Characterization of simian virus 40 T-antigen double hexamers bound to a replication fork. The active form of the helicase.

Large T-antigen (T-ag) is a viral helicase required for the initiation and elongation of simian virus 40 DNA replication. The unwinding activity of the helicase is powered by ATP hydrolysis and is critically dependent on the oligomeric state of the protein. We confirmed that the double hexamer is the active form of the helicase on synthetic replication forks. In contrast, the single hexamer cannot unwind synthetic forks and remains bound to the DNA as ATP is hydrolyzed. This inability of the T-ag single hexamer to release the DNA fork is the likely explanation for its poor helicase activity. We characterized the interactions of T-ag single and double hexamers with synthetic forks and single-stranded (ss) DNA. We demonstrated that DNA forks promote the formation of T-ag double hexamer. The lengths of the duplex region and the 3' tail of the synthetic forks are the critical factors in assembly of the double hexamer, which is bound to a single fork. We found that the cooperativity of T-ag binding to ss oligonucleotides increased with DNA length, suggesting that multiple consecutive subunits in the hexamer engage the ssDNA.

DNA helicases are ubiquitous enzymes that function as cellular motors to unwind DNA duplexes at the expense of NTP hydrolysis (1). Helicases must not only bind to DNA but also translocate along it. Translocation involves a series of binding and release events during which helicases cycle between DNA binding states varying in affinity using steps in the NTP hydrolysis cycle as "switches." Helicases translocate unidirectionally along ssDNA. 1 This directionality is also intrinsic to their unwinding activity and is revealed by a tight association to only one of the product single strands. Two general groups of helicases are thus distinguished: 3Ј35Ј and 5Ј33Ј.
Helicases are also classified according to their quaternary structure as monomeric, dimeric, and hexameric. Hexameric helicases form rings around DNA and function primarily in replication. Simian virus 40 (SV40) T-antigen (T-ag) is a 3Ј35Ј-hexameric helicase (2)(3)(4). Its essential role in DNA replication depends on its ability to bind specifically to the SV40 origin of replication and subsequently to power bi-directional unwinding via its helicase activity. Thus, as in other well studied DNA replication programs, the origin recognition and helicase loading steps may be studied separately. In the case of SV40 a single protein is responsible for all of these reactions. An initiation complex at the origin assembles through the binding of T-ag monomers to precisely spaced and oriented GAGGC repeats (5)(6)(7). T-ag hexamers can also form in the absence of DNA, but these cannot load productively onto origin-containing duplex fragments (8). The pre-replication origin complex consists of two bound T-ag hexamers in a head-to-head orientation (9). Once unwinding starts from the origin, the two T-ag hexamers could remain attached to one another or separate and move away from each other. Evidence for both scenarios has been obtained by electron microscopy (EM) of in vitro unwinding intermediates (10). The finding of unwinding intermediates with two hexamers linked to each other has been interpreted as suggesting that double hexamers are the active helicase machine.
T-ag is a complex DNA-binding protein. It binds double- (11), as well as single-stranded DNA (ssDNA) (12,13). Moreover, the duplex DNA binding can be specific to SV40 origin sequences or be sequence-independent (14). The nonspecific binding is crucial for the helicase activity of T-ag. EM analyses of unwinding intermediates in vivo (15) and in vitro (10) have shown that the physiological substrate of T-ag is the border between the melted ssDNA and the intact upstream duplex. In this study we will focus exclusively upon T-ag complexes bound to synthetic replication forks that mimic the structure of the in vivo substrate. The forks we constructed contain a duplex part and 3Ј and 5Ј ssDNA tails. In principle, T-ag could bind to any combination of these three regions. Moreover, the strength and distribution of these contacts are expected to change during ATP hydrolysis to bring about unwinding. Knowing how T-ag engages its substrate is essential for understanding the molecular mechanism of its helicase activity. Two previous footprinting studies have reached somewhat different conclusions as to how T-ag binds to forks. Wessel et al. (10) have presented evidence that T-ag protects both the duplex and the two ssDNA tails, and SenGupta and Borowiec (16) have proposed that T-ag engages the forks through interactions with only the 3Ј ssDNA tail.
Binding of T-ag to either synthetic forks or replication bubbles results in the formation of two distinct complexes, which differ in gel mobility and sedimentation velocity (17). We designate the faster migrating complex on gels as type I and the slower as type II. Smelkova and Borowiec (17) have found that type II complexes are 15 times more active helicases than type I complexes. Based on parallels with the origin binding activity of T-ag and the structures formed at the SV40 replication origin (18), the authors proposed that type I complexes are single hexamers, whereas type II complexes are dimers of type I complexes held together by protein-protein interactions. However, the evidence supporting this model for type II complexes is not strong, and little is known about what governs the formation of the two types of T-ag complexes. Recently Smelkova and Borowiec (19) found that efficient assembly of type II complexes on replication bubbles requires a ssDNA region of at least 40 nucleotides (nt).
To construct a model for how T-ag interacts with the replication forks during unwinding, we systematically characterized the binding of T-ag to ssDNA and to synthetic DNA forks, in the presence or absence of nucleotide cofactors. AMP-PNP (a non-hydrolyzable ATP analog) and ADP were used to mimic the pre-hydrolysis and post-hydrolysis states, respectively, of T-ag helicase. We varied independently the length of the three components of the forks, the duplex region and the 3Ј and 5Ј ssDNA tails, and found that extension of either the 3Ј ssDNA tail or the duplex markedly stimulated the formation of type II complexes. We further demonstrated that T-ag forms a double hexamer in type II complexes but that both type I and type II complexes contain only a single DNA fork. This result required a revision of the model for how double hexamers are arranged at a fork. We found that type I complexes are extremely stable, independent of the bound nucleotide cofactor. In contrast, type II complexes are bound tightly to their substrate in the presence of AMP-PNP but loosen their grip on the DNA when ADP is bound by the helicase. These differences in DNA binding affinities can explain why type II complexes have much higher helicase activity than do type I complexes.
DNA Substrates-Oligonucleotides were from Operon and purified by PAGE. All DNA forks contained a duplex part and two poly(dT) tails. DNA forks with duplex lengths of 25, 55, or 80 bp were used. The sequences of the 5Ј3 3Ј strand of the duplex are as follows: 1) 25  Preparation of Forks and Origin-containing Duplex-Oligonucleotides were 5Ј-32 P-labeled using T4 polynucleotide kinase and [␥-32 P]ATP (21) to a specific activity of ϳ1⅐10 8 cpm/g. Polynucleotide kinase was heat-inactivated by a 20-min incubation at 65°C, and unincorporated 32 P was removed by passage through a Biospin-6 column (Bio-Rad). Synthetic replication forks and origin duplex were prepared by annealing two partially complementary oligonucleotides in 50-l reaction mixtures containing 5 pmol of labeled oligo and 5 pmol of cold oligo, 50 mM Tris-HCl (pH 8.0), and 10 mM MgCl 2 . Reaction mixtures were heated to 96°C and cooled overnight.
Preparation of a Biotinylated Origin-containing Duplex-The origin duplex, formed by annealing two complementary oligonucleotides of 83 and 84 nt, contained a single nucleotide (A) 5Ј overhang. Klenow fragment was used to fill-in the end of the duplex with a single Biotin-16-dUTP (Roche Molecular Biochemicals). Four pmol of origin duplex with a 5Ј overhang were incubated with 5 units of Klenow fragment (New England Biolabs) at 30°C for 1 h in a 30-l reaction containing 1ϫ DNA polymerase buffer (New England Biolabs) and 100 M biotin-16-dUTP.
T-ag Binding Reactions and Cross-linking-Unless otherwise stated, T-ag⅐fork complexes were assembled in standard reaction mixtures (10 l) containing 50 mM triethanolamine (pH 7.6), 7 mM MgCl 2 , nucleotide cofactor AMP-PNP, ADP, or ATP at final concentrations indicated in the figure legends, 15 fmol (1.5 nM) of labeled fork, 15 fmol (1.5 nM) of labeled SV40 origin-containing fragment, or 100 fmol (10 nM) of ssDNA oligo and the appropriate amount of T-ag. T-ag was diluted in 10 mM Pipes (pH 7.6), 1 mM dithiothreitol, 0.5 mM EDTA, 5 mM NaCl, 10% glycerol and added last to the reaction mixture in 5-l aliquots. When needed, the standard binding reaction was scaled up. The final reaction volume for each experiment is indicated in the figure legends. Binding of T-ag to ssDNA in the absence of nucleotides was carried out in 1 mM MgCl 2 . Binding reactions were incubated for 30 min at 37°C. Glutaraldehyde (final concentration 0.04%) was used to cross-link the protein and to stabilize the T-ag⅐fork complexes before electrophoresis. Crosslinking was quenched by the addition of 10ϫ quenching buffer (100 mM glycine, 10 mM Hepes (pH 7.6)). AMP-PNP is a competitive inhibitor of the helicase activity of T-ag (data not shown).
Helicase Assay-Indicated amounts of T-ag were incubated with radiolabeled DNA forks under standard reaction conditions in the presence of 4 mM ATP. The helicase activity of T-ag leads to melting of the duplex part of the fork and formation of free radiolabeled ssDNA. The helicase reactions were terminated by the addition of SDS and EDTA to final concentrations of 0.5% and 15 mM, respectively. Reactions were further treated with proteinase K (final concentration of 1 mg/ml) for 10 min at 37°C and then analyzed by electrophoresis through an 8% native polyacrylamide gel. Alternatively, helicase reactions were stopped by glutaraldehyde cross-linking and analyzed by native PAGE (3.5% gel).
Gel-shift Analysis-Binding of T-ag to the synthetic forks or origincontaining duplex was analyzed by gel-shift analysis using non-denaturing PAGE (4 or 3.5% gels, 0.5ϫ TBE, 6 -8 V/cm). After electrophoresis, the gels were dried on Whatman paper and autoradiographed or scanned using a Fuji PhosphorImager. Binding was quantified using the Image Gauge version 3.3 software.
Isolation of T-ag⅐fork Complexes on Ferromagnetic Beads-Streptavidin-coated ferromagnetic particles were purchased from Dynal. Before use, the beads were washed 1 time with TE and 3 times with 1 M NaCl and then resuspended in 1 M NaCl at a concentration of 10 mg/ml. Five l of the stock were used for every 30 fmol of biotinylated fork. Beads were added to the reaction mixtures after the assembly of T-ag-DNA complexes was completed and incubated for 15 min at 37°C with occasional mixing. DNA or protein-DNA complexes bound to the beads were isolated from the reaction mixture on magnetic separator stands, washed, and processed as described in the individual experiments.
Assembly State of T-ag on Forks and at the SV40 Origin-Crosslinked T-ag-DNA complexes were pulled down on streptavidin-coated beads, washed in 50 l of TE, and allowed to stand for 3 min at room temperature. The TE wash was repeated 2 more times, and the beads were incubated with 20 l of 1ϫ SDS loading buffer (10 mM sodium phosphate (pH 7.2), 1% SDS, 0.1 M dithiothreitol, 0.005% bromphenol blue, 10% glycerol) for 5 min at 50°C and 15 min at room temperature. The beads were pulled down, and the supernatant, containing the cross-linked T-ag oligomers, was recovered and used for SDS-PAGE.
Denaturing Gel and Western Analysis of Cross-linked T-ag-Crosslinked T-ag was separated on SDS/PO 4 gels according to the procedure of Weber et al. (23). Gels consisted of 0.5% agarose, 3% acrylamide with a ratio of acrylamide/bisacrylamide of 80:1. Gels were run at 0.8 V/cm for 24 h with constant circulation of the buffer and then transferred to polyvinylidene difluoride or nitrocellulose membranes using the semidry transfer system (Bio-Rad). Transfer was carried out in Towbin buffer (without methanol) and lasted for 1.5 h. Membranes were processed for Western analysis using a mouse anti-SV40 antibody (PAb101) (BD Biosciences) and secondary horseradish peroxidase-coupled secondary antibody (Pierce). The protein was detected using the SuperSignal West Pico chemiluminescence detection kit (Pierce). The quantification of the signals was done using a ChemiImager 5500 by Alpha Innotech Corp.
Pull-down Assay-Cross-linked T-ag-DNA complexes were pulled down on streptavidin-coated beads, washed once with 20 l of 1 M NaCl and twice more with 100 l of TE, and resuspended in 20 l of TE. The bead suspension, the supernatants after the initial pull-down (S), and the NaCl wash (W) were treated with proteinase K (final concentration of 1 mg/ml) and SDS (0.5%) at 37°C for 15 min with occasional mixing to prevent sedimentation of the beads. The bead suspension was further treated with NaOH (final concentration of 0.4 N) and incubated for 5 min at room temperature. The beads were pulled down, and the supernatant (E) containing radiolabeled DNA was recovered. Salts and sodium hydroxide were removed from samples (S), (W), and (E) by passage through Biospin-6 columns. The flow-through were collected and analyzed by PAGE.
Gel shifts from the pull-down experiment were analyzed by Phosphor-Imager analysis, and the relative percentages of the individual complexes in the binding reactions were determined for each experiment. The percentages were used to calculate the corrected theoretical ratio (corer. Ratio), according to Equation 1, The calculation assumes that half of type I complexes are formed on biotinylated forks.
Thin Layer Chromatography-One 2-l aliquot containing radioactive nucleotides was spotted onto cellulose polyethyleneimine-thin layer chromatography plates (J. T. Baker Inc.), separated with 1 M formic acid, 0.5 M LiCl, air-dried, and quantified using a Fuji PhosphorImager.
Analysis of the Binding Curves with ssDNA-The total percentage of all T-ag-ssDNA complexes for each protein concentration was plotted against the free concentration of monomeric T-ag. Free T-ag was determined by subtracting the concentration of the bound T-ag from the total protein concentration, assuming that all complexes represent bound hexamers. Binding data were fit to sigmoidal binding curves by nonlinear least square analysis using Prism 3 (GraphPad Software, Inc., San Diego). The Hill coefficient (n app ) and the concentration at the mid-point of the titration curves (S 0.5 ) were determined.
Analysis of the Binding Curves with DNA Forks-For each experiment the relative percentages of type I, type II, type III, and aggregates in each lane were determined from the PhosphorImager analysis of the gel shifts as described above. The data were plotted as percent of complexes (type I or combined type II ϩ type III) versus the total protein concentration expressed as monomers. Under most conditions, type III complexes and aggregates represent a minor fraction of all assembled complexes, except at very high protein concentration. Data sets for each fork were fitted to a sigmoidal binding curve as described for the ssDNA analysis. The Hill coefficient (n app ) and the concentration at the midpoint of the titration curves (S 0.5 ) were determined. The best-fit values for S 0.5 and n app are presented in the legend to Fig. 2 and refer to the individual data sets. Standard errors for all best-fit values were less than 5% for n app and 2% for S 0.5 . The 95% confidence intervals of all best-fit curves were narrow, and R 2 values were greater than 0.99. The statistical significance of the differences between the curves was evaluated by a pairwise F test. We find that the differences between the curves shown are highly statistically significant, with p values less than 5 ϫ 10 Ϫ5 . Three of the presented seven different binding curves in the presence of ADP ( Fig. 2B forks with a 25-and 55-bp duplex, and Fig. 2C forks with a 10-and 30-nt 3Ј tail) and two of the curves in AMP-PNP (data not shown) were determined twice. We find that the difference between the duplicates is not statistically significant as p values using an F-test were greater than 0.03.

RESULTS
Binding of T-ag to ssDNA-As binding of T-ag to synthetic forks is expected to occur, at least in part, through interactions with the ssDNA tails, we began our study by investigating the length dependence of T-ag binding to ssDNA using gel-shift analyses (Fig. 1A). One predominant complex was formed independent of the length of the ssDNA. This complex was previously identified as a hexamer (12,13). Two minor complexes with slower electrophoretic mobility probably represent- ing higher oligomeric states of T-ag were also detected. We found that the affinity of T-ag binding increased with oligonucleotide length, saturating at 45-55 nt (Fig. 1B). The T-ag preference for long ssDNA is manifest in the presence of either AMP-PNP or ADP. In a competition binding experiment, T-ag bound predominantly to a 54-mer, even in the presence of a more than 1000-fold molar excess of a 24-mer (Fig. 1C).
We next compared quantitatively the binding curves for four ssDNA oligonucleotides with lengths of 24, 32, 44, and 54 nt with increasing amounts of T-ag (Fig. 1D). The binding data were fit to Hill equations, and the Hill coefficient (n app ) and S 0.5 values, the protein concentration at the mid-point of the titration curve, were determined ( Table I). The values of n app we obtained are all much greater than 1 (2.4 -4.8), which clearly shows cooperative binding to all the oligonucleotides. The molecular basis for the preference of T-ag for binding to long ssDNA seems to be the increase in cooperativity with increasing oligonucleotide length. Additionally, the Hill coefficient (n app ) generally gives a lower bound to the number of protomers that are recruited into cooperative binding. Thus, n app can be interpreted as the minimum number of T-ag protomers within the hexamer that are bound to ssDNA. We could therefore derive an approximate value for the maximal length of the ssDNA-binding site of a single T-ag protomer by dividing the length of the ssDNA oligo by n app . The result for all four oligonucleotides is that each T-ag protomer binds ϳ9 -10 nt (Table I). We also found an interesting dependence of T-ag binding on the absolute concentration of ssDNA oligonucleotide ( Fig. 1E and Table II). The cooperativity of binding to a 54-nt DNA decreased as its concentration was increased. This result can be explained if multiple sites on the T-ag hexamer simultaneously bind to the ssDNA. At low concentrations of ssDNA, successive protomers will be recruited to the same ssDNA ( Fig.  1F, dashed line). However, as the concentration of ssDNA in solution is increased, separate ssDNA molecules (solid line) would be bound, resulting in a loss of the free energy associated with the binding of adjacent sites on a single DNA molecule.
We examined whether the nucleotide cofactors affect the binding of T-ag to ssDNA (Fig. 1G). Although T-ag binds cooperatively to a 54-mer under all three conditions, the presence of nucleotide cofactors clearly makes binding stronger and more cooperative. Thus, we conclude that T-ag binds cooperatively, in a length-dependent fashion to ssDNA, in a reaction stimulated by nucleotide cofactors.
Binding of T-ag to DNA Forks, Effect of Fork Length and Nucleotide Cofactors-We next turned our attention to the analysis of how T-ag binds to synthetic forks, as a model for the in vivo substrate. As in the ssDNA binding experiments, we investigated first the effect of the length of the DNA forks and, second, the effect of nucleotide cofactors on the binding affinity of T-ag. We focused on data for the formation of type II complexes, as they are the active helicase species. Binding data for type I complexes are shown where they emphasize important distinctions from type II complexes. To determine the effect of the fork length, we independently varied the length of the 3Ј ssDNA tail, the 5Ј ssDNA tail, or the duplex region starting with a fork consisting of a 25-bp duplex, a 30-nt 5Ј ssDNA tail, and a 10-nt 3Ј ssDNA tail. We measured T-ag binding by gel shifts in protein titration assays. A representative experiment with this fork is shown in Fig. 2A. As the amount of T-ag was increased, type I complexes were formed first, but at higher protein concentrations type II complexes were the predominant species. A third minor complex, called type III, also formed at high protein concentration, as did aggregates. Type III complexes form when all of the free DNA fork is already bound and probably reflect binding of T-ag to type II complexes. For this reason we add the usually small amount of type III complexes to the type II complexes in quantifying the results (see "Experimental Procedures").
We first examined the effect of the duplex length on T-ag binding. The percentage of type II complexes formed at each protein concentration was plotted along with the binding curves for two different DNA forks (Fig. 2B). The length of the two ssDNA tails was unchanged, whereas the duplex was extended from 25 to 55 or 80 bp. Extending the duplex to 55 bp stimulated markedly the formation of type II complexes, whereas a further increase in length to 80 bp had no additional effect. Formation of type II complexes on all three forks is clearly cooperative, and the slope of the binding curves increases as the length of the duplex is extended from 25 to 55 bp, indicating increasing cooperativity of T-ag binding.
We then analyzed the effect of extending the 3Ј ssDNA tail. We compared forks with a 3Ј tail of 10, 30, or 60 nt, a 5Ј tail of 30 nt, and a duplex of 25 bp (Fig. 2C). Extending the 3Ј ssDNA tail stimulated the formation of type II complexes by increasing the cooperativity of T-ag binding, similar to what we observed when we increased the duplex length. However, with forks containing a 55-bp duplex, the assembly of type II complexes was only modestly enhanced by extending the 3Ј ssDNA tail from 10 to 60 nt (Fig. 2D). Together these results imply that in type II complexes, T-ag interacts extensively with both the duplex part of the fork and the 3Ј ssDNA tail.
The finding that the assembly of type II complexes is sensitive to the length of the DNA forks prompted us to investigated whether the formation of type I complexes is also DNA lengthdependent. The combined binding data for the formation of type I complexes on seven forks (Fig. 2, B-D) of different lengths in the low protein concentration range (in which the percentage of type II complexes was less than 15%) are plotted in Fig. 2F. The data can all be fit well to a single sigmoidal binding curve (R 2 ϭ 0.97), implying that type I complex formation is insensitive to an increase in the length of the fork beyond that of the shortest fork used (25 bp in the duplex, 10 nt in the 3Ј ssDNA tail, and 30 nt in the 5Ј ssDNA tail), even though formation of type I complexes is also cooperative (Hill coefficient of 3.3).
The second aspect of our studies of T-ag binding to synthetic forks was the analysis of the effect of different nucleotide cofactors. As we observed with ssDNA, we found that the binding affinity of T-ag was similar in the presence of ADP or   AMP-PNP. Comparative binding curves for type II complexes formed in the presence of either ADP or AMP-PNP on the highest affinity fork (55-bp duplex, 60-nt 3Ј tail, and 30-nt 5Ј tail) are presented in Fig. 2E. Moreover, we found that extending the length of the duplex and the 3Ј ssDNA tail stimulates the formation of type II complexes in the presence of both ADP and AMP-PNP (data not shown).
The final fork component we varied was the length of the 5Ј ssDNA tail. These experiments also revealed an interesting dependence of the ratio between type I and type II complexes on the concentration of the nucleotide cofactor. We compared the binding of T-ag to forks with a duplex of 25 bp, a 3Ј tail of 30 nt, and a 5Ј tail of 0, 8, 20, or 30 nt (Fig. 2G). In the presence of 1 mM AMP-PNP the complete removal of the 5Ј tail reduced the efficiency of type II formation, so that type I complexes were favored. As the length of the 5Ј tail was increased from 0 to 20 nt, type II assembly was stimulated ϳ5-fold. However, the stimulation was less than 2-fold if 4 mM AMP-PNP was used instead of 1 mM. The nucleotide cofactors could affect the binding of T-ag by chelating Mg 2ϩ and thus modulating the concentration of free Mg 2ϩ . To test this hypothesis we analyzed the binding of T-ag to forks in the absence of nucleotide cofactors.
The omission of the nucleotide cofactors from the binding reaction resulted in complete aggregation of T-ag associated with the DNA forks (data not shown). In order to prevent T-ag from aggregating, we lowered the concentration of Mg 2ϩ . From the binding of T-ag (400 ng) to a fork with a 55-bp duplex and two 30-nt ssDNA tails in the presence of increasing amounts of Mg 2ϩ but absence of nucleotide cofactors (Fig. 2H), we draw several conclusions. First, T-ag binds poorly to the forks without Mg 2ϩ . Second, the Mg 2ϩ concentration modulates the ratio of type I and type II complexes. At low Mg 2ϩ concentrations, type I complexes were formed exclusively. As the concentration of the divalent ions was increased, type II complexes, type III complexes, and aggregates formed. Consistent with our findings are the results of Wessel et al. (11) who found that Mg 2ϩ induces the aggregation of T-ag hexamers bound to distant sites on a non-origin containing duplex DNA. Interestingly, such aggregated T-ag could catalyze nonspecific duplex unwinding upon addition of ATP, suggesting that the Mg 2ϩinduced association could be physiologically significant.
Further insight into how type II complexes are assembled was gained through a reciprocal binding assay, in which we kept the amount of T-ag constant, while increasing the concentration of the fork (Fig. 2I). Even at the highest concentration of DNA, the concentration of T-ag hexamers was in excess over the DNA forks. At 1.5 nM, the fork was completely bound, and mostly type II complexes were formed. As we increased the fork concentration up to 22.5 nM (15-fold increase), the fraction of type II complexes gradually decreased, whereas the percentages of type I complexes and unbound fork steadily increased. This result argues against the model diagramed in Fig. 2J whereby type II complexes are formed by dimerization via protein-protein interactions of type I complexes bound to separate DNA forks (17). This model predicts that an increase in the concentration of type I complexes, brought about by increasing the total DNA concentration, will result in increased type II formation. We observe the opposite. We suggest instead that type II complexes are a double hexamer bound to a single fork (Fig. 2K) and provide direct evidence below.

Structure of Type II Complexes, Assembly State of T-ag and
Number of DNA Forks-To determine the stoichiometry of protein-to-DNA forks in type II complexes, we measured directly the number of forks in these complexes and whether they contained single or double hexamers. We examined the assembly state of T-ag in the oligomeric complexes formed on DNA forks by determining via denaturing gel electrophoresis the mass of the protein component of these complexes. To prepare size markers for these experiments, we partially cross-linked pre-assembled T-ag hexamers in the presence of nucleotide cofactors but absence of DNA, and then we separated the different species on a denaturing composite agarose-polyacrylamide gel (Fig. 3A). This generated a ladder of protein bands we identified as cross-linked dimers through hexamers. The mobility of the cross-linked species was linearly dependent on the logarithm of the molecular weight of the corresponding T-ag multimers, increasing our confidence in the assignments (Fig. 3B). The number of monomers assigned to each band was I complexes is included for comparison. All reactions contain 4 mM ADP and 400 ng of T-ag (5 ϫ 10 Ϫ7 M). J, a schematic representation of the previously proposed assembly pathway for type I and type II complexes. T-ag subunits are depicted as spheres. Six T-ag monomers assemble into a hexamer bound to a DNA fork to form a type I complex. The protomers in the bound T-ag hexamer are numbered 1-6. Two type I complexes were proposed to dimerize and form a type II complex. K, model for the structure of type II complexes. We propose that type II complexes contain a double hexamer of T-ag bound to a single DNA fork. further corroborated by comparison of the mobility of the T-ag multimers with the mobility of cross-linked phosphorylase b markers after silver staining on pure polyacrylamide gels (data not shown). We confirmed that in the presence of nucleotide cofactors T-ag forms mostly hexamers. The two additional bands migrating slower than the hexamer were designated DX1 and DX2, as we believe they are double hexamers (see "Discussion"). The apparent molecular masses of DX1 and DX2, 650 and 840 kDa, respectively, are both smaller than the predicted molecular mass of the double hexamer (942 kDa). A deviation from linear resolution is, however, not unexpected in this high molecular mass range.
We then analyzed the formation of T-ag oligomers at different protein concentrations after extensive cross-linking to trap hexamers and other higher order association states in the absence of DNA. T-ag oligomers smaller than the hexamer were barely detected due to the high cooperativity of selfassembly of T-ag (data not shown), ADP, or AMP-PNP-promoted oligomerization (Fig. 3C). We find that the T-ag hexamer is the predominant species at low protein concentration (0.3 and 0.4 g) in the presence of either ADP or AMP-PNP. A small amount of the DX1 oligomer is also present, consistent with our recent EM studies (24) that showed that 8% of all complexes formed at this protein concentration are double hexamers. ADP was much more effective than AMP-PNP in stimulating DX2 assembly at high protein concentration.
DX2 has a comparable electrophoretic mobility to the known double hexamer that assembles at the origin of replication. An 84-bp biotinylated fragment containing the core SV40 origin was used to form T-ag-origin complexes at increasing concentrations of T-ag (Fig. 4A). After cross-linking, a fraction of the reactions was run on a native PAGE gel to assess the extent of binding (Fig. 4B). The remainder of the complexes was purified on streptavidin-coated magnetic beads. The protein was extracted from the DNA and analyzed by SDS-PAGE and Western blotting (Fig. 4C). Consistent with previous studies (5,25), T-ag formed a double hexamer with the origin fragment (Fig. 4, B and C). The single hexamer (SX) complex, detected at high concentrations of T-ag, is a result of nonspecific binding to the ferromagnetic beads. The origin double hexamer has mobility similar to DX2, thus supporting the assignment of DX2 as a double hexamer.
With an understanding of the T-ag oligomers formed in the absence of DNA, we proceeded to ask what was the assembly state of T-ag bound to DNA forks in type II complexes. We used two types of biotinylated DNA forks. One had an 80-bp duplex and 10 nt 3Ј ssDNA tail (Fig. 4D), and supported the efficient formation of type II complexes. The second type of fork had a 55-bp duplex and 60-nt 3Ј tail (Fig. 4E) and promoted the assembly of both type II and type III complexes. The T-ag complexes with forks containing an 80-bp duplex were crosslinked with glutaraldehyde. Part of the reaction mixtures was analyzed by gel-shift analysis (Fig. 4D, left panel), whereas the rest was incubated with streptavidin-coated beads to purify bound from free T-ag. The DNA-bound T-ag oligomers were extracted with SDS-containing buffer, separated on a SDS/PO 4 gel, and analyzed by Western blotting (Fig. 4D, right panel). Under conditions where T-ag forms mostly type II complexes, the primary oligomeric species extracted off the DNA is DX2. We conclude that type II complexes are indeed double hexamers. At the same concentration of T-ag, but in the absence of DNA, DX2 is not formed (see Fig. 3C, AMP-PNP panel, 300 and 400 ng), indicating that DNA forks stimulate the formation of the double hexameric complex. Forks with a 60-nt ssDNA tail promote the assembly of type III as well as type II complexes (Fig. 4E, left panel). The Western blot showed an oligomeric species with apparent molecular mass of ϳ1200 kDa in addition to DX2 that we designate as TX because we believe it is a triple hexamer of T-ag (Fig. 4E, right panel). The TX oligomer is most probably the protein component of type III complex.
Knowing that the active helicase complex, type II, contains a double hexamer of T-ag, we next determined whether these complexes contain one or two forks (Fig. 2, J and K). To do so we SV40 T-ag Binding to DNA used a pull-down assay with two radiolabeled DNA forks (Fig.  5A) that similarly promote the formation of type II complexes, yet differ in length and therefore electrophoretic mobility. T-ag was bound to an equimolar mixture of the two forks, only one of which was biotinylated, under conditions favoring the formation of type II complexes (Fig. 5B). The complexes were cross-linked with glutaraldehyde to stabilize protein-DNA interactions and prevent disassembly during subsequent manipulations. The protein-DNA complexes were bound to streptavidin-coated ferromagnetic beads, which were then washed vigorously to reduce nonspecific binding. The attached complexes were then deproteinized by treatment with SDS/proteinase K and treated with 0.4 N NaOH to denature the DNA. The released radioactive DNA was analyzed by native PAGE. If type II complexes represent a double hexamer bound to a single fork, then only the biotinylated fork would be retained on the beads. However, if type II complexes are T-ag double hexamers bridging two forks, then we would pull down both forks in a ratio depending on the initial input.
A representative experiment is shown in Fig. 5C in which the 3Ј ssDNA tails of the two forks used differed by 20 nt (10 versus 30 nt). The longer fork was biotinylated at the 5Ј ssDNA tail. In addition to the sample extracted from the beads (E), we analyzed the supernatant (S) after the pull-down and the first wash (W). In the control reaction in the absence of T-ag we recovered only the labeled strand of the biotinylated fork, as expected (lane 8). In the presence of T-ag we recovered the radiolabeled 3Ј strands of both forks in a 9.5:1 (biotinylated/ non-biotinylated) ratio (lane 11) (see Table III, Experiment I). If type II complexes had two bound forks, the expected ratio is about 1.5:1. As type II complexes coexist with small amounts of free fork DNA and type I complexes, the corrected ratio is 1.6:1 in Table III, Experiment I. The results of four individual pulldown experiments presented in Table III show only small contaminating amounts of the non-biotinylated fork. Therefore, type II complexes contain only a single fork.
Stability of Type I and Type II Complexes-So far we have characterized the parameters that govern the assembly of type I and type II complexes. To gain further insights into the protein-DNA interactions in these complexes, we studied their dissociation kinetics. The dissociation of T-ag bound to a labeled fork was measured by gel-shift analysis after the addition of heparin (an ionic competitor for DNA binding) or excess cold fork. The amount of heparin was sufficient to block complex formation if added to the reaction before T-ag (Fig. 6B, lane 1). After heparin challenge type I complexes were stable in the presence of either AMP-PNP or ADP (Fig. 6, A and B). In contrast, type II complexes were stable in the AMP-PNP-bound form but readily dissociated in the ADP-bound form. Thus, the protein-DNA contacts within type I and type II complexes differ substantially.
The dissociation rate of type II complexes, containing labeled forks with a 55-bp duplex and two 30-nt ssDNA tails and the presence of ADP, was measured quantitatively. Net dissociation of labeled fork was started by addition of a 100-fold excess of cold fork. The spectrum of T-ag complexes remaining at times thereafter was measured by gel-shift analysis (Fig. 6C). The apparent rate constants and corresponding half-lives (t1 ⁄2 ) for the individual complexes and the appearance of protein-free fork were determined by fitting the results to a one-or twophase exponential (Fig. 6D). Based on the results of two independent experiments, we determined that type II complexes in the ADP-bound form disappear with a half-life of 1.3 min. Type III complexes and the aggregated material have a similar halflife of 1.5 min. Dissociation of type II, type III, and the aggregates leads mostly to formation of type I complexes, which, in contrast, are very stable and accumulate with time (Fig. 6C). A smaller part of the initial T-ag⅐fork complexes dissociates directly to free DNA. The rate of appearance of the free fork was fit to a FIG. 5. Analysis of the number of DNA forks bound in type II T-ag⅐fork complexes. A, outline of the assay. Schematic diagrams of the forks used (FBio30 and F10) are shown. The forks were radiolabeled at the 5Ј end of the duplex. The forks differed in the length of the 3Ј ssDNA tail and in that FBio30 was biotinylated at the end of the 5Ј tail. T-ag⅐fork complexes were assembled in 20-l reactions. 5 l were used for the gel-shift analysis, whereas the remainder was used for the pull-down assays. PrK, proteinase K treatment. B, gel-shift analysis of the assembled T-ag⅐fork complexes. The positions and relative percentages of the free DNA forks, type I, type II, type III complexes, and aggregates (Aggr.) are indicated next to the corresponding bands. C, PAGE analysis of the pull-down assay using two forks. The assay was performed in the absence (lanes 6 -8) or presence of T-ag (400 ng) (lanes 9 -11). Radioactivity recovered in the supernatant (S) after the initial binding to the ferromagnetic beads, the first wash (W) and the extracted DNA (E) after the NaOH treatment. I, input ratio of radiolabeled forks. Also presented are controls of the individual forks in the absence (lanes  1 and 2) or presence (lanes 3 and 4) of streptavidin, and the radiolabeled strands of the two forks (lanes 12 and 13). In lane 4 the shift in the position of the FBio30 fork is due to the binding of streptavidin. The small amount of non-biotinylated fork pulled down on the streptavidincoated beads in the presence of T-ag is most likely due to nonspecific binding of T-ag complexes to the beads, aggregation that traps together complexes assembled on either fork, or may represent a small fraction of a specific complex with two forks. biphasic curve with a fast phase (t1 ⁄2 ϭ 0.28 min), corresponding to the release of the fork from type II and type III complexes and aggregates, and a slow phase (t1 ⁄2 ϭ 98 min), corresponding most probably to the release from the more stable type I complexes, generated by the dissociation of higher order complexes.
We next performed a similar quantitative dissociation experiment of T-ag complexes assembled on a labeled 54-nt ssDNA in the presence of ADP, where type I complexes predominate. A 100-fold excess of cold 54-nt ssDNA initiated dissociation. The T-ag-ssDNA-ADP complexes were extremely stable (Fig. 6D). We conclude that type I T-ag-DNA complexes are very stable independent of whether they form on ssDNA or a synthetic fork.
Helicase Activity of Type I and Type II Complexes-So far we have investigated the DNA binding properties of T-ag in the presence of ATP analogs, which trap the helicase in different states of the unwinding cycle. We now turn our attention to the condition that allows helicase activity, the presence of ATP. Under these conditions, T-ag separates the two strands of the synthetic fork, which leads to accumulation of ssDNA (Fig. 7A). Similar to previous studies (17), we observed that the fraction of substrate unwound increases with increasing amounts of T-ag, but not all of the substrate is unwound. One explanation for the limited unwinding is that only a subset of the assembled T-ag⅐fork complexes is active. If type II complexes are the active helicase (17), then pre-assembling these complexes before the onset of the strand separation would increase the amount of unwinding. We tested this hypothesis by first forming type II complexes in a low concentration of AMP-PNP, after which they were made competent for unwinding by addition of excess ATP. The time course of formation of ssDNA for two concentrations of T-ag was followed by PAGE and is plotted in Fig. 7B. The indicated percent of pre-assembled type II complexes in each reaction was determined by gel-shift analysis. The helicase activity was robust, with the 25-bp duplex region unwound within less than a minute after the addition of ATP. The plateau value of the unwinding matched the amount of preformed type II complexes, as expected if they are the active helicase species.
This conclusion was corroborated and extended by analysis of the turnover rates of type II complexes. T-ag and forks with a 55-bp duplex were mixed in the presence of ATP. At times thereafter we determined by gel-shift analysis the relative amounts of labeled DNA fork bound in type I and type II complexes, as well as the amount of ssDNA product (Fig. 7, C  and D). Both type I and type II complexes assembled rapidly within less than a minute and then turned over. We fit the decay phase to an exponential equation and calculated the rates of turnover (k). From the average of two experiments we estimate that type II complexes (k ϭ 0.49) turn over 9 times faster than type I complexes (k ϭ 0.057), and the rate of disappearance coincided with the rate of appearance of ssDNA (k ϭ 0.52). It is clear that type II complexes are the active helicase assembly.
DNA unwinding relies on the coupling of cyclic changes in DNA binding to ATP binding and hydrolysis. The finding that type I complexes remain bound to synthetic forks for a longer time than type II complexes suggested to us that they may have a defect in the ATP hydrolysis cycle. T-ag bound in a type I complex may not be able to bind and/or hydrolyze ATP well, or TABLE III Predicted and observed fork ratios from the pull-down assays The corrected theoretical ratio reflects the ratio of forks predicted from the dimerization model (Fig. 2J), corrected for the presence of type I complexes and free DNA fork in the binding reactions as described under "Experimental Procedures."  the products of hydrolysis may release too slowly for the cycle to continue efficiently. To distinguish among these alternatives, we determined the nucleotide cofactors bound to T-ag in type I complexes. We purified these complexes in the following way. T-ag was incubated with a radiolabeled biotinylated fork in the presence of 4 mM ATP. After 5 min, a 50-fold excess of cold DNA was added to soak up free T-ag. After cross-linking with glutaraldehyde, an aliquot of the reaction mixture was analyzed by native PAGE (Fig. 7E). Most of the stable protein-DNA complexes are type I complexes. The remainder of the reaction was adsorbed to streptavidin-coated magnetic beads to purify the stable T-ag⅐fork complexes. The assembly state of T-ag in these complexes was analyzed by Western blotting (Fig.  7F), which demonstrated that these complexes are indeed single hexamers.
To determine the nucleotide-cofactors bound in the single hexameric T-ag⅐fork complexes, cold ATP was replaced by [␣-32 P]ATP, and the steps of the previously described experiment were repeated. At the end of the reaction, all DNAcontaining species were pulled down with streptavidin-coated ferromagnetic beads, and the bound nucleotides were analyzed by thin layer chromatography (Fig. 7G). Both ADP and ATP were bound to the T-ag⅐fork complexes, and the ATP/ADP ratio was 5.4:1. The simplest interpretation is that approximately one of the subunits in the T-ag hexamer is bound to ADP and the rest to ATP. Both ATP and ADP were stably bound because they were not washed away by cold ATP. The total number of bound nucleotides, ϳ6(5.4 ϩ 1), is consistent with our own (data not shown) and previously published estimates for the number of nucleotides bound in the T-ag hexamer (26).
Because the T-ag hexamer in type I complexes can clearly hydrolyze ATP, we suggest that the unwinding defect of these complexes is associated with failure to continue the hydrolysis cycle after the first of the six ATPs is hydrolyzed. Perhaps an inability to release ADP from the first nucleotide-binding site of the T-ag hexamer blocks the subsequent hydrolysis events.

DISCUSSION
To help understand the mechanism of the T-ag helicase, we undertook a thorough characterization of T-ag interaction with a synthetic DNA replication fork. The pattern of DNA binding by T-ag is complex, as the binding is linked to the assembly of higher order oligomeric structures of T-ag. In a previous study, Smelkova and Borowiec (17) detected by gel-shift analysis two forms of oligomeric T-ag associated with forks and showed that the slower mobility complex (referred to as type II in this paper) was the more active helicase. They speculated that this active form contained a double hexamer, with each hexamer loaded on a separate fork. Our data show that the type II complexes are indeed the active helicase assemblies, but they are double hexamers bound to a single fork. Additionally, we compared the DNA binding properties of the single and double hexameric complexes of T-ag in two ways. First, we identified the parameters that affect the assembly of the two major oligomeric complexes as follows: the concentrations of T-ag and of DNA; the type of DNA substrate (ssDNA, synthetic forks, and origin-containing duplex); the length of the DNA forks; nucleotide cofactors; and Mg 2ϩ concentration. Second, we studied the energetics of the interactions through measurements of the kinetics of dissociation of T-ag-DNA complexes. Our results give credence to the idea that the double hexamer, rather than the single hexamer, is the active helicase machine at the replication fork. Our explanation for the robust helicase activity of the T-ag double hexamer is its ability to cycle timely between a high and low affinity state for the DNA forks, coupled to the binding and hydrolysis of ATP, to allow the unwinding and subsequent translocation along the DNA. In contrast, the single hexameric complex locks tightly to the DNA fork for an inappropriately long time, only slowly progressing through the ATP binding/ hydrolysis cycle and thus failing to bring about fast unwinding (Figs. 6 and 7). The percent of preassembled type II complexes, as determined by gel-shift analysis, is indicated on the right side of the graphs. Four mM ATP was added to the reactions to initiate unwinding and product was measured after 10, 20, 30, 45, 60, and 120 s. The black lines represent best fits to first-order reactions. The plateau values for the best-fit curves are 39% () and 64% (f). C, a helicase assay was carried with 0.5 M T-ag using 6 nM forks with a 55-bp duplex region and 30-nt ssDNA tails. Aliquots were taken at 0.3, 0.7. 1, 2, 3, 4, 5, 6, 7, and 9 min, cross-linked with glutaraldehyde to stop the unwinding, and analyzed by PAGE. D, quantification of the results in C. Solid lines are single exponential fits to the data. E, a helicase reaction using 1.5 nM radiolabeled, biotinylated DNA forks was carried out with 0.6 M T-ag. After 5 min, cold non-biotinylated fork was added. After 5 more min the reaction was terminated by glutaraldehyde cross-linking. The single-stranded products of DNA unwinding, the fork substrate, and the remaining T-ag⅐fork complexes were separated by native PAGE. D, denatured fork. A diagram of the fork used is shown next to the gel panel. F, after crosslinking, the stable T-ag⅐fork complexes were pulled down on ferromagnetic beads and analyzed by Western blotting. A control reaction contained no DNA. The positions of the single hexamer (SX) and the double hexamer (DX1) are indicated. G, a standard helicase reaction in a 20-l volume was performed with a biotin-labeled fork but in the presence of [␣-32 P]ATP. The stable T-ag⅐fork complexes were purified from the reaction mixture using streptavidin-coated magnetic beads. The nucleotide cofactors bound by the protein were extracted and separated by thin layer chromatography (TLC). The observed ratio of ATP/ADP averaged in two independent experiments is 5.4:1.
The identification of the double hexamer as the active helicase assembly provides support for the model for bi-directional unwinding from the SV40 origin in vivo, in which the two hexamers remain attached to each other with the DNA actively spooled through the protein complex (10). The alternative model, in which the two hexamers each travel away from the origin of replication along with a growing fork, is not attractive in light of the finding that the single hexamers are poor at unwinding. Our data show that even with a single fork, a double hexamer is needed for efficient unwinding. We did not find two forks bound to a double hexamer in type II complexes most probably because forks were limiting in concentration. This would not be an issue at the SV40 origin in vivo, where the two forks would be in close proximity and their local concentration would be very high.
Our findings may have implications for the higher order organization of the replication machinery. In prokaryotes, cytological evidence supports the conclusion that the two forks of bi-directional replication from an origin remain juxtaposed through DNA synthesis and are coordinately regulated. This model has been dubbed the factory model (27). The observation of dense zones of replication in high eukaryotes has led to the speculation that many such factories may be localized and jointly controlled (28). Hexameric rings can form higher oligomers in several different ways (24). Therefore, they may provide some of the interacting surfaces that hold the factories together and may serve as platforms for assembly of the multimolecular replication complexes.
T-ag Double Hexamers-The profound effect of the oligomeric structure on the helicase activity of T-ag shows that the interactions between the hexamers are critical for efficient ATP hydrolysis and DNA unwinding. Unfortunately, the structure of T-ag double hexamer is not as well studied as the single hexamer. The latter is a planar ring with two different faces: an N-terminal face formed by the six J-domains and a C-terminal face formed by the C-terminal domains (24). The double hexamer can exist in several alternative configurations (24). Van-Loock et al. (24) observed association of single hexamers through the N-terminal faces (N-N double hexamer) and through the C-terminal faces (C-C double hexamer). A double hexamer, in which the two individual hexamers appear to interact through their J-domains, was observed by Valle et al. (9) bound to the SV40 origin of replication. However, mutational studies suggest that the linker between the J-domain and the origin DNA binding domain (OBD) and the OBD itself are the domains required for bridging the two hexamers (29). Interestingly, neither the OBD nor the linker contact each other in the two currently described structures of T-ag double hexamers (N-N and C-C). This finding raises the possibility that the helicase complex active in bi-directional DNA unwinding may represent still a third configuration of double hexamers.
We identified two T-ag oligomers, DX1 and DX2, that after cross-linking had a lower electrophoretic mobility than the single hexamers on denaturing PAGE. The identification of DX2 as a double hexamer is straightforward based on its apparent molecular weight (840 kDa) and an electrophoretic mobility similar to the origin-bound double hexamer. Our best argument that DX1 is also a double hexamer is that under conditions where we observe only DX1, EM shows that 8% of all protein complexes are double hexamers (24). The alternative is that DX1 is an oligomer intermediate between the hexamer and the double hexamer that arises because of its inherent stability or incomplete cross-linking. We do not favor this alternative for two reasons. First, formation of the T-ag hexamer is highly cooperative and intermediates in the assembly are rare (Refs. 30 and 31 and data not shown); EM did not reveal any evidence for the existence of intermediates between the hexamer and the double hexamer (24). Second, the extent of cross-linking did not influence the amount of DX1 (data not shown).
What do these multiple configurations of T-ag double hexamers represent? We speculate that the identified structures of the double hexamer may reflect intermediates in the conversion of the static double hexamer bound to the origin to a moving helicase. A more trivial explanation is that they arise from nonspecific association between two single hexamers.
Binding of T-ag to ssDNA and Synthetic DNA Forks-As ssDNA binding is essential for the helicase activity of T-ag, we studied how T-ag engages ssDNA. The increase in the cooperativity of T-ag binding with the length of the ssDNA, as well as the decrease in the cooperativity of binding of T-ag with increasing concentrations of ssDNA, indicate that multiple adjacent sites on the T-ag hexamer are involved in interactions with ssDNA. This finding was unexpected because hexameric DNA helicases are generally thought to interact with ssDNA via only one or two of their subunits (32). T-ag seems more similar to the hexameric RNA helicase rho, which interacts with about 70 nt of ssRNA by simultaneously engaging all six binding sites on the hexamer (33). ssDNA probably winds around the T-ag hexamer contacting each subunit.
Our data provide an upper limit for the size of the ssDNAbinding site per T-ag protomer of 9 -10 nt, similar to that found with other helicases. Quantitative binding studies with PriA (34) and rho (35), as well as the co-crystal structures of the Rep (36), hepatitis C virus NS3 (37), and PcrA (38) helicases with ssDNA or partial duplexes show that generally 7-8 nt are bound by each subunit.
Earlier biochemical analyses had suggested a much longer ssDNA-binding site for DnaB (39) and T7 gp4 helicase (22), 20 and 30 nt, respectively. Thermodynamic studies with DnaB, however, showed that the ssDNA-binding site is actually built of two subsites, each of which encompasses ϳ10 nt (40). The recently determined crystal structure of T7 gp4 helicase showed that the ssDNA-binding site could extend to even three subunits without any steric constraints (41). If so, the size of the binding site allocated to each subunit would be ϳ10 nt. We conclude that many helicases may bind to ssDNA in a similar way with a binding site size of 8 -10 nt per monomer. The questions that remain open, however, are why all subunits of the T-ag hexamer appear to engage the ssDNA simultaneously, and whether this represents an important distinction with other helicases.
The two major T-ag⅐fork complexes, type I and type II, differ as follows: 1) in protein-DNA contacts; 2) in T-ag assembly states; 3) in the Mg 2ϩ requirements for assembly; and 4) in helicase activities. Type II complexes are markedly favored with forks containing a long duplex or a long 3Ј ssDNA tail, suggesting that T-ag interacts extensively with both fork regions in these complexes. In contrast, for type I complexes the contacts between T-ag and the fork are more limited. The differences in DNA-protein contacts as directly measured by footprinting will be presented in another study. 2 In brief, the contacts in type II complexes are very different form type I complexes and are much more extensive in the duplex and in the 3Ј ssDNA tail. These results are consistent with our finding here that double hexamer formation is poorly supported by ssDNA alone, yet markedly promoted on forks with a long duplex region.
The differences in protein-DNA contacts between type I and type II complexes are manifested also by the contrasting kinetics of dissociation. The active type II complexes show a clear nucleotide dependence of the T-ag DNA binding affinity, whereas in type I complexes T-ag is tightly bound to the DNA fork independent of the nature of the bound nucleotide cofactor. Type I complexes contain T-ag that is trapped on DNA forks in a state in which at least one subunit has hydrolyzed ADP, whereas the other subunits are in an ATP-bound state (Fig.  6F). Presumably, in type II complexes the T-ag helicase moves rapidly along the DNA substrate (300 bp unwound per min, data not shown) apace with the ATPase cycle.
How does the oligomeric structure influence the ATPase activity of T-ag? The recently solved structures of hexameric helicases show that ATP is cradled between the subunits (41,42). ATP hydrolysis and release of the phosphate would untether the two subunits sharing the ATP and allow an intersubunit rearrangement, most probably a relative rotation. Such conformational changes could be associated with a high energetic barrier in single T-ag hexamers bound to DNA forks that is alleviated in double hexameric complexes.
We see two possible ways whereby the double hexameric structure could influence the rates of subunit rotations and thus affect the helicase activity of T-ag. First, the double hexameric structure could change the number of T-ag protomers bound to the DNA. The rate of ATP processing/conformational change at a pair of T-ag protomers could be affected by whether or not their immediate neighbors also contact DNA. Only two of the six protomers of type I complexes change their conformations in the presence of DNA and are presumably bound to the DNA fork (24). Given the requirement for long forks for the assembly of the double hexamer (Fig. 2, B-D), and that all six protomers of the single hexamer seem to bind long ssDNA (Fig.  1, A-D), more than two T-ag protomers may be bound to DNA in type II complexes.
Second, the formation of the double hexamer could lead to a global change in the conformation of the individual hexamers that affect the T-ag intersubunit contacts. Nucleotide cofactors, which probably bind at the interface between the subunits, influence differently the association of T-ag hexamers, with ADP promoting the formation of double hexamer (DX2) substantially better than AMP-PNP (Fig. 3C). Additionally, formation of the double hexamer on DNA depends strongly on the concentration of Mg 2ϩ (Fig. 2H). The finding that an increase of only 1 mM Mg 2ϩ is sufficient to shift the assembly of T-ag⅐fork complexes from type I to predominantly type II suggests a cooperative change in the structure of the oligomeric complex. Other hexameric helicases are known to exhibit substantial polymorphism at the level of quaternary structure. Interconverting populations of hexamers with 3-fold (C 3 ) or 6-fold (C 6 ) axes of symmetry have been observed by EM for Escherichia coli DnaB (43), bacteriophage SPP1 G40P helicase (44), and the close functional homolog of T-ag, the bovine papillomavirus E1 protein (45).