Replication by a single DNA of a stretched single-stranded DNA

Berenike Maier*, David Bensimon, and Vincent Croquette†

Laboratoire de Physique Statistique, Ecole Normale Supe´rieure, Unite´Mixte de Recherche 8550 associe´au Centre National de la Recherche Scientifique et aux Universite´s Paris VI et VII, 24 rue Lhomond, 75231 Paris, France

Edited by Nicholas R. Cozzarelli, University of California, Berkeley, CA, and approved August 21, 2000 (received for review November 8, 1999) A new approach to the study of DNA͞protein interactions has been replication is thought to involve a conformational change (4, opened through the recent advances in the manipulation of single 5), from an open to a closed conformation of the fingers, where DNA molecules. These allow the behavior of individual molecular the bound dNTP is matched to its homologous base on the motors to be studied under load and compared with bulk mea- template (9). surements. One example of such a motor is the DNA polymerase, Despite this similarity, recent structural and kinetic studies which replicates DNA. We measured the replication rate by a single suggest that the different characteristics of these (rate, of a stretched single strand of DNA. The marked difference fidelity, and ) depend on subtle conformational between the elasticity of single- and double-stranded DNA allows differences within the enzyme͞template complex (10, 11). X-ray for the monitoring of replication in real time. We have found that and footprinting studies have shown that different the rate of replication depends strongly on the stretching force make subtly different contacts with the minor groove of the applied to the template. In particular, by varying the load we double-stranded (ds) DNA (2, 12). T7 DNAP appears to interact determined that the biochemical steps limiting replication are with three bases, whereas the Klenow fragment has been re- coupled to movement. The replication rate increases at low forces, ported to contact four bases. Whether these differences have decreases at forces greater than 4 pN, and ceases when the direct functional implications, for example in the fidelity or single-stranded DNA substrate is under a load greater than Ϸ20 pN. processivity of the enzymes, is not known. The decay of the replication rate follows an Arrhenius law and In this paper we will address this issue by using single molecule indicates that multiple bases on the template strand are involved experiments to study the replication on a template under tension in the rate-limiting step of each cycle. This observation is consis- of two functionally different polymerases: Sequenase and Kle- tent with the induced-fit mechanism for error detection during now. We shall see that the aforementioned structural differences replication. appear to correlate with movement of the enzymes during the rate-limiting step of their cycle. That motion is coupled to a molecular motors ͉ DNA elasticity displacement of at least n bases on the template (n ϭ 2 for Sequenase and n ϭ 4 for Klenow). The requirement, at the NA polymerases (DNAPs) are responsible for the synthesis rate-limiting step of replication, of complementarity between the Dof a new DNA strand on a single-stranded (ss) template (1). last n bases of the two strands provides a mechanism for error They play a key role in the replication, repair, and proofreading detection during replication (should one of these n bases not of DNA by catalyzing the addition of a complementary dNTP to match). the 3Ј end of the growing strand. The rate of replication of a Single molecule experiments have been used to study a DNAP, its fidelity, and its processivity (its uninterrupted asso- number of molecular motors [kinesin (13, 14), myosin (15), ciation with the template) are related to the enzyme’s function F1-ATPase (16), RNA polymerase (17), II (18), in vivo. and DNAP (19)]. In these studies, the variation of the enzymatic Typically, fast and processive enzymes are associated with rate with load has been instrumental in providing a unique quick replication. The simplest example of such an enzyme is T7 insight into biochemical steps that are coupled to movement. It DNAP coupled to thioredoxin (2), which may is important to note that even though single turnover resolution incorporate several thousand nucleotides at a rate of Ϸ300 bases was not always achievable, it was nevertheless possible to gain per second (b͞s) without dissociating from its template (3). A insight into the constituent steps of the enzymatic cycle. For mutant of T7 DNAP (Sequenase), lacking 28 amino acids and example, with RNA polymerase the fact that the transcription the associated activity, retains these properties (4) rate is constant at low template tension indicates that the and was studied in this report. rate-limiting biochemical step does not generate movement (17). Slower, less processive enzymes are usually responsible for In the case of topoisomerase II, the exponential decay with load repair (and have an auxiliary role in replication). Possessing of the rate of supercoil removal suggests that the rate-limiting 5Ј33Ј as well as 3Ј35Ј exonuclease activity, DNAP I from E. step in the reaction involves a movement of the enzyme by about coli is representative of this group of DNAPs (1). The 5Ј33Ј 1 nm, presumably during closure of the DNA gate (18). Finally, exonuclease activity is situated on the smaller fragment of the recent experiments (19), although lacking single base resolution, tri-domain enzyme and may be eliminated by . The were able to show from an analysis of the rate vs. force remaining large (Klenow) fragment, which was also studied here, has a low processivity (1Ϫ100 bases) and a slow repli- cation rate (15 b͞s) (5). As with T7 DNAP, the 3Ј35Ј This paper was submitted directly (Track II) to the PNAS office. exonuclease activity of the Klenow fragment may be abolished Abbreviations: DNAP, DNA polymerase; ss, single-stranded; ds, double-stranded; b͞s, bases by a mutation without significantly altering its polymerase per second. activity (6, 7). *Present address: Physikdepartment E22, TU Mu¨nchen, James-Franck-Strasse 1, 85748 Garching, Germany. Although T7 DNAP and E. coli DNAP I have different †To whom reprint requests should be addressed at: LPS-ENS 24 rue Lhomond, 75231 Paris functions and replication rates, they are structurally similar: cedex 05. E-mail: [email protected]. they resemble a right hand with the active polymerization site The publication costs of this article were defrayed in part by page charge payment. This in the palm, the dNTP-binding region in the fingers, and the article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. processivity factor on the thumb (8). The rate-limiting step in §1734 solely to indicate this fact.

12002–12007 ͉ PNAS ͉ October 24, 2000 ͉ vol. 97 ͉ no. 22 Downloaded by guest on September 23, 2021 drifts of the microscope are corrected by the simultaneous differential tracking of two beads: the bead attached to the DNA and an identical reference bead immobilized on the surface. To measure the force, F, we monitor the transverse Brownian fluctuations ͗␦x2͘ of the DNA-tethered bead and use the equi- ϭ ͗͞␦ 2͘ partition theorem to determine F kBTl x (23). To replicate the ssDNA, an 18-base primer specific to a single locus of the DNA template (pX⌬II) is first hybridized to the template, and DNAP is then injected. Because ssDNA and dsDNA have different extensions for a given force, measuring the length of the chain allows for monitoring of the synthesis of the second strand in real time. Once replication is complete, the newly synthesized strand may be displaced from the duplex by stretching the DNA with a force F Ͼ 50 pN [in 10 mM phosphate buffer (pH 8)]. The force vs. extension curve after ejection of the new strand matches the initial curve before replication (see Fig. 2b). This step breaks Fig. 1. Sketch of the experimental system and the replication experiment. the link between the bead and the surface about 50% of the time. (Left) The setup, consisting of a ssDNA bound at one extremity to a small magnetic bead and at the other to the surface of a capillary tube. Small Ͻ High Stringency Primer Hybridization. A primer (5Ј-GTA CCG permanent magnets pull on the bead with a force F 100 pN. F is monitored Ј through the fluctuations ␦x2 of the bead. The extension l of the molecule is TTA AGT CAG GTG-3 ; MWG-Biotech France) was chosen at monitored in real time. (Right) The steps during the synthesis of the comple- 1,807 bases from the end of the pX⌬II template. Hybridization mentary DNA strand. For small stretching force, the ssDNA has a small exten- was performed on a 6 pN-loaded template for 15 min in the ͞ ͞ sion. Hybridization of a specific primer near one end of the molecule allows polymerization buffer [10 mM Tris (pH 8) 5 mM MgCl2 25 the polymerase to start replication. The progression of the polymerase leads mM KCl͞100 ␮M of each dNTP͞1 mg/ml BSA͞1 ␮M primer]. to an increase of the molecule’s extension. When the entire molecule has been Note that hybridization on a template under tension (like replicated, we increase the stretching force to 50 pN for a short time. This hybridization at high temperature) selects against noncomple- destabilizes the newly synthesized strand, which unpairs from the template mentary sequences and controls the stringency of hybridization. and diffuses in the buffer. The ssDNA molecule is ready for a new run of ͞ replication. The capillary was then rinsed with 10 mM Tris (pH 8) 100 mM NaCl, after which Klenow fragment (Roche Molecular Bio- chemicals), Klenow fragment minus exo3Ј35Ј (New England dependence that T7 DNAP organizes two template bases in the Biolabs), or SEQUENASE Version 2.0 (Amersham Pharmacia) was polymerization site during each catalytic cycle (19). In the injected at Ϸ20 nM in the polymerization buffer. The hybrid- experiments reported here we use the distinct mechanical prop- ization stringency was assessed by using a nonspecific primer on erties of ssDNA and dsDNA to follow the replication by a single the same molecule and checking that it did not initiate poly- DNAP of a ssDNA template under external load. While we have merization under the same conditions. obtained independently the same results as Wuite et al. [that Sequenase organizes two template bases (19)], the different rate Low Stringency Hybridization. Highly stringent priming requires vs. force dependence for Klenow DNAP suggests that this time-consuming preparations on a template under tension. Because the ssDNA construct is fragile and the full replication of the 17-kb enzyme has to organize at least four template bases during each ⌬ catalytic cycle. pX II DNA by Klenow may last up to6h(seeFig.3a), a quicker protocol on a shorter molecule (an 11-kb charomid) with lower Materials and Methods stringency of priming was generally used to determine the force ⌬ dependence of the replication rate for Klenow. A primer (5Ј-ATG DNA and Sample. dsDNA [17-kb pX II a gift of J.-F. Allemand, Ј Ecole Normale Supe´rieure(20); 11-kb charomid a gift of O. GAG GCG GAT AAA GTT-3 ; MWG-Biotech France) was chosen at 1,000 bases from the end of the template. The hybrid- Hyrien, Ecole Normale Supe´rieure(21)] was differentially end- ization was performed for 5 min in 10 mM Tris (pH 8)͞5mM labeled by ligating DNA fragments (about 800 bp long), multiply MgCl ͞25 mM KCl͞100 ␮M each dNTP͞1 ␮M primer. Klenow labeled with either biotin or digoxigenin, to its extremities (22). 2 was then injected without rinsing at a concentration of Ϸ10 nM. As This DNA construct was denatured by heating in 2 mM NaOH the tension on the template was not high enough to ensure that no at 100°C for 5 min. The resulting ssDNA was immediately bound ␮ nonspecific priming occurred, it cannot be excluded that polymer- to streptavidin-coated superparamagnetic beads (4.5 m; Dynal, ization started at different sites. However we did check in a few Oslo) by a 5-min incubation in PBS. The beads were then diluted cases that results in these conditions did not differ significantly from 20-fold in the incubation buffer [10 mM phosphate buffer (pH ͞ ͞ ͞ those in high stringency conditions. All experiments were per- 8) 0.1% Tween 20 0.1 mg/ml fish sperm DNA 1mMNaN3] formed at 25 Ϯ 1°C. and incubated for about 20 min on an anti-digoxigenin-coated ϫ ϫ square glass capillary (1 1 50 mm; Vitrocom, Rockaway, NJ) Data Processing. Determination of the replication rate requires ϫ placed on top of a 63 immersion objective of an inverted measuring the elongation of the molecule vs. time: l(t). The microscope. The unbound beads were washed by flushing the number of nucleotides added is then given by: N(t) ϭ [l(t) Ϫ ͞⌬ BIOPHYSICS capillary tube with a few milliliters of incubation buffer. l(0)]Ntot l(F), where Ntot is the total number of replicated bases and ⌬l(F) is the total change in extension at force F. Force and Extension Measurement. ␴ ϭ ͌͗␦ 2͘ A strong magnetic field gradi- The Brownian fluctuations of the DNA’s extension 1 z ent was generated on the sample by focusing the field (Ϸ0.5 (which vary between 70 nm at 1 pN and 25 nm at 19 pN) limit tesla) of two small NdFeB magnets with appropriate pieces the resolution in a single replication run (to 500 bases at 1 pN and separated by Ϸ1.5 mm. By varying the distance between the 175 bases at 19 pN). The absolute size of the replication domain magnets and the capillary tube, a force of up to 100 pN can be (or the absolute position of the replication complex along the exerted on the magnetic bead and its tethered DNA molecule molecule) may be determined with an accuracy of ␦n Ϸ 1,000 (22) (see Fig. 1). The molecule’s extension, ͗z͘ϭl, is measured bases, set by the precision (␧ Ϸ 0.3 ␮m) in the absolute length ␦ ͞ ϭ ␧͞⌬ by real-time video analysis of the bead’s image (22). Mechanical measurements: n Ntot l(F).

Maier et al. PNAS ͉ October 24, 2000 ͉ vol. 97 ͉ no. 22 ͉ 12003 Downloaded by guest on September 23, 2021 To determine the local rate of replication v(t), the average over the fluctuations in extension must be determined. Because filtering l(t) would smear out pauses in replication and thus yield lower average replication rates, we chose to fit the elongation Ͻ Ͻ signal l(t)(0 t tmax) to a polygon of M vertices (see Fig. 4a). The rationale for that fit is that l(t) is a broken line: replication bursts correspond to rising segments, pauses to horizontal ones. ϭ ͞⌬ The number of vertices M tmax t was chosen to tolerate for an error smaller than 5% on the average rate (⌬t ϭ 10 s for Sequenase, 100 s for Klenow). The coordinates {ti, li}ofthe polygon vertices were adjusted to minimize the fitting error (␹2 Ϫ Ϸ⌬ test) (24) while maintaining ti tiϪ1 t. The average number of bases replicated between two adjacent vertices of the polygon ⌬ ϭ Ϫ ͞⌬ is then ni (li liϪ1)Ntot l(F) and the replication rate is: ϭ⌬ ͞ Ϫ v(ti) ni (ti tiϪ1). In this paper the mean and standard deviations of measured replication rates are given with respect to the number of individual molecules studied K and the number of independent measurements J. The error in the estimate of the ␴ mean rate ͗v͘ (at a given force) has two contributions: (i)a ␴ statistical error equal to the spread in the measured rates v ͌ ␴ divided by J (24) and (ii) a systematic error s resulting from the precision ␧ ϭ 0.3 ␮m in the conversion from measured lengths to number of replicated bases: ␴ ϭ͗v͘␧͞⌬l(F). (This s Fig. 2. Force vs. extension curves F(l) for ssDNA. (a) Comparison of the contribution is significant near the point where the extensions of ⌬ ␴2 ϭ ␴2͞ ϩ ␴2 elasticity of a charomid and pX II ssDNA in the polymerization buffer and the ss- and dsDNA are equal.) Thus ͗v͘ v J s. (For the case ϭ ͗ ͘ϭ ͞ ␴ ϭ known dsDNA elasticity curve (WLC, worm-like chain). Length was normalized of F 10 pN described in Fig. 4, we have v 121 b s, v with respect to the dsDNA extension. (The error in this normalization is Ϸ10% ͞ ϭ ␴ ϭ ͞ ␴ ϭ ͞ 31 b s, J 12, s 13b s, thus ͗v͘ 16 b s.) for the charomid and Ϸ5% for pX⌬II.) Notice that the elastic behavior of these The replication rate versus force curves were fitted to the two molecules is different below Ϸ 10 pN, consistent with the difference in model described in the following discussion by using a Leven- GϩC content of the molecules which influences the stability of secondary berg–Marquardt algorithm (24) to determine the two parame- structures at low forces. (b) Determination of the extent of polymerization. ᭛ ters of the model (v , n). Because of the nonlinearity of the fit Extension vs. force curves in polymerization buffer for dsDNA lds(F)( ), ssDNA 0 ⅜ ᮀ ϩ and the likelihood that the probability distribution of the pa- lss(F) before replication ( ), partially replicated DNA ( ), and ssDNA ( ) produced by briefly pulling on the replicated template with a force F Ͼ 50 pN rameters of the model is non-Gaussian, we have used a Monte- to eject the newly synthesized strand. The full line is the superposition l (F) ϭ Carlo bootstrap method (24) to estimate the error in the fitted p plds(F) ϩ (1 Ϫ p)lss(F) with percentage replication p ϭ 0.7 which fits the parameters and the associated confidence levels. measured points (ᮀ) remarkably well. Results and Discussion ϭ ϭ ⌬ The experiment is conducted in a small capillary placed above ssDNA and dsDNA, where lds lss,isatF 3.2 pN for a pX II an oil immersion microscope objective. The inlet and outlet of DNA and at F ϭ 4.9 pN for a charomid. Because the extension– the capillary are connected to small reservoirs permitting buffer force curves lss(F) for ssDNA exhibit a complex variation with changes. A ssDNA molecule is anchored at one end to the ionic strength, the measurement of this curve in the enzymatic bottom surface of the capillary and at the other to a small assay conditions is required before the study of replication. magnetic bead. As shown in Fig. 1, magnets are placed just above the capillary and generate a vertical force F. Motion of the Measurement of the Activity of a Single DNAP. The replication of magnets on a millimeter scale in the vertical (z) direction is used ssDNA into dsDNA as a function of time is measured by to control the force F applied on the molecule. At a fixed magnet monitoring the relative change in either extension or rigidity of position, the force F is constant. A computer image analysis is the system. Thus the extension vs. force behavior lp(F)ofa used to track the bead’s position x, y, z in real time. From these molecule which has been partially replicated, p being the per- data we determine the extension of the molecule l and the centage of replication, is intermediate between that of ssDNA stretching force F. The replication experiment is carried out by l (F) and dsDNA l (F) as shown in Fig. 2b for p ϭ 70%. A monitoring the extension l of the molecule while the polymerase ss ds partially replicated molecule is well described by the superpo- is converting ssDNA to dsDNA. ϭ ϩ sition of dsDNA and ssDNA elastic behavior: lp(F) plds(F) Ϫ (1 p)lss(F). Inverting this expression allows us to determine the The Difference in the Elastic Behavior of ssDNA and dsDNA Allows for ϭ Ϫ ͞⌬ percentage of replication p(t) (l(t) lss(F)) l(F) with Monitoring of the Activity of DNAP. The extension vs. force curve ⌬ ϭ Ϫ for ssDNA in the polymerization buffer, l (F), is shown in Fig. l(F) lds(F) lss(F). The number of incorporated nucleotides ss at time t, N(t), is then the total number of nucleotides times the 2 and compared with the analogous curve for dsDNA, lds(F). The ϭ Ϸ percentage of replication N(t) p(t)Ntot. When lds(F) lss(F)at elasticity of the latter is well described by the worm-like chain ϭ model of a polymer with a persistence length ␰ ϭ 50 nm (22, 25, F 4 pN, the molecule’s extension does not change noticeably 26). ssDNA exhibits a more complex behavior dependent on the on replication. However, as ssDNA is replicated into stiffer strength of the secondary structures formed. Thus l (F) cannot dsDNA, replication can still be measured by monitoring the ss Ѩ ͞Ѩ be described by a simple statistical model, such as the modified increase in the system stiffness, lp(F) z [which is inversely freely jointed chain model proposed by Smith et al. (27). proportional to the longitudinal fluctuations of the bead (22): Although this model describes the experimental results obtained ͗␦z2͘]. in 10 mM phosphate buffer at high forces reasonably well, it fails By using random primers, the ability of the enzymes to at low forces. Indeed, below 10 pN the elastic behavior of pX⌬II replicate a 17-kb ssDNA was checked. Initiating synthesis at (70% AϩT) is significantly different from that of a charomid arbitrary sites, the replication rate v(t) for Sequenase exhibited (50% AϩT), as is evident in Fig. 2a. The crossing point between bursts of replication followed by pauses. Complete replication

12004 ͉ www.pnas.org Maier et al. Downloaded by guest on September 23, 2021 Fig. 3. Time evolution of the replication on a pX⌬II ssDNA under a tension of 1 pN (specific priming). (a) Number of bases N(t) replicated by Klenow minus exo3Ј35Ј as a function of time in two successive replication runs on the same template. Notice that pauses in replication occur at different positions in successive runs. The dots represents the raw data after low-pass filtering at Ϸ1 Hz. The extension of the DNA is converted into the number of added nucle- otides N(t) as explained in the text. The full lines are polygon fits (see Materials and Methods) with an average time duration of 100 s. (b) N(t) for Sequenase in two runs on different molecules. The full lines are polygon fits with an average time duration of 10 s. Fig. 4. Measurement of the replication rate by Sequenase. (a) Time course of replication after high stringency hybridization on pX⌬II at 10 pN. The raw was confirmed by checking that the replicated molecule could be data (dots) were fitted by a polygon with M ϭ 40 vertices (shown as ⅜ in the fit by a worm-like chain model with the persistence length of 50 Inset) as explained in Materials and Methods.(b) Instantaneous replication nm expected for dsDNA. At the end of the replication cycle the rate (velocity) as determined from the slope of the polygon segments drawn in a.(c) The histogram of the velocities determined in b displays two peaks. The newly synthesized strand may sometimes be ejected without peak at v ϭ 0 Ϯ 8.6 b͞s results from pauses in replication. The broad peak with rupturing the bonds binding the bead to the surface (see Figs. 1 Ͼ positive rates results from replication bursts. The boundary between the two and 2) by briefly pulling on the molecule with a force F 50 pN peaks is shown by the dashed line at thrice the standard deviation of the (28). Complete ejection was confirmed by verifying that the force pauses peak, i.e., 26 b͞s. (d) The instantaneous replication rate histogram at vs. extension curve of the molecule after this brief pull was 16 pN (M ϭ 56). The peak of replication bursts has a mean value of Ϸ50 b͞s. identical to the one before replication, see Fig. 2. On hybrid- The boundary between the pauses and replication peaks is at 20 b͞s, lower ization of a new primer and addition of DNAP, a new round of than in c because of the smaller noise at this higher stretching force. replication can then be initiated on the same DNA molecule. As previously reported (4), the polymerization rate is se- quence-dependent. To verify this in our experiments, we set a cantly shorter for Sequenase [hundreds of seconds; see Fig. 3b], unique starting point for replication by annealing a specific than for Klenow (thousands of seconds), we do not have primer to the template under stringent hybridization conditions. sufficient data to analyze these events more precisely, especially Because of the intrinsic thermal fluctuations (see Materials and because pauses are comparable in duration to the lifetime of Methods) at low force or for a low processivity enzyme, it is polymerase activity. difficult to ensure that the same polymerase is monitored Fig. 3b shows N(t) for Sequenase under a 1-pN load, whereas throughout the experiment. However, with specific priming we Fig. 4a corresponds to a 10-pN stretching force. In all cases, N(t) can be certain that only a single enzyme is active at a given time. displays a typical broken line shape with some added noise. We For the pX⌬II template we achieved specific priming of an thus fitted N(t) to a broken line with slightly varying time ⌬ 18-base primer at 1,807 bases from the 3Ј end by stretching the intervals, corresponding to an average integration time t. While template with a 6-pN force (in 5 mM MgCl2 and 25 mM KCl ionic significantly reducing the noise, this algorithm allows for abrupt conditions). changes in replication rate as observed between burst to pause ⌬ A striking feature of the replication by a single enzyme is the phases. The choice of the integration time, t, is a tradeoff existence of long pauses of varying duration (up to hours) and between a large ⌬t ensuring low noise in the replication rate and BIOPHYSICS position along the molecule. Fig. 3a shows the number of a small ⌬t allowing for the observation of short pauses. As F is replicated nucleotides, N(t), in two successive replication runs on increased, the experimental noise decreases and consequently ⌬t the same template under a 1-pN load for Klenow minus can be reduced. The replication rates in Fig. 4b are obtained by exo3Ј35Ј. Notice that replication on the same molecule doesn’t plotting the slope of the fitted segments in Fig. 4a with ⌬t ϭ 8s. always pause at the same sites. Hence these long pauses are The histogram of these replication rates (at 10 pN) is shown in apparently not due to defective bases on the template (e.g., Fig. 4c. A peak is observed at v ϭ 0 with a standard deviation ␴ ϭ ͞ depurination). We believe that they are a result of fluctuating 0 8.6 b s corresponding to the pauses and end of replication. hairpins, which may act as road blocks for the proceeding A second broad peak corresponds to the replication rates during polymerase (29). Although the duration of pauses was signifi- burst activity. To measure the polymerase properties, we disre-

Maier et al. PNAS ͉ October 24, 2000 ͉ vol. 97 ͉ no. 22 ͉ 12005 Downloaded by guest on September 23, 2021 ␴ gard pauses by considering only rates greater than 3 0 (here 26 ͞ ␴ b s). 0 decreases when the template stretching force F is increased as may be seen in the histogram in Fig. 4d obtained at F ϭ 16 pN. To compare with bulk experiments, we have measured the mean replication rate at low forces (F Ϸ 1 pN). As expected, the mean replication rate of Sequenase (210 Ϯ 10 b͞s; K ϭ 3, J ϭ 46) is higher than for Klenow (6 Ϯ 1b͞s; K ϭ 9, J ϭ 114) and Klenow minus exo3Ј35Ј (7 Ϯ 1b͞s; K ϭ 4, J ϭ 50). The value for Sequenase is consistent with its known activity (3). For Klenow, slightly higher values of 15 b͞s have been reported in single turn-over experiments (5). These data are, however, similar to the maximal replication rates measured here: 13.5 Ϯ 1.5 b͞s for Klenow at a load of F Ϸ 4 pN. This medium stretching force might help to suppress the possible secondary structures in the rather long ssDNA molecule used here.

Force Dependence of the Replication Rate. To study the force dependence of the replication rate v(F), we used a high strin- gency hybridization protocol with Sequenase. The replication rates measured at 1 and 2 pN with low stringency priming were slightly lower than with the high stringency protocol (see Fig. 5c) which may be attributed to the different (GϩC richer) template used. The mean replication rate of the two studied polymerases is strongly affected by the stress on the template (Fig. 5). Klenow stalls at a template stretching force of about 20 pN. Thus, at an initial force of 3 pN, the enzyme exhibits normal activity (Fig. 5a). Increasing the force to 24 pN inhibits polymerization, which can be reinitiated by reducing the stress back to 3 pN. Note that exposing the molecule to a 24-pN strain does not eject the newly synthesized strand, as the number of incorporated nucleotides does not decrease. Below the stalling force, polymerase slows down as the load is increased: Sequenase, for example, works slower at 16 pN than at 1 pN (Fig. 5b) and stalls at a force of 26 Ϯ 6pN(K ϭ 8, data not shown). For both enzymes, in the range of 0 Ͻ F Ͻ 4 pN, the mean rate increases with increased force, whereas above 4 pN it decreases exponentially with the force as the enzyme has to work against the load on the template (Fig. 5 c and d). How can the exponential decay of the mean replication rate at Fig. 5. Force dependence of the replication rate. (a) Replication by Klenow on a charomid is started at 3 pN. Increasing the force to 24 pN stops the high forces be explained? One might expect the load to interfere replication. After reducing the force back to 3 pN, the polymerization degree, with the binding of the enzyme to its template or with a N, has not changed and the enzyme starts replicating again. (b) Number of rate-limiting step of the enzymatic cycle. Because DNA is known bases N(t) replicated by Sequenase on pX⌬II at two different forces: 1 pN and to be bent by Ϸ80° to fit into the binding pocket of the 16 pN. (c) Mean replication rate, ͗v͘, versus force for Sequenase: high strin- polymerase (10, 12), stretching might indeed lower the affinity gency hybridization (K ϭ 13, J ϭ 154; ⅜) and random priming (K ϭ 5, J ϭ 82; ᭛ ␴ of the enzyme for its substrate and a higher enzymatic concen- ). The error bars represent ͗v͘ estimated as explained in Materials and tration might significantly increase the replication rate. To assess Methods. The full curve is a fit to the model described in the text ͗v(F)͘ϭ v0exp(ϪnF⌬h͞kBT), where v0 ϭ 200 b͞s and n ϭ 2.1 (only ⅜ were fitted). The this effect, the mean replication rate was measured at 15 pN for ϭ two concentrations of Sequenase: 20 and 200 nM. The mean rate dashed curve is obtained with the previously determined v0 and n 1 (as ϭ Ϯ ͞ ϭ ϭ ϭ Ϯ expected if only one base was rate determining). (d) Replication rate versus was v 37 5b s(K 2, J 54) at 20 nM and v 59 8 ϭ ϭ ᭛ ͞ ϭ ϭ force for Klenow: low stringency hybridization (K 24, J 298; ), high b s(K 2, J 32) at 200 nM. As these rate are of the same stringency hybridization (K ϭ 9, J ϭ 114; ⅜). Replication rate for Klenow minus magnitude despite a 10-fold difference in concentrations, we exo3Ј35Ј high stringency hybridization (K ϭ 4, J ϭ 50; ᮀ). The full curve is a deduce that for the enzymatic concentration used in our exper- fit to ᭛, where v0 ϭ 13.5 b͞s and n ϭ 4. iments (Ϸ20 nM) the binding efficiency was not significantly affected by the stretching force. Ϫ ⌬ ͞ In the following we shall therefore assume that the enzyme v0exp( nF h kBT) and determine v0 and n (Fig. 5 c and d). For performs an amount of work, F␦, while contracting the template Sequenase the best fit values (see Materials and Methods) are v0 by a distance ␦ against the force. The work performed during a ϭ 200 Ϯ 20 b͞s and n ϭ 2.1 Ϯ 0.5 (confidence level: P ϭ 95%), ϭ Ϯ ͞ ϭ rate-limiting step of the enzymatic cycle reduces the replication whereas for the Klenow fragment: v0 13.5 1.5 b s and n Ϫ ␦͞ ⌬ ϭ Ϯ ϭ rate by an Arrhenius factor exp( F kBT) (17, 30). Let h (lss 4.0 0.7 (P 95%). The values of v0 are consistent with the Ϫ ͞ ͞ lds) Ntot be the difference in the distance (at a given force F) published values for Sequenase and 15 b s for Klenow (4, 5). between successive bases in ssDNA and dsDNA (see Fig. 2b). As Moreover, results for v0 and n similar to the one for Sequenase this represents the largest length change per base (i.e., contrac- were recently obtained on wild-type T7 DNAP by Wuite tion for F Ͼ 4 pN) in our experiment, we define by n ϭ ␦͞⌬h et al. (19). the minimum number of contracted bases during the polymer- The values obtained for n are interesting because they relate ization rate-limiting step. We thus fit the measured mean to the minimal number of bases implicated in the replication replication rate to the following Arrhenius law: ͗v(F)͘ϭ rate-limiting step. Obviously, when a new nucleotide is added to

12006 ͉ www.pnas.org Maier et al. Downloaded by guest on September 23, 2021 the growing strand, the ssDNA has to contract to fit into the 31]. Thus, close interaction with 3 and 4 nucleotides of the tem- dsDNA structure. Thus one might have expected n ϭ 1tocome plate are reported for T7 DNAP and Klenow, respective- out naturally from a fit to the data. However, as can be seen from ly (2, 12), which is close to the values of n estimated in our Fig. 5 c and d, the predicted decay of the replication in that case experiments. does not fit our data. The hypothesis that a single nucleotide (the one added) is singly participating in the rate-limiting step can be Conclusions rejected with confidence P Ͼ 99% [Monte-Carlo bootstrap We have shown that the replication by a single DNAP can be method (24)]. Moreover, notice that the values of n determined followed due to the different elastic behavior of ssDNA and here are actually lower bounds as they assume the greatest possible conformational change: that between stretched ss- and dsDNA. The enzyme can convert part of the energy gained from dsDNA resulting from complete fraying of the dsDNA end over the hydrolysis of dNTP to work against an external force applied n bases. As this is unlikely to happen, the values of n imply that on the template (17). Both Sequenase (a replicase) and Klenow more nucleotides on the template strand than the one being (a repair enzyme) show similar behavior when their template is replicated are involved in determining the replication rate. stretched by an external force. The rate-limiting step in their This result is consistent with the induced-fit kinetic mecha- enzymatic cycle involves more than just the one nucleotide being nism for error detection during DNA replication (4, 5). Accord- incorporated. Although they perform different functions, both ing to this mechanism, the rate-limiting step in replication is due enzymes belong to the DNAP I family and are thus structurally to a conformational change in the enzyme–DNA complex related. It will be interesting to see whether there is a correlation characterized by a transition between a loose and tightly fit between n, the number of bases involved in the replication rate configuration of the two DNA strands. An error in the insertion limiting step, and the number of bases in close interaction with of the last n bases skews the equilibrium to the ‘‘loose’’ config- the polymerase as seen in the structural data. uration, slowing down replication and favoring error correction by the native DNAP exonuclease activity. Similarly in our We thank J.-F. Allemand, H. Buc, N. Crisona, N. Dekker, O. Hyrien, and experiment, by pulling on the template strand the equilibrium T. Strick for their suggestions and the preparation of DNA; C. Busta- should be skewed to the ‘‘loose’’ configuration, thus reducing the mante, S. Smith, G. Wuite, and D. Keller for discussing their results forward rate. Structural data indeed show that the polymerase before publication; and E. Sackmann and J. Ra¨dler for their support. has a strong interaction with the dsDNA minor groove over a B.M. acknowledges the support of a Deutscher Akademischer Austaus- length of several base pairs [with respect to the 3Ј end; refs. 2 and chdienst scholarship.

1. Kornberg, A. & Baker, T. (1992) DNA Replication (Freeman, San Francisco). 19. Wuite, G. J. L., Smith, S. B., Young, M., Keller, D. & Bustamante, C. (2000) 2. Kunkel, T. A. & Wilson, S. H. (1998) Nat. Struct. Biol. 5, 95–99. Nature (London) 404, 103–106. 3. Tabor, S., Huber, H. E. & Richardson, C. C. (1987) J. Biol. Chem. 262, 20. Cluzel, P., Lebrun, A., Heller, C., Lavery, R., Viovy, J.-L., Chatenay, D. & 16212–16223. Caron, F. (1996) Science 271, 792–794. 4. Patel, S. S., Wong, I. & Johnson, K. A. (1991) Biochemistry 30, 511–525. 21. Saito, I. & Stark, G. R. (1986) Proc. Natl. Acad. Sci. USA 83, 8664–8668. 5. Dahlberg, M. E. & Bencovic, S. J. (1991) Biochemistry 30, 4835–4843. 22. Strick, T., Allemand, J.-F., Bensimon, D. & Croquette V. (1998) Biophys. J. 74, 6. Derbyshire, V., Freemont, P. S., Sanderson, M. R., Beese, L., Friedman, L. M., 2016–2028. Joyce, C. M. & Steitz, T. A. (1988) Science 240, 199–201. 23. Strick, T., Allemand, J. F., Bensimon, D., Bensimon, A. & Croquette, V. (1996) 7. Eckert, K. A. & Kunkel, T. A. (1993) J. Biol. Chem. 268, 13462–13471. Science 271, 1835–1837. 8. Ja¨ger, J. & Pata, J. D. (1999) Curr. Opin. Struct. Biol. 9, 21–28. 24. Press, W. H., Teukolsky, S. A., Vettering, W. T. & Flannery, B. P. (1992) 9. Doublie, S. & Ellenberger, T. (1999) Curr. Opin. Struct. Biol. 8, 704–712. Numerical Recipes in C 10. Doublie, S., Tabor, S., Long, A. M., Richardson, C. C. & Ellenberger, T. (1998) (Cambridge Univ. Press, Cambridge, U.K.), pp. Nature (London) 391, 251–258. 609–699. 11. Wong, I., Patel, S. S. & Johnson, K. A. (1991) Biochemistry 30, 526–537. 25. Baumann, C., Smith, S., Bloomfield, V. & Bustamante, C. (1997) Proc. Natl. 12. Beese, L. S., Derbyshire, V. & Steitz, T. A. (1993) Science 260, 352–355. Acad. Sci. USA 94, 6185–6190. 13. Hua, W., Young, E. C., Fleming, M. L. & Gelles, J. (1997) Nature (London) 388, 26. Bouchiat, C., Wang, M. D., Block, S. M., Allemand, J.-F., Strick, T. R. & 390–393. Croquette, V. (1999) Biophys. J. 76, 409–413. 14. Visscher, K., Schnitzer, M. J. & Block, S. M. (1999) Nature (London) 400, 184–189. 27. Smith, S. B., Cui, Y. & Bustamante, C. (1996) Science 271, 795–799. 15. Finer, J. T., Simmons, R. M. & Spudich, J. A. (1994) Nature (London) 368, 113–119. 28. Rief, M., Clausen-Schaumann, H. & Gaub, H. E. (1999) Nat. Struct. Biol. 6, 16. Yasuda, R., Noji, H., Kinosita, K. & Yoshida, M. (1998) Cell 93, 1117–1124. 346–349. 17. Wang, M. D., Schnitzer, M. J., Yin, H., Landick, R., Gelles, J. & Block, S. (1998) 29. Bierne, H. & Michel, B. (1994) Mol. Microbiol. 13, 17–23. Science 282, 902–907. 30. Keller, D. & Bustamante, C. (2000) Biophys. J. 78, 541–556. 18. Strick, T. R., Croquette, V. & Bensimon, D. (2000) Nature (London) 404, 31. Baker, T. A. & Bell, S. P. (1998) Cell 92, 295–305. 901–904. BIOPHYSICS

Maier et al. PNAS ͉ October 24, 2000 ͉ vol. 97 ͉ no. 22 ͉ 12007 Downloaded by guest on September 23, 2021