The ATPase activity of BfpD is greatly enhanced by zinc and allosteric interactions with other Bfp proteins.

Type IV pilus biogenesis, protein secretion, DNA transfer, and filamentous phage morphogenesis systems are thought to possess similar architectures and mechanisms. These multiprotein complexes include members of the PulE superfamily of putative NTPases that have extensive sequence similarity and probably similar functions as the energizers of macromolecular transport. We purified the PulE homologue BfpD of the enteropathogenic Escherichia coli bundle-forming pilus (BFP) biogenesis machine and characterized its ATPase activity, providing new insights into its mode of action. Numerous techniques revealed that BfpD forms hexamers in the presence of nucleotide. Hexameric BfpD displayed weak ATPase activity. We previously demonstrated that the N termini of membrane proteins BfpC and BfpE recruit BfpD to the cytoplasmic membrane. Here, we identified two BfpD-binding sites, BfpE(39-76) and BfpE(77-114), in the N terminus of BfpE using a yeast two-hybrid system. Isothermal titration calorimetry and protease sensitivity assays showed that hexameric BfpD-ATPgammaS binds to BfpE(77-114), whereas hexameric BfpD-ADP binds to BfpE(39-76). Interestingly, the N terminus of BfpC and BfpE(77-114) together increased the ATPase activity of hexameric BfpD over 1200-fold to a V(max) of 75.3 mumol of P(i) min(-1) mg(-1), which exceeds by over 1200-fold the activity of other PulE family members. This augmented activity occurred only in the presence of Zn(2+). We conclude that allosteric interactions between BfpD and BfpC and BfpE dramatically stimulate its ATPase activity. The differential nucleotide-dependent binding of hexameric BfpD to BfpE(39-76) and BfpE(77-114) suggests a model for the mechanism by which BfpD transduces mechanical energy to the biogenesis machine.

Type IV pili (Tfps) 1 are filamentous appendages found on the surface of Gram-negative bacteria. Tfps are major virulence factors for many important human pathogens including Vibrio cholerae (1), Pseudomonas aeruginosa (2), Neisseria gonorrhoeae (3) and enteropathogenic Escherichia coli (EPEC (4)), mediating attachment to host cells and formation of microcolonies. These filaments are also involved in motility (5)(6)(7)(8), biofilm formation (9), and horizontal gene transfer (10 -13). Tfps are predominantly helical polymers of the pilin subunit that are defined by their shared structural and biochemical features and a highly conserved biogenesis machinery (14,15). Formation of Tfps is a complex process requiring many proteins besides the pilin protein, including the prepilin peptidase (which processes prepilin into pilin), prepilin-like proteins, proteins with nucleotide-binding motifs, and cytoplasmic membrane and outer membrane proteins. Evidence suggests that these proteins compose a sophisticated molecular machine that allows pilin polymerization, filament stabilization, and surface translocation. Most of the proteins involved in the assembly of Tfps have extensive sequence similarity with proteins of protein secretion, DNA transfer, and filamentous phage transport systems (16,17), suggesting that these systems possess similar structures and modes of action. Thus, Tfp biogenesis is a useful model for a variety of systems that transport biological macromolecules across membranes.
EPEC is a major cause of severe infantile diarrhea and a leading cause of infant mortality in developing countries (18). The BFP of EPEC is a particularly good system to study Tfp biogenesis because all of the genes necessary for BFP biogenesis are known. The 14-gene bfp operon contains 13 genes required for BFP biogenesis and function (4, 19 -26). The bfpA gene encodes prebundlin (27,28), which is processed into bundlin (the pilin protein) by the protein product of bfpP (the prepilin peptidase (19)). The precise functions of the other components of the biogenesis machinery are unknown. The bfpB gene encodes an outer membrane lipoprotein of the secretin family (20) that is hypothesized to form a pore in the outer membrane through which the BFP is extruded. The bfpD and bfpF genes encode putative cytoplasmic nucleotide-binding proteins BfpD and BfpF (22,23), which are thought to provide energy for BFP biogenesis and retraction, respectively. The bfpE gene encodes a polytopic cytoplasmic membrane protein with four transmembrane domains, a large cytoplasmic N terminus, two periplasmic domains, and a small cytoplasmic domain (26). We recently demonstrated that the N terminus of BfpE interacts with BfpD and the cytoplasmic N terminus of the bitopic cytoplasmic membrane protein BfpC, which also interacts with BfpD (29). Together, BfpC and BfpE recruit BfpD to the membrane (29).
BfpD is a member of the PulE superfamily of putative nucleotide-binding proteins involved in bacterial type II and IV secretion, Tfp assembly, natural competence, and assembly of archeal flagellae (17). These proteins contain several conserved sequence motifs, including Walker A and B boxes involved in ATP binding and hydrolysis (30). However, ATPase activity has been confirmed for only a few of them, mostly from type IV secretion systems (31)(32)(33)(34)(35)(36)(37)(38). Furthermore, this ATPase activity has been relatively weak, between 0.71 and 61 nmol of inorganic phosphate (P i ) released min Ϫ1 mg Ϫ1 of protein. Mutation of key residues in the Walker A box of PulE superfamily members abolishes activity of the cognate systems, for example, PulE (39,40), EpsE (41), OutE (42), PilB (43), and BfpD (4). Therefore, these proteins most likely share the function of supplying energy from ATP hydrolysis to drive organellar assembly or substrate translocation.
The characterization of the ATPase activity of BfpD is crucial for understanding its role in BFP biogenesis and mechanism of action. In this paper we describe a detailed study of the ATPase activity of BfpD in which we address its activity in the context of the BFP biogenesis machine, i.e. in the presence of components of the machinery with which it interacts. The results of our studies suggest a model for conversion of chemical energy into mechanical energy to power extraction of bundlin from the cytoplasmic membrane for assembly into BFP filaments.
The gel filtration column was calibrated with blue dextran (2,000 kDa), thyroglobulin (669 kDa), ferritin (440 kDa), catalase (232 kDa), aldolase (158 kDa), bovine serum albumin (67 kDa), ovalbumin (43 kDa), chymotrypsinogen A (25 kDa), and ribonuclease A (13.7 kDa; Amersham Biosciences). The Stokes radius (R) of each form of BfpD was determined from a plot of the partition coefficient (K av ) versus log R of standards as previously described (45). K av was calculated using the equation where V e is the elution volume, V o is the void volume, and V t is the bed volume.

Zonal Sedimentation Analysis
Purified BfpD (20 g) was applied to the top of a 2.5-ml 20 -40% sucrose gradient formed in 20 mM Tris-HCl buffer, pH 7.6, 100 mM NaCl, 10 mM MgCl 2 , with 1 mM ATP as appropriate. Gradients were centrifuged at 245,000 ϫ g for 12 h at 4°C. Fractions (100 l) were collected, precipitated with trichloroacetic acid, and subjected to SDS-PAGE and immunoblotting. Molecular mass standards run in parallel gradients were visualized by SDS-PAGE and Coomassie staining. Sedimentation coefficient (s 20,w ) values were determined by sedimentation relative to protein standards.
Molecular mass values were calculated from the equation molecular mass ϭ 6Nas 20,w 20,w /1 Ϫ 20,w in which N is Avogadro's number, a is the Stokes radius, 20,w is the viscosity of water at 20°C, is the partial specific volume of the protein calculated from its amino acid composition, and 20,w is the density of water at 20°C as described (45).

Electron Microscopy
Negative Staining-Carbon-coated copper grids were floated on 5-l drops of 100 g ml Ϫ1 BfpD in 20 mM Tris-HCl, pH 7.5, 100 mM NaCl, 0.1 mM EDTA, 10 mM MgCl 2 , 1 mM ATP␥S. After 5 min the sample was wicked off, and grids were floated consecutively on 3 drops (60 l) of 3% uranyl acetate, with blotting in between. Grids were floated on the third drop for 1 min, then blotted and air-dried. Images were recorded on a Philips/FEI CM100 electron microscope operating at 100 keV in low dose mode at ϫ52,000 magnification and a defocus of 500 nm.
Cryo-electron microscopy-Five-l drops of BfpD were applied to Quantifoil holey grids (Quantifoil Micro Tools GmbH) that had been glow-discharged in the presence of amyl amine. After 1 min the grids were blotted for 2.5 s and then plunged into an ethane slush cooled with liquid nitrogen, using the Vitrobot (FEI). Grids were transferred to a Gatan 626 cold stage (Gatan), and images were recorded on a Philips/ FEI CM120 electron microscope operating at 120 keV in low dose mode at ϫ50,000 magnification and a defocus of 1 m.

ATPase Assay
The ATPase activity of hexameric BfpD, purified in the presence of ADP and the absence of MgCl 2 , was measured as the P i released by the hydrolysis of ATP using a colorimetric malachite green assay (46). ATPase activity was determined from the initial rate of P i release, which was linear with time over a period of at least 1 h (Fig. 2B). Assays were carried out in 40 mM Tris-HCl, pH 7.0, 150 mM NaCl, 10 mM ZnCl 2 , and 5% glycerol at 37°C in a 50-l reaction volume. The metal dependence of the ATPase activity of BfpD was examined by substituting the ZnCl 2 in the reaction buffer with MnCl 2 , CaCl 2 , CuCl 2 , MgCl 2 , FeCl 2 , CoCl 2 , CdCl 2 , and NiCl 2 . In experiments to determine the influence of BfpC and BfpE on ATPase activity, different combinations of the N terminus of BfpC (6:1 molar ratio of BfpC to BfpD), BfpE 39 -76 (6:1 molar ratio of BfpE to BfpD), and BfpE 77-114 (3:1 molar ratio) were incubated with hexameric BfpD at 37°C for 30 min prior to initiation of the reactions.
Reactions were initiated by the addition of ATP to a 5 mM final concentration and allowed to proceed for 0, 5, 10, 15, 20, 25, and 30 min. For substrate saturation experiments, ATP was varied from 0.1 to 10 mM. Reactions were terminated and released P i quantified by the addition of 800 l of freshly prepared color reagent (0.034% malachite green, 0.1% Triton X-100, and 1.05% ammonium molybdate in 1 M HCl; filtered to remove insoluble material). After 1 min at room temperature, color development was stopped by addition of 100 l of 34% citric acid. Samples were incubated at room temperature for 30 min, and then the absorbance at 660 nm was measured. Values were calibrated against KH 2 PO 4 standards and corrected for P i released in the absence of BfpD and the absence of ATP. Data from substrate saturation experiments, where ATP concentration was varied, were fit by nonlinear regression to the Michaelis-Menten equation with Origin software (Microcal) and the parameters K m and V max were calculated. Similar values were obtained using Lineweaver-Burk and Eadie-Hofstee plots.

Yeast Two-hybrid Experiments
We used the Matchmaker GAL4 two-hybrid system 3 (Clontech) following the protocol suggested by the manufacturer. The cloning of DNA fragments encoding BfpD and the N terminus of BfpE into pG-BKT7 (GAL4 DNA-binding domain vector) and pGADT7 (GAL4 activation domain vector) has been described previously (29). In addition, PCR-amplified fragments encoding amino acids 1-38, 39 -76, 77-114, 1-76, and 39 -114 of the N terminus of BfpE (for primers see Table I) were cloned into these vectors. Saccharomyces cerevisiae strain AH109 was co-transformed with these constructs by the lithium acetate procedure and a minimum of 30 transformants for each vector combination were used to test protein interactions. The transformants were plated on medium containing X-␣-Gal and lacking tryptophan, leucine, histidine, and adenine and assayed for ␣-galactosidase activity to detect and quantify interactions.

Isothermal Titration Calorimetry
Isothermal titration calorimetry (ITC) experiments were carried out using a VP-ITC Microcalorimeter (Microcal). Peptides of BfpE 39 -76 (LYKFTSDEGRKKDNPDAFALQRWLIAVRNGKTLAEAMR) and BfpE 77-114 (GWVPFDELSIISAGEISGNVHQALDDIIYMNDTKKKVK), synthesized and high performance liquid chromatography-purified by Biopeptide, were titrated with hexameric BfpD purified in the presence of ATP␥S or ADP. Peptide corresponding to BfpE 39 -76 (0.018 mM) or BfpE 77-114 (0.009 mM) in 20 mM Tris-HCl buffer, pH 7.6, containing 100 mM NaCl, 10 mM MgCl 2 and either 1 mM ATP␥S or 1 mM ADP was loaded into the sample cell. BfpD (0.12 mM), in the same buffer as the peptide, was loaded into the injection syringe and titrated into the sample cell. A typical titration series involved 35 injections of 3 l of titrant. The baseline was allowed to stabilize for 210 s between injections to permit adequate time for reaction and equilibration. All experiments were performed at 37°C and the cell was stirred continuously at 280 rpm. For control experiments to determine the heats of dilution of BfpD, the peptides were substituted by buffer. The heat produced for each injection of BfpD into peptide or buffer was obtained by integration of the area under each peak of the titration plots with respect to time. The heats of reaction were calculated by subtraction of the integrated heats of dilution of BfpD from the heats corresponding to the injection of BfpD into peptide. Titration data, corrected for heat of dilution, were analyzed with Microcal Origin using a single-site model.

Fluorescence Measurements
The fluorescent nucleotide analogues 2Ј,3Ј-O-trinitrophenyl (TNP)-ATP and TNP-ADP were used to study nucleotide binding to BfpD. Experiments were performed at 37°C using a spectrofluorometer with bandwidths of 4 nm. Monomeric BfpD (0.72 M) was incubated for 30 min with various concentrations of TNP-nucleotide in 20 mM Tris-HCl buffer, pH 7.6, containing 100 mM NaCl and 10 mM MgCl 2 . The fluorescence emission was measured at 535 nm upon excitation at 408 nm (total fluorescence). Enhancement of fluorescence (⌬F) was calculated as the difference between total fluorescence and the fluorescence measured in incubations of TNP-nucleotide in buffer. The enhanced fluorescence in the presence of BfpD, proportional to the quantity of BfpDbound TNP-nucleotide, was plotted as a function of TNP-nucleotide concentration. Data were fitted to the equation where ⌬F is the change in fluorescence intensity of TNPnucleotide at a concentration [S], ⌬F max is the maximum change in fluorescence intensity, and K d is the dissociation constant. Fitting was performed using Origin and values for ⌬F max and K d were obtained.
The stoichiometry of TNP-nucleotide binding to BfpD was derived using the method of Stinson and Holbrook (47). Briefly, 1/(1 Ϫ ␣) was plotted against [S]/␣, where ␣ is the fractional saturation of the TNPnucleotide-binding sites (⌬F/⌬F max ). A straight line with a slope of 1/K d and an x-intercept of the concentration of TNP-nucleotide-binding sites was obtained. The stoichiometry of binding was then calculated by dividing the x-intercept with the BfpD concentration.

Proteinase K Digestion
All protease digestions were carried out at room temperature in 20 mM Tris-HCl buffer, pH 7.6, 100 mM NaCl, 10 mM MgCl 2 , containing 1 mM ATP␥S or 1 mM ADP as appropriate. Hexameric BfpD (5 g), purified in the presence of ATP␥S or ADP, was preincubated with BfpE 39 -76 (5 g), BfpE 77-114 (5 g), or neither of these peptides at room temperature for 10 min. Proteolysis was initiated by addition of proteinase K to give final concentrations of 0, 0.05, 0.1, and 1 mg ml Ϫ1 and the reaction was allowed to proceed for 30 min. Proteinase K digestion was terminated by adding Laemmli buffer (48) and boiling the samples for 10 min. Finally, proteolytic products were subjected to SDS-PAGE and detected by silver staining.

RESULTS
BfpD Is a Hexamer-BfpD synthesized with a N-terminal His tag in E. coli strain BL21pLysS,pRPA405 was purified to homo-geneity using two chromatographic steps (for BfpD purity, see Fig. 6). BL21pLysS,pRPA405 produces functional BfpD as plasmid pRPA405, encoding the His-tagged BfpD, complements the bfpD mutant UMD926 to restore autoaggregation, a phenotype that correlates with biogenesis of functional BFP (results not shown). The elution profile of BfpD during Sephacryl S-300 gel filtration chromatography, the second purification step, was dependent on ATP and MgCl 2 (Fig. 1A). In the absence of both ATP and MgCl 2 and in the absence of ATP but the presence of MgCl 2 , BfpD eluted as a sharp peak with an estimated molecular mass of 60 kDa (see below). Because this is the molecular mass of BfpD obtained when it is calculated from the amino acid sequence or estimated by SDS-PAGE, we concluded that BfpD exists as a monomer in the absence of ATP. In the presence of ATP and the absence of MgCl 2 and in the presence of both ATP and MgCl 2 , BfpD eluted as two sharp peaks with molecular masses of 60 (the monomeric form) and 369 kDa (see below). The homogeneity of the His-tagged BfpD within the faster peak was confirmed by SDS-PAGE and silver staining, indicating that BfpD forms a homohexamer in the presence of ATP. Interestingly, the majority of BfpD was in the hexameric form (88% hexameric, 12% monomeric) in the presence of both ATP and MgCl 2 , whereas approximately half of the BfpD was in this form (48% hexameric, 52% monomeric) in the presence of ATP and absence of MgCl 2 . Similar results were obtained when ATP was replaced with ATP␥S, TNP-ATP, ADP, and TNP-ADP, suggesting that nucleotide binding, but not its hydrolysis, is required for BfpD hexamerization.
To accurately determine the molecular masses of the two BfpD forms, parameters obtained from gel filtration (Stokes radii) and zonal sedimentation (sedimentation coefficients) were analyzed (Table II). These indicate that the BfpD in the slower peaks was monomeric BfpD (molecular mass ϭ 60 kDa), whereas that in the faster peaks was hexameric (molecular mass ϭ 369 kDa). In addition, DLS analysis of the high molecular mass form of BfpD in the presence of ATP, ATP␥S, or ADP showed that BfpD is a hexamer (Table III).
Electron microscopy studies revealed that hexameric BfpD had a ring-like structure ( Fig. 1, B-E). The electron micrograph of negatively stained BfpD in Fig. 1B shows ring-like structures in which subunits are evident. The diameters of these rings measure ϳ11.5 nm. Furthermore, the rings in the electron micrographs of frozen-hydrated BfpD (Fig. 1, C-E) display 6-fold symmetry, suggesting a hexamer structure consistent with the gel filtration, zonal sedimentation, and DLS results. Taken together, these findings demonstrate that BfpD assembles to form hexamers in the presence of ATP, ATP␥S, TNP-ATP, ADP, and TNP-ADP.
Hexameric BfpD Is an ATPase-We detected the ATPase activity of BfpD using a colorimetric malachite green assay to measure the amount of P i released upon incubation with ATP. BfpD purified in 40 mM Tris-HCl buffer, pH 7.0, containing 150 mM NaCl and 1 mM ADP was used for these ATPase assays. Use of ADP instead of ATP in the BfpD purification produces hexameric BfpD and has the advantage of eliminating high background P i levels. ADP does not inhibit BfpD ATPase ac- The italicized portion of the primer sequence indicates the restriction site present in a PCR product prepared with that primer. b The underlined portion of the sequence anneals to the gene indicated.
tivity as in assays with increasing concentrations of ADP (up to 1000-fold molar excess over the final ADP concentration in the reactions) and BfpD purified in 40 mM Tris-HCl buffer, pH 7.0, containing 150 mM NaCl and 1 mM ATP there was no effect on BfpD activity (data not shown).
In preliminary experiments to optimize the ATPase assay for BfpD, we examined the effects of different reaction conditions on the ATPase activity of hexameric BfpD. In experiments measuring the ATPase activity at different pH values, using 40 mM sodium acetate buffer for pH 5.0 to 6.0, 40 mM Mops for pH 6.5 to 7.0, and 40 mM Tris for pH 7.0 to 9.0, maximal activity occurred at pH 7.0 (data not shown). In additional experiments, the optimal concentration of NaCl was found to be 150 mM (data not shown). Furthermore, ATPase activity required divalent cations; no activity was detected in the absence of divalent cations ( Fig. 2A) and activity was strongly inhibited in the presence of the divalent cation chelator EDTA (data not shown). Interestingly, hexameric BfpD exhibited the highest activity in the presence of Zn 2ϩ (Fig. 2A). Substitution of Zn 2ϩ by Mg 2ϩ and Mn 2ϩ decreased the ATP hydrolyzing activity to 90 and 84%, respectively, of that obtained with Zn 2ϩ , whereas substitution by Ca 2ϩ , Cu 2ϩ , Fe 2ϩ , Co 2ϩ , Cd 2ϩ , and Ni 2ϩ reduced it to less than 10%. Therefore, ATPase assays were performed in 40 mM Tris-HCl, pH 7.0, 150 mM NaCl, 10 mM ZnCl 2 , and 5% glycerol.
When hexameric BfpD protein was incubated with ATP, release of P i occurred in a time-and BfpD concentration-dependent manner (Fig. 2, B and C), indicating that hexameric BfpD has ATPase activity. That this activity was because of BfpD rather than a contaminating ATPase is demonstrated by the fact that preparations obtained by an identical purification scheme from cells containing plasmid lacking bfpD did not display ATPase activity. Interestingly, monomeric BfpD only began to display ATPase activity after 30 min (data not shown). Using gel filtration chromatography, we found that monomeric BfpD hexamerizes in the presence of nucleotide, such as the ATP added to initiate ATP hydrolysis (data not shown). Therefore, we believe that the ATPase activity exhibited by monomeric BfpD was actually that of nascent hexamers. Finally, BfpD did not hydrolyze any other nucleotides (ADP, AMP, GTP, CTP, and UTP) tested in substrate specificity experiments (data not shown).
To determine the Michaelis-Menten constants V max and K m of the ATP hydrolyzing activity of hexameric BfpD, ATPase activity was measured over a range of ATP concentrations, under initial rate conditions. The dependence of the rate of ATP hydrolysis on ATP concentration exhibited typical Michaelis-Menten behavior (Fig. 2D) and showed a linear relationship in Eadie-Hofstee and Lineweaver-Burk plots (data not shown). The V max values obtained from these experiments (Table IV) are consistent with the results shown in Fig. 2, A-C. In conclusion, under optimal conditions, the ATPase activity was calculated as 62.2 Ϯ 2.2 nmol released P i min Ϫ1 mg Ϫ1 of protein (mean Ϯ S.E.). Similar results were obtained using an assay coupling regeneration of ATP to oxidation of NADH (data not shown). Because hexameric, but not monomeric, BfpD displays ATPase activity, it appears that BfpD ATPase activity is activated upon hexamerization.
Either Amino Acids 39 -76 or 77-114 of the N Terminus of BfpE Are Sufficient for Its Interaction with BfpD-We previously demonstrated that BfpD interacts with the N terminus of the polytopic membrane protein BfpE (29). Here we used a yeast two-hybrid system to more precisely define the region in the N terminus of BfpE involved in interacting with BfpD. Full-length BfpD and full-length and fragments of the N terminus of BfpE were expressed as fusions to both the GAL4 DNA-binding (pGBKT7) and activation (pGADT7) domains. S. cerevisiae AH109 was co-transformed with these constructs and subsequently analyzed for transcriptional activation of the three reporter genes, HIS3, ADE2, and MEL1. All of the transformants grew on medium lacking tryptophan and leucine (pG-BKT7 and pGADT7 carry TRP1 and LEU2 selectable marker genes respectively; results not shown), indicating no toxicity as a result of expression of the fusion proteins. Activation of the reporter genes, which is indicative of an interaction between the fusion proteins, was detected as blue colonies on medium containing X-␣-Gal and lacking tryptophan, leucine, histidine, and adenine and quantified by measuring ␣-galactosidase activity. We found reporter gene activation when fusion protein pairs murine p53 and simian virus 40 (Fig. 3). These interactions occurred irrespective of which protein was fused to the activation and which to the DNA-binding domains of GAL4. The remaining transformants did not grow on medium containing X-␣-Gal and lacking tryptophan, leucine, histidine, and adenine and exhibited low levels of ␣-galactosidase activity, similar to the negative control expressing human lamin C and SV40 large T-antigen fusion proteins (Fig. 3). Western analysis showed that the transformants expressed the Bfp fusion proteins at comparable levels (results not shown). Furthermore, S. cerevisiae AH109 cells co-transformed with each construct individually and either "empty" pGBKT7 or pGADT7 grew on the appropriate selective medium, i.e. they did not grow on medium containing X-␣-Gal and lacking tryptophan, leucine, histidine, and adenine and grew as white colonies on medium containing X-␣-Gal and lacking tryptophan and leucine (results not shown). This confirms that the individual fusion proteins cannot activate the reporter genes alone and that the interactions we observed were because of direct interactions between the Bfp domains of the fusion proteins. Thus, it appears that the N terminus of BfpE contains two sites to which BfpD binds, one in BfpE 39 -76 and one in BfpE 77-114 . However, these data do not distinguish whether these sites represent two independent BfpD-binding sites or whether these sites cooperate to form one binding site. Differential Nucleotide-dependent BfpD Affinity for BfpE 39 -76 and BfpE 77-114 -We speculated that BfpD preferentially binds to each of the BfpD-binding sites of the N terminus of BfpE at different stages of catalysis of ATP hydrolysis. To test this hypothesis and to confirm interactions between BfpD and BfpE 39 -76 and BfpE 77-114 , the thermodynamics of the interactions were measured by ITC. For these purposes we synthesized peptides corresponding to BfpE 39 -76 and BfpE 77-114 . Fig. 4 shows calorimetry data for titrations of these peptides with hexameric BfpD in the presence of ATP␥S (a non-hydrolysable ATP analogue) and ADP. Large exothermic enthalpies were observed in titrations of BfpD-ATP␥S into BfpE 77-114 and BfpD-ADP into BfpE 39 -76 . The titration heat was converted to heat mol Ϫ1 as a function of molar ratio and the data were fitted with one set of identical binding site models to give the association constant (K), stoichiometry (molar ratio of BfpD/peptide, n), and enthalpy change (⌬H) upon peptide-BfpD complex   (Fig. 4), even when peptide concentrations were up to 100 times higher (data not shown). Furthermore, we were unable to detect interactions between monomeric (nucleotide-free) BfpD and BfpE 39 -76 or BfpE 77-114 (data not shown). Similarly, no interaction could be detected between monomeric BfpD and the cytoplasmic domain of BfpC (data not shown). In contrast, hexameric BfpD has previously been demonstrated to bind this region of BfpC by ITC (29). To summarize, these results show differential binding of hexameric BfpD to BfpE 39 -76 and BfpE 77-114 depending on the phosphorylation of the nucleotide, i.e. hexameric BfpD binds to BfpE 39 -76 in the presence of ADP and to BfpE 77-114 in the presence of ATP␥S. Interestingly, these data suggest that up to six molecules of BfpE can bind to BfpD that contains ADP, whereas a maximum of three molecules of BfpE can bind to BfpD containing ATP. In addition, the low K d for these interactions suggest that BfpD-BfpE 39 -76 and BfpD-BfpE 77-114 complexes are likely to be biologically relevant.
Hexameric BfpD Binds Six Molecules of the Fluorescent Nucleotide Derivatives TNP-ATP and TNP-ADP-To determine whether the stoichiometry of binding of BfpD-ATP␥S to BfpE 77-114 and BfpD-ADP to BfpE 39 -76 reflects the stoichiometry of binding of BfpD to ATP␥S and ADP, respectively, we used the non-hydrolysable nucleotide analogues TNP-ATP and TNP-ADP to ascertain the parameters of BfpD binding to nucleotides. TNP-nucleotides are weakly fluorescent in aqueous solution, however, their fluorescence is greatly enhanced in a hydrophobic environment such as the nucleotide-binding site of a protein. Upon excitation at 408 nm, each TNP-nucleotide in solution exhibited a characteristic fluorescence emission with a maximum at 558 nm; addition of monomeric BfpD markedly increased the fluorescence and blue shifted the maximum to 535 nm (data not shown). When monomeric BfpD was incubated with the TNP-nucleotides, both TNP-ATP and TNP-ADP bound to BfpD with saturable enhancement in the fluorescence emission (Fig. 5). The data were analyzed as described by Stinson and Holbrook (47) and the K d and the concentration of TNP-nucleotide-binding sites were obtained from the reciprocal of the slope and the x-intercept, respectively, of the resulting plots (data not shown). BfpD bound TNP-ATP (K d 340 nM) with higher affinity compared with TNP-ADP (K d 760 nM). The stoichiometry of binding (molar ratio of TNP-nucleotide/BfpD), calculated by dividing the concentration of TNP-nucleotidebinding sites with the BfpD concentration, was 1.07 for TNP-ATP and 0.98 for TNP-ADP. These results indicate that one molecule of monomeric BfpD binds one molecule of either of these TNP-nucleotides. Because monomeric BfpD hexamerizes in the presence of either TNP-ATP or TNP-ADP, we can assume that one molecule of hexameric BfpD binds six molecules of the TNP-nucleotides, each BfpD monomer of the hexamer binding one TNP-nucleotide molecule. Thus, the inability of hexameric BfpD to bind more than three BfpE 77-114 molecules is not because of an inability of each subunit to bind ATP.

FIG. 3. Yeast two-hybrid analysis of binary interactions between fragments of the N terminus of BfpE and BfpD.
Yeast strain AH109 was co-transformed with expression vectors encoding GAL4 DNA-binding (pGBKT7) and activation (pGADT7) domains fused in-frame to fragments of the N terminus of BfpE and full-length BfpD. For purposes of clarity, only the Bfp protein encoded on vectors pG-BKT7 and pGADT7, respectively, is indicated. Numbers correspond to amino acid residue numbering of BfpE. ϩ and Ϫ are the positive and negative controls, respectively, described in the text. Transformants were plated on medium containing X-␣-Gal and lacking histidine and adenine and assayed for ␣-galactosidase activity to detect transcriptional activation of reporter genes MEL1, HIS3, and ADE2. Transformants that produced blue colonies on selective medium are shown in white, whereas those that did not grow are shown in black. Columns denote the mean Ϯ S.E. ␣-galactosidase activity from three experiments, each performed in triplicate.

FIG. 4. Titrations of hexameric BfpD into BfpE 39 -76 (A) and BfpE 77-114 (B) in the presence of ATP␥S (black) and ADP (red).
Top panels show the raw data for the calorimetric titrations. The area under each peak represents the heat produced at each injection. The lower panels show integrated areas that were corrected for heat of dilution and plotted against the molar ratio (BfpD/peptide), with the best-fit curve for a one set of sites model.

BfpE Binding Increases BfpD Susceptibility to Proteinase K Digestion-
We have previously shown that BfpD and the N terminus of BfpC induce reciprocal conformational changes upon binding to one another (29). We hypothesized that each BfpD subunit of hexameric BfpD-ADP binds one molecule of BfpE 39 -76 , whereas alternate BfpD subunits of hexameric BfpD-ATP␥S bind one molecule of BfpE 77-114 . If this is the case, it suggests that binding of one monomer of BfpD-ATP␥S to BfpE 77-114 induces conformational changes that preclude binding of adjacent monomers. To examine the effect of BfpE binding on the conformation of BfpD occupied by either ATP␥S or ADP, BfpD in the presence of peptides of BfpE 39 -76 or BfpE 77-114 was compared with BfpD in the absence of these peptides with respect to susceptibility to proteinase K proteolysis. Fig. 6 shows silver-stained gels of two representative proteinase K digestion experiments in which hexameric BfpD, purified in the presence of ATP␥S (top panel) or ADP (bottom panel), was incubated alone and together with BfpE 39 -76 or BfpE 77-114 and subjected to limited proteinase K proteolysis. Note that BfpE 39 -76 and BfpE 77-114 migrate too rapidly to be seen on the gels. BfpD-ATP␥S and BfpD-ADP alone were relatively resistant to proteinase K. The presence of BfpE 39 -76 and BfpE 77-114 did not appear to alter the resistance of BfpD-ATP␥S and BfpD-ADP, respectively, to proteinase K. In contrast, BfpE 77-114 significantly increased the sensitivity of BfpD-ATP␥S to digestion by proteinase K. In addition, BfpE 39 -76 modestly but reproducibly increased the sensitivity of BfpD-ADP to proteolysis. These results show that BfpE 39 -76 and BfpE 77-114 only had a detectable effect on the proteinase K sensitivity of the form of BfpD with which our ITC data demonstrates they interact. Furthermore, they suggest that BfpE 39 -76 and BfpE 77-114 induce changes in the conformation of BfpD-ADP and BfpD-ATP␥S, respectively, upon binding, which account for the observed increases in sensitivity to proteinase K digestion. However, consistent with our hypothesis, BfpD-ADP and BfpD-ATP␥S undergo different conformational changes as they display differing proteinase K sensitivities.
BfpD Binding to the N Terminus of BfpC and BfpE 77-114 Dramatically Increases Its ATPase Activity-Because hexameric BfpD interacts with the N terminus of BfpC, BfpE 39 -76 , and BfpE 77-114 (Ref. 29, and this study), we investigated the influence of these proteins on ATPase activity. Hexameric BfpD was preincubated with various combinations of the N terminus of BfpC, BfpE 39 -76 , and BfpE 77-114 at a molar ratio of BfpC or BfpE to BfpD of 6:1, 6:1, and 3:1, respectively, based on reaction stoichiometry determined by ITC, and the ATPase activity was assayed. The results in Fig. 7A demonstrate that the proteins with which hexameric BfpD-ATP␥S interacts, i.e. the N terminus of BfpC and BfpE 77-114 , had a dramatic effect on the ATPase activity of hexameric BfpD. The N terminus of BfpC and BfpE 77-114 together stimulated the ATPase activity more than 1200-fold. The N terminus of BfpC alone increased the ATPase activity to 58.0 mol of released P i min Ϫ1 mg Ϫ1 of protein and BfpE 77-114 further increased it to 75.6 mol of released P i min Ϫ1 mg Ϫ1 of protein. Consistent with these results, the N terminus of BfpC and BfpE 77-114 together increased the V max of BfpD over 1200-fold, with the N terminus of BfpC alone increasing it almost 1000-fold (Table IV). In addition, the N terminus of BfpC reduced the K m 6-fold (Table IV), indicating an increase in the affinity of BfpD for ATP in the presence of BfpC. The stimulation of ATPase activity only occurred in the presence of Zn 2ϩ (Fig. 7B), highlighting the importance of utilizing optimal reaction conditions. Although the total Zn 2ϩ content of E. coli is in the millimolar range, evidence suggests the free Zn 2ϩ concentration is femtomolar (50). Therefore, we examined the ATPase activity of BfpD with and without BfpC and BfpE across a broad range of Zn 2ϩ concentrations. Remarkably, under both conditions there was only a 2% change in activity over Zn 2ϩ concentrations that spanned many orders of magnitude. Thus, the BfpC-and BfpEmediated enhancement of BfpD activity still occurred in the presence of trace levels of Zn 2ϩ (data not shown). In the presence of other divalent cations, ATPase activity was enhanced only 2-fold by the N terminus of BfpC, BfpE 39 -76 , and BfpE 77-114 . Interestingly, BfpE 77-114 alone and BfpE 39 -76 did not stimulate ATPase activity (Fig. 7A). Furthermore, the N terminus of BfpC and BfpE 77-114 had similar allosteric effects on the delayed ATPase activity of monomeric BfpD (data not shown). In conclusion, the N terminus of BfpC and BfpE 77-114 induce conformational changes that greatly enhance the ATPase activity of hexameric BfpD. DISCUSSION Tfp biogenesis, protein secretion, DNA transfer, archeal flagellar assembly, and filamentous phage morphogenesis systems are multiprotein assemblies constructed in part from proteins conserved between these systems. Of the proteins composing these macromolecular transport systems, putative nucleotide-binding proteins of the PulE superfamily are widely distributed and highly conserved (17). The PulE superfamily belongs to the P-loop class of NTPases and is characterized by Walker A (or P-loop) and B boxes, an Asp box, and a His box (32,40,51). Although the general biochemical characteristics of PulE superfamily proteins are well documented, important questions regarding their specific role and mode of action remain unanswered. To decipher the mechanism by which proteins of the PulE superfamily harness energy from ATP hydrolysis for the function of their respective systems, we considered it extremely important to characterize the biochemical activities of proteins of the PulE superfamily in the context of the systems to which they belong. The BFP biogenesis machine of EPEC is an attractive model system because all of the proteins composing this machine and the interactions of the PulE homologue BfpD with the other components have been identified (22,23,29). Therefore in this study, we have characterized the ATPase activity of the EPEC BfpD protein and the effects of conditions related to its biological function on this activity.
We purified BfpD using an approach including Sephacryl S-300 gel filtration chromatography and observed that BfpD eluted as a hexamer as well as a monomer in the presence of nucleotide (ATP, ATP␥S, TNP-ATP, ADP, and TNP-ADP). A combination of gel filtration, zonal sedimentation, DLS, and electron microscopy confirmed that BfpD forms ring-shaped hexamers in the presence of nucleotide (ATP, ATP␥S, and ADP). The ability to form hexameric rings has been reported for other PulE superfamily members (33,34,37,(52)(53)(54) and seems to be a common characteristic of members of this superfamily. These hexameric ring assemblies are reminiscent of those formed by nucleotide-dependent molecular motors such as helicases and F1-ATPases (52,54,55), and membrane fusion ATPases such as N-ethylmaleimide-sensitive fusion protein and p97 (56,57).
The ATPase activity measured for purified hexameric BfpD (V max of 62.2 nmol of P i released min Ϫ1 mg Ϫ1 of protein) is similar in magnitude to the highest values previously measured for other members of the PulE superfamily (ranging from 0.71 to 61 nmol of P i released min Ϫ1 mg Ϫ1 of protein). We considered the weak ATPase activities of these PulE superfamily members to be because of the absence of stimulatory proteins of the cognate systems and/or the reaction conditions used. After optimizing our reaction conditions, we tested whether components of the BFP biogenesis apparatus that interact with BfpD stimulate its ATPase activity. We have previously shown that BfpD interacts with the cytoplasmic N termini of BfpC and BfpE (BfpE 1-114 ) and is recruited to the cytoplasmic membrane by both BfpC and BfpE (29). Here we further define the BfpD-binding sites of BfpE and show that BfpD binds to one of two sites in BfpE depending on the stage of the catalytic cycle. The ATP-bound form of BfpD binds with high affinity (K d ϭ 72 nM) to BfpE 77-114 and the ADP-bound form of BfpD binds with high affinity (K d ϭ 75 nM) to BfpE 39 -76 . We found that the N terminus of BfpC and BfpE 77-114 together stimulate the ATPase activity of BfpD more than 1200-fold to a V max of 75.3 mol of P i released min Ϫ1 mg Ϫ1 of protein. This activity exceeds that of any member of the family studied thus far by over 1200-fold. Moreover, to our knowledge BfpD is the first component of a Tfp biogenesis machinery or similar system whose ATPase activity has been shown to be modulated by other components of the system. The N terminus of BfpC alone increased the ATPase activity almost 1000-fold to a V max of 58.7 mol of P i released min Ϫ1 mg Ϫ1 of protein. However, BfpE 77-114 only increased ATPase activity in the presence of the N terminus of BfpC. Furthermore, BfpE 39 -76 had no detectable effect on BfpD ATPase activity, a result that is not surprising, as BfpD-ATP does not interact with BfpE 39 -76 . The binding of BfpC and BfpE to BfpD induce conformational changes in each of the proteins as indicated by protease sensitivity experiments conducted in this and prior studies (29). The allosteric effects of BfpC and BfpE on BfpD ATPase activity indicate that BfpC and BfpE indirectly play a major role in supplying the BFP biogenesis machine with the energy it requires for its assembly, bundlin polymerization. and/or transmembrane transport. Not only do they recruit BfpD to the biogenesis machinery, they also dramatically increase its ATPase activity. In this way, the ATPase activity of BfpD is regulated to be maximal at the site where energy is needed for BFP formation, making the process of producing BFP energy efficient and minimizing wasted energy expenditure by the free cytoplasmic enzyme. Interestingly, monomeric BfpD preparations exhibited delayed ATPase activity that was greatly increased by the N terminus of BfpC and BfpE 77-114 but not BfpE 39 -76 . This ATPase activity was most likely that of BfpD hexamers formed in the presence of the ATP added in the assay, because monomeric BfpD does not interact with the N terminus of BfpC, BfpE 39 -76 , or BfpE 77-114 . Our data suggest that nucleotide binding induces formation of BfpD hexamers, which are selectively recognized, recruited, and stimulated by membrane-bound BfpC and BfpE, thus maximizing the energy supplied to the BFP biogenesis machine.
We found that hexameric BfpD binds to a maximum of three molecules of BfpE 77-114 in the presence of ATP, but can bind to FIG. 7. BfpC and BfpE increase the ATPase activity of hexameric BfpD by more than 1000-fold. BfpD ATPase assays were carried out in 40 mM Tris-HCl, pH 7.0, 150 mM NaCl, 5% glycerol, 5 mM ATP, and 10 mM ZnCl 2 , in the presence of the N terminus of BfpC, BfpE 39 -76 , and BfpE 77-114 unless otherwise indicated (see "Experimental Procedures"). In A, assays were performed in the presence of the indicated combination of BfpC, BfpE 39 -76 , and BfpE 77-114 . In B, assays were performed in the presence of BfpC and BfpE 77-114 and the Zn 2ϩ in the reaction buffer was replaced with the indicated ion. The results are derived from at least three assays performed in triplicate, each using different BfpD preparations. six molecules of BfpE 39 -76 in the presence of ADP. This difference in stoichiometry is not because of differences in nucleotide saturation, as BfpD is able to bind six molecules of either ATP or ADP. Rather these data suggest that occupancy of the ATPactivated BfpD hexamers by BfpE occurs at every other subunit and induces a conformational change that precludes binding of adjacent subunits.
Hexameric BfpD exhibited the highest ATPase activity in the presence of Zn 2ϩ and the dramatic stimulation of its activity by the N terminus of BfpC and BfpE 77-114 occurred only in the presence of Zn 2ϩ . The crystal structure of a truncated form of the closely related type II secretion ATPase EpsE (missing ninety N-terminal amino acids) of V. cholerae has been solved, revealing a molecule with four domains, N2, C1, C M , and C2, and inferring a missing N1 domain. Electron density most consistent with the presence of a Zn 2ϩ ion was detected in the C M domain, coordinated by four cysteines. This location is remote from the active site. An alignment of BfpD and EpsE reveals that these cysteine residues are conserved. It is not clear therefore, whether Zn 2ϩ is catalytically relevant or merely structurally relevant for either EpsE or BfpD. For example, Zn 2ϩ binding might be necessary for conformational changes that allow binding of BfpE or BfpC. However, this hypothesis is not supported by our ITC data, which were acquired in the presence of Mg 2ϩ , not Zn 2ϩ . Therefore, we propose that the Zn 2ϩ ion might perform a critical catalytic role for BfpD and perhaps for other members of the PulE ATPase family. In our assays, we purified hexameric BfpD in buffer lacking divalent metal ions and measured ATPase activity in the presence of buffer containing various divalent metal ions. We cannot be certain that trace quantities of other metals were not present in any of the buffers used. However, we found using inductively coupled plasma optical emission spectroscopy that the enzyme purified in the absence of Zn 2ϩ already contained an equimolar ratio of Zn 2ϩ (data not shown). Therefore, the fact that the addition of Zn 2ϩ to the enzyme, which already contains Zn 2ϩ , stimulated activity more than 600-fold higher than any other metal, suggests that the Zn 2ϩ plays a catalytic role in addition to a structural role. Furthermore, if the role of Zn 2ϩ were merely conformational, then the structural analogue Cd 2ϩ would be expected to be a reasonable substitute. Instead the activity in the presence of Cd 2ϩ was less than 1/1000 of that in the presence of Zn 2ϩ when BfpC and BfpE 77-114 were also present. Thus, we propose that BfpD and perhaps other PulE ATPases use Zn 2ϩ to perform a catalytic role in ATP hydrolysis, which to our knowledge is unprecedented for ATPases. However, Zn 2ϩ is known to catalyze hydrolysis of phosphate bonds for other enzymes, for example, the hydrolysis of phosphodiester bonds by endonuclease IV (58), and therefore such a role is clearly plausible.
Based on the findings presented here, we envision the role of BfpD in BFP biogenesis is analogous to that of p97 and Nethylmaleimide-sensitive fusion protein in homotypic membrane fusion and organelle biogenesis in that it transduces mechanical force to the BFP biogenesis machine possibly for the extraction of bundlin from the cytoplasmic membrane. This theory is supported by the following findings: (i) nascent bundlin is an integral cytoplasmic membrane protein (59) and therefore must be extracted from the membrane for pilus biogenesis; (ii) the N terminus of BfpE is involved in interactions with other essential Bfp proteins that are necessary for BFP formation (29); (iii) hexameric BfpD exhibits high ATPase activity in the presence of other essential components of the BFP biogenesis machine, consistent with the idea that it is a molecular motor; and (iv) hexameric BfpD displays differential binding to two sites of the N terminus of BfpE depending on the phospho-rylation of its associated nucleotide. This last point indicates that the relative position of BfpD and BfpE change upon ATP hydrolysis. Thus, the resulting chemical energy can be converted to mechanical energy to change the position of BfpE and the proteins to which it binds. Fig. 8 illustrates the proposed mode of action of BfpD, showing the changes it induces in the cytoplasmic membrane subassembly of the BFP biogenesis machine upon hydrolysis of ATP. The figure is simplified to show the action of only a BfpD monomer, but presumably occurs simultaneously at alternating monomers. Hexameric BfpD-ATP interacts with the cytoplasmic N termini of BfpC and BfpE, which also interact with each other. BfpD-ATP binds to BfpE 77-114 , the domain of the N terminus of BfpE closest to the cytoplasmic membrane. Conformational changes in BfpD upon binding to the N terminus of BfpC and BfpE 77-114 greatly increase its ATPase activity. Subsequently, the BfpD-ADP resulting from ATP hydrolysis shifts to BfpE 39 -76 while remaining bound to the N terminus of BfpC. It seems unlikely that BfpD moves to its new binding site in BfpE as its movement is restricted by its interaction with the N terminus of BfpC. Therefore we propose that the N terminus of BfpE is raised allowing BfpD-ADP to interact with BfpE 39 -76 . This movement may force the most distal part of the N terminus of BfpE into the cytoplasmic membrane and perhaps into the periplasm. The interaction between the N terminus of BfpC and BfpE 39 -114 presumably limits the distance the N terminus of BfpE is pushed through the cytoplasmic membrane. According to this model, the BfpD motor energizes BfpE to act as a piston, whereas BfpC serves as a scaffold protein, anchoring BfpD and the N terminus of BfpE in place. Preliminary studies indicate that other essential Bfp proteins bind to BfpE 77-114 , 2 which could be transported across the cytoplasmic membrane in association with the N terminus of BfpE. In this way, BfpD provides the necessary mechanical force for recruitment and assembly of Bfp proteins into the BFP biogenesis FIG. 8. Schematic diagram of the proposed mechanism by which BfpD transduces mechanical energy to the BFP biogenesis machine. Hexameric BfpD-ATP binds to the N terminus of BfpC and to the last third of the N terminus of BfpE. This greatly increases its ATPase activity and the resulting Bfp-ADP releases the distal third of the N terminus of BfpE and instead binds to the middle third while remaining bound to the N terminus of BfpC. This movement drives the distal third of the BfpE N terminus through the membrane. Upon replacement of ADP with ATP, BfpE relaxes, re-establishing its resting state. machine, which results in the extraction of nascent bundlin from the cytoplasmic membrane and incorporation into the growing pilus filament. Finally, BfpD is replenished with ATP and released from BfpE 39 -76 and the cytoplasmic membrane subassembly of the BFP biogenesis machine relaxes, allowing the distal part of the N terminus of BfpE to slip back through the cytoplasmic membrane. Whereas further experiments are necessary to test this model of energy transduction during BFP biogenesis, the insights into the function and mode of action of BfpD we report here increase our understanding of the mechanism of Tfp biogenesis machines and have important implications for similar systems that transport biological macromolecules across membranes. We propose that this model is relevant to the entire PulE superfamily. Orthologues of BfpE (GspF) are highly conserved in systems that include PulE superfamily members (17). Although BfpC orthologues are not common, GspL and GspM proteins, which span the cytoplasmic membrane, interact with each other and recruit PulE family AT-Pases to the cytoplasmic membrane (41,60), are widespread and likely play roles similar to that carried out by BfpC. Furthermore, the unique properties of BfpD suggest that drug therapy targeting the ATPase activity of PulE superfamily proteins and their interactions with proteins of their cognate systems could be designed to disrupt these macromolecular machines and the virulence properties they control.