Ho endonuclease cleaves MAT DNA in vitro by an inefficient stoichiometric reaction mechanism.

Mating type switching in Saccharomyces cerevisiae initiates when Ho endonuclease makes a double-stranded DNA break at the yeast MAT locus. In this report, we characterize the fundamental biochemical properties of Ho. Using an assay that monitors cleavage of a MAT plasmid, we define an optimal in vitro reaction, showing in particular that the enzyme has a stringent requirement for zinc ions. This suggests that zinc finger motifs present in Ho are important for cleavage. The most unexpected feature of Ho, however, is its extreme inefficiency. Maximal cleavage occurs when Ho is present at a concentration of 1 molecule/3 base pairs of substrate DNA. Even under these conditions, complete digestion requires >2 h. This inefficiency results from two characteristics of Ho. First, Ho recycles slowly from cleaved product to new substrate, in part because the enzyme has an affinity for one end of its double strand break product. Second, high levels of cleavage in the in vitro reaction correlate with the appearance of large protein-DNA aggregates. At optimal Ho concentrations, these latter aggregates, referred to as “florettes,” have an ordered structure consisting of a densely staining central region and loops of radiating DNA. These unusual properties may indicate that Ho plays a role in other aspects of mating type switching subsequent to double strand break formation.

Genetic recombination is the intracellular process that moves DNA sequence information from one genomic location to another. A plethora of genetic and biochemical experiments, conducted primarily in prokaryotic and fungal systems, has led to a detailed understanding of the DNA intermediates that arise during this process. Examples of such intermediates include broken duplexes, heteroduplex DNA joints, and Holliday junctions. The first of these intermediates (broken duplexes or double-stranded breaks (DSBs) 1 ) are thought to initiate recombination. Thus, in the yeast Saccharomyces cerevisiae, an organism in which DSB formation has been intensively analyzed, the following lines of evidence indicate that DSBs initiate meiotic recombination. First, DSBs are repaired using gene products required for meiotic recombination (reviewed in Ref. 1). Second, certain meiotic segregation patterns of yeast genes are consistent with molecular models of recombination that posit a DSB initiation event (2). Third, DSBs occur at regions of the genome defined genetically as initiators of recombination (3)(4)(5)(6)(7)(8)(9)(10)(11). Fourth, DSBs introduced into DNA during meiosis stimulate nearby meiotic recombination (12). Finally, meiotic DSBs precede and are independent of pairing between homologs, implying that DSBs initiate chromosomal pairing (13,14). However, despite the intensive analysis of DSB formation in yeast, the endonuclease(s) that generates DSBs during meiotic recombination remains unidentified.
DSBs have also been shown to initiate yeast mating type switching, a mitotic recombination reaction (for a review of yeast mating type, see Ref. 15). Haploid yeast cells come in two mating types, designated a and ␣. These mating types are determined by a master regulatory locus on chromosome III called MAT. MAT contains five sequence blocks: W, X, Y (Ya or Y␣), Z1, and Z2 (Fig. 1A). The W, X, Z1, and Z2 blocks are identical in a and ␣ cells, but each mating type contains a unique MAT Y sequence: Ya in a cells and Y␣ in ␣ cells. Roughly once per cell generation, new Ya or Y␣ information transposes into MAT from two unlinked and transcriptionally inactive loci known as HML␣ and HMRa, thereby switching the cell's mating type.
Mating type switching initiates when an endogenous, sitespecific endonuclease known as Ho makes a DSB at the MAT YZ border (16 -18). The cleavage products and recognition sequence for Ho have been analyzed in detail by Heffron and co-workers (16,17,19,20). Ho cleaves in Z1 between Z3 and Z4 on one strand and between Z7 and Z8 on the other to produce 3Ј-overhangs that are four nucleotides long (base pairs are numbered going away from the YZ junction) (Fig. 1B). A 16-bp region spanning the YZ border (Y6 -Z10) can be cleaved in vitro by purified Ho endonuclease (Fig. 1B). Eight noncontiguous base pairs located within this region are necessary for cleavage. One located in Y (Y4) is an AT base pair conserved in both Ya and Y␣. The other seven are located in Z1 (Z1-Z4, Z6, Z7, and Z9). Surprisingly, however, the 16-bp region sufficient for in vitro cleavage (Y6 -Z10) and a larger, inclusive 20-bp region (Y8 -Z12) are insufficient for in vivo cleavage (19,20). Instead, a 24-bp fragment (Y11-Z13) is required for cleavage inside cells (Fig. 1B).
An analysis of Ho site recognition distinguishes Ho endonuclease from typical prokaryotic type II restriction endonucleases (19,20). Heffron and co-workers (16,17,19,20) showed that gross changes outside the recognition sequence affect Ho cleavage, while some changes inside of it are tolerated by the enzyme. These properties indicate that the Ho recognition site should not be thought of as a highly defined "restriction" site. In this respect, Ho would bind to DNA in a manner similar to transcription factors, an intriguing idea given that Ho contains several putative zinc fingers, structural motifs common to certain classes of transcription factors (21).
While its recognition sequence has been analyzed in great detail, Ho endonuclease itself has been the subject of little biochemical analysis. In this report, we characterize the fundamental biochemical parameters of the enzyme. This study has two purposes. The first is to increase our understanding of the mechanism by which Ho generates a DSB at MAT. Given the similar genetic requirements for mating type switching and "standard" meiotic recombination, this information should increase our understanding of DSB formation in other types of recombination reactions. The second purpose is to identify biochemical parameters of Ho that can be used as base lines in future experiments on accessory factors that may be required for in vivo cleavage. Our results indicate that Ho is not merely the yeast version of a bacterial restriction enzyme. In fact, Ho has some very unusual characteristics that may indicate that the enzyme has other roles in mating type switching subsequent to DSB formation.

MATERIALS AND METHODS
Strains and DNA-Plasmid pRK128 is a bacteriophage -derived vector containing the gene for the thermolabile repressor cI 857 and the P R promoter driving the complete coding sequence of Ho endonuclease (17). Plasmid pDC283 is a pBR322-based plasmid carrying the entire MATa locus and flanking DNA sequences (17). Both pRK128 and pDC283 were kindly supplied by Richard Kostriken. Plasmid pAV115, kindly supplied by Andrew Vershon, is a pBR322-based plasmid carrying the entire MAT␣ locus and flanking DNA sequences (22). Plasmid pRW17 is a YCp50-based plasmid containing MAT␣ DNA from Y22 to Z22 cloned between the HindIII and BamHI sites. Oligonucleotide RHo has the following sequence: GTA AAA TTT TAT AAA CTT TAT A-AA ATT TTA CTG TT. Oligonucleotide LHo has the following sequence: GCG GAA AGC TGA ACT TCA GCT TTC CGC AAC A. The Esche-richia coli strain used in this study was DH5␣FЈ (FЈ/endA, hsdR17 (r K Ϫ m K Ϫ ), supE44, thi-1, recA1, gyrA (Nal r ), relA1, ⌬(lacZYA-argF)U169, deoR (80dlac⌬(lacZ)M15)).
Chemicals and Enzymes-Sepharose CL-4B, DEAE-Sephacel, and Sephadex G-200 were from Pharmacia Biotech Inc. Bio-Rex 70 and Chelex 100 were from Bio-Rad. Phosphocellulose P-11 was from Whatman and was prepared by the manufacturer's instructions. Dithiothreitol (DTT), benzamidine hydrochloride, phenylmethylsulfonyl fluoride (PMSF), N-ethylmaleimide, and 1,10-phenanthroline were from Sigma. Ethylene glycol was from J. T. Baker Co. Low molecular weight protein standards were from Bio-Rad. High and low molecular weight calibration standards for gel filtration were from Pharmacia.
Assay for Ho Endonuclease-The standard reaction mixture (25 l) contained 20 mM Tris-HCl, pH 7.5, 80 mM NaCl, 8 mM MgSO 4 , 4 mM DTT, 2 mM benzamidine, 0.45 nmol of EcoRI-cleaved pDC283 (expressed in molar concentration of nucleotides), and Ho endonuclease. Reactions were incubated at 30°C for 60 min and then terminated by the addition of 5 l of stop solution (100 mM EDTA, 1% (w/v) SDS, 1 mg/ml proteinase K, 0.25% (w/v) bromphenol blue, and 50% glycerol), followed by a 30-min incubation at 65°C. Terminated reactions were loaded on a 0.8% (w/v) agarose gel made up in 0.04 M Tris acetate and 1 mM EDTA and electrophoresed at 7 V/cm for 2 h. Gels were stained in 0.5 g/ml ethidium bromide, and digitized pictures were generated on a GDS 7500 gel documentation system (UVP). Quantitation of digitized pictures was carried out using the IP Lab gel software package (Signal Analytics Corp.). One unit of activity is defined as the amount that cleaves 1 g of EcoRI-linearized pDC283 in 1 h at 30°C.
To test the effects of pH on Ho activity, the Tris-HCl in the standard reaction was replaced with 20 mM KH 2 PO 4 /K 2 HPO 4 for the pH range from 6.4 to 8.1 or with 40 mM Tris maleate/NaOH for the pH range from 8.6 to 9.2. All other reaction components were the same.
Cell Growth-The HO bacterial strain was grown at 30°C in 3.6 liters of LB medium supplemented with 50 g/ml ampicillin. When the culture reached an absorbance at 600 nm of 0.8, 2.4 liters of the same  (19,20). The bars marked Active in vivo and Inactive in vivo indicate sequences that function or fail to function, respectively, in an in vivo cleavage assay (see text). medium pre-equilibrated to 65°C were added to bring the temperature of the culture immediately to 42°C. The cells were then grown for an additional hour at 42°C and harvested by centrifugation. The resulting cell pellet was washed one time in ice-cold 50 mM Tris-HCl, pH 7.5, and pelleted again by centrifugation. The washed pellet was quick-frozen in liquid nitrogen and stored at Ϫ80°C. For a typical preparation, the above procedure was repeated four times to yield ϳ30 g of wet cell pellet.
Purification of Ho Endonuclease-Purification of Ho was carried out as described by Kostriken and Heffron (17) 2 with minor modifications. All procedures were performed on ice or at 4°C. Quantitation of the purification is given in Table I.
High-resolution Gel-filtration Chromatography-The Stokes radius of Ho was determined by gel filtration according to the method of Laurent and Killander (23). Typically, 50 l of Fraction III were loaded at 0.2 ml/min on a Sephadex 200 HR 5/20 column (Pharmacia) equilibrated with Buffer C. Eluted fractions (0.027 ml) were collected and analyzed for Ho by the standard cleavage assay. A calibration curve was generated using ribonuclease A, chymotrypsinogen A, ovalbumin, albumin, aldolase, catalase, and ferritin, with Stokes radii of 16.4, 20.9, 30.5, 35.5, 48.1, 52.2, and 61.0 Å, respectively. Blue dextran 2000 (Pharmacia) was used to mark the void volume.
Zinc Ion Dialysis-All dialysis and reaction buffers were pretreated with Chelex 100 chelating resin to remove zinc ions. Zinc ions were subsequently removed from Ho fractions by dialysis against buffer containing the Zn 2ϩ -specific chelator 1,10-phenanthroline. Typically, 100 l of Fraction III were dialyzed at 4°C for 2 h against 100 ml of Buffer C containing 2 mM 1,10-phenanthroline. This procedure was repeated two more times using fresh dialysate. In control experiments, 100 l of Fraction III were dialyzed under the same conditions, with the exception that the 1,10-phenanthroline was omitted from the dialysate.
Other Procedures-SDS-polyacrylamide gel electrophoresis of purified fractions was performed on 10% gels (24). Proteins were visualized by staining with Coomassie Brilliant Blue R-250. Protein concentrations were determined by the method of Bradford (25) using bovine serum albumin as the standard. Electron microscopy of Ho-DNA complexes was carried out according to the method of Inman and Schnös (26) with the following modifications. After adsorption of DNA and protein-DNA complexes, the grids were positively stained by brief immersion in 95% ethanol containing 3% uranyl acetate. A thin film of platinum-palladium was subsequently evaporated at low angle onto the grid to make the DNA and protein-DNA complexes clearly visible.

RESULTS
Purification of Ho Endonuclease-To isolate biochemically homogeneous Ho endonuclease, we used the bacterial overexpression system of Kostriken and Heffron (17). This system utilizes a plasmid containing the yeast HO gene driven by the P R promoter of bacteriophage . Also carried on the plasmid is the gene for the thermolabile repressor cI 857 , which represses the P R promoter at low but not high temperatures. Ho endonuclease was induced by heat shock, and the enzyme was purified by the method of Kostriken and Heffron (17) with slight modifications (see "Materials and Methods"). The assay for Ho monitored the cleavage of a linear DNA molecule that contained the MATa locus and flanking sequences; 1 unit of activity is defined as the amount of enzyme that cleaves 1 g of substrate in 1 h at 30°C (see "Materials and Methods"). The results of a typical purification are summarized in Table I and Fig. 2. The purified protein migrated as a single band of M r ϭ 65,000 on an SDS gel, in close agreement with the predicted molecular weight of the Ho polypeptide (M r ϭ 65,940). The final fraction retained Ͼ85% of its activity after 12 months of storage at Ϫ20°C.
Fundamental Biochemical Characteristics of Ho Endonuclease-We used the homogeneous protein to characterize the fundamental biochemical properties of Ho. The first step in this process was to determine the optimal in vitro conditions for the enzyme. Cleavage reactions were carried out under varying conditions, and the activity of Ho under these conditions was determined by the standard agarose gel assay. Several conclusions were drawn from these data. First, Mg 2ϩ ions were absolutely required for Ho activity, as no cleavage was observable 2 R. Kostriken, personal communication. in the absence of the ion (Fig. 3A). The optimal concentration of MgCl 2 was 5-30 mM (Fig. 3A). Second, NaCl was not required for activity, as reasonable cleavage was seen in its absence (Fig.  3B). Cleavage, however, was stimulated 2.5-fold when NaCl was present at concentrations between 50 and 100 mM (Fig.  3B). Third, the activity of Ho exhibited a relatively broad pH range, with optimal cleavage occurring between pH 7.5 and 9.0 ( Fig. 3C). Fourth, a reduced sulfhydryl group(s) on the enzyme may be involved in Ho activity, as the addition of N-ethylmaleimide to the cleavage reaction at a concentration of 50 mM eliminated product formation (data not shown). Finally, the enzyme appeared relatively stable, as the addition of protein stabilizers had no effect on cleavage (data not shown). These parameters are similar to those of standard type II restriction endonucleases.
Effect of Zinc Ions on Ho Endonuclease Activity-One interesting feature of the amino acid sequence of Ho is that the carboxyl-terminal 100 amino acids contain five putative zinc finger motifs (21). Recently, a mutation in the HO gene product that eliminates activity was mapped to one of these potential zinc fingers, providing genetic evidence for the importance of these motifs (27). To provide further biochemical evidence that these fingers are important for Ho activity, we asked whether zinc ions were required in the in vitro cleavage reaction. This was determined by carrying out two experiments (see "Materials and Methods"). In the first, an aliquot of Fraction III was dialyzed for 6 h against a buffer containing 1,10-phenanthroline, a Zn 2ϩ -specific chelating agent. In the second, an identical aliquot was dialyzed for the same time against a buffer lacking 1,10-phenanthroline. The two samples were then added back to reaction mixtures containing Mg 2ϩ ions, but no Zn 2ϩ ions, and the activities of the dialyzed proteins were determined by the standard agarose gel assay. The results showed that the aliquot dialyzed against 1,10-phenanthroline retained only 0.4% of its activity, as predicted for a protein that requires zinc. By comparison, the aliquot dialyzed against buffer lacking 1,10phenanthroline retained 95.5% of its activity, indicating that dialysis per se was not responsible for the inactivation of the first aliquot.
These results argue that the removal of zinc ions from the Ho preparation inactivates the enzyme. However, another interpretation is that the 1,10-phenanthroline inactivated the protein through some other nonspecific mechanism. To rule out this possibility, the enzyme that had been dialyzed against 1,10-phenanthroline was added to a standard reaction supplemented with 10 Ϫ5 M ZnCl 2 . Under these conditions, cleavage was stimulated 40-fold, i.e. the inactivated aliquot went from 0.4 to 16% wild-type activity. Thus, a large part of the inactivation was directly attributable to the loss of zinc ions. Further experiments showed that optimal activity required a final Zn 2ϩ concentration between 10 Ϫ6 and 10 Ϫ4 M. Even under optimal conditions, the reconstituted activity was 6-fold lower than the original wild-type activity. The reason for this partial reactivation remains unclear. However, it is possible that refolding a zinc finger after removal of the Zn 2ϩ ion is a kinetically unfavorable reaction. Alternatively, the 1,10-phenanthroline might have inactivated a fraction of Ho by other nonspecific means.
Ho Is a Monomer in Solution-We next determined whether Ho existed in solution as a monomer or some higher order multimer. (Given the asymmetry of the Ho recognition sequence, there was no a priori reason to expect that the enzyme would exist as a dimer like many type II restriction endonucleases.) Purified Ho was subjected to high-resolution gel-filtration chromatography (see "Materials and Methods"), and the FIG. 2. Purification of Ho from a bacterial overexpression system. Ho was purified as described under "Experimental Procedures." Aliquots from Fractions I-VI (see Table I) were electrophoresed on a 12% polyacrylamide gel and stained with Coomassie Blue. Lane M contains electrophoretic standards (Bio-Rad) with the indicated molecular weights. elution of Ho activity was compared with the elution of known molecular weight standards. From these data, the apparent molecular weight of Ho was calculated using the method of Laurent and Killander (23). Ho endonuclease activity eluted as a symmetrical peak that corresponded to a native molecular weight of 81,000 (Fig. 4). Ho therefore appears to be a slightly elongated monomer in solution.
In Vitro Cleavage by Ho Is Inefficient-We defined 1 unit of activity for Ho as the amount of enzyme necessary to cleave 1 g of MAT vector in 1 h. According to this definition, Fraction VI, which appeared to be electrophoretically pure, had a specific activity of 350 units/mg of protein. By comparison, using 1 unit defined in an analogous fashion, a purified fraction of EcoRI had a specific activity of ϳ1.5 ϫ 10 6 units/mg of protein (28). Thus, the specific activity of our purified Ho fraction was Ͼ4000-fold lower than that of a typical EcoRI preparation. We also observed that many of our in vitro reactions never went to completion at lower enzyme concentrations, even when the digestion period extended over several hours (data not shown). These data indicated that Ho was extremely inefficient in the in vitro reaction.
To better understand this property, we further characterized the stoichiometry requirements of Ho. Different amounts of enzyme were incubated with a fixed amount of DNA; aliquots were removed at various times; and the extent of cleavage was determined by the standard agarose gel assay (Fig. 5). Full cleavage occurred in the reaction when Ho was present at 3.45 M (2 units/25 l), i.e. 1 Ho molecule/2.6 bp of substrate DNA. For the 8100-bp MAT vector, this translated into ϳ3000 Ho molecules/recognition sequence. When Ho was present at a concentration of 1 molecule/6.5 bp (0.8 units/25 l), ϳ10% of the molecules remained uncleaved. Moreover, product formation plateaued after 2-3 h of incubation in all the reactions, indicating that Ho cleavage was exceedingly slow even at high concentrations of enzyme.
One way to account for this inefficiency is to hypothesize a gross inactivation of the enzyme during the extensive purification. Four observations, however, argue against this interpretation. First, the use of earlier chromatographic fractions led to similar kinetic parameters (data not shown). Second, as de-scribed above (see "Purification of Ho Endonuclease"), purified Ho remained stable after long periods of storage, which argued against any inherent instability of the enzyme. Third, as described above (see "Effect of Zinc Ions on Ho Endonuclease Activity"), purified Ho could be dialyzed for up to 6 h at 4°C and still retain full activity, again indicating that the activity was stable. Finally, endonucleases that are structurally and functionally homologous to Ho, such as I-SceI, I-SceII, I-SceIII, I-SceIV, and PI-SceI, are also much less active than standard restriction enzymes in in vitro cleavage reactions, arguing that the inefficiency seen in the Ho reaction is common to all members of this class of endonucleases (29 -38).
Ho Forms Large Aggregates with DNA-What then accounts for the inefficiency of Ho in the in vitro reaction? One possibility is that the majority of Ho is shunted down a cleavagedeficient pathway. Our first evidence in support of this possibility came during a band shift analysis. We observed that the addition of Ho to short double-stranded oligonucleotides carrying the Ho recognition sequence resulted in protein-DNA complexes that were too large to migrate into the gel (data not shown). This suggested to us that Ho, which is a monomer in solution, might bind as a multimer to DNA. If true, then at least some of the Ho molecules bound to substrate might not participate in cleavage.
We used electron microscopy to further analyze the formation of complexes between Ho and DNA. Two sets of reactions were conducted. In the first set, the substrate DNA was linearized pRW17, a YCp50-based vector carrying the Ho recognition sequence (see "Materials and Methods"). Ho was added at concentrations of 0.2 molecules/bp (Reaction 1), 0.02 molecules/bp (Reaction 2), and 0 molecules/bp (Reaction 3). In the second set, the substrate DNA was linearized pSF21, a YCp50based vector carrying non-MAT DNA. Ho was added again at concentrations of 0.2 molecules/bp (Reaction 4), 0.02 molecules/bp (Reaction 5), and 0 molecules/bp (Reaction 6). All six reactions were incubated at 30°C for 30 -45 min, and the resulting protein-DNA complexes were spread on collodion grids according to the protocol of Inman and Schnös (26). Electrophoretic examination indicated that 70% of the substrate DNA was cleaved in Reaction 1 and Ͻ5% in Reaction 2. No cleavage occurred in the other reactions.
These experiments confirmed that Ho forms aggregates with DNA (Table II) bp, DNA molecules in the solution aggregated into structures referred to here as "florettes" (Fig. 6C). A florette is defined as a large complex containing a densely staining central region and surrounding loops of DNA. The sizes of the florettes were somewhat variable, although most contained ϳ15-20 loops. The formation of florettes in Reaction 1 was very efficient, as they comprised 66% of the visible structures in the solution; the other 30% constituted mainly of discrete molecules, defined as unaggregated single molecules of DNA ( Fig. 6 and Table II). It is important to state, however, that since the florettes likely contain multiple molecules of DNA (see "Discussion"), significantly Ͼ66% of the total DNA molecules may be present within florettes.
The Ho recognition sequence was not necessary for florette formation. As seen in Table II, florettes formed as efficiently on linearized pSF21 as on pRW17. Moreover, electron microscopic examination of the florettes revealed no significant differences between the two substrates (data not shown). This lack of specificity indicated that florette formation depended upon some other feature of the DNA, possibly the ends of the linear substrate molecules. As seen in Fig. 6C, florettes contain loops of DNA. Even though the original substrate was linear, relatively few DNA ends were visible in florettes, suggesting that they were sequestered within the densely staining core region. If the formation of florettes was primarily dependent upon ends, it might be expected that they would form efficiently on any linear substrate, even those lacking an Ho recognition sequence. Further evidence for the binding of Ho to ends is described below (see "The Y-proximal End of the Break Competes Efficiently in the Ho Cleavage Reaction").
The amount of protein required to form florettes correlated with the amount of protein required for cleavage. Thus, when the protein/DNA ratio was decreased 10-fold to 1 Ho molecule/50 bp, cleavage and florette formation dropped to Ͻ5% (Table II). Aggregation of substrate still occurred under these conditions, but the aggregates were much more unstructured, forming what is referred to as "tangled aggregates" (Fig. 6B). Again, the frequency of tangled aggregate formation was very high, and an Ho recognition sequence was not required. The amount of DNA in tangled aggregates appeared more variable. These results suggest that florettes, rather than aggregation per se, are required for optimal cleavage in vitro.
Ho Cleavage Occurs by a Stoichiometric Reaction Mechanism-The kinetic analysis shown in Fig. 5 also indicates that Ho recycles inefficiently from product DNA to new substrate. At subsaturating amounts of Ho, product formation reached a series of plateaus below full cleavage. This characteristic is the hallmark of a stoichiometric mechanism, defined as a reaction mechanism in which a protein catalyst carries out one round of product formation and then fails to recycle to a new substrate.
By comparison, if Ho recycled, as is true of most enzymes, full cleavage would occur at subsaturating amounts, although the slope of product formation with time would decrease with decreasing amounts of enzyme. We hypothesize that Ho fails to recycle because of its sequestration within florettes (see "Discussion").
The Y-proximal End of the Break Competes Efficiently in the Ho Cleavage Reaction-Ho is a member of a class of endonucleases with similar protein sequences. One well studied member of this class is I-SceI, a Group I intron-encoded endonuclease present in yeast mitochondria. The studies described above indicate that Ho and I-SceI share not only structural homology, but also functional homology. A detailed kinetic analysis showed that I-SceI, like Ho, recycles inefficiently from cleaved product DNA to new substrate (39). Another interesting characteristic of I-SceI is that it has an affinity for one of the ends of its double strand break product (39). Given the similarities between the two enzymes, we wondered whether the same might be true for Ho.
To test this, we carried out standard cleavage reactions in the presence of oligonucleotide LHo or RHo, two self-complementary competitor oligonucleotides that mimicked the Y-proximal and Z-proximal ends, respectively, of the doublestranded break. Renaturation of LHo generated a "snap-back," double-stranded molecule consisting of the 16 bp normally Y-  proximal of the cleavage site. The competitor was also designed to maintain the appropriate 3Ј-overhang generated after Ho cleavage. Renaturation of RHo produced a similar doublestranded molecule, with the exception that the 16 bp were from the Z side of the break. The Y-proximal oligonucleotide strongly inhibited cleavage in the in vitro reaction, regardless of whether the substrate was a MATa or MAT␣ vector (Fig. 7). The Z-proximal oligonucleotide, however, had a much smaller effect (Fig. 7). We conclude that Ho, like I-SceI, has a higher affinity for one end of its double strand break product, in this case, the Y-proximal end. As described under "Discussion," this preferential binding is consistent with molecular studies of mating type switching in yeast.

DISCUSSION
Mating type switching in yeast is an ideal model system for analyzing genetic recombination (15). The reaction is relatively simple, highly efficient, and controllable (40 -42). Moreover, the mechanism of mating type switching shares the following important similarities with meiotic recombination. First, DNA intermediates that arise during mating type switching (heteroduplex DNA joints, Holliday junctions, etc.) are extremely similar to those that arise during meiotic recombination (42). Second, gene products required for optimal mating type switching are also required for meiotic recombination, such as the products of the RAD51, RAD52, RAD54, RAD55, RAD57, and XRS2 genes (41,(43)(44)(45). Finally, mating type switching, like meiotic recombination, initiates with a DSB (16 -18). Regarding this last point, the mating type system is particularly attractive in that Ho endonuclease, the enzyme responsible for generating the DSB, has been identified (16,17).
This laboratory is using the mating type system to analyze molecular mechanisms that regulate the formation of DSBs at MAT. To understand these mechanisms, it is crucial to understand the biochemical characteristics of Ho endonuclease. As a result of the studies of Heffron and co-workers (16,17,19,20), DNA sequences required for cleavage within the Ho recognition sequence have been identified and characterized (Fig. 1B). In the studies described here, we have characterized several fundamental biochemical parameters of the enzyme itself.
Perhaps the most salient feature of Ho is its extreme inefficiency in an in vitro cleavage reaction. We observed that maximal cutting occurred when Ho was present at a concentration of 1 Ho molecule/3 bp of DNA, which for our substrate translated into ϳ3000 Ho molecules/recognition sequence. Moreover, even when present at optimal concentrations, complete digestion in the in vitro reaction required Ͼ2 h. Given the affinity of Ho for the Y-proximal end of the break (Fig. 7), we wondered whether the ends of our linear substrates might nonspecifically titrate out the enzyme, thereby leading to the apparent low activity in the cleavage assays. Three results, however, argue against this interpretation. First, the right end of the break competed very inefficiently (Fig. 7). Second, linearized pBluescript DNA was not an efficient competitor in the in vitro cleavage reactions (data not shown). Finally, cleavage of relaxed circular DNA, i.e. DNA that lacks ends, did not cleave significantly more efficiently than linear substrates (data not shown). We suggest instead that Ho requires other proteins to cleave MAT DNA efficiently. Recently, we identified a protein called YZ-binding protein that acts as a positive activator of Ho cleavage inside cells (46). Other proteins may also be required.
An unusual characteristic of the in vitro cleavage reaction was the formation of large protein-DNA aggregates, referred to as florettes, which are defined as structures containing a densely staining core region and radiating loops of DNA. When reactions containing the optimal concentration of Ho were spread for electron microscopic analysis, the number of separate DNA molecules/grid was significantly lower than the number seen in the control reactions containing no Ho. This suggests to us that the florettes contain multiple DNA molecules. In support of this, Ho at suboptimal levels generated tangled aggregates that clearly contained multiple molecules. It is difficult to ascertain whether florettes result from an artifactual aggregation process. However, all of the florettes within a sample (and in fact between samples) were strikingly similar in their morphology, relative lack of free DNA ends, and number of loops. This morphological consistency might not be expected to result from a random aggregation process. Moreover, other proteins known to aggregate DNA do not generate the type of structure observed here (47)(48)(49). We therefore suggest that the ability to aggregate multiple DNA molecules may be intrinsic to Ho.
Florette formation may partially explain the unusual kinetics of the in vitro reaction. Ho cleavage occurred by a stoichiometric mechanism, in which the protein recycled poorly from product to new substrate. To account for this, we hypothesize that florettes, once formed, are relatively stable, i.e. their dissociation and reassembly on new molecules are slow. Inherent to this model is the assumption that florettes are required for cleavage. In support of this, florettes failed to form at lower concentrations of Ho (300 Ho molecules/recognition sequence), and this failure correlated with a low level of cleavage. It is important to state that an unstructured aggregate still formed under these latter conditions, indicating that aggregation per se was not required for cleavage. If we also hypothesize that florette formation is relatively slow, which seems reasonable given its complex structure, this might explain why cleavage, even in the presence of optimal amounts of Ho, is also slow.
Ho belongs to a family of nucleases that share a common dodecapeptide motif first identified in mitochondrial endonucleases. This family of enzymes includes intron-homing endonucleases, such as I-SceI, I-SceII, I-SceIII, I-SceIV, and Endo.SceI from yeast (29,30,(33)(34)(35)(36)(37)(38). Also included in this group are endonucleases encoded by inteins (internal protein sequences that are cleaved from larger primary translation products), such as PI-SceI (31, 32). These proteins share many biochemical characteristics with Ho. For instance, all recognize large, non-palindromic sequences, and several have specific activities significantly lower than those of type II restriction endonucleases. In the case of I-SceI, this lower activity has been studied in depth by careful kinetic analysis of the in vitro cleavage reaction (39). The I-SceI cleavage reaction was found to separate into two kinetically distinct phases, an early and rapid cleavage phase and a later and slower turnover phase.
Given the similarities between Ho and I-SceI, we wondered whether the functional homology between the enzymes might extend to other characteristics. One interesting characteristic of I-SceI is that one end of its DNA break was a more efficient competitor than the other end in an in vitro cleavage reaction (39). Based on this observation and the enzyme's slow turnover, Perrin et al. (39) hypothesized that I-SceI remains bound to one side of the break after cleavage. This may occur so that the protein can participate in later steps of the mitochondrial recombination reaction; alternatively, it may occur so that the protein can help regulate the RNA splicing reaction responsible for I-SceI production (39). Our experiments also show that the Y-proximal side of the Ho break is a better competitor than the Z-proximal side, further extending the similarities between these two proteins.
During mating type switching in vivo, the DSB formed by Ho is acted upon by an unknown exonuclease(s) that exposes single-stranded DNA for subsequent invasion into the silent loci (42). It is intriguing to note, however, that exonucleolytic degradation into Y is blocked after Ho cleavage, while degradation into Z occurs normally (42). Degradation into Y does occur, however, in strains lacking the silent loci. This latter observation makes two points regarding the asymmetric degradation. First, it is not intrinsic to the exonuclease that attacks the break; and second, it is dependent upon an interaction between MAT and the silent loci (42). One simple model to account for these phenomena hypothesizes that a protein complex remains bound to the Y side of the break, thereby blocking entrance of the exonuclease. If true, then formation of the complex depends upon an interaction between MAT and one of the silent loci. The properties of Ho defined in this study suggest that it might participate in such a complex. As required by the model, 1) Ho does not recycle efficiently from cleaved DNA; 2) it forms aggregates containing multiple DNA molecules; and 3) it binds more tightly to the Y-proximal side of the break than to the Z-proximal side. We therefore hypothesize that Ho remains bound to the Y-proximal side of break and participates in the formation of a synaptic complex between MAT and the silent loci.
The conclusion that Ho remains bound to the Y-proximal side of the break, however, must be tempered by another set of studies carried out by Sugawara and Haber (40). These authors showed that both ends of the break were susceptible to exonucleolytic degradation in vivo when the Ho recognition sequence was integrated into non-MAT DNA in the yeast genome. Thus, with respect to nuclease sensitivity, cleavage in this latter system is similar to cleavage at MAT in the absence of the silent loci (42). Therefore, Ho may remain stably bound in vivo to the Y-proximal side of the break only in the context of other factors specific to the mating loci.
Finally, the primary structure of Ho, predicted from the sequence of the HO gene, indicated several potential zinc finger motifs (21). Recently, Raveh and co-workers (27) have shown that a single mutation that maps to the first of these zinc fingers eliminates Ho cleavage activity in vivo. These latter results, in combination with our observation that zinc ions are required for Ho activity, strongly argue that Ho endonuclease contains a zinc finger(s) that is necessary for activity.