INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

Until a few years ago Cu,Zn superoxide dismutase (Cu,ZnSOD) was considered almost exclusively a eukaryotic enzyme, protecting the cytosol and the extracellular environment of higher organisms from damage by oxygen free radicals (1). Recently, Cu,ZnSOD has been identified in the periplasmic space of a wide range of gram-negative bacteria, including Brucella abortus (6), Haemophilus spp., Actinobacillus spp., Pasteurella spp., Neisseria meningitidis (24-26), Escherichia coli K-12 (7), Legionella pneumophila (40), Salmonella spp. (9), and Mycobacterium tuberculosis (45). This enzyme is thought to protect bacteria from toxic oxygen-free radicals generated outside the cell or in the periplasm itself, since superoxide is unable to cross the cytoplasmic membrane (21). Therefore, Cu,ZnSOD has been proposed to be a determinant of virulence in bacteria potentially exposed to toxic free radicals produced by the host in response to bacterial infection. In vivo experiments have demonstrated the role of bacterial Cu,ZnSOD in the virulence and pathogenicity of infecting microorganisms (15, 18, 19, 36, 42, 43), while in vitro models have provided conflicting data concerning Cu,ZnSOD involvement in bacterial resistance to macrophage killing (19, 42) or survival within nonprofessional phagocytes (42). However, more recent results have shown that this enzyme protects Salmonella enterica serovar Typhimurium (15) and an overproducing strain of E. coli (4) from macrophage killing and that neutropenia restores virulence to an attenuated Cu,ZnSOD-deficient strain of Haemophilus ducreyi in a swine model of chancroid (36).

It is well known that some bacteria, defined as facultative intracellular pathogens, are able to survive within host professional or nonprofessional phagocytes and that this ability plays a pivotal role in infection and disease (20). While it has been clearly demonstrated that the oxidative burst contributes to bacterial killing in phagocytic cells, it is unknown whether nonphagocytic cells are able to kill bacteria by an oxidative pathway.

Considering the wide occurrence of Cu,ZnSOD in facultative intracellular bacteria, we decided to investigate whether this enzyme could offer a selective advantage in survival within epithelial cells. To test this hypothesis, we have used E. coli strains bearing the inv gene from Yersinia pseudotuberculosis in the pRI203 plasmid (23). The expression of invasin, the product of the inv gene, renders noninvasive E. coli strains able to enter cultured mammalian cells but unable to replicate intracellularly. In fact, recombinant invasive E. coli HB101 resides in endocytic vesicles (23) and the number of intracellular viable bacteria significantly diminishes some hours after infection (13). However, this strain carries mutations in the recA and rpoS (39) genes, which are both expected to increase bacterial sensitivity towards oxidative stress. In particular, rpoS encodes a stationary-phase sigma factor, RpoS, which controls a regulon of over 30 genes required for survival in the stationary phase, including several genes providing protection against oxidative stress and resistance to low pH (17, 27). As rpoS is believed to play an important role in bacterial survival within phagocytes (12, 33, 41, 44) and has been shown to modulate sodC expression in E. coli (20a), in this work we have compared the intracellular survival of E. coli HB101(pRI203) with that of a different E. coli strain, GC4468(pRI203), expressing functional rpoS and recA genes.

We have found significantly higher intracellular survival in all invasive strains of E. coli bearing the sodC gene on a multicopy plasmid than in those containing the chromosomal copy or an inactivated sodC gene. These results suggest that bacteria encounter an oxidative stress upon invasion of epithelial cells and show that overproduction of Cu,ZnSOD offers a selective advantage.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Reagents. Ampicillin, bafilomycin A₁, bovine serum albumin, diethyldithiocarbamate, cytochrome c, gentamicin, kanamycin, penicillin, pyrogallol, streptomycin, trypsin, xanthine, and xanthine oxidase were purchased from Sigma. Restriction endonucleases, DNA-modifying enzymes, and catalase were obtained from Boehringer Mannheim. All other chemicals were purchased from BDH and were of the highest grade available. Oligonucleotides were synthesized by Genset. Culture tissue media were obtained from Seromed.

Plasmids, bacterial strains, and culture conditions. The plasmids, E. coli strains, and oligonucleotides used in this work are listed in Tables 1 and 2. Plasmid pRI203 was kindly provided by S. Falkow. This plasmid, which carries the inv gene of Y. pseudotuberculosis, renders E. coli cells able to invade animal cells (23). A DNA fragment containing the whole E. coli sodC gene (22) was obtained by PCR amplification of E. coli chromosomal DNA carried out with the oligonucleotides prom5' and prom3'. The approximately 920-bp amplified DNA was digested with EcoRI and HindIII and subcloned in the corresponding sites of pRI203 to obtain pRIEcSOD. In order to study the transcriptional regulation of sodC, its promoter region (corresponding to the 223 bp before the translation start site) was amplified with the oligonucleotides prom5' and lacZ. The amplified DNA fragment was digested with EcoRI and BamHI and inserted into promoter probe plasmid pMC1403 (11) to obtain pMCPromEcSOD. In this vector, the sodC promoter is cloned upstream of the lacZ coding region, allowing the possibility of analyzing sodC expression following the accumulation of -galactosidase (-Gal).

Host cells. HeLa S3 cells (from an epithelioid carcinoma of the human cervix) and Caco-2 cells (from a human colonic carcinoma) were grown as monolayers at 37°C in MEM supplemented with 1.2 g of NaHCO₃ per liter, 2 mM glutamine, 100 U of penicillin per ml, 0.1 mg of streptomycin per ml, and 10% heat-inactivated fetal calf serum (FCS) in a 5% CO₂ incubator. During the infection experiments, FCS was added at a concentration of 2%.

Invasion assay. Invasion of cultured cells was assayed by a modification of the technique of Isberg and Falkow (23). Briefly, semiconfluent monolayers of HeLa S3 or Caco-2 cells grown without antibiotics in 12-well plates (Costar) were infected with invasive E. coli strains, in either exponential or stationary phase, at a multiplicity of infection (MOI) of 100 (defined as 100 bacteria per cell). The infection was performed for 1 h at 37°C. Then, cells were washed extensively with PBS without Ca²⁺ and Mg²⁺, and 1 ml of fresh medium containing 200 or 100 µg of gentamicin per ml was added to each well, which was infected with E. coli GC4468(pRI203) or GC4468(pRIEcSOD), QC2725(pRI203) or QC2725(pRIEcSOD), and HB101(pRI203) or HB101(pRIEcSOD). After a further 2-h incubation period at 37°C, infected cells were washed extensively and subsequently treated with trypsin-EDTA (a mixture of 0.05% trypsin [1/250] and 0.02% EDTA) for 5 min at 37°C and lysed by the addition of 1.0 ml of cold 0.1% Triton X-100. Cell lysates were diluted in PBS and plated on LB medium containing 100 µg of ampicillin per ml to quantify the number of viable intracellular bacteria.

Intracellular survival assay. After bacterial infection, HeLa or Caco-2 cells were washed and fresh medium containing 50 µg of gentamicin per ml, 2% FCS, 1.2 g of NaHCO₃ per liter, and 2 mM glutamine was added. The monolayers were then incubated for 4, 6, 24, and 48 h at 37°C. At these times, cells were washed and lysed and the number of viable intracellular bacteria was evaluated by CFU counts.

Effect of ammonium chloride and bafilomycin on bacterial invasion and survival. Prior to the invasion assays, HeLa cell monolayers were preincubated for 30 min with or without 20 mM NH₄Cl or 100 nM bafilomycin A₁, either of which is known to inhibit vacuolar acidification (34). After bacterial infection, gentamicin- and NH₄Cl- or bafilomycin-containing medium was added to the cell monolayers. At different times, cultured cells were washed with PBS and lysed, and viable intracellular bacteria were counted, as described above.

Electron microscopy. Twenty-four-well tissue culture plates were seeded with 1 × 10⁶ HeLa cells/well and pulsed with ferritin (0.2 mg/ml) for labeling of lysosomes as described by D'Arcy et al. (14). After 3 h of incubation at 37°C, monolayers were washed five times with MEM to remove free ferritin; then, they were infected with E. coli strains bearing pRI203 and pRIEcSOD at an MOI of 100 (see "Invasion assay," above). At different time intervals (2, 24, and 48 h), cells were incubated with a 0.05 trypsin-0.02% EDTA solution, washed gently with PBS, and pelleted at 600 × g for 10 min. Pellets were fixed with 2.5 mM glutaraldehyde in cacodylate buffer for 1 h at room temperature and postfixed in 1% unbuffered OsO₄. Cells were then dehydrated with increasing concentrations of ethyl alcohol and embedded in Agar 100. Ultrathin sections were stained with uranyl acetate and lead citrate and examined with a Philips 208S transmission electron microscope.

Activity assays. Cu,ZnSOD activity was assayed by the pyrogallol method (30). The periplasmic fraction was obtained by a procedure described previously (3), with the only difference that cells were resuspended at an optical density at 600 nm (OD₆₀₀) of 100. The low expression level of Cu,ZnSOD and the presence of small amounts of cytoplasmic FeSOD and MnSOD in the periplasmic extracts prevent accurate measurements of Cu,ZnSOD activity in E. coli. Therefore, to characterize our model we have determined the Cu,ZnSOD activity in periplasmic extracts before and after a 15-min incubation with 2 mM diethyldithiocarbamate, a copper chelator which inactivates the Cu,ZnSOD enzyme without affecting the activity of MnSOD and FeSOD (7). Protein content was determined by the method of Lowry et al. (29). -Gal activity was measured by a previously described procedure (35).

Characterization of the sodC strain. Susceptibility to extracellular superoxide was evaluated by monitoring bacterial survival upon exposure to superoxide generated by the action of xanthine oxidase on xanthine. Bacteria grown until stationary phase were washed and suspended at a density of about 10⁵ cells/ml in PBS containing 0.1 to 1 mM xanthine and 1 U of catalase (Boehringer Mannheim) to remove hydrogen peroxide generated by xanthine oxidase and ensure that bacterial death was due to superoxide and not to the hydrogen peroxide formed by the spontaneous dismutation of the superoxide anion under assay conditions (37). Superoxide generation was initiated by the addition of xanthine oxidase to a final concentration of 0.1 to 0.5 U/ml. Effective superoxide formation under the above-mentioned conditions was checked by monitoring the rate of reduction of cytochrome c (previously purified by gel filtration) at 550 nm (31). Aliquots of the reaction mixture were diluted and plated at different times to determine the number of CFU per milliliter.

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

Cu,ZnSOD expression in E. coli. Measurements of Cu,ZnSOD activity in periplasmic extracts of E. coli HB101 and GC4468 cells bearing pRI203 or pRIEcSOD are reported in Table 3. Cu,ZnSOD was detectable in all strains only in the stationary phase and the enzyme activity was two- to threefold higher in strains bearing the sodC gene on the multicopy plasmid than in strains bearing the control vector. Strain GC4468 showed higher Cu,ZnSOD activity than HB101. It is worth noting that the Cu,ZnSOD activity values we have found in E. coli strains overexpressing sodC are well below those previously measured in some bacterial pathogens (22, 40). We have also studied sodC expression with a fusion between the sodC promoter and the nucleotide sequence which codifies for -Gal (Table 4). Our results confirm previously reported data showing that rpoS controls sodC (20a, 22) but in addition indicate that a low level of transcription from the sodC promoter may also occur in the absence of this specific sigma factor.

Invasiveness and intracellular survival of different invasive E. coli strains. The capacity of E. coli HB101 or E. coli GC4468 harboring pRI203 and pRIEcSOD to invade epithelial HeLa cells or enterocyte-like Caco-2 cells was assayed. Cell monolayers were infected at 37°C for 1 h with an MOI of 100 at different growth phases. Data of intracellular CFU counts, reported in Table 5, showed a different invasion efficiency (assayed at 2 h postinfection) of E. coli strains depending on the bacterial growth phase: as previously reported (13, 38), the highest level of invasiveness was obtained with logarithmically grown bacteria, which synthesized the greatest amount of invasin. Intracellular bacterial survival at 24 and 48 h after infection was also determined. The results showed that E. coli HB101(pRI203) and E. coli GC4468(pRI203) were susceptible to intracellular killing, whereas E. coli HB101(pRIEcSOD) and E. coli GC4468(pRIEcSOD) were resistant. When cell monolayers were infected with bacteria in the stationary growth phase, in addition to a significant decrease of invasive efficiency for all the strains tested, E. coli HB101(pRI203) displayed the lowest intracellular viability (no viable bacterial cell was recovered within HeLa cells after 24 h of infection). Gentamicin incubation times longer than 48 h of infected monolayers were excluded owing to antibiotic entry into cell monolayers and the consequent killing of intracellular bacteria. Similar results were obtained from the infection of Caco-2 cell monolayers, although the resulting invasive efficiency was at least 50-fold lower (data not shown).

Influence of endosome pH on bacterial survival. Acidification is an important event in several endocytic pathways. E. coli HB101 is known to be highly sensitive to low pH, due to a mutation in the rpoS gene (39). To verify if such a pH sensitivity was responsible for the higher mortality of E. coli HB101(pRI203) than of E. coli GC4468(pRI203), the effects of two different inhibitors of vacuolar acidification, the lipophilic weak base ammonium chloride, and the antibiotic bafilomycin A₁, were tested. The invasion efficiency and intracellular survival of E. coli HB101 were measured in cell monolayers treated with 20 mM NH₄Cl or 100 nM bafilomycin for 30 min prior to infection and during the whole experiment. Cell monolayers were infected with logarithmically grown or stationary E. coli HB101(pRI203). NH₄Cl or bafilomycin pretreatment had no effect on either the entry or the intracellular survival of E. coli strains (Table 6). These results rule out the possibility that intracellular killing was due to a susceptibility to low pH mediated by the altered rpoS gene in E. coli HB101(pRI203). Similar experiments were carried out with E. coli HB101(pRIEcSOD), and similar levels of bacterial survival within treated and untreated cells were observed (data not shown).

Electron microscopy analysis. The fate of E. coli HB101(pRI203) and E. coli HB101(pRIEcSOD) during cell infection was visualized by transmission electron microscopy, and ferritin was used as a lysosomal marker to demonstrate phagosome-lysosome fusion. After 2 h of infection, the strain appeared intact, while after 24 or 48 h E. coli HB101(pRI203) appeared damaged within the endosome-lysosome (Fig. 1). In contrast, cells infected with E. coli HB101(pRIEcSOD) showed that the endosome-lysosome still contained intact bacteria (Fig. 2a); in some observations, it was not possible to visualize a membrane surrounding E. coli pRIEcSOD cells that appeared undamaged, even though they were in close contact with ferritin-labeled material (Fig. 2b).

View larger version (167K): [in this window] [in a new window]   FIG. 1.   Electron micrograph of HeLa cells infected with E. coli HB101(pRI203) at 24 h postinfection. Bar, 250 nm.

View larger version (168K): [in this window] [in a new window]   FIG. 2.   Electron micrographs of HeLa cells infected with E. coli HB101(pRIEcSOD) 24 h postinfection. Bars, 450 (a) and 160 (b) nm.

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

Several authors have investigated the possible involvement of periplasmic Cu,Zn superoxide dismutase in the mechanisms of bacterial protection against the respiratory burst elicited by phagocytes. It has been shown that Cu,ZnSOD-deficient mutants of several bacterial genera are less virulent in in vivo models and that, at least in some cases, the enzyme plays a protective role against superoxide generated by macrophages. Some bacteria producing Cu,ZnSOD are facultative intracellular microorganisms. This lifestyle, in a protected cellular niche, is thought to allow bacteria to escape the host defense mechanisms, thus contributing to the establishment of chronic and recurrent diseases. In order to evaluate whether oxidative species participate in bacterial killing within epithelial cells and whether periplasmic Cu,Zn superoxide dismutase could afford protection, we have chosen an experimental model consisting of nonpathogenic E. coli strains able to invade eukaryotic cells by the expression of the inv gene from Y. pseudotuberculosis but unable to escape from the endosome.

Our data clearly show that overproduction of Cu,ZnSOD protects E. coli strains from killing upon infection by HeLa cell monolayers. In fact, E. coli HB101, GC4468, and QC2725 bearing pRIEcSOD were noticeably more resistant to intracellular killing than E. coli strains bearing pRI203 (Table 5). As also demonstrated by electron microscopy observations, 24 h after the infection E. coli HB101 cells harboring pRIEcSOD were shown to be intact within the endosome-lysosome (Fig. 2a), whereas E. coli HB101(pRI203) cells were damaged (Fig. 1). It is intriguing that in some observations, E. coli HB101(pRIEcSOD) appeared undamaged in close contact with ferritin-labeled material, without a visible endosome-lysosome membrane (Fig. 2b).

The increase of SOD activity in strains bearing pRIEcSOD was only two- to threefold higher than the basal level of strains containing only a single copy of the chromosomal sodC gene (Table 3). This finding is in apparent contradiction with the high expression of -Gal from a sodC-lacZ fusion carried on plasmid pMC1403 (Table 4). Whether this reflects a difference in plasmid copy number or an in vivo-reduced stability of Cu,ZnSOD compared to -Gal or whether the measurement of Cu,ZnSOD led to an underestimation of the amount of enzyme expressed from the plasmid is unclear. However, several factors could be responsible for an underestimation: the SOD assay used (which relies on the use of an inhibitor to discriminate between the activity of Cu,ZnSOD and that of contaminating cytoplasmic SODs), protease sensitivity, the low stability of E. coli Cu,ZnSOD (2, 3, 5), and the requirement of adequate amounts of copper to ensure full catalytic activity to the enzyme (2, 3).

Considering the effect of Cu,ZnSOD overproduction on E. coli survival within epithelial cells, we expected that a sodC-null mutant could be more sensitive to intracellular killing than its isogenic wild-type strain. However, this expectation was only partially supported by our results, which showed that a differential survival between the two strains could be detected only at 48 h postinfection. As the strains overexpressing sodC are much more resistant to intracellular killing, we suggest that the small difference in the intracellular survival of the wild type and the sodC mutant in HeLa cells could be due to the low level of expression of chromosomal sodC in E. coli.

Moreover, we have also shown that the intracellular survival of E. coli HB101(pRI203) infecting HeLa cells in the logarithmic or stationary phase is independent of the vacuole pH. In fact, intracellular survival of E. coli HB101(pRI203) did not change following treatment of monolayers with NH₄Cl or bafilomycin, either of which prevents endosome acidification (Table 6). These results indicate that the acid-sensitive phenotype of E. coli HB101 (39) is not responsible for the low viability of this strain within epithelial cells. The characterization of the peculiar features responsible for the different survival in epithelial cells of invasive E. coli HB101 and GC4468 is outside the aims of this work. However, it is interesting that overexpression of sodC enhances survival of both of these strains and appears to be sufficient to increase the survival of invasive E. coli HB101 (which is known to be highly sensitive to oxidative stress due to mutations in the genes encoding RecA and RpoS) to a level close to that of invasive E. coli GC4468.

Taken together, our findings suggest that oxyradical damage is involved in the mechanisms of bacterial killing by epithelial cells. Such a proposal is in agreement with a previous study which showed that Caco-2 and IEC-18 intestinal epithelial cells are able to kill bacteria by an oxidative pathway (16). While it is well established that superoxide production by an NADPH oxidase located on the cell membrane plays a pivotal role in the oxygen-dependent antimicrobial systems of phagocytic cells, no information is available about the presence of specific antimicrobial mechanisms based on free radicals generating enzymes in epithelial cells. The observation that Cu,ZnSOD protects E. coli from killing within epithelial cells, however, suggests that free radicals can also be produced on the endosomal membrane of nonphagocytic cells.

Although obtained with a model which uses nonpathogenic recombinant E. coli K-12 strains rendered invasive by the expression of a Y. pseudotuberculosis gene, our data encourage investigations of whether sodC plays a role in the intracellular survival of those pathogens that are naturally able to invade epithelial cells.