Identification of the VirB4-VirB8-VirB5-VirB2 Pilus Assembly Sequence of Type IV Secretion Systems*

Type IV secretion systems mediate the translocation of virulence factors (proteins and/or DNA) from Gram-negative bacteria into eukaryotic cells. A complex of 11 conserved proteins (VirB1-VirB11) spans the inner and the outer membrane and assembles extracellular T-pili in Agrobacterium tumefaciens . Here we report a sequence of protein interactions required for the formation of complexes between VirB2 and VirB5, which pre-cedes their incorporation into pili. The NTPase Walker A active site of the inner membrane protein VirB4 is required for virulence, but an active site VirB4 variant stabilized VirB3 and VirB8 and enabled T-pilus formation. Analysis of VirB protein complexes extracted from the membranes with mild detergent revealed that VirB2-VirB5 complex formation depended on VirB4, which identified a novel T-pilus assembly step. Bicistron expression demonstrated direct interaction of VirB4 with VirB8, and analyses with purified proteins showed that VirB5 bound to VirB8 and VirB10. VirB4 therefore localizes at the basis of a trans-envelope interaction sequence, and by stabilization of VirB8 it mediates the incorporation of VirB5 and VirB2 into extracellular pili. and Gel Filtration— Isolation of membranes and detergent extraction with 2% dodecyl- (cid:1) - D -maltopyranoside (DDM) were performed as previously described (20). Blue native gradient gel electrophoresis (7– 14% acrylamide, 16-cm-long gel) was performed for 20 h and 5 mA at 4 °C. Alternatively, for separation of small proteins, blue native gel electrophoresis was performed in a linear 15% polyacrylamide gel for 5–6 h and 5 mA at 4 °C in a mini-gel system (10-cm-long gel). Calibration was achieved by separation of reference proteins of known molec- ular masses: thyroglobulin (669 kDa), ferritin (440 kDa), catalase (232 kDa), lactate dehydrogenase (140 kDa), albumin (67 kDa), ovalbumin (43 kDa), chymotrypsinogen A (25 kDa), and ribonuclease A (14 kDa, Amersham Biosciences). For gel filtration, 100 (cid:2) l of 2% DDM-solubi- lized membrane proteins were applied to a Superdex 200 or Superdex 75 (XK 16/40; Amersham Biosciences), and chromatography was per- formed in 6-amino-caproic acid buffer (750 m M 6-amino-caproic acid, 50 m M BisTris, pH 7) containing 0.03% DDM performed at 4 °C with a flow rate of 0.5 ml/min in an ¨Akta fast protein liquid chromatography puri-fier (Amersham Biosciences). (cid:2) of calibration kit proteins (ferritin, mg/ml; all other proteins, 10 mg/ml) were separated in 6-amino-caproic acid with 0.03% Blotting— Cells were incubated sample for at followed with VirB2, which is a key step in the formation of the pilus assembly subcomplex. We suggest that the effect of VirB4 on the localization and stability of VirB3 is mediated via VirB5 properly localized in the pilus assembly subcomplex.

gen A. tumefaciens. The T4SS of the closely related animal pathogen B. suis is encoded by an operon of similar organization, and it is essential for survival and multiplication inside mammalian cells (6). The T4SSs of these bacteria share 11 proteins, which can be divided into three groups. The first group comprises two inner membrane-associated NTPases (VirB4 and VirB11), which reside mainly in the cytoplasm but may traverse the inner membrane and contact periplasmic T4SS components (7)(8)(9). They contain Walker A nucleotidebinding motifs and are believed to energize T4SS assembly or substrate transfer. The second group consists of inner membrane VirB6 and periplasmic VirB7, VirB8, VirB9, and VirB10. They form the T4SS core and may constitute the translocation channel (10 -13). The third group comprises the major T-pilus component VirB2, the minor component VirB5, and the pilusassociated protein VirB7 (14 -17). VirB3 has not been firmly assigned, but its outer membrane localization and binding to VirB5 suggest that it is a pilus-associated protein (18,19). Biochemical experiments based on extraction of VirB proteins with a mild detergent followed by separation under native conditions led to a model for T-pilus assembly (20). This model is refined here based on improved separation methods and the analysis of the contribution of VirB4.
VirB4 is the largest T4SS component (A. tumefaciens, 87 kDa; B. suis, 94 kDa) and exhibits the highest degree of conservation among T4SS components (31% identity and 52% similarity between A. tumefaciens and B. suis). It is homodimeric or homomultimeric and is essential for virulence (8,21). An important feature is its Walker A nucleotide-binding site, which is essential for virulence and plasmid transfer (22,23). ATPase activity of purified A. tumefaciens VirB4 was previously reported (24), but a more recent study argues against such an activity of the purified protein (25). Coordinated action of VirB4 with VirB11 and VirD4 was proposed to mediate the early DNA transfer reactions (26). In addition, VirB4 stabilized VirB3 and was required for its localization in the outer membrane (19). A study using the yeast two-hybrid system suggested that VirB4 binds to VirB1, VirB8, VirB10, and VirB11, but this was not substantiated with biochemical methods (27).
Interestingly, the VirB4 NTPase active site is essential for virulence, but it is not required for interactions with VirB11 and VirD4, for self-association, and for the stimulation of IncQ plasmid transfer into A. tumefaciens by the T4SS in the recipient (8). To unravel the role of VirB4, we constructed a virB4 deletion mutant and assessed the complementation with A. tumefaciens and B. suis VirB4 and NTPase active site variants. We refined the protocol for the separation of detergent-extracted VirB proteins and identified a novel step in T-pilus assembly. Purified components were used to study interactions in vitro, and this revealed an interaction sequence from the VirB4-VirB8 complex to the pilus component VirB5 and the core component VirB10. This work gives fundamental insights into the contribution of VirB4 to T4SS complex stabilization and T-pilus assembly.
Plasmid and Strain Constructions and Mutagenesis-DNA manipulations followed standard procedures (28). A. tumefaciens virB genes were amplified from pGK217, and B. suis virB4 genes were amplified from pUCvirB with oligonucleotides, cleaved with restriction sites, and cloned into vectors with compatible sites (Table II). The codons determining the ATP-binding site Lys residues in VirB4 and VirB4s were changed to Arg using the in vitro site-directed mutagenesis system (Promega). The sequences of PCR-amplified genes were confirmed by DNA sequencing.
Transmission Electron Microscopy-A. tumefaciens strains to be examined were cultivated on AB agar plates in the presence or absence of AS and ARA for gene induction. Cells were collected with 5 ml of 50 mM sodium potassium phosphate buffer, pH 5.5, and the cell density was adjusted to A 600 of 1.5-2. 10 l were applied onto UV-sterilized 200 mesh carbon-coated formvar copper electron microscopy grids and air dried in a laminar flow hood for 10 min. The grids were then stained with 2% phosphotungstic acid-0.01% glucose, pH 6, for 15 s prior to examination. Specimen images were taken with a JEOL 1200EX II transmission electron microscope. T-pili on 300 cells from three independent virulence induction experiments of each strain were counted (10 cells per visual field were analyzed).
SDS-PAGE and Western Blotting-Cells were incubated in Laemmli sample buffer for 5 min at 100°C followed by SDS-PAGE. Chromatography samples were incubated in Laemmli sample buffer buffer for 30 min at 37°C to avoid aggregate formation followed by SDS-PAGE using the Laemmli (for proteins larger than 20 kDa) or the Schä gger and von Jagow system (for proteins smaller than 20 kDa) (30,31). Western blotting was performed following standard protocols (32), with VirB protein-specific antisera.
Purification of Fusion Proteins-N-terminally hexahistidyl-thioredoxin (H 6 TrxA)-tagged proteins were overproduced in GJ1158 using pT7-H 6 TrxFus constructs (33). Cells from 400-ml cultures were suspended in 5 ml of lysis buffer (50 mM Hepes, 200 mM KCl, 5 mM MgCl 2 , pH 7.5) with 0.5 mM phenylmethylsulfonyl fluoride and lysed by a French Pressure Cell (Aminco) at 18,000 p.s.i. The lysate was centrifuged twice (SS34 rotor, 30 min, 13,000 rpm at 4°C), and the supernatant was applied to a high pressure liquid chromatography system (Äkta Purifier; Amersham Biosciences) with a Ni 2ϩ -charged IMAC column (Talon TM Superflow; Clontech). Tagged proteins were eluted using a step gradient. At a flow rate of 0.5 ml/min, the column was first washed for 5 column volumes in buffer A (50 mM Hepes, 0.3-1 M NaCl, pH 7.5), followed by washing with buffer A with 20 mM imidazole (2.5 column volumes) and elution with buffer B (buffer A with 400 mM imidazole). The fractions were dialyzed overnight in H1B (50 mM Hepes, 200 mM NaCl, pH 7.5), followed by a second dialysis overnight in H2B (H1B with 50% glycerol and 2 mM dithiothreitol) and storage at Ϫ20°C.
N-terminally StrepII-tagged proteins were overproduced in GJ1158 using pT7-7StrepII (34), and 400-ml cultures were lysed in 10 ml of S2B (0.1 M Tris-HCl, pH 8, 0.15 M NaCl, 1 mM EDTA, pH 8) with 0.5 mM phenylmethylsulfonyl fluoride. The cells were lysed, and the supernatant was collected as described above. Fusion proteins were purified with a 1-ml Strep-Tactin Superflow® column (IBA, Göttingen, Germany) following the instructions of the manufacturer using 2.5 mM desthiobiotin in the elution buffer. The fractions were subsequently purified by size exclusion chromatography using S2B at a flow rate of 0.5 ml/min. Superdex 75 or Superdex 200 (Amersham Biosciences) was used, depending on the molecular mass of the protein. The proteins were then dialyzed overnight in PSB (S2B buffer with 50% glycerol and 2 mM dithiothreitol) and stored at Ϫ20°C.
Assays for Protein-Protein Interactions: Pull-down and Cross-linking Experiments-For pull-down experiments, 10 l each of purified Stre-pII-and H 6 TrxA-tagged proteins concentrated at 5 pmol/l in PSB were mixed, followed by the addition of 80 l of S2B and incubation for 30 min at 22°C. Next, 20 l of Strep-Tactin Sepharose beads (50% suspension in S2B; IBA) were added, followed by a 15-min incubation at 22°C. The Sepharose beads were subsequently sedimented by centrifugation and washed three times with 200 l of S2B. Bound proteins were eluted with 50 l of S2B with 1 mM biotin, mixed with 1 volume of Laemmli sample buffer, and analyzed by SDS-PAGE and Western blotting.
For chemical cross-linking, 5 l each of 10 pmol/l stock solutions of purified StrepII-tagged proteins in PSB (or 5 l of PSB as negative control) were mixed for 5 min at 22°C, 90 l of CLB (50 mM MES-KOH, 150 mM NaCl, 1 mM EDTA, pH 6.5) were added, and the mixture was incubated for 30 min. The cross-linking agent disuccinimidyl suberate (10 mM stock in Me 2 SO; Pierce) was added in different concentrations (0.05 and 0.1 M), and the samples were incubated for 1 h at 22°C, followed by the addition of 1 volume of Laemmli sample buffer and analysis by SDS-PAGE and Western blotting.
Yeast Two-hybrid Assay-The Matchmaker 3 system (Clontech) was used for the analysis of protein-protein interactions with the yeast two-hybrid system following the manufacturer's protocols. The genes encoding the periplasmic domains VirB5sp, VirB8sp, VirB9sp, and VirB10sp were cloned into pGADT7 (GAL4 activation domain fusion) and pGBKT7 (DNA binding domain fusion) as described above. The plasmids were transformed into Saccharomyces cerevisiae AH109 using the lithium acetate method (Clontech Manual; Clontech), and plasmidcarrying cells were selected on minimal medium without leucine and tryptophan. Interactions between VirB proteins, which tethered both domains of GAL4 together, were identified by growth of plasmid-carrying cells on minimal medium without histidine and adenine and by ␤-galactosidase activity.

RESULTS
VirB4 Stabilizes VirB3 and VirB8 -To study the role of VirB4, we deleted virB4 in the Ti plasmid of wild-type C58, resulting in strain CB1004 (⌬virB4). Next, virulence genes were induced by cultivation on acidic minimal medium with acetosyringone followed by SDS-PAGE and Western blotting. Specific antisera were applied to compare the VirB protein levels in cells of non-induced and virulence gene-induced C58 (controls) and in CB1004. The levels of most VirB proteins were comparable in C58 and CB1004 (Fig. 1B), but VirB3, VirB8, and VirB4 were not detected or were detected only in very small quantities in CB1004 (Fig. 1A). Whereas the absence of VirB4 was expected, and reduced levels of VirB3 had been

Formation of VirB2-VirB5 T-pilus Assembly Complexes
reported previously (19), the reduced levels of VirB8 showed that VirB4 plays a role for the accumulation of this protein.
A previous study in a different A. tumefaciens strain suggested that the structure of VirB4, but not its NTPase activity, is required for VirB protein stabilization (8). To assess this possibility, A. tumefaciens virB4 was cloned behind the tightly controlled E. coli arabinose (BAD) promoter of broad host range vector pVSBAD. An active site change was engineered, and the cloning vector pVSBAD and constructs encoding VirB4 and its Walker A derivative (VirB4 K439R ) were introduced into CB1004. Analysis of the resulting strains demonstrated that production of VirB4 as well as production of VirB4 K439R restored the levels of VirB3 and VirB8 (Fig. 1A). To further study the requirement of VirB4 sequence and structure, we cloned the gene encoding the B. suis VirB4 homolog (VirB4s) into pVSBADNco and engineered an active site change (VirB4s K464R ). The production of both VirB4s and VirB4s K464R fully restored levels of VirB3 and VirB8 (Fig. 1A). When CB1004 complemented with plasmids encoding VirB4 or VirB4s (data not shown) was cultivated with gradually increasing arabinose inducer levels (0.01-0.5%), a gradual increase of VirB4 and concomitant increases of VirB3 and VirB8 levels were observed (Fig. 1C). These results showed that the NTPase activity is not important for the stabilization of VirB3 and of VirB8 and that the heterologous VirB4s provided the required structural information.
The Active Site of VirB4 Is Required for Substrate Transfer but Not for T-pilus Formation and pLS1 Recipient Activity-We next determined the functionality of A. tumefaciens and B. suis VirB4 and of their active site variants using four different complementation assays. First, the virulence was tested on wounded Kalanchoë diagremontiana leaves. Tumor formation was observed only at sites inoculated with C58 and CB1004 producing wild-type VirB4 and VirB4s, respectively ( Fig. 2A). The requirement for the NTPase active site(s) is in accord with previous reports, but complementation by B. suis  VirB4s shows that despite low sequence identity, the heterologous protein exerts all VirB4 function(s). Tumors induced by VirB4s-complemented CB1004 were smaller than those induced by the VirB4-complemented strain. We have not attempted to measure tumor formation and used the more readily quantifiable plasmid conjugation systems for this purpose. To this end, when T4SS-mediated IncQ group plasmid pLS1 transfer between Agrobacteria was employed as second assay, only A. tumefaciens and B. suis wild-type VirB4 complemented the virB4 defect ( Fig. 2B and Table III). As a third assay, we exploited the ability of T4SS to stimulate conjugative pLS1 transfer upon production in the recipient (35). The ability of a virB4 deletion strain to serve as pLS1 recipient was restored by VirB4 as well as VirB4 K439R , which is in accord with previous work, and production of the B. suis proteins had similar effects (Fig. 2C and Table IV). As a fourth assay, T-pili were isolated from the cells by shearing and ultracentrifuga-tion, followed by detection of major T-pilus component VirB2 and minor component VirB5, respectively. Both proteins were not detected in the pilus fractions from CB1004, but production of VirB4, VirB4s, and their active site variants fully restored T-pilus formation (Fig. 2D). The active site of VirB4 was obviously not required for T-pilus assembly, which constitutes a remarkable difference to previous reports (22). To directly assess T-pilus formation, we analyzed the cells from three independent induction experiments (300 cells of each strain were counted) by transmission electron microscopy. Analysis of wildtype strain C58 under virulence gene-inducing conditions revealed T-pili on 77% of the cells (Fig. 3), whereas no T-pili were observed on C58 grown under non-inducing conditions or CB1004 pVSBADNco. T-pilus formation in CB1004 was at least partly restored by production of VirB4 (61%), VirB4s (56%), VirB4 K439R (38%), and VirB4s K464R (36%). Whereas the formation of T-pili in CB1004 carrying the active site variants  was reduced as compared with the wild-type, the difference from the negative controls was very clear, showing that the intact active site was not required for the formation of this structure (Fig. 3B). The above results suggest that VirB4 contributes only structural information to T4SS assembly and T-pilus formation. In contrast, T-DNA translocation required a wild-type Walker A box, but heterologous VirB4s fully complemented those functions. Biochemical methods were used next to analyze the molecular basis of the stabilization phenomenon.
VirB4 Is Required for the Association of Pilus-associated Proteins with VirB3, VirB6, and VirB8 -Extraction of membrane proteins with mild detergents greatly contributed to the understanding of protein uptake machineries in mitochondria (36,37), and we have also found applications for the analysis of bacterial protein translocation systems (38). We have previously adapted this method to the extraction of VirB protein subcomplexes from the membranes of A. tumefaciens (20). Further analyses revealed two subassemblies, a low molecular mass complex of pilus-associated proteins and a high molecular mass complex containing the T4SS core components. This method was applied here to assess the contribution of VirB4. Membranes were isolated from C58, CB1004, and CB1004 producing VirB4, VirB4s, and their active site variants, followed by the extraction with DDM (2%, w/v) and analysis of subcellular fractions by Western blotting. Similar to previous reports, VirB2 was detected in the total cell lysate (Fig. 4A, T) and in the membranes (M). VirB5 was predominantly detected in the membranes, but a significant portion was present in the soluble fraction (Fig. 4A, S ϭ cytoplasmic and periplasmic). Differences were not apparent, showing that the absence of VirB4 did not impact the membrane association of VirB2 and VirB5 (Fig. 4A).
To characterize the formation of the characteristic low molecular mass VirB2-VirB5 complex, the DDM-solubilized proteins were separated by blue native electrophoresis in a 7-14% polyacrylamide gel. Similar to previous findings, VirB2 and VirB5 co-localized in complexes of 140 kDa in samples from C58, but only very low amounts of both proteins were detected in samples from CB1004 (Fig. 4B). Production of VirB4, VirB4s, and their active site variants restored the levels of VirB2 and VirB5 in these complexes (Fig. 4B). These results suggest that whereas the membrane association is not affected in CB1004, VirB4 is necessary for the incorporation of VirB2 and VirB5 into complexes of 140 kDa.
Because blue native PAGE separation has limited resolution, we separated DDM extracts by gel filtration next. In contrast to the previously described procedure, gel filtration was conducted in the presence of 0.03% DDM, which is higher than the critical micellar concentration (0.006 -0.007%), to avoid deter-gent dilution. Similar to previous observations in C58, a high molecular mass complex with the core components VirB4, VirB6, VirB7, VirB8, VirB9, VirB10, and the substrate VirE2 was detected (Fig. 5A). VirB3, VirB6, and VirB8 co-fractionated with VirB2, VirB5, and VirB7 in the low molecular mass complex of pilus-associated proteins (Fig. 5A). Analysis of samples from CB1004 revealed a drastically different distribution. Similar to the wild-type, VirB9, VirB10, and VirE2 were detected in the high molecular mass complex. In contrast, VirB6 and VirB7 eluted exclusively in the high molecular mass fractions (Fig. 5B). As expected, VirB3 and VirB8 were not detected. In addition, we did not detect VirB2 and VirB5 in the fractions eluting from the column. Analysis of samples from CB1004 producing VirB4, VirB4s, and their active site variants revealed VirB protein complexes as in the wild-type (data not shown). Thus, the NTPase activity of VirB4 was not required for the formation of the wild-type VirB protein complexes.
VirB2 and VirB5 Do Not Co-fractionate with Each Other and Other T4SS Components in CB1004 -The observation that VirB2 and VirB5 were not detected in the gel filtration fractions from CB1004 was unexpected because both pilus components associated with the membranes and were extracted by DDM similar to the wild-type (Fig. 4A). One explanation was that VirB2 and VirB5 do not interact with other VirB proteins in CB1004 and are therefore present in low molecular mass complexes. To assess this possibility, we adapted the fractionation techniques to separate low molecular mass proteins. First, blue native electrophoresis was conducted in linear 15% acrylamide gels, which efficiently separated only small proteins. Using this technique, both VirB2 and VirB5 were detected in low molecular mass complexes in extracts from CB1004 (Fig. 6). VirB2 had an apparent molecular mass of 13 kDa, and VirB5 had an apparent molecular mass of 30 kDa. In strains C58 and complemented CB1004, VirB2 and VirB5 were detected in complexes similar to the wild-type.
As a second approach to identify VirB2 and VirB5 in CB1004, we separated the DDM extracts on a Superdex 75 gel filtration column, which resolved proteins smaller than 100 kDa. In extracts from C58, VirB2 and VirB5 co-eluted in fractions of high molecular masses (Ͼ150 kDa), which could not be resolved (Fig. 7A). In extracts from CB1004, VirB2 and VirB5 eluted in separate fractions of low molecular masses of 13 and 30 kDa, respectively (Fig. 7B), and complementation of CB1004 restored the wild-type situation (data not shown). The results show that in the absence of VirB4, the interactions of VirB2 and VirB5 with each other and with other T4SS components are weakened or abolished.
VirB4s Binds to the Periplasmic Domain of VirB8s-The absence of VirB4 strongly impacts the formation of VirB2- VirB5 complexes, but it was not obvious whether this effect was direct or indirect via stabilization of VirB3 and/or VirB8. Interaction between VirB4 and VirB8 had been predicted, and our data are consistent with this finding (27). We directly tested this possibility using bicistron expression (39). Because biochemical work with A. tumefaciens VirB proteins proved to be very difficult (low solubility and yield upon overproduction), 2 we performed the following experiments with B. suis VirB homologs (named VirBs). The virB4s gene was cloned into pT7-7StrepII to produce an N-terminal fusion with the StrepII peptide for affinity purification and detection. Next, two frag-ments of the VirB8s-encoding gene (full-length and one encoding the periplasmic domain VirB8sp) were cloned downstream of virB4s. The resulting constructs pT7-7StrepIIVirB4s, pT7-7StrepIIVirB4sVirB8s, and pT7-7StrepIIVirB4sVirB8sp were transformed into E. coli strain GJ1158, followed by expression and detection of StrepII-VirB4s with a StrepII-specific mono-

FIG. 5. Analysis of DDM-extracted virulence protein complexes from C58 and CB1004 by Superdex 200 gel filtration.
A. tumefaciens wild-type C58 and virB4 deletion strain CB1004 were grown on AB minimal medium in the presence of AS and pBAD inducer arabinose. Cells were lysed, and membranes were isolated and incubated with 2% DDM, followed by gel filtration in the presence of 0.03% DDM and analysis of the fractions by SDS-PAGE and Western blotting. Low molecular weight (LMW) fractions contain the pilus assembly complex, and high molecular weight (HMW) fractions contain the core T4SS proteins; linker proteins are part of both complexes. A, analysis of extracts from C58. B, analysis of extracts from CB1004. Molecular masses of reference proteins are shown on the right (in kDa). clonal antibody and the VirB8s variants with a specific antiserum (data not shown). To study interactions, cell lysates were applied to a streptavidin affinity matrix, which enriched StrepIIVirB4s. VirB8s and VirB8sp co-eluted with StrepIIVirB4s, and controls showed that the column had very weak affinity for non-tagged VirB8sp (data not shown). To assess whether the co-elution reflects a stable interaction, we subjected samples to two separation procedures. First, the proteins were separated by blue native PAGE, and VirB4s was detected in complexes slightly larger than 440 kDa in all cases (Fig. 8A). A minor portion was detected in larger complexes in samples from strains carrying the bicistron constructs. VirB8s and VirB8sp were mainly detected in complexes of similar size such as VirB4s, but minor fractions were present in higher and lower molecular mass forms (Fig. 8A). As a control, we separated StrepII-VirB8sp, and it was detected as monomer of 20 kDa. The results suggest that VirB4s formed multimers and that VirB8s and VirB8sp bound to this complex when expressed from bicistron constructs. As a second approach, the affinity column-enriched samples were subjected to gel filtration over a Superdex 200 column. StrepIIVirB4s co-eluted with VirB8s and VirB8sp in fractions 8 (corresponding to 440 kDa) and 9, suggesting that they form a complex (Fig. 8B). StrepIIVirB4s enriched from pT7-7StrepIIVirB4s-carrying cells eluted exclusively in fraction 9, showing that the StrepIIVirB4s complex was smaller than the StrepIIVirB4s-VirB8s and StrepIIVirB4s-VirB8sp complexes (Fig. 8C). As a control, StrepIIVirB8sp was analyzed and proved to be much smaller than StrepIIVirB4s complexes (Fig. 8C). Finally, to assess the stability of the StrepIIVirB4s-VirB8s complex, it was applied to a Superdex 75 column, which would permit the detection of VirB8s, if it dissociated from StrepIIVirB4s. Both StrepIIVirB4s and VirB8s were exclusively detected in the void elution volume, indicating a large and stable complex (Fig. 8D). Taken together, our experiments show that StrepIIVirB4s forms multimers, which constitutes direct evidence for the formation of the hexameric complex recently predicted based on a bioinformatics and modeling approach (40). In addition, it forms stable complexes with full-length VirB8s as well as with its periplasmic domain VirB8sp.
VirB5s Binds to T4SS Core Components VirB8s and VirB10s-After demonstrating the interaction between VirB4s and VirB8s, we next assessed the possibility that VirB8s may bind to VirB5s and thereby impact its incorporation into pili. We also analyzed the core components VirB9s and VirB10s because their A. tumefaciens homologs were previously shown to interact with VirB8 (13). First, we performed pull-down assays with differentially tagged VirBs proteins. To this end, the genes encoding the periplasmic domains of VirB5s, VirB8s, VirB9s, and VirB10s were cloned into pT7-7StrepII and pT7-H 6 TrxFus, respectively, for overexpression and purification of StrepII-and H 6 TrxA-tagged proteins. The different sizes of the tags enabled identification based on their molecular masses and recognition by specific antisera. StrepIIVirB5sp was mixed with H 6 TrxAVirB5sp, H 6 TrxAVirB8sp, H 6 TrxAVirB9sp, and H 6 TrxAVirB10sp, respectively, and streptactin-Sepharose beads were added to isolate the StrepII-tagged bait and the H 6 TrxA-tagged preys. StrepIIVirB5sp pulled down H 6 TrxAVirB8sp and H 6 TrxAVirB10sp, but not H 6 TrxAVirB9sp (Fig. 9A), suggesting interactions with these two core T4SS components. H 6 TrxAVirB5sp bound non-specifically to the affinity matrix so that this assay could not be used to assess the selfinteraction of this protein (data not shown).
As an alternative method, we applied the cross-linking agent disuccinimidyl suberate to StrepII-tagged VirBsp proteins and mixtures thereof, followed by SDS-PAGE and Western blot analysis. Application of disuccinimidyl suberate led to the formation of higher molecular mass complexes, indicating homomultimer formation (Fig. 9B). When mixtures with StrepIIVirB5sp were subjected to cross-linking, we observed changes of the cross-linking patterns. Novel complexes appeared when StrepIIVirB5sp was cross-linked in the presence of StrepIIVirB8sp and StrepIIVirB10sp (Fig. 9B). Similar to the results of the pull-down experiments, this indicated interactions with StrepIIVirB5sp (Fig. 9B).
Third, the regions encoding the periplasmic domains were cloned into pGADT7 (fusion to GAL4 activation domain) and pGBKT7 (fusion to GAL4 DNA binding domain), and their interactions were tested using the yeast two-hybrid system. An interaction of the prey VirB5sp with the bait VirB8sp was shown by restoration of growth of the yeast strain AH109 on medium without adenine and on medium without histidine and by ␤-galactosidase activity (Table V). This approach also suggested self-interaction of VirB5sp. However, it did not show interactions with the baits VirB9sp and VirB10sp, and assays in the reverse order with VirB5sp as bait did not indicate interactions (data not shown). Nevertheless, we demonstrated the not previously reported VirB5sp-VirB8sp interaction with three independent methods, and two of those suggested that VirB5sp also interacts with VirB10sp. DISCUSSION In this study, we define the contribution of VirB4 to T4SS stabilization and pilus assembly via a VirB4-VirB8-VirB5-VirB2 interaction sequence. There is compelling evidence for the requirement of the Walker A nucleotide-binding site for virulence, but the enzymatic activity (presumably ATPase) has not been demonstrated. We show here that it is dispensable for T4SS stabilization and pilus assembly, suggesting that its role is to energize T-complex translocation. Similar to previous findings, we showed that VirB4 stabilizes VirB3, and here we reveal that it also stabilizes VirB8. The Walker A active site was not required for stabilization, and the B. suis homolog VirB4s fully complemented the virB4 deletion CB1004. In accord with previous work, VirB4 was required for T-pilus formation, and VirB4 and VirB4s restored T-pilus formation in CB1004. In contrast to previous findings, however, the active site variants VirB4 K439R and VirB4s K464R fully complemented T-pilus formation (22). This difference may be explained by the active site change we introduced, which differs from those in previous studies. The active site Lys was here changed to Arg, a conservative change, which is known to abolish the NTPase function of Walker sites but preserves active site structure and NTP binding (41,42). In contrast, in most previous studies, the Lys residue was changed to Glu, Gln, or Met, or small deletions were introduced (8,9,22). These changes may cause substantial alterations of the conformation, and the amounts of some VirB4 variants were reduced, which may explain the effects on T-pilus formation. Because the change introduced here abolished virulence but not pilus assembly, we conclude that the NTPase activity is exclusively required to energize T-complex translocation, which is in accord with recent studies using a transfer DNA immunoprecipitation assay (26).
The fact that T-pilus assembly depends on VirB4 but not on its NTPase activity was surprising at first, and we further dissected this process with biochemical methods. T-pili were not formed in CB1004, but the levels of VirB2 and VirB5 did not differ from C58, and they were extracted with similar efficiency from the membranes. Therefore, in contrast to VirB3 and VirB8, their stability is not reduced, and their membrane association is not affected in the absence of VirB4. We then analyzed the VirB protein subcomplexes. First, DDM extracts were separated by gel filtration. In our previous work, the separation had been conducted without detergent in the column buffer, which raised concerns about protein solubility upon dilution of the detergent (20). Here we conducted the separation in the presence of 0.03% DDM, which is higher than the critical micellar composition. Use of the modified procedure led to changes of VirB protein fractionation as compared with previous work. In the absence of DDM, VirB3, VirB6, and VirB8 had eluted exclusively with the high molecular mass core T4SS components. In the presence of 0.03% DDM, however, VirB3 eluted exclusively with VirB2 and VirB5 in the low molecular mass fraction, whereas VirB6 and VirB8 were equally distributed between the high molecular mass core complex and the low molecular mass complex with VirB2, VirB3, and VirB5. These results reflect the association between VirB proteins more appropriately because the inclusion of DDM prevents artifacts due to detergent dilution. Thus, VirB6 and VirB8 may link the core T4SS proteins to the pilus assembly complex of VirB2, VirB5, VirB7, and VirB3. VirB3 had previously been shown to localize to the outer membrane and to interact with VirB5, suggesting that it may indeed be part of the pilus assembly subcomplex (18,19).
Analysis of extracts from the virB4 deletion mutant further supported the notion that VirB6 and VirB7 link the core components and the pilus assembly subcomplex. In extracts from CB1004, VirB6 and VirB7 were exclusively detected in the high molecular mass fraction with VirB9 and VirB10. To our surprise, VirB2 and VirB5 were not detected in any of the fractions from CB1004 eluting from the Superdex 200 column. However, fractionation over a Superdex 75 column, which separates small proteins, showed that VirB2 and VirB5 eluted as dimers and monomers, respectively. Similarly, after separation of DDM extracts from CB1004 by blue native PAGE in gels of high acrylamide concentration, the pilus components were detected in small complexes. Complementation of CB1004 with plasmids expressing VirB4, VirB4s, or its active site variants restored the wild-type situation, showing that the NTPase activity is not required for the formation of VirB protein subcomplexes. Taken together, in CB1004, VirB2 and VirB5 did not co-fractionate with each other and with other VirB proteins. This may reflect a loss of interaction or a weakened interaction, which could be dissociated by native separation techniques. Our results show that VirB4, but not its NTPase activity, is essential for the formation of interactions between VirB2 and VirB5 and with VirB3, VirB6, VirB7, and VirB8.
The membrane fractionations did not show whether the effects of VirB4 on the stabilization of VirB3/VirB8 and on the VirB5-VirB2 interaction were direct or indirect. To address this question, we conducted in vitro experiments. When the genes encoding VirB4s and VirB8s were expressed as a bicistron in E. coli, the gene products formed a stable complex, and the periplasmic domain VirB8sp was sufficient. VirB4s therefore binds to and stabilizes VirB8s, which is in line with predictions by yeast two-hybrid analysis. In addition, analyses by gel filtration and blue native PAGE revealed that VirB4s forms multimers, and their size is in accord with the hexamers recently predicted based on a bioinformatics approach (40). Our results therefore support the notion that VirB4 proteins function as homo-hexameric complexes much like VirB11 and VirD4 (43,44). We next analyzed whether VirB8sp or other core T4SS components interact with VirB5sp. Using pull-down and cross-linking experiments and yeast two-hybrid analysis, we showed that VirB5sp binds to VirB8sp as well as to VirB10sp. This suggests that VirB4-stabilized VirB8 impacts VirB5-VirB2 complex formation via its direct interaction with VirB5 and perhaps in concert with VirB10. Based on the data in this study, we propose a refined model of T-pilus assembly, which takes the contribution of VirB4 into account (Fig. 10).
VirB4 resides at the inner face of the cytoplasmic membrane, its short N-terminal domain is exposed to the periplasm (9), and it binds to and stabilizes VirB8 via its periplasmic domain. VirB8 binds to VirB5, which is stabilized by VirB6 (12). Stabilized and properly oriented VirB5 forms a complex with VirB2, which is a key step in the formation of the pilus assembly subcomplex. We have not directly assessed whether VirB4 or VirB8 binds to and stabilizes VirB3, which may also be a component of the pilus assembly subcomplex. Nevertheless, B. henselae VirB5 was shown to directly bind VirB3 (18), and we suggest that the effect of VirB4 on the localization and Processed VirB2 and VirB5 subsequently associate with the membranes. VirB4 stabilizes VirB8, which binds to VirB5, possibly in concert with VirB10. Stabilized and properly oriented VirB5 forms a complex with VirB2, which is a key step in the formation of the pilus assembly subcomplex. We suggest that the effect of VirB4 on the localization and stability of VirB3 is mediated via VirB5 properly localized in the pilus assembly subcomplex. ϪAde Formation of VirB2-VirB5 T-pilus Assembly Complexes stability of VirB3 is mediated via VirB5 properly localized in the pilus assembly subcomplex. The above model is supported by the results of the gel filtration experiments, which showed that in C58, VirB6 and VirB8 partly co-fractionated with the pilus assembly subcomplex. VirB6 and VirB8 therefore link the core components to the pilus assembly subcomplex. In the absence of VirB4, the key component VirB8 is not stabile, which leads to a loss of VirB5-VirB2 and VirB5-VirB3 interactions, the pilus assembly subcomplex does not form, and VirB6 and VirB7 re-distribute to the core complex. Taken together, the experiments presented here reveal several novel features of the T-pilus assembly process in A. tumefaciens, which are likely conserved in other T4SSs. They provide a concise explanation for the observation that VirB4 stabilizes the T4SS but that its presumptive ATPase activity is not necessary. In addition, this work constitutes the basis for future experiments to study the activation of NTPase activity of VirB4. Despite efforts in different laboratories, nucleotide hydrolysis by this protein has not been conclusively demonstrated. A detailed biochemical analyses of purified VirB4 homologs TrbE from RP4 and TrwK from R388 showed that the purified proteins do not hydrolyze ATP or GTP (25). It is thus likely that hydrolysis depends on interaction(s) with other T4SS components or substrates. The bicistron approach we pursued here showed the interaction of VirB4s with VirB8s, and it will be interesting to assess whether VirB4s hydrolyzes nucleotides under these conditions. Similar approaches could be pursued to systematically study other putative VirB4 interaction partners, such as VirB11 and translocated substrates. This work may reveal in the future which of these interactions triggers the NTPase activity of VirB4.