ABSTRACT
A new Escherichia coli virulent clonal group, O45:K1, belonging to the highly virulent subgroup B21 was recently identified in France, where it accounts for one-third of E. coli neonatal meningitis cases. Here we describe the sequence, epidemiology and function of the large plasmid harbored by strain S88, which is representative of the O45:K1 clonal group. Plasmid pS88 is 133,853 bp long and contains 144 protein-coding genes. It harbors three different iron uptake systems (aerobactin, salmochelin, and the sitABCD genes) and other putative virulence genes (iss, etsABC, ompTP, and hlyF). The pS88 sequence is composed of several gene blocks homologous to avian pathogenic E. coli plasmids pAPEC-O2-ColV and pAPEC-O1-ColBM. PCR amplification of 11 open reading frames scattered throughout the plasmid was used to investigate the distribution of pS88 and showed that a pS88-like plasmid is present in other meningitis clonal groups such as O18:K1, O1:K1, and O83:K1. A pS88-like plasmid was also found in avian pathogenic strains and human urosepsis strains belonging to subgroup B21. A variant of S88 cured of its plasmid displayed a marked loss of virulence relative to the wild-type strain in a neonatal rat model, with bacteremia more than 2 log CFU/ml lower. The salmochelin siderophore, a known meningovirulence factor, could not alone explain the plasmid's contribution to virulence, as a salmochelin mutant displayed only a minor fall in bacteremia (0.9 log CFU/ml). Thus, pS88 is a major virulence determinant related to avian pathogenic plasmids that has spread not only through meningitis clonal groups but also human urosepsis and avian pathogenic strains.
Escherichia coli is the second cause of neonatal bacterial meningitis in industrialized countries, after group B streptococci. Despite advances in neonatal care, E. coli neonatal meningitis (ECNM) remains a potentially devastating illness, with a mortality rate of 10 to 15% and neurological sequelae in 30 to 50% of cases (8, 20, 28, 29, 62). Like most extraintestinal pathogenic E. coli (ExPEC) strains, ECNM strains mainly belong to phylogenetic group B2 and compose a small number of serotypes (O18:K1:H7, O1:K1, O83:K1, and O7:K1) (2, 11, 12, 33, 41, 49). Virulence genes in ECNM strains are involved in two major pathological steps, namely, intravascular multiplication (leading to high-level bacteremia) and blood-brain barrier penetration (38). A new virulent clone, O45:K1:H7, was recently identified in France, where it now accounts for one-third of ECNM isolates (13). The K1 capsular antigen and the salmochelin siderophore are the only two specific virulence traits known to be shared by this clone and the archetypal strains of clonal group O18:K1:H7, even though the two clonal groups belong to the same highly virulent subgroup B21, defined by a specific sequence type (ST) designated ST29Whittam (www.shigatox.net ), ST95Achtmann (www.mlst.net ), or B2-IX (in Denamur's scheme) (10, 40). Hence, this new clone may serve to identify new virulence determinants and other genetic factors involved in the pathogenesis of ECNM. Strain S88, representative of clone O45:K1:H7, has been sequenced as part of a whole-genome sequencing project (ColiScope [www.genoscope.cns.fr ]) at the Evry Genoscope in France. This strain harbors a large plasmid, designated pS88, that encodes the salmochelin and aerobactin siderophores and may be involved in the virulence of S88 (13).
Here we describe the complete DNA sequence, epidemiology, and function of pS88.
MATERIALS AND METHODS
Bacteria.Strain S88, a representative of the French clonal group (O45:K1:H7) which accounts for one-third of ECMN isolates in this country, was isolated in 1989 from the cerebrospinal fluid of a newborn (Robert Debré Hospital, Paris, France). This strain is known to harbor a large plasmid (>100 kb) encoding at least two iron uptake systems, namely, salmochelin and aerobactin (13). Other chromosomally encoded ExPEC traits are K1 antigen, P fimbriae with the adhesin PapGII, and yersiniabactin (13). Strain K-12, MG1655, was used as a colicin-sensitive strain. Strain J53, with natural rifampin (rifampicin) resistance, was used for conjugal transfer of plasmids as previously described (51).
To investigate the epidemiology of plasmid pS88, we also used 66 strains representative of the major meningitis clonal groups defined by the combination of their ST (www.shigatox.net ) and their serogroup (sequence O types) as previously described (9). Other human and avian E. coli strains (n = 15) belonging to the highly virulent subgroup B21, characterized by ST29 (10), were also studied (Table 1).
Molecular epidemiology of plasmid pS88 by multiplex PCR of pS88 ORFs performed on 81 K1 E. coli strains representative of extraintestinal pathogenic clones causing meningitis and strains of other origins belonging to the highly virulent ST29/B21 subgroup B2
Sequencing and annotation of the pS88 plasmid.The genome and plasmid of strain S88 were sequenced as part of a whole-genome sequencing project (ColiScope [www.genoscope.cns.fr ]) at the Evry Genoscope in France. Sequencing and assembly of pS88 were performed as previously described (7, 50). MaGe (Magnifying Genomes) software was used for gene annotation and comparative analysis of the S88 genome as described elsewhere (50, 63).
Phylogenetic analysis.To determine the genetic relatedness of the tra gene clusters in plasmid S88 by comparison with plasmids of other strains, the nucleotide sequences of traJ, traM, traS, traT, and traY from several E. coli conjugative plasmids were extracted from GenBank. The Clustal W program was used to align the sequences (59). Phylogenetic and molecular evolutionary relationships were examined by using the neighbor-joining method implemented with MEGA 3.1 software (39). Bootstrap confidence values for each node of the trees were calculated over 100 replicate trees.
Epidemiology of pS88-like plasmids by multiplex PCR and pulsed-field gel electrophoresis (PFGE).In order to investigate the epidemiology of pS88 among human and avian E. coli strains, PCR was used to screen for 11 open reading frames (ORFs) or genes scattered throughout the plasmid. The PCR targets and primers are listed in Table 2. The aerobactin and salmochelin genes were PCR amplified as previously described (14). The other nine ORFs or genes were detected with two new multiplex PCR methods (one hexaplex and one triplex). PCR was carried out in a 50-μl volume with 25 μl of 2× Qiagen Multiple PCR Master Mix (Qiagen, Courtaboeuf, France), 5 μl of 5× Q-solution, 5 μl of a primer mix (with final concentrations in the hexaplex PCR of 0.2 μM for cia, traJ, and ompTP; 0.1 μM for etsC and ORF143; and 0.05 μM for iss and in the triplex PCR of 0.2 μM for hlyF, cvaA, and sitA), 10 μl of distilled water, and 5 μl of bacterial lysate, using an iCycler thermal cycler (Bio-Rad, Marnes la Coquette, France) under the following conditions: DNA denaturation and polymerase activation for 15 min at 95°C; 30 cycles of 30 s at 94°C, 90 s at 55°C, and 90 s at 72°C; and a final extension step for 10 min at 72°C. Samples were electrophoresed in 3% Resophor gels (Eurobio, France) and then stained with ethidium bromide and photographed with UV transillumination. In each PCR run, a lysate of strain S88 was used as a positive control. Finally, all strains harboring the 11 pS88 genes were also screened for eitB and tsh genes using the primers described by Johnson et al. (35).
Oligonucleotides used for mutagenesis and PCR assaya
Plasmid sizes were determined by PFGE of undigested DNAs of strains harboring the 11 pS88 genes. PFGE followed by Southern blot hybridization with an iroN probe were performed as previously described (13).
Cure of the pS88 plasmid.Strain S88 was grown for 18 h in Luria-Bertani (LB) medium at 37°C with shaking. The culture was then diluted to 105 CFU/ml, and serial concentrations (2.5, 5, and 10%) of sodium dodecyl sulfate were added. After 18 h of growth with shaking, the cultures were plated on LB agar. After overnight incubation, 200 colonies were screened for colicin production by pricking them out on LB agar plates overlaid with a suspension of E. coli strain K-12, which is sensitive to colicin. Test strains were assumed to have lost the ColV plasmid if they were unable to inhibit the growth of strain K-12. This was confirmed, using the PCRs described above, by the loss of the 11 plasmid-related ORFs or genes and by the disappearance of extrachromosomal DNA on agarose gel electrophoresis after plasmid preparation with the relevant colonies (Qiaprep Minispin, Qiagen). For each variant thus obtained, we checked the expression of the K1 capsule antigen and the presence of chromosomal virulence genes, using multiplex PCR as previously described (14).
Mutant construction.S88 mutants were obtained with the PCR-based method of Datsenko and Wanner (19) as previously described, with plasmids kindly provided by Lionello and Nara Bossi (Centre de Génétique Moléculaire, CNRS, Gif sur Yvette, France) (47, 50). The primers used for insertional mutation are listed in Table 2. Correct introduction of the cat gene into the target was controlled by PCR using primers homologous to the cat gene and the flanking region of the target, as previously described (47, 50). For each mutant thus obtained, we checked the expression of the K1 capsule antigen, bacteriocin production, the presence of nondeleted plasmid genes, and chromosomal virulence genes, using multiplex PCRs.
Conjugal plasmid transfer.The cat gene was introduced into a nonencoding region of pS88 between ORFs 131 and 132, using the method of Datsenko and Wanner (19), so that the plasmid harbored a chloramphenicol resistance determinant (plasmid pS88cat) for the selection of transconjugant clones after conjugal transfer. The cured S88 variant was made resistant to nalidixic acid in order to select transconjugants for pS88cat reintroduction.
Donor and recipient cells were grown overnight at 37°C in LB broth. A 10−2 dilution was prepared with fresh LB medium. After incubation at 37°C for 2.5 h without shaking, 2.5-ml aliquots of each culture were mixed. After overnight incubation, the mating mixture was harvested and bacteria were plated on selective medium. The transconjugants thus obtained were checked for colicin production and the presence of plasmid-related genes by PCR and by gel electrophoresis of a plasmid extract (Qiaprep minispin; Qiagen).
Experimental models.We assessed the ability of the wild-type strain and mutants to induce high-level bacteremia in newborn rats, as previously described (30). Briefly, pathogen-free Sprague-Dawley rats were obtained from Charles River Laboratories at 4 days of age, with their mothers. At 5 days of age, the pups were inoculated intraperitoneally with a normal saline suspension containing ∼500 CFU of the test strain. A tail incision was made 18 h after inoculation, and 5 μl of blood was sampled. Serial dilutions were plated for CFU counting. Comparisons of bacterial counts in the animal model were based on a two-sample unpaired t test. Data are expressed as mean ± standard deviation. P values below 0.05 were considered to denote statistically significant differences.
Serum bactericidal activity against E. coli strains was determined using pooled sera of healthy volunteers as previously described (30). To measure bactericidal serum activity, 20 μl of a bacterial inoculum of 108 CFU/ml in physiological serum was added to 180 μl of freshly thawed pure serum. Quantitative cultures were done 5 h later. Experiments were repeated five times.
Nucleotide sequence accession number.The DNA sequence of pS88 has been deposited in GenBank under accession number CU928146.
RESULTS
Sequencing of pS88. (i) Overview.A complete circularized DNA sequence of plasmid pS88 was obtained. It was 133,853 bp long, with an overall G+C content of 49.28% (Fig. 1). A total of 133 protein-coding genes plus 11 fragments of protein-coding genes corresponding to six pseudogenes were identified in the sequence (Table 3). Globally, based on sequence homologies, the plasmid could be separated into two halves. The first, from base 20,000 to base 86,000, is mostly composed of the genes involved in plasmid-related functions (conjugal transfer, maintenance, and partition); all are almost exclusively transcribed in the clockwise direction, with an average G+C content of 51%. The other half contains most of the putative or known virulence-associated genes, such as those encoding iron uptake systems and factors involved in resistance to the innate immune system (see below); they were transcribed in both directions and showed sharp deviations from the average G+C content (29.9% to 63.6%), suggesting a foreign mosaic origin. Thirty-five ORFs of unknown function and 14 insertion sequence-like genes were scattered throughout the plasmid. Two different colicins, colicin Ia and colicin V, along with their immunity protein, were encoded by pS88.
Circular representation of the Escherichia coli strain S88 plasmid (pS88). Circles display (from the outside) (i) GC percent deviation (GC window − mean GC) in a 1,000-bp window, (ii) predicted ORFs transcribed in the clockwise direction, (iii) predicted ORFs transcribed in the counterclockwise direction, (iv) GC skew (G + C/G − C) in a 1,000-bp window, (v) transposable elements (pink), and (vi) coordinates in kilobase pairs (kbp) from the origin of replication. Genes displayed in circles ii and iii are categorized by color as follows: red, iron uptake systems; orange, other putative virulence factors; yellow, bacteriocin production and immunity; pink, mobile genetic elements; dark blue, plasmid transfer; green, plasmid replication; teal, plasmid maintenance; gray, unknown.
Predicted ORFs identified on the DNA sequence of plasmid pS88 with their homologs on plasmids pAPEC-O2-ColV and pAPEC-O1-ColBM
(ii) Replication, transfer, and maintenance regions of pS88.Plasmid pS88 harbors two replicons. The first replicon region, RepFIB, contains the typical replication gene repA and the site-specific integrase int (55). The second, RepFIIA, encodes the CopB repressor and the RepA1 replication proteins. Thus, pS88 appears to belong to both the IncFI and IncFII incompatibility groups. The complete F-like transfer region of pS88 spans 31,485 bp and contains 32 genes (from traM to finO). The DNA sequences of the three regulator genes traM, traJ, and traY, relative to publicly available databases, appear to be characteristic of an F-type plasmid rather than an R1-type plasmid (15) (Fig. 2). None of the other tra genes were discriminated (as exemplified by traT), except for traS, which was closer to that of R1-type plasmids (Fig. 2). Upstream of the transfer locus lie genes involved in single-stranded DNA transfer; ssb, the gene encoding single-stranded binding protein; and psiA and psiB, plasmid SOS inhibition genes (6). Two loci putatively involved in plasmid maintenance were identified. One, close to the transfer region, was composed of the sopA and sopB genes, coding for the plasmid partition proteins (45). Located downstream of the RepFIIA replicon, the yacABC operon may represent a toxin-antitoxin plasmid stability system of which a second copy, albeit truncated, lies between the colicin V and salmochelin operons.
Phylogenetic trees generated by the neighbor-joining method for traJ, traM, traS, traT, and traY sequences of several reference E. coli conjugative plasmids. Numbers at the branches are bootstrap proportions (displayed if >75) obtained from 100 replicates. Sequences were extracted from GenBank, except for plasmid pS88: pAPEC-O2-ColV (strain APEC O2; accession number AY545598), pAPEC-O1-ColBM (strain APEC O1; accession number DQ381420), plasmid F (accession number AP001918), p1658-97 (accession number AF550679), p53638 (accession number CP001064), pMAR7 (accession number DQ388534), pRK100 (accession number AY230886), pSMS-3-5 (accession number CP000971), pUTI89 (accession number CP000244), pVM01 (accession number EU330199), and plasmid R1 (accession number X13681) and its derepressed mutant pR1-19 (accession number M19710). Plasmid pS88 and the two APEC plasmids pAPEC-O2-ColV and pAPEC-O1-ColBM are highlighted in bold.
(iii) Putative virulence region of S88.The putative virulence region of pS88 harbored three different iron uptake systems, namely, aerobactin (iucABCD and iutA), salmochelin (iroBCDEN), and the sitABCD genes (35, 53). The sitABCD genes were also chromosomally integrated. The other putative virulence genes found on pS88 were the increased serum survival gene iss, involved in complement resistance (17); the etsABC genes, encoding a putative type 1 secretion system (35); ompT, encoding an outer membrane protease (58); and hlyF, encoding a hemolysin (44). The S88 ompT gene is 100% homologous to the APEC O2 and O1 plasmid orthologs (34, 35) but differs significantly from the common E. coli chromosomal ompT gene. Therefore, we designate this gene ompTP (ompT of plasmids).
(iv) Comparison with other plasmids.BLAST comparisons of the overall pS88 sequence with other sequences revealed that pS88 is closely related to pAPEC-O2-ColV and pAPEC-O1-ColBM (Table 3). These plasmids (184,501 and 174,240 bp, respectively) came from avian pathogenic E. coli (APEC) O2:K1 and O1:K1 strains causing colibacillosis in chickens (34, 35). Alignment of the three plasmid sequences using the online comparison tool WebACT Artemis (1) is shown in Fig. 3. The line plot revealed several large blocks of highly homologous DNA between pS88 and the two APEC plasmids. Depending on the region examined, the genes and their organization were more similar to pAPEC-O1-ColBM or to pAPEC-O2-ColV. The virulence region, from the locus iroBCDEN to the locus sitABCD, appears to be highly homologous to the virulence region of pAPEC-O2-ColV (the “conserved” virulence region described by Johnson et al. [35]) and to have a highly similar organization (Fig. 3). In contrast, a block of DNA containing the loci etsABC, hlyF, and ompTP, central to the virulence region of pS88, was inverted and relocated to the end of the virulence region in pAPEC-O1-ColBM. As expected, the transfer loci were highly homologous, but in pS88 the three transfer regulator genes traM, traJ, and traY (at the beginning of the tra operon) had a higher percent identity to those of pAPEC-O1-ColBM than to those of pAPEC-O2-ColV (100% versus 85%, 99% versus 25%, and 99% versus 53%, respectively). The contrary was observed for traS (Fig. 2). Between the locus tra and the virulence region in pS88 lies a small region containing the colicin Ia and V operons and the partially duplicated yacABC operon. This contrasts with the presence at this location of a larger region in both APEC plasmids, harboring putative virulence loci such as eit and tsh (the “variable” virulence region described by Johnson et al. [35]). However, these “variable” virulence regions contain the complete ColV operon in pAPEC-O2-ColV, as in pS88, while it is truncated in pAPEC-O1-ColBM. Finally, the maintenance region downstream of the virulence region appears to be more closely related to the corresponding region of pAPEC-O1-ColBM than to that of pAPEC-O2-ColV. For example, the partition system present in pAPEC-O2-ColV (parAB) differs from that of pAPEC-O1-ColBM and pS88 (sopAB).
Comparison of plasmid pS88 (133,853 bp) with plasmids pAPEC-O1-ColBM (174,241 bp) and pAPEC-O2-ColV (184,501 bp) using the line-plot representation of homologous regions. For convenience, the traM start codon was chosen as the beginning of the three sequences. Strand conservations are indicated in blue and strand inversions in red. Genes displayed are categorized using the color scheme described in Fig. 1.Variable and conserved virulence regions are defined as previously described by Johnson et al. (35).
Molecular epidemiology of the plasmid.The distribution of 11 ORFs or genes scattered throughout pS88 in a representative collection of worldwide human meningitis strains and in several nonmeningitis E. coli strains genetically related to S88 is shown in Table 1. Strains positive for all 11 screened ORFs were also screened for eitB and tsh genes, representing the “variable” virulence region. Strains harboring all 11 pS88 genes and lacking the “variable” virulence region were considered to harbor a pS88-like plasmid. Among human meningitis strains, a pS88-like plasmid was frequently observed in the highly virulent subgroup B21 defined by a specific ST (ST29Whittam [www.shigatox.net ], ST95Achtmann [www.mlst.net ], or B2-IX of Denamur's scheme) (10, 40). Indeed, a pS88-like plasmid was found in all the ST29O45 strains and about half the ST29O18 and ST29O1 strains. Outside of this subgroup ST29/B21, a pS88-like plasmid was observed in a few (2/11) O83:K1 strains (ST697O83 and ST692O83) and never in other major-sequence O-type strains (ST304O16, ST301O7, and ST100O1). As this plasmid appears to be highly frequent in group B2 and especially in subgroup ST29/B21, we investigated several nonmeningitis human and nonhuman strains belonging to the ST29/B21 subgroup. Interestingly, the four ST29O45 APEC strains from Spain also harbored a pS88-like plasmid. Among the ST29O2 strains, the pS88-like plasmid was found in several human strains, including strain ECOR62 but not strain ECOR61; the two ST29O2 APEC strains harbored a pS88-close plasmid that was PCR negative for cia and traJF-plasmid, suggesting the presence of a pAPEC-O2-ColV-like plasmid.
PFGE of undigested DNAs of several strains representative of different sequence O types positive for the 11 pS88 genes showed that all harbored a plasmid of about 130 kb hybridizing with the iroN probe (Fig. 4).
(A) PFGE of undigested DNAs from seven representative strains harboring the 11 ORFs characteristics of pS88. Lane 1, S145; lane 2, Ben1068; lane 3, S4; lane 4, S133; lane 5, S94; lane 6, HN30; lane 7, ECOR62; lane M, molecular size marker (50-kb DNA ladder). (B) Hybridization with iroN-specific probe on a Southern blot.
Overall, our molecular epidemiology studies indicate that the pS88-like plasmid is not restricted to France, as such plasmids can be found in strains from the United States, Spain, or Sweden, and not only to meningitis strains (Table 1).
Functional analysis. (i) Plasmid curation and experimental virulence.Exposure to sodium dodecyl sulfate yielded S88 variants at a high rate (99%). One of these variants (CH7) was markedly less virulent than the wild-type strain in the animal model, with a level of bacteremia at least 2 log CFU/ml lower in two separate experiments (Table 4). To confirm that this was due to the loss of pS88, we complemented the S88 variant with pS88cat, using a double-conjugal plasmid transfer method. Virulence was fully restored with the complemented strain, designated S88E (Table 4). Comparative experiments with growth in serum showed that S88 and S88E were more resistant to serum bactericidal activity than CH7, with a mean difference of 1.1 log CFU/ml (P < 0.05) after 5 hours of incubation (data not shown).
Mean bacteremia values in a neonatal rat model challenged with S88 and its mutants, variants, and transconjugants
(ii) Mutagenesis.As we have previously shown that iroN has a role in the virulence of strain C5 (O18:K1:H7) (47), we examined its involvement, as well as that of the other two iron uptake systems (aerobactin and the sit operon), in the virulence linked to pS88. Only sitABCD was also present on the chromosome. A CH1 mutant with the salmochelin receptor gene (iroN) disrupted was slightly less virulent than the wild-type strain, with a decrease in bacteremia of 0.9 log CFU/ml (Table 4). Two other mutants were also constructed: CH19, which lacks both the chromosome- and plasmid-encoded copies of the sit operon (double deletions were obtained in one step of mutagenesis), and CH23, which lacks the aerobactin receptor. The virulence of these two mutants was similar to that of the wild-type strain (Table 4).
DISCUSSION
S88, the representative strain of the recently described French meningitis clone O45:K1:H7, harbors only a few of the specific virulence factors, such as the K1 capsule antigen and salmochelin, known to influence the onset of meningitis in newborns (13). The E. coli genomic project ColiScope, which include the sequencing of S88, provides the opportunity to discover new genetic determinants involved in the virulence of this clone. Indeed, we recently demonstrated the major role of the O-antigen gene cluster, which, surprisingly, encodes a new O antigen slightly different from the O45 reference antigen (50). Sequencing of S88 revealed the presence of a large plasmid encoding two bacteriocins: colicin V and colicin Ia. Although ColV plasmids have long been known to contribute to the virulence of ExPEC strains (3, 4, 54, 57, 64, 66), pS88 is the first ColV plasmid of a human ExPEC strain to be entirely sequenced. However, ColV plasmids have also been linked to APEC virulence, and recently an APEC ColV plasmid (pAPEC-O2-ColV) and a related plasmid (pAPEC-O1-ColBM) have been sequenced (34, 35).
Comparison of pS88 with other sequenced plasmids revealed strong homologies with pAPEC-O2-ColV and pAPEC-O1-ColBM. One of the most striking homologies involves a “conserved” virulence region in APEC plasmids. This region contains numerous known or putative virulence determinants (iroBCDEN, iss, etsABC, ompTP, hlyF, iucABCD/iutA, and sitABCD) but also ORFs of unknown function, which are all present in the three plasmids in a DNA block less than 60 kb long. Several of these virulence genes have been described as chromosomally encoded in pathogenicity islands, including the salmochelin locus in PAI III536 (16), the aerobactin locus in PAI ICFT073 (65), and the sitABCD locus in Salmonella pathogenicity island 1 (31), or on phage-related DNA, such as iss in several wholly sequenced uropathogenic E. coli (UPEC) genomes (37). In contrast, the three contiguous putative virulence determinants etsABC, ompTP, and hlyF have never been described on the chromosomes of E. coli or related species, including those examined for the ColiScope project. These genes may therefore be plasmid specific and represent a signature for the presence of a virulence plasmid in a given strain. The absence of these three genes in E. coli chromosomes suggests that they were acquired more recently than other virulence traits that have been integrated in either the chromosome or the plasmid. Their marked %GC differences (48%, 43%, and 38%, respectively) indicate that this nucleus was constructed step by step rather than being acquired “en bloc.” This is supported by the fact that hlyF can be found alone in plasmid p1658-97 (accession number AF550679) and associated with ompTP (but without ets) in plasmid pSMS-3-5 (accession number CP000971).
The other region showing strong DNA homology with APEC plasmids was, as expected, the transfer region. However, several interesting features were noted. traY, one of the three regulatory genes of the tra locus, was used initially by Boyd, and since by several other authors, to discriminate between F- and R1-type plasmids (15, 32). Phylogenetic relatedness inferred from publicly available DNA sequences of traY and the two adjacent genes, traJ and traM, clearly show that pS88 may be considered an F-plasmid, like pAPEC-O1-ColBM, while pAPEC-O2-ColV is closer to an R1-type plasmid. None of the other genes, except for traS, could discriminate, owing to their strong homology (traT for instance). Interestingly, traS was paradoxically closer to that of pAPEC-O2-ColV than to that of pAPEC-O1-ColBM. This supports the mosaic structure of the pS88 transfer locus, in keeping with the results of previous tra sequence analyses (15). In contrast to genetic and biochemical mechanisms, we know little of the environmental and physiological factors that affect conjugal transfer, even though they are likely to play an important role (23). The striking segregation of the three regulatory genes, and especially traJ, into two groups may correspond to two different regulatory systems adapted to two different environments or two different lifestyles. Of note, traJ of the R1-like plasmid is known to have a role in the virulence of the major neonatal meningitis clone O18:K1:H7 (represented by strain UTI89 in Fig. 2) (5). In this clone, TraJ contributes significantly to the survival of E. coli K1 inside professional phagocytes (27). Therefore, it would be of interest to determine whether the distantly related pS88 traJ gene (related to an F-like plasmid) has a similar or a different role.
One particular feature of pS88 is the cooccurrence of the two bacteriocins ColV and ColIa. Cells producing two bacteriocins can kill single colicin producers and have fitness advantages, as bacteriocin can be produced under different conditions of stress (25). In a collection of 1,308 animal and human E. coli strains, Jeziorowski and Gordon showed that colicin Ia and colicin V were both present in a given strain more often than would be expected by chance and that the two bacteriocin genes were carried on a single plasmid (32). Based on sequence polymorphism analysis of the two operons and their rate of occurrence, it has been postulated that this coassociation may result from the bacteriocin V operon moving onto colicin Ia plasmids (32). The first complete sequence of a plasmid harboring both bacteriocins throws light on the coevolution of the two bacteriocins. First, we found that the two bacteriocin loci were located close together. This would facilitate their horizontal cotransfer on plasmid DNA segments. Although we observed no mobile genetic elements flanking the ColV and ColIa loci, the locus yacABC upstream of the ColIa-encoding genes was partially duplicated downstream of the ColV genes, possibly indicating a recombination event in which the two colicins were acquired in a single step. Therefore, pS88, which resembles pAPEC-O2-ColV and pAPEC-O1-ColBM, may have evolved by cotransfer of bacteriocins Ia and V with a recombination event between two conserved regions, the virulence region and the transfer region. This cotransfer might have provided the host strain with a fitness advantage and have favored the spread of this plasmid to different clonal groups.
Our molecular epidemiology studies of pS88 indicates that the pS88-like plasmid is not restricted to the virulent O45:K1:H7 clone but is also distributed, albeit less frequently, in other E. coli clonal groups such as O18:K1, O1:K1, O2:K1, and O83:K1. Intriguingly, these clones (except for the less extensively studied group O83:K1) may possess the same outer membrane protein profile, designed OMP9 (2, 48, 61, 67). OMP9 is linked to a particular protein (called PCP) encoded by a large IncFI plasmid that has not been further characterized (42). Moreover, OMP9 has never been described in clones O16:K1 and O7:K1, in which we found no pS88-like plasmid. This strongly suggests that pS88 harbors the genes encoding the PCP. We also found that all B21 O45:K1:H7 strains, both avian and human, harbor a pS88-like plasmid, further supporting the genetic relatedness of avian and human O45:K1:H7 strains (10). This is consistent with other observations suggesting that avian species may be the source of neonatal meningitis strains (22, 36, 46). The monomorphism of the plasmid content observed here in B21 O45:K1 strains and of the chromosomal determinants previously observed in these strains (13) suggests recent clonal emergence. Moreover the high prevalence of pS88-like plasmids among ST29/B21 strains (40/56, 71%) compared to other STs could represent an example of fine-tuning between a plasmid and a particular genetic background. Finally, the role of pS88 does not appear be restricted to meningeal virulence, as some members of the O2:K1 clonal group, known to cause urosepsis but not meningitis (10), also harbor a pS88-like plasmid (Table 1). Thus, the virulence mechanisms of this plasmid may have implications that go beyond a particular clone or pathotype.
To determine whether plasmid pS88 is directly involved in the ability of strain S88 to induce high-level bacteremia, the main step in the pathogenesis of neonatal meningitis, we obtained an S88 variant lacking pS88. The variant produced a far lower level of bacteremia in the rat model (at least 2 log units lower). In order to confirm that the fall in virulence was due to loss of pS88, we reinserted a tagged pS88 in the cured strain S88. The resulting variant was as virulent as the wild-type strain. These results demonstrate the key role of the pS88 plasmid in the virulence of strain S88. Although numerous studies have established a link between the ColV plasmid and the virulence of various avian and human ExPEC strains (56), this is the first direct evidence supporting a role of a ColV plasmid in a specific step of neonatal meningitis, namely, sustained high-level bacteremia. This property may be due in part to the contribution of pS88 in resistance to serum bactericidal activity that we observed.
In a first attempt to understand the mechanism by which the plasmid contributes to the virulence of S88, we examined the three iron uptake systems encoded by the plasmid, namely, the siderophores salmochelin and aerobactin and the Sit system. Iron is essential for bacterial growth, but its availability within the host is limited (26). We have previously shown that salmochelin is involved in the high-level bacteremia achieved by the representative O18:K1:H7 strain C5 (47), in which this factor is chromosome borne. Hence, we examined whether this siderophore was also involved in the virulence of strain S88. We also investigated the respective virulence roles of the other two plasmid-encoded iron uptake systems, namely, aerobactin and the Sit system. Aerobactin is the most frequent siderophore in neonatal meningitis E. coli (NMEC) strains, after yersiniabactin (18, 13). Although this iron uptake system has been implicated in virulence in several experimental models (21, 24, 43, 60) its role in neonatal meningitis remains to be determined. The sitABCD operon, encoding a member of the family of ATP-binding cassette (ABC) proteins initially described for Salmonella enterica, was first functionally characterized as a manganese and iron transporter in an APEC strain (53). This newly described system is required for full virulence of Salmonella enterica serovar Typhimurium and APEC (31, 52). However, little is known of its distribution in human ExPEC strains or its role in their virulence (36, 42, 52).
As previously reported with strain C5 (47), we found that the virulence of strain S88 was attenuated when iroN was lacking. In contrast, the absence of aerobactin and the Sit system had no discernible influence on the virulence of strain S88. Our results thus prove the key role of salmochelin in the bacteremic step of E. coli meningitis. However, the lower virulence of the iro-defective mutant (∼1 log CFU/ml lower) cannot alone explain the marked loss of virulence observed with the variant lacking the whole plasmid (>2 log CFU/ml). The fact that the loss of the other two siderophores did not influence the level of bacteremia does not totally exclude a role in the virulence of S88, as the iro locus may compensate for their loss. Further studies with double and triple mutants are under way. Alternatively, these siderophores may be involved in steps other than sustained bacteremia, such as gut translocation or cerebrospinal fluid invasion. Other iron uptake systems such as yersiniabactin, enterobactin, and the FecABCD system exist on the S88 chromosome (www.genoscope.fr ) and may be able to compensate for the loss of the plasmid-encoded systems. Finally, it is likely that the virulence of the ColV plasmid involves other determinants. The pS88 plasmid carries two genes, iss and traT, which play a role in serum survival and resistance to macrophage phagocytosis, respectively. Several other pS88 coding sequences have no homologs and could also encode new virulence determinants.
In conclusion, pS88 shares a similar structure with avian virulence plasmids, pointing to a possible common ancestor. It is an essential meningeal virulence determinant comparable to the K1 capsule and O45S88 somatic antigens and is not restricted to the French O45:K1:H7 clone. A comprehensive study of the molecular mechanisms responsible for its virulence may have important implications, possibly beyond the pathogenesis of neonatal meningitis.
FOOTNOTES
- Received 31 October 2008.
- Returned for modification 12 January 2009.
- Accepted 15 March 2009.
- Copyright © 2009 American Society for Microbiology
