Potentially pathogenic bacteria, such as Escherichia coli and Vibrio cholerae, become non‐culturable during stasis. The analysis of such cells has been hampered by difficulties in studying bacterial population heterogeneity. Using in situ detection of protein oxidation in single E. coli cells, and using a density‐gradient centrifugation technique to separate culturable and non‐culturable cells, we show that the proteins in non‐culturable cells show increased and irreversible oxidative damage, which affects various bacterial compartments and proteins. The levels of expression of specific stress regulons are higher in non‐culturable cells, confirming increased defects relating to oxidative damage and the occurrence of aberrant, such as by amino‐acid misincorporation, proteins. Our data suggest that non‐culturable cells are produced due to stochastic deterioration, rather than an adaptive programme, and pinpoint oxidation management as the ‘Achilles heel’ of these cells.
The existence of a viable but non‐culturable (VBNC) state, in which cells are apparently intact, but have lost the ability to form colonies on standard plate‐count media, led Roszak & Colwell (1987) to propose that the formation of VBNC cells is analogous to spore formation in differentiating bacteria. VBNC bacteria are a major concern in public‐health risk assessments because many pathogenic Gram‐negative bacteria, such as Vibrio cholerae, Vibrio vulnificus and Escherichia coli, have been reported to enter a VBNC state, from which they are able to return to the infectious state after passaging in animal hosts (Colwell, 2000; Huq et al., 2000; Olivier, 2000). However, the concept of VBNC formation as a programmed and adaptive phenomenon during nutrient starvation is controversial (for example, see Kell et al., 1998) and another model suggests that cells become non‐culturable due to cellular deterioration, and that these cells are moribund (Dukan & Nyström, 1999; Bogosian & Bourneuf, 2001). Experimental data that could settle this controversy would have an obvious impact on the evaluation of the public health hazard posed by non‐culturable bacteria. However, such studies have been hampered by a lack of techniques that would allow experiments to be carried out on single cells and on subpopulations of cells.
In this study, in situ detection of protein oxidation, and a density‐gradient centrifugation technique, allowed us to analyse cell‐specific oxidative deterioration and differential gene expression in culturable and non‐culturable cells from the same population.
Stasis‐induced carbonylation affects only a subpopulation
We used an immunological assay for the detection of protein carbonyls, and developed this method for use at the single‐cell level (see Methods; Fig. 1A,D). To confirm that the signal obtained using this detection assay reflected the oxidative damage to proteins, we exposed cells to hydrogen peroxide and found that the in situ signal was significantly increased in parallel with the signal obtained from slot‐blot analysis (Fig. 1B,E,H). Treatment of cells with proteinase K (Fig. 1C,F,I), but not with RNase, DNase, or lipase (not shown), abolished the signal. Next, we analysed growing and stationary‐phase (48‐h) E. coli populations. The in situ detection of protein carbonyls showed that the stationary‐phase‐induced increase in oxidative modification (Dukan & Nyström, 1998) affects only a subpopulation of cells (Fig. 2A,B). At this time during stationary phase, all cells have an intact membrane (Fig. 2F,G), and about 60% of the cells in the population have become non‐culturable (see next section).
Non‐culturable cells are specifically carbonylated
To determine whether the increased differential carbonylation of cells is correlated to the VBNC phenomenon, we used a gradient centrifugation technique that has been reported to separate culturable and non‐culturable subpopulations on the basis of a slight difference in cell densities (Siegele et al., 1993). Two density fractions were clearly visible at the time in stationary phase when part of the population had become non‐culturable (Fig. 3A,B). The isolation of cells from the two fractions showed that only a small percentage of the cells in the high‐density fraction were able to form colonies, whereas more than 90% of the cells in the low‐density fraction were culturable (Fig. 3C). Moreover, using the LIVE/DEAD BactLight kit (Molecular Probes) on separate populations, we found that more than 90% of the cells were intact in both subpopulations (Fig. 3C). As shown in Fig. 3, oxidative damage is mainly found in the fraction containing non‐culturable cells.
Specific protein carbonylation in non‐culturable populations
The pattern of protein oxidation in culturable and non‐culturable cells was similar, but several proteins were oxidized only in the non‐culturable cell fraction (Fig. 3D). Some of these proteins were identified, and included the histone‐like DNA‐binding protein H‐NS, glutamate synthase (GltD) and β‐ketoacyl (acyl carrier protein) synthetase (FabB) (Fig. 3D). Other qualitative differences between oxidation in non‐culturable and culturable cells included the significant oxidative damage in the ribosome particle and periplasmic proteins of non‐culturable cells (Fig. 3E).
Upregulating stress regulons in non‐culturable populations
We investigated whether the increased level of oxidative damage was accompanied by a change in the levels of expression of stress regulons. The levels of expression of indicator genes controlled by RpoS, SoxRS, RpoH, RpoE and CpxR were elevated in non‐culturable cells (Fig. 4A). The levels of σS (Hengge‐Aronis, 2000), but not β‐galactosidase from the transcriptional Φ(rpoS–lacZ) fusion, were higher in non‐culturable cells, indicating that σS levels are increased in non‐culturable cells by a post‐transcriptional mechanism (Fig. 4A,C). The increase in σS levels in non‐culturable cells was confirmed by measuring the activities of catalase and glutathione reductase (which are RpoS‐dependent; Fig. 4B) and of σS‐dependent promoters (sodCp, katEp and uspBp; Fig. 4A). In addition, the activities of the soxS promoter (Fig. 4A) and glucose‐6‐phosphate dehydrogenase (Fig. 4B) were higher in non‐culturable cells, indicating the activation of the SoxRS regulon in this subpopulation. Indicator promoters, PdnaK, regulated by the of the heat‐shock regulon, and P3rpoH and PcpxP regulated by σE of the extra‐cytoplasmatic stress response, were also upregulated in non‐culturable cells (Fig. 4A,C). The elevated activites of the rpoH3 and cpxP promoters suggests that non‐culturable cells have defects in the management of proteins in the extra‐cytoplasmatic compartments (Raivio & Silhavy, 2000; Fig. 4A). This is consistent with the data showing elevated carbonylation damage to the periplasmic proteins of non‐culturable cells (Fig. 3E).
In contrast to most of the stress genes analysed, expression of Φ(sodA–lacZ) and Φ(sodB–lacZ) was similar in culturable and non‐culturable cells (Fig. 4A). However, the levels and activities of the corresponding proteins were significantly lower in non‐culturable cells (Fig. 4B,C). Moreover, the gene encoding universal stress‐protein A (UspA) (Nyström & Neidhardt, 1994) was expressed at lower levels in non‐culturable cells. This is interesting because mutants devoid of uspA and sodA sodB have been shown to lose culturability at an increased rate during stasis (Nyström & Neidhardt, 1994; Benov & Fridovich, 1995).
sodA mutants mimic non‐culturable cells
We investigated whether the low abundance and activity of superoxide dismutases (Sods) in non‐culturable cells is a cause of increased stress‐regulon expression. As shown in Fig. 4D, all of the stress genes analysed, except uspA, showed higher levels of expression in sodA mutants compared with wild‐type cells. Thus, the RpoS, SoxRS, RpoH, RpoE and CpxR regulons seem to be activated in response to diminished Sod activity. By contrast, the inactivation of uspA had no such effects on the expression of stress‐regulon genes (data not shown).
The data presented in this paper support the theory that starving E. coli cells lose their reproductive ability due to deterioration, rather than due to a programmed and adaptive entry into a VBNC state, and that this sterility is associated with increased oxidative damage. Indeed, previously observed symptoms of bacterial damage in the global bacterial population (Dukan & Nystrom, 1998) might be associated specifically with sterility. It is possible that sterility might be reversible, depending on the amount of cellular damage. Nevertheless, we propose that, if starvation and oxidation damage are allowed to proceed for an extended period, the non‐culturable cells become moribund and irreversibly lose their life‐supporting mechanisms (see also Ericsson et al., 2000). We infer that the increased production of stress‐defence proteins in non‐culturable cells is a response to increased cellular damage, rather than a programmed developmental pathway that leads to VBNC formation. The increased oxidation damage in the non‐culturable cell population, which is of a highly detrimental and irreversible nature, supports this interpretation. This is also consistent with data showing that starving E. coli cells remain culturable for much longer periods of time in the absence of oxygen (Dukan & Nyström, 1999), and that V. vulnificus cells that were previously thought to be VBNC are, in fact, a subpopulation of the culture that fails to reproduce due to starvation‐induced damage and hydrogen‐peroxide sensitivity (Bogosian & Bourneuf, 2001).
One interesting question is how asymmetry in population damage is generated. Recent reports have shown that more genes than expected show changes in their expression levels during progression through the bacterial division cycle (Bechtloff et al., 1999; Laub et al., 2000; Weitao et al., 2000). It is possible that sudden starvation and growth arrest at a time in the cycle when specific gene products (for example, SodA, SodB and UspA) are present at low levels could generate a subpopulation of cells that undergo increased damage during prolonged stasis. There is no direct evidence for this, but it is interesting to note that the amount of Sod is much lower in non‐culturable cells, and that the pattern of protein carbonylation is similar in non‐culturable cells and cells that lack cytoplasmic Sod activity. For example, self‐inflicted oxidation of proteins is increased both in sod mutants (Dukan & Nyström, 1999) and in non‐culturable cells (this study). Moreover, the oxidation of specific proteins, such as H‐NS, GltD and FabB, is similar in sod‐deficient cells (Dukan & Nyström, 1999) and in non‐culturable wild‐type cells (Fig. 3). In addition, the elevated expression of specific stress regulons (RpoS, SoxRS, RpoH, RpoE and CpxR) in non‐culturable cells might be mimicked by decreased Sod activity. Last, sodA sodB mutants have been shown to lose culturability at an increased rate during stasis (Liochev & Fridovich, 1992; Benov & Fridovich, 1995; Dukan & Nyström, 1999). Thus, it is possible that the loss of reproductive ability of some cells as they enter stationary phase is linked to the abundance of Sod in these individual cells. If so, the longevity of stationary‐phase E. coli cells might, as in ageing fruitflies (Sun & Tower, 1999), be limited by the cellular availability of Sod.
Chemicals and reagents.
Anti‐DnaK mouse monoclonal antibodies were from StressGen Biotechnologies and anti‐RpoS mouse monoclonal antibodies were from Neoclone. Anti‐SodA and anti‐SodB rabbit polyclonal antibodies were gifts from D. Touati. Anti‐mouse and anti‐rabbit IgG peroxidase conjugates were from Sigma. The chemiluminescence blotting substrate (ECL+) and Hybond‐P polyvinylidene difluoride membrane (Amersham) were used in accordance with the manufacturers’ instructions. Radioselectan (76%) was from Schering.
Bacterial strains and media.
All strains used are E. coli K‐12 derivatives of strain MG1655. Cultures were grown aerobically at 37 °C in liquid Luria–Bertani (LB) medium, using Erlenmeyer flasks, in a rotary shaker.
Crude cell extracts were obtained using an SLM‐Aminco French Pressure Cell. Culture samples were processed to produce extracts for resolution on two‐dimensional polyacrylamide gels using the method of O'Farrell (1975) with the modifications described in VanBoegelen & Neidhardt (1990). β‐galactosidase levels were measured as described by Miller (1972), with the modifications described by Albertson & Nyström (1994). Fractionation of cellular compartments was carried out as described previously (Imlay & Fridovich, 1991).
Carbonylation and antioxidant measurements.
Detection of carbonylated proteins was carried out as described previously (Dukan & Nyström, 1998). Superoxide dismutase activity was assayed as described by Imlay & Fridovich (1991). Catalase, glucose‐6‐phosphate dehydrogenase and glutathione reductase activities were measured as described by Gonzalez‐Flecha et al. (1993), Fraenkel & Levisohn (1967) and Lopez‐Barea & Lee (1979).
Cells were fixed using 50% ethanol in PBS, and were then permeabilized with lysozyme (5 mg ml−1) in 100 mM Tris‐HCl, 50 mM EDTA, pH 8.0 for 15 min at 20 °C. The cells were then washed twice in buffer 1 (1.2 M sorbitol, 0.1 M potassium phosphate (pH 6.5), 1% 2‐mercaptoethanol) and resuspended in buffer 2 (1.2 M sorbitol, 0.1 M potassium phosphate (pH 6.5)). Cellular carbonyl groups were derivatized as described in Dukan & Nyström (1998). Cells were then placed on slides that had been pre‐treated with poly‐lysine, and were rinsed with PBS and immersed in cold methanol (−20 °C) for 6 min, followed by immersion in cold acetone (−20 °C) for 30 s. After drying, slides were incubated in blocking buffer (1% BSA in PBS–Tween) for 15 min at 20 °C. Samples were incubated with primary antibody for 2 h and then immersed for 5 min in PBS containing 0.05% Igepal CA‐360. The slides were then incubated with biotinylated secondary antibody (45 min) and, after immersion in PBS, 0.05% Igepal CA‐360 (5 min), were incubated with strepavidin–fluorescein‐isothiocyanate (2 min). Transmission images of cells were obtained, and fluorescence emission from fluorescently labelled antibodies was analysed using Radiance 2000‐MP confocal/multiphoton equipment (BioRad). All samples were laser‐scanned in three dimensions to ensure that any low signals obtained were not the result of the cells being out of focus.
Radioselectan equilibrium density‐gradient centrifugation.
Cells were grown in LB medium at 37 °C and collected after 48 h. Cells were washed in cold (4 °C) phosphate buffer (0.05 M, pH 7) and concentrated tenfold in cold (4 °C) 26.45% Radioselectan. Gradients were made as described previously (Siegele et al., 1993) using 10 ml of radioselectan in polycarbonate centrifuge tubes (25 mm × 89 mm). The gradients can be layered with 1 ml of bacterial solution (5 × 1010 cells) without loss of resolution. Gradients were spun at 55,000 r.p.m. in a Ti60 rotor in a Beckman tabletop ultracentrifuge at 4 °C for 4 h. After collection, the cells were pelleted, rinsed and resuspended in sterile MilliQ water.
This work was funded by a grant from the Swedish Natural Science Research Council and the Foundation for Strategic Research (to T.N.) and by the CNRS–INSU, ATI 2001 (to S.D.).
- Copyright © 2003 Nature Publishing Group