Bidirectional replication initiates at sites throughout the mitochondrial genome of birds.

Analysis of mitochondrial replication intermediates of Gallus gallus on fork-direction gels indicates that replication occurs in both directions around circular mitochondrial DNA. This finding was corroborated by a study of chick mitochondrial DNA on standard neutral two-dimensional agarose gels, which yielded archetypal initiation arcs in fragments covering the entire genome. There was, however, considerable variation in initiation arc intensity. The majority of initiation events map to regions flanking the major non-coding region, in particular the NADH dehydrogenase subunit 6 (ND6) gene. Initiation point mapping of the ND6 gene identified prominent free 5' ends of DNA, which are candidate start sites for DNA synthesis. Therefore we propose that the initiation zone of G. gallus mitochondrial DNA encompasses most, if not all, of the genome, with preferred initiation sites in regions flanking the major non-coding region. Comparison with mammals suggests a common mechanism of initiation of mitochondrial DNA replication in higher vertebrates.

The study of initiation of replication in prokaryotes, viruses, and plasmids led to the discovery of a single discrete origin for each genome, which was essential for replication (1). The large chromosome size of nuclear DNA necessitates multiple origins of replication (2); nevertheless, it was widely assumed that nuclear DNA would initiate replication at a discrete site for a given region of DNA. The identification of autonomously replicating sequence elements seemed to support this idea (3). However, nuclear DNA replication often initiates at a multiplicity of sites across a broad zone (4 -7). Even the EBV genome, which contains a site that behaves like a classic discrete origin when cut out of the genome and placed in a plasmid, is replicated from numerous origins distributed over a broad zone when it is intact (8). Nor is the initiation zone size-fixed; in flies, the initiation zone size is dependent on developmental stage (9). Recently, we reported that the 16.5-kb circles of mammalian mitochondrial DNA (mtDNA) initiate replication from multiple sites across a zone of ϳ4 kilobases (10).
Although there is great diversity in the size and organization of mitochondrial genomes of plants, fungi, and animals (11), within the animal kingdom they are very similar (12), particularly so among vertebrates (13). One might therefore anticipate a conserved mechanism of replication for vertebrate mtDNA.
In the 1970s and 80s, a series of studies of mammalian mtDNA gave rise to a strand-asynchronous model of mtDNA replication (14). The model proposed that replication of the two strands of the circle arose in each case from a single initiation site. These sites were designated the heavy and light strand origins of replication (O H 1 and O L ). The site of second-strand synthesis, O L , is a sequence element that can theoretically form a hairpin stem-loop and to which abundant free 5Ј ends map. Its primary sequence is poorly conserved in mammals, yet the ability to form a stem-loop is conserved (15).
In its simplest or most literal form, this model cannot apply to all vertebrates. Avian mtDNAs lack a convincing stem-loop structure that might act as a dedicated initiation site for second-strand synthesis, begging the question by what mechanism is avian mtDNA replicated? Birds do share with mammals the abundant short displacement or D-loop form of mtDNA (16,17), which until recently was widely regarded as supporting the strand-asynchronous model. As in mammals, the 3Ј end of the D-loop is close to one end of the major non-coding region (NCR), and the 5Ј end of the D-loop defines O H . At 16,775 nucleotide pairs (np), the mitochondrial genome of Gallus gallus is slightly larger than that of most mammals (18). Avian D-loops, which typically span almost 800 nucleotides (np 75-855 in G. gallus), account for much of the length difference. The high degree of similarity in gene content and arrangement between avian and mammalian mtDNA is illustrated in Fig. 1.
The strand-asynchronous model of mammalian mtDNA replication has been challenged in recent years (19,20). Duplex replication intermediates were identified that could not be reconciled with the strand-asynchronous model (19), and the co-existing population of partially single-stranded molecules was subsequently shown to arise from RNase H degradation occurring during isolation of replication intermediates, which contain extensive tracts of RNA-DNA hybrid (20). Extensive ribonucleotide incorporation has not been reported previously, and it was not known whether this feature was unique to mammalian mtDNA.
The contentious nature of recent findings in mammals (21, * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. 22), coupled with the dearth of information on mtDNA replication mechanisms in other vertebrates prompted us to investigate replication intermediates derived from purified mitochondria of the avian G. gallus. The results indicate that bidirectional replication occurs at sites dispersed throughout the mitochondrial genome of G. gallus giving rise to replication forks that travel in both directions around the circle of mtDNA. However, initiation of replication is not random, as most initiation events map to a region adjacent to the 3Ј end of the D-loop, centered on the NADH dehydrogenase 6 gene.

EXPERIMENTAL PROCEDURES
Purification of Chicken Liver Mitochondria-Chicken liver mitochondria were isolated first by a crude differential centrifugation procedure and subsequently on a single step sucrose density gradient. 20 -30 chick livers were minced finely with scissors and washed three or more times with 100 ml of homogenization buffer (HB) comprising 225 mM mannitol, 75 mM sucrose, 10 mM Tris-HCl, pH 7.6, 1 mM EDTA, and 0.1% bovine serum albumin (fatty acid-free). After washing the chopped liver was suspended in 10 ml/g of wet weight of HB buffer and subjected to six strokes of Dounce homogenization with a tight fitting pestle. All operations were performed at 4°C or on ice. The homogenate was centrifuged at 1,000 ϫ g max for 5 min, and the supernatant recentrifuged at 9,000 ϫ g max for 10 min. The crude mitochondrial pellet was resuspended in 30 -50 ml HB and loaded onto a sucrose gradient (1.0/ 1.5 M) and centrifuged at 40,000 ϫ g max for 120 min. Mitochondria were removed from the interface, diluted in HB, and pelleted by centrifugation at 12,000 ϫ g max for 15 min. Purified mitochondria were resuspended in 75 mM NaCl, 50 mM EDTA, pH 8.0, 1% SDS, 0.5 mg/ml proteinase K and incubated either for 2 h at 50°C. Nucleic acid was extracted from the protease-treated mitochondrial pellet by successive incubations with phenol and chloroform, ethanol precipitated, washed and dried, resuspended in 10 mM Tris-HCl, 1 mM EDTA, pH 8.0, and stored at Ϫ20°C. Mouse mtDNA was isolated by the same method (20).
DNA Digestions-DNA prepared from mitochondria was digested with one of a number of restriction endonucleases under conditions recommended by the manufacturer (New England Biolabs). Where indicated RNase H (Promega) treatment was 1 unit of enzyme for 1 h at 37°C with 0.1-1.2 g of mtDNA. RNase One (Promega) treatment was 5 units of enzyme for 10 min at 37°C with 0.1-1.2 g of mtDNA. S1 nuclease (Promega) treatment was 1 unit for 1.5 min at 37°C. For RNase H digestion samples were incubated in 20 mM HEPES-KOH, pH 7.8, 50 mM KCl, 10 mM MgCl 2 , 1 mM dithiothreitol. RNase One and S1 nuclease treatments employed buffers supplied by the manufacturer (Promega).
Two-dimensional Agarose Gel Electrophoresis and Hybridization-Neutral/neutral two-dimensional agarose gel electrophoresis (AGE) was performed by the standard method (23). For fragments of 3-4 kbp, first dimension electrophoresis was 0.7 V/cm for 20 h at room temperature in a 0.4% agarose gel, and second dimension electrophoresis was 6 V/cm for 4 h at 4°C through a 1% agarose gel, with 300 ng/ml ethidium bromide. In the case of fragments of Ͼ5 kb, first dimension electrophoresis was in a 0.35% agarose gel at 1.5 V/cm for 20 h, and second dimension electrophoresis was at 3 V/cm for 18 h in a 0.875% agarose gel.
In-gel digestion for fork-direction gels was carried out as follows, after separation of DNA in the first dimension, the gel lane was excised and washed twice with 10 mM Tris, 0.1 mM EDTA, pH 8.0, for 30 min at room temperature. The gel slice was twice equilibrated with 1ϫ restriction enzyme buffer (New England Biolabs) for 1 h at room temperature. Excess buffer was aspirated, and 100 units of restriction enzyme were added directly to the surface of the gel. After incubation at 37°C for 3 h, a further 50 units of enzyme were added, and the incubation was continued overnight. After overnight incubation, the gel slice was washed first with 10 mM Tris, 1 mM EDTA, pH 8.0, for 30 min at room temperature and then with Tris borate EDTA 1ϫ for 15 min at room temperature. Thereafter, the procedure was identical to standard second dimension electrophoresis for two-dimensional AGE.
Southern hybridization was overnight at 65°C in 0.25 M sodium phosphate, pH 7.2, 7% SDS. Post-hybridization washes were 1ϫ SSC followed by 0.1ϫ SSC, 0.1% SDS, both for 30 min at 65°C. Filters were exposed to x-ray film and developed after 0.5-10 days.

RESULTS
Origin Detection and Mapping-In a previous study, we reported that initiation of mtDNA replication encompasses a broad zone downstream of the 3Ј end of the D-loop, in mammals (10). Therefore, a series of fragments of chick mtDNA was separated by neutral two dimensional AGE and examined for evidence of initiation arcs, with particular emphasis on the region downstream of the 3Ј end of the D-loop. Restriction sites for the fragments analyzed in this study are illustrated on a schematic map of the G. gallus mitochondrial genome (Fig. 1C).
Analysis of a 7.8-kb AflIII fragment of G. gallus mtDNA, spanning np 10,540 -1,603 and including O H , revealed a prominent initiation arc and a weak simple Y arc terminating in a prominent spot above the linear duplex arc ( Fig. 2A, and interpreted in supplemental information). The prominent initiation arc suggests the majority of initiation events occur within the fragment, via a mechanism. The prominent spot above the linear duplex arc can be explained by arrest in the NCR of one of a pair of replication forks before the other fork exits the fragment at np 10,540 or by unidirectional replication in the vicinity of O H . In the former case, O H acts as the terminus of replication. The initiation arc comprises products of coupled leading and lagging-strand replication based on their mobility on two-dimensional AGE and their resistance to single strand nuclease (Fig. 2B).
A shorter DraI fragment (np 14,075-988), again including O H , also gave rise to an initiation arc more prominent than the simple Y arc (Fig. 2C), whereas a 7.4-kb fragment lacking O H yielded a Y arc considerably stronger than the initiation arc ( Fig. 2D). Hence, some initiation events must arise at sites other than O H , which excludes the unidirectional replication from a discrete origin (O H ) as the sole mechanism of replication for G. gallus mtDNA. Still weaker initiation arcs were associated with FauI (np 8,831-16,351) and MscI (np 8,334 -15,947) restriction fragments (Fig. 2, E and F) in which the O H proximal ends are more distant from O H than that of the BsmBI fragment (Fig. 2D). These data can be interpreted in one of two ways. Either G. gallus mtDNA contains a series of unidirectional origins of decreasing strength, extending from O H into the cytochrome b gene, or G. gallus mtDNA contains an initiation zone of bidirectional replication whose center lies close to the 3Ј end of the D-loop region.
Direction of Fork Movement-If the replication of chick mtDNA initiates bidirectionally across a zone adjacent to the 3Ј end of the D-loop, and replication terminates in the NCR, then replication forks will travel in one direction only through frag-ments outside the initiation zone. To test this proposition, fork-direction gels were produced. These gels require an in-gel restriction enzyme digestion treatment between the first and second electrophoresis steps (23,29). Fragments from several regions of the genome, including ones distant from the 3Ј end of the D-loop, gave rise to pairs of Y arcs (Fig. 3, A, D, G, and J) indicating that replication forks enter these fragments from both ends. Therefore, replication forks travel in both directions on G. gallus mtDNA.
In general, one arc was more prominent than the other, and wherever this was the case it was indicative of a majority of replication forks traveling away from the 3Ј end of the D-loop (Fig. 3, A, D, and G) (see supplemental information). The major arc is consistent with bidirectional replication arising from a zone proximal to the D-loop 3Ј end and terminating at or near O H , so that most of the genome is replicated by the counterclockwise-moving fork. It is equally consistent with counter- clockwise unidirectional replication originating in the NCR. In contrast, the other, fainter arc comprising forks traveling toward the 3Ј end of the D-loop (Fig. 3, A, D, and G) fits neither model, rather it necessitates initiation at sites far from the proposed initiation zone.
Unidirectional Versus Bidirectional Replication-Unidirectional replication originating in the NCR would contribute forks traveling exclusively in one direction, whereas a bidirectional initiation zone would generate forks traveling in two directions. Fork-direction analysis of a BsoBI-BsaHI fragment spanning np 12,777-15,152 indicated that most forks travel away from the 3Ј end of the D-loop (Fig. 3G). In contrast, fork-direction analysis of a smaller fragment with one end closer to the 3Ј end of the D-loop (MboI fragment np 14,866 -16,476) revealed two simple Y arcs of approximately equal intensity (Fig. 3J). The difference between the MboI-HincII (Fig. 3J) and the BsoBI-BsaHI (Fig. 3G)  This is entirely consistent with bidirectional initiation at dispersed sites, whereas unidirectional replication from the NCR would yield only one simple Y arc on fork direction gels, and there would be no difference in the results from the BsoBI/ BsaHI and MboI/HincII fragments.

FIG. 2. G. gallus mtDNA replication initiates across a region of several kilobases downstream of the 3 end of the D-loop.
Gradient-purified chick liver mtDNA was digested with a series of restriction enzymes, and the products were separated by two-dimensional AGE. A 7.8-kb AflIII fragment (np 10,540 -1,603) revealed by hybridization with probe 1, without (A) and with S1 nuclease treatment (B). C, probe 1 was also used to reveal a 4-kb DraI fragment (np 14,075-988) containing O H (np 855) near one end. Cleavage with BsmBI yielded a fragment (np 10,092-768) with one end close to but not including O H (D). The samples in E and F were digested and hybridized with probes that detected, respectively, a FauI fragment np 8,831-16,351, and an MscI fragment np 8,334 -15,947. G, the fragments analyzed in A-F are shown aligned with the relevant section of the G. gallus mitochondrial genome (np 7,000 -1,000). The D-loop region is depicted as a broad line whose 5Ј end is O H , the intensity of each initiation arc is indicated by a number of "ϩ" symbols to the right of each restriction fragment.

Mapping of the Initiation Zone Based upon Fork-direction
Gel Data-Comparison of the set of fork-direction gels also provided an independent means of mapping the initiation zone of G. gallus mtDNA. Based on the equal intensity of the two Y arcs associated with the MboI-HincII fragment (np 15,405-16,476) (Fig. 3J), replication initiates either side of the center of the fragment (np ϳ16,000) at equal frequency. That is, approximately half of all initiation events occur at least 1.8 kilobases downstream of the 3Ј end of the D-loop. In contrast, in the BsoBI-BsaHI fragment spanning np 12,777-15,125, the signal from the Y arc corresponding to forks traveling away from the 3Ј end of the D-loop was greater than the signal of the Y arc comprising forks traveling toward the 3Ј end of the D-loop, (Fig. 3G). Therefore, most initiation events originate upstream of the center of this fragment (np ϳ14,000); a conclusion fully compatible with the standard two-dimensional AGE analysis, which predicts an initiation zone adjacent to the NCR.
Initiation of Replication Occurs Genome-wide in Chick mtDNA-The detection of an arc comprising forks traveling toward the 3Ј end of the D-loop in fragments outside the proposed initiation zone was unexpected (Fig. 3, A, D, and G). These arcs could in theory have arisen by a quite distinct mechanism such as rolling-circle replication, which is believed to operate in yeast mtDNA (30 -32). Further analysis of fragments of chick mtDNA from around the genome revealed initiation arcs in fragments covering all regions, however (Fig. 4).
Hence, replication can account for all the replication intermediates of chick liver mtDNA.
Initiation arcs were faintest in fragments far from the initiation zone defined above, such as a DraI fragment spanning np 3,448 -10,746 and an XhoI fragment spanning np 5,456 -12,777 (Fig. 4, A and B). Analysis of fragments of G. gallus mtDNA that included the NCR and the ribosomal RNA genes also yielded an initiation arc (Fig. 4, C and D), which was S1 nuclease-resistant (Fig. 4E). Although these initiation arcs were less prominent than in fragments extending from the NCR in the opposite direction (e.g. Fig. 2A), they were more intense than for fragments far from the NCR (Fig. 4, A and B). Fragments including the rRNA genes, but lacking the NCR had a much reduced initiation arc (Fig. 4F). These data are again compatible with bidirectional initiation across a zone, accompanied by fork arrest in the NCR. In this case, the zone is located upstream of the D-loop 5Ј end (O H ) and includes the ribosomal RNA genes.
The absence of a descending Y arc in fragments in which the NCR is centered (Fig. 4G) indicates that replication forks rarely traverse the NCR in either direction. Therefore the NCR acts as a bidirectional, or bipolar, replication fork barrier.
In the light of these findings it was important to determine whether we had overlooked initiation upstream of the 5Ј end of the D-loop in mammals. Analysis of a Bsu36I fragment of mouse mtDNA spanning np 16,022-6,350 revealed a very faint initiation arc (Fig. 5A) as did another similar restriction fragment (MscI np 15,190 -4,130) (Fig. 5B). NspI cleaves mouse mtDNA at three sites yielding three fragments of roughly equal size. The most prominent initiation arc was associated with the fragment spanning np 10,758 -114 (Fig. 5C), which includes the NCR and the previously defined initiation zone (10). However, the NspI fragment of mouse mtDNA spanning np 6,092-10,758 also yielded an initiation arc, albeit a weak one (Fig.  5D). Nevertheless, unlike chick, some fragments of mouse mtDNA were not associated with an initiation arc, including the np 114 -6,092 NspI fragment (Fig. 5E). Faint termination arcs (33) were also detected in a number of fragments of G. gallus mtDNA (Figs. 2E, and 4, A, B, D, and F) suggesting that termination events may be as dispersed as initiation events.
Replication Initiation Point Mapping-LM-PCR has been used widely to map free 5Ј ends of DNA. These species are candidates for transition points from RNA to DNA synthesis and can thereby define sites of initiation of replication (34). exonuclease pretreatment degrades spurious ends created by nicking, making replication initiation point mapping a more reliable method for detecting start sites (35). Attempts to apply replication initiation point mapping to mammalian mtDNA have been confounded by the high ribonucleotide content of a majority of mammalian mitochondrial replication intermediates 2 ; however, conventional ribonucleotide-sparse replication intermediates were more abundant in G. gallus mtDNA, making the technique potentially applicable.
Although initiation zones allow origin firing at multiple positions there are recognized preferred start sites in nuclear DNA (36 -38). Therefore, we focused on the ND6 gene, which appeared to be the center of the initiation zone (see "Discussion"). Standard LM-PCR revealed a host of free 5Ј ends (Fig. 6,  lanes 1 and 3). Replication initiation point mapping simplified the results, revealing persistent prominent free 5Ј ends, at np 16,184,16,257,16,492,16,567, and 16,581 of the NADH 6 gene, and np 16,122, and 16,126 of tRNA Pro gene (Fig. 6, lanes 2 and  4). Note that these free ends are on the L-strand, which will often be the leading strand for the early arresting fork in our 2 M. Yang and I. J. Holt, unpublished experiments.

FIG. 4. Initiation events occur throughout G. gallus mtDNA.
A-H, Brewer-Fangman gels of fragments of chick liver mtDNA. A and B, a DraI fragment spanning np 3,448 -10,746 and an XhoI fragment np 5,456 -12,777, hybridized to probes 4 and 5, respectively. C, AvaIdigested chick mtDNA hybridized to probe 3, which detects the fragment spanning np 16,204 -5,389. D, BssSI/PvuII generated a fragment of np 420 -7,254 hybridized to probe 2 (double digestion with PvuII and AgeI generated a fragment spanning np 109 -7,254 that gave essentially the same result, data not shown). E, AvaI np 16,204 -5,389 after treatment with S1 nuclease. model. Thus, the prominent free 5Ј ends detected on the Lstrand are consistent with these loci being preferred sites of initiation within the zone defined by two-dimensional AGE mapping. Replication fork arrest or pausing will also generate prominent free 5Ј ends of DNA; however, frequent fork arrest can be discounted as an explanation of the prominent free 5Ј ends, as there was no evidence of replication pausing in this region of the G. gallus mitochondrial genome (Fig. 2). DISCUSSION Initiation arcs are diagnostic of origins of replication in all systems examined to date (8, 39 -43). The initiation arcs detected in a variety of fragments of G. gallus mtDNA (Figs. 2 and 4, A-F) account for a substantial majority of mitochondrial replication intermediates. Therefore, the major mechanism for propagating mtDNA in chick liver is strand-coupled replication, as per mammalian liver and placenta (10). Accordingly, higher vertebrates appear to employ a common mechanism of mtDNA replication.
The first model of strand-coupled mtDNA replication in mammals erroneously inferred that the association between initiation arcs and fragments containing O H was indicative of unidirectional replication from O H (19). Subsequently, initiation arcs were found in fragments lacking O H , and this led to characterization of an initiation zone of bidirectional replication downstream of the 3Ј end of the D-loop (10). The latest findings necessitate further revision of the model, as forkdirection gels indicate that replication occurs in both directions throughout the mitochondrial genome of G. gallus (Fig. 3), and moreover, standard two-dimensional gels revealed a fraction of initiation events mapping to the region immediately upstream of the 5Ј end of the D-loop in chick and mouse mtDNA (Figs. 4 and 5). Thus, there are two zones of initiation flanking the NCR, which can be represented as a single zone interrupted by a termination region, which may be synonymous with the NCR (Fig. 7A).
Center of the Initiation Zone-The decrease in initiation arc intensity with distance from O H was more pronounced in chick mtDNA (Fig. 2) than that of mouse (10), suggesting that the center of the initiation zone is displaced toward the D-loop 3Ј end in G. gallus relative to mammals (Fig. 7B). There is one major structural difference between the mtDNA of birds and mammals, which could explain the different map positions of the center of the initiation zone. In birds, including G. gallus, ND6 lies closer to the NCR than in mammals, because of the difference in gene order of ND6 and cytochrome b (Fig. 1 and Ref. 18). Therefore, if the center of the initiation zone is located in the ND6 gene in both classes of organism, then most initiation events will arise 150 -650 np from the 3Ј end of the D-loop in birds, whereas in mammals replication initiation events will be concentrated over a kilobase further downstream of the 3Ј end of the D-loop. Initiation point mapping in the region of the chick ND6 gene identified several prominent free 5Ј ends of DNA (Fig. 6). Although, free 5Ј ends of DNA are not diagnostic of initiation of replication, they are predicted products of replication initiation. Therefore, their presence in a region mapping to the center of the initiation zone makes them good candidates for preferred sites of initiation of replication, particularly given the absence of replication pausing in this region (Fig. 2).
ND6 is unique in vertebrates in being the only protein encoded on the L-strand of mtDNA. ND6 is also encoded on the opposite strand to all the other mitochondrial proteins of some invertebrates (e.g. Ref. 44), but other proteins are encoded on the same strand as ND6 in insects (45), whereas in urochordates all 13 proteins are encoded on one strand (46). Further study of mtDNAs of these other animals could help clarify the role of the ND6 gene in mtDNA replication.
Extent of the Initiation Zone in Mouse mtDNA-Initiation arcs were detectable genome-wide in chick and the new data from mouse indicate a significantly broader initiation zone than described previously (10). Nevertheless, several frag-ments of mouse mtDNA lacked a detectable initiation arc, the NspI fragment np 114 -6,092 (Fig. 5E) and the XbaI and BamHI fragments spanning np 953-8,529 and 4,275-11,167, respectively (10). These fragments define an initiation-free region of mouse mtDNA spanning approximately np 3,000 -7,700.
All the data in this report are consistent with an initiation zone centered close to the 3Ј end of the D-loop in G. gallus mtDNA. The faint initiation arcs in fragments far from the center of the initiation zone can be regarded as the diminishing effect of the zone with distance from the center. The influence of the initiation zone extends throughout the mitochondrial genome of the chick and most of that of mouse.
Mitochondrial DNA Replication in Other Animals-The similarities between avian and mammalian mtDNA replication reported here suggest that replication may well be the common mechanism for all vertebrate mtDNAs. As for invertebrates, it is known only that echinoderm (44) and malarial mtDNA (47) yield conventional simple replication fork arcs, implying that invertebrate mtDNA replication involves coupled leading and lagging strand synthesis. Therefore, it remains to be determined whether or not replication is the universal mechanism of mtDNA replication of animals.
In conclusion, the replication of an avian mitochondrial genome occurs via replication. G. gallus mtDNA replication is bidirectional and arises from an initiation zone straddling the NCR. The influence of the initiation zone pervades the entire FIG. 7. Initiation of replication in mitochondrial DNA of higher vertebrates. A, schemes I and II depict models of how replication initiation might operate in G. gallus mtDNA. I, multiple unidirectional origins of different strength. The arrowhead indicates the direction of replication; the size of the arrow corresponds to frequency of initiation attributable to the region assuming unidirectional replication (based on standard twodimensional AGE, see Fig. 2 and text for details). II, bidirectional replication initiates at multiple sites across a region straddling the major NCR or D-loop region. Two pairs of arrows pointing in opposing directions denote two potential sites of initiation of bidirectional replication. In all cases, the D-loop or NCR acts as the terminus. The data from standard two-dimensional AGE cannot discriminate between models I and II; however, only scheme II is compatible with the data from fork-direction gels, and moreover model II, but not model I, predicts prominent free 5Ј ends in the ND6 gene (see Figs. 3 and 6, and "Results" for details). B, the bidirectional initiation zone of avian and mammalian mtDNA. Circular maps of chicken and mouse mitochondrial genomes with the proposed prominent initiation zone (Ori-z) marked by an unfilled two-headed arrow inside the circle. The center of the initiation zone maps approximately to the ND6 gene in both classes of organism; however, the different arrangement of cytochrome b and ND6 genes places the center closer to the NCR in birds than mammals. ND, gene encoding a subunit of NADH dehydrogenase; cyt b, cytochrome b gene, 12S rRNA, 12S ribosomal RNA gene. mitochondrial genome in the chick and most of that of mouse. The center of the zone maps to the ND6 gene, and prominent free 5Ј ends of DNA are found in this region, which represent candidate start sites of DNA replication.