Branched unwinding mechanism of the Pif1 family of DNA helicases

Significance

DNA helicases are motor proteins that operate on nucleic acids during DNA replication, recombination, and repair. Repetitive translocation on and unwinding of nucleic acids is an emerging property of multiple helicases. How this repetitive enzymatic activity may be used for function remains unclear. For the 5′-3′ Pif1 family of helicases our data suggest a model in which a highly processive opening of short stretches of DNA is all that this class of nonreplicative helicases need to achieve when their function is coupled to DNA replication.

Abstract

Members of the Pif1 family of helicases function in multiple pathways that involve DNA synthesis: DNA replication across G-quadruplexes; break-induced replication; and processing of long flaps during Okazaki fragment maturation. Furthermore, Pif1 increases strand-displacement DNA synthesis by DNA polymerase δ and allows DNA replication across arrays of proteins tightly bound to DNA. This is a surprising feat since DNA rewinding or annealing activities limit the amount of single-stranded DNA product that Pif1 can generate, leading to an apparently poorly processive helicase. In this work, using single-molecule Förster resonance energy transfer approaches, we show that 2 members of the Pif1 family of helicases, Pif1 from Saccharomyces cerevisiae and Pfh1 from Schizosaccharomyces pombe, unwind double-stranded DNA by a branched mechanism with 2 modes of activity. In the dominant mode, only short stretches of DNA can be processively and repetitively opened, with reclosure of the DNA occurring by mechanisms other than strand-switching. In the other less frequent mode, longer stretches of DNA are unwound via a path that is separate from the one leading to repetitive unwinding. Analysis of the kinetic partitioning between the 2 different modes suggests that the branching point in the mechanism is established by conformational selection, controlled by the interaction of the helicase with the 3′ nontranslocating strand. The data suggest that the dominant and repetitive mode of DNA opening of the helicase can be used to allow efficient DNA replication, with DNA synthesis on the nontranslocating strand rectifying the DNA unwinding activity.

Growing experimental evidence indicates that multiple helicases/translocases display a peculiar behavior in which a single motor protein molecule can repetitively translocate on nucleic acid (1–6) or unwind double-stranded DNA (7–11) and G-quadruplex structures (6, 12–14). How this repetitive enzymatic activity is used for function remains unclear, but it may be linked to the specific pathway in which these helicases operate, the nature of the substrate itself, and the directionality of unwinding. Repetitive unwinding activity may be of particular significance for 5′-3′ helicases, such as those of the SF1B Pif1 family (15, 16), whose activity may be coupled to DNA replication. Pif1 from Saccharomyces cerevisiae, the best studied of this class of helicases and its founding member, displaces telomerase from telomeric ends (17–22), facilitates DNA replication at G-quadruplex–forming sequences during lagging strand DNA synthesis (23–26), cooperates with Dna2 helicase/nuclease in long flap processing during Okazaki fragment maturation (27–31), and functions with DNA polymerase δ during break-induced DNA replication (32–35). Similarly, Pfh1, the Pif1 homolog in Schizosaccharomyces pombe, functions in many of the same pathways and has biochemical activities similar to Pif1 (36–42).

Single-molecule experiments showed that Pif1 can repetitively reel in single-stranded DNA (ssDNA) or unwind G-quadruplex substrates (6, 12, 43). Repetitive unwinding of G-quadruplexes has been proposed to be a means to keep the formation of these structures under check (6). Also, once stable G-quadruplexes are formed, these can pose a significant obstacle to DNA replication, and the unwinding activity of Pif1 or Pfh1 is needed to allow significant DNA synthesis (44). Intramolecular G-quadruplexes are compact secondary structures held by a well-defined number of Hoogsteen base pairs, and limited unwinding by the helicase may be sufficient to lead to their full opening. However, the situation can be different for double-stranded DNA (dsDNA), where, depending on the length of the duplex region, repetitive helicase activity may not lead to full unwinding of the substrate. Indeed, the intrinsic dsDNA unwinding activity of Pif1 is counteracted by DNA rewinding or annealing activities (45, 46) that limit the amount of ssDNA product generated and lead to an apparently poor helicase. Moreover, under applied force, Pif1 displays a complex unwinding mechanism in which dsDNA unwinding is counteracted by either the helicase sliding back on the substrate or strand-switching and translocating back (47). In the absence of applied force, it remains unclear how these different activities contribute to the dsDNA unwinding mechanism of Pif1. In light of all these activities that compete with dsDNA unwinding, it is surprising that, when coupled to DNA replication, the helicase activity of Pif1 is enough to lead to kilobases of DNA synthesis and displacement of proteins tightly bound to dsDNA (32, 33, 48). It is possible that DNA synthesis on the nontranslocating strand may be a means to suppress DNA rewinding by Pif1, favoring efficient unwinding and removal of protein blocks (45, 48). In this scenario, one may expect that, when coupled to DNA synthesis, Pif1 does not need to operate as a processive helicase that can open large amounts of DNA. Rather, efficient and processive opening of short stretches of DNA is all that this class of helicases may need to achieve to allow synthesis on the 3′ nontranslocating strand to proceed efficiently. If this were the case, one would expect the DNA unwinding mechanism of these helicases to be dominated by this latter mode of unwinding, rather than by processive unwinding in sensu stricto. In this work, we set out to test this possibility and used single-molecule Förster resonance energy transfer (smFRET) approaches to study in real time the unwinding mechanism of S. cerevisiae Pif1 and S. pombe Pfh1.

Results

Individual Molecules of Pfh1 Exhibit Repetitive Attempts of Partial dsDNA Unwinding, Composed of Multiple ATP-Dependent Steps.

In ensemble experiments, ATPase active full-length Pfh1 helicase from S. pombe has limited intrinsic dsDNA unwinding activity in the absence of a trap to prevent reannealing of the unwound strands (SI Appendix, Fig. S1). This behavior is similar to the one reported for Pif1 (45, 46) and is consistent with a rewinding activity counteracting unwinding (45). In order to study the interplay between these 2 activities in real time, based on our previous ensemble work (45) we designed smFRET experiments using DNA substrates labeled with donor-acceptor couples at different positions (SI Appendix, Fig. S2). The substrates contain a short biotin-labeled 18-bp dsDNA handle for attachment to a pegylated surface via a biotin-neutravidin-biotin sandwich, a 30-bp dsDNA region to be unwound, a 5′ oligo-dT₁₀ tail as an entry point for the helicase, and variations in the 3′-tail (Fig. 1A). In the experiments, Pfh1 is loaded onto surface immobilized DNAs, and excess enzyme is removed by extensive washing. Upon addition of ATP, but not its nonhydrolyzable analogs ATPγS or AMP-PNP, the Cy3 and Cy5 intensities show clear anticorrelated changes and the corresponding FRET signal transitions between high and low values, consistent with Pfh1 opening the dsDNA without full unwinding (Fig. 1B). Indeed, at 10 μM ATP with this DNA substrate that contains a 3′-dT₂₂ tail only 7% of the DNA molecules show evidence of unwinding (SI Appendix, Fig. S3A), while the majority undergo a series of repetitive transitions before Pfh1 dissociates or the fluorophores photobleach. This behavior does not arise from additional Pfh1 molecules bound to the 22-nucleotide (nt)-long 3′-tail, as shown in experiments where the 3′-tail is substituted with either a 21-bp duplex region or a 3′-dT₂ tail (SI Appendix, Fig. S3 C and D). Furthermore, immobilization of Pfh1 on the surface reveals that the observed ATP-dependent FRET transitions originate from a Pfh1 monomer (SI Appendix, Fig. S3E). Importantly, ATP-dependent FRET transitions, from low to high, are observed also with a DNA substrate in which Cy5 is positioned 20 bp into the duplex region (SI Appendix, Fig. S3B), confirming that the dsDNA is being partially unwound. Analyses of experiments with Cy3 and Cy5 placed at different positions within the substrate or only Cy3 positioned 20 bp into the duplex strongly suggest that the processive and repetitive unwinding is limited to short stretches of dsDNA (<20 bp) (SI Appendix, Fig. S4) and that this is a property that originates from the helicase core of the enzyme (SI Appendix, Fig. S5).

Fig. 1.

Individual Pfh1 molecules display repetitive attempts of partial dsDNA unwinding. (A) Schematic of the smFRET experiment with D-A at the junction. (B) Representative anticorrelated changes in Cy3 and Cy5 intensities and corresponding FRET changes after starting the reaction with addition of ATP. For clarity, a 15-point moving average is overlaid with the raw data. (C and D) Distribution of FRET states visited during repetitive unwinding of the substrate with D-A at the junction in A or 20 bp apart in D. Solid lines are the fits to a Gaussian distribution. The white line indicates the distribution of FRET values before addition of ATP. (E) ATP dependence of the difference in mean FRET values of the closed and open states. (F–I) Distribution of the times spent in the different states during the repetitive cycle for the substrate with D-A at the junction and 10 μM ATP. The lines are fits to either a single exponential function or a gamma function with 2 equal rates. (J) Minimal kinetic model that combines the information from each step during the repetitive cycle of partial unwinding.

Repetitive unwinding/translocation activity on DNA, monitored by FRET and/or protein-induced fluorescence enhancement changes, has been reported for multiple helicases/translocases (1, 3, 4, 7, 8, 10, 13, 14, 49), including Pif1 (6, 12, 43, 47). The repetitive cycling behavior has been shown to be ATP dependent, as expected for a phenomenon driven by a motor protein; however, how ATP affects the multiple steps within the cycle has not been extensively characterized. To this end, we note that, during the repetitive cycles of partial DNA unwinding, the system appears to spend most of its time in 4 states: 2 well-defined states associated with high and low transfer efficiency and 2 transitions (from high to low and from low to high FRET) that connect these states. We identify these states by clustering the data based on their Euclidean distance, and, for each molecule, we estimate the transfer efficiency of the closed and open states, the dwell time in these 2 states, and the duration of the unwinding or rewinding transitions (SI Appendix, Fig. S6).

For the DNA construct with the donor-acceptor (D-A) pair at the junction, analysis of 225 molecules and ∼4,900 transitions shows that at 10 μM ATP the system transitions from a closed state with mean transfer efficiency (E = 0.77 ± 0.09) to an open state with a lower mean transfer efficiency (E = 0.4 ± 0.1) (Fig. 1C). Importantly, the system transitions from the open state back to a closed state that is the same as the Pfh1-DNA complex before addition of ATP (E = 0.78 ± 0.08), indicating that the entire dsDNA region that was opened has been reclosed. The same is true when the donor and acceptor are 20 bp apart (Fig. 1D). Moreover, independent of whether FRET is monitored with the acceptor at the junction or at 20 bp into the dsDNA, the difference in the average FRET value between the closed and open states does not change as a function of ATP concentration (Fig. 1E), indicating that on average the system opens the same number of base pairs before returning to its original closed state.

The times spent in the closed and open states are exponentially distributed (Fig. 1 F and G, and SI Appendix, Fig. S7 A and B), indicating that the system escapes from a single state, with no evidence of hidden intermediates. In contrast to the times spent in the closed and open states, the times of unwinding are not exponentially distributed. Rather, a gamma function with 2 steps of equal rate is sufficient to describe the distribution of the unwinding times (Fig. 1H and SI Appendix, Fig. S7C), independent of the position of the donor-acceptor couple. Such a distribution of times is indicative of the presence of an intermediate unwound state that precedes formation of the open complex. Similarly, the rewinding times do not appear to follow an exponential distribution (Fig. 1I and SI Appendix, Fig. S7D), suggesting the presence of a hidden intermediate. However, these events approach the time resolution of recording (30 Hz), and their number may be underestimated. Thus, assuming that the events in the first 2 time bins are within the time resolution limit of the measurement, the distribution of the rewinding times was fitted to an exponential function. The calculated rewinding rates show little ATP dependence (SI Appendix, Fig. S7H), suggesting that either the rewinding step is ATP independent or, if there is an ATP dependence, it must occur at lower ATP concentrations. Combining the information from analysis of the ATP dependence of each individual phase during the cycle of opening attempts (SI Appendix, Fig. S7 E–G) leads to the minimal kinetic model depicted in Fig. 1J.

Finally, we performed the same measurements for the Pif1 helicase (SI Appendix, Figs. S8 and S9). Pif1 behaves similarly to Pfh1 with the difference that the lifetimes of the different states appear to be slightly shorter. Shorter lifetimes of the closed and open state suggest either a higher affinity of Pif1 for ATP or a faster rate of escape to the next step. Also, the unwinding times for Pif1 are slightly faster than for Pfh1, suggesting that Pif1 is a faster helicase. However, at difference with Pfh1, the rewinding times for Pif1 appear to have some ATP dependence, as would be expected if this step were driven by the motor protein activity of Pif1.

During Rewinding Neither Pfh1 nor Pif1 Strand-Switch.

Strand-switching during unwinding has been proposed for different DNA helicases as a mechanism that leads to rezipping of the opened dsDNA (7, 11, 50). Recent single-molecule work, under applied force, provides evidence of a complex mechanism of unwinding by Pif1 and suggests that during unwinding Pif1 can switch to the opposite DNA strand, leading to DNA rewinding (47). Because this mechanism depends on the translocation activity of Pif1 on ssDNA, an ATP dependence is expected. Based on our observation that, for Pfh1, the rewinding step from the open to the closed state does not show an ATP dependence, we conclude that, for Pfh1, strand-switching does not significantly contribute to this step. The same may not be the case for Pif1, since it shows some ATP dependence for rewinding.

In order to provide a more direct test of whether strand-switching may contribute to rewinding of the DNA during the cycle of partial unwinding, we designed a simple experimental strategy. Our idea is based on the fact that Pif1 can unwind RNA-DNA hybrids only when the strand used for translocation is DNA and not RNA (51), indicating that the RNA strand cannot be used for translocation. Because the basic tenet of a strand-switching mechanism is translocation of the helicase on the opposite strand, one would expect that, if this strand were RNA, no FRET transitions should be observed. Therefore, we performed experiments with both Pfh1 and Pif1 and RNA-DNA hybrids with Cy3-Cy5 at different positions along the substrate. Independent of the position of the donor-acceptor couple, FRET transitions can be clearly observed for both Pfh1 and Pif1 (Fig. 2 A–D), as well as for a Pfh1 variant missing its first 291 amino acids (SI Appendix, Fig. S5C). These data provide direct experimental evidence that under our assay conditions strand-switching is not the major reason why, during the cycle of partial unwinding, the DNA rezips.

Fig. 2.

Neither Pfh1 nor Pif1 strand-switch. (A–D) Representative examples of repetitive FRET transitions observed at 10 μM ATP for both Pfh1 and Pif1 with RNA-DNA hybrids and the D-A couple at the indicated positions. (E) ATP dependence of lifetime in the different stages of the repetitive FRET transitions for Pfh1 and the DNA-RNA hybrid compared to dsDNAs with different 3′-tails.

Interestingly, for Pfh1 and the RNA-DNA hybrid with the donor-acceptor at the junction, the ATP dependence of the lifetime in the different states of the repetitive cycle of FRET transition is similar to the one observed for dsDNA and a 3′-dT₂₂ tail (Fig. 2E). This result strongly suggests that the 3′-tail of the substrate does not have a significant role in regulating the kinetics of repetitive transition from the closed to the open states. This point is further reinforced by the observation that neither the length (from 5 to 22 nt) nor the nature (oligo-dT, mixed sequence, dsDNA) of the 3′-tail significantly modulates the kinetics of the repetitive FRET transitions (Fig. 2E).

The Repetitive DNA Opening Attempts Are on a Separate Pathway from Full Unwinding.

Although the 3′-tail of the substrate does not appear to modulate the kinetics of the repetitive FRET transitions, we found that, in the presence of a 3′-tail of mixed sequence composition (3′-ss₂₂), the fraction of molecules that is fully unwound by Pfh1 increases. For example, at 10 μM ATP of 420 DNA molecules that show Pfh1-dependent activity, 58% show evidence of full unwinding of the dsDNA (Fig. 3 A and B). A significant fraction of full unwinding occurs also with a 3′-ss₅ tail, in stark contrast to the limited unwinding observed with a 3′-dT₂₂ tail (Fig. 3B). Also, an effect of the nature of the 3′-tail of the substrate is evident in experiments monitoring unwinding via disappearance of Cy3 signal from DNA substrates immobilized on a surface (SI Appendix, Fig. S10 A and B). This is the same phenomenon that we reported for Pif1 from ensemble studies (45), corroborated here by single-molecule experiments (SI Appendix, Fig. S10C), and indicates that this is a general property of members of the Pif1 family of helicases.

Fig. 3.

Partitioning between partial dsDNA opening and full unwinding. (A) Representative repetitive FRET transition observed at 10 μM for Pfh1 and a substrate with D-A at the junction and a 3′-tail composed of 22 nt of mixed sequence compositions. Full unwinding events can occur either after repetitive attempts (class 1) or directly (class 2). (B) ATP dependence of the fraction of full unwinding by Pfh1 of substrates with D-A at the junction and different 3′-tails. (C) Examples of class 1 and class 2 unwinding behavior showing the changes in Cy3 and Cy5 intensities. (D) ATP dependence of the fraction of class 1 and class 2 unwinding behaviors for a 3′-tail with 22 or 5 nt. (E) Schematic of 2 ways of calculating the average number of partial unwinding attempts before full unwinding. (F and G) ATP dependence of the average number of partial unwinding attempts, accounting or not for the presence of t_end.

These observations raise the question of how the nature of the 3′-tail of the DNA modulates the probability of unwinding and, importantly, whether the multiple partial unwinding attempts are on-pathway with formation of full-length unwound product or abortive off-pathway events. To this end, we note that the unwinding behavior of Pfh1 can be divided in 2 classes. For example, at 10 μM ATP, 50% of the molecules show repetitive FRET transitions followed by the presence of a clear final open intermediate state before full unwinding occurs (class 1), while 8% show the presence of this intermediate without evidence of FRET transitions preceding it (class 2) (Fig. 3 A and C). Importantly, the fraction of molecules that unwinds the substrate with class 2 behavior increases as the ATP concentration increases (Fig. 3D). Furthermore, the presence of 2 classes of unwinding behavior for Pfh1 does not depend on the position of the donor-acceptor couple (SI Appendix, Fig. S11 A and B) or the length of the 3′-tail (Fig. 3D) and is a property shared with Pif1 (SI Appendix, Fig. S11C).

The presence of 2 distinct classes of unwinding behavior provides evidence that the repetitive cycle of partial unwinding is not on-pathway with complete unwinding (model 1 in Fig. 4A). To test this point further, we calculated the average number of cycles that the system undergoes before unwinding (Fig. 3E and SI Appendix, Fig. S12). The average number of repetitive cycles increases slightly with increasing ATP concentration, independently of the length of the 3′-tail (Fig. 3 F and G). If the repetitive unwinding attempts were on-pathway, the increase in the fraction of full unwinding as ATP increases would be expected to lead to a decrease in the number of cycles of partial unwinding attempts. Importantly, the same behavior is observed also when the donor is placed 20 bp into the dsDNA (SI Appendix, Fig. S13A) and with Pif1 as well (SI Appendix, Fig. S13B), arguing that the phenomenon is conserved for both helicases.

Fig. 4.

The branch point in the mechanism is determined by conformational selection of an unwinding competent state. (A) Schematic of 4 possible kinetics schemes to explain the partitioning between abortive attempts and full unwinding. Model 1 is an on-pathway mechanism, while models 2 to 4 are off-pathway mechanisms with different branching points. (B) Distribution of the time spent by Pfh1 in the last closed state with a substrate with D-A at the junction and 10 μM ATP. The solid line is the fit to a gamma function with 2 equal rates. (C) ATP dependence of the rate of escape from the last closed state. (D) Distribution of the time spent in the final intermediate state before full unwinding at the indicated ATP concentrations. The solid lines are fits to a gamma function with 7 to 10 equal rates.

Next, we sought to identify the branching point at which the system either enters an off-pathway cycle of repetitive unwinding attempts or commits to enter the open intermediate from which full unwinding occurs. Model 2 in Fig. 4A depicts a simple mechanism where the branching point is at the ATP-bound closed state. Because in this model branching occurs from a single closed state, we would expect the times spent in this state to be exponentially distributed. However, the times spent in the final closed state are not exponentially distributed and can be fitted with a gamma function with 2 steps of equal rate (Fig. 4B). Such a distribution of times is inconsistent with model 2 and strongly argues for the presence of a second closed hidden intermediate. Models 3 and 4 in Fig. 4A depict 2 possible mechanisms whereby the branching point originates from a second closed state. The major distinction between these 2 models is that in model 3 the equilibrium is between 2 ATP-bound closed states (i.e., a sequential mechanism), while in model 4 the 2 closed states exist in equilibrium before ATP binding (i.e., conformational selection). In the sequential model 3 the equilibrium between 2 ATP-bound closed states is a first-order transition and is not expected to be ATP dependent. However, the rates calculated from fitting the distribution of times spent in the final closed state are clearly ATP dependent (Fig. 4C), lending strong support to model 4.

Lastly, the final open state is kinetically distinct from the open state visited during the cycle of partial unwinding attempts. First, the average time spent in this state is longer than the time spent in the open state during the repetitive cycle (SI Appendix, Fig. S12A). Second, independent of the ATP concentration, the times spent in the final open state are not exponentially distributed (Fig. 4D). These distributions can be fitted with a gamma function with 7 to 10 steps, indicating that, upon escape from the final open state, the dsDNA is unwound in multiple ATP-dependent steps.

Partial DNA Opening Is a Mode of Unwinding That Can Be Used during DNA Synthesis.

When coupled to DNA synthesis on the nontranslocating strand, both Pif1 and Pfh1 stimulate DNA replication through stable G-quadruplex structures (44). Moreover, the 5′-3′ unwinding activity of Pif1 stimulates kilobase pair of strand-displacement DNA synthesis, even when proteins are tightly bound to dsDNA (48). Similarly, DNA primer extension assays using an exonuclease-deficient DNA polymerase δ (SI Appendix, Fig. S14), under conditions in which it has limited activity (52), also show that Pfh1 stimulates DNA synthesis (Fig. 5A). We note that Pfh1 is less efficient than Pif1 in stimulating DNA synthesis, as evidenced by the higher enzyme concentration needed and the slower processing of intermediates and accumulation of full product when an 80-bp dsDNA is used. This is consistent with Pfh1 being a slower helicase and suggests that unwinding of the DNA rather than synthesis is the overall rate-limiting step in the reaction.

Fig. 5.

Cooperation between DNA synthesis and helicase activity. (A) DNA primer extension assays were performed with exonuclease-deficient DNA polymerase δ (Pol δ^DV), alone (black) or in the presence of Pfh1 (blue) or Pif1 (red). The reaction and substrates are schematically depicted, with the analyzed replication products binned in different groups and their quantification reported in the corresponding graphs. Error bars are the SD from 3 independent experiments. (B) Unwinding of Cy3-labeled DNA substrates immobilized on the surface by either Pfh1 or Pif1 in the absence or the presence of Pol δ^DV. Experiments to control for photobleaching or for the activity of Pol δ^DV alone are indicated.

The observation that DNA unwinding by Pfh1 and Pif1 occurs via 2 modes (one dominant, short ranged, and highly repetitive, the other less frequent and longer ranged) raises the questions of whether only one or both modes are functionally relevant during DNA synthesis. We note that the probability of full unwinding, by either class 1 and 2, is modulated by the nature of the 3′ nontranslocating strand (Fig. 3B). The presence of dsDNA at the 3′ arm of the substrate does not affect the kinetics of repetitive unwinding, as monitored via FRET transitions with the D-A couple at the junction (21-bp dsDNA, Fig. 2E); however, we detected a limited amount of full unwinding. This suggests that, similar to 3′-T₂₂, with a 21-bp dsDNA at the 3′ arm the long-ranged mode of unwinding is not a frequent event, and the helicase activity is dominated by the short-ranged and repetitive mode. Consistent with this, unwinding assays monitoring the disappearance of Cy3 signal from DNA immobilized on the surface show limited unwinding by either Pfh1 or Pif1 (Fig. 5B). If only the long-ranged mode of unwinding were to be used, because of its low frequency one may expect a limited effect of DNA synthesis on DNA unwinding. However, addition of DNA polymerase δ significantly increases the apparent rate and extent of full product formation (Fig. 5B), suggesting that the short-ranged mode of unwinding (e.g., partial opening) is being used.

Discussion

In summary, we showed that the unwinding of dsDNA by the S. pombe Pfh1 helicase, like S. cerevisiae Pif1, is dominated by highly processive and repetitive attempts of partial DNA opening. The presence of these abortive unwinding events explains the apparent DNA rewinding activity observed in ensemble experiments: repetitive opening of a limited number of base pairs (e.g., <20 bp) would not lead to unwinding of sufficiently long dsDNA. Interestingly, Pif1 has been proposed to unwind dsDNA in 1-bp steps (53, 54), and our data clearly point to an intermediate state visited during unwinding. However, during the partial unwinding attempts, both Pif1 and Pfh1 open more than 2 bp, yet only one intermediate is kinetically populated. Therefore, this intermediate must originate from the opening of multiple base pairs. Importantly, repetitive unwinding of dsDNA has been reported for other helicases, and multiple mechanisms that would lead to closure of the transiently opened dsDNA have been proposed. For example, strand-switching during unwinding, with the helicase being able to jump to the opposite ssDNA strand and translocate back, has been proposed for multiple helicases (7, 11, 50, 55), including Pif1 (47). The observation in this work that, for both Pfh1 and Pif1, repetitive unwinding occurs also on RNA-DNA hybrids provides strong experimental evidence that strand-switching is not a significant mechanism leading to closure of the partially opened dsDNA. On the one hand, a spring-loaded or snap-back mechanism (1, 8, 55), where the repetitive cycle of unwinding originates from the helicase remaining bound to a portion of the substrate, may explain closure of the partially opened DNA. While Pif1 has been shown to repetitively reel in ssDNA or unwind G-quadruplexes when bound with high affinity to a 5′-ds/ssDNA junction (6), neither ssDNA translocation nor dsDNA unwinding require such a site to occur (45, 56). For the DNA substrates in this work, the repetitive partial unwinding attempts occur independently of the 3′-ssDNA tail of the substrate, leaving the 5′-ssDNA as the potential anchor point. In this scenario, Pfh1 or Pif1 would have to remain bound to the 10-nt 5′-tail as they unwind the downstream duplex. On the other hand, closure of the partially unwound DNA could be due to the helicases slipping back on the substrate. This would be consistent with the same mechanism reported for Pif1 as an alternative pathway to strand-switching (47) and for other helicases (57–59). Although our data do not allow us to unambiguously discriminate between snap-back and slippage back, based on our observation that DNA synthesis on the nontranslocating strand stimulates DNA unwinding, we favor the latter explanation. This is not to say that neither Pif1 nor Pfh1 can use either strand-switching or spring-loaded mechanisms. It is possible that these mechanisms may become significant under different experimental conditions, such as when DNA is under tension (47) or a high-affinity site for the helicases is present (6).

Importantly, the observation that Pfh1, Pif1, and other helicases can undergo repetitive cycles of partial dsDNA unwinding raises the question of whether these attempts are on-pathway to full enzymatic activity (e.g., full product formation) or off-pathway events. We showed here that for both Pfh1 and Pif1 the latter appears to be the case. The data suggest a branched mechanism of DNA unwinding, where the major branching point is established by conformational selection of an unwinding competent complex. Such a mechanism would explain why the nature of the 3′-ssDNA tail of the substrate does not affect the kinetics of partial DNA opening, but instead modulates the probability of full product formation. We propose that interaction of the 3′-tail of the substrate with a secondary site on Pfh1 and Pif1 (45), separate from the primary interaction site with the 5′ translocating strand, leads to a conformation of the enzymes that is capable of unwinding longer stretches of dsDNA. The location of the secondary DNA site and how it may interact with 3′-tails of different sequence composition remain to be determined. Moreover, whether regulation of unwinding by the nature of the 3′-ssDNA tail of the substrate is unique to Pif1-family members or shared with other helicases is currently unknown. Crystal structures of Pif1-family members suggest that the 2B domain undergoes rotation upon DNA binding (60–63). We speculate that interaction of the 3′ nontranslocating strand with a secondary DNA site may modulate rearrangements of the 2B domain that lead to different modes of unwinding. For example, stabilization in a closed state of the 2B domain of Rep and PcrA helicases leads to highly processive DNA unwinding (64).

Finally, the observation that Pfh1 and Pif1 can switch between 2 modes of unwinding, one in which the enzyme processively opens short stretches of DNA and the other in which longer DNA stretches can be unwound, raises the question of whether both modes are functional or only one is predominantly used. The ability of the nature of the 3′-ssDNA tail of the DNA to shunt the system toward unwinding of longer stretches of DNA could be used to regulate, in a DNA-sequence–dependent context, the activity of these helicases. However, even in this mode the length of DNA that can be unwound remains limited. In cells, Pif1 and Pfh1 have been shown to function in multiple pathways that involve DNA synthesis: DNA replication across G-quadruplex sequence motifs (23–26); break-induced replication (32, 33); and processing of long flaps during Okazaki fragment maturation (27–31). Also, in vitro Pif1 increases strand-displacement DNA synthesis by Pol δ and allows DNA replication across arrays of proteins tightly bound to DNA (48). Similarly, here we showed that Pfh1 also stimulates strand-displacement DNA synthesis of Pol δ. Conversely, Pol δ activity stimulates the apparent rate and extent of DNA unwinding. Taken together, these observations suggest that DNA synthesis on the 3′-strand prevents the helicase from slipping back on the substrate, effectively working as an intrinsic trap in cis that suppresses DNA closure. That is to say, DNA synthesis on the nontranslocating strand rectifies the unwinding activity. We propose that efficient and processive opening of short stretches of DNA, the dominant mode of unwinding observed in vitro, is all that the Pif1 family of 5′-3′ helicases needs to achieve when their activity is coupled to DNA replication.

Methods

Single-molecule experiments were performed with an objective-type total internal reflection fluorescence (TIRF) microscope (Olympus IX71) with an oil-immersed, high-numerical-aperture TIRF objective (PlanApo N, 60×/1.45 N.A., Olympus) and exciting Cy3 with a 532-nm laser (CrystaLaser), as described previously (65). Cy3 and Cy5 emission intensities were split and recorded with an Andor iXon EMCCD camera (Model DU897E). The temperature of the slide was maintained at 25 °C using a temperature-controlled stage (BC-110 Bionomic controller; 20/20 Technology) and an objective heater (Bioptechs). A single-channel flow chamber was assembled from a coverslip (VWR, 24 × 50 mm N.1) and a predrilled slide (VWR, 75 × 25 × 1 mm). The channel was coated with a solution containing a 1:20 ratio (wt/vol) of biotin-polyethylene glycol (PEG)-succinimidyl valeric acid (SVA) (MW_avg 5000) and (wt/vol) mPEG-SVA (MW_avg 5000) (Laysan Bio, Arab, AL) in 0.1 M NaHCO₃ (pH 8.1) and incubated overnight in the dark at 4 °C. The flow channels were then washed with water, dried with N₂, and stored under vacuum conditions. Before the experiment, a 0.2-mg/mL solution of NeutrAvidin (Thermo Scientific) was flowed into the channel and incubated for 2 min, and excess NeutrAvidin was washed away by 200 µL of binding buffer (20 mM Tris⋅HCl, pH 8.0, 20 mM NaCl, 2% [vol/vol] glycerol, 8 mM MgAc_2, 1 mM dithiothreitol, 0.1 mg/mL bovine serum albumin). Next, 10 to 100 pM of biotinylated DNA substrate in binding buffer was added to the chamber for 5 min, free DNA was washed away with 400 µL of binding buffer, and the DNA on the surface was imaged in imaging buffer (binding buffer supplemented with 0.5% wt/vol glucose, 3 mM Trolox, 165 U/mL glucose oxidase, and 2,170 U/mL catalase). Next, 100 nM protein was incubated in binding buffer with the surface-immobilized DNA for 10 min, and excess unbound protein was washed away with 100 to 200 μL of imaging buffer (“wash” condition). Finally, the reaction was started by addition of the indicated concentration of ATP in imaging buffer, and images recorded at 30 Hz. Data were collected using SINGLE (provided by T. J. HA, Johns Hopkins University), processed with IDL (Exelis VIS) and analyzed with MatLab (Mathworks).

Experiments with the protein on the surface were performed by immobilizing His₆-tagged Pfh1 with a biotinylated PentaHis antibody (45). DNA unwinding experiments monitoring disappearance of the signal from Cy3-labeled DNA substrates, immobilized on a surface, were performed adding at the same time 1 mM ATP (or 1 mM ATP and 100 μM dNTP) and helicase (5 to 10 nM) in the absence or the presence of DNA polymerase (20 nM).

Further details on protein purification, DNA substrates, and ensemble assays are provided in SI Appendix.