Chi hotspot control of RecBCD helicase-nuclease: Enzymatic tests support the intramolecular signal-transduction model

Abstract

Repair of broken DNA is essential for life; the reactions involved can also promote genetic recombination to aid evolution. In Escherichia coli, RecBCD enzyme is required for the major pathway of these events. RecBCD is a complex ATP-dependent DNA helicase with nuclease activity controlled by Chi recombination hotspots (5’-GCTGGTGG-3’). During rapid DNA unwinding, when Chi is in a RecC tunnel, RecB nuclease nicks DNA at Chi. Here, we test our signal transduction model – upon binding Chi (step 1), RecC signals RecD helicase to stop unwinding (step 2); RecD then signals RecB (step 3) to nick at Chi (step 4) and to begin loading RecA DNA strand-exchange protein (step 5). We discovered that ATP-γ-S, like the small molecule RecBCD inhibitor NSAC1003, causes RecBCD to nick DNA, independent of Chi, at novel positions determined by the DNA substrate length. Two RecB ATPase-site mutants nick at novel positions determined by their RecB:RecD helicase rate ratios. In each case, we find that nicking at the novel position requires steps 3 and 4 but not step 1 or 2, as shown by mutants altered at the intersubunit contacts specific for each step; nicking also requires RecD helicase and RecB nuclease activities. Thus, altering the RecB ATPase site, by small molecules or mutation, sensitizes RecD to signal RecB to nick DNA (steps 4 and 3, respecitvely) without the signal from RecC or Chi (steps 1 and 2). These new, enzymatic results strongly support the signal transduction model and provide a paradigm for studying other complex enzymes.

Introduction

When their DNA is broken, cells must repair it or they die. Faithful repair occurs by the reactions that can promote homologous recombination if the cell is heterozygous for genetic markers. The enzymatic reactions of homologous recombination are thus important for continuation of life of the individual and for evolution of the species. The major pathway of repair of DNA double-strand breaks (DSBs) and recombination in Escherichia coli requires the large, multifunctional RecBCD enzyme [1–4]. This complex, three-subunit enzyme has eight known enzymatic activities, including two ATP-dependent helicases of opposite DNA strand polarities: slower 3’ → 5’ translocation by RecB and faster 5’ → 3’ translocation by RecD [5, 6]. RecB also contains a nuclease domain, and RecC contains a tunnel through which the 3’-ended strand passes and in which the Chi recombination hotspot (5’-GCTGGTGG-3’) is recognized during rapid DNA unwinding (Fig. 1A) [7–10].

Figure 1. Models for RecBCD structure, RecBCD-promoted recombination, and RecBCD’s control by Chi recombination hotspots.

(A) A cryoEM structure (PDB 6SJB) of RecBCD bound to hairpin-shaped DNA with long 5’ and 3’ ss tails, the latter containing Chi [10]. RecB (orange), RecC (blue), and RecD (green) have the indicated functional domains. Four amino acids (two are shown in red) of the RecB nuclease domain have charges opposite to those of four amino acids (yellow) on the RecC loop (cyan); each set is important for Chi hotspot activity and can be computationally fit to the other by swinging the nuclease domain on the RecB tether (magenta line delimited by arrows) [26]. (B) Model for genetic recombination promoted by RecBCD enzyme. RecBCD binds a dsDNA end (a) and unwinds the DNA, producing a loop and two tails (b) which can anneal to produce two growing loops (c). Upon encountering Chi (5’-GCTGGTGG-3’) on the top strand, RecBCD nicks that strand (d) and loads RecA onto it (e). The ssDNA-RecA filament invades intact homologous DNA, forming a D-loop (f), which can be converted into a Holliday junction (g) to produce reciprocal recombinants; alternatively, DNA synthesis (h) primed in the D-loop can lead to non-reciprocal recombinants. From [23]. (C) Signal transduction model for Chi’s control of RecBCD. When Chi is recognized in the RecC tunnel (step 1), RecC signals RecD to stop unwinding (step 2). When stopped, RecD signals RecB to swing Nuc (step 3), cleave the DNA (step 4) and begin loading RecA (step 5) (see Fig. 1B). Modified from [23]. (D) RecB nuclease domain (Nuc) swing model. Without DNA, RecBCD has the conformation of that in a crystal structure (PDB 1W36) [8] with Nuc at the RecC tunnel exit. Upon DNA binding, Nuc (grey) swings to the RecC loop; upon Chi’s encounter, Nuc swings back to the RecC tunnel exit and nicks DNA at Chi. A further change, such as Nuc rotation, allows it to load RecA. From [20].

After encountering Chi, RecBCD’s activities are dramatically changed in a manner that requires proper interaction and co-ordination among the three subunits. RecBCD cleaves the strand with Chi a few nucleotides to the 3’ side of Chi and begins to load the DNA strand-exchange protein RecA onto this newly generated 3’-ended single-stranded DNA (Fig. 1B) [11–13]. DNA unwinding transiently stops for a few seconds and then continues at a slower rate [14]. After nicking DNA at Chi, RecBCD loses the ability to cut at a subsequently encountered Chi, either on the same DNA or on another DNA molecule [15]. Failure to cut on the same DNA can account for the reduction of intracellular Chi hotspot activity by another Chi site inserted in cis [16]. At some point, such as at the end of the DNA substrate for purified RecBCD, the three subunits dissociate and remain inactive for an hour or more [17]. This inactivation is Chi-dependent and can account for Chi’s action in trans in cells (i.e., on another DNA molecule) [17–19]. Conformational changes are also induced by Chi, as evidenced by RecBCD’s increased sensitivity, at limited sites, to proteases [20]. These sites are protease-sensitive before DNA binding, become more resistant upon DNA binding and during initial unwinding, and become sensitive again after Chi’s encounter during continued unwinding.

These multiple responses to Chi require the RecC tunnel (for Chi recognition) and RecB (for Chi nicking). RecD is also required because recD nonsense and deletion mutants lack Chi’s stimulation of recombination (Chi hotspot activity), Chi nicking activity, and all detectable nuclease activity, but they retain DNA unwinding activity [21, 22]. These and other observations led to the “intramolecular signal transduction” model (Fig. 1C) [23, 24]. In this model, when Chi is recognized in the RecC tunnel, RecC signals RecD to stop unwinding. This cessation of unwinding, in turn, prompts RecD to signal the RecB nuclease domain (Nuc) to nick the DNA near Chi. This model is supported by mutations altering each of the points of intersubunit contact [24], determined in crystal and cryoEM structures [8–10, 25], between RecC-RecD, RecD-RecB, and RecB-RecC. These mutants lack Chi hotspot activity and presumably block the signal transduction at the indicated steps (Fig. 1C). The related nuclease swing model (Fig. 1D), which accounts for the changes of protease-sensitivity noted above, is supported by mutations in the 19-amino-acid tether connecting Nuc to the RecB helicase domain and in the 40-amino-acid “RecC loop” (Fig. 1A) at which Nuc is postulated to be in its inactive state during unwinding before Chi [20, 26, 27]. The nuclease swing model is also supported by physical analyses – the limited proteolysis noted above and small angle X-ray scattering (SAXS) analysis showing, upon RecBCD’s binding DNA, movement of mass from the RecC tunnel exit to the RecC loop (Fig. 1D) [20].

Here, we test the signal transduction model further by employing enzymatic analyses. We report our discovery that the ATP analog ATPγS induces RecBCD to nick the DNA at a novel position, likely by the mechanism previously proposed for the small-molecule inhibitor NSAC1003 and two RecB ATPase-site mutants (see Results section) [23, 28]. RecBCD mutants altered in points of direct contact between subunit pairs [24], such as RecC-RecD for step 2 and RecD-RecB for step 3 (Fig. 1C), differentially respond to these inhibitors and the RecB ATPase-site mutations, in the manner predicted by the signal transduction model. These enzymatic tests complement our previous genetic tests [23, 24, 26] and further support this model for the control of a complex, multisubunit enzyme. Our approaches may be useful in elucidating control mechanisms in other complex enzymes with multiple activities and subunits.

Results

To study RecBCD, we employ several of its multiple enzymatic activities that can be readily assayed using linear, end-labeled DNA and analyzing the products by gel electrophoresis and autoradiography. DNA unwinding, whether by RecBCD or by boiling, produces full-length single-stranded (ss) DNA that migrates faster than any remaining double-stranded (ds) DNA substrate. Nicking at Chi produces a shorter ssDNA fragment. RecA-loading at Chi makes this short fragment resistant to exonuclease I. Degradation to oligonucleotides produces a rapidly moving smear. Here, we used this type of assay to study cleavage of DNA by RecBCD at novel sites and the effects of the signal transduction mutations on these activities.

ATPγS Induces Chi-independent DSB Hotspots by RecBCD Enzyme.

During assays for RecA loading by RecBCD [11], we unexpectedly noted, in a control experiment lacking RecA, a novel DNA cleavage product when ATP and ATPγS were both present (Fig. 2A, lanes 9 and 10). This product and full-length (4.4 kb) ssDNA were not present with ATPγS alone, indicating that ATPγS does not detectably support DNA unwinding or hydrolysis (Fig. 2A, lanes 11 and 12; Tables S1 and S2) [6, 29, 30]. Chi-independent cleavage was also not seen with ATP alone (lanes 3 and 4), as reported before (e.g. [12, 13]). ATPγS-dependent DNA cleavage occurred at the same place on DNA with or without Chi (Fig. 2B, lanes 5, 6, 9, and 10), showing it is Chi-independent. Chi-dependent DNA cleavage was not observed with ATPγS present (lanes 9 and 10); we infer that after RecBCD makes an ATPγS-dependent cleavage, it does not cleave at a subsequently encountered Chi site, which would have produced the shorter 5’-end-labeled, Chi-dependent fragment observed without ATPγS (lanes 7 and 8). This result is reminiscent of RecBCD not cleaving DNA at a subsequently encountered Chi site after cutting at the first Chi site [15] and indicates a close relation between cutting at Chi hotspots and cutting at the novel ATPγS-dependent hotspot. To determine the nature of this unexpected, ATPγS-dependent DSB hotspot, we conducted additional experiments.

Figure 2. Small molecules ATPγS and NSAC1003 convert RecBCD to a Chi-activated state dependent on signal transduction steps 3 and 4 but not 1 or 2.

(A) ATPγS and NSAC1003 induce RecBCD to cleave DNA at a hotspot independent of Chi. pBR322 χ° DNA (0.4 nM) linearized with HindIII, labeled at the 5’ ends, and cleaved with ClaI (six bp from the right-end HindIII site) was reacted with purified RecBCD enzyme (0.3 or 0.11 nM) as described in Materials and Methods. Reactions contained ATP (5 mM except 2.5 mM with ATPγS), NSAC1003 (200 μM), or ATPγS (2.5 mM) or combinations as indicated. Reaction products were analyzed by electrophoresis in a 0.9% agarose gel and autoradiographed. DS, double-stranded DNA; SS, single-stranded DNA produced by boiling (+);✖: NSAC1003-induced cleavage product; ✦: ATPγS-induced cleavage product; M, single-stranded DNA markers with the indicated nucleotide lengths. (B) ATPγS-induced nicking occurs on χ° or χ⁺E224 DNA. Reactions were as described in (A). (C) Intramolecular signaling steps 3 and 4, but not 1 or 2, are required to respond to ATPγS. Extracts from RecBCD wild-type and mutants (0.5 or 0.25 mg of extract protein/ml of reaction mix) deficient in each step of the signal transduction pathway (Table 1) were assayed on pBR322 χ° DNA as described in (A). (D) As in panel C but with NSAC1003 (200 μM) in place of ATPγS and 0.33 mg of extract protein/ml of reaction mix. Mutants were, for step 1, RecBC^S39VD; step 2, RecBC^Δ541–544D^Δ97–99; step 3, RecB^{[634–635, 639, 643–644, 646]Ala}CD^Δ523–526; and step 4, RecB^Δ913–922 C^Δ599–608D (Table 1).

Position of DSB Hotspots Depends on DNA Substrate Length and ATPγS Concentration.

To determine the position of the novel cleavages, we varied the DNA substrate by cutting off the unlabeled end of the DNA, to make substrates lacking 2.1 or 3.0 kb on the “right” end and thus lacking the site of cleavage seen above (Fig. 3A). Unexpectedly, we found DSB hotspots with all these substrates, but the position of the hotspot varied (compare lanes 5 – 6 to lanes 11 – 12 and 17 – 18). These results indicate that the cleavage site is not determined by a specific DNA sequence, unlike Chi-dependent cleavage, which occurs a few nucleotides to the 3’ side of 5’-GCTGGTGG-3’ [13, 31]. Instead, we found that, under the conditions used here (2.5 mM ATPγS plus 2.5 mM ATP), cleavage was at ~32 – 36% of each substrate’s length, measured from the “right” (unlabeled) end.

Figure 3. Position of the ATPγS-dependent cut depends on DNA substrate length and ATPγS concentration.

(A) pBR322 χ⁺E224 DNA (0.4 nM) was digested with HindIII, labeled on the 5’ ends, and further digested with ClaI (substrate 1), NdeI (substrate 2) or StyI (substrate 3) to produce substrates 4354 bp, 2270 bp or 1350 bp long, respectively, with Chi 962 bp from the labeled end. Reactions were as described in Fig. 2A with 1 nM or 0.3 nM RecBCD enzyme. (B) pBR322 χ⁺E224 DNA (0.4 nM), labeled on the 5’-ends and cut with ClaI, was reacted with 0.3 nM RecBCD enzyme and the indicated concentrations of ATP and ATPγS. Products were analyzed as in Fig. 2A. ✦ and Chi (χ⁺E): ATPγS-induced and Chi-dependent cleavage products, respectively.

We next varied the ATPγS and ATP concentrations, keeping their sum constant (5 mM). As ATPγS was increased from 0.25 mM to 2 mM, the position of the cleavage moved closer to the unlabeled (3’) end, as indicated by the length of the 5’-labeled product becoming longer (more slowly migrating during agarose gel electrophoresis) (Fig. 3B, lanes 4 – 9). ATPγS, held constant at 2.5 mM, induced cleavage at the same position with 2.5 mM or 8 mM Mg²⁺ (Fig. S1, lanes 6 – 8 and 12 – 14), but cleavage was reduced and moved to the “left” as Mg²⁺ was reduced from 2.5 to 0.5 mM, below which cleavage was not detectable (Fig. S2, lanes 11 – 18). Chi-dependent cleavage, although remaining at the same position, was also reduced as Mg²⁺ was reduced to 0.5 mM and was undetectable below that concentration (lanes 3 – 9). Thus, Chi-dependent and ATPγS-dependent cleavage share multiple properties.

The ATPγS-dependent cuts occurred near the middle of the DNA substrate, which implies that under the conditions used here (2.5 mM ATP and 2.5 mM ATPγS) the ratio of the RecB and RecD helicase rates is about 0.5, the point on the DNA substrate at which RecB is when RecD reaches the end of the DNA. Examination by electron microscopy of DNA partially unwound by RecBCD in the presence of ATPγS (5 or 10 mM with 5 mM ATP) showed that ATPγS slowed the RecB and RecD helicases about equally (Table S1). The ratio of RecB:RecD helicase rates remained at about 0.5 at each ATPγS concentration (0, 5, or 10 mM with 5 mM ATP), as expected from the point of ATPγS-dependent cleavage at higher ATPγS concentration.

These results are reminiscent of those observed with two mutants altered at adjacent amino acids in the RecB ATP binding site (RecB Y803H and RecB V804E) [23]. These mutants nick DNA at a position a certain fraction of the length of the dsDNA substrate. For each mutant, this fraction is nearly equal to the ratio of the rates of RecB:RecD translocation on dsDNA, as determined by electron microscopy of single DNA molecules (unwinding intermediates; Fig. 1B). This result is consistent with RecD signaling RecB to nick the DNA where it is when RecD reaches the end of the DNA substrate and perforce stops unwinding, the model in Fig. 1C. Our current observations (Figs. 2A and 2B) suggest that ATPγS has two effects on RecBCD – slowing RecB and RecD helicases in a concentration-dependent manner and converting RecBCD into a Chi-activated state such that the RecB nuclease cuts the DNA where it is when RecD, the faster helicase, reaches the end of the DNA and stops unwinding. Below, we test this suggestion with further enzymatic assays using mutants altered in various steps of the signal transduction from Chi recognition by RecC to DNA cleavage by RecB (Fig. 1C) [24].

An alternative interpretation to be considered is that the ATPγS-dependent bands arise from collision of two RecBCD molecules, one starting at each end of the DNA substrate, and cutting at the middle of the molecule, as reported previously [32, 33]. In some cases, such as in Fig. 2A (lanes 9 – 10) and Fig. 2C (lanes 3 – 8), the mobility of the ATPγS-dependent product was clearly less than that of ss DNA ½ the length of the substrate (2180 nuc), although in some figures, such as Fig. 6B (lanes 5 – 6), it migrated nearer that of ½ length molecules. However, the mobility of the ATPγS-dependent band decreased smoothly up to 0.5 as ATPγS concentration was increased from 0.25 to 2 mM (Fig. 3B, lanes 4 – 9); similarly, the mobility increased smoothly as Mg²⁺ concentration was decreased from 2 to 0.5 mM (Fig. S2, lanes 11 – 15). Obviously, these bands cannot all reflect cutting in the middle of the substrate. Furthermore, under the conditions used here, increasing the RecBCD enzyme concentration to a very high level did not yield a detectable cut in the middle (a “collision” cut) (Fig. S3). Finally, we observed ATPγS-dependent cuts on hairpin DNA, on which only one RecBCD molecule can act, precluding collision cuts (Fig. S3). Thus, we see no evidence that the Chi-independent cuts result from collisions of two RecBCD molecules. Rather, they arise from one RecBCD enzyme being activated to cut at a novel position – where RecB is when RecD reaches the end of the DNA.

Figure 6. ATPγS-dependent cuts require RecD and the RecB nuclease active site.

(A) Purified RecBCD (0.3 or 0.1 nM) or RecBC (10 or 3.3 nM) was assayed with pBR322 χ⁺E224 DNA (0.4 nM) and either 5 mM ATP or 2.5 mM ATP plus 2.5 mM ATPγS and analyzed as in Fig. 2A. (B) As in panel A but with χ° DNA and purified RecB^D1080ACD (1 or 0.33 nM) instead of RecBC. ✦ and Chi (χ⁺E): ATPγS-induced and Chi-dependent cleavage products, respectively.

ATPγS-induced DNA Cleavage Depends on Signal Transduction Steps 3 and 4 but Not Steps 1 or 2.

We have inferred five steps in the signal transduction from Chi recognition by RecC to DNA nicking at Chi by RecB and loading of RecA [23]. In this model (Fig. 1C), when the Chi sequence is in a unique position in a tunnel in RecC, RecC recognizes Chi (step 1) [10]. RecC then signals RecD to stop unwinding (step 2) [14]. When stopped, RecD signals RecB Nuc to swing from its “storage” position at a surface loop on RecC (step 3) [20, 26]. After swinging back to the RecC tunnel exit, RecB nicks the DNA 4 – 6 nucleotides to the 3’ side of the Chi octamer (step 4) [13] and begins loading RecA onto the newly generated 3’ ssDNA end (step 5) [11]. Mutants altered at each of the first four steps behave as expected from the models in Fig. 1 – they lack Chi hotspot activity but retain intracellular nuclease activity and recombination-proficiency, though the latter is modestly reduced [24]. (Mutants specifically altered in step 5 have not, to our knowledge, been reported.) Here, we used mutants altered at each of steps 1 – 4 (Table 1) to test this model further.

Table 1.

Mutants in the RecBCD signal transduction pathway

Stepa	Rec subunit contact pointb	Contact point mutationsc	Reference
1	C	RecCS39V	[61]
2	CD	RecCΔ541–544 RecDΔ97–99	[24]
3	DB	RecDΔ523–526 RecB[634–635, 639, 643–644, 646]Ala	[24]
4	BC	RecBΔ913–922 RecCΔ599–608	[24]

^aStep altered in the signal transduction model for Chi hotspot control of RecBCD enzyme (Fig. 1C).

^bRecBCD subunit(s) involved in sending or receiving the proposed signal.

^cThe indicated amino acid(s) were altered by deletion (Δ) or substitution to valine (V) or alanine (Ala).

We observed that mutants altered in step 1 or 2 responded to ATPγS like wt RecBCD (Fig. 2C, lanes 3 – 8; Fig. S4A, lanes 6 – 8, 12 – 14, 18 – 20, and 24 – 26), but mutants altered in steps 3 or 4 lacked detectable ATPγS-dependent DNA cleavage (Fig. 2C, lanes 9 – 12). These results indicate that step 1 (Chi recognition) and step 2 (Chi-activated RecC signaling RecD to stop unwinding) are not necessary for ATPγS to induce RecB Nuc to cut the DNA. These results are predicted by the model, because Chi is not required for the ATPγS-dependent DSB hotspot (Fig. 2B, lanes 5 and 6) and RecD does not need to be signaled by another subunit to stop unwinding at the end of the DNA. Steps 3 and 4, however, are required for ATPγS to promote cleavage (step 4) when RecD is stopped at the end of the DNA substrate and signals RecB Nuc to swing (step 3). These results suggest that ATPγS converts RecBCD into a Chi-activated state, as does NSAC1003 [28], discussed next.

Potential Antibiotic NSAC1003-induced DNA Cleavage Also Depends on Steps 3 and 4 but Not Steps 1 or 2.

A search for potential antibiotics that block DNA break-repair by RecBCD yielded, among other compounds, NSAC1003 [28, 34]. This compound inhibits RecBCD nuclease in a manner competitively inhibited by ATP. NSAC1003 can be computationally docked onto the RecB ATPase site, near the sites of the two recB mutant amino acids (Y803H and V804E) noted above. More detailed analysis showed that NSAC1003 induces RecBCD to cleave DNA at novel hotspots with the properties shown above for ATPγS – cleavage occurs at a certain percent of the DNA substrate length, and the cleavage position occurs closer to the 3’ end as NSAC1003 concentration is increased, suggesting that NSAC1003 slows RecB in a concentration-dependent manner [28]. These results indicate that NSAC1003, like the ATPase-site mutants, sensitizes RecBCD to nick the DNA where RecB is when RecD reaches the end of the DNA. In other words, NSAC1003, like ATPγS, appears to convert RecBCD into a Chi-activated state (Fig. 2A, lanes 5 – 6).

To test this idea further, we tested the effect of mutations in steps 1 – 4 on NSAC1003-dependent cleavage (Fig. 2D). As for ATPγS (Fig. 2C), steps 1 and 2 were not detectably required for NSAC1003-dependent cleavage, but steps 3 and 4 were required – cleavage was the same in wt and step 1 and 2 mutants (Fig. 2D, lanes 4, 6, and 8), but undetectable in step 3 and 4 mutants (lanes 10 and 12).

Two RecB ATPase Site Mutants Behave Like ATPγS- and NSAC1003-altered RecBCD.

The signal transduction model (Fig. 1C) arose from studies of RecB ATPase site mutants, Y803H and V804E [23]. These mutants cleave DNA, with or without Chi, at a certain percent of the DNA length – 26% for RecB Y803H and 19% for RecB V804E under the conditions used; cleavage is on the strand with 3’ at the DNA end at which RecBCD initiated unwinding, as for ATPγS-, NSAC1003-, and Chi-dependent DNA cleavage (Figs. 1 – 3; [13]). With ATPγS, these two RecB mutants still made Chi-independent cleavages, but the positions of the cleavages changed – cleavage was closer to the entry end with RecB Y803H but farther from the entry end with RecB V804E in the presence of ATPγS (Fig. 4A, lanes 7 – 14; Fig. S5, lanes 3 – 6).

Figure 4. RecB ATPase site mutants respond to ATPγS and require intramolecular signal transduction step 3, but not 2, to make DNA length-dependent cuts.

(A) Purified RecBCD enzyme (0.3 or 0.1 nM; wt or the indicated mutant) was assayed with pBR322 χ⁺F225 DNA (0.4 nM) and analyzed as in Fig. 2A. (B) Extracts (0.5 or 0.17 mg of extract protein/ml of reaction mix) from RecB^Y803HCD and RecB^V804ECD mutants without (wt) or with additional mutations blocking step 2 or step 3 of the signal transduction pathway (Fig. 1C; Table 1) were assayed on pBR322 χ° DNA and analyzed as in Fig. 2C without ATPγS. ✦, ◈, and Chi (χ⁺F): ATPγS-induced, ATPase-site mutant-dependent, and Chi-dependent cleavage products, respectively. See also Fig. S5.

The similarity of the Chi-independent cleavages by the RecB ATPase site mutants to the cleavages induced in wt RecBCD by ATPγS and NSAC1003 led us to test the effect of step 2 and step 3 mutations when added to the RecB Y803H or RecB V804E mutations. For each RecB mutant, the step 2 mutations had no detectable effect on cleavage (Fig. 4B, lanes 3 – 6), but step 3 mutations abolished the Y803H- and V804E-induced cleavages (lanes 7 – 8 and 13 – 14). Thus, in all cases examined here, the novel cleavages require the signal from RecD for RecB to swing its nuclease (step 3) but not the signal from RecC for RecD to stop unwinding (step 2).

Both RecB and RecD ATPase Sites Are Required for Novel DSB Hotspot Activity.

Since the position of DNA cleavage in these RecB ATPase site mutants is determined by the ratio of the RecB:RecD helicase rates [23], we predicted that both ATPase sites, and thus both helicases, would be required for activity of these hotspots. This prediction was found to be correct. Changing an essential lysine to glutamine in either the RecB (K29Q) or RecD (K177Q) ATPase site [35, 36] resulted in undetectable cleavage induced by ATPγS or NSAC1003 (Figs. 5A, lanes 7 – 8 and 11 – 12, and 5B, lanes 9 – 10 and 13 – 14). Note that if RecD helicase is inactive, it never reaches the end of the DNA and thus does not signal RecB to cut the DNA. If RecB helicase is inactive, it never leaves the initial, unlabeled end, and any cut it makes would be indistinguishable from uncut 5’-end-labeled substrate.

Figure 5. ATPγS- and NSAC1003-dependent cuts require the RecB and RecD ATPases.

Purified RecBCD (0.1 or 0.3 nM), RecB^K29QCD (1 nM or 0.33 nM), or RecBCD^K177Q (1 nM or 0.33 nM) was assayed with pBR322 χ⁺E224 DNA (0.4 nM) and analyzed as in Fig. 2A. (A) Reactions contained either 5 mM ATP or 2.5 mM ATP plus 2.5 mM ATPγS. (B) Reactions with pBR322 χ° DNA (0.4 nM) contained ATP (5 mM) with or without NSAC1003 (200 μM). ✦, ✖ and Chi (χ⁺E): ATPγS-induced, NSAC1003-induced and Chi-dependent cleavage products, respectively.

Novel DSB Hotspots, Like Chi Hotspots, Require RecD, an Active RecB Nuclease, and a Tether Connecting the RecB Helicase and Nuclease Domains.

RecD is a rapid helicase that moves on the 5’ → 3’ strand [5, 8, 37], but it is also essential for RecBCD nuclease activity even though the nuclease domain is in RecB [38]. This latter feature is consistent with steps 3 and 4 of the signal transduction model – RecD is required to activate the RecB nuclease. We tested the role of RecD by using RecBC enzyme, which unwinds DNA but has no detectable nuclease activity in standard assays with ATP [21, 39]. RecBC also made no detectable cleavages induced by ATPγS (Fig. 6A, lanes 7 – 10). RecBCD enzyme with the RecB D1080A mutation, lacking an aspartate that binds Mg²⁺ essential for DNA cleavage [38], also made no detectable cleavages induced by ATPγS (Fig. 6B, lanes 11 – 14). RecBCD lacking the 19 amino-acid tether connecting the RecB helicase and nuclease domains [27] also made no detectable ATPγS-induced cleavages (Fig. S4B, lanes 10 – 12). These mutant enzymes also lack detectable nicking of DNA at Chi, and the mutants lack Chi hotspot activity in cells [7, 21, 27, 38, 40]. Thus, RecBCD’s DNA cleavages induced by Chi, ATPγS, and NSAC1003 share many properties.

Discussion

One of the most complex enzymes known, RecBCD has eight identified enzymatic activities – two ATPases, two DNA translocases, dsDNA exonuclease, ssDNA exonuclease, ssDNA endonuclease, and RecA loading – all contained within three polypeptides with a combined MW of 330 kDa and regulated by a specific 8 bp sequence Chi (reviewed in [1–4]). Proper coordination of these activities is critical for RecBCD enzyme to promote its important physiological roles – repairing broken DNA and forming genetic recombinants if the DNA is repairable by homologous recombination, or degrading into small oligonucleotides broken DNA that cannot recombine [1, 41].

RecBCD is the only enzyme to our knowledge with two different, separate helicases: RecD moves on the 5’ → 3’-ended strand and RecB on the 3’ → 5’-ended strand [5, 6, 37]. This feature allows the DNA to remain single-stranded immediately behind the enzyme, even in the absence of DNA binding proteins, because the complementary sequences are far apart (Fig. 1B, line b). This feature is likely important to direct the RecA-ssDNA filament to invade another (intact) DNA for repair, rather than reforming dsDNA behind RecBCD and not proceeding to repair and recombination.

The nuclease activities must be properly regulated for RecBCD to promote repair of DNA that can recombine (i.e., when there is homologous, intact DNA for strand invasion) but to convert DNA that cannot recombine into reusable nucleotides. This feature likely reflects the control of the nuclease activities, which can be viewed as a ss endonuclease that acts during unwinding of linear dsDNA (and thus by one definition is a ds exonuclease) or during translocation on linear ssDNA (and thus is a ss exonuclease). (Note that by one definition an exonuclease requires a ds or ssDNA end for nuclease activity, even though cleavages may occur internally, which by a different definition means an endonuclease.) Regulation of the nuclease activity is complex, for it requires ATP (for translocation on ds or ssDNA) and Mg²⁺ ion. The latter acts in a complex with ATP for the two helicases but alone for the nuclease [42]. Thus, the ATP and Mg²⁺ concentrations affect the observed nuclease activity in a complex way [1, 2, 41, 43]. The nuclease activity is also controlled by Chi hotspots to prompt or halt cleavage of the 3’-ended and 5’-ended strands, depending on the ATP and Mg²⁺ concentrations [1, 2, 4, 12, 13, 32, 33, 41, 44]. The available evidence indicates that, in cells, Chi prompts a cut on the 3’-ended strand, much like ATPγS and NSAC1003 (Figs. 2, 3, and 4) and two RecB ATPase site mutants (Fig. 4) [4, 23, 28]. We have proposed models (Fig. 1) to account for these complex activities and their coordination.

The signal transduction model (Fig. 1C) and the nuclease swing model (Fig. 1D) are closely related. The former model accounts for the coordination of the three subunits during RecBCD’s reaction cycle and its response upon encountering Chi during DNA unwinding [23, 24, 26]. The latter model accounts for the dramatic change of nuclease activity at Chi and the physical changes of RecBCD, first upon binding DNA and second upon encountering Chi [20, 26, 27]. Both models have been supported by genetic and physical analyses, as noted in the Introduction. Here, we provide additional support for these models by enzymatic tests.

Two small molecules, ATPγS and NSAC1003 (Figs. 2, 3, and 4) [28], prompt RecBCD to cleave DNA at special sites (DSB hotspots) that are a certain fraction of the length of the DNA substrate. That fraction depends on the concentration of ATPγS or NSAC1003 rather than a special DNA sequence, as for cleaving DNA at Chi. In other regards, the ATPγS- and NSAC1003-dependent hotspots behave like Chi hotspots – they require RecD, the RecB nuclease active site, the two ATPase sites, and signal transduction steps 3 and 4 (but not steps 1 or 2) (Figs. 2, 5, and 6). The ATPγS-dependent hotspot, like Chi hotspots, also requires an intact tether connecting the RecB helicase and nuclease domains (Fig. S4B) and on which the nuclease domain is proposed to swing (Fig. 1D). (NSAC1003 has not been similarly tested, but we expect it too would require the tether to prompt hotspot cleavage.)

Coordination of RecBCD’s multiple activities requires interaction among the three subunits. The signal transduction model employs contacts between a pair of subunits, or between RecC and Chi, at each of the four proposed steps. In the published crystal and cryoEM structures, these contact points are far apart, encompassing ~185 Å from step 1 to 2 to 3 to 4. This is greater than the maximal distance across RecBCD (~140 Å), implying that the signal transduction proceeds along a loop whose ends are ~40 Å apart (from the middle of the Chi octamer to the RecB nuclease active site). This outcome indicates a highly evolved system that is likely conserved among RecBCD enzymes from multiple species, many of which are regulated by special DNA sequences, such as Chi [45–51].

The nuclease swing and signal transudction models are amenable to further physical tests, for example by FRET or PET analysis using non-natural amino acids incorporated into the RecB nuclease domain and the RecC surface loop to which it is postulated to swing (Figs. 1A and 1D). Examination by cryoEM of RecBCD enzymes stopped during unwinding, before or after Chi, could provide direct visualization of the nuclease domain swinging upon binding DNA and upon encountering Chi; to our knowledge the structure of RecBCD stopped during active DNA unwinding has not been determined. The clear predictions of the signal transduction model can be similarly tested using RecBCD mutant enzymes blocked at one or another step. Success with these methods depends on development of RecBCD containing non-natural amino acids and microscopic techniques not yet available. The results reported here should motivate such endeavors, which could provide even further support for the signal transduction and nuclease swing models for RecBCD’s control by Chi.

Signal transduction over long distances by conformational changes of the type described here may apply to other complex proteins. Identifying mutations that strongly affect enzyme activity but are located far from an enzyme’s active site may indicate regulation by major conformational changes. Several such examples have been described. The transcription rate of Thermus thermophilus RNA polymerase is reduced >1000-fold when amino acids located ~20 Å from the polymerization site are altered in the β’ subunit [52]. On the other hand, the transcription rate of Saccharomyces cerevisiae RNA polymerase I is increased 2- to 3-fold by a single amino acid change ~50 Å from the active site [53]. The movement of myosin along an actin filament requires ATP hydrolysis [54]. The second-order rate constant of ADP binding to human myosin in the presence of actin is reduced ~85-fold by an amino acid change ~30 Å from the ATP hydrolysis site and even more distant from the actin binding site [55]. Long distance communication between distant sites elsewhere on these enzymes might be revealed by more extensive studies. Other large enzymes must properly coordinate their activities. The regulation of kinesin and ATP synthetase [56, 57] likely requires communication between distant points on the enzymes. Mutational analysis of specific contact points may reveal the mechanism of control. RecBCD is not essential for growth and coordinates many enzymatic activities during its complex reaction cycle. These factors make RecBCD an excellent model for further studies using combined genetic, enzymatic, and biophysical approaches.

Materials and Methods

Bacterial Strains and recBCD Mutant Plasmids

Supplementary Table S3 lists the E. coli strains with the plasmids bearing the mutant recBCD alleles, their allele numbers, descriptions of the mutations, and their sources. Plasmids were in strain V2831, bearing a complete deletion of the recBCD genes.

Enzymes and Chemicals

RecBCD enzyme (10.6 μM, 200,000 units/mg), RecBC (2.06 μM), RecB^D1080ACD (1.15 μM), RecB^Y803HCD (1.02 μM), RecB^V804ECD (1.38 μM), RecB^K29QCD (1.81 μM), RecBCD^K117Q (0.71 μM) were gifts from Andrew Taylor prepared as described [5]. For experiments assaying RecBCD wild-type and mutant enzymes in cell-free extracts, plasmid transformants of strain V2831 (ΔrecBCD2731) were grown in Terrific Broth to late log phase. Extracts were prepared as described [58].

Restriction enzymes, T4 polynucleotide kinase, and calf intestinal phosphatase (CIP) were from New England BioLabs. ATP (Sigma Aldrich), ATPγS (Fisher, Tocaris or Sigma-Aldrich) and NSAC1003 (gift from Ryan Cirz, Achaogen, Inc.) were used as indicated in the figure legends.

Assay for RecBCD Enzyme Unwinding and Cutting Activity

The substrate for DNA unwinding and Chi-dependent, ATPγS-induced, or NSAC1003-induced cutting was pBR322 χ° (lacking a Chi site), χ⁺F225 [Chi at basepair (bp) 1493–1500] or χ⁺E224 (Chi at bp 984–991) [59]. DNA was linearized with HindIII, treated with Quick CIP (New England BioLabs), labeled at the 5’ ends with [γ−³²P]ATP (3000 Ci/mmol; Perkin Elmer) using polynucleotide kinase, and digested with ClaI to remove the “right” end label except in Fig. 6A. The pBR322 χ⁺E224 capped substrate (Fig. S3) was prepared as described [33]. Unincorporated nucleotides and the six-bp ClaI – HindIII fragment were separated from the substrate with an SR-200 minicolumn (Cytiva). ssDNA markers were made by digesting pBR322 χ° DNA with HindIII, labelling the 5’ ends as above, and digesting with EagI, PvuII, EcoRV or StyI to produce two 5’-end labeled fragments from each molecule. The DNA was boiled for 3 min to produce ssDNA and cooled on ice before loading onto the gel.

DNA unwinding and nicking were assayed as described [27] in 12 μl reactions containing 25 mM Tris acetate (pH 7.5), 2.5 mM MgCl₂, 1 mM DTT, 1 μM SSB (Promega), and 0.4 nM DNA. Reactions were started by addition (to a final concentration) of 5 mM ATP, 2.5 mM ATP plus 2.5 mM ATPγS, or 5 mM ATP plus 200 μM NSAC1003 as indicated in the figure legends; in Figs. 3B, S1, and S2 the Mg⁺², ATP, and ATPγS concentrations were varied as indicated. After 2 min at 37°, 4 μl of stop buffer (0.1 M EDTA, 2.5% sodium dodecyl sulphate, 10% Ficoll) was added, and the tubes placed on ice. Reaction products were separated by electrophoresis (100 V; 2 hr) on 0.9% agarose gels (22 cm long) in Tris acetate buffer [60]. Dried gels were visualized on a Typhoon Trio Phosphorimager (GE Healthcare Life Sciences).

Keypoints.

• Chi hotspots contol RecBCD enzyme via complex interactions among its three subunits.

• Small molecules and RecB ATPase-site mutations convert RecBCD into a Chi-activated state.

• Chi-activated state requires steps 3 and 4, but not 1 or 2, of the signal transduction model.

• Enzymatic results here and previous genetic results strongly support the model.

• The approaches used here provide a paradigm for studying other complex enzymes.