The Clamp Loader Assembles the β Clamp onto Either a 3′ or 5′ Primer Terminus

Clamp loaders assemble sliding clamps onto 3′ primed sites for DNA polymerases. Clamp loaders are thought to be specific for a 3′ primed site, and unable to bind a 5′ site. We demonstrate here that the Escherichia coli γ complex clamp loader can load the β clamp onto a 5′ primed site, although with at least 20-fold reduced efficiency relative to loading at a 3′ primed site. Preferential clamp loading at a 3′ site does not appear to be due to DNA binding, as the clamp loader forms an avid complex with β at a 5′ site. Preferential loading at a 3′ versus a 5′ site occurs at the ATP hydrolysis step, needed to close the ring around DNA. We also address DNA structural features that are recognized for preferential loading at a 3′ site. Although the single-stranded template strand extends in opposite directions from 3′ and 5′ primed sites, thus making it a favorite candidate for distinguishing between 3′ and 5′ sites, the single-strand polarity at a primed template junction does not determine 3′ site selection for clamp loading. Instead, we find that clamp loader recognition of a 3′ site lies in the duplex portion of the primed site, not the single-strand portion. We present evidence that the β clamp facilitates its own loading specificity for a 3′ primed site. Implications to eukaryotic clamp loader complexes are proposed.

Chromosomal replicases are tethered to DNA by a circular ring-shaped clamp (1). The clamp encircles DNA and slides along the duplex behind the polymerase, holding it to the primed site for highly processive DNA synthesis. Sliding clamps are assembled onto primed sites by a clamp loader machine that harnesses the energy of ATP hydrolysis to open and close the sliding clamp around DNA (2). Bacterial sliding clamps are homodimers called ␤, while eukaryotic and archaeal sliding clamps are homotrimers referred to as PCNA.
Structural studies of clamp loader complexes show that they are circular AAAϩ heteropentamers (2)(3)(4). The N-terminal domains contain the AAAϩ region of homology and are arranged in a spiral to form a central chamber that binds double-stranded (ds) DNA. The C-terminal domains form a closed circular cap above the DNA binding chamber. This results in a screw cap architecture that underlies the structure-specific DNA binding of a primed site. Specifically, DNA cannot pene-trate the closed cap, and must make a sharp bend to exit the central chamber through the side of the clamp loader, where there is a gap between the AAAϩ domains of two subunits. Duplex DNA is far too stiff to make this sharp bend, but the single-stranded (ss) 2 DNA portion of a primed site has the required flexibility to accommodate this sharp bend.
The minimal ␥ complex clamp loader of Escherichia coli contains three ␥ subunits and one subunit each of ␦ and ␦Ј. The complete ␥ complex contains two additional small subunits, and , which are not essential to clamp loading but stimulate the reaction. The chromosomal replicase, Pol III*, contains the heterotrimeric Pol III core (␣, DNA polymerase; ⑀ 3Ј-5Ј exonuclease and ) attached to one clamp loader assembly. Attachment of Pol III core to the clamp loader is mediated by the subunit (71.1 kDa), which is encoded by the same dnaX gene as ␥. The subunit is the fulllength product of dnaX while ␥ (47.5 kDa) is truncated at the C terminus by a translational frameshift (5). Two to three subunits replace ␥ in the clamp loader within Pol III*, and each subunit binds to one molecule of Pol III core via the C-terminal sequence unique to (6,7). In the current report, the term "␥ complex" refers to the form of the clamp loader that contains three ␥ subunits and no subunit (i.e. ␥ 3 ␦␦').
ATP binding to the ␥ subunits promotes a conformational change that enables the clamp loader to bind and open the ␤ clamp (8,9). The opened clamp adopts a spiral "lock washer" shape that complements the spiral arrangement of the clamploading subunits (10). DNA binding triggers ATP hydrolysis, which ejects the clamp loader and allows the clamp to close around DNA (9,11,12). Ejection of the clamp loader is important because Pol III, attaches to the same site on the clamp to which the clamp loader binds (13,14).
Clamp-loading function appears specific to 3Ј primer template (P/T) junctions (15). It is hypothesized that the clamp loader binds to a P/T junction and recognizes the minor groove to position the template ssDNA of a 3Ј P/T junction adjacent to the exit channel from the central chamber (see Fig. 7A) (2). In this hypothesis, minor groove recognition is used to distinguish between 3Ј and 5Ј primed sites since the ssDNA at a 5Ј site would be positioned at the back wall of the central chamber, far from the exit channel. This predicts that a 5Ј P/T junction will not fit into the central chamber.
Surprisingly, the recent crystal structure of ␥ complex bound to a primed site reveals that the clamp loader does not recognize the minor groove, but instead only forms contacts to the template strand (16). In fact, superimposition of a 5Ј P/T junction into the structure indicates that it forms essentially the same contacts to the clamp loader as a 3Ј P/T junction (e.g. see Fig.  7A). Therefore, the basis by which the clamp loader distinguishes between 3Ј P/T and 5Ј P/T structures is not clear. Considering that the template ssDNA extends outward 3Ј-5Ј at a 3Ј P/T junction, and extends 5Ј-3Ј from a 5Ј P/T site, we have proposed that the clamp loader may distinguish 3Ј P/T and 5Ј P/T junctions by reading the direction of the template ssDNA (16).
The current study examines the underlying basis for specificity of clamp loading at a 3Ј P/T versus a 5Ј P/T in the E. coli system. We find that the ␥ complex binds tightly to a 5Ј P/T structure in the presence of the ␤ clamp and ATP, and therefore DNA binding does not underlie discrimination of a 3Ј versus a 5Ј primed site. Surprisingly, we even observe clamp loading at a 5Ј P/T junction, although the efficiency of this process is at least 20-fold reduced compared with loading at a 3Ј P/T junction. The rate-limiting step in 5Ј loading appears to be the ATP hydrolysis step, as the clamp loader-clamp complex at a 5Ј P/T junction is stabile, yet is very slow to hydrolyze ATP. Thus, preference for 3Ј loading is derived from a decreased rate in coupling ATP hydrolysis to clamp closure at a 5Ј P/T site.
We also examined the DNA structural basis for preferential loading at a 3Ј site. The most obvious difference in the structures of 3Ј and 5Ј P/T junctions is the polarity of the template ssDNA. However, we find that the polarity of template ssDNA is not a source of specificity in distinguishing a 3Ј P/T from a 5Ј P/T junction. Therefore, some feature of the duplex portion of a primed site is recognized to discriminate a 3Ј P/T from a 5Ј P/T junction. Evidence in support of duplex DNA recognition for 3Ј versus 5Ј loading is gained from use of ␤ clamp mutants, which implicate direct contacts between the clamp and DNA in distinguishing 3Ј from 5Ј sites. We propose that the ␤ clamp itself facilitates the productive orientation of a 3Ј P/T junction in the clamp loader.
Construction of a Fused Single-chain ␤ Dimer-Two copies of the dnaN gene were cloned in tandem, separated by a linker encoding 14 residues, (SG) 7 , in a pET11-based vector that incorporates a N-terminal tag containing ten histidines and a protein kinase sequence (MG(H) 10 SSGHIEGRHILRRASVSSGHIEGRH). Mutations in the hydrophobic pocket binding site of ␤ (H175S, F278A, Y323S, substitution of the C-terminal 5 residues to AGGGG) were introduced into subunit A (i.e. N-terminal protomer) of the wild-type (wt) single-chain ␤ construct. Expression plasmids encoding the single-chain ␤ and mutant singlechain ␤ were transformed into E. coli BL21(DE3) and grown in LB to OD 0.6 at 37°C, then lowered to 20°C and induced upon adding 1 mM isopropyl-1-thio-␤-D-galactopyranoside followed by incubation at 20°C for 10 h. Cells were lysed and singlechain ␤ was purified by nickel chelate chromotography.
Bead-based Clamp Loading and Clamp-Clamp Loader DNA Binding Assays-To prepare DNA immobilized to beads, streptavidin-coated magnetic beads were washed four times with binding buffer and then biotin end-labeled DNA was incubated with beads in binding buffer using 150 pmol of DNA/mg beads in a 50-l volume for 1.5 h at room temperature on a rotator. Afterward, beads were washed three times with binding buffer to remove unbound DNA and any excess oligonucleotide primer used to form the P/T substrate, and then two more times with a buffer for the subsequent reaction. Under these conditions, about 130 pmol of a primer/template of 30-mer/75-mer is immobilized per mg of beads.
Clamp-loading reactions were performed in 30 l of clamploading buffer containing 13 pmol of (430 nM) DNA, 12 pmol of (400 nM) 32 P-␤ 2 , 20 pmol of (660 nM) SSB 4 , 0.5 mM ATP, and the indicated amount of ␥ complex (or Pol III* ED, ⑀ 3Ј-5Ј exonuclease mutant). A mixture of ␤ 2 , SSB 4 , and ATP in 20 l of clamp-loading buffer was added to DNA-conjugated beads on ice, then 10 l of the indicated amounts of ␥ complex and 21 mM MgCl 2 (final concentration 7 mM) were added, followed by vigorous pipetting and placing the reaction at 37°C. At the times indicated, reactions were quenched upon adding 20 volumes of ice-cold Quench/Wash buffer. 32 P-␤ 2 loaded onto DNA was quantitated upon separating the beads on a magnet for 1 min on ice and removing the Quench/Wash buffer. Beads were then resuspended in 0.5% SDS and 20 mM EDTA, and incubated at 95°C for 5 min to release 32 P-␤ 2 from the DNAbead conjugate. 32 P-␤ 2 was quantitated from its known specific activity by liquid scintillation counting.
Stabile binding of the clamp-clamp loader complex at a 3Ј or 5Ј P/T junction was performed in a similar fashion as described above, except reactions were quenched after 20 s at 37°C by dilution upon adding a 25-fold volume of ice-cold clamp-loading buffer. 32 P-␤ 2 bound to DNA was recovered from the magnetic beads and quantitated as described above.
Stability of 5Ј Complexes-The stability of ␥ complex-␤ at a 5Ј P/T junction was determined by a dilution protocol. The initial clamp-loading reaction was performed with 5Ј P/T DNA in presence of 0.5 mM ATP or ATP␥S for 20 s at 37°C as described in above for clamp-loading assays. Then, reactions were diluted 25-fold using prewarmed clamp-loading buffer. Bead-DNA conjugates were collected on a magnetic separator at 37°C, aliquots were removed at the indicated times, and 32 P-␤ 2 bound to bead-DNA conjugates was quantitated by liquid scintillation.
ATPase Assays-ATPase assays contained 13 pmol of (430 nM) DNA, ␤ 2 (as indicated), 20 pmol of SSB 4 (660 nM), 2 pmol of (67 nM) ␥ complex, 2 nmol of (67 M) [␥-32 P]ATP, and 7 mM MgCl 2 in 30 l of clamp-loading buffer on ice, unless indicated otherwise in the figure legends. After finally adding ␥ complex, the reaction was shifted to 37°C. After 1 min, the reaction was quenched by adding 70 l of 1% SDS and 40 mM EDTA. 1.5 l of each quenched aliquot was spotted on a polyethyleneimine cellulose TLC sheet and developed in 0.5 M LiCl 2 and 0.5 M formic acid. The TLC sheet was dried, and [␥-32 P]ATP and 32 P i were quantitated using a Phosphor-Imager (GE Healthcare).

Sliding Clamps Can Be Loaded at a 5Ј P/T Junction-To facilitate our
studies of E. coli ␥ complex clamploading specificity for a 3Ј P/T site versus a 5Ј P/T site, we constructed synthetic 3Ј and 5Ј primed template (P/T) DNAs and immobilized them to magnetic beads through a streptavidin-biotin linkage (see schemes in Fig. 1A). Assembly of the clamp onto DNA was followed using 32 P-␤. To prevent the 32 P-␤ clamp from sliding off the short linear DNA substrates, SSB was added, which binds the ssDNA and traps 32 P-␤ on the linear DNA (see supplemental Fig. S1). Clamp loading is initiated upon adding ␥ complex to reactions containing immobilized DNA, 32 P-␤, and ATP. Timed aliquots are quenched with EDTA, which rapidly chelates magnesium and stops the ATP-dependent clamp-loading reaction. The magnetic beads and the associated 32 P-␤ -DNA complex are separated from unbound 32 P-␤ using a magnetic separator. Fig. 1B shows the time course of clamp loading on a 3Ј P/T at different concentrations of ␥ complex.
Next, we examined clamp loading onto a 5Ј P/T site. Previous studies indicate that ␥ complex cannot load ␤ onto a 5Ј P/T junction (15). Hence, we were surprised to find that ␤ is loaded at a 5Ј P/T junction, albeit substantially more ␥ complex is required to load ␤ onto a 5Ј P/T compared with a 3Ј P/T junction (Fig. 1C). The requirement for more ␥ complex explains why clamp loading at a 5Ј P/T junction was not observed in our previous study. A plot of the rate of ␤ loading at different concentrations of ␥ complex reveals a 20-fold difference between 3Ј and 5Ј loading (supplemental Fig. S2).
Characterization of Clamp Loading at a 5Ј P/T Junction-Clamp loading at a 3Ј P/T junction requires ATP hydrolysis, and thus one may expect that ␤ loading onto a 5Ј P/T junction will also require ATP hydrolysis. Experiments performed in the absence of ATP give no ␤ loading ( Fig. 2A). Nor is ␤ loaded at a 5Ј P/T junction in the presence of the non-hydrolyzable ATP analogue, ATP␥S. The positive control shows that ␤ is loaded at a 5Ј P/T junction in the presence of ATP. Therefore ␤ loading at a 5Ј P/T junction requires hydrolyzable ATP, similar to clamp loading at a 3Ј P/T junction.
Next we examined the response of ␥ complex ATPase activity to a 5Ј P/T junction. The ␤ clamp is known to stimulate the DNA-dependent ATPase activity of ␥ complex in the presence of a 3Ј P/T junction (23). In Fig. 2B we compare the ␥ complex ATPase activity in the presence or absence of ␤ using either a 3Ј P/T or a 5Ј P/T. The results show that in the absence of ␤, a 3Ј FIGURE 1. The ␤ clamp can be loaded at a 5 P/T junction. A, scheme of the synthetic 3Ј P/T and 5Ј P/T junctions, and the bead-based assay. B, clamp loading at a 3Ј P/T junction was performed using ␥ complex at 47 fmol, (triangles), 94 fmol (diamonds), 188 fmol (squares), and 375 fmol (circles). C, clamp loading at a 5Ј P/T junction using ␥ complex at: 0.75 pmol (triangles), 1.5 pmol (diamonds), 3 pmol (squares), or 6 pmol (circles). Reactions (30 l) were performed in duplicate (B and C). P/T induces substantially more ATP hydrolysis compared with a 5Ј P/T junction. The presence of ␤ enhances ␥ complex ATPase activity at the 3Ј P/T junction, as expected from previous studies (23). However, ␤ does not stimulate the ␥ complex ATPase in the presence of the 5Ј P/T junction.
The ␥ Complex Forms a Stabile Complex with ␤ at a 5Ј P/T Junction-It is presumed that the specificity of ␥ complex for a 3Ј P/T junction is based in an inability to bind to a 5Ј P/T junction. However, the crystal structure of ␥ complex bound to a 3Ј P/T reveals phosphodiester backbone interactions to only one strand, the template strand. Hence, it cannot recognize the minor (or major) groove for orienting the DNA. To assess if the template strand can suffice for sensing direction, a 5Ј P/T junction was modeled into ␥ complex (16). The modeling exercise shows that the template strand phosphates are located in very similar positions for both 5Ј and 3Ј P/T junctions, regardless of the reverse polarity of the template strand at a 5Ј P/T relative to a 3Ј P/T junction (16). These findings suggest that ␥ complex may be capable of binding to a 5Ј P/T junction.
To test whether the ␥ complex can bind a 5Ј P/T junction we used the bead-based assay in the experiment of Fig. 2C. The experiments in the left panel utilize ATP␥S, which is not hydrolyzed by ␥ complex (24). ATP␥S is known to enable ␥ complex to open ␤ and position it onto DNA, but ␥ complex remains bound to the open clamp on DNA, since it cannot hydrolyze ATP␥S to eject from ␤ (8,25).
The results show that ATP␥S promotes attachment of 32 P-␤ to the 5Ј P/T junction (Fig. 2C). In fact, a similar amount of 32 P-␤ is retained on the 5Ј P/T junction as on the 3Ј P/T junction (solid bars in lanes 1 and 2 of the left panel, Fig. 2C). Hence, the ␥ complex appears fully capable of binding a 5Ј P/T structure. To confirm that the ␥ complex is present with ␤ on the DNA, we performed similar reactions, but omitted SSB. If the clamp loader is present, it should hold the ␤ clamp on the linear P/T DNA even without the SSB protein block. Experiments performed in the absence of SSB show that this is in fact the case (hatched bars in lanes 1 and 2).
It is important to note that the results using ATP␥S are only obtained when EDTA is omitted from the wash buffer. When EDTA is present in the wash, 32 P-␤ no longer remains bound to DNA, whether SSB is present or not (lanes 3 and 4 in the left panel of Fig. 2C). We presume that EDTA chelates the magnesium and results in reversal of the ATP␥S mediated ␤-␥ complex-DNA interaction. This result supports earlier studies indicating that the clamp remains in the open configuration with ATP␥S, otherwise it would close around DNA and remain stabile on the magnetic beads during this wash step (right panel). To ensure that complexes containing open clamps are not detected in the assays of this report we routinely include EDTA in the wash steps of all magnetic bead assays, unless noted otherwise.
The right panel of Fig. 2C compares experiments performed in the presence of ATP with those using ATP␥S. ATP hydrolysis promotes ejection of the clamp loader from the clamp, allowing ␤ to close around the 3Ј P/T junction. Hence, use of ATP results in the stabile attachment of 32 P-␤ on DNA even upon washing with buffer containing EDTA (lane 7). The greater amount of 32 P-␤ on a 3Ј P/T in the absence of EDTA (lane 5) may reflect an additive effect, resulting from two complexes: 1) the open ␤-␥ complex bound to DNA, and 2) the ␤ ring closed around DNA. Interestingly, hydrolyzable ATP promotes retention of 32 P-␤ at the 5Ј P/T junction in the absence of EDTA, but not when EDTA is present in the wash (compare lanes 6 and 8). This result indicates that ATP is not hydrolyzed, and that ATP binding acts like ATP␥S in promoting formation of a stabile ␤-␥ complex at the 5Ј P/T junction in the presence of Mg. This interesting result may carry biological significance because both ATP and Mg are present in vivo, and therefore this finding implies that the open ␤-␥ complex binds to 5Ј P/T junctions in living cells (see "Discussion").
To address the stability of the ␤-␥ complex bound to a 5Ј P/T junction, we performed an experiment in which clamp loading was allowed to proceed for 20 s, and then the reaction was diluted 25-fold and aliquots were analyzed to determine the amount of 32 P-␤-␥ complex remaining after different intervals of time at 37°C (Fig. 2D, left panel). The 32 P-␤ signal after 20 s of reaction with ATP before dilution shows the amount of 32 P-␤ closed around DNA as "loaded" ␤ and the additional amount of open 32 P-␤ bound to DNA as a ␤-␥ complex that can be washed away by a wash buffer containing EDTA (Fig. 2D, left panel,  compare lanes 1 and 2). A control in which the reaction was first diluted, and then ATP was added, shows no significant clamp loading after dilution (Fig. 2D, right panel; reactions were washed with buffer containing EDTA). The result confirms that the amount of 32 P-␤ retained by the beads should decrease if the clamp-clamp loader complex dissociates from the 5Ј P/T junction under the dilute conditions. The time course measurement of 32 P-␤ signal after dilution shows no significant dissociation after 20 min, supporting the conclusion that the ␤-␥ complex forms a stabile complex at a 5Ј P/T structure.
Comparison of Clamp Loading at RNA versus DNA P/T Sites-The physiological P/T structure during chromosome replication is an RNA-primed site; DNA-primed sites are predomi-nately associated with repair. The sliding clamp is used in both of these processes. A proposal emanating from the structural studies suggests that the ␥ complex evolved to bind only to the template strand as means to accommodate both RNA-DNA and DNA-DNA primed sites (the template strand is DNA in both cases). In the experiments of Fig. 3, we compare the rate of clamp loading at RNA-and DNA-primed sites using a 25 s reaction time but different concentrations of ␥ complex (panels A and B, respectively). The results clearly demonstrate that the rate of clamp loading at a 3Ј RNA-DNA P/T site is about 5-fold more efficient then loading onto a 3Ј DNA-DNA P/T site. This rate difference persists upon using another sequence (supplemental Fig. S3). In contrast, the rate of clamp loading at a 5Ј P/T junction is essentially the same for RNA-DNA and DNA-DNA P/T sites. The faster rate of clamp loading at a 3Ј RNA-DNA P/T junction results in a 3Ј versus 5Ј specificity that is about 5-fold more selective than at DNA-DNA P/T junctions.
The Clamp Loader Does Not Obtain Directional Information from Template ssDNA-The crystal structure of ␥ complex bound to a primed site shows that template ssDNA contacts several residues in the ␦ subunit (16). Furthermore, mutation of side chains in ␦ that bind template ssDNA result in less efficient clamp loading (16). Thus it is hypothesized that these interactions with ssDNA may "read" the polarity of the ssDNA template strand and provide information for discriminating a 3Ј from a 5Ј P/T junction.
To determine whether the ␥ complex distinguishes among 3Ј and 5Ј P/T junctions by recognizing the polarity of the template ssDNA we flipped the polarity of ssDNA by incorporating a reverse polarity link in the template strand at either the 3Ј or 5Ј P/T junction (Fig. 4). If ␥ complex reads the polarity of template ssDNA, it should read the 5Ј P/T junction of the reverse polarity template strand similar to how it reads a 3Ј P/T junction. In this case, the clamp loader will efficiently load 32 P-␤ on the 5Ј terminus of the reverse polarity substrate rather than the 3Ј terminus. However, the result clearly shows that the clamp loader still prefers to load ␤ at the 3Ј terminus, even though the template ssDNA has the same polarity as a 5Ј P/T junction. Sideby-side studies of loading onto normal and reverse polarity link junctions show no significant effect of these reverse polarity links on the rates of clamp loading (supplemental Fig. S4). This result also reveals that any asymmetry due to binding of SSB to ssDNA does not participate in 3Ј versus 5Ј clamp-loading preference because the orientation of SSB will be inverted at a reverse polarity junction, yet clamp loading onto 3Ј and 5Ј P/T is not affected.
The above results indicate that the clamp loader does not derive directional information from the polarity of the template ssDNA. It is possible that the clamp loader somehow recognizes the 3Ј hydroxyl group of the primer strand, unique to a 3Ј P/T junction. To test this possibility, we constructed a 3Ј dideoxy terminated P/T junction, but the results show a similar rate of 32 P-␤ loading onto 3Ј dideoxy and 3Ј hydroxyl-terminated primed sites (supplemental Fig. S5).
Polymerase III Provides Exquisite Specificity for a 3Ј Flap Junction-DNA polymerases have an intrinsic affinity for a 3Ј primer terminus, and therefore one may consider whether Pol III confers additional specificity to the clamp loader for loading ␤ at a 3Ј versus 5Ј P/T junction. We address this possibility in the experiments of Fig. 5 using Pol III* ED, which contains an ⑀ 3Ј-5Ј exonuclease mutant subunit. We could not use wt Pol III* containing an active ⑀ subunit for clamp loading on a 5Ј P/T, because the exonuclease digests the 3Ј ssDNA termini and thus destroys the 5Ј P/T substrate. The results using Pol III* ED in the clamp-loading assay shows a similar ␤ loading rate to ␥ complex for both 3Ј and 5Ј P/T junctions (Fig. 5A, compare with Fig. 3B). Since the clamp loader within Pol III* ED contains three subunits in place of the three ␥ subunits (i.e. 3 ␦␦'), the results of Fig. 5A demonstrate that a clamp loader that contains in place of ␥ does not give a significantly different result in 3Ј versus 5Ј clamp loading.
There is an instance in which one may expect Pol III* to significantly enhance the specificity of 3Ј versus 5Ј clamp loading. Consider the case in which a 3Ј ssDNA flap is produced during recombination or repair. In this situation the ⑀ proofreading 3Ј-5Ј exonuclease in Pol III core may excise the 3Ј flap to produce a flush 3Ј P/T site for efficient clamp loading. The proofreading exonuclease cannot excise 5Ј ssDNA (27). Therefore, if ssDNA flaps inhibit clamp loading, Pol III core may specifically excise the 3Ј flap and stimulate clamp loading after forming a flush 3Ј P/T junction.
The ␥ complex-DNA structure suggests that the clamp loader will not efficiently accomodate a primed site having a ssDNA flap at either a 3Ј or 5Ј P/T, consistent with DNA binding studies that examine this issue (16). In Fig. 5B, we test the effect of a 9 nucleotide ssDNA flap on clamp loading at either a 3Ј or 5Ј P/T junction. The result confirms that the flap essentially prevents clamp loading by ␥ complex at both P/T structures. Use of smaller ssDNA flaps results in detectable, but diminished clamp-loading activity, and correlate with reduced ␥ complex DNA stimulated ATPase activity (supplemental Fig.  S6). An earlier study that detected clamp loading at a forked junction was performed using a gel filtration method that required over 15 min to perform (15). These long incubation times may explain the observed level of clamp loading in the earlier study.
Next, we tested Pol III* for ability to load ␤ onto a 3Ј flap. The result demonstrates that the rate of clamp loading onto a 3Ј flap approaches over time a similar rate as observed using a flush 3Ј P/T junction (Fig. 5C, circles). Presumably, the 3Ј-5Ј exonuclease within Pol III core efficiently excises the 3Ј flap prior to loading. To test whether the 3Ј-5Ј exonuclease activity is required for clamp loading at the 3Ј flap, we compared loading at a 3Ј flap and a flush 3Ј P/T using Pol III* ED. The result shows no observable clamp loading (Fig. 5C, squares), consistent with a requirement for an active ⑀ proofreading subunit that excises the 3Ј flap to form a flush 3Ј P/T. We were unable to test Pol III* containing an active ⑀ subunit for clamp loading on the 5Ј flap P/T, because the exonuclease destroys the 5Ј P/T substrate as described above. However, we presume that the ⑀ subunit of Pol III* will not improve the rate of clamp loading at a 5Ј flap junction because the ⑀ subunit lacks activity on a 5Ј terminus.
In overview, the ⑀ 3Ј-5Ј exonuclease activity in Pol III core provides an additional degree of specificity to Pol III* for loading clamps at a 3Ј flap junction compared with a 5Ј flap junction. The ␤ Clamp Participates in Clamp-loading Directionality-It was demonstrated earlier in this report that ␤ stimulates the ATPase activity of ␥ complex at a 3Ј P/T junction, but not at a 5Ј P/T junction (Fig. 2B). This result indicates that the ␤ clamp may participate in specifying loading at a 3Ј P/T junction.
Indeed, the ␤-DNA crystal structure shows interactions between ␤ and DNA at multiple points (18). The ␤ clamp binds both strands of the exiting duplex using residues Arg-24 and Gln-149, which are positioned on exposed loops. Contacts to both stands of DNA may enable the clamp to FIGURE 5. Pol III confers specificity to clamp loading at a flapped 3 P/T junction. A, clamp loading at a 3Ј P/T junction (circles) and at a 5Ј P/T junction (squares) was performed using Pol III* ED (Pol III* reconstituted using an inactive ⑀ 3Ј-5Ј exonuclease mutant subunit). Reactions were 25 s, in duplicate. The "slope" was calculated within the linear range (p Ͻ 0.05) using GraphPad Prismா software. B, clamp loading at a P/T with no flap (circles) and at a P/T with a 9 nucleotide (ntd) flap (diamonds). Duplicate reactions (30 l each) contained 7 pmol of ␤, and 1 pmol of ␥ complex for the 3Ј P/T (left panel) or 6 pmol ␥ complex for the 5Ј P/T (right panel). C, schematic illustrates Pol III* correcting the 3Ј flap using the 3Ј-5Ј exonuclease activity inherent in the ⑀ subunit of Pol III core. The ratio of clamp loading at a P/T with a flap and a P/T with no flap is shown in the plot for Pol III* (circles) and Pol III*ED (squares). The 30-l reactions contained 7 pmol of ␤, and 1 pmol of Pol III* or Pol III*ED and 3 dNTPs (dGTP, dCTP, and dTTP).
correctly align only a 3Ј P/T junction, and not a 5Ј P/T, for the ssDNA at a 3Ј site to exit the central chamber of the clamp loader.
Previous studies of ␤ have shown that replacement of Arg-24 and Gln-149 with alanines results in a clamp mutant (R24A/ Q149A) that is inefficiently loaded onto DNA (18). However, the residual activity is sufficient to measure the rates of loading onto 3Ј and 5Ј P/T junctions. The result shows that mutation of these dsDNA binding residues (Fig. 6A, left structure) affects the 3Ј loading rate considerably, but not the 5Ј loading rate (Fig.  6B). Hence, the ratio of 3Ј versus 5Ј P/T loading for the mutant ␤ is lowered to a discrimination factor of about 10-fold, as compared with about 30-fold for wild-type ␤ under the same conditions. This result supports the hypothesis that the ␤ clamp participates in clamp-loading polarity.
The ␤-DNA structure shows that ␤ binds the ssDNA portion of a P/T junction in the same hydrophobic pocket to which the clamp loader and polymerase attach to ␤ (18). We wished to examine the effect of mutations in the hydrophobic pocket of the ␤ clamp. However, mutation of the hydrophobic pocket in both protomers of ␤ inactivates the ␤ clamp. Therefore, to test whether ssDNA binding to ␤ plays a role in loading at a 3Ј versus 5Ј site, we mutated the hydrophobic pocket of only one protomer by constructing a fusion protein, in which the two protomers are connected by an amino acid linker (i.e. by cloning two dnaN genes in tandem; Fig. 6A, right structure). The "fused ␤" retains about 50% the activity of wild-type ␤. The singlechain fused ␤ clamp provides the opportunity to mutate only one protomer of the ␤ dimer, since a single gene encodes the two protomers. The hydrophobic pocket in protomer B cannot be mutated because it is needed for the ␦ subunit to bind and open the interface in the fused ␤ clamp (28). Therefore, we mutated residues in the hydrophobic pocket of protomer A and studied the rate of clamp loading at 3Ј and 5Ј P/T junctions. The result shows that clamp loading at a 3Ј P/T junction is compromised by the hydrophobic pocket mutant clamp relative to the wt fused ␤ (Fig. 6D). Moreover, clamp loading at a 5Ј P/T junction is enhanced by these mutations, lowering the discrimination ratio of 3Ј versus 5Ј loading to a factor of about 5 for the hydrophobic pocket mutant compared with about 30 for wildtype ␤ using similar assay conditions. Comparison of a time course of clamp loading at a 5Ј site using mutant ␤ clamps versus wt ␤ clearly shows that more mutant ␤ is loaded over time than wt ␤ (Fig. 6, C and E). A similar effect is seen using an RNA-primed site (supplemental Fig. S7). Consistent with the observed differences in clamp loading rates, the effect of wt ␤ on the ATPase activity of ␥ complex shows stronger discrimination between 3Ј and 5Ј P/T junctions than the hydrophobic pocket mutant ␤ (supplemental Fig. S8). In overview, these results support a role of the ␤ clamp in specifying clamp loading at a 3Ј versus a 5Ј P/T junction.

DISCUSSION
This report examines the basis for clamp-loading specificity at a 3Ј P/T junction (i.e. a primed site) relative to a 5Ј P/T junction. Both P/T junctions consist of a duplex region and a ssDNA extension, and both conceivably fit into the ␥ complex active site chamber. Whereas previous studies indicate there is no loading on a 5Ј P/T junction, this report demonstrates that such loading indeed occurs. Nevertheless, we observe at least 20-fold discrimination in clamp loading at a 3Ј site compared with a 5Ј site.
Basis for Clamp-loading Specificity at a 3Ј P/T Junction-The crystal structure of ␥ complex bound to a 3Ј P/T junction shows that the template ssDNA is positioned near the gap between ␦ and ␦Ј, facilitating exit of the template ssDNA out of the clamp loader (see Fig. 7A) (16). We previously proposed minor groove recognition as the basis for specificity in clamp loading at a 3Ј P/T junction, by orienting the DNA so that template ssDNA exits the central chamber (2). In this hypothesis, the ssDNA at a 5Ј P/T would be positioned deep within the central chamber, preventing the clamp loader from binding a 5Ј P/T site. However, the structure of ␥ complex bound to a 3Ј P/T shows that the clamp loader does not recognize the minor groove, but instead binds only the template DNA strand within both the duplex and ssDNA portions of the primed site (16). Furthermore, superimposition of the 5Ј 3 3Ј template strand of a 5Ј P/T junction, with the 3Ј 3 5Ј template strand of a 3Ј P/T junction shows that the phosphates of the template strand are in very nearly the same position regardless of their opposite polarity (middle illustration in Fig. 7A). Therefore, the structure suggests that the clamp loader may bind either a 3Ј or 5Ј P/T, and can orient either of them in a way that the template ssDNA extension is adjacent to the gap in the side of the clamp loader (see left and right illustrations in Fig. 7A). Indeed, we find here that ␥ complex forms a highly stable complex with a 5Ј P/T junction in the presence of ATP and ␤.
Perhaps the clamp loader recognizes the ssDNA extension at a primed site, and distinguishes the 3Ј 3 5Ј direction of ssDNA at a 3Ј P/T from the 5Ј 3 3Ј direction of the template strand at a 5Ј P/T site. In support of this hypothesis, the structure shows numerous contacts to the phosphodiester backbone of the template ssDNA, but whether this enables it to distinguish the directionality of the ssDNA is unknown. The current study appears to rule out recognition of the ssDNA template strand as a basis of clamp-loading polarity because clamp loading is not diminished by a reversed polarity link that reverses the direction of template ssDNA at a 3Ј P/T junction. If the clamp loader were to recognize template ssDNA directionality to discriminate a 3Ј from a 5Ј P/T junction, the rate of clamp loading at the reversed polarity 3Ј P/T junction would be severely diminished.
Likewise, reversal of the ssDNA at a 5Ј P/T junction does not "fool" the clamp loader into more efficiently loading ␤ onto the reverse polarity 5Ј P/T junction relative to an unmodified 5Ј P/T junction. Therefore, specificity of clamp loading at a 3Ј versus a 5Ј P/T junction would appear to derive from the duplex portion of a primed site.
We find that the DNA stimulated ATPase activity of the clamp loader is quite low at a 5Ј P/T junction compared with a 3Ј P/T junction. Furthermore, we demonstrate that the ␤ clamp does not stimulate the ␥ complex ATPase at a 5Ј P/T junction as it does at a 3Ј P/T junction. Without ATP hydrolysis the ␤-␥ complex remains bound to the 5Ј P/T site. Hence, inefficient 5Ј clamp loading derives from inability of a 5Ј P/T junction to trigger ATP hydrolysis, despite formation of a tight complex of the clamp loader with ␤ at the 5Ј P/T junction.
The ␤ Clamp Facilitates the Specificity of Clamp Loading-In experiments using mutant versions of the ␤ clamp we obtained unexpected evidence for participation of the clamp in specificity of 3Ј P/T loading. Binding of the clamp to the clamp loader may alter the conformation of the clamp loader and facilitate recognition of the minor groove of duplex DNA. Alternatively,  (16). The diagram to the left illustrates binding of a 3Ј P/T, and binding to a 5Ј P/T is proposed in the diagram to the right. The middle diagram shows the template strand phosphate atoms of the 3Ј P/T in the actual structure (green), and the location of phosphates in the template strand of a 5Ј P/T superimposed on the 3Ј P/T structure (red) (PDB code 3GLF). B, structure of DNA bound to ␤ (PDB code 3BEP). The two structures illustrate the different conformation of DNA bound to either the blue protomer or the green protomer, assuming the clamp is positioned on ␥ complex for opening at the indicated interface.
the clamp may recognize DNA directly and help position it into the clamp loader for specificity in clamp loading at a 3Ј versus 5Ј P/T junction.
The structure of ␤ bound to DNA shows two residues, Arg-24 and Gln-149, that bind the two strands of duplex DNA in an asymmetric fashion; only one face of DNA interacts with both Arg-24 and Gln-149 (Fig. 7B) (18). This may be the underlying basis by which ␤ participates in orienting DNA within the clamp loader for 3Ј versus 5Ј clamp loading. Study of a ␤ mutant, in which these residues are replaced with alanine, results in a 2-3-fold lower specificity in loading the mutant ␤ at a 3Ј versus a 5Ј P/T junction. This supports the hypothesis that the clamp acts directly on DNA to specify a 3Ј versus 5Ј P/T junction. Loss of specificity using the ␤ mutant is mostly due to a lower rate of clamp loading at a 3Ј P/T junction. Thus mutations in ␤ may reflect loss of necessary ␤-DNA interactions needed to discriminate 3Ј versus 5Ј primed sites.
The clamp binds to the clamp loader and to ssDNA via a hydrophobic pocket located on the C-terminal face of the clamp (18,28). Mutation of residues in the hydrophobic pocket of one protomer of the ␤ dimer reduces the specificity of clamp loading at a 3Ј versus 5Ј P/T junction. This may be the result of an altered interaction of the ␤ clamp with the clamp loader, or perhaps the residues of the pocket that are closest to the edge of the central cavity of the ␤ ring interact with DNA when the clamp is in the open configuration during the clamp-loading process.
In summary, use of ␤ mutants suggests that the ␤ clamp participates in determining clamp-loading polarity. The clamp may do so directly, by binding and positioning the DNA duplex, and/or indirectly by enabling ␥ complex to discriminate between 3Ј and 5Ј P/T junctions.
Does 5Ј P/T Clamp Loading Occur in the Cell?-This report documents that the ␤ clamp becomes loaded at a 5Ј P/T junction. Previous studies may not have detected 5Ј loading given the fact that clamp loading is reduced at least 20-fold at a 5Ј P/T relative to a 3Ј P/T (15). This may simply reflect a natural leakiness inherent in the clamp-loading mechanism. However, it also seems possible that 5Ј end loading may be physiological, and that the 5Ј-loaded ␤ clamp could serve a biological role.
Clamps that are loaded at a 5Ј P/T would not be expected to function with a DNA polymerase. However, other types of proteins may target to a 5Ј P/T junction. For example, DNA ligase functions to seal a 5Ј terminus to a 3Ј terminus, and ligase is known to bind ␤ (29). Certain proteins involved in recombinative repair are thought to target a 5Ј P/T junction, such as RecF, in which case an interaction with ␤ loaded at a 5Ј P/T terminus may increase the specificity of reaction (30,31). In fact, the amounts of ␥ complex (200 nM; 6 pmol) and ␤ (200 -400 nM) needed to observe a 5Ј loading are within the range of the intracellular concentrations of ␥ complex (200 nM) and ␤ (600 nM) (32).
This report also shows that the clamp loader forms a stabile complex at a 5Ј P/T junction with ␤ in the presence of ATP and magnesium. This interesting finding suggests that the ␤-␥ complex may bind 5Ј P/T junctions in living cells. A possible role for a protein complex bound to a 5Ј P/T could be to protect the 5Ј P/T and inhibit other proteins from acting upon it, such as nucleases, helicases, or even ligase. Further studies will be needed to determine whether clamp loader-clamp complexes perform intracellular functions by binding to 5Ј P/T junctions.
Why Does the Clamp Loader Prefer an RNA-primed Site?-Sliding clamps are utilized by many different proteins, including DNA polymerases and factors involved in DNA repair. The substrate for clamp loading during repair is a DNA-DNA hybrid, while clamp loading during replication mainly utilizes RNA-primed sites produced by primase. We show here that clamp loading at a RNA-DNA P/T junction is about 5-fold more rapid than clamp loading at a DNA-DNA P/T junction. This increased rate may reflect the fact that bacterial replication forks move at a rate of ϳ650 bp/s (33). Assuming an Okazaki fragment size of 1-2 kb, a new RNA primer will be formed every few seconds, and ␤ clamps must be efficiently loaded onto new RNA primers in order for lagging strand synthesis to keep pace with the leading strand.
The fact that ␥ complex binds only the template strand at a P/T junction may enable it to accommodate both DNA-DNA and RNA-DNA substrates, since the template strand is DNA in both structures. Indeed, modeling of an A-form RNA-DNA hybrid in place of the DNA-DNA hybrid in the ␥ complexprimed site structure indicates that an RNA-DNA hybrid fits quite well.
Implications of These Studies to Eukaryotes-The eukaryotic clamp loader, replication factor C (RFC), is a AAAϩ assembly composed of five distinct subunits arranged in a spiral similar to E. coli ␥ complex (3,4). The eukaryotic PCNA clamp has a similar structure to E. coli ␤, and is known to function with a wide variety of DNA polymerases and repair factors. Like E. coli ␥ complex, RFC loads the PCNA clamp onto 3Ј P/T junctions in an ATP dependent reaction. Unlike bacteria, eukaryotic cells contain alternate clamp loaders in which the RFC1 subunit is replaced by another protein. The Rad17 subunit (Rad24 in yeast) replaces RFC1 in an alternative clamp loader involved in the DNA damage checkpoint response (34 -36). Interestingly, Rad17-RFC loads a different type of clamp onto DNA, referred to as the 911 clamp. The 911 clamp is a heterotrimer ring involved in the DNA damage checkpoint signaling pathway, but its exact function in the pathway is not yet clear.
Studies of Rad17 RFC show that it has an opposite specificity of clamp loading from RFC and E. coli ␥ complex; it loads the 911 clamp onto a 5Ј P/T junction instead of a 3Ј P/T junction (35). The 5Ј direction of the 911 clamp suggests that the 911 clamp does not function with a DNA polymerase. In addition, the 911 clamp is not known to slide on DNA. Perhaps the 911 clamp protects a 5Ј P/T junction from the action of other proteins, or it may target a factor to a 5Ј P/T junction. It is tempting to speculate that the stabile formation of ␤-␥ complex at a 5Ј P/T junction in the bacterial system (or a 5Ј-loaded ␤ clamp) may fulfill a similar role as the eukaryotic 911 clamp at a 5Ј P/T junction.
The distinctive 5Ј P/T clamp-loading polarity of the 911 clamp is thought to be conferred by the replacement of the Rad17 for RFC1. However, the idea that the Rad17 subunit underlies the specificity of clamp loading at a 5Ј P/T junction may need to be revised in light of the current study which shows that the clamp itself participates in the specificity of clamp load-ing at a 3Ј versus a 5Ј P/T junction. Therefore, on the basis of the current study, one may predict that the 911 clamp plays a role in specifying its own assembly at a 5Ј P/T junction.