Initiation of Mitochondrial DNA Replication by Transcription and R-loop Processing*

The mitochondrial genome of eukaryotic cells is maintained by a mechanism distinct from that employed in the nucleus. Mitochondrial DNA replication at the leading-strand origin is coupled to transcription through the formation of an RNA-DNA hybrid known as an R-loop. In vivo and in vitro evidence has implicated an RNA processing enzyme, RNase MRP, in primer maturation. In our investigation of mammalian RNase MRP, we have analyzed its specific endoribonuclease activity on model R-loops. We demonstrate here that human RNase MRP cleaves this distinctly configured substrate at virtually all of the major DNA replication sites previously mappedin vivo. We further show that the processed RNA products remain stably base-paired to the template DNA strand and are functional for initiating DNA synthesis on a closed circular plasmid. Thus,in vitro initiation of leading-strand mtDNA synthesis requires only the actions of RNA polymerase and RNase MRP for the generation of replication primers.

The mitochondrial genome of eukaryotic cells is maintained by a mechanism distinct from that employed in the nucleus. Mitochondrial DNA replication at the leading-strand origin is coupled to transcription through the formation of an RNA-DNA hybrid known as an Rloop. In vivo and in vitro evidence has implicated an RNA processing enzyme, RNase MRP, in primer maturation. In our investigation of mammalian RNase MRP, we have analyzed its specific endoribonuclease activity on model R-loops. We demonstrate here that human RNase MRP cleaves this distinctly configured substrate at virtually all of the major DNA replication sites previously mapped in vivo. We further show that the processed RNA products remain stably base-paired to the template DNA strand and are functional for initiating DNA synthesis on a closed circular plasmid. Thus, in vitro initiation of leading-strand mtDNA synthesis requires only the actions of RNA polymerase and RNase MRP for the generation of replication primers.
Transcription of mtDNA is directed by cis-acting elements located at the 5Ј boundary of the regulatory sequence known as the D-loop, a region which also houses the leading-strand origin of replication, O H (1). Each of the two strands of the circular mtDNA is transcribed by bacteriophage-like mtRNA polymerase, producing polycistronic transcripts, which are processed into rRNAs, tRNAs, and mRNAs (2). Transcripts from the light (L)-strand promoter yield eight tRNAs, the message for NADH dehydrogenase subunit 6, and the primer precursors for leading-strand mtDNA replication. Earlier studies provided strong evidence for transcript-priming at the mtDNA O H in human and mouse organelles (3,4). Surprisingly, primer RNA species were found to co-purify with mtDNA in CsCl-ethidium bromide equilibrium density gradients (4); this result has been recently extended by the detection of in vivo RNA-primed DNA strands by polymerase chain reaction analysis (5). In vitro mitochondrial transcription assays have shown that a significant fraction of nascent transcripts in the direction of leading-strand replication remains base-paired at the mtDNA O H (6), a phenomenon that can explain the observed physical association of primer RNAs and mtDNA. Since the L-strand promoter transcripts represent genes located downstream of the DNA replication start sites, a specific RNA processing event was hypoth-esized for primer RNA 3Ј-end maturation. RNase MRP was originally identified as an activity in mitochondrial extracts that fulfilled at least one of the criteria for a specific primer maturation endonuclease (7). We have recently reported that mouse RNase MRP has the capacity to process a model R-loop substrate containing the O H (8). In this report, we extend these studies to examine this unusual RNA processing activity in the human mitochondrial system and further demonstrate that such processed RNA primers are capable of initiating DNA replication on a model double-stranded circular mtDNA template in vitro.

EXPERIMENTAL PROCEDURES
Construction of Substrates-RNA was synthesized by SP6 transcription generated from pK408SP which was digested with restriction endonuclease FokI. This RNA was radiolabeled with [5Ј-32 P]pCp (3000 Ci/mmol, NEN Life Science Products) and T4 RNA ligase as described elsewhere (7). Labeled RNA was purified by urea-polyacrylamide gel electrophoresis and used to assemble the R-loop with the supercoiled pK408SP template by the formamide annealing method described previously (9). For the construction of the RNA-cDNA heteroduplex hybrid, an oligonucleotide that is complementary to position 294 -314 on pK408SP (oligo HF314, 5Ј-d(TTGATTCCTGCCTCATCCTAT)) was annealed onto the purified radiolabeled pK408SP-derived RNA by heat denaturation at 95°C for 5 min and slow cooling to 37°C over 30 min. The reaction mixture was adjusted to 50 mM Tris-HCl, pH 8.0, 50 mM KCl, 10 mM MgCl 2 , 2 mM dithiothreitol, 0.2 mM spermidine, and 0.4 mM dATP, dCTP, dGTP, and dTTP (Amersham Pharmacia Biotech). 10 units of AMV 1 reverse transcriptase (Life Technologies, Inc.) were added and incubated at 37°C for 30 min. The resulting RNA-DNA hybrid was phenol-extracted, concentrated by ethanol precipitation, and then resolved in a 4% native polyacrylamide gel in 45 mM Tris borate-EDTA buffer at pH 8.3. Individual species were located by autoradiography, and full-length products were excised and eluted into 0.5 M ammonium acetate, 0.1% SDS, overnight at room temperature. The solution containing the hybrid was ethanol-precipitated, redissolved in water, and passed through a Sephadex G-50 (Amersham Pharmacia Biotech) spin column.
Enzymes and Activity Assays-RNase MRP was purified from human KB cells, and assays were carried out as reported previously (7,10). We detected no deoxyribonuclease activity nor any nonspecific ribonuclease activity in our final RNase MRP preparation. Oligonucleotide LRI, used previously for specific RNase MRP inhibitions (8), contains the following sequence, 5Ј-d(GGTCCGGGGACTTTCC). Murine and bovine endo G preparations were kindly provided by Dr. Michael J. Lieber. Reaction conditions were identical to those in Lee and Clayton (8) except endo G was added at various dilutions ranging from 1:1 to 1:1000. Digestion times at 37°C were also varied for 15-120 min. Reactions were quenched by phenol extraction and ethanol precipitation, and the products were resolved as described in Lee and Clayton (8). Micrococcal nuclease treatment of the endo G fractions had no effect on its nonspecific cleavage activity on the R-loop and RNA-cDNA substrates. No reproducible specific cleavage products were observed on either substrate. Sequencing-grade ribonucleases T1 and U2 were purchased from Amersham Pharmacia Biotech. Bacteriophage SP6 RNA polymerase, Escherichia coli RNase H, and RNase T2 were purchased from Life Technologies, Inc. Micrococcal nuclease and DNase I were purchased from Boehringer Mannheim. Modified bacteriophage T7 DNA polymerase (Sequenase® version 2) was purchased from Amersham Life Sciences.
Structural Probing Assays-1-10 pmol of radiolabeled RNA or Rloops were incubated in RNase MRP reaction buffer for 15 min at 37°C in a total reaction volume of 25 L. Serial dilutions ranging from 1:10 to 1:10 6 of RNase T1, T2, V1, and U2 were tested to generate a uniform cleavage pattern. 1 L of appropriately diluted ribonuclease (between 1:10 3 and 1:10 4 ) was added to the sample and incubated at 37°C for 10 s. RNase H reactions were carried out using 0.03 unit of enzyme and digested at 37°C for 1 min. All reactions were quenched by phenol extraction, and samples were recovered by ethanol precipitation. Products were resolved in 6% polyacrylamide gels containing 8 M urea.
In Vitro DNA Synthesis Reactions-General methods to generate R-loops via transcription using bacteriophage SP6 RNAP and supercoiled templates were described in Lee and Clayton (9). Briefly, 0.5-1.5 nM DNA template, 0.4 mM complete NTP mixture, and 10 units of SP6 RNAP were mixed in a total reaction volume of 100 L. When the RNA was to be internally radiolabeled, 100 Ci of [␣-32 P]UTP (300 Ci/mmol, NEN Life Science Products) were included in the transcription mixture. Reactions were quenched by phenol extraction and ethanol precipitated. Dried pellets were redissolved in 50 mM Tris-HCl, pH 8, and applied onto a 1.5 cm ϫ 3-cm benzoylated-naphthoylated-DEAE cellulose column as described in Lee and Clayton (9). After the ion-exchange fractionation, the mixture was passed through a 1.5 cm ϫ 6-cm Sepharose CL-4B (Amersham Pharmacia Biotech) column in 20 mM Tris-HCl, pH 8, 50 mM KCl. The substrate concentration was quantitated, aliquoted, and adjusted to standard RNase MRP reaction conditions. RNase MRP digestions were carried out in a total reaction volume of 25 L; reactions were stopped by the addition of 40 M oligo LRI. Thereafter, the entire reaction mixture was passed through a Sepharose CL-4B (equilibrated in 20 mM Tris-HCl, pH 8, 50 mM KCl) spin column constructed in a 3-ml syringe barrel. Negative RNase MRP controls were treated in the same manner. Samples were phenol extracted and recovered by ethanol precipitation.
For DNA synthesis, the gel-filtered R-loop sample was treated with RNase MRP or mock-treated (negative control) as above; the RNase MRP reactions were inhibited by the addition of 40 M of oligo LRI. Thereafter, the reaction mixture was adjusted to 125 M dATP, dCTP, dGTP, and dTTP (Pharmacia) and 13 units of modified T7 DNAP was added in a total reaction volume of 100 L (buffer containing 20 mM Tris-HCl, pH 8, 10 mM MgCl 2 , 50 mM KCl, and 5 mM dithiothreitol). For synthesis of radioactive DNA, the reaction mixture described above was supplemented with 25 Ci of [␣-32 P]dATP (3000 Ci/mmol, NEN Life Science Products). Reactions were incubated at 37°C for 15 min, after which the entire reaction mixture was passed through a 1.5 cm ϫ 3-cm Sepharose CL4B column. Thereafter, samples were phenol-extracted, concentrated by ethanol precipitation, and resolved in denaturing gels.
Primer extension analysis was carried out using oligo HF314 radiolabeled at the 5Ј terminus with T4 polynucleotide kinase (New England Biolabs) and [␥-32 P]ATP (6000 Ci/mmol, NEN Life Science Products). Oligo HF314 was annealed, and 10 units of AMV reverse transcriptase were added as described above for the cDNA synthesis reactions. Samples were incubated at 37°C for 30 min followed by phenol extraction and ethanol precipitation. Ribonuclease digestion of DNA synthesis reaction was carried out by incubation with 10 units of RNase T2, followed by phenol extraction. After ethanol precipitation, samples were analyzed by primer extension as described previously. Deoxyribonuclease treatment was carried out in an analogous manner but with 10 units of RNase-free DNase I. Sequenase-dideoxyribonucleotide DNA sequencing reactions were carried out as described by the kit manufacturer (U. S. Biochemical Corp.). Products were resolved in 6% polyacrylamide, 8 M urea gels.

Assembly of R-loops and Structural
Analysis-Model R-loops were reconstituted using a recombinant plasmid containing the human mtDNA O H sequence (pK408SP) and RNA synthesized in vitro using bacteriophage SP6 RNAP (10). The reconstituted human R-loop exhibited chromatographic and electrophoretic properties (Fig. 1, A and B), consistent with previously observed model R-loops (9). When the R-loop is digested with E. coli RNase H, an endoribonuclease specific for the RNA strand in an RNA-DNA heteroduplex, a biased cleavage pattern is observed, indicating a specific region of high cleavage frequency (Fig. 1C, lane 2). These cleavage sites correspond to the mapped positions of the RNA-DNA contacts within R-loops generated by transcription (6). Consistent with previous observations, these RNA-DNA contacts localized to the guanine-rich CSB II, one of three short elements conserved between human and mouse mtDNAs (11). Sequence mutations within this element completely abolish all transcription-coupled R-loop formation and prevent the assembly of stable R-loops by the formamide annealing method. 2 Furthermore, under our transcription reaction and formamide annealing conditions, R-loop formation was not observed with unrelated nucleic acid sequences of comparable size and base content (data not shown).
To assess the global structure of this stable R-loop, the RNA strand of the R-loop was systematically digested with ribo-2 D. Y. Lee and D. A. Clayton, manuscript in preparation. nucleases of various specificities. In an effort to maintain a cleavage frequency of no more than one nick per RNA molecule, ribonuclease reactions were carried out using dilute enzyme concentrations and brief digestion times (1 min for RNase H reactions; 10 s for all others). Because the structure of the RNA may change after an initial cleavage event, it was important to minimize potential experimental artifacts in obtaining structural information.
RNase V1 from cobra venom has been used to probe for regions of potential RNA-RNA base pairing because of its preferential cleavage of RNA in double helical or stacked base conformations (12). Aside from those regions in RNA-DNA pairings as demonstrated by E. coli RNase H sensitivity (Fig.  1C, lane 2), much of the full-length of the RNA was digested by RNase V1 in a semiperiodic pattern (Fig. 1C, lane 5), implicating the participation of those residues in RNA-RNA base-pairing. RNase T2 from Aspergillus oryzae, a ribonuclease which cleaves after any unpaired residue, cleaved the RNA strand in the R-loop at those sites generally protected from RNase V1 cleavage (Fig. 1C, lane 8), suggesting those regions to be singlestranded. Although the highly reproducible cleavage patterns may reflect a composite of multiple subpopulations of differing RNA conformations, these results indicate that the RNA strand in the R-loop is only partially base paired to the DNA template.
The surprisingly complex structure of the RNA strand in the R-loop was compared with the free RNA by probing with two highly specific ribonucleases (Fig. 1D). The cleavage patterns generated by RNase T1 from A. oryzae (specific for unpaired guanine) and RNase U2 from Ustilago sphaerogena (unpaired adenine Ͼ Ͼ unpaired guanine) revealed that the RNA strand in the R-loop was folded into a structure surprisingly similar to that of the free RNA. For RNase T1, differences between the free RNA and R-loop cleavage patterns were limited to the CSB II-CSB III region, where the RNA is base-paired with the DNA template as judged by the RNase H cleavage pattern. The cleavage sites located at the 3Ј boundary of CSB II in the free RNA were reduced in the R-loop; likewise, the relatively efficient cleavage site located between CSB II and CSB III is protected in the R-loop (Fig. 1D, lane 3 versus lane 4). Interestingly, the guanine residues of CSB II are well protected against RNase T1 cleavage even in the free RNA, implying that this region may be configured in a manner that is sterically occluded from the enzyme. For RNase U2, since there is only one adenine residue in CSB II, which is protected in the free RNA, there is little overall difference between the cleavage patterns of the free RNA and R-loop (Fig. 1D, lane 6 versus lane  7), a surprising observation that has been previously reported for the mouse mtRNA sequence (9). Hence, for both ribonucleases used in this assay, there is little discernible difference in the cleavage pattern between the free and R-loop form of the RNA. Although the physiological relevance for the extensively folded RNA structure in the R-loop complex is presently under investigation, our initial experimental results suggest it is likely to be important in RNase MRP substrate recognition. 2 This similarity in higher order structure may also explain the ability of RNase MRP to cleave the free RNA substrate as previously reported (8,10).
Human RNase MRP Cleaves the R-loop Substrate-Having obtained gross structural information on the R-loop, we then asked if human RNase MRP can cleave this substrate in a physiologically relevant manner. RNase MRP was purified from human KB cells as described previously (7,10), and a standard amount was added to reaction mixtures containing 0.5-1.5 nM R-loop substrate. Digestions were quenched by phenol extraction, and samples were recovered by ethanol precipitation. Reaction products were resolved in 6% polyacrylamide gels containing 5 M urea and visualized by autoradiography. Addition of the human RNase MRP fraction resulted in multiple cleavage products ( Fig. 2A, lane 3). To confirm that RNase MRP catalyzed this reaction, the ribonucleoprotein composition of the holoenzyme was targeted for selective inhibition. Micrococcal nuclease was previously used to digest the RNA subunit of RNase MRP and thereby abolish enzymatic activity (13). A standard amount of the RNase MRP fraction was treated with micrococcal nuclease in the presence of Ca 2ϩ (an essential metal cofactor), the digestion was quenched with EGTA (a calcium-specific chelator), and then assayed for R-loop processing activity by the addition of the substrate. Consistent with previous results, the observed R-loop processing activity was abolished ( Fig. 2A, lane 7), demonstrating a requirement for RNA in the R-loop cleavage reaction. The requirement for protein subunits in the reaction was tested by heat inactivation or by treatment of the enzyme fraction with proteinase K, a serine protease active on native polypeptides. After digestion of the fraction with proteinase K, all R-loop processing activity was again abolished (Fig. 2A, lane 9). In contrast, inactivated proteinase K had no inhibitory effect ( Fig. 2A, lane 8).
The specificity of the RNase MRP RNA subunit involved in the cleavage reaction was previously shown by using an antisense oligonucleotide directed against the 265-nucleotide MRP RNA (13). Addition of the antisense oligonucleotide, oligo LRI (8), inhibited the observed R-loop cleavage reaction in a concentration-dependent manner (Fig. 2B, lanes 3-5). In contrast, an unrelated oligonucleotide of comparable length, oligo 5.8S, had no effect at the concentrations tested (Fig. 2B, lanes 7-10).
Since we have previously shown that oligo LRI had no inhibitory effect on the related ribonucleoprotein endonuclease RNase P, these data suggest that the observed R-loop processing reaction is catalyzed specifically by RNase MRP (8).
Certain antisera from humans with clinical autoimmune syndromes are known to bind particular ribonucleoproteins selectively. The Th/To human autoantiserum was previously shown to immunoprecipitate the RNase MRP ribonucleoprotein and thereby deplete enzymatic activity (14). Crude Th/To serum was adsorbed onto Staphylococcus aureus protein A-Sepharose beads (PAS), washed, and incubated with a standard amount of the human enzyme fraction. After centrifugation, the supernatant was assayed for activity. Control experiments were carried out in parallel using serum from a normal subject. While R-loop processing was unaffected by immunoprecipitation with the normal serum (Fig. 2C, lane 2), the Th/To serum depleted the observed R-loop processing activity (Fig. 2C, lane 3).
To determine whether the unique structure of the R-loop is required for the cleavage reaction, we assayed a simple RNA-DNA heteroduplex constructed with the same RNA substrate and its cDNA made by reverse transcription (see "Experimental Procedures"). This RNA-DNA hybrid was digested by E. coli RNase H in a nonspecific manner, yielding no oligoribonucleotides of significant length (Fig. 2D, lane 2). In contrast, this substrate was not cleaved by RNase MRP (Fig. 2D, lane 3), an observation consistent with previous reports (7,8). This experiment also demonstrates the absence of nonspecific RNase H activities in the RNase MRP preparation. Interestingly, this type of RNA-cDNA heteroduplex was reported to be a substrate for endo G, an abundant mitochondrial endonuclease proposed to generate primers for mtDNA replication (15). Therefore, in an effort to rule out contamination of our RNase MRP preparations, purified endo G was used to obtain a cleavage pattern on our model substrates. Under our reaction conditions, endo G efficiently degraded the RNA strand as well as the DNA template of our R-loop and heteroduplex in a nonspecific manner, generating no population of oligoribonucleotides with a distinctive length (data not shown). Thus, the R-loop processing observed with our RNase MRP preparation is not due to the action of endo G. We conclude from these and the above results that the observed R-loop cleavage reaction is an activity of RNase MRP.
Mapping of the cleavage sites on the R-loop was achieved by resolving the RNase MRP reaction products next to RNA sequencing ladders in polyacrylamide-urea gels (Fig. 3A). The cleavage sites, arbitrarily labeled 1 through 7, were then compared with the positions of previously mapped sites of RNA priming and DNA synthesis initiation. Although the human in vivo mapping data are complex with regard to the extensive microheterogeneity of RNA and nascent DNA termini, we observed alignment of the majority of the RNase MRP cleavage sites with a significant number of the in vivo priming sites, most notably the major DNA synthesis initiation site positioned ϳ16 -20 bases downstream of CSB I, classically defined as the principal O H (Fig. 3B, cleavage site 6). Remarkably, RNase MRP cleavage site 6 corresponds precisely with the major H-strand 5Ј termini more recently mapped by Kang et al. (5) using a polymerase chain reaction method. By inspection, the cleavage site selection followed no apparent sequence rule. This finding is consistent with previous results in the mouse system, where R-loop cleavage sites do not bear any obvious sequence homology. More importantly, the sites of cleavage for each species, differing in spatial and sequence parameters, reflect the correct placement with respect to mapping of nascent in vivo DNA strands. Conservation of accurate processing across species supports the role of RNase MRP in primer RNA formation.
To address the issue of physiological relevance, we have examined whether the RNase MRP cleavage products are functional as replication primers. Although mtRNA polymerase and mtDNA polymerase (pol ␥) activities have been biochemically fractionated from mammalian cell extracts, such preparations are not of sufficient purity to characterize sequential in vitro enzymatic reactions on R-loop substrates reliably. Therefore the well defined activities of recombinant bacteriophage SP6 RNAP and T7 DNAP were employed to follow the fate of the transcription-generated RNA primers as a function of RNase MRP activity.
RNase MRP Processing Is Required for Initiating DNA Synthesis in Vitro-Transcription was carried out in the presence of [␣-32 P]UTP using SP6 RNAP on supercoiled plasmid pK408SP, and the resulting R-loop was purified by ion-exchange chromatography on benzoylated-naphthoylated-DEAE cellulose and gel-filtration (see "Experimental Procedures"). These purification steps were required to remove the abundant amount of variable-length free RNAs from those assembled into R-loops. Hybridized RNAs were demonstrated by their persistence after size-exclusion chromatography and sensitivity to E. coli RNase H (Fig. 4, lanes 1 and 2, respectively). The same procedure using the parental vector, pSP65, resulted in no detectable hybridized RNA (data not shown). Also, R-loop formation via transcription required superhelical tension in the template plasmid as previously reported (9). Consistent with earlier observations (6,9), the nonspecific smear of radioactivity in Fig. 4, lane 1 contains a group of strong transcriptional pause or termination products of ϳ130 nt. If we assume that these RNA fragments represent full-length strands initiating from the promoter, their 3Ј termini would map to the 5Ј boundary of CSB II. A shorter autoradiographic exposure (data not shown) indicated that this cluster of bands is of the same size as those observed in Fig. 4, lane 4, and, less intensely, in lane 5. We have noticed that this effect was dependent on the NTP and enzyme concentrations as well as reaction times (data not shown), suggesting that multiple rounds of transcription may be blocked at the CSB II site perhaps due to the strong RNA-DNA base pairing in that region. It is also possible that these abortive transcripts may have remained hybridized to the DNA template by base pairing at the CSB III region, an element previously reported to influence greatly the stability of the RNA-DNA interaction (6). We also note that due to the high specific radioactivity of the transcript, which was necessary to detect final products after the multiple reaction and purification steps, the RNA was unavoidably subject to radiochemical hydrolysis, contributing to the nonspecific background smear (Fig. 4, lanes 1 and 3-5).
Using the transcription-generated R-loops, we next assessed the fate of the hybridized RNA following the action of RNase MRP. Digestions were carried out as before, but RNase MRP reactions were quenched by the addition of oligo LRI (40 M) to permit rapid handling steps. Because we wished to know whether the processed RNAs remained base-paired to the DNA template, the reaction samples were passed over a high molecular weight exclusion column (Sepharose CL-4B). Recovered samples were then phenol extracted, ethanol-concentrated, and resolved in denaturing gels as before. As shown in Fig. 4, lane 3, RNase MRP generated only marginally visible discretelength RNA fragments against the smear of radioactivity. However, the relative amount of the ϳ130-nt cluster of bands is significantly reduced, indicating that these RNAs were not of sufficient stability to elute with the DNA template in the second gel filtration step.
To address whether the RNAs that remained following the action of RNase MRP could be extended by a DNAP, the R-loop substrate was treated with RNase MRP as above and, following the digestion, the mixture was adjusted to 125 M each of the four nonradioactive dNTPs and 13 units of modified T7 DNAP (Sequenase® version 2) in a total reaction volume of 100 l (see "Experimental Procedures"). We note that a relatively large amount of T7 DNAP was required to visualize final products, perhaps reflecting this enzyme's low activity on duplex DNA templates. After incubation at 37°C for 15 min, the entire reaction volume was again passed through the gel filtration column, followed by phenol extraction and precipitation. Of the many bands still within the radioactive smear, several discrete-length products ranging in size from ϳ130 to ϳ300 nt were readily visible (Fig. 4, lane 4). Since the DNAP reactions were carried out on a supercoiled plasmid template, these species probably represent DNA synthesis termination events due to a topological constraint. The ϳ300-nt product would require helix unwinding of ϳ9% of the total DNA template length (3.4 kilobase pairs), consistent with relaxation of a circle of this size (9). Therefore, T7 DNAP is not expected to be capable of extending beyond this point without accessory factors. A similar observation was reported for the yeast mtDNA supercoiled templates in sequential mtRNA polymerase and mtDNA polymerase reactions (16). The ϳ130-nt cluster of bands was more prominent in this particular autoradiogram (Fig. 4, lane 4), but this effect did not correlate with DNAP activity in repeated experiments (data not shown) and may represent some flow-through in the gel filtration step. Addition of T7 DNAP alone resulted in few discrete products (Fig. 4, lane  5). Hence, these results suggest that the appearance of nascent DNA products is dependent on the action of RNase MRP. In these experiements, since the only radioactive entity is RNA, the newly formed products must be composed of both RNA and DNA.
To confirm this observation, we carried out a parallel experiment in which the R-loop was generated by transcription in the presence of nonradioactive NTPs and subsequently digested with RNase MRP in the standard manner. DNA synthesis was then carried out in the presence of [␣-32 P]dATP (plus complete dNTP mix) and T7 DNAP as described above. Following the same procedure, the reaction samples were passed through a final gel filtration column, concentrated, and resolved in denaturing gels. Again, multiple bands were observed, indicating the synthesis of new DNA strands that matched the major bands observed in the RNA-labeled experiment (Fig. 4, lane 6; compare with lane 4). The cluster of ϳ130-nt bands is not visible in these experiments, suggesting that the presumed abortive RNA products are not substrates for T7 DNAP. This parallel experiment permitted the clear identification of nascent DNA products, devoid of the RNA breakdown that was observed when the transcripts were radiolabeled.
To probe further the structure of the RNA-containing nascent DNA, an equivalent amount of the completed DNAP reaction was each treated with RNase H or RNase T2, as described above. After the ribonuclease digestions, the samples were passed through the gel filtration column and further processed in the standard manner. Digestion of these DNA products with RNase H and RNase T2 each generated smaller fragments ranging from ϳ90 to ϳ150 nt in length, indicating that the length of the RNA component of the nascent chain is significant (Fig. 4, lanes 7 and 8, respectively). By comparing the largest bands in Fig. 4, lane 6, with those of lanes 7 and 8, we can estimate the length of the RNA component of the nascent DNA chains to be ϳ150 to 190 nt. Interestingly, the RNase H result demonstrates that the DNAP extension products contain RNA at their 5Ј termini in a configuration also accessible to RNase T2, a structural presentation similar to those reported by others (17,18). Close scrutiny of the bands in Fig. 4, lanes 7 and 8, reveals that the RNase H-generated products are larger by 1 or 2 nt relative to those generated by RNase T2; this is also consistent with the observations of other investigators (17,18). We also note that, after the ribonuclease treatments, the total amount of radioactivity that passes through the gel-filtration column is decreased by 75-90% relative to the untreated RNase MRP-DNAP reaction, suggesting that the loss of the RNA component destabilizes the nascent DNA strand from the template. These results demonstrate that the processed RNAs in the R-loops remain on the DNA template and are capable of providing 3Ј termini for initiating DNA synthesis on a closed circular model mtDNA template.
Because the in vitro DNA synthesis reactions were performed on circular DNA templates, we wished to obtain more precise positional information on the RNA-primed nascent DNA strands. Transcription, RNase MRP, and T7 DNAP reactions were performed as before, but in the absence of labeled nucleotides. RNA-primed DNA products were then analyzed by primer extensions using AMV reverse trancriptase and a 5Јend-labeled oligonucleotide complementary to the 3Ј region of the furthest RNase MRP cleavage site. Products were resolved next to dideoxyribonucleotide-terminated sequencing reactions using the same radiolabeled oligonucleotide. The experimental control demonstrates that the SP6 RNAP reaction yields essentially full-length RNA originating from the promoter (Fig. 5,  lane 2). However, when RNase MRP and T7 DNAP are included in the reaction, multiple primer extension products are visible, some minor species extending as far as CSB I (Fig. 5,  lane 3). This result is consistent with that observed with the internally labeled DNA synthesis reactions described above (Fig. 4). Although there is a background of multiple primer extension products due to the presence of the template plasmid (Fig. 5, compare lanes 1 and 5), the observed cluster of bands is reproducible and depends on the actions of RNase MRP and T7 DNAP. These products were treated with RNase T2 to determine the precise locations of the RNA to DNA transitions. As shown in Fig. 5, lane 4, the transitions map precisely to the RNase MRP cleavage sites 5, 6, and 7. By comparing the difference in size of the primer extension products, we deduce that the RNA component of the nascent DNA chains is approximately 157, 172, and 185 nt. Hence, it appears that only the RNase MRP-generated primer RNA termini that map downstream of CSB I are extended by T7 DNAP. This result is consistent with the in vivo distribution of RNA to DNA transitions observed in human mitochondria (3). We suggest that our reconstituted heterogeneous DNA replication system correctly recapitulates the in vivo replication mechanism. DISCUSSION The results of our studies suggest that the initiation of DNA replication at the mammalian mtDNA O H involves transcription-coupled synthesis of a structurally complex pre-primer RNA that remains stably hybridized to the DNA template. We have demonstrated the RNA processing activity of human RNase MRP on model R-loops either assembled in vitro or generated by transcription. Since we have found that RNase MRP cleaves at virtually all of the known priming sites, the action of this single enzyme can account for the heterogeneous primer RNA species observed in vivo. We have also demonstrated that the products of RNase MRP cleavage remain stably base paired to the DNA template strand, and these RNA fragments are capable of initiating DNA synthesis in vitro using a heterologous bacteriophage DNAP. Although relatively inefficient, we have shown that in vitro activation of the mtDNA leading-strand origin can be achieved simply by the actions of RNAP and RNase MRP.
Compelling observations support a conserved mechanism among animal mitochondria, the universal proximity of known transcriptional promoters and replication origins, and the ability of RNase MRP from various species to process the respective mtRNAs in all cases where it has been investigated. However, a true in vitro mtDNA replication system awaits the availability of well defined mammalian mitochondrial enzymes. The sequence of the putative human mtRNA polymerase catalytic subunit is now known (19); however, the precise composition of the holoenzyme remains ill defined. Characterizations of the mtDNA polymerase (pol ␥) from Drosophila indicate that it is a multimeric holoenzyme with at least two subunits (20,21), but the mammalian counterparts still remained poorly characterized. Pol ␥ appears to lack associated 5Ј to 3Ј exonuclease and RNase H activities (22), features that are compatible with the proposed R-loop mechanism since we have shown that RNase H activity destabilizes the RNA-DNA hybrid. This condition presents a potentially important consequence because of the strong RNA-DNA interaction which may impede the complete replication of the H-strand. Therefore we postulate a separate mitochondrial RNase H activity that is involved in primer RNA removal. It is also conceivable that RNase MRP, given its ability to cleave at multiple sites, may be involved in primer RNA excision once DNA synthesis has initiated; this remains to be tested. It is also interesting that the nascent DNAs in our heterologous DNA synthesis system appear to be unstable after ribonuclease treatment (Fig. 4, lanes  7 and 8). These findings suggest that the RNA-DNA hybrid may influence the stability of the nascent DNA chain on the template DNA strand, perhaps by functioning as a tether that resists branch migration. In this way, such an RNA-DNA association may contribute to the stability of nascent H-strands in vivo as manifested in the three-stranded "D-loop" structure of mtDNA (1). A more complete understanding of how initiation is controlled requires further characterization of the mitochondrial transcription and DNA replication machineries.
Our in vitro model system has permitted the clear demonstration of transcription-coupled DNA synthesis that requires a critical RNA processing step. The remarkable correlation of in vitro cleavage sites on the model R-loops with the known RNA to DNA transitions at the mtDNA O H suggests that the RNase MRP reaction proceeds in vivo. How substrate recognition is achieved by RNase MRP remains unclear, but our current work implies that a specific higher order RNA structure is important. Functionally, the mtDNA initiation mechanism is similar to those of certain prokaryotic systems such as the ColE1 replicon and bacteriophage T7 where RNA-DNA hybrids are critical intermediates of transcription-generated RNA primers. For ColE1, the large (555 nt) mature primer is produced by the action of RNase H (23,24); while for T7, replication occurs by the sequential actions of phage-encoded RNAP and DNAP presumably via the formation of transient but sufficiently stable RNA-DNA hybrids (25,26). The mitochondrial system is unique in this regard, requiring a ribonucleoprotein enzyme to generate the mature primer termini. Why such an enzyme has been evolutionarily selected to process the mitochondrial primer RNAs is enigmatic, but the highly organized structure of the preprimer RNA suggests a unique substrate-enzyme interaction.
Phylogenetic sequence comparisons have suggested the presence of tRNA-like structures at known or presumed O H regions of several mammalian species (27). The presence of these structures is intriguing because of biochemical and antigenic similarity of RNase MRP to the universally found ribonucleoprotein RNase P, which catalyzes the 5Ј maturation of pre-tRNAs (28). Our initial R-loop structural studies indicate that the RNA structure is not confined to a canonical RNA-DNA hybrid throughout its length, and it is likely to exist as a specifically FIG. 5. Primer extension analysis of in vitro synthesized nascent DNA chains. R-loop synthesis, RNase MRP processing, and T7 DNAP reactions were carried out as described but in the absence of radionuclides. Products were further processed as described and hybridized to a 5Ј-end-labeled oligonucleotide complementary to position 294 -314 of the H-strand (see "Experimental Procedures"). The oligonucleotide was extended with AMV reverse transcriptase (lanes 1-5). The same primer was used for dideoxyribonucleotide sequencing. Lanes G, A, T, and C indicate the respective chain terminator. Lane 1 is the circular DNA template (pK408SP) alone. Lane 2 is the SP6 RNAP reaction alone. Lane 3 is the RNase MRP ϩ T7 DNAP reaction. Lane 4 is the RNase MRP ϩ T7 DNAP product sample treated with RNase T2. Lane 5 is the RNase MRP ϩ T7 DNAP reaction treated with DNase I. Arrowheads indicate the positions of RNA to DNA transitions corresponding to RNase MRP cleavage sites 5, 6, and 7. folded entity. We suggest that this folded structure is the substrate feature recognized by RNase MRP. Additional speculation on a common ancestral past between RNase MRP and RNase P is based on the proposed nucleolar function of RNase MRP in rRNA processing in S. cerevisiae (29 -31). Indeed, it is probably more than coincidental that, for mammalian mtDNA, the two rRNAs are immediately separated and flanked by tRNAs. How the distinct nuclear and mtRNA processing roles of RNase MRP are coordinated and regulated remains a challenging question for further investigation.