Impact of Individual Proliferating Cell Nuclear Antigen-DNA Contacts on Clamp Loading and Function on DNA*

Background: Crystal structures reveal direct clamp-DNA contacts whose functions are unclear. Results: Mutations of single N-terminal arginines/lysines in the inner PCNA ring affect DNA binding-linked steps in the clamp loading mechanism. Conclusion: Loss of individual PCNA-DNA contacts disrupts clamp loading and its interaction with DNA. Significance: DNA binding to specific locations within circular clamps influences other proteins that work with clamps on DNA. Ring-shaped clamp proteins encircle DNA and affect the work of many proteins, notably processive replication by DNA polymerases. Crystal structures of clamps show several cationic residues inside the ring, and in a co-crystal of Escherichia coli β clamp-DNA, they directly contact the tilted duplex passing through (Georgescu, R. E., Kim, S. S., Yurieva, O., Kuriyan, J., Kong, X. P., and O'Donnell, M. (2008) Structure of a sliding clamp on DNA. Cell 132, 43–54). To investigate the role of these contacts in reactions involving circular clamps, we examined single arginine/lysine mutants of Saccharomyces cerevisiae proliferating cell nuclear antigen (PCNA) in replication factor C (RFC)-catalyzed loading of the clamp onto primer template DNA (ptDNA). Previous kinetic analysis has shown that ptDNA entry inside an ATP-activated RFC-PCNA complex accelerates clamp opening and ATP hydrolysis, which is followed by slow PCNA closure around DNA and product dissociation. Here we directly measured multiple steps in the reaction (PCNA opening, ptDNA binding, PCNA closure, phosphate release, and complex dissociation) to determine whether mutation of PCNA residues Arg-14, Lys-20, Arg-80, Lys-146, Arg-149, or Lys-217 to alanine affects the reaction mechanism. Contrary to earlier steady state analysis of these mutants (McNally, R., Bowman, G. D., Goedken, E. R., O'Donnell, M., and Kuriyan, J. (2010) Analysis of the role of PCNA-DNA contacts during clamp loading. BMC Struct. Biol. 10, 3), our pre-steady state data show that loss of single cationic residues can alter the rates of all DNA-linked steps in the reaction, as well as movement of PCNA on DNA. These results explain an earlier finding that individual arginines and lysines inside human PCNA are essential for polymerase δ processivity (Fukuda, K., Morioka, H., Imajou, S., Ikeda, S., Ohtsuka, E., and Tsurimoto, T. (1995) Structure-function relationship of the eukaryotic DNA replication factor, proliferating cell nuclear antigen. J. Biol. Chem. 270, 22527–22534). Mutations in the N-terminal domain have greater impact than in the C-terminal domain, indicating a positional bias in PCNA-DNA contacts that can influence its functions on DNA.

Sliding clamps encircle DNA and are best known for making DNA replication processive by limiting polymerase dissociation from the primer template junction (4). Clamps also tether and coordinate exchanges of many other proteins at different sites on DNA, including proteins responsible for DNA replication, repair, recombination, and cell cycle control; for example, PCNA 2 helps coordinate switching of replicative and translesion bypass polymerases at sites of DNA damage (5,6). All clamps have the same basic structure with six ␣/␤ domains distributed across two or three subunits in a planar ring (e.g., Escherichia coli ␤ and human PCNA, respectively) (7). The ␤ sheets line the outer rim and form interfaces between subunits; the ␣ helices line the inner rim and present positively charged residues to complement the negatively charged DNA backbone (see Fig. 1). The clamp is loaded onto DNA by the clamp loader, a five-subunit complex of AAAϩ family proteins that couple ATP binding and hydrolysis to mechanical work (8 -13). Structural and mechanistic analyses of clamp loaders such as E. coli ␥ complex (11,14), bacteriophage T4 gp44/62 (8,15,16), and Saccharomyces cerevisiae RFC (10,(17)(18)(19) among others, have shown that ATP binding enables the clamp loader to bind and open the clamp and bind ptDNA, and ATP hydrolysis leads to the release of the clamp-ptDNA product (see Fig. 1A), which can then be used by DNA polymerase and other proteins. In recent pre-steady state kinetic studies of S. cerevisiae RFC and PCNA, we explicitly measured the rates of individual steps in the reaction, including PCNA opening/closing, ptDNA binding/release, ATP hydrolysis, and phosphate (P i ) release (17,18). The data revealed a pair of slow events that bookend the loading reaction: the first occurs after ATP binding to RFC (RFC acti-vation) and involves conformational changes that enable PCNA opening and ptDNA binding to form an RFC-ATP-PCNA open -ptDNA intermediate primed for ATP hydrolysis; the second occurs after ATP hydrolysis (RFC deactivation) and involves conformational changes that enable PCNA closure and release of the PCNA-ptDNA product.
On ATP binding, the AAAϩ modules in the clamp loader subunits begin rearranging toward a right-handed spiral conformation (10,18). PCNA accelerates this activation process (17,18) as it interacts with the base of RFC and is held open at one subunit interface (Fig. 1B, a molecular dynamics generated model of S. cerevisiae RFC bound to open PCNA) (20). In the clamp loader-ATP-clamp open complex, the protein subunits are twisted in a spiral such that the inner surface of the chamber complements the double helix geometry (9,10,20) (Fig. 1C, bacteriophage T4 gp44/62-ADP-BeF 3 -gp45-ptDNA complex; (8). Structural data indicate that the clamp does not open wide enough to allow direct entry of the duplex portion of ptDNA (8, 19 -21). Kinetic data indicate that short linear ptDNA can bind closed or partially open RFC-ATP-PCNA complex and stimulate PCNA opening (17,22). These findings support a model in which the single-stranded template enters through a small gap between the clamp subunits first, and then ptDNA threads upward through the full protein complex. Cationic residues in both the clamp loader and clamp present sites for ionic interactions with the duplex, and the template exits near the top through a gap between the clamp loader subunits (8,9). ATP binding to at least the three central AAAϩ modules is necessary for rapid binding of ptDNA (22). In turn, interaction with duplex DNA appears essential for the modules to achieve the final symmetric spiral conformation set for ATP hydrolysis (8,9). Consistent with these structural features, kinetic data show that ATP binding to RFC initiates and PCNA binding accelerates formation of the spiral complex, but only DNA binding (ptDNA or linear duplex) triggers a rapid burst of ATP hydrolysis (17,18); ptDNA has the same effect on E. coli ␥ complex (23). ATP hydrolysis is followed by clamp closure and clamp-ptDNA, ADP, and phosphate release (17,24). The exact nature and sequence of all the events following ATP hydrolysis remain under investigation.
Although ATP binding and hydrolysis initiate the assembly and disassembly phases of the reaction, respectively, interactions with ptDNA accelerate key steps: PCNA opening, ATP hydrolysis, and subsequent product release. Conserved cationic residues inside the S. cerevisiae RFC and E. coli ␥ complex that contact DNA are known to be essential for clamp loading (25,26). Many conserved cationic residues line the inner rim of clamps as well (e.g., 9 arginines and lysines per human or S. cerevisiae PCNA monomer; i.e., 27 per clamp; Fig. 1D), but any functional significance of these potential contacts with DNA is less well understood. In the crystal structure of E. coli ␤-ptDNA complex, the duplex is tilted within the ring, and the backbone interacts with specific residues (1). Of these, mutation of Arg-24 and Gln-149 was shown to disrupt loading of ␤ onto ptDNA, indicating that these direct interactions can influence clamp function. MD analysis and a crystal structure of the S. cerevisiae PCNA-ptDNA complex also indicate that the duplex is tilted and contacts specific residues in the ring (2,27).
Individually mutating the arginines and lysines in human PCNA to alanine showed that each one of the nine is required for processive DNA replication by polymerase ␦ (3). The mutations suppressed initiation of processive DNA synthesis, implying a defect in PCNA loading by RFC. However, the mutations apparently had little effect on stimulation of RFC-PCNA ATPase activity by ptDNA, leaving open the question of how loss of particular PCNA-DNA contacts affects clamp loading and function on DNA (3). A more recent study of the same mutants in S. cerevisiae PCNA (Fig. 1D) also detected no significant reduction in ptDNA-stimulated ATPase activity; indeed the rate was only 2-fold lower with a quadruple mutant than with wild type PCNA (K20A/K77A/R80A/R149A; 12 of 27 arginines/lysines in the ring eliminated) (2). Thus, the earlier finding that loss of even one cationic residue per PCNA monomer ablates processive DNA replication remained unexplained (3).
The effect of ptDNA on the steady state RFC ATPase rate (k cat ) is not an appropriate measure of the potential effects of DNA binding mutations in PCNA. A major reason is that the effect of ptDNA on k cat is similar in the presence or absence of PCNA, rendering any effect of PCNA mutations undetectable (k cat for RFC ϭ 0.02 s Ϫ1 , RFC-PCNA ϭ 0.05 s Ϫ1 , RFC-ptDNA ϭ 0.7 s Ϫ1 , and RFC-PCNA-ptDNA ϭ 0.8 s Ϫ1 ). In contrast, stark differences are revealed in the ATPase mechanism with and without PCNA under pre-steady state conditions (18). Another reason is that rate-limiting steps that determine the k cat are not fully known for any clamp loader (e.g., the ADP dissociation rate is unkown). Thus, the k cat cannot reveal which steps, if any, in the clamp loading mechanism are altered by disruption of PCNA contacts with DNA. It has been speculated that clamp-DNA interactions may facilitate positioning of ptDNA inside the clamp loader-clamp complex, induce clamp closure around ptDNA, and possibly influence clamp interactions with target proteins (and protein switching) by altering the relative location and tilt angle of DNA within the ring (1,28). Given the potential functional importance of DNA binding residues within clamps, we examined six single arginine/lysine mutants of S. cerevisiae PCNA for effects on individual events constituting the clamp loading reaction. The results demonstrate that the loss of even one cationic residue can alter DNA binding-linked steps in the reaction mechanism, as well as movement of PCNA on DNA.
Primer template DNA was prepared by annealing a 40-nucleotide primer (5Ј-ATT TCC TTC AGC AGA TAG GAA CCA TAC TGA TTC ACA TGG C-3Ј) and a 65-nucleotide template (5Ј-TAG TTA GAA CCT AAG CAT ATT AGT AGC CAT GTG AAT CAG TAT GGT TCC TAT CTG CTG AAG GAA AT-3Ј). DNAs were purchased from Integrated DNA Technologies and purified by urea gel electrophoresis. ptDNA labeled with 5-(and 6-) carboxytetramethylrhodamine (TAMRA; Invitrogen) at the 3Ј primer end was prepared as described (18).
PCNA Opening and Closing-PCNA opening/closing was monitored by FRET under steady state conditions on a Fluoro-Max-3 fluorometer (Horiba Jobin-Yvon). Fluorescence spectra of 0.25 M PCNA WC AEDANS were measured in buffer A (30 mM Hepes-NaOH, pH 7.5, 10 mM MgCl 2 , 100 mM NaCl, 2 mM DTT) at 25°C, in the absence and presence of 0.6 M RFC, 0.25 M ptDNA, and 0.5 mM ATP ( EX ϭ 290 nm or 336 nm, EM peak ϭ 490 nm). The signal was converted to FRET efficiency as follows, where E T is FRET efficiency, and I AD and I A are fluorescence intensities of AEDANS acceptor in the presence (PCNA WC AEDANS ) and absence (PCNA FC AEDANS ), respectively, of the tryptophan donor in PCNA at EX ϭ 290 nm. ⑀ A (1800 M Ϫ1 cm Ϫ1 ) and ⑀ D (4100 M Ϫ1 cm Ϫ1 ) are AEDANS and tryptophan extinction coefficients at 290 nm (19). A small contribution by RFC tryptophans to the signal (interprotein FRET) was first corrected as described (17,31 , 0.25 M ptDNA (when present), and 0.5 mM ATP. Three or more kinetic traces were averaged, and the signal was converted to FRET efficiency as described above and plotted versus time. The data were fit to single or multiple exponential functions for initial estimation of rate constants.
ptDNA Binding-DNA binding was monitored by fluorescence anisotropy of ptDNA TAMRA labeled at the 3Ј primer end in equilibrium conditions on a FluoroMax-3 fluorometer (Horiba Jobin-Yvon). 4 nM ptDNA TAMRA was titrated with 0 -50 nM RFC in the absence or presence of PCNA WC (2:1 [PCNA]:[RFC] ratio) in buffer A containing 0.1 mM ATP␥S and 0.05 mg/ml BSA at 25°C. The samples were excited with vertically polarized light at EX ϭ 555 nm, and the anisotropy was calculated from vertically (I VV ) and horizontally (I VH ) polarized emission intensities at EM ϭ 580 nm (I VV Ϫ GI VH /I VV ϩ 2GI VH , where G is the grating correction factor). The anisotropy values were plotted versus RFC concentration and K d obtained by fitting with a quadratic equation modified to correct for 1.8-fold increase in ptDNA TAMRA quantum yield on binding to RFC (Ϯ PCNA) (32).
The kinetics of DNA binding/release were measured by double-mixing stopped flow experiments in which RFC and PCNA WC were preincubated with ATP for various times (⌬t ϭ 0.02-3 s) and then mixed with ptDNA TAMRA in buffer A containing 0.05 mg/ml BSA at 25°C. TAMRA fluorescence was measured over time ( EX ϭ 550 nm, EM Ͼ 570 nm) as described (18). Final reactant concentrations were: 0.1 M RFC, 0.4 M PCNA WC , 0.04 M ptDNA TAMRA , and 0.5 mM ATP. Three or more kinetic traces were averaged, and the change in fluorescence intensity (relative to free ptDNA TAMRA ) was plotted versus time. The data were fit to a double exponential function for initial estimation of rate constants.
Phosphate (P i ) Release-P i release from RFC was measured under pre-steady state conditions by double-mixing stopped flow experiments in which RFC and PCNA WC were preincubated with ATP for various times (⌬t ϭ 0.02-2 s) and then mixed with ptDNA and MDCC-PBP in buffer A containing a P i contaminant mopping system of 0.1 unit/ml purine nucleoside phosphorylase (Sigma-Aldrich) and 0.2 mM 7-methylguanosine (R. I. Chemical Inc). Changes in MDCC-PBP fluorescence were measured over time ( EX ϭ 425 nm, EM Ͼ 450 nm) as described (18). The final reactant concentrations were 0.5 M RFC, 1 M PCNA WC , 2.5 M ptDNA, 0.5 mM ATP, and 10 M MDCC-PBP. Three or more kinetic traces were averaged, and P i concentration, determined from calibration curves generated by parallel experiments with standard P i solution (Sigma-Aldrich), was plotted versus time after subtracting a small amount of P i formed during ⌬t in the absence of ptDNA. The data were fit to double exponential ϩ linear function for initial estimation of rate constants.

RESULTS
The side chains of nine arginines and lysines on the inner surface of the PCNA monomer are shown in Fig. 1D. Residues Lys-13, Arg-14, Lys-20, Lys-77, and Arg-80 are located in the N-terminal domain I, whereas Lys-146, Arg-149, Lys-210, and Lys-217 are located in the C-terminal domain II of PCNA. We chose a subset of six residues distributed across the entire surface (Arg-14, Lys-20, Arg-80, Lys-146, Arg-149, and Lys-217), mutated them individually to alanine, and assessed the effects on DNA binding-related steps in the PCNA loading reaction catalyzed by RFC. All of the point mutations were made in PCNA WC , in which the four naturally occurring cysteines are modified to serine and Phe-185 and Lys-107 are modified to tryptophan and cysteine, respectively, for FRET-based analysis of PCNA opening and closure (17,19).
Effect of Mutating Inner Rim Cationic Residues on PCNA Opening by RFC-First, the structural integrity of the PCNA WC mutants was assessed by measuring FRET across the subunit interface for closed and open clamps (alone and with RFC and ATP, respectively) ( Table 1). The FRET signal between Trp-185 and AEDANS-labeled Cys-107 is high when PCNA WC AEDANS is closed and low when the clamp is open. The signal was converted to FRET efficiency (E T ) using data from parallel experiments with tryptophan-free PCNA FC AEDANS and correcting for a small level of interprotein FRET caused by RFC tryptophans, as described (17,31). E T for PCNA WC alone is 0.95, as reported previously (19), and ranges from 0.92 to 0.98 for the mutant clamps, indicating they are also closed in the absence of RFC. PCNA WC E T drops to 0.74 in the presence of RFC and ATP, on formation of the RFC-ATP-PCNA open complex. Domain I mutants, PCNA WC -R14A, PCNA WC -K20A, and PCNA WC -R80A exhibit similar reduction in E T with RFC and ATP, indicating that these mutant clamps are opened as well. E T for domain II mutants, PCNA WC -K146, PCNA WC -R149A, and PCNA WC -K217A, also drops with RFC and ATP, but PCNA WC -K217A E T remains quite high at 0.84, suggesting some perturbation of its interaction with RFC and/or opening, which are unrelated to its interaction with ptDNA.
The kinetics of PCNA opening were measured next to determine the rate and extent to which RFC can open these clamps independently of ptDNA. In stopped flow experiments, a solution of RFC and PCNA WC AEDANS was mixed rapidly with a solution of ATP, and the decrease in FRET as a consequence of PCNA opening was measured over time; note: the order of mixing does not affect the PCNA opening rate (17). The kinetic trace for PCNA WC AEDANS ( Fig. 2A) fit empirically to a single exponential yielded k open ϭ 2.3 s Ϫ1 , consistent with previous reports of PCNA opening rates (19,33). Clamps containing domain I mutations R14A, K20A, or R80A all showed the same opening kinetics as PCNA WC ( Fig. 2A and Table 1). Opening rates for clamps containing domain II mutations K146A or R149A were also comparable with PCNA WC (Fig. 2B), although the extent of opening was lower for PCNA WC -K146A, consistent with the steady state data. PCNA WC -K217A, however, exhibited defects both in the rate and extent of opening (Fig. 2B); thus, this mutant was not included in analysis of subsequent ptDNA binding-related steps in the reaction.
ptDNA Binding and Release Are Influenced by PCNA Inner Rim Cationic Residues-ATP␥S binding to RFC leads to formation of stable RFC-ATP␥S-ptDNA and RFC-ATP␥S-PCNA open -ptDNA complexes. We measured formation of these complexes by fluorescence anisotropy of a 40/65-nucleotide ptDNA substrate labeled with TAMRA dye at the 3Ј primer   . S1A) (17). ATP␥S-bound RFC has high affinity for ptDNA in the absence or presence of PCNA; thus, K d measurements with PCNA mutants are unlikely to report any significant differences unless the mutations severely impact interaction between ptDNA and RFC as well. Therefore we examined the interaction in finer detail by measuring the kinetics of ptDNA association after ATP binding to RFC and dissociation after ATP hydrolysis. We have shown previously that an increase in fluorescence reports ptDNA TAMRA binding to RFC (Ϯ PCNA) and decrease reports dissociation of the complex after ATP hydrolysis (17). Sequential mixing experiments were performed in which RFC and PCNA WC were preincubated with ATP for varying times (⌬t ϭ 0.02, 0.5, 2, and 3 s) and then mixed with ptDNA TAMRA , and the signal was monitored over time; note that, because of slow activation of RFC after ATP binding, a preincubation period of at least 2 s is required for maximal ptDNA binding and ATP hydrolysis (see Fig. 6 and supplemental Fig. S1B) (17). Fig. 3A shows data at ⌬t ϭ 2 s for PCNA WC with the increase in signal reporting ptDNA binding to the RFC-ATP-PCNA complex at 14 s Ϫ1 under these conditions (bimolecular binding constant estimate ϳ1 ϫ 10 8 M Ϫ1 s Ϫ1 ). Following ATP hydrolysis, the decrease in signal at 4 s Ϫ1 reports ptDNA-PCNA (and/or ptDNA) release from RFC. The  signal levels out higher than that for free ptDNA in steady state because the short linear substrate slips out from PCNA and is bound again by excess protein. The clamps containing domain I mutations, R14A, K20A, or R80A, exhibit slightly slower ptDNA binding compared with PCNA WC and distinctly different release kinetics, resulting in higher levels of ptDNA-bound complex in steady state (Fig. 3A). Clamps containing domain II mutations, K146A or R149A, exhibit similar ptDNA binding as PCNA WC and slightly different release kinetics (Fig. 3B).
The data were globally fit to a highly simplified model of ptDNA binding/release using KinTek Explorer (34) to obtain initial estimates of key rate constants (supplemental Fig. S1C). The model contained four steps: (i) k 1 , slow RFC activation after binding ATP and PCNA opening (assumed irreversible because PCNA opening is comparable with ATP and ATP␥S, indicating minimal PCNA closure prior to ATP hydrolysis) (19); (ii) k 2 /k Ϫ2 , high affinity interaction of ptDNA with the ATP-bound protein complex (K d ϭ 2 nM; supplemental Fig. S1A) (17); (iii) k 3 , ATP hydrolysis, RFC deactivation and PCNA closure (assumed irreversible because of ATP hydrolysis); and (iv) k 4 , release of ptDNA (assumed irreversible because of ATP hydrolysis) to complete the catalytic cycle. All of the rate constants were allowed to float during fitting of the PCNA WC data, and the relative change in signal for free:bound ptDNA was set at 1:1.8 based on the fluorescence intensity of ptDNA TAMRA alone and bound to RFC-ATP␥S-PCNA (17). Supplemental Fig. S1B shows the best fits as black lines overlaying the experimental data, and the corresponding rate constants are listed in supplemental Fig. S1C. With PCNA WC , the RFC activation step occurs at an apparent rate of 8.9 s Ϫ1 , ptDNA binding at 1 ϫ 10 8 M Ϫ1 s Ϫ1 , RFC deactivation at 5.6 s Ϫ1 , and ptDNA release at 0.5 s Ϫ1 . The mutant data were fit in the same manner as PCNA WC except the ptDNA-independent RFC activation step was fixed at 8.9 s Ϫ1 (supplemental Fig. S1B shows the data and fits for PCNA WC -R80A; supplemental Fig. S1C lists rate constants for all the mutants). All the domain I mutants exhibit Ն2-fold differences in the rates of one or more ptDNA-associated steps in the reaction. The K20A and R80A mutants exhibit slower ptDNA binding (k 2 ), R14A and R80A mutants exhibit slower RFC deactivation (k 3 ), and all three domain I mutants R14A, K20A, and R80A exhibit faster ptDNA release than PCNA WC (k 4 ). In case of domain II mutants, all the best fit rate constants for K146A are almost the same as for PCNA WC , whereas R149A exhibits somewhat slower RFC deactivation and faster ptDNA release than PCNA WC . The ptDNA binding/release data were analyzed further by fitting to a more comprehensive model of the clamp loading mechanism (together with all other kinetic data), as described later.
PCNA Closure around ptDNA Is Influenced by Inner Rim Cationic Residues-As noted earlier, ptDNA binding triggers ATP hydrolysis, which in turn leads to PCNA closure and complex disassembly. The ptDNA binding/release data in Fig. 3 indicate that single arginine and lysine mutations in PCNA can disrupt post-ATP hydrolysis step(s) in the reaction. We investigated whether that includes PCNA closure around ptDNA. Sequential mixing experiments were performed in which RFC and PCNA WC AEDANS were preincubated with ATP for varying times (⌬t ϭ 0.02, 0.5, and 2 s) and then mixed with ptDNA, and the tryptophan-AEDANS FRET signal was monitored over time; note that at short ⌬t both FRET decrease (PCNA WC AEDANS opening after ATP binding) and subsequent increase (PCNA WC AEDANS closing after ATP hydrolysis) are observed, whereas at longer ⌬t PCNA WC AEDANS is fully opened, and only the closing phase is observed (see Fig. 6) (17). Fig. 4 shows data at ⌬t ϭ 2 s for all clamps, and an exponential fit of the PCNA WC trace yields an apparent rate of 8 s Ϫ1 for closure. All the mutant clamps close slightly slower than PCNA WC , although PCNA WC -K20A and PCNA WC -R80A are again the worst affected, with Ͼ2-fold decreases in closing rates. These data were analyzed further by fitting to a comprehensive model of the clamp loading mechanism as described later.
Phosphate Release toward the End of the PCNA Loading Cycle Is Also Influenced by Inner Rim Cationic Residues-Previous studies of the RFC mechanism indicate that a burst of phosphate (P i ) release occurs after PCNA closure around ptDNA and is associated with PCNA-ptDNA release from RFC (17, 18). If this model is correct, the P i release kinetics should reflect the defects in clamp closure (Fig. 4) and complex disassembly (Fig.  3) caused by the mutations. We measured P i release in the first catalytic turnover using a fluorescent reporter MDCC-PBP that binds free P i rapidly and with high affinity (k on ϭ 1.4 ϫ 10 7 M Ϫ1 s Ϫ1 and K d ϭ 0.1 M) (30). Sequential mixing experiments were performed in which RFC and PCNA WC were preincubated with ATP for varying times (⌬t ϭ 0.02, 0.5, and 2 s) and then mixed with ptDNA and MDCC-PBP, and the fluorescence signal was monitored over time; note that, because of slow activation of RFC after ATP binding, a preincubation period of at least 2 s is required for maximal ATP hydrolysis and P i release (Fig. 6). Fig.  5 shows data at ⌬t ϭ 2 s for all the clamps, and an exponential ϩ linear fit of the PCNA WC trace yields a burst rate k Pi Release ϭ 11 s Ϫ1 (following a short lag) and k cat ϭ 0. 8 (Fig. 5A). Nonetheless, the apparent steady state ATPase rate of RFC with these mutants is almost the same as with wild type PCNA (e.g., 0.6 s Ϫ1 for PCNA WC -R80A), confirming that the ATPase k cat is unsuitable as a primary measure of changes in the PCNA loading mechanism. Also, consistent with the generally milder impact of clamp II domain mutations, K146A and R149A, on earlier steps in the reaction, P i release with these mutants is comparable with PCNA WC , although PCNA WC -R149A exhibits a slightly lower burst rate and amplitude (Fig. 5B). These data were analyzed further by fitting to a comprehensive model of the clamp loading mechanism (together with all other kinetic data) as described below.
A Kinetic Model of the Impact of PCNA-ptDNA Contacts during and after Clamp Loading-To determine the rate constants of transient events in the reaction, all the experimental data were fit simultaneously to a current model of the PCNA loading mechanism using KinTek Explorer (17). The following is a description of the analysis for PCNA WC and PCNA WC -R80A, a representative of the domain I mutants that exhibit the most significant defects (data shown in Fig. 6; see kinetic scheme shown in Fig. 7A). Fig. 6 (A and B) provides a visual summary of the sequence of events following 2 s of RFC preincubation with ATP and PCNA WC or PCNA WC -R80A, respectively. The experimental data and the fits to each data set at ⌬t ϭ 0.02, 0.5, and 2 s with PCNA WC and PCNA WC -R80A are shown for ptDNA binding/release (Fig. 6, C and D), PCNA opening/closing (Fig. 6, E and F), and P i release (Fig. 6, G and H). The fits are shown as black lines overlaying the data, and the corresponding best fit parameters are listed in Table 2. The model begins with ATP binding to RFC with a bimolecular rate constant fixed at 1 ϫ 10 8 M Ϫ1 s Ϫ1 (k 1 ) and a dissociation rate fixed at 100 s Ϫ1 (k -1 ). We assumed a fast k 1 because the ATPase rates are independent of ATP concentration in our experiments (500 M ATP), and k Ϫ1 was based on the measured affinity of RFC for ATP␥S (K d1 ϭ ϳ1 M) (18,22,35). The same parameters were fixed for mutant PCNA, because this step is independent of ptDNA. The next step is PCNA binding to RFC (because PCNA accelerates ATP-induced RFC activation), and the rate was fixed at the reported value of 2 ϫ 10 8 M Ϫ1 s Ϫ1 (k 2 ) (33). The dissociation rate was fixed at 10 s Ϫ1 for a K d of 0.05 M (assuming that the K d prior to activation is similar to that in the absence of ATP, reported as 0.04 M) (33). The same parameters were fixed for mutant PCNA. The next step is ATP-induced activation of the complex. The rates, k 3 and k Ϫ3 , were allowed to float during data fitting but linked to maintain an equilibrium constant of 0.02 (1/K 3 ϭ k Ϫ3 /k 3 ), resulting in a net K d of 1 nM (K d2 ϫ 1/K 3 ), the reported value for the activated *RFC-ATP-PCNA complex (33,36) (Fig. 7A; an asterisk denotes ATP-activated RFC). The fit yielded k 3 ϭ 13 s Ϫ1 , and k Ϫ3 ϭ 0.26 s Ϫ1 for PCNA WC and identical values for PCNA WC -R80A. The next step, PCNA opening, was linked to RFC acti- We have shown previously that ATP-activated RFC binds ptDNA rapidly without PCNA or with closed, partially open or open PCNA (17,22). With PCNA WC , data fitting yielded rate constants of 1 ϫ 10 8 M Ϫ1 s Ϫ1 (k 5 ) and 1.2 ϫ 10 8 M Ϫ1 s Ϫ1 (k 6 ) for ptDNA binding to *RFC-ATP-PCNA WC and *RFC-ATP-PCNA WC open complexes, respectively (Fig. 6C and Table 2), FIGURE 6. Global fitting of kinetic data to a minimal mechanism of PCNA loading. A and B, overlay of kinetic data for PCNA WC (A) and PCNA WC -R80A mutant from DNA binding/release, PCNA opening/closing, and P i release experiments (B) provides a visual of rapid ptDNA binding to RFC-ATP-PCNA complex (formed during preincubation, ⌬t ϭ 2 s), which triggers ATP hydrolysis (not shown), PCNA closure, P i release, PCNA-ptDNA dissociation, and catalytic turnover. C-H, kinetic data measured at ⌬t ϭ 0.02, 0.5, and 2 s for ptDNA binding/release (C and D), PCNA opening/closing (E and F), and P i release for PCNA WC and PCNA WC -R80A (G and H), respectively, were all fit simultaneously to the model described in Fig. 7 and Table 2. The black lines are simulations generated by the model based on parameters listed in Table 2. consistent with the 14 s Ϫ1 rate measured at 0.1 M RFC (Fig.  3A). At this stage a striking difference is observed between PCNA WC and PCNA WC -R80A. ptDNA binds *RFC-ATP-PC-NA WC -R80A complex at 0.17 ϫ 10 8 M Ϫ1 s Ϫ1 (k 5 ) and *RFC-ATP-PCNA WC -R80A open complex at 0.4 ϫ 10 8 M Ϫ1 s Ϫ1 (k 6 ), which are 6-and 3-fold slower rates, respectively, than observed for PCNA WC (Fig. 6D and Table 2). ptDNA binding rates were similarly slow for the other domain I mutants: 0.22 ϫ 10 8 M Ϫ1 s Ϫ1 (k 5 ) and 0.44 ϫ 10 8 M Ϫ1 s Ϫ1 (k 6 ) for PCNA WC -R14A, and 0.15 ϫ 10 8 M Ϫ1 s Ϫ1 (k 5 ) and 0.23 ϫ 10 8 M Ϫ1 s Ϫ1 (k 6 ) for PCNA WC -K20A. The ptDNA dissociation rates, k Ϫ5 and k Ϫ6 , were linked to corresponding binding rates to maintain the measured K d of ϳ2 nM (supplemental Fig. S1A). The relative change in signal for free:bound ptDNA was set at 1:1.8 during data fitting for both clamps, based on the fluorescence intensity of ptDNA TAMRA alone and bound to RFC-ATP␥S-PCNA complexes with PCNA WC and the mutants (data not shown) (17). Consistent with the milder effect of domain II mutations, ptDNA binding rates were 1. The next step in the reaction, ATP hydrolysis, was set at an irreversible rate of 25 s Ϫ1 (k 7 ), in keeping with the experimentally measured range of 20 -50 s Ϫ1 for PCNA (18) (allowing k 7 to float during data fitting yields a rate of 17-30 s Ϫ1 ). After ATP hydrolysis, a step in the reaction limits P i release, as indicated by the lag phase in the data (Fig. 5). PCNA closure occurs during this lag, as reported by increase in PCNA WC AEDANS FRET (Fig.  6A). Complex dissociation (PCNA-ptDNA release), as reported by decrease in ptDNA TAMRA fluorescence, appears after PCNA closure (Fig. 6A). Thus, the model includes PCNA closing right after ATP hydrolysis, and the fit yields a rate of 5.1 s Ϫ1 (k 8 ) for PCNA WC (Fig. 6E and Table 2). PCNA WC -R80A closure around ptDNA is slower at 2.8 s Ϫ1 , indicating that this step is also affected by loss of a PCNA-DNA contact (Fig. 6F and Table  2). Clamp closure rates are 3 and 3.3 s Ϫ1 for the other domain I mutants, PCNA WC -R14A and PCNA WC -K20A, respectively. Note that these rates are consistent with the RFC deactivation/ clamp closure rates obtained from the simple model of ptDNA binding/release (supplemental Fig. S1C). In case of domain II mutants, the clamp closure rates were 5.1 s Ϫ1 for PCNA WC -K146A (comparable with PCNA WC ) and 4.1 s Ϫ1 for PCNA WC -R149A (intermediate between PCNA WC and domain I mutants). The remaining product dissociation steps, PCNA-ptDNA release (k 9 ) and P i release (k 10 ), were fixed at a high rate of 500 s Ϫ1 in the model for both clamps because these events could simply be limited by the preceding closure of PCNA (allowing these parameters to float during data fitting yielded rate constants Ͼ 200 s Ϫ1 ). Thus, for PCNA WC , relatively rapid clamp closure (Fig. 6E) is followed by correspondingly rapid P i release (Fig. 6G) and PCNA-ptDNA release (Fig. 6C), whereas for PCNA WC -R80A, relatively slow clamp closure (Fig. 6F) is followed by correspondingly slow P i release (Fig. 6H) and PCNA-ptDNA release (Fig. 6D). ADP release rates are completely unknown; hence this step is not included in the model.
The catalytic cycle is now complete, but two additional events were included in the model to account for the experimental conditions and data. A PCNA-ptDNA dissociation step, i.e., PCNA slipping off linear ptDNA, was added to allow recycling of these substrates in the steady state phase of the reaction. The fitting yields a dissociation rate of 0.7 s Ϫ1 (k 11 ) for PCNA WC (the reverse rate, k -11 was set to 0 because there is no available measure of closed PCNA slipping onto linear DNA). Interestingly, the dissociation rate for PCNA WC -R80A was 2.5fold higher at 1.8 s Ϫ1 , which suggests that loss of a PCNA-DNA contact enables the mutant to move on ptDNA and slip off faster than wild type PCNA (Table 2). Faster ptDNA release enables rebinding by excess RFC in the reaction, contributing to more ptDNA bound with PCNA WC -R80A compared with PCNA WC in steady state (Fig. 6, C and D). The dissociation rate was faster (1.5 s Ϫ1 ) for PCNA WC -R14A and PCNA WC -K20A as well. In case of domain II mutants, the dissociation rates were 1 s Ϫ1 for PCNA WC -K146A and 0.8 s Ϫ1 for PCNA WC -R149A, closer to the rate with PCNA WC . Finally, PCNA isomerization was included in the model to better fit a second slow phase observed in PCNA opening/closing kinetics. We have proposed previously that the clamp partitions between different conformations that bind RFC with high or low affinity (PCNA and **PCNA in Fig. 7A, respectively). The fit yields similar rate constants of 21 s Ϫ1 (k 12 ) and 0.9 s Ϫ1 (k Ϫ12 ) for PCNA WC and 17 s Ϫ1 (k 12 ) and 0.3 s Ϫ1 (k Ϫ12 ) for PCNA WC -R80A. Inclusion of this step is speculative; however, evidence from MD simulations indicates that clamps transiently access multiple conformations, one or more of which may be bound preferentially by the clamp loader (20,37). Finally, a scaling factor representing the number of ATPase sites per RFC complex was allowed to float during data fitting and indicated a burst of 2.7 ATP molecules hydrolyzed rapidly in one catalytic turnover with PCNA WC . This value is consistent with previous kinetic and structural studies of various clamp loaders, indicating that three ATPase active sites are necessary and sufficient for loading a clamp onto DNA (8,17,22,23,38).

DISCUSSION
The inner surface of circular clamps has positive electrostatic potential that favors enclosure of duplex DNA within the ring. Beyond this topological linkage, the role of direct contacts between clamps and DNA and how they might influence clamp position, movement and interaction with other proteins has been the subject of structural and computational studies in recent years. The reports all indicate that specific interactions between cationic residues and the phosphodiester backbone tilt the clamp relative to the DNA axis, affecting its orientation and sliding on the duplex, which in turn might affect the positions and functions of proteins bound to the clamp (1, 27, 28, 39 -41). The finding that Sulfolobus solfataricus Rad9-Rad1-Hus1 clamp (9-1-1) can bind DNA ligase, DNA polymerase, and FEN1 simultaneously (42), and a crystal structure of three human flap endonuclease molecules (FEN1) bound to one PCNA clamp (43) support the hypothesis that up to three proteins can bind a trimeric clamp and wait their turn to act on DNA. A recent EM and MD analysis of human FEN1-PCNA and FEN1-9-1-1 binary complexes indicates highly flexible linkage between the proteins in the absence of DNA, allowing them to adopt different relative orientations (28). In the DNAbound ternary complexes, however, the proteins are stabilized in one orientation with tilted DNA passing through the clamp and into the FEN1 active site. The authors suggested that contacts with PCNA help anchor DNA and, consequently, FEN1 in a conformation that enables catalysis (28). The importance of direct clamp-DNA contacts seems such that mutating even one of the nine arginine/lysine residues per human PCNA monomer to alanine prevents polymerase ␦ from initiating processive DNA synthesis (3). Moreover, structural alignments of diverse clamps such as E. coli ␤ and Pyrococcus furiosus, S. cerevisiae, and human PCNA show high positional conservation among these residues, indicating sustained functional significance through evolution.
Detailed understanding of the RFC-catalyzed PCNA loading mechanism enabled us to test whether and how a specific contact between clamp and DNA might influence the actions of another protein that works with the clamp on DNA, in this case the clamp loader. The assessment is complicated by the fact that the other protein also binds DNA. In case of clamp loaders, the central chamber of the five-subunit complex presents a large DNA binding surface (Fig. 1, A and B) (8 -10); hence, we did not anticipate substantive reduction in the DNA binding affinity of RFC-ATP-PCNA complex on removal of a single cationic residue in PCNA (supplemental Fig. S1A). The steady state RFC ATPase rate was also not expected to be informative, because the k cat values are similar with or without PCNA (18) and may be limited by events that are independent of PCNA-DNA interaction. However, with knowledge of kinetic and thermodynamic parameters governing distinct steps in the clamp loading reaction (17,18), it was possible to directly measure any consequences of disrupting PCNA-DNA contacts on events related to DNA binding and release in the reaction. It was necessary to examine the cationic residues individually because of the unresolved question of why the loss of any one of them in human PCNA results in a nearly complete loss of processive DNA replication (3). S. cerevisiae PCNA has nine arginines/lysines in the same positions as in human PCNA (Fig. 1D). We mutated a subset of six of these to alanine (Arg-14, Lys-20, Arg-80, Lys-146, Arg-149, and Lys-217) and measured the impact on the clamp loading mechanism.
The following is a summary of key steps in the clamp loading mechanism, based mainly on kinetic and structural analysis of S. cerevisiae proteins and supplemented with analogous information about E. coli and bacteriophage T4 proteins. Prior to ATP binding, the clamp loader subunits are in a relatively disorganized conformation (E. coli ␥ complex structure) (11). ATP binding initiates clamp loader activation, and the AAAϩ modules begin to organize in a right-handed spiral (Fig. 7B, I) that can bind a clamp at its base (Fig. 7B, II) (S. cerevisiae RFC-ATP␥S-PCNA closed structure) (10). The interaction stimulates completion of ATP binding and accelerates clamp loader activation to a conformation that holds the clamp open in a spiral (Fig. 7B, III) (S. cerevisiae RFC-ATP␥S-PCNA open MD structure) (20). ptDNA binds the clamp loader-ATP-clamp complex rapidly and with high affinity (Fig. 7B, IVa and IVb) and can induce clamp opening (17,22), implying that the singlestranded template can slip into the complex before the clamp is fully open. Indeed structural data suggest that the gaps in both the clamp loader and clamp may not open wide enough for direct entry of duplex DNA (8,20,21). Once inside, ptDNA can thread upward through the protein chamber, interacting with cationic residues in both the clamp and clamp loader. The 3Ј end of the primer is blocked at the roof of the chamber and the template bends and escapes through a gap between the clamp loader subunits (T4 gp44/62-ADP-BeF 3 -gp45-ptDNA and E. coli ␥ complex-ADP-BeF 3 -ptDNA structures) (8,9).
DNA binding locks the AAAϩ modules into catalytically active conformation triggering ATP hydrolysis (Fig. 7B, V); S. cerevisiae RFC and E. coli ␥ complex kinetics indicate three ATP molecules are hydrolyzed rapidly per catalytic turnover (17,18,23). ATP hydrolysis breaks the AAAϩ spiral symmetry (T4 gp44/62-ADP-BeF 3 -ADP-gp45-ptDNA structure) (8), deactivating the clamp loader and disrupting its interactions with both ptDNA and the open clamp. As a result, the clamp closes around ptDNA (Fig. 7B, VI), and the complex dissociates, releasing the clamp-ptDNA and P i products (Fig. 7B, VII); E. coli ␥ complex kinetics indicate that P i is released prior to clamp-ptDNA (24). The clamp can slide off the ends of short linear ptDNA used in the experiments, and both substrates become available for subsequent catalytic turnovers (Fig. 7B,  VIII). Notably, this recycling step provides an indirect measure of clamp movement on DNA.
Of the six residues examined in this study (Fig. 1D), R14A, K20A, or R80A mutations in the N-terminal domain I slowed binding of ptDNA to RFC-ATP-PCNA (Fig. 6, C and D; and 7B, IV). Notably, ptDNA binding prior to full PCNA opening was disrupted the most ( Table 2 and Fig. 7B, IVa). We interpret this result to mean that after the template strand slips into PCNA, these positively charged residues are important for guiding ptDNA upward through the clamp and into the clamp loader to form a tight complex. The domain I mutations also slowed PCNA closure around ptDNA after ATP hydrolysis (Fig. 6, E and F; and 7B, VI). This result supports the idea that the contacts aid transfer of ptDNA from clamp loader to clamp and transition of the clamp from open spiral to closed planar form, thus promoting clamp-ptDNA release from the clamp loader. All three mutants display correspondingly slower complex dissociation (Fig. 3A) and P i release (Fig. 5A) following ATP hydrolysis compared with PCNA. The third effect of the domain I mutations was to hasten the clamp slipping off ptDNA ( Table 2; Fig. 7B, VIII; and supplemental Fig. S1C). This result indicates that loss of even one of the contacts (three per homotrimer) can alter the position and movement of the clamp on DNA, which in turn could affect the activities of clampbound proteins. Thus, for the analogous K14A, K20A, and K80A mutants of human PCNA, the loss of processive polymerase ␦ activity can be explained by disruption of both clamp loading and position on DNA (3). Of the three S. cerevisiae PCNA C-terminal domain II mutations examined in this study, K217A slowed clamp opening (Fig. 2B), which led to a significant reduction in ptDNA binding (data not shown). This defect, and perhaps additional problems in subsequent steps of the reaction, can explain the lack of processive polymerase ␦ activity with the human PCNA K217A mutant (3). In contrast, the K146A mutant of S. cerevisiae PCNA exhibited similar clamp loading kinetics as wild type PCNA, and the differences were Ͻ2-fold for R149A (Figs. 3B, 4B, and 5B). For the analogous human PCNA mutants R146A and R149A, a defect after clamp loading (e.g., in productive interaction with the polymerase) may underlie the loss of processive polymerase ␦ activity (3). Contacts between the PCNA N-terminal domain I and DNA have a greater effect on the loading mechanism, suggesting that positional bias of DNA toward this domain is important for clamp function.
To summarize, we examined a current hypothesis that direct contacts between circular clamps and DNA influence the location and orientation of DNA within the ring, as well as clamp movement on DNA, which in turn could modulate the activities of other proteins that work with clamps on DNA. By measuring the RFC catalyzed PCNA loading mechanism at high resolution, it was possible to assess the functional significance of individual ionic interactions between the clamp and DNA. The results confirm that DNA binding to specific locations within the PCNA ring can modulate the function of a PCNA-binding protein and affect PCNA movement on DNA. Mutant clamps that can be loaded but exhibit altered dynamics on DNA present useful tools for investigating how clamps regulate and coordinate the functions of other DNA metabolic proteins.