A transaminase-mediated aldol reaction and applications in cascades to styryl pyridines

Transaminase enzymes are well established biocatalysts that are used in chemical synthesis due to their beneficial sustainability profile, regio- and stereoselectivity and substrate specificity. Here, the use of a wild-type Chromobacterium violaceum transaminase (CvTAm) in enzyme cascades revealed the formation of a novel hydroxystyryl pyridine product. Subsequent studies established it was a transaminase mediated reaction where it was exhibiting apparent aldolase reactivity. This promiscuous enzyme reaction mechanism was then explored using other wild-type transaminases and via the formation of CvTAm mutants. Application of one pot multi-step enzyme cascades was subsequently developed to produce a range of hydroxystyryl pyridines.


Transaminase expression
E. coli BL21 (DE3) was used as the expression host and plated out on agar plates supplemented with 50 µg/mL kanamycin.A single colony was then picked to inoculate into 5 mL of 2 x TY media supplemented with 50 µg/mL kanamycin and grown at 37 °C and 250 rpm overnight (8-16 h). 1 mL of the overnight cultures was then added into a 500 mL baffled shaking flask containing 100 mL of 2 x TY media supplemented with 50 μg/mL kanamycin at 37 °C, 250 rpm until an OD600 = 0.6.Enzyme expression was induced by the addition of 500 µM IPTG to the culture.Cultures were incubated overnight at 30 °C prior to harvesting, whilst shaking at 250 rpm.Cells were harvested by centrifugation (10,000 rpm, 15 mins) and the cell pellets was stored at −20 °C.

Transaminase cell lysate preparation
Cell pellets (50 mL culture) were resuspended in 5 mL of 50 mM HEPES buffer (pH 7.5) supplemented with 10 mM PLP and lysed by 10 cycles of sonication on ice (10 s on plus 10 s off, 12 watts output) equipped with a Process Timer.Cells lysates were centrifuged at 4 °C (10,000 rpm, 15 min).The supernatant was collected.The concentration of supernatant protein was measured following the standard Bradford procedure.The samples were duplicated and the average OD595 were used for cell lysate concentration calculations.

Transaminase purification
Cell lysates were filtered through a 0.2 μm cellulose acetate springe filter to remove insoluble cell components.A PD-10 column charged with Ni-NTA (5 mL) was washed with 10 mL of MilliQ TM water, followed by 10 mL of binding buffer (50 mM HEPES, 10 mM imidazole (Sigma-Aldrich), pH 7.5).The filtered supernatant was then passed through the Ni-NTA column, and the column was washed with wash buffer (2 × 5 mL, 50 mM HEPES, 20 mM imidazole, pH 7.5) to remove some background protein.The bound protein was then eluted with elution buffer (50 mM HEPES, 500 mM imidazole, 100 mM NaCl, pH 7.5) until all the protein was collected.The eluent containing pure enzyme was concentrated using a vivaspin (30,000 MW) at 4 °C, (8,000 rpm, 5 min) until 2.5 mL eluent remained.
Then the concentrated eluent was desalted into 3 mL of 50 mM HEPES (pH 7.5), using a Sephadex TM G-25 in PD-10 column (GE Healthcare Life Sciences, Germany).To store the pure enzyme, 10% (v/v) glycerol was added.The concentration of the pure protein was measured by OD280 using a Nanodrop.The protein was split into different tubes with 0.5 mL/each, and stored at −20 °C.To check the protein purity, the expression supernatant, flow through, wash, and eluents were examined using an SDS gel (Figure S1).  and IPTG were purchased from Alfa Aesar (Thermo Fisher Scientific, USA).All chemicals were purchased in the highest purity available.

Method for high performance liquid chromatography (HPLC)
These were performed with a Dionex

Analytical HPLC Method 1 (achiral)
Achiral quantitative analyses adopted a reverse phase analysis method.Separation was achieved with an ACE 5 C18 column (150 × 4.6 mm) with a flow speed of 1 mL/min at 30 °C.The injection volume was 10 μL.Substrates and products were measured via UV absorbance at 280 nm.Eluent A (H2O with (v/v) 0.1% TFA) and eluent B (acetonitrile) were used as a mobile phase over 10 mins.
The gradient is shown below (Figure S2).

Preparative HPLC Method 2
Methods were developed with a Agilent 1260 Infinity The separation was achieved with a Vydac TM 218TP1022 (C18, 10 µm, 2.2 cm ID x 25 cm L) preparative column or a Supelco TM Discovery BIO wide pore (C18, 10 µm, 2.12 cm x 25 cm) preparative column and a flow speed of 8 mL/min at 25 °C.The injection volume was 900 μL.
Products were identified via UV absorbances at 214 nm and 280 nm.Eluent A (H2O with 0.1% (v/v) TFA) and eluent B (acetonitrile with 0.1% (v/v) TFA) were used as a mobile phase over 28 mins.The gradient is shown below (Figure S3).

Methods for mass spectrometry (MS)
The molecular masses of new compounds were measured on an Agilent 1100 Series System with a Finnigan LTQ mass spectrometer.An ACE 5 C18 reverse phase column (50 mm × 2.1 mm, 5 μm) was adopted with a mobile phase of eluent A (H2O with 0.1% (v/v) formic acid) and eluent B (acetonitrile) over 5 min with a flow rate of 0.6 mL/min.The sample injection volume was 10 μL.
Chemical compounds were measured in a positive ion mode, and the operating conditions of the ESI interface were set to a capillary temperature 300 °C, capillary voltage 9 V, spray voltage 4 kV, sheath gas 40, auxillary gas 10, sweep gas 0 arbitrary units.The gradient of eluents was as follows (Figure S4).
The reaction was quenched by adding 1% TFA.The product was purified using preparative HPLC (method 2, Supelco TM Discovery BIO wide pore (C18, 10 µm, 2.12 cm x 25 cm) preparative column, retention time: 15.1 min, run time: 28 mins, flow rate: 8 mL/min)).Fractions containing the desired product were freeze-dried to give product 7b as an off-white powder (yield by HPLC (calibration curve) 36% (method

Single crystal X-ray diffraction studies
The diffraction data for compounds 3a-OEt and 3b were collected on a four-circle Agilent SuperNova (Dual Source) single crystal X-ray diffractometer using a micro-focus CuKα X-ray beam (λ = 1.54184Å) and an Atlas CCD detector.The sample temperatures were controlled with an Oxford Instruments cryojet.All data were processed using the CrysAlis Pro programme package from Rigaku Oxford Diffraction. 1 The crystal structures were solved with the ShelXT programme 2 and refined by least squares on the basis of F 2 with the ShelXL programme. 3Both programmes were used within the Olex 2 software suite. 4,5 note that only the crystal structure of 3b passes the checkCIF validations for data completeness and consistency.The single crystals of 3a-OEt were too small and diffracted only up to a data resolution of 0.99 Å.And since the collection of a full data set was not possible, and the available data was only used to guide NMR data interpretation.

Crystal structure refinement process for 3a-OEt
All non-hydrogen atoms were refined anisotropically by the full-matrix least-squares method.
Hydrogen atoms affiliated with oxygen and nitrogen atoms were refined isotropically [Uiso(H) = 1.5Ueq(O)] in geometrically constrained positions.Hydrogen atoms associated with carbon and nitrogen atoms were refined isotropically [Uiso(H) = 1.2Ueq(C/N)] in geometrically constrained positions.
The crystal structure of 3a is shown in Figure S5, and its crystallographic and refinement parameters are shown in Table S1.

Crystal structure refinement process for 3b
All non-hydrogen atoms were refined anisotropically by the full-matrix least-squares method.The locations of the hydrogen atoms affiliated with the N1 nitrogen atom and the C10 carbon atom were identified from the difference map and were refined isotropically [Uiso(N-H) = 1.2Ueq(N) and Uiso(C- The structure of the disordered trifluoroacetate anion in the major occupancy site was modelled using rigid-body fragment fitting (with the FRAG and FEND commands in SHELX 3 ).The structures of the anion in the other two minor occupancy sites were refined using the SAME restraints command in SHELX. 3e crystal structure of 3b is shown in Figure S6.The crystallographic and refinement parameters are shown in Table S1.

Molecular dynamic simulations and MM-PBSA calculation
To investigate in silico whether the hydroxystyryl pyridines generated had the potential to inhibit human pancreatic amylase (HPA), molecular dynamics modeling of 3a and HPA (PDB: 2QMK44) was conducted with GROMACS 2020.4,45-47 using the AMBER99SB-ILDN forcefield to investigate the structural flexibility of predicted protein-ligand complexes.48As comparisons, the natural HPA inhibitor montbretin A (MbA, PDB: 4W93)49 and dehydrodieugenol B (DDEB)50 were also investigated.A 30 ns molecular dynamics simulation for each ligand was performed in triplicate on the entire system at 300 K. Rootmean-square deviation (RMSD) of the backbone group was calculated using the initial structure as a reference to assess structural stability, and the molecular mechanics Generalised Born surface area (MM-GBSA) method was employed to investigate changes in binding free energy within a protein-ligand system.According to the dynamic simulation, 3a bound to HPA tightly and the RMSD fluctuated between 0.1 and 0.2 with the totalΔGBSA at −29.56 (Figure S8A).DDEB can bind to HPA with a slightly higher fluctuation between 0.1 and 0.25, and the total ΔGBSA was −18.62 (Figure S8B).Interestingly, MbA dissociated from HPA after 1 ns and the ΔGBSA reached up to 10.76 (Figure S8C and movie 3).These dynamic simulations indicated that the hydroxystyryl pyridine 3a could potentially be an inhibitor of such amylases.Molecular dynamic simulation software GROMACS 2020.4 with AMBER99SB-ILDN forcefield 14-17 was used to investigate the structural flexibility of predicted protein-ligand complexes.The complex topology files were prepared using Ambertools and ACPYPE. 18The starting structure for the molecular dynamic simulation was solvated in a cubic simulation box with water and neutralized using an adequate amount of Na + .The entire system was energy-minimized using the steepest descent method (2000 steps) followed by the conjugate gradient method (5000 steps).Two-phase equilibration was carried out under the NVT and NPT ensembles for 50 ns each.Finally, a 30-ns molecular dynamics simulation was performed in triplicate on the entire system at 300 K. RMSD of the backbone group was calculated using the initial structure as a reference to assess structural stability.
Molecular mechanics Generalised Born surface area (MM-GBSA) method was employed to investigate changes in binding free energy within a protein-ligand system.Trajectories derived from molecular dynamics simulations were analysed using the gmx_MMPBSA tool 19 after the elimination of PBC conditions.The per-residue effective free energy decomposition (prEFED) protocol was utilized to identify energetically significant residues located within 4 Å of the protein-ligand interface.The AMBER99SB force field was utilized to compute the internal energy term (ΔEint) as well as the
The yield of 5a refers to the yield at equilibrium.For the kinetic study of transamination activities (here conversion of the amine to the aldehyde), the concentration of 4a was varied from 0.5 mM to 20 mM (1 eq.).Reactions contain 1 eq. of pyruvate and 0.1 eq. of 1 in HEPES buffer (50 mM, pH 7.5) and purified CvTAm at a final concentration of 10 µg/mL was used.Reactions were performed at 37 o C, 700 rpm.Samples of each reaction were obtained at 2 min, 5 min, 10 min and 20 min, and quenched by flash freeze-drying.Samples were then measurement by HPLC at 280 nm.The apparent Km.app is 1.69 mM and kcat.app is 6.04 s -1 , giving kcat.app/Km.app= 3.57 s -1 mM -1 .
For the kinetic study of aldol addition activities, the concentration of 4a was varied from 0.1 mM to 15 mM (1 eq.).Reactions contained 1 eq. of pyruvate and 1.5 eq. of 1 in HEPES buffer (50 mM, pH 7.5) and purified CvTAm at a final concentration of 10 µg/mL was used.Reactions were performed at 37 o C, 700 rpm.Samples of each reaction were obtained at 2 min, 5 min, 10 min, 20 min, 30 min and 40 min, and quenched by flash freeze-drying.Samples were then measurement by HPLC at 280 nm.The apparent Km.app is 9.84 mM and kcat.app is 1.75 s -1 , giving kcat.app/Km.app= 0.18 s -1 mM -1 .

Figure S4 .
Figure S4.Gradient of the LC-MS method.

1 H and 13 C
NMR spectra were recorded respectively at 600 MHz and 150 MHz on a Bruker Avance 600 spectrometer or at 700 MHz and 175 MHz on a Bruker Avance 700 spectrometer in the stated solvent.Chemical shifts (in ppm) are quoted relative to tetramethylsilane and referenced to residual Formic acid) Acetonitrile protonated solvent.Coupling constants (J) are measured in Hertz (Hz) and multiplicities for 1 H NMR couplings are shown as s (singlet), d (doublet), t (triplet), and m (multiplet).
2Ueq(C)], as calculated positions resulted in molecular geometries with an unreasonably short C−H•••H-O contact.All other hydrogen atoms associated with carbon atoms were refined isotropically [Uiso(H) = 1.2Ueq(C)] in geometrically constrained positions.

Figure S8 .
Figure S8.Dynamic simulation of ligands (3a, DDEB AND MbA) with HPA. A. Dynamic simulation of 3a with HPA.B. Dynamic simulation of DDEB with HPA with DDEB fitted into the active site of HPA. C. Dynamic simulation of MbA with HPA.

Figure S9 .
Figure S9.Production of 3a and 5a using different equivalents of pyruvate.Reactions were performed with 4a (10 mM, 1

Figure S10 .
Figure S10.Production of 3a and 5a using different equivalents of PLP 1.Reactions were performed with 4a (10 mM, 1 TM UltiMate TM 3000 HPLC System, with a Dionex TM UltiMate TM 3000 RS Pump, a Dionex TM UltiMate TM 3000 Autosampler, a Dionex TM UltiMate TM 3000 Column Compartment and a UltiMate TM 3000 RS Diode Array Detector (Thermofisher Scientific, US).

Table S1 .
Crystal data and structure refinement for compounds 3a-OEt and 3b.
a.The ranking order followed the affinity energy.