The present invention relates to amino acid dehydrogenase-derived biocatalysts designed by site-directed mutagenesis to produce chirally pure, non-natural L-α- amino acids and to allow kinetic resolution of racemic DL- α-amino acid mixtures, to provide chirally pure, non-natural D-α-amino acids.

As used herein, the term "α-amino acid" means an entity in which the amino group and the COOH group are both attached to the same carbon atom. As used herein, the term "natural α-amino acids" is taken as those which normally occur in human or other mammal organisms, in vivo, either as a constituent of their proteins (L- alanine, L-arginine, L-asparagine, L-aspartic acid, L-cysteine, L-glutamic acid, L-glutamine, glycine, L-histidine, L-isoleucine, L-leucine, L-lysine, L-methionine, L- phenylalanine, L-proline, L-serine, L-tyrosine, L-threonine, L-tryptophan and L- valine) or else as mainstream metabolites (e.g. L-citrulline, LL- and meso- diaminopimelic acid, L-β-3,4-dihydroxy phenylalanine, 3-iodotyrosine, 3,5- diiodotyrosine, L-homocysteine, L-homoserine, 3-hydroxykynurenine, L- hydroxyproline, L-5-hydroxytryptophane, kynurenine, β-methylaspartic acid, L- ornithine, L-thyroxine and 3,5,3 '-triiodo-L-thyronine). All other α-amino acids are considered as non-natural amino acids.

Introduction.

Enantiopure drugs constitute an increasing proportion (35% of drugs on the market in 2000) of pharmaceuticals and there is substantial commercial interest in methods for asymmetric synthesis [R.M. Williams et al, Synthesis of Optically Active Amino Acids, VoI 7 of Organic Chemistry Series; Pergamon Press, Oxford, 1989; and Duthaler, R.O., Tetrahedron, 1994, 50, 1539-1650]. Non-natural α-amino acids, in enantiopure form, are of considerable interest in the synthesis of alkaloids, peptides and other compounds with therapeutic application e.g. in HIV-protease inhibitors [A. S. Bommarius et al Tetrahedron: Asymm., 1995, 6, 2851-2888]. Incorporation of non-natural α-amino acids into biologically active peptides and proteins can greatly improve their activity, stability, bioavailability and binding specificity [A. Boto et al Tetrahedron. Lett., 2002, 43, 8269-8272]. In this context, access to both of the enantiopure series, namely each of D- and L- series, is important. Non-natural α-amino acids are increasingly in demand by the pharmaceutical industry for peptidomimetic and other single-enantiomer drugs [P.P. Taylor et al, Trends in Biotechnology, 1998, 16, 412-418]. They are also in demand as precursors to ligands for asymmetric synthesis [A.S. Bommarras et al, Tetrahedron: Asymm., 1995, 6, 2851-2888]. While some enantiopure non-natural α-amino acids are commercially available, they are expensive.

One of the most important strategies for asymmetric synthesis involves biocatalysts - the application of biological species such as microbial cells or enzymes derived therefrom to catalyse organic reactions [K. Faber, Biotransformations in Organic Chemistry - A Text Book, 4th ed. Berlin: Springer- Verlag, 2000; and R.N. Patel, Stereoselective Biocatalysis, New York: Marcel Dekker, 2000]. Many biocatalysts exhibit high regio-, chemo- and stereo- selectivity making them superior to chemical catalysts for asymmetric synthesis [H. G. Davies et al, Crit. Rev. Biotechnol, 1990, 10, 129-153; and E. Santaniello et al, Chem. Rev., 1992, 92, 1071-1140]. Furthermore, biocatalysts are ideal energy-efficient, environmentally acceptable reagents, as virtually all reactions proceed under mild conditions and avoid the use of toxic reagents and disposal of byproducts. Thus, biocatalysts offer a. good opportunity to prepare industrially useful chiral compounds [A.L. Margolin, Enzyme Microb. Technol., 1993, 15, 266-280; K. Mori, Synlett, 1995, 11, 1097-1109; R.N. Patel, Adv. Appl. Microbiol, 1997, 43, 91-140; and R.N. Patel, Adv. Appl. Microbiol, 2000, 47, 33-78].

While naturally occurring enzymes have been widely employed in asymmetric synthesis [O.P.Ward and A.Singh, Current Opinion In Biotechnology, 2000, 11(6), 520-526], there are instances in which they cannot be readily employed owing to their limited substrate scope. Designer biocatalysts, where the naturally occurring protein is modified structurally to provide the desired reaction site selectivity, could potentially offer many advantages. In the specific case of the amino acid dehydrogenases [Smith, E. L. et al (1975) In

'The Enzymes' 3ld Edn. (Boyer, P.D. ed.) 11, 293], a reversible reaction is catalysed,

in which either an L-α-amino acid is oxidised by the coenzyme nicotinamide adenine

dήϊucleotide (NAD+) to release ammonia and an α-keto (2-oxo) acid with formation

of the reduced co factor NADH or, conversely, ammonia is incorporated by reductive

amination of an α-keto acid to foπn the corresponding α-amino acid entirely in the

L- form (Eqn. 1). Also occasionally with some amino acid dehydrogenases, the

coenzyme NAD+/NADH may be replaced by its phosphorylated counterpart

NADPVNADPH.

R CH. NH2. COOH + NAD+ R CO. COOH + NADH + NH3 + H+

These enzymes thus offer a means of cleanly introducing a stereogenic centre at the

α-carbon atom. From the standpoint of utilising these dehydrogenases as synthetic

tools, the general limitation noted above, namely that of circumscribed substrate

specificity, undoubtedly applies. Each can deal with its biological substrate and in

some cases close analogues but this set of acceptable substrates is usually quite

limited.

In the case of phenylalanine dehydrogenase, the range of activities with naturally-

occurring α-amino acids has been assessed and described by Asano et al. and Seah et

al. as shown in Table 1.

Table 1. Specific activities (s"1) were calculated using a subunit Mw of 41.6kDa. Reaction conditions were 2.5mM NAD+ and in 50 mM glycine -KOH supplemented with 10 mM KCl

adjusted to pH 10.4 at 250C. (a)- [Asano et al, J. Biol Chem. 1987, 262, 10346] test substrate was 1OmM except for tyrosine (0.3mM) (b) Kinetic analysis was done at 250C in the same buffer with amino acid concentrations varied to allow extrapolation of the Vmax value.[Seah et al, Biochemistry 2002, 41, 11390-11397]

Phenylalanine dehydrogenase (PheDH) nevertheless has been employed for the enantioselective synthesis of L-2-amino-4-phenylbutanoic acid with excellent enantiopurity by supplying the homologous appropriate α-keto acid [E. Santaniello et al, Chem. Rev., 1992, 92, 1071-1140; Y. Asano et al, J. Org. Chem., 1990, 55, 5567; and CW. Bradshaw et al, Bioorg. Chem., 1991, 19, 29].

Site-directed mutagenesis allows alteration of the amino acid residues surrounding the substrate-binding pocket of the enzyme to alter the size, shape and polarity of the pocket. Seah et al [FEBS Letters 1995, 370, 93-96] undertook site-directed mutagenesis on PheDH and reported in 1995 that the resulting mutant enzymes displayed reduced activity for L-phenylalanine compared to the wild type enzyme and enhanced activity towards other natural α-amino acid substrates, indicating that the substrate profile of the enzyme was varied by the mutations. Enhanced discrimination between phenylalanine and tyrosine in another set of engineered enzymes was subsequently reported [Biochemistry 2002, 41, 11390-11397]. The present invention concerns the hypothesis that engineered amino acid dehydrogenase mutants, for example PheDH mutants, might prove useful as biocatalysts for the asymmetric synthesis of a wide range of non-natural α-amino acids, such as phenylalanine analogues.

The basis of this site-directed mutagenesis of PheDH was initially a recognition of the common chemistry and shared structural features of the amino acid dehydrogenase family [K.L. Britton et al, J. MoI. Biol., 1993, 234, 938-945], with glutamate dehydrogenase as the archetype for which a high-resolution structure had been solved by X-ray crystallography [P.J. Baker et al, Proteins, 1992, 12, 75-86; and TJ. Stillman et al, J.Mσ/5zø/., 1992, 224, 1181-1184]. More recently, this strategy has been further strengthened by the direct solution of the structure of a PheDH [J.L.Vanhooke et al, Biochemistry, 1999, 38, 2326-2339]. The active site of phenylalanine dehydrogenase (PheDH) from Bacillus sphaericus has been mutated on the basis of homology modelling [S.Y. Seah et al, FEBS Letters 1995, 370, 93-96] and several mutants affecting the contact with the aromatic ring of the natural α-amino acid substrate have been studied.

Research by one of the present inventors during the 1980's and 1990's led to the first determination by X-ray crystallography of the 3-D structure of one of these enzymes at high resolution. This was the glutamate dehydrogenase (GIuDH) of Clostridium symbiosum [P. J. Baker et al, Proteins (1992) 12, 75-86; and Stillman et al, (1993) J. MoI. Biol. 234, 1131-1139]. A realisation a) that a number of other amino acid dehydrogenases (e.g. leucine, valine and phenylalanine dehydrogenases (PheDH)) showed amino acid sequence homology to GIuDH; b) that the 3-D structure in general reflects strong similarities in linear primary sequence; c) that the high-resolution structure of GIuDH provides an unambiguous structural interpretation of the basis of its amino acid specificity led to the prediction that our 3-D insight allowed us to interpret the basis of specificity also in other amino acid, dehydrogenases [K. L. Britton et al, J. MoI. Biol. (1993) 234, 938-945]. This in turn allowed specification of ways in which to alter amino acid specificity by mutating key amino acid residues in the active-site region. A number of these key residue positions were listed in US Patent No. 5,798,234, which is co-owned by one of the present inventors.

Subsequent research by some of the present inventors has established that mutant amino acid dehydrogenases may be successfully made and purified with the same ease as the normal 'wild-type' enzyme. (Similar protein engineering work has of course been carried out with many other enzymes, but each enzyme is unique and the general view that site-directed mutagenesis is a legitimate and feasible strategy cannot be unconditionally extrapolated; each case has to be tested and established). Secondly it has established with three different amino acid dehydrogenase starting points (GIuDH, VaIDH and PheDH) that it is possible to produce either subtle or radical changes to amino acid specificity [S. Y. K Seah et al, FEBS Letters (1995) 370, 93- 96; S.Y.K. Seah et al, Eur. J. Biochem. (2003) 270, 1-7; CG. Hyun et al, Antonie van Leeuwenhoek (2000) 78, 237; and X.-G. Wang et al, Eur. J. Biochem. (2001) 268, 5791-5799]. The present invention shows that engineered amino acid dehydrogenase biocatalysts can, surprisingly, handle a wide range of non-natural α-amino acids, often with high efficiency and always with perfect retention of the absolute discrimination between the L- and D- series of α-amino acids, the former being substrates, the latter being in some cases inhibitors but never substrates [P. Busca et al, Org. Biomol. Chem. (2004) 2, 2684]. These biocatalysts may be used for quantitative conversion of an α- ketoacid, achiral at the α-position, to the corresponding non-natural L-α-amino acid and vice versa, namely, for quantitative conversion of a non-natural L-α-amino acid to the corresponding α-ketoacid, achiral at the α-position.

According to a first aspect of the invention there is provided a process for preparing a non-natural L-α-amino acid of the formula

NH2 R ^^ ^X COOH

from an α-ketoacid, achiral at the α-position, the process comprising

reacting the α-ketoacid of the formula

O

R- X^ COOH

in which X is a substituted or unsubstituted, saturated or unsaturated C0-3 (optionally Co-Oalkylene radical, optionally containing a heteroatom (for example, O, S and/or N); and

R is a substituted or unsubstituted, branched or unbranched, saturated or unsaturated, lower alkyl of Cj-8 (optionally C] -6), preferably Ci-5 or Ci-3 or C3-5; a substituted or unsubstituted aryl radical; a substituted or unsubstituted fused aryl radical; a substituted or unsubstituted, saturated or unsaturated heterocyclic radical; a substituted or unsubstituted, saturated or unsaturated fused heterocyclic radical; or a substituted or unsubstituted cycloalkyl radical of C5-8(optionally C5-6), with the proviso that, when X is a C2 saturated alkylene radical, R is not selected from the group consisting of an unsubstituted phenyl radical, an unsubstituted C1-2 saturated alkyl radical or a C1 saturated alkyl radical substituted with -CONHNH2

with an ammonia source, a mutant amino acid dehydrogenase and a source of an appropriate reduced coenzyme (for example NADH or NADPH but other analogue coenzyme molecules able to satisfy the coenzyme specificity e.g. deamino NADH, reduced acetyl pyridine adenine dinucleotide etc. may also be employed) in a suitable reaction solvent, under reaction conditions sufficient to form the non-natural L- α-amino acid.

X may be substituted with a carbon-based substituent such as a C1-5 (optionally C1-3) alkyl, alkenyl or alkynyl group. Alternatively, X may be substituted with a non- carbon based substituent (by non-carbon based, is meant substituents containing no carbon atoms such as halide (such as fluoride or chloride), -OH, -NH2, -SH or -NO2, as well as, substituents containing carbon atoms but linked by at least one non-carbon atom to the rest of the moiety such as C1-5, optionally C1-3, alkoxy, alkenyloxy and alkynyloxy such as-OCH3 or C1-5, optionally C1-3;thioalkyl, thioalkenyl and thioalkynyl such as-SCH3.

When R is a substituted or unsubstituted, branched or unbranched, saturated or unsaturated, lower alkyl of C1-8 (optionally C1-6), preferably Cj-5 or C1-3 or C3-s; R may be substituted with a non-carbon based substituent (by non-carbon based, is meant substituents containing no carbon atoms such as halide (such as fluoride or chloride), -OH, -NH2, -NHNH2, -CO, -SH or -NO2, as well as, substituents containing carbon atoms but linked by at least one non-carbon atom to the rest of the moiety such as C1- 5, optionally Ci-3, alkoxy, alkenyloxy and alkynyloxy such as-OCH3 or C1-5, optionally Ci-3)thioalkyl, thioalkenyl and thioalkynyl such as-SCH3. Optionally, R is a substituted or unsubstituted aryl radical; a substituted or unsubstituted fused aryl radical; a substituted or unsubstituted, saturated or unsaturated heterocyclic radical; a substituted or unsubstituted, saturated or unsaturated fused heterocyclic radical; or a substituted or unsubstituted cycloalkyl radical of C5-8(optionally C5-6).

Preferably, R is a substituted or unsubstituted, simple or fused aryl radical, preferably a substituted or unsubstituted phenyl radical or a substituted or unsubstituted naphthyl radical, optionally a substituted phenyl or naphthyl radical. IfR is phenyl, it maybe substituted at one or more of ortho, meta or para positions (optionally the para position) with aldehyde; nitrile; nitro; halo, for example fluoro or chloro; lower C1-5 alkyl, for example Ci-3 alkyl; lower C1-5 alkoxy, for example C1-3 alkoxy; lower Ci-5 haloalkyl such as C]-3 haloalkyl, including, but not limited to, lower C1-5 perhaloalkyl such as C1-3 perhaloalkyl; lower C1-5 haloalkoxy such as Ci-3 haloalkoxy, including, but not limited to, lower C1-5 perhaloalkoxy such as C1-3 perhaloalkoxy; hydroxy or a mixture thereof.

Alternatively or additionally, R is a substituted or unsubstituted heterocyclic radical optionally selected from substituted or unsubstituted furan, pyran, pyrrole, imidazole, pyrazole, isothiazole, isoxazole, pyridine, pyrazine, pyrimidine, pyridazine, furazan, pyrrolidine, pyrroline, imidazolidine, imidazoline, pyrazolidine, pyrazoline, piperidine, piperazine, morpholine or thiophene. R may be substituted with a carbon- . based substituent such as a C1-5 (optionally Ci-3) alkyl, alkenyl or alkynyl group. Alternatively, R may be substituted with a non-carbon based substituent (by non- carbon based, is meant substituents containing no carbon atoms such as halide (such as fluoride or chloride), -OH, -NH2, -SH or -NO2, as well as, substituents containing carbon atoms but linked by at least one non-carbon atom to the rest of the moiety such as C1-5, optionally C1-3) alkoxy, alkenyloxy and alkynyloxy such as-OCH3 or C1-5, optionally C1-3;thioalkyl, thioalkenyl and thioalkynyl such as-SCH3.

Alternatively or additionally, R is a substituted or unsubstituted fused heterocyclic radical optionally selected from substituted or unsubstituted benzofuran, isobenzofuran, indole, isoindole, benzothiophene, benzo[c]thiophene, benzimidazole, purine, indazole, benzoxazole, benzisoxazole, benzothiazole, quinoline, isoquinoline, quinoxaline, quinazoline or cinnoline. R may be substituted with a carbon-based substituent such as a C]-5 (optionally Cj-3) alkyl, alkenyl or alkynyl group. Alternatively, R may be substituted with a non-carbon based substituent (by non- carbon based, is meant substituents containing no carbon atoms such as halide (such as fluoride or chloride), -OH, -NH2, -SH or -NO2, as well as, substituents containing carbon atoms but linked by at least one non-carbon atom to the rest of the moiety such as Cj-5, optionally Q-3; alkoxy, alkenyloxy and alkynyloxy such as-OCH3 or Q-5, optionally C1-3jthioalkyl, thioalkenyl and thioalkynyl such as-SCH3.

Alternatively or additionally, R is a substituted or unsubstituted, saturated or unsaturated, cycloCs-s alkyl radical such as cyclopentyl or cyclohexyl. R may be substituted with a carbon-based substituent such as a C1-5 (optionally C1-3) alkyl, alkenyl or alkynyl group. Alternatively, R may be substituted with a non-carbon based substituent (by non-carbon based, is meant substituents containing no carbon atoms such as halide (such as fluoride or chloride), -OH, -NH2, -SH or -NO2, as well as, substituents containing carbon atoms but linked by at least one non-carbon atom to the rest of the moiety such as C1-5, optionally C1-3, alkoxy, alkenyloxy and alkynyloxy such as-OCH3 or Cj-5, optionally Ci-3,thioalkyl, thioalkenyl and thioalkynyl such as- SCH3.

Preferably, the ammonia source is selected from ammonia or an ammonium salt. Optionally, the ammonium salt is selected from ammonium sulphate, ammonium esters (such as ammonium formate and ammonium acetate) and ammonium halides, for example, ammonium chloride.

The source of an appropriate reduced coenzyme is either the appropriate reduced coenzyme itself or, alternatively, the corresponding oxidised coenzyme and means for converting the corresponding oxidised coenzyme into the appropriate reduced coenzyme. For example, the converting means might comprise ethanol and an alcohol dehydrogenase. If such a converting means is used, reduction of the ethanol to ethanal, in turn, converts the oxidised coenzyme into the corresponding reduced coenzyme. The extent of conversion of an α-ketoacid substrate to the L-α-amino acid product is limited by the stoichiometry of reaction, so that x millimoles of amino acid, for instance, cannot be quantitatively formed from x millimoles of α-ketoacid unless x millimoles of reduced coenzyme and x millimoles of ammonium ions (as ammonia or preferably as an ammonium salt) are supplied over the timecourse of the reaction. The requisite supply of reduced coenzyme may be accomplished in two ways. The required x millimoles of reduced coenzyme may be supplied in total at the outset; alternatively and preferably in view of the high commercial cost of the reduced coenzyme, the reaction mixture may be supplemented with the components of the aforementioned converting means.

Where coenzyme recycling is not envisaged, therefore, the reaction mixture will require a clear stoichiometric excess of a reduced coenzyme such as NADH and ammonium ions over the starting ketoacid. Where recycling is envisaged, a stoichiometric excess of ammonium ions is still required, but the coenzyme may be supplied, for example either as NADH or as NAD+, in a greatly reduced, catalytic amount.

The mutant amino acid dehydrogenase useful in the processes of the first, second and third aspects of the invention is selected from a mutant of glutamate dehydrogenase, phenylalanine dehydrogenase, leucine dehydrogenase and valine dehydrogenase, preferably phenylalanine dehydrogenase. The mutant is prepared by site-directed mutagenesis of the corresponding wild type enzyme by following, for example, the teaching of US Patent No. 5,798,234, the contents of which are incorporated herein by reference. Nl 45 mutant amino acid dehydrogenases, as will be observed hereinafter, favour substrates with a hydrophobic substitution at the para position of an aryl radical, if an aryl radical is present.

In a further aspect of the invention there is provided a mutant phenylalanine dehydrogenase derived from Bacillus sphaericus and having the following amino acid sequence:

N145A/L307V MUTANT SEQUENCE MAKQLEKSSKIGNEDVFQKIANHEQIVFCNDPVSGLQAIIAIHDTTLGPALGGT RMYPYKNVDEALEDVLRLSEGMTYKCAAADIDFGGGKAVIIGDPEKDKSPAL FRAFGQFVESLNGRFYTGTDMGTTMDDFVHAQKETNFIAGIPEQYGGSGDSSI PTAQGVIY ALKATNQYLFGSDSLSGKTY AIQGLGKVGYKVAEQLLKAGADLF VTDIHENVLNSIKQKSEELGGSVTIVKSDDIYSVQADIF VPCAMGGIINDKTIPK LKVKAVVGSANNQLKDLRHANVLNEKGILYAPDYIVNAGGVIQVADELYGP NKERVLLKTKEIYRSLLEIFNQAALDCITTVEAANRKCQKTIEGQQTRNSFFSR GRRPKWNIKE.

The mutant amino acid dehydrogenase of the further aspect of the invention favours bulky substrates, as will be observed hereinafter.

Asymmetric Synthesis of L-series of non-natural amino acids

The above-mentioned mutant enzymes(s) may also be used in the opposite direction of reaction, starting with a racemic mixture of a non-natural α-amino acid, to effect quantitative removal of the L-enantiomer. This leaves the enantiomerically pure D- α-amino acid, readily separable from other reaction products.

Thus, according to a second aspect of the invention there is provided a method for resolving a racemic, non-natural α-amino acid, to provide the D-α-amino acid of the formula

, the method comprising

contacting the racemic non-natural α-amino acid of the formula

in which X is a substituted or unsubstituted, saturated or unsaturated Co-3, optionally Co-i,alkylene radical, optionally containing a heteroatom (for example, O, S and N); and

R is a substituted or unsubstituted, branched or unbranched, saturated or unsaturated, lower alkyl of C1-8 (optionally C1-6), preferably C1^; a substituted or unsubstituted aryl radical; a substituted or unsubstituted fused aryl radical;a substituted or unsubstituted, saturated or unsaturated heterocyclic radical; a substituted or unsubstituted, saturated or unsaturated fused heterocyclic radical; or a substituted or unsubstituted cycloalkyl radical of C5-8, with the proviso that, when X is a C2 saturated alkylene radical, R is not selected from the group consisting of an unsubstituted phenyl radical, an unsubstituted C1-2 saturated alkyl radical or a Cjsaturated alkyl radical substituted with -CONHNH2

with water, a mutant amino acid dehydrogenase and a source of an appropriate oxidised coenzyme (NAD+ or NADP but optionally other analogue coenzyme molecules able to satisfy the coenzyme specificity e.g. deamino NAD+, acetyl pyridine adenine dinucleotide etc.) in a suitable reaction solvent, under reaction conditions sufficient to convert the L-α-amino acid of the racemic non-natural α-amino acid into an α-ketoacid; and leaving the unreacted D-α-amino acid.

The extent of conversion of an L-α-amino acid substrate to the α-ketoacid product is limited by the stoichiometry of reaction, so that x millimoles of α-ketoacid, for instance, cannot be quantitatively formed from x millimoles of amino acid unless x millimoles of oxidised coenzyme and x millimoles of water are supplied over the timecourse of the reaction. The requisite supply of oxidised coenzyme may be accomplished in two ways. The required x millimoles of oxidised coenzyme such as NAD+ may be supplied either in total at the outset or over the timecourse of the reaction. The requisite water supply will normally be ensured by either using a reaction solvent comprising water or by using a reaction solvent which contains water and which optionally has a predominantly aqueous environment. It will be appreciated by those skilled in the art that pure water is 55.6M. Thus, in a reaction system containing 10 or 20% water, the water would still be present in a vast stoichiometric excess. In fact, even in a reaction system containing 1% water, the water would still be present in a substantial stoichiometric excess over any likely working concentration of the amino acid substrate.

Where coenzyme recycling is not envisaged, therefore, the reaction mixture will require a clear stoichiometric excess of an oxidised coenzyme such as NAD+ or NADP+ over the starting amino acid. Where recycling is not envisaged, a stoichiometric excess of water is still required.

Preferably, the reaction solvent comprises the water.

Preferably, the source of an appropriate oxidised coenzyme is the oxidised coenzyme itself. Alternatively, the source of an appropriate oxidised coenzyme is the corresponding reduced coenzyme, together with means for converting the corresponding reduced coenzyme into the appropriate oxidised coenzyme. Such a converting means might comprise diaphorase and a source of oxygen, each in an amount sufficient to convert, for example, NADH/NADPH to NAD+/NADP+.

X may be substituted with a carbon-based substituent such as a C]-5 (optionally C1-3) alkyl, alkenyl or alkynyl group. Alternatively, X may be substituted with a non- carbon based substituent (by non-carbon based, is meant substituents containing no carbon atoms such as halide (such as fluoride or chloride), -OH, -NH2, -SH or -NO2, as well as, substituents containing carbon atoms but linked by at least one non-carbon atom to the rest of the moiety such as C]-5, optionally C1-3, alkoxy, alkenyloxy and alkynyloxy such as-OCH3 or Ci-55 optionally C1-3ithioalkyl, thioalkenyl and thioalkynyl such as-SCH3 . When R is a substituted or unsubstituted, branched or unbranched, saturated or unsaturated, lower alkyl of C1-8 (optionally Q-6), preferably C]-5 or C1-3 or C3-5, R may be substituted with a non-carbon based substituent (by non-carbon based, is meant substituents containing no carbon atoms such as halide (such as fluoride or chloride), -OH, -NH2, -NHNH2, -CO, -SH or -NO2, as well as, substituents containing carbon atoms but linked by at least one non-carbon atom to the rest of the moiety such as C1- 5, optionally C1-3; alkoxy, alkenyloxy and alkynyloxy such as-OCH3 or C1-5, optionally C1-3jthioalkyl, thioalkenyl and thioalkynyl such as-SCH3.

Optionally, R is a substituted or unsubstituted aryl radical; a substituted or unsubstituted fused aryl radical; a substituted or unsubstituted, saturated or unsaturated heterocyclic radical; a substituted or unsubstituted, saturated or unsaturated fused heterocyclic radical; or a substituted or unsubstituted cycloalkyl radical of C5-8(optionally C5-6).

Preferably, R is a substituted or unsubstituted, simple or fused aryl radical, preferably a substituted or unsubstituted phenyl radical or a substituted or unsubstituted haphthyl radical, optionally a substituted phenyl or naphthyl radical. If R is phenyl R may be substituted at one or more of ortho, meta or para positions (optionally, the para position) with aldehyde; nitrile; nitro; halo, for example fluoro or chloro; lower C1-5 alkyl, for example Ci-3 alkyl; lower Ci-5 alkoxy, for example C1-3 alkoxy; lower C1-5 haloalkyl such as C1-3 haloalkyl, including, but not limited to, lower C1-5 perhaloalkyl such as C1-3 perhaloalkyl; lower C1-5 haloalkoxy such as Ci-3 haloalkoxy, including, but not limited to, lower C1-5 perhaloalkoxy such as C1-3 perhaloalkoxy; hydroxy or a mixture thereof.

Alternatively or additionally, R is a substituted or unsubstituted heterocyclic radical, optionally selected from substituted or unsubstituted furan, pyran, pyrrole, imidazole, pyrazole, isothiazole, isoxazole, pyridine, pyrazine, pyrimidine, pyridazine, furazan, pyrrolidine, pyrroline, imidazolidine, imidazoline, pyrazolidine, pyrazoline, piperidine, piperazine, morpholine or thiophene. R may be substituted with a carbon- based substituent such as a C]-5 (optionally Cj-3) alkyl, alkenyl or alkynyl group. Alternatively, R maybe substituted with a non-carbon based substituent (by non- carbon based, is meant substituents containing no carbon atoms such as halide (such as fluoride or chloride), -OH, -NH2, -SH or -NO2, as well as, substituents containing carbon atoms but linked by at least one non-carbon atom to the rest of the moiety such as Ci-5, optionally Cj-3, alkoxy, alkenyloxy and alkynyloxy such as-OCH3 or C1-5, optionally C1-3,thioalkyl, thioalkenyl and thioalkynyl such as-SCH3.

Alternatively or additionally, R is a substituted or unsubstituted fused heterocyclic radical optionally selected from substituted or unsubstituted benzofuran, isobenzofuran, indole, isoindole, benzothiophene, benzo[c]thiophene, benzimidazole, purine, indazole, benzoxazole, benzisoxazole, benzothiazole, quinoline, isoquinoline, quinoxaline, quinazoline or cinnoline. R may be substituted with a carbon-based substituent such as a C1-5 (optionally Cj-3) alkyl, alkenyl or alkynyl group. Alternatively, R may be substituted with a non-carbon based substituent (by non- carbon based, is meant substituents containing no carbon atoms such as halide (such as fluoride or chloride), -OH, -NH2, -SH or -NO2, as well as, substituents containing carbon atoms but linked by at least one non-carbon atom to the rest of the moiety such as C1-5, optionally C1-3, alkoxy, alkenyloxy and alkynyloxy such as-OCH3 or C1-5, optionally C1-3>thioalkyl, thioalkenyl and thioalkynyl such as-SCH3.

Alternatively or additionally, R is a substituted or unsubstituted, saturated or unsaturated, cycloC5-8alkyl radical such as cyclopentyl or cyclohexyl. R may be substituted with a carbon-based substituent such as a Cj-5 (optionally C1-3) alkyl, alkenyl or alkynyl group. Alternatively, R may be substituted with a non-carbon based substituent (by non-carbon based, is meant substituents containing no carbon atoms such as halide (such as fluoride or chloride), -OH, -NH2, -SH or -NO2, as well as, substituents containing carbon atoms but linked by at least one non-carbon atom to the rest of the moiety such as C1-5, optionally C1-3> alkoxy, alkenyloxy and alkynyloxy such as-OCH3 or C1-5, optionally C1-3jthioalkyl, thioalkenyl and thioalkynyl such as- SCH3.

Preferably, the reaction solvent comprises water and the water is an aqueous buffer. Kinetic Resolution of Racemic Amino Acids to provide D-series

racemic ammo aci ■dA H2° NH,

+

The byproduct of the above-recited method of the second aspect of the invention, namely, the α-ketoacid may, by sequential operation of the method of the first aspect of the invention, be converted to the L-α-amino acid, so that the biocatalyst has in effect resolved the racemic mixture into its two components. This combined strategy, putting together the two processes illustrated above, gives access to both pure enantiomers starting from the racemic α-amino acid. Given that the racemic mixture is often the most readily accessible synthetic product, this combined process, yielding two high-value products, is very attractive commercially. Combined strategy for access to both enantiomeric series of amino acids

racemic amino acid NH, +

It has further been shown that the biocatalyst, whether purified or in the form of whole cell, may be successfully immobilised on a solid support via adsorption, entrapment or cross-linking on Celite (Trade Mark), alginate, chitosan -or a mixture of the last two- silica gel or agarose beads, both for enhanced stability and for ease of separation from a reaction mixture and reuse [S. S. Betigeri et al, Biomaterials (2002) 23, 3627; S. Soni et al, Journal of Applied Polymer Science (2001) 82, 1299; and W. Limbut et al, Biosensors and Bioelectronics (2004) 19, 813]. Also the biocatalyst may be used in the presence of miscible organic solvents (short chain alcohols such as methanol, miscible ketones such as acetone, nitriles such as acetonitrile, miscible ethers, amides, alkyl sulfoxides, dioxane, tetrahydrofuran), facilitating reaction with substrates poorly soluble in purely aqueous reaction media. Finally the reaction may be successfully run in a 2-phase solvent system, where the biocatalyst and the coenzyme are in the predominantly aqueous phase but a second, immiscible phase (e.g. aliphatic or aromatic hydrocarbons such as hexane, immiscible ethers such as diethyl ether, chlorinated alkanes, esters, immiscible ketones, pyridine) contains the bulk of the poorly water-soluble α-ketoacid substrate. In the drawings: Fig. 1. HPLC separation of 4-Cl-phenyl pyruvate (3) + DL-4-Cl-Phe (A) and HPLC chromatogram of the reductive animation of 4-Cl-phenyl pyruvate (3) with the mutant amino acid dehydrogenase N 145 A (B).

Fig. 2. NMR spectrum of the reaction mixture of the reductive amination of 1 mmol 4-F-phenyl pyruvate (6) after 24hrs.

Fig. 3. HPLC chromatogram of the reaction mixture after 44hrs (peak at 3.0 is the α- ketoacid, peak at 4.9 min is small impurity, peak at 8.1 min is the D enantiomer unreacted. The L enantiomer has a RT of 6.9 min under these elution conditions.

Experimental General procedures All the reagents were reagent-grade and used without further purification. HPLC grade solvents were used for HPLC chromatographic separations. The NADH grade II was purchased from Roche. The wild type PheDH and the mutants were over- expressed in E. coli TGl cells and purified as described elsewhere [S.Y. Seah et al, FEBS Letters 1995, 370, 93-96]. The enzymes were stored as precipitates in 60% ammonium sulphate at 4°C and desalted before use through extensive dialysis against Tris buffer pH 8.0, 5OmM.

All solvents were distilled prior to use as follows: Dichloromethane and ethyl acetate were distilled from phosphorous pentoxide. Diethyl ether and THF were freshly distilled from sodium in the presence of benzophenone. Melting points were measured on a Thomas Hoover Capillary melting point apparatus and are uncorrected. Elemental analyses were performed by the Microanalysis Laboratory, University College Cork, using a Perkin Elmer 240 and an Exeter Analytical CE440 elemental analyser. Infrared (IR) spectra were recorded on a Perkin-Elmer FT-IR Paragon 1000. Liquid samples were examined as thin films interspersed between sodium chloride plates. Solid samples were dispersed in potassium bromide and recorded as pressed discs. 1H NMR spectra were recorded at 300MHz on a Bruker AVANCE 300 spectrometer. 13C NMR spectra were recorded on a Bruker AVANCE 300 instrument at 75MHz. Spectra were run in deuteriochloroform (CDCl3) with tetramethylsilane (TMS) as internal standard unless otherwise specified. Chemical shifts (5H and δc) are expressed as parts per million (ppm), positive shifts being downfield from TMS. Splitting patterns in 1H spectra are designated as s (singlet), bs (broad singlet), bd (broad doublet), d (doublet), t (triplet), q (quartet), dd (doublet of doublets), ABq (AB quartet) and (m) multiplet. Coupling constants are quoted in Hz. For the 13C NMR spectra, assignments are made from 13C DEPT spectra run in the DEPT-90 and DEPT-135 modes and with the aid of COSY and HETCOR experiments in some cases. Mass spectra were measured on a Kratos Profile spectrometer in electron impact (E.I.) mode with an ionisation voltage of 7OeV.

All organic solutions were dried using magnesium sulphate unless otherwise specified. Thin layer chromatography was performed on precoated silica gel (Merck HF254) plates and compounds were visualised under ultraviolet light. Column "wet flash" chromatography was performed using Merck silica gel 60 (typical ratio of silicaxrude reaction mixture ~30:l). Bulb to bulb distillations were carried out on a Buchi GKR-50 Kugelrohr.

For the Grignard reactions, Et2O was freshly distilled from LiAlH4 and the molarity of the reagent was determined by titration of aliquots with IM HCl. For the preparation of organocadmium reagents, CdCl2 was stored in a high temperature oven to keep it dry.

Methods and Materials. The genes. In all cases the starting point for creating the tailored biocatalysts is the cloned gene for a naturally-occurring amino acid dehydrogenase. The one used herein is the phenylalanine dehydrogenase of Bacillus sphaericus [S. Y. K Seah et al, FEBS Letters (1995) 370, 93-96] but it will be appreciated by those skilled in the art that the invention is not so limited. The DNA coding sequence for the enzyme in question is inserted into a cloning site in a suitable expression vector. This is a plasmid which provides an inducible promoter, allowing high-level production of the enzyme to be initiated as desired. It also contains a drug resistance gene so that the presence of the plasmid in a bacterial host population may be maintained by the selective pressure of including the drug in the growth medium.

Mutagenesis. In order to produce mutant biocatalysts, amino acid sidechains in the protein structure are targeted by studying the protein 3-D structure and selecting those -most .likely to affect the fit of the substrate to its binding pocket on the protein surface. In the case of clostridial GIuDH this can be done by reference to the published high-resolution structures [P. J. Baker et al, Proteins (1992) 12, 75-86; and Stillman, TJ. et al, (1993) J. MoI. Biol. 234, 1131-1139]. In the case of PheDH, it was done initially [S. Y. K Seah et al, FEBS Letters (1995) 370, 93-96; and S.Y.K. Seali et al, Eur. J. Biochem. (2003) 270, 1-7] by using a modelled structure based on the GIuDH structure and observed sequence homology [K. L. Britton et al, J. MoI. Biol. (1993) 234, 938-945]. However, subsequently the high-resolution structure has been published for a PheDH [J.L. Vanhooke et al, (1999) Biochemistry 38, 2326- 2339] and, although it is not the PheDH of B. sphaericus, it provides useful information. The relevant positions for site-directed mutagenesis include those listed in US Patent No. 5,798,234 and also another position [S.Y.K. Seah et al, Eur. J. Biochem. (2003) 270, 1-7], corresponding to Thr 193 in clostridial GIuDH and Asn 145 in PheDH of B. sphaericus. Changes at these positions, singly or in combination, affect the size, shape and local polarity or charge of the substrate binding pocket.

The changes in sequence in the biocatalysts are brought about by site-directed mutagenesis using commercially synthesised mismatch oligonucleotides and standard procedures [e.g. S.Y.K. Seah et al, FEBS Letters (1995) 370, 93-96; S.Y.K. Seah et al, Eur. J. Biochem. (2003) 270, 1-7; and X.-G. Wang et al, Eur. J. Biochem. (2001) 268, 5791-5799], and are then checked by DNA sequencing to ensure that the correct change, and no other, has been made. Alternatively, mutations may be introduced in a random way by using chemical mutagens, error-prone PCR and/or 'directed evolution', involving a nutritional selective pressure. In this case, mutants with desirable properties may be identified by activity screening of colony blots from Petri dishes. Such mutations may be at the positions listed above but not necessarily so since alterations elsewhere may produce small structural movements that affect catalytic performance. In this context, it is pertinent to note that catalysis requires not only adequate binding of the substrate but also appropriate orientation in relation to catalytic groups, so that a small change in geometry may have large effects on catalytic rate.

Growth of cells and induction of recombinant gene expression. Production of a biocatalyst is initiated by transforming a population of host cells with a preparation of the plasmid (or other vectors like P 12) carrying the mutated gene of interest. In our hands the host cells have been strains of Escherichia coli but are not intended to be so limited. Such cells are propagated in Petri dishes on layers of nutrient agar containing the selective drug, in our case usually the antibiotic ampicillin. The most efficient production of biocatalyst is achieved by taking a fresh colony from such a Petri dish and using it to inoculate larger scale cultures. The expression vectors used in our own studies carry the tac promoter and thus are inducible by addition of IPTG, a standard, non-metabolisable analogue of the natural biological promoter substance.

Typically, under laboratory conditions, a colony of cells may be used to inoculate a 5 or 10 ml culture which, after approximately 12 hours growth at 370C, is used as the inoculum for a 1 litre growth. Usually IPTG is added to the 1 litre culture under sterile conditions in mid-log phase - i.e. after about 9 hours, but the choice of timing may be varied to optimise yields of enzyme. Under these conditions the cells produce 20-40% of their total cell protein as the desired enzyme (this may be approximately assessed by the intensity of Coomassie Blue staining on SDS-PAGE gels), and a 1 litre culture may yield 20 - 150 mg of pure enzyme (typically 40 - 80 mg) following purification from the cells.

Growth can also be successfully scaled up e.g. to 20 litres in fermenter culture, and the use of a fed batch approach permits a large increase in cell density and hence an increase of up to 50-fold in the yield of biocatalyst per litre of vessel capacity. Once the cells enter stationary phase (typically about 18 hr.) they are harvested e.g. by centrifugation.

Cell breakage and enzyme purification. If cell breakage is required (see later) the cell paste is resuspended in a limited volume of liquid (aqueous buffer at pH 7-8) so that the cells in the resultant slurry may be broken open e.g. by sonic disintegration or by shear forces in an instrument such as the Manton-Gaulin homogeniser. The released soluble material is separated from cell walls and other solid debris e.g. by centrifugation. For some purposes the resulting crude extract may be an adequate source of biocatalyst without further treatment but, if required, the extract is further processed by standard column chromatographic procedures to yield homogeneous biocatalyst protein. The purification procedure for clostridial GIuDH is given in S.E.- H Syed et al, Biochim. Biophys. Acta (1991) 1115, 123-130 and that for PheDH in S.Y.K. Seah et al, FEBS Letters (1995) 370, 93-96 and thus far these procedures have also proved satisfactory for mutants with altered α-amino acid specificity.

Supporting the biocatalyst. Where it is advantageous to support the biocatalyst on or in a solid phase, the following procedures were followed.

Celite/purified enzyme: The purified enzyme (0.6 mg/ml) was taken straight from the eluate of the Procion Red P3BN chromatography column in potassium phosphate buffer (20 mM, pH 7.9) containing NaCl (0.5 M), [Y.K. Seah, K.L. Britton, PJ. Baker, D.W. Rice, Y. Asano, P.C. Engel, FEBS Lett. 1995, 370, 93] without precipitating with ammonium sulphate. 2 ml of this protein solution was added to Celite (1 mg) [M. Persson, E. Wehtje, P. Adlercreutz, Biotechnol. Lett. 2000, 22, 1571]. The preparation was dried for approx. 18 hr under vacuum and then stored at 40C. This form of the biocatalyst was used for all the reactions carried out in presence of organic solvents (see results).

Alginate beads/whole cells: A solution of alginate 2% was prepared by dissolving 100 mg of alginate in 5 mL of H2O under vigorous stirring. A volume of 1 mL of cell paste was added to the alginate solution and the mixture was then polymerised by dropping it with a syringe into a solution of CaCl2 (1 %). The beads form immediately under these conditions. They were left curing for 30 min and stored at 40C.

Chitosan beads/whole cells: A solution of chitosan 2% was prepared by dissolving 100 mg of chitosan in 5 mL of H2O acidified with CH3COOH 1% under vigorous stirring at 30°C. A volume of 1 mL of cell paste was added to the chitosan solution and the mixture was then polymerised by dropping it with a syringe into a solution of tripolyphosphate (0.1 M). The beads form immediately under these conditions. They are left curing for 10 min, rinsed with potassium phosphate (50 mM, pH: 8.0) and stored at 40C. .

Conditions for using the biocatalysts. In selecting conditions for using the biocatalysts, it is important to consider the overall thermodynamics of the reaction and also the particular pH dependence of the biocatalyst. The thermodynamics is represented by an equilibrium constant which is likely to be similar for most α-amino acids and has been accurately determined in the case of the GIuDH reaction [P. C. Engel & K. Dalziel, Biochem. J., 105 (1967) 691-695]. A crucial feature in relation to pH dependence is the fact that a proton is produced in the oxidative deamination (or consumed in the reductive animation). Therefore, all else being equal, higher buffered pH values favour the oxidative deamination, whilst lower values favour reductive amination. However, the performance of the catalyst depends on the ionisation state of catalytic groups in the. active site and therefore the optimum pH range varies slightly from enzyme to enzyme. Broadly, biocatalysts derived from glutamate dehydrogenase, phenylalanine dehydrogenase, leucine dehydrogenase and valine dehydrogenase (for example, PheDH of B. sphaericus) work best at around pH 9.8-11 (optionally about pH 10.4 - 10.6) for oxidative deamination and lower, at around pH 8.5-9.0, for reductive amination.

Recycling system. The cofactor used in the oxidative deamination (for example, NAD+) and in the reductive amination (for example, NADH) can be used in sub- stoichiometric (possibly catalytic) amounts to minimise the cost of the process. Several recycling systems may be used as long as the appropriate additional substrate is added to the reaction mixture. The recycling system adopted must be stable under the same reaction conditions as the mutant amino acid dehydrogenase enzyme. Typically, enzymes used to recycle NAD+/NADP+ to NADH/NADPH are alcohol dehydrogenases from different sources which convert NAD+/NADP+ and ethanol to NADH/NADPH and ethanal [K.F. Gu et al, Biotechnol Appl Biochem. (1990) 3, 227], formate dehydrogenases [H. Groger et al, Org. Lett. (2003) 5, 173] and amino acid dehydrogenases. Enzymes used to recycle NADH/NADPH to NAD+/NADP+ are generally known as NADH oxidases (NOX) which use molecular oxygen to re- oxidise the cofactor [W. Hummel et al, Org. Lett. (2003) 20, 3649]. Similar reactions may be used to recycle other suitable oxidised coenzymes to their corresponding reduced coenzymes, and vice versa.

Monitoring reaction. All the reactions may be monitored by observing the change in light absorbance at 340 nm as a result of the appearance or disappearance of the characteristic absorbance of NADH at this wavelength (This is a standard method for all such dehydrogenases). All coenzymes with an unmodified nicotinamide ring show the absorbance maximum at 340 nm in the reduced form. For other analogues such as acetyl pyridine adenine dinucleotide, the absorbance maximum may be shifted to a different wavelength, but this is easily established and the method accordingly adapted. However, this is not possible when a recycling system is used. For this reason, reaction mixtures, after removal of protein, were analysed by HPLC with appropriate standards applied to allow identification of the peaks in the elution profile.

Quantitative conversion of the α-ketoacid into the corresponding L-α-amino acid without coenzyme recycling. A reaction mixture would be prepared as follows: 1 mmol of α-ketoacid, KCl 1 mmol (this is an essential activator for PheDH), ammonium salt 4mmol, NADH 1.2-4 mmol and Tris buffer pH 8.5 to a final volume of 10 mL (the ammonium salt and NAD+ both need to be in stoichiometric excess over the α-ketoacid to be converted).

Quantitative conversion of the α-ketoacid into the corresponding L-α-amino acid with coenzyme recycling. A reaction mixture was prepared as follows: 1 mmol of α-ketoacid, KCl 1 mmol, ammonium salt 4mmol, NAD+ 20μmol, EDTA 10 μmol, ethanol 0.5mL and Tris buffer pH 8.5 to a final volume of 10 mL (the ammonium salt needs to be in stoichiometric excess over the α-ketoacid_to be converted; the coenzyme is present in a catalytic amount since NADH is recycled, and for the same reason may be supplied either as NADH or as NAD+, the latter being preferable on grounds of cost; EtOH is present at high concentration in order to drive the recycling reaction). The reaction was started by adding lmg alcohol dehydrogenase (ADH) (663 U/mg) and a suitable amount of the chosen biocatalyst (50 μg wild-type PheDH, 60 μg N145A, 10 μg N145V or 30 μg N145L). The reaction mixture was incubated at 25°C in an orbital shaker incubator and the formation of the amino acid was monitored by chiral HPLC (CHIROBIOTIC T column) over a period of 24 hours (the reaction could be performed at any temperature between 15 and 400C depending on the stability of the substrates, higher temperature will lead to an increase rate but may be deleterious for sensitive substrates; if necessary the timecourse of the reaction may be decreased by increasing the amount of biocatalyst added). The same reaction can be performed also by adding beads of chitosan or alginate prepared as described above.

Reactions in homogeneous co-solvents. For these experiments a freshly prepared stock reaction mixture was used consisting of KCl (200 raM), NH4Cl (800 mM), NADH (0.2 mM), phenylpyruvic acid (1 mM), ethanol (5% - all % as used herein are v/v)) and EDTA (1 mM) in Tris (50 mM, pH 8.5). To 3 mL of this solution, different amounts of homogeneous organic co-solvents: THF, CH3CN, acetone and MeOH were added. The reactions were initiated with the enzyme-impregnated Celite (3 mg) and yeast alcohol dehydrogenase (1 mg). Reaction mixtures were magnetically stirred in capped vials at room temperature.

Reactions in biphasic systems. The supported enzyme (3 mg) and yeast alcohol dehydrogenase (1 mg) were added to the aqueous Tris buffer stock solution (3 mL), followed by (20%) of tBuOMe, CH2Cl2, toluene or diethyl ether (0.75 mL), and in this case the stock solution, freshly prepared in Tris buffer (50 mM, pH 8.5), contained: KCl (200 mM), NH4Cl (800 mM), phenylpyruvic acid (1 mM), NADH (0.2 mM), EDTA (1 mM) and 5% EtOH.

Moving to the 80:20 organic: aqueous ratio we changed the stock solution as follows: freshly prepared stock solution in Tris buffer (50 mM, pH 8.5) contained KCl (1600 mM), NH4Cl (3200 mM), phenylpyruvic acid (4 mM), NADH (0.8 mM), EDTA (4 mM) and 5% EtOH. To the Tris stock solution (1.5 mL) the supported enzyme (6 mg), alcohol dehydrogenase from Saccharomyces cerevisiae (2 mg) and (80%) of one of the following solvents ^BuOMe, CH2Cl2, toluene, and diethyl ether (6 mL) were added. To minimize changes in the phase volumes, the organic solvent was previously saturated with the Tris solution before use. Since both solvent ratios gave satisfactory results, it is highly likely that any intermediate convenient ratio would also be suitable.

Resolution of D-α-amino acid. The system requires NAD+ which either must be supplied in stoichiometric excess or else, preferably, is constantly replenished by recycling NADH to NAD+ with bacterial diaphorase from Tliernius aquaticus [C. Logan et al, J Biol Chem. 2000, 275, 30019].

Resolution of D-α-amino acid in the absence of coenzyme recycling. The reaction would be performed at pH 10.5 (carbonate/bicarbonate buffer 0.1 M or similar; glycine buffer is preferably not used in order to avoid subsequent problems in separation). The choice of pH is governed by several considerations; pH optimum for enzyme activity, the equilibrium of the reaction, but also the stability of reaction components. pH 10.5 is good both for activity of this enzyme and for pulling the reaction in the desired direction, but it is less desirable in terms of coenzyme stability. Without recycling one is forced to go for the equilibrium advantage of the higher pH. With recycling, which works much better, one can employ the lower pH which stabilises the coenzyme, still knowing that the reaction will be driven to completion. The reaction mixture would contain 5-10 mM of the DL-α-amino acid, at least a 2- fold molar excess of NAD+, 10OmM KCl. The reaction would be started by adding a suitable amount of the appropriate biocatalyst and monitored either by spectrophotometric measurement of the production of NADH or by chiral HPLC, both as taught above. On page22 and in Table 6,the use of chiral HPLC (the CHIROBIOTIC T) column is mentioned. The solvent system is an isocratic elution with 30% methanol in water, and the same system is used both for following the formation and consumption of amino acids.

Resolution of D-α-amino acid with coenzyme recycling. The reaction is performed at pH 9.5 (ethanolamine-HCl, 20 mL, 50 mM). The reaction mixture contains 5-10 mM of the DL-α-amino acid, 1 mM NAD+, 100 mM KCl and 0.1 mg DCPIP (dichlorophenol indophenol). The reaction is started by adding a suitable amount of the appropriate biocatalyst, 0.1 mg diaphorase and monitored by chiral HPLC as reported above. Although high pHs will favour the equilibrium of the reaction as described above, the stability of NAD+ needs to be considered: for this reason the pH is lowered to a maximum of 9.5. The reaction is carried out in a vessel with a large surface area shaking or stirring to maximise the exchange with molecular oxygen. We carried out the reaction at room temperature (15-200C) but in principle higher temperature up to 30°C could be used.

Separation of the end product. In the processes of the first and second aspects of the invention, the separation of the end product from the reaction mixture can be achieved by ion-exchange chromatography [Y. Asano et al, Agric.Biol. Chem. (1987) 51, 2035], crystallisation or preparative HPLC. In the process of the second aspect of the invention, the separation of the end product from the reaction mixture can additionally be achieved with the use of trapping agents such as functionalised silica.

Synthetic Routes to Substrates

Three general synthetic routes to the α-keto acids were employed. The first route involved preparation through the addition of Grignard reagents to diethyl oxalate followed by hydrolysis of the resulting α-keto ester as illustrated in Scheme 1 below [L. M.Weinstock et al, Synth. Commun., 1981, 11(12), 943-946; and W. J. Middleton & E. M. Bingham, J. Org. Chem., 1980, 45, 2883-2887].

This method was employed for the synthesis of the chloro and fluoro substituted phenylpyruvate derivatives (2)-(6) and for the saturated analogue (12), see Table 2 below.

(14-19) (20-25) (2-6) (12) 1. Mg, Et2O, reflux, 10 min; 2. (EtOCO)2, Et2O, 00C, 2h; 3. AcOH, H2SO4 10%, Δ, Ih. Scheme 1 (The triangle above means delta T, as referring to the supply of heat.) It will be appreciated that the α-keto acids exist predominantly in the enol tautomeric form illustrated in, for example, Scheme 1 above.

Table 2 Synthesis of α-keto acids - Grignard route

a. Yields following chromatography b. Yield following distillation c. Yield following precipitation from EtOAc/DCM

Using a series of benzylic halide derivatives, the α-keto esters (20)-(25) were obtained in excellent yields. The Grignard reagents were prepared in Et2O instead of THF to avoid the Wurtz coupling side reaction [K. V. Baker et al, J. Org. Chem., 1991, 45, 698-703]. While this approach gave the α-keto esters in high yields, the recovery of the acids after hydrolysis was poor, which can be attributed to a number of side reactions. Decarboxylation can occur during hydrolysis at high temperature or in basic media during the extraction process. Rapid decomposition can occur in polar solvents during concentration of the organic layer at reduced pressure giving an unidentifiable mixture by NMR. However, the acids can be isolated in high purity by precipitation from an ethyl acetate/dichloromethane mixture. 1H and 13C NMR spectra in CD3OD show that only the enol form is present for (2)-(6) [T. Takai et al, Spectroscopy Lett., 1998, 31(2), 379-395]. This route also led successfully to the cyclohexyl keto acid (12), the saturated analogue of phenylalanine. In this case only the keto form is observed in d6-DMSO.

Preparation of α-keto-acids and their esters

3-(2-Chlorophenyl)-2-oxo-propionic acid (1) [H. Hongwen & F. Xianqi, Gaodeng Xuexiao Huaxue Xuebao, 1988, 9, 966-968]

To a suspension of Mg turnings (0.51 g, 20.9 mmol, 1.1 eq) in Et2O (3 mL) was added 1 mL of a solution of 2- chlorobenzyl chloride (2.50 mL, 19.0 mmol, 1 eq) in

Et2O (17 mL). The reaction was initiated by heating and the rest of the halide solution was added dropwise at a rate that maintains reflux. The mixture was stirred for 10 min, cooled to RT and CdCl2 (1.83 g, 10.0 mmol, 0.5 eq) added portion wise. The suspension was stirred at reflux for 40 min, cooled to RT and transferred to a dropping funnel. This was added dropwise to a solution of oxalyl chloride (4.30 mL, 38.5 mmol, 2 eq) in Et2O (18 mL) previously cooled to 0°C. After stirring for 1 h at 0°C, the reaction was quenched with ice (approx. 5OmL). The aqueous layer was extracted twice with Et2O (2 x 5OmL) and the combined organic layers were washed with brine (50 mL), dried with MgSO4 and finally concentrated under reduced pressure. Precipitation of the residue (EtOAc:DCM«l : 19) afforded the title compound as a pale yellow solid (1.2Og, 32%). 1H NMR δ (CD3OD, ppm) 6.90 (IH, s, H-3), 7.17-7.29 (2H, m, Ar-H), 7.38-7.41 (IH, dd, J= 1.4 Hz, 7.8 Hz, Ar-H), 8.33-8.37 (IH, dd, J= 1.4 Hz, 7.8 Hz, Ar-H); 13C NMR δ (CD3OD, ppm) 106.55 (CH-3), 128.52, 130.18, 131.09, 132.96 (4 x CH), 134.80, 134.98 (2 x C), 144.63 (C- 2), 168.78 (C-I); vmax /cm"1 (KBr) 3416 and 3200-2500 (OH), 1682 (C=O).

3-(3-Chlorophenyl)-2-oxo-propionic acid ethyl ester (20) [H. Yinglin & H. Hongwen, Gaodeng Xuexiao Huaxue Xuebao, 1991, 12(5), 617-620].

To a suspension of Mg turnings (1.02 g, 42.0 mmol, 1.1 eq) in Et2O (5 mL) was added 2 mL of a

solution of 3-chlorobenzyl chloride (5.00 mL, 38.0 mmol, 1 eq) in Et2O (35 mL). The reaction was initiated by heating and the rest of the halide solution was added dropwise at a rate that maintains reflux. The mixture was stirred for 10 min, cooled to RT and then transferred into a dropping funnel. This was added dropwise to a solution of diethyl oxalate (10.00 mL, 76.0 mmol, 2 eq) in Et2O (75 mL) previously cooled to 0°C. After stirring for 2 h at RT, the reaction was quenched with aq. NH4Cl (IM, 100 mL). The aqueous layer was extracted twice with Et2O (2 x 100 mL) and the combined organic layers were washed with brine (100 mL), dried with MgSO4 and finally concentrated under reduced pressure, maintaining the temperature below 25°C. The excess diethyl oxalate was removed with bulb to bulb distillation and the residue was purified by flash chromatography (hexane:EtOAc=8:2) to yield the title compound as a colourless oil (7.75g, 90%). 1H NMR δ (CDCl3, ppm) 1.35 (3H, t, J= 7.1 Hz, CH3), 4.33 (2H, q, J= 7.1 Hz, CH2), 6.43 (IH, s, H-3), 6.82 (IH, bs, OH), 7.12-7.98 (4H, m, Ar-H).

The procedure described above for (20) was employed for the synthesis of each of the esters (21)-(25) using the appropriate benzyl or alkyl halide in each case. Purification by flash chromatography using hexane:EtOAc (8:2) gave the pure esters.

3-(4-Chloropheny])-2-oxo-propionic acid ethyl ester (21) This was prepared as described for (20) using A-

chlorobenzyl chloride to give the title compound as

a colourless oil (87%). 1H NMR a (CDCl35 PPm) 1.35 (3H, t, J = 7.1 Hz, CH3), 4.33 (2H, q, J = 7.1 Hz, CH2), 6.43 (IH, s, H-3), 6.82 (IH, bs, OH), 7.29 (2H, J = 8.8 Hz, HA part of ABq, Ar-H), 7.64 (2H, J = 8.8 Hz5 HB part of ABq, Ar-H).

3-(2-FluorophenyI)-2-oxo-propionic acid ethyl ester (22)

as a

5

(3H, t, J = 7.1 Hz, CH3), 4.33 (2H, q, J = 7.1 Hz, CH2), 6.66 (IH, s, H-3), 6.79 (IH, bs, OH), 6.95-7.37 (3H, m, Ar-H), 8.22-8.24 (IH, m, Ar-H).

3-(3-Fluorophenyl)-2-oxo-propionic acid ethyl ester (23) [K. Hirai et al, Patent no. JP 06157488

1994]. This was prepared as described for (20)

using 3-fluorobenzyl chloride to give the title compound as a colourless oil (92%). 1H NMR δ (CDCl3, ppm) 1.36 (3H, t, J - 7.1 Hz, CH3), 4.34 (2H, q, J = 7.1 Hz, CH2), 6.48 (IH, s, H-3), 6.77 (IH, bs, OH), 6.95- 6.99 (IH, m, Ar-H), 7.29-7.33 (IH, m, Ar-H), 7.42-7.44 (IH, m, Ar-H), 7.61-7.80 (IH, m, Ar-H).

3-(4-FIuorophenyl)-2-oxo-propionic acid ethyl ester (24) [H. F. Schuster & G. M. Coppola, Journal of Heterocyclic Chem., 1994, 31(6), 1381- 1384]. This was prepared as described for (20) using 4- fluorobenzyl chloride to give the title compound as

a colourless oil (87%). 1H NMR δ (CDCl3, ppm)

1.37 (3H, t, J = 7.1 Hz, CH3), 4.35 (2H, q, J = 7.1 Hz, CH2), 6.48 (IH, s, H-3), 6.64 (IH, bs, OH), 7.01-7.07 (2H, m, Ar-H), 7.71-7.78 (2H, m, Ar-H).

3-Cyclohexyl-2-oxo-propionic acid ethyl ester (25) [S. Brandange et al, Acta. Chem. Scand.,1913, 27(10), 3668-3672].

This was prepared as described for (20) using bromomethyl cyclohexane to give the title compound as a colourless oil (81%). 1H NMR δ (CDCl3, ppm) 0.64-

1.34 (5H, m, ring CH2), 1.39 (3H, t, J= 7.1 Hz, CH3), 1.66-1.72 (5H, m, ring CH2), 1.90-1.98 (IH, m, CH), 2.70 (2H, d, J= 7.1 Hz, CH2), 4.27 (2H, q, J= 7.1 Hz, CH2).

Preparation of α keto acids from their esters 3-(3-Chlorophenyl)-2-oxo-propionic acid (2) [H. Yinglin & H. Hongwen, Gaodeng Xuexiao Huaxue Xuebao, 1991 , 12(5), 617-620].

A solution of (20) (7.00 g, 30.1 mmol) in AcOH (20 niL) and 10% aq. H2SO4 (20 mL) was stirred at reflux for 1 hour, under a Dean-Stark apparatus.

The mixture was cooled to RT, diluted with water (20 mL) and extracted with Et2O (3 x 50 mL). The combined organic layers were extracted twice with aq. NaHCO3 (2 x 100 mL). The combined aqueous layers were acidified to pH 1 and extracted three times with Et2O (3 x 100 mL). The combined organic layers were washed with brine, dried with MgSO4 and finally concentrated under reduced pressure, keeping carefully the temperature below 25°C. Precipitation of the residue (EtOAc:DCM∞l :19) afforded the title compound as a white solid (2.67g, 45%). 1H NMR δ (CD3OD, ppm) 6.43 (IH, s, H-3), 7.19-7.32 (2H, m, Ar-H), 7.57-7.59 (IH, dd, J= 7.5 Hz, 1.2 Hz, Ar-H), 7.89 (IH, t, J= 1.8 Hz, Ar-H); 13C NMR δ (CD3OD, ppm) 110.02 (CH-3), 128.46, 129.37, 130.51, 131.07 (4 x CH)5 136.28, 139.12 (2 x C), 144.78 (C-2), 162.03 (C-I); vmax /cm"1 (KBr) 3459 and 3200-2500 (OH), 1676 (C-O).

α-Keto acids (3)-(6), (12) were prepared following the procedure described for (2). Precipitation using EtOAc:DCM 1:19 was used to purify the acids in each case.

3-(4-Chlorophenyl)-2-oxo-propionic acid (3) [H. Yinglin & H. Hongwen, Gaodeng Xuexiao Huaxue Xuebao, 1991, 12(5), 617-620]. give acid (3) as a δ (CD3OD, ppm) 6.44 8.8 Hz, HA part of ABq,

HB part of ABq, Ar- H); 13C NMR δ (CD3OD, ppm) 110.68 (CH-3), 130.15, 132.85 (2 x CH), 133.17, 135.30 (2 x C), 143.98 (C-2), 168.83 (C-I); vmax /cm"1 (KBr) 3466 and 3200-2500 (OH), 1667 (C=O). 3-(2-Fluorophenyl)-2-oxo-propionic acid (4) [M. Tanaka & K. Otsuka, Patent no. JP63048244 1988].

Ester (22) was hydrolysed to give acid (4) as a white solid (42%). 1H NMR δ (CD3OD, ppm) 6.68 (IH, s, H- 3), 7.02-7.35 (3H, m, Ar-H), 8.28-8.32 (IH, m, Ar-H);

13C NMR δ (CD3OD, ppm) 102.20 (d, 3JC,F = 8.0 Hz, C- 3), 116.49 (d, 2Jc1F = 22.4 Hz, C-3'), 125.87 (d, 3JQF =3.5 Hz, C-6'), 130.65 (d, 3Jc1F = 8.0 Hz, C-4'), 132.75 (C-51), 146.59 (C-2), 166.17 (d, 1Jc1F = 283.0 Hz, C-21), 168.64 (C-I); vmax /cm"1 (KBr) 3478 and 3200-2500 (OH), 1694 (C=O).

3-(3-Fluorophenyl)-2-oxo-propionic acid (5) [M. Tanaka & K. Otsuka, Patent no. JP63048244 1988] Ester (23) was hydrolysed to give acid (5) as a white solid (27%). 1H NMR δ (CD3OD, ppm) 6.46 (IH, s, H-3), 6.94-7.04 (IH, m, Ar-H), 7.27-7.35

(2H, m, Ar-H), 7.43 (IH, d, J= 7.8 Hz, Ar-H); 13C NMR δ (CD3OD, ppm) 110.36 (CH-3), 115.26 (d, 2JQF = 21.8 Hz, C-4'), 117.09 (d, 2JQF = 23.0 Hz, C-2'), 127.02 (C-6'), 131.15 (d, 3JQF = 8.4 Hz, C-5'), 139.43 (d, 3JQF = 8.4 Hz, C-I '), 144.78 (C-2), 167.56 (C-I), C-F not detected; vmax /cm"1 (KBr) 3477 and 3200-2500 (OH), 1698 (C-O).

3-(4-Fluorophenyl)-2-oxo-propionic acid (6) [M. Tanaka & K. Otsuka, Patent no. JP63048244 1988]

Ester (24) was hydrolysed to give acid (6) as a white solid (39%). 1H NMR δ (CD3OD, ppm) 6.47 (IH, s, H-3), 6.98-7.08 (2H, m, Ar-H), 7.76-7.82

(2H, m, Ar-H); 13C NMR δ (CD3OD, ppm) 111.07 (CH-3), 116.78 (d, 2JC,F = 21.8 HZ, 2 X C-3'), 133.39 (d, 3JC,F = 8.0 Hz, 2 x C-2'), 132.90 (C-I'), 142.88 (C-2), 162.52 (d, 1Jc1F = 245.5 Hz, C-4'), 169.02 (C-I); vmax /cm"1 (KBr) 3476 and 3200-2500 (OH), 1699 (C=O). S-Cyclohexyl-^-oxo-propionic acid (12) [D.H.R. Barton et al, Tetrahedron, 1995, 51(7), 1867-1886]

bs, OH); 13C NMR δ (CD3OD, ppm) 26.30, 26.36, 33.82 (3 x CH2), 33.96 (CH), 45.49 ( (CCHH22)),, 1161.40 (C-I), 195.80 (C-2); vmax /cm"1 (KBr) 3428 and 3200-2500 (OH), 1694 (C=O).

The Grignard route mentioned above was less successful for the preparation of compounds containing electron-donating aryl substituents such as methyl and methoxy, as the benzyl Grignard reagents bearing these substituents were seen to undergo extensive Wurtz coupling. For these derivatives (7)-(ll), a second method was employed via the azlactone (Scheme 2 and Table 3), a route reported to provide an efficient synthesis of arylpyruvic acids containing electron-donating groups [H.N.C. Wong et al, Synthesis, 1992, 793-797]. In this synthesis, the starting aldehyde is condensed with iV-acetylglycine in the presence of sodium acetate in acetic anhydride. The resulting azlactone is then hydrolysed to yield the α-keto acid with 3M hydrochloric acid.

(26-30) (31-35) (7-11)

1. N-acetylglycine, AcONa, Ac2O, Δ Ih; 2. 3M HCl, Δ, 3h.

Scheme 2

Table 3 Synthesis of α-keto acids - Azlactone route

While the literature reports the condensation under reflux overnight, in this work the reaction was found to reach completion within 1 hour at reflux, monitoring by TLC. The reaction was then quenched with ice and filtration of the resulting precipitate afforded the desired azlactones as yellow solids in good yields and purity [H.N.C. Wong et al, Synthesis, 1992, 793-797; and V. DaIa et al, Tetrahedron Lett, 1997, 38 (9), 1577-1580].

The reaction time for the hydrolysis step was also reduced from a reported 24-48 hours to typically 3 hours again monitoring by TLC. Filtration of the resultant precipitate gave the α-keto acids in high yields and purity. Again 1H and 13C NMR showed that only the enol form of each compound (7)-(ll) was present in CD3OD or d6-DMSO.

Preparation of Azlactones

2-Methyl-4-[4-(methyI)benzylidene]-5(40)-oxazolone (31) [H. Hoshina et al, Heterocycles, 2000 53(10), 2661-2274] A mixture of jo-tolualdehyde (5.00 mL, 42.4 mmol, 1 eq), N-acetyl-glycine (6.45 g, 55.1 mmol, 1.3 eq) and sodium acetate (4.50 g, 55.1 mmol, 1.3 eq) in acetic anhydride (19.00 mL, 0.2 mol, 5 eq) was stirred at reflux for 1 h. The reaction was quenched with ice (approx. 50 mL) and vigorously stirred for 1 h in an ice bath to allow precipitation. Filtration afforded the title compound as a yellow solid (6.14g, 72%). 1H NMR δ (CD3OD, ppm) 1.90 (3H, s, CH3), 2.23 (3H, s, CH3-Ph), 7.23 (2H, HA part of ABq, J = 8.8 Hz, Ar-H), 7.48 (IH, s, H-3), 7.49 (2H, HB part.of ABq, J - 8.8 Hz, Ar-H); 13C NMR δ (CD3OD, ppm) 21.84 (CH3), 22.97 (CH3-Ph), 126.51 (C), 130.77, 131.41 (2 x CH), 132.57 (C), 136.25 (CH-3), 141.65 (C), 168.81 (C=N), 173.60 (C=O); vmax /cm"1 (KBr) 1794, 1773, 1667.

The procedure described above for (31) was employed for the synthesis of each of the azlactones (32)-(35) using the appropriate benzaldehyde in each case. Filtration afforded the pure azlactones in each case.

2-Methyl-4-[4-(methoxy)benzylidene]-5(4//)-oxazoIone (32) [H. Hoshina et al, Heterocycles, 2000 53(10), 2661-2274]

This was prepared as described for (31) using 4- methoxybenzaldehyde to give the title compound as a yellow solid (78%). 1H NMR δ (CDCl3, ppm) 2.39 (3H, s, CH3), 3.87 (3H, s, OCH3), 6.96 (2H, HA part of ABq, J = 7.1 Hz, Ar-H), 7.02 (IH, s, H-3), 7.49 (2H, HB part of ABq, J = 7.1 Hz, Ar-H); δ (CDCl3, ppm) 16.03 (CH3), 55.84 (OCH3), 114.84 (2 x CH), 126.53 (C), 130.78 (2 x CH), 131.91 (CH-3), 134.65 (C), 162.45 (C), 165.28 (C=N), 168.59 (C=O); vmax /cm"1 (KBr) 1798, 1775, 1665.

2-Methyl-4-[4-(trifluoromethyl)benzyHdene]-5(4ir)-oxa2olo ne (33) [H. Hoshina et al, Heterocycles, 2000 53(10), 2661-2274] This was prepared as described for (31) using 4- trifluoromethylbenzaldehyde to give the title compound as a yellow solid (81%). 1H NMR δ (CDCl3, ppm) 2.43 (3H, s, CH3), 7.13 (IH, s, H-3), 7.81 (2H, HA part of ABq, J = 7.1 Hz, Ar-H), 8.06 (2H, HB part of ABq, J = 7.1 Hz, Ar-H); vfflax /cm"1 (KBr) 1797, 1775, 1663.

2-Methyl-4-(4-pyridyI)-5(417)-oxazo]one (34) [K. Yamabe et al, Sasebo Kogyo Koto Senmon gakko Kenkyu Hokoku, 1988, 25, 1 19-122] This was prepared as described for (31) using 4- pyridinecarboxaldehyde to give the title compound as a grey solid (68%). 1H NMR δ (CDCl3, ppm) 2.45 (3H, s, CH3), 7.03 (IH, s, H-3), 7.98 (2H5 HA part of ABq, J= 6.2 Hz, ArH), 8.80 (2H, HB part of ABq, J= 6.2 Hz, ArH); vmax /cm-] (KBr) 1793, 1771,1662.

2-Methyl-4-(2-thienyl)-5(4£T)-oxazolone (35) [J. Meiwes et al, Tetrahedron: Asymmetry, 1997, 8(4), 527- 536] This was prepared as described for (31) using 2- thiophenecarboxaldehyde to give the title compound as a yellow solid (79%). 1H NMR δ (CDCl3, ppm) 2.41 (3H, s, CH3), 7.12-7.15 (IH, m, Ar-H), 7.38 (IH, s, H-3), 7.56 (IH, d, J = 3.6 Hz, Ar-H), 7.68 (IH, d, J = 5.1 Hz, Ar-H); v^ /cm"1 (KBr) 1793, 1779, 1665.

Preparation of α-keto acids

3-(4-Methylphenyl)-2-oxo-propionic acid (7) [K. Yamabe et al, Sasebo Kogyo Koto Senmon gakko Kenkyu Hokoku, 1988, 25, 119-122] A suspension of (31) (2.60 g, 12.9 rnmol) in aq. HCl (3M, 10 mL) was stirred at reflux for 3 h. The mixture was cooled to RT to allow crystallisation.

Filtration afforded the title compound as an orange solid (2.19g, 95%). 1H NMR δ (CD3OD, ppm) 2.34 (3H, s, CH3), 6.47 (IH, s, H-3), 7.14 (2H, HA part of ABq, J = 8.1 Hz5 ArH), 7.65 (2H, H3 part of ABq5 J = 8.1 Hz, ArH); 13C NMR δ (CD3OD, ppm) 21.76 (CH3), 112.16 (CH-3), 130.32, 131.17 (2 x CH), 133.96 (C), 138.86 (C), 141.98 (C-2), 168.90 (C-I); vmax /cm"1 (KBr) 3478 and 3200-2500 (OH), 1669 (C=O).

α-Keto acids (8)-(ll) were prepared following the procedure described for (7) Filtration gave the acids in each case. 3-(4-Methoxyphenyl)-2-oxo-propionic acid (8) [H. Yinglin & H. Hongwen, Gaodeng Xuexiao Huaxue Xuebao, 1991, 12(5),

was hydrolysed to give acid (8) lid (98%). 1H NMR δ (CD3OD,

s, OCH3), 6.48 (IH, s, H-3), 6.96 (2H, HA part of ABq, J = 7.1 Hz, ArH), 7.49 (2H, HB part of ABq, J = 7.1 Hz, ArH); 13CNMRδ(CD3OD,ppm)56.05(CH3), 112.15(CH-3),115.10(2xCH), 129.47 (C), 132.65(2xCH), 140.98(C-2), 160.91 (C), 169.00(C-I);vmax/cm"1 (KBr)3458 and3200-2500(OH), 1662(C=O).

3-(4-Trifluoromethylphenyl)-2-oxo-propionic acid (9) [Y. Ueno et al, Patent no. JP 10265464 1998] (33) was hydrolysed to give acid (9) as solid (93%). 1H NMR δ (d6-DMSO, (IH, s, H-3), 7.69 (2H, HA part of ABq,

ArH), 7.95 (2H, HB part of ABq, J - 8.1 Hz, ArH), 9.80 (IH, bs, OH); vmax /cm"1 (KBr) 3461 and 3200-2500 (OH), 1667 (C=O).

3-(4-Pyridyl)-2-oxo-propionic acid (10) [K. Yamabe et on gakko Kenkyu Hokoku,

ed to give acid (10) as a red DMSO, ppm) 6.57 (lH, s, H-3), 8.24 (2H, HA part of ABq, J = 6.7 Hz, ArH), 8.75 (2H, HB part of ABq, J = 6.7 Hz, ArH); 13C NMR δ (d6-DMSO, ppm) 103.71, 125.33, 141.03 (3 x CH), 152.08 (C), 152.36(C-2) 164.93 (C-I); vmax /cm"1 (KBr) 3449 and 3200-2500 (OH), 1660 (C=O).

3-(2-Thienyl)-2-oxo-propionic acid (11) [J. Meiwes et al, Tetrahedron: Asymmetry, 1997, 8(4), 527- 536] Azlactone (35) was hydrolysed to give acid (11) as a grey-green solid (92%). 1H NMR δ (d6-DMSO, ppm) 6.74 (IH, s, H-3), 6.99-7.02 (IH3 m, ArH), 7.22 (IH, d, J - 3.5 Hz, ArH). 7.50 (IH, d, J = 5.7 Hz, ArH), 9.49 (IH, bs, OH); 13C NMR δ (d6-DMSO, ppm) 105.93, 127.10, 128.18, 128.27 (4 x CH), 137.80 (C), 140.00 (C-2) 166.03 (C-I); vmax /cm"1 (KBr) 3458 and 3200-2500 (OH), 1664 (C=O).

The 2-chlorophenylpyruvate derivative (1) was synthesized as illustrated in Scheme 3. Based on earlier work describing acylation of organocadmium reagents with acid chlorides [J. Cason, Chem Rev., 1947, 40, 15-32; and D.A. Shirley, Org. Reactions, 1954, 8, 28-58], reaction of an organocadmium reagent with oxalyl chloride followed by hydrolysis was envisaged as a short synthetic route to the α-keto acid. However, while this route did produce the α-keto acid (1), this synthetic method is less satisfactory than the routes described above so it was not investigated further.

(iv) H2O, O0C

Scheme 3

2-Ketocaproic acid (13) is commercially available from Sigma and was included in this study as an acyclic derivative.

Evaluation of enzyme activity - UV assays

Each α-keto acid (4 mM) was dissolved to form a component of a reaction mixture containing NH4Cl (40OmM), KCl (10OmM), O.lmM NADH and Tris (5OmM). The pH was adjusted to 8.0 by adding a suitable amount of HCl. ImL of reaction mixture was incubated at 25°C. The reaction was followed at 340nm over 1 minute after adding an appropriate amount of enzyme to achieve an optimally measurable reaction rate (between 0.01-0.03 min"1). Each of the reactions was carried out in duplicate and the average value is reported. Example 1: Specificity for non-natural substrate resulting from a set of mutations created at position 145 of PheDH.

We focused in particular on mutations of the Asn in position 145 (N145) in which the asparagine has been substituted by an alanine (N145A), a leucine (N145L) or a valine (N 145Y) which resulted in an increased discrimination between phenylalanine and tyrosine [S.Y. Seah et al, Biochemistry 2002, 41, 11390-11397].

With the objective of synthesizing a series of non-natural analogues to phenylalanine from the appropriate α-keto acids through use of the engineered PheDH mutants, a series of substrates structurally related to phenyl pyruvate was selected. The compounds (1-13) were chosen for this study, including aromatic, aliphatic and heteroaromatic α-keto acids as precursors for the non-natural α-amino acids. Y = 2-Cl (1) 33--CCll (2)

4-Me (7) 4-OMe (8) 4-CF3 (9)

CH3(CH2)3COCO2H (13) These compounds were chosen to allow exploration of substrate scope in the modified active site of the enzyme. Both steric and electronic factors have been explored through this series. α-Keto acids, especially analogues of the naturally occurring α-amino acids are of major importance in intermediary metabolism. They have been used in the therapy of certain conditions such as uraemia and nitrogen accumulation disorders. They are also of interest as intermediates in chemical synthesis, in the development of enzyme inhibitors and drugs (and here the availability of non-natural α-amino acid analogues is of particular importance), as model substrates of enzymes and in other ways. A general review was published in 1983 [A. JX. Cooper et al, Chem. Rev., 1983, 83, 321-358].

Biocatalysis

The non-natural α-keto acids were initially screened for activity in the reductive amination of the α-keto acid with the wild type PheDH and three different mutants (Nl 45 A, N145L and Nl 45V). The results are reported in Table 4. The activity is measured under standard conditions at 250C [H. Hongwen & F. Xianqi, Gaodeng Xuexiao Huaxue Xuebao, 1988, 9, 966-968], by following the decrease in absorbance of NADH at 340nm (UV spectrophotometer, Gary 50) according to the reaction in Scheme 4. For convenience, each of the results in Table 4 is expressed both as a specific activity (a unit, U, of enzyme is defined as the amount of enzyme which converts one μmole of substrate per minute under standard conditions) and also normalized relative to activity with phenylpyruvate for each of the enzymes.

+ Enzyme + Enzyme

Scheme 4

These mutants were created with the aim of producing a less polar environment in the vicinity of the 4-position of the phenyl group of the substrate phenylalanine. These mutants were therefore tested for their acceptance of a variety of substrates with substitutions at position 4. Halogen substituents at positions 2 and 3 were also tested and the effect of replacing the phenyl ring with other rings, such as pyridine or cyclohexane was explored. Table 4 shows the results of this investigation carried out for the reductive amination reaction. In many cases substrates very poorly accepted by the wild-type precursor enzyme are good substrates for the new biocatalysts, and some of these compounds are also better substrates than the original phenylpyruvate. The pyridine and cyclohexane rings, but not thiophene, are well accepted. Halogen substitutions at positions 2 and 3 give lower rates than at position 4 (This set of biocatalysts was not specifically designed for substituents at positions 2 and 3), but nevertheless the rates of reaction are quite sufficient to make these biocatalysts suitable for handling these substrates in a production process.

Table 4. Activity in Reductive Amination of α-Keto Acids with WT (Wild Type)

PheDH and Engineered Mutants. The upper number is a relative activity with respect

to phenylpyruvate, while the number in brackets is the specific activity.

Example 2: Non-natural substrate specificity by a series of mutants, single and

double, at various positions in the PheDH sequence.

These mutants have been made with a view to creating more space in the binding

pocket and Table 5 shows that there is a very substantial enhancement of activity with

both neopentylglycine and phenylglycine as compared to the wild type enzyme.

Neopentylglycine Phenylglycine

Table 5. Screen of single and double mutants for increased activity towards non- proteinogenic α-amino acids

Example 3: Enantioselectivitv.

Determination of enantioselectivity - chiral HPLC conditions

Each α-keto acid (~0.5mM) was dissolved in a solution containing NH4Cl (40OmM), KCl (10OmM), ImM NADH and Tris (5OmM). The pH was adjusted to 8.0 by adding a suitable amount of HCl. The solution was filtered through a sterile filter Acrodisc® 0.45 μm. ImL of reaction mixture was incubated at 25°C and an appropriate amount of enzyme (N 145 V, N 145 A or N145L - each gave similar results) was added to allow approximate completion of the reaction within about 40min. The formation of the corresponding α-amino acid was followed by loading 20 μL of the mixture onto a CHIROBIOTIC T, chiral HPLC column. The elution mixture was MeOH/H2O 70/30 at a flow rate of 1.OmL/min.

Table 6: Retention times (min) of the α-amino acids on a Chirobiotic T (MeOH/H2O 70/30, flow rate of 1.OmL/min)

A critical issue in assessing the utility of these biocatalysts is the extent to which they retain the absolute chiral selectivity of the parent enzyme. To test the enantioselectivity of some of these reductive animations, a new experiment was set up for each α-keto acid substituted in position 4 of the aryl ring and also for the pyridine substrate (10): working with excess of NADH (1 mM) and with a suitable amount of enzyme, a reaction mixture containing 0.5 mM α-keto acid was taken to completion and the crude mixture was then loaded onto a chiral HPLC column. Each enzyme was tested with all the substrates. The D-α-amino acid was never detected, while the L-α-ainino acid was clearly identified in all cases. Figure 1 shows an example of HPLC outcome, in which all the peaks are clearly separated and the L enantiomer is the only one detectable. With the PheDH derivatives in particular this has been exhaustively tested by monitoring reaction mixtures by chiral HPLC analysis. In no case was there any detectable production of the D-isomer of the α-amino acid (Fig. — l)τ Fig. IA shows that chiral HPLC gives a good separation of the L- and D- enantiomers of the α-amino acid and separates both from other, early eluting components of a typical reaction mixture. Fig. IB shows that when the reaction is carried out using the α-ketoacid as the starting substrate, a clean peak is seen at the position corresponding to the L-α— amino acid product and there is no trace of a similar peak in the D-α-position.

Some α-amino acids have more than one chiral centre. All except glycine exhibit chirality at the α-carbon atom, but others, such as threonine and isoleucine, contain a second chiral atom in their sidechain. The ability to distinguish the enantiomers at these positions will also be a very valuable property of these biocatalysts.

Quantitative conversion. This has been carefully demonstrated in the direction of reductive amination with the PheDH-derived mutants and using recycling of cofactor with alcohol dehydrogenase as described above. Three non-natural α-amino acids were successfully synthesised via the enzymatic route. ADH (alcohol dehydrogenase from Saccharomyces cerevisiae) was chosen among the possible enzymes available to recycle the cofactor NADH because it presented significant advantages: with non- natural substrates, solubility may be a problem and the ethanol added to the reaction mixture acts both as a substrate for ADH and as a co-solvent. ADH is stable and active under our reaction conditions and furthermore is inexpensive. The choice of the optimal catalyst was based on initial screening of wild type PheDH and the mutants. N145A offered the best affinity for 4-CF3-phenylpyruvate (9), while N145V and N145L represented the best choice for 4-F-phenylpyruvate (6) and 4-MeO- phenylpyruvate (8). The reaction mixture analysed by HPLC and NMR confirmed a conversion >99%. Figure 2 is the NMR spectrum of a reaction mixture after 24hrs for synthesis of the 4- F-ρhenylalanine. This spectrum is very similar to that reported in the literature for the authentic compound. The peaks corresponding to the aromatic hydrogens and the single hydrogen attached to the chiral carbon atom are clearly seen.

Form of the biocatalvst. It is possible to envisage using the biocatalyst a) as the substantially purified protein; b) as a crude extract without purification; c) as whole cells without breakage or purification; d) as crude extract or as substantially purified protein immobilised on a solid support; e) as crude extract or as substantially purified protein or cells entrapped in (or cross-linked to) sol-gels (such as alginate, chitosan or agarose beads). Thus far a), c) and d) and e) have been carefully documented. The possibility of using the crude extract without purification arises because the biocatalyst is so highly over-expressed in the bacterial cells that it comprises typically 25-40% of total soluble protein. Whole cells with externally added substrates function very satisfactorily without any apparent problems resulting from permeability barriers. The PheDH-derived biocatalysts were also very effective and stable after absorption onto Celite. Either of these modes of use allow very simple physical separation of the biocatalyst from the reaction mixture at the end of reaction, allowing repeated use of the biocatalyst preparation.

Organic solvents. A potential limitation to the use of biocatalysts is the fact that they are generally used in aqueous solution, whereas some of the potentially interesting substrates are poorly, if at all, soluble in water. The results presented above in relation to recycling the cofactor already indicate, however, that the purified biocatalysts in solution tolerate 5-10% ethanol. Thus use of an organic solvent offers a promising route for extending the utility of the biocatalysts that form the subject of present invention.

Example 4: Organic solvents

The use of organic solvents has been more fully explored in the case of immobilised biocatalyst on a Celite support. The results, summarised in Tables A and B, lead to two important conclusions. Firstly, miscible organic solvents such as methanol and

acetone may be added to the reaction mixture without destroying the activity. Some

of the reaction timecourses clearly showed that activity was retained over a period of

days under these conditions. Secondly, a striking result was obtained with nonpolar

organic solvents such as hexane or diethyl ether. These are immiscible with water

and therefore form a two-phase system when added to an aqueous solution. It was

found that high percentage conversion of α-ketoacid to L-α-amino acid (and vice

versa) could be achieved in such a 2-phase system when the biocatalyst (Nl 45A) and

coenzyme were in the aqueous phase but the α-ketoacid (substrate of the second

aspect or product of the first aspect) was dissolved in a water-immiscible phase. The

water-immiscible phase appears to serve as a reservoir feeding α-ketoacid to the

aqueous phase. In the reverse reaction, the amino acid would remain predominantly

in the aqueous phase, but the poorly water soluble α-ketoacid product would be

continuously extracted into the organic phase. This makes it possible to use the

biocatalyst on substrates with poor water solubility. Under these conditions the

effectiveness of recycling with alcohol dehydrogenase was demonstrated, making it

possible again to use sub-stoichiometric amounts of coenzyme.

Table 7. Conversion % of phenylpyruvic acid into L-phenylalanine using Celite- supported enzyme in homogeneous system of aqueous Tris buffer solution (pH 8.5) with different co-solvents and co-factor regeneration.

Table 8. Conversion % of phenylpyruvic acid into L-phenylalanine with co-factor regeneration using Celite-supported enzyme in biphasic systems of aqueous Tris buffer solution (pH 8.5), 20% and 80% of different co-solvents.

As explained in experimental section, 20% hexane means 20:80 hexane:aqueous buffer (v/v). AU solvents in table 8 are immiscible, all solvents in table 7 are miscible.

Example 5 : Substrate Specificity of Oxidative Deamination Reaction for PheDH

Given the excellent enantioselectivity of PheDHs, it is possible to use these catalysts to separate the D and L enantiomers very efficiently via a kinetic resolution process.

Tables 9 and 10 represent preliminary data for the oxidative deamination reaction of WT PheDH and mutants with aromatic substrate analogues of phenylalanine (Phe) substituted at position 4 (36-40 and 44-48), with increased chain length (L- homophenylalanine 43) or substituted at the β-carbon (β-methyl-DL-Phe 51) and aliphatic α-amino acids corresponding to the α-keto -acids 12 (L-cyclohexylalanine 42 and DL-Cyclohexyl-alanine 50) and 13 (L-norleucine 41 and DL-norleucine 49) reported in Table 4 or with a smaller aliphatic ring (DL- cyclopentylalanine 52). These reactions are exceptionally sensitive to pH and the experiments were performed at pH 10.4 (The data given herein refer to this specific pH and could change quite sharply if someone used a slightly different pH. It is not intended to imply that it will not work at a slightly different pH.) and at 25 °C. Due to the low concentration of the substrates (0.2 mM in case of pure L-α-amino acids and 0.4 raM for the racemic . mixtures, unless otherwise indicated), consistent through the tables, we are possibly observing the activity of the enzymes under non-optimal conditions (i.e., the specific activity of WT PheDH with L-phenylalanine is 7.5 U/mg, lower than the mutants, and is explained by its higher Km [Biochemistiy 2002, 41, 11390-11397]). To be consistent with the data reported above for the reverse reaction, both relative

activities and absolute values are included in the same boxes.

4-NO2 (40)

The data in Table 10 concerning the racemic mixture of α-amino acids clearly show

that the D enantiomer is invariably inhibitory (all the relative activities are referred to

the corresponding pure L-α-amino acid and, if the D-α-amino acid were a substrate,

the values would be higher than 100).

Table 9: Preliminary results for activity of WT and mutants at 0.2mM concentration of substrate (* 2.5 mM). Assays are performed at pH 10.4 in Gly/NaOH buffer by following spectrophotometrically the production of NADH at 340nm. In this table we are comparing relative activities (top numbers) and specific activities (bottom numbers) with pure L-α-amino acids.

4-Cl (45) 4-Me (46) 4-OMe (47) 4-NO2 (48)

β-methyl-DL- n.a. n.a. n.a. n.a. n.a. Phe* (51) (0.9 U/mg) (0.8 U/mg) (2.5 U/mg) (0.4 U/mg) DL-cyclopentyl n.a. n.a. n.a. n.a. n.a. alanine* (52) (1.7 U/mg) (0.3 U/mg) (3.1 U/mg) Table 10: Preliminary results for activity of WT and mutants at 0.4mM concentration of racemic α-amino acids (* 10 mM). Where available, the relative activities (top numbers) refer to the activity shown in comparison to that of the corresponding L- α-amino acid (100). Reactions are performed at pH 10.4 in Gly/NaOH buffer by following spectrophotometrically the production of NADH at 340nm.

Example 6 ; Synthesis of D-α-amino acids.

This has been exemplified by the kinetic resolution of 4-Cl-DL-phenylalanine (45) with the Nl 45 A mutant of PheDH. The system requires NAD+ which is recycled with bacterial diaphorase from TJtermus aquations . The reaction is performed at pH 9.5 (ethanolamine-HCl, 20 mL, 50 mM). The reaction mixture contains 5-10 mM 4-C1- DL-phenylalanine (45), 1 mM NAD+, 100 mM KCl and 0.1 mg DCPIP (dichlorophenol indophenol). The reaction is started by adding 25 μg of Nl 45 A PheDH and 0.1 mg diaphorase and monitored by chiral HPLC as reported above. Figure 3 shows the total kinetic resolution reached after 44 hrs.

Discussion

The wild type enzyme is able to catalyse the reductive animation of α-keto acids bearing 4-substituted aromatic groups, although it is necessary to distinguish between halogens and bulky non-polar groups: while the activity with 4-fluoro (6) and A- chloro (3) substitution is certainly satisfactory (114 and 62.3%), it drops substantially with 4-methyl (7) or 4-methoxy (8) (14.6 and 14.9%) and it is relatively poor with A- trifluoromethyl (9) (0.9%).

In comparison with these results, the activities of the mutants are remarkable: while the mutants show decreased activity for phenyl pyruvate compared to the WT enzyme, the α-amino acid substitution in the binding pocket of these enzymes results ill generally increased tolerance of substitution at the 4-position of the aromatic ring. In the case of the 4-chloro substitution, one of the most striking results is that for N145L (359% as compared with 62% for the wild-type enzyme). However, in assessing this result it is important to note that this mutant shows much the poorest reference activity with phenyl pyruvate (only 22 U/mg). Thus 359% is only 79 U/mg which is in fact the lowest figure for the 4-chloro α-keto acid (3) with the four enzymes tested. With 4-methyl (7) or 4-methoxy (8) on the other hand, the improvement in catalytic activity is absolute, in that activities are not only higher in most cases than with phenyl pyruvate, but also uniformly much higher than the activity of the wild type enzyme with the same substrates (3.9-fold for N145L with 4- methoxy and 4.6-fold for N145A with 4-methyl). In the case of the 4-trifluoromethyl (9) substitution, the N 145 A mutant shows a remarkably high activity of 29.4 U/mg.

Despite the fact that these mutants were designed to discriminate between Phe and Tyr (different group in the 4-position), the effect of substitution in the 2- and 3- positions has also been investigated for the first time: all the enzymes show the same trend when the substrates are 2-substituted, namely a decrease of about 15-fold in specific activity, regardless of the size of the halogen (F or Cl). On the other hand, 3- fluoro substitution is much better tolerated whereas 3-chloro is a considerably less acceptable substitution, suggesting a more precise spatial fit at this position and reflecting the decreased steric demand of the fluoro substituent.

Finally, a series of substrates which differ more substantially from the natural substrate, phenyl pyruvate, has been tested: heteroaromatic α-keto acids (pyridine (10) and thiophene (11) in place of the phenyl ring), and cyclic (cyclopentyl or cyclohexyl (12) in place of the phenyl ring) and linear aliphatic derivatives (C-6) (13).

Pyridine α-keto acid (10) is well tolerated by N145A and the specific activity is very close to the activity shown by the wild type, while the mutant N 145V shows very . little activity. Quite surprisingly, each of the enzymes show activity with a cyclic aliphatic substrate indicating a good degree of freedom in the shape of the binding pocket, especially in N145A and N145V. The activity towards one linear aliphatic α- keto acid (13) is reported and, although the wild type is rather tolerant, all the mutants provide increased activity, offering more versatile catalysis. In conclusion, in this work we explored the possible application in catalysis of engineered enzymes which in several cases have markedly improved activities with novel non-natural α-keto acid substrates. In contrast, the activity of the engineered enzymes with the natural substrate phenyl pyruvate decreases significantly. Clearly, site substitution of the asparagine residue at position 145 for the less polar alanine, leucine or valine residues results in a binding site which can better accommodate substituted aromatic derivatives of phenyl pyruvate than the WT phenylalanine dehydrogenase enzyme. As a general rule, 5% of the wild-type activity with phenylpyruvate might be regarded as a borderline for discriminating between a "good" or a "poor" catalyst. In cases where the activity achieved at present falls well below this figure, such as 3-chlorophenylpyruvate (2), it seems likely that further site- directed mutagenesis may produce better biocatalysts. However, it should be borne in mind that even activities of 1 -3 U/mg represent perfectly usable biocatalysts, especially in view of the possibility of 'over-production' of the biocatalyst in the bacterial host.

We have preliminary data illustrating that the mutant enzymes have considerable potential to be employed as synthetic biocatalysts for the kinetic resolution of racemic non-natural α-amino acids to produce the D-α-amino acids. The potential for use of the engineered enzymes as biocatalysts for the production of non-natural D-α-amino acids and/or non-L-α-amino acids is clear from this work. The optional use of recycling systems for achieving quantitative conversion in both directions is an advantage in the production of non-natural D-α-amino acids and/or non-L-α-amino acids. Significantly, we have demonstrated that there is no decrease in enantioselectivity associated with the increased activity.