Comparison of Different Chemoenzymatic Process Routes to Enantiomerically Pure Amino Acids

INDUSTRIAL BIOCATAL YSIS 50

CHIMIA 2001,55, No.1/2

Chimia 55 (2001) 50-59 © Neue Schweizerische Chemische Gesellschaft ISSN 0009-4293

Andreas S. Bommariusa,b*, Michael Schwarma, and Karlheinz Drauza

Abstract: Five common biocatalytic process platforms forthe production of enantiomerically pure amino acids are compared along four different dimensions of merit: i) enantioselectivity, ii) overall yield, iii) biocatalyst operating stability, and iv) reactor space-time yield. All processes practiced on industrial scale utilize biocatalysts of very high enantioselectivity. Short efficient process routes are a necessity, giving an inherent advantage to routes based on lyase reactions or reduction of prochiral keto acids. For processes based on the splitting of a racemate, recycling the unwanted enantiomer is crucial. The pKa value of abstracting the a-proton serves as a first indication of the difficulty of achieving overall yields far in excess of 50% through dynamic resolution. High solubility of substrates and products was found to favor high reactor productivity (space-time- yield s.t.y. > 1 kg/(I·d)) and high biocatalyst productivity (enzyme consumption number e.c.n. < 500-1000 U/kg of product). Keywords: Amino acid· Enantioselectivity . Operating stability' pKa value· Volumetric productivity

1. Process Performance There are four dimensions of merit prochiral substrates and thus, in princi- Parameters Pertinent to that pertain to any process route, whether ple, allow up to 100% conversion, the Enantiomerically Pure Compounds biologically or chemically catalyzed, main three processes in operation (acyla- which are regarded as most important: se, amidase, and hydantoinase/carba- There are five common biocatalytic selectivity or rather, in the case of moylase) work on the basis of the separa- methods for the production of enantio- manufacture of enantiomerically pure tion of a racemate. Yields at or even be- merically pure amino acids; in rough his- products, enantioselecti vity, low 50% are increasingly unsatisfactory, torical order, these are: i) the acylase • overall yield, however, because they are uneconomical process, ii) the amidase process, iii) the • (bio )catalyst operating stability, and as well as environmentally unacceptable. lyase processes to L-aspartic acid via as- • reactor space-time yield. Various methods that are discussed in partase, to L-alanine with the help of L- Enantioselectivity. The enantiomeric Section 3 have been developed to cir- aspartate-~-decarboxylase (Asp-DC), and excess of any optically active product cumvent the limitation of 50% overall to L-phenylalanine with phenylalanine slated for use in the pharmaceutical or yield. ammonia lyase (PAL), iv) the hydantoin- food industry is a key product attribute. Biocatalyst Operating Stability. Suf- ase/carbamoylase process, most notably As the FDA requires separate toxicologi- ficient (bio)catalyst operating or process to D-amino acids, and v) the reductive cal data for each impurity (including the stability is a crucial parameter for the amination or transamination starting from undesired enantiomer) of greater than 1% success of any process. With such a pa- prochiral a-amino acids. of total composition, an enantiomeric ex- rameter, stability of the (bio)catalyst is cess of > 98% ee is necessary, usually scaled against product formation. The re- even> 99% ee. Use in other industries sult is a total turnover number (TIN) such as agriculture often does not neces- (Eqn. (1», a term which is most often sitate such high enantioselectivity; the used in the homogeneous catalysis field. 'Correspondence: Prof. Dr. A. Bommarius herbicide (S)-Metolachlor, one of the In enzyme technology, the enzyme con- "Research, Development and Applied Technology most successful examples of the applica- sumption number (e.c.n.) is more com- Fine Chemicals Division tion of homogeneous catalysis, is suffi- monly used (Eqn. (2»: degussa. AG P.O. Box 1345 ciently pure for application with 80% ee 0-63403 Hanau because even the (R)-enantiomer has her- IT = m J. f pr duct g 'n 'ralcdlm Ie bPresent address: School of Chemical Engineering bicidal activity but although weaker than of spent atal t I) Georgia Institute of Technology 315 Ferst Drive, Atlanta, the (S)-enantiomer [1][2]. GA 30332-0363, USA Overall Yield. While the lyase pro- II Tel.: +1 4043851334 e .. n. = rna r enz me preparati Fax: + 1 404 894 22 91 cesses (iii) and the reductive amination or pem {in g)/k.g pr ducl ( ) E-Mail: [email protected] transamination processes (v) start from INDUSTRIAL BIOCATALYSIS 51

CHIMIA 2001, 55, No. 1/2

For the operating stability of the rameter in the process to (-)-menthol -.l.)-. = am unt f pr duct g neratedl (bio)catalyst to be considered sufficient, based on the homogeneous catalytic (rea t r volume· time) () 10 000 is a useful threshold number for isomerization is the TIN in excess of TTN; TINs below that number can only 11 reactor volume IS taKen as eqUiva- 100 000. The LeuDH-catalyzed reduc- be sustained in case of a very inexpen- lent to the liquid volume of the reacting tive ami nation of trimethylpyruvate to sively produced (bio)catalyst or a high solution, an increase of s.t.y. necessarily L-tert-Ieucine (see below) scores high on value-added product. Provided the molar corresponds to an increase in reaction enantioselectivity and on stability. The masses of enzyme and product are rate. The rate, expressed in dimensions of different operating stability numbers for known, TTN and e.c.n. can be mutually [mole/(l . s)], in turn tends to be high ei- chloroperoxidase-(CPD-) catalyzed reac- interconverted: ther at high substrate concentration [S], tions demonstrate the possible depen- i.e. at high substrate solubility, or in case dence of TIN on substrate and product e.c.n = WCnJ)m~ I MW pro<.lu~I·IT ) ( ) of a short reaction timescale. Such a short instead of just on the catalyst. timescale is facilitated in the case of a fast Whereas the parameter TIN does al- catalyst, otherwise, a higher catalyst load low comparison between stabilities of is required. As in the case of enantiose- 2. Processes for the Manufacture of homogenous and biological catalysts, the lectivity (ee) and (bio)catalyst operating Enantiomerically Pure Amino Acids molecular mass especially of an enzyme stability (TTN), a threshold number can catalyst is not a suitable reference for the be provided for the specific activity of the 2.1. The Acylase Process complexity or cost of manufacture. In (bio)catalyst, the turnover frequency N-Acetyl-L,D-amino acids can be contrast, both contributions of e.c.n. in (tof): a tof of 1 s'] is considered the mini- transformed to L-amino acids and N- Eqn. (3) can be expressed in monetary mum number for a successful catalyst ex- acetyl-D-amino acids via L-aminoacylase terms so that the relevance of the biocata- cept in cases of extremely high stability. (Acylase I, Scheme 1). With L-aminoacy- lyst cost vis-a-vis other costs can be as- lase, Tanabe Seiyaku in 1969 introduced sessed. Assuming MW productof 200 and Overview of the Status of an immobilized enzyme into industrial MWenzymeof 50 000, a threshold of 10000 Development of Catalytic scale [9-13]. In 1982, Degussa began to for TTN translates into an e.c.n. of 25 g Processes operate the first enzyme membrane reac- enzyme preparation per kg of product. To assess different processes and bio- tor (EMR) for the manufacture of phar- The relevant parameter to measure for catalysts in the following sections, it is maceutical intermediates, again with L- the calculation of the e.c.n. and thus for useful to compare results for the dimen- aminoacylase [3][4][14][15]. Currently, the assessment of operating stabilities of sions of merit with pertinent benchmark the EMR process is operated for several enzymes is the product of active enzyme cases. Table 1 provides numbers that hundreds of tons annually of L-amino ac- concentration [E]activeand residence time have been achieved for mostly enantio- ids, especially L-methionine and L-valine 't, [ELc'ive''t . In a continuous stirred tank selective processes, both biologically and (Scheme 2). reactor (CSTR) the quantities [E]active homogeneously catalyzed. The EMR is operated as a continuous and 't are linked by the following equa- From Table 1 it can be discerned that stirred tank reactor (CSTR) with a recy- tion: only few processes meet all of the thresh- cle ratio (feed flow rate into reactor/rate old criteria mentioned above. In the case of recycled flow) of up to 200. On pilot ([ lucllvc·t)/[ oJ = Ir( ) (4) of (S)-metolachlor, 80% ee is sufficient and production scale the ultrafiltration for use as a herbicide as the unwanted membranes are configured as hollow where [So] signifies initial substrate con- enantiomer is not toxic and even some- fibers with a molecular weight cut-off centration, x the degree of conversion what active; however, both tof and TTN around 10 kDa and a retention rate of and rex) the conversion-dependent reac- are impressive. The key performance pa- considerably larger than 99.9%. tion rate [3][4]. Often, biocatalyst stability is reported as the half-life 't 1/2 of an enzyme catalyst. Under the following two conditions, half- Table 1. Benchmark numbers for dimensions of merit for catalytic processes lives 't1/2 are equivalent to enzyme consumption numbers: i) there are no diffu- Product Process Catalyst Use, tot [S'l] ee [%] TIN [-] Ref. sion limitations (the effectiveness factor company 11 :::: 1), and ii) the substrate fully saturates (5)- cata!. Ir, herbicide, S 100 80 108 [1][2] the enzyme ([S] » K , the reaction is M metolachlor hydro- ferrocene Novartis [6J zeroth order). Then, both in a CSTR and a genatlon plug-flow reactor (PFR), with both constant flow as well as constant conversion (-)-menthol cata!. Rh -(5)- food, >99 3 1()5 (7] policy, a true half-life is calculated if ei- isomeri- BINAP Takasago ther the degree of conversion (constant zation flow policy) or flow rate or enzyme con- L-tert. reductive LeuDH pharma, 50 > 99.9 107 (7] centration (constant conversion policy) leucine amlnation degussa. are at half the initial values [5]. butylene epoxldation CPD pharma « 1 > 98 4 1()3 [8] Reactor Space-Time Yield. Indepen- oxide intermediate dent of the catalyst, space-time yield is a 5 universal criterion for volumetric reactor oxindol epoxidalion CPD Crixivan « 1 n.a. 8.4 10 [8] intermediate producti vity: INDUSTRIAL BIOCATALYSIS 52

CHIMIA 2001,55, NO.1/2

Of the numerous investigations on 0 0 0 acylase I the reader is referred specifical- Acylase ly to the literature on substrate specificity Ryt-OH RrOH R?OH [16], reaction engineering [17], opera- pH 7.0 - 7.3 • + tional stability [18], and breadth of appli- H2N HNy HN)( cability [19]. 20-37 °C 0 0 2.2. The Amidase Process Scheme 1. L-Aminoacylase reaction. Amino acid ami des, generated from nitriles through chemical or enzymatic hydrolysis, in tum can be hydrolyzed to L-amino acids with the help of L-amino acid amidase (Scheme 3). The catalyst is used as crude permeabilized whole cells in the pH range of 8-10 and 20% substrate by weight [20]. The substrate specificity of L-amidase is wide [21], similar to L-acylase. The enzyme can be isolated from a wide variety of microorganisms; DSM has developed RyCOOH furthest those from Pseudomonas putida RyCOOH and Ochrobactrum anthropi. A compre- + CH hensive and fairly recent review on the ~ HNy 3 H20 NH2 amidase process has been published in dlv.r.nt metal Ion .cU •• tor o [22]. K.qu • 3-12 IN-Acetyl-O,L-emlno acid ~ I L-Amlno acid ~ IN-Ac.tyl-o-amlno acid ~ 2.3. The Lyase Processes While several more processes have t +_C_H3_C_O_O_H _ been developed with lyase enzymes such Rac.mlzatlon as a process to hydroxy and amino acids with oxynitrilase (hydroxynitrile lyase, HNL) [23-26], only three are discussed Scheme 2. L-Acylase process of degussa. here; the processes to L-aspartate from fumaric acid with L-aspartase, the reaction from L-aspartate on to L-alanine catalyzed by L-aspartate-~-decarboxylase, and the synthesis of L-phenylalanine 0 from cinnamic acid via phenylalanine RyCN Ketone NH3.HCN ~ ~ ammonia lyase, PAL [27]. R)lH Strecker NH 2 OH L-aspartate is generated with the help Synthesis of aspartase from inexpensive fumaric Cl-Aminonitrile acid, preferentially used as ammonium fumarate, via addition of ammonia 0 (Scheme 4). While use of living cells was L-specific 0 aminopeptidase COOH effective, higher productivities and sta- R0NH2 ~ R0 + bilities were achieved by using immobi- -- :: NH2 NH 2 NH H2 lized cells. In 1974, the first commercial (Pseudomonas putida) 2 1 process based on immobilized cells was introduced by Tanabe Seiyaku. Since D,L-Amino acid amide D-Amino acid amide then, a large number of immobilization protocols have been developed [28] I Benzaldehyde resulting in one of the most efficient en- t pHB-11 zyme processes known today (see below). Today, the aspartase reaction is virtually the only process by which L-aspartate, a precursor of Aspartame®, is 1) H30'~ RyCOOH manufactured in quantities of> 5000 tons annually by several producers around the 1) oW (pH 13) 2)OH" NH2 world. Racemization 2) H30' The L-aspartate-~-decarboxy]ase route is based on the thermodynamically favo- D-Amino acid I rable ~-decarboxylation of L-aspartate via L-aspartate-~-decarboxylase (L-Asp- Scheme 3. Amidase process (DSM). ~-DC, E.c. 4.1.1.12.) (Scheme 5). INDUSTRIAL BIOCATALYSIS 53 CHIMIA 2001,55, No. 1/2

This reaction has some features ideal for large-scale processing: HOOC~COOH ~COOH • The irreversible decarboxylase reac- Asoartase HOOC I tion allows complete conversion of L- NH + 3 NH2 aspartate to L-alanine and carbon dioxide in up to 100% isolated yield, in Scheme 4. Aspartase reaction. contrast to separation of racemates. Since the chiral a-C-atom is not in- volved in key reaction steps, the L-aspartate-~-decarboxylase reaction proceeds enantioconservatively and the ~COOH L-Asp-P..-DC yCOOH enantiomeric purity of the L-alanine HOOC 1 product is not compromised. The NH NH2 2 gene coding for L-aspartate-~-decarboxylase has been cloned and overex- pressed [29] so that the alanine race- Scheme 5. Aspartate-B-decarboxylase reaction. mase activity commonly occurring in the wildtype enzyme cannot be de- tected anymore. H • The CO2 generated acts as buffer at the pH optimum of the enzyme, pH 6, ~COOH NH40H so that only L-alanine is obtained be- ©Jt~OOH L ·Phenylalanine sides volatile co-products when am- ammonia lyase monium aspartate is used as substrate. (PAL) Different pH optima of L-aspartase trans-cinnamic acid L-Phe (pH 8.5) and L-aspartate-~-decarboxylase (pH 6.0) preclude use of a single reactor to convert fumaric acid directly to L- Scheme 6. Process to L-phenylalanine from cinnamic acid via PAL. alanine [30]. L-phenylalanine can be obtained by simple addition of ammonia to trans-cinnamic acid catalyzed by phenylalanine ammonia lyase, PAL [27] (Scheme 6). I L -specific branch ~ I D-speclflc branch ~ The process had been scaled up for the production of L-phenylalanine for Aspar- RyCOOH I L-Hydantoinase I R~O RHO I o-Hydantoinase I R...... ,...... COOH pH> 8 tame by Genex but ultimately had to be ..i +HzO , \ .H20 ""'" .: HNyNH2 ~ HNyNH HNyNH 00;:-- HNyNH2 abandoned because the costs to L-phenyl- ...H20 -H20 alanine via fermentation from glucose o 0 o o were lower than those for the enzymatic L-Carbamoyic acid L-Hydantoin I o-Hydantoin I o-Carllamoylic acid process. H.o,(L-Hydanloic acid) (o-Hydantoic acid) 2.4. The Hydantoinasel I L-Garllamoylase I Carbamoylase Process RyCOOH "yCOOH •• If II!'~ Hydantoins, which can be viewed as NH2 Mf2 cyclically protected amino acids, can be COl. Mil converted to D- or L-carbamoyl amino I L-Amino acid I I o-Arrm acid I acids through the action of D-or L-hydan- Racemization of hydantoin can be rate-llmltlng: toinases (Scheme 7). The carbamoyl ami- 't1/2.rac (pHS.5, 40·C): R=phenyl-:O.3h, R=benzyl-:5h, R=lsopropyl-:56h no acids can be further converted to the D- or L-amino acids with the respective Scheme 7. Hydantoinaselcarbamoylase process to D- or L-amino acids. carbamoylases. Only innocuous by-products such as ammonia and carbon dioxide are formed. The hydantoinase/carbamoylase process is widely used to produce D-phenylglycine and D-p-hydroxyphenyl-glycine, sidechains of the antibiotics '-l eOOH LeuDH '-l COOH ampicillin and amoxicillin, respectively, /Y (S)-Tle TrimethYIPyru::e g with annual production volumes exceed- ~ NH2 ing 5000 tons annually worldwide. NAOH + H+ NAO+ The o-hydantoinase/D-carbamoylase process has been described in detail [31]. e024"--~------HCOO-NH/ The commercially available D-hydanto- FDH inase has a wide substrate specificity [32] with already a high enantioselectivity. Scheme 8. Reductive amination process with cofactor regeneration. INDUSTRIAL BIOCATAL YSIS 54 CHIMIA 2001. 55. No.1/2

The D-carbamoylase is very enantiospe- dox cycle between L-aspartate and L- extended with additional purification and cific, thus enhancing the action of the glutamate catalyzed by aspartate amino- polishing steps which often compromise D-hydantoinase, but is rather unstable so transferase (AAT) is necessary. This efficiency, overall chemical yield, or it has not been used in isolated form. The latter cycle is driven by the spontaneous space-time yield. As an example, the L-branch of the system is much less well decarboxylation of the oxaloacetate gen- lipase-catalyzed ring opening of azlac- known [33]. L-hydantoinase is found to erated from aspartate into pyruvate and tones in toluene yielded the substituted possess a low enantiospecificity which CO2, Again, pyruvate has to be removed L-tert-Ieucine ester which upon direct often even favors the D-side with a partic- because it can serve as substrate for hydrolysis in 6N HCI yielded L-TIe with ular hydantoin substrate. Only recently BCAT, so it is condensed with itself with 73% ee (c.y. 80%), however, an addition- the L-hydantoinase was evolved towards the help of acetolactate synthase (ALS) al hydrolysis step with alcalase followed either higher L- or D-enantioselectivity which in turn spontaneously decarboxy- by treatment with HCI resulted in an [34]. For a review and comparison of lates again, to volatile acetoin which enantiomeric purity of 97% ee [47]. both D-and L-hydantoinase/carbamoylase can be removed by distillation. As in the It deserves to be noted that, in contrast systems the reader is referred to reference cases of aminoacylase and amidase, sub- to processes based on homogeneous ca- [35]. strate specificity is not very tight, allow- talysis, any process to enantiomerically ing for conversion of a variety of keto pure compounds that was developed to 2.5. Reductive Amination or acid substrates. This feature of transami- large scale, uses an enzyme with very Transamination Starting from nases and others has been reviewed in re- high to near perfect enantioselectivity, at Prochiral a-Amino Acids cent years [43][44]. least in the last step. As the example of a-Keto acids can be reduced to a-ami- hydantoinases demonstrates, enzyme cat- no acids with the help of amino acid alysts in any penultimate step do not have dehydrogenases. Leucine dehydrogenase 3. Assessment of Quality of the to be completely enantioselective to be (LeuDH) and phenylalanine dehydro- Different Processes considered for a large-scale process. genase (PheDH) are of particular interest However, as issues of overall yield and for synthesis. The area has been reviewed 3.1. Enantioselectivity space-time-yield arise, efforts are made by Hummel and Kula (1989) [36], Ohshi- A comparison of results regarding to improve even enzymes of penultimate rna and Soda (1989) [37] and, most re- enantiomeric excess of products for the processing steps [34]. cently, by Brunhuber and Blanchard different enzyme systems discussed in (1994) [38]. this article reveals that all of them offer 3.2. Overall Yield or Substrate Enzymatic reductive amination with enantioselectivities above 99% ee, at Utilization: Thermodynamic Yield NADH as cofactor can only be operated least in the last crucial step yielding the Limitations, Prochiral Selectivity or on large scale if the cofactor is regener- amino acid product (L-amidase: [45], L- Ease of Racemization ated. Wandrey and Kula have developed acylase: [46], L-carbamoylase [33], and The chemoenzymatic amino acid a regeneration scheme using formate as transaminase [42]). In the hydantoinase/ processes fall into two different catego- reductant ofNAD+ generated upon reduc- carbamoylase system, this function rests rIes: tive amination. The formate is oxidized on the carbamoylase which has been • The acylase, amidase, and hydantoin- irreversibly to CO2 by formate dehydro- found to be highly enantioselective. ase/carbamoylase processes utilize genase (FDH, E.C. 1.2.1.2.) (Scheme 8) However, the enantio-selectivity of hy- one enantiomer of a racemic mixture. [39]. dantoinase, the enzyme catalyzing the Per pass (i.e. without recycle), the For soluble reactants and products, penultimate step from hydantoin to car- maximum yield is limited to 50%. enzymes are preferentially immobilized bamoylic acid, varies with the substrate • The processes based on addition (as- in an enzyme-membrane reactor (EMR). and is always imperfect [34]. partase), elimination (L-asp-~-decar- To keep the cofactor from penetrating Routes that do not offer the required boxylase), or reductive amination of a through the membrane, it can be enlarged level of enantiomeric purity have to be keto group (amino acid dehydro- with polyethyleneglycol (PEG) [40]. With the advantage of quantitative use of keto acid substrate and with a suitable E.C. 2.6.1.x process of cofactor regeneration, enzy- + matic cofactor-dependent reductive ami- Aminotransferase nation of trimethylpyruvate is the route of choice for large-scale synthesis of enantiomerically pure (S)- TIe. Degussa operates an industrial process producing high-quality (S)-Tle on tons scale via this L-Asp AAT Oxalacetate - Pyruvate + CO2 route [41]. a-Keto acids can also be reduced to ~ ALS a-L-amino acids via transamination ~ (Scheme 9) [42]. In this process, L-aspar- a-KG L-Glu tate serves as the ultimate amino donor; however, only the broad-range trans- ~ yt! ferase can accept L-aspartate directly. L-AA BCAT keto acid o The branched-chain transferase (BCAT, branched-chain aminotransferase) only accepts L-glutamate, so an additional re- Scheme 9. Transamination process to a-amino acids (NSC). INDUSTRIAL BIOCATAL YSIS 55

CHIMIA 2001, 55, No. 1/2

genases and transaminases) can utilize Table 2. pKa Values for 5-substituted hydantoins up to 100% of the substrate. Sidechain Amino acid pKa Before investigating any other yield- limiting factor, possible thermodynamic Phenyl Phenylglycine 870 limitations should be ascertained. Among 3'-Pyndylmethyl 3'-Pyridylalanine 8.78 the processes based on a separation of Methylthloethyl Methionine 9.00 racemates, only the amidase process is Imidazolmethyl Histidine 9.03 not thermodynamically limited. The hy- Methyl Alanine 9.10 drolysis reaction of N-acetyl amino acids Benzyl Phenylalanine 9.11 is equilibrium-limited, however, the equi- Isopropyl Valine 9.14 librium is far on the side of the hydrolysis Carboxyethyl so that at low substrate concentrations Glutamic acid 9.26 conversion is nearly quantitative. For the p-Chlorobenzyl p-CI-Phenylalanine 9.28 case of N-acetyl methionine, the equilibrium constant K (K = [acetate][L-Met]/ Table 3. Racemization half-lives [N-Ac-L-Met] ) was determined to equal 2.75 at 37°C and pH 7 []7]; equilibrium Side chain Amino acid t 1/2 [h] conversion Xe (based on N-Ac-L-Met) at [So] = 100 mM is 96%, and decreases to 91 % at [So] = 300 mM and 86% [So] = Hydroxyphenyl p-OH-Phenylglycine 0.21 500 mM. In the hydantoinase/carbamoy- Phenyl Phenylglycine 0.27 lase system, there is an equilibrium be- Hydroxymethyl Serine 1.60 tween the hydantoin and the carbamoylic Benzyl Phenylalanine 5.00 amino acid (K "" I0 favoring the carbamoylic acid), however, as the car- Methylthioethyl MethiOnine 5.82 bamoylase reaction is irreversible Imidazolylmethyl Histidine 16.09 (Scheme 7), the combined enzymatic Isobutyl Leucine 21.42 system produces yields up to 100%. Methyl Alanine 33.98 Lyase-catalyzed reactions discussed Isopropyl Valine 55.90 in this atticle (aspartase, L-aspartate-~- tert-Butyl tert-Leucine 120 decarboxylase and phenylalanine ammonia lyase) are not thermodynamically Conditions: 40 cC, pH 8.5, no buffer added limited to any significant degree. The two Contents of Table partially excerpted from [35] amination processes, however, offer a stark contrast on this point: whereas the Accordingly, hydrogen abstraction al- 3. The latter then can be racemized in a reductive amination process with amino ready can be afforded by OH- ion at mod- facile manner at pH "" 13 and recycled acid dehydrogenases is not limited signif- erate pH values at which hydantoinases [55). icantly (for leucine/ ketoleucine at pH and carbamoylases are active and, at least 11.0, Keq = 9.1012 ([48)), the equilibrium in the o-hydantoinase/o-carbamoylase Ease of Racemization of Acetyl Amino constant of the transamination process is system, also stable. The pH optimum of Acids in the order of 1 (K = 1.86 for L-Alala- hydantoinases has been reported to be at The pKa of abstraction of a-hydrogen ketoglutarate and pyruvate/L-Glu [49]; 8.0 to 9.0 [31][33]. atoms from acetyl amino acids is much for additional equilibrium constants, see Despite pKa values suitable for pro- higher than 8-9, approaching ]5. There- [50)). ton abstraction with common bases such fore, either extremes of temperature and For the processes based on the separa- as hydroxide at moderate pH values, ra- pH or extraordinary catalysts are neces- tion of racemate, the yield can be in- cemization proceeds more quickly in the sary to afford racemization. Neverthe- creased beyond 50% if the unconverted presence of chemical or enzymatic cata- less, N-acetyl amino acid racemases enantiomer is racemized and recycled. lysts. When passed over strongly basic (N-AAARs) have been found, first by The pKa value of the labile a-hydrogen ion exchange columns, hydantoins ra- Tokuyama et al. [56-59] and recently by serves as an indicator of the ease of ra- cemize preferentially in the highly basic Verseck [60]. cemization. layer around the beads [51]. In addition, Scheme 10 depicts the desired pro- two hydantoin racemases have been cess scheme for the coupled acylase- Ease of Racemization of Hydantoins found and characterized with respect to acetyl amino acid racemase reaction The pKa values for the abstraction of sequence and substrate specificity [52-54). which increases overall yield from about a hydrogen atom in position 5 do not dif- 45% based on N-Ac-oL-amino acid to fer much (Table 2); they range from 8.5 Ease of Racemization of Amino acid nearly 90%. In principle, acetyl amino to 9.5. There is no discernible trend for Amides acid racemase can be coupled with either aromatic and aliphatic, functionalized While unsubstituted amino acid an L- or a o-acylase. The enzyme requires and unfunctionalized sidechains. Corre- amides are not easily racemized (pKa ;?: divalent metal ions to function, the order sponding racemization half-lives at 40°C 15) these compounds can be treated with of activation is C02+ > Mg2+ > Mn2+ > and pH 8.5 in the absence of any buffer aromatic aldehydes at pH 8-11 resulting 2n2+ [60). Its pH-optimum around pH 8 are listed in Table 3; values for half-lives in formation of a Schiff base in accord- is compatible with that from acylase at span more than two orders of magnitude. ance with the bottom branch in Scheme pH7. INDUSTRIAL BIOCATALYSIS 56

CHI MIA 2001, 55, No.112

dure is applied, i.e. either the constant flow policy with varying degree of conversion or, even less ambiguous, the constant conversion policy with either + changing enzyme concentration or L-Acylase changing residence time. 2 pH 7.0, Zn +,30°C For the amino acid processes discussed here, the numbers are presented in o,L-AA L-AA N-Ac-o-AA Table 4. The data in Table 4 demonstrate that half-lives on the order of one month need to be attained for sufficient operating stability. At prevalent high substrate con- N-Acetyl amino acid racemase centrations of often more than 1M, such half-lives correspond with enzyme con- pH 7.0, C02+, 30°C sumption numbers of between low 100s to low 1000s units per kg of product. Scheme 10. Yield increase by enzymatic racemization of N-acetyl-D-amino acid. While the enzyme consumption number is important, this quantity is modulated by the price to produce one unit of activity. With the advent of efficiently L-Asp ~ Oxalacetate Pyruvate + CO2 t produced recombinant enzymes, even e.c.n.'s of 1000 Ulkg can still mean a suf- ALS ficiently inexpensive biocatalyst for keto acid ~ t large-scale processing. ~ The operating stability of acylase I o-AAO, \ O2 has been particularly well characterized catalase \ I [16][18][66]: In aqueous solution, the If-t resting stability of acylase from Aspergil- o,L-AA o lus oryzae was found to depend much more on pH than on concentration: while Scheme 11. Coupling of transaminases with D-amino acid oxidase to yield solely L-amino acid. at room temperature (25 0c) and standard pH (7.0) half-life 't1/2 was around 60 days between concentrations of 30 to 120 g The physiological role of an acetyl amination reactions face enzymatic crude enzyme/I, 't1/2 dropped to 45 d at amino acid is unclear. There is evidence side reactions. Even if the aforemen- pH 6.5 and to about 30 d at pH 6.0 [66]. that the enzyme fortuitously acquired the tioned thermodynamic limitations are al- Also, in solubilized form, the fungal en- function of racemizing acetyl amino leviated through decarboxylation of oxa- zyme is quite stable whereas porcine kid- acids and that the protein's main function loacetate and condensation of pyruvate ney enzyme is sensitive to auto oxidation is as an O-succinylbenzoate synthase with acetolactate synthase (ALS), the and therefore should be kept under nitro- (OSBS) rather than an N-AAAR [61]. pyruvate formed can act as a substrate gen if stored in solubilized form [16]. Both the k~atas well as kcat/KM values for the branched-chain aminotransferase Tests for operating stability in repeated- suggest that the protein is a much better (BCAT) (Scheme 9), resulting in amino batch mode ([3][16][17][67]) reveal that OSBS than N-AAAR. acid mixtures that can be difficult to acylase from Aspergillus oryzae again Currently, however, racemization of purify. fares much better than the porcine kidney acetyl amino acids is effected chemically An interesting scheme to purify D,L- enzyme. Tests for operating stability in a at low pH values combined with high amino acids enantiomerically is given in continuous reactor with the acylase from temperatures and resulting in an overall Scheme 1I [63]. The D-amino acid is ox- Aspergillus oryzae [18] demonstrated yield on the order of 80% [15]. idized to the corresponding keto acid with D-amino acid oxidase, available - the superior stability of AA (616 U/kg Yield Limitations in Amination Processes from Trigonopsis variabilis; the keto acid L-methionine) over PKA (6000 U/kg The reductive amination and transam- is transaminated with L-glutamate to the L-met) both measured with [C02+] of ination processes from prochiral a-keto L-amino acid, and oxaloacetate decar- 5·1O-4M, acids to the corresponding amino acids boxylates to pyruvate which in turn is - the tighter binding of Zn2+ vs. C02+ at do not suffer from yield limitations re- converted to acetoin by condensation and 5·1O-4M (308 vs. 616 U/kg L-met), sulting from an unwanted enantiomer. decarboxylation as described earlier. - that loss of metal, commonly Zn2+, is However, even such process schemes can responsible for activity loss and suffer from other limitations affording 3.3. Biocatalyst Operating Stability - the possibility of reconstitution over a yields ofless than 100%. At high concen- Data on biocatalyst stability under timescale of several hours, whereas trations, keto acids and ammonia can industrially relevant conditions are not- the time constant of leaching is on the form dimeric-type by-products (N-acyl- oriously hard to obtain. Following the order of 48 h, as well as D,L-amino acid amide) from condensa- stability behavior of even a reasonably - the option of pulsing divalent metal tion reactions [62]. In addition to this stable biocatalyst is time-consuming and addition resulting in 477 U/kg L-met chemical side reaction, enzymatic trans- laborious, especially if the proper proce- at [Zn2+] of 4·1O-4M. INDUSTRIAL BIOCATALYSIS 57 CHIMIA 2001,55, No. 1/2

Porcine kidney acylase seems to have Table 4. Enzyme consumption numbers and half-lives for amino acid process biocatalysts a different spectrum of activation [3]: al- Enzyme (organism) e.c.n. [Ulkg] '112 [d] Substrate Ref. though C02+ activates PKA most strongly, Ca2+ is not far behind whereas Zn2+, Aspartase (E.CO/I) 17~90 18(H30 NH4 fumarate [64] just like Mg2+ or Fe2+, does not seem to L-Acylase (A. oryzae) 306-370 N-Ac-o.L-Met [4] exert a strong effect. Both the fungal and the porcine enzyme have moderate ther- Asp-~-decarboxylase 1000-2400 45 NH4 aspartate [64] mostability [68] and moderate stability in LeuDH, FDH 600-800 > 30 Trimethylpyruvate [65] the presence of organic cosolvents [16]. Hydantoinase 30 Phenyl hydantoin [64] Transaminase 30 not known [64] 3.4. Space- Time- Yield: The Influence of Solubility and Abbreviations: e.c.n. = enzyme consumption number; 1'1/2 = half-life. Dispersion

Conditions for using an EMR Reactor Table 5. Solubility of amino acids and derivatives As far as reactor configurations are concerned, retention can be achieved Amino Acid Xaa Solub. Ac-Xaa Solub. Hydantoin Solub. Carba- Solub. through immobilization (fixed- or fluid- [M] [M) [M) moyl-AA [M) ized-bed) or by ultrafiltration, either in successive batch mode, as dead-end fil- Alanine Ala 2.28 Ac-Ala 1.32 tration or as a recycle reactor [69]. The Amlnobutyric Abu 2.25 Ac-Abu 1.05 Ethylhyd. 0.04 enzyme membrane reactor has been re- acid viewed several times [15][70]. Methionine Met 1.01 Ac-Met 0.7 MTEH 0.2 Carb-Met 0.52 Employing soluble enzymes retained Leucine Leu 0.14 Ac-Leu 0.18 in an enzyme membrane reactor possess- Phenylalanine Phe 0.25 Ac-Phe 0.45 BH 019 Carb-Phe 0.105 es the advantages of i) absence of mass Tyrosine Tyr 0.004 Ac-Tyr 0.29 p-OH-BH 0.28 transfer limitations in connection with Homophenyl- Hph <0.002 Ac-Hph 0.07 nearly perfect retention, ii) ease of re- plenishment of fresh catalyst which is alanine much easier than in an immobilized en- Abbreviations: Xaa = amino acid, MTEH = Methylthioethylhydantoin, BH = Benzylhydantoin, zyme reactor column, and iii) ease of Carb = Carbamoyl. scale-up; fluidity F (F = V/(~P·t·A) [liters/ Conditions: Xaa: water, 40 ec, Ac-Xaa: pH 7.5, 40 ec, hydantoins: pH 8.5, 40 ec, carbamoylic (bar·hour·m2)], V = uItrafiItered volume, amino acids: pH 8.5, 40 °C. ~P = transmembrane pressure, t = filtra- tion time and A = membrane area) can be assumed to stay constant during scale-up. Table 6. Reactor productivities of biocatalytic processes Degussa-Hiils has wide experience with Product enzyme s.t.y. (g/(I d)] [5]0 [M] Ref. both continuous and batch membrane reactors on very different scales, ranging L-Aspartic acid L-Aspartase 60000 1.5 [64J from small-scale laboratory models with L-AJanlne Asp-DC, fixed-bed. Ideal, 4 bar 45600 1.35 [66] ]0 ml volume via the pilot-plant scale L-Alanlne Asp-DC. fixed-bed, Ideal, 1 bar 31200 1.3 [29] models at 500 cm2 to 0.5 m2 to the plant- L-Alanine Asp-DC, fixed-bed. dispersion 7500 1.35 size reactors at several cubic meters [15]. [66J L-A1anlne Glutamic-pyruvIc transaminase 4800 1.0 [64] Solubility of Substrates and Products L-tert-Leucine LeuDH, FDH 638 0.5 [65] An EMR reactor can only be used L-Methionine L-Acylase, soluble 600 0.6 [64] with soluble substrates and products. Ta- L-Valine L-Acylase. Immobilized 500 0.4 [11] ble 5 provides an overview of common L-p-F-phenylalanine broad-range aminotransferase 480 O.t [71] amino acids including their derivatives. L-VaJine L-Acylase, soluble 350 0.4 [171 As enzymes often do not work at all in suspension or are co-precipitated, or as Abbreviations: s.t.y.: space-time-yield, [S]o = initial substrate concentration, Asp-DC: L-aspar- the reaction is performed in an enzyme tate decarboxylase, LeuDH: leucine dehydrogenase, FDH: formate dehydrogenase. membrane reactor, solubility of substrates and products is important. From Table 4, however, it is apparent that there Table 6 reveals that experimental data Dependence of Space-Time-Yield on is no a priori preferred route that features on reactor productivity indeed demon- Reactor Configuration: L-Acylase more highly soluble substrates or inter- strate a dependence on substrate concen- The L-acylase process has been the mediates among the alternatives dis- tration as expected: productivities in the pioneering process in enzyme technology cussed in this paper. By combining Eqn. range of greater than 1-10 kg/(l·d) cannot for both the immobilized plug-flow col- (4) and (5) it can be seen that the maxi- be attained with fairly insoluble sub- umn reactor as well the continuous mum substrate concentration attainable strates; however, even productivities of stirred-tank membrane reactor. Some- sets a limit to reactor productivity. Some only several hundred grams per liter and times, results for different dimensions of such numbers of reactor productivities day should not be the limiting factor if merit for the same process do not point in are listed in Table 6. the product has an attractive price level. the same direction: while the half-life of INDUSTRIAL BIOCATALYSIS 58 CHIMIA 2001,55, No.1/2

acylase I immobilized on a nylon mem- ficient process routes are a necessity giv- [1] H.-V. Blaser, F. Spindler, Top. Catal. brane as measured by thermal stability of ing an inherent advantage to routes based 1997,4,275-282. 161 days [72] is superior to the data for on lyase reactions or reduction of prochi- [2] H.-V. Blaser, H.-P. Buser, K. Coers, R. immobilized acylase (65 d) [11] or solu- ral keto acids. Processes based on the Hanreich, H.-P. Jalett, E. Jelsch, B. Pugin, ble enzyme in an EMR [17], reactor pro- separation of a racemate have to empha- H.-D. Schneider, F. Spindler, A. Weg- mann, Chimia 1999, 53, 275-280. ductivity at 0.136 L-valine kg/(l·d) is size recycle of the unwanted enantiomer. [3] C. Wandrey, Habilitationsschrift, TU Han- lower than that for DEAE-Sephade£(-im- The pKa value of abstracting the a-hy- nover, 1977. mobilized acylase (0.5 kg/(l·d)) [11] or drogen on the remaining unconverted [4] A.S. Bommarius, K. Drauz, U. Groeger, that for a membrane reactor (0.35 kgl enantiomer serves as a first indication of C. Wandrey, in 'Chirality in Industry', (l·d)) [17]. However, as both soluble en- the difficulty of the prospect for achiev- Eds. A.N. Collins, G.N. Sheldrake, J. zyme and immobilized enzyme processes ing overall yields far higher than 50% Crosby, Wiley & Sons Ltd., London, New depend on soluble substrates and prod- through dynamic resolution. Volumetric York, 1992, Chapter 20,371-397. ucts, high solubilities of the reaction productivity figures should be critically [5] T. Yamane, P. Siricote, S. Shimizu, Bio- tech. Bioeng. 1987,30,963-969, components seem to be an almost neces- checked for the conditions employed for [6] U. Nagel, Chem. Ber. 1996, 129, 815- sary condition for high space-time- their generation: degree of conversion 821. yields. and limiting rate process are especially [7] A.S. Bommarius, M. Schwarm, K. Drauz, pertinent parameters towards this end. J. Mol. Cat B: Enzymatic 1998, 5, 1-11. L-Aspartate-f3-Decarboxylase in a Fixed- High solubility of substrates and prod- [8] M.P.J. van Deurzen, Biocatal. Biotrans- Bed Reactor: Influence of Dispersion ucts, in addition to a short process route, form. 1997, 15, 1-16. The dependence of space-time-yield favors high reactor productivity as ex- [9] T. Tosa, T, Mori, N. Fuse, I. Chibata, En- zymologia 1966, 31, 214-224. and thus volumetric productivity on fac- pressed by space-time yields and high [10] T. Tosa, T. Mori, N. Fuse, 1. Chibata, tors such as degree of conversion and biocatalyst productivity as indicated by Agric. Bioi, Chem. 1969,33, 1047-1052. detailed reactor configuration has been the enzyme consumption number for the [I I] 1. Chibata, T. Tosa, T. Sato, T. Mori, studied in detail for the decarboxylation operating stability. Meth. Enzymol. 1976,44, 746-759. of L-aspartate to L-alanine with L-as- [12] 1. Chibata, T. Tosa, T, Sato, Meth. Enzy- partate-p-decarboxylase in a fixed-bed Received: December 21, 2000 mol. 1976,44,739-746. reactor [66]. Whereas the s.t.y. at the [13] T. Takahashi, O. Izumi, K. Hatano, Kazu- nori, EP 0 304 021 (Takeda Chemical In- optimum operating point in a 15 ml-spi- dustries, Ltd.), 1989. ral-wound catalytic reactor at an initial [14] W. Leuchtenberger, M. Karrenbauer, U. substrate concentration [L-Asp]o of PlOcker, Enzyme Engineering 7, Ann. N. Y. 1.35M at 99% conversion was 45.6 kg L- Acad. Sci. 1984,434, 78-86. Ala/(I·d), the corresponding figure in an [15] A.S. Bommarius, M. Schwarm, K. Drauz, 83 ml-fixed bed reactor at 90% conver- Chimica Oggi 1996, 14(10), 61-64. sion but otherwise similar conditions was [16] K. Chenault, J. Dahmer, G.M. Whitesides, J. Am. Chem. Soc, 1989, 111, 6354-6364. only 7.5 kg L-Ala/(l·d). At longer resi- [17] C. Wandrey, E. Flaschel, Adv. Biochem. dence times and thus lower degrees of Eng., Eds. T.K. Ghose, A. Fiechter, N. conversion, higher s.t.y. can be attained; Blakebrough, Springer, Berlin, 1979, 12, however, these conditions are not rele- 147-218. vant for processing. [18] A.S. Bommarius, K. Drauz, H. Klenk, C. The cause of the lower s.t.y. was Wandrey, Ann. N. Y. Acad. Sci.(Enzyme found to be neither the difference in de- Eng. XI), 1992, 672, 126-136. gree of conversion nor internal (pore) [19] U. Groeger, A.S. Bommarius, in 'Enzyme Catalysis in Organic Synthesis', Eds. K. diffusion effects (the Wagner-Weisz- Drauz, H. Waldmann, YCH, Weinheim, Wheeler (WWW) modulus

CHIMIA 2001. 55, No. 1/2

[27] S.B. Mirviss, S.K. Dahod, M.W. Empie, [52] K. Watabe, T. Ishikawa, Y. Mukohara, H. Ind. Eng. Chern. Res. 1990,29,651-659. Nakamura, J. Bacteriol.1992, 174, 3461- [28] G.J. Calton, in 'Biocatalytic Production of 3466. Amino Acids and Derivatives', Eds. D. [53] K. Watabe, T. Ishikawa, Y. Mukohara, H. Rozzell, F. Wagner, Chap. I, Hanser, Nakamura, J. Bacteriol. 1992,174,7989- Munich, 1992, 1-21. 7995. [29] J.D. Rozzell, US-Patent 5019509,1991. [54] A Wiese, M. Pietzsch, C. Syldatk, R. [30] A.-S. Jandel, H. Hustedt, C. Wandrey, Mattes, 1. Altenbuchner, J. Biotechnol. Eur. J. Appl. Microbiol. Biotechnol. 1982, 2000,80,217-230. 15,59-63. [55] W.H.J. Boesten, D.H.N. Dassen, H.E .. [31] C. Syldatk, R. MUlier, M. Siemann, K. Shoemaker, Dutch Appl. 8.501.093,1985. Krohn, F. Wagner, in 'Biocatalytic Pro- [56] S. Tokuyama, H. Miya, K. Hatano, T. duction of Amino Acids and Derivatives', Takahashi, Appl. Microbiol. Biotechnol. Eds. D. Rozzell, F. Wagner, Chap. 5, 1994, 40, 835-840. Hanser, Munich, 1992,75-128. [57] S. Tokuyama, K. Hatano, Appl. Micro- [32] O. Keil, M.P. Schneider, P. Razor, Tetrah- biol. Biotechnol. 1995,42,853-859. dron Asymm. 1995, 6, 1257-1260. [58] S. Tokuyama, K. Hatano, Appl. Micro- [33] C. Syldatk, R. MUller, M. Pietzsch, F. bioi. Biotechnol. 1995,42,884-889. Wagner, in 'Biocatalytic Production of [59] S. Tokuyama, K. Hatano, Appl. Micro- Amino Acids and Derivatives', Eds. D. bioi. Biotechnol. 1996,44,774-777. Rozzell, F. Wagner, Chap. 6, Hanser, [60] S. Verseck, AS. Bommarius, K. Drauz, Munich, 1992, ]29-176. M.-R. Kula, DE 19935268.2 (to Degus- [34] O. May, P.T. Nguyen, F.H. Arnold, Nat. sa-Htils); S. Verseck, AS. Bommarius, Biotechnol. 2000,18, 3]7-320. M.-R. Kula, Appl. Microbiol. Biotechnol. [35] C. Syldatk, M. Pietzsch, in 'Enzyme 2001, in press. Catalysis in Organic Synthesis', Eds. K. [61] D.R. Palmer, 1.B. Garret, V. Sharma, R. Drauz, H. Waldmann, VCH, Weinheim, Meganathan, P.C. Babbitt, 1.A. Gerlt, Bio- Basel, Deerfield Beach, 1995, 409-430. chemistry 1999, 38, 4252-4258. [36] W. Hummel, M.-R. Kula, Eur. J. Bio- [62] A.S. Bommarius, AS., K. Drauz, W. chern. 1989, 184, ]-13. Hummel, M.-R. Kula, C. Wandrey, Bio- [37] T. Ohshima, K. Soda, Adv. Biochem. catalysis 1994, 10, 37-48. Eng./Biotech. 1990,42, 187-209. [63] 1.0. Rozzell, U.S. Patent 4518692,1989. [38] N.M.W. Brunhuber. J.S. Blanchard, Crit. [64] J.D. Rozzell, Chimica Oggi 1999, 42-47. Rev. in Biochem. and Mol. Biol. 1994, [65] U. Kragl, D. Vasic-Racki, C. Wandrey, 29(6),415-467. Bioproc. Eng. 1996,14,291-297. [39] R. Wichmann, C. Wandrey, A.F. Buck- [66] AS. Bommarius, Habilitationsschrift, mann, M.-R. Kula, Biotechnol. Bioeng. RWTH Aachen, 2000. 1981,23,2789-2802. [67] L Gentzen, H.-G. Laffler, F. Schneider, in [40] M.-R. Kula, C. Wandrey, Meth. Enzymol. 'Metalloproteins', Ed.: U. Eser, Thieme, 1988, 136, 34-45. Stuttgart, 1979, 270-274. [41] A.S. Bommarius, M. Schwarm, K. Stingl, [68] I. Gentzen, H.-G. Loffler, F. Schneider, Z. M. Kottenhahn, K. Huthmacher, K. Naturforsch. 1980, 35c, 544-550. Drauz, Tetrah. Asymm. 1995, 6, 285]- [69] A.S. Bommarius, in 'Bioprocessing', Ed. 2888. G.N. Stephanopoulos, Vol. 3, Chap. 17, in [42] P.P. Taylor, D.P. Pantaleone, R.F. Senk- Series 'Biotechnology', Series Eds. H.-J. peil, LG. Fotheringham, T1BTECH 1998, Rehm, G. Reed, 2. Ed., VCH, Weinheim, 16,412-418. Basel, Deerfield Beach, 1993, 427-466. [43] W. Leuchtenberger, in 'Enzyme Catalysis [70] U. Kragl, D. Vasic-Racki, C. Wandrey, in Organic Synthesis', Eds. K. Drauz, H. Chern. Ing. Tech. 1992,64 (6), 499-509. Waldmann, VCH, Weinheim, Basel, Deer- [71] 1.0. Rozzell, AS. Bommarius, in 'En- field Beach, 1995,470-479. zyme Catalysis in Organic Synthesis', [44] D.J. Ager, I.G. Fotheringham, T. Li, D.P. Eds. K. Drauz, H. Waldmann, S. Roberts, Pantaleone, R.F. Senkpeil, Enantiomer Wiley-VCH, Weinheim 2001. 2000,5, 235-43. [72] J.L. Iborra, J.M. Obon, A. Manjon, M. [45] W.H. Kruizing, J. Bolster, R.M. Kellogg, Canovas, Biotechnol. Appl. Biochem. J. Kamphuis, W.H.J. Boesten, E.M. Mei- 1992, 15, 22-30. jer, H.E. Shoemaker, J. Org. Chern. 1988, [73] P.B. Weisz, C.D. Prater, Adv. Catal. 1954, 53, ]826-1827. 6, 143-196. [46] A.S. Bommarius, K. Drauz, K. Gunther, [74] O. Levenspiel, 'Chemical Reaction Engi- G. Knaup, M. Schwarm, Tetrah. Asymm. neering', Wiley, New York, 1999, 3rd ed., 1997,8,3197-3200. Chap. 13. [47] N.J. Turner, 1.R. Winterman, R. McCague, J.S. Paratt, S.J.C. Taylor, Tetrahedron Lett. 1995,7, 1I 13-1116. [48] B.D. Sanwal, M.W. Zink, Arch. Biochem. Biophys.1961,94,430-435. [49] J.D. Rozzell, Methods in Enzymology 1987,136,479-497. [50] Y.B. Tewari, R.N. Goldberg, J.D. Rozzell, J. Chern. Thermodyn. 2000, 32, 138]- ]398. [5]] F. Wagner, C. Syldatk, M. Pietzsch, US 5449786 (to Degussa AG), 1998.