Separating Thermodynamics from Kinetics—A New Understanding of the Transketolase Reaction

Abstract Transketolase catalyzes asymmetric C−C bond formation of two highly polar compounds. Over the last 30 years, the reaction has unanimously been described in literature as irreversible because of the concomitant release of CO2 if using lithium hydroxypyruvate (LiHPA) as a substrate. Following the reaction over a longer period of time however, we have now found it to be initially kinetically controlled. Contrary to previous suggestions, for the non‐natural conversion of synthetically more interesting apolar substrates, the complete change of active‐site polarity is therefore not necessary. From docking studies it was revealed that water and hydrogen‐bond networks are essential for substrate binding, thus allowing aliphatic aldehydes to be converted in the charged active site of transketolase.


Molecular Biology and Enzyme Expression
Bacterial work was performed under sterile conditions using autoclaved consumables and solutions under a bunsen burner flame. Cells were broken using a Sonifier 250 (Branson) for volumes up to 5 mL of cell suspension, or a Multishot Cell Disrupter (Constant Systems Ltd) for larger volumes. Enzymes were purified on a NGC Quest 10 system (Biorad) using XK16/20 columns (GE Healthcare Life Sciences) packed with Ni Sepharose 6 FF resin (GE Healthcare Life Sciences). Analytical quantities of enzyme were purified using Ni-NTA spin columns (Qiagen). Protein overexpression and purity were analysed by SDS-PAGE on a PhastSystem Separation and Control Unit (Pharmacia) using Phastgel Gradient 10-15% precast gels (GE Healthcare Bio-Sciences). Protein concentrations were determined using a Bicinchoninic Acid (BCA) assay kit (Sigma-Aldrich) following the absorbance at 562 nm on a Ultrospec 2100 pro UV/Vis spectrometer (Pharmacia). Wild type Transketolase from S. cerevisiae (EC 2.2.1.1, gene accession number P23254) was previously cloned into the pBAD/His A plasmid adding an N-terminal His-tag to the protein. CaCl 2 competent cells (E. coli top10, Invitrogen) were used as hosts for cloning and corresponding mutants as previously reported 1 . Plasmids were isolated using a QIAprep Spin Miniprep Kit (QIAGEN) and DNA concentrations were determined using a ND-1000 Spectrophotometer (Thermo Fisher Scientific). A PCR Master Mix (2x) was bought from Thermo Fisher Scientific (K0171) and the reaction was carried out in a Tgradient PCR machine (Biometra Westburg). SYBR Safe stain (Thermo Fisher Scientific) was used for staining of DNA gels and SimplyBlue Safe stain (Invitrogen) was used for staining protein gels.

Stock Solutions
Growth medium: 10 g tryptone, 10 g NaCl and 5 g yeast extract were dissolved in 1 L of milli-Q and autoclaved the same day 1000x Ampicillin: the ampicillin sodium salt was dissolved at 100 mg/mL in sterile milli-Q and was kept at -20°C for max. 1 week 20% (w/v)arabinose stock: 400 mg were dissolved in 2 mL of sterile milli-Q 20 mM sodium phosphate buffer, pH=7.0: 1.32 g NaH 2  O + 115 mg thiamin diphosphate chloride were dissolved in 5 mM sodium phosphate buffer and the pH was adjusted to 7.0 using 0.1M NaOH to afford a final volume of 10 mL Permanganate stain: 1.5 g KMnO 4 + 10 g K 2 CO 3 and 1.25 mL 10% NaOH were dissolved in water and filled up to 200 mL

Protocols
Ampicillin plates (20x) were made by carefully boiling 400 mL Luria Bertani (LB) agar in the microwave until complete dissolution. It was allowed to cool down to approximately 40°C before ampicillin was added (400 µL, 1000x). The plates were allowed to solidify under the flame before the lid was closed. Plates were stored in the fridge for max. 2 weeks.
Cryostocks were prepared from freshly grown 5 mL overnight cultures by pelleting 1.5 mL thereof, decanting the supernatant, resuspending the cell pellet in 0.5 mL of fresh LB and subsequent addition of 0.5 mL 80% glycerol. The suspensions were mixed by vortex, transferred into cryotubes and were stored at -80°C.
Transformation of plasmid DNA into CaCl 2 competent cells (E. coli top 10, Invitrogen) was performed by thawing the cells on ice for 30 minutes. 100-200 ng of plasmid DNA was then added and incubated for 30 minutes on ice. The cells were subsequently exposed to a heat shock at 42°C for 90 seconds. They were placed directly on ice again for 2 minutes. LB (800 µL) was added and incubated at 37°C for 1h. The cells were pelleted (13500 rpm, 5 minutes), the supernatant removed and the cells resuspended in 100 µL LB. The suspension was plated on ampicillin plates and incubated overnight at 37°C. The mixture was distributed over ten PCR tubes and amplified using a gradient over the temperature range of 56 to 69°C. 30 µL of the product mixtures were each digested with DpnI (0.5 µL, 1h, 37°C). Of this, 15 µL were transformed and successful point mutation was confirmed by sequencing.
Enzyme overexpression: Pre-cultures were grown overnight from single colonies (100 mL LB, 100 µg/mL ampicillin, 37°C, 180 rpm). The inoculum was then added to 1 L of sterilised growth medium in a 5 L Erlenmeyer flask and grown until an OD 600 of 0.6-0.8 was obtained. The cells were then induced by addition of arabinose at a final concentration of 0.02% (w/v). Enzyme overexpression was carried out overnight (37°C, 180 rpm). The cells were harvested (10'000 rpm, 10 minutes). a) For preparation of cell free extract: The pellet was resuspended (20 mM sodium phosphate buffer, pH=7.0). The suspension was transferred into 50 mL tubes and was centrifuged again (5'000 rpm, 10 minutes). The supernatant was discarded and the pellet was stored at -80°C. b) For enzyme purification by affinity chromatography: The pellet was resuspended (Ni-NTA binding buffer) and lysosyme at 20 mg/g cell pellet was added and a spatula tip of DNAse.
Preparation of cell free extract: The cell pellets containing the overexpressed enzymes were thawed on ice (30 minutes). The cells were resuspended in sodium phosphate buffer (5 mM, pH=7.0, 10 mL/g cell pellet). A protease inhibitor (PMSF, 200 µL, 0.1M in EtOH) was added to each sample. Lysosyme was added at a concentration of 20 mg/g cell pellet and a spatula tip of DNAse was added to each sample. The obtained suspension was incubated on ice for another 30 minutes. The cells were then sonicated on ice (30% duty cycle, output = 3, 4-5 minutes). The samples were centrifuged (5000 rpm, 20 minutes, 4°C) and the supernatant stored in aliquots of 4 mL at -20°C. SDS-PAGE: 5 µL of protein solution was boiled with 15 µL of bromophenol blue loading dye at 90°C for one minute. Of this solution, 2 µL were then loaded on a 10-15% gradient gel and separated. The gel was subsequently stained by incubation overnight with an aqueous solution of simply blue safe stain (100 rpm, room temperature).
Native-PAGE: 2 µL protein solution were directly loaded on a 10-15% gel and separated. The gel was subsequently stained by incubation overnight with an aqueous solution of simply blue safe stain (100 rpm, room temperature).
Protein concentrations were determined by performing the bicinchoninic S5 acid assay (BCA) according to the protocol provided: a working solution was prepared by mixing the bicinchoninic acid stock solution with a 4% (w/v) solution of CuSO 4 H 2 O in a ratio of 50:1. For the analysis, 2 mL of working solution were mixed with 100 µL of protein solution, incubated at 37°C for 30 minutes and the absorbance measured in technical duplicate at 562nm for each sample. A calibration curve was made each time using BSA as standard at concentrations of 0-1000 µg/L in intervals of 200 µg/L in biological and technical duplicate (R 2 = 0.997). Protein samples were measured in biological and technical duplicate.
Preparative scale enzyme purification: The mixture of cells, lysosyme, DNAse and 1 mM PMSF in binding buffer (5 mM imidazole for R528K, R528Q, R528K/S527T, R528Q/S527T and 20 mM imidazole for WT, D477E and D477T) was incubated on ice for one hour. The cells were subsequently broken using a cell disrupter at 1.8 kbar. The cell debris was removed by centrifugation (10'000 rpm, 10 minutes, 4°C), the supernatant filtered (Whatman filter, 0.45 µm pore size) and charged on 10 mL of Ni-sepharose 6 Fast Flow resin. Fractions were stored on ice and the peaks analysed by SDS-PAGE.
DNA gel electrophoresis to analyse PCR products: 500 mg agarose was dissolved in 50 mL 1x TAE buffer (1% w/v) and was added 5 µL 10000x SYBR SAFE stain. The gel was allowed to solidify at room temperature and a mixture of 10 µL PCR product and 2 µL of 6x Loading Dye was loaded on the gel. A DNA ladder was added as reference and the gel was analysed (50 minutes, 100 V). Chemicals and solvents were obtained as reagent grade from Sigma-Aldrich. Aldehydes were freshly distilled and their purity confirmed by 1 H NMR before usage. Petrolether (boiling point 40-60°C) was freshly distilled before usage. Lithium hydroxypyruvate was both obtained commercially and synthesised as previously described 2 . Reaction progress was monitored by TLC (TLC Silica gel 60 F 254 , Merck) using UV light and a potassium permanganate stain for visualisation. NMR spectra were recorded using an Agilent 400 MHz ( 1 H, 9.4 Tesla) spectrometer at 298K and were subsequently interpreted using MestReC. A benzene-D 6 NMR insert capillary (Sigma-Aldrich) was used for external locking during water suppression experiments using the PRESAT pulse sequence. Preparative scale bioconversions were carried out in an Excella E24 Incubator Shaker (New Brunswick Scientific). Optical rotations were measured on a Model 343 Polarimeter (PerkinElmer).

Chemistry and Biotransformations
Analytical quantitation of products containing the dihydroxyketone motif was accomplished by reversed-phase chromatography using calibration curves on a Shimadzu LC-20AD prominence system equipped with an ICSep Coregel 87H3 column (0.4x25 cm, Transgenomic). The absorbance was followed at 210 nm using 0.1% (v/v) aqueous trifluoroacetic acid pH = 2.5 as mobile phase (60°C, 0.8 mL/min).
Chiral separation of the dibenzoylated enantiomers for the analysis of enantiomeric excess was determined on a Shimadzu LC-20AD prominence system equipped with a Chiralpak AD-H column (0.46x25 cm, Daicel) using n-heptane/i -PrOH 97:3 as mobile phase (35°C, 1 mL/min).
The Optical rotation of commercial L(+) Erythrulose was measured at 589 nm (sodium D line) in aqueous solution (200 mg/10 mL) at 20°C, the value processed according to equation (1) and the enantiomeric excess was determined according to equation (2). (2)

Lithium Hydroxypyruvate Synthesis
Bromopyruvic acid (30g, 180 mmol, 1.0 eq.) was dissolved in demi water (150 mL). The pH was adjusted to 9 using 5 M LiOH. The reaction was then continued in a pH stat using 1 M LiOH overnight to maintain a constant pH of 9 over the course of the reaction. The reaction mixture was then acidified to pH = 5 using glacial acetic acid. The solvent was reduced in vacuo until precipitation was initialised. The product was precipitated overnight in the fridge. It was filtered, washed with ice cold water (2x) and dried in a desiccator over silica. The product was obtained as a fine white powder (8.8g, 45%) 2 and was confirmed by 1 H NMR in agreement with the commercial product. Both the synthesised and commercially obtained lithium hydroxypyruvate still contained traces of the starting material.

Synthesis of Racemic Standards
Racemic standards were synthesised according to a method previously described 3 . N -methylmorpholine (330 µL, 3.0 mmol, 1.0 eq.) was dissolved in water (40 mL) and the pH was adjusted to 8.0 using 10% HCl. LiHPA (330 mg, 3.0 mmol, 1.0 eq.) and the corresponding aldehyde (3.0 mmol, 1.0 eq.) were added and the reaction was stirred overnight at room temperature under N 2 atmosphere. The conversion of the aldehyde was monitored by TLC (npentane / EtOAc 1:1). Silica powder was added, the water removed in vacuo and the crude product purified by flash chromatography (n-pentane / EtOAc 1:1).

Dibenzoylation Procedure for Racemic Standards
Racemic standards were dibenzoylated for chiral analysis using a general method: The dihydroxyketone substrate (1.0 eq.) was dissolved in dry dichloromethane (10 mL) under N 2 atmosphere in a flame dried round bottomed flask. Dry triethylamine (10.0 eq.) and benzoyl chloride (5.0 eq. per hydroxyl) were added and the reaction mixture was stirred for 2 hours at room temperature. The reaction was quenched by addition of saturated sodium bicarbonate solution (30 mL) and stirred for another 30 minutes. The phases were separated and the organic phase was washed (sat. NaHCO 3 , 2x, 50 mL, sat. NH 4 Cl, 1x, 50 mL, brine, 1x, 30 mL). The organic phase was dried over sodium sulphate, the solvent was removed in vacuo and the product purified by flash chromatography (petrolether / EtOAc 10:1). Purification by flash was omitted in the determination of the enantiomeric excess.   4c 1-cycloheyl-2-oxopropane-1,3-diyl dibenzoate: the product was not synthesised due to unsuccessful conversion of the substrate by the enzyme.

Glycolaldehyde Activity Assay
Cell free extract (50 µL) was incubated with the cofactors (ThDP: 1 mM, Mg 2+ : 4 mM, 20 min., 800 rpm, 25°C). Subsequently lithium hydroxypyruvate and glycolaldehyde were added at a final concentration of 50 mM in 300 µL total reaction volume. The reaction mixture was shaken (15 min., 800 rpm, 25°C), quenched by 1:1 addition of trifluoroacetic acid (300 µL, 0.2% v/v), the enzyme removed by centrifugation and the supernatant was analysed by RP HPLC. Erythrulose product concentrations were obtained using a calibration curve (R 2 = 0.998) and subsequently were converted into moles. The activity was determined according to the definition of 1U = 1µmol/minute. The activity was multiplied by 20 to correct for 50 µL cell free extract in order to obtain the volumetric activity in U mL 1 .

Preparative Scale Bioconversions
Cell free extract (20 U based on the glycolaldehyde activity assay) was incubated with the transketolase cofactors (20 min, room temperature, final concentrations: Mg 2+ 18 mM, ThDP 5 mM). Lithium hydroxypyruvate (110 mg, 1.0 mmol, 1.0 eq.) and the corresponding aldehyde (1.0 mmol, 1.0 eq.) were added and the reaction volume was adjusted to 10 mL by addition of buffer (5 mM sodium phosphate, pH = 7.0). The flask was sealed air tight and the reaction was carrried out overnight in a shaker (25°C, 200 rpm, ≈ 18h). The reaction mixture was transferred into a 50 mL falcon tube, extracted with MTBE (2x, 40 mL) and the solvent was removed in vacuo. The product was obtained in moderate to good purity for low activity mutants and excellent purity with D477E (determined by 1 H NMR).

Determination of Michaelis-Menten Parameters
Aqueous solutions of racemic dihydroxyketone products were prepared in duplicate and subsequently used in dilution series to obtain HPLC calibration curves. In order to measure the Michaelis-Menten parameters under credible initial rate conditions (<20% conversion), individual reaction times were determined for all enzymes and substrates by following the conversion over time. The buffered reaction mixture (600 µL, 5 mM sodium phosphate, pH = 7.0) containing holotransketolase (100 µg/674 pmol), lithium hydroxypyruvate (65 mM) and the corresponding aldehyde (30 mM) was incubated in a thermoshaker (500 rpm, 25°C). Samples were taken over time (50µL, quenched by 1:1 addition of 0.2% v/v TFA), the enzyme precipitated by centrifugation and the conversion was analysed by RP HPLC.

S14
In order to determine the Michaelis-Menten parameters, buffered reaction mixtures (300 µL, 5 mM sodium phosphate, pH = 7.0) containing the holoenzyme (50 µg, 1 mM ThDP, 4 mM Mg 2+ ), lithium hydroxypyruvate (100 mM) and the corresponding aldehyde at different concentrations (5, 10, 20, 40, 70, 100, 150 mM) were incubated (25°C, 500 rpm) in duplicate over the individual times previously determined. The reactions were quenched by 1:1 addition of 0.2% (v/v) TFA, the enzyme was precipitated by centrifugation and product concentrations were determined RP HPLC analysis. Product concentrations were correlated to molar amounts and converted into rates by taking the reaction time into account. The average from duplicate data was then plotted against substrate concentration. A Michaelis-Menten type non-linear fit was obtained by inserting literature based initial parameters for K M and v max into equation (3). The Excel built-in solver then automatically varied K M and v max in order to successively minimise the sum of the squared errors between the measured and fitted data points until the solver converged to the best solution for K M and v max . The parameter k cat was obtained according to equation (4) taking the total enzyme concentration (50 µg / 337 pmol) into account. A benzene-D 6 capillary was obtained commercially (Sigma-Aldrich) and was used as internal standard in NMR water suppression experiments using the PRESAT pulse sequence in sealed Wilmad screw-cap NMR tubes (Sigma Aldrich Enzyme (WT transketolase, 200 µg) was incubated with its cofactors (25°C , 500 rpm, 20 min, final concentrations: ThDP 5 mM, Mg 2+ 18 mM). a) Solutions of glycolaldehyde and lithium hydroxypyruvate were added to achieve a final concentration of 100 mM each and the reaction mixture was filled up to 500 µL with buffer (5 mM sodium phosphate, pH = 7.0). b) A solution of glycolaldehyde was added to achieve a final concentration of 200 mM and the reaction mixture was filled up to 500 µL with buffer (5 mM sodium phosphate, pH = 7.0). c) Three control reactions were prepared in buffered solutions containing the cofactors Mg 2+ and ThDP containing 1) glycolaldehyde (200 mM), 2) glycolaldehyde (100 mM) with LiHPA (100 mM) and 3) L(+)erythrulose (100 mM).

Thermodynamic Model
A thermodynamic model was created using MATLAB (MathWorks, Version R2014b) to describe the equilibrium concentrations of glycolaldehyde in aqueous solution as shown in figure (4) based on experimental data obtained from NMR studies by Kua et al. 5 . Experimental equilibrium concentrations were used to assign arbitrary kinetic parameters according to the Law of mass action as shown in equation (6) for each equilibrium.
The change in Gibbs free energy for the one-substrate transketolase reaction was calculated according to equation (7) using the Gibbs free energies of formation of the substrates and products shown in table (3). The formation of erythrulose from LiHPA via decarboxylation was implemented as irreversible in the model. The free energy change of the one-substrate reaction was then correlated to the corresponding equilibrium ratio using equation (9). S17 -29,2 6 -118,4 7 -237,1 8 -62,4 9 -586, 9 8 Table S 3: Gibbs Free Energies of substrates and products involved in the transketolase mediated synthesis of erythrulose.
The system was then described as a coupled set of first order (10) and second order (11) differential equations, allowing the system to converge towards thermodynamic equilibrium from any starting point of choice.