THE JOURNAL OF BIOLOGICAL CHEMISTRY 0 1987 by The American Society of Biological Chemists, Inc.

Vol. 262, No. 19, Issue of July 5, pp. 9160-9165, 1987 Printed in U.S.A.

Construction of a Synthetic Holliday Junction Analog and Characterization of Its Interaction with a Saccharomyces cereuisiae Endonuclease That Cleaves Holliday Junctions*

(Received for publication, August 12, 1986)

David H. Evans$ and Richard Kolodnerg From the D a m Farber Cancer Institute, Boston, Massachusetts 02115 and the Department of Biological Chemistry, Haruard Medical School, Boston, Massachusetts 02115

We describe the construction and characterization of an oligonucleotide Holliday junction analog and char- acterize its interaction with a Saccharomyces cerevis- iae endonuclease that cleaves Holliday junctions. A Holliday junction analog containing four duplex arms and 54 base pairs was constructed by annealing four unique synthetic oligonucleotides. Mixing curve anal- ysis showed that the complex contained a 1:l: 1:l mol ratio of the four unique sequence strands. In addition, a linear duplex with a sequence identical to two of the junction arms was also constructed for use as a control fragment. High resolution gel exclusion chromatogra- phy was used to purify and characterize the synthetic junction. The synthetic Holliday junction was found to be a specific inhibitor of a S. cerevisiae enzyme that catalyzes the cleavage of Holliday junctions. Under standard cleavage conditions, 50% inhibition was ob- served at a synthetic Holliday junction to substrate ratio of 711, whereas no inhibition by linear duplex was observed at molar ratios in excess of 150/1. Kinetic analysis showed that Holliday junction was a compet- itive inhibitor of the reaction and had an apparent Ki = 2.5 nM, although the mode of inhibition was complex. The synthetic Holliday junction was not a substrate for the enzyme, but was found to form a specific complex with the enzyme as evidenced by polyacrylamide gel electrophoresis DNA binding assays.

DNA structures in which two duplex DNA molecules are joined to each other by a reciprocal single-stranded cross-over have been postulated to be intermediates in a number of cellular processes. Replication (l), telomere resolution (2), and general (3, 4) and site-specific recombination (5) are all processes that may generate such structures now commonly referred to as Holliday junctions (6). Their importance has prompted physical searches for such structures and the iso- lation of molecules containing Holliday junctions has been described in a number of systems (7-11).

The processing of Holliday junctions could be catalyzed by specific endonucleases and several endonucleases that cleave Holliday junctions have been described. The Escherichia coli

* This work was supported by National Institutes of Health Grant GM29383 (to R. K.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked “aduertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

$ Postdoctoral Fellow of the Alberta Heritage Foundation for Med- ical Research.

§ T o whom correspondence should be addressed Dana Farber Cancer Inst., 44 Binney St., Boston, MA 02115.

bacteriophage T4 gene49 encodes endonuclease VI1 (12). This endonuclease appears to be required for packaging T4 DNA into phage heads and catalyzes the cleavage of multiply branched T4 replication intermediates and other structures containing Holliday junctions. The bacteriophage T7 gene3 product is required for both recombination and for the pro- duction of DNA precursors for DNA synthesis (13). The T7 gene3 encodes the T7 endonuclease I (14) which is a single- stranded DNA-specific endonuclease that also cleaves Holli- day junctions (13). Recently, Saccharomyces cereuisiae has been shown to contain a Holliday junction specific endonu- clease, although this enzyme has not yet been implicated in any particular aspect of nucleic acid metabolism (15,16). The bacteriophage X int protein is required for integrative recom- bination and has been shown to resolve Holliday junctions (5). However, the X int protein appears to differ from the former three enzymes in that it only resolves Holliday junc- tions constructed from att sites and the products of resolution do not contain any broken phosphodiester bonds.

Most studies characterizing enzymes that cleave Holliday junctions have focused on determining the position of the cleavage sites relative to the Holliday junction. The most extensively used substrates for these studies have been cova- lently closed circular DNAs containing palindromes. When heated, such molecules will extrude a cruciform structure that contains a Holliday junction at its base (17) and provides a convenient substrate for the enzymes described above. Using these types of substrates, T4 endonuclease VI1 and the T7 endonuclease I have been shown to cleave cruciform contain- ing DNAs at sites symmetrically located near the base of the Holliday junction (12, 13). The products formed were linear duplex molecules with hairpin ends containing single- stranded breaks that could be sealed with DNA ligase. The s. cereuisiae endonuclease is also thought to cleave cruciform containing DNA in a similar fashion except that this enzyme has not been as extensively studied as the T4 and T7 enzymes (15, 16). The exact distance between the cleavage sites and the Holliday junction remains somewhat uncertain because it is unclear if the bound enzyme alters the extent of cruciform extrusion and because in the case of the T4 endonuclease there may be preferred sequences that are cleaved (12).

Substrates lacking the sequence symmetry present in nat- ural Holliday junctions and cruciforms provide a way by which the problem of junction migration can be circumvented. Sub- strates capable of only very limited branch migration have been constructed by annealing phage X att sites (8) or M13 single strands (18). Kallenbach et al. (19) have constructed four-stranded junctions from synthetic oligonucleotides and characterized them extensively by NMR techniques (20). In order to prevent any exchange of base pairing, however, the

9 160

Synthetic Holliday Junction Analog 9161

junctions constructed by these latter authors lack the four symmetrical bases expected to be found at the core of a Holliday junction. We describe here the construction of a somewhat larger synthetic Holliday junction that contains these four symmetrical bases and illustrate how such an analog can be used to analyze the substrate specificity of an endonuclease that recognizes Holliday junctions.

MATERIALS AND METHODS

Synthesis of Oligonucleotides-Oligonucleotides were synthesized on an Applied Biosystems Model 380A DNA synthesizer using phos- phoramidities and other chemicals purchased from the manufacturer. Crude yields were typically 100-200 OD254 units of each strand with lengths of 25-34 nucleotides. Molar extinction coefficients were cal- culated from the base composition to be (see Fig. l): strand l, e254 = 216,000; strand 2, e254 = 216,000; strand 3, e254 = 227,000; strand 4, f254 = 195,000; strand 5, e254 = 268,000; strand 6, c2% = 268,000.

Purification of Oligonucleotides-Detritylated oligonucleotides ob- tained from the synthesizer were deprotected by incubating -1 ml of oligonucleotide with 3 ml 58% ammonium hydroxide overnight at 55 "C. The oligonucleotides were then taken to dryness in a vacuum centrifuge and resuspended with H,O. Samples containing 30 OD,,, units of oligonucleotide were denatured by heating one part oligonu- cleotide with three parts probe buffer (99% formamide, 11 mM NaOH, 0.05% xylene cyanol, 0.05% bromphenol blue) at 100 "C for 1 min and then fractionated by electrophoresis through a 20% acrylamide gel containing 50% (w/v) urea, 0.09 M Tris borate, pH 8.3, 2.5 mM EDTA. Full-length products were identified by UV shadowing, eluted from the gel, and purified on Waters C-18 Sep-Pak columns according to the method of Lo et al. (23). Samples were resuspended at 200 p M (oligonucleotide) in 10 mM Tris, pH 8.0, 1 mM EDTA, and stored frozen at -20 "C. Yields were typically 30-50% of material applied to the gels.

Hybridization of Strands-Strands were mixed in appropriate ra- tios at strand concentrations of 5-80 p~ and 5 M NaCl added to 0.1 M. The solutions were then heated at 65 'C for 5 min and then incubated at room temperature. Hybridization appeared to be com- plete within 5 min of returning to room temperature. In some exper- iments 5 mM MgC12 was included but this was later found to be unnecessary.

High Performance Liquid Chromatography-A Waters HPLC'sys- tem consisting of automatic gradient controller, Model 441 absorb- ance detector operating at 254 nm, Model 510 pumps, Model U6K liquid injector, and Protein-Pak 300 SW size-exclusion column was used in all chromatography experiments. Column buffer (10 mM Tris- HCl, 100 mM NaCl, 1 mM EDTA, pH 7.5) was filtered through type GSTF 0.22-pm filters (Waters Associates, Milford, MA) before use. To control the temperature the column, precolumn, and inlet tubing were immersed in a regulated water bath.

described by Symington and Kolodner (15) with the following modi- Enzymes-S. cerevisine endonuclease was purified essentially as

fications. Strain LBLl/n was grown on a rotary shaker at room temperature and pretreated 5 h with 0.01% (v/v) methyl methane sulfonate (Kodak). Methyl methanesulfonate was added when cells reached Am = 1 and cells harvested 5 h later at Am = 2-3 (late log). Most experiments described here used enzyme purified through the single-strand DNA cellulose step as described (15). Specific activity of the enzyme preparation (see below) was 39,000 units/mg and was approximately 300-fold purified relative to the crude lysate. The enzyme was further purified to a specific activity of -50,000 units/ mg by chromatography on double-strand DNA cellulose for oligonu- cleotide cleavage experiments. T4 polynucleotide kinase was purified as described by Panet et al. (21).

Enzyme Assays-S. cerevisiae endonuclease was assayed for its ability to cleave pallindromes extruded in plasmid pBR322::PAL114 (22). Standard assays (100 pl) contained 10 pg/ml DNA, 8 mM MgC12, 50 mM Tris-HC1, pH 7.8,O.l mg/ml acetylated bovine serum albumin, 10 m M dithiothreitol, and enzyme. After incubation at 30 "C for 1 h, the DNA was purified by phenol extraction, ethanol precipitated, and recut with P o d 1 in 15-pl reactions containing 6.7 mM Tris-HC1, pH 8.0, 60 mM NaCl, 6 mM MgCl,, 6 mM (3-mercaptoethanol, 0.1 mg/ml acetylated bovine serum albumin, 0.67 mM EDTA, and 5 units PvuII (New England Biolabs, Beverly, MA). After 1 h at 37 "C, samples

The abbreviations used are: HPLC, high pressure liquid chro- matography; bp, base pair.

were electrophoresed through 0.8% agarose gels in 40 mM Tris, 5 mM sodium acetate, 1 mM EDTA, pH 7.9, buffer containing 0.5 pg/ml ethidium bromide. Parallel lanes containing standard amounts of linear pBR322::PAL114 or pBR322 DNAs were also run. Gels were photographed with Polaroid Type 665 positive/negative film and DNA quantitated by densitometry. One unit of enzyme cleaves 1 ng PAL114 DNA in 1 h at 30 "C. This is 0.34-fmol cruciform junctions given a molecular weight of 2.96 x lo6 g/mol for pBR322::PAL114 (22).

Cleavage of Oligonucleotide and Sequencing Reactions-Oligonucle- otides were 5'-end-labeled in 50-pl reactions containing 50 mM Tris- HCl, pH 7.5, 10 mM MgCl,, 2 mM dithiothreitol, 0.1 mM spermidine, 0.1 mM EDTA, 200 pmol of [Y-~'P]ATP (290,000 dpm/pmol), 100 pmol of oligonucleotide, and 3.5 units T4 polynucleotide kinase. After 30 min at 37 "C, 2 pl of 100 p~ unlabeled ATP was added and the reaction continued 30 min. Reactions were stopped by the addition of 0.7 pmol of ATP and 1 pmol of EDTA in 4 pl, additional required oligonucleotides were added, the solution was heated 5 min at 65 "C, and after cooling the labeled oligonucleotide complex was isolated by HPLC. Specific activities were adjusted to 60,000 cpm/pmol with unlabeled material.

Radiolabeled oligonucleotides were digested under standard con- ditions in 100-pl reactions containing 140 units of s. cerevisiae endonuclease and 500 fmol of labeled oligonucleotide substrate. To inhibit a trace of exonuclease (present at a level that removes -0.2 nucleotide/end/h under these conditions) 5 pmol of unlabeled linear duplex was added to all reactions. The linear duplex was observed (see below) to have no effect on the endonuclease activity. Aliquots (50 p1) were removed at time 0 and after 2 h at 30 "C and the DNAs purified by Sep-Pak chromatography in the presence of 50 pg of carrier tRNA (23). The lyophilized products were redissolved in 10 pl of probe buffer and fractionated by electrophoresis through 0.3-mm thick 20% acrylamide/urea gels as described above.

Sequence markers were prepared from appropriate 5'-32P-labeled oligonucleotides as described (24) except that reaction times were adjusted as necessary to give good product distributions. Autoradiog- raphy was performed at -70 "C with Kodak X-Omat AR film and two fluorescent intensifying screens (Du Pont).

Polyacrylamide Ekctrophoretic DNA Binding Assays-Standard reactions (10 pl) contained 50 mM Tris-HC1, pH 7.8, 40 mM NaCI, 1 mM dithiothreitol, 0.1 mg/ml acetylated bovine serum albumin, 10 fmol of 32P-end-labeled polynucleotide complex (-3000 cpm), soni- cated E. coli DNA (a gift of Dr. M. A. Osley, Dana Farber Cancer Institute, Boston, MA, -500-bp in length) and 0.4 units (8 ng) of double-strand DNA cellulose purified S. cereuisine endonuclease. After adding enzyme, reactions were incubated 20 min at 20 "C and then 2 p1 of 50% (w/v) glycerol, 5 mM Tris-HCI, pH 8.0, 0.5 mM EDTA, 0.05% bromphenol blue added, and the samples electropho- resed 3 h at 4 "C at 200 V. Gels contained 8% acrylamide (145:5 acrylamide:bisacrylamide), 1 mM EDTA, 3.3 mM sodium acetate, and 6.7 mM Tris, pH 7.5, and were run in the same buffer.

RESULTS

Designing a Synthetic Holliday Junction-Fig. 1 illustrates the two oligonucleotide complexes that we have prepared. The Holliday junction makes use of two synthetic oligonucleotides that were originally prepared for other studies (strands 1 and 2 (25)) and two that were specifically designed to generate a unique base-pairing mrangement (strands 3 and 4). This structure is potentially capable of base pairing right to the junction as are true Holliday junctions. The base pairs sur- rounding the junction were chosen to prevent branch migra- tion into alternate base-pairing arrangements and the arm lengths are all different. Different arm lengths were used to prevent uniformly distributed exonuclease action from ap- pearing as symmetrical endonuclease incisions in cleavage experiments. I t should be noted that unlike true Holliday junctions the sequences of the four arms present in this structure are all different. In addition, a 34-bp long linear duplex was synthesized as a control. The sequence is that of the vertical arms of the Holliday junction (Fig. 1).

Construction of Oligonucleotide Complexes-HPLC sizing columns are capable of resolving single strands from duplex and four-stranded complexes. The limits of resolution on

9162 Synthetic Holliday Junction Analog

G C C G T A

A T

Y

5

G C G C

C G C G

A T T A

G C A T

C G T A

T A 5 C G 6

GC A T GC CG GC A T GC C G T A C G

C G GC

GC

A T T A

C G C G A T A T GC CG T A Y

FIG. 1. Structure of synthetic Holliday junction and control duplex. Single strands range in size from 25 to 34 bases in length. The sequence of the linear duplex is exactly the same as the long vertical arm of the Holliday junction. Although drawn as a planar structure the Holliday junction is probably tetrahedral in solution (18).

P

these columns are illustrated by the observation that 34-mers are fairly well separated from 25-mers. This technology can be exploited to quantitate complex concentrations and dem- onstrate that appropriate strands form two- and four-stranded complexes with the expected strand stoichiometry.

In order to demonstrate that the oligonucleotides can form a four-stranded Holliday junction, pairs of strands were first annealed in varying ratios. By fixing the amount of one strand and varying the amount of the second one would expect to see the amount of half-junction increases until the second strand is in excess. Then, the amount of half-junction becomes limited by the amount of the first strand and remains con- stant. Conversely, the total amount of monomer would be expected to decrease as half-junction is formed then increases again as the second strand becomes present in excess. Fig. 2A illustrates such a mixing experiment. As predicted, strands 2 and 3 anneal 1:l and similarly 1:l hybridization of strand 1 with strand 4 and strand 5 with strand 6 was also observed (not shown).

Formation of a four-stranded complex can be demonstrated in the same way, except 1:1 mixtures of half-junctions were mixed instead of single strands. Within the limits of error (extinction coefficients are not known with absolute accu- racy), a four-stranded complex consisting of one each of strands 1, 2, 3, and 4 was observed (Fig. 2B) .

Stability of the Holliday Junction-Rechromatography of purified Holliday junction does not give a single peak, but rather free dimer and monomer components reappear (Fig. 3). This is also apparent in Fig. 2 where at 1:l mol ratios of component strands not all of the components form higher molecular complexes. This is due to chromatographic sepa- ration of the strands that are in dynamic equilibrium with one another. The equilibrium distribution of the strand can be calculated from the peak heights and from these values the free energy can be estimated. The results of this analysis will be published elsewhere: but as can be seen by inspection of Fig. 3, this junction is stable at moderate temperatures and salt concentrations. This stability is enhanced by MgC12 (not

* D. H. Evans and R. Kolodner, manuscript in preparation.

0.

08 10 I2 14 Strand role ratio Half-junction mole ratio

FIG. 2. A, strands 2 and 3 anneal 1:l to form half of the Holliday junction. Strand 2 (150 pmol) was mixed with 100-200 pmol of strand 3 in a total volume of 25-35 pl and hybridized as described under “Materials and Methods.” Samples were then chromatographed and the relative amounts of dimer and monomer quantitated using peak elution heights. Breaks in the curves at 1:l mol ratio are observed as expected, demonstrating that the dimer contains strands 2 and 3 in equimolar quantities. E , formation of Holliday junctions from two half-junctions. Duplexes containing equimolar mixes of strands (1 + 4) or (2 + 3) were purified by HPLC and quantitated by UV spec- troscopy. Strands (2 + 3) (39-pmol complex) were mixed with 28-55 pmol of strands (2 + 3) in 40-60 pl, hybridized, and chromatographed as described above. Breaks at 1:l mol ratio confirm that the synthetic junction contains strends 1, 2, 3, and 4 in equimolar quantities. Monomers, (-O-); half-junctions (*-); and Holliday junction, (-X-).

,Tetramer

I I I I I I I I

0 2 4 6 8 1 0 1 2 1 4

Elution Volume (mL) FIG. 3. HPLC analysis of a synthetic Holliday junction.

Buffer temperature was controlled by warming the column and 30 cm of inlet tubing in a regulated water bath. Samples (16 pmol in 50 pl) of column-purified Holliday junction were rechromatographed at 30 “C. Half-junctions and free monomers that form as the junctions disassociate can be seen; the relative proportions of tetramer, dimer, and monomer species provides a qualitative measure of junction stability under these conditions.

shown) and under the reaction conditions described here the junction exists primarily in a four-stranded configuration.

Size Behauior-The unusual structure of the four-stranded complex affects its mobility during electrophoresis through agarose and acrylamide gels (data not shown). In order to minimize disassociation into component strands 15% acryl- amide gels were run at 4 “C in the presence of 0.5 pg/ml ethidium bromide. Under these conditions the Holliday junc- tion (which contains 54 bp) migrated as if it were a 172-bp

Synthetic Holliday Junction Analog 9163

linear DNA. Under other conditions (agarose gels at room temperature with or without ethidium bromide), the Holliday junction migrated as would linear fragments ranging in size from 91 to 230 bp. We have not studied the electrophoretic behavior of this molecule in detail; however, our results clearly show that, as expected, the four-stranded complex has unusual hydrodynamic properties. Similar behavior has been described by others (18, 19).

Interaction with a Saccharomyces cereuisiae Holliday Junc- tion Specific Endonuclease-If the synthetic Holliday junction has a structure analogous to naturally occurring Holliday junctions it should interact with enzymes that recognize such junctions. Initially, we have studied the cleavage of the syn- thetic Holliday junction by the S. cereuisiae Holliday junction specific endonuclease. 5’-32P-labeled Holliday junction was digested with S. cereuisiue endonuclease, and the reaction products were analyzed by electrophoresis through acrylamide gels. In a number of experiments (data not shown) no cleavage products were detected. This indicated that the synthetic Holliday junction was cleaved at (5% of the rate of the standard pBR322::PAL114 substrate and was therefore un- likely to be a substrate for the enzyme.

We have studied the ability of the synthetic Holliday junc- tion to inhibit the cruciform cleavage reaction catalyzed by the S. cerevisiae endonuclease. As illustrated in Fig. 4A, the synthetic Holliday junction is a specific inhibitor of this reaction. The linear duplex has no effect on the S. cereuisiue enzyme even at concentrations in excess of 100 nM, whereas only moderate concentrations of synthetic Holliday junction cause significant inhibition. The inhibition appears to be competitive and from the substrate concentration (3.4 nM) and K,,, (Fig. 4B, 2.3 nM) a K j of 2.5 nM under these conditions can be calculated. The mode of inhibition is complex and further kinetic analysis (Fig. 4B) indicated that the enzyme displayed substrate inhibition at substrate concentrations in excess of 4 nM and that the synthetic Holliday junction was a markedly better inhibitor at low substrate concentrations than at high substrate concentrations.

The binding of 5’-32P-labeled synthetic Holliday junctions to the S. cerevisiae endonuclease has been studied using

A. T 140 t

polyacrylamide gel electrophoretic DNA binding assays (26). Under the conditions of the binding assay, the synthetic Holliday junction exists in equilibrium with the denatured forms discussed above (Fig. 5, lane 1 ). Addition of the enzyme preparation resulted in the formation of several species that migrated more slowly than the unbound Holliday junction (Fig. 5, Zane 2 ) . When increasing amounts of nonspecific competitor E. coli DNA was added to the binding reactions along with the synthetic Holliday junction, the slowly migrat- ing species were gradually competed away (Fig. 5, lanes 3- IO). All except one protein-DNA complex were competed away by the addition of 22 ng of competitor DNA, and the most stable complex was still formed in the presence of 220 ng of competitor DNA. In control experiments in which binding reactions were carried out with linear duplex control DNA, all of the protein-DNA complexes formed were com- peted away by the addition of 22 ng of competitor DNA. This suggests that the most stable complex formed with the syn- thetic Holliday junction represents specific binding to the Holliday junction.

DISCUSSION

Design and Construction of a Synthetic Holliday Jumtion- Natural Holliday junctions that join duplex DNA molecules at regions of homology can move by branch migration due to the presence of sequence symmetry surrounding the joint. This feature complicates the ability to isolate naturally oc- curring Holliday junctions and study their structure because such structures are heterogeneous and sometimes unstable. The ability to branch migrate also complicates the study of enzymes that cleave Holliday junctions because the junction can exist in many positions relative to the sequence of the substrate. In addition, it is also possible that interaction with such enzymes can change the position of the Holliday junc- tion. Kallenbach et al. (19) have described the use of synthetic oligonucleotides to construct Holliday junction analogs that contain fixed junctions. In this communication we describe the construction of a synthetic Holliday junction analog for use in studying enzymes that interact with Holliday junctions. The Holliday junction analog described here differs from

B. i

/ I I -x) 0 10 x ) 100 -10 -a5 o 05 IO E 20 25

,,

C I I (nM) ’/lcrc!esl (nM-l) FIG. 4. A, Dixon plot illustrating specific inhibition of S. cereuisiue endonuclease by the synthetic Holliday

junction. Standard reactions (100 pl) containing 57 units of S. cereuisiue endonuclease and 0.34 pmol of PAL114 DNA were supplemented with oligonucleotide complexes as indicated and enzyme activity quantitated as described under “Materials and Methods.” Under these conditions a Kc of 2.5 nM can be calculated (see text). Holliday junction, (0); linear duplex, (0). B, Lineweaver-Burk analysis of the inhibition of S. cereuisioe endonuclease by a synthetic Holliday junction. Reactions (100 pl) containing 57 units of S. cereuisiue endonuclease were prepared as described except that plasmid concentrations were varied as illustrated and the initial rate of plasmid cutting measured in the presence or absence of 5 nM synthetic Holliday junction. Under these conditions K , and V , are 2.3 nM (eireks) and 56 fmol/h, respectively. No addition, (0); plus 5 nM synthetic Holliday junction, (0).

9164 Synthetic Holliday Junction Analog

1 2 3 4 5 6 7 8 910 1112 13141516171819 x)

,--Holliday junction7 r Linear duplex-i

I Enzyme I - + + + + + + + + + - + + + + + + + + +

Competitor DNA (ng/reaction) -I

FIG. 5. Specific binding of S. cerevisiae endonuclease to synthetic Holliday junction. Oligonucleotides were 5’-3ZP-end- labeled and incubated with or without 0.4 units of S. cereuisiae endonuclease in the presence of varying amounts of sonicated E. coli DNA as described under “Methods and Materials.” After incubation samples were electrophoresed on 8% polyacrylamide gels at 4 “C, dried, and autoradiographed with fluorescent intensifier screens for 2 days at -70 “C. Lanes 1-10, synthetic Holliday junction; lunes 11- 20, linear duplex control. S. cereuisiae endonuclease was omitted in lanes 1 and 11.

those of Kallenbach et al. (19) in that each duplex arm is a different length to aid in mapping specific cleavage products. In addition, we retained a one nucleotide symmetry at the core of the junction since such symmetry might be expected to play a role in the overall three-dimensional structure of the junction. From binding studies it is apparent that four strands anneal with 1:l:l:l stoichiometry to form a structure we presume to be base-paired as illustrated in Fig. 1. The peculiar size behavior is consistent with the suggestion that the arms of Holliday junctions are tetrahedrally arranged (18), but proof of this arrangement and the details of base pairing around the junction cannot be deduced without NMR or x- ray techniques.

Interaction of a Synthetic Holliday Junction with an Enzyme That Cleaves Holliday Junctions-The available physical evi- dence suggests that we have constructed an analog of the core of a Holliday junction. This structure is a competitive inhib- itor of an enzyme that cleaves such junctions (Fig. 4) and this effect is specific to the synthetic junction. Small linear oligo- nucleotides (or ends) are clearly not inhibitory (Fig. 4A). The synthetic junction is a very effective inhibitor with a Ki only slightly higher than the K,,, for plasmid borne junctions which is especially curious considering its small size which probably precludes association with the enzyme by a process of one- dimensional diffusion.

The mode of inhibition is complex. Substrate inhibition is observed at high substrate concentration and this effect is pronounced in the presence of the inhibitor (Fig. 4B) . Fur-

thermore, a t low substrate concentrations enhanced inhibi- tion is also observed. This could reflect a complex, possibly cooperative, mode of interaction between the enzyme and its substrate. Another possible explanation for these kinetics is that they reflect a binding competition which normally favors plasmid binding even if the actual affinities for plasmid and synthetic junctions are very similar. This is because nonspe- cific binding and facilitated diffusion can preferentially target enzyme molecules to plasmid junctions much as EcoRI pref- erentially associates with recognition sites inbedded in large pieces of DNA (27). This competion will, however, be critically dependent upon the relative concentration of the two types of binding sites and under conditions where binding to the oligonucleotide is favored ([SI < K,,, or [I] >> [SI) one would expect enhanced inhibition like that observed. This type of behavior could also be described as very slow, tight binding inhibition (28). More detailed kinetic analysis, additional information on the structure of the enzyme, and footprinting experiments are clearly required before this question can be resolved.

The synthetic Holliday junction was not a substrate for the S. cereuisiae Holliday junction specific endonuclease. There are several possible explanations for this observation. First, the junction may be too small. Thus, the enzyme may not bind it sufficiently well for it to serve as a substrate. A point that may argue against this again comes from observations of EcoRI. EcoRI binds with similar specific activity to the EcoRI recognition site in 34- or 4363-bp DNAs implying flanking nonspecific sequences and charges on oligonucleotides of this size provide few, if any, additional binding sites (29). Another affect of size is the possibility that the cleavage sites are located more than 14 bp away from the junction (the largest symmetrical cutting we could observe). This possibility cannot be ruled out based on plasmid cruciform cleavage experiments since the junction location is not known with precision in such experiments.

Another possible explanation for the failure to observe cleavage of the synthetic Holliday junction is that symmetry elements (homology) within the arms of the junction are recognized by the enzyme. Assuming the S. cereuisiae enzyme is involved in general recombination or recombinational re- pair, it should be nonsequence specific. However, local se- quence symmetry around the Holliday junction may still be important to the cleavage reaction. Nonsymmetric junctions would not normally be formed during genetic recombination unless the recombination substrate contained a heterology, in which case the junction would sometimes be half-symmetric. The S. cereuisiae enzyme cleaves plasmid borne cruciforms, which we presume to be half-symmetric. This is because the excess supercoiling present in small plasmid DNAs is nor- mally sufficient to completely extrude small inverted repeats. The cleavage sites of the S. cereuisiae (15, 16) and also the T7 gene3 endonuclease (13), appear to be located within the symmetrically related sequences. In contrast, T4 endonucle- ase VI1 cleavage sites have been mapped to nonsequence symmetric sites (12) but whether this is a general property of these nucleases is unclear. Assuming the small size of the synthetic junction is not a problem, the data presented here could be taken to suggest that sequence symmetry may, in- deed, be important to the s. cerevisiae endonuclease.

I t is very difficult, however, to separate sequence effects from symmetry effects. A third possible reason that the syn- thetic Holliday junction is not cleaved is because the sequence is one that this enzyme cannot cut. Sites of genetic recombi- nation in yeasts are known to be nonrandomly distributed (30, 31) and whether this is due to nonrandom initiation of

Synthetic Holliday Junction Analog 9165

recombination or resolution remains unknown, but it is quite possible that there are DNA sequences in which Holliday junctions either do not form (32) or if formed are poorly resolved. Recently we have been studying the resolution of other types of Holliday junctions and find quite dramatic effects of sequence on both the directionality of S. cerevisiae endonuclease cleavage and on the rate of cleavage,' and it requires further study to determine whether the lack of se- quence symmetry or the DNA sequence per se is responsible for the inability of the S. cerevisiae endonuclease to cleave the synthetic junction.

These experiments show that small oligonucleotides can be used to construct oligonucleotide complexes which by physical and enzymatic criteria are models of Holliday junctions. Ques- tions remain regarding what features of Holliday junctions are required for the cleavage reaction that is catalyzed by the class of enzymes that process Holliday junctions. By con- structing different synthetic Holliday junctions it should be possible to precisely define the features of Holliday junctions that are recognized by these types of enzymes.

Acknowledgments-We would like to thank Dr. C. Muster-Nassal for helpful advice and P. Morrison for the synthesis of oligonucleo- tides.

1.

2. 3.

4.

5.

6. 7.

8.

9.

REFERENCES Mosig, G. (1983) in Bacteriophage T4, pp. 120-130, American

Szostak, J. W., and Blackburn, E. H. (1982) Cell 29 , 245-255 Meselson, M. S., and Radding, C. M. (1975) Proc. Natl. Acud. Sci.

U. S. A. 72,35&361 Szostak, J., Orr-Weaver, T. L., Rothstein, R. J., and Stahl, F. W.

(1983) Cell 33, 25-35 Weisberg, R. A., and Landy, A. (1983) in Lambda ZZ, pp. 211-250,

Cold Spring Harbor Laboratory, Cold Spring Harbor, New York

Society for Microbiology, Washington, D. C.

Holliday, R. (1964) Genet. Res. 5 , 282-304 Bell, L., and Byers, B. (1979) Proc. Natl. Acud. Sci. U. S. A. 76,

Mizuuchi, K., Kemper, B., Hays, J., and Weisberg, R. A. (1982)

Wolgemuth, D. J., and Hsu, M.-T. (1981) Proc. Natl. Acud. Sci.

3445-3449

Cell 29,357-365

U. S. A. 78,5076-5080

10. Potter, H., and Dressler, D. (1978) Proc. Natl. Acud. Sci. U. S. A.

11. Thompson, B. J., Escarmis, C., Parker, B., Slater, W. C., Doniger, J., Tessman, I., and Warner, R. C. (1975) J. Mol. Biol. 91,409- 419

12. Kemper, B., Jensch, F., Depka-Prondzynski, M. V., Fritz, H.-J., Borgmeyer, U., and Mizuuchi, K. (1984) Cold Spring Harbor Symp. Quant. Biol. 49,815-825

13. deMassey, B., Studier, F. W., Dorgai, L., Appelbaum, E., and Weisberg, R. A. (1984) Cold Spring Harbor Symp. Quant. Biol.

75,3698-3702

49,715-726 14. Sadowski, P. (1972) Can. J. Biochem. 50 , 1016-1023 15. Symington, L. S., and Kolodner, R. (1985) Proc. Natl. Acud. Sci.

16. West, S. C., and Korner, A. (1985) Proc. Natl. Acud. Sci. U. S. A.

17. Mizuuchi, K., Mizuuchi, M., and Gellert, M. (1982) J. Mol. Biol.

18. Gough, G. W., and Lilley, D. M. J. (1985) Nature 313 , 154-156 19. Kallenbach, N. R., Ma, R.-I., and Seeman, N. C. (1983) Nature

20. Wemmer, D. E., Wand, A. J., Seeman, N. C., and Kallenbach, N. R. (1985) Biochemistry 24 , 5745-5749

21. Panet, A., van de Sande, J. H., Loewen, P. C., Khorana, H. G., Raae, A. J., Lillehaug, J. R., and Kleppe, K. (1973) Biochemistry

22. Warren, G. J., and Green, R. L. (1985) J. Bacteriol. 161, 1103-

23. Lo, K.-M., Jones, S. S., Hackett, N. R., and Khorana, H. G.

24. Maxam, A. M., and Gilbert, W. (1980) Methods Enzymol. 65,

25. Muster-Nassal, C., and Kolodner, R. (1986) Proc. Natl. Acud. Sci.

26. Strauss, F., and Varshavsky, A. (1984) Cell 37,889-901 27. Jack, W. E., Terry, B. J., and Modrich, P. (1982) Proc. Natl.

28. Williams, J. W., and Morrison, J . F. (1979) Methods Enzymol.

29. Terry, B. J., Jack, W. E., Rubin, R. A., and Modrich, P. (1983)

30. Orr-Weaver, T. L., and Szostak, J . W. (1985) Microbiol. Reu. 49,

31. Fogel, S., Mortimer, R. K., and Lusnak, K. (1981) in The Molec- ular Biology of the Yeast Saccharomyces-Life Cycle and Znher- itance, pp. 289-338, Cold Spring Harbor Laboratory, Cold Spring Harbor, New York

U. S. A. 82, 7247-7251

82,6445-6449

156,229-243

305,829-831

12,5045-5050

1111

(1984) Proc. Natl. Acud. Sci. U. S. A. 8 1 , 2285-2289

499-560

U. S. A. 8 3 , 7618-7622

Acud. Sci. U. S. A. 79,4010-4014

63,437-467

J. Biol. Chem. 258,9820-9825

33-58

32. Lilley, D. M. J. (1985) Nucleic Acids Res. 13, 1443-1465