Asymmetric Regulation of Bipolar Single-stranded DNA Translocation by the Two Motors within Escherichia coli RecBCD Helicase*

Background: RecBCD helicase is involved in repair of double-stranded DNA breaks. Results: The 5′ to 3′ ssDNA translocation rate of RecBCD is faster than the 3′ to 5′ rate in the absence of a CHI site, and the rates are coupled asymmetrically. Conclusion: RecBC controls 3′ to 5′ and 5′ to 3′ translocation, but RecD controls only 5′ to 3′ translocation. Significance: Asymmetric regulation may explain how RecBCD is regulated after CHI recognition. Repair of double-stranded DNA breaks in Escherichia coli is initiated by the RecBCD helicase that possesses two superfamily-1 motors, RecB (3′ to 5′ translocase) and RecD (5′ to 3′ translocase), that operate on the complementary DNA strands to unwind duplex DNA. However, it is not known whether the RecB and RecD motors act independently or are functionally coupled. Here we show by directly monitoring ATP-driven single-stranded DNA translocation of RecBCD that the 5′ to 3′ rate is always faster than the 3′ to 5′ rate on DNA without a crossover hotspot instigator site and that the translocation rates are coupled asymmetrically. That is, RecB regulates both 3′ to 5′ and 5′ to 3′ translocation, whereas RecD only regulates 5′ to 3′ translocation. We show that the recently identified RecBC secondary translocase activity functions within RecBCD and that this contributes to the coupling. This coupling has implications for how RecBCD activity is regulated after it recognizes a crossover hotspot instigator sequence during DNA unwinding.

RecBCD uses its helicase and nuclease activities to degrade foreign DNA, but upon encountering a crossover hotspot insti-gator ("CHI") 2 sequence (5Ј-GCTGGTGG-3Ј) within the E. coli genome, RecBCD promotes a recombination event by loading RecA onto single-stranded (ss) DNA (1,12). One current model proposes that before CHI recognition RecD is the faster motor (10,13), and the nuclease preferentially degrades the 3Ј-terminated ssDNA but cleaves the 5Ј-terminated strand infrequently (13,14). However, after CHI is recognized by the RecC subunit, DNA unwinding becomes slower, and RecB is proposed to act as the lead motor (13,15), whereas the nuclease preferentially degrades the 5Ј-terminated ssDNA. RecBCD then loads RecA protein onto the 3Ј ssDNA to initiate recombinational DNA repair (12,14,16).
It has been suggested that intersubunit communication exists between RecB and RecD (17). However, although it has been shown that the RecD and RecB motors do not act concertedly within RecBCD (10,13), it is not known whether the translocation activities of the two motors are independent or are functionally coupled. This is because the translocation rates of the individual RecB and RecD motors have not been measured within a functioning RecBCD holoenzyme but have only been inferred from DNA unwinding studies of RecBCD and variants possessing mutations within the RecB and RecD motors by assuming motor independence (10,16,18).
Using a fluorescence assay that allows direct monitoring of translocation along each of the single strands, it has recently been shown that RecBC alone (without RecD) possesses two distinct translocase activities that are controlled by the single ATPase motor within RecB (19,20). The primary translocase enables RecBC to move along ssDNA in the expected 3Ј to 5Ј direction, consistent with the directionality of RecB on ssDNA, whereas a secondary translocase facilitates translocation along the other DNA strand in the 5Ј to 3Ј direction, although this secondary translocase is not sensitive to the polarity of the ssDNA backbone (19). As such, RecBC can move along two non-complementary strands of ssDNA at the same rates in a concerted mechanism in which both translocases are tightly coupled to ATP hydrolysis within the RecB motor (20). Using an assay that enables direct measurement of the rates of ssDNA translocation of the RecBCD holoenzyme along each DNA strand in both the 3Ј to 5Ј and 5Ј to 3Ј directions, we show that the translocation rates along each DNA strand driven by the RecB and RecD motors are functionally coupled due to the action of the secondary RecBC translocase that operates within RecBCD.
Heparin stock solutions were prepared by dissolving heparin sodium salt (Sigma) in buffer M 30 and dialyzing extensively against buffer M using 3500 molecular weight-cutoff dialysis tubing. Heparin stock solutions were stored at 4°C until use, and its concentration was determined by titration with Azure A as described (21). ATP stock solutions were prepared and stored at Ϫ20°C as described (22).
Proteins-Wild type E. coli RecBCD was expressed and purified as a heterotrimer and stored in Ϫ80°C as described (23,24). The mutants RecB D1080A CD, RecB K29Q CD, RecB Y803H CD, RecBCD K177Q , and RecBCD Y567H were expressed in E. coli strain V2831 (a gift from Dr. Gerald R. Smith) in which the genes encoding wild type RecBCD have been deleted from the E. coli chromosome (17,(25)(26)(27). RecBCD and mutant enzymes were purified as described (24,28) except that the Mono Q column (Amersham Biosciences) was used as the final step to prevent possible RecBC contamination (29). RecB and RecC were purified separately as described previously (24,28). RecBC was reconstituted by mixing RecB and RecC at equimolar concentrations on ice. RecBC was shown to fully form heterodimers by sedimentation velocity experiments in buffer M. The enzymes were dialyzed against buffer M 30 at 4°C before use, and enzyme concentrations were determined spectrophotometrically using an extinction coefficient of ⑀ 280 ϭ 3.9 ϫ 10 5 M Ϫ1 cm Ϫ1 for the RecBC heterodimer and ⑀ 280 ϭ 4.5 ϫ 10 5 M Ϫ1 cm Ϫ1 for the RecBCD heterotrimer (24,30).
Oligodeoxynucleotides-Oligodeoxynucleotides, either unlabeled or labeled covalently with Cy3 or fluorescein, were synthesized and purified, and their concentrations were determined as described (22). DNA stocks were stored in buffer M 30 in the absence of Mg 2ϩ at Ϫ20°C until use. The sequences of the DNA substrates used are given in supplemental Table S1.
Stopped-flow Fluorescence Experiments-Stopped-flow kinetics experiments were performed in buffer M 250 plus 10 mM MgCl 2 at 25°C unless otherwise indicated using an SX.18MV stopped-flow apparatus (Applied Photophysics Ltd., Leatherhead, UK). RecBCD (or RecBC) was preincubated with DNA on ice for 20 min before loading into the stopped-flow syringe. Reactions were initiated by rapid 1:1 mixing of preformed enzyme-DNA complex in one syringe versus ATP and heparin in the other syringe. Final concentrations in the cuvette after mixing were 37.5 nM RecBCD, 50 nM DNA, and 7.5 mg/ml heparin in buffer M 250 plus 10 mM MgCl 2 and 5 mM ATP unless indicated otherwise. Cy3 fluorescence was excited at 515 nm, and its emission was monitored using a Ͼ570 nm-cutoff filter (Oriel Corp., Stratford, CT). Fluorescein fluorescence was excited at 492 nm, and its emission was monitored using a Ͼ520 nm-cutoff filter (Oriel Corp.).
Stopped-flow DNA unwinding experiments were performed in buffer M 250 plus 10 mM MgCl 2 at 25°C. DNA substrates with various duplex lengths possessing a high affinity binding site (a fork with a 3Ј-(dT) 6 and a 5Ј-(dT) 10 tail) were described previously (22). Final concentrations after mixing were 150 nM WT and mutant RecBCD, 50 nM DNA, and 7.5 mg/ml heparin in buffer M 250 plus 10 mM MgCl 2 and 5 mM ATP. Cy3 fluorescence was excited at 515 nm, its emission was monitored using a 570 nm interference filter (Oriel Corp.), and Cy5 fluorescence was monitored simultaneously using a Ͼ665 nm-cutoff filter (Oriel Corp.).
Analysis of DNA Translocation Time Courses-Rates of ssDNA translocation were obtained by performing translocation experiments as a function of ssDNA extension length and determining the inverse of the slope of a plot of lag time versus ssDNA extension length as described (19,31). We measure the "lag time" or duration of the lag phase as the time at the intersection of the two linear fits of the time course (see Fig. 1B). By performing this experiment with DNA substrates differing in the length of the oligodeoxythymidylate, (dT) L , extension, the translocation rate can be calculated from the inverse of the slope of a plot of lag time versus L. Table 1 lists the average rates ϮS.D. based on three or four measurements where S.D. is calculated as ͌(⌺(x Ϫ x ) 2 )/(n Ϫ 1) where x is the average rate. Any rate reported without uncertainties is based on a single set of measurements. The plot of the macroscopic translocation rate (V trans ) versus [ATP] was fit to Equation 1 using KaleidaGraph (Synergy Software, Reading, PA) to obtain K m and V max .
Analysis of DNA Unwinding Time Courses-Global non-linear least square analysis of DNA unwinding time courses was performed using Conlin (kindly provided by Dr. Jeremy Williams and modified by Dr. Chris Fischer) and the International Mathematics and Statistics Library C Numerical Libraries (Visual Numeric Inc., Houston, TX) as described previously (22,24,28,32). The entire time courses (up to 2 s) for the production of ssDNA, f ss (t), were fit to Scheme 1, using Equation 2 using numerical methods as described previously (22,24,33), where ᏸ Ϫ1 is the inverse Laplace transform operator with s as the Laplace variable, A T is the unwinding amplitude for a given DNA substrate, n is the number of DNA unwinding steps, k U is the unwinding step rate constant, and k NP is the rate constant for isomerization from a non-productive to a productive helicase-DNA complex. The average kinetic step size m is defined as the average number of unwound base pairs unwound between two consecutive rate-limiting steps with rate constant k U . The kinetic step size for unwinding, m, is obtained as the inverse of the slope of a plot of duplex length versus n.

RESULTS
DNA Substrates and Fluorescence Assay to Monitor ssDNA Translocation of RecBCD-The DNA substrates used (DNA sequences are given in supplemental Table S1) consist of a 24-base pair (bp) duplex possessing a high affinity loading site for RecBCD (a fork with a 3Ј-(dT) 6 and a 5Ј-(dT) 10 tail) or RecBC (a fork with a 3Ј-(dT) 6 and a 5Ј-(dT) 6 tail) (18,19,30,34). At the other end of the duplex, either one or both of the DNA strands possesses a (dT) L extension where L is the nucleotide length. Each DNA is labeled covalently with a Cy3 fluorophore on the terminal end of one (dT) L extension. We refer to the two DNA strands as the 3Ј-terminated strand or the 5Ј-terminated strand where the 3Ј or 5Ј designation refers to the DNA end containing the enzyme loading site. When RecBCD reaches the Cy3 fluorophore, the Cy3 fluorescence increases (19,22); hence, ssDNA translocation rates can be monitored independently either in the 5Ј to 3Ј direction or in the 3Ј to 5Ј direction depending on which (dT) L extension is labeled as described below.
All translocation experiments were performed in 250 mM NaCl (buffer M 250 at 25°C) with DNA (50 nM) in excess over RecBCD (37.5 nM). Under these conditions, RecBCD binds exclusively at the high affinity loading site and upon addition of ATP initiates unwinding of the 24-bp duplex (19) followed by ssDNA translocation along the (dT) L extension (supplemental Fig. S1 and supplemental Discussion). An excess of heparin (7.5 mg/ml after mixing) added with the ATP serves to trap any free RecBCD, thus ensuring a single round of translocation (i.e. no rebinding of dissociated enzyme). Fig. 1B shows a stopped-flow fluorescence time course from an experiment with DNA substrate (L ϭ 51 nucleotides) showing the expected "lag phase" reflecting the average time for RecBCD to translocate (3Ј to 5Ј) to the Cy3 label followed by a rapid increase in Cy3 fluorescence when the population of RecBCD reaches the Cy3 fluorophore. The final slow decrease in Cy3 fluorescence reflects RecBCD dissociation from the DNA.
We measure the lag time or duration of the lag phase as the time at the intersection of the two linear fits of the time course (Fig. 1B). By performing this experiment with a set of DNA substrates differing in ssDNA (i.e. (dT) L ) extension length but with a constant short duplex length (24 bp), the average rate of translocation along the (dT) L extension can be calculated from the inverse of the slope of a plot of lag time versus L (19,31). The amplitude of the fluorescence peak is dependent upon the frac-tion of DNA labeled with Cy3 as well as the processivity of the translocase. The Cy3 labeling of the DNA can vary from 75 to 98%. These differences have no effect on the lag times or the calculated rates of translocation (supplemental Fig. S2); however, they do preclude accurate estimates of translocation processivity.
5Ј to 3Ј ssDNA Translocation of RecBCD Is Faster than 3Ј to 5Ј Translocation-We first used the double (dT) L extension substrates I and II ( (Table 1). Under these conditions, the 5Ј to 3Ј rate (1922 Ϯ 72 nt/s) is faster than the 3Ј to 5Ј rate (1409 Ϯ 109 nt/s). Experiments comparing RecBC with RecBCD under these same conditions (buffer M 250 ) with the same DNA substrate (III) indicate that the 3Ј to 5Ј translocation rate of RecBCD (1627 Ϯ 103 nt/s) is faster than that of RecBC (909 Ϯ 51 nt/s) (19), and both are faster than the 3Ј to 5Ј translocation rate of the RecB motor alone (ϳ800 nt/s) (19) (supplemental Table S2 and supplemental Fig. S4A). Hence, the 3Ј to 5Ј translocation rate of RecBC is enhanced upon forming a complex with RecD. The 5Ј to 3Ј rate is also faster than the 3Ј to 5Ј rate with the  Table S3). Experiments performed with the double (dT) L extension substrates (I and II) under other conditions (ATP (5 mM) in excess over [Mg 2ϩ ] (1 mM) or in buffer T (supplemental Table S4)) also show the 5Ј to 3Ј rate to be faster than the 3Ј to 5Ј rate.
The 3Ј to 5Ј and 5Ј to 3Ј Translocation Rates Are Coupled Asymmetrically-We next examined whether the 3Ј to 5Ј and 5Ј to 3Ј ssDNA translocase activities within RecBCD are independent or show coupling. We examined the behavior of two well studied mutants (25,26), RecB K29Q CD and RecBCD K177Q , in which the conserved Lys within the Walker A sequence (helicase motif I) is mutated to Gln, thus eliminating the ATPase activity of the mutated motor. Hence, only the RecD ATPase is active in RecB K29Q CD, and only the RecB ATPase is active in RecBCD K177Q . As expected, no 3Ј to 5Ј translocase activity was detected for RecB K29Q CD, and initiation of DNA unwinding and translocation by RecB K29Q CD requires the 5Ј-(dT) tail of the enzyme loading site to be at least 5Ј-(dT) 10 (supplemental Fig. S5), consistent with the need for a 5Ј-ssDNA tail to reach the RecD subunit (18,30,34).
The 3Ј to 5Ј translocase activity of RecBCD K177Q (Fig. 2C) and the 5Ј to 3Ј translocase activity of RecB K29Q CD (Fig. 2D) were examined at saturating ATP (5 mM). Loss of the RecD ATPase activity has only a minor effect (if any) on the 3Ј to 5Ј translocase rate (1289 Ϯ 14 nt/s for RecBCD K177Q versus 1409 Ϯ 109 nt/s for WT RecBCD in Table 1). In contrast, loss of the RecB ATPase activity causes a nearly 40% reduction in the 5Ј to 3Ј translocase rate (1162 Ϯ 11 nt/s for RecB K29Q CD versus 1922 Ϯ 72 nt/s for WT RecBCD in Table 1). This asymmetry indicates that the RecB motor influences not only the 3Ј to 5Ј translocation rate as expected because RecB is a 3Ј to 5Ј translocase (19) but also the 5Ј to 3Ј translocation rate. However, the RecD motor only affects the 5Ј to 3Ј translocation rate. These conclusions are supported by experiments performed on the single extension DNA substrates III and IV (supplemental Figs. S5 and S6 and supplemental Table S3).
We next examined translocation as a function of ATP concentration at constant [Mg 2ϩ ] (10 mM). Fig. 2E shows little effect of knocking out the RecD ATPase (RecBCD K177Q ) on the 3Ј to 5Ј translocation rate. In contrast, deactivating the RecB ATPase (RecB K29Q CD) reduces the 5Ј to 3Ј translocation rate at all [ATP] (Fig. 2F). Therefore, the asymmetric coupling of the  All rates were determined in buffer M 250 plus 10 mM Mg 2ϩ , 5 mM ATP, and 7.5 mg/ml heparin. DNA I and DNA II were used to determine 3Ј to 5Ј and 5Ј to 3Ј rates, respectively. Three translocases, RecBC primary (BC 1°) (in red), RecBC secondary (BC 2°) (in blue), and RecD (D) (in green), are indicated in the schematic.
* Time courses are shown in supplemental Fig. S3.

Coupling of the Bipolar Translocation Rates within RecBCD
two translocation rates driven by the RecB and RecD motors is maintained at all [ATP]. Furthermore, deactivating the ATPase of one motor does not affect the K m for ATP of the other motor. Therefore, the reduction in the 5Ј to 3Ј translocase rate upon inactivating the RecB ATPase (RecB K29Q CD) does not appear to result from an effect on the ATPase activity of the RecD motor.
To further investigate the coupling of the two translocation rates, we examined two additional mutants, RecB Y803H CD and RecBCD Y567H . These mutations are in equivalent positions within helicase motif VI of the RecB and RecD motors (17,35). Previous studies showed that the DNA unwinding rate of RecB Y803H CD is slowed significantly (17), and this mutation also slows the rates of both the primary and secondary translocases within RecB Y803H C (20). As anticipated, the translocation rate of RecB Y803H CD in the 3Ј to 5Ј direction (Fig. 3A) is reduced to 441 Ϯ 13 from 1409 Ϯ 109 nt/s for WT RecBCD. In addition, the translocation rate of RecB Y803H CD in the 5Ј to 3Ј direction (Fig. 3B) is also reduced to 1454 Ϯ 75 from 1922 Ϯ 72 nt/s for WT RecBCD (Table 1). Therefore, this single mutation within RecB shows reductions in both the 3Ј to 5Ј and 5Ј to 3Ј translocation rates, providing further evidence that the RecB motor regulates both translocation rates.
We next examined the effect of the equivalent mutation (Y567H) within RecD. As anticipated, the 5Ј to 3Ј translocation rate of RecBCD Y567H (Fig. 3C) is reduced to 1046 Ϯ 12 from 1922 Ϯ 72 nt/s for WT RecBCD. However, the 3Ј to 5Ј translo-cation rate of RecBCD Y567H (Fig. 3D) is unchanged (1412 Ϯ 12 nt/s) relative to WT RecBCD (1409 Ϯ 109 nt/s). Hence, reducing the rate of RecD translocation only reduces the 5Ј to 3Ј translocation rate (Table 1). These effects were also observed using the single extension DNA substrates (supplemental Table  S3). These experiments show asymmetric coupling within RecBCD such that the 3Ј to 5Ј translocation rate depends solely on the RecB motor, whereas the 5Ј to 3Ј translocation activity is controlled by both the RecB and RecD motors.
The Secondary Translocase Activity of RecBC Is Functional within RecBCD K177Q and RecBCD-Although the 5Ј to 3Ј translocation activity of RecBCD depends on both the RecB and RecD motors, this coupling does not appear to be due to communication between their ATP binding sites. Recently, Wu et al. (19) have shown that RecBC has both a primary (3Ј to 5Ј) translocase activity and a secondary translocase activity that can operate in the 5Ј to 3Ј direction along the other DNA strand, although it is not sensitive to the ssDNA polarity. Hence, a possible explanation for the asymmetric coupling of the translocase activities driven by the RecB and RecD motors is that the secondary translocase activity of RecBC is functional within RecBCD so that the 5Ј to 3Ј translocation rate is controlled by both RecD and the secondary RecBC translocase.
To determine whether the secondary RecBC translocase is functional within RecBCD, we examined whether any 5Ј to 3Ј translocation activity is observed for RecBCD K177Q . Because RecD K177Q has no ATPase activity, any 5Ј to 3Ј translocation activity observed for RecBCD K177Q could result only from a functional secondary RecBC translocase activity. Using the single extension DNA substrate IV, the results in Fig. 4A Fig. S3).

. Slowing the RecB motor (RecB Y803H ) slows both the 3 to 5 and 5 to 3 translocation rates, whereas slowing the RecD motor (RecD Y567H ) slows only the 5 to 3 translocation rate.
A, monitoring 3Ј to 5Ј ssDNA translocation of RecB Y803H CD using DNA I. Inset, rate ϭ 441 Ϯ 13 nt/s. B, monitoring 5Ј to 3Ј ssDNA translocation of RecB Y803H CD using DNA II. Inset, rate ϭ 1454 Ϯ 75 nt/s. C, monitoring 3Ј to 5Ј ssDNA translocation of RecBCD Y567H using DNA I. Inset, rate ϭ 1412 Ϯ 12 nt/s. D, monitoring 5Ј to 3Ј ssDNA translocation of RecBCD Y567H using DNA II. Inset, rate ϭ 1046 Ϯ 108 nt/s. AU, arbitrary units. To examine whether the secondary RecBC translocase is functional within WT RecBCD, we used DNA substrate V (Fig.  4B). This substrate contains a 3Ј-3Ј-phosphodiester linkage in the top strand (red X) after the 24-bp duplex that reverses the DNA backbone polarity and thus should block the 5Ј to 3Ј translocation activity of the RecD motor at that point. However, such a reverse polarity linkage does not block the RecBC secondary translocase activity; hence, any Cy3 signal observed using this DNA substrate could only result if the secondary RecBC translocase activity functions within RecBCD. Fig. 4B (red curve) shows 5Ј to 3Ј translocase activity of RecBCD on this substrate. In contrast, RecB K29Q CD shows no 5Ј to 3Ј translocase activity on this same substrate (Fig. 4B, blue curve), indicating that the RecD motor cannot bypass the reverse polarity linkage. Hence, the 5Ј to 3Ј translocation activity observed for WT RecBCD on this DNA requires an active RecB ATPase, consistent with translocation being due to the secondary translocase activity within RecBC.
The RecB and RecD Slow Translocation Mutants Also Reduce DNA Unwinding Rates-RecBCD is a dual motor helicase, and it has been shown that inactivating either motor by mutating the ATPase site (RecB K29Q or RecD K177Q ) decreases the observed rate of duplex DNA unwinding (16,18). Our current results as well as a previous study (20) show that a mutation within RecD (Y567H) that slows its 5Ј to 3Ј ssDNA translocation rate does not affect the 3Ј to 5Ј translocation rate of RecB, whereas a mutation within RecB (Y803H) that slows its primary 3Ј to 5Ј translocation rate also slows the secondary 5Ј to 3Ј ssDNA translocation rate of RecBC. We therefore examined how these mutations affect the DNA unwinding rates of RecBCD.
We examined the DNA unwinding kinetics using a stoppedflow fluorescence assay under single round DNA unwinding conditions (22,24,28,36) under the same solution conditions used in the ssDNA translocation studies. The time courses were analyzed to obtain average macroscopic rates of DNA unwinding as well as a kinetic unwinding step size using a uniform n-step sequential kinetic model (Scheme 1 and Equation 2 under "Experimental Procedures") as described (22,24,28,36). DNA unwinding was performed using a series of DNA substrates with varying duplex lengths with each possessing a high affinity RecBCD binding site (a fork with a 3Ј-(dT) 6 and a 5Ј-(dT) 10 tail) on one end. The time courses are shown in supplemental Fig. S8, and the best fit kinetic parameters are summarized in Table 2.
We first examined WT RecBCD as well as the RecB K29Q CD and RecBCD K177Q mutants that knock out the respective ATPase activities of the individual motors. In agreement with previous reports at high Mg 2ϩ concentration (16,18) (Table 2). Hence, both of the slow motor mutants also slow DNA unwinding, although slowing the RecB motor has a greater impact than slowing the RecD motor. Therefore, both motors facilitate DNA unwinding, and the DNA unwinding rate of the complex is not solely determined by either motor.
DNA unwinding kinetic step sizes ranging from 3.1 Ϯ 0.7 to 5.3 Ϯ 1.7 bp were found for all of the RecBCD mutants (Table  2), and these are in the same range as observed previously for RecBCD and RecBC (22,37,38). This suggests that the mutant RecBCD enzymes use the same kinetic mechanism for DNA unwinding regardless of the speed of the individual motors and that the decreased unwinding rate is due to the decreased ssDNA translocation rate.

The Bipolar Translocation Rates within RecBCD Are Coupled Asymmetrically Due to the Presence of Three Translocase
Activities-The methods reported here enabled us to monitor directly the rates of ssDNA translocation by RecBCD in both the 3Ј to 5Ј and 5Ј to 3Ј directions. We infer the following conclusions from the results of our studies, summarized in Table 1.
First, under all conditions examined, including a range of ATP and Mg 2ϩ concentrations (supplemental Table S4), the 5Ј to 3Ј translocase rate is always faster than the 3Ј to 5Ј translocase rate on ssDNA without a CHI sequence. This agrees with previous DNA unwinding studies showing that an ssDNA loop forms ahead of RecBCD in the 3Ј-terminated strand along which the primary RecB translocase operates (10,13). Second, the sec-

rates and kinetic parameters for WT and mutant RecBCD
All parameters except RecBC were determined in buffer M 250 plus 10 mM Mg 2ϩ , 5 mM ATP, and 7.5 mg/ml heparin. Reaction time courses were fit to Scheme 1 using non-linear least square methods as described previously (22). DNA unwinding rates were calculated by k U timing step size m. Standard deviations are calculated based on three or four observations. ondary translocase activity previously identified within RecBC (19) also functions within RecBCD. Hence, RecBCD possesses three translocase activities: the primary RecBC (3Ј to 5Ј) translocase, the RecD (5Ј to 3Ј) translocase, and the recently identified secondary RecBC translocase that normally operates 5Ј to 3Ј and is driven by the RecB motor but is insensitive to ssDNA polarity (19). Third, the 3Ј to 5Ј and 5Ј to 3Ј translocase activities within the RecBCD holoenzyme are coupled asymmetrically. That is, the 3Ј to 5Ј translocation rate is regulated only by the RecB motor, whereas the 5Ј to 3Ј translocation rate is regulated by both the RecD and RecB motors due to the presence of the secondary RecBC translocase activity. If the RecD translocase activity is eliminated as in RecBCD K177Q , then the 3Ј to 5Ј and 5Ј to 3Ј translocation rates become equal because both rates are tightly controlled by the single RecB ATPase motor as shown for RecBC (19,20). Separately, neither the RecD nor the secondary RecBC translocation rate (5Ј to 3Ј) appears to be faster than the primary RecBC translocation rate (3Ј to 5Ј) ( Table 1). Therefore, before CHI, the 5Ј to 3Ј translocation rate of RecBCD may be faster simply due to the fact that two translocase activities (RecD and secondary RecBC) operate in the 5Ј to 3Ј direction, whereas only one translocase activity (primary RecBC) operates in the 3Ј to 5Ј direction. How the two translocase activities (RecD and secondary RecBC) coordinate to move at a rate (1922 Ϯ 72 nt/s) that is faster than the rates of the individual translocases (1144 Ϯ 62 and 1162 Ϯ 11 nt/s) remains to be determined. As discussed (19), the likely candidates are either the "arm" of RecB that interacts with the duplex region of DNA in the crystal structure (34) and/or the "dead nuclease domain" within RecC that interacts with the 5Ј-ssDNA tail that eventually threads into the RecD motor (Fig. 1A). The fact that the secondary RecBC ssDNA translocase activity is insensitive to the ssDNA backbone polarity (19) differentiates it from the canonical primary translocase activities of RecB and RecD. This suggests that the secondary translocase activity may normally operate during DNA unwinding as a dsDNA translocase ahead of the fork through the RecB arm.

Enzyme
Relationship between ssDNA Translocase Activity and Helicase Activity-Based on the experiments reported here, we can compare the DNA unwinding rates of RecBCD, RecBC, and the various mutants with both the 3Ј to 5Ј and 5Ј to 3Ј ssDNA translocation rates, and these are shown in Fig. 5. Fig. 5A sug-gests that the DNA unwinding rate correlates with the 3Ј to 5Ј ssDNA translocation rate, which represents both the primary and secondary RecBC ssDNA translocation rates because these are equal. Fig. 5B shows no correlation between the 5Ј to 3Ј ssDNA translocation rate, which represents contributions from both the secondary RecBC and RecD translocase activities, and the DNA unwinding rate. This suggests that the DNA unwinding rate under these conditions is influenced by the primary and secondary RecBC translocase rates and not by the RecD translocase rate. The fact that the RecB K29Q CD rate falls on the line in Fig. 5A even though only RecD can contribute to translocation may be fortuitous.
Previous ensemble studies (18) have attempted to infer information about the relative rates of the RecB and RecD motors by comparing DNA unwinding rates of WT RecBCD with the same mutants that we have studied here containing individual motor knockouts, RecB K29Q CD and RecBCD K177Q . Those studies (16,18) showed and our own experiments (  (16). Hence, it was also inferred from those studies that the relative rates of the two motors could be reversed by changing the ratio of ATP to Mg 2ϩ in the buffer (16). In the experiments reported here that monitor ssDNA translocation rates directly rather than DNA unwinding rates, we observe that the 5Ј to 3Ј rate is always faster than the 3Ј to 5Ј rate in WT RecBCD regardless of the ATP to Mg 2ϩ ratio (supplemental Table S4). The conclusions concerning the relative rates of the two motors drawn from the DNA unwinding studies were based on the assumption that the translocation rates of RecB and RecD are independent. However, we show here that the translocation rates are not independent but are coupled. The main reason that the DNA unwinding rate is reduced more when the RecB ATPase is inactivated is likely that RecB motor inactivation knocks out both the primary and secondary translocase activities, whereas inactivating the RecD ATPase only affects the 5Ј to 3Ј translocation rate. Thus, the fact that the 5Ј to 3Ј translocation rate is regulated by both RecB and RecD makes it impossible to infer rates for the two motors from DNA unwinding rates alone.
The observation that DNA unwinding activity is maintained even in mutants that knock out the ATPase activities of either of the individual motors (i.e. RecB K29Q CD and RecBCD K177Q ) has also been interpreted as an indication that either motor (RecB or RecD) can function as the helicase. However, DNA unwinding by RecBCD requires more than just the ability to translocate along ssDNA, and it is possible (likely) that DNA unwinding (bp separation) and translocation are separate processes (20). It is possible that base pair separation occurs in multiple (4 -6) base pair steps via the ATP-independent melting reaction that has been observed for initiation and that uses only the free energy from RecBCD (37) or RecBC (38) binding to the DNA end. If this mechanism applies, the ATPase activities of the motors would be used only for directional translocation along the ssDNA that is formed by the protein-induced DNA melting reaction. In such a mechanism, DNA melting is facilitated by a part of RecBC that is independent of its ATPase activities and thus remains functional regardless of whether one motor is knocked out. As such, the only role of the ATPase activity is to effect ssDNA translocation to move the enzyme to the new ss/dsDNA junction at which point the enzyme is reset so that the binding free energy can be used again to melt out another 4 -6-bp region of the DNA. Hence, it is not likely that the helicase activity switches between motors when one motor is knocked out but rather that a region of RecBC distinct from its ATPase provides the DNA melting activity and that the RecB and RecD ATPase motors only provide translocase activity, and this can be accomplished (albeit at different rates) by either only one or both motors.
How Might CHI Recognition Affect the Three Translocase Activities within RecBCD?-The finding that RecBCD possesses three translocase activities on non-CHI-containing DNA substrates raises the question of how these might be affected after RecBCD recognizes a CHI site. One current view is that, before CHI, RecD is the faster motor and thus operates as the helicase, whereas after CHI, recognition by RecB becomes the faster motor and thus serves as the helicase (13). However, the results reported here suggest that the faster 5Ј to 3Ј translocation rate of RecBCD before CHI is due to the activity of two translocases (RecD plus secondary RecBC) working 5Ј to 3Ј, whereas only one translocase (primary RecBC) works 3Ј to 5Ј (Fig. 6).
Post-CHI RecBCD unwinds DNA at half its pre-CHI rate (13), similar to the rate of RecBCD K177Q . As suggested previously (14,39), we also propose that post-CHI the RecD motor is inactivated and thus no longer contributes its translocase activity to RecBCD, resulting in a decrease in unwinding rate. However, our results (19,20) suggest that the primary and secondary RecBC translocase activities will both remain operational because they are powered by the RecB ATPase and will con-FIGURE 6. Schematic of RecBCD-DNA complexes pre-and post-CHI. A, RecBCD-DNA initiation complex. RecD (green) and the secondary RecBC translocation site (blue) operate on the 5Ј-ssDNA tail, and the primary RecBC translocase site (red) operates on the 3Ј-ssDNA tail. B, pre-CHI DNA unwinding by RecBCD. The 5Ј to 3Ј rate is faster due to the action of two translocases (RecD (green) and the secondary RecBC (blue)) versus only one translocase (primary RecBC) acting in the 3Ј to 5Ј direction, resulting in a loop in the 3Ј-terminated strand ahead of the enzyme. C, post-CHI DNA unwinding by RecBCD. Upon deactivation of RecD, the 3Ј to 5Ј and 5Ј to 3Ј translocation rates are equal due to the concerted action of the RecBC secondary (5Ј to 3Ј) (blue) and primary (3Ј to 5Ј) (red) translocases, resulting in maintenance of the loop formed in the 3Ј-terminated strand ahead of the enzyme. Two alternative pathways (I versus II) result depending on whether CHI remains bound to RecC (I) wherein a second loop would form at the interface between RecB and RecC as proposed (13,41). Alternatively, if the 3Ј-tail does not remain bound to RecC (II), then the 3Ј-ssDNA will pass through RecC, and a second loop would not form.
tinue to move at the same rates post-CHI; hence, both unwound DNA strands will be fed through RecBCD at the same rate. As a result, the ssDNA loop in the 3Ј-tailed strand that was formed ahead of the fork due to the faster pre-CHI 5Ј to 3Ј translocase activity (Fig. 6B) would remain in place (Fig. 6C). This ssDNA loop could serve as the site for loading RecA protein. As shown in Fig. 6C, two alternative pathways could ensue depending on whether the CHI site remains bound to RecC. If the CHI site remains bound to RecC as has been proposed (13), a second ssDNA loop would then be formed at the interface between RecB and RecC (Fig. 6C, I), and this could serve as the loading site for RecA protein (13). Alternatively, if the 3Ј tail does not remain bound to RecC, then the 3Ј-ssDNA tail would pass through RecC, and a second loop would not form (Fig. 6C,  II).
The 3Ј to 5Ј Translocation Rate of RecB Is Enhanced by Both RecC and RecD Binding-The experiments reported here and in previous studies (19) allow us to compare the primary 3Ј to 5Ј ssDNA translocation rates of RecB, RecBC, RecBCD K177Q , and RecBCD (supplemental Table S2). Formation of a RecBC complex increases the 3Ј to 5Ј translocase rate of RecB slightly from ϳ800 to ϳ1000 nt/s. However, a more substantial increase in the 3Ј to 5Ј rate to ϳ1600 nt/s occurs upon interaction of RecBC with the RecD motor. In fact, even RecD K177Q with an inactive ATPase enhances the 3Ј to 5Ј translocation rate of RecBC to ϳ1400 nt/s. Based on the crystal structures of RecBCD (34,40), there does not appear to be any direct contact between RecD and RecB, although not all of the residues are observable within RecD. Therefore, the observation that RecD binding increases the rate of RecB translocation appears to result from a long range communication through the RecC subunit. This allosteric effect may occur via the same pathway but in reverse when RecC recognizes a CHI sequence, resulting in inhibition or deactivation of the translocase activity of RecD.