ABSTRACT
Certain Escherichia coli strains bind the Fc fragment of immunoglobulin G (IgG) at the bacterial cell surface. Previous work established that this nonimmune Ig binding depends on several large proteins with apparent molecular masses that can exceed 200 kDa. ForE. coli strain ECOR-9, four distinct genes (designatedeibA, eibC, eibD, andeibE) are responsible for Ig binding. Two eibgenes are linked to eaa genes, which are homologous to genes for the autotransporter family of secreted proteins. With reference to the E. coli K-12 chromosome, theeibA-eaaA cluster is adjacent to trpA (min 28.3) while the eibC-eaaC cluster is adjacent toaspS (min 42.0). Sequence adjacent to theeibA-eaaA cluster converges with that of strain K-12 precisely as observed for the Atlas family of prophages, suggesting that eibA is part of one of these. All four eibgenes, when cloned into plasmid vectors, impart IgG binding to E. coli K-12 strains, and three impart IgA binding also. The IgG binding occurs at the bacterial cell surface, and its expression increases survival in serum by up to 3 orders of magnitude. Theeib sequences predict a C-terminal peptide motif that is characteristic of outer membrane proteins, and the protein sequences show significant similarity near the C terminus to both the YadA virulence factor of Yersinia species and the universal surface protein A II of Moraxella catarrhalis. The sizes predicted for Eib proteins from DNA sequence are much smaller than their apparent sizes on sodium dodecyl sulfate-polyacrylamide gel electrophoresis, possibly reflecting stable oligomerization.
Proteins which bind immunoglobulins (Ig) in a nonimmune manner have been identified in numerous bacteria, and some are thought to play a role in virulence (summarized in reference 32). We previously found that sixEscherichia coli strains from the ECOR (E. colireference) collection exhibit Ig-binding activity (32). The responsible proteins are located on the surface of the cells, where they can be destroyed by limited proteolysis. The nonimmune nature is emphasized by the fact that the binding requires only the Fc fragment of IgG. The binding activity takes the form of several large proteins, some with apparent molecular masses exceeding 200 kDa by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE). In immunoblots, these proteins are seen as multiple bands, with no two strains having the same banding pattern. Expression of these proteins in ECOR-9, theE. coli strain selected for this study, is favored by growth at 37°C and by entry into stationary phase. Interestingly, the material will bind Ig both as native protein on the cell surface and after being subjected to the denaturing conditions that accompany SDS-PAGE. The goal of the present work was to determine the genetic basis of the Ig binding and the complexity of the observed banding patterns. We report the characterization of a family of eibgenes (for “E. coli Ig binding”) from ECOR-9, a strain of E. coli isolated from the feces of a healthy Swedish schoolchild (24).
MATERIALS AND METHODS
Strains and culture conditions.The ECOR reference collection of E. coli (24) and representative strains of the DEC collection (38) were obtained from Robert Selander and Thomas Whittam, respectively. Enteropathogenic E. coli strain E2348-69 was obtained from Michael S. Donnenberg. TheE. coli K-12 strains DH5α and AB1157 were used for cloning and expression studies. E. coli C was used for propagation of phages derived from ECOR-9. For expression of Ig-binding activity, 24-h Luria-Bertani (LB) broth cultures grown at 37°C with agitation were used unless noted otherwise (32). Ampicillin (50 μg per ml) was added for maintenance of pUC21-based plasmids, and kanamycin (50 μg per ml) was added for pOK12 derivatives. Phage induction and propagation were done in LC broth (LB broth containing 2.5 mM CaCl2). Phages were plated in a soft-agar overlay consisting of 0.7% agar prepared in TC medium (10 g of tryptone per liter and 5 g of NaCl per liter, 2.5 mM CaCl2) and poured over TC medium containing 1% agar.
DNA cloning and analysis.The techniques used for DNA isolation, cloning, Southern analysis, and sequence analysis involved minor modifications of those described elsewhere (16, 40). The plasmid vectors for cloning were pOK12 and pUC21 (37). Cloning of the eibA gene utilized a partial Sau3A digest of ECOR-9 genomic DNA. Cloning of eibD utilized aBamHI digest and a BssHII digest of ECOR-9. Fragments in the desired size range were purified by agarose gel electrophoresis, ligated into the BamHI or MluI site of pOK12, and electroporated into E. coli DH5α. Colony blots of transformants were screened for Ig binding by procedures based on published protocols (12). Briefly, colonies were blotted to nitrocellulose and lysed in situ with 1% SDS at 65°C for 30 min. The membranes were blocked with 10% (wt/vol) nonfat dry milk in phosphate-buffered saline (PBS) for 1 h, washed, and incubated for 1 h with affinity-purified donkey anti-rabbit Ig conjugated with horseradish peroxidase (25 or 50 ng per ml) (Amersham). The blots were washed and used to expose film. Procedures similar to those described above were used to clone othereib genes after their infectious transfer to E. coli C (see below). Key plasmids are listed in Table1. DNA probes for Southern analysis were generated by PCR and are listed in Table2; for their locations, see Fig. 2. The ECL random-prime labeling and detection systems (Amersham) were used to label probes and detect hybrids. After exposure, the blots were stripped of probe by treatment with 0.1 M NaOH at 45°C before being reprobed with a second probe. Southern blots were scanned with a UMAX Astra 1200S scanner equipped with Vista Scan 2.4.0 software (Fremont, Calif.). Oligonucleotide synthesis and automated DNA sequencing were done by the Macromolecular Core Facility of the Pennsylvania State College of Medicine. For all new sequences, both strands were sequenced; for repetitive sequencing of essentially identical segments, sequencing was sometimes limited to a single strand. Amino acid sequence similarity searches were done with BlastP (3) without filters.
Plasmids containing segments of P-Eib prophages
PCR primers and templates for probe preparation and genomic sequence analysis
Infectious transfer of eib genes.An overnight LC broth culture of ECOR-9 was diluted into fresh broth and shaken at 37°C to mid-log phase. A sample was diluted and plated on TC agar to determine the cell concentration (CFU). The cells were harvested, suspended in 0.9% NaCl, and irradiated for 45 s with a UV light (15-W GE germicidal lamp) at a distance of 35 cm. The cells were resuspended in the initial culture volume of fresh LC broth and shaken at 37°C for 2 h. Several drops of chloroform were added to the culture, and shaking was continued for 5 min. Cells and cell debris were removed by centrifugation, and the supernatant was filtered (0.45-μm-pore-diameter filter). After dilution, the filtrate was plated on a lawn of E. coli C, and very small plaques were observed. Plugs containing individual plaques along with small portions of the bacterial lawn were inoculated into separate tubes of broth. The tubes were shaken to mid-log phase and treated with chloroform as above. A drop of undiluted supernatant was placed on another lawn ofE. coli C. The plate was incubated overnight at 37°C and then at room temperature until a turbid area of bacterial growth was evident within an area of lysis. Single colonies were isolated by streaking material from these turbid regions. DNA samples from candidates were tested for homology to the eibA gene by Southern hybridization, and whole-cell extracts were tested for Ig binding by immunoblotting (32). E. coli C derivatives bearing eibE (CH6249) and eibD(CH6383) were established in this way. Strains bearing eibB(CH6304) and eibC (CH6306) were derivatives of E. coli C isolated by a modified strategy as outlined in Results.
Protein extraction and Ig binding.Preparation of cell extracts, determination of protein concentration, SDS-PAGE, immunoblotting, trypsinization of intact cells, and fluorescent-antibody binding were as described previously (32). The EibA sample was treated with 88% phenol by the method Hancock and Nikaido (10), which has been used to dissociate heat-stable protein multimers (11). Briefly, a whole-cell extract was treated with 88% phenol at 70°C, cooled to 4°C, and centrifuged. The resulting interface and phenol layer were extracted once with distilled water at 70°C, cooled to 4°C, and centrifuged again. The lower phase was mixed with 2 volumes of acetone, and the resulting precipitate was collected by centrifugation at 4°C. The precipitate was rinsed sequentially with acetone and diethyl ether and then dissolved in SDS-PAGE sample buffer. A purified Fc fragment of human IgG conjugated with horseradish peroxidase (IgG Fc-HRP) (Rockland) was used at either 5 or 20 ng of antibody per ml; purified whole human IgA conjugated with HRP (IgA-HRP) (Jackson Immunoresearch Laboratories) was used at 100 ng per ml; a purified Fc fragment of human IgG conjugated with fluorescein isothiocyanate (IgG Fc-FITC) was used at 200 μg per ml.
Serum sensitivity assay.Cultures grown in LB broth containing kanamycin were washed in PBS and thoroughly resuspended in PBS by vortexing. Control cells harboring vector pOK12 alone or test cells harboring a cloned eib gene were seeded in 150 μl of 25% unheated human serum complement (Sigma) or PBS at 2 × 108 CFU per ml. Five replicate samples of cells were incubated with serum, and three were incubated with PBS. In control experiments, heating the serum at 56°C for 30 min completely eliminated subsequent killing of control cells harboring the cloning vector pOK12. In fact, heated serum supported a slight increase in cell number. The cells were shaken at 37°C for 1 h. Serial dilutions were prepared from each replicate and plated on LB agar containing kanamycin. The experiments were repeated several times.
RESULTS
Genetic basis of nonimmune Ig binding.As a first step toward understanding the genetic basis of Ig-binding, we cloned from E. coli strain ECOR-9 a segment of DNA which imparted the binding activity to E. coli DH5α. This clone was detected in a direct screen of colonies for Ig binding (see Materials and Methods). The recombinant plasmid pCS6102 contained a 22.5-kb fragment derived from ECOR-9 DNA (Table 1). Portions of the pCS6102 insert were isolated through a series of subclonings, and a 1.4-kbXmnI-MscI segment (pCS6379) was found to be sufficient to produce Ig-binding activity. This subclone contained a 392-codon open reading frame (ORF) designated eibA. TheeibA sequence predicted a protein with a molecular mass of 42 kDa, much smaller than was expected from observations of ECOR-9. The Ig-binding activity of cells harboring pCS6379 appeared as two bands of 121 and 131 kDa in SDS-PAGE (8.5% polyacrylamide) (Fig.1A, lane 2). Possible reasons for the large difference between predicted and observed sizes are considered later.
IgG Fc-binding activity in extracts of cells harboring cloned eib genes. Bacteria were grown, and whole-cell extracts were fractionated by SDS-PAGE, blotted to polyvinylidene difluoride, and probed with IgG Fc-HRP, as indicated in Materials and Methods. Each lane contains 10 μg of protein. (A) Gel containing 8.5% acrylamide; (B and C) gels containing 10% acrylamide. (A) Lanes: 1, negative control (pOK12 vector); 2, eibA (pCS6379); 3, ECOR-9; 4, eibC (pCS6431); 5, eibD (pCS6364); 6, ECOR-9; 7, eibE (pCS6432). (B) Lanes: 1 and 3,eibA (pCS6165); 2, 66-nucleotide deletion within theeibA ORF (pDC6246). (C) Lanes: 1, eibA (pCS6379) control; 2, phenol extracted (Materials and Methods). Plasmids were maintained in the DH5α background.
Genes linked to eibA.A 11.3-kb portion of the primaryeibA clone was sequenced, revealing seven ORFs in addition to eibA (Fig. 2). All eight ORFs had the same orientation. Three ORFs had significant amino acid similarity to proteins encoded by genes of bacteriophage lambda:J, lom, and 401. Lambda genesJ and 401 encode phage tail proteins, andlom encodes an outer membrane protein (14, 29). Three of the remaining ORFs were named according to the number of amino acids encoded, ORF-191, ORF-156, and ORF-60. The fourth, designatedeaaA, was striking because of the large size of its product (1,335 amino acids) and the homology of the product to numerous virulence factors of the autotransporter family of secreted proteins (see Discussion).
Alignment of eib-linked genes. The open bars indicate the portions of each cluster that was sequenced. Coordinates are in kilobases; 0 designates the start codon of each eibgene. Positions and orientations of ORFs are indicated by arrows. GenesJ, lom, and 401 are named according to their homology to lambda genes. Only the distal portion of the Jgene was determined. Positions of DNA probes (Table 2) are shown by solid bars. Restriction sites selected for illustration are as follows: B, BssHII; Ba, BamHI; D, DraI; E,Eco47III; Ev, EcoRV; G, BglI; H,HpaI; Hd, HindIII; Hc, HincII; K,KpnI; Ks, KasI; M, MscI; N,NcoI; R, EcoRI; X, XmnI. In EibA an Hd site occurs between E and H (not shown).
At the far right, 735 bp beyond the eaaA stop codon, the sequence became identical to the E. coli K-12 sequence. This joint corresponded to min 28.3, between trp andtonB, and it interrupted the K-12 gene yciD 116 bp before its C terminus. This was exactly the same insertion point at which a diverse group of prophages, collectively termed Atlas, had been identified in 23 ECOR strains, including ECOR-9 (21).
Multiple eib loci in ECOR-9.We used a DNA probe (EibA in Fig. 2 and Table 2) and Southern analysis (see Materials and Methods) to test whether a sequence similar to eibA was common to E. coli. We observed that eibA homology was absent from strains K-12 and C, which lack Ig binding. An unanticipated insight into eib genetics was obtained when the probe was applied to a BssHII-digest of ECOR-9 DNA itself: at least five fragments with eibA homology were observed (Fig. 3A, lane 1), only one of which, a 9.8-kb fragment, was predicted from our knowledge of the original eibA clone. (In lane 1, the 9.8-kb eibAfragment overlaps a 9.7-kb fragment later shown to carry a differenteib homolog, eibC.) Results described below showed that there were at least four orthologous eib genes in ECOR-9, each capable of producing Ig-binding activity. Stripping and rehybridizing the blots with a probe derived from the eaaAsequence (Fig. 2) gave one broad band, which, from other results, represented two distinct fragments. For example, an alternate digestion using BamHI rather than BssHII clearly resolved the two bands exhibiting eaaA homology (Fig. 3B, lane 2).
eibA and eaaA homologies in the ECOR-9 genome. Restriction enzyme digests of ECOR-9 DNA were fractionated by agarose gel electrophoresis and transferred for Southern analysis as indicated in Materials and Methods. Each panel shows a single lane from the blot sequentially probed, stripped, and reprobed: probes were EibA (lane 1) and EaaA (lane 2). (A)BssHII digest; (B) BamHI digest. The weakly hybridizing band labeled e (lane 1A) hybridized strongly with a probe derived from the eibE ORF (not shown).
Transfer of three eib homologs from ECOR-9 to E. coli C.The association of eibA with the attachment position of an Atlas prophage as well as its linkage to genes homologous to lambda genes (Fig. 2) raised the possibility that the eib genes could be transferred to other strains by phages resident in ECOR-9. A supernatant of UV-induced ECOR-9 cells was applied to E. coli C, and well-separated plaques were obtained. The phage titer observed was 0.1 PFU per UV-treated CFU. Application of the supernatant to strain K-12 failed to produce any plaques. Potential lysogens of E. coli C were isolated (as detailed in Materials and Methods) and screened for acquisition of sequence homologous to the EibA probe by Southern hybridization (Fig.2). Two positive strain C derivatives, CH6249 and CH6252, were isolated. Their eib homologies were named eibEand eibD, respectively, and the phages recovered from the two strains were named P-EibE and P-EibD. P-EibE made plaques on CH6252 but not on its source CH6249; conversely, P-EibD made plaques on CH6249 but not on its source, CH6252.
At this point, while three eib homologs had been identified either by cloning or by infectious transfer, at least one more was suspected to exist in ECOR-9. Candidates were isolated as follows. By infecting CH6249 (P-EibE) with P-EibD, strain CH6253 was obtained. Neither P-EibD nor P-EibE was able to form plaques on this strain, and Southern hybridization confirmed that it had two positive Eib bands, one like that present in CH6249 and one like that present in CH6252 (Fig. 4, lanes 2 to 4). A supernatant from UV-treated ECOR-9 was applied to CH6253, and plaques were seen, although at a frequency approximately 103-fold lower than that seen when the same extract was tested on E. coli C. Strains CH6254 and CH6256 were isolated, which had three eibhomology bands (lanes 5 and 6). Lysates of these were used to produce two additional strain C derivatives, CH6304 and CH6306, each now containing a single eib homology. The respectiveeib homologies were named eibB andeibC, and the associated phages were named P-EibB and P-EibC. After UV treatment, supernatants from CH6304 and CH6306 both made plaques on E. coli C, CH6249, CH6252, and on CH6253. However, neither P-EibB nor P-EibC made plaques on either CH6304 or CH6306. P-EibD and P-EibE did make plaques on both CH6304 and CH6306.
eib homologies in E. coli C derivatives. DNA preparations were digested with BamHI, and the blot was probed with the EibA probe. Other techniques were as indicated for Fig. 3. Lanes: 1, E. coli C; 2, CH6249; 3, CH6252; 4, CH6253; 5, CH6254; 6, CH6256.
Properties of the P-Eib phage.The above discussion describes four E. coli C derivatives harboring phages recovered from ECOR-9. The derivative strains were each judged to be lysogenic because each could be induced to produce PFU detectable when plated withE. coli C but not when plated with the derivative itself. This ability was retained through multiple single-colony isolations.
Two of the putative lysogens, CH6304 (P-EibB) and CH6306 (P-EibC), gave PFU titers similar to that of a lambda lysogen control (about 10 PFU per CFU) after UV treatment. In contrast, CH6249 (P-EibE) and CH6252 (P-EibD) gave titers of only 10−2 and 10−5, respectively. These titers were several orders of magnitude greater than those found if UV treatment was omitted, except for CH6249, for which little difference in titer was observed whether or not UV treatment was included.
Cloning and sequencing eib homologs.DNA homologous to the EibA-probe was cloned (see Materials and Methods) from each of the four putative P-Eib lysogens, CH6249, CH6252, CH6304, and CH6306 (Table 1). A selected portion of each cloned segment was sequenced. In each case, an ORF with distinct homology toeibA was observed. The ORFs so identified were designated as follows: eibB from CH6304, eibC from CH6306,eibD from CH6252, and eibE from CH6249. In addition, inserts containing eibD were cloned directly from ECOR-9 and served as the initial source of sequence data. No differences were found in the sequence of eibD from the two sources.
Southern analysis (based on digests using six separate restriction enzymes) indicated that ECOR-9 possessed only two eaahomologies, one of which was linked to eibA. The linkage of the second eaa homology was unknown. In contrast to CH6249 and CH6252, CH6304 and CH6306 both had acquired eaa homology in addition to eib. The following evidence showed thateibB, as carried by CH6304, was the product of recombination. Sequence comparison of the eib loci showed that the 5′ ends of eibA and eibB were identical but different from that of eibC, while the 3′ ends ofeibB and eibC were identical but different from that of eibA. Since eibA was cloned directly from ECOR-9 and was not recombinant by definition, we concluded thateibB was recombinant and that crossover had occurred within the Eib ORF itself close to the 3′ end. Once this was established, studies of eibB were deemphasized and efforts were focused on eibA, eibC, eibD, andeibE (Fig. 2).
We attempted to determine the chromosomal location of eibCby searching for sequence near it that was identical to the E. coli K-12 sequence. (Selection of the eibB clone for the initial phases of this effort preceded our knowledge thateibB was recombinant within the eib gene). Sequence identical to the K-12 genome between bisZ (which encodes biotin sulfoxide reductase) and aspS (which encodes aspartyl-tRNA synthetase) at min 42.0 (6, 31) was found at the end of a 9.7-kb BssHII fragment that also containedeaaC (Fig. 2). Insertion of the eibC-eaaC linkage had interrupted the yecE ORF of K-12. From the sequences ofeaaC and yecD, PCR primers were designed to produce a 1,250-bp PCR product covering the position of sequence divergence (Table 2). The same PCR product was obtained with ECOR-9 DNA and with CH6306 DNA as templates. Sequence comparison of the 994 bp preceding the start of K-12 identity showed that CH6306 and ECOR-9 were identical. This result established that the eibC-eaaClinkage was located at min 42.0 in ECOR-9 as well as in the E. coli C derivative bearing eibC. The eaaA andeaaC genes were separated from their respective boundaries with chromosomal DNA by spacers of 734 and 877 bp (Fig. 2). The spacers were dissimilar, except for a 300-bp internal segment with 87% identity.
The organization of sequences linked to eibC,eibD, and eibE is shown in Fig. 2. All three linkages contained homologs of genes already found linked toeibA: ORF-191, ORF-156, and ORF-60. However, the three new versions of ORF-191 actually had 193 codons and the version of ORF-60 linked to eibE was disrupted. The ORF-191 product was homologous to a Yersinia pestis protein encoded by the virulence plasmid pMT1 (38% amino acid identities). Just beyond the ORF-60 homolog linked to eibD lay the beginning of an ORF that was similar to L0124 of the Shiga toxin 2 phage 933W (27) (445 of 449 [99%] nucleotide matches).
The degree of sequence conservation observed for these foureib linkage clusters was complex. For example, at the DNA level, the ORF-191 versions linked to eibC, eibD, and eibE were 99 to 100% identical. However, alignment of these three with the corresponding gene linked to eibArevealed an uneven relationship. The first 80 amino acids were 25% identical, an internal segment of 93 amino acids was 99% identical, and the last 20 amino acids were 50% identical. As discussed in more detail below, eibA was also the phylogenetic outlier with regard to the four eib genes. Alignments of the ORF-156 to ORF-60 regions give a different picture in that the region linked toeibE rather than the region linked to eibA was the outlier. The region adjacent to eibE maintained only 65% identities at the DNA level compared to any one of the other three. Alignment of ORF-156 through ORF-60 copies adjacent toeibA, eibC, and eibD, by contrast, revealed 92.8 to 96.6% identities in pairwise comparisons.
The eib gene products.The IgG Fc-binding activity of ECOR-9 distributed as multiple bands, all of which had apparent molecular masses of ≥121 kDa in SDS-PAGE (8.5% polyacrylamide) (Fig.1A, lanes 3 and 6). Two of the forms migrated in the range of 121 to 131 kDa, while three others appeared to be >180 kDa in SDS-PAGE. Curiously, none of the eib sequences predicted a protein nearly this large. The predictions from sequence were as follows:eibA, 42.0 kDa; eibC, 53.2 kDa; eibD, 53.9 kDa; eibE, 51.6 kDa. An effort was made to correlate the cloned eib genes with individual IgG Fc-binding bands on an immunoblot and with Coomassie-stained bands in a gel. Immunoblotting revealed predominant IgG Fc-binding bands expressed by each clone which were unique in apparent size and corresponded to a form present in ECOR-9. When replicate gels were stained with Coomassie instead of being blotted, we observed bands which comigrated with bands observed in ECOR-9 but absent from DH5α (data not shown). The strain with cloned eibA expressed two closely migrating bands with apparent molecular masses of 121 and 131 kDa (lane 2). Strains harboring eibC, eibD, and eibE each expressed Ig-binding bands exceeding 180 kDa, with EibD (lane 5) being the largest, EibC (lane 4) being intermediate, and EibE (lane 7) being the smallest of this group.
The discrepancies between predicted and observed sizes could be explained if the cloned eib genes were not the actual structural genes for the respective Ig-binding activities but only controlled their synthesis. That eibA is indeed a structural gene for the binding protein was demonstrated by introducing a 66-bp deletion (between an Eco47III site and aHindIII site) near the center of the eibA ORF (Fig. 2). This deletion caused a significant reduction in the apparent size of the Ig-binding activity (Fig. 1B, lane 2). Consequently, we concluded that the Eib proteins assume a physical form that causes them to migrate in an unpredicted fashion.
The final nine amino acids at the C terminus of each Eib protein matched a motif associated with outer membrane proteins (see Discussion). A set of deletion mutations was prepared that eliminated the last 8 (eibA, eibC, and eibD) or last 5 (eibE) codons, replacing them with, respectively, 12 and 5 codons from vector sequence. In all four cases, these deletions abolished Ig-binding activity as well as the appearance of large proteins on the Coomassie-stained gels (data not shown).
Several treatments were used in attempts to produce forms of the Eib activity as small as those predicted from the sequences of theeib ORFs. For example, migration of the Ig-binding activities in SDS-PAGE was unaffected by boiling in sample buffer containing 6% SDS and 2-mercaptoethanol. For EibA, boiling of the extract in 8 M urea was also tried without effect. However, extraction of the EibA sample with 88% phenol at 70°C did have at least a partial effect (Fig. 1C). Whereas the bulk of the EibA-binding activity continued to migrate at 121 and 131 kDa, a portion appeared at a position slightly faster than the 42.7-kDa standard.
IgA binding.The preceding results showed that each of the three eib genes produced material that bound IgG. It was of interest to determine if binding to IgA occurred as well. The immunoblot shown in Fig. 5 was identical to that shown in Fig. 1A, except that the gel contained 10% acrylamide. The blot was first probed with human serum IgA-HRP (Fig.5A), washed but not stripped, and then probed with IgG Fc-HRP (Fig.5B). Testing with IgA revealed that eibD produced strong binding to IgA (Fig. 5A, lane 5) and that eibC produced clearly significant binding (lane 4). However, the binding produced byeibE to IgA was barely perceptible even with overexposure of lane 7 (data not shown), and eibA produced no observable binding to IgA at all (lane 2). Thus, the binding of IgA relative to IgG differed substantially for the four genes, being greatest foreibD and least for eibA. This seemed to correlate with the difference in Ig binding patterns seen with an ECOR-9 extract (compare lanes 3 or 6 in Fig. 5A and B).
IgA-binding activity in extracts of cells harboring cloned eib genes. Extracts were prepared and fractionated as described in the legend to Fig. 1A, except that SDS-PAGE (10% polyacrylamide) was used. The immunoblot was probed with human serum IgA-HRP (A), washed but not stripped, reblocked, and probed with IgG Fc-HRP (B). Lane assignments are the same as in Fig. 1A.
IgG Fc binding by intact cells.Previous studies demonstrated that fluorescent antibodies bind to intact cells of ECOR-9 in a nonimmune fashion (32). That work also revealed a heterogeneity of antibody binding within the population; i.e., a small percentage of cells fluoresced strongly, while most of them resembled the negative controls. To determine whether cells harboring the clonedeib genes had a similar phenotype, each gene was introduced into the K-12 strain AB1157. Unfixed cells were incubated with IgG Fc-FITC and prepared for microscopy (see Materials and Methods). The same fields were observed by both fluorescence microscopy (Fig.6, top row) and bright-field microscopy (bottom row). As previously reported (32), only a minority of ECOR-9 cells fluoresced brightly, and most cells showed little or no fluorescence (Fig. 6A). AB1157 cells harboring only the pOK12 cloning vector showed no fluorescence (Fig. 6B). Cells harboringeibA (Fig. 6C) and eibE (Fig. 6F) showed bright fluorescence concentrated at the cell surface. Cells harboringeibC (Fig. 6D) or eibD (Fig. 6E) exhibited a less bright and more even distribution of fluorescence. Each of theeib genes was capable of conferring Ig-binding activity on intact cells, but without the extreme cell-to-cell heterogeneity observed for ECOR-9. These experiments were repeated using E. coli C as the host for the recombinant plasmids with similar results (data not shown). Attempts to repeat the experiments inE. coli DH5α revealed only a very pale fluorescence for all clones (data not shown). The fluorescence observed was nevertheless greater than that of the vector control. This weak fluorescence was not expected, since immunoblotting of whole-cell extracts (Fig. 1) had indicated that significant amounts of Ig-binding proteins were being expressed, and limited proteolysis of intact cells (data not shown) had demonstrated that binding occurred at the cell surface of DH5α hosting the constructs.
Binding of human IgG Fc-FITC to intact, unfixed cells. Bacterial cells were incubated with IgG Fc-FITC as indicated in Materials and Methods. Fluorescing cells (top row) and total cells in the same field (bottom row) are shown. (A) ECOR-9; (B to F) AB1157 containing the following plasmids: pOK12 (B), eibA in pCS6379 (C), eibC in pCS6431 (D), eibD in pCS6364 (E), and eibE in pCS6432 (F). Plasmids were maintained in the AB1157 background. Microscopic images were captured with a Nikon Microphot-FX and processed with QED (Pittsburgh, Pa.) Camera software. Twelve individual images were assembled into a composite with Photoshop 5.0. Bar: 10 μM = 10 μm.
eib genes confer serum resistance.Some cell surface proteins enhance the survival of bacterial cells in serum. Whether Eib proteins have such an effect was evaluated using AB1157, a strain which is known to be particularly sensitive to serum (4). Cells were exposed to 25% human serum or control PBS for 1 h and then diluted and plated. The results of two representative experiments are shown in Fig.7. In experiment A, only 0.003% of control cells harboring vector survived serum treatment while 0.44% of cells harboring eibA and 11% of cells harboringeibD survived (Fig. 7A). In experiment B, 0.03% of cells containing vector survived serum treatment while 34% of cells harboring eibC and 6% of cells harboring eibEsurvived (Fig. 7B). A survival ratio was calculated for each by dividing its survival frequency by the survival frequency observed for the control. The survival ratios conferred by the eib genes in these experiments were 147 (eibA), 1,133 (eibC), 3,667 (eibD), and 200 (eibE).
Serum resistance imparted by eib genes. AB1157 cells containing various plasmids were treated with 25% human serum (open bars) or PBS (shaded bars) for 1 h. The cells were diluted and plated as indicated in Materials and Methods. The plasmids used were as follows: vector, pOK12; eibA, pCS6379;eibC, pCS6431; eibD, pCS6364; andeibE, pCS6432. Survivors are shown as CFU per milliliter. Each bar represents the mean of three (PBS) or five (serum) replicate treatments. Error bars indicate standard deviation. Panels A and B represent separate experiments.
Factors affecting expression.The work on eibexpression described above was done on genes inserted into a multicopy plasmid vector, and quite strong expression was achieved in bothE. coli K-12 and C backgrounds. In contrast, expression of Eib proteins in P-Eib lysogens of E. coli C was very weak relative to that in ECOR-9 (data not shown). This difference between expression in ECOR-9 and the E. coli C lysogens was not anticipated since both cases appear to involve expression from prophage.
Expression of Ig-binding proteins in ECOR-9 is maximal in cells grown at 37°C and undetectable in those grown at 27°C (32). It is also highly favored by entry into stationary phase. Expression of Ig-binding activity upon entry into stationary phase was found to increase significantly in strains harboring each of the four cloned genes (data not shown). To see if the thermal control was maintained for the cloned eib genes, their expression in AB1157 was tested (Fig. 8). Comparison of Fig. 8lanes 1 and 3 (eibA), 5 and 7 (eibD), and, to a lesser degree, 9 and 11 (eibE) shows that while Ig binding was greater in cells grown at 37 than at 27°C, it was still detectable when the cultures were grown at 27°C. The three constructs used in the preceding experiments retained 538 bp (eibA), 342 bp (eibD) and 337 bp (eibE) of leader sequence upstream from the respective eib ORFs. We also prepared variants that retained only 105 to 110 bp of leader sequence. Comparison of lanes 2 and 4 (eibA), lanes 6 an 8 (eibD), and lanes 10 and 12 (eibE) shows that these constructs also had greater expression at 37°C. However, in each comparison, regardless of the growth temperature, the deletion of a portion of the leader sequence enhanced the overall expression.
Effect of culture growth temperature and amount of leader DNA on expression of Ig-binding activity. AB1157 cells containing cloned eib genes were grown at 27°C (lanes 1, 2, 5, 6, 9, and 10) or 37°C (lanes 3, 4, 7, 8, 11, and 12). Whole-cell extracts were fractionated by SDS-PAGE (10% polyacrylamide), blotted to polyvinylidene difluoride, and probed with IgG FC-HRP as indicated in Materials and Methods. Plasmids present: pCS6374 (eibA with longer leader [537 bp]) (lanes 1 and 3); pCS6379 (eibA with shorter leader [105 bp]) (lanes 2 and 4); pCS6417 (eibD with longer leader [342 bp]) (lanes 5 and 7); pCS6364 (eibD with shorter leader [110 bp]) (lanes 6 and 8); pCS6418 (eibE with longer leader [336 bp]) (lanes 9 and 11); and pCS6412 (eibE with shorter leader [104 bp]) (lanes 10 and 12).
Ig binding in other strains.In our previous survey of the ECOR collection, five strains besides ECOR-9 scored positive for Ig binding while the remaining 66 were negative (32). The binding activity of all six positive strains (ECOR-2, ECOR-5, ECOR-9, ECOR-12, ECOR-43, and ECOR-72) appeared as multiple large bands, although the pattern of bands was different for each strain. Southern blots of the entire ECOR collection were tested for homology to the EibA probe. We found an absolute correlation between the Ig-binding activity and eibA homology: the six positive strains all exhibited two to five homologies (tested after BamHI digestion), while none of the other 66 strains had any detectable homology. Except for ECOR-72, the positive strains also showed homology to the EaaA probe. These results suggested that eiborthologs are responsible for the Ig binding of the six positive ECOR strains.
We have also screened 19 representative strains from the DEC (diarrheagenic E. coli) collection of pathogenic E. coli (38). These included at least one representative of each electrophoretic type. All were negative for both IgG Fc binding and EibA probe homology. EPEC strain E2348-69 was also tested for IgG Fc binding and found to be negative.
DISCUSSION
A set of at least four eib genes is responsible for the Ig-binding activity of ECOR-9. The four predicted peptide sequences are aligned in Fig. 9. They show clear similarity at the N terminus (Fig. 9, block 1) which is predicted to be part of a signal peptide for export across the cytoplasmic membrane (23). The peptides are also similar at the C terminus, where 124 of the last 167 residues (74%) are identical in all four (blocks 2 to 5). The last nine amino acids conform to a pattern of alternating hydrophobic residues terminating in phenylalanine (block 5). This motif has been implicated in targeting proteins to the outer membrane (35), and its presence is consistent with cell surface exposure of Eib. In pairwise comparisons, EibC and EibD are the most similar at 87% overall amino acid identity. EibA, at 392 amino acids, is significantly shorter than the other three, which range from 487 to 511 residues. EibA diverges from the others in the region between blocks 1 and 2, where it retains fewer than 10% identities relative to the others.
Alignment of amino acid sequences predicted from the four eib ORFs. Hyphens are introduced into the sequences to aid alignment. Amino acids that are identical in all four proteins are specified for the eibC sequence and indicated by ● for the other three. The blocks specify regions discussed in the text. The successive coiled-coil heptads predicted by the MultiCoil program for blocks 3 and 6 are indicated by abcdefg, while the alternating hydrophobic residues at the C termini are marked by asterisks beneath the alignments.
A search of the protein databases for sequences similar to Eib identified two outer membrane proteins: YadA (Yersiniaadhesin) from Yersinia pseudotuberculosis and Yersinia enterocolitica (30, 33, 36) and UspAII (universal surface protein AII) from Moraxella catarrhalis(2). The only region of high mutual similarity comprises 50 amino acids near the C terminus (block 4 in Fig. 9), where sequence identities range from 64 to 74% in pairwise comparisons of the three. Neither UspAII nor YadA has been reported to have nonimmune Ig-binding activity, but, like Eib, both confer serum resistance on the bacterial cell (1, 20, 26).
The Eib proteins have apparent molecular masses much greater than those predicted from the DNA sequences. EibA, for example is predicted to be 43 kDa, but the product takes two forms, of 121 and 131 kDa when observed by SDS-PAGE (8.5% polyacrylamide). The largest of the four, EibD, is predicted to be 54 kDa, but the major product appears to be approximately 210 kDa. We emphasize that the size estimates of the Ig-binding material, based on comparisons to conventional protein standards, vary with the acrylamide concentration (32); the estimated size becomes relatively greater with greater acrylamide concentrations. This dependence strongly suggests the contribution of a shape factor in addition to mass. Interestingly, both YadA and UspAII also appear as proteins with much greater molecular masses in SDS-PAGE than those predicted from their gene sequence (1, 18, 19, 33, 36). Like the Eib proteins, these proteins lack cysteine residues, and so disulfide cross-linking does not cause this effect.
One possibility for the aberrant behavior of Eib proteins in SDS-PAGE is that they are modified by the addition of large side chains or that they form a stable association with other proteins. In either case, cryptic components present in both E. coli C and K-12 would have to provide components essential for this association, since a single eib ORF cloned from ECOR-9 is sufficient to confer both Ig binding and apparent large size. While our data do not negate this possibility, we favor an explanation that the large sizes reflect stable oligomerization of the Eib proteins themselves. A 46-amino-acid coiled coil is predicted for the Eib proteins (block 3 in Fig. 9). Coiled-coil structures involve two or more α-helices, and the MultiCoil program (39) predicts an extended coiled-coil structure for the Eib sequences (probability of 0.66), favoring a trimer over a dimer (ratio, 1.65). Oligomerization through formation of coiled-coils could explain the apparent large size in SDS-PAGE, with the proviso that the interaction is sufficiently stable to survive boiling in 6% SDS. The only treatment found to convert even a portion of the EibA-associated activity to a size comparable to that predicted for the eibA ORF involved extraction with 88% phenol at 70°C (Fig. 1C). Interestingly, the Eib homolog YadA maintains an oligomeric structure, probably a trimer, under the standard preparation conditions used for SDS-PAGE (19, 33, 36), although dissociation of YadA was achieved by boiling in 10 M urea. Extreme treatment was also required to dissociate UspAII (18). We have applied the MultiCoil algorithm to the YadA and UspAII sequences and found that, like Eib, YadA and UspAII are predicted to assume coiled-coil structures. However, the segments of the YadA, UspAII, and Eib sequences predicted as coiled-coil structures are not the same as the segments showing the greatest sequence identity (i.e., block 4 of Eib [Fig. 9]). The MultiCoil algorithm predicts another coiled-coil structure for EibA (block 6) with a probability of 0.99, but similar regions are not identified for the other three Eib sequences. Block 6 of EibA has four Leu residues spaced at intervals of seven residues and thus has the form of a leucine zipper.
ECOR-9 expresses Ig binding preferentially in the stationary phase, and the expression is highly dependent on the temperature (32). In addition, expression observed at the level of individual cells is not uniform (Fig. 6A and B). Expression from eib genes on multicopy plasmids in strain K-12 was substantial and on the order of that seen with ECOR-9 (judged by examination of immunoblots and Coomassie-stained gels), yet the number of eib gene copies in ECOR-9 would be much smaller than it would be in cells with multicopy plasmids. This observation suggests that eibexpression in ECOR-9 is a complex phenomenon. For an eibAconstruct retaining as little as 105 bp of leader, expression increased with entry into stationary phase and with a shift in temperature from 27 to 37°C. Similar observations were made for a comparableeibD construct. Apparently, sites required for response to these environmental signals are within the 105-bp leader ofeibA. We note that each of the four eib ORFs is separated from the next ORF downstream by an 86- to 90-bp spacer which contains a GC-rich stem-loop sequence followed by several T nucleotides that might serve as a transcription terminator.
The first eib gene cloned was eibA, and it was obtained directly from the ECOR-9 genome. The cloned segment included typical prophage genes as well as a potential prophage-host boundary. This boundary was identical to the prophage boundary identified for the Atlas family of prophages, some of which are defective (21). Prophage location defines the Atlas family, so P-EibA is an Atlas prophage. However, the occurrence of Atlas prophages is more frequent among ECOR strains than is the occurrence of eib genes; consequently, most of the Atlas prophages do not carry eib. The presence of sequence with homology to lambda J,lom, and 401 also suggested that eibAmight be part of a prophage. Based on these observations, we hypothesized that other eib genes might be associated with prophage and could be transferred from ECOR-9 to a recipient strain in association with phage. A strategy based on this hypothesis was adopted for isolation of eib genes. The eibA locus, however, was never recovered from our UV treatment of ECOR-9, and so the putative ECOR-9 prophage may be defective. Nevertheless, the UV treatment did result in the isolation of E. coli C derivatives containing three other eib homologs. The phages that transferred them remain largely uncharacterized, but an obvious possibility is that eib occurrence in ECOR-9 is associated with multiple prophages. Caution must be used in equating the phage transferred to the E. coli C derivatives with the phage present in ECOR-9, since the possible presence of multiple prophages (some possibly defective) in ECOR-9 could result in recombinant phages among those that were recovered. We have presented evidence that theeibB ORF is recombinant (see Results). Also, we have evidence based on restriction site patterns (unpublished) that the sequence to the left of eibC (Fig. 2) is recombinant with respect to sequences resident in ECOR-9. However, the eibCORF contains many codons unique with respect to the other threeeib loci (Fig. 9), a circumstance that reduces the possibility that the ORF is recombinant.
Two of the four eib genes are linked to a large (1,335-codon) ORF designated eaa (Fig. 2). eaaAand eaaC are 99.4% identical at the nucleotide level. Of the 26 divergent nucleotides, only 8 are nonsynonymous, indicating strong selective pressure. The protein predicted for Eaa conforms to characteristics of a family of secreted autotransporter proteins identified in numerous gram-negative pathogens (reviewed in references13 and 17). Several autotransporters are virulence factors (5, 7-9, 13, 17, 22, 25, 28, 34). Specifically, residues 1056 to 1335 of EaaA exhibit 62 to 92% identity to the C-terminal β domains of various autotransporters of E. coli and Shigella. In contrast, the identities for the central α domains are only 34 to 40%. Typically, sequence identity among different autotransporters is high among β domains and low among α domains. Eaa shares numerous structural features of the autotransporter subfamily termed SPATE (for “serine protease autotransporters ofEnterobacteriaceae”) (13): large size (1,335 amino acids before putative processing); a long signal sequence (cleavage predicted after serine at residue 56) (23); a serine protease motif (GDSGS at residues 260 to 264); a candidate cleavage site (9) between the α and β domains (after Asn at residue 1058); and a C-terminal outer membrane localization motif consisting of alternating hydrophobic amino acids ending in F (VNAGFRYSF).
Perhaps one of the most curious aspects of eib genetics is the occurrence of multiple eib orthologs within singleE. coli strains. Of 72 ECOR strains, only 6 contain sequences detectable by probe EibA, yet all 6 have two or moreeib genes. Five of these also have eaa homology. The phylogenetic relationships of the ECOR collection have been delineated in considerable detail (15). Of the six positive strains, two, ECOR-72 and ECOR-43, are partitioned from the other four in separate ECOR groups. The remaining four, including, ECOR-9, belong to a 10-member clade within the ECOR phylogeny, but there is no obvious clustering of positives within the clade. We postulate that this distribution reflects selection for a bacterium to have either multipleeib genes or none at all. eib genes may benefit cells occupying a special ecological niche within the host, and/or their encoded functions may interact cooperatively.
ACKNOWLEDGMENTS
We thank Ira Ropson for helpful discussions and Du Chungen for expert technical assistance.
This work was supported by Public Health Service grant GM16329 from the National Institutes of Health.
Notes
Editor: P. E. Orndorff
FOOTNOTES
- Received 30 September 1999.
- Returned for modification 2 December 1999.
- Accepted 29 December 1999.
- Copyright © 2000 American Society for Microbiology
