Hmi1p from Saccharomyces cerevisiae mitochondria is a structure-specific DNA helicase.

Hmi1p is a Saccharomyces cerevisiae mitochondrial DNA helicase that is essential for the maintenance of functional mitochondrial DNA. Hmi1p belongs to the superfamily 1 of helicases and is a close homologue of bacterial PcrA and Rep helicases. We have overexpressed and purified recombinant Hmi1p from Escherichia coli and describe here the biochemical characteristics of its DNA helicase activities. Among nucleotide cofactors, the DNA unwinding by Hmi1p was found to occur efficiently only in the presence of ATP and dATP. Hmi1p could unwind only the DNA substrates with a 3'-single-stranded overhang. The length of the 3'-overhang needed for efficient targeting of the helicase to the substrate depended on the substrate structure. For substrates consisting of duplex DNA with a 3'-single-stranded DNA overhang, at least a 19-nt 3'-overhang was needed. In the case of forked substrates with both 3'- and 5'-overhangs, a 9-nt 3'-overhang was sufficient provided that the 5'-overhang was also 9 nt in length. In flap-structured substrates mimicking the chain displacement structures in DNA recombination process, only a 5-nt 3'-single-stranded DNA tail was required for efficient unwinding by Hmi1p. These data indicate that Hmi1p may be targeted to a specific 3'-flap structure, suggesting its possible role in DNA recombination.

Helicases are enzymes that utilize the free energy derived from the hydrolysis of nucleoside triphosphates to unwind duplex RNA or DNA. All of the organisms contain several helicases involved in the processes of nucleic acid metabolism, such as replication, transcription, translation, recombination, and so forth (1)(2)(3)(4).
All of the DNA helicases found in bakers' yeast Saccharomyces cerevisiae are encoded in the nucleus. To date, two of them have been shown to be involved in mtDNA 1 metabolism, Pif1p and Hmi1p (5)(6)(7). Both belong to the helicase superfamily 1 but to different subfamilies (8). Pif1p is conserved among eukaryotes, as homologues have been described in other yeasts, Caenorhabditis elegans, Drosophila melanogaster, and human (9). Pif1p is a 5Ј-3Ј-directional distributive DNA helicase that is functional in the monomeric state and is preferentially recruited to forklike DNA structures (10). Deletion of Pif1p affects the stability of mtDNA upon exposing to UV light and at 37°C but not at 28°C (11)(12)(13). The role of Pif1p in mtDNA metabolism has been suggested to be in recombination and repair of mtDNA (11)(12)(13).
The second DNA helicase described in yeast mitochondria, Hmi1p, is a close homologue of the bacterial Rep, UvrD, and PcrA helicases (5). To date, in silico analyses of eukaryotes reveal Hmi1p homologues only in evolutionarily close yeasts from the family of Saccharomycetaceae. The C terminus of Hmi1p contains a mitochondrial targeting signal peptide that is cleaved off after the protein is transported into mitochondria (14). Deletion of Hmi1p has a more severe effect on mtDNA metabolism than the deletion of Pif1p, because Hmi1p-deficient mutants do not retain functional mtDNA (5). We have shown that Hmi1p also influences the rho mtDNA metabolism, because in the absence of Hmi1p, the rho mtDNA concatemers are shorter (15). This shortening of the rho mtDNA molecules is linked to transcription, because the absence of Hmi1p has a much less effect on the mitochondrial RNA polymerase Rpo41p-deficient mutant. In addition, the role of Hmi1p on the functional mtDNA has been studied using a temperature-sensitive mutant of Hmi1p. We demonstrated that the cells containing the temperature-sensitive mutant of Hmi1p lost mtDNA through fragmentation during growth under restrictive temperatures, this being similar to the effect of Hmi1p on rho mtDNA metabolism (15).
In mammalian mitochondria, another DNA helicase, Twinkle, has been described, which has no homologue in S. cerevisiae (16). The Twinkle helicase has sequence similarity to the bacteriophage T7 gene 4 protein and other hexameric helicases/ primases (17). It is a 5Ј-3Ј-directional helicase that has been proposed to be the replicative helicase of mammalian mitochondria, as it has been shown to form a minimal mtDNA replisome in vitro together with the mitochondrial DNA polymerase and the mitochondrial ssDNA-binding protein (18,19).
Despite the fact that two DNA helicases have been described in S. cerevisiae mitochondria, the question of replicative helicases in mitochondria is still unsolved. For Pif1p, a role in mtDNA metabolism has been proposed. The function of Hmi1p in this context has remained ambiguous. We have ruled out the involvement of Hmi1p in mtDNA transcription (5). Whether Hmi1p is involved in mtDNA replication is not clear, but it cannot be the sole replicative helicase because, in the Hmi1pdeficient strain, the rho mtDNA is still maintained. To further elucidate the role of Hmi1p in mitochondria, we have purified the recombinant Hmi1p from Escherichia coli. Here we describe the helicase activity of Hmi1p, its processivity, nucleotide cofactor utilization, and DNA substrate requirements.

MATERIALS AND METHODS
Constructs Used for Protein Overexpression-The previously described plasmid pGEX41-HMI1 was used for the overexpression of Hmi1p (5). To change the conserved Glu-211 to Gln, the G nucleotide in the position 631 of the HMI1 open reading frame was changed to C by cloning the HMI1 open reading frame into pUC119 phagemid vector followed by site-directed mutagenesis conducted by the method of Kunkel et al. (20). To generate the plasmid pGEX41-HMI1(E211Q) that was used for the overexpression of the mutant Hmi1p(E211Q), the mutated HMI1 open reading frame was cloned into the BamHI site of the pGEX4T1 vector. Here and elsewhere, the restriction enzymes as well as other DNA-modifying enzymes were from Fermentas. Both the WT and the mutant Hmi1p(E211Q) expression constructs were verified by sequencing.
Expression and Purification of Hmi1p and Hmi1p(E211Q)-Hmi1p and Hmi1p(E211Q) were purified essentially as described previously (5,21). A 2-3-liter culture of E. coli strain DH5␣ harboring either pGEX41-HMI1 or pGEX41-HMI1(E211Q) was grown in 2ϫ YT broth (16 g/liter tryptone, 10 g/liter yeast extract, 5 g/liter NaCl) supplemented with 100 g/ml ampicillin at 30°C until an A 600 of 0.4 was reached. The culture was cooled to 25°C, and the expression of Hmi1p was induced by the addition of isopropyl ␤-D-thiogalactoside to the final concentration of 1 mM and incubation for 10 -15 h. The cells were harvested by centrifugation and resuspended in 25 ml of buffer A (50 mM Tris-HCl, pH 8.0, 300 mM NaCl, 10% sucrose) containing 1 mM phenylmethylsulfonyl fluoride, 1 mM DTT, and 1 mM EDTA. The cell suspension was frozen in liquid nitrogen and stored at Ϫ75°C. Subsequent purification was performed at 0 -4°C. The cells were thawed and lysed for 1.5 h with 1 mg/ml lysozyme followed by one freezing-thawing cycle. After thawing, buffer A was added to 50 ml and the cells were sonicated (5 ϫ 20-s pulses separated by 1-min intervals at 70% power, Ultrasonic homogenizer CP300, Cole-Parmer Instrument Co.). The cell debris was removed by centrifugation for 20 min at 12,000 ϫ g. 30 mg of glutathione-Sepharose 4B beads (Sigma catalogue number G4510) preequilibrated in buffer A were added to the extract and incubated for 3-6 h with gentle mixing. The beads were poured into a 5-ml plastic column and washed consecutively with 10 ml of buffer B (20% glycerol, 20 mM Tris-HCl, pH 8.0, 300 mM NaCl, 0.1 mM EDTA, 1 mM DTT), 2 ml of buffer B with 0.1% Nonidet P-40, 2 ml of buffer B, 2 ml of buffer B adjusted to 1 M NaCl, and again with 2 ml of buffer B. The protein was eluted with buffer B containing 10 mM glutathione. The fractions containing Hmi1p (2-4 mg of total protein) were pooled together and adjusted to 2.6 mM CaCl 2 , and the GST tag was cleaved with 40 units of thrombin (Sigma catalogue number T6634) by incubating overnight on ice. The protein concentration was determined here and elsewhere by the method of Bradford (22) using BSA as a standard. The sample was diluted 3-fold with buffer B without NaCl and loaded onto a 1-ml Q-Sepharose column equilibrated with buffer B containing 100 mM NaCl. Hmi1p was found to be in the Q-Sepharose flow-through that was further loaded onto a 0.3-ml S-Sepharose column equilibrated with buffer C (25 mM Mes-NaOH, pH 6.5, 20% glycerol, 0.1 mM EDTA, 1 mM DTT, 1 mM phenylmethylsulfonyl fluoride) containing 100 mM NaCl. The column was washed with buffer C containing 150 mM NaCl, and Hmi1p was eluted with buffer C containing 400 mM NaCl. Aliquots of the peak fraction were frozen in liquid nitrogen and stored at Ϫ75°C. The protein concentration in the peak fractions was estimated to ϳ0.1-0.4 mg/ml. The molarities of Hmi1p were calculated from the estimated protein concentrations and the approximate molecular mass of Hmi1p without the 15 C-terminal amino acids (78.5 kDa).
Sedimentation in Glycerol Gradient-A 200-l aliquot of Hmi1p (80 g) was loaded on a 5-ml 10 -30% glycerol gradient containing 25 mM Tris-HCl, pH 8.0, 500 mM NaCl, 0.1 mM EDTA, 0.1 mM DTT, and 0.1 mM phenylmethylsulfonyl fluoride and centrifuged in a Sorwall SW 55 rotor at 40,000 rpm for 13 h at 4°C. 26 fractions of 200 l were collected from the top of the tube. The protein concentration in the fractions was determined, and aliquots of the fractions were analyzed by SDS-PAGE. The fractions were also analyzed for the ATPase activity using the charcoal assay as described below. BSA, ovalbumin, aldolase, and catalase were run in parallel gradients as molecular mass markers.
To determine the kinetic parameters for ATP hydrolysis, the ATPase activity was assayed by linking the ADP formation to the oxidation of NADH as described previously (23). The assay mixture (0.5 ml) contained 30 mM Tris-HCl, pH 7.5, 7 mM MgCl 2 , 1 mM DTT, 0.1 mg/ml BSA, 0.3 M oligonucleotide G as ssDNA cofactor (see Table I), 40 units/ml pyruvate kinase (from rabbit muscle, Roche Applied Science), 40 units/ml lactate dehydrogenase (Calbiochem), 4 mM phosphoenol pyruvate, and 0.3 mM NADH. Six independent reactions were carried out at each of four different ATP concentrations used (1, 0.2, 0.1, and 0.05 mM). Samples were prewarmed for 3 min at 30°C, and the reaction was initiated by adding Hmi1p to the final concentration of 0.03 M. The hydrolysis of ATP was monitored by the decrease in absorbance at 340 nm using a UV-visible spectrophotometer (Shimadzu UV-1601PC). Values for the constants V max and K m were derived by nonlinear regression analysis according to the Michaelis-Menten equation.
Preparation of Substrates for Helicase Assays-The DNA sequences of all of the oligonucleotides used for preparing substrates are shown in Table I. Oligonucleotides annealed to give various substrates are listed in Table II. The annealing reaction mixture (10 l) containing 0.5-1 M oligonucleotides and buffer Y (33 mM Tris acetate, pH 7.9, 10 mM magnesium acetate, 66 mM potassium acetate) was incubated for 5 min at 80°C, slow-cooled to 25°C over a period of 2-3 h, and then incubated on ice for 15 min. The top strand oligonucleotides (for making substrates 2-6, 16 -21, and 23-27) and the bottom strand oligonucleotides (for making substrates 7-10) were 5Ј-end-labeled using [␥-32 P]ATP and bacteriophage T4 polynucleotide kinase before annealing to the complementary strand oligonucleotides. The top strand oligonucleotides of the substrates 1, 11-15, 22, and 28 were labeled with [␣-32 P]dCTP by using the DNA polymerase I Klenow fragment in the presence of 0.1 mM dNTPs as indicated in Table II. The substrates were purified by electrophoresis in 8% polyacrylamide gels (29:1, acrylamide:bisacrylamide) using 45 mM Tris, 45 mM boric acid, and 1 mM EDTA as electrophoresis buffer. The substrates were eluted from the gel by incubating the gel slices overnight at 4°C in 10 mM Tris-HCl, pH 8.0, 100 mM NaCl, and 0.1 mM EDTA. For preparing substrate 29, the oligonucleotide USP was 5Ј-end-labeled using [␥-32 P]ATP and bacteriophage T4 polynucleotide kinase and annealed to phagemid pUC119 ssDNA as described above.
To get a substrate with heterogeneous length of duplex DNA, a DNA synthesis reaction was performed in the presence of 0.5 mM dNTPs and 0.1 mM ddGTP using T7 DNA polymerase. After 5-min synthesis at 37°C, the reaction product was extracted with phenol/chloroform (1:1 v/v) and purified by gel filtration on a 1-ml Sephacryl S-400 column. Substrate 30 was prepared by annealing 5Ј-end-labeled oligonucleotide C to phagemid pUC119 ssDNA as described above. The reaction mixture was extracted with phenol/chloroform (1:1 v/v), and the DNA was precipitated with ethanol, resuspended in 10 l of buffer Y, and cut with 5 units of SmaI for 3 h. 0.1 mM dNTP and [␣-32 P]dCTP were added as indicated in Table II, and the 3Ј end of the oligonucleotide was extended by using the DNA polymerase I Klenow fragment. The substrate was extracted with phenol/chloroform (1:1 v/v) and purified by gel filtration on a 1-ml Sephacryl S-400 column.
Helicase Assay-The helicase assay mixture (10 l) contained 1.5 nM substrate DNA (this corresponds to 70 -160 nM in nucleotides for substrates made of oligonucleotides and ϳ3000 nM in nucleotides for substrates based on pUC119 ssDNA), 4 mM ATP, 30 mM Tris-HCl, pH 7.5, 7 mM MgCl 2 , 1 mM DTT, 0.1 mg/ml BSA, and Hmi1p as indicated. Reactions were performed at 30°C for 15 min, unless otherwise specified. The reactions were stopped by adding 1.5 l of stop solution containing 25% glycerol, 1% SDS, 0.4 M Tris, 0.4 M boric acid, 50 mM EDTA, and 0.1% bromphenol blue and loaded directly onto 12% polyacrylamide gels (29:1, acrylamide:bisacrylamide). 8% gels were used for the D loop substrates and the substrates based on pUC119 ssDNA. Electrophoresis was run at 4°C using 45 mM Tris, 45 mM boric acid, and 1 mM EDTA as electrophoresis buffer. The gels were dried and exposed to x-ray films and to Molecular Dynamics PhosphorImager screens (Amersham Biosciences) for quantitation. All of the helicase assays were repeated three times.

RESULTS
Purification of Hmi1p-Hmi1p was overexpressed in E. coli using the pGEX-4T1 expression vector that encodes for a GST affinity tag fused to the N terminus of Hmi1p. The GST tag was removed by thrombin cleavage, and Hmi1p was subjected to chromatographic fractionation by Q-Sepharose and S-Sepharose columns. The purified protein preparation separated by SDS-PAGE is shown in Fig. 1A, lane 3. In parallel with the WT Hmi1p and by the same method, the E211Q mutant of Hmi1p, Hmi1p(E211Q), was purified. Glu-211 of Hmi1p is a conserved residue in the helicase motif II (8), being a putative catalytic acid (24); thus, Hmi1p(E211Q) should be defective in regard to ATPase and helicase activities. The purified Hmi1p(E211Q) preparation separated by SDS-PAGE is shown in Fig. 1A, lane 2. Both the WT Hmi1p and the Hmi1p(E211Q) preparations contained the 75-kDa Hmi1p as the major component but also a small fraction of GST-Hmi1 fusion protein and proteins of ϳ50, 35, and 30 kDa as minor contaminations. Although 0.07-0.1 mg of WT Hmi1p was routinely purified from 2 liters of bacterial culture, only 0.02 mg of Hmi1p(E211Q) could be obtained from the same amount of the bacterial culture. We also overexpressed two other mutants, Hmi1p(K32M) and Hmi1p(R275M), but these proteins could not be purified using the method described above because the GST fusion proteins were in the insoluble fraction after cell lysis (data not shown).
ATPase and Helicase Activities of Hmi1p and Hmi1p(E211Q) Preparations-The ATPase activities of the WT Hmi1p and the mutant Hmi1p(E211Q) were examined at four different ATP concentrations in the presence and absence of an oligonucleotide ssDNA cofactor (for details see "Materials and Methods"). The WT Hmi1p preparation exhibited ssDNA-dependent ATPase activity with kinetic parameters for ATP hydrolysis: K m ϭ 0.185 Ϯ 0.037 mM and k cat ϭ 13 Ϯ 1 s Ϫ1 . The ATPase activity without ssDNA was below the limits of detection on all of the ATP concentrations used. The mutant Hmi1p(E211Q) preparation did not show detectable ATPase activity, either in the presence or absence of ssDNA (data not shown).
The helicase activities of the WT Hmi1p and the mutant Hmi1p(E211Q) were assessed in the presence and absence of ATP using substrate 1 that consisted of duplex DNA with a 3Ј-ssDNA overhang 22 nt in length (Fig. 1B). The WT Hmi1p showed ATP-dependent helicase activity, whereas ssDNA formation was not observed when the mutant Hmi1p(E211Q) was used. This result together with the result of the ATPase assay confirms that the ATPase and helicase activities of the Hmi1p preparation are properties of the Hmi1 helicase and not of any of the contaminating E. coli proteins.
The Directionality of Hmi1p-The directionality of the helicase activity of Hmi1p was determined with substrate 30 consisting of the linearized pUC119 ssDNA with 32 P-labeled oligonucleotides hybridized to both ends of the molecule. Hmi1p could unwind only the oligonucleotide from the 5Ј end of the phagemid ssDNA ( Fig. 2A). No oligonucleotide displaced from the 3Ј end of the linearized pUC119 ssDNA could be detected. This result shows that Hmi1p translocates in the 3Ј-5Ј direction on ssDNA.
The Usage of Nucleotide Cofactors-Because helicases utilize the free energy derived from the hydrolysis of nucleotides to separate DNA strands, the ability of Hmi1p to hydrolyze different nucleotides was determined using the helicase assay. Hmi1p and the 22-nt 3Ј-ssDNA overhang containing substrate 1 were incubated in the presence of different NTP(s) and dNTP(s) (Fig.  2B). The highest enzymatic activity of Hmi1p was detected in the presence of ATP and dATP, thus these nucleotides being the preferred cofactors. Incubation in the presence of dCTP and CTP resulted in Ͻ20% activity compared with ATP. None of the other nucleoside triphosphates could support the helicase activity of Hmi1p above the limits of detection.
The Processivity of Hmi1p-To estimate the processivity of Hmi1p, a heterogeneous substrate 29 was made. The substrate consisted of pUC119 ssDNA carrying oligonucleotides of different lengths synthesized in the presence of ddGTP (for details see "Materials and Methods"). The unwinding of this substrate by Hmi1p was assessed within 30 min (Fig. 3). The unwinding of 30-, 50-, 80-, and 120-nt oligonucleotides in length was quantified by phosphorimaging analysis. During a 30-min incubation, solely the 30-mer was completely unwound, whereas only ϳ60% 50-mer was displaced during the same time. The unwinding of the 80-and 120-mer was far less effective (10 -20% after 30 min of reaction). This demonstrates that Hmi1p preferentially unwinds short DNA duplexes.
Unwinding of Substrates with Different Structures by Hmi1p-Substrates with different structures were made of oligonucleotides to characterize the substrate specificity of the protein (Fig. 4). Of 10 different substrates, Hmi1p could unwind only the ones containing a 3Ј-overhang or a stretch of ssDNA. The 3Ј-overhang was either as an extension to duplex DNA (substrate 1), in a forked structure (substrate 4), or in a flap structure (substrate 6). The ssDNA stretch, which also appeared to form a sufficient structural determinant for Hmi1p loading, was either in a bubble structure (substrate 7) or in different D-loop structures (substrates 8 -10). D-loop substrates were unwound independently of the ssDNA overhangs of the invading strand, the only difference being at substrate 10 containing the invading strand with 3Ј-ssDNA overhang where the formation of bubble structure was detected (lane 28). This result shows that Hmi1p starts unwinding D-loop substrates from the 3Ј-flap structures that form in the 5Ј end of ssDNA. Hmi1p could not start unwinding from a blunt end, from a nick in one strand (substrate 3), or from a 5-nt gap in one strand (substrate 2). Consistent with the previous result that Hmi1p translocates in the 3Ј-5Ј-direction on ssDNA, Hmi1p could not unwind the substrate with a 5Ј-flap structure (substrate 5).
Minimal Length of 3Ј-ssDNA Overhang Necessary for Productive Hmi1p Recruitment-To determine the minimal required length of the 3Ј-overhang, we used substrates with 23-nt duplexes and 5-, 9-, 13-, 16-, 19-, or 22-nt 3Ј-overhangs, respectively. Hmi1p was incubated with these substrates, and the unwinding of the 32 P-labeled oligonucleotide was quantified by phosphorimaging analysis. Only substrates with 19-and 22-nt overhangs could be unwound by Hmi1p (Fig. 5). Substrates with shorter 3Ј-overhangs could not be unwound, even by 4-fold molar excess of the protein over the substrate.
Unwinding of Substrates with Forked Structures-We further examined the ability of Hmi1p to unwind different forked substrates. Three series of substrates were used: first, substrates with a 5Ј-overhang of constant length (23 nt) and the 3Ј-overhang of varying length; second, substrates with a 3Јoverhang of constant length (9 nt) and the 5Ј-overhang of varying length; and third, a set of substrates with both 5Ј-and 3Ј-overhangs of the same length (9-, 5-, 2-, and 0-nt, respectively). We found that, with the series of substrates that had 23-nt 5Ј-overhangs, a shorter 3Ј-overhang was required than in the series without the fork (Fig. 6A). Half of the maximum unwinding activity could be seen with a 2-nt 3Ј-overhang, whereas complete unwinding was reached with a 9-nt 3Ј-overhang. As demonstrated above (Fig. 5), at least a 19-nt 3Јoverhang was necessary for Hmi1p to be targeted to a substrate without fork structure. We next analyzed how the length of the 5Ј-overhang affects unwinding of the substrates with constant 9-nt 3Ј-overhangs (Fig. 6B). There was no difference between the substrates containing a 23-or a 9-nt 5Ј-overhang. However, the efficiency of unwinding of the substrates with shorter than 9-nt 5Ј-overhangs was lower. Thus, a 5-nt 5Ј-overhang could support ϳ50% of unwinding compared with a 9-nt 5Ј-overhang and a 2-nt 5Ј-overhang ϳ25%. Comparing the series in Fig. 6, A and B, the main difference could be seen between substrates with a 2-nt variable overhang. The substrate with a 2-nt 5Јoverhang was unwound only 14% by 6 nM Hmi1p, whereas the fork with a 2-nt 3Ј-overhang was unwound ϳ30% by the same amount of Hmi1p. This finding demonstrates that Hmi1p preferentially utilizes forked substrates that contain longer 5Јoverhang structures. Finally, substrates with equal length of both 3Ј-and 5Ј-overhangs were tested. Of such substrates, only the one with 9-nt overhangs could efficiently support the helicase activity of Hmi1p (Fig. 6C). The substrate with 2-nt overhangs was not unwound by Hmi1p, and the substrate with 5-nt overhangs was only 7% unwound by the 6 nM helicase. Thus, it can be concluded that Hmi1p can be targeted to much shorter overhangs of substrates with fork structures but the overhangs should be at least 9 nt in length to ensure efficient unwinding by the enzyme.
Unwinding of 3Ј-Flap-structured Substrates-The required length of the 3Ј-tail was determined also in flap-structured substrates (Fig. 7). With these substrates, even shorter 3Ј-overhangs than in forked substrates could support the helicase activity of Hmi1p. Substrates containing 3Ј-ssDNA tails 16, 9, and 5 nt in length were unwound with maximum efficiency. In comparison with the previous results (Figs. 5 and 6), this was the only type of substrate to which Hmi1p was efficiently targeted by a 5-nt 3Ј-tail. The 3Ј-flap substrate with a 2-nt 3Ј-tail was also slightly more efficiently unwound than the forked substrate with a 3Јoverhang of the same length (Figs. 6A and 7, compare lanes 10). In general, the substrates with flap structure were more efficiently unwound by Hmi1p than forked substrates (Figs. 6 and 7, compare graphs). In the reactions at 2 nM enzyme concentration, 45-50% flap substrates with at least 5-nt 3Ј-tails were unwound. In the reaction at 6 nM Hmi1p, ϳ70 -80% of the same substrates were unwound. At the same time, different forked substrates with at least 9-nt overhangs were unwound 30 -35% by 2 nM Hmi1p and 50 -60% by 6 nM Hmi1p. Taken together, these results show that Hmi1p can start the unwinding of duplex DNA from a specific 3Ј-flap structure. DISCUSSION We have purified recombinant Hmi1p from E. coli and described the NTPase and helicase activities of the purified protein. We also expressed three different mutant proteins with substitutions in conserved residues, but only one of them, Hmi1p(E211Q), could be purified by the same method as the WT Hmi1p. The GST fusion proteins of two other mutants, Hmi1p(K32M) and Hmi1p(R275M), were both in the insoluble fraction after cell lysis. All of the helicases of superfamily 1 contain seven conserved short amino acid sequence motifs that are essential for NTP hydrolysis and DNA unwinding (8). Glu-211 of Hmi1p is an absolutely conserved residue in motif II, being a putative catalytic acid (24); thus, Hmi1p(E211Q) should be defective with regard to ATPase and helicase activities. However, crystal structure analysis of PcrA, the closest homologue of Hmi1p in bacteria, suggests that mutations in this residue should not influence the binding of ATP (24). On the other hand, residues Lys-32 and Arg-275, conserved residues from motifs I and IV, respectively, are involved in ATP binding as demonstrated by crystal structure analysis of PcrA as well as substitution mutants of PcrA and UvrD (24 -26). Therefore, it is possible that the GST-Hmi1 fusion protein is able to take a stable conformation in E. coli only upon binding ATP. We demonstrated that Hmi1p translocates in the 3Ј-5Ј direction on ssDNA. This is consistent with our previous result that Hmi1p could unwind only the substrate with a 3Ј-ssDNA overhang and not the substrate with a 5Ј-ssDNA overhang (21). Compared with all of the other NTP(s) and dNTP(s), Hmi1p could efficiently hydrolyze only ATP and dATP. The determined kinetic parameters for ATP hydrolysis, K m and k cat (0.185 Ϯ 0.037 mM and 13 Ϯ 1 s Ϫ1 , respectively) were similar to the same parameters of PcrA (0.225 mM and 25 s Ϫ1 , respectively) (24). Hmi1p appeared to act as a distributive helicase, unwinding efficiently only short DNA duplexes. It has been demonstrated with different homologues of Hmi1p, Srs2p from yeast nucleus as well as PcrA and Rep helicases from bacteria, that they are distributive helicases by themselves (27)(28)(29). However, cofactors such as ssDNA-binding proteins and/or replication initiator proteins have been found to strongly enhance the processivity of these helicases (27)(28)(29). Thus, it remains to be studied whether some mitochondrial proteins can influence the processivity of Hmi1p as well.
The substrate requirements for the helicase activity of Hmi1p were characterized using oligonucleotide substrates with different structures. Our first conclusion was that Hmi1p unwinds only substrates with 3Ј-ssDNA overhangs or stretches of ssDNA. This is in accordance with the interpretation that Hmi1p translocates in the 3Ј-5Ј direction on ssDNA. The necessary length of 3Ј-overhang for the productive Hmi1p recruitment was studied with three different types of substrates. The required length appeared to differ depending on the substrate structure. In summary, the results were as follows. (i) For simple duplex DNA with 3Ј-ssDNA overhang, a ssDNA strand of at least 19 nt in length was needed for the targeting of Hmi1p. (ii) With forked substrates, the maximum unwinding activity was observed already with two 9-nt overhangs. (iii) At flap structures, even a shorter 3Ј-ssDNA tail (2-5 nt) was sufficient for the helicase activity of Hmi1p. These properties may represent different roles of Hmi1p in the mtDNA metabolism, although it is also possible that some of these properties do not find use in vivo.
In the aspect of the results with simple double-stranded DNA with a 3Ј-overhang, Hmi1p is similar to PriA, a bacterial DNA helicase that requires at least a 16-nt 3Ј-overhang for stable binding (30). Although PriA is not a homologue of Hmi1p by sequence, it has some functional redundancy with the Rep helicase, a close homologue of Hmi1p (31). Thus, it is possible that PriA shares some functional homology with Hmi1p as well. The less stringent requirements for the 3Ј-overhang in the forked substrates indicate that Hmi1p is a structure-specific helicase. The experiments with the forked substrates with one short (2 nt) and one long (at least 9 nt) overhang gave somewhat unexpected results. Such substrates were unwound with approximately half of the maximum efficiency; however, the substrate with the longer 5Ј-overhang was slightly more efficiently unwound than the substrate with the longer 3Ј-overhang. This might possibly be due to the sequence difference that may influence the fork structure recognized by Hmi1p in this case. Finally, the result with flap-structured substrates confirms the hypothesis that Hmi1p recognizes a higher-order structure of DNA. The 2-5-nt 3Ј-tails that were found to be enough for Hmi1p to start unwinding double-stranded DNA cannot promote stable binding by themselves. Therefore, these results indicate that Hmi1p is probably targeted to a 3Ј-flap structure. In comparison, PcrA has also been demonstrated to be a structure-specific helicase but it does not unwind forked substrates at all, preferentially unwinding substrates that  22,19,16,13,9, and 5 nt, respectively, and separated on a 12% PAGE. The structure of the substrates is shown above the radiograph, the variable ssDNA overhang is indicated with a dotted line, and the 32 P label is indicated with and asterisk. ϩ, reaction with 6 nM Hmi1p; Ϫ, substrate incubated in the reaction buffer without Hmi1p; D, heat-denatured substrate; S, substrate; P, unwound oligonucleotide. Results from the reactions with different Hmi1p concentrations were quantified and plotted in a graph.
have hairpin structures with a short overhang at one end (32).
Considering the characteristics of the helicase activity of Hmi1p reported here together with the results from in vivo experiments reported in our previous studies (5,15), we may propose different functions for Hmi1p in yeast mtDNA metabolism. Hmi1p may be necessary at some step of replication elongation as has been proposed for its bacterial homologue, Rep helicase (33). One possibility is that Hmi1p may facilitate replisome progression through regions containing hairpin or other secondary structures. This hypothesis is supported by the results of the experiments with temperature-sensitive mutants of Hmi1p where the mtDNA became fragmented after shifting the culture to the restrictive temperature (15). Such fragmentation may possibly be the result of ineffective elongation in the absence of Hmi1p. We also have detected some specific binding of Hmi1p to hairpin structures. 2 The other possibility is that Hmi1p is involved in recombination processes, because during recombination, such DNA structures may form to where Hmi1p is preferentially targeted. It has been demonstrated that a specific recombination-based replication mechanism is used in yeast mitochondria, because the mtDNA in yeast forms a network of branched molecules (34,35). The hypothesis of Hmi1p functioning in recombination finds also some support in the experiment with yeast cells deficient in Hmi1p as well as Pif1p, the other DNA helicase described in yeast mitochondria. We demonstrated that, in the absence of both helicases, some types of rho mitochondrial genomes could not be maintained in the mitochondria, suggesting that Hmi1p and Pif1p have some functional redundancy (5). Because the role of Pif1p is proposed to be in recombination and repair (11,12), it can be speculated that recombination is the process necessary for maintaining this type of rho genomes. Therefore, recombination may also be the process where Hmi1p and Pif1p share the functional redundancy. Because the replication system of yeast mtDNA in general has not been well described to date, the role of Hmi1p, particularly within this system, remains to be revealed by future investigations.  16,9,5,2, and 0 nt, respectively, and separated on a 12% PAGE. The structure of the substrates is shown above the radiograph, the variable 3Ј-ssDNA tail is indicated with a dotted line, and the 32 P label is indicated with an asterisk. ϩ, reaction with 6 nM Hmi1p; Ϫ, substrate incubated in the reaction buffer without Hmi1p; D, heat-denatured substrate; S, substrate; P1, unwound oligonucleotide with 3Ј-T-tail; P2, oligonucleotide unwound from the 5Ј end of the bottom strand oligonucleotide. Results from the reactions with different Hmi1p concentrations were quantified and plotted in a graph.