Triclosan is a broad-spectrum antibacterial agent that inhibits bacterial fatty acid synthesis at the enoyl-acyl carrier protein reductase (FabI) step. Resistance to triclosan inEscherichia coli is acquired through a missense mutation in the fabI gene that leads to the expression of FabI[G93V]. The specific activity and substrate affinities of FabI[G93V] are similar to FabI. Two different binding assays establish that triclosan dramatically increases the affinity of FabI for NAD+. In contrast, triclosan does not increase the binding of NAD+to FabI[G93V]. The x-ray crystal structure of the FabI-NAD+-triclosan complex confirms that hydrogen bonds and hydrophobic interactions between triclosan and both the protein and the NAD+ cofactor contribute to the formation of a stable ternary complex, with the drug binding at the enoyl substrate site. These data show that the formation of a noncovalent “bi-substrate” complex accounts for the effectiveness of triclosan as a FabI inhibitor and illustrates that mutations in the FabI active site that interfere with the formation of a stable FabI-NAD+-triclosan ternary complex acquire resistance to the drug. Triclosan is a broad-spectrum antibacterial agent that inhibits bacterial fatty acid synthesis at the enoyl-acyl carrier protein reductase (FabI) step. Resistance to triclosan inEscherichia coli is acquired through a missense mutation in the fabI gene that leads to the expression of FabI[G93V]. The specific activity and substrate affinities of FabI[G93V] are similar to FabI. Two different binding assays establish that triclosan dramatically increases the affinity of FabI for NAD+. In contrast, triclosan does not increase the binding of NAD+to FabI[G93V]. The x-ray crystal structure of the FabI-NAD+-triclosan complex confirms that hydrogen bonds and hydrophobic interactions between triclosan and both the protein and the NAD+ cofactor contribute to the formation of a stable ternary complex, with the drug binding at the enoyl substrate site. These data show that the formation of a noncovalent “bi-substrate” complex accounts for the effectiveness of triclosan as a FabI inhibitor and illustrates that mutations in the FabI active site that interfere with the formation of a stable FabI-NAD+-triclosan ternary complex acquire resistance to the drug. The fatty acid synthase system of Escherichia coli is the paradigm for the type II or dissociated fatty acid synthase systems (for reviews, see Refs. 1Rock C.O. Jackowski S. Cronan Jr., J.E. Vance D.E. Vance J.E. Biochemistry of Lipids, Lipoproteins, and Membranes. Elsevier Science Publishers B.V., Amsterdam1996: 35-74Google Scholar, 2Rock C.O. Cronan Jr., J.E. Biochim. Biophys. Acta. 1996; 1302: 1-16Crossref PubMed Scopus (291) Google Scholar, 3Cronan Jr., J.E. Rock C.O. Neidhardt F.C. Curtis R. Gross C.A. Ingraham J.L. Lin E.C.C. Low K.B. Magasanik B. Reznikoff W. Riley M. Schaechter M. Umbarger H.E. Escherichia coli and Salmonella typhimurium: Cellular and Molecular Biology. American Society for Microbiology, Washington, D. C.1996: 612-636Google Scholar). Distinct genes encode each of the individual enzymes, and the same basic chemical reaction is often catalyzed by multiple isozymes. There are four basic reactions that constitute a single round of elongation. The first step is the condensation of malonyl-ACP 1The abbreviations used are: ACP, acyl carrier protein; FabI, enoyl-acyl carrier protein reductase; triclosan, 2,4,4′-trichloro-2′-hydroxydiphenyl ether; IC50, concentration giving 50% inhibition of activity; 4:1-NAC, crotonyl-N-acetylcysteamine; 8:1-NAC, trans-2-octadecenoyl-N-acetylcysteamine. with either acetyl-CoA to initiate fatty acid synthesis (FabH) or with the growing acyl chain to continue cycles of elongation (FabB or FabF). The β-ketoacyl-ACP is reduced by an NADPH-dependent β-ketoacyl-ACP reductase (FabG). Only a single enzyme is responsible for this step (4Zhang Y. Cronan Jr., J.E. J. Bacteriol. 1998; 180: 3295-3303Crossref PubMed Google Scholar). There are two β-hydroxyacyl-ACP dehydrases (FabA and FabZ) capable of forming trans-2-enoyl-ACP. The product of the fabA gene is specifically involved in the introduction of a cis double bond into the growing acyl chain at the β-hydroxydecanoyl-ACP step and most efficiently catalyzes dehydration of short-chain β-hydroxyacyl-ACPs, whereas the FabZ dehydratase has a broader substrate specificity (5Heath R.J. Rock C.O. J. Biol. Chem. 1996; 271: 27795-27801Abstract Full Text Full Text PDF PubMed Scopus (211) Google Scholar). The last reaction in each elongation cycle is catalyzed by enoyl-ACP reductase (FabI). Contrary to the initial conclusion that there were two enoyl-ACP reductases, based on assays in crude extracts (6Weeks G. Wakil S.J. J. Biol. Chem. 1968; 243: 1180-1189Abstract Full Text PDF PubMed Google Scholar), E. coli cells possess only a single NADH-dependent enoyl-ACP reductases encoded by the fabI gene that utilizes all chain lengths (7Heath R.J. Rock C.O. J. Biol. Chem. 1995; 270: 26538-26542Abstract Full Text Full Text PDF PubMed Scopus (318) Google Scholar). The importance of fatty acid biosynthesis to cell growth and function makes this pathway an attractive target for the development of antibacterial agents. Two important control points in the cycle are the condensing enzymes and the enoyl-ACP reductase (8Heath R.J. Rock C.O. J. Biol. Chem. 1996; 271: 1833-1836Abstract Full Text Full Text PDF PubMed Scopus (196) Google Scholar, 9Heath R.J. Rock C.O. J. Biol. Chem. 1996; 271: 10996-11000Abstract Full Text Full Text PDF PubMed Scopus (193) Google Scholar), and both reactions are targeted by compounds that effectively inhibit fatty acid synthesis. Two natural products, cerulenin and thiolactomycin, are potent antibiotics that function by specifically inhibiting the condensing enzyme reactions (for reviews, see Refs. 3Cronan Jr., J.E. Rock C.O. Neidhardt F.C. Curtis R. Gross C.A. Ingraham J.L. Lin E.C.C. Low K.B. Magasanik B. Reznikoff W. Riley M. Schaechter M. Umbarger H.E. Escherichia coli and Salmonella typhimurium: Cellular and Molecular Biology. American Society for Microbiology, Washington, D. C.1996: 612-636Google Scholar, 10Jackowski S. Sutcliffe J.A. Georgopapadakou N.H. Emerging Targets for Antibacterial and Antifungal Chemotherapy. Chapman and Hall, New York1991: 151-162Google Scholar). The diazaborines, a class of heterocyclic antibacterials, inhibit fatty acid biosynthesis by blocking the FabI step (11Bergler H. Wallner P. Ebeling A. Leitinger B. Fuchsbichler S. Aschauer H. Kollenz G. Högenauer G. Turnowsky F. J. Biol. Chem. 1994; 269: 5493-5496Abstract Full Text PDF PubMed Google Scholar). Resistance to the diazaborines arises from a missense mutation in the fabIgene that leads to the expression of a FabI[G93S] mutant protein (12Turnowsky F. Fuchs K. Jeschek C. Högenauer G. J. Bacteriol. 1989; 171: 6555-6565Crossref PubMed Google Scholar,13Bergler H. Högenauer G. Turnowsky F. J. Gen. Microbiol. 1992; 138: 2093-2100Crossref PubMed Scopus (55) Google Scholar). Similarly, the fabI analog in Mycobacterium tuberculosis, the inhA gene, encodes a cellular target for isoniazid and ethionamide. A point mutation in the inhAgene confers resistance to the drugs (14Banerjee A. Dubnau E. Quémard A. Balasubramanian V. Um K.S. Wilson T. Collins D. de Lisle G. Jacobs Jr., W.R. Science. 1994; 263: 227-230Crossref PubMed Scopus (1236) Google Scholar). Both isoniazid and diazaborine bind at the substrate site of the respective enoyl-ACP reductases and covalently react with NAD+ to form tight binding bi-substrate complexes (15Baldock C. Rafferty J.B. Sedeinikova S.E. Baker P.J. Stuitje A.R. Slabas A.R. Hawkes T.R. Rice D.W. Science. 1996; 274: 2107-2110Crossref PubMed Scopus (215) Google Scholar, 16Rozwarski D. Grant G. Barton D. Jacobs W. Sacchettini J.C. Science. 1998; 279: 98-102Crossref PubMed Scopus (618) Google Scholar). Triclosan is a broad-spectrum antibacterial agent that enjoys widespread applications in a multitude of contemporary consumer products including, soaps, detergents, toothpastes, skin care products, cutting boards, and mattress pads (17Bhargava H.N. Leonard P.A. Am. J. Infect. Control. 1996; 24: 209-218Abstract Full Text PDF PubMed Scopus (441) Google Scholar). Triclosan is widely thought to be a nonspecific biocide that attacks bacterial membranes (18Regös J. Zak O. Solf R. Vischer W.A. Weirich E.G. Dermatologica (Basel). 1979; 158: 72-79Crossref PubMed Scopus (80) Google Scholar, 19Vischer W.A. Regös J. Zentbl. Bakt. Hyg. I. Abt. Orig. A. 1974; 226: 376-389PubMed Google Scholar, 20Regös J. Hitz H.R. Zentbl. Bakt. Hyg. I. Abt. Orig. A. 1974; 226: 390-401PubMed Google Scholar), and if triclosan does not have a discrete mechanism of action, the acquisition of cellular resistance is unlikely. However, recent work reveals that resistant E. coli strains arise from missense mutations in the fabIgene (21McMurray L.M. Oethinger M. Levy S. Nature. 1998; 394: 531-532Crossref PubMed Scopus (838) Google Scholar, 22Heath R.J. Yu Y.-T. Shapiro M.A. Olson E. Rock C.O. J. Biol. Chem. 1998; 273: 30316-30320Abstract Full Text Full Text PDF PubMed Scopus (323) Google Scholar) and that triclosan and other 2-hydroxydiphenyl ethers directly inhibit fatty acid biosynthesis in vivo and FabI catalysis in vitro (22Heath R.J. Yu Y.-T. Shapiro M.A. Olson E. Rock C.O. J. Biol. Chem. 1998; 273: 30316-30320Abstract Full Text Full Text PDF PubMed Scopus (323) Google Scholar). The goal of this study was to investigate the mechanism by which triclosan inhibits FabI. Sources of supplies were: American Radiochemicals, Inc., [adenine-2,8-3H]NAD (specific activity 25 Ci/mmol); Sigma, NADH and NAD+; and Millipore Corp., Ultrafree Probind centrifugal filtration units. The substrate analogs, 4:1-NAC and 8:1-NAC, were the generous gift of Rocco Gogliotti and John Domagala (Parke-Davis). Triclosan was the gift of KIC Chemicals, Inc. (Armonk, NY). All other chemicals were of the best grade available. The expression and purification of His-tag FabI was described previously (7Heath R.J. Rock C.O. J. Biol. Chem. 1995; 270: 26538-26542Abstract Full Text Full Text PDF PubMed Scopus (318) Google Scholar). ThefabI G93V allele was amplified from the chromosome of the triclosan-resistant strain RJH108 (22Heath R.J. Yu Y.-T. Shapiro M.A. Olson E. Rock C.O. J. Biol. Chem. 1998; 273: 30316-30320Abstract Full Text Full Text PDF PubMed Scopus (323) Google Scholar) and ligated into the pET15b expression vector. DNA sequencing of the expression construct confirmed the presence of the single point mutation. The His-tagged FabI[G93V] was then expressed and purified to homogeneity by Ni2+ affinity chromatography using the same methods as described for the wild-type FabI (7Heath R.J. Rock C.O. J. Biol. Chem. 1995; 270: 26538-26542Abstract Full Text Full Text PDF PubMed Scopus (318) Google Scholar). Protein was determined by the method of Bradford (23Bradford M.M. Anal. Biochem. 1976; 72: 248-254Crossref PubMed Scopus (217544) Google Scholar). FabI activity was assayed spectrophotometrically by monitoring the decrease in absorption at 340 nm using an adaptation of the spectrophotometric assay used by Bergleret al. (24Bergler H. Fuchsbichler S. Högenauer G. Turnowsky F. Eur. J. Biochem. 1996; 242: 689-694Crossref PubMed Scopus (133) Google Scholar). Standard reactions contained 100 μm 8:1-NAC, 12 μg of homogeneous FabI (7Heath R.J. Rock C.O. J. Biol. Chem. 1995; 270: 26538-26542Abstract Full Text Full Text PDF PubMed Scopus (318) Google Scholar, 8Heath R.J. Rock C.O. J. Biol. Chem. 1996; 271: 1833-1836Abstract Full Text Full Text PDF PubMed Scopus (196) Google Scholar), 200 μm NADH, 0.1 m sodium phosphate, pH 7.5, in a final volume of 300 μl. The reactions were performed at 24 °C in semimicro quartz cuvettes. The change in optical density was continuously monitored for 1 min, and the reaction rate was calculated from the slope of the trace. Triclosan was added to the final concentrations indicated in the figure legend from serially diluted stock solutions in Me2SO. The Me2SO concentration in all assays was maintained at 1.66%, which did not significantly affect FabI activity. The FabI specific activity in the absence of drugs under these experimental conditions was 0.36 μmol/min/mg. Data points were the mean of duplicate assays, and the individual values were within ± 5% of the average. Two methods were used to measure the binding of NAD+ to FabI in the presence and absence of triclosan. The first method employed gel filtration chromatography to separate [3H]NAD+ bound to FabI from free [3H]NAD+. A 0.7 × 8-cm column (3 ml) of Sephadex G-25 was equilibrated in 0.1 m sodium phosphate, pH 7.5, at 24 °C. Samples (50 μl) were applied, the column eluted with the same buffer at a flow rate of approximately 0.25 ml/min, and 5-drop (110 μl) fractions were collected. The second method used binding of FabI to a polyvinylidene difluoride membrane to separate free from FabI-bound [3H]NAD+. Assays were placed into a Ultrafree-Probind centrifugal filtration unit containing a 0.2-cm (2Rock C.O. Cronan Jr., J.E. Biochim. Biophys. Acta. 1996; 1302: 1-16Crossref PubMed Scopus (291) Google Scholar), 0.45-μm polyvinylidene difluoride membrane (Millipore, Corp.). The filters were washed once with 100 μl of 0.1 msodium phosphate and counted in 5 ml of ScintSafe scintillation solution. Both assays contained 0.1 m sodium phosphate, 1 μm [3H]NAD+ (specific activity 4 Ci/mmol), 2% Me2SO, the indicated concentrations of triclosan, and 7.6 μg of FabI or FabI[G93V] in a final volume of 50 μl. Samples were incubated at 24 °C for 15 min and then separated using one of the two methods outlined above. Crystals of FabI were grown from hanging drops (6 μl) containing 5 mg/ml FabI, 25 mm Tris, 150 mm NaCl, 0.5 mm dithiothreitol, 5 mg/ml NADH, 50 mm sodium acetate, 4% polyethylene glycol 4000 (w/v), pH 4.6. The drops were equilibrated against a reservoir (600 μl) containing 100 mm sodium acetate, pH 4.6, and 8% polyethylene glycol 4000 (w/v). Large hexagonal bipyrimidal crystals appeared after equilibration for 1 day at 26 °C and reached dimensions of 1.0 × 0.5 × 0.5 mm. Crystals of the FabI-NAD+-triclosan ternary complex were prepared by soaking pregrown crystals of FabI in solutions containing 10 mg/ml NAD+, 10 mg/ml triclosan, 100 mmsodium acetate, pH 4.6, and 10% polyethylene glycol 4000 (w/v) for 7 days. X-ray diffraction data were collected on the IMCA beamline (17-ID) at the Advanced Photon Source at the Argonne National Laboratory. The crystals used for data collection were first transferred to a cryoprotectant buffer containing 100 mmsodium acetate, pH 4.6, 10% polyethylene glycol 4000 (w/v), 25% glycerol (v/v) for 1 h and then flash-frozen in a stream of liquid nitrogen. X-ray data to 1.8-Å resolution were collected at 100 K using synchrotron X-radiation at a 1.0-Å wavelength and a Bruker 2 × 2 mosaic CCD x-ray detector. The x-ray data were then processed using the SAINT (Bruker x-ray, Madison, WI) program package. Crystals of the FabI-NAD+-triclosan complex were hexagonal, space group P6122 (number 178) with unit cell dimensionsa = b = 79.70 and c = 327.7 Å and two molecules of the ternary complex in the asymmetric unit. These crystals were isomorphous to those of the FabI-NAD+-diazaborine complex reported by Baldock et al. (15Baldock C. Rafferty J.B. Sedeinikova S.E. Baker P.J. Stuitje A.R. Slabas A.R. Hawkes T.R. Rice D.W. Science. 1996; 274: 2107-2110Crossref PubMed Scopus (215) Google Scholar). The FabI structure was initially refined as a rigid body using XPLOR, and the positions and conformation of the bound NAD+ and triclosan molecules were determined from Fo-Fc difference electron density maps. In addition, the positions of the 185 tightly bound water molecules were determined from subsequent residual electron density maps. The entire structure containing 2 molecules of the FabI-NAD+-triclosan complex and 185 water molecules was refined to a crystallographic R-factor of 0.232 using data from 8.0 to 1.8 Å resolution. Residues 194–210 were disordered in the structure. Solvent-accessible surface areas were calculated using a 1.4-Å probe with the SOLVATION module of the InsightII software (version 97.1, Molecular Simulations, Inc.). Previous work showed that FabI was inhibited by triclosan (22Heath R.J. Yu Y.-T. Shapiro M.A. Olson E. Rock C.O. J. Biol. Chem. 1998; 273: 30316-30320Abstract Full Text Full Text PDF PubMed Scopus (323) Google Scholar) and that a mutation in the fabI gene that results in the expression of a FabI[G93V] protein exhibited high level triclosan resistance (21McMurray L.M. Oethinger M. Levy S. Nature. 1998; 394: 531-532Crossref PubMed Scopus (838) Google Scholar, 22Heath R.J. Yu Y.-T. Shapiro M.A. Olson E. Rock C.O. J. Biol. Chem. 1998; 273: 30316-30320Abstract Full Text Full Text PDF PubMed Scopus (323) Google Scholar). FabI[G93V] was cloned into the pET15b expression vector and expressed and purified as described under “Experimental Procedures.” The apparent K mvalues for the two substrates were determined for both the wild-type and mutant protein. The wild-type protein had an apparentK m for NADH of 15 ± 1 μm, whereas the FabI[G93V] exhibited an apparent K mfor NADH of 8 ± 2 μm. Thus, there were not major differences between the kinetic constants for NADH in the wild-type and mutant protein. We also measured the apparent K m for the enoyl substrate using 4:1-NAC. This substrate analog was selected for these experiments because it had the higher solubility in the assay buffer than 8:1-NAC. The apparent K m for 4:1-NAC was 25 ± 5 mm for FabI compared with 15 ± 2.2 mm for FabI[G93V]. The V max for FabI was 4.4 μmol/min/mg, and for FabI[G93V], theV max was 3.6 μmol/min/mg. These data illustrated that FabI and FabI[G93V] had similar specific activities and substrate and cofactor binding characteristics. The observation that cells expressing the FabI[G93V] protein have a minimum inhibitory concentration for triclosan that was 64-fold higher than the wild-type strain (0.5 compared with 32 μg/ml) (22Heath R.J. Yu Y.-T. Shapiro M.A. Olson E. Rock C.O. J. Biol. Chem. 1998; 273: 30316-30320Abstract Full Text Full Text PDF PubMed Scopus (323) Google Scholar) suggested that the FabI[G93V] protein would be refractory to triclosan inhibitionin vitro. We directly tested this assumption using the spectrophotometric FabI assay and found that both FabI and FabI[G93V] were inhibited by triclosan (Fig. 1). The IC50 for FabI (2 μm) was lower than the IC50 for FabI[G93V] (10 μm). This ≈5-fold increase in the IC50 for FabI[G93V] was consistent with increased resistance of the enzyme to triclosan inhibition; however, the magnitude of this difference was significantly less than was predicted from the 64-fold difference in the effectiveness of triclosan against the mutant strain. This unexpected discrepancy between the minimum inhibitory concentration and IC50 data led us to examine the mechanism of FabI inhibition by triclosan in more detail. The data in Fig. 1 was obtained by starting the assay with the addition of FabI and measuring the initial rate of NADH oxidation from the linear portion of the trace within the first 1 min of the reaction. Examination of the time course of the reaction over a longer period of time (15 min) revealed a clear difference between the inhibition of FabI and FabI[G93V] by triclosan (Fig. 2). Although there was an initial burst of activity in the presence of triclosan, within 3 min of the start of the experiment, FabI catalysis essentially ceased (Fig.2 A). Thus, the inhibition of FabI by triclosan became irreversible as a function of time. In contrast, triclosan inhibition of FabI[G93V] was consistent throughout the 15-min time course, yielding a slower rate and no indication of increasing inhibition as a function of time (Fig. 2 B). These data suggested that the difference between the effectiveness of triclosan against wild-type strains compared with strain RJH108 (fabI G93V) was the ability of triclosan to irreversibly inhibit FabI, but not FabI[G93V], as a function of time. The data in Figs. 1 and 2 suggested that the irreversible inhibition of FabI by triclosan arose from the tight binding of the drug to the enzyme in conjunction with another component in the assay. This prediction was experimentally tested by measuring the association of [3H]NAD+ with either FabI or FabI[G93V] in the presence of triclosan by gel filtration chromatography (Fig.3). FabI bound approximately 80% of the [3H]NAD+ (0.2 mol/mol of FabI subunit) when triclosan was present, as evidenced from the shift in radioactivity from the included volume (fractions 18–40) to the excluded volume (fractions 11–17) of the column. In contrast, FabI[G93V] did not displace [3H]NAD+ from the included to the void volume of the column. The column was not equilibrated with triclosan; thus the appearance of label associated with FabI in the excluded volume indicated a stable association with the enzyme during the course of the separation (about 30 min). The binding of [3H]NAD+ to the two enzymes in the absence of triclosan was barely detectable and equivalent to the binding of [3H]NAD+ to FabI[G93V] in the presence of triclosan shown in Fig. 3 (not shown, see below). These data demonstrated that triclosan induced the high affinity binding of NAD+ to FabI but not to FabI[G93V]. This concept was experimentally verified using a filter binding assay developed to quantitate the binding of [3H]NAD+ to FabI in the presence and absence of triclosan. FabI exhibited saturable NAD+ binding in the presence of 4 μm triclosan (Fig.4 A). An apparent association constant (K) of 2.5 μm–1 for NAD+ in the presence of triclosan was calculated by analysis of the data using a double-reciprocal plot (Fig.4 B). There was no indication of cooperative binding or the existence of more than a single class of sites by analysis of the data using a Scatchard plot (not shown). In contrast, we detected virtually no binding of [3H]NAD+ to FabI in the absence of triclosan (Fig. 4 A). The affinity of FabI for NAD+ was so low that we measured only trace levels of [3H]NAD+ binding to the free enzyme by this technique. Therefore, we examined the effectiveness of NAD+as a product inhibitor of FabI using the spectrophototmetric assay. NAD+ was a poor inhibitor of the FabI reaction, with 50% inhibition occurring between 10 and 20 mm NAD+. These data indicated a very low apparent affinity of FabI for NAD+ in the 10–50 mm range (not shown). The filter binding assay was next used to examine the effect of triclosan concentration on NAD+ binding to FabI (Fig.5). The association of [3H]NAD+ with FabI was dependent on the concentration of triclosan and saturable (Fig. 5 A). The apparent K for triclosan binding was calculated as 4.2 μm–1 (Fig. 5 B). Thus, in the presence of triclosan, the affinity of FabI for NAD+increased by approximately 4 orders of magnitude (>10 mmto 2.5 μm–1).Figure 5Triclosan-dependent [3H]NAD+;1 binding to FabI. Panel A, dependence of [3H]NAD+binding to FabI on the concentration of triclosan. Panel B,double-reciprocal plot of the data in panel A. The apparent triclosan association constant (K) was 4.2 μm. The filter binding assay was performed as described under “Experimental Procedures.”View Large Image Figure ViewerDownload (PPT) The binding studies suggested that the basis for the potency of triclosan as a FabI inhibitor was because of the formation of a high affinity FabI-NAD+-triclosan ternary complex. Determining the structure of the FabI-NAD+-triclosan ternary complex by x-ray crystallography directly tested this idea (see “Experimental Procedures”). Triclosan bound noncovalently in a location adjacent to the nicotinamide portion of NAD+ (Fig.6 A). There was 494.7 Å2 of surface area buried between triclosan and the protein and 238.5 Å2 of surface area buried at the interface with NAD+. The diphenyl ether of triclosan adopted a conformation with a dihedral angle of about 90° between the two phenyl rings of the inhibitor. The 2-hydoxy-3-chlorophenyl ring formed a parallel stack with the nicotinamide ring of the co-factor with an interplanar stacking distance of 3.4 Å. In addition to the nicotinamide ring, the hydroxychlorophenyl ring was surrounded by hydrophobic side chains on the protein including Tyr-146 and Tyr-156. The 2′-hydroxyl group of triclosan was involved in two strong hydrogen bonds. One bond was with the 2′-hydroxyl group of the nicotinamide ribose, and the second was to the phenolic hydroxyl group of Tyr-156. The 2,4-dichlorophenyl ring of triclosan was rotated 90° of the plane of the hydroxychlorophenyl group. The chlorine substituent in the 4-position of the phenyl ring accepted a hydrogen bond from the backbone amide of Ala-95 and formed hydrophobic contacts with the side chain of Met-159. These data supported the concept that the formation of a stable, ternary FabI-NAD+-triclosan complex accounted for the effective inhibition of the FabI reaction by this drug. The formation of a ternary FabI-NAD+-triclosan complex accounts for the effectiveness of triclosan as an antibacterial agent. Triclosan binds to the enoyl substrate site on FabI, and tight binding of the drug requires interactions between both the protein and the NAD+ cofactor (Fig. 6 A). There is a strong similarity between the mode of triclosan NAD+ binding to FabI (Fig. 6 A) and the structure of the FabI-NAD+-diazaborine complex described by Baldock et al. (15Baldock C. Rafferty J.B. Sedeinikova S.E. Baker P.J. Stuitje A.R. Slabas A.R. Hawkes T.R. Rice D.W. Science. 1996; 274: 2107-2110Crossref PubMed Scopus (215) Google Scholar) (Fig. 6 B). The diazaborine compounds form covalent adducts with NAD+ at the FabI active site. The boron atom of the diazaborine is attached to the 2′-hydroxyl of the NAD+ ribose, and this is the same 2′-hydroxyl group that forms a hydrogen bond with the hydroxychlorophenyl hydroxyl of triclosan. The FabI-NAD+-triclosan ternary complexes can be superimposed with the diazaborine complex with a root mean square deviation of 0.3 Å for the observed α-carbons. Residues 194–210 were not well defined in the electron density map of the FabI-NAD+-triclosan complex. These residues are also not observed in the native structure of FabI, but the positions of these residues are evident in the FabI-NAD+-diazaborine structure (15Baldock C. Rafferty J.B. Sedeinikova S.E. Baker P.J. Stuitje A.R. Slabas A.R. Hawkes T.R. Rice D.W. Science. 1996; 274: 2107-2110Crossref PubMed Scopus (215) Google Scholar). If the loop seen in the FabI-NAD+-diazaborine complex is modeled onto the triclosan complex structure, there are no conflicting contacts between triclosan and the protein. In this model Ile-200 is in van der Waals contact with the 2,4-dichlorophenyl ring of triclosan, and Phe-203 is in contact with the hydroxychlorophenyl group. These data support the conclusion that triclosan forms a ternary complex that acts as a dead-end inhibitor that effectively removes protein from the catalytic cycle. These results lead to the general conclusion that all 2-hydroxydiphenyl ether FabI inhibitors (22Heath R.J. Yu Y.-T. Shapiro M.A. Olson E. Rock C.O. J. Biol. Chem. 1998; 273: 30316-30320Abstract Full Text Full Text PDF PubMed Scopus (323) Google Scholar) block enzyme activity through the formation of noncovalent bisubstrate complexes with NAD+. Missense mutations, like FabI[G93V], confer resistance by preventing the formation of the FabI-NAD+-triclosan ternary complex. Strains expressing either FabI[G93S] and FabI[G93V] have significantly elevated minimum inhibitory concentrations for triclosan (21McMurray L.M. Oethinger M. Levy S. Nature. 1998; 394: 531-532Crossref PubMed Scopus (838) Google Scholar, 22Heath R.J. Yu Y.-T. Shapiro M.A. Olson E. Rock C.O. J. Biol. Chem. 1998; 273: 30316-30320Abstract Full Text Full Text PDF PubMed Scopus (323) Google Scholar). Although FabI[G93V] activity is still inhibited at high triclosan concentrations (Fig. 2 B), the drug is not capable of forming a stable complex with the reductase. Triclosan is neither a dead-end inhibitor of FabI[G93V] (Fig. 2) nor does it form a FabI[G93V]-NAD+-triclosan ternary complex based on the inability of FabI[G93V] to bind [3H]NAD+(Fig. 3). These biochemical results are interpreted by modeling the Val-93 into the Gly-93 position in the structure of the triclosan complex. Gly-93 lies on one side of the substrate binding pocket, and the branched hydrophobic side chain of Val protrudes into the pocket and occupies the same space as the dichlorophenyl ring of triclosan (see Fig. 6). Therefore, we conclude that the substitution of Gly-93 with bulkier amino acid residues imparts resistance to 2-hydroxydiphenyl ethers because of steric interference. Our identification of a ternary FabI-NAD+-triclosan complex reinforces the conclusion that FabI is a specific intracellular target for triclosan. Triclosan permeabilizes the bacterial envelope (18Regös J. Zak O. Solf R. Vischer W.A. Weirich E.G. Dermatologica (Basel). 1979; 158: 72-79Crossref PubMed Scopus (80) Google Scholar, 19Vischer W.A. Regös J. Zentbl. Bakt. Hyg. I. Abt. Orig. A. 1974; 226: 376-389PubMed Google Scholar, 20Regös J. Hitz H.R. Zentbl. Bakt. Hyg. I. Abt. Orig. A. 1974; 226: 390-401PubMed Google Scholar), and these findings have justified the widespread and increasing use of triclosan in consumer personal care products based on the reasoning that bacteria would not acquire resistance to a nonspecific membrane disrupter. However, this type of evidence is not sufficient to conclude that a compound has nonspecific membrane effects, because agents that block fatty acid biosynthesis perturb membrane assembly by stopping phospholipid production. For example, the temperature-sensitivefabI mutant was initially designated envM because it was isolated through a genetic selection for strains with envelope-permeability defects (25Egan A.F. Russell R.R.B. Genet. Res. 1973; 21: 3603-3611Crossref Scopus (30) Google Scholar). Also, the diazaborine class of FabI inhibitors are known to perturb membrane function as well (12Turnowsky F. Fuchs K. Jeschek C. Högenauer G. J. Bacteriol. 1989; 171: 6555-6565Crossref PubMed Google Scholar). The first evidence that triclosan has a specific cellular target was provided by McMurray et al. (21McMurray L.M. Oethinger M. Levy S. Nature. 1998; 394: 531-532Crossref PubMed Scopus (838) Google Scholar) who found that triclosan-resistant strains have missense mutations in thefabI gene. The biochemical analysis of FabI inhibition by triclosan (22Heath R.J. Yu Y.-T. Shapiro M.A. Olson E. Rock C.O. J. Biol. Chem. 1998; 273: 30316-30320Abstract Full Text Full Text PDF PubMed Scopus (323) Google Scholar) and the delineation of its mechanism of binding (this study) provide definitive evidence that enoyl-ACP reductase is the primary site for triclosan action. Thus, the reported effects of triclosan on membrane structure and function are a consequence of its specific inhibition of fatty acid biosynthesis at the FabI step. The ability of E. coli to acquire genetic resistance to triclosan and related compounds through missense mutations in thefabI gene suggests that the widespread use of this drug will lead to the appearance of resistant organisms that will compromise the usefulness of triclosan. The ubiquitous occurrence of type II fatty acid synthase systems in bacteria and the essential nature of the FabI reaction make this enzyme an attractive target for antibacterial drugs. Accordingly, triclosan is effective against a broad spectrum of bacteria (17Bhargava H.N. Leonard P.A. Am. J. Infect. Control. 1996; 24: 209-218Abstract Full Text PDF PubMed Scopus (441) Google Scholar), including multi-drug-resistant Staphylococcus aureus (26Bartzokas C.A. Paton J.H. Gibson M.F. Graham F. McLoughlin G.A. Croton R.S. N. Engl. J. Med. 1984; 311: 1422-1425Crossref PubMed Scopus (103) Google Scholar, 27Webster J. J. Hosp. Infect. 1992; 21: 137-141Abstract Full Text PDF PubMed Scopus (43) Google Scholar). The design and development of second generation FabI inhibitors based on their ability to form ternary FabI-NAD+-drug complexes will supplement the arsenal against a broad spectrum of bacteria. We thank Amy Sullivan and Magdalena Kaminska for technical assistance, Rocco Gogliotti and John Domagala for the trans-2-acyl-NAC substrate analogs, Tom Mueller and Craig Banotai for the protein used in the crystal structure studies, Steve VanderRoestand for the fabI clone, and our colleagues for their help in editing the manuscript.