Combinatorial Library Based Engineering of Candida antarctica Lipase A for Enantioselective Transacylation of sec-Alcohols in Organic Solvent

A method for determining lipase enantioselectivity in the transacylation of sec-alcohols in organic solvent was developed. The method was applied to a model library of Candida antarctica lipase A (CalA) variants for improved enantioselectivity (E values) in the kinetic resolution of 1-phenylethanol in isooctane. A focused combinatorial gene library simultaneously targeting seven positions in the enzyme active site was designed. Enzyme variants were immobilized on nickel-coated 96-well microtiter plates through a histidine tag (His6-tag), screened for transacylation of 1-phenylethanol in isooctane, and analyzed by GC. The highest enantioselectivity was shown by the double mutant Y93L/L367I. This enzyme variant gave an E value of 100 (R), which is a dramatic improvement on the wild-type CalA (E=3). This variant also showed high to excellent enantioselectivity for other secondary alcohols tested.


Chemicals, enzymes and molecular biology kits:
All chemicals were purchased from Sigma-Aldrich, except Zeocin™ that was purchased from Invitrogen. Phusion Hot Start II DNA polymerase and FastDigest DpNI was purchased from Thermo Fischer Scientific Inc. Wizard® SV Gel and PCR product clean up PureYield system and Bacterial plasmid preparation kit was purchased from Promega. An E.Z.N.A.® Yeast Prep Kit was supplied from Omega-Bio Tek.

Equipment:
Bio-Rad Micropulser was applied for electroporation using electroporation cuvettes (0.2 mm) from Bio-Rad. GC analyses were performed on a Varian GC 3900 using an IVADEX-1 chiral column from IVA Analysentechnik.

Selection of mutation points by molecular modeling:
A library of CalA was designed based on knowledge from molecular modeling. The crystal structure of CalA [1] (2veo.pdb) in closed form [2] was previously opened by a 10 ns molecular dynamics simulation. [3] Starting from the catalytic Ser184, tetrahedral intermediates of the model reaction 1phenylethyl butyrate ( Figure S1) in both (S)-or (R)-configuration was built into the open structure using the Yasara software (www.yasara.org) [4] and further evaluated separately. The structures were energy minimized using Amber99 force field,V [5] 7,86 Å force cut-off point and particle mesh Ewald algorithm [6] to treat long range electrostatic interactions. After removal of conformational stress by a short steepest descent minimization, the procedure continued by simulated annealing (timestep: 2 fs, atom velocities were scaled down by 0.9 every tenth step) until convergence was reached, that is, the energy improved by less than 0.1 % during 200 steps. The structures were step-wise energy minimized, by firstly keeping all but the tetrahedral intermediate fixed followed by allowing everything to be free. Amino acid residues closer than 5 Å to the tetrahedral intermediates were identified. Seven amino acid residues (93, 183, 233, 336, 367, 370, 431) were selected to be included in a focused combinatorial library.

pBGP1-CalA plasmid:
The construction of the pBGP1-CalA plasmid has been described previously. [2] The template vector pBGP1 is an episomally replicating plasmid, useful for constitutively expressed enzyme libraries. [7] The α-factor secretion signal in the vector allows secretion of the produced enzyme into the supernatant.

Primers:
The following primers were used to construct the gene library. Primers are written in 5´-3´ and were ordered from Eurofins MWG Operon.

Preparation of enzyme library
Transformation into competent DH5α cells: Competent cells were made heat-shock competent and stored in -80 C until use. [9] Aliquots (100 μl) of the competent cells were thawed on ice. PCR product (1-5 μl) was added and gently mixed with the cells. The mixture was then left on ice for 30 min. The cells were heat shocked (42 °C, 45 s) in order for uptake of the plasmid to occur, held on ice for two minutes, where after SOC media (300 μl) was added. The cells were incubated in an orbital shaker (45-50 min, 37 °C, 250 rpm) after which 100 μl was plated on LA-cb plates and the rest shaken at 37 °C in 100 ml LB + 100 μl carbenicillin. After incubation overnight the library plasmid was extracted from this culture. Colonies appeared after incubation overnight (37 C) and a small fraction was cultivated for individual minipreps and subsequent sequencing, verifying the diversity of the library.
Transformation to Pichia pastoris: P. pastoris X33 (Invitrogen) was made electrocompetent by adhering strictly to the manufacturer's protocol. Until employed, cells were stored in 40 μL aliquots at -80 °C without any treatment. Thawed X33 cells were mixed with library plasmid (1-3 L) and electroporated (0.2 mm cuvettes, 1.5 kV, 1 pulse). Cells were incubated at 30 °C with YPDS (YPD with 1 M D-sorbitol, 1 mL) for 1-2 h followed by plating on YPD-agar plates containing zeocin (100 μg/mL), carbenicillin (100 μg/mL) and tributyrin (0.5 % v/v). After four days of incubation (30 C), colonies expressing active lipase displayed a clear halo around them. Those colonies were picked and inoculated in a deep conical 96 well plates containing YPD media (1,1 mL). The plates were incubated in an orbital shaker (270 rpm, 30 C, 8-10 days), where after they were harvested by centrifugation to obtain enzyme in the supernatant.

Organic solvent screening:
The enzyme supernatant was immobilized on Nunc Immobilizer Nickel-Chelate microtiter plates according to the manufacturers protocol. The plates were thereafter washed with isooctane (3 x 300 L) to remove residual water. A screening solution (20 mM phenylethanol, 200 mM vinyl butyrate and 20 mM dodecane in isooctane) was premixed before addition to the plate using a multi-channel pipette. To each well was added 150 L of the screening solution followed by incubation at 21 C on a shaker. After 2.5 h a sample (13 L) was removed from each well to a second microtiter plate containing 137 L isooctane for GC-analysis. The plate was screened by chiral gas chromatography (GC-FID). The total time to screen through one 96-well plate was around 24 h, when using a method of 14.2 min. The GC-spectra from one plate is quickly overviewed by using an overlay view. Only chromatograms indicating hits (i.e. product) require integration. By further calculation, both conversion and enantioselectivity can be determined. Variants showing an enantiomeric excess (ee p ) over 95 % were determined to be hits. In total, 17 hits were cultivated in 50 mL scale for further characterization (Table S1).

GC screen:
GC-analyses were performed using an IVADEX-1 chiral column. The screening was performed in 96 well microtiter plate format, where continuous integration and multi overlay of chromatograms simplified fast determination of possible hits. The GC program started at 90 C (4 min), continued to 172 C (10 C/min) and then to 200 C (100 C/min, then hold for 2 min).

Sequencing hits:
Pelleted cells from master plates were used to inoculate YPD containing zeocin (100 μg/mL) and carbenicillin (100 μg/mL), and cultures were shaken at 30 °C over night. The plasmids were extracted using Yeast plasmid kit (Omega Bio-Tek), and subsequently transformed into Escherichia coli DH5α to obtain a higher plasmid yield. The bacterial cells were cultivated, plasmids were extracted and sequenced using the sequencing primerspBGP1_for (gtccctatttcaatcaattgaac) and pBGP1_rev (gtaagtgcccaacttgaactgag).

TABLE S1. Conversion and enantiosselectivity shown by CalA variants found by library screening with product enantiomeric excess over 95% in kinetic resolution of the model reaction shown in Scheme 1. [a]
Values are determined by chiral GC. [b] The CalA variants are numbered according to the library number followed by the position in 96-well microtiter plate.

CalA variant [b] Conversion (%) [c] ee
[c] The conversion is determined after 2.5 hours of reaction time by the use of dodecane as an internal standard and shown in percent.
[d] The enantioselectivity for the product (ee p ) is determined from GC analysis by (A R(product) -A S(product) )/(A R(product) +A S(product) ) and shown in percent.
[e] The E-value is determined according to the equation shown by Rakel et al. [10] [f] Mutation determined by sequencing. [b] The conversion is determined after 2.5 hours of reaction time by the use of dodecane as an internal standard and shown in percent.
[c] The enantioselectivity for the product (ee p ) is determined from GC analysis by (A R(product) -A S(product) )/(A R(product) +A S(product) ) and shown in percent.
[d] The E-value is determined according to the equation shown by Rakel et al. [10] Sequencing hits: Pelleted cells from master plates were used to inoculate YPD containing zeocin (100 μg/mL) and carbenicillin (100 μg/mL), and cultures were shaken at 30 °C over night. The plasmids were extracted using Yeast plasmid kit (Omega Bio-Tek), and subsequently transformed into Escherichia coli DH5α to obtain a higher plasmid yield. The bacterial cells were cultivated, plasmids were extracted and sequenced using the primers pBGP1_for (gtccctatttcaatcaattgaac) and pBGP1_rev (gtaagtgcccaacttgaactgag)

Determination of kinetic parameters, K M and K I :
Kinetic constants for wild-type CalA and Y93L/L367I were determined for 1-phenylethanol under pseudo-one substrate conditions using the model reaction screening conditions. Enzyme supernatant was immobilized on Nunc Immobilizer Nickel-Chelate microtiter plates according to the manufacturers protocol. The plates were thereafter washed with isooctane (3 x 300 L) to remove residual water.
Solutions (3-100 mM 1-phenylethanol, 200 mM vinyl butyrate and 20 mM dodecane in isooctane) were premixed before addition to plate. To each well was added 150 L of this solution followed by incubation at 21 C on shaker. Samples were taken regularly (13 L) to a second microtiter plate containing isooctane (137 L) for GC-analysis. The kinetic constants were obtained by non-linear regression of the initial rates using the Michaelis-Menten equation for competitive substrate inhibition.

Exploring hit mutations by molecular modeling:
The effects of the mutation points on enantioselectivity were further explored by molecular modeling using a tetrahedral intermediate model of the model substrate in (R)-and (S)-configuration ( Figure S1). The structure was prepared using the YASARA software [4] . Mutant Y93L/L367I and the two single-point mutants (Y93L and L367I) were created by swapping the specific amino acid residues. The structures were step-wise energy minimized, by firstly keeping all but the mutation point and the intermediate fixed and then free everything.
Both the (R)-or (S)-configurations of the tetrahedral intermediate in the different variant structures were compared by alignments to the corresponding ones in the wild-type enzyme using MUSTANG in the YASARA software. [4,11] The alignments showed that the configurations of the tetrahedral (S)-and (R)-intermediate phenyl groups were affected to different extent by the mutations. The mutations induce movements in the configuration of the tetrahedral intermediate phenyl groups due to steric hindrance. To evaluate to which extent each carbon in the phenyl groups have to move, the distance to its original position in the wild-type structure was measured. Table S3 shows the distance that each carbon in the phenyl group has to move, compared to its position in the wild-type structure, to find the proper orientation in the mutant structures. Each carbon in the tetrahedral intermediate phenyl group is numbered (C no ) as shown in Figure S1.  (Table 3 entry 1 and Table 4).   Figure  S1.