Crystal Structure of T7 Gene 4 Ring Helicase Indicates a Mechanism for Sequential Hydrolysis of Nucleotides

Cell, Vol. 101, 589–600, June 9, 2000, Copyright 2000 by Cell Press

Crystal Structure of T7 Gene 4 Ring Helicase Indicates a Mechanism for Sequential Hydrolysis of Nucleotides Martin R. Singleton,* Michael R. Sawaya,† Tom Ellenberger,† and Dale B. Wigley*‡ * Sir William Dunn School of Pathology University of Oxford South Parks Road Oxford OX1 3RE United Kingdom † Department of Biological Chemistry and Molecular Pharmacology Harvard Medical School Boston, Massachusetts 02115

Summary We have determined the crystal structure of an active, hexameric fragment of the gene 4 helicase from bacteriophage T7. The structure reveals how subunit contacts stabilize the hexamer. Deviation from expected six-fold symmetry of the hexamer indicates that the structure is of an intermediate on the catalytic pathway. The structural consequences of the asymmetry suggest a “binding change” mechanism to explain how cooperative binding and hydrolysis of nucleotides are coupled to conformational changes in the ring that most likely accompany duplex unwinding. The structure of a complex with a nonhydrolyzable ATP analog provides additional evidence for this hypothesis, with only four of the six possible nucleotide binding sites being occupied in this conformation of the hexamer. This model suggests a mechanism for DNA translocation. Introduction Helicases are found in all living organisms and participate in almost every process that involves nucleic acids (Lohman and Bjornson, 1996). The basic activity of a helicase is the separation of nucleic acid duplexes into their component strands, a process that is coupled to the hydrolysis of NTPs. However, there are 12 known DNA helicases in E. coli that perform a variety of tasks in nucleic acid metabolism, ranging from simple strand separation at the replication fork (e.g., DnaB helicase) to more elaborate processes such as the migration of Holliday junctions (e.g., RuvAB complex). Helicases can be divided into two classes on the basis of mechanism, those that translocate in a 5⬘–3⬘ direction along singlestranded DNA and those that operate with the opposite polarity. For the 3⬘–5⬘ helicases, there has been a considerable amount of recent structural and biochemical data to enlighten our understanding of the mechanism of these enzymes. The first crystal structure of a helicase was that of PcrA (Subramanya et al., 1996). The structure revealed that the enzyme comprised several domains, including two domains that are structurally similar to ‡ To whom correspondence should be addressed (e-mail: wigley@

eric.path.ox.ac.uk).

the ATPase domain of the recombination protein RecA (Story and Steitz, 1992), despite minimal sequence homology with one another or with RecA. The E. coli Rep helicase (Korolev et al., 1997) and NS3 RNA helicase (Kim et al., 1998) were also crystallized with singlestranded DNA, giving the first glimpses of how the proteins interact with nucleic acids. Subsequent structures of PcrA helicase in substrate and product complexes with a DNA substrate (Velankar et al., 1999) have provided snapshots along the catalytic pathway for this enzyme family, which suggest an inchworm mechanism (Yarranton and Gefter, 1979). Biochemical data also support this mechanism and reveal that the step size for translocation along single-stranded DNA is one base for each ATP that is hydrolyzed (Dillingham et al., 2000). For the 5⬘–3⬘ helicases, there has been rather less structural information. Until recently, structural studies were limited to those utilizing electron microscopy techniques, such as reported for a variety of hexameric ring helicases including DnaB (Yu et al., 1996b; San Martin et al., 1998), Rho (Oda and Takanami, 1972; Gogol et al., 1991), and bacteriophage T7 gene 4 protein (Egelman et al., 1995; Yu et al., 1996a). These studies have demonstrated the conformational flexibility of the ring helicases in that species with both six-fold and three-fold symmetries, as well as many other forms, can be observed in a number of different systems. However, the functional significance of these conformations is unknown. The situation is complicated further by the finding that assembly of the hexamer involves stable dimeric and trimeric intermediates (Patel and Hingorani, 1993; Bujalowski et al., 1994; Bird et al., 1997), suggesting that the proteins could either be considered as “dimers of trimers” or “trimers of dimers.” For DnaB, there is clear evidence for a stable trimeric species in the absence of magnesium (Bujalowski et al., 1994), whereas for T7 gene 4 proteins the evidence is less clear and likely involves stable dimeric species as well as trimers and higher order oligomers (Patel and Hingorani, 1993; Bird et al., 1997). Again, the functional relevance of these observations to the mechanism of the enzymes is unclear. Ring helicases frequently contain domains in addition to those required for helicase activity. For example, the T7 gene 4 protein comprises separate helicase and primase domains (Bird et al., 1997; Frick et al., 1998). Crystal structures of the N-terminal “accessory” domains of DnaB (Fass et al., 1999) and Rho (Alison et al., 1998) helicases have been reported, but the structure of the core “helicase” domain remains undetermined in these systems. The first high-resolution glimpse of the structure of a helicase domain from this family was provided by the crystal structure of the bacteriophage T7 4E protein (Sawaya et al., 1999), a fragment of the full-length T7 gene 4 protein that is only slightly shorter than that required to retain helicase activity or to form hexamers (residues 272–566; Bird et al., 1997; Guo et al., 1999). The structure confirmed predictions that this region of the protein would have structural homology with the ATPase domain of RecA (Bird et al., 1998). The fragment

Cell 590

also retains the ability to bind and hydrolyze nucleoside triphosphates (Guo et al., 1999). Interestingly, although this protein has lost the ability to hexamerize, it forms helical filaments that are highly reminiscent of those formed by RecA (Story and Steitz, 1992). By examining the gene 4E filament down the screw axis, a model was proposed for hexamer formation that involved the N-terminal region of the protein fragment. Although structural studies of the T7 gene 4E helicase have revealed details about the protein fold, we know little about the physical basis for coupling nucleotide binding and hydrolysis to the DNA-unwinding activity. Biochemical studies of hexameric 5⬘–3⬘ helicases have shown that NTP binding and hydrolysis are highly cooperative processes (Bujalowski and Klonowska, 1993; Hingorani and Patel, 1996; Stitt and Xu, 1998) and that the activities of the six potential ATP binding sites within the hexamer are regulated differentially (Hingorani et al., 1997; Stitt and Xu, 1998). Based on such studies, a mechanism for nucleotide binding and hydrolysis has been proposed for T7 gene 4 protein (Hingorani et al., 1997) and the Rho helicase (Stitt, 1988; Stitt and Xu, 1998) that is akin to the “binding change” mechanism proposed for the F1-ATPase (reviewed in Boyer, 1993). In this model, asymmetry of the six subunits that comprise the ring is utilized to control the sequential binding and hydrolysis of nucleotides during catalysis, such that there are only three, rather than six, active NTPase sites in the hexamer. The functional asymmetry of the F1-ATPase is a consequence of the alternating arrangement of noncatalytic (␣) and catalytic (␤) subunits and their interactions with the central rotor, the ␥ subunit (Abrahams et al., 1994). How this might be accomplished by the six identical subunits of the hexameric helicases remains unclear. Using limited proteolytic digestion, we have previously identified fragments of the T7 gene 4 proteins that comprise the helicase domain that we refer to as 4C (residues 219–566) and 4D (residues 241–566) (Bird et al., 1997). Both fragments form stable hexamers and are active helicases, although the 4D fragment has reduced DNA-unwinding activity compared to 4C. We have now determined the crystal structure of the 4D fragment at 3.0 A˚ resolution. The structure of the hexameric protein shows unexpected deviations from sixfold rotational symmetry, suggesting a mechanism for sequential hydrolysis of bound NTPs. Further evidence for this mechanism is provided by the structure of a complex of the protein with a nonhydrolyzable analog of ATP (ADPNP) that binds to only a subset of the possible ATP binding sites in the hexamer. The model also suggests a mechanism for ssDNA translocation. Results and Discussion Structure of the Hexamer Using the 4E fragment structure as a model (Sawaya et al., 1999), the orientation and position of the equivalent regions of the 4D hexamer were determined by molecular replacement. The positions of three subunits of a hexamer were located initially and found to form one half of a hexamer, with the other half related by a crystallographic two-fold symmetry axis. Phases calculated

Figure 1. Hexamer and Subunit Contacts (A) Overall fold of the monomer. The chain is colored from blue at the N terminus to red at the C terminus. (B) The hexamer appears to be stabilized via interactions between the N-terminal region (residues 264–284) of one subunit and a pocket on its partner (residues 364–395). The four bound ADPNP molecules are overlaid in ball and stick represention. This is the view from the N-terminal side of the ring.

from the molecular replacement model were then used to to determine the positions of heavy atoms in a gold derivative (see Experimental Procedures). It was immediately evident from these difference electron density maps that a second set of gold sites was present, corresponding to another hexamer in the crystal. The coordinates of the gold sites were used to map the NCS operators relating the first hexamer to the second. Rigid body refinement was then used to determine the positions of each molecule more accurately. At this stage, the molecular replacement solution was confirmed by locating the heavy atom sites in three derivatives using phases calculated from the model coordinates, and checking that these sites (11 gold, 52 selenium, and 6 tungstate) were the same in each subunit. This initial model revealed how the closed, ringshaped hexamer is assembled (Figure 1), which, in general terms, was much as predicted from the structure of the 61-screw filament structure determined for the

T7 Gene 4 Ring Helicase 591

gene 4E fragment (Sawaya et al., 1999). The fold of each subunit is similar, with the N terminus of the polypeptide chain beginning on one face of the ring and the C terminus on the other. The subunit interface is stabilized by interactions between an ␣ helix in the N-terminal region (residues 271–280) and an adjacent subunit. However, the T7 4D helicase forms stable hexamers, unlike the 4E fragment, which is 31 residues shorter. The additional N-terminal extension of the 4D fragment includes a ␤ strand along the edge of a sheet (residues 265–267), but the remainder of this region of the structure (residues 241–260) is poorly ordered in the crystal structure. The T7 gp4 helicase and other 5⬘–3⬘ ring helicases have a preference for forked DNA substrates with two single-stranded tails of sufficient length to allow the protein to assemble on the DNA and begin unwinding the duplex (Ahnert and Patel, 1997; Kaplan and Steitz, 1999). The 5⬘ tail of the forked DNA passes through the center of the ring (Yu et al., 1996a), and the 3⬘ tail is thought to contact the outside of the ring (Ahnert and Patel, 1997). There is no obvious DNA binding surface on the outside of the gp4D ring, but several loops facing the central channel of the ring are implicated in binding to DNA. Mutational data (Notarnicola et al., 1995; Washington et al., 1996) suggest that residues within loop I (residues 424–439) and loop II (residues 464–475) interact with DNA, like the analogous loops of the related RecA structure (Story and Steitz, 1992). Consistent with this proposal, the loops contain a number of conserved basic residues that line the central cavity of the hexameric ring. In the structure of the 4E fragment, these loops are disordered but appear to become more ordered when the protein is complexed with nucleotide (Sawaya et al., 1999). By contrast, in the hexameric structure, loops I and II and a third loop (loop III; residues 503–513) are well ordered in both the apo and ADPNP complex structures (Figure 2). The loops extend into the central cavity, effectively reducing the space such that it would now appear impossible for a DNA duplex to pass through the center of the ring. Mutational data (Notarnicola et al., 1995) have also implicated loop II and the region adjacent to this loop as playing a role in hexamer stabilization. This loop contributes to a contact region between subunits that is not formed in the 4E structure and probably explains why this region is better ordered in the 4D protein. Asymmetry of the Hexamer Given that the protein comprises six identical protein subunits, we anticipated that the hexamer would have the six-fold symmetry shown previously in electron microscopy studies of the gene 4 proteins (Egelman et al., 1995). However, when we undertook molecular replacement with the 4E protein coordinates, the orientations of the subunits of the first hexamer indicated significant departures from the expected symmetry. The two independent hexamers both show similar departures from six-fold symmetry, despite their different packing contacts, suggesting that the conformation of the hexameric ring in the crystals represents a conformation that is of biochemical significance rather than simply a distortion due to crystal contacts. The nonisomorphism that we see between crystals (see Experimental Procedures) is

Figure 2. Structure and Location of the DNA Binding Loops (A) Ribbon diagram illustrating the structure of the proposed DNA binding loops that line the central cavity of the ring. The four bound ADPNP molecules are overlaid in black. Loop I (residues 424–439) is in blue, loop II (residues 464–475) is red, and loop III (residues 503–513) is green. The view is the same as that in Figure 1B. Lettering refers to the notation used in the text. (B) The same representation of the molecule, but rotated to show a view from the inside of the ring looking outward. The C-terminal face of the ring is uppermost in this view. Note that the loops form a spiral across the surface of the protein. For clarity, only three subunits are shown. The dotted line marks the symmetry axis of the hexamer.

a consequence of slightly differing orientations of the monomers around the ring indicating the conformational flexibility that is a feature of the hexamer. However, the variability between crystals relates to very small changes in subunit orientations (less than 2⬚) particularly when compared to the large distortions of the ring (up to 30⬚) that are discussed below. In order to analyze the conformation that we observe, we first compared our structure with that of the sixfold symmetric ring model proposed for the T7 gene 4E fragment proposed previously (Sawaya et al., 1999). While this was clearly an approximation for a ring, we thought that it would at least give us a template against which to examine the conformational changes observed in our structure (see Figure 3). However, our initial analysis revealed that none of the two-fold related pairs of subunits were a good fit on those of the modeled ring. Consequently, we instead prepared a six-fold symmetric ring based on the orientation of the “A” subunit. The relationship between the “B” subunits and equivalent

Cell 592

Figure 3. Deviations from Six-Fold Symmetry (A) A stereo diagram of the C␣-backbone of a perfect six-fold symmetric molecule (based on the A subunit and its symmetry equivalent) is in red, and the subunits of the 4D hexamer that deviate from this arrangement (the B and C subunits and their symmetry equivalents) are in blue. (B) The same model shown rotated by 90⬚ to give the same view as in Figure 2B. The deviations of each monomer from its six-fold symmetric equivalent are shown. For clarity, only three subunits are shown. (C) The GRASP molecular surface (Nicholls and Honig, 1991) of the 4D hexamer from the same view as in (A). Positively charged surface is blue, and negatively charged surface is red.

subunits of the modeled ring is a rotation of approximately 15⬚ about an axis lying in the plane of the ring, while that of the “C” subunits is 30⬚. Subunits B and C are also 5 A˚ closer to the center of the ring. The subunit rotations are a swiveling within the plane of the ring, with the rotations from A to B to C being a roll toward the central cavity of the ring as viewed from the C termi-

nus of the subunits. Consequently, the rotation from subunit C back to the symmetry equivalent of A constitutes a 30⬚ roll backward coupled with a 5 A˚ shift outward away from the center of the ring. This makes the NTP binding sites on subunits A and B most similar and those in the C subunits most different. Another important consequence of these subunit rotations is to place the

T7 Gene 4 Ring Helicase 593

DNA binding loops in a spiral pattern around the central cavity of the hexamer (Figure 2). Electron microscopy studies of some hexameric helicases report a variety of conformational states for a given protein sample suggestive of a dynamic equilibrium (Oda and Takanami, 1972; Gogol et al., 1991; Yu et al., 1996b; San Martin et al., 1998). For Rho helicase, many conformations exist including some that are referred to as “notched” or “lockwasher” structures (Oda and Takanami, 1972; Gogol et al., 1991). Other studies conclude that there are at least two major conformational states, one with six-fold rotational symmetry and another with three-fold symmetry (Yu et al., 1996b; San Martin et al., 1998). However, it is evident that other forms also exist that have been interpreted as being intermediates between the major conformational states. Almost all of the EM studies are on the enzymes in the absence of DNA or RNA. In these studies, the presence of any rotational symmetry was exploited to improve the resolution of the images. The exception is the study of the gene 4 protein complexed with ssDNA (Yu et al., 1996a). In this case, because the DNA only binds at one site in the hexamer, the six-fold symmetry could not be utilized to enhance the images of the complexes, as this would have resulted in averaging out of the density for the bound DNA. This observation reveals an underlying asymmetry in the protein, as only one site can bind ssDNA at any one time. Interestingly, a similar departure from three-fold symmetry is seen in the crystal structure of the bovine F1-ATPase (Abrahams et al., 1994), a protein that has structural and mechanistic similarities with hexameric helicases (Hingorani et al., 1997; Sawaya et al., 1999). For F1-ATPase, the expected three-fold symmetry of the ␣3␤3 component is broken due to the asymmetric interactions with the single ␥ subunit of the complex. Importantly, in the absence of the ␥␦⑀ subunits, the ␣3␤3 complex retains exact three-fold symmetry (Shirakihara et al., 1997). This analogy would suggest that in the absence of ssDNA hexameric helicases could also adopt three- or six-fold symmetric structures, but that once ssDNA is bound to the enzyme there has to be a departure from this symmetry. The implications for the interpretation of our structure are discussed below.

The NTP Binding Sites The gene 4 helicase can hydrolyze a number of different NTP and dNTP cofactors (Matson and Richardson, 1983; Matson et al., 1983), although dTTP appears to support most efficient DNA unwinding for unknown reasons (Hingorani and Patel, 1996; Sawaya et al., 1999). The NTP binding site is at the expected position adjacent to several of the conserved helicase motifs, and different nucleotides (e.g., dATP and dTTP) appear to bind in a similar fashion (Sawaya et al., 1999). We were able to obtain a complex of the protein with a nonhydrolyzable ATP analog (ADPNP) and Mg2⫹, either by cocrystallization or by soaking crystals (Figure 4). Although the open packing arrangement allows free access to all of the potential NTP binding sites, not all of these contained bound nucleotide. There was significant difference electron density to account for a bound ADPNP in the “A” and “B” sites of both hexamers, but there was no density

visible in the “C” sites (Figures 4A–4C). The binding of ADPNP to only four of the six possible sites in each hexamer reveals an important feature of the enzyme that is evidence for a binding change mechanism for the hydrolysis of NTPs (see below). Other than minor movements of side chains that contact the bound nucleotide, the structures of the apo protein and the ADPNP complex are very similar. Residues that contribute to nucleotide binding in site B are illustrated in Figure 4D. A striking similarity to the ATP binding site of RecA (Story et al., 1992; Sawaya et al., 1999) is evident, with a number of residues that appear to be equivalent in the two systems. The lysine of the conserved motif I (K318) contacts the ␤ phosphate and probably serves to stabilize the transition state during hydrolysis. There is density for a bound magnesium ion that is coordinated by the ␤ and ␥ phosphates and the side chains of S319 and D424. The adenine ring of the ADPNP is stacked between the side chains of Y535 and R504. The active site also shares considerable similarity with the ATP binding site of PcrA helicase (Velankar et al., 1999). However, interesting differences include a substitution of the conserved glutamine residue in PcrA (Q254) with a histidine (H465) in 4D. In PcrA, Q254 serves as a switch between the ATP binding site and a loop at the ssDNA binding site that includes the crucial W259 and R260 residues (Dillingham et al., 1999). H465 in the 4D protein contacts the ␥ phosphate, in a manner analogous to Q254 in PcrA, and is adjacent in sequence to an extended loop of the structure that contains a number of conserved basic residues (loop II). It is thought likely that this loop also contacts the bound DNA (Notarnicola et al., 1995; Washington et al., 1996), suggesting a common mechanism for communication between the ssDNA binding site and the nucleotide binding site in helicases. In both PcrA and gene 4D protein, the ␥ phosphate is also contacted by an arginine residue (R610 in PcrA and R522 in 4D). Interestingly, in both cases this arginine side chain comes from an adjacent RecA-like domain (domain 2A in PcrA and a different subunit in 4D), although the relative disposition of the two domains is very different in each of the systems. Surprisingly, the arginine arises from a different region in each of the structures, an ␣ helix in PcrA and a ␤ hairpin in gene 4D. Arginine residues have been reported to contribute to NTP binding and hydrolysis in a similar fashion in a variety of other NTP-hydrolyzing systems, interactions that have been termed the “arginine finger” (Wittinghofer, 1998). Replacement of R610 by an alanine in PcrA results in a protein with severely impaired ATPase activity (Soultanas et al., 1999). By analogy, it is likely that R522 contributes in a similar way in the 4D NTP binding site, and it is therefore significant that the position of R522 is not conserved in the three different sites (Figure 4). In the sites that bind ADPNP (sites A and B), R522 is in close proximity (ⵑ3.5 A˚) to the ␥ phosphate of the bound nucleotide, but in the C sites this residue is displaced by more than 10 A˚ from the equivalent position in the potential ␥ phosphate binding site and therefore cannot contribute either to NTP binding or hydrolysis. In the dATP and dTTP complexes of the gene 4E fragment, R522 is 4 A˚ from the ␥ phosphate, sufficiently close to allow binding of the nucleotide despite the distortion

Cell 594

Figure 4. ATP Binding Sites in the Hexamer (A) ADPNP difference density in an A site. (B) ADPNP difference density in an B site. (C) The C site showing a lack of difference density. (A–C) The difference electron density in (A)–(C) is contoured at 3␴. The difference density at site B is greater than that at site A, suggesting a higher occupancy of bound ADPNP at site B. The lower occupancy at site A might suggest that this site is actually in a conformation that would favor binding of NDP ⫹ Pi rather than NTP. The side chain of residue Tyr-535 marks the adenine binding pocket and is shown to illustrate the displacement of Arg-522 in the C site compared with the A and B sites. The ADPNP is green. The color scheme for the protein chains is the same as in Figure 1B. The views are chosen to be equivalent for each nucleotide binding site. (D) Residues that contribute to NTP binding (site B). Residues from subunit B are labeled in black, but Arg-522 comes from the adjacent C subunit and is labeled in pink. The protein chain of the B subunit is silver, the C subunit is blue, and the ADPNP is green.

of the active site due to filament rather than hexamer formation (Sawaya et al., 1999). The other components of the nucleotide binding sites are similar in the two structures. The asymmetry of the NTP binding sites has an important implication. In order to create a site that is competent to bind NTP, there has to be a rotation of the subunits of the ring to allow a 15⬚ displacement from a six-fold symmetric structure. However, if this rotation were to occur at every subunit interface around the ring, there would be a 90⬚ displacement of the subunits at the final position. Evidently, this would not be possible, with the consequence that in order to allow for the 15⬚ relative rotation at a particular site there has to be an equal and opposite rotation at another site or sites. For a six-fold symmetric ring, the strain is relieved evenly around the ring in the absence of bound nucleotide. However, on binding NTP, the rotation of subunits to make competent NTP binding sites sets up strain at the subunit interfaces. In the conformation that we observe, four subunit interfaces have the required 15⬚ rotation, but the strain this creates in the ring is absorbed by

the final two sites, which undergo 30⬚ rotations in the opposite direction. This has the consequence that these interfaces are displaced by 45⬚ from the alignment that would be competent to bind nucleotide. The observation that filaments are formed by the gene 4E protein (Sawaya et al., 1999), which is compromised in its ability to form hexameric rings, probably arises from the natural tendency of the subunits to associate with a 15⬚ twist. This also explains why nucleotide is able to bind at all of the sites in the filament. Similarly, the observation that RecA protein is also able to form both filaments and rings might be explained by this proposal (Yu and Egelman, 1997), as would the high proportion of “notched” and “lockwasher” images of Rho helicase that are observed by electron microscopy (Oda and Takanami, 1972; Gogol et al., 1991) representing rings that have failed to close. Consequently, ring formation takes place with an energetic cost with regard to NTP binding because there has to be distortion from six-fold symmetric ring contacts in order to be able to bind nucleotide. It is this energetic cost that gives rise to the negative cooperativity of NTP binding that is characteristic of ring

T7 Gene 4 Ring Helicase 595

Figure 5. A Binding Change Mechanism for Hexameric Helicases A four-site binding change model for hexameric helicases. The conformation that we observe in the crystal has four sites that are competent to bind nucleotides. In step I, we propose that the two sites we observe with high occupancy (the B sites) bind NTP, while the two sites with lower occupancy (A sites) would actually bind NDP ⫹ Pi. The remaining two sites are empty. Blue subunits contain bound NTP, yellow have NDP ⫹ Pi, and magenta have no bound nucleotide (empty). The notation of the subunits as A, B, or C is maintained as in the previous figures. To progress to step II, hydrolysis of the bound NTPs results in conformational changes around the ring such that bound NDP ⫹ Pi dissociate from the protein and the empty sites are now able to bind NTP. The same events are repeated in order to progress to step III but take place at different subunits. Three cycles of NTP hydrolysis are shown to illustrate the nucleotide binding state of each subunit during successive cycles of the reaction and that all six NTP sites are utilized at different points during sequential cycles of hydrolysis. Therefore, ATP hydrolysis can be regarded as a ripple going around the ring without requiring a rotation of the ring itself. Note also that, although we indicate that two NTPs are hydrolyzed per cycle, this need not be a requirement. The presence of ssDNA bound to one site, for example, could provide the required asymmetry.

helicases, that is, binding at the first sites disfavors binding of NTP at subsequent sites in the hexamer (Stitt, 1988; Hingorani and Patel, 1996; Hingorani et al., 1997; Picha and Patel, 1998; Stitt and Xu, 1998; Kim et al., 1999). Thus, the structural asymmetry seen in the 4D hexamer can explain how the sequential binding and hydrolysis of nucleotides might be controlled in hexameric helicases. The surprising observation that the truncated 4C and 4D fragments form more stable hexamers than the full-length protein (Bird et al., 1997; Guo et al., 1999) suggests that the strain resulting from hexamer formation can be relieved more easily in these proteins, probably because of the reduced subunit contacts. This proposal would explain why we have managed to trap the asymmetric conformation of the protein even in the absence of nucleotide because the energetic cost of distorting the ring is less in the 4D protein than in 4A, thus allowing a more dynamic equilibrium in solution. A Mechanism for Sequential Hydrolysis of NTPs It has been shown that hexameric helicases bind and hydrolyze NTPs in a highly cooperative manner. This property of the enzyme has been studied in detail for the T7 and Rho helicases (Stitt, 1988; Hingorani and Patel, 1996; Hingorani et al., 1997; Stitt and Xu, 1998), which has led to the proposal of a binding change mechanism for NTP hydrolysis that is generally similar to that proposed for the F1-ATPase (Boyer, 1993; Abrahams et al., 1994). Although the apparent similarity between a three-site binding change mechanism and that of F1-ATPase is seductive, there is a major difference between the two protein systems that needs to be considered. In both systems, there are potentially six NTP binding sites, but in F1-ATPase only three subunits are catalytic due to the ␣3␤3 stoichiometry of the complex. For hexameric helicases, with six identical subunits,

there is no reason a priori for an alternating arrangement of active sites or for the identities of the active subunits to remain fixed during multiple rounds of catalysis. These considerations, together with the unexpected structural asymmetry of the hexamer in our structure and our direct observation of four bound nucleotides in the complex, have led us to propose a modified version of the binding change mechanism (Figure 5). For this model, in common with the classical binding change mechanism, the binding and hydrolysis of NTPs is sequential, but instead of involving three NTP binding sites at any one stage in the cycle, our model utilizes four with the remaining two sites being unoccupied. The interactions in the NTP binding sites of the 4D structure demonstrate how this could be accomplished by repositioning the arginine finger (R522) in different sites, with only those sites that have an R522 in close enough proximity to the ␥ phosphate of the NTP having sufficient avidity for nucleotide. The loop of residues flanking R522 (K520–G525) also contributes to the edge of the nucleotide pocket. These residues are displaced in the empty site, probably contributing further to the low affinity of this site for nucleotide. We believe it to be likely that one of the pairs of sites, that which appears to have the lower occupancy (site A), most probably is set up to bind NDP ⫹ Pi rather than NTP. With this simple assumption, we can propose a cycling of occupancy of the sites between NTP bound, NDP ⫹ Pi bound, or empty in a concerted manner that could be coupled to the helicase activity of the protein (see below). A fundamental difference between the mechanism that we propose and the three-site model suggested previously is that in our model all six of the potential active sites are utilized at some point during multiple (processive) rounds of helicase activity, although a maximum of four sites have nucleotide bound at any one

Cell 596

time. In a three-site model, processive helicase activity only involves the same three of the six possible active sites once the initial (unexplained) selection of sites has been established. One advantage of the four-site model, therefore, is that there is no requirement for differential properties of identical polypeptide subunits. Although studies of the biochemical properties of hexameric helicase proteins consisting of mixtures of mutant and wild-type subunits ought to help distinguish between these models (Notarnicola and Richardson, 1993; Patel et al., 1994), in practice such experiments are very hard to interpret. The complexity of the analysis required to estimate the proportion of each of the seven different heterohexameric species that would be present in a dynamic mixture of proteins is not straightforward even assuming that they are in free exchange, which is not the case in the presence of ssDNA (Patel et al., 1994). However, these experiments are more revealing when considering the model for DNA translocation that is discussed below. Interpretation of other biochemical evidence that might distinguish between three- or four-site models is also complicated. The basic features of the kinetics of ATP hydrolysis are that there are two high affinity active sites on each hexamer, one of which turns over rapidly while the other is slower (Hingorani et al., 1997; Stitt and Xu, 1998). Equilibrium binding experiments show that there are either three or four sites that can be occupied by bound nucleotides at any one time (estimates include 2.7–4.3 for Rho helicase [Stitt, 1988; Stitt and Xu, 1998; Kim et al., 1999] or 2.4–4.3 for T7 gene 4 protein [Hingorani and Patel, 1996; Hingorani et al., 1997; Picha and Patel, 1998]), demonstrating that not all of the active sites that are able to bind NTP are able to catalyze hydrolysis at any one time. Distinguishing between three or four bound nucleotides becomes a critical point in light of our structural data. Significantly, when nonhydrolyzable analogs are used for the experiments, the estimated number of sites is most commonly four, but is nearer to three when hydrolyzable NTPs are used (Stitt, 1988; Hingorani and Patel, 1996). Experimental evidence from studies of Rho helicase shows that at least four sites contain bound NTP during catalysis (Kim et al., 1999). Consequently, interpretation of the biochemical data is not sufficiently clear cut to differentiate between mechanisms that involve three or four bound nucleotides at the present time. The structural data, however, clearly point to a four-site model. ssDNA Binding Site For a variety of hexameric helicases, it has been shown that there is a single site for the binding of ssDNA to each hexamer (Hingorani and Patel, 1993; Jezewska et al., 1996), which is located asymmetrically within the ring and involves only one or two subunits (Yu et al., 1996a). The ssDNA binding sites have been shown to involve loops I and II (Notarnicola et al., 1995; Washington et al., 1996). The location of these loops in the crystal structure is consistent with the ssDNA binding site proposed from electron microscopy studies (Yu et al., 1996a). Unfortunately, soaking ssDNA into our present crystals cannot help us to examine details of the interaction

Figure 6. A Mechanism for ssDNA Translocation (A) Representation of the hexameric ring such that the subunits are laid out flat in two dimensions. The color scheme and subunit notation used are the same as in Figure 5. The DNA binding loops are red, and the ssDNA is colored alternately orange and black to illustrate passage through the ring, with each band of color representing the step size for translocation. (B) The same mechanism but viewed from the side to show translocation of the DNA through the ring. The ssDNA segment that is contacting the protein is shown in cyan. The front three subunits of the ring have been made transparent for clarity. Note that the DNA tracks around the inside of the stationary protein ring, with neither molecule needing to rotate significantly.

with DNA. Both hexamers in the crystal are aligned with crystallographic two-fold symmetry axes, and therefore any bound ssDNA would be expected to occupy either of two possible sites with equal occupancy averaging out at a maximum of 50% occupancy at each site, resulting in electron density that would be too poor at the resolution of our present data. Although there are two potential ssDNA binding sites in each hexamer, only a single site for ssDNA can probably be occupied because steric constraints in the narrow central cavity would preclude binding of a second molecule. It seems most likely that either of the two sites site could be occupied by ssDNA initially, but that once a commitment has been made to a particular site only one site can operate during strand translocation. A Model for DNA Translocation The asymmetry of the hexamer has led us to consider how this conformation might relate to helicase activity. We propose above how NTP binding and hydrolysis can be coordinated around the ring, but it is also possible to relate these conformational changes to translocation along ssDNA (Figure 6). Nuclease protection studies (Hingorani and Patel, 1993) and electron microscopy (Egelman et al., 1995) have revealed that ssDNA passes through the gene 4 helicase ring with a contour length

T7 Gene 4 Ring Helicase 597

Table 1. Crystallographic Statistics Data Collection

Apo

ADPNP Complex

Au

Se

W

Resolution (A˚) Completeness (%) Rsym (%) Rderiv (%) Number of sites Phasing power

20–3.3 94.5 5.7 — — —

20–3.0 97.9 6.8 — — —

20–4.5 96.8 8.3 31.4 11 1.3

20–4.2 96.8 9.4 24.5 52 1.2

20–4.2 93.5 8.6 23.7 6 1.0

Overall mean figure of merit

0.36

Final Models

Apo

ADPNP Complex

R factor (%) (all data) Rfree (%) (5% of data) Rmsd bond length (A˚) Rmsd bond angle (⬚)

25.1 30.7 0.011 1.6

25.3 30.4 0.012 1.7

of about 3 A˚ per nucleotide (i.e., similar to that of B form DNA). The asymmetry of the hexameric ring results in relative displacements of the ssDNA binding loops to form a spiral along the inside of the ring (Figure 2). This spiral is complementary to the shape of the phosphodiester backbone of a ssDNA strand in approximate B form conformation, suggesting that the ssDNA could also follow a spiral track along the surface of the protein. However, the spiral pattern is broken at a pair of opposite subunit interfaces due to the 30⬚ distortion of adjacent subunits in the ring; thus, the ssDNA binding site can only extend across two, or at most three, subunits. There is a precedent for this mode of binding in other helicases. The crystal structure of the hepatitis C virus NS3 helicase complexed with ssDNA shows that eight bases of ssDNA bind across two RecA-like domains in a conformation that is very similar to one strand of a B form DNA duplex (Kim et al., 1998). When the protein binds and hydrolyzes NTP, undergoing the conformational changes described above, this could be coupled to ssDNA translocation as shown in Figure 6. In step I, ssDNA binds across adjacent subunits, spiraling over the inner surface of the ring to contact the DNA binding loops. At this stage, the leading subunit is the C subunit and the trailing subunit is A, using the same notation as described previously. This conformation of the protein allows binding of nucleotides at four sites, two with NTP and two with NDP ⫹ Pi. NTP hydrolysis triggers rotation of the subunits in the plane of the ring, causing the ssDNA to be pulled through the ring by a lever action. Rotation of the subunits displaces the loops by 6–7 A˚ along the axis of the hexamer, equivalent to about two bases along the ssDNA strand. Simultaneously, as a consequence of the subunit rotations, the leading subunit (now the subunit to the left of C in the diagram) would bind to DNA, while the trailing subunit (now subunit B) would release the DNA. The trailing subunit rotates by 15⬚ but also undergoes a translation of about 5 A˚ in the plane of the ring and away from the center, thus pulling down and back from the ssDNA to facilitate release. Thus, we arrive at the configuration shown in step II, where the conformational state of the complex would be identical to that in step I, except that the identity of the subunits has now altered. The cycle of changes would repeat to progress from step II to step III. Consequently, translocation can

be viewed as a wave of coupled subunit rotations that ripple around the ring accompanying NTP binding and hydrolysis. The bound ssDNA remains in contact with at least one subunit of the hexamer at all stages in the process, ensuring high processivity and efficient coupling of translocation to NTP hydrolysis. One interesting consequence of this model is that because the ssDNA spirals around the inside track of the ring, there need not be more than a minimal rotation of the protein about the DNA strand (see Figure 6). NTP hydrolysis is stimulated strongly by the binding of DNA or RNA, and affinity of the protein for ssDNA is greater when nucleotides are bound (Stitt, 1988; Hingorani and Patel, 1996; Picha and Patel, 1998). The asymmetric interaction of the ring with bound ssDNA that we propose in our model could explain why there is an initial burst of NTP hydrolysis that is equivalent to one, rather than two, NTPs per hexamer (Hingorani et al., 1997; Stitt and Xu, 1998). The observation that stoichiometric binding of ssDNA to gene 4A protein is complete when two of the total of four molecules of dTMP-PCP are bound per hexamer (Hingorani and Patel, 1996) is also consistent with our model. Experiments involving heterooligomers of wild-type and inactive mutant proteins show some interesting phenomena that are consistent with our model. The first, surprising observation is that addition of a dTTPasedeficient mutant gene 4 protein actually increases the dTTPase activity of the wild-type enzyme in the absence of a ssDNA cofactor (Patel et al., 1994). However, by contrast, it has been shown that addition of a dTTPasedeficient mutant protein will inhibit the dTTPase activity of wild-type protein quite dramatically in the presence of ssDNA (Notarnicola and Richardson, 1993), particularly with ssM13 DNA (Patel et al., 1994). Furthermore, since the mutant protein can bind but not hydrolyze dTTP, the degree of inhibition that is induced by a substoichiometric amount of mutant protein is greater than would be expected for a system in which only half of the sites in the hexamer ever need to be able to hydrolyze nucleotide (i.e., three-site model). These apparently conflicting results are explained by our model in the following way. The stimulation of dTTPase activity occurs in the uncoupled dTTPase assay in the absence of ssDNA. Nucleotide hydrolysis under these conditions is so dependent upon hexamer formation that it does not matter if some

Cell 598

of the subunits are not active in themselves, exactly as suggested previously (Patel et al., 1994). However, in the presence of ssDNA, NTP hydrolysis is coupled to DNA translocation. Under these conditions, when the ssDNA is passed onto an inactive subunit in a hexamer it cannot move on to the next until hydrolysis has taken place. The enzyme is therefore arrested at this point until the ssDNA, NTP, or the hexamer dissociate. For the model we describe above, inhibition would always occur in less than five turnovers, so inhibition would be rapid and significant. The three-site model, in which only half of the dTTPase sites need to be able to hydrolyze dTTP, predicts that a less dramatic inhibition would occur. Although our structures suggest a general mechanism for how NTP binding and hydrolysis might be controlled around the helicase ring and how this could be coupled to translocation along ssDNA, these are merely working hypotheses based on a conformation of the protein that appears to be caught at one particular stage in the catalytic cycle. It is likely that other conformations exist, and study of these structures will probably reveal further details about the mechanism of the enzyme. There are several questions that remain unanswered. For example, the mechanism of strand separation remains a mystery, and it is also unclear why hexameric helicases generally prefer substrates with two tails (i.e., a true fork) rather than a simple tailed duplex that is sufficient for other enzymes such as PcrA or Rep. These matters will hopefully be resolved by future structural studies. Experimental Procedures Structure Determination Crystals of T7 gene 4D protein were grown as described previously (Bird et al., 1997) and were of the tetragonal spacegroup P41212 with unit cell dimensions a ⫽ b ⫽ 120.7 A˚, c ⫽ 284.5 A˚. There are two independent half hexamers in the asymmetric unit. Data were collected from flash frozen crystals at 100 K and processed using the HKL programs (Otwinowski and Minor, 1997) (Table 1). Due to abysmal nonisomorphism of heavy atom derivatives, the structure was solved by molecular replacement using the CCP4 programs ALMN and TFFC (CCP4, 1994) with the coordinates of T7 gene 4 fragment 4E as a model. The individual subunits had to be located separately because of the large conformational changes between the model and the final structure. The validity of the structure was confirmed by location of heavy atom positions using data collected from three derivatives. Initial heavy atom positions were determined using phases based on the 4E coordinate model for one half hexamer. These sites were refined using MLPHARE, and additional sites were detected using difference Fourier calculations with phases calculated from the heavy atoms alone. Further heavy atom refinement was undertaken using SHARP (de la Fortelle et al., 1997), which identified some additional heavy atom sites in difference Fourier maps. Phasing from these derivatives was poor due largely to the low resolution of the diffraction and crystal nonisomorphism. However, the heavy atom positions revealed the location of two half hexamers that were related to each other by an approximate twofold noncrystallographic symmetry axis that was parallel to the fourfold crystallographic axis. Rigid body refinement was carried out in CNS (Brunger et al., 1998). Initial positional refinement was carried out using the maximum likelihood target as implemented in CNS, with a bulk solvent correction. Six-fold NCS constraints were applied to each of the individual domains during initial stages of the refinement but were released toward the end and deviations from noncrystallographic symmetry were accommodated. These changes were almost exclusively at the subunit interfaces. However, similarity in the conformations of the two independent hexamers allowed two-

fold NCS restraints to be retained. Rounds of manual model building were undertaken between refinement cycles using TurboFrodo (Roussel and Cambillau, 1989). Statistics concerning the quality of the final model are presented in Table 1. Nucleotide Soaks Although it is possible to grow crystals in the presence of ADPNP and magnesium chloride, the best diffraction was observed by soaking apo protein crystals. The ADPNP soaks were carried out overnight in harvest solution (20% PEG 4000, 0.8 M sodium acetate, 50 mM CHES [pH 9.5]) including 2 mM ADPNP and 5 mM MgCl2 before flash freezing in liquid propane in the same solution. Initial rigid body refinement was followed by least squares positional refinement in CNS, initially with six-fold NCS constraints but releasing to twofold restraints as the refinement progressed. The same set of reflections were used to monitor the free R factor for native and soak data to reduce bias. The bound nucleotides were omitted from the refinement until the final stage. Acknowledgments We thank S. Gamblin and S. Smerdon for advice on freezing crystals in liquid propane, S. Halford, A. Leslie, K. Marians, S. Patel, and M. Szczelkun for helpful discussions, D. Hall for assistance on synchrotron visits, J. Byrne for technical assistance, and the Daresbury and Hamburg synchrotrons (support from the TMR/LSF programme to the EMBL Hamburg Outstation—ERBFMGECT980134) for data collection facilities. This work was supported by the Wellcome Trust. Received March 29, 2000; revised May 11, 2000. References Abrahams, J.P., Leslie, A.G., Lutter, R., and Walker, J.E. (1994). Structure at 2.8 A˚ resolution of F1-ATPase from bovine heart mitochondria. Nature 370, 621–628. Ahnert, P., and Patel, S.S. (1997). Asymmetric interactions of hexameric bacteriophage T7 DNA helicase with the 5⬘- and 3⬘-tails of the forked DNA substrate. J. Biol. Chem. 272, 32267–32273. Alison, T.J., Wood, T.C., Briercheck, D.M., Rastinejad, F., Richardson, J.P., and Rule, G.S. (1998). Crystal structure of the RNA-binding domain from transcription termination factor rho. Nat. Struct. Biol. 5, 352–356. Bird, L.E., Ha˚kansson, K., Pan., H., and Wigley, D.B. (1997). Characterisation and crystallisation of the helicase domain of bacteriophage T7 gene 4 protein. Nucleic Acids Res. 25, 2620–2626. Bird, L.E., Subramanya, H.S., and Wigley, D.B. (1998). Helicases: a unifying structural theme? Curr. Opin. Struct. Biol. 8, 14–18. Boyer, P.D. (1993). The binding change mechanism for ATP synthase—some probabilities and possibilities. Biochim. Biophys. Acta 1140, 215–250. Brunger, A.T., Adams, P.D., Clore, G.M., DeLano, W.L., Gros, P., Grosse-Kunstleve, R.W., Jiang, J.-S., Kuszewski, J., Nilges, M., Pannu, N.S., et al. (1998). Crystallography and NMR system: a new software suite for macromolecular structure determination. Acta Crystallog. D 54, 904–925. Bujalowski, W., and Klonowska, M.M. (1993). Negative cooperativity in the binding of nucleotides to Escherichia coli replicative helicase DnaB protein—interactions with fluorescent nucleotide analogs. Biochemistry 32, 5888–5900. Bujalowski, W., Klonowska, M.M., and Jezewska, M.J. (1994). Oligomeric structure of Escherichia coli primary replicative helicase DnaB protein. J. Biol. Chem. 269, 31350–31358. Collaborative Computing Project 4 (CCP4) (1994). The CCP4 suite: programs for protein crystallography. Acta Crystallog. D 50, 760–763. de la Fortelle, E., and Bricogne, G. (1997). Maximum-likelihood heavy atom parameter refinement for multiple isomorphous replacement and multiwavelength anomalous diffraction methods. Methods Enzymol. 276, 472–494.

T7 Gene 4 Ring Helicase 599

Dillingham, M.S., Soultanas, P., and Wigley, D.B. (1999). Sitedirected mutagenesis of motif III in PcrA helicase reveals a role in coupling ATP hydrolysis to strand separation. Nucleic Acids Res. 27, 3310–3317. Dillingham, M.S., Wigley, D.B., and Webb, M.R. (2000). Unidirectional single stranded DNA translocation by PcrA helicase: measurement of step size and translocation speed. Biochemistry 39, 205–212. Egelman, E.H., Yu, X., Wild, R., Hingorani, M.M., and Patel, S.S. (1995). Bacteriophage T7 helicase/primase proteins form rings around single-stranded DNA that suggest a general structure for hexameric helicases. Proc. Natl. Acad. Sci. USA 92, 3869–3873. Fass, D., Bogden, C.E., and Berger, J.M. (1999). Crystal structure of the N-terminal domain of the DnaB hexameric helicase. Structure 7, 691–698. Frick, D.N., Baradaran, K., and Richardson, C.C. (1998). An N-terminal fragment of the gene 4 helicase/primase of bacteriophage T7 retains primase activity ion the absence of helicase activity. Proc. Natl. Acad. Sci. USA 95, 7957–7962. Gogol, E.P., Siefried, S.E., and von Hippel, P.H. (1991). Structure and assembly of the E. coli transcription termination factor rho and its interaction with RNA. I. Cryoelctron microscopic studies. J. Mol. Biol. 221, 1127–1138. Guo, S., Tabor, S., and Richardson, C.C. (1999). The linker region between the helicase and primase domains of the bacteriophage T7 gene 4 protein is critical for hexamer formation. J. Biol. Chem. 274, 30303–30309. Hingorani, M.M., and Patel, S.S. (1993). Interactions of bacteriophage T7 DNA primase/helicase protein with single-stranded and double-stranded DNAs. Biochemistry 32, 12478–12487. Hingorani, M.M., and Patel, S.S. (1996). Cooperative interactions of nucleotide ligands are linked to oligomerisation and DNA binding in bacteriophage T7 gene 4 helicases. Biochemistry 35, 2218–2228.

(1995). A domain of the gene 4 helicase/primase of bacteriophage T7 required for the formation of an active hexamer. J. Biol. Chem. 270, 20215–20224. Oda, T., and Takanami, M. (1972). Observations on the structure of the termination factor Rho and its attachment to DNA. J. Mol. Biol. 71, 799–802. Otwinowski, Z., and Minor, W. (1997). Processing of X-ray diffraction data collected in oscillation mode. Methods Enzymol. 276, 307–326. Patel, S.S., and Hingorani, M.M. (1993). Oligomeric structure of bacteriophage T7 primase/helicase proteins. J. Biol. Chem. 268, 10668– 10675. Patel, S.S., Hingorani, M.M., and Ng, W.M. (1994). The K318A mutant of bacteriophage T7 DNA primase-helicase protein is deficient in helicase but not primase activity and inhibits primase-helicase protein wild-type activities by heterooligomer formation. Biochemistry 33, 7857–7868. Picha, K.M., and Patel, S.S. (1998). Bacteriophage T7 DNA helicase binds dTTP, forms hexamers, and binds DNA in the absence of Mg2⫹. J. Biol. Chem. 273, 27315–27319. Roussel, A., and Cambillau, C. (1989). TURBO-FRODO. In Silicon Graphics Geometry Partner Directory, Silicon Graphics, ed. (Mountain View, California: Silicon Graphics), pp. 77–78. San Martin, C., Rademacher, M., Wolpensinger, B., Engel, A., Miles, C.S., Dixon, N.E., and Carazo, J.-M. (1998). Three-dimensional reconstructions from cryoelectron microscopy images reveal an intimate complex between helicase DnaB and its loading partner DnaC. Structure 6, 501–509. Sawaya, M.R., Guo, S., Tabor, S., Richardson, C.C., and Ellenberger, T. (1999). Crystal structure of the helicase domain from the replicative helicase-primase of bacteriophage T7. Cell 99, 167–177.

Hingorani, M.M., Washington, M.T., Moore, K.C., and Patel, S.S. (1997). The dTTPase mechanism of T7 DNA helicase resembles the binding change mechanism of the F1-ATPase. Proc. Natl. Acad. Sci. USA 94, 5012–5017.

Shirakihara, Y., Leslie, A.G.W., Abrahams, J.-P., Walker, J.E., Ueda, T., Sekimoto, Y., Kambara, M., Saika, K., Kagawa, Y., and Yoshida, M. (1997). The crystal structure of the nucleotide-free ␣3␤3 subcomplex of F1-ATPase from the thermophilic Bacillus PS3 is a symmetric trimer. Structure 5, 825–836.

Jezewska, M.J., Kim, U.-S., and Bujalowski, W. (1996). Binding of Escherichia coli primary replicative helicase DnaB protein to singlestranded DNA. Long range allosteric conformational changes within the protein hexamer. Biochemistry 35, 2129–2145.

Soultanas, P., Dillingham, M.S., Velankar, S.S., and Wigley, D.B. (1999). DNA binding mediates metal ion coordination and conformational changes in the active site of PcrA helicase. J. Mol. Biol. 290, 137–148.

Kaplan, D.L., and Steitz, T.A. (1999). DnaB from Thermus aquaticus unwinds forked duplex DNA with an asymmetric tail length dependence. J. Biol. Chem. 274, 6889–6897.

Stitt, B.L. (1988). Escherichia coli transcription termination protein rho has three hydrolytic sites for ATP. J. Biol. Chem. 263, 11130– 11137.

Kim, J.L., Morgenstern, K.A., Griffith, J.P., Dwyer, M.D., Thomson, J.A., Murcko, M.A., Lin, C., and Caron, P.R. (1998). Hepatitis C virus NS3 RNA helicase domain with a bound oligonucleotide: the crystal structure provides insights into the mode of unwinding. Structure 6, 89–100.

Stitt, B.L., and Xu, Y. (1998). Sequential hydrolysis of ATP molecules bound in interacting catalytic sites of Escherichia coli transcription termination protein Rho. J. Biol. Chem. 273, 26477–26486.

Kim, D., Shigesada, K., and Patel, S.S. (1999). Transcription termination factor Rho contains three noncatalytic nucleotide binding sites. J. Biol. Chem. 274, 11623–11628. Korolev, S., Hsieh, J., Gauss, G.H., Lohman, T.M., and Waksman, G. (1997). Major domain swiveling revealed by the crystal structures of complexes of E. coli Rep helicase bound to single-stranded DNA and ADP. Cell 90, 635–647. Lohman, T.M., and Bjornson, K.P. (1996). Mechanisms of helicasecatalyzed DNA unwinding. Annu. Rev. Biochem. 65, 169–214. Matson, S.W., and Richardson, C.C. (1983). DNA-dependent nucleoside 5⬘–triphosphatase activity of the gene-4 protein of bacteriophage T7. J. Biol. Chem. 258, 4009–4016. Matson, S.W., Tabor, S., and Richardson, C.C. (1983). The gene-4 protein of bacteriophage T7—characterisation of helicase activity. J. Biol. Chem. 258, 4017–4024. Nicholls, A., and Honig, B.J. (1991). A rapid finite-difference algorithm, utilizing successive over-relaxation to solve the Poisson-Boltzmann equation. J. Comput. Chem. 12, 435–445. Notarnicola, S.M., and Richardson, C.C. (1993). The nucleotide binding site of the helicase/primase of bacteriophage T7: interaction of mutant and wild-type protein. J. Biol. Chem. 268, 27198–27207. Notarnicola, S.M., Park, K., Griffith, J.D., and Richardson, C.C.

Story, R.M., and Steitz, T.A. (1992). The structure of the E. coli RecA protein monomer and polymer. Nature 355, 318–325. Story, R.M., Weber, I.T., and Steitz, T.A. (1992). Structure of the RecA protein–ADP complex. Nature 355, 374–376. Subramanya, H.S., Bird, L.E., Brannigan, J.A., and Wigley, D.B. (1996). Crystal structure of a DExx box helicase. Nature 384, 379–383. Velankar, S.S., Soultanas, P., Dillingham, M.S., Subramanya, H.S., and Wigley, D.B. (1999). Crystal structures of complexes of PcrA helicase with a DNA substrate indicate an inchworm mechanism. Cell 97, 75–84. Washington, M.T., Rosenberg, A.H., Griffin, K., Studier, F.W., and Patel, S.S. (1996). Biochemical analysis of mutant T7 primase/helicase proteins defective in DNA binding, nucleotide hydrolysis, and the coupling of hydrolysis with DNA unwinding. J. Biol. Chem. 271, 26825–26834. Wittinghofer, A. (1998). Signal transduction via ras. Biol. Chem. 379, 933–937. Yarranton, G.T., and Gefter, M.L. (1979). Enzyme-catalyzed DNA unwinding: studies on Escherichia coli rep protein. Proc. Natl. Acad. Sci. USA 76, 1658–1662. Yu, X., and Egelman, E.H. (1997). The RecA hexamer is a structural homologue of ring helicases. Nat. Struct. Biol. 4, 101–104.

Cell 600

Yu, X., Hingorani, M., Patel, S.S., and Egelman, E.H. (1996a). DNA is bound within the central hole to one or two of the six subunits of the T7 DNA helicase. Nat. Struct. Biol. 3, 740–743. Yu, X., Jezweska, M.J., Bujalowski, W., and Egelman, E.H. (1996b). The hexameric E. coli DnaB helicase can exist in different quaternary states. J. Mol. Biol. 259, 7–14. Protein Data Bank ID Codes The coordinates have been deposited at the RCSB Protein Data Bank with identifier codes 1EOK (apo enzyme) and 1EOJ (ADPNP complex).