The AddAB Helicase/Nuclease Forms a Stable Complex with Its Cognate χ Sequence During Translocation*

The Bacillus subtilis AddAB enzyme possesses ATP-dependent helicase and nuclease activities, which result in the unwinding and degradation of double-stranded DNA (dsDNA) upon translocation. Similar to its functional counterpart, the Escherichia coli RecBCD enzyme, it also recognizes and responds to a specific DNA sequence, referred to as Chi (χ). Recognition of χ triggers attenuation of the 3′- to 5′-nuclease, which permits the generation of recombinogenic 3′-overhanging, single-stranded DNA (ssDNA), terminating at χ. Although the RecBCD enzyme briefly pauses atχ, no specific binding of RecBCD toχ during translocation has been documented. Here, we show that the AddAB enzyme transiently binds to its cognate χ sequence (χBs: 5′-AGCGG-3′) during translocation. The binding of AddAB enzyme to the 3′-end of the χBs-specific ssDNA results in protection from degradation by exonuclease I. This protection is gradually reduced with time and lost upon phenol extraction, showing that the binding is non-covalent. Addition of AddAB enzyme to processed, χBs-specific ssDNA that had been stripped of all protein does not restore nuclease protection, indicating that AddAB enzyme binds toχBs with high affinity only during translocation. Finally, protection of χBs-specific ssDNA is still observed when translocation occurs in the presence of competitor χBs-carrying ssDNA, showing that binding occurs in cis. We suggest that this transient binding of AddAB toχBs is an integral part of the AddAB-χBs interaction and propose that this molecular event underlies a general mechanism for regulating the biochemical activities and biological functions of RecBCD-like enzymes.

The manner by which the RecBCD or AddAB complexes recognize and interact with their cognate sequence while translocating through DNA at rates of up to several hundred base-pairs (bp) per second is still unclear. In the case of RecBCD enzyme, early studies showed that under certain conditions, the RecBCD enzyme undergoes a -dependent inactivation, characterized by the reduced ability of the enzyme to catalytically process -containing DNA molecules (21,22). Under these conditions, the enzyme dissociates into subunits after DNA processing (23). However, recent single molecule microscopy experiments revealed that although recognition results in a brief pause of the enzyme at that is followed by a reduced rate of translocation (9), the heterotrimeric form of the enzyme persists and continues to translocate after recognition (24). Thus, subunit dissociation does not explain how recognition results in a persistent change in enzyme activity that can be maintained for distances of up to 30 kilobases but yet is fully reversible upon dissociation. Interestingly, single molecule studies showed that the newly generated -containing ssDNA remains associated with the translocating RecBCD enzyme (9), although the mechanistic significance of this observation is yet to be fully explored. Here, we report that AddAB enzyme also undergoes a Bs -dependent inactivation process, and we provide the first direct evidence that this behavior occurs by the noncovalent binding of the enzyme to the 3Ј-end of the newly generated Bs ssDNA that persists during the course of translocation. We suggest that this previously undisclosed interaction represents the underlying regulatory event that is responsible for all of the enzymatic changes that are elicited by interaction with and that this molecular event reversibly switches the RecBCD/AddAB family of enzymes from DNA-destroying to DNA-repairing enzymes.

EXPERIMENTAL PROCEDURES
Reagents and Proteins-Chemical reagents and ATP were purchased from Sigma. Shrimp alkaline phosphatase and [␥-32 P]ATP were purchased from United States Biochemical Corp and PerkinElmer, respectively. Proteinase K was from Roche Molecular Biochemicals. Restriction endonucleases and T4 polynucleotide kinase were from New England Biolabs (NEB). SSB protein was prepared as described previously (25). AddAB protein was purified as described below.
Cloning of the addA and addB Genes of B. subtilis for Overexpression Purposes-Cloning of the addA and addB genes from B. subtilis was achieved by PCR from genomic DNA. Primers were designed based on the published sequence of the B. subtilis genome (26), which allowed the amplification of the entire addA and addB genes immediately flanked by suitable restriction endonuclease recognition sequences (NdeI and XhoI for addA and NcoI and BamHI for addB). For addA, the forward and reverse primer sequences were 5Ј-GCG-GATGCATATGAACATTCCTAAACCGGCAG-3Ј and 5Ј-GCTT-ATGGCTCGAGCTATAATGTCAGAATGTGCCC-3Ј, respectively. For addB, the sequences were 5Ј-GTCTTCTGCCATGGGA-GCAGAGTTTTTAGTAGG-3Ј and 5Ј-TGCTGTCCGGATCCTT-AGGAATGTTCATTGTCATC-3Ј. The template DNA was purified from B. subtilis strain 168 supplied by Dr. Sabine Autret (Errington Laboratory, University of Oxford, UK). The addA and addB sequences were then inserted into the pET22b and pET28a vectors (Novagen), respectively, at the restriction sites listed above using standard cloning techniques to create plasmids pAddA-22b and pAddB-28a. DNA sequencing (MWG-Biotech) was used to confirm that the final sequences of the entire genes, as well as the promoter regions of the vectors, were identical to those published previously.
Expression and Purification of AddAB Protein-For expression of AddAB protein, the pAddA-22b and pAddB-28a plasmids were freshly transformed into E. coli B834 (DE3) cells (Novagen). An overnight starter culture of these cells was used to inoculate 12 liters of phosphate-buffered LB medium supplemented with 50 g/ml ampicillin and 30 g/ml kanamycin. The cells were grown at 37°C until they reached mid-log phase. At this point 1 mM IPTG and a further 25 g/ml ampicillin were added to the medium. The growth temperature was simultaneously dropped to 25°C, which increases the yield of soluble proteins. Growth was continued for a further 3-4 h before the cells were harvested, resuspended in 150 ml of 50 mM Tris-Cl pH 7.5, and 10% sucrose, and stored at Ϫ80°C. All subsequent purification steps were performed at 4°C. Cells were lysed in the presence of 0.1 mM phenylmethylsulfonyl fluoride using Brij lysis. Briefly, lysozyme was added to 0.2 mg/ml, followed by a 30-min incubation with stirring. Next, EDTA was added to 1.5 mM followed by stirring for 30 min. Finally Brij-58 was added slowly to a final concentration of 0.3% over 30 min with stirring. The soluble cell extract was recovered by centrifugation and ammonium sulfate was added to a final concentration of 50%. The precipitate was recovered by centrifugation, resuspended in buffer B (20 mM Tris-Cl, pH 7.5, 0.1 mM EDTA, 0.1 mM dithiothreitol) ϩ 100 mM NaCl, and dialyzed against the same buffer overnight. The protein sample was loaded onto a Fast-Flow Q-Sepharose column (Amersham Biosciences-GE Healthcare) in Buffer B ϩ 100 mM NaCl, washed with buffer B ϩ 300 mM NaCl, and eluted with a gradient to 600 mM NaCl. This eluate was dialyzed against 50 mM potassium phosphate buffer, pH 7.5, ϩ 1 mM dithiothreitol. It was then loaded in this buffer onto a Bio-Gel hydroxyapatite column (Bio-Rad) and eluted with a gradient to 300 mM potassium phosphate buffer. After dialysis against buffer B ϩ 100 mM NaCl, the sample was loaded in the same buffer onto a Hi-Trap heparin column (Amersham Biosciences-GE Healthcare). At this stage, the sample contains a molar excess of AddA protein but, unlike AddAB protein, AddA does not bind to the heparin. The AddAB protein was eluted with a gradient to 500 mM NaCl, and peak fractions were collected. The sample was diluted with B buffer to a conductivity equivalent to buffer B ϩ 100 mM NaCl and loaded onto a 10-ml mono-Q column (Amersham Biosciences-GE Healthcare). After washing with buffer B ϩ 200 mM NaCl, AddAB protein was eluted with a gradient to 600 mM NaCl. Most of the protein was present in a major peak at ϳ40% (equivalent to ϳ350 mM NaCl) through the gradient. This protein was collected, dialyzed against buffer B ϩ 100 mM NaCl ϩ 50% glycerol, and stored at Ϫ80°C. Concentration was determined using a theoretical extinction coefficient ( ϭ 251,900 M Ϫ1 ⅐cm Ϫ1 ), and the protein was judged to be greater than 95% pure by SDS-PAGE analysis. Functionality was assessed using a dye-displacement assay and protein titrations, as described earlier (27). The protein preparations used here were between 25 and 50% active.
Helicase Inactivation Assay-Standard reactions contained 25 mM Tris acetate (pH 7.5), 1 mM dithiothreitol, 0.5 mM magnesium acetate, 6.25 M nucleotides linear DNA (corresponding to 1.03 nM molecule), 2 M E. coli single-stranded DNA binding protein (SSB), 1 mM ATP, and 0.26 nM functional AddAB (AddAB:dsDNA ends ratio ϭ 1:8). The DNA substrates were linearized by BamHI digestion. Reactions were initiated by addition of AddAB enzyme and incubated at 37°C. Samples were taken at the indicated time points, deproteinized by digestion with Proteinase K, and processed by electrophoresis through 1% agarose gels at 4 V/cm for 4 h in 1ϫ TAE buffer. After separation of the reaction products, the gels were dried and exposed to PhosphorImager screens and quantified using the ImageQuaNT software (Molecular Dynamics).
-Protection Assay-Typical reactions contained 25 mM Tris acetate, pH 7.5, 1 mM magnesium acetate, 1 mM ATP, 1 mM dithiothreitol, 10 M nucleotide linear pADG6406-1 (corresponding to 1.6 nM molecules), 2 M SSB protein, and 3.2 nM (unless otherwise specified) functional AddAB enzyme. Assays were performed at 37°C. Tailed DNA substrates were prepared as described (18) so that AddAB can only access the linear dsDNA from one entry site. Two minutes after initiation of the reaction by the addition of the enzyme, Exonuclease I (NEB), a 3Ј-to 5Ј-ssDNA exonuclease, was added to a final concentration of 0.8 units/l. Samples were taken at the indicated time points and processed as described above. When indicated, a phenol extraction was carried out 30 s after the addition of Exonuclease I, followed by an ethanol precipitation. The pellets were then resuspended in water and further desalted by passage through a G-25 spin column (Amersham Biosciences-GE Healthcare) equilibrated with water. Standard buffer components were then added back to a final volume of about half of the initial volume to compensate for the loss of DNA incurring during the procedure. The samples were then treated with various combinations of enzymes, as indicated. Reaction mixtures were incubated for 1 min after each new addition, and the products were separated by agarose gel electrophoresis as described above.
Competition Experiment-The reaction conditions were similar to the ones described for the protection assay, except that a 100-fold molar excess of unlabeled oligonucleotides was added after the addition of AddAB enzyme. The oligonucleotides used here either contained or did not contain a Bs sequence at their 3Ј-end (SKNH55 or SKNH56, respectively). The concentration of the SSB protein was 4.5 M. The reaction mixture was incubated at 37°C for 2 min before the addition of each component. The reaction was initiated by addition of ATP. Exonuclease I was added to the reactions after 2 min. Samples were taken at the indicated time points and processed as described above.

RESULTS
The AddAB Enzyme Shows Bs -Dependent Inactivation-We first investigated the consequences of Bs recognition on the ability of the AddAB enzyme to process a linear DNA substrate. For this, we used three different DNA substrates carrying zero ( 0 ), one ( ϩ ), or three ( ϩϩϩ ) Bs sites (Fig. 1A). These DNA substrates were linearized, radiolabeled at their 5Ј-ends, and incubated with a limiting amount of enzyme (1 functional AddAB enzyme/8 DNA ends; 12.5% saturation). Substrate usage was then followed as a function of time after separation of the reaction products by agarose gel electrophoresis. In the case of the RecBCD enzyme, it is known that, at least under certain conditions (21), interaction with can result in a -dependent, reversible inactivation of the enzyme. As shown in Fig. 1B, processing of a 0 substrate by AddAB enzyme is catalytic, with substrate usage reaching near completion after 5 min, in agreement with results obtained using a more sensitive real time dye-displacement assay. 6 However, in the presence of three Bs sites, the efficiency of the reaction was markedly lower (Fig. 1B). Quan-tification of multiple independent experiments showed that substrate usage for 0 DNA reached 75% after 5 min, compared with 25% for ϩϩϩ DNA. An intermediate situation was observed for a ϩ substrate, for which substrate usage showed a modest but significant decrease compared with 0 DNA. The lower level of this reduction was expected for two reasons: 1) in one-half of the processing events, the AddAB enzyme approached the Bs site from the wrong orientation and did not recognize it, and 2) as for the E. coli RecBCD enzyme, Bs recognition for those AddAB enzymes that approached from the correct orientation is not 100% efficient. Therefore, a substrate carrying three Bs sites that can be recognized by the enzyme in either orientation is expected to lead to a much greater effect, as observed here. Thus, the AddAB enzyme, like its E. coli analogue, shows Bs -dependent inactivation with the lower yield of processed Bs -containing substrates resulting from a reduced ability of AddAB enzyme to cycle through multiple DNA molecules in a catalytic manner.
Bs -Specific ssDNA Is Protected from Exonuclease I Degradation due to Binding of AddAB to the 3Ј-End of Bs -To explain the mechanism by which Bs -dependent inactivation occurs, we speculated that the AddAB enzyme might remain bound to the DNA products generated during the unwinding reaction. To address this possibility, we used exonuclease I (Exo I) as a probe for the presence of a protein bound to the 3Ј-end of the ssDNA. Nucleolytic degradation of ssDNA by Exo I proceeds in a 3Ј-to 5Ј-direction and is sensitive to the presence of proteins bound to the 3Ј-end. This property was used to show that 6 F. Ché din and S. C. Kowalczykowski, unpublished observations. RecBCD enzyme loads E. coli RecA protein onto the 3Ј-end of the -containing ssDNA produced during dsDNA processing (10). Similarly, Exo I activity can be inhibited by the formation of secondary structure in ssDNA (28).
Here, we used a tailed DNA substrate either carrying or lacking a single Bs site ( Fig. 2A), and we initiated the reaction with a saturating amount of AddAB enzyme. Tailed DNA ensures that the AddAB enzyme enters and translocates through the DNA from only one end. As described earlier (18), processing of the 0 DNA by the AddAB enzyme resulted in the unwinding and degradation of much of the dsDNA substrate, but some full-length ssDNA is also produced (Fig.  2B, lanes 1-3). Addition of Exo I after completion of the reaction (2 min) led to the rapid and complete disappearance of this full-length ssDNA, indicating that this DNA species was fully available for degradation by Exo I. Unexpectedly, a novel prominent band appeared after treatment by Exo I (Fig. 2B, lanes 4 -6). Using alkaline gel electrophoresis, we determined that this DNA species migrates at a position consistent with a 5Ј-end-labeled DNA fragment ending at the primosome assembly site (PAS) present on these substrates (data not shown). This PAS site corresponds to the B. subtilis pAM␤1 PAS site, which was described as capable of adopting a strong secondary structure (29). We therefore conclude that this stable DNA fragment occurs because of inhibition of degradation by Exo I by the strong DNA secondary structure formed at PAS.
When a Bs -containing substrate was used, a 5Ј-end-labeled top strand Bs -specific DNA fragment was observed in addition to fulllength ssDNA (Fig. 2B, lanes 7-9), as described earlier (18). Addition of Exo I after 2 min led to the appearance of the PAS fragment generated by degradation of the full-length ssDNA species up to the PAS site, showing that Exo I was functional. In sharp contrast, however, the Bs -specific ssDNA was largely resistant to degradation (Fig. 2B, lanes 10 -12). Such protection of the Bs -specific fragment from Exo I degradation was unexpected. Unlike the RecBCD system, where the RecA protein is strictly required to observe protection of -specific fragments from degradation, RecA protein was absent from our assays. Quantification of the protection over time revealed that although Exo I degraded almost all of the full-length ssDNA and about one-half of the Bs -specific ssDNA within 30 s, the remaining half of the Bs -specific ssDNA persisted with a half-life of ϳ10 -15 min (Fig. 2C). Protection of the Bs -specific fragment was not limited to this particular Bs site but was found for all three Bs sites tested (data not shown). Finally, protection of the Bs -specific fragment FIGURE 2. Bs -Specific ssDNA is protected from exonuclease I degradation by binding of the AddAB enzyme. A, plasmids pADGF0 ( Bs 0 ) or pADG6406-1 ( Bs ϩ ) were tailed so that AddAB can only access the DNA from one end (18). The distance (in nucleotides) from the Bs site, or the PAS site, to the labeled 5Ј-end is indicated. B, the indicated substrates were reacted with AddAB enzyme (at an AddAB to dsDNA ends ratio of 1 to 2) for 2 min, after which 0.8 units/l exonuclease I enzyme were added. Aliquots were withdrawn at the indicated times, and the distribution of reaction products was analyzed by agarose gel electrophoresis. A schematic depiction of the main product species, as had been established previously by Ché din et al. (18), is shown on the right side. C, the relative percentages of Bs -specific and of full-length ssDNA remaining after addition of exonuclease I are plotted as a function of time. The results are the average of two independent experiments, and each point is presented with the observed variation. D, processing by AddAB enzyme was as described in B except that SSB was omitted from the time course shown in right panel.

Stable Binding of AddAB Enzyme to
from Exo I degradation was observed under all conditions tested (low and high magnesium ion concentrations; data not shown). Protection was also not because of the formation of a secondary DNA structure, because Exo I could clearly degrade the full-length ssDNA from its 3Ј-end down to the PAS site, which is located a few hundred base pairs downstream of the Bs studied in Fig. 2B. We therefore surmised that protection from Exo I degradation might be because of the binding of a protein to the 3Ј-Bs -specific end generated upon attenuation of the AddAB enzyme nuclease activity at Bs .
Only two proteins are present in our reactions, AddAB enzyme and SSB protein. To test whether protection could be due to binding of the AddAB enzyme to the end of the Bs -specific fragment, we performed reactions in the absence of SSB protein. As with the RecBCD enzyme, the catalytic activity of the AddAB enzyme is reduced in the absence of SSB protein (30), particularly at higher magnesium ion concentrations and at higher ratios of DNA ends to enzyme. Under the saturating conditions used here (1 enzyme/end), most of the linear dsDNA was processed in 2 min despite the absence of SSB (Fig. 2D, lanes 5 and 6). As had been described for the RecBCD enzyme (30), DNA degradation by AddAB enzyme is greater in the absence of SSB, as judged by the absence of full-length ssDNA (compare lane 2 with 6). Because of this more extensive degradation, the yield of Bs -specific ssDNA was reduced. Nevertheless, protection of the remaining Bs -specific ssDNA from Exo I degradation was observed (60 Ϯ 10% protection, compared with 50 Ϯ 4% by AddAB in the presence of SSB as judged 60 s after addition of Exo I; Fig. 2D, lanes 7 and 8). This, therefore, rules out the possibility that SSB was responsible for the protection of the Bs -specific ssDNA, a conclusion that is consistent with its documented property of stimulating DNA degradation by Exo I. We therefore conclude that Bs -specific ssDNA is being protected from Exo I degradation by the binding of the AddAB enzyme to the 3Ј-end of the Bs -containing ssDNA. Because the protection gradually decreases with time (Fig. 2C), we speculated that the binding of AddAB to the 3Ј-end was non-covalent and subject to slow dissociation.
Protection of Bs -Specific ssDNA Occurs by Binding of AddAB Enzyme during Translocation-Two models can explain the protection of the Bscontaining ssDNA from degradation by Exo I. In the first model, binding to the 3Ј-end of the Bs -specific ssDNA by AddAB enzyme is an integral part of the recognition process and occurs in cis during translocation. After binding, the enzyme would continue to unwind and degrade DNA while bound to the Bs -containing 3Ј-end until it reaches the end of the duplex DNA or dissociates. In the second model, processing of the duplex DNA into full-length ssDNA and Bs fragments occurs without any binding, but after dissociating from the DNA, AddAB enzyme would rebind in trans to the 3Ј-end of the Bs -specific ssDNA.
To distinguish between these two possibilities, we tested whether adding AddAB back to reactions from which all proteins had been removed after processing would result in protection of the Bs -containing ssDNA. For this, tailed Bs -containing DNA was first reacted for 2 min with a saturating amount of AddAB enzyme as described above, followed by addition of Exo I for 30 s to degrade any susceptible ssDNA. The reaction was then stopped by phenol extraction. The DNA, stripped of all proteins, was recovered by ethanol precipitation. We consistently noticed that only a portion (59 Ϯ 11% (n ϭ 8)) of the Bs -specific ssDNA was recovered after that treatment, compared with the PAS fragment (Fig. 3, lanes 3 and 4). This specific loss of Bs -specific fragments can be explained if some of the ssDNA that was bound by AddAB was directed to the phenol phase during the extraction, as has been observed for other DNA-binding proteins. Addition of the AddAB enzyme back to the recovered reaction products for 1 min did not lead to any detectable change in the product distribution (Fig. 3, lane 5). However, the addition of Exo I to the recovered reaction products resulted in the complete loss of the Bs -specific ssDNA, indicating that the factor that was protecting the 3Ј-end had been removed by the phenol extraction (a similar loss of protection was observed after treatment of the reaction products with Proteinase K; data not shown). Note that the PAS fragment was also partially degraded upon addition of Exo I (Fig. 3, lane 6). This was consistently observed in independent experiments, although the extent of the degradation was variable. We speculate that the secondary structure that confers protection to the PAS fragment can become at least partially unfolded as a result of phenol extraction and ethanol precipitation, thus allowing Exo I to degrade a subset of molecules. Importantly, incubation with AddAB for 1 min followed by addition of Exo I failed to restore protection to the Bs -specific ssDNA (Fig. 3, lane 7). This indicates that binding of the AddAB enzyme to the 3Ј-end of the Bs -specific fragment in trans after dissociation of the complex does not account for the protection that is observed simultaneously with processing by AddAB enzyme.
Our results suggest that the 3Ј-end of the Bs -specific ssDNA is being bound directly by the AddAB enzyme and that this binding occurs in cis, during translocation of the enzyme through DNA. Note that this binding does not result in the inhibition of the enzyme in cis because the reaction products are fully unwound. This observation, in turn, implies that the AddAB enzyme, although bound to the 3Ј-end at the Bs -specific ssDNA, can continue to travel through the DNA to which it is bound (i.e. in cis); however, these very same AddAB-Bs -specific ssDNA complexes cannot initiate unwinding on subsequent DNA molecules until they have dissociated, thus accounting for the inactivation observed with Bs -containing substrates.
To confirm that binding of AddAB to the Bs -specific ssDNA occurred during translocation, we also performed competition experiments with ssDNA in trans. In these experiments, a prebound AddAB-dsDNA complex was challenged with a 100-fold molar excess of unlabeled, 35-nucleotide-long, single-stranded oligomers either carrying or lacking the Bs sequence at their 3Ј-end. These oligonucleotides were FIGURE 3. Addition of AddAB enzyme does not protect Bs -specific ssDNA when added back in trans to deproteinized reaction products. The 5Ј-end-labeled, tailed pADG6406-1 plasmid was used as a substrate. Reaction conditions are as described for Fig. 2. Two minutes after AddAB enzyme addition, a phenol extraction was carried out, followed by ethanol precipitation. The recovered products were redissolved in the initial reaction buffer, and AddAB enzyme and/or Exo I were added, as indicated. The reaction was allowed to proceed for 1 min after addition of each component before the samples were analyzed by agarose gel electrophoresis.
readily bound by the AddAB enzyme in electrophoretic mobility shift assays (data not shown). The reactions were initiated by addition of ATP. If binding to the Bs -specific ssDNA only occurs by reassociation after dissociation from the unwound DNA, then the excess of cold ssDNA competitors should block rebinding to the 3Ј-end of the Bs -specific ssDNA. When present at a saturating concentration (Fig. 4), the processing of dsDNA by AddAB enzyme in the presence of excess competitor ssDNA resulted in the complete production of the expected full-length ssDNA and Bs -specific ssDNA. This is consistent with the behavior of the RecBCD enzyme, for which addition of ssDNA oligonucleotides after formation of the enzyme-dsDNA complex has little effect on enzyme activity (31). More importantly, however, the Bs -specific ssDNA was still protected from Exo I, whereas the full-length ssDNA was not, irrespective of whether the competitor oligonucleotides carried a Bs sequence at their 3Ј-extremity (Fig. 4, lanes 5-8) or not (lanes 9 -12). Thus, this result suggests that AddAB binds specifically to the 3Ј-end of Bs -specific ssDNA during the course of translocation and processing, and not subsequently after dissociating from the DNA processing products.

DISCUSSION
In this study, we showed that the AddAB helicase/nuclease binds to its cognate regulatory sequence, Bs , in cis during unwinding, after which it continues to process the downstream dsDNA while still bound to the 3Ј-end of the Bs sequence. This conclusion is based on two observations. First, we showed that the ability of AddAB enzyme to initiate new cycles of dsDNA processing was reduced in response to Bs , suggesting that the enzyme is inactivated by a prior encounter with its regulatory sequence. Second, we observed that the Bs -specific ssDNA, generated by processing of Bs -containing dsDNA, was protected from degradation by Exo I because of the binding of AddAB enzyme to the 3Ј-end of the Bs -specific ssDNA. Protection was specific to these fragments and was observed in reactions where AddAB was the only protein present. The full-length ssDNA that was generated during DNA unwinding was not protected by the AddAB enzyme against Exo I degradation, and the removal of proteins by phenol extraction or Proteinase K treatment rendered the Bs -specific ssDNA sensitive to Exo I. In the absence of any treatment, protection from Exo I was slowly lost over time, indicating that AddAB enzyme dissociated from the Bs -specific ssDNA with a half-life of 10 -15 min.
Our observation that AddAB remains bound to the end of the Bs -specific fragments provides a simple explanation for the Bs -dependent inactivation of the helicase activity of the enzyme (Fig. 1). Assuming that Bs -bound AddAB molecules are not competent to initiate a new round of processing, the number of functional AddAB molecules in the reaction is expected to gradually decrease as a consequence of Bs recognition in the previous cycle. The fact that recognition of Bs is not 100% efficient (it is estimated to be ϳ25% per individual site) and that dissociation of the AddAB-Bs -ssDNA complexes does occur, albeit slowly, explains why multiple rounds of DNA unwinding can, and do, occur. Interestingly, reversible -dependent inactivation was also described for RecBCD enzyme, although it is observed only at low free magnesium ion concentrations (21), and the RecBCD enzyme can be reactivated by the addition of magnesium salt. In contrast, Bs -dependent inactivation of AddAB was observed under all conditions tested, and attempts to reactivate the AddAB enzyme resulted in only very partial, if any, reactivation (data not shown). One reason behind this difference might be related to the strength of the interaction between AddAB and its Bs sequence. Indeed, in the case of the RecBCD enzyme, our attempts to detect protection of -specific DNA fragments from Exo I under inactivating conditions were unsuccessful (data not shown), indicating that even under such conditions, the RecBCD enzyme does not bind to with a sufficient stability to afford protection.
The binding of AddAB to Bs occurs only in cis during translocation. This is evidenced by two different results. First, treatment of reaction  The 5Ј-end-labeled, tailed, pADG6406-1 plasmid was used as a substrate. A 100-fold molar excess of single-stranded 35-mers carrying ( Bs ϩ ), or not ( Bs 0 ), at their 3Ј-end were added to preformed complexes of AddAB and dsDNA. The reaction mixture was incubated for 2 min at 37°C after each addition. The reaction was initiated after preincubation of all components by addition of ATP, allowed to proceed for 2 min after which exonuclease I was added to all reactions. The reactions were kept at 37°C for another minute before the separation of the products by agarose gel electrophoresis.

Stable Binding of AddAB Enzyme to
products with phenol, which extracts all non-covalently bound proteins from DNA, renders the Bs -specific fragments sensitive to treatment by Exo I even when AddAB was added back prior to addition of Exo I (Fig.  3). Second, the presence of competitor ssDNA did not affect the protection of Bs -specific ssDNA, confirming that binding to Bs probably occurs without prior dissociation of the enzyme from DNA (Fig. 4). Together, these results suggest that binding of AddAB to Bs probably occurs as part of the normal recognition mechanism. It is also clear that binding to Bs does not lead to any detectable impairment in the ability of AddAB enzyme to continue processing the dsDNA to which it is bound. This, in turn, suggests the existence of a growing loop of ssDNA extending from the point of contact of the translocating enzyme to Bs . In the case of the RecBCD enzyme, direct visualization of the enzyme in single molecule experiments revealed the formation of an ellipsoidal fluorescent spot after recognition that both grew in intensity and remained associated with the enzyme (9). This spot was presumed to correspond to the expected -specific ssDNA, but the fact that it translocated with the enzyme suggested that it was anchored to it, rather than being spooled out freely. Our observation here that the AddAB enzyme forms a stable complex with the 3Ј-end of the Bs -specific ssDNA provide the first direct evidence that such a transient complex with indeed exists.
These data for AddAB enzyme, together with the larger collection of genetic, biochemical, and structural (32) findings for the RecBCD enzyme that have been cited earlier, suggest a general model by which sequences regulate the biological function of their respective enzymes. Specifically, as depicted in Fig. 5, the 3Ј-terminated strand of a broken dsDNA molecule is scanned for the presence of a sequence during translocation and is fed toward the domain responsible for nucleolytic degradation. Upon recognition of , the sequence itself, which is now at the 3Ј-end of the processed ssDNA, binds tightly to the enzyme, preventing movement of this DNA strand into the nucleolytic site and thereby limiting degradation of the -containing ssDNA. This molecular event underlies the observed attenuation of the 3Ј-to 5Ј-nuclease activity by . Binding, and the expected enzymatic conformational changes, are correlated with a brief pause in the translocation of the enzyme at . Because the -containing ssDNA is bound to the enzyme via the sequence, continued translocation beyond causes a loop of ssDNA to form that travels with the enzyme and that serves as a substrate for the next step of homologous recombination. This next step requires the assembly of a RecA protein filament on the -containing ssDNA, to the exclusion of the ssDNA-binding protein; this "loading" of RecA protein is mediated by direct protein-protein interactions with RecBCD enzyme (it is presently unknown whether AddAB can load its cognate RecA protein onto ssDNA). Persistence of a loop of ssDNA during the course of translocation also means that RecA protein needs to assemble onto this ssDNA loop. The assembly of RecA protein filaments occurs in the 5Ј 3 3Ј direction; however, the RecBCD/AddAB enzyme is traveling in the opposite direction (as defined by translocation along the -containing DNA strand 3Ј 3 5Ј). Thus, as we had postulated previously, assembly of the RecA nucleoprotein filament must be discontinuous, with repeated loading events by RecBCD/AddAB enzyme serving as nucleation points that facilitate subsequent polymerization in the presumed standard 5Ј 3 3Ј direction. As long as the 3Ј-end of the -containing ssDNA is bound to the translocating enzyme, a homologous pairing event will produce a paranemic joint molecule, which can be converted into a more stable plectonemic joint by the action of a topoisomerase. Alternatively, the paranemic joint molecule can be converted to a plectonemic joint when the 3Ј-end is released by dissociation of the helicase complex, and then the 3Ј-end can be used to initiate DNA replication. This dissociation frees the AddAB enzyme to reinitiate another processing cycle, and, finally, the subsequent steps of recombinational repair ensue.
This simple model accounts for many observed features of the RecBCD family of enzymes. Namely, both recognition and binding are dependent on a properly oriented recognition site and the changes elicited by this interaction are only observed in cis. Furthermore, once a sequence has been recognized and bound, the enzyme becomes insensitive to the presence of additional sequences (33,34). Both genetic and biochemical findings show that both -binding and -elicited effects can persist for distances of ϳ30 kbp downstream of (9). However, although binding persists during translocation, it is fully reversible upon dissociation of the enzyme from DNA. In the case of the RecBCD enzyme, dissociation is typically rapid, allowing the enzyme to catalytically process additional dsDNA molecules. In contrast, in the case of the AddAB enzyme described here, dissociation is slower, causing inactivation of the enzyme and permitting detection of the AddAB-Bs complex directly. This model does not require that any subunit of the complex be modified or ejected in any way. In agreement, recent single molecule experiments conclusively show that the RecD subunit is not ejected from the RecBCD enzyme upon recognition (24). Thus, we concluded that the regulation of recombination activity by sequences is a direct consequence of binding by the regulatory sequence to the RecBCD/AddAB enzyme; this binding, in turn, elicits an allosteric change in the enzyme that persists for the duration of the translocation event needed to initiate recombination.