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

Article abstract of DOI:10.1111/febs.15108

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.

Full text of DOI:10.1111/febs.15108

DR HAORAN YU (Orcid ID : 0000-0001-9012-4688)
Received Date : 09-Aug-2019
Revised Date : 26-Sep-2019
Accepted Date : 23-Oct-2019
Color : Fig 1-7
This article has been accepted for publication and undergone full peer review but has not been through the
copyediting, typesetting, pagination and proofreading process, which may lead to differences between this
version and the Version of Record. Please cite this article as doi: 10.1111/FEBS.15108
This article is protected by copyright. All rights reserved

SuppInfo : 1 Supinfo
Engineering transketolase to accept both unnatural donor and acceptor
substrates and produce α-hydroxyketones
Haoran Yua, Roberto Icken Hernández Lópeza, David Steadmanb, Daniel Méndez-Sánchezb, Sally
Higsonb, Armando Cázares-Körnerb, Tom D. Sheppardb, John M. Warda, Helen C. Hailesb, Paul A.
Dalbya*
a Department of Biochemical Engineering, University College London, Gordon Street, London,
WC1H 0AH, United Kingdom
b Department of Chemistry, University College London, 20 Gordon Street, London WC1H 0AJ,
United Kingdom
*Correspondence and requests for materials should be addressed to Paul A. Dalby.
Department of Biochemical Engineering, University College London, Gordon Street, London,
WC1H 0AH, United Kingdom. Email: p.dalby@ucl.ac.uk; Tel: +44 (0)20 7679 9566
Running Title: Engineer substrate specificity of transketolase
Keywords: protein engineering, rational design, transketolase, two substrates specificity, docking
Abbreviations: DXS, 1-deoxy-D-xylulose-5-phosphate synthase; EE, enantiomeric excess; FBA,
formylbenzoic acid; HPA, hydroxypyruvate; NMR, nuclear magnetic resonance spectroscopy; PAC,
phenylacetylcarbinol; ThDP, thiamine diphosphate; TK, transketolase; TKgst, TK from Geobacillus
stearothermophilus; WT, wild type
This article is protected by copyright. All rights reserved

Abstract
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 optimization of
both substrate specificities is also required for the rapid expansion of accepted substrates.
Transketolase 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, transketolase from E. coli was engineered to accept both pyruvate as a novel donor
substrate, and unnatural acceptor aldehydes, including propanal, pentanal, hexanal and 3-
formylbenzoic acid. Twenty single-mutant variants were first designed and characterized
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-formylbenzoic acid. 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.
This article is protected by copyright. All rights reserved

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
optimization 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 lyophilized powder
has been established at Sigma-Aldrich [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 transketolase toolbox is desired, in which the
variants should be able to accept a diverse range of substrates including both acceptors and
donors.
Wild-type TK accepts a narrow range of hydroxyaldehydes as the aldol-acceptor substrate, and
with strict (2R)-specificity [8-10]. TK from E. coli has been engineered successfully for improved
and reversed enantioselectivity [11], as well as an expanded aldol-acceptor substrate range
including polar aliphatic [12], non-polar 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 (TKgst) [18-21].
This article is protected by copyright. All rights reserved

Wild-type 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-hydroxypyruvate 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 hydroxypyruvate only by the absence of the C-3-hydroxyl group, and yet
cannot function as a substrate in the transketolase 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 (DXS), catalyzes 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 TKgst 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 TKgst 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 wild-type 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.
This article is protected by copyright. All rights reserved

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 was 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 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]. Hydroxypyruvate 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 stabilize new donor substrates lacking the C-1-hydroxyl
This article is protected by copyright. All rights reserved

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, His473) within 5 Å of the first two carbons
of the donor substrate, as potential mutation sites (Fig. 1A).
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 Cα 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 the 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 stabilization
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-
hydroxy group in TK and played a role in anchoring the donor substrate. The equivalent residue
in TKgst was His474, and the mutations H474S and H474N were previously found respectively to
give a 6-fold and 3-fold improved activity towards glycolaldehyde 1 and pyruvate 2d. 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
This article is protected by copyright. All rights reserved

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 approximately
corresponding 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 Km for acceptor substrates, but
significantly increased 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 TKgst 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
This article is protected by copyright. All rights reserved

20 single mutants (summarised in Table S2) and measured their activities towards glycolaldehyde
1 or propionaldehyde 5a and pyruvate 2d experimentally.
Characterization 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, paired with glycolaldehyde 1 as this is known already to be one of the best
acceptor substrates for wild-type 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, Fig. 2). However,
erythrulose 4 was also detected as a side product at the retention time of 11.4 min (Scheme 1,
Fig. 2 & Fig. 3A), which was not reported in any previous mutagenesis studies, although the
reaction of two glycolaldehydes to form erythrulose is a known slow side reaction for wild-type
TK [35].
Considering the possible difference in expression level of the variants, the yields of both products
determined after 24-h 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
erythrulose byproduct, 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, 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
This article is protected by copyright. All rights reserved

enzyme form, compared to wild type (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 TKgst 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 TKgst (Fig. 3C). The contextual structure
differences between the active sites of E. coli TK and TKgst are a likely reason for these differences.
However, it should be noted that differences in the activity assays used in our work compared to
that in previous reports with TKgst, could also lead to similar observations. The pH-sensitive
colorimetric assay used for TKgst would not have detected the erythrulose by-product, nor
distinguish between the decarboxylation of pyruvate 2d versus true transketolase function
(decarboxylation followed by ketone unit transfer to acceptor) [25, 38].
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 byproduct erythrulose 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
This article is protected by copyright. All rights reserved

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 non-polar 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 byproduct 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 and a high Km of 140 mM [35]. In contrast, TKgst
H473N had a Km of 13.4 mM towards pyruvate 2d, lower than that of TKgst WT (40.2 mM) and
even the TKgst H100L (29.1 mM) [25]. These 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 byproduct 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 3,4-dihydroxy-2-butanone 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. Wild-type E. coli TK can accept a range of non-hydroxylated aliphatic aldehyde
substrates in the presence of hydroxypyruvate. However, the activity is typically only 5–35% of
This article is protected by copyright. All rights reserved

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 above (Fig. 3C), were first tested
for their activity in lysates, to identify any to add into the pool of purified variants. 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, 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 having 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-hours reaction. A peak was
identified by GC with the retention time of 8.9 min (Fig. S3A). The MS data confirm the successful
generation of the target product 3-hydroxy-2-pentanone 6ad. In this compound, fragmentation
between C-2 and C-3 lead to two species with m/z of 59 and 43. Further fragmentation of 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 hours using seven variants,
This article is protected by copyright. All rights reserved

including both the improved and non-improved 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-fold, 2.4-fold 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 & 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, 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 in both cases in approximately 5% yield, 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 NMR
analysis revealed a major singlet for the 1-H3 group at 2.21 ppm, and a minor peak at 2.13 ppm
(Appendix S4). The positions of these two signals corresponded to the (2S,3S) and (2S,3R)-esters,
This article is protected by copyright. All rights reserved

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 ppm
(major (2S,3S)-isomer) and 0.93 ppm (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 ppm
(major (2S,4S)-isomer) and 0.94 ppm (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 & 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
This article is protected by copyright. All rights reserved

towards the more hydrophobic pyruvate (relative to hydroxypyruvate), 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 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
This article is protected by copyright. All rights reserved

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, H100F, H100L, H100T,
D469E, D469T, H473N, H473S each 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
This article is protected by copyright. All rights reserved

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
the 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, 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
This article is protected by copyright. All rights reserved

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 characterized 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),
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 characterized at higher enzyme concentrations. When increasing the concentration of
enzyme from 0.067 mg/mL to 1.330 mg/mL, the conversion from 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 (Table 1 & S3). The specific activity, Km for pyruvate 2d and a kcat of
4M/H100L/D469T were also 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 Km [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
This article is protected by copyright. All rights reserved

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 wild-type and mutant TKs (Fig. S7A). ThDP-enamine had
the binding affinity energy of -9.5 kcal mol-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 pose of ThDP-enamine was similar in
the binding pockets of 4M/HL/DE and 4M/HL/DT with a RMSD of 0.3 Å, and yet slightly to the
binding in WT with a RMSD of 0.7 Å (Fig. S7A, Table 2). When the pyruvate 2d instead of
hydroxypyruvate 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 the position 100 would reduce the polarity of
this local region and form a hydrophobic interaction between L100 and C2 methyl group of the
This article is protected by copyright. All rights reserved

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.
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 Å, 4.6 Å 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, C-3 atoms of
propanal have 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, C-3 atoms of propanal are non-polar. 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 Å 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
This article is protected by copyright. All rights reserved

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 mol-1, 0.1 kcal mol-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
non-polar 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 generate the mutation that improves the activity towards pyruvate and 3-FBA.
Conclusion
In this study, we attempted to engineer the transketolase 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 towards two reactions
separately: 1) sodium pyruvate 2d and propionaldehyde 5a; 2) 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
This article is protected by copyright. All rights reserved

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
phenylacetylcarbinols, 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 thiamine diphosphate (ThDP), 9 mM MgCl2, 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 13000 rpm for 15 min to collect the supernatant as the cell lysates. The
protein expression level was assessed by 12% sodium dodecyl sulfate-polyacrylamide gel
electrophoresis (SDS-PAGE). Target bands were quantified using the GelDoc-It Imaging System
This article is protected by copyright. All rights reserved

(Cambridge, UK). Total protein concentrations in cell lysates were measured using the Bradford
method [45]. Cell lysates were used at 2 mg/mL 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 ketobutyric acid 2e. Reactions
between glycolaldehyde 1 and sodium pyruvate 2d were initiated by adding 50 µL or 100 µL
substrate solution containing 150 mM glycolaldehyde, 150 mM sodium pyruvate, 50 mM Tris-HCl,
pH 7.0, into 100 µL purified enzyme with the concentration of 0.2 mg/mL or 200 µL enzyme
lysates with the whole protein concentration of 2 mg/mL, to give a final substrate concentrations
of 50 mM. Triplicate vials containing reaction solutions were incubated at 30 oC for 24 h. Samples
were analysed by HPLC (Dionex, CA, USA) as previously [12] to determine the concentration of
3,4-dihydroxy-2-butanone 3 and erythrulose 4 against a standard curve.
Reactions between propionaldehyde 5a and sodium pyruvate 2d were initiated by adding 50 µL
or 100 µL substrate solution containing 150 mM propanal, 150 mM sodium pyruvate 2d, 50 mM
Tris-HCl, pH 7.0, into 100 µL purified enzyme with the concentration of 0.2 mg/mL or 200 µL
enzyme lysates with the whole protein concentration of 2 mg/mL, to give a final substrate
concentrations of 50 mM. Triplicate vials containing reaction solutions were incubated at 30 oC
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 the tetrazolium red, also removing the need to
add carbonate resin for extracting excess HPA as previously [46]. In a 96-well microplate, 50 µL
reaction mixture was diluted further with 50 µL 50 mM Tris-HCl pH 7.0, then 20 µL WST-1 (0.2%)
solution and finally 10 µL 3 M NaOH with good mixing. After incubation of 15 min at 25 oC, the
This article is protected by copyright. All rights reserved

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, 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 oC for 24 h, 20 µL reaction mixture were transferred to 380 µL
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 nm and 250 nm, and a mobile phase at a flow
rate of 1.0 ml/min 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-minute 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 of TK reactions, isolating the
products and measuring stereoselectivities were provided in the supporting information (see SI).
Site-directed mutagenesis and library construction
Site-directed mutagenesis was carried out with the QuikChange XL Site-Directed Mutagenesis Kit
(Agilent Technologies, US), 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 the 8 single mutants H100F, H100L, H100T, D469E,
D469T, D473N, D473S, R520Q, and combining them as described below to give 71 variants in
This article is protected by copyright. All rights reserved

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, divided into three
groups: group A were H100F, H100L, H100T, group B were D469E, D469T, D473N, D473S,
D469E/D473N, D469E/D473S, D469T/D473N, D469T/D473S and group C was 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 8 confirmed single mutants, 4
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 96 deep-well square plates (DWP)
filled with 900 µL 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 (VMR
International, US) and incubated at 37 °C, 400 rpm 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 4000 rpm 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 oC for 30 min.
Lysates were then transferred to a normal 96-well plate and centrifuged at 4000 rpm for 30 min,
and 100 µL supernatant was used in enzyme reactions in glass vials, initiated by adding 50 µL
aliphatic substrate solution containing 150 mM sodium pyruvate 2d, 150 mM propionaldehyde
This article is protected by copyright. All rights reserved

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 oC 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 0.1% TFA to quench the
reaction, centrifuged at 13000 rpm for 15 min, then 200 µL analysed by HPLC as 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 oC, with 2
mL/min 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), 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 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 chloroform was added and the mixture agitated, then
centrifuged at 4000 rpm 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,
MA, US), 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. The column temperature was kept at 40 oC for 3 min, then raised to 150
oC at a rate of 2 oC/min, then to 200 oC at a rate of 4 oC/min. The ion source temperature was 230
oC 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 x 0.25 mm, 0.25 µm) and enantiomeric excess measurements were determined with the
This article is protected by copyright. All rights reserved

following method: Starting temperature 100 °C and an increase of 5 °C per min until 135 °C, 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, MA, US) with an
SQ mass spectrometer (Waters, MA, US). The HPLC was performed with a Thermo Scientific
Hypersil GoldTM C4 reverse phase column (50 × 2.1 mm) with 0.2 mL/min 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 ionized using
electrospray ionization (ESI) and detected in positive mode. The operating condition of ESI
interface were set to a capillary temperature of 300 oC, 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 ppm to the nearest 0.01 ppm using the following abbreviations: s (singlet), d (doublet), t
(triplet), q (quartet), qn (quintet), sext (sextet), dd (doublet of doublets), dt (doublet of triplets),
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 ppm. Melting points were measured with a Gallenkamp apparatus
and were uncorrected.
This article is protected by copyright. All rights reserved

Structure alignment
The TK structure 2R8O.pdb and pyruvate decarboxylase structure 2VK1.pdb were aligned using
the server, TopMatch-Web [48, 49]. The aligned structure from output was visualized using
Pymol (Schrödinger, USA). Sequence identity was calculated online by the web server Sequence
Identity And Similarity (SIAS, 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 wild-type and mutant TKs with the grid centred at −11.25, 25.858, 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.
References
1. Reetz, M. T., Bocola, M., Carballeira, J. D., 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.
This article is protected by copyright. All rights reserved

2. Sprenger, G. A., Schorken, U., Sprenger, G. & Sahm, H. (1995) Transketolase A of Escherichia coli K12.
Purification and properties of the enzyme from recombinant strains, European journal of biochemistry /
FEBS. 230, 525-32.
3. Draths, K. M., Pompliano, D. L., Conley, D. L., Frost, J. W., Berry, A., Disbrow, G. L., Staversky, R. J. &
Lievense, J. C. (1992) Biocatalytic Synthesis of Aromatics from D-Glucose - the Role of Transketolase,
Journal of the American Chemical Society. 114, 3956-3962.
4. Demuynck, C., Bolte, J., Hecquet, L. & Dalmas, V. (1991) Enzyme-Catalyzed Synthesis of Carbohydrates
- Synthetic Potential of Transketolase, Tetrahedron Lett. 32, 5085-5088.
5. Toone, E. J. & Whitesides, G. M. (1991) Enzymes as Catalysts in Carbohydrate Synthesis, Acs Sym Ser.
466, 1-22.
6. Morris, K. G., Smith, M. E. B., Turner, N. J., Lilly, M. D., Mitra, R. K. & Woodley, J. M. (1996)
Transketolase from Escherichia coli: A practical procedure for using the biocatalyst for asymmetric
carbon-carbon bond synthesis, Tetrahedron-Asymmetr. 7, 2185-2188.
7. Wohlgemuth, R., Smith, M. E., Dalby, P. A. & Woodley, J. M. (2009) Transketolases, Encyclopedia of
Industrial Biotechnology.
8. Turner, N. J. (2000) Applications of transketolases in organic synthesis, Curr Opin Biotechnol. 11, 527-
531.
9. Kobori, Y., Myles, D. C. & Whitesides, G. M. (1992) Substrate-Specificity and Carbohydrate Synthesis
Using Transketolase, J Org Chem. 57, 5899-5907.
10. Schenk, G., Duggleby, R. G. & Nixon, P. F. (1998) Properties and functions of the thiamin diphosphate
dependent enzyme transketolase, Int J Biochem Cell Biol. 30, 1297-1318.
This article is protected by copyright. All rights reserved

11. Smith, M. E. B., Hibbert, E. G., Jones, A. B., Dalby, P. A. & Hailes, H. C. (2008) Enhancing and Reversing
the Stereoselectivity ofEscherichia coliTransketolaseviaSingle-Point Mutations, Advanced Synthesis &
Catalysis. 350, 2631-2638.
12. Hibbert, E. G., Senussi, T., Costelloe, S. J., Lei, W., Smith, M. E., Ward, J. M., Hailes, H. C. & Dalby, P. A.
(2007) Directed evolution of transketolase activity on non-phosphorylated substrates, J Biotechnol. 131,
425-432.
13. Hibbert, E. G., Senussi, T., Smith, M. E., Costelloe, S. J., Ward, J. M., Hailes, H. C. & Dalby, P. A. (2008)
Directed evolution of transketolase substrate specificity towards an aliphatic aldehyde, J Biotechnol. 134,
240-245.
14. Galman, J. L., Steadman, D., Bacon, S., Morris, P., Smith, M. E., Ward, J. M., Dalby, P. A. & Hailes, H. C.
(2010) alpha,alpha'-Dihydroxyketone formation using aromatic and heteroaromatic aldehydes with
evolved transketolase enzymes, Chem Commun (Camb). 46, 7608-7610.
15. Cazares, A., Galman, J. L., Crago, L. G., Smith, M. E., Strafford, J., Rios-Solis, L., Lye, G. J., Dalby, P. A. &
Hailes, H. C. (2010) Non-alpha-hydroxylated aldehydes with evolved transketolase enzymes, Org Biomol
Chem. 8, 1301-1309.
16. Payongsri, P., Steadman, D., Strafford, J., MacMurray, A., Hailes, H. C. & Dalby, P. A. (2012) Rational
substrate and enzyme engineering of transketolase for aromatics, Org Biomol Chem. 10, 9021-9029.
17. Payongsri, P., Steadman, D., Hailes, H. C. & Dalby, P. A. (2015) Second generation engineering of
transketolase for polar aromatic aldehyde substrates, Enzyme and Microbial Technology. 71, 45-52.
18. Zhou, C., Saravanan, T., Lorilliere, M., Wei, D., Charmantray, F., Hecquet, L., Fessner, W. D. & 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.
This article is protected by copyright. All rights reserved