NAD(P)H‐Dependent Dehydrogenases for the Asymmetric Reductive Amination of Ketones: Structure, Mechanism, Evolution and Application

Abstract Asymmetric reductive aminations are some of the most important reactions in the preparation of active pharmaceuticals, as chiral amines feature in many of the world's most important drugs. Although many enzymes have been applied to the synthesis of chiral amines, the development of reductive amination reactions that use enzymes is attractive, as it would permit the one‐step transformation of readily available prochiral ketones into chiral amines of high optical purity. However, as most natural “reductive aminase” activities operate on keto acids, and many are able to use only ammonia as the amine donor, there is considerable scope for the engineering of natural enzymes for the reductive amination of ketones, and also for the preparation of secondary amines using alkylamines as donors. This review summarises research into the development of NAD(P)H‐dependent dehydrogenases for the reductive amination of ketones, including amino acid dehydrogenases (AADHs), natural amine dehydrogenases (AmDHs), opine dehydrogenases (OpDHs) and imine reductases (IREDs). In each case knowledge of the structure and mechanism of the enzyme class is addressed, with a further description of the engineering of those enzymes for the reductive amination of ketones towards primary and also secondary amine products.

Abstract: Asymmetricr eductive aminations are some of the most importantr eactions in the preparation of active pharmaceuticals,a sc hiral aminesf eature in many of the worldsm ost important drugs. Although manye nzymes have been applied to the synthesiso fc hiral amines,t he development of reductive amination reactions that use enzymes is attractive,a si tw ould permit the one-stept ransformation of readily available prochiral ketones into chiral amines of high optical purity.H owever, as most natural "reductive aminase"a ctivitieso perate on keto acids,a nd many are able to use only ammonia as the amine donor, there is considerable scope for the engineering of natural enzymes for the reductive amination of ketones,a nd also for the preparation of secondary amines using alkylamines as donors.T his review summarises research into the development of NAD(P)H-dependent dehydrogenases for the reductive amination of ketones,i ncluding amino acid dehydrogenases (AADHs), natural amine dehydrogenases (AmDHs), opine dehydrogenases (OpDHs) and imine reductases( IREDs). In each case knowledge of the structurea nd mechanism of the enzyme class is addressed, with af urther description of the engineering of those enzymes for the reductiveamination of ketones towards primary anda lso secondary amine products.

1I ntroduction
Reductive amination, or the conversion of ac arbonyl group to an amine via an iminiumi on intermediate (Scheme 1), is one of the most important reactions for synthesising chiral amines,afunctional group that features in ac onsiderable proportion of small biologically active molecules.
Mahima Sharma is ap ostdoctoral research associateb ased in the group of Prof.G ideon Grogan at the York Structural Biology Laboratory,U niversity of York since May 2015. She completed her DPhil in chemical biology from the University of Oxford in 2015, working under the supervision of Prof. Benjamin G. Davis at the Department of Chemistry,w here she investigated the design of artificial metalloenzymes for C-C cross-coupling reactions.H er research interests include structure determination andengineering of enzymes for biocatalytic purposes.H er current project is focussedu pond iscoveringe nzymes enabling chiral amine synthesis,i np articular imine reductases (IREDs), and undertaking structural andb iochemical studies thereof.
Juan Mangas-Sanchez obtained his Ph.D.a tt he University of Oviedo (Spain) under the supervision of VicenteG otor-Fernandez working on newb iocatalytic routes to synthesise optically active alcoholse mployinga lcohol dehydrogenases and lipases.T henh em oved to Lund Universityi nS weden for twoy ears to worki nP rof.P atrick Adlercreutzs group on the optimisation of chemoenzymatic processes to obtain biodiesel, tailored triglycerides and prebiotics using hydrolases. Fort he last two years, he hasb een working as ar esearch associate at the Manchester Institute of Biotechnology in the group of Prof.N icholas Tu rner. His researchi nterests focus on the discovery,e ngineering, characterisation and applications of novelb iocatalystsf or the production of amines. Moreover, an increasing amount of research is being directedt owards the development of asymmetric processes for reductive amination, [1,2] as,i nm any cases,t he stereocentre bearing the amine is crucial in determining its biological activity.Examples of abiotic asymmetric reductivea mination include organometallic catalysis using different transitionm etal complexes, such as ruthenium, rhodium and iridium, employinge ither H 2 [3,4] or other reducing agents. [5] These techniques have been extensively applied and are well-established in industry.H owever, they also present serious environmental and safety issues due to the use of transitionm etals and hydrogeng as,f requently under high pressures.O rganocatalytic approaches [6] have also been employed using either hydrosilanes [7] or Hantzsch esters [8] as the hydrogens ource,u tilising chiral Brønsted acid species such as Akiyama-Terada catalysts to induce chirality.
When the optical purityo fp roductsi so fp aramounti mportance,t hen biocatalytic routes to these compounds also command consideration. [9] Thel ist of enzymest hat are used to catalyse the synthesiso f chiral amines now stretches from hydrolases [10] for the resolution of N-acylamines,t hrought of lavin-dependent monoamine oxidases (MAOs) for the deracemisation of chiral amines, [11] and w-transaminases (w-TAs), [12] which are able to synthesise chiral amines from ketones at the expense of an ammonia donor, such as alanine or isopropylamine.T he last example is interesting from the current perspective,i nt hat w-TAsc atalyse formal reductivea mination reactions,a lthoughn or eductivec hemistry is involved. However, despite many important examples of the application of w-TAs in amine production, the enzymes have limitationst hat are dictatedb yt heir mechanism.I n at ransamination, ammoniaf rom the donor is transferredt ot he enzyme cofactor pyridoxal 5-phosphate (PLP), which retainst he ammonia within the modified cofactorp yridoxamine 5-phosphate (PMP). The ammonia is transferred to the co-substrateo ft he reaction, in this case the prochiral ketone that will be converted into an amine.O nly transfer of ammoniai s permitted by this system, and, hence,t he syntheses using transaminases are limited to the synthesisofprimary amines.T he enzymatic preparation of chiral secondary amines has been achieved using MAOs, [11] althougharacemica mine itself is the starting material. Recenta dvances in imine reductase (IRED) biocatalysis, [13,14,15] addressed below,a lso permit the preparation of chiral secondary amines from preformed prochiral imines.H owever, in those cases,t he necessary preparation of imine substrates itself presents ac onstraint. Ap referredr eactionw ouldb et he enzymatic reductivea mination of ap rochiral ketone,i nw hich an enzyme wouldc atalyse bondf ormation between the ketonea nd amine,a nd subsequent reduction of the iminiumi on intermediate.T he mechanism of at rue reductive aminase enzyme would mirror that of the abiotic equivalent in which an enzyme would catalyse the coupling of ketone anda mine substrates, presented in a1 :1 ratio,t of orm ac arbinolamine intermediate,f rom which waterw as eliminated to form ai miniumi on ( Figure 1). Thee nzyme would then be able to asymmetrically reduce the prochiral iminium ion using an icotinamide cofactor[ NAD(P)H] to generate ac hiral amine product. Such an enzyme would also require am echanism to prevent reductiono ft he keto precursor to the alcohol. Reductive Aminases (RedAms) competent for the true reductivea mination of ketones,w ould therefore be extremelyu seful additionst ot he selection of biocatalytic agents for the production of chiral amines.
In this review we describe the inspiration for the design of such catalysts based on natural reductive amination enzymes that act on keto acid substrates. We also discuss examples in which native activities can be employed in reductive amination processes, and discuss how these might be evolved for improved RedAm activity.I ne ach case the structure,m echanism, and the applicationo ft he enzyme to amine synthesis,are consideredint he description.

2A mino Acid Dehydrogenases and their Evolution forthe ReductiveA mination of Ketones
Amino acid dehydrogenases( AADHs,E .C.1 .4.1.X) catalyse the NAD(P)H-dependent interconversion of keto and amino acids (Scheme 2). [16] In the amination direction, the amino acid is formed through the enzyme-catalysed coupling of the relevant keto acid and ammonia,f ollowed by reductiono ft he imino acid intermediate using hydride supplied by the cofactor. Av ast history exists in the literature on the biochemistry,b ut also on the applications of AADHs, [17,18] and manym echanistic studies on the chemistry through which the reductivea mination is achieved, have been published.
Further activities towards a-ketobutyrate and a-ketovalerate have been reported in homologues from Bacillus sphaericus [23] and Bacillus stearothermophilus. [24] Native LeuDHh as an established history in industrial biotransformations for the production of tertleucine,a ne ssential component of viralp rotease inhibitors. [25,26] Thes tructureo ft he LeuDHh omologue from B. sphaericus wasd eterminedb yR ice and co-workers (PDB 1LEH). [27] One subunit of the enzyme is 364 amino acids andi so rganised into two domains separated by ad eep cleft in which the cofactor binds (Figure 2A). In the crystal structure two subunits associate to form ad imer,b ut the quaternary structure of LeuDHi st hought to be an octamer in solution. It was thought that domain movements play ar ole in catalysis,w ith ac losure of the domains in the presence of the substrate required to bring the cofactor and substratessufficiently close for hydride transfer.
Althought he authors report ac omplexw ith the cofactor NAD(H), the atom coordinates for this molecule are not contained within the relevant PDBf ile. However, the report details the determinants of cofactor binding within the binding cleft, and ac omparison with the structures of phenylalanine dehydrogenase (vide infra)b ound to both NAD(H) andp henylalanine permits am odel to be constructed that places both NAD(H) and l-leucine within the active site ( Figure 2B).
Scheme3.Activity of wild-type LeuDH. Figure 2. A:S tructure of monomer of LeuDH( 1LEH), [27] with NADH (carbon atoms in grey) and l-leucine (carbon atoms in yellow)m odelled within the active site. B:D etail of active site of LeuDH showing significant interactions between active site side-chains and l-leucine,and residues targeted for mutation in directed evolutionstudies.

Scheme 2. Generalr eaction catalysed by amino acid dehydrogenases (AADHs).
Them odel suggestst hat l-leucinei ss ecured in the active site through the interaction of its amino group with aspartic acid residue D115 andi ts carboxylate with the side-chains of two lysine residuesK 68 and K80 ( Figure 2B). K80i sh ighly conserved amongst AADH sequences.A ni ncrease in K m for a-keto isocaproate,f rom 0.9 mM for the wild-type,t o9 .8 mM and 25 mM for mutants K80A andK 80Q, respectively,d etermined by Sekimoto andc o-workers, [28] was thought to be consistent with ar ole for K80i nk eto acid substrate binding. The K m for l-leucinea gain increasedf rom 5.1 mM for the wild-type to 17 mM for K80Q,but decreased slightly to 3.7 mM for K80A. Sekimotoa nd co-workers also determined that K80 acts as ag eneral base in catalysis,m ost likely involved in the activation of water for attack at the electrophilic carbon of the imininum ion in the deamination direction. Them odel mayb ec onsidered in conjunction with ap roposal for the oxidative deamination of leucine put forward by Sekimoto and co-workers shown in Figure 3. [28] Hydride abstraction from leucine (I) by NAD + + first creates an iminiumi on (II). K80 actsa s ab ase, activating aw ater molecule for attack at the electrophilic carbon atom, to form the carbinolamine intermediate (III). Intramolecular proton transfer from the intermediate hydroxyl to the amine formsa n oxyanion (IV), stabilised by the side chain of K80, and the ammonia leaving group,w hich departs to leave the keto acid product a-ketoisocaproate (V).

Structure-Guided Evolution of LeuDHfor the Reductive Amination of Ketones
Thea ctivity of LeuDHw as to provide the major inspiration for the engineering, by directede volution of an "amine dehydrogenase" (AmDH) enzyme that wouldb ec apable of reductive amination of the ketonea nalogue of a-ketoisocaproate 1,m ethyl isobutyl ketone 3 (MIBK, Scheme 4). [29] Bommariusa nd co-workers targeted the determinant residues of carboxylate recognition in aL euDH from Bacillus stearothermophilus for mutation. Al-thoughK 80 had to be retained, as it was essential for the mechanism, as aturation mutagenesis library of K68 yielded av ariant, K68M, that possessed low but measurable reductivea mination activity towards MIBK. [29] Other residues, homologous with those of LeuDHf rom B. sphaericus illustrated in Figure 2B, were targeted in am utagenesis strategy in which libraries of variants were screened for improved AmDH activity towards 3.T he most active variants Figure 3. Proposed mechanism of deaminationo fl-leucine by LeuDHa dapted from Sekimoto and co-workers. [28] I: l-leucine; II:i minium ion; III:carbinolamine; IV:oxyanion; V: a-ketoisocaproate.
Scheme4.Activity of AmDH evolved from LeuDH. [29] . resulting from the screen included those with mutations to positions K68, D114 and V291, but also N261, which is involvedi nt he recognition of the cofactor NADH. Thes uperior variant overall was K68S/ E114V/N261L/V291C, which displayed a k cat value of 0.46 s À1 for the transformation of 3 to (R)-1,3-dimethylbutylamine (1,3-DMBA) 4 with 92.5% conversion and 99.8% ee,a nd with negligible residual activity towards l-leucine. Activityw as also recorded towards other ketones,i ncluding cyclohexanone and acetophenone.

Structure andMechanism of PheDH
Furtherd evelopmentsi nt his area were achieved throughe volution of another AADH, phenyalanine dehydrogenase (PheDH, E.C. 1.4.1.20) from Bacillus badius, [30] which catalyses the interconversion of lphenylalanine and phenylpyruvate, again using NADH as the cofactor.P heDH shares 48% sequence identity with LeuDH,a nd as tructure of ab acterial enzyme from Rhodococcus sp.M 4, upon which engineering has been based, was determined by Holden, Thoden and co-workers. [31,32] Thestructureisvery similar to that of LeuDH, consisting of two domains,s eparated by al arge cleft.I nc ontrastt oL euDHh owever, structures have been obtained in which both cofactor and the substrates of amination and deamination reactions,p henylpyruvatea nd l-phenylalanine (1BW9,1 C1D)r espectively,p henyllactate (1C1X) and an inhibitor, b-phenylpropionate (1BXG),h ave been trapped in the active site.T he catalytic activity of PheDHwas again believed to be enabled by arelative closure of domains over the active site,i no rder to bring the substrate and cofactor into sufficiently close proximity for reaction.
Ar epresentation of the active site of Rhodococcus sp.P heDH in complexw ith l-Phei ss hown in Figure 4. Thed eterminants of amino acid binding are similart ot hat of LeuDH,w ith the amino group of the substrate secured by D118 andt he carboxylate makingi nteractions with the side-chains of two lysine residues,K 66 and K78, and also N262. However, kinetic measurements using the inhibitor 3-phenylpropionate suggest that the interaction of the amino group with D118 is not essential for substrate binding. [32] Thep resence of the ligandw as accompanied by the identification of aw ater molecule in the structure, in close proximity to the catalytic K68s ide-chain, and thought to be the one involved in formationo ft he carbinolaminei ntermediate.T he mechanism in the deamination direction wasd efineda s" orderedb iter", in which binding of the cofactorN AD + + was followed by that of l-Phe,f ollowed by ordered release of ammonium ions,phenylpyruvatea nd NADH.
Detailedk inetic measurements led to am echanistic proposalf or the deamination, which was similar to that proposed for LeuDH ( Figure 5), [32] although with many refinements,e specially with respect to proton transfer steps.I nt he first case,t he substrate is suggested to bind in the zwitterionic form, with the amine in the protonated state.K 78 activates the catalytic water molecule for deprotonation of the ammonium group (I), afterw hich hydride is transferred to NAD + + (II)w ith formation of the iminiumi on (III). K78 then activates waterf or attack at the iminium carbon (III), giving the carbinolamine intermediate (IV). K78now acts as ab ase,d eprotonating the carbinolamine (V), as D118 donates ap roton to the amine leaving group.T his leaves the phenylpyruvatep roduct with its carbonyl group bound to the side chain of K78 (VI)I ntriguingly, ac omparison of the structures of PheDH in complexw ith phenyllactate and l-Phe permitt he authors to propose am echanism whereby the keto acid substrate of reductive amination is prevented from beingreduced to phenyllactate by the cofactor. In the phenyllactate complex, binding of the keto group of the ligand to the side-chain of K78 positions the carbonyl carbon 5.1 ngstromf rom the C-4 of NAD(H), is too far for hydride deliveryt oo ccur between them.

Evolution of PheDHfor the Reductive Amination of Ketones
Bommarius and co-workers followed the in vitro evolution of LeuDHw ith experiments designed to evolve Figure4.Active site of PheDH from Rhodococcus sp.M 4i n complex with l-Phe and NAD( PDB code 1C1D) [32] showing interactions of the substrate with active site residues and also the catalytic water molecule.
PheDH for reductive amination of ketones. [30] Starting with the successfulL euDHm utations as ag uide, aP heDH from Bacillus badius was mutated to give the double variantK 77M/N276V,w hich displayed increaseda ctivity towards MIBK over the wild-type enzyme,b ut also for p-fluorophenylacetone. [30] Af ocused mutagenesis strategy, targeting these two positions furnished an improved variant K77S/N276L that displayed a k cat value 15-fold greater than that observed for the LeuDHa mine dehydrogenase mutant. Then ew variant( F-AmDH) displayed reductivea mination activity towards ar ange of ketones including phenoxy-2-propanone,2-hexanone and 3-methyl-2-butanone.
Thet ransformation of p-fluorophenylacetone 5 gave the amine product with 93.8% conversion (73.9% isolated yield) and > 99.8% ee (Scheme5). F-AmDH was also compatible with the NADH recycling system of glucose and glucose dehydrogenase.
Further functionality in the newa mine dehydrogenase scaffold was achieved by combiningt he N-termi-nal section (residues 1-149) of the F-AmDH with residues 140-166ofthe LeuDHvarianttogive achimeric enzyme,c FL1-AmDH that displayed activity towards acetophenone andadamantyl methyl ketone,a nd converted these to (R)-amine products with excellent enantioselectivity. [33] In af urther refinement, two adjacent asparagine residuesi nt he chimera, N270 and N271, were mutated to leucine in an effortt oa ffect amination activity.T he newv ariant displayed increasedank cat value towards p-fluoroacetophenone.
Some of the lower activitieso fF -AmDH were attributed to the poor solubility of substrates in aqueous solution. This prompted the formulation of ab iphasic reactions ystemi nw hich the enzyme operated in a1:4 ratio of heptane to water. [34] In such systems, the volumetric productivity of the F-AmDH-catalysed amination of p-fluorophenylacetonewas doubled.
In subsequent work by the group of Li, the Rhodococcus sp.P heDH was evolved for the asymmetric reductive amination of ketones. [35] In this report, the structure of Rhodococcus sp.P heDH again suggested that mutation of K66 and N262, known to interact with the natural substratesc arboxylate group,c ould be targeted in an effort to generate an enzyme that transformed ketone substrates.Afocused mutagenesis strategyr esulted in at riple mutant enzyme,K 66Q/ S149G/N262C,w hich catalysed the amination of phenylacetone 7 and 4-phenyl-2-butanone 9 to (R)-amphetamine 8 and (R)-1-methyl-3-phenylpropylamine 10 respectively, each with > 98% ee (Scheme6). Scheme 5. Reductive amination of p-fluorophenylacetone by F-AmDH, with cofactor recycling. [30] .
In addition to the established mutations changing the interaction with the substratesc arboxylate group, the addition of mutation S149G wast hought to enlarge the size of the entrance to the substrate binding pocket. The Rhodococcus sp.a mine dehydrogenase also proved to be compatible with the glucose/glucose dehydrogenase NADH recycling system, and 15 mM of 9 were transformed to (R)-1-methyl-3-phenylpropylamine 10 after6 0h,w ith 95.2%c onversion to the product.F urther workb yt his group saw the AmDH co-immobilised with GDH on magnetic nanoparticles (MNPs), [36] giving as uperior system for the asymmetric reductivea mination of 9 with at otal turnover number for NADH recycling of 2940. As ubsequent report by Mutti and co-workers revisitedt he Rhodococcus sp.M 4a mine dehydrogenase,a nd showed that, in addition to substrates 7 and 9,t he enzyme was also effective in transforming o-methoxyphenylacetone derivatives,a liphatic ketones,s uch as 2-octanone (99% conversion) anda lso "bulky-bulky" ketones such as 1-phenyl-butan-2-one( > 99%), 1-phenylpentan-2-one (71%) and 1-phenylpentan-3-one (83%). [37] Moreover, the Rhodococcus sp.A mDH provedt ob ec ompatible with the more atom-efficient formate dehydrogenase as am ethod for recycling the NAD + + cofactor. In ap reparative-scalee xperiment, 208 mg of (p-methoxyphenyl)acetone 11 were converted to the (R)-amine 12 with an 82% isolated yield and with > 99% ee in ammonium formate buffer (Scheme 6).

Application of AmDHs in Cascades
TheF -AmDH variant of PheDH hasb een applied in cascade reactions with alcohol dehydrogenases (ADHs) that permitt he conversion of alcohol substrates into amine productsw ith "closed-loop" recycling of the cofactor( Scheme 7). [38] In these experiments,A DHs from either Aromatoleum aromaticum or Lactobacillus brevis,w hich have stereocomplementary selectivities,w ere combined with the Bacillus badius PheDH variantK 78S/N277L in ammonium chloride buffera tp H8.7,w ith an alcohol substrate concentration of 20 mM and NAD + + cofactora t1mM. After 24 h, 85% conversion of the alcohol 13 to the (R)-amine 8 was achieved, with > 99% ee.
Thet echnique was applicable to aw ide range of 1phenyl-2-propanold erivatives.I nterestingly,t his "hydrogenb orrowing" systemp ermits the use of (S)-, (R)-or racemic alcohol substrates, depending on which ADH, or ac ombination of both, is used to catalyse the alcohol oxidation.
As imilar cascade was later reported by Xu andc oworkers. [39] In this case,aLeuDHh omologue from Exiguobacterium sibiricum (EsLeuDH) was engineered to contain equivalentm utations (K77S/ N270L)t ot hose engineered into the B. stearothermophilus LeuDHb yB ommariusa nd co-workers. [29] This variantw as coupled with an ADH from Streptomyces coelicolor that was selected on the basis of its lack of stereoselectivity for racemic substrate alcohols. A range of racemic alcohols,i ncluding 2-pentanol and sec-phenylethanol, was converted to (R)-amine products with 94% and 21% conversion, respectively,a nd in up to > 99% ee.

3O pine Dehydrogenases and their in vitro Evolution for the Reductive Aminationo fK etones
Opine dehydrogenases (OpDHs,E .C.1.5.1.28) [41] are ac lass of AADHs that catalyse the reversible reductive coupling of amino acid and keto acids to form Nderivatised amino acids called opines.T he model reaction catalysed by these enzymes in the BRENDA database (http://www.brenda-enzymes.org) is the NADH-dependent coupling of aminopentanoic acid 15 and pyruvate 16 into (2S)-2-{[1-(R)-carboxyethyl]-amino}pentanoate 17 (Scheme 9). They can be found in bacteria such as Agrobacterium, [42] where they can be flavin dependent,a nd known to catalyse the formation of opines that form in crown gallt umours in infectedp lants,b ut also in higher organisms,s uch as molluscsa nd sponges,i nw hich they are NAD(P)Hdependent, anda re thought to have ar ole in maintaining glycolytic flux during hypoxicc onditions. [43] They have aroused interest for applications in preparative biocatalysis,a s, unusually amongst AADHs, they are able to couple ac arbonylc ompound with an amine that is larger than ammonia, with possible consequences for the formationo fs econdary amines by engineered variants.
Thee nzyme -C ENDH -c atalysed the formation of 19 from l-phenylalanine, sodium pyruvate and NADH as cofactor.O ther hydrophobic l-amino acids including l-methionine,w ere tolerated as substrates, as well as d-leucine,a lthoughw ith only 3.4% of the activity observed with l-Met. Using l-Met as the amino acid substrate,t he enzyme also accepted keto acids such as oxaloacetate and, to al essere xtent glyoxylate.S ubsequent cloning of the gene revealed that it encoded apolypeptide of 359 amino acids. [45] Thes tructure of the enzyme was determined by Asano,R ice and co-workers, [46] and, in common with some other AADHs,r evealed at wo-domains tructure with ac left at the interface in which the nicotinamide cofactorw as observed. Further information has been obtained from as tructureo ft he relatedO pDH, octopine dehydrogenase,f rom the great scallop Pecten maximus ( Figure 6A), as separate complexes have been obtained with the amino acids arginine and also pyruvate. [47] Thesec omplexes have been used, along with NMR [48] andI TC measurements, [49] to begin to describe amechanism for OpDH activity.
In the l-arginine complex, among other interactions,t he alpha-carboxylate is bound to H212;t he guanidinium group to the side-chaino fE 142 (Figure 6B). [47] As these residues stem from the different domains it is thought that l-arginineb inding stimulates ar elativec losure of the domains to prime the enzyme for the coupling reaction. In the pyruvate complex, the carboxylate of pyruvateb inds to the side-chain of Q118 andH 212 makes an H-bond with the carbonylg roup.T he carbonylg roup is distant Scheme9.Reductive amination reactions catalysed by opine dehydrogenases OpDH such as CENDHf rom Arthrobacter sp.strain 1C. [44] . Scheme 8. Reductive aminationo f4 -oxopentanoic acid by naturally occurring amine dehydrogenase from Petrotoga mobilis. [40] . from the NADH cofactor, meaning that reductiono f pyruvate to lactate is not observed. Mutation of Q118 to alanine gave avariantofmarkedlyreducedactivity. As equential binding order for these two substrates was proposedb ased on these observations.Asuperimpositiono fl-arginine and pyruvate complexes reveals little difference in the orientation of Q118 and H212 and permits observation of the two ligands in their binding modes ( Figure 6B). However, each ligand is somewhat distant from the nicotinamide ring of the NADPH cofactor, suggesting that the active conformation, in which the imine bondb etween the substrate partners is reducedb yh ydride from the C4 atom of that ring, must incorporate different ligand binding modes.
Further evidence for the sequential, orderedb inding of l-argininea nd pyruvate came from NMR studies. [48] By monitoring five randomly selected peaks in 15 N-1 H-TROSY spectra of 15 Nl abelled apo-protein, it was observedt hat the binding of NADH resulted in perturbations to these chemical shifts. Further perturbations were observed upon the addition of l-arginine to ap re-formed OpDH-NADHc omplex, but not when pyruvate wasa dded to the same.T his was strongly suggestive of l-arginine as the second substrate,a fter NADH, to bind in the OpDH coupling reaction. Additional support for the mechanism was providedb yi sothermal calorimetry studies, [49] which again showedm easurable enthalpy changes upont he addition of l-arginine to the OpDH-NADHc omplex, but no binding when the same complex was presented with pyruvate. Interestingly,t he order of "amine" and "ketone"p artner binding in this reductive amination is therefore reversed from that which is observed in LeuDHa nd PheDH. However, although each of these experiments on OpDH sheds light on the binding order of substrates in the OpDH mechanism, no description of the chemical mechanism and the role of individual amino acid side-chains has been reported at this stage.

Application of OpDH
Following the discoveryo ft he OpDH from Arthrobacter sp.s train 1C,t his wasa pplied to the preparative formation of as econdary amine carboxylic acid by Asano and co-workers. [50] In addition to its native substrates, l-phenyalanine and pyruvate,s hortc hain neutral amino acids including (S)-enantiomers of 2aminobutyric acid, norvaline, norleucine and phenylglycinew ere accepted as the amino acid partner substrate,w ith yields of between 96 and > 99%, although (S)-methionine was preferred. Interestingly (S)-phenylalaninol, with no carboxylate group,w as also converted quantitatively by the enzyme,a ne arlyi ndication of the promiscuous activity of the enzyme for amine substrates.G lyoxylic acid and 2-ketobutyric acid were accepted in addition to pyruvic acid as keto acceptors.

Directed Evolution of OpDHs for the Reductive Amination of Ketones
Thea bility of OpDH and related enzymes to couple keto acids with amine partners larger than ammonia stimulated efforts on the part of Codexist oe ngineer these enzymesf or the coupling of non-carboxylated ketones and amines. [51] Using the CENDH from Arthrobacter sp.s train 1C as as tarting platform, an extensivep rogramme of directede volution experiments resulted in the first instance in CENDH variants that were not only capable of coupling pyruvate 17 to lnorvaline 20 to give (2S)-2-[1-(carboxyethyl)amino]pentanoic acid 21,b ut also to the non-carboxylated 1butylamine 22 to give 2-(butylamino)propanoic acid 23 (Scheme 10). Furtherv ariants catalysed the reductive amination of cyclohexanone,2 -pentanone and 2tetralone derivatives to av ariety of amine partners to give secondary amine products( Scheme 10) with enantiomeric excesses,w here relevant, of up to or greatert han 99.5% ee One variant, which was ranked first for the coupling of cyclohexanonea nd butylamine,f eatureds even amino acid mutations (A111M/ K156T/N198H/Y259M/Y280L/R292V/Y293H).
An examination of these mutations in the context of the CENDH structure,r eveals that the mutation sites were all within the active site cleft (Figure 7).
From these studies it is cleart hat the OpDH structure mayb ee ngineered for the catalysis of reductive amination reactions with ketone and amine partners, but more informationo nm echanism, and also the interaction of active site residues with the substrates, wouldassist in informing further engineering.
Scheme10. Selected reductivea minations catalysed by CENDH variants created using directed evolution. [51] Detailed information on the enantiopurity of relevant products was not available. Figure 7. Active site of CENDH( 1BG6), [46] showing some of the active siter esidues targeted for mutation in directed evolution libraries leading to variants competentf or the reductive aminationo fk etones. [51] Thep osition of NADH and pyruvate has been fixed using the structure of OpDH from Pecten maximus (3C7D). [47] .

Structure of IREDs
IREDsh ave been shown [54,55,56] to possess ad imeric structure in which one monomer of approximately 30 kDa is intimately associated with its partner to form an active site through domains haring. The active site is found at the interface of the N-terminal Rossman fold of one monomer,a nd the C-terminal bundleofits neighbour ( Figure 8A). Thet wo domains of each monomer are connected by al ong interdomainhelix, and the dimer structure is thought to be flexible throughout the reactionc oordinate,w ith significantc losure observed between domains upon substrate binding. Thes tructures of IREDs are most sim-ilar to those of hydroxyisobutyrate dehydrogenases (HIBDHs), typifiedb yP DB code 2CVZ, [57] although domains haring is not observed in the latter enzymes. In HIBDHs, ac atalytic lysine in the active site is thought to protonate the nascentalcohol in the reductive direction. [57] When the structures of HIBDHsa nd IREDsa re superimposed, the lysine of 2CVZ is observed to overlap with different residues according to the IREDs tudied,b eing in some cases aspartica cid or tyrosine.Arole for these residues in protonation of amine productsi nI RED-catalysed reactions has been suggested, [54,58] andi ndeed their mutation to alanine leads to enzymes of poor activity,a lthough their role in mechanism hasy et to be confirmed by structural studies.Ar ecent structure of the IRED from Amycolatopsiso rientalis is the first in which the amine product (R)-1-methyl-tetrahydroisoquinoline (R-MTQ) has been observed in the presenceo ft he cofactor( Figure 8B). [59] In this complex, the active site is indeed more closed than has previouslyb een observedi no ther IREDss tructures determinedi nt he absence of al igand, and the amine makes bonding interactions with residue side-chains in the active site including Y179 and N241. This complex is at least suggestive of ar ole for these residues in liganda nchoring, and places the electrophilic carbon of the amine within as uitable distance of the C-4 of the nicotinamide ring of NADPH for hydride exchange.T he ligand is somewhat distant from the pendant asparagine residue N171 that is the structural equivalento ft he catalytic lysine in HIBDH however, so the significance of the residue in this position, at least for the reduction of pre-formed iminesbyI REDs, remains unclear.

Reductive Amination Reactions using IREDs
Thep otential of IREDs for reductivea mination reactions was first described by Müllera nd co-workers, [55] Scheme 11. Enantiocomplementary reductions of 2-methyl-1-pyrroline by (S)-and (R)-IREDs from Streptomyces sp. GF3546 and GF5387, respectively. [52,53] .  [59] . who applied the (S)-selective IREDf rom Streptomyces sp.G F3546t ot he asymmetric reductivea mination of 4-phenyl-2-butanone 9 (Scheme 12), using methylammonium buffer as as ource of methylamine.A lthought he amine equivalents were very high, and al arge amount of enzyme hadt ob ee mployed, ac hiral amine product 10 resulted, albeit with al ow conversion andm odest ee of 8.8% and7 6%, respectively.I tw as not clear however if the IRED catalysed at rue reductive amination, both bringing the ketone and amine together in the active site,f ollowed by NADPH-mediated reduction, or merely reduced ap re-formed imine 30 that had formed owing to the large excess of amine in the reactionm edium.
Nonetheless,t he principle of IREDs being applied to reductive amination processesh ad been established.H auer andc o-workersf ollowed this initial report with as tudy of bimolecular reductive amination reactions in which an IREDf rom Streptosporangium roseum was challenged with benzaldehydei n the presence of ammonia, methylamine or aniline. [60] While reaction rates were slow with one equivalent of the amine,a ni ncrease to 10 or 50 equivalents gave, for example,a73% conversion of benzylaldehyde and methylamine to N-methylbenzylamine after 8hin the latter case.A cetophenone andc yclohexylacetone were coupled with methylamine to givea mine products with 39% and 53% conversion and 87% and 78% ee,r espectively.I ne ach case a5 0-fold excess of amine was required, and improved conversions were obtained at pH 9.0, favouring imine formation. NMR studies suggested that the IRED was very effective at drawing the imine intermediate from solution, as no imine intermediate in the conversion of acetophenone was detected.
Thet heme of IRED-catalysed asymmetric reductive amination was continued by the group at Roche, who had initially described al argef amily of IREDs, with complementarys electivity,c ompetent for the asymmetric reductiono fc yclici mines. [61] This library of enzymes, with nine further additions,w as then screened for an influence on the reductive amination of al ibrary of ketonea nd amine partners including acetophenone,2 -hexanone and cyclohexanone, with amine nucleophiles ammonia,m ethylamine or butylamine. [62] Conversions were observedt ov ary from 10% to > 90%, depending on the substrates,with aminations of acetophenone determinedt ob el ow-yielding. Tw oo ft he superior IREDs, IR_11 and IR_20 were applied to 100 mg scale reactions.I R_11 converted ketone 31 and methylamine to 32 in 71% yield and with 98% de (Scheme 13);I R_20 catalysed the conversion of 31 anda mmonia to 33 in 50% yield and with 94% de In addition,I R_20 catalysed the reductive amination of 34 with methylamine to form 35 with 55% yield and9 6% ee In each case the amine was suppliedi n1 2.5 molar equivalents andt he reactions were again carriedo ut at ap Ho f9 .3, to favour imine formation. Low activitiesw ere again attributed to the lowi nstance of imine formation in aqueous solution,a nd again, littleo rn one of the imine intermediates was detectableu sing NMR.
with 58% yield and 90% ee In addition, another IRED, IR-Sip was used in the production of the (S)enantiomer of rasagiline in 81% yield and 72% ee In each case,t he reactionw as buffereda tp H9.0 and a4 0-fold excesso fa mine was required, again suggestive of the pre-formation of the imine substrate as ap rerequisite for the activity of the IRED.
Thea pplication of IREDs in reductive amination reactions has thus farb een shown to be useful for directing the asymmetric reduction of pre-formed imines at high pH, but anyr ole of the enzyme in the catalysis of formation of the imine via ag eneral mechanism such as that showni nF igure 1i sf ar from clear. However, the increasing availability of IRED sequences and structures,a nd the availability of highthroughput methodso fe nzyme evolution, suggest that the investigation of improved IREDs for reductive amination reactions willb et he subject of considerable research in the future.

5C onclusions
Reductivea minases wouldb eaconsiderable and potenta ddition to the current portfolio of enzymes available for the preparation of optically active amines.R ecent research suggests that there may be numerous routes to the discovery of such activities, from the engineering of extant amino acid dehydrogenases, including opine dehydrogenases,t hrough to the discoveryo fn atural amine dehydrogenases,a nd the recruitment of imine reductases. In some cases, the mechanisticb asis for the reductivea minationr eaction is clear; in others the role of the enzyme in the reductivea mination process must be further investigated. These studies will help to inform the further engineering of reductive aminase activity for broadened substrate scope,a nd also for process suitability, so that the enzymes might be applied in the industrial synthesisofc hiral amines.