Thioesterase enzyme families: Functions, structures, and mechanisms

2.2.1 HotDog catalytic residues and mechanisms

Families with HotDog^{160, 161} fold structures (TE4–TE15, TE24, TE25, TE31) have highly similar tertiary structures, indicated by the consistently low RMSD_ave and high P_ave values.

HotDog-fold enzymes lack defined non-solvated binding pockets and conserved catalytic residues,⁴⁵ thus a variety of catalytic residues and mechanisms exist.

In TE4, Mycobacterium marinum TesB2 (3U0A) catalytic residues were predicted to be Asp194-Ser216-Gln266, based on comparison to an Escherchia coli TE II enzyme (1C8U) in which Asp204-Gln278-Thr228 orient a water molecule for nucleophilic attack on the substrate.¹¹⁵ This is consistent with the catalytic residues found in Y. pestis TesB (4QFW, 4R4U); a structure that presents an octameric quaternary structure, unique among HotDog families.³¹ A S. cerevisiae TE I structure (1TBU) contains only residues from the N-terminal domain that does not include the residues that could be compared to the catalytic triad. Catalytic residues for the remaining family members were predicted (see Table 3). Of note in these predictions are Mycobacterium avium MAV2540 (3RD7) and MAP1729c (4R9Z); these inactive TesB enzymes contain a mutation in which the highly conserved Asp residue is substituted for an Ala residue. Within TesB TEs, this mutation appears to be unique to Mycobacterium species.³²

In TE6, Mus musculus Acot7 N-terminal domain (2V1O) and C-terminal domain (2Q2B) catalytic residues are reported as Asn24 and Asp213, respectively.¹¹⁶ The structures for human Acot12 (3B7K, 4MOB, 4MOC) and M. musculus Acot7 (4ZV3, 6VFY) contain both N and C-terminal domains. Our alignment placed both 2V1O and 2Q2B over the C-terminal of these structures confirm catalytic residues in the C-terminal domain. Using this molecular symmetry, the N-terminal catalytic residues were predicted as well. This follows with literature which indicates that these structures form a functioning active site when joined as a dimer.³³ A study on N. meningitidis TE 12 (5SZU) supported these findings, pointing to a covalent disulfide bond dimer linkage that is requisite for enzymatic activity.²² The Asn-Asp catalytic motif is highly consistent in this family, recently supported by findings on a Bacillus cereus TE (7CZ3).¹¹⁹ Unique among the family is a S. aureus TE (4NCP) that also relies on a Thr residue for catalysis.¹¹⁸ Also in TE6, YciA structures have and aspartic acid catalytic residues in the same structural position as those in Campylobacter jejuni Cj0915 (3D6L) and Haemophilus influenzae Rd KW20 HI0827 (1YLI, 3BJK).^{46, 117}

Although TE7 has no known crystal structures, sequence analysis with other ACOT enzyme suggests that Asp120 and Asn305 are catalytic residues in the mouse ACOT9 enzyme.⁵⁰

It was proposed for TE8 enzymes, based on the crystal structure of a human Them2 enzyme, that Gly57 and Asn50 bind and polarize the thioester carbonyl group while Asp65 and Ser85 orient and activate the water nucleophile.^{120, 121} It was later proposed, based on mixed quantum mechanics/molecular mechanics simulations of the same human enzyme, that a His-Ser pair acts as the acid proton donor in a concerted mechanism where the Asp residue activates the water molecule.¹²² Based on superimposition with the crystal structure of the human Them2, the structures for M. musculus Acot13 (2CY9) and Danio renio Acot13 (4ORD) are predicted to have the same Asn50, His56, Gly57, Asp65 catalytic structure.¹²⁰ The position of these catalytic residues seem to be extremely highly conserved in this family; the position of the catalytic residues in 2CY9 and 2F0X are exactly the same and are only shifted by one position in 4ORD (e.g., Asp65 to Asp66).

In TE9, an E. coli enzyme (1S5U) is predicted to have catalytic residues Tyr14-Asp18-His25, based on a strong spatial correlation with the catalytic structure (Tyr7-Asp11-His18) of an Helicobacter pylori enzyme (2PZH) in the superimposed structures.⁵³

It was proposed for TE10 4-hydroxybenzoyl-CoA TEs (1LO7, 1LO8, 1LO9) that a helix dipole moment make the thioester carbonyl group more susceptible to a nucleophilic attack by Asp17.¹²³ We predict that Asp16 in an Alkalihalobacillus halodurans enzyme (5WH9) is catalytic, based on the Asp17 residue of a Pseudomonas TE (1BVQ).⁵⁶

TE11 TEs in Arthrobacter (1Q4S, 1Q4T, 1Q4U), E. coli K-12 (4K49), and A. thaliana At2g48320 (4K02) all have nearly identically positioned glycine and glutamic acid catalytic residues.^{35, 124} The crystal structures of other members of this family spatially align well, and are predicted to have the same Gly-Glu catalytic residues (Table 3). Members of TE11 may also act as chain elongation and cyclization domains in certain synthetic pathways.³⁴

TE12 enzymes from Synechocystis (4K00) and Prochlorococcus (2HX5) bacteria have been crystallized. In 4K00, Asp16 was proposed to act as a nucleophile, while it is also possible that it acts as a base to attack the thioester through activation of a water molecule. The thioester oxygen atom could be stabilized by the amide hydrogen on Phe23. Also, Pro57, which positioned above the substrate moiety, may contribute to substrate specificity.³⁵

From the structures 1WLU, 1J1Y, 1WM6, 1WLV, and 1WN3, a study proposed that TE13 Thermus thermophilus PaaI TE hydrolyze substrates with an Asp48-activated water nucleophile.¹²⁶ By comparison, an E. coli PaaI structure (2FS2) with the Arthrobacter TE11 structures and site-directed mutagenesis, a mechanism similar to that in TE11 was proposed: Gly53 prepares the thioester for a nucleophilic attack from Asp61.¹²⁷

TE14, which has many bacterial sequences that have been less characterized than their plant counterparts, has a surprising breath of substrate specificity.⁶⁰ In TE14, a site-directed mutagenesis study on a FatB enzyme from A. thaliana pointed to a Cys264, His229, and Asn227 papain-like catalytic triad.¹⁶² Another site-directed mutagenesis study on a FatB enzyme, from Umbellulaia californica, proposed a catalytic network of Asp281, Asn283, His285, and Glu319.¹²⁸ More recently, structural predictions and site-directed mutagenesis resulted in identifying the catalytic residues of the C. viscosissima acyl-ACP TE.¹⁶

In TE15, a mechanism based on CalE7 enzyme (2W3X), which has no acidic residues in the catalytic region, was proposed: Asn19 and Arg37 hold the substrate while a water molecule or hydroxide anion acts as a nucleophile, and Tyr29 assists in decarboxylation.⁶² Asn, Arg, and Tyr residues in a Micromonospora chersina tebC (2XEM, 2XFL), as well as Streptomyces globisporus (4I4J) and Actinomadura verrucosospora (5VPJ) TEs are predicted to be catalytic based on spatial correspondence with the superimposed M. echinospora structure (2W3X).

The crystal structure for TE24 is represented by Protein Data Bank (PDB) 2PFC and 3B18. The quaternary structure is formed by three dimers and has a long and narrow substrate-binding site. The catalytic site is formed by Asn83, Tyr87, Tyr33, and Met118 for subunit A and Tyr66, Thr70, His72, and Asn74 for subunit B.⁸⁰ Notably, the active site lacks acidic residues common to HotDog TEs, which is also observed in a TE24 Streptomyces enzyme.⁸²

In TE25, a T. thermophilus TE (2CWZ) is predicted to have Thr36, Glu44, and His70 as catalytic residues (see Figure 1) based on the spatial superimposition with the catalytic residues in Streptomyces cattleya fIK (3KUV).¹⁵⁸ The specificity for fluorine-containing compounds could arise from substrate binding through a hydrophobic pocket formed by a helical lid structure (side chains of Val46 and Val54), as well as by Val23, Leu26, Phe33, and Phe36 in S. cattleya fIK.⁸⁴

Family TE31 has Them4 and Them5 isoforms, which have been crystalized and are reported by the 4AE8 and 4AE7 structures, respectively, forming a homodimer unity. Their structures consist of a long central alpha helix surrounded by a six-stranded curved antiparallel beta-sheets. Both isoforms are formed by two active sites per homodimer at the end of each HotDog helix: His152, Gly153, Gly154/His158, Gly159, Gly160 (active site one), and Asp161, Thr177/Asp167, Thr183 (active site two).⁹²

2.2.2 α/β hydrolase catalytic residues and mechanisms

The α/β-hydrolase fold,¹⁶³ found in TE2, TE16 to TE22, and TE26 to TE28, shows higher variation in RMSD_ave and P_ave values than the HotDog fold. Most α/β-hydrolase fold proteins, not only TEs, are present in the ESTHER database.¹⁶⁴ Two families, TE29 and TE30, based on sequence similarity, are likely to have α/β-hydrolase-like folds; however, there are no available structures to confirm. α/β hydrolases have conserved catalytic residues: a nucleophile–histidine–acid triad.¹⁶³ Serine, cysteine, or aspartate can act as the nucleophile. There is a large variation of fold architecture and binding sites in α/β hydrolases.¹⁶⁵ In their catalytic mechanism, the acid stabilizes the histidine, which acts as a base by accepting a proton from the nucleophile, which forms a substrate intermediate that attacked by water. In PKSs or NRPs that make cyclic products, for example, in erythromycin biosynthesis,¹⁶⁶ a hydroxyl group from the substrate chain is used instead of a water molecule. Different cyclization mechanisms lead to a wide variety of PKS or NRP products.¹⁶⁷

The structure of TE2 is represented by 3HLK, which comes from human ACOT2, and 3K2I, which comes from human ACOT4. These structures are somewhat unique for this fold: in the primary structure for these enzymes the Asp residue precedes the His residue, where in all other α/β hydrolase TEs the His residue precedes the Asp residue.¹⁰⁹ The catalytic residues of 3K2I (Table 3) are predicted based on alignment with 3HLK.

In TE16, most structures show a consistent Ser-Asp-His catalytic triad: seen in the human FAS TE domain,^{129, 168-170} the TE domain in Bacillus NRPSs surfactin and fengycin synthetases,^{130, 131} the TE domain of the Aspergillus aflatoxin PKS,¹³² the TE domain of Mycobacterium PKSs involved in making mycolic acids,¹³³ and in the TE domain of NocB enzyme in Nocardia.¹³⁴ However, based on structural superimposition with TE16 structures with identified catalytic residues, we predict that the TE domain of an Acinetobacter baumannii NRPS enzyme (4ZXH, 4ZXI)¹⁷¹ has a Cys-Asp-His catalytic triad (Table 3).

TE17 has enzymes, which are the TE domain of macrocycle-forming PKSs, such as of 6-deoxyerythronolide B synthase from S. erythraea,^{135, 136, 172, 173} picromycin synthase from S. venezuelae,^{136, 174, 175} and tautomycin synthase.¹³⁷ They all show a consistent Ser-Asp-His catalytic triad.

Member of TE18 with crystal structures are type II TEs, a class of enzyme responsible for a variety of functions, primarily maintenance of biosynthetic pathways through release of undesired intermediates from carrier protein domains.^{28, 66, 67, 139-142, 176} A lid-flip conformational change is present in these enzymes and the Ser-Asp-His catalytic triad is conserved. This can be seen in the surfactin synthase from Bacillus subtilis,¹⁷⁷ from the rifamycin biosynthetic cluster from A. mediterranei,⁶⁷ the borrelidin biosynthetic cluster from Streptomyces,²⁸ in the prodiginine biosynthetic pathway in Streptomyces coelicolor,¹³⁹ and in ClbQ and YbtT enzymes in E. coli.^{141, 142} This also holds true in a human TE II and in a TesA from M. tuberculosis.^{66, 140}

In family TE19, a single structure is known, that of a Vibrio harveyi TE, which also has the Ser-Asp-His catalytic triad.¹⁴³

Families TE20, TE21, and TE22 all share the characteristic Ser-Asp-His catalytic triad. Comparison of tertiary structures within each family leads us to predict that this Ser-Asp-His catalytic triad is consistent for all structures (see Table 3 and Figure 2).

TE21 includes mainly eukaryotic acyl-protein hydrolases, as well as enzymes with different functions. The carboxylesterase from P. fluorescens has very little activity on triacylglycerides with fatty acids longer than four carbons, likely due to the loops constraining the active-site cleft.¹⁴⁵ A closely related human enzyme, hAPT1, originally thought to be a lysophospholipase, has been shown to have stronger TE activity.¹⁴⁶ Another APT, from Francisella tularensis, has a similar substrate specificity profile to both of the aforementioned enzymes, though unlike P. fluorescens, it lacks a lid domain.¹⁴⁸ This was confirmed by another study that examined the mechanism of isoform-selective inhibitors on human APT1.¹⁴⁹ The carboxylesterase from P. aeruginosa was shown to have no activity on triacylglycerols, and a preference for eight-carbon acyl substrates. The human lysophospholipase A2 is a cystolic serine hydrolase partially responsible for lysophospholipid metabolism.¹⁵⁰ All of these structures follow the Ser-Asp-His catalytic motif.

Members of TE22 are involved in glutathione-dependent formaldehyde detoxification, and many of the crystal structures in this family are of S-formylglutathione hydrolase (SFGH) enzymes. These have been studied in a variety of species: S. cerevisiae,¹⁵¹ Agrobacterium fabrum str. C58,¹⁵² P. translucida TAC125,¹⁵⁵ Shewanella frigidimarina,¹⁵⁷ and N. meningitidis MC58.¹⁵⁶ Other functions are present in this family as well: (a) a human esterase has been studied because it is relevant to retinoblastoma,¹⁵³ and (b) an oil-degrading bacterium, O. antarctica, expresses an enzyme with carboxylesterase and TE activity.¹⁵⁴ TE22 enzymes have the characteristic Ser-Asp-His catalytic triad. Based on this, the catalytic structure of a S. cerevisiae SFGH (4FLM) is predicted as Ser161-Asp241-His276 (Table 3).

A study on the only crystal structures found for this family, ybfF from E. coli (3BF7, 3BF8), suggests that this family is unique within the α/β hydrolase TEs: rather than the typical Ser-Asp-His catalytic triad, this family seems to have a Ser89-Asp113-Ser206-His234 catalytic tetrad. The α/β hydrolase domain of these structures gives good alignment with other canonical α/β hydrolases. However, the Asp113 residue, which normally lies above or parallel to the His234 imidizole rings, is located in the lower section of the His imidazole ring. The expected position for the Asp113 residue is instead occupied by Ser206, which is well conserved in the ybfF enzymes.⁸⁵

The structure of TE27 enzymes is described by a M. musculus ABHD10, which shows a Ser-His-Asp catalytic triad. The location of the catalytic serine residue suggests a hydrophobic interaction between the lipid substrate and the interior surface of the protein. A “cap domain” above the catalytic triad forms a binding pocket and affects substrate accessibility.⁸⁷ We predict that Ser113-Asp216-His246 is the catalytic triad in an A. vitis enzyme based on comparison to the M. musculus ABHD10 enzyme.⁸⁷

Families TE28 and TE29 have no crystal structures. TE28 shows sequence similarity with a putative α/β hydrolase fold enzyme, and their structure and mechanisms still unknown despite a close relationship with FASs.⁸⁸ TE29 may also have an α/β hydrolase fold, as was predicted from gene ABHD17C.⁸⁹

The structure of an CitA enzyme in TE30, predicted by homology from a co-expression of the PKS gene, suggests a Ser122-His235-Asp207 as catalytic triad.⁹¹

2.2.3 Catalytic residues and mechanisms in other folds

TEs are found in the NagB (TE1) and SGNH (TE3) folds.^{111-114, 110} In TE1, which also includes acyl-CoA transferases, we predict that the catalytic residues of a putative acetyl-CoA hydrolase from Porphyromonas givgivalis (2NVV) and a CoA transferase from P. aeruginosa (2G39) are Val259-Glu284-Asn337-Gly378 and Ile264-Glu288-Asn341-Gly382, respectively, based on those known from A. aceti AarCH6 structures (4EU3, 5DDK).^{107, 108}

In TE3, comparison to available structures—E. coli tesA (e.g., 1IVN, 1JRL)¹¹⁰ and Pseudoalteromonas estA (3HP4)¹¹¹—reveals the likely catalytic residues for an E. coil TE (6LFB, 6LFC) and A. indicum AlinE4 esterase (6IQ9, 6IQA, 6IQB) are Ser10-Asp154-His157 and Ser13-Asp162-His165, respectively. TesA enzymes were found to have a Ser-His-Asp catalytic triad similar to those in α/β hydrolases.¹¹⁰ The crystal structure of TesA from E. coli was found to be particularly compact and rigid, which likely pushes the substrate specificity toward smaller chain lengths.¹¹² It has also proved to be a useful candidate for attempts at engineering TEs to produce specific lengths of free fatty acids.¹¹³ Other SGNH fold TEs, CrmE10 and AlinE4 were similarly susceptible to engineering for increased enzymatic activity.¹¹⁴

Two families have the β-lactamase fold: TE23 and TE32. The structures in TE23 are significantly less well conserved than those in TE32. TE23 hydroxyglutathione hydrolases, which include glyoxalase II enzymes, have a metallo-β-lactamase fold, and their mechanisms are very different from the rest of TEs that do not have catalytic metal ions. Crystal structures of human glyoxalase II (1QH3, 1QH5) reveal two zinc ions with octahedral coordination, interacting with His and Asp residues. Based on this, a study proposed that a hydroxide ion bonded with both ions attacks the carbonyl carbon atom of the glutathione thioester substrate, forming a tetrahedral intermediate, followed by breakage of the CS bond.¹⁷⁸ In mitochondrial glyoxalase II from A. thaliana (1XM8, 2Q42), the zinc ions were also coordinated by His and Asp residues, but were in trigonal bipyramidal and tetrahedral geometries.¹⁷⁹ Another glyoxylase II enzyme, from Salmonella typhimurium (2QED), was proposed to have an uncommon metal affinity: a diiron, dimanganese, or hybrid Fe/Mn.¹⁸⁰ A unique member of the family, a persulfide dioxygenase from Myxococcus xanthus (4YSB), has a single ion in the active site with a two-His and one-carboxylate triad coordination pattern.¹⁸¹

Enzymes in TE32 have monomeric metallo-β-lactamase fold structures, with an Fe(II)Fe(III) center in the active site and an αβ/αβ sandwich core. All the resolved structures in this family are PqsE enzymes from P. aeruginosa, a human pathogen of particular interest due to its tendency for antibiotic resistance.¹⁸² The active center of the enzyme is covered by a lid formed by two α-helices in the C-terminal region, affecting substrate access.⁹⁴ It has also been demonstrated that PqsE has a role in alkylquinolone biosynthesis.¹⁸³

Although TE33 includes no crystal structures, a mechanism has been proposed, which shows an active site His acting as a base, with the substrate hydroxyl forming a hydrogen bond with a histidine residue.^184-186 A nucleophilic attack from a deprotonated hydroxyl at the carbonyl of an acyl-CoA thioester was described, as was the involvement of an Asp residue in the stabilization of the structure within the active site.^{96, 184, 185, 187}

Crystal structure 5VXS represents a member from TE34 and reveals a homotrimer with a substrate-bound cavity located between the N-terminal from one subunit and the C-terminal from the subsequent subunit. The N-terminal forms a β₈α₈-TIM barrel fold and the C-terminal is characterized by a lid-domain consisting of two helices connect by a β-hairpin loop. The β-hairpin loop presents a highly conserved Asp320 that removes a proton from the substrate during the catalytic activity.^{97, 103, 188, 189} In TE34, the catalytic residues for a M. tuberculosis (6AQ4), Cereibacter sphaeroides (4L9Y, 4L9Z), and M. extorquens (5UGR) enzymes are predicted to be Asp261, Asp299, and Asp304, respectively, based on comparison to human CLYBL structure (5VXS).⁹⁷ The catalytic residues for the remaining family members could not be confidently predicted by structural comparison. Two of these are CitE proteins from M. tuberculosis: one study (1U5H) predicts that the catalytic site is in a hydrophobic cavity formed by the C-terminal tips of the TIM β-barrel,¹⁹⁰ while another study (6AQ4) shows that the active site contains an Mg²⁺ ion coordinated by the ligand, Glu112, Asp138, and two water molecules.⁹⁹ Closely related to 1U5H is Y. pestis RipC (3QLL), for which the active site is similarly predicted. However, it is also suggested that the active site for 3QLL may be formed through an intermonomer interaction.¹⁰⁰

The structure 6AUN in TE35 is characterized by the presence of an Ankyrin domain, a 33-residue helix-turn-helix structure followed by a hairpin-like loop, and a catalytic domain. Regarding the catalytic mechanisms, a dyad formed by Ser-Asp is responsible for the lipid hydrolysis.^{159, 191}