Introduction

The heavy overuse and misuse of antibiotics due to their increased demand in several sectors have resulted in antimicrobial resistance as predicted by Alexander Fleming in 1945 (Levy 1992) Bacteria develop genetic resistance against antibiotics as an evolutionary response involving diverse mechanisms. While most of the studies performed in antimicrobial resistance focus on heritable resistance (that requires mutations), the ability of bacteria to resist antibiotics without any genetic alterations has been largely ignored. The non-heritable and transient ability of bacteria to resist antibiotics is termed as phenotypic antibiotic resistance (PAR) which primarily include drug indifference, persistence, biofilm formation, and reduced permeability of antibiotics or their increased efflux (Corona and Martinez 2013). PAR has been shown to be triggered by salicylate and other related compounds like acetate, acetyl salicylate (aspirin), benzoate, DMSO, and 1-methyl-2-pyrrolidinone in Escherichia coli (Rosner 1985). Salicylate has been shown to induce the marRAB (multiple antibiotic resistance) operon which results in the expression of MarA, a DNA binding transcription factor. MarA alters the expression of membrane proteins and efflux pumps like AcrAB–TolC (Cohen et al. 1993; Price et al. 2000). In a recent study, the initial observations have been further strengthened by demonstrating that sodium salicylate induces antibiotic resistance and alters growth in E. coli (Bhaskarla et al. 2016).

The bgl operon involved in the hydrolysis of plant-derived aromatic β-glucosides such as salicin and arbutin is silent and uninducible in most gut isolates (Mukerji and Mahadevan 1997b). A variety of cis- and trans-acting mutations can transcriptionally activate the operon (Reynolds et al. 1981, 1986; Lopilato and Wright 1990; Schnetz and Rak 1992; Schnetz 1995; Singh et al. 1995; Mukerji and Mahadevan, 1997a). In our laboratory, it has been previously observed that the activated bgl operon confers several advantages under selective conditions in addition to enabling the utilization of aromatic β-glucosides as sole carbon source. By enhancing the metabolic capability of the cell, it can confer a growth advantage in stationary phase (Madan et al. 2005). Also, Bgl+ bacteria actively metabolizing the aromatic β-glucosides salicin and arbutin can evade predators such as soil amoebae and nematodes (Sonowal et al. 2013).

Since salicylate is structurally similar to saligenin (SG) and hydroquinone (HQ) (aromatic part of salicin and arbutin respectively) and is known to induce phenotypic antibiotic resistance, we hypothesized that in the case of Bgl+ bacteria that are actively metabolizing salicin or arbutin, the saligenin or hydroquinone released might be involved in conferring phenotypic antibiotic resistance. In such a case, Bgl+ bacteria could exhibit a significant growth advantage in environments where both aromatic β-glucosides and antibiotics are present. To investigate this possibility, laboratory strains of E. coli and natural isolates of Klebsiella were used to perform assays to detect resistance against the two structurally and mechanistically different antibiotics nalidixic acid and ampicillin. Unlike E. coli isolates, Klebsiella aerogenes strains exhibit a Bgl+ constitutive phenotype by expressing a homolog of the bgl operon (Raghunand and Mahadevan 2003). Several Klebsiella strains also carry the casRAB genetic system conferring the ability to hydrolyze cellobiose, arbutin and salicin (Lai et al. 1997). The genes of the casRAB operon are significantly similar to the permease and phospho β-glucosidase genes of the bgl operon of E. coli. The active state of β-glucoside utilization in Klebsiella could be related to the fact that their primary habitat is soil where aromatic β-glucosides are known to be present.

The results of the experiments described below support the possibility that phenotypic antibiotic resistance is another selective advantage conferred by the activated form of the bgl operon. We also report that this low level of non-heritable phenotypic resistance in Bgl+ bacteria can further facilitate the emergence of a genetically resistant population in the presence of aromatic β-glucosides and sub-lethal concentrations of antibiotics.

Materials and methods

Chemicals, media and growth conditions

Bacterial strains (Table 1) were grown in L Broth (HiMedia, India/Difco, USA) or L Agar (1.5%) or M9 minimal salts (1X M9 salts, 1 mM MgSO4, 0.1 mM CaCl2, 0.01% vitamin B1, 0.2% Cas amino acids) supplemented with 0.4% sodium succinate (Sigma Aldrich) at 37 ℃ either static or shaking at 220 rpm, unless specified. Salicin and arbutin utilization phenotypes were checked on MacConkey medium (1% Tryptone, 0.6% Na Taurocholate, 1.5% Agar, 0.015% neutral red) supplemented with either 1% salicin or 1% arbutin. Sodium salicylate (NaSal), saligenin (SG), salicin, and arbutin were obtained from Sigma Aldrich, USA.

Table 1 List of bacterial strains and plasmids used in this study
Full size table

DNA manipulations

Plasmid DNA transformations and other DNA manipulations were performed as described previously (Sambrook et al. 1989, 2001). Linear DNA transformation was carried out by electroporation using a BIORAD electroporator as recommended by the manufacturer.

One-step chromosomal gene inactivation for strain construction

All chromosomal gene inactivations were generated by following the protocol described previously (Datsenko and Wanner 2000). Hybrid cassettes carrying Chloramphenicol or Kanamycin resistance marker were amplified from plasmid pKD3 or pKD4, respectively, using primers (Table 2) that carried a region of homology to the target gene. The amplified fragments were digested with DpnI to get rid of the vector backbone and the linear DNA of appropriate size was further purified from gel using GeneJET Gel Extraction Kit (Thermo Scientific/EU). The purified DNA was electroporated into the bacterial strain of interest carrying pKD46 plasmid. The pKD46 plasmid has a temperature-sensitive origin of replication and expresses λredgam genes under arabinose promoter. Electroporated cells were recovered by growing in L Broth for 18 h at 37 ℃, 220 rpm. The knockouts were selected on L Agar plates containing the appropriate antibiotic and confirmed by colony PCR using confirmation primers.

Table 2 List of primers used in this study
Full size table

Quantitative gene expression (qRT-PCR)

Total RNA was extracted using the acid phenol method (Harwani et al. 2012) from bacterial cells exposed to different growth conditions and treated with DNAse I (Thermo Fisher Scientific). cDNA was prepared from 1 μg of DNAse I-treated RNA using RevertAid First Strand cDNA synthesis kit (Thermo Fisher Scientific, K1621) as per the instruction manual. Transcript levels of target genes were analyzed by performing quantitative real time reverse transcription PCR (Quant Studio 6 Flex System Applied Biosystems) using gene-specific primers (Table 2). The primers were designed using EcoCyc database and NCBI BLAST. The reactions were prepared using Maxima SYBR green qPCR master mix kit. Analysis of gene expression was performed using comparative CT method (2−ΔΔCT) (Schmittgen and Livak 2008). Values were normalized to the 16S rRNA transcript. Graphs were generated using GraphPad Prism software.

Analysis of phenotypic antibiotic resistance on solid media

Cells were grown overnight in M9 salts + 0.4% succinate at 37 ℃, 220 rpm. The cultures were diluted to OD600 ~ 0.8. Cells were serially diluted and spotted on M9 succinate agar plates containing different compounds (2-mM NaSal, 1.5-mM SG, 5-mM Salicin, or 5-mM Arbutin) in the presence and absence of antibiotics (concentrations indicated in the respective figures). Plates were kept for overnight incubation at 37 ℃. Imaging was performed using ChemiDoc™ XRS + Imaging System (Bio-Rad). Experiments were repeated thrice.

Quantification of phenotypic antibiotic resistance

Cells were grown overnight in M9 salts + 0.4% succinate at 37 ℃, 220 rpm. The cultures were diluted to OD600 ~ 0.8 and plated on appropriate media as above. Strips containing a concentration gradient of antibiotic (Ezy MIC strip, HiMedia) were placed on the evenly spread bacterial lawns and kept for overnight incubation at 37 ℃. The MIC was read at the point of intersection of bacterial growth and antibiotic strip. Experiments were repeated twice or thrice.

Determination of frequency of genetic mutants

Cells were grown overnight in M9 salts + 0.4% succinate at 37 ℃, 220 rpm. The cultures were diluted to OD600 ~ 0.5 and 0.1% was inoculated in three different growth media i.e. (a) M9 salts + 0.4% succinate + 5 mM salicin (control) (b) M9 salts + 0.4% succinate + 1 μg/ml of nalidixic acid (control), and (c) M9 salts + 0.4% succinate + 1 μg/ml of nalidixic acid + 5 mM salicin (test experiment). For the control experiments, 12 independent cultures were inoculated for each strain and in the case of test experiments, 20 independent cultures were inoculated. Cells were grown for 24 h at 37 ℃, 220 rpm. Nalidixic acid resistant colonies were selected on LB agar plates containing 4, 8, 16, and 32 μg/ml of nalidixic acid. Fractions of nalidixic acid-resistant colonies were calculated as Fraction of NALR = CFU/ml on NAL plate/CFU/ml on LB agar plate. Plots were generated using MATLAB. A similar set of experiments was performed using 2 μg/ml of NAL (data not shown).

Confirmation of genetic resistance

Colonies from selection plates were randomly isolated and grown overnight in L broth at 37 ℃, 220 rpm without addition of nalidixic acid to OD600 ~ 1. Cultures were serially diluted in 96-well plates and spotted on LB agar plates containing higher concentrations of nalidixic acid (16, 32, 64, and 128 μg/ml). Plates were kept for overnight incubation at 37 ℃. Imaging was performed using ChemiDoc™ XRS + Imaging System (Bio-Rad). Experiments were repeated twice.

Sequencing of gyrA locus in mutants

The gyrA locus was amplified from 18 colonies (6 each from 3 nalidixic acid selection plates) using gene specific primers (Table 2) and Phusion high-fidelity DNA polymerase (Thermo Fisher Scientific) and Sanger sequencing was performed (Eurofins Scientific). The mutations were analyzed using various online tools (NCBI nucleotide BLAST, EXPASy, and ClustalW).

Co-culture experiments

Strains were individually grown overnight in M9 salts + 0.4% succinate at 37 ℃, 220 rpm. The cultures were diluted to OD600 ~ 0.5. The bacterial strains to be competed were mixed in M9 salts + 0.4% succinate medium in 1:1 ratio and grown at 37 ℃, 220 rpm. At every 4 h of interval, 1% of the previous culture was inoculated in fresh medium and grown to a total period of 12 h. At every sub-culture, aliquots were plated to determine CFU/ml. CFU/ml for the parent strain was calculated by subtracting the CFU on LB agar plates containing 16 μg/ml nalidixic acid (selection concentration) from CFU on LB agar plate (no antibiotic).

Results

Aromatic β-glucosides induce MarA-dependent PAR in Bgl+ strains

In an attempt to assess the induction of the marRAB operon upon hydrolysis of the aromatic β-glucosides salicin and arbutin, we initially measured the transcript levels of marA and its downstream target genes acrB and tolC in the Bgl+E. coli strain MA200. Total RNA was isolated from exponential phase cultures of MA200 exposed to specific growth conditions. The mRNA levels of marA, acrB, and tolC were quantified by qRT-PCR using gene specific primers. Figure 1a represents the fold change of specific mRNA normalized to the 16S rRNA gene rrnC (internal control). The mRNA levels of marA, acrB, and tolC were two–threefold higher when cells were grown in the presence of saligenin (SG), salicin and arbutin compared to M9 succinate control, indicating that saligenin, salicin and arbutin can induce the expression of the marRAB operon. To test whether this can lead to phenotypic antibiotics resistance, we examined the ability of the laboratory E. coli strain MA200 and pathogenic isolates of Klebsiella pneumoniae to grow in the presence of different concentrations of antibiotics. Overnight cultures of the strains were spotted on M9 succinate agar plates containing NaSal, SG, salicin, or arbutin along with nalidixic acid or ampicillin. The strain MA200 was able to tolerate 3 μg/ml of nalidixic acid and 1.5 μg/ml of ampicillin in the presence of aromatic compounds (Fig. 1b) compared to M9 succinate control. The Klebsiella pneumoniae strain was able to resist 3 μg/ml of nalidixic acid upon exposure to aromatic compounds (Fig. S1) as opposed to the M9 succinate control.

Fig. 1
figure 1

Aromatic β glucosides induce PAR in Bgl+ bacteria. a qRT-PCR analysis of marA, acrB, and tolC in MA200 exposed to different growth conditions indicated. Mean ± SEM is plotted. (n* = 3, ANOVA, p < 0.001 = *** p < 0.01 = **, p < 0.05 = * p > 0.05 = ns). b Growth assay of MA200 and MA200ΔmarA on M9 succinate agar plates containing NaSal, SG, salicin and arbutin in the presence and absence of nalidixic acid or ampicillin. Representative image of n* = 3 biological replicates

Full size image

To investigate the role of marA in the PAR phenotype observed, the strain MA200ΔmarA was constructed and growth assays were performed as before. The deletion strain MA200ΔmarA showed loss of the resistant phenotype upon exposure to antibiotics in the presence of saligenin, salicin (SG) and arbutin (Fig. 1b). Thus, Bgl+ bacteria exhibit marA-dependent PAR in the presence of aromatic β-glucosides and low drug pressure. Owing to technical limitations and toxicity issues, hydroquinone (aromatic byproduct of arbutin) was not included in the assays. In these studies, we have also used lower concentrations of sodium salicylate (positive control) and saligenin compared to the concentrations used in a previous study (Cohen et al. 1993).

Involvement of the bgl operon in conferring PAR

To address the question whether the bgl operon is involved in the phenotypic antibiotics resistance, the bgl operon was deleted in the strain MA200. The deletion was further confirmed by the loss of the Bgl+ phenotype on MacConkey salicin plates. Growth assays were carried out as described earlier using the strain MA200Δbgl and the permease defective strain MA200-1. Both strains showed loss of the resistant phenotype at 3 μg/ml of nalidixic acid and 1.5 μg/ml of ampicillin when grown in the presence of salicin (Fig. 2d). However, the strains were able to grow on the antibiotic plate in the presence of 1.5-mM SG (Fig. 2c) and 2-mM NaSal (Fig. 2b) indicating that the functional bgl operon is necessary to generate saligenin by the hydrolysis of salicin and can be bypassed if the hydrolyzed product is provided directly in the medium.

Fig. 2
figure 2

Involvement of the bgl operon and AcrAB–TolC efflux pump in conferring the PAR phenotype in Bgl+ strains. Growth assays of the bgl deletion strain MA200Δbgl, permease-deficient strain MA200-1, and acrB deletion strain MA200ΔacrB in the presence and absence of antibiotics (nalidixic acid or ampicillin).a Growth pattern of these strains in M9 succinate medium. b, c, d Growth of these strains in M9 succinate medium containing NaSal, SG, and salicin, respectively. Representative image of n* = 3 biological replicates

Full size image

Involvement of multidrug efflux transporters in the acquisition of PAR in Bgl+ bacteria

The results reported so far imply that the release of the aromatic moiety saligenin/ hydroquinone upon hydrolysis of salicin/arbutin is responsible for the phenotypic antibiotic resistance in a marA-dependent fashion. Consistent with this observation and the enhanced transcript levels of acrB and tolC genes (Fig. 1a), we presumed that the AcrAB–TolC efflux pump could be playing an important role in antibiotics resistance. AcrAB–TolC multidrug efflux pump, a member of RND superfamily (Resistance–Nodulation–Division) (Saier et al. 1994; Tseng et al. 1999), has been shown to be involved in antimicrobial resistance in bacterial pathogens (Poole 2004; Piddock 2006). It has also been observed earlier that AcrAB–TolC efflux pump is over-expressed in clinical isolates of E. coli in spite of the presence of other Acr pumps (Mazzariol et al. 2000). To test the possible role of AcrAB–TolC efflux pump in the resistance phenotype, we constructed an acrB deletion strain in MA200 background. Surprisingly, the patterns of growth in the presence of NaSal and SG were similar (Fig. 2b, c) in the case of MA200 and MA200ΔacrB indicating that AcrAB–TolC efflux pump is not solely responsible for the MarA-dependent phenotypic antibiotic resistance in Bgl+ bacteria. This was consistent with a previous report by Shuster et al. where they have demonstrated that in the absence of AcrB, other TolC-dependent MDTs (Multidrug Transporters) provide necessary support in the acquisition of high-level antibiotic resistance (Shuster et al. 2016).

To determine the involvement of other multidrug transporters, we estimated the transcript levels of acrD, acrF, mdtF, and macB by qRT-PCR using gene-specific primers. acrD codes for a multidrug efflux pump RND permease AcrD, a homolog of AcrB (Ma et al. 1994). AcrF, a key component of AcrEF–TolC efflux pump, is not highly expressed in wild-type E. coli strains (Ma et al. 1994). MdtF, an inner membrane subunit of MdtEF–TolC efflux pump, belongs to the RND superfamily. macB codes for the ATPase subunit of an ABC-type tripartite MacAB–TolC efflux pump. All these TolC-dependent multidrug efflux pumps have been shown to be involved in antibiotic resistance. Results of quantitative RT-PCR using RNA isolated from exponential phase cultures of MA200ΔacrB confirmed 1.5–2-fold higher expression levels of marA, acrD, acrF, and mtdF in the presence of the aromatic compounds (Fig. 3). However, the transcript levels of macB were unaffected in the presence of SG, salicin and arbutin indicating that macB is not involved in the phenomenon. These results suggest the involvement of other multidrug transporters apart from AcrAB–TolC in phenotypic antibiotic resistance in Bgl+ bacteria.

Fig. 3
figure 3

Involvement of multi-drug transporters in PAR. qRT-PCR analysis of marA, acrD, acrF, macB, and mdtF in MA200ΔacrB upon exposure to growth conditions indicated. Mean ± SEM is plotted. (n* = 3, ANOVA, p < 0.001 = ***, p < 0.01 = **, p < 0.05 = * p > 0.05 = ns)

Full size image

Since TolC serves as a common outer membrane channel for all the tested efflux pumps, its involvement was tested. To investigate its role, tolC was deleted in MA200 strain background and growth of the mutant strain was monitored as before. Growth of this strain was significantly reduced upon exposure to 3 μg/ml of nalidixic acid or 1.5 μg/ml of ampicillin even in the presence of the aromatic compounds (Fig. 4). Therefore, the PAR observed in the Bgl+ strain is through over-expression of TolC-dependent efflux pumps.

Fig. 4
figure 4

Role of tolC in PAR Growth assays of the tolC deletion strain MA200ΔtolC in the presence/absence of antibiotics. a M9 succinate agar (control). b, c, d, e M9 succinate agar containing NaSal, SG, salicin, and arbutin, respectively. Representative image of n* = 3 biological replicates

Full size image

Quantification of PAR using EzyMIC strip assay

To estimate the level of PAR in Bgl+ bacteria, we performed EzyMIC strip-based assays using the E. coli strain MA200 and two natural isolates of Klebsiella (strains OK8 and ANK). The objective was to determine the alterations in the MIC for antibiotics when Bgl+ bacteria were exposed to aromatic compounds. MA200 grown in the presence of salicin or arbutin showed an increase in MIC of antimicrobial agents (Table 3). MIC for the tested antimicrobial compounds were two–threefold higher when MA200 was exposed to NaSal (positive control) and SG (aromatic component of salicin) compared to control. Both the Klebsiella strains OK8 and ANK were found to be resistant to ampicillin and trimethoprim (MIC < 256 μg/ml and 32 μg/ml, respectively). OK8 showed a twofold increase in MIC for nalidixic acid when growth medium contained NaSal, SG, or salicin as compared to minimal medium (Table 3). A fourfold increase in MIC for nalidixic acid was observed in the presence of arbutin. Upon exposure to ciprofloxacin, OK8 was completely sensitive when grown in control medium (M9 succinate), but the strain was able to resist the antibiotic when supplemented with any of the compounds NaSal, SG, arbutin, or salicin (Table 3). Similar results were observed in the natural isolate Klebsiella ANK. Approximately, 1.5–1.6-fold increase in MIC of nalidixic acid was observed in the presence of NaSal, SG, and salicin. For unknown reasons, MIC was found to be slightly lower (4.5 ± 0.5 μg/ml) in the case of the ANK strain when exposed to arbutin as compared to control (5 ± 1 μg/ml). However, in the presence of ciprofloxacin, we observed an increase of ≈ 1.5- and 2-fold for the ANK strain upon exposure to NaSal, SG, salicin, and arbutin.

Table 3 Ezy MIC strip-based assay (n* = 3, mean ± SEM). Both the Klebsiella strains were found to be intrinsically resistant to ampicillin, oxacillin, and trimethoprim
Full size table

Effect of PAR on the emergence of genetic resistance in Bgl+ bacteria

Earlier studies have shown an enrichment of genetically resistant bacteria upon exposure to sub-MIC concentrations of antibiotics (Gullberg et al. 2011; Liu et al. 2011; Andersson and Hughes 2012, 2014; Sandegren 2014). Our observation of PAR in Bgl+ bacteria growing in the presence of aromatic β-glucosides and low doses of antibiotics raised the possibility of facilitation of the emergence of genetic resistance. To test this possibility, we designed an experimental set up (Fig. S2) under low drug pressure settings (0.3 X MIC) wherein bacterial growth is not attenuated. Multiple independent cultures of MA200 and MA200Δbgl strains were pre-exposed to sub-inhibitory concentration of nalidixic acid (NAL, 1 μg/ml) and salicin (SAL, 5 mM) individually or in combination in M9 succinate medium. Nalidixic acid-resistant variants were selected on LB agar plates containing higher concentrations of nalidixic acid (4, 8, 16, 32 μg/ml) and the fraction of resistant mutants was calculated. Both the strains showed similar frequency of mutants upon selection on nalidixic acid plates when grown in the presence of either NAL or SAL (Figs. S3 and S4). However, a significant increase in the frequency of mutants was observed in the case of the Bgl+ strain MA200 (Fig. 5a) grown in the presence of NAL and SAL compared to the isogenic Bgl− strain MA200Δbgl (Fig. 5b). For example, in the case of MA200, 17/20 and 15/20 cultures showed resistance to 8 and 16 μg/ml of nalidixic acid, respectively (Fig. 5c). A significant reduction in frequency of mutants was detected in the case of the Bgl− strain MA200Δbgl (Fig. 5c). For instance, only 7/20 and 2/20 cultures showed resistance to 8 and 16 μg/ml of nalidixic acid, respectively. Similar results were obtained when strains were pre-exposed to 2 μg/ml nalidixic acid (data not shown). These findings indicate that the presence of the activated form of the bgl operon in the genome facilitates the survival of bacteria in environments in which both aromatic β-glucosides and antibiotics are present by inducing PAR and also by enhancing the emergence of antibiotic resistant mutants.

Fig. 5
figure 5

Emergence of genetic mutants in a Bgl+ strain in the presence of salicin and low drug concentrations (a) Frequency of nalidixic acid resistant mutants obtained when strain MA200 was exposed to M9 succinate + 5 mM SAL + 1 μg/ml NAL. b Frequency of mutants when strain MA200Δbgl was exposed to similar culture condition.c Tabular summary of frequency of mutants at each concentration. X axis denotes individual culture tube number, Y axis denotes fraction of nalidixic acid resistant mutants normalized to LB control for individual culture tube (height of the bars), Z axis denotes nalidixic acid concentration used for selection. Shades of the bars represent individual concentration used for selection viz. 4, 8, 16, and 32 μg/ml. Empty spaces without bar represent no growth

Full size image

Confirmation of genetic resistance in Bgl+ mutant isolates

Detection of nalidixic acid resistant colonies on selection media at higher antibiotic concentrations (16 and 32 μg/ml) seen in the previous section can be explained only by invoking genetic resistance. To verify this, randomly isolated colonies from selection plates used in previous experiment were grown in L broth in the absence of antibiotic and plated on medium containing increasing concentrations of nalidixic acid. For instance, colonies isolated from selection plate containing 8 μg/ml of nalidixic acid were exposed to 16, 32, 64 and 128 μg/ml of nalidixic acid along with the parent strain MA200. Growth assays (Fig. 6a) showed that randomly picked isolates from selection plates containing 8 μg/ml of nalidixic acid can resist up to 128 μg/ml of nalidixic acid, while the parent showed sensitivity to even lower concentrations. Similar growth patterns were seen with colonies from the selection plates containing 16 and 32 μg/ml nalidixic acid (data not shown). Therefore, growth conditions promoting PAR can lead to the acquisition of mutations that can confer resistance higher concentration of antibiotics.

Fig. 6
figure 6

Characterisation of the NalR mutants a, Confirmation of heritable resistance by streaking randomly chosen colonies that appeared on selection plates containing 8 μg/ml of nalidixic acid on plates with higher concentrations of nalidixic acid (16, 32, 64, and 128 μg/ml). b List of mutations found in gyrA locus of 18 randomly isolated colonies from selection plates

Full size image

Characterization of Bgl+ nalidixic acid-resistant mutants

Since nalidixic acid belongs to the class of quinolones and is a known DNA gyrase inhibitor, we sequenced the gyrA locus of resistant strains using gene-specific primers. The gyrA subunit was amplified from 18 nalidixic acid resistant colonies which were randomly picked from selection plates (6 colonies each that appeared on 8, 16, and 32 μg/ml of nalidixic acid, respectively). The DNA sequences of mutants were compared with gyrA sequence of the sensitive parent (MA200 Bgl+). Mutations in the mar locus and efflux pumps were ruled out as cross resistance to ampicillin in the isolated mutants was not observed (data not shown). Importantly, DNA sequencing results revealed that all the 18 nalidixic acid-resistant colonies carried mutations in their gyrA locus (Fig. 6b). Most of the mutations were found in the QRD region (quinolone resistance determining region; amino acids 67–106) of GyrA subunit. Upon quantifying the resistance against nalidixic acid, it was observed that the mutants show MIC > 16 μg/ml of nalidixic acid (Ezy MIC strip assay). Some of the mutants were completely resistant to the highest concentration used (> 256 μg/ml of nalidixic acid). All the mutants were also able to resist ciprofloxacin as opposed to the parent which was completely sensitive (Table S1). A variety of mutations which include transitions, transversions and deletions were found (Table S1). Interestingly, several mutations viz. S83del, E875K, V54G, P79H, and A119E were novel, and not reported in literature earlier. We also identified various combinations of mutations like S83L and E875K, A67S and E875K, P79H and S83del in the isolates.

Absence of fitness cost in the isolated mutants

Acquisition of genetic antibiotic resistance is often associated with compromised fitness due to the impact of the mutations on the function of the target genes (Björkman and Andersson 2000; Hughes and Andersson 2015; Melnyk et al. 2015). To observe whether there is a fitness cost associated with antibiotic resistance in the mutants described above, exponential phase competition experiments were performed. Three sets of competitions were performed wherein MA200-8A (D87G), MA200-16F (D87N), and MA200-8D (S83L, E875K) were competed against MA200 (parent) in M9 succinate medium. Results indicate the absence of attenuated growth in the case of MA200-8D (S83L, E875K) when competed with the parent MA200 in exponential phase (Fig. 7a) in the absence of drug. The ratios of final versus initial viable cell count of both the strains were similar (Fig. 7b) confirming that Bgl+ nalidixic acid-resistant mutant MA200-8D does not suffer a fitness cost. None of the tested mutants viz. MA200-8A (D87G) and MA200-16F (D87N) showed reduction in growth during co-culture experiments (Fig. S5).

Fig. 7
figure 7

Absence of cost of resistance in the mutants a, Competition in the exponential phase between the parent MA200 and the mutant MA200-8D mixed in 1:1 ratio. b The respective ratio of final verse initial CFU/ml. (n* = 3)

Full size image

Discussion

The existence of cryptic or silent genetic elements such as the bgl operon of Escherichia coli is an evolutionary enigma because the bgl operon is maintained as an integral part of the genome over the course of evolution despite the fact that it is silent in most wild isolates. Specific growth conditions such as the presence of plant-derived aromatic β-glucosides can lead to the selection of the active state brought about by a single mutational event. The activated operon has been shown to provide additional fitness advantages under environmental conditions such as stationary phase and the presence of predators (Madan et al. 2005; Sonowal et al. 2013) indicating its highly dynamic role in survival. In the light of these observations it was important to explore other physiologically relevant conditions where the presence of an activated copy of the allele is beneficial to the bacterium.

The present study directed at the role of the bgl operon in conferring PAR was initiated based on the possibility that the release of the aglycones saligenin and hydroquinone upon the breakdown of salicin and arbutin by Bgl+ bacteria can derepress the marRAB operon. In support of this, growth assays revealed that laboratory strains of E. coli and natural isolates of Klebsiella carrying inducible and constitutive forms of the bgl operon, respectively, were able to resist sub-lethal doses of nalidixic acid and ampicillin upon exposure to NaSal, SG, salicin and arbutin. This finding also confirmed the role of MarA as loss of the phenotype was observed in the ∆marA strain. E. coli strains deleted for the bgl operon or deficient for transport and hydrolysis functions failed to show the phenotype in the presence of salicin or arbutin (data not shown) confirming the role of the activated bgl operon in the phenotype. The underlying mechanism for conferring PAR was suspected to be the increase in efflux due to enhanced transcript levels of the transcription factor MarA and its downstream targets acrB and tolC (AcrAB–TolC efflux pump) in the Bgl+ strain MA200 grown in the presence of aromatic compounds. Although earlier reports showed the enhanced expression of AcrAB–TolC efflux pump in E. coli clinical isolates (Mazzariol et al. 2000; Oethinger et al. 2000; Webber and Piddock 2001), interestingly, deletion of acrB in MA200 background did not affect the ability to resist the antibiotics. This suggested the involvement of additional Multi Drug Transporters (MDTs) of E. coli in conferring the phenotype in Bgl+ bacteria as a recent report (Shuster et al. 2016) also demonstrated the presence of a multidrug transporter interactome in E. coli. Consistent with the hypothesis transcript levels of other TolC-dependent efflux pumps, i.e., acrF, acrD, and mdtF were approximately 1.5–2-fold higher in the ∆acrB strain upon exposure to aromatic compounds in the growth medium. Since the identity of the efflux pump involved was unclear, deletion of the common transporter tolC was created in the Bgl+ background. The ∆tolC strains showed highly attenuated growth in the presence of nalidixic acid and ampicillin. These findings indicate that the function of AcrB is dispensable in specific environmental conditions; in the absence of AcrB, additional MDTs present can take over and provide necessary support for conferring the phenotype.

Upon quantification of PAR, an increase of two–threefold MIC for nalidixic acid, ampicillin and trimethoprim was observed in the case of E. coli MA200. The pathogenic isolate of Klebsiella pneumoniae OK8 showed fourfold increase in MIC when exposed to arbutin. This strain was completely sensitive to ciprofloxacin when grown in minimal medium; however in the presence of NaSal, SG, salicin and arbutin, low level of resistance was observed further strengthening the initial observations. In the case of another natural sewage isolate of Klebsiella ANK, a similar phenomenon was observed upon exposure to both antibiotics and inducers (NaSal, SG, salicin and arbutin). Both Klebsiella strains were found to be intrinsically resistant to ampicillin, trimethoprim, and oxacillin.

Globally, extensive usage of several antibiotics in various sectors (medical, animal husbandry, agriculture etc.) in an unregulated manner has led to a widespread distribution of several antibiotics in concentrations ranging from nanogram to microgram (Baquero et al. 2008; Fram and Belitz 2011; Jiang et al. 2013; Khan et al. 2013; Andersson and Hughes 2014). Furthermore, exposure to sub-lethal doses has been shown to facilitate the rise/enrichment of genetically resistant populations under laboratory settings (Gullberg et al. 2011; Liu et al. 2011). As the presence of phenotypic antibiotic resistance and the possible mechanism involved could be demonstrated in Bgl+ strains which are actively metabolizing salicin or arbutin, it was important to explore the possibility of emergence of genetic resistance triggered by phenotypic resistance at sub-lethal concentrations of the antibiotics. Towards this end, Bgl+ (MA200) and Bgl− (MA200∆bgl) strains were cultured under three different growth conditions (combinations of M9 succinate, salicin and sublethal dose of nalidixic acid). A remarkable increase in the frequency of mutants that show nalidixic acid resistance was observed when Bgl+ bacteria (MA200) were exposed to salicin and sub-lethal concentration of nalidixic acid as compared to the Bgl− strain (MA200∆bgl). However, the precise mechanism behind this observation is unknown. Interestingly, the frequency of mutants remained unaffected when cells were pre-exposed either to salicin or sub-lethal dose of nalidixic acid individually. In conclusion, these results suggest that in environments where aromatic -glucosides and low amount of antibiotics are present, bacteria harboring a functional bgl operon stand a better chance of survival as opposed to bacteria having the silent counterpart.

To characterize the nalidixic acid-resistant mutants further, the nucleotide sequence of the gyrA locus of 18 randomly isolated mutants was analysed as mutations in gyrA are more common in clinical isolates of quinolone resistant E. coli (Nakamura et al. 1989). The sequencing data revealed that indeed all the 18 mutants carried mutations in the gyrA locus and most of them were located in the QRDR—quinolones resistance determining region—spanning the amino acids 67–106 (Yoshida et al. 1990). Both single and combinations of clinically relevant mutations (Lindgren et al. 2003) such as S83L, D87N, and D87G were observed. Some of the identified mutations like S83del, E875K, V54G, P79H, and A119E were novel, not reported in literature earlier. A significant increase in the MIC of both nalidixic acid and ciprofloxacin was observed in the case of the characterized mutants.

Since antimicrobial resistance is often associated with a fitness cost in the absence of the drug because of the impact of the mutation on the functioning of the target gene or protein (Björkman and Andersson 2000; Hughes and Andersson 2015; Melnyk et al. 2015), it was possible that these mutants might carry a fitness cost of drug resistance. Exponential phase competition experiments between the sensitive parent and genetically resistant mutants indicated that the mutants do not carry a fitness cost imposed by resistance at least under the conditions tested. This observation is consistent with an earlier report which suggests that quinolone resistant mutations in gyrA along with other mutations in the genome do not exhibit a cost of resistance in Streptococcus pneumoniae (Gillespie et al. 2002). This can be a possible explanation for the prevalence of quinolone resistant pathogens in clinical settings (Redgrave et al. 2014). In the light of these observations, it is apparent that the presence of the bgl operon in its active form facilitates the enhanced survival of the bacteria in environments in which both aromatic β-glucosides and antibiotics are present, such as soil, thereby providing a strong selective force for its retention in the genome.