Mechanism of RecQ helicase mechanoenzymatic coupling reveals that the DNA interactions of the ADP-bound enzyme control translocation run terminations

The processing of various DNA structures by RecQ helicases is crucial for genome maintenance in both bacteria and eukaryotes. RecQ helicases perform active destabilization of DNA duplexes, based on tight coupling of their ATPase activity to moderately processive translocation along DNA strands. Here, we determined the ATPase kinetic mechanism of E. coli RecQ helicase to reveal how mechanoenzymatic coupling is achieved. We found that the interaction of RecQ with DNA results in a drastic acceleration of the rate-limiting ATP cleavage step, which occurs productively due to subsequent rapid phosphate release. ADP release is not rate-limiting and ADP-bound RecQ molecules make up a small fraction during single-stranded DNA translocation. However, the relatively rapid release of the ADP-bound enzyme from DNA causes the majority of translocation run terminations (i.e. detachment from the DNA track). Thus, the DNA interactions of ADP-bound RecQ helicase, probably dependent on DNA structure, will mainly determine translocation processivity and may control the outcome of DNA processing. Comparison with human Bloom's syndrome (BLM) helicase reveals that similar macroscopic parameters are achieved by markedly different underlying mechanisms of RecQ homologs, suggesting diversity in enzymatic tuning.


INTRODUCTION
Transmission and maintenance of genetic information requires the action of helicases, enzymes that catalyze nucleic acid (NA) strand separation and restructuring driven by nucleoside triphosphate (NTP) hydrolysis (1). RecQ-family enzymes, belonging to helicase superfamily (SF) 2, are conserved from bacteria to humans (2). Their key roles in homologous recombination (HR)-based DNA repair are indicated by severe cancer-causing hereditary syndromes resulting from their deficiency (2,3). Escherichia coli expresses a single RecQ homolog, with key roles in the suppression of illegitimate recombination (4), repair of DNA doublestranded breaks (5,6) and stabilization of stalled replication forks (7,8).
DNA-restructuring activities of RecQ (and many other) helicases are based on mechanoenzymatic coupling of their adenosine triphosphate (ATP)-hydrolytic activity to unidirectional movement (translocation) along single-stranded (ss)DNA. Previous studies showed that E. coli RecQ helicase effectively couples its ATP-hydrolytic activity to stepping cycles along ssDNA, and stepping occurs with a moderate processivity (36-310 nt/run) (9,10). Mechanistic understanding of the coupling mechanism requires knowledge of the kinetics of substeps of the adenosine triphosphatase (ATPase) enzymatic cycle (ATP binding, hydrolysis, product release) and the coupling of these processes to the DNA interaction of the helicase (11)(12)(13)(14)(15)(16)(17)(18)(19). To this end, we elucidated the ATPase kinetic mechanism of DNAfree and ssDNA-bound E. coli RecQ using rapid transient and steady-state kinetic analysis. We found that the marked (about 100-fold) activation of the RecQ ATPase by ssDNA is practically entirely due to the kinetic enhancement of the chemical step of ATP cleavage, which is rate-limiting in all conditions. ATP hydrolysis occurs productively due to rapid release of nascent inorganic phosphate (P i ). Synthesis of current results with earlier data on the RecQ-ssDNA interaction (11) allowed the construction of a mechanoenzymatic model of the translocation-coupled ATP hydrolytic cycle. The model reveals that the ADP-bound enzymatic state is populated by a small fraction of RecQ molecules during ssDNA translocation, but it is responsible for the majority of translocation run terminations (DNA detachment events) due to its relatively rapid dissociation from ssDNA. The DNA interactions of this state, which may strongly depend on the structure of the DNA substrate being processed, may therefore control the processivity and outcome of DNA processing events. Comparison of E. coli RecQ and human Bloom's syndrome (BLM) ATPase cycles reveals marked differences between DNA activation mechanisms of RecQ homologs that probably contribute to their functional diversity.

Kinetic experiments
Unless otherwise stated, all measurements were carried out in SF50 buffer at 25 • C. Stopped-flow experiments were carried out in KinTek SF-2004 and BioLogic SFM 300 apparatuses. Quenched-flow experiments were performed in a KinTek RQF-3 instrument. Post-mixing concentrations are stated in all experiments. In experiments requiring nucleotide-free RecQ, nucleotide contamination was removed by pre-incubation with 0.02-U/ml apyrase for 15 min at 25 • C. P i liberation from ATP was followed using a fluorescently labeled P i binding protein (MDCC-PBP) (12). MDCC-PBP calibration was performed as described earlier (13,14). mdATP and mdADP were excited at 280 nm and fluorescence emission was detected through a 420-nm long-pass filter utilizing FRET (Förster Resonance Energy Transfer) from aromatic residues of RecQ.

Oxygen exchange experiments
Samples were prepared by incubating 1-M RecQ and 1-mM ATP in SF50 buffer containing 40 ± 2% 18 O water for 4 h in the absence or for 30 minutes in the presence of 2 M dT 54 , to achieve complete ATP hydrolysis. The 18 O content of water was determined by measuring 18 O incorporation into P i upon hydrolysis of PCl 5 . Samples were then mixed with an equal volume of 10% activated charcoal in distilled water and centrifuged at 14 krpm in a tabletop microcentrifuge for 1 min at 25 • C to remove protein and residual nucleotides. The supernatant was loaded on a DOWEX column pre-equilibrated with distilled water. P i was eluted with 10-mM HCl and lyophilized to dryness. Samples were then analyzed by mass spectrometry (MS).
MS measurements were performed on a Waters/Micromass Q-TOF Premier mass spectrometer (Waters/Micromass, Manchester, UK) equipped with a microelectrospray ionization source. Samples were dissolved in MilliQ water and 2-l aliquots were diluted 20 times with methanol. Diluted samples were filled into gold-coated borosilicate glass nanospray capillary tips. During the measurement the spray voltage was set to 1 kV. The instrument was operated in negative ion mode observing differentially 18

Rapid and reversible nucleotide binding to RecQ is largely unaffected by DNA
We monitored nucleotide binding (steps 1 and 4 in Figure 1A) to RecQ helicase in stopped-flow experiments by rapidly mixing the enzyme with excess amounts of fluorescently labeled 3 -(N-methylanthraniloyl)-2 -deoxy (md) nucleotides (mdATP, mdADP) in the presence and absence of dT 54 ssDNA ligand ( Figure 1B). dT 54 was premixed with RecQ at saturating concentration, as determined in our earlier experiments (9,11). The time courses of the observed md-nucleotide fluorescence increase were fitted by singleexponential functions. The observed rate constants (k obs ) of the pseudo-first-order reactions showed linear dependence on nucleotide concentration ( Figure 1B and C). The slopes of the plots, reflecting nucleotide binding rate constants (k 1 and k -4 for ATP and ADP, respectively; Figure  1A), were similar in the absence and presence of DNA (Figure 1B and C and Table 1). The y intercepts of the linear fits represent the sum of the rate constants of possible breakdown pathways of the nucleotide-bound enzyme intermediates (RecQ.ATP and RecQ.ADP, respectively). In the case of mdADP, this intercept purely represents the k 4 ADP dissociation rate constant ( Figure 1C and Table 1). In mdATP, the intercept represents the sum of the k -1 (ATP dissociation) and k 2 (ATP hydrolysis) rate constants ( Figure 1A and B and Table 1; see also below). The determined parameters were characteristic of rapid and reversible ATP and ADP binding by RecQ, which was largely independent of the presence of DNA ( Table 1).
The kinetics of dissociation of mdADP from RecQ (k 4 ) was also measured independently in 'chasing' experiments. Upon rapid mixing of the RecQ.mdADP complex (with or without dT 54 added) with excess unlabeled ATP, singleexponential transients were recorded ( Figure 1C, inset). The k obs values were in agreement with the fitted intercepts of binding experiments ( Figure 1C and Table 1).

RecQ-catalyzed ATP hydrolysis is productive due to rapid release of phosphate
We assessed the reversibility of the ATP hydrolysis step and the coupled release of P i (inorganic phosphate) (steps 2-3 in Figure 1A) by 18 O exchange measurements ( Figure 1D).
The principle of this method is that, during ATP hydrolysis, the oxygen atom of the attacking water becomes incorporated in the detaching ␥ -phosphate. If hydrolysis is performed in 18 O-labeled water, oxygen incorporation can be monitored by MS. If ATP hydrolysis occurs irreversibly (k 3 >> k -2 ), a P i species with a single 18 O atom will be produced. In the case of reversible hydrolysis, P i species with two or more 18 O atoms will appear.
Upon the complete hydrolysis of 1-mM ATP by 1-M RecQ in 40 ± 2% 18 O-labeled water, we detected 35 ± 2% single 18 O-labeled and 0.8 ± 0.1% double 18 O-labeled P i in the absence of DNA ( Figure 1D). The ratio of the amounts of the two species reflects the probability of hydrolysis reversal (k -2 /(k 3 +k -2 )) as L 2 /L 1 = f * k -2 /(k 3 +k -2 ), where L 2 and L 1 are the amounts of double-and single-18 O-labeled P i , respectively, f is the fractional 18 O-content of water, k -2 is the rate constant of hydrolysis reversal and k 3 is the rate constant of P i release ( Figure 1A). The data showed that P i release is kinetically favored over hydrolysis reversal ((k 3 +k -2 )/k -2 = 17 ± 3), reflecting productive hydrolysis of enzyme-bound ATP ( Table 1). The situation was similar in the presence of dT 54 where we detected 40 ± 3% single 18 O-labeled and 1.4 ± 0.1% double 18 O-labeled P i , defining (k 3 +k -2 )/k -2 = 11 ± 2 ( Figure 1D and Table 1). Based on these results, the amount of triple 18 O-labeled P i is expected to be less than 0.05% in all cases, which is below the detection limit. In control experiments performed in the absence of enzyme and/or ATP, 5-10% of the low P i content of the samples was 18 O-labeled, and the ratio of the amount of single and double 18 O-labeled P i was between 2 and 3 (carrying contributions from spontaneous hydrolysis and/or endogenous P i contamination) ( Figure 1D).

The ATP hydrolysis step is rate-limiting and is markedly accelerated by DNA
We monitored the transient kinetics of ATP hydrolysis (step 2 in Figure 1A) in quenched-flow experiments by rapidly mixing 4-M RecQ with 3-M or 50-M radiolabeled ␥ -32 P-ATP to ensure single-and multiple-turnover conditions, respectively (Figure 2A and B). (In this context, 'single turnover' refers to a single ATP hydrolytic cycle, as opposed to single-run conditions referring to a sequence of ATP-driven translocation/unwinding steps upon a single encounter with the DNA track; see also below.) Time courses of ATP cleavage showed no observable burst phase, suggesting that the hydrolysis step is rate-limiting in the absence of DNA. To determine the rate constant of the hydrolysis step (k 2 ), we performed global fitting analysis of the quenched-flow data, constrained by k 1 , k -4 and k 3 /k -2 values determined in other experiments (Figure 1). In addition to k 2 , nucleotide dissociation and P i release/rebinding rate constants (k -1 , k 4 , k 3 , k -3 ) were floated to obtain the best global fit. The best-fit k 2 value was close to the steadystate ATPase k cat ((9); see also below), confirming that ATP hydrolysis is rate-limiting in the absence of DNA (Table 1).
We recorded stopped-flow transient profiles of P i release (step 3 in Figure 1) using a fluorescently labeled P i binding protein (MDCC-PBP) (12). Upon rapidly mixing RecQ with excess ATP, no burst was observed before the onset of the linear steady-state phase of the reaction ( Figure 2C).
This behavior is expected as the rate-limiting ATP hydrolysis step precedes P i release. Steady-state ATPase parameters, deduced from the ATP concentration dependence of the slope of P i release records, were in line with those determined earlier in PK-LDH assays ( Figure 2D and Table 1) (9). Similar to the situation in the absence of DNA, singleand multiple-turnover quenched-flow records obtained in the presence of dT 54 showed no burst phase, suggesting that ATP hydrolysis remains rate-limiting even in the DNAbound RecQ ATPase cycle ( Figure 3A and B). However, ATP hydrolysis was markedly accelerated by DNA: global fits to the data (performed as described above) yielded an ∼100 times greater k 2 value than in the absence of DNA ( Table 1). As expected, P i release (MDCC-PBP) stoppedflow transients did not show a burst phase even in the presence of DNA, in line with rate-limiting ATP hydrolysis ( Figure 3C). Steady-state parameters of the DNA-activated RecQ ATPase activity, deduced from P i release data, were in accordance with earlier PK-LDH results ( Figure 3D and Table 1) (9). As found earlier in PK-LDH experiments (9), Hill coefficients deduced from the MDCC-PBP data were close to 1 both in the absence and presence of DNA, indicating that RecQ active sites cycle independently ( Figures  2D and 3D and Table 1).
In the MDCC-PBP stopped-flow experiments, P i release occurred irreversibly due to quasi-irreversible binding of P i  Table 1. to MDCC-PBP. To assess the genuine irreversibility of P i release from RecQ, we measured the effect of the addition of 20-mM P i on the steady-state ATPase activity using a PK-LDH-linked assay. Both the DNA-free and DNA-bound ATPase activities of RecQ were largely unaffected by P i , indicating that the effective P i dissociation constant (K Pi,eff = K 3 K 2 /(K 2 +1)) is above 20 mM both in the absence and presence of DNA (Table 1).

Mechanoenzymatic model identifies the RecQ.ADP state as the major source of run terminations controlling translocation processivity
The presented results allowed the construction of a wellconstrained kinetic model of the RecQ mechanoenzymatic cycle during ssDNA translocation ( Figure 4A). This model reveals the distribution of DNA-bound enzymatic states, while being consistent with the macroscopic translocation parameters determined earlier (9,10,15). The model includes previously determined rate constants of dissociation of RecQ from ssDNA in different nucleotide states (11). Figure 4B shows the simulated time course of the distribution of enzyme intermediates during a single processive run of ssDNA translocation. (Note that single-run conditions are different from single ATP turnover conditions ap-plied in Figures 2A and 3A.) The run is initiated by mixing the RecQ.ssDNA complex with excess ATP and is terminated by dissociation of RecQ from ssDNA dictated by the specified kinetic constants. During translocation, the majority (∼76%) of ssDNA-bound RecQ molecules populate the ATP-bound (prehydrolytic) state due to the rate-limiting ATP hydrolysis step ( Figure 4B and C). The ADP-bound state is also significantly populated (to ∼15%), as ADP release is only about five times more rapid than ATP hydrolysis ( Table 1). The different enzymatic states, however, exhibit widely varying DNA dissociation kinetics, with the ADPbound state showing the most rapid ssDNA dissociation ( Figure 4A) (11). This feature causes the majority (∼76%) of run terminations to occur in the ADP-bound state, despite the moderate steady-state abundance of this species ( Figure 4C). Thus, the model implies that the strength of the DNA interaction of the RecQ.ADP state is an important determinant of translocation processivity. Accordingly, simulations imposing a moderate (2-fold) increase or decrease in the DNA dissociation rate constant of each individual enzymatic state revealed that the processive run length is markedly more sensitive to such changes in the RecQ.ADP state than in any other enzymatic species (Figure 4D). , with rate constants representing best-fits or limit values determined from experimental data (lower bounds for k 1 , k -2 , k 3 and k 4 ; upper bound for k -3 ; associative rates given for pseudo-first-order conditions) ( Table 1). Also shown are the rate constants of RecQ dissociation from ssDNA in different nucleotide states (downward arrows; (11)), resulting in the specified run termination probabilities. We note that the simulations of Figure 4A and B predict a lower mean number of mechanochemical cycles during a translocation run (∼23; Figure 4B, inset) than that determined in our earlier study (100-320, Table 2; (9)). The difference likely arises from the condition that the longer translocation run lengths were determined for quasi-infinite length DNA substrates (M13 phage DNA, poly-dT; (9)), whereas the ssDNA interaction (11) and ATPase transient kinetic data (current study) were determined using dT 54 . However, the simulated run length of Figure 4B inset is consistent with corresponding values obtained for dT 54 (9), and those determined by others under different conditions (10,15). The simulated run length is also influenced by k 4 (ADP release rate constant), as this value controls the exit from the weakest ssDNA-interacting state ( Figure 4A). Quenched-flow experiments using unmodified ATP yielded a higher k 4 value than did mdADP experiments, indicating that the fluorescent label may slightly strengthen the RecQnucleotide interaction (Table 1). However, the main conclu-sions drawn from the simulations (greatest abundance of the ATP-bound state, run terminations controlled through the ADP-bound state) were robustly reproduced and unaffected by these variations.
Earlier we detected that, during ssDNA binding, RecQ helicase and its nucleotide complexes undergo an isomerization strengthening DNA interaction (11). The DNA dissociation rate constants in the simplified model of Figure  4A are rates of exit from the strongly DNA-bound (i.e. isomerized) form of each RecQ-nucleotide state. Simulations including all parameters of the two-step DNA interaction of RecQ produced results that were indistinguishable from those of Figure 4, as the weakly DNA-bound form was populated by only ∼1% of RecQ-DNA complexes during ss-DNA translocation.  (9), determined in conditions identical to those in the current study and (13).

Energetic and structural implications for enzymatic tuning of DNA processing
Efficient coupling of the ATP-hydrolytic activity to translocation is generally associated with a marked DNA-induced allosteric activation of the ATPase activity of helicases (1).
Here we show that, in E. coli RecQ, this activation is almost exclusively brought about by kinetic enhancement of the ATP cleavage step (Table 1). Nonetheless, all steps of the DNA-bound RecQ enzymatic cycle appear fairly reversible with the exception of P i release, which therefore appears to be the thermodynamic driving force of the ATPase cycle, with energetic implications for ssDNA translocation. Earlier we proposed that RecQ complexed with the nucleotide analog ADP.AlF 4 adopts a strongly DNA-bound, 'clamped' conformation (11). This state may be analogous to the post-hydrolytic RecQ.ADP.P i complex, implying that P i release is triggered by the formation of extensive enzyme-DNA interactions, probably involving the accessory DNAbinding (winged helix and HRDC) domains. An intriguing implication of the model of Figure 4A is that translocation run terminations are largely controlled through the ADP-bound RecQ state. Even small changes in the DNA interaction (DNA dissociation rate constant) of RecQ.ADP will markedly influence run length, unlike in other nucleotide states ( Figure 4D). Various structural features of complex physiological DNA substrates (e.g. the presence of a displaced DNA strand and/or branched structures in the vicinity of the enzyme) may significantly influence dissociation of RecQ from DNA. Thus, the proposed mechanism creates the possibility for precise tuning of the processivity of RecQ-catalyzed physiological reactions including invasion disruption, branch migration and/or nu-cleoprotein displacement. It remains to be determined how the helicase core (two tandem RecA-like domains) and the accessory DNA binding domains of RecQ contribute to DNA structure-dependent enzymatic tuning.

Relation to other SF1-2 helicase mechanisms
The reversibility of nucleotide binding appears to be a common feature of SF1-2 helicases, which can be attributed to the two tandem RecA domains forming the ATP binding pocket of these enzymes (1,16). Nucleotide binding was observed to be a two-step process both in the absence and presence of DNA in Rep (17) and RecD2 helicases (18), whereas it occurred in a single step in the NA-free cycle and in two steps in the NA-bound cycle of PcrA (19), Ms116 (20) and DbpA (21) helicases. The second step of ATP binding was proposed to be a rate-limiting conformational change triggering ATP hydrolysis in RecD2 and PcrA helicases (18). Our data do not rule out that such an isomerization may occur also in DNA-bound RecQ helicase, facilitating tight mechanochemical coupling.
Similar to RecQ, rate-limiting ATP cleavage was proposed for RecG helicase (22). In DbpA the hydrolysis step appeared to be reversible and kinetically unfavorable in the absence of RNA, but activated by the NA strand (21).
As in RecQ, rapid and irreversible P i release was proposed for PcrA (19), RecD2 (18) and RecG (22). Contrary, P i release in DbpA contributes to the limitation of the RNA-activated steady-state ATPase activity (21).

Mechanistic differences between RecQ homologs
Comparison of the mechanism of DNA activation of the ATPase cycles of E. coli RecQ and human BLM enzymes (13) reveals the following important differences between