Among replication mutations that destabilize CAG repeat tracts, mutations of RAD27, encoding the flap endonuclease, and CDC9, encoding DNA ligase I, increase the incidence of repeat tract expansions to the greatest extent. Both enzymes bind to proliferating cell nuclear antigen (PCNA). To understand whether weakening their interactions leads to CAG repeat tract expansions, we have employed alleles named rad27-p and cdc9-p that have orthologous alterations in their respective PCNA interaction peptide (PIP) box. Also, we employed the PCNA allele pol30-90, which has changes within its hydrophobic pocket that interact with the PIP box. All three alleles destabilize a long CAG repeat tract and yield more tract contractions than expansions. Combining rad27-p with cdc9-p increases the expansion frequency above the sum of the numbers recorded in the individual mutants. A similar additive increase in tract expansions occurs in the rad27-p pol30-90 double mutant but not in the cdc9-p pol30-90 double mutant. The frequency of contractions rises in all three double mutants to nearly the same extent. These results suggest that PCNA mediates the entry of the flap endonuclease and DNA ligase I into the process of Okazaki fragment joining, and this ordered entry is necessary to prevent CAG repeat tract expansions.
REPETITIONS of the trinucleotide CAG are the known cause of nearly a dozen hereditary neurodegenerative diseases in humans (Everett and Wood 2004). A hallmark of these diseases is the bias toward an increase in the number of repeat units in the disease allele upon passage from the affected parent to child. While malfunctioning of many processes during replication, repair, and recombination could lead to the bias, malfunctioning of replication during the cell divisions preceding the meiotic division of gametogenesis may be the cause of many of the tract expansions. Consistent with this possibility are the results of a study that shows that many of the CAG repeat expansions that take place in male Huntington's disease patients do so before the completion of meiosis, presumably in the rounds of mitotic DNA replication that occur during spermatogenesis (Yoon et al. 2003).
Consequently, a major aim of our studies on CAG repeat tracts has focused on finding yeast DNA replication mutations that elevate the frequency of CAG repeat tract expansions (Schweitzer and Livingston 1997, 1998, 1999; Ireland et al. 2000). Mutations of CDC9 encoding DNA ligase I elevate the frequency of tract expansions. In particular, the cdc9-1 allele causes tract expansions when either CAG or CTG serves as the lagging-strand template (Ireland et al. 2000). The significant increase in the number of expansions when CTG is the lagging-strand template is particularly noteworthy because of the propensity of tracts to contract in this orientation. In addition, the cdc9-1 mutant is the only mutant to yield more expansions than contractions of a long repeat tract in which CAG serves as the lagging-strand template (Schweitzer and Livingston 1998, 1999; Ireland et al. 2000). Finally, we note that of all the temperature-sensitive mutations of essential replication genes that we have tested, cdc9-1 and cdc9-2 elevate the rate of tract expansion greater than do all the others (Schweitzer and Livingston 1999; Ireland et al. 2000). While making comparisons among temperature-sensitive mutations of multiple essential genes is difficult owing to potential differences in their functional impairment, the peculiarity is striking.
The ability of a cdc9 allele to cause CAG repeat tract expansions necessitates comparisons to rad27 mutations. RAD27 encodes the flap endonuclease, an enzyme that removes primers from Okazaki fragments before the fragments are joined by DNA ligase (Bambara et al. 1997; MacNeill 2001). Deletion of RAD27 causes expansion of both tandem and interrupted repeats (Johnson et al. 1995; Tishkoff et al. 1997; Freudenreich et al. 1998; Kokoska et al. 1998; Schweitzer and Livingston 1998; Spiro et al. 1999). In searching for the cause of tract expansions, we find making direct comparisons between rad27 and cdc9 mutations difficult because RAD27 is not essential for cell viability while CDC9 is. Nevertheless, one aim of this study is to make additional comparisons between cdc9 and rad27 mutations using alleles of each gene with orthologous amino acid changes.
Comparison of cdc9 mutations with rad27 mutations is also necessary to achieve our primary goal of defining a mechanism by which DNA ligase I mutations lead to CAG repeat tract expansions. In our report on the discovery of the tract expansion phenotype of cdc9 mutations, we speculated that tract expansions occur in a process in which mutant cells that are tardy in ligation attempt to reiterate flap formation and are subsequently subject to inhibition of the flap endonuclease activity and to the resultant expansions (Ireland et al. 2000). What we pointed out in that report and explore further in this study is the relationship that each enzyme has with proliferating cell nuclear antigen (PCNA). Immunoprecipitation of both the flap endonuclease and DNA ligase I precipitate PCNA in yeast extracts (Ho et al. 2002). Interaction of each enzyme with PCNA is promoted by a conserved PCNA-interaction motif (QXXLXXFF in each enzyme) (Jonsson et al. 1998; Vivona and Kelman 2003). Mutation of the two phenylalanine residues to alanine residues in the yeast flap endonuclease greatly reduces its ability to bind to PCNA (Gary et al. 1999). Human DNA ligase I is prevented from binding to PCNA by a peptide containing its PCNA interaction peptide (PIP) box. Changing the phenylalanine residues in the peptide renders it incapable of inhibiting the binding of DNA ligase I to PCNA (Montecucco et al. 1998). The recent publication of the crystal structure of human DNA ligase I with a DNA substrate shows that DNA ligase I encircles the DNA as does PCNA, is as large as the homotrimeric PCNA, and is shaped to make multiple contacts with PCNA (Pascal et al. 2004). Unfortunately, the crystal structure was determined on a truncated protein lacking the PCNA interaction domain, making conjecture on the role of the PCNA interaction domain difficult. (See supplemental Figure S4 from Pascal et al. (2004) at http://www.nature.com/nature/journal/v432/n7016/suppinfo/nature03082.html for a model of the docking between the two proteins.) In addition, a recent report shows that human DNA ligase I binds to multiple subunits of replication factor C (RF-C), suggesting that DNA ligase I loading onto DNA may depend on the clamp loader as well (Levin et al. 2004). The crystal structure of the flap enodnuclease complexed with PCNA has also been determined (Chapados et al. 2004; Sakurai et al. 2005). It shows that the flap endonculease may rotate upon PCNA binding to reach its substrate. Biochemical studies that measure the activity of DNA ligase I and the flap endonculease in the presence and absence of PCNA suggest a more prominent role for PCNA in exonuclease cleavage than in nick ligation. PCNA stimulates the exonucleolytic activity of the flap endonuclease (Li et al. 1995; Tom et al. 2000). Studies on the effect of PCNA on DNA ligase I activity have yielded results that range from inhibitory to slightly stimulatory (Jonsson et al. 1998; Tom et al. 2001; Jin et al. 2003). While the exact role of the interaction of PCNA with DNA ligase I is not completely understood, the results of this study suggest that the interactions among PCNA, DNA ligase I, and the flap endonuclease are important in maintaining CAG repeat stability and in preventing tract expansions.
MATERIALS AND METHODS
All strains are isogenic to strain CAG101 (Schweitzer and Livingston 1999). Strain CAG101 is a derivative of strain SSL204 in which a CAG repeat tract of 78 repeat units in length was embedded in a cloned copy of ADE2 and integrated at ARO2 (Dornfeld and Livingston 1991; Maurer et al. 1996).
Strain construction, determination of tract length stability, and statistical tests:
To introduce replication mutations into the strain containing the CAG repeat tract, we first construct mutant derivatives of our parental wild-type strain, mate the derivatives to the isogenic wild-type strain containing the CAG repeat tract, and sporulate the resulting diploids. To reduce the possibility of unwanted modifiers, we characterize two or more spores with the replication mutation and the CAG repeat tract. To examine CAG repeat tract stability, we spread independent yeast colonies of ∼4 × 106 cells on agar dishes to generate sibling colonies arising from single cells of the parental colony (Schweitzer and Livingston 1999; Ireland et al. 2000). We sample 32 sibling colonies from each of the parents and record the number of sibling colonies that have tract lengths that are different from the parental colony (Ireland et al. 2000). The values for the number of changes, expansions, and contractions for each set of siblings are given in Table 1. These values form the basis of the statistical tests (Table 2). The summation of the number of expansions and contractions as well as the mean sizes of the expansions and contractions are given in Table 3. Because the samples of 32 cells are independent, the frequencies recorded for one strain are compared to the frequencies in a second strain. Previously we used a one-sided Fisher's exact test to compare frequencies of mutant and wild-type strains. In this test, the values (from each set) are ranked in order and the exact probability of the highest values ending up in one set is calculated. In this study, we have used the nonparametric Mann-Whitney test (http://vassun.vassar.edu/∼lowry/VassarStats.html). This is similar to the Fisher's exact test in that the values are assigned an ordinal ranking and the exact probability of choosing sets of values is calculated. We used this test because it is more efficient when the frequency values of the two sets overlap. We judge comparisons between two strains to be significant when P < 0.05 on a two-sided test. The statistical comparisons are listed in Table 2. Further discussion of the reproducibility of our results is given in the appendix.
The cdc9-p allele (F44A/F45A) was created using a mutagenic PCR primer (5′-CCTAAACAAGCCACTTTGGCTAGAGCTGCAACTAGTATGAAAAATAAGCCAACAGAAGGT) that changed phenylalanine codons 44 and 45 into alanine codons and created a SpeI restriction site downstream of the codon changes without altering the amino acid sequence. The rad27-p (F346A/F347A) allele was a gift from Dmitry Gordenin (Gary et al. 1999). The pol30-90 (P252A/K253A) allele was provided by Peter Burgers (Eissenberg et al. 1997). Each mutation was substituted for the wild-type sequence in the chromosomes by two-step disruption. Because many of the mutations used in this study do not have growth defects, the presence of all mutations in mutant strains was confirmed by direct examination of the chromosomal DNA either by DNA sequencing or by restriction site analysis of appropriate PCR products.
Methyl methanesulfate sensitivity:
All strains were grown and all assays were performed as previously described (Boundy-Mills and Livingston 1993).
To epitope tag DNA ligase I with three HA epitopes, a portion of CDC9 extending from nucleotide 1701 of the open reading frame to the end of the last codon (position 2265) was used to replace CDC17 in pRS406-CDC17-3HA (Ricke and Bielinsky 2004). The plasmid was made linear with AflII and integrated into both the wild type and the cdc9-p mutant. Cells were grown to a density of 2 × 107 cells/ml and extracts were prepared as previously described (Cobb et al. 2003; Ricke and Bielinsky 2004). To precipitate PCNA, we used a mouse monoclonal antibody directed against PCNA (Labvision, Fremont, CA). The blotting antibodies were the HA-7 anti-HA antibody (Sigma, St. Louis) and a rabbit polyclonal antibody (PC-10) directed against PCNA (Labvision, Fremont, CA). In treated samples, DNase I (Sigma) was added to 3 units/μl and MgCl2 to 10 mm and incubated on ice for 30 min. To precipitate HA-tagged DNA ligase I, we used anti-HA-agarose (clone HA-7 Sigma). The beads were washed four times with 1 ml cold RIPA buffer (50 mm Tris base, 0.25% w/v deoxycholate, 1% v/v Igepal CA-630, 150 mm NaCl, 1 mm EDTA with a final pH of 7.4) as described in the instructions from the supplier (Sigma). The blotting antibodies were the same as those used for the PCNA co-immunoprecipitation. Quantification was carried out by scanning the exposed film with a Fuji FLA-5000 imager (Stamford, CT).
Alleles of DNA ligase I, the flap endonuclease, and PCNA that alter their interactions:
Previous studies have shown that both rad27 and cdc9 mutations elevate the expansion frequency of CAG repeat tracts. Those studies have used either deletion or catalytically impaired alleles of RAD27 or temperature-sensitive mutations of CDC9 (Freudenreich et al. 1998; Schweitzer and Livingston 1998; Spiro et al. 1999; Callahan et al. 2003; Liu et al. 2004). In this study we have characterized alleles of CDC9 and RAD27 that have mutations in their PCNA interaction domain (PIP box) to understand whether alteration of interactions between the enzymes and PCNA affect CAG repeat tract stability.
To further draw out the role of PCNA binding in the maintenance of repeat tract stability, we have also incorporated in our studies a PCNA allele that alters its ability to bind to proteins that contain PIP boxes. In addition, we have made pairwise combinations of the CDC9, RAD27, and POL30 alleles to determine whether any epistatic, additive, or synergistic relationships exist among them.
The rad27-p allele (F346A/F347A) that has changes in the two phenylalanine residues of the flap endonuclease PIP box has been characterized with respect to a number of phenotypes, including methyl methanesulfonate (MMS) sensitivity and PCNA binding (Gary et al. 1999). It has not been established whether these mutational changes alter CAG repeat stability. The PCNA allele pol30-90 (P252A/K253A) alters amino acid residues in the hydrophobic binding pocket in which the PCNA interaction motif of DNA ligase I, the flap endonuclease, and other PIP box-containing proteins bind (Gulbis et al. 1996; Jonsson et al. 1998; Vivona and Kelman 2003; Chapados et al. 2004; Pascal et al. 2004; Sakurai et al. 2005). Both the pol30-90 allele and a similar PCNA allele, pol30-104 (A251V), have been exploited to study interactions between PCNA and enzymes involved in replication and repair (Amin and Holm 1996; Flores-Rozas et al. 2000; Schmidt et al. 2002). PCNA purified from the pol30-90 mutant has been characterized as lacking the ability to stimulate the flap endonuclease in a biochemical assay (Eissenberg et al. 1997). We have previously found that this allele reduces CAG repeat tract stability (Schweitzer and Livingston 1999).
Characterization of the cdc9-p allele:
To begin our study, we have made a mutation in the PIP box of the yeast CDC9 gene in which the two phenylalanines have been mutated to alanine residues. The mutant allele, cdc9-p, was substituted for the wild-type copy. The cdc9-p mutant is viable at 30° (as well as at 18°, 25°, and 35°) and exhibits no obvious growth defect such as reduced growth rate. Because CDC9 is an essential gene, the mutant protein must retain sufficient catalytic function to support robust cell growth.
Before investigating the effect of this cdc9 allele on CAG repeat tract stability, we carried out additional characterizations of its phenotypes, some in combination with the rad27-p and pol30-90 alleles of the flap endonuclease and PCNA, respectively. We first examined the MMS sensitivity of cdc9-p and compared it to the MMS sensitivity of rad27-p. Both the deletion of the RAD27 gene and missense mutations of CDC9 cause MMS sensitivity (Montelone et al. 1981; Gary et al. 1999). Both rad27-p and cdc9-p alleles are negligibly sensitive to exposure to MMS (Figure 1A). However, the double mutant has a sensitivity that is slightly more than that expected by addition of the results of the single mutants (Figure 1A). This is the first indication of a genetic interaction between the two alleles. Additionally, the rad27-p allele increases the MMS sensitivity of a rad51 deletion mutation (Gary et al. 1999). The cdc9-p allele shows a similar behavior in increasing the MMS sensitivity of a rad51Δ mutant (Figure 1B). This increased sensitivity demonstrates that the cdc9-p allele either introduces lesions that require the action of the recombinational repair system or prevents completion of recombinational repair or both.
Next, we combined both cdc9-p and rad27-p with the mutant allele of PCNA, pol30-90 (P252A/K253A), that alters amino acid residues in the hydrophobic binding pocket in which the PCNA interaction motif of DNA ligase I and the flap endonuclease bind (Eissenberg et al. 1997). Both cdc9-p and rad27-p can be combined with pol30-90. Both double mutants and the triple mutant are viable, although the triple mutant grows more slowly than either double mutant. Each double mutant, cdc9-p pol30-90 and rad27-p pol30-90, is more sensitive to exposure to MMS than either single mutant (Figure 1C). In addition, the rad27-p pol30-90 double mutant is slightly more sensitive than the cdc9-p pol30-90 double mutant. The triple mutant is more sensitive than either double mutant and has a similar sensitivity to that of a rad27Δ strain (compare Figure 1, A and C). The increasing severity of MMS sensitivity in progressing from single to double to triple mutants indicates the importance of PCNA-mediated complexes in recovery from MMS-damaged DNA.
Finally, to investigate whether the mutational changes in cdc9-p abrogate the interaction of DNA ligase I with PCNA, we investigated the interaction by co-immunoprecipitation. For these studies we have used HA-tagged DNA ligase I in strains harboring either epitope-tagged CDC9 or cdc9-p. The crystallographic determination of the structures of PCNA and DNA ligase I show that both are able to encircle duplex DNA (Gulbis et al. 1996; Pascal et al. 2004). Furthermore, DNA ligase I appears to complement the shape of the front face of PCNA, suggesting that the context in which they are found within the cell, and the context in which they participate in DNA replication, is when both are bound to DNA (Pascal et al. 2004). Consequently, we examined the interaction between the two proteins in extracts without enzymatically digesting the cellular DNA. PCNA is precipitated better by the anti-HA antibody directed against DNA ligase I from the extract made from wild-type cells than by that from the extract made from cdc9-p cells (Figure 2A). The reduced amounts are neither the consequence of reduced quantities of mutant protein in the extract (“Input” in Figure 2, A and B) nor the consequence of reduced immunoprecipitation of the mutant ligase from the mutant extracts (Figure 2A). The reduced interaction between the mutant DNA ligase I and PCNA is also apparent when PCNA is precipitated and the blot is probed for HA-tagged DNA ligase I (Figure 2B). The results show that the two phenylalanine residues in the PIP box of yeast DNA ligase I play a role in keeping DNA ligase I close to PCNA when bound to DNA.
To examine whether the interaction between DNA ligase I and PCNA occurs in the absence of DNA, we also treated samples with DNase I (Figure 2C). Antibody precipitation of PCNA in treated extracts precipitated wild-type DNA ligase, showing that the interaction of the two proteins survives DNase I digestion. Significantly less mutant DNA ligase I was precipitated under these conditions, again showing the necessity of the two hydrophobic residues for interaction between DNA ligase I and PCNA (Figure 2C). We could not detect consistently an interaction between DNA ligase I and PCNA when HA-tagged DNA ligase I was precipitated from the treated extracts. This inability may have resulted from the use of a different type of precipitating beads that required additional detergents in the wash buffer.
cdc9-p and rad27-p destabilize CAG repeat tracts:
Next, we began our characterization of the cdc9-p and rad27-p alleles to learn whether they destabilize CAG repeat tracts as do other alleles of the two genes. The cdc9-p allele destabilizes the CAG repeat tract at 30°, yielding a mixture of tract expansions and contractions (Table 3). This allele yields approximately one-third the number of tract expansions as the cdc9-1 allele does, but about the same number of tract contractions as the cdc9-1 mutant does (Table 3). The mean length of the expansions is similar to that found in the cdc9-1 mutant (Table 3), while the mean length of the contractions is longer than that found in other replication mutations that we have studied (Schweitzer and Livingston 1999; Ireland et al. 2000).
Next, we compared our cdc9-p result with that of a mutant strain harboring rad27-p (Gary et al. 1999). As with the cdc9-p allele, the rad27-p allele destabilizes the CAG repeat tract, yielding a mixture of expansions and contractions (Table 3). The frequencies of both tract expansions and tract contractions are slightly greater than the values for the cdc9-p mutant (Table 3). This allele does not destabilize the CAG repeat tract nearly to the extent to which the rad27Δ mutant does, nor does it cause as a high a ratio of expansions to contractions as does the rad27Δ mutant (Schweitzer and Livingston 1998; Ireland et al. 2000). In addition, the mean length of the tract expansions in the rad27-p mutant is shorter than the value that we had previously determined for rad27Δ [17 repeat units ± 10 repeat units (Ireland et al. 2000)].
Our results show that weakening the interaction of either enzyme with PCNA destabilizes the CAG repeat tract. Each allele has a relatively mild phenotype in the degree to which it destabilizes the repeat tracts, and neither allele creates more tract expansions than tract contractions. Greater destabilization and higher ratios of tract expansions to tract contractions have been recorded in the rad27Δ and cdc9-1 mutants.
To examine their epistatic relationship, we combined the two alleles. The frequency of tract length changes in the double mutant is greater than the sum of the frequencies of the single mutants and is significantly different from each single mutant (Tables 2 and 3). Furthermore, this additive relationship extends to both the expansions and the contractions (Tables 2 and 3). Thus, neither allele is epistatic to the other in permitting the occurrence of expansions or contractions. The number of expansions in the double mutant is nearly twice the sum of the values of the single mutants. A comparison of the number of expansions in the double mutant and the sum of the numbers of expansions in each of the single mutants by a chi-square test shows them to be different (P = 0.023). In addition, there has been a small increase in the ratio of expansions to contractions in the double mutant in comparison to the single mutants, suggesting that the double mutant affects tract expansions more than tract contractions. The results suggest that in single mutants one enzyme might compensate for the other in preventing tract expansions.
The PCNA mutation pol30-90 is epistatic to cdc9-p and acts additively with rad27-p:
The results of the studies on the DNA ligase I and flap endonuclease alleles that reduce the binding affinity of each enzyme to PCNA show that PCNA binding is needed to maintain CAG repeat tract stability. To perturb further the interaction between PCNA and the two enzymes, we have combined the ligase and flap endonuclease alleles with the PCNA allele, pol30-90 (Eissenberg et al. 1997). Like the cdc9-p and rad27-p mutants, the pol30-90 mutant also destabilizes the CAG repeat tract, yielding a few tract expansions and more tract contractions (Tables 1–3⇑).
The behaviors of the two double mutants in destabilizing the CAG repeat tract is different. The combination of cdc9-p with pol30-90 increases the CAG repeat tract instability found in the cdc9-p mutant by increasing the contraction frequency (Tables 2 and 3). A comparison of the pol30-90 single mutant to the cdc9-p pol30-90 double mutant shows no significant increases in expansions or contractions. Indeed, in this double mutant, the number of tract expansions and tract contractions are less than additive. Thus, this double mutant behaves much like the pol30-90 allele.
In contrast, the combination of rad27-p with pol30-90 creates nearly double the number of tract length changes found in the cdc9-p pol30-90 double mutant. This number of tract length expansions and contractions in the rad27-p pol30-90 double mutant is significantly different from the frequencies in either single mutant (Tables 2 and 3). Inspection of the number of expansions and contractions (Table 3) shows that both are nearly double the sum of the events occurring in the single mutants. Although both tract expansions and contractions increase in the double mutant, the ratio of expansions to contractions increases slightly over the values for the single mutants, showing that the double mutant may affect the expansion process more than the contraction process. As before, a comparison of the numbers of expansions in the double mutant to the sum of the numbers of expansions in the single mutants by a chi-square test shows them to be different (P = 0.042).
The difference between the cdc9-p pol30-90 and rad27-p pol30-90 double mutants is illuminated by the observation that the two are significantly different from each other both in the number of expansions and in the number of contractions (Table 2). The results show that tract stability is greatly affected more by perturbations in the flap endonuclease-PCNA interaction than by perturbations in the DNA ligase-PCNA interaction. This overall difference includes an increase in expansions.
A proofreading exonuclease mutation in pol δ does not alter the frequency of CAG repeat tract expansions or contractions in cdc9-1:
To account for tract expansions in the cdc9-1 mutant, we proposed that expansions occur because 5′-flaps at the end of Okazaki fragments either persist or are an outcome of a 5′-flap that is recreated upon unsuccessful ligation in DNA ligase I mutants (Ireland et al. 2000). Recent studies have suggested that 5′-flaps at the ends of Okazaki fragments may branch migrate to produce 3′-flaps (Liu et al. 2004). Furthermore, flap endonuclease mutations, including rad27Δ and rad27-p, are synthetically lethal with pol3-01, a mutation that destroys the proofreading 3′-5′ exonuclease of pol δ (Morrison et al. 1993; Kokoska et al. 1998; Gary et al. 1999). The synthetic lethality may arise from the inability to remove 3′-flaps that occurs in the double mutants because of the incapacity of the flap endonuclease to remove 5′-flaps.
To test whether flaps that might be progenitors of CAG repeat tract expansions could result from 3′-flaps rather than from 5′-flaps in DNA ligase I mutants, we combined cdc9-1 with pol3-01. In contrast to flap endonuclease mutations, cdc9-1 and cdc9-p can be combined with pol3-01. Furthermore, the CAG repeat tract is as unstable in the cdc9-1 pol3-01 double mutant as in the cdc9-1 mutant and yields more expansions than contractions as does the single cdc9-1 mutant (Table 3). That mutational elimination of the 3′-5′ exonuclease of pol δ does not have a large effect on CAG repeat tract expansions in the DNA ligase I mutant argues that the expansion process does not involve a 3′-flap susceptible to proofreading exonuclease digestion.
Our major focus has been to understand how DNA ligase I prevents CAG repeat tract expansions. In this study we have examined mutations affecting the binding of DNA ligase I and the flap endonuclease to PCNA. The cdc9-p and rad27-p mutations behave similarly in destabilizing a long CAG repeat tract. Both elevate the frequencies of tract expansions and tract contractions, showing that PCNA-mediated binding of each protein is important in maintaining repeat tract stability. To determine the epistatic, additive, and synergistic relationships among cdc9-p, rad27-p, and pol30-90 in CAG repeat tract destabilization, we created pairwise double mutants. On the basis of the double-mutant studies we can conclude that DNA ligase I and the flap endonuclease can partially compensate for each other in preventing tract expansions. In addition, the interaction between the flap endonuclease and PCNA is more important in preventing tract expansions than is the interaction between DNA ligase I and PCNA.
Our most important result is that the number of CAG tract expansions in the cdc9-p rad27-p double mutant is greater than additive. An additive relationship might imply that the flap endonuclease and DNA ligase I each prevent tract expansions by different means. Because we found that the number of expansions in the cdc9-p rad27-p double mutant is greater than additive, we reject the possibility that each enzyme acts independently to prevent tract expansions. Instead, we interpret the result to mean that each is a leaky mutation in a single pathway that prevents tract expansions (Game and Cox 1972; Brendel and Haynes 1973). The single pathway is, of course, the final steps in Okazaki fragment maturation.
Another way of interpreting the double mutant result is that one enzyme partially suppresses the expansion phenotype of the other mutation. For example, failure to ligate Okazaki fragments in the cdc9-p mutant might allow for reiterative flap formation by pol δ strand displacement [as we have previously suggested (Ireland et al. 2000)], but the flaps are mostly removed by the flap endonuclease or the Dna2 nuclease before they generate tract expansions. In contrast, in the rad27-p mutant, DNA ligase I might be initially prevented from catalyzing the joining of nascent replication fragments by the flap, but the flap is removed by another nuclease, such as Dna2 or Exo1, and ligation is then successful, preventing tract expansions.
To account for our results, we have developed a model in which PCNA mediates an ordered entry of the flap endonuclease and DNA ligase I into Okazaki fragment maturation (Figure 3). Our model has been formulated on the recent crystallographic studies that have provided structural data on how DNA ligase I and the flap endonuclease bind to PCNA. While our results do not prove this model, we point out how our co-immunoprecipitation and repeat tract stability results are consistent with the structural results.
The crystallographic study of human DNA ligase I lacking its PIP box reveals an enzyme that is as large as the homotrimeric PCNA and encircles the DNA (Pascal et al. 2004). The authors of the study noted that DNA ligase I and the front face of PCNA are complementary in shape. They combined their structure with the solved structure of PCNA to depict the two proteins forming tight-fitting concentric rings around the DNA. In addition, the structure of human DNA flap endonuclease bound to PCNA has been determined (Sakurai et al. 2005). This enzyme forms extensive contacts with PCNA also. The structure shows that the flap endonuclease must rotate across the front face of PCNA to position itself for catalysis, possibly kinking the DNA substrate (Chapados et al. 2004; Sakurai et al. 2005). At this point, whether both enzymes can bind simultaneously to PCNA is difficult to discern. What is interesting about the structures is that they reveal additional interactions between both proteins and PCNA in addition to those that occur between their respective PIP boxes and the hydrophobic pocket on PCNA. The additional interactions may provide for simultaneous binding as has been observed among the archael proteins (Dionne et al. 2003).
Our co-immunoprecipitation and genetic studies are best interpreted by the structural studies. They show that interaction of the flap endonuclease and DNA ligase I through their respective PIP boxes is important for successful interaction between each enzyme and PCNA. Our co-immunoprecipitation studies from yeast extracts that have or have not been treated with DNase I prior to precipitation substantiate the need for the aromatic residues in binding and comment on the role of DNA binding in the association of DNA ligase I with PCNA. Reduced quantities of cdc9-p protein are precipitated with PCNA relative to wild-type DNA ligase I in samples treated with DNase I. In samples left untreated, we consistently observed less coprecipitation from extracts containing the cdc9-p protein than from the wild-type extract, but the reduction is less than that seen in the DNase I-treated samples. Furthermore, less wild-type DNA ligase I is coprecipitated from the DNase I-treated samples than from the untreated samples. We interpret these results to indicate that the contribution of the aromatic residues in maintaining DNA ligase I bound to DNA may be relatively small. Once encircled around the DNA, DNA ligase I may remain bound to the DNA and retain the ability to seal nicks in the absence of PCNA. The potential to remain bound to the DNA in the absence of tight binding to PCNA may explain why enzymological studies do not consistently reveal a stimulatory effect of PCNA on DNA ligation (Jonsson et al. 1998; Tom et al. 2001; Jin et al. 2003). Whether DNA ligase I loading is self-promoted or whether either PCNA or RF-C facilitate DNA ligase I loading is uncertain. What is certain is that the interaction between DNA ligase I and PCNA is abrogated by the mutation of the aromatic residues, and this small diminution is sufficient to destabilize CAG repeat tracts and cause MMS sensitivity when combined with rad51Δ or rad27-p.
Our result on CAG repeat instability in the cdc9-p mutant suggests that the strength of the DNA ligase-PCNA interaction is important for repeat tract stability. While not preventing DNA binding, weakening the strength of the DNA ligase-PCNA complex might disrupt the timing of the entry of DNA ligase I into Okazaki fragment joining, and this difference could be sufficient to cause repeat tract instability. The epistasis of pol30-90 to cdc9-p in the maintenance of tract stability suggests that DNA ligase I may not be dependent solely on PCNA to locate nicked DNA if it can bind to DNA in the absence of PCNA. In addition, DNA ligase I and PCNA appear to make extensive contacts when bound to DNA, and these are likely maintained in the cdc9-p pol30-90 double mutant. Consequently, the combination of cdc9-p with pol30-90 does not exacerbate the repeat tract instability past the effect observed in the more severe of the single mutants. We note that while this epistatic relationship is true for the repeat tract stability phenotype, the cdc9-p pol30-90 double mutant is more MMS sensitive than either single mutant, suggesting that recovery from DNA damage is not wholly similar to preventing repeat tract changes.
In contrast to the role PCNA binding plays in the prevention of repeat tract instability in cdc9-p mutants, the phenotypes of the rad27-p and rad27-p pol30-90 mutants show that the ability of the flap endonuclease to bind to PCNA is crucial in maintaining repeat tract stability. By itself, rad27-p has a phenotype similar to cdc9-p in destabilizing the CAG repeat tract. But, in combination with pol30-90, rad27-p behaves differently from cdc9-p in exacerbating repeat tract instability. Structural studies show that the flap endonuclease appears dependent on PCNA to reach its substrate (Chapados et al. 2004; Sakurai et al. 2005), and enzymological studies show that binding greatly stimulates its nuclease activity (Li et al. 1995; Tom et al. 2000). Furthermore, enzymological studies show that flap endonuclease mutations that impair its flap cleavage activity correlate well with the ability of the mutations to cause CAG repeat tract expansion in vivo (Liu et al. 2004). Consequently, reducing the interaction by combining rad27-p with pol30-90 in our studies exacerbates its phenotype by destabilizing the CAG repeat tract, particularly by increasing the frequency of tract expansions. In essence, we have mimicked the phenotype of rad27Δ in our double mutant.
Our genetic studies do not correlate with all the enzymological studies. In one study, DNA ligase I and the flap endonuclease have been allowed to compete on a substrate containing a flap of CTG repeats (Henricksen et al. 2002). This study shows that in a competition between the two enzymes, DNA ligase trumps the flap endonuclease in capturing bubbles and creating expanded tracts. In contrast, our studies show that wild-type DNA ligase I prevents CAG repeat tract expansions and that mutant copies of DNA ligase I lead to tract expansions. Enzymological studies have also suggested that repeat-containing flaps branch migrate to yield both 3′ and 5′ flaps, particularly in rad27 mutants (Liu et al. 2004). The synthetic lethality of rad27-p with pol3-01 also suggests that branch migration of 5′-flaps into 3′-flaps can occur. The synthetic lethality might occur because 3′-flaps that form in the absence of the flap endonuclease might be removed by the 3′-5′ proofreading exonuclease of pol δ. The rad27-p pol3-01 lethality can be suppressed by high-copy expression of DNA2 (Jin et al. 2003). This suppression might occur by Dna2 facilitating the delivery of the flap endonuclease to flaps or by providing helicase and nuclease activities that compensate for the deficiency of the flap endonuclease (Budd et al. 1995; Bae et al. 1998). Our study on the incidence of tract expansions in the cdc9-1 pol3-01 double mutant suggests that expansions are not created from 3′-flaps. If 3′-flaps serve as a reservoir for tract expansions, then eliminating the proofreading exonuclease might increase the incidence of tract expansions. We interpret the phenotype of the cdc9-1 pol3-01 mutant to mean either that 3′-flaps are not a reservoir of expansion causing flaps in the cdc9-1 mutant, because they never form in the presence of the flap endonuclease, or that the proofreading exonuclease of pol δ is not capable of removing them if they are. Clearly, additional genetic and enzymological studies will be needed to reconcile these differences.
We note that the mutations in the genes encoding the enzymes that catalyze primer removal and joining of Okazaki fragments have phenotypes that are similar to each other and unique among replication mutations. First and foremost they cause a higher rate of CAG repeat tract expansions than other replication mutations do. Second, CDC9, RAD27, and POL30 were genes identified by Merrill and Holm (1998, 1999) in a screen for mutations incompatible with mutations in the recombinational repair system. This implies that mutations in these genes lead to double-strand breaks, and this implication is borne out by studies that show that mutations of these genes lead to a higher incidence of CAG repeat tract “fragility” (Callahan et al. 2003). Why failure to carry out Okazaki maturation should be more susceptible to double-strand breakage and how this “fragility” relates to CAG repeat expansion are not obvious at this time.
Our model suggests that the final steps in Okazaki fragment maturation may be concerted (Figure 3). By this we mean that PCNA does not bind randomly to the myriad of replication and repair proteins that contain PIP boxes. Possibly, the DNA substrates themselves, the participating enzymes, or additional protein factors signal PCNA to order the entry of the enzymes into lagging-strand replication. We see two possibilities. The first is that PCNA does not permit the removal of flaps until DNA ligase I signals its presence by binding tightly to the clamp (Figure 3). Supporting this model is the observation that nascent DNA fragments purified from cdc9-1 mutant cells retain RNA primers (Bielinsky and Gerbi 1999). While the expectation is that all steps of primer removal are carried out in the absence of DNA ligase I, the presence of RNA primers suggests the necessity of DNA ligase I for their removal. The second possibility is that the failure of DNA ligase I to reach a nick or to seal a nick elicits a control reaction in which PCNA directs pol δ, Dna2, and the flap endonuclease to attempt the process of flap formation and removal a second time. In either scenario, the longevity of repeat-containing flaps leaves the template as a single strand capable of collapsing into the hairpins that give rise to tract contractions, makes the single-strand template susceptible to breakage (“fragility”), and, most importantly, provides the precursor for repeat tract expansions.
APPENDIX: REPRODUCIBILITY OF RESULTS
To investigate the reproducibility of our assay, we increased our sample size from 6 to 12 sets of sibling colonies for some of the experiments reported in this study. Supplemental Figure S1 at http://www.genetics.org/supplemental/ shows the frequencies of expansions and contractions in wild-type and mutant strains in the first and second sets of six samples. With a few exceptions at higher values of tract contractions, the frequencies are similar.
These results are reproducible because the rates of tract expansions and contractions are between 3 × 10−3 and 3 × 10−2 events/cell/generation. At these rates the product of the rate times the number of cells is >1 when colonies reach a size of a few hundred to a thousand cells. As the value of this product rises above 1, the incidence of tract length changes is no longer governed by stochastic fluctuation but becomes proportional to the number of cells. The period during which the appearance of tract length changes becomes proportional to the number of cells has been called the Luria-Delbruck time (Rosche and Foster 2000). This name was given because in their first description of the use of multiple cultures to determine mutation rates Luria and Delbruck (1943) combined the results of their multiple cultures, recognizing that at low mutation rates the number of mutations would be proportional to the number of cells if the culture size was very large. A representation of this phenomenon using our analyses is shown in supplemental Figure S2 at http://www.genetics.org/supplemental/.
An indication of the precision of our results can be observed in the expansion results for cdc9-1. Each value represents a sample size of 32 cells (sibling colonies) from a colony (parental colony) of ∼4.2 × 106 cells. The two sets of six values were done 3 years apart by two different researches, using two different incubators in two different buildings. The first six values were 1, 2, 3, 3, 3, 4 and the second were 0, 1, 2, 3, 5, 8. Combining both sets, the 95% confidence limit encompasses the values 2, 2, 3, 3, 3, 3, 4, 5 (Dixon and Massey 1969).
Another indication of the reproducibility of our data comes from analysis of the sizes of the expansions that we measured in cdc9-1. The sizes of expansions and contractions are often different even within a single sample. For example, the expansions in tract length, measured in additions to the number of repeat units, for the first set of six cdc9-1 samples were 1; 34, 36; 1, 3, 5; 1, 8, 16; 16, 17, 38; 3, 9, 14, 19 and for the second set were no expansions; 44; 22, 25; 14, 15, 19; 1, 2, 7, 8, 19; 20, 20, 20, 20, 29, 29, 29, 80, respectively, where the semicolon separates the values for the six determinations and the sets appear in the order of the preceding paragraph. The spectrum of expansion lengths gives insight into the rate and independence of tract length changes. Changes in different lengths are likely the result of independent events. For cdc9-1, if the mean, rmean, value for expansions is 3 in a sample of 32 cells from a colony that has grown 22 generations to reach a size of 4.2 × 106 cells, the rate of expansion in the single colony is ∼1.8 × 10−2 expansions/cell/generation using the Luria and Delbruck (1943) derivation of rmean = m ln(m). The number of tract length expansions that have occurred in the colony, m, is ∼37,800. (The rate and mutation number, m, have been calculated taking into account the fact that tract length changes that initiate as heteroduplex cannot be ascertained by our method until a cell has gone through a subsequent round of DNA replication.) This value signifies that the colony contains ∼37,800 sectors of cells having tract length expansions. Depicting the sectoring within a colony is difficult. We previously used a simple diagram showing a very small number of sectors (Schweitzer and Livingston 1999). Possibly, the best representation of the sectoring might be the visualization of the loss of a centromere plasmid using the ADE3 color assay (supplemental Figure S3 at http://www.genetics.org/supplemental/).
Having so many independent expansions occurring during colony growth makes choosing two examples that have arisen from the same event unlikely. In eight of the nine sets where two or more expansions have been found, no two examples are of the same expansion length. One set is unusual in two respects. First, this set has two expansion lengths that are represented more than once. Second, it has an event where the tract has doubled in size. Possibly, in this set, expansion events occurred early in colony growth, giving rise to clonal expansions of these events. The expansion that doubles the size of the tract may be a coincident event in which two independent expansion events have occurred. Such coincident events should arise in proportion to the square of the rate. We also point out the mean expansion length for the first and second sets of six samples are 15 repeat units and 19 repeat units (excluding the 80-repeat-unit expansion), respectively. Again, the similarity is an indication of the reproducibility.
We conclude that if we can achieve statistical significance from six parental colonies, the results are reliable. We note that it has been known for more than three centuries that among humans there are more male births than female births. The ratio is thought to be ∼106:100 (Davis et al. 1998). This becomes significant (P = 0.023) using a chi-square test when the ratio is 3180:3000. The point is that significance can always be achieved from a small difference in a large sample, but achieving significance with the small sample sizes that we employ is more difficult.
We thank Dmitry Gordenin for the rad27-p allele. We are appreciative of Anja-Katrin Bielinsky, Robin Ricke, and Sapna Das Bradoo for providing the epitope-tagged construct of DNA ligase I and for their help in carrying out the co-immunoprecipitation experiments. We were helped by discussions with Dennis Cook of the Department of Applied Statistics and Robert Herman of the Department of Genetics, Cell Biology and Development at our institution as well as by Patricia Foster at the University of Indiana. Anja-Katrin Bielinsky of our department made many helpful suggestions. We apologize to Louise and Satya Prakash for the hegemony of the yeast community that forced us to change the name of a gene that most of the scientific world calls FEN1. The National Ataxia Foundation and the Minnesota Medical Foundation funded the early parts of these studies. Continued support has been received from the National Institute of Neurological Disorders and Stroke.
Communicating editor: F. Winston
- Received March 17, 2005.
- Accepted July 22, 2005.
- Copyright © 2005 by the Genetics Society of America