Determination of Selectivity and Efficacy of Fatty Acid Synthesis Inhibitors*

Type II fatty acid synthesis (FASII) is essential to bacterial cell viability and is a promising target for the development of novel antibiotics. In the past decade, a few inhibitors have been identified for this pathway, but none of them lend themselves to drug development. To find better inhibitors that are potential drug candidates, we developed a high throughput assay that identifies inhibitors simultaneously against multiple targets within the FASII pathway of most bacterial pathogens. We demonstrated that the inverse t½ value of the FASII enzyme-catalyzed reaction gives a measure of FASII activity. The Km values of octanoyl-CoA and lauroyl-CoA were determined to be 1.1 ± 0.3 and 10 ± 2.7 μm in Staphylococcus aureus and Bacillus subtilis, respectively. The effects of free metals and reducing agents on enzyme activity showed an inhibition hierarchy of Zn2+ > Ca2+ > Mn2+ > Mg2+; no inhibition was found with β-mercaptoethanol or dithiothreitol. We used this assay to screen the natural product libraries and isolated an inhibitor, bischloroanthrabenzoxocinone (BABX) with a new structure. BABX showed IC50 values of 11.4 and 35.3 μg/ml in the S. aureus and Escherichia coli FASII assays, respectively, and good antibacterial activities against S. aureus and permeable E. coli strains with minimum inhibitory concentrations ranging from 0.2 to 0.4 μg/ml. Furthermore, the effectiveness, selectivity, and the in vitro and in vivo correlations of BABX as well as other fatty acid inhibitors were elucidated, which will aid in future drug discovery.

Infectious disease is a global problem, and the development of drug resistance is a major issue for all classes of antibiotics. Therefore, development of new high throughput assays for simultaneous screening of multiple targets to rapidly identify novel antibiotics is crucial. Fatty acids are essential for viability. The significant organizational and structural differences between the fatty acid synthesis of bacteria and humans make this system an attractive target for antibacterial drug discovery. The human fatty acid synthase (FASI) 1 is a multifunc-tional single polypeptide composed of distinct enzyme domains. In contrast, bacterial fatty acid synthesis (FASII) is carried out by a series of individual enzymes, which has been extensively reviewed (1,2) and is schematically described in Fig. 1. Initially, acetyl-CoA is carboxylated by AccABCD (3) to form malonyl-CoA, which is, in turn, transferred to ACP (4) by FabD (5). Fatty acid synthesis is initiated by FabH (6) supplying substrates (acetoacetyl-ACP) to the fatty acid elongation cycle, which includes FabG (7), FabA/Z (8,9), FabI (L/K) (10 -12), and FabF/B (13,14) enzymes. In the cycle, the keto group of ␤-ketoacyl-ACP is reduced to a hydroxyl group by NADPH-dependent reductase FabG. ␤-Hydroxyacyl-ACP is dehydrated by dehydratase FabA or FabZ. The double bond of trans-2-enoyl-ACP is reduced by NADH-dependent reductase FabI(K/L), which feeds the substrate back to FabF/B, which in turn adds an additional acetate unit (two carbons), and the cycle iterates.
Two marketed antibacterial agents, triclosan (antiseptic) and isoniazid (anti-TB), target the FabI enzyme of fatty acid synthesis (15,16). Two natural products, cerulenin (4) and thiolactomycin (17) inhibit the condensation enzymes FabH and FabF/B, with cerulenin showing selectivity for FabF/B, whereas thiolactomycin and its analogs (18,19) inhibit FabH and FabF/B. Thiolactomycin mimics malonyl-ACP in its binding site, whereas cerulenin forms a covalent bond with the active site cysteine in FabF/B with its tail occupying a long hydrophobic cavity, which normally contains the growing acyl chain of the natural substrate (13,21). Lately, inhibition of FabH has drawn significant attention in drug discovery efforts (22,23).
In this study, we developed and validated a reliable high throughput fatty acid synthesis pathway assay that can simultaneously identify inhibitors of multiple targets, including FabD, FabF/B, FabG, FabA/Z, and FabI. We screened natural product extract libraries and identified and characterized bischloroanthrabenzoxocinone (BABX) as a new inhibitor of fatty acid synthesis. We also investigated the kinetics of pathway enzymes, the selectivity of fatty acid synthesis inhibitors, as well as the relationship of biochemical activities and antibacterial activity.
other radiolabeled chemicals used in this study were from PerkinElmer Life Sciences. ACP (Sigma, A7303) was pretreated with 3 mM DTT on ice for 20 min, aliquoted, and stored at Ϫ80°C.
Preparation of Type II Fatty Acid Synthesis Enzymes-The procedure described previously (24) was used with some modification. Briefly, K12-derivative Escherichia coli strains, Staphylococcus aureus, and Bacillus subtilis were grown to stationary phase in 6 liters of LB medium. The cultures were centrifuged at 8,000 rpm for 10 min using a Beckman JA-10 rotor. The pellets were washed twice with ice-cold buffer A (0.1 M sodium phosphate, pH 7, 1 mM EDTA, and 5 mM ␤-mercaptoethanol) and resuspended in 500 ml of the same buffer. The cells were lysed in a cold microfluidizer (Microfluidics Corp., M-110EH) at 18,000 lbs per square inch and centrifuged at 20,000 rpm for 15 min at 4°C using a Beckman JA-20 rotor. The supernatant was collected, and its volume was precisely measured. Ammonium sulfate (129 g) was added to each 500 ml of the supernatant in small quantities with low speed stirring at 4°C to reach 45% ammonium sulfate saturation. The mixture was centrifuged at 10,000 rpm for 5 min, and the supernatant was collected. Ammonium sulfate (113 g) was then added to each 500 ml of the supernatant to reach 80% ammonium sulfate saturation. The mixture was centrifuged again, and the supernatant was then discarded. The 45-80% ammonium sulfate saturated protein fraction pellet, containing all necessary fatty acid synthesis (FASII) enzymes, was dissolved in 20 ml of buffer A, dialyzed at 4°C against four changes of buffer A using 10-kDa molecular mass cutoff dialysis tubing (Invitrogen, 15961-022), and then concentrated. The protein concentration was determined using the standard Bio-Rad protocol. The protein was aliquoted, flash frozen using liquid nitrogen, and stored at Ϫ80°C.
FASII Assay-The assay was performed in a Phospholipid 96-well Flashplate. Routinely, 3 g of the partially purified protein containing fatty acid synthesis enzymes were preincubated with a serial dilution of natural products or synthetic compounds at room temperature for 20 min in 50 l of buffer containing 100 mM sodium phosphate (pH 7.0), 5 mM EDTA, 1 mM NADPH, 1 mM NADH, 150 M DTT, 5 mM ␤-mercaptoethanol, 20 M n-octanoyl-CoA (or lauroyl-CoA), 4% Me 2 SO, and 5 M of the pretreated ACP. The reaction was initiated by addition of 10 l of water-diluted [ 14 C]malonyl-CoA (the label is at C-2 of the malonyl group), which gave a final concentration of 4 M malonyl-CoA with total counts of about 20,000 (84 cpm/pmol) or 10,000 CPM (42 cpm/pmol) using Beckman Coulter LS6500 and Packard TopCount NXT scintillation counters, respectively. The reaction was incubated at 37°C for 30 min for E. coli and 60 min for S. aureus and B. subtilis. The reaction was terminated by adding 100 l of 14% perchloric acid. The plates were sealed, incubated at room temperature overnight with mild shaking, and counted for 5 min using the TopCount counter. Through hydrophobic interactions, long hydrophobic acyl chains of acyl-ACP bind to the phospholipids on the surface of the well, which are coated with scintillant. This binding brings the incorporated 14 C into proximity of the scintillant resulting in the emission of a photon, which is captured by a scintillation counter. All data were analyzed using Prism (GraphPad Software, Inc.).
Whole Cell Labeling Assay-The assay was performed as previously described (25). Briefly, mid-log (A 600 ϭ 0.5-0. 6 H]glycine (S. aureus) was performed in the presence of 100 g/ml chloramphenicol, which blocks protein synthesis. The reaction was stopped by adding 10% trichloroacetic acid and harvested using a glass fiber filter (PerkinElmer Life Sciences, 1205-401). The filter was dried and counted with scintillation fluid.
Minimum Inhibitory Concentration-The MIC against each of the strains was determined as previously described (26). Cells were inoculated at 10 5 colony-forming units/ml followed by incubation at 37°C with a serial dilution of compounds in LB broth for 20 h. MIC is defined as the lowest concentration of antibiotic inhibiting visible growth.
E. coli Strain Construction-All strains in this study are listed in Table I. MB4902 E. coli (lpxC), MB5747 E. coli (tolC), and MB5746 E. coli (lpxC and tolC) strains were constructed using standard P1 transduction methodology (27). CAG12184 was obtained from the Yale University E. coli Genetic Stock Center (28). To bring in a second Tn10-linked mutation (tolC) via P1 transduction, the MB5008 strain was made by curing MB4902 of tetracycline resistance using a variation of the technique described (29). Concentrations of quinaldic acid and chlortetracycline were adjusted due to the increased sensitivity of the lpxC mutant.

Development of High Throughput Assay for Type II Fatty
Acid Synthesis (FASII Assay)-The conventional type II fatty acid synthesis pathway assay uses acetyl-CoA and malonyl-CoA as substrates (30). However, the FASII assay that we developed uses long-chain acyl-CoA (octanoyl-CoA or lauroyl-CoA) instead of acetyl-CoA. Therefore, our FASII assay measures the cumulative activities of enzymes involved in the fatty acid elongation cycle only, avoiding the complexity of the chain initiation reaction steps catalyzed by acetyl-CoA carboxylase (AccABCD) and FabH. To determine enzyme activity and set up proper conditions for the assay, we titrated the FASII enzyme mixture from E. coli (S. aureus and B. subtilis, data not shown) in a time course study ( Fig. 2A)  respectively, for 1.5, 3, 6, and 12 g of the FASII enzyme mixture added per assay. The plot of t1 ⁄2 values against each concentration of the enzyme mixture results in a linear correlation (Fig. 2B), indicating that t1 ⁄2 (or apparent rate constant) gives a measure of the activity of the FASII enzymes (31).
Determination of Enzyme Kinetics-The assay was performed under standard conditions, and the apparent K m K m(app) and V max of the FASII enzymes were determined. The K m(app) and V max of S. aureus FASII enzymes for octanoyl-CoA were 1.1 Ϯ 0.3 M and 88.9 Ϯ 4.5 pmol/min/mg, respectively (Fig. 3A). The K m(app) and V max of B. subtilis FASII enzymes were 10.0 Ϯ 2.7 M and 516 Ϯ 31 pmol/min/mg, respectively, for lauroyl-CoA (Fig. 3B). Similar results were obtained by interchanging octanoyl-CoA and lauroyl-CoA (data not shown). The experiments were repeated at least twice with duplicates.
The Effect of Metals and Reducing Agents on FASII-To ascertain the robustness of the assay for use with natural product extracts which may contain metals and chelators, the effect of divalent cations and EDTA on FASII activity was examined (Fig. 4). With the addition of EDTA, inhibition of FASII activity by Mg 2ϩ , Ca 2ϩ , Zn 2ϩ , and Mn 2ϩ exhibited IC 50 values of 81 Ϯ 17, 9.1 Ϯ 6.2, 5.6 Ϯ 3, and 19.4 Ϯ 7.6 mM, respectively. Without the addition of 5 mM EDTA, the inhibition of FASII activity by these metals was reduced, resulting in IC 50 values of 76 Ϯ 16, 4.0 Ϯ 0.9, 0.3 Ϯ 0.09, and 14 Ϯ 4.5 mM for Mg 2ϩ , Ca 2ϩ , Zn 2ϩ , and Mn 2ϩ , respectively. Neither potassium nor sodium demonstrated any inhibitory effects on FASII activity until the concentration reached 300 mM (data not shown). These experiments were repeated at least three times in duplicate.
Screening, Isolation, and Characterization of New Inhibitors-Using the FASII assay, we screened both natural product extracts and a collection of synthetic compounds (not shown). The assay performed adequately to identify inhibitors of fatty acid synthesis for enzymes involved in chain elongation. The hit rate was 0.13%, and the ZЈ factor (32) was 0.73. From actinomycete active broth S18 (Fig. 6), we isolated a new compound, BABX (Fig. 7A), which accounted for the FASII inhibitory activity of the extract. The details of isolation and structure elucidation will be published elsewhere. BABX inhibited FIG. 2. Determination of FASII activity. A, the assay was performed at 37°C as described under "Experimental Procedures" with increasing amounts of the partially purified FASII enzymes from E. coli (Ⅺ, 1.5 g; f, 3 g; E, 6 g; and •, 12 g). The fatty acid was slowly captured by phospholipids on the flash plate surface, and the elongated fatty acid was determined by measuring the incorporation of [ 14 C]malonyl-CoA at different times shown on the abscissa. Data are normalized to the background at zero time (202 cpm) determined by adding 100 l of 14% perchloric acid before the assay was initiated. B, FASII activity, presented as an inverse of each t1 ⁄2 , which was obtained from each data set of A, was plotted against the amount of FASII enzymes added. The goodness-of-fit of liner regression (r 2 ) is 0.9994. Data are means (ϮS.D.) of duplicate determinations, and similar results were obtained from repeated experiments. The specific activity of [ 14 C]malonyl-CoA was 56.9 cpm/pmol using Packard TopCount NXT as described under "Experimental Procedures." The maximum counts obtained from this experiment, however, were 6417 cpm, which implies the efficiency of flash plate versus liquid scintillation is 67.7%, giving 38.5 cpm/pmol.  7B). BABX exhibited good antibacterial activities against S. aureus (MIC ϭ 0.2 g/ml) and poor antibacterial activities against wild type E. coli (MIC Ͼ 250 g/ml). To determine the reason for the failure of this compound to reach its intracellular target in E. coli, we generated permeable E. coli (lpxC) and/or efflux-negative E. coli (tolC) strains (Table I). This compound was tested against these strains, and MIC values were determined. It showed MIC values of 0.24, Ͼ250, and 0.39 g/ml against E. coli (lpxC), E. coli (tolC), and E. coli (lpxC and tolC), respectively (Table II). Similar results were also observed in the Kirby-Bauer assay (Fig. 7D). In the whole cell labeling assay, BABX inhibited S. aureus phospholipid synthesis with an IC 50 value of 0.21 g/ml. However, it also inhibited DNA, RNA, protein, and cell wall syntheses with similar IC 50 values of 0.2-0.3 g/ml. Due to the lack of penetration of BABX in E. coli, the whole cell labeling assay was carried out using E. coli (lpxC), showing 51%, 35%, 34%, 8%, and no inhibition of phospholipids synthesis, DNA synthesis, protein synthesis, RNA synthesis, and cell wall synthesis at 100 g/ml, respectively (Table II). To determine the possible target of BABX, the FASII elongation assay was performed. When acetyl-CoA, a natural substrate, was used, it produced longer chains of acyl-ACP but weaker signals in S. aureus compared with E. coli (Fig. 7, lanes  2 and 10). However, when we used n-octanoyl-CoA, a substrate used in the FASII assay, as a substrate, it produced better signals compared with acetyl-CoA in S. aureus (Fig. 7, lanes 2  and 7). As a result, it afforded a better signal to background window that helped in the screening of natural product extracts against S. aureus enzymes. At 200 g/ml, BABX com-pletely blocked fatty acid elongation (Fig. 7, lanes 4 and 8) similar to that of cerulenin, unlike triclosan where an extra band can be seen (Fig. 7, lane 9). BABX did not inhibit the FabD reaction which catalyzes malonyl-CoA and holo-ACP to malonyl-ACP (Fig. 7C, lanes 1, 4, and 8).
Effects of Cell Permeability and Efflux on Fatty Acid Inhibitors-In the process of discovery and optimization of new antibacterials, it is evident that whole cell activity may vary independently of enzyme inhibition. Thus it is important to track their accumulation in cells by measuring their ability to specifically inhibit the target pathway (phospholipid synthesis, in this case) and cell growth as well as their enzyme inhibitory activity. The effects of cerulenin, thiolactomycin, and triclosan on macromolecular synthesis and MIC were examined as described under "Experimental Procedures." All three compounds selectively inhibited phospholipid synthesis. The IC 50 values of whole cell labeling for RNA, protein, and cell wall syntheses (data not shown) were similar to those of DNA synthesis. The results of MICs, whole cell labeling and IC 50 values in the FASII assay against E. coli (wild type), E. coli (lpxC), E. coli (tolC), E. coli (lpxC and tolC), and S. aureus are summarized in Table II. Rifampicin is a typical positive control against permeable Gram-negative bacteria. Novobiocin is a positive control substrate of the AcrA/B TolC efflux pump of E. coli. Erythromycin is subject to both the outer membrane barrier and efflux. The MIC results against the E. coli strains indicated that efflux played the primary role in modulating accumulation of the three compounds. In contrast, BABX had permeability characteristics similar to rifampicin, where TolC mediated efflux did not play a role.

DISCUSSION
Acetyl-CoA and malonyl-CoA have been used as substrates for the in vitro fatty acid synthesis assay for decades. Acetyl-CoA and malonyl-ACP are condensed by FabH to make acetoacetyl-ACP, which initiates the fatty acid elongation cycle. However, E. coli and S. aureus FabH can utilize short (up to six carbons) fatty acid chains as a substrate (22,33), which complicates the quantification of product formation. For example, cerulenin is a selective FabF/B inhibitor with an IC 50 range of 0.1-2 g/ml. However, the IC 50 of cerulenin for FabH is ϳ150 g/ml (21) (data not shown). As such, a concentration of cerulenin that is sufficient to completely inhibit FabF activity may still allow FabH to continue the elongation of fatty acid chains to a maximum length of eight carbons. Unless mass spectrometry is used, the eight carbon fatty acids can not be separated from long-chain fatty acids using ether extraction (data not shown). Using assay conditions described under "Experimental Procedures," with the substitution of acetyl-CoA for octanoyl-CoA or lauroyl-CoA, chain elongation inhibitors could be distinguished from fatty acid chain initiation inhibitors (Fig. 7C, data not shown).
As previously described (34), long-chain fatty acids can be converted to acyl-ACP by acyl-acyl carrier protein synthase (Aas) in vitro. However, using long-chain fatty acids as substrates for a high throughput assay is not easy due to their poor solubility. Conveniently, when water-soluble octanoyl-CoA or lauroyl-CoA and [ 14 C]malonyl-CoA were used as substrates, the assay worked very well. To confirm that the assay is truly using either octanoyl-CoA or lauroyl-CoA as a substrate, we titrated these substrates using the extracts of FASII enzymes from S. aureus or B. subtilis (Fig. 3). As the concentration of octanoyl-CoA or lauroyl-CoA increased, the [ 14 C]malonyl-CoA incorporation into longchain fatty acid increased in a concentration-dependent manner. The supporting data are visualized in Fig. 7C (lane 6). The affinity of S. aureus FASII enzymes for the substrates octanoyl-CoA or lauroyl-CoA was 10-fold higher than that of B. subtilis, whereas the V max of B. subtilis FASII enzymes for octanoyl-CoA or lauroyl-CoA was 5.8-fold higher than that of S. aureus. The S. aureus enzyme that catalyzes octanoyl-CoA or lauroyl-CoA to their ACP counterpart, which can be utilized by the enzymes for the fatty acid chain elongation, is unknown. Although it has been reported (35) that E. coli can carry out this reaction using condensing enzymes and that B. subtilis FabH1 can utilize octanoyl-CoA as a substrate, albeit poorly (36), there is no reported evidence that identifies which S. aureus enzyme converts long-chain (Ͼ6C) acyl-CoA to acyl-ACP. Further investigation could lead to a better understanding of fatty acid metabolism. It was not obvious that the FASII system of the important Gram-positive pathogen S. aureus would work under these conditions, because we have been unable to carry out the reaction with enzymes from Streptococcus pneumoniae and Streptococcus pyogenes.
Enzyme activities are often regulated by divalent cations. This high throughput assay has been developed for use in screening natural product extracts, as shown in Fig. 3. These extracts may contain metal ions, including divalent metals that modulate many enzyme activities, as well as chelators. To understand the effect of metals on FASII enzyme activity, we titrated Mg 2ϩ , Ca 2ϩ , Zn 2ϩ , and Mn 2ϩ (Fig. 4). The FASII enzyme activity was inhibited by all divalent cations tested in the order Zn 2ϩ Ͼ Ca 2ϩ Ͼ Mn 2ϩ Ͼ Mg 2ϩ . The addition of EDTA reduces free divalent metal concentrations, which can be calculated using Equations 1 and 2,  (37,38). Although the mechanism of action of regulation by divalent cations remains a scientific interest, addition of a significant concentration of EDTA to the assay screen eliminates false positives due to metal effects and chelation. Three known fatty acid synthesis inhibitors were evaluated in the high throughput screening format using extracts from E. coli, B. subtilis, and S. aureus as enzyme sources. All compounds inhibited FASII activity (Fig. 5). Because cerulenin is a covalent inhibitor (13,39) and triclosan is a slow binding inhibitor, having a very slow k off (40), their IC 50 values are completely dependent on assay conditions. The IC 50 values of thiolactomycin against FASII enzymes from B. subtilis and S. aureus are 110 and 13 g/ml, which are similar to the IC 50 values obtained in a single enzyme assay using S. aureus FabF and B. subtilis FabF (41,42). Because E. coli contains two ␤-ketoacyl-acyl carrier protein synthases (FabF and FabB), comparison of the IC 50 of thiolactomycin against FASII enzymes with a single enzyme (FabF or FabB) assay is required to ascertain each enzyme contribution. The IC 50 of thiolacto-mycin against E. coli FASII enzymes is 17.3 g/ml, which is similar to that obtained from the single enzyme assay with E. coli FabB (IC 50 ϭ 5.3-8.4 g/ml) (21,41). This is in agree-FIG. 6. Screening of natural product extracts. An example of performance of FASII assay in 96-well plates. S18 shows inhibition greater than 50%. The background noise is below 30% inhibition. Cerulenin was used as a positive control in each plate. Data are means (ϮS.D.) of duplicate determinations. The percent inhibition was calculated using the equation %INH ϭ 100 -100 ϫ (data Ϫ BG)/(MAX Ϫ BG). Background (BG) is determined using 1 mM of cerulenin and maximum enzyme activity (MAX) is obtained by using 3.3% Me 2 SO .   FIG. 7. Identification and characterization of a new inhibitor, bischloroanthrabenzoxocinone (BABX). A, structure of BABX isolated from S18 (Fig. 6).  6 -9). After the reaction, 10 l of each sample was directly applied to and resolved by a l6% polyacrylamide gel containing 4 M urea. The gel was blotted to a polyvinylidene difluoride membrane and visualized by using a PhosphorImager. Lane 1, a control of malonyl-ACP; lanes 2, 6, and 10, without inhibitors; lanes 3 and 7, 200 g/ml cerulenin; lanes 4 and 8: 200 g/ml BABX; lanes 5 and 9, 10 g/ml triclosan. The same samples were also resolved on l6% polyacrylamide gel containing 0.5, 2, and 3.7 M urea (data not shown), which was used for confirming the results. Similar experiments were repeated four times with reproducible results. D, Kirby-Bauer assay with minor modifications. Briefly, 20 ml of melted LB agar was maintained at 44°C, seeded with 0.75 ml (A 600 ϭ 0.3) of overnight culture and plated into an Omni plate (NUNC-Nalgene). After agar solidified and dried for 15 min, 10 g/ml of serial dilutions of BABX in LB containing 20% Me 2 SO were placed on the seeded agar plate and incubated at 37°C for 20 h. Zone sizes represent antibacterial activity. The results were confirmed by repeating three times. ment with the finding that FabB is the major target for thiolactomycin in E. coli (24). Inhibition of FASII activity by triclosan reached a plateau, but failed to reach 100% inhibition. A possible explanation for this is that the FabI inhibitor does not block the first cycle of [ 14 C]malonyl-CoA incorporation, and this hypothesis was confirmed in the FASII elongation assay shown in Fig. 7C. Comparing lanes 6 -9, lane 9 shows the inhibition by triclosan and contains an extra band, between C10:0-ACP and C12:0-ACP, that is most likely 10:1(⌬2t)-ACP, which is observed in all replications.
Whole cell activity of enzyme inhibitors may vary independently of enzyme inhibition. This is often due to variation in the ability of the compounds to reach their intracellular targets due to poor penetration (43) and/or active efflux (44,45). It is thus important in characterization and optimization of these compounds to ascertain their ability to accumulate in cells in parallel with their enzyme inhibitory activity. It is also important to show that the whole cell activity is due to specific enzyme inhibition (as opposed to off-target activity). Thiolactomycin, cerulenin, and triclosan resistance showed an association with efflux in Pseudomonas aeruginosa (20,46) as well as in E. coli (42). However, the role of outer membrane permeability and its relationship with the efflux of these fatty acid inhibitors have not been addressed. Therefore, we constructed the outer mem-brane-permeable and/or efflux-negative E. coli strains, which were used in this study in whole cell labeling, MIC, and cell free biochemical FASII assays. As described under "Experimental Procedures," both FASII and whole cell labeling assays were performed for 20 min, and MIC assays were done overnight. Therefore, the MIC presumably reflects the long term effects of accumulation of inhibitors inside the cell, whereas the IC 50 of whole cell labeling reflects the combination of potency and the net accumulation of inhibitors in a limited time. Direct comparison of FASII IC 50 values with those of PL labeling gives a measure of the effects of permeability and/or efflux. Comparison of IC 50 of PL labeling with MIC is more complex because multiple factors, such as rate of penetration versus efflux, mode of entry, nature of binding, feedback regulation, and others, play a role.
In this study, we provide both in vitro and in vivo evidence that the three fatty acid inhibitors are not only pumped out by efflux in a short period of time but are also subject to outer membrane permeability barriers. As expected, the IC 50 values of the cell-free biochemical FASII assay for each of the fatty acid synthesis inhibitors showed similar results with FASII enzymes from the four E. coli strains. MICs of the three inhibitors decreased (antibacterial activity increased) 2-fold against the outer membrane-permeable E. coli (lpxC) strain compared   c ND, not determined. d BABX was titrated starting at 100 g/ml and did not reach a plateau or 100% inhibition. At 100 g/ml, 51% of phospholipids synthesis was inhibited, showing some selectivity compared to cell wall (no inhibition), RNA (8%), protein (34%), and DNA (35%) syntheses. e More details on controls and BABX for macromolecular synthesis are in the Supplemental Data.
with its wild type parent strain. In contrast, MICs decreased 32-, Ͼ64-, and 125-fold against the efflux negative E. coli (tolC) strain for cerulenin, thiolactomycin, and triclosan, respectively, indicating that efflux plays a critical role for their antibacterial activities. Although it is possible that outer membrane-permeable (lpxC) E. coli may lack efflux mechanisms in addition to those requiring TolC, and hence also reflect efflux rather than a simple barrier to entry, our findings indicate that the MICs of the compounds are not strongly affected by the lpxC mutation, as would be expected if the lpxC effect were exerted via efflux.
In the whole cell labeling assay, however, IC 50 values of phospholipid (PL) labeling showed a 4.4-fold decrease for cerulenin, a 50-fold decrease for thiolactomycin and a 267-fold decrease for triclosan with the permeable E. coli (lpxC) strain compared with its wild type parent strain. Thus, the outer membrane appears to be a barrier to short term accumulation. With the efflux-negative E. coli (tolC) strain, IC 50 values of PL labeling for cerulenin, thiolactomycin, and triclosan also showed 7-, 120-, and 267-fold decreases, respectively. These findings demonstrate that both outer membrane permeability and efflux play roles in preventing the inhibitors from reaching their intracellular targets during short term exposure to the inhibitors. Consistent with this conclusion, the MICs of the permeable efflux-negative E. coli (lpxC and tolC) strain were similar to that of E. coli (tolC) for all three inhibitors, whereas the IC 50 values of PL labeling of the E. coli (lpxC and tolC) strain showed apparent additive effects caused by both the lack of efflux and increased permeability. The permeable efflux negative E. coli (lpxC and tolC) strain provides a tool to study the correlation of in vitro and in vivo activities of enzyme inhibitors in E. coli, whereas the single mutants (lpxC or tolC) are useful in tracking chemical optimization of enzyme inhibitors with antibacterial activity. In E. coli (lpxC and tolC), the IC 50 for cerulenin in PL labeling (3.55 g/ml) is close to that of the cell-free FASII (0.75 g/ml), which is consistent with the MIC (3.1 g/ml). For thiolactomycin, the IC 50 of PL labeling (0.28 g/ml) is closer to the MIC (3.1 g/ml) than that of FASII (24.6 g/ml), because the E. coli FASII assay does not involve FabH, one of the targets for thiolactomycin. On the other hand, although MICs for cerulenin and thiolactomycin are the same (3.1 g/ml) in this strain, their effect on short term PL labeling is disparate (3.55 and 0.28 g/ml, respectively). The reason for this is unknown but might be due to feedback up-regulation of FabH over time (affecting thiolactomycin), differential effects of covalent (cerulenin) versus non-covalent (thiolactomycin) binding, or off-target activities of cerulenin. The IC 50 of triclosan for PL labeling (0.0002 g/ml) is similar to its MIC (0.0004 g/ml). However, it is more potent than that for FASII (0.05 g/ml), suggesting the possible involvement of other unknown targets for triclosan activity in whole cells.
Interestingly, both IC 50 values of PL labeling (12.5 g/ml) and MICs (64 g/ml) are identical for cerulenin and thiolactomycin against S. aureus. However, the IC 50 values of FASII are 9-fold apart (1.5 and 13 g/ml, respectively), which likely reflects the fact that FabH does not play a role in the S. aureus FASII assay, although differential effects of permeability/efflux cannot be ruled out. The IC 50 of triclosan for PL labeling and its MIC against S. aureus are identical (0.002 g/ml), which is 4-fold better than the IC 50 for FASII (0.008 g/ml). Because little information about efflux and permeability on the fatty acid synthesis inhibitors for Gram-positive bacteria has been documented to date, the correlation of potency between in vitro and in vivo activities against S. aureus could not be determined. Further investigation of permeability and efflux would help to understand the mechanism of action of fatty acid synthesis inhibitors and resistance emergence on Gram-positive bacteria.
When we screened the natural product extract libraries with the FASII assay, we used an assay for inhibition of FtsZ polymerization as a counter screen (26). The S18 broth (Fig. 6) showed selective inhibition in FASII assay, leading to the discovery of a new compound, BABX, as an inhibitor of fatty acid synthesis. This compound showed in vitro enzyme inhibitory activity with IC 50 values of 11.4 and 35.3 g/ml in the S. aureus and E. coli FASII assays, respectively. BABX also showed potent antibacterial activities against S. aureus and permeable E. coli strains with MICs ranging from 0.2 to 0.4 g/ml. When evaluated in the whole cell labeling assay, BABX inhibited E. coli (lpxC) phospholipid synthesis (51%) with some selectivity compared with cell wall (no inhibition), RNA (8%), protein (34%), and DNA (35%) syntheses at 100 g/ml. BABX inhibited S. aureus phospholipid synthesis with an IC 50 of 0.21 g/ml, which supported the MIC and Kirby-Bauer assay results. However, the compound also inhibited DNA (as well as RNA, protein, cell wall) synthesis at similar concentrations. This indicates that inhibition of fatty acid synthesis is only one of the possible mechanisms for growth inhibition and illustrates the importance of evaluating the whole cell activity of inhibitors discovered in biochemical assays. In the E. coli case, the MIC in the lpxC strain was much lower than the IC 50 values for whole cell labeling of all macromolecular synthesis (phospholipid, DNA, protein, RNA, and cell wall) tested. Thus the primary target may be another system or growth inhibition may occur over a longer term than the labeling period.
Which fatty acid synthesis enzyme is a possible target for BABX? As we discussed earlier, any fatty acid synthesis inhibitor that inhibits elongation enzymes, other than condensation reactions, in the FASII assay initiated with octanoyl-CoA cannot stop the addition of first two carbons to form 10 carbons of ␤-ketoacyl-ACP, ␤-hydroxyacyl-ACP, or trans-2-enoyl-ACP (Fig. 1). From the FASII assay (Fig. 7B) and the FASII elongation assay (Fig. 7C), BABX did not inhibit FabD but fully inhibited acyl-ACP elongation, which is similar to that of cerulenin but different with that of triclosan, suggesting BABX is an inhibitor of the condensation enzyme in the elongation cycle. Further screening or potential chemical modification of BABX may lead to a better inhibitor of fatty acid synthesis and a more selective antibacterial agent.