A narrow substrate range is a major limitation in exploiting enzymes more widely as catalysts in synthetic organic chemistry. For enzymes using two substrates, the simultaneous optimisation of both substrate specificities is also required for the rapid expansion of accepted substrates. Transketolase (TK) catalyses the reversible transfer of a C₂-ketol unit from a donor substrate to an aldehyde acceptor and suffers the limitation of narrow substrate scope for industrial applications. Herein, TK from Escherichia coli was engineered to accept both pyruvate, as a novel donor substrate, and unnatural acceptor aldehydes, including propanal, pentanal, hexanal and 3-formylbenzoic acid (FBA). Twenty single-mutant variants were first designed and characterised experimentally. Beneficial mutations were then recombined to construct a small library. Screening of this library identified the best variant with a 9.2-fold improvement in the yield towards pyruvate and propionaldehyde, relative to wild-type (WT). Pentanal and hexanal were used as acceptors to determine stereoselectivities of the reactions, which were found to be higher than 98% enantiomeric excess (ee) for the S configuration. Three variants were identified to be active for the reaction between pyruvate and 3-FBA. The best variant was able to convert 47% of substrate into product within 24 h, whereas no conversion was observed for WT. Docking experiments suggested a cooperation between the mutations responsible for donor and acceptor recognition, which would promote the activity towards both the acceptor and donor. The variants obtained have the potential to be used for developing catalytic pathways to a diverse range of high-value products.

Abbreviations

ee: enantiomeric excess
FBA: formylbenzoic acid
HPA: hydroxypyruvate
NMR: nuclear magnetic resonance spectroscopy
PAC: phenylacetylcarbinol
ThDP: thiamine diphosphate
TK: transketolase
TK_gst: TK from Geobacillus stearothermophilus
WT: wild-type

Introduction

The limited degree of substrate acceptance has been a major limitation in exploiting enzymes as catalysts in synthetic organic chemistry [1]. For enzymes having two substrates, the simultaneous optimisation of both substrate specificities is desired for rapidly expanding the scope of substrate acceptance, and the range of products formed. However, few studies have reported the engineering of both substrate specificities.

Transketolase (TK) catalyses the reversible transfer of a C₂-ketol unit from D-xylulose-5-phosphate to either D-ribose-5-phosphate or D-erythrose-4-phosphate, linking glycolysis to the pentose phosphate pathway in all living cells [2, 3]. The stereospecifically controlled carbon-carbon bond-forming ability of TK makes it very promising as a biocatalyst in industry, for the synthesis of complex carbohydrates and other high-value compounds [4-6]. The large-scale fermentation, downstream processing and purification of TK to give a stable lyophilised powder have been established at Sigma-Aldrich (Buchs, Switzerland) [7], and TK has also been used in large-scale biocatalytic reactions for the production of unusual sugars [8]. However, to achieve wider industrial applications in large-scale processes, a robust TK toolbox is desired, in which the variants should be able to accept a diverse range of substrates including both acceptors and donors.

WT TK accepts a narrow range of hydroxyaldehydes as the aldol-acceptor substrate, and with strict (2R)-specificity [8-10]. TK from Escherichia coli has been engineered successfully for improved and reversed enantioselectivity [11], as well as an expanded aldol-acceptor substrate range including polar aliphatic [12], nonpolar aliphatic [13], heteroaromatic [14, 15] and aromatic aldehyde derivatives [16, 17]. Related variants were also introduced by others into a thermostable TK from Geobacillus stearothermophilus (TK_gst) [18-21].

WT TKs have very limited tolerance for different donor substrates, which require both an oxo group adjacent to the C–C bond undergoing cleavage and a hydroxyl group at C-1. They also show strong preference towards the donor substrates with a D-threo configuration of hydroxyls at the asymmetric C-3 and C-4 positions [22]. In addition to its natural phosphorylated donors, TK can also accept hydroxypyruvate (HPA). The use of β-HPA as the ketol donor renders the donor half-reaction quasi-irreversible, thus increasing the atom efficiency of the reaction favourably for industrial syntheses. E. coli TK accepts HPA with 30-fold higher activity than other orthologs such as from yeast and spinach, thus resulting in higher specific enzyme activity [23].

Pyruvate 2d differs from HPA only by the absence of the C-3-hydroxyl group, and yet cannot function as a substrate in the TK reaction, indicating the critical role of the hydroxyl group in serving as a donor substrate [24]. Unlike TK, the closely related enzyme 1-deoxy-D-xylulose-5-phosphate synthase catalyses the transfer of the acetyl group from pyruvate 2d to glyceraldehyde-3-phosphate to yield 1-deoxy-D-xylulose 5-phosphate. This was used recently as a template to guide the engineering of TK_gst for acceptance of pyruvate and aliphatic analogues, as the donor substrate [25]. However, this enzyme variant has not been shown to work also with the aliphatic or aromatic aldehydes as acceptor substrates. The best variants of TK_gst were able to catalyse the preparative scale reactions between novel donors and hydroxylated acceptors with isolated yields of 48–88%. However, the acceptor substrate used during the process of directed evolution was the hydroxylated aldehyde, glycolaldehyde 1. Therefore, as the WT TK already utilises hydroxylated acceptors with high activity, it remains unknown whether TK variants can be engineered to accept a combination of both pyruvate 2d as donor and aliphatic or aromatic aldehydes as acceptors. The reactions between pyruvate 2d and aliphatic or aromatic aldehydes would allow access to a widely diverse range of products such as analogues of phenylacetylcarbinol (PAC), an important pharmaceutical intermediate, and precursors to drugs such as spisulosine and phenylpropanolamine.

In this study, we engineered E. coli TK variants to accept both novel donor pyruvate 2d and unnatural aliphatic and aromatic acceptor aldehydes, including propanal 5a, pentanal 5b, hexanal 5c and 3-formylbenzoic acid (3-FBA 7). Single variants capable of accepting pyruvate 2d were firstly obtained using a structure-guided strategy and previous knowledge. A small library was then constructed by combining the single mutants showing improved activity towards pyruvate 2d and glycolaldehyde 1 or propionaldehyde 5a. After screening this library, we successfully obtained the TK variants that could catalyse the reactions between sodium pyruvate 2d and propionaldehyde 5a, or 3-FBA 7, with significantly improved efficiency compared to WT. The potential mechanisms for the enhanced activity were then evaluated using molecular docking simulations.

Results and Discussion

Design of single mutants

Several enzymes are capable of accepting pyruvate 2d for carboligation reactions among the superfamily of thiamine diphosphate (ThDP)-dependent enzymes [26, 27]. For example, pyruvate decarboxylase (PDC) can catalyse a carboligation side reaction that has been exploited in the asymmetric synthesis of phenylacetyl carbinol from pyruvate 2d and benzaldehyde. However, it also efficiently catalyses the decarboxylation of pyruvate 2d into acetaldehyde and CO₂ showing catalytic promiscuity [28, 29]. Therefore, we attempted to design variants of TK based on a structural alignment with PDC.

The E. coli TK crystal structure complexed with the natural donor substrate D-xylulose-5-phosphate and cofactor ThDP has been solved previously [30]. HPA readily acts as a donor substrate, whereas pyruvate 2d cannot, indicating a critical role of the C-3-hydroxyl group for catalysis. The TK structure in complex with the donor substrate showed that it is anchored by hydrogen bonds between the donor C-1-hydroxyl group, and His100 and His473 in the enzyme active site (Fig. 1A). Therefore, in order to stabilise new donor substrates lacking the C-1-hydroxyl group, new interactions may be required with other local residues, such as hydrophobic interactions with the methyl group in pyruvate. Therefore, we identified nine residues (His26, His66, His100, Leu116, Ile189, His261, Ser385, Asp469 and His473) within 5 Å of the first two carbons of the donor substrate, as potential mutation sites (Fig. 1A).

Details are in the caption following the image — **Figure 1**
Open in figure viewer PowerPoint

Structure alignment between TK (2R8O.pdb) and PDC (2VK1.pdb). (A) Structure of TK in the complex with D-xylulose-5-phosphate: green, the substrate; cyan, thiamine pyrophosphate (ThDP). Distances are shown in Å. (B) Active-site alignment between TK (blue) and PDC (pink). Only the residues of PDC are labelled. The cofactor ThDP of two enzymes are shown as lines.

Although the TK and PDC studied only shared 6.7% sequence identity, 559 pairs of residues that are structurally equivalent were found using the TopMatch tool (Fig. S1, Table S1). The root-mean-square error of the superposition for the structurally equivalent 559 C_α atoms was 2.66 Å and the sequence identity in the equivalent regions increased to 12.3% (Table S1). Interestingly, only three target residues (Leu116, Ser385 and His473) were located in the structurally equivalent regions (Fig. S1B). Based on structure alignment, Leu116 of TK is equivalent to Ile415 of PDC, and so the mutation L116I was proposed. However, for Ser385 and His473, the equivalent structural alignment with PDC showed the same residues at the equivalent positions (Fig. S1B). Ser385 of TK is equivalent to Ser32 of PDC, which is close to Asp28 in both the sequence and 3-D structure (Fig. 1B). Asp28 in PDC was reported previously to play a critical role in the stabilisation of the enzyme-bound cofactor, and in the stereochemical control of the decarboxylation reaction [31], and so the mutation S385D was proposed. His473 formed a hydrogen bond with the C-3-hydroxyl group in TK and played a role in anchoring the donor substrate. The equivalent residue in TK_gst was His474, and the mutations H474S and H474N were previously found to give a 6- and 3-fold improved activity towards glycolaldehyde 1 and pyruvate 2d, respectively. Therefore, we included the corresponding mutations H473S and H473N in E. coli TK to investigate whether these mutations would also improve the activity of E. coli TK towards pyruvate.

As for the target residues that were not in structurally equivalent regions, we proposed a range of mutations at each target site based on the types of residues found in the equivalent spatial location in PDC to enhance the chances of success. The active-site residues around the cofactor and donor substrate for the aligned TK and PDC are shown in Fig. 1B. Ile189 of TK was located in a space occupied by Glu477 in PDC, and also close to Asn488, and so the mutations I189N/E were proposed (Fig. 1B). I189Q was also included as glutamine has a similar structure to asparagine. Ile476 is a conserved hydrophobic residue among PDCs from various organisms and plays an important role in substrate recognition and enantioselectivity [32]. Three target histidines 26, 66 and 261 were all within 2 Å of Ile476 in PDC, and so the mutations H26I, H66I and H261I were also proposed. In addition, the mutations H26Q, H66T and H261E were proposed to approximately correspond to residues Gln552, Thr390 and Glu477 of PDC.

His100 in TK also formed a hydrogen bond with the C-1-hydroxyl group of the donor substrate and its mutation to alanine previously had little effect on the K_m for acceptor substrates, but significantly increased than that of the donor substrate, suggesting its important role in accepting the donor substrate [33]. Based on the structural alignment to PDC, His100 was close to residues His114 and Phe292 of PDC (Fig. 1B), and so we proposed the mutation H100F. Considering the importance of this site, we also included alternative mutations. H100Y was included as tyrosine has a similar structure to phenylalanine. In addition, the TK_gst mutations H102F/L/T previously showed improved performance towards glycolaldehyde 1 and sodium pyruvate 2d [25]. Therefore, we included corresponding mutations H100L and H100T in E. coli TK to investigate whether these mutations would improve activity in E. coli TK towards glycolaldehyde 1 or propionaldehyde 5a and pyruvate 2d.

When predicting the potential mutations, previous knowledge was also considered. Asp469 is a hotspot for engineering TK towards novel acceptor substrates. While using HPA as the donor, D469E led previously to increased activity towards aliphatic aldehyde acceptors, and D469T increased the activity towards both aliphatic and aromatic aldehydes [16, 34]. Preliminary work suggested that these two variants may have accepted pyruvate 2d as donor substrates with pentanal as the aldehyde acceptor, although with a very low reaction yield. Therefore, in addition to the D469H mutation suggested by the proximity of His114 in the structurally aligned PDC, we also included D469E and D469T into our screens. In total, we designed and constructed 20 single mutants (summarised in Table S2) and measured their activities towards glycolaldehyde 1 or propionaldehyde 5a and pyruvate 2d experimentally.

Characterisation of single variants

Reaction with glycolaldehyde and pyruvate

Clarified lysate preparations of the twenty single mutants were first tested for catalytic efficiency towards pyruvate 2d, and paired with glycolaldehyde 1, as this is known already to be one of the best acceptor substrates for WT TK. This allowed us to evaluate the acceptance of the novel donor substrate in the context of a favourable hydroxylated aldehyde acceptor. The target product, 3,4-dihydroxy-2-butanone (DHB) 3 of the reaction between glycolaldehyde 1 and pyruvate 2d was detected by HPLC at a retention time of 13.7 min (Scheme 1 in Fig. 2). However, erythrulose (ERY) 4 was also detected as a side product at the retention time of 11.4 min (Scheme 1 in Figs 2 and 3A), which was not reported in any previous mutagenesis studies, although the reaction of two glycolaldehydes to form ERY is a known slow side reaction for WT TK [35].

Considering the possible difference in expression level of the variants, the yields of both products determined after 24-h of reaction were normalised by the TK concentration as determined from the relative density of SDS/PAGE bands, to give an approximate measure of specific activity (Fig. S2). Each of the mutations tested influenced the yields of both the target product and the ERY by-product, and also the ratio of the two products, relative to WT (Fig. 3B). The variants H100F/L/T/Y, L116I, D469E and H473N in particular gave markedly increased reaction yields of the target product relative to that of WT. All of these improved variants were subsequently purified and re-evaluated, except for H100T which was the lowest ranked. Two negative variants, I189N and H473S, were also purified and tested as controls (Fig. 3C). Interestingly, H100Y, L116I, D469E and H473N no longer showed an improvement in activity in their pure enzyme form, compared to WT (Fig. 3B). The same variants also had low protein expression levels (Fig. S2) and so their poor specific activity when purified may have been due to an adverse impact of the mutations on enzyme stability, or otherwise resulted from the increased error in measuring low protein concentrations by densitometry on lysates.

The product concentration of the reaction catalysed by H100L was 22.3 mm, with a conversion rate of 45% (Table 1), a 2.1-fold improvement compared to WT, consistent with previous observations with equivalent TK_gst variants [25]. H100F also improved the catalytic efficiency by 1.2-fold. These indicate that a hydrophobic interaction has played an important role in anchoring the pyruvate 2d, consistent with other ThDP-dependent enzymes that have hydrophobic residues such as valine or phenylalanine for interactions with the methyl group of pyruvate 2d [36, 37]. However, H473N and H473S did not improve the target product yield in E. coli TK, in contrast to the effect of the equivalent mutations in TK_gst (Fig. 3C). The contextual structure differences between the active sites of E. coli TK and TK_gst are a likely reason for these differences. However, it should be noted that differences in the activity assays used in our work compared to those in previous reports with TK_gst could also lead to similar observations. The pH-sensitive colorimetric assay used for TK_gst would not have detected the ERY by-product, nor would distinguish between the decarboxylation of pyruvate 2d and true TK function (decarboxylation followed by ketone unit transfer to acceptor) [25, 38].

Table 1. Reactions between novel acceptor and donor substrates catalysed by TK variants. ND, Not determined.

Acceptor substrates	Donor substrates	TK variant	Product	Yield (%)	Improvement relative to WT	Configuration	ee (%)
1	2d	H100L	3	45	2.1-fold	ND	ND
5a	2d	4Ma/H100L/ D469E/R520Q	6ad b	ND	9.2-fold	ND	ND
5b	2d	D469E/T	6bd	5	+∞d	3S	> 98c
5b	2e	D469T	6be	4	+∞d	4S	> 98c
5c	2d	D469E/T	6cd	5	+∞d	3S	> 98c
5c	2e	D469E	6ce	7	+∞d	ND	ND
7	2d	4M/H100L/ D469T	8	47	+∞d	ND	ND

^a 4M contains mutations H192P/A282P/I365L/G506A
^b Difficult to isolate due to its high volatility
^c Determined by GC
^d No reaction was observed for WT

The variants also changed the relative reaction yields for the two products. WT produced equal concentrations of target product and by-product, at 10.5 mm with a conversion rate of 21% (Fig. 3C), whereas all purified variants except I189N gave higher amount of the target product DHB 3, than the by-product ERY 4. Both I189N and I189Q showed a higher amount of by-product than the target product, indicating that the substitution of isoleucine at position 189 to residues with amide side-chains could potentially increase the specificity towards glycolaldehyde 1 over pyruvate 2d. Ile189 interacts with the thiazolium ring of ThDP via a hydrophobic interaction and also forms the substrate channel, through which the donor substrate approaches the C-2 atom of the thiazolium ring to form the dihydroxyethyl-ThDP intermediate. The mutants with amide groups could hence facilitate the entry of the more polar glycolaldehyde 1, compared to a preference from Ile189 for the nonpolar methyl group within pyruvate 2d. Interestingly, the neighbouring mutation S188Q showed a 2.3-fold improvement in activity towards Li-HPA and glycolaldehyde [12], implying the importance of this region of the active site in accepting glycolaldehyde 1.

H473N gave the highest selectivity towards pyruvate, with no by-product detected after 24-h reaction, even though the target product accumulated to 5.5 mm (Fig. 3C). It has been found that in the one-substrate reaction, TK accepted glycolaldehyde 1 to form the dihydroxyethyl-ThDP intermediate with a low rate of 0.23 U·mg⁻¹ and a high K_m of 140 mm [35]. In contrast, TK_gst H473N had a K_m of 13.4 mm towards pyruvate 2d, lower than that of TK_gst WT (40.2 mm) and even the TK_gst H100L (29.1 mm) [25]. These results suggested that the H473N could have marked specificity towards pyruvate 2d over the glycolaldehyde 1 as the donor substrate. The ratios of the target product DHB 3 relative to by-product were 8.3, 5.3, 5.2, 2.9, 2.4 and 1.5 for H100L, H100F, H110Y, H473S, L116I and D469E, respectively.

Most of the variants based on structural comparison to PDC failed to improve the yield of the target product DHB 3 for the reaction between glycolaldehyde 1 and pyruvate 2d, which is perhaps not surprising given only 6.7% sequence identity between TK and PDC and the likely need for multiple simultaneous mutations to generate a change in function.

Reaction with propionaldehyde and pyruvate and extension to other aliphatic analogues to determine stereoselectivities

All variants were also tested for their activities towards propionaldehyde 5a and sodium pyruvate 2d. WT E. coli TK can accept a range of nonhydroxylated aliphatic aldehyde substrates in the presence of HPA. However, the activity is typically only 5–35% of those for hydroxylated substrates such as glycolaldehyde 1. Propionaldehyde 5a is therefore a more challenging acceptor for TK than glycolaldehyde 1, especially when using pyruvate 2d as the donor substrate. Initially, the 12 variants not already purified by the protocol mentioned above (Fig. 3C) were first tested for their activity in lysates, to indentify whether any of them were also worth purifying and investigating further. After reaction for 24 h using the cell lysates (Fig. 4A), the target product 3-hydroxy-2-pentanone 6ad was detected indirectly using the colorimetric assay (Scheme 2 in Fig. 2). Four variants, H26Q, H100T, H261I and D469T, gave 2- to 4-fold improved specific activities relative to WT (Fig. 4A). These four variants were subsequently purified and re-evaluated together with the other 8 variants already purified above.

The absorbance at 450 nm increased with reaction time, indicating a significant time-dependent formation of 3-hydroxy-2-pentanone 6ad (Fig. 4B). However, the purified variants H26Q and H261I no longer improved the activity, even though their lysates had shown improved performance, which could also be attributed to their low expression as discussed above. Seven variants, including H100F, H100L, H100T, D469E, D469T, H473N and H474S, gave improved activity relative to WT, with the most active variant D469E showing more than 2-fold improvement in the catalytic efficiency compared to WT (Fig. 4B).

Based on the colorimetric assay, WT also appeared to show activity towards propionaldehyde 5a and pyruvate 2d, which had never been observed previously. In order to confirm the reaction product, GC/MS was applied to analyse the reaction mixture after 24-h of reaction. A peak was identified by GC with the retention time of 8.9 min (Fig. S3A). The MS data confirmed the successful generation of the target product 3-hydroxy-2-pentanone 6ad. In this compound, fragmentation between C-2 and C-3 lead to the formation of two species with m/z of 59 and 43, respectively. Further fragmentation of the compound with m/z 59 generated fragments with an m/z of 31 [39](Fig. S3B).

The peak area in GC was used to quantify the relative amounts of product generated by the enzyme reactions. We carried out the enzyme reactions for 48 h using seven variants, including both the improved and nonimproved ones in their purified form. The GC peak area and absorbance at 450 nm gave a good linear correlation with an R² of 0.93, reflecting that the colorimetric assay was a reliable activity measurement method (Fig. S4). The negative intercept on the GC peak area axis indicated that some of the colorimetric assay response was due to background. This would reduce the apparent activity of WT considerably, although the presence of a peak by GC confirmed that it could indeed form the target product. Consistent with the measurements of the colorimetric assay, the GC peak area of the 3-hydroxy-2-pentanone 6ad showed that D469E was the most active variant, with a 4.3-fold improvement in yield relative to that of WT. H100F, H473S, H100L and H473N also revealed 1.4-fold, 1.5-, 2.4- and 2.6-fold improvements in yield respectively, relative to WT (Fig. 5). Larger improvements were observed by GC relative to WT, as a result of the better removal of background signal compared to in the colorimetric assay, and so the results from GC would be more accurate.

3-Hydroxy-2-pentanone 6ad is hard to isolate due to its volatility, so larger scale TK reactions were performed with pentanal 5b and hexanal 5c as acceptors with a view to isolating the products (Appendix S1 and S2). Both pyruvate 2d and keto-butyric acid 2e were also used as donors to highlight general applicabilities of the reaction at this stage (Scheme 2 in Fig. 2). Following the screening of variants with propanal and pyruvate described above, D469E was selected for use together with D469T as a lower performing mutant for comparison purposes. As shown in Scheme 2 (Fig. 2), isolated products 6bd and 6cd were formed in reactions with both D469T and D469E with an approximate yield of 5% in both cases, although some material was lost due to volatility (Table 1). (3R)-Hydroxy-2-octanone 6cd, the product arising from hexanal 5c and 2d, is an insect pheromone produced by males of the longhorn beetle Anaglyptus subfasciatus and a Mosher’s derivatisation method has been reported to determine enantiomeric excesses (ees) [40]. Following this precedent, the TK product 6cd was reacted with (2R)-3,3,3-trifluoro-2-phenylpropanoyl chloride ((R)-MTPA) to give the (2S)-Mosher’s ester (Appendix S3). ¹H nuclear magnetic resonance spectroscopy (NMR) analysis revealed a major singlet for the 1-H₃ group at 2.21 p.p.m., and a minor peak at 2.13 p.p.m. (Appendix S4). The positions of these two signals corresponded to the (2S,3S) and (2S,3R)-esters, respectively, following the literature and indicated that the major (3S)-isomer was formed by the TK mutants: integration of the peaks revealed an ee of > 95%. A second set of diastereomeric signals could also be observed for the 8-H₃ triplet, and integration of the signals at 0.83 p.p.m. [major (2S,3S)-isomer] and 0.93 p.p.m. [minor (2S,3R)-isomer] confirmed the ee as > 95% (Appendix S4). The absolute stereochemistry was further confirmed by measurement of the optical rotation compared to reported data. In addition, racemic samples of 6bd and 6cd were also formed via chemical methods as chemical standards (Appendix S3) to determine the isomeric compositions of the products via chiral GC which confirmed the formation of the major isomer in > 98% ee (Table 1). The reaction product 6bd formed from pentanal 5b and pyruvate 2d was then derivatised using (R)-MTPA, and integration of the 1-H₃ signal again indicated an ee of > 95% for the (S)-isomer. Chiral GC of the racemic product compared to the TK products with D469T and D469E similarly revealed ees of > 98% (Appendix S3). These results highlighted the highly stereoselective nature of the reactions.

From these results, the same aldehydes were used with keto-butyric acid 2e to give 6be and 6ce using D469T and D469E, respectively, which gave the highest isolated yields. Racemic samples of 6be and 6ce were again prepared to confirm the characterisation data. To determine again the stereoselectivity for one example 6be, it was analysed by chiral GC and reacted again with (R)-MTPA (Appendix S3). Analysis of the ester product 8-H₃ signal revealed two triplets at 0.82 p.p.m. [major (2S,4S)-isomer] and 0.94 p.p.m. [minor (2S,4R)-isomer], consistent with the method described above, with integrations indicating the ee as > 95%, and GC analysis gave an ee of > 98% (Appendix S3 and S4).

The improvements in yield for the reaction between propionaldehyde 5a and pyruvate 2d from E. coli TK variants D469E, H473N, H100L, H473S and H100F were particularly interesting given that they typically improved the activity towards glycolaldehyde 1 and sodium pyruvate 2d by a lesser extent, or even not at all for D469E, H473N and H473S. This highlights the complexity of engineering an enzyme towards two substrates simultaneously, and also that improvements towards the more hydrophobic pyruvate (relative to HPA) may be structurally linked to the better acceptance of the more hydrophobic aldehyde acceptor propionaldehyde 5a (relative to glycolaldehyde 1).

The catalytic reaction of TK proceeds in two stages. In the first half of the reaction, cleavage of the donor substrate is accompanied with the formation of a covalent intermediate. In the second half of the reaction, nucleophilic attack of the acceptor substrate by the enamine intermediate leads to the product formation and release. The improvement of the final yield of the reaction could be attributed to both the first and second halves of the reaction. D469E showed improved reaction yields towards propionaldehyde 5a and pyruvate 2d, and yet failed to significantly increase the activity towards glycolaldehyde 1 and pyruvate 2d. Interestingly, D469E also showed improved activity towards propionaldehyde 5a but decreased the activity towards glycolaldehyde 1 when HPA was used as the donor [12, 15, 38]. This suggested that D469E only improved the efficiency of the second half of the reaction which involves acceptance of the aldehyde substrate, but did not affect acceptance of pyruvate in the first half. By contrast, the mutations of His100 could have only influenced the first half of the reaction, accepting the pyruvate. We tested the correlation between reaction yields for the purified variants, when reacting with the two different acceptors and sodium pyruvate (Fig. S5). It clearly showed that the mutations of residue His100 were responsible for the enzyme activity towards both the glycolaldehyde and propanal, with an R² of 0.99 when only H100L, H100F, H100Y and WT were considered, suggesting that mutations of His100 were critically required to accept the pyruvate, while not being important for controlling the acceptor specificity, consistent with previous observations [33]. We thus combined the variants that improved donor acceptance with those that improved acceptor acceptance, to generate a new library.

Design of a combined library

In order to further improve the activity of E. coli TK towards propionaldehyde 5a and pyruvate 2d, eight single variants including H100F, H100L, H100T, D469E, D469T, H473N, H473S and R520Q were selected for recombination to construct the library. Among them, each of the strains H100F, H100L, H100T, D469E, D469T, H473N and H473S showed improvement in the activity towards propionaldehyde 5a and sodium pyruvate 2d. The R520Q mutation was also included as it was previously found to improve stability, and also the catalytic efficiency towards HPA and propionaldehyde 5a, within certain other variants [34]. This strategy kept the library size small while still sampling for potential cooperative effects between mutations. In previous work with TK [34], recombination of single active-site mutants, including at Asp469, nearly always led to the loss of soluble expression. Therefore, we used a previously stabilised variant H192P/A282P/I365L/G506A (4M) as the template for constructing the new library to promote evolvability [34, 41, 42]. Introduction of these same stabilising mutations was recently found to restore the stability loss of another TK variant (S385D/D469T/R520Q) that was engineered to accepted aromatic aldehydes [42, 43].

Improvement of activity towards propionaldehyde and pyruvate

The library was screened towards propionaldehyde 5a and sodium pyruvate 2d using the colorimetric assay. Ten promising variants were identified, purified and tested for catalytic activity. The reaction yield relative to WT was determined for the variants using GC after reaction for 24 h (Fig. 5). As the quadruple variant H192P/A282P/I365L/G506A (4M) was used as the template for constructing the library, its activity was also measured. The 4M maintained a comparable activity with WT, whereas its combination with other mutations influenced their impact. The D469E variant previously gave a 4.3-fold improvement in product yield relative to the WT, whereas 4M/D469E had improved the yield by 6.7-fold relative to 4M, demonstrating the benefit of TK stabilisation even for the single active-site mutations (Fig. 5). However, the addition of the compensatory stabilising 4M mutations did not significantly change the impact of H100L (Fig. 5), presumably as this variant did not significantly decrease the stability of TK relative to the WT stability. Interestingly, addition of 4M influenced the activity of H473N in a negative way. 4M/H473N only showed 1.2-fold improvement in reaction yield compared to WT, whereas H473N itself revealed 2.4-fold improvement. Thus, the relationship between template stability and the fitness of the new variant is not always predictable.

The combination between H100L and D469E generated the variant 4M/HL/DE that gave an 8.1-fold improvement in the reaction yield compared to WT. Addition of the mutation R520Q to this variant improved the reaction yield further and generated the best variant overall, 4M/HL/DE/RQ, which had a 9.2-fold improvement in the reaction yield compared to WT (Fig. 5, Table 1). R520Q has been reported to increase the activity of D469T towards propionaldehyde 5a and HPA due to improved soluble protein expression. The combination of H100L and H473N also led to a variant showing 4.2-fold improvement in the reaction yield compared to the WT, confirming the important role of H100L in accepting pyruvate 2d.

Novel enzyme activity towards 3-FBA and pyruvate

In our previous work, variants showing improved performance towards propionaldehyde 5a and HPA also improved the catalytic efficiency towards aromatic substrates with HPA, including 3-FBA 7 or 4-FBA [16, 17]. For instance, D469T improved the specific activity towards propionaldehyde 5a and HPA by 5-fold and also improved conversion yield of 3-FBA 7 and 4-FBA from 0% to 65% and 30%, respectively. We therefore explored whether any of the new library variants had improved activity towards the aromatic aldehyde 3-FBA 7 and pyruvate 2d (Scheme 3 in Fig. 2).

The whole library was screened for activity towards 3-FBA 7 and sodium pyruvate 2d as determined by HPLC. Most variants had no activity towards these two substrates, with the exception of 4M/H100L/D469T, 4M/D469T/H473N and 4M/D469T/H473N/R520Q. After reaction for 24 h, a new peak at the retention time of 4.1 min was identified that correlated with the consumption of the 3-FBA 7 in the HPLC chromatogram for these three variants, but not for WT or any other variants (Fig. 6A).

In order to confirm if this peak was the product expected, the peak was isolated by semi-preparative HPLC and characterised by ¹H NMR and LC/MS. In the LC/MS analysis of the extracts, a peak with an m/z of 195.1 was observed, corresponding to the expected product 3-(1-hydroxy-2-oxopropyl) benzoic acid 8 (Fig. S6). The NMR spectrum also confirmed the generation of the target product (Appendix S4). These data demonstrated that the E. coli TK variants catalysed the unprecedented reaction between the aromatic aldehyde and sodium pyruvate 2d.

Three variants including 4M/H100L/D469T (4M/HL/DT), 4M/D469T/H473N (4M/HL/DT/HN) and 4M/D469T/H473N/R520Q (4M/HL/DT/RQ) were also purified and tested for the conversion of 3-FBA 7 with sodium pyruvate 2d as the donor substrate. The variant 4M/H100L/D469T showed the highest activity of all, with an initial rate around 2-fold higher than those of 4M/D469T/H473N and 4M/D469T/H473N/R520Q (Fig. 6B). The variant 4M/H100L/D469T was then characterised at higher enzyme concentrations. When increasing the concentration of enzyme from 0.067 to 1.330 mg·mL⁻¹, the conversion of 3-FBA 7 to product 3-(1-hydroxy-2-oxopropyl) benzoic acid 8 within 24 h was increased from 2.5% to 47% based on the substrate depletion (Tables 1 and Table S3). The specific activity, K_m for pyruvate 2d and k_cat of 4M/H100L/D469T were found to be 0.048 µmol·mg⁻¹·min⁻¹, 90.4 mm and 0.22 s⁻¹, respectively (Table S3).

Addition of the mutation S385Y into the double-mutant D469T/R520Q previously improved the catalytic efficiency towards 3-FBA 7 and HPA by significantly decreasing the K_m [17]. We therefore introduced the S385Y mutation into the two mutants 4M/D469T/H473N and 4M/D469T/H473N/R520Q to test whether their activities could be improved. The variant 4M/H100L/D469T/H473N/R520Q was also purified for activity measurements. Interestingly, the mutation S385Y failed to improve the activity for the variants of 4M/D469T/H473N and 4M/D469T/H473N/R520Q (Fig. 6B). The residue Ser385 was within 5 Å around the first two carbons of the donor substrate (Fig. 1A). The mutation S385Y, hence, could easily have interfered with the binding of pyruvate 2d and decreased the activity towards 3-FBA 7 and sodium pyruvate 2d. The variant 4M/H100L/D469T/H473N/R520Q also showed limited activity, consistent with the activity measured using lysates (Fig. 6B).

Elucidating the mechanism of enhanced activity

To gain insights into the influence of the mutations on enzyme activity, the substrates were docked into the binding pocket of the WT and several TK variants using the Autodock Vina program [44]. 4M/H100L/D469E and 4M/H100L/D469T were selected for docking as they had significantly improved enzyme catalytic efficiencies and yet different activities towards propionaldehyde 5a and 3-FBA 7 with the only difference in their substitution of position 469. The first half of the TK reaction consists of the cleavage of the donor substrate, pyruvate 2d, and the formation of the first product, carbon dioxide and a covalent intermediate, ThDP-enamine. The intermediate was first docked into WT and mutant TKs (Fig. S7A). ThDP-enamine had the binding affinity energy of −9.5 kcal⁻¹mol⁻¹ when bound into the WT, 0.3 kcal·mol⁻¹ higher than when it was bound into the 4M/HL/DE and 4M/HL/DT, indicating that the intermediate had stronger binding with the variants than with the WT. The poses of ThDP-enamine in the binding pockets of 4M/HL/DE and 4M/HL/DT were similar to each other, with an RMSD of 0.3 Å, but each differed slightly to the binding in WT with an RMSD of 0.7 Å (Fig. S7A, Table 2). When the pyruvate 2d instead of HPA was used as the donor substrate, the (α-hydroxyethyl)-ThDP intermediate instead of (α, β-dihydroxyethyl)-ThDP intermediate was generated. In the WT, the C2-methyl group of the hydroxyethyl was located in a highly polar environment surrounded by six charged residues including H26, H66, H100, H261, D469 and H473, which was potentially unfavourable (Fig. S7B). Substitution of histidine with leucine at position 100 would reduce the polarity of this local region and form a hydrophobic interaction between L100 and C2-methyl group of the intermediate (Fig. S7C,D). This could result in the slight difference of binding orientation of ThDP-enamine and stronger binding affinity in the two variants compared to WT.

Table 2. Substrates binding pose and affinity in WT and mutant TKs.

	ThDP-enamine			Propionaldehyde				3-FBA
	Affinity (kcal·mol⁻¹)	RMSDa (Å)		Affinity (kcal·mol⁻¹)	Distanceb (Å)	RMSD (Å)		Affinity (kcal·mol⁻¹)	Distanceb (Å)	RMSD (Å)
WT	−9.5	0	0.7	−2.4	5.2	0	2.0	−5.4	5.0	0	1.1
4M/HL/DE	−9.8	0.7	0	−2.4	4.6	2.0	0	−5.6	6.1	1.1	0
4M/HL/DT	−9.8	0.7	0.3	−2.4	5.0	0.9	2.4	−5.5	5.1	2.1	2.5

^a The RMSD was calculated with WT or 4M/HL/DE as the reference
^b The distance is the nucleophilic attack distance of α-carbanion of the ThDP-enamine on the aldehyde group of the acceptor substrate

Two acceptor substrates, propionaldehyde 5a and 3-FBA 7, were then separately docked into the WT and variants containing the ThDP-enamine intermediate. According to the elucidated catalysis mechanisms of TK, the α-carbanion of the (α-hydroxyethyl)-ThDP intermediate would attack the aldehyde group of the acceptor substrates to form the products with a carbon skeleton extended by two carbon atoms. The distances between the C-1 atom of the hydroxyethyl group of the ThDP-enamine and the carbon atom of the aldehyde group of propionaldehyde 5a were 5.2, 4.6 and 5.0 Å for the WT, 4M/HL/DE and 4M/HL/DT, respectively, consistent with the activity measurements towards propionaldehyde 5a and pyruvate 2d (Table 2, Fig. 5). The higher activity of 4M/HL/DE thus potentially resulted from the slightly shorter distance between the acceptor substrate propionaldehyde 5a and the enamine intermediate, compared to 4M/HL/DT and WT. As shown in Fig. 7, the D469T and D469E mutations apparently led to differences in the shape of the tunnel around the intermediate. The D469E mutation led to a narrower tunnel because of the longer side chain of glutamic acid (Fig. 7B,E). The longer side chain also has greater steric hindrance, which result in C-2 and C-3 atoms of propanal having dramatically different positions when docked into WT, 4M/HL/DE and 4M/HL/DT (Fig. S8). Compared to the side chain of threonine, that of glutamic acid will have greater polarity, whereas the C-2 and C-3 atoms of propanal are nonpolar. The strong polarity and long side chain of Glu469, hence, may lead propanal to have a shorter distance towards the ThDP-enamine compared to WT and 4M/HL/DT (Table 2 Fig. S8).

3-FBA 7 is larger than propanal, and so the narrow substrate tunnel of 4M/HL/DE influenced the access of 3-FBA 7 to the enamine intermediate as evidenced by the distance of 6.1 Å, which was 1 and 0.9 Å longer than those for 3-FBA 7 docked within WT and 4M/HL/DT, respectively (Table 2). This is consistent with the observation that 4M/HL/DE did not show any activity towards pyruvate 2d and 3-FBA 7. By contrast, the D469T mutation expanded the size of the substrate tunnel towards the intermediate (Fig. 7C,F), which allowed the occurrence of nucleophilic attack of the intermediate on the aldehyde group of 3-FBA 7. This was similarly observed previously when we docked the 3-FBA 7 into D469T containing the (α, β-dihydroxyethyl)-ThDP intermediate [16]. 4M/HL/DT also showed a binding affinity of −5.5 kcal⁻¹mol⁻¹, 0.1 kcal⁻¹mol⁻¹ lower than that of WT. A possible explanation is that replacing the negatively charged residue aspartic acid with the more hydrophobic and uncharged threonine provides more favourable interactions with the nonpolar acceptor substrates.

Consistent with the activity test, the docking experiments suggest a picture of cooperation between different residues to improve the catalytic efficiency towards different substrates. Specifically, combination between H100L, responsible for the donor acceptance, and D469E, responsible for propionaldehyde 2 acceptance, lead to the mutation showing enhanced activity towards pyruvate and propanal, while the combination of H100L and D469T responsible for 3-FBA acceptance generates the mutation that improves the activity towards pyruvate and 3-FBA.

Conclusion

In this study, we attempted to engineer the TK from E. coli to accept both the unnatural acceptor substrates, propanal 5a, pentanal 5b, hexanal 5c or 3-FBA 7, and a novel donor substrate, sodium pyruvate 2d and its analogue keto-butyric acid 2e, thus decreasing the polarity of both substrates. Twenty single-mutant variants were firstly designed based on structural alignment with PDC and previous knowledge of variants in E. coli and G. stearothermophilus. The catalytic efficiency was then measured for these single-mutant variants experimentally. Beneficial mutations were then recombined using a previously obtained stable variant as the template to construct a library. Library screening was carried out for two reactions separately: (a) sodium pyruvate 2d and propionaldehyde 5a; and (b) sodium pyruvate 2d and 3- FBA 7. The best variant 4M/H100L/D469E/R520Q showed a 9.2-fold improvement in the catalytic yield towards sodium pyruvate 2d and propionaldehyde 5a compared to WT. Pentanal and hexanal were also used as acceptors in the reactions with pyruvate 2d or keto-butyric acid 2e to determine stereoselectivities, which were found to be higher than 98% ee for the S configuration. Also, three unprecedented variants were identified to be active towards the reaction between sodium pyruvate 2d and 3-FBA 7. The best variant 4M/H100L/D469T was able to convert 47% of substrate into product within 24 h, whereas no conversion was observed for WT. Docking was applied to illustrate the cooperation between mutations responsible for donor acceptance and the mutation responsible for acceptor acceptance to promote the activity towards both the acceptor and donor, and this provides a guide for the future work about simultaneously engineering two substrate specificity. The hits obtained in this study have the potential to be used for developing catalytic pathways to a diverse range of products such as the PACs, an important pharmaceutical intermediate in the synthesis of drugs including spisulosine and phenylpropanolamine. This study also suggested that, in order to engineer enzymes for simultaneously accepting two novel substrates, a useful strategy is to identify the variants improving the single substrate acceptance first and then combine them together.

Materials and methods

All chemicals were obtained from Sigma-Aldrich, UK, unless mentioned otherwise.

Preparation of lysates for enzyme reaction

The cell pellets from 12-mL cell culture were suspended using 1.2-mL cofactor solution containing 2.4 mm ThDP, 9 mm MgCl₂ and 50 mm Tris/HCl, pH 7.0. The suspended cells were then transferred to a 2-mL Eppendorf tube for cell lysis by sonication. The sonicated cells were centrifuged at 18 900 g for 15 min to collect the supernatant as the cell lysates. The protein expression level was assessed by 12% SDS/PAGE. Target bands were quantified using the GelDoc-It Imaging System (Cambridge, UK). Total protein concentrations in cell lysates were measured using the Bradford method [45]. Cell lysates were used at 2 mg·mL⁻¹ concentration in enzyme reactions.

Enzyme reactions

Enzyme reactions were carried out with lysates or purified proteins. Five acceptor substrates including glycolaldehyde 1, propanal 5a, pentanal 5b, hexanal 5c and 3-FBA 7 were used for reaction with the donor substrates sodium pyruvate 2d or keto-butyric acid 2e. Reactions between glycolaldehyde 1 and sodium pyruvate 2d were initiated by adding 50 or 100 µL of substrate solution containing 150 mm glycolaldehyde, 150 mm sodium pyruvate and 50 mm Tris/HCl, pH 7.0, into 100 µL of purified enzyme with the concentration of 0.2 mg·mL⁻¹ or 200 µL of enzyme lysates with the whole protein concentration of 2 mg·mL⁻¹, to give a final substrate concentration of 50 mm. Triplicate vials containing reaction solutions were incubated at 30 °C for 24 h. Samples were analysed by HPLC (Dionex, Sunnyvale, CA, USA) as described previously [12] to determine the concentration of DHB 3 and ERY 4 against a standard curve.

Reactions between propionaldehyde 5a and sodium pyruvate 2d were initiated by adding 50 or 100 µL of substrate solution containing 150 mm propanal, 150 mm sodium pyruvate 2d and 50 mm Tris/HCl, pH 7.0, into 100 µL of purified enzyme with the concentration of 0.2 mg·mL⁻¹ or 200 µL of enzyme lysates with the whole protein concentration of 2 mg·mL⁻¹, to give a final substrate concentrations of 50 mm. Triplicate vials containing reaction solutions were incubated at 30 °C for 24 h. A colorimetric assay detected the generation of the reaction product 3-hydroxy-2-pentanone 6ad using WST-1 instead of 2,3,5-triphenyltetrazolium chloride reported previously [46]. Compared to HPA, pyruvate does not reduce tetrazolium red, also removing the need to add carbonate resin for extracting excess HPA as described previously [46]. In a 96-well microplate, 50 µL of reaction mixture was diluted further with 50 µL of 50 mm Tris/HCl, pH 7.0, then with 20 µL of WST-1 (0.2%) solution, and finally with 10 µL of 3 m NaOH with good mixing. After incubation of 15 min at 25 °C, the microplate was scanned in a plate reader (BMG Fluostar Optima, Offenburg, Germany) for absorbance at 450 nm.

Reactions between aromatic aldehyde 7 and sodium pyruvate 2d were initiated as above by adding solution containing 150 mm 3-FBA 7, 150 mm sodium pyruvate 2d and 50 mm Tris/HCl, pH 7.0, into purified enzyme solution, to give final concentration of 50 mm 3-FBA 7 and 50 mm sodium pyruvate 2d. After reaction at 30 °C for 24 h, 20 µL of reaction mixture was transferred to 380 µL of 0.1% TFA to quench the reaction. Samples were then analysed by HPLC with an ACE5 C18 reverse-phase column (150 × 4.6 mm), UV detection at 210 and 250 nm, and a mobile phase at a flow rate of 1.0 mL·min⁻¹ starting with 85% of A (0.1% TFA) and 15% of B (acetonitrile). After 0.5 min, mobile phase B was linearly increased to 72% over 9 min, followed by a 2-min re-equilibration at 85% of 0.1% TFA and 15% acetonitrile.

The reactions between acceptor substrates pentanal 5b or hexanal 5c and the donor substrates sodium pyruvate 2d or keto-butyric acid 2e were also carried out for isolating the products and determining the stereoselectivities. The detailed procedures regarding TK reactions, isolation of the products and measurement of stereoselectivities are provided on p11 of the supporting information.

Site-directed mutagenesis and library construction

Site-directed mutagenesis was carried out with the QuikChange XL Site-Directed Mutagenesis Kit (Agilent Technologies, Santa Clara, CA, USA), according to the manufacturer’s instructions, on the TK gene in plasmid pQR791 [47]. XL10-Gold ultracompetent cells contained in the kit were transformed by the mutated plasmids. Mutagenic primers were designed using the web-based QuikChange Primer Design Program (www.agilent.com/genomics/qcpd).

A small library was constructed based on eight single mutants, H100F, H100L, H100T, D469E, D469T, D473N, D473S and R520Q, and combining them as described below to give 71 variants in total. Three rounds of mutagenesis were applied to construct the library. In the first round, site-directed mutagenesis was used to construct each of the following 12 variants, which were divided into three groups: group A included H100F, H100L, H100T; group B D469E, D469T, D473N, D473S, D469E/D473N, D469E/D473S, D469T/D473N, D469T/D473S; and group C R520Q. In the second round, pairwise recombination of the groups as AB, BC and AC was achieved by mixing templates from one group and primers from the other, to give 35 further variants. For instance, to combine group A and B mutations, the mixed PCR products of group A from the first round were used as the template, and the mixed mutagenic primers of group B were used as the primers for the Quikchange protocol. In the third round of PCR, the 24 possible new recombinants of groups ABC were constructed. PCR products from each round were mixed and transformed into ultracompetent cells. The total library contained eight confirmed single mutants, four confirmed double mutants and 180 random colonies containing double, triple and quadruple mutations. The library was then screened and the hits were sequenced.

Library screening

Colonies from agar plates were picked and transferred to 2 mL in 96 deep-well square plates (DWP) filled with 900 µL of LB Amp⁺ medium. Three wells were inoculated with blank cells without the TK gene as the negative control. DWP plates were sealed with breathable sterile film (VWR International, Radnor, PA, USA) and incubated at 37 °C, 400 r.p.m., for 18 h in a humidified shaker. The final OD₆₀₀ was measured from 50-µL cell culture transferred into a 96-well microplate. The remaining cell culture was centrifuged at 1792 g for 30 min to collect cell pellets, frozen at –80 °C and thawed at room temperature for freeze-thaw lysis of cells. The cell pellets were also suspended in 200 µL lysis buffer containing 9 mm MgCl₂, 2.4 mm ThDP, 1 × bugbuster and 1% lysonase (Merck, Hertfordshire, UK) in 50 mm Tris/HCl, pH 7.0, and then incubated at 25 °C for 30 min. Lysates were then transferred to a normal 96-well plate and centrifuged at 1792 g for 30 min, and 100 µL of supernatant was used in enzyme reactions in glass vials, initiated by adding 50 µL of aliphatic substrate solution containing 150 mm sodium pyruvate 2d, 150 mm propionaldehyde 5a, 50 mm Tris/HCl, pH 7.0, or 50 µL aromatic substrate solution containing 150 mm sodium pyruvate 2d, 150 mm 3- FBA 7, 50 mm Tris/HCl, pH 7.0. Reactions were incubated at 30 °C for 24 h before detection. For reactions with sodium pyruvate 2d and propionaldehyde 5a, the colorimetric assay described above was used to detect the product. For reactions with sodium pyruvate 2d and 3-FBA 7, 20-µL samples were transferred to 380 µL of 0.1% TFA to quench the reaction, centrifuged at 18 900 g for 15 min and then 200 µL of the solution was analysed by HPLC as mentioned above.

Semi-preparative HPLC

The target peak detected by analytical HPLC was collected using semi-preparative HPLC on an Agilent ZORBAX 300SB-C18 (250 mm × 9.4 mm, 5 µm) reverse-phase column at 25 °C, with 2 mL·min⁻¹ flow rate and detection at 250 nm. The two-column mobile phases were water with 0.1% TFA (A) and acetonitrile with 0.1% TFA (B), and the column was pre-equilibrated with 85% A, 15% B. The reaction mixture was diluted 5-fold using 0.1 % TFA and then 200 µL was loaded onto the column. A gradient was then applied as 85% A, 15% B at 0–5 min, then linearly ramped to 28% A, 72% B over 15 min, and then linearly back to 85% A, 15% B over 5 min.

Gas chromatography

After the enzyme reaction, 500 µL of chloroform was added and the mixture agitated, and then centrifuged at 1792 g for 3 min. The organic phase was then pipetted out carefully and analysed by GC/MS with an ISQ series single Quadrupole GC/MS system (ThermoFisher Scientific, Waltham, MA, USA), equipped with a Rxi-5Sil MS column (30 m × 0.2 mm I.D.) and the helium carrier gas at a flow rate of 1 mL·min⁻¹. The column temperature was kept at 40 °C for 3 min, then raised to 150 °C at a rate of 2 °C per min and then to 200 °C at a rate of 4 °C per min. The ion source temperature was 230 °C and the range of molecular weight scanned was 20–350 m/z. GC analyses were also carried out with an Agilent 7820A GC-system and a chiral stationary phase [Supelco RT BetaDEX225 column (30 m × 0.25 mm, 0.25 µm)], and ee measurements were determined with the following method: starting temperature 100 °C and an increase of 5 °C per min until 135 °C, and then 40 °C per min to 210 °C.

Liquid chromatography/mass spectrometry

MS analysis was performed in the UCL Chemistry Mass Spectrometry Facility. The molecular masses of the compounds were measured on an Acquity UPLC System (Waters, Milford, MA, USA) with an SQ mass spectrometer (Waters). The HPLC was performed with a Thermo Scientific Hypersil Gold™ C4 reverse-phase column (50 × 2.1 mm) with 0.2 mL·min⁻¹ flow rate. The mobile phases were 0.1% (v/v) formic acid (A) and 95% acetonitrile-5% water (B), equilibrated at 95% A, 5% B, and then a gradient applied after sample injection to 5% A, 95% B over 4 min, then back to 95% A, 5% B over 0.5 min, and held for a further 0.5 min. Compounds were ionised using ESI and detected in positive mode. The operating conditions of ESI interface were set to a capillary temperature of 300 °C, capillary voltage 9 V and spray voltage 4 kV.

Nuclear magnetic resonance spectroscopy

¹H NMR spectra were recorded on Bruker Avance 400, 500, 600 and 700 MHz spectrometers at 25 °C, using the residual protic solvent stated as the internal standard. Chemical shifts are quoted in p.p.m. to the nearest 0.01 p.p.m. using the following abbreviations: s (singlet), d (doublet), t (triplet), q (quartet), qn (quintet), sext (sextet), dd (doublet of doublets), dt (doublet of triplets) and m (multiplet) defined as all multi-peak signals where overlap or complex coupling of signals make definitive descriptions of peaks difficult. The coupling constants are defined as J and quoted in Hz. ¹³C NMR spectra were recorded at 100, 125, 150 or 175 MHz on Bruker Avance 400, 500, 600 and 700 MHz spectrometers at 25 °C using the stated solvent as standard. Chemical shifts are reported to the nearest 0.1 p.p.m. Melting points were measured with a Gallenkamp apparatus and were uncorrected.

Structure alignment

The TK structure 2R8O.pdb and PDC structure 2VK1.pdb were aligned using the server, TopMatch-Web [48, 49]. The aligned structure from output was visualised using Pymol (Schrödinger, New York, NY, USA). Sequence identity was calculated online by the web server Sequence Identity And Similarity (http://imed.med.ucm.es/Tools/sias.html) with the input of 2R8O.pdb and 2VK1.pdb.

Computational Molecular Docking

The structures of the variants H192P/A282P/I365L/G506A/H100L/D469E (4M/HL/DE) and H192P/A282P/I365L/G506A/H100L/D469T(4M/HL/DT) were obtained using SWISS-MODEL and 1QGD.pdb as the template [50]. ThDP-enamine, propionaldehyde 5a and 3- FBA 7 structures were drawn in Chem3D Ultra v.13.0 and energy minimised by MM2 calculations. Autodock Vina was used to sequentially dock the ThDP-enamine and propionaldehyde 5a or ThDP-enamine and 3- FBA 7 into WT and mutant TKs with the grid centred at −11.25, 25.858 and 40.198 and the exhaustiveness of 24 [51, 52]. The grid sizes of 30 Å × 30 Å × 30 Å, 10 Å × 10 Å × 10 Å and 28 Å × 28 Å × 28 Å were used for the docking of ThDP-enamine, propionaldehyde 5a and 3-FBA 7, respectively. Five replicate dockings were carried out and the conformations with lowest energy were selected for analysis.

Acknowledgements

This work was supported in part by a Chinese Scholarship Council stipend (to HY) and Engineering and Physical Sciences Research Council Grants EP/N025105/1 and EP/P006485/1. We gratefully acknowledge the Department of Chemistry at University College London for funding SH and DS, the Biotechnology and Biosciences Research Council (BBSRC) (BB/N01877X/1) for funding DMS and CONACYT is thanked for a PhD studentship to ACK.

Conflicts of interest

The authors declare no conflict of interest.

Author contributions

PAD designed the research. HY, RIHL, DS, DMS, SH and ACK performed the experiments. HY, PAD and HCH interpreted the data. PAD, HCH, TDS and JMW supervised the studies. HY, PAD and HCH wrote the paper. All authors reviewed, revised and approved the manuscript for publication.

Supporting Information

Filename

Description

febs15108-sup-0001-Supinfo.zipZip archive, 12.6 MB

Table S1. TK (2R8O.pdb) and PDC (2VK1.pdb) structure alignment parameters.

Table S2. Design of single variants.

Table S3. Characteristics of the variant 4M/HL/DT.

Fig. S1. Structure alignment between transketolase (2R8O.pdb) and pyruvate decarboxylase (2VK1.pdb).

Fig. S2. Densitometry to measure TK variants expression in the lysates with the protein concentration of 2 mg·mL⁻¹.

Fig. S3. GC/MS analysis of 3-hydroxy-2-pentanone.

Fig. S4. Correlation between colorimetric assay and the GC/MS measurement for the reactions yield.

Fig. S5. Activity correlation between reactions with sodium pyruvate as donor and with glycolaldehyde or propionaldehyde as acceptor.

Fig. S6. LC/MS analysis of the reaction product between 3-FBA and sodium pyruvate.

Fig. S7. Pose of ThDP-enamine intermediate in the binding pockets of wild-type and mutant TKs.

Fig. S8. Structural alignment of TK wild-type and two variants 4M/HL/DE, 4M/HL/DT.

Appendix S1. General experimental for synthesis and analysis.

Appendix S2. General procedures for TK reactions.

Appendix S3. Stereoselectivity measurements.

Appendix S4. NMR spectra.

Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.

References

1Reetz MT, Bocola M, Carballeira JD, Zha D & Vogel A (2005) Expanding the range of substrate acceptance of enzymes: combinatorial active-site saturation test. Angew Chem Int Ed Engl 44, 4192–4196.
10.1002/anie.200500767
CASPubMedWeb of Science®Google Scholar
2Sprenger GA, Schorken U, Sprenger G & Sahm H (1995) Transketolase A of Escherichia coli K12. Purification and properties of the enzyme from recombinant strains. Eur J Biochem 230, 525–532.
10.1111/j.1432-1033.1995.0525h.x
CASPubMedWeb of Science®Google Scholar
3Draths KM, Pompliano DL, Conley DL, Frost JW, Berry A, Disbrow GL, Staversky RJ & Lievense JC (1992) Biocatalytic synthesis of aromatics from D-Glucose - the role of transketolase. J Am Chem Soc 114, 3956–3962.
10.1021/ja00036a050
CASWeb of Science®Google Scholar
4Demuynck C, Bolte J, Hecquet L & Dalmas V (1991) Enzyme-Catalyzed synthesis of carbohydrates - synthetic potential of transketolase. Tetrahedron Lett 32, 5085–5088.
10.1016/S0040-4039(00)93434-8
CASWeb of Science®Google Scholar
5Toone EJ & Whitesides GM (1991) Enzymes as catalysts in carbohydrate synthesis. Acs Sym Ser 466, 1–22.
10.1021/bk-1991-0466.ch001
CASWeb of Science®Google Scholar
6Morris KG, Smith MEB, Turner NJ, Lilly MD, Mitra RK & Woodley JM (1996) Transketolase from Escherichia coli: a practical procedure for using the biocatalyst for asymmetric carbon-carbon bond synthesis. Tetrahedron-Asymmetr 7, 2185–2188.
10.1016/0957-4166(96)00266-2
CASWeb of Science®Google Scholar
7Wohlgemuth R, Smith MEB, Dalby PA & Woodley JM (2010) Transketolases. In Encyclopedia of Industrial Biotechnology: Bioprocess, Bioseparation and Cell Technology, ( MC Flickinger, ed), Vol. 7, pp. 4746–4752. John Wiley and Sons Inc, Hoboken.
Google Scholar
8Turner NJ (2000) Applications of transketolases in organic synthesis. Curr Opin Biotechnol 11, 527–531.
10.1016/S0958-1669(00)00140-3
CASPubMedWeb of Science®Google Scholar
9Kobori Y, Myles DC & Whitesides GM (1992) Substrate-specificity and carbohydrate synthesis using transketolase. J Org Chem 57, 5899–5907.
10.1021/jo00048a023
CASWeb of Science®Google Scholar
10Schenk G, Duggleby RG & Nixon PF (1998) Properties and functions of the thiamin diphosphate dependent enzyme transketolase. Int J Biochem Cell Biol 30, 1297–1318.
10.1016/S1357-2725(98)00095-8
CASPubMedWeb of Science®Google Scholar
11Smith MEB, Hibbert EG, Jones AB, Dalby PA & Hailes HC (2008) Enhancing and reversing the stereoselectivity of Escherichia coli transketolaseviasingle-point mutations. Adv Synth Catal 350, 2631–2638.
10.1002/adsc.200800489
CASWeb of Science®Google Scholar
12Hibbert EG, Senussi T, Costelloe SJ, Lei W, Smith ME, Ward JM, Hailes HC & Dalby PA (2007) Directed evolution of transketolase activity on non-phosphorylated substrates. J Biotechnol 131, 425–432.
10.1016/j.jbiotec.2007.07.949
CASPubMedWeb of Science®Google Scholar
13Hibbert EG, Senussi T, Smith ME, Costelloe SJ, Ward JM, Hailes HC & Dalby PA (2008) Directed evolution of transketolase substrate specificity towards an aliphatic aldehyde. J Biotechnol 134, 240–245.
10.1016/j.jbiotec.2008.01.018
CASPubMedWeb of Science®Google Scholar
14Galman JL, Steadman D, Bacon S, Morris P, Smith ME, Ward JM, Dalby PA & Hailes HC (2010) α, α'-Dihydroxyketone formation using aromatic and heteroaromatic aldehydes with evolved transketolase enzymes. Chem Commun (Camb) 46, 7608–7610.
10.1039/c0cc02911d
CASPubMedWeb of Science®Google Scholar
15Cazares A, Galman JL, Crago LG, Smith ME, Strafford J, Rios-Solis L, Lye GJ, Dalby PA & Hailes HC (2010) Non-alpha-hydroxylated aldehydes with evolved transketolase enzymes. Org Biomol Chem 8, 1301–1309.
10.1039/b924144b
CASPubMedWeb of Science®Google Scholar
16Payongsri P, Steadman D, Strafford J, MacMurray A, Hailes HC & Dalby PA (2012) Rational substrate and enzyme engineering of transketolase for aromatics. Org Biomol Chem 10, 9021–9029.
10.1039/c2ob25751c
CASPubMedWeb of Science®Google Scholar
17Payongsri P, Steadman D, Hailes HC & Dalby PA (2015) Second generation engineering of transketolase for polar aromatic aldehyde substrates. Enzyme Microb Technol 71, 45–52.
10.1016/j.enzmictec.2015.01.008
CASPubMedWeb of Science®Google Scholar
18Zhou C, Saravanan T, Lorilliere M, Wei D, Charmantray F, Hecquet L, Fessner WD & Yi D (2017) Second-generation engineering of a Thermostable Transketolase (TKGst ) for aliphatic aldehyde acceptors with either improved or reversed stereoselectivity. ChemBioChem 18, 455–459.
10.1002/cbic.201600609
CASPubMedWeb of Science®Google Scholar
19Zabar JA, Lorilliere M, Yi D, Thangavelu S, Devamani T, Nauton L, Charmantray F, Helaine V, Fessner WD & Hecquet L (2015) Engineering a thermostable transketolase for unnatural conversion of (2S)-hydroxyaldehydes. Adv Synth Catal 357, 1715–1720.
10.1002/adsc.201500207
CASWeb of Science®Google Scholar
20Yi D, Saravanan T, Devamani T, Charmantray F, Hecquet L & Fessner WD (2015) A thermostable transketolase evolved for aliphatic aldehyde acceptors. Chem Commun (Camb) 51, 480–483.
10.1039/C4CC08436E
CASPubMedWeb of Science®Google Scholar
21Saravanan T, Reif ML, Yi D, Lorilliere M, Charmantray F, Hecquet L & Fessner WD (2017) Engineering a thermostable transketolase for arylated substrates. Green Chem 19, 481–489.
10.1039/C6GC02017H
CASWeb of Science®Google Scholar
22Fiedler E, Golbik R, Schneider G, Tittmann K, Neef H, König S & Hübner G (2001) Examination of donor substrate conversion in yeast transketolase. J Biol Chem 276, 16051–16058.
10.1074/jbc.M007936200
CASPubMedWeb of Science®Google Scholar
23Sprenger GA & Pohl M (1999) Synthetic potential of thiamin diphosphate-dependent enzymes. J Mol Catal B-Enzym 6, 145–159.
10.1016/S1381-1177(98)00107-6
CASWeb of Science®Google Scholar
24Esakova OA, Meshalkina LE, Kochetov GA & Golbik R (2009) Halogenated pyruvate derivatives as substrates of transketolase from Saccharomyces cerevisiae. Biochemistry (Mosc) 74, 1234–1238.
10.1134/S0006297909110091
CASPubMedWeb of Science®Google Scholar
25Saravanan T, Junker S, Kickstein M, Hein S, Link MK, Ranglack J, Witt S, Lorilliere M, Hecquet L & Fessner WD (2017) Donor promiscuity of a thermostable transketolase by directed evolution: efficient complementation of 1-Deoxy-d-xylulose-5-phosphate synthase activity. Angew Chem Int Ed Engl 56, 5358–5362.
10.1002/anie.201701169
CASPubMedWeb of Science®Google Scholar
26Widmann M, Radloff R & Pleiss J (2010) The thiamine diphosphate dependent enzyme engineering database: a tool for the systematic analysis of sequence and structure relations. BMC Biochem 11, 9.
10.1186/1471-2091-11-9
CASPubMedWeb of Science®Google Scholar
27Hailes HC, Rother D, Muller M, Westphal R, Ward JM, Pleiss J, Vogel C & Pohl M (2013) Engineering stereoselectivity of ThDP-dependent enzymes. FEBS J 280, 6374–6394.
10.1111/febs.12496
CASPubMedWeb of Science®Google Scholar
28Sandford V, Breuer M, Hauer B, Rogers P & Rosche B (2005) (R)-phenylacetylcarbinol production in aqueous/organic two-phase systems using partially purified pyruvate decarboxylase from Candida utilis. Biotechnol Bioeng 91, 190–198.
10.1002/bit.20513
CASPubMedWeb of Science®Google Scholar
29Meyer D, Walter L, Kolter G, Pohl M, Muller M & Tittmann K (2011) Conversion of pyruvate decarboxylase into an enantioselective carboligase with biosynthetic potential. J Am Chem Soc 133, 3609–3616.
10.1021/ja110236w
CASPubMedWeb of Science®Google Scholar
30Asztalos P, Parthier C, Golbik R, Kleinschmidt M, Hubner G, Weiss MS, Friedemann R, Wille G & Tittmann K (2007) Strain and near attack conformers in enzymic thiamin catalysis: X-ray crystallographic snapshots of bacterial transketolase in covalent complex with donor ketoses xylulose 5-phosphate and fructose 6-phosphate, and in noncovalent complex with acceptor aldose ribose 5-phosphate. Biochemistry 46, 12037–12052.
10.1021/bi700844m
CASPubMedWeb of Science®Google Scholar
31Wang J, Dong H, Li S & He H (2005) Theoretical study toward understanding the catalytic mechanism of pyruvate decarboxylase. J Phys Chem B 109, 18664–18672.
10.1021/jp052802s
CASPubMedWeb of Science®Google Scholar
32Pohl M, Siegert P, Mesch K, Bruhn H & Grotzinger J (1998) Active site mutants of pyruvate decarboxylase from Zymomonas mobilis–a site-directed mutagenesis study of L112, I472, I476, E473, and N482. Eur J Biochem 257, 538–546.
10.1046/j.1432-1327.1998.2570538.x
CASPubMedWeb of Science®Google Scholar
33Wikner C, Meshalkina L, Nilsson U, Backstrom S, Lindqvist Y & Schneider G (1995) His103 in yeast transketolase is required for substrate recognition and catalysis. Eur J Biochem 233, 750–755.
10.1111/j.1432-1033.1995.750_3.x
CASPubMedWeb of Science®Google Scholar
34Strafford J, Payongsri P, Hibbert EG, Morris P, Batth SS, Steadman D, Smith ME, Ward JM, Hailes HC & Dalby PA (2012) Directed evolution to re-adapt a co-evolved network within an enzyme. J Biotechnol 157, 237–245.
10.1016/j.jbiotec.2011.11.017
CASPubMedWeb of Science®Google Scholar
35Sevostyanova IA, Solovjeva ON & Kochetov GA (2004) A hitherto unknown transketolase-catalyzed reaction. Biochem Biophys Res Commun 313, 771–774.
10.1016/j.bbrc.2003.11.164
CASPubMedWeb of Science®Google Scholar
36Steinmetz A, Vyazmensky M, Meyer D, Barak ZE, Golbik R, Chipman DM & Tittmann K (2010) Valine 375 and phenylalanine 109 confer affinity and specificity for pyruvate as donor substrate in acetohydroxy acid synthase isozyme II from Escherichia coli. Biochemistry 49, 5188–5199.
10.1021/bi100555q
CASPubMedWeb of Science®Google Scholar
37Wille G, Meyer D, Steinmetz A, Hinze E, Golbik R & Tittmann K (2006) The catalytic cycle of a thiamin diphosphate enzyme examined by cryocrystallography. Nat Chem Biol 2, 324–328.
10.1038/nchembio788
CASPubMedWeb of Science®Google Scholar
38Yi D, Devamani T, Abdoul-Zabar J, Charmantray F, Helaine V, Hecquet L & Fessner WD (2012) A pH-based high-throughput assay for transketolase: fingerprinting of substrate tolerance and quantitative kinetics. ChemBioChem 13, 2290–2300.
10.1002/cbic.201200364
CASPubMedWeb of Science®Google Scholar
39Pechal M, Cvengrosova Z, Holotik S, Malik L & Hrusovsky M (1981) Identification and determination of alpha-hydroxyketones in a reaction mixture. J Chromatogr 206, 541–547.
10.1016/S0021-9673(00)88923-2
CASWeb of Science®Google Scholar
40Shi X, Leal WS & Meinwald J (1996) Assignment of absolute stereochemistry to an insect pheromone by chiral amplification. Bioorg Med Chem 4, 297–303.
10.1016/0968-0896(96)00006-5
CASPubMedWeb of Science®Google Scholar
41Bloom JD, Labthavikul ST, Otey CR & Arnold FH (2006) Protein stability promotes evolvability. Proc Natl Acad Sci USA 103, 5869–5874.
10.1073/pnas.0510098103
CASPubMedWeb of Science®Google Scholar
42Yu H & Dalby PA (2018) Coupled molecular dynamics mediate long- and short-range epistasis between mutations that affect stability and aggregation kinetics. Proc Natl Acad Sci USA 115, E11043–E11052.
10.1073/pnas.1810324115
CASPubMedWeb of Science®Google Scholar
43Yu H & Dalby PA (2018) Exploiting correlated molecular-dynamics networks to counteract enzyme activity-stability trade-off. Proc Natl Acad Sci USA 115, E12192–E12200.
10.1073/pnas.1812204115
CASPubMedWeb of Science®Google Scholar
44Trott O & Olson AJ (2010) AutoDock Vina: improving the speed and accuracy of docking with a new scoring function, efficient optimization, and multithreading. J Comput Chem 31, 455–461.
10.1002/jcc.21334
CASPubMedWeb of Science®Google Scholar
45Bradford MM (1976) A rapid and sensitive method for the quantitation of microgram quantities of protein utilizing the principle of protein-dye binding. Anal Biochem 72, 248–254.
10.1016/0003-2697(76)90527-3
CASPubMedWeb of Science®Google Scholar
46Smith ME, Kaulmann U, Ward JM & Hailes HC (2006) A colorimetric assay for screening transketolase activity. Bioorg Med Chem 14, 7062–7065.
10.1016/j.bmc.2006.06.008
CASPubMedWeb of Science®Google Scholar
47Martinez-Torres RJ, Aucamp JP, George R & Dalby PA (2007) Structural stability of E. coli transketolase to urea denaturation. Enzyme Microb Technol 41, 653–662.
10.1016/j.enzmictec.2007.05.019
CASWeb of Science®Google Scholar
48Sippl MJ & Wiederstein M (2008) A note on difficult structure alignment problems. Bioinformatics 24, 426–427.
10.1093/bioinformatics/btm622
CASPubMedWeb of Science®Google Scholar
49Sippl MJ & Wiederstein M (2012) Detection of spatial correlations in protein structures and molecular complexes. Structure 20, 718–728.
10.1016/j.str.2012.01.024
CASPubMedWeb of Science®Google Scholar
50Waterhouse A, Bertoni M, Bienert S, Studer G, Tauriello G, Gumienny R, Heer FT, de Beer TAP, Rempfer C, Bordoli L et al. (2018) SWISS-MODEL: homology modelling of protein structures and complexes. Nucleic Acids Res 46, W296–W303.
10.1093/nar/gky427
CASPubMedWeb of Science®Google Scholar
51Forli S, Huey R, Pique ME, Sanner MF, Goodsell DS & Olson AJ (2016) Computational protein-ligand docking and virtual drug screening with the AutoDock suite. Nat Protoc 11, 905–919.
10.1038/nprot.2016.051
CASPubMedWeb of Science®Google Scholar
52Subrizi F, Cárdenas-Fernández M, Lye GJ, Ward JM, Dalby PA, Sheppard TD & Hailes HC (2016) Transketolase catalysed upgrading of l-arabinose: the one-step stereoselective synthesis of l-gluco-heptulose. Green Chem 18, 3158–3165.
10.1039/C5GC02660A
CASWeb of Science®Google Scholar

Citing Literature

Volume287, Issue9

May 2020

Pages 1758-1776

Journal list menu

Tools

Follow journal

Engineering transketolase to accept both unnatural donor and acceptor substrates and produce α-hydroxyketones

Abstract

Abbreviations

Introduction

Results and Discussion

Design of single mutants

Characterisation of single variants

Reaction with glycolaldehyde and pyruvate

Reaction with propionaldehyde and pyruvate and extension to other aliphatic analogues to determine stereoselectivities

Design of a combined library

Improvement of activity towards propionaldehyde and pyruvate

Novel enzyme activity towards 3-FBA and pyruvate

Elucidating the mechanism of enhanced activity

Conclusion

Materials and methods

Preparation of lysates for enzyme reaction

Enzyme reactions

Site-directed mutagenesis and library construction

Library screening

Semi-preparative HPLC

Gas chromatography

Liquid chromatography/mass spectrometry

Nuclear magnetic resonance spectroscopy

Structure alignment

Computational Molecular Docking

Acknowledgements

Conflicts of interest

Author contributions

Supporting Information

References

Citing Literature

Figures

References

Related

Information