The Birth of Genomic Enzymology: Discovery of the Mechanistically Diverse Enolase Superfamily

Abstract

Enzymes are categorized into superfamilies by sequence, structural, and mechanistic similarities. The evolutionary implications can be profound. Up until the mid-1990s, the approach was fragmented largely due to limited sequence and structural data. However, in 1996, a paper by Babbitt et al. was published in Biochemistry that demonstrated the potential power of mechanistically diverse superfamilies to identify common ancestry, predict function, and in some cases, specificity. This perspective describes the findings of the original work and reviews the current understanding of structure and mechanism in the founding family members. The outcomes of the genomic enzymology approach have reached far beyond the functional assignment of members of the enolase superfamily, inspiring the study of superfamilies, adoption of sequence-similarity networks and genome context and yielding fundamental insights into enzyme evolution.

Graphical Abstract

INTRODUCTION

In 1996, a bioinformatics, mechanistic, and structural team led by Patricia Babbitt and John Gerlt that included George Reed, Ivan Rayment, Dagmar Ringe, George Kenyon, and others published a paper in Biochemistry titled “The Enolase Superfamily: A General Strategy for Enzyme-Catalyzed Abstraction of the α-Protons of Carboxylic Acids”¹. The paper reported a superfamily of enzymes related by their ability to catalyze the abstraction of the α-proton of a carboxylic acid to afford an enolic intermediate (later refined to be a carboxylate and enolate ion). The reaction is mediated by a common active-site scaffold that had been modified in the course of evolution to allow partitioning of the intermediate to different products. The authors argued that a comparative analysis in a superfamily context could be used to determine mechanism, assign function, correct misassigned function, and predict possible reactions for members with unknown functions. The approach provided the framework for tackling and making sense of the tsunami of sequences that was about to flood the databases. The work also supported the idea that enzymes evolve by recruiting a scaffold that can carry out the required chemistry, which was a direct challenge to the existing dogma that new catalytic functions arise from a common ability to bind a substrate. Over the past 25 years, the superfamily approach has changed the way we think about and study enzymes. Thus, this paper ushered in the era of genomic enzymology.

Prior to this work, mechanistic enzymologists explored an enzyme of interest, having intellectual and medical relevance, in great detail, and teased out how the individual active-site residues contributed to the chemical and catalytic mechanism of the overall reaction. The enzyme of interest could be part of a biosynthetic (e.g., thymidylate synthase)² or catabolic pathway (e.g., glyceraldehyde 3-phosphate dehydrogenase)³ or in a group of enzymes (serine, cysteine, aspartic, and metalloproteases)⁴. The strategy was reasonably straightforward and used kinetic, mechanistic, site-directed site mutagenesis, and X-ray crystallography to establish structure-function relationships for the reaction. Although detailed mechanistic studies most certainly contributed to the development of the superfamily concept, the continued study of each individual enzyme and assignment of function would be impractical and, in fact impossible, in the face of a vast number of sequences and structures.

The paradigm shift resulted from the convergence of three events. First, the mechanisms of mandelate racemase (MR) and muconate-lactonizing enzyme (MLE) were being studied in great detail, with many structure-function relationships having been established. Research groups led by George Kenyon, John Kozarich, John Gerlt, Dagmar Ringe, and Greg Petsko, working in close collaboration, contributed to this overall body of work^5–8. Second, the mechanism of enolase was under scrutiny because (among other things) it was a model system for the facile enzymatic removal of a non-acidic proton from carbon⁹. The role of the metal ion in the stabilization of the carbanion was of particular interest and was being studied by research teams led by George Reed and Ivan Rayment¹⁰. Finally, the increase in computational power enhancing speed of X-ray crystal structure determination and concurrent increase in community gene-sequencing efforts gave the needed information. The overlay of the three representative structures of the MR, MLE and enolase subgroups validated the sequence alignment of these representative members with one another and with the more divergent members of the superfamily.

In the original report, the enolase superfamily consisted of 16 members that were divided into three subgroups (MR, MLE, and enolase) based on the identities of the active-site bases. All use an enzyme-liganded metal ion to stabilize the enolate. At last count, the superfamily has grown to more than 48,000 members, divided into seven subgroups based on the identities of the acid-base catalysts and metal ion ligands¹¹. The active sites are located at the C-terminal ends of β-strands in a (β/α)₇β-barrel domain, and indeed, common active-site location is a hallmark of mechanistically diverse superfamilies. Almost all enolase superfamily members have conserved acidic ligands (Asp or Glu) to bind the essential Mg²⁺ (or Mn²⁺) ion cofactor at the C-terminal ends of the third, fourth, and fifth β-strands. The acid/base catalysts are located at the C-terminal ends of the second, third, fifth, sixth, and or seventh β-strands (as discussed below).

This perspective will discuss the members of the three original subgroups and their roles in developing the comparative approach along with an update of their mechanisms and function. Characterization of the original subgroup members led to the determination of mechanisms and functional assignment for enzymes in newly sequenced genomes. This, in turn, led to the discovery of new facets of bacterial metabolism, a better understanding of divergent evolution including the structural bases for the evolution of new activities and the driving role of chemistry in it, and a model for enzyme design and redesign.

The spark for the original work^{12, 13} came from the serendipitous observation that the three-dimensional structures of MR and MLE were “remarkably” superimposable (Figure 1a). The active-site carboxylate groups that bind the essential metal ion (usually Mg²⁺, but Mn²⁺, Cu²⁺ and Ni²⁺ will also support catalysis¹⁴) and the general-base catalysts that mediate proton transfer from the α-position of the carboxylate group are highly conserved. Once the metal-ion ligands were aligned, the candidate general bases responsible for α-proton abstraction could be identified (Figure 1b). MR and MLE along with fourteen additional enzymes that were identified initially are related by their ability to abstract the α-proton from a carboxylate group thus initiating the reaction. The abstraction of the α-proton of a carboxylate substrate is an exceedingly difficult reaction to do in the active site of an enzyme because of the high pK_a (~29 for mandelate)^{15, 16}, Thus, a major insight was that the mechanism for stabilizing the enolate intermediate to catalyze this abstraction is retained by evolution throughout the superfamily and so the relative positions of the ligands to the essential metal cofactor within the active site is retained.

Figure 1. The (β/α)₇β-barrel fold and active site of the enolase superfamily.

A. Ribbon diagram of enolase (PDB 1ONE) with metal-binding residues (gold stick) and metal (Mg²⁺; blue tint sphere) which are conserved and identifying features of the enolase superfamily. Catalytic residues in enolase at conserved positions in the active site (grey stick). Strands of the (β/α)₇β-barrel fold are highlighted in blue with strands numbered S1-S8 and N-terminal excursion from the barrel made transparent B. Comparison of the active-site residues in MR (blue; PDB 1MNS), MLE (pink; PDB 1BKH) and enolase (grey) with conserved metal position shown for MR (grey sphere).

Superimposition of the structures showed that MR and MLE are homologous even though their sequence identity from pairwise alignment are less than 26%¹³. A few years later, yeast enolase was reported to share the same structure as well as some of the essential active-site residues and became the third member of the superfamily^{1, 17}. Enolase became the namesake of the superfamily because of its central role in primary metabolism. In the original three subgroups, members of the MR subgroup were identified as having a His-Asp dyad and often a lysine residue, those in the MLE I subgroup utilize two lysine residues, and those in the enolase subgroup have one lysine residue (Figure 1b). With the discovery and exploration of new members, it soon became apparent that these “canonical” acid/base catalysts were not sufficient to carry out the range of reactions in the superfamily, and new ones were identified (as discussed below)¹⁸.

Overview of MR, MLE, and Enolase

Comparison of the three representative members shows the diversity of the roles of the catalytic residues among the subgroups. Although the ligands to the essential metal ion are the same in position and identity among MR, MLE and enolase, no other catalytic residue is conserved in identity. Residues critical to catalysis are often found in the same position in the protein scaffold but differ in identity. Notably, even when there is conservation of the position of a catalytic moiety, the residues may originate from a different secondary structural element in the protein fold.

MR interconverts the R- and S-isomers of mandelate and allows S-mandelate to feed into the β-ketoadipate pathway in Pseudomonas putida (Scheme 1a)⁸. MLE is an enzyme in the β-ketoadipate pathway that catalyzes the interconversion of cis,cis-muconate and muconolactone involving reversible protonation/deprotonation (an intramolecular addition/elimination reaction); Scheme 1b)^{19, 20}. The entire suite of enzymes allows the organism to use both isomers of mandelate as sole sources of carbon.

Scheme 1.

Substrates, Intermediates, and Products for MR, MLE I, and Enolase

A combination of studies with competitive and irreversible inhibitors, site-directed mutagenesis, and crystallography identified the core elements of the MR mechanism, which uses a two-base mechanism⁸. Lys166 is the general-base catalyst that abstracts the α-proton from the S-isomer, becoming known as the S-specific base, and Lys164 is the general acid that neutralizes the anionic charge of the carboxylate substrate⁵. These two lysines are part of a KXK motif found in sequence alignments of the MR subfamily. His297 (part of a His-Asp dyad) is the general base that abstracts the α-proton from the R-isomer (the R-specific base) and Glu317 is the general-acid catalyst that donates a proton to the carboxylate group of substrate (as the α-proton is abstracted)⁷. The essential Mg²⁺ cofactor is coordinated by Asp195, Glu221, and Glu247, as well as one carboxylate oxygen and the α-hydroxy of the bound inhibitor (S-atrolactate). The ε-ammonium group of Lys164 forms a hydrogen bond to the same carboxylate group.

In MLE, Lys167, Lys169, (in the KXK motif)^{21, 22}, and Glu327 are functional equivalents to Lys164, Lys166, and Glu317 in MR. Accordingly, Lys167 stabilizes one oxygen of the negatively charged carboxylate group, Lys169 is the general-base catalyst, and Glu327 stabilizes the second oxygen of the negatively charged carboxylate group^{1, 20}. Lys273 was definitively identified as the acid catalyst several years later when a liganded structure became available.²³Although both MR and MLE catalyze the abstraction of an α-proton to yield enolate intermediates, the intermediate has different fates: protonation in MR and vinylogous elimination in MLE I (which catalyzes ring-opening of muconolactone to cis,cis-muconate). A homologous residue to His297 (in MR) is not present in MLE I, but there are counterparts to the residues that bind the metal-ion cofactor (Asp198, Glu224, and Asp249).

Enolase is also a Mg²⁺-dependent metalloenzyme that plays a central role in glycolysis and gluconeogenesis⁹. It interconverts 2-phosphoglycerate (2-PGA) and phosphoenolpyruvate (PEP) by the elimination or addition of a water molecule (Scheme 1c). Mutagenic analysis indicates that Lys345 (equivalent to His297 in MR and Lys273 in MLE I) is the general base that abstracts the α-proton and Glu211 is the general acid that facilitates the departure of the β-hydroxide leaving group. Although Glu211 has no equivalent group in MR or MLE I, His373 homologous to the general base His297 in MR, positions the γ-carboxylate of Glu211 for catalysis. The high-resolution X-ray crystallographic structure of yeast enolase complexed with substrate, 2-PGA^{17, 24} shows two Mg²⁺ ions, where the metal that is conserved in position all superfamily members, is bound by Asp246, Glu295, and Asp320, homologous residues to the metal-ion ligands in MR and MLE I. This Mg²⁺ also forms a coordinate bond with both carboxylate oxygens of the 2-PGA substrate. One of the carboxylate oxygens also ligands the second Mg²⁺ ion. Thus, the interactions of the substrate carboxylate group with the active site differ in considerable detail from those of MR and MLE.

Despite the limited numbers of sequences and three-dimensional structures available, the group was able to identify enolase superfamily members by the presence of the conserved metal-ion ligands and construct a primary sequence alignment¹. The alignment included sequences for five homologs with defined functions: galactonate dehydratase (GalD), glucarate dehydratase (GlucD), β-methylaspartate ammonia-lyase (β-MAL), N-acylamino acid racemase (NAAAR), and o-succinylbenzoate synthase (OSBS). Four open reading frames (orfs) with no functional assignments were also identified as enolase superfamily members: RspA, Spa2, reverse thymidylate synthase (rTS), and SynORF. Two homologs with known function were identified: carboxyphosphonoenolpyruvate synthase (CPEPS), and a DNA-binding protein (MBP1, c-Myc promoter-binding protein). The latter is a homolog of enolase, but no catalytic function is associated with the protein. It is not uncommon for non-enzymatic function to evolve in enzyme families, with loss of catalytic residues (eg. binding of phosphatidic acid without phosphohydrolase activity in the haloalkanoate dehalogenase superfamily²⁵ or binding of NAD metabolites in the Nudix hydrolase superfamily²⁶). Although it is also true that binding proteins may have diverse ligands that are unlike the substrates of homologous enzyme²⁷.

Based on this rudimentary sequence alignment, the authors determined the common catalytic strategies for multiple reactions in the superfamily along with phylogenetic relationships among the proteins. The relationships were inferred by clustering the sequences into the MR, MLE, and enolase subgroups. In all cases, the alignment is sufficient to predict/assign the general bases as well as the ligands for metal-ion binding. It is not a sufficient basis to predict/assign the general acids. Accordingly, the identities and properties of the putative general bases of the newly identified members were discussed.

MR Subgroup

The MR subgroup initially consisted of those members with a His297 equivalent (in a His-Asp dyad) that would function as a general-base catalyst. These members included galactonate dehydratase (GalD), glucarate dehydratase (GlucD), and three of the reading frames of unknown function: RspA, Spa2, and rTS. Of these, a function has not yet been reported for Spa2.

GalD catalyzes an anti-elimination of water from D-galactonate to afford 2-keto-3-deoxy-(D)-galactonate (Scheme 2). Alignment of E. coli GalD with MR shows the conserved metal ligands (Asp183, Glu209, Glu235) as well as His285 (in a dyad with Asp258), which functions as the (R)-specific base at C-2 of substrate¹². In MR, His297 is the (R)-specific base, and the agreement of the stereochemistry supports the assignment (Figure 2a). Abstraction of the α-proton (at C2) by His285 initiates the vinylogous β-elimination of water with the assistance of an unknown general acid catalyst to result in an anti-dehydration. The KXK motif in MR is replaced by a KXN motif in GalD where Asn146 is a non-functional replacement for Lys166 in MR. A few years later, a “new” general acid catalyst (i.e., not the acidic KXK lysine), His185, was identified based on crystallographic data (proximity to OH leaving group), mutagenesis (substantially decreased k_cat and k_cat/K_m values for H185N and H185Q variants), and chemical analysis (the comparable elimination of fluoride from a fluoro-substituted substrate) evidence²⁸. [Tautomerization, after loss of H₂O, could be catalyzed by His185 and His285 acting as the general base and acid, respectively].

Scheme 2.

The GalD-catalyzed Reaction

Figure 2. Comparative active sites in the MR subgroup.

Comparison of the active-site residues in MR (blue; PDB 1MNS), and A. GalD (gold) with conserved metal position shown for GalD (gold sphere) and D-lyxose (gold-stick) and B. GlcD (wheat; PDB 1ECQ) with metal position shown for GlcD (grey sphere) and 4-deoxyglucarate (wheat stick). C. Overlay of GalD (gold) with RspA (yellow; PDB 4IL2) with metal position (gold sphere) and D-lyxose (gold stick) shown for GalD.

The mechanism of glucarate dehydratase (GlucD) from P. putida (and later E. coli) is more complicated because both (S)- and (R)-specific bases are present¹ (See Scheme 3). Alignment of the sequences (E. coli numbering) shows the conserved KXK motif where Lys207 is the (S)-specific base (equivalent to Lys166 in MR) and the His-Asp dyad where His339 is the (R)-specific base (equivalent to His297 in MR).¹ GlucD catalyzes the syn-dehydration of both D-glucarate and L-idarate to form 5-keto-4-deoxy-(D)-glucarate as well as the epimerization of both at C-5 (in competition with dehydration) (Scheme 3). D-glucarate and L-idarate are epimers of one another at C-5 where D-glucarate is 5-(S) and L-idarate is 5-(R). His339 abstracts the proton at C-5 of D-glucarate and Lys207 abstracts the proton at C-5 of L-idarate (Figure 2b).

Scheme 3.

The GlucD-catalyzed Reactions

This analysis gave a general picture of the mechanism, but a more complete picture emerged in 2001 after completion of crystallographic and mutagenesis studies on the E. coli enzyme. The new picture suggested that Lys207 of the KXK motif would be positioned to abstract the α-proton from C-5 of L-idarate and His339 (in the His339-Asp313 dyad) would be in position to abstract the α-proton from C-5 of D-glucarate (Scheme 3). His339 is proximal to the C4 OH leaving group, such that it may act as the general-acid catalyst. The most reasonable scenario is one in which Asn341 is required to position His339, such that it functions as a general-base and general-acid catalyst^{29, 30}. [Again, tautomerization, after loss of H₂O, could be catalyzed by His339 and Lys207 for D-glucarate and L-idarate, respectively].

Of the original orfs identified, three fell into the MR subgroup. RspA, Spa2, and rTS were identified based on the metal-ion ligands homologous to those found in MR and a histidine equivalent to the (R)-specific His297 of MR. RspA is encoded by a gene that is involved in the onset of the stationary phase when metabolites are depleted (also known as starvation-sensing protein RspA). Spa2 is encoded by an orf in a region of the Streptomyces ambofaciens that is associated with chromosomal instability^{1, 31}. They share 65% sequence identity, suggesting that they might catalyze the same reaction. The protein rTS has been shown to accumulate in tumor cells and to be associated with methotrexate resistance. It was later identified as a reverse thymidylate synthase with three isoforms: α, β, and γ (vide infra)³². There was less clarity about homologous residues to Lys164 and Lys166 (in the KXK sequence motif) in RspA and Spa2, but there is a homologous KXK motif in rTS. Similar to GalD, both RspA and Spa2 have a His185 homologue (i.e., the acid catalyst) which suggested that they catalyze a dehydration reaction²⁸ (though a substrate for Spa2 has not been identified) (Figure 2c).

The specific chemical reactions and substrates for RspA were not known in 1996. However, the accumulation of sequence data over the next several years identified orthologues of the RspA gene in Novoshingobium aromaticivorans and E. coli. In N. aromaticivorans, the RspA orthologue is in a cluster of genes encoding the enzymes that degrade D-glucuronate and D-galacturonate³³. In E. coli, the same catabolic pathways are found where distinct enzymes catalyze the dehydration of D-mannonate (from D-glucuronate) and the 3-epimer, D-altronate (from D-galacturonate) to yield the same product, 2-keto-3-deoxy-(D)-gluconate. In N. aromaticivorans, the enzymes producing this pair of sugars have promiscuous substrate specificities and the dehydratase genes are absent, but the RspA orthologue is present in the gene cluster. Based on these observations, it was concluded that RspA is a bifunctional dehydratase (AltD/ManD) that processes both D-altronate (AltD) and D-mannonate (ManD) to 2-keto-3-deoxy-(D)-gluconate (Scheme 4). Sequence analysis of RspA highlighted the histidine of the MR His-Asp dyad as well as two conserved histidine residues. One histidine (His₁) acts as the base that initiates both dehydration reactions, whereas the other histidine is the general-acid catalyst that facilitates the elimination of the 3-OH group from D-mannonate (His₂). The histidine of the MR His-Asp dyad would provide the general acid catalyst that facilitates the elimination of the leaving group from D-altronate³³. However, to date it not yet known if these activities are biologically relevant or are promiscuous activities of the enzyme.

Scheme 4.

Reactions catalyzed by RspA (Alt D/Man D)

rTS turned into a much more complex story. In humans, rTSβ is the medically relevant isoform because the overexpression is correlated with methotrexate and 5-fluorouracil resistance. There was some evidence suggesting that human rTSβ and rTSγ share similar enzymatic functions, such that the identification of activity in bacterial orthologs might suggest functions for rTSβ and rTSγ. rTSβ is one member of a group of orthologous proteins that includes the bacterial enzyme, L-fuconate dehydratase (FucD) (Scheme 5a).³⁴ rTSγ has significant FucD activity (k_cat/K_m ~ 2.5 × 10³ M⁻¹ s⁻¹), which is 10-fold less than the bona fide bacterial enzyme from Xanthomonas campestris. The β isoform did not exhibit any FucD activity. Based on these observations, it was concluded that the mechanism of resistance does not involve catalysis, but perhaps interactions with thymidylate synthase and dihydrofolate reductase.³² The subsequent crystal structure confirmed membership in the MR subgroup and showed the conservation of the metal binding residues and catalytic groups (Figure 3a). Accordingly, the KXK motif (Lys220-Val221-Lys222), metal binding ligands (Asp250, Glu276, Glu305), and His-Asp dyad (His355-Asp328) are all present. Lys222 is responsible for the abstraction of the α-proton and His-Asp dyad functions at the general acid catalyst, specifically His355³².

Scheme 5.

Reactions catalyzed by FucD and ManD

Figure 3. Conservative and diverged active sites in the MR subgroup.

A. Comparison of the conserved active-site residues in rTSγ (magenta; PDB 4A35) and FucD (PDB 2HXT; brown) with conserved metal position (silver sphere) and D-erythronohydroxamate (brown-stick) shown for FucD. Residues labelled according to the sequences in Babbitt et al.¹ B. The active sites of MR (blue; PDB 1MNS) and ManD (purple; PDB 2QJJ) with metal position shown for ManD (grey sphere) and 2-keto-3-deoxy-D-gluconate (yellow stick) from PDB 2QJN. Note that the coordinates of ManD were used from 2QJJ as the electron density was resolved for the long loop from β-strand 2 bearing Tyr159; the position of 2-keto-3-deoxy-(D)-gluconate was obtained by superposition of ManD from PDB 2QJN.

A new catalytic strategy in the MR subgroup is represented by D-mannonate dehydratase (ManD) (Scheme 5b)³⁵. ManD catalyzes the syn-dehydration of L-mannonate, following by ketonization of the enolate intermediate to afford the product, 2-keto-3-deoxy-(D)-mannoate. It retains the ligands that bind Mg²⁺ (Asp210, Glu236, and Glu262), but lacks the Lys acid/base catalyst and has only the His312 of the His-Asp dyad. A Tyr159-Arg147 dyad is positioned to abstract the α-proton and Tyr159 and His212 facilitate the dehydration and subsequent ketonization. The retention of the metal-binding ligands confirmed the centrality of the metal cofactor in the mechanism for stabilization of the enolate anion intermediate in the enolase superfamily³⁵(Figure 3b). However, ManD represents extreme divergence in the enolase superfamily, with some catalytic residues in the canonical positions in the scaffold, but others contributed by new secondary structural elements (eg. Tyr159 resides on a long loop from beta strand 2 and His 212 on the end of beta strand 3). With the structural exploration of ManD and other divergent members of the family, it is recognized that the only strictly conserved residues are the divalent cation ligands at the ends of the third, fourth, and fifth strands of the (β/α)₇β-barrel.

Over the years, many more MR subgroup members have been identified (48,000 at present) and assigned functions, with the majority acting as acid sugar dehydratases³⁶. In all, the structural strategy for stabilization of the enolate anion (generated by abstraction of the α-proton) is present, but the strategies for generating the enolate and directing it to product are not. The “new” catalytic strategies partitioned the MR subgroup into at least three additional subgroups. Almost all MR subgroup members have a His-Asp dyad (His at the end of the seventh β-strand and Asp at the end of the sixth β-strand), if present, along with an acid/base (Lys, Arg, His, Tyr) at the end of the second β-strand.

MLE Subgroup

The MLE I subgroup initially consisted of MLE II, N-acylamino acid racemase (NAAAR), o-succinylbenzoate synthase (OSBS), a sequence of unknown function designated SynORF, and β-methylaspartate ammonia-lyase (β-MAL)¹. NAAAR and SynORF were determined later to fall into the OSBS subgroup³³. All members in the original subgroup bear a KX(K/R) motif, but no member has the His-Asp dyad that in the MR subgroup is responsible for the abstraction of a proton from the (R)-substrate of mandelate. Instead, MLE subgroup members have Gly/Ser (in place of His297) and exclusively lysine (in place of Asp270) (Figure 1b).

MLE II functions as a chloromuconate lactonizing enzyme (also referred to as chloromuconate cycloisomerase) (Scheme 6a). In the original sequence alignment, MLE II is shown to possess the same key residues as MLE I¹ (metal ligands Asp194, Glu220, and Asp245, as well as Lys165 and Lys269). However, unlike MLE I, MLE II performs two reactions: the cycloisomerization of chloromuconates, notably, 2-chloro-cis,cis-muconate, and the subsequent dehalogenation to yield a dienelactone (Figure 4a)^{37, 38}. MLE I does not catalyze dehalogenation. The mechanism for lactonization is the same for MLE I and MLE II, but the latter shows increased specificity for chloromuconates (k_cat/K_m of 3.3 × 10⁵ M⁻¹ s⁻¹ versus that of MLE I 6.4 × 10⁴ M⁻¹ s⁻¹ using 2-chloro-cis,cis-muconate)³⁸ along with the ability to carry out the elimination of chloride. The structural features responsible for the dehalogenation step have been the object of much study and speculation^37–39. The general consensus is that the same general acid/base (Lys269) is involved in the lactonization and elimination of chloride^{38, 39}.

Scheme 6.

The MLE II- and NAAAR-catalyzed Reactions

Figure 4. Conservative and diverged active sites in the MLE subgroup.

A. Comparison of the conserved active-site residues in MLEI (pink; PDB 1BKH) and A. MLEII (PDB 2CHR; green) with Mn²⁺ metal position shown for MLEII ((silver sphere); B. OSBS (purple; PDB 1SJB) with o-succinylbenzoic acid and Mg²⁺ shown as silver sphere and C. Beta-MAL (grey; PDB 1KKR) with (2S,3S)-3-methylaspartic acid. In B, C, and D and Mg²⁺ is shown as silver sphere.

NAAAR was categorized as a MLE I subgroup member based on the identities of the active site bases, Lys163 and Lys263, which are the (S)-specific base and (R)-specific base, respectively (Scheme 6b). Lys163 aligns with Lys169 (in the KXK) of MLE I and Lys263 aligns with Lys273 in MLE I (although Lys273 is not thought to be directly involved in α-proton abstraction/delivery in MLE I)¹. Thus, Lys263 is the functional equivalent of His297 (in MR) (Figure 4a).

NAAAR was first identified in a screening program as an activity that might be useful for the commercial (and environmentally friendly) production of chiral amino acids^{40, 41}. The “best” substrate, N-acetylmethionine, has a k_cat/K_m value of 370–590 M⁻¹s⁻¹ (depending on the bacterial source).^{33, 42}. The physiological role for this enzyme had always been questionable for two reasons: there was no apparent metabolic step that required the racemization of an N-acylamino acid and the very poor k_cat/K_m value for the “best” substrate. Indeed, NAAAR was later found to function as a highly efficient o-succinylbenzoate synthase (OSBS) with a k_cat/K_m value of ~ 10⁵ M⁻¹s^{−1 42}.

Indeed, a bona fide OSBS from E. coli along with Synechocystis sp SynORF were also categorized as MLE subgroup members in the 1996 paper¹. OSBS catalyzes the β-elimination of water from the substrate, 2-succinyl-6-hydroxy-2,4-cyclohexadiene-1carboxylate (SHCHC), to generate o-succinylbenzoate (OSB) (Scheme 7)(Figure 4b)⁴². The enzyme is in a pathway for the biosynthesis of menaquinone, which is an essential cofactor for anaerobic growth. The OSBS-catalyzed reaction is initiated by the abstraction of an α-proton using Lys163 (in the KXK motif; Amycolatopsis sp. numbering) or Lys263, homologs of Lys169 and Lys273 in MLE, respectively^{42, 43}. The identity of the general-base catalyst could not be established unequivocally when the family was initially classified because the structure of the substrate and orientation of the α-proton relative to Lys163/Lys263 was not known. Likewise, SynORF was designated as most similar to OSBS, but no function was assigned, and the chemistry was not predicted. A few years later, a preponderance of evidence implicated SynORF as an OSBS⁴².

Scheme 7.

The OSBS-catalyzed Reaction

These functional annotations came together with the increased availability of sequences and structures^42–44 coupled to the insights from the superfamily analysis. The OSBS reaction is initiated by Lys163 (in the KXK motif) acting as the general-base catalyst (Amycolatopsis sp. numbering system as above). Abstraction of the proton generates a transient Mg²-stabilized enediolate anion intermediate that collapses to product with the assistance of the conjugate acid of Lys163. Lys263 is in close proximity to the substrate aromatic ring and may assist in electrostatic stabilization of the enolate-anion intermediate through a π-cation interaction.⁴²

The NAAAR activity of Amycolatopsis OSBS can be rationalized because the N-acetyl-(R)-or (S)-methionine substrate binds in the same pocket as the substrate for the OSBS reaction, and is positioned between Lys163 and Lys263 so that the racemization can occur, although at a modest rate⁴³. Modifications of the N-acylamino acid substrate (i.e., from N-acetyl to N-succinyl to mimic the succinyl side chain of SHCHC and replacing methionine with phenylglycine to yield N-succinyl-L-phenylglycine) so that it more closely resembles the OSBS substrate SHCHC, gave kinetic parameters for the NAAAR activity of the promiscuous OSBS enzyme of k_cat/K_m = 1.9 × 10⁵ M⁻¹ s⁻¹ (Scheme 8). This value is close to that for racemization of SHCHC by NAAAR (k_cat/K_m = 2.5 × 10⁵ M⁻¹ s^{−1 43}. In addition, X-ray crystal structures confirmed that these various N-acylamino acids substrates and the aromatic ring of the OSB product occupy the same hydrophobic pocket⁴⁴.

Scheme 8.

NAAAR-catalyzed racemization of M-acylamino acids

The observation that MLE subgroup members could catalyze 1,1 proton transfers (i.e., the reaction catalyzed by NAAAR) prompted a search for members that catalyzed a 1,1 proton transfer in a biologically-relevant pathway³³. This search resulted in the discovery of the L-Ala-D/L-Glu epimerases (AE epimerase)⁴⁵. The proposed function (based on genome context⁴⁶) of the AE epimerases is to break down the murein (peptidoglycan) peptide and recycle its components⁴⁵. The kinetic parameters are consistent with the assignment of function where the L-Ala-D-Glu peptide is the “best” substrate (k_cat/K_m = 4.7 – 7.7 × 10⁴ M⁻¹ s⁻¹ depending on the bacterial species).

As part of an effort to find a route for the evolution of new activities in the enolase superfamily, single-site mutations (the easier evolutionary route as opposed to multiple simultaneous changes) were explored that introduced OSBS activity into L-Ala-D/L-Glu (AE) epimerase and MLE II^{45, 47}. These enzymes were chosen as scaffolds because neither one catalyzes the OSBS reaction. By rational design (AE epimerase) and directed evolution (MLE-II), single-site mutations were identified that introduced significant OSBS activity. Of note, the changes made did not simply introduce the corresponding OSBS catalytic residues into these active sites. For a D297G variant of AE epimerase and the E323G variant of MLEII the increase over the non-enzymatic activity was 10⁶ and 10⁹-fold respectively. By comparison, the rate enhancement of native OSBS is 1.8 × 10¹¹ fold over the uncatalyzed rate⁴⁸. The results are all the more remarkable because they changed substrate specificity and chemical mechanism, as opposed to contemporary efforts in other enzyme families that changed specificity, but not mechanism.^{33, 45}

The final member of the original MLE I subgroup is β-methylaspartate ammonia-lyase (β-MAL), which catalyzes the β-elimination of ammonia from L-threo-(2S,3S)-3-methylaspartate (instead of an oxygen leaving group) to give mesconate¹ (Scheme 9). There is long-standing interest in β-MAL because of its potential application in the asymmetric synthesis of aspartic acid derivatives, which have utility as biological tools and chiral building blocks for pharmaceuticals⁴⁹. β-MAL uses Lys331 (as the (S)-specific base), a homolog of Lys273 in MLE, but has no homolog of the active-site base (Lys169 in MLE) or of the diagnostic KXK motif. It was noted that there is a RDR and KLR motif nearby and that the Arg might function homologously, based on mutagenesis work with MR^{1, 18}. More recent structural and mutagenic analysis implicates His194 as the (R)-specific base and identified the metal-ion binding ligands as Asp238, Glu273, and Asp307^{50, 51} (Figure 4c). β-MAL does not have an obvious acid catalyst and likely does not need one because the ammonia group does not require protonation for elimination¹⁸. As such, β-MAL became the founder of a new subgroup within the MLE subgroup¹¹.

Scheme 9.

The MAL-catalyzed Reaction

In summary, the original MLE subgroup members catalyzed three transformations: cycloisomerization (intermolecular addition/elimination) in the MLE I- and MLE II-catalyzed reactions; β-elimination (dehydration or deamination) in the OSBS- or β-MAL-catalyzed reactions, respectively; and 1,1 proton transfers in the NAAAR- and (now) L-Ala-D/L-Glu epimerase-catalyzed reactions⁵². Members carry out these reactions by the presence of two conserved lysine residues at the end of the second and sixth β-strands, where the lysine at the second β-strand initiates the reaction by abstraction of the α-proton of a carboxylate anion. In some reactions, a general acid catalyst is not required, and in these cases, the lysine at the end of the sixth β-strand stabilizes the enediolate intermediate.

Enolase Subgroup

The active site of enolase contains a single general base (Lys345, Saccharomyces cerevisiae numbering) that initiates catalysis by abstraction of the α-proton from C2 of 2-phospho-D-glycerate. His373 (which aligns with His297 in MR) positions the γ-carboxylate group of Glu211, the proposed general-acid catalyst (Figure 5) (Scheme 1c).^{1, 9} On the basis of sequence similarity and conservation of residues associated with enolase activity, carboxyphosphonoenolpyruvate synthase (CPEPS) and cMyc promoter binding protein were assigned to the enolase subgroup.¹

Figure 5. Active site of enolase complexed with substrate/product.

Enolase (PDB 1ONE) shown as ribbon with catalytic residues and phosphoenolpyruvate (grey stick) and Mg²⁺ (silver sphere); 2-phosphoglycerate (2-PGA) not shown.

CPEPS is an interesting case as sequence data placed it squarely in the enolase superfamily¹. It was first suggested to catalyze a transesterification between phosphoenolpyruvate (PEP) and phosphoformate (PF) (Scheme 10a) (⁵³. The product, carboxyphosphoenolpyruvate (CPEP) would then undergo an intramolecular rearrangement followed by decarboxylation to form phosphonopyruvate (PPA). The series of reactions is part of biosynthetic pathway for phosphinothricin (PT), which is a non-proteinogenic amino acid found in various peptide antibiotics such as PT-Ala-Ala (or bialaphos)⁵⁴. However, this proposed, transesterification reaction does not look like any of those catalyzed by enolase superfamily members indicating that the suggested reaction for CPEPS and its substrates were likely incorrect¹. Instead, it was proposed that the product is generated by the β-elimination of water from carboxy 2-phosphoglycerate) (Scheme 10b). More recent mutagenic and sequence data support this proposal^{54, 55}. CPEPS bears a homologous residue to Lys345 (in yeast enolase) as well as for His373 and the metal binding ligands (Asp166, Glu208, Asp235). There is no general acid catalyst (homolog of Glu211 in enolase). Notably, this example shows the power of comparative studies to correct apparently misassigned functions.

Scheme 10.

Misassigned and Corrected Reactions for CPEPS

cMyc promoter binding protein (MBP1) was identified as an enolase subgroup member due to the presence of two of the three conserved metal-ion coordinating residues found throughout the enolase superfamily (with no homologous residue to Asp320 in enolase)¹. There is a homologous residue for the general base (Lys345), but no homolog for the general acid (Glu211). Although it has a DNA-binding activity; to date, it has not been shown to have catalytic activity.

With the exception of the two enzymes described above, all members of the enolase subgroup, identified thus far function as enolases³⁴. The general base, Lys345, is at the end of the sixth β-strand. E211, the general acid, is found in a loop following the second β-strand.

IMPLICATIONS

The enolase superfamily is the first, and one of the most extensively characterized mechanistically diverse superfamilies^{14, 56, 57}. It is replete with examples of how structure-function relationships in the context of a mechanistically diverse superfamily can yield significant biological insights. An understanding of the chemical mechanisms of the annotated enzymes and the assignment of particular roles in the mechanism to conserved residues enables the prediction of functions for new members, the correction of misannotated functions, and the design of protein engineering experiments. Many different reactions, metabolic pathways and niches were identified and many more remain to be discovered. This same approach has since been demonstrated repeatedly in a number of superfamilies^{14, 56–58}. One important goal is the rapid assignment of function to members in newly sequenced genomes without benchtop experimentation, which would be impossible in view of the numbers of sequences⁵⁹.

The description of the enolase superfamily spawned the discovery and study of other superfamilies including the class I aldolase⁶⁰, amidohydrolase⁶¹, crotonase⁶², haloacid dehalogenase (HAD)⁶³, Nudix hydrolase⁶⁴, tautomerase⁶⁰, thiyl radical¹⁴, and vicinal oxygenase chelate (VOC)⁶⁵ superfamilies. These superfamilies represent the early ones, but many others have joined the list. All have a common mechanistic feature or conserved partial reaction. For example, members of the crotonase superfamily share the ability to generate and stabilize an oxyanion intermediate derived from a thioesters (as found in the namesake enzyme, crotonase) or tetrahedral intermediates formed in the hydrolysis of peptide bonds. Other than the residues comprising the oxyanion hole, no active-site functional groups are conserved. Members of the amidohydrolase superfamily catalyze diverse hydrolytic reactions using a mono- or binuclear metal-ion center to activate the water molecule. The common reactions of representative superfamilies can be found in Table I.

Table 1.

Representative superfamilies and signature common partial reactions

Superfamily	Common Partial Reaction
class I aldolase	protonated Schiff base as electron sink
amidohydrolase	metal-assisted hydrolysis
crotonase	oxyanion-hole stabilization of enolate anion intermediate
haloacid dehalogenase (HAD)	covalent enzyme-substrate intermediate via conserved aspartate
Nudix hydrolase	metal-assisted hydrolysis
tautomerase	N-terminal proline-assisted tautomerization
thiyl radical	conserved thiyl radical derived from cysteine
vicinal oxygen chelate (VOC)	stabilization of oxyanion intermediates from metal-dependent catalysis

The insights yielded for the enolase superfamily also led to widespread implementation and popularization of sequence similarity networks (SSNs)⁶⁶. The calculation and visualization of SSNs, adapted from the analysis of protein-proteins interactions on a large scale, and developed by Patsy Babbitt and co-workers, is a facile method to visualize extremely large sets of related sequences⁶⁶. SSNs are highly effective for the generation of testable hypotheses for sequences with unknown function by allowing a view of the interrelation of sequences with facile adjustment of clustering based on similarity^{66, 67}.

The 1996 paper also prompted the development of new strategies to assign function⁵⁹. The majority of the enolase superfamily members are found in microorganisms, where the metabolic pathways are frequently encoded by operons. In addition, bacteria are subject to straightforward genetic manipulation so that knock-outs can be generated to test predictions. Hence, genome context can be used to assign function. The enolase superfamily is rich in examples where function was identified by operon context, but also demonstrates the many complications of this approach^{59, 68, 69}. A second strategy adopted in the enolase family is the screening of a library of compounds as substrates. This approach was most successful in the MR subgroup where many of the members are acid sugars, but there are also many cases where this approach has not worked^{35, 36}. Paramount amongst the difficulties is addressing whether low activity is biologically relevant or whether it is a failure to have the “right” substrate in the library. Most recently, computational approaches are used to implement high-throughput screening. One approach uses ligand docking of known or potential metabolites^{59, 68, 69}. This circumvents some of the practical limitations of libraries such as the relatively small sizes and potential synthetic liabilities. Another approach screened and explored the properties of solute-binding proteins (SBPs) which assist transport of small solutes from the outer to the inner membrane of Gram-negative bacteria⁶⁹. Study of these proteins has several advantages including the fact that they afford tight ligand binding (with high nM to low μM binding affinities) and do not carry out chemistry. Because SBPs transport the initial metabolite into the cell the chemical space is greatly narrowed. The feasibility of these approaches has been demonstrated for enzymes in the enolase superfamily and other superfamilies⁶⁹. These methodologies were invented and/or implemented as part of the Enzyme Function Initiative, which was a massive, multidisciplinary effort to reliably predict functions of unknown enzymes⁷⁰.

Finally, study of the enolase superfamily addressed a number of issues in divergent evolution as well as providing fundamental principles. How new enzyme activities have evolved has been the subject of enormous speculation^{1, 18, 71}. Previous to this work, it was thought that homologous enzymes differ in specificity, but not mechanism (e.g., serine proteases)⁷². Studies of the enolase superfamily underscored the importance of chemistry in the evolution of new enzymes in which the catalytic machinery for a partial reaction (i.e., the abstraction of an α-proton from a carboxylate group) is hard-wired into a scaffold (i.e., (β/α)₇β-barrel). The active-site architecture of this new enzyme is modified to channel the common intermediate/transition state to different products.

The currently accepted hypothesis for the evolution of a new enzyme involves gene duplication of the progenitor so that the original function is retained while the duplicate gene undergoes divergence of sequence and function¹⁸. Examples from the enolase superfamily illustrate two potential means for selection of the progenitor. In the first, the progenitor fortuitously catalyzes the new reaction at a low, but sufficient rate, as to confer a selective advantage. This is known as catalytic promiscuity. Catalytic promiscuity is widespread in the enolase superfamily (and many other superfamilies), but is well illustrated by the observation that OSBS was first identified by its low NAAAR activity in the enolase superfamily²³. In the second, the progenitor does not catalyze the new reaction, but can with a limited number of mutations. This is exemplified by the introduction of OSBS activity into L-Ala-D/L-Glu epimerase and MLE II²³.

The 1996 paper had a significant impact on the field, representing a paradigm shift in the approaches to the study of enzymes, introducing a “comparative anatomy” approach. The impact of this method is that it allows the community to tackle the vast (and exploding) volume of sequence data from biological systems. From the origin of a 16-member alignment, new tools and approaches are available to the community for the discovery and description of enzymes categorized into superfamilies.

ABBREVIATIONS

AE epimerase

L-Ala-D-Glu-epimerase

AltD

altronate dehydratase

FucD

fuconate dehydratase

GalD

galactonate dehydratase

GlucD

glucarate dehydratase

HAD

haloacid dehalogenase

MR

mandelate racemase

ManD

mannonate dehydratase

β-MAL

β-methylaspartate ammonia-lyase

MLE I

muconate lactonizing enzyme

MLE II

2-chloromuconate lactonizing enzyme

MBP

cMyc promoter binding protein

NAAAR

N-acylamino acid racemase

orf

open reading frame

OSB

o-succinylbenzoate

OSBS

o-succinylbenzoate synthase

PEP

phosphoenolpyruvate

PF

phosphoformate

2-PGA

2-phosphoglycerate

PT

phosphinothricin

PPA

phosphonopyruvate

rTS

reverse thymidylate synthase

SSN

sequence similarity network

SBP

solute-binding protein

SHCHC

2-succinyl-6-hydroxy-2,4-cyclohexadiene-1-carboxylate

RspA

bifunctional altronate/mannonate dehydratase

VOC

vicinal oxygenase chelate