The production of l- and d-phenylalanines using engineered phenylalanine ammonia lyases from Petroselinum crispum

The biocatalytic synthesis of l- and d-phenylalanine analogues of high synthetic value have been developed using as biocatalysts mutant variants of phenylalanine ammonia lyase from Petroselinum crispum (PcPAL), specifically tailored towards mono-substituted phenylalanine and cinnamic acid substrates. The catalytic performance of the engineered PcPAL variants was optimized within the ammonia elimination and ammonia addition reactions, focusing on the effect of substrate concentration, biocatalyst:substrate ratio, reaction buffer and reaction time, on the conversion and enantiomeric excess values. The optimal conditions provided an efficient preparative scale biocatalytic procedure of valuable phenylalanines, such as (S)-m-methoxyphenylalanine (Y = 40%, ee > 99%), (S)-p-bromophenylalanine (Y = 82%, ee > 99%), (S)-m-(trifluoromethyl)phenylalanine (Y = 26%, ee > 99%), (R)-p-methylphenylalanine, (Y = 49%, ee = 95%) and (R)-m-(trifluoromethyl)phenylalanine (Y = 34%, ee = 93%).

as they provided relatively high conversions (61.8% for 2c and 76.9% for 2l) and moderate medium viscosity for the further optimization steps.
The effect of substrate concentration. The solubilities of model substrates 2l and 2c were tested in the reaction buffers providing the highest conversions (6 M NH 4 OH and 2 M ammonium carbamate, Fig. 2a,b). Besides the conversion values (Fig. 2a,b), the solubilities of the model substrates 2l and 2c were also higher in 6 M NH 4 OH in comparison with the 2 M ammonium carbamate solution, providing substrate concentrations below 5 mM.
Therefore, using the optimal reaction medium (6 M NH 4 OH at pH 10), and maintaining the biocatalysts:substrate concentration ratio (OD 600 : mM) at the optimal value of 1, the ammonia additions onto all substrates 2a-l were performed varying the substrate concentration in the limit of their solubilities (Figs. 4a,b and S2-S11).
In most cases the progression of the conversions slowed down with increasing substrate concentration, leading to longer reaction times at high substrate concentrations.
The ortho-substituted substrates 2a,g,j, were transformed with high conversions of 66-91% after 48 h reaction time, even substrate concentrations as high as 30-70 mM (Figs. S2, S7, S9). In contrast, the ammonia addition onto o-OCH 3 -substituted 2d was seriously affected by the increase of substrate concentrations above 10 mM. In this case the stationary conversions decreased from the moderate ~46% to lower values of ~25%, when using substrate concentrations of 2-10 mM and 20-70 mM, respectively (Fig. S5).
Similar tendency was registered for meta-substituted substrates, where concentrations higher than 2 mM substantially decreased the conversion values (Figs. S3, S8, S10 for 2b,h and k), except for m-methoxycinnamic acid 2e (Fig. 4a), where conversion values maintained at ~71% even at high 50 mM substrate concentration.
In case of para-substituted substrates, substrate concentrations not affecting the final conversions were higher for substrates with electron donating substituents (30 mM for 2c and 2f -Figs. S4, S6) than for substrates with electron withdrawing substituents (10 mM and 2 mM for 2i and 2l, respectively -Figs. 4b, S11).
In case of m-CH 3 -cinnamic acid 2b, o-OCH 3 -cinnamic acid 2d, p-CF 3 -cinnamic acid 2l the ammonia additions stopped at low conversions (15-42%) even in case of low substrate concentrations (2-10 mM, Figs. S3, S5,  In all points of the reactions the ee values were >99%. S11), suggesting the occurrence of substrate inhibition. This was further supported by the inhibitory effect of increasing substrate concentration on the reaction velocities of the ammonia addition reaction of 2l, determined through the UV-based PAL-activity assay, using purified I460V PcPAL as biocatalyst (Fig. S22).
Using the optimal conditions specific for each substrate, the ammonia additions onto all substrates 2a-l were performed, with the aim to determine the final, stationary conversions and ee values. Most of the reactions proceeded with high or moderate conversions and provided the l-Phe derivatives l-1a-l in excellent ees ( Table 1). As exception, the ammonia addition reactions of 2b and 2f stopped at low stationary conversions of 26.4% and 19.1%, respectively (Table 1), which were maintained even after adding fresh batch of PcPAL-whole cells (preceded by the removal of the initial whole cell-batch by centrifugation). The significantly decreased conversions and reaction velocities of the ammonia addition onto 2b and 2f, performed in presence of increasing concentrations of 1b and 1f, respectively, supports the occurrence of product inhibition in these cases (Figs. S24-S26).
In terms of comparison with similar PAL mediated procedures, the optimized biotransformations provide superior conversions and enantiomeric excess values for the ammonia additions onto o-, m-, p-OCH 3 -cinnamic acids to those reported for the wild-type PcPAL 8 or PAL from Anabaena variabilis (AvPAL) 10,40 , and approximates, in case of o-OCH 3 -cinnamic acid even surpasses, the conversions obtained with the improved variants of PAL from Planctomyces brasiliensis (PbPAL), reported for its remarkable high activity towards cinnamic acids bearing electron donor ring substituents 25,40 . The high enantiomeric excess values of the trifluoromethyl-substituted l-phenylalanines l-1j-k obtained with moderate to high conversions (Table 1) exceed the ee values previously obtained through wild-type AvPAL, PcPAL or PAL from Rhodotorula glutinis (RgPAL) 10,39 . In case of p-Br-substituted substrate 2i, the obtained results are comparable with the procedures based on improved    24 . While AvPAL-mediated ammonia additions onto methyl-substituted cinnamic acids were also reported 25 , the optimized PcPAL procedures provide superior conversions (in case of o-and p-methylcinnamic acids) and enantiomeric excess values (for o-and m-methylphenylalanine) ( Table 1).

Ammonia elimination optimizations.
For the optimization process of the kinetic resolution-type ammonia eliminations from racemic phenylalanine analogues, rac-1c and rac-1k were chosen as model substrates. Besides aiming to achieve conversion approximating the optimal 50% of a highly selective kinetic resolution, we monitored the enantioselectivity of the resolution process, by the correlation of the experimentally determined ee values of the unreacted d-enantiomer with the theoretical ee values, calculated from the obtained conversions, considering a fully enantioselective resolution process. Since during a kinetic resolution of high enantioselectivity (with E»200), only the ammonia elimination from the l-Phe derivative occurs, the ee of the unreacted d-enantiomer increases upon the progress of the reaction and reaches the optimal value of 100% at conversion values approximating 50%.
In case of rac-m-(trifluoromethyl)phenylalanine (rac-1k), the conversions reached 50% in all cases after 30 h reaction time with the exception of ammonium acetate buffer, where the maximum conversion was not reached even after 48 h (43.1%). The highest conversions and ee (>99%) was obtained using Tris, borax or phosphate buffer (Fig. 5a).
Probably, the reason for the incomplete reactions obtained using ammonium acetate or 0.1 M NH 3 buffers (selected for their advantage of easy removal during the reaction work-up), is the occurrence of the reverse ammonia addition reaction, under the low ammonia content. While borax and Tris-buffer systems provide complete conversions for both model substrates, the slight differences in the optimal reaction medium suggest that for rigorous fine-tuning of the efficiency of resolution process, reaction medium tests should be performed for each individual substrate of interest.
The effect of biocatalysts: substrate ratio and substrate concentration. Further, the impact of the increased biocatalyst and substrate concentrations on the reactions of model substrates were tested. Firstly, whole-cell PcPAL biocatalysts with various cell densities (OD 600 of 1, 2, and 4) were used at fixed 2 mM substrate concentration. Generally, the increase of biocatalyst:substrate ratio (2:1 ratio of cell density (OD 600 ): mM) resulted shorter reaction times. Thus, when using cell densities of OD 600 4, the optimal conversion of ~50% was reached in 5 h in case of rac-1c and 24 h in case of rac-1k (Table 2), while lower cell densities of OD 600 1 or 2 afforded similar conversions of 50% in longer reaction times (30 h in case of rac-1k and >48 h in case of rac-1c).
The influence of substrate concentration on the resolution process of model substrates was also tested. When increasing the substrate concentration, using the optimal biocatalyst:substrate ratio of 2:1 (OD 600 : mM), the corresponding increase of cell densities was also necessary, leading to the appearance of viscosity issues, hindering reaction monitoring. Therefore, evaluation of the effect of substrate concentrations, ranging between 2 mM and the maximum solubility of model substrates rac-1k and rac-1c (70 mM and 15 mM, respectively), was performed with biocatalyst:substrate ratio of 1:2 (OD 600 : mM).
In case of ammonia eliminations from rac-1c, the conversion reached the optimal 50% with similar time progressions at each substrate concentrations tested (Fig. 6a). However, ee D-1c slightly decreased with the increase of substrate concentration, 95% was obtained at 2 and 5 mM concentration of rac-1c and conversions approximating the 50% optimal value, while at higher 15 mM substrate concentration ee D-1c was 89% (Fig. 6a). Notable, that the obtained ee values are slightly lower than the theoretical ees (Figs. S14-S16), supporting a not fully enantioselective resolution process, of which selectivity decreases by the increase of substrate concentration.
In the ammonia elimination of rac-1k, using high substrate concentrations, low (c 30mM = 6%) or no conversions (at substrate concentration >30 mM) were obtained (Fig. 6b). The decrease of stationary conversion values upon increasing the concentration of rac-m-(trifluoromethyl)phenylalanine rac-1k (Fig. 6b), similarly to the ammonia addition onto 2l (Fig. S11), suggested the occurrence of substrate inhibition. This was confirmed by the kinetic measurements performed using purified PcPAL I460V (Fig. S23). Besides the significant decrease of conversions, at substrate concentrations exceeding 10 mM of rac-1k, the difference of the ee values registered for the unreacted d-1k and the theoretical ee values also increased, supporting the decrease of the enantioselectivity upon the increase of substrate concentration, similarly as in the case of model substrate rac-1c (Fig. 7).
Therefore, due to the influence of substrate concentration on the enantioselectivity of the resolution process, and the occurrence of substrate inhibition, the selection of the optimal substrate concentration based primarily on the final ee d-1 values, followed by the conversion values. Performing the ammonia eliminations of rac-1c and rac-1k under the optimal, relatively low, substrate concentrations (Table 3), high conversions, reaching the theoretical 50% of a kinetic resolution process, and high ee values (95% and 93%, respectively) were obtained.
Using the optimal reaction conditions set-up by analytical scale biotransformations, the preparative scale biotransformation provided similarly high conversion and ee values, affording the corresponding d-and l-amino acids in good conversions, isolated yields and optical purities ( Table 4). The efficiency of ammonia eliminations, in terms of yields and enantiomeric excess values, are comparable with the kinetic resolutions performed with purified/commercial enzymes, such as the penicillin G acylase mediated enantioselective acylation for d-p-methylphenylalanine d-1c 44 , or the l-AAO/catalase catalyzed kinetic resolution process yielding d-m-(trifluoromethyl)phenylalanine d-1k 45 , a key chiral intermediate for (R)-PFI-2 33 . Definitely, in terms of conversions the kinetic resolution-type ammonia elimination of rac-1c is surpassed by the asymmetric synthetic routes, such as the reductive amination of the corresponding keto-acid catalyzed by DAADH (d-amino acid dehydrogenase) 46 , necessitating co-factor (NADPH) regeneration system or the l-amino acid deaminase and engineered aminotransferase coupled cascade, that uses l-or racemic amino acid 1c as starting material 47 . In this context the asymmetric synthesis-type ammonia additions, starting from inexpensive cinnamic acids, represent a more attractive synthetic route. Accordingly, in case of l-m-(trifluoromethyl)phenylalanine l-1k, intermediate for kinesin KIFC1 inhibitors 32 , the developed PAL-ammonia addition provides higher yield than the claimed enantioselective hydrolysis of corresponding carboxylic esters 48 . The yields for l-m-methoxyphenylalanine l-1e are comparable with the yields of the recently reported, PbPAL mediated semi-preparative scale reaction (61% yield, ee > 99%) 40 , but also with the aspartate aminotransferase catalyzed reductive amination 49 . The optically pure l-p-bromophenylalanine l-1i, as versatile chiral intermediate 12 was also obtained in high, 82% yield (Table 4), similarly as in case of using improved AvPAL variants 12 , but somewhat lower than the 94% reported yield when using the 31E variant of Rhodotorula graminis PAL (RgrPAL) 24 .

Methods
Preparation of whole cell biocatalysts for enzymatic reactions -general procedure. The overnight preculture was prepared in Erlenmeyer flasks containing LB (Luria Bertani) medium supplemented with carbenicillin (50 µg/mL) and chloramphenicol (30 µg/mL) and inoculated with glycerol stocks of E. coli Rosetta (DE3) pLysS cells harbouring the pET19b vector carrying the wt-or mutant pcpal gene 50 , followed by overnight incubation at 37 °C and shaking at 200 rpm. The obtained preculture (5 mL) was further used to inoculate shake flasks (2l) containing 500 mL LB. Cultures were grown at 37 °C, 200 rpm until OD 600 reached 0.6-0.8, at which point protein production was induced via the addition of 0.1 mM IPTG (final concentration), and the cell growth was maintained at 25 °C for another 16 h. Cell densities of OD 600 were measured after for each mutant variant and wild-type PcPAL. The cells were harvested by centrifugation at 4000 rpm (1751 × g) and 4 °C for 20 min and washed with PBS buffer (20 mM phosphate, 150 mM NaCl, pH 8.0) (4000 rpm, 1751 × g, 4 °C, 20 min) and stored at -20 °C in 1.5 ml polypropylene tubes with OD 600 of ~2 until further use. www.nature.com/scientificreports www.nature.com/scientificreports/ Biotransformation screenings for ammonia addition reactions -general procedure. The ammonia addition reactions were performed in duplicate using 1.5 ml polypropylene tubes, containing the whole cell PcPAL-biocatalysts, prepared as described above. The bacterial pellet was resuspended to an OD 600 of ~2, in 1 mL of ammonia source (2, 4, 6 M NH 4 OH pH 10 adjusted with CO 2 ) and ammonium carbamate (2, 4, 6 M NH 4 [H 2 NCO 2 ], pH 9.6-10 without adjustment). For the biotransformations, 2 mM substrate (cinnamic acids 2a-l) concentrations were used, and the reaction mixtures were incubated at 30 °C, 250 rpm for specified reaction times. Conversions were monitored using reversed-phase high performance liquid chromatography (HPLC). Reaction samples were quenched by adding an equal volume of MeOH, vortexed and centrifuged (13400 rpm, 12000 × g, 10 min). The supernatant was filtered through a 0.22 μm nylon membrane filter and analyzed by HPLC.  www.nature.com/scientificreports www.nature.com/scientificreports/ Biotransformation screenings for ammonia elimination reactions-general procedure. The ammonia elimination reactions were performed in duplicate using 1.5 ml polypropylene tubes, containing the whole cell PcPAL -biocatalysts, prepared as described above. The bacterial pellet was resuspended to an OD 600 of ~2, in 1 mL of different buffers: Tris (20 mM Tris.HCl, 120 mM NaCl, pH 8.8,), NH 3 -buffer (0.1 M NH 4 OH, pH 9.5, adjusted with CO 2 ), borax (0.1 M Na 2 [B 4 O 5 (OH) 4 ], pH 9.5), ammonium acetate buffer (0.1 M, pH 9.5,) sodium carbonate (0.1 M, pH 9.0) and phosphate-buffer (0.1 M phosphate, pH 8.8). For the biotransformations 2 mM substrate (racemic amino acids rac-1a-l) concentrations were used, and the reaction mixtures were incubated at 30 °C, 250 rpm for different reaction times. Conversions were monitored using reversed-phase HPLC. Reaction samples were quenched by adding an equal volume of MeOH, vortexed and centrifuged (13400 rpm, 12000 × g, 10 min). The supernatant was filtered through a 0.22 μm nylon membrane filter and analyzed by HPLC.
Preparative scale ammonia additions. In a 500 mL flask, E. coli Rosetta (DE3) pLysS cells harbouring the pET19b vector carrying the corresponding mutant pcpal gene were resuspended in 6 M NH 4 OH-solution (pH 9.8 adjusted with CO 2 ) to give a final OD 600 of ~10 for 2i and 2k (13.26 g wet cells in 221 mL reaction volume and 13.92 g wet cells in 232 mL reaction volume, respectively) and a final OD 600 of ~30 (16.92 g wet cells in 94 mL reaction volume) for 2e. 0.5 g cinnamic acid 2e, 2i, 2k (2.8, 2.2, 2.3 mmol, respectively) was added to the cell suspension in a final concentration of 30 mM for 2e and 10 mM for 2i, 2k and the reaction was incubated at 200 rpm, 30 °C for 48 hours, monitoring the conversion values by reversed-phase HPLC. When stationary conversions were reached (Figs. 4a,b and S9) the reaction mixture was acidified to pH 1.5 by dropwise addition of aqueous H 2 SO 4 (50% w/v). The formed precipitate was removed by centrifugation at 10000 rpm (10947 × g), 4 °C for 20 min., while the non-reacted cinnamic acid was removed by extraction with ethyl acetate (3 × 10 mL). The amino acid found within the aqueous phase was purified by ion exchange chromatography using Dowex ® 50WX2 resin, using 2 M NH 4 OH solution for elution. The fraction containing the final product was evaporated in a centrifugal evaporator affording the pure l-amino acid (0.29 g, 1.63 mmol, 59% isolation yield for (S)-1e; 0.4 g, 1.76 mmol, 80% isolation yield for (S)-1i and 0.26 g, 1.2 mmol, 52% isolation yield for (S)-1k).
Preparative scale ammonia eliminations. In a 500 mL flask, E. coli Rosetta (DE3) pLysS cells harbouring the pET19b vector carrying the corresponding or mutant pcpal gene were resuspended in Borax buffer (pH 10 unadjusted) to give a final OD 600 of ~7.5 (12.51 g wet cell mass /278 mL for rac-1c and 6.43 g wet cell mass /143 mL rac-1k). Racemic phenylalanine analogue rac-1c or rac-1k (0.25 g, 1.4 mmol, or 0.5 g, 2.1 mmol, respectively) was added to the cell suspension in a final concentration of 5 or 15 mM, respectively and the reactions were incubated at 200 rpm, 30 °C for 48 hours, monitoring the conversion values by reversed-phase HPLC (Figs. S11, S12). When stationary conversions were reached, the reaction mixtures were purified using the method described above for the preparative scale ammonia addition, providing the d-amino acids (0.245 g, 1.34 mmol 49% isolation yield for (R)-1c and 0.19 g, 0.82 mmol and 39% isolation yield for (R)-1k).

conclusions
Within this study, we developed highly efficient biocatalytic procedures for the synthesis of valuable enantiopure l-and d-Phe, using as whole-cell biocatalyst recombinant E. coli cells harbouring the plasmid of PcPAL mutants (L256V, L134A, I460V), specifically tailored towards the targeted ring-substituted cinnamic acid and phenylalanine substrates.
In case of ammonia addition reactions, regarding to substrate solubility and reaction time, the optimal reaction medium was found to be 6 M NH 4 OH, while in several cases the substrate concentration significantly affected the final conversions. Substrates bearing ortho-substituents or para-electron donating substituents were transformed with high conversions (66-91%) even at high substrate concentrations (30-70 mM), while in case of substrates with meta-or electron withdrawing para-substituents high substrate concentrations (>2-10 mM) significantly lowered the final conversion values, substrate inhibition occurring in several cases.
In case of ammonia elimination reactions, the reaction medium affected slightly, but differently the kinetic resolution of the model substrates, suggesting specific conditions for each substrate of interest. Using an optimal biocatalyst:substrate ratio, inhibitory effect occurred at high substrate concentration (>2-5 mM), while the enantioselectivity of the resolution process also decreased, forcing the use of relatively low substrate concentrations.