At 10.0% (w/w) and 4.05% (w/w), respectively, ethanol and 1-propanol killed the mar as well as the parent E. coli strains exponentially, with single-hit kinetics (Figure 1). Despite the rapid loss of viability, optical density of the cultures increased slightly during the first 30–60 min after alcohol addition (not shown), which is indicative of a growth in the size of viable bacteria, with no apparent osmotic lysis.

No statistically significant differences between any strains in 30-, 60- or 90-minute survival fractions could be detected, neither in the ethanol-treated (Figure 1a), nor in the 1-propanol-treated (Figure 1b) cultures. The Mar phenotype expressed in E. coli AG112 does not protect this strain against ethanol- or 1-propanol-mediated killing, compared to the parent strain. The deletion of marA in E. coli AG100/Kan does not render this strain more susceptible to ethanol or 1-propanol, thus mar induction may not significantly contribute to tolerance towards these alcohols.

Growing E. coli are generally more solvent-sensitive than non-growing cells [2], therefore cultures in mid- to late exponential phase were chosen for subsequent solvent addition. In the stationary phase, solvent tolerance mechanisms other than Mar would take effect. Stationary-phase-dependent expression of the alternative sigma factor RpoS up-regulates uspB, which is required for stationary-phase resistance to ethanol [30].

Growth is but one indicator of the physiological status of bacterial cells. E. coli generation times increase in the presence of low concentrations of ethanol [31], presumably through the diminished cellular energy status. At 5% (v/v), ethanol causes growth arrest after one mass doubling with incomplete septation, also indicating an effect on cell wall synthesis [31]. This growth arrest could be alleviated by dilution with fresh medium. In contrast, the ethanol- and 1-propanol effect observed here is irreversible, killing the cells. In minimal medium, 5% (v/v) of ethanol has been shown to cause cellular lysis in the absence, but not in the presence, of 0.1 M NaCl. Peptidoglycan cross-linking was also shown to decrease [2]. It was proposed that peptidoglycan cross-linking enzymes may be inhibited because ethanol generally weakens hydrophobic interactions within the periplasm as well as the membrane interior, and that this may be overcome by NaCl, as well as other antichaotropic salts. Similar lysis has been reported in E. coli grown in Luria-Bertani broth containing up to 5% (v/v) of ethanol [32]. That no gross lysis ensued in our case may be surprising, especially since NaCl was omitted from the growth medium. However, actively growing cells, simultaneously synthesizing poorly cross-linked peptidoglycan and displaying autolysin activity, would be expected to readily lyse. Rapid cellular inactivation due to other mechanisms, at the relatively high alcohol concentrations used here may hence mask any impairment of cell wall synthesis, or autolysins may simply be just as rapidly inhibited.

E. coli AG100/Kan is a mar deletion mutant lacking most of the mar operon including the mar repressor and Mar activator genes. Salicylate can elicit no MarA-dependent solvent tolerance in this strain, but hypothetically so in the parent strain AG100 as cyclohexane tolerance was earlier shown to increase by salicylate addition (or marA overexpression) in E. coli W3110; not carrying any mar mutation [25]. Addition of 10.0% ethanol to E. coli AG100 and AG100/Kan grown in the presence of 2 mM sodium salicylate results in an increased killing rate, compared to AG100 grown without salicylate (Figure 2a). Similarly, salicylate increases the killing rate of 4.05% 1-propanol in both strains (Figure 2b). The 60- and 90-minute survival fractions of salicylate/solvent-exposed AG100 as well as AG100/Kan differ significantly from non-salicylate treated AG100, for both solvents. In contrast, no statistically significant differences between AG100 and AG100/Kan survival rates in the presence of salicylate can be shown. Therefore, salicylate (2 mM) significantly enhances ethanol- and 1-propanol mediated killing of E. coli K-12, irrespective of any salicylate-mediated Mar induction. This further strenghtens our view that the Mar phenotype cannot protect E. coli against lower primary alcohol (ethanol, 1-propanol) mediated killing.

Interestingly, synergism between a weak acid and ethanol in killing of E. coli has been described before [33]. 50 mM of lactate in combination with 5% (v/v) ethanol enhanced killing of E. coli O157:H7 about four log units in the exponential phase (first hour after addition), compared to ethanol only. It was also found that both lactate and ethanol independently lowered the cytoplasmic pH of the cells. We have not investigated cellular pH, but salicylate at 2.5 mM, close to the concentration used here, has been shown to increase bacterial sensitivity towards aminoglycosides [34,35]. This has been proposed to partly result from intracellular deprotonation following salicylic acid diffusion across the cytoplasmic membrane [36,37].

Ma and coworkers [14] have described the induction of acrAB transcription by 4% ethanol, even in a Δmar strain; thus independently of the Mar response. In this study they did establish whether ethanol was a substrate of the AcrAB pump. AcrAB and TolC are both necessary to extrude solvents with a P_ow greater than 3.4 [38]. We measured the killing rate of ethanol and 1-propanol in Δ(acrAB) as well as Δ(acrAB) marR strains (Figure 3). Again, no statistically significant differences between AG100, AG100A and AG102/K survival rates after ethanol or 1-propanol exposure could be shown. This suggests that the AcrAB efflux pump cannot pump ethanol or 1-propanol across the membrane.

In summary, the fact that salicylate similarly enhanced ethanol- and 1-propanol-mediated killing, and that a lower concentration of the more hydrophobic 1-propanol was required to reach a comparable killing rate, suggest a common, synergistic relationship between salicylate and the alcohols, as well as a common, general killing mechanism presumably based on the partitioning of solvent into the plasma membrane. Neither the Mar phenotype, a well-characterised multi-drug and organic solvent tolerance phenotype, nor any stress-dependent induction of Mar or its central component AcrAB could alleviate the effect of ethanol- and 1-propanol. Collectively, these results establish a lower limit in terms of size and functionality for substrates for the AcrAB pump, information which should prove valuable for understanding resistance mechanisms, and for the use of microorganisms in biotechnological processes involving organic solvents [39,40].

E. coli AG100 (K-12 argE3 thi-1 rpsL xyl mtl Δ(gal-uvrB) supE44) [27], AG112 (marR 5-bp deletion) [28], AG100A (Δ(acrAB)::Kan) [19], AG102K (marR Δ(acrAB)::Kan) [19] and AG100/Kan (Δ(marCORAB)::Kan) [29] were all kindly donated by Dr. S. B. Levy, Center for Adaptation Genetics and Drug Resistance, Tufts University School of Medicine, Boston MA, USA. All strains were maintained on Luria-Bertani (LB) Agar (10 g Tryptone (Merck), 5 g Yeast extract (Merck), 5 g NaCl, 12 g agar (Difco) and 10 mL 10% (w/v) d-glucose, separately autoclaved l⁻¹); and where appropriate supplemented with 12.5 μg kanamycin mL⁻¹.

Overnight cultures of E. coli in modified LB Broth (10 g Tryptone (Merck), 5 g Yeast extract (Merck) and 10 mL 10% (w/v) d-glucose, separately autoclaved l⁻¹) were re-inoculated (1:100) into the same medium with or without sodium salicylate (2 mM) and shaken vigorously at 37 °C. In mid- to late exponential phase (≈5 × 10⁸ cfu mL⁻¹; 3–4 h), 1-propanol (4.05%, w/w) or ethanol (10.0%, w/w) was added. Shaking was continued at 37 °C with samples being removed every 30 min for subsequent dilution in cold Dil (1 g Nutrient Broth (Merck), 5 g NaCl L⁻¹) before plating on LB Agar. Plates were incubated overnight at 37 °C for the determination of viable counts. Survival at each time point was calculated as fractions of the viable count at the time of solvent addition. Data were statistically treated using the unpaired t-test. Optical density of liquid cultures was measured at 635 nm using an Aquanal-plus Spectro™ water analyser (Riedel-de-Haën, Germany).