Old Yellow Enzyme-Catalysed Asymmetric Hydrogenation: Linking Family Roots with Improved Catalysis

: Asymmetric hydrogenation of activated alkenes catalysed by ene-reductases from the old yellow enzyme family (OYEs) leading to chiral products is of potential interest for industrial processes. OYEs’ dependency on the pyridine nucleotide coenzyme can be circumvented through established artiﬁcial hydride donors such as nicotinamide coenzyme biomimetics (NCBs). Several OYEs were found to exhibit higher reduction rates with NCBs. In this review, we describe a new classiﬁcation of OYEs into three main classes by phylogenetic and structural analysis of characterized OYEs. The family roots are linked with their use as chiral catalysts and their mode of action with NCBs. The link between bioinformatics (sequence analysis), biochemistry (structure–function analysis), and biocatalysis (conversion, enantioselectivity and kinetics) can enable an early classiﬁcation of a putative ene-reductase and therefore the indication of the binding mode of various activated alkenes.


Introduction
The 2001 Nobel prize in chemistry awarded to William S. Knowles and Ryōji Noyori internationally highlighted the importance of catalysed asymmetric hydrogenation reactions [1]. Particularly, the creation of one to two chiral centres through asymmetric hydrogenation of C=C bonds is a highly valuable reaction in organic synthesis [2]. Common synthetic routes for cis-hydrogenation are accomplished via homogeneous chiral catalysts composed of precious metals such as rhodium (Rh), ruthenium (Ru) or iridium (Ir), and phosphine ligands such as chiral monoand di-phosphines, C 2 -symmetric bisoxazoline ligands or C 2 -symmetric N-heterocyclic carbenes, respectively [3]. In comparison, synthetic methods for asymmetric trans-hydrogenation to afford the stereo-complementary products are scarce [4].
OYEs are flavin mononucleotide (FMN)-containing ERs and catalyse the selective asymmetric reduction of activated C=C bonds at the expense of the pyridine nucleotide coenzyme NAD(P)H, following a bi-bi ping-pong kinetic mechanism (Scheme 1). In the reductive half-reaction, FMN is reduced through hydride transfer from NAD(P)H (C4) [37]. In the oxidative half-reaction, a hydride is transferred from the N5-atom of the reduced flavin to the Cβ-atom of the activated alkene. A tyrosine residue provides a proton to the Cα-atom, thus completing the reduction of the C=C bond [9,37,38]. This mechanism leads to an anti-addition (trans-fashion) hydrogenation and is supported by recent quantum mechanics/molecular mechanics calculations [39]. The reductive half-reaction was experimentally investigated in detail for OYE1 by Massey and co-workers [40]. Binding of NADPH to the oxidised enzyme-FMN complex led to the observation of a transient concentration-dependent Michaelis complex. After NADPH binding, generation of the reduced enzyme-NADP + complex was noticed as a long wavelength absorbance band. Formation of this charge-transfer complex indicated that the electron and subsequent hydride transfer requires π-π stacking between the pyridinium ring of the nicotinamide cofactor and the isoalloxazine ring of the FMN [40,41].
OYEs are flavin mononucleotide (FMN)-containing ERs and catalyse the selective asymmetric reduction of activated C=C bonds at the expense of the pyridine nucleotide coenzyme NAD(P)H, following a bi-bi ping-pong kinetic mechanism (Scheme 1). In the reductive half-reaction, FMN is reduced through hydride transfer from NAD(P)H (C4) [37]. In the oxidative half-reaction, a hydride is transferred from the N5-atom of the reduced flavin to the Cβ-atom of the activated alkene. A tyrosine residue provides a proton to the Cα-atom, thus completing the reduction of the C=C bond [9,37,38]. This mechanism leads to an anti-addition (trans-fashion) hydrogenation and is supported by recent quantum mechanics/molecular mechanics calculations [39]. The reductive half-reaction was experimentally investigated in detail for OYE1 by Massey and co-workers [40]. Binding of NADPH to the oxidised enzyme-FMN complex led to the observation of a transient concentration-dependent Michaelis complex. After NADPH binding, generation of the reduced enzyme-NADP + complex was noticed as a long wavelength absorbance band. Formation of this charge-transfer complex indicated that the electron and subsequent hydride transfer requires π-π stacking between the pyridinium ring of the nicotinamide cofactor and the isoalloxazine ring of the FMN [40,41]. Scheme 1. OYE-catalysed asymmetric hydrogenation of activated alkenes through a bi-bi ping-pong mechanism producing one to two chiral centres. AD = adenine dinucleotide; R = ribose phosphate; EWG = electron withdrawing group.
Since 2013, several OYEs were found to exhibit high catalytic activities with different NCB analogues [52,[57][58][59]. In a first study, YqjM, TsOYE and RmOYE were screened against NCBs 1-5 ( Figure 1) [52], followed by more extensive kinetic studies with a panel of OYEs [59]. Depending on the applied NCBs, OYEs differ in their reaction rates and the catalytic efficiency (kcat/Km) is at times even higher with NCBs than with the natural coenzyme, as discussed further in Section 4 [22,58,59].
In this review, we suggest linking the evolutionary history of OYEs with their activity with NCBs and their use as chiral catalysts on various substrates. To this end, we start with a phylogenetic classification of OYEs characterised thus far, and relate this classification to their structural and biocatalytic properties.

Phylogenetic Classification of OYEs
OYEs are ubiquitous in Nature [21,60]. Many ERs from the OYE family have been (recombinantly) expressed and characterised over the last 25 years. Currently, we have access to approximately 63 characterised and "ready-to-use" OYEs from plantae, fungi and bacteria. Tables 1 and 2 indicate the distribution of those well-characterised OYEs, with respect to their domain eukaryota (Table 1) and prokaryota (Table 2).
The bacterial OYEs investigated until now have been categorised in classical and thermophilic-like (formerly YqjM-like) enzymes [9]. In 2016, Nizam and co-workers performed a comprehensive study of 424 putative OYEs from 60 fungal species and indicated a novel group among fungal OYEs [63]. However, none of these enzymes has been characterised thus far.
Since 2013, several OYEs were found to exhibit high catalytic activities with different NCB analogues [52,[57][58][59]. In a first study, YqjM, TsOYE and RmOYE were screened against NCBs 1-5 ( Figure 1) [52], followed by more extensive kinetic studies with a panel of OYEs [59]. Depending on the applied NCBs, OYEs differ in their reaction rates and the catalytic efficiency (k cat /K m ) is at times even higher with NCBs than with the natural coenzyme, as discussed further in Section 4 [22,58,59].
In this review, we suggest linking the evolutionary history of OYEs with their activity with NCBs and their use as chiral catalysts on various substrates. To this end, we start with a phylogenetic classification of OYEs characterised thus far, and relate this classification to their structural and biocatalytic properties.

Phylogenetic Classification of OYEs
OYEs are ubiquitous in Nature [21,60]. Many ERs from the OYE family have been (recombinantly) expressed and characterised over the last 25 years. Currently, we have access to approximately 63 characterised and "ready-to-use" OYEs from plantae, fungi and bacteria. Tables 1 and 2 indicate the distribution of those well-characterised OYEs, with respect to their domain eukaryota (Table 1) and prokaryota ( Table 2).
The bacterial OYEs investigated until now have been categorised in classical and thermophilic-like (formerly YqjM-like) enzymes [9]. In 2016, Nizam and co-workers performed a comprehensive study of 424 putative OYEs from 60 fungal species and indicated a novel group among fungal OYEs [63]. However, none of these enzymes has been characterised thus far.
The discovery of a second OYE subclass twelve years ago led Macheroux and co-workers to publish the structure of YqjM, an OYE from the Bacillus subtilis strain 168 [75,106]. In contrast to all other OYEs known at the time, YqjM exhibited unique structural properties, such as its occurrence as a homotetramer, and the presence of an arginine at the C-terminus involved in the substrate binding of the adjacent monomer. Next, the thermostable TsOYE (formerly CrS) and TOYE were isolated from Thermus scotoductus SA-01 and Thermoanaerobacter pseudethanolicus, respectively [77][78][79]. Based on a sequence alignment of known and putative OYE homologues from mesophilic and thermophilic organisms, the renaming of the "YqjM" subclass into the "thermophilic-like" was proposed by Scrutton and co-workers in 2010 [77][78][79]. Subsequently, the number of characterised OYEs in this subclass has risen to sixteen ( Table 2, highlighted in green).
OYEs from the thermophilic-like class possess shorter protein sequences (between 337 and 371 amino acid residues) than classical OYEs (between 349 and 412 amino acid residues). Thermophilic-like OYEs have an average increased thermal stability compared to classical ones. ). The thermostability of TsOYE and TOYE was assigned to a high proline content within loops and turns (typical for Thermus species) as well as to strong inter-subunit interactions through hydrogen bonding and complex salt bridge networks at the dimerization interface [77,78].
We recently described FOYE-1 as a thermostable OYE [23]. Surprisingly, FOYE-1 showed highest sequence identity (55% and 50%), and therefore closest phylogenetical relationship, to the mesophilic counterparts RmER and DrER [23]. DrER and RmER are not an exception with respect to the non-thermostable OYE relatives YqjM, XenA, and OYERo2, all clustering in the thermophilic-like subclass. These OYEs have a proline content below 7%, and are mostly stabilized through single salt bridges. Due to these varieties in the thermophilic-like subclass, we suggest the classification be updated through a phylogenetic analysis of all biochemically characterized OYEs ( Figure 2).  Tables 1 and 2. Class I (yellow) contains classical OYEs originating from plants and bacteria. Class II (grey) contains classical OYEs originating only from fungal species. Class III (green) contains the thermophilic-like and mesophilic OYEs originating from various bacteria. The maximum likelihood distance tree (Mega7-mac computed) was calculated with replications of 500 bootstraps. Corresponding values are shown as nodes at all branches. The corresponding alignment was produced via ClustalW alignment applying the GONNET protein weight matrix.
Calculating the phylogenetic distance tree of 63 characterised OYEs revealed three instead of two comprehensive branches ( Figure 2). Branch 1 (highlighted in yellow) contains many classical OYEs from plants, actino-and proteobacteria, flavobacteria, but also another distinct subclade from cyanobacteria. We designate this branch "class I". Branch 2 ("class II", highlighted in grey) appears closely related to branch 1 and contains exclusively classical enzymes from fungi. Branch 3 ("class III", highlighted in green) is further away from branch 1 and 2 and contains-like class I-not only bacterial OYEs from actino-, proteo-and cyanobacteria, but also OYEs from deinococci, flavobacteria and firmicutes. This branch contains the traditional thermophilic-like OYEs. Furthermore, several sequences (SYE4, Chr-OYE1, Nox, LacER and YqiG) occurred in a position where they cannot be assigned yet. They may represent evolutionary intermediates/predecessors of class I and class III enzymes as they are of bacterial origin. The short distance between class I and class II hints towards a close relatedness and therefore a co-driven evolution. Our phylogenetic analysis allowed no  Tables 1 and 2. Class I (yellow) contains classical OYEs originating from plants and bacteria. Class II (grey) contains classical OYEs originating only from fungal species. Class III (green) contains the thermophilic-like and mesophilic OYEs originating from various bacteria. The maximum likelihood distance tree (Mega7-mac computed) was calculated with replications of 500 bootstraps. Corresponding values are shown as nodes at all branches. The corresponding alignment was produced via ClustalW alignment applying the GONNET protein weight matrix.
Calculating the phylogenetic distance tree of 63 characterised OYEs revealed three instead of two comprehensive branches ( Figure 2). Branch 1 (highlighted in yellow) contains many classical OYEs from plants, actino-and proteobacteria, flavobacteria, but also another distinct subclade from cyanobacteria. We designate this branch "class I". Branch 2 ("class II", highlighted in grey) appears closely related to branch 1 and contains exclusively classical enzymes from fungi. Branch 3 ("class III", highlighted in green) is further away from branch 1 and 2 and contains-like class I-not only bacterial OYEs from actino-, proteo-and cyanobacteria, but also OYEs from deinococci, flavobacteria and firmicutes. This branch contains the traditional thermophilic-like OYEs. Furthermore, several sequences (SYE4, Chr-OYE1, Nox, LacER and YqiG) occurred in a position where they cannot be assigned yet. They may represent evolutionary intermediates/predecessors of class I and class III enzymes as they are of bacterial origin. The short distance between class I and class II hints towards a close relatedness and therefore a co-driven evolution. Our phylogenetic analysis allowed no distinction between thermostable and non-thermostable OYEs within class III since thermostability does not display a logical pattern.
Among conventional phylogeny methods, we considered the amino acid composition of all analysed OYEs regarding their early/simple (particularly Ala, Thr, and Val) and late/sophisticated (particularly Cys, Leu, Phe, Trp, and Tyr) amino acid residues [107,108]. The quotient of late over early amino acids was used to determine a comparable time of evolvement factor. Therefore, a low number represents a primary chain mostly containing early/simple amino acids. Regarding the OYEs from class I without the subclade from cyanobacteria, an average factor of 0.76 was determined. By contrast, class II's average factor amounted to 0.94. This result would indicate a successive evolution during which class II evolved from class I. However, the more remote class III could have evolved in a convergent evolution with class I since the average factor was similar (0.77). The convergent evolution hypothesis is strengthened by the difference in average amino acid length as well as by the biochemical and structural properties, which both differ among class I and class III. OYEs from cyanobacteria (included in class I) seem also to be recruited later due to the average factor of 0.88. To intensify this finding, main properties of class I, II and III regarding their sequences and structures are discussed in the following section dealing with the structural classification of those three classes. Furthermore, this information enables to subclassify class III OYEs and to assign the lost proteins SYE4, Chr-OYE1, Nox, LacER and YqiG to one of the three classes or to confirm their independent status, respectively.

Structural Classification of OYEs: From Sequence to Structure
OYE homologues show high conservation of amino acids involved in the binding of flavin, substrates and/or inhibitors, as well as for those involved in the formation of the dimeric interface. However, there are many differences in the conservation of distinct residues depending on the class of OYEs mentioned above. A multiple sequence alignment was performed with five representative sequences from each class ( Figure S1). This section compares the sequences of OYE classes I-III enzymes and discusses the significant effects on the structure of OYEs.

Multiple Sequence Alignment
Approximately 15% (56 amino acids (aa)) are OYE-conserved residues in all three classes ( Figure S1, highlighted in black). Moreover, 40 additional aa are conserved especially in class I (highlighted in yellow), 32 aa in class II (highlighted in grey) and 43 aa in class III (highlighted in green). Evidently, class I and class II share significantly more conserved residues when compared to class III, resulting in the minor distance ( Figure 2).
The alignment including the five non-assignable sequences also allows matching of these OYE enzymes with the defined classes. The sequences SYE-4 and Chr-OYE1 are more closely related to class I since they share 63% and 45% of the class I conserved residues but only 34% (25%) for class II and 9% (21%) for class III, respectively. Nox and YqiG are more related to class III, sharing 37% and 40% of class III conserved residues whereas they jointly own only 23% (34%) of class I, and 25% (19%) of class II conserved residues. LacER shares 39% (class I), 19% (class II) and 30% (class III) and is somewhere in between. Interestingly, all five proteins do not possess the Cys26 and the Arg336 (YqjM numbering), which are highly conserved in class III and are involved in flavin and substrate binding. The alignment shows they contain many general OYE motifs. However, all of them comprise motifs from class I/II but also from class III generating a reasonable alignment in between of classes I-III. Therefore, small substitutions of essential amino acids may change binding properties or oligomeric state performance making those enzymes promising candidates for bioengineering.

Monomeric Structure and Dimeric Interface
Members of the OYE family were found to exist in different oligomeric states. A remarkable property observed for class III OYEs is their occurrence in solution as homodimers and homotetramers [22,76,105]. Even higher species as octamers and dodecamers, emerged from functional homodimers, were observed for TOYE and TsOYE [77,79]. A shift between higher multiple oligomeric states was noticed depending on the protein concentration [22,77]. On the other hand, members of class I and class II were found to exclusively occur as monomers or homodimers [14,83,109]. Nevertheless, the basic monomer structure is very similar. A typical single domain is represented by an (α,β) 8 -barrel structure (TIM barrel) with additional secondary structure core elements, in which FMN is bound at the C-terminal end (Figure 3). Secondary structure prediction showed that these core elements are similar in location and length in all three classes ( Figure S1, SecStruc). An N-terminal hairpin of two short β-strands (βA and βB) builds the bottom of the barrel prior to strand β1 (OYE1 numbering) (Figure 3, blue hairpin) [75,78,83]. functional homodimers, were observed for TOYE and TsOYE [77,79]. A shift between higher multiple oligomeric states was noticed depending on the protein concentration [22,77]. On the other hand, members of class I and class II were found to exclusively occur as monomers or homodimers [14,83,109]. Nevertheless, the basic monomer structure is very similar. A typical single domain is represented by an (α,β)8-barrel structure (TIM barrel) with additional secondary structure core elements, in which FMN is bound at the C-terminal end (Figure 3). Secondary structure prediction showed that these core elements are similar in location and length in all three classes ( Figure S1, SecStruc). An N-terminal hairpin of two short β-strands (βA and βB) builds the bottom of the barrel prior to strand β1 (OYE1 numbering) (Figure 3, blue hairpin) [75,78,83]. Additional secondary structure elements occur on surface loops between core building blocks of β-strands and α-helices. The largest loop occurs between β3 and α3 and differs greatly between all OYE representatives (Figure 3, green surface loop) [75,78,83]. Class III enzymes are up to 40 amino acids shorter and therefore more compact. This is due to the shortening of the N-terminal surface loops between β3 and α3, but also between β5 and α5 and at the C-terminal end of the protein ( Figure S1). However, the capping domain between β3 and α3, which is partly covering the entrance of the active site, differs greatly in length and type of structural units within subclass III [78]. Each monomer within the functional dimer is facing the central hole with the same side and contributes the same residues to the hydrogen bond network that keeps monomers together [77]. Residues Gln330 and Tyr331 (TOYE numbering) are highly conserved and seem to play a role in monomer interaction [77]. While Gln330 is specific for class III enzymes, Tyr331 is conserved for all OYEs ( Figure S1). A variety of additional residues assist in the formation of the functional dimer-dimer interface such as Thr45, Ser28/His42/Arg46 and Tyr315 (TOYE numbering) [77,78]. Many of them are conserved only for class III enzymes. The most influential factor for the formation of the functional dimer-dimer interface of class III is Arg333 (TOYE numbering). This "arginine finger" stretches into the active site of the adjacent monomer and interacts with the respective flavin cofactor. An additional stabilization of some class III enzymes originating from extremophiles is associated with an increased proline content of surface loops [23,78], and with the presence of three complex salt Additional secondary structure elements occur on surface loops between core building blocks of β-strands and α-helices. The largest loop occurs between β3 and α3 and differs greatly between all OYE representatives (Figure 3, green surface loop) [75,78,83]. Class III enzymes are up to 40 amino acids shorter and therefore more compact. This is due to the shortening of the N-terminal surface loops between β3 and α3, but also between β5 and α5 and at the C-terminal end of the protein ( Figure S1). However, the capping domain between β3 and α3, which is partly covering the entrance of the active site, differs greatly in length and type of structural units within subclass III [78]. Each monomer within the functional dimer is facing the central hole with the same side and contributes the same residues to the hydrogen bond network that keeps monomers together [77]. Residues Gln330 and Tyr331 (TOYE numbering) are highly conserved and seem to play a role in monomer interaction [77]. While Gln330 is specific for class III enzymes, Tyr331 is conserved for all OYEs ( Figure S1). A variety of additional residues assist in the formation of the functional dimer-dimer interface such as Thr45, Ser28/His42/Arg46 and Tyr315 (TOYE numbering) [77,78]. Many of them are conserved only for class III enzymes. The most influential factor for the formation of the functional dimer-dimer interface of class III is Arg333 (TOYE numbering). This "arginine finger" stretches into the active site of the Catalysts 2017, 7, 130 9 of 24 adjacent monomer and interacts with the respective flavin cofactor. An additional stabilization of some class III enzymes originating from extremophiles is associated with an increased proline content of surface loops [23,78], and with the presence of three complex salt bridge networks at the dimerization interface that increases the subunit interaction strength. The complex five residue salt bridge between four residues of α2 and an asparagine from α1 is not conserved for class III enzymes. However, incorporation of this salt bridge within another class III enzyme by site-directed mutagenesis indeed increased the thermal stability [22].

FMN Binding
The FMN prosthetic group is bound at the C-terminal end of the β-strands, a typical location for the active sites of TIM-barrel enzymes (Figure 3) [110]. The re side of the flavin is facing the protein and is completely hidden from the active site but in contact with strand β1 [111,112]. The si side of the flavin is facing a solvent filled access channel (20 Å length in PETNR) and therefore forms the bottom of a wide-open active pocket mainly assembled with aromatic residues [9]. The redox potential of the FMN cofactor is controlled by the different interactions between the protein and the flavin. The N3 and O2 atoms of the flavin pyrimidine ring are interacting with a glutamine (Gln102 in YqjM) and the N1 and O2 with an arginine (Arg215 in YqjM) ( Figure 4) [75]. Both residues are strictly conserved in OYEs. Furthermore, a conserved histidine pair is described for all OYEs to be in hydrogen bonding distance. In YqjM (class III), His167 and His164 donate a hydrogen bond for the N1 or N3 atom of the flavin, respectively (Figure 4) [75]. However, in PETNR (class I) the pursuant histidines (His184 and His181) donate the hydrogen bond to the flavin O2 as well as to the activating group of the substrate/inhibitor [111]. While both histidines are highly conserved in class III, the second histidine is replaced by an asparagine in several class I and class II enzymes (Figure 4 and Figure S1). For example, the corresponding asparagine (Asn194) in OYE1 is known to play a key role in ligand binding [83,111]. Moreover, residues Ala60 and Cys26 (YqjM) were found to be in hydrogen bonding distance to the FMN O4 atom (Figure 4) [75]. While Ala60 is conserved within all classes (sometimes glycine in class II), the cysteine residue is a unique feature only for class III enzymes. Cys26 was shown to interact not only with the O4-atom of the flavin but also with Tyr28 (Figure 4), which is also conserved in class III. In class I/II, Cys26 is replaced by a conserved threonine (Thr37 in OYE1), which is as well within hydrogen bonding distance to the O4-atom of the isoalloxazine ring [112]. It was shown that both residues modulate the flavin reduction potential after they were exchanged by mutagenesis [112][113][114].
The dimethylbenzene moiety of the flavin isoalloxazine ring was described to interact with the hydrophobic residues Met25, Leu311 and Arg308 (YqjM [75]). The methionine is conserved only for class III enzymes as well as the leucine, which is replaced by isoleucine in class I/II. The arginine is also highly conserved for class II but not for class I enzymes. In PETNR, a tyrosine (Tyr351) is in van der Waals contact with the edge of the dimethylbenzene nucleus [111]. This tyrosine is conserved for class I/II. In OYE1 the hydrophobic area around the dimethylbenzene ring is more compact and contains two interacting phenylalanine residues (Phe250 and Phe296). Phe250 is highly conserved in class II enzymes but also occurs in PETNR (Phe240). However, Phe296 is neither conserved in class I nor in class II. Instead of the phenylalanines, class III enzymes possess an arginine (Arg336 in YqjM), a residue involved in both, flavin binding and formation of the dimeric interface (vide supra).
The ribityl chain of the flavin is stabilized in all classes by a conserved proline and an arginine (Pro24 and Arg215: YqjM numbering). Additionally, class III enzymes were also found to anchor the cofactor in this part to Ser23, Ser249 and Gln265 (YqjM numbering) [75]. However, Figure S1 shows that Ser23 is not conserved in class III OYEs, as previously published [9]. classes (sometimes glycine in class II), the cysteine residue is a unique feature only for class III enzymes. Cys26 was shown to interact not only with the O4-atom of the flavin but also with Tyr28 (Figure 4), which is also conserved in class III. In class I/II, Cys26 is replaced by a conserved threonine (Thr37 in OYE1), which is as well within hydrogen bonding distance to the O4-atom of the isoalloxazine ring [112]. It was shown that both residues modulate the flavin reduction potential after they were exchanged by mutagenesis [112][113][114].

Coenzyme and Inhibitor Binding
In co-crystallisation studies of OYEs, the F o -F c electron density map of the uncomplexed, oxidized enzyme gives a strong positive peak above the flavin isoalloxazine ring. This observation is due to the binding of an anion from the crystallisation solution such as sulphate [75], chloride [83,111], acetate or formate [77]. Addition of NADPH resulted in the replacement of the sulphate by two water molecules and in the reduction of the flavin [75]. In agreement with a bi-bi ping-pong mechanism, great similarity exists among OYEs in the binding mode of the nicotinamide moiety of NADPH and phenolic inhibitors such as para-hydroxybenzaldehyde (p-HBA) ( Figure 5) or para-nitrophenol (p-NP). Both aromatic rings are oriented through π-π stacking with the FMN isoalloxazine ring and hydrogen bonding with His167/Asn194 and His164/His191 (YqjM/OYE1) [115]. The crystal structure of OYE1 (class II) with an NADP + analogue showed that the oxygen of the amide on the pyridinium ring is (hydrogen) bonding with the two conserved histidine residues, thus positioning the C4 atom close to the N5 atom of FMN for the hydride transfer [83]. The same position was observed for the nicotinamide ring of tetrahydro-NADH in MR (class I) ( Figure 5) and TOYE (class III) [77,116].
Analogues of histidine/asparagine residues are found in all classes of OYE homologues. A remarkable difference is the binding of the functional group of the phenolic ligand. For instance the aldehyde group of p-HBA interacts with Tyr375 (OYE1) [83], or Tyr351 (PETNR) [75], respectively. This tyrosine residue is conserved within all OYE classes ( Figure S1), but was never mentioned to play a role in the catalysis of class III enzymes, which use a different strategy to build their active sites. A reorientation of the C-terminal end causes the formation of a part of the active site of the adjacent monomer. Therefore, only an arginine (Arg333 in TOYE) contributes to the opposed active site. The "arginine finger" causes the formation of a strong hydrogen bond with the nitro-group of p-NP [75], or the O1P/O1N atoms of tetrahydro-NADH [77]. Replacement of this arginine might be a biocatalytic tool to broaden/change the substrate spectrum or cofactor specificity of class III enzymes. Another N-terminal class III conserved tyrosine (Tyr28 in YqjM) is involved in binding of the aldehyde oxygen of p-HBA [75,78]. Furthermore, Tyr175/Tyr196 (TsOYE/OYE1) was confirmed to be a proton donor for the substrate 2-cyclohexenone [75,78], and is conserved within all subclasses. To summarize, the binding partner for the functional group of aromatic substrates is always a tyrosine, which is in hydrogen bonding distance. Class I and class II use two tyrosines from the central part (Tyr196 in OYE1) and the C-terminal protein part (Tyr375 in OYE1), whereas class III enzymes involve the central part tyrosine (Tyr175 in TsOYE) as well, but a second N-terminal tyrosine (Tyr25 in TsOYE) takes over the role of the class I/II C-terminal tyrosine due to reorientation. sulphate by two water molecules and in the reduction of the flavin [75]. In agreement with a bi-bi ping-pong mechanism, great similarity exists among OYEs in the binding mode of the nicotinamide moiety of NADPH and phenolic inhibitors such as para-hydroxybenzaldehyde (p-HBA) ( Figure 5) or para-nitrophenol (p-NP). Both aromatic rings are oriented through π-π stacking with the FMN isoalloxazine ring and hydrogen bonding with His167/Asn194 and His164/His191 (YqjM/OYE1) [115]. The crystal structure of OYE1 (class II) with an NADP + analogue showed that the oxygen of the amide on the pyridinium ring is (hydrogen) bonding with the two conserved histidine residues, thus positioning the C4 atom close to the N5 atom of FMN for the hydride transfer [83]. The same position was observed for the nicotinamide ring of tetrahydro-NADH in MR (class I) ( Figure 5) and TOYE (class III) [77,116].  The majority of OYEs display a preference for NADPH over NADH as the source of hydride, as indicated from catalytic efficiencies (k cat /K M ) or apparent dissociation constants (K D ). The specificity ratio of NADPH:NADH, obtained from specific activity with trans-2-hexenal by Bruce and co-workers [117], can vary from 0.02 (MR) to 10 (OYE1), although the cofactor preference for KYE1, XenA and Yers-ER was shown to differ depending on the substrate, yielding different NADPH:NADH specificity ratios, a surprising result given the bi-bi ping-pong mechanism described above [6]. Interestingly, TsOYE displays a similar K D for NADPH and NADH, but a five times higher reaction rate when using NADPH with respect to NADH [59]. The only OYEs displaying a higher affinity for NADH are from class I: NerA/GTNR [98], MR [85], SYE1 and SYE3 [103]. LacER, which falls short of being assigned to a class, although the closest class it relates to seems to be class I as seen above, also shows a preference for NADH [10].
The group of Hauer recently showed examples of grafted β/α surface loop regions between OYEs MR, NCR and OYE1 that led to altered as well as new reaction activities and a change in NADH interactions [118][119][120]. One example was showcased with NCR (class I), which displayed lower activity with increasing NAD + present in the reduction reaction of cinnamaldehyde [118][119][120]. Through β/α surface loop grafting from OYE1 (class II) and MR (class I) to form various loop variants of NCR, the loss of activity with higher presence of NAD + was significantly reduced [118][119][120].

Biocatalytic Conversions
From the full biocatalytic reaction for the reduction of the model substrate ketoisophorone, moderate to high conversions were obtained with NADPH and NADH and low to high conversions with NCBs 1-5 ( Figure 6) [59]. NCB 5 gave very low conversions (1-10%), except for XenA (80%), TsOYE (59%) and RmER (43%) [59]. With our new classification, we noted the enzymes accepting NCB 5 were all from class III, and that OYE2-3 from class II gave lower conversions with the NCBs in general.

Kinetic Data: Steady State and Pre-Steady State
The reduction of 2-cyclohexenone under steady-state conditions gave kcat, KM and catalytic efficiency kcat/KM for PETNR (class I), TOYE and XenA (class III) [59]. Noting that class III TOYE and XenA gave the highest rates with the NCBs, XenA in particular displayed high catalytic efficiency (kcat/KM) with NCB 2 (Figure 7).

Biocatalytic Conversions
From the full biocatalytic reaction for the reduction of the model substrate ketoisophorone, moderate to high conversions were obtained with NADPH and NADH and low to high conversions with NCBs 1-5 ( Figure 6) [59]. NCB 5 gave very low conversions (1-10%), except for XenA (80%), TsOYE (59%) and RmER (43%) [59]. With our new classification, we noted the enzymes accepting NCB 5 were all from class III, and that OYE2-3 from class II gave lower conversions with the NCBs in general.

Kinetic Data: Steady State and Pre-Steady State
The reduction of 2-cyclohexenone under steady-state conditions gave k cat , K M and catalytic efficiency k cat /K M for PETNR (class I), TOYE and XenA (class III) [59]. Noting that class III TOYE and XenA gave the highest rates with the NCBs, XenA in particular displayed high catalytic efficiency (k cat /K M ) with NCB 2 (Figure 7). The rates of the reductive half-reaction showed that PETNR from class I displayed lower rates when compared to TOYE, XenA and TsOYE from class III ( Figure 8A,B) [59]. TOYE gave the highest reduction rates followed by XenA and TsOYE. NCB 3 clearly afforded the highest reduction rate values, but also the highest dissociation constant KD for the enzyme-reduced nicotinamide complex ( Figure 8C,D). For each OYE, the order of the best to poorer NCB differed ( Figure 8A,B). Fast reaction kinetics with MR and PETNR (class I) by the group of Scrutton led to the conclusion that quantum mechanical tunnelling is crucial for the hydride transfer from the nicotinamide cofactor to FMN [121]. They proposed that sub-angstrom differences in the donor-acceptor distance affect the probability of hydride transfer [121][122][123]. From molecular modelling studies with TsOYE and NCBs 1-5 [52], as well as with the crystal structures of XenA [59], The rates of the reductive half-reaction showed that PETNR from class I displayed lower rates when compared to TOYE, XenA and TsOYE from class III ( Figure 8A,B) [59]. TOYE gave the highest reduction rates followed by XenA and TsOYE. NCB 3 clearly afforded the highest reduction rate values, but also the highest dissociation constant K D for the enzyme-reduced nicotinamide complex ( Figure 8C,D). For each OYE, the order of the best to poorer NCB differed ( Figure 8A,B). The rates of the reductive half-reaction showed that PETNR from class I displayed lower rates when compared to TOYE, XenA and TsOYE from class III ( Figure 8A,B) [59]. TOYE gave the highest reduction rates followed by XenA and TsOYE. NCB 3 clearly afforded the highest reduction rate values, but also the highest dissociation constant KD for the enzyme-reduced nicotinamide complex ( Figure 8C,D). For each OYE, the order of the best to poorer NCB differed ( Figure 8A,B). Fast reaction kinetics with MR and PETNR (class I) by the group of Scrutton led to the conclusion that quantum mechanical tunnelling is crucial for the hydride transfer from the nicotinamide cofactor to FMN [121]. They proposed that sub-angstrom differences in the donor-acceptor distance affect the probability of hydride transfer [121][122][123]. From molecular modelling studies with TsOYE and NCBs 1-5 [52], as well as with the crystal structures of XenA [59], Fast reaction kinetics with MR and PETNR (class I) by the group of Scrutton led to the conclusion that quantum mechanical tunnelling is crucial for the hydride transfer from the nicotinamide cofactor to FMN [121]. They proposed that sub-angstrom differences in the donor-acceptor distance affect the probability of hydride transfer [121][122][123]. From molecular modelling studies with TsOYE and NCBs 1-5 [52], as well as with the crystal structures of XenA [59], the NCBs were observed to be in the correct position, π stacking with the FMN isoalloxazine ring, and the amide oxygen within hydrogen bonding distance of the two conserved histidine residues (His181 and His178 in XenA). The exception was NCB 5, which contains a nitrile substituent in place of the amide, thus missing the oxygen and only hydrogen bonding with one of the two histidines. This analogue was used as a hydride donor by the OYEs but with less efficiency. Although the coenzyme biomimetics bind to the OYEs' active site in a similar way as NAD(P)H, small changes in the orientation of the nicotinamide ring could affect the rate of hydride transfer. Indeed, different kinetic data was observed for all analogues and OYEs across the three classes. The NCB giving the highest reduction rate for one OYE, NCB 2 for PETNR (class I), was different for another OYE, NCB 3 for TOYE, XenA and TsOYE (class III) [59].
The overlaid crystal structures of XenA (from P. putida II-B, class III) with the tetrahydro form of cofactors NADPH 4 and NCB 1-2,4 showed only minimal changes. One exception in all three NCBs structures was the alternative conformation of the tryptophan residue (Trp302 in XenA), which reduced the volume of the active site [59].

Classification of OYEs with Respect to Substrates
As listed in the introduction, OYEs can reduce a variety of α,β-unsaturated activated alkene substrates, which are continuously being extended with new activating groups explored for industrial purposes. In general, the review by Toogood, Gardiner and Scrutton gives an overview of substrate family scopes for OYEs [9]. Looking at the compilation of substrates, there is a large overlap in the substrate scope and stereochemical outcome of OYEs within and across the three classes, and also wide differences within one class. Unfortunately, reported data often varies; in some cases, the specific activity was reported for each substrate, in other cases, percentage conversion of the biocatalytic reactions was reported. Furthermore, reaction times vary (4-48 h) and a range of enzyme concentrations is used, which makes identification of preferred substrates more difficult.
Other α-substituted unsaturated cyclic ketones such as ketoisophorone and carvones are also alternative substrates for all three classes. However, several non-thermostable class III enzymes such as RmER, OYERo2 [22,80], and FOYE-1 (unpublished data) gave very low conversions on those substrates. Poor conversions are supported by very poor activities towards these typical substrates (ketoisophorone or cyclohexenone), mainly in class III (YqjM, FOYE-1, OYERo2, XenA, TsOYE, DrOYE, and RmOYE) [22,23]. In contrast, these enzymes showed highest specific activities on several maleimides (up to 70 U/mg) [22,23]. Comparably, the specific activities on maleimides for classes I and II enzymes are much lower [6,8,21], but all classes reduce maleimides with high conversion yield ( Figure 10). YqiG is highly active towards maleimide (56 U/mg) and shows little activity toward carboxylic acids and cyclic ketones (<3 U/mg) [104]. This non-assigned OYE exhibits high temperature stability and is closely related to class III (see Section 3). With ketoisophorone, (R)-carvone, 2-methyl-N-phenylmaleimide and 2-methylmaleimide, all investigated OYEs led to the (R)-enantiomer ( Figure 11). In general the enantiomeric excess was lower with ketoisophorone due to known product racemisation [8]. (Dimethylated) dicarboxylic acids show remarkable differences between the three classes. Itaconic acid was only converted by class II enzymes (average conversion: 25%). The two isomers, mesaconic acid and citraconic acid, were transformed by all three classes.
However, class II showed on average 50% higher conversions with respect to the other two classes and always preferred a distinct enantiomeric product: (S)-enantiomer with itaconic and citraconic acid; (R)-enantiomer with mesaconic acid. Only with the cis-substrate (citraconic acid), class III was the more active enzyme class ( Figure 10). Class I enzymes exclusively led to the (R)-enantiomer with dimethylated dicarboxylic acids, whereas the stereochemical outcome for class III enzymes was not investigated until now. The aliphatic dimethylated dienal substrate citral was shown to be a great substrate for class I and class II. Only low to modest citral conversions were obtained within class III with YqjM and DrER ( Figure 10) [12,80], as well as with TsOYE [52].
To summarise, class I and class II prefer cyclic ketones, especially 2-methylcyclohexenone, ketoisophorone and carvones. On dicarboxylic acids, class II enzymes show best conversion rates. Maleimides are excellent substrates for the entire OYE family, class III enzymes especially, displaying remarkable activities on maleimides (about 50-times higher than on all other substrates).
A correlation between the protein structure and the sequence patterns for the stereopreference with respect to reduction of 1-nitro-2-phenylpropene was attempted to be established for 12 independent OYE structures [125]. LeOPR3, NCR, YqjM, XenA, TOYE, TsOYE, and GkOYE reduced 1-nitro-2-phenylpropene to the S-enantiomeric product whereas LeOPR1, OYE1, PETNR and NemR produced the (R)-enantiomer ( Figure 11). However, the group of Pietruszka showed that the stereoselectivity of an uncharacterised OYE could not be predicted solely based on the primary sequence pattern of the surface loop β2 [126]. Stereoselectivity was also investigated by the group of Stewart, who determined how subtle changes in the residues control the orientation of substrate binding in OYEs [127,128], and showed that mutations can lead to inverted enantioselectivity [129,130]. The stereochemical outcome was influenced by mutagenesis of Trp116 in OYE1. For instance, (S)-carvone displayed an inverted stereoselectivity for six different variants. The same observation was made for (Z)-β-aryl-β-cyanoacrylate when Trp116 was changed. Substitution of this amino acid against smaller residues resulted in a classical orientation of the substrate and gave the (S)-enantiomer. Substitution against more bulky residues resulted in a switch of stereoselectivity affording the (R)-enantiomer. This tryptophan is conserved for classes I and II enzymes and might be a major residue for changing enantioselectivity of different prochiral substrates [25]. Class III enzymes possess a conserved alanine at this position. Directed evolution performed on YqjM by Reetz and co-workers enabled control of the enantioselectivity for the bioreduction of substituted cyclopentenone and cyclohexenone substrates [131]. A recent review by Stewart and Amato on protein engineering of OYEs gives a good overview on the possibilities of stereochemical outcome [130].

Conclusions
We provide a new classification for the currently available and ready to use 63 OYEs, divided into three classes. The structural classification of OYEs displayed the conserved as well as the differing residues among the three classes.
Class I are monomeric OYEs originating from plants and bacteria. They prefer α-methylated cyclic enones as substrates and showed highest reduction rates with NCB 2. Class II enzymes are fungal OYEs likely developed in a co-driven evolution from class I. Both classes are phylogenetically and structurally closely related resulting in similar substrate preferences, although class II members display higher conversions on certain substrates and those tested did not accept NCBs as well.
Class III OYEs are likely driven by a convergent evolution to class I/II resulting in a substantial distance on sequence level according to the dendrogram. Due to remarkable variances in conservation compared to the two foreign subclasses but great conservation within their own members, class III OYEs have a different structure, biochemical behaviour and substrate preference. They form oligomers built from homodimers and are often thermostable. They highly prefer maleimides as substrates and NCB 3 as the hydride donor, and accept the nitrile-substituted NCB 5 best of all classes.
The symbiosis of bioinformatics, biochemistry and biocatalysis enables a closer analysis of an industrially attractive biocatalyst, OYE, to determine from which class it should be chosen for a targeted product, and to allow further improvements with specially designed NCBs.