Analysis of substrate specificity of the RuvC holliday junction resolvase with synthetic Holliday junctions.

The Escherichia coli RuvC protein endonucleolytically resolves Holliday junctions, which are formed as intermediates during genetic recombination and recombination repair. Previous studies using model Holliday junctions suggested that a certain size of central core of homology and a specific sequence in the junction were required for efficient cleavage by RuvC, although not for binding. To determine the minimum length of sequence homology required for RuvC cleavage, we made a series of synthetic Holliday junctions with various lengths of homologous sequence in the core region. It was demonstrated that a monomobile junction possessing only 2 base pairs of the homology core was efficiently cleaved by RuvC. To study the sequence specificity for cleavage, we made 16 bimobile junctions, which differed only in the homologous core sequence. Among them, 6 junctions were efficiently cleaved. Cleavage occurred by introduction of nicks symmetrically at the 3'-side of thymine in all cases. However, the nucleotide bases at the 3'-side of the thymines were not always identical between the two strands nicked. These results suggest that RuvC recognizes mainly topological symmetry of the Holliday junction but not the sequence symmetry per se, that the thymine residue at the cleavage site plays an important role for RuvC-mediated resolution, and that a long homologous core sequence is not essential for cleavage.

The Escherichia coli RuvC protein endonucleolytically resolves Holliday junctions, which are formed as intermediates during genetic recombination and recombination repair. Previous studies using model Holliday junctions suggested that a certain size of central core of homology and a specific sequence in the junction were required for efficient cleavage by RuvC, although not for binding. To determine the minimum length of sequence homology required for RuvC cleavage, we made a series of synthetic Holliday junctions with various lengths of homologous sequence in the core region. It was demonstrated that a monomobile junction possessing only 2 base pairs of the homology core was efficiently cleaved by RuvC. To study the sequence specificity for cleavage, we made 16 bimobile junctions, which differed only in the homologous core sequence. Among them, 6 junctions were efficiently cleaved. Cleavage occurred by introduction of nicks symmetrically at the 3-side of thymine in all cases. However, the nucleotide bases at the 3-side of the thymines were not always identical between the two strands nicked. These results suggest that RuvC recognizes mainly topological symmetry of the Holliday junction but not the sequence symmetry per se, that the thymine residue at the cleavage site plays an important role for RuvC-mediated resolution, and that a long homologous core sequence is not essential for cleavage.
The Holliday junction (HJ) 1 is an important intermediate in the proposed pathway for genetic recombination and recombination repair (Holliday, 1964). In it, two homologous DNA molecules are joined at the single strand crossover by a fourarmed DNA structure. Although there have been reports about putative Holliday junction cleavage activities in yeast and mammals , the Escherichia coli RuvC protein is the first cellular enzyme that has been highly purified and shown to resolve HJs in vitro (Dunderdale et al., 1991;Iwasaki et al., 1991), and its three-dimensional structure has been determined by x-ray crystallography (Ariyoshi et al., 1994). Studies using HJ analogs have demonstrated that RuvC-mediated resolution occurs by the introduction of symmetrical nicks into two of the four DNA strands around the junction, resulting in two nicked duplex DNAs (Dunderdale et al., 1991;Iwasaki et al., 1991;Bennett et al., 1993;Takahagi et al., 1994). Dimerization is essential for activity, with subunits being related by a dyad axis. The two catalytic centers of endonuclease activity in the RuvC dimer are symmetrically located at the bottom of the DNA binding cleft, consistent with the relative position of the nicks observed biochemically (Ariyoshi et al., 1994;Saito et al., 1995). Although recognition of HJs by RuvC is structure-specific, the cleavage reaction additionally requires a divalent cation such as Mg 2ϩ and a homologous sequence core with a certain length, such as 12 bp in two symmetrically related strands of the HJs (Bennett et al., 1993;Takahagi et al., 1994). Immobile synthetic HJs that had no homology core sequences were not resolved by RuvC (Dunderdale et al., 1991;Takahagi et al., 1994). Shah et al. (1994) have shown that RuvC has a sequence preference for cleavage at 5Ј-A/TTT2G/C-3Ј around the crossover point, as determined in a HJ with crossover points that could branch migrate along a several hundred-base pair-long sequence.
The physical structure of HJs is also a question of great interest. With small synthetic HJs as model substrates, it was shown that the HJ forms a "stacked X structure" in which the arms of the junction are antiparallel (Duckett et al., 1988;Hagerman, 1987, 1989;Murchie et al., 1989;von Kitzing et al., 1990). The stacked X structure exhibits 2-fold symmetry, with two strands approximating B-form DNA, whereas the complementary strands are sharply bent as they pass from one helix to the other (Duckett et al., 1988;Churchill et al., 1988;Cooper and Hagerman, 1987;von Kitzing et al., 1990). The sequence of the crossover point determines which strand is bent (Duckett et al., 1988). However, the HJs used in these studies (defined here as immobile HJs) were not substrates for RuvC, because they did not contain a homologous core sequence, which allows spontaneous branch migration of the junctions. Therefore, a question remains whether the structures of the immobile junctions are the same as the catalytically competent substrates. On the other hands, synthetic HJs that are cleavable by RuvC can branch migrate spontaneously within the homologous core sequence (mobile HJs), and experimental determination of a single mobile (catalytically competent) HJ structure among the many isomers would be difficult.
In this work, to understand the RuvC cleavage requirement for the homologous core sequence and to obtain new insights into molecular mechanisms of the HJ resolution by the protein, we designed a systematic series of synthetic HJs. First, we varied the size of the homologous core sequence from 10 bp with five possible steps of branch point migration (pentamobile HJ) to 2 bp with only a single possible step (monomobile) and analyzed cleavability by RuvC. Second, the sequence specificity for RuvC cleavage was examined in bimobile HJs (4 bp of * This work was supported by Grants-in-aid for Scientific Research from the Ministry of Education, Science, Sports, and Culture of Japan (to H. I., Y. K., and H. S.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
¶ Supported by a Taniguchi Fellowship from the Research Foundation for Microbial Diseases, Osaka University.

MATERIALS AND METHODS
Oligonucleotides-The oligodeoxyribonucleotides were synthesized with an automated synthesizer (ABI 380B). These oligonucleotides were purified by reverse-phase C18 column chromatography.
Enzymes and Chemicals-T4 polynucleotide kinase was purchased from Takara Shuzo (Kyoto, Japan). All other reagents (highest available grade) were purchased from Wako Pure Chemicals (Osaka, Japan). RuvC protein was purified as described (Iwasaki et al., 1991). The concentration of RuvC was determined by using a molar extinction coefficient of 7.0 ϫ 10 3 M Ϫ1 cm Ϫ1 at 280 nm.
Construction of Synthetic HJs-Synthetic HJs were prepared by annealing stoichiometric amounts (10 M each in terms of the junction molecule) of four 24-mer oligonucleotides as described (Shida et al., 1996). Higher concentrations of oligonucleotide were required to form stable HJs from 24-mer oligonucleotides (Shida et al., 1996). One of the four oligonucleotides was uniquely labeled at the 5Ј-end using [␥-32 P]ATP and T4 polynucleotide kinase before annealing. Sequences of the oligonucleotides are shown in the appropriate figures.
RuvC Cleavage Assay-The standard reaction mixture (10 l) containing a 32 P-labeled synthetic DNA junction (5 M) and RuvC (20 M) in the reaction buffer containing 20 mM HEPES-NaOH, pH 7.5, 8 mM MgCl 2 , 50 mM sodium glutamate, 2 mM dithiothreitol, and 10% glycerol was incubated for 20 min at 37°C. The reactions were stopped by the addition of 2.5 l of 100 mM EDTA, and the reaction mixtures were evaporated to dryness. The reaction products were dissolved in the loading solution (8 M urea, 0.25% bromophenol blue, and 0.25% xylene cyanol), denatured by heat, and analyzed by 20% PAGE containing 7 M urea and 0.4 ϫ Tris borate/EDTA buffer (90 mM Tris borate, pH 8.3, and 2 mM EDTA), followed by autoradiography. Relative efficiencies of the cleavage reactions with the different substrates were determined by densitometric scanning of the autoradiography with a densitometer (Shimadzu CS9000). RuvC cleavage sites were identified by comparison of the cleavage product bands with 5Ј-32 P-labeled synthetic markers (11-13 mer with sequences identical to the possible products).
Protein-DNA Binding Assay by Gel Retardation-The assay was performed as described (Shida et al., 1996).

Minimization of Homologous Sequence Size in Synthetic HJs
for RuvC Cleavage-In previous studies, the central core of homology in synthetic HJs was required to be cleavable by RuvC, and the predominant cleavage sites by RuvC were 5Ј-T2C-3Ј or 5Ј-T2G-3Ј (Dunderdale et al., 1991;Bennett et al., 1993;Takahagi et al., 1994;Shah et al., 1994). To determine the minimum homologous core still cleavable by RuvC, we made a series of HJs keeping the sequence TC in the homologous core and varying the size of homology (Fig. 1). The synthetic mobile HJs, M1 (bimobile), M2 (trimobile), M3 (tetramobile), and M4 (pentamobile) contain, respectively, 4, 6, 8, and 10 bp of homologous sequences around the crossover, which allows, respectively, two through five steps of possible branch point migration freedom. An immobile junction, IM, which does not possess a homologous core sequence, was used as a control. RuvC bound to all of the HJs with the same affinity, as judged by a gel retardation assay, and mediated the cleavage of all mobile HJs at almost the same high efficiency (ϳ95% of the substrates cleaved) except M4, which was less efficiently cleaved (ϳ70% cleaved), and IM, the cleavage of which was not detected (Fig. 2). To determine the cleavage sites in the junctions, the products were analyzed by denaturing PAGE (Fig. 2). All of the mobile HJs were cleaved by the introduction of single nicks into strands 2 and 4. These cleavage sites were symmetrically related within the homologous core and mapped to the sequence 5Ј-T2C-3Ј in all cases. The cleavage sites are indicated in Fig. 1.
Sequence Specificity for Cleavage of Bimobile HJs-To determine sequence specificity, if any, for RuvC cleavage, we constructed 16 possible bimobile HJ derivatives of M1 (Fig. 3). Only the homologous core sequence differed. Gel shift assay showed that RuvC bound to all of the bimobile HJs with the same affinity (data not shown). However, only 6 HJs (A, 60%; D, 60%; E, 39%; H, 50%; I, 54%; and P, 93%) were cleaved by RuvC with relatively high efficiency, and the other bimobile HJs were not or very poorly cleaved by RuvC (Fig. 4). The cleavage sites of these HJs were determined and are summarized in Fig. 5. All of the nicks were introduced into exactly symmetrical positions at the 3Ј-side of thymine within the homologous region. However, there was no specificity for the base at the 3Ј-side of the thymine. In addition, in the case of A, D, H, and P junctions, the rule of symmetrical sequence iden-FIG. 2. Detection of RuvC-mediated cleavage products of the mobile HJs. HJs containing a uniquely 5Ј-32 P-labeled strand were incubated with RuvC. The DNA products were analyzed by 20% denaturing PAGE and autoradiography. Size markers in lanes S2 and S4 consisted of three possible products of strands 2 and 4, respectively, and were 11-13 bases long. Note that all of the cleaved products of strand 2 of M2-M4 showed the same mobility as that of strand 2 of M1.
FIG. 1. Synthetic HJs containing various lengths of homologous core sequence and the sites of cleavage by RuvC. IM is an immobile HJ without a homologous core sequence. In all mobile HJs (M1-M4), to increase the homology core size, 1 or more base pairs in strands 1 and 2 were changed in the core regions, as depicted in the box, keeping the sequences of strands 3 and 4 unchanged. M1 is a bimobile HJ containing 4 bp of the homologous core sequence, which gives two steps of branch point migration freedom. M2 is a trimobile HJ (6 bp of the homology core, three steps of possible freedom), M3 is a tetramobile HJ (8 bp of the homology core, four steps of possible freedom), and M4 is a pentamobile HJ (10 bp of the homology core, five steps of possible freedom). The crossover points are arbitrarily positioned at the center of each strand. Arrowheads, cleavage sites by RuvC. tity at the 3Ј-side of the cleavage site between the opposing strands was not observed (e.g. in the A junction, T2T of strand 2 and T2A of strand 4 were the cleavage sites, respectively). This demonstrates that the nucleotide base specificity at the 3Ј-side of the nicking sites is not a major determinant of RuvCmediated cleavage. Notably, some of the noncleavable HJs (B, C, L, M, N, and O junctions) also contain thymine in their homologous core sequence; the presence of thymine is obviously not sufficient for cleavage.
Cleavage of a Monomobile HJ-Since sequence identity at the 3Ј-side of the cleavage site was not absolutely required for RuvC cleavage, certain monomobile HJs may be cleaved efficiently by RuvC. We constructed a monomobile HJ, Q, in which only the T-A bond was exchangeable (Fig. 6A). When junction Q was incubated with RuvC, it was resolved efficiently (ϳ90%) by the introduction of symmetrical nicks at the 3Ј-side of thymine in the strands 2 and 4 (Fig. 7). As previously seen in the cleavable bimobile HJs, the bases of the 3Ј-side of cleavage site were different (cytosine in strand 2 and guanine in strand 4). This result also shows that RuvC introduces nicks into the phosphodiester bonds exactly at or 1 base away from the point of strand exchange, because only two junction points are possible in this monomobile junction; the precise distance between the nicking sites and the junction points could not be determined by the previous studies using HJs with larger sizes of homologous core sequences (Dunderdale et al., 1991;Iwasaki et al., 1991;Bennett et al., 1993;Takahagi et al., 1994).
Cleavage of immobile HJs that had 5Ј-TC-3Ј or 5Ј-TG-3Ј sequences in the junctions, as shown in Fig. 6B, were not detected under our experimental conditions (data not shown); the presence of thymine at the junction is apparently not sufficient for RuvC-mediated cleavage.

DISCUSSION
The present study has provided several novel clues to the molecular mechanism of HJ resolution by RuvC. We have shown that a monomobile HJ was cleavable by RuvC, and this demonstrates that RuvC introduces nicks at or 1 base away from the point of strand exchange (Figs. 6 and 7). In retrospect, the large homology cores (e.g. 12 bp in synthetic HJs) previously used as substrates (Dunderdale et al., 1991;Bennett et al., 1993;Takahagi et al., 1994) were not essential. However, not all of the bimobile HJs were cleaved; only 6 bimobile HJs among 16 possible combinations were cleaved by RuvC (Figs. 3  and 4). The size of homology core sequence per se is not the key determinant of cleavage. Introduction of nicks occurred symmetrically at the 3Ј-side of thymine in all cases, with the nucleotide at the 3Ј-side of the thymine of cleavage sites varying. In addition, nucleotide bases at the 3Ј-side of the thymine were not always identical between the two cleaved strands (A,D,H,and Q in Figs. 6 and 7). Therefore, these results suggest that the thymine residue at the cleavage site plays an important role in RuvC-mediated resolution and that RuvC recognizes mainly topological symmetry of the Holliday junction but not DNA sequence symmetry at both sides of the thymine at the cleavage site. Since some bimobile HJs with thymine in the homologous core (B,C,L,M,N,and O in Figs. 3 and 4) and immobile HJs (S-X in Fig. 6B) were not cleaved, other molecular requirements in addition to the presence of thymine in the homologous core are critical for resolution by RuvC. Unlike restriction enzymes, which recognize and cleave specific primary sequences per se, RuvC may require a certain sequence context for catalysis to occur. This context may, for example, induce a certain structure that can interact productively with RuvC on forming an RuvC-HJ complex.
The three-dimensional structure of RuvC has been revealed at a 2.5-Å resolution by x-ray crystallography (Ariyoshi et al., 1994). The subunits in the RuvC dimer are related by a dyad axis, and thus, the DNA binding cleft and catalytic center in each subunit are located symmetrically in the dimer (Saito et al., 1995). This relative positioning ensures the introduction of symmetrical nicks by RuvC. The computer docking model of RuvC structure with DNAs suggested that the HJ in RuvC consists of two antiparallel, quasicontinuous DNA duplexes linked with at least two unpaired nucleotides. Each DNA du- plex is inclined by about 80°to the dyad axis of the dimer (Ariyoshi et al., 1994). This model of HJ differs from that of the protein-free stacked X structure of HJ proposed previously (Duckett et al., 1988;Cooper and Hagerman, 1987;von Kitzing et al., 1990). The stacked X structure exhibits 2-fold symmetry with two strands approximating B-form DNA, whereas the complementary strands are sharply bent where they pass from one helix to the other; as a result, all of the base pairings are maintained around the junction. However, RuvC binding can convert the protein-free junction with a folded stacked X structure into an unfolded form with 2-fold symmetry with base pairings around the crossover disrupted (Bennett et al., 1993;Bennett and West, 1995a). These results suggest that configuration of protein-free HJs is converted into a productive form by RuvC binding.
Using synthetic HJs that are constrained to adopt defined isomeric configurations, it has been shown that the nicks are preferentially introduced into the continuous (noncrossing) pair of strands rather than the bent (crossing) pair of strands of antiparallel forms of HJs (Bennett and West, 1995b). The RuvC-HJ computer docking model agrees with this biochemical evidence. In it, the thymines at the cleavage sites are on the continuous strands of antiparallel forms of HJ and are positioned in the hydrophobic pocket just beside the active sites (Ariyoshi et al., 1994). It has been demonstrated by a study using immobile HJs that the choice of partners of crossing or noncrossing strands is governed by the base sequences at the junction (Duckett et al., 1988). In the case of mobile HJs the partners of crossing or noncrossing strands may also be changeable during spontaneous branch migration in solution. The cleavable bimobile and monomobile HJs used in this study may be able to form such a catalytically competent structure, whereas the noncleavable ones, including immobile HJs, may not.
In summary, the cleavable synthetic HJs may have the following properties: 1) thymines are positioned at or very close to the junction; 2) the strands containing the thymines are continuous (noncrossover) strands, which is governed by the sequence context around the crossover; and 3) the HJ can be converted into a productive form in the complex with RuvC, in which the thymine is in the hydrophobic pocket just beside the active site of RuvC, and base pairings at the crossover are disrupted so that the phosphodiester bond at the 3Ј-side of the thymine should become closer to the catalytic site. The thymine-containing sequence context around the crossover in the synthetic HJs may determine the ability to adopt such a catalytically competent configuration. This model could explain why RuvC does not cleave all of the synthetic HJs that contain a thymine at the junction. Shah et al. (1994) have shown that RuvC cleaves the HJs at hot spots with a consensus sequence of 5Ј-A/TTT2G/C-3Ј in vitro. One of our efficiently cleavable bimobile HJs, P, possesses this consensus sequence, and nicking occurred at 5Ј-TTT2G-3Ј in strand 1 and 5Ј-ATT2C-3Ј in strand 3 in this junction (Fig. 5). Therefore, their result does not necessarily conflict with our results. We have found here that HJs that do not possess the consensus sequences were still cleaved by RuvC. They used a HJ made by the RecA protein and then deproteinized. Thus, their substrate can branch migrate over a large DNA sequence, whereas our synthetic HJs have a very limited distance of branch migration, in which two junction points are possible for the monomobile HJ and three junction points are possible for the bimobile HJ. The equilibria among isomers (or choice of junction points) are governed by the sequence contexts around the encountering junction points (Duckett et al., 1988), and they may be quite variable. The consensus sequence might be selected by the two factors: relative stability and cleavable configuration of the junctions with the sequence (with equilibrium biased to such an isomer). However, our cleavable junctions can take only limited numbers of isomers due to the limited junction mobility, which results in relative abundance of the cleavable isomer. It is conceivable that the favorable isomers occur infrequently and rather unstably in a large Holliday structure with freely mobile junctions. FIG. 6. Monomobile and immobile HJs and their cleavage by RuvC. A, monomobile HJ Q was constructed from four 24-mer oligonucleotides with the same sequences in the flanking regions of four arms as those of IM in Fig. 1. The monomobile homologous core is enclosed in the box. Two isomers with different crossover points are possible by spontaneous branch migration. Arrowheads, cleavage sites by RuvC. B, immobile four-way junctions were constructed from four 24-mer oligonucleotides with the same sequences in the flanking regions of four arms as those of IM in Fig. 1. None of the immobile junctions were resolved by RuvC. FIG. 7. Resolution of the monomobile HJ by RuvC. The monomobile HJ Q, as shown in Fig. 6, was uniquely 5Ј-32 P labeled and incubated with RuvC. The DNA products were analyzed by 20% denaturing PAGE and autoradiography. Size markers in lanes S2 and S4 consist of three possible products (11-13 bases long) for strands 2 and 4, respectively.