Topology of the VirB4 C Terminus in the Agrobacterium tumefaciens VirB/D4 Type IV Secretion System*

Gram-negative type IV secretion systems (T4SSs) transfer proteins and DNA to eukaryotic and/or prokaryotic recipients resulting in pathogenesis or conjugative DNA transfer. VirB4, one of the most conserved proteins in these systems, has both energetic and structural roles in substrate translocation. We previously predicted a structural model for the large C-terminal domain (residues 425-789) of VirB4 of Agrobacterium tumefaciens. Here we have defined a homology-based structural model for Agrobacterium VirB11. Both VirB4 and VirB11 models predict hexameric oligomers. Yeast two-hybrid interactions define peptides in the C terminus of VirB4 and the N terminus of VirB11 that interact with each other. These interactions were mapped onto the homology models to predict direct interactions between the hexameric interfaces of VirB4 and VirB11 such that the VirB4 C terminus stacks above VirB11 in the periplasm. In support of this, fractionation and Western blotting show that the VirB4 C terminus is localized to the membrane and periplasm rather than the cytoplasm of cells. Additional high resolution yeast two-hybrid results demonstrate interactions between the C terminus of VirB4 and the periplasmic portions of VirB1, VirB8, and VirB10. Genetic studies reveal dominant negative interactions and thus function of the VirB4 C terminus in vivo. The above data are integrated with the existing body of literature to propose a structural, periplasmic role for the C-terminal half of the Agrobacterium VirB4 protein.

required for pathogenesis by bacteria that cause many human diseases, such as whooping cough, brucellosis, gastric ulcers, and Legionnaires disease (reviewed in Refs. 1 and 2). Genomic sequence mining reveals many more T4SSs of unknown function.
Agrobacterium tumefaciens, a soil-borne plant pathogen, encodes one of the best-understood and most flexible T4SSs (reviewed in Refs. [3][4][5]. In nature, Agrobacterium uses the virencoded T4SS to genetically modify a large number of dicotyledonous plants by transferring a DNA portion, termed the T strand, from itself to plant cells. The cellular changes set in motion by the integration of the T strand give rise to crown gall disease. In the laboratory, the vir-encoded T4SS transfers DNA and protein substrates to bacteria, plants, fungi, and mammalian cells. Agrobacterium is commonly used to create transgenic plants for research and commercial agriculture. The twelve proteins of the vir-encoded T4SS (reviewed in Refs. [3][4][5] can be divided into several subgroups: 1) the transfer pilus (T pilus) composed of VirB2 (the major pilin subunit), VirB5 (a minor pilin subunit), and VirB7 (a lipoprotein that may help anchor the pilus to the outer membrane); 2) the trans envelope proteins VirB6 -VirB10; 3) three inner membrane-associated proteins with nucleotide binding domains (VirD4, VirB4, and VirB11), which likely energize transporter assembly or substrate translocation; and 4) proteins with additional specific or unknown functions such as (a) VirB1 (a two-domain protein with two proposed functions, peptidoglycan lysis to assemble the trans envelope core structure and plant interaction via the secreted VirB1 fragment VirB1*), (b) VirD4 (thought to couple the DNA substrate to its cognate T4SS transporter), and (c) VirB3 (a necessary protein of unknown function).
All three ATPases (VirB4, VirB11, and VirD4) physically interact and are proposed to mediate initial T strand substrate transport (13). Although VirD4 and VirB11 have been placed at the entrance of the T4SS (14), the relative position and orientation of VirB4 with regard to these two hexamers are not known. We previously proposed that VirB4 was located in the cytoplasm, acting in concert with VirD4 (8). This assignment was based solely on the striking structural similarity between the C terminus of VirB4 and the cytoplasmic domain of TrwB, a VirD4 homolog. This hypothesis was without experimental support. Here, we have used a combination of yeast two-hybrid, biochemical, bioinformatic, and genetic methods to address the topology of the C terminus of VirB4.
We present two-hybrid interactions between the C terminus of VirB4 and the periplasmic domains of several proteins, VirB1, VirB8, and VirB10. Cell fractionation studies reveal that the C-terminal half of VirB4 is localized to the periplasm, and genetic studies suggest it is physically involved in transporter function in vivo. We have further oriented VirB4 in relation to VirB11. To enable these latter studies, we have generated a structural model of Agrobacterium VirB11 based on a bioinformatics comparison with the solved structure of the Helicobacter pylori homolog HP0525 (10) and presented a refined model of our recently predicted VirB4 structure (8). By combining our findings with existing biochemical data (9,13,15), we suggest that the C-terminal VirB4 hexamer resides in the periplasm directly above an inner membrane-embedded VirB11 hexamer.
Protein Expression-Agrobacteria grown on LB agar Ϯ 200 M acetosyringone were harvested, resuspended in denaturing sample buffer, and boiled for 5 min. Proteins from 10 8 cells were loaded into each lane, resolved on Laemmli (20) or tricine (21) protein gel systems, and transferred to Immobilon-P polyvinylidene difluoride 0.45 m membranes (Millipore Corporation). Standard protocols for Western blotting were used to detect VirB4 (see below for description of antibodies) or HA-tagged VirB4 constructs. HA-tagged proteins were detected with monoclonal HA.11 antibody (Covance Research Corporation). Fractionation-Cells were fractionated based on the procedure of De Maagd and Lugtenberg (22). Briefly, cells were harvested and resuspended in 50 mM potassium phosphate buffer, pH 5.5, and then pelleted by centrifugation at 7000 ϫ g. An aliquot of total cells was obtained and frozen in denaturing sample buffer. The remaining cells were resuspended in an osmotic shock solution of 50 mM Tris, pH 8, 20% sucrose, 2 mM EDTA, 87 g/ml phenylmethylsulfonyl fluoride, and 200 g/ml lysozyme and incubated for 30 min at room temperature. The mixture was centrifuged at 7000 ϫ g, and the supernatant (periplasmic fraction) was concentrated by acetone precipitation. The pellet was resuspended in 50 mM Tris, pH 8, containing 87 g/ml phenylmethylsulfonyl fluoride and then sonicated. Unbroken cells were pelleted and discarded. The membrane and cytoplasmic fractions were separated by treating the lysate with KCl to a final concentration of 0.2 M and centrifugation at 262,000 ϫ g. The supernatant (cytoplasmic fraction) was concentrated by acetone precipitation. The pellet (membrane fraction) was resuspended in a buffer of 20 mM Tris, pH 7.4, 150 mM KCl, 10% glycerol, and 0.05 mM EDTA. The samples were boiled in denaturing buffer and quantified by visual inspection of a Coomassie-stained SDS-polyacrylamide gel. Approximately equal amounts of protein were loaded in each lane of a 10% Laemmli gel (20), run at 200 V, and then transferred to a polyvinylidene difluoride membrane. Blots were blocked in a solution of 5% milk in TBST (Tris-buffered saline and Tween 20) and then incubated with primary antibodies. The N-terminal VirB4 antibody is a polyclonal rabbit antibody to the N-terminal 103 amino acid residues of VirB4 (23). The C-terminal VirB4 antibody is a polyclonal rabbit antibody to the C-terminal 157 amino acid residues of VirB4 (23), which was purified with acetone-powdered (24) A. tumefaciens cells that had been cured of their Ti plasmid.
Transient Expression Assay-Nicotiana benthamiana plants were greenhouse grown for five to seven weeks. In any one experiment, plants from the same cohort were chosen when leaves were ϳ2 cm in diameter. Single colonies of A. tumefaciens were grown for 2 days at 28°C in 2 ml of liquid LB under selection. 1 ml of the overnight cultures was plated onto LB plates with antibiotics and grown for 2 days at 28°C. Cells were resuspended and diluted to a final concentration of A 600 ϭ 0.6 in a solution containing 10 mM MgCl 2 , 10 mM MES, pH 5.5, and 200 M acetosyringone. The cells were incubated at 19°C for 3 h prior to infiltration. Six 26-gauge needle holes were made on each N. benthamiana leaf and infiltrated with ϳ400 l of suspension of A. tumefaciens/leaf as described previously (19). The GFP-T-DNA plasmid was a gift from the Staskawicz laboratory and is a derivative of pMD1 (19) modified to express GFP under the control of the cauliflower mosaic virus 35 S promoter. Plants were placed at 19°C for 12-24 h and then transferred to a 22°C incubator with a 16-h light and 8-h dark cycle. The plants were assayed for GFP expression at different times during a 24 -72-h period post-infiltration. 10 -18 inoculations were examined/construct/experiment, and experiments were repeated three times.
Microscopy and Image Processing-Plant leaves were visualized on a Zeiss Axiophot epifluorescence microscope with a GFP filter set (Chroma Technology Corporation, set number 41017). Half-second exposures were captured through a Roper Quantix camera by the IP Laboratory program (Scanalytics). The color table and brightness levels were adjusted in Adobe Photoshop in a consistent manner across all exposures and experiments.
Homology Modeling-We used the model of the monomer of the C terminus of VirB4 (8) to create an improved model of the VirB4 hexamer. As previously, each subunit of the VirB4 monomer was aligned to the corresponding subunit of the solved TrwB hexamer structure (Protein Data Bank code 1E9R (11)). However, instead of using the program Modeler (25), here we used the Fit function of the Deep View program (26), which led to an emphasis toward resolved residues while de-emphasizing residues with no known coordinates. Models of VirB4 can be accessed at the phylogenomics.berkeley.edu web site.
The VirB11 protein sequence from A. tumefaciens was aligned with the HP0525 protein sequence from H. pylori using the alignment given in Yeo et al. (10). The Modeler program (25) was used to build a homology model of two VirB11 subunits based on the structure of two subunits of HP0525 found in the asymmetric unit of the crystal structure of HP0525 bound to ADP (Protein Data Bank (PDB) code 1G6O; chains A and B (10)). The hexamer was then assembled by applying crystallographic 3-fold symmetry operators using the program O (27). Molecular images were produced using the UCSF Chimera package from the Resource for Biocomputing, Visualization, and Informatics at the University of California, San Francisco (28).

RESULTS
The topology of the C terminus of VirB4 was defined using several experimental approaches. We first present results defining subdomains of protein-protein interactions between VirB4 and other components of the T4SS. As described below, the C terminus of VirB4 interacts with periplasmic domains of several components of the T4SS.
The C-terminal Domain of VirB4 Interacts with Periplasmic Proteins-To begin to address VirB4 topology, we defined interactions between the C terminus of VirB4 (residues 329 -650) and other components of the T4SS using the yeast twohybrid system. The C-terminal portion of VirB4 is of particular interest, because it contains a clear structural domain (the TrwB-like domain) with Walker A and B boxes characteristic of many ATPases. Fig. 1 summarizes VirB1-VirB4, VirB8-VirB4, VirB10-VirB4, and VirB11-VirB4 two-hybrid interactions. We previously reported interactions between VirB4 and VirB1, VirB8, VirB10, and VirB11 (17); however, (except for the VirB4-VirB8 interaction) the data presented here are derived from new two-hybrid studies that define smaller bait and prey and lead to a finer map of the regions of interaction.
Supplemental Fig. 1 describes our two-hybrid analyses using a high resolution prey library corresponding to fusions after every nucleotide of the virB operon and virD4. The high representation of the library defines positive interactions against a highly significant population of negative interactions as controls. The data in Fig. 1A display regions of interaction between two proteins as swaths of lavender and green. When two independent interactions overlap, they are shown as darker swaths of color. These latter regions of overlap then define the minimal regions of interaction between two proteins; these minimal regions are summarized in Fig. 1B.
VirB1 has been proposed (29) and now demonstrated to have lytic transglycosylase activity (30) that likely functions in the periplasm to cleave the peptidoglycan layer, allowing localized establishment of other periplasmic T4SS proteins. In support, 1) VirB1 has a signal sequence for transport across the inner membrane via the general secretory pathway, and removal of this sequence abolishes complementation of a virB1 deletion mutant (18); 2) alkaline phosphatase fusions along the length of the VirB1 coding sequence are active (31,32), and 3) the VirB1 C-terminal portion VirB1* is exported to the cell surface after periplasmic processing (18,33). Previous two-hybrid data show multiple VirB1-VirB8 and VirB1-VirB10 interactions (17) consistent with a periplasmic localization for this subgroup of proteins. Here we have demonstrated reciprocal interactions between VirB1-VirB4 that support periplasmic exposure of the C terminus of VirB4 (Fig. 1A, middle).
The interactions of the C terminus of VirB4 with VirB8 and VirB10 further strongly suggest that this region of the VirB4 protein resides in the periplasm. The region of VirB8 interaction (residues 60 -237) represents the entire periplasmic domain of VirB8 (9, 31, 32, 34 -36). We have also found that VirB4 residues 198 -401 interact with the same periplasmic domain of VirB8 (17); thus, the minimal VirB8 interaction domain occurs between amino acids 329 and 401 of VirB4 (Fig.  1A, top). Though this region of VirB4 (residues 329 -401) has no known structure or function, its interaction with VirB8 supports a periplasmic localization of this portion of the VirB4 protein.
Further evidence for a periplasmic location of the C terminus of VirB4 is its interaction with the N terminus of VirB11 (Fig.  1A, bottom). The orientation of the solved structure of the VirB11 homolog HP0525 from H. pylori (10) predicts that VirB11 resides in the inner membrane with its N terminus facing the periplasm and its C terminus facing the cytoplasm. The Agrobacterium VirB11-VirB4 interaction is analyzed in detail below.
Homology Model of VirB11 Based on H. pylori HP0525-To test whether two-hybrid data can be used to determine how full-length T4SS proteins interact in the cell, we analyzed twohybrid interacting domains and structural data to predict the three-dimensional relationship between VirB4 and VirB11. Because the structures of VirB4 and VirB11 are unknown, we first needed to build homology models of these proteins. A structure of VirB11 was predicted using bioinformatics approaches similar to those utilized to create a homology model of the VirB4 C terminus (8).
Our VirB11 homology model is based on HP0525 from H. pylori, which shares 31% amino acid identity (and 52% similarity over 301 aligned residues) with A. tumefaciens VirB11. Two subunits of HP0525 were found in the asymmetric unit of its crystal structure (10) and form the basis for our structural comparison with VirB11. Two monomers of our predicted VirB11 structure ( Fig. 2A, right) are highly similar to two monomers of HP0525 ( Fig. 2A, left). Compared with our VirB4 model, the VirB11 model was easier to construct because of the higher degree of sequence identity and similarity between VirB11 and HP0525. Based on the relatedness in sequence and function with HP0525, VirB11 is predicted to form a hexamer (Fig.  2B). If VirB11 adopts a similar membrane orientation as HP0525 (10), then its N-terminal domain would be embedded in the inner membrane facing the periplasm (P) with its C-terminal domain exposed to the cytoplasm (C) (Fig. 2C, bottom left).
In addition to our new VirB11 model, we improved our hexameric model of the VirB4 C terminus (residues 425-789) (8). This new model emphasizes resolved residues leading to a more compact and symmetrical three-dimensional structure (Fig. 2C, upper right and left images).
Faces of VirB4 and VirB11 Hexamers Interact-We mapped VirB4-VirB11 two-hybrid interactions onto the three-dimensional models of VirB4 and VirB11 to assess whether two-hybrid interactions include residues that are accessible for protein-protein interaction. In the yeast two-hybrid assay, the first third of VirB11 (residues 1-106) retrieved a relatively small region of VirB4 (residues 443-564) (Fig. 1A, bottom). These domains of interaction between VirB4 (residues 443-564) and VirB11 (residues 1-106) were mapped in teal and blue, respectively, onto the ribbon (side view) and space filling the structural models of VirB4 and VirB11 (Fig. 2C). Strikingly, the regions of interaction between VirB4 (Fig 2C, top, in teal) and VirB11 (bottom, in blue) map to a single face of each hexamer. Knowing the orientation of VirB11 in the inner membrane allowed us to orient VirB4. By analogy to the topology HP0525 (10), the N terminus of VirB11 faces the periplasm (toward the exterior of the cell). Therefore, the C terminus of VirB4 must reside in the periplasm with the surface that interacts with the N terminus of VirB11 facing the interior of the cell. These results demonstrate the complementary utility of twohybrid and structural data.
Thus, yeast two-hybrid interactions between VirB4 and periplasmic T4SS components (VirB1, VirB8, VirB10) combined with the mapping of VirB4-VirB11 interaction domains onto structural models of VirB4 and VirB11 consistently support a periplasmic location of the C terminus of VirB4. Biochemical studies below confirm this assignment.
VirB4 C-terminal Fragment Is Found in the Periplasm-To address directly where in the cell the C terminus of VirB4 protein is located, cell fractionation and Western blotting were performed on Agrobacterium cells expressing both the wildtype VirB4 and a HA-tagged VirB4 C terminus (residues 432-789). We monitored fractions using antibodies to three differ- Regions of interaction between two proteins are shown by swaths of lavender and green. The darker swaths occur when two interactions overlap; these reflect the minimal zones of two-hybrid interaction. B, regions of VirB4 that interacted with other VirB proteins are mapped below a linear representation of fulllength VirB4. In parentheses are the VirB4 residues that interacted with the named VirB protein; for example, VirB1-(432-522) indicates that VirB4 (residues 432-522) interacts with VirB1. DECEMBER 8, 2006 • VOLUME 281 • NUMBER 49 ent peptides, VirB4 N-terminal residues 1-103, VirB4 C-terminal residues 632-789, and the HA epitope (Fig. 3). The N-terminal antibody detected the full-length VirB4 protein in total cells and in the membrane fraction (Fig. 3A). The C-ter-minal antibody, however, detected two different VirB4 species whose sizes correspond to the full-length protein and the C-terminal fragment (Fig. 3B). The full-length protein again was detected in the total cells and in the membrane fraction (Fig. 3B, upper arrowheads). The C-terminal fragment was detected in the total cells and in the membrane fraction as well, but unlike the full-length protein, it was found also in the periplasm fraction (Fig. 3B,  lower arrowheads). The HA antibody confirmed this latter result by again detecting the C-terminal VirB4 fragment in total cells as well as in the periplasm and membrane fractions (Fig. 3C). We never observed VirB4 (or its C-terminal fragment) in the cytoplasm fraction under any conditions. In control experiments, VirB9, a core periplasmic protein of the T4SS, was found in total cells, periplasm, and membrane fractions (Fig. 3D), whereas the T4SS substrate protein VirE2 was detected in total cells, cytoplasm, and membrane fractions (data not shown). Taken together, these results indicate that the fulllength VirB4 protein is anchored in the membrane, possibly by its N-terminal domain, whereas the C-terminal domain is located in the periplasm.

Fragments of VirB4 Exert a Dominant Negative Effect on T-DNA
Transfer-A fragment of VirB4 starting at residue 432 defined the smallest region of VirB4 interaction with itself (data not shown) (17) and with VirB1, whereas a fragment starting at residue 443 was the smallest region interacting with VirB11 (Fig. 1). Significantly, the canonical Walker A box sits precisely between amino acids 432 and 443, suggesting that there is a distinct differentiation in T4SS component interaction between the region C-terminal of the Walker A box and one including the Walker A box. With this in mind, we assayed the function of two VirB4 fragments, one from the Walker A box through to the end of VirB4 (residues 432-789) and one just C-terminal of the Walker A box through to the end of VirB4 (residues 443-789).

VirB4 Topology
To promote uniformity in detecting VirB4 deletion derivatives, an HA tag was added to constructs of the VirB4 regions cloned into a vir-inducible plasmid with an Agrobacterium-specific ribosome binding site. Fig. 4 shows steady-state levels of synthesis of HA-tagged full-length VirB4 and the two C-terminal VirB4 constructs. In tumorigenesis assays, VirB4 and VirB4HA equally complement a virB4 deletion, whereas the C-terminal fragments do not (data not shown). Tagged versions of full-length VirB4 (residues 1-789) and the two C-terminal VirB4 fragments (residues 432-789 and 443-789) were introduced in trans to an Agrobacterium C58 derivative containing a plasmid with T-DNA borders flanking the coding sequences of soluble GFP under the control of a cauliflower mosaic virus 35 S promoter.
A "wild-type" phenotype shows a high level of transient GFP expression in N. benthamiana leaves, whereas a reduction of GFP expression results from a failure in T-DNA transfer. Protein domains that are sufficient for interaction but cannot complement the wild-type protein likely disrupt VirB/T4SS transporter function by binding to some but not all of the components necessary for an intact transporter. In effect, the non-productive protein fragments titrate out productive components, leading to incomplete and unstable transporters. A reduction of in vivo function due to non-productive physical interactions between transporter components would confirm our two-hybrid results that indicate the C-terminal domain of VirB4 is a key region for protein-protein interactions. Each construct was assessed for function at least 10 times in three separate experiments, and representative data are shown in Fig. 5.
Full-length VirB4 expressed in trans to the wild-type virB operon had almost as high a level of GFP expression as the control strains (Fig. 5A versus D and E). This reproducible difference between VirB4 merodiploid and wild-type VirB4 levels may be due to the sensitivity of the system to the correct stoichiometry of all T4SS components. Note, we never observed expression of GFP in vir-induced Agrobacterium; this control is also evidenced by the fact that constructs that poison the transporter do not result in high GFP expression in planta (see below). Furthermore, lowered T-DNA transfer was not due to    DECEMBER 8, 2006 • VOLUME 281 • NUMBER 49 an overall decrease in T4SS component protein expression or stability, as monitored by unchanged levels of VirB1 in merodiploid strains by Western blotting (data not shown).

VirB4 Topology
Expression of VirB4 residues 432-789 containing the Walker A box reduced GFP transfer, and expression of the VirB4-(443-789) fragment without the Walker A box greatly reduced GFP transfer (Fig. 5, B and C). These latter two results suggest that the interaction of these two VirB4 fragments with wild-type T4SS components leads to poisoning of T-DNA transfer. The dominant negative effect of the construct lacking the Walker A box (Fig. 5C) is stronger than that of the construct containing the Walker A box (Fig. 5B). This result is especially notable given that the expression of the construct without the Walker A box (Fig. 4, lanes 5 and 6) is lower than that of the construct with the Walker A box (Fig. 4, lanes 3 and 4). The stronger poisoning by the Walker A-minus construct makes sense, as this construct is likely to be less active and thus exhibit stronger interference than the construct that still contains the Walker A box.

DISCUSSION
The results presented address the location and topology of the C terminus of VirB4 in a framework of several other Agrobacterium T4SS components. Our findings support a T4SS structure wherein the hexameric VirB4 C-terminal domain resides in the periplasm and interacts with the VirB11 hexamer embedded in the inner membrane. This conclusion supercedes our previous T4SS model (8), which predicted a cytoplasmic position for VirB4 based solely on its homology to the type IV coupling protein VirD4.
VirB4 topology based on sequence analyses predicts from zero (8) to four transmembrane domains (38), whereas alkaline phosphatase studies report either no alkaline phosphatase transposon insertions in VirB4 (31,32) or periplasmic exposure of the region around residue 450 (15). Our VirB4 two-hybrid interactions with the periplasmic domains of VirB8 and VirB10 and our fractionation and Western blot data strongly indicate periplasmic exposure of the VirB4 C-terminal hexameric domain. This corresponds with the above data on the alkaline phosphatase fusion at residue 450. In addition, studies with detergent-solubilized Brucella suis T4SSs demonstrate that the VirB8 periplasmic domain and a VirB4 hexamer exist in the same complex (9). Our VirB4 C-terminal periplasmic placement is further supported by degradation of VirB4 in Agrobacterium spheroplasts treated with proteinase K (14,15). Finally, an independent and elegant transfer DNA immunoprecipitation assay, which defines the order of interaction between the T strand and components of the T4SS, suggests that the T strand first interacts with VirD4 and VirB11 in the cytoplasm (14); these authors have additional genetic data that place the requirement for VirB4 downstream of these initial contacts. These latter data are consistent with our assignment of the C-terminal ATPase domain of VirB4 to the periplasm.
The present studies do not address the topology of the N terminus of VirB4. Early work suggested a possible periplasmic location of a small N-terminal region of VirB4 (residues 76 -84) predominantly based on findings in E. coli (15); however, the same authors demonstrate that a region around amino acid 237 is cytoplasmic in Agrobacterium. We favor an N-terminal cyto-plasmic assignment in our model described below (Fig. 6). Potentially, predictions of membrane topology based on fusions to other proteins may not always represent the native state. This caution is especially relevant to a large protein such as VirB4 (789 residues), which may not fold properly when fused to alkaline phosphatase or ␤-galactosidase. Thus, biochemical fractionations of native proteins described herein, as well as by other groups (9,13,37), may more accurately reflect their native localization.
We offer one further point of discussion. Proteins may arrive in the periplasm by three mechanisms. The first two, a membranespanning tether that anchors one or more domains of the protein in the periplasm or cleavage of an N-terminal signal peptide, are well established. However, there is a third possibility. A protein may arrive or orient itself in the periplasm by means of interaction with either membrane-anchored or periplasmic proteins. Thus, the C terminus of VirB4 may arrive in the periplasm by interaction with VirB11, as well as VirB8 and VirB10.
The periplasmic localization of the VirB4 ATP binding domain poses a potential question for how this domain gains accesstoATP.Weproposetwoscenarios.Inthefirst,theATPasedependent function and/or conformational change occurs in the cytoplasm prior to arrival of the C-terminal portion of VirB4 at its periplasmic destination. In the second, VirB4 is active solely in the periplasmic space, and ATP is transferred through the central channel of the VirB11 hexameric transmembrane complex to the periplasmic hexameric VirB4 C terminus, embedded within the tightly knit T4SS complex (Fig. 6). In this scenario, ATP would not need to cross a membrane per se. As the Agrobacterium T4SS transports DNA, it may also transport nucleotides with similar biophysical properties to nucleic acid polymers. For example, the VirB11 hexameric ATPase likely interacts with ATP during transfer of the T strand DNA through its pore (14). Thus, we favor a model where ATP is also transferred through this pore directly to VirB4, thereby enabling VirB4 activity in the periplasm.
VirB11 associates tightly with the inner membrane; 70% of VirB11 fractionates with the inner membrane, and 30% of VirB11 is found in the cytoplasm (39). Proteinase K susceptibil-

VirB4 Topology
ity studies demonstrate that VirB11 is not exposed to the periplasm (15). Our two-hybrid data superimposed onto our VirB4 and VirB11 structural models show one face of the VirB4 hexameric domain interacts with the N-terminal face (residues 1-106) of the VirB11 hexamer. This interaction would then "cover" any potential periplasmic exposure of VirB11 and may explain why VirB11 is immune to protease digestion. The N-terminal 106 residues of HP0525, the H. pylori template for our VirB11 homology model, co-crystallized with a polyethylene glycol nonamer, which indicates that the N-terminal ring of HP0525 (and likely VirB11) is embedded in the phospholipid bilayer of the inner membrane (10). VirB4-VirB11 complex formation is confirmed also by co-immunoprecipitation (13). The topology, protease susceptibility, crystallization, fractionation, and two-hybrid data support our proposal that the C-terminal periplasmic VirB4 hexamer stacks on top of the inner membrane-embedded VirB11 hexamer. Barrel-forming stacked ATPases exist in other biological systems, including the ClpP-ClpA and ClpP-ClpX (protease-ATPase complexes) (40) and the p97 AAA (ATPases associated with diverse cellular activities) ATPase and its homologs (41).
VirB4 exists as a highly conserved component in T4SSs (38), indicating its evolutionary and functional importance. Deletions of VirB4 abolish both T-DNA tumor and T pilus formation (42,43), whereas some VirB4 Walker A box mutations lead to loss of T-DNA transfer but maintenance of T pilus formation (9), indicating that VirB4 contributes differentially to pilus formation and DNA transfer.
It is striking that a large number of physical interactions exist between the C terminus of VirB4 and four components of the T4SS (VirB1, VirB8, VirB10, and VirB11). These data imply a highly dynamic functional and/or an integral structural role of VirB4. The model in Fig. 6 illustrates how the VirB4 C terminus may interact with the periplasmic domains of VirB8 and VirB10, the inner membrane face of VirB11, and cytoplasmic VirD4. Our model incorporates previously demonstrated VirB8-VirB10, VirB8-VirB11, and VirB10-VirB11 yeast two-hybrid interactions (17,44) and a VirD4-VirB10 interaction determined by co-immunoprecipitation in Agrobacterium (13). We hypothesize that the C terminus of VirB4 interacts with VirB8 and VirB10 in the periplasm, whereas the N terminus of VirB4 interacts with VirD4. In support, VirD4, VirB4, and VirB11 are co-immunoprecipitated (13) and the N terminus of VirB4 interacts with itself (7) in the bacterial cytoplasm.