Characterization of the enzymatic properties of the yeast dna2 Helicase/endonuclease suggests a new model for Okazaki fragment processing.

The Saccharomyces cerevisiae Dna2, which contains single-stranded DNA-specific endonuclease activity, interacts genetically and physically with Fen-1, a structure-specific endonuclease implicated in Okazaki fragment maturation during lagging strand synthesis. In this report, we investigated the properties of the Dna2 helicase/endonuclease activities in search of their in vivo physiological functions in eukaryotes. We found that the Dna2 helicase activity translocates in the 5' to 3' direction and uses DNA with free ends as the preferred substrate. Furthermore, the endonucleolytic cleavage activity of Dna2 was markedly stimulated by the presence of an RNA segment at the 5'-end of single-stranded DNA and occurred within the DNA, ensuring the complete removal of the initiator RNA segment on the Okazaki fragment. In addition, we demonstrated that the removal of pre-existing initiator 5'-terminal RNA segments depended on a displacement reaction carried out during the DNA polymerase delta-catalyzed elongation of the upstream Okazaki fragments. These properties indicate that Dna2 is well suited to remove the primer RNA on the Okazaki fragment. Based op this information, we propose a new model in which Dna2 plays a direct role in Okazaki fragment maturation in conjunction with Fen-1.

Biochemical and genetic studies of DNA replication in viruses and lower eukaryotes have contributed substantially to our understanding of eukaryotic DNA synthesis (1)(2)(3)(4)(5). The various steps in eukaryotic DNA synthesis are basically similar to those in prokaryotes, requiring many common enzymatic functions. Despite these striking similarities, certain eukaryotic replication components differ from the prokaryotic counterparts. Most noteworthy is that three essential DNA polymerases (pol) 1 ␣, ␦, and ⑀ are required for DNA replication in eukaryotes (5)(6)(7). In contrast, pol III is the only replicative enzyme in prokaryotes involved in DNA synthesis. Unlike its prokaryotic counterpart, the eukaryotic primase is complexed with DNA pol ␣. Thus, the role of the pol ␣-primase complex appears to function solely in the synthesis of a short RNA-DNA primer (referred to as primer DNA in this paper). The primer DNAs are then utilized by pol ␦ for the initiation of leading strand synthesis and by pol ␦ or pol ⑀ for each Okazaki fragment synthesis for lagging strand DNA replication.
The short and discontinuous Okazaki fragments at replication forks are processed and then joined to generate a continuous DNA strand through a series of complex enzymatic reactions that require a number of enzymes that include a 5Ј-to 3Ј-exonuclease (5,8). However, none of the eukaryotic polymerases possess intrinsic 5Ј-to 3Ј-exonuclease activity for Okazaki fragment processing, unlike the well characterized prokaryotic polymerase, Escherichia coli DNA pol I (1). In the current model of replication in eukaryotes, Fen-1 provides the 5Ј-to 3Ј-exonuclease activity, and with the assistance of RNase HI removes the RNA segments on Okazaki fragments (9 -11). RNase HI hydrolyzes the initiator RNA of the primer DNA leaving a single ribonucleotide at the RNA-DNA junction, which is subsequently removed by the 5Ј-to 3Ј-exonuclease activity of Fen-1. In mammalian cells, DNA pol ␦, replication protein-A (RPA), proliferating cell nuclear antigen (PCNA), replication factor-C (RFC), RNase HI, Fen-1, and DNA ligase I are necessary and sufficient to reconstitute lagging strand synthesis in vitro (11). Recently, it was shown that Fen-1 contains a structure-specific endonuclease activity that cleaves the 5Јunannealed single-stranded (ss) DNA or RNA at the duplex junction (8,(12)(13)(14). In addition, mammalian RNase HI can cleave 5Ј of the last ribonucleotide of ssRNA-DNA hybrid molecules (15). These findings suggest that Okazaki fragment maturation could occur through a more complex process than inferred previously. For example, Okazaki fragment maturation is likely to require formation of a 5Ј-tail prior to the action of Fen-1 and/or RNase HI, a process that is poorly understood at present.
Although the role of Fen-1 and RNase HI in Okazaki fragment maturation has been well established in vitro, the phenotypes observed in Saccharomyces cerevisiae strains lacking a gene for the yeast Fen-1 (yFen-1) or RNase HI (RAD27/RTH1 or RNH35, respectively) reveal that, unlike other genes whose products are essential for DNA replication, RAD27 and RNH35 are dispensable. RNase HI is a two-subunit enzyme (16), and the deletion of its catalytic subunit did not affect cell viability (17). Yeast strains carrying a deletion of RAD27 are viable at 30°C and compromised for growth at 37°C, producing a yeast strain with a terminally arrested phenotype, an indication of a defect in DNA replication (18,19). At 30°C, this strain shows elevated rates of spontaneous mutation, mitotic recombination, and chromosome loss (19 -22), consistent with its critical roles in chromosome maintenance. In support of these findings, Fen-1 has been shown to participate in long patch base excision repair in human cells (23)(24)(25)(26)(27), maintenance of direct repeats (22) or CTG trinucleotide repeats (28), nonhomologous DNA end joining (29), and mitotic gene conversion (30) in yeasts. Therefore, the viability of the RAD27-deleted yeast mutant strain is puzzling if yFen-1 is the only enzyme that plays such a crucial role in the completion of DNA replication. This is especially true considering that deletion from the yeast genome of the DNA ligase I gene, which is required for the joining of Okazaki fragments processed by Fen-1/RNase HI, renders cells inviable (31). Thus, the viability of yeast cells lacking Fen-1 argues strongly for the existence of a more critical pathway for the in vivo processing of Okazaki fragments.
Although Dna2 has been implicated in Okazaki fragment maturation on the basis of its genetic and physical interaction with yFen-1 (32), its precise role in DNA replication remains unclear. Recently, we reported that Dna2 of S. cerevisiae contains a potent ssDNA-specific endonuclease activity (33). In an attempt to define the role of Dna2 in yeast DNA replication, we have carried out genetic and biochemical studies to examine whether Dna2 participated in the maturation of Okazaki fragments. In this paper, we present evidence that Dna2 is well suited to remove completely the primer RNA on Okazaki fragments. Based on the biochemical studies presented here and genetic studies published elsewhere (34), we propose a novel model in which Dna2 plays a key role in the processing of Okazaki fragments.

EXPERIMENTAL PROCEDURES
Oligonucleotides, DNA, and Nucleoside Triphosphates-All oligonucleotides used for the construction of various DNA and RNA substrates were synthesized commercially (Life Technologies, Inc., and Gemini Biotech) and gel-purified prior to use. The sequences of the oligonucleotide used in this study are listed in Table I. Oligonucleotides used to prepare substrates, the position of radioisotopic labels in the substrates, and the substrate structures are as described in each figure. The oligonucleotides 2 and 12 (Table I)  Proteins and Enzymes-The following proteins were obtained commercially: restriction endonuclease HpaII, T4 polynucleotide kinase, and terminal transferase were from New England Biolabs. Streptavidin, the Klenow fragment of E. coli DNA polymerase I, and T4 DNA polymerase were from Roche Molecular Biochemicals. Schizosaccharomyces pombe DNA pol ␦ (35) was a generous gift from Dr. J. Hurwitz (Sloan-Kettering Institute). The recombinant Dna2 protein (HX-Dna2), with six histidine residues and the Xpress epitope at its amino terminus, was prepared as described previously (33). The PCNA gene of S. cerevisiae was cloned as described (36), and recombinant PCNA was overexpressed and purified according to the procedure described previously (37). RPA was purified from a protease-deficient yeast strain BJ2168 (MATa, ura3-52, trp1-⌬63, leu2-⌬, prb1-1122, pep4-3, prc1-407, gal2) as described (38). For RFC purification we combined two procedures that were described previously (37,39) to obtain a preparation devoid of detectable nuclease activity. Briefly, crude extracts were prepared from a 60-liter culture of BJ2168, and RFC was purified using several column chromatographic steps that included an adsorption to and an elution from Affi-Gel Blue (39), hydroxylapatite (37), and ssDNA-cellulose (37). The latter fraction was dialyzed, loaded onto a PCNA-agarose column (1.5 ϫ 6 cm, 10 ml), and eluted as described (39). RFC-containing fractions were pooled and concentrated using a Mono S column (Amersham Pharmacia Biotech). The RFC fractions (ϳ100 g) obtained from the Mono S column were aliquoted and stored at Ϫ80°C. Nuclease activities were barely detectable in this RFC preparation, and silver-stained gels revealed that the preparation was more than 95% pure (data not shown).
Preparation of DNA Helicase and Nuclease Substrates-The DNA substrates used to determine the directionality of Dna2 helicase movement were constructed as described previously (40) with following modifications. The 3Ј-end of oligonucleotide 1 (52-nt oligomer) annealed to X174 sscDNA (4 pmol) was first labeled by incorporating [␣-32 P]dCTP and then chased with excess unlabeled dCTP in the presence of Klenow fragment. The partial duplex region was then cleaved by HpaII to generate linear X174 ssDNA. This procedure yielded a substrate labeled only at the 3Ј-end of the 29-nt fragment. To prepare a substrate labeled only at the 5Ј-end of the 23-nt fragment, oligonucleotide 1 (10 pmol) was labeled at the 5Ј-end in the presence of [␥-32 P]ATP by T4 polynucleotide kinase, annealed to 4 pmol of X174 sscDNA, and then digested with HpaII. The various partial duplex substrates used to characterize the endonuclease activity of HX-Dna2 were prepared as described previously (33) using the synthetic oligonucleotides listed in Table I. Among the DNA substrates constructed, we examined whether the Y-and flap DNA substrates (5Ј-end labeled in their ssDNA tails) were cleaved by Fen-1 in order to confirm the presence of the desired specific structure. The 5Ј-tail of the Y substrate was resistant to Fen-1 as expected, whereas the flap substrate was cleaved by Fen-1 (data not shown). This result as well as the mobility change upon heat treatment during gel electrophoresis revealed that the substrates possessed a specific structure as intended.

Dna2
Translocates in the 5Ј to 3Ј Direction-Early efforts to establish the polarity of translocation of the Dna2 helicase activity were hampered by its intrinsic endonuclease and weak helicase activities. We determined the DNA unwinding polarity of Dna2 by blocking the nuclease activity through using high ratios of ATP to Mg 2ϩ (33) and by using a partial duplex DNA substrate that consisted of linear X174 ssDNA containing a 29-nucleotide (nt) fragment at the 3Ј-end and a 23-nt fragment

Enzymatic Role of Dna2 in Okazaki Fragment Maturation
at the 5Ј-end (Fig. 1). Due to its limited unwinding activity, the HX-Dna2 protein failed to produce ssDNA from the 52-base pair duplex DNA substrate (data not shown). In contrast, the enzyme displaced a shorter duplex region in the linearized substrate in the presence of ATP ( Fig. 1, lanes 4 -6). HX-Dna2 displaced the 29-nt fragment at the 3Ј-end but not the 23-nt fragment at the 5Ј-end (Fig. 1, compare lanes 5 and 6 with lanes 11 and 12). This result indicates that Dna2 translocates in the 5Ј to 3Ј direction in contrast to the previous report that Dna2 translocated in the opposite direction (41). In the previous report, the directionality of Dna2 was not experimentally determined. The directionality of Dna2 was inferred based upon the observation that it required a 5Ј-tailed forked substrate for unwinding activity (41). By using a purified recombinant nuclease-deficient Dna2 (42), however, we found that Dna2 was not dependent on the presence of 5Ј-tailed fork structure for its unwinding activity (data not shown). Dna2 Acts Preferentially on Free Ends of ssDNA-Although the addition of a molar excess of ATP (2 mM) over Mg 2ϩ (1 mM) reduced the ssDNA-specific endonuclease activity, it stimulated the release of the 3Ј-end label (Fig. 1, compare lanes 2 and 3 with lanes 5 and 6). This observation suggests that Dna2 endonuclease activity is likely to prefer ssDNA with free ends as substrate. For this reason, we examined the substrate specificity of the endonuclease activity of Dna2 by using three different DNA substrates that possessed either a 5Ј-or 3Ј-ssDNA overhang or ssDNA flanked by partial duplexed regions at both ends ( Fig. 2A). The cleavage products were first analyzed using a low resolution gel for the convenience of precise quantification.
The enzyme degraded the 5Ј-ssDNA tails most efficiently. The amount of the 5Ј-tail (69 nt) substrate cleaved with 0.5 ng of HX-Dna2 was 10.4 fmol ( Fig. 2A, lane 4), whereas the amount of 3Ј-tail (69 nt) substrate cleaved with the same amount of enzyme was slightly reduced (7.2 fmol) ( Fig. 2A, lane  10). We estimated that the efficiency of cleavage of the 5Ј-tail substrate by HX-Dna2 was about 2-fold greater than that of the 3Ј-tail (Fig. 2, lanes 5 and 12). In contrast, cleavage of internal ssDNA regions flanked by partial duplexes at both ends was extremely poor ( Fig. 2A, lanes 16 -18), requiring much higher enzyme levels (Ͼ200-fold) to accumulate comparable amounts of products. The addition of 100 ng of HX-Dna2 to the reaction resulted in the cleavage of 9.3 fmol of the substrate with partial duplexes at both ends ( Fig. 2A, lane 18). This is not because the internal 48-nt ssDNA between the two partial duplexes is too short for Dna2 to utilize. The labeled, but unannealed, 21-nt ssDNA present as background in the substrate-alone control ( Fig. 2A, lanes 13 and 15) was degraded efficiently by HX-Dna2 to yield faster migrating products, indicated by a closed arrowhead ( Fig. 2A, lanes 16 -18; see also Figs. 3 and 4). We repeated this in an identical way but analyzed the reaction products formed using a high resolution sequencing gel. As shown in Fig.  2B, the cleavage products that migrated as a uniform band in the low resolution gel ( Fig. 2A) consisted of oligonucleotides of varying sizes that were separated on the high resolution sequencing gel. In addition, we noted that Dna2 cleaved the 5Ј-overhang ssDNA more efficiently than the 3Ј-overhang DNA judging from the amount of substrate uncleaved (Fig. 2B), in keeping with the observations described in Fig. 2A. These results demonstrate that Dna2 prefers to cleave ssDNA with free ends to the internal ssDNA regions flanked by duplex regions.
This finding suggests that Dna2 binds to or uses free ssDNA ends as an entry site for its subsequent endonucleolytic action. This prompted us to examine whether modification of the ends of ssDNA affected the DNA-dependent ATPase activity of Dna2. For this purpose, we used 5Ј-or 3Ј-biotinylated oligonucleotides coupled to streptavidin in order to block access of Dna2 to the ends of ssDNA substrate (Fig. 2C). In control experiments, the addition of free streptavidin to reaction mixtures containing nonbiotinylated oligonucleotides did not affect ATP hydrolysis by Dna2 (Fig. 2C, open squares). However, in the presence of the 5Ј-biotinylated oligonucleotide, ATPase activity was inhibited partially by the addition of streptavidin (Fig. 2C, open circles). Due to the relatively slight difference in activity, this experiment was repeated four times, and the average of the results obtained are presented with error bars that represent the standard deviation about the mean. Inhibition was greatest (2-2.5-fold lower than control values) at Ն50 ng of streptavidin, which saturated Ն1 pmol of the oligonucleotide substrate used. In contrast, the binding of streptavidin at the 3Ј-end of the oligonucleotide stimulated the ATPase activity of HX-Dna2 nearly 2-fold over control values (Fig. 2C, closed circles). ATP hydrolysis by HX-Dna2 was inhibited when we created a partial duplex at the 5Ј-end of the ssDNA template but stimulated by a partial duplex at the 3Ј-end (data not shown). If this is not due to the fact that the partial duplex present at the 3Ј-end of the ssDNA effector is not really a large bulky complex, one alternative explanation for the increased ATP hydrolysis is the inability of Dna2 to interact with a DNA duplex. Since Dna2 does not utilize double-stranded DNA as an effector for ATP hydrolysis (data not shown), it is likely that the presence of a partial duplex at the 3Ј-end causes Dna2 to interact more efficiently with the available 5Ј-ssDNA end, resulting in the increased ATP hydrolysis. We conclude that the binding of streptavidin to the 5Ј-end or the presence of a partial duplex at the 5Ј-end blocks the entry of Dna2 onto the DNA substrate, thus decreasing ATP hydrolysis. On the other hand, formation of a large complex with streptavidin or a partial duplex structure at the 3Ј-end of the template could increase the effective concentration of available 5Ј-ends of ssDNA, thereby stimulating the ATPase activity of Dna2. This notion is consistent with the observations that the enzyme prefers free 5Ј-ends of ssDNA as substrate ( Fig. 2A) and that Dna2 translocates in the 5Ј to 3Ј direction (Fig. 1).
ATP Alters Cleavage of the 5Ј-ssDNA Tail by Dna2 but Not the 3Ј-End ssDNA Tail-In order to investigate further whether Dna2 cleaves 5Ј-ssDNA tails more efficiently than 3Ј-ssDNA tails in the presence of ATP, we prepared a Y-structured partial duplex substrate that contained 5Ј-and 3Ј-oligo(dT) tails. Only one of the two tails present in the Y-structured substrate was labeled as shown in Fig. 3. In order to follow only the fate of labeled strands, we used a limited amount (3 fmol) of HX-Dna2 in the presence of excess substrate (15 fmol) and preincubated the enzyme with substrates in the absence of Mg 2ϩ to allow only the binding events. Under this condition, it was expected that most of the enzyme would be bound to the substrate and, therefore, the addition of Mg 2ϩ would cause initial cleavage of only the Dna2-bound strand. We first examined the products generated by the action of HX-Dna2 with the 5Ј-end-labeled substrate, either in the absence or presence of ATP. The DNA cleavage reaction was completely dependent on the addition of Mg 2ϩ (Fig. 3A, lane 1). In the absence of ATP, the 5Ј-tail of the substrate was cleaved, yielding predominantly oligonucleotides that were 7-12 nt in length (Fig. 3A, lanes 2-4). Higher levels of Mg 2ϩ affected the cleavage reaction qualitatively, resulting in a slight decrease in the average size of the products (Fig. 3A, lanes 2-4). The addition of 2 mM ATP resulted in a significant increase in the size of the cleavage products, which varied from 16 to 22 nt in length (Fig.  3A, lanes 5-7). At a higher ratio of ATP to Mg 2ϩ (Fig. 3A, lane  Table I. Reaction mixtures (20 l) were preincubated with 15 fmol of each substrate at 37°C for 5 min and initiated by the addition of the indicated amounts of Dna2, followed by incubation at 37°C for 2 min. The reaction products were analyzed as in Fig. 1. S and B denote the substrate alone and boiled substrate controls, respectively. The arrows indicate the migration positions of the labeled oligonucleotides. Closed arrowheads indicate positions of cleavage products containing either 5Ј-or 3Ј-end labels; the open arrowhead denotes the migration position of duplex DNA resulting from cleavage of internal ssDNA region in the substrate. The amounts of substrate cleaved are indicated at the bottom of the figure. B, the experiment described in A was repeated, but the cleavage products formed were analyzed using a high resolution sequencing gel. The reaction products were boiled for 1 min and subjected to electrophoresis for 1.5 h at 35 watts through a 20% denaturing polyacrylamide gel (7 M urea) as described previously (33). M denotes molecular size markers prepared by labeling a synthetic oligo(dT) mixture (2-4, 6, 8, 10, and 12 nt) and commercial size markers ((dGATC) n , where n denotes oligonucleotides 8 -32 nt in length, Amersham Pharmacia Biotech) at their 5Ј-ends with T4 polynucleotide kinase. The numbers shown on the left of the figure indicate the size of the markers. C, ssDNA-dependent ATPase activity was measured in the presence of unmodified oligonucleotide (open squares) or oligonucleotide modified with biotin (indicated by the capital letter B) at either the 5Ј-(open circles) or 3Ј-end (closed circles). The oligonucleotides used were identical in size and sequence except for their modified end. Reaction mixtures (20-l) containing 0.3 mM MgCl 2 , 150 M ATP, 20 nM [␥-32 P]ATP, and 1 pmol of each oligonucleotide were preincubated on ice for 15 min with increasing amounts of streptavidin as indicated, and then Dna2 (10 ng) was added to each reaction. The mixtures were incubated at 37°C for 10 min, and the amount of ATP hydrolyzed was measured as described previously (33). The experiment was carried out four times, and the average of the four experiments is shown with error bars as indicated. 5), the endonuclease activity was inhibited significantly as previously observed (see Ref. 33 and Fig. 1). Under this condition, the size of the products was also altered significantly. At equimolar concentrations of Mg 2ϩ and ATP, or at a higher ratio of Mg 2ϩ /ATP (2 or 5 mM Mg 2ϩ and 2 mM ATP), products larger (Ͼ25-nt) than the length of the 5Ј-tail were observed (Fig. 3A,  lanes 6 and 7), which must have resulted from cleavage events within the duplex region. The data above may not imply the 5Ј to 3Ј directionality of Dna2, since the same finding could have resulted if Dna2 moved in the 3Ј to 5Ј direction along the unlabeled strand and cleaved the strand not bound by the protein. We believe that this is not the case, however, since we obtained a result identical to this when we used a substrate that lacked the 3Ј tail (data not shown). Therefore, the fact that the cleavage at distal sites required ATP is consistent with the observation that Dna2 translocates in the 5Ј to 3Ј direction, partially melting the DNA duplex because of its helicase activity.
In contrast, the cleavage products derived from the 3Ј-endlabeled substrate tail differed significantly from those formed from the 5Ј-end-labeled DNA. Products formed from the 3Ј-endlabeled substrate were significantly shorter, ranging from 2 to 5 nt in length (Fig. 3A, lanes 9 -11). Moreover, the addition of ATP did not alter the pattern of cleavage of the 3Ј-end labeled substrate (Fig. 3A, compare lanes 9 -11 with 12-14), whereas Mg 2ϩ affected the cleavage reaction in a manner similar to that observed with the 5Ј-end-labeled substrate. At equimolar concentrations of ATP and Mg 2ϩ , cleavage of the 3Ј-tail ssDNA decreased about 2-fold. These results demonstrate that in the presence of ATP, Dna2 acts preferentially on the 5Ј-tail ssDNA, reflecting its 5Ј to 3Ј helicase activity.
An identical series of experiments was carried out in which the mutant Dna2 protein, Dna2K1080E, was used rather than the wild type enzyme. This mutant enzyme is devoid of ATPase activity but contains wild type levels of endonuclease activity (33). Unlike wild type Dna2, the addition of ATP to reactions with the mutant enzyme did not alter the cleavage pattern observed with the 5Ј-tail ssDNA (Fig. 3B). These observations again demonstrate that the increased size of the cleavage products was due to the ATP-dependent 5Ј to 3Ј translocation of the enzyme along the template DNA.
The Presence of an RNA Segment in an RNA-DNA Chimeric Oligonucleotide Stimulates the Endonuclease Activity of Dna2-The finding that Dna2 acts efficiently to cleave an unannealed 5Ј-tail in the presence of ATP prompted us to examine whether Dna2 can hydrolyze an RNA segment on primer DNA that exists as a flap structure. We first determined whether Dna2 cleaved RNA and whether RNA stimulated ATP hydrolysis by the enzyme. Neither native, boiled tRNA, nor a poly(U) (a homopolymeric synthetic RNA) supported detectable ATP hydrolysis by Dna2. In addition, a synthetic RNA oligonucleotide labeled at its 5Ј-end was not degraded by Dna2 (data not shown), demonstrating that Dna2 is devoid of both RNase and RNA-dependent ATPase activities. In an electrophoretic gel shift assay, however, Dna2 bound to ssRNA with an affinity homopolymer at either the 5Ј-or 3Ј-end. The 5Ј-end-labeled (lanes 1-7) or the 3Ј-end-labeled (lanes 8 -14) substrate (15 fmol each) was preincubated with 0.5 ng of Dna2 at 37°C for 5 min in the presence (ϩ) or absence (Ϫ) of 2 mM ATP. Reactions were initiated by the addition of the indicated amount of MgCl 2 and incubated at 37°C for an additional 2 min; products were then analyzed as described in Fig. 2B. M denotes molecular size markers. The numbers shown on the left and right of the figure refer to the size of markers. B, analysis of the cleavage products generated by the mutant Dna2K1080E protein (devoid of both ATPase and helicase activities) in the presence and absence of ATP. The reactions were carried out, and the products were analyzed as described in A.  Table I) to construct the substrates contain a 25-nt oligo(dT) ((dT) 25 )

Enzymatic Role of Dna2 in Okazaki Fragment Maturation
greater than or at least equivalent to its binding to ssDNA (data not shown).
From these preliminary results, one possible prediction was that the 5Ј-end of an ssRNA segment present in an RNA-DNA chimeric flap structure would inhibit the endonucleolytic activity of Dna2. If this were the case, the 5Ј-terminal ssRNA would probably render such a flap structure resistant to endonucleolytic degradation by Dna2. In order to test this prediction, we prepared a substrate containing a 12-nt oligo(U) segment at its 5Ј-end in the RNA-DNA chimeric flap (Fig. 4A). Contrary to the prediction, this substrate was efficiently cleaved within the DNA sequence. Cleavage occurred exclusively within the DNA sequence of the RNA-DNA chimeric flap. Although the products formed with the RNA tail are mostly longer than 19 nt (Fig. 4A, lanes 2 and 3), this does not indicate that the nuclease is likely to cleave off a segment of that length from an Okazaki fragment, since Dna2 is able to cleave starting 2 nt from the RNA-DNA junction (Fig. 4A, lanes 2 and 3). When we used a substrate containing a flap consisting of a random sequence instead of homopolymeric (U-dT), the endonucleolytic cleavage occurred preferentially 2-3 nt from the RNA-DNA junction (data not shown), indicating that the cleavage site can be varied in the sequence context of DNA present in the flap. These findings suggest that Dna2 can completely remove the primer RNA segment in Okazaki fragments, resulting in the complete removal of ribonucleotides in the newly replicated DNA. Noteworthy was the observation that the chimeric flap tail region was cleaved more efficiently by HX-Dna2 than was the flap tail that solely contained DNA (Fig. 4A, compare lanes  2 and 3, and 5 and 6). The simplest explanation for this observation is that the RNA segment provided a more efficient initial binding or entry site for Dna2. The chimeric flap substrate was also efficiently cleaved within the DNA sequence in the presence of 2 mM ATP (data not shown) and the cleavage extended into the duplex region in the presence of 2 mM ATP. This indicates that the presence of RNA at the terminus of the flap did not inhibit the unwinding of the strand by Dna2 (data not shown).
In order to confirm further the stimulatory effects of a terminal RNA on the Dna2 cleavage reaction, we used two levels (15 and 30 fmol) of the two substrates in the presence of the fixed amount of Dna2 (0.25 ng and 1.5 fmol), and we determined the rate of substrate hydrolysis. In the presence of 0.5 mM Mg 2ϩ , the chimeric flap tail substrate was cleaved 5-7 times more efficiently than the ssDNA flap substrate (Fig. 4B), regardless of the amount of substrates used. As shown, 1.5 fmol of enzyme cleaved more than 9 fmol of substrate in 1 min in the presence of 30 fmol of substrate, indicating that Dna2 cycles efficiently especially with an RNA-initiated substrate. At lower Mg 2ϩ (Ͻ0.1 mM), the disparity between the rates of cleavage of the two substrates was even greater (Ͼ10-fold) (data not shown). These results suggest that a displaced RNA segment in an Okazaki fragment may provide a highly efficient entry site for Dna2.
Dna2 Cleaves the 5Ј-End of Duplex DNA in Conjunction with a DNA Polymerase Capable of Displacement DNA Synthesis-We have shown that Dna2, like Fen-1, can remove an RNA segment from a flap structure consisting of an RNA-DNA chimera. If this structure were a preferred substrate in vivo for the two enzymes, how could an unannealed flap structure be generated? Several mechanisms have been proposed in which the 5Ј-end regions of pre-existing Okazaki fragments are displaced for processing by Fen-1 (5,8). One postulated mechanism suggested that Dna2 generates the flap structure from the 5Ј-end region of a pre-existing Okazaki fragment through its ability to unwind in the 3Ј to 5Ј direction along the template, preceding the action of pol ␦, which extends the newly synthesized upstream Okazaki fragment. The 5Ј to 3Ј directionality of HX-Dna2 demonstrated in this study, however, questions this mechanism. Another possible mechanism postulated is strand displacement synthesis by a polymerase that extends the newly synthesized pre-Okazaki fragment.
We decided to examine whether Dna2, in conjunction with a polymerase capable of displacement DNA synthesis, could remove the 5Ј-end label of duplex DNA that is normally resistant  Table I. Note that the ss 5Ј-tail (30-nt in length) consisted of a homopolymeric 12-nt oligonucleotide (U) segment ((U) 12 , wavy line), a 13-nt oligo(dT) segment ((dT) 13 , thin line), and an additional 5-nt oligonucleotide ((CG-GAC), thick line). Reaction conditions with these substrates were as described in Fig. 3  to endonucleolytic cleavage by Dna2. For this purpose, we annealed two oligonucleotides to X174 sscDNA to generate a substrate that contained a gap of 226-nt between the two annealed oligonucleotides. In this substrate, the downstream oligonucleotide, in the form of an RNA-DNA chimera, was labeled at its 5Ј-end as shown in Fig. 5. We used two different polymerases, the Klenow fragment of DNA pol I and T4 DNA pol. The former is known to perform displacement DNA synthesis, whereas the latter lacks this property. Both polymerases are devoid of 5Ј to 3Ј exonuclease activity. In the absence of a polymerase, the labeled primer was not extended as expected, and the downstream 5Ј-end-labeled oligonucleotide was not cleaved by Dna2 (Fig. 5, lanes 1-4). When the reaction was supplemented with 0.2 unit of the Klenow fragment, cleavage of the 5Ј-end-labeled oligonucleotide occurred 3-4 nucleotides from the RNA-DNA junction, generating major products 15-16 nt in length (Fig. 5, lanes 5-8, closed arrowheads). The extent of 5Ј-end duplex DNA degradation was directly proportional to the amount of Dna2 added to the reaction, reaching Ͼ95% removal of the label (Fig. 5, lane 8) in the presence of 10 ng of Dna2. Degradation products shorter (Ͻ12 nt) than the length of the RNA segment were also observed but were not Dna2-dependent, as the Klenow fragment alone produced this signal (Fig. 6, lane 5, open arrowheads). In the presence of T4 DNA pol, cleavage was hardly observed (Fig. 6, lanes 9 -12), despite the fact that more efficient DNA synthesis was observed with this polymerase than with the Klenow enzyme. This demonstrates that cleavage at the 5Ј-end requires a displacement reaction and that Dna2 can efficiently cleave displaced 5Ј-ends as soon as they are generated.
In order to determine whether our observations with the prokaryotic polymerases also applied to a eukaryotic system, we examined whether Dna2 could cleave an RNA segment conjointly with the extension of the upstream Okazaki fragment by pol ␦ (Fig. 6). In this experiment, we used recombinant pol ␦ of S. pombe (a gift from Dr. J. Hurwitz) in conjunction with RFC, PCNA, and RPA purified from S. cerevisiae. DNA synthesis by S. pombe pol ␦ was dependent on S. cerevisiae RFC and PCNA with either singly primed (Fig. 6, lanes 4 and 5) or multiply primed DNA templates (data not shown), similar to previous findings that calf thymus pol ␦ was interchangeable with human pol ␦ (43). As shown, HX-Dna2 removed the 5Јlabeled RNA segment by cleaving 3-4 nucleotides beyond the RNA-DNA junction (Fig. 6, lanes 8 -10, closed arrowheads). This was hardly observed when the elongation of the primer DNA was prevented by omitting any of the pol ␦ processivity factors (Fig. 6, lanes 4 and 5). With HX-Dna2 alone as a control (Fig. 6, lane 2) or when PCNA, RFC, or pol ␦ was omitted from the reaction mixture (Fig. 6, lanes 4 -6), the size of the labeled downstream oligonucleotide was shortened by HX-Dna2. This is due to the partial melting and subsequent cleavage of the unwound 3Ј-end region as we showed previously (33). When the mutant Dna2 (Dna2K1080E) protein devoid of helicase activity was used in place of the wild type enzyme, cleavage of the RNA-initiated primer was still observed only in conjunction with displacement DNA synthesis (Fig. 6, lanes 11-14). This result suggests that yeast cells may rely on displacement DNA synthesis to generate unannealed flap structure during Okazaki fragment maturation. This may account for the recent observation that the helicase activity of Dna2 is not required constantly for cell viability (44), whereas the nuclease activity is essential for cell viability regardless of growth conditions (42,45). These findings emphasize that the endonuclease activity of Dna2 plays indeed a more critical biological role than the DNA helicase activity.
RPA Protects Template DNA from Cleavage by Dna2-Although we have shown that a short internal ssDNA flanked by duplex DNA region is poorly cleaved by Dna2 ( Fig. 2A), this weak activity could potentially result in chromosomal breakage during DNA replication. We found that the addition of RPA strongly inhibited the degradation of the bulk of the ssDNA present in X174 DNA by Dna2 (Fig. 7A, compare lanes 3 and  4 and 5 and 6). The slight increase in the amount of 3Ј-end label released in the presence of RPA (Fig. 7A, lanes 5 and 6) is most likely due to the partial melting of the 3Ј-duplex end caused by the destabilizing activity of RPA. The cleavage of the internal ssDNA region franked by partial duplex stretches at both ends was also inhibited by RPA (Fig. 7B, lanes 8 -10). In contrast to this, cleavage of 5Ј-ssDNA ends was hardly affected by RPA (Fig. 7B, lanes 3-5). This effect may be due to the fact that a short ssDNA with free ends is not sufficient for the stable and cooperative binding of RPA required for inhibition of Dna2 endonuclease activity. Thus, RPA binding to ssDNA may protect the template ssDNA in either lagging or leading strand from the endonucleolytic action of Dna2, which is critical to preserve the integrity of replication forks. DISCUSSION In this study, we have examined the properties of the ssDNA-specific endonuclease activity of Dna2 in order to evaluate its role in Okazaki fragment maturation. By demonstrating that Dna2 translocates in the 5Ј to 3Ј direction, we have ruled out the possibility that Dna2 acts to create an unannealed flap structure that facilitates Fen-1-catalyzed removal of the 5Ј-terminal RNA-DNA primer segment in the Okazaki fragments. The properties of the endonuclease activity of Dna2 described here suggest that Dna2 plays a more direct role in Okazaki fragment maturation. This is supported by the following observations. (i) Dna2 cleaved unannealed ssDNA tails more efficiently (Ͼ200-fold) than ssDNA regions flanked by duplex DNA at both ends. (ii) Dna2 acted preferentially on unannealed 5Ј-tail ssDNA in the presence of ATP. (iii) Dna2 completely removed the ssDNA tail from the 5Ј-end but not from the 3Ј-end of a duplex DNA structure (33). (iv) The Dna2catalyzed cleavage reaction was significantly stimulated by the presence of an RNA segment at the displaced 5Ј-end of the ssDNA tail; RNA alone neither stimulated the ATPase activity of Dna2 nor was cleaved by the enzyme. (v) Finally, when combined with a DNA polymerase capable of displacement synthesis, Dna2 catalytically and efficiently removed the 5Јend-labeled region. These data imply that Dna2 alone can remove an RNA-DNA primer from a duplex DNA structure without the participation of other nucleases such as Fen-1 or RNase HI.
Genetic Evidence of DNA2 Involvement in Okazaki Fragment Maturation-In an S. cerevisiae DNA2 temperature-sensitive mutant strain (dna2-1), low molecular weight DNA accumulates at the nonpermissive temperature (41). This observation was interpreted as a potential defect in the elongation stage of DNA replication due to the lack of fork propagation. An equally likely interpretation is that the accumulation of low molecular DNA is due to a defect in the joining of Okazaki fragments. Our data support the latter possibility. Consistent with this possibility, other DNA2 mutant alleles (for example, dna2-2) under nonpermissive conditions do not affect bulk DNA synthesis but arrest cells at late S-phase of cell cycle (46). When the temperature-sensitive allele of S. pombe dna2 ϩ was combined with a mutant allele (hus1-14; see Ref. 47) of a DNA replication checkpoint gene, the double mutant cells underwent aberrant mitosis at the restrictive temperature, resulting in a catastrophic consequence (34). This result suggests that the replicated DNA cannot be converted to the complete duplex DNA in the absence of a functional Dna2. In addition, the temperature lethality of the dna2 mutant cells was specifically rescued by the presence of multiple copies of genes such as cdc17 ϩ (S. pombe DNA ligase I), cdc1 ϩ , or cdc27 ϩ (subunits of S. pombe pol ␦), and rad2 ϩ (S. pombe Fen-1) that are involved directly in Okazaki fragment synthesis/maturation (34). These genetic observations, together with enzymatic properties of Dna2 described above, support the hypothesis that Dna2 plays a direct and essential role in Okazaki fragment maturation.
A Two-enzyme Processing Model for Okazaki Fragment Maturation-Although Dna2 and Fen1 share many common enzymatic properties and appear to act in the same pathway but in parallel, their properties are not completely overlapping in several regards. (i) A clear difference in the endonucleolytic properties is that the endonuclease activity of Dna2 is markedly stimulated by terminal RNA and cleaves within DNA, a property unique to Dna2 that can be utilized for the removal of RNA primers in Okazaki fragments. (ii) Dna2 acts inefficiently in catalyzing the removal of a short flap DNA. This suggests that although Dna2 can efficiently remove a growing branch structure containing the RNA primer, it is likely to leave a short DNA flap region behind. (iii) In contrast, Fen-1 can catalyze the removal of a short flap DNA, ultimately converting a branched structure to a nicked duplex (26).
Based on these distinct properties and the results presented here, we propose a novel model for Okazaki fragment processing (Fig. 8) involving the sequential action of these two processing enzymes for the removal of primer RNA and DNA. In this model, (i) pol ␦ first extends the newly synthesized Okazaki fragment, and (ii) displaces the RNA segment of the preexisting downstream Okazaki fragment, creating the intermediate flap structure through displacement synthesis until the RNA segment is completely displaced. Dna2 is then targeted to the displaced RNA-DNA junction prior to the removal of the primer RNA. (iii) Dna2 releases the RNA segment of the downstream Okazaki fragment by cleaving several nucleotides away from the RNA-DNA junction. It would cleave ssDNA as well if available. The fact that Dna2 is unable to digest the RNA segment, but cleaves within ssDNA, ensures the complete removal of the RNA segment. In this step, the flap junction generated while DNA is being displaced by pol ␦ may not be sensitive to cleavage by Fen-1 for the following reasons. (a) Steric hindrance caused by the replication proteins engaged at the junction might prevent Fen-1 from recognizing the ssDNA-dsDNA junction. (b) The presence of an ssDNA gap of a few nucleotides between the 3Ј-end extended by pol ␦ and the flap junction generated ahead may not be suitable for cleavage by Fen-1. These suggestions are hypothetical at present and await future biochemical analysis. If this is the case, replication proteins that prevent the access of Fen-1 have to be disassem-bled prior to the action of Fen-1. The disassembly could also allow the annealing of displaced strand to the ssDNA template, if present, restoring a flap structure consisting only of DNA. (iv) Subsequently, the flap junction becomes susceptible to Fen-1, and then Fen-1 cleaves the remaining flap DNA. The second cleavage reaction in this step, however, could be accomplished by Dna2 or another enzyme possessing 5Ј to 3Ј exonuclease activity. (v) Finally, DNA ligase I is recruited to complete Okazaki fragment maturation by sealing the resulting nick, generating a continuous DNA strand to complete lagging strand maturation. Our model suggests that the endonuclease activity of Dna2 is a critical function of the enzyme. In support of this, it was recently shown that point mutations that either abolished or reduced the endonuclease activity of Dna2 rendered the mutant cells inviable (42,45).
Although our model appears to require redundant enzymatic functions, it is supported by many of the genetic and biochemical observations obtained to date. First of all, the finding that Dna2 formed a physical complex with Fen-1 in carrying out its essential function (32) strongly favors our model that Dna2 and Fen-1 act in a concerted fashion. Our model is also consistent with genetic observations that overexpression of Fen-1 suppressed the temperature sensitivity of the dna2-1 allele, and overexpression of Dna2 partially compensated for the growth defect of the rad27 deletion at the restrictive temperature (32,44). This could account for the synthetic lethality observed when a rad27 deletion was combined with several mutant alleles of dna2 (32,44), which suggests that Fen-1 becomes essential in the absence of active Dna2 protein. Recently, the structures of DNA lesions were assessed in yeast strains deficient in lagging strand DNA synthesis (48). This study showed that dna2 and rad27⌬ mutant strains contained DNA lesions characteristic of aberrant lagging strand synthesis, in keeping with our model that Dna2 and Fen-1 collaborate in the processing of Okazaki fragments. The hypothesis that more than two enzymes could be involved in Okazaki fragment maturation is also supported by recent in vivo studies that demonstrated that double-strand break (DSB) repair in yeast requires both leading and lagging strand DNA synthesis (30). In this study, DSB-induced gene conversion at the MAT locus of S. cerevisiae was analyzed in mutant strains thermosensitive for essential replication factors. Virtually all of the known replication fork proteins affected DSB-induced gene conversion events, although to varying degrees. Gene conversion was decreased 50% in RAD27 null strains, compared with wild type strains, regardless of growth conditions. This confirms that Fen-1 is required for Okazaki fragment processing but at the same time presents strong evidence for the existence of redundant functions for this process, most likely Dna2. Recent genetic analyses of rad27 mutant alleles showed that rad27-n (a nuclease-deficient allele of RAD27) inhibited the growth of mutant cells, whereas rad27-p (a PCNA-binding defective allele) did not (49). Interestingly, intragenic combination of both mutations (rad27-n,p) nullified the deleterious effect of rad27-n on cell growth. These results suggest that the interaction of Rad27-n with PCNA allows the mutant protein to occupy its normal position within a multiprotein complex, preventing access of an alternative enzyme capable of processing the Okazaki fragment to the incompletely processed DNA, whereas Rad27-n,p cannot be assembled into the complex, allowing free access of the redundant enzyme. These results support the model that the two processing enzymes collaborate to process efficiently Okazaki fragments.
The Possibility of Parallel Pathways of Dna2 Versus Fen-1/ RNase HI in Okazaki Fragment Processing-The deletion of rad2 ϩ in S. pombe did not result in the temperature-sensitive phenotype, unlike S. cerevisiae (data not shown). Instead, the deletion mutant cells were capable of growing at both 25 and 37°C, emphasizing again that Fen-1 is not absolutely required and other enzyme(s) can replace its function in DNA replication. Although our model could account for many genetic and biochemical observations, it does not necessarily rule out the possibility that two nucleases, Dna2 and Fen-1, constitute parallel pathways under some circumstances. If elongation of Okazaki fragments by pol ␦ were inefficient, for example, Okazaki fragments could be susceptible to RNase HI and Fen-1 before the upstream polymerase arrives. In this case, the 5Ј to 3Ј exonuclease activity of Fen-1 comes into play and can contribute to the removal of the RNA-initiated primer with the aid of RNase HI, which is completely independent of Dna2. Why eukaryotic cells are equipped with multiple pathways for Okazaki fragment maturation deserves a brief speculation. A single enzyme may not be sufficient to deal with every Okazaki fragment processing event. Since Okazaki fragment maturation is a major and complicated enzymatic step in DNA replication, the absence of a back-up or redundant mechanism(s) may compromise the survival of eukaryotic organisms.
Possible Roles of Dna2 Helicase Activity in Okazaki Fragment Maturation-In this model, we can postulate that Dna2 plays a number of roles. First of all, the unique feature of Dna2 is its limited unwinding activity (Ref. 33 and this study) and its ability to interact with pol ␦ (34). The unwinding activity of the Dna2 helicase is likely to be tightly coordinated with displacement synthesis by pol ␦, regulating the extent of unwinding of the 5Ј region of the Okazaki fragment. In fact, pol ␦ can catalyze DNA displacement synthesis up to 274 base pairs (24,50), longer than what is required for Okazaki fragment processing. Considering the relatively small size (100 -150 nt) of an average Okazaki fragment, the limited unwinding activity of Dna2 is probably an important feature of the enzyme. This activity could facilitate the timely disassembly of the pol ␦ complex from the template, when no further displacement is required. If this is not the case, uncontrolled displacement synthesis catalyzed by pol ␦ would lead to (i) unnecessary extensive degradation of the pre-existing Okazaki fragment and (ii) the formation of lengthy ssDNA segment with the potential to form secondary structures that would be resistant to cleavage by Dna2 or Fen-1. Thus, one essential role of Dna2 could be to prevent unnecessary displacement synthesis by pol ␦. Another possible role of the Dna2 helicase activity could be to help remove transient secondary structures in displaced strands, which in turn could facilitate the sliding of Fen-1 to the duplex junction (51). It may also help to load Fen-1 onto cleaved ends of ssDNA generated by Dna2, thus coordinating and coupling the two endonucleolytic reactions. The unwinding activity of Dna2 may be required for such a coordination for maximal efficiency, since inactivation of Dna2 leads to cell death (52) only when cells are grown in rich media (44). The roles for Dna2 described above may not be substituted by other nucleases, accounting for the fact that Dna2 is essential in vivo, whereas Fen-1 is not.
In order to test our model, we plan to develop an in vitro system for Okazaki fragment maturation. Such a system may clarify the role played by Dna2 in Okazaki fragment synthesis as well as help to define the functional interactions between Dna2 and other proteins that act in lagging strand synthesis.