Structural characterization of site-specific discontinuities associated with replication origins of minicircle DNA from Crithidia fasciculata.

The kinetoplast DNA of trypanosomes is comprised of thousands of DNA minicircles and 20-50 maxicircles catenated into a single network. Replication intermediates of minicircle DNA from the trypanosomatid species Crithidia fasciculata contain site-specific discontinuities in both heavy (H) and light (L) strands. These discontinuities map to two small regions situated 180 degrees apart on the minicircle; each region has two sites at which a discontinuity can occur, one on each strand. We have determined the position of these discontinuities on the minicircle DNA sequence and have characterized their structure. H-strand discontinuities occur within a 4-5-nucleotide sequence and consist of single nicks, only one of which appears to be a DNA-DNA junction. Characterization of the remaining H-strand nicks indicates a structure other than a typical DNA-DNA or DNA-RNA junction. Discontinuities on the L-strand can be either a nick or a short gap which overlaps a 12-nucleotide sequence universally conserved among minicircles from various trypanosome species. Up to 6 nucleotides are hydrolyzed from the 5' terminus facing the gap upon treatment with alkali, suggesting the presence of an RNA primer. Based on the structures of minicircle replication intermediates, we present a model for replication of minicircle DNA in which the site-specific discontinuities closely coincide with the origins of replication.

The DNA minicircles of most kinetoplastid species are heterogeneous in sequence. However, in the C. fasciculuta strain Cf-C1 greater than 90% of the minicircles belong to a single class which is nearly homogeneous in DNA sequence (7). Therefore, it is possible to localize sequences of possible biological importance precisely and to determine the structure of minicircle replication intermediates at the nucleotide level.
The general mechanism for replication of kinetoplast DNA networks has been revealed through in uiuo studies, primarily on C. fasciculata. Minicircles detach from the network, replicate as free molecules (8,9), and then reattach to the network periphery (10,11). The newly replicated network-associated minicircles contain both nicks and gaps. Thus, as replication progresses, a concentric wave of replication sweeps the kinetoplast DNA network, doubling its size, and transforming it from a covalently closed form to a nicked-relaxed form (5,12). I n uiuo, the free minicircle population contains a class of replication intermediates termed smear or gapped DNA; in these molecules the nascent H-strand has numerous nicks and short gaps (9,13). However, no such intermediate has been observed for the L-strand which appears to be synthesized as a continuous DNA chain. Kinetoplasts isolated from C. fasciculuta and incubated in an ATP-dependent DNA synthesis reaction give rise to minicircle DNA replication intermediates very similar to those described above. Furthermore, minicircle molecules containing nascent half-length (1.25 kb) H-strands are generated during the kinetoplast reaction (14,15). Similar size nascent DNA chains of unknown strand specificity are associated with replicating kinetoplast DNA networks in vivo (13).
Minicircles replicated in the kinetoplast system contain discontinuities in both H and L nascent DNA strands. Each strand has two sites spaced 180" apart on the minicircle at which a discontinuity can occur. The sites of discontinuity on opposite strands do not overlap but are separated from one another by about 100 base pairs (14,15). In Trypanosoma equiperdum minicircles, a gap of approximately 10 nucleotides is observed in one of the nascent DNA strands. This gap overlaps a 12-nucleotide sequence conserved between species and found in all minicircles sequenced to date (16).
Replication intermediates of animal cell mtDNA have been shown to contain site-specific discontinuities and to accumulate distinct size nascent strands (17,18). Such structures are remarkably similar to those described above for minicircle The abbreviations used are: kb, kilobases; L-strand, light strand; H-strand, heavy strand. DNA replication intermediates and may reflect similarities between their modes of replication.
Detailed characterization of the site-specific discontinuities found in C. fasciculata minicircles is presented here. These results show that the discontinuities coincide with sites of initiation of minicircle DNA replication, indicating that both chain initiation and termination occur at the replication origins.

EXPERIMENTAL PROCEDURES
Materiak-Brain heart infusion medium was purchased from Difco and hemin from Sigma. The unlabeled deoxynucleoside triphosphates and ATP were obtained from Pharmacia  Cell Growth-C. fasciculata Cf-Cl cells (7) were cultured in Difco brain heart infusion medium supplemented with 10 pg/ml of hemin at 28 "C with shaking.
Kinetoplust DNA Synthesis Reaction-Kinetoplasts from C. fasciculata Cf-C1 cells were isolated and incubated in a DNA synthesis assay for 30 min at 30 "C in the presence of unlabeled deoxynucleoside triphosphates (14,15). The reaction was quenched, and the free minicircle fraction was isolated on sucrose gradients as described (14). The free minicircle fraction was electrophoresed on an agarose gel, and the nicked-circular form (RF 11) and the covalently closed relaxed circular form (RF IV) of free minicircles, the starting material for all subsequent experiments, were isolated separately by electroelution.
Polyacrylamide Gel Electrophoresis-Ethanol-precipitated DNA samples were resuspended in 99% formamide, 0.1 X TBE, 0.1% xylene cyanol, and 0.1% bromphenol blue. Just prior to electrophoresis the samples were heated for 5 min at 100 "C and quenched on ice. DNA was electrophoresed in 8% polyacrylamide gels in the presence of 7 M urea at 40-45 W constant power. The acrylamide/bisacrylamide ratio was 19:1, the gel dimensions were 37 cm X 17 cm X 0.3 mm, and the buffer system used was TBE.
DNA Sequencing-The right-hand StuI-XhoI minicircle DNA fragment (see Fig. la) was previously inserted into the M13 vectors mplO and mpll, producing the clones M13 CFKl2O (L-strand) and M13 CFKl2OC (H-strand), respectively (7)? Single strand M13 CFK120 phage DNA was primed with a complementary 21-base oligonucleotide (5'CGCGTCGTTGTTAATTTTGCC), whereas that of M13 CFK120C was primed with a complementary 15-base oligonucleotide (5' CTAGAAATCAAAGCT). The oligonucleotides were synthesized on a model 380A DNA Synthesizer from Applied Biosystems, purified on a 20% polyacrylamide gel, and isolated using the crush and soak method (21). Sequencing of the primed templates was performed by the dideoxy chain termination method of Sanger et al. (22).
Enzymatic Reactions-Unless stated otherwise, all reactions were treated as follows. Each reaction contained 0.1-0.3 pg of DNA in a reaction volume of 20 pl. After incubation the reactions were quenched by addition of 5 p1 of EDTA (0.2 M). The reaction volume was then brought up to 100 p1 with TE, and the sample was extracted one time with chloroform/isoamyl alcohol (241) and ethanol-precipitated in the presence of 0.3 M sodium acetate.
Restriction enzyme digests were carried out in a volume of 40 p1 as recommended by the supplier. Ligations were performed for 4 h at room temperature. T4 DNA ligase reactions contained 50 mM Tris, pH 7.4, 10 mM Mg& 10 mM dithiothreitol, 1  Alkaline Hydrolysis-After the first ethanol precipitation to eliminate unincorporated label (see above), specified DNA samples were resuspended in 10 p l of water. An equal volume of 0.6 M KOH and 2 mM EDTA was added, and the sample was incubated for 12-20 h at 37 "C. The sample was neutralized by addition of 10 pl of 0.6 M HCl and ethanol-precipitated for a second time in the presence of 2 M ammonium acetate. Mock hydrolysis reactions were performed in TE buffer at 37 "C, and the acid neutralization step was omitted.
Autoradiography-The polyacrylamide gels were dried prior to exposure on Kodak x-ray film at -70 "C in the presence or absence of a single intensifying screen. Longer autoradiographic exposures of the gels allowed visualization of each step on the nucleotide ladder. Therefore, the exact size difference between fragments containing the same 5'-or 3'-end could be determined.

RESULTS
Site-specific Discontinuities Lie within Conserved Regwns-Kinetoplasts isolated from the C. fasciculata strain Cf-C1 replicate minicircle DNA when incubated in a DNA synthesis reaction (14,15). The nascent DNA chains produced during the reaction contain site-specific discontinuities which were localized on a restriction enzyme cleavage map ( Fig. la) of the major minicircle class found in this strain (14). DNA sequence analysis has shown that the discontinuities are contained within imperfect direct repeats of approximately 170 base pairs. These direct repeats are present in all minicircles that have been sequenced from this strain.' The regions containing the conserved direct repeats, designated A and B, are depicted in expanded scale in Fig. 1 b and c, respectively, along with a detailed restriction map for each region deduced from the DNA sequence. Most of the cleavage sites between and immediately adjacent to each pair of Hand L-strand discontinuity sites are seen to be conserved between regions A and B.
Determination of 3'-End Points of B Region Discontinuities-Nicked minicircle DNA (RF 11), prepared from kinetoplasts reacted in the presence of unlabeled dNTPs, was cleaved with MluI or XbaI and kinase-labeled at the resulting 5'-ends. A schematic of the B region discontinuities with the adjacent MluI and XbaI sites cleaved and labeled on the Hand L-strands, respectively, is shown in Fig. 2a. In reality, an individual minicircle DNA molecule would not contain both discontinuities, since it would have only one nascent strand.
Utilizing the known sequence from the minicircle B region,' oligonucleotide primers, a 21-mer and a 15-mer (see "Experimental Procedures"), were synthesized to be complementary to the DNA sequences directly adjacent to the MluI or XbaI sites, respectively. Furthermore, the 5'-ends of the primers are identical in sequence to those generated by cleavage of minicircle DNA with MluI or XbaI. Single-stranded DNA minicircle clones M13 CFK120 (L-strand) and M13 CFKl2OC (H-strand), containing the B region, were primed with the complementary oligonucleotide (Fig. 2b) and sequenced.
The H-strand-sequencing ladder, primed from the 21-mer, was electrophoresed on a sequencing gel alongside the minicircle sample that had been end-labeled only at the MluI site (Fig. 3a). A series of labeled H-strand DNA fragments is observed spanning a 4-5-base sequence and differing in length by 1-nucleotide increments. The two largest fragments are more abundant than the shorter ones, forming a sharp cutoff to the upper size limit of the fragments.
In an analogous experiment, the L-strand-sequencing ladder was electrophoresed alongside the minicircle sample that had been end-labeled only at the XbaI site (Fig. 3b). Two distinct groups of L-strand fragments are seen. The groups are about 11 nucleotides apart and each group contains two prominent fragments separated by 1 nucleotide. A small degree of sequence heterogeneity exists within the major minicircle class (7). There is a large subpopulation of the major minicircle class containing a single base deletion with respect to the minicircle clone M13 CFKl2OC in the region between the XbaI site and the L-strand discontinuity (data not shown). This most likely results in the L-strand doublets observed in Fig. 3b, the upper fragment of each pair being derived from M13 CFK120C-like molecules and the lower fragment from the subpopulation containing the one base deletion. Labeling of the other 5'-ends produced upon cleavage with MluI or XbaI also occurs in these experiments, but the fragments labeled are so large that they do not interfere with the analysis.
Structural Characterization of B Region Discontinuities-The actual DNA structure(s) at the sites of H-and L-strand discontinuity in the B region was probed by enzymatic means. Fig. 4 shows DNA samples which were treated with T4 DNA ligase, E. coli DNA ligase, or T4 DNA polymerase prior to cleavage and then 5'-end-labeled at both the MluI and XbaI sites. All of the H-strand fragments and the larger pair of Lstrand fragments disappear upon treatment with T4 DNA ligase, indicating that the discontinuities at those locations are nicks and not gaps (Fig. 4a). The smaller set of L-strand fragments is unaffected by T4 DNA ligase but is elongated about 7 nucleotides by treatment with T4 DNA polymerase, generating a pair of fragments intermediate in size between the larger and smaller L-strand fragments. Thus, the smaller L-strand fragments seem to border a short gap and terminate with 3'-hydroxyl groups capable of priming T4 DNA polymerase.
Stepwise treatment of the DNA sample with T4 DNA ligase and T4 DNA polymerase, in either order, generates L-strand fragments of only the intermediate size (data not shown). Therefore, the intermediate size fragments represent the full extent of L-strand elongation by T4 DNA poymerase, and the junction so formed can not be sealed by T4 DNA ligase. The polymerase treatment also appears to cause a slight amount of strand displacement at the H-strand nicks, resulting in reduced levels of all but the largest fragment (Fig. 4a) and generating a junction which is resistant to sealing by T4 DNA ligase (data not shown).
In Fig. 4b the DNA sample was treated with E. coli DNA ligase prior to cleavage and end labeling. The absence of the largest H-strand fragment indicates that only the 3'-most Hstrand nick is suitable for ligation by the E. coli enzyme, whereas the others remain open. L-strand nicks are closed by E. coli DNA ligase and, as expected, L-strand fragments bordering the gap are not affected. Thus, the L-strand nicks and the 3'-most H-strand nick appear to be DNA-DNA junctions with juxtaposed 3'-hydroxyl and 5'-phosphoryl end groups.
E. coli DNA ligase differs from the T4 enzyme in that it can not ligate a nick in which the 5' terminus is a ribonucleotide and the 3' terminus is a deoxyribonucleotide (24). Incubation of the DNA sample under alkaline conditions, after treatment with T4 DNA ligase and end labeling, did not regenerate any of the H-strand fragment (data not shown). Therefore, those H-strand nicks which are closed by T4 DNA ligase but not by E. coli DNA ligase do not represent simple DNA-RNA junctions, a structure which should be hydrolyzed under these conditions.
Elongation of the lower L-strand fragments by T4 DNA polymerase stops short of the L-strand nick sites (Fig. 4a).
To examine the structure of the 5' terminus facing the Lstrand gap, the DNA samples in Fig. 5 AuaII sites when labeling with dATP, all other labeled fragments were either too large or too small to interfere with the subsequent analysis. In Fig. 5, the 123-and 155-nucleotide fragments are 3'end-labeled at the AvaII site mentioned above. Their 5'-ends are generated upon cleavage of the MspI (123-nucleotide fragment) or AuaII (155-nucleotide fragment) sites directly flanking the B region L-strand discontinuity. Since these fragments share the same 3'-end and DNA sequence as those generated by the L-strand discontinuities, they serve as accurate size markers from which the exact 5'-end(s) of the discontinuity can be determined. The sample in the righthand control lune is derived from covalently closed minicircle DNA. Fragments common to the experimental lanes and to this lane are the result of cleavage of the major and minor minicircle classes and are not due to strand discontinuities. The untreated DNA sample in Fig. 5 contains a prominent L-strand fragment 3 nucleotides longer than the 123-nucleotide marker. Based on the data obtained by mapping from the opposite side using XbaI (Fig. 3b), this is exactly the position predicted for the 5' terminus of the L-strand nick. After treatment with T4 DNA ligase, this fragment is no longer present, confirming that its 5'-end lies at the L-strand nick. In addition to the nick fragment, a pair of faint L-strand fragments (arrowheads) is observed with 5'-ends lying 6-7 nucleotides past the nick. T4 DNA ligase has little or no effect on this pair of fragments. However, incubation of the ligase-treated sample under alkaline conditions results in their loss and the appearance of a new prominent L-strand fragment migrating as though it was a little more than 1 nucleotide longer than the nicked fragment in the untreated sample. These results are consistent with the 5' terminus of the L-strand gap being comprised of an RNA primer of 5-6 nucleotides.
This alkali sensitive 5' extension is most likely responsible for limiting elongation of the 3' terminus of the L-strand gap by T4 DNA polymerase, thus generating the intermediate size L-strand fragments (Fig. 44. The 3' termini of the intermediate size L-strand fragments lie beyond the 5' termini of the alkali-sensitive nucleotides. Therefore the T4 DNA polymerase seems to have strand-displaced the first 1 or 2 alkalisensitive nucleotides. A  Region Discontinuities-Fig. 6 shows the analysis of the 3'-ends of the H-strand discontinuities in the A region. DNA samples in Fig. 6 were treated as indicated in the figure legend and 5l-end-labeled on the H-strand at the RsaI site designated in Fig. lb. The 3'-end of the 138-nucleotide marker lies directly adjacent to the H-strand discontinuity and was generated by cleavage with AuaI. Two fragments flank the position of the 138-nucleotide marker in the experimental samples. They are also present in the control sample derived from RF IV minicircle DNA and so are not due to strand discontinuities. To eliminate a minicircle DNA fragment which overlaps the size range of fragments to be analyzed, the indicated samples were digested with TuqI.

Sequence Localization and Structural Characterization of
Two H-strand fragments, terminating 9 and 10 nucleotides past the 3'-end of the 138-nucleotide marker, are observed in the untreated sample (Fig. 6). This places the 3'-ends of these H-strand fragments at the same relative sequence position as was determined for the two largest B region H-strand fragments (Fig. 9a). Incubation of the DNA sample with T4 DNA ligase resulted in the loss of both H-strand fragments, demonstrating the presence of a nick at the 3' terminus of each fragment. E. coli DNA ligase only sealed the nick at the end of the larger fragment, leaving the smaller fragment uneffected. The results are the same as those obtained from the H-strand discontinuities in the B region.
Analysis of the 3'-ends of the L-strand discontinuities in the A region is presented in Fig. 7. The DNA samples in Fig.  7 were treated as indicated in the figure legend and 5'-endlabeled on the L-strand after cleavage at the Ah1 site designated in Fig. lb. The 3'-end of the 86-nucleotide marker fragment was produced by digestion at the MspI site located between the AluI site and the L-strand discontinuity. A pair of larger fragments present in the same lane are derived from minicircle sequence outside the A region. The remaining DNA samples were digested with Hinff, thus eliminating any confusion of the A region L-strand fragments with similar size fragments extending between the AluI site and L-strand discontinuity within the B region (Fig. IC).
In the untreated sample (Fig. 7), two predominant L-strand fragments are observed (arrowheads). They are separated by 11-12 nucleotides and their 3'-ends lie at 12 nucleotides (smaller fragment) and 23-24 nucleotides (larger fragment) past the 86-nucleotide marker. Exact placement of the 3'-end of the larger fragment is not possible in this experiment due to the extended distance between it and the marker. The

Fragments derived from heavy (H) or light (L) minicircle strands are indicated by the brackets. Panels a and b represent separate experiments.
larger fragment is absent in DNA samples treated with T4 or E. coli DNA ligase, indicating that its 3'-end lies at a nick. The previously stated substrate specificities of these enzymes suggest that this nick represents a DNA-DNA junction. Within the resolution of this experiment, the L-strand nick in the A region occurs at the same nucleotide sequence as was observed for the B region (Fig. 9, b and c, respectively).
Ligase treatment has no effect on the smaller L-strand fragment. However, this fragment is extended 6-7 nucleotides by T4 DNA polymerase, generating a pair of intermediate size L-strand fragments (arrows) similar to those seen in the B region (Fig. 4a). Therefore the 3'-end of the smaller L-strand fragment in region A appears to border a short gap. Fig. 8 shows the analysis of the 5'-end(s) of the L-strand discontinuity in region A. Minicircle RF I1 DNA was digested with both XhoI and Sac11 and the resulting 0.75-kb minicircle Ava II fragment containing the A region (Fig. l a ) was gel purified for the subsequent analysis. This was done to ensure that the L-strand fragments obtained would be derived from the A region and not the B region. The 0.75-kb fragment was aliquoted into several tubes and treated with T4 DNA ligase where indicated. Intermolecular joining between fragments during such ligations will not affect the results. All DNA samples were then cleaved and 3'-end-labeled on the L-strand of the AuaI site designated in Fig. lb. The 5'-end of the 98nucleotide marker was generated by cleavage at the MspI site directly adjacent to the L-strand discontinuity.
In the untreated control (Fig. 8), a prominent L-strand fragment is observed. Its 5'-end lies 3 nucleotides past that of the 98-nucleotide marker and at the position expected for the L-strand nick. This fragment is absent after reaction with T4 DNA ligase confirming that its 5'-end indeed lies at a nick. Therefore, the L-strand nick can now be precisely localized. Its position with respect to the A region sequence is the same as for the L-strand nick within the B region (Fig. 9,  b and c).
In addition to the nick fragment, a series of faint L-strand fragments are detected with 5'-ends lying mostly 6-7 nucleotides past the nick (arrowheuds). These fragments are not affected by ligase treatment. However, they are absent after incubation with alkali and a new fragment is observed, terminating a little more than 1 nucleotide 5' of the nick. Thus, the structures of the A and B region L-strand discontinuities appear to be identical.  Fig. 3. The 138-nucleotide marker fragment is discussed in the text.

DISCUSSION
The C. fasciculata strain Cf-C1 contains a major class of minicircles which are nearly homogeneous in DNA sequence (7). This has made possible the sequence localization and structural characterization of previously identified site-specific discontinuities (14) within minicircle DNA replication intermediates. The regions designated A and B on the minicircle map shown in Fig. lu, contain imperfect direct repeat sequences which encompass the discontinuities.' Fig. 9 presents a summary of the sequence position and structural characterization data of the discontinuities within each region.
Molecules containing an H-strand nick are observed in both the A and B regions (Fig. 9a). Since the sequence shown is identical between the A and B regions, the results from both regions are presented together. The exact site of the nick can vary by a few nucleotides between molecules, but it always lies at one of the positions indicated by the solid arrows. Sites at which nicks occur more frequently are represented by the larger arrows. Both of these major nicked species are always detected. Molecules nicked at the sites indicated by the smaller arrows are less common, and their actual abundance varies between separate minicircle preparations. Although each nick must have juxtaposed 3"hydroxyl and 5'-phosphoryl groups in order to be sealed by ligase, there are differences between them. T4 DNA ligase seals all the Hstrand nicks while E. coli DNA ligase seals only the nick on the right (Figs. 4 and 6). The inability of E. coli DNA ligase to seal the remaining nicks, and the resistance to alkaline hydrolysis of bonds formed at these nicks by T4 DNA ligase (data not shown), indicate that they do not represent simple DNA-DNA or DNA-RNA junctions. Although their structure remains unknown, modified nucleotides may be present at these junctions. Closure of the nick on the right by both DNA ligases is consistent with it being a DNA-DNA junction. The fact that this nick lies the furthest 3' in the sequence suggests that the series of H-strand nicks are the result of a sequential 86 -

5' to 3' excision of modified nucleotides presumably involved in priming of H-strand DNA synthesis.
The sequence location and structure of L-strand discontinuities within the A and B regions are summarized in Fig. 9, b and c, respectively. Numbering of the nucleotides above the sequences is solely for reference. In region A, the L-strand discontinuity can be either a nick or a short gap. A molecule containing both a nick and a gap in region A is shown in Fig.  9b, but in reality such discontinuities would exist at separate sequences. Unlike the H-strand, the L-strand nick occurs at a single site (arrow). This nick is sealed by either T4 or E. coli DNA ligase and so appears to represent a DNA-DNA junction with a 3'-hydroxyl and B'-phosphoryl group.
Molecules containing an L-strand gap have a 3' terminus at nucleotide 20 and predominant 5' termini at nucleotides 13 and 14. Most gaps, therefore, span 5-6 nucleotides, but for simplicity only molecules containing a 5-nucleotide gap are shown. T4 DNA polymerase elongates the 3' terminus of the gap by 6-7 nucleotides indicating that the first 1-2 nucleotides at the 5' terminus of the gap are strand displaced.
Alkaline hydrolysis causes the release of nucleotides from the 5' terminus of the gap, down to and including nucleotide 9, generating a unique 5'-end that maps between nucleotides 8 and 9. The slightly retarded mobility of this DNA fragment on a sequencing gel could be due to the loss of its 5'-phospho- ryl group upon alkaline hydrolysis or to the presence of an otherwise unusual nucleotide at the 5'-end. Whether all nucleotide linkages in the 5"terminal region are alkali-sensitive, or just the linkage between nucleotides 8 and 9, has not been determined.
DNA sequence surrounding the L-strand discontinuities in the B region (Fig. 9c) differs only slightly from that in region A. Therefore, it is not surprising that the B region L-strand discontinuities are identical to those in the A region with respect to sequence location and structure. The only difference is the gap length which is 4-5 nucleotides in region B versus 5-6 nucleotides in region A. This is apparently the result of the B region being shorter by 1 nucleotide than the A region in the area spanned by the gap. Alkali-sensitive nucleotides at the 5' termini of L-strand gaps suggest that L-strand DNA synthesis is initiated by an RNA primer of at least 6 nucleotides. The gap could result from termination of chain elongation prior to encountering the 5' terminus of the "primer" or from partial excision of a more extensive RNA primer. L-strand nicks would represent the final step prior to ligation in which the primer has been excised and the gap filled. Excision of the primers in both Hand L-strands seems ultimately to result in a nick between adjacent deoxyribonucleotides. The adenine residue at nucleotide position 8 in L-strand discontinuities from both regions is not released by alkaline hydrolysis. If this residue is part of the L-strand primer, its structure may resemble the nonhydrolyzable nucleotides observed in the H-strand primer. Another possibility is that the transition from DNA to RNA occurs at the junction of nucleotides 8 and 9 and that the primer excision mechanism results in the removal of the first  deoxynucleotide encountered (i.e. nucleotide 8).

' -A G A G C C C A T C C C C G C A A G
A 12-nucleotide universal minicircle sequence found in all minicircles sequenced to date (from T. brucei (25), T. equiperdum (26), Leishmnia tarentolae (27), T. lewki (28), and C.
fasciculata (15)) is designated by the boxed regions in Fig. 9, b-d. The L-strand nick, the primer, and the 5' terminus of the gap all fall within this sequence. This conserved sequence has been shown to overlap a 10-nucleotide gap in a nascent DNA strand of minicircles from T. equiperdum (16). The 5' terminus of the gap in this species corresponds precisely with the nick site in C. fasciculata, but the presence of a primerlike structure was not reported. This may reflect more effcient primer excision due to the difference in species or to the fact that minicircle replication intermediates analyzed from T. equiperdum were generated in vivo. More  Cf-C1 is presented in Fig. 10. The principal assumption, that site-specific discontinuities in nascent DNA strands coincide with origins of minicircle DNA replication, is strongly supported by results reported here and by Ntambi and Englund (16). In most aspects the model is analogous to the D-loop replication model proposed for unicircular dimeric mouse mtDNA (17). Such dimeric mtDNA molecules contain four replication origins, two on each strand, with origins on the same strand located 180" apart on the circular genome. This is very reminiscent of the minicircle structure in C. fasciculata.
Furthermore, except for the discontinuous synthesis of the second strand (i.e. H-strand), replication intermediates analogous to those observed in minicircles are also observed in the dimeric form of mouse mtDNA. These include site-specific discontinuities and strand-specific nascent chains of several distinct sizes (17).
We propose that replication begins by initiation of L-strand synthesis within region A or B (Fig. loa, u'). Molecules containing newly initiated L-strands within both regions are most likely very rare or nonexistant. Elongation of the nascent L-strand displaces the parental L-strand, exposing the adjacent origin of H-strand synthesis. This allows initiation of Hstrand DNA synthesis which is then elongated about 100 nucleotides to the site of L-strand initiation where further synthesis may be blocked by the duplex nature of the molecule. The displaced parental L-strand also serves as template for multiple, apparently random, secondary initiations of Hstrand synthesis (Fig. lob, b ' ) . Priming of these Okazaki-like fragments would differ from the site-specific priming of Hstrand synthesis in that their primers are readily excised and the fragments joined. As L-strand synthesis proceeds around the minicircle, the second origin of H-strand synthesis is uncovered. A second site-specific initiation of H-strand DNA synthesis occurs at this point, and the chain is elongated until it encounters and is joined to the 5'-end of one of the Okazakilike H-strand fragments (Fig. lOc, c ' ) . When L-strand synthesis nears completion, the daughter molecules segregate. Thus, multiple discontinuities are present in the nascent Hstrand of one daughter molecule (Fig. 10e, e') whereas the other contains a single discontinuity at the origin of L-strand synthesis in the A or B region (Fig. 10d, d'). Complete joining of the Okazaki-like H-strand fragments results in an intermediate containing two site-specific H-strand discontinuities, one at each origin of H-strand synthesis (Fig. lOf). Excision of the primers and sealing of the H-strand breaks is not simultaneous but stepwise, generating molecules with a single H-strand discontinuity at the origin of H-strand synthesis in the A or B region (Fig. log, g'). The final step is the excision of the single remaining H-or L-strand primer and repair of the discontinuity, producing a newly replicated covalently closed minicircle.
This model is consistent with all observed site-specific discontinuities, and the highly discontinuous nature of Hstrand synthesis. It also provides an explanation for the existence of minicircle replication intermediates containing half-length nascent H-strands. Electronmicroscopic analysis of replicating minicircles reveals 8-type structures (8, 30, 31). In some of these structures, one arm of the replication bubble appears to be at least partially single-stranded (8, 31). This type of structure is predicted by the model during early stages of replication. A few examples of what appear to be single-stranded minicircles have been observed in electronmicrographs of the free minicircle fraction (31). Such a structure is not predicted by our model, but could arise from a replicating molecule in which H-strand synthesis has not been primed, thereby displacing a single-stranded parental L-strand circle. Further characterization of these rare circles is needed to determine whether they are, in fact, completely singlestranded or are just extensively gapped double-stranded molecules. It is possible that the spreading and staining procedures used (31) may not clearly distinguish between these two alternatives.
It is not known what prevents the discontinuities in newly replicated minicircles from being closed before all molecules have been replicated and reattached to the network. However, the presence of what appear to be modified nucleotides within the primers at these sites may be involved.