Dynamic kinetic resolution of phenylglycine esters via lipase-catalysed ammonolysis

Article abstract of DOI:10.1016/S0957-4166(99)00143-3

Ammonolysis of D,L-phenylglycine methyl ester catalysed by Novozym 435 at 40°C in tert-butyl alcohol gave D-phenylglycine amide in 78% ee at 46% conversion, corresponding to an enantiomeric ratio (E) of 16. Lowering the temperature improved the enantioselectivity (E = 52 at -20°C). Combination of ammonolysis with pyridoxal-catalysed in situ racemisation of the unconverted ester (dynamic kinetic resolution), at -20°C, gave D- phenylglycine amide with 88% ee at 85% conversion. The amide racemised much slower than the ester at this low temperature.

Full text of DOI:10.1016/S0957-4166(99)00143-3

Pergamon
Tetrahedron: Asymmetry 10 (1999) 1739–1750
Dynamic kinetic resolution of phenylglycine esters via
lipase-catalysed ammonolysis
M. A. Wegman, M. A. P. J. Hacking, J. Rops, P. Pereira, F. van Rantwijk and R. A. Sheldon∗
Laboratory of Organic Chemistry and Catalysis, Delft University of Technology, Julianalaan 136, 2628 BL Delft,
The Netherlands
Received 7 April 1999; accepted 15 April 1999
Abstract
Ammonolysis of D,L-phenylglycine methyl ester catalysed by Novozym 435 at 40°C in tert-butyl alcohol
gave D-phenylglycine amide in 78% ee at 46% conversion, corresponding to an enantiomeric ratio (E) of 16.
Lowering the temperature improved the enantioselectivity (E=52 at −20°C). Combination of ammonolysis with
pyridoxal-catalysed in situ racemisation of the unconverted ester (dynamic kinetic resolution), at −20°C, gave D-
phenylglycine amide with 88% ee at 85% conversion. The amide racemised much slower than the ester at this low
temperature. © 1999 Elsevier Science Ltd. All rights reserved.
1. Introduction
As a result of recent developments in the manufacture of penicillin and cephalosporin antibiotics,1
multi-step chemical procedures for the coupling of the D-phenylglycine and D-(4-hydroxyphenyl)glycine
side-chains with the β-lactam nuclei are being replaced by enzymatic alternatives (Scheme 1).
In the chemical procedure, D-phenylglycine and its 4-hydroxy derivative are key intermediates.
Two processes are used for the commercial production of these amino acids: classical resolution via
diastereomeric salt crystallisation2 or enantioselective hydrolysis of the corresponding hydantoin.3,4 In
the former, the L-isomer has to be racemised while the latter process is a dynamic kinetic resolution and
produces the D-isomer as the sole product. The enzymatic coupling (Scheme 1) is successful only with an
activated side-chain, e.g. an amide or an ester.5 Hence, when the side-chain is produced as the free acid
(see above) additional steps are required to convert it to the corresponding ester or amide. Consequently,
an efficient enzymatic coupling process requires a direct method for the synthesis of the D-ester or D-
amide.
∗
Corresponding author. E-mail: r.a.sheldon@stm.tudelft.nl
0957-4166/99/$ - see front matter © 1999 Elsevier Science Ltd. All rights reserved.
PII: S0957-4166(99)00143-3

1740
M. A. Wegman et al. / Tetrahedron: Asymmetry 10 (1999) 1739–1750
Scheme 1. Enzyme-catalysed coupling of D-phenylglycine amide and 6-aminopenicillic acid
An enzymatic route to D-phenylglycine amide has been developed and commercialised by DSM.6
This process involves a kinetic resolution of the racemic amide by an L-specific aminopeptidase and,
consequently, suffers from the inherent disadvantage of producing L-phenylglycine as the coproduct. The
latter needs to be racemised and converted, via the ester, to the D,L-amide substrate. A direct method for
the enantioselective conversion of racemic ester to D-amide would have obvious advantages, especially
if it could be performed with in situ racemisation, i.e. as a dynamic kinetic resolution.
We recently showed7 that the lipase (Novozym 435) catalysed ammonolysis of racemic phenylglycine
methyl ester 1a affords D-phenylglycine amide 2 (Scheme 2). This method would meet the criteria
outlined above if the slow racemisation of the ester, which we observed7 (5% over 24 h) in the course of
the reaction, could be sufficiently accelerated to allow for a dynamic kinetic resolution. Obviously, for
such a scheme to be effective the product amide should racemise much slower than the ester substrate.
Scheme 2. Enantioselective ammonolysis of phenylglycine esters
The racemisation of amino acid esters is known to be effficiently catalysed by aromatic aldehydes such
as 2-hydroxybenzaldehyde (salicylaldehyde) and 3-hydroxy-2-methyl-5-hydroxymethyl-pyridine-4-
carboxaldehyde (pyridoxal) under basic conditions.8,9 The use of aldehyde-based racemisation catalysts
under ammonolytic conditions would seem questionable, however, because reaction of the aldehyde
group with ammonia8,9 might be expected to interfere. The problem of undesired racemisation of
D-phenylglycine amide 2 would seem more manageable; as the amide 2 is racemised much slower and
at high concentrations, it precipitates and is thus effectively removed from the reaction mixture.
Indeed we found previously10 that pyridoxal and salicylaldehyde readily racemise phenylglycine

M. A. Wegman et al. / Tetrahedron: Asymmetry 10 (1999) 1739–1750
1741
methyl ester 1a in ammonia-saturated tert-butyl alcohol, thus providing the basis for a dynamic kinetic
resolution process, which we have now investigated in more detail.
2. Results and discussion
2.1. Ammonolysis of phenylglycine esters
Because Candida antarctica lipase B (Novozym 435) emerged from our initial study7 as the catalyst
of choice, it was used throughout the present work. Initially we investigated the influence of the chain
length of the ester group on the course of the reaction. The data in Table 1 show that the butyl ester
1c reacted at one third of the rate of the methyl ester 1a. The effect on the enantioselectivity ratio E,11
which varied between 16 and 19 was negligible.†,‡ In view of these results all further experiments were
performed with 1a.
Table 1
Effect of the chain length of the ester in the ammonolysis of phenylglycine esters 1a
The concentration of D,L-methyl ester 1a was varied between 50 and 800 mM. The initial reaction
rate of L-1a was linearly proportional to the concentration over the whole range measured, whereas
D-1a followed Michaelis–Menten kinetics (Fig. 1). The apparent Michaelis constant Km, which is
approximately 830 mM for D-1a, was deduced by fitting the experimental data to the Michaelis–Menten
rate equation.
Due to the deviation of D-1a from first order kinetics, an increased starting concentration of D,L-
1a resulted in longer reaction times and a slight decrease in E. The concentration of ammonia had no
influence on either E or the reaction rate between 1.7 and 2.5 M (data not shown).
†
The enantiomeric ratios for the ester (Ee) and the amide (Ea) were calculated from the conversion (c) and the enantiomeric
excesses of the ester (eee) and the amide (eea) as follows:
For the ester:
ln[(1 − c)(1 − eee)]
Ee =
ln[(1 − c)(1 + eee)]
For the amide:
ln[1 − c(1 + eea)]
Ea =
.
ln[1 − c(1 − eea)]
Experimental values for E were lower than those found previously7 which can be attributed to improvements in the analytical
‡
procedure.

1742
M. A. Wegman et al. / Tetrahedron: Asymmetry 10 (1999) 1739–1750
Figure 1. Initial reaction rate as function of the concentration of D-1a (ꢀ) and L-1a (ꢁ)
The E-value of 16 that we consistently observed in the ammonolysis of 1a in tert-butyl alcohol is rather
low for an efficient kinetic resolution. We note, however, that in an ideal dynamic kinetic resolution
this would result in product ee of 88%.§ Nevertheless, there is an obvious incentive to improve the
enantiomeric ratio.
Since it is known that the reaction medium can dramatically influence enantiopreferences of
enzymes,12–14 the effect of solvent on the ammonolysis of 1a was studied. However, little variation in
the E-value was observed (Table 2). The reaction rate, however, is strongly dependent on the reaction
medium, but there was no apparent correlation with any solvent parameter such as hydrophobicity (log
P), dielectric constant or dipole moment.
Table 2
Ammonolysis of 1a in various solventsa
§
In an ideal dynamic kinetic resolution the racemisation of the reactant is fast compared with its transformation into product,
whereas the product racemisation is negligible.

M. A. Wegman et al. / Tetrahedron: Asymmetry 10 (1999) 1739–1750
1743
The reaction temperature is also known to affect E.15,16 Since the reaction rate decreases with
temperature, only solvents with a sufficiently high initial rate at 40°C were used in the low-temperature
reactions (e.g. tert-butyl alcohol, tert-amyl alcohol and tert-butyl methyl ether). From the solvents tested,
the upward trend in E was most pronounced in tert-butyl alcohol (Table 3), resulting in an E-value up to
52 at −20°C. Measurements at even lower temperatures were impractical owing to the low reaction rates.
Table 3
Ammonolysis of 1a at low temperaturea
2.2. Racemisation via Schiff base intermediates
D-Methyl ester 1a and D-amide 2 were subjected to pyridoxal or salicylaldehyde catalysed racemisa-
tion under ammonolysis conditions (tert-butyl alcohol, 2.5 M NH3 at 40°C). In all cases the ee decreased
to 0% according to first order kinetics. The initial racemisation rate obeyed Michaelis–Menten kinetics
according to:
krac · [D-PGX]
v =
[rac. cat.]
(1)
Kdis + [D-PGX]
It has generally been assumed that racemisation takes place via Schiff base formation.8,9,17 We
found, however, saturation kinetics, which imply a fast pre-equilibrium with dissociation constant Kdis
.
We propose that racemisation takes place via a hemiaminal which undergoes slow dehydration with
catalytic rate constant kcat followed by fast racemisation of the Schiff base (Scheme 3). The kinetic
parameters were determined from Lineweaver–Burk plots (Table 4). It should be noted that the measured

1744
M. A. Wegman et al. / Tetrahedron: Asymmetry 10 (1999) 1739–1750
racemisation rate constant is half the kinetic rate constant kcat. (The hemiaminal or Schiff base can revert
equally well to either D- or L-isomer.)
Scheme 3. Racemisation via Schiff base intermediates
The catalytic rate constants show that both catalysts racemise 1a 30 times as fast as 2 under Vmax
conditions. At lower concentrations this rate advantage of 1a will be less, as can be deduced from
the kcat/Kdis values. Moreover, due to its lower Kdis value, 2 would be expected to compete effectively
for the racemisation catalyst, whose effect would be less pronounced for salicylaldehyde (Table 4). We
tentatively concluded that both racemisation catalysts appear to be suitable for in situ racemisation.
2.3. Ammonolysis with in situ racemisation
We next combined ammonolysis and racemisation of 1a in tert-butyl alcohol (2.5 M NH3 at 40°C).
From the time–course and ee of reactant and product (Fig. 2) we conclude that L-1a is efficiently
racemised by pyridoxal. Its ee at 50% conversion is just over 40%, compared with 70% in the absence of
racemisation catalyst. The racemisation catalyst also accelerated the reaction from a 50% conversion in
4 h to 60%, because it counteracts the depletion of fast-reacting D-1a. On the other hand, the downward

M. A. Wegman et al. / Tetrahedron: Asymmetry 10 (1999) 1739–1750
1745
Table 4
Kinetic data of the racemisation of 1a and 2
Figure 2. Composition and ees of ester 1a and amide 2 as function of time: (ꢂ) ee of ester 1a, (ꢃ) amount of ester 1a, (◦) ee of
amide 2, (ꢀ) amount of amide 2. Reaction conditions: D,L-phenylglycine methyl ester 1a (200 mM) was shaken with Novozym
435 (50 mg) and pyridoxal (2 mM) in ammonia-saturated tert-butyl alcohol (5 ml, 12.5 mmol NH3) at 40°C
trend of the ee of 2 in the course of the reaction shows that pyridoxal-catalysed racemisation of 2 is quite
significant.
In order to establish the optimum conditions for a dynamic kinetic resolution, we varied the concentra-
tion of pyridoxal between 40 µM and 2 mM. Similar experiments were performed with salicylaldehyde
(1 to 4 mM) as racemisation catalyst. From the results (Table 5) no clear optimum with regard to
concentration and ee is apparent. Racemisation catalyst concentrations exceeding 1 mM mainly affect
the conversion rate. Pyridoxal and salicylaldehyde at 1 and 2 mM perform similarly with regard to ee,
but in the case of pyridoxal a somewhat higher conversion is reached.
In order to gain more insight into the effect of the racemisation catalysts, reactions were followed
over time and the ee values were plotted against the conversion (Figs. 3 and 4). Up to 50% conversion,
the results appeared to be independent of the concentration of the racemisation catalyst. Even when
no racemisation catalyst was added, the same result was obtained. For conversions higher than 50%
we observed that at low pyridoxal concentrations (40 to 200 µM, see Fig. 3) the racemisation of 1a
was too slow to be effective. When the pyridoxal concentration was increased to 400 µM the results
improved slightly, but a further increase had no effect. For salicylaldehyde a very similar pattern was
observed (Fig. 4). It would seem that, once the racemisation catalyst concentration exceeds a critical
value, higher concentrations affect 1a and 2 equally. Consequently, the reaction proceeds faster due

1746
M. A. Wegman et al. / Tetrahedron: Asymmetry 10 (1999) 1739–1750
Table 5
Ammonolysis with in situ racemisationa
Figure 3. Enantiomeric excess of amide 2 as a function of the conversion at different pyridoxal concentrations (mM): (∆) 2, (◦)
1, (ꢂ) 0.4, (ꢀ) 0.2, (ꢃ) 0.04
to faster racemisation of L-1a, but the racemisation of D-2 is also accelerated and hence, no optimum
becomes apparent.
Because 2 would be expected to precipitate and hence to be effectively withdrawn from the reaction
mixture, we would expect that a high concentration of 1a would accelerate its racemisation relative
to 2. However, when we increased the concentration of 1a and pyridoxal fourfold (to 800 and 4 mM,
respectively) no effect on ee vs concentration became apparent (Table 5). This is probably due to the

M. A. Wegman et al. / Tetrahedron: Asymmetry 10 (1999) 1739–1750
1747
Figure 4. Enantiomeric excess of amide 2 as a function of the conversion at different salicylaldehyde concentrations (mM): (◦)
4, (ꢂ) 2, (∆) 1
Figure 5. Enantiomeric excess of amide 2 as a function of the conversion in different organic solvents: (ꢄ) tert-amyl alcohol,
(ꢀ) tert-butyl alcohol, (ꢀ) dioxane, (◦) 1,2-dimethoxy ethane, (∆) acetonitrile, (ꢂ) tert-butyl methyl ether
lower E-value at high concentrations of 1a (see Fig. 1). Rapid exchange of precipitated D-2 with the
product in solution may also contribute to this unsatisfactory result.
To improve the results of the dynamic kinetic resolution, the racemisation rate of 1a should be increas-
ed with respect to 2. We assumed that the racemisation rate could be influenced by the reaction medium.
Hence, we studied ammonolysis with in situ racemisation in various solvents at 40°C. Experiments
were carried out with pyridoxal (1 mM) as racemisation catalyst. As can be seen from Fig. 5, the
substrate is most efficiently racemised in tert-butyl alcohol and tert-amyl alcohol. In all other — non-
protic — solvents, the racemisation of the substrate was so slow that L-1a accumulated, which caused
low yields with low ee.
Encouraged by the good results at low temperature we next combined the ammonolysis of 1a with in
situ racemisation at −20°C. tert-Butyl alcohol:tert-butyl methyl ether (70:30% v/v) was used as solvent
and pyridoxal (4 mM) as racemisation catalyst. The reaction was followed over time and after 66 h
we obtained 85% yield with 88% ee. This result shows that at low temperature the substrate is indeed
racemised much faster than the product and a dynamic kinetic resolution is feasible whereby the product
is obtained in good yield and high ee.

1748
M. A. Wegman et al. / Tetrahedron: Asymmetry 10 (1999) 1739–1750
3. Conclusion
Pyridoxal or salicylaldehyde catalysed racemisation of phenylglycine methyl ester 1a is compatible
with lipase-catalysed enantioselective ammonolysis. The catalyst became more enantioselective at low
temperature. At −20°C D-phenylglycine amide 2 was obtained with 88% ee at 85% conversion.
4. Experimental
4.1. General
The reaction mixtures of phenylglycine methyl ester were analysed by chiral HPLC on a Daicel
Chemical Industries Ltd, 4.6×150 mm 5 µ Crownpak CR (+) column using a Waters 625 LC pump
and a Waters 486 UV detector. The eluent was aqueous HClO4, pH 1.5 at a flow of 0.6 ml/min, the
column temperature was 18°C. The reaction mixtures of the ethyl and butyl esters of phenylglycine
were analysed by chiral HPLC on a 4.6×250 mm 10 µ Chiracel OD column with a Waters 510 pump,
and a Shimadzu SPD-6a UV detector. Hexane:iso-propyl alcohol:dimethyl amine (90:10:0.1, v/v) at 0.5
ml/min was used as eluent. 1H and 13C NMR spectra were recorded using a 400 MHz Varian-VXR 400S
spectrometer.
Immobilised Candida antarctica lipase B, Novozym 435, was a gift from Novo Nordisk A/S (Bags-
værd, Denmark). Immobilised Candida Antarctica lipase B on Accurel EP 100, SP 611 (ex Novo Nordisk
A/S) was a gift from Uniqema (Gouda, The Netherlands). Salicylaldehyde was purchased from Janssen
Chimica, pyridoxal hydrochloride from Aldrich. Racemic phenylglycine was obtained from Acros.
Enantiomerically pure phenylglycine, phenylglycine methyl ester and phenylglycine amide were kindly
donated by DSM (Geleen, The Netherlands). Solvents were dried on Zeolite CaA (Uetikon, activated at
400°C for 24 h before use). The methyl, ethyl, and butyl esters of phenylglycine were synthesised from
the amino acid according to the literature.18 Phenylglycine methyl ester was used as free base, obtained
by bulb to bulb distillation of a mixture of phenylglycine methyl ester hydrochloric acid salt and sodium
methoxide (30% solution in methanol). The free base distilled at 100°C (1 mbar) as a colourless liquid,
which solidified by standing at room temperature.
4.2. Ammonolysis of phenylglycine esters
A mixture of 1.0 mmol phenylglycine ester (165 mg methyl ester 1a, 216 mg ethyl ester hydrochloric
salt 1b, or 232 mg butyl ester hydrochloric salt 1c) and Candida antarctica lipase B in 5.0 ml ammonia
saturated solvent was shaken in a 40 ml reaction vessel at the selected temperature (see Tables 1–3 for
further details). Reactions were monitored over time. The partial pressure of ammonia was maintained
at 1 atmosphere in all reactions, which guarantees a constant thermodynamic activity of ammonia. The
reactions were stopped by adding concentrated formic acid and the reaction mixture was dissolved by
adding water. When tert-amyl alcohol or tert-butyl methyl ether was used as the solvent methanol was
also added; when hexane was used as solvent, ethanol was added to dissolve the reaction mixture. A
sample was taken for HPLC analysis.

M. A. Wegman et al. / Tetrahedron: Asymmetry 10 (1999) 1739–1750
1749
4.3. Racemisation of phenylglycine methyl ester and phenylglycine amide
All kinetic measurements were made in 40 ml reaction vessels at 40°C; ammonia saturated tert-butyl
alcohol was used as solvent. When pyridoxal was used as racemisation catalyst the concentration of 1a
was varied between 50 and 400 mM, with salicylaldehyde between 30 and 300 mM. The concentration
of phenylglycine amide 2 was varied between 40 and 160 mM in both cases. The concentration of
pyridoxal as well as salicylaldehyde was kept constant at 1.3 mM. Samples were periodically withdrawn
and analysed by HPLC. The kinetic constants were determined by fitting the experimental data in a
Lineweaver–Burk plot.19
4.4. Ammonolysis with in situ racemisation
Compound 1a (1.0 mmol), racemisation catalyst and Candida antarctica lipase B in 5.0 ml solvent
saturated with ammonia were shaken in 40 ml reaction vessels at the selected temperature (see Table 5
for further details). The reactions were carried out under 1 atmosphere of ammonia by bubbling dry
ammonia gas through the reaction mixtures. The concentration of the racemisation catalyst was varied
between 0.04 and 4 mM for pyridoxal and between 1 and 4 mM for salicylaldehyde. The reactions
were followed over time. The reactions were stopped by adding formic acid and the reaction mixture
was dissolved by adding water. When tert-amyl alcohol or tert-butyl methyl ether was used as solvent,
methanol was also added. A sample was taken for HPLC analysis.
Acknowledgements
A generous donation of Novozym 435 (Candida antarctica lipase B) by Novo Nordisk (Bagsværd,
Denmark) is gratefully acknowledged. SP 611 (Candida Antarctica lipase B, immobilised on Accurel EP
100) ex Novo Nordisk A/S was kindly donated by Uniqema (Gouda, The Netherlands). This work was
supported by a grant from the Netherlands Ministry of Economic Affairs and carried out in cooperation
with Chemferm Industrial Pharmaceuticals (Breda, The Netherlands).
References
1. Bruggink, A.; Roos, E. C.; De Vroom, E. Org. Process Res. Dev. 1998, 2, 128–133.
2. Sheldon, R. A. Chem. Ind. 1990, 212–219.
3. Olivieri, R.; Fascetti, S.; Nagamachi, T.; Yoneda, K.; Yamada, H. Agric. Biol. Chem. 1981, 45, 831–838.
4. Takahashi S.; Ohashi, T.; Kii, Y.; Kumagai, H.; Yamada, H. J. Ferment. Technol. 1979, 57, 328–332.
5. Diender, M. B.; Straathof, A. J. J.; Van der Wielen, L. A. M.; Ras, C.; Heijnen, J. J. J. Mol. Catal. B: Enzym. 1998, 5,
249–253.
6. Elferink, V. H. M.; Breitgoff, D.; Kloosterman, M.; Kamphuis, J.; Van den Tweel, W. J. J.; Meijer, E. M. Recl. Trav. Chim.
Pays-Bas 1991, 110, 63–74.
7. De Zoete, M. C.; Kock-van Dalen, A. C.; Van Rantwijk, F.; Sheldon, R. A. Recl. Trav. Chim. Pays-Bas 1995, 114, 171–174.
8. Pugnière, M.; Commeyras, A.; Previero, A. Biotechnol. Lett. 1983, 5, 447–452.
9. Chen, S.-T.; Huang, W.-H.; Wang, K.-T. J. Org. Chem. 1994, 59, 7580–7581.
10. Hacking, M. A. P. J.; Wegman, M. A.; Rops, J.; Van Rantwijk, F.; Sheldon, R. A. J. Mol. Catal. B: Enzymatic 1998, 5,
155–157.
11. Chen, C.-S.; Fujimoto, Y.; Girdaukas, G.; Sih, C. J. J. Am. Chem. Soc. 1982, 104, 7294–7298.
12. Fitzpatrick, P. A.; Klibanov, A. M. J. Am. Chem. Soc. 1991, 113, 3166–3171.
13. Yang, F.; Weber, T. W.; Gainer, J. L.; Carta, G. Biotechnol. Bioeng. 1997, 56, 671–680.

1750
M. A. Wegman et al. / Tetrahedron: Asymmetry 10 (1999) 1739–1750
14. Parida, S.; Dordick, J. S. J. Org. Chem. 1993, 58, 3238–3244.
15. Sakai, T.; Kawabata, I.; Kishimoto, T.; Ema, T.; Utaka, M. J. Org. Chem. 1998, 62, 4906–4907.
16. Sakai, T.; Kishimoto, T.; Tanaka, Y.; Ema, T.; Utaka, M. Tetrahedron Lett. 1998, 39, 7881–7884.
17. Jacob, P. J. Org. Chem. 1994, 61, 2916–2917.
18. Guttmann, St. Helv. Chim. Acta 1961, 85, 733.
19. Engel, P. C. Enzymology Labfax; Bio Scientific Publishers: Oxford, 1996; pp. 84.