ATP Binding to the Escherichia coli Clamp Loader Powers Opening of the Ring-shaped Clamp of DNA Polymerase III Holoenzyme*

The Escherichia coli γ complex serves as a clamp loader, catalyzing ATP-dependent assembly of β protein clamps onto primed DNA templates during DNA replication. These ring-shaped clamps tether DNA polymerase III holoenzyme to the template, facilitating rapid and processive DNA synthesis. This report focuses on the role of ATP binding and hydrolysis catalyzed by the γ complex during clamp loading. We show that the energy from ATP binding to γ complex powers several initial events in the clamp loading pathway. The γ complex (γ2δδ′χψ) binds two ATP molecules (one per γ subunit in the complex) with high affinity (K d = 1–2.5 × 10−6 m) or two adenosine 5′-O-(3-thiotriphosphate)(ATPγS) molecules with slightly lower affinity (K d = 5–6.5 × 10−6 m). Experiments performed prior to the first ATP turnover (k cat = 4 × 10−3 s−1 at 4 °C), or in the presence of ATPγS (k cat = 1 × 10−4s−1 at 37 °C), demonstrate that upon interaction with ATP the γ complex undergoes a change in conformation. This ATP-bound γ complex binds β and opens the ring at the dimer interface. Still prior to ATP hydrolysis, the composite of γ complex and the open β ring binds with high affinity to primer-template DNA. Thus ATP binding powers all the steps in the clamp loading pathway leading up to the assembly of a γ complex·open β ring·DNA intermediate, setting the stage for ring closing and turnover of the clamp loader, steps that may be linked to subsequent hydrolysis of ATP.

Rapid and efficient duplication of genomic DNA depends on biomolecular machines known as DNA replicases. In diverse organisms these multicomponent biological machines exhibit varying degrees of complexity, including substantial differences in subunit composition. It is apparent, however, that certain mechanisms of action and the associated protein tools are conserved through evolution. One prominent example is the mechanism for processive DNA replication common among replicases from bacteriophage T4, Escherichia coli, yeast, and humans. Briefly, a clamp loader uses energy from ATP to assemble a circular protein clamp around DNA; the clamp, now topologically linked to DNA, tethers DNA polymerase to the template, and by sliding freely on duplex DNA it allows continuous replication of several thousand nucleotides without a dissociation event (reviewed in Refs. 1-3).
The E. coli clamp loader, ␥ complex, is a composite of five different proteins, ␥, ␦, ␦Ј, , and , of which ␥ is the ATPbinding subunit essential for clamp loading (4 -7). The clamp, ␤, is a dimeric ring with a 35-Å inner diameter, large enough to encircle DNA (8). ␤ and ␥ complex form part of DNA polymerase III holoenzyme, the replicative DNA polymerase of E. coli. The holoenzyme assembly also includes a core polymerase (␣⑀), composed of ␣, the DNA polymerase (9), ⑀, the proofreading exonuclease (10,11), and of unknown function (12,13), as well as , a dimeric protein that holds together two polymerase cores and binds one ␥ complex (14 -16). The core polymerase is a nonprocessive, inefficient DNA polymerase that extends a primer by only 10 -20 nucleotides before dissociating from the template (17). When the ␤ clamp tethers the core polymerase to template DNA, however, the enzyme develops high processivity and extends DNA by several thousand nucleotides per binding event (18,19).
In the current model for E. coli DNA replication, after the assembly of an initiation complex in which the polymerase III holoenzyme is tethered to a primed DNA template by circular protein clamps, the two core polymerases in the holoenzyme synthesize leading and lagging strand DNA synchronously (20 -23). Synthesis of the leading strand occurs in the direction of replication fork movement, but the lagging strand is synthesized in the opposite direction in discrete 1-2-kilobase pairlong Okazaki fragments. Therefore on the leading strand, after one clamp loading event and initiation complex formation, the polymerase can extend an RNA primer for several thousand nucleotides. On the lagging strand, however, the polymerase must initiate synthesis at new primers every few seconds, and a protein clamp must be loaded at each primer for initiation complex formation (24). Thus, clamp loader action is required continuously through the entire process of genome replication.
The clamp loader belongs to the category of proteins known as molecular matchmakers. These proteins, by definition, use their ATPase activity to promote assembly of a stable complex between target macromolecules (25), as the clamp loader uses ATP to assemble a complex between a clamp and DNA. Recent studies of the E. coli clamp loader have revealed substantial information about its subunit composition and its mechanism of action. The ␥ complex contains one each of ␦, ␦Ј, , and subunits, and two or three ␥ subunits (6,26). ␦ is the only subunit that binds ␤ (27); ␥ binds and hydrolyzes ATP (5,28); ␦Ј appears to bury ␦ within the ␥ complex and block its interaction with ␤ 1 ; interacts with SSB, 2 facilitating DNA chain extension at physiological ionic strength, and the function of is unknown (29,30). Initial insight into how the clamp loader works was provided by studies showing that the free ␦ subunit binds ␤ in the absence of ATP, but when ␦ is in the ␥ complex ATP is required for its interaction with ␤. It has been postulated that in the presence of ATP a conformational change in ␥ complex presents ␦ for interaction with ␤ (27). Once this interaction is established, the clamp loader must assemble the ␤ ring around DNA, and it must release the ␤⅐DNA complex to complete its job as a molecular matchmaker. At present there is not a clear understanding of how these steps occur and how the ATPase activity of ␥ complex powers ␤ assembly on DNA.
An earlier study with an ATPase mutant ␥ complex demonstrated that inhibition of ATP hydrolysis blocks DNA replication, and this effect was attributed to the inability of the mutant complex to assemble a ␤ clamp on DNA (7). In the present study we have continued to investigate how the clamp loader uses ATP, and we were surprised to find that ATP binding powers the clamp loading process almost to completion. Initially, quantitative nucleotide binding assays were used to characterize ␥ complex interaction with ATP and its slowly hydrolyzing analog, ATP␥S. Partial proteolysis assays, DNA binding experiments, and a novel ring-opening assay were performed prior to and subsequent to ATP/ATP␥S hydrolysis to separate the role of ATP binding from hydrolysis in clamp loading. We show here that ATP binding induces a conformational change in the ␥ complex, promotes its interaction with ␤, and powers ␤ ring opening. The ATP-bound ␥ complex⅐␤ complex binds primer-template DNA leading to formation of a stable ␥ complex⅐open ␤ ring⅐DNA composite that is an important intermediate in the clamp loading pathway (Fig. 10).
Clamp Loading Assay-␤ was 3 H-labeled by reductive methylation as described (36), to a specific activity of 5 ϫ 10 4 cpm/pmol (dimer), and found to be as active as ␤ in DNA replication. [ 3 H]␤ (500 fmol of dimer) was incubated in Reaction buffer A with singly primed M13mp18 ssDNA (400 fmol) coated with SSB (140 pmol as tetramer), and ␥ complex (␥ or K51R-␥; 500 fmol) in a 75-l reaction volume. The mixture was incubated for 5 min at 37°C when 1 mM ATP was added to the reaction, or for up to 1 h at 37°C when ATP␥S was added, followed by gel filtration over a 5-ml Bio-Gel A-15m column (Bio-Rad) in Column buffer A containing 100 mM NaCl. Fractions of 200 l were collected, and 150-l aliquots were subjected to liquid scintillation counting to detect and quantitate [ 3 H]␤.
Replication Assay-A (dT) 35 -primed poly(dA) template was used to measure DNA replication in the presence of various nucleotides. A poly(dA) template (120 M total nucleotide; average length, 4000 nts) was primed with a (dT) 35 DNA (20:1 ratio) and incubated with ␥ complex (0.025 M) and ␤ (0.025 M dimer), in the absence of nucleotides or in the presence of 1 mM ATP or ATP␥S, for 5 min at 23°C in Reaction buffer A (30 l final volume). Core polymerase (␣⑀) was added (0.025 M), and following incubation for another 1 min the reaction was shifted to 37°C and replication initiated by addition of 500 M [␣-32 P]dTTP. At various times the reaction was quenched by mixing 7 l of the reaction mix with 7 l of 1% SDS ϩ 100 mM EDTA. Aliquots were spotted onto DE81 filters (Whatman), and free [␣- 32  , with or without ␤ (5 M) and primer-template DNA (5 M), as described above for ATPase activity. ATPase activity of the ␥ subunit (5 M) was also measured by similar assays performed at 37°C in the absence of ␤ and DNA.
Quantitative Nucleotide Binding-ATP and ATP␥S binding to ␥ complex was measured at 23°C using nitrocellulose membrane binding assays in which a constant amount of ␥ complex was titrated with increasing concentrations of ATP or ATP␥S. Nitrocellulose membrane circles (25 mm) were washed with 0.5 N NaOH, rinsed immediately with water, and equilibrated in membrane wash buffer (20 mM Tris-HCl, pH 7.5, 10 mM MgCl 2 , and 50 mM NaCl) prior to use. The reactions contained ␥ complex (2 or 5 M), ␤ (2 or 5 M dimer), and [␣-32 P]ATP (0 -40 M) or [␥-35 S]ATP (0 -50 M) in Reaction buffer B in a total volume of 15 l. Samples (10-l aliquots) were filtered through nitrocellulose membranes on a glass microanalysis filter assembly (Fisher) immediately after addition of ATP or after a 10-min incubation of reactions containing ATP␥S. The membranes were washed before and after filtration with 1 ml of membrane wash buffer. 1-l aliquots of the samples were spotted on a separate nitrocellulose membrane to determine total nucleotide in the reaction.
or to the Hill equation to estimate cooperativity in ligand binding (Equation 2).
where ML is the amount of ligand bound to protein; M T is the total protein concentration; K 1 and K 2 are association constants; L F is the free ligand concentration (ATP or ATP␥S); n is the Hill coefficient; K is a dissociation constant; and ML/M T is the fraction of ligand bound per ␥ complex, or ligand bound per ␥ subunit in the ␥ complex (in the Hill equation). The binding curves were fit by a nonlinear least squares fitting program using KaleidaGraph (Synergy Software). Partial Tryptic Digest of ␥ Subunit and ␥ Complex-Tryptic digestion of ␥ was performed at 37°C in the absence and in the presence of various nucleotides. The reaction contained ␥ subunit (10 M) and no nucleotide, ATP, ADP, or ATP␥S (2 mM) in Reaction buffer B in a total volume of 12 l. Trypsin was added to a final concentration of 0.05 g, and after 10 min at 37°C the reaction was quenched with 10 l of SDS loading buffer and boiled for 5 min. The proteolytic products were analyzed on a 12% gel by SDS-PAGE and visualized by staining with Coomassie Blue.
Tryptic digestion of ␥ complex was performed with 5 M ␥ complex in Reaction buffer B in the absence or in the presence of ATP, ADP, ATP␥S (500 M) in a total reaction volume of 10 l. Trypsin was added to a final concentration of 0.8 g, and following incubation for 10 min at 37°C the reaction was quenched with an equal volume of SDS loading buffer. The products were analyzed on a 12% polyacrylamide gel as described above. Some of the proteolytic fragments were N-terminally sequenced at the Rockefeller University Protein/DNA Technology Center.
Gel Filtration Analysis of ␥ Complex-␤ Interaction-Interaction of ␥ complex with ␤ was examined by gel filtration essentially as described in an earlier report (27). Briefly, ␥ complex (5 M) was incubated with ␤ (2.5 M dimer) in 200 l of Column buffer B containing no nucleotide, 0.3 mM ATP, or 0.3 mM ATP␥S. The protein mix was gel-filtered at 4°C on a Superose 12 column equilibrated with the same buffer. After the first 7 ml, fractions of 200 l were collected, and migration of the proteins was monitored by analyzing aliquots (40 l) of the column fractions by SDS-PAGE on 12% gels.
Protein Footprinting Analysis of ␥ Complex-␤ Interaction-␤ PK , a modified ␤ with a C terminus kinase recognition site, was 32 P-labeled by cAMP-dependent protein kinase as described earlier, and was found as active as ␤ in DNA replication (36). [ 32 P]␤ PK (0.1 M dimer) was added to Reaction buffer B (40-l final volume) containing 100 mM NaCl, with no nucleotide, ATP, ADP, or ATP␥S (0.5 mM). The total protein was adjusted to 100 g with BSA. ␥ complex (0.3 M) was added to the reaction followed by immediate initiation of digestion with 0.2 g of Pronase E (23°C). After 2 min the reaction was quenched with 20 l of stop solution (0.5 M Tris-HCl, pH 6.8, 0.3 mM DTT, 14% SDS, 14% glycerol, 70 mM EDTA, and 2 mM Pefabloc SC). 30-l aliquots were analyzed on the discontinuous Tricine/SDS-polyacrylamide gel system of Schagger and von Jagow as described (37,38). The 32 P-labeled proteolysis products were visualized on a PhosphorImager (Molecular Dynamics).
A similar assay was performed to obtain a footprint of ␤ PK prior to ATP hydrolysis. At 4°C, [ 32 P]␤ PK (0.1 M dimer) was incubated for 1 min with ␥ complex (0.2 M) and BSA (100 g) in Reaction buffer B ϩ 100 mM NaCl. ATP was added to initiate the reaction, and after 5 s Pronase E was added to a final concentration of 2 g. After another 5 s, the ␥ complex ATPase and Pronase E activities were quenched by addition of 20 l of HCl (1 N) and 20 l of stop solution, followed by neutralization with base. The proteolysis products were analyzed as described above.
DNA Binding-Gel filtration assays were performed on Bio-Gel A-15m columns to detect interaction between ␥ complex, ␤, and M13mp18 ssDNA. [ 3 H]␥ complex (specific activity, 1.1 ϫ 10 4 cpm/pmol) was prepared by tritium-labeling the ␦Ј subunit and incorporating it into the ␥ complex as described (39). Following incubation for 1 min at 23°C, the reaction mixture was gel-filtered on a 5-ml column equilibrated in Column buffer A containing 0.1 mM ATP, ADP, or ATP␥S (when in reaction). 200-l fractions were collected and quantitated by liquid scintillation counting. Complementary assays with radiolabeled ␤ were performed as described above except with [ 32 P]␤ PK (36) and unlabeled ␥ complex.
Gel shift assays were performed under nondenaturing conditions to detect interaction between ␥ complex, ␤, and short radiolabeled DNAs. 100-nt ssDNA1 was 32 P-labeled using T4 polynucleotide kinase and [␥-32 P]ATP. After free nucleotide was removed by spinning through a Bio-Gel P-30 column (Bio-Rad), the DNA was annealed to primer 28 or 100-nt ssDNA2 to prepare primer-template DNA or 100-nt dsDNA, respectively. Radiolabeled DNA (1 M) was mixed with increasing concentrations of ␥ complex and ␤ (0 -5 M; 0 -10 M with dsDNA) in 10 l of Reaction buffer B, in the presence of 0.5 mM ATP␥S. Following incubation for 2 min at 23°C, the samples were subjected to gel electrophoresis on a 1% agarose gel containing 20 mM MgCl 2 and 0.1 mM ATP␥S in TBE buffer. Stability of ␥ complex⅐␤ bound to primer-template DNA versus ssDNA was tested in similar assays in which 1 M 32 P-labeled DNA was bound to 5 M ␥ complex⅐␤ followed by incubation with 60 M unlabeled 70-nt ssDNA (1 min at 4°C) and analyzed by nondenaturing gel electrophoresis. The gels were dried onto DE81 paper, and the 32 P-labeled DNAs were visualized on a PhosphorImager.
Ring-opening Assay-A modified version of ␤, with Leu-273 (at the dimer interface) changed to cysteine and Cys-333 (at the protein surface) changed to serine, was prepared and tested for activity as described. 1 L273C-␤ activity in ATPase assays, DNA replication assays, and clamp loading assays was identical to that of wild-type ␤. ␤ ring opening was detected by measuring the labeling of Cys-273 with eosin-5-maleimide. A mixture of L273C-␤ (2 M) and ␥ complex (2 M) in 20 mM Tris-HCl, pH 7.5, ϩ 10 mM MgCl 2 (10 l final volume) was incubated in the dark at 4°C for 1 min. ATP or ATP␥S (0.5 mM) was added, followed in 10 s by addition of 1 l of eosin-5-maleimide stock solution (5 mM in Me 2 SO). Control reactions were performed with L273C-␤ alone and with ␥ complex ϩ L273C-␤ in the absence of nucleotides. The reactions were quenched after 30 s with an equal volume of 0.5 M DTT ϩ SDS loading buffer; the samples were boiled for 2 min and subjected to SDS-PAGE on a 10% gel. Fluorescence-labeled ␤ was visualized on a Fluor-S MultiImager (Bio-Rad). The ring-opening assay was performed similarly in the presence of 3 M primer-template DNA. The proteins were incubated with primer-template DNA in the presence of ATP␥S (1 min at 23°C), and the labeling reaction and analysis were performed as described above.

RESULTS
The E. coli clamp loader uses ATP to convert the inefficient polymerase III core enzyme into a highly processive DNA replicating machine. Earlier studies have demonstrated that ␥ complex loads the circular clamp ␤ onto a primed DNA template, and ␤ tethers the polymerase to DNA, facilitating highly processive DNA synthesis. Here we investigate how ␥ complex uses its ATPase activity for clamp loading. To assemble ␤ on DNA the ␥ complex must bind ␤ as well as DNA, open the ␤ ring and place it around DNA, and finally dissociate from the complex to allow polymerase access to the clamp. The question is: what powers various steps in the clamp loading pathway or, alternately, how is energy from ATP binding and hydrolysis used for clamp loading?
ATP Hydrolysis Requirement for Clamp Loading: an ATPase Mutant Study-Initially we used an ATPase-inactive ␥ complex to investigate the role of ATP in clamp loading. Mutants of the clamp loader subunits ␦ and ␥ had been prepared previously to determine which subunit harbors the ATPase activity essential for clamp loading (40). According to data from UV cross-linking experiments, the ␦ subunit may interact with ATP (41); however, mutation of Lys-225 to Ala in the nearconsensus nucleotide-binding site in the ␦ subunit did not inhibit the ATPase activity or the clamp loading activity of ␥ complex reconstituted with mutant ␦ (40). On the other hand, mutation of Lys-51 to Ala in the nucleotide-binding Walker A site of ␥ subunit inactivated the ATPase activity of ␥ complex, indicating that ␥ serves as the clamp loader ATPase (7). The mutant ␥ complex did not support DNA replication, and it was hypothesized that the defect was due to a loss of clamp loading activity resulting from the loss in clamp loader ATPase activity.
In this study we tested the above hypothesis by assaying directly the clamp loading activity of a mutant ATPase ␥ complex. Fig. 1, panel A, shows that the mutant K51R-␥ complex has severely inhibited ATPase activity. At 37°C, wild-type ␥ complex hydrolyzes ATP with a k cat ϭ 5 s Ϫ1 in the presence of ␤ and a primed DNA template, whereas the ATPase activity of K51R-␥ complex is much slower with a k cat ϭ 6.9 ϫ 10 Ϫ2 s Ϫ1 . The clamp-loading activity of the mutant ␥ complex was assayed directly by following the assembly of 3 H-labeled ␤ on SSB-coated, primed M13mp18 DNA (Fig. 1, panel B). On a Bio-Gel A-15m gel filtration column, [ 3 H]␤ assembled on DNA elutes in the early fractions (fractions 10 -15) and is resolved from free ␤ protein (fractions 18 -35). Wild-type ␥ complex assembles [ 3 H]␤ around DNA in the presence of ATP, but under the same conditions K51R-␥ complex is incapable of loading clamps. These data together with the earlier results confirm that ATP hydrolysis is necessary for loading ␤ clamps on DNA and is therefore essential for processive DNA synthesis.
ATP Hydrolysis Requirement for Clamp Loading: an ATP␥S Study-The apparent requirement of ATP hydrolysis for the assembly of ␤ onto DNA as well as for processive DNA synthesis was investigated further by measuring ␥ complex activity in the presence of ATP␥S, an ATP analog that is poorly hydro-lyzed by most ATP hydrolases. First we examined the effect of ATP␥S on replication of a (dT) 35 -primed poly(dA) template ( Fig. 2, panel A). Utilization of a poly(dA) template allowed measurement of DNA synthesis in the absence of dATP which, like ATP, can support ␥ complex-catalyzed loading of clamps on DNA. The clamp loader, clamp, and DNA were incubated with ATP or ATP␥S to allow clamp loading, followed by addition of core polymerase (␣⑀) and [␣-32 P]dTTP to initiate DNA replication. As shown in Fig. 2, panel A, [␣-32 P]dTMP is incorporated into DNA at a rate of 2.2 M min Ϫ1 when ATP is present in the reaction. In the presence of ATP␥S, however, DNA replication is indistinguishable from the basal level of replication in the absence of any nucleotide (0.2 M min Ϫ1 ). We then examined the effect of ATP␥S on the clamp loading activity of ␥ complex. Clamp loading assays using [ 3 H]␤ were performed in the presence of ATP or ATP␥S, and reactions containing ATP␥S were sampled at longer times to account for potentially slow ATP␥S hydrolysis. As shown in Fig. 2, panel B, when ATP␥S was present in the reaction, assembly of [ 3 H]␤ onto DNA was barely detectable over the background, even after 1 h at 37°C.
To determine why ATP␥S does not support clamp loading, we tested the ability of ␥ complex to hydrolyze [ 35 S]ATP␥S under various conditions. The assays were performed at high ATP␥S concentrations to measure the maximum steady-state rate of hydrolysis. Fig. 2, panel C, shows that ␥ complex (5 M) hydrolyzes ATP␥S very slowly (k cat ϭ 1 ϫ 10 Ϫ4 s Ϫ1 ) both in the absence and in the presence of ␤. There is no burst activity at initial time points in the reaction, and the k cat reflects the slowest rate in the reaction pathway; therefore one turnover of ATP␥S hydrolysis requires more than 2 h at 37°C. There is also no stimulation of ATP␥S hydrolysis on addition of primertemplate DNA, a substrate in the clamp loading reaction. Inability of the clamp loader to hydrolyze ATP␥S efficiently, coupled with its inability to load clamps or stimulate DNA synthesis in the presence of ATP␥S, is consistent with earlier evidence that ATP hydrolysis is essential for clamp loading and processive DNA replication (Ref. 7; Fig. 1). 3 It is possible, however, that the clamp loader cannot bind ATP␥S as it binds ATP and therefore cannot use ATP␥S to assemble ␤ onto DNA. To address this possibility we used nitrocellulose membrane binding assays to measure and compare interactions of ␥ complex with ATP and ATP␥S. Fig. 3, panel A, shows data from the titration of a constant amount of ␥ complex and ␤ (2 M) with increasing concentrations of ATP, where the ␥ complex⅐␤-bound ATP is plotted versus total ATP in the reaction. The binding isotherm reaches saturation at 4 M ATP, demonstrating quite strikingly that two ATP molecules bind ␥ complex with high affinity. Earlier evidence from high pressure liquid chromatography and SDS-PAGE analysis of the ␥ complex has indicated that the E. coli clamp loader is likely composed of 2-3 ␥ subunits, 1 ␦, 1 ␦Ј, 1 , and 1 subunit (26). The ATP binding data shown here are most consistent with the presence of two ATP-binding ␥ subunits in the ␥ complex. 3 Like ␥ complex the complex (␦␦Ј) also assembles ␤ around DNA in the presence of ATP, but unlike ␥ complex, it stimulates replication of the dT-primed poly(dA) template by core polymerase in the presence of ATP␥S ([␣-32 P]dTMP is incorporated into DNA at 6 M min Ϫ1 in the presence of ATP, 2 M min Ϫ1 in the presence of ATP␥S, and 0.5 M min Ϫ1 in the absence of nucleotides). Similar complex activity has also been observed in an assay measuring replication of M13 G ori DNA (59). Perhaps the complex can hydrolyze ATP␥S and use this ATP analog for clamp loading. It should be noted, however, that unlike ␥, the subunit binds both DNA and the core DNA polymerase; therefore, the complex may bring the clamp, DNA, and polymerase in close proximity and facilitate their interaction even in the absence of ATP hydrolysis. See "Discussion" for details of the clamp loading pathway. , and the linear slope yields k cat ϭ 5 s Ϫ1 for wild-type ␥ complex and k cat ϭ 6.9 ϫ 10 Ϫ2 s Ϫ1 for K51R-␥ complex (panel A). Clamp loading activity was assayed by incubating 500 fmol of ␥ complex (wild-type, with 1 mM ATP, q; without nucleotide, Ⅺ; K51R, with 1 mM ATP, ࡗ) with [ 3 H]␤ (500 fmol) and SSB-coated circular M13 ssDNA (400 fmol) for 5 min at 37°C. The reactions were gel-filtered as described in the "Experimental Procedures." ␤ assembled on M13 ssDNA elutes in fractions 10 -15 and free ␤ elutes in fractions 18 -35 (panel B).
The data were fit to to an equation describing the interaction of two ligands with one macromolecule (Equation 1), and Fig. 3, panel B, shows that the two ATP molecules bind ␥ complex with slightly differing affinities, with dissociation constants 1/K 1 ϭ 2.6 ϫ 10 Ϫ6 M and 1/K 2 ϭ 1.12 ϫ 10 Ϫ6 M. The binding isotherm fits the Hill equation (Equation 2) with a coefficient n ϭ 1.5, which also indicates slight positive cooperativity in the interaction of two ATP with ␥ complex⅐␤ (data not shown). The results from nitrocellulose membrane binding assays are consistent with previous equilibrium gel filtration experiments, which yielded a dissociation constant K d ϳ1 ϫ 10 Ϫ6 M for ATP binding to ␥ complex (40). Interestingly, although the ␥ subunit is an inefficient ATPase, it binds ATP with the same affinity as the ␥ complex, K d ϭ 1.2 ϫ 10 Ϫ6 M (5). Presumably, assembly of ␥ into a complex with other subunits does not affect ATP binding but stimulates ␥ subunit ATPase activity by influencing other steps in the ATPase pathway.
The high affinity interaction between ATP and ␥ complex was compared with ATP␥S binding. Like ATP, two ATP␥S molecules bind per ␥ complex⅐␤ (Fig. 3, panel C). The Hill coefficient, n ϭ 1.4, indicates that the two molecules bind with slight positive cooperativity (data not shown), and Fig. 3, panel D, shows that the two dissociation constants for ATP␥S binding are similar 1/K 1 ϭ 6.5 ϫ 10 Ϫ6 M and 1/K 2 ϭ 5 ϫ 10 Ϫ6 M. Most notably, the data show that ATP␥S binds ␥ complex with only 2-5-fold lower affinity than ATP; therefore, the inability of the clamp loader to use ATP␥S for clamp loading is likely not due to a defect in ATP␥S binding.
The data presented in Figs. 2 and 3 indicate that ATP binding is not sufficient, and ATP hydrolysis is necessary for ␥ complex-catalyzed assembly of circular clamps around DNA. We have also found that the E. coli clamp loader binds ATP␥S but does not hydrolyze it efficiently; therefore, ATP␥S can be considered a nonhydrolyzable ATP analog in experiments performed within a short time. The assays described in the following sections utilize ATP␥S in conjunction with ATP to determine whether ATP binding plays an important role in clamp loading.
ATP Binding to ␥ Promotes a Conformational Change in the ␥ Complex-It has been postulated that ATP facilitates interaction between ␥ complex and ␤ by inducing a conformational change in ␥ complex. This idea is supported by the observation that the Pronase digestion pattern of ␦ subunit within the ␥ complex is altered in the presence of ATP (27). Does ATP binding effect a change in ␥ complex conformation? To address this question, a partial tryptic digest of the ␥ subunit was performed in the absence of ATP and in the presence of various nucleotides. The digest products were analyzed by SDS-PAGE, and the Coomassie Blue-stained gel is shown in Fig. 4, panel A. In the absence of nucleotide almost all the ␥ in the reaction is proteolyzed in 10 min at 37°C (lane 2). Several proteolytic fragments range between masses 20 -30 kDa, and some are the result of cleavage at Arg-208 or Arg-215 and varying C-terminal cleavage sites (data from N-terminal sequencing and mass spectroscopic analysis not shown). In the presence of ATP, the digest pattern is significantly different than in the absence of nucleotides (Fig. 4, panel A, lanes 2 and 3). The small proteolytic products observed in lane 2 are not formed when ATP is present in the reaction (lane 3), and ␥ appears to be relatively resistant to tryptic digest. A similar change in digest pattern is detected when ADP or ATP␥S are present in the reaction (Fig.  4A, lanes 4 and 5). The ␥ subunit hydrolyzes ATP (k cat ϭ 3 ϫ 10 Ϫ3 s Ϫ1 at 37°C) but not ATP␥S (hydrolysis undetectable; data not shown) during the tryptic digest. Both nucleotides, however, induce similar changes in the ␥ subunit digest pattern indicating that the change occurs on ATP binding to ␥. In control assays performed with other isolated proteins, it was determined that trypsin activity is generally unaffected by the presence of nucleotides (data not shown).
The proteolytic digest assay was also performed with ␥ complex to determine if it changes conformation on ATP binding, as does the free ␥ subunit. The complex was subjected to tryptic digestion in the absence and in the presence of nucleotides, and the products were analyzed by SDS-PAGE (Fig. 4, panel B).
Comparison of the peptide products in lane 3 (no nucleotide in the reaction) with lanes 4 and 6 (ATP and ATP␥S in the reactions, respectively) shows that the digest pattern of ␥ complex changes in the presence of nucleotides. Again, ATP␥S is not hydrolyzed in the time scale of the assay, indicating that the change occurs on nucleotide binding. ADP binding also induces changes in both ␥ and ␥ complex, similar to those observed on ATP binding (Fig. 4, panel A, lane 4, and Fig. 4,  panel B, lane 5, respectively). These data imply that after ATP binding and hydrolysis the complex may not revert back to its original conformation until after ADP dissociation.
ATP Binding Facilitates Interaction between ␥ Complex and ␤-Next we analyzed the role of ATP binding and hydrolysis in promoting the association of ␥ complex with ␤. In an earlier study, interaction of ␥ complex with ␤ was analyzed by gel filtration on a Superose 12 column, and it was found that ATP is necessary for stable ␥ complex-␤ interaction (27). In a similar assay used here, ␥ complex (5 M) was incubated with ␤ dimer (2.5 M) in the absence of nucleotide and in the presence of ATP or ATP␥S; the mixture was analyzed by gel filtration and SDS-PAGE of the column fractions. Panels A and B in Fig. 5 confirm the earlier report that ATP is required for ␥ complex binding to ␤, and panel C shows that like ATP ATP␥S supports stable interaction between ␥ complex and ␤.
A more rapid assay based on protein footprinting was used to determine whether the ␥ complex-␤ interaction occurs upon ATP binding and whether ATP hydrolysis is required for this interaction. A 32 P-labeled ␤ clamp was incubated with ␥ complex in the absence and in the presence of nucleotides and proteolyzed with Pronase E, and the protein footprint was examined for changes due to ␥ complex-␤ interaction. In an earlier study, a similar protein footprinting assay demonstrated that ATP-induced interaction between ␤ and ␥ complex protects a C-terminal fragment of ␤ from proteolytic digest (38). Fig. 6, panel A, shows the expected [ 32 P]␤ footprint in the presence of ATP and ␥ complex (lane 4; the protected fragment is indicated by an arrow). ATP␥S facilitates similar protection of the ␤ C terminus, and because it is not hydrolyzed in the 2-min time scale of the assay nucleotide binding appears sufficient for ␥ complex interaction with ␤ (lane 6). To confirm this observation, the assay was also performed at 4°C within 10 s, prior to the first turnover of ATP (k cat ϭ 4 ϫ 10 Ϫ3 s Ϫ1 ; no presteady-state burst of hydrolysis in the absence of DNA). In the presence of ATP, ␤ is protected by ␥ complex even under these conditions (Fig. 6, panel B, lane 2); therefore, ATP binding but not hydrolysis is necessary for ␥ complex-␤ interaction. The assay also shows that the ␤ C terminus is not protected by ␥ complex in the presence of ADP (Fig. 6, panel A, lane 5). Therefore, although ADP binding can lead to a change in ␥ complex conformation, it does not support ␥ complex-␤ interaction. This result implies that after ATP is hydrolyzed, ␥ complex⅐␤ may dissociate, resulting in ␥ complex turnover and its catalytic loading activity.
ATP Binding Facilitates Interaction between ␥ Complex⅐␤ and DNA-The E. coli clamp loader must bring ␤ and primed DNA together to assemble the clamp around DNA. Consequently, one intermediate that must form during the clamp loading process is the three-component composite of ␥ complex, ␤, and DNA. To detect and study this intermediate, we examined the interactions between ␥ complex, ␤, and DNA, as well as the effects of nucleotide ligands on these interactions. [ 3 H]␥ complex was incubated with single-stranded M13mp18 DNA, in the absence and in the presence of ␤ and nucleotides, and analyzed by gel filtration on an agarose A-15m column (Fig. 7,  panels A and B). [ 3 H]␥ complex bound to DNA elutes early (fractions 11-16) and is resolved from free protein (fractions 16 -31) during gel filtration. All ␥ complex in the reaction binds to M13 ssDNA in the presence of ␤ and ATP␥S (panel A). ␤ appears essential for stable complex formation with DNA, because no interaction is detectable in its absence (panel A). In a complementary assay performed with 32 P-labeled ␤ and unlabeled ␥ complex the clamp co-elutes with the clamp loader on DNA, although appearance of free ␤ in the gel filtration profile suggests that the clamp may be less stable in the protein⅐DNA complex as compared with the clamp loader (panel C). Since the DNA binding assay is complete within 13 min and prior to ATP␥S hydrolysis (refer Fig. 2, panel C), the results demonstrate that ATP␥S binding is sufficient for the interaction of ␥ complex⅐␤ with DNA.
Interestingly, when the assay is performed using ATP instead of ATP␥S only a small fraction of 3 H-labeled ␥ complex elutes with ssDNA, and when ADP is used there is no detectable interaction between the proteins and DNA (panel B). The results with ATP␥S, indicating that ATP binding increases the affinity of ␥ complex⅐␤ for ssDNA, coupled with the fact that ADP does not support stable DNA binding suggest that ATP binding and hydrolysis modulate interaction of ␥ complex and ␤ with DNA; thus ATP binding to ␥ complex promotes DNA binding, and ATP hydrolysis triggers DNA release. Consequently, the ATP turnover rate likely controls turnover of the protein⅐DNA composite, which explains why the composite is stable to gel filtration in the presence of the slowly hydrolyzing analog ATP␥S but not in the presence of ATP.
It should be noted that ␤ can be assembled as a ring around double-stranded DNA, or RNA-DNA, e.g. a primed ssDNA template, but not around ssDNA. Furthermore, ATP hydrolysis is essential for the assembly of ␤ around primed DNA, and this final product of the clamp loading pathway is stable during gel filtration even if no ATP is present in the column buffer (refer Fig. 2, panel B). In contrast, ATP␥S must be present in the column buffer to maintain a stable interaction between ␥ complex, ␤, and single-stranded M13mp18 DNA during gel filtration. Thus the protein⅐DNA composite formed in the presence of ATP␥S is different from the final product of the clamp loading pathway. Evidence from an earlier study also indicates that after hydrolyzing ATP and loading ␤ around DNA, the ␥ com- DNA binding properties of ␥ complex⅐␤ were characterized further by testing different DNA substrates. Fig. 8 shows the results of nondenaturing gel-shift assays used to detect interactions between ␥ complex⅐␤ and primer-template DNA, ssDNA, or dsDNA. In the presence of ATP␥S, ␥ complex⅐␤ binds with high affinity to a 32 P-labeled primer-template DNA substrate (28-nt primer annealed to a 100-nt ssDNA of random sequence), resulting in a gel-shifted complex (Fig. 8, panel A). ␥ complex⅐␤ also binds to a 100-nt single-stranded DNA (Fig. 8,  panel B), but no gel shift can be detected with the 100-nt double-stranded DNA even at 10 M protein concentration (Fig.  8, panel C). Binding to dsDNA was also examined by gel filtration (as described in Fig. 7) using Bluescript plasmid DNA; however, no interaction could be detected between ␥ complex⅐␤ and dsDNA (data not shown). The data from experiments performed with ss-and dsDNAs suggest that ␥ complex⅐␤ binds predominantly the singlestranded region on primer-template DNA. The primer-template substrate used here has a 36-nt single-stranded overhang on either side of the primer. Interestingly, binding assays performed with various ssDNAs showed that a minimum length of about 70 nts is required to detect a stable gel-shift from ␥ complex⅐␤ binding to DNA (data not shown). Presumably on our primer-template substrate the ss/ds junction structure confers additional stability to the protein⅐DNA composite beyond that provided by binding to a 36-nt ssDNA alone. In an assay designed to compare ␥ complex-␤ interaction with primer-template DNA to single-stranded DNA, the protein⅐DNA complexes were challenged with 60-fold excess 70-nt ssDNA. As shown in Fig. 8, panel D, the composite of ␥ complex⅐␤ and 32 P-labeled primer-template DNA remains intact following addition of the unlabeled ssDNA trap to the reaction (lane 3). The composite of ␥ complex⅐␤ and 32 P-labeled 100-nt ssDNA, however, dissociates easily, and the proteins are trapped effectively by unlabeled DNA, whereas the 32 P-labeled 100-nt ssDNA migrates as free DNA (Fig. 8, panel D, lane 6). The same results were obtained when the assay was performed by mixing 32 P-labeled DNA with the trap DNA prior to incubation with proteins (data not shown). Thus ␥ complex⅐␤ binds primer-template DNA with both greater affinity and stability than ssDNA, which may be a result of specific interaction between ␥ complex⅐␤ and the ss/ dsDNA junction on the primer-template.
ATP Binding Powers ␥ Complex-catalyzed ␤ Ring Opening-As mentioned earlier, it has been hypothesized that clamp loading involves ␥ complex-catalyzed opening of the ␤ ring at the dimer interface followed by assembly around a primed DNA template. To investigate this hypothesis, an assay was developed that directly measures ␤ ring opening. 1 A modified ␤ clamp, L273C-␤, was prepared by changing Leu-273, a residue buried at the dimer interface, to Cys, and by changing a reactive Cys-333 on the protein surface to Ser (note: L273C-␤ activity was found identical to that of wild-type ␤ in DNA replication, clamp loading, and ATPase assays). When the ring is closed Cys-273 is buried within the L273C-␤ dimer interface, inaccessible to solvent, and therefore unreactive. When the ring opens, however, Cys-273 can be labeled with thiol-reactive agents. In the current study, the role of ATP binding in clamp loading was investigated further with L273C-␤ and eosin-5maleimide, a fluorescent reagent that reacts covalently with Cys-273 when the ring is open.
L273C-␤ was incubated with eosin-5-maleimide in the absence and in the presence of ␥ complex and nucleotides, and the proteins were analyzed by SDS-PAGE followed by fluorescence detection on a Fluor-S MultiImager (Bio-Rad). L273C-␤, alone or with ␥ complex and no nucleotide, reacts weakly with eosin-5-maleimide (Fig. 9, panel A, lanes 1 and 2, respectively). When both ␥ complex and ATP are present in the reaction, fluorescent labeling of L273C-␤ is stimulated significantly, indicating that the reactive cysteine at the ␤ dimer interface is exposed under these conditions (lane 3). Lane 4 in Fig. 9, panel A, shows similar stimulation of ␤ labeling in the presence of ATP␥S, and this occurs prior to ATP␥S hydrolysis (reaction conditions, 30 s at 4°C). Therefore, when it binds ATP the ␥ complex facilitates opening of the ␤ clamp, presumably to place DNA in the clamp through the open ring interface.
The ring opening assay was also used to examine if the clamp is open in the three-component ␥ complex⅐␤⅐DNA composite that is formed on ATP␥S binding. As demonstrated earlier, the interaction between ␥ complex⅐␤ and primer-template DNA is very stable in the presence of ATP␥S (Fig. 8, panel D); therefore, the ring opening assay was performed in the presence of the primer-template DNA substrate. Fig. 9, panel B, shows that fluorescence labeling of ␤ increases significantly over the background when ␥ complex and ATP␥S are present in the reaction with primer-template DNA (lane 3). Together, the data in Fig. 9 demonstrate that ATP binding powers ␥ complexcatalyzed opening of the ␤ clamp as well as formation of a stable open-clamp complex on a primed DNA template.

DISCUSSION
The ring-shaped ␤ clamp does not assemble on DNA by itself. The clamp loader, ␥ complex, catalyzes assembly of clamps around primed template DNA for use by the core DNA polymerase. The ␥ complex requires ATP for clamp loading, and consistent with this observation, mutation of Lys-51 in the Walker-A ATP-binding site in ␥ inhibits the clamp loading activity of ␥ complex and blocks processive DNA replication (7).
But how does the clamp loader couple its ATPase activity to clamp loading? In this study, we have examined the effects of ATP on ␥ complex and ␤ (on protein conformation and proteinprotein/protein-DNA interactions), and we demonstrate that ATP binding powers several initial steps in the clamp loading pathway (Fig. 10).
The E. coli clamp loader functions as a "protein topoisomerase," cracking open the ring-shaped ␤ clamp to place DNA in the center of the ring. ATP binding powers changes in ␥ complex that lead to formation of a ␥ complex⅐open ␤ ring⅐DNA composite, which is likely an important intermediate in the clamp loading pathway. This information has been incorporated into a model for the role of ATP binding in the clamp loader pathway which will be discussed in detail later (Fig. 10). Furthermore, based on the effects of ATP and ADP on ␥ complex activity, we have hypothesized that after ATP binding powers formation of the ␥ complex⅐open ␤ ring⅐DNA composite, ATP hydrolysis may trigger placement/closing of the clamp around DNA or simply dissociate the proteins and DNA so that ␤ is released as a ring around DNA. In rendering the model we have also engaged in some speculation, imagining that the protein topoisomerase clamp loader functions like DNA topoisomerase II (42), taking in substrate DNA through one gate and releasing it into the ␤ ring through a second gate to form a topological link between the clamp and DNA.
Nucleotide Binding-Nucleotide binding and hydrolysis assays demonstrated that ␥ complex binds ATP␥S with similar affinity as ATP, and it hydrolyzes ATP␥S very slowly so that one turnover can take up to 2 h at 37°C (k cat ϭ 1 ϫ 10 Ϫ4 s Ϫ1 with no presteady-state burst of hydrolysis); thus, ATP␥S can be considered nonhydrolyzable in all assays that are completed within a few minutes. These results led to the choice of ATP␥S for our study of the role of ATP binding in clamp loading, instead of a truly nonhydrolyzable ATP analog like AMP-PNP, which binds ␥ complex with almost 100-fold lower affinity than ATP. 4 Both ATP and ATP␥S bind ␥ complex with a stoichiometry of two nucleotides per one ␥ complex, as can be expected for a clamp loader comprising two ␥ subunits (Fig. 3, panels A and C,  respectively). The free ␥ subunit can form dimers, tetramers, and even higher order oligomers (5,16). These higher order ␥ oligomers break down in the presence of the subunit, and further addition of yields a subassembly estimated to contain 2.5 ␥, 0.7 , and 1 subunits (43). In various other studies using gel filtration (5), densitometry of stained SDS-polyacrylamide gels (6), or sedimentation equilibrium analysis (44), the molar ratio of ␥ has been reported as low as 2 or as high as 4 subunits per ␥ complex. The subunits in ␥ complex have also been resolved by high performance liquid chromatography under denaturing conditions and quantitated by absorbance at 280 nm (using known extinction coefficients). This analysis indicated the presence of 2-3 ␥ subunits and 1 each of ␦, ␦Ј, , and per ␥ complex (26). Recently, we have analyzed the molecular mass of ␥ complex by multi-angle laser light scattering, and the molar ratio of ␥ subunits per ␥ complex is 2.5, a value closer to 2 rather than 4. 5 Finally, high pressure liquid chromatography analysis of polymerase III* purified from cell lysates has shown that 2.2 ␥ subunits are present per polymerase III* assembly (polymerase III* indicates polymerase III holoenzyme minus ␤; see Ref. 16). Thus, our measure of 2 ATP or ATP␥S-binding sites per ␥ complex is consistent with earlier evidence that ␥ complex composition in the holoenzyme is likely ␥ 2 ␦␦Ј rather than ␥ 4 ␦␦Ј.
Changes in Conformation-A recent study shows that in order to forge a topological link between the ␤ ring and DNA, ␥ complex acts as a protein topoisomerase rather than a "DNA topoisomerase," cracking open the ␤ ring to assemble it around DNA. 1 This process likely involves interaction between the ␦ subunit of ␥ complex and ␤ 1 (27). Free ␦ binds ␤ in the absence of ATP, but ␥ complex requires ATP to bind ␤, suggesting that ATP promotes a change in ␥ complex that induces its interaction with ␤ (27). By using a partial proteolysis assay, we have shown that the ␥ subunit undergoes a change in conformation on binding ATP, both alone and in complex with other clamp loader proteins. Particularly striking are the changes in tryptic cleavage of ␥ at Arg-208/Arg-215 after it binds ATP (Fig. 4).
Analysis of the recently solved structure of ␦Ј provides an interesting interpretation of the proteolytic assay results. The ␦Ј subunit is a "C"-shaped protein that shares sequence identity and homology with the ␥ subunit (45). A three-dimensional model structure of ␥ created with the MODELLER program suggests that the entire ␦Ј structure may be a good model for the structure of ␥ (45-47). Arg-215 in ␥ has been identified as a highly conserved residue in the "Sensor-2 Motif," common to E. coli, bacteriophage T4, and human clamp-loader proteins. In both ␦Ј and ␥, the Sensor-2 motif is in a domain that forms a "hinge" between the top and bottom domains of the C (45). This motif packs closely against the ␤ strand-P loop-␣ helix structural segment containing the ATP-binding Walker-A sequence, GTRGVGKT, in ␥. The Arg-215 residue may even serve as a ligand for the nucleotide phosphates; therefore, a tryptic cleavage site in this region is likely in a position of sensitivity to nucleotide binding. Trypsin activity at this site may be directly inhibited by nucleotide binding, but perhaps, more interestingly, tryptic cleavage is blocked because of a conformational change in the Sensor-2 motif. In ␥, an ATP binding-mediated change in hinge conformation could facilitate movement of top and bottom domains, possibly even an opening/closing of the "jaws" that may be coupled to opening/closing of the ␤ ring.
ADP has the same effect on the proteolytic digest pattern of ␥ and ␥ complex as ATP does, implying that ADP binding induces similar changes in protein conformation (Fig. 4). ADP binding, however, does not facilitate ␥ complex-␤ interaction like ATP does; therefore, the ADP-bound ␥ complex must differ from the ATP-bound complex. Presumably the proteolytic digest assay is not sensitive enough to pick up these differences. These results do suggest that when the clamp loader changes its conformation on ATP binding, it does not not revert back to its original shape until after ATP hydrolysis and dissociation of the ADP product. Another interesting observation is that while the ␥ subunit in ␥ complex is digested by trypsin, the other 4  subunits, ␦, ␦Ј, , and , remain mostly intact as if immune to proteolysis. Perhaps the two C-shaped ␥ subunits in the ␥ complex are arranged so as to limit exposure of the other subunits to trypsin. The ␥ complex sketch in our model clamp loading pathway reflects this speculative arrangement with all the subunits except ␦ ensconced deep within the complex (Fig.  10). The ␦ subunit is depicted as partially accessible because at high protein concentrations ␤ can interact with ␥ complex even in the absence of ATP. 6 Ring Opening-Experiments performed with ATP␥S and with ATP (prior to and post-hydrolysis) revealed that the clamp loader interacts with ␤ on binding ATP. Following interaction with ␤, and still prior to ATP hydrolysis, ␥ complex opens the ␤ ring at the dimer interface. A novel assay based on labeling a reactive cysteine at the ␤ dimer interface was used to detect opening of the ␤ ring. ␤ was modified by changing Leu-273 at the interface to cysteine; at the same time a surface-exposed Cys at position 333 was changed to Ser, creating a ␤ clamp that can be labeled with a thiol-reactive reagent only when the ring is open and the interface exposed. 1 As shown in Fig. 9, panel A, the clamp is opened and fluorescence-labeled with eosin-5maleimide when ␥ complex binds ATP or ATP␥S, and ATP hydrolysis is not necessary for ring opening.
Recent reports have suggested that the bacteriophage T4 clamp loader, gp44/62, hydrolyzes ATP to open the gp45 clamp ring (48,49). In one detailed study, photo-cross-linking tech-niques were used to track changes in conformation of the gp45 clamp in various stages of the loading cycle. Conformational changes were detected in the presence of ATP, ATP␥S, ADP, as well as primer-template DNA. These data were presented in a model proposing that on ATP binding there occurs a rearrangement in intersubunit contacts between gp45 and gp44/62, and on ATP hydrolysis further changes occur that lead to gp45 ring opening (48). Our direct labeling assay with eosin-5-maleimide shows that ␥ complex cracks open the ␤ dimer interface on binding ATP, and ATP hydrolysis is not required for this activity. However, it remains possible that E. coli shares similarity with T4 bacteriophage in that ␥ complex-catalyzed ATP hydrolysis may lead to further opening of the ␤ ring. This idea will be interesting to test in future studies in which transition state analogs of ATP, such as ADP-AlF 4 , may be used to analyze clamp loading intermediates that form during ATP hydrolysis.
DNA Binding-ATP binding to the clamp loader also promotes high affinity binding of ␥ complex⅐␤ to both primertemplate DNA and single-stranded DNA substrates. Although in our gel shift assays we do not detect interaction between ␥ complex alone and DNA, a more sensitive assay measuring changes in rotational anisotropy of fluorescence-labeled DNA has shown weak interaction between an 80-nt ssDNA and ␥ complex (50). Therefore, ␥ complex may bind DNA and ␤ in random order to form the ␥ complex⅐␤⅐DNA composite.
␥ complex and ␤ bind single-stranded DNA but not doublestranded DNA in the gel-shift assays implying that on primed 6  ssDNA these proteins may interact with only the singlestranded regions. However, a study using photoreactive primers on template DNA demonstrated that the ␥ and ␦ subunits in ␥ complex link covalently to the primer at position Ϫ2 (51), which indicates its close proximity to the double-stranded region and the primer-template junction. Our DNA binding experiments also show that ␥ complex⅐␤ binds primer-template DNA with higher affinity and stability than ssDNA (Fig. 8,  panel D). Furthermore, primed ssDNA stimulates ␥ complex ATPase activity substantially more than other DNA substrates 1 (28). Presumably the ss/ds DNA junction structure influences the interaction of ␥ complex and ␤ with DNA, and the ability of the clamp loader to recognize specifically the primer-template junction may play an important role in targeting clamps to primed sites on template DNA.
Interaction between ␥ complex⅐␤ and DNA was not detectable in the absence of nucleotides or in the presence of ADP (Fig.  7). Given that ␥ complex also does not bind ␤ in the presence of ADP, these results suggest that ATP binding and hydrolysis modulate the interaction of ␥ complex with both its clamp loading substrates, ␤ and DNA. Thus in each catalytic cycle, ATP binding to ␥ complex brings together the DNA and proteins in preparation for clamp loading, and ATP hydrolysis leads to clamp assembly around DNA and dissociation of all the components by a mechanism that is not yet clearly understood. Concurrent with this study, the interaction of ␥ complex⅐␤ with DNA has been examined by fluorescence anisotropy measurements of rhodamine-labeled primer-template DNA. These data also indicate that a stable ternary complex of ␥ complex⅐␤⅐DNA is formed on ATP binding and suggest that the complex turns over when ATP is hydrolyzed (Ref. 52 (accompanying paper)).
A ␥ Complex⅐Open ␤ Ring⅐DNA Intermediate in the Clamp Loading Pathway-The ␤ clamp in the ␥ complex⅐␤⅐DNA composite is an open ring. In a reaction containing this stable composite, eosin-5-maleimide labels L273C-␤ at the dimer interface indicating the clamp loader holds the ␤ dimeric ring open. Because all the components necessary for clamp loading are assembled at this point, our current model of the clamp loading cycle shows this ternary composite as the final intermediate before ␤ is released onto DNA (Fig. 10D). At the start of the loading cycle, ATP binding induces a change in ␥ complex (Fig. 10, A 3 B) that facilitates its interaction with ␤ and DNA, perhaps in random order (Fig. 10, B 3 C). ␥ complex ATPase activity is greatly stimulated by DNA (28), and the next step in the pathway reflects this property by showing a DNA-induced change in ␥ complex that may speed up its catalytic activity (Fig. 10, C 3 D). Perhaps after the ␥ complex⅐open ␤ ring⅐DNA intermediate is formed (D), ATP hydrolysis powers other changes such as further opening of the ␤ ring to place it around DNA. On the other hand, ATP hydrolysis may induce conformational changes that close the ring around DNA. Equally possible is the pathway in which ATP binding powers the work necessary for clamp loading up to and including placement of the open ␤ ring in the right configuration around DNA; ATP hydrolysis then simply triggers dissociation of the composite, allowing the ␤ dimer to assume its lowest free energy state, i.e. a closed ring, and the close proximity of primer-template junction ensures that the ring closes around DNA.
Although we speculate here about the effects of ATP hydrolysis, there is no detailed information available as yet that definitively explains the role of ATP hydrolysis in E. coli ␥ complex-catalyzed ␤ loading. However, based on our results with ADP and the slowly hydrolyzing analog ATP␥S, it does appear that turnover of ATP is essential for turnover of the ␥ complex⅐open ␤ ring⅐DNA composite. This is shown in the next step of the model pathway where ATP hydrolysis and product dissociation complete the clamp loading cycle with formation of ␤⅐DNA and free ␥ complex (Fig. 10, panels D 3 E). Finally, evidence from proteolytic digestion of ␥ complex suggests that ADP must dissociate before ␥ complex returns to its original free state; therefore, the pathway shows that after ATP is hydrolyzed and ADP is released by the clamp loader, the clamp loading cycle can start again (Fig. 10, panels E 3 A).
A recent report on the mechanism of bacteriophage T4 clamp loader activity proposed that gp44/62 requires energy from ATP hydrolysis to physically open or close the gp45 ring around DNA (53). The study employed an elegant strand-displacement assay that follows the formation of active T4 holoenzyme by measuring its processive DNA synthesis activity (54). The assay was performed with ATP and analogs ATP␥S and AMP-PCP, and the data showed that ATP hydrolysis is absolutely necessary for processive DNA replication. The interpretation of the results was that nonhydrolyzable ATP analogs do not support processive T4 holoenzyme activity because energy from the ATP hydrolysis step is required to open the clamp and assemble it on DNA (53). Our study with the E. coli clamp loader shows that the ATP␥S-bound ␥ complex⅐open ␤ ring⅐DNA composite is quite stable and may not break up until the the nucleotide dissociates or is hydrolyzed. The core DNA polymerase competes with the clamp loader for ␤ (38); therefore, a stable interaction between ␥ complex and ␤ can block processive DNA polymerase activity. Similarly in T4, in the presence of nonhydrolyzable ATP analogs, gp44/62 may be trapped in a complex with gp45 and DNA and block the access of T4 polymerase to the clamp, thus inhibiting DNA synthesis (earlier studies have shown that T4 polymerase and gp44/62 compete for the gp45 clamp; see Refs. 55 and 56). In the T4 bacteriophage, ATP hydrolysis may lead to ring opening, or ring closing around DNA, or simply dissociation of the clamp loader⅐clamp⅐DNA complex, and inhibition of any of these steps in the clamp loading pathway would inhibit processive DNA replication; therefore, the exact mechanistic role of ATP hydrolysis in the clamp loading pathway still remains an open question.
ATP␥S-mediated formation of the clamp loader⅐clamp⅐DNA composite has been observed also in the eukaryotic DNA replication systems of both humans and Saccharomyces cerevisiae. In humans, ATP␥S supports stable interaction of the clamp loader, RFC, with its clamp, PCNA, and primed DNA template (57). The authors noted that the DNA polymerase, pol␦, binds to the protein⅐DNA composite and catalyzes DNA synthesis only when the composite is formed in the presence of ATP but not when it is formed in the presence of ATP␥S. They also found that the ATP␥S-bound RFC⅐PCNA⅐DNA composite can be activated for DNA synthesis by adding ATP to the reaction. Based on information from the E. coli clamp loading pathway, we speculate that pol␦ does not bind the ATP␥S-bound composite because it is blocked by RFC, and when ATP␥S dissociates, the subsequent turnover of RFC in the presence of ATP allows pol␦ access to PCNA⅐DNA, thereby allowing it to synthesize DNA processively.
Experiments performed with the S. cerevisiae RFC and PCNA provide yet additional information (58). As in the sys-tems described above, in the presence of ATP␥S an RFC⅐PCNA⅐DNA composite is formed, which does not bind pol␦ or stimulate DNA synthesis. As expected, addition of ATP to this composite allows the polymerase to synthesize DNA processively (most likely because after the composite dissociates, RFC can load PCNA on DNA and release PCNA⅐DNA in the presence of ATP). However, the authors also found that if they isolated ATP␥S-bound RFC⅐PCNA⅐DNA by gel filtration, addition of pol␦ to the reaction resulted in some DNA synthesis even without addition of ATP; furthermore, addition of excess ATP␥S to this isolated composite completely inhibited pol␦ activity. To explain these results, the authors speculated that after the ATP␥S-bound composite was formed and isolated, perhaps it could be reactivated for DNA synthesis by the other nucleotides added to the reaction (dCTP, dGTP, or dTTP, but not AMP-PNP), and further addition of ATP␥S to the reaction blocked this reactivation. We offer another explanation based on our study of the ␥ complex and presuming some similarity between the S. cerevisiae and E. coli clamp loading pathways. Perhaps in the ATP␥S-bound RFC⅐PCNA⅐DNA composite the PCNA ring is open and ready for release around DNA. If this composite breaks up when ATP␥S dissociates from RFC and the PCNA ring closes, it may sometimes snap closed around DNA, because of the close proximity of DNA within the complex, resulting in stimulation of pol␦ activity. If there is excess ATP␥S present in the reaction and it binds RFC rapidly, the composite will stay predominantly ATP␥S-bound and retain its integrity, thereby blocking pol␦ access to PCNA and processive DNA synthesis. We were unable to isolate a stable ␥ complex⅐open ␤ ring⅐DNA intermediate without having excess ATP␥S present in the reaction; therefore, we could not test directly if ATP␥S dissociation allows the composite to break up such that ␤ snaps closed around DNA and can be used by core polymerase. However, the data from S. cerevisiae lends substance to the idea that ATP hydrolysis may simply trigger dissociation of the clamp loader⅐clamp⅐DNA intermediate, resulting in release of the clamp as a ring around DNA.
In conclusion, our model mechanism for clamp loading shows that ATP binding to the clamp loader powers most of the work involved in the assembly of a ring-shaped clamp on DNA. Such efficient use of ATP binding energy is reminiscent of the mechanism by which ligands activate allosteric enzymes. In multisubunit allosteric enzymes, the energy derived from ligand binding leads to specific conformational changes, in the component subunits, which are coupled to modulation of the catalytic activity. Similarly the energy from ATP binding to the clamp loader leads to conformational changes, in the component subunits, that are coupled to the work of opening the clamp ring and holding it in close proximity to DNA. Then, ATPase activity of the clamp loader functions as a switch mechanism and changes the clamp loader from a form that FIG. 10. ATP binding powered activity of ␥ complex. On binding ATP ␥ complex undergoes a change in conformation (A 3 B), binds ␤ and opens the dimeric ring, and binds DNA to form a ␥ complex⅐open ␤ ring⅐DNA composite (B 3 C 3 D). We propose that this intermediate is poised for release of the clamp around DNA (D 3 E), and ATP hydrolysis triggers ␤ loading and dissociation of the composite. Following release of the hydrolysis products, ADP and P i , the clamp loader cycles back to start loading again (E 3 A).
prefers binding the clamp and DNA into a form that prefers releasing the clamp and DNA, topologically linked to each other.