Inhibition of flap endonuclease 1 by flap secondary structure and relevance to repeat sequence expansion.

Recent genetic evidence indicates that null mutants of the 5'-flap endonuclease (FEN1) result in an expansion of repetitive sequences. The substrate for FEN1 is a flap formed by natural 5'-end displacement of the short intermediates of lagging strand replication. FEN1 binds the 5'-end of the flap, tracks to the point of annealing at the base of the flap, and then cleaves. Here we examine mechanisms by which foldback structures within the flap could contribute to repeat expansions. Cleavage by FEN1 was reduced with increased length of the foldback. However, even the longest foldbacks were cleaved at a low rate. Substrates containing the repetitive sequence CTG also were cleaved at a reduced rate. Bubble substrates, likely intermediates in repeat expansions, were inhibitory. Neither replication protein A nor proliferating cell nuclear antigen were able to assist in the removal of secondary structure within a flap. We propose that FEN1 cleaves natural foldbacks at a reduced rate. However, although the cleavage delay is not likely to influence the overall process of chromosomal replication, specific foldbacks could inhibit cleavage sufficiently to result in duplication of the foldback sequence.

Recent biochemical and genetic data have served to illustrate the dual roles of many proteins in DNA metabolism. Enzymes first identified as central components of DNA replication forks, such as the flap endonuclease 1 (FEN1), 1 are now understood to play critical roles in DNA repair pathways (1)(2)(3)(4)(5). Our increasing understanding of the mechanisms of these enzymes is revealing the nature of their multiple functions in the cell.
FEN1, a member of the RAD2 superfamily, is a structurespecific nuclease known to be involved in both lagging strand synthesis during DNA replication (6 -10) and long patch base excision repair (11). Genetic studies highlight the central role of FEN1 in these cellular processes (12)(13)(14). In Saccharomyces cerevisiae, a null mutant of the FEN1 homologue (RAD27/ RTH1) is conditionally lethal at high temperatures producing a cellular morphology indicative of an S phase arrest. At the permissive temperature, FEN1 mutants exhibit slow growth and hyper-recombination phenotypes consistent with defects in DNA replication and recombination. Null mutants also have an increased sensitivity to the alkylating agent methyl methanesulfonate but are only moderately affected by UV or ionizing radiation. These characteristics are consistent with participation of FEN1 in base excision repair.
Biochemical analyses have clarified the nature of the FEN1 catalyzed reactions during DNA replication and repair. Reconstitution of lagging strand DNA synthesis in vitro showed that FEN1 is needed to remove the initiator RNA primers of Okazaki fragments (3). FEN1 assists in primer removal through two proposed pathways. First, the nuclease RNase HI cleaves within the RNA primer leaving a single ribonucleotide remaining at the 5Ј-end of the Okazaki fragment. The FEN1 nuclease removes this ribonucleotide prior to ligation with an upstream Okazaki fragment (9). Alternately, synthesis from an upstream Okazaki fragment may cause the displacement of the RNA primer generating an unannealed 5Ј-tail or flap structure (15,16). FEN1 cleaves endonucleolytically at the base of the flap, thereby removing the entire segment of RNA (17)(18)(19).
The endonucleolytic activity of FEN1 is thought to be important for the removal of damaged nucleotides in long patch base excision repair (11,20). During repair of an abasic site, an apurinic/apyrimidinic endonuclease cleaves on the 5Ј-side of the abasic sugar generating a nick within the DNA (21). We have previously demonstrated that FEN1 cannot remove the abasic sugar (22). Instead, when the damaged sugar and an additional downstream nucleotide are displaced to generate a flap, FEN1 endonucleolytically removes the site of damage as part of an oligomer. The resulting short gap is filled and ligated to complete the repair process.
FEN1 employs a unique cleavage mechanism for substrates containing unannealed 5Ј-tails or flap structures (Fig. 1). FEN1 removes the flap by recognizing the 5Ј-end, tracking the length of the tail, and cleaving at the point of annealing (1,18). Flap substrates composed of either RNA or DNA are readily cleaved by FEN1 (19). However, the unannealed 5Ј-tail must be singlestranded as the presence of large adducts (18) or annealed primers prevent FEN1 cleavage (18,23).
The FEN1 family of nucleases is conserved throughout evolution with homologues identified from archaebacteria, yeast, Xenopus, and mammals (1, 24 -28). Biophysical analysis of homologues has begun to provide a structural context for understanding the molecular mechanism of FEN1. Crystal structures of FEN1 homologues from T5 exonuclease (29), T4 RNase H (30), Methanococcus jannaschii FEN1 (31), and Pyrococcus furiosus FEN1 (32) reveal a helical arch or loop above a globular domain containing the active site. This arch may be utilized by the nuclease to track onto the 5Ј-end of a flap structure (29). Mutational analyses of the loop region in the M. jannaschii FEN1 indicate that this physical structure is critical for both the binding and cleaving of flap substrates (31).
Previous results show that the ability of the FEN1 endonuclease to cleave flap structures is affected by a range of modi-fications of the unannealed 5Ј-tail. These modifications include the addition of large adducts such as biotin-streptavidin complexes (18), small adducts (cis-platinum) (33,34), and oligonucleotide primers annealed to the tail (18,23). These results strongly suggest that FEN1 tracks down the length of the tail before reaching the cleavage point. However, it still remains unknown whether FEN1 "tracks" by threading the length of the flap through an arch structure (35) or whether the nuclease grips the flap in the manner of enzymes with thumb domains (36). Recent studies utilizing flap substrates containing a branch structure permit cleavage by FEN1 (34). These results suggest a tracking process in which the flap is not completely encircled.
In addition to DNA replication and repair, genetic and biochemical data also connect FEN1 with repeat sequence expansion. Many human genetic diseases are characterized by a substantial alteration of the genome. One class of changes observed involve a unique group of sequences referred to as repetitive sequences. These stretches of the genome contain either single, dinucleotide, or trinucleotide sequences repeated in an array of several to about 60 nucleotides (microsatellites) or longer 4 -100-nucleotide sequences repeated to greater lengths (minisatellites) (37). Repeat expansions have attracted attention particularly because of their link with a host of genetic conditions called human triplet repeat disorders (38). These disorders include several neurodegenerative diseases such as Huntington's disease, Friedreich's ataxia, and some colon cancers. A notable characteristic of these sequences is their ability to adopt higher ordered structures such as hairpin loops both in vitro (39,40) and in vivo (41).
Although the expansion of repetitive sequences is well described with regard to the alterations seen in patients, information on the mechanism of expansion only recently has become available. First, the nature of the sequence itself influences expansion. Examination of families with Huntington's disease indicates that the rate of expansion is strongly linked to DNA sequence (38). As a result of the repeating nature of the sequences, regions that expand are able to adopt secondary structure. In addition, an increase in expansion rate is seen if the sequences are located on the lagging strand during DNA replication in yeast (42,43). It has been observed in yeast that mutations in the proteins FEN1 and replication factor C lead to an increase in the length of repetitive sequences (37, 44 -49). These observations suggest that repeat sequences expand during the RNA removal and joining steps of Okazaki fragment processing. Most likely foldbacks formed from the repeat sequences interfere with the required cleavage reaction of FEN1. Because primers annealed to a flap inhibit cleavage, a foldback in the flap is anticipated to inhibit by a similar mechanism. The resulting delay in the removal of flaps with foldbacks leads to an expansion of the repeat. These phenotypes of yeast are relevant to human FEN1 not only because the two nucleases exhibit a high degree of sequence homology but also because human FEN1 will rescue the defects of the null mutant in yeast (50).
The link between the structure of repetitive sequences and the mechanism of FEN1 has led to the proposal of a model to explain repeat expansion (37). During DNA synthesis, displacement by a polymerase or helicase allows repetitive sequences to self-anneal forming secondary structure such as a hairpin loop or foldback within the DNA. Equilibration of this intermediate with the template would form a "bubble" intermediate. Synthesis from the upstream primer and ligation would prevent any further degradation of the extra repeats. Resolution of this structure during replication would lead to expansion (Fig. 1). The model explains expansion of both trinucleotide repeats and the longer minisatellite repeats.
Using a series of substrates designed to adopt secondary structure, we have examined the ability of FEN1 to resolve the foldback and bubble structures formed in the model. In addition, we examined the ability of PCNA and RPA, which stimulate FEN1 activity, to assist in the removal of these intermediates. Results suggest that the model correctly describes a mechanism for repeat expansion.

EXPERIMENTAL PROCEDURES
Materials-Oligonucleotides were synthesized either by Integrated DNA Technologies (Coralville, IA) or by Genosys Biotechnologies (The Woodlands, TX). Radionucleotides [␥-32 P]ATP (6000 or 3000 Ci/mmol) and [␣-32 P]dCTP (3000 Ci/mmol) were obtained from NEN Life Science Products. T4 polynucleotide kinase and Klenow fragment of DNA polymerase I (labeling grade) were from Roche Diagnostics. All other reagents were the best available commercial grade.
Oligonucleotide Substrates-Oligonucleotide primers were designed to form a series of flap or bubble substrates. For flap substrates, the 3Ј-end of each downstream primer is complementary to the 5Ј-end of its appropriate template. The 5Ј-end of the downstream primer forms the unannealed 5Ј-tail or flap. Within the 5Ј-end of downstream primers containing secondary structure is an inverted repeat that forms a hairpin loop or foldback. The length of the stem varies from 6 -24 nucleotides. Upstream primers anneal to the 3Ј-end of the template forming a nick at the base of the flap. Bubble substrates contain 25 nucleotides of complementarity at both the 5Ј-and 3Ј-ends of the primers, generating an internal unannealed region. Upstream primers anneal to this internal region, forming a nick at the 3Ј-end of the bubble. Oligomer sequences are listed in Table I. Substrates were constructed as described in the figure legends.
Substrates were generated by annealing a downstream primer, template, and upstream primer at a molar ratio of 1:2.5:5, respectively. A downstream primer and template were placed in 50 l of TE (10 mM Tris-Cl, pH 8 and 1 mM EDTA) and heated to 100°C for 5 min. The reaction was placed at 70°C and allowed to slowly cool to 25°C. After an upstream primer was added, the mixture was incubated at 37°C for 30 min to 1 h. Enzyme Assay-Assays contained the indicated amounts of substrate and FEN1 in reaction buffer (30 mM HEPES, pH 7.6 (diluted from a 1 M stock), 40 mM KCl, 8 mM MgCl 2, 5% glycerol or 0.01% Nonidet P-40, and 0.1 mg/ml bovine serum albumin) in a final volume of 20 l. Assays were incubated at 37°C for 15 min and stopped by the addition of 10 l of termination dye (95% formamide (v/v) with bromphenol blue and xylene cyanol). After heating to 95°C for 5 min, samples were separated on a 12% polyacrylamide, 7 M urea denaturing gel. Products were detected by PhosphorImager (Molecular Dynamics) and analyzed using ImageQuant v1.2 software from Molecular Dynamics. All assays were performed at least in triplicate.

RESULTS
We propose that certain FEN1 substrates form secondary structures that influence the kinetics and specificity of cleavage. We examined FEN1 cleavage on substrates with foldbacks in the flap and bubble structure intermediates. This permits us to assess their importance in the context of the model for repeat expansions.
Cleavage of Flap Substrates Containing Secondary Structure-To determine whether FEN1 cleaves flaps containing secondary structure, we designed substrates with an inverted repeat within the 5Ј-end of the flap that forms a stem-loop structure. The length of the annealed region of the stem varied from 6 to 24 nucleotides, and the loop contained 6 nucleotides. To control for the effect of sequence, each of the stem loops had similar nucleotide composition and GC content. The control substrate contained a 26-nucleotide flap consisting mostly of thymidine residues producing an unannealed 5Ј-tail devoid of secondary structure. All other sequences were identical for each substrate.
The downstream primers of the substrates were radiolabeled at the 5Ј-end. The substrates were exposed to increasing amounts of FEN1 (Fig. 2). Addition of FEN1 to the control substrate releases a 25-nucleotide product (lanes 1-5) representing cleavage at the base of the flap. If the presence of secondary structure had no effect on FEN1 activity, the expected products would be 24, 36, and 60 nucleotides in length for the 6, 12, and 24 nucleotide stem substrates, respectively. FEN1 is only able to readily cleave the 6-nucleotide stem (lanes 6 -10) with similar specificity as the flap with no foldback. Cleavage of the 12-nucleotide stem (lanes 11-15) yields both a full-length product from release of the 5Ј-tail and more significant levels of mono-and dinucleotide products. Cleavage of the 24-nucleotide stem resulted in only the smaller mono-and dinucleotide products (lanes 16 -20).
As the foldback stem became more stable with increasing length, FEN1 was blocked from cleaving directly at the base of

FEN1 Cleavage of Flap Secondary Structures
the flap. Instead, FEN1 cleaved the 5Ј-end of the self-annealed tail, generating mono-or dinucleotide products. These products are expected because the foldback is a substrate for the 5Јexonuclease activity of FEN1 (1,17). Appearance of the monoand dinucleotide products also confirms the ability of each sequence to adopt the predicted secondary structure, because FEN1 is an obligate double-stranded exonuclease. As the length of the annealed region increases, so does the stability of the foldback, allowing the nuclease to more efficiently attack the 5Ј-end. However, cleavage of the 5Ј-radiolabeled nucleotide prevents further analysis regarding the fate of the stem loop present within the flap. It is possible that FEN1 continued to cleave as an exonuclease until the double-stranded region was sufficiently small that the stem melted, forming a more favorable flap structure for FEN1.
For further examination of the effect of secondary structure, substrates were radiolabeled at the 3Ј-end of the downstream primer and incubated with increasing amounts of FEN1. To determine more precisely the length of secondary structure necessary to influence FEN1 activity, the annealed regions of the stems were made 6, 12, 18, and 24 nucleotides long. Because the length of the annealed downstream region of each substrate is identical, cleavage by FEN1 resulted in the generation of a 19-mer for all substrates (Fig. 3A). FEN1 cleavage was easily detectable on the control substrate and substrates containing 6 and 12 nucleotides (Fig. 3A, lanes 1-15). The ability of FEN1 to cleave the 18-and 24-nucleotide stem substrates is significantly reduced (Fig. 3A, lanes 16 -25). At higher levels of FEN1, additional cleavage products smaller than the 19-mer are observed. These are the result of exonucleolytic cleavage within the annealed double-stranded region of the downstream primer. We also observed some exonucleolytic removal of the 5Ј-end as expected from the results of Fig.  2. Production of the 19-mer does not appear to derive from progressive exonucleolytic cleavage through the stem, because there are no detectable intermediates between the full-length substrate and the product. These data were quantitated and plotted as the percentage of substrate converted to the 19nucleotide and smaller products versus the amount of FEN1 (Fig. 3B). It is readily apparent that there is a transition between 12 and 18 nucleotides within the stem that has a major inhibitory affect on FEN1 cleavage. Kinetic analysis of the degradation of 3Ј-radiolabeled substrate over time produced results consistent with those obtained here by increasing FEN1 concentration (data not shown).
Effect of the Position of the Secondary Structure Relative to the Cleavage Site-We were concerned that the inhibitory effect of foldbacks reflects a steric hindrance by the stem loop of FEN1 interaction with the cleavage site. This would prevent FEN1 from cleaving the substrate and would most likely correlate with the stability of the stem-loop structure. We examined this possibility by varying the distance from the 5Ј-end of the stem to the first nucleotide annealed to the template (designated b in Fig. 4). Substrates containing the 18-nucleotide stem were lengthened to place 0, 6, or 18 nucleotides between the 5Ј-end of the stem and cleavage site. Again, the expected product for all substrates is a 19-mer representing the annealed portion of the downstream primer. The control was easily cleaved by FEN1 (Fig. 4, lanes 1-5) as compared with the less efficient cleavage observed for the 18-nucleotide stem at the 6-nucleotide distance (Fig. 4, lanes 11-15). This distance is equivalent to that of the 18-nucleotide stem used in Fig. 3. Placement of the stem at a distance of 0 or 18 nucleotides had little additional affect on FEN1 activity (Fig. 4, lanes 6 -10 and  16 -20). The cleavage of all three of these substrates is reduced to a similar extent. We conclude that the inhibitory nature of secondary structure results from the requirement of a free 5Ј-end by FEN1. These results continue to demonstrate the importance of the 5Ј-end of the substrate for the cleavage mechanism of FEN1.
Cleavage of Trinucleotide Repeats by FEN1-Recently, there has been significant interest in the mechanism of trinucleotide repeat expansion. One model suggests that the ability of these repeats to adopt secondary structure prevents their removal during DNA replication and repair (37). Given that secondary structure is inhibitory to FEN1, we examined the effect of trinucleotide repeats within flap structures. The substrates that we designed contained an increasing number of CTG repeats at the 5Ј-end of the flap.
Substrates with either 5, 10, or 20 CTG repeats in the flap ((CTG) 5 , (CTG) 10 , or (CTG) 20 ) were incubated with FEN1 over time (Fig. 5A). The control substrate, (CTG) 5 and (CTG) 10 were cleaved efficiently by FEN1 (lanes 1-6 and 13-24). The ability of FEN1 to remove the (CTG) 20 was significantly reduced as compared with the other substrates (lanes 7-12). The data were quantitated and plotted as the percent of substrate converted to product versus time (Fig. 5B). During the reaction at 10 min, FIG. 3. Increasing length of secondary structure inhibits FEN1 cleavage. The ability of FEN1 to cleave foldback structures was examined using 3Ј-radiolabeled substrates. Each substrate (10 fmol) was incubated with an increasing amount of FEN1 (10, 50, 100, and 500 fmol) at 37°C as described under "Experimental Procedures." A, reaction products were separated on a 12% denaturing polyacrylamide gel. A schematic diagram of the substrates is depicted above the figure. The length (in nucleotides) of the annealed region of the secondary structure is designated as a. The control substrate contains a standard flap structure. Reactions in lanes 1, 6, 11, 16, and 21 contain only substrate. The size (in nucleotides) of substrates and location of cleavage products are as indicated. Annealing primers D control , T stem , and U 25 generated the control substrate. Each foldback substrate contains D 6 stem , D 12 stem , D 18 stem , or D 24 stem annealed to T stem and U 25 . B, the results in A were quantitated by PhosphorImager analysis. The percentage of substrate converted to product was plotted against the amount of FEN1. FEN1 cleaved ϳ 45, 41, and 35% of the control, (CTG) 5 , and (CTG) 10 , respectively. By comparison, only 3% of the (CTG) 20 substrate was converted to product by FEN1. The inhibition of FEN1 by the (CTG) 20 most likely reflects the ability of this sequence to form a stable secondary structure (54). As with the stem-loop structures used in the previous experiments, we do not observe any intermediates resulting from the exonucleolytic degradation of the 5Ј-end of the substrate. However, we were able to observe exonucleolytic degradation of the (CTG) 20 at extremely high levels of FEN1 (data not shown).
Cleavage by FEN1 of Intermediates Proposed to Form during Repeat Expansion-The model of repeat expansion (Fig. 1) includes an intermediate in which the flap anneals to the template forming a bubble. Because a free 5Ј-end of the flap appears to be required for efficient FEN1 activity (23), the bubble is expected to be a poor substrate. It may not be completely inert, however, because FEN1 can degrade the annealed 5Ј-region exonucleolytically until the flap becomes single-stranded.
To examine the course of events when FEN1 encounters such an intermediate, we created a model bubble substrate. The top strand has 25-nucleotide regions at both its 5Ј-and 3Ј-ends annealed to the bottom strand, with a central 30-nucleotide unannealed bubble region. The bottom strand has a central 25-nucleotide unannealed region. The different lengths of the top and bottom single strands are designed to encourage the strands to physically separate into structures resembling flaps. The control substrates contain a different bottom strand lacking the sequences that allow the 5Ј-end of the top strand to anneal to the bottom strand, resulting in a conventional flap structure. The top strands of each substrate were labeled at the 3Ј-end. This allowed us to detect exonucleolytic cleavage at the 5Ј-end of the top strand and to detect endonucleolytic cleavage at the site equivalent to the base of the flap.
Control substrates incubated with FEN1 were efficiently cleaved over time. Cleavage of the control substrate in the presence of an upstream primer resulted in the release of a specific 25-nucleotide product, corresponding to cleavage at the base of the flap (Fig. 6A, lanes 25-30). We previously observed that the presence of an upstream primer sometimes stimulates FEN1 cleavage depending on the sequence surrounding the base of the flap (55). Comparison of the control substrate in the presence or absence of an upstream primer (lanes 1-6) indicated that the upstream primer moderately increases cleavage of the downstream primer. Cleavage of the control substrate lacking an upstream primer resulted in products of differing sizes. These products arise because nucleotides within the long flap transiently anneal to the template upstream of the base of the flap, generating alternate cleavage sites. The presence of the upstream primer blocks transient annealing of the downstream primer to the template, eliminating the longer cleavage products seen in the absence of the upstream primer.  5. Cleavage of CTG repeats by FEN1. The effect of a triplex sequence on FEN1 activity was examined using the sequence (CTG) that shows preferential expansions in vivo. FEN1 cleavage of flaps containing (CTG) repeated 20, 10, or 5 times was compared with a control lacking secondary structure over time. Assays were assembled at 4°C and contained 80 fmol of FEN1 and 80 fmol of substrate in a total volume of 160 l. Reactions were placed at 37°C, and 20-l aliquots were removed at the indicated time (min) and stopped by the addition of 10 l of termination dye at 95°C. A, products were separated on a 10% denaturing polyacrylamide gel and detected by Phos-phorImager. A schematic diagram of the substrates is depicted above the figure. Reactions in lanes 1, 7, 13, and 19 contain only substrate. The control substrate contains a standard flap structure. The number of CTG repeats is noted as n. The size (in nucleotides) of substrates and location of cleavage products are as indicated. Annealing primers D control , T stem , and U 25 generated the control substrate. Each CTG substrate contains D (CTG)20 , D (CTG)10 , or D (CTG)5 annealed to T stem and U 25 . B, the data in A were quantitated and presented as the percentage converted to product versus time.

FEN1 Cleavage of Flap Secondary Structures
Compared with the control substrate, we observed very inefficient cleavage of the bubble substrates (Fig. 6A, lanes 7-24). Generally products followed the same pattern as with the control substrates. The low level of cleavage observed resulted from a small portion of the substrates that fail to completely anneal. We again considered whether a primer annealed to the template would be stimulatory. A primer as long as the singlestranded region of the bottom strand (lanes 13-18) or a shorter 15-nucleotide primer (lanes 19 -24) had little effect on the efficiency of cleavage. However, the presence of either primer resulted in a more precise cleavage product as with the control substrates. At these levels of FEN1 we did not observe any noticeable cleavage of the 5Ј-end of the bubble substrate. At higher levels of FEN1, exonucleolytic degradation of the substrate was apparent. In fact, only exonucleolytic products are observed at high levels of FEN1 when the top strand is radiolabeled at the 5Ј-end (data not shown) (23).
We also considered that FEN1 might require an extension of the 3Ј-end of the bottom strand for most efficient exonucleolytic cleavage of the annealed region. The presence of an extended segment more closely models the proposed expansion intermediate. The top strand was annealed to a template generating either a blunt end or a 10-nucleotide 3Ј-overhang. At increasing levels of enzyme, FEN1 more efficiently cleaved the substrate containing the upstream extension compared with the blunt end (Fig. 6B, lanes 8 -21). However, cleavage of either substrate was significantly reduced when compared with the level of cleavage seen with the control flap substrate (lanes 1-7). For these proposed intermediates, exonucleolytic cleavage precedes the endonucleolytic cut on each strand. Almost certainly it is this need for two classes of cleavage that delays cleavage in the bubble compared with the flap substrates. Overall, the results show that FEN1 has evolved to deal with the bubble intermediate but necessarily suffers a delay associated with the extra exonucleolytic step.
Ability of Accessory Proteins to Assist FEN1 in the Removal of Foldbacks-Several proteins have been identified that interact with and affect FEN1. We considered that addition of these proteins might reduce the level of inhibition seen with stemloop structures by stimulating FEN1. PCNA physically interacts with FEN1 and increases FEN1 activity ϳ30-fold (23, 56 -58). We have recently shown that PCNA decreases the K m of FEN1 for its substrates, thereby increasing its ability to bind and cleave (59). Addition of PCNA to stem loops may allow FEN1 better access to the substrate and facilitate enhanced resolution of the secondary structure.
Increasing amounts of PCNA were incubated with FEN1 in the presence of a control substrate or substrates containing 6 -24-nucleotide stems (Fig. 7A). At these levels of FEN1, very low amounts of cleavage are observed on the control substrate. However, addition of PCNA results in a dramatic increase in cleavage products (compare lane 2 with lane 5). Stimulation of FEN1 by PCNA is seen for the 6-and 12-nucleotide stem substrates (lanes 6 -15). However, little enhancement occurs on the 18-and 24-nucleotide stem structures (lanes 16 -25). Although PCNA can stimulate FEN1, PCNA is unable to alter the substrate specificity for cleavage and reduce the negative effects of secondary structure. The data for the control and the 12-and 24-nucleotide stems were quantitated and plotted as the fold stimulation of FEN1 cleavage versus PCNA concentration (Fig. 7B). The results indicate that although FEN1 cleavage is substantially increased by PCNA, the level of cleavage of a 12-nucleotide foldback substrate was not enhanced relative to that of the control. In fact, the presence of PCNA was unable to generate any cleavage from the nearly inert 24-nucleotide foldback substrate. It is readily apparent that the capacity for PCNA to stimulate FEN1 is limited by the intrinsic ability of the nuclease to cleave a particular substrate. PCNA improves the reactions performed by FEN1 without altering its specificity. High levels of PCNA are needed for these reactions because loading of PCNA onto a linear substrate is dependent on diffusion.
In addition to PCNA, the single-stranded DNA-binding protein RPA also stimulates FEN1, although to a lesser degree (ϳ2-3-fold) (60). The ability of RPA to bind DNA and cause local unwinding may aid in the formation of a single-stranded flap better suited to FEN1 cleavage. We tested the ability of RPA to stimulate cleavage of foldbacks under standard conditions for FEN1 cleavage and under varying salt and MgCl 2 concentrations. Although we observed a moderate increase in the activity of FEN1 in the presence of RPA on control sub- Because FEN1 can exonucleolytically degrade double-stranded DNA, we determined whether FEN1 is able to resolve these structures in model substrates. A, FEN1 (150 fmol) was incubated with bubble substrates (60 fmol) in the absence or presence of upstream primers in a total volume of 120 l at 37°C. Aliquots (20 l) were removed at 1, 3, 5, 10, and 20 min and stopped as described under "Experimental Procedures." A schematic diagram of the substrates is depicted above the figure. The size (in nucleotides) of substrates and location of cleavage products are as indicated. Primers D bubble and T bubble.1 were annealed in the absence (lanes [7][8][9][10][11][12] or presence of either U 25 (lanes [13][14][15][16][17][18] or U 15 (lanes 19 -24) to generate the bubble substrates. Control reactions contained primers D bubble and T control annealed in the absence (lanes [1][2][3][4][5][6] or presence of U 15 (lanes [25][26][27][28][29][30]. Reactions in lanes 1,7,13,19, and 25 contain only substrate. B, FEN1 was also incubated with a bubble substrate containing a 3Ј-extension on the template. Increasing amounts of FEN1 (10, 50, 100, 500, and 1000 fmol) were incubated with 10 fmol of a bubble substrate with either a blunt end or a 10-nucleotide extension at 37°C as described under "Experimental Procedures." Primer D bubble was annealed to either T bubble.2 (lanes 8 -14) or T bubble2.extend (lanes [15][16][17][18][19][20][21] to generate the blunt end and extension substrates, respectively. Reactions in lanes 1, 8, and 15 contain only substrate. The control contains primer D control annealed to T stem (lanes  1-7).
strates, the addition of RPA did not improve FEN1 cleavage of flaps containing secondary structure (data not shown). DISCUSSION Our results demonstrate that foldback structures in flaps and bubble structures formed by annealing the 5Ј-end region of the flap to the template are both inhibitory to cleavage by FEN1. The results are consistent with a mechanism proposed for repeat sequence expansions in vivo (37). In this mechanism (Fig. 1), a foldback on the flap created during removal of an Okazaki fragment primer delays cleavage. The long-lived flap can equilibrate to other structures, including annealing to the template at a location different from its original position. The bubble formed by that annealing also inhibits cleavage, providing time for synthesis from an upstream primer and ligation. This process causes the duplication of sequence on the newly synthesized strand. A repair step then duplicates the sequence on the parental strand.
The model is consistent with expansion of triplet repeats or the duplication of the larger minisatellite repeats. Both expansions are observed to accelerate in S. cerevisiae harboring a deletion in RAD27 (37, 44 -48). In the first case, triplets such as CTG can form secondary structures sufficiently stable to decrease the rate of cleavage. This promotes the remaining steps of the mechanism. In the second case, an inverted repeat sequence, such as those we have tested, would cause the flap to resist cleavage. This would provide an opportunity for the 5Ј-end region to find a new position of complementarity or partial complementarity on the template. These reactions must be very infrequent because an opportunity exists for them to occur in humans at the 5Ј-ends of millions of Okazaki fragments, each time that the chromosome is duplicated.
Our results show that the degree of inhibition of cleavage is strongly influenced by the length of the annealed region of the foldback. This suggests that at least one reason for the infrequency of sequence expansion in wild type cells is that short foldback sequences are inefficient inhibitors of cleavage. For example, 20 -30% inhibition does not occur in our assays until the complementarity of the foldback is 12 nucleotides. The likelihood of a perfect hairpin of 12 nucleotides forming in a region of, for example, 60 nucleotides is about 1 in 1000. This makes the opportunity for formation of a large foldback infrequent within the genome. Of course, because of their large numbers, the 5Ј-ends of Okazaki fragments will not be far from any site that can form a very stable hairpin.
Considering the deleterious effects of sequence expansions, we expect that the normal processes of DNA replication and repair have evolved to resist expansion by preventing steps in the expansion mechanism. Taking this view suggests explanations for two puzzling features of the current model of Okazaki fragment processing. First, why does FEN1 have an exonuclease activity? It would appear to be able to perform all of its functions in both Okazaki fragment processing and DNA repair as an endonuclease. The exonuclease function may be retained to degrade the annealed regions of flap foldbacks and bubble substrates. Our results show that the double-stranded portions of these substrates can yield to progressive exonucleolytic degradation from the 5Ј-end. This produces shorter complementary regions, which are more prone to transient unannealing that allows entry of FEN1. It is possible that this function alone provides the necessary evolutionary pressure to retain exonuclease activity.
Second, reconstitution of efficient Okazaki fragment processing in vitro strongly suggests a role for RNase H. Nevertheless, deletion of RNase H activities in yeast have little effect on the efficiency of Okazaki fragment processing (61). Because the primer displacement pathway apparently backs up the use of RNase H, is the role of RNase H in DNA replication simply to provide an alternative pathway for RNA removal? A different possibility is that it is necessary to efficiently degrade foldbacks when Okazaki fragments still containing the initiator RNA are displaced. In this case, the foldback would be partially RNA. Annealed 5Ј-triphosphorylated RNA segments resist the exonuclease activity of FEN1. RNase H, however, readily degrades them. In this case, RNase H would not appear to be necessary for cellular viability but might help to control genome stability.
Results presented here provide further evidence that FEN1 tracks down the flap to the position of cleavage, because blockage of the 5Ј-end region during either a foldback or bubble formation is inhibitory. Cleavage by FEN1 previously was shown to be inhibited by primers annealed to the flap, by biotin-streptavidin complexes on the flap and by some chemical adducts (18,23,34). Analysis of the structures of members of the FEN1 family of nucleases has revealed the presence of a helical arch or loop in the protein just adjacent to the proposed site of catalytic activity. This led to the proposal that FEN1 threads the flap through the hole created by this structure, sliding to the site of cleavage. Measurements showing that the hole was appropriate in size for single-stranded but not double- FIG. 7. Ability of PCNA to stimulate FEN1 cleavage of foldbacks. Increasing amounts of PCNA (100, 500, and 1000 fmol) were incubated with FEN1 (1 fmol) and substrates containing stem-loop structures. All components were assembled at 4°C and then placed at 37°C as described under "Experimental Procedures." A, reaction products were separated on a 12% denaturing polyacrylamide gel. Reactions in lanes 1, 6, 11, 16, and 21 contain only substrate. A schematic diagram of the substrates is depicted above the figure. The length (in nucleotides) of the annealed region of the secondary structure is designated as a. The control substrate contains a standard flap structure. The size (in nucleotides) of substrates and location of cleavage products are as indicated. Substrates were annealed as described in the legend to Fig.  3. B, the results in A were quantitated by PhosphorImager analysis. The fold stimulation of FEN1 by PCNA was plotted against the amount of PCNA. stranded DNA appeared to explain inhibition by annealed primers (29). However, we have recently shown that branch structures on the flap, far from the site of cleavage, are not inhibitory (34). This result would appear to rule out a threading mechanism, in favor of a tracking process that allows for certain large flap modifications. Proving with absolute certainty that a tracking process occurs remains an elusive goal. Virtually every relevant experiment involves blocking the movement of FEN1 on the flap. Results presented here also fall into that category. However, any modification of the flap could also potentially influence interaction of FEN1 with the cleavage site. At this point, a preponderance of evidence, rather than a definitive experiment, supports the tracking mechanism.
Overall, our results show how flap secondary structure and bubble formation can promote steps in a proposed mechanism for repeat sequence expansion. A recent report by Spiro et al. (54) also provides convincing evidence that triplex repeat sequences form structures resistent to FEN1 activity, providing convincing support for the proposed model. Furthermore, the inhibition of FEN1 by these structures provides explanations for the exonuclease activity and the apparent role of RNase H in Okazaki fragment processing.