Cross-resistance to antibiotics of Escherichia coli adapted to benzalkonium chloride or exposed to stress-inducers


Solveig Langsrud, MATFORSK, Norwegian Food Research Institute, Oslov 1, N-1430 Ås, Norway (e-mail:


Aims:  To study the effects of adaptation and stress on the resistance to benzalkonium chloride (BC) and cross-resistance to antibiotics in Escherichia coli.

Methods and Results:  Precultivation of E. coli ATCC 11775 and E. coli DSM 682 in the presence of subinhibitory concentrations of BC or stress inducers (salicylate, chenodeoxycholate and methyl viologen) resulted in higher minimum inhibitory concentration (MIC) of BC and chloramphenicol (CHL). Adaptation to growth in sixfold of the initial MIC of BC resulted in stable BC resistance and enhanced tolerance to several antibiotics and ethidium bromide (EtBr). The MIC of CHL increased more than 10-fold for both strains. Enhanced efflux of EtBr in adapted E. coli ATCC 11775 indicated that the observed resistance was due to efflux. Changes in outer membrane protein profiles were detected in the BC-adapted cells. There were no indications of lower membrane permeability to BC.

Conclusions:  Induction of stress response or gradual adaptation to BC or CHL results in acquired cross-tolerance between BC and antibiotics in E. coli. Enhanced efflux was one of the observed differences in adapted cells.

Significance and Impact of the Study:  Provided not taking due precautions, extensive use of disinfectants could lead to emergence of antibiotic-resistant isolates.


Disinfectants based on benzalkonium chloride (BC) and other quaternary ammonium compounds (QACs) have been used for decades to improve hygiene in hospitals, in food-processing environments, in the veterinary field and other areas. Bacteria employ a variety of strategies to avoid the lethal and inhibitory effects of antimicrobial agents. Resistance to QACs has been described (for an overview, see McDonnell and Russell 1999). Resistance to disinfectants may be intrinsic or acquired. In general, Gram-negative bacteria are intrinsically more resistant to disinfectants such as QACs than Gram-positive, mainly because of their relatively impermeable outer membrane (McDonnell and Russell 1999). Acquisition of resistance in Escherichia coli and Pseudomonas aeruginosa has been related mainly to changes in the membrane composition (Sakagami et al. 1989; Ishikawa et al. 2002). Ishikawa et al. (2002) have reported that spontaneous acquisition of resistance to QACs in E. coli may be due to alteration of the lipopolysaccharide (LPS) composition and reduction in OmpF. Active efflux of cationic surfactants as an alternative mode of acquisition of resistance has been demonstrated in staphylococci (Littlejohn et al. 1992; Heir et al. 1995, 1998) and Listeria monocytogenes (Aase et al. 2000).

Acquisition of resistance to antibiotics is often followed by alterations in the membrane and active efflux. Enhanced resistance to QACs in bacteria with acquired resistance to antibiotics has been demonstrated (Váczi et al. 1957; Edgar and Bibi 1997). In recent years there has been a growing concern that the use of antibacterial agents in the household, food industry and in hospitals may contribute to the emergence of bacteria resistant to antibiotics (Moken et al. 1997; Russell 2000; Schweizer 2001). We have previously demonstrated a genetic linkage between genes encoding resistance to QACs and β-lactams in food, clinical and veterinary isolates of staphylococci (Sidhu et al. 2001; Sidhu et al. 2002). The phenomenon of cross-resistance between antibiotics and disinfectants has also been demonstrated in E. coli (Ishikawa et al. 2002). A membrane protein (MdfA) belonging to the major facilitator superfamily of transport proteins, renders E. coli more resistant to several antibiotics as well as BC when expressed on a multicopy plasmid (Edgar and Bibi 1997, 1999).

In its natural habitats E. coli has to tolerate high levels of bile salts or survive noxious agents. Introduction of salicylic acid (SAL), bile salts, chloramphenicol (CLH) and other compounds triggers global regulatory stress systems such as mar RAB and sox RS (Grkovic et al. 2002). These systems promote transcriptions of genes encoding proteins of diverse functions, among them efflux pumps and porins (OmpF) (Barbosa and Levy 2000; Pomposiello et al. 2001). Exposure to SAL results in resistance to several hydrophobic antibiotics, oxidative agents and organic solvents (Cohen et al. 1993; Martin et al. 1996; Asako et al. 1997; Pomposiello et al. 2001). Similar phenotypic changes can be obtained by exposure to methyl viologen (MV), which is involved in the superoxide-stress response (Pomposiello et al. 2001). It is not known whether the stress response leads to higher resistance to disinfectants such as BC and whether bacteria stressed by low concentrations of BC acquire higher resistance to antibiotics.

The aim of this work was to study the effects of adaptation and stress on the resistance to BC and cross-resistance to antibiotics in E. coli and to see whether changes of permeability or efflux were among the possible resistance mechanisms.

Materials and methods


BC was provided by Norwegian Medical Depot (Oslo, Norway). Ampicillin (AMP), Penicillin G (PEN), norfloxacin (NOR), nalidixic acid (NAL), kanamycin (KAN), gentamicin (GEN), CHL, tetracycline (TET), erythromycin (ERY), proflavine (PRO), ethidium bromide (EB), N-phenylnaphthylamine (NPN), SAL, MV and CDC were from Sigma (St Louis, MO, USA).


The bacteria used in this study were E. coli ATCC 11775 and E. coli DSM 682. Escherichia coli ATCC 11775 is the type strain and E. coli DSM 682 is a strain often used for antibiotic testing. The species identification was confirmed by 16S rDNA analysis. The adapted cultures were obtained by daily serial transfers in Mueller Hinton (MH; Oxoid, Hampshire, UK) broth containing stepwise higher concentrations of BC or CHL, using increments of 5 μg ml−1, to a concentration of 150 or 100 μg ml−1, respectively. No precipitation of BC in the medium could be observed at this concentration. The strains were stored in Microbank vials (PRO-LAB Diagnostics, Ontario, Canada) at −80°C as described by the manufacturer. Every month, one bead was spread on MH agar, incubated overnight at 37°C and kept at 4°C. The cell cultures were prepared by inoculation of one colony in 5 ml MH broth (containing 150 μg ml−1 BC for cultures adapted to BC) and incubation at 37°C with shaking to early stationary phase. The stability of the acquired resistance was determined by three successive transfers (10 μl) in 5 ml MH broth without BC and incubation at 37°C.

Induction of stress

Overnight cultures in MH broth were inoculated (50 μl) in 5 ml MH broth containing 5 mm SAL, 0·1 mm MV, 2·4 mm CDC, 5 μg ml−1 BC or 20 μg ml−1 BC and grown to late logarithmic phase (4–5 h) at 30°C with shaking.

Bactericidal test

The cultures were grown to late exponential phase in MH broth. The bacterial test suspensions were prepared by diluting the cultures 10-fold in peptone water (saline with 0·1% Bacto-peptone; Oxoid), to a cell concentration of ca 108 cells ml−1. The test was performed by a modified version of the Council of Europe suspension test (Anon. 1987) for testing antimicrobial activity of disinfectants. The microbial test suspension (1 ml) was added to 4 ml distilled water. After 2 min, 5 ml BC prepared in distilled water was added to twice the final concentration. The disinfectant was neutralized by 10-fold dilution in Letheen broth (Difco, Detroit, MI, USA) after 5-min exposure at 20°C. A control sample was treated in the same way, but with distilled water instead of disinfectant. The number of CFU was determined by serial dilution in peptone water and spreading on MH agar. The plates were incubated at 37°C for 1 and 3 days before determination of CFU. The experiment was carried out three times on different days and with freshly prepared solutions. The reduction in viability was calculated as RV = Nc/Nt where Nc is the number of CFU in the control sample and Nt is the number of CFU after exposure to disinfectant. The differences between means were tested using the two sample t-test in MINITAB (release 11·21; Minitab Inc., State College, PA, USA). The effectiveness and toxicity of the neutralizing agent has been tested in an earlier investigation (Langsrud and Sundheim 1998).

MIC of antibiotics and BC

The minimum inhibitory concentration (MIC) of BC was determined in volumes of 5 ml and an inoculum of 105–106 cells ml−1. Growth was recorded by visual inspection after incubation at 37°C with shaking. The MIC of antibiotics and dyes in MH broth was determined using microtitre plates and an inoculum of 105–106 cells ml−1. Growth was recorded after 48 h at 37°C, using an ELISA reader at 600 nm. The lowest concentration preventing growth was taken to be the MIC. All MIC determinations were repeated twice (three times if the results from the first two tests differed).

Efflux and permeability studies

Ethidium bromide (EtBr) transport was measured as described by Turner et al. (1997) unless otherwise specified. Briefly, 1 ml of an overnight culture in MH medium was inoculated in 100 ml MH broth. The cultures were incubated at 37°C to an O.D.600 of 0·4. The cells were harvested by centrifugation (3000 g, 10 min) and washed in 20 mm HEPES (pH 7). Washed cells were resuspended in buffer to O.D.600 of 10 and placed on ice. Transport experiments were performed with a 1-cm path-length fluorescence cuvette containing 1·7 ml of buffer to which 100 μl EB (final concentration 10 μm) and 200 μl cell suspension were added to load the cells with EtBr. Efflux was initiated by adding 100 μl of a 1·0% glucose solution. Fluorescence was measured using an excitation wavelength of 520 nm and emission of 590 nm. The significance of the difference in efflux activity was found by calculation of the regression lines and comparison of the mean slopes using the two sample t-test in MINITAB. Membrane permeability was determined using the method of Bengoechea et al. (1996). Briefly, the bacteria were grown to an O.D.600 of 0·5, washed twice (3000 g, 10 min) with 2 mm HEPES (pH 7·2) with 1 mm KCN and resuspended in the washing buffer to an O.D.600 of 0·5. Fluorescence measurements were performed with a 1-cm path-length fluorescence cuvette which contained 1·8 ml cell suspension to which 20 μl NPN was added to a final concentration of 10 μm. BC (200 μl) was added and accumulation of NPN monitored using an excitation wavelength of 350 nm, emission wavelength of 420 nm and a slit width of 10 nm. The concentrations of BC tested were in the range of 0·5–500 μg ml−1. The strains were grown at 30°C on MH agar containing 1 μg ml−1 EtBr and inspected for fluorescence under u.v.-light after 24 h.

Investigation of outer membrane proteins

Late exponential phase cells grown in MH broth were harvested by centrifugation (5000 g, 10 min) and washed once with 25 mm Tris–HCl buffer (pH 7·4). Cell membranes were isolated by sonication and sarcosyl solubilization (Chart 1995). The inner and outer membrane protein fractions were analysed using SDS-PAGE as described previously (Laemmli and Favre 1973) with 12% (w/v) acrylamide in the running gel. Outer membrane proteins of E. coli ATCC 11775 were blotted on polyvinylidene difluoride membranes according to the method of Matsudaira (1987). Amino acid sequencing was carried out by use of a gas phase 477A automatic microsequencer (Applied Biosystems, Foster City, CA, USA). The sequences were compared with sequences in the GenEMBL databank and to the E. coli chromosomal database.


Adaptation to BC and CLH

Figure 1 shows the rate of adaptation to BC for E. coli ATCC 11775 and DSM 682. The MIC of BC was increased from 25 to 150 μg ml−1 after 24 passages in gradually higher concentrations of the disinfectant. A similar pattern was found in a replicated experiment. At concentrations above 150 μg ml−1 the growth was not consistent and the cells tended to die early in their growth phase. Generally, adapted strains grew slower than control cells, even without BC (data not shown). During the adaptation a few colonies became small pinpoint-like and more colourless, but they regained the original size and colour when restreaked. Fully adapted cells generally produced somewhat smaller colonies than control cells. The appearances of pinpoint colonies were the same in the replicate experiments.

Figure 1.

Adaptation of ATCC 11775 (open symbols) and DSM 682 (filled symbols) to BC. MIC plotted against time

Adaptation to growth in the presence of 150 μg ml−1 BC resulted in higher resistance to the bactericidal activity of 20 μg ml−1 BC as measured by the suspension test (Table 1). The acquired resistance was relatively stable through three reinoculations in medium without the disinfectant. After storing BC-adapted cells at 4°C on MH agar not containing BC they were able to grow in medium containing 150 μg ml−1 BC. After being stored frozen, cells of both strains could grow in 150 μg ml−1 BC after two or three reinoculations in subsequently higher concentrations of the disinfectant. Statistically significant increase in resistance was also obtained after growth in subinhibitory concentrations (5 μg ml−1) of BC, but the resistance was lesser than that for the cultures adapted to 150 μg ml−1 BC (Table 1).

Table 1.  Log10 reduction of Escherichia coli ATCC 11775 and DSM 682 after exposure to 20 μg ml−1 benzalkonium chloride (BC) for 5 min at 20°C
Cell statusPrecultivation
  1. *Mean values of three replicates are given.

  2. Significantly different from the control at †1 and ‡5% level.

AdaptedMH broth with 150 μg ml−1 BC2·4†1·8‡
AdaptedMH broth last three cultivations2·1†3·1‡
ControlMH broth with 5 μg ml−1 BC3·0‡2·4†
ControlMH broth3·54·3

It was possible to grow E. coli ATCC 11775 and DSM 682 in 100 μg ml−1 CHL after five passages in gradually higher concentrations. Adaptation to higher concentrations was possible but resulted in slow growth and low cell yields.

Cross-resistance of adapted strains

Table 2 shows the MIC of antibiotics, proflavine and EtBr for the control and BC-adapted strains. The MICs were 1·5–20-fold higher for the adapted strains than for controls with two exceptions. The highest effect of adaptation to BC on the MIC was obtained with CHL.

Table 2.  MIC of antibiotics and dyes for BC-adapted Escherichia coli ATCC 11775 and DSM 682
E. coli strain*MIC (μg ml−1)†
  1. *C, nonadapted strain; A, strain adapted to grow in the presence of 150 μg ml−1 BC.

  2. †The median value of three experiments is given. The antimicrobial agents were tested in concentration steps of 10 μg ml−1 (AMP), 50 μg ml−1 (PEN), 0·05 μg ml−1 (NOR), 10 μg ml−1 (A) or 2 μg ml−1 (C) (NAL), 4 μg ml−1 (KAN), 2 μg ml−1 (A) or 1 μg ml−1 (C) (GEN), 20 μg ml−1 (A) or 5 μg ml−1 (C) (CHL), 2 μg ml−1 (TET), 20 μg ml−1 (ERY), 10 μg ml−1 (PRO) and 20 μg ml−1 (EB).

ATCC 11775 (C)102500·1588210414050120
ATCC 11775 (A)504000·4301642401618090>220
DSM 682 (C)51000·14102541002020
DSM 682 (A)202000·153010460616040180

Adaptation to CHL resulted in about a threefold increase in the MIC of BC from 25 to 70 and 90 μg ml−1 in CHL-adapted E. coli ATCC 11775 and DSM 682, respectively.

The effect of stress inducers on the resistance to BC, CLH and tetracycline

Escherichia coli ATCC 11775 and DSM 682 differed in their response to exposure to stress-inducers and sub-lethal concentrations of BC (Fig. 2). Pregrowth of nonadapted E. coli ATCC 11775 and DSM 682 in 20 μg ml−1 BC resulted in a two- and five-fold increase in the MIC of CHL, respectively, while the increase in BC tolerance was less than twofold for both strains. A small (30–80%), but consistent, increase in the MIC of TET was found for E. coli ATCC 15775, but not DSM 682 (not shown). Exposure to SAL had a reproducible effect on the tolerance to BC leading to an increase in the MIC from 25 to 35 μg ml−1 for E. coli ATCC 11775 (Fig. 2), but the effect of MV and CDC was smaller. However, exposure of E. coli DSM 682 to MV and CDC resulted generally in increased tolerance to BC and CHL (Fig. 2). With SAL, only increased tolerance to CHL was observed for this strain.

Figure 2.

MIC of BC (filled) and chloramphenicol (open) for nonadapted Escherichia coli after pre-exposure to salicylate (SAL), methyl viologen (MV), chenodeoxy cholate (CDC) and benzalkonium chloride (BC)

Efflux of ethidium bromide and permeability studies

EtBr accumulated in E. coli ATCC 11775 in the absence of glucose. Addition of glucose resulted in active transport of EtBr out of the loaded cells, indicating the presence of efflux pumps (Fig. 3). The efflux was higher in adapted cells (P = 0·04). The efflux of EtBr was partly inhibited in the presence of subinhibitory concentrations of CHL (not shown). Similar experiments with E. coli DSM 682 did not give reproducible results. Both adapted strains showed decreased fluorescence when grown on nutrient agar containing EtBr (not shown).

Figure 3.

Efflux of ethidium in BC-adapted (dashed line) and control (solid line) Escherichia coli ATCC 11775

Addition of BC to a suspension of E. coli resulted in accumulation of NPN. The fluorescence increased with higher concentrations of the disinfectant. Accumulation of NPN in the absence and presence of BC was similar for control and adapted E. coli ATCC 11775 and DSM 682 (Fig. 4). The drop in fluorescence during addition of disinfectant was due to removal of the cuvette.

Figure 4.

Permeability of BC-adapted (dashed line) and control (solid line) Escherichia coli by measuring uptake of NPN after addition of 5, 50 and 500 μg ml−1 BC. The arrows indicate addition of BC

Outer membrane protein profiles of adapted strains

The presence of an outer membrane protein with an apparent molecular mass of 27 kDa was increased in BC-adapted E. coli ATCC 11775 and DSM 682 (Fig. 5). The N-terminal amino acid sequence was S G/S T L T V/L G A G V G V V E Q S Y K G C/Y X A K A Y L I P A V (X, not known). The sequence was determined for two independently adapted cultures of ATCC 11775. No complete match was found when the sequence was compared with sequences at the NCBI ( The sequence showed 52% similarity to the MltA-interacting protein precursor of Salmonella typhimurium. This 26 kDa outer membrane protein encoded by the mipA gene is a scaffolding protein involved in murein synthesis. Increased amounts of a 54-kDa protein was also observed in one of the independently adapted E. coli ATCC 11775. The N-terminal amino acid sequence of this protein was A Q V I N X X X L L T Q N N L N K S Q X X X X X A I. The known part of the sequence was equal to the N-terminal sequence of the flagellin proteins of several Salmonella spp. in the EMBL library. The adaptation did not seem to affect the inner membrane protein profile (not shown).

Figure 5.

SDS-PAGE of outer membrane proteins of control (lanes 2 and 5) and BC-adapted (lanes 3 and 6) Escherichia coli ATCC 11775 (lanes 2 and 3) and DSM 682 (lanes 5 and 6). Molecular mass standards (lanes 1, 4 and 7) are: phosphorylase b, 97·4 kDa; bovine serum albumine, 66·2 kDa; ovalbumine, 45 kDa; anhydrase, 31 kDa; trypsin inhibitor, 21·5 kDa. The arrow indicates the position of the 27 kDa protein


The present investigation does not support the general opinion that a decrease in membrane permeability is the major resistance mechanism in Gram-negative bacteria adapted to QACs. Earlier studies have demonstrated alterations in the membranes of QAC-adapted E. coli and Ps. aeruginosa, leading to the speculation that resistance was linked to decreased permeability (Sakagami et al. 1989; Mechin et al. 1999; Ishikawa et al. 2002). There were no indications of decreased rate of accumulation of EtBr in the absence of glucose in the adapted strains compared with the original strains. This indicated that adaptation to BC did not affect the permeability to EtBr significantly. Further investigations were performed using an assay based on uptake of the fluorescent probe NPN, which fluoresces strongly in phospholipid environments, but only weakly in aqueous environments. This method has been used extensively to determine bacterial permeabilization (Helander and Mattila-Sandholm 2000) and it has been demonstrated that BC increases the permeability of the outer membrane to NPN (Hancock and Wong 1984). There were no indications of reduced permeability of the adapted E. coli strains and it was therefore concluded that other mechanisms were involved.

Exposure of E. coli to stress-inducers (SAL, MV and CDC) generally increased the tolerance to BC and CHL. This indicated that the adaptation to BC could be partly attributed to general stress responses. Tolerance to CHL was also induced by subinhibitory concentrations of BC. This implied that BC could initiate general stress reactions comparable with the effect of stress-inducers. It is well known that exposure of E. coli to SAL leads to increased expression of the multiple-antibiotic resistance (mar RAB) operon (Cohen et al. 1993). The resultant Mar phenotype includes resistance to several antibiotics, organic solvents and triclosan (Ma et al. 1995; Miller and Sulavik 1996; Asako et al. 1997; McMurry et al. 1998). The efflux pump encoded by acrAB is one of the key elements in the Mar stress response giving enhanced resistance to antibiotics (Miller and Sulavik 1996). Moken et al. (1997) demonstrated that a deletion in acrAB rendered a strain of E. coli hypersensitive to a disinfectant containing BC and they suggested that AcrAB is involved in efflux of QACs. Therefore, expression of drug pumps could have contributed to the observed low-level increase in tolerance to BC after growth in SAL (ATCC 11775), MV and CDC (DSM 682). The level of tolerance induced by SAL, CDC and MV in this study is difficult to compare with the values obtained in the literature, because of differences in the methods applied. Usually, the MIC of antibiotics is tested in the presence of stress-inducing agents, but this was not the case in this investigation because of possible interactions with BC.

Active efflux of QACs has been demonstrated for staphylococci (McDonnell and Russell 1999) and efflux of EtBr has been used for model systems for studying QAC-efflux pumps in Gram-positive bacteria (Heir et al. 1998; Aase et al. 2000) and E. coli (Edgar and Bibi 1999). Addition of glucose resulted in EtBr efflux in both adapted and nonadapted cells. The efflux was, however significantly higher in the adapted E. coli ATCC 11775. It was not possible to obtain reproducible results in E. coli DSM 682. However, both adapted strains showed decreased fluorescence when grown on nutrient agar containing EtBr. This is compatible with enhanced efflux. This method has been used to separate between staphylococci with and without genes encoding QAC efflux proteins in an earlier study (Heir et al. 1995).

The level of response when exposed to BC differed between the two strains. The initial MIC of BC was similar for the two E. coli strains and both could be adapted to grow in the presence of 150 μg ml−1 BC. However, the effect of adaptation on the resistance to BC (Table 1) and antibiotics (Table 2) differed. We have also found that adaptation of E. coli K12 to BC does not result in CHL resistance (not shown). Exposure to sublethal concentrations of BC had different effects on resistance to BC and CHL for the two strains (Fig. 2). The effect of the stress-inducers on BC resistance was also not equal (Fig. 2). Therefore, one should be careful before drawing general conclusions about resistance mechanisms from investigations with a limited number of strains.

Reduction of the expression in the porin OmpF has been associated both with spontaneous acquisition of BC resistance in E. coli and exposure to SAL (Ramani and Boakye 2001; Ishikawa et al. 2002). The OmpF porin functions as a nonselective pore through which small hydophilic molecules can diffuse, and the relation to BC resistance is therefore not obvious. However, if global stress responses were involved in adaptation to BC, one would expect expression and repression of proteins that are not necessarily involved in the observed BC resistance. In the present study, a protein putatively involved in murein synthesis was found in outer membrane fractions of adapted strains. Evidently bacteria show several responses to adaptation. The significance of the induced protein is an open question.

The present investigation supports the results of Ishikawa et al. (2002) that exposure to biocides in the laboratory can enhance antibiotic tolerance. A low-level tolerance to a range of antimicrobial agents of BC-adapted cells was found. The practical clinical significance and significance for the food industry of these findings is as yet not known.

Cross-resistance between CHL and BC was demonstrated using both BC- and CHL-adapted strains implying that a common mechanism was involved. Generally, the main mechanisms behind acquired CHL resistance are inactivation of the drug by acetyl transferases, lower permeability caused by loss of outer membrane porins and active efflux (Russell and Chopra 1996). It is not likely that a specific mechanism such as CHL inactivation also applies to QACs, and we were not able to detect changes in the membrane permeability. Therefore, in the light of the observed cross-resistance between QACs and CHL enhanced efflux appears to be the main resistance mechanism involved. This was supported by the competitive inhibition of EtBr efflux in the presence of subinhibitory concentrations of CHL. Similar results were obtained by Edgar and Bibi (1997) and Mine et al. (1998), who studied the multidrug resistance protein MdfA. Whole genome DNA-expression analysis will be used to further elucidate mechanisms behind QAC resistance in E. coli.


We thank Vibeke Høst, Janina Berg, Dorota Dynda, Birgitta Baardsen and Tove Maugesten for excellent technical assistance. We also want to thank Dr Knut Sletten at the Department of Biochemistry, University of Oslo, for carrying out the automatic amino acid sequencing.