Interactions between Brucella suis VirB8 and Its Homolog TraJ from the Plasmid pSB102 Underline the Dynamic Nature of Type IV Secretion Systems

Abstract

The proteinVirB8 plays a critical role in the assembly and function of the Agrobacterium tumefaciens virB type IV secretion system (T4SS). The structure of the periplasmic domain of both A. tumefaciens and Brucella suis VirB8 has been determined, and site-directed mutagenesis has revealed amino acids involved in the dimerization of VirB8 and interactions with VirB4 and VirB10. We have shown previously that TraJ, the VirB8 homologue from pSB102, and the chimeric protein TraJB8, encompassing the cytoplasmic and transmembrane (TM) domains of TraJ and the periplasmic domain of VirB8, were unable to complement a B. suis mutant containing an in-frame deletion of the virB8 gene. This suggested that the presence of the TraJ cytoplasmic and TM domains could block VirB8 dimerization or assembly in the inner membrane. By bacterial two-hybrid analysis, we found that VirB8, TraJ, and the chimeras can all interact to form both homo- and heterodimers. However, the presence of the TM domain of TraJ resulted in much stronger interactions in both the homo- and heterodimers. We expressed the wild-type and chimeric proteins in wild-type B. suis. The presence of proteins carrying the TM domain of TraJ had a dominant negative effect, leading to complete loss of virulence. This suggests that the T4SS is a dynamic structure and that strong interactions block the spatial flexibility required for correct assembly and function.

Brucellosis is a major worldwide zoonosis primarily affecting developing countries and causing them severe economic losses (7). Bacteria of the genus Brucella, the causative agent, are gram-negative facultative intracellular pathogens of various wild and domestic mammals, as well as humans, where it causes a very debilitating disease known as Malta fever (48). In addition, these bacteria are also a focus of concern as possible biological warfare agents (23).

The key aspect of Brucella virulence is its ability to survive and proliferate within professional and nonprofessional phagocytes (7). Once phagocytosed, this bacteria subverts the vesicular traffic in the host cell to establish a niche in a compartment derived from the endoplasmic reticulum, where it multiplies (3, 34, 39, 40, 41). Several factors have been reported to be essential for the virulence of this bacterium (16, 17, 18, 19, 21, 22, 29, 31). Strikingly, we (36) and others (44) have demonstrated the presence in Brucella of a type IV secretion system (T4SS) that is encoded by the virB operon and whose integrity is required for virulence (8, 14, 19). Several other species of gram-negative bacteria have been found to rely on the presence of a T4SS for full virulence (10, 13). Both extracellular (Helicobacter) and intracellular (Legionella, Bartonella) pathogens use their T4SSs to inject effector proteins directly into the target cell, where they affect the biology of the cell. Bordetella pertussis uses its T4SS to secrete the pertussis toxin into the extracellular medium, where it is taken up by cells. The T4SS of Agrobacterium tumefaciens translocates both effector proteins and a nucleoprotein complex into target plant cells through a mechanism reminiscent of bacterial conjugation through T4SS.

The A. tumefaciens VirB T4SS, which is considered the T4SS paradigm, is composed of 11 different proteins named VirB1 to VirB11 plus VirD4. These proteins can be functionally subdivided in three different groups. The proteins VirB4, VirB11, and VirD4 are inner membrane ATPases with a large cytoplasmic domain and are believed to provide the energy required for T4SS assembly and for the translocation of effectors. VirB2 and VirB5 form an extracellular bacterial appendage believed to anchor the bacteria to the host cell (4, 25). Finally, VirB3 and VirB6 to VirB10 are believed to form a channel-like structure spanning both the inner and outer membranes of the bacteria.

Among these structural proteins, VirB8 has been shown to play a key role in the assembly of the T4SS. Recent studies with A. tumefaciens demonstrated that VirB8 acts as a nucleation center required to recruit VirB9 and VirB10 into clusters in the outer membrane (15, 30) and to localize VirB proteins at the cell pole (26). VirB8 is a protein spanning the bacterial inner membrane, with the first 67 amino acids forming a short cytoplasmic tail, followed by a single hydrophobic transmembrane (TM) domain. The carboxy-terminal moiety of the protein, of 172 amino acids, is believed to be entirely periplasmic. Recently, the three-dimensional structures of the periplasmic domains of VirB8 from Brucella suis (46) and A. tumefaciens (5) have been determined. Using these structural data, site-directed mutagenesis has been performed on the periplasmic part of B. suis VirB8, showing that changes in amino acids that inhibit the dimerization of VirB8 or its interactions with VirB4 or VirB10 also affect T4SS assembly and B. suis virulence (37). Among all of the homologs of B. suis VirB8, the closest are the proteins TraJ, encoded in the tra operons of broad-host-range plasmids pSB102 and pIPO2 (43, 45). The TraJ protein from pSB102 shares more than 50% identity with B. suis VirB8 at the amino acid level, and this percentage increases to more than 60% when only the periplasmic domain is considered. In a previous study, we have taken advantage of this close similarity between VirB8 and TraJ to examine the possibility of a functional heterologous complementation of VirB8 by TraJ in BS1008, a B. suis mutant carrying an in-frame deletion of the virB8 gene (38). From our results, it appeared that the protein TraJ was unable to complement BS1008. As the major similarities between VirB8 and TraJ were found in their respective periplasmic domains, we constructed chimeric genes encoding proteins in which the major part of the periplasmic domain of one protein (amino acids 77 to 241 of TraJ and amino acids 76 to 239 of VirB8) was replaced with the corresponding part of the other protein (these proteins are described in Fig. 1) and studied the ability of these chimeric proteins to restore the virulence of BS1008. The TraJB8 chimera, where the periplasmic part of VirB8 replaces the corresponding part of TraJ, was also unable to complement BS1008. In contrast, B8TraJ, the reverse chimera in which the periplasmic part of TraJ replaces the corresponding part of VirB8, partially restored the virulence of BS1008. These results show that, when fused to the cytoplasmic and TM parts of VirB8, the periplasmic part of TraJ can functionally replace the corresponding part of VirB8 in T4SS assembly. In contrast, the cytoplasmic and TM parts of TraJ cannot replace the corresponding part of VirB8, whether fused to the periplasmic part of TraJ or VirB8. To further elucidate these points, we undertook a more detailed study of the interactions of these proteins by bacterial two-hybrid (BACTH) analysis. Here we show that the proteins VirB8, B8TraJ, TraJB8, and TraJ display strikingly different abilities to interact with themselves, as well as with VirB8 itself. Further, the TM domain of these proteins plays a crucial role in determining the strength of these interactions. Finally we show that, when overexpressed in a wild-type B. suis strain, these various proteins are able to modulate its virulence, even leading to complete loss of virulence. These data give interesting clues concerning the mechanisms of type IV secretion.

FIG. 1.

Schematic representation of the proteins used in this study. Light and dark gray parts represent protein domains from VirB8 and TraJ, respectively. All proteins were synthesized as fusion proteins with the T18 or T25 subunit of the B. pertussis adenylate cyclase domain fused to the amino-terminal end of the VirB8- and/or TraJ-containing part. In the case of the periplasmic domains of VirB8 (VirB8p) and TraJ (TraJp), synthesis of these proteins as fusions with the subunits of B. pertussis adenylate cyclase likely targets them to the bacterial cytoplasm, as suggested by fractionation studies. The rightmost column indicates whether the corresponding proteins have (+) or have not (−) been detected in Western blotting experiments.

MATERIALS AND METHODS

Bacterial and cell culture.

All cloning and plasmid amplifications were done with the Escherichia coli DH5α or XL₁-Blue strain by using standard cloning protocols. Transformants were selected on 2YT agar containing the appropriate antibiotic (ampicillin, 100 μg/ml; kanamycin, 25 μg/ml; chloramphenicol, 30 μg/ml), and cultures were grown in 2YT medium containing the same antibiotic. BACTH experiments were performed with the adenylate cyclase-deficient strain E. coli BTH101 as described previously (27). All experiments included a positive control (plasmids encoding fusion proteins with a leucine zipper from Saccharomyces cerevisiae GCN4) and a negative control (empty plasmids). Wild-type strain B. suis 1330 was transformed by electroporation as previously described (38). All of the resulting Brucella strains were used to infect J774 murine macrophage-like cells in a standard gentamicin protection assay as described previously (38).

Plasmids.

The PCR primers used in this study are described in Table 1, and the plasmids used are described in Table 2. Plasmid pIN73 was derived from pUT18C (27) by replacing a PstI-EcoRI fragment containing the polylinker with the corresponding fragment of pKT25. Plasmid pIN94 was obtained from pKT25 by filling in the unique XbaI site of the polylinker with the Klenow fragment of E. coli DNA polymerase. All of the other plasmids used in the BACTH studies were derived from pIN73 and pIN94 by inserting DNA fragments obtained by PCR between the unique BamHI and EcoRI sites. The construction of some derivatives of pBBR1-MCS used in infection studies has already been described (38). Other plasmids were designed specifically for this study from plasmids used for BACTH analysis.

TABLE 1.

Primers used in this study

Oligonucleotide	Sequencea
18CF	5′-GTTCGAAGTTCTCGCCGGA-3′
18CR	5′-GATTGTACTGAGAGTGCACC-3′
KT25F	5′-GCGATTCGGTGACCGATTAC-3′
KT25R	5′-GTTGTAAAACGACGGCCAGT-3′
B8pF	5′-GCGGATCCCGCGTCAACGCACAG-3′
B8F	5′-GCGGATCCATGTTTGGACGCAAACAATC-3
B8R	5′-CCGAATTCATTGCACCACTCCC-3′
TraJpF	5′-GCGGATCCCGCGTGGATAACACC-3′
TraJF	5′-GCGGATCCATGTTCAGACGCAAAAAGG-3′
TraJR	5′-CCGAATTCATTGTCCACCCCC-3′
B8CTM	5′-CGGAATTCATTGCTTTAGCGGTAACATCC-3′
B8CTMBgl	5′-GCAGATCTATTGCTTTAGCGGTAACATCC-3
B8CTMFLAG	5′-CGTCTAGAGACTACAAGGACGACGATGACAAGGGATCCATGTTTGGAC-3′
B8CTMHis	3′-CGTCTAGACACCACCACCACCACCATATGGGATCCATGTTTGGAC-3
B8TM	5′-GCGGATCCATCGCCGGGGCTTTCG-3′
TraJCTM	5′-CGGAATTCATTCCTTGAGCGGCATCAT-3′
TraJCTMBgl	5′-GCAGATCTATTCCTTGAGCGGCATCAT-3′
TraJTM	5′-GCGGATCCGTGGCCTATGGAGCCGT-3′
B8CyJTM	5′-GGCTCCATAGGCCACTTTCCACGCGCGGCG-3′
JcyB8TM	5′-GAAAGCCCCGGCGATGTACCACGCCCGTTG-3′
NdeFLAG	5′-GCCATATGGACTACAAGGACGACGAT-3′
KTFLAG	5′-CGTCTAGAGGACTACAAGGACGACGATGACAAGGGATCCATGTTTGGACG-3′
KTHis	5′-CGTCTAGAGCACCACCACCACCACCATATGGGATCCATGTTTGGACG-3′

^aBoldface sequences are restriction sites used for cloning.

TABLE 2.

Plasmids used in this study

Plasmid	Description	Source
pUT18C	BACTH plasmid conferring resistance to ampicillin and encoding the T18 subunit of B. pertussis adenylate cyclase	26
pKT25	BACTH plasmid conferring resistance to kanamycin and encoding the T25 subunit of B. pertussis adenylate cyclase	26
pIN34	pBBR1 with a 1.1-kb BamHI-NdeI PCR-amplified fragment containing the B. suis virB promoter and the NdeI-ClaI fragment of pFLAG-CTC	37
pIN38	pIN34 with a 700-bp NdeI-BglII PCR-amplified fragment encompassing B. suis virB8 with a silent mutation at Val76 creating a unique Mlu1 restriction site and a STOP codon upstream of the FLAG sequence; expresses VirB8 in Brucella	37
pIN54	pIN34 with a 700-bp NdeI-BglII PCR-amplified fragment encompassing pSB102 traJ with a silent mutation at Ile77 creating a unique Mlu1 restriction site and a STOP codon upstream of the FLAG sequence; expresses TraJ in Brucella	37
pIN56	pIN54 with the Mlu1-BglII fragment exchanged with the corresponding fragment from pIN38; expresses TraJ-VirB8 chimera in Brucella	37
pIN57	pIN38 with the Mlu1-BglII fragment exchanged with the corresponding fragment from pIN54; expresses VirB8-TraJ chimera in Brucella	37
pIN109	pIN34 with an NdeI-BglII PCR-amplified fragment coding for the cytoplasmic and TM domains of VirB8 from pIN82	This study
pIN110	pIN34 with an NdeI-BglII PCR-amplified fragment coding for the cytoplasmic and TM domains of TraJ from pIN83	This study
pIN111	pIN34 with an NdeI-BglII PCR-amplified fragment coding for the FLAG epitope, followed by the cytoplasmic and TM domains of VirB8 from pIN86	This study
pIN112	pIN34 with an NdeI-BglII PCR-amplified fragment coding for the FLAG epitope, followed by the cytoplasmic and TM domains of TraJ from pIN88	This study
pIN73	pUT18C plasmid with the PstI-EcoRI fragment encompassing the multiple cloning site replaced with the corresponding fragment from pKT25	This study
pIN76	pIN73 with a BamHI-EcoRI PCR fragment encoding the periplasmic domain of VirB8 (VirB8p)	This study
pIN77	pIN73 with a BamHI-EcoRI PCR fragment encoding the periplasmic domain of TraJ (TraJp)	This study
pIN78	pIN73 with a BamHI-EcoRI PCR fragment encoding the whole VirB8 protein	This study
pIN79	pIN73 with a BamHI-EcoRI PCR fragment encoding the whole TraJ protein	This study
pIN80	pIN73 with a BamHI-EcoRI PCR fragment encoding the cytoplasmic and TM domains of TraJ, followed by the periplasmic domain of VirB8 (TraJB8)	This study
pIN81	pIN73 with a BamHI-EcoRI PCR fragment encoding the cytoplasmic and TM domains of VirB8, followed by the periplasmic domain of TraJ (B8TraJ)	This study
pIN82	pIN73 with a BamHI-EcoRI PCR fragment encoding the cytoplasmic and TM parts of VirB8 (B8CyTM)	This study
pIN83	pIN73 with a BamHI-EcoRI PCR fragment encoding the cytoplasmic and TM parts of TraJ (JCyTM)	This study
pIN84	pIN73 with a BamHI-EcoRI PCR fragment encoding the cytoplasmic domain of TraJ, followed by the TM domain of VirB8 (JCyB8TM)	This study
pIN85	pIN73 with a BamHI-EcoRI PCR fragment encoding the cytoplasmic domain of VirB8, followed by the TM domain of TraJ (B8CyJTM)	This study
pIN86	pIN73 with an XbaI-EcoRI PCR fragment encoding the FLAG epitope, followed by the cytoplasmic and TM parts of VirB8 (FLAGB8CyTM)	This study
pIN87	pIN73 with an XbaI-EcoRI PCR fragment encoding the six-His epitope, followed by the cytoplasmic and TM parts of VirB8 (HisB8CyTM)	This study
pIN88	pIN73 with an XbaI-EcoRI PCR fragment encoding the FLAG epitope, followed by the cytoplasmic and TM parts of TraJ (FLAGJCyTM)	This study
pIN89	pIN73 with an XbaI-EcoRI PCR fragment encoding the six-His epitope, followed by the cytoplasmic and TM parts of TraJ (HisJCyTM)	This study
pIN90	pIN73 with a BamHI-EcoRI PCR fragment encoding the TM domain of VirB8	This study
pIN91	pIN73 with a BamHI-EcoRI PCR fragment encoding the TM domain of TraJ	This study
pIN92	pIN73 with a BamHI-EcoRI PCR fragment encoding the cytoplasmic domain of VirBB8	This study
pIN93	pIN73 with a BamHI-EcoRI PCR fragment encoding the cytoplasmic domain of TraJ	This study
pIN94	pKT25 plasmid with the unique XbaI restriction site filled in with the Klenow DNA polymerase	This study
pIN95	pIN94 with a BamHI-EcoRI PCR fragment encoding the periplasmic domain of VirB8	This study
pIN96	pIN94 with a BamHI-EcoRI PCR fragment encoding the periplasmic domain of TraJ	This study
pIN97	pIN94 with a BamHI-EcoRI PCR fragment encoding the whole VirB8 protein	This study
pIN98	pIN94 with a BamHI-EcoRI PCR fragment encoding the whole TraJ protein	This study
pIN99	pIN94 with a BamHI-EcoRI PCR fragment encoding the cytoplasmic and TM domains of TraJ, followed by the periplasmic domain of VirB8 (TraJB8)	This study
PIN100	pIN94 with a BamHI-EcoRI PCR fragment encoding the cytoplasmic and TM domains of VirB8, followed by the periplasmic domain of TraJ (B8TraJ)	This study
pIN101	pIN94 with a BamHI-EcoRI PCR fragment encoding the cytoplasmic and TM domains of VirB8 (B8CyTM)	This study
pIN102	pIN94 with a BamHI-EcoRI PCR fragment encoding the cytoplasmic and TM domains of TraJ (JCyTM)	This study
pIN103	pIN94 with a BamHI-EcoRI PCR fragment encoding the cytoplasmic domain of TraJ, followed by the TM domain of VirB8 (JCyB8TM)	This study
pIN104	pIN94 with a BamHI-EcoRI PCR fragment encoding the cytoplasmic domain of VirB8, followed by the TM domain of TraJ (B8CyJTM)	This study
pIN105	pIN94 with a BamHI-EcoRI PCR fragment encoding the TM domain of VirB8	This study
pIN106	pIN94 with a BamHI-EcoRI PCR fragment encoding the TM domain of TraJ	This study
pIN107	pIN94 with a BamHI-EcoRI PCR fragment encoding the cytoplasmic domain of VirB8	This study
pIN108	pIN94 with a BamHI-EcoRI PCR fragment encoding the cytoplasmic domain of TraJ	This study
pIN117	PIN73 with the HindIII-BamHI fragment from pIN86 encompassing the T18 peptide and the FLAG epitope replacing the corresponding fragment of pIN73	This study
pIN118	pIN73 with the HindIII-BamHI fragment from pIN87 encompassing the T18 peptide and the six-His epitope replacing the corresponding fragment of pIN73	This study
pIN119	pIN117 with a BamHI-EcoRI fragment encoding the full-length VirB8 protein	This study
pIN120	pIN117 with a BamHI-EcoRI fragment encoding the full-length TraJ protein	This study
pIN121	pIN117 with a BamHI-EcoRI fragment encoding the BCy8JTM chimera	This study
pIN122	pIN117 with a BamHI-EcoRI fragment encoding the JCyB8TM chimera	This study
pIN123	pIN117 with a BamHI-EcoRI fragment encoding the periplasmic domain of VirB8	This study
pIN124	pIN117 with a BamHI-EcoRI fragment encoding the periplasmic domain of TraJ	This study
pIN125	pIN117 with a BamHI-EcoRI fragment encoding the cytoplasmic domain of VirB8	This study
pIN126	pIN117 with a BamHI-EcoRI fragment encoding the cytoplasmic domain of TraJ	This study
pIN127	pIN1117 with a BamHI-EcoRI fragment encoding the TM domain of VirB8	This study
pIN128	pIN1117 with a BamHI-EcoRI fragment encoding the TM domain of TraJ	This study
pIN133	pIN117 with a BamHI-EcoRI fragment encoding the B8 TraJ chimera	This study
pIN134	pIN117 with a BamHI-EcoRI fragment encoding the TraJ B8 chimera	This study

β-Galactosidase assay.

Colonies of transformed BTH101 (generally three for each combination of plasmids) were inoculated into 1 ml of LB medium containing ampicillin, kanamycin, and isopropyl-β-d-thiogalactopyranoside (IPTG). After overnight culture at 30°C or 37°C, β-galactosidase activity was measured as described previously (33).

Western blotting.

To study the stability of the proteins used for BACTH analysis, we transformed BTH101 simultaneously with the pIN94-derived plasmid expressing the protein of interest fused to the T25 subunit of B. pertussis adenylate cyclase and the pIN-derived plasmid expressing the same protein fused to the T18 subunit, followed by a FLAG epitope. One colony of each transformation was cultured overnight at 37°C in 1 ml LB medium containing ampicillin, kanamycin, and IPTG. Bacteria were pelleted (4,000 × g, 5 min, 4°C). In some experiments, a fraction corresponding to the bacterial periplasm was first isolated from the bacterial pellet by osmotic shock in 5 mM MgSO₄ (35). Otherwise, the pellets were resuspended in 100 μl of 50 mM Tris-HCl buffer (pH 8) containing 5 mM EDTA, a cocktail of protease inhibitors (leupeptin and antipain both at a 1-μg/ml final concentration and phenylmethylsulfonyl fluoride at a 100-μg/ml final concentration), and lysozyme at a final concentration of 0.25 mg/ml until the cells were lysed. When the suspension became viscous, 10 μl of a solution containing 0.02 mg/ml DNase I, 1.5 M NaCl, 0.1 M CaCl₂, and 0.1 M MgCl₂ was added. When the suspension was no longer viscous, it was centrifuged (25,000 × g, 30 min, 4°C). Aliquots of the supernatant and the resuspended pellet were diluted with an equal volume of gel loading buffer, boiled, and loaded onto a 12% polyacrylamide gel. The resolved proteins were transferred to a polyvinylidene difluoride membrane (Immobilon PVDF; Millipore). The membranes were then assayed for the presence of proteins containing the FLAG epitope by using a monoclonal antibody directed against this epitope and coupled to horseradish peroxidase (HRP; Sigma). HRP activity was detected with a chemiluminescent substrate (Immobilon Western HRP substrate; Millipore).

Growth kinetics of Brucella strains.

Wild-type B. suis strain 1330 and the strains used in the macrophage infection study were grown overnight at 37°C in 2YT medium. The resulting cultures were diluted in 2YT to obtain an optical density at 600 nm (OD₆₀₀) of 0.1. The diluted cultures were further grown at 37°C, and the OD₆₀₀s were measured every 2 h for 8 h. The doubling time (T) of each strain was calculated from the slope (m) of the straight line obtained by plotting the ln of the OD₆₀₀ as a function of the incubation time by using the formula T = ln 2/m.

RESULTS

Expression of VirB8, TraJ, and different chimeras in the BACTH system.

The different proteins were cloned in phase with the T18 or T25 subunit of B. pertussis adenylate cyclase and transferred to E. coli BTH101 (the proteins studied in this work are described in Fig. 1). The expression levels of the different proteins with a FLAG epitope tag were determined by Western blotting (Fig. 2). The fusion proteins containing VirB8p or TraJp, the full-length proteins, or the chimera B8TraJ or TraJB8 were all expressed at comparable levels (Fig. 2A). Crude fractionation experiments showed that the fusion proteins containing VirB8p or TraJp were detected mainly in the cytoplasm, while fusion proteins containing full-length VirB8 or TraJ or the full-length chimeras were found exclusively in insoluble membrane-enriched fractions (not shown). The fusion proteins with TraJCyTM or B8CyJTM, which both contain the TM domain of TraJ, were detected in large amounts and exclusively in membrane fractions (Fig. 2B and data not shown); in contrast, the fusion proteins with B8CyTM or JCyB8TM, which both contain the TM domain of VirB8, were not detected at all. Finally, the expression of the fusion proteins containing either the TM domain or the cytoplasmic domain of VirB8 or TraJ alone could not be detected (Fig. 2B). Therefore, when expressed, the fusion proteins were detected in the correct cellular compartment; however, protein instability can explain the apparent lack of interaction for certain fusion protein interactions described below.

FIG. 2.

Expression and stability of various FLAG-tagged proteins expressed in BTH101 bacteria. Bacterial extracts were obtained as described in Materials and Methods. After sodium dodecyl sulfate-polyacrylamide gel electrophoresis of crude bacterial extracts, Western blotting revealed FLAG-tagged proteins with a specific monoclonal antibody against the FLAG epitope. The fusion proteins synthesized by the bacteria in panel A are B8p (lane 1), TraJp (lane 2), VirB8 (lane 3), TraJ (lane 4), B8TraJ (lane 5), and TraJB8 (lane 6), and those in panel B are TraJCyTM (lane 1), B8CyTM (lane 2), JCyB8TM (lane 3), B8CyJTM (lane 4), B8Cy (lane 5), TraJCy (lane 6), B8TM (lane 7), and TraJTM (lane 8).

Homodimeric interactions between periplasmic domains.

Dimerization has been shown to be important for VirB8 function (37). As a first step in understanding why the VirB8-TraJ chimeras had different capacities to restore virulence to the virB8 mutant BS1008, we studied the formation of homodimers by BACTH analysis with β-galactosidase activity as an indicator of the strengths of the different interactions. We first studied the periplasmic domains of VirB8 and TraJ, as this domain of VirB8 is assumed to be responsible for the interactions of this protein with itself and with other proteins of the B. suis VirB complex. Using the standard BACTH analysis protocol with bacterial culture at 30°C (27, 28), we found that the strengths of the homodimeric interactions of VirB8p or TraJp were not significantly different (Fig. 3 left panel). Both gave close to 75% of the β-galactosidase activity measured for the positive control (labeled ZIP). Our laboratory has recently shown (1) that culture at 37°C, rather than 30°C, potentiates the differences in β-galactosidase activity which reflect the strength of interaction of the proteins of interest. At 37°C, the activity in the positive and negative controls remained unchanged compared to the activity measured following culture at 30°C (Fig. 3, compare left and right panels). In contrast, under the same conditions the homodimeric interaction of VirB8p or TraJp gave rise to an activity decreased to 25 to 50% of the positive control value (Fig. 3, right panel). As 37°C is the temperature encountered by Brucella in infected animals and in vitro virulence assays, we tested all further interactions at this temperature. For the data on interaction at 30°C, see the supplemental material.

FIG. 3.

Homodimeric and heterodimeric interactions between the periplasmic domains of VirB8 and TraJ. The first protein was synthesized from high-copy-number plasmid pIN73, and the second was synthesized from low-copy-number plasmid pIN94. Following overnight culture at 30°C (left panel) or 37°C (right panel), β-galactosidase activities were measured as described in Materials and Methods. Results are the mean ± the standard error of the mean of at least three different cultures.

Full-length proteins and chimeras form homodimers.

We then studied the homodimeric interactions of the VirB8 and TraJ full-length proteins and the B8TraJ and TraJB8 chimeras. Homodimeric interactions of VirB8 and the chimeric protein B8TraJ both produced the same β-galactosidase activity. In contrast, both the homodimeric interactions of TraJ and the TraJB8 chimera reproducibly produced β-galactosidase activity 5 to 10 times higher (Fig. 4 left panel), suggesting a much stronger interaction.

FIG. 4.

Homodimeric interactions between full-length proteins (left panel) or cytoplasmic and TM domains (right panel). Following overnight culture at 37°C, β-galactosidase activities were measured as described in Materials and Methods. Results are the mean ± the standard error of the mean of at least three different cultures.

To assess the contribution of the cytoplasmic and TM domains of these proteins to the homodimeric interactions, we cloned the corresponding DNA fragments into pIN73 and pIN94 (Table 2). We found that the cytoplasmic and TM domains of TraJ alone were sufficient to produce strong homodimeric interactions for TraJCyTM (Fig. 4, right panel). We were not able to detect the expression of B8CyTM (Fig. 1), and no β-galactosidase activity could be measured (Fig. 4, right panel). To further dissect the roles of the different parts of these proteins, we used chimeras composed either of the cytoplasmic part of VirB8 and the TM part of TraJ (B8CyJTM) or, conversely, of the cytoplasmic part of TraJ and the TM part of VirB8 (JCyB8TM). B8CyJTM formed homodimers interacting as strongly as homodimers of TraJCyTM. No β-galactosidase activity could be measured for the other proteins, the expression of which could not be detected by Western blotting (Fig. 1).

Heterodimeric interactions with VirB8.

Sequence comparisons suggest that the periplasmic domain of TraJ could functionally replace the corresponding domain of VirB8 (see Discussion). We first used BACTH analysis to investigate whether the periplasmic domains of VirB8 and TraJ could interact to form heterodimers (Fig. 3). At 37°C, heterodimers were formed with β-galactosidase activity higher than that of the homodimers, again suggesting a stronger interaction. We then examined the formation of heterodimers between full-length VirB8 and full-length TraJ or the chimeras. Here again, all of the combinations could form heterodimers. As with the homodimers, the presence of the cytoplasmic and TM regions of TraJ (TraJ and TraJB8) resulted in at least 10-fold higher activities, suggesting much stronger interactions than with the corresponding regions of VirB8 (VirB8 and B8TraJ) (Fig. 5).

FIG. 5.

Heterodimeric interactions between full-length wild-type proteins or chimeras (pIN73) and full-length VirB8 (pIN94). Following overnight culture at 37°C, β-galactosidase activities were measured as described in Materials and Methods. Results are the mean ± the standard error of the mean of at least three different cultures.

Expression of TraJ and chimeras in Brucella has a dominant negative effect on virulence.

These strong interactions between various proteins and the wild-type VirB8 protein at the temperature normally encountered by Brucella during an infection suggest that overexpression of some of these proteins in a strain synthesizing wild-type VirB8 could have effects on virulence. We examined the effects on the virulence of the wild-type strain of B. suis of overexpressing one of these proteins from a multicopy plasmid. The six different strains of B. suis we used were wild-type strain 1330 (S1) without any plasmid or the same strain containing the empty vector (pIN34) or a plasmid expressing one of the full-length proteins described previously (pIN38, pIN54, pIN56, and pIN57) (38). Intracellular survival and multiplication were assessed at different times following the beginning of the infection (Fig. 6). Two hours after the beginning of the infection, no difference could be observed between the different Brucella strains tested. After that time point, the Brucella strain containing plasmid pIN34, without any gene, showed a small decrease in virulence at 24 h, compared to the wild type, and was only slightly less virulent than the wild type at 48 h. The strain containing plasmid pIN38, which encodes the wild-type VirB8 protein, displayed a consistent about 1-log decrease in virulence at 24 h, compared to the wild-type strain, and an about 20-fold decrease in virulence at 48 h. The strain containing plasmid pIN57, which encodes a chimera with the cytoplasmic and TM parts of VirB8 and the periplasmic part of TraJ, displayed a behavior qualitatively similar to that observed with VirB8 but with a more marked decrease (around 40-fold) at both 24 and 48 h. The last two strains, both containing plasmids which encode proteins with the cytoplasmic and TM parts of TraJ, were clearly strongly attenuated. Both displayed decreases in virulence of more than 2 logs compared with the wild-type strain at 24 h and a similar attenuation (pIN56) or even a complete lack of virulence (pIN54) at 48 h.

FIG. 6.

Comparison of the virulence of wild-type B. suis 1330 (S1) and S1 containing the empty pIN34 plasmid with S1 overexpressing different VirB8 proteins. J774 murine macrophage-like cells were infected in a standard gentamicin protection assay as described previously (37). Intracellular survival and multiplication of the bacteria were followed at different times after the beginning of the infection. The results shown are representative of at least three independent experiments.

To ascertain that the differences in virulence observed between the different strains were not due to differences in their growth rates, we measured the doubling time of each strain in 2YT medium. Table 3 shows clearly that the wild-type S1 strain grows slightly faster than any other strain, displaying a doubling time significantly shorter than those displayed by the other strains. Except for the wild-type strain, all of the other strains grew at the same rate, displaying doubling times not statistically significantly different from one another.

TABLE 3.

Growth rates of bacterial strains used in macrophages infection studies

Brucella strain	Mean doubling time (h) ± SEMa	P valueb
S1 (1330)	2.18 ± 0.08
S1/pIN34	2.56 ± 0.06	0.003
S1/pIN38	2.49 ± 0.13	0.023
S1/pIN54	2.62 ± 0.09	0.003
S1/pIN56	2.64 ± 0.18	0.015
S1/pIN57	2.62 ± 0.09	0.003

^aResults of three independent determinations are shown.

^bThe significance of the difference from the S1 doubling time was determined by Student's t test.

DISCUSSION

BACTH analysis is a powerful tool to study protein-protein interactions in bacteria (27). One big advantage over the yeast-based systems is that it allows the study of interactions between membrane proteins (1, 28). We have taken advantage of this property to study the formation of homo- and heterodimers of various protein chimeras made of different parts of two inner membrane proteins, B. suis VirB8 and TraJ from pSB102. As a measure of the strength of the interactions between these proteins, we have used the measurement of β-galactosidase activity (33). As recently shown by our laboratory, performing the overnight culture at 37°C instead of 30°C dramatically potentiated the differences in the strength of interactions (1) and may be more relevant to proteins from mammalian pathogens.

Despite a high degree of sequence similarity, the full-length TraJ protein could not restore the virulence of a B. suis virB8 mutant (38); however, a B8TraJ chimera could do so, at least partially. The results of our previous study with TraJ and the chimeras could be interpreted as meaning that the cytoplasmic and/or TM domains of TraJ do not allow proper interactions between the two subunits of the homodimer TraJ-TraJ or TraJB8-TraJB8 to take place, thus leading to a nonfunctional T4SS. BACTH analysis showed that VirB8, TraJ, TraJB8, and B8TraJ can form homodimers and that the interactions of either TraJ or TraJB8 are, in fact, much stronger than those of VirB8 or B8TraJ (Fig. 4). Western blotting showed that, at least for the periplasmic domains and the full-length proteins and chimeras, the differences observed in β-galactosidase activity are not due to a lack of stability or a mislocalization of these proteins.

Sequence comparisons suggest why the periplasmic domain of TraJ could functionally replace the corresponding domain of VirB8 in B. suis. Almost all of the amino acids in the periplasmic domain which have been shown to play an important role in VirB8 function are strictly conserved in TraJ (37, 38, 46). Amino acids Met₁₀₂, Tyr₁₀₅, and Glu₂₁₄, implicated in VirB8 dimerization, have as their counterparts Met₁₀₃, Tyr₁₀₆, and Glu₂₁₅ in TraJ, suggesting that the periplasmic domain of TraJ should also form heterodimers with the periplasmic domain of VirB8. BACTH analysis confirmed that VirB8p and TraJp can form heterodimers with interactions apparently stronger than in homodimers. Amino acids Thr₂₀₁ and Arg₂₃₀ which are, respectively, involved in the interactions between the periplasmic domain of VirB8 and the periplasmic domain of VirB10 and with VirB4, correspond to amino acids Thr₂₀₂ and Arg₂₃₁ in TraJ. This suggests that the periplasmic domain of TraJ should also be able to interact with these proteins in the VirB T4SS, allowing correct assembly of the T4SS. The amino acids located close to the deep groove found in the structure of the periplasmic domain of VirB8, such as Trp₁₁₉, Tyr₁₂₆, Leu₁₅₁, Asp₁₅₂, and Lys₁₈₂, are all conserved in TraJ. The only exception is the replacement of Gln₁₄₄ of VirB8 with Leu₁₄₅ in TraJ. Finally, one notable difference between the periplasmic domains of VirB8 and TraJ is the replacement of Ile₁₁₂, located at the interface between the two subunits of a homodimer, with Arg₁₁₃ of TraJ. It is worth noting that in the periplasmic domain of A. tumefaciens VirB8, the Arg₁₀₇ residue is found at the same position (5).

The proteins JCyTM and B8CyJTM, containing the TM domain of TraJ preceded by the cytoplasmic domain of either TraJ or VirB8, were both expressed at high levels and interact strongly, forming homodimers (Fig. 4). In contrast, certain proteins containing the TM domain of VirB8 (JCyB8TM and B8CyTM) appeared unstable, explaining the lack of interaction in the BACTH assay. A closer examination of the amino acid sequence of the TM domains of VirB8 and TraJ revealed striking features. Clustal alignment of the two sequences shows that Tyr₅₀ in TraJ corresponds to a gap in the sequence of VirB8 and that Trp₆₀ in TraJ corresponds to Gly₅₉ in VirB8 (see Fig. 3 in reference 38). Both Tyr and Trp residues are much more hydrophobic and bulky than Gly or Ala, the residues located in VirB8 (positions 48 to 50 and 59), suggesting that the presence of these bulky amino acids in the TM domain could be, at least in part, responsible for the stronger interactions.

The full-length proteins containing the TM domain of TraJ formed very strong interactions with VirB8, suggesting that an excess of one protein containing this TM domain in a wild-type B. suis strain could adversely affect its virulence. Overexpression of TraJ or TraJB8 in a wild-type Brucella strain was found to strongly attenuate its virulence. We believe that this is due to the formation of strongly interacting heterodimers. Overexpression of VirB8 or B8TraJ also had a small attenuating effect, suggesting that stoichiometry is also important. Our results raise several questions concerning the assembly and function of the T4SS. The inability of a protein containing the TM domain of TraJ to complement a VirB8 mutant, as well as the dominant negative effects on virulence, may be due to incorrect assembly of the T4SS. We have no way to test T4SS assembly in Brucella; however, we could test the effects of the proteins carrying the TraJ TM domain in the recently described heterologous system in which the whole B. suis T4SS expressed in A. tumefaciens increases its ability to serve as a recipient in T4SS-mediated plasmid conjugation (9). This is thought to be a measure of T4SS assembly (9, 32).

The T4SS can also be viewed as a dynamic structure. Using the TrIP assay, the Christie group has been able to follow the pathway that the T-DNA complex takes across the VirB channel (11). A combination of immunoprecipitation of selected VirB/D4 proteins and PCR amplification of the Ti plasmid was used to identify contacts between the T-DNA and components of the virB/D4 conjugation system. The T-DNA could be found first bound to the protein VirD4 and then successively to the proteins VirB11, VirB6, VirB8, VirB9, and VirB2. In addition, although not directly binding the T-DNA, proteins VirB3, VirB4, VirB5, VirB7, and VirB10 were absolutely required for the T-DNA to pass through the different steps of this pathway. If we assume a similar role for VirB8 in Agrobacterium and Brucella, we can predict that this protein will interact with effectors during translocation. VirB7, -9, and -10 have recently been shown to form a double-walled channel spanning both membranes of the bacterial envelope (20). One can imagine that the T4SS must be “closed” to stop leakage and then “open” transiently to allow substrate passage; we can speculate that VirB8 may play the role of “gatekeeper,” controlling the passage of the substrate through the channel. Further support for the dynamic nature of T4SS comes from data showing that the VirB10 protein undergoes ATP-dependent conformational changes which result in the formation of transient “bridges” linking inner and outer membrane-associated subassemblies of the T4SS (12). In addition to playing an essential role in the translocation of effectors, VirB8 is involved in numerous protein-protein interactions, particularly with other VirB proteins (15, 17, 22, 24, 26, 30, 46, 47, 49). However, it has been suggested (6) that, given its relatively small size, VirB8 is unlikely to engage simultaneously in such a large number of interactions. This implies that these interactions might occur transiently at defined time points during the translocation process, with possible conformational changes in VirB8 allowing these different interactions to take place. Dimerization of VirB8 can therefore be viewed as a transient state during the translocation process, resulting in a dynamic equilibrium between monomers and dimers of this protein in the T4SS.

T4SSs have been very elusive structures in most animal pathogens and to date have only been visualized on Helicobacter pylori (2, 42). Taking the notion of a dynamic structure to an extreme, could this be because they are, in fact, very transient, only fully assembled when needed and then rapidly disassembled once the effector molecules have been translocated?