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

Bacterial strains and plasmids.

The bacterial strains and plasmids used in this study are presented in Table 1. The construction of gene deletion mutants of E. coli MC4100 (2) was performed by the gene replacement method previously described, using the pKO3 plasmid (9). E. coli cells were cultured in Luria-Bertani (LB) medium, supplemented with appropriate antibiotics when necessary.

Construction of reporter plasmids.

The reporter plasmids were constructed as follows. DNA fragments including the putative promoter region were amplified by PCR using the primers listed in Table 2. Chromosomal DNA of E. coli MG1655 (1) was used as a PCR template. The DNA fragments were cloned in front of the lacZ reporter gene in a single-copy pNN387 vector, which carries chloramphenicol resistance as a marker (3). Since the emrAB, cusBA, and mdtEF genes are transcribed in the emrRAB (10), cusCFBA (4), and gadE-mdtEF (14; our unpublished data) operons, respectively, emrR, cusC, and gadE promoter-fused lacZ are reporters for the respective exporter expression. The resulting plasmids were introduced into host cells for β-galactosidase activity measurements.

Reporter gene assay.

To investigate the growth phase-dependent transcription of various reporter constructs, each bacterial strain was grown at 37°C in LB broth containing 15 mg/liter chloramphenicol until the optical density at 600 nm (OD₆₀₀) reached 0.4 (early logarithmic phase), 0.8 (late logarithmic phase), 3.0 (early stationary phase), or 6.5 (late stationary phase). When we determined the effect of indole on the transcription of mdtEF, cells were grown at 37°C in LB broth until the OD₆₀₀ reached 0.8 with 15 mg/liter chloramphenicol and 1 mM indole or only with 15 mg/liter chloramphenicol and the solvent (dimethyl sulfoxide) as a control. β-Galactosidase activity in cell lysates was assayed using o-nitrophenyl-β-d-galactopyranoside (ONPG) as a substrate as described by Miller (16), with a slight modification.

Quantitative real-time RT-PCR.

Quantitative real-time reverse transcription-PCR (RT-PCR) was performed as follows. Cells were grown at 37°C in LB broth until 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 were performed by the methods described previously (6). The specific primer pairs are listed in Table 3. Real-time PCR was performed with each specific primer pair, using SYBR green PCR master mix (PE Applied Biosystems). The E. coli rrsA gene was chosen as a control for normalization of the cDNA loading in each PCR. The reactions were performed with an ABI PRISM 7000 sequence detection system (PE Applied Biosystems).

Drug tolerance assay.

Each bacterial strain, grown at 37°C in LB broth until OD₆₀₀ reached 0.8 (logarithmic phase) or 6.5 (stationary phase), was diluted to an OD₆₀₀ of 0.1. Then, growth was measured in the presence or absence of crystal violet or kanamycin.

Survival assay.

Each bacterial strain 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 with fresh medium, and then crystal violet was added to each bacterial cell culture (final concentration, 50 or 200 mg/liter). After incubation for 30 min at 37°C, aliquots of the cell cultures were spread on YT agar plates. After overnight incubation, the numbers of colonies were determined and percent survival was calculated in comparison with that of untreated cells.

Indole production assay.

The extracellular indole concentration was determined by high-performance liquid chromatography (HPLC). The E. coli strain was cultured at 37°C and then pelleted by centrifugation at 20,000 × g. The resulting supernatant was extracted twice with ethyl acetate. The ethyl acetate phase was loaded onto a Symmetry C₁₈ column (5 μm, 4.6 by 150 mm; Waters Corp.) attached to an L2130 HPLC system (HITACHI). The loaded samples were eluted with acetonitrile-H₂O (1:1) at the flow rate of 0.8 ml/min. Then the indole peak was detected relative to the absorbance at the wavelength of 276 nm and was identified by the corresponding peak of the purified indole (Sigma). The indole concentration was calculated from the ratio of the detection peak area to the standard peak one.

The growth phase-dependent expression of the intrinsic drug exporter genes in Escherichia coli.

In order to profile the growth phase-dependent expression levels of intrinsic drug exporter genes, we examined their reporter enzyme activities at different growth phases using single-copy plasmids containing lacZ-fused gene promoters. Because the emrAB, cusBA, and mdtEF genes are transcribed as a part of the operons of emrRAB (10), cusCFBA (4), and gadE-mdtEF (14 and our unpublished data), respectively, emrR, cusC, and gadE promoter-fused lacZ constructs were used as reporters for these three genes. Each bacterial strain was grown 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 β-galactosidase activities were measured. Among the 20 exporter genes, only expression of the mdtE gene greatly increased with growth cessation (Fig. 1). Compared with the late logarithmic phase, 14- and 41-fold increases in the mdtE expression were detected at the early and late stationary phases, respectively. No other exporter genes showed such increases in expression at the stationary phase.

At the logarithmic phase, the gene that showed the highest expression level was ydgF (109 Miller units); however, YdgF confers low-level drug 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 genes showed relatively high expression levels at the logarithmic phase, next to that of ydgF. The deletion of these genes is known to increase drug susceptibility (29). The expression level of mdtE at the logarithmic phase was lower than those of emrE, mdfA, and acrA. In addition, RND (resistance nodulation cell division)-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 cell growth. Therefore, it seems that the drug tolerance of E. coli cells under laboratory conditions mainly reflects the expression levels of these drug exporters at the logarithmic phase except for ydgF.

On the other hand, at the late stationary phase, the expression level of mdtE (260 Miller units) was the highest out of those of the 20 drug exporter genes. The second highest was that of emrE, but the level of its activity (50 Miller units) was far lower than that of mdtE. Such a growth phase-dependent increase in mdtE gene expression was also confirmed by determination of transcripts by quantitative RT-PCR analysis, although the maximum level was observed at the early stationary phase with respect to the mRNA level. The mdtE gene transcripts showed 380- and 76-fold increases at the early and late stationary phases, respectively, compared to the logarithmic phase. This indicates that the promoter activity of mdtEF is highest at the early logarithmic phase. Although the reporter enzyme was accumulated during the stationary phase, the amount of mRNA of mdtEF was gradually decreased, probably due to its high turnover rate. In summary, MdtEF is greatly induced at the stationary phase 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 induction of mdtEF contributes to drug tolerance, cell growth was compared in the presence of drugs after the stationary phase and in logarithmic-phase cells being diluted to the same density with fresh medium. If cells at two different phases have different susceptibilities to the drug, the growth rate must reflect their initial viability. At first, the growth rate was compared in the ΔacrB background because the high-performance housekeeping drug exporter AcrAB may mask the contribution of MdtEF. MC4100ΔacrB and MC4100ΔacrBΔmdtEF cells were first grown to the logarithmic phase (OD₆₀₀ of 0.8) or to the stationary phase (OD₆₀₀ of 6.5) in the absence of drugs, and then the cells were diluted to the same density with fresh medium. The growth was monitored in the absence or presence of several drugs, dyes, detergents, and antiseptics. Our previous studies revealed that MdtEF confers resistance to erythromycin, doxorubicin, crystal violet, ethidium bromide, rhodamine 6G, tetraphenylphosphonium bromide (TPP), benzalkonium, SDS, and deoxycholate when overexpressed (19). We used these compounds for the drug tolerance assay. In addition to these compounds, kanamycin, nalidixic acid, and norfloxacin, which are not substrates of MdtEF, were used as negative controls (19). In Fig. 2A and B, the growth curves in the presence or absence of crystal violet and kanamycin are shown as examples for MdtEF substrates and negative controls, respectively. All logarithmic-phase and stationary-phase cells grew at about the same rate in the absence of drugs, while the growth of logarithmic-phase cells was greatly retarded with 1.56 mg/liter crystal violet or 6.25 mg/ liter kanamycin (Fig. 2A and B). Although the growth of MC4100ΔacrBΔmdtEF stationary-phase cells was also greatly retarded in the presence of crystal violet, the growth of MC4100ΔacrB stationary-phase cells was significantly recovered in the presence of crystal violet (Fig. 2A). Because MC4100ΔacrB stationary-phase cells did not exhibit growth recovery in the presence of kanamycin (Fig. 2B), which is not a substrate of MdtEF, the drug tolerance of the stationary-phase cells to crystal violet was certainly due to MdtEF in the ΔacrB background. Similarly, we observed mdtEF-dependent drug tolerance of the stationary-phase cells to erythromycin, doxorubicin, rhodamine 6G, ethidium bromide, TPP, benzalkonium, SDS, and deoxycholate, but not to nalidixic acid or norfloxacin (data not shown).

Then, in order to determine whether MdtEF contributes to drug tolerance even in the presence of the acrB gene, we measured the growth of wild-type MC4100 and MC4100ΔmdtEF. In the presence of 1.56 mg/liter crystal violet, both logarithmic- and stationary-phase cells showed full growth independent of mdtEF. However, in the presence of high concentration (12.5 mg/liter) of crystal violet, an increase in the mdtEF-dependent drug tolerance of the stationary-phase cells of the wild-type strain was observed (Fig. 2C), although the degree of the relative increase in drug tolerance (MC4100/MC4100ΔmdtEF) was lower than that in the ΔacrB background. In contrast, wild-type stationary-phase cells did not exhibit tolerance to kanamycin (Fig. 2D). These observations indicate that MdtEF actually contributes to the multidrug tolerance of E. coli at the stationary phase.

Subsequently, in order to confirm the increase in drug tolerance at the stationary phase, the viability of the cells was measured after short exposure to bactericidal compounds. MC4100ΔacrB and MC4100ΔacrBΔmdtEF cells were first grown to the logarithmic phase (OD₆₀₀ of 0.6) or to the stationary phase (OD₆₀₀ of 6.5). Then the stationary-phase cells were diluted to the same density as 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 was calculated as described in Materials and Methods (Fig. 3A). The logarithmic-phase cells of MC4100ΔacrB and MC4100ΔacrBΔmdtEF showed very low survival rates (0.7% and 1.6%, respectively). On the other hand, the MC4100ΔacrB stationary-phase cells showed very high viability (101%), whereas MC4100ΔacrBΔmdtEF cells still exhibited low viability (2.2%) at the stationary phase. In the acrB⁺ background, both logarithmic- and stationary-phase cells were fully viable in the presence of 50 mg/liter of crystal violet independent on the presence or absence of mdtEF. However, in the presence of 200 mg/liter of crystal violet, the viability of the stationary-phase cells (57%) was significantly higher than that of the logarithmic-phase cells (4%) and the viability of the stationary-phase ΔmdtEF cells (25%) was significantly lower than the wild-type cells. These results indicated that the induction of mdtEF gene expression at the stationary phase contributes to the drug tolerance, while in the high drug concentration, the drug resistance mechanisms other than MdtEF also partly contribute to the stationary-phase drug tolerance.

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 positively controls mdtEF expression (20). In order to determine whether the growth phase-dependent mdtEF induction is controlled by the EvgSA system, the expression level of mdtEF was measured in the ΔevgSA background. It was found that the level of induction of mdtEF at the stationary phase was not affected by evgSA deletion (Fig. 4A), indicating that the growth phase-dependent regulation of mdtEF was not mediated by the EvgSA system.

We also previously reported that indole, which is a toxic metabolite synthesized from tryptophan by a tryptophanase, TnaA (26), up-regulates mdtEF expression (5). The extracellular indole concentration in wild-type cells increased with cell growth and reached about 500 μM after 24 h of culture, whereas there was no detectable level of indole in the culture medium of ΔtnaAB cells even after 24 h of culture (Fig. 5). In order to determine the contribution of indole to the expression level of mdtEF, we measured the growth phase-dependent mdtEF induction in the ΔtnaAB background. The tnaAB deletion significantly decreased the level of induction of mdtEF at the stationary phase to about 65% of the wild-type level (Fig. 4B). When 500 μM indole was added in the medium, the expression level of mdtEF in the ΔtnaAB strain was restored to almost the same level as the stationary-phase tnaAB⁺ strain (Fig. 4B). Thus, indole plays some role in the growth phase-dependent induction of mdtEF.

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 transcriptional regulator GadX (5), we examined whether the growth phase-dependent induction of mdtEF is mediated by gadX. In the gadX deletion mutant (MC4100ΔgadX), growth phase-dependent induction of mdtEF was completely abolished (Fig. 4C). Recently, a small-RNA regulator, GadY, which binds to Hfq protein, was found to be induced at the stationary phase in a sigma factor RpoS-dependent manner (22). The overexpression of GadY enhances the mRNA level of gadX (22). Therefore, we then deleted the hfq and rpoS genes. In the resultant strains, the growth phase-dependent induction of mdtEF was completely abolished, like on the gadX deletion. Thus, the growth phase-dependent mdtEF induction is mediated by the RpoS-GadY(Hfq)-GadX signaling pathway.

Then we investigated the effect of rpoS, hfq, and gadX deletion on drug tolerance of E. coli. As shown in Fig. 6, the deletion of these genes greatly reduced drug tolerance of the stationary-phase cells to crystal violet in the ΔacrB background. The deletion of these genes did not affect the crystal violet sensitivity of the logarithmic-phase cells.

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 expression of the rpoS, gadY, hfq, and gadX genes by means of the β-galactosidase reporter assay. As shown in Fig. 7A, the expression of rpoS, gadY, and gadX showed growth phase dependence and the maximum expression was observed at the late stationary phase, except in the case of gadX, of which the expression levels at the early and late stationary phases were almost equal to each other within experimental error. On the other hand, the expression level of hfq did not show such significant growth phase dependence. Compared with the expression levels at the early logarithmic phase (OD₆₀₀ of 0.4), the expression levels of rpoS, gadY, and gadX at the late stationary phase were increased by factors of 8.8, 2.5, and 3.6, respectively. That the degree of increase in the expression levels of regulator genes was relatively lower than that of mdtEF (41-fold) may be due to the fact that GadX acts as a dimer in transcriptional regulation. In summary, the growth phase-dependent control of mdtEF was mediated through modification of the amounts of RpoS-dependent small-RNA GadY and transcriptional regulator GadX. One exception is that, although rpoS, gadY, and gadX expression was increased from early to late log phase (Fig. 7A), the mdtEF expression was slightly decreased during this period (Fig. 1). The slight decrease of mdtE expression might be within the experimental deviation; however, there may be a possibility of the postexponential activation of RpoS.

The growth phase-dependent expression of rpoS, gadY, and gadX was confirmed on quantitative PCR analysis. At the early stationary phase, 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 significant change in the hfq mRNA level was not observed (Table 4).

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 mdtEF expression 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-h culture of these cells, the indole concentration in the culture medium of the respective deletion mutants was the same as that of the wild type, except in the case of the ΔtnaAB mutant (data not shown), in which indole production was not observed (Fig. 5), indicating that RpoS-GadY(Hfq)-GadX signaling does not affect indole production.

Then the mdtEF reporter gene expression by these strains was measured in the presence or absence of externally added indole. In the wild-type and evgSA deletion mutant, the expression was similarly increased by the addition of 1 mM indole, whereas when the gadX, hfq, or rpoS gene was deleted, the expression of mdtEF was no longer increased by indole at all (Fig. 7B); therefore, the induction of mdtEF by indole is also mediated by 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 indicate that indole controls mdtEF gene expression via increasing the amounts of RpoS, GadY, and GadX. Quantitative PCR analysis of rpoS, hfq, gadY, and gadX gave similar results (data not shown).