Racemization-free and scalable amidation of L-proline in organic media using ammonia and biocatalyst only

Efficient amide formation is of high importance for the chemical and pharmaceutical industry. The direct biocatalytic one-pot transformation of acids into amides without substrate activation is a highly desirable but...

Solubility of L-proline in different solvents at various temperatures (data provided by Thermo Fisher Scientific, Linz). The effect of adding L-prolinamide or ammonia on the solubility of Lproline was investigated. 2M2B = 2-methyl-2-butanol. Enzyme engineering

Rational engineering of CalB
Interesting amino acid exchanges were selected from literature focussing on diverse strategies to improve product formation.

Solvent stability
Increased stability in organic solvents is a desired parameter. In literature enzyme variants with improved stability in MeOH have been reported. 1,4-dioxane is hydrophilic and miscible with water. In a paper by Park et al. the stabilization of CalB in hydrophilic organic solvents was performed by rational design of hydrogen bonds based on information of previously found stabilized variants by directed evolution. 2 The concept behind that study was to increase hydrogen bonds between the enzyme surface and surrounding water molecules to reduce enzyme deactivation by stripping of water molecules from the enzyme surface. Variants were tested with up to 80% methanol and residual activities were higher than that of the wild type. However, no absolute activities were given, only the residual activity of each variant normalized to 0% methanol in aqueous systems. Purely organic solvent systems have not been used. Other publications also describe inhibiting effects by methanol rather than enzyme destabilization. 3 Recently, the obtained variants with increased stability in hydrophilic solvents were combined with amino acid exchanges for modulating the flexibility within the substrate binding region. 4 The aim of this study was to increase the stability as well as activity of CalB by multiple-site mutagenesis.

Thermostability
Next to engineering for improved solvent stability, thermostability was another engineering goal in order to enable increased reaction temperatures and prolonged reaction times. Several variants, which have been reported in literature have been selected for investigation under the desired reaction conditions including V210I/V221D/A281E, 5 R249L, 6 A162C/K308C, 7 A251E, 8 P218N/L219K/F220T/V221S and T57A/A89T/G226R/R168K. 9 Interestingly the amino acid exchanges T57A and A89T are already naturally occuring in a CalB homologue from ATCC 32657 Pseudozyma aphidis used as parental enzyme for engineering by circular permutation. 10 Docking study of L-proline into the active site of CalB A docking study was performed to visualize the location of the substrate in the active site. For the modeling the structure of CalB PDB-ID: 5A71, chain A in 'open' form was used. Energy minimization was performed with YASARA Structure utilizing AMBER03 Force Field. Based on the modeling of the substrate in the active site, surrounding amino acid residues, including Val190, Ile189, Thr138, Asp134 and Gln157, were selected to evaluate their influence on substrate binding. The docking experiment indicated that there was a lot of free space around the substrate in the active site. This was in agreement with the fact that bulkier substrates, such as Cbz-protected L-proline, were very well converted. One strategy consisted in increasing the size of the amino acid side chains to test whether it was possible to achieve tighter binding of the substrate when the available space was reduced. The opposite hypothesis, to enable alternative binding modi by insertion of smaller amino acid residues, was tested as well.

Water tunnel
Several alternative approaches to increase amidation activity of CalB were pursued. One strategy consisted in closing the water tunnel towards the active site of the enzyme to avoid unwanted hydrolysis. Although this might also decrease the accessibility for NH3, due to the similar size of water and ammonia, it may allow the use of increased water concentrations in the reaction media to improve the substrate solubility without increasing the effect of hydrolysis. In a study by the group of K. Hult, the double variant Q46A/S47L was designed which resulted in strongly reduced hydrolysis as well as increased transacylation activity. 11 All variants were immobilized and stored for roughly the same time (several months) before lyophilization and analysis of amidation of L-proline in 2M2B. In the initial screening of the variants, the beads were applied without lyophilization in 1,4-dioxane. In this screening, CalBopt-24 exceeded all other variants and the wild type significantly. Therefore, this variant was selected for subsequent random mutagenesis experiments.

Combination of amino acid exchanges
Based on the results of the first CalB variants (obtained by rational as well as random engineering), the amino acid exchanges of interesting variants were combined.
For industrial applications, high product concentrations are desired. The substrate loading was increased to 200 mM (only partially dissolved) and the enzyme loading to 4 mg immobilized beads. After 120 h, a product concentration of ~100 mM L-prolinamide was obtained with CalBopt-24 (shown in Figure S1). Under the same conditions, a 20% increase was observed with CalBopt-24 T245S (named CalB-S). The obtained results indicated, that the currently best variant for production of L-prolinamide is CalB-S (CalB T57A/A89T/G226R/R168K/T245S). The positions of the exchanged amino acid residues in the protein structure (shown in Figure S2) showed that the residues are spread all over the enzyme and not focused on the same region, e.g. not located in the active site, which suggests that involvement in substrate binding interactions seems to be less likely than a stabilizing effect on the structure.

Testing alternative hydrolases Commercially available enzymes
Reports in literature suggest that alternative enzymes or enzyme formulations can be applied for amide formation beside immobilized CalB. [27][28][29] Different enzyme preparations were tested with Lproline and Cbz-L-proline as substrate with 2M2B and NH3 at 70 °C. Low levels of L-prolinamide formation were obtained with CalB in a liquid preparation. Only trace amounts were obtained with CalB in lyophilized form (DSM) and lyophilized lipases from Candida rugosa and Mucor miehei, as well as immobilized Lipolase 100T and Lipozyme TL IM (shown in Table S3).  A 'Hydrolase Enzyme Screening Kit (HESK-4800)' from Almac Group (Craigavon, UK) was purchased, which contained 48 hydrolases. The enzymes were provided in lyophilized, liquid or immobilized form. The latter were directly used for the target amidation reaction. The liquid enzyme formulations were added to the beads (ECR8285) and the buffer for immobilization. The lyophilized enzyme preparations were incubated in the immobilization buffer over night under gentle mixing for solvation prior to the immobilization on the epoxy beads. Those enzymes, which were soluble in the immobilization buffer, were immobilized on resin ECR8285 and their capability to catalyze the amidation of Cbz-L-proline or L-proline was investigated. The experiments were run for 62 h in 1,4dioxane at 40 °C with 8.3 mg enzyme (immobilized, without lyophilization). The results showed that substantial amidation of L-proline was only achieved with AH 39 'Lipase from Candida antarctica' (15% conversion) and traces with AH 2 'Lipase B from Alcaligenes sp.' (0.14% conversion).
To reduce the water content of the beads after immobilization of the enzymes from the Almac Hydrolase Kit, the beads were lyophilized for 24 h. No amidation activity was detected in 2M2B at 40 °C, 700 rpm, 16 h with NH3 (0.5 M) and L-proline (33 mM) with the immobilized and lyophilized preparations (2 mg). Furthermore, Lipase Amano PS-IM (immobilized on diatomaceous earth) was tested and the following commercial preparations were immobilized on ECR8806 and tested for amidation: Acylase I from porcine kidney and Acylase I from Aspergillus melleus, Acylase Amano 30000, Penicillinase from Bacillus cereus and Amidase from Rhodococcus erythropolis (2 mg). In none of these cases formation of L-prolinamide was obtained.

Enzyme preparation
The effect of different methods for the stabilization and immobilization of CalB WT in organic solvent were investigated. Highest amounts of product formation were obtained with Novozym 435 and CalB expressed in P. pastoris and immobilized on resin ECR8806 from Purolite (CalB 8806). Both preparations are based on immobilization via adsorption. According to respective product data sheets, Novozym 435 is adsorbed on macroporous Lewatit® VP OC 1600, consisting of poly(methacrylic acid) with divinylbenzene for crosslinking. Numerous applications of this preparation have been reported, often due to its exceptional activity and stability. Nevertheless, recycling of Novozym 435 can be problematic due to enzyme leaching and mechanical instability. 30 According to the supplier, the octadecyl activated methacrylate resin of ECR8806 allows reversible but at the same time very strong adsorption on the hydrophobic material. 31 In comparison to covalent binding adsorption causes less conformational change, which may explain the higher conversions of adsorbed preparations. When CalB is immobilized on resin ECR8285 (CalB 8285), covalent binding between the amine groups of the enzyme and the epoxy groups of the epoxy/butyl methacrylate carrier takes place. Furthermore, cross-linked enzyme aggregates (CLEAs), 32,33 sol-gel entrapped CalB 34 and EziG TM , immobilized on controlled porosity glass, 35 were tested. In Table S4 a comparison with the same amount of enzyme preparation, is shown. Novozym 435 and CalB 8806 led to highest amide formation. According to literature, the enzyme loading on the support of Novozym 435 ranges from 8.5 to 20 w/w%. 30 The varying enzyme loadings make a comparison difficult. When amide formation was normalized to the actual amount of enzyme, highest product concentrations were received with CalB Solgel, Novozym 435 and CalB 8806.

Microscopy pictures after the reaction
The shape of the beads after the reaction was analyzed by microscopy (by Thermo Fisher Scientific, Linz, Austria) to assess whether mechanical damage has occurred. After shaking, the beads of resin ECR8806 were still largely intact ( Figure S4). Figure S4: Microscopy pictures of CalB immobilized on resin ECR8806 after biotransformation of Lproline in 2M2B with NH3. Two uncharacterized fragments can be seen in the middle and the lower part of the picture, which are marked with blue circles. The surface of the beads showed no cracks.
The next picture by transmitted-light microscopy revealed, that the surface was not damaged. Solid matter precipitated on the surface of the beads ( Figure S5). The beads themselves remained intact.

Recyclability of the immobilized enzyme preparations
The enzyme preparations were reused for several amidation reactions of L-proline in 2M2B with NH3 for 20 days and washed with 2M2B in between. Comparing the conversion reached with the reused preparation with a fresh preparation showed, that reused Novozym 435 resulted in half of the conversion of fresh Novozym 435. In contrast, CalBopt-24 showed still higher conversion than fresh Novozym 435 (see Figure S6).

Concept
The rational variants of CalB were generated based on the two-step site-directed mutagenesis protocol 36,37 using Q5® High-Fidelity DNA Polymerase (New England Biolabs GmbH). For expression of CalB a 500 bp fragment of the CAT1 promoter was used. 38 The first step of the workflow was the insertion of the desired mutation(s) into the CalB wild type gene by PCR. After transformation of E. coli Top10 F' cells and sequencing, the DNA was used to transform P. pastoris by integration into the genome. Figure S7: Vector map of pBSY3S1Z_CalB. The vector contains a pUC origin for replication in E. coli, an EM72_syn and ILV5 promoter for expression of a Zeocin resistance marker in E. coli and P. pastoris, respectively, and an AOD terminator. The CAT1 promoter (500 bp) and AOX1_syn terminator were used for expression of CalBopt-24 and the D signal sequence for secretion into the supernatant. Restriction sites (XhoI and NotI) and primers used for amplification and cloning of CalB are shown.

Cloning of CalB mutants in the pBSY3S1Z_CalB expression vector
The vector backbone of pBSY3S1Z_CalB (shown in Figure S7) was cut with XhoI and NotI. This vector included a deletion variant of the S. cerevisiae mating factor alpha signal sequence (D). Two PCRs per mutation were started (PCR1: P16123 CalBopt-(Hind)_Gibsonfor + rev primer for the desired mutation and PCR2: P16124 CalBopt-(Not)_Gibsonrev + fw primer of the desired mutation). The plasmids were assembled via Gibson assembly or oePCR and used to transform E. coli Top10 F' cells. Cells were plated on LB-Zeocin (Life Technologies, Carlsbad, CA, USA; 25 µg/mL) plates, plasmid DNA was isolated and sent to Microsynth for sequencing.
When the correct sequence has been confirmed, plasmids were linearized with SwaI and used to transform P. pastoris BSY BG11 cells. After addition of DNA (1 µg) to the cells (30 µL), incubation on ice >5 min and electroporation with 1.5 kV, sorbitol (500 µL) and YPD (500 µL) were added to the cells. They were regenerated for 2 h at 28 °C and plated on YPD-Zeocin (100 µg/mL) plates.

Directed evolution of CalB Concept
Next to the rational design of CalB variants, a random engineering approach based on directed evolution was performed. CalBopt-24 (T57A/A89T/G225R/R168K), the best variant identified in the initial screening of the rational variants with 1,4-dioxane as solvent (data not shown), was chosen as starting point for mutagenesis by error-prone PCR (epPCR).
An ARS (autonomous replicating sequence) plasmid system was used, which consisted of inserting the gene in a plasmid containing an ARS by homologous recombination in P. pastoris. The advantages of this system compared to integration into the genome are that significantly lower amounts of DNA are required for transformation and that cloning in E. coli and linearization can be omitted. On the other hand, in some cases integration of the plasmid in the genome may occur and distort the results and addition of antibiotics to the growth media is required to avoid plasmid loss. When the landscapes of the different cloning procedures were compared, highest activity and uniformity of the clones was obtained with the pBSYA3Z ARS plasmid assembled by homologous recombination in P. pastoris BSY11dKU70 (obtained from bisy, Austria).

Cloning
The ARS plasmid pBSYA3S1Z (shown in Figure S8) was used. The original stuffer fragment between the CAT1 promoter and AOX1TT_Syn terminator was removed by SapI digestion. For homologous recombination by P. pastoris primers with long overhangs to PCAT1 and TTAOX1_Syn were designed and the plasmid pBSY3S1Z_CalBopt-24 including a D signal sequence was used as template for epPCR. Figure S8: Vector map of pBSYA3S1Z_CalBopt-24. The vector contains a pUC origin for replication in E. coli, an EM72_syn and ILV5 promoter for expression of a Zeocin resistance marker in E. coli and P. pastoris, respectively, an rrnC terminator and a CbAOD1 ARS sequence as terminator and autonomous replicating sequence in P. pastoris. The CAT1 promoter (500 bp) and AOX1_syn terminator were used for expression of CalBopt-24 and the D signal sequence for secretion into the supernatant. Restriction sites (XhoI and NotI) and primers used for homologous recombination of the CalB insert with the plasmid are shown.

Random mutagenesis
Different epPCR conditions (0.1-0.5 mM MnCl2, varied dNTP ratios) were tested to determine a suitable error rate for mutagenesis of CalBopt-24 (T57A/A89T/R168K/G226R) and the secretion signal D. It turned out, that addition of MnCl2 caused too many inactive variants. Therefore, the standard GoTaq® Flexi polymerase protocol (with MgCl2) was used once with equimolar dNTP ratios and once with an imbalanced ratio of dTTP+dCTP to dATP+dGTP of 3 to 1. The epPCR inserts (150 ng) were assembled with the cut backbone of pBSYA3Z (100 ng) by homologous recombination in yeast. BSY11dKU70 cells (obtained from bisy, Austria) were transformed with the DNA fragments (2kV; 500 µL sorbitol and 500 µL YPD-medium were added for 2 h regeneration at 28 °C, cells were plated on YPD Zeocin plates).

Cultivation and screening
After transformation of P. pastoris, individual clones were picked and cultivated in 96-well DWPs in 250 µL YPD-Zeocin (50 µg/mL Zeocin, 1% glucose) for transformants containing ARS plasmids. Positive controls (CalBopt-24) and negative controls (blank, only buffer) were added on each plate. After cultivation for 60 h the cells were stamped on YPD-Zeocin plates to conserve the transformants and 250 µL YPM (10% MeOH) was added for induction. Another induction step was performed after 72 h with 50 µL YPM (10% MeOH). After 84 h total cultivation time, the cells were harvested by centrifugation for 10 min at 3000 xg.
The protein activity in the supernatant was determined with the pNPB assay before and after incubation with 1,4-dioxane as cosolvent. For the screening without co-solvent, substrate solution in Tris/HCl buffer (160 µL) was added to the supernatant (40 µL) and absorbance was followed over 10 min at 405 nm. Furthermore, supernatant (100 µL) was incubated in 1,4-dioxane in Tris/HCl buffer (40 µL and 60 µL, respectively, in total 20% 1,4-dioxane) for 2 h at 60 °C in a polypropylene microtiter plate (MTP) sealed with aluminium foil. After centrifugation of the MTP at 4000 xg for 5 min, 40 µL were added to pNPB substrate solution in Tris/HCl buffer (160 µL) for activity measurement.

Analysis of best variants
For isolation of plasmid DNA from P. pastoris, yeast lysis buffer (1 M sorbitol, 100 mM EDTA, 14 mM DTT in ddH2O) was prepared and the clones with highest activity were cultivated in ONCs (5 mL YPD-Zeocin). The ONCS were harvested by centrifugation (5 min, full speed) and the pellet was washed with ddH2O (1 mL). After another centrifugation step, the pellet was resuspended in yeast lysis buffer (1 mL) and zymolyase stock solution (4000 U/mL, 20 µL) was added. The reaction was incubated for 2 h at 30 °C and the obtained spheroblasts were harvested by centrifugation (5 min, full speed). After removal of the supernatant the pellet was used for standard plasmid DNA isolation with the Miniprep Kit and the DNA was used for transformation of E. coli cells, for preservation and sequencing of the genes. Most genes contained one mutation and sometimes one or two additional silent mutations. After sequencing of clones with beneficial mutations and retransformation of P. pastoris, a rescreening of 20 clones per construct was performed to verify that the variants are indeed more active and contain only one plasmid per cell. The hydrolytic activity before and after incubation in organic solvent was measured with the pNPB assay and normalized to the protein concentration (determined with the Bradford assay). Consequently, the best variants were inserted in the integration vector pBSY3S1Z (XhoI and NotI) and the plasmid was used for transformation of E. coli and P. pastoris BSY BG11 cells (see description above). After cultivation in shaking flasks (450 mL BMG1%) and induction with MeOH, the cell cultures were harvested and the protein supernatant was concentrated and immobilized. The immobilized CalB variants were applied for amidation of the target substrate in organic solvent.

Protein concentration
The total protein concentration in the samples was determined using the Bradford Assay (Bio-Rad; Art: 5000006). The Bradford solution was diluted 1:5 with dH2O. The samples (5 µL) were pipetted in the wells of a transparent 96-well MTP and Bradford solution (200 µL) was added. BSA standards with concentrations of 0.0625 to 2 mg/mL BSA and a blank sample were measured as well. All determinations were made in duplicates. After incubation of at least 5 min the absorbance at 595 nm was measured with a plate reader (with a 5 sec shaking period before measurements).
Buffer exchange PD10 desalting columns from GE Healthcare Life Sciences (Little Chalfont, UK) were used for buffer exchange at small scale according to the recommendations of the supplier. After washing the columns with ddH2O and target buffer, the sample (2.5 mL) was loaded onto the column. Target buffer (3.5 mL) was then applied to elute the protein and the fraction was collected. Depending on the total sample volume, the last two steps were repeated when required. At the end, the column was washed with ddH2O and stored in 20% EtOH. The obtained protein solutions were concentrated by centrifugation using Vivaspin 20, 10,000 MW CO centrifugation tubes from Sartorius at full speed and 8 °C.

S 23
Optical purity: D-/L-prolinamide racemisation For analysis of D-/L-prolinamide racemization, derivatization with ethyl chloroformate and DMAP was performed. This method allowed the differentiation between the two enantiomers after derivatization and GC-measurements. In the biotransformations with L-proline as substrate, Lprolinamide was obtained.  In order to determine the limit of detection of the GC analysis, L-prolinamide solutions were spiked with 1% as well as 0.1% of D-prolinamide and analyzed by GC. In both cases it was possible to distinguish between the peaks of the two enantiomers and to detect the D-enantiomer.