Enzymes and reactions at the eukaryotic DNA replication fork.

Many recent reviews of eukaryotic DNA replication have emphasized our current understanding of either the initiation and regulation of eukaryotic nuclear DNA replication (1, 2), the DNA polymerases and other proteins involved (3–11), or the entire range of knowledge of the replication process (12–17). Here, we will focus on recent findings concerning specific enzymatic reactions necessary for the growth of the eukaryotic replication fork.

subunit (␣) form a complex at the primer terminus. ␣ is at the 3Ј-end and immediately behind are the other subunits contacting the double-stranded region of the primer-template (28). By both analogy and preliminary experimental evidence, RFC is retained in the complex of PCNA and DNA pol ␦. 2

Synthesis of the Lagging Strand
Priming by DNA pol ␣ and switching to DNA pol ␦ occur in essentially the same way on both the lagging and leading strand of SV40 (29,30) as illustrated in Fig. 1A. In fact, polymerase switching during synthesis of Okazaki fragments was found to be a necessary prerequisite for complete gap filling (30). Analysis of SV40 replication intermediates indicates that priming on the lagging strand is very frequent, with initial placement of primers ϳ 50 nucleotides apart (31). Okazaki fragment intermediates consist of initiator RNAs averaging about 10 nucleotides in length, extended with 10 -20 additional deoxyribonucleotides (31). Because Okazaki fragment intermediates are made in the absence of ATP, and RFC requires ATP for polymerase switching, the deoxyribonucleotides of the intermediates appear to be added by DNA pol ␣ prior to the switch. After loading of PCNA and DNA pol ␦, an additional 10 -20 nucleotides are added prior to the position of the next downstream initiator RNA primer. There must be further extension of the upstream primer during or after removal of the initiator RNA and possibly nick translation synthesis through the first deoxyribonucleotides of the downstream primer. DNA pol ␦ should then dissociate. Results from the homologous E. coli system for DNA pol III (32) indicate that encountering a downstream primer induces the dissociation of the processive complex. If eukaryotes operate similarly, the necessary frequent dissociations of DNA pol ␦ would be promoted by contact with the downstream primer.
Numerous RNA primers must be removed. The action of two nucleases is sufficient to perform this task (33). RNase H1 endonucleolytically cleaves the initiator RNA one nucleotide upstream of the RNA-DNA junction. The initiator RNA is removed intact leaving a single ribonucleotide on the downstream DNA portion of the Okazaki fragment. RNase H1 cleavage leaves the junctional ribonucleotide irrespective of the length of the initiator RNA. The position of cleavage is not affected by either the RNA or DNA sequence around the cleavage site. In fact, cleavage remains specific even if mismatched nucleotides are present at positions at and near the RNA-DNA junction. 3 This indicates that RNase H1 recognizes the junction and not the intermediate structure between the A-form helix of the RNA-DNA duplex and the B-form helix of the DNA duplex. The structure of these regions is likely distorted in the mismatched substrates.
The remaining ribonucleotide is removed (30,33) by FEN1/RTH1 (30, 34 -37), which contains both an exonucleolytic and endonucleolytic capability (38,39). As an exonuclease, it is specific for DNA or 5Ј-RNA terminated DNA primers annealed to templates (40). Depending on the sequence of the region, it may be stimulated, unaffected, or inhibited by the presence of a primer bound immediately upstream of the cleavage site (40). Removal of the junctional ribonucleotide is generally inhibited by an upstream primer. This suggests that the initiator RNA removed by RNase H1 dissociates before FEN1/RTH1-directed cleavage of the junctional ribonucleotide. During synthesis, the extension of the upstream primer by polymerization could stimulate nick translation. When a sequence is encountered such that the upstream primer is inhibitory, nick translation would stop, allowing ligation of the Okazaki fragments. We found that the simultaneous action of a DNA polymerase, RNase H1, FEN1/RTH1, and DNA ligase I results in correct Okazaki fragment processing in vitro (41) as depicted in Fig. 1B.

Alternative Pathways for Okazaki Fragment Processing
Genetic analyses in yeast suggest that there are alternative means of initiator RNA removal. Null mutants of the primary RNase H in yeast are not significantly defective in DNA replication. 4 One possibility is that the yeast RNase H is not equivalent to mammalian RNase H1. Alternatively, there is an efficient second pathway for RNA removal that does not require RNase H. Null mutants of FEN1/RTH1 in yeast are temperature-sensitive for growth and have a hyper-recombination phenotype (42). This is indicative of long lived regions of ssDNA in the chromosome and is symptomatic of a defect in Okazaki fragment processing. This result suggests that a backup pathway which still involves FEN1/ RTH1 could compensate for defective or absent RNase H1. In the absence of FEN1/RTH1, both pathways would fail, reverting to a third alternative pathway inefficient enough to produce the temperature-sensitive phenotype. Recent discovery of at least one additional gene in yeast that encodes a nuclease with a similar structure to FEN1/RTH1 (43,44) suggests the possibility of a third pathway for RNA primer removal.

Possible Role of the Endonuclease Activity of the FEN1/RTH1 Nuclease in Initiator RNA Removal
The mammalian FEN1/RTH1 class nucleases also contain a unique, structure-specific endonucleolytic activity (38,39). A favored substrate consists of two primers on a template having the annealed portions directly adjacent to each other. However, the 5Ј-end of the downstream primer forms an unannealed 5Ј-tail. Then, FEN1/RTH1 could cleave the tail endonucleolytically in the DNA region downstream of the initiator RNA. After the gap between the remaining DNA portions of the two primers is filled, ligation would complete Okazaki fragment processing.
Endonucleolytic cleavage by FEN1/RTH1 occurs by a unique mechanism. The nuclease slides over the 5Ј-end of the unannealed tail and traverses the entire length of the tail before arriving at the point of cleavage near the annealing point of the tail (39,45). One of the most definitive experiments demonstrating an obligatory sliding mechanism involved use of a tail 73 nucleotides long. Annealing of a 20-nucleotide-long primer anywhere on the tail inhibited cleavage (46). Additionally, modification of the most 5Ј-nucleotide with a biotin-containing side chain, followed by binding of streptavidin, prevented entry of the nuclease and cleavage (46).
Calf FEN1/RTH1 readily recognizes the triphosphorylated 5Јend region of a displaced initiator RNA and cleaves in the downstream DNA (47). Surprisingly, several of the substrates used in this study were cleaved in the absence of a nick-like structure. Influences of nucleotide sequence on nick-dependent stimulation and specificity of cleavage were examined; however, no specific criteria could be discerned (47). Overall, results demonstrate that the endonucleolytic activity of the nuclease is capable of bypassing the need for RNase H1 in Okazaki fragment processing. The proposed cleavage mechanism of a displaced Okazaki fragment by FEN1/RTH1 nuclease is shown in Fig. 2.
The structure of the mammalian 5Ј-3Ј-exo/endonuclease has not yet been determined. However, eukaryotic and prokaryotic nucleases of this type have homologous sequences (48). Recently, the x-ray crystal structures of three of the prokaryotic nucleases, Taq 5Ј-exonuclease (49), the bacteriophage T5 5Ј-3Ј-exo/endonuclease (50), and the bacteriophage T4 5Ј-nuclease (also called RNase H) (51), were determined. While the overall structures are similar, Ceska et al. (50) specifically pointed out an arch shape in the T5 nuclease, rising above the globular main section of the protein. Two ␣-helical structures make up the supports of the arch, and the keystone region is an area of random coil. Working from known mechanistic information about this class of nucleases, Ceska et al. (50) proposed a way for the nuclease to interact with its substrate. Their 4 R. Crouch, personal communication. model shows how the double-stranded region of a 5Ј-tailed primer could bind a wide cleft in the globular portion of the protein and the unannealed tail could fit through the arch. Significantly, the opening in the arch is sufficiently wide to allow passage of single but not dsDNA. The single strand then threads through the archway until the region around the annealing point of the tail can bind the cleft for catalysis. The free unannealed strand can then drift through the archway, and the rest of the substrate would dissociate from the cleft.
Both exo-and endonucleolytic mechanisms of initiator RNA processing allow for the removal of some of the DNA downstream of the initiator RNA. This could occur through a nick translation process involving the exonucleolytic action of FEN1/RTH1. Alternatively, a tail longer than the initiator RNA would be displaced and then cleaved endonucleolytically by FEN1/RTH1 near the point of annealing. In either case it is possible that all of the riboand deoxyribonucleotides added by DNA pol ␣ would be removed. This would explain why the high fidelity of chromosomal DNA replication could be maintained, even though the fidelity of synthesis of both RNA and DNA by DNA pol ␣ is relatively low (52). It also could explain why DNA pol ␣ (53) had no need for 3Ј-5Јexonuclease activity.

How Can the Unannealed Tail Substrate Be Generated?
Separation of DNA strands is usually performed by a DNA helicase and a ssDNA-binding protein (RPA) to stabilize the ssDNA. Human RPA is a three-subunit protein necessary for SV40 DNA replication (19,21,54). Studies indicate that RPA is able to unwind both short oligomers and long double-stranded regions in the absence of magnesium and ATP (55,56). However, the rate of unwinding by RPA is significantly slower than the unwinding rate of DNA helicases (56). Therefore, it seems unlikely that RPA alone is responsible for efficient strand separation in vivo. However, RPA does stimulate several DNA helicases, suggesting that it aids in the formation of 5Ј-tails on Okazaki fragments. There are at least three current candidates for a helicase that would create the 5Ј-tail. All track 3Ј to 5Ј on the template strand.
The first is the DNA2 helicase from Saccharomyces cerevisiae (57). This helicase is a 172-kDa, single subunit protein that copurifies with the yeast FEN1/RTH1. The DNA2 gene is essential for viability and complements a temperature-sensitive yeast strain defective in the elongation stage of DNA replication. When an invariant lysine in the nucleotide binding sequence of the helicase domain was altered to glutamic acid, complementation no longer occurred. These results suggest that the helicase activity is at least part of what makes this protein essential. Its specific substrate is a primer with an unannealed 5Ј-tail. This originally suggested a role in separating the replication fork. It would appear that DNA2 helicase could propagate, but not initiate, formation of a 5Ј-tail of an Okazaki fragment. However, such initiation could be performed by polymerases, RPA, another helicase, or other replication proteins. High expression of FEN1/RTH1 complements a temperature-sensitive dna2 mutation (58) supporting the likelihood of a functional interaction in vivo. Together these results argue that the DNA2 helicase is responsible for the creation of unannealed tails on Okazaki fragments.
The second is helicase E from calf thymus (59 -61). This helicase also copurifies with FEN1/RTH1, as well as with DNA pol ⑀. It is apparently a monomer of approximately 100 kDa. It can act on a fully annealed primer and is capable of displacing DNA up to several hundred nucleotides in length (59 -61). The third is the Ku helicase (62). This enzyme associates with human DNA pol ⑀ (63) and has two subunits of 72 and 80 kDa (62). It can also displace fully annealed primers. Ku has recently been identified as part of the complex of proteins that binds the yeast origin of DNA replication ARS121 (64). It is the protein previously designated OBF2. This observation supports a significant role for the Ku protein in DNA replication and possibly the separation of the replication fork. Either of these helicases could have a role in the creation of an unannealed 5Ј-tail on Okazaki fragments.

Stimulation of FEN1/RTH1 Nuclease Action by PCNA
Recently Burgers, Lieber, and colleagues (65) demonstrated that the catalytic action of FEN1/RTH1 is stimulated approximately 10-fold by PCNA. This was shown on substrates having a primer with a unannealed 5Ј-tail. PCNA and FEN1/RTH1 were the only two proteins in the reaction, proving that their interaction created the stimulation. These two proteins can interact in the absence of substrate, as indicated by the binding of FEN1/RTH1 to PCNAcontaining affinity resin. Stimulation requires a large excess of PCNA over substrate molecules, presumably because PCNA freely slides on and off of the linear template. When PCNA was loaded using RFC and ATP onto M13 DNA having two adjacent primers, PCNA stimulated FEN1/RTH1 nuclease-directed cleavage of the terminal 5Ј-nucleotide of the downstream primer (65). A PCNA mutant protein (pcna-52), incapable of trimerization, was inactive (65). The results show that PCNA only stimulates FEN1/RTH1 after encircling the substrate. Blocking the unannealed strand with a primer prevents FEN1/RTH1 nuclease-directed cleavage even in the presence of PCNA (45). This shows that PCNA does not forgive FEN1/RTH1 of its obligation to slide down the unannealed tail. Overall these results suggest that PCNA acts downstream of the displaced tail and possibly stabilizes FEN1/RTH1 at the position of cleavage. If PCNA is always loaded directionally by RFC, one face of PCNA may be designed to interact with polymerases while the other binds FEN1/RTH1. This implies that there are two PCNAs at the replication fork, one serving as the sliding clamp for DNA pol ␦ and the other interacting with FEN1/RTH1. These interactions are part of the model of the proposed pathways for Okazaki fragment processing as shown in Figs. 1 and 2.

Why Does FEN1/RTH1 Nuclease Have Such a Unique
Sliding Mechanism? The ssDNA sliding mechanism exhibited by the FEN1/RTH1 nuclease may have evolved because the nuclease has a dual function in DNA replication and repair. The null mutant of FEN1/ RTH1 in S. cerevisiae has an increased sensitivity to methyl methanesulfonate, an alkylating agent, implicating FEN1/RTH1 in removal of adduct-damaged nucleotides (42).
The sliding mechanism of FEN1/RTH1 suggests a means by which it could participate in repair of adduct-damaged DNA. It could slide past damage on an unannealed 5Ј-tail and then cleave the tail removing the damaged nucleotides (Fig. 3). This approach allows cleavage of a variety of types of damage, without needing to specifically recognize the structure of the damaged nucleotide. To assess the mechanism of nuclease tracking, and its ability to cleave modified DNA, adducts were placed at various locations on the tails of substrates (66). Footprint analysis using micrococcal nuclease indicates that after tracking, but before cleavage, FEN1/RTH1 protects a region of the tail 25 nucleotides long, adjacent to the cleavage site. When cis-diamminedichloroplatinum(II) (CDDP) adducts were placed within or beyond the region protected by FEN1/ RTH1, the 5Ј-tail was cleaved. A CDDP adduct bound to the last two nucleotides at the very 5Ј-end of an eight-nucleotide-long tail was also cleaved by FEN1/RTH1. The nuclease also removes tails containing adducts on the 2Ј-position of the ribose. In contrast, a CDDP adduct just adjacent to the expected cleavage point was inhibitory to the nuclease.
As discussed earlier, biotin adducts can be traversed but not if they are conjugated with streptavidin (46). This suggests an ultimate limit on adduct size, possibly imposed by the size and flexibility of the anticipated arch structure in the nuclease. It appears that the nuclease is designed to tolerate a variety of adduct structures. However, damage could inhibit catalysis if it is too close to the point of cleavage (66). This problem could be remedied in vivo by further strand displacement.
DNA having abasic lesions is cleaved in a similar fashion. Abasic sites on chromosomal DNA are sensitive to type II abasic endonucleases. These cleave upstream of the damage to create a strand with a phosphorylated abasic sugar at its 5Ј-end, annealed to a template just downstream of a second primer. FEN1/RTH1 cannot remove the abasic sugar exonucleolytically. However, strand displacement synthesis from the upstream primer can create a tail, over which the nuclease can slide, and then remove the abasic deoxyribose as part of an endonucleolytically cleaved oligomer (67).

Why Are There Three Nuclear DNA Polymerases?
Genetic studies in yeast demonstrate that DNA polymerases ␣, ␦, and ⑀ are all essential (68 -70). These determinations prompted suggestions that all three performed specific functions at the replication fork (4,71,72). DNA pol ⑀, although capable of highly processive DNA synthesis, was stimulated at high salt concentrations by interaction with PCNA (72). This suggests that it also participates in polymerase switching, having a similar role to that of DNA pol ␦. Consistent with these observations, DNA pol ⑀ was proposed to perform the majority of elongation of either the leading or lagging strand. However, reconstitution of SV40 DNA replication with purified proteins showed that both leading and lagging strand DNA replication is performed efficiently with only DNA pol ␣ and ␦. Recent cross-linking experiments were done to determine the DNA polymerases bound to nascent DNA during replication in cell extracts (73). For SV40 replication, DNA pol ␦ but not ⑀ was found to cross-link. For cellular DNA, both were found to cross-link, but mitogenic stimulation induced pol ⑀ to a considerably lesser extent than DNA polymerases ␣ and ␦. The authors concluded that only DNA pol ␣ and ␦ participate in SV40 DNA replication. Measurements with chromosomal DNA suggest that DNA pol ⑀ is not participating as a sole elongating enzyme for either the leading or lagging strand. Instead it is likely to be performing an essential process that occurs during DNA replication.
Recent years have seen a major expansion of our knowledge of the eukaryotic replication fork. As summarized here, nearly all of the necessary reactions and enzymes involved have been identified, characterized genetically, and reconstituted in vitro. Nevertheless, the exact complement of components, their contacts and interactions, and structure of the complex remain to be determined.