Subinhibitory concentrations of antibiotics affect stress and virulence gene expression in Listeria monocytogenes and cause enhanced stress sensitivity but do not affect Caco-2 cell invasion

Authors


Correspondence

Gitte M. Knudsen, National Food Institute, Technical University of Denmark, Soeltofts Plads Bldg. 221, DK-2800 Kongens Lyngby, Denmark. E-mail: gmkn@food.dtu.dk

Abstract

Aims

Antibiotics can act as signal molecules and affect bacterial gene expression, physiology and virulence. The purpose of this study was to determine whether subinhibitory antibiotic concentrations alter gene expression and physiology of Listeria monocytogenes.

Methods and Results

Using an agar-based screening assay with promoter fusions, 14 of 16 antibiotics induced or repressed expression of one or more stress and/or virulence genes. Despite ampicillin-induced up-regulation of PinlA-lacZ expression, Caco-2 cell invasion was not affected. Subinhibitory concentrations of ampicillin and tetracycline caused up- and down-regulation of stress response genes, respectively, but both antibiotics caused increased sensitivity to acid stress. Six combinations of gene-antibiotic were quantified in broth cultures and five of the six resulted in the same expression pattern as the agar-based assay.

Conclusions

Antibiotics affect virulence and/or stress gene expression; however, altered expression could not predict changes in phenotypic behaviour. Subinhibitory concentrations of antibiotics led to increased acid sensitivity, and we speculate that this is attributed to changes in cell envelope or reduced σB-dependent gene expression.

Significance and Impact of the Study

Although subinhibitory concentrations of antibiotics affect gene expression in L. monocytogenes, the changes did not increase virulence but did enhance the acid sensitivity.

Introduction

Listeria monocytogenes is a Gram-positive, facultative intracellular bacterium that can cause severe food-borne infections in immunocompromised humans, elderly and neonates (Vazquez-Boland et al. 2001; Wagner and McLauchlin 2008). Clinical cases of listeriosis are rare; however, the infection is of societal concern owing to a mortality of 20–30% (Vazquez-Boland et al. 2001). The disease is treatable with antibiotics if discovered in time, and first choice of treatment is a β-lactam in combination with an aminoglycoside (Hof 2003; Aarestrup et al. 2007; Kvistholm Jensen et al. 2010). Isolates of L. monocytogenes are susceptible to many antibiotics but L. monocytogenes is intrinsically resistant to nalidixic acid, fosfomycin and third generation cephalosporins (Hof et al. 1997; Hof 2003). Most antibiotics that are effective against L. monocytogenes have a bacteriostatic rather than a bacteriocidal effect (Espaze and Reynaud 1988; Hof et al. 1997). As opposed to the alarming increase in antibiotic resistance in Staphylococcus aureus and Pseudomonas aeruginosa (Rice 2009), clinical and environmental L. monocytogenes isolates continue to have high susceptibility to antibiotics (Hof 2003; Aarestrup et al. 2007; Chen et al. 2010); however, an increase in antibiotic resistance may be developing (Conter et al. 2009; Morvan et al. 2010).

Besides being an opportunistic pathogen, L. monocytogenes is a saprophytic bacterium that is commonly isolated from soil and decaying plant material, including silage (Ivanek et al. 2006; Wagner and McLauchlin 2008; Freitag et al. 2009). In these environments, the bacterium coexists with a range of antibiotic-producing micro-organisms, such as Streptomycetes sp. and filamentous fungi (Driehuis and Oude Elferink 2000; Martinez 2009). It is possible that this potential exposure to low concentration of antibiotics has lead to indigenous resistance to antibiotic produced by coexisting micro-organisms (Martinez 2009) or can explain why most antibiotics are bacteriostatic against L. monocytogenes rather than bactericidal. Clearly, from an evolutionary point of view, survival in the environmental niche would be difficult if L. monocytogenes was sensitive to the antibiotics encountered in nature.

Antibiotics, especially at subinhibitory concentrations, can act as signal molecules aside from their antibacterial effect (Davies et al. 2006; Yim et al. 2007; Ratcliff and Denison 2011). Several studies have shown that antibiotics in subinhibitory concentrations alter the transcriptomic and phenotypic response of pathogenic bacteria such as Bacillus subtilis, Campylobacter jejuni, Escherichia coli, Pasteurella multocida, Ps. aeruginosa and Staph. aureus (Lin et al. 2005; Davies et al. 2006; Linares et al. 2006; Aminov 2009; Nanduri et al. 2009; Almofti et al. 2011; Subrt et al. 2011). Of special concern is the potential increased expression of virulence genes by antibiotics. For instance, in Staph. aureus, cell wall inhibitors induce expression of spa, encoding staphylococcal protein A, and lukE, encoding leukotoxin E, and cause a denser biofilm formation, a virulence trait of Staph. aureus (Subrt et al. 2011). Little is known about the transcriptomic and phenotypic effects of subinhibitory antibiotics concentrations on L. monocytogenes, but ampicillin, vancomycin and gentamicin reduce the expression of hly, encoding listeriolysin O, in L. monocytogenes (Nichterlein et al. 1996, 1997). It is not known whether the changes in gene expression affect virulence in more complex virulence model.

Listeria monocytogenes is capable of sensing antibiotics through a range of master regulators including the two-component systems, CesRK, LisRK and LiaSR that respond to cell wall inhibitors (Cotter et al. 2002; Kallipolitis et al. 2003; Gottschalk et al. 2008; Collins et al. 2012; Nielsen et al. 2012). The expression of orf2420 and lmo0443 genes that are CesR-dependent is induced by ampicillin, vancomycin and cefuroxime (Kallipolitis et al. 2003; Gottschalk et al. 2008; Nielsen et al. 2012) and the LisR-dependent lmo2210 expression is increased by cefuroxime (Gottschalk et al. 2008; Nielsen et al. 2012). Furthermore, the alternative sigma factor, σB, is induced upon antibiotic exposure as are a number of σB-dependent genes (Begley et al. 2006; Shin et al. 2010; Palmer et al. 2011).

Listeria monocytogenes is in its environmental niche likely to be constantly exposed to low levels of antibiotics and because it does harbour antibiotic-sensing systems, it likely modulates the expression of both virulence and stress genes upon exposure to subinhibitory antibiotic concentrations as observed in other pathogens. Understanding how L. monocytogenes may respond to low concentrations of antibiotics is important, because successful treatment of listeriosis, especially in immunocompromised humans, is dependent on antibiotic treatment, and maintaining lethal antibiotic concentrations during the entire treatment is not always possible (Hof et al. 1997). We have previously demonstrated that subinhibitory concentrations of biocides alter L. monocytogenes virulence gene expression (Kastbjerg et al. 2010), and the purpose of the present study was to determine whether subinhibitory concentrations of different classes of antibiotics affect gene expression and physiology of L. monocytogenes. Promoter fusions of 11 different genes were included and represented expression of PrfA, σB, CesR and LisR-dependent genes, that is, PrfA-regulated virulence genes (i.e. hly, inlA and prfA), σB-dependent stress response genes (i.e. hfq, opuCA, lmo0911, lmo1526 and uspL-3) and antibiotic-sensing genes controlled by two different antibiotic-sensing two-component systems CesRK (i.e. orf2420 and lmo0443) and LisRK (lmo2210).

Materials and methods

Bacterial strains and growth conditions

Plasmids carrying lacZ and bgaB fusions to promoters of three virulence genes, five stress response genes and three known antibiotic-sensing genes were transformed into L. monocytogenes EGD wild-type as well as ΔprfA, ΔsigB and ΔcesR mutants (Table 1 and Table S1). Bacterial stock cultures were stored at −80°C and inoculated on brain–heart infusion (BHI; Oxoid CM 1135) agar and grown at 37°C overnight. Antibiotics to maintain plasmids were used in the following concentrations: 50 μg ml−1 kanamycin and 5 μg ml−1 erythromycin. A standard inoculum culture for the agar-based screening assay was obtained by inoculating one colony in 5 ml BHI broth and incubating aerobically at 37°C overnight with shaking (250 rev min−1). A second subculture was prepared by inoculating 5 μl overnight culture into 5 ml BHI broth and incubating for 18 h at the same conditions.

Table 1. Listeria monocytogenes strains and plasmids used in this study
Strains/plasmidsCharacteristicSource or reference
  1. a

    lmo2673 have recently been renamed to uspL-3 (Gomes et al. 2011).

Strains
EGD Chakraborty et al. (1992)
ΔprfA Chakraborty et al. (1992)
ΔsigB Brøndsted et al. (2003)
EGDe (Hain) Hain et al. (2008)
ΔsigB (Hain) Hain et al. (2008)
ΔcesR Williams et al. (2005)
Plasmids
pTCV-lacCloning vector for transcriptional fusions to lacZ; aphA-3, ermBPoyart and Trieu-Cuot (1997)
PprfA-lacZpTCV-lac derivative containing the 5′-UTR prfA region from EGD; aphA-3, ermBLarsen et al. (2006)
PinlA-lacZpTCV-lac derivative containing the 5′-UTR inlA region from EGD; aphA-3, ermBLarsen et al. (2006)
Phly-lacZpTCV-lac derivative containing the 5′-UTR hly region from EGD; aphA-3, ermBLarsen et al. (2006)
Porf2420-lacZpTCV-lac derivative containing the 5′-UTR orf2420 region from LO28; aphA-3, ermBGottschalk et al. (2008)
Plmo0443-lacZpTCV-lac derivative containing the 5′-UTR lmo0443 region from LO28; aphA-3, ermBGottschalk et al. (2008)
Plmo2210-lacZpTCV-lac derivative containing the 5′-UTR lmo2210 region from LO28; aphA-3, ermBGottschalk et al. (2008)
Phfq-lacZpTCV-lac derivative containing the 5′-UTR hfq region from EGD; aphA-3, ermB 
pSOG30222Cloning vector for transcriptional fusions to bgaB; ermCHain et al. (2008)
PopuC-bgaBpSOG30222 derivative containing the 5′-UTR opuC region from EGD; ermCHain et al. (2008)
Plmo0911-bgaBpSOG30222 derivative containing the 5′-UTR lmo0911 region from EGD; ermCHain et al. (2008)
Plmo1526-bgaBpSOG30222 derivative containing the 5′-UTR lmo1526 region from EGD; ermCHain et al. (2008)
PuspL-3-bgaBapSOG30222 derivative containing the 5′-UTR uspL-3 region from EGD; ermCHain et al. (2008)

General molecular techniques

Plasmids were isolated from L. monocytogenes using QIAprep Spin Miniprep (Qiagen 27104), and transformation of plasmids into L. monocytogenes EGD was performed according to Schäferkordt et al. (1998).

Influence of antibiotics on gene expression in anagar-based screening assay

The effects of subinhibitory concentrations of antibiotics on gene expression were tested as described by Kastbjerg et al. (2010) with few modifications. A 1000-fold dilution of a second subculture was mixed with melted BHI agar containing 150 μg ml−1 X-gal (dissolved in dimethylformamide; Roche 10703729001) but no antibiotic was added to maintain the plasmids in the agar-based screening assay. Antibiotic discs (Oxoid) were placed on the agar plates after drying (45 min followed by 15 min in laminar air flow). Antibiotics tested included cell wall inhibitors (ampicillin, amoxicillin, penicillin, cefuroxime, vancomycin, teicoplanin, fosfomycin and bacitracin), protein synthesis inhibitors (gentamicin, streptomycin, tetracycline and erythromycin) and DNA/RNA synthesis inhibitors (ciprofloxacin, nalidixic acid, co-trimoxazole, that is, sulfamethoxazole and trimethoprim, and rifampicin). Five μl 30% H2O2 and 5 μl sterile H2O were spotted on 6-mm AA discs (Whatman) as positive and negative controls, respectively. Plates were incubated at 37°C for 48 h, and each promoter fusion was tested in a minimum of two independent biological replicates. Zones of clearing appeared around most antibiotics indicative of bacteriostatic or bacteriocidal effect. In the circle just outside the clearing zone, L. monocytogenes would grow and be exposed to subinhibitory concentration of the antibiotic.

Determination of subinhibitory antibiotic concentrations in liquid broth

Expression of several virulence and stress genes was affected by antibiotics, and to determine whether this altered bacterial physiology, we determined subinhibitory concentrations of ampicillin (causing up-regulation) and tetracycline (causing down-regulation) in liquid broth. For ampicillin, a second subculture was diluted 100-fold in 60 ml BHI broth and grown at 37°C with 250 rev min−1 to optical density at 600 nm (OD600) of 0·4. The culture was split in three by diluting 15 ml culture with 45 ml prewarmed BHI broth. At OD600 = 0·4, ampicillin (0·01–10 μg ml−1) or sterile MilliQ was added and 0·05 μg ml−1 selected as the subinhibitory concentration causing a slight reduction in maximum optical density. A bacteriostatic effect was obtained using 0·5 μg ml−1. Tetracycline was tested in 10 ml BHI broth to which twofold dilution of tetracycline (0·0039–2 μg ml−1) was added at OD600 = 0·4. The subinhibitory tetracycline concentration was 0·065 μg ml−1 causing a minor inhibition of the maximum optical density.

Phenotypic effects of exposure to subinhibitory concentrations of antibiotics: Invasion into Caco-2 cells

Virulence gene expression, including inlA, was changed by antibiotics in the agar-based screening assay, and we therefore determined whether L. monocytogenes exposed to subinhibitory concentrations of antibiotics were affected in Caco-2 cell invasion (Jensen et al. 2008). In brief, 2 × 105 Caco-2 cells ml−1 were grown in a 24-well tissue culture plate for 24 h to reach a monolayer. L. monocytogenes were grown as described above. At OD600 = 0·4, 5 ml bacterial culture was transferred to tubes and exposed to subinhibitory concentration of ampicillin, tetracycline or sterile MilliQ as control and incubated at 37°C with 250 rev min−1 for 3 h. The bacterial cultures were adjusted to 1 × 107 CFU ml−1 by dilution in modified Eagle's medium (MEM; Invitrogen, 41090028) supplemented with 20% foetal bovine serum (Lonza, DE14-830) and 0·1 mmol l−1 nonessential amino acids (Lonza, BE13-114E). After addition of 1 ml bacterial culture to each well, the plates were incubated 1 h at 37°C allowing the bacteria to invade the Caco-2 cells. After incubation and wash of the cells in saline water, 1 ml of supplemented MEM with 50 μg ml−1 gentamicin was added to each well and the plate was incubated for 1 h at 37°C. Following incubation, cells were washed again in saline water and lysed by addition of 0·1% TritonX-100. Bacterial counts were determined by plating 10 dilutions on BHI agar plates. The experiment was performed with three independent biological replicates with two technical replicates.

Phenotypic effects of exposure to subinhibitory concentrations of antibiotics: stress tolerance

Antibiotics changed expression of stress-related genes in the agar-based screening assay, and we determined whether these changes affected stress response in L. monocytogenes. Listeria monocytogenes were grown to OD600 = 0·4 and exposed to subinhibitory concentration of antibiotics for 3 h as described above. Acid tolerance of bacteria grown for 3 h with antibiotics was tested by exposing them to low pH: 800 μl of antibiotic-stressed or control cells were inoculated in 7·2 ml normal BHI or BHI adjusted to pH 3·3 with 2 mol l−1 HCl, which caused a final pH of 3·6. Samples for bacterial counts, pH and OD600 measurements were taken at time zero and after 1.5 h at 37°C with 250 rev min−1. Oxidative stress tolerance of L. monocytogenes grown for 3 h with antibiotics was tested by exposure to H2O2: 600 μl of antibiotic-stressed or control cells were inoculated to 5·4 ml BHI or BHI with 33·33 mmol l−1 H2O2, which would generate a final concentration of 30 mmol l−1 H2O2. Samples for bacterial counts and OD600 measurements were taken at time zero and after 1.5 h.

Quantitative measures of gene expression in liquid culture using β-galactosidase assay

We quantified the effects on gene expression of ampicillin and tetracycline using a virulence gene (PinlA-lacZ), a stress response gene (Phfq-lacZ) and an antibiotic-sensing gene (Porf2420-lacZ) by β-galactosidase assay. Bacterial cultures were grown to OD600 = 0·4 and exposed to subinhibitory concentration of antibiotics as described above. Samples for β-galactosidase assay were harvested at time zero and 3 h after addition of tetracycline or at time zero, 1 and 3 h for ampicillin by mixing 1 ml culture with 15 μl 10 mg ml−1 chloramphenicol, before centrifugation at 10 000 g for 2 min. Samples were stored at −20°C. β-galactosidase assay was performed and Millers units calculated as described by Kallipolitis et al. (2003). The experiment was performed with three (tetracycline) or two (ampicillin) independent biological replicates.

Verification in liquid culture using quantitative real-time PCR (qRT-PCR)

The expression of the inlA, hfq and orf2420 genes upon ampicillin exposure was also investigated by qRT-PCR. Bacteria were grown as described above, however, samples for qRT-PCR were quenched in RNAprotect Bacteria Reagent (Qiagen 76506) and stored at −80°C prior to extraction of total RNA. Total RNA was extracted using RNeasy Mini Kit (Qiagen 74104) with a mechanically lysis step using a Big Beater for 2 × 5 mins and a DNase treatment step on the column (RNase-Free DNase Set; Qiagen 79254). Total RNA quantity and quality were assessed using Nanodrop and on an RNA 6000 Nano chip (2100 Bioanalyser; Agilent 5067-1511, Santa Clara, CA), respectively. Generation of cDNA and qRT-PCR was performed as described by van der Veen et al. (2007) with few modifications. In brief, 1 μg of total RNA was treated with DNase I (Invitrogen 18068-015) prior to the generation of cDNA using Superscript III reverse transcriptase (Invitrogen 18080-044). RNA samples were controlled for DNA contamination by omitting reverse transcriptase in this step (RT- samples). 2× SYBR Green PCR Master Mix (Applied Biosystems 4309155) was used for qRT-PCR and reactions were run on Mx3000P (Stratagene, La Jolla, CA) on the following programme: 1 cycle at 95°C for 10 min, followed by 40 cycles at 95°C for 30 s and 60°C for 1 min followed by a dissociation curve. Water was included as nontemplate Controls (NTC) and positive control consisted of genomic DNA from L. monocytogenes EGD. Primers (Table S2) were either previously published or designed using Primer3 (http://frodo.wi.mit.edu/primer3/input.htm). For each primer set, the primer efficiency was calculated based on serial dilution of genomic DNA using data from four biological replicates and they ranged from 93·7 to 101·4% (Table S2). The qRT-PCR experiment was performed with three independent biological replicates. Expression levels were normalized using the geometric mean of the rpoB and 16S rRNA housekeeping genes and calculated using the comparative Ct method (2−ΔΔCt; Schmittgen and Livak 2008). Expression levels of each biological replicate were calculated prior to calculating average and standard error of mean using Microsoft Excel.

Statistical analysis

Bacterial counts and OD600 measurements from each biological replicate were log10-transformed prior to statistical analysis using Microsoft Excel. A paired t-test with a significance level of P < 0·05 was used to determine whether antibiotic exposure affected survival upon low acid and oxidative stress. Quantitative expression data calculated as Millers unites from each biological replicate were compared using paired t-test with a significance level of < 0·05. To determine statistical difference in Caco-2 invasion, the invasion efficiency into Caco-2 cells was calculated from the number of invading bacteria relative to the inoculation level for each treatment and was log10-transformed prior statistical analysis.

Results

Effect of subinhibitory concentrations of antibioticson virulence gene expression in Listeria monocytogenes wild-type background

Of the 16 antibiotics tested, 14 were antibacterial (Tables S3 and S4), and consistent with previous reports (Hof et al. 1997; Hof 2003), L. monocytogenes was resistant to nalidixic acid and fosfomycin (Table S4). We found no difference in the diameter of the inhibition zones in L. monocytogenes with or without plasmid expect for erythromycin caused by erythromycin resistance encoding genes on both plasmids (Table 1). Following Subrt et al. (2011) and Kastbjerg et al. (2010), we defined the circle of growth immediately outside the growth/no growth boundary as the zone with subinhibitory antibiotic concentrations and up- or down-regulation of the lacZ and bgaB fusions were seen as either a blue or white circle (Fig. 1 and Table S3). Listeria monocytogenes with and without empty control vectors had a very weak blue colour ring when exposed to subinhibitory concentrations of β-lactam and glycopeptides (Table S5). However, only results, where the colour difference was clearly larger when comparing strains with and without promoter fusions, are reported as either up- or down-regulated.

Figure 1.

Listeria monocytogenes with PinlA-lacZ, PopuCA-bgaB and Porf2420-lacZ fusions in response to ampicillin or tetracycline. Bacterial cells were diluted 1000-fold prior to being cast in brain–heart infusion agar. The results presented are representative of three independent experiments.

Six of eight of antibiotics targeting the bacterial cell wall induced the expression of the virulence gene promoters investigated (Table 2), whereas the antibiotics targeting protein or DNA/RNA synthesis did not systematically alter virulence gene expression (Table 2). The virulence gene expression was repressed by tetracycline, whereas no effect was observed for gentamicin (Table 2), although both antibiotics target the 30S protein. Subinhibitory concentrations of co-trimoxazole increased the virulence gene expression, whereas rifampicin decreased the virulence expression. When an antibiotic caused either increased (ampicillin, vancomycin, co-trimoxazole etc.) or decreased (tetracycline) virulence gene expression, it did so consistently for all virulence genes. The results reported here are based on screenings in BHI agar but similar results were obtained in BHI agar with active charcoal that induce virulence gene expression in L. monocytogenes (data not shown).

Table 2. The effect of different classes of antibiotics on the gene expression of either virulence or stress genes in Listeria monocytogenes EGD wild-type strain
GeneRegulatorCell wallProteinDNA/RNAH2O2H2O
AMPAMXPENCXMVANTECFOFBACGENSTRTETERYCIPNALSXTRIF
  1. AMP, ampicillin; AMX, amoxicillin; PEN, penicillin; CXM, cefuroxime; VAN, vancomycin; TEC, teicoplanin; FOF, fosfomycin; BAC, bacitracin; GEN, gentamicin; STR, streptomycin; TET, tetracycline; ERY, erythromycin; CIP, ciprofloxacin; NAL, nalidixic acid; SXT, co-trimoxazole, that is, sulfamethoxazole and trimethoprim; RIF, rifampicin. Symbols: ↑ indicates up-regulation, that is, more intense blue colour in proximity to the growth/no growth boundary, when compared with the background gene expression. (↑) indicates a weak up-regulation. ↓ indicates down-regulation, that is, less intense blue colour or white colour in proximity to the growth/no growth boundary, when compared with the background gene expression. (↓) indicates a weak down-regulation. –, no difference in the expression between the background expression and the area in proximity to the growth/no growth boundary. ND, not determined because of erythromycin resistance cassette on the bgaB plasmid.

  2. a

    Results of the EGDe wild-type obtained from T. Hain. Similar PopuCA-bgaB expression profile was obtained with the EGD wild-type.

PhlyPrfA(↑)(↑)(↑)(↑)
PinlA(↑)
PprfA(↑)(↑)(↑)(↓)
PhfqSigB
PopuCAa(↑)ND(↑)
Plmo0911aND
Plmo1526a(↑)ND(↑)
PuspL-3a(↑)ND(↑)
Plmo0443CesR(↑)(↑)
Porf2420
Plmo2210LisR(↑)

Effect of subinhibitory concentrations of antibiotics on stress response and antibiotic-sensing gene expression in Listeria monocytogenes

Promoter fusions of L. monocytogenes stress response genes, opuCA, lmo0911, lmo1526 and uspL-3, were highly induced by vancomycin and teicoplanin but also by antibiotics belonging to the β-lactam family (Table 2). Similar to the virulence genes, no common expression pattern of the stress response genes was observed in response to antibiotics of the protein and DNA/RNA synthesis inhibitor classes. Gentamicin and tetracycline consistently induced and repressed stress genes investigated, respectively (Table 2). Ciprofloxacin and co-trimoxazole induced the stress response genes, whereas rifampicin reduced expression of these genes. The expression of Phfq-lacZ was very low, and no changes were observed in the agar-based screening assay (Table 2).

The promoter fusions of the cell wall antibiotic-responding genes, orf2420, lmo0443 and lmo2210, were induced in response to cell wall-acting antibiotics such β-lactams (Table 2). When testing protein and DNA/RNA synthesis inhibitors, tetracycline and co-trimoxazole, respectively, repressed and induced orf2420 and lmo0443, but not lmo2210. Similar to the virulence gene expression, the CesR-dependent Porf2420-lacZ and Plmo0443-lacZ fusions responded in the same manner to all antibiotics tested, whereas the LisR-dependent Plmo2210-lacZ varied in expression pattern.

Effect of subinhibitory concentrations of antibiotics on virulence, stress response and antibiotic-sensing gene expression in Listeria monocytogenes in ΔprfA, ΔsigB and ΔcesR mutants

To determine whether the antibiotic-induced changes in L. monocytogenes gene expression were a direct effect via the known regulators of the three groups of genes, PrfA, σB and CesR, we tested selected promoter fusions (PinlA-lacZ, PopuC-bgaB and Porf2420-lacZ) in relevant regulator mutants. As expected, the target gene expression was reduced in the regulator mutants as compared with wild-type background although a few exceptions were observed (Table S6). Hence, the target gene expression in response to subinhibitory concentrations of antibiotics appeared to be both dependent and independent of the three known regulators of the promoter fusions investigated.

Effect of subinhibitory antibiotic concentration on in vitro virulence in Caco-2 cells

We investigated invasion of antibiotic-exposed L. monocytogenes in Caco-2 cells including ampicillin that induced expression of the virulence genes including inlA and tetracycline that reduced virulence expression. A ΔprfA mutant was included as control and compared with the wild-type, and this mutant invaded Caco-2 cells in significantly lower numbers (= 0·028; Table 3). Exposure to ampicillin or tetracycline prior to cell invasion did not significantly change L. monocytogenes invasion of Caco-2 cells (P ≥ 0·27; Table 3).

Table 3. Comparison of the invasiveness of Listeria monocytogenes when pre-exposed to subinhibitory antibiotic concentrations
StrainAntibiotic treatmentaPercentage of entry into Caco-2 cellsb
  1. a

    Bacterial cultures were exposed to antibiotics for 3 h as described for the acid tolerance assay.

  2. b

    Per cent entry was calculated from the number of bacteria recovered after 1 h treatment of the Caco-2 cells with 50 μg ml−1 gentamicin with respect to the total number of inoculated bacteria. Data are average of three biological replicates showing standard deviation.

  3. c

    Statistical significance is denoted by (< 0·05).

EGD 0·16 ± 0·05
EGD ΔprfA 0·05 ± 0·02c
EGD0·05 μg ml−1 ampicillin0·14 ± 0·09
EGD0·065 μg ml−1 Tetracycline0·11 ± 0·05

Effect of ampicillin and tetracycline exposure on subsequent stress survival

As ampicillin increased and tetracycline reduced expression of the σB-dependent genes in the agar-based screening assay, we tested whether ampicillin-exposed L. monocytogenes was more stress tolerant than control cells and vice versa with tetracycline-exposed L. monocytogenes using acid and oxidative stress that are both σB-dependent (Ferreira et al. 2001; Oliver et al. 2010). Exposure to subinhibitory concentration of tetracycline significantly (= 0·015) reduced L. monocytogenes survival at low pH as compared with nonexposed cells (Fig. 2a). However, the survival of ampicillin-stressed L. monocytogenes was also significantly reduced as compared with the control (= 0·007; Fig. 2a) and was as sensitive as tetracycline-stressed cells (= 0·77), indicating that antibiotic-exposed cells, independently of stress gene expression, were more stress sensitive. Similarly, ampicillin- and tetracycline-exposed bacterial cells had a lower survival when exposed to 30 mmol l−1 H2O2; however, this trend was not statistically significant from nonstressed control cultures (= 0·24 and = 0·20, respectively; Fig. 2b).

Figure 2.

Survival of antibiotic-stressed Listeria monocytogenes during (a) pH 7 (grey) or pH 3·6 (dark grey) or (b) 0 (grey) or 30 (dark grey) mmol l−1 H2O2. Data are average of three biological replicates showing standard error of mean.

Expression of virulence and stress genes in a liquid culture measured by β-galactosidase assay

As the changes in promoter gene expression observed in the agar-based screening assay did not translate to the expected phenotypic change, we quantified the effects on gene expression in liquid cultures to verify the agar-based screening assay. The expression of PinlA-lacZ and Phfq-lacZ was significantly decreased with 0·0625 μg ml−1 tetracycline compared with nontreated control (P = 0·015; Fig. 3a and P = 0·004; data not shown). Also Porf2420-lacZ was repressed by tetracycline; however, this was not significant (≥ 0·09; Fig. 3b).

Figure 3.

Expression of selected lacZ fusions in Listeria monocytogenes in response to tetracycline (a and b) and ampicillin (c and d). (a and b) Twofold concentrations of tetracycline at 3 h exposure and gene expression are shown as bars. The secondary axis shows the optical density at 600 nm of the culture at 3 h shown as lines. Data are average of three biological replicates with standard error of mean. (a) Expression of PinlA-lacZ in wild-type (dark grey), ΔprfA mutant (bright grey) and ΔsigB (middle grey). (b) Expression of Porf2420-lacZ in wild-type. (c and d) Expression of selected lacZ fusions in Listeria monocytogenes in response to two concentrations of ampicillin exposed for either 1 or 3 h. Data are average of two biological replicates with standard error of mean. (c) Expression of PinlA-lacZ in wild-type. (d) Expression of Porf2420-lacZ in wild-type.

Expression of the antibiotic-sensing gene, Porf2420-lacZ, was significantly induced by both concentrations of ampicillin and at both time points (= 0·004 at 3 h with 0·05 μg ml−1 ampicillin; Fig. 3d) but the timing of the orf2420 induction at the two different ampicillin concentrations was different (Fig. 3d). Ampicillin decreased the PinlA-lacZ expression compared with the control; however, the reduction was not significant (P ≥ 0·17; Fig. 3c). Consistent with the agar-based screening system, the expression level of Phfq-lacZ was very low and no effect of ampicillin was observed (data not shown), whereas other stress and virulence genes including PinlA-lacZ were slightly induced by ampicillin in the agar-based screening assay.

To investigate whether the significant tetracycline effect on gene expression was mediated via known regulators, we investigated the PinlA-lacZ expression in the wild-type as well as ΔsigB and ΔprfA mutants. In the non-tetracycline-treated control, the PinlA-lacZ expression was reduced in both the ΔsigB and ΔprfA mutants as compared with the wild-type EGD. Similar to the wild-type, the PinlA-lacZ expression was decreased by increasing tetracycline concentration for both the ΔsigB and ΔprfA mutants (Fig. 3a).

Expression of virulence and stress genes in liquid culture measured by qRT-PCR

As the PinlA-lacZ expression upon ampicillin exposure varied between the agar-based screening assay and the liquid culture, we used qRT-PCR to verify the β-galactosidase assay. In line with the agar-based screening, the β-galactosidase assay and published data (Kallipolitis et al. 2003; Gottschalk et al. 2008), the orf2420 expression was increased in ampicillin-stressed cultures as compared with the control (Fig. 4a). Furthermore, orf2420 inductions observed by β-galactosidase assay and qRT-PCR were similar. Also the expression of hfq was consistently found not to be affected by ampicillin by all three methods, that is, agar-based screening, β-galactosidase assay and qRT-PCR (data not shown). Measured by qRT-PCR, the inlA expression was slightly decreased compared with the control when exposed to ampicillin (Fig. 4b) but reductions were lower than the reductions found by the β-galactosidase assay. In summary, the expression data obtained by β-galactosidase assay and qRT-PCR were consistent but as expected with slight differences in the fold changes observed.

Figure 4.

Relative expression of orf2420 and inlA in response to ampicillin comparing expression in treated with nontreated control measured by quantitative RT-PCR at 1 or 3 h. Data are average of three biological replicates showing standard error of mean. (a) Relative expression of orf2420 in wild-type. (b) Relative expression of inlA in wild-type.

Discussion

We have shown that subinhibitory concentration of several antibiotics, including antibiotics used for treatment of listeriosis, affects expression of virulence and stress response genes in L. monocytogenes. This is in agreement with other studies investigating virulence gene expression in L. monocytogenes and other pathogenic bacteria (Nichterlein et al. 1996, 1997; Davies et al. 2006; Kastbjerg et al. 2010; Nielsen et al. 2010; Subrt et al. 2011). Based on the work with the regulator mutants, we show that antibiotics are able to affect gene expression both dependently and independently of the known regulatory cascade consistent with previously published work (Kastbjerg et al. 2010; Kindrachuk et al. 2011; Subrt et al. 2011).

We used an agar-based screening approach to investigate a high number of interactions but found that β-lactam and glycopeptides caused weak inductions when investigating L. monocytogenes with empty control vector encoding lacZ or bgaB. This could either be caused by altered growth pattern in the zone of subinhibitory antibiotic concentrations or by altered expression of X-gal cleaving enzymes encoded by L. monocytogenes. X-gal is an analogue of lactose, which Listeria can ferment (Pine et al. 1989). Similar agar system using either lacZ or lux fusions has been used previously with B. subtilis (Cao et al. 2002), Staph. aureus (Nielsen et al. 2010; Subrt et al. 2011) and L. monocytogenes (Kastbjerg et al. 2010). Results from agar-based assays have been verified using Northern blotting from cells grown in liquid culture (Kastbjerg et al. 2010), or using qRT-PCR of agar-embedded Staph. aureus (Subrt et al. 2011). Consistent with these studies, the altered gene expression we observed in the agar-based screening was confirmed in five of six gene-antibiotic combinations in liquid systems using either lacZ fusions or qRT-PCR. Further, the expression of the CesR, LisR and σB-dependent fusions in our study is consistent with previous studies performed in liquid cultures (Kallipolitis et al. 2003; Gottschalk et al. 2008; Shin et al. 2010), showing that vancomycin induces expression of σB-dependent genes (Shin et al. 2010) and several cell wall inhibitors induce expression of CesR-dependent genes, orf2420 and lmo0443 (Kallipolitis et al. 2003; Gottschalk et al. 2008).

inlA expression following exposure to ampicillin was not verified in liquid culture. This could be caused by the difference between agar-embedded and planktonic cells as is observed in Salmonella, where heat stress induces virulence gene expression during immobilized growth but reduces expression during planktonic growth (Nielsen et al. in press). In agreement with our liquid culture results, Nichterlein and co-workers found that the expression of Phly-lacZ was decreased in response to subinhibitory concentrations of several antibiotics including ampicillin (Nichterlein et al. 1996, 1997). Also, they demonstrated that this translated into a downstream phenotypic response as the haemolytic titre was decreased by ampicillin and vancomycin (Nichterlein et al. 1996). However, in our study, the observed changes in virulence expression did not cause ampicillin- or tetracycline-exposed L. monocytogenes to be more or less invasive in a Caco-2 assay. Nichterlein et al. (1996, 1997) found that some antibiotics caused a reduced haemolytic titre but the antibiotic-treated cells were not tested in a more complex virulence model in their study. Whilst the liquid cultures using quantitative methods, that is, β-galactosidase assay or qRT-PCR give the most accurate measure of gene expression, one could speculate that the agar-embedded L. monocytogenes are in a physiological state closer to the infectious lifestyle. In the human body, L. monocytogenes is more likely to be associated with surfaces hence this mode of growth potentially provide a response mimicking better the infectious stage.

Other studies have, as ours, found that exposure to subinhibitory antibiotic concentrations affects the bacterial physiology such as acid tolerance. Walkmycin C reduced the acid tolerance in Streptococcus mutans (Eguchi et al. 2011), and E. coli O157:H7 pre-exposed to trimethoprim, ampicillin or ofloxacin was more sensitive to simulated gastric fluid compared with the non-antibiotic-exposed control (Azizoglu and Drake 2007). However, this pattern of reduced acid tolerance is not universal as pre-exposure to kanamycin and erythromycin does not alter acid resistance in Streptococcus mutants (Eguchi et al. 2011). Consistent with Azizoglu and Drake (2007), we found that ampicillin reduced the acid tolerance although we expected increased acid tolerance based on the stress gene expression. We suspect this discrepancy is caused by increased activity of the two-component system, LisRK, indicated by ampicillin-induced expression of Plmo2210 (Table 2). Besides being involved in cell response to β-lactam and nisin (Cotter et al. 2002; Gottschalk et al. 2008), LisRK affects acid tolerance, and a ΔlisK mutant is more resistant to pH 3·5 than the wild-type (Cotter et al. 1999). Thus, the increased LisRK activity would reduce the acid tolerance as observed in ampicillin-treated cells. In contrast, we expected the down-regulation of the σB-dependent genes by tetracycline (Table 2) to render these cells more stress sensitive. Our data indicate that if L. monocytogenes encounter subinhibitory concentration of ampicillin or tetracycline in the environment such as silage or the food productions, this pre-exposure to these antibiotics would not have a protective effect against low pH from silage, gastric acid or preservatives.

All of the genes selected in the present study are under the control of master regulators (PrfA, σB, CesRK and LisRK) and some antibiotics, for example tetracycline, affected gene expression of a specific gene both dependently and independently of the regulator (Table S6 and Fig. 3a). Such non-regulator-mediated antibiotic effects on gene expression have been found in L. monocytogenes and in other bacterial species such as Ps. aeruginosa and Staph. aureus (Kastbjerg et al. 2010; Kindrachuk et al. 2011; Subrt et al. 2011). The cause of these unspecific effects on expression is unknown; however, we speculate that antibiotic potentially could stabilize mRNA and affect translation efficiency as found for erythromycin and tetracycline (Bechhofer and Zen 1989; Flache et al. 1992).

In our agar-based screening, most of the antibiotics that affected gene expression did so in a consistent manner, that is, expression of all genes was induced (e.g. ampicillin, vancomycin) or reduced (e.g. tetracycline). These observations could indicate that each antibiotic affects the general transcription and/or translation efficiency. Alternatively, we speculate that the antibiotic molecule may not be acting as a specific signal molecule causing alterations in expression of specific genes but that it affects the physiology of the bacterial cell, for example cell wall and/or membrane, and thereby affect gene expression as a secondary response. Linares et al. (2006) suggested that it is the general burden to bacterial physiology that affects the gene expression upon antibiotic exposure rather than a specific antibiotic-gene effect. This is supported by microarray study of the B. subtilis global response to a range of antibiotics where the primary effect of antibiotics, which affected the target area, was only observed after 10 min exposure, but was not detectable at 40 min, where a more global secondary response was observed (Freiberg et al. 2005). We are currently investigating the global response using transcriptomic analysis upon exposure to subinhibitory antibiotic concentrations.

First choice of antibiotic for listeriosis-treatment is a β-lactam in combination with an aminoglycoside, most often ampicillin and gentamicin (Hof 2003), and these antibiotics had no or minor effect on virulence. Second choice of antibiotic for treatment (vancomycin, erythromycin or co-trimoxazole) (Wagner and McLauchlin 2008) all induced virulence expression in the agar-based screening assay. We can only speculate why L. monocytogenes alter gene expression in response to antibiotics. Nevertheless, several of the investigated antibiotics are produced by soil and saprophytic micro-organisms (Hendlin et al. 1969; Shoji et al. 1986; Chater 2006), which L. monocytogenes are likely to coexist within the environment. Protozoan can be an environmental vector for foodborne pathogens including L. monocytogenes (Hilbi et al. 2007; Zhou et al. 2007; Anacarso et al. 2012), and virulence expression is vital for survival of L. monocytogenes in the protozoan, Acanthamoeba castellanii (Zhou et al. 2007). Hence, increased virulence expression in response to environmental antibiotics could increase internalization and survival of L. monocytogenes within the protozoan. Other escape strategies from antibiotics in the environment could be increased motility or biofilm formation as observed in other bacteria (Linares et al. 2006; Yim et al. 2006; Subrt et al. 2011).

In conclusion, we have shown that subinhibitory concentrations of antibiotic can alter gene expression and affect the cell physiology of L. monocytogenes. Subinhibitory concentration of some antibiotics reduced the acid tolerance independently of changes in stress gene expression. However, this was not a universal stress response because oxidative stress tolerance was not altered and the invasion level of ampicillin or tetracycline-exposed L. monocytogenes in Caco-2 cells was not affected as compared with non-antibiotic-exposed controls. Additionally, our data indicate that antibiotics do not always mediate the expression through the known regulatory cascade. The fact that some of the antibiotics of clinical relevance enhanced expression of virulence genes calls for further studies owing to the obvious potential clinical side effects.

Acknowledgements

We would like to thank Yin Ng for expert technical assistance as well as Louise Feld and other members of the Gram Lab for good discussions. The work was supported by the Danish Council for Independent Research, Technology and Production Sciences (FTP) grant 09-066098 (274-08-0531).

Ancillary