Racemases and epimerases operating through a 1,1-proton transfer mechanism

Racemases and epimerases catalyse changes in the stereochemical configurations of chiral centres and are of interest as model enzymes and as biotechnological tools. They also occupy pivotal positions within metabolic pathways and, hence, many of them are important drug targets. This review summarises the catalytic mechanisms of PLP-dependent, enolase family and cofactor-independent racemases and epimerases operating by a deprotonation/reprotonation (1,1-proton transfer) mechanism and methods for measuring their catalytic activity. Strategies for inhibiting these enzymes are reviewed, as are specific examples of inhibitors. Rational design of inhibitors based on substrates has been extensively explored but there is considerable scope for development of transition-state mimics and covalent inhibitors and for the identification of inhibitors by high-throughput, fragment and virtual screening approaches. The increasing availability of enzyme structures obtained using X-ray crystallography will facilitate development of inhibitors by rational design and fragment screening, whilst protein models will facilitate development of transition-state mimics. and the His-126/Glu-241 pair, with Glu-241 contributed by the second monomer subunit. His-126/Glu-241 pair removes the a -proton of the S -2-methylacyl-CoA substrate whilst Asp-156 protonates enolate residues for R -2-methylacyl-CoA substrate.


Maksims Yevglevskis
Maksims Yevglevskis obtained his BSc in Chemical Engineering from Riga Technical University (Latvia) in 2009 and MRes in Chemical Research from University of Bath in 2011. He then joined Matthew Lloyd's group at the University of Bath, where he obtained a PhD in 2015 working on developing assays and inhibitors of a-methylacyl-CoA racemase (AMACR). This was followed by a three-year PDRA position on the same project. He is currently working at CatSci Ltd in Cardiff and has an interest in enzymes as drug targets and biomarkers.

Introduction
Chirality is at the very heart of Chemical Biology. Proteins, nucleic acids, carbohydrates and many lipids are all chiral molecules, as are the overwhelming majority of their monomer precursors. In addition, many cellular metabolites also contain chiral centres. It is well-known that, for most chiral biomolecules, one particular configuration is preferred; thus proteins contain predominantly chiral amino-acids with L-configuration 1,2 (S-configuration in the Cahn-Ingold-Prelog system 3 except for R-cysteine and achiral glycine). Similarly, carbohydrates are or contain predominantly D-sugars, with L-ascorbic acid (vitamin C) being a well-known exception. An important consequence of the chiral nature of proteins is that, when they interact with other chiral molecules, a diastereomeric situation arises; thus, most proteins will be highly selective for a particular configuration of their interacting partners (substrate, inhibitor, allosteric effector). An important consequence of this is that different stereoisomers of chiral drugs are effectively different drugs, which will generally have different protein targets (enzyme, receptors etc.) and different pharmacokinetics. 4,5 Finally, many drugs are known to undergo metabolic changes of chiral configuration in vivo, 4,5 e.g. ibuprofen and related 'profens' (reviewed in ref. [6][7][8] and mandelic acid. 9,10 In addition the 2-(aryloxy)propanoic acid herbicides undergo changes in chiral configuration which are mediated by soil bacteria. [11][12][13] Notwithstanding the fact that most biological molecules exist overwhelmingly in one stereochemical configuration, there are many examples where minor stereoisomers play an essential role. The most well-known example of this is proteinogenic amino-acids such as alanine and glutamate, which are

Amit Nathubhai
Amit Nathubhai obtained his BSc (hons) and MSc in Biological Chemistry from the University of Leicester and a PhD from the University of Bath with Dr Ian Eggleston. Amit performed postdoctoral studies in Medicinal Chemistry, Biochemistry and in vitro Cell Biology at the University of Bath. He is currently a Senior Lecturer at the University of Sunderland. His research interests lie at the chemistry and biology interface and emphasises development of chemical tools and drug-like molecules to dissect biological mechanisms involved in cancer, fibrosis, diabetes and obesity. He has a keen interest in fresh-water aquatics, and writing and playing music. found in their D-configuration (R-configuration) within bacterial peptidoglycan. [14][15][16][17] In most cases, these minor stereoisomers are not biosynthesised de novo but are obtained by changing the stereochemical configuration of the most abundant isomer into that of the less abundant isomer.
The enzymes which perform these changes in stereochemical configuration are known as racemases and epimerases, which have been shown to have a pivotal position in metabolism, and thus have gained significant interest as drug targets for diseases such as bacterial infections, 14,[18][19][20][21][22][23][24][25] Chagas disease, [26][27][28] cancer, 6,7,18,29 Alzheimer's disease and other dementias, 1,2,30-32 formation of cataracts 1,33 and diabetic retinopathy; 34 racemase levels are also a marker of ischaemic stroke. 35 Inhibition of diaminopimelate epimerase activity also potentiates cephem antibiotic activity by compromising the integrity of the bacterial cell wall. 36 Low activity or concentrations of racemases/epimerases (AMACR, 37 methylmalonyl-CoA epimerase 38 ) are associated with inherited errors in metabolism and may also be associated with stroke and dementia 39 and neurodegenerative diseases, 40 such as Amyotrophic Lateral Sclerosis (ALS, a.k.a. motor neurone disease, Lou Gehrig's disease). Increased methylmalonic acid levels in the aging population (resulting from a decrease in methylmalonyl-CoA epimerase activity) is suggested to promote an aggressive cancer phenotype by upregulation of the SOX4 transcription factor. 41 Increased levels of aspartate/ glutamate racemases protect Salmonella enterica from aminoacrylate metabolic stress. 42 Increased activity of the bifunctional enzyme UDP-N-acetylglucosamine 2-epimerase/N-acetylmannosamine kinase results in sialuria, an extremely rare genetic disorder, while knockout of the corresponding gene is lethal in mice. 43,44 Mutations in this epimerase are linked to hereditary inclusion body myopathy (HIBM). 44,45 In addition, O-ureidoserine racemase is involved in the biosynthesis of the antibiotic D-cycloserine, 46 while a peptide epimerase is found in funnel web spider (Agelenopsis aperta) venom which interconverts two 48 aminoacid peptides differing only in the configuration at a single serine residue (Ser-46). 47 Finally, racemases and epimerases are used in dynamic kinetic resolutions and other biotechnological applications. [48][49][50][51][52][53][54] Racemases and epimerases use several different strategies to bring about changes in stereochemical configuration of their substrates, including the use of radical reactions, [55][56][57][58][59][60] elimination and re-addition of nucleotides 20,61 and the use of redox cofactors. 18,19,[62][63][64] An important example of a 'epimerase' utilising redox cofactors is decaprenylphosphoryl-b-D-ribose epimerase (DprE); however, this is not a true epimerase reaction as the oxidative and reductive reactions are catalysed by separate enzymes (DprE1 and 2, respectively) using different cofactors [flavin adenine dinucleotide (oxidised) and nicotinamide adenine dinucleotide (reduced)]. [65][66][67][68] By far the most common mechanism used by racemases and epimerases is the deprotonation/reprotonation 14,18,20,63 (1,1proton transfer) reaction. These enzymes fall into three classes: those which are pyridoxal 5 0 -phosphate (PLP)-dependent; 5,50,69,70 those which use metal ions (enolase enzymes 14,49,71,72 ); and those which are cofactor-independent (Scheme 1). 6,7,18,20,63 The PLP-dependent enzymes (Scheme 1A) catalyse exchange between PLP in the internal aldimine 1 (catalytic Lys sidechain) and the external aldimine 2 (substrate a-amino group). Deprotonation 69,70 of 2 results in the ylide intermediate 3 which is subsequently reprotonated from the other face to produce the external aldimine 4 with opposite configuration. In contrast, the metal-dependent enzymes, e.g. mandelate racemase, apparently perform a concerted reaction (Scheme 1B; 5-7). Solvent isotope experiments show that label is incorporated into product with little incorporation into recovered substrate, 73,74 which is consistent with a concerted mechanism. However, kinetic isotope effect measurements on mandelate racemase are consistent with a stepwise reaction and a discrete deprotonated intermediate. 75 Finally, most cofactor-independent racemases/ epimerases utilise a concerted mechanism, 5,76-78 as illustrated by glutamate racemase (Scheme 1C; [8][9][10]. However, some cofactor-independent enzymes using substrates with acidic a-protons perform their reactions with a stepwise mechanism via a discrete enolate intermediate (e.g. a-methylacyl-CoA racemase [79][80][81]. Enzymes which use metal ions as Lewis acids (enolase family enzymes) or are cofactor-independent are of particular interest, since they are able to perform the apparently simple 1,1-proton transfer using active site amino-acid residues and thus are model systems for understanding enzymatic reactions in general. Several of these enzymes are also important as drug targets, 6,7,[20][21][22][23] potential drug targets, 84 or are used in biotechnological applications. 48,49,51 This review will consider racemases/ epimerases utilising deprotonation and deprotonation mechanisms, their reactivity and the strategies used to inhibit them. protonated and deprotonated, respectively. 5 These studies also show that protonation of the amino-acid carboxylate further decreases the C a -H pK a value to 6 but crystal structures suggest that this does not occur during the enzyme catalytic cycle. This is contrast to the situation in cofactor-independent racemases/ epimerases, where substrate carboxylate groups are held within a hydrogen-bonding network 20 or transiently protonated during the reaction. 21 The pK a of the external aldimine C a -H in the alanine racemase reaction is estimated to be 9, which is intermediate between those for the catalytic bases, Tyr-265 and Lys-39. 69 The mechanism of some PLP-dependent enzymes, e.g. ornithine decarboxylase, is thought to go via a quinoid intermediate 69,70 resulting from protonation of the pyridoxal nitrogen by a glutamic acid residue. The equivalent residue in alanine racemase is an arginine and the pyridoxal nitrogen is not extensively protonated 69,70 (Scheme 2). Therefore, alanine racemase is thought to catalyse its reaction via a carbanion 3 not a quinoid 11 intermediate. 69 Kinetic isotope effect experiments on alanine racemase are consistent with a carbanion rather than quinoid intermediate. 70,87 The situation is different for enolase-family racemases and epimerases and those which are cofactor-independent. The fundamental problem for these enzymes is how to deprotonate a substrate carbon acid (typical pK a = B21-23 86,88,89 ) using active site bases with pK a values in the range 6-9 without the enhancement afforded by a PLP cofactor. Many racemase/ epimerase substrates possess carboxylic acids (pK a 2-5), which are deprotonated to the negatively charged carboxylate (e.g. substrates of amino-acid racemases/epimerases 20 and methylmalonyl-CoA epimerase 90 ). Consequently, the apparent pK a of the C a -H for these substrates will be B29. 20,86 This effect is illustrated by chemical systems, which show that the pK a for the C a -H of glycine in water is 28.9, while the corresponding pK a for glycine methyl ester is 21.0. 86 Racemases and epimerases utilising a negatively charged substrate generally hold the carboxylate group within a hydrogen-bonding network or ion pair to disperse the negative charge. 20 In some cases, the enzyme also transfers the incoming proton onto the substrate carboxylate group before it is transferred onto the C a of the product (e.g. glutamate racemase 21 ). Scheme 2 Carbanion 3 and quinoid 11 intermediates in the alanine racemase reaction. 69 Scheme 1 Example mechanisms of racemases and epimerases operating by a 1,1-proton transfer mechanism. (A) PLP-dependent amino acid racemases, as shown by alanine racemase; 5,69 (B) Metal-dependent (enolase) enzymes, as shown by mandelate racemase; 74,75 (C) Cofactorindependent racemases as shown by glutamate racemase. 63 Exceptions to this strategy are seen with methylmalonyl-CoA epimerase 90 and mandelate racemase, 91,92 where the carboxylate group is ligated to the active site Co 2+ or Mg 2+ ion which acts as a Lewis acid and diminishes the pK a of the C a -H. 21 Typically the carboxylate group is also held within a hydrogen-bonding network with active-site residues. 5 Some racemase/epimerase substrates also contain further destabilising groups, such as ammonium groups (amino-acid racemases/epimerases), 18,20 amide carbonyl groups (N-succinylamino acid racemases, dipeptide epimerases and other enolase family enzymes 18,49 ) and OH (mandelate racemase, 18,91 various sugar epimerases 18 ). Both ammonium and OH groups are more easily deprotonated than the C a -H. Chemical models 86,93 show that the pK a of the C a -H is diminished by 9-15 units by protonation of an adjacent amine and a number of amino-acid racemases/epimerases, 20,94 including diaminopimelate epimerase and glutamate racemase, appear to protonate the amine of the substrate during the reaction. In the case of mandelate racemase 18,91 and N-succinylaminoacid racemases, 49 the OH or amide carbonyl groups are ligated to active-site metals such as Mg 2+ (mandelate racemase 18,91,92 ) or Co 2+ , Mn 2+ or, occasionally, Mg 2+ (N-succinylamino-acid racemases 49 ). The rates of proton transfer for the deprotonation and reprotonation steps are generally high, with rate constants of the order of 5 Â 10 9 to 100 Â 10 9 M À1 s À1 . 86 Recent analysis 18 of racemase/epimerase crystal structures, obtained in the presence of ligands, suggests that the vast majority of enzymes bind the two substrate stereoisomers using 'mirror-image packing', that is functional groups are held within the same position with the C a -H on opposite sides in the different stereoisomers. In some cases, e.g. amino-acid racemases/epimerases, 20 the positions of the substrate sidechain and functional groups show remarkably small differences in their positions between the stereoisomers. In other cases, e.g. AMACR/MCR 6,7,18,95 which utilises substrates with large hydrophobic side-chains, the different epimers are accommodated by fixing two of the function groups (the methyl group and acyl-CoA moiety in this case) whilst the side-chain is accommodated in discrete binding sites on a hydrophobic surface at the entrance of the active site.
The active-site bases sit immediately adjacent to the C a -H. In the vast majority of cases, the active-site bases are located on both sides of the substrate (the so called 'two-base enzymes'), while, in a few cases (the 'one-base' enzymes), a single activesite base mediates catalysis. 18,20,96 Many racemases/epimerases are dimers, with the active site located at the dimer interface and active-site bases contributed by both subunits; 18,20,95 binding of substrate often triggers movement of the subunits from an 'open' to a 'closed' conformation, moving the active-site bases into position and desolvating the active site. 20,94,97,98 In some enzymes (e.g. glutamate racemase 21 ), this conformational change triggers a change in the conformation of the deprotonating activesite base as part of the pre-activation step which results in protonation of the substrate carboxylate group. It has also been suggested that conformational changes by 'capping domains', which result in the closed form of the racemase, activate the enzyme for catalysis, are important, e.g. in mandelate racemase. 94 In other cases, little or no conformational changes are observed in the protein upon binding of substrate and the enzyme active site is substantially desolvated in the unbound state. 20,95 Active-site bases PLP-dependent enzymes use several different active-site bases. In alanine racemase, these are generally thought to be Tyr-265 and Lys-39 (Fig. 1). 5,69,70,99 Chemical models suggest that the pK a of these active-site bases are increased to B21 (Lys) and B28 (Tyr), respectively, in the hydrophobic active site. 100 This is considerably higher than the experimentally-determined C a -H pK a value of 11 101 and 9.94. 87 Hence, deprotonation of the substrate is expected to be facile. In serine racemase, the corresponding active-site bases are Lys-57 and Ser-82 69 and the experimentally determined external aldimine C a -H pK a value is 9.26. 87 Chemical models suggest that their active site base pK a values will be B21 and 33-39, 100 the latter being extremely high. These pK a values will be modified by hydrogen-bonding networks within the active site, to allow deprotonation of the active-site residues and reprotonation of the carbanionic intermediate (vide supra, Schemes 1 and 2, 3).
The N-succinylamino acid racemases and related enolase enzymes, e.g. O-succinylbenzoate synthase, 102,103 utilise a pair of lysine residues as catalytic bases 49,104 (Fig. 2). Chemical models 100 suggest that the pK a for these lysine residues within the active site will be B21. The C a -H pK a for these substrates  ligated to active-site metals appears not to have been calculated, though studies on other metal-dependent enzymes (mandelate racemase) 101 suggests that this will be B15.
Several different active-site bases are used by the cofactorindependent racemases and epimerases. In most amino-acid racemases/epimerases, both active-site bases are Cys, which act as a thiolate base/thiol acid pair, catalysing deprotonation and deprotonation, 18,20 e.g. Cys-74 and Cys-185 in B. subtilis glutamate racemase 105 (Fig. 3). Cys is favoured as an active-site base in amino-acid racemases and epimerases because it is more easily desolvated and has a lower pK a than Ser or Thr. 106 Chemical models suggest that desolvation raises the pK a of the Cys residue thiol to thiolate conversion to B28, 100 matching the expected pK a of the C a -H of B29. 20,86 This allows deprotonation of the C a -H by the Cys thiolate. In contrast, the pK a values of the active-site Cys residues acting as an acid appear to be B6-7 to enable protonation from the opposite side. This change in pK a appears to be mediated by a dipole on the a-helices bearing the Cys thiol (at least in diaminopimelate epimerase 20,107 ). Exceptions to this rule include aspartate/ glutamate racemase from a pathogenic E. coli strain (Ecl-DER), in which one of the catalytic Cys is replaced by Thr. This enzyme catalyses irreversible conversion of S-Asp to R-Asp, which arises partly because of differences in the pK a values of the Cys and Thr side-chains and partly because of differences in the distance between the C a -H and the catalytic bases on either side of the substrate. 20,108,109 Similarly, MMP0739 aspartate/glutamate racemase from Methanococcus maripaludis possesses active-site Cys and Thr residues and is predicted to catalyse unidirectional enantiomerisation 42 (the opposite catalytic base is replaced compared to the aspartate/glutamate racemase exception noted above 20 ). The H. sapiens trans-3-hydroxy-S-proline epimerase 110 also possesses an equivalent Cys-to-Thr substitution to that in MMP0739. 42 However, biochemical analysis shows that this Cysto-Thr substitution converts the latter enzyme from an epimerase into a dehydratase, 110 i.e. the enzyme catalyses elimination rather than racemisation/epimerisation (vide infra).
An often-overlooked consideration in the catalytic mechanism is the hydrogen bonding between the electron-deficient C a -H (which are activated by adjacent carbonyl groups) and active-site bases. This is of relevance for all proteins, since all protein amino-acid residues are capable of forming such bonds. 113 These hydrogen bonds tend to be moderately weak (8 to 10.6 kJ mol À1 when bonding to water compared to 18.9 kJ mol À1 for an 'typical' intra-molecular bond 114 ). In addition, amino-acids and other racemase/epimerase substrates will also be able to form such bonds. The case of the C a -H/His/ Glu hydrogen bond in AMACR/MCR is particularly interesting in this regard (Fig. 4), as the hydrogen bond resembles that in the catalytic triad of chymotrypsin and related hydrolytic enzymes which has been studied in detail. 115

Concerted versus stepwise reactions
The PLP-dependent enzymes have been extensively studied and a series of mechanistic and computational studies show the  presence of a carbanionic intermediate (vide supra, Schemes 1 and 2, 3), 5,50,69,70,101 indicating a step-wise reaction. Kinetic isotope effect studies on alanine racemase are also consistent with a carbanionic intermediate. 87 Alanine racemase catalyses C a -H exchange but the stereochemical course of this reaction was not determined, 116 although non-stereoselective incorporation of label into substrate is expected because of the stability of the carbanionic intermediate.
Studies investigating isotopic incorporation from solvent into substrates have been particularly informative about the concertedness of mechanism in other enzymes. For the majority of enolase family and cofactor-independent racemases and epimerases, isotopic incorporation is observed into the product but very little incorporation is observed into the substrate, e.g. glutamate racemase, 78 proline racemase, 77 mandelate racemase, 74 2-methylmalonyl-CoA epimerase 73 and a racemase mediating post-translational modification of peptides. 117 This is consistent with a concerted reaction. Monitoring the progress of the reaction by these enzymes in isotopically labelled solvent using circular dichroism typically results in an overshoot of the equilibrium position, e.g. as has been observed for mandelate racemase. 74 This results from isotopic incorporation into product only with a significant kinetic deuterium isotope effect affecting the reverse reaction. These results further support a mechanism in which two-base enzymes catalyse a microscopic enantiomerisation reaction, with asynchronously concerted deprotonation and reprotonation. 76 Such a mechanism minimises the formation of a highly unstable doubly deprotonated intermediate and hence partly overcomes the effect of destabilising groups adjacent to the C a -H (i.e. the carboxylate).
In contrast to the above is the observation that incubation of substrates with AMACR in 2 H 2 O results in a near 1 : 1 incorporation of deuterium into substrate and product. This has been interpreted as formation of a discrete deprotonated intermediate followed by deuteration from either side. 79,80 Analysis of the crystal structure of the M. tuberculosis homologue, MCR, shows catalytic residues on both sides of the substrate (the His-126/Glu-241 pair and Asp-156; Fig. 4) and are consistent with the formation of an enolate intermediate. 81,95,112 Thus, AMACR and MCR fundamentally differ in their mechanisms from most other cofactorindependent racemases and epimerases, in that they catalyse microscopic racemisation rather than epimerisation. 8,79,80 Incorporation of deuterium from solvent is also catalysed by hydantoin racemase via an enolate intermediate 118 and is expected to be non-stereoselective but this has not yet been verified.
The above results can be rationalised based on the pK a values for the deprotonation of the substrate. The pK a of C a -H for a thioester is 21, 86,88 while the pK a values for C a -H for amino-acid zwitterions is 29, 20,86 for simple carboxylates is 33 and for simple amides 28.4. 86 Therefore, concerted reactions occur with substrates containing relatively unactivated C a -H (high pK a values), with consequent asymmetrical isotopic incorporation. This explains the behaviour of peptide epimerases, 117 which are observed to undergo concerted reactions. These peptide substrates have pK a values of B26-31 for C a -H, although these values are dependent on both N-and C-substituents and the protonation status of amine groups. 86 This model also allows prediction of enzymatic behaviour for uncharacterised racemases/ epimerases, e.g. hydantoin racemase, 118 based on pK a values for C a -H. The proposed model also casts doubt on the use of isotopic labelling studies to differentiate between 'two-base' and 'one-base' enzymes (reviewed in ref. 92). It has previously been proposed that near-symmetrical isotopic incorporation into substrate and product is indicative of 'internal return', i.e. a 'one-base' mechanism. The results on AMACR 79,80 (reviewed in ref. 6 and 7) show that this behaviour is also observed with 'two-base' enzymes with activated C a -H, as it is known that AMACR/MCR possesses appropriate active-site bases on both sides of the substrate.
There have also been several studies on the elimination reaction catalysed by H. sapiens serine racemase. [129][130][131]133 The wild-type enzyme has a ca. 4-fold preference for b-elimination over racemisation of S-serine. 129,131 Other substrates can also undergo b-elimination, including S-serine-O-sulfate and S-threo-hydroxyaspartate. 131 The enzyme is allosterically activated by divalent metal ions (with Mn 2+ being the strongest) and ATP, 129,133 and activity is potentiated by halide anions. 130 The elimination reaction catalysed by serine racemase is thought to control levels of R-serine in neurons 133 and, hence, modulate the activity of NMDA receptors; 129,131,133 over-activation of the NMDA receptor has been shown to result in neuronal cell death. 133 This is, however, at the expense of producing highly electrophilic 2-aminoacrylate 16. 129,131,133 Enolase family enzymes, such as P. putida mandelate racemase 126 and L. aggregata cis-3-hydroxy-S-proline racemase/ dehydratase (IAM 12614), 119 are also able to catalyse elimination reactions. Mandelate racemase was able to catalyse elimination of chlorine from 3-chlorolactate 18 to give pyruvate 19 (Scheme 4A). 126 The mechanistic details of the reaction was not determined but it is assumed to occur by E2 anti-elimination to give the enol 20 followed by tautomerisation. 126 However, the possibility of a E1cb-type mechanism via an enediolate type intermediate cannot be discounted. The elimination of chlorine from 3-chlorolactate 18 by mandelate racemase is reminiscent of the elimination of HCl from 3-chloroalanine 13 by glutamate racemase, which also gives pyruvate 17 as a product (vide infra, Scheme 12). 134 This result contrasts with the earlier observation on P. putida mandelate racemase with 3,3,3-trifluorolactate 21, which undergoes racemisation. b-Elimination to give 22 is not observed (Scheme 4B). 91 L. aggregata cis-3-hydroxy-S-proline racemase/dehydratase (IAM 12614) 119 catalyses both racemisation and b-elimination reactions with its substrate 23, in a 3 to 2 ratio (Scheme 5). The b-elimination reaction is proposed to go via an enediolate intermediate 24, although it may be a more concerted E2-like reaction. The cis substrate allows for anti-elimination of the hydroxy group to give the enamine product 26, which subsequently tautomerises to D-pyrroline-2-carboxylate 27 (Scheme 5). Alternatively, epimerisation to give 25 can occur. It is notable that 27 is a known inhibitor of T. cruzi proline racemase. 135 The cofactor-independent enzymes diaminopimelate epimerase 124 and glutamate racemase 128 are able to eliminate N-hydroxy substrates. In the case of glutamate racemase, deprotonation of substrate 28 results in elimination of hydroxide or water with formation of imine 29, which is hydrolysed to 2-oxoglutarate 30 (Scheme 6).
With aliphatic substrates containing b-fluorine or b-chlorine, the presence of the halogen increases the acidity of the C a -H, 120,136 and, hence, these elimination substrates tend to be converted with somewhat higher efficiency than their Scheme 4 Reaction of halogen substrates with P. putida mandelate racemase. 91

Scheme 5
The racemisation and elimination reactions catalysed by cis-3-hydroxy-S-proline racemase/dehydratase. 119 Scheme 6 Elimination of N-hydroxy-R-glutamate 28 by an E2 mechanism followed by hydrolysis of imine 29 to form 2-oxoglutarate 30. 128 racemisation/epimerisation equivalents. 120 With diaminopimelate epimerase, 121 only stereoisomers allowing an antiperiplanar conformation between the C a -H and the fluorine underwent elimination, with substrates not allowing an antiperiplanar conformation undergoing epimerisation instead. Similarly, mutant glutamate racemases (in which the active-site Cys bases were mutated to Ser) eliminated either 2R,3R-or 2S,3S-3chloroglutamate stereoisomers 31R and 31S with antielimination (Scheme 7), depending on which active site Cys residue was still present. 82 The resulting enamine 32 tautomerises to imine 29, which is hydrolysed to 2-oxoglutarate 30. These results are consistent with a substantially concerted (E2) mechanism. 137 The above results contrast with those observed with AMACR, in which epimeric substrates 33 and 34 were eliminated to the same product 35, 120 consistent with an E1cb mechanism through the enolate intermediate 36 (Scheme 8). 137 These results are inconsistent with an E2-elimination because the substrate requires a conformation in which the a-H and the fluorine are anti-; epimer 33 can adopt such a conformation but epimer 34 cannot. Interestingly, compounds closely related to 33 and 34 were synthesised 136 and tested as inhibitors of native rat AMACR and no elimination of fluoride was observed. 136 These inhibitors 136 had the same configuration as 34 (and its epimer with opposite C2 and C3 configurations) but, in view of the subsequent report, 120 this is a surprising observation.
However, this is not the only example where an expected elimination reaction did not take place. Nagar et al. investigated trifluorolactate (2-hydroxy-3,3,3-trifluoropropanoate) 9R and 9S as substrates for mandelate racemase (vide supra, Scheme 4). 91 Kinetic analysis showed that K m values for trifluorolactate 9 were unexpectedly similar to the natural substrate, mandelate (1.2-1.74 mM and 1.0-1.2 mM, respectively) and lower than the predicted K m value of B10 mM. In contrast, k cat values were reduced by B318-fold, with k cat /K m reduced by B430-fold. 19 F NMR analysis showed that no fluoride was eliminated during the reaction and, hence, 10 was not formed. 91 The lack of a b-elimination reaction is unexpected, because the trifluorolactate 9 is able to take up the required anti-conformation for an E2 reaction when in a staggered conformation. Clearly, the mandelate racemase-catalysed racemisation of trifluorolactate is faster than the elimination reaction. The reasons for this are unclear but it could be related to the presence of multiple fluorine atoms within the substrate. 138 If loss of fluoride is asynchronous with abstraction of the C a -H, this will result in generation of a positive charge on the b-carbon. The presence of two additional fluorine atoms will destabilise formation of this transition state. However, it is notable that E. coli dipeptide epimerase (YcjG) eliminates fluoride from S-alanyl-R,S-difluoroalanine in preference to epimerisation, 125 suggesting that other factors are also at play such as the extent of d+ charge stabilisation in the transition state.

Methods for determining racemase and epimerase activity
Racemases and epimerases are simple enzymes, in the sense that they only have one substrate and product (a uni-uni reaction), which is a characteristic they share with other isomerases. A consequence of racemases and epimerases accepting both stereochemical configurations of their substrates is that their reaction will, in most cases, be readily reversible and k cat /K m values are likely to be similar for the reactions in both directions (which is required by the Haldane relationship 79,80,139 for an equilibrium constant of B1). Hence the rate for the reverse reaction when determining initial rates is likely to be significant and this must be corrected for in any kinetic study, such as the determination of inhibitor potency.
Several different assays exist for measuring racemase/ epimerase activity (Table 1). One approach is to measure rates Scheme 7 b-Elimination of 2S,3S-3-chloroglutamate 31S by Lactobacillus glutamate racemase to give enamine 32. Tautomerisation to imine 29 followed by hydrolysis gives the resulting 2-oxoglutarate 30. 82 Scheme 8 Elimination of fluoride from substrates by a-methylacyl-CoA racemase. 120 For substrates 33 and 34, n = 6. For inhibitors (34 and its epimer with opposite C2 and C3 configuration) tested on native rat enzyme reported not to eliminate, 136 n = 12. Enzyme catalytic residue numbers are those for human AMACR. 6,7 This journal is © The Royal Society of Chemistry 2021 Chem. Soc. Rev., 2021, 50, 5952-5984 | 5961 at very early time points where the reverse reaction will have less impact. Typically, the enzyme reaction is followed using techniques such as optical rotation 46,[140][141][142][143][144][145] or circular dichroism, 23,91,107,[146][147][148][149][150] allowing a time-course to be determined. These assays are ideal, in that they allow much more accurate determination of initial rates, 151 although correction for the reverse reaction should still be performed. Indeed, circular dichroism is by far the most common method of detecting enzymatic activity (Table 1), although it is noted that these are kinetic studies designed to measure K m , k cat and k cat /K m (vide infra).
A second alternative when analysing substrates undergoing racemisation or epimerisation is to use HPLC of a diastereoisomeric substrate/product mixture at a fixed time point, 37,136 although the differences in energies between diastereoisomers means that these substrates may behave differently from natural enantiomeric substrates. Alternatively, a product containing one chiral centre can be derivatised using a chiral reagent and analysed by HPLC, GC or NMR. 8,79,80 The latter approach is time-consuming, as several time-points for each reaction should be analysed and can be technically challenging, especially when working with the low amounts of product typically obtained from enzymatic reactions. Chiral HPLC is an option for separating enantiomeric substrates, although there appear to be no examples of this having been used.
A second approach is to measure exchange of the a-proton with isotopically labelled substrates 82,95,[152][153][154] or solvent, 8,73,[78][79][80]107 measuring reaction extent by scintillation counting, mass spectrometry or NMR. Such approaches will introduce significant kinetic isotope effects [155][156][157][158] and deprotonation and reprotonation rates will be markedly different from each other although the extent of this will depend on levels of conversion of substrate and whether the transition states are early or late. 107 Consequently, careful design of experiments is needed, especially where precise rate measurements are required. These approaches are often used in mechanistic studies where isotopic distribution in substrate and product is measured (vide supra).
A third approach is to make the enzymatic reaction irreversible. This can be achieved using an irreversible coupled enzyme to remove the reaction product. 107,159,160 There are a number of examples of the use of coupling enzymes in kinetic studies determining K m and k cat values ( Table 1). The most common coupling enzymes used are D-amino acid oxidase and NAD-dependent oxidoreductases. Coupled enzyme assays are the second most common method of assessing enzymatic activity. Similarly, an unnatural substrate which undergoes an irreversible elimination reaction 82,83,120,127,161 can also be used. Typical examples of eliminated groups include water (from amino-acid hydroxamate derivatives 83,162 ), bromide, 122,123 chloride 47,82,124 and fluoride 120,121,124,125,161 as described above. The products from these elimination reactions often need to be assayed using coupling enzymes 83,121 or low-throughput spectroscopic techniques such as NMR. 120,121,161 Attempts to use fluoride sensors to measure enzymatic activity with substrates eliminating fluorine has met with limited success. 161 A notable example of this approach is the elimination reaction of an unnatural acyl-CoA substrate 37 by AMACR to give 2,4dinitrophenoxide 38 and acyl-CoA 39 (Scheme 9); 127 this assay was used in a high-throughput screening campaign of 20 387 compounds which identified novel pyrazoloquinolines and pyrazolopyrimidines as inhibitors 163 and also in the first extensive inhibitor structure-activity relationship studies on any racemase/epimerase (vide infra). 164,165 Catalytic efficiency of racemases and epimerases Kinetic parameters for racemases/epimerases can vary quite widely (  166 In contrast, O-ureidoserine racemase which has reported K m values 46 of 12 and 32 mM for S-and R-O-ureidoserine, respectively. In contrast, the K m for S,S-diaminopimelate (2,6-diaminoheptanedioic acid) for M. tuberculosis diaminopimelate epimerase is only 166 mM, 167 significantly lower than the K m values for other amino-acid racemases/epimerases. The relatively high K m values for most amino-acid racemases/epimerases are undoubtedly a consequence of these enzymes converting small and relatively unfunctionalised substrates. The same trend is observed for mandelate racemase, which has K m values of 1.0 and 1.2 mM for S-and R-mandelate (2-hydroxyphenylacetate), respectively. 91 Racemases/epimerases with larger substrates tend to have lower K m values, as there is more opportunity for binding interactions. For example, human AMACR has a K m value of B86 mM for pristanoyl-CoA, 152   This journal is © The Royal Society of Chemistry 2021 homologue. 146 These lower K m values are generally accompanied by lower k cat values (Table 1).
Catalytic efficiency is quantified using k cat /K m values (Table 1). Again, these can vary quite widely but many racemases/epimerases have relatively modest efficiencies. For example, O-ureidoserine racemase is quite efficient, with reported k cat /K m values 46 of 39 583 and 45 312 M À1 s À1 . Similarly, k cat /K m is reported to be 16 730 and 15 294 M À1 s À1 for F. nucleatum glutamate racemase for S-and R-Glu, while the corresponding values are 3000 and 3806 M À1 s À1 for the B. subtilis enzyme. 166 In contrast, M. tuberculosis diaminopimelate epimerase 167 has a very modest k cat /K m of 883 M À1 s À1 . On the other hand, RacX 168 has extremely low k cat /K m values of 2.86 and 3.23 M À1 s À1 for S-and R-Lys, while YgeA 168 has k cat /K m values of 45.8 and 45.8 M À1 s À1 for S-and R-His.
k cat /K m values for other racemases and epimerases tend to be higher and this is often related to the lower K m values observed for these larger substrates. For example, mandelate racemase (6.2 and 6.5 Â 10 5 M À1 s À1 for S-and R-mandelate 91 ) and the M. tuberculosis homologue of AMACR (5.23 Â 10 6 and 6.0 Â 10 6 M À1 s À1 for S-and R-ibuprofenoyl-CoA, 146 respectively). Finally, N-succinylamino acid racemases and N-acetylamino acid racemases exhibit highly variable k cat /K m values (Table 1).
It is noteworthy that even the most efficient racemases/ epimerases have k cat /K m values well below the theoretical diffusion-controlled maximum of B1 Â 10 9 M À1 s À1 . 169 As proton-transfer reactions are extremely fast (between 5 Â 10 9 and 1 Â 10 11 M À1 s À1 ), 86 rates may be limited by binding of substrate, release of product or conformational changes in the protein. A survey of k cat /K m values for various enzymes 169 shows that, for most enzymes, they are around 10 5 to 10 9 M À1 s À1 , with the most efficient enzyme (superoxide dismutase) having a k cat /K m of 7 Â 10 9 M À1 s À1 . Moreover, k cat /K m values for most enzyme-catalysed reactions appear to be diffusion-limited. 169 There have been few detailed kinetic studies on racemases/ epimerases but studies on mandelate racemase using mandelate as a substrate show that both k cat and k cat /K m are affected by increasing the viscosity of the solvent. 170,171 This indicates that both binding of substrate and release of product are partly rate-limiting, although the effects on k cat are more extensive than those on k cat /K m indicating that that release of product is more sensitive to solvent viscosity than binding of substrate. 172 In contrast, poorer substrates of mandelate racemase 91 or less active mutants of the enzyme 173 tend to be unaffected by increasing solvent viscosity, suggesting that rates are limited by the chemical reaction or other processes, e.g. conformational changes in the protein. Although the k cat /K m values for mandelate racemisation is relatively modest (6.2 and 6.5 Â 10 5 M À1 s À1 ) 91 compared to these other enzymes, it should be noted that racemisation of mandelate is a 'difficult' reaction as judged by the estimated half-life for the spontaneous uncatalysed reaction of 9.8 Â 10 4 year. 169 Thus, mandelate racemase is providing a considerable enhancement (an effective molarity of B4.87 Â 10 6 M). It is unclear whether racemases/ epimerases with lower k cat /K m values are limited by diffusion, chemical reactivity or other processes, or whether these low efficiencies result from a low amount of active enzyme within the enzyme preparation.

Drug design strategies for inhibiting racemases and epimerases
As noted above, many racemases and epimerases are drug targets for various diseases. The following is a survey of different strategies for the development of inhibitors.

Substrate/product analogues
Exploiting the differences in side-chain conformation of different racemase/epimerase substrate stereochemical isomers can be a particularly fruitful strategy for the development of inhibitors. A significant advantage of these inhibitors is that they are achiral when identical sidechains are used. The substrate/product analogue approach works particularly well for racemases/epimerases possessing discrete side-chain-binding pockets for the different stereoisomers, e.g. mandelate racemase 139 and M. tuberculosis a-methylacyl-CoA racemase (MCR). 149 It can also work for enzymes with more subtle changes in side-chain conformation, e.g. aspartate racemase 174 and glutamate racemase, 175 although the potency of inhibition tends to be more modest. In many respects, substrate/product analogues are the equivalent of bisubstrate inhibitors of other enzymes, 176 which often give rise to potent inhibition.
Several substrate/product analogues have been reported as inhibitors of amino-acid racemases (Fig. 5). For example, citrate 40 was shown by X-ray crystallography to bind as a substrate/ product analogue to aspartate racemase. 174 Citrate 40 behaves as a competitive inhibitor, although the potency was very low (K i = 7.4 mM vs. K m = 0.74 mM for L-aspartate). Pal et al. designed cyclic inhibitors of glutamate racemase, in which the ring mimicked the side-chain positions for the different stereoisomers of glutamate, including compound 41. 175 This proved to be a partial non-competitive inhibitor, although potency was modest (K i = 3.1 mM vs. K m = 1.41 mM for Fig. 5 Structures of representative substrate-product analogues which are inhibitors of aspartate racemase (40), 174 glutamate racemase (41), 175 serine racemase (42), 188 proline racemase (43), 188 mandelate racemase (44) 139 and M. tuberculosis a-methylacyl-CoA racemase (MCR) (45). 149 This journal is © The Royal Society of Chemistry 2021 Chem. Soc. Rev., 2021, 50, 5952-5984 | 5965 substrate S-glutamate). 175 In contrast, substrate/product analogues were poor inhibitors of serine racemase (e.g. 42, mixed competitive inhibition; K i = 167 mM and K i ' = 661 mM vs. K m = 19 mM) 188 and proline racemase (e.g. 43, non-competitive inhibition; K i = 111 mM vs. K m = 5.7 mM). 188 Proline racemase is known to have an extremely confined active site in the 'closed form' of the enzyme, 26,27 which binds substrates and inhibitors.
There have only been two substrate/product analogue studies on non-amino-acid racemases. Mandelate racemase substrate/product analogues 139 bind with similar affinity to the substrate [e.g. benzilate (2,2-diphenyl-2-hydroxyacetate) 44, K i = 0.67 mM vs. K m = 0.70 and 0.54 mM for R-and S-mandelate, respectively 139 ]. Similarly, a substrate-product analogue of ibuprofenoyl-CoA (Fig. 5, 45a) was a competitive inhibitor of the M. tuberculosis homologue of AMACR (MCR) and showed about a 6-fold increase in binding affinity (K i = 16.9 mM vs. K m = 106 mM) compared to ibuprofenoyl-CoA, undoubtedly due to the sidechain of the inhibitor binding to both the R-and S-subsites. 149 Enhancing acidity of the a-proton and alternative substrates A number of racemases/epimerases have alternative substrates which undergo changes in stereochemical configuration 8,91,127,164,189 or elimination. 63,82,[120][121][122][123][124]128,161,164 Efficiency of inhibition is dependent on concentrations of inhibitor and their catalytic efficiency as substrates (k cat /K m ). Alternative substrates are usually competitive inhibitors (for example see 127 ), which means that inhibition can be overcome by high concentrations of the substrate whose conversion is being inhibited.
Efficient inhibition can be achieved by increasing the acidity of the C a -H, e.g. by use of trifluoromethyl group (Fig. 6, 47 and  48). 136,190 The trifluoromethyl group lowers the energy of the enolate intermediate 81 in the AMACR reaction; 136 intermediates generally bind tightly to enzymes and more closely resemble the transition states of the reaction. 94 The presence of a sulfur atom immediately adjacent to the substrate C a -H is also an effective strategy for increasing acidity (vide infra, Fig. 29). 165 Preventing the removal of the a-proton These types of inhibitors fall into two types: those in which the C a -H has been replaced by an alternative group and those with neighbouring groups which decrease the acidity of the C a -H. A number of different groups have been used to replace the C a -H (in addition to the substrate/product analogues with a second side-chain noted above), including fluorine atoms, e.g. 49,191 and methylene groups, e.g. 50 192 (Fig. 7).
Inhibitors can also have substituents adjacent to the C a -H, which raise the energy of the deprotonated intermediate, such as hydroxy groups as exemplified by 51 9,164 and 52 164 (Fig. 8).
Exchange of the C a -H was shown not to occur by incorporation studies in 2 H 2 O and 1 H NMR analyses when 51 and 52 were tested as substrates for AMACR. 9,164 In all cases, these approaches tend to give rise to moderate inhibitors, as judged by the ratio of IC 50 /K m or K i /K m values.

Transition-state and intermediate analogues
Transition-state analogues are widely recognised as potent drugs. 135,193,194 This approach has been relatively under-used as a strategy for inhibition of racemases and epimerases, although the few examples show that highly potent inhibitors can be obtained.
An early example is proline racemase, which is inhibited by pyrrole-2-carboxylate 53 and D-pyrroline-2-carboxylate 27 (reviewed in ref. 135) (Fig. 9A). Relatively high concentrations Fig. 6 Representative inhibitors with increased C a -H acidity. 136,165,190 Fig. 7 Representative inhibitors in which the C a -H is replaced. 191,192 Fig. 8 Representative inhibitors in which the acidity of C a -H is decreased. 164,191 Fig. 9 (A) Structures of proline racemase transition-state analogues. 135 Log P values were calculated using: https://www.molinspiration.com/ cgi-bin/properties. Log P, log 10 (ratio of concentrations of drug in octan-1-ol and water at equilibrium); (B) X-ray crystal structure of pyrrole-2carboxylate 53 bound within the active site of proline racemase from T. cruzi, 195  of these compounds are required for inhibition in vitro of the enzyme (about 10 Â that of substrate) and they should be considered as inhibitors in which the C a -H is replaced (vide supra). X-ray crystallographic analysis showed that pyrrole-2carboxylate binds within the T. cruzi active site (Fig. 9B) between the catalytic bases Cys-130 and Cys-300. 195 However, despite its relatively low potency, pyrrole-2-carboxylate 53 reduced invasion of T. cruzi in infected mammalian cell models and also reduced differentiation of the parasite from the amastigote form into trypomastigotes. 28 A number of more water-soluble analogues (e.g. 54 and 55) were tested for their ability to inhibit the enzyme but these proved to have lower potency. 26 Compounds 53, 54 and 55 had similar lipophilicity (calculated log P values of À2.41, À2.33 and À2.20, respectively) and the loss of inhibitory activity is likely to be related to the difficulties of accommodating the bulky halogen in the highly restricted active site. The halogen atom in 54 and 55 is also likely to force the carboxylate group out of plane, and this is expected to have a significant impact on binding affinity.
A more recent example of the use of transition-state analogues are the carbamate inhibitors of a-methylacyl-CoA racemase (Fig. 10), 191 which mimic the transition state (or enolate intermediate), giving rise to highly potent inhibition. 127,164,191 Although the carbamate inhibitor 56 is by far the most potent AMACR inhibitor reported to date, it has limited utility because acyl-CoAs violate Lipinski guidelines and inhibitors are delivered to cells as the carboxylic acid pro-drug. Unfortunately, the acid pro-drug in this case would be a carbamate 57 which may readily decompose 164 especially under acidic conditions or in the presence of cellular nucleophiles.
There are several other examples of using analogues of the deprotonated intermediate as inhibitors. For example, the conversion of mandelate by mandelate racemase is proposed to go through an aci-carboxylate intermediate (Fig. 11, 58). 196 Several mandelate racemase inhibitors of this type have been reported, including the phosphonate inhibitors 196,197 such as the highly potent inhibitor 59 (K i = 4.7 mM vs. K m of 1.0 and 1.2 mM for R-and S-mandelate, respectively). The phosphonate group in 59 possesses two negatively charged oxygen atoms, and hence resembles the aci-carboxylate intermediate 58.
Similarly, cupferron 60 and N-hydroxyformanilide 61 also act as analogues of the deprotonated intermediate 58 because they have an extended planar system of sp 2 -hybridised atoms, whilst benzohydroxamate 62 is a hydroxamate. Inhibitors 59-62 (Fig. 11) strongly ligate to the metal in the active site of mandelate racemase (K i values of 2.7, 2.8 and 9.3 mM, respectively). 198

Allosteric inhibition
Allosteric inhibition arises from inhibitors binding somewhere other than at the enzyme active site. The uncompetitive type of inhibition observed through enzyme kinetics arises from binding of the inhibitor to the enzyme-substrate complex with (almost) no binding to unoccupied enzyme 199 and, hence, must arise from binding at an allosteric site.
Glutamate racemase is the only racemase/epimerase for which confirmed allosteric inhibitors have been reported. Lundqvist et al. identified an uncompetitive inhibitor 63 during a high-throughput screening campaign on the H. pylori enzyme (Fig. 12). 200 The inhibitor-binding site is remote from the active site. 21,200 A second cryptic inhibitor-binding site was subsequently identified in the B. anthracis enzyme by virtual screening, which led to the identification of pyridine-2,6-dicarboxylate (dipicolinate) 64 as an inhibitor (Fig. 12), with K i = 1.9 mM. 25 Further studies on 37 showed that inhibitor binding resulted in the active-site Cys 185 adopting a conformation in which the SH group points away from glutamate C a -H. 21 It is also noted that some uncompetitive inhibitors of a-methylacyl-CoA racemase were recently identified (vide infra, Fig. 15), 163 implying that they bind to an allosteric site, although the exact binding site has not been confirmed.

Covalent inhibition
Inhibitors which form a covalent bond to their targets are enjoying a resurgence because of their potential for longlasting effects and strong affinity for the target, amongst other benefits. [201][202][203][204][205][206][207] Indeed, around 30% of all approved clinical drugs acting on enzymes are covalent inhibitors. 202,207 Covalent inhibitors can cause either reversible or irreversible inhibition of their target. [204][205][206][207][208] There is a perception that covalent inhibitors are non-selective and hence are less useful. However, studies    200 and pyridine-2,6-dicarboxylate 64 (B. anthracis glutamate racemase). 25 The ionisation state of 64 which is shown is that used in the virtual screen. have shown that high selectivity for the target enzyme can be achieved. [203][204][205][206][207][208][209][210] Modification of the substituents around the electrophile can also further enhance selectivity, 210-212 especially for electrophiles modifying cysteine residues [212][213][214][215] (which are the catalytic bases in many cofactor-independent amino-acid racemases and epimerases 5,18,20,22 ). Electrophilic properties can be predicted using the 'electrophilicity index'. 209 Covalent inhibition of racemases and epimerases has been previously investigated. Both diaminopimelate epimerase 20,124 and a-methylacyl-CoA racemase 127,152 have been shown to be inhibited by non-specific protein-modification agents. In each case, these are cysteine-reactive compounds such as iodoacetamide, ebselen (2-phenyl-1,2-benzoisoselenazol-3(2H)-one) and ebselen oxide. It is also noted that mandelate racemase undergoes covalent inhibition by 3-hydroxypyruvate because of formation of an imine between the inhibitor and Lys-166, one of the active-site bases. 216 There have also been several attempts to design irreversible inhibitors rationally, most notably the aziridine inhibitors of diaminopimelate epimerase. 20,124,217,218 A recent example of rational covalent inhibitor design is seen with O-ureidoserine racemase (which interconverts S-and R-O-ureidoserine 65), which is irreversibly inhibited by oxiranes R-and S-66 to give covalent adducts (Scheme 10). 46 An irreversible inhibitor of B. subtilis glutamate racemase was also identified by virtual screening (vide infra), 23,24 and was proposed to bind close to the catalytic cysteine residues. 22 The inhibitor is proposed to modify irreversibly one of these thiols by conjugate addition (a.k.a. Michael addition). 202,205,206 The hit compound 67a and several analogues 67b-67d (Scheme 11A) were subsequently shown to be irreversible inhibitors. 22 Compounds 67a and 67c proved to be non-saturating inhibitors. In contrast, 67b and 67d displayed saturating inhibition, consistent with modification of active site residues. Further experiments showed that inhibition was reversible, consistent with a reversible conjugate addition via enolate 68a to give the product 69a (Scheme 11B). Mass spectrometric analysis of wild-type and C74A mutant glutamate racemase following incubation with 67a confirmed modification of Cys-74, one of the active-site bases. Compound 67a was unreactive with 2-mercaptoethanol under the assay conditions, 22 showing that conjugate addition to thiols only occurred in the presence of the high nucleophilic Cys-74 in the enzyme active site. The rhodanine warhead (Scheme 11A) is recognised as a common motif found in panassay interference compounds (PAINs), which give rise to false positive or intractable leads in high-throughput screening campaigns. 219 These rhodanine glutamate racemase inhibitors showed activity against various bacterial strains, including various methicillin-resistant S. aureus strains. 22 In a second study, 3-chloroalanine 70 (b-chloroalanine; a poor inhibitor of PLP-dependent alanine racemase 134 ) was shown to irreversibly inactivate glutamate racemase from M. tuberculosis. 134 Non-saturating kinetics where observed for the S-isomer with a second-order rate constant of 2.7 M À1 s À1 . Mass spectrometric analysis of peptides showed that 3-chloro-S-alanine (70S) reacted at Cys-185, while 3-chloro-R-alanine (70R) reacted at Cys-74. In the glutamate racemase reaction, R-glutamate is deprotonated by Cys-74 whilst S-glutamate is deprotonated by Cys-185 during enantiomerisation, i.e. the active-site Cys acting as an acid is derivatised by 3-chloroalanine 70.
The authors proposed 134 that the adduct was a pyruvate derivative, based on the observation that pyruvate 71 was generated upon treatment of the enzyme with 3-chloroalanine 70 but their proposed mechanism is very unlikely. Two more likely mechanisms can be envisaged (Scheme 12, pathways A and B) based on the observed increase in mass of B87 Da. In pathway A, removal of the C a -H of 70S by Cys-74 results in elimination of HCl, yielding the aminoacrylate   220 Similarly, two covalent inhibitors (77 and 78) of T. cruzi proline racemase were identified by virtual screening. 26 The inhibitors were proposed to modify the active-site cysteine residues 26 by conjugate addition 202,205,206 and this was subsequently confirmed by X-ray crystallography. 27 The active compounds (Scheme 13A) each have a double bond in conjugation with a carboxylate and a ketone 26,27 and X-ray crystallography showed that conjugate addition occurred towards the ketone. 27 This is unsurprising as ketone carbons are more d+ than carboxylic acids/carboxylates and hence conjugate addition is expected to occur towards the ketone. The most active compound of those subsequently investigated (NG-P27, 79) 27 was active against T. cruzi in infected mammalian cells. It is also notable that one of the original compounds 26 (5-bromo-4oxopent-2-enoate 78) is divalent and reacts with both active-site cysteine residues, cross-linking the enzyme to give adduct 80 (Scheme 13B). 27 Virtual screening and structure-based fragment screening Virtual screening of drug targets with a compound library is a well-established method in drug discovery. [221][222][223][224][225] These approaches can utilise artificial intelligence to optimise the process 223 or negative design 222 to remove compounds which are poor prospects.
There are only a few examples of virtual screening being used for identification of inhibitors of racemases/epimerases and all have been for amino-acid racemases. For example, Skariyachan et al. conducted a screen of a virtual natural products library against diaminopimelate epimerase, amongst several other targets, identifying limonin 81 as a hit (Fig. 13). 226 Limonin 81 and several other hits showed dose-dependent activity against a clinical strain of multi-drug resistant Acinetobacter baumannii.
Studies on B. subtilis glutamate racemase used ab initio quantum mechanics/molecular mechanics to probe transition states in the reaction. 23 A strong correlation between computational and experimentally determined binding of known inhibitors was observed. The same study 23 used a enzymatic transition state conformation in a virtual screen of over one million compounds followed by experimental testing. Although no tight-binding inhibitors were identified, several common motifs for competitive inhibitors were identified. A subsequent virtual screening study on the same enzyme 24 identified several competitive inhibitors such as 64 and 82, with the two most potent inhibitors, 83 and 84, having K i values of 59 and 42 mM (Fig. 14).
Virtual screening has also been used against proline racemase from T. cruzi, the causative agent of Chagas disease (vide supra). 26 Proline racemase without a ligand exists in an 'open' conformation. Upon binding of an inhibitor, a 'closed' conformation is adopted, such that a very restricted active site is produced, which prevents design of inhibitors using standard

High-throughput screening and related approaches
High-throughput screening is an under-utilised approach to discovering inhibitors of racemases and epimerases. Highthroughput screening offers a number of advantages, including the possibility of discovering inhibitors which are not competitive (which is the mode of inhibition often observed for activesite-directed inhibitors). 227 Several different in vitro assays have been used in discovery campaigns. Release of tritium ( 3 H + ) from a radiolabelled substrate into solvent was used in a screen of B5000 compounds against human a-methylacyl-CoA racemase (AMACR). 152 Crucially, the assay requires several steps, including chromatographic separation of residual (acyl-CoA) substrate from tritiated water product. Therefore, this assay is not ideally suited for high-throughput or fragment-screening campaigns. This study 152 identified a number of non-specific protein-modifying and degrading agents, such as ebselen, ebselen oxide and Rose Bengal. 127,152 A subsequent high-throughput screen on human AMACR 163 made use of an eliminating substrate 37 (vide supra, Scheme 9) producing 2,4-dinitrophenoxide 38. 127 Conveniently, this allowed identification of inhibitors based on absorbance changes over the time course in a continuous assay. The screen identified a series of mixed competitive and uncompetitive pyrazoloquinolines, e.g. 85, and pyrazolopyrimidines, e.g. 86 (Fig. 15). The use of a chromogenic substrate 127 allows real-time monitoring of the enzymatic reaction but substrates of this type can only be used with a few racemases/epimerases. Two studies have made use of high-throughput screens with coupling enzymes. The first study, by Lundqvist et al., conducted a high-throughput screen of 385 861 compounds against H. pylori glutamate racemase. 200 No details of the actual screen are given but the authors used two different assays for assessing identified inhibitor activity: conversion of S-to R-glutamate was coupled to UDP-N-acetyl-muramic acidalanine: R-glutamate ligase (MurD) and purine nucleoside phosphorylase with monitoring of the reaction at 360 nm. In the second assay, conversion of R-to S-glutamate was coupled to S-glutamate dehydrogenase with spectrophotometric monitoring of conversion of NAD + to NADH. These screens led to the identification of an uncompetitive inhibitor (vide supra Fig. 12,  63), which was subsequently shown to exert its effect by changing the conformation of the catalytic base, Cys-185, such that it points away from the glutamate substrate C a -H.
In a second example 228 of a coupled assay, dTDP-6-deoxy-Dxylo-4-hexopyranosid-4-ulose 3,5-epimerase (RmlC) was coupled to the subsequent enzyme in the biosynthetic pathway, which is NADP + -dependent, with activity being followed by decreasing fluorescence of NADPH at 460 nm. This led to the identification of a series of inhibitors, including some with potency in the nM range e.g. 87 (Fig. 16). 228 Use of coupling enzymes, such as in these examples, is a standard approach in high-throughput and fragment-screening campaigns 227 (vide infra), although it is always necessary to check if the hits are inhibiting the desired target or the coupling enzyme.
Racemases used in biotechnological applications have been assayed in several different ways, typically using oxidase enzymes of various types. These assays could be adapted for high-throughput screening for inhibitors. For example, mutant mandelate racemases were assayed using mandelate dehydrogenase, which uses NAD + . Conveniently, the ketoacid product   can be assayed using 2,4-dinitrophenylhydrazine (2,4-DNPH) at alkaline pH with the final 2,4-dinitrophenylhydrazone product absorbing at 450 nm. 54 Ketoacids are also be produced by the action of amino-acid racemases and epimerases on amino-acid hydroxamate and other eliminating substrates 82,128 (vide supra, Schemes 3, 4, 6 and 7). Alternatively, the NAD(P)H product from dehydrogenases can be assayed using diaphorases 23,125 or by direct monitoring of absorbance or fluorescence.
Similarly, hydrogen peroxide is produced by several oxidative enzymes, including D-amino-acid oxidase, which can be conveniently assayed using horseradish peroxidase. 174,229 Notably, an assay based on D-amino-acid oxidase/horseradish peroxidase was used to evaluate rationally designed inhibitors of proline racemase 26 and in the high-throughput screening of alanine racemase. 230 Similarly, a continuous assay for Nacetylamino-acid racemases was developed using the R-substrate 88R (Scheme 14). 181 A stereoselective deacetylase was used to convert the S-product 88S to the corresponding S-amino acid 89S and acetate 90. L-Amino acid oxidase was used to produce H 2 O 2 from 89S, which was quantified by the horseradish peroxidase-catalysed oxidation of dianisidine 90 to give the coloured oxidation product 91.
In additional, several racemases/epimerases eliminate HF from fluorine-containing substrates (vide supra, Scheme 8), 120,121 which could potentially be assayed using fluoride sensors, although this can be challenging in aqueous systems. 161 Finally, microscale, medium-throughput polarimetric assays offer the possibility of direct observation of the change in chirality 142 during screening.

Inhibitors of racemases and epimerases
Racemases and epimerases play pivotal roles in metabolism and are excellent drug targets. The following is a survey of recent advances in drug development, focussing on recently reported small-molecule inhibitors. Inhibition of many aminoacid racemases/epimerases has been the subject of a recent excellent review 20 and readers are referred to this and the above sections for details of studies on inhibition of diaminopimelate racemase, 20,124,217,218,226 proline racemase, 26,27,135,188 hydroxyproline epimerase, 20 aspartate racemase, 20 serine racemase, 188 isoleucine epimerase 20 and O-ureidoserine racemase. 46 Studies on inhibition of mandelate racemase 91,139,[196][197][198]216 are detailed in the section on inhibition strategies above. There have been no reported studies on inhibition of EcL-DER, RacX, YgeA, McyF, YcjG or N-acetylmannoseamine-6-phosphate 2-epimerase or N-acetylamino-acid or N-succinylamino-acid racemases since 2015.

PLP-dependent racemases
Alanine racemase. Alanine racemase is involved in cyclosporine biosynthesis 179 and is a well-established antibacterial drug target. 18,20 The enzyme has been the subject of extensive inhibitor studies, 231,232 including the early studies with 3-fluoroalanine and 3-chloroalanine noted above. 132 Other inhibitors include several peptide and halogen-containing peptides, phosphonic acid derivatives (fosfalin) and various halovinylglycines and thiadiazolidinones. 231,232 An important inhibitor of alanine racemase is D-cycloserine, a natural product used in the treatment of drug resistant tuberculosis. 233,234 It is notable that the biosynthetic pathway for D-cycloserine contains a reaction catalysed by a racemase, O-ureidoserine racemase 46 (vide supra, Scheme 10A). D-Cycloserine 93 is a relatively non-specific antibiotic and targets M. tuberculosis alanine racemase and D-Ala-D-Ala ligase. 233,234 Inhibition of alanine racemase is proposed to occur by formation of the external aldimine 94 followed by reversible rearrangement to the ketimine 95 and isoxazole 96 (Scheme 15). 233 Stereoselective isotope exchange with solvent is observed when the reaction is carried out in 2 H 2 O, with D-cycloserine 93 incorporating deuterium at the a-position without a change in stereochemical configuration. Deuteration appears to arise by exchange of the a-proton. Incubations of alanine racemase with D-cycloserine 93 also result in the formation of isoxazole 97. 233 The authors propose a complex rearrangement of the keto tautomer of 96 but this seems unlikely. A simpler explanation is that alanine racemase catalyses hydrolysis of aldimine 94 using a hydrogen-bonded water molecule, to form the linear aldimine 98 directly. This could undergo imine exchange to form 99 but a more likely scenario is that 97 is released from 98 which uses its hydroxylamine to form the aldimine complex 99 directly (the pK a for the conjugate acid of Scheme 14 Continuous assay of N-acetyl-amino acid racemases using a three-enzyme coupling system to give a coloured product 92 absorbing at 436 nm. 181 This journal is © The Royal Society of Chemistry 2021 Chem. Soc. Rev., 2021, 50, 5952-5984 | 5971 the a-NH 2 and g-O-NH 2 groups are 9.14 and 3.16, 235 respectively, and, hence, the g-O-NH 2 group will be uncharged at neutral pH). Thus, D-cycloserine 93 is both a substrate (undergoing a-proton exchange and hydrolysis) and an irreversible inhibitor of alanine racemase, consistent with the observation that M. tuberculosis alanine racemase is not fully inhibited even by high concentrations of 93.
A series of tetrazole-peptide derivatives were also designed and synthesised as inhibitors of alanine racemase in bacterial cells (Fig. 17). 236 The tetrazole group is a well-established bioisostere of the carboxylate group, 236 and hence 5-(1-aminoethyl)tetrazole 100 should behave as an analogue of alanine. S-and R-5-(1-aminoethyl)tetrazole 100 (AET) were inactive when tested against a series of Gram-negative and -positive bacteria 236 but this is unsurprising as it is known that alanine is imported into bacterial cells as an oligopeptide. Indeed, fosfalin 101 is delivered to bacterial cells as a ''dipeptide'' alafosfalin 102. 231,236 Therefore, a series of di-and tripeptide derivatives were synthesised and tested. 236 The SS-Ala-Ala analogue 103a was active against several Gram-negative species, whilst the SS-norvalene-AET 103b, SS-Leu-AET 103c and SS-Phe-AET 103d analogues were active against several Gram-positive species. The SSS-Ala-Ala-AET analogue 104 showed similar activity to 103a. N-Succinyl derivatives of these peptide analogues were largely inactive. It was not determined if S-and R-5-(1-aminoethyl)tetrazole 100 or any of the peptide analogues were inhibitors of alanine racemase and hence the mechanism of antibacterial activity has not been confirmed.
High-throughput screening of small-molecule and fungal extract libraries against Aeromonas hydrophila alanine racemase has also been performed. This screen identified several previously unknown inhibitors (Fig. 18) of moderate potency (IC 50 = 6.6 to 18.5 mM), including homogentisic acid 105 and hydroxyquinone 106. 230 D-Cycloserine 93 (the control inhibitor) had an IC 50 of 5.4 mM under the conditions used in this screen. Kinetic analysis showed that homogentisic acid 105 was a competitive inhibitor (K i = 51.7 mM) whilst hydroxyquinone 106 was a non-competitive inhibitor (K i = 212 mM). These two compounds showed antibacterial activity against A. hydrophila, a Gram-negative anaerobic pathogen. Anabellamide 107 was a potent inhibitor of alanine racemase in vitro (IC 50 = 6.6 mM) but was inactive in cellular assays (Fig. 18).
Scheme 15 Inhibition of alanine racemase by D-cycloserine 93. Note that 96 was proposed to be directly converted into 99 233 but it is more likely that 94 undergoes hydrolysis to form 98, which releases 97. This directly forms two different external aldimines, 98 or 99, respectively (see text for details). Isoleucine 2-epimerase. Isoleucine 2-epimerase is a novel anti-bacterial drug target. There are only two studies on the inhibition of isoleucine 2-epimerase. 180,189 Mutaguchi et al. noted, in their original characterisation of the enzyme, that it was inhibited by non-specific inhibitors of other PLPdependent enzymes, such as hydroxylamine, aminooxyacetate and phenylhydrazine. 180 The effect of hydroxylamine was reversed upon dialysis and addition of PLP to the buffer, providing good evidence that the enzyme is a PLP-dependent epimerase.
Subsequently, a study investigating inhibition of Lactobacillus buchneri isoleucine 2-epimerase by substrate/product analogues was reported. 189 Two groups of inhibitors (Fig. 19) were investigated, based on the structure of the substrate S-isoleucine 108. These inhibitors fell into two classes: those with a single modified sidechain (109)(110)(111); and those with dual sidechains (112), which are similar to the substrate-product analogues that inhibit other racemases and epimerases. 139,149,150,174,188,197,198 The 2R-and 2S-enantiomers of 109 were substrates of the enzyme, although these are converted with an efficiency of only B50% to 80% of that of the natural substrates (as judged by k cat /K m values 180,189 ) and, hence, would not be very effective competitive inhibitors. On the other hand, 110 which possesses an additional methyl group on the sidechain was not a substrate but instead behaved as a pure competitive inhibitor (K i values of 1.5 and 2.9 mM for the 2R and 2S enantiomers, respectively). Compound 111 which has a cyclic sidechain was also an alternative substrate and was converted with very similar efficiencies to 109 (as judged by k cat /K m values). The synthesised compounds with dual sidechains (112, n = 1, 2 and 3) were rather poor inhibitors, although potency increased with increasing size and hydrophobicity of the sidechain (K i = 144, 19 and 11 mM, respectively).
Serine racemase. Serine racemase catalyses the formation of R-serine from S-serine as well as the elimination of water. 69 R-Serine binds to the N-methyl-D-aspartate (NMDA) receptor glycine-binding site. The NMDA receptor is associated with several neurological diseases, including Alzheimer's disease, amyotrophic lateral sclerosis, Huntington's disease, Parkinson's disease, epilepsy and eye disease, amongst others, and psychiatric diseases such as schizophrenia, and depression. 35,[237][238][239][240][241] Hence, inhibition of serine racemase is of interest as a strategy for the manipulation of levels of R-serine.
A large number of studies have appeared on the biological and pathological role of serine racemase in the last five years 35,69,129,[241][242][243][244] but only one study has reported synthesis and evaluation of inhibitors. 239 This study 239 elaborated a potent hit (Fig. 20, 113) identified in a previous virtual screen. 245 Of the synthesised compounds, five showed potent inhibition of serine racemase in vitro. Two of these compounds were similar in structure to the original hit and IC 50

Cofactor-independent racemases and epimerases
Glutamate racemase (MurI). Glutamate racemase catalyses the interconversion of S-and R-glutamate. R-Glutamate is a key component of the bacterial cell wall 17 and the enzyme is an important drug target. Readers are also referred to the review on amino-acid racemases 20 and the previous section for details of substrate/product analogues, 175 allosteric inhibitors, 21,25,200 covalent inhibitors, [22][23][24]134 high-throughput screening 200 and virtual screening. [23][24][25] Malapati et al. have reported a series of medium-throughput screening studies on M. tuberculosis glutamate racemase using thermal-shift assays (Fig. 21). [246][247][248] Structure-activity relationship (SAR) studies led to inhibitors 117-119 with low mM IC 50 values. Non-competitive inhibition was assigned based on the observed changes within the thermal shift assay, although this was not confirmed by enzyme activity assays. Docking studies suggested that these compounds bound to an allosteric binding site.
In addition, Duvall et al. reported phenotypic screening of a diversity-orientated synthetic collection (B100 000 compounds)  against Clostridium difficile and other bacterial strains under anaerobic conditions. 249 One of the hits (BRD0761, Fig. 22, 120) showed minimum inhibitory concentrations of 0.06-0.25 mg mL À1 (0.13-0.55 mM) against various C. difficile strains, with much higher MIC values against other anaerobes, while its epimer BRD3141 121 was also active (Fig. 22). BRD0761 inhibited uptake of [ 14 C]-N-acetylglucosamine into bacteria in a dose-dependent manner, suggesting that it targeted bacterial cell wall biosynthesis. The target was identified from resistance mutants as glutamate racemase and a binding model was produced based on the X-ray crystal structure of H. pylori enzyme. Dosing of mice with 120 protected them from C. difficile infection. UDP-N-acetylglucosamine 2-epimerase. UDP-N-acetylglucosamine 2-epimerase is one of the first enzymes in the teichoic acid biosynthetic pathway, 250 which is required for the integrity of the bacterial cell wall. In addition, Zika virus uses 2,3-linked sialic acid residues to enter mammalian cells and CRISPR-Cas9 knock-out of this enzyme reduces viral infection. 251 The use of N-acetylmannosamine analogues as inhibitors is especially favourable, as N-acetylmannosamine is used solely for biosynthesis of sialic acid; in contrast, UDP-N-acetylglucosamine is also used in the biosynthesis of other glycans 252 and analogues are likely to suffer from lack of selectivity.
A series of N-acetylglucosamines and N-acetylmannosamines, some with modified UDP moieties, have been previously developed as inhibitors but had modest potency (reviewed in ref. 253). Nieto-Garcia et al., reported a series of inhibitors in which the C6 hydroxy group was replaced with sulfur or selenium (Fig. 23). 253 The diselenide inhibitor 122 proved to be highly potent (IC 50 = 8.5 mM) compared to the other inhibitors (IC 50 values of 1.9 to 410 mM). The dimeric monoselenide inhibitor 123 was much less potent (IC 50 = 3.0 mM). The corresponding disulfide analogue 124 was also much less active than 122 (IC 50 = 4.2 mM), showing the importance of the diselenide unit for potent inhibition. The much higher potency of 122 compared to the other inhibitors could be due to bond length or flexibility of the linker. 253 Small-molecule diselenide bonds have been reported as having a bond length of 2.29 Å, 254 while disulfide bonds (in proteins) have a corresponding bond length of 2.05 Å. 255 It has also been suggested that van der Waals interactions and hydrogen bonding potential may also be important in determining inhibitory potency. 253 Diselenide 122 was a competitive inhibitor with a K i value of 15.7 mM. 253 Hinderlich et al. reported a high-throughput screening campaign using a library of 41 536 compounds and a luciferase assay to measure ATP depletion. 252 The N-acetylmannosamine substrate was used at 33 mM, close to its K m value, with an average Z 0 value (a measure of the ability of the assay to discriminate between a hit and random noise 227   UDP-N-acetylglucosamine epimerase/N-acetylmannosamine kinase is also one of only two racemases or epimerases to be subjected to a fragment-screening campaign. 257 A library of 281 fluorinated fragments were screened at 50 mM using 19 F NMR and binding of inhibitor was confirmed by competition with N-acetylmannosamine and ATP, yielding 23 hits. Of these, compound 129 was also shown to inhibit in a coupled enzyme assay and so was chosen for development. Analogues of 129 were screened, leading to identification of 130 which was optimised to 131 (Fig. 25). Modelling of the binding of the inhibitor suggested that 131 bound to the active-site Mg 2+ used in the kinase reaction, near the catalytic site. However, these compounds were not developed into more potent leads.
dTDP-4-keto-6-deoxyglucose 3,5-epimerase (RmlC). dTDP-4keto-6-deoxyglucose 3,5-epimerase (RmlC) is involved in biosynthesis of L-rhamnose in M. tuberculosis and other bacteria. 111,228 L-Rhamnose is biosynthesised from D-glucose-6-phosphate in a four-step pathway. The third step of this pathway is epimerisation at both carbons C3 and C5 of the 4-ketosugar moiety catalysed by RmlC, followed by reduction of the keto group by RmlD in the final step. 111 Because L-Rhamnose is essential for the integrity of the bacterial cell wall, RmlC and the other enzymes in the pathway are drug targets. [258][259][260] RmlC is also responsible for activation of the virulence factor in the marine pathogen Vibrio vulnificus. 261 Several inhibitors of RmlC have been previously characterised (including the high-throughput screening inhibitors noted above; 228 vide supra, Fig. 16), although many have limited aqueous solubility. 259 van der Beek et al. conducted a fragmentscreening campaign with a commercial library using bio-layer interferometry. 259 A library of B1000 fragments was screened at 200 mM with twelve hits. Of these, seven compounds showed dose-dependent enzyme inhibition and inhibited bacterial growth. Three hits (Fig. 26, 132-134) with diverse structures inhibited both RmlB (the preceding enzyme in the biosynthetic pathway) and RmlC.
Sasikala et al. also conducted a virtual screen of RmlC and identified several potential inhibitors of the Vibrio vulnificus enzyme (also known as WbpP). 261 However, none of these compounds were confirmed as hits in biochemical or biophysical screens.
a-Methylacyl-CoA racemase. a-Methylacyl-CoA racemase (AMACR; P504S) is a metabolic enzyme involved in the degradation of branched-chain fatty acids and the activation of ibuprofen and related drugs. 6,7 Levels of the AMACR protein are increased in prostate cancer and many other cancers and the reader is referred to previous reviews on the subject. 6,7 The M9V single-nucleotide polymorphism (SNP) is well known to increase risk of prostate cancer (reviewed in ref. 7) but recent analysis showed interaction of this SNP with SNPs in serine/ threonine kinase AKT1 which are also involved in prostate cancer. 262 Levels of AMACR protein have also been shown to be downregulated by microRNA miR200, resulting in decreased proliferation and migration of prostate cancer cells. 263 Interestingly, a recent epidemiological study showed that AMACR levels were diminished in men with prostate cancer who supplemented their diet with extracts from cruciferous vegetables, such as broccoli, which contains the isothiocyanate compound, sulforaphane. 264 AMACR levels are also increased in glioblastoma 265,266 and high AMACR levels are correlated with poor prognosis for patients. 265 Hence, AMACR is a potentially a novel biomarker for glioblastoma. 265 siRNA knock-down of AMACR levels led to reduced proliferation of glioblastoma cells. 265 Increased AMACR levels are thought to indicate an increase in fatty acid b-oxidation, in a similar way to that observed in prostate cancer. 266 AMACR has been the subject of several previous inhibitor studies as well as structural studies on the M. tuberculosis homologue (MCR), with literature up to the end of 2012 having been previously reviewed. 6,7 Following on from previous reports, 136 Carnell    hydrogen with fluorine is commonly used in drug design because of the similar atomic radii (1.10 vs.  191 although this was not actually proven. Model studies suggest the C a -H pK a for 136 should be B14. 5-16.5 86,191 compared to a C a -H pK a of B21 for standard acyl-CoA esters. 86,88 Inhibitors 49 and 135 had IC 50 values of 324 and 570 nM. 191 N-Dodecanoyl-R-alanyl-CoA 136 was less potent, with an IC 50 value of 2300 nM, which is probably be due to lower stabilisation of the negatively charged intermediate. 191 The known inhibitor (AE)-2-trifluoromethyltetradecanoyl-CoA 136 47 had an IC 50 of 156 nM. 191 Significantly, two potent carbamate inhibitors 56 and 137 as analogues of the intermediate enolate were reported (IC 50 = 98 and 1000 nM). Later studies 127,164 showed that the carbamate inhibitor 56 was highly potent compared to other inhibitors (IC 50 = B0.4 nM using the colorimetric assay 127 ).
Also following on from the Carnell et al. study in 2007, 136 Festuccia et al. 190 reported the synthesis and testing of trifluoroibuprofenoyl-CoA (Fig. 28, 48). Limited kinetic analysis suggested non-competitive inhibition by this compound with a K i = 1.7 mM. This result is notable because non-competitive inhibition of enzymes is rather rare (reviewed in ref. 227) and this is the only example of a non-competitive inhibitor reported for AMACR (and one of only a few for racemases/epimerases in general). Non-competitive inhibition is inconsistent with the inhibitor acting as an alternative substrate but instead arises through allosteric inhibition, stabilisation of an inactive conformation or covalent modification of the target. 227 The basis for inhibition of AMACR by trifluoroibuprofenoyl-CoA is unclear, although elimination of fluoride is not reported. Treatment of cultured androgen-dependent and -independent prostate cancer cells with the pro-drug trifluoroibuprofen (Fig. 28, 138) resulted in arrest at G2/M in the cell cycle and a host of other changes, including induction of apoptosis. 190 Tumour growth in androgen-dependent and -independent prostate cancer xenograft mouse models was also significantly reduced by treatment with this agent. 190 The advent of the AMACR colorimetric assay (vide supra, Scheme 9) 127 has enabled much more thorough testing of inhibitors than had been previously possible, including determination of IC 50 and K i values and of reversibility of inhibition. This also enabled the first structure-activity relationship studies to be conducted. The first studies 127,164 looked at a series of known AMACR inhibitors and substrates. A second study 165 looked at a focussed series of 2-(arylthio)propanoyl-CoA inhibitors; the presence of the side-chain sulfur atom resulted in increased acidity of the C a -H (previous studies on straight-chain acyl-CoAs and their 3-thia analogues showed that the presence of the sulfur reduces the pK a of the C a -H to B15-16.5, 269,270 compared to B21 for the corresponding acyl-CoA 86,88 ). Many of these 2-(arylthio)propanoyl-CoA inhibitors were equipotent to fenoprofenoyl-CoA but optimisation of the inhibitor side-chain resulted in increased potency, e.g. 139, IC 50 = 22.3 nM. 165 A 2-(arylsulfonyl)propanoyl-CoA inhibitor 140 was also synthesised in the hope that the presence of the sulfonyl group would further increase C a -H acidity but this proved to be a poor inhibitor (Fig. 29). 165 Plotting pIC 50 values for all inhibitors 127,164,165 characterised by the AMACR colorimetric assay 127 against calculated log P values (Fig. 30) showed that inhibitor potency was positively correlated with log P. The 2-(arylsulfonyl)propanoyl-CoA inhibitor 140 was highly hydrophilic, 165 suggesting that this was the reason for its unexpected low potency. Although the 2-(arylthio)propanoyl-CoA inhibitors, such as 139, were highly potent in enzyme assays in vitro (IC 50 = 22-520 nM), the carboxylic acid pro-drugs did not show any appreciable inhibition of androgen-dependent or -independent prostate cancer cells, 165 possibly due to oxidation of the inhibitor pro-drug sulfur to the sulfoxide or sulfone.
Since the last review, 7 two studies featuring rational design of inhibitors for the M. tuberculosis AMACR homologue, MCR, have been published. The first study 149 describes the synthesis and testing of several substrate/product acyl-CoA inhibitors   7a-7n, 10b). 165 Log P values were calculated using: https://www.molinspiration.com/cgi-bin/properties. Log P, log 10 (ratio of concentrations of drug in octan-1-ol and water at equilibrium); pIC 50  (vide supra), in which the a-proton is replaced by a second sidechain in the inhibitor. The presence of the second sidechain increases potency of inhibition by B6-fold, although the measured absolute potency is relatively modest (e.g. 16.9 cf. 106 mM for 45a vs. ibuprofenoyl-CoA). One of these inhibitors (Fig. 31A,  45b) has a a-proton in place of the a-methyl group and, as predicted, this does not undergo enzyme-catalysed exchange with solvent consistent with the a-proton being located in the methylbinding site of the enzyme. The study is also notable in that several carboxylic acid precursors are also inhibitors, albeit with IC 50 values in the mM range. 149 Similar to the above AMACR inhibitors (vide supra, Fig. 30), potency of inhibition of MCR is also related to calculated log P values (Fig. 31).
Following from the observation that carbamate analogues are highly potent AMACR inhibitors, 127,164,191 Pal et al. 150 synthesised and tested carbamate analogues 141-144 of their substrate/product inhibitors against MCR (Fig. 31A). 149 Inhibition is reported to be competitive, although the Lineweaver-Burk plots for some analogues, e.g. 143 (n = 5), suggested mixed competitive inhibition. Surprisingly, several of these analogues show irreversible inhibition, in marked contrast to the carbamate AMACR inhibitors which are fully reversible. 127,164 Inhibitors with long alkyl chains (n = 9 and 11) show saturating loss of activity with maximum k inact values of B0.4 min À1 consistent with being active-site directed. Analogues with less lipophilic side-chains (143, n = 3) or a single sidechain (144) showed a non-saturating loss of enzymatic activity with a rate constant of 0.016-0.04 min À1 . 150 Inhibition was not reversed upon dialysis but no protein modification was observed by mass spectrometry. This observation is consistent with either non-covalent slow-binding inhibition, resulting in a long-lived enzyme-inhibitor complex, or irreversible inhibition resulting in a covalent modification of the protein, which is labile under mass spectrometric conditions. Identification of AMACR inhibitors by high-throughput screening has also been reported. 163 Unlike the previous study, 152 the identified inhibitors were not non-specific protein modification agents. 163 A number of pyrazoloquinolines and pyrazolopyrimidines were identified (vide supra, Fig. 15, 85a, 85b and 86), and some structure-activity relationships were observed. 163 The identified inhibitors displayed either mixed competitive or uncompetitive inhibition. The latter is a rare type of inhibition and arises from binding of inhibitor to the enzyme-substrate complex.

Conclusions
Racemases and epimerases occupy a unique position in metabolism, in that they are the only major class of enzymes which can use substrates with both configurations at a chiral centre. Because of this, many racemases and epimerases are excellent drug targets and several have been extensively investigated as such, e.g. glutamate racemase. [20][21][22][23][24][25]134,175,200,249 Use of inhibitors with the same configuration as the less abundant substrate (often D-or R-enantiomer) potentially offers additional benefits in that these isomers may be less prone to off-target binding and may have reduced drug metabolism, with consequent reductions in toxicities and longer durations of action.
However, efforts to develop drugs targeted against racemases and epimerases have been largely limited to rational design campaigns, with the few notable exceptions detailed above. Development of inhibitors which are alternative substrates has met with limited success, in part because several of the effective inhibitors are rapidly depleted in vivo whilst the effects of less effective inhibitors are readily overcome by the physiological substrate. Moreover, these inhibitors are necessarily chiral and there had been a move away from chiral drugs towards drugs with fewer sp 3 carbons. 227 This is despite a growing realisation that the attrition rate is higher for 'flatter' drugs 271,272 and that licenced drugs have a higher average proportion of sp 3 centres than molecules published in The Journal of Medicinal Chemistry. 273 Consequently there has been a more recent move towards structures with a higher proportion of sp 3 and chiral centres. 274 Similarly, racemase/epimerase (B) acyl-CoA inhibitor potency as measured by pIC 50 as a function of calculated Log P values. Log P values were calculated using: https://www. molinspiration.com/cgi-bin/properties. Log P, log 10 (ratio of concentrations of drug in octan-1-ol and water at equilibrium); pIC 50 , Àlog 10 IC 50 . inhibitors in which the C a -H is replaced or where deprotonation is made more difficult tend not to be highly potent. The use of substrate-product analogues as inhibitors has also met with variable success. This strategy tends to work relatively well for racemases and epimerases in which large changes in substrate side-chain position occur during the reaction. Enzymes catalysing reactions resulting in limited changes in side-chain position and/ or with small or sterically hindered active sites tend to be poorly inhibited by this type of compound.
There are relatively few developed inhibitors which are analogues of the transition state/deprotonated intermediate, perhaps because the early inhibitors developed against proline racemase were not highly effective and one of these inhibitors was an unstable imine. 135 Some covalent inhibitors have been developed by rational design or identified by screening techniques. 20,[22][23][24]26,27,46,124,134,217,218 There has been renewed interest in the development of covalent inhibitors in recent years, prompted by a number of covalent drugs coming into clinical use. 201,205,207,208 Covalent drugs acting on racemases and epimerases have all been directed against enzymes using active-site cysteine thiols 20,[22][23][24]26,27,46,124,134,217,218 (amino-acid racemases and epimerases). Covalently reacting drugs containing electrophiles reacting with other active-site bases 205,208 have been under-explored. Both transition-state analogues 135,193,194 and covalent inhibitors 201,[204][205][206]208 offer the potential for high potency and long duration of action and are potentially fertile ground for the future development of inhibitors.
Screening approaches 227 have also been under-used to identify novel inhibitors. There are only five high-throughput screening campaigns in the literature 152,163,200,228,252 and only two fragment-screening campaigns. 257,259 Almost all racemases and epimerases catalyse reversible reactions and this places restriction on these assays but these can be overcome by using an elimination substrate or irreversible coupling enzyme (see section on enzyme assays for examples). The use of coupling enzymes also enables assays based on fluorescence or absorbance to be used, which are readily adaptable to highthroughput screening formats. 227 Direct assaying of racemase or epimerase activity may also be possible using fluorescence anisotropy to monitor ligand binding. 227 Fragment screening using assays of enzyme activity 227,275-279 or biophysical techniques [275][276][277][278][279][280] (particularly X-ray crystallography [274][275][276][277][279][280][281] ) hold significant promise, although the different screening techniques have advantages and disadvantages 280 and different tendencies towards false positive and negative results. 279 There is also a balance to be struck between fragment complexity and affinity to maximise chances of success. 279 Screening of fragment libraries for direct identification of inhibitors is particularly appealing for enzymes with small, enclosed active sites, e.g. proline racemase, 26,27 as the amino-acid substrates are small fragments themselves (M w = 89-204 Da). There have been a number of studies on the screening of small fragments (which will generally have low affinity 279 ), including one using virtual screening initially to triage compounds which resulted in a 40% hit rate for a very small fragment library (fifteen compounds). 282 Similarly, fragment-based screening holds promise for development of inhibitors of enzymes with larger active sites, 227,[275][276][277][278][279][280] although there are challenges associated with identification of different fragments which bind simultaneously and also in the elaboration of fragment hits into leads. 274,283 It is important that inhibitors produced by rational design, identified by screening and other approaches are fully characterised to determine if covalent modification of the target is occurring. There are examples of rationally designed racemase inhibitors intended to be reversible which appear to exert their effects by covalent modification of the racemase target. 150 Several inhibitors identified by screening approaches 239,252,259 could also potentially inhibit their targets by covalent modification. Unselective modification of off-target proteins or other biological molecules could give rise to significant toxicities. [201][202][203][204][205][207][208][209]215 Therefore, it is important to balance this potential draw-back with the advantages of covalent inhibition. Microcystis aeruginosa aspartate racemase MMP0739

Conflicts of interest
There are no conflicts of interest to declare.