Growth Phase-Dependent Expression of Drug Exporters in Escherichia coli and Its Contribution to Drug Tolerance

Drug exporters contribute to the intrinsic drug resistance inmany organisms. Although there are at least 20 exporter genesin Escherichia coli, most of them apparently do not confer drugresistance in complex laboratory media except for the AcrAB,EmrE, and MdfA efflux systems. In this study, we comprehensivelyinvestigated the growth phase-dependent expression of drug exportergenes. The expression of acrAB, emrAB, emrD, emrE, emrKY, mdfA,and ydgFE is stable at moderate levels during any growth phase,whereas mdtEF promoter activity greatly increased with cellgrowth and reached the maximum level at the late stationaryphase. The growth phase-dependent increase in mdtEF expressionwas also observed on quantitative reverse transcription-PCRanalysis. As expected from the transporter expression, the stationary-phasecells actually showed MdtEF-dependent tolerance to drugs andtoxic dyes. Growth phase-dependent elevation of mdtEF expressionwas found to be mediated by the stationary-phase ${sigma}$ factor rpoSand the RpoS-dependent signaling pathway, Hfq, GadY, and GadX.The induction level was decreased by tnaAB deletion, suggestingthat indole sensing stimulates this process.

INTRODUCTION

Top

Introduction

Bacterial intrinsic tolerance to a wide range of antimicrobialagents is often caused by active efflux systems, such as AcrABin Escherichia coli and MexAB in Pseudomonas aeruginosa (17,18, 23), and multidrug-resistant mutants due to the overexpressionof efflux pumps have been isolated in clinical settings (15).Our previous studies revealed that there are at least 20 drugexporter genes in E. coli that confer drug resistance when theyare overexpressed (19); however, the previous study showed thatmost of them do not contribute to drug tolerance in complexlaboratory media, probably due to their low expression levels,except for acrAB, emrE, and mdfA (29). In that study, drug resistancewas determined as MICs, which would not reflect the actual expressionlevels of drug exporters at different growth phases.

In E. coli, drug exporter gene expression is affected by variousenvironmental stresses. For instance, the acrAB gene is knownto be induced by ethanol, osmotic shock, oxidizing agents (11),and bile salts and fatty acids (24). Throughout bacterial growth,the bacterial cell density, nutrient conditions, pH, and otherfactors are changing. Therefore, it would be important to studythe growth phase dependency of the expression of drug exportersthat may facilitate understanding of drug exporter-mediatedmultidrug resistance at actual infection sites.

The growth phase-dependent expression of various genes in E.coli has been reported. For instance, quorum-sensing signalmolecule autoinducer 2, which is produced and secreted intothe culture medium at the logarithmic phase, influences theexpression of type III secretion system-related genes (27) andmotility-related genes (28). The ${sigma}$ factor, RpoS, and RpoS-dependentgenes are known to be induced at the stationary phase (7). However,there is little information available on growth phase-dependentexpression of drug exporter genes.

In this study, we comprehensively investigated the expressionof the 20 drug exporter genes at different growth phases andtheir contribution to growth phase-dependent drug tolerance.We found that out of the 20 drug exporter genes in E. coli,the expression levels of the emrA, emrD, emrK, and ydgF genesare relatively stable at moderate levels without a significantchange throughout the bacterial growth phase as well as thoseof the acrA, emrE, and mdfA genes. In contrast, mdtEF gene expressionwas significantly increased at the late cell growth phase, followedby mdtEF-dependent drug tolerance. A possible regulation schemeis discussed.

MATERIALS AND METHODS

Top

Materials and Methods

Bacterial strains and plasmids. The bacterial strains and plasmids used in this study are presentedin Table 1. The construction of gene deletion mutants of E.coli MC4100 (2) was performed by the gene replacement methodpreviously described, using the pKO3 plasmid (9). E. coli cellswere cultured in Luria-Bertani (LB) medium, supplemented withappropriate antibiotics when necessary.

View this table:
[in this window]
[in a new window]
TABLE 1. Bacterial strains and plasmids used in this study

Construction of reporter plasmids. The reporter plasmids were constructed as follows. DNA fragmentsincluding the putative promoter region were amplified by PCRusing the primers listed in Table 2. Chromosomal DNA of E. coliMG1655 (1) was used as a PCR template. The DNA fragments werecloned in front of the lacZ reporter gene in a single-copy pNN387vector, which carries chloramphenicol resistance as a marker(3). Since the emrAB, cusBA, and mdtEF genes are transcribedin the emrRAB (10), cusCFBA (4), and gadE-mdtEF (14; our unpublisheddata) operons, respectively, emrR, cusC, and gadE promoter-fusedlacZ are reporters for the respective exporter expression. Theresulting plasmids were introduced into host cells for ß-galactosidaseactivity measurements.

View this table:
[in this window]
[in a new window]
TABLE 2. Oligonucleotides used for plasmid construction

Reporter gene assay. To investigate the growth phase-dependent transcription of variousreporter constructs, each bacterial strain was grown at 37°Cin LB broth containing 15 mg/liter chloramphenicol until theoptical density at 600 nm (OD₆₀₀) reached 0.4 (early logarithmicphase), 0.8 (late logarithmic phase), 3.0 (early stationaryphase), or 6.5 (late stationary phase). When we determined theeffect of indole on the transcription of mdtEF, cells were grownat 37°C in LB broth until the OD₆₀₀ reached 0.8 with 15mg/liter chloramphenicol and 1 mM indole or only with 15 mg/literchloramphenicol and the solvent (dimethyl sulfoxide) as a control.ß-Galactosidase activity in cell lysates was assayedusing o-nitrophenyl-ß-D-galactopyranoside (ONPG) asa substrate as described by Miller (16), with a slight modification.

Quantitative real-time RT-PCR. Quantitative real-time reverse transcription-PCR (RT-PCR) wasperformed as follows. Cells were grown at 37°C in LB brothuntil the absorbance at 600 nm reached 0.8 (logarithmic phase),3.0 (early stationary phase), or 6.0 (late stationary phase).The purification of total RNAs and the synthesis of cDNAs wereperformed by the methods described previously (6). The specificprimer pairs are listed in Table 3. Real-time PCR was performedwith each specific primer pair, using SYBR green PCR mastermix (PE Applied Biosystems). The E. coli rrsA gene was chosenas a control for normalization of the cDNA loading in each PCR.The reactions were performed with an ABI PRISM 7000 sequencedetection system (PE Applied Biosystems).

View this table:
[in this window]
[in a new window]
TABLE 3. Oligonucleotides used for quantitative real-time PCR

Drug tolerance assay. Each bacterial strain, grown at 37°C in LB broth until OD₆₀₀reached 0.8 (logarithmic phase) or 6.5 (stationary phase), wasdiluted to an OD₆₀₀ of 0.1. Then, growth was measured in thepresence or absence of crystal violet or kanamycin.

Survival assay. Each bacterial strain was grown at 37°C in LB broth untilthe OD₆₀₀ reached 0.6 (logarithmic phase) or 6.5 (stationaryphase). The stationary-phase cells were diluted to an OD₆₀₀of 0.6 with fresh medium, and then crystal violet was addedto each bacterial cell culture (final concentration, 50 or 200mg/liter). After incubation for 30 min at 37°C, aliquotsof the cell cultures were spread on YT agar plates. After overnightincubation, the numbers of colonies were determined and percentsurvival was calculated in comparison with that of untreatedcells.

Indole production assay. The extracellular indole concentration was determined by high-performanceliquid chromatography (HPLC). The E. coli strain was culturedat 37°C and then pelleted by centrifugation at 20,000 xg. The resulting supernatant was extracted twice with ethylacetate. The ethyl acetate phase was loaded onto a SymmetryC₁₈ column (5 µm, 4.6 by 150 mm; Waters Corp.) attachedto an L2130 HPLC system (HITACHI). The loaded samples were elutedwith acetonitrile-H₂O (1:1) at the flow rate of 0.8 ml/min.Then the indole peak was detected relative to the absorbanceat the wavelength of 276 nm and was identified by the correspondingpeak of the purified indole (Sigma). The indole concentrationwas calculated from the ratio of the detection peak area tothe standard peak one.

RESULTS

Top

Results

The growth phase-dependent expression of the intrinsic drug exporter genes in Escherichia coli. In order to profile the growth phase-dependent expression levelsof intrinsic drug exporter genes, we examined their reporterenzyme activities at different growth phases using single-copyplasmids containing lacZ-fused gene promoters. Because the emrAB,cusBA, and mdtEF genes are transcribed as a part of the operonsof emrRAB (10), cusCFBA (4), and gadE-mdtEF (14 and our unpublisheddata), respectively, emrR, cusC, and gadE promoter-fused lacZconstructs were used as reporters for these three genes. Eachbacterial strain was grown until the OD₆₀₀ reached 0.4 (earlylogarithmic phase), 0.8 (late logarithmic phase), 3.0 (earlystationary phase), or 6.5 (late stationary phase), and thenß-galactosidase activities were measured. Among the20 exporter genes, only expression of the mdtE gene greatlyincreased with growth cessation (Fig. 1). Compared with thelate logarithmic phase, 14- and 41-fold increases in the mdtEexpression were detected at the early and late stationary phases,respectively. No other exporter genes showed such increasesin expression at the stationary phase.

View larger version (23K):
[in this window]
[in a new window]
FIG. 1. Expression of the 20 drug exporter genes at different growth phases. The expression of the drug exporter genes was determined by the ß-galactosidase reporter enzyme assay. Single-copy reporter plasmids were transformed into E. coli strain MC4100. E. coli cells were cultured until the OD₆₀₀ reached 0.4 (early logarithmic phase), 0.8 (late logarithmic phase), 3.0 (early stationary phase), or 6.5 (late stationary phase), and then ß-galactosidase activity was measured. pNN387 indicates the vector control.

At the logarithmic phase, the gene that showed the highest expressionlevel was ydgF (109 Miller units); however, YdgF confers low-leveldrug resistance only to deoxycholate and sodium dodecyl sulfate(SDS), even when overproduced from a multicopy plasmid (19).The ydgF-knockout strain did not show hypersensitivity to drugs,including deoxycholate and SDS. The emrE, mdfA, and acrA genesshowed relatively high expression levels at the logarithmicphase, next to that of ydgF. The deletion of these genes isknown to increase drug susceptibility (29). The expression levelof mdtE at the logarithmic phase was lower than those of emrE,mdfA, and acrA. In addition, RND (resistance nodulation celldivision)-type multidrug exporter genes acrE, acrD, and mdtABC,which cause high multidrug resistance when they are overexpressed(19), showed very low expression levels throughout the cellgrowth. Therefore, it seems that the drug tolerance of E. colicells under laboratory conditions mainly reflects the expressionlevels of these drug exporters at the logarithmic phase exceptfor ydgF.

On the other hand, at the late stationary phase, the expressionlevel of mdtE (260 Miller units) was the highest out of thoseof the 20 drug exporter genes. The second highest was that ofemrE, but the level of its activity (50 Miller units) was farlower than that of mdtE. Such a growth phase-dependent increasein mdtE gene expression was also confirmed by determinationof transcripts by quantitative RT-PCR analysis, although themaximum level was observed at the early stationary phase withrespect to the mRNA level. The mdtE gene transcripts showed380- and 76-fold increases at the early and late stationaryphases, respectively, compared to the logarithmic phase. Thisindicates that the promoter activity of mdtEF is highest atthe early logarithmic phase. Although the reporter enzyme wasaccumulated during the stationary phase, the amount of mRNAof mdtEF was gradually decreased, probably due to its high turnoverrate. In summary, MdtEF is greatly induced at the stationaryphase and contributes the intrinsic drug tolerance.

Drug tolerance mediated by up-regulation of mdtEF at the stationary phase. In order to determine whether the growth phase-dependent inductionof mdtEF contributes to drug tolerance, cell growth was comparedin the presence of drugs after the stationary phase and in logarithmic-phasecells being diluted to the same density with fresh medium. Ifcells at two different phases have different susceptibilitiesto the drug, the growth rate must reflect their initial viability.At first, the growth rate was compared in the ${Delta}$ acrB backgroundbecause the high-performance housekeeping drug exporter AcrABmay mask the contribution of MdtEF. MC4100 ${Delta}$ acrB and MC4100 ${Delta}$ acrB ${Delta}$ mdtEFcells were first grown to the logarithmic phase (OD₆₀₀ of 0.8)or to the stationary phase (OD₆₀₀ of 6.5) in the absence ofdrugs, and then the cells were diluted to the same density withfresh medium. The growth was monitored in the absence or presenceof several drugs, dyes, detergents, and antiseptics. Our previousstudies revealed that MdtEF confers resistance to erythromycin,doxorubicin, crystal violet, ethidium bromide, rhodamine 6G,tetraphenylphosphonium bromide (TPP), benzalkonium, SDS, anddeoxycholate when overexpressed (19). We used these compoundsfor the drug tolerance assay. In addition to these compounds,kanamycin, nalidixic acid, and norfloxacin, which are not substratesof MdtEF, were used as negative controls (19). In Fig. 2A and B,the growth curves in the presence or absence of crystal violetand kanamycin are shown as examples for MdtEF substrates andnegative controls, respectively. All logarithmic-phase and stationary-phasecells grew at about the same rate in the absence of drugs, whilethe growth of logarithmic-phase cells was greatly retarded with1.56 mg/liter crystal violet or 6.25 mg/ liter kanamycin (Fig.2A and B). Although the growth of MC4100 ${Delta}$ acrB ${Delta}$ mdtEF stationary-phasecells was also greatly retarded in the presence of crystal violet,the growth of MC4100 ${Delta}$ acrB stationary-phase cells was significantlyrecovered in the presence of crystal violet (Fig. 2A). BecauseMC4100 ${Delta}$ acrB stationary-phase cells did not exhibit growth recoveryin the presence of kanamycin (Fig. 2B), which is not a substrateof MdtEF, the drug tolerance of the stationary-phase cells tocrystal violet was certainly due to MdtEF in the ${Delta}$ acrB background.Similarly, we observed mdtEF-dependent drug tolerance of thestationary-phase cells to erythromycin, doxorubicin, rhodamine6G, ethidium bromide, TPP, benzalkonium, SDS, and deoxycholate,but not to nalidixic acid or norfloxacin (data not shown).

View larger version (35K):
[in this window]
[in a new window]
FIG. 2. Growth phase dependence of drug tolerance. Each strain (MC4100 ${Delta}$ mdtEF, MC4100 ${Delta}$ acrB, and MC4100 ${Delta}$ acrB ${Delta}$ mdtEF) was grown until the OD₆₀₀ reached 0.8 or 6.5 and then diluted to an OD₆₀₀ of 0.1 with fresh medium. Cell growth was monitored in the absence or presence of drugs. (A) MC4100 ${Delta}$ acrB and MC4100 ${Delta}$ acrB ${Delta}$ mdtEF with or without crystal violet. (B) MC4100 ${Delta}$ acrB and MC4100 ${Delta}$ acrB ${Delta}$ mdtEF with or without kanamycin. (C) MC4100 and MC4100 ${Delta}$ mdtEF with or without crystal violet. (D) MC4100 and MC4100 ${Delta}$ mdtEF with or without kanamycin.

Then, in order to determine whether MdtEF contributes to drugtolerance even in the presence of the acrB gene, we measuredthe growth of wild-type MC4100 and MC4100 ${Delta}$ mdtEF. In the presenceof 1.56 mg/liter crystal violet, both logarithmic- and stationary-phasecells showed full growth independent of mdtEF. However, in thepresence of high concentration (12.5 mg/liter) of crystal violet,an increase in the mdtEF-dependent drug tolerance of the stationary-phasecells of the wild-type strain was observed (Fig. 2C), althoughthe degree of the relative increase in drug tolerance (MC4100/MC4100 ${Delta}$ mdtEF)was lower than that in the ${Delta}$ acrB background. In contrast, wild-typestationary-phase cells did not exhibit tolerance to kanamycin(Fig. 2D). These observations indicate that MdtEF actually contributesto the multidrug tolerance of E. coli at the stationary phase.

Subsequently, in order to confirm the increase in drug toleranceat the stationary phase, the viability of the cells was measuredafter short exposure to bactericidal compounds. MC4100 ${Delta}$ acrB andMC4100 ${Delta}$ acrB ${Delta}$ mdtEF cells were first grown to the logarithmic phase(OD₆₀₀ of 0.6) or to the stationary phase (OD₆₀₀ of 6.5). Thenthe stationary-phase cells were diluted to the same densityas the logarithmic-phase cells (OD₆₀₀ of 0.6) with fresh medium.Both types of cells were exposed to 50 mg/liter crystal violet.After incubation for 30 min at 37°C, the survival rate wascalculated as described in Materials and Methods (Fig. 3A).The logarithmic-phase cells of MC4100 ${Delta}$ acrB and MC4100 ${Delta}$ acrB ${Delta}$ mdtEFshowed very low survival rates (0.7% and 1.6%, respectively).On the other hand, the MC4100 ${Delta}$ acrB stationary-phase cells showedvery high viability (101%), whereas MC4100 ${Delta}$ acrB ${Delta}$ mdtEF cells stillexhibited low viability (2.2%) at the stationary phase. In theacrB⁺ background, both logarithmic- and stationary-phase cellswere fully viable in the presence of 50 mg/liter of crystalviolet independent on the presence or absence of mdtEF. However,in the presence of 200 mg/liter of crystal violet, the viabilityof the stationary-phase cells (57%) was significantly higherthan that of the logarithmic-phase cells (4%) and the viabilityof the stationary-phase ${Delta}$ mdtEF cells (25%) was significantlylower than the wild-type cells. These results indicated thatthe induction of mdtEF gene expression at the stationary phasecontributes to the drug tolerance, while in the high drug concentration,the drug resistance mechanisms other than MdtEF also partlycontribute to the stationary-phase drug tolerance.

View larger version (19K):
[in this window]
[in a new window]
FIG. 3. Growth phase-dependent increase in cell viability with the bactericidal drugs. (A) Each bacterial strain (MC4100 ${Delta}$ acrB and MC4100 ${Delta}$ acrB ${Delta}$ mdtEF) was grown at 37°C in LB broth until the OD₆₀₀ reached 0.6 (logarithmic phase) or 6.5 (stationary phase). The stationary-phase cells were diluted to an OD₆₀₀ of 0.6. Then crystal violet was added to each bacterial cell culture (final concentration, 50 mg/liter). After incubation for 30 min at 37°C, viability was measured as described in Materials and Methods. (B) Each bacterial strain (MC4100 and MC4100 ${Delta}$ mdtEF) was grown at 37°C in LB broth until an OD₆₀₀ of 0.6 (logarithmic phase) or 6.5 (stationary phase). Then the bacterial cell viability with medium containing crystal violet (200 mg/liter) was measured as described above.

The effect of the evgSA and tnaAB gene deletions on the growth phase-dependent induction of mdtEF gene expression. Previously, we found that the EvgSA two-component system positivelycontrols mdtEF expression (20). In order to determine whetherthe growth phase-dependent mdtEF induction is controlled bythe EvgSA system, the expression level of mdtEF was measuredin the ${Delta}$ evgSA background. It was found that the level of inductionof mdtEF at the stationary phase was not affected by evgSA deletion(Fig. 4A), indicating that the growth phase-dependent regulationof mdtEF was not mediated by the EvgSA system.

View larger version (26K):
[in this window]
[in a new window]
FIG. 4. Effects of deletion of mdtEF regulatory genes on the expression of the gadE-mdtEF genes. The expression of gadE-mdtEF in the wild type (MC4100) or each gene deletion mutant (MC4100 ${Delta}$ evgSA, MC4100 ${Delta}$ tnaAB, MC4100 ${Delta}$ gadX, MC4100 ${Delta}$ hfq, and MC4100 ${Delta}$ rpoS) was determined by means of the ß-galactosidase reporter enzyme assay. E. coli cells cultured for 2, 3, 4, 12, and 24 h were collected, followed by ß-galactosidase activity measurement. (A) Effect of evgSA deletion. (B) Effect of tnaAB deletion and addition of indole (500 µM). (C) Effect of deletion of the gadX, hfq, and rpoS genes. At least three independent experiments were performed in each case.

We also previously reported that indole, which is a toxic metabolitesynthesized from tryptophan by a tryptophanase, TnaA (26), up-regulatesmdtEF expression (5). The extracellular indole concentrationin wild-type cells increased with cell growth and reached about500 µM after 24 h of culture, whereas there was no detectablelevel of indole in the culture medium of ${Delta}$ tnaAB cells even after24 h of culture (Fig. 5). In order to determine the contributionof indole to the expression level of mdtEF, we measured thegrowth phase-dependent mdtEF induction in the ${Delta}$ tnaAB background.The tnaAB deletion significantly decreased the level of inductionof mdtEF at the stationary phase to about 65% of the wild-typelevel (Fig. 4B). When 500 µM indole was added in the medium,the expression level of mdtEF in the ${Delta}$ tnaAB strain was restoredto almost the same level as the stationary-phase tnaAB⁺ strain(Fig. 4B). Thus, indole plays some role in the growth phase-dependentinduction of mdtEF.

View larger version (17K):
[in this window]
[in a new window]
FIG. 5. Indole accumulation with cell growth. The extracellular indole concentrations of the wild-type and ${Delta}$ tnaAB strains were measured by HPLC analysis as described in Materials and Methods. Black squares, MC4100 (wild type); white squares, MC4100 ${Delta}$ tnaAB.

The effect of the gadX, hfq, and rpoS genes on growth phase-dependent induction of mdtEF gene expression. Since indole induces the expression of mdtEF via transcriptionalregulator GadX (5), we examined whether the growth phase-dependentinduction of mdtEF is mediated by gadX. In the gadX deletionmutant (MC4100 ${Delta}$ gadX), growth phase-dependent induction of mdtEFwas completely abolished (Fig. 4C). Recently, a small-RNA regulator,GadY, which binds to Hfq protein, was found to be induced atthe stationary phase in a sigma factor RpoS-dependent manner(22). The overexpression of GadY enhances the mRNA level ofgadX (22). Therefore, we then deleted the hfq and rpoS genes.In the resultant strains, the growth phase-dependent inductionof mdtEF was completely abolished, like on the gadX deletion.Thus, the growth phase-dependent mdtEF induction is mediatedby the RpoS-GadY(Hfq)-GadX signaling pathway.

Then we investigated the effect of rpoS, hfq, and gadX deletionon drug tolerance of E. coli. As shown in Fig. 6, the deletionof these genes greatly reduced drug tolerance of the stationary-phasecells to crystal violet in the ${Delta}$ acrB background. The deletionof these genes did not affect the crystal violet sensitivityof the logarithmic-phase cells.

View larger version (39K):
[in this window]
[in a new window]
FIG. 6. Effect of deletion of the RpoS-dependent signaling pathway on drug tolerance of the stationary-phase cells. Each strain (MC4100, MC4100 ${Delta}$ acrB, MC4100 ${Delta}$ acrB ${Delta}$ gadX, MC4100 ${Delta}$ acrB ${Delta}$ hfq, and MC4100 ${Delta}$ acrB ${Delta}$ rpoS) was grown until an OD₆₀₀ of 0.8 or 6.5 and then diluted to an OD₆₀₀ of 0.1 with fresh medium. Cell growth was monitored in the absence or presence of crystal violet. (A) MC4100 ${Delta}$ acrB and MC4100 ${Delta}$ acrB ${Delta}$ gadX. (B) MC4100 ${Delta}$ acrB and MC4100 ${Delta}$ acrB ${Delta}$ hfq. (C) MC4100 ${Delta}$ acrB and MC4100 ${Delta}$ acrB ${Delta}$ rpoS.

Growth phase-dependent expression of rpoS, hfq, and gadX. In order to characterize the RpoS-GadY(Hfq)-GadX signaling pathway,we measured the growth phase-dependent changes in the expressionof the rpoS, gadY, hfq, and gadX genes by means of the ß-galactosidasereporter assay. As shown in Fig. 7A, the expression of rpoS,gadY, and gadX showed growth phase dependence and the maximumexpression was observed at the late stationary phase, exceptin the case of gadX, of which the expression levels at the earlyand late stationary phases were almost equal to each other withinexperimental error. On the other hand, the expression levelof hfq did not show such significant growth phase dependence.Compared with the expression levels at the early logarithmicphase (OD₆₀₀ of 0.4), the expression levels of rpoS, gadY, andgadX at the late stationary phase were increased by factorsof 8.8, 2.5, and 3.6, respectively. That the degree of increasein the expression levels of regulator genes was relatively lowerthan that of mdtEF (41-fold) may be due to the fact that GadXacts as a dimer in transcriptional regulation. In summary, thegrowth phase-dependent control of mdtEF was mediated throughmodification of the amounts of RpoS-dependent small-RNA GadYand transcriptional regulator GadX. One exception is that, althoughrpoS, gadY, and gadX expression was increased from early tolate log phase (Fig. 7A), the mdtEF expression was slightlydecreased during this period (Fig. 1). The slight decrease ofmdtE expression might be within the experimental deviation;however, there may be a possibility of the postexponential activationof RpoS.

View larger version (23K):
[in this window]
[in a new window]
FIG. 7. Expression levels of rpoS, hfq, gadY, and gadX and effect of their deletion on induction of mdtEF genes by indole. (A) The growth phase-dependent expression of the rpoS, hfq, gadY, and gadX genes was determined by means of the ß-galactosidase reporter enzyme assay. E. coli cells were cultured until the OD₆₀₀ reached 0.4 (early logarithmic phase), 0.8 (late logarithmic phase), 3.0 (early stationary phase), or 6.5 (late stationary phase), and then ß-galactosidase activity was measured. (B) Effect of deletion of mdtEF regulatory genes on the induction of mdtEF by indole. The wild-type and mutant strains ( ${Delta}$ evgSA, ${Delta}$ gadX, ${Delta}$ hfq, and ${Delta}$ rpoS) were grown until the OD₆₀₀ reached 0.8 in LB broth with (black bars) or without (white bars) 1 mM indole. The ß-galactosidase activity of the lacZ fusion of the gadE-mdtEF promoter was measured. (C) Effect of indole on expression of rpoS, hfq, gadY, and gadX. The expression of the rpoS, hfq, gadY, and gadX genes was determined by means of the ß-galactosidase reporter enzyme assay. E. coli cells were grown until the OD₆₀₀ reached 0.8 in LB broth with (black bars) or without (white bars) 1 mM indole, and then the ß-galactosidase activity of the lacZ fused to each promoter was measured.

The growth phase-dependent expression of rpoS, gadY, and gadXwas confirmed on quantitative PCR analysis. At the early stationaryphase, the mRNA of mdtE was drastically increased (380-fold),and those of rpoS, gadY, and gadX were also moderately increased(2.5-, 18-, and 23-fold, respectively), whereas a significantchange in the hfq mRNA level was not observed (Table 4).

View this table:
[in this window]
[in a new window]
TABLE 4. Induction of mdtE, rpoS, hfq, gadY, and gadX gene transcripts attributed to the growth phase, as determined by amplification of cDNA samples

GadX, Hfq, and RpoS are essential for indole-induced mdtEF expression, and the expression is induced via rpoS, gadY, and gadX up-regulation. We investigated the relationship between the induction of mdtEFexpression by indole and the RpoS-GadY(Hfq)-GadX signaling pathway.At first, indole production was measured in wild-type and tnaAB,evgSA, gadX, hfq, and rpoS deletion mutant cells. After a 24-hculture of these cells, the indole concentration in the culturemedium of the respective deletion mutants was the same as thatof the wild type, except in the case of the ${Delta}$ tnaAB mutant (datanot shown), in which indole production was not observed (Fig.5), indicating that RpoS-GadY(Hfq)-GadX signaling does not affectindole production.

Then the mdtEF reporter gene expression by these strains wasmeasured in the presence or absence of externally added indole.In the wild-type and evgSA deletion mutant, the expression wassimilarly increased by the addition of 1 mM indole, whereaswhen the gadX, hfq, or rpoS gene was deleted, the expressionof mdtEF was no longer increased by indole at all (Fig. 7B);therefore, the induction of mdtEF by indole is also mediatedby the RpoS-GadY(Hfq)-GadX signaling pathway.

The effects of indole on the expression levels of rpoS, hfq,gadY, and gadX were examined by means of the reporter gene assay.The expression of these genes was also increased by 1 mM indole,except in the case of hfq (Fig. 7C). These results indicatethat indole controls mdtEF gene expression via increasing theamounts of RpoS, GadY, and GadX. Quantitative PCR analysis ofrpoS, hfq, gadY, and gadX gave similar results (data not shown).

DISCUSSION

Top

Discussion

In this study, we comprehensively investigated the growth phase-dependentexpression of drug exporter genes in E. coli. We found thatout of the 20 drug exporter genes, only the expression of mdtEFgreatly increased with cell growth (Fig. 1 and Table 4). Theinduction of mdtEF expression actually conferred drug toleranceto E. coli at the stationary phase (Fig. 2 and 3). The growthphase-dependent activation of the mdtEF promoter was mediatedby the RpoS-GadY(Hfq)-GadX signaling pathway and enhanced byindole (Fig. 4 and 7). As expected, the deletion of rpoS, hfq,and gadX caused the lack of the MdtEF-dependent drug toleranceof the stationary-phase cells.

Schellhorn et al. (25) reported that yhiUV, which is an oldname for mdtEF, is one of the RpoS-dependent genes that areinduced at the stationary phase. Our previous study revealedthat YhiUV is a multidrug exporter system (20), and thus itwas renamed MdtEF (21). In this study, we revealed that thegrowth phase-dependent expression of mdtEF confers drug toleranceat the stationary phase. In addition, we revealed that the growthphase regulation of mdtEF is mediated by RpoS-dependent small-RNAGadY and transcriptional regulator GadX, which is the same signalingpathway as indole signaling.

Sulavik et al. reported that most drug exporter genes did notcontribute to drug tolerance under laboratory conditions, exceptfor acrAB, emrE, and mdfA, in an exporter gene knockout experiment(29). The results of this study indicate that the drug hypersensitizationof exporter gene deletion mutants reflects the expression levelsof drug exporter genes at the logarithmic phase, except in thecase of ydgF, which is a weak exporter having a very narrowsubstrate range. Since the expression level of mdtEF is verylow at the logarithmic phase, on deletion of the mdtEF gene,no hypersensitization might be seen on MIC measurement. On theother hand, at the stationary phase, MdtEF is a major drug exporterand certainly confers drug tolerance, although the contributionof AcrAB to the stationary-phase drug tolerance is still significant.

As for the possibility that MdtEF confers tolerance againstindole, experimental detection was difficult because the indoletoxicity is very low. However, the fact that the ${Delta}$ mdtEF mutantshowed a somewhat reduced indole concentration in the stationary-phasemedium and increased accumulation of indole in the cells whenindole was externally added (data not shown) suggests the possibilitythat MdtEF plays some role in indole export.

The mdtEF genes are cotranscribed with acid response regulatorgadE, which is encoded upstream of mdtEF in the same operon.The gadE expression is controlled as a response to acid stress(22, 30, 31, 32). However, in our experiments, the pH of themedium at the stationary phase was moderately alkaline (aboutpH 8.5). Therefore, the signal causing growth phase-dependentinduction must be different from acid stress. We examined theeffect of alkaline pH on mdtE expression by means of quantitativePCR analysis. The expression level of mdtE was not altered whenMC4100 was cultured until the logarithmic phase (OD₆₀₀ of 0.8)in the LB broth at pH 7.0 or in the LB broth that had been preparedto pH 8.5 (data not shown).

In our previous study, we reported that 2 mM indole added tothe culture medium significantly induced the expression of mdtE,acrD, acrE, emrK, yceL, and cusB at the logarithmic phase, whereaswhen the indole concentration was 1 mM, only mdtE inductionwas significant (5). The concentration of indole in the MC4100culture medium at the stationary phase was around 500 µMwith the complex laboratory medium. That is the reason why theinduction of the drug exporters other than mdtE by intrinsicindole at the stationary phase was not observed in this study.Of course, this fact does not exclude the possibility that indole-dependentdrug exporter genes other than mdtEF may be actually inducedand play some roles at infection sites due to the high localconcentration of indole produced by other bacteria.

In P. aeruginosa, N-(3-oxododecanoyl)-L-homoserine lactone andN-(butyryl)-L-homoserine lactone are known as quorum-sensingsignal molecules at the stationary phase, while E. coli doesnot produce acylhomoserine lactones. On the other hand, theindole concentration increases with cell growth as a by-productof pyruvate production in E. coli cells (33) (Fig. 5). Besides,it has been reported that indole regulates biofilm formationby E. coli and is associated with the virulence of Haemophilusinfluenzae (12, 13). Our results confirmed the finding of Wanget al. (33) that indole acts as a stationary-phase signal molecule.

It is well known that the drug resistance of pathogens at thesites of infection is generally higher than that under laboratoryconditions (8). We believe that elucidation of the mechanismsunderlying the growth phase-dependent induction of multidrugexporter genes is important for understanding such acquireddrug resistance mechanisms at the site of infection.

ACKNOWLEDGMENTS

We wish to thank George M. Church for plasmid pKO3 and RonaldW. Davis for plasmid pNN387.

H. Hirakawa was supported by a research fellowship from theJapan Society for the Promotion of Science for Young Scientists.This work was supported by a Grant-in-Aid from the Ministryof Education, Culture, Sports, Science and Technology of Japanand the Zoonosis Control Project of the Ministry of Agriculture,Forestry and Fisheries of Japan.

FOOTNOTES

* Corresponding author. Mailing address: Institute of Scientific and Industrial Research, Osaka University, 8-1 Mihogaoka, Ibaraki-shi, Osaka 567-0047, Japan. Phone: 81-6-6879-8545. Fax: 81-6-6879-8549. E-mail: akihito{at}sanken.osaka-u.ac.jp.

Present address: Department of Microbiology and Molecular Genetics,Graduate School of Pharmaceutical Sciences, Chiba University,Chiba 263-8522, Japan.

REFERENCES

Top