FabH Mutations Confer Resistance to FabF-Directed Antibiotics in Staphylococcus aureus

Abstract

Delineating the mechanisms for genetically acquired antibiotic resistance is a robust approach to target validation and anticipates the evolution of clinical drug resistance. This study defines a spectrum of mutations in fabH that render Staphylococcus aureus resistant to multiple natural products known to inhibit the elongation condensing enzyme (FabF) of bacterial type II fatty acid synthesis. Twenty independently isolated clones resistant to platensimycin, platencin, or thiolactomycin were isolated. All mutants selected against one antibiotic were cross-resistant to the other two antibiotics. Mutations were not detected in fabF, but the resistant strains harbored missense mutations in fabH. The altered amino acids clustered in and around the FabH active-site tunnel. The mutant FabH proteins were catalytically compromised based on the low activities of the purified enzymes, a fatty acid-dependent growth phenotype, and elevated expression of the fabHF operon in the mutant strains. Independent manipulation of fabF and fabH expression levels showed that the FabH/FabF activity ratio was a major determinant of antibiotic sensitivity. Missense mutations that reduce FabH activity are sufficient to confer resistance to multiple antibiotics that bind to the FabF acyl-enzyme intermediate in S. aureus.

INTRODUCTION

The type II fatty acid synthesis (FASII) pathway of bacteria has attracted attention as a viable target for the development of new drugs against multidrug-resistant Staphylococcus aureus (1). Although there was a debate regarding the effectiveness of FASII inhibitors in the presence of extracellular fatty acids (2–4), it is clear from biochemical analyses (5), and murine infection models (6–12) that FASII inhibitors are effective in vivo. Small molecules, such as AFN-1252 and CG02390, engineered to target the enoyl-acyl carrier protein (ACP) reductase (FabI) of S. aureus are being tested in human clinical trials (13, 14). The natural products cerulenin, platensimycin, platencin, and thiolactomycin (TLM) have antimicrobial activity against a broad range of bacteria and target the FabF elongation condensing enzyme (1, 6, 15–19) in S. aureus (Fig. 1). With the exception of cerulenin, which also inhibits mammalian fatty acid synthesis, each of the compounds has efficacy in animal infection models (6, 9, 16). Platencin is considered a dual FabH-FabF inhibitor, although the affinity for S. aureus FabH (SaFabH) is 10-fold lower than that for SaFabF (16). The most potent of the natural products targeting the condensing enzymes is platensimycin, a metabolite of Streptomyces platensis and a nanomolar inhibitor of bacterial elongation condensing enzymes. Platensimycin has a MIC of between 0.5 and 1.0 μg/ml for several important pathogens, including S. aureus, Enterococcus faecium, and Streptococcus pneumoniae. The potency of platensimycin and lack of toxicity to mammalian cultured cells (>1 mg/ml) has spurred several groups to design analogs with more favorable pharmacokinetic properties and easier synthesis pathways (20, 21). The condensing enzymes operate by a ping-pong mechanism. The first step is the acylation of the active cysteine by acyl-ACP, followed by the release of ACP. Next, malonyl-ACP binds, and the condensation reaction occurs. Crystal structures show that each of the antibiotics interacts with specific residues within the substrate binding tunnel and bind to the acyl-enzyme conformation of the elongation condensing enzymes (6, 16, 19, 22).

FIG 1.

Structures of the four natural product FabF inhibitors examined in this study. Platensimycin, platencin, and thiolactomycin are reversible inhibitors that all bind to the FabF acyl-enzyme intermediate. Cerulenin binds to the free enzyme, and activation of the epoxide in the active site leads to the covalent attachment of the drug to the active-site cysteine.

Common mechanisms for genetically acquired resistance to antibiotics include (i) alterations in the target gene that confer drug resistance while retaining catalytic activity, (ii) inactivation of transcriptional repressors controlling either target gene expression or drug efflux systems, (iii) changes in cell permeability, and (iv) enzymatic degradation of an inhibitor (23). Previous research showed that cerulenin resistance can arise from a missense mutation of the fabF gene of Bacillus subtilis (24), and resistance to platensimycin and platencin in the producing Streptomyces platensis strain is attributed to an elongation condensing enzyme isoform that is refractory to the compounds (25). Thiolactomycin resistance in Escherichia coli arises either from an inactivating mutation in a transcriptional repressor (EmrR) resulting in the activation of the EmrAB efflux pump (26) or through a specific missense mutation in the FabB target that reduces its affinity for thiolactomycin (27). Increased expression of the FabF condensing enzyme target also increases resistance to the antibiotics (6, 28, 29). The goal of this study was to identify the mechanisms of genetically acquired resistance to these natural products. We anticipated uncovering specific changes in FabF for each antibiotic tested that would reflect the distinct binding modes for the antibiotics. However, we found that genetically acquired resistance in S. aureus arises from missense mutations in the fabH gene that compromise the catalytic activity of the initiating condensing enzyme. Manipulating FabH and FabF expression levels shows that lowering the FabH/FabF activity ratio is sufficient to confer resistance to all FabF-targeted antibiotics. Thus, compromising FabH activity is a unique mechanism that confers resistance to antibiotics that target the FabF acyl-enzyme intermediate.

MATERIALS AND METHODS

Strains and materials.

The strains used in this study are listed in Table 1, and the primers used for plasmid construction and molecular analysis are listed in Table 2. Platensimycin and platencin were purchased from BioAustralis, and thiolactomycin was purchased from Sigma-Aldrich. [2-³H]isobutyryl coenzyme A ([2-³H]isobutyryl-CoA) (25 Ci/mmol) and [¹⁴C]malonyl-CoA (55 mCi/mmol) were purchased from American Radiolabeled Chemicals. When appropriate, strains were stored and propagated in the absence of inducer. Strains engineered to express individual proteins were handled with care, because IPTG (isopropyl-β-d-thiogalactopyranoside) induction often resulted in lower growth rates and if the strains were passaged in the presence of inducer, the plasmid expression was silenced. Specifically, overexpression of fapR, fabH, fabHF, and the antisense plasmids resulted in decreased growth rates, providing the selection pressure for silencing.

TABLE 1.

Bacterial strains and plasmids

Strain or plasmid	Description	Source or reference
Bacterial strains
RN4220	Restriction-negative strain	51
MWF55	RN4220 FabH(G203S)	This study
MWF56	RN4220 FabH(T146R)	This study
MWF61	RN4220 FabH(W32C)	This study
MWF62	RN4220 FabH(T37I)	This study
MWF64	RN4220 FabH(R212I)	This study
CS34	RN4220 ΔfapR	5
SA178R1	RN4220 derivative containing T7 polymerase gene (Tetr)	52
CS61	SA178R1 ΔfapR	This study
Plasmids
pET28b	Kanr	Novagen
pET15b	Ampr	Novagen
pPJ174	pET15b with the SafabF gene ligated into the BamHI site.	This study
pCS44	pET28b with the SafabH gene ligated into the NheI/BamHI restriction sites	This study
pG164	E. coli-S. aureus shuttle vector	52
pfabH	pG164 with the SafabH gene ligated into the BamHI and AvaI sites	This study
pfabF	SafabF cloned with a C-terminal FLAG tag, ligated into the EcoRI and HindIII sites of pG164	53
pfabHF	pG164 with the fabHF operon from S. aureus ligated into the EcoRI/HindIII sites of pG164	This study
pfapR	pG164 with the SafapR gene ligated into the BamHI and HindIII sites.	This study
pfabH-AS	SafabH antisense ligated into the BamHI and HindIII sites of pG164	This study
pfabF-AS	SafabF antisense ligated into the BamHI and NdeI sites of pG164	This study

TABLE 2.

Primers used in this study

Primer name	Sequence (5′→3′)a	Target vector
fabF-pet-for	GGATCCGAGTCAAAATATAAGAGTAGTTATTACA	pET15b
fabF-pet-rev	GGATCCTGTGCTGTCGCTCATCTTAG	pET15b
fabH-pet-for	ATGCTAGCGGTATTAAAGGTTTTGGTGC	pET28a
fabH-pet-rev	ATGGATCCCTATTTTCCCCATTTTATTG	pET28a
fabH-pg164-for	GGATCCATGAACGTGGGTATTAAAGGTTTTGGTGCATATGCG	pG164
fabH-pg164-rev	CTCGAGTTACTTGTCGTCATCGTCTTTGTAGTCTTTTCCC	pG164
fabF-pg164-for	GGATCCATGAGTCAAAATAAAAGAGTAGTTATTACAGGTATGGG	pG164
fabF-pg164-rev	GAATTCTTACTTGTCGTCATCGTCTTTGTAGTCTTTTCCC	pG164
fabHF-pg164-for	GGGGGGGTAGGAGGAGGATCCATGAGTCAAAATAAAAGAGTAGTTATTACAGGTATGGG	pG164
fabHF-pg164-rev	GCGCCGCCGCCGCGCAAGCTTTTATGCTTCAAATTTCTTGAATACTAATACTGC	pG164
fapR-pg164-for	AGAGGATCCATGGCAAGGGGTGAGACGTTGAAAC	pG164
fapR-pg164-rev	TCGAAGCTTACTTGTCGTCATCGTCTTTGTAGTCTCCTCGCTTATCATAAAACATTTTAAAATTTCC	pG164
fabF-AS-pg164-for	GGCTGTGGATCCTGCGCAATTAGATATCAATGAAAATACTG	pG164
fabF-AS-pg164-rev	CAATTGCATATGGCCCAGTTGCCATATCAGGAATTAAC	pG164

^aBoldface indicates the restriction sites used for cloning.

Selection of resistant mutants.

S. aureus strain RN4220 was grown to an absorbance at 600 nm (A₆₀₀) of 1.6 (2.0 × 10⁹ CFU/ml). A 200-μl (3.2 × 10⁸-CFU) aliquot was plated on Luria broth (LB) plates containing 4 μg/ml platensimycin, 4 μg/ml platencin, or 100 μg/ml TLM. These antibiotic concentrations were 4 times the MICs for these compounds. Selections performed at 1 and 2 times the MIC gave a lawn of cells, and individual clones could not be isolated. Plates were incubated at 37°C for 48 h and colonies restreaked on LB plates supplemented with appropriate compound. The fabHF operons of the resulting colonies were sequenced.

MIC.

The MICs for the different compounds against S. aureus were determined using a broth microdilution method. The medium was LB supplemented with 1 mM IPTG where indicated. For RN4220 strains and derivatives, cells were grown to an A₆₀₀ of 1.0 and diluted 30,000-fold in medium. A 10-μl aliquot of diluted cells was added to each well of a U-bottom 96-well plate containing 100 μl of medium with the appropriate concentration of compound. The plate was incubated at 37°C for 20 h and read using a Fusion plate reader at 600 nm. Cells grown in medium containing dimethyl sulfoxide (DMSO) were used as 100% growth. For SA178R1 strains expressing plasmids, a modified protocol was used due to plasmid instability when grown with inducer for extended periods of time. Cultures of SA178R1 were also grown to an A₆₀₀ of 1.0 but were diluted to 50-fold into the microtiter plate and grown at 37°C for 6 h.

Cloning of S. aureus genes into pET15b, pET28b, and pG164.

The mutant genes listed in Table 3 were amplified from S. aureus RN4220 genomic DNA using primers listed in Table 2 and were verified by sequencing. The fabH antisense plasmid was generated by Gene Art Gene Synthesis (Life Technologies) using the sequence 5′-ATAGAGATAGGATCCCTGTTATTTTAGATAATTTATCTGCACCGACAACTAAAATGTTATGATAATCTCCAGATTGAACATATTGTTTAGCTGTAATCATTGAATACATAAATCCAGAACATGCTGCAAGTTGATCCATAGAGGCAACTTTGCCCGTCCCTAAACGTTCTTGCAACATATTTGCGACAAAGCTTTGTAAGTAA-3′.

TABLE 3.

Mutations in fabH conferring resistance to FabF inhibitors^a

Amino acid change	No. of mutations				MIC (μg/ml) for:
	Total	From:
		Plcn	Pltm	TLM	Plcn	Pltm	TLM
None (strain RN4220)					0.78	0.78	25
−8, G to Ab	1			1	6.25	6.25	1000
Phe 22 Cys	1	1			50	25	1000
Trp 32 Cys	2	2			12.5	12.5	1000
Thr 37 Ile	2	1	1		50	25	1000
Gly 83 Val	1			1	12.5	12.5	1000
Thr 146 Lys	2		2		50	25	1000
Thr 146 Arg	1		1		25	12.5	1000
Met 201 Ile	1			1	6.25	6.25	1000
Gly 203 Ser	5	3	2		25	12.5	1000
Lys 208 Asn	1		1		25	12.5	1000
Ala 210 Thr	1			1	25	12.5	1000
Arg 212 Ile	3	2	1		50	25	1000
Pro 237 Leu	1			1	6.25	6.25	1000
Ala 240 Val	1	1			50	25	1000
Ala 248 Thr	1		1		50	25	500
Gly 301 Ser	1		1		25	12.5	1000

^aPlcn, platencin; Pltm, platensimycin; TLM, thiolactomycin.

^bThis nucleotide change was at position −8 with respect to the ATG initiation codon and is predicted to alter the ribosomal binding site.

Expression and purification of SaFabH and EcFabH.

Expression strains were cultured in LB medium until the A₆₀₀ reached 0.6, and isopropyl-β-d-thiogalactopyranoside (IPTG) was added to a final concentration of 1 mM. Cells were incubated with shaking at 16°C for 18 h. Soluble proteins were applied to an Ni-nitrilotriacetic acid (NTA) agarose column and washed with increasing concentrations of imidazole. FabH eluted with 400 mM imidazole and was subsequently applied to a chitin agarose column to remove contaminating SlyD protein. The flowthrough from the chitin column was further purified by separation on a Superdex S200 column (Amersham Biosciences) equilibrated and eluted with 20 mM Tris-HCl (pH 7.5), 0.1 M NaCl, 3 mM dithiothreitol (DTT), 3 mM EDTA, and 10% glycerol. E. coli FabH (EcFabH) was expressed in BL21(DE3) and purified as described previously (30).

Expression and purification of SaFabF.

The expression strain was inoculated in LB medium and grown to an A₆₀₀ of 0.6. The flask was heat shocked at 42°C for 20 min before the addition of 1 mM IPTG and growth at 25°C for 3 h. The cells were harvested by centrifugation, resuspended in lysis buffer [20 mM Tris-HCl (pH 8.0), 0.5 M NaCl, 10% glycerol, and 2 mM tris(2-carboxyethyl)phosphine)], and lysed. After centrifugation to remove debris, the supernatant was applied to Ni-NTA agarose. After elution from the column, FabF was purified further using a Superdex S200 column equilibrated in 20 mM Tris-HCl (pH 7.5), 1 mM EDTA, and 1 mM DTT.

Synthesis of acyl-ACP thioesters.

Lauryl-ACP was synthesized using Vibrio harveyi acyl-ACP synthetase and S. aureus ACP (5, 31). Malonyl-ACP was synthesized using S. pneumoniae AcpS (32, 33). Products were purified from reactions by anion-exchange chromatography and dialyzed overnight against 20 mM bis-Tris, pH 6.8.

FabH radiochemical assay.

S. aureus apo-ACP was expressed, purified, and converted to ACP as described previously (33). FabH was assayed in a reaction mixture containing 50 μM malonyl-ACP, 50 μM isobutyryl-CoA, 0.1 μl [2-³H]isobutyryl-CoA, 0.1 M sodium phosphate (pH 7.0), and 5 ng of purified FabH in a total volume of 50 μl. Reactions were initiated by the addition of FabH unless preincubation steps were required, in which case reactions were initiated by the addition of malonyl-ACP and isobutyryl-CoA. After initiation, the assay mixtures were incubated at 37°C for 12 min and the reaction mixture transferred to a Whatman No. 3 MM filter paper disc. The discs were washed 3 times for 30 min each in 10%, 5%, and 1% trichloroacetic acid before drying and scintillation counting.

FabF radiochemical assay.

FabF reaction mixtures contained 0.1 M sodium phosphate (pH 7.0), 50 μM [¹⁴C]malonyl-CoA, 100 μM lauryl-ACP, 1 μg SaFabD, 200 μM ACP, and 0.3 mM DTT in a volume of 20 μl. Dilutions of compound were made in 33% DMSO to maintain a constant DMSO concentration. Assays were initiated by the addition of 37.5 ng of SaFabF, and reaction mixtures were incubated at 37°C for 20 min. Samples were processed as described previously (19).

Liquid chromatography-MS/MS of cerulenin-treated FabH.

Two reaction mixtures were prepared to examine the covalent modification of FabH by cerulenin. Reaction mixture 1 contained 43 mM FabH, 66 mM sodium phosphate (pH 7.0), and 0.5 μl DMSO. Reaction mixture 2 contained the same components but with 0.5 μl of 447 mM cerulenin instead of DMSO (7.45 mM total cerulenin concentration). Reaction mixtures were incubated at 37°C for 1 h before the addition of 1 μl of 1 M DTT. Reaction mixtures were allowed to incubate for a further 20 min before the addition of 1 μl of 600 mM iodoacetamide and a second incubation at 37°C for 20 min. Finally, excess iodoacetamide was quenched through the addition of 3 μl of 1 M DTT and incubating the reaction mixture at 37°C for 20 min. The protein sample was digested with trypsin, and mass spectrometric analysis was performed using an Orbitrap mass spectrometer from Thermo Electron (San Jose, CA) by the St. Jude Proteomics Core Facility. The digest was introduced into the instrument via on-line chromatography using reverse-phase (C₁₈) ultra-high-pressure liquid chromatography on a nanoAcquity instrument (Waters, MA). Data acquisition involved acquiring the peptide mass spectrum (MS), followed by fragmentation of the peptide to produce a tandem mass spectrum (MS/MS) that provides information about the peptide sequence. Database searches were performed using raw files in combination with the Mascot search engine. Protein/peptide assignments were made on the basis of MS/MS.

Anti-FabH and anti-FabF immunoblotting.

FabF and FabH polyclonal antibodies were generated by Rockland Immunochemicals using the whole protein as an antigen. Cultures of the specified strains were grown to an A₆₀₀ of 1.0 and harvested by centrifugation at 4,000 × g for 10 min. Pellets were resuspended in 100 μl of phosphate-buffered saline (PBS) (pH 7.5) containing 1× Roche EDTA-free protease inhibitors, 1 mg/ml lysostaphin, and 0.1 mg/ml DNase I. Cell suspensions were incubated on ice for 3 h and centrifuged at 20,000 × g for 10 min. The supernatant was removed and protein quantified by Bradford assay. Samples were normalized by protein concentration and boiled with 1× Laemmli sample buffer before separation by SDS-PAGE. Proteins were transferred to a polyvinylidene difluoride (PVDF) membrane and blocked overnight in Tris-buffered saline containing 5% milk with 0.1% Triton X-100 (TBSTM). The primary antibody (anti-FabF or anti-FabH antisera) was added at a 1:200 dilution in TBSTM and incubated for 1 h. After washing 3 times for 15 min, the secondary antibody (goat anti-rabbit–alkaline phosphatase) was added at a 1:5,000 dilution and incubated for 1 h. After a final wash, the membrane was imaged using the Typhoon 9200 PhosphorImager and electrochemical fluorescence substrate (GE Healthcare).

qRT-PCR measurements.

Bacterial strains were grown in LB to an A₆₀₀ of 1.0, and the cells were harvested by centrifugation and washed with 1 ml of RNAlater solution (Ambion). Total RNA was then isolated and analyzed as described previously (34). Briefly, RNA was isolated immediately after cell harvesting with the Ambion RNAqueous kit (Applied Biosystems) per the manufacturer's instructions but with the inclusion of a lysostaphin treatment to lyse the cells. The Turbo DNA-free kit (Ambion) was used to remove genomic DNA, and cDNA was synthesized with Superscript II reverse transcriptase (Invitrogen). Quantitative real-time PCR (qRT-PCR) was run in triplicate using the ABI Prism 7700 sequence detection system with distilled water and RNA samples prepared without the reverse transcriptase step as negative controls. The gyrB gene was selected as the calibrator to compare mRNA levels between strains. The primers used were described previously (34). The expression levels of various housekeeping genes (proC, gyrB, gmk, glyA, rpoD, rho, recF, and pyk) were checked (35), and gyrB was determined to be the calibrator least changed by drug treatment and subsequently was used as the control to normalize mRNA levels between samples (see also reference 33). The values were compared using the threshold cycle (C_T) method (36) and the amount of cDNA present (2^−ΔCT) was reported with respect to the gyrB calibrator.

RESULTS

Mutations in fabH confer resistance to FabF-directed antibiotics.

S. aureus strain RN4220 (2 × 10⁹ cells) was applied to LB plates containing 4 μg/ml platensimycin, 4 μg/ml platencin, 100 μg/ml thiolactomycin, or 250 μg/ml cerulenin and allowed to grow for 48 h. Colonies (2 or 3 per plate) appeared with each compound used except cerulenin. We anticipated that each clone would have a missense mutation in fabF that would selectively confer resistance to the antibiotic used in the selection. However, DNA sequencing of the fabF genes in these resistant clones showed no mutations in the target gene, but rather each clone harbored a missense mutation in the fabH gene that resulted in a modified FabH protein (Table 3). In total, 15 different mutations in fabH were identified, and 5 of them were obtained more than once. The MICs for each of the mutants were tested against the panel of FabF inhibitors (Table 3), and each mutant was equally resistant to thiolactomycin, platensimycin, or platencin regardless of which compound was used to select for resistance. The isolates were resistant to only the FabF-targeted antibiotics. The MICs of the parent and of mutant strains MWF55, MWF61, and MWF64 (Table 1) were determined for the following: AFN-1252 (3.9 ng/ml), a FASII inhibitor at the FabI step; rifampin (7.8 ng/ml), an RNA synthesis inhibitor; amoxicillin (156 ng/ml), a cell wall synthesis inhibitor; clarithromycin (625 ng/ml), a protein synthesis inhibitor; and levofloxacin (156 ng/ml), a DNA synthesis inhibitor. There was no difference between the MIC for the wild-type strain RN4220 and that for any of the three mutant strains for these antibiotics. The simple idea that these FabF inhibitors actually block FabH in S. aureus was tested by the biochemical analysis of FabH and FabF inhibition by these antibiotics in vitro. As reported by others (6, 16, 19, 22), they were selective FabF inhibitors.

Cerulenin-resistant strains were not isolated.

Cerulenin is a fungal epoxide that irreversibly inhibits the elongation condensing enzymes (FabF) by covalent modification of the active-site cysteine (37). Thus, cerulenin was a natural product that was included in our screen; however, repeated attempts to isolate cerulenin-resistant S. aureus mutants failed. This was a frustrating outcome in light of the prior isolation of a cerulenin-resistant B. subtilis strain harboring a missense mutation that produced a FabF(I108F) mutant (24). Cerulenin differs from the other inhibitors in this study in that it binds to the free enzyme and not the acyl-enzyme intermediate. There are many reports describing cerulenin as a selective inhibitor of FabF with a much lower affinity for FabH (19, 38–40). To address this issue, we tested the potency of cerulenin in SaFabF and SaFabH enzyme assays. The 50% inhibitory concentration (IC₅₀) of cerulenin against S. aureus FabF (SaFabF) was 17 μM, consistent with a previously reported value of 22 μM (40). Cerulenin had a low affinity in the standard FabH assay, but when SaFabH and cerulenin were incubated for 15 to 60 min before addition of the substrates, cerulenin effectively blocked SaFabH activity up to 95% (Fig. 2A). Previous reports have indicated that cerulenin is a slow-binding inhibitor of E. coli FabB (EcFabB) (19). The requirement for the preincubation of cerulenin with FabH but not FabF perhaps reflects the differences in the on-rates for cerulenin between the two enzymes. To ensure that the ability of cerulenin to inhibit FabF and FabH is not solely a property of SaFabH, we also tested E. coli FabH (EcFabH). Cerulenin also inhibited the activity of EcFabH but only if the enzyme was incubated with cerulenin for 30 min before the addition of substrates (Fig. 2B). Cerulenin-treated SaFabH was subjected to matrix-assisted laser desorption ionization–time of flight (MALDI-TOF) MS/MS to determine if the inhibitory mechanism against SaFabH also functions by covalent modification of the catalytic cysteine. A mass of 223 was attached to the catalytic Cys112 in SaFabH, which corresponded to the loss of a proton and the addition of cerulenin, providing convincing evidence that cerulenin is an irreversible inhibitor of SaFabH (Fig. 2C). Cerulenin acting as a dual inhibitor explains the inability to isolate cerulenin-resistant mutants using our approach, because mutations would be required in both the fabH and fabF genes. An explanation for the isolation of a cerulenin-resistant B. subtilis FabF mutant (24, 41) was that B. subtilis expresses two redundant FabH enzymes (42), whereas S. aureus has only one.

FIG 2.

Inhibition of FabH by cerulenin. (A) Effect of preincubating SaFabH with 100 μM cerulenin before initiating the radiochemical assay by the addition of substrates. (B) Inhibition of SaFabH and EcFabH by 100 μM cerulenin with either no preincubation or a 30-min preincubation with the antibiotic. The 100% activity corresponds to the rate without a preincubation. The maximal activity for SaFabH (no preincubation) is 305 nmol/min/mg, and that for EcFabH (no preincubation) is 134 nmol/min/mg. (C) Mass spectrometry analysis of trypsinized FabH treated with cerulenin. Cerulenin was covalently attached to Cys112 of SaFabH as indicated by the increase in the mass of the y12, y13, and y14 fragments by 223.1 mass units.

The FabH mutants were catalytically defective.

The growth rates of strains harboring five of the most frequently occurring fabH mutations were determined (Fig. 3A). Each of these mutant strains grew significantly slower than the wild-type parent. In all cases, the growth defect was overcome by adding fatty acids to the growth medium (Fig. 3B). These data suggested that the reduced growth rate was due to a reduction in the rate of fatty acid synthesis that arose from defects in the FabH enzyme. All of the mutations except one resulted in an alteration in the fabH coding sequence. The other mutation was a G-to-A transformation that was predicted to occur in the ribosome binding site for fabH translation. The FabH mutations were mapped onto a published structure of SaFabH (43) (Fig. 4). The mutations clustered in and around the substrate binding tunnel. Several mutations were in residues directly interacting with the acyl-CoA substrate, either in the tunnel accommodating the acyl-phosphopantetheine moiety or in Trp32, a residue that stacks with the adenine of CoA (Fig. 4). FabH(G203S), the most frequent mutant identified in the screen, was purified. FabH(G203S) had a 10-fold decrease in FabH specific activity [1.38 ± 0.06 μmol/min/mg for FabH(G203S) versus 17.3 ± 0.90 μmol/min/mg for the wild type], illustrating that the mutant FabH(G203S) was catalytically defective. The apparent K_m for isobutyryl-CoA was reduced from 30.4 ± 2.0 μM in the wild type to 4.6 ± 0.61 μM in the FabH(G203S) mutant, whereas the apparent K_m for malonyl-ACP increased from 10.7 ± 0.86 μM in the wild type to 31.5 ± 2.7 μM in the FabH(G203S) mutant. We also purified FabH(W32C) and FabH(R212I) mutants. However, neither of these mutant proteins possessed sufficient catalytic activity in vitro for kinetic analyses. These data showed that the primary effect of the mutations was to significantly compromise the overall activity of FabH.

FIG 3.

Growth phenotypes of fabH mutant strains resistant to FabF antibiotics. (A) Growth of S. aureus wild-type strain RN4220 and five selected fabH mutant strains in LB medium, illustrating the slow-growth phenotype of the resistant clones. (B) Growth of strains RN4220 and MWF55 in LB medium with and without 1 mM anteiso15:0-anteiso17:0 fatty acids (2:1). The strains used with mutant FabH proteins expressed were MWF55 [FabH (G203S)], MWF56 [FabH (T146K)], MWF61 [FabH (R212I)], MWF62 [FabH (T37I)], and MWF64 [FabH (W32C)].

FIG 4.

Locations of the FabH mutations on the S. aureus FabH. (A) Alignment of S. aureus FabH with E. coli FabH. The black background indicates identical residues and the gray background conserved residues. Red boxes mark the locations of the mutations found in this study. Blue boxes mark residues involved in substrate binding (54). (B) S. aureus structure (PDB 3IL7). The SaFabH structure is represented by a cartoon model with the length of the substrate binding tunnel oriented with the entrance to the top rotated toward the viewer. The rear-facing monomer has a light gray surface model overlay. The mutated residues are in cyan with a stick presentation and are mapped onto the other monomer. Acetyl-CoA is placed into the active site by structure alignment with the acetyl-CoA-bound E. coli FabH structure (PDB 1HND). CoA carbons are green, phosphorus is orange, nitrogen is blue, and oxygens are red. FabH mutations are located either within the substrate binding tunnel or in residues associated with the helices that form the active site.

Gene expression in strains with mutant fabH.

Expression of the fab genes in S. aureus is regulated by the FapR transcription factor (44–47). In S. aureus, the fabH and fabF genes are organized into an operon that is tightly controlled by FapR (44–46). FapR is a feed-forward regulator that detects changes in the FASII precursor malonyl-CoA and depresses the fabHF operon in response to the inhibition of the pathway (44). Therefore, we used qRT-PCR to measure the transcript levels of fabH and fabF to determine if their expression levels were elevated, indicating a defect in FASII in vivo. Three mutant strains were analyzed, and in each case fabH and fabF expression was significantly increased (Fig. 5). The accD gene is not part of the FapR regulon and was not altered in the fabH mutant strains. However, the increase in fabH and fabF expression in the mutant strains was not as great as observed in strain CS34 (ΔfapR) (Fig. 5). This suggested that strain CS34 may be more resistant to the FabF inhibitors due to the increased expression of FabF; however, the MIC for all three antibiotics was the same in the fapR knockout strain as in its wild-type parent (platensimycin, 0.78 μg/ml; platencin, 0.78 μg/ml; and thiolactomycin, 25 μg/ml). This observation was on the surface inconsistent with the reports that the MICs for platensimycin and platencin were increased by overexpressing fabF or decreased by expressing small interfering RNA (siRNA) directed against fabF (6, 15, 16). Also, FabB overexpression increases thiolactomycin resistance in E. coli (27). These data suggested that resistance was simply related to the level of FabF expression. However, when fabF expression was increased by inactivating the FapR repressor in strain CS34 (ΔfapR), the strain did not exhibit increased resistance to FabF-targeted antibiotics. The plasmid-driven expression of FabH(G203S) in strain RN4220 (platensimycin MIC = 0.78 μg/ml) did not alter the MIC for the FabF antibiotics, showing that FabH(G203S) was a recessive rather than a dominant mutation. The expression of wild-type FabH using the same plasmid system did not alter the MIC in the wild-type strain (platensimycin MIC = 0.78 μg/ml), but when introduced into strain MWF55 [FabH(G203S)], it reduced the MIC for the FabF antibiotics to wild-type levels (platensimycin MIC = 0.78 μg/ml). Thus, restoring FabH activity to the fabH mutant strains returned their sensitivity to platensimycin to the low MIC characteristic of wild-type strains.

FIG 5.

Elevated expression of the fabHF operon in the mutant strains. qRT-PCR of fabH, fabF, and accD mRNAs in strains MWF55 [FabH(G203S)], MWF61 [FabH(R212I)], and MWF64 [FabH(W32C)] was performed. Expression levels were normalized to that of the wild-type protein, from strain RN4220.

Independent manipulation of FabH and FabF levels.

Constructs were designed to selectively manipulate the expression of fabH and fabF, and the protein levels were monitored by Western blot analysis. The strain used for these experiments was SA178R1, which had a slightly lower MIC for platensimycin than strain RN4220. The large elevation in fabHF mRNA in the ΔfapR strain (Fig. 5) resulted in a less substantial increase in both FabH (1.7-fold) and FabF (1.6-fold) protein levels (Fig. 6A). This coordinate increase in both proteins did not affect the susceptibility to platensimycin (Fig. 6B). Superrepression of the fabHF operon by elevated FapR expression resulted in a 50% decrease in both FabH and FabF proteins (Fig. 6A). The lower levels of these key condensing enzymes resulted in a significant IPTG-dependent growth defect but did not change the MIC for platensimycin (Fig. 6B). Substantially increased FabH protein by expression of pfabH (3.5-fold) (Fig. 6A) did not alter the platensimycin MIC (Fig. 6B). Increasing fabF expression using pfabF resulted in a 4-fold increase in resistance that correlated with a substantial increase (3.2-fold) in FabF protein (Fig. 6). The experiments with the pfabHF plasmid were complicated by the presence of an internal promoter in the fabHF operon that led to a 2.7-fold increase in the levels of FabF and a 16-fold increase in the MIC for platensimycin in the absence of inducer (Fig. 6). The expression of the fabH siRNA construct (pfabH-AS) decreased FabH protein by 50% but had a negligible effect on FabF levels (Fig. 6A). The expression of this antisense construct might be expected to decrease both FabH and FabF equally because they are in the same operon; however, the existence of a fabF promoter within the fabHF operon means that FabF was not as affected as FabH when expression of the antisense construct was induced (Fig. 6A). This small decrease in the FabH/FabF ratio resulted in a 4-fold increase in the MIC (Fig. 6B). The expression of the pfabF-AS antisense construct decreased the levels of both FabH and FabF by 50% (Fig. 6A) and led to a 2-fold decrease in the platensimycin MIC (Fig. 6B). Taken together, these data were consistent with the FabH/FabF protein ratio being a key determinant of a strain's sensitivity to platensimycin.

FIG 6.

Correlation between the levels of FabH and FabF expression and the platensimycin MIC. (A) Immunoblot of S. aureus strain SA178R1 expressing different genes cloned into the IPTG-regulated expression plasmid pG164. Immunoblotting was performed on strains grown in either the presence or absence of 1 mM IPTG inducer as indicated. FabH and FabF proteins were detected using specific polyclonal antibodies as described in Materials and Methods. The protein levels were compared using ImageQuant software to integrate the intensity of the bands. (B) MICs of strain SA178R1 to platensimycin and the effect of modulating FabH and FabF protein levels on the MIC for the antibiotic. Strains used were as follows: pG164, parental strain SA178R1 with an empty vector; ΔfapR, strain SA178R1 with a fapR deletion; pfapR, strain SA178R1 with the plasmid expressing FapR; pfabH, strain SA178R1 with the plasmid expressing FabH; pfabF, strain SA178R1 with the plasmid expressing FabF; pfabHF, strain SA178R1 with the plasmid expressing the fabHF operon; pfabH-AS, strain SA178R1 with the plasmid expressing antisense against fabH; and pfabF-AS, strain SA178R1 with the plasmid expressing antisense against fabF.

DISCUSSION

This work identifies a unique mechanism for genetically acquired resistance to the natural product FabF inhibitors platensimycin, platencin, and thiolactomycin in S. aureus. At the onset, we anticipated detecting specific mutations in the fabF gene that would render the mutant FabF enzymes refractory to inhibition by the drug while retaining some activity. These mutations would be specific for the antibiotic used for selection, because each antibiotic has a different mode of binding to the acyl-enzyme intermediate. However, every resistant clone isolated had a wild-type fabF gene, and instead each harbored one of several different mutations in the fabH gene. Regardless of which antibiotic was used for selection, the resulting mutant strains acquired resistance to all three antibiotics (platensimycin, platencin, and thiolactomycin). The fabH mutations give rise to defective enzymes based on in vitro biochemical analysis, in vivo gene expression profiles, and impaired cell growth that is rescued by exogenous fatty acids. Whether or not this impairment would impact the proliferation of the pathogen in infection models is unknown. S. aureus can incorporate host fatty acid into its membrane phospholipids (5, 48), and strains harboring the fabH mutations grew normally in the presence of a fatty acid supplement in the laboratory, suggesting that the fabH mutant strains may retain their pathogenic potential. However, mammalian fatty acids are not well tolerated by S. aureus, and the fatty acid transport system is not robust (12), meaning that this conclusion cannot be made without experimental validation.

The individual manipulation of FabH and FabF expression levels showed that resistance arises from an increase in the FabH/FabF activity ratio. As expected (6, 15, 16, 27), increasing FabF expression increases resistance to all three antibiotics and decreasing FabF expression using antisense technology increases sensitivity to the drugs. When the FabF levels are increased or decreased coordinately with FabH by manipulating the FapR repressor, there is no alteration in the MIC, demonstrating that resistance is not determined solely by the expression level of FabF. Our finding that compromising FabH activity through either mutations or antisense technology also increases S. aureus resistance to all FabF-targeted antibiotics tested illustrates the role of FabH activity in determining resistance. Plasmid-driven FabH expression will convert resistant strains with mutant fabH alleles to wild-type sensitivity for FabF drugs but will not render wild-type cells more sensitive to the FabF inhibitors. This effect is attributed to the regulatory role of FabH in controlling the amount of acyl-ACP substrate in the elongation cycle (30, 49, 50). Stringent biochemical regulation at this step means that more fatty acids are not made in cells overexpressing FabH, whereas a decrease in FabH expression leads to reduced fatty acid synthesis and growth rate.

There are two potential processes by which FabH activity may alter the sensitivity of FabF to the natural product inhibitors. FabF uses a ping-pong kinetic mechanism with an acyl-FabF intermediate. Platensimycin, platencin, and thiolactomycin selectively bind to the FabF acyl-enzyme intermediate (“pong”) conformation and have very low affinity for the FabF (“ping”) conformation (6, 16, 22). Cocrystal structures commonly employ mutant FabF enzymes with the active-site cysteine mutated to a glutamine to lock the protein in the “pong” conformation (6, 16, 22). Because FabH controls the amount of acyl-ACP in the elongation cycle, it directly controls the proportion of FabF that is present as the acyl-enzyme intermediate and available for binding this class of antibiotics. Decreasing FabH activity is predicted to limit the amount of acyl-ACP available to FabF, effectively reducing the amount of acyl-enzyme, the actual antibiotic target, in vivo. A second possibility is that lower FabH activity increases the malonyl-ACP level. The FapR transcription factor is a malonyl-CoA sensor (44–46). Thus, the activation of fabHF transcription in the strains with mutant fabH alleles suggests that malonyl-CoA levels are elevated. Malonyl-CoA and malonyl-ACP are interconverted by the FabD transacylase, suggesting that malonyl-ACP levels rise with malonyl-CoA. Malonyl-ACP is the second substrate for FabF and competitively blocks antibiotic binding to the FabF acyl-enzyme intermediate. There is not a reversible FabF inhibitor that targets the “ping” rather than the “pong” FabF conformation, which would allow a direct test of this idea. The mechanism for acquired resistance to multiple FabF antibiotics through mutations in another gene in the pathway is similar to the report that inactivation of acetate kinase confers platensimycin resistance (29). The understanding of how metabolism is altered in these two mutants to increase the resistance to FabF inhibitors will involve quantification of the FabF acyl-enzyme, acetyl-CoA, malonyl-CoA, malonyl-ACP, and acyl-ACP pool levels and is beyond existing technical capabilities. The discovery that FabH mutations are the most common mechanism for resistance reveals a unique mechanism that should be considered by research programs developing elongation condensing enzyme inhibitors for clinical use.