Conjugative plasmid protein TrwB, an integral membrane type IV secretion system coupling protein. Detailed structural features and mapping of the active site cleft.

Bacterial conjugation is an example of macromolecular trafficking between cells and responsible for the spreading of antibiotic resistance among bacteria. It involves translocation of single-stranded DNA across membranes through a type IV secretion system. A coupling protein links the DNA-processing nucleoprotein complex, the relaxosome, with the transport apparatus during cell mating. In Escherichia coli plasmid R388 such a protein is TrwB, a basic integral inner-membrane nucleoside-triphosphate-binding protein. TrwB is the structural prototype for the type IV secretion system coupling proteins, a family of proteins essential for macromolecular transport between cells and export. The structure of a soluble TrwB variant unveils an elongated molecule with six equivalent protein units featuring a spherical quaternary structure, leaving a central channel. The structures of the non-liganded protein and four different complexes with substrate analogues and products allow the precise description of the active site architecture. The active sites are located at the interface between protomers, each of them shaped mainly by residues of one monomer, but including two crucial arginine residues belonging to the adjacent molecule. Upon substrate binding and putative hydrolysis, conformational changes are transferred from the external surface to the interior central channel.

Transport of DNA across cell membranes occurs in different biological processes such as viral infection, cell division, spor-ulation, or conjugation. The latter is mediated by conjugative plasmids aimed to transfer DNA unidirectionally between bacteria, from a donor to a recipient cell, which then becomes a potential donor. It is the main process by which bacterial pathogens acquire antibiotic resistance and constitutes a widespread tool to shuttle genes between different species of bacteria. Even trans-kingdom gene transfer can be achieved through a conjugation-like system, from bacteria to fungal or plant cells (1)(2)(3)(4)(5).
The transfer utilizes a type IV secretion system (6), a multiprotein transmembrane organelle, and requires physical contact between donor and recipient cells. Genes contained in the plasmid transference (tra) region, further subdivisible into dtr and mpf, control conjugation (5,7). Mpf proteins are required for mating pair formation and assembly of the conjugative or sex-pilus. The region of DNA transfer and replication genes, dtr, includes oriT, the origin of transfer, and genes encoding proteins involved in plasmid DNA processing. Establishment of stable intercellular contacts may trigger a signal that is transmitted to a specialized nucleoprotein complex, the relaxosome (1), made up by the DNA region at oriT, and Dtr proteins that cleave and unwind DNA. The relaxosome moves to the transport site, and the type IV DNA transport apparatus (formed by Mpf proteins) transfers single-stranded (ss) 1 DNA or T-strand to the recipient cell.
Escherichia coli plasmid R388 (8) has a size of 33 kb and contains genes that code for sulfonamide and trimethoprim resistance (9,10). Its mpf region encodes a functional type IV secretion system, featuring 11 trw genes (trwD to trwN), which are similar to other conjugative mpf genes and to Agrobacterium tumefaciens virB genes (11,12). 2 R388 displays the shortest dtr region known (12,13), just made up by oriT, trwA, trwB, and trwC. TrwC acts as a relaxase and helicase. It is responsible for both nick cleavage at oriT and T-strand unwinding (14,15). TrwA is a small-sized tetrameric protein, which binds two sites at oriT and enhances TrwC relaxase activity, while repressing transcription of the trwABC operon (16). These two proteins constitute, together with oriT DNA and the hostencoded integration host factor, the relaxosome of plasmid R388 (12). Integration host factor has a modulator role in conjugation (17), because it inhibits TrwC nic cleavage, probably by affecting the topology of the TrwC DNA target site (18).
The third protein encoded by R388 dtr region, TrwB, is a basic (pI ϭ 10) integral inner-membrane protein of 507 residues (12,19). It features a nucleosyl-triphosphate (NTP)-binding site (NBS), reminiscent of those of helicases (20,21) and the ␣ and ␤ subunits of F 1 -ATPase (1,21). Sequence analysis predicts the transmembrane part to comprise the seventy Nproximal residues (two transmembrane helices and a small periplasmic domain in between), so TrwB was overproduced and purified as a soluble fragment lacking these residues (TrwB⌬N70). This fragment has been shown to bind ATP (19). Furthermore, TrwB⌬N70 interacts with both ss and doublestranded (ds) DNA with approximately the same affinity as supercoiled dsDNA (19). This DNA binding is nonspecific, independent of NTP binding, and associated with TrwC nick cleavage enhancement, possibly due to a topological modification of DNA around oriT, favoring its ss nature. The protein couples the unwound T-strand in its relaxosome complex with the type IV transport pore (19) and is the prototype for a family of membrane-bound type IV secretion system coupling proteins (T4CPs).
T4CPs bridge the relaxosome with the structures responsible for cell-to-cell contact in all conjugative DNA-transfer systems studied so far (5,(22)(23)(24). They are also present in as distinct systems as agrobacterial transfer-DNA translocons or pathogenicity-related protein secretion systems. This family has a similar arrangement of possible transmembrane segments, and its members possess at least one NTP-binding domain (NBD; see Fig. 1a). Besides TrwB, it further includes TraD of plasmid F, TraG of plasmid RP4, VirD4 from the Ri and Ti plasmids (25), and TrsK from pGO1 (26), among more than 40 additional homologues. Interestingly, some of these proteins can be functionally exchanged for another member of the family (27).
Finally, a further interesting functional similarity is found between TrwB-like proteins and the SpoIIIE/FtsK bacterial protein family (19). SpoIIIE and FtsK are supposed to drive DNA across the annulus during sporulation and cell division, respectively (19, 28 -31). TrwB has the same molecular organization as SpoIIIE/FtsK-like proteins (see Fig. 1a) and shares sequence similarities, in particular around the helicase NTPbinding motifs, suggesting a relationship in function.
We briefly described the structure of non-liganded TrwB⌬N70 (32). Here we present its fully refined structure and its binary complexes with non-hydrolyzable GTP and ATP an-alogues, with ADP⅐Mg 2ϩ and with a sulfate anion. Accordingly, we show the enzyme in different states, the apo protein, two equivalent "substrate" complexes, and two different "product" complexes, respectively. These data unveil detailed structural features of the active site.

EXPERIMENTAL PROCEDURES
Purification, Crystallization, Complex Formation, and Data Collection-The soluble 437-residue variant of TrwB, TrwB⌬N70, was produced by overexpression and purified as described previously (19). The protein was crystallized in two crystal forms (P3 1 21 and P2 1 ) from equivolumetric drops of protein solution and precipitant agent solution employing the hanging-drop vapor-diffusion method (32,33). Complexes between TrwB⌬N70 and adenosine-and guanosine-5Ј-(␤,␥-imido)triphosphate (ADPNP and GDPNP, respectively) and ADP were obtained by a 20-to 30-min soak of trigonal crystals grown from tartrate in 0.1 M GDPNP, ADPNP, or ADP further 0.25 M in HEPES (pH 7.5) and 10 mM in MgCl 2 . Crystals treated in this way were immediately flash cryo-cooled in liquid nitrogen after application of a cryo-protecting strategy as described (33).
Complete diffraction data sets were collected from a single cryocooled crystal each on a 345-mm MAR Research Imaging plate, 165-mm marCCD and ADSC Quantum4 CCD detectors at DESY (Hamburg) and ESRF (Grenoble) synchrotrons. All data were processed with MOSFLM 6.01 (34) and scaled, merged, and reduced with SCALA (35). Table I provides a summary for the data collection and processing.
Structure Solution and Refinement-The structure of the trigonal crystal form of TrwB⌬N70 grown from sulfate (product P complex) was solved by multiple-wavelength anomalous dispersion as described previously (33), revealing a toroidal homohexamer in the asymmetric unit (32). The structure of the monoclinic crystal form (apo form) was solved by molecular replacement using a whole TrwB⌬N70 hexamer as a searching model and AMoRe (33,36) unveiling twelve monomers arranged as two hexamers in the asymmetric unit. The models were built and improved in successive cycles of manual model building on a Silicon Graphics workstation with TURBO-FRODO (37) and posterior bulksolvent and anisotropic temperature factor correction, followed by maximum-likelihood positional and temperature factor refinement with CNS version 1.0 (38). The structures of the GDPNP, ADPNP, and ADP complexes were solved by difference-Fourier synthesis using the trigonal crystal structure as a starting model. The coordinates were subjected to rigid-body and maximum-likelihood positional and temperature factor refinement with CNS version 1.0 against the diffraction data of each of the three complexes. The subsequently calculated electron density omit maps before and after 6-fold averaging clearly revealed the presence of bound molecules in the six NBSs of a particle (see Fig. 5, b and d). Additional density for a HEPES molecule on the molecular surface for a solvent-exposed chloride anion (tentatively assigned in base of strong electron density) and for a second, partially disordered GDPNP molecule in the central hexamer pore were found in the first complex. The ADPNP complex structure revealed an additional chloride   (27) are more closely related between them than they are to the FtsK/SpoIIIE-proteins (30,73). Transmembrane segments (TM) are represented in red and NBDs in blue with A and B walker motifs depicted. AADs present in the TrwB and VirD4 families are colored in green. b, topology scheme displaying the regular secondary structure elements of TrwB⌬N70. Green arrows represent ␤ strands (labeled ␤1 through ␤12), and rods stand for ␣-helices (labeled ␣A through ␣R). The N and C termini, the relative position of the central channel, the inner-membrane and the AAD, the NTP-binding site and the NBD are further displayed, as is the amino acid positions that make up each secondary structure element. c, stereo ribbon plot of a TrwB⌬N70 monomer. Helices are magenta, strands are cyan arrows, and coils and turns are yellow ropes. A bound GDPNP (substrate complex) molecule is shown as a green stick model to position the NBS. d, surface representation of the complete TrwB hexamer, including the modeled transmembrane parts (lower stem). a R-factor ϭ {⌺ hkl ʈF obs ͉ Ϫ k͉F calc ʈ/⌺ hkl ͉F obs ͉} ϫ 100, with F obs and F calc as the observed and calculated structure factor amplitudes; free R-factor same for a test set of 7% reflections (Ͼ500) not used during refinement (until the penultimate cycle). Data sets of same space group share the test set in the common resolution range.
anion, and the ADP complex revealed an ion of Mg 2ϩ in each NBS (except in the NBS of chain F), besides two HEPES molecules at the molecular surface. Model building and computational refinement were assessed as mentioned. The final refinement statistics are presented in Table II.
The polypeptide chains of the different monomers within each hexamer are defined from their N termini at residues Asn 72 to Glu 78 until the C-terminal amino acids Glu 504 to Ile 507 . Residues Arg 437 -Thr 444 to Thr 450 -Glu 455 from a surface-exposed region (strands ␤9 and ␤10) probably interacting with the inner part of the chopped-off transmembrane region, are flexible and have not been traced. All residues, excepting His 125 (⌽/⌿ ϭ ϳ70°/ϳ160°) of each polypeptide chain in all structures, are placed in allowed regions of the Ramachandran plot. This residue is, however, clearly defined by electron density and located at the beginning of strand ␤2, because its (and the preceding residue's) carbonyl oxygen is involved in ␤ sheet interactions with amide nitrogen atoms of residues Arg 408 and Ser 409 .

RESULTS AND DISCUSSION
The Monomer Structure-The TrwB⌬N70 monomer consists of two domains: a nucleotide-binding domain (NBD) attached to the inner-membrane and a membrane-distal all-␣ domain (AAD; Fig. 1, b and c). This NBD displays an ␣/␤ P-loop containing NTP-hydrolase core structure. This domain runs from the membrane-proximal N terminus, anchor point of the excised transmembrane domain (TMD), to residue Lys 183 and continues from position Asp 298 to the surface-located C terminus. It is composed of a central, highly twisted (ϳ90°) ninestranded pleated ␤-sheet of mixed parallel/antiparallel topol-ogy (strands ␤2-␤8, ␤11, and ␤12; Fig. 1b), flanked on its concave side by four helices (termed ␣A through ␣D) and on the other side by seven helices (␣L through ␣R). The convex side faces an interior channel. On the membrane-proximal edge of the central ␤-sheet, a small three-stranded antiparallel sheet (strands ␤1, ␤9, and ␤10; see Fig. 1b) is inserted, almost perpendicularly. Strands ␤9 and ␤10 are only partially defined by electron density, probably due to the absence of interactions with the (excised) TMD preceding strand ␤1.
On top of NBD, the smaller AAD is inserted between strand ␤4 and helix ␣L, comprising residues Gly 184 -Gly 297 (Fig. 1, b and c). This domain contains seven helices: a 3 10 -helix (helix ␣E), followed by a two-helix hairpin (helices ␣F and ␣G) and by a second, curved, helical hairpin segment (helices ␣H through ␣K). The latter mainly protects and interacts with the central helix ␣G. The interaction between NBD and AAD is mainly of hydrophilic nature and encompasses a common surface of 2314 Å 2 (about 36% of the total surface of AAD). It comprises mainly segments Leu 175 -Tyr 178 , Thr 182 -Lys 183 , Asp 298 -Glu 306 , Ala 324 -Ser 342 , Glu 362 -Ala 365 , and Asp 393 -Val 394 of NBD and Gly 184 -Tyr 195 , Arg 199 -Lys 209 , and Asp 286 -Gly 297 of AAD. We analyzed the 56 close contacts between domains and observed 18 hydrogen bonds, 4 salt bridges, and 6 van der Waals interactions.
TrwB Forms Hexamers-As derived from both crystal forms obtained (see "Experimental Procedures"), six TrwB⌬N70 protomers intimately associate employing a local 6-fold axis (see Figs. 1d and 2b). One hexamer is observed in the asymmetric unit of the trigonal crystals (monomers A to G), and two are present in the monoclinic system (monomers A to G and H to M, respectively). This agrees with the native protein migrating as a hexamer in gel filtration column chromatography (data not shown). The protomers are practically indistinguishable from each other within a hexamer. The hexamer, with overall dimensions of 110-Å diameter and 90-Å height (see Fig. 2a), displays an orange-like shape, somewhat flattened at both poles along the axis. A central channel runs from the cytosolic pole (made up by the AADs) to the membrane pole (formed by the NBDs), ending at the (tentatively modeled) transmembrane pore shaped by the 12 transmembrane helices, rendering an overall mushroom-like structure (Fig. 1d). This channel is delimited at its entrance mainly by gatekeeping residues Asn 271 of loop ␣I␣J from each subunit and is restricted to a diameter of ϳ7-8 Å. This is the closest point of the channel that, at its exit to the TMD, has an opening of ϳ22 Å. The interaction between TrwB⌬N70 monomers is mainly of hydrophilic nature with the central twisted ␤-sheets arranged almost in parallel (C-edge to N-edge). It is characterized by a common surface of 4588 Å 2 as calculated for monomers A and B (about 25% of the total surface of a monomer). This contact surface is lined up mainly by segments Gly 80  The total intermolecular contact surface of a monomer (e.g. A with both B and G) is 9100 Å 2 , almost 50% of the total surface. Upon inspection of the 95 close contacts between these monomers A and B, 28 hydrogen bonds, 6 salt-bridges, and 19 van der Waals interactions are ascertained. In particular, a close contact is observed between strands 9 and 10 (partially disordered at their ends, see above) and residues of the N-terminal segments of each monomer.
Structural Similarities in the NBD-TrwB reveals a stringent structural similarity of its NBD with the equivalent part of RNA and DNA helicases, molecular motors involved in DNA metabolism that use the energy from NTP hydrolysis for nucleic acid unwinding or strand separation/translocation (45). RecA (Protein Data Bank access code (PDB) 2reb; see Fig. 3a), essential for genetic recombination and repair, is the structural prototype for these enzymes that can adopt varying (functional) oligomerization states (46 -48). The family further includes toroidal hexameric DNA ring helicases, like T7 phage gene 4 replicative helicase/primase (PDB 1e0j (49) and 1cr0 (50)) and RepA (PDB 1g8y (51)), and monomeric enzymes like 3Ј35Ј PcrA DNA helicase (PDB 2pjr and 3pjr (52)).
The family of AAA proteins (ATPases associated with a variety of cellular activities) displays an NBD with a similar fold to TrwB NBD, as reported for eukaryotes, prokaryotes, and archaebacteria. It is an essential family of specialized, chaperone-like enzymes that participate in membrane fusion and trafficking, organelle biogenesis, proteolysis, and protein folding (53,54). It includes the ␦Ј subunit of the clamp loader complex of E. coli DNA polymerase III (PDB 1a5t (55)) and N-ethylmaleimide-sensitive fusion protein (NSF) domain 2, a cytosolic ATPase required for intracellular vesicle fusion reactions (PDB 1d2n (56)). AAA ATPase p97 is involved in homotypic membrane fusion (PDB 1e32 (57)), and ATP-dependent protease HsIU-HsIV (PDB 1e94 (58)).
Striking structural similarity is further found between TrwB and both ␣ and ␤ subunits of F 1 -ATPase, part of the membraneassociated F 0 F 1 -ATPase complex responsible for energy conversion through axial rotation movement in mitochondria, chloroplasts, and bacteria (see Fig. 3b; PDB 1bmf (59)). Finally, the recently reported six-clawed, grapple-shaped structure of homohexameric Helicobacter pylori Cag525/HP0525 traffic ATPase, an inner-membrane-associated part of the bacterial type IV secretion system involved in pathogenic protein CagA export, also reveals structural similarity in the NBD core (60).
Nevertheless, on examining the hexameric toroidal quaternary structures of the mentioned protein families (Fig. 2), helicases, AAA ATPases, and Cag525 appear more flat-topped than TrwB, with much less interaction surface between the constituting protomers. This weaker interaction is required in helicases to permit interchanging aggregation stages and helicoidal protein filament formation (50,61,62). The oligomeric structure of TrwB displays an almost spherical shape, which, together with the overall hexamer dimensions, is much more reminiscent of the F 1 -ATPase ␣ 3 ␤ 3 heterohexamer (Fig. 2, a  and b), a membrane protein (membrane-associated) like our bacterial protein. Furthermore, both particles share the presence of a crown made up by ␤-strands, although placed on opposite faces of the central NBD (see Fig. 3b); i.e. cytosolic in ATPase, membrane-proximal in TrwB⌬N70.
Is the AAD a DNA-binding Domain?-Several representatives of the helicase/ATPase-like proteins display, besides a NBD, also an AAD, as shown for RecA, AAA ATPases, helicases, and the ␣ and ␤ subunits of F 1 -ATPase (see Fig. 3, a and  b), among others, although arranged in distinct ways with respect to their NBDs. However, detailed topological inspections reveal that TrwB AAD bears only significant structural similarity with N-terminal domain 1 of the site-specific recombinase, XerD, of the integrase family (see Fig. 4a; PDB access code 1a0p (63) displays four helices arranged such that there are two parallel helix hairpins arranged at 90°to each other. It has been proposed to contribute to the shape of the DNA-binding site with two of its helices, equivalent to helices ␣G and ␣J in TrwB, being considered for interaction with the major groove of the inner part of the recombinase specific site. This domain is positioned over the DNA binding region blocking the access to it, so that a large conformational rearrangement has to occur for target DNA binding. Such a rearrangement, which could also happen in TrwB, could be implicated in a putative working mechanism. Xer site-specific recombination has been associated with FtsK activity (64), a protein that displays functional and structural similarities with TrwB and other T4CPs (see above).
TrwB AAD displays further topological similarity in a ϳ40residue segment encompassing three ␣-helices with the recently reported solution structure of the 56-residue DNA-binding domain of TraM, the first of a (putative) endogenous component of the relaxosome encoded by the dtr region ((65); see Fig. 4, b and c). This protein enhances relaxase activity in plasmid R1 and is capable of binding DNA (66). TraM specifically interacts with the TrwB orthologue TraD of plasmid R1, suggesting that TraM links the relaxosome with the DNA transfer apparatus (67). Plasmid R388 lacks such a TraM or-thologue. Therefore, an appealing hypothesis is that TrwB uses a DNA-binding domain similar to that of TraM, but imbedded in its structure, to directly recruit the relaxosome. Unfortunately, experimental proof of sequence-specific interactions between TrwB and oriT DNA is still unavailable.
NTP-binding Site Location-We managed to obtain the protein structure in four states, non-liganded (monoclinic crystal form), in complexes with the NTP analogues adenosine-and guanosine-5Ј-(␤,␥-imido)triphosphate (ADPNP and GDPNP; trigonal crystal form), with ADP in the presence of Mg 2ϩ ions and with a sulfate anion (both trigonal crystal form; see Figs. 5-7). By analogy to PcrA (52,68), we suggest that the protein is trapped in states that emulate its apo form, two equivalent substrate complexes and two distinct, and possibly alternative or mutually excluding, product complexes (ADP and phosphate (P) complexes).
As predictable from the topology of the ␤-sheet and confirmed by the ability of TrwB to bind a fluorescent ATP analogue (19), the active-site crevice (in this case the NBS) is located at the C-terminal edge of the sheet and is shaped by loops ␤2␣C and ␤6␣N (see Fig. 1b Glu 357 -Leu 358 -Ala 359 ) described for NTP-binding proteins (20,21), so as a putative Asp box (Met 154 -Val 155 -Ile 156 -Val 157 -Asp 158 -Pro 159 -Asn 160 -Gly 161 ). If compared with other NTPbinding pockets of RecA-like family members, including F 1 -ATPase or PcrA (47,49,52,59), the NBS is not very deep and readily accessible to bulk solvent. Considering the whole particle, the six NBSs form a belt around the hexamer at half height of the particle, and they are located ϳ32 Å apart on superficial cavities at interfaces between vicinal protomers. Therefore, residues coming from two vicinal subunits contribute to feature each NBS. There are no significant differences between the NBSs, contrary to what has been described for helicases or F 1 -ATPases (49,59); therefore, they can be considered as equivalent.
Active Site Architecture: A Model for the Substrate Complex-The presently described structures of TrwB⌬N70, non-liganded and with ADPNP and GDPNP, make possible the definition of the residues involved in the binding and catalysis of the substrate. No unambiguous Mg 2ϩ ion could be assigned in the electron density maps of either complex despite it having been added to the crystallization mixture. Both complexes (see Figs. 5-7) are very similar (r.m.s.d. for all common C␣ atoms is 0.34 Å), thus only the GDPNP complex structure will be discussed in detail. GDPNP binding provokes only minor rearrangements when compared with the apo form. The overall r.m.s.d. between A monomers is 0.25 Å; no significant deviation is observed in the overall hexamer structure either (see Fig. 7b).
The main interaction of the bound nucleotide analogue is made with the P-loop, located between strand ␤2 and helix ␣C. Protein⅐substrate interactions involve the main-chain amide groups from Gly 135 to Val 138 of the beginning of helix ␣C accommodating the substrate P␣-group. Ser 137 O␥ contacts both P␣ and P␤ oxygen atoms. This phosphate group interacts further with the main-chain nitrogen of Gly 133 and P␥ with Thr 132 O␥. The substrate base moiety resides on a small hydrophobic pillow made up by the side chains of Ile 492 , Val 138 , and Leu 473 . The less well defined base (located on the molecular surface, see Figs. 5b, 5c, and 6a) may further interact with Gln 494 (Figs. 5-7). Contacts between the substrate and a vicinal protomer include the side chains of Arg 375 (with P␥ oxygen atoms) and Arg 124 (with the nucleotide sugar moiety O3* atom and P␥ oxygen atoms). Therefore, the side chains of both Arg 124 and Arg 375 (the latter putatively imbedded in a helicase superfamily II motif VI (20)) both rotate on a complex formation to FIG. 5. Electron density maps. Stereoview of a detail of the final models of the analyzed TrwB⌬N70 complexes around the P-loop and the NBS (displaying the protein chain from residue 128 to 141) and/or showing the bound nucleotide. The corresponding final sigmaa-weighted (2F obs Ϫ F calc ) type maps (1 contour; a, c, e, f) or the initial sigmaa-weighted 6-fold averaged (F obs Ϫ F calc ) type omit maps (2 contour; b, d) are further displayed. The resolution is indicated in parenthesis in each case. a, apo structure (2.5 Å); b, GDPNP (2.8 Å); c, GDPNP-bound substrate complex (2.8 Å); d, ADP and the putative Mg 2ϩ ion (3.0 Å); e, ADP⅐Mg 2ϩ -bound product complex (3.0 Å); f, product P complex-mimicking structure of TrwB⌬N70 with sulfate (2.4 Å).
accommodate and bind the substrate P␥ group (see below). These two residues do not have their net charges compensated by surrounding acidic residue side chains in the non-liganded structure (although they are both partially accessible to bulk solvent).
Interestingly, PcrA displays also a pair of basic residues in charge of liganding P␥ oxygen atoms, Arg 610 (from a helicase superfamily motif VI (20,52)) and Arg 287 (of motif IV). In the structures of both TrwB and PcrA, the NBSs are located at the interface between two domains, duplicated in PcrA (1A and 2A) and corresponding to two vicinal protomers in TrwB. Similarly, the arginine residues are provided by the vicinal domain (2A in PcrA, highly homologous to the P-loop containing domain 1A and, putatively, a product of a gene duplication) or protomer (TrwB). Arg 610 of PcrA and Arg 375 of TrwB are localized at the end of a topologically equivalent ␣ helix (␣N in TrwB), with their C␣ atoms just 2.3 Å apart. Arg 287 (PcrA) and Arg 124 (TrwB) are somewhat more far away (7 Å for their C␣ atoms), but their side chains come closer together and are localized on the same side of the NBS. Cag525 also displays a cluster of arginine residues contiguous to the ADP-binding site, also made up by two vicinal subunits, at whose interface the NBS is located (60). Furthermore, basic residues in the proximity of the phosphate oxygen atoms have been associated with transition state stabilization in G proteins and F 1 -ATPase (59,69,70) and to act as P␥ sensors in AAA ATPases (71). Two arginine residues have been also suggested to have important roles in ATP binding or hydrolysis in hexameric ATP-dependent helicase RuvB, which functions as a motor for branch migration at the Holliday junction during homologous recombination (72).
A strong salt bridge is observed in the apo form between P-loop Lys 136 N and Glu 357 O⑀1 (of DEXX motif II), which is weakened upon NTP binding, as inferred from the change in the position of the side-chain amine group (see Fig. 7b). The former atom moves toward the backbone carbonyl oxygen Gly 130 O, whereas the side chain of Glu 357 remains almost invariant (see Fig. 6b). This lysine residue is part of motif I and essential for the function of TrwB, because the single point mutant K136T abolishes conjugation (19). Other movements are observed upon NTP binding. Gln 386 , weakly binding P-loop Thr 132 O␥1 in the apo form, rotates around its C␣-C␤ bond toward the interior of the molecule contacting Ser 137 O␥ and Ser 406 O␥ (of a vicinal molecule). On the other hand, the side chain of Thr 132 rotates around its C␣-C␤ bond, so that its O␥1 atom contacts a substrate's P␥ oxygen atom.
The Product Complexes-Product ADP complex traps the structure in a hypothetical state, where an ADP molecule is still bound to the NBS once ATP has been (putatively) hydrolyzed and the P␥ has left the cleft. This structure is very similar to those of the substrate complexes, although it is distinct in that a bound Mg 2ϩ ion could be assigned (see Figs. 5d, 5e, and 6b). Similar to the NTP complex, the nucleotide sugar is liganded by the side chain of vicinal Arg 124 and the base resides on the same hydrophobic side chains. The cation interacts with Ser 137 O␥ and P␤ oxygen atoms, as proposed for the topologically equivalent catalytically active sites of F 1 -ATPase. The side chain of vicinal Arg 375 , which contacts the P␥ group in the substrate complex, somewhat penetrates into the free remaining space in the ADP complex missing P␥. In this way, the guanidinium group approaches the ADP P␤ group (Fig. 6a). Significant changes were observed when comparing the substrate complex with the product P complex, where a strongly bound sulfate anion (100% occupation; see Fig. 5f) may be interpreted as a leftover phosphate. This complex would constitute an alternative to the ADP complex, because during the hydrolytic mechanism one of the two products, ADP or phosphate, must leave the NBS first. When molecules A of each complex are superimposed, some strongly deviating regions appear despite the close overall similarity (r.m.s.d. of 0.47 Å; note negligible differences in the overall hexamer structure). In particular and besides the N-and C-terminal regions, segments Gly 130 -Arg 141 (encompassing the P-loop; r.m.s.d. of 1.1 Å), Gly 384 -Gln 390 (1.0 Å), and Gly 414 -Gly 415 (1.6 Å) strongly diverge, even for their main chains (Fig. 6c). The sulfate anion is located 2 Å away from the P␤ position of the substrate (Fig.  7, a and c). Vicinal monomer residues Arg 124 and Arg 375 move away, when compared with the GDPNP complex structure. The P-loop suffers a major rearrangement from Gly 130 to Arg 141 as does the surface-located C-terminal strand between Leu 490 and Phe 495 . The side chain of Ser 137 rotates, so that its O␥ atom points in the direction of the sulfate. Most importantly, the side chain of DEXX box Glu 357 rotates around its C␣-C␤ bond. This is accompanied by a main-chain and side-chain rearrangement of loop ␤7␣O. The region of TrwB mainly involved in these conformational changes and facing the interior channel displays a striking similarity in topology, but not in sequence, including the same regular secondary structure elements, with a structural feature described for AAA proteins, the N-terminal part of the second region of homology (72), or AAA minimum consensus region (57). This ϳ20-residue region mainly distinguishes AAA ATPases from other Walker-type ATPases (53,72,74) and is surface-accessible from the central channel, as ob- 6. Bound nucleotides. a, scheme of the interactions made by GDPNP (label GNP 701); b, ADP⅐Mg 2ϩ (labels ADP701 and MG2 702) with protein atoms. The involved residues are depicted together with their labels, with hydrogen bonds as dashed lines. (N) stands for residues from a neighboring protomer. The spiked circle segments surround hydrophobic van der Waals partners. served in NSF (56). This segment is proposed to play a role in ATP hydrolysis but not in binding (72). In TrwB this region runs from the end of strand ␤7 to the beginning of helix ␣P (see Fig. 1b). Although TrwB does not show any of the conserved residues, it is, however, noteworthy that this region appears to be involved in transmitting a movement from the exterior surface to the interior channel, activated by nucleotide-binding/hydrolysis. These changes could play a role in moving the ssDNA through the central channel during conjugation.
Considerations About the Hydrolytic Mechanism-ATP hydrolysis in DEXX box and other NTP-hydrolases implies a covalent bond formation between P␥ (or P␤) and an attacking solvent molecule, giving rise to a trigonal bipyramidal, negatively charged transition state species, and subsequent breakage of the O3␤-P␥ (or O3␤-P␤) scissile bond. This intermediate may be stabilized by a Mg 2ϩ cation, coordinated to the P␤ and/or P␥ oxygen atoms of the substrate, the highly conserved P-loop lysine side chain, and possibly the main-chain amide groups of P-loop motif I (20,21,75). Further basic residues in the pocket may act additionally as P␥ sensors (see above). Two acidic residues may act as a general base, polarizing the attacking solvent molecule, and may coordinate the catalytic cation, respectively (56, 76 -78). These two residues are possibly Glu 96 and Asp 144 , the latter of which if from the DEXX box   FIG. 7. Structure comparison. a, superimposition of the P-loops (running from residues Ala 131 to Ser 137 in each case) of the several analyzed structures after optimal least squares fitting of a monomer and of the vicinal Arg 124 and Arg 375 residues. The positioned molecules occupying the NBS are also displayed. Color coding: apo structure in yellow, AD-PNP complex in blue, GDPNP complex in violet, ADP⅐Mg 2ϩ complex in green, and P-mimicking sulfate complex in magenta. Superimposition of the final models as sticks of the (b) apo structure (yellow) and the substrate complex emulating GDPNP complex structure (pink) and (c) the substrate complex emulating structure (violet) and the product P complex mimicking structure (cyan). The relative position of the central channel and the exterior surface are shown. Some residues are labeled as are the bound GDPNP and SO 4 2Ϫ molecules.

Structure of Type IV Conjugative Coupling Protein TrwB
motif I (20), in the archetypal RecA enzyme (46,47). This motif is also essential in monomeric eIF4A RNA helicase ATPase domain, where Asp 169 of a DEAD box (Asp 169 -Asp 172 ) is shown to bind the Mg 2ϩ cation and proposed to act also as the general base (76), and in the structurally related monomeric DNA helicases Rep and PcrA (here the DEXX box encompasses Asp 223 and Glu 224 ). In these latter cases, the aspartate residues are binding the cation and the glutamate residues are suggested to function as the general base (52,68,79), although spatially at a different position from that of Glu 96 in RecA.
Our protein displays a DEXX box motif II (including Asp 356 and Glu 357 ), which is topologically equivalent to this motif in PcrA. Both residues, however, do not point their side chains toward the bound GDPNP P␥ group moiety. A rotation around the C␣-C␤ and C␤-C␥ bonds of Asp 356 would position its O␦2 atom 6 Å away from a P␤ oxygen. The same distance (ϳ6 Å) holds for the immediately vicinal Glu 357 , involved, prior to NTP-binding, in a salt bridge with the tip of Lys 136 of the P-loop, as observed in eIF4 ATPase domain (76). Accordingly, these residues would only play the postulated catalytic role for PcrA (52) or eIF4 ATPase domain (76) if the substrate was bound in an extended conformation, thus penetrating deeper into the molecular structure than we see in our complex (Fig.  5e). Also, TrwB Asp 158 (of a putative Asp box) would be too far away from the substrate. Thus, we do not find a candidate in the surroundings of the bound inhibitor susceptible of acting as a general base in a hydrolytic mechanism, apart from Thr 132 of the P-loop. A similar lack of general base candidates has been described for p21 ras , and, in this case, the ␥ phosphate was proposed to be the catalytic base (80). In NSF domain 2 (56), also displaying the DEXX box, the (essential) glutamate is also, as in TrwB, too far away to polarize a single attacking solvent molecule. This led to the proposal that the glutamate of this conserved motif could very well play the role of the general base but through solvent molecules in a kind of cascade reaction (56). This could also be a working hypothesis for ATP-hydrolysis in TrwB. Mutation experiments are under way to address this issue.