Site-directed Mutations of T4 Helicase Loading Protein (gp59) Reveal Multiple Modes of DNA Polymerase Inhibition and the Mechanism of Unlocking by gp41 Helicase*

The T4 helicase loading protein (gp59) interacts with a multitude of DNA replication proteins. In an effort to determine the functional consequences of these protein-protein interactions, point mutations were introduced into the gp59 protein. Mutations were chosen based on the available crystal structure and focused on hydrophobic residues with a high degree of solvent accessibility. Characterization of the mutant proteins revealed a single mutation, Y122A, which is defective in polymerase binding and has weakened affinity for the helicase. The interaction between single-stranded DNA-binding protein and Y122A is unaffected, as is the affinity of Y122A for DNA substrates. When standard concentrations of helicase are employed, Y122A is unable to productively load the helicase onto forked DNA substrates. As a result of the loss of polymerase binding, Y122A cannot inhibit the polymerase during nucleotide idling or prevent it from removing the primer strand of a D-loop. However, Y122A is capable of inhibiting strand displacement synthesis by polymerase. The retention of strand displacement inhibition by Y122A, even in the absence of a gp59-polymerase interaction, indicates that there are two modes of polymerase inhibition by gp59. Inhibition of the polymerase activity only requires gp59 to bind to the replication fork, whereas inhibition of the exonuclease activity requires an interaction between the polymerase and gp59. The inability of Y122A to interact with both the polymerase and the helicase suggests a mechanism for polymerase unlocking by the helicase based on a direct competition between the helicase and polymerase for an overlapping binding site on gp59.

The T4 replisome is considered to be a model for the more complex replication systems (1,2). The eight protein replisome can be divided into three modules, the leading strand holoenzyme, the lagging strand holoenzyme, and the primosome (3). The leading and lagging strand holoenzymes are formed by the polymerase (gp43) 4 and the polymerase clamp (gp45) with the aid of the clamp loader protein (gp44/62). Contained within gp43 polymerase are two distinct active sites. One is the polymerization site that catalyzes the 5Ј to 3Ј incorporation of deoxynucleotides (dNTPs) into the growing primer strand, and the other is the exonuclease site that is responsible for the removal of dNMPs in the 3Ј to 5Ј direction (4). gp45 is bound to the polymerase and encircles the DNA duplex, thereby increasing its processivity during DNA synthesis (5,6). gp44/62 aids in the assembly of the polymeraseclamp complex by positioning the clamp at the 3Ј-OH of the primertemplate junction and acting as a chaperone for the interaction between gp45 and gp43 (7,8). The primosome is made up of the helicase (gp41), primase (gp61), single-stranded DNA-binding protein (gp32), and the helicase loading protein (gp59) (9). gp41 is a hexameric protein that travels along the ssDNA in a 5Ј to 3Ј manner using the energy of ATP hydrolysis to unwind duplex DNA (10,11). gp61 is also hexameric and is responsible for synthesis of pentaribonucleotides on the lagging strand template that are used as primers for the lagging strand polymerase (12,13). gp32 coats the ssDNA that is produced by the helicase and is thought to aid in the coupling of leading and lagging strand replication (14,15). gp59 is responsible for the recruitment and proper loading of gp41 at the replication fork and appears to be involved in coordinating the assembly of the replisome (16 -18).
The assembly of the T4 replisome at a replication fork has been the subject of extensive investigation. Functional and physical interactions have been described for all components of the replisome (2,9,19,20). The polymerase physically interacts with gp32, which has important consequences for the properties of gp43, such as higher affinity for the primer-template junction and increased processivity during DNA synthesis (21). The polymerase is in turn attached to the clamp protein through a well described interaction between the C terminus of the polymerase and the subunit interfaces of the trimeric clamp protein (5,6,22). Based on functional evidence, an additional interaction between gp41 helicase and the polymerase has been proposed (23,24).
gp59 is a central component of the primosome, making protein-protein contacts with all other primosomal proteins. Protein cross-linking experiments have detected molecular contacts between gp59, gp61, gp41, and gp32 proteins (25)(26)(27). Single molecule and ensemble fluorescence resonance energy transfer experiments (FRET) have confirmed the cross-linking results for gp59-gp32 and gp59-gp41 interactions (9). Recently, we have characterized a protein-protein interaction between gp59 and the gp43 polymerase that inhibits the polymerase and exonuclease activities of gp43. Cross-linking and FRET studies indicate that a complex between these two proteins is formed at the replication fork prior to the initiation of DNA replication (9,16,27). With computational methods, a molecular model of the complex was generated that rationalizes the inhibition of the polymerase activities by gp59 where a protein segment of gp59 blocks partitioning of the DNA substrate between the polymerase and exonuclease sites of the enzyme (17). Unlocking of the polymerase is achieved through the loading of gp41 at the replication fork in the presence of ATP (28). In addition to the many protein-protein interactions involving gp59, it also binds to several types of DNA substrates, possessing the highest affinity for forked DNA structures (29). A molecular model has been proposed for the interaction between gp59 and a forked DNA substrate (30). As predicted by this model, a site-directed mutation at residue Ile 87 renders gp59 defective in regard to the binding of forked DNA (31).
In vivo, gp59 is indispensable for the loading of gp41 helicase onto a D-loop, which is formed through the action of UvsX/Y-catalyzed strand invasion of ssDNA into a homologous dsDNA template (32). T4 phage employs two strategies for the initiation of replication (33). At the early stages of replication, origin-dependent initiation dominates, using stably bound RNA transcripts called R-loops to serve as the primer and scaffold for replisome assembly (33). Subsequently, in the middle and late stages of the T4 infection, R-loop-dependent initiation ceases and the D-loopdependent mode takes over (33).
Here we present the functional effects of site-directed mutation of several possible "hot spots" of protein-protein interactions involving the helicase loader protein gp59 (Fig. 1). Of the four mutants generated, a single mutation, Tyr 122 to alanine (Y122A), has drastically altered the properties of the protein. The previously proposed gp59-gp43 interaction model places residue Tyr 122 near the interface of these two proteins. Y122A could no longer load gp41 helicase onto DNA and failed to support coupled leading and lagging strand synthesis unless extremely high concentrations of gp41 are used. Moreover, FRET-based assays indicated that Y122A does not form a specific protein-protein contact with DNA polymerase. As a consequence of this deficiency, Y122A no longer inhibits the exonuclease activity of the polymerase but still retains the ability to inhibit the polymerization activity of the polymerase, revealing two distinct modes of polymerase inhibition by gp59. We propose a further refined model that features a competition between gp41 and gp43 for an overlapping binding site on gp59 that rationalizes the loss of both the helicase loading and polymerase inhibition activities upon mutation of Tyr 122 .
Mutagenesis, Expression, and Purification of Wild-type and gp59 Mutants-Site-directed mutants were introduced into the gp59-intein fusion pet-IMPACT vector (12). Mutations were made in gp59 using the Quickchange mutagenesis method (Stratagene), and the entire gp59 open reading frame was sequenced using the dideoxy terminator method. The sequences of the mutagenic primers were as follows with the boldface underlined letters indicating the mutation site (reverse primers are the reverse complement of the forward): F111A, 5Ј-GGACG-CTTAAAGCAAATTAAAGCTAAGTTTGAAGAAGATATTCGC; Y217A, 5Ј-GAAACTGTGAAATCTTGCAAGGCTTGCTTTGCCAA-GGGTAC; Y146A, 5Ј-ATCCAAAAGTTCAATCAAGTGCTATTTTT-AAACTTCTGCAATCG; and Y122A, 5Ј-GATATTCGCAACATTTAT-GCTTTTAGTAAAAAAGTTGAAGTTTC.
Wild-type and gp59 mutants were transformed into BL21(DE3) cells and grown in 10 ml of Luria broth overnight at 37°C. The overnight cultures were diluted 100-fold into 1 liter of LB and grown at 37°C to an A 600 of 0.8. The cultures were then allowed to cool to 18°C, and protein expression was induced with 0.1 mM isopropyl 1-thio-␤-D-galactopyranoside. After 16 h of shaking, cells were collected by centrifugation at 6,000 ϫ g and resuspended in 10 ml of chitin column binding buffer containing 1/5 of a protease inhibitor pellet (Roche Applied Science). Cells were lysed using sonication, and cell debris was pelleted at 25,000 ϫ g. Cell-free extract was loaded onto 1 ml of chitin (New England Biolabs) columns and washed with 50 column volumes of chitin binding buffer. The chitin resin was resuspended in binding buffer plus 75 mM ␤-mercaptoethanol and incubated overnight at 4°C to facilitate intein-mediated cleavage. Following overnight cleavage, full-length gp59 was eluted in binding buffer, dialyzed into storage buffer, and analyzed for purity using SDS-PAGE. Protein concentrations were determined by measuring the absorbance at 280 nM using an extinction coefficient of 37,800 M Ϫ1 cm Ϫ1 .
DNA Constructs-Single-stranded M13 phage DNA (ssM13) was purified from infected Xl1-Blue cells by polyethylene glycol precipitation and phenol extraction as described (35). Double-stranded M13 phage DNA (dsM13) was purified from the infected Xl1-Blue cells using the Qiagen TM Qiaprep Spin miniprep kit according to the manufacturer's instructions. The etheno modification of the ssM13 was carried out as described (36). Briefly, ssM13 DNA (0.5 mg/ml) was incubated with 2 M chloroacetaldehyde in 20 mM potassium phosphate (pH 5.5) for 8 h. The reaction was then dialyzed against 10 mM Tris-HCl (pH 8.0) and 1 mM EDTA at 4°C for 16 h and transferred to a 55°C oven for an additional 8 h to allow for maturation of the derivatized bases. The integrity of the modified DNA (⑀M13) was confirmed by Tris acetate-EDTA (TAE)-agarose electrophoresis. The absorbance maximum and minimum were identical to published values (36), and the concentration of the ⑀M13 was calculated using an extinction coefficient of 16,268 M Ϫ1 cm Ϫ1 .
The 3540-bp nicked substrate used for rolling circle replication (pGEM_nick) was made by specifically nicking a modified pGEM vector at position 1353 with the enzyme N.BbvC IB (New England Biolabs). The backbone of gp59 protein is shown as a ribbon. The sites of mutation are indicated as labeled. The illustration was generated using the Protein Data Bank coordinates 1C1K using DeepView software.

gp59 Helicase Loader Protein-Protein Interactions
The nicking site was introduced through Quickchange TM mutagenesis of a pGEM vector containing a 530-bp insertion of phage T4 DNA corresponding to positions 114,754 -115,284 of the T4 genome. The sequence of the forward mutagenic primer (reverse primer is the reverse complement of the forward) was 5Ј-GCT TAA TCA GTG AGG CAC CTA CC^T CAG CGA TCT GTC TAT TTC GTT, where the boldface underlined letter indicates the mutation site, and^indicates the nicking position. The nicking reaction was carried for 2 h at 37°C in buffer 4 (New England Biolabs) containing 10 g of the pGEM_nick plasmid and 20 units of N.BbvC IB. Following digestion, the reaction was heated to 70°C for 20 min to completely denature the nicking enzyme. The substrate was used without further treatment. The efficiency of the nicking reaction was monitored with 1% TAE-agarose electrophoresis. As a control, pGEM vector without the nick site was treated in an identical fashion, and no change in mobility was observed using TAE electrophoresis.
The sequences of the oligonucleotide substrates used in the unwinding assays were as follows: fork lead, 5Ј-CAT CAT GCA GGA CAG TCG GAT CGC AGT CAG ATT TAC TGT GTC ATA TAG TAC GTA TTC AG; fork lag, 5Ј-TAA CGT ATT CAA GAT ACC TCG TAC TCT GTA CTG ACT GCG ATC CGA CTG TCC TGC ATG ATG; and trap, 5Ј-CTG ACT GCG ATC CGA CTG TCC TGC ATG ATG. The unwinding fork was made by mixing fork lead and fork lag DNA in equal molar amounts.
The 80-base oligonucleotide (ss80mer) used for the D-loop experiments is homologous to the coding strand of phage M13 between positions 6240 and 6320. PCR was used to amplify a 1500-bp region of M13 phage (dsM13 1500 ) using primers annealing to positions 6200 -6222 and 7229 -7250 for the forward and reverse primers, respectively.
The forked DNA substrate used in the FRET assays was identical to those used previously (17). The sequences of the primer and template strands of the Bio25/75 substrate used in the polymerase idling assays were 5Ј-Bio-CAA CAT TTC CTC CAC CTC ACC CTTЈ and 5Ј-AGG TGG AGG GAT GGT GGT AAT CCC GTG TGA GGT GGA GGT GAG TAG GTT GGA AGG GTG AGG TGG AGG AAG ATG TTG, respectively. All DNA strands were annealed by heating to 95°C for 3 min followed by slow cooling (ϳ1 h) to room temperature. Where indicated, 5Ј labeling was performed with polynucleotide kinase using [␥-32 P]ATP under standard conditions. DNA Binding of Wild-type and gp59 Mutants-Etheno-derivatized single-stranded M13 DNA (⑀M13) binding experiments were performed in 20 mM Tris acetate (pH 7.8), 5 mM magnesium acetate, 8 M ⑀M13 (nucleotides), and a gp59 concentration of 0.75 M (37). Steadystate fluorescence was recorded using an ISA FluoroMax-2 spectrofluorometer at 25°C. Protein and DNA were incubated at room temperature for 10 min prior to fluorescence readings. The excitation and emission wavelengths used were 305 and 410 nM, respectively. Slit widths of 2 nm were used for both excitation and emission. Each data point represented an average of 10 1-s acquisitions. For the NaCl titration experiments, the system was allowed to come to equilibrium for 5 min before acquisition. The fluorescence signal was corrected for protein fluorescence and dilution upon addition of NaCl. The total volume change throughout the NaCl titration was no more than 7%. To determine the IC 50 values for NaCl for the wild-type and mutant gp59s, the equilibrium fluorescence data were fit to Equation 1, where ⌬F is the change in fluorescence caused upon addition of NaCl; F 0 is the fluorescence in the absence of ligand; ⌬F max is the maximum fluorescence change at an infinite concentration of NaCl; X is the concentration of NaCl; K d is the dissociation constant; and n is the Hill coefficient.
Helicase Unwinding Assay-The helicase unwinding assays were performed in the standard replication buffer (25 mM Tris acetate (pH 7.8), 125 mM KOAc, and 10 mM Mg(OAc) 2 ) with 10 nM unwinding fork DNA, 50 nM trap ssDNA, 400 nM gp32, 5 mM ATP, 150 nM gp41, and 150 nM gp59 (monomeric concentrations). The reaction was carried out at 37°C, and various aliquots were removed at the time points indicated and quenched with an equal volume of 250 mM EDTA, 0.2% SDS, and loading buffer (50% glycerol, 1 g/ml bromphenol blue, 1 g/ml xylene cyanol FF). Reaction products were separated by 10% PAGE in TBE buffer and analyzed using a PhosphorImager.
Labeling gp43 and gp59 with Fluorophores-Labeling of WT and mutant gp59s (F111A, Y122A, Y146A, and Y217A) with CPM was carried out essentially as described previously (17). Briefly, after dialyzing against the labeling buffer (20 mM Tris-HCl (pH 7.3), 150 mM NaCl, 10% glycerol), WT and gp59 mutants were labeled with a 4-fold excess of CPM dye, and the labeling reactions were allowed to proceed in dark for 4 h at 4°C. Excess dye was removed by buffer exchange with a Centricon (Millipore) until no free dye was detectable in the filtrate. The protein concentration and the amount of dye were measured by the absorbance at 280 and 384 nm, respectively. The labeling efficiency ([dye]/[protein]) was generally between 60 and 70%. The labeled proteins were frozen in aliquots and stored at Ϫ70°C. The procedure for labeling the N terminus of gp43 with Oregon Green 488 maleimide (OG) is as described previously (17).
Steady-state FRET Experiments-Steady-state FRET experiments were carried out on an ISA FluoroMax-2 spectrofluorometer at 25°C. gp59-CPM (600 nM) was mixed with gp43-OG (400 nM) in the presence of a forked 32/64/73-mer DNA substrate (100 nM) constructed as described previously (17). An excitation wavelength of 390 nm was used. Donor (gp59-CPM) quenching and acceptor (gp43-OG) sensitization due to FRET were observed over a wavelength range between 430 and 600 nm. Slit widths were adjusted between 1 and 3 nm to keep the spectrum on scale. FRET signals between gp43 and WT/mutant gp59s were measured.
Three reactions were performed for each gp43/gp59 pair. In the first reaction, CPM-labeled gp59 was mixed with OG-labeled gp43 to measure the energy transfer between the donor and the acceptor. In the second reaction, unlabeled gp59 was mixed with OG-labeled gp43 to measure the background signal of acceptor sensitization under the experimental conditions. This spectrum was then subtracted from the one obtained in the first reaction to correct the FRET signal. In the third reaction, CPM-labeled gp59 was mixed with unlabeled gp43 to measure the amount of donor quenching by gp43 and was subtracted from the first reaction.
Wild-type Polymerase Idling-The polymerase idling reactions were performed in the standard replication buffer containing 0.5 M Bio25/ 75, 100 nM gp43, 500 nM gp44/62, 500 nM gp45, 5 mM ATP, 50 M dCTP, 50 M dTTP, 50 M dATP, and 10 Ci of [␣-32 P]dCTP in a total volume of 20 l. Because of the omission of dGTP, the holoenzyme will idle at the guanosine base preceding the stall site (turnover of dCTP). Bio25/75 was mixed with holoenzyme components and nucleotides for 15 s prior to adding either wild-type or Y122A gp59. Aliquots of 2 l were removed from the reaction mixture at various time points and quenched with an equal volume of 250 mM EDTA. 1 l from each time point was spotted on PEI-cellulose TLC plates and developed with 300 mM KP i (pH 7.0). After drying, the TLC plate was exposed to a PhosphorImager cassette for 1 h and then quantitated using the PhosphorImager. IC 50 values were deter-gp59 Helicase Loader Protein-Protein Interactions MARCH 31, 2006 • VOLUME 281 • NUMBER 13 mined by using Equation 1, with the fluorescence signal being replaced with counts/min.
Rolling Circle Replication-The rolling circle replication reactions were performed in the standard replication buffer containing 5 nM pGEM_nick, 200 nM gp43, 200 nM 44/62, 200 nM gp45, 400 nM gp41, 400 nM gp61, 100 nM gp59 (WT or mutant as indicated), 4 M gp32, 100 M each of CTP, GTP, and UTP, 2 mM ATP, 100 M dNTPs, and 10 Ci of [␣-32 P]dCTP, in a reaction volume of 45 l. The reactions were quenched at the time points indicated with an equal volume of 500 mM EDTA. Titration of WT-gp59 and Y122A was performed in a total volume of 10 l, and the reaction was quenched after 10 min. IC 50 values were determined by using Equation 1, with the fluorescence signal being replaced with counts/min. In experiments where the concentration of gp41 was varied, a total reaction volume of 5 l was used, and reactions were quenched after 7 min. The DNA products were analyzed through 0.8% alkaline-agarose gel electrophoresis (30 mM NaOH and 5 mM EDTA) for 48 h.
D-loop Initiated Replication-For the polymerase inhibition reactions, the 5Ј-32 P-labeled ss80mer (10 nM) was preincubated with 1.5 M UvsX and 0.25 M UvsY for 10 min at 37°C prior to mixing with a 3-fold molar excess of dsM13 1500 (30 nM) in the presence of 2 mM ATP, 50 M dNTPs, 10 mM creatine phosphate, and 5 units/ml of creatine kinase followed by the immediate addition of the replisomal proteins at the following concentrations: 200 nM gp43 exo Ϫ , gp44/62, and gp45; 400 nM gp41 (where indicated); 2 M gp32; and 100 nM wild-type or Y122A gp59 mutant. Aliquots were removed at the indicated time points, quenched with 500 mM EDTA, and separated on 0.8% alkaline-agarose gels for 24 h. For the exonuclease inhibition reactions, the 5Ј-32 P-labeled ss80mer (10 nM) was preincubated with 1.5 M UvsX and 0.25 M UvsY for 10 min at 37°C prior to mixing with a 3-fold molar excess of supercoiled dsM13 (30 nM) in the presence of 2 mM ATP, 50 M dNTPs, 10 mM creatine phosphate, and 5 units/ml of creatine kinase. D-loop formation was allowed to occur for 20 min before the addition of the replisomal proteins (same as above except for replacement of gp43 exo Ϫ with WT-gp43). Aliquots were removed at the indicated time points, quenched with 500 mM EDTA, and separated on 6% denaturing PAGE for 2.5 h.

RESULTS
DNA Binding Experiments-Because we are mainly concerned with gp59 protein-protein interactions that occur while bound to DNA, it is essential that the mutant proteins have the same DNA binding properties as the wild-type enzyme. ⑀M13 DNA has been extensively used to characterize the binding properties of DNA-binding proteins, including T4 phage replication proteins such as gp59 and UvsX (37,38). The binding of a protein to etheno-modified DNA is known to produce a conformation change in the etheno-modified bases, resulting in an increase in fluorescence intensity when monitored at 405 nM (35). Under conditions of low salt ionic strength, WT and all four gp59 mutants bind to ⑀M13 and cause a 2.2-2.5-fold increase in fluorescence ( Table 1). The binding of gp59 to ssDNA is highly sensitive to ionic strength, and NaCl titrations can be used as a measure of binding affinity (36). All mutants display a similar resistance to NaCl as compared with the wild-type protein (IC 50 values for NaCl vary by Ͻ1.5-fold), indicating that the mutants are not altered in their affinity for ssDNA. There is a moderate amount of cooperativity in the dissociation of WT-gp59 and mutant gp59s from ⑀DNA by NaCl (Hill coefficients from 1.4 to 2.1). Ionic strength-dependent changes in cooperativity of gp59 binding to ⑀DNA have been observed previously and are thought to reflect a transition between two different binding modes of gp59 on ssDNA (36).
Unwinding Assays-A simple helicase unwinding assay (39) was employed to measure the stimulatory effect of wild-type and gp59 mutants on gp41 helicase activity (Fig. 2). Cross-linking and FRET studies have shown that this effect is mediated through a direct contact between the two proteins at the replication fork (26). As shown, under conditions of our assay, gp41 helicase itself has an extremely slow onrate with only 10% of duplex unwound in 1 min. Inclusion of WT-gp59 increases the binding rate of gp41 by ϳ6-fold (60% of duplex unwound in 1 min). F111A and Y217A increase the unwinding rate by an extent similar to that of the WT-gp59. Y146A stimulates unwinding to a lesser extent (30% unwound in 1 min), whereas Y122A does not increase the unwinding rate as compared with gp41 alone.
Rolling Circle Replication-It has been suggested that gp43 is involved in the loading of gp41 at the replication fork (23). If so, it may be possible that the helicase loading defect in Y122A could be overcome by the inclusion of gp43 in the reaction. To test this possibility, we employed a nicked 3.6-kb circular plasmid as a substrate for rolling circle replication (Fig. 3). Upon treatment of the pGEM_nick plasmid with the N.BbvC IB enzyme a single 3Ј-OH is created at position 1353, which serves as a primer for the holoenzyme polymerase. Following the incorporation of 30 -50 nucleotides by the polymerase, a sufficient amount of ssDNA is produced, which allows the loading of the helicase onto the lagging strand (40). In this assay, nucleotide incorporation by the polymerase is used as an indicator of productive helicase loading. As shown in Fig. 3, the efficient loading of gp41 helicase onto the replication fork is achieved when gp32 and WT-gp59 are included in the reaction mixture. Consistent with the helicase unwinding assay, F111A and Y217A stimulate rolling circle replication to an extent equal to the wild-type protein. Y146A only partially stim-

gp59 Helicase Loader Protein-Protein Interactions
ulates rolling circle replication as compared with the wild-type enzyme, whereas the Y122A gp59 mutant does not support helicase-dependent DNA synthesis in the rolling circle reaction. The similarity between the results from the unwinding and replication assays indicates that the effect of gp43 on the loading of helicase at the replication fork is either minimal or completely absent. Even though Y122A is not able to function as a helicase loader, it still effectively inhibits the strand displacement activity of the holoenzyme. This is readily apparent because both the overall length and total amount of the DNA synthesis products are reduced in the reaction containing Y122A as compared with the reaction where gp59 is omitted (Fig. 3, reaction 4 versus reaction 2). The inhibition of gp43 polymerase activity by Y122A clearly demonstrates that it is fully capable of binding to the DNA substrate but is unable to aid in efficient loading of the helicase.
We also explored the ability of the gp59 mutants to inhibit the strand displacement activity of the holoenzyme polymerase in the absence of gp41 (data not shown). At a concentration of 0.1 M, wild-type and all four gp59 mutants inhibit strand displacement activity to a similar degree (5-fold decrease in total DNA synthesis). A titration of WT-gp59 and Y122A gave IC 50 values of 0.46 Ϯ 0.039 and 0.398 Ϯ 0.032 M, respectively. In both cases the maximal amount of inhibition was ϳ95% at the highest concentration of protein tested (2 M).
Next we tested if high concentrations of gp41 would rescue holoenzyme that was stalled by the presence of Y122A at the replication fork (Fig. 4). Because of the assay conditions, even elevated levels of gp41 do not lead to gp41-dependent synthesis in the absence of gp59. Upon addition of WT-gp59, a sharp increase in DNA product formation is observed between 0.3 and 0.9 M gp41, which is indicative of helicasedependent synthesis. In the presence of Y122A, the concentration of gp41 required to observe helicase-dependent synthesis is increased by 3-4-fold.
Steady-state FRET Experiments-Steady-state FRET was used to characterize the ability of the mutant gp59s to interact with gp32 and gp43. As shown in Fig. 5A, the interaction of WT-gp59 with gp32 results in the expected amount of FRET between the labeled gp59 and gp32. All four labeled gp59 mutants display a FRET signal with gp32 (Table 2), indicating the ability to interact with gp32 has not been affected by the mutations. The cause of the increased FRET signal displayed by F111A is unclear. Because all function assays carried out on F111A suggest that it is identical to the wild-type enzyme, it may be due to a change in the environment of the CPM fluorophore and not its interaction with gp32. Using a similar assay, F111A, Y146A, and Y217A retain their interaction with gp43 (Table 3). However, using the FRET assay as a measure, Y122A is unable to form a close association with gp43 (Table 3 and Fig. 5B).
Inhibition of gp43 Polymerase by WT-gp59 and Y122A-In order to further characterize the interaction between gp59 and gp43, the ability of WT-gp59 and Y122A to inhibit the idling reaction of the polymerase was explored (Fig. 6A). The template strand of Bio25/75 is made up of only A, T, G, except for 3 Cs located at positions 50 -52. Upon the omission of dGTP, the wild-type enzyme undergoes repeated cycles of incorporation of dCTP and excision of dCMP at the base prior to the stall site. The k off of the polymerase holoenzyme that is stalled by nucleotide omission is 0.005 s Ϫ1 (see Ref. 41; data not shown). At the time of gp59 addition to the reaction (15 s after mixing of holoenzyme with DNA), 100% of the holoenzyme has assembled and replicated up to the stall site, and because of the extremely slow dissociation of the polymerase, it remains bound to the DNA template for the duration of the assay (1 min). As shown, wild-type gp59 potently inhibits the WT-gp43 idling reaction (IC 50 ϭ 0.45 Ϯ 0.06 M, n ϭ 1.3 Ϯ 0.2); however, Y122A has no effect over the concentration range tested.
D-loops are formed by the invasion of single-stranded DNA into homologous dsDNA templates and serve as the primer for the polymerase (32). gp59 is absolutely essential for loading gp41 helicase at D-loop and is thought to play a role in preventing the bubble migration synthesis that occurs in the presence of Dda helicase (42). We tested the ability of WT-gp59 and the Y122A to inhibit the polymerization activity of gp43 at a D-loop structure (Fig. 6B). D-loop formation was catalyzed through a combination of the T4 recombination proteins UvsX and UvsY. A linear dsDNA template was used, eliminating the requirement for topoisomerase activity. In the absence of gp59 and gp41, a significant amount of replication occurs, which is produced by a strand displace-   MARCH 31, 2006 • VOLUME 281 • NUMBER 13 ment reaction performed by the holoenzyme polymerase and gp32 without the aid of gp41 helicase. Upon inclusion of WT-gp59 or Y122A the strand displacement reaction is significantly inhibited, indicating that the D-loop binding properties of Y122A have not been affected by the mutation. Addition of gp41 to the WT-gp59 reaction results in a large increase DNA synthesis, demonstrating the ability of WT-gp59 to load gp41 onto a D-loop. However, much less DNA synthesis is observed when gp41 is added to the reaction containing Y122A, consistent with the gp41 loading defect seen in Figs. 2 and 3.

gp59 Helicase Loader Protein-Protein Interactions
The exonuclease activity of wild-type gp43 polymerase can remove D-loops (Fig. 6C). In this case, the homologous dsDNA template was supercoiled dsM13, which stabilizes the D-loop after strand invasion. The 20-min incubation prior to the addition of holoenzyme ensured that the strand invasion reaction was complete. Wild-type gp43 holoenzyme alone cannot perform strand displacement synthesis when initiating from a D-loop; instead the D-loop is removed by repeated cycles of nucleotide excision and fork regression. In this situation, presumably, the polymerization activity of gp43 is overcome by its exonuclease activity, resulting in the removal of a single nucleotide, which is the first half of the polymerase idling reaction. Next, rather than the incorporation of a nucleotide, the DNA strands reanneal causing the polymerase to remain at its current position. The following nucleotide excision by gp43 results in another round of fork regression. Upon repeated cycles of this process, the net result is the removal of the D-loop. Inclusion of WT-gp59 in the reaction mixture increases the lifetime of the D-loop by inhibiting the exonuclease activity of the polymerase (Fig. 6C, reaction 2). The Y122A mutant, however, does not bring the same amount of stability to the D-loop (Fig. 6C, reaction 3).

DISCUSSION
Following the first few rounds of origin-initiated replication, the majority of DNA replication in T4 phage is initiated through strand invasion of ssDNA into homologous dsDNA templates catalyzed through the combined action of UvsX, UvsY, and gp32 (43). The result of this strand invasion is a DNA structure termed a D-loop, which serves a dual role as a primer for DNA synthesis and a scaffold for the assembly of the replisome. This strand invasion mechanism is also thought to play a major role in recombination-mediated DNA repair pathways such as double strand break repair and single-stranded lesion bypass (44). In vivo, the helicase loader protein gp59 is absolutely required for initiation of replication at D-loops. This is supported by in vitro studies showing that gp59 abrogates bubble migration synthesis and promotes the semiconservative type of replication that is the hallmark of normal DNA synthesis (40). These results suggest a crucial role of gp59 during replisome assembly.
Based on recent studies, gp59 appears to be a central player in the T4 replisome, making contacts with components of the leading strand holoenzyme, primosome, and the DNA substrate (9,17,29). For the most part, the functional consequences of these interactions are speculative. We have characterized a series of mutations with the goal of elucidating the effects of the loss of one or more of these protein-protein interactions. Guided by the crystal structure, we chose conserved hydrophobic residues that have a high degree of solvent exposure. Residues with these properties are thought to be hot spots of protein-protein contacts (45). Site-directed mutations were introduced separately into each of the four positions chosen: F111A, Y122A, Y146A, and Y217A. Each mutant was tested for its helicase loading activity, ssDNA binding, gp32 ssDNA-binding protein interaction, polymerase interac-

gp59 Helicase Loader Protein-Protein Interactions
tion, and inhibition of strand displacement synthesis by the polymerase holoenzyme.
Two mutants, F111A and Y217A, showed no altered functionality in the assays used and were not subjected to a more extensive characterization. Also, recently Jones et al. (31) demonstrated that mutation of these two residues did not affect the affinity of gp59 for forked DNA substrates. Similarly, Y146A displayed only minor changes in helicase loading activity and was fully active in all other assays. For these reasons, the majority of this Discussion will focus on the properties of the Tyr 122 3 Ala gp59 mutant (Y122A). When standard conditions are employed, Y122A is completely defective in helicase loading activity. This deficiency cannot be attributed to a lack of DNA binding capability because Y122A binds ⑀M13 DNA as well as the WT-gp59 and displays a similar resistance to increased NaCl concentrations. It is noteworthy that much less DNA synthesis is observed in a reaction containing gp41 and Y122A as compared with a reaction with gp41 helicase alone. This is likely due to the inhibition of strand displacement activity that is observed in reactions containing WT-gp59 and strongly suggests that Y122A is able to bind to the replication fork, but cannot serve as a scaffold for the productive loading of gp41 helicase. Additionally, when gp41 concentrations were elevated, Y122A supported a moderate amount of helicase-dependent DNA synthesis consistent with the interaction between Y122A and gp41 having been altered and not the interaction between gp59 and the DNA substrate.
We have employed two different FRET-based assays to characterize interactions between gp59 and other components of the T4 replisome (9,17). These assays were used to test the mutants with respect to gp32 and gp43 interactions. All mutants interacted with gp32 normally, both on and off DNA. This result is consistent with the structural model for gp59-gp32 interaction based on cross-linking data (25). The residues chosen for mutation lie on the opposite face of gp59 from where gp32 interacts. With the exception of Y122A, the mutants displayed similar levels of energy transfer with gp43 as compared with the wild-type protein. Y122A showed very little FRET signal with labeled gp43, demonstrating that the interaction between these two proteins has been disrupted by the introduction of the mutation at position 122.
The disruption of the gp43-Y122A interaction was unexpected, because Y122A retains the ability to inhibit the strand displacement activity of gp43 using the nicked circular substrate. It is likely that strand displacement inhibition only requires that gp59 be capable of binding to the forked DNA, indicating that the mode of inhibition is steric to impede the forward movement of the polymerase holoenzyme. What then is the purpose of a direct protein-protein interaction between gp59 and gp43? To address this question we tested the ability of Y122A to inhibit a holoenzyme polymerase that is idling because of nucleotide omission. The order of addition was controlled to ensure that all substrate was saturated by the holoenzyme before addition of WT-gp59 or Y122A. Under this condition, gp59 would only inhibit a polymerase activity through a direct protein-protein interaction and not through any other mechanism such as competition for the primer-template junction or sequestering of gp43 from the DNA substrate. As expected, WT-gp59 potently inhibits idling of the polymerase holoenzyme; however, Y122A has no effect at concentrations up to 1.2 M.
We further tested the inhibition of polymerase activity by Y122A at a D-loop, which is the in vivo substrate for gp59 helicase loading. The in vitro system for studying D-loop-dependent initiation of replication enabled us to study the effect of WT-gp59 and Y122A on the exonuclease activity of gp43(exo ϩ ) and the polymerization activity of gp43(exo Ϫ ) using the similar substrates. The results mirror that of the experiments using other templates. Wt-gp59 inhibits both the exonuclease and polymerase activities of gp43 at the D-loop. Although the Y122A mutant inhibits the polymerization activity of the holoenzyme, it is unable to prevent the removal of the D-loop by the exonuclease activity. This finding, in conjunction with the result from the idling experiment, strongly suggests that a direct interaction between gp59 and gp43 is necessary to inhibit the exonuclease activity of gp43. However, binding of gp59 to the replication fork alone will result in the inhibition of the strand displacement activity of the holoenzyme.
The failure of the Y122A mutant to interact with the polymerase is consistent with the proposed interaction model for these two proteins (17). This model was generated through the use of an online docking program, Cluspro, using the Protein Data Bank files for gp59 and gp43.

gp59 Helicase Loader Protein-Protein Interactions
Several preliminary models were generated, and the most favorable model was chosen on the basis of cross-linking and FRET data. On the other hand, the mutations in this paper were chosen on the basis of hydrophobicity and solvent exposure and not the interaction model. Tyr 122 of gp59 is located directly between the two regions (helix H7 and loop H6-H7) that were predicted to interact with gp43. On the basis of this central location, the role of Tyr 122 may be to stabilize helix H13 and the loop H6-H7 in their interactions with gp43. Indeed, helix H7, which contains Tyr 122 , makes contacts with both helix H6 and H13. Because the Y122A mutant is defective in polymerase binding and Tyr 122 is positioned very close to the predicted interface between gp43 and gp59, it represents an independent confirmation of the interaction model.
The fact that binding of gp59 to the replication fork is the only requirement for inhibition of the polymerase activity of gp43, yet both proteins have co-evolved to maintain the gp59-gp43 interaction, suggests that the inhibition of the exonuclease activity of the polymerase is of critical importance. gp59 inhibition of D-loop removal by the polymerase may be required for efficient origin-independent initiation of replication. In the absence of gp59, loading of gp41 onto fork structures is relatively slow (on the order of minutes). This allows enough time for the highly active exonuclease activity of gp43 to remove either a portion of or the entire invading strand of the D-loop. The invading strand of the D-loop used in this study was 80 bp and could be completely degraded within 1 min. Based on recombination frequencies, the invading strand in vivo can vary between 30 and several hundred base pairs (46). It seems quite plausible that the invading strand of smaller D-loops could be completely removed prior to gp41 loading unless gp59 binds to the D-loop and locks the polymerase in place until the helicase can load.
It is quite surprising that the mutation of a single residue on gp59 eliminates interactions with both the polymerase and the helicase. This result could be rationalized if the mutation altered the overall structure of the protein. However, this seems unlikely because several other properties such as DNA binding and the interaction with gp32 have not been affected by the mutation. It is plausible that the polymerase and the helicase share overlapping binding sites on gp59. If this is the case, how then does gp59 break its interaction with the polymerase to interact with gp41? We propose two possible models for polymerase unlocking and helicase loading that are based on a direct competition between gp41 and gp43 for binding gp59. In the first model (Fig. 7A), a hexameric gp59 is proposed to be the functional form. This is based on the crosslinking studies indicating that gp59 has the ability to form higher order structures on DNA or in the presence of gp41 and gp32. Moreover, gp59 stimulates the unwinding activity of the hexameric gp41 maximally at a 1:1 gp59:gp41 ratio. In this scenario, a single subunit of a hexameric gp59 is involved in a protein-protein interaction with the polymerase. This type of arrangement leaves the other five subunits exposed and gp59 Helicase Loader Protein-Protein Interactions available for helicase binding. The binding of helicase subunits to the gp59 monomers not involved in the polymerase interaction induces a conformation change throughout the gp59 hexamer, which breaks the interaction with the polymerase and exposes the last gp59 monomer to gp41. Loading of the final gp41 monomer triggers ATP hydrolysis and causes gp59 to dissociate from the fork, which initiates helicase-dependent DNA synthesis.
The second model is similar in many respects to the first; however, it can accommodate either a gp59 monomer or hexamer bound to the replication fork (Fig. 7B). In this model, a portion of the gp41-binding site on gp59 remains exposed even while gp59 is bound to gp43. The binding of gp41 to the ternary complex of DNA-gp59-gp43 causes a conformational change in gp59 that releases gp43 and creates a larger binding interface, which includes residue 122. Following the conformational change by gp59, gp41 repositions itself and achieves the final state that is competent for ATP hydrolysis, gp59 dissociation, and DNA unwinding. In this scenario, the gp41 can still bind to Y122A at the replication fork but its active, final form is not achieved because of the absence of the Tyr 122 interaction.
The role of gp32 was not explicitly mentioned in either of the above models. However, it is clear that the ssDNA is coated with gp32 prior to the binding of gp59. Morrical et al. (47) has demonstrated that gp32 and gp59 can simultaneously co-occupy ssDNA and that a ternary complex between ssDNA-gp32-gp59 is a required intermediate in the gp59-assisted loading of gp41. Furthermore, cross-linking and FRET experiments indicate that the stoichiometry between gp59 and gp32 is 1:1 on ssDNA (9,24). Based on these and other results, Morrical and co-workers (14) have proposed that the complex of gp59-gp32-ssDNA forms a unique topological structure referred to as a helicase loading complex. The hexameric conformational change mechanism (Fig. 7A) is best suited for accommodating the proposed helicase loading complex structure. Recently, Nossal and co-workers (49) have confirmed the requirement of gp32 for efficient helicase-dependent DNA synthesis when gp59 is used to assist in the loading of the helicase. The need for gp32 in gp59-assisted helicase loading is specific to bona fide replication reactions and is not required for gp59-assisted helicase loading when gp43 is absent and DNA replication is not used as a reporter for helicase loading (38) (Fig. 2). Based on this result, the authors proposed that one of the roles of gp32 at the replication fork is to move gp59 away from the polymerase during the act of gp41 loading. In light of our current results, we suggest that gp32 is required to facilitate the conformational change in gp59 that is required for both the loading of gp41 in the presence of gp43 and the establishment of a functional interaction between gp43 and gp41.
Several details of the mechanism for polymerase unlocking remain to be determined. The loading of gp41 onto the lagging strand template requires that gp41 binds to the replication fork as individual subunits or that the hexameric form is capable of breaking at least one subunit interface before encircling the DNA. The oligomeric structure of gp41 prior to loading onto the replication fork is unclear. Because ATP and GTP are known to stabilize the hexameric structure of gp41 off DNA, it is thought that majority of the gp41 in solution is hexameric (10). The interaction between gp59 and gp43 would create asymmetry in the putative gp59 hexamer, which would force a hexameric gp41 to lose its symmetry upon binding to gp59. This loss in symmetry may lead to the breaking of a single subunit interface allowing the hexamer to encircle the ssDNA template upon unlocking of gp43. Recent electron microscopy experiments performed on gp41 with ssDNA have revealed both "open" and "closed" forms of the gp41 hexamer (50). It is quite possible that the open form binds to the gp59-gp43-DNA ternary complex, and upon ATP hydrolysis and unwinding of DNA, the closed form is stabilized. The precise details of how and when gp59 dissociates from the replication fork are not clear. Single molecule experiments have demonstrated that gp59 leaves the replication fork upon binding of active helicase in the presence of ATP (28). However, electron microscopy experiments have shown that in some cases gp59 is retained as a complex of gp32-gp59 in the lagging strand loop in some replicating molecules (48). Clearly, further investigation into the mechanism of polymerase unlocking and helicase loading is required, and the two mechanisms presented here will serve as good working models for additional experimentation.