DsbA and DsbC Are Required for Secretion of Pertussis Toxin by Bordetella pertussis

Abstract

The Dsb family of enzymes catalyzes disulfide bond formation in the gram-negative periplasm, which is required for folding and assembly of many secreted proteins. Pertussis toxin is arguably the most complex toxin known: it is assembled from six subunits encoded by five genes (for subunits S1 to S5), with 11 intramolecular disulfide bonds. To examine the role of the Dsb enzymes in assembly and secretion of pertussis toxin, we identified and mutated the Bordetella pertussis dsbA, dsbB, and dsbC homologues. Mutations in dsbA or dsbB resulted in decreased levels of S1 (the A subunit) and S2 (a B-subunit protein), demonstrating that DsbA and DsbB are required for toxin assembly. Mutations in dsbC did not impair assembly of periplasmic toxin but resulted in decreased toxin secretion, suggesting a defect in the formation of the Ptl secretion complex.

Pertussis toxin is a major virulence factor of the gram-negative bacterium Bordetella pertussis, which is the causative agent of whooping cough (34, 43). The toxin is a member of the AB₅ family of toxins, which includes Shiga toxin, cholera toxin, and Escherichia coli heat-labile toxin. It plays a crucial role in virulence by mediating ADP-ribosylation of host GTP-binding proteins (G_i, G_o, and G_t), thereby disrupting normal host cellular regulation (19, 26). The systemic effects on the infected host include blocking of antimicrobial activity in a number of immune effector cells, resulting in a less effective immune response by the host (20, 42).

Five structural toxin genes (S1 to S5) and the nine ptl (for pertussis toxin liberation) genes (ptlA to ptlI), encoding the secretion complex, are cotranscribed from a single operon (28, 45) that is positively regulated by the Bvg two-component regulatory system (22). S1 is the enzymatic A subunit of the toxin, while the B subunit or B pentamer binds mammalian cells and delivers the toxin into the mammalian cytoplasm (25, 40). Pertussis toxin is a somewhat atypical AB₅ toxin. The B pentamer is not a homo-oligomer but rather consists of subunits S2, S3, S4, and S5, in a 1:1:2:1 ratio (30, 31, 39), which associate with the C terminus (2, 49) of the S1, or A, subunit. Each toxin subunit is translated with its own signal sequence (30, 31), and the subunits are secreted to the periplasm presumably via the Bordetella equivalent of the E. coli Sec machinery. Once in the periplasm, the subunits are assembled into the 105-kDa holotoxin that is recognized by the Ptl type IV secretion complex (10, 15, 35), which then moves the holotoxin through the outer membrane into the extracellular milieu (12, 14, 45). Additionally, it has been demonstrated that only the holotoxin is targeted for secretion, as expression of B subunit (S2 to S5) or A subunit (S1) in isolation did not result in extracellular secretion of toxin subunits (15). A region of the S1 subunit which might act as a recognition domain for this interaction of holotoxin with the Ptl secretion apparatus has been described (10).

Interestingly, AB₅ toxins have been described only for gram-negative bacteria, which suggests the periplasm may supply specialized functions required for assembly and secretion of these toxins. A candidate function is disulfide bond formation. Key to disulfide bond formation in the periplasm are the disulfide bond-forming enzymes, or Dsb proteins, of the thio:oxidoreductase family, which include DsbA, DsbB, DsbC, and others (3, 4). In E. coli, DsbA promotes disulfide bond formation in the periplasm (6) and the integral membrane protein DsbB regenerates oxidized DsbA (5, 13). Periplasmic DsbC promotes disulfide bond exchange (which is often needed for proteins that possess more than one disulfide bond) in previously oxidized proteins (37, 48). Dsb proteins have been shown to be essential for correct folding or assembly of a number of proteins and proteinaceous complexes, including enteropathogenic E. coli type IV pili (50), the E. coli flagellar apparatus (13), the Klebsiella oxytoca type II secreton (36), the E. coli PapD P-pilus chaperone (24), Pseudomonas aeruginosa lipase (LipA) (29, 41), E. coli heat-labile toxin (47), and cholera toxin (32, 47).

The crystal structure of pertussis toxin reveals one or more intrachain disulfide bonds that stabilize each of the subunits in the mature toxin (39). The S1 subunit of pertussis toxin contains a disulfide bond between cysteine-41 and cysteine-201 (Fig. 1). This cystine linkage stabilizes the subunit, and mutations that prevent the formation of this bond result in degradation of S1 (2, 49). In addition, mutations that prevent association of the S1 subunit with the B subunit result in instability of the S1 protein (2, 15, 35). The predicted sequences of the Ptl proteins also reveal a number of cysteine residues, some of which likely participate in intra- or extrachain disulfide bonds. Indeed, the PtlF-PtlI complex is stabilized by disulfide bonds in the outer membrane of B. pertussis (14).

FIG. 1.

Intramolecular disulfide bonds in pertussis toxin subunits. The positions of the disulfide bonds in the five pertussis toxin subunits are indicated by loops, with the numbers indicating the positions of the participating cysteine residues.

To further understand the role of disulfide bond formation in the assembly and secretion of pertussis toxin, we investigated B. pertussis strains mutant for members of the thiol:oxidoreductase enzyme family. Using the Sanger Centre preliminary genomic sequence of B. pertussis, we identified and mutated the B. pertussis dsbA, dsbB, and dsbC genes. Our results demonstrate that DsbA and DsbB are required for toxin assembly, and although DsbC is not required for the assembly of holotoxin, it is important for extracellular toxin secretion.

MATERIALS AND METHODS

Bacterial strains and plasmids.

The bacterial strains and plasmids used are listed in Table 1. Bordetella strains were grown on Bordet-Gengou agar (BGA) (Difco, Detroit, Mich.) containing 15% sheep's blood (Colorado Serum, Denver, Colo.). E. coli strains were grown on Luria-Bertani (LB) agar or in LB broth (Difco). When necessary, the following antibiotics at the indicated concentrations were added to the media: nalidixic acid, 30 μg/ml; gentamicin, 10 μg/ml (for maintenance of B. pertussis and E. coli strains) or 30 μg/ml (for selection of transconjugates); ampicillin, 100 μg/ml; kanamycin, 50 μg/ml; and streptomycin, 300 μg/ml. Plasmids were isolated using either the midiprep kit or the miniprep kit from Qiagen (Valencia, Calif.).

TABLE 1.

Strains and plasmids

Strain or Plasmid	Relevant characteristicsa	Source or reference
Strains
B. pertussis
BP338	Wild type; Tohama I background, Nalr	44
BPRA	Pertussis toxin deletion mutant; Nalr Strr	2
DsbA-1	BP338 dsbA::pTS8; Nalr Genr	This study
DsbB-1	BP338 dsbB::pTS24; Nalr Genr	This study
DsbC-1	BP338 dsbC::pTS2; Nalr Genr	This study
B. bronchiseptica
RB54	bvg mutant of RB50	9
RB54/DsbA	RB54 dsbA::pTS8; Genr	This study
RB54/DsbC	RB54 dsbC::pTS2; Genr	This study
E. coli K-12
One Shot	High-efficiency transformation	Invitrogen
DH5α	High-efficiency transformation; Nalr	Gibco BRL
MM294     (pRK2013)	Conjugation helper strain; carries Kmr mobilizing plasmid; IncP1 tra oriE1	17
Plasmids
pBluescript SK(+)	Cloning vector; Ampr	Stratagene
pCR2.1	TA cloning vector; Ampr Kanr	Invitrogen
pUW2138	Genr-oriT cassette in pBluescript SK(+)	17
pTS4	372-bp internal fragment of putative dsbA ORF in pCR2.1	This study
pTS8	pTS4 with Genr-oriT cassette in the HindIII site	This study
pTS19	263-bp internal fragment of putative dsbB ORF in pCR2.1	This study
pTS24	pTS19 with Genr-oriT cassette in the HindIII site	This study
pTS1	378-bp internal fragment of putative dsbC ORF in pCR2.1	This study
pTS3	pTS19 with Genr-oriT cassette in the HindIII site	This study

^aNal^r, naladixic acid resistance; Str^r, streptomycin resistance; Gen^r, gentamicin resistance; Amp^r, ampicillin resistance; Kan^r, kanamycin resistance.

Reagents.

Restriction enzymes and T4 DNA ligase were purchased from Invitrogen Life Technologies (Carlsbad, Calif.) or New England BioLabs (Beverly, Mass.) and used according to the manufacturer's recommendations. Shrimp alkaline phosphatase was purchased from U.S. Biochemical Corp. (Cleveland, Ohio) and used according to the manufacturer's recommendations. SeaKem and SeaPlaque (low-melting-point) agarose were obtained from BioWhittaker Molecular Applications (Rockland, Md.). Antibiotics and bacterial protease inhibitor cocktail were purchased from Sigma Chemical Co. (St. Louis, Mo.). Sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) reagents were obtained from Bio-Rad Laboratories (Hercules, Calif.). Tissue culture media, antibiotic supplements, and fetal bovine serum were acquired from Invitrogen Life Technologies. DNA and protein molecular weight markers were purchased from Invitrogen Life Technologies.

PCR cloning.

Primers were designed to amplify an internal fragment of the putative open reading frames (ORFs) of dsbA, dsbB, and dsbC: 5′-GTACGTGAACATCAACCCGCCGATG-3′ (dsbA forward) and 5′-GTCTGCACGCTGAACGAATCGAATAC-3′ (dsbA reverse), 5′-TTCCTGATCGCCATCCTGTGCTTTG-3′ (dsbB forward) and 5′-GTCTGGTCGCACGACAGCATCTTGG-3′ (dsbB reverse), and 5′-TGTTCGAGGTGCAGATCGGAAC-3′ (dsbC forward) and 5′-CCATCCAGTCTTTCCAGACCTTGGC-3′ (dsbC reverse). The sizes of the putative ORFs and the internal cloned fragments were 627 and 372 bp, 555 and 263 bp, and 879 and 378 bp for dsbA, dsbB, and dsbC, respectively. PCR was performed using the Advantage-GC cDNA PCR kit (Clontech Laboratories, Palo Alto, Calif.) according to the manufacturer's recommendations with B. pertussis strain BP338 DNA as the template. The PCR products were cloned into pCR2.1 using the TOPO TA cloning kit (Invitrogen) according to the manufacturer's recommendations. Cloning of the appropriate fragment was confirmed by restriction digestion, gel electrophoresis, and sequencing using T7 and M13 universal primers.

DNA sequencing.

DNA sequencing was performed by the Department of Molecular Genetics, Biochemistry, and Microbiology DNA Core Facility, University of Cincinnati, using an ABI 373 automated sequencer (PE Applied Biosystems, Foster City, Calif.)

Insertional duplication mutagenesis.

An insertional duplication mutagenesis strategy was developed to rapidly isolate mutants for each of the putative B. pertussis thiol:oxidoreductases. Internal gene fragments were introduced into B. pertussis or Bordetella bronchiseptica by using triparental matings to mobilize DNA as previously described (7). This was enabled by cloning a DNA cassette encoding gentamicin resistance and the origin of transfer (oriT) for the P-plasmid incompatibility group (from pUW2138 [Table 1]) into each plasmid harboring the cloned PCR products (17). Transconjugates were selected on BGA plates containing nalidixic acid and gentamicin. When nalidixic acid-sensitive Bordetella strains were used as the parental strains, colicin B was used as previously described (8) to counterselect against E. coli. Since these plasmids cannot replicate in Bordetella, gentamicin selection yields strains in which the plasmid has integrated into the chromosome via homologous recombination between the PCR-cloned gene fragments and their corresponding chromosomal loci (16, 45). The cloned sequences are thus duplicated in the genome but are interrupted by the vector. Previous studies have shown these mutations to be very stable (10, 12). Independent transconjugates were generated for each assay, and the presence of integrated plasmid in the chromosomes of the mutants was confirmed by PCR analysis at the end of each assay as previously described (10). The confirmatory primers were designed after determining the orientation of the cloned PCR products in the pCR2.1 vector. These primers flank the insertion: a T7 forward primer based on the pCR2.1 vector was used as the forward primer (5′-TTGTAATACGACTCACTATAGGGCGA-3′), and a reverse primer designed to be outside the original PCR amplification and downstream of the T7 primer was designed for each mutagenized locus. The reverse primers were 5′-CTGGAGTCGCCCGAGGATGTCATTG-3′, 5′-CGTGTTTCGACGGCGAGTACGTGAC-3′, and 5′-CCATCCAGTCTTTCCAGACCTTGGC-3′ for dsbA, dsbB, and dsbC insertions, respectively.

Five independent mutants were isolated from B. pertussis strain BP338 for each of the dsbA, dsbB, and dsbC loci. Three independent mutants were isolated from B. bronchiseptica strain RB54. Isolated mutants were not subcultured more than twice before each assay to minimize the likelihood of second-site compensatory mutations, and they were maintained and tested using media containing gentamicin.

Motility assay.

Strain RB54 and its derivatives were spotted on LB agar containing 0.5% agar. The size and morphology of the resulting colonies were observed at 24 and 48 h.

Secretion assay.

For secretion assays, B. pertussis was grown in a thin layer of Stainer-Scholte broth on nutrient-rich BGA plates (10-12). Growth on this biphasic medium is reproducible and vigorous due to the large surface/volume ratio of the culture and the rich nutrient base of the semisolid agar. To assay de novo production of pertussis toxin, strains were passed twice on BGA containing 40 mM MgSO₄ to modulate the bacteria and turn off transcription of the pertussis toxin operon (21), and bacteria from 24-h cultures were suspended in Stainer-Scholte broth to an optical density at 600 nm (OD₆₀₀) of 0.1. Six milliliters was plated on BGA plates containing appropriate antibiotics and incubated at 37°C for 30 h. The amounts of secreted and periplasmic pertussis toxin were determined as previously described (10-12). Briefly, the bacteria were harvested and pelleted by centrifugation, and the supernatants were filter sterilized for determination of secreted toxin. The cells were suspended to the original volume in phosphate-buffered saline (PBS), and the OD₆₀₀ was determined from an appropriate dilution. Periplasmic toxin was released from the cell suspensions by treatment with lysozyme and EDTA and filter sterilized for determination of intracellular pertussis toxin.

CHO cell assay.

The Chinese hamster ovary (CHO) cell assay was used to determine pertussis toxin activity as previously described (10, 23, 45). Pertussis toxin-treated CHO cells lose contact inhibition and clump together. The limit of detection for purified pertussis toxin (List Biological Laboratories, Campbell, Calif.) was approximately 1 to 2 ng/ml, and the last positive well for an unknown sample was assigned that value. Each sample was assayed in duplicate. The Student t test was used to analyze the data.

SDS-PAGE and immunoblotting.

SDS-PAGE and immunoblotting were performed as previously described (10, 46). B. pertussis cells were grown and harvested as for the secretion assay. Reducing SDS-PAGE was performed by resuspending the cells in PBS to the equivalent of an OD₆₀₀ of 8, adding an equal amount of loading buffer containing 4% β-mercaptoethanol, and boiling for 7 min. Nonreducing SDS-PAGE was performed by omitting β-mercaptoethanol from the loading buffer. In PtlF stability experiments, the indicated samples were resuspended in PBS containing bacterial protease inhibitor cocktail added according to the manufacturer's instructions prior to addition of loading buffer. Pertussis toxin subunits were detected by probing with monoclonal antibody C3X4 to S1 (27) or monoclonal antibody 11B7 to S2 (18). PtlF was detected by probing with polyclonal rat anti-PtlF antiserum. This antiserum was raised against a PtlF-maltose binding fusion protein containing residues 73 to 205 of the predicted PtlF protein sequence. Peroxidase-conjugated goat anti-mouse immunoglobulin G secondary antibody and peroxidase-conjugated goat anti-rat immunoglobulin G secondary antibody were purchased from Cappel (West Chester, Pa.). Antibody interactions were visualized by chemiluminescence using the Dupont Western blot Renaissance kit (NEN Research Products, Boston, Mass.) according to the manufacturer's recommendations. Apparent molecular weights were determined by comparison with prestained molecular weight markers and purified pertussis toxin.

RESULTS

Genomic analysis of B. pertussis Dsb proteins.

Key to proper disulfide bond formation in the gram-negative periplasm are the periplasmic disulfide bond-forming enzymes commonly known as the Dsb proteins. The preliminary genomic sequence of B. pertussis was searched for putative B. pertussis Dsb proteins by BLAST analysis. These sequence data were produced by the B. pertussis Sequencing Group at the Sanger Centre and were obtained from ftp://ftp.sanger.ac.uk/pub/pathogens/bp/BP.dbs. The resulting predicted protein sequences were inspected to ensure that each contained the catalytic CX₁X₂C thioredoxin motif (33). Single best-fit homologues were identified for DsbA, DsbB, and DsbC (Table 2). Insertional mutations were created in each of the putative ORFs.

TABLE 2.

 Characteristics of putative B. pertussis thiol:oxidoreductases

Protein	Amino acid sequencea		Size (aab)	Closest homologue (GenBank accession no.)	% Identityc
	N terminal	C terminal
DsbA	MQSTTFTRLL	VDQLIVQSRK	209	DsbA, Pseudomonas aeruginosa (P95460)	37
DsbB	MQPIAAPPAL	ALPAALRRNA	185	Dsb1, Pseudomonas aeruginosa (P21482)	42
DsbC	MPCRSSSGNS	LPRDELEASL	293	DsbC, Neisseria meningitidis (E81185)	36

^aPredicted sequences of the putative protein.

^baa, amino acids.

^cPercentage of identity between the BlastP-aligned sequence of the B. pertussis protein and its closest homologue.

Motility in B. bronchiseptica DsbA and DsbC mutants.

Motility is a complex phenotype, which has been shown to be sensitive to Dsb enzyme deficiency in E. coli (13). B. bronchiseptica is closely related to B. pertussis, but unlike B. pertussis it produces pertrichous flagella and is motile (1). To assess the effect of dsb null mutations on motility, we isolated null mutants for both dsbA and dsbC in B. bronchiseptica strain RB54. Null mutations in dsbB would be expected to give an phenotype identical to that of dsbA mutations, and dsbB was not included in these analyses. The parent and mutant strains were analyzed for motility by colony morphology after growth on soft agar (Fig. 2). The parental strain was motile, growing as a large diffuse colony. However, the dsbA and dsbC null mutants produced small colonies on this medium after 24 and 48 h (Fig. 2). The colony of one of three dsbA mutants spread but was smaller than wild type, suggesting that this isolate acquired a second-site mutation which compensated for the original dsbA deficiency. These results support the identity of the DsbA and DsbC homologues identified in this study.

FIG. 2.

Motility of DsbA- and DsbC-deficient B. bronchiseptica. RB54, parental strain; dsbA, RB54/DsbA; dsbC, RB54/DsbC.

Pertussis toxin secretion by Dsb mutants.

Assembled pertussis toxin contains 11 intramolecular disulfide bonds: one in the S1 subunit, three in subunits S2 and S3, and two in subunits S4 and S5 (Fig. 1). Pertussis toxin secretion in the dsb mutant strains was monitored by CHO cell assay of culture supernatants with parental strain BP338 as the wild-type positive control and the pertussis toxin deletion mutant BPRA as the negative control (Fig. 3). Mutations in dsbA, dsbB, and dsbC resulted in a 4- to 15-fold reduction in secretion of pertussis toxin (P values of 0.01, 0.04, and 0.01, respectively), indicating that these mutations caused a defect in assembly or secretion of pertussis toxin.

FIG. 3.

Toxin secretion by dsb mutants. The amount of secreted toxin after 30 h was determined by CHO cell assay. Error bars represent standard errors of the means. BP338, wild-type B. pertussis; BPRA, pertussis toxin deletion strain. ∗, P < 0.05 for mutant compared to wild type.

To determine whether the toxin deficits in the dsbA, dsbB, and dsbC mutants were secondary to growth defects, we measured growth rates. No significant difference was found between any of the strains versus the wild type at 30 h (P values of ≥0.43), and the growth curves were indistinguishable, indicating that the secretion deficits were not a result of an overall growth defect and that loss of Dsb function did not affect growth in the medium tested.

Pertussis toxin assembly by Dsb mutants.

Pertussis toxin assembles in the periplasm prior to secretion (10), and the decrease in secreted toxin could be due to a defect in assembly of the toxin and not a secretion defect per se. Cellular toxin was extracted from the secretion-deficient strains and the control strains, and the amount of active, assembled toxin was measured in the CHO cell assay (Fig. 4). Mutations in the dsbA and dsbB loci resulted in a marked (13- to 17-fold) reduction in the amount of periplasmic toxin compared to parental strain BP338 (P values of 0.02), indicating that these mutants were unable to assemble pertussis toxin. The DsbC mutants produced periplasmic toxin levels of 202 ± 58 ng/ml, which were not significantly different from those for the wild type (P = 0.78), indicating that their secretion defect was not due to lack of assembly of active toxin.

FIG. 4.

Toxin assembly by dsb mutants. The amount of periplasmic toxin after 30 h was determined by CHO cell assay. Error bars represent standard errors of the means. BP338, wild-type B. pertussis; BPRA, pertussis toxin deletion strain. ∗, P < 0.05 for mutant compared to wild type.

Immunoreactive S1 in insertional mutants.

Western blots were probed with a monoclonal antibody to the pertussis toxin A subunit (S1) after SDS-PAGE (Fig. 5). Wild-type BP338 produced full-length 28-kDa S1 subunit and a breakdown product, in accordance with previous observations (10, 35). The pertussis toxin deletion strain BPRA did not produce S1. Full-length S1 was seen in the DsbC mutant, as were the breakdown product and another slightly larger breakdown product not seen in the other strains. Markedly reduced amounts of full-length S1 were seen in the DsbA and DsbB mutants, suggesting that in the absence of these enzymes the S1 subunit is unstable, likely due to a failure to fold properly.

FIG. 5.

Western blot of cellular S1. BP338, wild-type B. pertussis; BPRA, pertussis toxin deletion strain; PT, 100 ng of purified pertussis toxin; MW, molecular weight markers (in thousands). S1, migration point of intact 28-kDa S1 subunit.

Immunoreactive S2 in Dsb mutants.

The effect that the Dsb mutations have on a B-subunit protein was investigated using Western blots probed with a monoclonal antibody to the pertussis toxin B-subunit protein (Fig. 6). The wild-type strain, BP338, produced S2 subunit that comigrated with a pertussis toxin control. As expected, mutant BPRA (with the pertussis toxin deleted) did not produce any S2. S2 was detected in the DsbC mutant, consistent with the ability of this mutant to produce wild-type levels of periplasmic pertussis toxin; however, the S2 subunit was not detected in the DsbA and DsbB mutants, suggesting that, like for S1, in the absence of these enzymes the S2 subunit is unstable. The S2 signal is present in the dsbC mutant but appears slightly fainter than in the wild type. While we cannot exclude the possibility that there is a slightly reduced amount of antigenic S2 in the dsbC mutant, the amounts of antigenic S1 are equivalent in the wild type and the dsbC mutant (Fig. 5), and the CHO cell assay (Fig. 4) demonstrates that there is no significant difference in the amount of active toxin. In addition, the assay is near the limit of detection for the antibody used.

FIG. 6.

Western blot of cellular S2. BP338, wild-type B. pertussis; BPRA, pertussis toxin deletion strain; PT, 25 ng of purified pertussis toxin; MW, molecular weight markers (in thousands).

Analysis of PtlF.

PtlF is known to be associated with PtlI in the outer membrane of B. pertussis in an interaction that is stabilized by disulfide bonding (14). The presence of PtlF was also analyzed using Western blotting. A high-molecular-weight band was determined to be nonspecific due to its presence in BPRA, a mutant lacking expression of pertussis toxin and the Ptl proteins (Fig. 7B; lane BPRA). In the absence of reducing agents, the PtlF-PtlI complex was detected in wild-type BP338 at approximately 35 kDa (Fig. 7A, lane WT-ox), in agreement with the predicted molecular mass of the complex and previous observations (14). When the samples were treated with reducing agents, PtlF was not detectable in the wild-type strain (Fig. 7A, lane WT-red), and the addition of bacterial protease inhibitors prior to reduction in sample buffer did not increase recovery of PtlF band (Fig. 7A, lane WT-PI). This concurs with previous observations (14) suggesting that PtlF is stable only when it is complexed with PtlI. The presence of PtlF was also analyzed in the dsb null mutants (Fig. 7B). The PtlF-PtlI complex was detected at similar levels in the wild-type strain and the dsb null mutants, demonstrating that the PtlF-PtlI disulfide bond was formed in strains lacking DsbA, DsbB, or DsbC.

FIG. 7.

Western blot of PtlF. (A) PtlF stability. WT-ox, wild-type B. pertussis BP338 boiled in sample buffer lacking reducing agents; WT-red, BP338 boiled in sample buffer with reducing agents; WT-PI, BP338 treated with protease inhibitors and boiled in sample buffer with reducing agents. (B) Analysis of PtlF under nonreducing conditions. BP338, wild-type B. pertussis; BPRA, pertussis toxin deletion strain; MW, molecular weight markers (in thousands). PtlF, migration point of the PtlI-PtlF complex; NS, nonspecific band.

DISCUSSION

The Dsb enzymes are required for folding of a number of proteins in gram-negative bacteria. DsbA catalyzes disulfide bond formation (6), while DsbC shuffles preformed disulfide bonds, an activity that is often needed for proteins that possess more than one disulfide bond (37, 48). In this study, we have generated null mutants for the Bordetella dsbA, dsbB, and dsbC homologues. Motility assays indicate that the dsbA and dsbC alleles that we have targeted are needed for proper protein assembly. The DsbB enzyme regenerates the active form of DsbA, and mutations in either allele have been shown to generate an identical phenotype (5, 13). We observed indistinguishable pertussis toxin phenotypes for the dsbA and dsbB mutants, and for simplicity, we will henceforth refer only to the dsbA mutant and DsbA. Of prime interest to us was the effect that these mutations would have on the assembly and secretion of the complex AB₅ toxin, pertussis toxin. Our results indicate that DsbA is needed for periplasmic pertussis toxin assembly, and while DsbC is not needed for toxin assembly, it is necessary for extracellular toxin secretion.

Instability of proteins that lack their proper disulfide bonds is well documented. For example, E. coli type IV pilin, the pullulanase secreton pilot protein PulS, and P. aeruginosa lipase all require correct disulfide formation for stability (29, 36, 50). The absence of the major periplasmic oxidant, DsbA, was shown to result in instability of both S1 and S2. Like S2, the other B-pentamer subunits have similarly positioned disulfide bonds along their lengths; thus, it is possible that they may also be misfolded and degraded in the absence of DsbA. S1 contains a disulfide bridge between C-41 and C-201 (2, 39, 49) that has been shown to stabilize the subunit, and the formation of this stabilizing disulfide bond may be DsbA dependent. Alternatively, the reduced amount of the S1 subunit in DsbA-deficient strains could be due to the lack of B pentamer, since S1 mutants lacking the regions of the C terminus that promote association with the B subunit have also been reported to be unstable (2, 15). We are unable to distinguish between these explanations from the present study.

PtlF of the pertussis toxin liberation machinery has also been shown to be stabilized by disulfide bonding, which is required for formation of the PtlF-PtlI complex in the outer membrane (14). We have confirmed that reduction of this disulfide bond results in instability of PtlF. Surprisingly, the intermolecular association of PtlF with PtlI in the outer membrane is not dependent on the presence of DsbA or DsbC. An enzymatic function other than DsbA or DsbC may be required; alternatively, it is possible that an effect was not seen with single mutations because the presence of one of these Dsb enzymes could compensate for the other in PtlF-PtlI disulfide formation.

DsbC has been reported to promote exchange between preformed disulfide bonds (37, 48). The absence of DsbC did not affect the assembly of pertussis toxin. Thus, DsbA is apparently sufficient to generate the multiple disulfide bonds in S2 and S3, and DsbC-catalyzed disulfide isomerization is not required. Similarly, it has been shown that alkaline phosphatase of E. coli (which contains multiple disulfide bonds necessary for its activity) can form some correct disulfides in the absence of DsbC, as activity in a dsbC null background is reduced to only about 10% of the wild-type level (37). Furthermore, it is important to note that the B pentamer disulfide bonds are formed between adjacent cysteines within the S2 to S5 peptides. It has previously been suggested that disulfide bonds may not require DsbC isomerization if they are formed correctly between adjacent cysteines of a peptide as it is translocated to the periplasm (38).

The lack of pertussis toxin secretion in the presence of functional pertussis toxin assembly in the dsbC mutants suggests a defect in the folding or assembly of the Ptl secretion machinery. We have ruled out failure to form the PtlF-PtlI complex, as this complex is formed in all of the dsb mutants we have examined. It is plausible that another Ptl protein, many of which contain multiple cysteines, is not correctly folded or assembled in the absence of DsbC, resulting in the secretion defect. Examination of the roles of other proteins in this process is presently limited by a lack of antibodies to the other pertussis toxin subunits or Ptl proteins.

The process of disulfide bond formation in the assembly of proteins in the gram-negative periplasm has been an area of intense research over the past decade. These bacteria possess a number of enzymes that are involved in the proper genesis of disulfides in periplasmic and extracellularly secreted proteins. Among the proteins whose disulfide catalysis is paramount for their physiological folding, assembly, and secretion are the AB₅ toxins. We have shown that the most complex of these toxins, pertussis toxin, also has an absolute requirement for Dsb disulfide catalysis in its biogenesis.