doi:10.1006/jmbi.2000.3556 available online at http://www.idealibrary.com on J. Mol. Biol. (2000) 298, 7±20

The Bacillus subtilis AddAB Helicase/ is Regulated by its Cognate Chi Sequence in Vitro

FreÂdeÂric CheÂdin1,2, S. Dusko Ehrlich2 and Stephen C. Kowalczykowski1*

1Sections of Microbiology and The AddAB enzyme is important to homologous DNA recombination in of Molecular and Cellular Bacillus subtilis, where it is thought to be the functional counterpart of the Biology, University of RecBCD enzyme of . In vivo, AddAB responds to a speci®c California, Davis, CA 95616- ®ve-nucleotide sequence (50-AGCGG-30 or its complement) in a manner 8665, USA analogous to the response of the RecBCD enzyme to interaction with w sequences. Here, we show that puri®ed AddAB enzyme is able to load at 2Laboratoire de GeÂneÂtique a double-stranded DNA end and is both a DNA helicase and nuclease, Microbienne INRA, Domaine whose combined action results in the degradation of both strands of the de Vilvert, 78352, Jouy en DNA duplex. During translocation, recognition of the properly oriented Josas, Cedex, France sequence 50-AGCGG-30 causes attenuation of the AddAB enzyme nucle- ase activity that is responsible for degradation of the strand 30-terminal at the entry site. Therefore, we conclude that 50-AGCGG-30 is the B. subtilis

Chi site and it is hereafter referred to as wBs. After encountering wBs, both the degradation of the 50-terminal strand and the helicase activity persist. Thus, processing of a double-stranded DNA end by the AddAB enzyme produces a duplex DNA molecule with a protruding 30-terminated single-stranded tail, a universal intermediate of the recombination process. # 2000 Academic Press *Corresponding author Keywords: Chi; nuclease; ; helicase; AddAB

Introduction cal for the initiation of this repair pathway (for a review, see Smith, 1988; Kowalczykowski et al., The repair of genotoxic lesions is a central aspect 1994). RecBCD enzyme will bind to dsDNA ends of DNA metabolism in living cells, and many com- and proceed to unwind the DNA while extensively 0 plex enzymatic processes are devoted to maintain- degrading the strand terminating 3 at the entry ing the integrity of the . Homologous site (referred to as the ``top-strand''; see Figure 1) recombination is one important pathway by which (Dixon & Kowalczykowski, 1993). This destructive potentially lethal lesions, such as double-stranded mode, however, is switched to a recombinogenic DNA breaks, can be repaired. These lesions arise mode after the enzyme recognizes a pro- frequently during normal cell growth if DNA repli- perly oriented w sequence. The w octamer, cation is impeded (Michel et al., 1997) or, in eukar- 50-GCTGGTGG-3', initially de®ned as a recBC- yotes, at the onset of meiosis (Baudat & Nicolas, dependent recombination hotspot (Lam et al., 1974; 1997; Dernburg et al., 1998; McKim & Hayashi- Smith et al., 1981; Stahl et al., 1975), elicits a num- Hagihara, 1998; Sun et al., 1989). In Escherichia coli, ber of biochemical changes in RecBCD enzyme's double-stranded DNA (dsDNA) breaks are behavior: the enzyme brie¯y pauses at w before repaired by homologous recombination between unwinding resumes, at which point the top-strand the damaged DNA and an intact copy of the nuclease is down-regulated, and a weaker nuclease chromosome. Numerous genetic and biochemical activity is activated on the opposite strand data have shown that the RecBCD enzyme is criti- (Anderson & Kowalczykowski, 1997a; Dixon & Kowalczykowski, 1991, 1993). These modi®cations produce a processed dsDNA with a 30 single- Abbreviations used: ssDNA, single-stranded DNA; dsDNA, double-stranded DNA; SSB protein, single- stranded DNA (ssDNA) tail terminating at w that stranded DNA-binding protein. can be bound by the RecA protein to initiate the E-mail address of the corresponding author: homologous pairing phase of genetic recombina- [email protected] tion. The formation of a RecA ®lament on ssDNA

0022-2836/00/010007±14 $35.00/0 # 2000 Academic Press 8 AddAB-Chi Interactions in Vitro

Figure 1. Four potential wBs- speci®c fragments can be produced upon interaction of the AddAB

enzyme with wBs-containing DNA. The region of the DNA between

wBs and the enzyme's entry point is referred to as the upstream region

while the region past wBs to the opposite end of the DNA is the downstream region. The top-strand is the 30-terminating strand at the entry point; the opposite strand is referred to as the bottom-strand.

The various putative wBs-fragments are indicated.

is a crucial step in this initiation reaction, and the to interact with an enzyme possessing both heli- w-activated RecBCD enzyme is directly responsible case and exonuclease activities. In H. in¯uenzae, for coordinating the loading of the RecA protein this enzyme is a clear homolog of RecBCD. How- onto w-containing ssDNA. This ensures maximal ever, for the two Gram-positive L. lactis participation of this w-containing ssDNA in the and B. subtilis, the corresponding enzymes seem to process of genetic recombination (Anderson & belong to a novel class of enzymes that are func- Kowalczykowski, 1997b). tionally, but not structurally, homologous to the The interaction between the E. coli RecBCD RecBCD class of enzymes. enzyme and w has been extensively studied, and it In B. subtilis, the model organism for Gram-posi- serves as a model for understanding how homolo- tive bacteria, the counterpart of the RecBCD gous DNA recombination is initiated in bacteria. enzyme is thought to be the AddAB enzyme. However, until very recently, evidence for conser- Genetically, mutations in either addA or addB genes vation of an interaction between a helicase/exonu- result in reduced cell viability, increased sensitivity clease and a w-like regulatory sequence as a means to DNA-damaging agents, and reduced levels of to initiate homologous recombination was strik- DNA recombination in certain strains (Kooistra ingly sparse. RecBCD homologs have been found et al., 1988; Alonso et al., 1988), which are pheno- in several enterobacteria as well as in other, more typic characteristics of recB or recC mutations in distantly related bacteria, such as Haemophilus E. coli. In addition, expression of AddAB enzyme in¯uenzae, Pseudomonas aeruginosa, Neisseria menin- in E. coli complements RecBC phenotypes gitidis and Mycobacterium tuberculosis. However, (Kooistra et al., 1993) and, in vitro, the AddAB RecBCD-w interactions, as de®ned by the pro- enzyme possesses exonuclease and helicase activi- duction of a diagnostic w-speci®c fragment by the ties (Haijema et al., 1996a). From previous in vivo E. coli enzyme, seem to be conserved only in work, we showed that a speci®c ®ve-nucleotide enteric bacteria that are closely related to E. coli sequence, 50-AGCGG-30 allowed formation of a (McKittrick & Smith, 1989). This suggests that high molecular weight form of plasmid DNA, pre- either the RecBCD-w interaction is restricted to a sumably by protecting this DNA from degradation small group of microbes or that sequences other by AddAB (CheÂdin et al., 1998). Although hotspot than the canonical w sequence might be used in activity was not demonstrated for this sequence, it different organisms. This latter view is supported was proposed that this pentamer was the B. subtilis by the recent identi®cation of speci®c short DNA analog of the well-described Chi site of E. coli (wEc). sequences (5-8 bp) that protect dsDNA from exo- Below, we establish that this supposition was cor- nucleolytic degradation in the bacteria Lactococcus rect, and 50-AGCGG-30 will therefore be referred to lactis (Biswas et al., 1995), H. in¯uenzae (Sourice as wBs. et al., 1998) and Bacillus subtilis (CheÂdin et al., 1998). Despite these parallels with the E. coli system, In all cases, those putative w sequences are believed the biochemical consequences of the interaction AddAB-Chi Interactions in Vitro 9

between AddAB enzyme and its putative wBs concentrations (Mg(OAc)2). As expected, the initial sequence are unclear. Although the in vivo obser- dsDNA is rapidly processed by the enzyme and is vations suggest that the enzyme loses its ability to converted to full-length ssDNA and a variety of degrade double-stranded DNA after wBs is recog- degradation products. The amount of full-length nized (CheÂdin et al., 1998), the exact nature of the ssDNA, however, varied strongly with the concen- changes elicited by this interaction could not be tration of magnesium ions. At 0.5 mM Mg(OAc)2, inferred. Here, we show that the puri®ed AddAB about 80 % of the unwound DNA was full-length enzyme is a potent helicase and exonuclease ssDNA after ®ve minutes, whereas at 2 mM which, in contrast to the RecBCD enzyme, Mg(OAc)2, only 10 % of the unwound DNA was degrades both strands of the DNA duplex almost full-length. Similar results were obtained with symmetrically. When a properly oriented wBs dsDNA that was linearized with other restriction sequence is present, exonucleolytic degradation of endonucleases or with dsDNA that was 30-end the 30-terminated strand at the entry point ceases, labeled (data not shown). These results show that, while degradation of the 50-terminated strand con- at low magnesium ion concentrations, the nuclease tinues. This alteration in the behavior of the activity of AddAB enzyme is low, so that proces- enzyme results in the production of a dsDNA mol- sing of linear dsDNA under these conditions ecule carrying a long 30-overhang, an intermediate mostly produces fully unwound ssDNA. However, that is universal for the initiation of DNA recombi- at higher magnesium ion concentrations, AddAB nation. Our data also imply that the interaction both unwinds and degrades dsDNA, as evidenced between an exonuclease/helicase and a regulatory by the decrease in production of full-length sequence is an evolutionarily conserved means to ssDNA. This behavior is similar to that of the initiate DNA recombination, and that the AddAB RecBCD enzyme (Dixon & Kowalczykowski, 1993, enzyme de®nes a novel class of enzymes that are 1995; Taylor & Smith, 1995), whereby the free mag- functionally, but not structurally, homologous to nesium ion concentration determines the overall the RecBCD class of enzymes. level of nuclease activity, but has little effect on helicase activity (Roman & Kowalczykowski, 1989b). Results When dsDNA is processed by the RecBCD enzyme, most of the full-length ssDNA produced The AddAB enzyme degrades both strands of 0 dsDNA during unwinding originates from the 5 -terminated strand at the entry point of the enzyme (i.e. ``bottom-strand''; Before investigating the interaction between the see Figure 1). This behavior is due to the intrinsic AddAB enzyme and its putative Chi sequence, we asymmetry of DNA degradation by the RecBCD wanted to examine the processing of DNA that is enzyme: the top-strand of the duplex is degraded devoid of Chi sites (w0). To this end, plasmid more vigorously than the bottom-strand (Dixon & pADGF0 was linearized using ClaI and 50-end Kowalczykowski, 1993). To determine if exonucleo- labeled with [g-32P]ATP. Figure 2 shows a time lytic degradation by the AddAB enzyme is also course of processing of this substrate by the asymmetric, we designed DNA substrates that AddAB enzyme at two different magnesium ion would restrict entry of the AddAB enzyme to only

Figure 2. AddAB enzyme both unwinds and degrades DNA

devoid of wBs. Plasmid pADGF0 was cut by ClaI, 50-end labeled with [g-32P]ATP, and treated with subsaturating amounts of AddAB enzyme (one functional AddAB enzyme per three ends) at two different magnesium concen- trations. Aliquots were withdrawn at 30 seconds and ®ve minutes, and reaction products were sub- jected to electrophoresis through native 1 % agarose gel. The initial double-stranded DNA (dsDNA) and full-length single-stranded DNA (ssDNA) are indicated by an arrow. A control lane correspond- ing to heat denatured dsDNA is also shown. 10 AddAB-Chi Interactions in Vitro one end of the linear dsDNA. For this, we gener- ated dsDNA with 50-ssDNA overhangs (approxi- mately 50 nt long) using exonuclease III, and then treated the DNA with a restriction enzyme to cre- ate a nearly blunt end at one end of the molecule to permit entry of the enzyme. As with RecBCD enzyme, overhangs block the entry of AddAB enzyme to the dsDNA at all magnesium ion con- centrations tested (data not shown and see Figure 5 below). Either the top-strand or the bottom-strand was labeled as described in Materials and Methods, and the action of AddAB enzyme was assayed at different magnesium ion concentrations, since this parameter controls the overall level of nuclease activity. Any strand asymmetry in the degradation process would be manifest as a differ- ential loss of one of the strands during unwinding.

Figure 3 shows that, as the Mg(OAc)2 concen- tration increases, the amount of remaining full- length ssDNA strongly decreases for both top- strand and bottom-strand labeled substrates. More- over, it is apparent that both strands of the duplex Figure 3. The AddAB enzyme degrades both strands are degraded with similar ef®ciencies, with a sharp of dsDNA symmetrically. Plasmid pADGF0, devoid of any w site, was treated so that the AddAB enzyme has transition occurring around 1 mM Mg(OAc)2. Bs Although the bottom-strand is slightly more sensi- access to only one end of the dsDNA (see Materials and Methods). The top-strand (30-terminated at the entry tive to degradation than the top-strand for this par- 0 ticular substrate, we conclude that, unlike the point) and the bottom-strand (5 -terminated at the entry point) were then radiolabeled separately (asterisks), and RecBCD enzyme, which shows a strong strand their respective fates were followed after processing by asymmetry in the same assay (F.C. & S.C.K., the AddAB enzyme. The graph represents the percen- unpublished data), the AddAB enzyme degrades tage of remaining full-length ssDNA relative to the both strands of the duplex nearly symmetrically. amount of dsDNA that was processed by AddAB after three minutes of reaction (typically >90 % of the initial substrate) as a function of magnesium acetate concen- Interaction with wBs results in the production 0 tration. of a 5 -terminated wBs-specific fragment When the RecBCD enzyme encounters a prop- erly oriented w site, the nuclease activity on the top-strand is attenuated, thus allowing for the one 50-end labeled fragment is produced in preservation of ssDNA downstream of w on the response to wN. This fragment migrates as ssDNA top-strand. With plasmid-size DNA substrates, a with an approximate size of 1900 nt. For pADGF1, 0 5 -terminated, w-speci®c, top-strand, downstream a second, wF-speci®c band is produced, in addition fragment is produced (Figure 1). It was recently to the wN-speci®c fragment. The size of this frag- shown that the sequence 50-AGCGG-30, referred to ment, 1700 nt, is expected due to the position of as wBs, allows double-stranded DNA to resist exo- wF and the fact that wF has an opposite orientation nucleolytic degradation by the AddAB enzyme in compared to wN. Although the two wBs-speci®c B. subtilis, thus suggesting that at least the nuclease fragments observed with pADGF1 could not activity of the complex is regulated (CheÂdin et al., be resolved on the native gel, they can clearly be

1998). To investigate the effect of wBs on the proces- distinguished on the alkaline gel. Under the con- sing of dsDNA by the AddAB enzyme, two differ- ditions used here, the yield of wF-speci®c fragments ent plasmids were used. pADG6406-1 (CheÂdin produced relative to the amount of dsDNA pro- et al., 1998) carries a single wBs sequence, desig- cessed was close to 40 %. The yield of wF-speci®c nated as wN. Its derivative, pADGF1, carries two fragments was higher than with wN, in agreement additional tandem wBs sequences (collectively with wF being composed of two tandem wBs sites. designated as wF) present in the opposite orien- When other restriction endonucleases were used to tation (Figure 4(a)). When these two DNA mol- linearize either pADG6406-1 or pADGF1, the size 0 ecules are linearized and labeled at their 5 ends, of the wBs-speci®c fragments shifted accordingly processing by AddAB enzyme produces novel (data not shown).

DNA bands (Figure 4(b)). Native agarose gel elec- Although in vivo data show that the wBs sequence trophoresis, however, does not permit an accurate is active in only one orientation (CheÂdin et al., assessment of the number, size, or single-stranded 1998), the exact polarity of recognition by AddAB nature of these new bands. For this reason, the was not known. Because of this uncertainty and same samples were analyzed by alkaline gel elec- because the AddAB enzyme could enter these trophoresis (Figure 4(c)). For pADG6406-1, only DNA substrates from either end, we could not AddAB-Chi Interactions in Vitro 11

0 Figure 4. Interaction between AddAB enzyme and wBs produces a novel 5 -terminated, wBs-speci®c fragment. (a) Structures of the DNA substrates used. pADG6406-1 and pADGF1 were linearized with ClaI, and 50-end radio- labeled (asterisks). The respective sizes of the fragments upstream and downstream of the wBs sites are indicated in nucleotides. pADG6406-1 carries one wBs site (wN) whereas pADGF1 carries three wBs sites (wN and wF, which is actu- ally composed of two tandem wBs sites). Analysis of the reaction products on a (b) native or (c) alkaline 1 % agarose gel. The major reaction products are indicated by arrows. A heat denatured DNA sample and a radiolabeled 1 kb ladder are also present as control lanes. The sizes of the DNA ladder bands are given in nucleotides.

attribute the observed wBs-speci®c fragments to the incomplete tailing of a small fraction of the portion of DNA that was processed before or after molecules). In contrast, substrates with one entry wBs. However, if the analogy to RecBCD enzyme site (Figure 5(b), middle and right panels) are applies, and because wBs protects linear dsDNA rapidly processed. A wBs-speci®c fragment is seen from degradation by AddAB enzyme in vivo only for the DNA substrate that allows entry of the (CheÂdin et al., 1998), we expect that the observed AddAB enzyme from the right side (SpeI-treated). 0 wBs-speci®c fragments correspond to a DNA frag- In this orientation, the 3 terminated strand at ment located downstream of wBs with respect to the entry point (top-strand) carries the sequence the enzyme's entry point. 50-AGCGG-30. Hence, for recognition to occur, the AddAB enzyme must approach the sequence 0 0 0 The active orientation of w is 50-AGCGG-30 5 -AGCGG-3 from the 3 side. According to the Bs E. coli nomenclature, 50-AGCGG-30 is therefore the

To unambiguously determine in which orien- equivalent of the E. coli w site. The size of the wBs- tation wBs is recognized by the translocating speci®c fragment produced in this experiment was AddAB enzyme, we used the strategy shown in con®rmed using alkaline gel electrophoresis, Figure 3 to limit entry of AddAB to one end of the demonstrating that it corresponds, as expected, to

DNA substrate. Plasmid pADG6406-1 was linear- the top-strand downstream wBs-speci®c fragment ized using ClaI, treated with exonuclease III as (data not shown). Interaction between AddAB 0 described in Materials and Methods, and 5 -end enzyme and its cognate wBs sequence therefore pro- labeled. Figure 5(b) (left panel) shows that a DNA duces a top-strand ssDNA fragment downstream molecule carrying two recessed ends is almost of wBs. completely resistant to processing by the AddAB enzyme (the slight amount of full length ssDNA Interaction with wBs causes degradation to and wBs-speci®c fragment produced at the later attenuate at the last base on the 30 side of w time points probably results from slow Bs 0 degradation of the ssDNA tails by the ssDNA To map precisely the 3 -end of the wBs-speci®c exonuclease activity of AddAB enzyme or from fragment, we produced DNA substrates carrying 12 AddAB-Chi Interactions in Vitro

Figure 5. Blocking access of the AddAB enzyme to DNA with ssDNA tails de®nes the active orientation of wBs. (a) pADG6406-1 was cut with ClaI, treated with exonuclease III to create 50-overhanging tails and radiolabeled at the 50-ends. Further treatment with MslIorSpeI allowed AddAB enzyme to access DNA from one or the other end. The 0 0 0 0 wBs sequence (wN) is shown (5 -AGCGG-3 on the top-strand, 5 -CCGCT-3 on the bottom-strand). (b) Time course of the reaction analyzed on a native 1 % agarose gel. The major products of the reaction are indicated by arrows.

0 wBs about 100 bp from the 5 -end (Figure 6(a)) and ever, the AddAB enzyme generated DNA analyzed the products of a reaction with AddAB fragments whose 30 ends were located in a G-rich using denaturing 8 % polyacrylamide gels. The stretch of DNA (Figure 6(c), FSK1516). Interest- mobility of the various products was then com- ingly, these G-rich stretches contained a wBs-like pared to a sequence ladder. As can be seen from sequence with four matches out of ®ve positions (underlined with broken line), suggesting the Figure 6(b), the w0 control DNA displays a non- speci®c pattern of ssDNA fragments (although the AddAB enzyme retained some ability to recognize banding pattern suggests that the enzyme prefer- wBs variants, as was described for the RecBCD entially cuts DNA before A residues and, to a les- enzyme (Cheng & Smith, 1984, 1987). ser extent, T residues). For the wF-containing DNA, one major band is observed for each wBs site con- A second wBs-specific fragment is observed 0 tained within wF. Each wBs-speci®c fragment termi- with 3 -end labeled substrates nates at the last base to the 30 side of w Bs In principle, four potential w -speci®c fragments (Figure 6(b)). Since magnesium ion concentration Bs can be produced upon interaction of the AddAB strongly in¯uences the location of the 30 extremity enzyme with its cognate Chi sequence. As shown of the w-containing ssDNA fragment produced by in Figure 1, two of these putative fragments have RecBCD enzyme (Taylor & Smith, 1995), we the original 50 terminus (top-strand downstream repeated this experiment over a range of Mg(OAc)2 and bottom-strand upstream), while the remaining concentrations (1 mM to 10 mM) and found that two have the original 30 terminus (top-strand 0 the location of the 3 -end of the wBs-speci®c frag- upstream and bottom-strand downstream). So far, ment was invariant for AddAB (data not shown). we showed that interaction between the AddAB To examine the effect of DNA sequence context enzyme and wBs results in the production of a on the interaction of w with AddAB enzyme, 0 Bs single 5 -terminated, top-strand downstream wBs- other DNA substrates were generated (see speci®c fragment (Figure 4). To investigate whether Materials and Methods; Figure 6(c)). In all cases, additional wBs-speci®c fragments are produced, we processing by AddAB enzyme produced a compared the reaction products obtained upon wBs-speci®c fragment that terminated one nucleo- processing of wBs-containing substrates that were 0 0 0 tide to the 3 -side of wBs. In some instances, how- either 5 -end or 3 -end labeled (Figure 7). As AddAB-Chi Interactions in Vitro 13

expected, a top-strand downstream wBs-speci®c a pause, resulting in a high probability of cleavage fragment is observed when the substrates are at this site, as suggested for the RecBCD enzyme 0 5 -end labeled. However, wBs-speci®c fragments are (Dixon & Kowalczykowski, 1995). also detected with 30-end labeled substrates. With pADG6406-1, one novel fragment is produced. Its Discussion size (1900 nt), as judged from alkaline gel electro- phoresis (data not shown), indicates that this frag- Here, we show that the B. subtilis AddAB ment corresponds to the top-strand upstream, wBs- enzyme interacts with the ®ve-nucleotide sequence, speci®c fragment. With pADGF1, a second novel 0 0 5 -AGCGG-3 , which we de®ne as wBs. We further fragment is detected, in addition to the one demonstrate that this interaction results in the described. Its size (600 nt) is also consistent with regulation of AddAB nuclease activities in a way it corresponding to the top-strand upstream frag- that allows for the formation of a DNA substrate ment. Note, however, that the yield of the shorter suitable for the initiation of homologous recombi- 600 nt fragment is greater than that of the longer nation. 1900 nt fragment. This result is expected if the top- strand is subjected to infrequent non-speci®c clea- B. subtilis AddAB defines a novel class of vages, such that a small fragment has a high prob- enzymes that are functional, yet distinct, ability of escaping cleavage, while a longer homologs of RecBCD enzymes fragment is more likely to be cleaved (see Discus- sion). We therefore conclude that, under the con- Evidence gathered from the complete genome ditions used here, the AddAB enzyme does not sequences of various organisms indicates that the fully degrade the DNA upstream of wBs and that E. coli RecBCD enzyme is the prototypic member interaction with this site results in a high prob- of a large family of enzymes that are found ability of cleavage on the top-strand at wBs. throughout the eubacteria. Close homologs of RecBCD enzyme are found in genera such as Spir- ochaetales, Green sulfur bacteria, Chlamydiales, The yield of the upstream wBs-specific fragment depends strongly upon the level of and Proteobacteria. Although Chi recognition by AddAB enzyme nuclease activity RecBCD-type enzymes has only been demon- strated for E. coli and H. in¯uenzae, it is reasonable To further characterize the conditions under to assume that this interaction is conserved in 0 which this 3 -terminated, upstream wBs-speci®c these other organisms. Our results show that the fragment is produced, we examined the effect of AddAB enzyme from B. subtilis is a functional magnesium ion concentration on the yield of this homolog of the RecBCD enzyme. However, the fragment relative to the 50-terminated downstream AddAB enzyme possesses only two subunits and, wBs-speci®c fragment. Figure 8 shows that the while the AddA subunit shows homology to the downstream wBs-speci®c fragment is detected with RecB subunit, the AddB subunit has only limited a similar yield over a wide range of magnesium homology to the RecC subunit and no homology ion concentrations (0.5 to 15 mM, Figure 8 and to the RecD subunit. It therefore appears that data not shown). In contrast, the upstream wBs- AddAB is not strictly homologous to RecBCD. speci®c fragment behaves similarly to the full Rather, we propose that AddAB de®nes a novel length ssDNA and is no longer detected when the class of enzymes. magnesium concentration is raised above 2-3 mM Our proposal follows from the fact that homo- (Figure 8 and data not shown). From these results, logs of AddAB have been discovered in several we conclude that under conditions which promote other bacteria. The RexAB enzyme of L. lactis is a higher nuclease activity, the AddAB enzyme vigor- two subunit complex showing extensive sequence ously degrades the top-strand before encountering similarity to AddAB. This enzyme was shown a correctly oriented wBs site, after which its nuclease genetically to interact with a speci®c seven-nucleo- activity on the top-strand is turned off, thus produ- tide sequence (Biswas et al., 1995; el Karoui et al., cing a ssDNA fragment downstream of wBs. How- 1998). Furthermore, an analysis of all publicly ever, at lower nuclease conditions, the probability available eubacterial permits identi®- of a non-speci®c cleavage before wBs is low, leaving cation of AddAB homologs in Clostridium acetobu- the top-strand upstream of wBs intact. In agreement, tylicum, Enterococcus faecalis, Staphylococcus aureus, we observed that the yield of the top-strand Streptococcus pyogenes, Streptococcus mutans and upstream fragment was inversely proportional to Streptococcus pneumoniae. All of those organisms the length of DNA before wBs suggesting that long- belong to the Bacillus/Clostridium subdivision of er have an increased probability of receiving the Gram-positive bacteria, which therefore seems at least an endonucleolytic cut even under low to be a specialized ``niche'' where the AddAB-class nuclease conditions (Figures 7, 8 and data not of enzymes evolved. shown). Similar magnesium ion concentration- dependent appearance of the top-strand upstream Processing of DNA substrates devoid of w w-speci®c fragment was observed for the RecBCD Bs enzyme (Dixon & Kowalczykowski, 1995). The We show here that the AddAB enzyme is both a interaction with wBs is presumably associated with powerful DNA helicase, able to unwind a linear 14 AddAB-Chi Interactions in Vitro

Figure 6 (legend opposite)

double-stranded DNA fragment of at least three lent cation. The intensity of the nuclease activity is kilobases, and a powerful exonuclease. Both activi- strongly dependent on the concentration of mag- ties require the hydrolysis of ATP as a source of nesium ion (Figures 2, 3 and 8). This dependence energy and the presence of magnesium as a diva- on the magnesium ion concentration was AddAB-Chi Interactions in Vitro 15 also observed for the RecBCD enzyme in E. coli regulatory DNA sequence has been evolutionarily (Dixon & Kowalczykowski, 1995). However, in conserved. The outcome of this interaction depends contrast to the RecBCD enzyme, AddAB degrades on the initial strength of the nuclease activity. both strands of the duplex symmetrically even Under low nuclease conditions, interaction with under conditions where the RecBCD enzyme wBs results in the production of two wBs-speci®c shows a marked preference for degrading the 30- fragments derived from the cleavage of the top- terminated strand at the entry point (Figure 2; strand at wBs (Figures 7 and 8). However, as the Dixon & Kowalczykowski, 1993; F.C & S.C.K., strength of the nucleolytic degradation is unpublished observations). increased, the top-strand upstream fragment Recently, a domain required for all of RecBCD rapidly disappears while the top-strand down- nucleolytic activities was identi®ed (Yu et al., stream fragment remains protected (Figure 8). This 1998b). This domain, present at the C terminus of indicates that interaction results in the attenuation the RecB subunit, is conserved at a similar position of the exonuclease activity on the top-strand, a in the AddA subunit (Haijema et al., 1996b). Inter- estingly, we were able to identify a second nucle- behavior which parallels that of the RecBCD ase domain in the C-terminal portion of the AddB enzyme (Dixon & Kowalczykowski, 1993). Under subunit by the virtue of its strong homology to the all conditions tested, no discrete fragment was previously de®ned RecB nuclease domain and its recovered from the bottom-strand, which indicates counterpart in AddA. This domain is highly con- that the enzyme resumes unwinding after wBs and served among members of the AddAB family of that the 50 to 30 degradation of that strand is unaf- enzymes (data not shown). A similar search, how- fected by wBs. Hence, interaction with wBs trans- ever, failed to reveal any additional nuclease forms the AddAB enzyme from a helicase which domain in the RecB, RecC or RecD subunits of the degrades both DNA strands to one with only 50 to E. coli enzyme. The AddAB enzyme therefore car- 30 exonuclease. ries two putative nuclease domains, while RecBCD Mapping of the wBs-speci®c nuclease-attenuation carries only one; this distinction could explain the sites to nucleotide resolution revealed that the ability of the AddAB enzyme to degrade both AddAB enzyme predominantly stops degrading strands of dsDNA, where RecBCD enzyme 0 DNA one nucleotide to the 3 side of wBs (Figure 6). degrades only one or the other. Varying the sequence context in which wBs is located did not lead to a signi®cant modi®cation of

Interaction of AddAB enzyme with wBs the distribution of the cleavage sites (Figure 6(c)) and the pattern did not change over a wide range From in vivo experiments, it was proposed that of conditions. This behavior is in contrast to that of AddAB was able to interact with a speci®c ®ve- the RecBCD enzyme, which shows an array of nucleotide sequence, 50-AGCGG-30, its comp- attenuation sites distributed from four to six lement, or both, referred to as wBs (CheÂdin et al., 1998). Here, using puri®ed components, we nucleotides before w at low nuclease conditions, to demonstrated that AddAB is indeed able to inter- four nucleotides inside w at high nuclease con- ditions (Taylor & Smith, 1995). This difference may act with, and to be regulated by, wBs. Furthermore, suggest that the recognition of wBs by AddAB we de®ned the active orientation of wBs as the sequence 50-AGCGG-30, which must be approached enzyme results in a longer pause than for the by an AddAB enzyme travelling from the 30 side RecBCD enzyme, producing a more speci®c clea- for recognition to occur (Figure 5). This observation vage event, or alternatively, that the exonuclease represents the ®rst biochemical analysis of a sys- domain of AddAB is more tightly tethered than tem other than the RecBCD system of E. coli, and the equivalent ``swing'' domain of the RecBCD provides direct support to the notion that the inter- enzyme (Yu et al., 1998a). In some instances, how- action between a DNA helicase/nuclease and a ever, strong attenuation sites were detected in wBs-

Figure 6. Determination of the precise size of the wBs-speci®c fragment. (a) The substrates used in the reactions in (b) corresponded to 650 bp PCR fragments with or without the wF locus. The respective sizes of the fragments upstream and downstream of wF are indicated in nucleotides. The arrow above wF indicates the direction from which the AddAB enzyme must approach to recognize wBs based on previous results (see Figure 5). As described in Materials and Methods, only the 50-end of the top-strand was labeled (asterisk). (b) Analysis of the reaction products on an 8 % sequencing gel. The sequence ladders were generated using pADGF1 (containing wF, left) or pADG6406-1 (devoid of wBs site, right) as templates, and the FSK7 oligonucleotide as a primer (small arrow). Lanes indicate wF and w0 correspond to the reaction products obtained after the AddAB enzyme processed the DNA substrates. Two timepoints (30 seconds and ®ve minutes) are shown. The location of the fragments can be deduced by comparison with the sequence ladders. Two regions of the gel have been expanded. The location of the cleavage sites (arrows) as 0 well as the position of the two wBs sites are indicated. (c) Mapping of the 3 -ends of various wBs-speci®c fragments. The position of the 30-ends were deduced by comparison with a sequence ladder; major fragments (large arrows); minor fragments (small arrows). The numbers refer to the primers used to generate the fragments (see Materials and

Methods). The Bacillus subtilis Chi site (wBs) is underlined. Two wBs-like sites in FSK 1516 are marked by a broken line, the asterisk indicates the position that is mutated compared to wBs. 16 AddAB-Chi Interactions in Vitro

Figure 7. A30-terminated,

upstream wBs-speci®c fragment is also produced by AddAB proces- sing. (a) pADG6406-1 or pADGF1 were linearized by BamHI and either 50-end or 30-end labeled as described in Materials and Methods. The size of the fragments upstream and downstream of the

different wBs sites are indicated in nucleotides. Arrows above the w sites show the direction from which the AddAB enzyme must approach for recognition to occur. (b) Analysis of the reaction pro- ducts on a native 1 % agarose gel. The AddAB enzyme was added to the different DNA substrates for the indicated time before aliquots were withdrawn. The major reac- tion products are indicated by arrows. The left part of the gel cor- responds to 50-end labeled sub- strates, the right part, to 30-end labeled substrates. pADG refers to pADG6406-1. M corresponds to a radiolabeled 1 kb DNA ladder.

like sequences, especially if those were embedded types of ssDNA substrates in RecA protein-pro- in stretches of G residues (Figure 6(c)). moted reactions showed that 30 ends are more The mechanism by which Chi sequence recog- invasive than 50 ends (Konforti & Davis, 1987, nition is transformed into a change of the enzyme's 1990). Similarly, physical analysis of recombination activity remains elusive. For the RecBCD enzyme, intermediates showed that dsDNA with a protrud- the changes resulting from the interaction with w ing 30-terminated ssDNA is produced during the are thought to occur through the modi®cation of initiation phase of recombination both in Bacteria the RecD subunit. One model hypothesizes that and Eucarya (Anderson & Kowalczykowski, 1997a; the RecD subunit is ejected from the complex, a Friedman-Ohana & Cohen, 1998; Sun et al., 1991; proposal which is consistent with various in vivo White & Haber, 1990). These DNA intermediates, observations (Koppen et al., 1995; Myers et al., therefore, are universally used to initiate homolo- 1995; Stahl et al., 1990; Thaler et al., 1988), whereas gous DNA recombination. In all cases, the proces- another simply proposes that the RecD subunit is sing of dsDNA is followed by the binding of a modi®ed in an as yet unknown way (Anderson DNA strand exchange protein to the ssDNA inter- et al., 1997; Churchill et al., 1999; Dixon et al., 1994). mediate. Interestingly, the translocating RecBCD For the AddAB enzyme, however, physical ejection enzyme also functions to load the RecA protein is impossible since the complex comprises two onto the downstream w-containing ssDNA subunits, both of which are strictly essential to the (Anderson & Kowalczykowski, 1997b). Whether enzyme's function (Kooistra et al., 1993). Therefore, this level of coordination between the AddAB for AddAB, it is more likely that the interaction enzyme and the RecA protein is observed in with wBs results in an as yet unde®ned confor- B. subtilis is now under investigation. mational change. Materials and Methods Regulation by wBs allows the production of a universal recombination intermediate Purification of AddAB As discussed above, interaction of the AddAB Plasmid pWSK2978 (gift from J. Kooistra; Kooistra enzyme with wBs transforms AddAB from a 0 0 et al., 1993), carrying both the addA and addB genes double-stranded exonuclease to a 5 to 3 exonu- under the control of the natural promoter for the operon, clease. The net result of this modi®cation is, there- was introduced into E. coli strain AB1157 recB::Tn10 (gift fore, the production of dsDNA with a 30-ssDNA from B. Michel; Uzest et al., 1995). The protocol for puri- tail. Studies of the relative invasiveness of various ®cation was based on a previously published protocol AddAB-Chi Interactions in Vitro 17

Figure 8. Differential sensitivity of the upstream and downstream top-strand wBs-speci®c fragments to conditions that affect the AddAB nuclease activity. (a) Structure of the substrates used. pADGF1 was cut with BamHI and either 50-end (left panel) or 30-end labeled (right panel). Reactions were performed at varying magnesium ion concentrations and aliquots were withdrawn three minutes after addition of the enzyme. Reaction products were analyzed after elec- trophoresis through native 1 % agarose gels, as depicted in (b). The major reaction products are indicated by arrows.

for puri®cation of the RecBCD enzyme (Roman & After mixing, the tubes were placed on ice, and spun at Kowalczykowski, 1989a; Eggleston & Kowalczykowski, 12,000 rpm for 20 minutes; 500 ml of the supernatant was 1993). The presence of AddAB enzyme was followed added to 5 ml of scintillation cocktail, and then counted using ATP-dependent exonuclease activity, as described using a Beckman scintillation counter. below. AddAB enzyme concentration was determined using a molar extinction coef®cient of 2.7 Â 105 M1 cm1 at 280 nm (derived from Mach et al., 1992) and a DNA substrates molecular mass of 275,531 Da as determined from its Plasmid pADG6406-1 was described previously nucleotide sequence. The A280/A260 ratio was 1.35. The preparation's purity appeared to be greater than 90 % as (CheÂdin et al., 1998). This 3 kb plasmid carries a wBs judged from Coomassie-stained polyacrylamide gels. sequence at position 1805. Plasmid pADGF1 is a deriva- Two other bands could be detected and probably corre- tive of pADG6406-1 containing two additional wBs spond to proteolytic degradation fragments of AddB. sequences cloned in tandem, in the opposite orientation, Because of this, the subunit ratio was found to be at the unique EcoRI site (the wBs-containing linker cloned slightly in favor of AddA. here was described in Figure 2 of CheÂdin et al., 1998). Plasmid pADGF0 is a derivative of pADG6406-1 in which two targeted silent mutations were introduced in 0 0 0 ATP-dependent exonuclease assays the natural wBs sequence (from 5 -AGCGG-3 to 5 - AATGG-30), thus inactivating the site (this study and The speci®c activity of the enzyme in nuclease units CheÂdin et al., 1998). All plasmid DNA was puri®ed by was determined as described by Eichler & Lehman two rounds of cesium chloride density-gradient ultracen- (1977) using sonicated 3H-labeled E. coli chromosomal trifugation (Sambrook et al., 1989). The molar concen- DNA. The reaction mixture (300 ml) consisted of 100 mM tration of the dsDNA in nucleotides was determined glycine-NaOH (pH 9.2), 50 mM MgCl2, 0.1 mg/ml BSA, spectrophotometrically at 260 nm using an extinction 0.67 mM DTT, 20 mM[3H]DNA and 75 mM ATP, when coef®cient of 6290 M1 cm1 at 260 nm. Plasmid DNA present. Incubation was for 20 minutes at 37 C, and the was cut with various restriction enzymes as rec- reaction was stopped by addition of 15 ml of sonicated ommended by the suppliers and radiolabeled at either calf thymus DNA (2.5 mg/ml) and 300 ml of 15 % TCA. the 30-end using the Klenow fragment of DNA polymer- 18 AddAB-Chi Interactions in Vitro ase I and appropriate [a-32P]dNTPs or the 50-end by NaOH, 1 mM EDTA). After electrophoresis, gels were sequential action of shrimp alkaline phosphatase and neutralized for one hour in 0.5 M Tris-HCl (pH 7.5), phage T4 polynucleotide kinase in the presence of dried and analyzed as described above. [g-32P]ATP. Labeled DNA was further puri®ed from unincorporated radionucleotides by passage through a S-200 MicroSpin column (Pharmacia). Tailed substrates Analysis of reaction products on sequencing gels carrying 50 single-stranded overhangs were generated using exonuclease III (1 unit/mg DNA, Promega) for one Aliquots (25 ml) were taken at the indicated timepoints minute at room temperature as described (Anderson & and added to a mixture of phenol and chloroform (24:1) Kowalczykowski, 1997a). The 50-end labeling of these to stop the reaction and deproteinize DNA. The aqueous phase was withdrawn (10 ml), added to 4 mlof4Â load- resected substrates was achieved using T4 polynucleo-  tide kinase as described above. 30-End labeling was per- ing buffer containing formamide, heated to 95 C for formed using Klenow polymerase and only three of the three minutes, and loaded onto an 8 % sequencing gel. four deoxynucleotides, including [a-32P]dATP. Unincor- The sequence ladder was generated using the Sequenase porated radionucleotides were removed as described kit II from USB as recommended by the supplier. Elec- above. trophoresis was performed at 2000 V for four to ®ve hours and the gel was analyzed directly using a Storm PCR fragments were generated using plasmids 840 PhosphorImager and the Image-QuaNT software. pADG6406-1 or pADGF1 as templates and oligonucleo- tides FSK7 (50-AGCATTTTAACATTATACTTTG-30) and FSK8 (50-GGGGGATCCTCTAGCTAGCA-30) as primers, or plasmid pBR322 and oligonucleotides FSK9 (50- TTTGGATCCACCGATTTAGCTTTATGC-30) and FSK10 (50-ATATTCATTGAATCCCCCTC-30), FSK11 (50-TGCC Acknowledgments GCAAAAAAGGGA ATAAG-30) and FSK12 (50-ACA We are grateful to the many members of the GGGATCCGAGAACGAGTGCGC-30), FSK15 (50-GTGC Kowalczykowski laboratory for critical reading of the GGATCCTCGTGCTCCTGTCGT-30) and FSK16 (50-AAG manuscript: Dan Anderson, Rick Ando, Deana Arnold, CTCATCAGC GTGGTCGT-30) as primers. Reactions Piero Bianco, Carole Bornarth, Frank Harmon, Julie were performed in 100 ml volume using the Pfu DNA Kleiman, Alex Mazin, Jim New, Erica Seitz and Susan polymerase (Stratagene). The resulting fragments were Shetterley. This work was supported by funds from the puri®ed by agarose gel electrophoresis, 50-end labeled as National Institutes of Health grant GM-41347. described above, treated with BamHI, which recognizes a unique site in the wBs-distal primer (bold letters), and passed through a S-200 microspin column. The 1 kb DNA ladder was purchased from Gibco-BRL and labeled References using T4 polynucleotide kinase as described. Alonso, J. C., Tailor, R. H. & Luder, G. (1988). Charac- terization of recombination-de®cient mutants of Reaction conditions and electrophoresis Bacillus subtilis. J. Bacteriol. 170, 3001-3007. Anderson, D. G. & Kowalczykowski, S. C. (1997a). The Standard reaction conditions corresponded to 25 mM recombination hot spot w is a regulatory element Tris-acetate (pH 7.5), 2 mM magnesium acetate, 1 mM that switches the polarity of DNA degradation by ATP, 1 mM dithiothreitol, 10 mM nucleotides DNA the RecBCD enzyme. Genes Dev. 11, 571-581. (equivalent to 1.6 nM DNA molecules), 2 mM E. coli SSB Anderson, D. G. & Kowalczykowski, S. C. (1997b). The protein, and 0.5 nM functional AddAB enzyme (the translocating RecBCD enzyme stimulates recombi- amount of functional enzyme was determined from a nation by directing RecA protein onto ssDNA in a titration of enzyme helicase activity against a ®xed con- w-regulated manner. Cell, 90, 77-86. centration of DNA (Roman & Kowalczykowski, 1989b); Anderson, D. G., Churchill, J. J. & Kowalczykowski, S. C. the preparation used here was determined to be 50 % (1997). Chi-activated RecBCD enzyme possesses active). In certain instances, the magnesium acetate con- 50 ! 30 nucleolytic activity, but RecBC enzyme does centration was varied and is indicated accordingly. not: evidence suggesting that the alteration induced Assays were performed at 37 C and were initiated by by Chi is not simply ejection of the RecD subunit. addition of the AddAB enzyme after a two-minute incu- Genes Cells, 2, 117-128. bation of all other components at 37 C. Aliquots of the Baudat, F. & Nicolas, A. (1997). Clustering of meiotic reaction mixture (20-25 ml) were withdrawn at the indi- double-strand breaks on yeast chromosome III. cated times and added to an equal volume of stop buffer Proc. Natl Acad. Sci. USA, 94, 5213-8. (0.1 M EDTA, 2.5 % (w/v) SDS, 10 % (v/v) Ficoll, Biswas, I., Maguin, E., Ehrlich, S. D. & Gruss, A. (1995). 0.125 % (w/v) bromophenol blue, and 0.125 % (v/v) A seven-base-pair sequence protects DNA from xylene cyanol). Samples were subjected to electrophor- exonucleolytic degradation in Lactococcus lactis. Proc. esis in 1 % native agarose gels for approximately Natl Acad. Sci. USA, 92, 2244-2248. 15 hours at 1.4 V/cm in TAE buffer. Gels were dried on CheÂdin, F., Noirot, P., Biaudet, V. & Ehrlich, S. D. DEAE paper and analyzed on a Molecular Dynamics (1998). A ®ve-nucleotide sequence protects DNA Storm 840 PhosphorImager using the Image-QuaNT soft- from exonucleolytic degradation by AddAB, the ware. Samples which were subjected to denaturing alka- RecBCD analogue of Bacillus subtilis. Mol. Microbiol. line agarose gel electrophoresis were stopped as 29, 1369-1377. described above, and 10 ml of alkaline loading buffer Cheng, K. C. & Smith, G. R. (1984). Recombinational (300 mM NAOH, 6 mM EDTA, 18 % Ficoll, 0.15 % (w/v) hotspot activity of Chi-like sequences. J. Mol. Biol. bromocresol green, 0.25 % xylene cyanol) was added. 180, 371-377. The samples were mixed and subjected to electrophor- Cheng, K. C. & Smith, G. R. (1987). Cutting of Chi-like esis through 1 % (w/v) alkaline agarose gels for 20 hours sequences by the RecBCD enzyme of E. coli. J. Mol. at 1.4 V/cm in alkaline electrophoresis buffer (50 mM Biol. 194, 747-750. AddAB-Chi Interactions in Vitro 19

Churchill, J. J., Anderson, D. G. & Kowalczykowski, S. C. Koppen, A., Krobitsch, S., Thoms, B. & Wackernagel, W. (1999). The RecBC enzyme loads RecA protein onto (1995). Interaction with the recombination hot spot ssDNA asymmetrically and independently of chi, w in vivo converts the RecBCD enzyme of Escherichia resulting in constitutive recombination activation. coli into a w-independent recombinase by inacti- Genes Dev. 13, 901-911. vation of the RecD subunit. Proc. Natl Acad. Sci. Dernburg, A. F., McDonald, K., Moulder, G., Barstead, USA, 92, 6249-6253. R., Dresser, M. & Villeneuve, A. M. (1998). Meiotic Kowalczykowski, S. C., Dixon, D. A., Eggleston, A. K., recombination in C. elegans initiates by a conserved Lauder, S. D. & Rehrauer, W. M. (1994). Biochemis- mechanism and is dispensable for homologous try of homologous recombination in Escherichia coli. chromosome synapsis. Cell, 94, 387-398. Microbiol. Rev. 58, 401-465. Dixon, D. A. & Kowalczykowski, S. C. (1991). Homolo- Lam, S. T., Stahl, M. M., McMilin, K. D. & Stahl, F. W. gous pairing in vitro stimulated by the recombina- (1974). Rec-mediated recombinational hot spot tion hotspot, Chi. Cell, 66, 361-371. activity in bacteriophage lambda. II. A mutation Dixon, D. A. & Kowalczykowski, S. C. (1993). The which causes hot spot activity. Genetics, 77, 425-433. recombination hotspot w is a regulatory sequence Mach, H., Middaugh, C. R. & Lewis, R. V. (1992). Stat- that acts by attenuating the nuclease activity of the istical determination of the average values of the E. coli RecBCD enzyme. Cell, 73, 87-96. extinction coef®cients of tryptophan and tyrosine in Dixon, D. A. & Kowalczykowski, S. C. (1995). Role of native proteins. Anal. Biochem. 1, 74-80. the Escherichia coli recombination hotspot, w,in McKim, K. S. & Hayashi-Hagihara, A. (1998). mei-W68 RecABCD-dependent homologous pairing. J. Biol. in Drosophila melanogaster encodes a Spo11 homolog: Chem. 270, 16360-16370. evidence that the mechanism for initiating meiotic Dixon, D. A., Churchill, J. J. & Kowalczykowski, S. C. recombination is conserved. Genes Dev. 12, 2932- (1994). Reversible inactivation of the Escherichia coli 2942. RecBCD enzyme by the recombination hotspot w McKittrick, N. H. & Smith, G. R. (1989). Activation of in vitro: evidence for functional inactivation or loss Chi recombinational hotspots by RecBCD-like of the RecD subunit. Proc. Natl Acad. Sci. USA, 91, enzymes from enteric bacteria. J. Mol. Biol. 210, 485- 2980-2984. 495. Eggleston, A. K. & Kowalczykowski, S. C. (1993). Bio- Michel, B., Ehrlich, S. D. & Uzest, M. (1997). DNA chemical characterization of a mutant recBCD 2109 double-strand breaks caused by replication arrest. enzyme, the recB CD enzyme, which lacks w- EMBO J. 16, 430-438. speci®c, but not non-speci®c, nuclease activity. Myers, R. S., Kuzminov, A. & Stahl, F. W. (1995). The J. Mol. Biol. 231, 605-620. recombination hot spot w activates RecBCD recom- Eichler, D. C. & Lehman, I. R. (1977). On the role of bination by converting Escherichia coli to a recD ATP in phosphodiester bond hydrolysis catalyzed mutant phenocopy. Proc. Natl Acad. Sci. USA, 92, by the recBC deoxyribonuclease of Escherichia coli. 6244-6248. J. Biol. Chem. 252, 499-503. Roman, L. J. & Kowalczykowski, S. C. (1989a). Charac- el Karoui, M., Ehrlich, D. & Gruss, A. (1998). Identi®- terization of the adenosinetriphosphatase activity of cation of the lactococcal exonuclease/recombinase the Escherichia coli RecBCD enzyme: relationship of and its modulation by the putative Chi sequence. Proc. Natl Acad. Sci. USA, 95, 626-631. ATP hydrolysis to the unwinding of duplex DNA. Friedman-Ohana, R. & Cohen, A. (1998). Heteroduplex Biochemistry, 28, 2873-2881. joint formation in Escherichia coli recombination is Roman, L. J. & Kowalczykowski, S. C. (1989b). Charac- initiated by pairing of a 30-ending strand. Proc. Natl terization of the helicase activity of the Escherichia Acad. Sci. USA, 95, 6909-6914. coli RecBCD enzyme using a novel helicase assay. Haijema, B. J., Meima, R., Kooistra, J. & Venema, G. Biochemistry, 28, 2863-2873. (1996a). Effects of lysine-to-glycine mutations in the Sambrook, J., Fritsch, E. F. & Maniatis, T. (1989). Molecu- ATP-binding consensus sequences in the AddA and lar Cloning: A Laboratory Manual, 2nd edit., Cold AddB subunits on the Bacillus subtilis AddAB Spring Harbor Laboratory Press, Cold Spring enzyme activities. J. Bacteriol. 178, 5130-5137. Harbor, NY. Haijema, B. J., Venema, G. & Kooistra, J. (1996b). The C Smith, G. R. (1988). Homologous recombination in pro- terminus of the AddA subunit of the Bacillus subtilis caryotes. Microbiol. Rev. 52, 1-28. ATP-dependent DNase is required for the ATP- Smith, G. R., Kunes, S. M., Schultz, D. W., Taylor, A. & dependent exonuclease activity but not for the heli- Triman, K. L. (1981). Structure of Chi hotspots of case activity. J. Bacteriol. 178, 5086-5091. generalized recombination. Cell, 24, 429-436. Konforti, B. B. & Davis, R. W. (1987). 30 homologous Sourice, S., Biaudet, V., El Karoui, M., Ehrlich, S. D. & free ends are required for stable joint molecule for- Gruss, A. (1998). Identi®cation of the Chi site of mation by the RecA and single-stranded binding Haemophilus in¯uenzae as several sequences related proteins of Escherichia coli. Proc. Natl Acad. Sci. USA, to the Escherichia coli Chi site. Mol. Microbiol. 27, 84, 690-694. 1021-1029. Konforti, B. B. & Davis, R. W. (1990). The preference for Stahl, F. W., Crasemann, J. M. & Stahl, M. M. (1975). a30 homologous end is intrinsic to RecA-promoted Rec-mediated recombinational hot spot activity in strand exchange. J. Biol. Chem. 265, 6916-6920. bacteriophage lambda. III. Chi mutations are site- Kooistra, J., Vosman, B. & Venema, G. (1988). Cloning mutations stimulating rec-mediated recombination. and characterization of a Bacillus subtilis transcrip- J. Mol. Biol. 94, 203-212. tion unit involved in ATP-dependent DNase syn- Stahl, F. W., Thomason, L. C., Siddiqi, I. & Stahl, M. M. thesis. J. Bacteriol. 170, 4791-4797. (1990). Further tests of a recombination model in Kooistra, J., Haijema, B. J. & Venema, G. (1993). The which Chi removes the RecD subunit from the Bacillus subtilis addAB genes are fully functional in RecBCD enzyme of Escherichia coli. Genetics, 126, Escherichia coli. Mol. Microbiol. 7, 915-923. 519-533. 20 AddAB-Chi Interactions in Vitro

Sun, H., Treco, D., Schultes, N. P. & Szostak, J. W. DNA Damage Processing (Friedberg, E. & Hanawalt, (1989). Double-strand breaks at an initiation site for P., eds), pp. 413-422, Alan R. Liss, Inc., New York. meiotic gene conversion. Nature, 338, 87-90. Uzest, M., Ehrlich, S. D. & Michel, B. (1995). Lethality of Sun, H., Treco, D. & Szostak, J. W. (1991). Extensive 30- rep recB and rep recC double mutants of Escherichia overhanging, single-stranded DNA associated with coli. Mol. Microbiol. 17, 1177-1188. the meiosis-speci®c double-strand breaks at the White, C. I. & Haber, J. E. (1990). Intermediates of ARG4 recombination initiation site. Cell, 64, 1155- recombination during mating type switching in Saccharomyces cerevisiae. EMBO J. 9, 663-673. 1161. Yu, M., Souaya, J. & Julin, D. A. (1998a). The 30-kDa Taylor, A. F. & Smith, G. R. (1995). Strand speci®city of C-terminal domain of the RecB protein is critical for nicking of DNA at Chi sites by RecBCD enzyme. the nuclease activity, but not the helicase activity, Modulation by ATP and magnesium levels. J. Biol. of the RecBCD enzyme from Escherichia coli. Proc. Chem. 270, 24459-24467. Natl Acad. Sci. USA, 95, 981-986. Thaler, D. S., Sampson, E., Siddiqi, I., Rosenberg, S. M., Yu, M., Souaya, J. & Julin, D. A. (1998b). Identi®cation Stahl, F. W. & Stahl, M. (1988). A hypothesis: Chi- of the nuclease active site in the multifunctional activation of recBCD enzyme involves removal of RecBCD enzyme by creation of a chimeric enzyme. the recD subunit. In Mechanisms and Consequences of J. Mol. Biol. 283, 797-808.

Edited by M. Gottesman

(Received 20 September 1999; received in revised form 12 January 2000; accepted 20 January 2000)