Strand specificity of nicking of DNA at Chi sites by RecBCD enzyme. Modulation by ATP and magnesium levels.

RecBCD enzyme is essential for the major pathway of homologous recombination of linear DNA in Escherichia coli. It is a potent nuclease and helicase and, during its unwinding of double-stranded DNA, makes single-strand scissions in the vicinity of Chi recombination hot spots. We report here that both the strand that is cut and the position of the cuts relative to Chi depended on the ATP to Mg2+ ratio. With ATP in excess, Chi-dependent nicks occurred, as we have previously reported, four to six nucleotides to the 3′-side of the Chi octamer (5′-GCTGGTGG-3′) and were detected only on the strand bearing that sequence. Three differences were seen with Mg2+ in excess. 1) Chi-dependent 3′-ends were produced on the GCTGGTGG-containing strand closer to and within the Chi octamer. 2) Chi-dependent cuts occurred on the complementary DNA strand. 3) RecBCD enzyme destroyed the 3′-terminated strand of DNA from its entry point up to the vicinity of the Chi site, as others have previously reported. We show here that, with Mg2+ in excess, the enzyme continued to travel along DNA, after encountering a Chi site, releasing both strands of the DNA distal to Chi as single strands. We discuss potential biological consequences of these two modes of RecBCD enzyme-Chi interaction.

RecBCD enzyme (EC 3.1.11.5), encoded by the recB, recC, and recD genes of Escherichia coli, is the best characterized enzyme specific to the major (RecBCD) pathway of homologous recombination in E. coli conjugation and transduction (reviewed in Ref. 1). Purified RecBCD has many activities. It has potent ATP-dependent ss 1 and ds DNA exonuclease activities and a weak ATP-stimulated endonuclease activity on ss, but not ds, circular DNA. Under conditions that reduce these nuclease activities (e.g. 5 mM ATP and 1 mM Mg 2ϩ ) the enzyme totally unwinds ds DNA, up to 40 kilobase pairs or longer in the presence of SSB (2)(3)(4). Electron microscopy revealed that the enzyme travels unidirectionally and highly processively from a ds DNA end through the DNA, with either transient or permanent unwinding of the DNA behind the enzyme (5,6). Under such conditions the enzyme has a potent site-specific cleavage activity at Chi sequences (5Ј-GCTGGTGG-3Ј) (7), a recombinational hot spot specific to the RecBCD pathway (reviewed in Ref. 8). This cleavage occurs only as the enzyme is unwinding the DNA from right to left, as Chi is written here (9). The enzyme makes Chi-dependent ss nicks a few bases to the 3Јside of Chi on the GCTGGTGG-containing ("Upper") strand 2 but makes no detectable Chi-dependent nicks on the complementary strand (7,9) (Fig. 1, left). The activities of the purified enzyme, under these conditions, correlate well with the properties of Chi deduced from genetic studies and support a previously proposed model of Chi-stimulated recombination by the RecBCD pathway (8, 10) (see "Discussion").
RecBCD enzyme has been coupled with RecA and SSB proteins to produce Chi-dependent joint molecules from linear Chi-containing ds DNA and supercoiled Chi-free ds DNA (11)(12)(13). This reaction requires [Mg 2ϩ ] to be in excess over [ATP] (e.g. 8 mM Mg 2ϩ and 1 mM ATP) to allow RecA protein to promote homologous strand exchange between the linear DNA unwound by RecBCD enzyme and the supercoiled ds DNA. Under this condition RecBCD enzyme has a different reaction at Chi: it degrades the Upper strand from its 3Ј-end up to Chi, which attenuates the exonuclease active under this condition ( Fig. 1, right). Unwinding continues, to produce ss DNA with a 3Ј-end near Chi, which is used by RecA and SSB proteins to make a joint molecule. the Upper strand at Chi, and continues unwinding to produce ss DNA with Chi near its 3Ј-end ( Fig. 1, left). With [Mg 2ϩ ] Ͼ [ATP] the enzyme degrades the Upper strand up to Chi, ceases degradation, and continues unwinding to produce a similar ss DNA fragment (Fig. 1, right). To explore which reaction more nearly reflects that in E. coli cells, we have further characterized the reaction products under the two conditions. Markedly different products were observed, and we discuss their relationship to the types of genetic recombinants produced by the RecBCD pathway.

Proteins
Fraction V of RecBCD enzyme (14) was used throughout. Molar enzyme concentrations were calculated from the ds DNA exonuclease activity of the fraction (3) and the specific activity of fraction VI, which is nearly homogeneous (14) (3.3 ϫ 10 5 units/mg of protein). SSB was a gift from S. Kowalczykowski, University of California at Davis (12). Restriction enzymes and DNA-modifying enzymes were from New England BioLabs, Life Technologies, Inc., or Boehringer Mannheim.

DNA Substrates
Plasmids-Plasmid pBR322 and its Chi ϩ derivatives were purified as described elsewhere (15). Where necessary, RNA was removed by centrifugation (16) or by agarose gel electrophoresis. DNA concentrations are given as molarities of DNA molecules, initially determined from the A 260 of the purified plasmid DNAs. During DNA substrate constructions, DNA concentrations were calculated from the specific activity of labeled DNAs, as determined by trichloroacetic acid precipitation of samples taken immediately after labeling.
Numbering and Strand Designation Scheme-The first T in the unique EcoRI site of pBR322 is designated nucleotide number 1, with the numbering continued around the molecule through tet and then amp, using the corrected sequence (17). ϩ E arose from the deletion of an A at nt 989, while ϩ F arose from a C to T transition at nt 1498 (18). Restriction cuts are identified by the position of the 5Ј-terminus of the Upper strand of the DNA fragment which corresponds, for ϩ E and ϩ F, to the strand proceeding 5Ј to 3Ј with increasing nucleotide numbers.
5Ј-End-labeled Full-length pBR322-Plasmid pBR322 and its ϩ E or ϩ F derivatives were linearized with NdeI and then 32 P-labeled at the 5Ј-termini by treatment with calf intestinal phosphatase followed by incubation with [␥-32 P]ATP (DuPont NEN, 3000 Ci/mmol) and T4 polynucleotide kinase. Unincorporated ATP was removed either by repeated ethanol precipitation or by filtration through MicroSpin S-200 HR columns (Pharmacia Biotech Inc.).
Fragments with 5Ј-End Label on One Strand-°and ϩ F fragments were isolated (for convenience) from ϩ H and ϩ F ϩ H derivatives of pBR322 (12,13). Plasmid DNAs were digested with DdeI or AvaI, appropriate fragments were recovered with GeneClean (BIO 101, La Jolla, CA) after agarose gel electrophoresis, and their concentrations were measured by A 260 . Following 5Ј-32 P labeling, as above, the fragments were digested with appropriate restriction enzymes, and the desired fragments were recovered by polyacrylamide gel electrophoresis and electroelution.
Strand-specific Mid-labeled Substrates-pBR322 DNAs were linearized with StyI and labeled at the 3Ј-end of one or the other strand by incubation with the Klenow fragment of DNA polymerase I and [␣-32 P]TTP or dATP (DuPont NEN, 800 Ci/mmol), in the presence of the three other, unlabeled, dNTPs. After removal of enzyme and unincorporated nucleotides, DNAs were ligated, using T4 DNA ligase, and (after heat inactivation of the ligase) were digested with EcoRI. DNA molecules the length of linear pBR322 DNA were purified by agarose gel electrophoresis and purified with GeneClean. Hairpin-shaped oligonucleotide linkers bearing an EcoRI site (14) were ligated onto the DNA, which was then cut with NdeI, and the desired capped DNA fragment was purified as previously described (15). Gel electrophoresis of samples, following the EcoRI digestion, showed that 70 -90% of the incorporated label was in the expected fragments.
Size Standards-A BstEII digest of phage DNA (New England BioLabs) was labeled as described elsewhere (14). Maxam-Gilbert degradation reactions were as previously described (19).

Sample Preparation and Electrophoresis
Native Gels-Reactions were stopped as described elsewhere (13) and loaded directly onto horizontal 1.2% agarose gels in Tris acetate-EDTA buffer (16). Where indicated, samples were denatured by boiling immediately before electrophoresis. Gels were run at 1 V/cm for 16 h and dried onto DEAE paper before autoradiography at Ϫ70°C with intensifying screens.
Denaturing (Alkaline) Gels-Reactions were stopped by addition of an equal volume of 10 mM EDTA, 0.6 M sodium acetate, 10 g/ml tRNA. After addition of 5 volumes of ethanol, DNA was recovered by centrifugation for 15 min and resuspended in 10 l of alkaline loading buffer (21), before electrophoresis on 1% agarose gels in 50 mM NaOH, 1 mM EDTA, run at 1 V/cm for 16 h at room temperature. Gels were dried and exposed as above.
"Sequencing" Gels-Reactions were stopped by addition of 9 volumes of 10 mM EDTA, 0.3 M sodium acetate, 0.2% SDS, 80 g/ml sonicated carrier DNA, 4 g/ml tRNA. Following phenol and chloroform extractions to remove SSB, DNA samples were recovered by ethanol precipitation and dissolved by boiling in formamide loading buffer (19). Samples were analyzed by electrophoresis on 12% urea-Tris borate-EDTA polyacrylamide gels (19), followed by autoradiography of the undried gel at Ϫ70°C with intensifying screens.

Identification of Cleavage Sites
Examination of the RecBCD enzyme cleavage products on ureapolyacrylamide gels allowed identification of bands corresponding to every nucleotide position expected for the region of the gel between the restriction enzyme markers in adjacent lanes. These identifications were confirmed by reference to the Maxam-Gilbert chemical degradation products, after allowance had been made for the migration differences caused by the two terminal phosphates on the chemical degradation products.

RESULTS
Mg 2ϩ to ATP Ratio Determines Nuclease Activity-The interactions of Chi sites on ds DNA with purified RecBCD enzyme have been investigated under two different reaction conditions, one designed to minimize nonspecific nuclease activity by the enzyme (7), the other to allow RecBCD enzyme and RecA protein to function simultaneously (11). While there are several differences between the reaction conditions, the ratio of ATP to Mg 2ϩ ions is the factor that determines the level of the nonspecific nuclease activity of RecBCD enzyme (22), and hence the two conditions will be referred to as "Low Mg 2ϩ " and "High Mg 2ϩ ," respectively. They are defined under "Experimental Procedures." We presume the major factor is the presence of free Mg 2ϩ ions in High Mg 2ϩ conditions, most of the Mg 2ϩ ion being chelated to ATP in Low Mg 2ϩ conditions.
During the course of this work we examined the differences between these two reaction conditions, using Chi cleavage assays similar to those described below. We found that, in the High Mg 2ϩ reactions, MOPS buffers with pH values between 7.0 and 7.5 would substitute for the Tris acetate and that omission of the ATP-regenerating system did not affect the  (13). SSB, necessarily present under High Mg 2ϩ conditions, is omitted for clarity. The previously unreported Chidependent nicking of the Lower strand at Chi is described in this report.
results. Likewise we observed that omission of SSB from the Low Mg 2ϩ reactions did not alter the products observed (23). 3 Using a 3Ј-end-labeled substrate, we examined the survival of ds DNA, and its Chi-dependent cleavage products, under conditions related to High Mg 2ϩ conditions. 3 Neither DNA unwound by RecBCD enzyme nor Chi-dependent ss fragments were observed if Mg 2ϩ exceeded ATP either at high ATP concentrations (5 mM ATP, 8 mM Mg 2ϩ ) or at low ATP concentrations (1 mM ATP, 8 mM Mg 2ϩ ), even in the presence of SSB, as observed by Dixon and Kowalczykowski (13). When ATP was in excess, ss DNA reaction products survived both at high ATP concentration (5 mM ATP, 3 mM Mg 2ϩ ) and at low ATP concentration (1 mM ATP, 0.5 mM Mg 2ϩ ). The ratio of ATP to Mg 2ϩ (presumably free Mg 2ϩ ion) is thus the salient feature distinguishing the two reaction conditions, in accord with previous observations (22,24). In the experiments reported below, either 5Ј or internal 32 P labels were used, to allow detection of Chispecific reaction products under either reaction condition.
Cuts on Both Strands at Chi under High Mg 2ϩ Conditions-We show here that under High Mg 2ϩ reaction conditions Chi-dependent cutting of both strands of the DNA occurs. As seen in Fig. 2, two distinct reaction products resulted from the action of RecBCD enzyme on 5Ј-end-labeled ds DNA bearing a Chi site. These reaction products (which are indeed ss; Fig. 2B) migrated at the positions expected for ss fragments resulting from cutting of the two strands of the DNA at, or near, Chi.
The Chi sequence is uniquely determined by the sequence 5Ј-GCTGGTGG-3Ј (18), and no significant effects of flanking sequences have previously been found for the nicks on the Upper strand (7,9,23). Fig. 2 shows that the two Chi sequences tested, which are about 500 base pairs apart and share no obvious flanking sequence similarities (18), stimulated equivalent levels of cutting on the Lower strand of the DNA under High Mg 2ϩ conditions.
The fragments just described were Chi-dependent, since they were produced at much lower levels from DNA devoid of Chi sites (lanes 10 -13). In the absence of Chi the DNA was degraded, even in the presence of SSB, to generate both intermediates (mostly less than 800 nt long) and the final products of degradation by the enzyme (short oligonucleotides, seen at the bottom of the gel). Fragments apparently equivalent in size to those resulting from Lower strand nicking at ϩ F were seen in the °lanes but were severalfold enhanced by ϩ F.
The products shown in Fig. 2A resulted from a single round of reaction of RecBCD enzyme with ds DNA. Heparin prevents binding of RecBCD enzyme to the ends of ds DNA but does not disrupt preformed complexes (25). In the experiments in Fig. 2, RecBCD enzyme was bound to DNA ends in the absence of ATP, and the reactions were started by addition of a mixture of ATP and heparin. In the control lanes (1, 8, and 20), addition of heparin to the DNA prior to the addition of RecBCD enzyme prevented any detectable reaction on ds DNA. It is unlikely that the reaction products resulted from the subsequent action of RecBCD enzyme on ss products released in the first round of reaction, as SSB strongly inhibits the activities of the enzyme on ss DNA (2).
The yields of the Chi-dependent fragments depended upon the RecBCD enzyme concentration. The yield of fragments resulting from nicking at Chi on the Upper strand was maximal when there was approximately one RecBCD enzyme per DNA molecule. Presumably, at higher enzyme concentrations the reaction products were destroyed during the collision between enzyme molecules traversing the DNA from opposite directions, as previously observed (13).
Products resulting from the cutting of the Lower strand at Chi persisted, even at the highest enzyme concentrations used. We infer that the ss product was released by the enzyme, prior to its collision with another enzyme molecule, and was resistant to degradation by RecBCD enzyme, due to its coating of SSB. The maximal yield of the Lower strand cut fragment appears to be at least equivalent to that of the Upper strand cut.
In a previous report on Chi cleavage under these conditions, only faint bands are visible (Fig. 3, A and B, of Ref. 13) in positions consistent with cutting on the Lower strand as reported here. More recently, fragments corresponding to Lower strand cuts at Chi were reported (24); these fragments were Chi-dependent (cited in Ref. 24). Our results appear more consistent with the latter than with the former report.
Reaction Products Are Released as ss DNA-Native agarose gel electrophoresis was used to determine if the reaction prod-3 A. F. Taylor, unpublished results.

FIG. 2. Reaction of linear ds DNA with RecBCD enzyme under
High Mg 2؉ conditions. A, Chi-dependent cleavage of both strands of duplex DNA. DNA substrates were plasmid pBR322, and its ϩ E or ϩ F derivatives, linearized at the unique NdeI site and 32 P labeled at both 5Ј-ends. RecBCD enzyme was incubated with DNA, in reaction mixtures lacking both ATP and heparin, to allow binding of enzyme to DNA ends. After 5 min or more at 37°C, reactions were started by the addition of a mixture of ATP and heparin and were stopped after 90 s. The reaction mixture (10 l) contained 1 nM DNA and 1 mg/ml heparin in High Mg 2ϩ reaction buffer (see "Experimental Procedures") and the indicated concentrations of RecBCD enzyme. In lanes 1, 8, and 20, heparin was added to the reaction mixture prior to addition of RecBCD enzyme. Size markers, chosen to produce fragments approximately the same size as those produced by RecBCD enzyme cutting at Chi sites, were restriction digests of labeled substrate DNA. They were denatured before electrophoresis, as were the unreacted DNA samples in lanes 2, 9, and 19. EagI (lane 7) cuts pBR322 at base pair 940, close to the ϩ E site (nt 984 -992). In this figure, but not in other experiments (not shown), overdigestion by PpuMI (lane 14) produced cuts at sites other than the canonical site at base pair 1481, near the ϩ F site (nt 1493-1500). Samples were analyzed on a 1.2% native agarose gel. RecBCD per DNA denotes the ratio of RecBCD enzyme molecules to DNA molecules. B, release of reaction products as single strands. Reactions (as above but without heparin) were for 60 s with 0.5 nM RecBCD enzyme and 1 nM ϩ E DNA. RecBCD enzyme was omitted from lanes 25 and 26. The indicated samples were denatured by boiling immediately before analysis on a 1.2% native agarose gel. Lanes marked M contain marker DNA, an EagI digest of pBR322, as in A.
ucts were ss or ds DNA. Native (ds) and boiled (ss) size markers corresponding to the sizes expected for cutting at ϩ E or ϩ F were well separated by the gel (Fig. 2B, lanes 21 and 22). Migration of the RecBCD reaction products (lanes 23 and 24) was unaffected by boiling, showing that both the Upper strand and the Lower strand cut products were released as ss DNA.
Given the ss nature of the products, the comigration of the Chi-dependent species with the ss size markers confirms that the Chi-dependent cuts are in the immediate vicinity of the Chi sequences. Their exact locations are described below.
Strand Nicking during Unidirectional Travel by RecBCD Enzyme-Under Low Mg 2ϩ conditions, RecBCD enzyme nicks the Upper strand of DNA only if the enzyme has approached the Chi site from the biologically appropriate direction (26), that is by entering the DNA at the 3Ј-terminus of the strand containing the GCTGGTGG sequence (9). We considered the possibility that Lower strand Chi cleavage might be the result of RecBCD enzyme encountering Chi from the opposite direction, either by itself or as a result of a collision with a RecBCD enzyme molecule that was "paused" at the Chi site.
We tested this possibility by using synthetic hairpin-shaped oligonucleotides ligated onto one end of the DNA substrates. Hairpins prevent the entry of RecBCD enzyme under Low Mg 2ϩ reaction conditions (14). We first tested whether such hairpin-shaped molecules were resistant to RecBCD enzyme under High Mg 2ϩ conditions in the absence of SSB. Hairpin oligonucleotides, labeled at their 5Ј-termini, were self-ligated, and the double-length products purified (Fig. 3A). Removal of the tips, by restriction enzyme digestion, provided positive controls for RecBCD enzyme sensitivity. The "double-length" DNA was reacted with increasing concentrations of RecBCD enzyme and reaction products assayed by trichloroacetic acid precipitation (Fig. 3B). The small fraction of 32 P released by low concentrations of RecBCD enzyme presumably came from double-length open-ended ds DNA molecules formed by annealing of two hairpin molecules (as shown in Fig. 3A, top). However, the remainder of the DNA was resistant to 100 times the concentration of RecBCD enzyme needed to solubilize the control DNA species (unligated hairpin oligonucleotides or doublelength molecules whose tips had been removed by restriction enzyme digestion). Thus, hairpin oligonucleotide caps are indeed resistant to cleavage by RecBCD enzyme under High Mg 2ϩ conditions, even in the absence of SSB.
Substrates for Chi cutting were made by ligating hairpin oligonucleotides onto fragments of pBR322 with or without Chi sites (Fig. 4) to produce DNA molecules in which RecBCD enzyme could approach Chi only in the active orientation. DNA molecules were prepared uniquely labeled in the Upper or the Lower "strand" of the DNA, in the position labeled A or T in Fig. 4, to allow separate examination of the two strands of the DNA. In the ϩ E substrates the 32 P label was between the ds DNA end and Chi, allowing detection of fragments extending from Chi toward the ends of the DNA strands, while in the ϩ F substrates the 32 P label was distal to Chi, allowing detection of fragments extending from Chi toward the hairpin.
Reaction of these substrates under Low Mg 2ϩ conditions produced products similar to those previously observed using substrates labeled on both strands (15). Reaction of either ϩ DNA substrate produced half-length (2300 nt) molecules, resulting from RecBCD enzyme molecules passing Chi without cutting it, then cutting the hairpin from within; fragments of this length were the only specific reaction products seen with °s ubstrates (Fig. 4A, lanes 4 -6 and 11-13). Such fragments were seen with substrates labeled either in the Upper or the Lower strand, confirming that there is little or no preferential degradation of either strand under Low Mg 2ϩ conditions. Un-der High Mg 2ϩ conditions, however, the half-length fragments are seen with Lower (5Ј-terminated) strand labeled substrates (lanes 2 and 3) but not with Upper (3Ј-terminated) strand labeled substrates (lanes 9 and 10), confirming the strandspecific degradation previously reported under these conditions (13).
Nicking of the Upper strand at ϩ E under Low Mg 2ϩ conditions produced a 1300-nt fragment with the T-labeled substrate (Fig. 4B, lanes 11-13) and a 3300-nt fragment with the Alabeled DNA (Panel B, lanes 4 -6). The absence of a specific 1300-nt fragment with A-labeled DNA under Low Mg 2ϩ conditions (Panel B, lanes 4 -6) is further evidence of the lack of Lower strand cutting under these conditions. Under High Mg 2ϩ conditions, a 1300-nt fragment was produced with A-labeled DNA (Fig. 4B, lanes 2 and 3), demonstrating that RecBCD enzyme molecule(s) that encounter Chi from only one direction can indeed produce Lower strand cuts at Chi. The product was Chi-dependent (compare lanes 2 and 3 of Panel B with those of Panel A) and was produced at low enzyme concentrations (lane 2 with 2 enzyme molecules per DNA end).
Experiments with 3Ј-end-labeled DNA molecules bearing Chi sites have previously been interpreted as evidence that RecBCD enzyme degrades the 3Ј-terminated strand until it encounters a Chi sequence (13). Such data were consistent with FIG. 3. Resistance of hairpin ends to RecBCD enzyme degradation. A, construction of the substrates. The self-complementary oligonucleotide shown was 5Ј-end labeled with 32 P, and some of it was self-ligated (14). Monomer and dimer length molecules were purified by gel electrophoresis. Some of the purified dimer DNA was treated with ClaI or HaeIII to produce linear ds DNA with internal labels. The intermolecular dimer shown results from isomerization of the hairpin monomer material and is not separated from the ligated dimer on gel purification. B, reaction of hairpin DNAs with RecBCD enzyme. Reactions (10 l) contained, in High Mg 2ϩ reaction mix lacking SSB, the indicated concentrations of RecBCD enzyme and monomer (0.11 nM) or dimer (0.09 nM) hairpin DNAs or restricted dimer hairpin DNAs (0.14 nM). After 10 min of incubation at 37°C, the reactions were analyzed by trichloroacetic acid precipitation (3). Precipitate and supernatant fractions were counted by Cerenkov radiation. The graph plots the percentage of the counts precipitable before RecBCD enzyme reaction subsequently released by RecBCD enzyme. the alternative view that RecBCD enzyme removed only a few 3Ј-terminal-labeled nucleotides, rendering invisible (in the reported experiments) a postulated DNA fragment extending from the Chi site to near the 3Ј-terminus. Results in Fig. 4B, with a ϩ E substrate labeled internally in the 3Ј-terminated strand, under High Mg 2ϩ conditions, failed to reveal either any fragment of the postulated size (1300 nt) or any Chi-dependent smear indicative of limited degradation. The experiment thus shows that degradation of the 3Ј-terminated strand extends at least 900 nt from the terminus, and presumably up to Chi.
If the RecBCD enzyme molecule that nicked the Lower strand at ϩ E under High Mg 2ϩ conditions also failed to degrade the 3Ј-terminated strand, then a 3300-nt T-labeled fragment would be observed (analogous to the 3300-nt A labeled fragment seen in lanes 4 -6 of Panel B under Low Mg 2ϩ conditions). The failure to observe such a fragment (Panel B, lanes 9 and 10) implies that RecBCD enzyme molecules that cut the Lower strand do indeed also degrade at least part of the Upper strand of the DNA.
A capped DNA substrate, with the label distal to the Chi site, permits investigation of the nuclease activities of RecBCD enzyme after it has encountered Chi ( ϩ F in Fig. 4C). The oligonucleotide cap is essential to prevent RecBCD enzyme molecules approaching from the opposite direction and destroying the 3Ј-end of the Lower strand. Cleavage of the Upper strand at ϩ F would produce, in the absence of other cuts, a 3800-nt fragment. Such fragments were the most prominent product observed, under Low Mg 2ϩ conditions, with the 32 P label in either strand (Panel C, lanes 4 -6 and 11-13). Fragments of this size were much less prominent under High Mg 2ϩ conditions ( Panel C, lanes 2, 3, 9, and 10) and may be no more frequent than a contaminant band present in the substrate (lanes 1 and 8). The most prominent reaction product with ϩ F DNA, under High Mg 2ϩ conditions, was 1500 nt long, resulting from RecBCD enzyme cutting both at ϩ F and at the tip of the DNA. The occurrence of this 1500-nt fragment, and the absence of a 2300-nt fragment with Upper strand labeled DNA, confirms that the nuclease activity of RecBCD enzyme on 3Јterminated strands is attenuated at Chi (13). The yield of this fragment was, however, much lower than that of equivalent 3800-nt fragment produced under Low Mg 2ϩ conditions, showing that the attenuation under High Mg 2ϩ conditions is only partial (Panel C, lanes 9 and 10 versus lanes [11][12][13]. Observation of the 1500-nt fragment with Lower strand labeled DNA shows that RecBCD enzyme neither gains a nuclease activity nor loses its hairpin cutting activity upon nicking the Lower strand at Chi. In summary, the experiments in Fig. 4 demonstrate that nicking of the Lower strand at Chi, under High Mg 2ϩ conditions, is not a result of collisions between RecBCD enzyme molecules coming from opposite directions, or from enzyme molecules encountering Chi from the "opposite" direction. These results, when taken with the results from Fig. 2, imply that a single RecBCD enzyme, encountering a Chi in the "active" orientation, is sufficient to catalyze nicking of the Lower strand of the DNA at Chi. The scission at Chi on the Lower strand is indeed a nick (or two or more closely spaced nicks), as shown by the recovery of fragments to both sides of it, while that on the Upper strand results from degradation of the 3Јterminated strand up to Chi.
Position of Upper Strand Cuts at Chi-Analysis of RecBCD enzyme reaction products on "sequencing" gels allows determination of the position of the cut sites at Chi, using restriction enzyme digestion and chemical degradation products of the same substrate DNAs as size markers. As seen in Fig. 5 (lanes  23 and 24), RecBCD enzyme cleavage of the Upper strand at ϩ F under Low Mg 2ϩ (5 mM ATP, 3 mM Mg 2ϩ ) conditions was at the same two nucleotides as previously reported (9) for somewhat different reaction conditions (5 mM ATP, 1 mM Mg 2ϩ , 100 mM NaCl). The SSB, present here but not in the previously published results (9), did not influence the positions of the cuts. Reaction under High Mg 2ϩ conditions, however, caused a marked change in the distribution of Chi-specific Upper strand cuts (Fig. 5, lanes 1 and 2). The cuts were less pronounced, distributed over more nucleotides, and closer to (or within) the Chi octanucleotide.
The High Mg 2ϩ and Low Mg 2ϩ reactions were carried out with different buffers and at different pH values, and with different ATP concentrations, but the salient difference was the ATP to Mg 2ϩ ratio. As the ATP concentration in a High Mg 2ϩ reaction was increased (lanes 1 through 12 versus lane 23), the Chi-dependent cut sites moved toward the positions seen in the Low Mg 2ϩ reactions. Similarly, when the ATP concentration in a Low Mg 2ϩ reaction was decreased, the reaction products changed to resemble those of a High Mg 2ϩ reaction ( lanes  16 -24 versus lane 2).  7 and 14) were a 3Ј-end-labeled BstEII digest of phage DNA. RecBCD:DNA denotes the ratio of RecBCD enzyme molecules to DNA molecules. The diagram below the figure shows the locations of the Chi sites and 32 P labels, and their distances in nucleotides from the hairpin. The small diagrams next to the autoradiograms denote the parts of the molecules present in the adjacent fragments. Half-length molecules, cut at the hairpin, are not indicated in Panels B and C.
A qualitative summary of the cut sites under High and Low Mg 2ϩ conditions is shown on the right of Fig. 5 and is later compared to Lower strand cuts.
Position of Lower Strand Cuts at Chi-Investigation of Lower strand cuts (Fig. 6) revealed a somewhat different picture. With 10 mM ATP, 8 mM Mg 2ϩ (lanes 2 and 3), Chi-dependent cuts were barely detectable, but under High Mg 2ϩ conditions (1 mM ATP, 8 mM Mg 2ϩ ) Chi-dependent cuts at at least 9 positions were prominent. Outside that region, the patterns of cuts were indistinguishable in °and ϩ F substrates. Most of the High Mg 2ϩ reactions shown used heparin to prevent multiple rounds of RecBCD enzyme reaction. The enzyme titration (lanes 11-25) revealed that Chi-enhanced cuts were seen at low enzyme doses and reached a maximum at about one RecBCD enzyme per DNA, as shown by PhosphorImager analysis. 3 Lower strand cuts were thus the result of a single round of reaction with one RecBCD enzyme molecule.
The results of PhosphorImager analysis of the °and ϩ F reaction products at the highest RecBCD enzyme concentration are shown in Fig. 6C. Maximal Chi-enhanced cleavage (ϳ3fold) was after the C (in the center of the Chi sequence) at nt 1497, with enhancement disappearing within 9 nt to either side. These results are shown schematically in Fig. 6 and are compared, in Fig. 7 (Figs. 4 -7) (7, 9). This reaction produces three ss DNA frag-ments (Fig. 1, left). Under High Mg 2ϩ conditions ([Mg 2ϩ ] Ͼ [ATP]) RecBCD enzyme degrades the Upper strand from its 3Ј-terminus to Chi, nicks the Lower strand within or near either side of the Chi sequence 3Ј-CGACCACC-5Ј, and continues unwinding (Figs. 2 and 4 -7) (13,24). This reaction produces three ss DNA fragments and oligonucleotides (Fig. 1,  right). Upon cutting at Chi the enzyme loses its Chi-nicking activity (Low Mg 2ϩ conditions) (15) or its 3Ј to 5Ј degrading activity (High Mg 2ϩ conditions) (13). The yields of Chi-cut products are considerably greater under Low Mg 2ϩ conditions than under High Mg 2ϩ conditions (Figs. 4 -6); SSB is required for detection of Chi-cut products under High Mg 2ϩ conditions but not under Low Mg 2ϩ conditions (Figs. 4 -6) (7, 9, 23) (data not shown).
ATP and Mg 2ϩ Levels in E. coli-Knowledge of the levels of unbound ATP and Mg 2ϩ within E. coli would enable us to decide which of the reactions reported here is more pertinent to the recombination reactions catalyzed by RecBCD enzyme within the cell. The total ATP concentration in E. coli is reported to be 1.6 mM (27, 28) or 2.7 mM (29) and varies little with growth rate (27) but can vary from 0.6 mM to 2.2 mM depending on growth conditions (28). While the total Mg 2ϩ ion concentration in E. coli is about 100 mM (30), only between 1 and 2 mM is estimated to be "free in solution in the cellular sap" (31). This estimate of free Mg 2ϩ was somewhat indirect, and it is unclear whether it would have included Mg 2ϩ ions bound to ATP. A more direct measurement using fluorescent indicators estimates the free intracellular concentration of Mg 2ϩ in mouse 3T3 fibroblasts to be less than 1 mM (32) but has not, to our knowledge, been reported for E. coli. 31 P NMR (33) can distinguish free ATP from Mg-bound ATP; it was used with anaerobic E. coli but was unable to separate the ATP signals from those of other nucleoside triphosphates (34). The authors observed that all detectable NTP was bound to Mg 2ϩ and suggested that free Mg 2ϩ was 1 mM. Values obtained in eu- karyotes, in which the ATP signal can be separated from that of other nucleoside triphosphates, are also typically less than 1 mM free Mg 2ϩ (33). The free ATP and Mg 2ϩ concentrations in E. coli thus appear to be approximately equal, and we cannot decide on this basis which condition studied here more nearly reflects that in E. coli.
Protection of Linear DNA by Chi in E. coli-Purified RecBCD enzyme is differentially inactivated when it encounters a Chi site (13,15). Such inactivation persists for greater than 20 min under Low Mg 2ϩ conditions (15,35) but is rapidly reversed if the reactions are switched to High Mg 2ϩ conditions (35). The fate of intracellular linear DNA molecules bearing Chi sites, coupled with these observations, may reveal the conditions inside the cell.
RecBCD enzyme rapidly destroys DNA molecules with ds DNA ends generated within E. coli by the intracellular action of a type I restriction enzyme (36), by rolling circle plasmid replication (37,38) or by the intracellular induction of phage 's terminase (39). Such degradation is reduced by Chi on the linear DNA molecules (37)(38)(39). The requirement for RecA pro- FIG. 6. Cleavage sites at Chi on the Lower strand of DNA. Substrates were fragments of pBR322, and its ϩ F derivative, extending clockwise from the unique EagI site at nt 940 to the DdeI site at nt 1582, bearing a 5Ј-32 P label on the DdeI end. A, DNA (0.25 nM) was reacted with 0.1 nM RecBCD enzyme for 1 min at 37°C, using High Mg 2ϩ conditions, but with 10 mM ATP, prior to electrophoresis on an 8% polyacrylamide sequencing gel. B, DNA (2 nM) was reacted with RecBCD enzyme under High Mg 2ϩ conditions, using heparin to prevent reinitiation as in Fig. 2. In lanes 8,9,27, and 28, heparin was omitted, while in lanes 5 and 6 it was added prior to RecBCD enzyme. A mixture of restriction enzyme fragments of the substrate was run in each of the unmarked lanes. Lanes 16 and 17 contain Maxam-Gilbert degradation products of the Chi°s ubstrate, with its sequence superimposed (white and black letters are used merely to aid legibility). The corresponding Chi ϩ sequence is given to the right of the figure. C, PhosphorImager analysis of Chi-enhanced cuts (lanes 24 and 25). The entire width of each lane was analyzed, using area analysis and manual baseline adjustment (ImageQuant Version 3.3, Molecular Dynamics Inc.). Numbers on the graph represent the enhancement of Chi cutting at each position, as measured by the relative areas under the Chi ϩ and Chi°peaks at each position. The widths of the arrows in part B are proportional to the relative magnitude of the Chi enhancement at each position, as determined by subtracting the area under each Chi ϩ peak from that under the corresponding Chi°peak.
tein (37)(38)(39) and SSB (39) suggests that RecBCD-mediated homologous recombination, rather than simply the inactivation of the exonuclease activity of RecBCD enzyme by Chi or the titration of the enzyme by the large number of ds DNA ends in the cell, is responsible for the apparent loss of exonuclease activity.
Two groups have investigated the trans effect of a plasmid bearing a Chi site on the survival of a compatible Chi-free plasmid in the same cell. Zaman and Boles (38) did not find extensive protection of a Chi°plasmid by a compatible Chi ϩ plasmid in the same cell. Kuzminov et al. (39), however, found that intracellular linearization of a plasmid bearing Chi largely prevented the degradation of an unrelated linearized Chi°plasmid in the same cell. If, in contrast to the conclusion in the preceding paragraph, this protection reflects inactivation of RecBCD enzyme, such inactivation must be long lived. Since, however, the Chi-mediated inactivation of purified RecBCD enzyme is rapidly reversed by High Mg 2ϩ conditions (35), the conditions within the cell may more nearly resemble the Low Mg 2ϩ conditions described here. Fig. 8 we explore the consequences, for a Chi initiation model of recombination (10), of the two modes of RecBCD enzyme action at Chi. We have not yet determined whether a single RecBCD enzyme can effect scissions on both strands, and so we consider several possible actions of RecBCD enzyme at Chi.

Genetic Consequences of Upper versus Lower Strand Cuts at Chi-In
First, in Fig. 8A we diagram the essence of the previously proposed model, under conditions in which the Upper, but not the Lower, strand is nicked. RecBCD enzyme enters the right end of the black parental ds DNA, and travels along the DNA, unwinding and rewinding it (5). This rewinding occurs more frequently with long (Ͼ10 kilobase pairs) DNA used for electron microscopy (5), and likely with the DNA substrates typical in E. coli recombination, than with the short (Ͻ5 kilobase pairs) DNA used here and in other studies with Chi and purified RecBCD enzyme (7, 9, 12-15, 23, 24, 35). When the enzyme encounters Chi, it nicks the Upper strand of the DNA, and continues to unwind the DNA, resulting in extrusion of a 3Јended single strand. SSB and RecA protein then promote invasion of the homologous region of the gray parental ds DNA, to form a D-loop. Cleavage of the D-loop, followed by annealing and ligations, produces a Holliday junction (central panel) which in turn is cleaved by one or more enzymes in E. coli to produce two reciprocal recombinants (40).
Second, if intracellular conditions allow the exonuclease activity of RecBCD enzyme to degrade the Upper strand up to the 3Ј-side of Chi, the same basic model still holds (Fig. 8B). After the Holliday junction has formed, the 3Ј-terminated strand of gray DNA can prime DNA synthesis, resulting in the recreation of the degraded 3Ј-terminated strand. Hence reciprocal recombinants could be formed, regardless of the intracellular conditions, if the Chi-mediated strand cleavages are restricted to the Upper strand of the DNA.
Third, if the Lower strand, but not the Upper strand, is nicked at Chi, then recombination could occur as shown in Fig.  8A, except that the invading single strand extending to the left of Chi will have a 5Ј-end, rather than a 3Ј-end. Experiments with purified RecA protein and substrates similar to those in Fig. 8A suggest that the 3Ј-terminus is preferred in such reactions (41). Cleavage of the Lower strand produces a 3Ј-ended fragment extending to the right of Chi (Fig. 8C), which would lead to Chi stimulation of recombination to the right of Chi; this has not been observed (e.g. see Refs. [42][43][44]. These observations question whether Lower strand cutting and, hence, High Mg 2ϩ conditions do indeed occur in E. coli. Fourth, Chi-mediated cuts on both strands (Fig. 8C) preclude the recovery of reciprocal recombinants, for markers bracketing Chi, from just the two DNA duplexes in the diagram. If the Upper strand is degraded up to Chi (or merely nicked at Chi) and the Lower strand is nicked at Chi, then a Holliday junction can still be formed (Fig. 8C). However, as the joint molecule is missing part of one of the parental DNAs (black DNA to the right of Chi), reciprocal recombinants cannot be formed but could be recovered if the incomplete arm of the Holliday junction recombines with a third ds DNA (45). The partially unwound DNA structure in the upper panel of Fig. 8C, produced by continued unwinding by RecBCD enzyme after cutting both strands at Chi, is equivalent to the "split end" structure, hypothesized to be a recombination intermediate (46).
In summary, the models in Fig. 8, A and B, are essentially equivalent to that previously proposed (10), which can account for conjugational and transductional recombination in E. coli (47) and the stimulation at and to the left of Chi (e.g. see Refs. [42][43][44]. The model in Fig. 8C does not allow reciprocal recombinants to emerge from a single interaction between only two parental DNA molecules but reciprocal recombinants could arise when three or more DNA molecules are involved, as in phage crosses (45). This model could also account for the integration of ds DNA fragments into the E. coli chromosome during conjugation or transduction. The model in Fig. 8C, however, predicts stimulation to the right of Chi, which has not been detected (e.g. see Refs. [42][43][44].
Following the fate of DNA molecules during recombination in E. coli, for example by Southern blot hybridization of DNA extracted from cells, may reveal which mode of RecBCD enzyme action prevails in E. coli and, hence, support one or another model of recombination. Knowledge of the products of RecBCD enzyme reaction on Chi-containing DNA, as reported here, will aid the design and interpretation of such experiments.