|Home | About | Journals | Submit | Contact Us | Français|
The virulence of many human pathogens does not seem to be an evolutionarily selected trait, but an accidental by-product of the selection that operates in another ecological context. We investigated the possibility that virulence of the extraintestinal pathogenic Escherichia coli (ExPEC) strains, which frequently cause disease in the host in which they asymptomatically colonize the intestine, is the consequence of commensalism. Most of the ExPEC virulence factors are clustered on genomic islands called pathogenicity-associated islands (PAIs). We constructed and characterized several mutants of the ExPEC 536 strain with either (i) deletions of each single PAI or (ii) a complete deletion of all seven PAIs. In vitro phenotypic characterization of 536 mutants showed that the seven PAIs were dispensable for growth in the absence of external stress, as well as under a range of biologically relevant stressors, i.e., serum, bile, and oxidative, nitrosative, hyperosmotic, and acidic stress. However, challenge against the wild-type (WT) strain in a murine model shows that the deletion of all seven PAIs drastically reduces the fitness of 536 during persistent intestinal colonization. This defect seems to be linked to the hypermotility observed for mutants devoid of all seven PAIs. In addition, we show that PAIs diminish fitness of their carrier during growth in urine, suggesting that urinary tract infections are unlikely to provide selective pressure for the maintenance of ExPEC PAIs. Our results are in accordance with the coincidental-evolution hypothesis postulating that extraintestinal E. coli virulence is a by-product of commensalism.
Theoretical and experimental investigations suggest that increased virulence, i.e., parasite-mediated morbidity and mortality for infected hosts (38), could be selected for when it enhances the chance of a microorganism to be transmitted to a new host (38). However, virulence and transmission are not always correlated, and some forms of virulence did not coevolve with the host (39). For example, in an anthropo-centered point of view, Escherichia coli O157:H7 is an obligate intestinal pathogen. However, the normal habitat of this bacterium is the bovid gastro-intestinal tract, where it lives as a commensal rarely causing disease (24). E. coli O157:H7 virulence against humans presumably results from coincidental evolution, i.e., the symptoms provoked by the bacteria seem to be a maladjusted response to its accidental presence in the human gastro-intestinal tract rather than an evolutionarily selected means of transmission. Experimental evidence suggests that some of the genetic determinants of E. coli O157:H7 virulence, such as stx genes coding for the Shiga toxins, could be selected for to face off predators in the bovid intestinal tract (33, 59). Generally, the E. coli species challenges the threshold between commensal and pathogen (for a review, see reference 32). Hence, it constitutes an interesting model to address the question of the evolution of virulence. This is especially true for the extraintestinal pathogenic E. coli (ExPEC) strains, as they most frequently cause disease in the host in which they first asymptomatically colonize the intestine as commensal bacteria.
ExPEC strains are defined as bacteria with enhanced ability to cause infection outside the intestinal tract, such as in the bloodstream, the cerebrospinal fluid, or the urinary tract of mammalian and human hosts (29). The ExPEC ability to proliferate in these environments makes them the causative agents of devastating diseases, including septicemia, meningitis, cystitis, and pyelonephritis. ExPEC strains belong mainly to the B2 and D E. coli phylogenetic groups to which the prevalent intestinal colonizers of healthy individuals living in industrialized countries also belong (13, 45, 49, 62). Epidemiological studies showed that E. coli strains causing extraintestinal diseases carry specific genetic determinants called ExPEC virulence factors (VFs), which code for toxins, fimbriae, adhesins, and iron acquisition systems and confer utilization of alternative nutrients or resistance to biological stresses (reviewed in reference 30).
Most of the ExPEC VFs are clustered on mobile genetic elements, such as plasmids, or on genomic islands called pathogenicity-associated islands (PAIs) (21). PAIs are horizontally acquired genetic elements, whose involvement in extraintestinal pathogenesis was experimentally demonstrated (41). Under laboratory growth conditions, most PAIs can be spontaneously deleted from the chromosome (44). It is not clear if PAIs were acquired and maintained in vivo because they increase the fitness of their carriers in the intestine or in the extraintestinal environments. Having the intestine as a main reservoir provides optimal conditions for transmission because commensal bacteria are shed daily in the external environment and hence can reach new hosts. On the other hand, within-host extraintestinal environments do not provide such efficient modes of transmission. In fact, some can be considered dead ends, e.g., blood and cerebrospinal fluid. Therefore, it is plausible that ExPEC virulence is a by-product of commensalism resulting from an intrahost evolution. According to this hypothesis, the acquisition of PAIs should increase the recipient strains' fitness inside the host intestine but also coincidentally enhance the virulence of their carriers in extraintestinal niches. Such a scenario has previously been proposed (22, 37, 39) and is supported by both epidemiological (49-52, 62) and experimental (1, 18, 25, 26) studies showing that virulence factors such as adhesins could favor intestinal colonization. However, the consequences of the complete elimination of all PAIs have never been determined for commensal and pathogenic lifestyles. To gain insight into the selective pressures that drive the acquisition and/or maintenance of PAIs, we chose the uropathogenic strain E. coli 536 (5) as a model organism. Genomic sequencing has revealed the presence of seven PAIs in the chromosome of this bacterium (6).
We constructed seven single-PAI deletion mutants, as well as mutants with the deletion of all seven PAIs (I to VII). These mutants were characterized both in vitro, for growth, resistance to biologically relevant stressors, motility, and biofilm formation, and in vivo, for virulence and asymptomatic intestinal colonization. Our results strongly suggest that PAIs are fitness elements cooperating to enhance persisting intestinal colonization. This seems to be linked to the regulation of flagellum gene expression.
The uropathogenic strain E. coli 536 (O6:K15:H31) was isolated from a patient with acute pyelonephritis (5). This strain is naturally resistant to streptomycin. Therefore, all the strains used in this study are also resistant to streptomycin. Mutants deleted for all seven PAIs, the ΔI-ΔVII mutants, were constructed starting from the E. coli 536 ΔIII strain kindly provided by U. Dobrindt (Institut für Molekulare Infektionsbiologie, Universität Würzburg, Germany) (44). Bacteria were grown overnight under aeration at 30°C or at 37°C in LB broth (Difco) from frozen samples kept at −80°C in 20% glycerol. The antibiotics ampicillin, chloramphenicol, and kanamycin were added when needed at final concentrations of 100 μg/ml, 25 μg/ml, or 50 μg/ml, respectively. Serial dilutions of bacteria were performed in 10−2 M MgSO4 unless otherwise mentioned. A complete list of strains and plasmids used in this study is provided in Table S1 in the supplemental material.
Deletions of fliC and PAIs from the chromosome of E. coli 536 were performed using gene replacement methods based on the lambda Red recombinase system (11). Primers designed to amplify a recombinogenic PCR fragment consisting of a removable antibiotic resistance cassette flanked by 40-bp sequences homologous to the 5′ (upstream [up]) or 3′ (downstream [dw]) extremity of the target region that they were to replace are listed Table S2 in the supplemental material. The removable chloramphenicol resistance cassettes flanked by flipase recognition target (FRT) sequences were amplified from the template plasmid pKD3 (11). The same cassettes flanked by loxP sequences were amplified from an NdeI/XmnI digestion product of the plasmid pLoxPcat. Removable kanamycin resistance cassettes flanked by FRTs were amplified from pKD4. PCR and enzymatic digestion products were purified after electrophoresis on agarose gel with commercial kits (Promega). Strains carrying the temperature-sensitive helper plasmid pKD46 (11) coding the lambda Red recombinase system were treated as described previously (11). Purified recombined clones were verified by genotypic analysis with the PCR primers listed in Table S3 in the supplemental material and tested for the loss of pKD46 on LB agar with ampicillin. These primers were specially designed for multiplex PCR in order to detect several deletions in the same strain (melting temperature [Tm], 55°C, and primers used at a final concentration of 0.4 μM each). In order to remove the antibiotic resistance cassette from the chromosome, resistant clones were transformed with the helper plasmid pCP20 or pMDcre46 encoding the flipase (FLP) or the Cre recombinase, respectively. To remove the cassette flanked by FRT sites, strains transformed with pCP20 were treated as described previously (11). Cre was expressed from pMDcre46 under the control of the Para promoter inducible by arabinose but presenting an important leakiness. For this reason, removal of cassettes flanked by loxP sequences was conducted by incubating cells carrying pMDcre46 at 42°C overnight on nonselective LB agar plates without arabinose. The loss of helper plasmids and resistance cassettes was verified by plating clones on selective medium plates. In addition, each PAI deletion and antibiotic resistance cassette loss was verified by PCR with the primers listed in Table S3. The sequential deletion of the seven PAIs implied the presence of an increasing number of genomic scars consisting of FRT sequences. These scars provide substrates for homologous recombination with the recombinogenic fragment used to delete the next PAI carrying two of these FRT sequences. We noticed that after four successive deletions, the replacement efficiency of the following PAI decreased drastically because of improper insertion of the recombinogenic fragment in a previous scar, giving more than 99.9% false-positive clones. We resolved this problem by adapting the Cre/loxP recombination system (20) to the original gene replacement technique, and loxP sequences were used for some deletions instead of FRT sites during the construction of the septuple mutants. Because the seven PAIs represent 7.7% of the whole genome of E. coli 536, it was technically impossible to complement these deletions. Consequently, to be sure that the observed phenotypes were due to the deletion of the PAIs, two independent ΔI-VII mutants (#1 and #2) were constructed and tested in parallel in all experiments. Moreover, it has to be noticed that six sequential deletions represent 17 rounds of competent-cell preparation and electro-transformation, plating, streaking, overnight growth in liquid medium, and six rounds of expression of the lambda Red recombinase system. During these different steps, with each deletion representing around 200 generations, some mutations could be fixed. For this reason, an E. coli 536 strain derivative, referred to as the wild-type (WT) strain in the text, was obtained by submitting the ancestral 536 strain to six blank rounds of deletions in which the preparation of recombinogenic fragments was replaced by distillated sterile water. This evolved WT strain was used as a reference in all experiments performed in this study. A list of characterized and putative functions encoded on E. coli 536 PAIs is presented in Table S4 in the supplemental material.
The ΔI-VII #1 mutant was labeled by integrating the pMD2YFP (derived from pAH63) plasmid carrying the yfp gene under the control of the constitutive promoter P2rrnB in the phage lambda attB site on the chromosome as described in reference 23. The integration of the plasmid was verified as described in reference 23. The resulting strain was named ΔI-VII #1 (L).
The fitness of the ΔI-VII mutants relative to the WT strain was estimated during competitions in chemostat. The chemostats were constructed as described in reference 9, with some modifications (see http://www.necker.fr/tamara/pages/lab.html for the design and photographs). The growth rate, equal to the dilution rate (48), was set to 1 h, with a maximal volume of 16 ml and a fresh nutrient flow of 16 ml/h. The maximal volume was limited by an overflow consisting of a flushing glass tube linked to a pump sucking up the exhausted medium and alive or dead cells. Agitation was ensured by air or by N2 bubbling during aerobic or anaerobic experiments, respectively. Maintenance of the temperature at 37°C was ensured by maintaining chemostats in thermostated heating blocks. A total of 200 μl of a 1:1 mixture of mutants and WT strains coming from overnight cultures was added in chemostats continuously supplied with LB broth supplemented with streptomycin to a final concentration of 100 μg/ml. From each chemostat, 1 ml of culture was taken daily, the optical density at 600 nm (OD600) was measured, and the relative proportion of each strain was evaluated by plating dilutions on selective and nonselective LB agar plates. The competitive index (CI) was calculated as the logarithm of the number of antibiotic-resistant CFU divided by the inferred number of sensitive CFU after a given time and divided again by the same ratio in the inoculum for each chemostat.
We determined the relative fitness of the ΔI-VII mutants compared to the WT in LB and human urine over 100 generations in batch cultures. The human urines correspond to a pool of 24-h urines obtained from several healthy male donors that did not take any drug and were devoid of urinary tract infection (UTI) history. Overnight cultures of E. coli 536 WT and the ΔI-VII mutants were diluted a thousandfold in LB or urine at a final ratio of 1:1. Every 24 h, cultures were diluted a thousandfold in fresh medium during 10 days. Each of these cycles corresponds to 10 generations. The relative amount of each strain was determined after plating on selective agar plates. Neutrality of the chloramphenicol resistance cassette carried by the ΔI-VII mutants was estimated by using competing ΔI and ΔI::Cm strains (see Table S1 in the supplemental material).
We measured the impact of PAIs on the motility of E. coli 536 by measuring both the abilities to swarm and to swim as described in reference 19. Overnight cultures were inoculated on swimming and swarming plates with a sterile toothpick. The swimming motility plates were prepared with 0.3% agar (Difco), 1.0% Bactotryptone (Difco), 0.5% yeast extract (Difco), and 0.5% NaCl (Sigma). The swarming motility plates were prepared with 0.5% agar (Difco), 1.0% Bactotryptone (Difco), 0.5% yeast extract (Difco), 0.5% NaCl (Sigma), and 0.5% D-(+)-glucose (Sigma). The plates were air dried at room temperature for 30 min under a laminar-flow hood before being inoculated. The plates were photographed after incubation for 8 h for swimming and 24 h for swarming at 37°C. Competitions in structured medium (LB 0.3% agar or 0.75% agar) allowing or not allowing for swimming were performed as described in reference 14 between the WT strain and the ΔI-VII #1 (L) mutant labeled with the yfp gene.
For real-time PCR, the cDNA synthesis was performed using 2 μg RNA with random hexamers (12.5 ng/ml) and the Superscript II RNase H− kit 5 (Invitrogen) according to the manufacturer's instructions. The real-time PCR experiments were performed using the SYBR green PCR Master Mix (Applied Biosystems) to quantify the expression level of the fliA and motA genes. The rpoD gene was chosen as a reference gene for data normalization. Primer pairs were designed using the Primer Express software (Applied Biosystems): for motA, 5′-GCGATTAAAGGCACGCTGAAGG and 5′-GAAGGTGTTCATGTGACCGCTG; for fliA, 5′-GCGTGGAACTTGACGATCTGC and 5′-TCCGCTACCTCAGTTTCCGTG; and for rpoD, 5′-CGTCTGATCATGAAGCTCTG and 5′-GTATCGCTGGTTTCGTTG. Amplification and detection of gene-specific products were carried out with the 7300 real-time PCR system (Applied Biosystems). Data analysis was performed with the 7300 system software. For each target gene, the average cycle threshold (CT) value was calculated from triplicate RNA samples. The difference between the CT value of the target gene and the CT value of the endogenous reference gene (rpoD) was defined as the ΔCT. The ΔΔCT value described the difference between the ΔCT of the WT strain and the mutant strain. The difference in expression levels was calculated as 2ΔΔCT.
To assess the ability of the bacterial strains under study to form biofilm, an in vitro test was conducted as described in reference 12. Briefly, bacteria from overnight cultures were diluted in minimal medium supplemented with glucose and grown for 24 h at 37°C without shaking in 96-well polystyrene plates. Planktonic bacteria were washed out with distilled water, and bacteria that remained attached to the well walls were dried and fixed for 30 min at 80°C. The amount of biofilm was estimated using 1% crystal violet (Sigma) solution. The crystal violet trapped in biofilms was solubilized with dimethyl sulfoxide (Sigma). The concentration of crystal violet was then measured with a spectrophotometer at 600 nm. For each strain, at least six independent experiments were performed.
The ability of bacterial strains to cause sepsis was determined using 16- to 18-g female mice (Charles River OF1). A total of 200 μl of a suspension of 109 bacteria/ml in physiological serum was inoculated by subcutaneous injection, and the number of surviving mice was monitored during 7 days (53).
Six-week-old female mice (Charles River CD-1) treated with streptomycin were used to monitor the ability of the different strains to colonize the intestine of a mammalian host as described in reference 46, with modifications. Five days before inoculation, mice were isolated and streptomycin was added to the sterile drinking water at a final concentration of 5 g/liter. Streptomycin was maintained until the end of the experiment. The antibiotic treatment efficiency against the coliform intestinal population was controlled by plating a pure suspension of feces in physiological water on Drigalski selective agar medium. Mice free of coliform flora were inoculated per os with 106 bacteria in 200 μl of physiological water. Every day postinoculation, the intestinal population of E. coli was estimated by plating dilutions of weighted fresh feces on LB agar with or without appropriate antibiotic. All animal experiments were performed in compliance with the recommendations of the French Ministry of Agriculture and approved by the French Veterinary Services (accreditation A 75-18-05).
Statistical analyses were performed using JMP IN (SAS Institute, Inc.) and Prism (GraphPad Software, Inc.).
Deletion of the seven PAIs reduced the E. coli 536 genome size by 7.7%. The loss of some PAI-encoded functions and/or such an important reduction of the genome size could modify the fitness of the ΔI-VII mutants relative to the WT strain. Therefore, we tested the ability of these mutants to grow in LB medium or in human urine in competition against the WT. In order to follow bacterial growth at constant rates during several days, we first conducted experiments in chemostat. Growth conditions were set to obtain a generation time of about 60 min, which is close to the generation time of bacteria growing in the intestinal tract of streptomycin-treated mice (54). Growth under anaerobic conditions, mimicking the lack of oxygen in the intestinal mucus (31), was performed in chemostat in which air bubbling was replaced by N2 bubbling. After 4 days of competition in the presence of oxygen (about 96 generations) or 7 days of competition under anaerobic conditions (about 168 generations), there was no significant deviation from the initial ratio between the WT strain and the ΔI-VII mutant strains (Fig. (Fig.1;1; CIO2 = 0.23 ± 0.45; CIN2 = 0.22 ± 0.25; Wilcoxon signed-rank test, P > 0.05).
We also performed competitions with aerobic batch cultures in LB or in human urine (Fig. (Fig.2).2). This experimental setup allowed for testing of bacterial strain fitness during 10 cycles of passages through lag phase, exponential growth, and stationary phase. Under such competition conditions, after about 100 generations, the complete PAI deletion was advantageous in LB medium and in urine (CILB = 0.51 ± 0.08; CIurine = 1.32 ± 0.18; Wilcoxon signed-rank test, P < 0.05). It should be noted that the relative fitness of the ΔI-VII mutants compared to the WT was significantly higher in human urine than in LB (Mann Whitney test, P < 0.01). This effect was not due to the presence of the chloramphenicol resistance cassette in the ΔI-VII mutants, as the ΔI::Cm strain had the same fitness relative to the isogenic ΔI mutant (CILB = 0.08 ± 0.18; CIurine = −0.01 ± 0.21; Wilcoxon signed-rank test, P > 0.05). The fact that PAIs diminish the fitness of its carrier during growth in human urine is in line with the hypothesis according to which ExPEC PAIs could be selected and maintained in an ecological context other than the urinary tract.
The ability of the WT and ΔI-VII mutant strains to use 95 substrates as a sole carbon source was determined using Biolog GN2 galleries. Bacterial growth was evaluated by measuring the increase of the OD600 due to tetrazolium reduction after incubation for 24 h and 48 h at 37°C. A given substrate was considered metabolized when the OD600 was >0.2 for at least two repetitions. WT and ΔI-VII mutant strains differed only in their abilities to use d-serine, as it took more time for the ΔI-VII mutants to reach the threshold OD600 than for the WT strain (see Table S5 in the supplemental material). This presumably results from the absence in the ΔI-VII mutants of the dsd gene cluster carried by PAI II. This gene cluster codes for DsdA, a d-serine dehydratase, DsdX, a transporter, and DsdC, a transcriptional regulator. Nevertheless, the duplication of dsdA and the presence of an alternative transporter coded by cycA (3) in the chromosome of E. coli 536, outside the seven PAIs, should allow ΔI-VII mutants to use d-serine, albeit at a lower rate.
In order to identify selective pressures that might lead to the acquisition and maintenance of PAIs, several in vitro stress resistance tests were conducted. All of the chosen stressors were of biological relevance, i.e., the ability to grow in bovine bile or human serum, the resistance to oxidative and nitrosative stress generated, respectively, by paraquat or NaNO2, and resistance to hyperosmotic stress and acidic pH. For all conditions, the responses of the ΔI-VII mutants and of the WT strain were compared. For some stressors, resistance of the mutants with a single PAI deletion was also tested. No significant difference between resistance levels of the ΔI-VII mutants, single mutants, and WT strain to any of these stressors was observed (see Fig. S1, S2, and S3 in the supplemental material). Therefore, among the genes carried on the seven PAIs of E. coli 536, none seemed to play an essential role in the resistance to tested biological stressors which are normally encountered by this bacterium during both commensal and pathogenic life cycles.
Motility is a major trait involved in both the pathogen and commensal lifestyles of E. coli (16, 34). Different orthologous genes present on several PAIs of E. coli 536 were shown to be involved in the regulation of motility in other ExPEC strains (57, 58). These genes encode putative MarR-like transcriptional regulators. They are located downstream of the adhesin-coding gene in some fimbrial operons. According to the nucleotide sequence homology and synteny (determined using Microbial Genome Annotation System; https://www.genoscope.cns.fr), at least three of such genes are present in E. coli 536, i.e., ECP_3776 on PAI I, papX (ECP_4531) on PAI II, and sfaX (ECP_0301) on PAI III. The role of PAIs in the motility of E. coli 536 was addressed by measuring the swarming and the swimming abilities of the WT and the mutant strains. Single mutants did not present any alteration of their swimming and swarming abilities compared to the WT (Fig. (Fig.33 A and B). On the other hand, the swimming and swarming abilities of the ΔI-VII mutants were clearly enhanced (Fig. 3A and B). Redundancy may explain why only the ΔI-VII mutants had an altered motility, unlike the single PAI deletion mutants. This result was confirmed by placing the ΔI-VII #1 (L) mutant in competition against the WT strain in a structured environment. When the agar percentage was low enough to allow for cell motility, the WT strain was rapidly outcompeted by the mutant (Fig. (Fig.3C).3C). On solid LB agar, an environment on which bacteria cannot swim, no difference in fitness between the ΔI-VII #1 (L) and WT strains was observed. Because regulators potentially responsible for alteration of E. coli 536 motility act at the transcriptional level, the expression of the class II flagellar gene fliA and the motA gene coding for the proton conductor component of the flagellum motor was quantified using qRT-PCR (Fig. (Fig.3D).3D). The expression of fliA was found to be upregulated >4-fold and that of motA >2-fold in the ΔI-VII mutants compared to the WT.
In both the gastrointestinal (42) and urinary tracts (2), ExPEC cells are found in biofilm-like communities, which could ensure their persistence. Therefore, we assayed the ability of the mutants deleted for one or the seven PAIs to form a biofilm in polystyrene 96-well plates during overnight growth in minimal medium glucose at 37°C (Fig. (Fig.4).4). The amount of biofilm was revealed using crystal violet. The results obtained for the different strains show that only genes carried by PAI II and PAI III play a significant role in biofilm formation by E. coli 536. Both the ΔII and ΔIII mutants formed thinner biofilms than the WT (Mann Whitney test, P values corrected for multiple comparisons by the Bonferroni-Holm procedure were <0.0001). The impact of those genes seems cumulative, as the ΔII-ΔIII double mutant and ΔI-VII strain were equivalently impaired in their capacities to form biofilm (Mann Whitney test, P > 0.05).
The pathogenicity of different mutants relative to the WT strain was evaluated using the murine septicemia model (53). Analysis of the survival data by the Kaplan-Meyer method showed that the ΔI-VII mutants had significantly attenuated virulence compared to the WT strain (Fig. (Fig.5,5, log rank test, P < 0.0001). This result confirms that some PAIs carry genes that enhance the capacity of the WT strain to kill mice during sepsis. However, none of the mutants deleted for a single PAI had a significant defect in their ability to kill mice under the experimental conditions used. This could be explained by the redundancy of functions encoded on several different PAIs (e.g., the alpha-hemolysin encoded on both PAI I and PAI II ) and/or the additive effects of VFs on virulence (28, 60). It should be noted that the attenuation of the ΔI-VII mutants cannot be related to a loss of serum resistance (see Fig. S1B in the supplemental material).
The hypothesis that the genes on PAIs could be involved in intestinal colonization was verified using streptomycin-treated mice. This is a well-established model for determination of the fitness of enteric bacteria during intestinal colonization (8). The efficiency of the streptomycin treatment in elimination of the Enterobacteriaceae from the mice intestine was verified by plating feces on Drigalski medium plates.
First, the ability of individual bacterial strains to colonize the gut of mice without interference from coliform competitors was determined. Pure cultures of WT or ΔI-VII mutant strains were inoculated in streptomycin-treated mice by intragastric administration of 106 CFU per animal. During 1 week, the amount of E. coli bacteria in the intestine was evaluated daily by plating serial dilutions of fresh feces on LB agar plates. The WT and the mutant strains colonized the gut of mice with the same efficiency. After 48 h, the amount of 109 to 1010 CFU/g of feces was achieved and maintained throughout the following period of 5 days (Fig. (Fig.66 A). Therefore, without coliform competitors and the streptomycin-sensitive part of the intestinal flora, the absence of PAIs did not affect the ability of E. coli 536 to initiate and maintain intestinal colonization.
In natura, the intestinal ecological niche never stays free of competitors. Thus, in a second experiment, we measured the fitness of mutants without one or the seven PAIs relative to the WT strain during competitions for intestinal colonization (Fig. (Fig.6B).6B). A nonparametric Wilcoxon test on paired data comparing the number of CFU of antibiotic-resistant strain with the number of CFU of sensitive strain was conducted for each competition, and the P values was corrected for multiple comparisons by the Bonferroni-Holm procedure (27). The expression of the cassettes conferring resistance to chloramphenicol or to kanamycin that were used to differentiate strains during competitions did not have any significant cost during intestinal colonization (Fig. (Fig.6B6B).
No defect due to the absence of PAIs for the initial phase of the colonization, i.e., 24 h postinoculation, was observed (Fig. (Fig.6B).6B). Forty-eight hours postinoculation, populations of both competing strains reached 109 to 1010 CFU/g of feces (data not shown). However, the mean competitive index of the ΔI-VII mutants decreased by three orders of magnitude 7 days postinoculation (Fig. (Fig.6B).6B). Both ΔI-VII mutants (#1 and #2) gave similar results against the WT (Mann Whitney test, P > 0.05). Therefore, no mutation elsewhere in the core genome other than the PAI deletions seemed to be responsible for the ΔI-VII colonization defect. None of the single PAI deletion mutants showed significant intestinal fitness loss. This suggests that the maintenance of the colonization was compromised only by the absence of several PAIs. Redundant functions encoded by different PAIs promoting intestinal maintenance could render difficult the detection of effects resulting from the deletion of a single PAI. It is also again possible that genes on different PAIs encode proteins with additive and/or synergistic activities.
In vitro characterization of different PAI mutants showed that the ΔI-VII mutant had two phenotypes that differ significantly from those of the WT, i.e., a reduced capacity to form biofilm and an enhanced motility. We tested a possible involvement of these phenotypes in the intestinal colonization. First, we used the double PAI II-PAI III deletion mutant (the ΔII-ΔIII mutant), which is as deficient in its capacity to form biofilms as the ΔI-VII mutant (Fig. (Fig.4),4), to test the involvement of this phenotype in intestinal colonization. This strain did not show any intestinal colonization defect relative to the WT (Fig. (Fig.6B),6B), suggesting that the capacity to form biofilm on an abiotic surface is not correlated with the ability to durably colonize the gut of mice. Second, we deleted fliC, coding for flagellin, from the genomes of the WT and the ΔI-VII mutant in order to test the role of motility in the intestinal colonization. Both strains were completely immobile (data not shown). Competition between the fliC and ΔI-VII fliC mutants showed that there is no difference in their capacities to colonize and persist in intestine (Fig. (Fig.6B),6B), indicating a role of flagella in intestinal colonization by E. coli.
ExPEC strains are characterized by an enhanced ability to cause infections outside the intestinal tract compared to other E. coli strains (29). Most of these strains belong to the E. coli phylogenetic groups B2 and D (53). The virulence of ExPEC strains depends on VFs, most of which are clustered on mobile genetic elements, such as plasmids, or on PAIs (reviewed in reference 30). Nevertheless, E. coli strains from groups B2 and D, carrying ExPEC VFs, are prevalent in the feces of healthy individuals living in industrialized countries (13, 45, 49, 62). These observations indicate that the intestine could be the primary reservoir of ExPEC strains. As a corollary, it has been proposed that ExPEC PAIs could be fitness elements in the intestinal ecological context. Consequently, the virulence of ExPEC strains could be seen as the by-product of the commensal lifestyle (37, 39). The aim of this study was to provide experimental demonstration of this hypothesis. The archetypal uropathogenic strain E. coli 536 was chosen as a model organism (21). The vast majority of 536 VFs are clustered on seven PAIs. Mutants with a complete deletion of all seven PAIs (the ΔI-VII mutants) and a set of mutants with only one or two PAIs deleted were constructed and characterized both in vitro and in vivo.
The capacity of the ΔI-VII mutants to colonize the host intestine relative to the WT strain was evaluated using streptomycin-treated mice. For more than 7 days after intragastric inoculation, the WT strain colonized the mice, reaching populations of 109 to 1010 bacteria per gram of feces, without any evidence of clinical symptoms. The ΔI-VII mutants colonized the mice intestine with the same efficiency as that of the WT strain (Fig. (Fig.6A).6A). We concluded that PAI-encoded functions are not essential for the colonization of the streptomycin-treated mice. In order to colonize the mice intestine, bacteria must survive acidic pH in the stomach and then the bile unloaded in the duodenum. Therefore, it is not surprising that the ΔI-VII mutants and the WT strain resist equivalently to both stresses in vitro (see Fig. S1 and S3 in the supplemental material). In competition against the WT strain, the ΔI-VII mutants were able to initiate colonization, reaching the intestinal carrying capacity corresponding to 109 to 1010 bacteria/g of feces. However, 7 days after inoculation, the population size of the ΔI-VII mutants decreased to a level on average three orders of magnitude lower than that of the WT (Fig. (Fig.6B).6B). These results clearly showed that ΔI-VII mutants were impaired in their ability to maintain intestinal colonization but only in competition with the WT strain.
Maintenance of intestinal colonization requires many properties, among which metabolic competence is of the utmost importance. When two strains are in competition for a limited nutrient, the one that is able to use it more efficiently should outcompete the other (15). However, several pieces of evidence rule out metabolic competence as an important factor in the outcome of the intestinal competition between the ΔI-VII mutants and the WT strain. First, in vitro competitions in rich medium in chemostat showed that the ΔI-VII mutants were at least as fit as the WT under both aerobic and anaerobic conditions (Fig. (Fig.1)1) and in batch culture competitions the ΔI-VII mutants were even fitter than the WT (Fig. (Fig.2).2). Second, comparison of the metabolic capacities of both strains in Biolog GN2 galleries (see Table S5 in the supplemental material) showed that the ΔI-VII mutants were impaired only in their ability to use d-serine as a sole carbon source. This phenotype probably results from the deletion of PAI II that carries the dsd gene cluster as published for E. coli CFT073 (3). However, the WT strain did not outcompete the ΔII mutant in the intestine. Finally, the WT strain and the ΔI-VII mutants grow equally well in minimal medium with gluconate, i.e., the principal carbon source used by E. coli in the intestine (7) (see Table S5).
When competitors have the same ability to use available nutrients, the issue of competition can be determined by the ability to resist different stresses or the outflow by, for example, forming biofilms. The WT strain and the ΔI-VII mutants showed no significant difference in their abilities to resist different biologically relevant stressors (see Fig S1, S2, and S3 in the supplemental material). However, the ΔI-VII mutants had a completely abolished ability to form biofilm on an abiotic surface compared to the WT strain (Fig. (Fig.4).4). Some of the genetic determinants allowing for biofilm formation in vitro seemed located on PAI II and III (Fig. (Fig.4).4). Several genes on PAI III were known to play a role in biofilm formation, i.e., flu coding for antigen 43 (10) and the foc operon coding for the F1C fimbriae (35), while further work must be done to identify genes in PAI II involved in biofilm formation. F1C fimbriae were shown to improve the maintenance of the E. coli commensal strain Nissle 1917 in the intestine of infant mice during competition (35). However, the capacity to form biofilms did not contribute to fitness during intestinal colonization, with the ΔII-ΔIII double mutant being as fit as the WT, even though it was unable to form biofilm (Fig. (Fig.44 and and6B6B).
In vitro motility was enhanced in the ΔI-VII mutants, but not in single mutants (Fig. (Fig.3).3). Despite the fact that their enhanced motility provided a selective advantage to ΔI-VII mutants in competition with the WT strain in an in vitro structured environment (Fig. (Fig.3C),3C), it did not contribute to any fitness increase during intestinal colonization (Fig. (Fig.6B).6B). Actually, when both the WT and ΔI-VII mutants were deleted for fliC, the gene encoding flagellin, no difference between the two strains was observed regarding their relative fitness in the mouse intestinal tract. This suggests that several PAIs could be maintained in the chromosome of E. coli 536 because they carry genes negatively regulating flagellum expression. This result corroborates the previous observation that motility is dispensable for intestinal colonization by E. coli F18 (43). Furthermore, selective pressures in the intestinal tract tend to counterselect the motility of E. coli. Hence, it was shown that mutations in the flhDC genes and in the two-component system envZ-ompR, drastically affecting the expression of the flagella, give a selective advantage in the intestine of mice (17, 36). A parsimonious regulation of the expression of genes involved in motility is required probably because proteins constituting the flagella are immunogenic (4) and costly to produce (14). However, neither the cost of motility nor the potential immunogenicity of the flagella seemed to be involved in the reduction of fitness of the ΔI-VII mutants relative to the WT strain during intestinal colonization as suggested by two observations: (i) the cost of hypermotility for the ΔI-VII mutants was not detected under conditions allowing elevated growth rates in vitro (Fig. (Fig.11 and and2)2) and (ii) no macroscopic or microscopic cecal inflammation was observed during colonization with either the ΔI-VII mutant or the WT strain (data not shown). Hence, the fact that fliC-dependent hypermotility reduces fitness of the ΔI-VII mutant remains to be explained.
The purpose of this study was to show that PAIs could be selected in the intestinal ecological context. Our results strongly suggest that this could be the case. The ability to durably colonize the intestinal tract guaranties the long-term fitness of commensal E. coli in terms of transmission. In adult humans, E. coli is the main facultative anaerobic species, reaching around 109 bacteria per gram of colonic content (56), meaning that up to 200 g of feces per individual per day, containing about 1011 E. coli bacteria, reach the external environment. It is undeniable that PAIs are an important contributor to the pathogenicity of ExPEC strains in extraintestinal niches. The loss of the seven PAIs clearly affected the ability of E. coli 536 to kill mice during sepsis (Fig. (Fig.5).5). However, extraintestinal infections are unlikely to provide a mode of transmission as efficient as intestinal colonization. The impact of the extraintestinal infections on the transmission of bacteria between hosts depends on the type of infection. Host death resulting from sepsis or neonatal meningitis is an evolutionary dead end for bacteria. Urinary tract infections could provide means of transmission via urine. On the other hand, induction of the host immune response results in clearance of the bacteria, thus limiting the time of shedding. Some E. coli strains provoking urinary tract infections can stay hidden in a quiescent state in the bladder urothelium, from which they can spread and trigger new infections once conditions are again favorable (55). However, even when urinary tract infections are recurrent, the quantity of bacterial cells shed should be several orders of magnitude lower than that disseminated via feces. Some E. coli strains can durably and asymptomatically colonize the human urinary tract and can be shed daily in quantities up to 109 bacteria/ml of urine (40). Such adaptation to the human urinary tract could increase transmission. However, full armament of ExPEC VFs seems not to be selected for during the evolution of urinary tract commensalism, as it was reported that some VFs of these strains are inactivated by point mutations and recombination events (61). Last, our results showed that PAIs represented a burden when E. coli 536 was cultivated in urine (Fig. (Fig.22).
In conclusion, we demonstrated that ExPEC PAIs are fitness elements involved in intestinal colonization. Our results are in accordance with the coincidental evolution hypothesis that regards extraintestinal E. coli virulence as a by-product of commensalism. This work lays the ground for further research on new strategies preventing extraintestinal diseases by acting directly on E. coli inside the gut before it can reach extraintestinal sites.
We are grateful to U. Dobrindt for the generous gift of a strain. We thank A. Swidsinski for his technical help, O. Bouvet for useful discussions, and M. Leroy and L. Robert for comments on the manuscript.
M. Diard was supported by a grant from Fondation pour la Recherche Médicale.
Published ahead of print on 23 July 2010.
§Supplemental material for this article may be found at http://jb.asm.org/.