino acid substitutions, showed a 60-fold increase in activity for the target β-keto ester, but the substitutions led to a complete loss of activity for keto acids. The third example employed the same ω-TA as the one used in the second example and focused on improving activities for arylalkyl ketones. To the best of our knowledge, active-site engineering of a ω-TA to relieve the canonical steric constraint for α-keto acids and α-amino acids has not yet been reported.

In this study, we aimed to create a variant of OATA capable of accommodating bulky α-keto acids. To this end, we performed substrate docking simulations to select key active-site residues and carried out alanine scanning mutagenesis to identify a hot spot responsible for the narrow substrate specificity for α-keto acids. The resulting variant, carrying only a single point mutation in the large pocket, showed a desirable relaxation of the canonical steric constraint and afforded efficient stereoselective amination of bulky α-keto acids using isopropylamine as an amino donor.

MATERIALS AND METHODS

Chemicals.

Pyruvic acid was obtained from Kanto Chemical Co. (Tokyo, Japan). Isopropylamine was purchased from Junsei Chemical Co. (Tokyo, Japan). l-Alanine was purchased from Acros Organics Co. (Geel, Belgium). All other chemicals were purchased from Sigma-Aldrich Co. (St. Louis, MO). Materials used for the preparation of the culture media, including yeast extract, tryptone, and agar, were purchased from BD Biosciences (Franklin Lakes, NJ).

Site-directed mutagenesis of ω-TA.

Single point mutations of OATA were carried out using a QuikChange Lightning site-directed mutagenesis kit (Agilent Technologies Co., Santa Clara, CA) according to the guidelines in the instruction manual. The template used for mutagenesis PCR was pET28-OATA, which was previously constructed (31). Mutagenesis primers were designed by a primer design program (Agilent). The primer sequences are provided in Table 1. The intended mutagenesis was confirmed by DNA sequencing.

TABLE 1.

PCR primers used for preparing variants of OATA

Mutation	Orientation	Mutagenesis primersa
Y20A	Forward	5′-CGCGATATCCGTTATCATCTCCATTCTGCGACCGATGCTGTCCG-3′
	Reverse	5′-CGGACAGCATCGGTCGCAGAATGGAGATGATAACGGATATCGCG-3′
L57A	Forward	5′-CGAAGCGATGTCAGGAGCGTGGAGTGTTGGCGTG-3′
	Reverse	5′-CACGCCAACACTCCACGCTCCTGACATCGCTTCG-3′
L57V	Forward	5′-CGAAGCGATGTCAGGAGTGTGGAGTGTTGGC-3′
	Reverse	5′-GCCAACACTCCACACTCCTGACATCGCTTCG-3′
L57G	Forward	5′-TATCGAAGCGATGTCAGGAGGCTGGAGTGTTGGCGTGG-3′
	Reverse	5′-CCACGCCAA CACTCCAGCCTCCTGACATCGCTTCGATA-3′
W58A	Forward	5′-GCGATGTCAGGACTGGCGAGTGTTGGCGTGGG-3′
	Reverse	5′-CCCACGCCAACACTCGCCAGTCCTGACATCGC-3′
F86A	Forward	5′-AGATGAAGAAGCTGCCTTTCTACCATACAGCGTCCTACCGTTCGCAT-3′
	Reverse	5′-ATGCGAACGGTAGGACGCTGTATGGTAGAAAGGCAGCTTCTTCATCT-3′
Y151A	Forward	5′-CTCACGCAAGCGCGGCGCGCACGGTGTGACGATTG-3′
	Reverse	5′-CAATCGTCACACCGTGCGCGCCGCGCTTGCGTGAG-3′
V154A	Forward	5′-CGGCTATCACGGTGCGACGATTGCCTCTG-3′
	Reverse	5′-CAGAGGCAATCGTCGCACCGTGATAGCCG-3′
I261A	Forward	5′-TCTGCTGATCGCCGACGAGGTTGCGTGCGGCTTCGGA-3′
	Reverse	5′-TCCGAAGCCGCACGCAACCTCGTCGGCGATCAGCAGA-3′
F323A	Forward	5′-ACGCTTGGCACGGGCGCGACGGCATCTGGCCAT-3′
	Reverse	5′-ATGGCCAGATGCCGTCGCGCCCGTGCCAAGCGT-3′
T324A	Forward	5′-GGCACGGGCTTCGCGGCATCTGGCC-3′
	Reverse	5′-GGCCAGATGCCGCGAAGCCCGTGCC-3′

^aThe mutation sites are underlined.

Expression and purification of ω-TAs.

Overexpression of His₆-tagged ω-TAs was carried out as described previously with minor modifications (32). Escherichia coli BL21(DE3) cells carrying the expression vector [i.e., pET28a(+) harboring the ω-TA gene] were cultivated in LB medium (typically, 1 liter) containing 50 μg/ml kanamycin. Protein expression was induced by IPTG (isopropyl-β-d-thiogalactopyranoside) at an optical density at 600 nm of 0.4, and the cells were allowed to grow for 10 h. The culture broth was centrifuged, and the resulting cell suspension was subjected to ultrasonic disruption. Protein purification was carried out as described previously (32). Protein purity was confirmed by SDS-PAGE (see Fig. S1 in the supplemental material). The molar concentrations of the purified ω-TAs were determined by measuring the UV absorbance at 280 nm.

Enzyme assay.

All enzyme assays were carried out at 37°C and pH 7 (50 mM phosphate buffer). Unless otherwise specified, the reaction conditions for the activity assay were 10 mM (S)-α-MBA and 10 mM pyruvic acid. The typical reaction volume was 50 μl. The enzyme reaction was stopped after 10 min by adding 300 μl acetonitrile. The acetophenone produced from the reactions was analyzed by high-pressure liquid chromatography (HPLC). For initial rate measurements, reaction conversions were limited to less than 10%.

Substrate specificity.

To examine the substrate specificity of the OATA variants, initial rate measurements (i.e., <10% conversion) were independently obtained in triplicate. The reaction conditions for examination of the substrate specificity for amino acceptors were 20 mM α-keto acid and 20 mM (S)-α-MBA in 50 mM phosphate buffer (pH 7). The acetophenone produced was analyzed by HPLC. To measure activities for α-amino acids, 20 mM amino donor and 20 mM propanal were used as the substrates, and the α-keto acids produced were analyzed by HPLC. To measure amino donor activities for isopropylamine, reaction conditions were 20 mM isopropylamine and 20 mM pyruvic acid. l-Alanine was analyzed by chiral HPLC after derivatization with Marfey's reagent (33, 34).

Molecular modeling.

Using four X-ray structures of (S)-selective ω-TAs as the templates, an homology model of OATA was constructed by use of the Modeler module (version 9.8) of the Discovery Studio package (version 3.5.0; BIOVIA, San Diego, CA). The X-ray structures used as the templates were ω-TAs from P. denitrificans (PDB accession number 4GRX) (35), Chromobacterium violaceum (PDB accession number 4A6T) (36), Mesorhizobium loti (PDB accession number 3GJU) (37), and Rhodobacter sphaeroides (PDB accession number 3I5T) (37). The outward-pointing arginine of the structure with PDB accession number 4GRX was set to be conserved in the homology model, resulting in a dimeric structure of OATA in which each subunit has a different conformation of the active-site arginine. To construct a holoenzyme structure, the PLP moiety was copied from the structure with PDB accession number 4A6T. Only 0.4% nonglycine residues were found by a Ramachandran phi-psi analysis to lie in the disallowed region.

The active-site models of the L57A and L57G variants were built by amino acid substitution and then energy minimization (2,000 steps; dielectric constant = 4) of the mutation site, the internal aldimine, and the neighboring residues within 3 Å from L57 (i.e., S55, G56, L57, W58, S59, F82, H84, and T324) until the root mean squared gradient reached 0.1 kcal/mol/Å.

Docking simulations with l-alanine and l-norleucine as ligands were accomplished using the CDOCKER module under a default setting within the active site defined by the Binding-Site module. The most stable docking pose was chosen as a docking model.

Kinetic analysis.

A pseudo-one-substrate kinetic model was used to obtain apparent kinetic parameters for pyruvic acid and 2-oxohexanoic acid as described previously (38). The range of pyruvic acid concentrations used for initial rate measurements was 0.5 to 5 mM. The ranges of 2-oxohexanoic acid concentrations were 7 to 200 mM and 1 to 30 mM for the wild type and the L57A variant, respectively. The concentration of the cosubstrate [i.e., (S)-α-MBA] was fixed at 20 mM. The acetophenone produced was analyzed by HPLC to measure the initial rates. The initial rate data were fitted to a Michaelis-Menten equation, and the K_m and k_cat values were calculated from the slopes and y intercepts of the double-reciprocal plots.

Measurement of enzyme stability.

To examine the effect of the L57A mutation on enzyme stability, 10 μM enzyme was incubated in 50 mM potassium phosphate buffer (pH 7) at 37°C. Aliquots of the enzyme solution were sampled at predetermined incubation times and were mixed with reaction buffer containing 10 mM pyruvic acid and 10 mM (S)-α-MBA. Initial reaction rates were measured by analyzing the acetophenone produced. Inactivation constants were obtained by curve fitting of the residual activity data to a single exponential function.

Enzyme reactions to produce unnatural amino acids.

The reaction volume for the enzyme reactions was 1 ml, and the reaction mixture in 50 mM potassium phosphate buffer (pH 7) was incubated at 37°C. The reaction conditions for the asymmetric synthesis of l-norvaline were 50 mM 2-oxopentanoic acid, 100 mM isopropylamine, and 40 μM ω-TA. The synthesis of l-norleucine was carried out under the reaction conditions of 100 mM 2-oxohexanoic acid, 200 mM isopropylamine, 0.1 mM PLP, and 200 μM ω-TA. The kinetic resolution of racemic norvaline (rac-norvaline) was performed with 50 mM rac-norvaline, 50 mM glyoxylic acid, and 40 μM ω-TA. Aliquots of the reaction mixture (50 μl) were taken at predetermined reaction times and mixed with 10 μl HCl solution (5 N) to stop the reaction. The reaction mixtures were subjected to HPLC analysis for measurement of conversion and enantiomeric excess.

HPLC analysis.

All HPLC analyses were performed on a Waters HPLC system (Milford, MA). Analysis of acetophenone was performed using a Sunfire C₁₈ column (Waters Co.) with isocratic elution with 60% methanol–40% water–0.1% trifluoroacetic acid at 1 ml/min. UV detection was done at 254 nm. Quantitative chiral analyses of alanine, norvaline, and norleucine were carried out using a Crownpak CR(−) column (Daicel Co., Japan) or a Sunfire C₁₈ column after chiral derivatization with Marfey's reagent (33, 34). Κeto acids were analyzed using an Aminex HPX-87H column (Bio-Rad, Hercules, CA) with isocratic elution of a 5 mM H₂SO₄ solution at 0.5 ml/min. The column oven temperature was set to 40°C, and UV detection was done at 210 nm.

RESULTS AND DISCUSSION

Substrate docking.

We set out to construct a homology model of OATA to identify key active-site residues responsible for rejecting the entry of bulky substituents in the small pocket. The X-ray structures used as the templates for the homology modeling were ω-TAs from P. denitrificans (PDB accession number 4GRX; sequence identity, 41%) (35), C. violaceum (PDB accession number 4A6T; sequence identity, 41%) (36), M. loti (PDB accession number 3GJU; sequence identity, 43%) (37), and R. sphaeroides (PDB accession number 3I5T; sequence identity, 34%) (37). The alignment of the OATA sequence with the sequences of the template ω-TAs is shown in Fig. S2 in the supplemental material. The four template ω-TAs adopt a homodimeric structure where both active-site arginines assume an inward conformation (i.e., they point toward the large pocket), except for the structure with PDB accession number 4GRX, where one subunit harbors an outward arginine (i.e., it points toward a solvent side). It is known that the active-site arginine, responsible for recognition of a carboxylate of an incoming substrate, undergoes a gross movement and assumes the outward conformation when the large pocket is taken up by a hydrophobic substituent (39).

We performed substrate docking simulations using the subunit structure where the active-site arginine (i.e., R417) adopts the inward conformation (Fig. 1). For the docking simulation, we used l-alanine, instead of pyruvic acid, as a ligand because all the active-site lysines in the template ω-TAs form a Schiff base with PLP (i.e., an internal aldimine capable of deamination of an amino donor), and thereby, the resulting homology model adopts the internal aldimine structure. It is notable that the binding of l-alanine is coordinated by a hydrogen bond between the amino group of the substrate and the phenolic oxygen of the PLP moiety as well as multiple hydrogen bonds between the α-carboxylate and the active-site arginine. Compared to the active-site structure, where R417 adopts an outward conformation, the large pocket becomes contracted owing to the inward conformation of R417 to recognize the carboxyl group. The small pocket accommodating the methyl group of l-alanine is wrapped by side chains of the five active-site residues (i.e., Y20, F86, Y151, V154, and T324), a phosphate group of the PLP, and a backbone chain ranging from G322 to T324. The Connolly surface of the active site clearly explains why the small pocket can accept a substituent whose bulk is only up to that of an ethyl substituent.

FIG 1.

Model of l-alanine docking in the active site of OATA. l-Alanine is shown in Corey-Pauling-Koltun (CPK) representation. The internal aldimine, formed between PLP and K287, is shown as thick sticks. K287 is positioned right behind the bound substrate. The cyan dotted lines represent hydrogen bonds. The colors used for the active-site residues are consistent with the colors of the labels. The active site is visualized by use of a Connolly surface.

Alanine scanning mutagenesis of the active-site residues.

The docking model indentified 13 active-site residues that participate in the formation of the active-site surface and lie within a 7-Å distance from the bound l-alanine (i.e., Y20, L57, W58, F86, Y151, V154, A230, I261, K287, G322, F323, T324, and R417). We carried out alanine substitution of the active-site residues one by one to examine which residue could be reduced in size to lead to the generation of the additional room required for bulky substituents to be accepted in the small pocket. A230 and G322 were excluded from the mutagenesis because alanine substitution of the two residues would not generate more room in the active site. K287 and R417 were also excluded because these are essential residues responsible for catalytic turnover and recognition of a carboxyl group, respectively. Therefore, nine active-site residues were subjected to alanine scanning mutagenesis, and the enzyme activities of the resulting variants toward 2-oxopentanoic acid were measured (Fig. 2). Among the nine variants, an L57A variant showed an exceptionally improved activity for 2-oxopentanoic acid (i.e., a 48-fold increase in activity relative to that of the wild-type enzyme). This was an unexpected result because the L57 residue participates in the large pocket. Besides the L57A variant, the V154A variant was the only alanine substitution variant that allowed an improvement in activity for 2-oxopentanoic acid (i.e., a 3-fold increase). This result is in line with our previous observation that an alanine substitution at the same position of the ω-TA from P. denitrificans achieved excavation of the small pocket (29).

FIG 2.

Enzyme activities for 2-oxopentanoic acid of the alanine mutants that were scanned. The reaction rate represents the initial rate per 1 μM enzyme and was measured with 20 mM (S)-α-MBA and 20 mM 2-oxopentanoic acid. WT, wild-type OATA.

Site-directed mutagenesis of L57.

We presumed that the remarkable improvement in activity for 2-oxopentanoic acid achieved with the L57A mutation would result from altered binding of a carboxyl group owing to the generation of more room in the large pocket because of the reduction in the size of L57. To examine whether a greater improvement in activity could be induced by substitution of L57 with small amino acids other than alanine, we prepared L57V and L57G variants by site-directed mutagenesis. Enzyme activities for various α-keto acids were measured and compared with those of the wild-type enzyme and the L57A variant (Table 2). The wild-type OATA showed negligible activities (i.e., <1% relative to that for pyruvic acid) for bulky α-keto acids carrying a side chain larger than an ethyl group. The L57V mutation did not induce any improvement in activity for the bulky α-keto acids but caused drastic losses of activity toward native substrates (i.e., substrates whose bulks ranged from the bulk of glyoxylic acid to the bulk of 2-oxobutyric acid). In contrast, the L57A variant showed improvements in activity for α-keto acids carrying side chains larger than a methyl group (i.e., 2-, 48-, and 39-fold increases in activity for 2-oxobutyric acid, 2-oxopentanoic acid, and 2-oxohexanoic acid, respectively). The improvements in activity for bulky α-keto acids were so remarkable that the L57A variant exhibited substantial activity even for 2-oxooctanoic acid (i.e., 2% relative to that for pyruvic acid). The L57G substitution also induced improvements in activity for bulky α-keto acids but to much less of a degree than those observed with the L57A substitution. Despite the improvements in activity toward α-keto acids carrying bulky linear alkyl groups, neither the L57A variant nor the L57G variant displayed substantial activities for branched-chain α-keto acids, including 3-methyl-2-oxobutyric acid, 3-methyl-2-oxopentanoic acid, 4-methyl-2-oxopentanoic acid, and trimethylpyruvic acid (data not shown).

TABLE 2.

Effect of mutations of L57 residue on substrate specificity of OATA for α-keto acids

Substrate	Side chain	Reaction rate (μM/min)a
		Wild type	L57V variant	L57A variant	L57G variant
Glyoxylic acid	—H	618 ± 31	60 ± 2	354 ± 23	113 ± 12
Pyruvic acid	—CH3	478 ± 25	66 ± 12	367 ± 37	132 ± 6
2-Oxobutyric acid	—CH2CH3	69 ± 5	13 ± 3	161 ± 6	30 ± 2
2-Oxopentanoic acid	—(CH2)2CH3	4 ± 2	6 ± 1	190 ± 8	47 ± 4
2-Oxohexanoic acid	—(CH2)3CH3	2 ± 1	2 ± 1	77 ± 1	10 ± 1
2-Oxooctanoic acid	—(CH2)5CH3	NDb	ND	8 ± 4	ND

^aThe reaction rate represents the initial rate per 1 μM enzyme and was measured with 20 mM (S)-α-MBA and 20 mM α-keto acid in 50 mM phosphate buffer (pH 7).

^bND, not detectable (i.e., the reaction rate was <1 μM/min).

As shown in Table 3, the relaxed steric constraint for α-keto acids was also observed for α-amino acids (i.e., 56- and 5-fold improvements in the activities of the L57A and L57G variants, respectively, for l-norvaline). Despite the nondetectable activity of the wild-type OATA for l-norleucine, the relaxed steric constraint allowed the L57A and L57G variants to show substantial activities for l-norleucine. The L57V mutation led to drastic decreases in activity for α-amino acids, as was also observed with α-keto acids. Interestingly, all the OATA variants, in addition to the wild type, showed very low activities for glycine, despite the high levels of activity for the cognate keto acid. These extremely opposite activities for glycine and glyoxylic acid seem to be ascribable to the thermodynamically unfavorable conversion of glycine to its keto acid (40). As was observed with the branched-chain α-keto acids, both the L57A and the L57G variants showed nondetectable activities for α-amino acids carrying branched-chain substituents, including l-valine, l-leucine, l-isoleucine, and l-tert-leucine (data not shown).

TABLE 3.

Effect of mutations of L57 residue on substrate specificity of OATA for α-amino acids

Substrate	Side chain	Reaction rate (μM/min)a
		Wild type	L57V variant	L57A variant	L57G variant
Glycine	—H	18 ± 1	3 ± 2	9 ± 4	3 ± 2
l-Alanine	—CH3	559 ± 4	75 ± 5	203 ± 26	39 ± 5
l-Homoalanine	—CH2CH3	42 ± 4	2 ± 1	52 ± 1	9 ± 1
l-Norvaline	—(CH2)2CH3	2 ± 1	NDb	111 ± 7	14 ± 1
l-Norleucine	—(CH2)3CH3	ND	ND	26 ± 2	3 ± 1

^aThe reaction rate represents the initial rate per 1 μM enzyme and was measured with 20 mM amino acid and 20 mM propanal in 50 mM phosphate buffer (pH 7).

^bND, not detectable (i.e., the reaction rate was <1 μM/min).

The synthetic utility of the L57A variant for the asymmetric amination of bulky keto acids is contingent upon a high level of activity for isopropylamine as well as no loss of stereoselectivity in the parent. We measured the amino donor reactivity of isopropylamine with the L57 variants (Fig. 3). In contrast to the findings for the L57V and L57G variants, the L57A variant conserved 80% of the parental activity for isopropylamine. In addition to the wild-type OATA, all L57 variants showed the nondetectable formation of d-alanine during the transamination between isopropylamine and pyruvic acid (i.e., enantiomeric excess [ee] of l-alanine produced, >99.9%). This result indicates that the three amino acid substitutions of L57 do not affect the stringent stereoselectivity of the parental enzyme.

FIG 3.

Effect of mutations of the L57 residue on amino donor activity for isopropylamine. The reaction rate represents the initial rate per 1 μM enzyme and was measured with 20 mM isopropylamine and 20 mM pyruvic acid.

Structural modeling of the L57A variant.

To provide structural insight into how the L57A mutation relieves the steric constraint in the small pocket, we performed a docking simulation with the L57A variant using l-alanine as a ligand (Fig. 4A). Compared to the docking pose of l-alanine in the wild-type active site, the L57A mutation allowed docking of l-alanine slightly away from the small pocket with 0.71- and 0.73-Å translocations of the C-α and C-β of the substrate to the large pocket, respectively. These translocations are attributed to a 1.35-Å movement of the α-carboxyl carbon to the W58 residue owing to the room created by the L57A substitution. The altered binding orientation of the α-carboxylate in the large pocket permits the formation of an additional hydrogen bond with the indole group of W58, which leads to a concurrent loss of one of the preexisting hydrogen bonds with R417.

FIG 4.

Docking models of the L57A variant using l-alanine (A) and l-norleucine (B) as ligands. The ligands that docked in the active site, visualized by use of a Connolly surface, are represented by ball-and-stick models. Purple sticks and cyan dotted lines, the mutation site and the hydrogen bonds, respectively; yellow sticks in the binding pocket in panel A, the docking pose of l-alanine in the wild-type OATA, as shown in Fig. 1.

The altered binding of the α-carboxylate in the active site of the L57A variant led to the substrate docking closer to the large pocket, which provides more effective room to the small pocket even without any structural change in the small pocket. In the substrate specificity experiments, we could test neither α-keto acid nor α-amino acid substrates carrying an n-pentyl side chain because these substrates were not commercially available. To examine how large of a substituent that the extended small pocket can accommodate, we carried out a docking simulation using l-norleucine as a ligand (Fig. 4B). The findings obtained with the docking model are consistent with the significant enzyme activity for l-norleucine and predict that the small pocket cannot accept an n-pentyl group. The n-butyl group of the bound l-norleucine does not assume a staggered conformation due to a steric interference with the residues of the small pocket, resulting in an intramolecular steric strain of the bound substrate. This could explain the 87% loss of activity for l-norleucine compared to that for l-alanine.

Contrary to expectation, the L57G variant did not show higher activities toward bulky substrates than the L57A variant did. To examine how much the L57G mutation allowed the altered binding of the α-carboxyl group to the large pocket, we performed a docking simulation with l-norleucine in the active site of the L57G variant (see Fig. S3 in the supplemental material). The docking pose of l-norleucine in the L57G variant turned out to be very close to that generated with the L57A variant, indicating that creation of more room in the large pocket by the L57G mutation did not elicit a greater translocation of the α-carboxyl group toward the large pocket. This is in agreement with the findings obtained with the model of l-norleucine docking with the L57A variant, where the α-carboxyl group is already in a close contact with W58, and thereby, further translocation of the α-carboxyl group is prohibited because of a steric clash against W58 (Fig. 4B). It is presumed that the lower activities of the L57G variant for bulky substrates compared to those of the L57A variant result from an increased conformation flexibility of the backbone chain, which might be deleterious to a catalytic step after the formation of a Michaelis complex. This is supported by the drastic losses of activity of the L57G variant even for the native substrates, such as glyoxylic acid, pyruvic acid, and l-alanine. Note that two consecutive glycine residues happened to be created immediately before W58 in the L57G variant because the 56th residue is glycine (see Fig. S2 in the supplemental material).

Kinetic analysis of the L57A variant.

To provide a mechanistic understanding of how the L57A mutation improves activities for bulky substrates, we compared the kinetic parameters of the wild-type and the L57A variant for pyruvic acid and 2-oxohexanoic acid (Table 4). Compared with the specificity constant (i.e., k_cat/K_m) of the wild-type enzyme for pyruvic acid, the L57A variant showed a 50% reduction in the specificity constant. This loss of activity for the native substrate was caused by a 2.8-fold decrease in binding affinity, although the L57A mutation induced a 1.4-fold increase in the turnover number. The weakened binding to pyruvic acid by the variant with the L57A mutation seems to result from the loss of a native hydrogen bond between R417 and the α-carboxylate of pyruvic acid (Fig. 1 and 4B).

TABLE 4.

Kinetic parameters of the wild-type and the L57A OATA for α-keto acids^a

α-Keto acid	Wild type			L57A variant
	Km (mM)	kcat (s−1)	kcat/Km (M−1 s−1)	Km (mM)	kcat (s−1)	kcat/Km (M−1 s−1)
Pyruvic acid	0.12 ± 0.01b	2.6 ± 0.1b	22,000 ± 3,000b	0.34 ± 0.05	3.7 ± 0.2	11,000 ± 2,000
2-Oxohexanoic acid	35 ± 1	0.023 ± 0.001	0.66 ± 0.05	5.7 ± 0.3	1.1 ± 0.1	190 ± 20

^aKinetic parameters represent the apparent rate constants determined at a fixed concentration of (S)-α-MBA.

^bThese kinetic parameters were taken from a previous study (38).

As the size of the side chain of the α-keto acid substrate extends from that of a methyl group to that of an n-butyl group (i.e., from the side chain found in pyruvic acid to that found in 2-oxohexanoic acid), the wild-type enzyme showed 290- and 110-fold decreases in binding affinity and the turnover number, respectively. The deterioration in both the binding and the catalytic steps renders the wild-type enzyme nearly inactive toward 2-oxohexanoic acid. The L57A mutation turned out to promote both steps (i.e., 6- and 48-fold increases in binding affinity and catalytic turnover, respectively), leading to a 290-fold increase in the specificity constant for 2-oxohexanoic acid. It is notable that the fold change in k_cat is even larger than that in K_m, indicating that the L57A-induced alleviation of steric interference in the small pocket benefits the catalytic step much more than the binding step. This result is in accordance with the findings obtained with structural model of a quinonoid intermediate (i.e., the most unstable reaction intermediate), where formation of a covalent linkage between the substrate and PLP rearranges the substrate moiety so that it is close to the PLP side and thereby results in the steric interference of a bulky substituent in the reaction intermediate stronger than that in the Michaelis complex (38).

Synthetic utility of the L57A variant.

For the practical applicability of enzyme variants engineered to attain a desirable functionality, it is essential for the amino acid substitution to disturb the protein stability in a minimal way. We performed time course monitoring of the enzyme activities under the incubation conditions of 50 mM phosphate buffer (pH 7.0) and 37°C, which were the same as those used with the reaction mixture used for preparative purposes (Fig. 5). Intriguingly, we found that the L57A mutation remarkably increased the stability of the enzyme. The inactivation constants obtained from curve fitting of the residual activity data to a single exponential function were 1.1 × 10⁻² h⁻¹ (r² = 0.97, half-life = 62 h) and 8.1 × 10⁻⁴ h⁻¹ (r² = 0.89, half-life = 860 h) for the wild type and the L57A variant, respectively. The striking improvement in stability renders the L57A variant highly promising for use in industrial applications.

FIG 5.

Effect of the L57A mutation on enzyme stability. Purified enzymes were incubated in 50 mM potassium phosphate buffer (pH 7.0) at 37°C.

To examine how the L57A mutation improves the catalytic potential of OATA for the production of bulky unnatural amino acids, we carried out the asymmetric synthesis of l-norvaline, which is a key intermediate of perindopril (i.e., an angiotensin-converting enzyme inhibitor) (41) and a potential inhibitor of arginase (42). As expected, the L57A variant afforded a much faster synthesis of l-norvaline from 50 mM 2-oxopentanoic acid and 100 mM isopropylamine than its parental enzyme did under the same reaction conditions (Fig. 6A). The level of conversion reached 99.3% at 2 h, with the ee of the resulting l-norvaline obtained using the L57A variant being >99.9%, whereas the wild-type OATA permitted only 22.6% conversion at 2 h. As shown in Fig. 6B, we also performed the asymmetric synthesis of l-norleucine from 100 mM 2-oxohexanoic acid and 200 mM isopropylamine using an enzyme concentration 5-fold higher than that used in the assay whose results are presented in Fig. 6A. The L57A variant completed the reaction within 2 h (i.e., 99.1% conversion and a >99.9% ee of l-norleucine), while use of the wild-type OATA led to only 24.6% conversion at 2 h.

FIG 6.

Enzymatic reactions to produce unnatural amino acids by use of the L57A variant in comparison with the use of parental ω-TA. (A) Asymmetric synthesis of l-norvaline. Reaction conditions were 50 mM 2-oxopentanoic acid, 100 mM isopropylamine, and 40 μM ω-TA. (B) Asymmetric synthesis of l-norleucine. Reaction conditions were 100 mM 2-oxohexanoic acid, 200 mM isopropylamine, and 200 μM ω-TA. (C) Kinetic resolution of rac-norvaline. Reaction conditions were 50 mM rac-norvaline, 50 mM glyoxylic acid, and 40 μM ω-TA.

To demonstrate the catalytic utility of the L57A variant for the production of d-amino acids, we performed the kinetic resolution of racemic norvaline to prepare d-norvaline for use as a chiral building block of macrolide drugs, such as pamamycin-607 (43) and epilachnene (44) (Fig. 6C). Starting with 50 mM rac-norvaline and 50 mM glyoxylic acid using the same enzyme concentration as the one used in the reaction whose results are presented in Fig. 6A, the kinetic resolution was completed by the L57A variant within 5 h (i.e., 51.6% conversion and a 99.2% ee of d-norvaline). In contrast, the same reaction catalyzed by the wild-type enzyme led to only 0.8% conversion and a 2.5% ee at 5 h.

Conclusions.

To expand the synthetic utility of ω-TAs for the production of diverse unnatural amino acids via either asymmetric synthesis or kinetic resolution, it is indispensable to overcome the canonical steric constraint in the small pocket. To the best of our knowledge, this study is the first example of the engineering of the canonical substrate specificity of a ω-TA for α-keto acids and α-amino acids. Our results indicate that a single point mutation introduced in the large pocket rather than the small pocket could be more effective in relieving the steric constraint. However, further engineering of the substrate specificity toward more structurally demanding α-keto acids should include additional mutations in the small pocket of the L57A variant, e.g., the V154A substitution identified to be beneficial for relieving the steric constraint.