An Engineered Cholesterol Oxidase Catalyses Enantioselective Oxidation of Non‐steroidal Secondary Alcohols

Abstract The enantioselective oxidation of 2° alcohols to ketones is an important reaction in synthetic chemistry, especially if it can be achieved using O2‐driven alcohol oxidases under mild reaction conditions. However to date, oxidation of secondary alcohols using alcohol oxidases has focused on activated benzylic or allylic substrates, with unactivated secondary alcohols showing poor activity. Here we show that cholesterol oxidase (EC 1.1.3.6) could be engineered for activity towards a range of aliphatic, cyclic, acyclic, allylic and benzylic secondary alcohols. Additionally, since the variants demonstrated high (S)‐selectivity, deracemisation reactions were performed in the presence of ammonia borane to obtain enantiopure (R)‐alcohols.

Experimental Procedures

General chemicals
Substrates and product standards were purchased from Sigma Aldrich in the highest grade possible unless otherwise stated. (R)phenyl-2-propanol was from Fluorochem, and 4-phenyl-2-butanol, 3-methyl-2-butanol, and 2-pentanone were purchased from Alfa Aesar. The source of any other reagents that were used are referred to in the relevant section below. Solvents were either analytical or HPLC grade -tert-butyl methyl ether, ethyl acetate, and cyclohexane were purchased from Fisher Scientific.

Biotransformations and analysis
2.1 Biotransformations 500 µl biotransformation reactions were set up (final concentration enzyme: 17µM, final concentration substrate: 10 mM or 50 mM or 100 mM). The biotransformations were made up in air-saturated 100 mM potassium phosphate buffer, pH 8.0. Biotransformations were typically carried out in a 2 mL eppendorf tube. Substrates were added either neat or from a 1 M solution made up in DMSO (Sigma).
Reactions were incubated at 30 °C, 200 rpm for 24 hrs. Deracemisations additionally contained ammonia borane (Sigma) in a four-fold excess over substrate. Biotransformations were extracted by addition of 1 mL tert-butyl methyl ether (TBME), vortexing for 1 min followed by centrifugation at 13000 rpm for 2 min to separate the phases. The organic phase was analysed by GC-FID.

GC-FID methods
GC-FID analysis was performed on an Agilent 6850 GC with a Gerstel Multipurposesampler MPS2L. The methods used are given in Table SI2. Two columns were used for analysis: β-dex 325 (Supelco) with dimnsions 30 m x 0.25 mm x 0.25µm, and Chirasil β-dex CB (Agilent) with dimensions 25 m x 0.25 mm x 0.25µm with a detector temperature 250°C and injector temperature 250 °C. The methods used are described in Table S1 with Figure S1 showing which method was used for which substrate.

100 mg scaled reactions
To 20 mg of purified ShCO variant enzyme in 20 mL air-saturated 100 mM potassium phosphate buffer pH 8.0, in a 200 mL Duran bottle, was added 2 mg catalase (Sigma) and 100 mg of substrate, followed by 20 mL of TBME. The reaction was shaken at 150 rpm, 30 °C overnight. When GC-FID analysis showed complete conversion the product was extracted with 3-fold excess of TBME and dried over magnesium sulphate. TBME was removed under reduced pressure and no further purification was carried out.

Cloning
The gene for cholesterol oxidases from Streptomyces hygrospinosus, Brevibacterium sterolicum and Rhodococcus erythropolis were codon optimised for expression in E. coli and synthesised by GeneArt® (Life Technologies). The gene was sub-cloned into the vector pET28b (also digested with NdeI/XhoI and then CIP (New England Biolabs (NEB)) to remove phosphates) following the protocol in the Quick ligation kit (NEB) and transformed into E. coli NEB10β (NEB). Clones were sequenced at MWG Eurofins using T7 and T7term primers to verify the insert and then transformed into E. coli BL21 (DE3) (NEB) for expression.

Expression and purification
Overnight cultures were prepared by inoculating 6 mL of LB (complemted with 50 µg / mL kanamycin) with a single colony and grown for 16 hours at 37 °C at 200 rpm. The overnight culture was then added to a 2 L flask containing 600 mL autoinduction media (LB based with trace elements (Formedium)) and 50 µg / mL kanamycin and grown at 14 °C with shaking at 200 rpm for 96 hours. Cells were harvested by centrifutgation at 4000 rpm for 20 min and stored as pellets at -20 °C. Purification was carried out as in reference 2. [2] Table S2. N means any base and S = C/G.

Solid phase screen
Libraries were transformed into E. coli BL21 (DE3) (NEB) according to the manufactureres instructions, except that the whole transformation reaction was plated on top of a membrane (HyBond) on LB agar containing 50 μg mL-1 kanamycin and grown overnight at 30 °C. The membrane was then transferred to a second LB agar plate (containing 50 μg mL -1 kanamycin and 1 mM IPTG) for induction of prottein expression. After six hours at 20 °C the membranes were transferred to petri dishes and stored at -20 °C until use.
To perform the screen, filter paper containing 0.1 mg mL-1 Horse-radish peroxidase (HRP) (Sigma) in pH 8.0, 100 mM phosphate buffer was prepared. The membranes were freeze-thawed three times (using liquid N2) in order to partially lyse the colonies and were then placed on the filter paper. After one hour at room temperature (to ensure removal of any cellular H2O2) the membrane was transferred to another filter paper which had been soaked in assay solution. (The assay solution consited of of 0.1 mg mL-1 HRP, 3,3'-Diaminobenzidine (DAB) (1 tablet per 15 mL, SigmaFast, Sigma), and 100 mM substrate in 100 mM potassium phosphate buffer, pH 8.0.). A positive hit was indicated by colonies that changed to a dark red/brown colour. These colonies were culture and the plasmid DNA extracted and sequenced using T7 or T7 terminator primers (MWG Eurofins).

Kinetics
Cholesterol oxidase produces hydrogen peroxide as it turns over. Hydrogen peroxide production can be detected by horse-radish peroxidase (Type VI, Sigma) and the dye ABTS (Sigma) (ε = 36000 L mol -1 cm -1 ). The assay was carried out in a 96 well assay plate to which was added 50 µL substrate (dissolved in DMSO to 1 M then diluted in buffer), 50 µL ABTS (0.7 mg mL -1 ) and 50 µL HRP (0.4 mg mL -1 ). The assay was started by adding 50 µL purified enzyme (typically 0.1 -0.3 mg mL -1 ). Initial rates were calculated by following the absorbance at 420 nm over time on a TECAN Infinite M200 spectrophotometer at 30 °C. Typically, eight different substrates concentrations were examined (typically 250 mM down to 0.5 mM). A plot of rate vs substrate concentration ( Figure S17) allowed the extraction of Vmax and KM values using a non-linear curve fit analysis using the Hill model (OriginPro 9.1). kcat and KM error values represent a 95 % confidence interval based on the standard error of the regression.

T50 calculations
Purified enzyme was incubated at the given temperature for 15 minutes. An initial rates experiment was then conducted with 100 mM hexanol (see 3.5 for assay components) and the highest initial rate for each variant was set to 100%. Other rates were then compared to this to give a relative activity. A sigmoidal dose-response (variable slope) non-linear fit was applied in GraphPad Prism.

Results and Discussion
Hits from screening libraries The hits from each library are described in Table S3 with the most active variant confirmed in the right-hand column. These libraries were constructed because we found hits in the initial libraries and proposed that we could gain better activity by taking an iterative approach where one library is made in the background of a variant that is already active.
Additionally we combined E404C/P409S and P387W to make a three point variant but activity with 2-cyclohexen-1-ol went from full conversion (with E404C/P409S) to 25% conversion. A four point variant E404C/P409S/L418A/L420F was constructed and conversion went from full conversion (E404C/P409S) to 65% with 2-cyclohexen-1-ol although it was apparent that the enzyme was selective. F122C was combined with E404N/P409L but activity remained the same. Activities were comparable between E404N/P409L, and E404A/P409I on cyclohexanol but E404A/P409I showed better convesrions with the other substrates.

Kinetics of variants with hexanol and cylcohexanol
The kinetics of the variants were established with hexanol and cyclohexanol (Table S4)

Modelling of inactive substrate
Modelling of substrate 13 ( Figure S19) in the active site shows a potential clash with the FAD cofactor, suggesting why 13 is not a good substrate.

Temperature stability
The LogEC50 values returned (in this case the temperature at which 50% activity remained) were defined as the T50 for the enzyme of 50.9 ± 0.5 °C. ( Figure S20). Figure S20: Non-linear fit to data from temperature stability experiments using GraphPad Prism