Effect of Fructooligosaccharide Metabolism on Chicken Colonization by an Extra-Intestinal Pathogenic Escherichia coli Strain

Extra-intestinal pathogenic Escherichia coli (ExPEC) strains cause many diseases in humans and animals. While remaining asymptomatic, they can colonize the intestine for subsequent extra-intestinal infection and dissemination in the environment. We have previously identified the fos locus, a gene cluster within a pathogenicity island of the avian ExPEC strain BEN2908, involved in the metabolism of short-chain fructooligosaccharides (scFOS). It is assumed that these sugars are metabolized by the probiotic bacteria of the microbiota present in the intestine, leading to a decrease in the pathogenic bacterial population. However, we have previously shown that scFOS metabolism helps BEN2908 to colonize the intestine, its reservoir. As the fos locus is located on a pathogenicity island, one aim of this study was to investigate a possible role of this locus in the virulence of the strain for chicken. We thus analysed fos gene expression in extracts of target organs of avian colibacillosis and performed a virulence assay in chickens. Moreover, in order to understand the involvement of the fos locus in intestinal colonization, we monitored the expression of fos genes and their implication in the growth ability of the strain in intestinal extracts of chicken. We also performed intestinal colonization assays in axenic and Specific Pathogen-Free (SPF) chickens. We demonstrated that the fos locus is not involved in the virulence of BEN2908 for chickens and is strongly involved in axenic chicken cecal colonization both in vitro and in vivo. However, even if the presence of a microbiota does not inhibit the growth advantage of BEN2908 in ceca in vitro, overall, growth of the strain is not favoured in the ceca of SPF chickens. These findings indicate that scFOS metabolism by an ExPEC strain can contribute to its fitness in ceca but this benefit is fully dependent on the bacteria present in the microbiota.


Introduction
Extra-intestinal pathogenic E. coli (ExPEC) strains are responsible for a wide range of diseases in humans and animals. These strains have been isolated either from urinary tract infections, neonatal meningitis, cases of septicaemia of various origins, pneumonia, deep surgical wound infections and mastitis, or from colibacillosis in poultry, a systemic infection that starts in the respiratory tract [1][2][3][4][5]. This respiratory disease is characterized by fibrinopurulent lesions of internal organs such as air sacculitis, perihepatitis and pericarditis and is often associated with septicaemia and mortality [6,7]. These extra-intestinal diseases represent a serious economic, medical and veterinary burden [8].
ExPEC strains can asymptomatically colonize the intestinal tract of humans and animals as commensal bacteria. Consequently, the intestine serves as a reservoir for pathogenic strains. Intestinal colonization by ExPEC is thus a potential risk factor for a subsequent extra-intestinal infection in the same host or for dissemination of pathogens in the environment, thus leading to a potential zoonotic risk [9][10][11][12][13][14]. The establishment of ExPEC in the intestine appears to play an important role in their future establishment in the urinary or respiratory tract. For instance, strains involved in urinary tract infections gain access to the periurethral area from the anus and establish infection in an ascending manner [15]. Strains involved in neonatal meningitis could translocate from the intestinal lumen of the neonate to the blood stream, and poultry inhale pathogenic E. coli in dust derived from faeces [6,7,16].
ExPEC strains mainly belong to the phylogenetical lineage B2 [2,17]. It has been shown that strains belonging to this phylogenetic group have a greater ability to persist in the intestinal tract of healthy or infected people [18][19][20][21]. It has been suggested that virulence factors of ExPEC such as adhesins, or iron acquisition factors could confer a higher capacity to colonize their reservoir [18,19,22]. For example, the K5 capsule and P fimbriae enhance intestinal colonization in gnotobiotic rats [23,24]. Moreover, the frz operon of the avian ExPEC strain BEN2908 and the pathogenicity islands of the uropathogenic strain 536 are fitness elements involved in intestinal colonization [25,26].
Maintenance of intestinal colonization thus requires many properties, one of the most important being metabolic competence, in addition to virulence factors. When two strains compete for a limited nutrient, the one that is able to use it more efficiently should outcompete the other [26,27].
We have recently shown that the ability of the strain BEN2908 to metabolize short-chain fructooligosaccharides (scFOS) enhances colonization of the chicken intestine by bacteria during the first 8 days post-inoculation [28]. scFOS are natural linear polymers comprising two to four b-(2-1)-linked fructosyl units, usually attached to a terminal glucose residue [29]. Like many complex plant carbohydrates, these sugars are not hydrolyzed by digestive enzymes, and therefore they reach the distal parts of the intestine intact where they are assimilated by the intestinal microbiota, particularly probiotic bacteria [30,31].
The genomic region responsible for scFOS metabolism in the BEN2908 strain, called the fos locus (GenBank accession no. AY857617), is found on the AGI-3 pathogenicity island [32]. This locus is composed of six genes organized as an operon encoding for a putative MFS (Major Facilitator Superfamily) sugar transporter (FosT), two glycoside hydrolases of family 32 (FosE 1 and FosE 2 ), two proteins of unknown function (FosX and FosY), a fructokinase (FosK), and of a divergently transcribed gene encoding for a putative transcriptional regulator of the LacI/ GalR family (FosR). We previously defined a regulatory model of scFOS metabolism in BEN2908 [33]. In the absence of scFOS, FosR is able to bind to the promoter of the fos operon on two operator sequences, suppressing fos gene expression. Moreover, fos gene expression relies on catabolite repression, and the presence of glucose represses this expression. It has also been shown that fos gene expression depends on the presence of scFOS in the medium.
Due to the presence of the fos locus of the ExPEC strain BEN2908 on the AGI-3 pathogenicity island, one aim of this study was to investigate a possible role of these genes in the virulence of BEN2908 for chickens by analysing fos gene expression in extracts of target organs of avian colibacillosis and by performing a virulence assay in chickens. Moreover, in order to study the involvement of the fos locus in the colonization of the strain's reservoir, we analyzed fos gene expression and the implication of these genes in the growth ability of BEN2908 in different intestinal extracts. Finally, we performed intestinal colonization assays in axenic and SPF (Specific Pathogen-Free) chickens. We found that the fos locus is not involved in the virulence of BEN2908 for chickens and that, even if this locus is strongly involved in axenic chicken intestinal colonization in the ceca, it does not significantly contribute to cecal colonization of SPF chickens.

Results
The fos locus is not involved in virulence As the fos locus is located on the AGI-3 pathogenicity island [32], we investigated a possible role of this locus in the virulence of the BEN2908 strain for chicken. Firstly, we examined fos gene expression in minimal media containing extracts of target organs of avian colibacillosis including the lung, liver, spleen and pericardial fluid. As the average body temperature of chicken is 41.5uC [34], we first checked the growth ability of the strain BEN2908 in M9 minimal medium containing scFOS at 41.5uC. Surprisingly, this strain was impaired in its ability to grow in this minimal medium at 41.5uC compaired to 37uC (Fig. 1). This impairment was also observed in M9 minimal medium containing glucose but not in a more complex medium such as LB-Miller medium (data not shown). As growth is altered in M9 minimal medium at 41.5uC, we thus performed the experiments at 37uC.
Expression of fos genes was monitored by measuring the fos promoter-mediated expression of luciferase from plasmid pQF52 during the growth of the BEN2908, BEN2908DfosT (unable to metabolize scFOS) and BEN2908DfosR (unable to repress fos gene expression) strains. As shown in Fig. 2, luciferase expression in BEN2908 and BEN2908DfosT was always lower than in BEN2908DfosR in all the media tested (5-to 31-fold higher in BEN2908DfosR than in BEN2908 in the lung, 3-to 14-fold higher in the liver, 14-to 36-fold higher in the spleen, and 3-to 14-fold higher in pericardial fluid; 5-to 17-fold higher in BEN2908DfosR than in BEN2908DfosT in the lung, 2-to 11-fold higher in the liver, 7-to 22-fold higher in the spleen, and 2-to 12-fold higher in pericardial fluid). This indicates that FosR repressed fos gene expression in these organ extracts. Nevertheless, fos genes were expressed slightly more in both BEN2908 and BEN2908DfosT in minimal media containing liver extract and pericardial fluid (a maximum of 1.2260.11610 6 and 1.7460.1610 6 RLU/OD 450 in the liver, 160.42610 6 and 1.3560.3610 6 RLU/OD 450 in pericardial fluid, versus 3.3460.4610 5 and 7.0760.7610 5 RLU/OD 450 in the lung and 2.760.3610 5 and 5.3861.19610 5 RLU/OD 450 in spleen extracts, respectively) ( Fig. 2). This expression was not due to the presence in the media of sugars metabolized by the fos locus because it was not higher in BEN2908 than in BEN2908DfosT and there was no difference in the growth of these strains. We then verified that the fos locus, whose genes were slightly expressed in a minimal medium containing liver extract, did not give a growth advantage to strain BEN2908 in the presence of this colibacillosis target organ. Before, we checked that the impairment of the BEN2908DfosT strain in scFOS metabolism was only due to deletion of the fosT gene. We then introduced in this strain a pGEM-T derivative containing the whole fos locus [28]. As shown in Fig. 3, the recombinant BEN2908DfosT strain was able to grow with scFOS as the sole carbon source. However, this plasmid is unstable in M9 minimal medium containing organ extracts (70-80% of the recombinant bacteria maintain the plasmid during the growth, data not shown). As we performed co-cultures to determine the implication of the fos locus in the growth on such media, this instability did not allow us to realize complementation assays in the presence of organ extracts. To check the involvement of the fos locus in the presence of liver extract, BEN2908 and BEN2908DfosT were inoculated together in equal amounts in a minimal medium containing liver extract, and the proportion of each strain was monitored during their growth. As shown in Fig. 4, BEN2908 did not outcompete BEN2908DfosT in the presence of liver extract.
While these results strongly suggest that the fos locus is not involved in virulence, the expression of fos genes was not studied in blood and air sacs, which are other target organs/fluids of colibacillosis. We thus examined the impact of the fos locus on the virulence of the BEN2908 strain in vivo. To that end, equal amounts of BEN2908 and BEN2908DfosT were co-inoculated into the air-sac of SPF chickens (5610 6 CFU of each strain/chicken), and the proportion of each strain in organs was monitored. As shown in Fig. 5, there was no difference in the colonization of target organs of avian colibacillosis (lung, thoracic air sac, liver, spleen and pericardial fluid) by BEN2908 or BEN2908DfosT, as the competitive indexes did not differ significantly from 1. Overall, these results demonstrate that the fos locus is not involved in the virulence of BEN2908 for chicken.
fos genes are expressed in chicken cecal content and not in intestinal mucus The involvement of the fos locus in chicken intestinal colonization has previously been demonstrated by monitoring the proportion of BEN2908 and BEN2908DfosT in faeces [28]. However, identifying viable bacteria in faeces does not enable the site(s) along the gastrointestinal tract in which the fos locus is important to be determined. We thus investigated whether fos genes were expressed in different parts of the distal intestine, i.e. in intestinal mucus collected from the ileum or colon and in cecal content of axenic chickens. As shown in Fig. 6A and B, luciferase expression in intestinal mucus from the ileum and colon was much lower in BEN2908 and BEN2908DfosT than in BEN2908DfosR (7.5-to 12-fold higher in BEN2908DfosR than in BEN2908 in the ileum, and 7-to 12.5-fold higher in the colon; 5-to 7-fold higher in BEN2908DfosR than in BEN2908DfosT in the ileum, and 5.5-to 7fold higher in the colon). This indicates that no inducer was present in these media to lift the FosR repression and to allow fos gene expression. In cecal content, luciferase expression in BEN2908 and BEN2908DfosT was much higher (a maximum of 2.360.25610 6 and 1.5360.2610 6 RLU/OD 450 in cecal content, respectively, versus 5.3260.49610 5 and 9.1962.21610 5 RLU/ OD 450 in mucus from the ileum and 7.0361.45610 5 and 1.0560.4610 6 RLU/OD 450 in mucus from the colon, respectively), although lower than expression in BEN2908DfosR (1.5-to 7fold higher, and 1.5-to 11-fold higher in BEN2908DfosR, respectively) (Fig. 6C). This indicates that expression in BEN2908 and BEN2908DfosT was not fully activated. Moreover, growth of BEN2908DfosT was less than that of BEN2908 and BEN2908DfosR and luciferase expression was significantly higher in BEN2908 than in BEN2908DfosT at different times ( . This suggests that inducers of fos gene expression are present in cecal content to support BEN2908 growth and that this strain is able to metabolize cecal nutrients via the fos locus. For a better comparison of activation of fos gene expression under all the conditions tested, we calculated the ratio of maximal luciferase activity in BEN2908 to the maximal luciferase activity in BEN2908DfosR. As shown in Fig. 7, maximal activation of the fos promoter was observed in cecal content (54.767.3%). The fos promoter was less activated in the presence of mucus from the ileum and colon (9.560.4% and 10.662.2%, respectively) than in the liver (23.761.4%), and was activated to the same extent in the presence of pericardial fluid (13.965.1%). It was activated least in the presence of lung and spleen extracts (5.860.5% and 6.561.5%, respectively).
The fos locus is involved in growth on cecal content from axenic chicken in vitro and in vivo As fos genes were expressed in cecal content, we then investigated whether the fos locus benefited BEN2908 growth in cecal content. We also investigated a possible role of this locus in the growth of BEN2908 on intestinal mucus from the ileum and colon. Indeed, although fos genes were not expressed in these media, BEN2908 grew better than BEN2908DfosT in these biological extracts when monitored independently during expression analyses (Fig. 6). To monitor the involvement of the fos locus in the growth of BEN2908 in the presence of samples from the   . Competition between E. coli strains BEN2908 and BEN2908DfosT for colonization of chicken organs during avian colibacillosis. Twenty five-day-old SPF White Leghorn chickens were inoculated in the right air sac with strains BEN2908 and BEN2908DfosT together (each at 5610 6 CFU/chicken). Animals were euthanized 48 h post-inoculation by injection of Nesdonal and then necropsied. The proportion of each strain in organs was monitored and CI were calculated. Horizontal red bars indicate the median of CI and diamonds indicate individual CI. Statistical analyses were conducted using the Mann-Whitney U-test, measuring the difference between CI in organs and in the inoculum. The calculated P values are presented. doi:10.1371/journal.pone.0035475.g005 distal intestine, equal amounts of BEN2908 and BEN2908DfosT were inoculated together into a minimal medium containing either intestinal mucus from the ileum or colon, or cecal content from axenic chickens. In the medium containing intestinal mucus from the ileum, there was no difference between BEN2908 and BEN2908DfosT during the exponential growth phase. From six hours of growth, strain BEN2908 had a significant advantage over BEN2908DfosT [8.5960.04 and 8.4360.05 log 10 CFU/ml at 6 hours (p = 0.017); 8.6360.02 and 8.4860.06 log 10 CFU/ml at 7 hours (p = 0.015); 8.7360.06 and 8.5660.06 log 10 CFU/ml at 8 hours (p = 0.023), respectively]. However, at 24 hours of growth, this difference was no longer significant (Fig. 8A). In the medium containing intestinal mucus from the colon, there was no significant difference between BEN2908 and BEN2908DfosT during the entire growth curve (Fig. 8B). Finally, in the medium containing cecal content, there was a significant difference from five hours of growth with a distinct advantage of BEN2908 over BEN2908DfosT [8.2160.06 and 8.0360.02 log 10 CFU/ml at 5 hours (p = 0.008); 8.2860.02 and 8.1460.03 log 10 CFU/ml at 6 hours (p = 0.002); 8.3860.03 and 8.1460.02 log 10 CFU/ml at 7 hours (p = 0.0003); 8.4460.01 and 8.1960.04 log 10 CFU/ml at 8 hours (p = 0.0003); 8.6260.02 and 8.260.01 log 10 CFU/ml at 24 hours (p = 9.9610 26 ), respectively] (Fig. 8C). These results indicate that the fos locus confers a slight growth advantage to BEN2908 in ileal mucus and a more marked advantage in cecal content from axenic chicken.
To confirm this involvement in this intestinal compartment, we performed an in vivo experiment. BEN2908 and BEN2908DfosT were fed together in equal amounts to axenic chickens (5610 7 CFU of each strain/chicken), and the proportion of each was monitored in ceca. The results showed that the BEN2908 strain had a strong advantage in cecal colonization compared to the BEN2908DfosT strain up to eight days post-inoculation (median of CI of 12.73 at 2 days post-feeding; 215 at 3 days; 2382.5 at 6 days and 844.5 at 8 days) (Fig. 9A). This clearly indicates that the fos locus is involved in the colonization of axenic chicken ceca.
The fos locus is involved in growth on cecal content in the presence of a microbiota in vitro but not in vivo All the previous experiments were conducted with biological extracts from axenic chicken and do not represent the real conditions encountered by the bacteria in vivo. Indeed, sugars present in these extracts have not undergone any catabolism or degradation by bacteria. The sugar contents in the chicken intestine with a normal microbiota could be different. We thus analyzed the involvement of the fos locus in the growth of BEN2908 in cecal content from SPF chicken. When BEN2908 and BEN2908DfosT were inoculated together in a minimal medium containing previously sterilized cecal content from SPF chicken, there was no competition between the two strains, indicating that the sugars metabolized via the fos locus had previously been metabolized by the cecal microbiota (Fig. 8D). This also suggests that competition observed in cecal content from axenic chicken is a consequence of sugar metabolism mediated by the fos locus.
In order to avoid the use of sugars metabolized by the fos locus before sampling of cecal content but while looking at whether the presence of cecal microbiota could inhibit the benefit conferred by the fos locus, equal amounts of BEN2908 and BEN2908DfosT were inoculated together into a minimal medium containing cecal content from axenic chicken supplemented with cecal bacteria from SPF chicken. We first verified that these cecal bacteria were able to grow in a minimal medium containing cecal content from axenic chicken, although the experiments were not carried out in anaerobic conditions (8.97 log 10 CFU/ml culturable on LB-Miller medium at 24 h) (data not shown). As shown in Fig. 8E, there was a significant difference after five hours of growth, with BEN2908 having a distinct advantage over BEN2908DfosT [8.3260.03 and 8.260.02 log 10 CFU/ml at 5 hours (p = 0.007); 8.4160.02 and 8.2660.04 log 10 CFU/ml at 6 hours (p = 0.003); 8.4560.03 and 8.360.02 log 10 CFU/ml at 7 hours (p = 0.002); 8.5060.03 and 8.3260.04 log 10 CFU/ml at 8 hours (p = 0.002); 8.7360.02 and 8.6360.004 log 10 CFU/ml at 24 hours (p = 0.0005), respectively]. These results demonstrate that the BEN2908 strain is able to compete with a complex microbiota that also metabolizes the substrate of the fos locus.
We then checked if these results could be observed in vivo. To that end, BEN2908 and BEN2908DfosT were fed together in equal amounts to SPF chickens (5610 7 CFU of each strain/chicken), and the proportion of each was monitored in ceca. We also enumerated the total E. coli population in cecal content of each animal. This population varied from one animal to the other (from 3.41610 5 to 1.24610 8 CFU/g of cecal content) ( Table 2). As shown in Fig. 9B, the BEN2908 strain had a slight advantage, but not significant, in cecal colonization compared to the BEN2908D-fosT strain at 3 days post-feeding but no advantage was observed on other days (median of CI of 1 at 2 days post-feeding; 5.96 at 3 days; 1.24 at 6 days and 0.85 at 8 days) (Fig. 9B). These results indicate that, in the presence of a complex microbiota, the fos locus does not provide a benefit to the BEN2908 strain to colonize the intestine. Figure 6. fos operon promoter activity in extracts of chicken intestine. Growth curves (OD 450 nm ) and relative luminescence intensities (RLU/ OD 450 nm ) of strains BEN2908, BEN2908DfosR and BEN2908DfosT carrying pQF52 grown without shaking at 37uC in M9 minimal medium supplemented with (A) intestinal mucus from the ileum at 2 mg/ml of proteins, (B) intestinal mucus from the colon at 2 mg/ml of proteins, (C) 2% of cecal content. The RLU average values and standard deviations result from three independent experiments. Asterisks indicate significant differences in mean luciferase activity between BEN2908 and BEN2908DfosT determined by a Student's t-test. *** P,0.005; ** P,0.02; * P,0.05. doi:10.1371/journal.pone.0035475.g006

Discussion
In this study, we demonstrated both in vitro and in vivo that the E. coli strain BEN2908 had a strong growth advantage in cecal content from axenic chicken over a strain that does not metabolize scFOS (Fig. 6C, 8C and 9A). In the presence of a microbiota in cecal content, we observed that, in vitro, the ability to metabolize scFOS gave significant competitive advantage to the BEN2908 strain (Fig. 8E). However, in vivo in SPF chickens, overall, no significant competitive advantage was observed but it is notable Figure 8. Competition between E. coli strains BEN2908 and BEN2908DfosT to grow in the presence of extracts of chicken intestine. Growth curves (log 10 CFU/ml) of strains BEN2908 (pink circle) and BEN2908DfosT (blue triangle) grown without shaking at 37uC in M9 minimal medium supplemented with (A) intestinal mucus from the ileum at 2 mg/ml of proteins, (B) intestinal mucus from the colon at 2 mg/ml of proteins, (C) 2% of cecal content from axenic chicken, (D) 2% of cecal content from SPF chicken, (E) 4% of cecal content from axenic chicken supplemented with cecal bacteria from SPF chicken. The average values and standard deviations result from three independent experiments. Asterisks indicate significant differences in mean number of CFU between BEN2908 and BEN2908DfosT determined by a Student's t-test. *** P,0.005; ** P,0.02; * P,0.05. doi:10.1371/journal.pone.0035475.g008 that competitive indexes considerably varied from one animal to another (Fig. 9B). As the cecum is an intestinal site where nondigestible carbohydrates such as cellulose, inulin and FOS are assimilated by the microbiota [35,36], we can assume that the substrates giving a growth advantage to BEN2908 are scFOS. These results thus suggest that scFOS metabolism by the BEN2908 strain could contribute to its implantation in ceca. Nevertheless, one of the most important factors determining the nutrients found in the gastrointestinal tract is the presence of the intestinal microbiota. The number of bacteria found in the ceca of chicken is approximately 10 11 CFU/g, including Clostridiaceae, Sporomusa, Enterobacteriaceae, Fusobacterium, Bacteroides, Lactobacillus, Streptococcus, Ruminococcus and Bifidobacterium [37][38][39][40]. The presence of these bacteria allows extensive bacterial fermentation, resulting in further nutrient absorption, detoxification of harmful substances and prevention of pathogen colonization [35,38,41]. Among these bacteria, some strains of Lactobacillus, Bifidobacterium, Bacteroides and Fusobacterium prausnitzii are able to metabolize nutrients such as FOS [37,[42][43][44][45][46][47]. These strains can therefore compete with the BEN2908 strain to metabolize FOS nutrients in vivo. We have previously shown that the fos locus of the BEN2908 strain was able to give an advantage to this strain to colonize the intestine of SPF chickens [28]. In this latter experiment, SPF chickens were obtained by inoculating axenic chickens with a complex bacterial inoculum consisting of a suspension of faeces from adult SPF hens that have been conserved during several years at 280uC. In this case, it is likely that there were no strict anaerobes in the faeces, therefore any anaerobes inoculated to chickens. Similarly, in the in vitro experiment performed in this study with the cecal content from axenic chickens supplemented with cecal bacteria from SPF chickens, it is possible that strict anaerobes did not survive during sampling of cecal content and/or did not survive in the medium because experiment was not achieve in anaerobic conditions. This suggests that a strict anaerobic bacterium is able to metabolize scFOS more efficiently than E. coli strain BEN2908.
During the intestinal colonization experiment in SPF chickens, we observed that the E. coli population varied from one animal to the other (Table 2). This suggests that, despite the fact that chickens had the same diet and same environmental conditions, the intestinal colonization by E. coli is highly variable. This fluctuation is also observed for the strains BEN2908 and BEN2908DfosT, the proportions of which in this total population were inconstant (from 0.07 to 61.43% for BEN2908 and from 0.01 to 25.62% for BEN2908DfosT) ( Table 2). Finally, we can also observe that, in some animals, the BEN2908 strain was highly dominant in the total E. coli population (61.43%, 43.12%, 50.70% and 56.23% in chickens 75, 78, 83 and 85, respectively) whereas the BEN2908DfosT strain proportion was lower (a maximum of 25.62% in the chicken 94) ( Table 2). This observation suggests that the fos locus, according to the bacteria present in the microbiota, and more particularly in the absence of a particular anaerobic bacterium able to metabolize scFOS most efficiently, is able to contribute to intestinal colonization. It can nevertheless be noted that, in few animals, and especially in chicken 90, the BEN2908DfosT strain is able to outcompete the BEN2908 strain ( Table 2). As we have never observed this trend before, even in minimal media containing only one carbon source, we cannot currently explain this result in these animals.
It is noteworthy that the E. coli strain BEN2908 is fully able to colonize experimentally the chicken intestine, even in SPF animals. Conversely, commensal E. coli colonization cannot be studied experimentally in conventional animals due to colonization resistance, which results when all niches are filled by the microbiota. Streptomycin-treated animals are thus routinely used to study E. coli intestinal colonization [48]. The fact that BEN2908 is able to colonize SPF chicken intestine, without antibiotic treatment and on a long range, strongly suggests that ExPEC strains are better intestinal colonizers than commensal E. coli strains. This is consistent with previous observations demonstrating that strains belonging to phylogenetical lineage B2 have a greater ability to persist in the intestine [18][19][20][21].
In the present study, we investigated fos gene expression in different biological extracts and we showed that fos gene expression in cecal content was high in both the BEN2908 and BEN2908D-fosT strain (Fig. 6C). This result is surprising since we previously showed that fos gene expression depends on the presence of scFOS [with b-(2-1) links] in the medium and that BEN2908DfosT is not able to metabolize these carbohydrates [28,33]. Several hypoth- eses can therefore be considered. Different types of fructans exist: inulin and its derivatives, with one linear b(2-1)-linked fructosyl chain attached to the fructosyl residue of the sucrose starter; neoseries inulin with two linear b(2-1)-linked fructosyl chains, one attached to the fructosyl residue of the sucrose, the other to the glycosyl residue; levan with one linear b(2-6)-linked fructosyl chain attached to the fructosyl residue of the sucrose starter; and graminan with both b(2-1) and b(2-6) types of fructosyl linkages [49]. One interesting hypothesis is that another type of fructan induces fos gene expression. This inducer could enter via a transporter other than FosT, explaining fos gene expression in BEN2908DfosT. Another hypothesis to explain this result could be the action of an activator of fos gene expression. We previously demonstrated that fos gene expression also depends on catabolite repression and the binding of CRP-cAMP complex to the fos promoter region [33]. This binding enhances the ability of RNA polymerase to bind and initiate transcription of fos genes. cAMP synthesis is mediated by adenylate cyclase which is activated by phosphorylated EIIA Glc (IIA component of the glucose-specific phosphoenolpyruvate:carbohydrate phosphotransferase system) [50][51][52][53]. The phosphorylation state of EIIA Glc depends on the [phosphoenolpyruvate]/[pyruvate] ratio and is thus completely dependent on the substrates metabolized by the cell and differs according to the substrate [54]. In this study, fos gene expression in the BEN2908DfosT strain was up to 36% in cecal content and 33% in liver extract (compared to the expression in the BEN2908DfosR strain). Therefore, some substrates found in these media, entering via the FosT transporter in the BEN2908 and BEN2908DfosR strains, could lead to a lower [phosphoenolpyruvate]/[pyruvate] ratio and thus lower concentrations of phosphorylated EIIA Glc and cAMP synthesis in these strains. This could also explain the high levels of expression in the BEN2908DfosT strain compared to the expression in the BEN2908DfosR strain. Another surprising result in this study is that fos gene expression was always lower in BEN2908 than in BEN2908DfosR, even in cecal content ( Fig. 2 and 6). This indicates that in none of the biological extracts tested, corresponding to the conditions encountered by the bacteria in vivo, was there a sufficient inducer concentration to allow a complete lift of FosR repression, whereas previously we observed this situation in vitro [33].
When this study was initiated, the fos locus had not been identified in bacteria other than E. coli and its prevalence was low (only 10 of the 133 E. coli strains tested possessed this locus [28]). The subsequent release of newly discovered genome sequences in databases enabled us to identify a truncated locus similar to the fos locus in the genome of several pathogenic bacteria. Several Klebsiella pneumoniae strains and Klebsiella variicola strain At-22 possess a truncated locus including a transcriptional regulator sharing 77% identity with FosR, an MFS transporter sharing 85% identity with FosT, and a glycosyl-hydrolase sharing 73% identity with FosE 1 . This locus also comprises an intergenic region between the regulator and the transporter that is very similar to that of the fos locus (sharing 54% identity throughout the whole region), including the same regulatory elements of operator 1 and 2 sequences and a CRP-cAMP recognition sequence. Moreover, the Enterobacteriaceae bacterium 9_2_54FAA strain, isolated from inflamed biopsy tissue from a patient with Crohn's disease [Enterobacteriaceae bacterium 9_2_54FAA Sequencing Project, Broad Institute of Harvard and MIT (http://www.broadinstitute.org/)], possesses a truncated locus similar to the fos locus with a transcriptional regulator (73% identity with FosR), an MFS transporter (89% identity with FosT), a glycosyl-hydrolase (67% identity with FosE 1 ), a protein of unknown function, and a fructokinase (66% identity with FosK). It is well known that the gastrointestinal tract of humans and animals is the reservoir of commensal and pathogenic Enterobacteriaceae strains such as Klebsiella [55][56][57][58]. This suggests that other pathogenic bacteria could assimilate scFOS and that this metabolism could be controlled by the same regulation mechanism as that of the fos locus of BEN2908. Finally, we can thus postulate that, like the BEN2908 strain, the ability to metabolize scFOS could also enhance the ability of these pathogens to colonize their reservoir according to the bacteria present in the intestinal microbiota.
In sum, as the importance of the fos locus is dependent on the presence of the microbiota, and particularly on the presence of specific anaerobic bacteria, it could be of interest to study the involvement of this locus on intestinal colonization in different avian lineage and in chicken fed with different diets, thus with different intestinal microbiota, to determine if some conditions are more favorable to a pathogenic strain to colonize its reservoir then leading to greater dissemination of this strain in the local environment, dust for example, and hence colonization of the respiratory tract and pathogenesis.

Ethics statement
The housing, husbandry and slaughtering conditions complied with European Union guidelines for the care and use of laboratory animals. The experimental protocol for experimental colibacillosis was approved by the regional ethics committee [Comité d'Ethique en Expérimentation Animale (CEEA) Val de Loire] under number CL2007-44. The experimental protocol for intestinal colonization of chickens was approved by the regional ethics committee (CEEA Val de Loire) under number CL2007-43.

Bacterial strains, plasmids and growth conditions
Bacterial strains and plasmids used in this study are listed in Table 1.
E. coli strain BEN2908, O2:K1:H5 is a nalidixic acid-resistant derivative of strain MT78 isolated from the trachea of a chicken with a respiratory infection [59,60]. The fosT and fosR isogenic mutants of BEN2908 were also used in this study [28,33]. Strains were routinely grown in LB-Miller medium at 37uC with agitation [61]. When necessary, antibiotics were used at the following concentrations: nalidixic acid 30 mg/ml, ampicillin 100 mg/ml, or kanamycin 50 mg/ml. For growth monitoring experiments, overnight LB-Miller cultures of strains were washed twice and resuspended in the same volume of M9 minimal medium [61]. The strains were then inoculated to an optical density (OD) at 450 nm of 0.05 and cultured at 37uC or 41.5uC with agitation in 20 ml of M9 medium supplemented with 0.2% scFOS (Profeed P95; Beghin Meiji, France). scFOS powder is a mixture containing small quantities of glucose, fructose and sucrose (5%), and larger amounts of kestose (37%), nystose (63%) and fructofuranosyl nystose (10%).
For expression analyses, overnight LB-Miller cultures of strains carrying the pQF52 plasmid [33] were washed as described previously. The strains were then inoculated to an OD at 450 nm of 0.05 and cultured at 37uC without agitation in 20 ml of M9 medium supplemented with either 2% of cecal content, 4% of liver or spleen extract, or 10% of lung extract, in 2.5 ml of M9 medium supplemented with 4% of pericardial fluid, or in 1 ml of M9 medium supplemented with intestinal mucus adjusted at a final protein concentration of 2 mg/ml as described by Edelman et al. [62]. In these experiments, biological extracts were collected from axenic chickens. For media containing pericardial fluid and intestinal mucus, bacterial growth was monitored by plating serial dilutions onto LB-agar plates supplemented with ampicillin at 100 mg/ml at different times.
For co-cultures of BEN2908 and BEN2908DfosT, overnight LB-Miller cultures were washed as described previously. The strains were then inoculated in equivalent numbers (each corresponding to an OD at 450 nm of 0.025) and cultured at 37uC without agitation in 10 ml of M9 medium supplemented with either 2% of cecal content from axenic or SPF chicken, 4% of cecal content from axenic chicken supplemented with cecal microbiota of SPF chicken, or 4% of liver extract, or in 0.3 ml of M9 medium supplemented with intestinal mucus adjusted to a final protein concentration of 2 mg/ml. At different times, serial dilutions were plated onto LB-agar plates supplemented with nalidixic acid at 30 mg/ml (selection of BEN2908 and BEN2908DfosT strains) or with kanamycin at 50 mg/ml (selection of the BEN2908DfosT strain) for bacterial quantification. The number of CFU of BEN2908 was calculated by subtracting the number of kanamycin-resistant bacteria from the number of nalidixic acid-resistant bacteria.
In all the experiments, the media supplemented with the biological extracts were vigorously vortexed, centrifuged for 5 min at 4,0006 g, and supernatants were sterilized by filtration.
For the experiment with cecal bacteria, 100 ml of M9 minimal medium containing cecal microbiota were inoculated in 10 ml of M9 minimal medium supplemented with cecal content from axenic chickens (representing 5610 5 CFU/ml, culturable on LBagar medium).  [59,60] BEN2908DfosT Isogenic deletion mutant of BEN2908; Nal r Kan r Fos 2 [28] BEN2908DfosR Isogenic deletion mutant of BEN2908; Nal r Fos + [33] Plasmids pQF52 oripMB1 oripRO1600 Amp r ; luc under the control of the fos promoter [33] pGEM::fos pGEM-T easy vector containing the whole fos locus; Amp r [28] doi:10.1371/journal.pone.0035475.t001 Table 2. Proportion of the BEN2908 and BEN2908DfosT strains in the total E. coli population in the cecal content of each SPF chicken. Collection of chicken organs, pericardial fluid, cecal content, intestinal mucus and cecal microbiota Liver, lung, spleen, pericardial fluid, cecal content and intestinal mucus were collected from 25-day-old axenic PA12 White Leghorn chickens from the Institut National de la Recherche Agronomique experimental platform for infectious diseases. Axenic PA12 White Leghorn chickens were obtained using the method described by Le Bars [63]. Cecal content was also collected from 25-day-old SPF PA12 White Leghorn chickens. None antibiotic resistant E. coli strain was present in the intestine of SPF chickens. Animals were euthanized by Nesdonal (Rhône-Mérieux, Lyon, France) injection in the occipital sinus. Organs were collected and homogenized in sterile saline. Pericardial fluid was collected with a Pasteur pipette and ceca were collected and drilled to recover cecal content. Aliquots containing extracts of organ, pericardial fluid or cecal content from axenic chicken were stored at 220uC. Mucus was isolated from the ileum and from the colon of chicken as described previously [64,65]. The animals were fasted 20 h before isolating intestinal mucus. Briefly, colon and ileum were collected and gently rinsed with PBS to remove intestinal content. Mucus was isolated from the intestinal walls by gentle scraping with the back of a scalpel, diluted 1:3 with 25 mM HEPES (pH 7.2 to 7.5; Invitrogen) and vigorously vortexed. Epithelial cells and large cell components were removed by centrifugation at 11,0006 g for 10 min at 4uC. The supernatant was then centrifuged at 26,0006g for 15 min at 4uC, and aliquots of the supernatant containing the crude mucus were stored at 220uC. The protein concentration of the mucus preparation was determined using a Bradford protein assay (Biorad).
Cecal content from SPF chicken was preserved in glycerol at 280uC to conserve cecal bacteria. To collect cecal microbiota, this cecal content was washed to remove cecal matter while maintaining cecal bacteria. To that end, one volume of cecal content was first diluted 1:10 in M9 minimal medium and centrifuged 1 min at 2006 g. Supernatant was recovered, centrifuged for 1 min at 3006 g, then 2 min at 3006 g. Finally, the supernatant was centrifuged for 2 min at 16,0006 g and the pellet was resuspended with the same volume of M9 minimal medium.

Luciferase measurements
Promoter activities of the fos operon in different media were determined by firefly luciferase expression along the growth curve as described previously [33]. Briefly, samples of 100 ml were taken every 45 min, and light emission (relative light unit, RLU) was recorded with a luminometer (Lumat LB 9507, Berthold). A luciferase Assay System kit (Promega) was used with some modifications. Samples were incubated with 300 ml of freshly lysed buffer [16 CCLR (Cell Culture Lysis Reagent, Promega), 1.25 mg/ml lysozyme (Sigma), 2.5 mg/ml BSA (Sigma)] for 10 min with agitation at room temperature. Solutions were quick-frozen in liquid nitrogen and then immediately incubated at 37uC. After thawing, samples were incubated for 10 min with agitation at room temperature. Finally, RLU was measured by incubating 25 ml of cell lysate with 50 ml of Luciferase Assay Reagent.
To compare levels of luciferase expression, a correlation curve between OD at 450 nm and the number of bacterial CFU was plotted for each strain. Numbers of CFU obtained in media containing pericardial fluid and intestinal mucus were then converted into OD values to normalize results.

Experimental colibacillosis
An in vivo virulence assay was conducted as described previously with some modifications [32]. SPF chickens were obtained from the Institut National de la Recherche Agronomique experimental platform for infectious diseases. Fifteen 25-day-old SPF PA12 White Leghorn chickens were inoculated in the right thoracic air sac with a 0.1 ml suspension containing a mixture of equal numbers of BEN2908 and BEN2908DfosT (each approximately at 5610 6 CFU) in LB-Miller medium. The inoculum was prepared from an overnight culture of each strain grown in 20 ml LB-Miller medium at 37uC without agitation. Animals were euthanized 48 h post-inoculation by injection with Nesdonal (Rhône-Mérieux, Lyon, France) in the occipital sinus and necropsied. A swab of the left thoracic air sac was taken, and samples of the left lung, liver, spleen and pericardial fluid were collected. After homogenization in sterile saline, serial dilutions were plated onto Drigalski agar plates (Biorad) supplemented with nalidixic acid at 30 mg/ml (selection of BEN2908 and BEN2908DfosT strains) or with kanamycin at 50 mg/ml (selection of the BEN2908DfosT strain) for bacterial quantification. The number of CFU of BEN2908 was calculated by subtracting the number of kanamycin-resistant bacteria from the number of nalidixic-resistant bacteria. Competition indices (CI) were calculated following Freter et al.'s method using BEN2908 as the reference strain [CI = (number of BEN2908 CFU/number of BEN2908DfosT CFU)/(number of BEN2908 CFU/number of BEN2908DfosT CFU in the initial inoculum)] [66]. By definition, a CI of .1 indicates out-competition of the mutant strain (BEN2908DfosT) by the wild-type reference strain (BEN2908). A CI equal to 1 indicates no difference in colonization of organs, and a CI of ,1 indicates out-competition of the wildtype reference strain (BEN2908) by the mutant (BEN2908DfosT).

Intestinal colonization of chickens
An intestinal colonization assay was conducted as described previously with some modifications [28]. Briefly, 16 axenic and 20 SPF 11-day-old PA12 White Leghorn chickens from the Institut National de la Recherche Agronomique experimental platform for infectious diseases were fed with a 0.5 ml suspension containing a mixture of equal numbers of BEN2908 and BEN2908DfosT (each approximately at 5610 7 CFU) in LB-Miller medium. None antibiotic resistant E. coli strain was present in the intestine of SPF chickens. The inoculum was prepared from an overnight culture of each strain grown in 20 ml LB-Miller medium at 37uC without agitation. Animals were euthanized on days 2, 3, 6 and 8 postinoculation by injection with Nesdonal in the occipital sinus and necropsied. Ceca were collected and cecal contents were weighed and then homogenized in sterile saline (9 ml/g of cecal content). Viable E. coli cells were counted by plating 10-fold dilutions in sterile saline on Drigalski agar plates (Biorad) supplemented with nalidixic acid at 30 mg/ml (selection of BEN2908 and BEN2908D-fosT strains) or with kanamycin at 50 mg/ml (selection of the BEN2908DfosT strain). The number of CFU of BEN2908 was calculated by subtracting the number of kanamycin-resistant bacteria from the number of nalidixic-resistant bacteria. The numbers of cecal CFU were calculated per gram of cecal content. Competition indices (CI) were calculated as described before for experimental colibacillosis.
In the intestinal colonization experiment of SPF chicken, total E. coli population of cecal contents was counted by plating 10-fold dilutions of cecal contents in sterile saline on Drigalski agar plates (Biorad). In this experiment, we also verified that the kanamycinresistant E. coli population was also nalidixic acid-resistant and unable to metabolize scFOS. To that end, kanamycin-resistant colonies were picked on LB-Miller agar plates containing nalidixic acid at 30 mg/ml. All the colonies tested were nalidixic acidresistant. Moreover, for each animal, kanamycin-resistant colonies were pooled, resuspended in 10 ml of LB-Miller medium containing kanamycin at 50 mg/ml and incubated overnight at 37uC with agitation. Overnight LB-Miller cultures were washed twice in M9 minimal medium and resuspended in the same volume of M9 minimal medium. Five milliliters of M9 minimal medium supplemented with 0.2% scFOS were then inoculated with 50 ml of the washed culture and incubated at 37uC with agitation. No growth was observed for any of the pools tested.

Statistical analysis
Statistical analyses of CI were performed using the Mann-Whitney U test. Exact P values were calculated with StatXact software (version 5.0; Cytel Inc., Cambridge, MA). Statistical analyses of luciferase expression and growth ability were performed using a Student's t-test.