FabI is the major and most studied enoyl-ACP reductase enzyme of the FAS II system, which is found in several microorganisms, such as Staphylococcus aureus, E. coli, Bacillus subtilis, Francisella tularensis, Burkholderia pseudomallei, Haemophilus influenza, Moraxella catarrhalis, Acinetobacter baumannii, Bacillus anthracis, Chlamydia trachomatis, Pseudomonas aeruginosa, Plasmodium falciparum, and Toxoplasma gondii. Some bacterial species possess only FabI, while others contain an additional ENR from the FabL, FabV or FabK types (Khan et al., 2018). InhA (FabI homologue) is also present in Mycobacterium tuberculosis and other mycobacterial species (Banerjee et al., 1994; Quémard et al., 1995).

ENR FabI enzymes catalyze the reduction of trans-2-acyl-ACP substrates (enoyl-ACP) to acyl-ACP using either NADH or NADPH (depending on the enzyme) as reductant species (Figure 4; Massengo-Tiassé and Cronan, 2009). The chemical mechanism in bacterial FabI enzymes involves a nucleophilic conjugate addition of a hydride ion from the 4S hydrogen position of the nicotinamide ring of either NADH or NADPH to the C3 position (Cβ) of the α, β-unsaturated thioester of the enoyl-ACP substrate, followed by protonation at C2 (Cα) of the enzyme-stabilized enolate intermediate (Figure 4A; Saito et al., 1980; Quémard et al., 1995; Parikh et al., 1999). The conformation of the cofactor is modulated by the enzyme to facilitate the hydride transfer, with the 4S hydrogen pseudoaxially oriented relative to the nicotinamide ring (Bell et al., 2007). A breakthrough in the understanding of the reaction mechanism of enzymes catalyzing hydride transfers was made with the MDR superfamily member Ccr, a crotonyl-CoA reductase/carboxylase. It was shown that NADPH forms a covalent ene adduct intermediate with crotonyl-CoA (Rosenthal et al., 2014). Subsequently, a similar covalent adduct was found between NADH and an octenoyl-CoA substrate during the reaction cycle of InhA (Figure 4B), raising the interesting possibility that covalent adduct formation can be a conserved hydride transfer mechanism among dehydrogenases/reductases (Vögeli et al., 2018). This intermediate species has a covalent bond between the C2 of the NADH nicotinamide ring and the C2 atom of the acyl substrate. As it is already reduced at the C3 position, but not yet protonated at C2, this species is an intermediate between the two half-reactions of hydride transfer and protonation. Conceivably, the C2-ene adduct could be produced directly from an ene-shaped transition state by a pericyclic ene reaction or after hydride transfer by a Michael addition (Figure 4B). Apart from improving our understanding of the reaction mechanisms of FabI enzymes and dehydrogenases in general, the discovery of this covalent intermediate opens exciting new avenues for the development of novel high-affinity mechanistic inhibitors.

The protonation mechanism is still a matter of debate. Initially, it was suggested that the C1 carbonyl oxygen of the enolate anion intermediate could be protonated, with subsequent tautomerization leading to the reduced acyl product. The conserved tyrosine residue from the active site signature motif of E. coli FabI (Y156 in this enzyme—EcFabI—and Y158 in M. tuberculosis InhA) was suggested to be the proton donor to the enolate anion, while the conserved lysine (K163 and K165, respectively) would stabilize the negative charge of the transition state (Baldock et al., 1998b). However, as subsequently found, the reduction reaction follows a specific stereochemical course, with hydride transfer at C3 and protonation at C2 occurring in the same face (syn addition), yielding (2R,3S)-[2,3-²H₂]octanoyl-CoA from trans-2-octenoyl-CoA after ²H₂ addition (Fillgrove and Anderson, 2000). As already observed, this defined stereochemical outcome is not compatible with a mechanism involving enol formation by protonation of the enolate oxygen (Schiebel et al., 2015), implicating a direct protonation of C2 (Cα carbon).

The structural elucidation of a ternary complex containing InhA associated with both NAD⁺ and a C16 fatty acyl substrate (trans-2-hexadecenoyl-(N-acetylcysteamine)-thioester) gave new insights into the reaction mechanisms of ENR FabIs (Rozwarski et al., 1999). The fatty acyl substrate lies on the top of the cofactor and adopts a U-shaped conformation. Both fatty acyl and NAD⁺ binding sites lie within a pocket containing a major and a minor portal, with the substrate-binding loop (SBL) contributing with most of the hydrophobic residues that surrounds the fatty acyl chain (Figures 5A,B). The SBL was found to be longer in InhA from M. tuberculosis and mycobacterial orthologs than that of other ENR FabIs, comprising α-helix 6, a connecting loop, and part of α-helix 7 (residues 196–219). A longer binding crevice accommodates longer fatty acyl substrates, consistent with the fact that mycobacterial ENR FabIs, which are involved in mycolic acid biosynthesis, preferentially reduce long-chain substrates (>12 carbons).

Residue F149 is structurally analogous to S146 of the canonical SDR E. coli 7α-HSDH, a residue that is part of the catalytic triad of this enzyme (S146-Y159-K163). This led to the proposal that the catalytic triad in bacterial FabIs is F149-Y158-K165 (InhA positions), with F149 replaced by tyrosine for many FabIs (Figure 5C; Rozwarski et al., 1999). Interestingly, the only hydrogen bond between the fatty acyl substrate and the protein is between the hydroxyl oxygen from the conserved Y158 and the thioester carbonyl oxygen, with both of them making additional hydrogen bonds with the 2′-hydroxyl oxygen of the nicotinamide ribose of NAD⁺ (Figure 5D). The interaction between Y158 and the thioester carbonyl oxygen was unanticipated, since this residue adopts a different conformation in the structure of InhA bound to NADH elucidated previously (Dessen et al., 1995), referred to as the out conformation (Chollet et al., 2018). The ternary structure revealed that residue Y158 from InhA rotates upon fatty acyl substrate binding, orientating in a similar position adopted by the correspondent Y156 residue from EcFabI (in conformation), and in this way being able to be hydrogen bonded with the thioester carbonyl oxygen. Kinetic isotope effect (KIE) studies of wild-type and mutant InhA (Y158F, Y158A, Y158S) corroborated this model, with Y158 functioning as an electrophilic catalyst, stabilizing the carbonyl enolate but not acting as a proton donor (Parikh et al., 1999). The ternary structure also revealed that the catalytic function of the conserved K165 residue is to hold the NADH in place and not to form an oxyanion hole for the thioester carbonyl oxygen, as previously suggested (Figure 5D; Baldock et al., 1998b). The role of K165 in cofactor binding was also confirmed in enzymatic studies of site-directed mutants (Parikh et al., 1999).

Residue T196 was also implicated in catalysis, by hydrogen bonding a water molecule that is also held by phosphate oxygen of NAD⁺ (Figure 5D). This structure also revealed the absence of any polar or ionizable residue close enough to C2 to serve as a proton donor, suggesting the proton could come from the solvent (Rozwarski et al., 1999). Aside its role modulating the conformation of NADH into the active site, contributing to the pseudoaxial positioning of the 4S hydrogen for hydride transfer (Bell et al., 2007), residue F149 would also serve as a barrier to a water channel and, upon rotation, would expose the active site to a network of hydrogen-bonded water molecules contained within this water channel that, ultimately, would lead to protonation (Figures 5E,F). This mechanistic model was further explored for SaFabI, the ENR FabI from S. aureus. The catalytic efficiency of SaFabI was compromised when residues stabilizing this conserved water wire were mutated. A pathway of proton transport was proposed, starting in a water chamber located inside the homo-tetrameric structure and followed by the network of water molecules inside the water channel (Schiebel et al., 2015). Moreover, the proposition that NAD(P)⁺ amide could be the source of the proton (Fillgrove and Anderson, 2000) was rejected because its hydrogen atoms are involved in a tight hydrogen bonding network (Schiebel et al., 2015).

Nevertheless, this picture for the protonation reaction is still controversial. In contrast to Parikh et al. (1999); Vögeli et al. (2018) found that Y158 is directed involved in protonation and is essential for the stereospecificity of the protonation reaction. The differences in data between the two studies were attributed to the method employed in the latter that directly measures KIE on protonation (Vögeli et al., 2018). The involvement of the intermediate covalent adduct, which was reported in this same study, was invoked to reconcile the syn addition of hydride transfer and protonation with a mechanism involving direct protonation from Y158. The (2S)-C2-ene adduct (Figure 4B) would lead to a change in hybridization of the C2 from sp2 to sp3, bringing the C2 closer to Y158 (Vögeli et al., 2018). A clear picture of the reaction mechanisms employed by ENR FabIs will require new mechanistic studies addressing these issues.

Contrary to early expectations (Weeks and Wakil, 1968), FabI is the single trans-2-enoyl-ACP reductase enzyme in E. coli (Bergler et al., 1996). It is encoded by fabI and utilizes substrates with different chain lengths (Campbell and Cronan, 2001). E. coli FabI is a NADH-dependent FabI. The lower NADPH-dependent activity previously reported for this enzyme (Bergler et al., 1996) could not be reproduced in later experiments (Heath et al., 2000b). B. subtilis and Burkholderia. pseudomallei FabI have also a marked preference for NADH over NADPH (Heath et al., 2000c; Liu et al., 2011). Moreover, M. tuberculosis InhA is strictly dependent on NADH (Quémard et al., 1995). However, S. aureus FabI (SaFabI) have a clear preference for NADPH cofactor, despite being highly homologous to the others FabI enzymes (Heath et al., 2000b). The structural basis for this difference in cofactor specificity was found to be the presence of two positively charged residues (R40 and K41) lying close to the 2′-phosphate bound to the adenosine ribose of NADPH in SaFabI (Xu et al., 2008; Priyadarshi et al., 2010; Figure 6). Residues at correspondent positions in NADH-dependent FabI enzymes from E. coli (Baldock et al., 1998b) and M. tuberculosis (Dessen et al., 1995) are polar (Q40 in EcFabI) or nonpolar (F41 in M. tuberculosis InhA). Moreover, the importance of R40 and K41 for NADPH binding was directly evaluated by site-directed mutagenesis (Xu et al., 2008). Both R40Q and K41N mutants show at least 50-fold decrease in k_cat/K_m for NADPH, with correspondent 5-7-fold increases in k_cat/K_m for NADH (Xu et al., 2008).

Steady-state kinetics showed that the two substrates can bind to FabI via a sequential mechanism with few differences in the order of the substrate binding among microorganisms. InhA kinetic mechanism is not strictly ordered, but there is a preferred order of addition, with NADH binding first (Quémard et al., 1995). B. pseudomallei FabI-1 follows an ordered Bi-Bi kinetic mechanism in which NADH is the first substrate to bind in the active site (Liu et al., 2011), while a rapid equilibrium random kinetic mechanism has been proposed for Haemophilus influenzae FabI (Marcinkeviciene et al., 2001).

FabI has been shown to be essential for bacterial survival (Heath and Rock, 1995). The effect of FabI deficiency in a heat-sensitive FabI conditional mutant of E. coli impaired the completion of the elongation of long-chain saturated or unsaturated fatty acids without supplementation with exogenous FabI. The fabI(Ts) strains accumulate β-hydroxybutyryl-ACP at the nonpermissive temperature indicating that FabI is the only reductase required to complete both saturated and unsaturated fatty acid synthesis in the type II synthase system of E. coli (Heath and Rock, 1995). In addition, FabI seems to be an important control point in the fatty acid elongation cycle (Bergler et al., 1996). Due to the essential role that the fatty acids play in bacterial cell survival and the low degree of ENR sequence homology with mammalian targets, it has become one of the most attractive new antibiotic targets for the treatment of multidrug-resistant infections. FabI is well-studied and a validated target, in which several potent inhibitors with diverse chemical scaffolds have been discovered with potential antibacterial activity (Rana et al., 2020).

Current ENR FabI inhibitory compounds can be broadly considered as belonging to one of the following categories: (1) competitive inhibitors of the fatty acyl substrate, (2) molecules that form covalent adducts with the cofactor, and (3) bisubstrate inhibitors (Chollet et al., 2018). Different molecular structures that act as competitive inhibitors and bind into the fatty-acyl substrate-binding site were structurally characterized (Chollet et al., 2018; Rana et al., 2020). Among them, triclosan (TCL) and derivatives have been studied in more detail (Figure 7A). TCL, a trichlorinated biphenyl ether, is a common antibacterial additive used in consumer products and acts as a non-covalent FabI inhibitor (Heath et al., 1999). Triclosan is a slow, tight binding, reversible inhibitor of the EcFabI, binding preferentially to the E-NAD⁺ form of the wild-type enzyme, with an inhibition constant value for this complex (K₁) of 23 pM (Sivaraman et al., 2003). The crystal structure has shown that TCL forms a tightly associated ternary complex with the protein and the nicotinamide cofactor stabilized by hydrogen bonds and hydrophobic interactions (Figure 7B; Stewart et al., 1999). EcFabI mutants (G93V, M159T and F203L) that confer resistance to triclosan showed a substantially reduced affinity binding to triclosan with K₁ values of 0.2 μM, 4 nM, and 0.9 nM, respectively (Sivaraman et al., 2003). All the three FabI mutant residues are involved in the formation of the cofactor binding site, located close to the inhibitor (Qiu et al., 1999; Stewart et al., 1999). In addition, a mutation of Phe204 (equivalent to Phe203 in EcFabI) to Cys in S. aureus that disrupts the formation of the FabI-NAD⁺-triclosan ternary complex confers resistance to triclosan in the S. aureus mutant (Fan et al., 2002). Triclosan also inhibits InhA from M. tuberculosis (Parikh et al., 2000) and Mycobacterium smegmatis (McMurry et al., 1999). Interestingly, triclosan is a weaker inhibitor of InhA compared to FabI. While triclosan binds preferentially to the E-NAD⁺ form of InhA, the affinity of triclosan for InhA is 10,000-fold lower than that for the EcFabI (Parikh et al., 2000). On the other hand, S. pneumoniae (FabK), B. subtilis (FabI and FabL), and P. aeruginosa (FabI and FabV) are resistant to triclosan (Heath and Rock, 2000a; Heath et al., 2000b). Overexpression of the AcrAB-encoded multidrug efflux pump and MarA and SoxS, both positive regulators of acrAB, can also confer resistance to triclosan (Massengo-Tiassé and Cronan, 2009).

The poor inhibitory activity of TCL against InhA and its low antimicrobial activity against Mtb strains prompted efforts to discover more potent TCL derivatives. This led to the development of the most potent inhibitor of InhA currently available, the TCL derivative 2-(o-Tolyloxy)-5-hexyphenol (PT70; Figure 7A). This compound has an inhibition constant for the InhA-NAD⁺ complex of 22 pM (Luckner et al., 2010). As in the case of TCL, the ternary complex of PT70, NAD⁺ and InhA revealed that the compound occupies the same space of the fatty acyl substrate (Figure 7C; Luckner et al., 2010). The hydroxyl oxygen from the PT70 A-ring establishes hydrogen bounds with the catalytic Y158 and the 2′-nicotinamide ribose oxygen, which, in turn, also forms hydrogen bonds with the PT70 ether oxygen and with the catalytic K165 (Figure 7C). Tyrosine 158 is in the in conformation, pointing toward K165, in the same orientation found in the ternary complex with C16 fatty acyl substrate, opposed to the out conformation, found in all binary complexes of InhA with NAD⁺/NADH (Chollet et al., 2018). Residue F149, which together with Y158 and K165 makes up the catalytic triad of bacterial FabIs, is also found in two different conformations (Figure 7D). Interestingly, F149 in and Y158 in conformations are mutually exclusive, as they overlap in space (Figure 7D), and some inhibitors stabilize F149 in conformation, restricting Y158 into the catalytically inactive out conformation (Chollet et al., 2018).

Inhibition studies with TCL and TCL derivatives acting on ENR FabIs are interesting examples in which efforts were undertaken to optimize for compounds with longer residence times (t_R) rather than necessarily for a lower equilibrium binding constant K_d, an idea originally proposed by Copeland and colleagues (Copeland et al., 2006). Despite being a rapid reversible inhibitor for mycobacterial InhAs, TCL is a slow-onset inhibitor of various ENR FabIs (Xu et al., 2008; Lu et al., 2009; Schiebel et al., 2012; Chang et al., 2013). Moreover, TCL derivatives that are “slow-binders” of M. tuberculosis InhA were also developed, including the aforementioned PT70 (Luckner et al., 2010). The kinetic mechanism of these inhibitors was suggested to involve an induced-fit step, wherein the initial protein-ligand complex (EI) is followed by the ordering of the SBL H6 helix (196–206) into the closed conformation, which represents the EI* state (Li et al., 2014). Importantly, this segment is found in a wide range of conformations that can be described as disordered, widely open, open or closed (Figure 7E; Chollet et al., 2018). In the case of PT70, the rationale for its development was to reduce the ligand conformational flexibility to compensate for the entropic penalty associated with H6 helix ordering. The introduction of a methyl group ortho to the diphenyl ether linkage of the rapid reversible inhibitor 6PP was sufficient to generate a compound (PT70) with a residence time t_R 14,000 times longer. In accordance with the proposed kinetic mechanism, the ternary complex of InhA, NAD⁺ and PT70 reveals that H6 helix is in the closed conformation (Luckner et al., 2010).

The second category of ENR FabI inhibitors includes adduct-forming compounds such as diazaborines (DAZ; Figures 8A,B) and the TB drugs isoniazid (INH; Figures 8C,D), ethionamide (ETH) and prothionamide (PTH). Diazaborines are a class of compounds that have a heterocyclic 1,2-diazine ring containing a boron atom (Baldock et al., 1998a). Several DAZ derivatives were synthesized in which the boron-containing diazine ring is fused to a five or six-membered ring. Five-membered thieno-fused derivatives (Figure 8A) are among the most potent antibacterial agents of this class, followed by the benzodiazaborines (Grassberger et al., 1984). Both thieno- and benzodiazaborines were found to inhibit ENR FabI by forming a covalent bond between the boron atom and the 2′-hydroxyl of NAD⁺ ribose. The diazaborine group binds in the enoyl substrate active site to form the diazaborine-NAD adduct (Figure 8B; Baldock et al., 1996). Diazaborine inhibitors also showed antibacterial activity against M. tuberculosis (Davis et al., 1998).

INH, a frontline antitubercular drug that targets InhA from M. tuberculosis, is a prodrug that requires activation by mycobacterial KatG catalase/peroxidase into an activated form of the drug (suspected to be an isonicotinic-acyl radical; Zhang et al., 1992, 1993; Heym et al., 1993; Johnsson and Schultz, 1994). Structural and mass spectrometry data revealed that the activated form of INH is found covalently linked to NADH within the active site of INH-inhibited InhA (Rozwarski et al., 1998). The INH-NAD adduct has an isonicotinic acyl group attached in the 4S position of the nicotinamide ring, replacing the hydrogen required for hydride transfer in the first half-reaction step of ENR FabIs (Figure 8C). INH-NAD adduct is a slow, tight binding competitive inhibitor for InhA (Rawat et al., 2003). The structure of InhA bound to INH-NAD adduct was found to be very similar to the binary InhA:NADH complex, with the main exception being the conformation of Phe149, which is rotated to accommodate the isonicotinic acyl group, going from the “out” to the “in” conformation (Figure 8D), according to the nomenclature proposed by Chollet et al. (2018). Rozwarski et al. (1998) also suggested that the INH-NAD adduct formation occurs within the enzyme’s active site, because they were unable to produce it without the addition of InhA. However, the formation of the INH-NAD adduct in the absence of InhA was achieved later on (Chollet et al., 2015). The preformed adduct was able to displace the cofactor NADH bound to InhA in vitro and only the reaction product with 4S configuration was found inside the structure of a binary complex between InhA and adduct (Chollet et al., 2015). Therefore, it is still unknown whether, in mycobacterial cells, INH-NAD adduct forms before binding to InhA or inside its active site.

Most INH resistance is related to mutations in the KatG and InhA enzymes. Mutation of InhA active site residues, such as Ser94 to Ala (S94A equivalent to G93V of EcFabI) and Ile16 to Thr, decreases the enzyme affinity to NADH and INH, and strains with such mutations are resistant to INH. Structural studies of the InhA mutants reveal that INH resistance is directly associated with disruption of hydrogen bounds that stabilize NADH binding in the wild-type enzyme (Dessen et al., 1995).

ETH and PTH are thioamide analogues of INH that are used as second-line drugs to treat TB. They are structurally similar to INH, also require an activation step inside cells, and form NAD adducts that inhibit InhA (Wang et al., 2007). However, the activation mechanism was found to differ, with the monooxygenase EthA being required to activate ETH and PTH (Baulard et al., 2000; Wang et al., 2007).

The third category of ENR FabI inhibitors that act as bisubstrate inhibitors is represented by pyridomycin (Figure 8E). This compound is a natural product of the cyclodepsipeptide family, originally isolated as an antimycobacterial agent from Streptomyces pyridomyceticus (Maeda et al., 1953), which was later reclassified as Dactylosporangium fulvum (Shomura et al., 1986). InhA was identified as the target of this compound in M. tuberculosis after whole-genome sequencing of selected pyridomycin-resistant mutants (Hartkoorn et al., 2012). It was found as the first nonmimetic competitive inhibitor for the NADH-binding site of InhA (Hartkoorn et al., 2012). The structural elucidation of the binary complex containing pyridomycin bound to InhA revealed that the compound interacts with both the NADH- and lipid-binding sites (Figure 8F; Hartkoorn et al., 2014).

Several other classes of synthetic molecules, such as indoles, acrylamides, imidazoles, napthyridones, coumarins, pyrrolidine, pyridines, among others, have been reported as FabI inhibitors (in a nanomolar to micromolar range), which represent starting-point structures for the development of new antibacterial agents (Rana et al., 2020). A series of piperazine derivatives showed potent inhibition against InhA from M. tuberculosis (Rotta et al., 2015). Some natural products like cephalochromin (Zheng et al., 2007), verrulactone A and B (Kim et al., 2012b) are also capable of inhibiting the SaFabI from S. aureus, with excellent in vitro antibacterial activity. The efforts devoted to the development of effective FabI-targeting antibiotics lead to the discovery of AFN-1252, CG400549 and FAB-1, currently in various stages of clinical development for the treatment MDR infections (Rana et al., 2020).

The enzyme FabV is an enoyl-ACP reductase (ENR) originally identified in Vibrio cholera (VcFabV), a Gram-negative bacterium that causes cholera in humans (Massengo-Tiassé and Cronan, 2008). Crystal structures from ENR FabV from Xanthomonas oryzae (Li et al., 2011), Yersinia pestis (Hirschbeck et al., 2012; Neckles et al., 2016), B. pseudomallei (PDB: 4BKO, not published) and Vibrio fischeri (Kwon et al., 2017) are available. ENR FabVs are 60% larger than other SDR members (except FabMG) but contain the typically conserved Rossmann fold, with an eight-stranded parallel β-sheet with three α-helices on one side and additional four helices on the other (Figure 9A; Hirschbeck et al., 2012). The structures available for the four FabVs are very similar (Figure 9B). However, when compared to other SDR ENRs, such as the EcFabI, there are many regions exclusively found on FabVs (Figure 9C). In particular, there is a FabV-specific N-terminal region that includes a β-hairpin (β1/β2) and an extended α-helix (α1), more central structures, such as the extended β-hairpin (β9/β10) that lies at the top of the substrate-binding pocket and four helices that surround the active site (α11 to α14), and an additional C-terminal β-strand (β14; Hirschbeck et al., 2012). The catalytic tyrosine and lysine residues do not conform with the signature motif Yx₂(x)Mx₃K of the divergent subfamily of SDR ENRs (Kavanagh et al., 2008). In FabVs, they are spaced by eight residues (Yx₈K), without the internal methionine. However, the two additional residues found in the signature motif of FabVs do not change the relative positions of the catalytic tyrosine and lysine, which lie in very similar positions relative to E. coli EcFabI (Figure 9D).

FabV is expressed in other organisms than V. cholera, including important pathogens, such as Burkholderia mallei, P. aeruginosa and Y. pestis (Lu and Tonge, 2010; Zhu et al., 2010; Hirschbeck et al., 2012). Some organisms, such as B. mallei and P. aeruginosa, have both FabI and FabV, while V. cholera and Y. pestis contain only FabV (Rana et al., 2020). FabV catalyzes the reduction of the enoyl-ACP using NADH as the hydride donor. Kinetic studies have shown that the FabV from B. mallei (BmFabV) features a sequential bi-bi catalysis mechanism, in which the NADH binding occurs first followed by the enoyl substrate (Lu and Tonge, 2010), like the mechanism reported previously for FabI (Fawcett et al., 2000; Marcinkeviciene et al., 2001). The Rossmann fold on the N-terminal part of FabV binds the nucleotide cofactor (Figure 9A) while the catalytic site that provides substrate specificity is located on the C-terminal part of the enzyme (Hirschbeck et al., 2012). Furthermore, FabV are found as monomers, differently from the others SDR ENRs FabI and FabL, that form tetrameric structures (Hirschbeck et al., 2012).

The deletion of fabV in P. aeruginosa results in a lower production of fatty acids and slower growth of these strains. Furthermore, mutation of fabI has no effect on growth or the fatty acid profiles, suggesting that FabV is the main ENR of this organism (Huang et al., 2016). Considering the successful use of FabI as a drug target, the development of FabV inhibitors in drug discovery efforts is underway. Triclosan is a potent inhibitor of FabI from different organisms, but its inhibition of FabV is weak and reversible (Massengo-Tiassé and Cronan, 2008; Lu and Tonge, 2010). Accordingly, it was shown that P. aeruginosa is resistant to TCL due to the presence of FabV, since the deletion of the fabV gene causes sensitivity to this compound (Zhu et al., 2010; Huang et al., 2016).

Inhibitory studies with compounds containing the 2-pyridone scaffold revealed that they behave as competitive inhibitors against FabV from Y. pestis (YpFabV), with Ki values of 1–2 μM, being considerably more potent than TCL (Ki value of 71 μM; Hirschbeck et al., 2012). Inhibitory studies with the YpFabV T276S mutant have shown that the activities of different classes of inhibitors are affected when compared to the wild-type enzyme. The affinity is lower for diphenyl ether inhibitors, while it is higher for 2-pyridone inhibitors and do not change for 4-pyridone compounds (Neckles et al., 2016). The T276 residue is located at the SBL on the N-terminus, and mutations in this residue change the size of the active site. This study was important to elucidate the importance of residue T276 for the activity of FabV, providing a framework for the development of novel inhibitors based on structure–activity relationships for this enzyme (Neckles et al., 2016).

FabL (or YgaA or enoyl-ACP reductase III) is an ENR encoded by the ygaA gene that was first described by Heath et al. (2000c) in B. subtilis (BsFabL). This enzyme is a tetramer comprised of protomers containing 250 amino-acid residues in length (subunit MW = 27,178 Da) and has a strict requirement for NADPH as cofactor (Heath et al., 2000c). The crystallographic structure reveals that BsFabL has the classical Rossmann fold (Figure 10A) and, albeit only 25% identical in sequence with EcFabI, they are very similar in structure (Figure 10B; Kim et al., 2011). FabL enzymes have a signature motif with the expected spacing for the divergent subfamily of SDR ENRs (Yx₆K), but without the internal methionine (Yx₂(x)Mx₃K). The extended signature Ser-Xn-Tyr-Xn-Tyr-X6-Lys, including the highly conserved Ser and Tyr residues involved in a hydrogen bond network in FabIs (White et al., 2005), is replaced by Asn-Xn-Ser-Xn-Tyr-X6-Lys in the BsFabL, as is also found for FabG and other SDR members. The positions of the functional groups suggest that the active site of FabL is a hybrid between that of FabI and FabG (Kim et al., 2011).

The enzyme BsFabL is inhibited by triclosan (TCL), presenting an IC50 for this compound of 16 μM. This value is eight times higher than what is found for EcFabI (2 μM), but identical to the IC50 of TCL for B.subtilis BsFabI (Heath et al., 2000c). However, TCL is a slow-onset inhibitor for BsFabI, forming a stable NAD–triclosan ternary complex in a time-dependent manner, while for BsFabL it is a reversible inhibitor (Heath et al., 2000c). In fact, FabL is responsible for triclosan resistance in B. subtilis and the inclusion of the fabL gene in high-copy-number plasmids leads to increased triclosan tolerance in E. coli (Heath et al., 2000c). Noteworthy, FabL is being used as a triclosan selection marker in B. subtilis/E. coli shuttle vectors (Mordukhova and Pan, 2020).

The comparison of the apo form with BsFabL bound to TCL and NADP revealed that some structural regions surrounding the lipid-binding site are disordered in the apo form, and, upon NADH and TCL binding, these regions are drastically reoriented (Figure 10C). The catalytic tyrosine (Y151) and lysine (K158) of the signature motif are both completely reoriented upon NADP and TCL binding, positioning these residues for cofactor and fatty acyl binding (Figures 10D,E). Indeed, the analysis of ligand binding interactions reveals the pivotal role of Y151 and K158 in cofactor binding and in the interaction with TCL, positioned into the fatty acyl binding site (Figure 10F). The hydroxyl oxygen from Y151 hydrogen bonds with both S141 and the hydroxyl oxygen from the chlorophenol ring of TCL. On the other hand, K158 hydrogen bonds with both 2′- and 3′-hydroxyl oxygens from the nicotinamide ribose, as expected from its known role in cofactor stabilization in other SDR enzymes. Direct TCL-NADP interactions were also detected, with 2′-hydroxyl oxygen forming hydrogen bonds with both the hydroxyl oxygen of the TCL chlorophenol ring and ether oxygen (Figure 10F).

To explore the possibility that the extensive and long-term use of TCL have led to the emergency of TCL-resistant bacteria in the environment, and that the soil could represent an environmental source of antibiotic resistant genes for human pathogenic bacteria, Khan et al. (2016) investigated the TCL resistome using a functional metagenomics-based screening. Samples from alluvial soil and industrially contaminated soil were used for metagenomic library construction. TCL-resistant clones were selected, subcloned, and the gene responsible for TCL resistance identified. Aside the identification of homologues of prototypic FabI, FabV, FabK or FabL ENRs, novel candidate ENR sequences were identified, presenting similarity with 7-α hydroxysteroid dehydrogenase (7-AHSDH), FabG or containing an YX₇K signature motif (Khan et al., 2016).

The 7-AHSDH-like protein was characterized in more detail in another study and referred to as FabL2. This sequence presented significant homology with sequences annotated as 7-AHSDH in Epsilonproteobacteria (69 to 96%), followed by FabL (41%). Its coding sequence functionally complemented the fabI(Ts) E. coli mutant JPP1111 strain and conferred complete TCL tolerance. Interestingly, FabL2-like sequences were found among pathogenic Epsilonproteobacteria, in many cases along with FabI ENRs (Khan et al., 2018). The recombinant enzyme possesses NADPH-dependent ENR activity, but, albeit the higher similarity, did not exhibit 7-AHSDH activity. The modeled structure has a typical Rossmann fold containing a 7-stranded parallel β-sheet between two sets of three α-helices. The structural comparison between FabL and FabL2 revealed the presence of an extra 6-residue long loop in the latter (96–101). Based on molecular dynamics (MD) simulations, it was proposed that this loop might be involved in TCL resistance. Effectively, removal of the extra loop by site-directed mutagenesis resulted in loss of TCL resistance (Khan et al., 2018).

Another alternate SDR ENR obtained from the same metagenome screening (Khan et al., 2016) was characterized in more detail (Khan et al., 2019). This sequence is similar to FabI sequences (30%) but contains the Yx₇K signature motif; hence, the name FabI2. FabI2 complemented ENR activity and in vitro is a NADH-dependent ENR. Additionally, a model of FabL2 revealed a Rossmann fold with a 7-stranded parallel β-sheet flanked by three α-helices on each side.

A novel hypothetical protein conferring resistance to TCL (protein AH4-3), obtained in the same functional screen, was also studied in more detail. Differently to the previous sequences described above (FabL2 and FabI2), this protein does not show any sequence similarity to proteins of known function (Kim et al., 2020). However, crystal structures were obtained and revealed that this protein is a novel SDR-type ENR protein. As is the case for FabV members, FabMG proteins are considerably larger than typical SDR proteins. The metagenome-derived FabMG has 438 residues, with a core Rossmann fold structure from which long α-helices project upwards (Figure 11A). The superposition of FabMG with EcFabI reveals the conserved Rossmann fold above which the cofactor lies and a fatty acyl-binding site with extensive differences (Figure 11B). This ENR displays a signature motif with a shorter spacer, Yx₅K. The ENR activity of FabMG was also confirmed in complementation studies using the fabI(Ts) E. coli mutant JPP1111 strain. Moreover, site-directed mutants Y260F and K266A were unable to complement this same strain, confirming the functional importance of Yx₅K for its ENR activity.

Although members of the SDR superfamily make up the most common type of ENRs in bacteria, a non-SDR ENR is found in some organisms, like S. pneumoniae, C. difficile and other Gram-positive bacteria (Marrakchi et al., 2002; Massengo-Tiassé and Cronan, 2009). FabK, a non-SDR ENR protein, was first described in S. pneumoniae (Heath and Rock, 2000a). This bacterium has a complete set of fab genes, except for fabI. FabK was found as an open reading frame (ORF) encoding a 34 KDa protein, demonstrated to have activity in an enoyl-ACP reductase coupled-assay system and to be resistant to triclosan inhibition (Heath and Rock, 2000a). FabK has been identified in other organisms, such as in Streptococcus mutans, Thermotoga maritima, C. difficile, Porphyromonas gingivalis and Enterococcus faecalis, the latter having both FabI and FabK enzymes, and has been studied as a drug target (Hoang and Schweizer, 1999; Heath and Rock, 2000a; Zhu et al., 2010; Nakashima et al., 2011; Khan et al., 2018).

This protein is structurally different from other ENRs, displaying a TIM barrel fold with eight α-helices and eight β-sheets alternated (Figure 12A). It is a flavin mononucleotide (FMN)-containing enzyme that belongs to the NAD(P)H-dependent flavin oxidoreductase family (Massengo-Tiassé and Cronan, 2009). S. pneumoniae FabK (SpFabK) was the first to have its crystal structure elucidated (Saito et al., 2008). FabK forms homodimers, and the protomers are highly similar to each other, with low mean square deviation (rmsd) values. Some differences between FabK and SDR ENR are visible, such as the α-helical insertion domain near C terminus, that is not present in FabK (Saito et al., 2008). FabK shows a two-step ping-pong catalytic mechanism that couples the reduction of a flavin mononucleotide (FMN) to the oxidation of the NADH. The NADH acts indirectly reducing the flavin cofactor which will reduce the double bond of the enoyl-ACP subtract (Marrakchi et al., 2003; Saito et al., 2008; Massengo-Tiassé and Cronan, 2009; Kim et al., 2012). The structures of bacterial FabK from P. gingivalis (PgFabK) and T. maritima (TmFabK) were also elucidated. They share a very similar structure (Figure 12B), but some differences among FabK proteins can be found (Saito et al., 2008; Kim et al., 2012; Ha et al., 2017; Hevener et al., 2018). PgFabK show selectivity for NADPH over NADH, and it occurs because the 2′-ribosyl phosphate group of NADPH make electrostatic interactions with Lys263 and His46, while NADH binds the SpFabK at corresponding residues Ala267 and Pro47. TmFabK shows the same NADH preference as the S. pneumoniae enzyme (Saito et al., 2008; Ha et al., 2017; Hevener et al., 2018). The residues among three species (S. pneumoniae, P. gingivalis and T. maritima) are 34% identical and the active site similarity is significantly higher, with most of the differences found for PgFabK. The FMN-binding region from Ala188 to Phe195 is conserved among all the species, except for a methionine substitution in TmFabK at the position 191 (Saito et al., 2008; Hevener et al., 2018).

First described as a triclosan-resistant protein, organisms that present FabK are typically more resistant to TCL than the organisms that present FabI, being the fabK gene considered one TCL-resistant determinant mechanism, as well as fabV, fabL and acrB, an efflux pump (Heath and Rock, 2000a; Khan et al., 2018). The expression of FabK in E. coli can rise the inhibitory concentration of TCL by 4 times, evidencing the ability of FabK to confer resistance to TCL (Marrakchi et al., 2003; Massengo-Tiassé and Cronan, 2009). Although E. faecalis has both FabI and FabK, none of them mediate TCL resistance, being the fabK knockout viable without fatty acid supplementation and having no difference in the inherent TCL resistance (Zhu et al., 2010).

Some naturally occurring compounds were found to efficiently inhibit FabK, as Atromentin and Leucomelone, the first specific SpFabK inhibitors. Atromentin and Leucomelone inhibited this enzyme with IC₅₀ values of 0.24 and 1.57 μM, respectively, while not inhibiting the FabI of either E. coli or S. aureus (Zheng et al., 2006; Massengo-Tiassé and Cronan, 2009). Leucomelone showed the ability to inhibit SpFabK with Ki and Km values of 4.1×10⁻⁷ M and 1.9×10⁻⁴ M, respectively. However, neither of the compounds displayed antibacterial activity, possibly due to the activities of efflux pumps (Zheng et al., 2006). In addition to natural compounds, AG205 was described as an efficient FabK inhibitor, showing IC₅₀ of 1.5 μM, exhibiting antibacterial activity of 1–8 μg/ml against most of the tested S. pneumoniae isolates while no inhibition was observed against bacteria possessing FabI. This antibacterial activity was found to be bacteriostatic rather than bactericidal (Takahata et al., 2006).

Phenylimidazoles derivatives have also been studied as FabK inhibitors. These compounds inhibit recombinant FabK and are potent antibacterial agents against S. pneumoniae. Some of them are specific inhibitors for FabK, while others were found to be dual inhibitors, acting on both FabK and FabI (Kitagawa et al., 2007a,b). The most active specific-FabK phenylimidazole revealed an IC₅₀ value of 0.14 μM and a MIC value against S. pneumoniae of 0.5 μg/ml. On the other hand, the most active 4-pyridone derivative compound showed IC₅₀ of 0.38 ± 0.07 μM and 0.0045 ± 0.0007 μM for EcFabI and SpFabK, respectively. The MIC against S. aureus and S. pneumoniae of this same compound was >32 and 0.5 μg/ml, respectively (Kitagawa et al., 2007a,b). Phenylimidazoles also showed activity against a clinically dominant ribotype 027 C. difficile, with MIC values of 8 and 2 μM in the absence and presence of palmitic acid or fatty acid mix, respectively. The enzymatic inhibition of the same compound for C. difficile FabK was demonstrated with IC₅₀ of 3.31 ± 0.046 μM (Marreddy et al., 2019). The structure of SpFabK was elucidated in the presence of a phenylimidazole derivative (compound 1; Figure 12C; Saito et al., 2008). A conserved histidine (H144), suggested to play a role in catalysis as a proton donor/acceptor, is found to rotate upon complex formation with compound 1 (Figure 12C). As observed for the catalytic tyrosine in SDR ENRs, this rotation could be relevant in catalysis to position histidine toward the reactants, but the catalytic mechanism of FabKs was not studied in more detail yet.

In prokaryotes, aside from the Archaea domain, where FA synthesis pathways seem to be ACP-independent (Lombard et al., 2012), Enoyl-ACP reductase (ENR) orthologs are widespread across the genomic landscape of most bacteria. The first one to be identified, FabI, was thought to be the only bacterial ENR, due to it being widely distributed among bacteria, until the discovery of FabL in B. subtilis (Heath et al., 2000c) and FabK in S. pneumoniae (Marrakchi et al., 2003). While FabI and FabL origins are related, FabK is structurally related to a 2-nitropropane dioxygenase (NPD). FabV, a novel ENR reported in 2008 (Massengo-Tiassé and Cronan, 2008) shows low sequence identity to both FabI and FabL and, while it belongs to the same superfamily as them, it constitutes a new ENR class. FabV was also discovered in other bacteria afterward, such as B. mallei (Lu and Tonge, 2010) and P. aeruginosa (Zhu et al., 2010). In fact, more than 3,800 FabVs are deposited on NCBI RefSeq already. ENRs diversity and phylogeny are shown in the tree depicted in Figure 13, along with FabI, FabL, FabL2, and FabK. Some sequences were grouped within an ENR class different from its annotated one, which could be due to annotation errors.

In contrast to bacteria, eukaryotic fatty acid biosynthesis is catalyzed by a type I fatty acid system, where different enzyme activities are performed by the domains of a large polypeptide (FAS-I; Smith et al., 2003), aside from plants, which use the FAS-II system, just like bacteria. Eukaryotes have two considerably distinct type I FAS: the metazoan (Maier et al., 2006) and the fungal FAS (Jenni et al., 2006). The former is a 540 kDa homodimer comprising two sets of functional domains, whose structure is common to those of bacterial polyketide synthases (Smith and Tsai, 2007). Fungal FAS form a 2.6 megadalton structure with 48 domains, and may contain a phosphopantetheinyl transferase domain, necessary for attachment to ACP (Smith and Tsai, 2007; Grininger, 2014). Evolutionary origins for this fungal FAS can be traced back to the bacterial ENRs, as it seems to share a common ancestrality with FabK from S. pneumoniae and other bacterial ENRs, while forming a monophyletic sister group to CMN-FAS enoyl-reductases, which are common across the groups of Corynebacteria, Mycobacteria and Nocardia (Bukhari et al., 2014).

Even though FAS-I systems are the most common among eukaryotes, a recent study presents phylogenetic evidence of an ENR present in bacteria, lower eukaryotes from the group Apicomplexa, and some higher eukaryotes as well (Khan et al., 2019). This protein, designated as FabI2 due to its high sequence similarity with FabI, was identified during a soil metagenomics analysis (Khan et al., 2016), and seems to be predominant in Apicomplexa and intracellular pathogenic bacteria. These novel ENRs shared by both bacteria and eukaryotes can be traced back to a common ancestral ENR closely related to FabI (Khan et al., 2019).

Domains for the SDR superfamily are found within the sequences of FabI, FabI2, and FabL. While sharing a common Rossmann fold with other SDR ENRs, a search on the Conserved Domains Database shows that FabV contains its own domain: PRK13656. The novel FabI2, spread from bacteria to higher eukaryotes (Khan et al., 2019), also contains a domain for the Rossmann-fold NAD(P)(+)-binding protein superfamily, just like its FabI counterpart. This protein, obtained from metagenomic studies, despite sharing a relatively high homology and some structural similarity with FabI, contains a unique Yx₇K catalytic signature motif, which was thought to be specific to higher eukaryotes with plant origins (Massengo-Tiassé and Cronan, 2009; Khan et al., 2019). The most common catalytic signature among members of the SDR superfamily is Asn-x_n-Ser-x₁₀-Tyr-x₃-Lys. In M. tuberculosis, the catalytic signature of its FabI, which is annotated as InhA, is Phe-x₈-Tyr-x₆-Lys. These catalytic signatures vary considerably amidst members of the SDR superfamily and, while higher eukaryotes such as plants are also part of it, they contain a unique signature (Thr-x₁₀-Tyr-x₇-Lys) as well (Massengo-Tiassé and Cronan, 2009).

The evolution of mycobacterial ENRs is particularly interesting, in the sense that the genome of Mtb seems to encode more than 250 enzymes involved in the metabolism of fatty acids, a vast amount when compared to the 50 contained within the genome of E. coli (Cole et al., 1998), despite the size similarities between these two genomes. In Mtb, though, fatty acid metabolism is vital to allow survival of the bacterium while inside its host, as its cell wall contains a lipid layer derived from the production of mycolic acids which, in turn, result from elongated fatty acids, thus justifying the presence of such a wide array of FAS enzymes (Kinsella et al., 2003). The three main explanations for this are either the birth of genes, horizontal gene transfer, or gene duplication resulting in paralogues. Mtb also has one of the highest numbers of interkingdom gene fusions—19 Mtb genes have been proposed to have a eukaryotic origin (Wolf et al., 2000; Gamieldien et al., 2002). The mycobacterial InhA, though, does not show clear evidence of horizontal gene transfer, and there are few indications of paralogy across members of its family. The enzyme also does not seem to be the result of positive selection (Kinsella et al., 2003).