The HRDC domain of E. coli RecQ helicase controls single-stranded DNA translocation and double-stranded DNA unwinding rates without affecting mechanoenzymatic coupling

DNA-restructuring activities of RecQ-family helicases play key roles in genome maintenance. These activities, driven by two tandem RecA-like core domains, are thought to be controlled by accessory DNA-binding elements including the helicase-and-RnaseD-C-terminal (HRDC) domain. The HRDC domain of human Bloom’s syndrome (BLM) helicase was shown to interact with the RecA core, raising the possibility that it may affect the coupling between ATP hydrolysis, translocation along single-stranded (ss)DNA and/or unwinding of double-stranded (ds)DNA. Here, we determined how these activities are affected by the abolition of the ssDNA interaction of the HRDC domain or the deletion of the entire domain in E. coli RecQ helicase. Our data show that the HRDC domain suppresses the rate of DNA-activated ATPase activity in parallel with those of ssDNA translocation and dsDNA unwinding, regardless of the ssDNA binding capability of this domain. The HRDC domain does not affect either the processivity of ssDNA translocation or the tight coupling between the ATPase, translocation, and unwinding activities. Thus, the mechanochemical coupling of E. coli RecQ appears to be independent of HRDC-ssDNA and HRDC-RecA core interactions, which may play roles in more specialized functions of the enzyme.

and the HRDC domain is conserved among RecQ helicases and this interaction moderates the rate of ATPase-driven activities, but its effect on mechanochemical coupling may vary among RecQ homologs, according to the differing physiological roles of the enzymes.

Results
Protein constructs. To dissect the role of the HRDC domain of Ec RecQ helicase in the enzymatic mechanism, we used three protein constructs: wild-type protein (RecQ WT ), a construct in which the previously characterized Y555A point mutation abolishes the ssDNA-binding ability of the HRDC domain (RecQ Y555A ) 16 , and a construct lacking the entire HRDC domain (RecQ 523 , comprising amino acids (a.a.) 1-523) (Fig. 1a). Circular dichroism measurements indicated proper folding of all constructs ( Supplementary Fig. S1).
ATPase suppression by the HRDC domain does not greatly affect ssDNA translocation processivity or ATPase-translocation coupling. To monitor the ATPase activity of helicase constructs, we followed the kinetics of inorganic phosphate (P i ) generation from ATP in a stopped-flow instrument. The amount of P i produced was quantified using a fluorescently labeled phosphate binding protein (MDCC-PBP) ( Supplementary Fig. S2) [24][25][26]28,29 . In the absence of DNA, traces of P i generation were linear for all constructs in the presence of saturating ATP concentration (shown for RecQ 523 in Fig. 2a). Steady-state DNA-free ATPase slopes for RecQ WT were consistent with previous results determined by other methods 25,30 . The DNA-free ATPase activity (k basal ) was not influenced by either the Y555A mutation or HRDC deletion ( Table 1).
As shown previously, ssDNA markedly activates RecQ ATPase activity 30,31 . To assess the role of the HRDC domain in ssDNA-induced ATPase activation, we rapidly mixed ATP with RecQ constructs pre-mixed with increasing amounts of dT 54 , and monitored P i release kinetics. Kinetic traces showed a lag followed by quasi-linear P i generation for all constructs (shown for RecQ 523 in Fig. 2b). The onset of the linear steady-state became faster with increasing DNA concentration with no sign of saturation, suggesting that the lag originated from an enzyme-DNA binding process induced by mixing with ATP (Fig. 2b). Our earlier study showed that ATP binds very rapidly to the enzyme 30 , ruling out that the lag in Fig. 2b originates from ATP binding. However, we recently showed that the presence of ATP, or its analogs, markedly increases the ssDNA affinity of RecQ 17 . Therefore, the lag can be explained by an ATP-induced increase in the fraction of ssDNA-bound RecQ molecules occurring upon stopped-flow mixing. This leads to a concomitant increase in the rate of P i generation until reaching the steady-state.
We determined the apparent rate constant of the ATP-induced DNA binding process (k b ) and the steady-state ATPase rate (k ss ) at each dT 54 concentration from the P i generation transients (Fig. 2b) using the model described in Supplementary equation (S1). Fits to the dT 54 concentration dependence of k ss using the Hill equation revealed ssDNA-activated k cat values (i.e. maximal k ss at saturating dT 54 concentration) in line with those determined previously using an NADH-coupled assay (Fig. 2c) 17 . The analysis indicated small positive cooperativity for the RecQ-ssDNA interaction, which was not systematically altered by the mutations (Hill coefficients were 1.7 ± 0.3, 1.0 ± 0.4 and 1.7 ± 0.3 for RecQ WT , RecQ Y555A and RecQ 523 , respectively). In agreement with our previous results, k cat was almost identical for RecQ WT and RecQ Y555A , but was 3 fold-higher for RecQ 523 (Table 1) 17 . The apparent dissociation constants for  dT 54 (K d,app,dT54 ), also determined from the dT 54 concentration dependence of k ss were slightly higher for RecQ Y555A and RecQ 523 compared to RecQ WT ( Table 1).
The apparent rate constant of ATP-induced DNA binding (k b , determined from lag kinetics using Supplementary equation (S1)) increased linearly with dT 54 concentration ( Supplementary Fig. S3). This dependence allowed another, independent means of determination of the ssDNA binding K d values of RecQ constructs during ATPase cycling ( Supplementary Fig. S3, Supplementary Table S1). The determined values (Supplementary Table S1) were slightly higher than those determined from k ss values (Table 1), but were in line with those determined previously for the RecQ WT .AMPPNP (non-hydrolyzable ATP analog) complex 17 . The mutations caused a slight reduction in the ssDNA affinity of RecQ (Table 1,  Supplementary Table S1). In summary, these results suggest that the ssDNA interaction of the HRDC domain has a minor contribution to the overall ssDNA binding affinity of RecQ, but the presence of the HRDC domain-regardless of its ssDNA binding capability-significantly suppresses the ssDNA-activated ATPase activity of the motor core. To dissect how the HRDC domain affects ssDNA translocation, we monitored the kinetics of P i generation from ATP hydrolysis during single-round translocation, which was previously shown to be suitable for the determination of translocation processivity and ATPase-translocation coupling (ATP hydrolyzed per nucleotide (nt) traveled) [24][25][26]29 .
To ensure single-round translocation conditions, we used dextran sulfate (DxSO 4 ) as a DNA-mimicking protein trap 25 . First we assessed the trapping efficiency of DxSO 4 by rapidly mixing RecQ with a pre-mixture of ATP, poly-dT and different concentrations of DxSO 4 , and monitoring P i generation using MDCC-PBP in a stopped-flow instrument. Poly-dT was used instead of oligo-dT to minimize DNA end effects. Similar to those recorded in the absence of DxSO 4 (Fig. 2b), traces showed a lag followed by linear steady-state P i generation (Fig. 3a). In the absence of DxSO 4 , the steady-state rate of P i generation was 3 times higher for RecQ 523 than for RecQ WT (Table 1). As expected, the steady-state rate markedly decreased with increasing DxSO 4 concentration (Fig. 3a). Surprisingly, we found that RecQ 523 required 30-fold higher DxSO 4 concentration for half-maximal inhibition of the ssDNA-activated steady-state ATPase activity  Table 1.
We characterized the ssDNA translocation processivity of RecQ constructs by determining the mean number of ATPase cycles during a single translocation run (< n ATP > , Table 1). In these experiments, RecQ was preincubated with poly-dT and then rapidly mixed with ATP and different concentrations of DxSO 4 in a stopped-flow instrument. The kinetics of P i generation during ATP hydrolysis was followed by MDCC-PBP. DxSO 4 concentrations were chosen from the regime where the trapping efficiency was above 95% (cf. Fig. 3b). P i generation kinetic traces comprised an exponential phase (characteristic of single-round ssDNA translocation) followed by a slow steady-state that ensued upon dissociation of RecQ from ssDNA after the translocation run (Fig. 3c). Thus, the amplitude of the exponential phase (mol P i /mol RecQ) equals < n ATP > . This amplitude decreased with increasing DxSO 4 concentration for all RecQ constructs, indicating that DxSO 4 actively facilitates RecQ dissociation from ssDNA during translocation. In the case of RecQ 523 , the amplitudes showed a shallower DxSO 4 concentration dependence than for RecQ WT and RecQ Y555A (Fig. 3d), as expected based on the experiments of Fig. 3a-b. To determine the genuine (DxSO 4 -free) < n 0 ATP > value, we extrapolated to zero DxSO 4 concentration using Supplementary equation (S3) 24,25,29 . RecQ WT and RecQ Y555A showed similar < n 0 ATP > values, whereas this value was slightly higher in RecQ 523 (Table 1), reflecting that the HRDC domain does not have a profound effect on the ssDNA translocation processivity of RecQ.
As described earlier, the ATPase-translocation coupling stoichiometry (C trans , number of ATP molecules hydrolyzed per nt traveled) can be determined from the ssDNA (oligo-dT) length dependence of P i generation amplitudes during single-round ssDNA translocation 24,25,29 . In these experiments, we preincubated RecQ with oligo-dT substrates of different lengths and then rapidly mixed these premixtures with ATP plus DxSO 4 in a stopped-flow instrument. Kinetic traces showed exponential and linear phases (Fig. 4a), similar to those observed for poly-dT (Fig. 3c). As expected, the single-round (exponential) P i generation amplitudes increased with increasing oligo-dT length, showed saturation, and their maximal value decreased with increasing DxSO 4 concentration (Fig. 4b). Fits to these data using Supplementary equation (S4) revealed a C trans value of 1.0 ± 0.1 ATP/nt and an occluded site size (b) of 13 ± 1 nt for RecQ 523 along ssDNA, both of which were independent of DxSO 4 concentration in the assessed range (2−4 mg/ml) ( Table 1). Comparison of these values with those determined earlier for RecQ WT under identical or similar conditions revealed that deletion of the HRDC domain does not significantly affect the ATPase-translocation coupling but it decreases the occluded site size (Table 1) 25,26 .
The rate of ATP hydrolysis during translocation along poly-dT (k ATP,trans , determined from the initial slopes of the exponential phase in the experiments of Fig. 3c) was similar to those determined for DxSO 4 -free steady-state ATP hydrolysis for each construct (Fig. 3a, Table 1). The HRDC domain suppresses the rate of ssDNA translocation in parallel with that of ATP hydrolysis. The above results showed that the HRDC domain does not influence ATPase-translocation coupling (C trans ), but it suppresses the rate of ATP hydrolysis during translocation (k ATP,trans ). These findings imply that the rate of ssDNA translocation (k trans (expressed in nt/s) = k ATP,trans /C trans ) must also be suppressed by the HRDC domain, in parallel with the ATPase rate. Furthermore, we found that < n 0 ATP > was not greatly affected by the HRDC domain, implying that this domain neither affects the mean processive run length (nt traveled in a single run, < n nt > = < n 0 ATP > /C trans ). As < n nt > = k trans /k off,trans where k off,trans is the rate constant of RecQ dissociation from ssDNA during translocation, one will expect that the latter parameter will be suppressed by the HRDC domain in parallel with the ATPase rate. Thus, the experimental determination of k off,trans provides an independent means of verification of the proposed mechanochemical effects.
The intrinsic (tryptophan, Trp) fluorescence intensity of each ssDNA-bound RecQ construct is markedly lower than that of the DNA-free forms 17 . This signal is thus suitable for monitoring the kinetics of dissociation of RecQ molecules (i.e., k off,trans ) from ssDNA upon completing single-round translocation 25 . We determined the k off,trans values of RecQ constructs upon rapidly mixing the RecQ-poly-dT complex with excess ATP and varying concentrations of DxSO 4 (Fig. 5a). The observed rate constants (k obs ) of the transients increased linearly with DxSO 4 concentration (Fig. 5b). The slopes and intercepts of the plots, reflecting the DxSO 4 sensitivity of the reaction and the DxSO 4 -free k off,trans value, respectively, were similar for RecQ WT and RecQ Y555A (Fig. 5b, Table 1). Importantly, the DxSO 4 -free k off,trans of RecQ 523 was 3 times higher than that of RecQ WT and RecQ Y555A , providing independent verification of the above mechanochemical considerations.
As expected based on the experiments of Fig. 3, the ssDNA dissociation k obs values of RecQ 523 were less sensitive to DxSO 4 compared with the other constructs (Fig. 5b, Table 1). In control experiments performed in the absence of ATP, k obs values were several times higher than those in the presence of ATP at the corresponding DxSO 4 concentration for all constructs (Fig. 5b). This finding indicates that the Trp fluorescence data obtained in the presence of ATP reliably report dissociation after single-round translocation. Calculated < n nt > values (= k trans /k off,trans ) were slightly higher compared to those determined from MDCC-PBP experiments (< n nt > = < n 0 ATP > /C trans ), but were practically identical for RecQ WT and RecQ 523 in both cases ( Table 1).

The HRDC domain does not affect the coupling between ATP hydrolysis and dsDNA unwinding.
To determine the coupling between the ATPase and dsDNA unwinding activities, we devised experiments to measure the rate of DNA unwinding and the rate of ATPase activity in the presence of a forked duplex DNA substrate comprising 33 bp dsDNA with two 21-nt arms. First we determined the binding affinity of RecQ constructs to this DNA substrate by fluorescence anisotropy titrations utilizing  Table 1). The slopes of the plots, characterizing the DxSO 4 -sensitivity of ssDNA dissociation, were 2.0 ± 0.3, 1.1 ± 0.1, and 0.04 ± 0.01 s −1 (mg/ml) −1 for RecQ WT , RecQ Y555A , and RecQ 523 , respectively. Symbols marked by * indicate k obs values determined in the absence of ATP. Symbols are as in Fig. 2c. AU, arbitrary units. a fluorescein label placed on the dsDNA-forming 3′ -end of one strand (Fig. 6a). Compared to RecQ WT , the binding affinity (K d, forked duplex ) was reduced about 4-5-fold in both RecQ Y555A and RecQ 523 , reflecting a modest contribution of the HRDC-ssDNA interaction (Table 1).
To determine dsDNA unwinding rates, we performed rapid kinetic single-turnover unwinding experiments in which we rapidly mixed the RecQ-forked DNA complex with ATP and excess unlabeled ssDNA trap strand in a quenched-flow instrument, and monitored the time course of fluorescently-labeled ssDNA generation from forked duplex via gel electrophoresis of reaction products (Fig. 6b). Traces comprised a short (about 0.1-s) initial lag followed by two quasi-exponential rise phases in the case of all constructs (Fig. 6c), and were analyzed based on a previously described n-step kinetic model (Supplementary equation (S5)) 32 . This model assumes that unwinding occurs as a result of n consecutive rate limiting steps that have a uniform rate constant. The model is suitable for the calculation of the macroscopic dsDNA unwinding rate (k unw ) (Supplementary equation (S5)). The lag and the rapid unwinding phase of RecQ WT was similar to that observed earlier using a Förster resonance energy transfer (FRET)-based assay 33 . We  found that k unw was slightly accelerated in RecQ Y555A and RecQ 523 compared to RecQ WT , indicating that the HRDC-ssDNA interaction moderately hinders dsDNA unwinding (Fig. 6c, Table 1).
The unwinding traces of all constructs contained an additional slow exponential phase (apparent after 5 s in Fig. 6c) ( Table 1). As discussed earlier for RecBCD and UvrD helicases, this phase may result from a fraction of enzyme molecules bound to DNA non-productively, necessitating a rate-limiting initiation of unwinding 32,34 . Alternatively, if multiple helicase molecules are initially bound to the 3′ -ssDNA overhang, the slow phase may result from the action of one helicase proceeding in the trail of the leading one 33,35 .
To assess ATPase-dsDNA unwinding coupling, we measured the steady-state ATPase rate of RecQ constructs (k ATP,unw ) during unwinding of the same forked dsDNA substrate, using a pyruvate kinase-lactate dehydrogenase (PK-LDH) coupled assay (Fig. 6d). Compared to RecQ WT , the ATPase activity of RecQ Y555A was slightly elevated, while RecQ 523 had a significantly higher (k ATP,unw ) value (Fig. 6d, Table 1). The calculated macroscopic ATPase-unwinding coupling stoichiometry (C unw (expressed as ATP hydrolyzed per bp unwound) = k ATP,unw /k unw ) was close to 1 ATP/bp for all constructs, suggesting that unwinding is tightly coupled to ATP hydrolysis and is not influenced by the HRDC domain (Table 1). Taken together, these results show similar trends to those found for ssDNA translocation: the HRDC domain hinders the rate of unwinding in parallel with that of the ATPase activity, without influencing the tight coupling between these processes.

Discussion
Although the HRDC domain has generally been considered as an auxiliary ssDNA-binding element, recent crystal structures of human BLM constructs showed that the HRDC domain can fold back onto and interact with both RecA domains, both in the presence and absence of DNA (pdb ids.: 4CGZ, 4CGD, 4O3M) (Fig. 1c) 23 . These interdomain interactions were proposed to influence the ATPase activity and the coupled ssDNA translocation and dsDNA unwinding by BLM 23 . Indeed, suppression of DNA-activated ATPase activity by the HRDC domain was observed for various RecQ helicases (Table 2) 14,17,23,36 . Moreover, in the case of Ec RecQ, ATPase suppression is independent of the ssDNA-binding ability of the HRDC domain, further indicating interdomain interactions (RecQ Y555A data in Table 1) 17 .
Importantly, however, the present study shows that the coupling of the RecQ ATPase activity to ssDNA translocation is unaffected by either the HRDC-ssDNA or the HRDC-motor core interactions. Moreover, the processivity of translocation is also unaffected by HRDC deletion in RecQ (Table 1), similar to the lack of effect of WHD-HRDC deletion on ATPase-translocation coupling and processivity in human BLM 14 .
We also found that the rate of forked duplex DNA unwinding, determined explicitly in transient kinetic experiments (Fig. 6c), is suppressed by the RecQ HRDC domain in parallel with the ATPase activity, indicating that coupling between these processes is not affected by the HRDC domain (Table 1)  that HRDC-ssDNA interactions slightly suppress the rate of unwinding, unlike that of ssDNA translocation (Table 1). Explicit kinetic rates of dsDNA unwinding have not been measured for HRDC-deletion constructs of other RecQ helicase homologs. Available steady-state unwinding kinetic 23 and end-point measurements 14,36 are not directly informative of possible changes in mechanochemical coupling, as they can be influenced by more complex features (unwinding processivity, unproductive initiation, reversal during unwinding etc.). Nevertheless, such data reflect the relative unwinding efficiencies of helicase constructs and indicate that the HRDC domain may affect ATPase-driven unwinding of forked duplex DNA by BLM in a way that is different from that in Ec and Deinococcus radiodurans (Dr) RecQ enzymes ( Table 2) 22,36 . BLM constructs lacking the HRDC 23 or both the WHD and HRDC domains 14 showed decreased efficiency of forked duplex DNA unwinding despite increased ATPase activities (Table 2). Further studies are needed to clarify whether and how the varying properties of HRDC domains of different RecQ homologs contribute to mechanochemical coupling and/or more complex dynamic processes during dsDNA unwinding (see below).
RecQ helicases are thought to translocate along ssDNA and unwind dsDNA via ATP-driven inchworm-type stepping [24][25][26] . Thus, the interactions and the relative positions of the two RecA-like domains are likely to undergo coordinated changes during the ATP hydrolytic cycle, as reported for various other helicases harboring a similar motor core 37,38 . Based on the lack of direct polar interactions between the two RecA domains of BLM, it was proposed that the interactions of the HRDC domain with both RecA domains contribute to mechanochemical coupling 23 . HRDC deletion will therefore increase the flexibility of the RecA core, thereby accelerating the ATPase cycle but possibly decreasing ATPase-unwinding coupling efficiency 23  Crystal structures of human RECQ1, a RecQ homolog naturally lacking the HRDC domain, also suggest inter-RecA communication via transient polar interactions 39 . RECQ1 bound to a dsDNA substrate with a 5-nt 3′ -ssDNA overhang exhibits no direct polar inter-RecA interactions, while the C-core RecA domain interacts with DNA 39 . On the other hand, RECQ1 bound to dsDNA with a 4-nt 3′ -ssDNA overhang contains an inter-RecA salt bridge, but shows no C-core RecA-DNA interaction (PDB code 4U7D). These findings reflect that inter-RecA coordination can be modified by interactions with the DNA substrate. Interestingly, the two RecA domains form numerous direct as well as water-mediated polar contacts in ADP-bound DNA-free RECQ1 39 . Together with the markedly (about 10 times) lower DNA-activated ATPase activity of RECQ1 40 compared to Ec RecQ and human BLM 14,41 , this finding provides further indication for the inverse relationship between inter-RecA rigidity and ATPase kinetics. Thus, the HRDC-induced suppression of the ATPase activity of various HRDC-containing RecQ helicases (Table 2) is likely brought about by HRDC-mediated coordination of the RecA domains.
Our previous work showed that the similar macroscopic mechanochemical properties of Ec RecQ and human BLM result from different underlying kinetic mechanisms 24,30 . The steady-state rate of DNA-activated ATP hydrolysis by Ec RecQ is limited by the ATP cleavage step, whereas in BLM a transition between two ADP bound states is rate-limiting 24,30 . ATP hydrolysis by RecA-type ATPases is thought to be triggered by the so-called "arginine finger" residue that interacts with the γ -phosphate of the bound nucleotide in SF1 helicases 42,43 . The putative arginine finger located in the C-core RecA domain of different RecQ helicases shows variation in its nucleotide interactions, suggesting its possible role in kinetic tuning. The arginine finger does not interact with the bound nucleotide in DNA-free Ec RecQ·ATPγ S 22,44 and human RECQ1·ADP complexes 39 . In contrast, the arginine finger of DNA-bound BLM 45 interacts with both phosphate groups of ADP (PDB code 4CGZ). These differences may indicate that the dynamic interaction of the arginine finger with the bound nucleotide during the hydrolytic cycle, which can in turn be influenced by HRDC-RecA core interactions, contributes to limiting the rate of enzymatic activity.
The HRDC domain of RecQ helicases is connected to the rest of the protein through a long and flexible loop, raising the possibility of dynamic interactions with the RecA core of the protein and/or the ssDNA regions of the DNA substrate. Amino acids involved in ssDNA binding by the isolated HRDC domain of BLM 18 are buried in crystal structures of HRDC-containing BLM constructs 23 due to the interaction with the RecA core, suggesting that the HRDC-ssDNA and HRDC-RecA core interactions are mutually exclusive.
Synthesis of our current findings on Ec RecQ with earlier data on BLM reveals that in both enzymes, depending on the structure of the DNA substrate encountered, the HRDC domain is prone to interact with ssDNA regions outside the DNA segment tracked by the RecA core. We found that the unwinding of forked DNA substrates by Ec RecQ is noticeably slowed by HRDC-ssDNA interactions (cf. RecQ WT vs. RecQ Y555A profiles in Fig. 6c of RecQ Y555A and RecQ 523 in the same experiment). Consistent with this observation, we found that the deletion of the HRDC domain drastically reduced the DxSO 4 sensitivity of ssDNA translocation kinetics by Ec RecQ (Figs 3d and 5b), suggesting that the HRDC domain greatly assists ssDNA-RecQ-DxSO 4 ternary complex formation. In parallel with these observations, the HRDC domain of BLM was suggested to bind to ssDNA regions outside the one tracked by the RecA core during unwinding of G4-containing DNA substrates 46 , and the HRDC domains of Dr RecQ and BLM were found to greatly contribute to recognition and processing of Holliday junction structures 36,47 .
On the other hand, in the presence of simple ssDNA substrates and absence of trap, the HRDC domain appears to dominantly interact with the RecA core, as inferred from the very small effects of the Y555A point mutation but marked effects of HRDC deletion under these conditions ( Fig. 2c; and trap-free dissociation rate constants in Fig. 5b; Table 1). Taken together, available data suggest that the HRDC domain may dynamically switch between ssDNA-and RecA core-interacting modes, thereby fine-tuning DNA-restructuring processes.

Materials and Methods
Reagents. All reagents were from Sigma-Aldrich unless otherwise stated. ATP was from Roche Post-reaction mixtures were incubated at 25 °C for additional 3 min and were held on ice until further processing. Samples were then loaded on 12% non-denaturing polyacrylamide gels in TBE buffer (89 mM Tris-HCl pH 7.5, 89 mM boric acid, 20 mM EDTA). Electrophoresis was carried out at 4 °C. Fluorescently-labeled DNA was detected by using a Typhoon TRIO+ Variable Mode Imager (Amersham Biosciences). The intensities of bands corresponding to the DNA substrate and unwinding products were quantified by densitometry (GelQuant Pro software (DNR Bio Imaging Ltd.)).

Data analysis.
Means ± SEM values (n = 3) are reported in the paper, unless otherwise specified.