All chemicals were purchased from Sigma Aldrich, unless stated otherwise. All enzymes for the genetic work were purchased from ThermoFisher Scientific. The expression vector pEamTA was produced as stated before (Reisinger et al., 2007). The Escherichia coli strain BL21(DE3) was obtained from BioLabs (New England). Analytical HPLC was carried out with an Agilent 1,200 HPLC-MS system equipped with a 6,120 quadrupol mass spectrometer. UV-vis scans and single wavelength kinetics were measured on a Perkin Elmer Lambda 2 spectrophotometer.

The wild-type LTAaj (for sequence and alignment with PDB:3WGB template see Supplementary Information) was cloned in pEamTA vector with a C-terminal His-tag as previously described (Fesko et al., 2015). LTAaj mutants were generated by site-directed mutagenesis according to Agilent's QuikChange II protocol using PfuUltra High-Fidelity DNA Polymerase (Agilent Technologies). The pEamTA-LTAaj plasmid was used as the template. Primers were ordered at Integrated DNA Technologies. PCR was carried out under the following conditions: 50 μL total volume, 200 μM of each dNTP, 0.2 μM of each primer, 5 ng of template plasmid DNA, and 2.5 U of PfuUltra High-Fidelity DNA polymerase. Cycling conditions were as follows: initial denaturation at 95°C for 2 min, followed by 21 cycles of denaturation at 95°C for 30 s, annealing at 55°C for 60 s, extension at 68°C for 8 min. Amplification was controlled by electrophoresis of 10 μL of the PCR reaction on an agarose gel. After the digestion of parental non-mutated methylated plasmid DNA with DpnI at 37°C for 1 h, the samples were desalted and electro-competent Escherichia coli BL21(DE3) cells were transformed with the resulting nicked vector DNA containing the desired mutations. The introduction of mutations was confirmed by sequencing analysis (LGC genomics).

The constructs pEamTA-LTAaj-mut contained the tac promoter and the LacI repressor, which control the protein expression level. 300 mL 2xYT medium (in a 2 L flask) containing ampicillin (50 μg/mL) was inoculated with 5 mL of an E. coli overnight culture. Cells were grown at 37°C to an OD₆₀₀ of 0.6–0.8. Protein expression was induced by the addition of 1 mM IPTG and the culture was allowed to grow overnight at 20°C. The biomass of the bacterial cultures was harvested by centrifugation at 4500x g for 10 min at 4°C. The cell pellets were re-suspended in sodium phosphate buffer (0.1 M, pH 7.4) and disrupted by ultrasonic treatment for 7 min. The crude lysate was cleared by centrifugation at 20,000 x g for 1 h at 4°C and subjected to further purification. Lysates were loaded onto a Ni-NTA Superflow resin (Qiagen), washed with sodium phosphate buffer (50 mM, pH 7.4) containing imidazole (5 mM) and NaCl (0.3 M) and subsequently eluted with imidazole (0.3 M) in sodium phosphate buffer (50 mM, pH 7.4). To remove the imidazole, the pooled and concentrated protein fractions were loaded on a Sephadex desalting column PD-10 (GE Healthcare) and eluted with sodium phosphate buffer (50 mM, pH 7.4) containing 0.1 mM PLP. The purified proteins were used for the activity measurements and biocatalytic transformations.

The activity of the LTAaj mutants was verified by monitoring the formation of benzaldehyde upon cleavage of DL-syn-β-phenylserine (30 mM) or L-α-methyl-β-phenylserine (55:45 syn:anti mixture; 30 mM) at 279 nm (ε = 1.4 × 10³ M⁻¹ cm⁻¹). The kinetic assays were carried out at 25°C in sodium phosphate buffer (50 mM, pH 8) containing PLP (50 μM) and LTA (10–50 μM). One activity unit (U) is defined as the amount of enzyme that catalyzes the formation of 1 μmol of benzaldehyde per minute under these conditions. Controls without the addition of TAs were carried out under the same conditions.

The spectral scanning of purified LTAaj and mutants was performed in 1 mL sodium phosphate buffer (50 mM, pH 8.0) with LTA (50 μM) in a 1 cm light path cuvette at the wavelength range of 300–550 nm at 22°C. The formation of an external aldimine (λ = 420 nm) and quinonoid intermediate (λ = 495 nm) was monitored after addition of a donor substrate (0.1 M final concentration).

For substrate and product quantification by HPLC, reference DL-syn-β-phenylserine and L-α-methyl-β-phenylserine were measured in different concentrations (0.05–2.5 mM) to calculate standard curves for each compound. Concentrations were calculated using linear interpolation.

The HPLC-UV with pre-column derivatization was performed on an Agilent 1,200 Series HPLC system for e.e. and d.e. determination. The reaction was first quenched by diluting 1:50 with NTBB buffer (0.1 M of Na₂B₄O₇, pH 10.5), and the samples were treated with ortho-phthaldialdehyde/N-acetyl-L-cysteine (OPA/NAC). The OPA/NAC reagent was obtained by dissolving of 15 mg of NAC in 300 μL of distilled water and 300 μL of NTBB, and then 10 mg of OPA in 400 μL of MeOH was added. A freshly prepared solution was always used. 5 μL of OPA/NAC reagent was mixed with 1 μL of diluted reaction solution for 5 or 15 min (for α-quaternary products) in the HPLC system before injection. The molar ratio of OPA and the total of amino acids was maintained at 3.5: 1. Detection of the analytes was accomplished with a diode array detector at a wavelength of λ = 340 nm. The separations were carried out on an EC 150/3 Nucleodur® C18 Gravity column (150 × 3.0 mm, 3.0 μm; Macherey-Nagel, Düren, Germany). The mobile phases were 85% potassium phosphate buffer (20 mM, pH 6.8) and 15% acetonitrile (ACN) isocratic flow at 2.5 mL/min.

For HPLC-MS general monitoring of the reactions, eluent mixtures consisting of methanol, and water containing 0.01 vol% formic acid as additive were used. UV (210 nm) and ESI-MS detection, typically in positive mode, was used to identify the compounds. Following settings of the Agilent 6,120 MSD were applied: drying gas: N₂, 10.5 L/min; drying gas temperature: 350°C; nebulizer pressure: 35 psi; capillary voltage: 3,000 V; mode: negative; scan parameters: mass range: 100–450, fragmentor: 30 V, gain: 1.0, threshold: 150, step size: 0.10. The separation of the analytes was carried out using a Poroshell® 120 EC-C18, 3.0 × 100 mm, 2.7 μm (Agilent, Vienna, Austria) reversed-phase column. Samples were either dissolved in methanol and any insoluble particles were removed prior to analysis by centrifugation or filtration through 0.2 μm GHP Acrodic® 13 mm syringe filters (PN 4,567, Pall Corporation, Ann Arbor, USA). A stepwise gradient was used: 0.0 min: 98 % water/ 0.1% HCOOH and 2% CH₃OH, linear gradient to 100% CH₃OH in 6 min, 6.0–8.0 min 100% CH₃OH; 0.7 mL/min, 30°C.

The enzymatic transformations were performed in 1 mL reaction volume in sodium phosphate buffer (50 mM, pH 8.0), containing PLP (50 μM), DMSO (10%, v/v), benzaldehyde (50 mM) and glycine (0.5 M). Transformations were performed at 30°C with 0.5 U (based on phenylserine cleavage activity) of protein. To stop the reaction, 50 μL of the reaction mixture was diluted in 950 μL sodium tetraborate buffer (0.1 M, pH 10.5) and used for the determination of conversion by RP-HPLC after a pre-column derivatization with OPA/NAC.

L-β-phenylserine 3a: t_L−syn = 4.2 min, t_L−anti = 7.5 min.

L-β-(3-nitro)-phenylserine 4a: t_L−syn = 6.7 min, t_L−anti = 10.1 min.

The enzymatic transformations were performed in 1 mL reaction volume in sodium phosphate buffer (50 mM, pH 8.0), containing PLP (50 μM), DMSO (10%, v/v), 50 mM of benzaldehyde or 3-nitro-benzaldehyde and 0.5 M of D-amino acid donor (D-alanine, D-serine, D-aminobutanoic acid, D-valine, D-norvaline). Transformations were performed at 25°C with 10 U (based on phenylserine cleavage activity) of protein. To stop the reaction, 50 μL of the reaction mixture was diluted in 450 μL of 50/50 methanol/formic acid 0.01% and used for the HPLC-MS analysis.

L-α-methyl-β-phenylserine 3b: t_L−anti = 2.7 min, t_L−syn = 4.1 min; [M+1]⁺ 195.9

L-α-hydroxymethyl-β-phenylserine 3c: t_L−anti = 1.3 min, t_L−syn = 1.8 min; [M+1]⁺ 212.1

L-α-ethyl-β-phenylserine 3d: t_L−anti = 2.1 min, t_L−syn = 2.3 min; [M+1]⁺ 210.1

L-α-methyl-β-(3-nitro)-phenylserine 4b: t_L−anti = 3.1 min, t_L−syn = 4.4 min; [M+1]⁺ 241.2

L-α-hydroxymethyl-β-(3-nitro)-phenylserine 4c: t_L−anti = 1.6 min, t_L−syn = 2.1 min; [M+1]⁺ 257.2

L-α-ethyl-β-(3-nitro)-phenylserine 4d: t_L−anti = 2.3 min, t_L−syn = 2.5 min; [M+1]⁺ 255.0

L-α-propyl-β-(3-nitro)-phenylserine 4f: t_L−anti = 2.6 min, t_L−syn = 2.8 min; [M+1]⁺ 269.2

L-α-(2-hydroxyethyl)-β-(3-nitro)-phenylserine 4g: t_L−anti = 2.1 min, t_L−syn = 2.2 min; [M+1]⁺ 271.2

According to the phylogenetic analysis of LTA sequences, LTAs can be grouped into two major functional families—L-allo-TA and L-low specificity TA, which correlates with their specificity at C_β of L-threonine (Liu et al., 2015; Fesko et al., 2018). The sequence similarity is low between functional families and only a few conserved residues were found in all LTA sequences (G60, T61, N64, H85, E90, E138, D168, R171, and R313 in LTAaj). It is not surprising, that their mutation led to a significant loss of activity in the cleavage of L-threonine and L-β-phenylserine (Fesko et al., 2018). In this study, we have tested the performance of these and other active site mutants in the aldol addition reactions. Interestingly, the variants N64I, H85F, E90A, E138A, D168S, C196T, G200S, R231A, R313F with a significantly decreased retro-aldol activity catalyse the formation of 3a with conversions similar or even higher than the wild type enzyme although with slower rates (Figure 5, Table 1). For example, the R313F mutant with 500 fold reduced catalytic activity in the cleavage reaction, produces 3a with 40% conversion after 24 h and 70% after 5 days.

As was shown before, these conserved residues are essential to control the electrophilic strength of the PLP cofactor to act as an electron sink stabilizing the formed carbanion after C_α-C_β bond cleavage by the formation of quinonoid species of the PLP-glycine complex (Figure 2) (Fesko et al., 2018). This seems to be important for the retro-aldol reaction, but less essential in the opposite reaction—formation of the β-hydroxy-α-amino acid. Previously, we have shown, that the rate-limiting step in the LTA catalyzed reactions is the C_α-H deprotonation of glycine donor (aldol direction) or protonation of the C_α carbanion to form glycine (retro-aldol direction) (Figure 6) (Fesko et al., 2008). Thus, in the retro-aldol reaction, the life-time of the glycine carbanion is longer in comparison to the aldol reaction and has to be stabilized via delocalization of electron density in the PLP-ring.

The significantly higher energy barrier for the C_α-H protonation step in comparison to the C_α-C_β bond cleavage step requires a more stable carbanion in the case of retro-aldol cleavage (Fesko et al., 2008). The above described catalytic residues influence carbanion stability and thus suppress the retro-aldol reaction. Due to inhibition of reverse step, in the aldol addition reaction the equilibrium is reached later leading to the accumulation of the aldol products, when an excess of a donor is applied. Similar behavior was observed earlier for D-threonine aldolases, when aldol reactions were performed at low temperatures to suppress the cleavage reaction and to obtain aldol products under kinetic control (Steinreiber et al., 2007b). However, this was not possible to achieve in LTA catalyzed reactions due to its very high retro-aldol activity and fast equilibration even at low temperatures (Gutierrez et al., 2008). Here we observed mutations with intrinsically low retro-aldol activity, what allowed running the aldol reaction under kinetic control. Thus, conversion up to 60% was obtained for LTAaj variants with 1a and 2a, whereas only 20% was previously reported for the wild type enzyme under standard conditions (Figure 7, Table 1) (Blesl et al., 2018).

Moreover, in previous studies high excess of LTA was usually applied for the biocatalytic aldol reaction (5–10 U in a 1 mL reaction according to different studies) (Steinreiber et al., 2007a,b; Gutierrez et al., 2008; Dückers et al., 2010). This led to faster equilibration and the thermodynamically-controlled mixtures of products (syn/anti) was obtained. In the case of “low active” mutants or reducing the amount of wild type enzyme in the reaction mixture to 0.5 U, the products can be obtained under kinetically controlled conditions, when syn-isomer (in case of L-β-phenylserine) is formed faster than anti-isomer, leading to d.e. up to 80% and conversion of 20% after 5 h of reaction time (e.g., T10A mutant, Figure 7). Interestingly, some mutations cause the inversion of stereopreferences in the kinetically-controlled mode, thus anti-product was obtained as the major isomer for N64I, H85F, and R313F mutants. To conclude, for an efficient retro-aldol cleavage process catalyzed by LTA, the C_α-carbanion intermediate has to be stabilized, whereas in the aldol process this is much less important. Thus, lowering the retro-aldol activity opens the possibility to run the aldol reaction under kinetic control to reach high diastereoselectivities with good conversion.

The active site positions, which are located in the aldehyde substrate binding pocket (Y32, F89, K224), as well as positions important for the cofactor binding (G60, N142, A170, R171, R231) or participate in the C_β-OH protonation (H85, Y87, H128) reduce the conversion in the aldol addition reaction catalyzed by LTAaj (Figure 5, Table 1). The strongest negative impact on the aldol addition of benzaldehyde 1a to glycine 2a was observed for the mutations at residues responsible for the binding and fixation of the PLP-cofactor in the active site. For example, G60C perturbs the hydrogen bonding with the phosphate group of PLP and reduces the retro-aldol activity 10-fold, but gives only 5% conversion in the synthesis of 3a. The large cavity on the re-face of the PLP-ring is beneficial for accommodating large aliphatic and aromatic aldehydes, whereas the introduction of the polar residues (e.g., Thr, Ser, or Asn) in positions 32, 89, and 224 results in the decrease of conversion. The positions Y87 and H128 are responsible for the protonation of C_β-OH group, as was shown previously (Qin et al., 2014). Their mutation led to significantly decreased conversions in the aldol synthesis of 3a, whereas for the retro-aldol cleavage the effect was not significant.

Besides the natural donor glycine 2a, wild type LTAaj accepts also D-alanine 2b, and D-serine 2c in the aldol addition reaction with aldehydes to produce highly valuable L-α-quaternary-α-amino acids (Fesko et al., 2010). In order to improve the performance of biocatalytic transformations, reaction engineering was recently performed and a wide range of L-α-alkyl-β-hydroxy-a-amino acids were produced (Blesl et al., 2018). To further improve the donor specificity and conversions in the reactions, enzyme engineering can be additionally used. Thus, here we have investigated the influence of mutations of active site residues on the aldol synthesis of α-alkyl-β-phenylserines (Figure 8).

Recently, serine hydroxy-methyltransferase from Streptococcus thermophiles (SHMTst) was engineered to accept D-serine as a donor in the aldol reaction with aldehydes (Hernandez et al., 2015). The mutation Y55T was found to be beneficial to introduce the novel donor specificity presumably due to the generated space which allows to accommodate the (hydroxy)methyl group of the D-serine donor. Additionally, the hydroxy-group of T55 stabilizes the donor in the proper orientation for C_α-H proton abstraction via hydrogen bonding. In LTAaj T10 occupies the same space as Y55 in SHMTst (Figure 9A). Indeed, the T10A LTAaj mutant increased the conversion by 1.5 fold in the aldol addition of 1a to 2b to produce 3b with up to 70% conversion (Figure 8, Table 1). Besides T10, we have found positions D29, V31, Y32, H128, R231, and K224, which interact with the side-chain of an amino-acid donor, to have an impact on the donor specificity. The mutations V31T, Y32M, and K224S increased the substrate binding pocket to accommodate the side chain of a donor and thus improved the conversion in the aldol reaction with 2b and 2c (Figure 9B).

The change of arginine at position 231 to a smaller residue would create even more space and one would expect increased donor specificity for R231X mutants (Figures 9C,D). We performed the site-saturation mutagenesis at R231 in order to find an appropriate residue for this position and tested the variant for the synthesis of 3b, 3c, and 3d. The substitution to the positively charged amino acids (e.g., Lys, Gln) did not significantly influence the conversion. However, small aliphatic residues or residues bearing a negative charge decreased the conversion despite enlarged space upon mutation in comparison to the wild type (Table 2). This might be due to the important role of the residue R231 in the catalytic mechanism. It is a conserved residue among L-allo specific threonine aldolases, which participates in the hydrogen bonding with the phosphate group of PLP, as well as maintains the pKa of Schiff-base forming lysine K199 and acts as the general acid/base in the external aldimine formation (Fesko et al., 2018). Moreover, the spectrophotometric analysis showed that R231A mutation hinders the formation of the quinonoid complex with glycine (Figure 10) and, consequently, reduces the natural retro-aldol activity toward L-β-phenylserine by 10 fold. This confirms that R231 is an important catalytic residue, which participates in maintaining the hydrogen network in the active site for donor stabilization.

Interestingly, the quinonoid species were formed in the R231A mutant with non-natural donors, e.g., 2b and 2c (Figure 10). This might be due to the divergent mechanism of the fixation of the donors in the active site for C_α-H abstraction and carbanion stabilization. Indeed, different mutations improve the conversion in the aldol reactions with different donors. Thus, for donor 2b, the best conversion was obtained with the T10A mutant (Table 1). The conversion with 2c donor was increased upon mutation to V31T, where T31 might stabilize the donor in a proper orientation via hydrogen bonding with the hydroxy group of serine substrate. For 2d, the double mutant R231A/Y32M increased the conversion from 0.5 to 8%, presumably due to improved hydrogen stabilization between the side chain of 2d and M32. Further, H128Y increases the conversion in the aldol reaction with both 2b and 2c by providing a hydrophobic environment for the side chain of an amino acid product. Interestingly, H128 participates in the C_β-OH protonation in the reaction with natural donor glycine 2a (Qin et al., 2014; Fesko et al., 2018), therefore conversion toward 3a is reduced for H128F and H128Y mutants (Table 1). In the case of non-natural donors 2b-2d, the reactions were not affected by the H128F mutation, which indicates that other residue protonate C_β-O⁻ toward α-quaternary amino acids.

As was shown before, aldehyde 1a is not optimal substrate for LTA-catalyzed reactions and higher conversions can be obtained with the aromatic substrates bearing electron-withdrawing groups (Fesko et al., 2015; Blesl et al., 2018). Thus, we have applied 3-nitrobenzaldehyde (1b) as substrate to test other bulky donors with the mutant variants of LTAaj. Consequently, the conversion with the donor 2d was improved from 2% for the wild type to 33% for the R231A/Y32M mutant (Table 3). Further, the double mutant accepts also 2f and 2g as donors, albeit with low efficiency, which were not accepted by the wild type. The decrease of conversion with an increasing donor's side chain can be enforced due to its interaction with the phosphate group of PLP and destabilization of the overall enzyme-substrate complex. Thus, more rational active site design with simulation studies is beneficial for further engineering of donor specificity of LTAaj. To conclude, in order to get other D-donors accepted, the active site should be designed toward a specific target to enlarge the pocket, respectively, and to fix a donor in the optimal position for C_α-H deprotonation.

The conserved residue H85 in LTAaj has an important multifunctional role to (1) coordinate a catalytic water molecule in the active site (di Salvo et al., 2014), (2) to control the electrophilic properties of the PLP ring (Fesko et al., 2018), and (3) to maintain the diastereoselectivity. Although the H85F mutant was active in the aldol addition of 1a to 2a, the mutant was inactive in the reaction with non-natural donors. The same is true for other variants, which have an impact on the electrophilic properties of the PLP cofactor (N64I, H85F, E138A, R171T, and R313H)—they were not active in the synthesis of α-alkyl-β-phenylserines 3b-c (Table 1). This indicates that C_α-C_β bond formation in the case of C_α-substituted donors occurs with significantly lower rates in comparison to the glycine donor, thus C_α carbanion stabilization by the conjugated PLP system is essential for the catalytic activity with non-natural donors. Therefore, a different set of mutations is needed for improving the synthesis of either L-β-phenylserine or L-α-alkyl-β-phenylserines.

The author declares that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.