Engineering the 2-Oxoglutarate Dehydrogenase Complex to Understand Catalysis and Alter Substrate Recognition

The E. coli 2-oxoglutarate dehydrogenase complex (OGDHc) is a multienzyme complex in the tricarboxylic acid cycle, consisting of multiple copies of three components, 2-oxoglutarate dehydrogenase (E1o), dihydrolipoamide succinyltransferase (E2o) and dihydrolipoamide dehydrogenase (E3), which catalyze the formation of succinyl-CoA and NADH (+H+) from 2-oxoglutarate. This review summarizes applications of the site saturation mutagenesis (SSM) to engineer E. coli OGDHc with mechanistic and chemoenzymatic synthetic goals. First, E1o was engineered by creating SSM libraries at positions His260 and His298.Variants were identified that: (a) lead to acceptance of substrate analogues lacking the 5-carboxyl group and (b) performed carboligation reactions producing acetoin-like compounds with good enantioselectivity. Engineering the E2o catalytic (core) domain enabled (a) assignment of roles for pivotal residues involved in catalysis, (b) re-construction of the substrate-binding pocket to accept substrates other than succinyllysyldihydrolipoamide and (c) elucidation of the mechanism of trans-thioesterification to involve stabilization of a tetrahedral oxyanionic intermediate with hydrogen bonds by His375 and Asp374, rather than general acid–base catalysis which has been misunderstood for decades. The E. coli OGDHc is the first example of a 2-oxo acid dehydrogenase complex which was evolved to a 2-oxo aliphatic acid dehydrogenase complex by engineering two consecutive E1o and E2o components.


Strategies for Protein Engineering
To achieve altered substrate specificity, thermostability and organic solvent stability, enzymes require engineering prior to their use as an industrial biocatalyst. At present, many enzymes which are conserved across species may not have their properties coupled. Thus, an altered substrate specificity may come at the cost of thermostability [1,2]. Rational design is one of the earliest approaches applied for enzyme engineering [3,4]. However, a detailed knowledge of enzyme structure and its active site configurations is required. The enzyme residues targeted for alteration of substrate specificity or catalysis may need to be located within 6Å or less from the substrate-or cofactor-binding sites [5][6][7][8][9]. At present, there is still limited knowledge about the optimization and enzymes engineering by a rational design approach. The complexity of enzyme structure-function relationships is a major limiting factor for this approach.
These limitations could be resolved by directed evolution, where the rapid engineering of enzymes could be achieved without detailed structural knowledge. Thus, it has been reported that substitutions of the residues located far apart from the active site may contribute to substrate recognition [10][11][12]. The directed evolution approach could identify amino acids for substitution that would be impossible to predict by rational design. Frances Arnold et al. pioneered directed evolution in one of the earliest instances, achieving improved stability of subtilisin in dimethyl sulfoxide solvent [13][14][15][16]. Another notable example reported by Arnold et al. was the formation of novel C-Si bonds with an efficiencỹ 15 times over that of a chemical catalyst used for the same purpose. This reaction is not found in any biological system [17]. A combination of rational design and directed evolution can also be an effective engineering strategy. For example, focused libraries could be created by using site saturation mutagenesis (SSM) with detailed knowledge about the active site structure and of the key residues responsible for substrate recognition and catalysis.
SSM is a site saturation random mutagenesis technique used in protein engineering to convert a certain amino acid to all other 19 amino acids and is employed to engineer the substrate specificity and enantioselectivity of enzymes, to improve their thermostability, as well as to introduce a new enzyme function. The application of SSM to ThDP-dependent enzymes is reviewed by Andrews and McLeish [18]. Briefly, benzoylformate decarboxylase (BFDC) catalyzes the non-oxidative decarboxylation of benzoylformate with the formation of CO 2 and benzaldehyde. It also catalyzes the carboligation reaction with the formation of various 2-hydroxy (also called α-hydroxy) ketones [18]. The combination of directed evolution, site-directed mutagenesis and SSM enabled those authors to identify the Leu476Gln and Leu476Ser BFDC variants with improved carboligase activity in organic solvents, while the Leu476Gln and Met365Leu/Leu461Ser variants could accept benzaldehyde derivatives. The SSM on Leu476 led to variants with improved carboligase activity [18]. Another ThDP-dependent enzyme, transketolase (TK), is perhaps more attractive for biocatalysis because it has stereo-specifically controlled C-C bond cleavage and formation. TK from bacterial sources has been successfully engineered to improve activity and stereoselectivity by using double-site saturation mutagenesis [19]. By using a targeted mutagenesis approach in combination with computational design, an improved thermostability of the E. coli TK was demonstrated [20]. While a detailed review of TK as a biocatalyst is beyond the scope of this review, numerous references to the engineering of E. coli TK by using novel approaches could be found in recent publications from Dalby's group [20,21] and in earlier reviews [18,22].
In cases of multienzyme complexes, catalysis involves two or more enzymes, which are related through sequential, coupled, divergent or convergent reactions [23][24][25]. Recently, using the blueprint of natural models for enzyme co-localizations, a scaffolding approach was employed to generate macromolecular structures with improved catalytic efficiencies [26]. This approach is based on the principle of adjusting the spatial distance and molecular ratio between a few enzymes using DNA scaffolds [27], protein scaffolds [28], polymeric particle immobilizations [29], organometallic frameworks [30] and cross-linked enzyme aggregates [31]. For example, glucose oxidase and horseradish peroxidase were immobilized on a DNA scaffold and as a result they exhibited improved catalytic efficiency compared to free enzymes [27]. Recently, a strategy for the construction of a synthetic multienzyme assembly inside E. coli cells was reported, where the genetically engineered virus-like particles were used as protein scaffolds to assemble three enzymes in the terpene biosynthetic pathway into a multienzyme complex shifted to the production of amorpha-4,11-diene, the key precursor of the antimalarial drug artemisinin [32]. At present, the co-immobilization of multiple cascade enzymes remains challenging. The different systems for the spatial organization of the multi-enzyme pathways in vitro have been discussed in detail in a recent review by M-J. Seo and C. Schmidt-Dannert [33], including nucleic acidbased scaffolds such as DNA origami, peptide and protein building blocks, self-assembling protein arrays and nanostructures and others.
The spatial arrangement of multiple enzymes to improve their activities in cascade reactions is still debated as the scaffold used could potentially alter the enzyme activity by changing the environment surrounding the assembled enzymes (for review, see ref. [34][35][36] The E. coli 2-oxoglutarate dehydrogenase complex (OGDHc, also known as α-ketoglutarate dehydrogenase complex) is a highly conserved multienzyme complex in the tricarboxylic acid (TCA) cycle and catalyzes the formation of succinyl-coenzyme A (succinyl-CoA) by utilizing 2-oxoglutarate (2-OG) as a substrate according to the overall reaction in Equation (1) and mechanism in Scheme 1 [37][38][39].  [40][41][42]. The E2o monomer has a multidomain structure consisting of the flexible lipoyl domain (LD,~8 kDa), a subunit binding domain (~4 kDa) to which E1o and E3 components are flexibly tethered, and the E2o core or catalytic domain (CD), which determines the stoichiometry of the E1o and E3 binding and the quaternary organization of the E. coli OGDHc (Figure 1) [39][40][41][42][43]. The individual E2o domains are connected by flexible linkers of~28-50 residue long, which are rich in Ala and Pro [40][41][42]. The E. coli E2o assembles into a 24-mer molecular assembly with octahedral symmetry [41,[44][45][46][47]. X-ray and NMR structures are available for the E. coli globular domains of the E1o [39], E2o core domain and E3 ( Figure 1) [48,49]. However, there is no atomic resolution structure reported of an intact 2-oxoacid dehydrogenase complex from any sources. The earlier studies on E. coli pyruvate dehydrogenase complex (PDHc) and OGDHc [50,51] revealed a distant and flexible attachment of the E1 and E3 components, with gaps of 2-4 nm and of 3-5 nm between the E2o core and E1 and E3 components, respectively. Later, Murphy and Jensen showed that, in E. coli PDHc and OGDHc, the E1 and E3 components are flexibly tethered~11 nm from the E2o core rather than being specifically bound to the faces and edges of the E2o core as had been generally accepted [43]. More recently, low resolution cryo-EM reconstruction of the entire native E. coli PDHc has been reported and confirmed the presence of the outer components to the E2o core shell components; however, the individual E1p and E3 components could not be resolved [52]. The EM reconstruction of the human E2o cubic core with octahedral symmetry has been reported by two groups, providing insight into E2o trimers assembly and into formation of the multimeric core [53,54]. However, the electron density in the outer shell could not be assigned due to conformational flexibility [53]. In the absence of a molecular organization of 2-oxo acid dehydrogenase complexes, the molecular mechanism of intermediates channeling between components remains elusive. The generally accepted mechanism of active center coupling is that the lipoyl moieties covalently bound to the ε-amino group of lysine residues of the E2 lipoyl domain interact with the catalytic sites of the E1, E2 and E3 components while transferring the intermediates between the active sites in accord with the so-called "swinging arm" mechanism [55][56][57][58]. Recent studies on human PDHc by cryo-EM revealed more insight into the binding of the peripheral components and intermediates channeling and suggested the "division-of-labor" mechanism [59]. According to a "division-of-labor" mechanism, the formation of local clusters of E2p-E1p and E3BP-E3 with specific functions for E1p-E2p clusters (oxidative decarboxylation of pyruvate and acetyl transfer) and for E3BP-E3 clusters (regeneration of reduced lipoamide) was suggested. The studies also suggested an asymmetric architecture of the human PDHc which could be modulated by the CoA substrate binding [59].
An additional complexity in our understanding of the assembly of 2-oxo acid dehydrogenase complexes originated from the remarkable ability of the bacterial OGDHc to form hybrid complexes with PDHc. Thus, in the cases of Cornybacterium glutamicum and of Mycobacterium smegmatis, the E1o and E2o components are fused into one protein named OdhA with no lipoyl domain present. Here, the lipoyl domain is shared between the OdhA and the E2p component of the PDHc [60][61][62]. Identification of the functional and regulatory cross-talk between the human OGDHc and 2-oxoadipate dehydrogenase (E1a) from the final degradation pathway of L-lysine, L-hydroxylysine and L-tryptophan was demonstrated at Rutgers with collaborators in vitro [63,64] and in vivo [65], suggesting that the well-known OGDHc in the TCA cycle can exist as a hybrid 2-oxoglutaric acid and 2-oxoadipic acid dehydrogenase complex where the E2o and E3 components are shared between the E1o and E1a components [65]. While challenges remain in our understanding of the assembly of 2-oxo acid dehydrogenase complexes, their most important feature for biocatalysis is that this is a self-assembled system for the channeling of intermediates.
The challenge of the studies summarized below, as undertaken jointly in the Jordan and Farinas laboratories, involved engineering of the E1o and E2o components of the E. coli OGDHc by using the SSM method with two goals: to understand the mechanism of the E. coli OGDHc, and to enable chemoenzymatic synthesis of chiral α-hydroxy ketones and of acyl-CoA analogues. The authors also anticipate that this review will serve as a guide for the future engineering of the E. coli OGDHc for biocatalytic applications.

Engineering of the E1o Component of the E. coli OGDHc that Leads to Acceptance of Substrate Analogues Lacking the 5-Carboxyl Group
In this work, we were interested in the engineering of the E. coli E1o active centers to accept substrates lacking the 5-carboxylate group, such as 2-oxovaleric acid (2-OV, Figure 2, top), by constructing saturation mutagenesis libraries at the E1o active site residues suggested to interact with the distal carboxylate of 2-OG [39,68]. According to the X-ray structure of the E. coli E1o [39], the E1o active site has three prominent His residues at positions 260, 298 and 729 ( Figure 3) [39]. These His residues are positioned near the thiazolium ring of the thiamin diphosphate (ThDP) cofactor, suggesting that they are likely involved in substrate binding or catalysis. It was also reported that substitution of His260 and His298 to Ala dramatically reduced the overall activity of NADH production by E. coli OGDHc [39]. These structural and kinetic findings made these residues prime targets for E. coli E1o engineering and optimization [68]. Sitesaturation mutagenesis libraries were constructed for His260, His298 and for the double His260/His298 substitution in E. coli E1o. The libraries were then screened for the E1ospecific activity where the E2o and E3 components were replaced by an artificial electron acceptor, 2,6-dichlorophenolindophenol (DCPIP), towards 2-OG and a non-natural substrate, 2-oxovalerate (2-OV), where a nonpolar methyl group replaces the charged carboxylate.  [39]. The ThDP coordinates were added by superimposing the X-ray coordinates for the E. coli E1o structure with that of the E. coli pyruvate dehydrogenase (E1p) [68].
The screening results revealed that His260 is a crucial residue for E1o function. Randomization at His260 yielded only one E1o variant (His260Glu) that could rescue the activity for 2-OG, albeit at a very low specific activity of~0.001% compared to wild-type E1o (Table 1). DNA sequencing identified several His298 E1o variants with the ability to decarboxylate 2-OG (His298Thr and His298Leu) as well as 2-OV (His298Asp and His 298 Val) ( Table 1). Remarkably, the His298Asp substitution converted E. coli E1o to 2-oxovalerate dehydrogenase with an efficiency about 38-fold higher than that of wild-type E1o, while no activity was detected for this variant with 2-OG (Table 1) [68]. An important observation from these studies was that wild-type E1o could use both pyruvate and 2-OV as substrates; however, the NADH production could not be detected in the overall assay on assembly with E2o and E3 into OGDHc implying discrimination at the E2o level [68]. Next, the 2-oxophosphonate/2-oxophosphinate analogues of 2-OG and 2-OV (succinylphosphonate (SP 2− ) and its monomethyl phosphonate ester (SPME − ), both analogues of 2-OG, and propionylphosphinate (PP − ), the phosphinate analogue of 2-oxovalerate ( Figure 2, middle)) were tested for their ability to form the analogue of the pre-decarboxylation covalent intermediate with ThDP on E1o and its variants according to CD spectroscopic methods established in the Jordan laboratory [71,72]. This experiment would inform about stabilization of the pre-decarboxylation intermediate analogue, and inference of the true short-lived intermediate, in the E1o active centers (see Scheme 1, upper right-hand corner).
It was evident that the His260Asp E1o variant and His260/His298Asn E1o with dual substitution did not display any measurable circular dichroism (CD) band at 297 nm on titration by SPME − , the usual signature reflecting binding of the 2-oxophosphonate analogue. This experiment − provided strong evidence that His260 is crucial and indispensable for 2-OG recognition in accord with the randomization experiment and kinetic studies [68]. In CD studies with the His298 E1o variants, the His 298Asp (K d , SPME− = 659 µM), and His298Val (K d , SPME− = 144 µM) substitutions were not favorable for SPME − binding, but the His298Tyr E1o (K d, SPME− = 6.1 µM) could bind it successfully. However, all His298 substitutions were favorable for PP − binding, the phosphinate analogue of 2-OV (K d, PP− in the 5.2-22.3 µM range). Furthermore, the His260Asp/His298Asn E1o dual substitution was favorable for PP − binding. With an average value of K d, PP− of~7.0 µM, the binding was approximately sixfold weaker compared to the wild-type E1o. The randomization at His298 E1o led to E1o variants being active toward 2-OG; however, the His298Asp and His298Val E1o variants yielded much better activity toward 2-OV [68].
In contrast to the E. coli E1o findings, the randomization at His281 in benzoylformate decarboxylase (BFDC) (the ThDP-dependent enzyme in the mandelate pathway which catalyzes the non-oxidative decarboxylation of benzoylformate to CO 2 and benzaldehyde), led to variants with activities not very different from the wild-type enzyme. These results were inconsistent with earlier predictions by kinetic studies of the Ala-substituted BFDC variants [18,73], suggesting that His281 could be the proton donor to the enamine intermediate. Those studies emphasized an important role of the SSM approach in helping interpret enzyme mechanisms, as Ala scanning solely is incapable of predicting the function of the residues [18]. A detailed summary of the application of SSM to other ThDP-dependent enzymes such as E. coli transketolase and 2-succinyl-5-enolpyruvyl-6hydroxy-3-cyclohexene-1-carboxylate synthase (MenD) can be found in the review by Andrews and McLeish [18]. A major contribution of these SSM studies is not only to aid the understanding of mechanistic studies, but to also serve as guides in the search for novel asymmetric catalytic targets involving ThDP-dependent enzymes.

Scheme 2.
Mechanism of the carboligase side reaction forming acetoin and acetolactate suggested for YPDC and for E. coli pyruvate dehydrogenase [71].
The E. coli E1o, catalyzes the initial formation of the ThDP-bound post-decarboxylation intermediate (C2α-carbanion/enamine) from 2-OG which is reactive towards electrophilic acceptor substrates, such as aromatic substrates and aliphatic aldehydes with different chain lengths but with moderate to excellent enantioselectivity [75]. In our laboratories the carboligation reaction of E. coli E1o was studied by using the natural substrate 2-OG as well as the non-natural substrates 2-OV and 3-methyl 2-oxobutanoic acid (2-OiV) ( Table 2) with no second charge in the 2-oxo acid: one of them with a straight chain (2-OV) and the other with a branched one (2-OiV) [97]. Three aldehyde acceptors, glyoxylate, ethyl glyoxylate and methylglyoxal, were successfully employed with these donor substrates that lead to a variety of chiral products according to CD and 1 H NMR spectra and lead to products in good yield with up to 80-90% enantiomeric excess according to chiral gas chromatography (Table 2) [97]. It was evident that the acetoin-like products displayed different stereospecificities depending on the donor substrates and acceptors employed in the E. coli E1o reaction. Thus, with 2-OG as a substrate and glyoxylate or ethylglyoxylate as acceptors, the (R)-2-hydroxy-3-oxohexanedioic acid and (R)-6-ethoxy-5-hydroxy-4,6-dioxohexanoic acid were formed (Figure 4, panel A). However, when methylglyoxal was used as an acceptor, the (S)--5-hydroxy-4,6-dioxoheptanoic acid was produced (Figure 4, panel A). When 2-OV was used as a substrate and glyoxylate or ethylglyoxylate as acceptors, the E. coli E1o produced (S)-enantiomers (Table 2). Furthermore, with 2-OiV as a substrate and employing either glyoxylate or ethylglyoxylate or methylglyoxal as an acceptor, E. coli E1o produced the (S) enantiomers. The scope of the E. coli E1o carboligation reaction was extended by our groups to the synthesis of chiral carboligation products with multiple functional groups by employing varying substrates, such as 2-oxoadipate (2-OA), 2-oxohexanoic acid and 2-oxo-5-hexenoic acid, which are structurally similar to the carboxylic acids (Figure 2, top) and the straight chain aldehydes as acceptors (Figure 2, bottom) [69]. Formation of chiral products on an analytical scale by E. coli E1o was confirmed by recording the CD spectra of the products formed and by 1 H NMR spectroscopy (see Figure 4, panels A, B for chiral products identified) [69]. More recently, the studies were also extended at Rutgers to a carboligation-type synthesis using human E1o from 2-OG and from 2-OA [63,98]. With 2-OG as a substrate donor and glyoxylate as an acceptor, (R)-2-hydroxy-3-oxohexanedioic acid ((R)-2-hydroxy-3-oxoadipate) (Figure 4, panel A) was detected in the CD experiments with an estimated catalytic efficiency (k cat /K m,OG ) of about 4.8% compared to the E1o-specific activity (Figure 4, panel B). With 2-OA as a substrate and glyoxylate as an acceptor substrate, (S)-2hydroxy-3-oxoheptanedioic acid (2-hydroxy-3-oxopimelic acid) could be detected (Figure 4, panel B). It is notable that, while 2-OA could serve as a substrate for human OGDHc, the human E1o efficiency toward 2-OA (k cat /K m,OA ) is about 18-19-fold lower compared to that for 2-OG in both the overall and E1-specific assay [63]. With regard to the carboligation reaction, the human 2-oxoadipate dehydrogenase (human E1a, also known as DHTKD1) in the L-lysine degradation pathway is the most promiscuous enzyme, which could successfully decarboxylate substrates with 5 (2-OG), 6 (2-OA) and 7 (2-oxopimelic acid) carbon atoms [63,65]. By analogy with the carboligation reaction of human E1o, it produced (R)-2-hydroxy-3-oxoadipate in the 2-OG: glyoxylate carboligation reaction and (S)-2-hydroxy-3-oxoheptanedioic acid in the 2-OA: glyoxylate carboligation reaction, similar predominant enantiomers as produced by human E1o. Formation of the carboligation products by human E1a represents a potential value for chemoenzymatic synthesis. Considering the broad substrate specificity of the human E1a component, it could be effectively employed in the chemoenzymatic synthesis of chiral α-hydroxy ketones, valuable synthetic intermediates. These findings expand the scope of both substrate donors and aldehyde acceptors to be employed in carboligation reactions. Different enantioselectivities in product formation could be achieved by using three enzymes: the E. coli and human E1o and human E1a.  [69,97]. The formation of products 1 and 3 by using 2-OG and 2-OA as substrates and glyoxylate as an acceptor was also demonstrated for the human E1o [63,64,69]. Panel A represent all chiral products corresponding to Table 2. Panel B represent chiral products using carboxylic acids which are structurally similar to 2-oxo acids (2-oxoadipic acid, 2-ketohexanoic acid and 2-oxo-5-hexenoic acid) and acceptors (glyoxalate, butyraldehyde and propanal).

Site-Saturation Mutagenesis Studies Clarified Catalysis of the Transthioesterification Reaction on the E. coli E2o and Corrected a Long-Misunderstood Mechanism Relevant to All E2 Components
The engineering of the E. coli E2o had two objectives: (a) to understand the catalytic mechanism of the E. coli E2o core domain and (b) to assign a role to the key residues in the mechanism. It needs to be noted that after years of multiple structural studies on the E2 catalytic domain by X-ray [46,99,100] and by cryo-EM [52][53][54]59,101,102], the catalytic mechanism of the E2 component remains elusive. Earlier, based on a comparison of the primary sequences and predicted secondary structures of the E2 enzymes with that of chloramphenicol acetyltransferase (CAT), the mechanism proposed for E2 [103] included an active site histidine as a general-base catalyst responsible for the deprotonation of the reactive thiol of CoA, analogously to that proposed for the active site His195 in CAT [104].
The CAT and the subtilisin family of serine proteases are considered as gold standards for acid-base catalysis [105]. It was predicted that E2o [99] belongs to the same class of enzymes, with the catalytic His having a similar proton-donor/acceptor role in the acyltransfer reaction [99]. Understanding the role of His375 in E. coli E2o was pivotal to gain insight into the catalytic mechanism of the acyl transfer reaction for production of acylCoA derivatives. Furthermore, with the increasing resolution of the cryo-EM structures of the E2 core domains and an attempt to understand the molecular mechanism of substrate channeling by the lipoyl domains [52], an understanding of the E2o catalytic mechanism becomes critical.
The X-ray structure of the E. coli E2o catalytic/core domain had been reported earlier ( Figure 5A) [66,99]. Briefly, the E. coli E2o core domain assembles into a 24-subunit oligomer where tightly associated trimers are arranged at the corners of the cube [99]. The substratebinding channel is~30 Å long, and it is formed by the interaction between two of the threefold related subunits. One side of the channel in the E. coli E2o core domain is exposed to solvent [99]. The catalytic residues His375 from one subunit and Thr323 from another subunit are positioned in the middle of the channel [99]. In our studies, the single-site substitution at the highly conserved His375, Asp374 and Thr323 from the E. coli E2o active center revealed that His375 and Asp374 but not Thr323 are catalytically the most important residues for succinyl transfer in both the physiological and reverse directions (see Figure 5B for steady-state kinetic data) [106]. The calculated value of the rate retardation resulting from the His375Ala substitution in E. coli E2o of 54-fold (this number represents the ratio of (k cat /K m ) wild-type /(k cat /K m ) variant in Figure 5B) was much smaller compared to thẽ 500,000 value calculated for His195Ala CAT, suggesting different functions for the highly conserved active center histidine on the two enzymes. More insight into the possible role of His375 came from SSM studies, in which the His375Trp E2o variant displayed a significant retention of the overall NADH activity (~60%) and showed similar pH dependence of the NADH activity, as well as similar binding ability for CoA and for succinyl-CoA, in comparison with unsubstituted E2o [106]. Although it is unclear why an aromatic ring substitution is tolerated at the active site histidine in multiple instances, it is safe to speculate that a possible steric effect of this histidine imidazole is to ensure intermediate tunneling through the active site [106]. Altogether, these findings question the general base role assigned to the conserved catalytic histidine of the acyl transferase class of enzymes. Two plausible mechanisms could be suggested for the transfer of an acyl group between two thiols by E. coli E2o. Lower pathway is a general acid-base mechanism, were the His375 converts the thiol of the attacking nucleophile (CoASH) to a thiolate anion (CoAS − ) ( Figure 5C, pathway B). The activated thiolate then attacks the carbonyl atom of succinyldihydrolipoyl-E2o to form a tetrahedral intermediate, which is stabilized by a hydroxyl side chain of Thr323 in accord with a mechanism suggested for CAT [103,104,107]. In the upper pathway, as an alternative to the acid-base catalysis mechanism, a direct attack by a thiolate anion (pKa~7.0 for the SHgroup) on the thiol ester carbon of the S8-succinyldihydrolipoyl-E2o was suggested with the formation of a symmetrical tetrahedral oxyanionic intermediate, where the central carbon atom is flanked by two C-S bonds with a near equal probability for cleavage ( Figure 5C, pathway A) [106]. The His375 and Asp374 residues in E. coli E2o are more likely involved in the stabilization of the oxyanionic intermediate by two hydrogen bonds. The His374Asn substitution in E. coli E2o (0.5% remaining activity) led to a further reduction in the OGDHc catalytic efficiency by~fourfold compared to the His375Ala substitution (2.1% remaining activity) ( Figure 5B). In contrast, the Asp374Gln substitution rescued the E2o activity (9.7%) compared to the Asp374Ala substitution (2.1%), suggesting that Asp374 is involved in hydrogen bond formation. Additionally, according to SSM studies, the His375Trp E2o substitution led to an OGDHc activity of about 60% compared to wild-type E2o, arguing strongly against the acid-base role of His375 [106]. This conclusion is consistent with data reported earlier from McLeish's group on the ThDP-dependent enzyme BFDC, where SSM studies on the active site His281 led to His281Phe and His281Trp variants with significant remaining activity compared to the wild-type enzyme, also arguing against an acid-base role for His281 [18]. In addition to elucidating this long-misunderstood mechanism of the transthioesterification reaction carried out by E. coli E2o, the findings are most likely applicable to all E2 enzymes and provide crucial information for further engineering of the E. coli E2o for producing a variety of acyl-CoA analogues.
A further important aspect of OGDHc assembly was revealed when recombinant E2o domains (LD and CD) were individually expressed, purified and mixed with E1o and E3 to assemble OGDHc in vitro. The complex, so assembled, displayed overall activity with 2-OG, leading to the conclusion that covalent bonding between the domains is not required for OGDH-complex assembly and activity.

Expansion of E2o Substrate Specificity and Creation of the 2-oxo Aliphatic Acid Dehydrogenase Complex with Goals of Acyl CoA Analogue Synthesis
Engineering of the E. coli E1o by single amino acid substitution led to changing its substrate specificity [68]. Thus, the His298Asp and His298Val substitutions converted the E. coli E1o to 2-oxovalerate dehydrogenase with activities comparable to those observed with 2-OG. However, when E1o variants were assembled with E2o and E3 into OGDHc, no NADH formation was detected with 2-OV as a substrate, implying discrimination at the E2o level [68]. In view of these findings, the E. coli E2o catalytic domain was next engineered to accept substrates other than succinyllysyldihydrolipoamide. Earlier, based on the alignments of known E2o, E2p and E2b enzymes, it was suggested that Gly325, Ser333 and Arg381 might contribute to the substrate specificity of the E. coli E2o core [108,109]. In the absence of the E2o core domain structure with succinyllysyldihydrolipoamide bound, the residues that may be implicated in substrate binding were identified based on molecular modeling studies [99]. According to those model studies, the carboxyl group of the S-succinyllysyldihydrolipoamide forms hydrogen bonds with the hydroxyl groups of Ser330 and/or Ser333 [99]. The His348, conserved among known E2o sequences, is located near the proposed succinyl-binding site and was predicted to be responsible for recognition of the succinyl group ( Figure 6) [66,99]. Figure 6. The succinyl-binding site of the E. coli E2o catalytic domain with succinyldihydrolipoamide modeled into it [110]. The succinyl-binding site contains residues Ser330 and Ser333 within hydrogenbonding distance of the carboxylate group of succinyldihydrolipoamide. The His348 which may be responsible for substrate recognition is shown. The modeling was performed by using UCSF Chimera Autodock program [111] and coordinates of the E. coli E2o catalytic domain deposited under the accession number 1C4T [99].
In the Farinas lab, SSM libraries were constructed for the E. coli E2o at positions Ser330, Ser333 and His348 were constructed [110]. The E2o variants with the His348Phe, His348Tyr, His348Gln and Ser333Met substitutions were identified from the screen. On reconstitution of the purified E2o proteins with substitutions identified with His298Asp-E1o and E3 into OGDHc, the formation of NADH from 2-OV could be detected in vitro (Table 3). Moreover, the formation of butyryl-CoA by the newly designed complexes could be detected by mass spectrometry [110]. Surprisingly, the identified His348 E2o variants retained the OGDHc activity toward the natural substrate 2-OG with the maximum reduction of catalytic efficiency (k cat /K m,2-OG ) by only 1.6-fold (His348Gln E2o). The E2o variants were also found to be active towards an even more hydrophobic substrate, 2-oxo-5-hexenoic acid (2-OHe), which contains a terminal alkene group that would be useful for further functionalization to increase product diversity (Figure 2, top) [110]. It needs to be noted that when assembled from unsubstituted E1o, E2o and E3 components, no NADH could be detected for OGDHc with 2-OV or 2-OHe as substrates. Related to residue His348, proposed to interact with the distal carboxylate of S-succinyllysyldihydrolipoamide [99], the OGDHc assembled with the His348Phe or His348Tyr E2o variants displayed higher activity compared to the His348Gln E2o variant with both 2-OV and 2-OHe as substrates. One might simply attribute this to the presence of a hydrophobic amino acid in that position being favorable to recognize the butyryl group of the substrate. Modeling the solvent-accessible volume for the binding channel revealed that phenylalanine and tyrosine substitutions at His348 do not alter the solvent-accessible volume, thus providing a higher catalytic efficiency versus 2-OV and 2-OHe (Table 3). With His348Gln E2o, where the histidine and glutamine have a similar size, the glutamine substitution creates a smaller solvent-accessible volume compared to unsubstituted E2o, leading to a lower catalytic efficiency versus 2-OG, 2-OV and 2-OHe (Table 3) [110]. The Ser333Met E2o substitution led to only a threefold-lower catalytic efficiency of the OGDHc with 2-OG compared to wild-type E2o and retained the highest activity with 2-OHe among the other E2o variants ( Table 3). The Met or Gln substitutions led to a decrease in the pocket volume, which possibly leads to unfavorable interactions with 2-OG. However, such an environment was favored for 2-OHe to be the substrate of choice. Thus, this study elucidates another aspect of the reactivity, where the importance of the microenvironment in the active site channel is extremely important and affects substrate recognition in E. coli OGDHc. Furthermore, the changes in the active site channel would be otherwise impossible to predict other than from extensive kinetic studies.

Conclusions and Perspectives
The major focus of the current review was to explore the power of protein engineering methods to investigate active site catalysis and further expansion of the substrate range of multienzyme complexes. This was done in view of the strategies implemented for evolution of the rudimentary case study of OGDHc. Notably, modern day enzymes have evolved by recruiting primitive singular enzymes performing sequential, divergent or convergent chemical reactions. Furthermore, under selection pressure, the complexes have achieved exquisite substrate-channeling efficiencies. This is extremely difficult to mimic inside a laboratory using traditional scaffolding/colocalization approaches [23][24][25]. Further, to broaden the substrate range of the entire complex, engineering is required at each component level, followed by assembly of the overall complex. Thus, a thorough understanding of both substrate binding and catalysis is essential in order to implement the complex for chemoenzymatic synthetic purposes. Although recent advances in machine learning and artificial intelligence can accurately predict protein structures from an evolutionary database, the actual protein-protein interactions and structure-function relationships baffle scientists. A notable example in this aspect has been demonstrated by the use of AlphaFold, an artificial-intelligence-based rapid theoretical protein prediction program [112]. The method appears remarkable in its predictive power for the structure of monomeric proteins. For the present, however, the prediction of interactions among different components in a multienzyme complex, consisting of multiple copies of each of three components, such as the 2-OGDHc, is still on hold for the near future [26]. For such multimeric assembly (~4-5 MDa) enzyme complexes, the pieces of the puzzle can be put together mainly by mutagenesis (directed or site saturation) approaches. Only then can the overall complex characterization be expressed as a summation of catalysis and substrate binding/specificities at individual component levels. This knowledge is the fundamental stepping stone of the engineering process to further move forward with its broad chemoenzymatic synthetic goals.
Over the years, the pharmaceutical, food, detergent and biofuel industries have reaped the advantages of enzyme catalysis on industrial scale applications. In cases of complex chemical reactions, the most significant advantages of biological (enzyme) catalysts include high stereo-, chemo-and regioselectivities [113]. The pharmaceutical industry has successfully exploited engineered-enzyme-mediated catalysis to synthesize complex compounds that are difficult to synthesize by traditional chemical catalysis. A notable example is the drug sitagliptin, marketed by Merck for type II diabetes treatment [114][115][116]. Acyl-CoA and its analogs are considered important intermediates in organic chemistry and drug development pipelines. Traditional organic synthetic pathways for the acylation of thiols or transthioesterification involve the use of acyl sources from carboxylic acids, acid anhydrides or acid chlorides. The reactions proceed in the presence of strong bases such as ethylamine, pyridine or DMAP for refluxing for several hours. In short, these are multistep chemical reactions, energy intensive and involving extreme reaction conditions, which can be hazardous to the environment [117]. Incidentally, the final product of the main transthioesterification reaction catalyzed by OGDHc is a CoA-linked thioester (succinyl-CoA). Further, the side-reaction of the E1o component is carboligation, which is also of immense importance in the chemistry world for the synthesis of chiral carbon compounds. This makes OGDHc a suitable candidate for one-pot multistep green chemical synthesis under much milder aqueous conditions, thus providing a sustainable and green synthetic chemical pathway.
OGDHc is a well-evolved multienzyme complex of a conserved metabolic pathway across many different species. Site saturation mutagenesis in the active sites of E1o and E2o effectively assigned roles to the pivotal residues contributing to catalysis. It is important to recognize that such assignments using structural homology modelling would be difficult to predict. The principal achievements in this approach that differentiate between our research and work by others are summarized below.
(a) The E1o component of the E. coli OGDHc was engineered to accept substrates lacking the 5-carboxylate group by subjecting His260 and His298 in the active center to site saturation mutagenesis. The results rule out an acid-base or hydrogen-bonding role for His298; however, they confirmed a hydrogen-bonding role for His260. It was suggested that the E2o component is a 'gatekeeper' for 2-OG channeling through the pathway. (b) It was demonstrated that the E. coli E1o can accept 2-oxovalerate and 2-oxoisovalerate in addition to its natural substrate 2-oxoglutarate in the carboligation reaction and that glyoxylate, ethyl glyoxylate and methylglyoxal can serve as aldehyde acceptors, producing acetoin-like compounds with good enantioselectivity. The novel products formed by the carboligation reaction include chiral 2-hydroxy-1,3-diketones and 2-hydroxy-β-ketoesters. The E. coli E1o offers a good starting point for protein engineering and optimization to synthesize stable chiral intermediates for fine chemical synthesis. (c) The fundamental mechanism of the transthioesterification reaction carried out by the E2o component of the E. coli OGDHc was elucidated, leading to the identification of Asp374 and His375 residues in the E2o catalytic center that are important for succinyl transfer in both the physiological and the reverse directions. The evidence provided ruled out a role of acid-base catalysis by His375. It was suggested that His375 and Asp 374 are involved in the stabilization of the tetrahedral oxyanionic intermediate through hydrogen bond donation. It was concluded that a rate-limiting step on E. coli OGDHc is succinyl transfer to CoA, thus providing crucial information for producing a variety of acyl-CoA analogues. (d) 6.4. The engineering of two consecutive components, E1o and E2o, was accomplished.
The promiscuous variant components assembled into OGDHc were able to catalyze reactions with both natural and unnatural substrates, leaving significant room for optimization. Similar to natural evolution, the new pathway could be engineered to suit particular user-defined goals. These designed routes can function together with the existing pathways and could be utilized for chemical and synthetic purposes [118].
Cellular organization and function have evolved over millions of years under natural selection pressure. The 2-oxoacid-dehydrogenase complexes are the molecular machinery that have been fine tuned to carry out cellular reactions with maximum efficiency. Catalytic promiscuity has been the likely driving force for development of specialized enzymes such as human E1a (2-oxoadipate-dependent component), catalyzing novel reactions compared to human E1o [119]. Furthermore, the approach to evolve a multienzyme complex takes into consideration the recruitment of sequential enzymes into self-assembled multicomponent supramolecular structures and performance of multistep metabolic reactions effortlessly. The knowledge of this organization by recent cryo-EM studies is a major effort that would provide molecular biologists with a platform from which structure-function relationships can be probed and fabricated towards user-defined interests [120,121]. Future extensions of the SSM methods here explored could focus on the demonstrated promiscuity of the (a) Human E1a component of a multienzyme complex related to OGDHc in the L-lysine degradation pathway, as it accepts, in addition to 2-oxoadipate, 2-oxoglutarate and 2oxopimelate as substrates, all capable of participating in ThDP-dependent carboligation reactions as 'acceptor' partners to produce novel chiral α-hydroxyketones; and (b) Human E2o active center with its ability to synthesize, in addition to its physiological product succinyl-CoA, glutaryl-CoA and adipoyl-CoA.
In the next step, directed evolution strategies could be implemented to improve the new catalytic activities and yield of products [122,123]. Traditionally, metabolic pathways are not considered fertile grounds for biocatalyst development [124]. However, the potential of the development of orthogonal novel pathways from existing metabolic pathways could be an exciting avenue in the world of green synthesis. In the post-genomic era and, with the help of protein engineering, the potential utility of such complexes to a chemist is invaluable.

Conflicts of Interest:
The authors declare no conflict of interest with the work reviewed.