Concerted Action of Exonuclease and Gap-dependent Endonuclease Activities of FEN-1 Contributes to the Resolution of Triplet Repeat Sequences (CTG)n- and (GAA)n-derived Secondary Structures Formed during Maturation of Okazaki Fragments*

There is much evidence to indicate that FEN-1 efficiently cleaves single-stranded DNA flaps but is unable to process double-stranded flaps or flaps adopting secondary structures. However, the absence of Fen1 in yeast results in a significant increase in trinucleotide repeat (TNR) expansion. There are then two possibilities. One is that TNRs do not always form stable secondary structures or that FEN-1 has an alternative approach to resolve the secondary structures. In the present study, we test the hypothesis that concerted action of exonuclease and gap-dependent endonuclease activities of FEN-1 play a role in the resolution of secondary structures formed by (CTG)n and (GAA)n repeats. Employing a yeast FEN-1 mutant, E176A, which is deficient in exonuclease (EXO) and gap endonuclease (GEN) activities but retains almost all of its flap endonuclease (FEN) activity, we show severe defects in the cleavage of various TNR intermediate substrates. Precise knock-in of this point mutation causes an increase in both the expansion and fragility of a (CTG)n tract in vivo. Taken together, our biochemical and genetic analyses suggest that although FEN activity is important for single-stranded flap processing, EXO and GEN activities may contribute to the resolution of structured flaps. A model is presented to explain how the concerted action of EXO and GEN activities may contribute to resolving structured flaps, thereby preventing their expansion in the genome.

Trinucleotide repeats (TNRs) 2 are a member of a class of DNAs termed microsatellites, which are composed of multiple repeats of a 1-6-bp DNA motif in a head to tail configuration (1). TNR expansion at specific gene regions interferes with the expression or properties of their gene product and can manifest in disease. The expansion of TNRs is associated with at least 20 neurological disorders, including Huntington disease, fragile X syndrome, myotonic dystrophy, and Friedreich's ataxia (2). Stretches of (CAG) n and (CGG) n are more prone to secondary structure formation than (AGG) n , (AGT) n , and (CAA) n (3). NMR studies show that both CAG or CTG (4) and CGG (5) form stable hairpin structures, comprising a repeat unit of two GC pairs and a mismatched pair under physiological conditions. However, CAA/GTT repeats form no hairpins in vitro (6,7). All but one of the diseases, Friedreich's ataxia, occurs due to the expansion of GAA repeats in the first intron of the human frataxin gene (8). Repeating GAAs can adopt unusual DNA conformations, primarily a structure containing pyrimidinepurine-pyrimidine (YRY) triple helix with non-Watson-Crick pairs (9 -11) and duplexes that associate strongly with each other called sticky DNAs (12,13). Heidenfelder et al. (14) showed for the first time that expanded GAA and TTC repeats form hairpin structures as well. Thus, multiple structures may play a role in genomic instability. The location of the secondary structure with respect to leading/lagging and nascent/template strands is also important in determining the repeat stability (15,16). In vivo studies of TNR stability have taken advantage of model systems in Escherichia coli, yeast, mice, and cell lines. Processes implicated in repeat tract instability include replication slippage, the direction of replication fork progression, lagging strand errors, Okazaki fragment processing, mismatch repair, gap filling, double strand break repair, and recombination (17). These mechanisms are not mutually exclusive, and any combination of them may result in significant TNR expansion and a disease phenotype in higher organisms.
Much attention has been focused on aberrations in DNA replication, a major source of TNR instability in proliferating cells. Chromosomal replication faces many obstacles during replication fork progression that could destabilize the genome and prove fatal if not overcome. One such obstacle stems from DNA secondary structure formation during replication. Studies have shown that TNRs associated with human diseases form stable secondary structures in vivo. Using an assay for hairpin formation, Moore et al. (3) showed that CAG, CTG, CCG, or CGG heteroduplex loops are inefficiently repaired during meiotic recombination and were copied during the next round of DNA replication. In contrast, AAG/CTT or CAA/TTG loops did not form hydrogen-bonded structures and were therefore

FEN-1 in DNA Secondary Structure Resolution
removed (3). GAA/TTC repeats that cause Friedreich ataxia have also been shown to be destabilized during DNA replication (18 -21). However, non-structure-forming GTT repeats have a 1,000-fold lower expansion rate relative to hairpin forming CNG repeats (6). This suggests that secondary structures can form in vivo and impair replication/repair enzymes. Studies in model organisms, such as S. cerevisiae, have shown that mutations in DNA replication enzymes, particularly Rad27 (yeast FEN-1 homolog), lead to TNR instability (6,(22)(23)(24)(25)(26)(27)(28). Rad27/FEN-1 is a multifunctional nuclease that plays a critical role in maintaining genome stability through RNA primer removal and long patch base excision repair. The role of FEN-1 in replication is well studied. During Okazaki fragment maturation, a 5Ј-flap is generated by polymerase ␦ (pol ␦) stranddisplacement synthesis. FEN-1 recognizes and cleaves the 5Ј-flap to create a nick, which is then ligated. FEN-1 interacts and works in a coordinated manner with proliferating cell nuclear antigen, pol ␦, replication protein A, DNA ligase I, and DNA2 for efficient Okazaki fragment processing. In addition to its 5Ј-flap endonuclease activity (FEN), FEN-1 is also known as an obligate double-stranded DNA 5Ј to 3Ј exonuclease (EXO) that cleaves nick, gap, and 5Ј-recessed DNA as well as blunt ended DNA to a lesser extent (41,42). Recently, FEN-1 was also shown to possess a gap endonuclease (GEN) activity (29). The GEN activity of FEN-1 cleaves the template strand of a gapped DNA fork and bubble substrates that mimic a stalled DNA replication fork (29,30). All three activities are reported to be important for the function of FEN-1 in replication, recombination, repair, and probably apoptosis (29,30,34,41,43,44). It has been proposed that FEN activity can only occur when FEN-1 can track along a single-stranded DNA flap but not doublestranded DNA to reach the cleavage site (31,32). If a hairpin or loop structure is formed by the displaced ␣-segment (the RNA primer and DNA segment that is synthesized by polymerase ␣), the FEN activity is inhibited. If that is the case in vivo, then the presence or absence of the Rad27 nuclease in yeast should make no difference in the rate of expansion during replication of the repeated sequence, such as (CTG) n . However, the expansion of (CTG) n in rad27 null mutants was 180-fold higher than in wild type yeast cells in the assay system employed by Spiro et al. (6). The results illustrate the dynamics of the hairpin structure formation and more importantly demonstrate that Rad27 nuclease plays a role in resolving the intrinsic secondary structure of the displaced ␣-segment of Okazaki fragments. We therefore hypothesize that Rad27 may use an alternative pathway to resolve the ␣-segment when a stable secondary structure is built in. We propose that the concerted action of EXO and GEN activities of Rad27/FEN-1 can resolve various secondary structures in eukaryotic cells.
By screening a pool of more than 100 point mutations of FEN-1 (30), a human FEN-1 allele, FEN-1-E178A was characterized to abolish over 95% of the GEN activity and some of its EXO activity while it retained most of the FEN activity on 5Ј double flap substrates. The substitution of glutamate at position 178 with alanine may cause a slight protein conformational change that alters substrate binding affinity, thereby biasing its nuclease activity profile. The E178A mutation is an ideal point mutation to study the effect of the loss of GEN and EXO activities on TNR expansion. The particular residue Glu 178 is highly conserved across different species and lies near the active center important for catalysis. Since hFEN-1 and Rad27 share considerable sequence homology and are functionally conserved, we knocked the analogous point mutation E176A into the yeast genome. The point mutation led to considerable TNR instability and fragility but showed no mutator phenotype or sensitivity to UV or DNA-alkylating agents. We also purified the mutant and wild type yeast proteins to correlate the phenotype with the loss of each biochemical activity. Our study suggests that FEN activity is important for flap processing, whereas EXO and GEN activities may contribute to the resolution of the structured flap during Okazaki fragment maturation.

EXPERIMENTAL PROCEDURES
Protein Expression and Purification-The construction of the protein expression vector encoding the His 6 -tagged wild type Saccharomyces cerevisiae Rad27 is as previously described (18). The plasmid for expression of the His 6 -tagged Rad27-E176A mutant protein was generated by the QuikChange sitedirected mutagenesis protocol (Stratagene, La Jolla, CA) using primers E176AF (TATGCCGCAGCAAGTGCAGATATG-GACACACTC) and E176AR (GAGTGTGTCCATATCTG-CACTTGCTGCGGCATA) (the substituted codon is underlined). The pET28b vector containing Rad27 and the mutant gene was transformed into E. coli Rosetta cells (Stratagene, La Jolla, CA) for overexpression.
The protein expression was performed as previously described, and all purification steps were carried out at 4 C (18). To purify His 6 -tagged Rad27 or E176A, the harvested cells were lysed in 50 ml of Bugbuster protein extraction reagent (Novagen, San Diego, CA) containing two tablets of complete EDTA free protease inhibitor mixture (Roche Applied Science), 50 l of benzonase nuclease (Novagen, San Diego, CA), 5 mg of lysozyme, and 5 mM ␤-mercaptoethanol. The cell lysate was subsequently spun at 13,000 rpm for 15 min. An aliquot of 5 M NaCl was added to the cleared lysate to bring the final concentration of NaCl to 1 M and incubated on ice for 15-20 min. The lysate was again spun to obtain a clear supernatant containing recombinant Rad27 or E176A, which was then purified using a HisTrap TM FF column (Amersham Biosciences) according to a previously published protocol (18). The fraction of interest was dialyzed overnight into buffer A (30 mM HEPES, pH 7.8, 0.5% inositol, 0.25 mM EDTA, 0.01% Nonidet P-40, 1 mM dithiothreitol, and 30 mM KCl). The sample was then applied to a SP Sepharose column equilibrated with buffer B (50 mM MES, pH 6.0, with 0.02% sodium azide and 1 mM dithiothreitol). His 6tagged Rad27 or E176A was then eluted with buffer B using a linear gradient of NaCl (0 -1 M). The fraction of interest was dialyzed overnight into buffer C (20 mM Tris, pH 7.5, 15% glycerol, 200 mM NaCl, and 2 mM ␤-mercaptoethanol). The protein was quantified by Bradford Protein Assay kit (Bio-Rad) using the microtiter plate assay and IgG as the protein standard.
Nuclease Assays-32 P-Labeled DNA substrates were prepared according to a previously described protocol (34). Oligo-nucleotides used for substrate preparation are specified in each experiment, and corresponding sequences are summarized in Table 1. The FEN, GEN, and EXO nuclease activities of Rad27 and E176A were assayed as previously described (30). Briefly, an indicated concentration of Rad27 or E176A was incubated with FEN, GEN, or EXO DNA substrates, respectively, in 50 mM Tris-Cl (pH 8.0), 5 mM Mg 2ϩ , and 1 mM dithiothreitol. Reactions were carried out at 37°C for 30 min if not specified. DNA substrates and cleavage products were resolved by DNA-sequencing PAGE, visualized with a PhosphorImager, and semiquantified with the ImageQuant software (GE Healthcare, Piscataway, NJ). The percentage of products was calculated by dividing the intensity of product bands by the total intensity of substrate and product bands.

FEN-1 in DNA Secondary Structure Resolution
Mutator and Sensitivity Assays in Yeast-Mutation rates for the accumulation of Can r mutants and Hom ϩ , Lys ϩ revertants were determined by fluctuation analyses by the method of the median (38) using 10 independent cultures per experiment as previously described (39). All rates are the average of two or three independent experiments. Mutation spectra were deter-mined by DNA sequence analysis as previously described (39). UV and methyl methanesulfonate sensitivity assays were conducted as described by Reagan et al. (40).

Rad27-E176A Is Deficient in EXO and GEN Activities-
FEN-1 possesses three activities that vary in efficacy, with FEN activity being the strongest and GEN activity being the weakest. As a result, increasing amounts of the proteins have to be used to visualize these activities in vitro. For biochemical characterization of Rad27-E176A and Rad27 proteins, the wild type and mutant proteins were overexpressed in E. coli cells and purified to Ͼ95% purity. Various substrates (i.e. double flap, nick, and gap) were tested. A double-flap substrate with a 1-nt 3Ј-flap is the optimal substrate for FEN-1 where the enzyme cuts at the base of the flap using its FEN activity (45). Two sets of experiments (concentration-dependent and time course) were done

FEN-1 in DNA Secondary Structure Resolution
to quantify the three activities. The results showed that the mutant E176A lost ϳ40% of EXO (Figs. 1B and 2B) and ϳ95% of GEN activity (Figs. 1C and 2C), whereas the FEN-1 activity was almost the same as wild type Rad27 protein (Figs. 1A and 2A).
The preliminary tests indicate that the mutant has intact FEN activity. To test if the flap length has any effect on the FEN activity of WT and mutant proteins, substrates with flaps of different lengths (5, 20, and 40 nt) were tested. In the presence of a 3Ј-flap, the activity of the wild type protein is similar on flaps of all three lengths (5, 20, and 40 nt). The mutant protein has no significant difference in the activity as compared with the wild type protein, suggesting that E176A has almost intact FEN activity (Fig. 3).
Mutant Enzyme E176A Cannot Process (CTG) 20 Hairpin Structures as Efficiently as Wild Type Protein-We further tested the defect in EXO and GEN activities of Rad27-E176A with various probable TNR intermediate substrates (Fig. 4). Repeats that form DNA secondary structures are more likely to undergo TNR expansion (16), because these structures inhibit recognition and processing by the repair machinery (46,47 20 has been demonstrated by NMR to have stable secondary structure in solution (6).
Here we employ the same (CTG) 20 hairpin substrate labeled at 5Ј-and 3Ј-ends to compare the cleavage efficiencies of WT and mutant proteins. To test the activity on hairpin substrates in our experiments, 400 fmol of protein was used. This substrate is an intermediate for TNR expansion that, if not resolved, can integrate into the genome. On this substrate, EXO activity cleaves nucleotides from the 5Ј-end of the hairpin (Fig. 4A), creating a substrate for gap-dependent endonuclease activity. When the substrate is labeled at the 5Ј-end, only EXO activity can be visualized on the gel. A 100% cleavage is obtained at the 20 min time point in the wild type reaction, whereas only 60% of the product is cleaved for the E176A protein at the end of the 60-min reaction (Fig. 4B). The deficiency is more apparent when the hairpin substrate is labeled at the 3Ј-end, where the final products resulting from both EXO and GEN activities can be observed (Fig. 4C). When WT protein is used, the 5Ј to 3Ј processive EXO cleavage products (10 -14 nt) accumulate faster, as compared with E176A, where, due to deficient GEN and EXO cleavage, higher molecular weight products are observed (16 nt). The activity of the mutant protein on hairpin substrate demonstrates that, as a result of deficient EXO and GEN activities, the mutant protein is unable to resolve the complete fold-back/hairpin substrate with high efficiency, probably resulting in an increased half-life of the structured flap in the reaction and increasing its chances of integrating into the genome.
FEN-1 removes the 5Ј-flap generated due to pol ␦ strand displacement synthesis using its 5Ј FEN activity. If the displaced flap contains trinucleotide repeats, the flap might fold back to form an incomplete and later a complete hairpin structure. These fold-back substrates are inhibitory to FEN activity, and a high concentration of FEN-1 is needed to resolve these structures in vitro (48,49). These unresolved structures could also form internal loop or bubble structures, which can integrate into the genome, leading to duplications or repeat expansions.
To test whether the GEN activity of FEN-1 plays a role in preventing repeat expansions, two substrates representing possible intermediates in repeat expansions were tested (Fig. 5). The GEN efficiency of the WT and Rad27-E176A was first tested on a double-stranded DNA substrate with a 4-nt gap. The template strand was labeled at the 3Ј-end (Fig. 5A). The GEN activity was observed on the template strand at the 3Ј-end of the gap. The mutant E176A retains almost negligible amounts of the GEN activity. GEN activity was also assayed on incomplete fold-back and internal loop substrates. The incomplete fold-back flap substrate with a 2-nt gap at the 5Ј-end is labeled at the 3Ј-end. As seen in Fig. 5B, E176A cleaves the substrate but with less efficiency. In addition, the cleavage sites are restricted to the single-stranded DNA/double-stranded DNA junctions, and no cleavage was detected at the 5Ј-end of the hairpin flap, ruling out the possibility that FEN-1 uses its 5Ј to 3Ј exonuclease activity to digest the 5Ј-end. In the case of the internal loop substrate, the GEN activity cleaves at the 3Ј singlestranded DNA/double-stranded DNA junction, followed by processive 5Ј to 3Ј exonuclease cleavage (Fig. 5C). It can be concluded that the Rad27 mutant E176A is deficient in GEN activity and thus processes the repeat expansion intermediates with less efficiency.
To test if secondary structure rather than the repeat sequence inhibits the FEN activity of FEN-1, (GTT) 20 flap substrate was synthesized (6). These GTT repeats form no hairpins (7). As shown in Fig. 6, A and B, the (GTT) 20 flap substrate was labeled at both the 5Ј and 3Ј-ends, respectively. When labeled at the 5Ј-end, the FEN activity cleaves at the base of the flap, resulting in a 60-nt product. Both the wild type and the mutant proteins process the long flap with almost equal efficiency, underscoring the fact that the mutant protein Rad27-E176A has FEN activity comparable with that of the wild type protein, and the non-structure-forming GTT repeats are efficiently processed by both of the proteins. Therefore, secondary structure rather than the repeat sequence underlies the inhibition of FEN-1. When labeled at the 3Ј-end, the products resulting from FEN followed by processive EXO activity can be observed (Fig.  6B). When wild type protein was used, the FEN cleavage products (17,16, and 15 nt) can be observed, followed by EXO cleavage products (10 -14 nt), which increase in intensity as the reaction progresses. When E176A is used, only FEN cleavage products are observed, along with a low fraction of EXO cleavage products. The results suggest that on a GTT flap substrate, both of the enzymes use the predominant FEN activity for processing the long flap, which is in contrast to the (CTG) 20 hairpin flap, where the enzyme uses EXO and GEN activities to resolve the secondary structure.
Concerted Action of EXO and GEN Activities Resolves (GAA) n Flaps, Which Equilibrate to Form Secondary Structures-Ruggiero and Topal (50) showed that short (GAA) n repeat-containing flaps can rearrange to form stable internal loops that are resistant to FEN activity but can be resolved by using EXO activity. The same nine-repeat template U a (TTC) 9 U b was annealed to oligonucleotides with 0, 10, 12, and 15 GAA repeats (50) to generate substrates with 0, 1, 3, and 6 extra GAA repeats, respectively, as shown in Figs. 7 and 8. The substrates were labeled both at the 5Ј and 3Ј-ends. When wild type protein was added to the repeat substrate containing a 5Ј-recessed end, (Fig.  7A), only EXO cleavage products were observed (1, 2, and 3 nt). The same products were also obtained with Rad27-E176A, but the cleavage was less efficient. On a substrate with one extra GAA repeat (Fig. 7B), both EXO and FEN cleavage products were obtained (2, 3, and 4 nt), suggesting that the substrate equilibrates between 5Ј-flaps and 5Ј-recessed ends. The substrates containing three and six extra GAA repeats (Fig. 7, C and  D) were more sensitive to the EXO cleavage than FEN activity. Small quantities of 4-, 7-, and 10-nt products were also obtained with the substrate with three extra GAA repeats, whereas 21and 45-nt products were also observed for substrates with six extra GAA repeats, indicative of FEN and GEN cleavages, respectively. When the substrates were labeled at the 3Ј-end (Fig. 8), only 10-, 11-, and 12-nt products were obtained, which are indicative of GEN cleavage at the single-stranded DNA/double-stranded DNA junction on the 3Ј-end of the loop, followed by EXO cleavage. The mutant protein showed almost negligible GEN cleavage. This is suggestive of the fact that, irrespective of the number of extra GAAs present in the substrate, the GAA repeats rearranged to form predominantly internal loops, which are more amenable to EXO and GEN cleavage but resistant to FEN cleavage. This result further strengthens the hypothesis that FEN-1 uses EXO and GEN activities rather than FEN activity to resolve structured flaps formed due to TNR expansion.
Defective GEN and EXO Cleavage Increases TNR Instability and Fragility-To determine if loss of EXO and GEN activities in the Rad27 mutant E176A affects CTG tract stability, a YACbased assay with various CTG tract lengths was used (25,51,52). CTG tracts are fragile sites in yeast and have been shown to cause an increase in chromosome breakage by both genetic and physical assays (25). In this assay, CTG tracts of various lengths are placed on the right arm of a YAC, proximal to a URA3 gene. Breakage followed by telomere addition leads to loss of the URA3 gene and generates Ura3 Ϫ and FOA R cells. Thus, measurement of the rate of generation of FOA R cells can be used as a monitor of the rate of breakage/fragility. Along with the assay for fragility, stability of the CTG repeat on the YAC can be monitored by a PCR assay, using primers on either side of the repeat to amplify the tract and monitor changes in size (51). The yeast strains with CTG 0, 85, and 155 were kar crossed into the canavinine-resistant WT RDKY2672, rad27⌬, and E176A mutant backgrounds.
For the stability assay, colony PCR was used to identify starting colonies with correct tract lengths, CTG-85 and CTG-155. The colonies were allowed to grow in liquid cultures for expansion or contraction to occur. The cultures were then plated, and

FEN-1 in DNA Secondary Structure Resolution
tract lengths of the resulting colonies were determined by PCR. In the wild type strain, the percentage of CTG tract expansion was very low (0.8% for CTG-85 and 1.3% for CTG-155) as compared with contractions, which were much higher than expansions and increased from a tract length of 85 to 155 (5.3% for CTG-85 and 24.1% for CTG-155) ( Table 2). This result is consistent with the previous reports that show a bias toward contraction in yeast cells (15,53). As has been reported earlier, in the rad27⌬ strain, the rate of both expansion and contraction were higher as compared with the wild type, with a strong bias toward expansion (6,25,51). The Rad27-E176A strain showed a significant increase in expansion (11.2% for CTG-85 and 13.8% for CTG-155) with a frequency of expansion (15.6%) almost identical to that of rad27⌬ for CTG-155 ( Table 2). The results suggest that defects in EXO and GEN activity caused by the E176A mutation and its inability to process structured flaps increased the percentage of expansion for both CTG-85 and CTG-155 tracts in vivo. Thus, the instability phenotype can be attributed to the loss of these activities.
The YAC-based fragility assay is based on the fact that CTG repeats cause chromosomal breakage. A Ura3 gene is in close proximity of the CTG tracts in the strains used for this study (51). Instability and breakage events due to the presence of CTG repeats cause the loss of the Ura3 marker, resulting in colonies resistant to 5-fluoroorotic acid. The rate of the generation of FOA R colonies can be considered a determinant of breakage events. YACs bearing CTG 0, 85, and 155 in the WT and mutant backgrounds were assayed for these breakage events. The rad27⌬ strains showed a significant increase in the rate of FOA R cell generation over wild type: 3-fold increase as compared with wild type for CTG-85 and 9-fold as compared with wild type for CTG-155 (Table 3). In all three strains, the rate of FOA R cell generation increased with increasing tract length, which is consistent with the fact that increasing CTG repeats will make the tracts prone to breakage. The Rad27-E176A mutation also increased the rate of FOA R cell generation compared with wild type in a CTG length-specific manner ( Table 3). The fragility of the YACs with CTG-85 is almost similar to rad27⌬ strain, indicating that this mutant is defective in processing CTGcontaining flaps, resulting in their expansion and thereby breakage. However, the rate of breakage did not increase significantly for CTG-155 repeats, suggesting a difference in the fate of CTG-containing flaps in the two strains (52).
The mutant E176A has almost intact FEN activity, suggesting that it has an intact RNA primer removal function in DNA replication (i.e. it can still process unstructured flaps). In order to test if E176A showed an induced or spontaneous mutator phenotype, the WT, rad27⌬ and E176A strains were subjected to treatment with methylmethane sulfonate and UV irradiation. The mutant E176A is resistant to both DNA-damaging agents, indicating that it has an intact DNA repair function, as compared with rad27⌬, which shows sensitivity to both of them (40). E176A strain also did not show any spontaneous mutations when subjected to hom3-10, lys2-Bgl, and canavinine assays, whereas rad27⌬ strain showed a high rate of spontaneous mutations, as previously reported (23) ( Table 4).

DISCUSSION
Numerous studies indicate that a free 5Ј-end is required before FEN-1 can recognize and cleave at the base of the flap, and if the 5Ј-end of the flap is blocked, FEN-1 activity is inhibited (48,54). Taking these studies into account, Liu et al. (52) FEBRUARY 9, 2007 • VOLUME 282 • NUMBER 6

FEN-1 in DNA Secondary Structure Resolution
proposed a model by which FEN-1 removes triplet repeats by a unique tracking mechanism. They proposed that the presence of trinucleotide repeats on the displaced flap results in the formation of fold-backs and bubbles, which are inhibitory to FEN-1 cleavage. These structures can re-equilibrate to form long 3Ј-flaps and short 5Ј-flaps. FEN-1 loads onto the 5Ј-flap, but cleavage is inhibited due to the presence of a longer 3Ј-flap. FEN-1 tracks down the 5Ј-flap, facilitating further equilibration until a suitable substrate with a 1-nt 3Ј-flap and a longer 5Ј-flap is formed. FEN-1 subsequently removes the 5Ј triplet repeatcontaining DNA using its endonucleolytic activity. Formation of hairpin structures is a dynamic process in vivo. FEN activity will certainly act on equilibrating flaps. However, specific elimination of EXO and GEN rather than FEN activity had a major impact on the stability of trinucleotide repeats. The purified yeast E176A protein was found to have defects in exonuclease and gap endonuclease activities but retained most of the flap endonuclease activity. On the nick substrates, the mutant protein showed a severe exonuclease defect. When a substrate with hairpin flaps was used, Rad27 alone was able to cleave them, albeit weakly. The 5Ј-exonuclease activity cleaved the hairpin substrate, resulting in low molecular weight cleavage products, whereas the mutant protein showed only 9% cleavage. This result demonstrates that FEN-1 can resolve hairpin flaps using its 5Ј-to 3Ј-exonuclease activity. When the same substrate was labeled at the 3Ј-end, the exonuclease activity was again evident, resulting in a laddering pattern of the products. The removal of nucleotides from the 5Ј-end generates a gap, creating an ideal substrate for GEN activity. In Fig. 4C, GEN activity results in 16-and 17-nt products, showing that the hairpin flap has been removed. The lower molecular weight bands are due to processive 5Ј-to 3Ј-exonuclease activity on the nick substrate generated. The mutant protein also showed a defect in GEN activity (Fig. 5). E176A showed almost negligible GEN activity on a 4-nt gap substrate. Unprocessed long flaps can fold back to form incomplete hairpin or loop/bubble intermediates. In order to assess the GEN activity, a hairpin flap and internal loop substrates were used (34). On both of the substrates, cleavage occurs at the single-stranded DNA/ double-stranded DNA junctions. E176A was found to be deficient on both of these repeat expansion intermediates. On the contrary, when non-structure-forming TNR repeats were used as flap substrates, only FEN activity was used to process the long unstructured flaps by both the wild type and mutant enzymes (Fig. 6A), further confirming the hypothesis that EXO and GEN activities and not FEN are used to resolve secondary structures. This observation is more pronounced when (GAA) n substrates (50) are used (Figs. 7 and 8). The GAA repeats in the flaps equilibrate to form loops, which are processed by EXO and GEN activities but are resistant to FEN cleavage.
Deletion of Rad27 in yeast results in a dramatic effect on TNR stability (6). The defect in EXO and GEN activities caused by the Rad27-E176A mutation, which was precisely knocked into the rad27 locus, results in increased percentage expansion for both CTG-85 and CTG-155 repeat tracts in vivo. We hypothesize that the presence of secondary structures on the unprocessed flap and deficiency in EXO and GEN activities add up to show this phenotype.
In view of these results, a new model can be proposed for the role of FEN-1 in inhibiting TNR expansions (Fig. 9). Unstructured long flaps with random sequences and TNR can be processed by endonucleolytic cleavage of FEN-1 (Fig. 9, 1 and 2). However, in the presence of trinucleotide repeats, which can base-pair and fold back to form hairpin structures (3) or loops (4), FEN-1 shows a weak endonucleolytic cleavage. This increases the half-life of the flap, which can form an internal loop and ligate into the genome, resulting in TNR expansion. However, if the EXO and GEN activities of FEN-1 are intact, nucleotides can be removed from the 5Ј-end (5), resulting in a gap or single-stranded DNA/double-stranded DNA junction, on which GEN cleavage can occur (6) and (7). The concerted action of EXO and GEN activities of FEN-1 may result in the resolution of flaps with secondary structures. In the presence of wild type Rad27/FEN-1, the 5Ј-flap with random sequences (Rnd) displaced due to pol ␦ synthesis is recognized and cleaved (1). The presence of TNRs on the flap results in longer or structured flaps. If the flaps contain a TNR-like (GTT) n , which results in longer flaps, Rad27/FEN-1 can still resolve it, probably with the help of DNA2 helicase/nuclease (2). However, if the displaced flaps contain (CTG) n or (GAA) n repeats, they might fold back to form hairpins or loops, which are inhibitory to FEN cleavage (3) and (4). Two possible pathways and their consequence for repeat expansions are shown. If the EXO activity of FEN-1 is intact, it removes nucleotides from the 5Ј-end of the hairpin, thereby creating a gap (5), which is an ideal substrate for GEN activity of FEN-1 (6). In case of a loop, both EXO and GEN activities resolve the secondary structure (7). However, if both EXO and GEN activities are defective, the hairpin flap and loop substrate remains unresolved and ligates into the genome, resulting in TNR expansion.