Structural Elucidation of the Specificity of the Antibacterial Agent Triclosan for Malarial Enoyl Acyl Carrier Protein Reductase*

The human malaria parasite Plasmodium falciparum synthesizes fatty acids using a type II pathway that is absent in humans. The final step in fatty acid elongation is catalyzed by enoyl acyl carrier protein reductase, a validated antimicrobial drug target. Here, we report the cloning and expression of the P. falciparum enoyl acyl carrier protein reductase gene, which encodes a 50-kDa protein (PfENR) predicted to target to the unique parasite apicoplast. Purified PfENR was crystallized, and its structure resolved as a binary complex with NADH, a ternary complex with triclosan and NAD+, and as ternary complexes bound to the triclosan analogs 1 and 2 with NADH. Novel structural features were identified in the PfENR binding loop region that most closely resembled bacterial homologs; elsewhere the protein was similar to ENR from the plant Brassica napus (root mean square for Cαs, 0.30 Å). Triclosan and its analogs 1 and 2 killed multidrug-resistant strains of intra-erythrocytic P. falciparum parasites at sub to low micromolar concentrations in vitro. These data define the structural basis of triclosan binding to PfENR and will facilitate structure-based optimization of PfENR inhibitors.

Treatment of Plasmodium falciparum malaria has depended for decades on the use of the aminoquinoline chloroquine or the synergistic antifolate combination pyrimethamine-sulfadoxine. These drugs were characterized by their potency against the P. falciparum asexual intra-erythrocytic stages (responsible for malaria pathogenesis), their affordability and their safety. The occurrence and spread of drug-resistant strains of P. falciparum have largely contributed to a recent resurgence of malaria, which claims 1 to 3 million lives annually and which is endemic in inter-tropical areas representing 40% of the world's population (1). The current situation of antimalarial chemotherapy and resistance, in conjunction with the reappearance of malaria in formerly well-controlled areas, reinforces the need for new, highly potent antimalarials.
Recent investigations into Apicomplexan parasites, including Plasmodium and Toxoplasma, have uncovered the exist-ence of a unique organelle, the apicoplast (2)(3)(4). The finding that ciprofloxacin-mediated inhibition of plastid replication in Toxoplasma gondii tachyzoites blocked parasite propagation provided evidence for the indispensable nature of this nonphotosynthetic plastid organelle (5). Studies of plastid inhibitors and apicoplast mis-segregation mutants confirmed the essential requirement of this organelle for normal parasite development and indicated that inhibition of apicoplast function or loss of this organelle resulted in parasite death following reinvasion of host cells (5)(6)(7). This organelle appears to derive ultimately from a cyanobacterial endosymbiont (4,8,9) and as such was postulated to contain prokaryotic-type metabolic pathways, of significant interest from the perspective of developing antiparasitic drugs (10). Recent studies indicate that these pathways include fatty acid and isoprenoid biosynthesis (11,12). Proteins involved in these pathways are often the products of nuclear genes that encode N-terminal signal and transit peptide sequences for apicoplast localization (12)(13)(14).
Fatty acids play a critical role in providing metabolic precursors of biological membranes and represent an important form of metabolic energy, making their biosynthetic pathway an excellent target for antimicrobial agents. In higher eukaryotes and yeast the biosynthetic enzymes are integrated into large multifunctional single polypeptides, commonly referred to as type I fatty acid synthases (FAS-I). 1 The FAS-I system utilizes acetyl CoA for iterative 2-carbon elongation of fatty acids. In contrast to the large eukaryotic FAS-I enzyme, plants and most prokaryotes perform the same enzymatic steps using separate, discrete enzymes. This system is referred to as type II fatty acid synthase (FAS-II) (15)(16)(17)(18). The first evidence in favor of a FAS-II pathway in malaria parasites (see Scheme 1) came from the work of Waller et al. (12). These investigators reported the presence of nuclear genes encoding the FAS-II proteins acyl carrier protein (ACP), and FabZ (␤-hydroxyacyl-ACP dehydratase) in Toxoplasma gondii and ACP, FabH (␤-ketoacyl-ACP synthase III), and FabF (␤-ketoacyl-ACP synthase II) in P. falciparum and provided evidence for their targeting to the apicoplast (12,14). Recent in vitro studies confirm that P. falciparum actively synthesizes fatty acids, predominantly C10 to C14 (19).
Inhibition of fatty acid biosynthesis has been repeatedly validated as an appropriate target for antimicrobials. Specific inhibitors of the FAS-II pathway include triclosan and thiolactomycin. Triclosan, a specific inhibitor of FAS-II trans-2-enoyl-ACP reductase (ENR, also known as inhA or FabI) is effective against a broad spectrum of bacteria (20), including Escherichia coli (21, 22), mycobacteria (23), and multidrug-resistant Staphylococcus aureus (24,25) and is widely used as an antimicrobial in household formulations, including soaps and toothpaste. Recently, triclosan was found to inhibit P. falciparum growth with an IC 50 of ϳ1 M (19,26). This compares with a reported IC 50 against P. falciparum of about 50 M for thiolactomycin, which inhibits the condensing enzymes FabB, FabF, and FabH in plants and bacteria (Ref. 12 and references therein). In vivo efficacy studies using Plasmodium berghei in mice showed that subcutaneous administration of 3 mg/kg triclosan for 4 days resulted in 75% reduction in parasitemia (19). Full parasite clearance was achieved with a single injection of 38 mg/kg given to mice that already had a parasitemia of 13-27%. Liver and kidney function tests were normal even at this highest dose, indicating that the further development of triclosan and its analogs may result in pharmacologically suitable compounds for use in humans.
In this report, we present the cloning of the pfenr gene and the three-dimensional structure of its translation product, PfENR, bound to triclosan and analogs that show biological activity. These data provide a framework for understanding the inhibitory mechanisms of fatty acid biosynthesis of P. falciparum and a model for undertaking structure-based drug development of selective FAS-II antimalarials.
Full-length pfenr gene was amplified using primers W1 (5Ј-AACGT-CCCATGGATAAAATATCACAACGGTTATTATTCCTCTTTCTACAT) and W2 (5Ј-ATATGGATCCTCATTCATTTTCATTGCGATATATATCA-TCTGGTAAAAACAT), which contain NcoI and BamHI sites, respectively (underlined). Four silent mutations (shown in lowercase letters) were introduced with mutagenic primers M1 (5Ј-GAgAAGGAAGAgAA-gAAgAATTCAGCTAGCCAAAATTATACATTTATAGATTAT and M2 (5Ј-GAATTcTTcTTcTCTTCCTTcTCACCTGAATTGTTCATAATATTA-TGAACATC) using a two-step megaprimer PCR method (27,28). In the first step, the cDNA library was used as a template to amplify both a 5Ј fragment with the primers W1 and M2 and a 3Ј fragment with the primers M1 and W2. Both reactions used the PCR conditions: 1ϫ (94°C for 2 min); 30ϫ (94°C for 20 s, 53°C for 40 s, and 60°C for 3 min). For the second step, both fragments were gel-purified and combined in a PCR reaction with primers W1 and W2, yielding the full-length pfenr. After restriction digestion, the gene was ligated into the pET28a vector (Novagen) and transfected into E. coli (NovaBlue, Novagen). A construct containing the four silent mutations was identified and verified by restriction digestion, PCR, and automated sequencing with internal primers.
This construct harboring the stabilized pfenr gene was used as a template to prepare a N-terminal and C-terminal truncated version, using expression primers E1 (5Ј-ACGTCCCATGGTGCATCATCATCA-TCATCATAATGAAGATATTTGTTTTATTGCTGGTATAGG) and E2 (5Ј-ATATGGATCCTCAATCATCTGGTAAAAACATTATATTTAATCC-GTTATCCACATATATTGTCTG) (NcoI and BamHI sites underlined) and the PCR conditions described above. This truncated gene was ligated into pET28a, and its sequence was verified.
Recently, the full-length pfenr gene sequence also appeared in the P. falciparum genome data base as sequence from the P. falciparum "blob" chromosomes that comigrate on pulse-field gels. This has been confirmed independently by two other publications (19,26). When compared with these recently published sequences, our data from multiple independent PCR products show a Gln at position 35 instead of His, and an Asn instead of a Tyr at position 88 of the complete amino acid sequence. Both changes occur at the N terminus of the enzyme in a region that is structurally distant to the site of enzymatic function.
PfENR Expression and Purification-BL21(DE3) Codon ϩ -RIL cells (Novagen) harboring the expression plasmids were grown in Terrific broth. When the A 600 reached 0.8, the cells were induced with 1 mM isopropyl-1-thio-␤-D-galactopyranoside for 5 h at 37°C. Cell pellets were resuspended in buffer A (20 mM Tris/HCl, pH 8.0, 500 mM NaCl, 50 mM imidazole) and disrupted using a French press. The filtered supernatant was applied to a metal chelate affinity column loaded with nickel. The column was washed with buffer B (20 mM Tris/HCl, pH 8.0, 500 mM NaCl, 150 mM imidazole) and eluted with buffer C (20 mM Tris/HCl, pH 8.0, 500 mM NaCl, 400 mM imidazole). The protein was SCHEME 1. Enzymes of fatty acid elongation in E. coli. There are four reactions in each cycle of fatty acid elongation. The first reaction is initiated by ␤-ketoacyl-ACP synthase III, which condenses malonyl-ACP with acetyl-CoA. Reduction of the resulting ␤-ketoester is accomplished by ␤-ketoacyl-ACP reductase. The third reaction is catalyzed by ␤-hydroxyacyl-ACP dehydratase. The final reaction is catalyzed by a single NADH-dependent enoyl-ACP reductase that converts trans-2enoyl-ACP to acyl-ACP. The malonyl-ACP used by the condensing enzymes arises from the transacylation of malonyl-CoA to ACP catalyzed by malonyl-CoA:ACP transacylase. Malonyl-CoA is synthesized by acetyl-CoA carboxylase (not shown). All GenBank TM accession numbers for the corresponding enzymes of P. falciparum are shown in parentheses. concentrated using Centriprep 30 and applied to a Superdex 75 sizeexclusion column equilibrated with buffer D (20 mM Tris/HCl, pH 7.5, 150 mM NaCl).
Crystallization and Data Collection-Using hanging drop and vapor diffusion methods, PfENR was crystallized as a binary complex with NADH bound to the enzyme and as a ternary complex with NAD ϩ and triclosan. The protein in buffer D (20 mg/ml) was incubated with 4 mM NAD ϩ and 1 mM triclosan for the ternary complex and with 4 mM NADH for the binary complex. Two microliters of these mixtures was mixed with 2 l of well solution consisting of 2.35 M (NH 4 ) 2 SO 4 and 100 mM buffer, pH 5.6 (sodium acetate for the ternary, MES for the binary complex) and equilibrated against the reservoir solution at 18°C. The crystals of both complexes were isomorphous, belonged to the space group P4 3 2 1 2 (cell dimensions a ϭ b ϭ 134.0 Å, c ϭ 84.0 Å), and contained a dimer (half of the functional tetramer) in the asymmetric unit.
Crystals of ternary complexes with NADH and the triclosan analogs 1 and 2 were prepared by soaking binary ENR⅐NADH crystals. The inhibitors were dissolved in acetonitrile, directly added to the drops containing crystals of binary complexes, and incubated for a week.
Diffraction data was collected at room temperature to 2.35-to 2.50-Å resolution from single crystals using a MacScience DIP2030 image plate detector with double-focusing mirrors coupled to a Rigaku x-ray generator utilizing a copper rotating anode (CuK ␣ wavelength ϭ 1.54 Å). The data were processed and scaled using DENZO/SCALEPACK (29).
Structure Determination and Refinement-The structure of the ternary ENR⅐NAD ϩ ⅐triclosan complex was solved by molecular replacement with AMORE (30) using only protein coordinates of the Brassica napus ENR structure (Protein Data Bank entry 1ENO) as a search model. The initial solution was used as a template for the Automated Protein Modeling Server (available at www.expasy.ch/swissmod/) to generate a three-dimensional model of the PfENR sequence. The resulting model was then used to calculate an initial electron density map at 2.43 Å, which showed strong and continuous density for NAD ϩ and triclosan. Several rounds of model refinement included the addition of missing amino acids. In a late stage, water was automatically added and the final refinement was carried out without any noncrystallographic symmetry restraints. This yielded a final R work of 17.1% and a value for R free of 21.3%. The first 9 amino acids (including the His 6 -tag) were not resolved, and 40 of the 43 amino acids comprising an insertion next to the binding loop area did not show any density. There were no additional breaks in the main chain, although the density was weak for residues Ile 153 -Lys 155 and Glu 179 -Asn 183 that form two small loop regions. The average B-value for protein atoms was 36 Å 2 . The final model contained a total of 289 amino acids, one NAD ϩ molecule, one triclosan molecule, and 57 water molecules in each monomer. PROCHECK analysis showed 90% of all residues in the most favored and 10% in the generously allowed regions of the Ramachandran diagram.
Because crystals of the binary complex with NADH were isomorphous to the ternary complex, the protein coordinates of the latter were used to calculate the initial binary complex density maps. The first map calculated at 2.40 Å clearly identified the NADH cofactor with strong and continuous density. Subsequent refinement led to an R work of 17.6% and an R free of 22.4%. Again, no density was observed for the first 9 amino acids and the same 40 amino acids of the binding loop insertion of each monomer, and the same areas for the loops showed weak density. The average B-value for main-chain atom positions of the binary structure was 31 Å 2 . The final model contained a total of 289 amino acids, one NADH molecule, and 77 water molecules in each monomer.
The ternary structures with bound inhibitors 1 and 2 were solved using the method described above. Initial maps showed strong density for NADH, and additional differences in electron density at the inhibitor binding site. Good density for 1 was only observed in monomer B, whereas 2 showed excellent density in both monomers. The inhibitors were built into the model, and subsequent refinement for 1 led to an R work of 18.7% and an R free of 23.2%, with a total of 289 amino acids, one NADH, and 69 solvent molecules in each monomer and one 1 molecule in subunit B. The ternary structure with bound 2 was refined to an R work of 17.6% and an R free of 22.7%. The model comprised 289 amino acids, one NADH molecule, one 2 molecule, and 64 solvent molecules in each monomer. Both structures lacked density for the initial 9 amino acids and the same residues of the large loop insertion. The density for the two small loops was weak. The average B-values of the main-chain atoms of the ternary ENR⅐NADH⅐1 and ENR⅐NADH⅐2 complexes were where F obs and F calc are observed and calculated structure factors, respectively. c R free was calculated on 10% of the data omitted at random.  Table I. Enzyme Assay-All experiments were carried out on a Shimadzu UV-1201 UV-visible spectrophotometer at 25°C in 20 mM Tris/HCl, pH 7.6, 150 mM NaCl. Kinetic parameters were determined spectrophotometrically by following the oxidation of NADH to NAD ϩ at 340 nm (⑀ ϭ 6.3 mM Ϫ1 cm Ϫ1 ). K m and V max values for crotonoyl-CoA were determined at a fixed and saturating concentration of NADH (200 M) and by varying the substrate concentration (0 -500 M). K m and V max values for NADH were determined at variable concentrations of NADH and a fixed and saturating concentration of crotonoyl-CoA (500 M). Kinetic parameters were obtained by fitting the initial velocity data to the Michaelis-Menten equation.
Inhibition constants were determined under conditions of saturating substrate (500 M crotonoyl-CoA, 200 M NADH) and variable inhibitor concentration. Values for K i were determined from the x-intercept of a Dixon plot, assuming uncompetitive inhibition. Mean values of two independent experiments are reported for kinetic parameters and inhibition data.
Whole Cell Assay-The inhibitory activities of triclosan and its analogs against P. falciparum asexual blood stages were determined using a 72-h in vitro assay that measures decreases in [ 3 H]hypoxanthine uptake as a marker of growth inhibition (31,32).
Compound 2, 4-chloro-2-hydroxyphenyl 6Ј-hydroxynaphthyl ether, was synthesized with a 46% overall yield by copper-catalyzed coupling of 2-bromo-4-methoxynaphthalene with 4-chloro-2-methoxyphenol using cesium carbonate as a base and 1-naphthoic acid as an additive to increase the efficiency of the coupling of the less soluble phenoxide (34), followed by demethylation of the resulting aryl methyl ether.

RESULTS AND DISCUSSION
Identification, Cloning, and Expression of PfENR-Fulllength P. falciparum ENR was cloned using PCR primers designed on the basis of sequence alignments of reported microbial and plant ENRs as well as contig data from the P. falciparum genome project. Primers were chosen to conserved regions and were used in combination with vector-specific primers to PCR amplify overlapping fragments of the pfenr gene from a P. falciparum gametocyte stage cDNA library. This yielded a single-exon open reading frame of 1299 bp with an A/T content of 72.0%. The predicted start codon was preceded by stop codons in all three reading frames located in a 0.5-kb 5Ј-untranslated region with an increased A/T content of 85.0%.
A stretch of 10 contiguous adenosines turned out to be particularly vulnerable for deletion mutations and was associated with minimal expression of full-length or truncated PfENR. Four silent mutations were introduced in this region, and the resulting construct was stable in E. coli and was used as a template for all further PCR reactions and expression studies.
To increase protein yield and facilitate the crystallization process, the PfENR protein was expressed without the N-terminal signal and translocation peptide and the following 18 residues, as well as the C-terminal 7 amino acids, which were predicted to extend into solvent and potentially interfere with crystallization but not contribute to enzyme function. When BL21(DE3) Codon ϩ -RIL cells (Novagen) were used for expression, the purification typically resulted in 20 -30 mg of truncated PfENR per liter of media. The K m and V max values of the truncated PfENR (Table II) were indistinguishable from the full-length enzyme carrying silent point mutations.
PfENR Sequence Characteristics-pfenr encoded a predicted The degree of sequence identity was 48, 16, and 30%, respectively. Green indicates completely conserved residues, yellow indicates two or more highly conserved residues, and blue indicates at least one similar amino acid residue. The secondary structure of P. falciparum ENR is shown above the sequences. The putative signal sequence cleavage site located between Cys 20 and Phe 21 is marked by the downward arrow (2). The first amino acid (Glu 78 ) of the mature enzyme is marked the double arrow (s). The residues proposed to be catalytically important are indicated by asterisks (*). The 43-amino acid low complexity region is underlined in red.
protein of 432 amino acids with an expected molecular mass of 49.8 kDa. Sequence alignments (Fig. 1) revealed that PfENR showed much greater overall sequence similarity to plant ENRs than to microbial ENRs. Regions of homology with plant enzymes were divided by a 43-amino acid insert (residues 325-367) that was enriched in the polar residues asparagine (30%), lysine (12%), glutamine (9%), and serine (9%). This low complexity insertion was demonstrated to be coding, because it was routinely identified in cDNA libraries (generated from oligo-dT-primed DNase I-treated poly(A) ϩ RNA and prepared from either asexual or sexual stage intra-erythrocytic parasites), it did not have typical splice acceptor and donor sequences (35), and it maintained the same A/T content as the other coding regions. Similar insertions have previously been reported in P. falciparum enzymes and are thought to generally have minimal impact on function (36 -39).
PfENR has a long N-terminal extension (similar in length to plant ENRs) that is characteristic of bipartite N-terminal presequences found in Plasmodium and Toxoplasma parasite proteins targeted to the apicoplast (12,14). Using prediction programs SIGNALP and PSORT, a putative cleavage site for the signal peptide could be detected between residues Cys 20 and Phe 21 (40,41). The size of the adjacent apicoplast translocation signal was recently discovered to be 56 amino acids long, with Glu 78 being the first amino acid of the mature enzyme (19).
Sequence alignments revealed 33 completely conserved and 44 highly conserved residues in the PfENR sequence when compared with the ENRs of E. coli, B. napus, and Mycobacterium tuberculosis. Mapping of the completely conserved residues onto the three-dimensional structure of PfENR showed Gly 104 , Tyr 277 , Met 281 , Lys 285 , Ala 312 , and Pro 314 located immediately adjacent to the nicotinamide ring of NADH. These residues are likely to play an important role in substrate recognition and/or the catalytic function of the enzyme. The remaining conserved residues were dispersed throughout the structure, occurring mainly at the interfaces between the subunits of the tetramer (Fig. 2b) or in positions where they were predicted to stabilize the orientation of secondary structure.
Prokaryotic FAS-II enoyl reductases catalyze the NADH/ NADPH-dependent reduction of 2-trans-enoyl-ACP by the direct transfer of the 4S hydrogen atom of NADH to the C 3 position of the ␣,␤-unsaturated thioester, via an intermediate enolate anion. Protonation of C 2 , by an as yet unknown group, would lead to the collapse of the enolate intermediate and yield the saturated product. Enoyl reductase activity assays on purified PfENR demonstrated a strict NADH dependence and substrate specificity for short-chain and medium-chain fatty acids, including crotonoyl-CoA (C4:1), as well as octenoyl-CoA (C8:1) and dodecenoyl-CoA (C12:1). Kinetic studies using crotonoyl-CoA gave K m and V max values of 48 Ϯ 3 M and 16 Ϯ 2 M min Ϫ1 , respectively. This observed K m value for PfENR was similar to that measured for the homologous plant enzymes from B. napus and Spinacia oleracea but significantly lower than the K m of 2700 M determined for the E. coli ENR (Table  II) (42). Using crotonoyl-CoA as substrate, analysis of NADH oxidation gave a K m of 30 Ϯ 3 M and a V max of 77 Ϯ 6 M min Ϫ1 .
The Tertiary Structure of the PfENR Subunit-Overall, the structure was reminiscent of the Rossmann fold (43) and was similar to all other structurally defined homologous enzymes. The PfENR subunit comprised a single domain of ϳ55 ϫ 50 ϫ 50 Å (Fig. 2a). Each subunit was composed of seven ␤-strands (␤1-␤7) that formed a parallel ␤-sheet and nine ␣ helices (␣1-␣9) that were connected to the ␤-strands by a number of loops of varying length. The parallel ␤-sheet was flanked by helices ␣1, ␣2, ␣4, ␣5, ␣6, and ␣9, with ␣3 arranged along the top of ␣2 and ␣4. Helix ␣8 was located at the C termini of strands ␤6 and ␤7. Comparison of the C␣ positions of the E. coli, M. tuberculosis, and B. napus ENRs showed that the overall structure of all enzymes was very similar (the C␣ root mean square for superimposition of PfENR with ENR from B. napus: 0.30 Å; from E. coli: 0.78 Å; from M. tuberculosis: 0.75 Å; Fig. 3, a, c, and e). Whereas the core region built of the ␤-sheet was nearly identical in all structures, major differences were nevertheless discernible between PfENR and the ENRs of E. coli and M. tuberculosis. The loop regions between ␣2-␤3, ␤3-␣3-␣4, and ␤4 -␣5 of PfENR were longer due to insertions in the sequence, and helix ␣2 was shifted away from the protein, toward the solvent, relative to the bacterial ENRs. Helix ␣3, a small ␣ helix in the loop region between ␤3 and ␣4, was not observed in the microbial structures but was present in the plant ENR.
The aforementioned 43-amino acid low complexity insertion in PfENR sequence localized to an important loop region (␣7 and ␣8), near the catalytic center of the protein. This region is thought to be a determinant in substrate specificity, because it participates in acyl substrate binding, as shown in the M. tuberculosis structure in which the bound fatty acyl substrate was held in place by the substrate binding loop (44). Only 3 amino acids of the low complexity insertion were visible in the electron density maps, indicating that most of this region was disordered even in the presence of bound substrate and inhibitor. Nonetheless, the last visible amino acids just before (Lys 325 ) and after (Tyr 366 ) the low complexity region were in nearly the exact same position as the comparable loop residues in the E. coli enzyme structure (Fig. 3d). For the E. coli ENR, this loop was disordered when NAD ϩ was bound, but became ordered upon binding of the inhibitor triclosan (45,46). In the M. tuberculosis ENR this loop was shifted away in a more open conformation (Fig. 3f), presumably for the binding pocket to accommodate the binding of the longer fatty acid substrates FIG. 2. Tertiary structure of PfENR. a, representation of subunit B of the PfENR tetramer with the cofactor NADH and inhibitor triclosan bound to their active sites. Helices are shown in gold, the ␤-strands in green, NADH and triclosan are colored by atom type. The tertiary structure shows the Rossmann fold typical of dinucleotide-binding enzymes (43). The chain break visible at the top of the inhibitor binding site is due to the PfENR substrate binding loop that was not resolved in the crystal structures. b, front view of the PfENR tetramer, in which each subunit is represented as a differently colored tube. The bound NADH is colored by atom type. Three perpendicular 2-fold symmetry axes intersect in the center, creating a molecule of internal 222 symmetry.
(C16 -C56, precursors of mycolic acid) used by the mycobacterial enzyme. The loop region of B. napus ENR was very different because ␤6 makes a turn in the opposite direction when compared with all other structures, and connects ␤6 with ␣7 of B. napus ENR, resulting in a substrate binding pocket that is more solvent exposed (Fig. 3b).
Earlier investigations into E. coli, B. napus, and M. tuberculosis ENRs have found a correlation between the length of the substrate binding loop and the fatty acyl substrate chain length (44). If this held for P. falciparum ENR, one could expect that very long-chain fatty acids would serve as substrate and, by analogy to the M. tuberculosis ENR, the PfENR would not use very short-chain acyl-CoAs as substrate. However, kinetic studies indicated that the P. falciparum ENR can use crotonoyl-CoA (C4:1) (Table II), with kinetics in line with those observed for S. oleracea and B. napus ENRs.
The Quaternary Structure of P. falciparum ENR-In gel filtration studies, PfENR formed a tetramer in solution (data not shown), in agreement with all other bacterial and plant ENRs reported to date. In support of this, the packing in the crystal showed an obvious homotetramer possessing internal 222 symmetry (Fig. 2b). The estimated dimensions for the tetramer were 60 ϫ 85 ϫ 85 Å. The solvent-accessible surface areas for the subunits and the tetramer were calculated using DSSP (47) and were determined to be about 15,000 Å 2 for each subunit and 43,000 Å 2 for the tetramer. Approximately 1600 Å 2 (11%) of the surface area of subunit A was buried, making intermolecular contacts with subunit B, 1700 Å 2 (12%) with subunit C, and 900 Å 2 (6%) with subunit D. Thus, the total surface area involved in intermolecular contacts of each subunit was 4200 Å 2 or 29%. This type of organization for PfENR was comparable to the crystal structures elucidated for enoyl-ACP reductases from E. coli (48), B. napus (49), and M. tuberculosis (44).
Analysis of the Nucleotide Binding Site-Both the ENR⅐NADH binary complex and the ENR⅐NAD ϩ /NADH⅐inhibitor ternary complexes showed excellent electron density for the cofactors (Fig. 4a). NAD ϩ and NADH were localized to the enzyme in an extended conformation at the C-terminal end of the ␤-sheet with both ribose sugar rings found as C2Ј-endo conformers and the nicotinamide moiety in the syn conformation (Fig. 4b).
The adenine ring was located in a pocket on the surface of the protein, formed by the side chains of Trp 131 , Phe 167 , Ala 169 , Ser 170 , Asn 218 and the main chain between residues Phe 167 and Asp 168 . Hydrogen bonds were formed between the adenine nitrogen atoms at position N1 with the peptide nitrogen atom of Ala 169 and at position N6 with the side chain of Asp 168 . Both the 2Ј and 3Ј adenine ribose hydroxyl groups were hydrogenbonded to the same ordered water molecule, which in turn was hydrogen-bonded with the peptide nitrogen of Trp 131 . This mode of hydrogen bonding in P. falciparum was different from many NAD ϩ /NADH-linked dehydrogenases for which the common mechanism for cofactor recognition involves hydrogen bonding between an acidic residue and both ribose hydroxyl groups (50). However, the recognition of the adenine ribose hydroxyl groups was similar to that observed in B. napus and E. coli ENRs. In PfENR, the 2Ј-hydroxyl group of the adenine ribose occupied a small depression flanked by Gly 106 , Trp 131 and Val 134 , which resulted in a tight fit for the NAD ϩ /NADH cofactor. This spatial arrangement leaves no room for the extra phosphate group of NADPH and is consistent with the previous observation that NADH was a much more efficient cofactor than NADPH for PfENR function (19).
The pyrophosphate moiety of NAD ϩ (NADH) lay close to the C-terminal part of the ␤-sheet and interacted with the glycinerich region of the loop connecting ␤1 and ␣1, with ␣1 being the nucleotide binding helix. Contacts through hydrogen bonds were made by the pyrophosphate oxygen atoms with the peptide nitrogen of Tyr 111 and, mediated by a solvent molecule, with the main-chain carbonyl of Gly 104 , the peptide nitrogen of Gly 112 and the side-chain hydroxyl of Ser 215 . By way of an additional water molecule, hydrogen bonding also occurred with the peptide nitrogen of Gly 104 . The cofactor was bound by this series of hydrogen bonds and was not further supported by positively charged side chains close to the nucleotide binding site.
The nicotinamide binding pocket was composed of the side chains of Tyr 111 , Leu 265 , Tyr 267 , Tyr 277 , Ala 312 , Gly 313 , Pro 314 , Leu 315 , and Ile 369 . Both 2Ј-and 3Ј-ribose hydroxyl groups hydrogen-bonded to the amino group of the Lys 285 , whereas only the 3Ј-group interacted with a solvent molecule that was contacted by the side chain of His 214 . The nicotinamide ring was completely ordered on the enzyme, where it interacted via specific hydrogen bonds formed by the oxygen and nitrogen of the carboxamide moiety and the Leu 315 peptide nitrogen and cofactor pyrophosphate moiety. These interactions appeared to stabilize the packing of the A-face of the nicotinamide ring against the phenolic ring of Tyr 111 , exposing the B-face to the active site. Thus, the cofactor adopted the same conformation ENRs. c, comparison of PfENR (red) and the E. coli ENR (green) is shown on the left, whereas the lid region has been magnified and is presented in d. Both lid regions adopted very similar orientations and conformations. e, major structural differences were found for the lid region when the ENR of P. falciparum (red) and M. tuberculosis (gold) were superimposed.
The structural characteristics of PfENR cannot account for the recent finding by Surolia and Surolia (19) of a reduced but significant incorporation of [ 14 C]malonyl-CoA in the presence of NADPH in a P. falciparum cell-free fatty acid synthesis system. This raises the question whether there might exist additional, so far not described malonyl-CoA-and NADPH-dependent metabolic pathways in P. falciparum. Just recently, the structure of another FAS-II enzyme that uses NADPH has been solved. This ␤-ketoacyl-ACP reductase (FabG) structure was reported from B. napus (56), and we have obtained partial sequence for the homologous P. falciparum protein as an initial step toward enzyme characterization.
Location of the Fatty Acyl Binding Pocket-Earlier structural studies of M. tuberculosis ENR complexed to the C16 fatty acyl substrate analog and NAD ϩ revealed that the C16 substrate bound in a U-shaped conformation, with the trans double bond position directly adjacent to the nicotinamide ring of NAD ϩ and the side chain of Tyr 277 interacting directly with the C16 substrate thioester carbonyl oxygen (44). The binding crevice for the fatty acyl portion of the substrate was built of hydrophobic residues that were derived primarily from the substrate binding loop. The ENR structures from P. falciparum, B. napus, and E. coli showed a similar patch of predominantly hydrophobic side chains adjacent to the position of the nicotinamide ring and the fatty acid chain binding area. With PfENR, the corresponding amino acids and side chains flanking the putative binding site were Tyr 267 , Gly 276 , Tyr 277 , Met 281 , Pro 314 , Ala 319 , Ala 320 , Ala 322 , Ile 323 , Ile 369 , and Ala 372 . Most residues were located in helices ␣7 and ␣8 and formed a hydrophobic fingershaped cavity with the approximate dimensions of 10 ϫ 8 ϫ 6 Å. One side of the cavity was accessible to solvent.
Based on volumetric measurements of the fatty acyl binding cavity in the PfENR⅐NADH binary complex, there was only enough space to accommodate a substrate of six to eight carbon atoms in length. This implies that there must be sufficient flexibility in the PfENR pocket to bind longer substrates (up to at least C16), which probably occurs via an opening movement of the flexible loop. In view of the present biochemical and structural information, it is possible that the extensive substrate binding loop of Pf ENR, which includes the low complexity region, would allow for broader specificity in the fatty acyl chain.
Analysis of the Triclosan Binding Site-Three distinct classes of chemically synthesized agents have been shown to act by inhibiting this enzyme in the bacterial FAS-II pathway. Isoniazid targets the ENR homolog, InhA, from M. tuberculosis after activation and covalent attachment to the nicotinamide ring of NADH (57). The mechanism of inhibition of the diazaborines is through a covalent bond between a boron atom in the diazaborine and the 2Ј-hydroxyl group of the nicotinamide ribose moiety in this enzyme (17,42,48). Another extremely potent bacterial ENR inhibitor is triclosan, which, in contrast to the other antimicrobials, forms a non-covalent complex with NAD ϩ and protein primarily via hydrogen bonds.
The crystal structure of PfENR solved with protein incubated with NAD ϩ and triclosan revealed the mode of triclosan binding (Fig. 4b). Comparison of the P. falciparum ENR⅐NAD ϩ ⅐triclosan structure with the corresponding E. coli and B. napus structures (45,53) demonstrated that, for the malaria enzyme, the binding mode for triclosan was very similar, showing the same stacking interaction with the nicotinamide ring of NAD ϩ and comparable hydrogen-bonding pattern with the 2Ј-hydroxyl group of the nicotinamide ribose and with Tyr 277 . Ring A (the phenol ring) of the inhibitor interacted face-to-face with the nicotinamide ring of NAD ϩ allowing -cation interactions. The same ring formed additional van der Waals interactions with the side chains of Tyr 267 , Tyr 277 , Pro 314 , Phe 368 , and Ile 369 . The phenolic hydroxyl hydrogenbonded to the 2Ј-hydroxyl moiety of the nicotinamide ribose and the oxygen atom of Tyr 277 , and the amino nitrogen of Lys 285 was within 4.6 Å. These residues are completely con- PfENR⅐inhibitor complexes. Residues involved in the formation of the binding pocket are shown. The most important amino acids for interactions are labeled. Triclosan is shown in red, 1 in blue, and 2 in green. The corresponding cofactor of each inhibitor complex is colored accordingly. Hydrogen bonds with Tyr 277 and the 2Ј-hydroxyl group of the nicotinamide ribose, as well as three additional hydrogen bonds mediated through the hydroxyl group of the naphthalene ring of 2 are shown as yellow dotted lines. The binding mode of triclosan, 1, and 2 showed the same stacking interactions to the nicotinamide ring of the cofactor with respect to ring A. Inhibitor 2 also exhibited three additional hydrogen bonds mediated through the hydroxyl group of the naphthalene ring that could interact with the side-chain nitrogen of Asn 218 and the main-chain oxygen and nitrogen of Ala 219 . served in all known ENRs and have been implicated in the enzyme's catalytic mechanism (49,58). The 4-chloro atom of ring A was surrounded by mainly hydrophobic residues making van der Waals contacts with the side chains of Tyr 267 , Pro 314 , and Phe 368 . The ether oxygen atom of triclosan interacted with the 2Ј-hydroxyl group of the nicotinamide ribose, and it approached to within 3.65 Å of one of the oxygen atoms of the nicotinamide ribose phosphate group. Ring B (2,4-dichlorophenoxy ring) of triclosan was located in a pocket bounded by the pyrophosphate and nicotinamide moieties of NAD ϩ , by the peptide backbone residues 217-231 and by the side chains of Asn 218 , Val 222 , Tyr 277 , and Met 281 . Although the 4-chloro atom of ring B was placed adjacent to the side chains of Val 222 and Met 281 and residues 218 -219, the 2-chloro atom was surrounded by the ␣-carbon atom, the side chain of Ala 217 , and atoms of the nicotinamide ribose pyrophosphate moiety.
One ordered water molecule was observed in the inhibitor binding site of the binary ENR⅐NADH complex, interacting through hydrogen bonds with the 2Ј-hydroxyl group of the nicotinamide ribose and with Tyr 277 . This water molecule was very close to the position of the phenolic hydroxyl group of the inhibitors and was displaced upon triclosan binding. Superposition of the binary and ternary complex structures revealed subtle conformational changes in the protein upon inhibitor binding, with the most pronounced change being a slight shift of helix ␣7 by 0.5 Å toward the solvent.
The Triclosan Analogs 1 and 2-Recently, the antibacterial activities of several 2-hydroxydiphenylethers as well as hexachlorophene and 2-hydroxydiphenylmethanes were determined (55). Studies with a des-hydroxyl analog of triclosan showed a more than 10,000-fold reduced affinity for E. coli ENR and implicated a critical antibacterial role for the triclosan 2-hydroxy group (55). Moreover, it was proposed that the ether oxygen of triclosan might be critical to the formation of the ternary complex, because corresponding 2-hydroxydiphenylmethanes did not result in tight binding (59), whereas the replacement of the ether oxygen with sulfur abolished the inhibitory activity (21). Ring B of triclosan was considered to be of minor importance, because variations in this region had less effect on inhibitor activity.
Synthesis of 20 triclosan analogs (originally designed to target the closely related M. tuberculosis ENR) and subsequent screening against purified PfENR revealed that PfENR inhibition was sensitive to the hydroxyl group at position 2 in ring A, which could not be replaced with methoxy groups or sulfur derivatives, as observed for the bacterial ENRs. These studies led us to identify a diphenylamine derivative (1, Fig. 5) with moderate inhibitory activity (K i 14.3 Ϯ 1.4 M). Bound to PfENR, this inhibitor, which carried nitrogen as a bridging atom, adopted a very similar conformation compared with triclosan (2,4-dichloroaniline ring). The extensive hydrogen bonding network involving the 2-hydroxy group of ring A and the stacking interaction with the nicotinamide ring system were completely maintained, emphasizing the importance of these interactions (Fig. 4b).
Another analog, 2, was found to have a K i of 150 Ϯ 14 nM against purified PfENR, close to the 50 Ϯ 6 nM K i observed with triclosan (Fig. 5). Structural analysis of this 6-hydroxy naphthalene derivative complexed to PfENR in the presence of NADH revealed an electron density that was clearly detectable after the first round of refinement, showing strong and continuous difference density in both monomers of PfENR. Ring A of 2 again stacked to the nicotinamide ring system in the same way as observed for triclosan. The extended ring B of 2 was oriented in the same direction as the corresponding ring in triclosan but was tilted by 10 degrees out of the plane found for ring B of triclosan and 1. This now allowed for three potential new hydrogen bonds mediated through the hydroxyl group of the naphthalene ring, the side-chain nitrogen of Asn 218 and the main-chain oxygen and nitrogen of Ala 219 (Fig. 4, a and b). Reduced affinity of 2 for PfENR may be a result of the missing chlorine in this analog. Because most bacterial ENRs have a Phe at the position comparable to Asn 218 of PfENR, it is likely FIG . 5. Structures of triclosan and derivatives tested for inhibitory activity. Biochemical activity assays were performed using purified recombinant PfENR. Out of the presented compounds, only triclosan, 1, and 2 showed significant biochemical inhibition (cut-off 100 M) of purified PfENR. The corresponding K i values for the active compounds are indicated. ]hypoxanthine uptake curves plotted as a function of compound concentration. All compounds were tested in duplicate on two to five separate occasions. that 2 will be specific for the malarial protein.
In Vitro Whole Cell Test-Whole cell assays against P. falciparum parasites propagated in human erythrocytes in vitro demonstrated 50% killing at a concentration of 1.8 -6.0 M for 1 and 5.9 -19.5 M for 2 (Table III). Compound efficacy was similar against drug-sensitive and multidrug-resistant strains. Triclosan proved even more effective with IC 50 values against multiple P. falciparum lines in the 0.3-1.5 M range, values that are consistent with reports from other groups (19,26). We note that 1 was more effective than 2 in killing P. falciparum asexual blood stages in culture even though the latter displayed a better K i value against purified PfENR. These findings raise the issue of compound solubility and uptake into the parasitized erythrocyte and support the requirement to combine both tests in subsequent studies that aim to define a structure-activity relationship and identify compounds as active as triclosan and more suitable for systemic administration in humans.
These data report critical structural information for a new and promising target for the development of novel antimalarial compounds. The elucidation of triclosan interactions with this target, PfENR, and the identification of two triclosan analogs that bind to this enzyme and that are active in vitro against P. falciparum provide the first data on structure-activity relationships of PfENR inhibitors. Further structure-based and medicinal chemistry approaches are warranted to leverage the uniqueness of this enzymatic pathway in malaria parasites as a strategy to counter the spread of lethal, drug-resistant P. falciparum infections.