Recombination and microsatellite mutation in humans contribute to disorders including cancer and trinucleotide repeat (TNR) disease. TNR expansions in wild-type yeast may arise by flap ligation during lagging-strand replication. Here we show that overexpression of DNA ligase I (CDC9) increases the rates of TNR expansion, of TNR contraction, and of mitotic recombination. Surprisingly, this effect is observed with catalytically inactive forms of Cdc9p protein, but only if they possess a functional PCNA-binding site. Furthermore, in vitro analysis indicates that the interaction of PCNA with Cdc9p and Rad27p (Fen1) is mutually exclusive. Together our genetic and biochemical analysis suggests that, although DNA ligase I seals DNA nicks during replication, repair, and recombination, higher than normal levels can yield genetic instability by disrupting the normal interplay of PCNA with other proteins such as Fen1.

EXPANSION of trinucleotide repeat (TNR) tracts can cause a significant number of hereditary neurological disorders such as Huntington disease (HD), myotonic dystrophy, and fragile X syndrome (reviewed in Cleary and Pearson 2003; Lenzmeier and Freudenreich 2003). Studies in budding yeast have supported a number of different models for the expansion and contraction of TNR tracts including classical “slippage” and recombination (Maurer et al. 1996, 1998; Freudenreich et al. 1997, 1998; Miret et al. 1997, 1998; Cohen et al. 1999; Richard et al. 1999, 2000, 2003; Schweitzer and Livingston 1999; Ireland et al. 2000; Jankowski et al. 2000; Schweitzer et al. 2001; Jankowski and Nag 2002; Callahan et al. 2003; Cleary and Pearson 2003; Lenzmeier and Freudenreich 2003). In all models, the ability of TNR sequences to produce intermediates stabilized by secondary structure formation is thought to be critical to the mutation process (reviewed in Mitas 1997; Cleary and Pearson 2003; Lenzmeier and Freudenreich 2003).

Studies with yeast strains lacking the gene coding for flap endonuclease Fen1 (rad27Δ strains) show enhanced levels of genetic instability that are thought to result from inefficient processing of single-strand DNA flaps formed during lagging-strand DNA replication (Tishkoff et al. 1997). During Okazaki fragment synthesis, the 5′-end of a downstream Okazaki fragment may be made single stranded by strand displacement synthesis as Pol δ extends the adjacent upstream Okazaki fragment. The 5′-flap thus formed would normally be eliminated by a combination of enzymes including Fen1 (Budd et al. 1995; Bae and Seo 2000; Bae et al. 2001; Maga et al. 2001; Ayyagari et al. 2003). The absence of Fen1 could yield DNA duplications by slipped mispairing of the flap followed by repair or replication prior to repair (Tishkoff et al. 1997). Unprocessed flaps could also cause an increase in DNA breaks (Symington 1998) and lead to greater genetic instability (Tishkoff et al. 1997).

This model has been extended to TNR mutations in wild-type cells (Gordenin et al. 1997). Expansion mutations were suggested to arise when a newly formed and an as-yet unprocessed TNR-containing flap forms a secondary structure that makes it possible for its 5′-end to directly ligate to the 3′-end of the upstream Okazaki fragment. Support for this model comes from data showing that single-stranded flaps containing a TNR secondary structure [e.g., (CTG)n] are resistant to the action of Fen1 in vitro (Spiro et al. 1999; Henricksen et al. 2000). A delay in flap processing in vivo could increase the chance that the unprocessed flap could be ligated back into the nascent strand resulting in an expansion mutation (Gordenin et al. 1997). In fact, delayed flap processing in rad27Δ cells results in enhanced rates of TNR expansion and contraction mutations (Freudenreich et al. 1998; Schweitzer and Livingston 1998, 1999; Spiro et al. 1999; Ireland et al. 2000; Rolfsmeier et al. 2000, 2001; Callahan et al. 2003; Liu et al. 2004). An unprocessed flap could also give rise to a double-strand break (DSB) that, during recombinational repair, might lead to expansions or contractions (Gordenin et al. 1997). The idea that unprocessed flap can lead to DSB is strongly supported by the synthetic lethality of rad27rad52 double mutants (Tishkoff et al. 1997; Symington 1998).

With regard to TNR expansion, the flap ligation model implies that competition between removal of the flap and ligation of the flap to the 3′ OH group of the upstream Okazaki fragment before its removal, determines whether a mutation can occur by this mechanism. We tested whether overexpression of DNA ligase I (CDC9 in budding yeast) can influence the competition in favor of flap ligation in a TNR-containing substrate. We found that overexpression of CDC9 increases the expansion rates (∼3.5-fold) when CTG (but not CAG) repeats are in the lagging daughter strand. Surprisingly, we found that CDC9 overexpression does not increase the expansion rate as a result of increasing ligase enzymatic activity, as the model would have predicted. Instead, the observed increase in instability was dependent upon the presence of a functioning proliferating cell nuclear antigen (PCNA) interaction motif in the overproduced Cdc9p. We also found an increase in the rates of TNR contraction mutations and mitotic recombination when Cdc9p was overproduced and this too depended on PCNA interaction rather than on enzymatic activity. We conclude that overproduction of DNA ligase I protein can induce genetic instability, presumably by means of its interaction with PCNA.


Strain construction:

The genotypes of all the strains used in this study are listed in Table 1. All the components for yeast growth media were obtained commercially (Q-Biogene, Irvine, CA). The strains used for TNR expansion and contraction assays as well as for the recombination assays at the LYS2 locus were derived from BY4741 and 4963 (Brachmann et al. 1998). To construct yeast strains for the TNR instability assay, plasmids containing TNR sequences (pBL69, pBL70, pBL144, and pBL145; described later) were digested with Bsu36I (for targeting to the LYS2 locus) and used for the transformation (Fast Track Yeast Transformation kit; Bioworld, Dublin, OH) of BY4741 or 4963. The transformants were selected on media lacking histidine.

View this table:

Yeast strains used in this study

The above strategy was used for the construction of the strains yJS1 (BY4741 with pBL69 integration), yJS2 (BY4741 with pBL70 integration), yJS3 (BY4741 with pBL144 integration), yJS4 (BY4741 with pBL145 integration), yJS5 (4963 with pBL69 integration), yJS6 (4963 with pBL70 integration), yJS7 (4963 with pBL144 integration), and yJS8 (4963 with pBL145 integration). In each case, Southern blotting was used to screen for transformants containing a single integration of the plasmid (Sambrook and Russell 2001).

yJS9, a host strain used for recombinational cloning of CDC9 plasmids, was obtained by sporulating a diploid derived by crossing PS807 (Ura Cdc9) and OAy619 (Leu Cdc9+). The two strains were a kind gift from Oscar Aparicio. A spore colony that was temperature sensitive (at 30°) on YPD and unable to grow on media lacking leucine or uracil was chosen to be yJS9 (Ura Leu Cdc9).

The strains JB3/P68 and HFY870-12A used for recombination assays at the HIS3 locus (Klein 1988) and LEU2 locus, respectively, were a kind gift from Hannah Klein. LP2915-8D (ATCC no. 208560) used for CDC9 complementation assays was obtained commercially (ATCC, Manassas, VA).

TNR-containing plasmids:

The TNR-containing plasmids pBL69, pBL70, pBL144, and pBL145 were a kind gift from Bob Lahue. The construction of these plasmids has been described previously (Miret et al. 1998; Rolfsmeier et al. 2001). Briefly, these are integration plasmids based on vector pRS303 (Sikorski and Hieter 1989) containing a HIS3 selectable marker. In addition, they also contain a 1141-bp portion of the LYS2 gene (coordinates 750–1891, relative to the coding sequence) and a URA3 gene under the control of the Schizosaccharomyces pombe adh1 promoter. The TNR sequence (CTG/CAG)25 is located at an SphI site between the adh1 promoter and the URA3 coding region. The integration of the TNR sequence at the SphI site resulted in the duplication of the site, leading to the formation of an out-of-frame ATG codon upstream of the initiator ATG codon of the URA3 reading frame. The TNR sequence is oriented in such a way that pBL69 and pBL145 contain CAG repeats in the same strand as the transcribed strand of URA3 (orientation I) whereas pBL70 and pBL144 contain CTG repeats in that position (orientation II). The plasmid pBL145 differs from pBL69 in having a randomized 24-bp C, A, G sequence upstream of the TNR sequence. Similarly, pBL144 differs from pBL70 in having a randomized 24-bp C, A, G sequence downstream of the TNR sequence. In all of these plasmids, the URA3 gene is placed in such a way that its direction of transcription will be the same as the LYS2 gene after integration.

CDC9-containing plasmids:

All plasmids containing CDC9 (except the ones used for protein purification) were derived from the plasmid pTW268 (a kind gift from Mike Lieber). The construction of pTW268 has been described previously (Wilson et al. 1997). Briefly, pTW268 is a CEN-ARS plasmid derived from the vector pRS316 (Sikorski and Hieter 1989) containing a URA3 marker. Between the KpnI-NotI sites of this plasmid is a Saccharomyces cerevisiae ADH1 promoter driving the expression of a GST-CDC9 coding sequence that is followed by the 3′-UTR of HDF1 gene (KpnI-ADH1-HindIII-GST-BamHI-CDC9-SalI-3′-UTR-NotI). The coding sequences of all CDC9 deletion or point mutation constructs described below were confirmed by DNA sequence analysis. Also, all the enzymes necessary for cloning and PCR were obtained from Promega (Madison, WI).

Deletion of the mitochondrial targeting sequence of CDC9:

CDC9 in pTW268 was amplified (35 cycles of 30 sec at 94°, 45 sec at 59°, and 3 min at 72°) with a mixture of Pfu (1 unit) and Taq (1.5 unit) polymerases using primers oJS3 (TACGTAGGATCCGCATGTCATCCTCATTACCTTC; underlined sequence is a BamHI site) and oJS2 (ATAGCCGTCGACGAGAAAAATAGTGTG; underlined sequence is a SalI site). The PCR product was cloned in pTW268 in place of the CDC9 by digesting both the vector and the PCR product with BamHI and SalI. This gave rise to the plasmid pTW268-GST-CDC9-NΔ23 (containing GST-CDC9-NΔ23 and URA3). This CDC9 derivative lacks the mitochondrial targeting sequence [MTS; amino acids 3–23 of Cdc9p in our GST-tagged version of Cdc9p (Willer et al. 1999)].

Deletion of the PCNA-interacting motif of CDC9:

We deleted amino acids 3–60 of Cdc9p using a similar PCR strategy (described above). pTW268-GST-CDC9-NΔ60 (GST-CDC9-NΔ60, URA3) was constructed by replacing CDC9 in pTW268 with a PCR product lacking the first 180 nt of the CDC9 ORF obtained with a pTW268 template and primers oJS1 (TACGTAGGATCCGCAAATCATCCAAACATATGCT; underlined sequence is a BamHI site) and oJS2.

Deletion of the GST tag in CDC9-NΔ23:

CDC9-NΔ23 lacking the GST tag was amplified from pTW268-GST-CDC9-NΔ23 using primers oJS35 (TACGTAAATAAGCTTCTCACCATGTCATCCTCATTACCTTC; underlined sequence is a HindIII site) and oJS2 (same PCR condition as above). This PCR product was used to replace the HindIII-SalI fragment containing GST-CDC9-NΔ23 in pTW268-GST-CDC9-NΔ23 to make pTW268-CDC9-NΔ23 (CDC9-NΔ23, URA3).

Site-specific mutagenesis of CDC9:

We followed the QuikChange Site-Directed Mutagenesis protocol (Stratagene, La Jolla, CA) for creating point mutations in the CDC9 coding region. The plasmids created using this protocol and the primers and DNA template used for making these plasmids are as follows:

  • pTW268-GST-CDC9-K419A (GST-CDC9-K419A, URA3): Primers oJS4 (GGCGAAACTTTTACGTCAGAATACGCTTACGATGGTGAAAGG) and oJS5 (reverse complement of oJS4) with the plasmid template pTW268.

  • pTW268-GST-CDC9-K598A (GST-CDC9-K598A, URA3): Primers oJS12 (GGTCAAGAAACTGGTTGAAATTAGCTAAAGATTACTTGGAGGGGG) and oJS13 (reverse complement of oJS12) with the plasmid template pTW268.

  • pTW268-GST-CDC9-FF44,45AA (GST-CDC9-FF44,45AA, URA3): First, primers oJS8 (GCCACTTTGGCTAGAGCTTTCACTTCCATGAAAAATAAGCC) and oJS9 (reverse complement of oJS8) were used with template pTW268 to make pTW268-GST-CDC9-F44A (GST-CDC9-F44A, URA3). Second, primers oJS10 (GCCACTTTGGCTAGAGCTGCTACTTCCATGAAAAATAAGCC) and oJS11 (reverse complement of oJS10) were used with template pTW268-GST-CDC9-F44A to create pTW268-GST-CDC9-FF44,45AA.

Recombinational cloning of LEU2 in pTW268 and pTW268-GST-CDC9-NΔ60:

pGST-CDC9 (GST-CDC9, LEU2) was created through recombinational cloning by cotransforming yJS9 (cdc9-1) with pTW268 (URA3) and the HindIII-digested fragment of pULA (contains LEU2 flanked by partial URA3 sequences; Gottschling et al. 1990). Transformants were selected for growth at restrictive temperature of 30° (Leu+ Ura Cdc9+ strain). Plasmids extracted from these transformants were used to transform Escherichia coli. Correct clones were identified by restriction digestion of the plasmids extracted from E. coli. pGST-CDC9-NΔ60 (GST-CDC9-NΔ60, LEU2) was also created using the same strategy except that the plasmid pTW268-GST-CDC9-NΔ60 was used in the place of pTW268.

CDC9 ORF cloning in LEU2-containing plasmids:

pTW268-GST-CDC9-K419A, pTW268-GST-CDC9-NΔ23, pTW268-GST-CDC9-K598A, and pTW268-GST-CDC9-FF44,45AA were cut with BamHI and NotI. The fragments containing these mutated versions of CDC9 were used to replace the BamHI-NotI fragment containing the CDC9 gene in pGST-CDC9 to give rise to plasmids pGST-CDC9-K419A (GST-CDC9-K419A, LEU2), pGST-CDC9-NΔ23 (GST-CDC9-NΔ23, LEU2), pGST-CDC9-K598A (GST-CDC9-K598A, LEU2), and pGST-CDC9-FF44,45AA (GST-CDC9-FF44,45AA, LEU2), respectively. pCDC9-NΔ23 (CDC9-NΔ23, LEU2) was created by cloning the HindIII-NotI fragment containing CDC9-NΔ23, obtained from pTW268-CDC9-NΔ23, in the place of GST-CDC9-NΔ23 in pGST-CDC9-NΔ23.

CDC9 plasmids used for protein purification:

GST-CDC9 was constructed by PCR amplification of yeast genomic DNA with primers 5′-AAGGATCCATGCGCAGATTACTGACC-3′ and 5′-CGGTCGACTAATTTTGCATGTGGGAT-3′ and cloning it in-frame downstream of the GST open reading frame at BamHI and SalI sites in pGSTag (Ron and Dressler 1992). GST–CDC9-FF44,45AA was generated by site-directed mutagenesis using QuikChange Site-directed mutagenesis kit (Stratagene) and was similarly cloned into pGSTag. The plasmids were verified by DNA sequencing before overexpression in E. coli.

TNR expansion and contraction assays:

For expansion assays, RAD27 (yJS1, yJS2) or rad27Δ (yJS5, yJS6) strains with an integrated (CTG/CAG)25 tract were transformed with a control plasmid (pRS415; Sikorski and Hieter 1989) or with plasmids expressing one of the various alleles of CDC9 (pGST-CDC9 series or pCDC9-NΔ23). Similarly, for contraction assays, RAD27 (yJS3, yJS4) or rad27Δ (yJS7, yJS8) strains were transformed with the same plasmids used for expansion assays. In both cases, the transformants were selected on media lacking histidine and leucine.

The transformants were resuspended in water and plated on media lacking histidine and leucine. The plates were incubated at 30° for 3 days (for both expansion and contraction assays). Nine colonies were picked up from these plates and resuspended in water. Appropriate dilutions were plated on media lacking histidine and leucine (control for expansions and contractions) and on media lacking histidine and leucine but containing 5-FOA (selective for expansions) or on media lacking uracil and leucine (selective for contractions). Plates were incubated for 3–4 days at 30°. Colonies were counted and the rate of instability was calculated using the method of the median (Lea and Coulson 1948). Software for calculating mutation rate was provided by Hannah Klien and modified by Pete Calabrese. At least three independent original transformants were tested for each strain and in each test; nine colonies from that strain were examined.

Recombination assays:

RAD27 or rad27Δ strains transformed with a control plasmid (pRS415) or the plasmids expressing one of the various alleles of CDC9 (pGST-CDC9 series or pCDC9-NΔ23) were selected on media lacking histidine and leucine (for gene conversion and pop-out recombination assays at the LYS2 locus), on media lacking leucine and tryptophan (for gene conversion assays at the HIS3 locus), or on media lacking leucine (for pop-out recombination assays at the LEU2 locus). The transformants to be used for the gene conversion assays were replated on the same media used previously. For pop-out recombination assays, the transformants were replated on media lacking leucine (for the assay at the LYS2 locus) or YPD (for the assay at the LEU2 locus).

For each trial, nine individual colonies from these plates were analyzed. Appropriate dilutions of each colony were plated on a control media [media lacking histidine and leucine (gene conversion assay at the LYS2 locus), media lacking leucine (pop-out assay at the LYS2 locus), media lacking leucine and tryptophan (gene conversion assay at the HIS3 locus), or YPD (pop-out assay at the LEU2 locus)] and a selective media [media lacking histidine, lysine, leucine (gene conversion assay at the LYS2 locus), media lacking lysine and leucine (pop-out assay at the LYS2 locus), media lacking leucine, tryptophan, and histidine (gene conversion assay at the HIS3 locus), and media containing 5-FOA (pop-out assay at the LEU2 locus)].

Plates were incubated for 3–4 days. Cells were always grown at 30°. Colonies were counted and the rate of recombination was calculated using the method of the median (Lea and Coulson 1948).

Measurement of CTG/CAG repeat size:

Single colonies containing TNR tracts were resuspended in 100 μl water, heated at 96° for 5 min, and cooled in ice for 5 min. For PCR 10 μl was used (35 cycles of 1.5 min at 93°, 2 min at 55°, and 3 min at 72°) with primers oBL91 (5′-D4Pa-AAACTCGGTTTGACGCCTCCCATG) and oBL157 (AGCAACAGGACTAGGATGAGTAGC) and Taq polymerase (1 unit; Miret et al. 1998). Fragment size analysis using 0.1–1 μl PCR product was carried out on a Beckman CEQ-8000 instrument and analyzed with CEQ-8000 fragment analysis software (Beckman, Fullerton, CA).

Proteins and antibodies:

The purification of Rad27p with an N-terminal poly (histidine) tag (his-Rad27p) has been described previously (Tseng and Tomkinson 2004). In short, E. coli cell extract containing his- tagged Rad27p was incubated with nickel-nitrilotriacetic acid agarose beads. his-Rad27p was eluted with lysis buffer containing 250 mm imidazole and further purified by Resource S and Resource Q chromatography. GST and GST fused to both wild-type Cdc9p and the FF44,45AA mutant version (GST-Cdc9p-FF44,45AA) were purified from E. coli cell extracts by affinity chromatography using glutathione-sepharose beads as described (Levin et al. 2004) with elution using reduced glutathione. A modified version of the purification scheme described by Bauer and Burgers (1988) was used to purify wild-type yeast PCNA (Pol30). Briefly, extracts of E. coli M15 cells harboring pQE16-PCNA were fractionated using Q sepharose (Amersham Biosciences, Piscataway, NJ), hydroxyapatite (Bio-Rad, Hercules, CA), and Mono Q and Superdex 200 (Amersham Biosciences) chromatography.

Polyclonal antibody against Cdc9p has been previously described (Tomkinson et al. 1992; Ramos et al. 1997). Polyclonal antibodies against Fen1 and yeast PCNA were kindly provided by Paul Modrich and Satya Prakash, respectively. A monoclonal anti-α-tubulin antibody (Piperno and Fuller 1985) was kindly provided by Susan Forsburg. Stabilized goat anti-rabbit HRP conjugate secondary antibody and an anti-mouse HRP conjugate secondary antibody were obtained commercially (Amersham Biosciences).

Pull-down assays:

To prepare GST, GST-Cdc9p, or GST-Cdc9p-FF44,45AA mutant beads, 10 μg of each purified protein was incubated with a 50-μl slurry of glutathione sepharose beads (Amersham Biosciences) equilibrated in binding buffer (50 mm KH2PO4, pH 7.5, 10% glycerol, 100 mm NaCl, 0.2% NP-40, 1 mm β-mercaptoethanol, 10 mm imidazole, and 1 mg/ml BSA) for 30 min at 4° with constant agitation. After washing with binding buffer, the beads were resuspended in 100 μl binding buffer. GST, GST-Cdc9p, or GST-Cdc9p-FF44,45AA mutant beads (5 pmol of each protein) were incubated with 5 pmol of PCNA and then incubated at room temperature for 30 min with constant agitation. Beads were collected by centrifugation and then washed extensively in binding buffer prior to being resuspended in 10 μl of SDS-PAGE sample buffer. After separation by SDS-PAGE, proteins were detected by immunoblotting (see below).

To prepare Rad27p beads, 50 μg of purified his-Rad27p was incubated with a 50-μl slurry of Ni-nitrilotriacetic acid agarose beads (Qiagen, Valencia, CA) equilibrated in binding buffer (50 mm KH2PO4, pH 7.5, 10% glycerol, 100 mm NaCl, 0.2% NP-40, 1 mm β-mercaptoethanol, 10 mm imidazole, and 1 mg/ml BSA) for 30 min at 4° with constant agitation. After washing with binding buffer, the beads were resuspended in 100 μl binding buffer. GST-Cdc9p or GST-Cdc9p-FF44,45AA mutant protein (5 pmol) was incubated with either 10 μl Ni-nitrilotriacetic acid agarose beads alone or 10 μl of his-Rad27p beads (5 pmol of his-Rad27p) in presence and absence of PCNA (5 pmol) and then incubated at room temperature for 30 min with constant agitation. Beads were collected by centrifugation and then washed extensively in binding buffer prior to being resuspended in 10 μl of SDS-PAGE sample buffer. After separation by SDS-PAGE at 200 V, proteins were detected by immunoblotting. Western transfer was done at 350 mÅ constant current for 1 hr. Anti-Cdc9p and anti-hFen1 were each used at 1:5000 dilutions. Anti-yeast PCNA was used at 1:1000 dilution.

Yeast cell extract preparation:

A single colony of BY4741 transformed with pRS415 or one of the CDC9 overexpressing plasmids was grown in 100 ml of media lacking leucine to a OD600 of ∼0.5. Cells were centrifuged, washed, and resuspended in 1 ml of 1× lysis buffer [50 mm Tris-Cl (7.5), 5 mm EDTA (8.0), 0.1% Triton X-100, 300 mm KCl, 10% glycerol] with freshly added DTT (2 mm) and protease inhibitors (1 mm PMSF, 1 μg/ml pepstatin, 1 mm benzamidine, 1 μg/ml leupeptin, and 5 μg/ml TPCK; Sigma-Aldrich, St. Louis). After adding an equal volume of glass beads, cells were vortexed six times at high speed for 30-sec intervals spaced by 90 sec in ice and finally centrifuged at 1500 RPM for 3 min at 4°. The supernatant was centrifuged again at 13,000 RPM for 15 min and protein concentration measured using a Bio-Rad protein assay kit I.

Immunoblotting of Cdc9p from cell extracts:

Approximately 30 μg of extracted proteins was separated on 6% SDS-PAGE at a constant current of 25 mÅ and transferred to a PVDF membrane (Millipore, Billerica, MA) using a wet transfer apparatus for 3 hr at a constant current of 200 mÅ at 4°. After transfer, the membrane was cut into two parts; one part was used for the detection of Cdc9p and the other for α-tubulin detection. Detection of Cdc9p was carried out using anti-Cdc9p polyclonal antibody used at a 1:2000 dilution. Stabilized goat anti-Rabbit HRP conjugate secondary antibody was used at a 1:10,000 dilution. For α-tubulin, a monoclonal anti-α-tubulin antibody was used at a 1:500 dilution and an anti-mouse HRP conjugate secondary antibody was used at a 1:10,000 dilution. In both cases, a chemiluminescent substrate (Super signal West Femto maximum sensitivity substrate, Pierce Chemical, Rockford, IL) was used to detect the bands. The image is acquired using a ChemiDoc-XRS system (Bio-Rad) and quantified using Quantity One software (Bio-Rad).


Analysis of TNR instability:

We integrated (CTG/CAG)25 at the LYS2 locus of RAD27 (wild type) and rad27Δ strains in two orientations and measured the rates of TNR expansions and contractions, as described previously (Miret et al. 1998; Rolfsmeier et al. 2001). In this system, the URA3 gene is expressed from the S. pombe adh1 promoter. For expansion assays, a (CTG/CAG)25-ATG sequence is located between the promoter and the reading frame of the URA3 gene. Transcription initiates downstream of the (CTG/CAG)25-ATG sequence and therefore, translation would start from the authentic ATG codon in the URA3 transcript, producing a functional protein. These cells cannot grow in a 5-FOA-containing media. However, an expansion of five TNRs or greater would increase the distance between the original transcription initiation site and the promoter enough so that transcription is initiated from an ATG codon that is out-of-frame with respect to the original ATG codon of the URA3 gene. Translation from the new ATG codon would produce a nonfunctional protein and therefore, the cells would be ura and grow in the presence of 5-FOA.

The contraction assay uses a construct containing a 24-bp random sequence immediately downstream or upstream of the TNR sequence [(CTG/CAG)25(C, T, G/C, A, G)8ATG or (C, A, G/C, T, G)8 (CAG/CTG)25ATG]. This is similar to an expansion size of eight repeats and cells cannot grow on media lacking uracil. If there is a contraction of five or more repeats, the transcript will not contain the out-of-frame ATG codon, the URA3 gene will be expressed, and cells can grow in the absence of uracil in the media (Ura+).

On the basis of the direction of DNA replication at the LYS2 locus (Freudenreich et al. 1997), the lagging-strand template contains CAG repeats in orientation I and CTG repeats in orientation II. Our wild-type tester strains transformed with pRS415 (empty vector; first row, Table 2) gave an expansion rate ∼360-fold higher in orientation I (yJS1) than in orientation II (yJS2). Contractions were ∼90-fold more common in orientation II (yJS3) than in orientation I (yJS4). Overall, the results are similar to previous studies (Maurer et al. 1996; Freudenreich et al. 1997; Miret et al. 1998; Rolfsmeier et al. 2001; Dixon and Lahue 2004) with the exception of a few that failed to find an orientation effect for contractions. In some of these cases, this might be because the repeats are alternatively replicated from two different origins of DNA replication present on either side of the repeat sequence (Miret et al. 1997; Rolfsmeier et al. 2001).

View this table:

TNR expansion and contraction rates in wild-type (BY4741 derivatives) cells

When rad27Δ tester strains carrying a TNR tract in orientation I (yJS5) were studied, we observed a rate 28-fold higher (4.0 ± 1.1 × 10−3/cell division) than that in yJS1. In the opposite orientation, the rad27Δ strain (yJS6) had a rate ∼177-fold higher (6.9 ± 2.8 × 10−5/cell division) than that in yJS2. When contractions in rad27Δ strains were measured in orientation I (yJS8) we observed a rate 6.7-fold higher (1.4 ± 0.2 × 10−5/cell division) than that in yJS4. In the opposite orientation (yJS7) we found a rate 6.8-fold higher (1.3 ± 0.5 × 10−3/cell division) than that in yJS3. This is consistent with previous studies that also show rad27Δ strains exhibit higher levels of TNR instability than wild-type cells (Freudenreich et al. 1998; Schweitzer and Livingston 1998, 1999; Spiro et al. 1999; Ireland et al. 2000; Rolfsmeier et al. 2001; Callahan et al. 2003). We next measured the effect of CDC9 overexpression on TNR expansion mutation in both orientations in wild-type strains to test whether favoring flap ligation over removal would increase TNR expansion rates. Our CEN-ARS plasmid overexpression system uses the S. cerevisiae ADH1 promoter to drive the expression of the CDC9 gene fused with an upstream GST tag. The CDC9 gene product is normally targeted either to the mitochondria or to the nucleus depending on the translational start site (Willer et al. 1999). Both the full-length GST-CDC9 and CDC9 lacking its mitochondrial targeting sequence (GST-CDC9-NΔ23) were able to complement the temperature sensitivity of a cdc9-2 mutant (LP2915-8D; Figure 1A). Note that when GST-CDC9 (or any of its variants) is overexpressed in any of our yeast strains, the endogenous CDC9 copy is always present.

Figure 1.—

Characterization of CDC9 overexpression. (A) Complementation of a cdc9-2 mutant was performed by transforming LP2915-8D with a control vector (pRS316) or plasmids (pTW268 series; URA3) overexpressing different CDC9 alleles listed on the left. The transformants were resuspended in water and 10-fold dilutions were spotted on YPD at 30° (left) and 35° (right). Spot tests used three different dilutions (highest concentration on the left). (B) Western blot of Cdc9p. Cell extracts were obtained from BY4741 transformed with a control vector (pRS415; lane 1) or a plasmid overexpressing CDC9-NΔ23 (lane 2), GST-CDC9 (lane 3), GST-CDC9-NΔ23 (lane 4), GST-CDC9-NΔ60 (lane 5), GST-CDC9-K419A (lane 6), GST-CDC9-K598A (lane 7), or GST-CDC9-FF44,45AA (lane 8). Following 6% SDS-PAGE and transfer to a PVDF membrane, the membrane was cut and the proteins were probed with either a polyclonal anti-Cdc9p antiserum or a monoclonal anti-α-tubulin antibody. The top and middle rows represent the overexpressed ∼118-kDa GST-Cdc9p and the ∼85-kDa endogenous Cdc9p probed with the anti-Cdc9p antibody, respectively. The bottom row represents the ∼50-kDa α-tubulin protein (loading control). The numbers in parentheses represent the fold difference in the volume (intensity × square millimeters) of overexpressed GST-Cdc9p or its mutant versions compared to that of endogenous Cdc9p within the same lane. The endogenous Cdc9p in lane 2 cannot be distinguished from the overexpressed protein because they have the same molecular weight. (C) Purified PCNA (5 pmol of PCNA trimer) was incubated with glutathione sepharose beads liganded by 5 pmol of GST, GST-Cdc9p, or GST-Cdc9p-FF44,45AA. Proteins bound to GST beads (lane 2), GST-Cdc9p beads (lane 3), and GST-Cdc9p-FF44,45AA beads (lane 4) were separated by SDS-PAGE. GST-Cdc9p (top) and PCNA (bottom) were detected by immunblotting with antibodies to Cdc9p and yeast PCNA, respectively. Lane 1 represents 10% of the input PCNA used in the pull-down assay.

When GST-CDC9 was overexpressed (level of expression 21-fold higher than that of the endogenous CDC9, Figure 1B) in wild-type cells containing the TNR tract in orientation I (yJS1), we repeatedly observed a 2.5- to 3.5-fold increase in the expansion rate compared to the control (yJS1 transformed with pRS415; Table 2). A similar increase was also seen when GST-CDC9-NΔ23 was overexpressed. Interestingly, we saw no effect of GST-CDC9 overexpression on expansions with the repeats in orientation II (yJS2).

To confirm that the effect of GST-CDC9 overexpression (higher expansion rate in orientation I) was due to increased ligase activity (and therefore, increased flap ligation) we overproduced GST-Cdc9p carrying mutations at the active site. Several critical amino acids have been identified to be necessary for ligase activity (Tomkinson et al. 1991; Sriskanda et al. 1999). Lys-419 is charged by ATP in the first step of ligation and Lys-598 is critical to the last step. We made the two different active site mutants (K419A and K598A), both carrying alanine in the place of lysine. As expected, neither allowed LP2915-8D to grow at the nonpermissive temperature (Figure 1A).

Unexpectedly, overexpressing either ligase activity mutant in wild-type cells (yJS1) did not lower the TNR expansion rate to control levels (Table 2). In the most unstable orientation for expansion (I), the mutation rates for GST-CDC9-K419A and GST-CDC9-K598A were 3.2 and 3.8 times that observed for the control, close to that seen when GST-CDC9 is overexpressed (3.7 times greater; Table 2). We conclude that the increase in TNR expansions that accompanies CDC9 overexpression is a consequence of a property of DNA ligase I that is not related to its enzymatic activity but, possibly, to its interaction with other proteins.

We next asked whether GST-CDC9 overexpression also increases TNR contraction rates in wild-type cells. When GST-CDC9 was overexpressed, a ∼22-fold increase in the rate of contraction in orientation I (yJS4) and a ∼10-fold increase in orientation II (yJS3) compared to the control levels (both strains transformed with pRS415) was observed (Table 2). The increase in contraction rates is also not due to ligase enzymatic activity because the contraction rates obtained with GST-CDC9-K419A and/or GST-CDC9-K598A in either orientation (yJS3 or yJS4) were not significantly different from the rates obtained with GST-CDC9.

CDC9 overexpression increases TNR instability by interacting with PCNA:

DNA ligase I is one of many proteins that have been shown to interact with PCNA (Levin et al. 1997; Jonsson et al. 1998; Montecucco et al. 1998; Tom et al. 2001; Ho et al. 2002) by means of a conserved binding motif (QxxI/L/MxxFF/Y, where x is any residue) that it shares with other PCNA-binding proteins (reviewed in Maga and Hubscher 2003). Genetic and biochemical studies have demonstrated the biological and functional significance of the PCNA binding motif at the N terminus of human DNA ligase I (Levin et al. 1997; Montecucco et al. 1998; Rossi et al. 1999; Levin et al. 2000; Tom et al. 2001). Studies on human DNA ligase I also showed that the adjacent phenylalanine residues in the conserved motif are necessary for ligase-PCNA interaction (Montecucco et al. 1998). In Cdc9p, the yeast homolog of DNA ligase I, the consensus PCNA binding motif is located within the polypeptide (residues 38–45) rather than at the N terminus. We examined the interaction between yeast Cdc9p and yeast PCNA by replacing the adjacent phenylalanine residues in the conserved PCNA binding motif of Cdc9p with alanine residues. In binding and Western blot experiments, beads containing GST-Cdc9p (lane 3; Figure 1C) but not GST-Cdc9p-FF44,45AA (lane 4; Figure 1C) specifically retained PCNA, confirming that the two adjacent phenylalanine residues at positions 44 and 45 of Cdc9p are essential for Cdc9p-PCNA interaction in yeast. Two additional experiments also gave the same results (data not shown).

If altering Cdc9p-PCNA interaction were responsible for the increase in TNR mutation rate then overexpression of CDC9 lacking the PCNA binding motif should bring the mutation rates back to the control level. We used a deletion (GST-CDC9-NΔ60) and a site-specific (GST-CDC9-FF44,45AA) mutant version of the conserved PCNA interaction motif of CDC9. In both TNR orientations in wild-type cells, overexpression of either of the PCNA interaction mutants gave the same expansion (yJS1 and yJS2) and contraction rates (yJS3 and yJS4) as control cells (four strains transformed with pRS415; Table 2). We were not surprised that both CDC9 constructs were able to complement the temperature sensitivity of LP2915-8D when overexpressed (Figure 1A) since a PCNA-DNA ligase I interaction is not required for viability in yeast (Sriskanda et al. 1999; Martin and Macneill 2004). Increasing the level of DNA ligase I enzymatic activity, by overexpressing mutant versions of GST-CDC9 that lack PCNA binding, does not enhance expansion rates. Thus, ligation does not appear to be the rate-limiting step in TNR expansions generated by flap ligation in wild-type cells. Similarly, TNR expansion rates (2.5 ± 0.7 × 10−3/cell division in yJS5) and contraction rates (1.7 ± 0.2 × 10−3/cell division in yJS7) in rad27Δ strains overexpressing mutant PCNA binding (but enzymatically active) versions of GST-Cdc9p are not significantly different from the rates obtained with the two rad27Δ strains transformed with pRS415 (P = 0.1 for expansions and P = 0.2 for contractions).

Since both the GST and PCNA interaction motifs are located at the N terminus of the protein, we also asked whether the increase in instability associated with PCNA interaction is influenced by the presence of GST. The new construct (CDC9-NΔ23; lacking GST and the MTS) was able to complement the temperature sensitivity of LP2915-8D (data not shown). When tested for its effect on TNR expansion (in yJS1) and contraction rates (in yJS3), the same levels of increase seen with overexpression of GST-CDC9 (or GST-CDC9-NΔ23) were observed (Table 2).

Finally, we used a single-colony PCR assay to test whether overexpression of Cdc9p with a functional PCNA binding motif (PBM-OE colonies; cells transformed with pGST-CDC9, pGST-CDC9-NΔ23, pGST-CDC9-K419A, or pGST-CDC9-K598A) changes the size distribution of TNR expansion compared to cells expressing normal levels of the PCNA binding site (PBM-WT colonies; cells transformed with pRS415, pGST-CDC9-NΔ60, or pGST-CDC9-FF44,45AA). For expansions, we tested a total of 220 yJS1 colonies with the TNR in orientation I (the most unstable orientation for expansions). A total of 92% of PBM-OE colonies (110/119 colonies) and 96% of PBM-WT colonies (97/101 colonies) had a real expansion mutation. The mean expansion size for PBM-OE colonies (10.3 ± 3.5 repeats) and PBM-WT colonies (10.3 ± 2.9 repeats) is not significantly different (P = 0.95; t-test, Figure 2A). For contractions, we analyzed a total of 161 yJS3 colonies (82 PBM-OE colonies and 79 PBM-WT colonies) with TNR in orientation II (most unstable orientation for contractions). All had TNR contractions. The mean contraction size for PBM-OE colonies (17 ± 2 repeats) and PBM-WT colonies (17.3 ± 2 repeats) were not statistically different (P = 0.33; t-test, Figure 2B). This result is similar to previously published values for contraction size distribution of (CTG)25 repeats (Dixon and Lahue 2004).

Figure 2.—

TNR size distributions estimated by single colony PCR analysis. (A) yJS1 colonies (with or without overexpression of CDC9) growing on media containing 5-FOA, but lacking histidine and leucine (expansion colonies), were used for PCR analysis. Data obtained from yJS1 colonies transformed with pRS415 or a plasmid overexpressing GST-CDC9-NΔ60 or GST-CDC9-FF44,45AA (collectively called PBM-WT colonies; shaded bars) were pooled and compared with the pooled data obtained from colonies transformed with a plasmid overexpressing GST-CDC9, GST-CDC9-NΔ23, GST-CDC9-K419A, or GST-CDC9-K598A (collectively called PBM-OE colonies; solid bars). The x-axis represents the number of repeats added to the original (CTG/CAG)25 and the y-axis represents the percentage of colonies carrying the expanded repeat size. For each of the two groups, PCR data from ∼100 individual colonies were pooled. (B) yJS3 colonies (with or without overexpression of CDC9) growing on media lacking leucine and uracil (contraction colonies) were used for PCR analysis. Data obtained from yJS3 colonies transformed with pRS415 or a plasmid overexpressing GST-CDC9-NΔ60 or GST-CDC9-FF44,45AA were pooled (PBM-WT colonies; shaded bars) and compared with the data pooled from colonies transformed with a plasmid overexpressing GST-CDC9, GST-CDC9-NΔ23, GST-CDC9-K419A, or GST-CDC9-K598A (PBM-OE colonies; solid bars). The x-axis represents the number of repeats deleted from the original (CTG/CAG)25 and the y-axis represents the percentage of colonies carrying the contracted repeat size. For each of the two groups, PCR data from ∼80 individual colonies were pooled.

Cdc9p and Rad27p compete for PCNA interaction in vitro:

Since both Cdc9p and Rad27p contain a PCNA binding motif, it is possible that both of these proteins can bind to the same PCNA homotrimer or their interaction with PCNA could be mutually exclusive. A mutually exclusive interaction of PCNA with Rad27p and Cdc9p could be a basis for the instability observed in CDC9 overexpressing cells because an excess of Cdc9p could decrease the steady-state levels of Rad27p-PCNA complex. Although a mutually exclusive interaction has been predicted for human ligase I and Fen1 on the basis of the X-ray structure of human ligase I complexed with DNA (Pascal et al. 2004), no direct data are available in yeast. To test this idea in yeast, we performed pull-down assays with Rad27p beads and both PCNA and GST-Cdc9p (Figure 3). As expected, PCNA was specifically retained by the Rad27p beads (lanes 5 and 7) but no binding of GST-Cdc9p was observed either in the absence or in the presence of PCNA (lanes 6 or 7). The same results were found in two additional experiments (data not shown). The failure to detect PCNA-dependent binding of GST-Cdc9p to Rad27p beads in our experiments and PCNA-dependent binding of Rad27p to Cdc9p beads (data not shown) is consistent with the formation of mutually exclusive PCNA-Rad27p and PCNA-Cdc9p complexes. Studies with the functionally homologous human proteins provide further support for this model. In co-immunoprecipitation experiments with PCNA, untagged DNA ligase I, and his-tagged FEN1, we did not observe either PCNA-dependent immunoprecipitation of FEN1 by DNA ligase I antibody or PCNA-dependent immunoprecipitation of DNA ligase I by FEN1 antibody (J. Varkey and A. E. Tomkinson, unpublished results). A prediction of this competitive binding model is that increasing the amount of one PCNA binding partner should reduce the amount of the other partner complexed with PCNA. In accord with this model, the addition of fivefold more GST-Cdc9p reduced the amount of PCNA bound to Rad27p beads (compare lanes 7 and 8) but only when the PCNA binding site of GST-Cdc9p was functional (compare lanes 12 and 13).

Figure 3.—

Analysis of complexes formed by Rad27p, Cdc9p, and PCNA. To detect complexes formed among Rad27p, Cdc9p, and PCNA, we performed pull-down assays with Rad27p beads (referred to as B-R) as described in materials and methods. The presence (+) or absence (−) of any component (listed on the left) is indicated above. Lanes 1, 2, and 9 show purified PCNA, GST-Cdc9p, and GST-Cdc9p-FF44,45AA (GST-Cdc9p-M) proteins, respectively. These lanes contain 10% of the input proteins used in the pull-down assays. PCNA (5 pmol trimer) was incubated with Rad27p beads (lane 5). GST-Cdc9p or GST-Cdc9p-FF44,45AA (5 pmol of each) was incubated with Rad27p beads either alone (lanes, 6 and 11) or with PCNA (lanes 7 and 12, 5 pmol of trimer) as indicated. Similarly, a 5× molar excess (25 pmol) of either GST-Cdc9p (lane 8) or GST-Cdc9p-M (lane 13) was incubated with Rad27p beads in the presence of PCNA (5 pmol trimer). To control for background binding, PCNA (lane 3), GST-Cdc9p (lane 4), and GST-Cdc9p-FF44,45AA (lane 10) were each incubated with Ni-NTA agarose beads (B). Proteins eluted from the beads were separated by SDS-PAGE and detected by immunoblotting with Cdc9p (top), Rad27p (middle), and PCNA antibodies (bottom).

CDC9 overexpression increases mitotic recombination:

TNR expansions could arise by both flap ligation and recombination but TNR contractions can arise only by the latter pathway (Gordenin et al. 1997). Since CDC9 overexpression affects TNR contractions more than expansions, it is possible that CDC9 overexpression may have increased TNR instability by increasing recombination rates, possibly as a result of altering the dynamics of the Cdc9p-PCNA interaction. If true, this altered interaction might be expected to affect recombination at different sites in the genome. We used a pop-out recombination and a gene conversion assay at several loci in the genome to study the effect of CDC9 overexpression on mitotic recombination. A DSB between direct repeats can lead to either pop-out recombination or gene conversion. Pop-out recombination (reviewed in Paques and Haber 1999) results in the deletion of the intervening sequence and one of the repeats. The first recombination substrate tested contains duplicate lys2 segments (Figure 4A). One of the duplicates has a ∼2.2-kb deletion at the 3′-end of LYS2 (lys2-3′Δ) and the other one has a deletion of ∼0.75 kb at the 5′-end (lys2-5′Δ) leaving ∼1.2 kb of identical sequence in each copy. Between the duplicates lie HIS3 and URA3 with the (CTG)25 repeat (and other plasmid sequences). Both lys2-3′Δ and lys2-5′Δ are not functional and therefore, cells carrying this construct are Lys His+ Ura+. A pop-out recombination event between the duplicates will result in the production of a functional LYS2 gene and the deletion of the intervening sequences, allowing the cells to grow in the absence of lysine. The phenotype of these cells will be Lys+ His Ura. Pop-out recombination frequency is the ratio of the number of colonies that are Lys+ His Ura to the total number of colonies.

Figure 4.—

Recombination substrates. A detailed description of the constructs used for recombination assays can be found in the text. All constructs are represented with boxes and lines (not drawn to scale). Each box represents a gene listed on top. Similar boxes in the rest of the figure represent the same gene. The solid box indicates the location of the mutation. The broken lines represent the intervening sequence in the construct. (A) Substrate used for recombination assays at the LYS2 locus (not drawn to scale) contains a duplicated lys2 segment separated by sequences that include the HIS3 and URA3 genes. Pop-out recombination (left pathway) would result in the deletion of the intervening sequences and the formation of a functional LYS2 gene. Gene conversion (right pathway) would also produce a functional LYS2 gene without the loss of intervening sequence. (B) The pop-out recombination substrate used at the LEU2 locus contains a duplicate leu2-k allele, each with the same mutation (loss of a KpnI site). Therefore, pop-out recombination between the duplicates would still yield a mutated leu2-k allele. However, the ADE2 and URA3 genes present in the sequence between the two leu2-k fragments will be lost. (C) Recombination substrate used at the HIS3 locus for measuring gene conversion involves duplicate his3 fragments (nonfunctional). Gene conversion can generate a HIS3 fragment that does not carry both the mutations and, therefore, a functional gene.

When GST-CDC9 was overexpressed in yJS1, a sevenfold increase in the rate of pop-out recombination compared to the control was detected (yJS1 transformed with pRS415; Table 3). Among a sample of 100 colonies that grew in the absence of lysine, none were able to grow on media lacking histidine, indicating that the colonies were true pop-outs and not the result of gene conversion. The increased rate of pop-out recombination due to GST-CDC9 overexpression was about one-third the rate of a rad27Δ strain (yJS8 transformed with pRS415; 1.1 ± 0.1 × 10−2 recombination events/cell division).

View this table:

Recombination rates

The requirement for the increase in the rate of pop-out recombination was the same as that for increasing TNR mutation: a functional PCNA-interaction motif. Ligase enzymatic activity was not required (Table 3). Therefore, the increased recombination rate is not a result of more efficient completion of pop-out recombination by ligation. Also, single-strand annealing, the mechanism by which pop-out recombination occurs, is known to be nearly 100% efficient when the homology between recombining sequences is >400 bp (Sugawara and Haber 1992; Sugawara et al. 2000). Therefore, the approximately sevenfold increase in pop-out recombination is likely a result of increased initiation (possibly strand breaks) rather than enhanced completion of the process. Finally, the observed increase in pop-out recombination was not related to the presence of GST since CDC9-NΔ23 gave the same results as GST-CDC9 (Table 3).

We also tested the rate of pop-out recombination at the LEU2 locus in a different strain background containing a different construct without a TNR tract. In this strain (HFY870-12A) duplicate segments of leu2-k (LEU2 carrying a 7-bp deletion at the KpnI site; see Klein 1988) is separated by a sequence that includes URA3 and ADE2 genes (Figure 4B). The phenotype of this strain is Leu Ura+ Ade+. A pop-out recombination between the leu2-k duplicates would result in the loss of URA3 and ADE2 in addition to one of the leu2-k duplicates. Therefore, these cells will be Leu Ura Ade. Ura cells can be selected on a 5-FOA-containing media and the true pop-outs would be red in color due to the loss of ADE2 gene. Overexpression of GST-CDC9 enhanced this recombination rate at this locus by ∼22-fold above the level of the control (HFY-870-12A transformed with pRS415; Table 3).

Using the same construct and strain (yJS1) used for the pop-out recombination assay at the LYS2 locus, the rate of gene conversion between the lys2 segments when CDC9 is overexpressed was also measured. Gene conversion events between lys2-3′Δ and lys2-5′Δ segments can also result in a functional LYS2 gene (Figure 4A). Gene conversion is not accompanied by the deletion of the intervening sequence (reviewed in Paques and Haber 1999). The progeny of cells that had undergone gene conversion at this locus will be Lys+ His+ Ura+. To avoid pop-outs from growing, we did not add histidine to the media. We also confirmed the “conversion” status of ∼100 Lys+ His+ Ura+ colonies by testing for the presence of intervening sequence between the LYS2 duplicates using PCR.

The conversion rate was 292-fold higher in GST-CDC9 overexpressing cells than in the control (yJS1 transformed with pRS415; Table 3). Mutating the Cdc9p active site (GST-CDC9-K419A) gave the same result. Interestingly, this gene conversion rate is ∼9-fold higher than that obtained with our rad27Δ strain (yJS8 transformed with pRS415; 2.9 ± 1.3 × 10−6 recombination events/cell division) carrying the same recombination substrate. Mutating the function of the PCNA binding site using GST-CDC9-FF44,45AA reduced gene conversion from 292- to ∼18-fold above the control cells rather than to the same level as the control cells (Table 3). The role of GST in influencing gene conversion was examined by comparing GST-CDC9 and CDC9-NΔ23 conversion rates. In the absence of GST, overexpression of ligase (CDC9-NΔ23) was associated with a 7.4-fold increase compared to the control strain, not the 292-fold increase observed when GST was present (GST-CDC9, Table 3). Thus, a portion of the increase in gene conversion due to overexpression of GST-CDC9 appears due to an as-yet undetermined effect of GST when present at the amino-terminal end of Cdc9p.

The effect of GST-CDC9 overexpression on gene conversion was also studied in a different chromosomal context and in the absence of a TNR sequence in a third strain background, JB3/P68. In this system, a direct repeat of two different HIS3 alleles separated by a sequence that includes the TRP1 gene was used (Figure 4C; Klein 1988). One of the HIS3 alleles, his3-513, carries a loss of a KpnI site and the other allele, his3-537, is deleted for the most distal HindIII site. These mutations are 316 bp apart. Neither of the his3 alleles is functional and the cells carrying this construct are His Trp+. A gene conversion event between these two alleles would produce a functional HIS3 gene without the loss of the TRP1 gene. Thus, the phenotype of these cells would be His+ Trp+. Gene conversion frequency can be measured as a ratio of the number of His+ cells to the total number of cells.

In contrast to our result at LYS2, the increase in gene conversion with GST-CDC9 overexpression was a modest 5.5-fold compared to the control levels (JB3/P68 transformed with pRS415; Table 3). Note that the background level of gene conversion is ∼100-fold higher in this his3 than that in our lys2 system. Mutating the PCNA interaction site but not the Cdc9p active site made a small but statistically significant difference to the observed increase. Taken together, all the recombination results suggest that the genetic instability induced by DNA ligase I-PCNA interaction is not restricted to a specific genetic locus or construct.


Our results show that overproduction of Cdc9p leads to TNR instability and increased recombination. TNR instability appears to result only from the Cdc9p-PCNA interaction and not from any other interactions mediated by the remainder of Cdc9p polypeptide chain including the presence of GST. Only the rate of gene conversion (but not pop-out recombination) at LYS2 was partially affected by GST at the N terminus of Cdc9p. The Cdc9p-PCNA interaction may lead to instability by different pathways. Since PCNA helps to coordinate Okazaki fragment maturation by interacting with a consensus sequence for binding present on DNA polymerase δ, Fen1, and DNA ligase I proteins (reviewed in Maga and Hubscher 2003), we hypothesize that the overexpression of CDC9 could disrupt the steady-state levels of the PCNA-Cdc9p and PCNA-Rad27p complexes by a competitive mechanism (Figure 5A). In support of this hypothesis, we have provided in vitro evidence that the interactions of Cdc9p and Rad27p with PCNA are mutually exclusive. Studies of the complexes formed by the functionally homologous human proteins also support this model. Although three FEN1 molecules can bind to a PCNA trimer (Sakurai et al. 2005), the complexes formed between human DNA ligase I and PCNA on DNA contain only one DNA ligase I molecule per PCNA trimer (Levin et al. 1997), suggesting that the binding of DNA ligase I to one of the interdomain connector loop sites on the PCNA trimer occludes the other two sites. Finally, the size and shape of the C-shaped clamp formed by the catalytic domain of DNA ligase I bound to nicked DNA provides a molecular explanation for the blocking of the interdomain connector loop sites when the DNA ligase and PCNA clamps are tethered together (Pascal et al. 2004).

Figure 5.—

Model for TNR instability in wild-type yeast with and without CDC9 overexpression. (A) TNR instability induced by CDC9 overexpression. Once a flap is formed, overproduced DNA ligase I protein may out-compete Fen1 for binding to PCNA (1), which could increase the chance of TNR secondary structure formation (2) and lead to expansions by flap ligation (3). If Fen1 were able to cut the flap, mutation would be prevented (4, 5). Alternatively, delayed flap processing could lead to DSB formation (6) and a TNR contraction by recombination (7). Note that once the DSB is formed and the 5′-ends resected, the intermediate is identical regardless of repeat orientation (see text). The TNR sequence in the template strand is indicated by a broken line. Arrowheads mark the 3′-end of a nascent strand. The open ovals represent PCNA, the solid circles Fen1, and open circles DNA ligase I. (B) Orientation dependence of TNR instability in wild-type yeast (without CDC9 overexpression). This figure illustrates replication fork movement through a TNR sequence. Okazaki fragments (numbered) move counterclockwise as the replication fork advances to the left. (Path a) CTG flaps are formed in the lagging daughter strand and, following secondary structure formation, can cause expansions by flap ligation. (Path b) In the opposite orientation, secondary structure formation (represented as a U in the broken line) by CTG repeats in the single-stranded region (intermediate steps not shown) of the lagging-strand template leads to a DNA gap that can be repaired by recombination after a DNA break on the other strand and can produce a contraction mutation. Arrowheads correspond to the 3′-end of the daughter strands.

If this is true for Cdc9p, whose catalytic activity domain is similar in size to human DNA ligase I (Barnes et al. 1990), then an excess of Cdc9p competing for PCNA binding might disturb the equilibrium level of the DNA-bound PCNA-Rad27p complex resulting in delayed or inefficient flap processing. Inefficiently processed flaps could lead to increased instability by promoting DSB formation followed by repair resulting in increased gene conversion, pop-out recombination, and TNR expansion and contraction mutations. DNA breaks initiated by inefficient flap processing are thought to be responsible for increased recombination rates in rad27 null mutants (Symington 1998; Tishkoff et al. 1997). Alternatively, direct ligation of an unprocessed flap would uniquely lead to TNR expansion but not contraction mutations. A second possibility is that CDC9 overexpression could alter the steady-state levels of the complex between PCNA and DNA pol δ. Studies on human proteins have shown an inhibitory effect of very high levels of DNA ligase I on DNA pol δ activity in vitro (Levin et al. 1997; Mackenney et al. 1997; Mossi et al. 1998). Also, interaction of DNA ligase I and DNA pol δ with PCNA has been shown to be mutually exclusive (Riva et al. 2004). A consequence of perturbing the steady-state levels of the PCNA-DNA pol δ complex could be stalled replication forks, a precursor to genetic instability leading to increased recombination activity (reviewed in Michel et al. 2001). Finally, we cannot exclude the possibility that excess DNA ligase I protein levels could decrease the pool of PCNA unbound to DNA or block the interaction of PCNA with other PCNA binding proteins.

It should be noted that Cdc9p is not the only PCNA-interacting protein that induces genetic instability when overproduced. In fact, overproduction of Rad27p (only if its PCNA binding domain is present) using a GAL1 promoter in a 2μ plasmid, results in a 20-fold higher rate of +1 poly(A) frameshift mutations and 2.4-fold higher rates of interchromosomal recombination than those in wild type (Greene et al. 1999).

A model for TNR instability in yeast:

rad27Δ strains exhibit increased TNR contractions and are also hyperrecombinogenic. This led to the idea that the two processes were causally related (Gordenin et al. 1997) and it has been frequently proposed that DSB repair could lead to TNR mutation (Gordenin et al. 1997; Richard et al. 1999; Jakupciak and Wells 2000; Jankowski et al. 2000; Richard et al. 2000; Bzymek and Lovett 2001; Richard et al. 2003; Hebert et al. 2004). We have shown (see Table 2) that in wild-type cells overexpressing CDC9, the contraction rates (relative to wild type without CDC9 overexpression) are enhanced in addition to having increased rates of mitotic recombination.

In both these hyperrecombinogenic strains, the TNR contraction rates are increased above the control levels to the same extent regardless of the orientations of the TNR tracts. This lack of an orientation effect supports recombination as the source of contractions because, no matter what the proximate cause of a DSB, the repair intermediates after resection will be the same [one (CTG)n and one (CAG)n single-strand tail, each containing a 3′ OH] regardless of the orientation of the triplet repeats. In both these hyperrecombinogenic strains, the strand breaks leading to increased contraction may have resulted from inefficient flap processing (Figure 5A, steps 6 and 7). We note that in the case of both the rad27Δ strains and CDC9 overexpression in wild-type cells, flap processing is abnormal although due to two different reasons; in one case the protein is absent whereas in the other the protein is in competition with Cdc9p.

Do contractions in nonhyperrecombinogenic wild-type strains also result from DSBs initiated by abnormal flap processing? Since these strains have normal levels of Rad27p, abnormal flap processing would result only if the flap forms a secondary structure. The orientation that would have a lagging daughter strand flap with a CTG strand (capable of forming a more stable secondary structure than a CAG strand; reviewed in Mitas 1997 but see also Gacy et al. 1995) should also be the orientation that has the highest contraction frequency. However, most of the existing data on RAD27 strains (and wild-type strains not overexpressing CDC9; this work) suggest that the contraction rate is greatest when CAG flaps form on the lagging daughter strand (Maurer et al. 1996, 1998; Freudenreich et al. 1997; Schweitzer and Livingston 1997, 1998, 1999; Ireland et al. 2000). Thus, it would appear that abnormal flap processing cannot fully explain the formation of DSB.

What then is the proximate cause of DSB formation? A more stable secondary structure formed by CTG rather than CAG repeats on the lagging-strand template could prevent that region from being replicated, resulting in single-strand interruptions (Figure 5B) that may eventually lead to DSBs (Tishkoff et al. 1997; Kuzminov 2001).

If recombination were responsible for expansions, it is difficult to explain why the rad27Δ strain showed greater enhancement of the expansion rate compared to wild type than did wild-type strains overexpressing (GST-CDC9 or CDC9-NΔ23). Note that with respect to contraction rates, the enhancement seen in all three strains was comparable. In wild-type cells with CDC9 overexpression, we favor flap ligation (see Figure 5A, steps 2 and 3) over recombination as the major cause of expansion since the orientation that increases the expansion mutation rate places the CTG rather than the CAG tract on the lagging daughter strand making a more stable secondary structure possible (reviewed in Mitas 1997 but also see Gacy et al. 1995).

Since increased levels of only DNA ligase I enzymatic activity (pGST-CDC9-NΔ60 and pGST-CDC9-FF44,45AA) in either wild-type or rad27Δ cells do not increase the rate of TNR expansions, we conclude that ligation is not the rate-limiting step in the flap ligation pathway. Some other step in the flap ligation process (e.g., secondary structure formation) would have to be rate limiting. Note that the flap ligation model (Gordenin et al. 1997) for TNR expansion mutation and its dependence on orientation is similar to the classical DNA slippage explanation, but thermodynamically more attractive, since strand displacement synthesis, rather than spontaneous “breathing” at the 3′-end of the nascent strand followed by reannealing upstream, leads to the synthesis of extra repeats.

Our model for TNR expansions and contractions in wild-type cells differs in its details from previous models in several ways. In the case of contractions, we propose that DSBs initiated from an unreplicated single-stranded region containing a secondary structure in the lagging-strand template may be the major source of contractions (Figure 5B, path b). This is in contrast to the model where a DSB forms at the site of an unprocessed DNA flap created by strand displacement synthesis during lagging-strand replication (Gordenin et al. 1997).

With regard to TNR expansions in wild-type cells, we favor the idea that secondary structure formation by TNR sequences in the flaps created during lagging-strand synthesis (Gordenin et al. 1997) is resolved preferentially by flap ligation rather than by recombination as discussed above.

Variation in DNA ligase I expression:

Null mutants of DNA ligase I gene in yeast and mice are lethal (Bentley et al. 1996; Giaever et al. 2002). Conditional inactivating mutations of CDC9 are known to be hyperrecombinogenic and have increased TNR mutation rates (Game et al. 1979; Schweitzer and Livingston 1998) and mice homozygous for a partially inactivating DNA ligase I mutation derived from a human cancer cell line show increased cancer susceptibility (Harrison et al. 2002). That overproduction of Cdc9p, with or without enzyme activity, can increase the rate of not only TNR instability, but also pop-out recombination and gene conversion, emphasizes the importance of balancing the levels of interacting proteins at the replication fork for maintaining genomic stability.

The fact that moderately higher levels of DNA ligase I (∼21-fold) could lead to genetic instability is interesting regardless of the actual mechanism by which it occurs. It has been reported that DNA ligase I protein levels increase sixfold in human fibroblasts after UV irradiation (Montecucco et al. 1995). The extent of variation in DNA ligase I levels among individual human cells in response to other stimuli is not known nor is the minimum amount of overexpression required to affect genetic instability. Expression of an individual gene has been shown to be variable from cell to cell within clonal populations of normal yeast and E. coli, presumably, because of stochastic effects on gene expression (Elowitz et al. 2002; Blake et al. 2003). This raises the possibility that in a mammalian cell population, cells with higher levels of DNA ligase I (or possibly other proteins containing PCNA binding motifs) may also have higher levels of genetic instability.


We thank Dmitry Gordenin and Richard Kolodner for their helpful comments on the manuscript and Bob Lahue, Mike Lieber, Hannah Klein, Rolf Sternglanz, and Richard Pelletier for providing plasmids and/or advice. We are especially grateful for Oscar Aparicio's assistance throughout this project. This study was supported in part by grants GM57479 (to A.E.T.) and GM36745 (to N.A.) from the National Institutes of Health.


  • Communicating editor: M. Lichten

  • Received March 2, 2005.
  • Accepted May 19, 2005.


View Abstract