Fission yeast exhibits a homothallic life cycle, in which the mating type of the cell mitotically alternates in a highly regulated fashion. Pedigree analysis of dividing cells has shown that only one of the two sister cells switches mating type. It was shown recently that a site‐ and strand‐specific DNA modification at the mat1 locus precedes mating‐type switching. By tracking the fate of mat1 DNA throughout the cell cycle with a PCR assay, we identified a novel DNA intermediate of mating‐type switching in S‐phase. The time and rate of appearance and disappearance of this DNA intermediate are consistent with a model in which mating‐type switching occurs through a replication–recombination coupled pathway. Such a process provides experimental evidence in support of a copy choice recombination model in Schizosaccharomyces pombe mating‐type switching and is reminiscent of the sister chromatid recombination used to complete replication in the presence of certain types of DNA damage.
Mating‐type switching in Schizosaccharomyces pombe involves a process of genetic recombination (Leupold, 1950; Egel, 1977; Egel and Gutz, 1981; Beach, 1983). The three mating‐type loci, mat1, mat2P and mat3M (Beach, 1983; Kelly et al., 1988), are located on the right arm of chromosome II. The loci are separated by the L (15 kb) and K (11 kb) regions (Beach and Klar, 1984). Two homology regions H1 (59 bp, distal) and H2 (135 bp, proximal) are common to all loci, whereas H3 (57 bp, proximal to H2) is present only at the mat2P and mat3M loci (Figure 1A).
Mating‐type switching results from a genetic transfer of the silent loci, mat2P or mat3M, to the active locus, mat1. This transfer is highly regulated by a process called directionality (Figure 1), which restricts switching to the opposite mating type (Egel and Gutz, 1981; Klar and Bonaduce, 1991; Thon and Klar, 1993).
The transcriptionally active mat1 cassette contains either the P or M alleles. Pedigree analysis of mitotic dividing cells has shown that two consecutive asymmetric divisions are required to produce one switched cell among four cousins (Miyata and Miyata, 1981; Egel, 1984). Furthermore, the sister of a switched cell is competent for switching during the next division (Egel and Eie, 1987; Klar, 1987, 1990). In the strand segregation model proposed by Klar (1987), a semi‐heritable chromosomal modification marks, in a strand‐specific manner, one of the two sister chromatids at the first cell division, restricting the mating‐type switching to only one of the two sister chromatids at the second cell division. Recently, it was shown that the mat1 locus does not contain a site‐specific, DNA double strand break (DSB) (Beach, 1983), but instead a fragile chromosomal site (Arcangioli, 1998). It was proposed that the fragility is due to a site‐specific, DNA single strand break (Arcangioli, 1998) or an alkaline‐labile (Dalgaard and Klar, 1999) DNA modification. This DNA modification is located at the junction of the mat1‐specific alleles and the H1 region on the upper strand (Nielsen and Egel, 1989; Arcangioli, 1998), and is responsible for the DSB appearing during standard genomic DNA purification.
Cis‐ and trans‐acting elements have been involved in the marking, or switching, process. Mutations in the SAS1 and SAS2 elements, located 140 and 70 bp away from the modification site, have a reduced DNA modification at mat1 (Arcangioli and Klar, 1991; Klar et al., 1991; Styrkarsdottir et al., 1993) and might be responsible for a particular nucleosomal structure. Three classes of genes (defined as swi genes) when mutated exhibit a reduced rate of mating‐type switching (Egel et al., 1984; Gutz and Schmidt, 1985). Class Ia (swi1, ‐3 and ‐7) mutants might be required for the marking process. This hypothesis is supported by the finding that swi7 encodes the catalytic subunit of the DNA polymerase α (Singh and Klar, 1993) involved in lagging strand DNA synthesis and associated with the primase activity. The latter activity is an excellent candidate in support of alkaline‐labile modification, as a primer RNA left from the previous replication round (Dalgaard and Klar, 1999). Class Ib (swi2, ‐5 and ‐6) mutants have a normal rate of DNA modification at mat1, but switch less frequently. Class II (swi4, ‐8, ‐9 and ‐10 and rad22) mutants also have a normal rate of DNA modification, but produce a high proportion of heterothallic progeny containing extensive DNA rearrangements at the mating‐type loci. Although the asymmetric distribution of the switching potential is well documented, little is known about how asymmetric mating‐type switching leads to only one switched cell among two sisters.
Since the mat1 locus is replicated preferentially from a centromere‐distal origin (Arcangioli, 1998; Dalgaard and Klar, 1999), the leading strand synthesis should be stalled at the DNA modification site, producing an invading strand containing a 3′‐OH. We reveal here a mechanism of gene conversion coupled with DNA replication, producing a switched chromatid and an unswitched chromatid that is marked for switching in the next generation.
To detect mitotic gene conversion intermediates, we devised a PCR assay using yeast genomic DNA, prepared in solid agarose plugs, preserving the integrity of the DNA throughout the purification steps. Figure 1 shows the mating‐type genomic region, two hypothetical chromosome foldings required for switching and the PCR primers used in this work. The PCR products obtained with the P1 and P2 primer pair are shown in Figure 2. Digestion with BamHI or EcoRI restriction enzymes should discriminate between the P or M alleles (Figure 1A). Since the wild‐type homothallic cell population is composed of 50% P and 50% M cell types, each of the enzymes should digest half of the initial fragment, as confirmed in Figure 2A, lanes 2–5. The P1 and P3 primers were devised to detect potential covalent DNA intermediates bridging the mat1‐distal locus with mat2P‐ or mat3M‐proximal loci (Figure 2A, lanes 6–9), while the P1 and P4 primers should detect only a DNA intermediate bridging mat1P‐distal with mat3M‐proximal loci (lanes 10–13). A similar strategy (see Materials and methods) has been used previously (White and Haber, 1990). Five additional amplification cycles were required to reveal the PCR products with the P1–P3 and P1–P4 primer pairs compared with the P1–P2 primer pair, indicating that the starting material is ∼30 times less abundant than the mat1 DNA. Similar results were obtained when the PCR assay was conducted with fewer cycles and labeled primers (data not shown and see below).
In order to relate the DNA intermediates observed with the process of mating‐type switching, we analyzed genomic DNA isolated from two heterothallic (non‐switching) mutant strains by PCR (Figure 3). The mat1M, smt0 strain contains a deletion of 262 bp (see Figure 1), removing the cis‐acting elements essential for the DNA modification at mat1 (Arcangioli and Klar, 1991; Stykarsdottir et al., 1993; Arcangioli, 1998; Dalgaard and Klar, 1999), and the mat1M, Δmat2‐3 strain contains a deletion of the mat2P and mat3M loci (see Figure 1) but still retains the DNA modification at mat1 (Klar and Miglio, 1986). Note that the P1 primer overlaps the smt0 deletion and was therefore replaced by the P′1 primer located 240 bp distal from P1 (Figure 1A). In order to rule out potential artifacts due to abortive DNA amplifications and run‐off products extending from the P′1 primer to the DNA modification site, genomic DNAs from both mutants were mixed and analyzed. P′1–P2 primers amplified the expected DNA fragments (Figure 3, lanes 1–4). The faster migrating PCR product obtained with the smt0 genomic DNA is due to the 262 bp deletion. As expected, the P′1–P3 and P′1–P4 primers amplified a detectable PCR product after five additional cycles (Figure 3A, lanes 5 and 9). However, no product was obtained from both heterothallic mutant DNA templates (Figure 3, lanes 6–7 and 10–11), notably in the mixed reactions (lanes 8 and 12). In Figure 3B, PCR products were analyzed by native polyacrylamide gel electrophoresis. In this experiment, PCR primers were labeled and fewer amplification cycles were required. Under these conditions, the two alleles mat1M and mat1P could be separated, without further enzymatic restriction, the M allele migrating more slowly than the P allele (Figure 3B, lanes 1 and 5). Again no product was obtained with the P′1 and P3 or P4 primers and the DNA of non‐switching mutants. Taken together, these data indicate that PCR products observed with the P′1–P3 and P′1–P4 primer pairs and DNA from the homothallic strain are most likely to reflect the existence of gene conversion DNA intermediates. Since the P3 and P4 primers are located outside the H2 sequences, then the cellular DNA polymerases involved in the gene conversion process do not stop at the homologous H2 sequences of the silent loci, but go beyond the H3 homology box.
It is well established that bulky lesions in DNA cause the replication machinery to stall. However, DNA replication eventually resumes without the repair of the lesion (Friedberg et al., 1995). It is presumed that a second intact DNA template will serve to bypass the lesion. We propose that in S.pombe the opposite silent mating‐type locus provides, in cis, the undamaged template for gene conversion. The DNA break or alkaline‐labile modification is stable during the entire length of the cell cycle, is located on the upper strand, at the junction of mat1 and H1 DNA sequences, and only one of the two sister cells switches its mating type efficiently. Hence, we propose that the replication fork, coming from the mat1‐distal side (Dalgaard and Klar, 1999), produces a 3′‐OH‐ended DNA strand when the leading strand DNA synthesis stalls at the DNA modification site (see Arcangioli, 1998), thereby initiating mating‐type switching (Figure 6, step b). To investigate this possibility further, we analyzed the recombination intermediates throughout the cell cycle. To this end, a homothallic, cdc25 conditional mutant strain (Russel and Nurse, 1986) was synchronized (see Materials and methods), and genomic DNA prepared as a function of time was analyzed by PCR. To ensure quantitative PCR amplification, the P1 primer was 32P labeled and the number of amplification cycles was reduced. As before, three separate amplifications using the same three pairs of primers were carried out, and the PCR products were quantitated and plotted over time (Figure 4A and B). Cell cycle position and degree of synchrony were followed by flow cytometry analysis (Figure 4C) and microscopic observations (data not shown). The amount of P1–P2‐generated products roughly doubled between 60 and 120 min, indicating the timing of mat1 locus replication (Figure 4A), which takes place in S‐phase observed by flow cytometry (Figure 4C). The P1–P3‐ and P1–P4‐generated products were only transitory; an abundant product appeared and peaked at the 120 min time point and almost disappeared in early G2‐phase (Figure 4A and B), before mitosis (data not shown), which began after 220 min. Quantitation of the level of PCR products present at the 120 min time point indicated that the DNA intermediates represent ∼20% of the total mat1 DNA and correlates with the steady‐state level observed in an asynchronous population (see above). Furthermore, 20% is the order of magnitude of the estimated proportion of switching cells in an exponentially growing population.
The disappearance of the switching DNA intermediates, in early G2‐phase, is likely to result from an efficient resolution of the gene conversion intermediates. Eleven genes have been implicated in mating‐type switching. Depending on their functions they were assigned to different classes (Egel et al., 1984). Importantly, none of these mutants abolish mating‐type switching. The Ia and Ib classes exhibit a slow mating‐type switching. The class II mutants are unstable and frequently produced stable heterothallic segregants. It was proposed that the heterothallic mutants originate from errors in the resolution of recombination intermediates during mating‐type switching (Egel et al., 1984). The steady‐state level of the gene conversion intermediate was analyzed in the three classes of switching mutants. As shown in Figure 5, class Ia and Ib swi mutants exhibit a small but significant reduction in the amount of PCR products with the P1–P3 and P1–P4 primer pairs, compared with wild type, with the exclusion of the swi6− mutant. The reduction in DNA intermediates correlates with the reduction of mating‐type switching in these mutants. The behavior of swi6− can be explained by a highly skewed switching ratio in favor of P‐to‐M events (Thon and Klar, 1994) as well as by a low spontaneous frequency of chromosomal rearrangements (Lorentz et al., 1992; see below). In contrast, the swi4− and swi10− mutants exhibit strong signals. A large body of evidence indicates that the class II genes (swi4, ‐8, ‐9 and ‐10 and rad22) are required for processing the gene conversion intermediates (Egel et al., 1984; Gutz and Schmidt, 1985; Fleck et al., 1992; Carr et al., 1994; Paques and Haber, 1997; Rodel et al., 1997). Unfortunately, we cannot distinguish between accumulation of gene conversion intermediates and spontaneous DNA rearrangements frequently occurring in the class II mutants. Furthermore, the swi4− and swi10− mutant strains produce ∼10–20% elongated and dead cells (data not shown). This issue was not analyzed further and will await the construction of conditional swi mutant strains.
In this study, we have revealed the existence of a covalent DNA molecule, bridging the acceptor mat1 locus with the distal donor mat2P or mat3M loci mostly during S‐phase. Such DNA intermediates are present in homothallic wild‐type strains but are reduced or absent in slow or non‐switching heterothallic mutant strains (except for the class II mutants). Taken together, we conclude that this novel DNA species is involved in the gene conversion event responsible for the mating‐type switching during the DNA replication period in S.pombe.
The kinetics of the appearance and disappearance of the gene conversion intermediates are consistent with a replication‐recombination coupled process. In Figure 6, we propose that the leading strand replication complex, coming from the distal side of mat1, is stalled by the DNA modification on the upper strand at the junction of the H1 region and the mat1‐specific allele (Arcangioli, 1998; Dalgaard and Klar, 1999) and provides the 3′‐OH‐ended invading strand (Arcangioli, 1998; Figure 6, step b). The nature of the DNA modification (nick or RNA) at mat1 is an important question, which is not addressed in the present study. However, it is assumed that DNA replication up to a single‐stranded nick will produce a double‐stranded DNA end (Arcangioli, 1998). Alternatively, a double‐stranded DNA end can be produced upon DNA replication arrest (Michel et al., 1997). The superimposition of recombination on replication is reminiscent of Belling's model known as the copy choice process, proposed many years ago as an alternative model for homologous recombination (Belling, 1933), a form of recombination‐dependent DNA replication, somewhat similar to the break‐induced replication mechanism (Mosig, 1987; Kogoma, 1997; Haber, 1999, and references therein).
The 30 min delay between the beginning of mat1 replication and the appearance of the DNA intermediates may reflect the time required for the initial steps of mating‐type switching. These steps include positioning of the opposite mating‐type cassette next to the mat1 locus, strand invasion by the newly synthesized 3′‐OH‐containing strand into the homologous H1 sequences, followed by polymerase assembly and priming (Figure 6, steps b–d). DNA synthesis of the opposite mating type occurs as a migrating D‐loop (Ferguson and Holloman, 1996) (Figure 6, step e) or as a small replication bubble coordinating leading and lagging strand synthesis (Holmes and Haber, 1999). In the absence of silent loci (Klar and Miglio, 1986), it is assumed that another process involving breakage or reorganization of the replication fork will bypass the DNA modification. This may occur by recombination with the newly replicated DNA formed on the opposite strand brought about by the converging replication fork, which lacks the modification. This is a variation of the template‐switching model proposed for trans‐lesion DNA synthesis (Friedberg et al., 1995), which is analogous to the type of sister chromatid exchange observed in bacteria, genetically in yeast and cytologically in mammalian cells (Wolff et al., 1974; Formosa and Alberts, 1986; Kadyk and Hartwell, 1995; Zou and Rothstein, 1997).
The process of mating‐type switching recruits many proteins (at least the products of the swi genes) that are normally required in other pathways. The class Ia genes are most likely to be required in the early steps of switching, preparing mat1 DNA, one generation before gene conversion (Klar, 1987; Singh and Klar, 1996; Dalgaard and Klar, 1999). The class Ib genes seem to facilitate the assembly of the mating‐type region into a higher order chromatin structure (Thon and Klar, 1993). Numerous studies have indicated that class II genes are involved in general DNA repair processes. The swi4 and swi8 genes share strong homology with the MSH3 and MSH2 subgroups, respectively, involved in the mismatch repair pathway (Fleck et al., 1992; Rudolph et al., 1999). The swi9 and swi10 genes share strong homology with the RAD1 and RAD10 genes, respectively, from Saccharomyces cerevisiae involved in the nucleotide excision repair pathway (Schmidt et al., 1989; Schlake et al., 1993; Tomkinson et al., 1993; Bardwell et al., 1994; Carr et al., 1994; Rödel et al., 1997). Finally, rad22 is homologous to RAD52, which is indispensable in the recombinational repair of DSB in S.cerevisiae. However, rad22 phenotypes do not match the severity of rad52, especially with regard to mating‐type switching (Ostermann et al., 1993). A hypothetical intramolecular secondary DNA structure between the H2 and H3 regions has been proposed recently (Rudolph et al., 1999). Our data support the notion that the gene conversion event goes beyond the H2 and H3 sequences in order to allow the formation of this hairpin structure (Figure 6, step f). The generation of this secondary DNA structure might be facilitated by a looping out occurring when the gene conversion process is copying the H2 and H3 region and serves as a target for the swi4/swi8 gene products, constituting a signal for correct termination of DNA synthesis. The resolution of the gene conversion intermediates might be carried out by the endonuclease activity of the products of the swi9/swi10 genes (Figure 6). This last scenario is reminiscent of what is proposed for mating‐type switching resolution in S.cerevisiae (Saparbaev et al., 1996; Sugawara et al., 1997).
Depending on the switching direction and on the efficiency of gene conversion resolution, two types of DNA rearrangements were observed (Beach and Klar, 1984; Fleck et al., 1990). When the cell switches from P to M (Figure 1B) and gene conversion does not terminate at the H2–H3 boundary (from mat3M), the 11 kb between mat2 and mat3 are duplicated in a direct configuration. On the other hand, when the cell switches from M to P (Figure 1C) and gene conversion does not terminate at the H2–H3 boundary (from mat2P), the 15 kb between mat1 and mat2 are also duplicated in a direct configuration, or produce an extrachromosomal mat2:1° circle. The frequency of DNA rearrangements increases in the class II mutants, and strains with up to seven cassettes in the mating‐type region were found (Fleck et al., 1990).
Zou and Rothstein (1997) have shown recently that Holliday junction recombination intermediates (xDNA) are detected in the rDNA repeats in S.cerevisiae, only in S‐phase. The major consequences of this recombination event are the homogenization of the rDNA sequences, variation of the copy number and formation of extrachromosomal rDNA circles (ERCs). The appearance and accumulation of ERCs strongly correlate with aging (Sinclair and Guarente, 1997). Notably, the absence of the FOB1 gene, which participates in the unidirectional replication fork block in the rDNA, decreases the number of ERCs in aging cells (Defossez et al., 1999). The molecular basis of the generation of ERCs is still under investigation (Defossez et al., 1999; Rothstein and Gangloff, 1999). Defects in the general DNA repair processes may increase the frequency of the spontaneous appearance of ERCs in S.cerevisiae, and decrease the efficiency of reintegration into the chromosomal rDNA cluster (Kim and Wang, 1989; Fleck et al., 1990), together leading to the accumulation of ERCs (Rothstein and Gangloff, 1999). It is possible that in both systems, stalled replication forks might be involved in the formation of extrachromosomal DNA circles.
Our replication–recombination model proposes that the mating type of a switching cell does not replicate its sexual locus by a conventional DNA replication process, although, as in S.cerevisiae, all the components necessary for genomic replication might be necessary for mating‐type switching (Holmes and Haber, 1999). For clarity, we did not present the progression of the lagging strand synthesis in Figure 6. Finally, in every generation, ∼20% of the cells delay the replication of the mating‐type locus by 30 min in a cell cycle period that can formally be considered as a late S‐phase (Figure 4B).
Despite growing understanding of asymmetric mating‐type switching, generating only one switched cell among two sister cells, we still do not know how the DNA modification at mat1 is asymmetrically distributed and maintained on only one of the two sister chromatids after each DNA replication cycle and how the cell can maintain two heterologous sequences at mat1 during the G2‐phase. These remarks indicate the intriguing properties of this locus, since at each generation the asymmetry of the two sister chromatids seems to be undetected by the recombination–repair machinery.
Materials and methods
The S.pombe strains used were originally derived from Leupold's culture: SP62 h90 leu1‐32 ura4‐D18; SP714 mat1M mat2,3Δ::LEU2 leu1‐32; SP807 mat1‐Msmt0, leu1‐32 ade6‐210 (Klar and Miglio, 1986; Engelke et al., 1987). SP838 h90 swi1 ade6‐M216; EG371 h90 swi2‐S73 lys1; EG743 h90 swi4‐1 ura4‐Δ18; EG372 swi5‐S39 lys1; PG1 h90 swi6‐115 leu1‐32 ura4‐Δ 18 ade6‐M216; SP112 h90 swi7‐ leu1‐32 ade6‐M216; EG741 h90 swi10::ura4 ura4‐D18; gifts from Genevieve Thon, Richard Egel, Amar Klar, Henning Schmidt and Oliver Fleck.
The homothallic h90 cdc25 strain (Russel and Nurse, 1986; Arcangioli, 1998) was shifted to the restrictive temperature (36°C) for 4 h (arrested at the G2–M transition) and re‐incubated at 24°C, allowing a synchronous entry into the cell cycle. Samples of the cells were harvested at intervals and either treated for FACS analysis or genomic DNA was prepared in agarose plugs and analyzed by the PCR assay in which the P1 primer was 32P labeled. The reactions were analyzed by 6% native polyacrylamide gel electrophoresis, dried and exposed to a phosphor screen (PhosphorImager 445SI) and quantified using ImagequantNT. Flow cytometry analysis was described earlier (Arcangioli, 1998), and the propidium iodide‐stained cells were observed microscopically in parallel in order to distinguish G2 and M phases unambiguously.
Genomic DNA preparation and PCR assay
Genomic DNA was prepared in agarose blocks as described earlier (Arcangioli, 1998), digested by the HindIII restriction enzyme, melted at 95°C and the resulting single‐stranded genomic DNA fragments were extracted and purified using a QIAquick gel extraction kit (Qiagen). The strategy consisting of melting the double‐stranded genomic DNA was used to avoid any preferential shearing of recombination intermediates during DNA manipulation. A similar strategy has been used previously with no success, presumably due to the DNA extraction procedure used (White and Haber, 1990). The PCR assay consisted of a first incubation of 5 min at 95°C, then Taq polymerase (Promega) was added, followed by the cycle steps (1 min at 95°C, 1 min at 52°C and 4 min at 72°C). The PCR products were resolved through agarose gel electrophoresis and stained with ethidium bromide. The primer sequences are: P1, CGAAGCAAATATCTCGTTAGAGG; P′1, ATGTAAAGAGTGGGTAGATGCAT; P2, TTATATGTAGTTTATAATTGTTGTGTC; P3, TTGGCAGCCTCGTAGGCTT; and P4, AGCAACTCCTGATACTTTACA. When P1 or P′1 were 32P end labeled, the PCR products were resolved through a 4% native polyacrylamide gel, fixed, dried and exposed on a phosphor screen and the signals detected by PhosphorImager 445SI (Molecular Dynamics) and quantified using ImagequantNT. All PCR amplification experiments were conducted at least three times, using independent sources of genomic DNA.
We thank G.Thon for DNA sequence information at the mat3 locus and S.Gangloff, G.Langsley, G.‐F.Richard, S.Schaper and M.Yaniv for comments on the manuscript. We are grateful for the careful and insightful comments of the referees, whose suggestions improved the quality of our manuscript. Supported by a grant from the Association pour la Recherche sur le Cancer. R.d.L. is a trainee of the Ministere de l'Éducation National, de la Recherche et de la Technologie.
- Copyright © 2000 European Molecular Biology Organization