Single-molecule measurements of RecQ helicase unwinding activity were performed with 174- or 584-base pair DNA hairpins using an MT apparatus (Figure 1A). DNA hairpin substrates were attached to the flow-cell surface and to a 1- or 2.8 µm magnetic bead via a 1.1 kbp double-stranded DNA handle and 60-nucleotides of single-stranded poly-dT, respectively (Figure 1A). Measurements with the DNA hairpin were performed at a constant force of 8 pN under which the hairpin did not open spontaneously. In the presence of RecQ helicase (20–100 pM), unwinding activity was monitored in real-time by tracking the three-dimensional position of a tethered bead at 60 or 200 Hz. Trajectories of the bead extension as a function of time were analyzed by fitting with a T-test based step finding algorithm to obtain the mean unwinding rate, the ‘step’ unwinding rate between pauses, the pause positions, and the pause durations (Harami et al., 2017; Seol et al., 2016).

Figure 1

Download asset Open asset

Figure 1—source data 1

Source data for Figure 1.

https://doi.org/10.7554/eLife.45909.004

Download elife-45909-fig1-data1-v2.xlsx

As described previously (Harami et al., 2017), frequent pausing and strand-switching by WT RecQ (RecQ^WT) is caused by the HRDC as the HRDC deletion mutant (RecQ-dH) shows significantly less pausing during DNA hairpin unwinding (Figure 1B). Pausing is attributed to transient binding of the HRDC domain to the displaced single-stranded DNA behind RecQ. Since both the displaced and the translocation strands of ssDNA are under tension in the hairpin substrate, binding of the HRDC to the displaced strand will prevent forward motion of the helicase. HRDC binding to either duplex DNA ahead of the helicase, or to the translocation strand of ssDNA behind the helicase, are ruled out by the lack of pauses during the unwinding of a ‘gapped’ DNA substrate in which the displaced strand is not constrained (Harami et al., 2017). Given the mechanical origin of the pausing associated with transient binding of the HRDC, the pause positions would be expected to be random, dependent on the stochastic kinetics of the interaction between HRDC and the displaced ssDNA.

Interestingly, the dwell-time histogram as a function of position for RecQ^WT unwinding traces exhibits peaks at distinct positions along the hairpin (Figure 2A; top). The peaks in the dwell-time histogram of unwinding traces arise from long and/or frequent pauses at specific positions during DNA hairpin unwinding by RecQ helicase (Figure 2—figure supplement 1). To identify the sequence context of the pauses, the extension change associated with DNA hairpin opening was converted to base-pairs via the worm-like chain (WLC) model of DNA (Manosas et al., 2010). Each unwound base pair resulted in the increase of the molecular extension by two ssDNA nucleotides, which at an applied force of 8 pN corresponds to ~0.8 nm assuming a 1 nm persistence length and a 0.65 nm inter-phosphate distance. With this conversion factor, the extension change for the fully open hairpin was 174 bp, consistent with the actual DNA hairpin size (174 bp). To determine if pausing is related to DNA base-pair energy, we compared the unwinding dwell time histogram (Figure 2A; top) with the DNA base-pair stability calculated by performing a running average (6 bp window) of the exponential of the DNA base-pair energy for the 174 bp DNA hairpin sequence based on the nearest neighbor base-pair energy model (Patten et al., 1984; SantaLucia, 1998; Huguet et al., 2010). We found that the peak locations of pausing and duplex stability were highly correlated (Figure 2A; bottom). The exact locations of peaks were identified by globally fitting the dwell-time histogram and the exponential of the average DNA melting energy with the sums of Gaussian distributions (Figure 2—figure supplement 2). The relationship between pausing during unwinding and the peaks in the dwell-time histogram is explained in Figure 2—figure supplement 2B (top). Consistent with this observation, pause positions from the dwell time histogram of RecQ^WT were linearly correlated with peak positions from the DNA base-pair energy profile (Figure 2B; top) with a slope of 0.99 ± 0.03, linear correlation coefficient (Pearson’s r) of 0.97, and χ² = 0.85, indicating a strong linear correlation. The sequence around the peak positions (±4 bp) contained a high percentage of GC (~70%), consistent with the finding that the pause positions are related to the duplex stability of the DNA. This finding raises the question of how HRDC-dependent pausing is correlated with DNA base-pair melting despite the fact that the HRDC itself does not play a role in unwinding DNA or exhibit sequence-specific ssDNA binding. We hypothesized that the HRDC may amplify or stabilize transient pauses associated with RecQ core domain (RecQ-dH) encountering regions of increased duplex stability (high GC content).

Figure 2 with 2 supplements see all

Download asset Open asset

Figure 2—source data 1

Source data for Figure 2—figure supplements 1 and 2.

https://doi.org/10.7554/eLife.45909.008

Download elife-45909-fig2-data1-v2.xlsx

To test this hypothesis, we determined if the transient pausing positions of RecQ-dH correlated with the peaks in the DNA base-pair stability curve (Figure 2B). The pause positions for RecQ-dH were obtained from dwell time histograms following the same procedure used for RecQ^WT (Figure 2—figure supplement 2C) and plotted as a function of the peak positions of DNA base-pair stability (Figure 2B). The pause positions of RecQ-dH were linearly correlated with the duplex stability peaks, returning a slope of 0.96 ± 0.06, Pearson’s r = 0.95, and χ² = 1.1. Moreover, the pause positions of RecQ-dH are statistically identical to those of RecQ^WT, confirming that HRDC-dependent pausing likely originates from stabilization of sequence-dependent unwinding kinetics of RecQ helicase. Sequence-dependent pausing by RecQ-dH reveals important mechanistic insights into the unwinding and translocation mechanism. If the enzyme unwinds one base pair per each kinetic step, the largest energy difference for a single base-pair opening (G/C vs A/T) is ~2.0 k_BT so the pause duration ratio of G/C to A/T will be a maximum of ~7 fold. However, the roughly 20-fold difference in the time the enzyme requires to unwind DNA at the longest pause duration sites in comparison to the average unwinding rate, suggests that more than a single base pair is being opened by the enzyme during each kinetic step. Following this simple analysis, we suggest that pausing is governed by a combination of the DNA base-pair stability and the number of base pairs melted by the helicase during each kinetic step. During processive unwinding, this melting step is the rate limiting step that determines the unwinding rate and pause durations.

To distinguish among possible models for the unwinding mechanism of RecQ-dH based on its pausing behavior, we simulated unwinding trajectories comprising a series of pauses and translocations (Figure 3A).

Figure 3 with 1 supplement see all

Download asset Open asset

Figure 3—source data 1

Source data for Figure 3 and Figure 3—figure supplement 1.

https://doi.org/10.7554/eLife.45909.011

Download elife-45909-fig3-data1-v2.xlsx

Based on previous studies of helicases (Manosas et al., 2010; Cheng et al., 2007; Neuman et al., 2005; Cheng et al., 2011; Lin et al., 2017; Myong et al., 2007), we considered two scenarios for RecQ-dH unwinding with an n-bp kinetic step size: either the enzyme unwinds n base-pairs simultaneously then rapidly translocates along the unwound DNA (simultaneous melting model), or it sequentially unwinds n base-pairs then releases the newly melted ssDNA (delayed release model) (Figure 3A). We exclusively simulated RecQ-dH unwinding and pausing kinetics rather than RecQ^WT due to the significantly more complex behavior of the RecQ^WT unwinding trajectories (Figure 1B).

In the simultaneous melting model, the pause duration, τ is related to the sum of n base-pair energies at the position of the i^th kinetic step,

Here G_1bp is the free energy required for 1 base-pair melting at the i^th position calculated using the nearest-neighbor energy parameters (Patten et al., 1984; SantaLucia, 1998; Huguet et al., 2010), s is a step index ranging from 1 to n, and A_RecQ is a pre-factor used to adjust the simulation to give the same average unwinding (46 nt/s) and translocation rate (~100 nt/s) as the RecQ-dH construct (Manosas et al., 2010). In the delayed release model,${\tau }_p\left(i\right)$ is the sum of the pause times associated with melting each of n base pairs at the i^th kinetic step,

Stochastic simulations of both models were run with different step-sizes, n. For each value of n, the pre-factor A_RecQ was adjusted to match the measured average rate of RecQ-dH, and the single-strand DNA translocation rate was 100 bp/s (Manosas et al., 2010; Bagchi et al., 2018). Simulated unwinding traces were generated for different kinetic step-sizes for the two different models (100 traces per each condition, example traces are shown in Figure 3—figure supplement 1). Simulated traces were analyzed with a T-test based step finding algorithm with the same parameters used for experimental data analysis. Pause durations were binned over 5 bp intervals for simulation and experimental traces and the mean pause duration for each bin was calculated (Figure 3B). Simulation results were compared with experimental data for RecQ-dH by calculating the reduced χ² (χ_ν²) between the simulated and experimental traces (Figure 3C). For the delayed release model, χ_ν² reached a minimum around 5 bp (χ_ν² = 0.9x10⁻¹), lower than the minimum for simultaneous melting model that reached a minimum at 2 bp (χ_ν² = 1.4x10⁻¹). This suggests that a delayed release scenario may describe the unwinding mechanism of core RecQ. To confirm this finding, we investigated how the RecQ-dH unwinding rate was affected by Na⁺ concentration and compared the results with the two unwinding models. As DNA base-pair melting energy increases with Na⁺ concentration (SantaLucia, 1998; Huguet et al., 2010), the average unwinding rate predicted by the simultaneous melting model should decrease more rapidly than that predicted by the delayed release model (Figure 3D). We varied the Na⁺ concentration from 25 to 500 mM while maintaining Mg²⁺ at 5 mM under otherwise identical buffer conditions. The unwinding rate of RecQ-dH decreased with increasing Na⁺ concentration. The relative decrease in unwinding rate was much better described by the delayed release model with a 5 bp kinetic step. than the simultaneous 2 bp DNA melting model (Figure 3D). The small deviation between the delayed release model and the measured Na⁺ concentration dependence of the unwinding rate suggests that although the duplex unwinding remains the rate-limiting step, the Na⁺ concentration effects other aspects of unwinding such as protein-DNA interactions, which are beyond the scope of the simple model. Thus, to further test and confirm the delayed release unwinding model, we performed two additional experiments as explained below.

The significant DNA sequence dependence of the RecQ-catalyzed DNA unwinding rate and pausing characteristics detected in MT single-molecule experiments should be reflected in ensemble unwinding kinetic measurements, which are suitable for the determination of the kinetic step size and the macroscopic dsDNA unwinding rate (Lucius et al., 2003). In these experiments unwinding kinetics are monitored via the appearance of fully unwound reaction products. Thus, ensemble unwinding experiments complement MT experiments, in which individual unwinding steps are monitored. Importantly, these techniques together should allow determination of the microscopic unwinding mechanism of RecQ helicase constructs based on the proposed base-pair energy dependent unwinding models.

To test this idea, we performed single-turnover unwinding kinetic experiments in which we rapidly mixed complexes of RecQ^WT or RecQ-dH with forked DNA substrates of varying GC content with ATP and excess unlabeled ssDNA traps in a quenched-flow instrument and monitored the time course of fluorescently-labeled ssDNA generation via gel electrophoresis of reaction products (Figure 4A). Forked DNA substrates used in the experiments comprised two 21-nt ssDNA arms and a 33 bp dsDNA segment containing 12 (gc36), 16 (gc48) or 26 GC (gc79) bps (sequences described in Supplementary file 1Table S1). Unwinding traces comprised a short (~0.1 s) initial lag, followed by a biphasic appearance of the labeled ssDNA reaction product (Figure 4B). The rapid rise originated from single unwinding runs of initially DNA-bound helicase molecules. The slow rise phase originates from premature dissociation, followed by slow rebinding, of the enzyme to the DNA substrate (hindered but not totally inhibited by the ssDNA trap strand) that eventually led to full unwinding of the DNA fork (Harami et al., 2017). To obtain parameters of unwinding, data were analyzed with a modified version of a previously described n-step kinetic model (Lucius et al., 2003). In its simplest form the model assumes that DNA unwinding occurs as a result of n consecutive rate limiting steps that have a uniform rate constant. This model is generally suitable for the determination of the macroscopic dsDNA unwinding rate, the kinetic step size and the number of intermediates in the unwinding reaction (Lucius et al., 2003).

Figure 4

Download asset Open asset

Figure 4—source data 1

Source data for Figure 4.

https://doi.org/10.7554/eLife.45909.013

Download elife-45909-fig4-data1-v2.xlsx

Using a modified version of the n-step model (Figure 4C), global fitting of the unwinding kinetics of the gc36, gc48 and gc79 substrates using an integer series of n ranging from 1 to 7 revealed smallest χ_v² values for an apparent kinetic step size of 5 bp for both RecQ and RecQ-dH (Figure 4B–D) with all DNA substrates, similar to that suggested by our MT results (Figure 3A–B) and by previous findings (Lin et al., 2017; Harami et al., 2015). However, the n-step model does not consider the sequence dependence of the rates of elementary unwinding steps, precluding the distinction between different microscopic mechanisms producing the same kinetic step size. Therefore, we used the same physical framework as described for the MT experiments (Equations 1 and 2) and performed global kinetic fitting to all transients of a given helicase construct (RecQ^WT or RecQ-dH) unwinding the different forked DNA substrates, based on the DNA sequence-dependent simultaneous melting and delayed release unwinding models (Figure 4B–C). For both models, fitting was done using an integer series of n ranging from 1 to 7. In agreement with the results of the MT analysis (Figure 3C), the smallest χ_v² value was obtained for the delayed release model with a kinetic step size of 4 bp for RecQ-dH and 5 bp for RecQ^WT (Figure 4B and C, other parameters are listed in Supplementary file 1 Table S2).

If RecQ-dH takes a certain kinetic step size, it could in principle be directly observed in the single-molecule unwinding traces. However, the enzyme unwinds DNA too rapidly at high ATP concentrations for steps to be routinely and accurately detected, given the spatial resolution limits of the measurement. Under our experimental conditions, the average baseline noise was ~14 nm at 200 Hz data collection rate. Thus, in order to observe, for example, a 4 bp step (i.e. a 3 nm change in DNA extension), the average pause duration should be >300 ms or the unwinding rate should be less than 13 bp/s (~3 fold slower than 42 bp/s). We tried three different conditions to decrease the unwinding rate of RecQ: lowering the ATP concentration (Figure 5—figure supplement 1) and including non-hydrolysable ATP analogues, ATPγS or AMP-PNP, in the assay (Figure 5—figure supplement 2). We found that decreasing the ATP concentration (sufficiently lowering the unwinding rate) resulted in frequent and extensive enzyme backsliding (observable as rapid partial rezipping of the hairpin during an unwinding event), which complicates kinetic step size measurements (Figure 5—figure supplement 1). AMP-PNP showed extremely slow dissociation kinetics from RecQ-dH that were inappropriate for unwinding assays (Figure 5—figure supplement 2). On the other hand, ATPγS, showed a comparable binding affinity to ATP with a significantly shorter binding time (~1 s) than AMP-PNP (Figure 5—figure supplement 2). In addition, ATPγS binding transiently locks RecQ in the strong DNA-binding ATP bound state without backsliding, leading to long duration pauses that effectively increased the spatial resolution by permitting longer averaging times (Figure 5A). We measured the unwinding activity of RecQ-dH at different fractions of ATPγS (0.05–0.5 mM) while keeping the total combined concentration of ATP and ATPγS constant at 1 mM. The unwinding rate decreased with increasing ATPγS fraction (Figure 5A). We reason that when the concentration of ATPγS is such that it is bound at least once per kinetic step, then the predominant physical step-size measured in the hairpin unwinding trajectories will correspond to the kinetic step-size. Step-sizes were estimated with two different step finding algorithms: a step finding program originally developed by Kerssemakers and coworkers (Kerssemakers et al., 2006) and the T-test based step finding analysis (Seol et al., 2016). To determine the average kinetic step-size for each condition, the estimated step-sizes were histogrammed and fit with Gaussian distributions (Figure 5B and C). We found that the estimated step size of RecQ-dH from both step-finding algorithms were comparable, converging from ~8 bp at a low ATPγS fraction to 5 bp at higher ATPγS fractions, suggesting that the average kinetic step size of RecQ-dH is 5 bp [T-test: 5.3 ± 0.1 (center); 3.0 ± 0.6 (Standard Deviation); Kerssemakers: 5.2 ± 0.1(center); 2.1 ± 0.1 (Standard Deviation), errors correspond to the standard deviations from Gaussian fitting]. The broad step-size distribution could reflect the stochastic nature of ssDNA release by RecQ. Also, it is likely that the two ssDNA strands are released asynchronously by RecQ. In line with this, we occasionally observed a 2.5 bp kinetic step at 500 µM ATPγS, and the step-size distribution at lower ATPγS fractions included peaks at 7.5, 10, and 12.5 bp, consistent with a fundamental step-size of 2.5 bp corresponding to the release of one ssDNA strand of 5 nt (Figure 5—figure supplement 3).

Figure 5 with 4 supplements see all

Download asset Open asset

Figure 5—source data 1

Source data for Figure 5 and Figure 5—figure supplements 1–4.

https://doi.org/10.7554/eLife.45909.019

Download elife-45909-fig5-data1-v2.xlsx

The prolonged pause state due to ATPγS binding instead of ATP at the cleft between two RecA domains of RecQ enabled us to probe the mechano-chemical coupling, C, of RecQ helicase that is a measure of the number of ATP hydrolyzed per kinetic step. For C = m/n (m ATP hydrolysis per n kinetic step size), the average number of bound ATPγS, l, can be estimated based on the binomial probability distribution.

P is the probability of ATPγS binding per each cycle. k_ATP is the ATP on-rate, k_ATPγS is the ATPγS on-rate, and [ATPγS] and [ATP] are the concentrations of ATPγS and ATP, respectively. The hydrolysis rate of ATPγS by RecQ in the presence of excess dT₄₅ was measured by monitoring thiophosphate production (Saran et al., 2006) and estimated to be < 0.2/s (Appendix 1 and Figure 5—figure supplement 1). This is significantly slower than the measured pause escape rate (2.8 ± 0.1/s) at 500 µM ATPγS, suggesting that k^off_ATPγS is much faster than the rate of ATPγS hydrolysis. Thus, we could simplify the mean pause duration per kinetic step, τ and the mean unwinding rate, $v$ as,

 where k_step is the mean kinetic stepping rate without ATPγS. We obtained the average pause durations for different fractions of ATPγS from 5 to 50 % and globally fitted the pause durations and the average unwinding rates as a function of ATPγS concentration with Eq. 4 and 5 respectively (Figure 5D). From this global fitting, we found that C = 1.0 ± 0.2 bp/ATP, k_ATP/ k_ATPγS = 1.2 ± 0.2, and 1/k^off_ATPγS = 0.4 ± 0.1 s suggesting a tight mechano-chemical coupling in agreement with previous ensemble measurements (Harami et al., 2015; Sarlós et al., 2012). We note that rebinding of ATPγS was not taken into account for simplicity in Equations 2 and 3, which is reasonable as ATPγS concentration is lower than ATP except for 50% ATPγS, and because the relative on rate (k_ATP[ATP] vs k_ATPγS[ATPγS]) of ATPγS is lower than that of ATP. To ensure that this simplification is reasonable, we simulated how many ATPγS molecules are bound instead of ATP per base pair based on the fitting parameter, k_ATP/k_ATPγS = 1.2 at 50% ATPγS. We found that the average is less than one ATPγS per base pair at this condition indicating that repetitive ATPγS binding at the same site is rare.

The sequence-dependent unwinding mechanism of RecQ consisting of a 5 bp kinetic step results in transient pauses that are further stabilized by the HRDC, which results in the long-lived sequence-dependent pausing of RecQ^WT (Figure 2). In addition to the sequence dependence, HRDC-dependent pausing exhibits two interesting features: occasional repetitive rezipping and unwinding (shuttling) around the pause position and significantly prolonged pausing durations for certain pausing positions (Figure 2 and Figure 2—figure supplement 1). It appears that HRDC-binding triggers this shuttling behavior in which 5–10 bp are repetitively unwound and rezipped at the relatively long-lived (>0.14 s) intrinsic pause positions. Shuttling activity repeats until RecQ passes the sequence-dependent roadblock. This complex shuttling behavior was significantly enhanced at those regions where long pauses of RecQ-dH are clustered, such as at 55, 90, and 120 bps (Figure 2—figure supplement 1) resulting in the apparent high dwell probabilities at these sites (Figure 2A). Since these positions also exhibit the highest base-pair stabilities (Figure 2A), the average dwell time of RecQ^WT is strongly correlated with the base-pair stability. Indeed, the average dwell times for RecQ^WT are highly non-linearly correlated with the base-pair stability (Figure 6A). In contrast, the dwell-times for RecQ-dH scale linearly and much less dramatically with the base-pair stability (Figure 6A), indicating that the HRDC-stabilized pausing can be described as a non-linear amplifier of the intrinsic sequence-dependent unwinding kinetics. We found that a simple kinetic competition model in which HRDC binding is in kinetic competition with the forward motion of the helicase (Appendix 1) cannot reproduce the dramatic changes in pause probability observed for RecQ^WT hairpin unwinding (Figure 6—figure supplement 1). In line with this, the pausing duration distribution for RecQ^wt is better described by a double rather than single exponential distribution (Figure 6—figure supplement 1B). Both pause lifetimes (1.8 ± 0.3 s and 0.4 ± 0.1 s) are longer than the average pause duration (0.14 ± 0.03 s, see SI) for the core RecQ (RecQ-dH), indicating that there are multiple HRDC-dependent pause states.

Figure 6 with 2 supplements see all

Download asset Open asset

Figure 6—source data 1

Source data for Figure 6 and Figure 6—figure supplements 1–2.

https://doi.org/10.7554/eLife.45909.023

Download elife-45909-fig6-data1-v2.xlsx

To test the proposal that base-pair stability-dependent RecQ^WT pausing underlies a potential mechanism of homology sensing, we investigated the effect of introducing single mismatches at high probability pause sites. We tested hairpin substrates containing 1, 2, or three single mismatches: (i) a mismatch introduced at 90 bp, (ii) mismatches introduced at 90 bp and 104 bp and (ii) mismatches introduced at 90 bp, 104 bp, and 124 bp. All mismatches were generated by changing G to T on the displaced strand (detailed sequence information is in Appendix 1). We found that pausing of RecQ^WT around the 90 bp unwound hairpin position was significantly reduced compared to intact 174 bp hairpin when a mismatch was present at 90 bp and additional mismatch at 124 bp further suppressed pausing around 120 bp (Figure 6B and Figure 6—figure supplement 2). The effect of mismatches on pausing of RecQ^WT can be clearly demonstrated by comparing the dwell-time histograms of three DNA substrates (Figure 6B). The prominent peaks in the dwell time histogram of the intact hairpin DNA were diminished one by one with the introduction of mismatches at the corresponding positions confirming the correlation between pausing and homology (Figure 6B and Figure 6—figure supplement 2).

The fundamental activity of helicases is the unwinding of duplex nucleic acids. In general, the unwinding mechanism has been classified as either passive or active depending on the degree to which the enzyme ‘actively’ destabilizes the duplex rather than ‘passively’ waiting for a thermal fluctuation to expose ssDNA (Betterton and Jülicher, 2005; Lohman et al., 2008). For a purely passive helicase, the enzyme does not provide external work to destabilize duplex DNA and translocates only when ssDNA is exposed by thermal fluctuations. On the other hand, an active helicase is actively involved in disrupting the DNA duplex, and in principle, is less sensitive to base-pair energy or sequence. In previous single molecule experiments, E. coli RecQ helicase was identified as an active helicase based on the minimal force and DNA sequence dependence of duplex unwinding (Manosas et al., 2010). In that study, following theoretical work by Betterton et al (Betterton and Jülicher, 2005), RecQ unwinding was modeled as one base-pair melting followed by 1–2 bases translocation. However, we found that the pause durations were generally longer than would be expected for melting of 1 base pair when we compared our results with simulations. We considered two different scenarios: RecQ either destabilizes multiple base-pairs (≥2 bp) during each kinetic step similar to NS3 helicase (Cheng et al., 2007) or delays releasing of multiple unwound base-pairs similar to speculative models suggested in previous studies (Cheng et al., 2011; Lin et al., 2017; Myong et al., 2007; Ma et al., 2018). However, the minimal dependence of the unwinding rate on Na⁺ concentration in addition to the sequence-dependent pauses cannot be explained by multi-base-pair melting. Rather, we found that an alternative scenario in which RecQ delays the release of nascent single-strand DNA (delayed release) was a better fit to the pause duration and Na⁺-dependent unwinding rate data, though the associated kinetic step size (number of bp unwound prior to release) was not uniquely constrained by the pause duration or Na⁺-dependent unwinding rate measurements (Figure 3). This finding is consistent with previous studies revealing ‘asynchronous’ release of nascent ssDNA (Lin et al., 2017; Ma et al., 2018). Nonetheless, the mechanism of delayed release of newly melted nucleotides remains unclear. Previous results suggest that a putative electrostatic interaction between newly melted ssDNA and RecQ sequesters several nucleotides of ssDNA. We consider a similar possibility in which RecQ releases the nascent ssDNA only when the accumulated torsion or tension on bound ssDNA is high enough to disrupt the interaction (Myong et al., 2007).

We further refined the delayed release model by directly measuring a 5 bp kinetic step size for DNA unwinding using ATPγS, which sufficiently slows down the unwinding rate without inducing the frequent back-sliding observed at reduced ATP concentrations (Figure 5 and Figure 5—figure supplement 1)). Recent single molecule fluorescent studies showed 2–4 bp kinetic step (Lin et al., 2017; Ma et al., 2018). This smaller and more random nature of the kinetic step size is likely due to the low ATP concentration (2–5 µM), at which ATP binding likely becomes the dominant rate-limiting step slower than or on the same order as the intrinsic off-rate of the nascent DNA. Consistent with this model, the study found a correlation between the ATP concentration and the measured kinetic step size.

The mechano-chemical coupling and kinetic analysis of ssDNA translocation of RecQ have been studied in detail (Sarlós et al., 2012; Rad and Kowalczykowski, 2012). Our unwinding kinetic step is consistent with a recent a study in which a five nucleotide kinetic step for RecQ translocation was estimated (Rad and Kowalczykowski, 2012). Other helicases display multi base-pair kinetic unwinding steps under sufficient ATP concentrations (above K_M ~ 20 µM) (Lohman et al., 2008). The mechano-chemical coupling is a measure of how many chemical cycles an enzyme completes to take one mechanical step. In the case of RecQ or other helicases, it corresponds to how many ATP molecules are consumed per one base translocation (or base-pair unwound for unwinding). For translocation, RecQ shows a tight coupling close to one nucleotide step per ATP hydrolysis (Sarlós et al., 2012). Our study reveals that the mechano-chemical coupling for unwinding maintains one base-pair melting per ATP hydrolysis (Figure 4C), which is also supported by the results of a recent single-molecule florescence study of RecQ unwinding (Lin et al., 2017). The proposed kinetic model based on our ATP dependent kinetic analysis (Appendix 1; Figure 5—figure supplement 1) suggests that DNA melting precedes ATP hydrolysis. In this model, ATP binding stabilizes the DNA-RecQ interaction and facilitates DNA melting presumably coupled to an ATP binding-dependent conformational change of RecQ such as rotation of the helicase domains relative to one another, which explains more frequent backsliding under lower ATP concentration (Bernstein et al., 2003; Manthei et al., 2015; Pike et al., 2009). Recent structural results suggest that RecQ binding may melt two base-pairs of DNA before ATP binding (Manthei et al., 2015). This may occur at the initial binding of RecQ (or rebinding) as the initiation of unwinding, but not the unwinding rate, is highly dependent on Na⁺ concentration.

Whereas we establish that pausing arises from the stability of DNA duplex, recent work by Voter et. al suggests an alternative mechanism for sequence-dependent pausing. In their work, they identify a ‘Guanine binding pocket’ located in the helicase domain that specifically interacts with guanine bases to destabilize G-quadruplex structures. It is possible that these interactions could also slow down the unwinding rate at clusters of guanine bases in the translocation strand by inducing short pauses (Voter et al., 2018). However, the translocation sequence at the strong pause locations of our DNA hairpin is mixture of G and C bases, suggesting that the pauses we observed originate from the duplex stability. Nevertheless, we cannot entirely rule out the possibility that these specific guanine interactions contribute slightly to the pausing of RecQ core over and above the dominant effect of DNA duplex stability.

One of the essential aspects of RecQ is that it processes diverse, non-canonical, DNA substrates in which the HRDC plays an important role in modulating substrate-specific unwinding of RecQ. It has been shown that the HRDC regulates the binding orientation of RecQ core to promote disruption of D-loop structures, early homologous recombination intermediates (Harami et al., 2017). However, it was not clear how it can regulate unwinding of RecQ to selectively disrupt illegitimate or non-homologous invading DNA strands since the HRDC presumably cannot directly sense DNA sequence homology (Harami et al., 2017). Our present study reveals that the HRDC-ssDNA interactions are strongly coupled to DNA sequence-dependent pausing of the RecQ helicase core: ssDNA binding by the HRDC is not random but occurs at DNA sequences where the helicase core pauses due to the high duplex stability (Figure 2). On the other hand, either a low homology (base-pair mismatches) or low duplex stability (low GC regions) strongly reduces RecQ pausing (Figures 2 and 6B and Figure 6—figure supplement 2). Importantly, this feature can support discrimination between legitimate and illegitimate recombination events by RecQ helicases, in accordance with the increased illegitimate recombination frequencies detected in vivo upon compromising RecQ HRDC function (Harami et al., 2017; Wang et al., 2016). Recombination events proceed through the formation of a displacement loop (D-loop) flanked by genomic DNA, which, due to the limited mobility of these large DNA domains, mimics the hairpin geometry of the magnetic tweezers experiments in which the displaced DNA strand is constrained (Figure 6C and D). Previously we showed that the HRDC both targets RecQ to D-loop intermediates and orients the enzyme in a configuration favoring D-loop disruption (Harami et al., 2017). The results obtained here provide a mechanistic basis for the subsequent discrimination between legitimate and illegitimate recombination based on the length and stability of the D-loop structure. RecQ-catalyzed unwinding of long and stable D-loops will be frequently interrupted by HRDC-stabilized pauses that drastically decrease the average unwinding rate. This slow average unwinding rate potentially permits the initiation of down-stream recombination processes associated with DNA synthesis resulting in extension and further stabilization of the D-loop. Conversely, RecQ unwinding of short and/or unstable D-loops will proceed rapidly (60–80 bp/s) resulting in the efficient disruption of the D-loop before it can be further extended. Our study reveals that the strategic location of the HRDC relative to the core domain, combined with sequence-dependent DNA unwinding, enable RecQ helicase to control pausing and shuttling in a substrate-dependent manner and expand its biological activity beyond simple duplex DNA unwinding. Whereas this study focused exclusively on E. coli RecQ, the homology sensing mechanism we propose is potentially applicable to the suppression of illegitimate, or so called ‘homeologous recombination’ by BLM (Wang et al., 2016).

Another biological role of HRDC domain-mediated pausing and shuttling could be linked to the role of RecQ helicases in G-quadruplex secondary DNA structure processing. G-quadruplex structures were shown to act as replication road blocks and these regions were shown to be recombinational hot spots (van Wietmarschen et al., 2018; Rhodes and Lipps, 2015). RecQ helicases can efficiently unwind G-quadruplex structures, possibly to aid DNA replication, suppress genome instability and to influence transcription of various genes (Voter et al., 2018; Mendoza et al., 2016). Prolonged shuttling at G-quadruplex sites could ensure that these secondary structures remain unfolded until further steps of replication or DNA repair can proceed. In line with this idea, the HRDC domain of human BLM helicase was shown to be essential for efficient, repetitive unwinding of G-quadruplexes Chatterjee et al., 2014).

In this study, we focused on elucidating the sequence-dependent unwinding and pausing mechanism of RecQ helicase in vitro with purified proteins. Whereas our results indicate a possible mechanism for homology sensing by RecQ helicases, the translocation and pausing kinetics on which the model is based could be modulated in vivo due to the interactions with other DNA binding and processing enzymes. For example, single-strand binding protein (SSB) would likely compete with the HRDC for ssDNA binding. However, we recently demonstrated that SSB is displaced by RecQ despite the much higher apparent binding affinity of SSB for ssDNA (Mills et al., 2017). Furthermore, the high local concentration of the HRDC, which is tethered to the RecQ core by a flexible linker, likely results in the HRDC out-competing other ssDNA binding proteins for the newly melted ssDNA. Nonetheless, as is often the case, RecQ helicases play diverse roles in DNA processing through the interaction with other proteins, thus, future experiments in the presence of other proteins that interact with RecQ in vivo including, SSB, RecJ, RecA, and Topoisomerase III are warranted to test our homology model in a context that more closely approximates physiological conditions.

The production of RecQ^WT and RecQ-dH were previously described in detail (Seol et al., 2016).

Forked DNA substrates (described in Supplementary file 1 Table S1) were generated and single-turnover unwinding experiments were performed as in ref (Harami et al., 2015). Global fitting kinetic analysis was performed using KinTek Global Kinetic Explorer 4.0.

The magnetic tweezers and the experimental set-up were previously described (Seol and Neuman, 2011). A mixture of DNA hairpin (3 pmol) and anti-digoxigenin (0.5 µg) was incubated in phosphate buffered solution (PBS, pH 7.5) for 20 min and introduced into the sample chamber, which was incubated overnight at 4°C. The sample chamber was washed with 1 ml of wash buffer (WB, 1X PBS, 0.02 % v/v Tween-20, and 0.3 % w/v BSA) to remove unbound DNA molecules and 40 µl of magnetic beads (MyOne, Invitrogen) were introduced to form DNA hairpin tethers. Correct DNA hairpins were identified by the sharp DNA extension change upon DNA hairpin unfolding under high force (~19 pN). Upon finding a proper DNA substrate, the chamber was washed with 200 µl of RecQ buffer (30 mM Tris pH 8, 50 mM NaCl, 5 mM MgCl₂, 0.3 % w/v BSA, 0.04 % v/v Tween-20, 1 mM DTT, and 1 mM ATP). After washing, RecQ was added at a concentration of 20–100 pM in 200 µl RecQ buffer. DNA unwinding measurements were done by tracking a DNA tethered magnetic bead in real-time with custom written routines in Labview. During the measurement, a 1 µm polystyrene stuck bead was tracked to correct sample cell drift by adjusting the sample cell position using 3-D piezo stage (Physik Instrumente) to compensate for the drift. The unwinding traces were analyzed with a custom-written T-test based algorithm in Igor Pro 6 (Wavemetrics) and the Kerssemakers step finding program in MatLab (Seol et al., 2016; Kerssemakers et al., 2006; Carter and Cross, 2005).

Usage of limiting ATP concentrations in hairpin unwinding experiments revealed a kinetic state of RecQ-dH that branched off the unwinding pathway, which enabled us to formulate a kinetic scheme relating ATP binding and hydrolysis by RecQ to its unwinding and stepping activities.

We considered three potential schemes for the coupling between the ATP binding and hydrolysis cycle and enzyme translocation. Each scheme consists of a different arrangement of four steps: DNA melting, ATP binding, hydrolysis, and RecQ translocation along ssDNA. In scheme A, RecQ binding triggers DNA melting that allows ATP binding and hydrolysis, followed by translocation of RecQ. In scheme B, DNA melting and immediate translocation of RecQ are driven by ATP hydrolysis. In scheme C, DNA melting occurs after ATP binding followed by ATP hydrolysis and translocation of RecQ.

Upon decreasing the ATP concentration from 1 mM to 5 µM, we observed sudden decreases in DNA hairpin extension, corresponding to rapid re-zipping of the hairpin that terminated in a pause followed by the resumption of unwinding. The probability and duration of these backsliding and pausing events increased with decreasing ATP concentration, consistent with an off-pathway weakly DNA-bound RecQ state in kinetic competition with ATP binding. We obtained the off-pathway probability (P_off) by counting the number of sudden extension decrease events (Figure 5—figure supplement 1A) normalized by the number of unwound base-pairs for each trace. To obtain the off-state duration (τ_off), we measured individual pause durations between sudden DNA extension decrease (re-zipping) events and the resumption of DNA unwinding (Figure 5—figure supplement 1A). For each kinetic scheme, P_off can be calculated by summing P_off for all possible pathways while the off-state duration (τ_off) can be calculated using a recursive relation (Shaevitz et al., 2005).

As some of the kinetic rates are common in P_off, τ_off, and the unwinding velocity, v, we globally fitted the experimentally measured P_off, τ_off, and v as a function of ATP concentration with the analytical expressions for P_off and τ_off for each scheme and v (see below) resulting in k_on, k_off, K_M (=k_ra/k_a), k_p, n and k_cat. The kinetic parameters are defined as follows:

In order to decrease the free parameters, we set k_a, and k_h and k_m to be 2 (s⁻¹µM⁻¹), 200 (s⁻¹), and 121 (s⁻¹), respectively, based on previous measurements (Sarlós et al., 2012) and the average unwinding time from the experimental measurements and used ${\displaystyle K_M{k}_a}$ instead of ${\displaystyle k_{ra}}$ in the fitting. The Chi-squared value (χ²) from the global fitting with scheme A is the minimum among the three as χ² values for schemes A, B, and C are 82, 604, and 9420, respectively. This suggests that ATP binding and DNA melting of RecQ may precede ATP hydrolysis, as encompassed in scheme A.

For all schemes, R is RecQ, D is DNA, n indicates the current DNA base-pair position where RecQ binds and performs its catalytic activity and D_n* represents a DNA state in which the n+1 st DNA base-pair is melted prior to translocation, which results in D_n+1.

ATPγS hydrolysis by RecQ-dH was estimated by measuring the level of thiophosphate in the reaction mixture using Malachite Green (BioAssay System; POMG-25H). First, the thiophosphate standard was generated by measuring the OD at 620 nm of 8 different thiophosphate concentrations (0, 1, 2, 4, 8, 10, 20, and 40 µM) in RecQ activity buffer. Using this standard, the amount of thiophosphate production from ATPγS hydrolysis by RecQ was estimated by measuring OD at 620 nm of the reaction mixtures containing 200 nM dT₅₄ DNA, 1 mM ATPγS and 200 nM RecQ at 0, 30 min, 1, 2, and 3 hr reaction times. Concurrently, the OD of the same reaction mixture without RecQ was also measured to correct any effects of auto-hydrolysis of ATPγS.

In a simple kinetic scenario, HRDC-dependent pausing can be considered as an off-pathway state with a single rate-limiting step, k_HP (the rate of entering an HRDC-stabilized pause state) that is in kinetic competition with the forward unwinding rate, k_step. The rate of escaping from the pause, that is the inverse lifetime of the HRDC-stabilized pause state or unbinding rate is k_-HP (Herbert et al., 2006). As there are no potential long pauses before or after the peak at 40 bp (Figure 2), we can estimate the probability and kinetics of HRDC-dependent pausing related to the sequence-dependent pausing. Although the average pause duration at 40 bp for RecQ^WT (0.16 ± 0.03 s) is comparable to that of RecQ-dH (0.14 ± 0.03 s), the long pauses (>0.5 s) were observed only in the traces of RecQ^WT but not RecQ-dH hairpin unwinding suggesting that they represent HRDC binding (Figure 6—figure supplement 1). At the 40 bp hairpin position, the step rate, k_step (inverse of pause duration for RecQ-dH) is ~7 s⁻¹ and the HRDC-stabilized pause efficiency is 0.25, estimated from the fraction of the RecQ^WT pause duration distribution that exceeds the exponential fit with k_step (Figure 6—figure supplement 1). The estimated k_HP is then 2.4 s⁻¹ and the estimated k_-HP is 0.89 s⁻¹. Based on the sequence-dependent stepping rate k_step and the calculated k_HP and k-_HP rates, HRDC-dependent pauses were calculated over the entire 174 bp hairpin. We found that the calculated pause durations and pause locations based on these simple assumptions fail to reproduce the RecQ^WT pause behavior (Figure 6—figure supplement 1). In general, the pause locations are less localized, with HRDC-stabilized pauses occurring randomly throughout the hairpin sequence, and the average lifetimes of the specific pauses are shorter, in the simulations as compared to the RecQ^WT data. This suggests that the pathway into and out-of the HRDC-stabilized pause may not be a simple single-step process, but rather a multi-phasic kinetic step. In support of this possibility, we found that the pausing duration distribution is better described by double-exponential (χ_ν² = 2.1) than single-exponential (χ_ν² = 3.2) (Figure 6—figure supplement 1B).

The sequence information indicates the 3´ to 5´ translocating strand (green region) for all DNA hairpins.

Appendix 1—figure 1

Download asset Open asset

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

We thank Yasuharu Takagi, Sarah Heissler and Jim Sellers for assistance with ATP hydrolysis assays and Dr. Jonathan Silver for comments. This research was supported by the Human Frontiers Science Program (RGY0072/2010) and the Intramural Research Program of the National Heart, Lung, and Blood Institute, National Institutes of Health (to KCN); the ‘Momentum’ Program of the Hungarian Academy of Sciences (LP2011-006/2011), ELTE KMOP‐4.2.1/B‐10‐2011‐0002, NKFIH K-116072, NKFIH ERC_HU 117680, and NKFIH K-123989 grants (to MK); and a Premium Postdoctoral Fellowship of the Hungarian Academy of Sciences (to GMH).

This is an open-access article, free of all copyright, and may be freely reproduced, distributed, transmitted, modified, built upon, or otherwise used by anyone for any lawful purpose. The work is made available under the Creative Commons CC0 public domain dedication.