Article Text


Escherichia coli strains colonising the gastrointestinal tract protect germfree mice againstSalmonella typhimuriuminfection
  1. S Hudault,
  2. J Guignot,
  3. A L Servin
  1. Institut National de la Santé et de la Recherche Médicale, Unité 510, Faculté de Pharmacie Paris XI, 92296 Châtenay-Malabry, France
  1. S Hudault. sylvie.hudault{at}


BACKGROUND Escherichia coli is part of the normal gastrointestinal microflora which exerts a barrier effect against enteropathogens. SeveralE coli strains develop a protective effect against other Enterobacteriaceae.

AIMS TwoE coli strains, EM0, a human faecal strain, and JM105 K-12 were tested for their ability to prevent in vivo and in vitro infection by Salmonella typhimurium C5.

METHODS Inhibition of C5 cell invasion by E coli was investigated in vitro using Caco-2/TC7 cells. The protective effect ofE coli was examined in vivo in germfree or conventional C3H/He/Oujco mice orally infected by the lethal strain C5.

RESULTS EMO expresses haemolysin and cytotoxic necrotising factor in vitro. In vitro, the two strains did not prevent the growth of C5 by secreted microcins or modified cell invasion of C5. In vivo, establishment of EM0 or JM105 in the gut of germfree mice resulted in a significant increase in the number of surviving mice: 11/12 and 9/12, respectively, at 58 days after infection (2×106/mouse) versus 0/12 in control germfree group at 13 days after infection. Colonisation level and translocation rate of C5 were significantly reduced during the three days after infection. In contrast, no reduction in faecal C5 excretion was observed in C5 infected conventional mice (1×108/mouse) receiving the EM0 or JM105 cultures daily.

CONCLUSIONS Establishment of E coli strains, which do not display antimicrobial activity, protects germfree mice against infection and delays the establishment of C5 in the gut. Possible mechanisms of defence are discussed.

Statistics from

The normal resident intestinal microflora is the major factor protecting animals and humans against intestinal colonisation by pathogenic bacteria.1 2 In rare cases it has been possible to isolate the bacterial species responsible for protection and to elucidate the mechanism of protection.Escherichia coli is one of the first bacterial genera, along with Streptococcus, to colonise the intestine of human3-6 and animal newborns, including mice, rats, piglets, and chickens. It has been reported that several E coli strains develop a protective effect against antibiotic resistant, colicin sensitive, and enterotoxigenic E coli.7-13 Barrow and Tucker14demonstrated strong inhibition of intestinal colonisation of chicken caecum with Salmonella typhimurium by pretreatment (24 hours before) with strains of threeE coli strains but the strains were not equally effective against other strains ofSalmonella.

Here we examined the efficacy of an E coliK-12 JM105 strain and a human faecal E colistrain, EM0, against S typhimuriuminfection in mice and in vitro in human cultured cells. Salmonella can be divided in two groups—those that typically produce typhoid-like infections in humans or animals and those that produce non-life threatening gastroenteritis (food poisoning). The severity of the disease depends on the serovar and the host. In nearly all Salmonellainfections, the bacteria multiply in the lumen in the small intestine, rapidly penetrate the intestinal mucosa, and reach the mesenteric lymph nodes (MLN) where they multiply. Most infections do not proceed beyond the local lymph nodes. Some more invasive strains can spread into the circulatory system and to deeper tissue such as the spleen and the liver. Infected bile causes a secondary intestinal infection and leads to the carrier state.15 In mice, S typhimurium induces typhoid-like disease and its primary site of invasion is the terminal ileum.16 17 Peyer's patches and M cells of the small intestine are the first to be invaded18 allowing theSalmonella to reach the MLN, and then the spleen and the liver.16 It is known that mouse genotype plays a critical role in the sensitivity to the lethal effect in systemic salmonellosis induced by S typhimurium. We have chosen C3H/He/Oujco mice which have the same characteristics as C3H/He mice. This breed19 is resistant to S typhimurium C5 given subcutaneously (LD50 >105). We observed that germfree C3H/He/OuJIco mice died in approximately eight days after oral infection by 2×106 bacteria/mouse whereas their conventional counterparts survived.20

The E coli strain EM0 was originally isolated from the faecal flora of a healthy human volunteer.9 It has been shown to have a protective effect against establishment of E coli antibiotic resistant strains in germfree mice,9 against antibiotic resistant enterobacteria in human newborns,13 21 and gnotobiotic mice.22 It has a protective effect against ETEC K88 in piglets.9 12 The mechanism suggested is either “adaptation” of the EM0 strain inoculated first to germfree mice as ultrastructural differences in cell morphology were observed in vivo but disappeared in in vitro cultures,9 or a slower generation time for the resistant enterobacteria mutant.22Moreover, this strain plays a role in the development of the immune system in germfree mice (MC Moreau, personal communication), inducing oral tolerance to ovalbumine23 and the production of cytokines in peritoneal and bone marrow macrophages.24 25

To investigate the antagonistic activity of the E coli strains against S typhimuriumoral infection in vivo, two experimental designs were used:E coli were given to either: (i) germfree animals to study the effect of the extensive gut colonisation byE coli on protection againstS typhimurium infection; or (ii) conventional C3H/He/Oujco mice to examine if an E coli strain, given daily as a probiotic that can transit along the gut, exerts an antagonistic effect in situ against infectingS typhimurium. In vitro,S typhimurium invades enterocyte-like Caco-2 cells.26 We next examined if the E coli strains exerted an antibacterial effect in vitro—that is, whether they could inhibit invasion of cultured human enterocyte-like Caco-2/TC7 cells by S typhimurium.



The E coli strain EM0 was originally isolated from the faecal flora of a healthy human volunteer9 and was kindly supplied by Y Duval (INRA, Jouy-en-Josas, France). The strain JM10527 is an E coli K-12 strain28 kindly provided by M Fons (INRA). S typhimuriumstrain C5 was kindly provided by M Popoff (Institut Pasteur, Service des Entérobactéries, Paris, France).29 E coli strains and S typhimurium C5 were grown in Luria-Bertoni broth (LB) (Difco, Detroit, Michigan, USA) for 18 hours at 37°C under aerobic conditions. After centrifugation, the bacteria were resuspended and adjusted to the appropriate concentration in sterile phosphate buffered saline (PBS) for assay. Viable counts of E colior S typhimurium were obtained after plating suitable dilutions on tryptic soy agar (TSA) (Difco) and incubation at 37°C for 18 hours. When differential counts ofS typhimurium from E coli were necessary, viable bacteria were enumerated onShigella andSalmonella agar (SS agar) (Difco). After incubation for 24–48 hours at 37°C, colonies ofS typhimurium were black and easily distinguishable from E coli colonies or normal resident enterobacteria which gave red or pink colonies.


Polymerase chain reaction (PCR) detection of α-haemolysin (hlyA), cytotoxic necrotising factor 1 and 2 (cnf 1-2), and cytolethal distending toxin (cdt) sequences were performed with gene Amp PCR system 2400 (Perkin-Elmer Applied Biosystems, Courtabeouf, France). Primers hlyA1 (5′-CTC ATT GGC CTC ACC GAA CGG-3′) and hlyA2 (5′-GCT GGC AGC TGT GTC CAC GAG-3′) were designed to amplify a 299 bp internal fragment fromhlyA gene. Primers cnfA (5′- CTG AGC GGC ATC TAC TAT GAA G-3′) and cnfB (5′-CCT GTC AAC CAC AGC CAG TAC-3′) were designed to amplify a 626 bp internal fragment fromcnf gene. Primers cdt1 (5′- GTW GCR ACY TGG AAY YTK CAR GG -3′) and cdt2 (5′-KCM GGY KMR CGR TTR AAA TCW CC -3′) were designed to amplify a 500 bp internal fragment fromcdt gene. Colony PCR was carried out using “PCR beads ready to go” (Amersham Pharmacia, Saclay, France) according to the manufacturer's protocol. After initial denaturation (five minutes at 94°C), the samples were subjected to 30 cycles of amplification, each of which consisted of the following steps: 30 seconds at 94°C, 30 seconds at 57°C, and one minutes at 72°C. A final extension step of 10 minutes at 72°C was conducted. PCR products were examined on 1% agarose gels. As positive controls, we used E coli SE124hly +,30 E coli J96hly +, andcnf+,31 kindly provided by P Boquet (Inserm U452, Nice, France), andE coli DH5αpOMEO1,32 kindly provided by E Oswald (Ecole Nationale Vétérinaire, Toulouse, France).


For qualitative evaluation of haemolysin production, isolates were inoculated onto Columbia agar plates (Biomérieux, Marcy l'Etoile, France) containing 5% sheep blood.33 Haemolysis was defined as a distinct zone of clearing around and under isolated bacterial colonies after three or 18 hours at 37°C.


Cell integrity was determined by measuring release of lactate dehydrogenase (LDH) from cells in the culture medium (Enzyline LDH kit; Biomérieux, Dardilly, France). Infection was conducted with the bacterial suspension of E coli(1×108 cfu/well) or with the filtered spent culture supernatants (0.22 μm filter unit, Millex GS; Millipore, Saint Quentin, France) GmbH). Aliquots (20 μl) of the incubation medium were sampled one, two, three, or four hours after infection. Results are given as percentage LDH released calculated as follows: LDH release (UI) found in test well/LDH release (UI) found in control well after cell lysis with distilled water (3000 UI). Assays were run at least in triplicate for three successive cell passages. Multinucleating effect due to CNF was studied on HeLa cells and on Caco-2/TC7 cells, as described by de Rycke and colleagues.34 Briefly, a 100 μl volume of a cell suspension containing 4×104cells/ml (HeLa cells) or 6×104/ml cells (Caco-2/TC7 cells) was distributed in culture cell well plates (Corning Glass Works, Corning, New York, USA) with a glass coverslip, and 22.5 μl of a fourfold dilution of the bacterial extracts in PBS was added. The multinucleating effect was observed after 72 hours of incubation after the cells were stained (Giemsa stain) by light microscopy (Leica Aristoplan, Rueil-Malmaison, France).


The cultured Caco-2/TC7 clone cells35 established from the parental human colonic adenocarcinoma Caco-2 cell line36 37 were used. Cells were routinely grown in Dulbecco's modified Eagle's minimal essential medium (25 mM glucose) (Eurobio, Paris, France), supplemented with 10% inactivated fetal calf serum (Boehringer, Mannheim, Germany) and 1% non-essential amino acids. For inhibition of cell association and cell invasionof S typhimurium, monolayers of Caco-2/TC7 cells were prepared in six well Corning tissue culture plates (Corning Glass Works). Cells were seeded at a concentration of 1.4×104 cells/cm2. Maintenance of cells and all experiments were carried out at 37°C in a 10% CO2/90% air mixture. The culture medium was changed daily. Cells were used at late post confluence—that is, after 15 days in culture. HeLa cells were cultured at 37°C with 5% CO2 in RPMI 1640 with 2 mM l-glutamine (Life Technologies, Cergy-Pontoise, France) supplemented with 10% inactivated fetal calf serum.


The cell infection assay was conducted as previously reported.38 Briefly, prior to infection, the Caco-2/TC7 monolayers were washed twice in PBS. Monolayers were then incubated for 15–30 minutes with PBS before infection. A suspension of 2×108 cfu/ml (0.5 ml) of bacteria in PBS and 0.5 ml of the cell culture medium (containing 1% mannose to prevent pili-1 mediated adhesion) were added to each well of the tissue culture plate. The plates were incubated for different times at 37°C in 10% CO2 /90% air and washed three times with sterile PBS.S typhimurium internalisation was determined by quantitative determination of bacteria located within the infected monolayers using the aminoglycoside antibiotic assay. After one hour of infection, monolayers were washed twice with sterile PBS and incubated for 60 minutes in medium containing 100 μg/ml of gentamicin to kill extracellular bacteria. The monolayers were washed three times with PBS and lysed with sterilised distilled water. Appropriate dilutions were plated on SS to determine the number of viable intracellular bacteria by bacterial colony counts. Assays were conducted several times with three successive passages of Caco-2/TC7 cells.


Inhibition of S typhimurium C5 cell invasion by E coli was determined as follows: S typhimurium (1×108cfu/ml) suspended in PBS with the E colistrain in its supernatant (1×108 cfu/ml) were added to each Caco-2/TC7 culture well for one hour at 37°C. For the EM0 strain, the experiment was conducted in the presence of dextran 4 (Sigma) at a final concentration of 30 mM for osmotic protection against the pores induced by E coli EM0 haemolysin,39 for one hour at 37°C. Control experiments were conducted with PBS, pH 7. Determination of viable intracellular S typhimurium was conducted as described above.


Detection of colicin was performed as previously described.11 Briefly, aliquots of 12 hour cultures ofE coli EM0 or JM105 were placed on hydrophobic membranes (0.45 μm) (HGMF QA Life Science Inc., San Diego, USA) which were themselves placed on brain heart infusion agar plates (Difco). After incubation for 24 hours at 37°C, membranes were removed and plates were overlayed with 10 ml of agar. This overlay agar consisted of minimum medium containing 6 g/l of agar, inoculated with a fresh overnight culture of S typhimurium C5. After incubation for 18–24 hours at 37°C, the clear zone surrounding the membranes indicated the presence of antibacterial compounds. Co-culture of S typhimurium C5 andE coli strains EM0 or JM105 were performed by incubating approximately 1×108 cfu/ml of each bacteria at 37°C in LB broth. Initially, and at predetermined intervals, aliquots were removed, serially diluted, and plated on SS agar to determine bacterial colony counts, as described above.


Both germfree (Cesal, Orléans, France) and conventional mice (Iffa Credo, L'Arbresle, France) were adult female C3H/He/Oujco mice, 7–8 weeks of age. They were housed, fed, and sacrified in accordance with the highest standards of humane animal care and the relevant national legislation. Germfree mice were reared in Trexler type isolators fitted with a rapid transfer system (La Calhène, Vélizy Villacoublay Cedex, France). Germfree mice were checked for freedom from bacterial contamination by culture of fresh faeces aerobically and anaerobically. They were given ad libitum a commercial diet RO3 40 (UAR, Villemoisson/Orge, France) sterilised by gamma irradiation (40 kGy) and autoclaved demineralised water. Conventional mice were fed an identical but not sterile diet.


Germfree mice were infected orally by drinking aSalmonella suspension in bottled water. Animals were deprived of water from the day before. Each mouse drank approximately 5 ml of the suspension containing 4×105cfu/ml (that is, approximately 2×106 cfu/mouse) within a few hours. This protocol was used to avoid the risks of injuring with an intragastric gavage needle (difficult to manage in isolators) and oral and systemic inoculation. E coli EM0 or JM105 strains were inoculated in germfree mice in a similar manner as that for Salmonella. Where the effect of colonisation of E coli on establishment of C5 was studied, Salmonella was inoculated seven days after E coli inoculation. Germfree mice infected by one strain only,Salmonella or E coli, were termed monoassociated mice. Diassociated mice harboured two bacterial strains (E coli andSalmonella). Mono or diassociated germfree mice were termed gnotobiotic mice.

Conventional mice were infected intragastrically via a gavage needle. On day 0, mice received either 0.2 ml of LB broth (control) or 0.2 ml of E coli culture (assay, 108cfu/mouse). A few minutes later they were infected with a single oral dose of S typhimurium C5 (1×108 cfu/mouse). For seven days after infection of C5, treatment with LB broth or E coli culture was administrated daily at the same dose as on day 0.


Fresh faecal samples were collected from the anus of each mice. Faecal samples were obtained one, four, and seven days after infection in C5 infected conventional mice. They were collected daily in the case of gnotobiotic mice. Faeces were weighed and diluted 10-fold in PBS. Viable C5 bacteria were determined by plating serial decimal dilutions on SS agar medium (Difco) to differentiateSalmonella counts from those of the resident enterobacteria, as described above. Results are expressed as the mean value of viable Salmonella counts (log cfu/g of faeces).

Mono or diassociated mice were killed by cervical dislocation. Tissue samples were obtained in the following order: mesenteric lymph nodes (MLN), spleen, and liver. The intestines were unfolded gently to collect MNL and a sterile swab was passed over the intestinal cavity to verify that the intestinal wall had not been damaged. This swab was soaked in 1 ml of PBS which was plated on SS medium. Whenever it proved positive (>5 cfu/ml), the results of the corresponding mouse for MNL, spleen, and liver were discarded. The different segments were removed in the following order: small intestine (divided into three segments corresponding approximately to the duodenum, jejunum, and ileum) and caecum. In most experiments, only the contents were sampled. However, in one experiment the intestinal wall of the segment was obtained. After removal of the contents, the intestinal wall was gently washed with eight successive 5 ml sterile PBS aliquots and drained before being weighed. All content samples were weighed and diluted 10-fold in PBS. Organs were weighed and homogenised with 2 ml of PBS by Ultraturrax for two minutes and diluted 10-fold. The number of viable bacteria in the samples was estimated by plating serial decimal dilutions on TSA or SS agar when differential counts ofS typhimurium and E coli were necessary. Bacterial counts are given per gram of intestinal wall per organ (MLN, spleen, and liver) or per gram of contents.


Numbers of viable bacteria were compared by variance analysis using the Student's t test. Numbers of surviving mice were compared by the exact Fisher test.40



Although the E coli strain EM0 was isolated from the faeces of a healthy human volunteer and was described as a non-pathogenic strain,13 we found that it had virulence factor genes such as haemolysin (hly) and cytotoxic necrotising factor (cnf). As shown by PCR (fig 1), theE coli strain EM0 showed specific amplification with the primers hly and cnf as did the positive controlE coli strains J96,hly+ and cnf+, and SE124, hly+. In contrast, no amplification was found with the cdt primers. In addition,hly, cnf, andcdt were not detected by PCR inE coli JM105 strain.

Figure 1

Polymerase chain reaction (PCR) amplification of hly, cnf, and cdt from E coli EM0 strain. Lane 1: 100 bp molecular weight marker; lanes 2–4: hly primers; lanes 5–7: cnf primers; lanes 8–10: cdt primers; lane 11: 1 kb molecular weight markers; lanes 3, 6, and 9: E coli EM0; lanes 4, 7, and 10: E coli JM105; lane 2: E coli SE124, hly+; lane 5: E coli J96, cnf +; lane 8: E coli DH5αpOMEO1, cdt+. The E coli strain EM0 showed specific amplification with the hly and cnf primers whereas the E coli JM105 strain showed no amplification.

Functional haemolysin was revealed by observation of blood haemolysis on sheep blood Columbia agar plates. Concurrently, we found that cell lysis developed in EM0 infected Caco-2/TC7 cells. Indeed, cell infection was followed by release of intracellular LDH in the culture medium which develops as a function of time after infection. In EM0 infected cells at two and three hours after infection, 23 (2)% and 98.7 (6)% of intracellular LDH was released, respectively. The two positive control E coli strains J96 and SE124 promoted similar LDH release. Observation of infected cells showed that the monolayers become fragile at one hour after infection and were entirely destroyed at three hours after infection. The culture supernatants of the EM0 strain and the haemolysin positive control strains were inactive, indicating bacterial contact dependent cell lysis.

Cytotoxicity related to CNF has been reported to be characterised by the appearance of multinucleated cells.34 In agreement with this, we observed that infection by the E coli strains EM0 and J96 both in HeLa and Caco-2/TC7 cells was followed by the appearance of a high level of multinucleated cells (not shown).


To study intestinal colonisation in germfree C3H/He/Oujco mice byE coli EM0 and JM105 strains, two groups of six mice, reared in separate isolators, received E coli EM0 or JM105 bacteria as a single dose (106cfu/ml). Each strain became rapidly established in the gut. EM0 and JM105 bacteria were detected at a level of 9.7 (0.0) and 9.4 (0.0), respectively, (mean (SEM) log 10 cfu/g of fresh faeces) one day after oral inoculation, and at 10.5 (0.1) and 10.3 (0.1) on and after day 2 post inoculation. By day 8 after inoculation, mice were killed and the population levels of E coli EM0 and JM105 were determined in intestinal contents (table 1). The twoE coli populations similarly increased from the proximal to the distal intestine. E coliEM0 and JM105 were also detected in MLN at about 100 and 50 bacteria per organ, respectively, indicating translocation. The number of tissue associated E coliwas measured in washed organs. E coli EM0 and JM105 were only partly associated with the mucosa of the gastrointestinal tract of monoassociated mice as the number of associated bacteria was always 10 or 100 times lower. This association on the mucosa increased from the proximal to the distal intestine, as for intestinal contents. The number of E coli in aliquots of the last wash of each organ was always 102–104 times lower than the number of E coli remaining in the washed organ itself. Although the EM0 strain has two specific virulence genes, such as haemolysin and CNF demonstrated by PCR, and shows functional effects in vitro, no adverse effect was observed in germfree mice colonised by this strain in our experimental conditions, as previously reported by others.9 12 22

Table 1

Translocation and distribution in the gut of E coli JM105 or EM0 in monoassociated mice


We examined the effect of each E colistrain established in the gut of germfree C3H/He/Oujco mice onS typhimurium C5 induced mortality. Two groups of 12 germfree C3H/He/Oujco mice, reared in separate isolators, were inoculated with EM0 or JM105 E coli.Eight days after E coli inoculation, they were orally infected with S typhimuriumstrain C5 (2×106 cfu/mouse) and compared with a group of 12 germfree mice infected singly by C5 under the same conditions. As shown in fig 2, establishment of E coli EM0 or JM105 in germfree mice significantly increased survival of C5 infected mice. Indeed, at 13 days after infection, all diassociated C5 infected mice survived while all control C5 infected mice were dead (p<0.0001). The EM0 strain apparently gave better protection than the JM105 strain as 11 of 12 and nine of 12 mice, respectively, survived 58 days post C5 infection. However, the difference between the EM0 and JM105 groups was not significant.

Figure 2

Survival of control germfree mice (n=12) and ex germfree mice monoassociated with E coli JM105 (n=12) or EM0 (n=12), after S typhimurium C5 oral infection. C5 oral infection (2×106 cfu/mice) occurred eight days after inoculation of E coli. Significant difference between control germfree group and EM0 and JM105 diassociated groups (p<0.0001).


The kinetics of establishment of S typhimurium in faeces of EM0 or JM105 monoassociated mice (12 per group) were determined and compared with those obtained in the control germfree group (six mice) after infection by C5 (table 2). On the first day after infection, the level of C5 in faeces was significantly lower in both EM0 and JM105 E coli diassociated groups compared with the C5 infected germfree control group (p<0.0001 and p<0.001, respectively). Afterwards, the level of C5 in faeces evolved differently for each group of E coli monoassociated mice. This inhibition was maintained in the EM0 associated group because by day 5 after C5 infection, the C5 population level had increased and reached the level observed on day 2 after infection in the germfree infected control group. In contrast, for JM105 associated mice, inhibition was no longer significant by days 2 and 3 after infection.

Table 2

Comparative kinetics of colonisation of S typhimurium C5 in germfree control mice and in monoassociated mice with E coli EM0 or JM105

The translocation rate and levels of C5 in the contents of the successive digestive segments were determined in two groups ofE coli associated mice and in control germfree mice on day 3 after infection by C5 (table 3). Increasing levels of C5 were found from the Si1 segment to the caecum of the control group and from those of the two groups of E coli diassociated mice. However, a significant difference was observed between C5 levels in segments Si2, Si3, and the caecum of EM0 diassociated mice compared with C5 infected control mice. In JM105 diassociated mice, the population level of C5 decreased significantly only in distal digestive segments (Si3 and the caecum). The translocation rate of C5 was lowered significantly in both EM0 and K-12 JM105 diassociated mice compared with the control C5 infected group. No significant difference was observed between the two groups ofE coli associated mice.

Table 3

Intestinal colonisation and translocation of S typhimurium C5 in germfree mice associated or not with E coli EM0 or JM105, three days after C5 infection, and in surviving mice at 58 days after infection

In surviving JM105 or EM0 monoassociated C5 infected mice at day 58 after infection, the Si3 and caecal C5 population, and tissue translocation of C5 were determined (table 3). The intestinal and caecal population level of C5 was significantly lower in the EM0 associated mice group than in JM105 associated mice. In contrast, the translocation rate of C5 was identical for both groups ofE coli associated mice. When examining the level of E coli in the faeces of gnotobiotic C5 infected mice from day 5 to day 58 after C5 infection, we found that the populations of E coli remained at a similarly high level, which varied from 9.2 (0.1) on day 5 to 8.8 (0.2) log 10 cfu/g of faeces on day 58 for the EM0 monoassociated C5 infected group. In the case of the JM105 monoassociated C5 infected group, populations of E coli decreased from 9.1 (0.2) on day 5 to 8.2 (0.4) log 10 cfu/g of faeces on day 58.


To determine if the decrease in C5 excretion observed in JM105 and EM0 monoassociated C5 infected germfree mice developed in the presence of a resident microflora, we used three groups of six conventional C3H/He/Oujco mice. Concomitantly with infection withS typhimurium C5 (1×108cfu/mouse), conventional mice were orally inoculated daily with one of the E coli strains (1×108cfu/day/mouse) or LB broth for the control group. Faeces were collected one, four, and seven days after infection. Figure 3 shows that the level of viable C5 bacteria in faeces was not significantly different in faeces of E coli treated C5 infected groups compared with the untreated control C5 infected group. This result demonstrated that in contrast with the germfree situation,E coli strains JM105 and EM0 did not decrease Salmonella excretion in conventional mice.

Figure 3

Faecal excretion of S typhimurium C5 from conventional mice treated daily with E coli strains EM0 and JM105, after C5 oral infection. Control and treated mice were infected with a single dose of S typhimurium C5 (1×108 cfu/mouse) on day 0. The treatment consisted of 0.2 ml of LB broth (control) or E coli culture (2×108 cfu/mouse) and was given on day 0 and then daily for seven days after infection. Infection and treatment were achieved using a gastric probe.


Enterobacteriaceae strains exert inhibitory activities to closely related bacteria by producing either microcins or colicins.10 We have previously reported thatLactobacillus casei GG andL johnsonii La1 strains produced an antibacterial substance that was active in vitro againstSalmonella.20 41 42 These twoLactobacillus strains, once established in the gastrointestinal tract of germfree mice, can protect them againstS typhimurium infection and accelerate faecal elimination of Salmonella from the gut of conventional mice treated with these bacterial cultures.20 42 We investigated if theE coli strains EM0 and JM105 could produce antimicrobial substances by two methods. Firstly, co-culture ofE coli EM0 with S typhimurium C5 in LB showed no inhibition ofSalmonella growth. The inoculum of 7×107 cfu/ml of each strain grew simultaneously and reached the level of 4×108 and 6×108 (log 10 number of cfu/ml) for C5 and EM0, respectively, after five hours of culture and 3×108 for the two strains after 24 hours of culture. The same growth curve as Salmonellawas obtained with E coli JM105. Growth of C5 co-cultured with E coli was not different from growth of C5 alone in LB broth. Moreover, using the method described by Portrait and colleagues,11 it was observed that these two strains did not produce colicin, which is active againstS typhimurium C5 (V Portrait, personal communication).

We further investigated if E coli EM0 or JM105 could prevent invasion of S typhimuriumC5 within cultured Caco-2/TC7 cells. For EM0 strain, the inhibition assay of C5 cell invasion was conducted in the presence of dextran 4 for osmotic protection against the pore forming lesions induced by the E coli EM0 haemolysin.39 Under these experimental conditions,E coli EM0 haemolysin induced LDH release of cells was strongly reduced (10% of LDH release versus 100% in control conditions). In the presence of dextran 4, penetration ofS typhimurium C5 within Caco-2/TC7 cells was identical to that obtained in control conditions: 6.7 (0.1) (mean (SEM) log10 cfu/ml for seven trials). The results show that none of theE coli strains (EM0 and JM105) exerted an inhibitory effect against invasion of C5 within Caco-2/TC7 cells. The level of C5 invasion was 6.6 (0.1) and 6.6 (0.2), respectively, in the presence of JM105 and EM0.


Our results showed that virulent and non-virulentE coli strains established in the intestine of germfree mice exerted a protective effect againstSalmonella infection. In contrast, no protective effect was found in the conventional mouse model, suggesting that the resident microflora by itself exerts a protective activity and does not allow the display of the protective activity exerted byE coli. These results are different from those recently reported by us showing that the L casei GG and L johnsonii La1 strains established in the gut of germfree mice delayed the mortality of mice orally infected with S typhimurium. In the conventional mouse model, elimination ofSalmonella from the gut was accelerated in groups treated daily with the GG or La1 cultures20 42compared with the control group. TheseLactobacillus strains produced an antibacterial substance(s) which was active in vitro againstSalmonella.41-43 However, the mechanism by which these Lactobacillusexerted a protective effect in vivo against bacterial infection has not yet been elucidated. It has been reported that someE coli strains secrete antimicrobial substances termed microcins and colicins which are active against Gram negative bacteria10 such as S typhimurium. 11 Barrow et al showed that some strains of Enterobacteriaceae such asE coli,Citrobacter,Klebsiella, andSalmonella can induce growth suppression of their isogenic antibiotic resistant mutants in early stationary phase LB broth cultures.8 This taxon specific inhibition was not related to lack of nutrients and was not the result of lysogenic bacteriophage or bacteriocin activity as the strains were isogenic.7 It was suggested that the inhibition was supported by a diffusible but labile chemical factor. When examining if the EM0 and JM105 strains produced an antimicrobial substance(s) active in vitro against S typhimurium, we found that there was no inhibitory activity of the twoE coli strains in the co-culture test and diffusion test. As a consequence, we propose that the protective effect developed by these E coli in monoassociated animals could be due to mechanisms other than production of an inhibitory substance.

The protective effect induced by the in vivo establishedE coli in germfree mice could be related to occupancy of Salmonella attachment sites as attachment is a prerequisite step before invasion. Indeed, we observed that the two E coli strains efficiently colonised the gastrointestinal tract of germfree mice. As a consequence, the established E coli could form an efficient biofilm of bacteria protecting the intestinal epithelium against attachment of S typhimurium, thus reducing the invasion step. However, our results showed that the number of E coliassociated with the intestinal mucosa was 100 times lower than the number of E coli in the luminal contents, indicating that there is no attachment of the bacteria to the intestinal wall itself.12 44 E colidetected on the washed intestinal wall probably corresponded to the bacteria embedded in the mucus layer, as observed by others.45 Moreover, using a cellular model of human enterocyte-like Caco2/TC7 cells which could be infected byS typhimurium, we found that the twoE coli strains had no effect on the invasion capacity of S typhimurium. The protective effect of E coli strains cannot be explained by competition for the invasion sites of S typhimurium.

E coli EM0 was described as a non-virulent strain as it was recovered from a healthy human donor and has been administered to newborns.13 In this study, we showed that EM0 had the genes of two virulence factors (haemolysin and CNF) whereasE coli K-12 JM105 does not. However, in our experiments, no adverse effect was observed in germfree mice colonised by this strain, as already shown by other authors in germfree piglets and in mice.9 13 22 The insertion sites of these toxins could also be species dependent and absent in the intestine of C3H/He/Oujco mice. There is probably no direct interaction in situ between bacteria and enterocytes which would prevent the effect of haemolysin or CNF, since as mentioned above, the twoE coli strains do not adhere to the mucosa. Interestingly, it has been shown previously that 6–22% of faecalE coli strains from healthy donors have haemolytic activity33 46 without producing pathology in the donors.

Another mechanism usually attributed to probiotics or to bacterial interactions in vivo is competition for nutrient sources. The following results obtained in chickens7 47 48 and pigs45 show a protective effect againstSalmonella intestinal colonisation by previous oral inoculation (one day before) of an avirulent mutant strain of Salmonella. As they observed it in vitro, the inhibitory effect appears to work across the serovar boundary but is more efficient with antibiotic resistant isogenic strains.45 48 The mechanism involved is related to competition for nutrients or electron receptor as resistance obtained is non-immunological and colonisation by the second bacteria is partially or totally reduced48 depending on the mutation of the first strain. Our results showed that mice were protected against clinical disease despite the absence of inhibition of colonisation of Salmonella except that there was a significant delay in the establishment of C5 during the two first days after C5 oral inoculation. This delay may be the result of adaptation of the Salmonella strain to growth conditions in the presence of E colibut competition does not lead to elimination of the strain. Filho-Limaet al showed recently that EM0 associated to probiotics (L acidophilus andSaccharomyces boulardii) was necessary to intestinal elimination of an antibiotic resistant mutant ofS flexneri but had no effect on intestinal levels of S enteritidissubspeciestyphimurium.22

Our results showed protection against systemic infection and lower contamination at remote sites such as the liver and spleen without inhibition of intestinal colonisation. These results are similar to those of Barrow et al (Prevention of salmonellosis in gnotobiotic pigs by precolonisation with avirulentSalmonella strains. “Salmonella and Salmonellosis”.Proceedings of Ploufragan, 20–22 May 1997:519–20). They showed a protective effect againstS typhimurium infection by inoculating gnotobiotic pigs with an avirulent strain of S infantis. Gut colonisation by S typhimurium was delayed for five days in pretreated animals which survived to oral infection. In these surviving pigs, translocation rates of S typhimurium in MLN, liver, and spleen were decreased significantly as we observed in surviving mice pretreated with E coli(tables 3, 4). Diminution of C5 translocation rate observed here in the MLN could be a consequence of diminution of the intestinal colonisation rate in situ, as Berg and Owens described in conventional mice that the level of translocation to the MLN is related to intestinal levels greater than 108/g.49 Diminution ofSalmonella at remote sites (liver and spleen) suggests an immunological mechanism. Rapid intestinal colonisation of E coli one week before infection and delay in S typhimuriumestablishment could allow progressive development of an immunological mechanism of defence. We showed that both EM0 and JM105 strains translocated from the gastrointestinal tract to the MLN, as found previously by others for indigenous E colistrains in mice50 51 and piglets.45 52 The clinical significance of translocation of indigenous bacteria and its role in priming the host immune response to improve host defences against overt or opportunistic pathogen(s) is not yet known.50 Some authors have reported thatE coli participates in the establishment of the population of IgA plasmocytes in the lamina propria in adult germfree mice, in suckling mice,53 and in germfree piglets52 to the level obtained in conventional animals.E coli EM0 has several immunological properties whereas those of E coli K-12 have not yet been well studied. In a germfree mice model,25 EM0 stimulated secretion of the cytokines interleukin (IL)-1 and IL-6 by peritoneal macrophages, to the level observed in conventional mice and cytokine production of bone marrow derived macrophages indicating an effect on macrophage precursors.24 Other strains ofE coli had a similar effect in germfree mice exposed to these strains: stimulation of IL-6 and tumour necrosis factor (TNF).54 Production of TNF-α was also stimulated by the EM0 strain in gnotobiotic mice.25 This cytokine in conjunction with interferon γ are the main cytokines implicated in the first steps of Salmonellainfection.55 As a consequence, it can be hypothesised that concerning infection of C3H/He/Oujco mice by S typhimurium, the presence of an establishedE coli in vivo could stimulate a non-specific immunological response.

Further experiments are required to establish the mechanism by which these two virulent and avirulent E colistrains directly affect Styphimurium pathogenicity. In particular, the immunological aspects of the protective role ofE coli should be studied.


We thank V Portrait for colicin determination, C Sandré for revision and council in redaction, and M Métioui for his revision of the English language of the manuscript. J Guignot is supported by a doctoral fellowship from the Ministère de l'Education Nationale, de la Recherche et de la Technologie (MENRT).


View Abstract


  • Abbreviations used in this paper:
    LB broth
    Luria-Bertoni broth
    cytolethal distending toxin
    phosphate buffered saline
    tryptic soy agar
    SS agar, Shigella and Salmonella agar
    PCR, polymerase chain reaction
    lactate dehydrogenase
    cytotoxic necrotising factor
    mesenteric lymph nodes
    tumour necrosis factor α

Request permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.