ABSTRACT
Enteroaggregative Escherichia coli (EAEC) is an important cause of diarrhea worldwide. We analyzed 17 Danish EAEC strains, isolated in the course of a case control study, for phenotypic and genotypic properties. The strains belonged to at least 14 different serotypes. Using PCR to investigate the prevalence of various putative virulence genes, we found that all but two strains were typical EAEC, as they harbored all or part of the previously described AggR regulon. The majority of the strains harbored genes encoding aggregative adherence fimbriae (AAF). The most common was AAF/I, found in nine strains; eight strains carried no known AAF-related genes. We utilized TnphoA mutagenesis to localize the aggregative adherence (AA) adhesin from one typical EAEC strain, C1010-00, which lacked a known AAF. We identified a TnphoA insertion in a hypothetical Dr-related pilin deposited in GenBank as HdaA. Four additional Danish strains harbored HdaA, and all but one displayed AA to HEp-2 cells. By using PCR primers derived from the pilins and ushers from the three AAF and Hda, we found that 16 of 17 strains exhibited evidence of one of these factors; importantly, the one negative strain also lacked the aggR gene. Cloning of the complete Hda gene cluster and expression in E. coli DH5α resulted in AA and complementation of the C1010-00 nonadherent mutant. Four related adhesins have now been found to confer AA in typical EAEC strains; our data suggest that, together, these variants may account for AA in the large majority of strains.
Enteroaggregative Escherichia coli (EAEC) has emerged as an important pathogen in several clinical scenarios, including traveler's diarrhea (1, 2, 18); pediatric diarrhea, which is endemic among children in industrialized (63) and developing countries (42); and persistent diarrhea among human immunodeficiency virus (HIV)-infected patients (12, 17, 31, 67, 68). A recent study suggested that it may be the most common bacterial cause of diarrhea in the United States (37). Nevertheless, much of the epidemiology of EAEC infection remains unknown, e.g., transmission, the distribution of types in populations worldwide, and the significance of the great heterogeneity of virulence genes among isolates (21).
EAEC strains are defined as E. coli strains which do not secrete the heat stable or heat labile enterotoxins of enterotoxigenic E. coli and are recognized by their characteristic “stacked-brick” patterns or aggregative adherence (AA) to HEp-2 cells in culture (36). It is likely that this definition encompasses a spectrum of clones, spanning from the highly pathogenic (likely represented by the globally most common serotypes, such as O126:H27 and O44:H18) to nonpathogenic E. coli strains (24), though the molecular basis for any such distinction is not fully understood.
Pathogenesis and molecular studies suggest the presence of a package of coinherited virulence-related genes carried by chromosomal islands and virulence plasmids (reviewed in reference 33). Included in this set of EAEC-specific genes is a large array of factors controlled by the global regulator called AggR. AggR is encoded on a large virulence plasmid called pAA, which encodes most of the previously described EAEC factors. Strains carrying elements of the AggR regulon have been termed typical EAEC (51), though which of these factors is responsible for pathogenicity remains unknown. Recently, a cluster of AggR-activated genes was identified on the chromosome of strain 042; these genes have a type VI secretion system and are carried within a chromosomal island called AAI (11). Of note, a DNA probe (often called CVD432), derived empirically from the pAA virulence plasmid of EAEC, has been reported to be specific for this pathotype (3). We have shown that this probe corresponds to the aatA gene, which is required for the surface translocation of dispersin, a common EAEC protein required for the proper function of fimbrial structures called aggregative adherence fimbriae (AAF) (41). Both of these factors are under AggR control.
The prevailing pathogenic paradigm for EAEC includes colonization of the intestinal mucosa followed by the elaboration of one or more cytotoxins and enterotoxins, of which three have been described (15, 16, 19, 20, 30, 46, 52, 66): Shigella enterotoxin 1 (ShET1), the plasmid-encoded toxin (Pet), and the enteroaggregative E. coli heat-stable toxin (EAST1). Interestingly, none of these three are under the control of the AggR regulator, though they are found more commonly among aggR-positive strains than aggR-negative strains (23).
A key step in the pathogenesis of nearly all bacterial enteric pathogens is the colonization of the gastrointestinal mucosa, which is mediated by specific adherence factors. EAEC prototype strains adhere to HEp-2 cells and to the intestinal mucosa by virtue of AAF (5, 9, 34, 38, 39, 53), which are encoded on pAA plasmids; all AAF biogenesis genes are controlled by AggR. AAF are distantly related to the Dr family of adhesins (55), the biogenesis of which requires a dedicated periplasmic chaperone, an outer membrane usher protein, and two surface-expressed subunits (a major subunit and a putative cap subunit). AAF and Dr adhesins display a high level of conservation of the usher and chaperone genes but greater divergence of the fimbrial subunit genes (55).
Three distinct AAF variants have been described based on the sequences of the major fimbrial subunits. The subunits of AAF/I, AAF/II, and AAF/III are encoded by the aggA, aafA, and agg3A genes, respectively (5, 9, 34). Some diversity in the alleles within the AAF/I (49) and AAF/II (7) pilin subunit variants has also been described. However, in most epidemiologic studies, the majority of EAEC strains do not express any of the three known AAF variants, despite the presence of the AggR regulator and other pAA-borne genes. This suggests to us that additional undiscovered AAF variants may exist (9, 10, 13). Given the relevance of surface adhesins to enteric vaccine development, a more complete characterization of EAEC fimbrial alleles is necessary.
We focused our study on a collection of EAEC strains from Denmark. In such industrialized countries, diarrhea-related mortality is low. Nevertheless, the social burden and economic cost as a result of the care of ill children and the subsequent parental absence from work are considerable (44). To clarify the infectious etiology of diarrhea among Danish children less than 5 years old, Olesen et al. conducted a 2-year prospective case control study (44). Nineteen EAEC strains were isolated in the course of this investigation: seven strains were derived from domestic cases, four strains from cases acquired through overseas travel, and eight strains from the stools of healthy controls. EAEC strains were isolated more frequently from cases than from controls, but the differences were not statistically significant (44).
In this study, we included 17 EAEC strains from the Danish case control study based on their positive reactions with the CVD432 probe. We eliminated from our analysis cases where other pathogens were isolated. We evaluated the presence of EAEC-specific genes among these EAEC strains, with emphasis on the fimbrial loci. One strain isolated from a case of diarrhea, designated C1010-00, was not found to express a known AAF allele, despite the presence of other AggR-related loci. Further investigation revealed the presence of genes encoding a previously unpublished adhesin termed Hda. Hda was found in several other strains isolated from Denmark, suggesting that this adhesin represents an important variant.
MATERIALS AND METHODS
Bacterial strains, plasmids, and growth conditions.The bacterial strains and plasmids that were used in this study are described in Table 1. The 17 EAEC strains were isolated in the course of a recent case control study of diarrheal illness in Denmark (44) and currently compose part of the collection of strains from the Statens Serum Institut (SSI), Denmark. Each strain was isolated as the sole pathogenic bacterium in the patient. Several additional well-characterized strains were obtained from the collections of the SSI and the Center for Vaccine Development of the University of Maryland School of Medicine.
Bacterial strains and plasmids used in this work
Stock cultures were frozen at −80°C in Luria broth (LB) or SSI broth containing 10% (vol/vol) glycerol. All strains were grown at 37°C. Streptomycin (30 μg ml−1), tetracycline (15 μg ml−1), kanamycin (50 μg ml−1), chloramphenicol (20 μg ml−1), nalidixic acid (50 μg ml−1), and 5-bromo-4-chloro-3-indolylphosphate (XP; 50 μg ml−1) were added when indicated. Serotyping of the EAEC strains was performed at the SSI using standard methods (45).
PCR.All primers and corresponding product sizes are listed in Table 2. The DNA template was obtained by InstaGene matrix purification (Bio-Rad Laboratories A/B, Copenhagen, Denmark) according to the manufacturer's instructions. The PCR cycles comprised (i) denaturation for 2 min at 94°C, (ii) denaturation for 30 s (depending on the size of the product, with a 30-s increase for each 500 bp) at 94°C, (iii) annealing for 1 to 2 min at the primer-specific temperature (aap, 55°C; aggR, 50°C; aatA, 58°C; aggA, 50°C; aafA, 50°C; agg3A, 50°C; hdaA, 59°C; pet, 62°C; setA, 59°C; cdt, 57°C; and astA, 56°C; all usher gene primers annealed at 55°C), and (iv) extension for 1.5 min at 72°C (depending on the product size) with 35 cycles of step ii. The final extension was for 10 min at 72°C.
List of primers and probes used in this work
Single-primer PCRs were run as previously described (25) by using the LA PCR kit (version 2.1; TaKaRa Bio, Inc., Shiga, Japan). A single-primer PCR is based on the use of a specific primer from one DNA strand, which nonspecifically primes from the opposite strand at low temperatures. The reactions were amplified in 50 μl of the following PCR mixture: 0.025 units of Taq polymerase, 1× LA PCR buffer II (25 mM Mg2+), 1,600 μM deoxynucleoside triphosphate (400 μM each), 1 μg of template DNA, 25 pmol of primer 1, 25 pmol of primer 2, and distilled water. The PCR conditions were denaturation for 1 min at 94°C; 20 cycles of denaturation for 30 s at 94°C, annealing for 30 s at 54°C, and extension for 15 min at 68°C; 30 cycles of denaturation for 30 s at 94°C, annealing for 30 s at 30°C, and extension for 15 min at 68°C; and 30 cycles of denaturation for 30 s at 94°C, annealing for 30 s at 54°C, and extension for 15 min at 68°C. The final extension was for 10 min at 68°C.
RNA extraction and reverse transcriptase PCR (RT-PCR).mRNA was quantitated from EAEC C1010-00 and C1010-00aggR::pJP5603 grown in Dulbecco's modified Eagle's medium (DMEM) containing 0.5% glucose. Whole-cell RNA was isolated from a 1-ml sample of the bacterial cultures using the RNeasy kit (Qiagen, Inc., Valencia, CA) according to the manufacturer's instructions. cDNA was synthesized from 500 ng of bacterial RNA by using random hexamer primers and the Thermoscript RT enzyme (Invitrogen, Inc., Carlsbad, CA). The PCR was performed using 2 μl of cDNA. The primers used are listed in Table 2.
Antibiotic resistance gene tagging of the pAA plasmid.To introduce an antibiotic resistance gene into the pAA plasmid of strain C1010-00 (herein named pAA4), we inserted suicide plasmid pJP5603 into the aggR gene of the plasmid by single crossover homologous recombination as previously described (13). For this experiment, E. coli S17-1λpir was employed as the donor strain. An internal fragment of the aggR gene from strain 042 was cloned into suicide plasmid pJP5603 (previously constructed by I. Henderson and J. Nataro [unpublished data]).
Transposon TnphoA mutagenesis.Insertion mutagenesis of strain C1010-00 was performed by using transposon TnphoA (28, 34, 62), which was delivered on suicide plasmid pRT733 as previously described (58, 62). Strain SM10λpir(pRT733) was cross-streaked against Danish EAEC strain C1010-00 on Luria agar plates lacking any antibiotics; mating occurred at 37°C overnight. The growth from each plate was harvested and plated on Luria agar containing streptomycin and kanamycin and 50 μg of indicator XP (Sigma Chemical Co., St. Louis, MO). This analysis strategy yielded 200 alkaline phosphatase-expressing fusions as a result of 31 separate matings. The fusions were numbered so that the products of different matings could be identified. The 200 fusions were screened for their ability to form a biofilm in DMEM containing 0.5% glucose, which has been shown to correlate with the expression of AAF adhesins (57).
The quantitative measurement of alkaline phosphatase activity in hdaA::phoA fusions under the control of AggR was carried out by transforming plasmid pBADaggR (56) into C1010-00hdaA::phoA and selecting for ampicillin- and kanamycin-resistant colonies. A total of 100 μl of overnight bacterial culture was added to 70 ml of LB and grown to an optical density at 600 nm (OD600) of 0.5. Either glucose or arabinose was added to separate aliquots in triplicate to a final concentration of 0.2%, and the cultures were incubated at 37°C for 3 h (to an OD600 of 1.7). The cells were collected by centrifugation, 50 μl of the pellet was resuspended in 0.9 ml of buffer AP (0.1 M Tris-HCl [pH 9.5], 0.1 M NaCl), and 100 μl of the substrate solution consisting of 0.2 M p-nitrophenol phosphate was added (N6876; Sigma). The reaction proceeded for 40 s at 37°C until a color change was observed. The reaction was stopped by the addition of 100 μl of solution containing 0.8 M KH2PO4 and 0.1 M EDTA. The cells were separated by centrifugation at 13,000 rpm for 2 min, and the supernatant was read spectrophotometrically at 410 nm. E. coli strain HS was used as a blank. Enzyme activity was expressed in arbitrary units of OD410/ml of culture/min.
Identification of TnphoA insertion sites.Plasmid DNA was isolated from TnphoA mutants by using the mini-AX kit (A&A Biotechnology, Gdynia, Poland) according to the manufacturer's instructions. Plasmid DNA was digested with BamHI or SalI and ligated into the corresponding site of pUC18. The ligation reaction was used to make E. coli strain Epi100 kanamycin resistant. The plasmid DNA purification of chimeras was performed by using the Midi plasmid kit (Qiagen, Inc., Valencia, CA).
Cloning of Hda.The Hda gene cluster was amplified from strain C1010-00, using a PCR with Pfx Platinum polymerase (Invitrogen, Inc., Carlsbad, CA). KpnI and HindIII restriction sites were introduced at the 5′ ends of the PCR primers, whose sequences were as follows: forward, 5′-GCG CGG TAC CTG TAG GAA GGA GAT ATA CAT ATG ACA CAG ATG ACT TCT A; reverse, 5′-GCG CGC AAG CTT TTA ATT ACT CCA AGT GGT CAA GTT, respectively. The PCR products were digested with the respective restriction enzymes and ligated into the corresponding sites of pUC18.
DNA colony hybridization.The DNA probes used for the colony hybridization are described in Table 2. The PCR products employed in the probe studies were labeled with digoxigenin, according to the manufacturer's instructions (Roche Diagnostics, A/S, Hvidovre, Denmark). The K-12 strain, MC1061, was used as a negative control. The colony blot protocol has been described previously (60).
HEp-2 adherence assay.The HEp-2 adherence assay was performed as previously described (36, 64) with modifications. Briefly, cells were grown overnight to 50% confluence in Eagle's minimal essential medium (plus HEPES and gentamicin [4 mg/100 ml]) and 10% fetal calf serum on Sonic Seal slides (Nunc Intermed, Biotech Line A/S, Denmark) at 37°C. Bacteria were grown overnight without shaking in LB. A total of 50 μl of the overnight culture was added to 5 ml DMEM containing 10% fetal calf serum and 1% d-mannose and then applied to the monolayers for 3 h at 37°C. After incubation, the cells were washed, fixed with 10% (vol/vol) formalin for 10 min, and stained with crystal violet (Sigma Chemical Co., St. Louis, MO) for 5 min. The cells were visualized under light microscopy (64), and adherence was determined as previously described (36).
Biofilm assay.The EAEC biofilm assay was performed as previously described by Sheikh et al. (57). Briefly, cultures were grown in DMEM supplemented with 0.45% glucose (high-glucose DMEM). The biofilm assays were performed by using either polystyrene Costar 24-well culture dishes with glass coverslips or Sonic Seal slides (Nunc Intermed). Two milliliters of high-glucose DMEM containing 1% d-mannose was added to each well simultaneously with 40 μl of the bacterial overnight culture. The culture dishes were incubated overnight at 37°C. In addition, 1 mM of IPTG (isopropyl-β-d-thiogalactopyranoside) was added to the wells to induce recombinant gene expression from DH5α(pNBO1), HB101(pJPN45)(pNBO1), and C1010-00hdaA::phoA(pNBO1). After 20 h, the culture medium was aspirated and the substratum was washed three times with phosphate-buffered saline (PBS). Biofilm formation was visualized by first fixing the bacteria with 10% (vol/vol) formalin for 10 min, followed by staining with 300 μl 0.5% crystal violet (Sigma Chemical Co., St. Louis, MO) for 5 min, washing once with PBS, and then air drying. The cover slides were removed from the wells and mounted on microscopic slides. Biofilms were observed by inverted light microscopy as previously described. Biofilm formation was quantified spectrophotometrically by adding 1 ml of 70% ethanol to the wells to solubilize the crystal violet stain; after a 5-min incubation at room temperature, absorbance was determined at 470 nm (Ultrospec 2100 pro; Amersham Bioscience, Piscataway, NJ) (57).
Hemagglutination of red blood cells by whole bacteria.Hemagglutination (HA) has been shown to correlate with AAF adhesin expression by EAEC strains (9, 47). Fresh erythrocytes were obtained from the senior author (J.P.N.), and the HA assay was performed as previously described (34). HA was similarly performed using sheep, rabbit, guinea pig, and bovine erythrocytes as described previously (34) with some modifications. E. coli was grown overnight in LB without shaking. A total of 100 μl of the bacterial culture was added to 10 ml of high-glucose DMEM and incubated for 4 h at 37°C; the bacteria were pelleted by centrifugation and then washed twice in sterile PBS. The pellets were resuspended in PBS to a concentration of 108 CFU/ml (estimated by McFarland nephelometry). Serial twofold dilutions of the bacterial suspension in PBS were prepared; 100 μl of each dilution was mixed with an equal volume of a 3% (vol/vol) erythrocyte suspension containing 1% d-mannose (Sigma Chemical Co., St. Louis, MO) in PBS. The solution was briefly mixed in a 24-well tissue culture dish and allowed to stand for 20 min at 4°C. HA was evaluated under ×20 magnification with an inverted microscope. Any degree of HA discerned to be greater than that of the negative bacterium-free control by two investigators in a blinded manner was considered to be positive. The highest dilution of the bacterial suspension producing HA was recorded for each strain tested.
Nucleotide sequence and phylogenetic analyses.BLAST searches and comparisons were conducted using the databases of the National Center for Biotechnology Information (http://www.ncbi.nlm.nih.gov/BLAST/ ). The primer used for the nucleotide sequence analysis of the C1010-00hdaA::phoA insertion was 5′-CAGTAATTTCCGAGTCCCCCATCC-′3. Sequencing was performed at MWG-Biotech AG.
The sequences of the Dr pilins were compared with that determined for the predicted C1010-00 HdaA pilin after removal of the putative signal sequence predicted by the SignalP algorithm (http://www.cbs.dtu.dk/services/SignalP/ ). Dr pilins truncated for their proven or predicted signal sequences were analyzed using ClustalW (http://www.ebi.ac.uk/Tools/clustalw/index.html ) at default settings. The Clustal alignment was subjected to neighbor-joining analysis using Phylip software available at http://bioweb.pasteur.fr/seqanal/phylogeny/phylip-uk.html . A consensus tree was calculated from 100 replicates and plotted using Phylodendron software (http://iubio.bio.indiana.edu/treeapp/treeprint-form.html ).
Nucleotide sequence accession numbers.Sequences were deposited in GenBank under accession number EU637023. The complete sequence of plasmid pO86A1 was also deposited (GenBank accession number NC008460).
RESULTS
Characterization of Danish EAEC isolates.Seventeen EAEC strains isolated in the course of a previously published case control study of Danish pediatric diarrhea patients (44) were selected for further characterization. All EAEC strains were positive with probe CVD432, carrying the aatA gene (3). Strains were obtained from children with diarrhea (nine) and from healthy children (eight).
The strains belonged to 14 different serotypes (Table 3); these strains included common EAEC serotypes such as O44:H18 and O86:NM (43), as well as several rare serotypes. No single serotype was found in a large proportion of strains. Strain C1046-00 belonged to serotype O44:H18, which has been implicated in diarrheal outbreaks in the United Kingdom and to which prototype EAEC strain 042 belongs (54, 59). One strain (C1059-00) was serogrouped as O73, which is known to cross-react with the O44 antigen. The most frequent O group was O86 (in four strains).
Origin, serotype, adherence, and genotyping results from the 17 Danish EAEC strainsa
We employed PCR to determine the frequencies of previously described chromosomal and plasmid-borne putative EAEC virulence genes and other E. coli virulence genes in our Danish collection. The distribution of these genes among the EAEC isolates is shown in Table 3. Several patterns are notable. As EAEC strains in this study were detected using the probe carrying the aatA gene, all strains were positive for aatA by PCR. As expected, all but two of the strains were positive for the gene encoding AggR (which controls aatA) (41), with perfect concordance between aggR positivity and the presence of the aap gene, which encodes the surface protein called dispersin (56); dispersin translocation requires the aatA product (41). The cosegregation of these plasmid-borne genes has previously been reported (10) and may correlate with highly virulent EAEC strains (50).
Twelve of the 17 strains carried genes encoding either the ShET1 or EAST toxins, though none harbored the pet gene. As suggested in a previous report, we observed a high concordance between the presence of ShET1-encoding genes and those encoding the AggR-regulated type VI secretion system encoded by the aai cluster (11); these two loci are adjacent on the AAI chromosomal island inserted at pheU in strain 042.
Adhesive properties of EAEC strains.The results of the HEp-2 adherence and biofilm assays for the Danish strains are presented in Table 3. Despite the presence of genes defining typical EAEC strains, three isolates, C790-00, C1046-00, and C1082-00, did not exhibit the AA pattern, which is still used to define the pathotype, and therefore, these strains cannot be considered EAEC sensu stricto. Notably, strains C790-00 and C1082-00 did not harbor the aggR gene, suggesting that perhaps the defect in adherence is due to the absence of a functional transcriptional activator and/or a functional adhesin. Among the adherent strains, the AA patterns varied, ranging from the typical honeycomb formation as exhibited by strain 042 to a very weak adherence phenotype. Strain C1010-00 exhibited an AA pattern but did not form a honeycomb on the glass substratum.
The biofilm-forming properties of all 17 EAEC isolates were determined after 20 h of incubation (Table 3). The presence of biofilm morphology did not correlate with the expression of specific fimbrial variants. Strains C809-00, C247-01, and C390-01 did not adhere to glass coverslips.
Adherence factors of the EAEC collection.Nine Danish EAEC isolates were found to be positive for AAF/I-encoding genes, making it the most common adhesin variant. PCR using primers generated from the published pilin sequences of AAF/II and AAF/III did not yield products for any of the Danish strains. Therefore, consistent with previous reports, genes representing known AAF pilin alleles were found in only a minority of the strains; strains C790-00, C1010-00, C1059-00, C1082-00, C246-01, C247-01, C252-01, and C254-01 carried no known AAF pilin genes. The results of the PCRs were confirmed by colony dot blotting using probes specific for AAF/I, AAF/II, and AAF/III pilins (data not shown). Therefore, we hypothesized that these isolates harbored either distant homologues of known AAF alleles or AAF pilins that were not previously described.
In contrast to the results generated with the pilin gene primers, strains C246-01, C247-01, C252-01, and C254-01 yielded PCR products using the primers derived from the usher-encoding gene agg3C from the AAF/III cluster. To confirm this result, we sequenced the PCR product from each of the four strains. However, the sequences of the PCR products from strains C246-01 and C247-01 revealed 98% and 82% predicted amino acid identity, respectively, with aafC (the usher-encoding gene of AAF/II), while the products of strains C252-01 and C254-01 yielded 88% and 98% identities, respectively, with the predicted protein product of agg3C (encoding the usher of AAF/III). Of note, the AAF/II and AAF/III ushers have previously been reported to be phylogenetically related (5), thus explaining the former's reactivity with the agg3C primers.
In an effort to generate additional sequences for some of these strains, we performed single-primer PCR extending outwards from the usher regions. By nucleotide sequence analysis, we found that C252-01 and C254-01 harbored the agg3D gene flanking agg3C; the sequence for C254-01 revealed that agg3C and agg3D were bordered by a gene predicted to encode the transposase of IS629 (GenBank accession number AJ85189). We therefore consider these four strains to be positive for AAF/II- and AAF/III-related genes, reducing the number of strains negative for all known AAF-encoding genes to 4 of the 17. The single-primer PCRs for C246-01 and C247-01 did not generate products.
HA properties of EAEC strains negative for known AAF.AAF have been shown to confer HA to erythrocytes from different animal species (34); this agglutination varies according to the AAF adhesin variant (47). All of the Danish isolates (except strain C247-01) that were negative for known AAF exhibited HA to both human and bovine erythrocytes (Table 4). HA for bovine erythrocytes was particularly strong for strain C1010-00.
HA phenotypes of Danish EAEC strains and E. coli K-12 constructs
Identification of the C1010-00 hdaA gene.In order to identify putative AAF adhesins other than AAF/I, AAF/II, and AAF/III, we selected one isolate, C1010-00, for further study. C1010-00 was isolated from a two-year-old girl hospitalized with diarrhea and was included in the Danish case control study (44). This isolate harbored the genes encoding AggR, AatA, and dispersin, as well as the chromosomal genes aaiC and setAB, thus suggesting that it carries both the complete pAA virulence plasmid and the AAI island (11). C1010-00 was found to carry one large plasmid of >100 kb (data not shown). To identify the adhesin of strain C1010-00, a TnphoA insertion library was constructed. Two hundred phoA-expressing mutants were screened for loss of the ability to form a biofilm.
The phoA insertion of one biofilm-deficient mutant was sequenced and localized to an open reading frame (ORF) that bore 90% identity to a gene designated hdaA (GenBank accession number YP787988). HdaA is situated on plasmid pO86A1. No experimental details for this plasmid were available as of this writing. Importantly, the Hda gene cluster was so named from “HUS-associated diffuse adherence” and ascribed to a clinical isolate termed DIJ1.
Several additional proteins revealed significant homology with the hdaA gene product. Blast analysis suggested 61% identity to M-agglutinin, an afimbrial adhesin encoded by a gene from the bma gene cluster found in some uropathogenic E. coli strains (GenBank accession number AAA23523) (48). The analysis also reported 57% amino acid identity with the product of the afaE gene, encoding the major subunit of the Dr adhesin designated AfaE-VIII; afaE is a component of the afa-8 gene cluster (GenBank accession number AAD44024) (26). AfaE-VIII is commonly produced by E. coli strains isolated from animals with diarrhea and septicemia and by human-derived isolates associated with extra-intestinal infections (27). Upstream of the putative pilin gene in strain C1010-00, our analysis predicted another ORF, displaying 98% identity with hdaB, also carried on plasmid pO86A1 (GenBank accession number BAF33891). This ORF protein exhibited 54% predicted amino acid identity with the minor pilin subunit protein Agg3B (GenBank accession number AAM88297), whose gene is part of the AAF/III gene cluster (5).
Primers were constructed from each gene of the hda gene cluster as deposited in GenBank (accession numbers BAF33891, BAF33890, BAF33889, and BAF33888); the genes are predicted to encode the putative major pilin, minor pilin, usher, and chaperone. Strain C1010-00 yielded PCR products from primers designed from each predicted hda gene, suggesting that this isolate harbored the complete Hda gene cluster.
To determine the prevalence of the hdaA gene among the Danish EAEC strains, we derived primers from our nucleotide sequence and performed PCR on our collection. Isolates C1059-00, C1082-00, C246-01, and C254-01 also harbored the hdaA gene. As shown in Table 3, these strains were negative for AAF/I to AAF/III genes but were positive for other EAEC genes, and all exhibited AA to HEp-2 cells (data not shown), biofilm formation, and HA (Table 4). Isolate C246-01 also harbors the aafC gene, whereas C254-01 carried agg3C and agg3D, indicating that these isolates might contain more than one AAF gene cluster.
Characterization of the Hda gene cluster.We sought to elucidate the phenotypes attributable to the pAA plasmid of strain C1010-00 by introducing the plasmid (designated pAA4) into E. coli HB101. pAA4hdaA::phoA was first introduced into both HB101 and HB101(pJPN45), the latter expressing the full-length aggR gene on high-copy-number plasmid pJPN45. No AA or biofilm formation was conferred on these strains, suggesting that pAA4 carrying a mutation in hdaA was not sufficient to mediate adherence. To transfer the plasmid with an intact Hda locus, pAA4 was tagged with antibiotic resistance by the single-crossover homologous recombination of suicide plasmid pJP5603 (kanamycin resistant) into its aggR gene. AggR function was complemented by transforming pAA4aggR::pJP5603 into HB101(pJPN45).
HB101(pAA4aggR::pJP5603)(pJPN45) formed a biofilm on glass coverslips and adhered to HEp-2 cells in a phenotype similar to isolate C1010-00 (Fig. 1A). In addition, HB101(pAA4aggR::pJP5603)(pJPN45) induced strong HA of bovine and human erythrocytes, whereas neither HB101(pAA4aggR::pJP5603) nor HB101(pJPN45) expressed this phenotype (Table 4).
Biofilm formation on glass surfaces after 20 h of growth. (A) C1010-00; (B) C1010-00hdaA::phoA; (C) C1010-00hdaA::phoA(pNBO1); (D) DH5α(pNBO1). Scale bars = 50 μm. The inset in panel D shows aggregates of bacteria.
In order to confirm that the Hda adhesin of strain C1010-00 is under the control of the AggR transcriptional activator, we transformed plasmid pBADaggR (56) into C1010-00hdaA::phoA and quantified the alkaline phosphatase expression. Though C1010-00hdaA::phoA carries a functional aggR gene, we performed these experiments under conditions known to repress native aggR and activated the heterologous aggR via the addition of arabinose to the medium. As predicted, heterologous aggR expression increased the expression of the hdaA::phoA product (Fig. 2A). We confirmed these results by RT-PCR: C1010-00 yielded the hdaA message under aggR-inducing conditions, whereas this message was undetectable in aggR mutant C1010-00aggR::pJP5603 (Fig. 2B).
The hdaA gene is under the control of AggR. (A) Alkaline phosphatase activity assay. C1010-00hdaA::phoA was complemented with pBADaggR, and the alkaline phosphatase activity was measured under repressing conditions (0.2% glucose) or activating conditions (0.2% arabinose). Shown are the mean absorbances of the results from three independent experiments with error bars representing one standard deviation. (B) RT-PCR for the hdaA transcript. RNA was extracted from EAEC strains C1010-00 (lane 1) and C1010-00aggR::pJP5603 (lane 2) and subjected to reverse transcription and cDNA amplification by PCR for hdaA (top) or the constitutive chloramphenicol acetyltransferase gene (cat; bottom).
To determine the phenotype conferred by the Hda product, the complete Hda cluster of C1010-00 (comprising the predicted chaperone, usher, invasin, and pilin genes) (Fig. 3) was cloned into pUC18, yielding plasmid pNBO1. pNBO1 was transformed into E. coli DH5α, HB101(pJPN45), and the C1010-00hdaA::phoA mutant. HB101(pJPN45)(pNBO1), C1010-00hdaA::phoA(pNBO1) (Fig. 1C), and DH5α(pNBO1) (Fig. 1D) exhibited heavy biofilm formation typical for EAEC, as did HB101(pJPN45)(pNBO1) (data not shown); the biofilms comprised thick aggregates with classic AA and an abundant stacked-brick pattern on glass surfaces (Fig. 4). Importantly, no aggregation was observed for DH5α (data not shown) and C1010-00hdaA::phoA (Fig. 1B). The results of the HA assays with bovine and human erythrocytes correlated with the biofilm results (Table 4).
Annotation of the hda biogenesis gene cluster as determined by nucleotide sequence analysis of the product amplified from strain C1010-00. Gene designations follow from GenBank accession number EU637023. All ORFs encoding >50 predicted amino acids are indicated. The nucleotide positions of the predicted translational start and stop codons are indicated below the line. The inverted triangle indicates the position of the TnphoA insertion into the hdaA gene of C1010-00, which corresponds to nucleotide 378 downstream of the predicted hdaA translational start codon.
Quantitation of biofilm formation by Hda-expressing clones. Bacteria were cultivated in DMEM-0.45% glucose for 20 h at 37°C in 24-well dishes. Biofilms were fixed and stained with crystal violet, and then the stains were solubilized and quantitated spectrophotometrically at 470 nm. The bars represent the means of the results from triplicate wells; error bars indicate one standard deviation. 1, C1010-00hdaA::phoA(pNBO1); 2, HB101(pJPN45)(pNBO1); 3, DH5α(pNBO1); 4, C1010-00; 5, HB1010(pJPN45); 6, C1010-00hdaA::phoA; 7, DH5α.
DISCUSSION
EAEC is an increasingly recognized diarrheal pathogen. The pathotype has been associated with endemic pediatric diarrhea worldwide and has been implicated in numerous diarrhea outbreaks. EAEC has recently been shown to be the second most common cause of traveler's diarrhea, after only enterotoxigenic E. coli (2, 61). However, EAEC pathogenesis is not fully understood, partly because there is substantial genetic heterogeneity among EAEC isolates (6, 23, 32).
We studied a collection of EAEC strains isolated in the course of a case control study in Denmark (44). The strains were characterized by serotype and by the presence of various phenotypes and genotypes potentially relevant to virulence. In agreement with previous reports (10), the 17 Danish EAEC strains we studied were heterogeneous with regard to plasmid-borne and chromosomal genes. Nevertheless, as previously reported, our data also suggest that certain factors are coinherited, presumably because they are coordinately regulated by the transcriptional activator AggR (as is known) and/or because they act in concert to execute a common pathogenic strategy. At the same time, however, studies have shown variable frequencies for several of the genes included in this study (42). The significance of this observation remains unknown, but geographic variation among EAEC strains may be expected. Nonetheless, the coinheritance of plasmid-borne and chromosomal EAEC virulence genes is similar to observations made for other diarrheagenic E. coli pathotypes, specifically enteropathogenic, enteroinvasive, and enterohemorrhagic strains (35).
EAEC strains belong to a diverse range and combination of O:H serotypes (47, 65, 69). Some EAEC strains, however, do not express recognizable O antigens (such as C1010-00) and may be nonmotile. The Danish strains were distributed over 16 different serotypes and included nonmotile and nontypeable strains. However, some strains belonged to common EAEC serotypes, including O44:H18 and EAEC O groups such as O86 and O111 (43).
Colonization of the intestinal mucosa is thought to be the first step in the pathogenesis of nearly all bacterial enteric pathogens. We have previously shown that the EAEC AAF confer adherence to the human intestinal mucosa (9), suggesting that they are important virulence factors. Moreover, our data show that antibodies to the AAF abolish EAEC adherence (J. Sheikh and J. Nataro, unpublished data), suggesting their potential use as EAEC immunogens. For this reason, we have embarked on an effort to identify the full complement of AAF variants among a collection of EAEC strains. Heretofore, three main variants of AAF (AAF/I to AAF/III) have been described, and each is present in only a minority of EAEC isolates. No large epidemiologic study has yet succeeded in assigning an AAF variant to the majority of its clinical isolates.
Consistent with previous reports, eight of the Danish isolates did not harbor any known AAF pilin-encoding genes as determined by PCR. Among these eight strains, six were negative for the known AAF pilin genes, despite carrying the transcriptional activator of all the AAF variants, AggR. However, we found that four of these isolates were positive for the ushers of AAF/II and AAF/III, suggesting that these strains harbor adhesins related to these previously described variants. Of the four remaining strains, three exhibited evidence of a fourth adhesin variant. The genes encoding this new adhesin have previously been deposited in GenBank as Hda genes, but as of this writing, no experimental characterization of this organelle is available, and the GenBank accession describes the strain as diffusely adherent. Our data suggest that like the AAF, the Hda adhesin confers AA, and its expression requires the activity of the AggR activator.
It is as yet unclear why some strains possess the AAF/II or AAF/III ushers yet do not apparently harbor genes homologous to the pilins of these adhesins. Jenkins et al. (23) suggested that recombination occurred between and among AAF variants and/or that variation of the fimbrial subunits exceeds that of the usher-encoding genes. Consistent with the latter hypothesis, the AAF thus far described display a high level of conservation of their accessory genes but with much greater divergence among the pilin genes. However, Czeczulin et al. (9) reported that the conservation of accessory proteins extends to the functional level, as AAF/I accessory proteins could assemble AAF/II pilin subunits. Thus, these strains may have pilins very different from those already described, and further characterization is warranted. Interestingly, isolates C246-01 and C254-01 were found to be positive for the hdaA gene and a known AAF; thus, C246-01 could carry genes encoding AAF/II as well as Hda, and C254-01 harbors AAF/III as well as Hda-encoding genes.
Though it is not routinely recognized, the AAF pilins and (even more strikingly) their accessory genes are related to members of the Dr family of adhesins, some of which are found in uropathogenic and diarrheagenic E. coli strains. The analyses presented here show that HdaA is more similar to the afa-8 pilin, and its related M-agglutinin, than it is to products of AAF genes. A clearer picture of what we now call the Dr superfamily is thus emerging. Neighbor-joining analyses of all Dr-related pilins suggest the existence of three distinct adhesin families (Fig. 5). The pilins of AAF/I, AAF/II, and AAF/III (AggA, AafA, and Agg3A, respectively) make up a distinct phylogenetic cluster, as do the original Dr family, comprising F1845 and the AFA adhesins. HdaA and Hda-C1010-00 form a clearly distinct phylogenetic cluster along with Afa-8 and the M-agglutinin. Of note, we found that one of the Danish strains positive for hdaA belonged to serotype (O73:K5:H18), recently shown to belong to an emerging new clonal group A of uropathogenic E. coli (F. Scheutz, personal correspondence).
Neighbor-joining tree analysis of the predicted pilin protein of the Dr superfamily. Pilin-encoding gene sequences were downloaded from GenBank, translated, and truncated at the sites of proven or predicted signal peptidase cleavage. Derived amino acid sequences were aligned using ClustalW, and a neighbor-joining consensus tree was constructed using Phylip. An unrooted phylogram including weighted branch lengths was plotted using Phylodendron software. Afa variants are derived from urinary tract isolates, DaaE is from diffusely adherent E. coli, Agg and Aaf represent variants from EAEC, and M-agglutinin and AfaE-8 are derived from veterinary isolates.
Pilins of the Hda/Afa-8 cluster exhibit several potentially important differences from those of the AAF cluster. Prominently, members of the Hda/Afa-8 cluster lack the predicted N-terminal disulfide bond found in the AAF and the Dr clusters, which is shown to exist in the published structure of AfaE-III. The significance of the N-terminal disulfide bond in Dr adhesins is not known, though a disulfide bond in the Escherichia coli type 1 pilin FimH (8, 22) may be important for the structural integrity of the receptor binding site. Moreover, Nilsson et al. suggested that the cysteine bonds in bacterial adhesins could enable the bacteria to bind to target surfaces under high-shear conditions (40). We have recently shown that all AAF pilins exhibit a surprisingly high pI (>9.0) and that the presence of AAF confers positive surface charge upon the host bacterium. In contrast, the reported Hda/Afa-8 pilins feature pIs close to neutral. Interestingly, however, the predicted amino acid sequence of HdaA from strain C1010-00 yields a predicted pI of 8.17, whereas the pI of HdaA from strain DIJ1 is 7.0, in spite of a high level of identity between the predicted amino acid sequences. Thus, the AA adhesin of strain C1010-00 may be functionally different from that of strain DIJ1, perhaps accounting for the presence of the former in a typical EAEC strain and the latter in a strain described as diffuse adherent. Considering the AA phenotype of strain C1010-00, we speculate whether our allele may be more appropriately designated AAF/IV. Of note, Afa-8 was identified among the strains considered to be animal pathogens; thus, the possibility that Hda is an adhesin of animal EAEC strains should be evaluated. This hypothesis could help to explain the highly variable association of EAEC with human disease in various epidemiologic studies.
Reports by Nataro (32) and Jenkins et al. (23) speculated that a package of coordinately regulated virulence genes is important for EAEC pathogenesis. Thus, the term typical EAEC was proposed to describe strains harboring sets of plasmid and chromosomal virulence-related genes under the control of the master regulator AggR. Importantly, these qualities can be ascribed to isolate C1010-00 and the majority of the Danish strains used in this study. It is widely reported that enterotoxigenic E. coli adhesins are highly diverse, with over 25 colonization factor antigens (or CS adhesins) described (35). This diversity presents a formidable challenge to the use of colonization factor antigens for enterotoxigenic E. coli vaccine development. Our data, however, suggest that EAEC adhesins may be considerably less diverse, presenting the plausible prospect of incorporating all four AAF in an EAEC vaccine, presumably as part of a more complete polyvalent antidiarrheal product. We believe that the data presented here could represent a significant advance in vaccine development against this emerging pathogen.
ACKNOWLEDGMENTS
We thank Fernando Ruiz, Jalaluddin Sheikh, Edward Dudley, Mauricio Farfan, and Ian Henderson for their contributions to this project.
This project was funded by U.S. PHS grant AI33096 to J.P.N. N.B. was supported in part by SSI, Copenhagen, Denmark.
FOOTNOTES
- Received 11 December 2007.
- Returned for modification 7 January 2008.
- Accepted 17 April 2008.
- Copyright © 2008 American Society for Microbiology
