INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

Enteropathogenic Escherichia coli (EPEC) is a leading cause of infantile diarrhea in developing countries (12). EPEC is recognized by two phenotypes that are apparent in vitro and in vivo, attaching and effacing and localized adherence (LA). The attaching and effacing phenotype is characterized by the destruction of microvilli and the formation of cup-like pedestals upon which the bacteria rest and is mediated by the products of genes present on a chromosomal pathogenicity island (15, 26). LA is the distinctive pattern of EPEC adherence to epithelial cells, in which the bacteria form tightly packed clusters (11, 37). LA is dependent upon the synthesis of a type IV fimbria, the bundle-forming pilus (BFP), the expression of which is dependent on genes found on a large plasmid (13, 19).

Type IV pili are expressed by a wide variety of pathogenic organisms. These fimbriae mediate adherence to epithelial cells and autoaggregation of bacteria, act as receptors for bacteriophages (8, 44), and mediate a type of surface translocation known as twitching motility (9). Although many organisms express type IV pili on their surfaces, very little is known about the biogenesis of these organelles. Type IV pili are usually composed of a single repeating pilin protein. With the exceptions of Neisseria gonorrhoeae, which has an adhesin protein located at the tip of the pilus (35), and E. coli strains that express the R64 thin pilus, which has a minor pilus component (48), type IV pili appear to be comprised solely of a single pilin protein. Pilin has a short leader sequence that is cleaved by a specific prepilin peptidase (43). Type IV pilus biogenesis systems also contain an outer membrane protein belonging to a family of proteins called secretins. These proteins assemble into rings composed of 12 to 14 monomers and are believed to form channels in the outer membrane through which pilin subunits pass (7). The functions of the other proteins involved in type IV pilus biogenesis are not known. The most extensively studied type IV pilus system is that of Pseudomonas aeruginosa. This system requires the protein products of more than 30 genes for formation of functional pili (4). Many of the genes that have been implicated in type IV pilus biogenesis in one system have homologues in other type IV systems, as well as in systems involved in protein secretion, DNA uptake, and filamentous phage assembly (21, 42). Some of these genes encode putative nucleotide-binding proteins and inner or outer membrane proteins. The precise functions of these genes in pilus biogenesis are not known.

LA by EPEC is dependent upon the presence of a 95-kb plasmid, termed the EPEC adherence factor (EAF) plasmid. A cluster of 14 genes from the EAF plasmid, when transformed along with an additional fragment of the EAF plasmid containing regulatory genes, is sufficient to confer both BFP biogenesis and LA on a laboratory strain of E. coli (39, 40). Thus, the BFP system is one of the few in which all the genes required for the biogenesis of a type IV pilus have been identified. The first gene in the bfp cluster is bfpA, which encodes prebundlin, the prepilin protein (13, 38). The bfpP gene encodes the prepilin peptidase which is believed to cleave prebundlin to its mature form, is homologous to prepilin peptidase genes in other type IV pilus systems, and is capable of complementing a P. aeruginosa prepilin peptidase mutant (51). The bfpB gene encodes an outer membrane lipoprotein (34) that belongs to the secretin family of outer membrane proteins including PilQ of the P. aeruginosa type IV pilus system. The bfpF gene encodes a putative nucleotide-binding protein which is homologous to the PilT protein of P. aeruginosa and is not required for BFP biogenesis or LA (5, 6). Although bfpF mutant strains autoaggregate, the aggregates are morphologically distinct from those of wild-type EPEC, and while the wild-type aggregates disperse over time, the bfpF mutant aggregates do not (6). Knutton et al. have shown that, when wild-type EPEC infects tissue culture cells in vitro, BFP undergoes a dramatic change in morphology, with the individual pilus filaments forming much longer and thicker bundles (23). BFP in a bfpF mutant strain does not undergo this transformation (23). In addition, volunteer studies have shown that the virulence of a bfpF mutant strain of EPEC is greatly attenuated compared to that of wild-type EPEC (6). Thus, while BfpF does not play a role in BFP biogenesis, it does play a role in BFP function. Additional genes of the bfp cluster are predicted to encode a putative nucleotide-binding protein, BfpD, which is homologous to PilB of P. aeruginosa (28); a transmembrane protein, BfpE, which is homologous to the PilC protein of P. aeruginosa (28); and a lytic transglycosylase, BfpH. The cluster also contains several genes that encode putative prepilin peptidase substrates (bfpI, bfpJ, and bfpK) (39, 40). The remaining genes of the cluster, bfpG, bfpC, bfpU, and bfpL, show no sequence homology to known genes (39, 40).

To date, mutations in genes involved in type IV pilus biogenesis in other organisms have resulted in nonpiliated bacteria with the accumulation of pilin protein within membrane fractions (1-3, 28), with the exception of a family of nucleotide-binding proteins related to and including the bfpF gene product, in which mutations result in hyperfimbriated bacteria (6, 47). In an effort to define the genes of the bfp cluster that are involved in BFP biogenesis or LA, we intend to create nonpolar mutations in each of the bfp genes. Here we report the mutation of six bfp genes, bfpG, bfpB, bfpC, bfpD, bfpP, and bfpH, and the effects of these mutations as well as that of a previously described bfpA mutation on BFP biogenesis, autoaggregation, and LA.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Strains and plasmids. Bacterial strains and plasmids used in these experiments are listed in Table 1. Strains were grown on Luria-Bertani (LB) agar or in LB broth at 37°C except where indicated and stored at 80°C in 50% LB broth-50% (vol/vol) glycerol stock. Antibiotics, when necessary, were added at the indicated concentrations: ampicillin, 200 µg/ml; kanamycin, 50 µg/ml; chloramphenicol, 20 µg/ml; and nalidixic acid, 50 µg/ml.

DNA cloning and mutant construction. DNA restriction digestions, electrophoresis, and ligations were performed by standard procedures (36). Plasmids were introduced into strains by CaCl₂ transformation or electroporation. Electroporation was carried out in 10% glycerol in a 0.1-cm cuvette, using an E. coli pulser (Bio-Rad, Hercules, Calif.) set at 1.8 kV.

Tissue culture and adherence assay. Adherence assays were performed with HEp-2 cells (ATCC CCL 23) in eight-well chamber slides (Becton Dickinson, Franklin Lakes, N.J.) as previously described (16).

Autoaggregation assay. Overnight bacterial cultures were diluted 1:100 in 20 ml of Dulbecco's modified Eagle's medium (DMEM)-F-12 with 15 mM HEPES and 2 mM glutamine and grown for at 37°C with shaking at 250 rpm until bacterial aggregates were visible in strains expressing BFP. A 1-ml aliquot of each culture was removed, and the A₆₀₀ of each culture was measured. The samples were then vortexed for 30 s, and the A₆₀₀ was measured again. The percent increase in A₆₀₀ after vortexing was recorded as a quantitative aggregation index. Mean values were compared using Student's t test for paired samples (two tailed).

Immunoblotting. Overnight bacterial cultures were diluted 1:100 in 20 ml of DMEM and grown for 5 h at 37°C with shaking at 250 rpm. Samples were centrifuged (3,400 × g, 4°C, 5 min) and resuspended in 350 µl of water. Fifty microliters of each sample was saved for determination of the protein concentration by the bicinchoninic acid protein assay (Pierce, Rockford, Ill.) in a multititer plate reader with bovine serum albumin as a standard. One hundred microliters of 4× sodium dodecyl sulfate (SDS)-polyacrylamide gel electrophoresis (PAGE) loading buffer (0.25 M Tris-HCl [pH 6.8], 8% SDS, 40% [vol/vol] glycerol, 0.008% bromphenol blue, 20% 2-mercaptoethanol) was added to the remaining 300 µl of each sample. Samples were denatured by boiling for 10 min in the SDS-PAGE buffer, and 5 µg of total protein per lane was loaded on 15% polyacrylamide gels and separated by electrophoresis. Samples were transferred to Immobilon-P polyvinylidene difluoride membranes (Millipore, Bedford, Mass.). The blots were blocked with phosphate-buffered saline (PBS) containing 0.5% (vol/vol) Tween 20 and 5% milk, reacted with mouse antibundlin monoclonal antibody ICA4 (20) at a 1:5,000 dilution, incubated with a goat antimouse antiserum conjugated to horseradish peroxidase (Sigma, St. Louis, Mo.) at a 1:30,000 dilution, and developed by enhanced chemiluminescence (ECL) (Amersham Life Science, Arlington Heights, Ill.).

Transmission electron microscopy. For viewing BFP, overnight bacterial cultures were diluted 1:100 in 20 ml of DMEM and grown at 37°C for 5 h. Samples were centrifuged (3,400 × g, 5 min) and resuspended in 1 ml of DMEM. A 10-µl aliquot of each sample was spotted on Formvar-carbon-coated copper grids (Electron Microscopy Sciences, Fort Washington, Pa.) for 5 min. Grids were blotted, washed with PBS, stained with phosphotungstic acid in PBS, and examined in a JEOL JEM-1200 EXII transmission electron microscope. Samples were coded and examined without knowledge of their identity.

Sucrose density flotation gradient fractionation. EPEC strains were grown overnight in LB medium and diluted 1:100 into 10 50-ml conical tubes containing 20 ml of DMEM-F-12 medium with 15 mM HEPES and 2 mM glutamine. Recombinant HB101 strains were grown overnight in LB medium and diluted 1:100 into 1 liter of LB medium. Cultures were grown for 5 h, centrifuged (4,000 × g, 4°C, 15 min), and resuspended in 3 ml of 25 mM Tris-HCl, pH 7.4. Samples were then lysed in a French press at 20,000 lb/in², centrifuged twice (3,000 × g, 4°C, 5 min), and stored at 20°C. Sucrose (3 g) was added to 2 ml of lysate to create a 60% (wt/wt) solution, and 1.3 ml of this solution was placed in the bottom of an ultracentrifuge tube. A sucrose density gradient was created by overlaying the sample with 1.3 ml of sucrose solutions of decreasing concentrations in 3% decrements, from 56 to 28% (wt/wt). Samples were centrifuged (245,000 × g, 10°C, 72 h), and 540-µl fractions were removed from the tubes and placed into microcentrifuge tubes. To determine the density, the tubes were weighed before and after removing 100 µl from fractions. These 100-µl aliquots were saved, and NADH oxidase activity was assayed as previously described (30). The rate of decline in A₃₄₀ was used as an indicator of relative enzyme activity in each fraction. The remaining 440 µl was precipitated by adding an equal volume of cold 25% trichloroacetic acid, incubating on ice for 30 min, and centrifuging (6,000 × g, 4°C, 30 min). The samples were resuspended in 80 µl of 10% saturated Tris base in 4× SDS-PAGE loading buffer. Twenty-microliter aliquots of each fraction were loaded on SDS-15% PAGE gels and separated by electrophoresis. The gels were transferred as described above, and the blots were stained with Ponceau stain (Sigma) to identify fractions containing soluble and insoluble proteins and those containing the outer membrane protein OmpA (32). The blots were blocked and immunoblotted for bundlin as described above.

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

Construction of bfp mutants. To examine the roles of the Bfp proteins in BFP biogenesis and function, the bfpG, bfpB, bfpC, bfpD, bfpP, and bfpH genes were mutated in this study. The genes were disrupted by the addition of a nonpolar kanamycin cassette (27). The cassette begins with stop codons in all three reading frames followed by the aphA3 gene. Immediately following this is a consensus ribosome binding site and an ATG start codon to reinitiate translation of the downstream portion of the interrupted gene. The cassette has been modified by the addition of either one or two nucleotides immediately following the start codon, so that it is available in all three reading frames. The disrupted gene was subcloned into pCVD442 (14), a positive-selection suicide vector, and mobilized into wild-type EPEC by triparental conjugation, and allelic exchange was carried out as previously described (17). PCR analysis of each mutant confirmed the addition of the 850 bp of the kanamycin cassette in the intended gene, with no changes in nearby loci (data not shown). Each mutation that resulted in a discernible phenotype was complemented by the addition of a wild-type copy of the mutant gene in trans on a low-copy-number vector to confirm that the mutations had no polar effects on prebundlin processing, BFP expression, or the ability to autoaggregate or perform LA. Control plasmids were also introduced into each mutant to verify that complementation was due to the presence of the wild-type allele of the gene (Fig. 1).

View larger version (15K): [in this window] [in a new window]   FIG. 1.   Location of bfp mutants and complementing fragments. Lines beneath the genes show the fragments of the cluster used to construct complementing and control plasmids. For each mutant except bfpA and bfpP mutants, a plasmid containing the fragment ending within the corresponding mutated gene served as the control plasmid and a plasmid containing the fragment ending within the gene immediately downstream of the mutated gene served as the complementing plasmid. For the bfpA and bfpP mutants, the vector served as the control plasmid and the vector containing the corresponding gene served as the complementing plasmid. The upper row of restriction sites shows the beginning and endpoint of each fragment. The lower row of restriction sites indicates where the kanamycin cassette was inserted into each gene. Restriction enzyme abbreviations: B, BamHI; BB, BstBI, H, HindIII; Hp, HpaI; K, KpnI; L, BalI; M, MfeI; N, NruI; P, PstI; S, SpeI; SB, SnaBI; X, XbaI.

Effects of bfp mutations on expression and processing of prebundlin. bfpA, the first gene in the bfp cluster, encodes bundlin, the major structural subunit of BFP. Bundlin is expressed as a preprotein with a short leader sequence. To determine if any of the mutations that we made affected prebundlin expression or processing, immunoblotting for bundlin was performed on all mutant strains and on wild-type EPEC (Fig. 2). In all mutant strains tested except the bfpP mutant strain UMD932, bundlin was expressed and processed to mature bundlin in a manner similar to that of wild-type EPEC. The addition of the bfpP gene on plasmid pRPA107 restored the ability of UMD932 to process prebundlin to bundlin, while a control plasmid had no effect. This indicates that only BfpP is required for the processing of prebundlin to mature bundlin. We found no consistent effect of any bfp mutation on the amount of bundlin expression in repeated experiments.

View larger version (21K): [in this window] [in a new window]   FIG. 2.   Expression and processing of bundlin in wild-type EPEC and mutant strains. Whole-cell lysates of cells grown in DMEM were prepared, separated by SDS-PAGE electrophoresis on a 15% polyacrylamide gel, and analyzed by immunoblotting with an antibundlin monoclonal antibody. Lanes 2 to 9 contain wild-type EPEC and mutant strains, lanes 10 to 15 contain complemented mutant strains, and lanes 16 to 21 contain mutants with control plasmids. Lane 1, pMSD205, unprocessed bundlin control; lane 2, wild-type (WT) EPEC strain E2348/69; lanes 3, 10, and 16, bfpA mutant strain UMD901; lanes 4, 11, and 17, bfpG mutant strain UMD928; lanes 5, 12, and 18, bfpB mutant strain UMD923; lanes 6, 13, and 19, bfpC mutant strain UMD924; lanes 7, 14, and 20, bfpD mutant strain UMD926; lanes 8, 15, and 21, bfpP mutant strain UMD932; lane 9, bfpH mutant strain UMD918.

Effects of bfp mutations on LA. We tested the ability of each mutant strain to adhere to HEp-2 cells. LA is a qualitative adherence pattern recognized as compact microcolonies on the surface of epithelial cells. With the exception of the bfpH mutant strain, all mutants were unable to perform LA. In contrast, the bfpH mutant strain adhered in a manner indistinguishable from that of wild-type EPEC (Fig. 3). In all other mutant strains, addition of the wild-type allele to complement the mutation restored the ability to perform LA. Mutants containing negative control plasmids did not regain the ability to perform LA. These results indicate that, with the exception of bfpH, all the genes tested are required for EPEC to perform LA.

View larger version (111K): [in this window] [in a new window]   FIG. 3.   Localized adherence of wild-type EPEC and mutant strains to epithelial cells. HEp-2 cells were incubated with bacteria for 3 h, washed, fixed, and stained with Giemsa stain. Slides were examined by light microscopy with a 63× objective lens. (A) Wild-type EPEC strain E2348/69. (B) bfpH mutant strain UMD918. (C) bfpC mutant strain UMD924. (D) bfpC mutant strain UMD924 containing plasmid pRPA103. Mutant and complemented strains not shown were indistinguishable from strains UMD924 and UMD924(pRPA103).

Effects of bfp mutations on autoaggregation of EPEC. EPEC strains grown under optimal conditions for BFP expression aggregate into large clusters, and this aggregation is dependent on BFP. When observed by microscopy, these clusters disaggregate in a matter of minutes as the culture cools (6). In contrast to LA, autoaggregation can be quantified, although the precise relationship between the aggregation index and the amount of pili produced is not known since we currently have no method of measuring the latter. Strains were grown in tissue culture medium and tested for BFP-mediated aggregation. To quantify the ability of mutant strains to aggregate in comparison to wild-type strains, the A₆₀₀ of a 4-h culture was measured before and after vortexing and the percent increase after vortexing was calculated as an aggregation index. For strains that aggregate, disaggregation after vortexing was confirmed by microscopy. Table 2 shows the aggregation indices for all the strains tested. All mutant strains had aggregation indices significantly lower than that of wild-type EPEC, and in each case, these indices increased significantly upon addition of complementing plasmids. However, in no case did the aggregation indices of complemented mutants increase to near-wild-type levels, raising the possibility that this failure of full complementation was due to polar effects of the mutations. To determine whether complementing plasmids could interfere with BFP biogenesis, aggregation assays were performed on the wild-type EPEC strain, the wild-type EPEC strain containing pRPA103, the bfpC mutant strain UMD924, and the bfpC mutant strain complemented with pRPA103. Plasmid pRPA103, the complementing plasmid for the bfpC mutant strain UMD924, was chosen for these experiments because strain UMD924 containing this plasmid had the lowest aggregation index of the complemented mutants. In two experiments, each with quadruplicate samples, the wild-type EPEC strain had a mean aggregation index (± standard deviation) of 150.1% ± 70.7% while strain UMD924 had an aggregation index of 2.0% ± 0.3% (P = 0.005 versus wild type). Wild-type EPEC containing plasmid pRPA103 had an index of 13.8% ± 8.0% (P = 0.008 versus wild type), and UMD924 containing pRPA103 had an index of 8.7% ± 3.6% (P = 0.001 versus UMD924; P = 0.19 versus wild type containing pRPA103). Thus, the presence of the complementing plasmid pRPA103 reduced the aggregation index of the wild-type strain to a level similar to that of the bfpC mutant containing the same plasmid. We attribute the decreased aggregation indices seen in the complemented mutant strains not to polar effects of the mutations but rather to stoichiometry problems involving the extra copy of the genes on the complementing plasmids. In this case, pRPA103 introduces extra copies of bfpA, bfpG, bfpB, and bfpC. The resulting overexpression of one or more of these proteins may partially interfere with BFP biogenesis and result in the lower aggregation index. In contrast to the increased aggregation indices seen in all the mutants containing complementing plasmids, aggregation indices of mutants did not change upon addition of control plasmids (data not shown). The bfpH mutant was capable of both aggregation and spontaneous disaggregation (data not shown). These results indicate that all of the genes tested except bfpH are required for EPEC autoaggregation.

Effects of bfp mutations on the biogenesis of BFP. To examine the effects of the bfp mutations on BFP biogenesis, mutant strains and wild-type EPEC were examined by transmission electron microscopy. BFP was detected in samples of wild-type EPEC and bfpH mutant strain UMD918 (Fig. 4). In contrast, we could not detect BFP in the remaining mutant strains. In every case, the addition of the complementing plasmid containing the wild-type allele of the mutated gene to each strain restored the ability of each mutant strain to express BFP, while the addition of the control plasmid did not. These results indicate that bfpH is not required for BFP biogenesis while the remaining mutated genes are.

View larger version (130K): [in this window] [in a new window]   FIG. 4.   Expression of BFP by wild-type EPEC and mutant strains. Strains were grown in DMEM, spotted on Formvar-copper-coated grids, negatively stained with phosphotungstic acid, and examined by electron microscopy. (A) Wild-type EPEC strain E2348-69. (B) bfpH mutant strain UMD918. (C) bfpG mutant strain UMD928. (D) bfpG mutant strain UMD928 containing plasmid pRPA104. Mutant and complemented strains not pictured were indistinguishable from strains UMD928 and UMD928(pRPA104), respectively. Bars, 500 (A to C) and 200 (D) nm.

Prebundlin does not require processing or the presence of any other Bfp proteins to localize to both the inner and outer membranes. To determine the role of each Bfp protein in the localization of bundlin, sucrose density flotation gradients of wild-type EPEC and the various mutant strains were analyzed. In all strains tested, bundlin was found both in fractions that had high activity of NADH oxidase, an inner membrane protein, and in fractions that contained the outer membrane protein OmpA (data not shown). Thus, it appeared that the protein products of none of the bfp genes examined in this study were required for bundlin to localize to fractions containing inner and outer membrane proteins and in fact that prebundlin did not require processing to localize to these fractions. To determine whether localization to fractions containing both membranes is an intrinsic property of bundlin, three additional strains were tested. Sucrose density flotation gradient analyses were performed on recombinant E. coli strain DH5 containing the entire bfp gene cluster, containing only bfpA and bfpP, and containing bfpA alone. The results of these studies using recombinant E. coli are shown in Fig. 5. Peak NADH oxidase activity was found in fractions with densities of 1.17 to 1.20 g/ml. In contrast, Ponceau staining revealed that soluble and insoluble cellular proteins were found in fractions with densities of 1.33 to 1.44 g/ml and that the outer membrane protein OmpA was found in fractions with densities of greater than 1.26. Bundlin (or, in the case of the strain containing bfpA alone, prebundlin) was found in all fractions from the gradient, including those with high NADH oxidase activity, those containing outer membrane proteins, and those with intermediate densities (Fig. 5). Localization to fractions containing both inner and outer membrane proteins is thus independent of the presence of other bfp genes and appears to occur whether or not prebundlin is processed to mature bundlin.

View larger version (67K): [in this window] [in a new window]   FIG. 5.   Distribution of bundlin in membrane fractions from recombinant E. coli strain DH5 containing the entire bfp gene cluster on plasmid pKDS302 (A), containing bfpA on plasmid pMSD230 and bfpP on plasmid pDN19PB (B), and containing bfpA on plasmid pMSD230 and a bfpP deletion on plasmid pDN19PB (C). Fractions collected from the sucrose flotation density gradients were analyzed by immunoblotting with antibodies against bundlin. NADH oxidase activity was measured and displayed as a percentage of the total NADH oxidase activity per fraction. Fractions were loaded on SDS-15% PAGE gels in order from the top to the bottom of the gradient (left to right, respectively). The density of each fraction is indicated above each blot. Variations from the linear trend of increasing density can be ascribed to pipetting error. Data are representative of two separate experiments with similar results. Numbers to the left of each panel indicate molecular mass(es) in kilodaltons.

View larger version (34K): [in this window] [in a new window]   FIG. 6.   Redistribution of previously fractionated prebundlin from low-density (A) and high-density (B) fractions after a second round of sucrose flotation density gradient fractionation. Fractions collected from the sucrose flotation gradient were analyzed by immunoblotting with antibodies against bundlin. NADH oxidase activity was measured and displayed as a percentage of the total NADH oxidase activity per fraction. Fractions were loaded on SDS-15% PAGE gels in order from the top to the bottom of the gradient (left to right, respectively). The density of each fraction is indicated above each blot. Variations from the linear trend of increasing density can be ascribed to pipetting error. Numbers to the left of the panels indicate molecular mass in kilodaltons.

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

Type IV pilus biogenesis is currently a poorly understood process. For many species, the essential genes for type IV pilus systems are scattered at several sites in the bacterial chromosome, making the study of all the genes required for pilus biogenesis very difficult. As all the genes required for the expression of BFP in a recombinant E. coli host are known, the BFP system is an attractive model system for the study of type IV pilus biogenesis. In this study, we examined the roles of seven bfp genes, bfpA, bfpG, bfpB, bfpC, bfpD, bfpP, and bfpH, in BFP biogenesis and LA. Previous studies have shown that mutations in bfpA block BFP biogenesis and LA (13). Ramer et al. reported that a bfpB mutation also blocked BFP biogenesis, autoaggregation, and LA (34), and Bieber et al. reported similar findings for a bfpD mutant (6). Our results confirm these findings and also show that bfpG, bfpC, and bfpP are required for BFP biogenesis and LA, while bfpH is not required for either BFP biogenesis or LA.

To examine the role of each Bfp protein in BFP biogenesis and related phenotypes, mutations were introduced into the wild-type EPEC background in the bfpG, bfpB, bfpC, bfpD, bfpP, and bfpH genes. The mutations were made by disrupting the genes with an aphA3 kanamycin cassette (27) which is designed to be nonpolar. Each mutation that resulted in a discernible phenotype was complemented by the addition of a wild-type copy of the mutant gene in trans on a low-copy-number vector. Fragments of the bfp cluster beginning 270 bp upstream of bfpA and ending within the gene directly downstream of the mutated gene were used to complement mutations in the bfpA, bfpG, bfpB, bfpC, and bfpD genes. This strategy was used to include the native bfp promoter in each complementation plasmid in an effort to restore wild-type levels of the gene products of the mutated genes, as earlier attempts to complement some bfp mutants with plasmids containing only a single gene were unsuccessful (data not shown). Complementation analysis of the mutant strains confirmed that all mutations constructed were nonpolar. All complemented mutants were able to express mature bundlin, and although we were unable to obtain a single Western blot showing equivalent amounts of bundlin in all strains, we have observed no consistent decreases in levels of bundlin in any of the mutants compared to the wild type. All complemented mutants were also able to perform LA and to express BFP. Assays for LA and BFP expression are qualitative, not quantitative. The autoaggregation index is an attempt to quantify a phenotype associated with BFP expression. Autoaggregation studies of wild-type EPEC containing plasmid pRPA103 suggest that stoichiometry problems posed by additional copies of bfp genes on the complementing plasmids are responsible for the lower aggregation indices of complemented mutant strains. These stoichiometry problems may result in decreased levels of BFP expression by altering the ratio of components of the Bfp biogenesis machinery and competing for required interactions. Negative control plasmids for complementation, containing fragments of the bfp cluster that end within the mutated gene, showed that the restoration of the wild-type phenotype was solely due to the addition of the wild-type copy of the mutated gene.

Western blotting showed that in the bfpG, bfpB, bfpC, bfpD, and bfpH mutant strains prebundlin is expressed and processed as in wild-type EPEC, while in the bfpA mutant strain no bundlin is seen and in the bfpP mutant strain prebundlin is expressed but not processed into mature bundlin. These results were expected, as the mutation in bfpA renders bundlin unstable (50) and bfpP encodes a prepilin peptidase (51). We previously showed that BfpP was capable of processing the P. aeruginosa prepilin. In the current report, we detected no processed bundlin in the bfpP mutant. Therefore, BfpP is the only protein in EPEC that processes prebundlin. The remaining mutations did not have a consistent effect on either bundlin expression or processing. These results indicate that the protein products of the bfpG, bfpB, bfpC, bfpD, and bfpH genes are not involved in the expression, stability, or processing of bundlin.

Studies of the abilities of the mutant strains to autoaggregate, perform LA, and express BFP all gave similar results. The bfpA, bfpG, bfpB, bfpC, bfpD, and bfpP mutant strains were unable to autoaggregate into clusters, perform LA, or express BFP. Since all mutations that disabled BFP biogenesis also disabled both autoaggregation and LA, it appears that both these phenotypes require mature pili. The addition of a wild-type copy of the mutated gene in trans restored the ability of the mutant strains to express these phenotypes. In contrast, the bfpH mutant strain was similar to wild-type EPEC in all three phenotypes. In addition, the autoaggregates formed by the bfpH mutant were able to disperse, thus distinguishing this mutant from bfpF mutants, which express pili and form aggregates that do not disperse (6).

Of the genes examined in this study, the functions of few are known. Prebundlin is encoded by the bfpA gene, and we show here that bfpP encodes the prepilin peptidase that cleaves prebundlin to its mature form. Ramer et al. reported that bfpB encodes a member of the secretin family of proteins (34). The functions of bfpG, bfpC, and bfpD are not known. BfpD contains functional motifs found in nucleotide-binding proteins, so it may be required to provide the energy required for BFP biogenesis. BfpG appears to be a small lipoprotein, and BfpC has a single putative transmembrane domain. Based on the phenotypes tested in this study, no function can be ascribed to BfpH. The modest reduction seen in the ability of the bfpH mutant to autoaggregate, while statistically significant, is of questionable biological significance. The bfpH gene is predicted to encode a lytic transglycosylase belonging to a large family. The presumed function of BfpH would be to hydrolyze a region of the peptidoglycan layer to allow for either bundlin or mature BFP to reach the outer membrane. However, several observations suggest that BfpH is not fully functional in EPEC strain E2348/69. First, results reported here show that a bfpH mutation does not disable BFP biogenesis, autoaggregation, or LA. Second, of all the BFP proteins, BfpP and BfpH are the only ones that we were unable to detect by T7 RNA polymerase expression (40). Furthermore, a comparison with other transglycosylases indicates that BfpH lacks several highly conserved residues (24). Thus, it is possible that bfpH has accumulated mutations because it is superfluous or poorly expressed. There are other transglycosylases encoded on the EPEC chromosome (18), and so it is possible that in the absence of BfpH, one of these proteins functionally replaces BfpH. An alternative explanation could be that BFP evolved from or still is a bacteriophage (22), which required or requires the action of the peptidoglycan hydrolase to enter the bacterial cell rather than for pilus biogenesis. The self-transmissible plasmid R64 encodes a type IV pilus and has a BfpH homologue, PilT. Mutation of pilT did not affect pilus biogenesis or sensitivity to pilus-specific bacteriophages (49), and thus, pilT appears to be similar to bfpH in that it is not required for pilus biogenesis. However, pilT mutants exhibit a reduced transfer frequency of the R64 plasmid. The possibility that bfpH is required for an as-yet-unidentified BFP function cannot be excluded.

The method that we chose to study membrane localization, sucrose flotation gradient centrifugation, is considered to be highly reliable, since membrane vesicles must float up to reach the fractions of equivalent density (31). Therefore, insoluble material does not contaminate membrane fractions as it can with sedimentation gradients, and proteins with unusual detergent solubility do not yield spurious results, as they can when differential detergent solubility is used. Previous studies indicated that prebundlin is at least transiently a cytoplasmic transmembrane protein accessible simultaneously to both BfpP and DsbA (50). However, its fate subsequent to the posttranslational modifications catalyzed by these enzymes and prior to incorporation into a pilus is unknown. The results of sucrose flotation gradients appear to show that bundlin localizes to both membrane fractions independent of any Bfp proteins including BfpP. Similar results were obtained in analysis of P. aeruginosa PAK pilin (41, 46) as well as the PulG and Xcp type IV pilus-like proteins from type II secretion systems (29, 33). The localization of prebundlin to fractions containing outer membrane proteins may be surprising since the protein would be predicted to be anchored in the inner membrane by two positively charged residues near the amino terminus followed by a stretch of hydrophobic amino acids. By subjecting fractions containing prebundlin to repeat flotation gradient centrifugation, we demonstrated that localization to inner and outer membrane fractions is not a transient, reversible phenomenon. However, it remains possible that the cell disruption techniques employed could result in cytoplasmic prebundlin or prebundlin complexes becoming irreversibly associated with membrane vesicles. As we are unable to prove that these results are not an artifact of cell disruption, we can conclude only that cell fractionation studies do not shed new light on the possible export of bundlin to the outer membrane during a phase of pilus biogenesis.

Based on the results of this study, we are able only to separate BFP biogenesis into three stages. The first stage is the synthesis of prebundlin. In the second stage, prebundlin simultaneously acquires a disulfide bond in the periplasmic space and is processed to bundlin through the loss of its leader sequence. The DsbA protein is required for the disulfide bond formation (50), and we have shown here conclusively that BfpP is required for the processing of prebundlin to bundlin. In the third stage, mature bundlin is assembled into BFP by a process that is not understood but which does not appear to involve sequential passage from inner to outer membrane. We have shown here that BfpG, BfpB, BfpC, and BfpD are involved in this process, as are BfpE and BfpU (T. E. Blank and M. S. Donnenberg, unpublished data). While our results do not elucidate the roles of any of these proteins in BFP biogenesis, they do show that these proteins do not have a role in the first or second stage of BFP biogenesis and probably act after processing by BfpP. Our results also indicate that BfpH does not play a critical role in BFP biogenesis or function for any of the phenotypes that we tested. While mutations in four bfp genes (bfpI, bfpJ, bfpK, and bfpL) which encode proteins with some resemblance to pilin subunits remain to be analyzed, it appears unlikely that analysis of the effects of single mutations using established approaches will yield new insights into type IV pilus biogenesis. Other approaches that examine interactions between individual components or the localization of Bfp components other than bundlin will be required to form a more coherent picture of the type IV pilus assembly machine.