Evolving the Promiscuity of Elizabethkingia meningoseptica Oleate Hydratase for the Regio‐ and Stereoselective Hydration of Oleic Acid Derivatives

Abstract The addition of water to non‐activated carbon–carbon double bonds catalyzed by fatty acid hydratases (FAHYs) allows for highly regio‐ and stereoselective oxyfunctionalization of renewable oil feedstock. So far, the applicability of FAHYs has been limited to free fatty acids, mainly owing to the requirement of a carboxylate function for substrate recognition and binding. Herein, we describe for the first time the hydration of oleic acid (OA) derivatives lacking this free carboxylate by the oleate hydratase from Elizabethkingia meningoseptica (OhyA). Molecular docking of OA to the OhyA 3D‐structure and a sequence alignment uncovered conserved amino acid residues at the entrance of the substrate channel as target positions for enzyme engineering. Exchange of selected amino acids gave rise to OhyA variants which showed up to an 18‐fold improved conversion of OA derivatives, while retaining the excellent regio‐ and stereoselectivity in the olefin hydration reaction.


Experimental Procedures General
Unless stated otherwise, standard laboratory reagents were obtained from Sigma-Aldrich® (Steinheim, Germany) or Carl Roth GmbH & Co. KG (Karlsruhe, Germany) with the highest purity available. Oleic acid (OA) and esters thereof (methyl, ethyl, i-propyl), oleyl alcohol and oleyl amine were purchased from Sigma-Aldrich® (Steinheim, Germany). The OA methyl and ethyl ester and oleyl alcohol were distilled prior to use to a purity >90% according to GC-FID analysis. (R)-10-hydroxy stearic acid was obtained from DSM Innovative Synthesis B.V. (Geleen, Netherlands). Other starting materials used for the investigations were synthesized from suitable precursors as described below or used as received from Sigma-Aldrich®. Flash column chromatography was performed on Acros Organics silica gel 0.035-0.070 mm, 60 Å. Analytical thin layer chromatography (TLC) was performed using TLC-plates from Merck (TLC aluminium foil, silica gel 60 F254) and subsequent visualization with cerium ammonium molybdate stain. The specific optical rotations were determined on a Perkin Elmer Polarimeter 341 with an integrated sodium vapor lamp. All samples were measured at the D-line of the sodium light (λ = 589 nm).

Amino acid sequence alignments
The protein sequence of OhyA was compared to the amino acid sequences in the Hydratase Engineering Database (HyED), in which a total of 2046 sequences are collected. [3] Since OhyA is categorized in homologous family 11 (HFam11) of the HyED, all amino acid sequences from HFam11 were selected for the multiple sequence alignment. Sequences were extracted from the database for a multiple sequence alignment with the Clustal Omega sequence alignment tool using default settings, [4] and were visualized with the Unipro UGene software.

Site-directed mutagenesis
Amino acid exchange variants of OhyA were generated by site-directed mutagenesis using a modified Stratagene QuikChange™ site-directed mutagenesis protocol. Twenty-five µL of two separate PCR reactions containing forward and reverse primers, respectively, were prepared (Fehler! Verweisquelle konnte nicht gefunden werden. Table S1). After five cycling steps, PCR reactions were combined and the PCR was continued for 20 additional cycles. Mutated plasmids were verified by sequencing of the coding regions of the constructs.

Whole cell bioconversions
Bioconversion assays of 1a-j were performed with E. coli BL21 Star (DE3) cells immediately after expression of recombinant OhyA. Fifty OD 600 units, which are corresponding to a cell dry weight of 50 mg, were resuspended in 50 mM HEPES, pH 6.0, supplemented with 100 mM glucose and 0.2 mM FAD in Pyrex® glass culture tubes (Corning, NY). Biotransformations at 1 mL scale were started by adding substrate to a final concentration of 2 mM from an ethanolic stock solution (100 mM). n-pentadecanoic acid (1 mM) was used as internal standard. The reactions were conducted in the presence of 2% (v/v) of ethanol as co-solvent at 30°C and shaking at 150 rpm at a defined angle of the Pyrex® tubes (55°). Biotransformations were performed for 22 h or 96 h. Whole cell bioconversions of OA esters 1f and 1g with alcohol additives were performed with 50 mg of E. coli BL21 Star (DE3) after recombinant expression of OhyA Gln265Ala/Thr436Ala/Asn438Ala. Biotransformations of 1f were coincubated with equimolar concentrations of the substrate and either ethanol or i-propanol, and biotransformations of 1g were co-incubated with equimolar concentrations of the substrate and either methanol or i-propanol. Otherwise, assay conditions were maintained as described above.

GC-MS analyses
Free fatty acids and derivatives thereof were initially analyzed and identified by gas chromatography-mass spectrometry (GC-MS). A HP-5 column (crosslinked 5% Ph-Me Polysiloxane; 30 m length, 0.25 mm in diameter and 0.25 µm film thickness) on a Hewlett-Packard 6890 Series II GC equipped with a mass selective detector was used. Sample aliquots of 1 µL were injected in split mode (split ratio 30:1) at 240°C injector temperature and 290°C detector temperature with N 2 as carrier at a flow rate set to 36 cm s -1 in constant flow mode. The temperature program was as follows: 100°C for 1 min, 15°C min -1 to 300°C, hold for 5 min. The total run time was 19.33 min. The mass selective detector was operated in a mass range of 50-400 amu at an electron multiplier voltage of 1765 V. Results were evaluated with the GC-MS Data Analysis software (Agilent Technologies, Austria).

GC-FID analyses
Product formation was quantified by GC after derivatization of extracted samples with BSTFA. A Shimadzu GC-2010 Plus instrument equipped with a flame ionization detector and a Phenomenex Zebron ZB-5 column (crosslinked 5% Ph-Me Polysiloxane; 30 m length, 0.32 mm in diameter and 0.25 µm film thickness) was used. Sample aliquots of 1 µL were injected in split mode (split ratio 10:1) at 240°C injector temperature and 320°C detector temperature. N 2 was used as carrier gas at a flow rate set to 20 cm s -1 in constant flow mode. The oven temperature program was as follows: 70°C for 4 min, 35°C min -1 to 300°C, hold for 5 min. The total run time was 15.57 min.

Preparation of OA derivatives
Oleamide (1c) was obtained via a literature procedure [5] and the material was purified through recrystallization from acetone [6] to a purity of 95% as checked by rp-HPLC at 210 nm. n-propyl (1i) and n-butyl oleate (1j) were synthesized from OA (1a) via Fischer esterification as described in the literature [7] and purified via flash chromatography on silica gel using cyclohexane/ethyl acetate 20:1 as eluent.

Preparative-scale hydration of OA derivatives
OA derivatives 1c-1j were hydrated to 2c-2j in a semi-preparative scale. Twenty to 150 mg of non-physiological substrates were converted in 1 mL scale whole cell bioconversions. Each reaction contained 200 mg of E. coli cells in Pyrex® glass tubes, after over-expression of OhyA Gln265Ala/Thr436Ala/Asn438Ala, resuspended in 50 mM HEPES, pH 6.0, containing 100 mM glucose and 0.2 mM FAD. Biotransformations were incubated for 96 h at 30°C and 150 rpm at a defined angle of the Pyrex® tubes (55°). After quenching by acidification to pH 2.0 with 0.12 M HCl, the suspensions were extracted with ethyl acetate (3 × 2 mL for 30 min) with intermittent centrifugation for 5 min at 2,900 × g and 22°C to improve phase separation. The organic phases were quantitatively collected and concentrated under a stream of N 2 .

Purification of crude reaction products
The products extracted with ethyl acetate were purified via flash chromatography (9.5 g silica gel, 20 × 1 cm column size) using eluent mixtures of cyclohexane/ethyl acetate in ratios from 10:1 to 1:1 (v/v) dependent on the polarity of the respective derivative. All fractions containing the desired product were pooled and evaporated to dryness.

Modeling of OA and derivatives to the OhyA 3D structure
Docking of OA to the OhyA 3D structure was performed using AutoDock implemented in YASARA structure as described previously. [2] Briefly, receptor (chain A of OhyA; PDB code: 4uir) and ligand (OA; formal charge of -1) were prepared and energy minimized with the Schrodinger package. The receptor was kept rigid, and the ligand had full 6 conformational flexibility around each single bond. The docking box (x=25 Å, y=27 Å, z= 25 Å) was set to be close to the flavin cofactor covering the elongated active site cavity. After 50 individual runs, the best docking modes were sorted by binding energies and chemical plausibility and were finally clustered according to a maximum root-mean-square deviation (r.m.s.d.) of the heavy atoms of 2.0 Å. OA derivatives 1c-1j were prepared using YASARA and AM1 charges were applied accordingly. The derivatives 1c-1j were docked to the 3D structure model of OhyA after introducing amino acid exchanges that resulted in the best activity on each substrate. The binding mode with the lowest docking energy was visually inspected. Amino acid exchanges were introduced in silico using YASARA. Docking was performed using VINA [8] implemented in YASARA structure in analogy to OA after substitution of the carboxylate for the different head groups of 1c-1j using 20 independent docking runs and a 2.0 Å r.m.s.d cluster deviation. The docking with the lowest energy for 1i and 1j did not result in a productive binding mode, even in the triple variant (visual inspection). This may be due to the rigid receptor docking, which resulted in higher docking energies for the larger derivatives in the productive binding mode. By overlaying of 1i and 1j with the docked OA and by performing an energy minimization step (Force Field: Amber03), the ligands 1i and 1j fitted quite well into the larger pocket introduced by the mutations. It can be assumed that small changes in the structure on the site of the mutations, although not optimal, do also provide enough space for the larger derivatives 1i and 1j.

Nucleotide information
Open reading frame of the codon-optimized E. meningoseptica oleate hydratase gene (OhyA, GenBank: ACT54545.1) used in this study:

GC-MS monitoring of the hydration of OA derivatives
Bioconversions of 1a-1j were performed with whole E. coli cells. Representative GC-chromatograms of technical triplicates of an authentic substrate standard, an OhyA-free E. coli strain (empty vector control, EVC) and a biotransformation with cells after over-expression of OhyA are overlaid ( Figure S1-Figure S10). In case a substrate was not converted with the wild type enzyme (1h and 1j), a representative chromatogram of the OhyA Gln265Ala/Thr436Ala/Asn438Ala bioconversion is shown. The OA-derived N-hydroxy oleamide (1d) and the hydrated reaction product (2d) were both detected as the respective isocyanates after a Lossen rearrangement occurring under GC-MS analysis conditions. [9] Moreover, conversion of 1d with the E. coli EVC and the strain expressing OhyA led to the unexpected formation of oleamide (1c), with a subsequent hydration to 10-hydroxy octadecanamide (2c) only in OhyA biotransformations. Since N-hydroxy oleamide (1d) was initially oleamide-free, one must assume that the oleamide (1c) was formed by degradation of N-hydroxy oleamide (1d) in E. coli. [10] Figure S1

Amino acid sequence alignment
To support our selection of amino acid residues involved in substrate binding in OhyA, we performed alignments of the OhyA protein sequence with all sequences of HFam11 in the HyED ( Figure S29). [3] The high degree of conservation of Gln265, Thr436, Asn438 and His442 is highlighted by the red boxes, and perfectly in line with our docking and sitedirected mutagenesis analyses. [2] 27 a) b) Figure S29. A segment of the multiple sequence alignment of the 116 amino acid sequences from HFam11 collected in the hydratase engineering database (HyED), highlighting the conserved residues involved in binding of a carboxylate (red boxes). a) Gln265 is conserved among all members of HFam11. b) Thr436 and Asn438 are conserved among all members of HFam11, while His442 is conserved among all but one member, where it is substituted with a Gln.

Conversion of OA derivatives with OhyA wild type and substrate binding variants
We compared the activity of OhyA wild type and all solubly expressed variants for hydration of 1a-1j by performing bioconversions with whole E. coli cells for 22 h ( Figure S32 -Figure S39).

Determination of the enantiomeric excess of reaction products by 1 H-NMR analysis
All products from the enzyme-catalyzed hydration reactions were O-acylated with (S)-(+)-O-acetylmandelic acid using a known procedure and purified by flash chromatography on silica gel using cyclohexane/ethyl acetate mixtures. [2] A reference material was obtained from methyl rac-10-hydroxy stearic acid for comparison in the 1 H-NMR analyses. [2] Figure S42. 1   Based on the clear NMR proof of (R)-selectivity in case of the hydrated compounds 2a and 2f, the strict enzymatic reaction mechanism involved and the fact that all other compounds 2c-2e and 2g-2j show a similarly shaped sharp NMR signal at 5.90 ppm, we conclude that the reaction proceeds in all cases with high stereoselectivity (ee ≥ 95 %) and that the products are the expected (R)-10-alcohols.

Determination of OhyA wild type and variant conversions by GC-FID
Hydration reactions of 1a-1j were quantified via GC on a Shimadzu GC-2010 Plus instrument equipped with a flame ionization detector and a Phenomenex Zebron ZB-5 column under the conditions described in the Experimental Procedures section of the Supporting Information. The improvement in catalytic activity is illustrated by representative chromatograms from wild type and the best variant conversions of each OA derivative ( Figure S49-S56). Integration results are shown as peak area values for each substrate and product, respectively.      Figure S56. Overlay of GC-FID chromatograms from bioconversions of n-propyl oleate (1i) to 2i with E. coli whole cells over-expressing OhyA wild type and OhyA Gln265Ala/Thr436Ala/Asn438Ala. Retention times and peak area values for the TMS-derivative of the internal standard n-pentadecanoic acid (11.21 min), 1i (12.30 min) and 2i (13.07 min) are highlighted. The insert shows a zoom-in to the section in which the reaction product 2i is eluting Figure S57. Overlay of GC-FID chromatograms from bioconversions of n-butyl oleate (1j) to 2j with E. coli whole cells over-expressing OhyA wild type and OhyA Gln265Ala/Thr436Ala/Asn438Ala. Retention times and peak area values for the TMS-derivative of the internal standard n-pentadecanoic acid (11.21 min), 1j (12.66 min) and 2j (13.50 min) are highlighted. The insert shows a zoom-in to section in which the reaction product 2j is eluting.

Control reactions with OA esters
In vivo hydrolysis of fatty acid esters is inherent to essentially all microbes, including E. coli. [12] To exclude any side reactions, in particular hydration of the corresponding free fatty acids after ester cleavage and re-esterification with available alcohols, respectively, we co-incubated 1f ( Figure S57) and 1g ( Figure S58)  Owing to the non-formation of 10-hydroxy stearic acid acid ethyl and i-propyl esters in biotransformations of 1f, as well as the absence of any 10-hydroxy stearic acid acid methyl and i-propyl esters in biotransformations of 1g, generation of free acid-derived hydration products was highly unlikely.