Identification and Characterization of Assembly Proteins of CS5 Pili from Enterotoxigenic Escherichia coli

Abstract

This study investigated the role of three genes comprising part of the operon which encodes CS5 pili from enterotoxigenic Escherichia coli. In-frame gene deletions were constructed, and the effects on biogenesis of the pili were examined. A deletion in csfB abolished CsfA major subunit accumulation in the periplasm, which could be restored by trans-complementation with a complete copy of the csfB gene. Localization studies using an antibody against CsfB showed that this protein was periplasmically located, and thus CsfB is likely to function as the specific chaperone for CsfA. An in-frame deletion mutation in the csfE gene resulted in pili approximately three times longer than those of the wild-type strain, thereby indicating a role for CsfE in pilus length regulation. Localization studies using an antibody generated against CsfE showed low-level CsfE accumulation in the outer membranes. Modulation of csfE expression in trans did not reduce the mean length of the pilus below that of the wild type, which indicated that CsfE is not rate-limiting for termination of pilus assembly. Interestingly, a deletion in the csfF gene also resulted in an elongated pilus morphology identical to that of the csfE deletion strain. However, unlike CsfE, CsfF was shown to be rate-limiting for termination of assembly, since overexpression of CsfF in a csfF deletion strain resulted in a significant decrease in the mean length of the pilus compared to that of the wild type. When the same construct was introduced into the wild-type strain, pilus expression was abolished. Since CsfF bears significant homology to the proposed CsfB chaperone, CsfF was predicted to act as the specific chaperone for CsfE. A double deletion in the csfB and csfF genes was shown to abolish the periplasmic accumulation of both CsfA and CsfD pilins, which could be restored individually only when the strain was trans-complemented with a wild-type copy of csfB or csfF, respectively. Therefore, CsfF may chaperone not only CsfE but also CsfD. A model for CS5 biogenesis is also proposed based on these and previous observations.

The assembly of cell surface pili on bacteria relies on the delivery of the individual pilin subunits from the periplasm by specific chaperones to an outer membrane assembly protein for translocation onto the cell surface. CS5 pili from human enterotoxigenic Escherichia coli (ETEC) are believed to assemble by a similar mechanism. The DNA region required for pilus assembly consists of six csf genes. Of these, csfA encodes the major subunit, csfD encodes a minor pilin subunit, and csfC encodes the outer membrane usher protein (4, 6, 7).

A specific deletion in the csfD gene showed that CsfD is required for initiation of pilus biogenesis, since bacteria with this deletion showed no translocation of detectable CsfA subunits across the outer membrane and no visible cell surface pilus structures (6). CsfD was not required for the stability of CsfA in the periplasm. Morphological studies on purified CS5 by deep-etch freeze fracture electron microscopy showed that the pilus was a 2-nm flexible fibrillar structure, which was devoid of any tip-associated structures (6). CS5 pili are therefore morphologically similar to the K88 and K99 fibrillae of animal ETEC. This verified the conflicting reports as to the true morphology of the CS5 pilus, which ranged from a semirigid pilus structure to a 5- to 6-nm structure consisting of two fine fibrils wrapped around each other (11, 13). In the latter case, this was likely a consequence of the hydrophobicity of the pilus itself, causing multiple interactions between individual pili, which are frequently observed under the electron microscope (6, 7). Identification of either CsfA or CsfD as the adhesive component of CS5 pili remains unconfirmed.

Previous studies have shown that the csfC gene shares protein sequence similarity with the CooC and CfaC ushers from CS1 and colonization factor antigen I (CFA/I) pili, respectively, but does not share structural similarity by hydropathy plot alignments (6). Cell fractionation experiments and the construction of an in-frame deletion in the csfC gene showed that CsfC is outer membrane located and is responsible for directing both the CsfA and CsfD pilin subunits across the outer membrane, since release of these pilins did not occur in a csfC deletion strain (6). The absence of CsfC did not result in any alteration in the periplasmic accumulation of either pilin. As with the ushers of the Pap, CS1, and 987P biogenesis systems (3, 5, 16), CsfC has been hypothesized to direct the ordered translocation of pilin subunits across the outer membrane, since CsfD is translocated in a csfA deletion strain, while CsfA can be translocated only when CsfD is present (a csfD deletion strain does not permit the release of CsfA) (6). This implies that a specific recognition event between a CsfD-chaperone complex and CsfC is first required before CsfC is able to translocate CsfA subunits across the outer membrane. The functions of csfB, csfE, and csfF remain largely unknown, since no sequence homology with other proteins has been identified from database searches (7). However, CsfB is homologous to CsfF, sharing 30% identity and 51% sequence similarity as well as significant structural similarity, which suggests that these two proteins are likely to exhibit similar functions (7).

The purpose of the present study was to investigate the roles of the three unknown genes of the csf cluster, namely, csfB, csfE, and csfF, by creating a set of in-frame deletions in each gene and determining their effects on CS5 pilus biogenesis. A double in-frame deletion in the csfB and csfF genes was also constructed to examine the combined effect of deleting two genes with similar proposed functions in biogenesis.

MATERIALS AND METHODS

Bacterial strains and culture conditions.

Recombinant E. coli K-12 strain DH5α (Bethesda Research Laboratories, Gaithersburg, Md.) harboring the various plasmid constructs was routinely grown in Luria-Bertani (LB) broth or agar plates, with aeration, at 37°C. For hemagglutination, immunogold electron microscopy, or other pilus expression experiments, strains were grown on CFA agar (9) at 37°C overnight. The E. coli K-12 strain BL21(DE3) (Novagen, Madison, Wis.) harboring the lacI^q plasmid pREP4 (Qiagen, Hilden, Germany) (Table 1) was used as a source for specific His₆-tagged protein expression. Antibiotics were used at the following concentrations: ampicillin, 100 μg/ml; kanamycin, 50 μg/ml. Both 5-bromo-4-chloro-3-indolyl-β-d-galactopyranoside (X-Gal) and isopropyl-β-d-thiogalactopyranoside (IPTG) were used at 40 μg/ml.

TABLE 1.

Plasmids used in this study

Plasmid	Description	Source or reference
pREP4	Kanr; lacIq constitutive	Qiagen
pWKS130	Kanr; low-copy-number cloning vector	22
pQE-31	Ampr; His tag vector	Qiagen
pBAD18-Kan	Kanr; expression vector	10
pPM5631	Ampr; csf region in pGEM-7Zf+	7
pPM5633	Ampr; PstI-XbaI PCR fragment of csfE in pBS-SK+	7
pPM5654	Ampr; ΔcsfE (nt 5428 to 5806) from pPM5631	This study
pPM5657	Kanr; PstI-XbaI PCR fragment of csfE in pWKS130	This study
pPM5684	Kanr; KpnI-XbaI PCR fragment of csfE in pBAD18-Kan	This study
pPM5655	Ampr; ΔcsfF (nt 6070 to 6397) from pPM5631	This study
pPM5659	Kanr; EcoRI-XbaI PCR fragment of csfF in pWKS130	This study
pPM5672	Kanr; EcoRI-XbaI fragment of csfF from pPM5659 in pBAD18-Kan	This study
pPM5662	Ampr; ΔcsfB (nt 2157 to 2602) from pPM5631	This study
pPM5665	Kanr; EcoRI-XbaI fragment of csfB in pWKS130	This study
pPM5679	Ampr; BamHI-KpnI fragment of csfB in pQE-31	This study
pPM5690	Ampr; ΔcsfB (nt 2197 to 2591) ΔcsfF (nt 6070 to 6397) from pPM5655	This study

DNA methods and plasmid construction.

PCR amplifications were carried out according to standard protocols (17). Amplitaq DNA polymerase used in the reactions was purchased from Roche (Mannheim, Germany). Plasmid DNA was transformed into E. coli by using CaCl₂-treated cells as previously described (2). Plasmid DNA preparations, restriction endonuclease digestions, agarose gel electrophoresis, and ligations were performed using standard techniques (17). Nested deletions were performed using an Erase-a-Base kit (Promega, Madison, Wis.). Sequencing reactions on deleted genes were carried out as previously described (7).

Construction of an in-frame deletion mutation in the csfE gene utilized a single NruI site located within the csfE gene across nucleotides (nt) 5597 to 5602 of the characterized sequence (GenBank accession no. AJ224079). This was utilized for exonuclease III-mediated bidirectional deletions using the csf operon plasmid pPM5631 (Table 1) (7), since NruI digestion produces blunt-end fragments that are susceptible to exonuclease III. One such sequenced deletion, which juxtaposed nt 5428 and 5806, thereby deleting 378 nt of the csfE gene and retaining the correct reading frame, was selected for further use and was designated pPM5654 (Table 1). A csfE complementing plasmid was constructed by using the wild-type csfE gene from pPM5631 in a PCR with oligonucleotides 2973 (5"-GGCTGCAGGGAAATATGTAAGCATTACC-3") and 2974 (5"-GGTCTAGACGATTAAAATCGCTCTAAAAA-3"), which contain PstI and XbaI sites, respectively (underlined); the fragment obtained was digested with the appropriate restriction enzymes and cloned into digested pWKS130 (22) to produce pPM5657, with the csfE gene oriented from the inducible lac promoter (Table 1). Plasmid pPM5684 was constructed by excising the csfE gene from pPM5633 (7) with KpnI and XbaI and then cloning it directly into digested pBAD18-Kan (10), such that the csfE gene was under the control of the P_BAD promoter from the araBAD (arabinose) operon (Table 1).

An in-frame deletion mutation in the csfF gene was constructed from pPM5631 by utilizing a unique StuI site located across nt 6229 to 6234 (accession no. AJ224079) for exonuclease III-mediated deletion as described above. One such deletion obtained resulted in the juxtaposition of nt 6070 and nt 6397, thereby deleting 327 nt of the csfF gene and retaining the correct reading frame. This plasmid was selected for further use and was designated pPM5655 (Table 1). To construct a complementing plasmid, the wild-type csfF gene was PCR amplified from pPM5631 using oligonucleotides 2971 (5"-GGAATTCACTATGTAGGGGGAGTAT-3") and 2972 (5"-GGTCTAGATGAACCATAAAG GAAAAAAAG-3"), which contain EcoRI and XbaI sites, respectively (underlined). The PCR fragment obtained was digested with the appropriate restriction enzymes and cloned into EcoRI- and XbaI-digested pWKS130 to produce pPM5659, with the csfF gene oriented from the inducible lac promoter (Table 1). The plasmid was used in trans to complement the introduced in-frame deletion of csfF in pPM5655. To construct pPM5672 with the csfF gene under the control of the P_BAD promoter of pBAD18-Kan (10), pPM5659 was digested with EcoRI and XbaI, and the csfF gene fragment obtained was cloned directly into EcoRI- and XbaI-digested pBAD18-Kan to produce pPM5672 (Table 1).

An in-frame deletion mutation was constructed in the csfB gene by utilizing a double PstI site located within the csfB gene at nt 2318 to 2323 and nt 2327 to 2232 of the characterized sequence (accession no. AJ224079). PstI restriction digestion produces 3" overhangs which are resistant to exonuclease III; therefore, to overcome this, the 3" overhangs were blunt ended by using the 3"-5" exonuclease activity of the Klenow enzyme (1 U/μg of DNA) without deoxynucleoside triphosphates for 10 min at 37°C. Blunt-end DNA fragments are susceptible to exonuclease III-mediated deletions. One csfB deletion mutant obtained resulted in the juxtaposition of nt 2157 and 2602, thereby deleting 444 nt of the csfB gene and retaining the correct reading frame. This plasmid was selected for further use and was designated pPM5662 (Table 1). To construct the csfB complementing plasmid pPM5665, the wild-type csfB gene was PCR amplified from pPM5631 using oligonucleotides 2944 (5"-GGAATTCCCAAGGCAGCTGCTGC-3") and 2945 (5"-GGTCTAGAGCCAGCTCACTTTATCAGC-3"), which contain EcoRI and XbaI sites, respectively (underlined). The PCR fragment obtained was restricted and cloned into EcoRI- and XbaI-digested pWKS130 to produce the csfB gene oriented behind the inducible lac promoter.

To construct an expression plasmid for the CsfB protein in order to generate a polyclonal antiserum, the csfB gene was PCR amplified from pPM5631 using oligonucleotides 3110 (5"-GGCGGATCCGTTCAGTGTTGATTCAATGATA-3") and 3111 (5"-GGGGTACCGGTGTCTTTTAGAGTCATA-3"), which contain BamHI and KpnI sites, respectively (underlined). The csfB fragment was then digested with the same restriction enzymes and cloned into BamHI- and KpnI-digested pQE-31 (Qiagen) such that the csfB gene was cloned in-frame with an N-terminal His₆ tag to produce plasmid pPM5679 (Table 1).

Finally, a double in-frame deletion mutation of the csfB and csfF genes was constructed by utilizing the csfF deletion plasmid pPM5655 to construct a further deletion in the csfB gene. The generation of a specific in-frame deletion in csfB was carried out as described above. One such deletion obtained resulted in the juxtaposition of nt 2197 and 2591, thereby deleting 393 nt of the csfB gene and retaining the correct reading frame. This plasmid was selected for further use and was designated pPM5690 (Table 1).

Western blot analysis.

Preparative protein samples were boiled for 5 min and then separated by sodium dodecyl sulfate (SDS)-15% polyacrylamide gel electrophoresis (PAGE) and transferred onto nitrocellulose membranes (Schleicher and Schuell, Dassel, Germany). For detection of CS5 pili, a rabbit anti-native CS5 pilus antiserum (11) was used at a dilution of 1:5,000. Rabbit anti-CsfE and rabbit anti-CsfB were used at a 1:1,000 dilution. Antigen-antibody complexes were reacted with a horseradish peroxidase-conjugated goat anti-rabbit antibody at a dilution of 1:10,000 and visualized by using a chemiluminescence kit according to the manufacturer's instructions (Roche). Blots were exposed to Cronex 10T medical X-ray film (Sterling Diagnostic Imaging, Stevenage, Hertfordshire, United Kingdom).

Overexpression and cell fractionation.

Specific overexpression of His₆-CsfB from pPM5679 initially involved transforming the plasmid into BL21(DE3) harboring pREP4. Bacteria were grown to the mid-exponential phase of growth (optical density at 600 nm, 0.6) with IPTG added to a final concentration of 1 mM for 3 h at 37°C with vigorous aeration. A 50-ml sample of induced bacterial cells was then fractionated into whole cells, periplasm, cytoplasm, and inclusion bodies as previously described (15). A sample of the inclusion body fraction (1 ml) was subjected to metal exchange chromatography using Ni-nitrilotriacetic acid agarose according to the manufacturer's instructions (Qiagen) in order to purify the His₆-tagged CsfB protein.

Overexpression of CsfE from the T7 promoter has been described previously (7). Cells overexpressing CsfE were fractionated according to the procedure described by Morona et al. (15).

Antiserum preparation.

Antibodies to both CsfE and CsfB were raised in New Zealand White rabbits. Samples of both CsfE and CsfB were run on SDS-15% PAGE gels, and bands corresponding to these proteins were excised from the gel and were each emulsified with 0.5 ml of Freund's complete adjuvant and an equal volume of 1× phosphate-buffered saline (PBS) in an Ultra-Turrax blender (Janke and Kunkel, Staufen, Germany). At day 0 each rabbit was immunized subcutaneously with 10 injections of the protein homogenate at different sites (100 μl per site). Periodic boosters were given in Freund's incomplete adjuvant. Each rabbit was exsanguinated by cardiac puncture, and the serum obtained was stored at −20°C. The antiserum was absorbed against live E. coli K-12 strains harboring the specific plasmid vectors used in the cloning process in the presence of 0.02% sodium azide. Absorbed antiserum was also incubated across nitrocellulose membranes impregnated with whole-cell extracts in order to further remove nonspecific antibodies. The specificities of the antiserum against CsfB and the antiserum against CsfE were examined by Western blot analysis.

Rapid pilus isolations.

Bacterial strains were grown to confluence at 37°C on CFA plates (9), harvested in 1 ml of PBS, and incubated with vigorous shaking at 56°C for 20 min in an Eppendorf reaction tube. The cell suspension was clarified by centrifugation for 5 min at 8,000 × g. Samples of the heat extracts were then subjected to SDS-15% PAGE as previously described (6).

Hemagglutinations and slide agglutinations.

Hemagglutination and slide agglutination experiments were performed as previously described (6).

Immunogold electron microscopy.

Immunogold labeling and negative staining of bacterial strains were carried out essentially as previously described (6) by using an antiserum against CS5 pili added at a 1:10 dilution and protein A conjugated with 20-nm-diameter gold beads (ICN, Costa Mesa, Calif.) at a 1:40 dilution.

RESULTS

Effect of a deletion in csfE on CS5 pilus biosynthesis.

The constructed in-frame deletion mutation of the csfE gene on pPM5654 was initially assessed by slide agglutinations for any phenotypic effects on CS5 pilus assembly. E. coli K-12 harboring pPM5654 did not agglutinate with the rabbit anti-CS5 pilus antiserum, whereas strong agglutination was observed in E. coli K-12 harboring pPM5631 and in E. coli K-12 harboring pPM5654 complemented with pPM5657. This indicated that the introduced csfE deletion may alter the assembly of CS5 pili. Immunogold electron microscopy was then performed on the strains by using a rabbit anti-CS5 pilus antiserum. E. coli K-12 harboring pPM5631 or pPM5654 complemented with pPM5657 produced peritrichous CS5 pili, while the negative-control strain, E. coli K-12 harboring pGEM-7Zf⁺, was devoid of pili (Fig. 1A to C). E. coli K-12 harboring pPM5654 did produce CS5 pili on the cell surface, but in far lower numbers, while the morphology of the bundled pili suggested that they were much longer than the corresponding CS5 pili in E. coli K-12 with pPM5631 or the complemented csfE mutant (Fig. 1D and E). These data show that CsfE may function in the control of the length of CS5 pili. The apparent drastic reduction in the number of cell surface-associated CS5 pili may explain why E. coli K-12 with pPM5654 did not readily agglutinate with the rabbit anti-CS5 pilus antiserum.

FIG. 1.

Immunogold electron micrographs of 1% uranyl acetate-stained E. coli K-12 harboring either pGEM-7Zf⁺ (A), pPM5654 (ΔcsfE) plus pPM5657 (csfE) (B), pPM5631 (C), or pPM5654 (D and E). The different bacterial strains were reacted first with a rabbit antiserum raised against native CS5 pili (1:10) and then with protein A conjugated to 20-nm-diameter gold particles (1:40). Bars, 500 nm.

The csfE deletion strain was also tested for the ability to cause mannose-resistant hemagglutination of human group A⁺ erythrocytes. Despite the ability of E. coli K-12 with pPM5654 to assemble CS5 pili on the cell surface, although they were fewer and much longer than wild-type pili, no hemagglutination ability was conferred. The longer CS5 pili produced by the csfE mutant strain are likely to be more susceptible to mechanical shearing caused by resuspension of the bacteria in PBS, which may explain why this strain was unable to mediate hemagglutination.

Western immunoblot analysis following SDS-15% PAGE using a rabbit anti-CS5 pilus antiserum was also performed on whole-cell, periplasmic, and heat extract samples prepared from E. coli K-12 harboring either pGEM-7Zf⁺, pPM5631, pPM5654, or pPM5654 complemented with pPM5657. E. coli K-12 harboring pPM5654 showed a small reduction in the accumulation of CsfA and CsfD pilin subunits in heat extract samples compared to E. coli K-12 harboring pPM5631 (Fig. 2). This reduction was examined further by preparing whole-cell and periplasmic extracts to ascertain whether the accumulation of CsfA and CsfD pilin subunits is reduced in a csfE deletion strain. E. coli K-12 harboring pPM5654 did not result in any significant change in the accumulation of either CsfA or CsfD in the periplasm compared to E. coli K-12 harboring pPM5631 or E. coli K-12 harboring pPM5654 complemented with pPM5657 (Fig. 2). This suggested that CsfE is not required for the stability of CsfA or CsfD in the periplasm of E. coli K-12. Increased amounts of CsfA and CsfD were detected in whole cells of E. coli K-12 with pPM5654 compared to E. coli K-12 harboring pPM5631 or pPM5654 complemented with pPM5657 (Fig. 2). This is likely a reflection of the overall hydrophobicity of CS5 pili (6, 7), where detached longer CS5 pili reassociate with cell surface pili and are centrifuged with the whole cells. Examination of electron micrographs of the csfE mutant strain shows a mesh-like network of associated pili on these strains (Fig. 1D and E), which provides some evidence for long, cell-free pili reassociating with cell surface pili attached to the bacteria.

FIG. 2.

Effect of a csfE deletion on CsfA and CsfD accumulation in heat extract, whole-cell, and periplasmic samples by Western blot analysis. Western blotting was performed using a rabbit anti-CS5 pilus antiserum (1:5,000). Lanes 1, pGEM-7Zf⁺; lanes 2, pPM5631; lanes 3, pPM5654 (ΔcsfE); lanes 4, pPM5654 plus pPM5657 (csfE). Samples represent 1 × 10⁹ cells (heat extracts and whole cells) or 2 × 10⁹ cells (periplasms).

Overexpression of CsfE, antiserum generation, and cellular location.

The CsfE protein was overexpressed from pPM5633 (Table 1), which contains the csfE gene behind the T7 promoter, using the T7 promoter-expression system protocol (20) as previously described (7). Bacterial cells overexpressing CsfE were then fractionated (15), and CsfE was found to localize exclusively in the inclusion body fraction (Fig. 3A). CsfE was then excised from the gel and used to immunize New Zealand White rabbits for specific antiserum generation.

FIG. 3.

Overexpression of CsfE and localization in E. coli K-12 harboring pPM5631. (A) Coomassie brilliant blue-stained SDS-15% PAGE gel of inclusion body fractions of E. coli K-12 harboring pBS-SK⁺ or pPM5633. Overexpression of CsfE from the T7 promoter in pPM5633 has been described previously (7). The CsfE protein is indicated by an asterisk. The preparation of the antiserum against CsfE is described in Materials and Methods. (B) E. coli K-12 cells harboring pPM5631 or pGEM-7Zf⁺ were fractionated into whole-cell (w/c), inclusion body (i/b), cytoplasm (cyto), periplasm (peri), whole-membrane (w/m), outer-membrane (o/m), and inner-membrane (i/m) fractions, and CsfE was specifically detected using an anti-CsfE antiserum (1:1,000). A Coomassie brilliant blue-stained SDS-15% PAGE gel of the inner- and outer-membrane fractions, showing the two major E. coli porins almost exclusively localized in the outer membrane, is also shown. A 20.1-kDa molecular mass standard is indicated.

The antiserum generated was used to localize CsfE in E. coli K-12 by fractionation experiments. E. coli K-12 bacteria harboring pPM5631 or pGEM-7Zf⁺ were fractionated into whole cells, inclusion bodies, cytoplasms, periplasms, whole membranes, outer membranes, and inner membranes. Western immunoblot analysis on the fractions using the rabbit anti-CsfE antiserum showed that CsfE localized in the outer-membrane fraction (Fig. 3B). There was no contamination between the outer and inner membranes, since the outer-membrane porins of E. coli (OmpC/OmpF) localized exclusively in the outer-membrane fraction as judged by Coomassie brilliant blue staining (Fig. 3B). The low band intensity of CsfE in this fraction and its absence in the periplasm suggest that CsfE is expressed at much lower levels in the cell than CsfA or CsfD, although it must also be taken into account that the CsfE protein may be far less immunogenic than the other CS5 pilins, thereby rendering the antiserum used in detection far less effective.

The question of whether CsfE is assembled as a minor component of the CS5 pilus was investigated. CsfE could not be detected in denatured pure CS5 pili weighing up to 100 μg or in heat extracts representing 5 × 10⁹ cells, a number that has been shown by Coomassie brilliant blue gel analysis to release more than 100 μg of CS5 pili (data not shown). Instead, CsfE remains in the outer membranes of these cells, which indicates either that the association of CsfE with the CS5 pilus is a relatively weak interaction or that CsfE directly associates with the CsfC usher (6) in the outer membrane rather than with the pilus itself.

csfF deletion mutants produce longer CS5 pili morphologically similar to those produced by csfE deletion mutants.

The effect of a deletion mutation in the csfF gene on CS5 pilus biogenesis was initially assessed by slide agglutinations using a rabbit anti-CS5 pilus antiserum. E. coli K-12 harboring pGEM-7Zf⁺ or pPM5655 did not agglutinate with the antiserum. Conversely, E. coli K-12 with pPM5631 or E. coli K-12 with pPM5655 complemented with pPM5659 showed immediate agglutination at room temperature. This indicates that the csfF deletion mutation may alter the cell surface assembly of CS5 pili. Examination of the same strains by immunogold electron microscopy revealed that E. coli K-12 harboring pPM5655 showed cell surface assembly of CS5 pili, but in far lower numbers, while the pili appeared to be far longer overall than those of the wild-type or the complemented mutant (Fig. 4A and B). This phenotype was very similar to that observed in a csfE deletion strain (Fig. 1D and E), which suggested that both CsfE and CsfF may control pilus length. E. coli K-12 containing pPM5655 complemented with pPM5659 produced peritrichous CS5 pili (Fig. 4C). No hemagglutination of the erythrocytes was observed with E. coli K-12 harboring pPM5655; however, hemagglutination of the erythrocytes was restored when E. coli K-12 harboring pPM5655 was complemented with pPM5659. The reduction in the number of cell surface pili, along with the increased susceptibility to shearing forces on longer pili, may also explain the negative slide agglutination and hemagglutination results obtained for the csfF deletion mutant strain.

FIG. 4.

Immunogold electron micrographs of 1% uranyl acetate-stained E. coli K-12 harboring pPM5655 (ΔcsfF) (A and B) or pPM5655 plus pPM5659 (csfF) (C). The different bacterial strains were reacted first with a rabbit antiserum raised against native CS5 pili (1:10) and then with protein A conjugated to 20-nm-diameter gold particles (1:40). Bars, 500 nm.

Western blot analysis of the csfF deletion mutant.

Deletion mutation of the csfF gene did not result in any obvious difference in the periplasmic accumulation of either the CsfA or the CsfD pilin from that for E. coli K-12 harboring pPM5631 or E. coli K-12 harboring pPM5655 complemented with pPM5659, as observed by Western analysis using an anti-CS5 pilus antiserum (Fig. 5). An increase in the concentration of CsfA in whole cells was evident in the csfF deletion strain (E. coli K-12 with pPM5655) over that in E. coli K-12 harboring pPM5631 or pPM5655 complemented with pPM5659 (Fig. 5). This is similar to the observation for whole cells of the csfE deletion mutant strain (Fig. 2) and is also likely a reflection of the hydrophobic nature of CS5 pili. Interestingly, no parallel increase in the accumulation of CsfD in whole cells was detected.

FIG. 5.

Effects of a csfF deletion on CsfA and CsfD accumulation in heat extract, whole-cell, and periplasm samples by Western blot analysis. Western blotting was performed using a rabbit anti-CS5 pilus antiserum (1:5,000) on E. coli K-12 strains harboring either pGEM-7Zf⁺ (lane 1), pPM5631 (lane 2), pPM5655 (lane 3), or pPM5655 plus pPM5659 (lane 4). Samples represent 1 × 10⁹ cells (heat extracts and whole cells) or 2 × 10⁹ cells (periplasms).

Examination of heat extracts derived from these strains revealed that E. coli K-12 harboring pPM5655 produced pili with apparently little or no CsfD minor pilin compared to those produced by E. coli K-12 with pPM5631 or E. coli K-12 with pPM5655 complemented with pPM5659 (Fig. 5). The amount of CsfA detected remained relatively constant. These data showed that CsfF may also play a role in delivering CsfD across the outer membrane during CS5 biogenesis but is not necessarily required for stabilizing CsfD in the periplasm, since CsfD levels remained unchanged in this fraction in a csfF deletion mutant (Fig. 5). Clearly, some CsfD must be assembled into the pilus, since it has been shown that CsfD is required for pilus initiation (6).

Expression of CsfE and CsfF and their effects on pilus length and number.

The results have indicated that CsfE and CsfF appear to play a role in regulating the length of the CS5 pilus. csfE and csfF were cloned into pBAD18-Kan under the control of the tightly regulated P_BAD promoter from the araBAD (arabinose) operon to produce pPM5684 and pPM5679, respectively. The presence of arabinose (at 0.002, 0.01, or 0.05%) in CFA agar induced expression of the genes. Following overnight incubation, bacteria were examined under the electron microscope after immunogold labeling and negative staining with 1% uranyl acetate, and numbers and lengths of pili were recorded. The mean number of surface-located pili ± standard error of the mean (SEM) was calculated on 50 randomly selected bacteria, while the mean length of the pili ± SEM was calculated by recording the lengths of 5 pili on each of 10 randomly selected bacteria (n = 50). Since CS5 pili bundle, the mean number and length are representative of discrete bundles of pili.

It must be noted that the presence of the pBR322 ori of pBAD18-Kan leads to a reduction in copy number when coexpressed with the incompatible pGEM-7Zf⁺ ColE1 ori. When this incompatibility was examined in the wild-type E. coli K-12 strain harboring pPM5631 and pBAD18-Kan in the presence of 0.01% arabinose (Table 2), it led to a reduction in the mean number of cell surface pili (11.82 versus 41.34 in the wild type) but did not significantly affect mean pilus length (0.522 μm versus 0.55 μm in the wild type) (P = 0.42). Similarly, no reduction in pilus length was observed in the presence of increasing or decreasing concentrations of the inducer (data not shown). Therefore, this two-plasmid system is suitable for examining the effects of modulating CsfE and CsfF expression on pilus length.

TABLE 2.

Effects of expression of CsfE or CsfF on CS5 pilus number and length

Plasmid(s) harbored by E. coli K-12	No. of pili per cella	Pilus length (μm)b
pPM5631	41.34 ± 1.349	0.55 ± 0.024
pPM5631 + pBAD18-Kan + 0.01% arabinose	11.82 ± 1.090	0.522 ± 0.075c
CsfE deletions
pPM5654	5.22 ± 0.475d	1.859 ± 0.069d
pPM5654 + pPM5657 + IPTG	22.72 ± 1.085d	0.771 ± 0.045d
pPM5654 + pPM5684 + 0.002% arabinose	10.62 ± 0.603d	1.036 ± 0.079d
pPM5654 + pPM5684 + 0.01% arabinose	9.88 ± 0.579d	0.721 ± 0.035e
pPM5654 + pPM5684 + 0.05% arabinose	6.36 ± 0.447d	0.735 ± 0.038d
pPM5631 + pPM5684 + 0.01% arabinose	18.46 ± 1.128d	0.539 ± 0.018f
CsfF deletions
pPM5655	14.08 ± 0.893d	1.540 ± 0.091d
pPM5655 + pPM5659 + IPTG	19.30 ± 0.703d	0.771 ± 0.023d
pPM5655 + pPM5672 + 0.002% arabinose	4.96 ± 0.536d	0.247 ± 0.015d
pPM5655 + pPM5672 + 0.01% arabinose	NDg	ND
pPM5631 + pPM5672 + 0.002% arabinose	0.480 ± 0.104	ND

^aMean number of pili on 50 randomly selected bacteria ± SEM.

^bMean length of 5 pili on 10 randomly selected bacteria (n = 50) ± SEM.

^cP < 0.42 compared to the wild type (pPM5631) by the unpaired two-tailed t test.

^dP < 0.0001 compared to the wild type (pPM5631) by the unpaired two-tailed t test.

^eP = 0.002 compared to the wild type (pPM5631) by the unpaired two-tailed t test.

^fP = 0.68 compared to the wild type (pPM5631) by the unpaired two-tailed t test.

^gND, not determined.

In-frame deletions in csfE (pPM5654) and csfF (pPM5655) resulted in E. coli K-12 mutants with significantly longer pili (means, 1.859 and 1.540 μm, respectively) than those of the wild type (mean, 0. 55 μm) (P < 0.0001) (Table 2). There were also large decreases in the mean numbers of surface pili assembled for the csfE (mean, 5.22) and csfF (mean, 14.08) deletion strains compared to those for the wild type (mean, 41.34) (Table 2).

When the csfE mutant was complemented in trans with pPM5657 in the presence of IPTG, the mean number of pili per cell increased (22.72 ± 1.085), while the mean pilus length decreased (0.771 ± 0.045 μm), compared to those for E. coli K-12 harboring pPM5654 (Table 2). However, neither the mean number of pili nor the mean pilus length was restored to that observed in the wild-type strain (P < 0.0001), which may be a consequence of expression of csfE from a low-copy-number plasmid versus the high copy number of pPM5631. Expression of CsfE from the P_BAD promoter on pPM5684 was carried out by addition of either 0.002, 0.01, or 0.05% arabinose, which specifically induces CsfE expression. For E. coli K-12 harboring pPM5654 and pPM5684 in the presence of 0.002% arabinose, the mean pilus length was 1.036 μm; this decreased further when the concentration of arabinose was increased to 0.01% (0.721 ± 0.035 μm) or 0.05% (0.735 ± 0.038 μm). The pili produced were still significantly longer than those observed in the wild type (P < 0.0001). When pPM5684 was expressed with pPM5631 in the presence of 0.01% arabinose, the mean pilus length was 0.539 ± 0.018 μm, which was not significantly different from that for pPM5631 alone (P = 0.68) (Table 2). The data showed that CsfE is specifically involved in length regulation but is not rate-limiting for length control, since increased expression of CsfE does not reduce mean pilus length below that observed in the wild-type strain.

In contrast, CsfF is rate-limiting for determination of pilus length. E. coli K-12 harboring both pPM5655 and the compatible low-copy-number plasmid pPM5659 grown in the presence of IPTG reduced the mean pilus length to 0.771 ± 0.023 μm, but pili were still significantly longer than those observed in the wild type (P < 0.0001) (Table 2). However, E. coli K-12 harboring pPM5655 complemented with pPM5672 grown in the presence of 0.002% arabinose resulted in a significant decrease in mean pilus length, to 0.247 ± 0.015 μm (P < 0.0001). When the same strain was grown in the presence of 0.01% arabinose, no surface-associated pili were evident. Moreover, when E. coli K-12 harboring pPM5631 was complemented with pPM5672 in the presence of 0.002% arabinose, the bacteria were virtually devoid of pili (mean number per cell, 0.480 ± 0.104), and thus no mean pilus length was recorded (Table 2). Taken together, these results indicate a direct association between the level of expression of CsfF in the periplasm and the mean length of the pilus. Therefore, CsfF is rate-limiting for the determination of pilus length.

Effect of an in-frame csfB deletion mutation on CS5 pilus biogenesis.

To examine the role of csfB in pilus assembly, an in-frame deletion mutation in the csfB gene was constructed in E. coli K-12 and designated pPM5662. Initial slide agglutination experiments using an anti-CS5 antiserum showed that strains harboring pPM5662 did not agglutinate. When the same strains were examined by immunogold electron microscopy, E. coli K-12 harboring pPM5662 was devoid of CS5 pili, while E. coli K-12 containing pPM5662 complemented with pPM5665 produced peritrichous CS5 pili (Fig. 6). The csfB deletion mutant strain was also tested for the ability to cause mannose-resistant hemagglutination of human group A⁺ erythrocytes. As expected from the immunogold electron microscopy data, no hemagglutination ability was conferred by E. coli K-12 with pPM5662, while hemagglutination of the human erythrocytes was restored when the csfB mutant was complemented with pPM5665. Therefore, CsfB is required for cell surface assembly of CS5 pili.

FIG. 6.

Immunogold electron micrographs of 1% uranyl acetate-stained E. coli K-12 harboring pPM5662 (A) or pPM5662 plus pPM5665 (B). The different bacterial strains were reacted first with a rabbit antiserum raised against native CS5 pili (1:10) and then with protein A conjugated to 20-nm-diameter gold particles (1:40). Bars, 500 nm.

CsfB acts as a chaperone for the major subunit CsfA.

Western immunoblot analysis using a rabbit anti-CS5 pilus antiserum following SDS-15% PAGE was performed on heat extract, periplasm, and whole-cell samples prepared from the strains to examine the effects of a csfB deletion on CsfA and CsfD accumulation. E. coli K-12 harboring pPM5662 showed an abolition of CsfA, but not CsfD, in heat extract, periplasm, and whole-cell samples compared to E. coli K-12 harboring pPM5631 and E. coli K-12 with pPM5662 and pPM5665 (Fig. 7). Therefore, the absence of the CsfB protein in the cells affects the periplasmic accumulation and hence the release of CsfA, but not those of CsfD, since CsfD is readily detected. There does, however, appear to be a slight reduction in CsfD accumulation in the periplasm of the csfB deletion strain compared to that in the wild type (Fig. 7). Although CsfB bears no significant similarity to any member of the immunoglobulin-like chaperone family (12), CsfB is predicted to act as a specific chaperone for CsfA.

FIG. 7.

Effects of a csfB deletion on CsfA and CsfD accumulation in heat extract, whole-cell, and periplasm samples as determined by Western blot analysis. Western blotting was performed using a rabbit anti-CS5 pilus antiserum (1:5,000) on E. coli K-12 strains harboring either pGEM-7Zf⁺ (lane 1), pPM5631 (lane 2), pPM5662 (ΔcsfB) (lane 3), or pPM5662 plus pPM5665 (csfB) (lane 4). Samples represent 1 × 10⁹ cells (heat extracts and whole cells) or 2 × 10⁹ cells (periplasms).

CsfB is periplasmically located in E. coli K-12 expressing CS5 pili.

The CsfB protein was overexpressed from pPM5679 and was found to localize in the inclusion body fraction (data not shown). The CsfB protein was subsequently purified by His₆-specific chromatography and SDS-15% PAGE and was then used for antiserum generation. The rabbit anti-CsfB antiserum was used to localize CsfB in the CS5 pilus-expressing strain, E. coli K-12 harboring pPM5631. The majority of chaperones described are located in the periplasms of pilus-expressing bacteria; however, the CooB chaperone of CS1 pili has been shown to associate directly with the assembly protein CooC in the outer membrane and therefore is not strictly periplasmically located (21). Western analysis of fractionated E. coli K-12 harboring pPM5631 or pGEM-7Zf⁺ showed that CsfB could be detected only in the periplasmic fraction and was not outer membrane associated (Fig. 8). Furthermore, no CsfB could be detected within purified denatured CS5 pili ranging from 10 to 150 μg of protein, which indicates that CsfB is also unlikely to form part of the mature pilus structure.

FIG. 8.

Localization of CsfB in E. coli K-12 harboring pPM5631 by use of an anti-CsfB antiserum. E. coli K-12 bacteria harboring pGEM-7Zf⁺ or pPM5631 were fractionated into periplasmic (peri), whole-membrane (w/m), outer-membrane (o/m), and inner-membrane (i/m) fractions, and CsfB was specifically detected by use of an anti-CsfB antiserum (1:1,000). Molecular mass standards are indicated.

Characterization of a csfB csfF double-deletion mutant and trans-complementation.

A double in-frame deletion mutation in both the csfB and csfF genes was constructed and designated pPM5690 in order to identify the effect of removing these two similar proteins on pilin subunit accumulation in the periplasm. Therefore, periplasmic fractions were prepared from E. coli K-12 harboring pPM5631 and from E. coli K-12 containing pPM5690 either alone or complemented in trans with a wild-type copy of csfB (pPM5665) or csfF (pPM5659). The double in-frame deletion mutation in csfB and csfF resulted in the complete absence of CsfA subunits in the periplasm by Western blot analysis, along with a drastic reduction, to almost undetectable levels, in the accumulation of CsfD (Fig. 9). The virtual absence of CsfD in this fraction was unexpected, given that an in-frame deletion of csfF from plasmid pPM5655 does not affect the periplasmic accumulation of CsfD (Fig. 9).

FIG. 9.

Western blot analysis of the effect of a csfB csfF deletion, and of complementation with csfB or csfF provided in trans, on the accumulation of pilin subunits in the periplasm. Periplasmic extracts were prepared from E. coli K-12 harboring either pGEM-7Zf⁺, pPM5690 (ΔcsfB csfF), pPM5690 plus pPM5659, or pPM5690 plus pPM5665, and CsfA and CsfD were detected with a rabbit anti-native CS5 pilus antiserum (1:5,000). A very faint CsfD band was detected in the double-deletion-mutant (pPM5690) lane. Each sample represents 2 × 10⁹ cells.

E. coli K-12 harboring pPM5690 complemented with pPM5665 restored the periplasmic accumulation of CsfA, which was absent in E. coli K-12 with pPM5690 (Fig. 9). This is consistent with the observation that CsfB is likely to act as a periplasmic chaperone for CsfA. The level of restoration achieved was similar to that observed for E. coli K-12 with pPM5631. When E. coli K-12 harboring pPM5690 was complemented in trans with the csfF plasmid pPM5659, only CsfD and not CsfA accumulation was restored in the periplasm (Fig. 9), albeit at a lower level than that in the wild type. These data show that CsfF is required for CsfD stability in the periplasm in the absence of CsfB. Therefore, CsfF is predicted to be a specific chaperone for CsfD. Given that CsfF has chaperone-like properties for CsfD, CsfF is also predicted to chaperone CsfE to terminate pilus assembly, since csfF mutants produce a phenotype identical to that of csfE mutants and CsfE encodes an outer-membrane-located minor pilin.

DISCUSSION

This study sought to identify and characterize the three remaining unknown genes of the csf gene cluster, csfB, csfE, and csfF, which are required for biogenesis of CS5 pili. Previous studies had established that csfA encodes the major subunit, csfD encodes the minor subunit, and csfC encodes the outer membrane usher protein (4, 6, 7). The present study has verified that the csf gene cluster of CS5 is more complicated than the other known human ETEC pilus operons, in that it consists of six contiguous genes, each possessing an indispensable function in the biogenesis of the mature pilus.

A constructed in-frame deletion mutation of the csfE gene resulted in pili that were more than three times longer than those of the wild-type strain, highlighting a specific role for CsfE in controlling the length of assembled CS5 pili. However, expression of CsfE did not reduce the length of the pilus from that of the wild type, which suggested that CsfE is not rate-limiting for pilus length regulation. Specific overexpression of other, analogous proteins such as PapH from Pap pili and MrpB from MR/P pili of Proteus mirabilis resulted in significant reductions in pilus length compared to that of the wild type (1, 14). A recent study has shown that the use of araBAD promoters to express genes of interest may not be modulated as first described (10); rather, at subsaturating levels of induction of the araBAD promoter, gene expression is not uniform with respect to individual cells (19). Therefore, a certain amount of caution must be applied to these results; however, a previous study using the minor pilin gene cooD under the control of the araBAD promoter found induction of CS1 pili from a mean of 28 pili/cell to >150 pili/cell in the majority of cells examined, suggesting that induction may have been uniform (16).

CsfE was localized to the outer membrane in E. coli K-12, as with other pilus length regulators, and as such is predicted to encode a minor pilin subunit, although no CsfE could be detected in pure CS5 pilus preparations. The difficulty in this study was determining a link between CsfE and the anchoring of CS5 pili to the cell surface. In other systems such as Pap, the PapH protein was shown to regulate the length of the pilus and was predicted to anchor the pilus to the cell surface, since an increased number of pili was detected in the supernatant of a papH mutant (1). Although this may be the case for such studies, the increased shearing forces on longer pili must be taken into account in harvesting the bacteria. The longer a pilus becomes, the greater, presumably, is the corresponding increase in shearing forces on the pilus. Based on the results obtained and on analogy with other systems, it is hypothesized that the incorporation of CsfE with the outer membrane usher CsfC results in a stable binding complex which cannot be displaced by further chaperone-subunit complexes, thereby leading to termination of assembly. This event may also provide a stable platform for association of the pilus with the cell envelope. The low level of expression of CsfE in the cell suggests that association of CsfE with other pilins would be a rare event.

A deletion mutation in the csfB gene was found to completely abolish CsfA subunit accumulation in the periplasm, and therefore bacteria with this deletion did not produce detectable cell surface CS5 pili. However, the deletion in CsfB did not affect the transport of the minor pilin subunit CsfD into the periplasm and across the outer membrane, since CsfD subunits were detected in heat extract samples from the csfB deletion strain. CsfB is therefore likely to function as the major subunit-specific chaperone. Localization studies using an anti-CsfB antiserum on fractions derived from the wild-type strain confirmed the periplasmic location of CsfB and showed that CsfB did not form part of the mature pilus structure.

A deletion mutation constructed in the csfF gene resulted in two distinct changes in the pilus structure. Firstly, the pili produced in this strain were approximately three times longer than those observed in the wild type, and secondly, no CsfD was detected in the longer pilus structures. Specific modulation of CsfF expression in a csfF deletion strain resulted in significantly shorter pili, and when CsfF was introduced into the wild-type strain, cell surface pilus expression was abolished. This showed that CsfF is rate-limiting for the control of pilus length, and unlike CsfE, increasing expression of CsfF results in pili that are significantly shorter than those of the wild type. Based on its amino acid similarity to the predicted CsfB chaperone, CsfF is predicted to function as a specific chaperone for delivering CsfE to the outer membrane assembly protein CsfC in order to terminate pilus elongation.

Although no CsfD minor subunits could be detected in the pili produced in the csfF deletion strain, CsfD was detected at levels comparable to those of the wild type in the periplasm. Therefore, the absence of csfF abrogates delivery of CsfD across the outer membrane. This implies that CsfF may also function as the specific chaperone for CsfD but may not be required for the stability of CsfD in a csfF deletion strain. Clearly, some CsfD must be assembled onto the cell surface, since CsfD has previously been shown to be absolutely required for initiation of CS5 pilus biogenesis (6). However, the level of CsfD is undetectable under the conditions used.

To explain this observation, a double in-frame deletion mutation in the csfB and csfF genes was constructed, thereby establishing directly whether some functional redundancy may exist between CsfB and CsfF. This double-deletion mutant resulted in the absence of CsfA and a large decrease in the level of detectable CsfD subunits in the periplasm. When a wild-type copy of csfF was provided in trans, CsfD levels in the periplasm were stabilized. Likewise, when csfB was reintroduced in trans, CsfA accumulation in the periplasm was restored. Therefore, CsfF appears to be directly responsible for the stability of CsfD in the periplasm, but probably also for that of CsfE, and thus functions as a chaperone.

So why is CsfD detected in the periplasm, but not in the final pilus structure, in the absence of CsfF but not in the absence of CsfB? The presence of CsfD in the periplasm in a csfF mutant can be explained in terms of the relative copy number of the plasmid system used and the altered binding specificity of CsfD in the absence of CsfF. Since the regulator of CS5 pili expression has never been directly demonstrated, significant expression of this pilus can be achieved only from high-copy-number plasmids (7). In the absence of CsfF, CsfD may interact weakly with CsfB alone or possibly with CsfB-CsfA complexes in the periplasm. It is unclear how pilus initiation occurs in this case, but it is possible that a CsfB-CsfA-CsfD complex, or indeed CsfD aggregates alone, may be able to associate with CsfC in the outer membrane, resulting in subunit translocation and the targeting of further complexes to the usher. Clearly, complexes of the CsfA-CsfB type are favored for assembly in this case, since CsfD is not detected. When the double deletion was complemented with csfB and csfF on low-copy-number plasmids in trans, the decreases in levels of these proteins relative to the subunits originating from the high-copy-number plasmids could lead to the chaperones binding only the pilins for which they have the greatest affinity. This may explain why only CsfA accumulates when CsfB is provided in trans and likewise why CsfD predominates when CsfF is provided. Since no CsfA is retained in a single csfB mutant, it is unlikely that CsfF shows any binding specificity for CsfA to help stabilize this protein in the periplasm.

The CsfB and CsfF chaperones do not show any discernible sequence or structural similarity with the prototypic member of the immunoglobulin-like chaperones, PapD (P = 0.89). CsfB and CsfF did, however, share 4 out of 10 and 3 out of 10, respectively, of the invariant residues associated with the 26 known members of the superfamily of the immunoglobulin-like chaperones (12). The present report on the CS5 pilus biogenesis operon is the first description of a dual-chaperone system for any human ETEC pilus system, in which one chaperone appears to be directly responsible for stabilizing and delivering the major subunit protein and the other is specifically responsible for stabilizing and delivering minor subunits to the outer membrane. Only one other pilus biogenesis system, 987P, which is an important colonization factor on porcine ETEC strains, contains two chaperones with similar functions (8). FasB is the periplasmic chaperone for the major pilin subunit FasA, and similarly, FasC chaperones the 987P adhesin FasG (8). The CsfB and CsfF chaperones bear no homology to the 987P chaperones. Binding experiments will need to be performed to specifically show that a protein interaction occurs between the CS5 chaperones and their cognate pilins.

Interestingly, the csf cluster also shows a genetic organization similar to that of 987P from porcine ETEC (18). Both systems utilize two chaperones, one for minor subunits and one for major subunits, and if the fasH regulatory gene is not considered, both contain the same number of genes encoding structural and assembly proteins. Furthermore, a large stem-loop structure (ΔG = −21.1) is located between the fasA major-subunit gene and the fasB chaperone gene (8). This is identical in organizational terms to the stem-loop structure located between the csfA major-subunit gene and the csfB major-subunit chaperone gene, which is predicted to act as an attenuator sequence to reduce expression of downstream genes (4).

A summary of the proposed model for CS5 assembly is shown in Fig. 10. All of the Csf proteins are translocated across the inner membrane via the Sec-dependent pathway. The major pilin subunit, CsfA, is then bound by the CsfB chaperone, and the minor pilin subunits, CsfD and CsfE, are bound by the CsfF chaperone. This serves to protect the pilin subunits from misfolding, premature aggregation, or degradation by periplasmic proteases. Pilus biogenesis is thought to be initiated by CsfD-CsfF complexes binding to the outer membrane assembly protein CsfC, which results in the translocation of CsfD across the outer membrane and leaves CsfC with an altered conformation, in an assembly-competent state. Pilus elongation occurs by means of multiple CsfA-CsfB interactions with CsfC in the outer membrane, driving the assembly of CsfA into the growing pilus, along with further CsfD-CsfF interactions. The rate of incorporation of CsfA versus CsfD is thought to depend on the stoichiometric ratio of the two pilins in the periplasm. CsfD is thought to add flexibility to the CS5 structure (Fig. 10, magnified region). Pilus termination occurs when CsfE-CsfF complexes are targeted to CsfC, which is predicted to result in the irreversible association of CsfE with CsfC, thereby preventing further polymerization of the pilus.

FIG. 10.

Model of CS5 pilus assembly. CS5 pilus biogenesis is hypothesized to be initiated by CsfD-CsfF complexes binding to the outer membrane assembly protein CsfC, resulting in the translocation of CsfD across the outer membrane. This is immediately followed by subsequent rounds of pilus elongation, in which CsfA subunits, delivered via a CsfA-CsfB complex to CsfC, are incorporated. Although the majority of the pilus consists of CsfA, further CsfD-CsfF complexes are also targeted to CsfC. The net result is a flexible CS5 pilus consisting predominantly of CsfA with a minor amount of CsfD. Pilus termination occurs when the rare CsfE-CsfF complex is targeted to CsfC, which is predicted to result in the irreversible incorporation of CsfE with CsfC, thereby preventing further rounds of pilus elongation. All of the Csf proteins have been assigned patterned structures as shown at the bottom. OM, outer membrane; IM, inner membrane; PERI, periplasm.