A Maturase-like Subunit of the Sequence-specific Endonuclease Endo.Sce1 from Yeast Mitochondria*

Some yeast strains possess a sequence-specific en- donuclease, Endo.Sce1, which is a heterodimeric enzyme localized in mitochondria. The larger subunit (75 kDa) of Endo.Sce1, encoded by a nuclear gene (ENSI), is transported from the cytosol into the mitochondria. In this study, we determined the partial amino acid sequence of the smaller subunit (50 kDa) of Endo.Sce1. The determined sequence matched well the partial se- quence deduced from a mitochondrial open reading frame (RF3). The RF3 locus is known to exhibit poly- morphism since this reading frame in some yeast strains is supposed to encode a maturase-like protein, whereas in other strains, the frame is interrupted by GC clusters, which thus break the frame. Southern blot analysis of various yeast strains showed that the continuity of RF3 is correlated with the presence of Endo.Sce1 activity. These data indicate that the continuous RF3 sequence is a functional gene (ENS2) coding for the smaller subunit of Endo.Sce1. The results of cytoduction, by which the continuous RF3 sequence was transferred into a yeast strain lacking mitochondrial DNA, confirmed this conclusion. This study sug- gests the involvement of Endo.Sce1 in genetic recombination of mitochondrial DNA.

There is a growing amount of evidence showing that sequence-specific endonucleases which cause double-strand breaks are involved in site-specific gene conversion (Szostak et al., 1983;Colleaux et al., 1986;Xiong and Eickbush, 1988;Wenzlau et al., 1989;Delahodde et al., 1989). In addition, recent data showed that a double-strand break also appears at the initiation sites for general recombination (Nicolas et al., 1989;Sun et al., 1989).
In Saccharomyces cereuisiue, three sequence-specific endonucleases have been identified as initiators of the site-specific gene conversion process: HO-endonuclease for mating type switching (Kostriken et al., 1983) and w -and a14-endonucleases for mitochondrial intron propagation (Jaquier and Dujon, 1985;Macreadie et al., 1985;Wenzlau et al., 1989;Delahodde et al., 1989). These endonucleases display several common enzymatic properties (Colleaux et al., 1986;Wenzlau et a t , 1989;Delahodde et al., 1989). First, they show complex sequence specificities, unlike prokaryotic restriction endonucleases, because the sequences recognized and required for diges-* This work was supported in part by grants from the RIKEN on Life Science (t.0 T. S.) and from the Frontier Research Program for Plant Biological Regulation (to T. S. and N. M.) and by a grant for special research promotion (to N. M.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore he hereby marked "aduertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
$ To whom correspondence should be addressed. tion are longer than 18 base pairs with no apparent dyad symmetry. Second, DNA is cleaved by these enzymes to create cohesive ends with 4-base 3"overhangs.
Endo.Sce1, purified from S. cereuisiae IAM4274 (Watabe et al., 1981(Watabe et al., , 1983, was the first example of a eukaryotic sequence-specific endonuclease which causes in vitro doublestrand scission at well-defined sites on DNA. The cleavage mode of Endo.Sce1 is similar to those of the other sequencespecific endonucleases described above, and its cutting sites are expected to be located at every several thousand base pairs on the genome, suggesting that this enzyme is involved in general genetic recombination in yeasts . Endo.SceI has been revealed to be a heterodimer of 75and 50-kDa subunits, both of which are required for full enzymatic activity (Watabe et al., 1983;Nakagawa et al., 1988).' In our previous study, the 75-kDa subunit was shown to be encoded by an essential nuclear gene, ENSI, which is a member of the 70-kDa heat shock protein family (Morishima et al., 1990). The sequence data suggested that the 75-kDa subunit is not a catalytic subunit and that the 50-kDa subunit thus bears the active site of the endonuclease. This conjecture has been confirmed by the results of recent biochemical analyses.' The 75-kDa subunit has an N-terminal leader sequence which is a potential targeting signal for mitochondrial localization of the protein. As expected from this result, both the 75-kDa subunit and the Endo.Sce1 activity were detected almost exclusively in the mitochondria, indicating the presence of the 50-kDa subunit in the mitochondria (Morishima et al., 1990). T o elucidate the molecular properties of the 50-kDa subunit, particularly its primary sequence and subcellular localization, we identified and analyzed the gene for the 50-kDa subunit.
Yeast Strains and Microbial Techniques-The yeast strains used in this study are listed in Table  I. Yeast media were essentially as described by Sherman et al. (1986).
Cytoductants were constructed by the method of Lancashire and Mattoon (1979) as follows. First, a po strain (N548-1) was isolated from a karl strain (deficient in nuclear fusion during mating), N548, by treatment of the cells with ethidium bromide (Sherman et al., 1986). The absence of mitochondrial DNA was checked by Southern blot analysis of total DNA using purified mitochondrial DNA from S. cereuisiae CG379, fragmented by EcoRI digestion and radiolabeled K. Kawasaki, M. Takahashi, T. Ando, and T. Shibata, unpublished observations. 1977  , as a probe. The po strain (his4) was crossed with the a-type haploid progenies of S. cereuisiae IAM4274, MT371a-6B, or X1049-9C for the generation of cytoductants without nuclear fusion according to Lancashire and Mattoon (1979). Since NCYC74 is a MATnIMATa diploid strain, mitochondria from NCYC74 were transferred into N548-1 by protoplast fusion (Spencer et al., 1989). Four isonuclear karl cytoductants, CD548-1, CD548-2, CD548-3, and CD548-4, harboring the mitochondria from IAM4274, NCYC74, MT371a-6B, and X1049-9C, respectively, were isolated. Selection of these strains with an identical nuclear background was based on their his4 phenotype. The genotypes of mitochondrial DNAs from the cytoductants were examined by Southern blot analysis of EeoRI/HapII double digests of the mitochondrial DNA as described below. KARI cytoductants (CD6-1, CD6-2, CD6-3, and CD6-4) were obtained by second cytoduction of the karl cytoductants with CG378-1, a po derivative of CG378 (HIS+), with haploid progenies of the HIS+ phenotype being isolated. Diploid strains were constructed by mating of the KARI cytoductants with a po derivative of CG379.
E. coli cells were transformed essentially as described by Hanahan (1983) using strain JM109. Minipreps of plasmid DNAs were obtained according to the method of Holmes and Quigley (1981).
N-terminal Sequence Analysis of Smaller Subunit-For N-terminal sequence analysis, proteolytic fragments of the smaller subunit were obtained by in situ digestion on a nitrocellulose membrane according to the method of Aebersold et al. (1987) as follows. Purified Endo.Sce1 (-400 pmol)  was applied to an SDS2-polyacrylamide gel (Laemmli, 1970) for the separation of the smaller subunit from the larger one, followed by electrophoretic transfer onto a Trans-Blot nitrocellulose membrane (Bio-Rad). SDS (final concentration: 0.01%) was included in the transfer buffer (25 mM Tris-HC1 (pH 8.3), 192 mM glycine, and 10% (v/v) methanol) because the smaller subunit was not efficiently transferred in the presence of methanol. The smaller subunit on the membrane was located by Ponceau S (0.1%) staining and subsequently digested with a lysine-specific endopeptidase (Wako Pure Chemicals, Tokyo) in 10 mM Tris-HC1 (pH 7.8)/ acetonitrile (95:5, v/v) at 37 "C for 16 h. The molar ratio (protease/ 2The abbreviations used are: SDS, sodium dodecyl sulfate; kb, kilobase pair(s). smaller subunit) was 1:25. Proteolytic fragments were eluted from the nitrocellulose membrane by the addition of 10% trifluoroacetic acid (final concentration: 3.5%). Mixtures of peptides were separated with a Model 130A separation system (Applied Biosystems, Inc.) equipped with an RP-300 microbore reverse-phase column (2.1 X 30 mm). The following buffer systems were used buffer A (0.1% trifluoroacetic acid) and buffer B (70% (v/v) acetonitrile in 0.1% trifluoroacetic acid). A linear gradient was formed, over 45 min, between 0% (buffer A) and 60% acetonitrile in trifluoroacetic acid at a flow rate of 100 rl/min. Elution of peptides were monitored by their absorbance at 215 nm, and peptide-containing fractions were collected manually for amino acid sequence analysis.
Peptide Sequence Analysis-Amino acid sequence analysis was performed with an Applied Biosystems 470A Protein Sequencer.
Computer Analysis-The partiai amino acid sequence of the smaller subunit was used to search for homology using the LSRCHP program (Wilbur and Lipman, 1983) at the National Biomedical Research Foundation Protein Identification Resource (Release 23.0).
Oligonucleotide Probe-An oligonucleotide probe was designed based on a partial amino acid sequence (that of peptide 11 shown in Table 11). The sequence of the complete degenerate 20-mer comprised the first letter for the glutamic acid codon through the second letter for the asparagine codon as follows: Southern Hot Analysis"Southern blot analysis of mitochondrial DNA using an ohgonucleotide probe was carried out as follows. Mitochondrial DNA was prepared from S. cereuisiae IAM4274 cells by the method of Hudspeth et al. (1980). The DNA was digested with restriction enzymes, and then the digestion products were subjected to agarose gel electrophoresis for Southern blot analysis (Southern, 1975). DNAs were blotted onto a Magnagraph nylon membrane (Micron Separations) as described by the manufacturer. The blotted membrane was hybridized to the 5'-end-labeled oligonucleotide in 6 X SSPE (SSPE: 10 mM NaH2P04 (pH 7.7), 1 mM EDTA, and 0.18 M NaCl). 10 x Denhardt's solution (0.2% Ficoll 400 (Pharmacia LKB Biotechnology Inc.), 0.2% polyvinylpyrrolidone, and 0.2% bovine serum albumin), and 1% SDS at 39°C for 15 h. After hybridization, the membrane was washed twice with 6 X SSPE and 0.1% SDS at room temperature, and finally with the same solution at 39 "C for 1.5 min.
For examination of the continuity of the ENS2 frame, total DNAs were prepared from cells of various yeast strains according to the method described by Sherman et al. (1986). The DNAs were digested with EcoRI and Hap11 and then subjected to 2% agarose gel electrophoresis with a Tris/borate/EDTA buffer system (Maniatis et al., 1982). The digests were blotted onto a BA85 nitrocellulose membrane (Schleicher & Schuell) and then probed with the 1.7-kb EcoRI fragment of mitochondrial DNA radiolabeled with [n-32P]dCTP using the Multiprime labeling kit. Hybridization was performed at 65 "C overnight, and the membrane was washed with 0.5 X SSC (SSC: 0.15 M sodium chloride containing 15 mM sodium citrate) and 0.1% sodium lauroylsarcosine at 65 "C.
DNA Sequencing-A 1.7-kb EcoRI fragment of mitochondrial DNA from S. cerevisiae IAM4274 was cloned into phagemid pUC118. A series of truncated clones was produced by DraI partial digestion or exonuclease I11 treatment (Henikoff, 1984) of the cloned fragment. Sequencing was performed by the chain termination method (Sanger et al., 1977;Messing et al., 1981) with L Y -~~S -~C T P and a 7-deaza sequencing kit (Takara Biomedicals).
Enzyme Assays-For assaying the sequence-specific endonucleases from various yeast strains, cell-free extracts were prepared with a French press (Morishima et al., 1990). The extracts were loaded on a TOSOH high performance liquid chromatography column (TSK G3000SW, 7.5 x 600 mm). The column was eluted with 50 mM Tris-HCl (pH 7.5), 1 mM EDTA, 0.2 M ammonium sulfate, 1 mM phenylmethylsulfonyl fluoride, 10 mM P-mercaptoethanol, and 10% glycerol at 0 'C. Fractions containing proteins with apparent molecular masses of 120-130 kDa were pooled and then dialyzed against 50 mM Tris-HC1 (pH 7.5), 1 mM EDTA, 1 mM phenylmethylsulfonylfluoride, 10 mM B-mercaptoethanol, and 50% glycerol prior to the assay. EndoSceI activity was measured using phage @lo5 DNA as a substrate as previously described .

RESULTS
Amino Acid Sequence of Smaller Subunit-The smaller subunit from purified Endo.Sce1 was prepared for amino acid sequence analysis . The smaller subunit (50 kDa) was well separated from the larger subunit (75 kDa) by SDS-polyacrylamide gel electrophoresis. N-terminal sequence analysis of the smaller subunit blotted from an SDSpolyacrylamide gel onto a membrane, however, was unsuccessful, probably because the N terminus was blocked. To obtain proteolytic fragments for amino acid sequencing, the smaller subunit was subjected to in situ digestion on a nitrocellulose membrane (Aebersold et al., 1987) with a lysinespecific protease. After elution of the mixture of proteolytic fragments from the membrane, eight fragments were purified by high performance liquid chromatography (Fig. 1). These fragments were selected because they showed symmetrical elution profiles, indicating their purity as high enough for the sequence analysis. The primary sequences of these peptides were determined by automated Edman degradation. Table I1 shows the partial amino acid sequences of the proteolytic fragments. These sequences were used to search for homology using the LSRCHP program (Wilbur and Lipman, 1983) at the National Biomedical Research Foundation Protein Identification Resource. All the sequences completely matched a probable maturase-like protein, which is encoded by a mitochondrial open reading frame (RF3). RF3 is a part of a multigenic transcription unit and is located downstream of oli2 (subunit 6 of the mitochondrial ATPase) (SBraphin et FIG. 1. Reverse-phase high performance liquid chromatography of proteolytic digest of SO-kDa subunit. A proteolytic digest of the 50-kDa subunit was loaded onto an RP-300 reversephase column (2.1 X 30 mm) and eluted with a 0-60% gradient of acetonitrile in 0.1% trifluoroacetic acid 0-5.2 min (0%) and 5.2-50.2 min (0-60%). The elution of peptides was monitored with regard to the absorbance at 215 nm. The flow rate was 100 pllmin. The fractions collected are indicated by the numbers. ' Numbers correspond to the purified peptides shown in Fig. 1. bThe sequences of proteolytic fragments of the 50-kDa subunit are in single-letter notation. X in the sequences represents an undetermined amino acid.
The locations indicated were determined through translation of the RF3 sequence (SBraphin et al., 1987) and were confirmed by this study (Fig. 3).
al., 1987). These partial sequences, in addition, are preceded by lysines in the whole sequence of the maturase-like protein (SBraphin et al., 1987). We tentatively concluded therefore that the smaller subunit of Endo.Sce1 is a maturase-like protein encoded by RF3 in the mitochondrial genome.
Southern Blot Analysis of Yeast Mitochondrial DNA with Mixed Oligonucleotide Probe-RF3 was physically mapped on a 1.7 kb-EcoRI fragment of mitochondrial DNA from several yeast strains (SBraphin et al., 1985(SBraphin et al., , 1987. However, some strains do not possess RF3 (S6raphin et al., 1987). To identify the RF3 sequence (probable BO-kDa subunit gene) within mitochondrial DNA from S. cereuisiue IAM4274, from which Endo.Sce1 was purified (Watabe et al., 1983), Southern blot analysis of mitochondrial DNA was carried out. On the basis of information obtained by amino acid sequence analysis, an oligonucleotide probe was prepared by chemical synthesis. The probe was a completely degenerate 20-mer including both the nuclear and mitochondrial codons, based on the primary sequence of peptide 11 (Table 11). The probe hybridized to a single DNA fragment obtained by EcoRI single or EcoRIIPstI double digestion of the purified mitochondrial DNA (1.7 and 1.2 kb, respectively) (Fig. 2). The sizes of the hybridized bands are consistent with the restriction map of mitochondrial RF3, which contains a unique PstI site (SBraphin et al., 1987). The results of Southern analysis indicate the presence of the gene for the smaller subunit of Endo.Sce1 (designated as ENS2) in mitochondrial DNA from strain IAM4274.
Nucleotide Sequence of Probable ENS2 Gene"RF3 in the mitochondrial genome has been shown to exhibit polymorphism among yeast strains because a Saccharomyces uuarum strain possesses a continuous RF3 sequence, whereas in some other yeast strains, the reading frame is interrupted by GC clusters, which thus break the frame (SBraphin et al., 1987). To confirm that S. cereuisiue IAM4274 possesses a functional RF3 sequence, we cloned and sequenced the 1.7-kb EcoRI fragment of mitochondrial DNA, which was found to hybridize to an oligonucleotide probe, as stated above. Fig. 3 shows the nucleotide sequence of the EcoRI fragment and the predicted amino acid sequence. The EcoRI fragment consisted of 1671 base pairs with two Ado deletions, compared to the equivalent region of S. uuarum NCYC74 (SBraphin et al., 1987), in Ado stretches within the 5"noncoding region. As expected, the fragment contained a continuous RF3 sequence, a probable ENS2 gene, with no GC clusters, which encodes a polypeptide of 476 amino acid residues with a molecular mass of 58 kDa. The coding sequence is almost identical to that of the continuous RF3 sequence reported by SBraphin et al. FIG. 2. Southern blot analysis of mitochondrial DNA with oligonucleotide probe. Mitochondrial DNA from S. cereuisiae IAM4274 was digested with EcoRI ( l a n e a ) and EcoRIIPstI (hne b) and then subjected to a 1.2% agarose gel electrophoresis. The digests were blotted onto a nylon membrane and then hybridized with a '*Plabeled oligonucleotide mixture. Autoradiographs of the hybridization bands are presented. The lengths (in kilobase pairs) of the hybridization bands are indicated to the right. are indicated in the shaded bones. In the 5*-upstream region of ENS2, the 3'-end of oli2 corresponding to the last 2? amino acid residues is also indicated. derscoring in Fig. 3; Guo'~' for Ado in NCYC74, Guo"~ for Ado, Guoao8 for Ado, Cyd'043 for dThd, Ado11g4 for Guo, dThdlZ4" for Cyd, and Ado14'jg for Cyd). Of these point mutations, three caused changes of two amino acids, Gly217 (GGA in S. cerevisiae IAM4274) for Lys (AAA in S. uuarum NCYC74) and Asn346 (AAT) for Asp (GAT).
Correlation between Continuity of ENS2 Frame and Presence of Endo.SceI Activity-In our previous survey for sequence-specific endonucleolytic activity, yeast strains, including Saccharomyces and Pichia, displayed strain polymorphism with regard to Endo.SceI activity (Watabe et al., 1981). The larger subunit of Endo.SceI is encoded by a nuclear gene, ENSl, which is indispensable for cell growth (Morishima et al., 1990). The gene for the smaller subunit, ENSZ, on the other hand, was strongly suggested to be identical to RF3 by the results of our amino acid sequence and Southern blot analyses. Considering that RF3 exhibits strain polymorphism, it is reasonable to suppose that the EndoSceI activity of yeasts is dependent on the presence of a continuous RF3 sequence. To clarify the relationship between Endo.SceI activity and the organization of the RF3 region, we classified several strains of S. cereuisiae on the basis of the RF3 polymorphism and measured the Endo.SceI activity in extracts of cells from these yeast strains. Sbraphin et al. (1987) classified the organization of the RF3 region on the basis of two criteria: (i) the presence or absence of RF3 in the mitochondrial genome and (ii) the presence or absence of GC clusters in the RF3 sequence. Therefore, S. cereuisiae could be classified into three allelic groups with regard to RF3: group A, RF3 devoid of GC clusters; group B, RF3 containing GC clusters, and group C, lack of an RF3 sequence. IAM4274, an Endo.SceI-producing strain, belongs to group A, as demonstrated by nucleotide sequence analysis. Southern blot analysis of mitochondrial DNA after double digestion with EcoRI and HpaII, which cleaves DNA at the CCGG sequence, was shown to be useful for distinguishing these groups (S6raphin et al., 1987). The total DNAs from yeast cells were digested with EcoRI and Hap11 (an isoschizomer of HpaII) and then probed with the 1.7-kb EcoRI fragment of NCYC74 mitochondrial DNA (group A). Fig. 4 shows a typical autoradiograph obtained by Southern analysis distinguishing the three types of RF3 region. Two strains of S. cereukiae, including IAM4274, displayed the pattern of group A, in which a 1.7-kb fragment hybridized to the probe (lanes 1 and 3). As to three other strains, a 1.7-kb fragment was not detected; instead, smaller fragments (-590 and -450 base pairs (doublet) in size) hybridized to the 1.7-kb probe, with results similar to those reported by S6raphin et al. (1987) (group B) (lanes 4-6). These fragments were shown to be generated from the 1%kb EcoRI fragment through the action of Hap11 because a 1.8-kb fragment was detected with the same probe when DNA was prepared for Southern analysis by single EcoRI digestion (data not shown). For the strains showing the pattern of group C, the 1.7-kb probe did not hybridize to their mitochondrial DNAs (lanes 7-9). These three autoradiographic patterns correspond to the three allelic groups (A-C, respectively) described above.

Measurement
,of the Endo.SceI activity in these yeast strains showed that specific endonucleolytic activity was only detected in strains belonging to group A (Fig. 5, lanes 1 and  3). The profiles of fragments produced through the action of endonucleases in these strains were apparently identical to of Yeast Endonuclease Endo.SceI FIG. 5. Site-specific endonuclease activity in cell-free extracts from various yeast strains. Cell-free extracts from yeast cells were prepared with a French press. Fractions containingproteins with apparent molecular masses of 120-130 kDa were obtained from the extracts by high performance liquid chromatography and were used for the assay. Endonuclease activity was assayed using phage 6105 DNA. DNA (14 p~) was treated with cell-free extracts containing -45 pg of proteins at 37 "c for 30 min in 100 pl of 50 mM Tris-HCI (pH 7.5) containing 10 mM MgCI2 and 5 mM 8-mercaptoethanol as described by Watabe et al. (1984). The DNA digests were deproteinized by phenol extraction; and subsequently, a one-sixth of each sample was run on a 0.8% agarose gel. The lanes correspond to the cell-free extracts prepared from the strains shown in Fig. 4  each other (for unknown reasons, the substrate DNA could not be completely digested by Endo.Sce1) . These profiles were identical to the characteristic band pattern obtained with purified Endo.Sce1 (lane S ) . In addition, the specific endonucleolytic activity was detected almost exclusively in the mitochondria (data not shown), as reported previously for strain IAM4274 (Morishima et al., 1990). This result indicates the presence of Endo.Sce1 in mitochondria from the group A strains. For the other strains (groups B and C), sequence-specific endonuclease activity could not be detected in the fractions corresponding to apparent molecular masses of 120-130 kDa prepared from these strains (lanes 4-9). There was also no detectable Endo.Sce1 activity in the cell-free extracts which had not been fractionated by high performance liquid chromatography (data not shown). Because there was clear correlation between the RF3 region in group A and the presence of Endo.Sce1 activity (we have confirmed the absence of Endo.Sce1 activity in another 15 strains of groups B and C), this result is consistent with our conclusion that the continuous RF3 is the functional gene (ENS2) encoding the smaller subunit of Endo.Sce1.
In a similar way, several strains of S. uvarum were analyzed with regard to the gene organization of the RF3 region and the sequence-specific endonuclease activity because RF3 of S. uvarum is also polymorphic (data not shown). Among the S. uvarum strains tested, a group A strain (NCYC74) alone (Fig.  4, lane 2 ) , like the group A strains of S. cerevisiae, was shown to exhibit sequence-specific endonuclease activity under these conditions (Fig. 5, lane 2 ) . This sequence-specific activity was also detected in isolated mitochondria (data not shown). The level of endonucleolytic activity was apparently the same as that of Endo.Sce1 from S. cerevisiae. This result suggests that the continuous RF3 sequence in S. uvarum is also a functional gene encoding a subunit of the endonuclease. The profile of fragments produced by the endonuclease from S. uvarum, however, was not identical to those in the case of S. cerevisiae group A strains (lanes 1-3). Two bands, which did not clearly appear with the Endo.Sce1 from S. cerevisiae, were detected for the digests with the S. uvarum endonuclease (indicated by arrowheads in lane 2). In contrast, an intense band was visible in the S. cerevisiae digests (indicated by arrowhead in lane 1 ), whereas there was only a faint band at the corresponding position in lane 2. These results suggest therefore that the group A strain of S. uvarum possesses another sequencespecific endonuclease(s) (probably homologue(s) of Endo.-SceI) whose sequence specificity is different from that of Endo.Sce1. It remains to be determined whether or not the endonuclease activity detected here is due to a single enzyme species.
Cytodmtion Experiment Involving ENS2 and po Strains-To confirm that the Endo.Sce1 activity is dependent on ENS2 in the mitochondrial genome and independent of nuclear genes other than ENS1 (the gene for the larger subunit of Endo.Sce1, which is essential for cell growth) (Morishima et al., 1990), we carried out mitochondrial transfer through cytoduction (Lancashire and Mattoon, 1979;Spencer et al., 1989). By this technique, mitochondria were transferred from a group A strain (possessing Endo.Sce1 activity) to a po strain (CG378-1), which lacks mitochondrial DNA and EndoSceI activity. A haploid cytoductant (CD6-1) was constructed to contain the mitochondria from S. cerevisiae IAM4274. Fig. 6 shows the results of Endo.Sce1 assaying of the cytoductant and the original strain, CG378 (p+, but having neither an RF3 sequence nor Endo.Sce1 activity, as shown in Fig. 4 (lane 9 ) and Fig. 5 (lane 9 ) , respectively). Although these two strains were isonuclear with regard to each other, Endo.Sce1 activity was detected in CD6-1, but not in CG378 (Fig. 6, lanes 9, 13,  and 1 ) . In a similar way, another cytoductant, CD6-2, was constructed using CG378-1 and S. uvarum NCYC74. Consistent with the results of sequence-specific endonuclease assaying of NCYC74, Endo.Sce1-like activity was detected in CD6-2 (lanes 14 and 2 ) . Essentially the same results were obtained with the use of diploid strains harboring mitochondria of either IAM4274 or NCYC74 origin, excluding the possibility that the ploidy of cells affects the Endo.Sce1 activity (see "Experimental Procedures"). Negative control experiments showed that neither group B (CD6-3) (lane 15) nor group C (CD6-4) (lane 16) mitochondria yielded sequence-specific endonuclease activity in the CG378-1 strain. These results, of Yeast Endonuclease Endo.SceI together with the results described above, demonstrated that the occurrence of Endo.Sce1 (or its homologue) activity is dependent on ENS2 in the mitochondrial genome.

DISCUSSION
We obtained consistent results from amino acid sequencing, Southern blot analysis, and cytoduction experiments demonstrating that the smaller subunit of Endo.Sce1 is a maturaselike protein encoded by a mitochondrial open reading frame (RF3). The open reading frame contains two conserved dodecamer sequences which have been observed in several maturases (Michel et ul., 1982;Waring et ul., 1982;Hensgens et al., 1983), that is, Tyr203 through Ile2I4 and Trp320 through Phe3" (Fig. 7). Maturases, encoded by mitochondrial intron reading frames, have been identified as essential factors in vivo for splicing of the respective introns by which each maturase is encoded (Lazowska et al., 1980;De La Salle et al., 1982;Anziano et al., 1982;Carignani et al., 1983). Recent work has shown that the maturase encoded by intron 4 of the yeast cytochrome c oxidase gene (aI4) is also a sequencespecific endodeoxyribonuclease which could be involved in the mobility of introns at the DNA level (Wenzlau et al., 1989;Delahodde et al., 1989). This finding raises the possibility that some other maturases and maturase-like proteins having conserved dodecamer sequences could have sequence-specific endonuclease activity. This study revealed another example of a sequence-specific endonuclease containing conserved motifs. Fig. 7 is a comparison of the conserved sequences in the deduced amino acid sequence of the 50-kDa subunit with those of yeast endonucleases: HO-endonuclease (Russell et al., 1986), w-endonuclease (Dujon, 1980), and aI4-endonuclease (Bonitz et al., 1980). These three sequence-specific endonucleases, like Endo.Sce1, recognize DNA sequences longer than 18 base pairs with no obvious dyad symmetry and make staggered cuts in their substrate DNAs, producing 4-base 3'overhangs (Kostriken et al., 1983;Shibata et al., 1984;Colleaux et al., 1988;Wenzlau et d., 1989;Delahodde et d., 1989).  (Dujon, 1980), and aI4-endonuclease (Bonitz et al., 1980). The numbers next to the amino acids (a.a.) represent the positions in each amino acid sequence.
The positions of conserved sequences in aI4-endonuclease protein are not known yet because the endonuclease is supposed to be generated from a precursor protein by proteolysis (Wenzlau et al., 1989). The consensus sequences for the two motifs are shown at the bottom. Amino acid residues which matched the consensus sequences are shaded. Note that HO-endonuclease possesses the motifs in reversed order (see the text).
In the case of HO-endonuclease, the second motif could not be readily found; instead, a sequence displaying weak homology to the second motif was found at -100 residues on the Nterminal side of the first motif. Therefore, the structural domain of HO-endonuclease may be distinct from those of other endonucleases.
The molecular mass of the smaller subunit calculated from its deduced amino acid sequence (58 kDa) is slightly larger than the apparent molecular mass of the subunit determined by SDS-polyacrylamide gel electrophoresis (50 kDa) (Watabe et al., 1983). The difference between the calculated and measured molecular masses could be explained by the fact that the 50-kDa subunit shows anomalously high mobility on SDSpolyacrylamide gel, as do other mitochondrial proteins (Groot et al., 1978;Nobrega and Tzagoloff, 1980;Weiss-Brummer et al., 1982). The proteolytic fragments of the subunit that we identified by amino acid sequence analysis are dispersed within the region from Asp32 to Lys4'j1 in the deduced sequence (Fig. 3). Therefore, the possibilities that processing of the de novo synthesized subunit occurs and/or that the translation of ENS2 mRNA is initiated at codon 12 for the second methionine cannot be excluded.
As previously suggested by us (Morishima et al., 1990), this study strongly supports the idea that the 50-kDa subunit bears the active site of the endonuclease because the subunit contains the conserved sequence found in maturase-related endonucleases. Supporting data have recently been obtained from biochemical analysis of Endo.Sce1.' An apparent difference between the 50-kDa subunit of Endo.Sce1 and other sitespecific endonucleases from yeasts is that the subunit alone exhibits little endonuclease activity. The results of our previous biochemical and immunochemical experiments (Watabe et al., 1983;Nakagawa et al., 1988) showed that the smaller subunit (maturase-like protein) exhibits full enzymatic activity when the protein is associated with a larger subunit of 75 kDa, which has now been revealed to be a nuclear-encoded heat shock protein (Morishima et al., 1990). Therefore, the 75-kDa subunit is likely required to function as, for example, a modulator of the protein structure of the 50-kDa subunit for the expression of full enzymatic activity.
In mitochondria from S. uuarum NCYC74, we detected a new sequence-specific endonuclease activity which is apparently different from the Endo.Sce1 activity in its sequence specificity (Fig. 5). Since both S. cereukiue and S. uvarum exhibit mitochondrial sequence-specific endonuclease activity A Maturase-like Subunit of Yeast Endonucleose Endo.SceI which is dependent on or correlated at least with the continuity of the ENS2 reading frame, it is possible that the endonuclease encoded by the reading frame in S. uuarum has a different sequence specificity from that of Endo.SceI. On comparison of the deduced amino acid sequences of ENS2 in S. cerevisiae IAM andRF3 in S. uvarum NCYC74 (Skraphin et al., 1985, 1987), two amino acid changes were observed (Gly217 and Asn346 in the former for LyP7 and Asp346 in the latter). The mutations at these sites, which are close to the conserved dodecamer sequences described above, could alter the sequence specificity of the endonuclease.
As to a genetic study on the 75-kDa subunit, this work reveals an apparent discrepancy as to the requirement of EndoSceI for the growth of yeasts. A null mutation in ENS2 (groups B and C in Figs. 5 and 6) does not affect cell viability, whereas ENSl, which encodes the 75-kDa subunit, is essential for growth (Morishima et al., 1990). The tester strain we used in a previous genetic study (diploid constructed through the mating of CG378 and CG379) has null alleles of ENSB, as judged by Southern blot analysis (Fig. 5).3 This suggests that the Endo.SceI activity is not essential, but that the 75-kDa subunit (heat shock protein) plays another unidentified role, i.e. besides functioning as a subunit of Endo.SceI. Yeast mitochondria contain several maturases, among which was found an essential protein for the splicing of the introns in the cytochrome oxidase subunit I gene (Labouesse and Slonimski, 1983). There are also two maturase-like sequences other than RF3 within the mitochondrial genome (named RF1 and RFZ) (Coruzzi et al., 1981;Michel, 1984). It is therefore possible that the 75-kDa protein is associated with some of these proteins. To examine this possibility, we have initiated a survey for such protein(s).
Whether Endo.SceI has a latent maturase activity or not remains to be examined. Unlike maturases, however, the 50-kDa subunit of Endo.SceI is not encoded within an intron of another gene (this study and S6raphin et al., 1987). A possible target RNA for Endo.SceI is not apparent. Considering that aI4-endonuclease, which cleaves an intron at the DNA level, has latent maturase activity for the intron at the RNA level (Delahodde et al., 1989;Wenzlau et al., 1989), information about the DNA sequence specificity of Endo.SceI would be suggestive. In vitro cleavage experiments on mitochondrial DNA have shown that Endo.SceI has many cleavage sites (possibly at every 2-3 kb) on the yeast mitochondrial DNA.4 This result indicates that Endo.SceI-cleavable sites in mitochondrial DNA appear over the length of the mitochondrial genome (discussed below). Therefore, a unique RNA could not be readily deduced as a target for Endo.SceI.
Our present knowledge suggests that Endo.SceI plays a role in the genetic recombination of mitochondrial DNA from yeasts. As to its biochemical characteristics, the mode of cleavage of Endo.SceI is similar to that of HO-endonuclease and w-endonuclease (Kostriken et al., 1983;Shibata et al., 1984;Colleaux et al., 1986). The latter two enzymes are involved in the gene conversion process of the mating type locus and rRNA locus, respectively. In addition, aI4-endonuclease, which is supposed to be involved in the propagation of an intron through a unidirectional gene conversion, also exhibited a similar mode of cleavage (Wenzlau et al., 1989;Delahodde et al., 1989). Therefore, Endo.SceI could cleave mitochondrial DNA at the site(s) where genetic recombination occurs. This work provides further supporting data, on a ' K. Nakagawa, N. Morishima, and T. Shibata, unpublished observations. ' N. Morishima, K. Nakagawa, and T. Shibata, unpublished observations. molecular basis, indicating that Endo.SceI possesses a function related to that of other sequence-specific endonucleases. There is a significant difference, however, between the enzymatic characteristics of EndoSceI and other endonucleases. Unlike HO-and o-endonucleases, which are initiators of sitespecific gene conversion, Endo.SceI has multiple recognition sequences (Shibata et aZ., 1984) and actually exhibits multiple digestion of yeast mitochondrial DNA in vitro (as described above), suggesting that it is involved in general recombination. As a step to identify the target sequences for Endo.SceI in vivo, we are trying to identify the cleavage sites within the mitochondrial DNA for the endonuclease.