Replication Factor C Clamp Loader Subunit Arrangement within the Circular Pentamer and Its Attachment Points to Proliferating Cell Nuclear Antigen*

Replication factor C (RFC) is a heteropentameric AAA+ protein clamp loader of the proliferating cell nuclear antigen (PCNA) processivity factor. The prokaryotic homologue, γ complex, is also a heteropentamer, and structural studies show the subunits are arranged in a circle. In this report, Saccharomyces cerevisiae RFC protomers are examined for their interaction with each other and PCNA. The data lead to a model of subunit order around the circle. A characteristic of AAA+ oligomers is the use of bipartite ATP sites in which one subunit supplies a catalytic arginine residue for hydrolysis of ATP bound to the neighboring subunit. We find that the RFC(3/4) complex is a DNA-dependent ATPase, and we use this activity to determine that RFC3 supplies a catalytic arginine to the ATP site of RFC4. This information, combined with the subunit arrangement, defines the composition of the remaining ATP sites. Furthermore, the RFC(2/3) and RFC(3/4) subassemblies bind stably to PCNA, yet neither RFC2 nor RFC4 bind tightly to PCNA, indicating that RFC3 forms a strong contact point to PCNA. The RFC1 subunit also binds PCNA tightly, and we identify two hydrophobic residues in RFC1 that are important for this interaction. Therefore, at least two subunits in RFC make strong contacts with PCNA, unlike the Escherichia coli γ complex in which only one subunit makes strong contact with the β clamp. Multiple strong contact points to PCNA may reflect the extra demands of loading the PCNA trimeric ring onto DNA compared with the dimeric β ring.

Replicases of cellular chromosomes utilize a circular sliding clamp protein that encircles DNA and tethers the polymerase to the template for high processivity in DNA synthesis (1). An example of this protein class is the Escherichia coli ␤ subunit, which confers high processivity onto the chromosomal replicase, DNA polymerase III holoenzyme (2,3). The eukaryotic equivalent is the PCNA 1 ring, which has essentially the same shape and chain fold as ␤ despite lack of sequence similarity between the two (4,5). These ring-shaped proteins require an ATP-fueled multiprotein clamp loader for assembly onto primed DNA.
The eukaryotic clamp loader is the heteropentameric replication factor C (RFC). The five subunits of RFC are each different proteins, but they are homologous to one another (6,7) and are members of the AAAϩ family of ATPases (8). The recent crystal structure of E. coli ␥ complex, the prokaryotic counterpart of RFC, has facilitated detailed hypothesis regarding RFC structure and mechanism (9,10). The ␥ complex (␥ 3 ␦␦Ј) consists of a minimal core of five proteins (␥ 3 ␦␦Ј), which contain the clamp loading activity (11). The remaining two subunits, and , are involved in recruiting an RNA primed DNA site from the primase, and they bind singlestranded DNA-binding protein (SSB) to assist polymerase elongation but are not essential to the clamp loading activity of ␥ complex (12)(13)(14). Biochemical studies of ␥ complex (15)(16)(17), combined with crystal structures of ␥ 3 ␦␦Ј (10) and ␦-␤ 1 complex (18,19), reveal a highly detailed view of ␥ 3 ␦␦Ј clamp loader form and function. The five subunits of the ␥ 3 ␦␦Ј complex are arranged in a circular formation (see Fig. 1). Like RFC, the ␥ 3 , ␦, and ␦Ј subunits are members of the AAAϩ family; they share a characteristic chain-folding pattern consisting of three domains each. The main intersubunit contacts in the heteropentamer are made via the C-terminal domains, which form a tight circular collar (Fig. 1A, top view). The N-terminal domains contain the ATP binding sites, and there is a gap between the N-terminal domains of the ␦ and ␦Ј subunits (Fig. 1A, front view). Only the ␥ subunits contain ATP binding and hydrolysis activity and therefore serve as the motor of this machine; both ␦ and ␦Ј lack a consensus ATP binding sequence, and neither of them bind ATP. The ␥ and ␦Ј subunits contain an SRC motif that is highly conserved from ␥ complex to RFC (20). The arginine residue (Arg finger) within the SRC motif is positioned such that it may participate in hydrolysis of ATP bound to the neighboring subunit (Fig. 1A, back view). Hence, the Arg finger within the ␦Ј SRC motif functions with ATP bound to ␥ 1 (site 1), the ␥ 1 SRC functions with ATP bound to ␥ 2 (site 2), and the ␥ 2 SRC functions with ATP bound to ␥ 3 (site 3) (see Fig. 1B). Biochemical studies confirm that these conserved Arg residues are catalytic and suggest that ATP must first be hydrolyzed in sites 2 and/or 3 before ATP in site 1 is hydrolyzed (21).
Study of ␥ complex and its subunits shows that the ␦ subunit forms the strongest attachment to the ␤ clamp, and in fact can open the ring by itself, leading to unloading of ␤ clamps from closed circular DNA (19). The ␥ trimer can also bind ␤ and unload it, but it is feeble in these actions compared with ␦ (15). Although ␦ binds ␤ 2 tightly, the ␥ complex does not bind ␤ 2 in the absence of ATP, indicating that one or more subunits of ␥ complex block the ␦-to-␤ 2 interaction (22). ATP binding to the ␥ subunits promotes tight interaction between ␥ complex and ␤, implying that ATP binding induces a conformation change in ␥ complex that exposes ␦, and presumably ␥ subunits as well, for interaction with ␤ and opening of the ␤ 2 ring. The ␦Ј subunit appears to be a rigid protein and has been termed the stator, the stationary part of a machine upon which other pieces move (10,20). Results of mutational studies are consistent with the idea that the ATP-induced conformation change of ␥ complex requires ␦Ј and that it may serve as a "backboard" to direct the ATP-induced changes in ␥ complex (23).
The similarities in RFC and ␥ complex subunit sequences and their common function in loading circular clamps onto DNA suggest that the RFC subunits may also be arranged in a circular fashion like ␥ 3 ␦␦Ј (10). Electron microscopy and atomic force microscopy studies of RFC are consistent with a circular arrangement of RFC subunits (9,24). The human RFC1 subunit (p140), like ␦, binds to PCNA (25), leading to the proposal that RFC1 may act to open PCNA just as the ␦ wrench opens ␤ (9, 18). RFC5, like ␦Ј, contains an SRC motif, and the putative ATP site deviates from the consensus sequence (GKKT instead of GKT) suggesting that if it can bind ATP, the ATP may not hydrolyze efficiently. These similarities suggest that RFC5 may play a similar role as the ␦Ј stator. On the basis of the ␥ 3 ␦␦Ј structure, it is proposed that the RFC1 (wrench) and RFC5 (stator) subunits may bracket the RFC2, 3, and 4, subunits, which, like ␥ 3 , contain both ATP binding and SRC motifs and thus may act as the motor of the RFC clamp loader (9). However, the order of the RFC2, 3, and 4 subunits within the pentamer and the identity of the subunits that bind RFC5 and RFC1 are not certain.
The aim of this study is to define the arrangement of the subunits within the RFC complex and determine which subunits form the major contact(s) to PCNA. The results indicate the arrangement RFC5:RFC2:RFC3 RFC4:RFC1, and in an orientation looking down the C-terminal plane, RFC5 contributes an arginine finger (SRC) to ATP in RFC2 (site 1), RFC2 contributes an arginine finger to the RFC3 ATP site (site 2), RFC3 contributes an arginine finger to the ATP site in RFC4 (site 3), and RFC4 contributes an arginine finger to the RFC1 ATP site (site 4). Moreover, we find that the RFC(3/4) complex, RFC(2/3) complex, and RFC1 subunit form major contacts to PCNA, unlike the case of the single major contact between ␦ subunit of ␥ complex and the ␤ clamp.
Mutations in the RFC1 Gene-Codons TAT and TTC in pET(11a)-RFC(1) corresponding to residues Tyr-404 and Phe-405 in RFC1 were mutated to GCT and GCC, respectively, by Commonwealth Biotechnologies, changing both hydrophobic residues to alanine. The mutated RFC(1) gene was then cloned into the pLANT (2)-RFC[1ϩ5] to replace the wild-type copy of the RFC(1) gene.
Mutations in the RFC3 and RFC4 Genes-Mutations were introduced into the RFC(3) and RFC(4) genes using the QuikChange method Top view shows the circular pentameric arrangement of the subunits looking down (i.e. perpendicular to) the plane defined by the C-terminal domains. Side view shows the helix (gray) of the ␦ subunit that interacts with the ␤ clamp. The C-terminal domains of the subunits are located at the top, and the N-terminal domains are at the bottom of this view. Back view is a 180°pivot around the vertical axis of the side view. This view shows the P-loops (yellow) and arginine fingers (white) located at the interface between subunits. The elements that are labeled in the figure are those of ␥ 1 (Arg finger) and ␥ 2 (P-loop) defining ATP site 2. The same elements are also present in ATP sites 1 and 3. B, schematic diagram illustrating the polarity of the P-loops and arginine fingers forming the three ATPase sites in E. coli ␥ complex. This diagram corresponds to the top view looking down onto the C-terminal face of ␥ complex. In this view, the Arg finger of one subunit points clockwise toward the P-loop of the adjacent subunit to form a hydrolysis-competent ATP site.

Over-expression of S. cerevisiae RFC Proteins in E. coli
The pET(11a)-RFC derived expression plasmids were transformed into BL21(DE3) codon plus (Stratagene) E. coli cells, which contain a secondary plasmid encoding genes for rare transfer RNAs, as well as chloramphenicol resistance. Transformants were selected using ampicillin (100 g/ml) and chloramphenicol (25 g/ml). Co-transformations involving pLANT (2) and pET(11a)-derived plasmids were performed with BLR(DE3) (Novagen) E. coli-competent cells. These transformants were selected using ampicillin (100 g/ml) and kanamycin (50 g/ml). Fresh transformants were grown in 12-24 liters of LB media containing the appropriate antibiotics at 30°C until reaching an A 600 value of 0.6. The cultures were then briefly chilled on ice before adding 1 mM isopropyl-1-thio-␤-D-galactopyranoside and then were incubated at 15°C for ϳ18 h. For RFC(2/3) and RFC(2/5), induction was at 37°C for 3 h. The cells were harvested by centrifugation.

Purification of S. cerevisiae RFC Complexes, Subcomplexes, and Individual Subunits
Purification of S. cerevisiae RFC 1mut and RFC wt Complex-RFC wt was overexpressed from the single plasmid, pET(11a)-RFC[1ϩ2ϩ3ϩ4ϩ5]. The RFC 1mut was expressed by co-transformation of pLANT (2)-RFC[1 mut ϩ5] and pET(11a)-RFC[2ϩ3ϩ4]. The harvested cells (from 24 liters of culture) were resuspended in ϳ200 ml of Buffer A containing 1 M NaCl and then lysed using a French press at 22,000 p.s.i. The cell lysate was clarified by centrifugation, diluted with Buffer A to ϳ150 mM NaCl, and then applied to a 30-ml SP-Sepharose Fast Flow (Amersham Biosciences) column equilibrated with Buffer A containing 150 mM NaCl. The column was eluted with a 300-ml gradient of 150-600 mM NaCl in Buffer A. Presence of proteins was followed in 10% SDS-polyacrylamide gels stained with Coomassie Blue. The peak of RFC (which eluted at ϳ365 mM NaCl) was pooled and diluted with Buffer A to ϳ150 mM NaCl. The protein was then applied to a 40 ml Q-Sepharose Fast Flow (Amersham Biosciences) column equilibrated with Buffer A containing 150 mM NaCl and eluted with a 400-ml gradient of 150-600 mM NaCl in Buffer A. The fractions containing RFC complex (which eluted at ϳ300 mM NaCl) were saved individually and stored at Ϫ70°C. Protein concentration was measured by Bradford assay, and a typical final yield was ϳ1 mg purified RFC complex/1 liter of culture.
Purification of S. cerevisiae Wild-type and Mutant RFC(3/4) Subcomplexes-RFC(3/4) subcomplexes (wild-type or SAC mutants) were expressed using the pET(11a)-RFC[3ϩ4] plasmid containing either wildtype or mutant genes. The cell pellet (harvested from 12 liters of culture) was resuspended in 200 ml of Buffer A containing 800 mM NaCl, lysed with a French press, and clarified by centrifugation. The clarified cell lysate was diluted with Buffer A to ϳ180 mM NaCl before being applied to a 100-ml SP-Sepharose column and eluted with a 1-liter gradient of 150 -600 mM NaCl in Buffer A. The peak of RFC(3/4) (which eluted at about 350 mM NaCl) was pooled, diluted with Buffer A to ϳ150 mM NaCl, and then applied to a 50-ml Heparin-agarose (Amersham Biosciences) column followed by a 500-ml elution gradient of 150 -600 mM NaCl in Buffer A. The peak fractions of RFC(3/4) subcomplex eluted at ϳ300 mM NaCl and were stored at Ϫ70°C. The final yields for RFC(3/4) were ϳ4 mg of RFC(3/4) wt , 4 mg of RFC(3/4 SAC ), and 15 mg of RFC(3 SAC /4) per liter of cell culture.
Purification of S. cerevisiae RFC(2/3) Subcomplex-The RFC(2/3) protein complex was expressed by co-transformation of pLANT (2)-RFC(2) and pET(11a)-RFC(3). The clarified cell lysate was diluted to 150 mM NaCl with Buffer A and then purified over a 20-ml SP-Sepharose Fast Flow column with a 200-ml gradient of 300 -700 mM NaCl in Buffer A. The peak of RFC(2/3) was pooled and diluted with Buffer A to 100 mM NaCl before being further purified over a 6-ml Q-Sepharose Fast Flow column with a 60-ml gradient of 200 -400 mM NaCl in Buffer A. The purified protein was dialyzed against Buffer A containing 100 mM NaCl before being stored at Ϫ70°C. The yield was ϳ1 mg RFC(2/3) per liter of cell culture.
Purification of S. cerevisiae RFC(2/5) Subcomplex-The RFC(2/5) subcomplex was coexpressed upon co-transformation of pLANT (2)-RFC(2) and pET(11a)-RFC (5). The purification protocol for this subcomplex is the same as for the RFC(2/3) complex (refer to above), except that a 150 -600 mM NaCl gradient in Buffer A was used with the SP-Sepharose Fast Flow column, and a 100 -450 mM NaCl gradient in Buffer A was used with the Q-Sepharose Fast Flow column. Yield was also ϳ1 mg of RFC(2/5) per liter of cell culture.
Purification of S. cerevisiae RFC2 Protein-The individual subunit, RFC2, was expressed from the pET(11a)-RFC2 plasmid. The cells were lysed and then the lysate was clarified and diluted to 180 mM NaCl as described above for RFC complex purification. The protein was first fractionated over a 100-ml SP-Sepharose Fast Flow column with a 1-liter gradient of 150 -600 mM NaCl in Buffer A. The peak of RFC2 (which eluted at about 300 mM NaCl) was pooled and diluted with Buffer A to ϳ200 mM NaCl, before being applied to a 50-ml Q-Sepharose Fast Flow column. The flow through fraction containing RFC2 was precipitated by addition of ammonium sulfate (0.5 g/ml). After centrifugation, the pellet was resuspended in Buffer A and dialyzed against Buffer A containing 250 mM NaCl at 4°C overnight. The yield of protein was ϳ4 mg of RFC2 per liter of cell culture. Trace contaminants containing ATPase activity were separated away from the RFC(2) protein by gel filtration. Approximately, 700 g of RFC2 protein was applied to a 24-ml Superdex-75 (Amersham Biosciences) column equilibrated with Buffer A containing 200 mM NaCl. After collecting 240 drops (6.1 ml), 5-drop (120 l) fractions were collected (79 fractions). Aliquots (3 l) from every other fraction were assayed for ATPase activity (refer to protocol below), and 16-l aliquots from the same fractions were analyzed on a 10% SDS-polyacrylamide gel. ATPase activity was detected in fractions 17-27, whereas RFC2 protein was present in fractions 25-37. RFC2 free from contaminating ATPase activity (fractions 29 -37) was used in the ATPase assays reported in this study.
Purification of S. cerevisiae RFC4 Protein-Expression of RFC(3/4) subcomplex from the pET(11a)-RFC[3ϩ4] plasmid results in the production of excess soluble RFC4 protein. Clarified cell lysate was diluted to ϳ100 mM NaCl with Buffer A before being fractionated over an 18-ml SP-Sepharose Fast Flow column with a 180-ml gradient of 100 -500 mM NaCl in Buffer A. Fractions containing both RFC(3/4) and excess RFC4 were pooled and dialyzed for 2 h against Buffer A until the conductivity was approximately equal to 50 mM NaCl. The protein was then applied to a 7-ml Q-Sepharose Fast Flow column and washed with 50 ml of Buffer A. Pure RFC4 eluted in the wash. Yield was ϳ1 mg RFC4 per liter of cell culture.

ATPase Assays
Wild-type and mutant RFC(3/4) subcomplexes were tested for ATPase activity in the presence or absence of a synthetic primed template. The primed template was formed by mixing the following two oligonucleotides: 79-mer, 5Ј-GGG TAG CAT ATG CTT CCC GAA TTC ACT GGC CGT CGT TTT ACA ACG TCG TGA CTG GGA AAA CCC TGG CGT TAC CCA ACT T-3Ј; and 45-mer, 5Ј-GGG TTT TCC CAG TCA CGA CGT TGT AAA ACG ACG GCC AGT GAA TTC-3Ј. The 79-mer (800 pmols) and 45-mer (2.4 nmol) were mixed in 100 l of 5 mM Tris-HCl, 150 mM NaCl, and 15 mM sodium citrate (final pH 8.5). The mixture was brought to 95°C and then cooled to room temperature over a 30-min interval. The ATPase reactions contained 2 M RFC(3/4), 2 mM [␣-32 P]ATP, 500 nM primed template (when present), and 2 M RFC2 (when present) in a final volume of 70 l of Buffer C. Reactions were incubated at 30°C, and 10-l aliquots were removed at intervals (0 to 4 min) and quenched with an equal volume of 5% SDS/40 mM EDTA mixture. One microliter of each quenched reaction was spotted on a polyethyleneimine cellulose TLC sheet (EM Science) and developed in 1 M formic acid/0.5 M lithium chloride buffer. After the TLC sheet was dried, [␣-32 P]ATP and [␣-32 P]ADP were quantitated using a Phospho-rImager (Amersham Biosciences) and the ImageQuant software (Molecular Dynamics).

Gel Filtration Analysis
Gel filtration was performed at 4°C using a 24-ml Superdex-200 column (Amersham Biosciences) equilibrated in Buffer B. Subunit mixtures (protein concentrations indicated in the figure legends) were incubated at 16°C for 15 min in 250 l Buffer B and, when present, included 1 M ATP, 8 mM MgCl 2 , and 6.5 M PCNA (as trimer). Subunit mixtures were applied to the column, and after collecting the first 5.6 ml (void volume), fractions of 180 l were collected. Column fractions were analyzed in 7.5% SDS-polyacrylamide gels stained with Coomassie Blue. Bovine serum albumin (Sigma) (66 kDa) was added to the protein mixtures to serve as an internal molecular mass marker for each gel filtration analysis. The fraction numbers were normalized to the 66-kDa standard, which was set at peak fraction 49.

Steady-State Fluorescence Measurements
Steady-state fluorescence intensity measurements were performed using a PTI spectrofluorimeter (Photon Technology International). Fluorescence emission spectra were obtained from 500 to 600 nm using an excitation wavelength of 490 nm. The band-pass for excitation and emission was 2 and 4, respectively. All samples were in 10 mM Tris-HCl (pH 7.5), 5 mM DTT, 1 mM EDTA, 8 mM MgCl 2 , 1 mM ATP, and 150 mM NaCl, as indicated. Fluorescent measurements utilized PCNA labeled at its two exposed Cys residues (22 and 62) with Oregon Green 488 maleimide (Molecular Probes). Measurements were performed at 100 nM, 200 nM, or 1 M PCNA 3 and increasing amounts of RFC wt or RFC 1mut .

Replication Assays
The primer-template was prepared by annealing a synthetic DNA 30-mer oligonucleotide (M13mp18 map position 6815-6847) to M13mp18 ssDNA as described (2). The large subunit of Pol ␦ used in this assay was a truncated version of the full-length protein; it contained the N-terminal 807 amino acids (from a total of 1098 amino acids in the full-length Pol ␦) and an additional 14 amino acids (LRDPLIIS-PKRNHV). The gene for this subunit of Pol ␦ was cloned into a pET(11a) vector, along with the other two subunits of Pol ␦ holoenzyme. Induced cell lysate was fractionated over a Q-Sepharose Fast Flow column. The enzyme activity required the presence of both PCNA and yeast RFC and had a specific activity of ϳ1.0 fmol dNTP incorporated per g of protein per min. Assays contained 14.1 fmol of primed ssDNA, 3.55 pmol of E. coli SSB, 60 M deoxyribonucleoside triphosphates (dGTP, dCTP, dATP), and 20 M [␣-32 P]TTP in 12.5 l of Buffer E. First, 50 fmol of RFC and PCNA (ranging from 25 fmol to 2.5 pmol) were incubated together at 30°C for 5 min and then 1 l of recombinant yeast Pol ␦ was added to initiate replication. After 5 min at 30°C, the reaction was quenched with an equal volume of 40 mM EDTA/1% SDS and then spotted onto DE81 paper followed by quantitation of DNA synthesis by scintillation counting as described (2).

RESULTS
We have reported previously a compatible two-vector expression plasmid system for coexpression of multiple proteins in E. coli (26). In that study we described the expression and purification of five-subunit yeast RFC in which an unessential Nterminal region of RFC1 is deleted. In the current study we express and purify RFC in which all subunits are full-length and unmodified. The earlier study indicated that expression of individual RFC subunits provided only insoluble protein. In the current study we coexpress two subunits at a time and find that three heterodimeric RFC subcomplexes are soluble and can be purified. In fact, during preparation of an RFC(3/4) complex, isolated RFC4 is also obtained from one of the purification steps. Moreover, we find that RFC2 expression results in some RFC2 that remains soluble and can be purified as an isolated subunit. In the experiments to follow, these subunit preparations are used to examine the architecture of RFC and its interactions with PCNA.
Subunit-Subunit Connections in RFC-Based on the E. coli ␥ complex circular pentamer, the five subunits of RFC are likely arranged in a similar circular fashion. In the case of a circular heteropentamer one may expect that neighboring subunits, when expressed together in the absence of other subunits, may form heterodimers and thereby identify which subunits are next to one another. To test this approach we used two compatible T7-based expression vectors described in our earlier study (26). One vector is the commercially available pET11a plasmid, which uses a ColE1 origin and contains an ampicillin resistance gene for selection. The other is the pLANT vector, which contains a p15A origin and a kanamycin resistance gene. The pLANT vector used in this study includes genes encoding tRNAs that are prevalent in yeast but underrepresented in E. coli (codons AUA (Ile), CUA (Leu), AGG (Arg), and AGA (Arg)) and utilizes the same inducible T7 promoter and cloning sites of the pET11a vector. The two subunit combinations for the five subunits of RFC were tested by placing one gene in pLANT and another in pET. Cells carrying both plasmids were then induced, and the cell lysate was examined for the presence of soluble heterodimer. Using this approach three combinations provided soluble proteins, RFC(2/3), RFC(2/5), and RFC(3/4). These findings are supported by an earlier observation that coexpression in E. coli of RFC2 and 3 and of RFC3 and 4 appeared to increase their solubility (27). In theory, for a circular heteropentamer, there should be five combinations of two subunits that are adjacent to one another. Unfortunately, we observed no soluble complexes containing RFC1, which may account for the two complexes that could not be obtained in soluble form (i.e. RFC1 with either of its two neighboring subunits).
Next we asked whether the three apparently soluble RFC heterodimers can be purified intact, and if so, whether they remain stably associated with each other during analysis in a gel filtration sizing column. For each heterodimer, induced cells were grown and lysed, and the cell lysate was fractionated using an SP-Sepharose cation exchange column. After this step, methods for the three preparations diverged, as explained under "Experimental Procedures." However, we succeeded in obtaining each of the RFC heterodimers in pure form as illustrated in the SDS-PAGE analysis of Fig. 2 (RFC(2/3), RFC(2/5), and RFC (3/4), lanes 3, 4, and 5, respectively).
The RFC heterodimers were tested for true stable complex formation by asking whether they remain together during analysis in a gel filtration sizing column (Fig. 3). The individual RFC2 and RFC4 subunits were also examined by this approach. Individual RFC2 and RFC4 eluted in fractions 54 -58 and 56 -58, respectively, slightly later than the 66-kDa size marker, indicating that these subunits are monomeric and consistent with the presence of only one of each these proteins in the RFC complex (Fig. 3, A and B). Next the RFC(3/4) preparation was analyzed (Fig. 3C). The RFC3 and RFC4 subunits comigrated with one another and eluted earlier than either RFC2 or RFC4 alone, indicating that the RFC(3/4) sub- units indeed exist as a complex. The two subunits stain with similar intensity suggesting that the complex probably contains one of each subunit. The RFC(2/3) preparation also behaved as an equimolar complex except the complex eluted slightly earlier than RFC(3/4) (Fig. 3D). Likewise the RFC(2/5) preparation coeluted earlier than either RFC2 or RFC4 indicating that they form a stable complex, although the two subunits did not resolve well from one another in the polyacrylamide gel because of their very similar mass (Fig. 3E).
A Model for Subunit Arrangement in RFC-The fact that RFC(2/3) and RFC(3/4) form complexes indicates that RFC3 binds to both RFC2 and RFC4. If RFC is a circular pentamer in which each subunit has only two neighbors, then the results indicate a subunit order RFC2:RFC3:RFC4 where RFC3 serves as a bridge between RFC2 and RFC4. This arrangement makes two predictions: first that RFC2 will not form a complex with RFC4, and second, that RFC2 and RFC(3/4) will form a RFC2/ (3/4) heterotrimer. The test of whether RFC2 and RFC4 interact is shown in Fig. 4A. Although the two subunits appear to coelute, their elution position in fractions 54 -58 is similar to their individual elution positions (compare Fig. 3, A and B) indicating that they do not form a complex. Had they bound to one another the complex should have eluted earlier, as a larger molecular weight complex. Next, the mixture of RFC2 with RFC(3/4) was examined for ability to form an RFC2/(3/4) complex. The results, in Fig. 4B, show that the RFC2 subunit coelutes with RFC(3/4) several fractions earlier than RFC2 alone, indicating formation of the RFC2/(3/4) complex. Hence, the predictions listed above hold true and indicate that the subunit order of these three proteins is RFC2:RFC3:RFC4.
The RFC(2/5) complex, taken together with the formation of RFC(2/3) complex, suggests that RFC2 binds to both RFC5 and RFC3. Thus RFC4, which is positioned on the other side of RFC3 from RFC2, should not form a complex with RFC(2/5). The test of this prediction, in Fig. 4C, demonstrates that RFC4 does not coelute with the RFC(2/5) complex. Taken together with the formation of a RFC2/(3/4) complex described above, the subunit order of the four RFC subunits is RFC5:RFC2: RFC3:RFC4. Consistent with this assignment, when the RFC(2/5) and RFC(3/4) complexes are mixed they coelute at a much earlier position than either of the subcomplexes alone. Given the arrangement of the RFC5/2/3/4 subunits, and the presumed circular arrangement of the RFC pentamer, the RFC1 subunit probably fits between RFC5 and RFC4. We tried to test this hypothesis by co-expression of RFC1 with each of the other subunits, but in no case could we obtain a soluble complex. Nor could we obtain RFC1 alone. We also tested Nand C-terminal truncated versions of RFC1 for solubility and also constructed glutathione S-transferase and maltose-binding protein fusions of RFC1 (and various truncated RFC1 fusion subunits), but what little soluble protein was obtained behaved as an aggregate in gel filtration and did not bind the other subunit preparations. We also tried immobilizing the RFC1 fusion proteins on microtiter plates and on beads but still could not obtain reliable information pertaining to which subunit(s) it binds to. Hence, we were unable to demonstrate directly that RFC1 is adjacent to RFC5 and RFC4. However, protein-protein interaction studies of human RFC have demonstrated that RFC1 (p140) binds to both RFC4 (p40) and RFC5 (p38) subunits (28,29) thus supporting the above position assigned to RFC1.
ATP Site Polarity within the RFC Pentamer-Given the subunit arrangement described above for the RFC pentamer, it may exist in either of two orientations as illustrated in the diagrams of Fig. 4E. These schemes reflect the view looking down the C-terminal plane of the pentamer and result in two very different consequences for how the complex may utilize ATP. The location of ATP sites in AAAϩ multimeric machines are at subunit interfaces and require an Arg residue from the adjacent subunit. In the ␥ complex pentamer, this Arg finger is contained within a highly conserved SRC motif in RFC box VII (10). Only four RFC subunits contain the SRC motif, RFC2, 3, 4, and 5 (6). Also, only four subunits contain consensus Walker A motifs for ATP hydrolysis, RFC1, 2, 3, and 4. In the subunit polarity shown in Model 1 of Fig. 4E, the placement of these motifs is maximized so that each ATP binding subunit is placed clockwise next to an SRC-containing subunit. This arrangement could provide four competent ATP hydrolysis sites. The other subunit polarity, in Model 2 of Fig. 4E, produces only three competent ATP sites.
To determine the polarity of the RFC pentamer we first focused on the RFC(3/4) complex and then extended the study to the RFC2/(3/4) complex. We find that the RFC(3/4) complex is an ssDNA stimulated ATPase, as illustrated in Fig. 5A. In the experiments to follow, we used this activity to examine the polarity of these two subunits. Two mutant RFC(3/4) complexes were constructed, containing a mutated SRC motif in either RFC3 or RFC4 (R replaced by A and referred to here as SAC). The mutant complexes were then analyzed for ssDNA-dependent ATPase activity. If the subunit polarity is as shown in In Fig. 6 we examine RFC2 for ATPase activity and its affect on the RFC(3/4) complex. We find that RFC2 alone lacks ATPase activity Ϯ DNA (data not shown) but provides about 4-fold stimulation to the RFC(3/4) ATPase (see Fig. 6A). The subunit arrangement and polarity of RFC, depicted in Model 1 of Fig. 4E, makes the prediction that RFC2 contributes a catalytic Arg residue to the ATP binding site in RFC3, which may explain how RFC2 stimulates the RFC(3/4) ATPase activity. Namely, RFC(3/4) has only one complete ATP site (site 3), and RFC2 generates a second complete ATP site (site 2). In this case, the RFC(3/4 SAC ) mutant should still be stimulated by RFC2 to the same extent as wild-type RFC complex. If RFC2 were to bind RFC4, instead of RFC3, addition of RFC2 would not stimulate the activity of the resulting RFC(2/3/4 SAC ) com- RFC (3/4) and RFC(2/3) Bind PCNA-RFC, like prokaryotic ␥ complex, binds the clamp in an ATP-dependent fashion (30). ATP binding is sufficient to promote the RFC⅐PCNA complex, hydrolysis is not required. We examined the individual RFC subunits and heterodimers for ability to form a stable complex with PCNA and tested whether the interaction requires ATP. The RFC-PCNA interaction study of Fig. 7 uses gel filtration analysis, which is a non-equilibrium technique, and only the tightest protein complexes will be observed. PCNA was mixed with the various RFC subunit preparations in the presence of ATP and then the mixture was assayed for complex formation. The result shows an interaction between RFC(3/4) and PCNA, and RFC(2/3) and PCNA, but no interaction between PCNA and RFC(2/5), RFC2, or RFC4. It is interesting to note that neither RFC4 nor RFC2 stay attached to PCNA, yet PCNA stably binds to both the RFC(3/4) and RFC(2/3) complexes. Thus RFC3 may form the main contribution between PCNA and these heterodimeric complexes. This prediction that RFC3 can bind PCNA is supported by studies in the human system indicating that the p36 subunit (yeast RFC3 homolog) of human RFC binds PCNA (31). It is still possible that other RFC subunits bind PCNA but do not form a complex with sufficient

FIG. 4. Interaction between RFC subunits and subcomplexes. Panels
A-D are protein-protein interaction studies of RFC subunit mixtures analyzed by gel filtration. Approximately 1.75 nmol of each subunit or heterodimer were incubated together and then analyzed on a Superdex-200 gel filtration column. Column fraction numbers are indicated above the top gel, and RFC subunits are identified on the left of each gel. E, two possible models of RFC subunit arrangement in the RFC pentamer, clockwise or counter clockwise, and the corresponding implications for placement of the Arg finger and P-loop, which form the ATPase site at the interface between adjacent subunits. Both diagrams depict the top view looking down onto the C-terminal face of the pentamer (e.g. see Fig. 1A). stability to be isolated by this nonequilibrium technique.
Study of these same protein mixtures in the absence of ATP revealed no new complexes with PCNA, and in fact the RFC(3/ 4)⅐PCNA complex was no longer observed (Fig. 7A). Possible explanations underlying the ATP dependence of the RFC(3/ 4)⅐PCNA complex and ATP-independent RFC(2/3)⅐PCNA complex are presented under "Discussion." We lack the RFC1 subunit in isolation or as a heterodimer with another RFC subunit, but we presume it also forms a major contact to PCNA based on studies of human RFC (25,32). The next few experiments take a different approach to assess the importance of RFC1 in RFC⅐PCNA complex formation.
RFC1 Contributes to PCNA Interaction-We have no subassembly of RFC that contains RFC1. Therefore, to address whether it contributes to PCNA binding we mutated two hydrophobic residues of RFC1 (Tyr-404 and Phe-405) to alanine and expressed it, along with the other subunits, to produce RFC 1mut . These two residues are located between the P-loop (Walker A) and DEXX (Walker B) motifs and have been hypothesized to form the main contact between RFC1 and PCNA based upon the crystal structure of prokaryotic ␦⅐␤ complex (18). Like RFC1, the ␦ subunit of ␥ complex is a member of the AAAϩ family, and it binds to ␤ via two hydrophobic residues that are located in the same relative position in ␦ as residues Tyr-404 and Phe-405 of RFC1. Previous studies of ␥ complex mutated in these two residues showed that the ␥ (␦ mut ) complex is less active than wild-type ␥ complex in replication assays, and its interaction with ␤ is compromised (23).
To determine whether RFC 1mut binds PCNA with lower affinity than wild-type RFC we developed a fluorescence assay by labeling PCNA at its two exposed Cys residues using the Oregon Green fluorophore. Titrations of RFC and RFC 1mut into PCNA OG reveal that RFC 1mut binds PCNA OG (K d app ϭ 3.1 M) about 10-fold weaker than wild-type RFC (K d app ϭ 0.34 M), demonstrating that the RFC1 subunit is involved in binding to PCNA. A second experiment was performed to test the above conclusion based on the PCNA-stimulated activity of Pol ␦. In Fig. 8, bottom panel, RFC 1mut is examined for activity with PCNA on a singly primed M13mp18 ssDNA template using a recombinant derivative of Pol ␦. The results show that the RFC 1mut is significantly less active than wild-type RFC, consistent with RFC1 forming an important contact site between RFC and PCNA. DISCUSSION RFC Subunit Arrangement and ATP Site Architecture-This study examines the subunit arrangement of the heteropentameric RFC clamp loader using recombinant heterodimeric subassemblies of RFC and individual RFC subunits to obtain information about subunit-subunit interactions within RFC. These protein reagents also made possible the reconstitution of three-and four-subunit RFC subassemblies. The results are all consistent with the presumed circular subunit arrangement, as no one subunit interacted with more than two other subunits. The five RFC subunits have been proposed to be arranged in a circle by analogy to the crystal structure of the circular pentamer E. coli ␥ 3 ␦␦Ј clamp loader (9,10) and is consistent with electron microscopy and atomic force microscopy studies of RFC (24). The final arrangement of RFC, illustrated in Fig. 9, confirms earlier proposals based on homologies between RFC and ␥ complex (9, 33) and on previous protein-protein interaction studies of human RFC (28,29).
The ATP sites of AAAϩ multimers are located at subunit interfaces, and the adjacent subunit contributes a catalytic Arg residue to facilitate ATP hydrolysis. With this in mind, there are two ways to order the five RFC subunits in a circle, clockwise and counter clockwise (i.e. Models 1 and 2 in Fig. 4E). The two arrangements result in RFC pentamers having either four competent ATP sites (Model 1) or three competent ATP sites (Model 2). The correct arrangement of subunits around the ring was determined by exploiting a finding herein showing that the RFC(3/4) complex contains DNA-dependent ATPase activity, and this ATPase is stimulated further by RFC2. The catalytic Arg residue of the SRC motif in either RFC3 or RFC4 was mutated to alanine, and the effect of these mutations on the ATPase activity of the RFC(3/4) complex was determined. The clockwise or counter clockwise arrangements of RFC predict opposite results, and only one of the two results was experimentally observed (i.e. RFC2 contributes an Arg finger to the ATP site of RFC3, and RFC3 correspondingly contributes an Arg finger to the ATP site of RFC4). The experimentally determined polarity of the RFC2/3/4 subunits and knowledge of the positions of RFC subunits clearly revealed Model 1 to be the correct structure of RFC complex (i.e. Model 1 of Fig. 4, and as illustrated in Fig. 9).
Interaction between RFC and PCNA-RFC1 is involved in a main connection to PCNA based on previous studies in the human system (25). The current study confirms this interaction in the yeast system and also demonstrates that the Tyr-404 and Phe-405 residues of RFC1 are involved in PCNA binding. The position of these two residues in RFC1 is analogous to that of the two hydrophobic residues in E. coli ␦ that bind the ␤ clamp and thus were hypothesized to be important to RFC1-PCNA interaction (18). This report also documents that RFC(3/4) and RFC(2/3) also form a main attachment site between RFC and PCNA. Because neither RFC2 nor RFC4 bind stably to PCNA it is possible that RFC3 alone forms the main contact to PCNA within these complexes (illustrated in Fig. 9). We observe here that the RFC(3/4) heterodimer displays ATP- dependent behavior in binding to PCNA, whereas the RFC(2/3) heterodimer binds PCNA in the absence of ATP. We propose that the PCNA binding site(s) within the RFC(3/4) complex is occluded in this heterodimer, and ATP results in a conformation change that makes these sites more accessible to PCNA. The ATP-independent interaction of RFC(2/3) with PCNA indicates that the PCNA binding site(s) of this heterodimer are more exposed to PCNA and thus do not require ATP to power a conformation change for PCNA interaction. However, the intact 5-subunit RFC requires ATP to observe PCNA binding (29), and thus the PCNA binding site(s) of the RFC(2/3) heterodimer must be obscured in the intact RFC, presumably by one or more of the other three subunits, and that ATP is required to expose them for PCNA interaction. The studies of this report do not rule out contact between PCNA and RFC2 or 5, as only strongly interacting complexes are observed by the non-equilibrium gel filtration technique used here. Indeed it has been suggested that all five RFC subunits contact PCNA in the human system (32).
Comparison of RFC with ␥ Complex-The E. coli ␥ complex, like RFC, contains five AAAϩ subunits that are required for clamp loading activity (10). However, unlike RFC, ␥ complex has only one strong subunit contact to the ␤ clamp, mediated by the ␦ subunit (22). Although the ␥ subunits contact ␤, they do so only with very weak affinity, and there is no detectable interaction thus far between ␦Ј and ␤ (15). Consistent with ␦ forming the main contact to ␤, the ␥ 3 ␦Ј subassembly lacking the ␦ subunit does not form a stable complex with ␤ (22). The fact that more than one RFC subunit binds PCNA tightly may underlie a divergence in mechanism between prokaryotic and eukaryotic clamp loaders. This difference may simply be a consequence of clamp loaders that function with either a dimer (␤) or trimer (PCNA) ring. For example, PCNA dissociates into monomers at low concentrations (34), and RFC may need multiple contact points with the PCNA trimer to prevent it from falling apart when it opens the ring at one interface.
The ␥ trimer has been referred to as the "motor" of ␥ complex, as these subunits are the only ones that bind ATP (9,10,35). The RFC(2/3/4) subassembly has been proposed to act like the ␥ 3 motor subassembly as they are the only RFC subunits that contain both the ATP site and the SRC arginine finger motifs, like ␥ (9, 35). Results of this study are consistent with this assignment as they demonstrate that RFC2, 3, and 4 bind one another and contain an inherent DNA-stimulated ATPase activity. Likewise, the analogous human RFC p40/p37/p36 complex contains DNA-dependent ATPase activity (28,29). Moreover, the current study also confirms the hypothesis that these RFC subunits require the SRC arginine finger for ATP hydro- lytic activity, as has been shown previously for the ␥ subunits of ␥ complex (21).
The E. coli ␦Ј stator contains an SRC motif that donates a catalytic arginine to the ATP site of ␥ 1 (21) but lacks the ATP binding site consensus sequence. Similarly, RFC5 contains an SRC motif and has a modified ATP binding site sequence suggesting that even if it binds ATP it may not easily hydrolyze it (6). According to the subunit arrangement determined here, RFC5 contributes a catalytic arginine to the ATP site of RFC2. The ␦Ј stator is thought to be a rigid subunit, perhaps acting as a backboard to direct the ATP-induced changes in ␥ subunits necessary for ␤ to bind ␥ complex (9, 10). It is not clear at this time whether RFC5 is rigid, and thus assigning a stator role for this subunit is still preliminary.
It has been proposed that RFC1 acts like the ␦ wrench to open the ring, although no clear evidence for this exists at the present time (9,18). This speculation derives from the fact that RFC1, like ␦, is the most divergent in sequence from the other subunits, lacks an SRC motif, and binds directly to the PCNA ring. Furthermore, as ␦ is positioned between the stator and a motor subunit in ␥ complex, so is RFC1 positioned in RFC, based upon the results in this report. If RFC1 is the wrench, one may predict that there will be a gap between the N-terminal domains of RFC1 (wrench) and RFC5 subunits, by analogy to the gap between the ␦ and ␦Ј subunits of E. coli ␥ 3 ␦␦Ј (see Fig.  9). Unlike ␦, RFC1 contains an ATP site, but mutational analysis demonstrates that this site is not required for clamp loading function and thus may be involved in some other activity (36).
It has been noted previously (18) that RFC1 contains two adjacent hydrophobic residues that are located in the same position as the two hydrophobic residues of ␦ involved in binding to ␤. Studies in this report show that RFC mutated in these residues of RFC1 binds less tightly to the clamp and that the RFC 1mut is impaired in replication activity, similar to E. coli ␥ complex containing a ␦ subunit that is mutated in the analogous hydrophobic residues (23). However, it also remains possible that the mutations in RFC1 examined here causes some alteration of the RFC structure that causes indirect effects that change PCNA interactions.
The apparent similarity between RFC1 and ␦ may imply that RFC1 opens the clamp like ␦, but until this is demonstrated, other possibilities also exist. For example, it has been pointed out that RFC3 and RFC5 also contain two adjacent hydrophobic residues in a similar position as ␦ and RFC1 (33). Therefore it is also possible that either of these other subunits acts like ␦. Top panel, wild-type RFC or RFC 1mut was titrated into PCNA labeled with a fluorescent reporter. Open circles represent reactions with wild-type RFC, and closed triangles represent reactions with RFC 1mut . Bottom panel, PCNA is titrated into replication assays containing a recombinant Pol ␦ derivative and either wild-type RFC or RFC 1mut . RFC and PCNA were preincubated with an SSBcoated singly primed M13mp18 ssDNA template coated with SSB before initiating replication with recombinant yeast Pol ␦ as described under "Experimental Procedures." Open squares represent reactions containing wild-type RFC, and open triangles represent RFC1 mut . Closed circles are reactions in which no RFC was added.
FIG. 9. RFC architecture and contacts to PCNA. A, cartoon of the subunit arrangement in the RFC pentamer looking down the C-terminal plane corresponding to the top view of ␥ complex in Fig. 1. ATP sites are located at subunit interfaces. B, cartoon of RFC side views. ATP sites of RFC and major subunit contacts to PCNA are shown in the two side views. The two side views illustrate the identity of each subunit that contributes an Arg finger to the four different ATP sites.
More importantly, RFC may simply function quite differently than ␥ complex, although the fact that both clamp loaders are composed of five AAAϩ proteins suggests that many aspects of their mechanism will be similar. For example, both RFC and ␥ complex bind ATP to associate with their respective clamps. However, the different oligomeric structures of their rings may require different mechanistic approaches to the problem of binding, opening and closing the rings around DNA. A detailed comparison between the prokaryotic and eukaryotic clamp loaders awaits crystal structure studies of RFC.