Ciprofloxacin-Resistant Salmonella enterica Serovar Typhimurium Strains Are Difficult To Select in the Absence of AcrB and TolC

Abstract

It has been proposed that lack of a functional efflux system(s) will lead to a lower frequency of selection of resistance to fluoroquinolones and other antibiotics. We constructed five strains of Salmonella enterica serovar Typhimurium SL1344 that lacked efflux gene components of resistance nodulation cell division pumps (acrB, acrD, acrF, acrBacrF, and tolC) plus three strains that lack genes that effect efflux gene expression (marA, soxS, and ramA) and a hypermutable strain (mutS::aph). Strains were exposed to ciprofloxacin at 2× the MIC in agar, in the presence and absence of Phe-Arg-β-naphthylamide, an efflux pump inhibitor. Mutants were selected from all strains except those lacking acrB, tolC, or acrBacrF. For strains from which mutants were selected, there were no significant differences between the frequencies of resistance. Except for mutants of the ramA::aph strain, two phenotypes arose: resistance to quinolones only and multiple antibiotic resistance (MAR). ramA::aph mutants were resistant to quinolones only, suggesting a role for ramA in MAR in S. enterica serovar Typhimurium. Phe-Arg-β-naphthylamide (20 μg/ml) had no effect on the frequencies of resistance or ciprofloxacin MICs. In conclusion, functional AcrB and TolC in S. enterica serovar Typhimurium are important for the selection of ciprofloxacin-resistant mutants.

Salmonella enterica serovar Typhimurium is an important cause of food poisoning and is the second most common cause of bacterial diarrhea in the western world (8, 21). The fatality rate is dependent upon the country and serovar but is usually 1 to 4%; for antibiotic-resistant strains, this is increased at least twofold (20). Antibiotic resistance is an increasing problem in many countries, especially those where antibiotics have been intensively used in farming (31). Resistance is partly related to food animal reservoirs, and the pattern of antimicrobial resistance is due to the use of antimicrobials in specific animal species. For example, S. enterica serovar Typhimurium has a strong bovine association and multiple antibiotic resistance (MAR) is known to be associated with the use of antimicrobials in cattle (9) and clonal spread. Antibiotic use in veterinary medicine may lead to the emergence of resistance in bacteria pathogenic to humans, thus presenting a risk to public health from zoonotic infections such as those by S. enterica serovar Typhimurium (31).

Resistance to quinolones in S. enterica is chromosomal in origin and is typically caused by mutations in gyrA that result in amino acid substitutions in the GyrA subunit of DNA gyrase (31). Mutations in the genes encoding the beta subunit of DNA gyrase (gyrB) and the gene encoding the alpha subunit of topoisomerase IV (parC) have also been reported (5, 12, 19). Other mechanisms, such as a decrease in drug permeability and active efflux mechanisms, can also contribute to quinolone resistance (31). These latter mechanisms give rise to MAR. Extrusion (efflux) of antibiotics and other substances toxic to the bacterial cell is one of the mechanisms employed by S. enterica to resist the action of antibiotics. In S. enterica, the main efflux system is thought to be AcrAB-TolC (16, 32). AcrAB is a member of the resistance nodulation cell division family of transporters and is encoded by acrAB. The pump has three components: a transporter protein in the inner membrane (AcrB), a periplasmic accessory protein (AcrA), and an outer membrane channel (TolC) (38, 41). AcrB captures its substrates within the phospholipid bilayer (41) and transports them into the external medium via TolC (15). Cooperation between AcrB and TolC is mediated by the periplasmic protein AcrA. Expression of acrAB is controlled by either acrR, the MarRAB operon, including MarA, a positive regulator, or the SoxRS operon (36). It has been found that overproduction of the AcrAB-TolC efflux pump confers MAR to a variety of compounds in S. enterica serovar Typhimurium (16, 32). It has been postulated that expression of a functional AcrAB-TolC efflux system in Escherichia coli and S. enterica serovar Typhimurium is the first step in these organisms' becoming ciprofloxacin resistant (3, 4, 29, 30). Therefore, we hypothesized that S. enterica serovar Typhimurium that lacked components of tripartite efflux pumps would give rise to mutants with one of two phenotypes: (i) where the pump was important in resistance and only GyrA mutants would be selected and (ii) where the pump played little or no role in resistance both MAR mutants and GyrA mutants would be selected. A low-molecular-weight dipeptide amide, Phe-Arg-β-naphthylamide (MC-207,110), has been described as a broad-spectrum efflux pump inhibitor (EPI) for Pseudomonas aeruginosa (25). We hypothesized that inhibition of S. enterica serovar Typhimurium efflux pumps by 20 μg/ml EPI would lead to a similar phenotype as deletion of efflux pump genes. This concentration has been shown by others to reduce the MIC of fluoroquinolones for S. enterica serovar Typhimurium and other species of gram-negative bacteria (14, 25, 35).

The aims of this study were to determine the effect of a lack of efflux pump genes or their regulators upon the ability to select ciprofloxacin-resistant mutants of S. enterica serovar Typhimurium and to determine whether inhibition of efflux by a putative EPI had a similar effect as a lack of efflux pump genes. Finally, a strain lacking MutS was examined, as this is a mismatch repair gene leading to a higher frequency of mutation.

MATERIALS AND METHODS

Construction of knockout strains.

Eight knockout strains with single deletions in marA, ramA, soxS, mutS, tolC, acrB, acrD, and acrF and one double-deletion strain (acrBacrF) were constructed from the widely used strain S. enterica serovar Typhimurium SL1344 (40) by the one-step technique previously described for E. coli (10) and S. enterica serovar Typhimurium (11). An acrA-disrupted strain was not constructed, as acrA is cotranscribed with acrB (26).

Mutant selection.

Spontaneous mutants with decreased susceptibility to fluoroquinolones were selected as follows. The parent strains were grown overnight in antibiotic-free broth, concentrated by centrifugation, and resuspended in sterile broth to give a range of inocula (10⁶ to 10¹⁰ CFU/ml). Agar plates containing ciprofloxacin at 2× the MIC were inoculated with 100 μl (10⁵ to 10⁹ CFU) of each cell suspension and incubated at 37°C in air for up to 7 days. The limit of detection was ∼10⁻¹⁰ CFU/ml. The ciprofloxacin exposure was repeated in the presence of 20 μg/ml of Phe-Arg-β-naphthylamide (Sigma) in agar. Ten colonies with the typical size and morphology of the original strain were chosen randomly from each selecting plate and subcultured onto antibiotic-free media. These were examined for susceptibility to antibacterial agents, and one mutant selected under each condition for each observed resistance phenotype was retained for further study of the mechanism of resistance. All strains were cultured on Iso-Sensitest agar at 37°C in air. Bacteriological media were supplied by Unipath (Basingstoke, United Kingdom) and chemicals by BDH or Sigma (Poole, United Kingdom). Mutant selection was repeated on at least three separate occasions.

Determination of MICs of antibiotics, dyes, detergents, and disinfectants.

The MIC of each agent was determined by the doubling agar dilution method as previously described by the British Society of Antimicrobial Chemotherapy (http://www.bsac.org.uk) (2). All MICs were determined on at least three independent occasions. The MIC breakpoints used were those recommended by the British Society of Antimicrobial Chemotherapy (see Table 3), except for ciprofloxacin, where a cutoff value of 0.25 μg/ml was used to define resistance, as suggested by other workers (1, 39). Isolates that were resistant to at least three agents of separate classes of antibiotics (e.g., quinolone, chloramphenicol, tetracycline, or ampicillin) were deemed multiply antibiotic resistant.

TABLE 3.

Phenotypes and genotypes of all of the strains used in this study^a

Strain	Mutants selected	Selecting conditionb	Cipc (1)	Nal (16)	Chl (16)	Tet (2)	Amp (16)	Tric	Cycl	n-Fold change in acrB expressionf	Phenotyped	Substitution in QRDR of gyrA
L108 (tolC::aph)	−	None	0.015	1	1	0.5	0.5	0.008	Si	(0.7)		WTh
L644 (acrB::aph)	−	None	0.015	1	1	0.5	1	0.015	S			WT
L646 (acrB::aph/ΔacrF)	−	None	0.015	1	1	0.5	2	0.015	S			WT
L354 (SL1344)	+	None	0.03e	2	2	1	0.5	0.12	S	1		WT
L663		CIP	0.12	>256	4	2	1	0.5	S	0.8	FQr	Asp87 → Gly
L664		CIP	0.12	16	16	4	0.5	1	Tj	1.3	MAR	WT
L106 (acrD::aph)	+	None	0.03	8	4	2	1	0.06	S	1 (1.9)g		WT
L676		CIP	0.12	8	16	8	8	0.5	S	1	MAR	WT
L677		CIP	0.5	>256	16	2	2	0.5	T	1	MAR	Asp87 → Gly
												Ala131 → Gly
L561 (acrF::aph)	+	None	0.03	8	2	1	2	0.06	S	1 (2)		WT
L662		CIP	0.25	>256	4	1	2	0.5	S	0.9	FQr	WT
L678		CIP	0.25	>256	16	2	1	0.5	S	0.9	MAR	Asp87 → Gly
												Ala131 → Gly
L679		CIP	0.12	16	16	2	2	0.5	T	1	MAR	WT
L101 (marA::aph)	+	None	0.03	4	4	2	1	0.06	S	1 (0.9)		WT
L658		CIP	0.12	16	16	8	0.5	1	T	1	MAR	WT
L670		CIP	0.25	16	16	2	2	0.5	S	1.1	MAR	Asp87 → Gly
												Ala131 → Gly
L107 (soxS::aph)	+	None	0.03	4	4	2	1	0.5	S	1 (0.7)		WT
L680		CIP	0.12	16	4	2	2	0.25	S	0.9	FQr	Ala119 → Glu
L681		CIP	0.25	>256	16	2	2	0.25	S	0.8	MAR	Ser83 → Tyr
L103 (ramA::aph)	+	None	0.03	8	4	2	1	0.06	S	1 (1.1)		WT
L660		CIP	0.25	>256	4	2	1	0.5	S	0.7	FQr	Ser83 → Tyr
L674		CIP	0.25	>256	4	2	1	0.5	S	1.1	FQr	Ala119 → Glu
L102 (matS::aph)	+	None	0.03	8	4	2	2	0.06	S	1 (0.8)		WT
L659		CIP	0.5	32	8	4	1	0.5	T	0.9	MAR	WT
L672		CIP	0.25	>256	8	2	2	0.5	S	0.8	FQr	WT

^aCip, ciprofloxacin; Nal, nalidixic acid; Chl, chloramphenicol; Tet, tetracycline; Ami, amikacin; Gen, gentamicin; Amp, ampicillin; Tric, triclosan; Cycl, cyclohexane. Values in parentheses following drug names are British Society for Antimicrobial Chemotherapy breakpoint concentrations (4a).

^bCIP: 2 × the MIC of ciprofloxacin.

^cThe ciprofloxacin MIC was unaltered in the presence of Phe-Arg-β-naphthylamide (20 μg/ml).

^dFQ^r, fluoroquinolone resistant.

^eBold type indicates MICs for parent strains.

^fBoldface type in this column indicates statistically significant differences in acrB expression compared with the respective parent strain (=1 [100%]), as determined by Student's t test (P < 0.05).

^gValues in parentheses indicate expression compared with SL1344.

^hWT, wild type.

ⁱS, sensitive.

^jT, tolerant.

Detection of mutations in topoisomerase genes.

The quinolone resistance-determining regions of gyrA, gyrB, parC, and parE were amplified by PCR from the DNA of all isolates as described previously (11). Mutations were detected by denaturing high-performance liquid chromatography (HPLC) as previously described by Eaves et al. (11). DNA sequencing was performed at the Functional Genomics Laboratory at the University of Birmingham.

Expression of 16S rRNA, gyrB, tolC, and acrB.

RNA isolation; reverse transcriptase multiplex PCR amplification of 16S rRNA, gyrB, tolC, and acrB; and PCR product quantification by denaturing HPLC analysis were done exactly as described previously (11). The primers used in this study are listed in Table 1. Within bacterial cells, the level of 16S rRNA and gyrB was assumed to be transcribed at a constant rate throughout the growth conditions used in this study (11). Differences in amplification efficiencies among cDNA preparations were normalized by comparing the peak area of each amplicon resulting from the 16S rRNA and gyrB amplification to that of each amplicon resulting from the mean value of the control strain 16S rRNA amplifications. To gain an accurate mean value for 16S rRNA and gyrB levels on which to base the normalization, five independent PCR amplifications from the cDNA of three separate RNA preparations from each control strain were used. The peak areas for tolC and acrB were adjusted as necessary to compensate for variations between the mRNA levels in each RNA preparation (11). Data are presented as means ± standard deviations from the independent PCR amplifications. All comparisons between mutants and parent strains were compared by Student's t test. A P value of less than 0.05 was considered significant. In a separate study, protein expression was determined as described by N. G. Coldham et al. (unpublished data) by two-dimensional HPLC-mass spectrometry as described elsewhere (9). Proteins present in the cell envelope were identified and filtered with the SEQUEST (13) and DTAselect (37) algorithms, respectively, and exported to Excel. The data from the replicates were further sorted by protein and proteomes compared with Microsoft access to determine those common to all replicates. The relative abundances of the cell envelope proteins were compared with the spectrum count (24), and the statistical significance of changes in expression was evaluated by a two-tailed Student t test.

TABLE 1.

Primers used in this study

Gene	Primer	Primer sequence (5′→3′)	Amplimer size (bp)
acrB	StmacrB1	GACGTCCTATTTTCG	397
	StmacrB2	CGAAGACGCCTCTGT
tolC	StmtolC1	TGCCGACTACACCTACAGCA	341
	StmtolC2	GCTCTTCTACCGCGTTATCC
gyrB	StmgyrB1	GCGCTGTCCGAACTGTACCT	181
	StmgyrB2	TGATCAGCGTCGCCACTTCC
16S rRNA	Stm16SF	CCTCAGCACATTGACGTTAC	247
	Stm16SR	TTCCTCCAGATCTCTACGCA

RESULTS

Mutant selection.

Spontaneous mutants less susceptible to ciprofloxacin were selected from parent strain SL1344 and all knockout strains except L644 (acrB::aph), L108 (tolC::aph), and double-knockout strain L646 (acrB::aph/ΔacrF) (Table 2). Where spontaneous mutants were selected, the mutation frequencies were between 9.5 × 10⁻⁹ and 2.35 × 10⁻⁸, except for the mutS::aph strain, which gave rise to a higher frequency of spontaneous mutants. The presence of Phe-Arg-β-naphthylamide had no effect upon the frequency of mutation to ciprofloxacin resistance (data not shown).

TABLE 2.

Frequency of mutation for each strain

Strain	Frequency of mutationa
SL1344	5.5 × 10−9
L644 (acrB::aph)	0
L106 (acrD::aph)	1.2 × 10−9
L561 (acrF::aph)	4.3 × 10−9
L646 (acrB::aphlΔacrF)	0
L108 (tolC::aph)	0
L101 (marA::aph)	2.9 × 10−9
L107 (soxS::aph)	1 × 10−9
L103 (ramA::aph)	7.2 × 10−8
L102 (mutS::aph)	3.4 × 10−6

^aCiprofloxacin resistance.

Phenotypes of knockout strains and mutants.

The same MIC of ciprofloxacin was observed for all knockout strains, except L644 (acrB::aph), L646 (acrB::aph/ΔacrF), and L108 (tolC::aph), which were more susceptible (Table 3). These strains were also more susceptible to nalidixic acid, chloramphenicol, tetracycline, and triclosan. The mutants selected from SL1344, L106 (acrD::aph), L561 (acrF::aph), L101 (marA::aph), L107 (soxS::aph), and L102 (mutS::aph) were less susceptible to ciprofloxacin and gave two phenotypes. The first phenotype included mutants less susceptible to quinolones only (e.g., L660 and L663). The second phenotype comprised mutants that were multiply antibiotic resistant, i.e., resistant to three different classes of antimicrobials (e.g., L659 and L679). Mutants selected from L103 (ramA::aph) all had the first phenotype, and none had MAR. The MICs of ciprofloxacin and nalidixic acid were two- to fivefold higher for all of the ciprofloxacin-selected mutants than for the respective parent strains (Table 3).

Mutation of topoisomerase genes.

Denaturing HPLC and DNA sequencing revealed that mutants that were only resistant to quinolones had a single mutation in gyrA (Table 3). Those mutants that had MAR either had the wild-type quinolone resistance-determining region (QRDR) of GyrA or two substitutions in this protein; one was at Ser83 or Asp87, and the second was at Ala131. No mutations were observed in the gyrA gene of any of the mutants selected from L102 (mutS::aph). No mutations were found in gyrB, parC, or parE of any mutant.

Expression of acrB and tolC.

Irrespective of whether these data were normalized to 16S rRNA or gyrB expression, the strains with marA, soxS, ramA, or mutS disrupted expressed acrB at the same level as in parent strain SL1344. In the strains with acrD or acrF disrupted, expression of acrB was increased twofold (Table 3), as previously reported (12). Expression of tolC by all knockout strains was similar to that of parent strain SL1344 (data not shown). Ciprofloxacin-resistant mutants which arose from the knockout strains all had similar levels of acrB expression as their respective parent strains (Table 3). One MAR mutant (L664), arising from wild-type strain SL1344, showed a statistically significant increase in acrB expression. Data for eight knockout strains were confirmed by two-dimensional HPLC-mass spectroscopy (N. G. Coldham and M. J. Woodward, unpublished data).

DISCUSSION

Deletion of efflux pump genes acrD (L106) and acrF (L561) had no effect upon the mutation frequency. The strains lacking AcrD and AcrF have similar susceptibilities to fluoroquinolones as isogenic parent strain SL1344, suggesting that these proteins play little or no role in the efflux of these agents, despite the fact that AcrD and AcrF are 80% and 64% identical to AcrB (16). No mutants were selected from the strains lacking acrB (L644 and L646). Eaves et al. (11) and others (4, 22) previously showed that S. enterica serovar Typhimurium lacking AcrB is hypersusceptible to antibiotics. Recently, Randall et al. (unpublished data) successfully selected ciprofloxacin-resistant mutants from L644 but not L108. However, very high organism concentrations were required (>10¹¹ CFU/ml) and the frequency of mutation was very low (∼10⁻¹³). These data indicate that functional AcrB is important in the development of ciprofloxacin resistance.

No spontaneous mutants were selected from L108 (tolC::aph). Taken with the data for L644 (acrB::aph), this indicates that a functional tripartite efflux pump, AcrAB-TolC, is important in surviving ciprofloxacin treatment, and lack of any component affects this ability. We have previously shown that when acrB is deleted, expression of acrD and acrF is increased, suggesting coordinate regulation of these genes (12). In addition, these three efflux pump proteins have overlapping substrate profiles; therefore, lack of AcrB can in part be compensated for, hence the modest MIC changes observed with knockout mutants. However, expression of tolC was unaffected by lack of acrB, acrD, or acrF. Deletion of tolC gives rise to a strain that grows poorly compared with SL1344 (data not shown) and exposure to ciprofloxacin at twice the MIC was lethal, suggesting that S. enterica serovar Typhimurium requires functional TolC to become less susceptible to ciprofloxacin. E. coli lacking acrAB did not have MAR, despite overexpression of marA, indicating that none of the other pumps in E. coli can compensate for acrAB in conferring MAR (29). Baucheron et al. (3, 4) showed that quinolone resistance and decreased fluoroquinolone susceptibility in different S. enterica serovar Typhimurium strains are highly dependent on the AcrAB-TolC efflux system and that single mutations in the QRDR of gyrA are of little relevance in mediating this resistance. They also recently observed that chloramphenicol and tetracycline resistance also appeared to be highly dependent upon the presence of AcrAB-TolC. These observations are consistent with this study.

Deletion of the global regulators marA and soxS had no effect upon the mutation frequency or on the selected phenotype of the mutant. These data suggest that neither marA nor soxS is critical for S. enterica to generate a MAR mutant. Deletion of the ramA gene gave mutants that were quinolone resistant only. This finding requires further study, as it suggests that ramA is important in the development of MAR in S. enterica serovar Typhimurium. L102 had a deletion of the mutS gene, which is a mismatch repair gene. Deletion of this gene will lead to inefficient DNA repair and an increase in the mutation frequency (6). The frequency of mutation to ciprofloxacin resistance was higher for L102 than that for parent strain SL1344. However, no gyrA mutants were isolated, in contrast to the study by Levy et al. (23).

Where ciprofloxacin-resistant mutants were selected from the knockout strains, two phenotypes were observed, i.e., (i) resistance to quinolones only and (ii) MAR. L103 (ramA::aph) was the only exception, giving rise to quinolone-resistant mutants only. The mutants which were only resistant to quinolones all had a single substituting mutation in gyrA. This has previously been shown by complementation to be sufficient to confer the observed MIC (18). In addition, mutations at this locus have been shown to be the most common mechanism of fluoroquinolone resistance in S. enterica serovar Typhimurium (31). Those mutants that were MAR had either the wild-type QRDR of GyrA or two substitutions in this protein. Mutations in gyrA have not been previously associated with conferring MAR. However, in E. coli DNA supercoiling by DNA gyrase influences expression of ompF (17); it is also known that decreased expression of ompF confers MAR (27). The mechanism of MAR and any relationship with substitutions in gyrA in these mutants are being further explored.

MAR mutants without a mutation in gyrA could be resistant due to a decrease in antibiotic influx or increased efflux, or a combination of both. Previous work in this and other laboratories has shown the importance of the AcrAB efflux pump in quinolone resistance and MAR, as mutants of S. enterica serovar Typhimurium with reduced accumulation of ciprofloxacin have been shown to overexpress AcrA or AcrB (12, 16, 32). Furthermore, S. enterica serovar Typhimurium mutants that overexpress AcrAB were more resistant than the parent strain to a wide variety of compounds such as fusidic acid, norfloxacin, tetracycline, chloramphenicol, and penicillin (28). Mutants that had MAR fell into two groups, those that were tolerant to cyclohexane and those that were sensitive. Previous evidence suggested that cyclohexane tolerance in salmonellae may occur via a mar-dependent pathway (33). It has also been proposed that cyclohexane tolerance in salmonellae is associated with active efflux pump mechanisms (34).

In the presence of 20 μg/ml of Phe-Arg-β-naphthylamide, the frequency of resistance was similar to the frequencies in the absence of this agent. In addition, the MIC of ciprofloxacin was unaltered when Phe-Arg-β-naphthylamide was present. It may be that Phe-Arg-β-naphthylamide has no effect on the AcrAB-TolC efflux pump or that much higher concentrations are required to elicit an effect, as seen by Baucheron et al. (3). Lomovskaya et al. (25) showed that 20 μg/ml of Phe-Arg-β-naphthylamide suppressed the emergence of levofloxacin-resistant P. aeruginosa, so it could be that at this concentration Phe-Arg-β-naphthylamide does not affect the AcrAB-TolC efflux pump in S. enterica serovar Typhimurium SL1344.

Both human and veterinary strains of S. enterica serovar Typhimurium that have MAR due to increased expression of the AcrAB-TolC system have been isolated (31). It has previously been shown that overexpressing mutants can be easily selected by fluoroquinolone exposure in the laboratory, and it has been proposed that this can also easily occur in vivo. Such S. enterica serovar Typhimurium strains with MAR are of public health concern especially if they occur in animals and are transferred through the food chain. However, if inhibitors of the AcrAB-TolC system can be developed that can be administered to animals if a fluoroquinolone is used to treat an animal infection, selection of quinolone-resistant S. enterica serovar Typhimurium may be unlikely.