Mapping the sites of initiation of DNA replication in rat and human rRNA genes.

To study the organization of DNA replication in mammalian rRNA genes, the sites of initiation of DNA synthesis in rat and human rRNA genes were mapped by two independent techniques. In rat cells the growth of the nascent DNA chains was blocked by Trioxsalen cross-links introduced in vivo. The fraction of “restricted” nascent DNA chains labeled in vivo was isolated, and the abundance in this fraction of cloned ribosomal DNA sequences was determined by hybridization. In the experiments with human cells, the nascent DNA chains were allowed to grow unrestricted for a certain period of time and the movement of the replication forks along the rRNA genes was followed by hybridization of cloned ribosomal DNA sequences to the “unrestricted” nascent DNA fragments fractionated according to size. The results show that in both rRNA genes there are two well defined regions of initiation of DNA synthesis. The first one is located upstream of the transcription units and the second one is located at the 3′-end of the coding regions of the ribosomal DNA repeats.

A number of techniques for mapping sites of initiation of DNA synthesis in vivo have been developed during several recent years. However, their application has led to conflicting results as to whether origins of DNA replication are located at defined positions in mammalian genomes. In the cases when methods that analyze the newly synthesized DNA chains were applied to the dihydrofolate reductase (DHFR) 1 gene domain in Chinese hamster ovary cells, which is the most studied model for initiation of mammalian DNA replication, the investigators were able to determine the presence of well defined origins of bidirectional replication (1). On the other hand, the analysis of genomic DNA for the presence of replication bubbles and forks by the method of two-dimensional gel electrophoresis failed to indicate the presence of such origins (2). The rRNA genes are repeated tandemly about 400 times in mammalian genomes. This makes them a suitable model for studies of initiation of DNA replication without the need for synchronization of the cells and amplification of the nascent DNA fragments. Three recent papers have analyzed the initiation of DNA replication in the rRNA gene cluster of human cells. In the first one, the authors used two-dimensional gel electrophoresis and concluded that initiation of replication takes place throughout most of the nontranscribed spacer, but not in the transcription unit and the adjacent regulatory elements (3). In the second paper, the authors used a method called "nascent strand abundance analysis," based on a combination of sedimentation and electrophoretic fractionation of DNA. They reached the conclusion that although initiation of DNA replication was more frequent a few kilobase pairs upstream of the transcribed region, it could occur everywhere in the ribosomal DNA repeat, including the transcription unit itself (4). In the third paper, the authors studied the in vitro replication of plasmids containing cloned human ribosomal DNA sequences and showed that replication initiated specifically within two 7-kb DNA fragments located upstream of the promoter and downstream of the 3Ј-end of the coding region (5). The results of the first two papers imply that initiation of DNA synthesis may be a random process. However, there is evidence indicating that initiation of DNA replication follows a well defined pattern in vivo. This pattern is specific for each cell line and varies during embryogenesis and development. Thus, certain genes or sequences are replicated early in S-phase in some cells, while the same genes and sequences are replicated late in the S-phase of other cells (6). A second argument against the concept of random initiation of DNA replication is that a completely random initiation, regardless of the number of initiation events and the length of the S-phase, would leave a part of DNA unreplicated in each cell cycle.
Since the problem of the organization of initiation of DNA replication in the rRNA gene cluster or elsewhere in the mammalian genome has not been satisfactorily solved so far, in the present article we used two well established biochemical procedures developed by Anachkova and Hamlin (7), and by Vassilev and Johnson (8), to map replication origins in rat and human rRNA genes, respectively. We modified these procedures to avoid certain drawbacks of the original protocols and the results we obtained did not support the hypothesis of random initiation of DNA replication in vivo. We came to the conclusion that in both rat and human rRNA genes, DNA replication most probably initiates at two well defined areas located a few kilobases upstream of the promoter and at the 3Ј-end of the transcription unit, and that no initiation of DNA synthesis normally occurs outside these zones.

EXPERIMENTAL PROCEDURES
Cell Cultures-Guerin ascites tumor cells were propagated in Swiss male albino rats. At days 7-9 after inoculation (late log phase), the ascites liquid was withdrawn under sterile conditions and the cells were further cultured in minimal essential medium for suspension cultures (Sigma), buffered with 50 mM HEPES. HeLa cells were grown as suspension cultures in minimal essential medium, supplemented with 5% fetal bovine serum (Sigma). Genomic DNA was randomly labeled with 0.025 Ci/ml [ 14 C]dT (DuPont, 50 mCi/mmol) for 24 h.
Isolation of "Restricted" Nascent DNA Fragments-Cells were spun down at 800 ϫ g and resuspended in fresh medium without serum to make 5 ϫ 10 6 -10 7 cells/ml. DNA was cross-linked by four successive treatments with Trioxsalen and near UV light as described previously (9) (20 -40 Ci/mmol, DuPont) for 1 h to label the nascent DNA fragments synthesized between crosslinks. Cells were lysed in 0.5% SDS, 1 M NaCl, 10 mM EDTA, 50 mM Tris-HCl, pH 8, and the proteins were digested with 200 g/ml Proteinase K (Merck) at 37°C for 4 h. After deproteinization with phenol/ chloroform (1:1) and with chloroform, 1 volume of ethanol was overlayered and the high molecular weight chromosomal DNA was recovered by spooling on a glass rod. DNA was dissolved in 10 mM Tris-HCl, 1 mM EDTA, pH 7.4, to give 200 -500 g/ml, made 0.2 M in NaOH by adding 1 M NaOH, and centrifuged in 5-20% sucrose density gradients prepared in 0.2 M NaOH, 1 mM EDTA in Beckman SW 27 rotor at 25,000 rpm, 10°C, for 18 h. Aliquots were counted, the fractions containing the nascent DNA chains were pooled together, and DNA was immunoprecipitated as described later in the text.
Isolation of "Unrestricted" Nascent DNA Fractions-Exponentially growing HeLa cells were spun down and resuspended in fresh medium without serum to make 5 ϫ 10 6 -10 7 cells/ml. They were incubated with 50 M BrdUrd and 10 Ci/ml [ 3 H]dC, or [ 3 H]dT for 10 min. Labeling was terminated by pouring the cell suspension in 10 volumes of ice-cold 0.14 M NaCl, 0.01 M phosphate buffer, pH 7, and DNA was isolated as in the previous paragraph. To size fractionate nascent DNA chains, DNA was dissolved in 10 mM Tris-HCl, 1 mM EDTA, pH 7.4, to make approximately 500 g/ml, made 0.2 M in NaOH and applied on top of linear 5-20% sucrose density gradients prepared in 0.2 M NaOH, 1 mM EDTA. Gradients were centrifuged in Beckman SW 27 rotor at 20,000 rpm at 10°C for 18 h. Tubes were unloaded from the bottom and the respective size fractions of DNA were pooled together and recentrifuged under the same experimental conditions using Beckman SW 41 rotor.
Immunoprecipitation of the Nascent DNA Strands-The nascent DNA fractions recovered from the gradients were dissolved in 0.14 M NaCl, 0.01 M phosphate buffer, pH 7, containing 0.5% Tween 20 and 100 g/ml bovine serum albumin in a final volume of 400 l. An equal volume of monoclonal anti-BrdUrd antibody (Beckton and Dickinson) was added, and after 1 h at room temperature the antigen-antibody complex was precipitated with an excess of second antibody (anti-mouse IgG rabbit IgG fraction, Sigma). After another hour at room temperature the samples were kept at 4°C overnight and the precipitate was collected by centrifugation in an Eppendorf microcentrifuge for 10 min. It was washed with 0.14 M NaCl, 0.01 M phosphate buffer, pH 7, resuspended in 200 l of 1 mM NaCl, 0.5% SDS, 50 mM Tris-HCl, 10 mM EDTA, pH 7, and digested with 200 g/ml Proteinase K (Merck). After deproteinization with phenol/chloroform (1:1) and chloroform, DNA was precipitated with 2.5 volumes of ethanol.

RESULTS
Experimental Approach-The restricted nascent chain growth technique was originally developed in our laboratory (9), and has been successfully used to map "ori-␤" in the DHFR domain of Chinese hamster ovary cells (7) and to isolate a number of mouse replication origins (14). The rationale behind this approach is depicted schematically in Fig. 1. Cells whose DNA had been uniformly labeled with [ 14 C]dT were treated with Trioxsalen and long wave ultraviolet light to introduce cross-links in DNA in vivo and were incubated with BrdUrd and [ 3 H]dC, or [ 3 H]dT for 1 h. In the course of the incubation BrdUrd and the radioactive precursors are incorporated into three classes of newly synthesized DNA. (i) As a result of excision repair short stretches of labeled DNA are synthesized (15), which are covalently attached by both their ends to the high molecular weight DNA and during all subsequent procedures remain bound to it (16,17). (ii) Elongation of the already initiated DNA chains continues until the replication forks reach Trioxsalen cross-links, where they will stall. This produces labeled DNA stretches covalently linked to the high molecular weight DNA by their 5Ј-ends. This process will take place only during the first 2-3 min of the labeling period, since the distances the replication forks have to travel to reach the first cross-link are a few kilobases at the most. (iii) Short DNA sequences initiated at origins of replication, located between Trioxsalen cross-links, are synthesized. They are not ligated to the high molecular weight DNA and under denaturing conditions are released from it to form a fraction of newly synthesized fragments. Upon centrifugation in alkaline sucrose density gradients, cross-linked genomic DNA sediments to the bottom, while the fragments synthesized at origins of replication sediment as a well defined peak in the upper half of the gradient (16,17). The material of this peak was collected and further purified by immunoprecipitation with anti-BrdUrd antibody. The course of purification of the nascent DNA chains was monitored by measuring the 3 H/ 14 C ratio and the specific H]dT, and the short nascent DNA fragments synthesized at origins of replication located between the cross-links (zone 2) were isolated by alkaline sucrose gradient centrifugation. The low molecular weight fraction was purified by immunoprecipitation with anti-BrdUrd antibody and used for hybridization with dot-blotted in excess DNA probes. Hybridization signal was obtained only with probes located at, or close to the origin of replication (zone 2), and not with probes located far from the origin region (zones 1 and 3). radioactivity of DNA. After the immunoprecipitation, the specific radioactivity of the nascent DNA was equal to that of control genomic DNA, uniformly labeled under the same conditions. This was an indication that the purification procedure had efficiently eliminated any contaminating genomic DNA. The purified labeled in vivo nascent DNA was hybridized to dot-blotted cloned DNA sequences spanning the ribosomal DNA repeat. By monitoring the hybridization signals it was possible to determine the sequences that were selectively represented in the fraction of nascent DNA fragments, meaning that these sequences contained sites that function as DNA replication origins in vivo (Fig. 1).
The unrestricted nascent chain growth technique was developed by Vassilev and Johnson (8) and was used to map ori-␤ in the single copy DHFR domain in Chinese hamster ovary cells (18), and a number of other mammalian origins of replication (19 -23). The rationale behind the unrestricted nascent chain approach is schematically represented in Fig. 2. Exponentially growing cells, uniformly labeled with [ 14 C]dT, were labeled with BrdUrd and [ 3 H]dC, or [ 3 H]dT for 10 min. During this time any DNA chains initiated at the beginning of the labeling period will grow to approximately 20 kb; DNA chains initiated later will grow only to fractions of this length and DNA chains initiated at the end of the labeling period will be very short. Thus for each active replication origin a set of nascent DNA fragments of different length, centered at the origin site, will be synthesized. By hybridizing the different size classes of nascent DNA chains to genomic probes, it is possible to determine the area where DNA synthesis has initiated. Genomic DNA was isolated and size-fractionated by centrifugation in alkaline sucrose density gradient. The different size fractions were purified by a second round of alkaline sucrose density gradient centrifugation and immunoprecipitation with anti-BrdUrd antibody. Their specific radioactivity, determined by their 3 H/ 14 C ratio was practically identical with that of control genomic DNA, uniformly labeled under the same conditions. This was an indication that the purification procedure had efficiently eliminated contaminating DNA. The purified nascent DNA fragments were hybridized with dot-blotted cloned unique DNA probes spanning the ribosomal DNA repeat. In this experiment the probes adjacent to the origin would hybridize both with the short and long nascent DNA fragments, while probes distal to the origin would hybridize with the long DNA fragments only.
We have modified the original protocol of Vassilev and Johnson (8) in a number of ways. The method has been introduced for detection of origins of replication in single copy sequences and that is why the selected unique DNA probes were amplified by the polymerase chain reaction. Since the rRNA gene family occurs as naturally multicopy, this step was omitted in our protocol. The most important modification was that we used the size-fractionated nascent DNA fragments labeled in vivo, rather then labeled in vitro. For this reason the label was 3 H and not 32 P and the specific radioactivity of the fragments was lower than it would have been if we had labeled them in vitro. However, in this way we were sure that we had eliminated the effect of any contaminating genomic DNA and detected the newly synthesized DNA only.
Mapping Replication Origins in the Rat rRNA Genes- Fig. 3 shows the physical map of the rat ribosomal DNA repeat and the positions of the DNA sequences we have used as probes. In the first experiment the in vivo 3 H-labeled immunoprecipitated origin fraction, isolated after cross-linking of exponentially growing rat cells with Trioxsalen, was hybridized to dot-blotted fragments RrII and RrIV, spanning most of the rat ribosomal DNA repeat. Both clones gave positive signals. Fragment RrII, which contained the gene for 18 S rRNA and part of the adjacent nontranscribed spacer upstream of the promoter region, gave a stronger signal than RrIV, which contained the gene for 28 S RNA and part of the nontranscribed spacer downstream of the 3Ј-end of the gene. Furthermore, DNA probes subcloned from RrII and RrIV were dot-blotted on membranes and hybridized with the origin DNA fraction. In this case only probes Rr56, a 5.2-kb DNA sequence located upstream of the promoter sequence, and Rr151, a 3.9-kb sequence located at the 3Ј-end of the transcription unit, gave good hybridization signals (Fig. 4A). This showed that there may be two initiation zones in the rat ribosomal DNA repeat. One of them is more pronounced (which would probably mean that it is used more often) and is located upstream of the promoter region, and the second one is less well pronounced and spans the 3Ј-end of the 28 S RNA gene and part of the adjacent nontranscribed spacer region. In an attempt to map more precisely the position of the stronger initiation site upstream of the 5Ј-end of the gene, we cleaved pRr56 with the restriction endonucleases EcoRI, EcoRV, and SauI, isolated the restriction fragments designated A, B, and C, and hybridized them with the origin fraction. The strongest signal was obtained with fragment A. Fragment C gave a very weak signal and fragment B gave a signal with intermediate strength (Fig. 4B). These experiments were performed with in vivo labeled [ 3 H]dT or [ 3 H]dC nascent DNA fragments, and in both cases we obtained absolutely the same result with the only exception that in the case of the [ 3 H]dT-labeled origin fraction the signals were between two and three times weaker, due to the lower specific radioactivity of the [ 3 H]dT-labeled DNA. In all experiments a 1.2-kb mouse DNA fragment containing the DHFR gene sequence was blotted along with the ribosomal probes and simultaneously hybridized to estimate the background level hybridization. These experiments indicated that an origin of DNA replication was located about 4 kb upstream of the promoter region of the rat rRNA gene, most probably within the 838-base pair EcoRI/EcoRV fragment A.
Mapping Replication Origins in the Human rRNA Genes-Exponentially growing HeLa cells were cross-linked with Trioxsalen and labeled as above. The nascent DNA fraction was isolated and hybridized to five dot-blotted probes that span the length of the human ribosomal DNA repeat. In this case we were able to detect the presence of a replication origin in the nontranscribed spacer 5Ј of the rRNA gene, but the results were not statistically significant because of the weak hybridization signal. For an unknown reason HeLa cells incorporated labeled precursors less readily than Guerin cells and the specific radioactivity of the labeled DNA was low. Moreover, both the cloned probes and the nascent DNA fragments were relatively short, which could also lead to the low hybridization signal. To get more conclusive results about the presence and location of replication origins in the human ribosomal DNA repeat we applied the unrestricted nascent chains growth technique in which longer nascent DNA fragments are used. We consider this technique comparable to the restricted nascent chains growth approach since in two different laboratories the positions determined for the DHFR ori-␤ by the two techniques coincided (7,18). DNA from exponentially growing HeLa cells, pulse-labeled with [ 3 H]dC and BrdUrd, was size fractionated by two successive alkaline sucrose density gradient centrifugations to obtain five DNA fractions, relatively homogeneous in respect to chain length. 28 S and 18 S rRNA were run as size markers, and it was calculated that the five fractions had average lengths of 30,15,8,4, and 1.5 kb. The nascent DNA fractions were immunoprecipitated with anti-BrdUrd antibody. Five cloned DNA fragments spanning the ribosomal DNA repeat, designated by numbers 1 through 5 (Fig. 5), were dotblotted in excess and were hybridized with the five size fractions of the in vivo labeled nascent DNA to determine the relative abundance of the cloned fragments in the fractionated nascent DNA. To compare the results obtained with the different size fractions we had to normalize them. This could be done either by using a constant amount of labeled DNA, regardless of the size of the fragments, or by using increasing amounts of DNA with increasing size of the fragments. In the first case we would measure the hybridization capacity in nucleotides per dot, while in the second case it would be expressed in chains per dot. We tried both approaches and found that the first approach gave more reproducible results. Therefore a constant amount of counts of the five size fractions was used for hybridization. A 211-base pair long DNA fragment from the 5Ј-end of the human ␤-actin gene was dot-blotted as a negative control and was hybridized with all size fractions. It gave background hybridization with all of them. Probe 1, located at the initiation site for transcription of ribosomal DNA, exhibited high and uniform hybridization with all size fractions of nascent DNA. Probe 2, located at the 3Ј-end of 28 S RNA, also gave a strong and reproducible hybridization signal with all size fractions, although the absolute strength of the signal was lower than with probe 1. The signal obtained with probe 3, located in the middle of the nontranscribed spacer, was very weak when hybridized with short nascent DNA fragments and approached the strength of the signals obtained with probes 1 and 2 when hybridized with the longest nascent DNA fragments. This was an indication that this probe was not located near an origin of replication. Probe 4 gave a hybridization pattern similar to that of probe 3, which showed that this sequence was also not located near an origin of replication. Probe 5 showed a hybridization pattern similar to that of probe 1 (Table I). It is difficult to express the relationship between fragment lengths, hybridization signals, and probe locations in precise analytical form. Nevertheless, by empirically arranging the nascent DNA fragments on the physical map of the ribosomal DNA repeat, it was possible to obtain an alignment satisfying the data in Table I (Fig. 5). Thus, it could be argued that since both probes 1 and 5, which are located 7 kb apart, hybridize with size fragments 4-kb long, an initiation zone at least 4 kb in length should be located between these two probes. On the other hand, since probe 4, which is located 5 kb upstream of probe 5, does not hybridize with the 8-kb nascent fragments, the length of this initiation zone should not exceed 4 kb. This is an important finding and the fact that the same size fragments hybridized with some probes and did not hybridize with others showed that the initiation zones are well defined and that no initiation occurs outside these zones. Similar arguments lead to the conclusion that a second replication origin is contained within an approximately 3 kb long sequence centered at the 3Ј-end of the transcription unit (Fig. 5). DISCUSSION To map the sites of initiation of DNA synthesis in the rat and human ribosomal DNA repeats, we applied two biochemical approaches, proven to be adequate for localization of mammalian origins of replication. They have been tested on the yeast ARS1 (14) and the SV40 (8,24) origins of replication as model systems, and were applied to a number of other mammalian replication origins (7, 14, 19 -23). The positions determined by these approaches excellently concur with the positions determined by other methods (1). The application of these two techniques to the rat and human rRNA genes enabled us to map a well defined zone of initiation of bidirectional DNA replication a few kilobases upstream of the coding regions, and also a less well expressed zone of initiation near the 3Ј-end of the transcribed units. These results are fully consistent with the findings of Coffman et al. (5), that a 1.38-kb sequence, located immediately upstream of the promoter, and to a lesser extent a 7-kb sequence, located at the 3Ј-end of the 28 S coding region of human ribosomal DNA, serve as efficient substrates in an in vitro replication system involving proteins from human cells (Fig. 5). The location of the upstream initiation zone agrees with the major initiation sites obtained with two-dimensional gel electrophoresis (3) and nascent strand abundance analysis of human rRNA genes (4) (Fig. 5). However, the conclusions that lower frequency initiation sites are distributed throughout most of the nontranscribed spacer (3) and the coding region (4) are not consistent with our results. The reasons for this discrepancy are not clear at present. The two-dimensional gel electrophoresis is the technique of choice for mapping origins of replication in genomes with low complexity and genomes that do not take up DNA precursors readily, since the method does not involve labeling of nascent DNA. When applied to mammalian genomes, it relies on a number of assumptions that are not proven to be always valid (25). For instance, the method involves enrichment for replication bubbles by using their presumed association with the nuclear matrix and assumes that no structural changes or nicks will occur in DNA during the isolation procedure. The reason the nascent strand abundance technique leads to the conclusion that initiation can occur throughout the ribosomal DNA repeat unit could be that the nascent DNA fractions were not well purified from genomic DNA, or that DNA have been fragmented in the course of the purification procedure. In this way random genomic DNA fragments will be present in the nascent DNA fraction, or fragments from longer nascent chains will be present in the population of shorter chains, both cases leading to a certain randomization of the results.
In our experiments the careful purification of the nascent fragments from random genomic DNA was critical for obtaining meaningful results. For this reason we isolated the nascent DNA fragments by immunoprecipitation. In this way we avoided the risk of compromising the results by using nascent DNA fractions not purified from genomic DNA, because immunochemical specificity can be considered almost absolute. In addition, we have followed the hybridization signal of the in vivo labeled nascent DNA fragments, thus eliminating the effect of any incidentally present nonlabeled contaminating DNA. Finally, in our experiments the fractionation of the newly synthesized DNA was done at the very first step of the isolation procedure and thus any possible artifacts from DNA degrada-  Table I   tion were avoided. A possible source for errors in the experiments with the human rRNA genes could be the use of a single nucleoside precursor to label DNA since the G ϩ C content of probes 1 (75%) and 2 (71%) was higher than that of probes 3, 4, and 5 (51, 47, and 56%, respectively). However, this fact should not affect our results since we deduce the positions of the replication origins from the length of the nascent fragments that hybridize with the different probes, rather than from the strength of the hybridization signal itself. Nevertheless, to make sure that our conclusions have not been biased because of the different G ϩ C content of the probes, the same experiments were performed with a different precursor, [ 3 H]dT. We obtained practically the same results with the only difference that in the case of [ 3 H]dT the specific radioactivity of DNA was lower, because dT competed with BrdUrd. A support for our conclusion that there are specific replication origins in the ribosomal DNA repeat came from the primary structures of the predicted origin regions. Analyses of the primary structure of other known mammalian chromosomal origin regions have revealed the existence of certain sequence and structural elements that are found in zones of initiation of DNA replication more frequently than would be expected by chance (26 -28). These include A ϩ T-rich tracts, sequences similar to the yeast ARS and sequences similar to Drosophila SAR, transcription factor binding sites, and DNA unwinding elements. These sequence elements are found at origins of replication in simple eukaryotic genomes such as yeast and animal viruses and their role in initiation of DNA replication has been determined by biochemical and functional assays. The primary structures of the upstream initiation zones of both rat and human ribosomal DNAs have been published and we searched a 2.52-kb EcoRI/SauI rat nontranscribed spacer sequence (29) and a 4.58-kb BamHI/EcoRI human nontranscribed spacer sequence (30) for such common features (Fig. 6). The numberings of the nucleotides given hereafter are according to Financhek et al. (29) and Sylvester et al. (30), respectively. Both sequences exhibit A ϩ T-rich tracts that contain yeast ARS-like sequences (31) and Drosophila SAR sequences (32). In the rat origin region there is a cluster of three SAR sequences starting at nt position 198, one at 862, and one at 1943. In the human replication initiation zone two SAR sites occur at nt positions Ϫ4830 and Ϫ1780 and three start at nt position Ϫ4290. Two sequences similar to yeast ARS at nt positions 1624 and 2106 are found upstream of the rat rRNA genes, and the human initiation zone contains four ARS-like sequences at nt positions Ϫ3785, Ϫ3764, Ϫ1374, and Ϫ1708, respectively. The initiation zones contain potential binding sites for at least two proliferation-specific transcription factors that may function in initiation of DNA replication, Oct-1/NFIII, and p53. The rat initiation zone contains a 9/10 match to the consensus binding site of Oct-1/NFIII (36), while the human one contains two perfect matches at nt positions Ϫ3968 and Ϫ3962. A putative binding site for the protein p53 (37) is located at nt position 1041 in the rat DNA and at Ϫ4509 in the human DNA. In addition, the sequences contain extensive polypyrimidine tracts, which are identified as sequence-specific start sites for the DNA polymerase-primase complex in vitro (38).
The analysis of the organization of origins of replication in Escherichia coli, yeast, and SV40 has shown that the minimal essential cis-acting sequence required to initiate DNA replication contains a genetic component that is easily unwound (33,34). The DNA unwinding elements are determined by base stacking interactions between nearest-neighbor dinucleotides and is not simply a function of A ϩ T content, but depends on the specific DNA sequence (34). We used the computer program Thermodyn (35) to perform a sliding window analysis of the helical stability (⌬G) of the initiation zones. The analysis of the rat sequence reveals three local minima of helical stability: around nt positions 180, 1180, and 1610, respectively. The first and lowest minimum is located in the 838-base pair EcoRI/ EcoRV fragment that showed the strongest hybridization signal with the origin fraction, and is therefore the most likely origin-containing candidate. The human sequence displays two local minima of helical stability, from nt Ϫ4822 to Ϫ4678 and around nt position Ϫ4318. Both are situated within the 1.38kb BamHI/SmaI fragment which in the in vitro replication reaction was evaluated as the most likely origin-containing fragment (5).
The existence of these origin-related sequence elements in the predicted zones of initiation of DNA synthesis strongly supports the conclusion that they function as replication origins in vivo. These data are consistent with the hypothesis that in eukaryotes replication initiates within clusters of redundant modular elements associated with DNA unwinding function (39). We propose that eukaryotic cells have evolved a limited number of short sequence elements that are used as modules to build different control regions, including regions of low helical stability that can serve as origins of DNA replication. Further studies are necessary to elucidate whether all, or only a subset of these sites, are used as origins and how, in the course of development and differentiation, chromatin is organized in such a way that some of the potential initiation sites are blocked, while others are made accessible for the assembly of the replication initiation complex.