Nuclease-Deficient FEN-1 Blocks Rad51/BRCA1-Mediated Repair and Causes Trinucleotide Repeat Instability

Abstract

Previous studies have shown that expansion-prone repeats form structures that inhibit human flap endonuclease (FEN-1). We report here that faulty processing by FEN-1 initiates repeat instability in mammalian cells. Disease-length CAG tracts in Huntington's disease mice heterozygous for FEN-1 display a tendency toward expansions over contractions during intergenerational inheritance compared to those in homozygous wild-type mice. Further, with regard to human cells expressing a nuclease-defective FEN-1, we provide direct evidence that an unprocessed FEN-1 substrate is a precursor to instability. In cells with no endogenous defects in DNA repair, exogenous nuclease-defective FEN-1 causes repeat instability and aberrant DNA repair. Inefficient flap processing blocks the formation of Rad51/BRCA1 complexes but invokes repair by other pathways.

Trinucleotide repeats give rise to disease when the number of repeated units of a simple trinucleotide, such as CAG, expands beyond an unaffected range (43). In the unaffected range, the number of repeated trinucleotide units, though polymorphic, is stable. An increase in the number of repeated units beyond a certain threshold, however, causes instability as well as disease, and expansion in subsequent generations becomes likely (21, 43). While the molecular events that give rise to repeat instability remain obscure, studies during the last several years have given sufficient insight to allow the development of working models for expansion. Emerging data suggest that expansion arises in the process of break repair (15, 35, 50, 51) and is mediated by aberrant interaction of key DNA repair enzymes. Important among these is the flap endonuclease (FEN-1).

FEN-1 is a structure-specific nuclease that participates in DNA replication and repair (39). Flap endonuclease (human FEN-1 and/or Saccharomyces cerevisiae Rad27 protein) is required for normal maturation of Okazaki fragments during replication (31, 39, 49, 68). FEN-1 has also been implicated in several repair processes, including base excision repair (BER) (13, 32, 33, 47, 49, 56, 64), nonhomologous end joining (70), recombination (46), nucleotide excision repair (NER) (57), and removal of bulky UV lesions by an alternative excision pathway (72). The details of the steps by which flap endonuclease protects cells from repeat instability or causes expansion remain unclear. However, yeast lacking Rad27 shows both an increased rate of expansion of trinucleotide repeats (15, 52, 60) and duplications flanked by direct repeats (29, 62, 71). These data have been used to support the notion that defective flap processing plays an important causative role in expansion. The failure of FEN-1 to process flaps folded into aberrant hairpin structures is thought to cause expansion at CAG repeats (26, 60). FEN-1 is inhibited by hairpins comprising CAG triplets, because the 5′ end needed for FEN-1 loading is concealed within the hairpin structure (26, 60).

Despite the compelling data that loss of yeast Rad27 causes expansion, the relevance to human instability is uncertain. The human FEN-1 and yeast Rad27 protein are homologous, and the human enzyme can complement the defect of yeast null for Rad27 protein (22, 23). Yet the features of expansion induced by loss of Rad27 in yeast are very different from those observed in human disease.

In yeast, expansions are rare compared to deletions, while in human diseases, expansions exceed contractions (16, 42-44, 52). Deletion of Rad27 from yeast causes a severe growth defect at 37°C, instability throughout the genome (mutator phenotype), and increased recombination (30, 45, 61, 62). In mammals, however, loss of FEN-1 is embryonically lethal (37). In human expansion disease, there is neither exchange nor alterations of flanking sequence around the CAG expanded repeat region, suggesting that expansion is not associated with increased homologous recombination. Further, in human repeat expansion, the mutation is limited to a single site (20), inconsistent with a mutator phenotype as occurs with the loss of flap endonuclease in yeast (30, 51). Expansion in human disease must, therefore, occur in the presence of a normal FEN-1 protein. The majority of expansion studies of yeast have relied on models from which the flap endonuclease (Rad27 gene product) is absent. Taken together, the features of rad27 null mutants in yeast do not recapitulate many of the features of expansion in mammalian disease. This raises a question as to the role of FEN-1 in expansion in mammals.

We report here studies designed to improve understanding of the role of FEN-1 in repeat instability in mammals. We have crossed a transgenic mouse lacking one functional allele for FEN-1 (37) with transgenic animals harboring an expanded CAG repeat within the human Huntington's disease (hHD) gene (40). Mice heterozygous for FEN-1 develop and reproduce normally, but they have a rapid tumor progression phenotype (37). Experiments using these animals, therefore, have the advantage that expansion in the hHD allele can be evaluated with a reduced level of normal FEN-1 rather than in its absence. Additionally, we evaluated trinucleotide instability of endogenous repeats in human cells stably transformed with a nuclease-defective FEN-1 that retains its ability to support protein-protein interactions.

FEN-1 protects mammalian cells from repeat instability. Expression of nuclease-defective enzyme destabilizes microsatellites. In addition, cells expressing nuclease-defective FEN-1 show aberrant foci containing proteins of break repair and NER pathways.

MATERIALS AND METHODS

Transgenic and heterozygous knockout mice.

The transgenic mouse R6/1 has exon 1 of an hHD allele containing about 120 CAG repeats, as described by Mangiarini et al. (40). Heterozygous knockout FEN-1 mice were from R. Kucherlapati (Harvard Medical School) (37); FEN-1 mice contain one wild-type (wt) and one nonfunctional flap endonuclease allele (37). To produce progeny that contained one hHD allele and were heterozygous for FEN-1, male R6/1 mice were bred with female heterozygous FEN-1 mice as described previously (36). A duplex PCR was used to determine the FEN-1 genotype. Primer A (5′ GGGAGTGAGATGGCAGTGTT, corresponding to a sequence within the FEN-1 coding region, nucleotides [nt] 5539 to 5558 in GenBank AY014962) and primer C (5′ GGCACTCAGGGTGTTTTCAA, corresponding to a sequence in the 3′ flanking region, nt 5832 to 5813) give a 293-bp product from the wt allele; primer B (5′ TGGAAGGATRTGGAGCTACGGC, which corresponds to a sequence within the targeting vector) and primer C give a 373-bp product from the interrupted allele. The presence of the hHD allele and the length of the allele were determined using primers described previously (40). GraphPad Prism was used for statistical calculations.

Plasmids, cell cultures, and transformation.

HCC1937, a BRCA1 mutated breast cancer cell line (65), and the HCC1937 + 5′ mycBRCA1-revertant line (7, 53, 65) were from J. Chen (Mayo Clinic). HCC1937 and SK-N-MC cells (American Type Culture Collection) were maintained in RPMI 1640-10% fetal calf serum. wt FEN-1 cDNA with a hemagglutinin [HA] epitope tag encoded at the 3′ end and D181A mutant FEN-1 cDNA were gifts of M. S. Park (Los Alamos National Laboratory). The coding region for nuclease-deficient FEN-1 D181A contains a C-for-A substitution at nt 542, which changes codon 181 from GAC (aspartate) in the wt to GCC (alanine) (54-56) (see Fig. 2.) A 625-bp Bpu1102I fragment (nt 433 to 1058) within the FEN-1 D181A coding region was substituted for the Bpu1102I fragment in the wt FEN-1-HA cDNA to create the D181A-HA cDNA. The wt and D181A coding regions were inserted into the vector pcDNA3.1/Hygro (Invitrogen), and plasmids were sequenced to confirm that they differed at the single nucleotide only. Cells were transformed by plasmid DNA-Lipofectamine (Invitrogen). At 24 h after transformation, growth medium was replaced by growth medium supplemented with 100 to 150 μg of hygromycin/ml, a concentration toxic to the SK-N-MC cells. Colonies surviving in a single well were pooled. Using Lipofectamine 2000 (Invitrogen) for transient assays, SK-N-MC cells were cotransformed with the appropriate FEN-1 construct and pEGFP-N1, a green fluorescent protein (GFP) expression vector (Clontech).

FIG. 2.

Stable transformation of SK-N-MC human neuroblastoma cells with wt or nuclease-defective FEN-1. (A) A single amino acid change from aspartate to alanine at amino acid 181 (D181A) in the nuclease domain of FEN-1 renders enzymes nuclease defective but does not alter binding properties (54, 55). (B) Expression of exogenous (HA-tagged) FEN-1 in SK-N-MC cells. S, parental SK-N-MC. Numbers indicate individual isolated colonies. Cell extracts were analyzed by immunoblotting. Proteins separated by sodium dodecyl sulfate gel electrophoresis were transferred to nitrocellulose and probed with anti-HA antibody. (C) Binding and cleavage by wt FEN-1. The structure-specific nuclease clips displaced strands at the three-way junction, leaving a substrate for ligation. (D) Mutant FEN-D181A binds normally but does not cleave, so that even in the presence of wt enzyme some flaps may not be cleaved. Uncleaved flaps may be ligated, leading to expansion (bottom left panel), or uncleaved flaps may trigger strand break and/or subsequent repair (bottom right panel) (50). (E) FEN-1 activity in nuclear extracts from transformed cells. The 5′ ³²P-labeled flap substrate (top panel) was incubated with extract prepared from stably transformed cells or from SK-N-MC parent cells and resolved from products on denaturing gels. Numbers indicate individual isolated cell lines. Products were 19 and 21 bp (arrows). −, BSA replaces nuclear extract. (F) Survival of SK-N-MC cells cotransformed with FEN and GFP expression plasmids. A plot of average numbers per field ± SEM of fluorescent cells is shown. Green cells were counted in 10 fields. Tests of averages indicate a statistically significant difference between D181A-transformed and parental SK-N-MC at each time point (P < 0.002).

Assay for FEN-1 protein.

Expression of exogenous FEN-1 was determined by immunoblotting using a rabbit anti-HA antibody (Zymed). Cells were harvested into sodium dodecyl sulfate sample buffer, cellular proteins were resolved by 12% polyacrylamide gel electrophoresis, and proteins were transferred to nitrocellulose by semidry blotting as described previously (58). After incubating membranes with anti-HA, HA-containing proteins were visualized using goat anti-rabbit horseradish peroxidase-conjugated antibody and West Pico chemiluminescent substrate (Pierce). The calculated molecular mass of the FEN wt-HA fusion protein was 44,164 Da.

FEN-1 nuclease activity.

For analysis of FEN-1 activity, nuclear extract from FEN-1-transformed and parent SK-N-MC cells was prepared using the method of Andrews and Faller (2), diluted to a final NaCl concentration of 200 mM, and assayed (using the modified Bradford method [Bio-Rad]) for total protein. Flap-1 substrate was used to measure FEN-1 activity in a modification of the previously described assay (24, 60). The reaction mixture included 50 mM Tris-Cl (pH 8), 10 mM MgCl₂, 0.5 mM β-mercaptoethanol, 100 μg of bovine serum albumin (BSA)/ml, 33 μg of sonicated salmon sperm DNA/ml, 80 fmol of 5′-end ³²P-labeled substrate, and nuclear extract. The substrate and cleaved product were resolved on 6% sequencing gels and analyzed by PhosphorImager (Molecular Dynamics).

Microsatellite instability assays.

SK-N-MC and FEN-transformed cells were harvested, and DNA was purified as described previously (59). Repetitive alleles were amplified using conditions that have been previously described and confirmed by independently repeating the assay (18, 20). For GeneScan (version 3.0; PE Applied Biosystems) analysis, PCR was carried out with a single fluorescein-labeled primer and the products were resolved by polyacrylamide gel electrophoresis. The products were identified by scans of the gel, so that peaks corresponded to DNA molecules. The prominent peak determines the repeat length of the allele. Using values from a GeneScan table, a fraction in a prominent band was calculated for each allele to compare samples. For example, the fraction under peak 191 equals the value for 191 divided by the sum of the values for 191 and 188 [see Fig. 3]). For the ACTC and HD alleles, instability was tested by amplification with alternative sets of primers. For ACTC, the first primer set was 51 to 70 with 140 to 121 (20); the second set was 51 to 70 with 152 to 133 (numbering according to GenBank accession no. HSAC06). For the HD allele, the first primer set was 337 to 366 with 485 to 465 (40); the second set was 308 to 332 with 485 to 465 (numbered according to GenBank accession no. HUMHDA).

FIG. 3.

Microsatellite instability is caused by expression of nuclease-deficient FEN-1. (A) GeneScan analysis of ACTC and HD alleles. Left panels, ACTC alleles amplified with primer set 1; middle panels, ACTC alleles with primer set 2; right panels, HD alleles with primer set 1; top scans, SK-N-MC cells; middle scans, FEN wt-transformed cells; bottom scans, D181A-transformed samples. Dashed lines indicate boundaries between two alleles, arrows indicate prominent peaks, and dots represent the secondary peak positions. Numbers indicate the length of the amplified DNA. (B to D) Fraction of area under prominent peak for the indicated allele. Values were from a GeneScan analysis table. Averages ± SEM are shown. A t test was used to determine whether averages were the same (GraphPad Prism). (B) For longer HD alleles, the fraction of the allele under the prominent peak was determined using primer set 1 (filled symbols) or primer set 2 (unfilled symbols). (C) For ACTC alleles, the fraction of the allele under the prominent peak was determined using primer set 1 (filled symbols) or primer set 2 (unfilled symbols). (D) For shorter HD alleles, the fraction of the allele under the prominent peak was determined using primer set 1 (filled symbols) or primer set 2 (unfilled symbols).

Drug treatment and cell survival.

Approximately 300 to 600 cells were plated in series either in the wells of 6-well dishes or in 60-mm-diameter dishes. Cells were allowed to attach overnight. Methyl methanesulfonate (MMS) and cisplatin were diluted into phosphate-buffered saline (PBS), and the stocks were sterilized by filtration through a 0.2-μm-pore-size filter. Drugs were added to cultures from the concentrated stock. After a 1-h exposure to drugs, the medium was aspirated and replaced with complete growth medium lacking drugs. At 10 to 12 days after drug treatment, cells were fixed and stained in a solution of 2% methylene blue in 50% methanol. Colonies were counted, and the fraction of surviving cells was determined (colonies in treated dish divided by colonies in zero-drug control for that series). Data are shown for series that used the same drug dilutions. Data for cisplatin represent averages for six series. For MMS, data are averages for three series.

Immunofluorescence assay of nuclear foci.

Anti-FEN-1 (N-17), anti-ERCC1 (FL-297), anti-XPG (N-17), anti-p21 (C-19), anti-CBP (A-22), and anti-Rad51 (H-92) were from Santa Cruz Biotechnology. Anti-rabbit Cy5 was from Amersham. Anti-goat Cy5 was from Zymed. Cells were plated on coverslips in the wells of 6-well dishes 1 day before drug treatment. Cisplatin stock (2 mM) was prepared in PBS and added (15 μM final concentration) to cultures at the indicated time before time zero. At time zero, coverslips were washed with PBS, fixed in ice-cold methanol, washed, and then incubated sequentially in blocking buffer (Tris-buffered saline [TBS] containing 0.02% Tween 20 [Sigma], 5% skim milk [Difco], 5% BSA [Roche]) and in primary antibody diluted in blocking buffer. Cells were washed in TBS-0.02% Tween 20 and then incubated in secondary antibody conjugated to Cy5. The integrity of compartments and primary antibody-dependent staining was confirmed (see Results and Fig. 5). As a control for primary antibody specificity, mock experiments in which a primary antibody raised in a different species was used with the same secondary antibody were carried out in parallel (see Fig. 5). Coverslips were then washed in TBS-0.02% Tween 20, in TBS, and in water containing 0.2 μg of Hoechst 33258/ml to stain nuclei. To estimate the number of nuclear foci, coverslips were examined for Cy5 fluorescent dots within the area of Hoechst staining. Foci in 50 unselected nuclei were counted. The average number is reported (± the standard error of the mean [SEM]).

FIG. 5.

Defective flap processing alters recruitment of repair proteins to nuclear foci. (A to D) Foci are primary antibody specific. Cells were fixed and incubated with specific antibodies after treatment with cisplatin or PBS carrier (Materials and Methods). Cells were treated and fixed in parallel, differing only in the primary antibody used. (A) No primary antibody; (B) ERCC1; (C) CBP; (D) p21. Cells were washed with 0.2 μg of Hoechst 33258/ml to stain nuclei. (E and F) Foci reflect functional changes in cells. (E) BRCA1 in HCC1937 cells that lack BRCA1. (F) BRCA1 in HCC1937 + 5′mycBRCA1 cells that are transformed with functional BRCA1. (G) FEN-1-positive nuclear foci after cisplatin treatment (average ± SEM). Number of foci at time zero (shown in parentheses) is defined as 100%. Symbols are defined as for Fig. 4.

RESULTS

Instability at a disease-length allele in mice heterozygous for FEN-1.

The R6/1 mouse contains exon 1 of the hHD gene with an expanded CAG repeat (about 120 CAG [40]). The repeat is unstable during intergenerational inheritance and in somatic tissue (in an age-dependent manner) (35, 40, 41). To test the hypothesis that FEN-1 contributes to trinucleotide expansion in mammals, we crossed the FEN-1^+/− mice (37) with R6/1 mice (Fig. 1A) and evaluated the behavior of the expanded CAG trinucleotide tract in the FEN-1^+/−/hHD^+/− progeny (Fig. 1B; Table 1).

FIG. 1.

CAG repeats in the human HD allele are unstable in mice heterozygous or homozygous for wt FEN-1. (A) Breeding scheme of the hHD/Fen-1 mice, indicating patterns for F₁ and F₂ animals. (B) GeneScan of CAG repeats in an hHD allele from offspring of homozygous wt (left panel) and heterozygous FEN-1 knockout (right panel) males. The FEN-1 genotype of offspring is indicated to the right of each scan; the HD repeat length (midpoint of the distribution) in offspring is indicated to the left of each scan. (C) GeneScan of hHD allele from somatic tissues of homozygous wt (top panel) and heterozygous knockout (bottom panel) littermates. Tissues were harvested at 21 weeks.

TABLE 1.

Repeat length change in FEN-1^+/− mice

Genotype	Effect of genotype on repeat length change in offspring
	No. of mice	Mean change	Median change	No. showing contraction	No. showing expansion	No. showing no change
F1
+/+	18	−0.3 ± 0.6	0	5	6	7
+/−	16	−1.4 ± 1.2	0	7	3	6
Parentala
+/+	34	−0.8 ± 0.6	0	12	11	11
+/−	22	0.6 ± 0.2	0.5	2	11	9

^aFor the parental genotype, Mann-Whitney rank analysis of the individual repeat length changes indicates that the medians of the groups differ (U = 234; P = 0.02). Repeat length changes: −9, 0, −1, 0, 1, 1, −1, 1, 2, −2, 2, 0, 0, 0, 1, 0, 0, −1, 0, −1, 0, 1, 0, −1, −2, −1, 0, 2, 1, −19, 0, −2, 0 (+/+ fathers; sum of ranks = 829); 1, 2, 2, 1, 1, 1, 0, 0, 2, −1, 0, 0, 0, 0, 1, 0, −1, 2, 2, 1, 0, 0 (+/− fathers; sum of ranks = 767). χ² = 5.001 for the averages (P = 0.08). GraphPad Prism 3 was used for calculations (9).

DNA from tails of 3-week-old offspring was analyzed for CAG repeat length at the hHD allele (36). The repeat length variability characteristics of F₁ hHD progeny homozygous (wt) or heterozygous for FEN-1 did not differ (Table 1). However, we found that consistent with a role of FEN-1, CAG instability characteristics in F₁ progeny and F₂ progeny were different and depended on the genotype of the father (Table 1) (Fig. 1A and B). Fathers heterozygous for FEN-1 differed from fathers wt for FEN-1 in that wt fathers developed with their full complement of FEN-1 while heterozygous fathers developed with FEN-1 haploinsufficiency. Among offspring of males homozygous for FEN-1, we observed that about equal numbers of mice showed deletions, expansions, and no net change. In contrast, deletions were rare in offspring of males heterozygous for FEN-1 (Table 1). Thus, FEN-1 appeared to play a role in stability of CAG repeats in repetitive alleles by preventing deletions and modestly increasing expansions. Further, stability was a function of FEN-1 haploinsufficiency in the parent.

Previous studies with R6/1 mice have revealed that repeats appear to be stable in somatic tissues until 11 weeks but that expansion occurs with aging (35, 40, 41). The degree of age-dependent expansion varies with the tissue type (35, 40, 41). We therefore examined tissues of littermates at 21 weeks of age to determine whether the degree of somatic instability differs depending on FEN-1 genotype. However, we found no significant differences in CAG repeat length in the hHD allele among tissues from FEN-1^+/−/hHD^+/− and FEN-1^+/+/hHD^+/− littermates sacrificed at different ages. Therefore, CAG repeats in somatic cells did not show FEN-1 genotype-dependent changes (Fig. 1C). FEN-1 effects on instability in progeny were limited to the developing germ cells.

Analyses of mice, then, indicate that haploinsufficiency of FEN-1 can alter trinucleotide repeat length stability by inhibiting the frequency of deletion events and modestly increasing expansion events at long disease-length alleles in mammals. In mice, expansion occurs in the haploid germ cells in the process of repair (35). FEN-1 effects on instability may also arise during repair.

Expression of nuclease-defective FEN-1 in human cells.

To test the hypothesis that deficient FEN-1 activity could create precursors to expansion, we created a stable human cell line transformed with nuclease-defective FEN-1 (54, 55) (Fig. 2A). Normal levels of wt enzyme are present, but the mutant can bind and interfere with processing by the wt enzyme at some flaps.

SK-N-MC cells were stably transformed with an expression vector containing the cDNA for nuclease-deficient FEN-1 (54, 55) fused to an HA tag (Fig. 2B). The mutant FEN-1 contains a single aspartate-to-alanine change (D181A) in its nuclease domain (Fig. 2A) (see Materials and Methods). This amino acid change does not alter binding of FEN-1 to its DNA substrate but does eliminate the endonuclease catalytic activity there (54, 55). Stable transformants contained wt and nuclease-defective FEN-1 with structure-dependent affinity for displaced strands. wt FEN-1 can efficiently cleave at the flap junction, leaving a substrate for DNA ligase (Fig. 2C) (27). However, binding of the nuclease-deficient FEN-1 at the substrate can cause failure to cleave the flap and thereby leave a precursor either to expansion or to a strand break (Fig. 2D). The coexpression of the mutant FEN-1 in the SK-N cells was expected to produce a competition with wt enzyme for binding to the flap template (Fig. 2D).

Transformed cells were analyzed for exogenous protein expression (Fig. 2B) and FEN-1 nuclease activity (Fig. 2E). Immunoblot analysis using anti-HA antibody showed that a protein of the appropriate size (the predicted mass of FEN-HA fusion protein is 44,164 Da) was detected in whole-cell extract from hygromycin-resistant cells but not in control parent cells (SK-N) (Fig. 2B). These data confirmed that the mutant enzyme was efficiently expressed in the human cell lines. The level of wt enzyme appears to be regulated in cells, since levels of the mutant enzyme were consistently higher than those of the exogenous wt enzyme in independent clones (Fig. 2B).

We tested extracts from stably transformed SK-N-MC cells to investigate whether expression of the nuclease-deficient FEN-1 inhibited flap processing. FEN-1 activity was measured in vitro by incubating a Flap-1 substrate (Fig. 2E, top panel) in the presence or absence of nuclear extract from the transformed cells (see Materials and Methods). FEN-1 cleaves at either base adjacent to the junction of the displaced strand, leaving fragments of 19 and 21 nt (Fig. 2E) (24, 60). FEN-1 D181A binds as well as the wt but does not cleave (54-56).

If nuclease-defective FEN-1 in extracts imparted a dominant-negative effect, then its presence might substantially lower FEN-1 cleavage activity by binding to substrate and interfering with processing by endogenous enzymes. Despite robust expression of the mutant FEN-1 enzyme (Fig. 2B), however, we found no apparent decrease in FEN-1 cleavage activity in the model template in any of the D181A-transformed cells compared to levels in the parent SK-N-MC (Fig. 2E). Although contribution of another endonuclease to the total activity cannot be ruled out, the pattern of cleavage is identical to that of FEN-1 previously reported in both purified form and in cell lysates (24, 60). Therefore, the concentration of normal enzyme in vitro is sufficient to carry out cleavage in the presence of nuclease-defective FEN-1. In vivo, however, FEN-1 must distribute throughout the genome and particular sites might be sensitive to the presence of the nuclease-defective enzyme. Therefore, we tested whether expression of the mutant FEN-1 caused a selective and/or subtle phenotype in vivo.

Human cells expressing nuclease-defective FEN-1 are susceptible to cell death.

We found that expression of the nuclease-deficient FEN-1 impaired survival. To measure the effects of nuclease-defective FEN-1 in human cells, we cotransformed SK-N-MC cells with GFP and FEN-1 expression vectors and monitored the survival of green cells. For transformants expressing the exogenous wt enzyme, we found no obvious effect on cell survival (Fig. 2F). The number of green cells in the FEN wt culture was similar to that in the SK-N-MC culture as late as 72 h after transfection. In contrast, expression of the nuclease-defective FEN-1 rendered cells more susceptible to death as early as 20 h after transfection, as indicated by the lower number of GFP-positive cells in cultures expressing nuclease-defective FEN-1 (Fig. 2F). Thus, the nuclease-defective enzyme caused the increase in cell death.

Expression of nuclease-defective FEN D181A causes repeat instability.

We hypothesize that site-specific failure to process flaps can cause repeat instability (Fig. 2D). The nuclease-defective FEN D181A should bind at some flaps, thereby interfering with cleavage by the endogenous wt enzyme and leaving a precursor for expansion (Fig. 2D). We tested whether expression of the nuclease-deficient enzyme can alter the stability of repeat tracts of several endogenous loci (Table 2) (18, 20) (see Materials and Methods). Two of the examined loci (Sca1 and HD) are associated with trinucleotide repeat diseases, and three of the loci (ACTC, D6S305, and D6S264) have been shown to be unstable in colon tumors (20). In normal cells, however, all of these microsatellites are stable (as is confirmed by their usefulness as highly informative markers for genetic analysis) (34).

TABLE 2.

Summary of microsatellites tested in SK-N cells^a

Locusb	Repeat unit	No. of repeat units (GenBank)	No. of interrup- tionsc	Length of longest uninterrupted tract (repeat units)	Repeat length of alleles in SK-Nd (repeat units)	Altered in D181A cells
dyn	CAG	10	1	10	10, 10	No
HD	CAG	21	0	21	23, 26	Yes
Sca1	CAG	30	2	15	NDe	No
TBP	CAG	38	4	18	37, 38	No
ACTC	CA	25	0	25	24, 26	Yes
D6S264	CA	20	1	17	ND	No
D6S305	CA	21	0	21	14, 13	No

^aRepeat regions at each of the sites indicated were amplified and sequenced in SK-N-MC parent and FEN-transformed cells.

^bdyn, prodynorphin; Sca1, spinocerebellar ataxia type 1 gene; TBP, TATA binding protein; ACTC, D6S264, and D6S305, CA microsatellite markers (20).

^cImperfections in the repeating units were sometimes present, and the number of these at each locus as well as the length of the longest uninterrupted tract is indicated.

^dThe length as determined by amplification in SK-N-MC cells at some loci differed from that of the GenBank sequence.

^eND, not determined.

We analyzed DNA from parental SK-N-MC cultures, five independent FEN wt-transformed cultures, and five independent D181A-transformed cultures. Cells were not treated with drugs nor was any selection applied. All cells in the population were genetically identical at the start of the experiment. Therefore, detection of a change originating in a single cell would be significant, since it must be distinguished from the background of the population with the parental allele. A change in repeat length would not be expected to confer selective advantage to a cell. Therefore, we relied on the sensitivity of GeneScan analysis to detect changes (35, 36) (Fig. 1C).

In the GeneScan method, repeat-containing regions are PCR amplified using one primer that is fluorescein labeled. The lengths of DNA products can be determined after resolution on denaturing polyacrylamide gels, and each amplified product is seen as a peak in the GeneScan trace (35, 36) (Materials and Methods). At the HD allele, for example, separation of the two prominent peaks in SK-N-MC cells indicates that the long and short HD alleles differ by 9 bp or three repeats (Fig. 3A, HD [right panel]). Alterations in repeat length cause changes in the sizes of PCR product. Since repeat lengths change by a discrete number of nucleotides, peaks that shift position are often coincident with another peak in the trace. Thus, repeat length changes are identified either as the appearance of a new peak or by changes in peak areas (Fig. 3A; compare results for parent SK-N-MC and D181A).

Amplification of repeat-containing regions typically yields a prominent band with adjacent smaller bands (Fig. 3A) (20, 69). SK-N-MC and FEN wt-transformed cultures are indistinguishable by microsatellite analysis. Reproducible deviations from the parental pattern were detected in D181A cultures at two test loci, ACTC (a microsatellite within the gene encoding cardiac muscle actin) and HD (Table 2; Fig. 3A). Changing peak areas within the GeneScan traces indicated that in the D181A sample, there were numbers of repeats different from those in the SK-N-MC parent or FEN wt at both loci (Fig. 3A). To confirm that the altered patterns were due to repeat-length changes within the target gene, both the ACTC and HD alleles were amplified with two different primer sets (primer 1 and primer 2) (Fig. 3A). For both ACTC and HD, amplification with alternative primer sets yielded the same pattern of products but shifted by the appropriate distance (Fig. 3A; results are shown for ACTC only [compare primer 1 and primer 2 results]). Thus, the observed microsatellite instability within the PCR products was gene specific.

At both ACTC alleles and the long HD allele in D181A-transformed cultures, the ratio of the prominent peak area (Fig. 3A) is altered relative to that of the secondary peak in the trace (Fig. 3A). Using both sets of primers (Fig. 3B to D), we calculated the fraction under the prominent peak (most-abundant repeat length) for the ACTC alleles and the long HD allele in independent amplification reactions (Fig. 3B to D). The analysis demonstrates that the repeat length of one HD allele was indistinguishable among the SK-N-MC and FEN wt-transformed cultures but differed significantly in FEN D181A-transformed cultures (P < 0.0001) (Fig. 3B to D). Both sets of primers yielded similar results, indicating that the repeat lengths reflected a property of the specific alleles (Fig. 3B and C). As an internal control, we found that the shorter HD allele was the same in all of the samples (Fig. 3D).

Therefore, expression of nuclease-defective FEN-1 is sufficient to cause microsatellite instability. An uncleaved flap may be a direct precursor to instability and may disrupt repair activities in transformed cells.

Expression of nuclease-deficient FEN-1 causes a reduced capacity to correct strand breaks.

Experiments with nuclease-defective FEN-1 confirmed that repair of displaced flaps is essential for genome stability (Fig. 3). Instability must, therefore, arise from faulty repair of flaps. To test relevant processing events, we treated cells with DNA-damaging agents that might directly or indirectly create flaps. MMS increases the number of alkylated bases, which are typically repaired by BER (10). BER causes single-strand DNA breaks in the process of removal of alkylated bases, and this process can be FEN-1 dependent (13, 32, 33, 47, 49, 56, 64). However, in cell cultures, a high degree of MMS-mediated single-strand breaks can also result in double-strand breaks (66).

We also used cisplatin to induce strand breaks. Cisplatin creates bulky lesions that can be removed by NER (3, 10, 11, 28). Cisplatin can also form interstrand cross-links that are corrected by double-strand break repair (4, 73). After either treatment, survival of drug-treated cells was measured by colony survival assay (Materials and Methods).

Expression of the exogenous wt FEN-1 and D181A FEN-1 caused cells to be hypersensitive to MMS (Fig. 4A) compared to the parent SK-N-MC cells. However, cells expressing the nuclease-defective FEN-1 did not differ in MMS sensitivity relative to the cells exogenously expressing wt FEN-1 (Fig. 4A). The activity of FEN-1 is regulated by its direct interaction with many cellular enzymes, including proliferating cell nuclear antigen, a DNA polymerase δ, and ɛ processivity factor (14, 19, 22, 31, 38, 39, 56, 63, 64, 68); the transcriptional coregulator and acetylase P300 (25); and the WRN helicase (6). Furthermore, the cell cycle regulator p21^waf1/CIP shares a proliferating cell nuclear antigen interaction domain with FEN-1 (64, 67), and the competition between FEN-1 and p21 modulates the response to DNA damage (64). That cells expressing wt and nuclease-defective FEN-1 were equally susceptible to MMS suggests that the levels of exogenous protein (expressed under control of the cytomegalovirus promoter) can disrupt the protein-protein interactions required for proper cellular metabolism.

FIG. 4.

Nuclease-deficient FEN-1 renders SK-N cells differentially sensitive to cisplatin-induced death. (A) Survival curve of SK-N cells after treatment with MMS. Open circles, SK-N cells; filled squares, SK-N cells expressing wt FEN-1; filled circles, SK-N cells expressing D181A FEN-1. Points represent the average number of colonies (± SEM) per dish. Open circles, SK-N-MC cells; filled squares, FEN wt-transformed cells; filled circles, D181A-transformed cells. (B) Survival curve of SK-N cells after treatment with cisplatin. Points are defined as described for panel A.

After treatment with cisplatin, in contrast, cells expressing the mutant D181A FEN-1 were more sensitive than those expressing the exogenous wt FEN-1 (Fig. 4B). At levels of cisplatin sufficient to kill all of the D181A-transformed cells, more than 80% of the parent SK-N-MC cells survived. Thus, cells were differentially sensitive to repair of cisplatin-induced lesions depending on the nuclease capability of FEN-1. MMS and cisplatin damage are repaired by double-strand break enzymes, yet only the cisplatin-induced damage created a differential sensitivity to the mutant FEN-1. These data raise the possibility that NER, which can remove bulky lesions, might dominate when flap processing is impaired. Repair downstream of the unprocessed flap might be required to complete the expansion (Fig. 2D).

Defective flap processing blocks assembly of Rad51/BRCA1 in repair foci after cisplatin treatment.

We hypothesized that expression of the exogenous enzyme might perturb the behavior of key repair enzymes. To test the hypothesis, we analyzed the recruitment of repair proteins to nuclear foci. Fixed cells (see Materials and Methods) were incubated with specific antibodies to identify components of nuclear foci. Analyses of control proteins confirmed that they were present in the correct subcellular compartment. Cells were treated and images were taken in parallel under identical conditions, with the exception of the primary antibody. For example, CREB binding protein, ERCC1, and p21 showed distinctive patterns and intensities. All utilized the same secondary antibody, which, in the absence of primary antibody, generated no fluorescence signal (Fig. 5A). These data indicated that the signal was specific to the primary antibody under the conditions of our experiments (Fig. 5A to F). The analysis was also able to detect functional differences in subcellular localization. BRCA1 was not detected in the nuclei of HCC1937 cells that lack BRCA1 (Fig. 5E) (7, 53, 65). In contrast, BRCA1 foci were observed in revertant HCC1937 + 5′mycBRCA1 cells that were transformed with functional BRCA1 (Fig. 5F).

To test repair activity of cisplatin lesions, SK-N-MC parent cells and FEN wt and D181A-transformed cells were fixed after treatment and then stained with the appropriate antibodies. In cells expressing exogenous wt FEN-1 or D181A, cisplatin damage caused FEN-1 to be recruited to repair foci in the nucleus (Fig. 5G). Therefore, we monitored repair enzymes that redistributed with FEN-1 under different conditions of flap processing. We hypothesized that relevant repair pathways would be perturbed when the mutant FEN-1 protein was overexpressed.

Rad51 is required for homologous repair and recombination at strand breaks. Previous studies have demonstrated its role in repair of strand breaks and of cisplatin-induced damage (4). We found that at 30 min after cisplatin treatment, recruitment of Rad51 to nuclear foci decreased in the D181A-transformed cells but increased in SK-N-MC and FEN wt-transformed cells (Fig. 6A). These data raised the possibility that FEN-1 plays a role in strand break repair.

FIG. 6.

Defective flap processing blocks recruitment of Rad51, and loss of BRCA1 blocks recruitment of FEN-1. (A) Rad51-positive nuclear foci (conditions and labels are defined as for Fig. 5G) after cisplatin treatment. (B) FEN-1-positive foci in HCC1937 cells that lack BRCA1 (open circles) and in revertant HCC1937 + 5′mycBRCA1 cells (filled squares). (C) ERCC1-positive nuclear foci after cisplatin treatment. (D) FEN-1-positive foci in Rovid cells or in NER-defective XPC 671 cells (8).

We hypothesized that if Rad51 and FEN-1 are normally recruited in response to cisplatin-induced damage, then cells defective in Rad51 pathway should show abnormal recruiting of FEN-1. We, therefore, tested whether FEN-1 was recruited to repair foci in the HCC1937 cell line, which lacks a functional BRCA1 (7, 53, 65). In BRCA1-deficient cells treated with cisplatin, Rad51-containing complexes do not assemble and DNA repair is severely impaired (4). In HCC1937 cells, the number of FEN-1-containing foci decreased after cisplatin treatment (Fig. 6B). However, recruitment of FEN-1 to nuclear foci was restored after cisplatin treatment of the revertant HCC1937 + 5′mycBRCA1 line, which expresses a functional BRCA1 transgene (Fig. 6B). Thus, the presence of wt FEN-1 in repair foci was restored when Rad51 assembly was competent. These data suggested that FEN-1 normally plays a role in BRCA1/RAD51-dependent repair but that impaired flap processing invoked an alternative pathway to compensate for a Rad51 deficit.

A substantial body of evidence has shown that NER can remove cisplatin intrastrand cross-links (3, 10, 11, 28). Therefore, we tested whether key components of the NER pathway were recruited to repair foci when flap processing was impaired. ERCC1 forms a nuclease with XPF that cleaves 5′ to a bulky lesion (3, 10, 11). We found that within 30 min of cisplatin treatment, there was a marked increase in the number of ERCC1 nuclear foci in D181A-transformed cells (Fig. 6D and 5F). Under the same conditions, cells overexpressing the wt FEN-1 displayed little to no change (Fig. 6C and 5F). It appeared that some components of the NER pathway were recruited to repair foci when flap processing was impaired. For cells deficient in NER (XPC671 cells) (8) (Fig. 6D), in contrast, there was no effect on FEN-1 focus formation relative to control cells (Fig. 6D). The staining of ERCC1 appeared to be specific. We tested other components of the NER pathway. For XPG, focus formation in D181A-transformed cells differed from that of FEN wt-transformed or from parental SK-N-MC cells but with a somewhat different pattern from that seen with ERCC1 (data not shown). The percentage of XPG foci in D181A-transformed cells decreased (56.2% ± 2.0%) relative to parent SK-N cells (74.9% ± 3.5%) and to FEN wt-transformed cells (99.9% ± 5.4%) within 30 min of cisplatin treatment.

DISCUSSION

We have shown that in human cells, expression of nuclease-defective FEN-1 has profound effects, including instability at endogenous repetitive sites and alterations in the response of RAD 51. Inefficient flap processing appears to prevent Rad51-mediated repair and may invoke components of NER to overcome the Rad51 deficit. Taken together, the data indicate that inefficient flap processing by FEN-1 contributes to CAG instability and confirms that wt FEN-1 has a ubiquitous role in maintaining genome stability in eukaryotes from yeast to humans.

Model for nuclease-defective FEN-1-mediated instability in mammals.

The data suggest a model by which FEN-1 maintains genome stability and aberrant FEN-1 processing underlies instability (Fig. 7). Normal processing by FEN-1 removes flaps that might give rise to instability (Fig. 3 and 7A). Although the mechanism of interaction is unknown, FEN-1 repair appears to depend on a complex comprising Rad51 and BRCA1 (Fig. 6B and 7A). Previous studies have not identified a role for FEN-1 in double-strand break repair (56), but Rad51 and BRCA1 are required for repair of cisplatin-induced DNA damage (4, 5). We show here that repair of cisplatin damage also requires functional FEN-1 (Fig. 4B) and that recruitment of FEN-1 depends on BRCA1 (Fig. 6B). Furthermore, defective FEN-1 nuclease activity inhibits recruitment of Rad 51 to repair foci after cisplatin treatment (Fig. 6A). Cells can survive in the presence of the nuclease-defective enzyme, but our studies show that its expression causes a failure to assemble Rad51-containing repair complexes even in the presence of wt FEN-1. This implies that repair in the absence of intact Rad 51 complexes occurs via other pathways (Fig. 5F and 6A).

FIG. 7.

A possible model to explain FEN-1-dependent genome stability and how defective flap processing causes triplet expansion. (A) Displaced strand is cleaved to leave precursor to ligation. wt FEN-1 is required for formation of Rad 51/BRCA1 complexes during normal flap processing. The question mark indicates that the mode of interaction of these components with FEN-1 and the sequence of the binding events are unknown. (B) A stable hairpin forms at the CAG repeat with mismatched base pairs in the stem. The stable hairpin blocks FEN-1 processing (Cleavage block) and formation of Rad51/BRCA1 repair complex and recruits a mismatch repair complex. Defective flap processing causes recruitment of another repair pathway to excise the hairpin. Expansion occurs in the process.

The absence of flap endonuclease (Rad27) in yeast causes a mutator phenotype with genome-wide instability. In trinucleotide expansion disease patients, however, FEN-1 is not defective, since only repeats at the disease locus are unstable. Thus, single-site mutation in trinucleotide diseases does not reflect a general fault of the protein machinery but reflects the inability of this machinery to function normally at the repeat site (20). Flaps comprising CAG repeats form stable hairpins with mismatched bases in their stems that are recognized by an Msh2 complex (17, 35, 48). We and others have previously shown that stable hydrogen bonds within the hairpin can prevent FEN-1-mediated flap processing (26, 60). Therefore, at the disease locus, FEN-1 is essentially nuclease defective and leaves a precursor for expansion (Fig. 7B). We speculate that in disease-related expansion, the proteins that repair the precursor flap are similar to those recruited during expression of the nuclease-defective FEN-1. If this were correct, then Rad51/BRCA1 complexes would fail to assemble at the blocked FEN-1 site (Fig. 7B) and an alternative pathway involving ERCC1 would be recruited under conditions in which flap cleavage is inefficient.

The alternative repair pathway is unlikely to be canonical NER, since XPG focus formation in D181A cells does not follow the same pattern as that of ERCC1 within a similar time frame after cisplatin treatment. ERCC1 has a role in NER but also has other functions in the cell, notably in strand break repair (1). In fact, evidence for a non-NER pathway that involves ERCC1 has been recently reported (12). Chinese hamster ovary lines defective in individual components of NER were unable to repair both cisplatin-induced interstrand and intrastrand cross-links. ERCC1- and XPF-deficient lines were extremely sensitive to cisplatin damage, however, in contrast to lines deficient in other components of NER. Under these conditions, cisplatin adducts did not appear to involve the formation of DNA double-strand breaks. EFCC1-deficient cells cannot support NER. However, the high level of cisplatin sensitivity of ERCC1- and XPF-deficient cells indicates a defect in another pathway, possibly single-strand annealing.

The alternative pathway that gives rise to trinucleotide instability is unclear. However, one of two mechanisms may underlie the mutation. A repair complex may be recruited but fail to excise the hairpin, in which case the hairpin is incorporated into the DNA (Fig. 7B). Alternatively, a repair complex may succeed in excising the hairpin; then extra repeats are incorporated after slippage during repair synthesis (Fig. 7B). Further studies employing multiple repair enzymes will be required to firmly establish a mechanism.