Interaction of a four-way junction in DNA with T4 endonuclease VII.

The binding of a synthetic four-way junction in DNA by T4 endonuclease VII has been studied using gel retardation and footprint analysis. Two specific protein-DNA complexes have been observed, but only one is stable in the presence of moderate concentrations of salt. The footprint of T4 endonuclease VII in the salt-resistant complex has been probed using hydroxyl radicals generated by the reaction of iron(II)/EDTA with hydrogen peroxide. The hydroxyl radical cleavage pattern indicates protection of approximately 5 residues in two strands that are diametrically opposed across the junction point.

The binding of a synthetic four-way junction in DNA by T4 endonuclease VII has been studied using gel retardation and footprint analysis. Two specific protein-DNA complexes have been observed, but only one is stable in the presence of moderate concentrations of salt. The footprint of T4 endonuclease VII in the saltresistant complex has been probed using hydroxyl radicals generated by the reaction of iron(II)/EDTA with hydrogen peroxide.
The hydroxyl radical cleavage pattern indicates protection of approximately 5 residues in two strands that are diametrically opposed across the junction point.
General genetic recombination (l-5) and many site-specific recombination events (6-9) take place by the formation and resolution of intermediates in which two DNA helices are interlinked by a crossover or Holliday junction. The structure of these four-way junctions in DNA and of complexes they form with enzymes is of central importance to our understanding of the process of genetic exchange. Endonucleases capable of processing four-way junctions to produce duplex helices have been isolated from Escherichia coli bacteriophages (10, 11) and from yeast (12)(13)(14). Here we demonstrate the binding of a four-way junction by the most extensively characterized enzyme of this class, T4 endonuclease VII (10,(15)(16)(17)(18)(19).
Binding results in the formation of two specific protein-DNA complexes that show differential stability to the presence of low concentrations of monovalent cations. The structure of the most stable form has been probed by hydroxyl radical attack to reveal a footprint that occupies two of the four strands at the junction point. MATERIALS AND METHODS Oligonucleotides-Junction DNA was made by annealing oligonucleotides 1(5'-GACGCTGCCGAATTCTGGCGTTAGGAGA TACCGATAAGCTTCGGCTTAA-3'), 2(5'-CTTAAGCCGAAGCT-TATCGGTATCTTGCTTACGACGCTAGCAAGTGATC-3'), 3(5'-TGATCACTTGCTAGCGTCGTAAGCAGCTCGTGCTGTCTAG-AGACATCGA-3), and 4(5'-ATCGATGTCTCTAGACAGCAC GAGCCCTAACGCCAGAATTCGGCAGCGT-3').

Construction of 3'-Labeled
Substrates-Synthetic four-way junctions were prepared by a method similar to that described by Kallenbach et al. (20). 9 pg of each of four complementary oligonucleotides were annealed in SSC buffer (150 mM NaCl, 15 mM sodium citrate) by incubation for 2 min at 95 "C, followed by 10 min at 65 "C, 10 min at 37 "C, and finally 10 min at room temperature (in a total volume of 40 ~1 Hydroxyl radicals were generated using hydrogen peroxide and iron/EDTA as described (22), and incubation was continued for 2 min hefore the addition of thiourea. The DNA products were ethanol-precipitated, denatured, and electrophoresed on a 12% denaturing acrylamide gel.

AND DISCUSSION
To detect the binding of a DNA junction by purified T4 endonuclease VII, we used gel retardation assays (23-25). In binding reactions in which "'P-labeled junction DNA was incubated with increasing amounts of T4 endonuclease VII, two bands of reduced mobility were observed (Fig. lA, bands a and b). Band a, which migrated slightly behind unbound junction DNA (Fig. lA, band X), was produced at low enzyme concentrations whereas band b was seen at higher enzyme levels. Under identical binding conditions and similar concentrations of enzyme, little binding was seen to a duplex control ( Fig. 1A). The formation of band b was observed only in the absence of monovalent cation, and the inclusion of concen-trations of KC1 or NaCl greater than 25 mM resulted in the formation of band a exclusively (Fig. 1B). This salt-resistant protein-DNA complex was stable up to 300 mM NaCl (data not shown).
To prevent endonucleolytic cleavage of the junction, it was necessary to perform endonuclease VII binding experiments in the presence of EDTA. When EDTA was replaced by 10 mM Mg'+, the junction was cleaved to form fragments with gel mobility equivalent to that of duplex DNA molecules (Fig.  1C). In these reactions, the presence of monovalent, cations had a stimulatory effect, leading to complete cleavage of the substrate.
It has been shown previously that cleavage of a synthetic four-way junction occurs by the introduction of symmetrical cuts about the junction point (26,27). To determine the sites of cutting in the junction used in the experiments described here, four identical junctions were prepared in which one 5'-""P-labeled oligonucleotide was annealed with three unlabeled strands. The junctions were treated with endonuclease VII, and the DNA was denatured and analyzed on polyacrylamide gels containing 7 M urea. As reference markers, 5'-""P-labeled oligonucleotides were sequenced using the chemical method  (21). The results, shown in Fig. 2, A and B, indicate that endonuclease VII cut the DNA predominantly in one orientation by cleavage of a single phosphodiester bond in strand 1 and strand 3. The sites of incision were symmetrically opposed across the junction and were located two nucleotides to the 3'.side of the base of the junction.
Minor sites of cleavage in the opposite orientation (strands 2 and 4) were observed only upon overexposure of the autoradiographs.
Our observation of unidirectional cleavage is consistent with previous experiment,s describing the cleavage of synthetic junction DNA by T4 endonuclease VII (26,27) and is indicative of the junction adopting one of two potential isomeric forms. Choice of isomeric form is thought to he dependent upon stacking interactions which are governed by the base sequences at the junction point (26)(27)(28). The formation of one isomeric structure was advantageous and allowed us to investigate the footprint of T4 endonuclease VII on the junction. Initial studies in which enzyme-junction complexes, formed in the presence of Mg", were probed by DNase I (29) were unsuccessful since the base sequences close t,o the junction point were inaccessible to DNase I and because the enzyme levels necessary to observe binding resulted in excessive cleavage of the substrate DNA. The inaccessibility ofjunction sequences to DNase I has recently been reported hy others (30).
However, the hydroxyl radical method of Tullius and Domhroski (22) or related chemical methods have been used successfully to investigate many protein-DNA interactions (31), including those of sequence-or structure-specific nucleases (32-36). The patterns of Fe(lI)/EDTA/H,OL cleavage on all four strands of the junction in the absence and presence of T4 endonuclease VII are shown in Fig. 3A. The reaction conditions chosen for footprint analysis include 50 mM NaCl in order to favor the formation of only the more stable of the two possible protein-DNA complexes observed in the earlier gel retardation experiments (Fig. 1, A and H, hand a). In the absence of endonuclease VII, we observed that strands 1 and 3 showed reduced cleavage in 2 residues at the branch point, in agreement with the data of Churchill c't al. (28). In the presence of the junction binding protein, this repion of protection was extended to approximately 5 residues (Fig 3A, brachets). Although it was necessary to carry out the hydroxyl radical cleavage reaction in the presence of EDTA, the high levels of T4 endonuclease VII required for binding resulted in some cleavage of the substrate by the enzyme. The sites of cutting under these conditions were located within the protected region and were identical to those observed when EDTA was replaced by Mg" (data not shown). The presence of these cleavage sites served as internal markers, and conparison with Maxam-Gilbert sequencing ladders allowed the precise determination of the residues that were protected by endonuclease VII (Fig. 3H). These results indicate that the cleavage of the junction in one orientation (by cutting of diametrically opposed strands 1 and 3) is a consequence of the way in which endonuclease VII binds tightly to two of the four strands that comprise the junction.
In a series of experiments over a range of T4 endonuclease VII concentrations, we were unable to unequivocally demonstrate binding to strands 2 and 4. However, in some experiments, a weak footprint was observed about the junction point of strand 2 and at a region located approximately 8 residues to the ii'-side of the junction (Fig. 3A, broken Linus). However, since this binding was weak and less reproducible than that observed in strands 1 and 3, it is possible that it may result from the formation of the second protein-DNA complex, as observed by the gel retardation assay (Fig. lA, band b). Alternatively, it might reflect the formation of transient contacts that are unstable during the hydroxyl radical reaction.
Synthetic four-way junctions, such as the one used in this study, have been used as models for Holliday junctions since the chemical structures of their junction points are analogous. However, in contrast to Holliday junctions, these synthetic substrates do not possess homologous arm sequences and this may have important structural consequences. Recently, it was shown that T4 endonuclease VII resolves Holliday junctions made in uitm by RecA protein, the recombination enzyme from E. coli (37). In these reactions, two DNA substrates were recombined such that homologous helices were joined at a single junction point. The presence of endonuclease VII resulted in resolution of the Holliday junctions made by RecA to give rise to the two expected types of recombinant product. The ahility of the endonuclease to recognize a Holliday junction as it moves along DNA, either by branch migration or by unidirectional strand exchange promoted by RecA protein, indicates that the basis of recognition is DNA structure rather than sequence specificity. However, once hound to a junction, the sites of cleavage are influenced by the local DNA sequence about the junction point (17,38 . .

B
in the present study may be best regarded as "frozen" Holliday junctions to which endonuclease VII can bind, and the observed protein-DNA contacts will be composed of the structural and sequence specificities of the nuclease.
Although the protein-DNA contacts detected in the present experiments were obtained in the presence of EDTA, the observed parallels between protection and cleavage are of interest with respect to the possible structures that a fourway junction may adopt. Fig. 4 shows three possible DNA configurations in which the cleavage and footprinting data are indicated. A tetragonal junction in which the four arms are unstacked and maximally extended in a square planar configuration is indicated in Fig. 4A. The adoption of such a structure, originally proposed to occur in the absence of Mg" (27), is inconsistent with recent evidence (39) and with the 2fold symmetric patterns of cleavage and protection observed here. Two-fold symmetric DNA structures are shown in Fig.  4, B and C. In the Sigal-Alberts junction (40) in which helical arms are aligned in parallel, the symmetrical axis of the complex is parallel to the helical axis (Fig. 4B). In contrast, the antiparallel alignment of arm sequences (27) has 2-fold symmetry perpendicular to the plane of the helical axis (Fig.  4C).
Gel filtration experiments (10) indicate that T4 endonuclease VII is a multimeric protein (relative M, 43,000) composed of two identical peptide subunits (M, 18,000 (41)). The hydroxyl radical protection data indicating that endonuclease VII protects 5 deoxyribose residues, in strands that are diametrically opposed, is consistent with the binding of a protein (or proteins) of this size. However, at the present time, the stoichiometry of the enzyme-junction complex is unknown. Parallel helices of the type shown in Fig. 4B would require that each of the crossover strands binds one enzyme unit (i.e. a monomer or a dimer), as shown in Fig. 4E. In contrast, cleavage of antiparallel helices could be accomplished by two enzyme units or by a single dimeric protein (Fig. 4F). In either case, it is likely that the structural distortion of the DNA at the crossover point will be the recognition target for the nuclease.