Cytosine DNA methylation influences drug resistance in Escherichia coli through increased sugE expression
Escherichia coli K-12 strains contain the orphan cytosine-5 DNA methyltransferase enzyme Dcm (DNA cytosine methyltransferase). Two recent reports indicate that Dcm has an influence on stationary phase gene expression in E. coli. Herein, we demonstrate that dcm knockout cells overexpress the drug resistance transporter SugE, which has been linked to ethidium bromide (ETBR) resistance. SugE expression also increased in the presence of the DNA methylation inhibitor 5-azacytidine, suggesting that Dcm-mediated DNA methylation normally represses sugE expression. The effect of Dcm on sugE expression is primarily restricted to early stationary phase, and RpoS is required for robust sugE expression. Dcm knockout cells are more resistant to ETBR than wild-type cells, and complementation with a plasmid-borne dcm gene restores ETBR sensitivity. SugE knockout cells are more sensitive to ETBR than wild-type cells. These data indicate that Dcm influences the sensitivity to an antimicrobial compound through changes in gene expression.
Modified DNA bases are found in almost every organism (Zemach et al., 2010) and are generated by postreplicative chemical modification of existing bases (often methylation) (Jeltsch, 2002). In Escherichia coli, 5-methylcytosine is generated by Dcm (DNA cytosine methyltransferase). Dcm methylates the second cytosine in the sequence 5′CCWGG3′ (Marinus & Lobner-Olesen, 2009). Escherichia coli K-12 dcm knockout strains have no detectable 5-methylcytosine, indicating Dcm is the only enzyme that generates 5-methylcytosine in strains lacking restriction–modification systems (Kahramanoglou et al., 2012; Militello et al., 2012). The methylation of cytosine bases by DNA methyltransferases increases the mutation rate due to deamination of 5-methylcytosine to thymine, and this phenomenon has been observed in E. coli (Lieb, 1991; Bandaru et al., 1996). The dcm gene is in an operon with the vsr gene (Sohail et al., 1990). Vsr is an endonuclease that nicks DNA 5′ to the thymine in a thymine–guanine mismatch generated by deamination of 5-methylcytosine (Hennecke et al., 1991; Robertson & Matson, 2012). The Vsr-generated nick is required for removal of the thymine and DNA repair by DNA polymerase I and DNA ligase, which ultimately maintains 5′CCWGG3′ sequences (Lieb & Bhagwat, 1996; Bhagwat & Lieb, 2002). DNA methyltransferases have a role in restriction-modification plasmid biology. In the case of Dcm, Dcm-dependent methylation of phage DNA increases phage infection frequencies in cells that harbor a restriction enzyme that cuts at the Dcm recognition site (Hattman et al., 1973). Dcm also enhances the loss of plasmids with restriction enzymes that cut at 5′CCWGG3′ sites and protects cells against postsegregational killing (Takahashi et al., 2002; Ohno et al., 2008). However, Dcm is often present in cells that do not harbor a restriction enzyme that cuts the same site and is therefore considered an orphan methyltransferase that may have additional functions.
In higher eukaryotes, 5-methylcytosine plays an important role in gene expression. Methylation of promoter DNA is typically associated with gene silencing, whereas gene body DNA methylation is often correlated with active gene transcription (Zemach et al., 2010). In prokaryotes, the generation of N6-methyladenine via DNA adenine methyltransferase has been linked to gene expression changes important for numerous processes including pili expression and virulence (Marinus & Lobner-Olesen, 2009). However, a role for cytosine DNA methylation in prokaryotic gene expression is less well defined. Some restriction-modification plasmids have DNA methyltransferases that influence the timing of restriction enzyme expression (O'Driscoll et al., 2005). It has recently been reported that transcription factors bind to regions lacking 5-methylcytosine in the Vibro cholerae genome and prevent methylation (Dalia et al., 2013), implicating a link between 5-methylcytosine, the DNA methyltransferase, and transcriptional control of chromosomal genes (Banerjee & Chowdhury, 2006). In E. coli, our group and Kahramanoglou et al. investigated a potential role for Dcm in transcriptional regulation (Kahramanoglou et al., 2012; Militello et al., 2012). In summary, the two reports indicate that the loss of Dcm causes an increase in gene expression of several categories of genes, most notably ribosomal protein genes. These observations were important as they indicate that Dcm can influence gene expression and that Dcm is normally repressive. In these studies, the effect of Dcm on gene expression was primarily restricted to stationary phase. Kahramanoglou et al. proposed that Dcm represses expression of the stationary phase sigma factor rpoS, and the loss of Dcm-mediated repression results in the up-regulation of rpoS and a downstream change in stationary phase gene expression (Kahramanoglou et al., 2012). Yet, the relationship between 5-methylcytosine and gene expression is still relatively unexplored, and many questions remain. In particular, we are interested in phenotypes associated with loss of Dcm. Dcm does not seem to have an effect on growth rate, the ability of cells to enter stationary phase, or the ability of cells to persist in stationary phase [K.T. Militello & R.D. Simon, unpublished data and (Kahramanoglou et al., 2012)].
We previously identified genes that have 5′CCWGG3′ recognition sites in the promoter region, and these targets are potentially regulated by cytosine DNA methylation (Militello et al., 2012). One identified target from this analysis was the sugE gene. The sugE gene has one Dcm site c. 10 nucleotides upstream from the transcription start site and three in the gene body (Supporting Information, Fig. S1A). The E. coli sugE gene was originally identified as a suppressor of groEL (Greener et al., 1993). SugE is a membrane transporter with four predicted membrane spanning regions and is a member of the small multidrug resistance family (Bay et al., 2008). SugE RNA is expressed at stationary phase (Table 2) (Greener et al., 1993) and thus could potentially be regulated by Dcm. The function of the E. coli sugE gene is a bit of a mystery. In an initial study of E. coli drug transporters, SugE-mediated resistance to quaternary ammonium compounds (QACs) and ethidium bromide (ETBR) was not observed when sugE was overexpressed (Nishino & Yamaguchi, 2001). However, Chung and Saier reported that sugE overexpressing cells have increased resistance to several QACs (Chung & Saier, 2002), but not ETBR. Thus, the original model has been that the E. coli SugE protein generates resistance to a narrow range of QACs, but does not generate cells resistant to other compounds such as ETBR. However, not all subsequent data fit with this simple model, especially with respect to a lack of an effect of SugE on ETBR resistance. For example, overexpression of the Citrobacter freundii sugE homolog in E. coli generated cells with slightly increased resistance to lipophilic cationic compounds (Lip. Cat. Cmpd) including ETBR (Son et al., 2003). Overexpression of the Enterobacter cloacae sugE homolog in E. coli generated cells with increased resistance to several QACs and ETBR (He et al., 2011). Overexpression of the Aeromonas molluscorum sugE homolog in E. coli generated resistance to ETBR, but not the QAC cetylpyridinium chloride (Cruz et al., 2013). Also, when the E. coli SugE protein was assembled in membrane mimics, it bound to ETBR with a Kd in the low micromolar range, which is consistent with a role in ETBR transport (Sikora & Turner, 2005). Thus, in this work, we directly tested the model that Dcm influences sugE expression and thereby affects SugE-mediated resistance to antibacterial compounds.
Materials and methods
The bacterial strains used in this study are shown in Table 1 (Baba et al., 2006; Militello et al., 2012); plasmids were a gift from Ashok Bhagwat (Sohail et al., 1990). The lack of 5-methylcytosine in the dcm knockout strain JW1944-2 has been previously reported (Militello et al., 2012). The absence of the sugE gene in JW5738-1 and the rpoS gene in JW5437-1 was confirmed by PCR (data not shown). Liquid bacterial cultures were grown at 37 °C at 250 r.p.m. in either Luria Broth (LB) or M9 minimal media containing 0.4% glucose (Difco). Ampicillin was added to liquid cultures containing dcm plasmids at 25 μg mL−1. Solid cultures were grown at 37 °C in the same media containing 15 grams of agar per liter, and when necessary ampicillin was added at 50 μg mL−1. All experiments to assess the sensitivity of strains to antibacterial compounds were performed in minimal media containing glucose as many QACs precipitate in LB.
Table 1. Bacterial strains used in this study
|Wild-type||Wild-type strain BW25113 [F-, Δ(araB-araD)567, ΔlacZ4787(::rrnB-3), λ-, rph-1, Δ(rhaD-rhaB)568, hsdR514]||Baba et al. (2006)|
|Δdcm||dcm knockout strain JW1944-2 [F-, Δ(araB-araD)567, ΔlacZ4787(::rrnB-3), λ-, Δdcm-735::kan, rph-1, Δ(rhaD-rhaB)568, hsdR514]||Baba et al. (2006)|
|ΔsugE||sugE knockout strain JW5738-1 [F-, Δ(araB-araD)567, ΔlacZ4787(::rrnB-3), λ-, ΔsugE775::kan, rph-1, Δ(rhaD-rhaB)568, hsdR514]||Baba et al. (2006)|
|ΔrpoS||rpoS knockout strain JW5437-1 [F-, Δ(araB-araD)567, ΔlacZ4787(::rrnB-3), λ-, ΔrpoS746::kan, rph-1, Δ(rhaD-rhaB)568, hsdR514]||Baba et al. (2006)|
|Wild-type/pDcm-9||Wild-type strain with truncated dcm allele on plasmid||Militello et al. (2012)|
|Wild-type/pDcm-21||Wild-type strain with functional dcm allele on plasmid||Militello et al. (2012)|
|Δdcm/pDcm-9||dcm knockout strain with truncated dcm allele on plasmid||Militello et al. (2012)|
|Δdcm/pDcm-21||dcm knockout strain with functional dcm allele on plasmid||Militello et al. (2012)|
Quantitative PCR (qPCR)
Bacteria were grown in LB at 37 °C at 250 r.p.m. to early logarithmic phase (A600 nm of c. 0.45) and early stationary phase (A600 nm c. 3.0). Total RNA was isolated from 3–4 mL of bacteria cultures using the MasterPure RNA Isolation Kit (Epicentre). For 5-azacytidine experiments, the drug (Sigma-Aldrich) was dissolved in 1X phosphate-buffered saline (PBS), and PBS was added to untreated samples as a control. RNA quality was assessed using bioanalysis at the University of Rochester Genomics Research Center. Prior to reverse transcription, RNA was treated with RQ1 RNase-free DNase (Promega). One microgram of total RNA was used for reverse transcription using the New England BioLabs Protoscript kit with random primers. cDNA was used as a template for qPCR reactions on a Stratagene MX3000p machine. All reactions were run in triplicate or quadruplicate (technical replicates), and each experiment was performed 3–4 times (biological replicates). Data were normalized to the levels of malate dehydrogenase (mdh) using the ΔΔCt method (Livak & Schmittgen, 2001). The primer sequences are listed in Table S1.
Disk diffusion assays
The following filter sterilized compounds were used in the assays: 5% dimethylethylhexadecyl ammonium bromide (DAB), 5% hexadecyltrimethylammonium bromide (CTAB), 10% benzalkonium chloride (BZA), 10 mg mL−1 ETBR, 10% tetraphenylphosphonium chloride (TPPC), 1 mg mL−1 chloramphenicol, 1 mg mL−1 kanamycin, 1 mg mL−1 gentamicin, and 1 mg mL−1 tetracycline. Overnight cultures were diluted to an A600 nm of 3.0 and swabbed onto M9 minimal media plates. Plates included ampicillin as appropriate for the strains. Antibacterial compounds (20 μL) were spotted onto 6-mm sterile disks and placed on the plates. The plates were incubated at 37 °C for 20 h, and zones of inhibition were measured. As a control, a tetracycline disk was placed on every plate. All compounds were tested at least four separate times (biological replicates).
Minimal inhibitory concentration (MIC) assays and growth determinations
MICs were performed as described as Wiegand et al. (2008) in 48 or 96 well plates with representative compounds used in the disk diffusion assays. Cetylpyridinium chloride (CPC) was only tested in the MIC assays as it precipitated on agar plates. Minimal media containing glucose were used for all experiments, and MICs were determined after 20 h of incubation at 37 °C. Strains were tested a minimum of two times. There was no variation in the MIC for a particular strain and antibacterial compound. Bacterial growth assays were initiated by inoculating M9 minimal media containing glucose with a 1 : 100 dilution of bacteria at A600 nm of 3.0. Bacteria were grown at 37 °C with shaking at 250 r.p.m. and read at A600 nm. Growth analysis was performed four times (biological replicates). Percent growth was calculated by dividing the A600 nm, in the presence of ETBR by the A600 nm in the absence of ETBR and multiplying by 100.
All statistics were performed in R (http://www.R-project.org). A P-value of < 0.05 was considered significant.
Results and discussion
Initially, we determined whether the presence of the dcm gene influences sugE expression. Strains expressing different levels of the dcm gene were constructed (Militello et al., 2012). These included a wild-type strain containing a plasmid with a truncated dcm gene (wild-type/pDcm-9), a wild-type strain with a functional dcm plasmid that overexpresses the dcm gene (wild-type/pDcm-21), a dcm knockout strain with a plasmid with a truncated dcm gene (Δdcm/pDcm-9), and a dcm knockout strain with a functional dcm plasmid (Δdcm/pDcm-21). These strains were grown to early logarithmic phase and early stationary phase, and sugE RNA levels were determined via reverse transcriptase qPCR (Table 2A). SugE RNA levels were c. 7 × higher in the dcm knockout strain with a plasmid with a truncated dcm gene at early stationary phase, and sugE RNA levels returned to normal in the dcm knockout strain with a functional dcm plasmid (complementation; P < 0.05). Thus, the presence of Dcm normally represses sugE expression at early stationary phase. At early logarithmic phase, robust up-regulation of sugE was not observed in the dcm knockout strain with a plasmid with a truncated dcm gene. Overexpression of the dcm had little effect on sugE expression at both early logarithmic and early stationary phase, and may be due to the fact that the E. coli has very few unmethylated sites in the presence of a genomic copy of the dcm gene (Kahramanoglou et al., 2012). In total, the data demonstrate that Dcm influences sugE expression, and the main effect is at stationary phase. This repressive effect of Dcm on gene expression is similar to the repressive effect observed by our laboratory and Kahramanoglou et al. on ribosomal protein genes at stationary phase and suggests that DNA methylation is normally repressive and has an important role during stationary phase (Kahramanoglou et al., 2012; Militello et al., 2012).
Table 2. SugE RNA levels in Escherichia coli strains with different dcm backgrounds (A), following treatment of wild-type cells with 5-azacytidine (5-azaC) (B), and in an rpoS knockout strain (C)
|A||Wild-type/pDcm-21||–||0.54 ± 0.31||0.79 ± 0.31|
|A||Δdcm/pDcm-9||–||1.73 ± 1.61||6.82 ± 3.44|
|A||Δdcm/pDcm-21||–||1.05 ± 0.52(P = 0.45c)||0.65 ± 0.12(P = 0.0060c)|
|B||Wild-type||5 μg mL−1 5-azaC||4.25 ± 1.97(P = 0.023d)||3.43 ± 1.10(P = 0.031d)|
|C|| ΔrpoS ||–||0.076 ± 0.056(P = 0.00061e)||0.048 ± 0.022(P = 8.5E-5e)|
The only known activity of Dcm is methylation of 5′CCWGG3′ sites in DNA. However, some DNA methyltransferases can influence gene expression in a DNA methylation-independent manner. For example, a mutant EcoRII methyltransferase that is not able to methylate DNA can still repress transcription of its own gene (Som & Friedman, 1994). Also, the human DNMT2 enzyme, which has weak DNA methyltransferase activity (Hermann et al., 2003), is able to methylate tRNAAsp and a limited set of other tRNAs (Schaefer et al., 2010). To directly test the model that Dcm-mediated DNA methylation represses sugE expression, wild-type cells were grown in the absence and presence of the DNA methylation inhibitor 5-azacytidine to both early logarithmic phase and early stationary phase, and sugE RNA levels were quantified by qPCR (Table 2B). We observed c. 3–4 × more sugE RNA in the 5-azacytidine treated cells at both early logarithmic and early stationary phase (P < 0.05). Although it was not surprising to observe up-regulation of sugE in the presence of 5-azacytidine at stationary phase based on the qPCR data given above, we were surprised to see an effect at early logarithmic phase, and the magnitude of the effect was similar to that at early stationary phase. This may be due to the fact that there is indeed a small repressive effect of DNA methylation on sugE expression at early logarithmic phase, and/or stationary phase cells that are not rapidly dividing are not as likely to incorporate as much 5-azacytidine into DNA. In addition, 5-azacytidine is known to be toxic to E. coli in killing assays (Bhagwat & Roberts, 1987; Betham et al., 2010). In our experiments, there are lower A600 nm readings only after c. 2.5 h of growth (Fig. S2), which is after the point in which the early logarithmic phase RNA was isolated. As a whole, the 5-azacytidine data are consistent with the dcm knockout data which suggest Dcm-mediated DNA methylation represses sugE expression. Yet, we cannot rule out that sugE expression is also increased by cell stress, changes in growth rate, and/or Dcm has both DNA methylation dependent and independent functions.
Next, we were interested in determining how Dcm influences sugE expression. Although we were originally intrigued by the presence of 5′CCWGG3′ sites in the sugE promoter and gene body, Kahramanoglou et al. proposed a model for Dcm function where Dcm normally represses expression of the sigma factor rpoS (Kahramanoglou et al., 2012). Loss of the dcm gene then leads to increased expression of rpoS and rpoS-dependent genes. The model was supported by increased expression of rpoS in the absence of the dcm gene in microarray, qPCR, and Western blot experiments (Kahramanoglou et al., 2012). To determine whether this model could apply to sugE, we determined whether the sugE gene is under control of RpoS itself by measuring sugE RNA levels via qPCR in an rpoS knockout strain. In the rpoS knockout strain, sugE RNA levels were c. 14-fold lower at logarithmic phase and c. 25-fold lower at stationary phase (Table 2C, P < 0.05). Thus, a simple model is that Dcm normally represses rpoS expression, which is required for robust sugE expression. In the absence of the dcm gene, sugE is expressed at a higher level in an rpoS-dependent manner. This model does not preclude Dcm directly influencing sugE expression via methylation of 5′CCWGG3′ sites. Determining the precise mechanism by which Dcm influences rpoS expression will be a high priority. Kahramanoglou et al. have identified 5′CCWGG3′ sites that could be required for direct Dcm-mediated repression of rpoS expression (Kahramanoglou et al., 2012). 5′CCWGG3′ sites are found in the gene body, and 5′ flanking region that harbors multiple promoters (Fig. S1B).
Next, we were interested in determining whether Dcm influences sensitivity to antibacterial compounds via increased expression of sugE. We characterized the sensitivity of the wild-type strain, dcm knockout strain, and sugE knockout strain to several antibacterial compounds using disk diffusion assays (Table 3) and MIC assays (Table 4). The compounds were chosen based on potential SugE substrates that are QACs (BZA, CTAB, CPC, DAB), Lip. Cat. Cmpds (ETBR, TPPC), and antibiotics that have not been associated with SugE-mediated resistance in most reports (chloramphenicol, gentamicin, kanamycin, tetracycline) (Nishino & Yamaguchi, 2001; Chung & Saier, 2002; He et al., 2011; Cruz et al., 2013). Significant differences were not observed for the majority of compounds including QACs. It should be noted that in E. coli, SugE-mediated resistance to QACs such as CTAB in previous studies was generated by overexpression from high copy number plasmids (e.g. pUC series) (Chung & Saier, 2002). SugE knockout cells may not have the reverse phenotype of sugE overexpressing cells as the levels of SugE protein in overexpressing cells may be extremely high. However, there was a statistically significant difference (P < 0.05) in ETBR sensitivity in the disk diffusion assays, and the same differences were found in the MIC assays. In these assays, the sugE knockout strain was more sensitive to ETBR indicating that SugE normally protects the cell against this compound. The simplest model is that SugE is able to pump ETBR out of the cell, as SugE has been shown previously to bind to ETBR (Sikora & Turner, 2005). The dcm knockout strain was more resistant to ETBR in both assays and growth curve analysis (Fig. S3). Increased sensitivity to ETBR was also observed in complemented dcm strains (Table 3). These data indicate that there is an inverse relationship between the presence of the dcm gene and ETBR resistance. Based on the qPCR and drug susceptibility data, our model is that increased sugE expression in the absence of Dcm is responsible for ETBR resistance.
Table 3. Sensitivity of Escherichia coli strains with different dcm and sugE backgrounds to antimicrobial compounds in disk diffusion assays
|QACs||BZA||16.4 (0.55)||16.9 (2.88)||–||–||16.4 (0.55)||0.245|
|CTAB||6.8 (1.10)||7.3 (0.84)||–||–||7.3 (0.45)||0.598|
|DAB||6.9 (0.89)||7.0 (0.00)||–||–||7.0 (0.71)||0.997|
|Lip. Cat. Cmpds||ETBR||17.6 (1.52)||16.0 (1.41)||–||–||19.2 (0.84)||0.016|
|ETBR||–||–||14.2 (1.66)||17.8 (2.76)||–||0.043c|
|TPPC||13.0 (2.92)||13.2 (3.83)||–||–||15.7 (0.98)||0.290|
|Antibiotics||Chloramphenicol||21.5 (1.94)||22.7 (2.99)||–||–||22.6 (1.95)||0.494|
|Gentamicin||26.3 (1.71)||26.0 (1.47)||–||–||26.4 (2.43)||0.957|
|Tetracycline||24.4 (0.89)||25.0 (0.71)||–||–||24.8 (0.45)||0.490|
|Tetracycline||–||–||25.1 (2.10)||24.9 (0.74)||–||0.849c|
Table 4. MIC values (μg mL−1) of different antibacterial compounds for different Escherichia coli strains
|ETBR||Lip. Cat. Cmpd||70||100||60|
The results of Sulavik et al. and Nishino et al. indicate that there are several transporter genes that are linked to ETBR resistance via overexpression and knockout studies including acrAB, acrEF, emrE, mdfA, tolC, yhiUV, and ydhE (Nishino & Yamaguchi, 2001; Sulavik et al., 2001). The biggest effect was with acrAB, as the MIC increased > 32-fold when acrAB was overexpressed and decreased > 250-fold when acrAB was disrupted. Thus, we were interested to know if there are other transporters in addition to SugE that are up-regulated in the absence of cytosine DNA methylation that could contribute to ETBR resistance. We are currently using DNA microarrays to generate gene expression profiles of wild-type cells, dcm knockout cells, and wild-type cells treated with 5-azacytidine at both logarithmic phase and stationary phase. In initial experiments, we observed no transporters from the list above that were up-regulated both in the absence of dcm and presence of 5-azacytidine (> twofold) (K.T. Militello, R.D. Simon, A.H. Mandarano, S.M. Hennick & A.C. DiNatale, in preparation). Moreover, none of the transporters listed above were up-regulated > twofold in the absence of dcm alone. Thus, our model is that SugE is responsible for the ETBR resistance observed, but it is not possible at this point to rule out the effect of other transporters on ETBR resistance or small contributions by multiple transporters that result in a detectible change in ETBR resistance. In total, our experiments have uncovered a new and unexpected phenotype for the loss of Dcm; changes in sensitivity to ETBR. Our data also brings up the possibility that potential changes in DNA methylation levels due to nutritional status, presence of restriction-modification systems, and/or epigenetic mechanisms may influence the sensitivity of prokaryotes to antibacterial compounds through changes in gene expression and thus link specific environments to differential antibiotic resistance.
We thank the Geneseo Foundation for funding, Ashok Bhagwat (Wayne State University) for plasmid DNAs, and Devin Chandler-Militello, Sarah Ackerman, Leanne Chen, and Erika Valentine for manuscript editing.