Introduction

Transketolase is a key enzyme of the pentose phosphate pathway (PPP), is ubiquitous in all organisms, and provides a unique link between glycolysis and the non-oxidative phase of the PPP. Transketolase is a thiamine pyrophosphate (TPP)-dependent enzyme that reversibly transfers a two-carbon ketol group from a donor substrate (usually a five-carbon ketose) to an acceptor aldehyde substrate (usually a five-carbon aldose) via a ping-pong reaction mechanism, forming a new asymmetric C-C bond with high regio- and stereo-specificity. In biocatalysis, β-hydroxypyruvate (HPA) is often used as the donor substrate due to the irreversible, concomitant release of CO2 as a by-product. Strong substrate inhibition has been observed above 25 mM HPA with an inhibition constant of around 42 mM1,2. The cause of this substrate inhibition is currently unknown and is addressed in this study.

The inactive, apo-form of transketolase is in a monomer-dimer equilibrium that is dependent on protein concentration. Upon cofactor binding, both the inactive apo-monomer and apo-dimer are converted into the catalytically active, dimeric holo-form of, until recently, seemingly structurally-identical subunits, with two active-sites per homodimer, located at the subunit interface3,4. Each active site is comprised of one divalent cation, such as Mg2+, and one TPP molecule. In the S. cerevisiae transketolase apo-dimer, even after removal of free Ca2+, one Ca2+ ion was bound extremely tightly to one active site and could only be removed using harsh treatment, while the second Ca2+ ion dissociated easily. Subsequent addition of Mg2+ to the medium led to a transketolase dimer with one Ca2+- and one Mg2+-reconstituted active site, with positive cooperativity observed between active sites upon TPP-binding5. It is unclear if E. coli transketolase behaves in a similar way, since both positive and negative cooperativity have been observed at different [Mg2+]6.

Transketolase displays ping-pong kinetics and catalyses two sequential half-reactions; formation of the dihydroxyethyl-TPP (DHE-TPP) carbanion intermediate, followed by transfer of this two-carbon ketol group from the carbanion intermediate to an acceptor aldehyde, thus returning the enzyme to its starting state. It is thought that there is considerable communication between active sites in order to coordinate the ping-pong kinetics between active sites, such that catalysis alternates between active sites, giving rise to the ‘half-of-the-sites reactivity’ phenomenon that has been observed in transketolase7 and other TPP-dependent enzymes8,9,10,11.

There is evidence that in the E1 subunit of the pyruvate dehydrogenase complex (PDC) (EC 1.2.4.1), the two TPP-containing active-sites communicate via a 20 Å proton wire that shuttles a proton between active sites, enabling the cofactors to operate reciprocally as general acid-base catalytic moieties, thus synchronising the ping-pong mechanism across active sites12. It was suggested by the authors that many thiamine-dependent enzymes may function in a similar way. Recently, the expected structural asymmetry between the two monomers of apo-dimeric E. coli TK was finally observed at the proton scale and was implicated as a key feature of the proton wire and hence cooperativity between active sites13.

Transketolase is unusual in that incubation at between 40–55 °C for 1 h increases the residual activity significantly, when measured after re-cooling the samples to 25 °C. For example, incubation at 42 °C for 1 hr increases activity by 50%14,15. It was postulated that this curious phenomenon may be the result of an inactive form of transketolase being physically altered or activated by temperature, although no such species had been detected at that time.

Until recently, it was thought that purified transketolase existed in a single form with a high affinity for TPP, which is essential for catalytic activity. However, we previously found that transketolase exists with two distinct subpopulations of subunits in purified samples, TKhigh and TKlow, with over 200-fold different affinities for TPP at high [Mg2+]6. TKhigh was found to form via at least one specific oxidation of the unmodified TKlow. The TKlow subunit had only 4.5% of the activity of TKhigh in the presence of saturating concentrations of TPP, and was also found to have disrupted the cooperativity between TPP-binding sites observed for TKhigh. The two distinct monomeric subunits were found to combine into at least two distinct dimer forms, the TKhigh-TKhigh and TKlow-TKlow dimers, while the TKhigh-TKlow mixed dimer was not observed directly, but not ruled out as possible.

The fraction of TPP that can bind to TKhigh relative to TKlow, %Bmax(high)TPP (33.6 ± 2.9%), was found to be invariant to changes in [Mg2+] and [TK], while the addition of thiamine during fermentation to increase cellular [TPP] also had no effect. It was hypothesized that oxidative stress may play a role in determining the ratio of TKhigh:TKlow, and detection of post-translational oxidations by mass spectrometry in TKhigh supported our hypothesis6.

The pentose phosphate pathway runs parallel to glycolysis and has several cellular functions, including the generation of pentose sugars as well as ribose-5-phoshate, the latter the precursor for nucleotide biosynthesis. It is therefore a key branch-point in the diversion of metabolic flux towards biosynthetic pathways. However, one of the PPP’s most important utilities is to respond to and negate oxidative stress, often due to reactive oxygen species (ROS), through production of an anti-oxidant, NADPH16.

Intracellular NADPH is continuously used as a reducing agent to replenish the reduced glutathione pool to protect against oxidative stress and to maintain a stable cellular redox potential17,18,19. Recently, it has been demonstrated that in the short-term, oxidative stress can be mediated by redox-sensitive enzymes in lower glycolysis20,21,22,23. In these instances, oxidative post-translational modifications (PTMs) can provide regulation with a rapid response time by cysteine oxidation and subsequent diversion of glycolytic flux through the PPP to generate NADPH and nucleotide precursors for DNA repair. Another separate cysteine-based regulatory system that protects against oxidative stress by diverting flux from NAD+-dependent ALDS to NADH-producing ALDHs was also recently elucidated24. While a few control points have been identified in lower glycolysis20,21,22,23 and in the oxidative phase of the PPP16, to the authors’ knowledge redox regulation of enzymes in the non-oxidative phase of the PPP are yet to be identified or fully characterised. The identification of such regulatory mechanisms may also inform studies into the role of transketolase in tumour progression, given its increased activity found in many cancer cells25.

Here, we examine directly the binding of substrate to the transketolase species subpopulations for both wild type and the variant S385Y/D469T/R520Q, and provide further evidence that transketolase is a redox-regulated enzyme, that could therefore potentially play a major role in the control of flux through the PPP during oxidative stress. The new substrate inhibition and heat-activation data indicated that a mixed dimer species, TKhigh-TKlow, may be the cause of donor-substrate inhibition and heat-activation. Finally, we propose an updated Two-Species Model for transketolase activation, regulation and inhibition that includes explanations of the origin of %Bmax(high), HPA inhibition, active-site cooperativity, redox control and heat-shock activation.

Results

The Two-species Transketolase model is persistent across variants

Previously, we measured the TPP-binding parameters of wild-type transketolase over a range of cofactor concentrations6. Here, we report the TPP-binding parameters for S385Y/D469T/R520Q, a variant that was previously engineered for activity towards aromatic aldehydes26. The rationale behind these measurements was to (a) confirm the existence of both TKhigh and TKlow across variants; and (b) to demonstrate a direct correlation between changes in the % substrate inhibition, %Bmax(high), and %TKmodified. The TPP-binding parameters of TKhigh and TKlow for S385Y/D469T/R520Q at between 0–18 mM Mg2+ are summarised below (Fig. 1a–d; Tables 1 and 2; Fig. S1, Supplementary Information). The double-Hill function6 was again utilised to determine the independent TPP-binding parameters for TKhigh and TKlow. As previously for wild-type TK6, at 0 mM Mg2+, the binding parameters of TKhigh and TKlow for S385Y/D469T/R520Q were too similar to deconvolve with accuracy (Fig. S2, Supplementary Information), therefore the dissociation constants and Hill coefficients were constrained to equal each other within the double-Hill function.

The trends in the TPP-binding parameters of TKhigh and TKlow of S385Y/D469T/R520Q were similar to those of wild-type transketolase6. A significant decrease and comparatively small increase in the dissociation constants were observed for TKhigh (Fig. 1a) and TKlow (Fig. 1b), respectively, at higher [Mg2+]. However, TKhigh and TKlow of S385Y/D469T/R520Q each bound to TPP with a lower affinity than the wild-type (4.5-fold and 1.6-fold lower at 18 mM Mg2+, respectively), and TPP-binding to TKhigh (Fig. 1c) was less cooperative in S385Y/D469T/R520Q. Furthermore, the binding cooperativity changed from being positive, to non-cooperative as [Mg2+] was increased, but no longer peaked at the physiologically relevant [Mg2+] of 4 mM as observed in wild-type TK6. The trends were most likely non-linear in nature, but a linear relationship could not be ruled out (Fig. 1c,d). These results indicated that while changes in maximum affinity may change between variants, the response of TKhigh and TKlow to Mg2+ was persistent across variants, although the cooperativity was impacted quite considerably.

Adaption of the fluorescence quenching-based TPP-binding assay for donor substrate binding

Similar to our previous fluorescence quenching-based cofactor-binding assay6, binding of HPA to holo-transketolase was shown to further quench the intrinsic fluorescence of holo-transketolase (λex = 240 nm; λem = 330 nm) (Fig. S3, Supplementary Information). Like TPP, HPA absorbed relatively strongly at 240 nm. The sample signal was therefore also corrected for the inner filter effect (IFE) by generating a correction factor, determined empirically from the change in fluorescence intensity of 0–80 mM free HPA in 50 mM Tris-HCl buffer, 9 mM Mg2+ and 0.3 mM TPP (Fig. S4, Supplementary Information).

Characterisation of E. coli transketolase binding to HPA

The dissociation constant of HPA-binding to transketolase, KdHPA, has never been previously determined directly, although the closely-related Michaelis-Menten constant of HPA-binding to wild-type transketolase, KmHPA, has been reported as 5.5 ± 0.5 mM27 and 5.3 mM28. Substrate inhibition by HPA, KiHPA, has also been measured as 42.2 mM and 43 mM1,2, providing suitable benchmarks for comparison. The double-Hill function6 was again used to determine the HPA-binding parameters for both wild-type transketolase and S385Y/D469T/R520Q, as the fit to the data was superior to a single Hill function (Fig. 2a–b; Table 3), as reported previously6.

Two distinct HPA-binding events, TKhigh and TKlow, were detected, with an 11-fold difference in affinity, supporting our previously postulated Two-Species Model of transketolase. The three mutations within the S385Y/D469T/R520Q variant appeared to have no significant effect on the affinity of HPA binding, but significantly decreased the cooperativity of both TPP- and HPA-binding in TKhigh, and increased the cooperativity of HPA-binding in TKlow (Table 3). In addition, binding of HPA to TKlow resulted in a decrease in fluorescence quenching in S385Y/D469T/R520Q. This was presumably related to the introduction of the fluorescent tyrosine at residue 385. Importantly, the proportion of HPA bound to TKhigh relative to all TK at saturation, %Bmax(high)HPA, matched that for TPP bound to TKhigh, %Bmax(high)TPP, for each variant (Tables 4 and 5), indicating that the structural difference between TKhigh and TKlow impacted both TPP and HPA binding, and confirmed that the two species remained distinct from each other. Furthermore, the %Bmax(high)HPA and %Bmax(high)TPP also matched the %TKmodified determined from the mass spectra of wild-type transketolase6 and the variant S385Y/D469T/R520Q (Fig. 3a; Table 4), demonstrating that the enzyme affinity parameters and the oxidation to form TKhigh were correlated.

The KdHPA of wild-type TKhigh, Kd(high)HPA, was slightly lower than the previously reported KmHPA 27,28. This was expected because Km is measured via enzyme kinetics, and is therefore a chemical pseudo-equilibrium which is convoluted with the additional forward reaction for formation of product, in other words substrate turnover (kcat). By comparison, KdHPA was obtained as a direct equilibrium measurement for binding of HPA. The KdHPA of wild-type TKlow, Kd(low)HPA, was very similar to the previously reported inhibition constant of HPA, KiHPA1,2, which suggested a possible relationship between TKlow binding to HPA and substrate inhibition by HPA, despite the fact that the KiHPA was obtained with no knowledge of the two-species model of transketolase.

HPA appeared to bind to the wild-type TKhigh active-sites with slight positive cooperativity (n = 1.38 ± 0.19), while HPA binding to TKlow was highly cooperative with a Hill coefficient of 3.10 ± 0.16. This high Hill coefficient suggests a potentially important role of TKlow inhibition by HPA in the regulation of TK activity.

Heat-induced activation and conversion of TKlow to a TKhigh-like state

The previous study into heat-activation of TK used cofactor concentrations of 0.5 mM Mg2+ and 50 μM TPP, rather than 9 mM Mg2+ 14. At these concentrations, the [TPP] was semi-saturating for TKhigh (approximately 60% saturated) but too low for TKlow saturation (approximately 20% saturated)6. In other words, the majority of TKhigh was in the catalytically active holo-form while TKlow was mostly in the catalytically inactive apo-form. Therefore, we hypothesised that heat exposure may convert TKlow to TKhigh, hence increasing the concentration of holo-transketolase and overall activity of the sample.

We tested our hypothesis by taking fluorescence quenching measurements of 0.05 mg/mL TK, 9 mM Mg2+ and a range of [TPP] before6 and after incubation at 42 °C for 1 hour (Fig. 3b; Table 4). As hypothesised, the %Bmax(high) increased from 33.7% to 51.7% after heat-activation. In addition, the affinity of both TKhigh and TKlow increased significantly. Performing the same heat activation for a second time on the same sample increased the %Bmax(high) only slightly to 53.5% (Fig. S5, Supplementary Information), indicating that no further change could be induced. Conversely, heat-activation had negligible impact on %TKmodified (29.4 ± 4.1%), determined from the mass spectra of heat-activated wild-type transketolase (Fig. 3c). Taken together, these results suggested heat-activation may occur through formation of a TKhigh-like conformational state but via a different mechanism to that of oxidation of TKlow to TKhigh.

The activity of 0.05 mg/mL TK, 9 mM Mg2+ and 50 μM TPP towards 50 mM GA and 50 mM HPA pre- and post-incubation was determined to calculate the activity of heat-activated TKhigh relative to pre-incubated TKhigh (Fig. 3d; Table 5). Overall, a 24.6% increase in transketolase activity was observed after heating, consistent with our hypothesis, though approximately only 50% of the activity enhancement expected from the %Bmax(high) increase. This may indicate lower activity in the heat-induced TKhigh-like state compared to the oxidised TKhigh, or some partial unfolding and inactivation during heating. These and other possible mechanisms were not investigated further here.

The existence of a TKhigh-TKlow mixed dimer species that mediates HPA substrate inhibition

The equivalence between the Kd(low)HPA and the KiHPA measured in previous activity assays suggested that binding of HPA to TKlow gave rise to the observed overall inhibition of transketolase activity. However, the activity of TKlow was already only 4.5% relative to that of TKhigh6, implying that the TKlow-TKlow dimer was effectively inactive already, and so binding of HPA to that dimer species could not have contributed significantly to the observed reaction inhibition. Therefore, the interaction between HPA and TKlow must inhibit the TKhigh activity, which in turn suggested that inhibition occurred within a TKhigh-TKlow mixed dimer form. We therefore attempted to estimate the relative proportions of the three dimeric species, TKhigh-TKhigh, TKhigh-TKlow, and TKlow-TKlow, from the heat-activation, HPA-binding, and enzyme activity data available.

Estimation of the % dimer forms from heat-activation data

We first assumed that conversion of TKlow to a TKhigh-like state via heat-activation was only possible for the proportion of TKlow present within a TKhigh-TKlow mixed dimer. We assumed all TKlow subunits of the mixed dimer were completely converted to a TKhigh-like state, and calculated the relative proportions of each dimeric species before heat-activation from the change in %Bmax(high) upon heat-activation:

$$({\rm{A}})\, \% {{\rm{TK}}}_{{\rm{high}}}-\,{{\rm{TK}}}_{{\rm{low}}}=( \% {{B}_{max(high)}}^{heated}- \% {{B}_{max(high)}}^{unheated})\ast 2=(51.7 \% -33.6 \% )\ast 2=36.2 \% ;$$
$$({\rm{B}})\, \% {{\rm{TK}}}_{{\rm{high}}}-{{\rm{TK}}}_{{\rm{high}}}= \% {{B}_{max(high)}}^{unheated}-\,({\rm{A}})/2=33.6 \% -(36.2/2)=15.5 \% ;$$
$$({\rm{C}})\, \% {{\rm{TK}}}_{{\rm{low}}}-{{\rm{TK}}}_{{\rm{low}}}=100 \% -\,({\rm{A}})\mbox{--}({\rm{B}})=100 \% -36.2 \% -15.5 \% =48.3 \% .$$

Estimation of the % dimer forms from HPA-inhibition data

Previous studies into the inhibition of transketolase determined the inhibition constants and the maximum inhibition of wild-type transketolase activity by HPA, %Imax, for wild-type (KiHPA = 43 mM; %Imax = 48.1 ± 5.1%)2 and also the D469T variant of transketolase (KiHPA = 40 mM; %Imax = 46.8 ± 11.5%)29.

Making the assumption that HPA-binding to the TKlow subunit of the mixed dimer resulted in total inhibition of the TKhigh subunit in that mixed dimer, and that TKhigh accounted for 91% of total activity, we predicted the relative proportions of the three dimeric species:

$$({\rm{A}})\, \% {{\rm{TK}}}_{{\rm{high}}}-{{\rm{TK}}}_{{\rm{low}}}=( \% {{B}_{max(high)}}^{unheated}\ast ( \% {I}_{max}/91 \% )\ast 2=\,(33.6 \% \ast (48.1 \% /91 \% )\ast 2=35.5 \%$$
$$({\rm{B}})\, \% {{\rm{TK}}}_{{\rm{high}}}-{{\rm{TK}}}_{{\rm{high}}}= \% {{B}_{max(high)}}^{unheated}-\,({\rm{A}})/2=15.8 \% ;$$
$$({\rm{C}})\, \% {{\rm{TK}}}_{{\rm{low}}}-{{\rm{TK}}}_{{\rm{low}}}=100 \% -({\rm{A}})\mbox{--}({\rm{B}})=100 \% -35.5 \% -15.8 \% =48.7 \% .$$

The similarity between the relative proportions of the three dimer species, calculated from the %Imax obtained through activity data, and from TPP-binding data after heat-activation (Table 6), provides compelling evidence for our reasoning above that HPA binding to the TKlow subunit within the mixed dimer resulted in inhibition of the associated TKhigh subunit in that mixed dimer. It also implies that heat-activation may happen via the removal of the donor substrate inhibition when the TKlow subunit in the mixed dimer is converted into a TKhigh-like state. The fact that inhibition via the mixed dimer species persisted into a single-mutant TK variant, and with a different acceptor substrate, also suggested that this phenomenon is fundamental to the TK structure and mechanism when utilising HPA as the donor substrate.

The unified two-species model of transketolase activation, regulation and inhibition

The development of assays capable of detecting both TKhigh and TKlow has facilitated investigations into donor substrate inhibition, active site cooperativity and heat-activation, and finally led to the discovery of a novel mechanism of transketolase regulation. This can be summarised in the unified Two-Species Model of transketolase activation, regulation and inhibition (Fig. 4).

In summary:

• Transketolase exists as two distinct transketolase species; TKhigh and TKlow.

• Inactive TKlow is the reduced, unmodified form of transketolase that is converted to TKhigh via oxidation of one or more methionine or cysteine residues, in response to cellular oxidative stress.

• TKhigh is significantly more active and also has a 35-fold increased affinity for TPP at physiologically relevant [Mg2+].

• Oxidation improves the cooperativity between active sites. Cys157 oxidation is a strong candidate6, possibly by providing the final residue of a proton wire between active sites.

• Redox regulation of transketolase potentially provides an important mechanism of control in diverting flux from glycolysis to the PPP during oxidative stress.

• Heat-activation of transketolase converts the TKlow subunit of the mixed dimer to a TKhigh-like state, which relieves the substrate inhibition of the associated TKhigh subunit.

• Heat-activation may offer cells a degree of heat shock protection by activating transketolase - a key enzyme in central metabolism - without the energetic or time-cost of protein biosynthesis.

Discussion

In this study, we attempted to resolve several unanswered questions remaining from our previous study into transketolase activation6; does the two-species phenomenon persist across variants and from cofactor to donor substrate; what is the cause of substrate inhibition; and what is the origin and physiological relevance of heat-activation?

Though wild-type, D469T and S385Y/D469T/R520Q all exist as a mix of TKhigh and TKlow, the susceptibility of TKlow to oxidation could potentially be enhanced through protein mutations in order to maximise [TKhigh] and hence activity. Alternatively, controlled oxidation of purified transketolase samples may be possible, but this may require careful control to avoid over-oxidation of Cys157, or at other sites, that eventually would cause inactivation.

It is worth noting that this study and our previous study6 only investigated TPP binding in the presence/absence of Mg2+ and not Ca2+. It is possible that Ca2+-reconstituted transketolase is activated differently via different oxidation-activation pathways. However, all mass spectra in these studies were derived from transketolases in their apo-form and the formation of modified, oxidised species were hence cation-independent. It is both possible that Ca2+-reconstituted transketolase is activated in a similar way to Mg2+-constituted transketolase, or active site coordination with Ca2+ bypasses the requirement of oxidation for activation. Further studies into TPP- and HPA binding to TKhigh and TKlow in the presence of Ca2+ would be able to determine which hypothesis is correct, but are beyond the scope of this study.

It is interesting that HPA binding to TKlow had such a high degree of positive cooperativity, and may indeed result from multiple HPA molecules binding to a single active-site to inhibit activity in that active site. Furthermore, it may also inhibit active-site synchronisation by inhibiting the shuttling of protons along the proton wire. The existence of the TKhigh-TKlow mixed dimer potentially complicates further the analysis of Hill coefficients for TKhigh and TKlow. However, the true Hill-coefficient of TKhigh is likely to be higher than the reported apparent value, whereas the true Hill-coefficient of TKlow is presumably lower. The most abundant dimer, TKlow-TKlow, may also have previously masked the recently-elucidated asymmetric structure of the TK apo-dimer13, given the reduced cooperativity of TPP-binding between active sites in this dimer.

The heat-activation of TK appeared to increase activity through a slightly different mechanism to that of oxidation of TKlow to TKhigh. Conformational rearrangements might result from heat-activation and give rise to a TKhigh-like conformational state that apparently relieves HPA inhibition of the TKhigh subunit of the mixed dimer. The heat-activation of transketolase may even be part of the cellular response to thermal stress, resulting in the upregulation of the PPP and hence NADPH production, while also increasing flux through critical biosynthetic pathways such as nucleotide biosynthesis. The heat-sensitivity of transketolase therefore has a potential role in facilitating rapid global changes in metabolism without the additional energetic burden of protein synthesis.

Finally, the discovery of a redox-sensitive regulatory system that activates transketolase, potentially during oxidative stress, could have important implications for future advances in cancer treatments. Cancer cells are often exposed to higher levels of oxidative stress than normal cells, and elevated transketolase activity has been reported in a number of cancer types in order to reduce ultimately catastrophic damage from high oxidative stress30. The redox-sensitivity of transketolase not only makes transketolase itself an antioxidant, but also an important mediator of the biosynthesis of a second anti-oxidant, NADPH. Reversal or prevention of transketolase oxidation at residue Cys157 by drug delivery or gene therapy may offer an effective way to prevent antioxidant production and hence proliferation in cancer cells.

Methods

Materials

TPP, MgCl2, glycolaldehyde (GA) and Erythrulose [Ery] were purchased from Sigma-Aldrich; Tris-HCl was purchased from VWR International and Guanidine-HCL was purchased from Life Technologies Ltd. HPA was synthesised from bromopyruvic acid and LiOH, as described previously31.

Enzyme preparation

Wild-type transketolase with an N-terminal His6 tag was expressed in E. coli XL10-gold cells (Agilent Technologies Ltd) from the plasmid pQR791. The resulting cell pellet was lysed and purified as described previously32. Purified transketolase was ultrafiltrated four times using an Amicon Ultra-4 10k MWCO (Millipore, US) centrifugal filter to remove excess imidazole and cofactors and subsequently dialysed overnight at 4 °C in 50 mM Tris-HCl, pH 7.0 to obtain apo-TK. Protein concentration was determined by absorbance at 280 nm in 6 M Guanidine-HCl and 20 mM Sodium Phosphate, pH 6.5. Absorbance was measured using a Nanodrop spectrophotometer (Thermo Fisher Scientific, Wilmington, DE), assuming a monomeric molecular weight of 73035.5 g mol−1 and an extinction coefficient (ε) of 92630 L mol−1 cm−1.

For TPP-binding assays, series of 2x concentrated cofactor solutions were prepared and purified TK was added to a final concentration of 0.05 mg/mL. The samples were incubated at 22 °C for 45 minutes to allow TK-TPP binding to reach equilibrium. For HPA-binding assays, 2x concentrated, purified holoTK was prepared at 0.1 mg/mL TK, 18 mM Mg2+ and 0.6 mM TPP and incubated at 22 °C for 45 minutes to allow TK-TPP binding to reach equilibrium. The purified holoTK was added to a series of 2x concentrated HPA solutions and incubated at 22 °C for 10 minutes to allow holoTK-HPA binding to reach equilibrium. For heat-activation studies, TK samples were incubated at 42 °C for 1 hour and subsequently re-equilibrated at 4 °C for 30 minutes and at 22 °C for 30 minutes prior to assays.

Fluorescence assay to detect TPP binding

TPP-binding was measured using a Fluoromax-4 (Horiba, UK) spectrofluorometer (λex = 240 nm; λem = 330 nm; integration time = 0.1 s; slit width = 8 nm), as described previously6.

Transketolase activity assay

Purified, dialysed apo-transketolase (0.2 mg/mL) was incubated with 2.4 mM TPP and 9 mM Mg2+ for 45 minutes at 22 °C. 50 μL was added to 100 μL 150 mM GA, 150 mM HPA, giving final substrate concentrations of 50 mM. The reaction was performed in triplicate at 22 °C in a 96 well plate with shaking at 300 rpm using a Thermomixer Comfort shaker. 10 μL of the reaction was quenched with 190 μL 0.1% trifluoroacetic acid (TFA) after 3, 5, 10, 15, 20, 30, and 40 minutes. Samples were subsequently analysed by a Dionex HPLC system (Camberley, UK) with a Bio-Rad Aminex HPX-87H reverse phase column (300 × 7.8 mm2) (Bio-Rad Labs., Richmond, CA, USA), via Chromeleon client 6.60 software, to separate and analyse the change in the concentration of substrate (GA) and product (Ery) over the course of the reaction using the method described previously29.

Fluorescence assay to detect HPA binding

The same methodology was used as the TPP-binding fluorescence assay, except cofactor concentrations were kept constant at 9 mM Mg2+ and 0.3 mm TPP, and an IFE CF was generated between 0–80 mM HPA.

Mass spectrometry (LC-ESI-MS)

LC-MS was performed using an Agilent 1100/1200 LC system connected to a 6510 A QTOF mass spectrometer (Agilent, UK), as described previously6. Triplicate fermentations were performed for each sample.