A Unique DNA Recombination Mechanism of the Mating/Cell-type Switching of Fission Yeasts: a Review

Abstract

Cells of the highly diverged Schizosaccharomyces (S.) pombe and S. japonicus fission yeasts exist in one of two sex/mating types, called P (for plus) or M (for minus), specified by which allele, M or P, resides at mat1. The fission yeasts have evolved an elegant mechanism for switching P or M information at mat1 by a programmed DNA recombination event with a copy of one of the two silent mating-type genes residing nearby in the genome. The switching process is highly cell-cycle and generation dependent such that only one of four grandchildren of a cell switches mating type. Extensive studies of fission yeast established the natural DNA strand chirality at the mat1 locus as the primary basis of asymmetric cell division. The asymmetry results from a unique site- and strand-specific epigenetic “imprint” at mat1 installed in one of the two chromatids during DNA replication. The imprint is inherited by one daughter cell, maintained for one cell cycle, and is then used for initiating recombination during mat1 replication in the following cell cycle. This mechanism of cell-type switching is considered to be unique to these two organisms, but determining the operation of such a mechanism in other organisms has not been possible for technical reasons. This review summarizes recent exciting developments in the understanding of mating-type switching in fission yeasts and extends these observations to suggest how such a DNA strand-based epigenetic mechanism of cellular differentiation could also operate in diploid organisms.

INTRODUCTION

Cells of the highly diverged Schizosaccharomyces (S.) pombe and S. japonicus fission yeasts exist in one of the two sex/mating types, called P (for plus) or M (for minus), specified by which allele, M or P, resides at mat1 (Fig. 1). The fission yeasts have evolved an elegant mechanism for switching P or M information at mat1 by a programmed DNA recombination event with a copy of one of the two silent mating-type genes residing nearby in the genome. The switching process is highly cell-cycle and generation dependent such that only one of four grandchildren of a cell switches mating type, and switching occurs in nearly half the cells of a population. Such a change of cell type is analogous to the stem-cell division found in higher eukaryotes whereby sister cells differ in their fate. Extensive studies of fission yeast established the natural DNA strand chirality at the mat1 locus as the primary basis of asymmetric cell division. This asymmetry results from a unique site- and strand-specific epigenetic “imprint” at mat1 installed in one of the two chromatids during DNA replication. The imprint is inherited by only one daughter cell, maintained for one cell cycle, and then used for initiating recombination during mat1 replication. The progression through two replication cycles and two cell divisions leads to the “one-in-four” switching proportion among granddaughter cells. This mechanism of cell-type switching is considered to be unique to these two organisms, but determining the operation of such a mechanism in other organisms has not been possible for technical reasons. Thus, the validity of this mechanism for development in general remains untested. This review summarizes recent exciting developments in understanding the mechanism of mat1 switching in fission yeasts and extends these observations to suggest how such a DNA strand-based mechanism of cellular differentiation could also operate in diploid organisms. Although the analogous cell-type switching found in the Saccharomyces (Sa.) cerevisiae budding yeast by HO-endonuclease cleavage of the mating-type locus appears superficially similar to that of fission yeast, the mechanistic details are very different in these organisms. Thus, studies with diverse single-celled model yeast organisms have been helpful to appreciate how different paradigms of cellular differentiation have evolved. Considering that the ultimate basis of cellular differentiation in yeast is the double-helical structure of DNA, it is likely that such a mechanism operates in higher eukaryotes as well.

FIGURE 1.

The mating-type region and mat1-switching process. The mating-type region comprises three genetic cassettes located on chromosome 2 at the indicated distances from each other in the order shown. The P-specific region, existing in both mat1-P and mat2-P loci, is 1,113-bp long and the M-specific sequence in mat1-M and mat3-M is 1,127-bp long (6). Recombination is facilitated by short homology boxes, which flank all cassettes. The cassette proximal homology box H1 is 135-bp long; the distal H2 element is 59-bp long. mat1 switching occurs when genetic information copied from the mat2 or mat3 locus is unidirectionally transmitted to the mat1 locus (horizontal arrows), where it displaces the previously existing mat1 allele. A unique epigenetic imprint (star) found at mat1 initiates the recombination leading to mat1 switching. Whereas the mat1 locus is transcriptionally active and dictates cell type, the mat2-K-mat3 region is transcriptionally silenced by another kind of epigenetic mechanism (reviewed in reference 5). Thus, the cell type is defined only by the mat1 locus even though genetic information for both mating types resides elsewhere in the chromosome. doi:10.1128/microbiolspec .MDNA3-0003-2014.f1

FISSION YEAST IS A LOWER EUKARYOTE IDEAL FOR BIOLOGICAL RESEARCH

S. pombe displays structural features that define eukaryotic organisms in that it contains a nucleus and mitochondria; however, unlike most other diploid eukaryotes, it exists primarily as a haploid, rod-shaped single cell. The cells are ~7 to 14-μm long and ~3 to 4-μm wide, nearly three times the size of Escherichia coli cells. The genome consists of three chromosomes with ~14.0 megabases of DNA, also about three times the size of the E. coli genome. Mitotically growing yeast cells elongate at the tips and divide by fission at the middle of the cell, and hence the name “fission yeast.” Many biological processes, such as cell cycle, gametogenesis, meiosis, cellular differentiation, and cell biology, are evolutionarily conserved in eukaryotes. Yeasts are superb eukaryotic models, providing a well-established demonstration of classical Mendelian genetics as well as being an inexpensive and abundant source of material for biochemical analyses. For inclusive descriptions of earlier studies on yeast genetics, meiosis, and mating-type switching, previous review articles are available (1, 2, 3, 4, 5). This review primarily summarizes research on the cellular differentiation of yeast cells and highlights more recent advances for the mechanisms of DNA recombination employed for cellular differentiation, genome integrity, and gene silencing.

HAPLOID CELLS CONSIST OF P OR M CELL TYPE

The yeast cell exists as one of two sex/cell types, defined by which cassette of genetic information, P or M, resides at the mat1 locus (1). These mat1 alleles are composed of nonhomologous sequences of about 1.1 kilobases (kb) and each allele encodes two divergently transcribed transcripts (6). As these transcripts are induced by nutritional starvation, especially for nitrogen, cells do not express their mating type while growing in a rich medium. When mixed on a solid starvation (sporulation) medium, adjoining cells of opposite mating type fuse to produce a zygotic cell. Usually, the zygotic cell proceeds through conventional nuclear fusion and meiosis followed by sporulation, in which four haploid meiotic spore “segregants” are produced in an ascus, a zygotic cell sac. Mutational and molecular studies have demonstrated that the specific mat1 transcripts are required for mating, meiosis, and sporulation (6). Two of the haploid spore segregants, or “ascospores,” in each ascus are of the P cell type and the other two are of the M cell type. This 2:2 meiotic segregation pattern in each ascus establishes that a single locus with two alleles genetically specifies the cell’s mating type (7).

In rare instances, the zygotic cell fails to sporulate, and remains in a diploid condition. Growing the culture in a rich medium that prohibits sporulation may perpetuate the diploid state, which is useful for genetic analyses. Once the diploid culture is transferred to a sporulation medium, cells with the mat1-P/mat1-M constitution proceed through meiosis and sporulation without further mating. As both mat1 alleles are necessary for meiosis and sporulation, diploid cells homozygous for mating type (mat1-P/mat1-P and mat1-M/mat1-M) do not sporulate, but they can do so once they have become heterozygous because of mat1 switching (described in the next section) or have mated with each other and then produce diploid ascospores (1).

The ascospores synthesize a starchy compound but mitotically growing cells do not. This property of yeast has provided a simple and powerful diagnostic tool to differentiate colonies that contain spores from those that do not: those with spores stain black when exposed to iodine vapors and nonsporulating colonies that lack starch fail to stain (8) (Fig. 2). This feature has become the single most useful tool in studies of the phenomenon of mat1 switching and in the search for cis- and trans-acting factors that play a role in the switching process.

FIGURE 2.

Staining phenotype of patches of cells growing on sporulation medium with iodine vapors. The patches are exposed to iodine vapors for about two minutes. Cell patches (or colonies, not shown) of the wild-type (swi⁺) strain are composed of a mixture of P and M cells that engage in efficient mating and sporulation. Such patches stain black because they contain asci with spores that synthesize a starch-like compound that is stained by iodine vapors. The patch of swi1⁻ cells fails to stain, indicating their switching defect. doi:10.1128 /microbiolspec.MDNA3-0003-2014.f2

SWITCHING OCCURS BY THE STEM-CELL-LIKE ASYMMETRIC CELL DIVISIONS

A remarkable phenomenon exhibited by yeast cells is that the mating cell type is inherently unstable such that the two cell types readily interconvert by a phenomenon known as homothallism (9). Yeast populations that are capable of switching are designated h⁹⁰ (for homothallic, 90% sporulation), but those incapable of efficient switching are called heterothallic. Heterothallic cells express very stable P (h⁺) or M (h⁻) cell types. Genetic and molecular analyses of heterothallic strains were fundamental in defining the structure of the mating-type locus and the mechanism of its switching (10). These studies showed that a heterothallic cell results from rare (~1 × 10⁻⁵) and spontaneous mitotic recombination events that occur at the mat1 locus. As the origin of the heterothallic state has been described in several review articles (1, 4, 5, 11) and as no new research has been conducted in this area recently, this aspect of mating-type research is not addressed here. Rather, this review highlights recent progress in mat1 switching that reveals a mechanism of DNA recombination that is unusual in many respects compared to the canonical double-stranded DNA break repair mechanism operating elsewhere in this yeast and other eukaryotes.

Owing to the efficient mat1-switching process, a colony originating from a single cell of either mating type becomes a mixture of P and M cells in a 1:1 ratio (12). Usually, daughter/sister cells generated from a single cell can mate while growing on a poor growth medium, indicating that one of the sister cells has switched to the opposite mating type. Miyata and Miyata (13) observed an unusual pattern of switching in cell pedigrees. They found one zygote among four grandchildren of a cell in about 72 to 94% lineages and two zygotes were never observed. This pattern is now known as Miyata’s one-in-four granddaughters switching rule and derives from the following switching restrictions imposed in cell pedigrees: (i) an unswitchable cell (for example, Pu, u for unswitchable [Fig. 3]) in ~80% of cell divisions produces one switchable daughter (Ps cell) while the sister cell is always unswitchable Pu; (ii) the Ps cell produces one switched Mu cell and one Ps cell in ~80% of cell divisions; and (iii) the Ps cell switches productively to the opposite mating type in ~80% of cell divisions. As a result of the last feature, chains of recurrent switching can be found by cell pedigree studies (14, 15). The same cell-lineage restrictions apply when cells switch from M to P. Thus, two consecutive asymmetric cell divisions are required to conform to the one-in-four pattern of switching (Fig. 3). This asymmetric cell division is very much analogous to the stem-cell asymmetric division reported in many eukaryotic cell divisions in which one of the two daughter cells changes fate, whereas the other maintains the fate exhibited by the parental cell. Remarkably, cellular differentiation decisions require two generations and this feature might be useful to help de-fine mechanisms of cellular differentiation in multicellular organisms, such as vertebrates.

FIGURE 3.

The one-in-four granddaughters switching pattern of yeast cells. The star indicates a cell that inherits the mat1 imprint from the parent cell, making it switching competent so that it will produce one of the daughters that is switched. The same pattern is observed when cells switch from M to P. Pu, unswitchable cell; Ps, switchable cell. doi:10.1128/microbiolspec .MDNA3-0003-2014.f3

SWITCHING OCCURS VIA mat1 ALLELE REPLACEMENT BY A PROGRAMMED RECOMBINATION MECHANISM

As individual cells can interconvert between the two cell types and mat1 alleles contain different DNA sequence information, genetic information for both mat1 alleles must reside in the genome (Fig. 1). However, only one of the two alleles occupies the mat1 locus at a given time in a haploid cell. Before the mat1 genes were molecularly cloned, classical genetic studies guided research for determining the mechanism of mat1 switching. Perhaps influenced by the promoter flip-flop mechanism of genetic switches first discovered in prokaryotic organisms, interconversion of yeast mat1 was initially proposed to be mediated by inversion of a shared promoter between the two mating-type genes of opposite allelic information residing in chromosome 2 at the mat1 locus by recombination (16). However, after the mating-type gene replacement mechanism of MAT switching in the unrelated budding yeast Sa. cerevisiae was discovered (17, 18) (reviewed in reference 19), genetic analysis of two mutations that produced meiosis-deficient mat1 alleles suggested that the gene-replacement model applied to switching in fission yeast as well (20, 21). These meiosis-defective mutations genetically defined the mat2 and mat3 loci, proposed as the source for copies of the genetic information that replaces the mat1 allele by recombination (Fig. 1).

A DNA FRAGILE SITE INITIATES RECOMBINATION FOR mat1 SWITCHING

Southern analysis of DNA extracted from yeast cells (10, 22) revealed that 20 to 25% of mat1 DNA exhibits a site-specific double-stranded DNA break (DSB) (Fig. 4). Analogous to the previously described similar finding in the S. cerevisiae cell-type switching system (23), the DSB in S. pombe was proposed to initiate recombination required for mat1 switching. The association between the DSB and switching was supported by genetic analysis of three switching-defective mutants, swi1 (swi for switch), swi3, and swi7, in which the switching deficiency was paralleled by a correspondingly reduced DSB level at the mat1 locus (24, 25). The swi1, -3, and -7 gene mutations played a fundamental role in deciphering the switching mechanism and help explain the biological basis of the switching pattern found in cell pedigrees (Fig. 3).

FIGURE 4.

The imprint creates a fragile site in mat1 DNA. Southern blot analysis of DNA extracted from a wild type and from an imprint-deficient swi1⁻ yeast strain is shown. The DNA was digested with HindIII endonuclease, and the resulting blot was probed with a radiolabeled DNA fragment containing the mat1-P cassette. The intensity of the signal of each fragment reflects the extent of DNA sequence homology between the cassettes. The imprint causes a fragile site in DNA that results in a double-stranded break during conventional methods of DNA extraction. The imprint level is much reduced in the swi1⁻mutant. doi:10.1128/microbiolspec.MDNA3-0003-2014.f4

GENETIC ANALYSIS SUPPORTS THE STRAND-SPECIFIC IMPRINTING MECHANISM OF SWITCHING

Cell-pedigree experiments have established that two consecutive asymmetric cell divisions are required to generate the one-in-four pattern of switching (Fig. 3). Clearly, the first decision for a switch of the granddaughter cell must be made in the grandparental cell and the second one in one of its daughter cells. Why do some cells switch while their sisters never do? This question led to the suggestion that a single-stranded inheritable imprint could be installed during DNA replication in the grandparental cell, and this imprint is subsequently used in one daughter cell to induce recombination in one of the two chromatids during replication (26). This scenario raises the problem of how to prohibit repair of the imprint for the length of an entire cell cycle. Compounding the problem is the observation that a constant level of DSB is maintained in cells regardless of the stage of the cell cycle. A molecular model was envisioned in which the Watson and Crick strands of DNA are nonequivalent in their ability to acquire the developmental potential for switching (26). That is, the double-helical structure of DNA might constitute the primary basis for cellular differentiation in this organism. To test this hypothesis genetically, another mat1 gene in an inverted orientation was introduced by a DNA-mediated transformation adjacent to the indigenous mat1 locus. According to the strand-specific imprinting/segregation model, two (cousin) cells among four grandchildren should switch as opposed to the one-in-four pattern found in standard strains that contain a single mat1 gene. The predicted pattern of cousins switching was observed in ~32% of cell lineages having the inverted duplication (15). Additionally, Southern analysis showed that the DSB occurs in either one mat1 gene or the other, but never in both genes on the same chromosome (26). Both of these results precisely conformed to the molecular and developmental predictions of the model. This conclusion required that the DNA sequence employed to make the mat1 inverted duplication carried all the sequences necessary for switching and that the inverted gene did not interfere with the indigenous mat1 locus’s ability to switch. The two-in-four granddaughter cells switching result indicated a gain in switching proficiency according to the strand-segregation model in predicted ways. However, a control experiment with a direct mat1 duplication could not be performed because the duplication is too efficiently removed by recombination at the mat1 locus (26).

THE IMPRINT ALSO EFFICIENTLY INITIATES mat1 GENE CONVERSION IN MEIOSIS

Analysis of strains engineered to be deleted for donor loci has significantly enhanced analyses of molecular events that occur at the mat1 locus. The lack of donor loci “freezes” the switching mechanism and directs repair of the imprint by a different pathway. The two donor loci and the intervening K region were deleted by replacing this ~17.0-kb region with the S. cerevisiae LEU2 gene (Fig. 1). The transgene was named mat2/mat3Δ::LEU2 (7). The mat1-P, mat2/mat3Δ::LEU2 cells exhibit only the P cell type and the mat1-M, mat2/mat3Δ::LEU2 cells exhibit only the M cell type. Although these heterothallic strains do not switch mating type, they retain the usual level of DSB at mat1. Therefore, the DNA break can be repaired somehow without undergoing conventional mat1 switching. It was initially hypothesized that repair might occur by recombination with the intact mat1 gene of the sister chromatid (7). This suggestion was supported by the finding that donor-deleted cells die if defective in the DSB repair gene rad22, just as the h⁹⁰, rad22 cells do (27, 28).

Having donor-deleted strains allowed investigators to ask if the imprint (normally used for mat1 switching in mitosis) could act as a recombination “hotspot” to induce mat1 gene conversion in meiosis as well. Crossing mat1-M, mat2/mat3Δ::LEU2 cells with mat1-P, mat2/mat3Δ::LEU2 cells followed by conventional meiotic tetrad analysis tested this idea. In these crosses, ~80% of the tetrads obtained were of the expected 2P:2M class type, which indicates conventional segregation of the mat1 alleles, but ~10% of the tetrad genes converted mat1-P into mat1-M (1P:3M tetrad type), and another ~10% of the genes converted mat1-M into mat1-P (3P:1M tetrad type). Thus, ~20% of the asci underwent mat1 gene conversion in meiosis, a high frequency of gene conversion compared to the level normally found at other loci (7). This result was interpreted to mean that the imprint normally used for conventional mat1 switching (Fig. 3) could also efficiently induce meiotic gene conversion by recombination between mat1 loci residing on chromosome 2 homologs. When S. pombe cells mate, the resulting zygotic cell does not divide further; instead, the zygote directly proceeds to meiosis and sporulation. This unique feature of the S. pombe life cycle was exploited to show that the imprint installed at the mat1 locus induces mat1 recombination in cis and segregates as an epigenetic entity, similar to a conventional Mendelian marker in mitosis and in meiosis.

swi1, swi3, AND swi7 GENES FUNCTION TO INSTALL THE mat1 cis-ACTING IMPRINT

Although a correlation had been observed between low switching efficiency and low DSB level in the swi1, -3, and -7 mutants, further analysis was required to ascertain whether the lower level of DSB caused the switching deficiency, or whether both deficiencies stemmed from an upstream source (29). To test whether these swi gene factors function to form the mat1 imprint and to determine whether the imprint is stable for one generation as stipulated by the one-in-four switching rule, crosses of donor-deleted strains were conducted in which one of the strains was swi⁻ while the other was swi⁺. Control crosses of mat1-M, mat2/mat3Δ::LEU2 cells with mat1-P, mat2/mat3Δ::LEU2 cells produced ~20% tetrads that were meiotically gene converted for both mat1 alleles, but a cross between donor-deleted mat1-P, swi3⁻and donor-deleted mat1-M swi⁺ strains generated an aberrant tetrad class of predominantly 3P:1M type (29). Conversely, a cross between donor-deleted mat1-M swi3⁻ and donor-deleted mat1-P swi⁺ strains generated a predominantly gene-converted class of 1P:3M type. Similar crosses with a swi1⁻ or a swi7⁻ parent generated meiotic mat1 gene conversions in which only the allele contributed by the swi⁺ parent converted. These results established several principles that further our understanding of the switching mechanism: (i) the competence for meiotic mat1 gene conversion segregates in cis with the mat1 locus; (ii) the swi1, -3,- and -7 genes had already installed the competence in a fraction of mitotic cells that participated in the cross; (iii) the swi⁺ gene present in the zygotic cell was incapable of conferring gene conversion potential to the mat1 allele that was replicated in a respective swi⁻ background; and (iv) because only tetrads with a 3:1 plus 1:3 ratio were observed, and none with a 0:4 ratio, only one of the two sister chromatids of the previously imprinted chromosome had undergone gene conversion. These results also established that chromosomally imprinted functions are installed by swi gene-encoded factors at least a generation before a gene conversion at mat1 can occur. Such meiotic analysis of the mat1 gene-conversion potential of individual chromosomes and their replication histories with respect to the swi gene constitution in mitotic cells was the key to defining the pattern of switching found in cell pedigrees (Fig. 3). These meiotic mat1 gene conversion results, combined with the two-in-four granddaughters switching result of the mat1-inverted duplication, were central in deciphering this unusual mechanism of mat1 switching by programmed recombination.

ORIENTATION OF mat1 REPLICATION DICTATES THE SWITCHING PATTERN

Two-dimensional gel analysis of replication intermediates at the mat1 locus had shown that the locus is replicated unidirectionally from centromere-distal origin(s) (3, 30). This is controlled by a polar replication termination site (RTS1) situated 0.7 kb to the centromere-proximal side of mat1 that blocks DNA replication proceeding from the centromeric direction (31). Notably, imprinting/switching was drastically reduced when the mat1 gene was inverted at its indigenous location in the chromosome or when the RTS1 element was trans-placed from its indigenous location to the centromere-distal side of the mat1 gene. However, once both these rearrangements, each individually reducing switching drastically, were combined in cis, both imprinting and switching proficiency were restored. These results clearly indicate that the imprint is installed only during the lagging-strand synthesis in one cell cycle in a chromatid-specific fashion where it is maintained for the entire length of the cell cycle. The ensuing DNA replication of the imprinted strand by the leading-strand replication complex stalls to create a DSB that induces switching in the specific chromatid (29, 32). Another possibility to consider is that the inherent difference in the sequence of the two DNA strands alone might form the biological basis of differentiated sister cells. However, if this were true, then all cells could produce one switched and one unchanged daughter instead of only one switched cell in four grandchildren (Fig. 3). These studies defining the DNA replication intermediates further established the orientation of replication, and led to the strand-specific imprinting/segregation models (15, 26) that explained the pattern of switching in cell pedigrees. However, this analysis initially created a major inconsistency regarding the result obtained when the additional inverted mat1 placed ~5.0 kb to the centromere-proximal side of the existing mat1 was switching proficient (15, 26), whereas the inverted cassette at its indigenous location was switching deficient (30). The discovery of the RTS1 element between the inverted mat1 cassettes resolved this inconsistency (31). Indeed, the strand-specific imprinting/segregation model precisely predicted such a genetic outcome.

swi1, swi3, AND swi7 FACTORS PLAY MULTIPLE ROLES IN IMPRINTING

The visualization of extracted DNA by Southern blot, probed with a mat1 fragment and subsequent quantitation of the DSB, indicated that nearly one-quarter of the DNA is broken at a specific site and suggested that the DSB (Fig. 4) is maintained unrepaired through the entire cell cycle (10, 22). As a persistent reactive site, during the following cell cycle it induces recombination in one of the chromatids resulting from DNA replication of the imprinted chromosome. Although early molecular analyses of switching mechanisms relied on the level of the DSB, this model had to be modified upon the discovery that the DSB observed in vitro was an artifact of the DNA preparation procedure (30, 32). The DSB was not found in DNA purified by embedding cells in agarose plugs instead of the conventional method using phenol/chloroform extraction, vortexing, and RNase treatment (10, 22, 33). Furthermore, denaturing the DNA by formamide and formaldehyde indicated that the DNA strands at mat1 were intact, but denaturing with sodium hydroxide produced a significant level of strand-specific break at mat1 (30). As the DSB is an artifact of the DNA extraction procedure because of the presence of a fragile site at the imprint, the level of the single-stranded nick is a better quantitation of the imprint. This method indicates that nearly 50% of the chromosomes are nicked, in agreement with the strand-segregation model (32, 34).

Imprinting requires several mat1 cis-acting sequences (Fig. 5): nearby replication origins, a Sap1 (switch activation protein 1)-binding site (SAS1) (35), a polar replication pause site (MPS1) (36), and the centromere proximal RTS1 (31). The SAS1 site is about 150-bp centromere distal from the imprint site, the MPS1 is situated near the imprint site, and the RTS1 is about 0.7-kb centromere proximal to the mat1 gene (31). Another recently described cis-acting element is a 104-bp “spacer region” that is essential for imprinting and lies proximal to the imprinting site (37).

FIGURE 5.

mat1 cis -acting sequence elements and trans-acting factors required for imprinting. See the text for details. doi:10.1128/microbiolspec.MDNA3-0003-2014.f5

At least six trans-acting proteins are necessary for imprinting as well (Fig. 5): Sap1 (38), Swi1, Swi3, and Swi7 (24), and the lysine-specific demethylases Lsd1 and Lsd2 (39). The first trans-acting factor identified as essential for imprinting was the catalytic subunit of DNA polymerase α encoded by swi7 (40). The Swi1 protein sequence resembles a mammalian timeless protein (Tim) (36). Swi3 is highly similar to the mammalian and mouse timeless interacting protein (Tipin) (41). Tim and Tipin are replisome-associated proteins that prevent replication-fork collapse (42). A complex including Swi1 and Swi3 (41) is required for replication-fork pausing by binding to the MPS1 and for blocking replication forks of the wrong polarity at RTS1. Interestingly, a specific mutant allele of swi1 indicates that different moieties of Swi1 function at MPS1 and RTS1 sites (36). Lsd1 and Lsd2 also facilitate imprinting by pausing replication at MPS1 (39). The requirement for swi7 (Polα), whose major function is synthesizing the RNA primer for the Okazaki fragment of the lagging strand, suggested the possibility that the primase function of the Polα/primase complex is responsible for imprint installation and that the imprint consists of ribonucleotides that are not removed from the Okazaki fragment primer by the replication machinery. Interestingly, swi1 and swi3 mutations compromised MPS1 pausing, whereas a swi7 mutation did not (36). Therefore, MPS1 pausing might facilitate the placement of the postulated RNA primer at a specific position by lagging-strand replication.

Recent studies have extended the mechanism of installing the imprint (reviewed in references 39 and 43). The function of the MPS1 site is to promote site-specific RNA priming at mat1 by the primase/Polα complex at the imprint site so that the 5′ end of the stalled fork is dictated by the replication pause (37). Work from the Arcangioli laboratory suggests that the imprint is a single-stranded nick that contains unphosphorylated 3′-OH and 5′-OH termini (32, 44, 45). It is possible that different outcomes, such as alkali lability of the imprint, a strand-specific nick, and/or DSB, derive from inserted ribonucleotides forming the imprint. Regardless of whether the imprint consists of ribonucleotides and/or a nick, all biochemical studies on the imprinting mechanism have supported the strand-specific imprinting/ segregation mechanism first proposed years ago from the genetic evidence (15, 26). Overall, several cis-acting sites spread around ~200 bp on both sides of the imprint site are required for imprinting. Such an arrangement suggests that a specific Okazaki fragment synthesized at the imprint site holds the key for imprinting. Accordingly, this yeast has adapted the usual DNA replication machinery to accomplish cellular differentiation by evolving an exquisitely regulated recombination mechanism at the mat1 locus.

mat1 SWITCHING REPLACES BOTH DNA STRANDS

As no endonuclease activity has been identified that produces the nick or DSB at mat1, an alternative hypothesis is that when the replication fork synthesizing the leading strand reaches the imprint, a polar one-ended DSB is transiently formed (29). In donor-deleted cells, the DSB is repaired using the intact sister chromatid (7), but cells preferentially use the mat2/mat3 donor loci, if available, for mat1 switching. Supporting the sister chromatid hypothesis, sister chromatid recombination intermediates have been found by molecular methods (46). Interestingly, each recombination pathway uses a unique en-donuclease to resolve the recombination intermediates; Mus81/Eme1 is used for sister chromatid-associated repair, and Swi9/Swi10 effect donor-associated repair (46).

The initiation step of repair remains unclear. One suggestion is that the broken end invades mat2 or mat3, and extends the 3′-OH end by the copy-choice mechanism (29, 30). Another possibility is that repair is initiated by template switching through the replication–recombination coupled process (32, 34, 47, 48), similar to that of the canonical DSB repair mechanism (23, 49, 50). Following the fate of newly synthesized ¹³C- and ¹⁵N-labeled strands during switching showed that both mat1 DNA strands are synthesized de novo when the opposite mat1 information is substituted (51). However, when switching between two P cassettes with homologous sequences was tested, one at mat1 and the other at mat2, each cassette marked with a different mutation, the wild-type mat1-P cassette was recreated through heteroduplex formation and repair (48). In switching between mat1 and a donor, a single strand of the donor is synthesized and then transmitted to mat1, where it serves as a template for the complementary strand. An artificially introduced reversible imprinting mechanism shows that the imprint occurs in one cell cycle, remains stable throughout the next cell cycle, and causes switching in the following cell cycle (34). These molecular results support the strand-segregation model derived from previous genetic studies (15, 26).

DIRECTIONALITY OF SWITCHING

The donors that replace the mat1 cassette are selected in a highly biased and counterintuitive manner. For example, mat1-P changes mostly by copying DNA from the nonhomologous mat3-M, and mat1-M from the non-homologous mat2-P in ~90% of the switching events. Donor preference, called directionality of switching, is determined by location rather than by sequence of the donor loci (52). This was demonstrated using a strain with swapped genetic contents of the donors, mat2-M and mat3-P, referred to as h⁰⁹. The switching to opposite mating-type was very inefficient (~18%) in the h⁰⁹ strain, which suggests that the directionality is working correctly and that most switches result in homologous replacement. Recently, it was reported that cis-elements adjacent to mat2 and mat3, the Swi2-dependent recombination enhancer (SRE) adjacent to mat2 (SRE2), and SRE3 at mat3, govern the directionality (53). When mat2 and mat3 are swapped, accompanied by their respective SRE2 and SRE3 elements, the cells switch as efficiently as wild type, indicating that the SRE elements are critical for determining the donor choice. How is the directionality determined by the location? There are two mechanisms identified for directionality; one is a cell-type specific distribution of a protein complex that mediates homologous recombination and the other is a chromosomal conformation to facilitate selective intra-chromosomal interactions.

The recombination-promoting complex (RPC), containing Swi2 and Swi5, promotes the recombination between mat1 and donor loci (54). Swi5 is also required for general recombination (28) and Swi2 recruits the recombination factor Rhp51 to the mat region (55). In P cells, Swi2 binding is mostly localized to the mat3 region dictating mat3-M as the donor, whereas in M cells Swi2 binding is distributed throughout the mat2/3 region (54). Swi2 binding to the mat region requires the SRE3, and it was proposed that in M cells Swi2 would reach to the mat2 region by spreading from the SRE3 through a heterochromatin region between mat2 and mat3 via physical interaction with Swi6 (54). Two more proteins, Abp1 and Mc, are required for Swi2 binding to the mat2 region, perhaps by promoting Swi2 spreading (56, 57, 58). The Mc protein, a transcription factor encoded within the M cassette, provides proper distribution of the RPC by positively regulating the expression of swi2 and swi5 genes in M cells (57). In addition, Mc stimulates expression of a shorter transcript (swi2S) of the swi2 gene that facilitates mat2 utilization in M cells (58). However, in conflict with the spreading hypothesis, another study suggests that Swi2 binds directly to the mat2 cis-acting element in M cells (53).

Heterochromatin in the mat region is another factor required for directionality (59). Defects of genes required for heterochromatin formation, swi6, clr4, rik1, clr7, and clr8, compromise directionality (52, 60, 61, 62). Therefore, heterochromatin may mediate Swi2 spreading and/or provide proper chromosomal conformation for promoting directionality.

mat1 IMPRINT IS A HOTSPOT OF MITOTIC RECOMBINATION BETWEEN HOMOLOGS

In addition to functioning in mat1 switching, the imprint also creates a mitotic recombination hotspot in diploid cells that causes homozygosis of all the centromere-distal markers (Fig. 1). The efficiency of homozygosis loosely reflects the switching rate of a mat1-cis-acting mutation carried on one of the homologs, even though the other homolog carries the recombination-competent wild-type mat1 allele (63, 64); thus, recombination only occurs when both mat1 loci in the diploid are capable of switching (Fig. 6). This feature is in sharp contrast to the general recombination mechanism, which is limited by the frequency of the recombination-initiating event, such as the double-stranded chromosomal break (23, 49, 50), but once initiation has occurred in one of the chromosomes/chromatids, recombination and DNA repair ensue efficiently. This difference in recombination between imprint-associated recombination, which requires both mat1 loci to be capable of switching, and general recombination, which is governed by the rate of the initiating event, led to the proposal of the selective chromatid recombination and arms-swapping model wherein chromatids that transiently and simultaneously carry DSBs and/or a single-stranded nick at mat1 could recombine (Fig. 6). In this model, chromatid arms can recombine or segregate without recombination at an equal frequency and both without involving DNA repair. Thus, the hotspot recombination is proposed to occur through a pathway other than the conventional DSB repair model. In support of this hypothesis, the hotspot recombination is not reduced in the swi5 mutant, encoding a function required for general mitotic recombination in yeast (65).

FIGURE 6.

The selective chromatid recombination and arms-swapping model of the mat1 hot-spot recombination. Recombination occurs only between nonsister chromatids numbers 1 and 3, and only when they simultaneously have a DNA break/nick (\\) in the S/G2 phase of the cell. To depict specific strand distribution, template Watson (W) strands are colored blue and Crick (C) strands are colored red. Strands synthesized in the present the replication cycle are indicated in black. The symbol X represents the crossover point at mat1. The figure is modified from reference 65. doi:10.1128/microbiolspec.MDNA3 -0003-2014.f6

A recent study (65) has reported several unusual features of the mechanism of hotspot recombination: (i) recombination occurs in the S/G2 phase of the cell; (ii) recombination occurs at an appreciable rate in about 4.0% of cell divisions; (iii) one-half of recombination events causes homozygosis, while the other half only changes linkage of markers flanking the mat1 locus; (iv) recombination occurs only between previously imprinted nonsister chromatids; and (v) recombination occurs only between chromatids that simultaneously contain DSBs/ nick. These results support the arms-swapping model (Fig. 6) and advance our understanding of the mechanism of mitotic recombination at the mat1 hotspot. Moreover, unique features, such as recombination occurring only in the S/G2 phase and only between specific nonsister chromatids, have helped define the mode of segregation of chromosome 2 strands to daughter cells. By genetic analyses, it is argued that chromosome 2 strands of one homolog are segregated randomly and independently with respect to those of the second homolog during cell division. Overall, selective chromatid recombination occurs at the hotspot and it is followed by random chromatid distribution to daughter cells. It is generally thought that the DSB-induced recombination occurs without any chromatid bias in eukaryotes. Overall, hot-spot recombination varies from the general DSB recombination/repair mechanism in several respects (discussed above).

S. JAPONICUS FISSION YEAST ALSO USES THE DNA STRAND-BASED MECHANISM TO GENERATE ASYMMETRIC CELL DIVISION

Chromosomally borne epigenetic differences between sister cells that change cell type have been demonstrated only in the S. pombe fission yeast. For technical reasons, it has been impossible to determine the existence of such a mechanism operating elsewhere in biology, especially during the embryonic development of multicellular organisms. Thus, it is easy to perceive that such a DNA-based mechanism does not operate in biology at large. However, unraveling the elegant system of mat1 switching in S. pombe motivated the search for another experimentally favorable system that permits observation of a similar DNA strand-based mechanism of asymmetric cell division. With that goal in mind, a study was initiated with the recently sequenced fission yeast S. japonicus (44% GC content) (66), which is highly diverged from the well-studied S. pombe species (36% GC content) having only a 50% identity at the amino acid level in their protein orthologs. Remarkably, S. japonicus cells also switch cell/mating type after undergoing two consecutive cycles of asymmetric cell divisions and, as in S. pombe, only one of four granddaughter cells switches. The DNA strand-specific epigenetic imprint at mat1 initiates the recombination event that is required for cellular differentiation (67). Therefore, the S. pombe and S. japonicus mating systems provide the first two examples in which the intrinsic chirality of the double-helical structure of DNA was found to confer asymmetric cell division. This conservation of the strand-specific mechanism of asymmetric cell division of S. japonicus is all the more interesting considering that the DNA sequences and genomic arrangement of cassettes are fundamentally different from those of S. pombe (67).

mat1-SWITCHING PARADIGM ADVANCED TO EXPLAIN ASYMMETRIC CELL DIVISION AND DEVELOPMENT IN HIGHER ORGANISMS

We learned from the fission-yeast studies that two consecutive asymmetric cell divisions are required to produce one switched granddaughter cell of a nonswitchable cell and that in both divisions the developmental program is advanced by the act of DNA replication itself. Also, DNA replication produces nonequivalent chromatids of the parental chromosome in the Watson versus Crick strand-specific fashion, and stable epigenetic states of gene expression are replicated along with the DNA and inherited both in mitosis and meiosis as conventional Mendelian factors (68, 69, 70). Could these lessons be exploited to explain cellular differentiation and development in higher eukaryotes?

In principle, differential gene regulation of developmentally important gene(s) may be accomplished by epigenetic entities somatically installed in a strand-specific fashion at specific stages in development (Fig. 7). This may occur by differential heterochromatin assembly and/or by altering cytosine base methylation of the relevant gene to turn off its transcription or to activate its transcription by disrupting the repressing epigenetic control. In this fashion, nonequivalent sister chromatids might be produced. However, for the hypothesized chromatid asymmetry to be a useful mechanism for cellular differentiation in diploid organisms, selective segregation of differentiated chromatids of both homologs is required. The somatic strand-specific imprinting and selective segregation (SSIS) model predicts the evolution of such a mechanism for producing developmentally non-equivalent sister cells in diploid organisms (Fig. 7). This model was advanced to explain visceral left–right axis laterality development in mice (71) and for human-brain hemisphere specialization (72). The SSIS model postulates selective chromatid segregation of specific chromosomes at a specific cell division to accomplish development during embryogenesis. Although the experiments involved are not straightforward, the selective chromatid segregation phenomenon appears to operate in mouse stem cells (73, 74) and in asymmetrically dividing Drosophila cells (75). For discovering aspects of the SSIS mechanism, we owe thanks to the model organisms of fission yeast for providing us with a paradigm possibly applicable to investigations of developmental biology in higher eukaryotes.

FIGURE 7.

The SSIS model. The model (71, 72) predicts the evolution of two phenomena. One causes production of nonequivalent sister chromatids through epigenetic means to express differentially a developmentally important gene in the Watson (W) versus Crick (C) strand/chromatid-specific fashion. The other causes nonrandom segregation of all four chromatids belonging to a pair of homologous chromosomes, or sets of chromosomes, in mitosis. To depict the diagrammed nonrandom segregation pattern, termed “W,W::C, C,” DNA strands are color-coded. The strands synthesized in the parent cell are depicted in black. A hypothetical trans-acting segregator factor is proposed to function by acting at the centromere to mediate selective chromatid segregation at a specific cell division in development to promote asymmetric cell division. Lateralized organs could be derived from the progeny of thus difierentiated daughter cells. doi:10.1128/microbiolspec.MDNA3 -0003-2014.f7