Mapping The Interaction Of DNA With The Escherichia Coli DNA Polymerase .

2005 Nature Publishing Group http://www.nature.com/nsmbARTICLESFigure 2 Effects of γBCD mutants. (a) Relative steady-state fluorescence emission spectra obtained with excitation at 460 nm. Fluorescein (500 nM) attachedto D106C δA in clamp loader plus 4 µM D253C β labeled with TMR. A schematic representation of the experiment is shown, and components in theindividual spectra are labeled. (b) Clamp binding and release as monitored by the reduction in donor fluorescence at 515 nm. Top, clamp binding (red bars)as calculated by Fbind-mutant / Fbind-WT where Fbind Fno ATP – FATP (compare black and red traces in a). Bottom, clamp release (green bars, 6.6 µM DNA;blue bars, 5 mM ADP plus DNA) calculated by ( Frelease-mutant / Fbind-mutant) / ( Frelease-WT / Fbind-WT) where Fbind is as described above and Frelease Frelease – FATP (compare green traces for DNA-dependent release or blue traces for ADP DNA-dependent release in a). (c) ATPase assays. Dataplotted are the stimulation in ATPase rates from addition of primer-template DNA where one-fold is equivalent to no stimulation. Error bars, s.e.m. of at leasttwo measurements. The relative basal ATPase rates (in the absence of DNA) are indicated below the graph.(Table 1 and Fig. 1b,c). Arginine and lysine residues were individuallyreplaced by glutamate. Ser132 in the γBCD subunits and Thr121 in theδ′E subunit were mutated to alanine. We created two additional mutations (R133A and K161A) to assess whether replacement of a positivelycharged residue by alanine instead of glutamate has a measurable effect.Finally, two additional control mutations (S33A and R80E) were introduced at sites in the γBCD subunits that were not expected, on the basis ofthe model, to interact with DNA. Because the δA subunit is highly divergent in sequence and structural detail with respect to the other clamploader subunits17,26, we excluded the δA subunit from our analysis.FRET assay for clamp binding and releaseWe can monitor the binding and release of the β clamp by the clamploader complex through fluorescence resonance energy transfer(FRET)25. In our assay, clamp loaders were reconstituted with δAsubunits labeled with fluorescein (donor). These complexes weremixed with β clamps labeled with tetramethylrhodamine (TMR,acceptor). In a typical experiment, the intensity of donor fluorescence was reduced after the addition of ATP, owing to resonanceenergy transfer to the acceptor on the β clamp (Fig. 2a, wild type).As noted previously25, the addition of primer-template DNA leadsto a partial reversal of this effect, presumably owing to the DNAstimulated hydrolysis of ATP. The experiments were done in the presence of excess ATP, and we assumed that the clamp loader rebindsATP and the clamp upon DNA-stimulated release. This is a probableexplanation for the observation that the FRET signal does not revertcompletely to its original value upon the addition of DNA. The addition of a five-fold excess of ADP over ATP did indeed cause completereversal of the FRET signal, indicating that these labeled complexesare not bound together irreversibly.Experiments using clamp loaders labeled with donor in the presenceof unlabeled clamps showed no substantial changes upon the addition ofATP or nucleotides, nor did acceptor-labeled clamps in the presence ofunlabeled clamp loaders (Supplementary Fig. 1 online). This indicatesthat the signals we observed are due to resonance energy transfer and notmerely donor quenching. Because the fluorescein donor emission signaltrailed into the acceptor emission region ( 550–615 nm), the concomitant reduction in donor fluorescence led to the somewhat misleadingappearance that the acceptor signal decreases when energy transfer isobserved from the donor. When corrected for the decreases in donorfluorescence in this region, acceptor fluorescence showed the reciprocalincreases expected when FRET occurs (Supplementary Fig. 1 online).Given that the quantum yield of protein conjugates with fluoresceinis typically four times higher than that of those with TMR27, we didnot expect to see overall changes in acceptor intensity that are as largeas those in the donor. Notably, this assay provides a means to evaluatewhether DNA can stimulate the release of clamp from various mutantforms of the clamp loader.NATURE STRUCTURAL & MOLECULAR BIOLOGY VOLUME 12 NUMBER 2 FEBRUARY 2005185

2005 Nature Publishing Group http://www.nature.com/nsmbARTICLESFigure 3 DNA-binding affinity for γBCD mutants. (a) Fluorescence anisotropy of 5′-TAMRA-labeled 40-nucleotide template annealed to 30-nucleotide primer.TAMRA-labeled DNA (100 nM) was excited at 550 nm with emission signal collected at 580 nm in the presence of 1 µM wild-type β, 1 mM ADP-BeFx(1 mM ADP, 2 mM BeCl, 10 mM NaF). Fits were obtained using the equation as described in Methods. (b) Dissociation constants of γBCD mutants for DNA.Error bars are the standard error of the fit.Mutations in the BCD subunitsWe examined the effects of the mutations in the γBCD subunits on DNAstimulated clamp release using our FRET assay (Fig. 2a). The reductionin the peak intensity of the donor fluorescence can be used to estimate therelative extent of clamp binding and subsequent release (Fig. 2b). Slidingclamps are highly negatively charged, and so the replacement of positivelycharged residues with glutamate in the γBCD subunits of the clamp loadermight be expected to disturb the loader-clamp interaction independentof the effects on DNA binding. Indeed, two mutations in residues nearthe clamp-interacting helix (R105E and R98E) had the largest effect onclamp binding ability. Nevertheless, all mutations (Table 1) had at mosta small effect on the ATP-dependent ability of the complex to bind the βclamp (Fig. 2). Instead, the results suggest that the residues we altered arecritical for interaction with DNA (Figs. 2 and 3).Clamp loaders with mutations near the clamp-interacting and central helices (R98E, K100E, R105E, S132A, R133A and R133E) were alldeficient in DNA-dependent clamp release, whereas those with mutations near the SRC helix (K161A and K161E) showed relatively modest decreases in this regard. The only mutant clamp loader complexesthat showed no change in DNA-dependent clamp release relative to thewild-type complex were those containing the control mutations (S33Aand R80E). None of the mutations affected the ability of excess ADPto release the clamp loader complexes from the clamps, indicating thatirreversible aggregation is not a complicating factor.A characteristic property of clamp loader complexes is the abilityof DNA to stimulate their ATPase activity. We used a kinase-coupledassay (see Methods) to measure the ATPase activity of the wild-typeand variant clamp loader complexes in the presence and absence ofDNA (Fig. 2c). The results for the mutant clamp loader complexes inthis ATPase assay were consistent with the effects of these mutationson DNA-dependent clamp release. Mutations that disrupted the clamprelease response to DNA also showed deficiencies in the stimulationof ATPase activity by DNA. One distinctionbetween the ATPase data and the clamp releasedata involves the response of the S132A andR133A γBCD mutants. These mutants wereimpaired in DNA-dependent clamp release,but clamp loader complexes containing thesemutations showed substantial, althoughreduced, ATPase stimulation by DNA. Thesedifferences may reflect limitations of the clamprelease assay that obscure the effects of intermediate levels of ATPase activity. It is also possible that clamp release and ATPase activity arenot always directly coupled in the clamp loadermechanism.Previous studies have monitored the changesin the fluorescence anisotropy of labeled DNAas a reporter for binding of the E. coli clamploader to duplex DNA28,29. Use of the poorlyhydrolyzable nucleotide analog ATPγS insteadof ATP was found to promote much morestable binding of the clamp loader complex toFigure 4 Effects of δ′E mutants. (a) Clamp binding and release from FRET experiments compared withclamp loaders containing wild-type δ′E subunits. Clamp binding (red bars) and clamp release (greenDNA, presumably because ATP hydrolysis trigbars, 6.6 µM DNA; blue bars, 5 mM ADP plus DNA) was calculated by changes in donor fluorescencegers release of the complex. Following a simiby the same methods as in Figure 2b. (See Supplementary Fig. 2 online for raw data.) Error bars representlar approach, we used DNA labeled with thethe s.e.m. of at least two measurements. (b) DNA-binding affinity for δ′E mutants. TAMRA-labeledfluorophore TAMRA (Invitrogen) to assess theDNA (100 nM) was excited at 550 nm with emission signal collected at 580 nm in the presence ofeffects of the mutations in the γBCD subunits on1 µM wild-type β, 1 mM ADP-BeFx. Fits were obtained using the equation as described in Methods.clamploader affinity for DNA. As suggested(c) Dissociation constants of δ′E mutants for DNA. Error bars are the standard error of the fit. All clampby other studies30–34, we used ADP-berylliumloaders examined contained wild-type γBCD subunits unless noted otherwise.186VOLUME 12 NUMBER 2 FEBRUARY 2005 NATURE STRUCTURAL & MOLECULAR BIOLOGY

ARTICLES 2005 Nature Publishing Group http://www.nature.com/nsmbFigure 5 Specificity of DNA-stimulated clamprelease. (a) Relative steady-state fluorescenceemission spectra obtained with excitation at460 nm. Fluorescein (500 nM) attached toD106C δA in clamp loader plus 4 µM D253Cβ labeled with TMR. (b) Clamp release basedupon return of donor fluorescence at 515 nmupon addition of 4 µM hairpin DNAs to donorfluorescence values obtained before clamp bindingby addition of 1 mM ATP. Error bars, s.e.m. of atleast two measurements; nuc, nucleotide.fluoride (ADP-BeFx) as the ATP analog rather than ATPγS because wefound that ADP-BeFx gives more durable and robust changes in fluorescence anisotropy (data not shown).Clamp loader complexes containing the various mutant forms ofthe γBCD subunits were titrated into solutions containing a constantamount of TAMRA-labeled primer-template DNA, β clamp and ADPBeFx (Fig. 3). The wild-type protein and the control γBCD mutations(S33A and R80E) bound strongly to the labeled DNA, with apparentdissociation constants (Kd) of 100–200 nM. All other variant forms ofthe γ complex exhibited weakened binding. Consistent with the clamprelease data, the clamp loader complexes with the K161A and K161Emutations in the γBCD subunits showed modest weakening of affinity(Kd 0.8–1.2 µM), whereas mutations in the other potential DNAbinding residues resulted in substantially reduced binding (Kd 25 µM).spiral. Its position is crucial for the recognition of the last base pair in duplex DNA and,presumably, for distinguishing between DNAmolecules with 3′ and 5′ overhanging ends. δ′Econtains a Ser-Arg-Cys (SRC) ‘arginine finger’motif that promotes ATP hydrolysis in theneighboring γD subunit35,36, and δ′E is therefore likely to be involved in sensing when DNAis available for clamp loading.Mutant δ′E subunits were mixed with wildtype γBCD subunits and labeled δA subunits forFRET clamp binding and release assays (rawdata in Supplementary Fig. 2 online, summarized in Fig. 4a). The presence of mutantδ′E subunits had little effect on the extent ofclamp binding as measured by the FRET signal obtained by adding ATP. Clamp loadercomplexes that contain only single mutationsin the δ′E subunit were minimally deficient inDNA-dependent clamp release. We therefore checked whether the δ′Emutations could potentiate the effects of mutations in the γBCD subunitsthat also have weak effects on their own. We mixed the variant forms ofthe δ′E subunit with γBCD subunits containing a mutation (K161A) nearthe SRC helix and found that clamp loaders with these combined mutations showed a synergistic reduction in the ability of DNA to stimulateclamp release (Fig. 4a).We also measured the ATPase activity of clamp loader complexescontaining these mutant δ′E subunits (Supplementary Fig. 3 online).These assays demonstrate a small reduction in the ability of DNA toMutations in the ′E subunitThe δ′E subunit is present in only one copy in the complex, where it ispositioned at the far end of the clamp loader spiral relative to the subunits that make the closest contact with the clamp (Fig. 1). The functionof the δ′E subunit has been unclear, as it has low affinity for the β clampand does not bind nucleotide. The DNA recognition model under consideration here includes an important role for the δ′E subunit, as it is thelast of the AAA modules to make contact with DNA in the clamp loaderFigure 6 DNA competition assays with hairpin DNAs measured byfluorescence anisotropy. (a) 5′ overhangs. (b) 3′ and blunt overhangs.5′-labeled template DNA was excited at 550 nm with emission signalcollected at 580 nm. Primer-template DNA (100 nM) was prebound to 1 µMwild-type clamp loader in the presence of 1 µM wild-type β clamp and1 mM ADP-BeFx (1 mM ADP, 2 mM BeCl, 10 mM NaF); nuc, nucleotide.NATURE STRUCTURAL & MOLECULAR BIOLOGY VOLUME 12 NUMBER 2 FEBRUARY 2005187

2005 Nature Publishing Group http://www.nature.com/nsmbARTICLESFigure 7 Steric-fit mechanism of clamp loader specificity.stimulate nucleotide hydrolysis, and the three δ′E mutations exhibitedthe same order of severity in this assay as in the clamp release assay(R94E T121 R150E). The combination of K161A γBCD subunitswith these mutant δ′E subunits further reduced the DNA-dependentstimulation of ATPase activity. We also examined the affinity for DNAof clamp loaders containing δ′E mutations by monitoring changes influorescence anisotropy of labeled DNA in the presence of β clamp andADP-BeFx (Fig. 4b,c). The results are in agreement with the clamp releaseand ATPase data, showing the strongest effects for the R94E mutant,with reductions in affinity that were enhanced by the inclusion of theK161A γBCD subunit.In summary, point mutations in both the γBCD and δ′E subunits of theE. coli clamp loader gave rise to defects in DNA binding and the abilityof nucleic acid to stimulate ATPase activity and clamp release. In bothsubunits, mutations in basic residues that precede the SRC helix (Lys161in γBCD and Arg150 in δ′E) gave only moderate effects even when theseresidues were replaced with negatively charged side chains. Much moresevere effects were found when the residues mutated were clustered nearthe clamp-interacting and central helices (Arg98, Lys100, Arg105, Ser132and Arg133 in γBCD, and Arg94 and Thr121 in δ′E).Clamp loader DNA-binding specificityOur current model for the interaction of clamp loader, clamp and DNAsuggests that the clamp loader should have complementarity for duplexDNA with 5′ overhangs but that DNA with 3′ overhangs would not fitwithin the central chamber (Fig. 1). We therefore sought to compare theability of DNA duplexes with different kinds of overhangs to stimulateclamp release. Owing to their small size, annealed duplex oligonucleotides are often partially denatured at their ends. This could lead tononuniform properties and potentially misleading results in our assays.DNA molecules designed to form hairpin structures can minimize theseproblems by having the ‘primer’ and ‘template’ regions be a part of thesame molecule37. Clamp release assays with hairpin DNAs having 3′or 5′ overhangs of various lengths were examined (Fig. 5). The FRETdata show that oligonucleotide hairpins with 5′ overhangs stimulatedclamp release whereas hairpins with 3′ overhangs did not. A blunt-endedhairpin with no overhanging single-stranded region also stimulated188clamp release. In contrast, a DNA moleculewith a sequence unlikely to produce internalbase pairing (DNA expected to be singlestranded) did not.We measured the relative affinity of theclamp loader complex for these various DNAsby using a competition assay in which fluorescently labeled primer-template DNA wasprebound to the clamp loader complex in thepresence of β clamp and ADP-BeFx (Fig. 6).The ability of DNA hairpins with 5′ overhangsto displace the prebound DNA seems to correlate roughly with the length of the overhang:the hairpin DNA with the ten-nucleotide 5′overhang was the best competitor whereasthat with a one-nucleotide overhang was theleast effective.Notably, hairpin DNA with 3′ overhangswas unable to compete out the prebound DNA(Fig. 6b). The addition of a single overhangingnucleotide on the 3′ end of the duplex severelyaffected its ability to act as a suitable competitor for prebound primer-template. Because wedid not see evidence for clamp loader interaction with hairpin DNAs having 3′ overhangs, the five-base loop in thehairpin does not seem to interact with the clamp loader. Blunt-endedhairpin DNA was also able to compete away the labeled primer-template (Fig. 6b), consistent with the ability of blunt-ended duplexes to fitwithin the central chamber of the clamp loader. Mirroring the clamprelease results, the randomized (and presumably single-stranded) DNAwas completely unable to compete away the prebound primer-templateDNA. These competition data have been confirmed with another hairpin DNA molecule of different size, duplex and overhang compositionfrom the one described above (Supplementary Fig. 4 online).DISCUSSIONThe polymerase clamp loader complex can recognize and bind freeclamps and bring them to primed DNA in an ATP-dependent manner.The recognition of primed DNA results in the stimulation of the ATPaseactivity of the clamp loader and, because the interaction with the clampis ATP-dependent, this triggers the release of clamps around nucleicacid. The determination of the structure of the RFC complex boundto PCNA19 led to a markedly simple hypothesis for how this ejectionmechanism works: the spiral organization of the ATPase domains of theclamp loader recognizes DNA, but also configures the active sites forcatalysis by bringing interfacial subunits into close proximity aroundnucleotide. We report here that the clamp loader residues identifiedthrough analysis of the eukaryotic clamp loader are critical for DNArecognition in the bacterial clamp loader. This emphasizes the functionalconservation across evolution of the clamp loader mechanism.Previous work on clamp loader complexes has suggested that slidingclamps used for DNA replication are preferentially loaded onto DNAwith recessed 3′ ends16,29,38 but these results have not yet been reconciledwith the structural organization of the clamp loader. We show here usingDNA hairpins with a variety of overhangs that DNA binding and clamprelease by the E. coli clamp loader is promoted by primer-template DNAhaving 5′ overhangs or blunt ends, but not by those having 3′ overhangs.This suggests that DNAs containing elongation-proficient 3′ hydroxylscan bind most strongly to the clamp loader. Our structural model of aternary complex of clamp loader, sliding clamp and DNA is consistentwith these biochemical results because a blunt-ended piece of duplexVOLUME 12 NUMBER 2 FEBRUARY 2005 NATURE STRUCTURAL & MOLECULAR BIOLOGY

2005 Nature Publishing Group http://www.nature.com/nsmbARTICLESDNA would be expected to fit well into the DNA-binding region ofthe complex, whereas duplex DNA with a 3′ extension of the primerstrand would not. Notably, the E. coli clamp loader has been shown toload clamps onto forked DNA16 and EM studies of an archaebacterialclamp loader suggest that DNA may pass through the C-terminal collardomains39, indicating that the DNA recognition process might be moreversatile than described here.Our steric fit mechanism provides a potential explanation for how theclamp loader discerns the proper place on DNA to release sliding clamps(

BCD Arg105 Within clamp-interacting helix 92.6 R105E Reduced clamp binding, eliminated DNA binding γ BCD Ser132 Before central helix 85.1 S132A Eliminated DNA binding γ BCD Arg133 Within central helix 48.0 R133A, R133E Reduced DNA binding γ BCD Lys161 Before SRC-containing helix 94.2 K161A, K161E Reduced DNA binding δ′