Regulation of type III secretion in Bordetella


  • Seema Mattoo,

    1. Department of Microbiology, Immunology and Molecular Genetics, David Geffen School of Medicine at UCLA, 10833 Le Conte Ave.
    Search for more papers by this author
  • Ming H. Yuk,

    1. Department of Microbiology, Immunology and Molecular Genetics, David Geffen School of Medicine at UCLA, 10833 Le Conte Ave.
    Search for more papers by this author
    • Present address: Department of Microbiology and Immunology, 201C Johnson Pavilion, University of Pennsylvania Medical Center, Philadelphia, PA 19104, USA.

  • Lisa L. Huang,

    1. Department of Microbiology, Immunology and Molecular Genetics, David Geffen School of Medicine at UCLA, 10833 Le Conte Ave.
    Search for more papers by this author
  • Jeff F. Miller

    Corresponding author
    1. Department of Microbiology, Immunology and Molecular Genetics, David Geffen School of Medicine at UCLA, 10833 Le Conte Ave.
    2. Molecular Biology Institute, University of California, Los Angeles, CA 90095-1747, USA.
    Search for more papers by this author


The BvgAS virulence control system regulates the expression of type III secretion genes in Bordetella subspecies that infect humans and other mammals. We have identified five open reading frames, btrS, btrU, btrX, btrW and btrV, that are activated by BvgAS and encode regulatory factors that control type III secretion at the levels of transcription, protein expression and secretion. The btrS gene product bears sequence similarity to ECF (extracytoplasmic function) sigma factors and is required for transcription of the bsc locus. btrU, btrW and btrV encode proteins predicted to contain PP2C-like Ser phosphatase, HPK (His protein kinase)-like Ser kinase and STAS anti-sigma factor antagonist domains, respectively, which are characteristic of Gram-positive partner switching proteins in Bacillus subtilis. BtrU and BtrW are required for secretion of proteins that are exported by the bsc type III secretion system, whereas BtrV is specifically required for protein synthesis and/or stability. Bordetella species have thus evolved a unique cascade to differentially regulate type III secretion that combines a canonical phosphorelay system with an ECF sigma factor and a set of proteins with domain signatures that define partner switchers, which were traditionally thought to function only in Gram-positive bacteria. The presence of multiple layers and mechanisms of regulation most likely reflects the need to integrate multiple signals in controlling type III secretion. The bsc and btr loci are nearly identical between broad-host-range and human-specific Bordetella. Comparative analysis of Bordetella subspecies revealed that, whereas bsc and btr loci were transcribed in all subspecies, only broad-host-range strains expressed a functional type III secretion system in vitro. The block in type III secretion is post-transcriptional in human-adapted strains, and signal recognition appears to be a point of divergence between subspecies.


Type III secretion is a highly regulated process that allows Gram-negative pathogens to deliver bacterial effector proteins into eukaryotic cells and alter host cell signalling functions (Hueck, 1998; Francis et al., 2002). Although most of the secretion apparatus components are well conserved among type III systems, the collection of effector proteins they deliver is often unique for different bacterial species. While significant advances have been made in understanding the functions of various type III effectors, the exact signals activating type III secretion and the regulatory mechanisms involved in responding to these signals during infection are largely uncharacterized. However, some common themes have emerged – two-component phosphorelay systems, AraC-like transcriptional activators and nucleoid-associated proteins are commonly involved in co-ordinating the expression of type III systems (Plano et al., 2001; Francis et al., 2002). Yersinia sp. undergo an additional level of control in which Yop transcription is coupled to secretion via a negative feedback regulatory loop (Pettersson et al., 1996). Another group of pathogens, which include Pseudomonas syringae, Erwinia amylovora and Ralstonia sp., controls type III secretion via alternative sigma factors that are, in turn, regulated by RpoN and response regulators that function as enhancer-binding proteins (Plano et al., 2001; Collmer et al., 2002).

A highly conserved type III secretion locus is shared by members of the Bordetella genus that cause respiratory infections in mammals (Yuk et al., 1998). Bordetella pertussis, the aetiological agent of whooping cough, and Bordetella parapertussis(Hu), which causes similar respiratory infections, are human-restricted variants of Bordetella bronchiseptica, which naturally infects a variety of four-legged mammals (Gerlach et al., 2001; Parkhill et al., 2003). The Bordetella type III secretion system has been most extensively studied in B. bronchiseptica (Yuk et al., 2000; Stockbauer et al., 2003), which appears to represent the evolutionary progenitor for most Bordetella subspecies (van der Zee et al., 1997; von Wintzingerode et al., 2002). In B. bronchiseptica, type III secretion is positively controlled at the transcriptional level by the products of the bvgAS locus, which encodes the master regulatory system that controls the infectious cycles of Bordetella subspecies (Yuk et al., 1998; Cotter and Jones, 2003). Under Bvg+ phase growth conditions, the membrane-bound sensor kinase, BvgS, phosphorylates the BvgA transcriptional activator via a His-Asp–His-Asp phosphorelay, and phosphorylated BvgA directs the expression of a large network of virulence and colonization factors (Uhl and Miller, 1996a,b; Mattoo et al., 2001).

The bsc locus includes 22 genes that encode components of the Bordetella type III secretion apparatus, secreted proteins and putative chaperones (Fig. 1A). Type III secreted polypeptides, which are easily detected in culture supernatants, are encoded by bopB, bopD, bopN and bsp22 (Yuk et al., 2000). BopB and BopD are orthologues of the Yersinia YopB and YopD translocation proteins (Cornelis, 2002). BopN is orthologous to YopN, which functions as a plug to prevent Yop secretion in Yersinia in the absence of host cell contact (Cheng et al., 2001). Bsp22, the most abundantly exported polypeptide, is unique. Although an in frame deletion in bsp22 has no effect on secretion of other polypeptides, it eliminates all known in vitro and in vivo effects of the Bordetella type III secretion system (Yuk et al., 2000). In vivo, type III secretion contributes to persistent tracheal colonization of rats and mice, and secreted proteins are highly immunogenic (Yuk et al., 2000). In vitro phenotypes associated with type III secretion include the induction of non-apoptotic cytotoxicity in epithelial and phagocytic cells, nuclear translocation of MAP kinases ERK1 and ERK2 and aberrant localization of NFκB (Yuk et al., 2000; Stockbauer et al., 2003).

Figure 1.

A. Genomic organization and open reading frames of type III secretion genes in B. bronchiseptica. The stretch of genes from bscF to bscC constitutes the bsc locus. Genes encoding secreted proteins are shown as grey block arrows. bscN encodes the putative ATPase. The genes of the btr type III regulatory locus, btrS, btrU, btrX, btrW and btrV, are represented as chequered arrows. The sequence data for these genes have been submitted to the GenBank database under accession numbers NP_888183, NP_888187, NC_002927, 3047274–3048016, NP_888190 and NP_888191 respectively.
B. Predicted protein structures of BtrS and HrpL from P. syringae. The protein domains are shown. BtrS contains an ECF sigma factor domain (E-value = 6.4e−16).
C. The B. subtilis RsbU–RsbV–RsbW partner-switching cycle that regulates the stress-responsive σB (see text).
D. Predicted protein structures of Bordetella BtrU, BtrW and BtrV and comparison with their Bacillus orthologues. Protein domains and conserved residues predicted to be important for protein function are shown. BtrU contains a HAMP domain (E-value = 6.4e−4) and a PP2C-like Ser phosphatase domain (E-value = 3.18e−42). BtrW contains a Ser kinase (S-KIN) domain (E-value = 4e−9), with conserved N-, D- and G-boxes for σ factor/antagonist binding, Mg2+ binding and ATP binding respectively. BtrV contains an anti-σ factor antagonist domain (STAS) (E-value = 4e−19).
E. Predicted protein structures of the Chlamydia partner switcher orthologues, CT588, CT549 and CT424. Like BtrU, CT588 contains a HAMP domain (E-value = 4.35e−5) and a PP2C-like Ser phosphatase domain (E-value = 3.09e−72). CT549 contains a Ser kinase domain (E-value = 8e−60). Chlamydia have two RsbV orthologues, CT424 and CT765, each with a STAS domain (E-values = 1.7e−28 and 1.2e−26 respectively).

DNA sequence analysis of the Bordetella genome in regions surrounding the bsc locus revealed five well-conserved open reading frames (ORFs), btrS, btrU, btrX, btrW and btrV (Fig. 1A). We report the characterization of these five genes, collectively referred to as the btr type III regulatory locus, and propose a novel cascade for the regulation of type III secretion in Bordetella.


Identification of btr genes

The btrS, btrU, btrX, btrW and btrV loci are located directly adjacent to the bsc locus (Fig. 1A). BtrS bears sequence similarity to the ECF (extracytoplasmic function) family of alternative sigma factors, which includes HrpL, an activator of type III secretion in the phytopathogen P. syringae, and σE, which activates heat shock and protein degradation responses in Escherichia coli (Fig. 1B) (Xiao and Hutcheson, 1994; Xiao et al., 1994; De Las Penas et al., 1997; Missiakas et al., 1997). ECF sigma factors typically regulate functions related to extracytoplasmic compartments, and their activity is often controlled by cognate membrane-bound anti-sigma factors (Missiakas and Raina, 1998; Campbell et al., 2003). BtrX is not orthologous to any known protein. BtrU, BtrW and BtrV have orthologues in partner-switching complexes that function to regulate σB activity in the stress response pathway and σF activity in the sporulation pathway of B. subtilis (Hughes and Mathee, 1998; Helmann, 1999). Partner-switching modules typically consist of a phosphatase, a protein kinase/anti-sigma factor and an antagonist protein, and represent a well-documented paradigm for signal transduction in Gram-positive bacteria (Fig. 1C and D). BtrU is orthologous to the RsbU and SpoIIE serine phosphatases and contains a PP2C-like serine phosphatase domain at its C-terminus (Voelker et al., 1995; Kang et al., 1996; Schroeter et al., 1999). BtrW bears sequence similarity to RsbW and SpoIIAB, which act as anti-sigma factors for σB and σF respectively (Benson and Haldenwang, 1993; Garsin et al., 1998). Like its orthologues, BtrW contains an HPK (histidine protein kinase)-like serine kinase domain. BtrV bears sequence similarity to RsbV and SpoIIAA, which act as antagonists of the RsbW and SpoIIAB anti-sigma factors (Dufour and Haldenwang, 1994; Garsin et al., 1998). RsbV, SpoIIAA and BtrV all contain anti-sigma factor antagonist (STAS) domains (Kovacs et al., 1998). In B. subtilis, the RsbW and SpoIIAB switch proteins are the primary regulators of σB and σF, and they function by binding and inhibiting their respective σ factors. The anti-σ factor activity of RsbW/SpoIIAB is countered by the RsbV/SpoIIAA antagonists, which, in their unphosphorylated state, can bind RsbW/SpoIIAB, thereby freeing their cognate sigma factors to interact with RNA polymerase and promote transcription. The phosphorylation states of RsbV/SpoIIAA are, in turn, controlled by opposing kinase (RsbW/SpoIIAB) and phosphatase (RsbU/SpoIIE) activities (Yang et al., 1996; Hughes and Mathee, 1998; Helmann, 1999). The RsbUVW partner-switching cycle is summarized in Fig. 1C.

For Bordetella, we refer to the genes encoding the BtrS, BtrU, BtrX, BtrW and BtrV regulatory factor orthologues as the btr (Bordetellatype III regulation) locus. The composition of this locus is unique, as it incorporates components of ECF family regulators that typically control extracytoplasmic events and proteins containing phosphorylation, kinase and phosphatase domains that are also present in partner switchers that regulate sigma factors in Gram-positive bacteria.

BvgAS regulation of btr genes

As type III secretion in Bordetella is a Bvg+ phase-specific phenomenon, we assessed the role of BvgAS in controlling the expression of btr genes. Transcript levels of btrS, btrU, btrX, btrW and btrV as well as several members of the bsc locus were estimated from cDNA isolated from Bvg+ (RB53, bvgS-C3) and Bvg (RB54, ΔbvgS) phase-locked derivatives of the wild-type strain RB50 (Fig. 2A). Transcript levels for fimD, a Bvg+ phase gene, and recA, which is Bvg independent, were measured as controls (Fig. 2B). As expected, primers specific for the type III apparatus-associated genes bsp22 and bopN produced polymerase chain reaction (PCR) products only from Bvg+ phase RNA (Fig. 2A, lanes 2 and 4). Transcripts for btrS, btrU, btrX and btrW were also detected only in Bvg+ phase samples. A transcript for btrV was detected in cDNA isolated from both Bvg+ and Bvg phase-locked strains; however, quantitative reverse transcription (RT)-PCR revealed that btrV transcription was 13-fold higher in the Bvg+ phase (Fig. 2C). Transcripts for btrS, btrU, btrX, btrW and btrV were also expressed in the type III secretion-defective strain WD3, which contains a deletion in bscN, the gene encoding the ATPase required for type III secretion (Fig. 2A, lane 6; Yuk et al., 1998). The data presented in Fig. 2 show that btrS, btrU, btrX, btrW and btrV are positively regulated by BvgAS. They also suggest that a block in secretion does not result in a block in transcription of type III genes, including those encoding secreted factors. The absence of negative feedback control was confirmed by comparing the expression of lacZ transcriptional fusions to the bsp22, bopN, btrU, btrW and btrX promoters in wild-type and ΔbscN strains (data not shown).

Figure 2.

A. RT-PCR analysis with primers specific for btrS, btrU, btrX, btrW, btrV, bsp22 and bopN. Genomic DNA (g) from wild-type B. bronchiseptica (RB50) and cDNAs (+) made from total RNA isolated from Bvg+ (RB53) and Bvg (RB54) phase-locked strains, the ΔbscN strain (WD3), each of the type III secretion system mutants, as well as the btrS ectopic strain (RB54.pbtrS) were used as templates for the RT reaction. Mock reactions (m), which did not contain reverse transcriptase, were used as controls against genomic contamination of the RNA preparations. Deletion mutations remove primer binding sites; therefore, the cognate transcript is not observed. Expression of btrS, btrU, btrX, btrW and the bsc locus (represented by bsp22 and bopN) is Bvg+ phase specific (compare lanes 2 and 4) and not subject to negative feedback in WD3 (lane 6), whereas btrV is constitutively expressed albeit at much higher levels in the Bvg+ phase (see Fig. 2C). Further, BtrS is necessary and sufficient for expression of btrU, btrX, btrW and the bsc locus (lane 18).
B. Control RT-PCR with primers specific for fimD, a Bvg+ phase-specific gene, and recA, a Bvg-independent, constitutively expressed gene. These primers also gave fimD and recA specific products with cDNA from WD3, confirming that expression of these genes was independent of type III secretion (lanes 5 and 11 respectively).
C. Quantitative RT-PCR analysis to determine btrV transcript levels. btrV transcription is 13-fold higher under Bvg+ versus Bvg phase conditions (compare RB53 and RB50 versus RB54). It is twofold lower in the ΔbtrS strain; however, ectopically expressed btrS does not significantly alter btrV transcription (compare RB54 versus RB54.pbtrS).

Deletion analysis of btr genes

Deletion mutations were introduced into btrS, btrU, btrX and btrW to investigate their roles in expression of the bsc locus as well as their effects on each other. A deletion in btrS eliminated the expression of all bsc genes, including bsp22, bopN, bopB, bopD, bscN, bscC, bcrD, bcrH1 and bcrH2 (Fig. 2A, lane 8; data not shown). The deletion in btrS also eliminated the expression of btrU, btrX and btrW. btrV transcript levels were twofold lower in the ΔbtrS strain compared with the wild-type parent (Fig. 2A, lane 8, and Fig. 2C). In contrast, deletions in btrU, btrX or btrW did not affect the expression of any btr or bsc locus tested (Fig. 2A, lanes 10, 12 and 14).

We were unable to introduce an in frame deletion into the btrV gene of wild-type strain RB50 grown under Bvg+ phase conditions (see Experimental procedures). In contrast, an identical deletion could be constructed in ΔbtrS or ΔbscN strains, or in a ΔbvgS (Bvg phase-locked) derivative of the wild-type strain. These observations suggested that deletion of btrV has a deleterious effect on cells expressing type III secretion. We next attempted to introduce the ΔbtrV mutation into RB50 by maintaining Bvg phase growth conditions during allelic exchange and then shifting to Bvg+ phase conditions for phenotypic analysis. This approach was successful (see Experimental procedures); however, the ΔbtrV mutant did not show altered expression of bsc or btr genes (Fig. 2A, lane 16).

The data shown in Fig. 2 indicate that BtrS is required for transcription of btrU, btrX, btrW and the bsc type III secretion locus. To determine whether BtrS is sufficient to activate the expression of type III secretion genes in the Bvg phase, we expressed btrS ectopically under the control of a constitutively expressed promoter in the ΔbvgS strain RB54. Ectopic expression of btrS allowed expression of all bsc locus genes tested including bsp22 and bopN as well as the type III regulatory genes btrU, btrX and btrW (Fig. 2A, lane 18; data not shown). As expected, ectopic expression of btrS had a relatively minor effect on btrV transcription (Fig. 2A, lane 18, and Fig. 2C). Thus, BtrS is necessary in the Bvg+ phase and sufficient in the Bvg phase to activate transcription of the bsc type III apparatus locus as well as btrU, btrX and btrW. BvgAS exerts control over type III secretion by directly or indirectly regulating BtrS transcription. The btrU, btrW and btrV gene products are apparently not required for transcription of type III secretion loci.

Export and expression of type III secretion substrates

As no phenotype was evident for deletions in btrU, btrX, btrW or btrV at the transcriptional level, we examined the roles of these loci in protein expression and secretion. Polypeptides isolated from supernatant and pellet fractions of wild-type bacteria, mutant derivatives and the btrS ectopic strain were immunoblotted for Bsp22 and BopD, two type III secreted proteins (Fig. 3A). Bsp22 and BopD were expressed and secreted by the Bvg+ phase-locked strain RB53 (Fig. 3A, lane 1) and wild-type strain RB50 (lane 2), whereas the ΔbscN strain failed to secrete them (Fig. 3A, lane 4). In agreement with RT-PCR data, Bsp22 and BopD were not expressed in ΔbvgS or ΔbtrS strains (Fig. 3A, lanes 3 and 5).

Figure 3.

A. Western immunoblot analysis of whole-cell pellet lysates (P) and culture supernatants (S) of wild-type B. bronchiseptica (RB50) and its various mutant derivatives grown under Bvg+ phase conditions. Polypeptides contained in pellet and supernatant fractions were separated by SDS-PAGE, transferred to PVDF membranes and probed with a 1:4000 dilution of a 1:1 mixture of polyclonal antibodies generated against type III secreted factors, Bsp22 and BopD (Yuk et al., 2000). The presence of a Bsp22-specific 22 kDa band or a BopD-specific 30 kDa band in the pellet fraction indicates that these proteins are expressed in the cell, while detection in supernatant fractions indicates secretion. Strains carrying pBBR1MCS, a replicating plasmid used as a vector for carrying type III regulatory genes for complementing deletions, are followed by ‘p’. Strains followed by ‘c’ represent complementation clones of the specific deletion. RBA1.pbtrS and RBA5.pbtrS are flagellar mutants carrying ectopically expressed btrS on pBBR1MCS. FF1 is an RBA5 derivative expressing the FHA adhesin.
B. Western immunoblot analysis of whole-cell lysates of wild-type B. bronchiseptica (RB50) and its various mutant derivatives grown under Bvg+ phase conditions. Polypeptides were separated by SDS-PAGE, transferred to PVDF membranes and probed with a 1:5000 dilution of serum collected from a rabbit infected with wild-type RB50, 5 weeks after inoculation. The bracket indicates a 60–90 kDa cluster of proteins that require both BtrS and active BvgAS for expression. The arrow marks Vag8.

The ΔbtrU and ΔbtrW mutants displayed a particularly interesting phenotype that appears to involve an uncoupling of protein expression from secretion. As observed with the ΔbscN mutant, the ΔbtrU and ΔbtrW strains expressed Bsp22 and BopD in nearly normal amounts but were unable to secrete them (Fig. 3A, lanes 8 and 12). The phenotype of the ΔbtrV mutation was even more striking, as it eliminated detection of Bsp22 and BopD in whole-cell pellet and culture supernatant fractions (Fig. 3A, lane 15). As bsp22, bopD and other type III secretion loci are transcribed at normal levels in ΔbtrV strains (Fig. 2A; data not shown), the lack of detectable protein could result from a defect in translation and/or a marked increase in the rate of degradation. The secretion and expression defects in the ΔbtrU, ΔbtrW or ΔbtrV mutant strains could be complemented by providing the corresponding wild-type alleles in trans (Fig. 3A, lanes 10, 14 and 17). Finally, Bsp22 and BopD were both expressed and secreted in a ΔbtrX strain, albeit at levels consistently higher than in the wild-type parent (Fig. 3A, lane 11).

Taken together, these data support the hypothesis that, although BtrS is required for expression of type III loci, BtrU and BtrW are required for secretion and BtrV is essential for translation and/or stability. To our knowledge, this is the first example of a functionally active set of partner switcher orthologues in a Gram-negative bacterium. Furthermore, they regulate type III secretion.

Additional BvgAS-dependent factors required for type III secretion

Ectopic expression of btrS in the Bvg phase-locked strain RB54 allowed Bsp22 and BopD expression, but not secretion (Fig. 3A, lanes 18 and 19). btrU and btrW expression is activated under these conditions, whereas btrV is expressed at a basal level. Lack of secretion of Bsp22 and BopD is most probably caused by a requirement for BtrS-independent Bvg+ phase-specific factors for the secretion process, or inhibition by one or more Bvg phase gene products. Flagellar biosynthetic systems are the evolutionary progenitors of type III secretion, and cross-talk between type III and flagellar systems has been suggested for Salmonella and Yersinia species (Schmitt et al., 1996; Young et al., 1999; Eichelberg and Galan, 2000; Bleves et al., 2002). In B. bronchiseptica, motility genes are specifically expressed in the Bvg phase. Ectopic expression of flagella in the Bvg+ phase is deleterious to infection (Akerley et al., 1995), and prevents secretion of Bsp22 and BopD (M. H. Yuk and J. F. Miller, unpublished). We therefore tested whether ectopic expression of btrS in a Bvg phase-locked strain devoid of flagella (RBA5), or the entire flagellar pathway as a result of deletion of the flagellar master regulatory locus frlAB (RBA1), would allow type III secretion. Type III secretion could not be activated by ectopically expressing btrS even in the absence of the flagellar pathway (Fig. 3A, lanes 22 and 23), supporting the hypothesis that additional Bvg+ phase factors are required for the secretion process.

To investigate this idea further, pellet fractions from wild-type or mutant strains were probed with serum from a B. bronchiseptica-infected rabbit (Fig. 3B). A cluster of polypeptides migrating between 60 and 90 kDa that were expressed in Bvg+ phase bacteria was missing in ΔbvgS (RB54) or ΔbtrS mutants (Fig. 3B, lanes 1–5). Interestingly, although complementation with btrS restored expression in the ΔbtrS strain under Bvg+ phase conditions (Fig. 3B, lane 7), the identical plasmid did not restore expression in RB54 (Fig. 3B, lane 9). Expression of these proteins therefore appears to require both BtrS and BvgAS. We have identified one of these polypeptides as a Bordetella autotransporter, Vag8 (Finn and Amsbaugh, 1998). In keeping with the results in Fig. 3B, vag8 transcription requires both BvgAS and BtrS. Furthermore, a deletion in vag8 eliminates export but not expression of type III secretion substrates (A. K. Foreman-Wykert et al., in preparation).

In vitro cytotoxicity of btr mutants

The most prominent phenotype associated with type III secretion in vitro is the induction of a caspase-independent cell death pathway that has an unusually broad cell type specificity (Yuk et al., 2000; Stockbauer et al., 2003). We tested the ability of type III secretion regulatory mutants to induce cytotoxicity in cultured HeLa cells using lactate dehydrogenase (LDH) release assays (Fig. 4A). As expected, B. bronchiseptica killed HeLa cells in a Bvg+ phase-specific, type III secretion-dependent manner. Infection with wild-type RB50 resulted in extensive cell death, whereas the Bvg phase-locked strain RB54 (ΔbvgS) and the type III secretion-defective strain WD3 (ΔbscN) had relatively little effect. Likewise, the ΔbtrS, ΔbtrU, ΔbtrW and ΔbtrV mutants were also defective in inducing cytotoxicity. Addition of wild-type alleles in trans resulted in partial (btrS, btrW) or near-total (btrU, btrV) levels of complementation. As expected, deletion of btrX did not diminish cytotoxicity. Although ΔbscN, ΔbtrS, ΔbtrU, ΔbtrW or ΔbtrV mutants adhere to HeLa cells as well as wild-type strains (data not shown), RB54 is defective in attachment. In contrast, FF1 is a Bvg phase-locked strain that ectopically expresses the FHA adhesin and efficiently adheres to cultured cells (Cotter et al., 1998). Expression of btrS in FF1 did not result in detectable cytotoxicity (Fig. 4), a result consistent with the inability of this strain to secrete Bsp22 or BopD (Fig. 3A, lane 25).

Figure 4.

Cytotoxicity of Bordetella subspecies on HeLa cells, as measured by LDH release. HeLa cells were infected with bacteria at an MOI of 40:1 for 3 h. Results are representative of three independent experiments.
A. Cytotoxicity induced by wild-type B. bronchiseptica (RB50) and its various type III secretion mutant derivatives.
B. Comparative analysis of cytotoxicity induced by various Bordetella subspecies. Compared with RB50, B. pertussis strains (GMT1, Tohama I, Bp536 and Bp18323) and human-adapted B. parapertussis strains (A168 and 12822) do not kill HeLa cells, whereas the ovine B. parapertussis strain FR107 does. All bacteria were grown under Bvg+ phase conditions except when modulated to the Bvg phase, indicated by (–).

In all cases, an inability to export or express type III secretion substrates detected by immunoblotting (Fig. 3A) correlates with a functional defect in type III secretion, as measured by the induction of cytotoxicity (Fig. 4A). Contact with eukaryotic cells therefore does not over-ride the requirement for BtrS, BtrU, BtrW or BtrV in the expression of type III secretion.

Type III secretion in human-adapted Bordetella

We conducted a comparative phylogenetic analysis of type III secretion in B. bronchiseptica, B. pertussis as well as human and ovine strains of B. parapertussis. In vitro, human-adapted Bordetella strains (GMT1, Tohama I, Bp536, Bp370, 18323 and 12822) failed to confer significant levels of cytotoxicity to mammalian cells (Fig. 4B). Immunoblot analysis of pellet and supernatant fractions revealed that B. bronchiseptica and ovine B. parapertussis strains expressed and secreted type III proteins (Fig. 5A, lanes 1, 7 and 8). In contrast, none of the human-adapted Bordetella strains tested displayed Bsp22 or BopD expression in vitro (Fig. 5A, lanes 2–6, 9 and 10). We hypothesized that the lack of type III protein expression could result from the lack of expression of bsc and/or btr genes. Surprisingly, RT-PCR analysis of cDNA isolated from each of these strains grown under Bvg+ phase conditions showed that bsc-encoded genes, including bsp22, bopN, bcrH1, bcrH2, bscN, bcrD and all the btr genes, were transcribed (Fig. 5B, lanes 1–15; data not shown). Furthermore, in the Tohama I derivative Bp536, expression was Bvg+ phase specific (Fig. 5B, lanes 6–9).

Figure 5.

Comparative phylogenetic analysis of type III secretion in Bordetella subspecies.
A. Western immunoblot analysis of whole-cell pellet (P) and supernatant (S) fractions from B. bronchiseptica Bvg+ (RB53) and Bvg (RB54) phase-locked mutants, B. pertussis strains (Bp536, Bp370, GMT1, Tohama I and 18323), human B. parapertussis strains (A168 and 12822) and ovine B. parapertussis strains (FR107 and H1). Polypeptides contained in pellet and supernatant fractions were separated by SDS-PAGE, transferred to PVDF membranes and probed with a 1:4000 dilution of a 1:1 mixture of polyclonal antibodies generated against type III secreted factors, Bsp22 (22 kDa) and BopD (30 kDa).
B. RT-PCR analysis of expression of various bsc and btr type III regulatory genes. Genomic DNA (g) from wild-type B. bronchiseptica (RB50) and cDNAs (+) made from total RNA isolated from B. pertussis strains (Tohama I, Bp536 and 18323), ovine B. parapertussis FR107 and human B. parapertussis 12822 were used as templates for PCR. Mock reactions (m), which did not contain reverse transcriptase in the RT reaction, were used as controls against genomic contamination of the RNA preparations. All bacteria were grown under Bvg+ phase conditions except when modulated to the Bvg phase, indicated by (–).

The observation that bsc and btr loci are expressed and regulated in B. pertussis is intriguing considering the high degree of sequence conservation with B. bronchiseptica. The bsc loci shown in Fig. 1 are intact in B. pertussis and are predicted to encode polypeptides with 93–100% amino acid sequence identity to those expressed by B. bronchiseptica. BtrS, BtrU, BtrX, BtrW and BtrV display 100%, 99%, 98%, 100% and 98% identity respectively. The vast majority of nucleotide substitutions observed between B. bronchiseptica and B. pertussis type III secretion loci are silent or result in conservative amino acid substitutions. These observations suggest that selective pressures exist to maintain functional type III secretion genes in B. pertussis.


We propose a unique cascade through which multiple regulatory factors are integrated to control type III secretion in Bordetella(Fig. 6). The master virulence regulatory locus, BvgAS, sits at the top of the hierarchy and activates btrS expression, either directly or indirectly. BtrS then activates the expression of genes encoding components of the type III secretion apparatus, secreted factors such as Bsp22, BopD and BopB and loci encoding additional regulatory factors. Expression of BtrS alone is sufficient for expression of genes encoded by the bsc locus in the absence of BvgAS. Stable synthesis and secretion of type III proteins, however, requires an array of proteins with domains that define partner switchers in Gram-positive bacteria: BtrV exerts post-transcriptional control required for translation and/or protein stability, whereas BtrU and BtrW specifically govern the secretion process. Additional factors subject to both BvgAS and BtrS control, such as the autotransporter Vag8, are also required for secretion. Secretion control could be exerted at the level of apparatus assembly and/or delivery of substrates to the type III secretion complex.

Figure 6.

Model for type III secretion regulation in Bordetella (see text).

Contrary to our expectations, there is no evidence suggesting that BtrU, BtrW or BtrV regulates BtrS activity. Therefore, as is true for other ECF sigma factors, BtrS function may be controlled by a cognate anti-sigma factor that remains to be identified. Similarly, by analogy with the Rsb system in B. subtilis, the BtrU, W and V proteins could function to regulate one or more additional sigma factors required for the expression of loci governing secretion. The sequence of BtrU includes two predicted transmembrane motifs separated by a hydrophilic stretch of 106 amino acids, followed by a HAMP domain (Fig. 1D). HAMP domains are typically involved in transmitting signals between input and output modules in signalling proteins (Aravind and Ponting, 1999). We hypothesize that BtrU could function as a membrane-bound sensor and transmit signals directly to the partner-switching complex. In an alternative model, BtrU, BtrW and/or BtrV could form part of the actual type III secretion apparatus.

Although btrV showed elevated transcription in the Bvg+ phase, it was the only btr locus with appreciable expression in the Bvg phase. Furthermore, the ability to isolate null mutations in btrV was greatly facilitated by mutations that eliminate type III secretion (ΔbtrS, ΔbtrU, ΔbscN or ΔbvgS). We were able to circumvent the apparently deleterious effects of deleting btrV in wild-type Bvg+ phase bacteria by modulating cultures to the Bvg phase during mutant construction, followed by a shift back to Bvg+ conditions. Our results demonstrate that BtrV is required for translation and/or stable expression of type III secreted polypeptides. The role of BtrV in the type III regulatory hierarchy is intriguing and reminiscent of data from other systems demonstrating that secretion and translation of exported proteins can be coupled (Karlinsey et al., 2000; Cambronne and Schneewind, 2002; Trcek et al., 2002).

We propose that BtrU, BtrW and BtrV are likely to operate in basic accordance with the predicted biochemical activities of their conserved PP2C-like serine phosphatase, HPK-like serine kinase and STAS domains. The phenotypes of ΔbtrU, ΔbtrW and ΔbtrV mutants, however, indicate significant differences in comparison with partner-switching models based on the Rsb and SpoII systems as summarized in Fig. 1 (Helmann, 1999). For example, on the basis of the B. subtilis paradigms, a deletion in btrW would be expected to have the opposite phenotype to a deletion in btrU. According to our analysis, the phenotypes are the same (Fig. 3A). Although numerous possibilities exist, one alternative mechanism consistent with our results predicts that non-phosphorylated BtrV binds BtrW, and the resulting BtrV–BtrW complex is required for type III secretion. Preliminary results using a bacterial two-hybrid system show that BtrW and BtrV interact, and in vitro assays with purified proteins indicate that BtrW phosphorylates BtrV (N. Kozak, S. Mattoo and J. F. Miller, in preparation). The mechanistic basis for the unique phenotypes associated with these partner switcher orthologues, however, remains to be determined. Phylogenetic trees for BtrU, BtrW and BtrV orthologues are shown in Fig. 7. In each case, Bordetella and Bacillus orthologues belong to distinct clades, indicating that they evolved along divergent branches of evolution. As such, it is not surprising that the orthologues of partner-switching proteins in these species may have evolved to perform different functions. Further biochemical and genetic studies of BtrU, BtrW and BtrV are likely to expand our appreciation of the range of functions that these related proteins can perform.

Figure 7.

Phylogenetic tree derived from BtrU (A), BtrW (B) and BtrV (C) based on alignment of whole sequences. The root for each of the tree diagrams is indicated as a bold line. Bordetella, Bacillus and Chlamydia species are indicated by arrows. In each case, Bordetella and Chlamydia species evolved along divergent clades when compared with Bacillus sp. and are more closely related to each other than to Bacillus. Bordetella subspecies are the only β-proteobacteria predicted to encode proteins orthologous to partner switchers.

Although the bsc and btr loci are highly conserved and fully transcribed in B. bronchiseptica, B. pertussis and B. parapertussis, human-adapted strains do not readily display type III secretion-associated phenotypes in vitro. Interestingly, the block in type III secretion observed with B. pertussis in vitro is apparently post-transcriptional, resembling the phenotype of a ΔbtrV mutant in B. bronchiseptica. In comparing B. pertussis with B. bronchiseptica, BtrW is identical whereas BtrV and BtrU differ at two and seven positions respectively. Of the nucleotide differences observed in bsc or btr loci, substitutions are dramatically skewed towards those that are silent or conservative. This implies evolutionary pressure for maintaining type III secretion in B. pertussis. It therefore seems possible that B. pertussis does express type III secretion, but in a manner that is regulated differently from that observed in B. bronchiseptica. Signal recognition, mediated by unidentified factors, is a possible source of divergence between subspecies. If this hypothesis is correct, it will dramatically change the way we view B. pertussis–host interactions.

The incorporation of what had previously been considered to be a Gram-positive system of gene regulation to control type III secretion is intriguing, and Bordetella species may not be the only example. Homologues of BtrU, BtrW and BtrV have been identified in Chlamydia species that also display type III secretion (Kalman et al., 1999; Fields et al., 2003). Interestingly, the Chlamydia BtrU homologue is more closely related to Bordetella BtrU than to its Bacillus counterparts (Fig. 1E). Its predicted protein sequence suggests that it also contains two transmembrane domains separated by a hydrophilic stretch of 274 amino acids followed by a HAMP domain. The phylogenetic analyses in Fig. 7 indicate that Chlamydia partner-switching orthologues diverged early from the root, along evolutionary clades closer to Bordetella than to Bacillus. Understanding the regulation of type III secretion in Bordetella may have implications for understanding how this process is regulated in Chlamydia species for which tractable genetic systems have not yet been developed. The Btr regulatory cascade may therefore represent a new class of mechanisms for controlling type III secretion in Gram-negative bacterial pathogens.

Experimental procedures

Bacterial strains and growth conditions

Wild-type B. bronchiseptica strain RB50 was isolated in our laboratory from a naturally infected rabbit (Cotter and Miller, 1994). RB53 is an RB50 derivative that contains a point mutation in bvgS corresponding to the B. pertussis bvgS-C3 allele, which confers a Bvg+ phase constitutive phenotype (Cotter et al., 1998). RB54 is an RB50 derivative with an in frame deletion in bvgS that confers a Bvg phase constitutive phenotype (Akerley et al., 1992). RBA5 and RBA1 are RB54 derivatives with deletions in the flagellar flaA or frlAB genes respectively (Akerley et al., 1992). FF1 is an RBA5 derivative that ectopically expresses filamentous haemagglutinin (FHA), a Bvg+ phase-specific adhesin that is sufficient for attachment of Bordetella to mammalian epithelial cells in vitro (Cotter et al., 1998).

Tohama I is a well-studied B. pertussis clinical isolate that was sequenced recently (Parkhill et al., 2003). Bp536 is a streptomycin- and nalidixic acid-resistant derivative of Tohama I (Stibitz and Yang, 1997). Bp370, 18323 and GMT1 are B. pertussis strains described previously (Kendrick et al., 1947; Martinez de Tejada et al., 1996; Stibitz and Yang, 1997). 12822 and A168 are human-adapted B. parapertussis clinical isolates from children with pertussis-like illnesses (Heininger et al., 2002). H1 and FR107 are B. parapertussis strains isolated from sheep in Scotland and New Zealand respectively (Heininger et al., 2002).

Bordetella bronchiseptica strains were grown in Stainer–Scholte medium or on Bordet–Gengou (BG) agar (Becton Dickinson Microbiology Systems) containing 7.5% defibrinated sheep blood (Mission Laboratories). B. pertussis and B. parapertussis strains were grown in Stainer–Scholte medium containing 0.1% heptakis or on BG agar containing 15% sheep blood. Antibiotics were used at the following concentrations: streptomycin, 20 µg ml−1; kanamycin, 40 µg ml−1; chloramphenicol, 50 µg ml−1 (5 µg ml−1 for B. pertussis); and ampicillin, 100 µg ml−1.

PCR, RT-PCR and quantitative RT-PCR

Total RNA isolated from mid-log bacterial cultures was reverse transcribed into cDNA using 2 µg of RNA, 200 ng of random hexamers and Superscript II (Gibco). PCR was performed according to the manufacturer's instructions (1 U of Master Taq polymerase, 200 µM each of the four dNTPs and 1 µM each primer) and supplemented with 5% dimethyl sulphoxide (DMSO). Primer pairs amplifying internal fragments were used to analyse gene expression. Cycling parameters were: one cycle of 94°C for 4 min; 30 cycles of 95°C for 1 min, 55°C, 57°C or 60°C for 1 min and 72°C for 1 min; and a final incubation at 72°C for 5 min.

Quantitative RT-PCRs were conducted using the iCycler iQ Real-Time PCR detection system from Bio-Rad. The system uses a 96-well thermal cycler with attached optics to measure fluorescence in each well and proprietary software to interpret the resulting signal. The iQ SYBR Green Supermix kit from Bio-Rad, which comes optimized for real-time PCR applications, was used to provide the intercalating SYBR Green fluorescent dye for detection of double-stranded DNA products resulting from each extension step of the PCR. The PCR was carried out according to the manufacturer's instructions using 10% of the cDNA obtained from various mid-log bacterial cultures (preparation described above) as template, 300 nM each primer and supplemented with 5% DMSO. The primers were designed according to the manufacturer's guidelines to generate amplicons of 100–150 bp. PCR conditions were 3 min at 95°C, followed by 40 cycles of 45 s at 60°C. Melt-curve analysis was performed immediately after the amplification protocol under the following conditions: 1 min denaturation at 95°C, 1 min annealing at 55°C, 80 cycles of 0.5°C increments (10 s each) beginning at 55°C (data collection step). The critical threshold (the first cycle in which fluorescence is determined to be above background by statistical significance) was determined, normalized to the signal output for the constitutively expressed gene recA for each bacterial cDNA sample and inversely correlated with the starting concentration of target DNA. All real-time PCRs were performed in duplicate in three independent experiments.

Construction of deletion mutants

All mutations were delivered to the chromosome by allelic exchange using the pRE112 allelic exchange system (Edwards et al., 1998). Deletions were constructed in btrS, btrU, btrX and btrW, but a deletion in btrV alone in the Bvg+ phase was lethal. The allelic exchange vector used to mobilize the ΔbtrV mutation into wild-type, ΔbtrU, ΔbtrS, ΔbscN and RB54 B. bronchiseptica strains yielded co-integrants from single recombination events occurring at either the 5′ or the 3′ end of btrV. However, a second homologous recombination event yielding ΔbtrV mutants was observed only in ΔbtrU, ΔbtrS, ΔbscN and RB54 backgrounds. A second homologous recombination event in co-integrants with a wild-type background always reconstituted the wild-type allele. Thus, we used an indirect approach to obtain a ΔbtrV mutation in wild-type Bvg+ phase bacteria: by modulating wild-type B. bronchiseptica to the Bvg phase with 20 mM nicotinic acid or growth at 25°C and maintaining Bvg phase conditions throughout the experiment, we used the allelic exchange strategy described above to obtain a ΔbtrV mutation in a wild-type background. Four such ΔbtrV mutants were obtained. These mutants were then switched to Bvg+ phase growth conditions, and the survivors were analysed. The results of assays conducted with one of the surviving clones, ΔbtrV-G8, have been reported here.

The deletion of btrS extends from 80 bp upstream of the btrS translational start site to 15 bp downstream of the stop codon, thereby deleting the entire gene. Deletions in btrU and btrX extend from codons 132 to 600 and 33 to 170 respectively. Deletions in btrW and btrV are in frame and extend from codons 12 to 388 for btrW and from 16 to 143 for btrV (where the predicted BtrV translational start corresponds to nucleotide 1748413 of the B. bronchiseptica genome sequence). All the above deletions, except ΔbtrS, leave intact the putative transcriptional and translational start sites, as well as the translational stop site. Allelic exchange plasmids containing deletion fragments were mobilized into RB50 via triparental mating using DH5α containing plasmid pRK2013, which provides mobilization functions in trans (Fee et al., 1980). Primer pairs annealing to sequences ≈ 100–250 bp upstream and downstream of the respective 5′ and 3′ ends of the gene of interest were used to analyse deletions by PCR.

To construct complementing plasmids for the deletion strains, the minimal ORFs of btrS, btrU, btrX, btrW and btrV plus an additional 40 bp upstream of the respective start codons were amplified from wild-type genomic templates by PCR and ligated downstream of the lacZ promoter in pBBR1MCS, a pBBR1CM derivative that replicates autonomously in B. bronchiseptica (Kovach et al., 1994). The same strategy was used to construct a Bvg phase-locked B. bronchiseptica strain that ectopically expresses btrS (a Bvg+ phase-specific gene) except that the btrS ATG site remained in frame with the lacZ ATG. Transcription of btrS under Bvg phase conditions and expression of type III proteins in the Bvg phase confirmed ectopic expression of btrS.

Immunoblot analysis

Proteins were extracted from pellet and supernatant fractions of wild-type and mutant bacteria grown overnight under Bvg+ phase conditions. SDS-PAGE was performed using 4–12% gradient gels (acrylamide–bisacrylamide, 29:1). OD equivalents of B. bronchiseptica whole-cell lysates and supernatants were transferred to Immobilon P polyvinylidene difluoride (PVDF) membranes (Millipore) and incubated with either polyclonal antibody generated against Bsp22 and BopD or serum collected from an RB50-infected rabbit, 5 weeks after inoculation. Anti-Bsp22 and anti-BopD antibodies were used at a dilution of 1:4000. Anti-B. bronchiseptica rabbit serum was used at a dilution of 1:5000. Antigen–antibody complexes were detected with horseradish peroxidase (HRP)-conjugated anti-mouse or anti-rabbit immunoglobulin (Amersham) at a 1:5000 dilution and visualized by an enhanced chemiluminescence technique (Amersham).

Cytotoxicity assays

All LDH release assays were performed in 24-well plates with cells at ≈ 70% confluency. HeLa human epithelial cells (ATCC, CCL-2) were cultured in EMEM (Gibco) supplemented with 5% fetal bovine serum (FBS) and 10 mg ml−1 penicillin–streptomycin solution. All cytotoxicity assays were conducted in the corresponding antibiotic-free media supplemented with 5% FBS. Bordetella strains grown to log phase were centrifuged on to monolayers of HeLa cells (600 g, 5 min) at a multiplicity of infection (MOI) of 40:1 and incubated for 3 h at 37°C. LDH is a stable eukaryotic cytoplasmic enzyme that is released upon cell lysis; the amount of LDH released thus correlates with the level of cytotoxicity. The LDH release was measured with the CytoTox 96 kit (Promega) according to the manufacturer's protocol with the exception that 50 µl of sample was transferred to a 96-well plate for the assay. The LDH release (% cytotoxicity) was calculated using the equation: [(OD480 sample–OD480 cells alone)/(OD480 total LDH control–OD480 cells alone)] × 100.

Phylogenetic analysis

Orthologues of BtrU, BtrW and BtrV were obtained using protein blast ( Orthologues of BtrU, BtrW and BtrV with E-values ≤ 2e−14, 0.66 and 8e−04, respectively, were chosen for building phylogenetic trees using clustalx ( and the phylip maximum likelihood method (


We thank Natalia A. Kozak for compiling sequence information for Fig. 1, Amy K. Foreman-Wykert for sharing unpublished results relating to Vag8, and Ekaterina M Panina for technical assistance with phylogenetic analyses. S.M. was supported by NIH Tumor Immunology Training Grant T32-CA009120 and the UCLA Dissertation Year Fellowship. This work was supported by NIH grant RO1-A138417 and USDA grant 02298 to J.F.M.