Localization and Properties of a Silencing Element Near the mat3-M Mating-Type Cassette of Schizosaccharomyces pombe

Corresponding author: Geneviève Thon, Department of Genetics, Institute of Molecular Biology, University of Copenhagen, Øster Farimagsgade 2A, DK-1353 Copenhagen K, Denmark., gen{at}biobase.dk (E-mail)

Communicating editor: G. R. SMITH

	ABSTRACT

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Transcription is repressed in a segment of Schizosaccharomyces pombe chromosome II that encompasses the mat2-P and mat3-M mating-type cassettes. Chromosomal deletion analysis revealed the presence of a repressor element within 500 bp of mat3-M. This element acted in synergy with the trans-acting factors Swi6, Clr1, Clr2, Clr3, and Clr4 and had several properties characteristic of silencers: it did not display promoter specificity, being able to silence not only the M mating-type genes but also the S. pombe ura4 and ade6 genes placed on the centromere-distal side of the mat3-M cassette; it could repress a gene when placed further than 2.6 kb from the promoter and it acted in both orientations, although with different efficiencies, the natural orientation repressing more stringently than the reverse. Following deletion of this element, two semistable states of expression of the mat3-M region were observed and these two states could interconvert. The deletion did not affect gene expression in the vicinity of the mat2-P cassette, 11 kb away from mat3-M. Conversely, deleting 1.5 kb on the centromere-proximal side of the mat2-P cassette, which was previously shown to partially derepress transcription around mat2-P, had no effect on gene expression near mat3-M. A double deletion removing the mat2-P and mat3-M repressor elements had the same effect as the single deletions on their respective cassettes when assayed in cells of the M mating type. These observations allow us to refine a model proposing that redundant pathways silence the mating type region of S. pombe.

IN eukaryotes, specialized chromatin structures influence the availability of certain chromosomal regions to transcription. Thereby, two copies of the same gene occasionally display different levels of expression even though they are located within the same nucleus, as do, for example, allelic genes located in the female X chromosomes of mammals (for reviews, see RIGGS and PORTER 1996 ; HEARD et al. 1997 ) and various imprinted genes (for review, see AINSCOUGH and SURANI 1996 ; BARTOLOMEI and TILGHMAN 1997 ). Furthermore, the location and extent of the inactive chromosomal areas can vary from organism to organism and from cell to cell, causing variegated phenotypes (reviewed for Drosophila by WEILER and WAKIMOTO 1995 ). Such flexibility suggests that turning on or off large chromosomal regions that contain clusters of genes involved in a common process might be a means of regulating cellular differentiation and development. This type of regulation occurs, as best exemplified by the regulation of the Drosophila homeotic genes (reviewed by PARO and HARTE 1996 ; PIRROTTA 1996 ).

Where and how are inactive regions formed? The features that trigger transcriptional inactivation of specific regions appear diverse, as do the modes of subsequent inactivation. Inactivation does not always originate from a well-defined repressor element. For example, gene silencing in Neurospora and position effect variegation in Drosophila can be caused by repeats of a gene that is not silenced when present in single copy (for review, see HENIKOFF 1996 ; RUSSO et al. 1996 ). In other cases, specialized DNA elements or silencers are able to interact with nuclear proteins to create a zone of repression in the chromosomes of cells meeting specific criteria. The Polycomb response elements in Drosophila (reviewed by PARO and HARTE 1996 ; PIRROTTA 1996 ), and the HML and HMR silencers in Saccharomyces cerevisiae (reviewed by HOLMES et al. 1996 ) are examples of such elements.

In the fission yeast Schizosaccharomyces pombe, position effects are observed near centromeres and telomeres and in the mating-type region (for review, see ALLSHIRE 1996 ). Prototrophic markers introduced at these locations are repressed in a variegated fashion. The repression is only partial in telomeric regions (NIMMO et al. 1994 ). It is tighter for some centromeric insertions (ALLSHIRE et al. 1994 , ALLSHIRE et al. 1995 ) and within the mating-type region (THON and KLAR 1992 ; THON et al. 1994 ). The products of swi6, rik1, clr1, clr2, clr3, and clr4 are required for the repression of transcription in these places, which indicates that the mechanisms of repression are related (LORENTZ et al. 1992 ; THON and KLAR 1992 ; EKWALL and RUUSALA 1994 ; THON et al. 1994 ; ALLSHIRE et al. 1995 ; GREWAL and KLAR 1997 ). Swi6, Rik1, and Clr4 appear more important for centromeric silencing than the other three factors, and mutations in swi6, rik1, and clr4 affect chromosome segregation in addition to transcription (ALLSHIRE et al. 1995 ). In the mating-type region, the six trans-acting factors repress not only transcription but also meiotic recombination (EGEL et al. 1989 ; KLAR and BONADUCE 1991 ; LORENTZ et al. 1992 ; THON et al. 1994 ). These pleiotropic phenotypes suggest an action at the level of chromatin structure.

The mating-type region comprises three linked loci in the right arm of chromosome II (Figure 1). The centromere-proximal locus, mat1, is expressed, whereas the two other loci, mat2-P and mat3-M, are not transcribed. In wild-type homothallic strains, designated h⁹⁰, mat2-P and mat3-M donate genetic information to mat1 in an efficient process akin to gene conversion. This process leads to interconversion of mat1 between two allelic forms, mat1-P and mat1-M, which determine, respectively, the P and M mating types of S. pombe (for review, see KLAR 1992 ). The two mating-type genes present within mat2-P and the two mating-type genes present within mat3-M, as well as prototrophic markers introduced near the cassettes or in the 10.9-kb interval that separates them (K region), are subject to transcriptional silencing (EGEL and GUTZ 1981 ; BEACH 1983 ; EKWALL et al. 1992 ; THON and KLAR 1992 ; THON et al. 1994 ; GREWAL and KLAR 1997 ). The borders of the silenced region are not known. An essential gene is located between mat1 and mat2-P, indicating the repression extends less than 10 kb on the centromere-proximal side of mat2-P (MICHAEL et al. 1994 ). The first known gene on the centromere-distal side of mat3-M is his2 and the distance between mat3-M and his2 as inferred from a cosmid map (HOHEISEL et al. 1993 ) is ~50 kb.

View larger version (44K): In this window In a new window Download PPT slide   Figure 1. Mating-type region of S. pombe. (A) General organization. The three mating-type cassettes of S. pombe, mat1, mat2-P, and mat3-M, are separated by a DNA segment where transcription and recombination can occur (L-region) and a segment where both transcription and recombination are inhibited (K-region). A single-strand break (SSB) is thought to initiate switching of the mat1 allele (ARCANGIOLI 1998 , and references therein). The block to transcription and recombination extends outside of the mat2-P-K-mat3-M region for an unknown distance. The K-region displays homology with centromeric sequences [dh dg homology; GREWAL and KLAR 1997 ]. Stars represent putative origins of replication: two near mat2-P allow plasmid replication (OLSSON et al. 1993 ) and two near mat3-M match the core sequence of S. pombe ARS elements (GREWAL and KLAR 1997 ). Restriction sites mentioned in the text are indicated: Bg, BglII; Bs, BssHII; H, HindIII; RI, EcoRI; RV, EcoRV; X, XbaI. Cells with the 7.5-kb deletion in the K-region (GREWAL and KLAR 1996 ; THON and FRIIS 1997 ) lack DNA between the mat2-P centromere-distal and the mat3-M centromere-proximal HindIII sites. (B) Representation of the three cassettes. The three mating-type cassettes contain a P- or M-specific core flanked by short homology boxes. Two transcripts originate from each mat1-P and mat1-M (adapted from KELLY et al. 1988 ). (C) mat3-M flanking sequence. A portion of mat3-M flanking sequence is depicted, along with the H3 homology box and part of the H2 homology box. The mat3-M centromere proximal deletions presented in this study originated at the junction of H2 and H3 at the base pair labeled 1. The extent of the (482) deletion is indicated, as well as the position and orientation of primers used to amplify and reintroduce DNA fragments in the (1185) and (482) deletions as shown in Figure 3. Boldface letters in the sequence represent differences between our clones and the published sequence (U57841; GREWAL and KLAR 1997 ). The sequence in the gray box is a putative Mts1/Mts2 (Atf1/Pcr1)-binding site.

Sequences located within the K region, which separates mat2-P from mat3-M, are implicated in silencing (GREWAL and KLAR 1996 ; THON and FRIIS 1997 ). The K region contains 4.3 kb of homology with centromeric repeats (GREWAL and KLAR 1997 ), suggesting that it interacts with the trans-acting factors shared by the mating-type region and centromeres. A deletion of 7.5 kb that removes the region with homology to centromeres causes the cells to interconvert between two epigenetic states: one similar to the wild type and one partially deficient for mating-type switching and transcriptional silencing (GREWAL and KLAR 1996 ; THON and FRIIS 1997 ). The mutant state resembles the phenotype caused by mutations in the trans-acting factors swi6, rik1, clr1, clr2, clr3, and clr4. Furthermore, there is no cumulative effect when mutations in the trans-acting factors are combined with the 7.5-kb deletion in the K region. This is consistent with an element located within the deleted fragment, possibly the centromeric repeats, either catalyzing the assembly of the trans-acting factors in the wild-type mating-type region, or preventing loss of the silenced state.

Elements close to the mat2-P cassette are also important for silencing, as inferred from deletion studies (EKWALL et al. 1991 ; THON et al. 1994 ). One or several elements on the centromeric side of the mat2-P cassette appear to act in a silencing pathway that is distinct from the pathway mediated by swi6, rik1, clr1, clr2, clr3, and clr4 (THON et al. 1994 ). The esp1, esp2, and esp3 genes were proposed to act in that second pathway (THON and FRIIS 1997 ).

We investigated whether elements similar to the element present near the mat2-P cassette were present near the mat3-M cassette. To this end, we introduced nested deletions in the chromosomal DNA flanking mat3-M. We determined the effect of these deletions on the expression of mat3-M and the region around it, as well as on gene expression in the mat2-P area. Conversely, we examined whether deletion of the mat2-P cis-acting element affected expression around mat3-M. We also tested whether cumulative effects were generated by combining deletions at the two loci or by combining deletions near mat3-M with mutations in the trans-acting factors swi6, clr1, clr2, clr3, clr4, and esp3.

	MATERIALS AND METHODS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Plasmid constructions and DNA sequencing
Escherichia coli strains, plasmid preparation, and cloning techniques: The E. coli strains DH5 (HANAHAN 1983 ) and S1754 (F^- lacI^Q metA endA hsdR17 supE44 thi-1 relA1 gyrA96; a gift from Stanley Brown) were used for cloning. Plasmid DNA was prepared and purified in cesium gradient according to SAMBROOK et al. 1989 . DNA fragments were isolated in agarose/TBE gels and electroeluted. Restriction enzymes (New England Biolabs, Beverly, MA), shrimp alkaline phosphatase (USB), and T4 DNA ligase (Pharmacia, Piscataway, NJ) were used as directed by the manufacturers. AmpliTaq (Perkin Elmer, Norwalk, CT) and Native Pfu (Stratagene, La Jolla, CA) were used for amplification by polymerase chain reaction (PCR). Nucleotides were purchased from Pharmacia.

mat3-M cassette with a deletion of the H3 homology box: pSM10 consists of an S. pombe 4.2-kb HindIII genomic fragment containing the mat3-M cassette cloned in pUC119 (BEACH 1983 ; KELLY et al. 1988 ). There are two EcoRI sites in pSM10: one in the polylinker, on the centromere-distal side of the cassette, and one within the cassette, which divides the HindIII genomic fragment in fragments of 2.5 kb (centromere-proximal) and 1.7 kb (centromere-distal; Figure 1). The 2.5-kb HindIII-EcoRI restriction fragment was cloned in M13mp19 (YANISH-PERRON et al. 1985 ) and mutagenized using an in vitro mutagenesis system version 2 from Amersham (Arlington Heights, IL) and an oligonucleotide whose sequence (5'GCATACAGAAAAATACTCGAGTGTAAAGTATCAGGA3') was designed to replace the mat3-M H3 homology box with an XhoI recognition site. A mutagenized HindIII-EcoRI insert was partially sequenced and cloned in Bluescribe(-) (BSG43 construct; Stratagene). The mat3-M cassette was reconstituted but for the deletion of the H3 box by ligating the 1.7-kb centromere-distal EcoRI fragment of pSM10 in the EcoRI site of BSG43, creating pGT79. A large portion of pGT79 was subsequently sequenced (see below).

mat3-M proximal ExoIII deletions: The double-stranded oligonucleotide produced by annealing 5'TCGAAGATCTCCATGGCCATGGACGCGTGGGCCC3' and 5'TCGAGGGCCCACGCGTCCATGGCCATGGAGATCT3' was cloned into the XhoI site of pGT79. A clone in which the oligonucleotide was oriented with its ApaI site close to the mat3-M cassette was digested with the restriction enzymes BglII and ApaI, and deletions were introduced using ExoIII and S1 nucleases as described (HENIKOFF 1984 ). This resulted in clones with mat3-M proximal deletions starting at the junction between the H2 and H3 homology boxes, in which the deleted fragment was replaced with an XhoI site. The deletion series was used for sequencing and further cloning (see below). Clones with deletions of 482 bp (pGT79d482) and 1185 bp (pGT79d1185) were used to transform S. pombe.

mat3-M distal ExoIII deletions: pGT70 contains a 4.2-kb HindIII fragment with a modified mat3-M cassette in which an NcoI restriction site was introduced by a single bp substitution at the centromere-distal border of the cassette (THON and KLAR 1993 ). The double-stranded oligonucleotide produced by annealing 5'CATGAGATCTCCACCGCGGTGGACGCGTGGGCC3' and 5'CATGGGCCCACGCGTCCACCGCGGTGGAGATCT3' was cloned in the NcoI site of pGT70 with the ApaI recognition site close to the cassette. Nested ExoIII deletions were introduced after digestion with the restriction enzymes ApaI and BglII. A resulting plasmid with a deletion of 424 bp (pGT70d424) was used for further construction and to transform S. pombe.

Combination of mat3-M distal and proximal deletions: The 1.7-kb EcoRI fragment of pGT79d1510 (an ExoIII deletion product of pGT79) was replaced with the 1.3-kb EcoRI fragment of pGT70d424 to create a plasmid with a mat3-M proximal deletion of 1510 bp and a mat3-M distal deletion of 424 bp designated pGT181.

Cloning of PCR fragments in pGT79d482 and pGT79d1185: The oligonucleotides GTO-112: 5'CCACATGTCTCGAGCGCTCTCGCCAAACCTA3', GTO-113: 5'CCACATGTCTCGAGGTAAGTCACGTTTGAGAAATC3', GTO-114: 5'CCACATGTCTCGAGGATTTCTCAAACGTGACTTAC3', GTO-115: 5'CCACATGTCTCGAGCAAACATATTGTTTGCTGC3', GTO-116: 5'CCACATGTCTCGAGGCAGCAAACAATATGTTTG3', GTO-117: 5'CCACATGTCTCGAGCAAAAGGAGTAACAAGTTAC3', GTO-118: 5'CCACATGTCTCGAGGTAACTTGTTACTCCTTTTG3', and GTO-119: 5'CCACATGTCTCGAGTGTAAAGTATCAGGAGTTG3' can anneal near mat3-M as shown in Figure 1C, and the following pairs were used to amplify mat3-M flanking DNA by PCR: GTO-112 and GTO-119, creating the 424-bp fragment 1; GTO-112 and GTO-114, creating the 134-bp fragment 2; GTO-113 and GTO-116, creating the 136-bp fragment 3; GTO-115 and GTO-118, creating the 88-bp fragment 4; GTO-117 and GTO-119, creating the 126-bp fragment 5. PCR amplification was performed with Native Pfu (Stratagene) in reaction volumes of 80 µl using the buffer provided by the supplier, 1 µg of pSM10 as template in each reaction, primer concentrations of 750 nM, nucleotide concentrations of 250 µM, and five cycles of 1 min at 94°, 1 min at 54°, 2 min at 72°). The PCR products were cleaved with XhoI, purified from agarose gels, and cloned into the XhoI site of pGT79d482 either in their natural orientation relative to mat3-M, creating pGT132 with fragment 1, pGT124 with fragment 2, pGT125 with fragment 3, pGT126 with fragment 4, and pGT127 with fragment 5, or in the opposite orientation, creating pGT133 with fragment 1 and pGT131 with fragment 5. In addition, fragment 1 and fragment 5 were cloned in the XhoI site of pGT79d1185 in their natural orientation creating, respectively, pGT142 and pGT137. The PCR inserts and insertion areas were sequenced.

mat3-M(EcoRV)::ade6 allele: The 4.2-kb HindIII fragment from pSM10 was filled in at the ends with the Klenow fragment of DNA polymerase I and ligated into the EcoRV site of pBluescript (Stratagene). The resulting plasmid was designated pPB16. The 3.0-kb SpeI-Asp700 fragment from pAS1 (SZANKASI et al. 1988 ), which contains the ade6 ORF, was filled in the ends with Klenow and cloned into the EcoRV site of pPB16, 150 bp away from the mat3-M cassette. A clone with the ade6 gene inserted with its promoter near the mat3-M cassette was saved as pPB17.

Combination of (H3)mat3-M with (EcoRV)::ura4: pGT77 contains the mat3-M 4.2-kb HindIII fragment with the S. pombe ura4 gene inserted at the EcoRV site 150 bp distal to the cassette (THON and KLAR 1992 ). The 1.7-kb EcoRI fragment of pGT79 was replaced with the 3.5-kb EcoRI fragment of pGT77, to create a mat3-M cassette with a deletion of the H3 box and an EcoRV insertion of ura4. The resulting plasmid was designated pGT180.

DNA sequencing: DNA sequencing reactions were performed with an ABI Prism Dye Terminator Cycle Sequencing Ready Reaction kit (PE Applied Biosystems, Foster City, CA) and run with an ABI sequencing system model 377. The mat3-M proximal deletion series and synthetic primers were used to sequence mat3-M flanking DNA from the centromere-proximal HindIII site to the H2 box. Three differences were observed in a comparison with the published sequence (GREWAL and KLAR 1997 ): the G at position 10315 was missing in our clones, an additional C was found at position 10439, and an A was present instead of a G at position 10493. Our isolate of pSM10 had the same three differences with the published sequence.

S. pombe strain constructions and manipulations
Media: YES (THON and FRIIS 1997 ) was used as rich, complete medium to propagate S. pombe. MSA (EGEL et al. 1994 ) supplemented with 100 mg adenine, 100 mg uracil and 200 mg L-leucine per liter unless specified otherwise was used as sporulation medium. Drop-out media (AA-ura, AA-ade, AA-leu, AA-ade-ura; ROSE et al. 1990 ) were used to identify and select prototrophic strains and FOA [AA-ura drop-out medium supplemented with 1 g 5-fluoroorotic acid (5-FOA) and 50 mg uracil per liter] was used to select ura4 mutant cells. YE (5 g yeast extract, 2 g casamino acids, 30 g glucose per liter) was used to monitor ade6 expression in colonies. Yeast extract, casamino acids, and yeast nitrogen base were purchased from Difco (Detroit). Amino acids and nucleotides were purchased from Sigma (St. Louis). Salts were purchased from Merck. 5-FOA was purchased from United States Biological.

Transformation: S. pombe cells were transformed using the lithium acetate protocol described by HEYER et al. 1986 with the modifications suggested by MORENO et al. 1991 . The strains and plasmids used for the transformations are listed in Table 1. Chromosomal integrations in the mating-type region were obtained in mutant backgrounds (swi6-115 or clr2-E22) that allow for, first, higher efficiency of recombination than the wild type and, second, positive and negative selections for prototrophic markers placed in the normally silenced region. All integrations were analyzed by Southern blot.

DNA preparation and Southern blot analysis: S. pombe DNA was prepared according to MORENO et al. 1991 . The DNA preparations were digested with HindIII, HindIII + XhoI, and EcoRI and size fractionated in 0.7% agarose/TBE gels (SAMBROOK et al. 1989 ). The gels were blotted to Hybond-N nylon membrane as directed by the manufacturer (Amersham). Hybridization was performed overnight at 65° in 5x SSC, 5x Denhardt's, 1% SDS, 100 µg/ml sonicated salmon sperm DNA. The 4.2-kb mat3-M HindIII fragment from pSM10 was radiolabeled using a Promega (Madison, WI) Random Priming kit and 3000 Ci/mmol [-³²P]dCTP from Amersham and used as probe. After hybridization, the blots were washed at 65° as follows: for 10 min in 2x SSC, 1% SDS; for 60 min in 2x SSC, 1% SDS; and for 30 min in 0.1x SSC, 1% SDS. They were autoradiographed on Agfa Curix X-ray films.

Genetic crosses: All strains listed in Table 1 as not originating from a transformation were obtained by tetrad dissection, with the exception of PG1, PG445, PG447, PG1049, PG1560, PG1562, PG1564, PG1566, PG1570, PG1615, PG1654, PG1655, PG1656, PG1657, PG1658, PG1659, PG1671, PG1672, PG1681, PG1682, PG1690, SP1122, SP1124, SP1125, SP1126, SP1138, and SP1151, which were obtained from random spore preparations.

RNA preparation and Northern blot analysis: Cells in liquid cultures were starved for nitrogen as described by NIELSEN and EGEL 1990 . RNA was prepared according to SCHMITT et al. 1990 . A total of 5 µg of each sample was run in 1.5% agarose 2.2 M formaldehyde gels in 3-[N-morpholino]propane-sulfonic acid buffer (SAMBROOK et al. 1989 ) and blotted onto Hybond-N membrane (Amersham) according to the manufacturer's instruction. Antisense RNA probes were prepared from a 665-bp XbaI-HindIII fragment of the cdc2 gene (cdc2 probe; HINDLEY and PHEAR 1984 ; NIELSEN and EGEL 1990 ), a mat1-M 1016-bp BclI-TaqI DNA fragment, and a mat2-P 904-bp HinPI-MluI DNA fragment (Mc and Pm probes; KELLY et al. 1988 ; NIELSEN and EGEL 1990 ) using a Riboprobe II core system (Promega) and 3000 Ci/mmol [-³²P]UTP (Amersham). Hybridization was allowed to occur for 24 hr at 42° in 0.25 M NaHPO₄, pH 7.2, 0.25 M NaCl, 7% SDS, 1 mM EDTA, 50% formamide, 10% polyethylenglycol (4000), 5x Denhardt's solution, 100 µg/ml yeast RNA. Washes and autoradiography were as for the Southern blots.

	RESULTS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Deletion analysis of the chromosomal region flanking mat3-M:
We introduced nested deletions on both sides of the mat3-M cassette. The extent of the deletions and their effect on mat3-M expression in both swi6⁺ and swi6-115 background, where the integrations were obtained, are shown in Figure 2. Expression of mat3-M was monitored by the amount of haploid meiosis occurring in cells with the stable mat1-P17::LEU2 allele. This assay relies on the observation that coexpression of the P and M mating-type genes triggers meiosis, not only in zygotes or diploid cells where the situation naturally occurs, but also in haploid cells (KELLY et al. 1988 ). The amount of haploid meiosis observed in mat1-P17::LEU2 cells reflects the expression of the M genes from mat3-M. Haploid meiosis was monitored by exposing colonies to iodine vapors, upon which spore-containing colonies are stained darkly whereas colonies containing no spores are stained yellow (BRESCH et al. 1968 ). According to this assay, deletion of 482 nucleotides on the centromere-proximal side of the mat3-M cassette increased expression of the M genes. Larger deletions that included these 482 bp had the same effect. A smaller deletion removing only the H3 homology box and a deletion distal to the cassette failed to derepress. In the cases where derepression was observed, the effect was small in a wild-type background but very pronounced in a swi6-115 background.

View larger version (62K): In this window In a new window Download PPT slide   Figure 2. Effect of mat3-M flanking deletions on the expression of mat3-M. Sporulated colonies of mat1-P17::LEU2 cells with the mat3-M flanking deletions represented by bars on the side of the photographs were stained with iodine vapors to monitor expression of mat3-M in both swi6+ and swi6-115 backgrounds as indicated. The intensity of staining reflects the amount of haploid meiosis occurring in the colonies. swi6+ strains, from top to bottom: PG1397, PG1404, PG1398, PG1403, PG1690, and PG445; swi6-115 strains, from top to bottom: PG1195, PG1194, PG1193, PG1192, PG1615, and PG1584. (1) and (4) represent two potential Mts1/Mts2 (Atf1/Pcr1)-binding sites. (2) and (3) represent two ARS consensus sequences.

Additional constructs were introduced in the chromosome to localize more precisely the element(s) responsible for the repression (Figure 3). In these constructs, PCR products representing portions of mat3-M flanking DNA were introduced in either the 1185- or 482-bp deletions. This analysis revealed the following. First, a 424-bp fragment representing the portion of DNA deleted in the 482-bp deletion except for the H3 box was sufficient to restore silencing in the 1185-bp deletion, indicating the remaining 761 bp were not required for silencing mat3-M. Second, part of the repression was mediated via an element located within 126 bp of the cassette. The 126-bp DNA fragment partially restored silencing when introduced in the large centromere-proximal deletion of 1185 bp or in the 482-bp deletion. An adjacent fragment of 88 bp also increased the repression although not as efficiently as the 126-bp fragment.

View larger version (100K): In this window In a new window Download PPT slide   Figure 3. Refined analysis of the centromere-proximal mat3-M flanking region. (A) Insertion of DNA elements in (1185). DNA fragments containing 424 or 126 nucleotides of mat3-M flanking DNA were produced by PCR using the indicated primers (GTO-112, -117, -119; sequences shown in Figure 1) and introduced in place of the (1185) deletion in mat1-P17::LEU2 swi6-115 cells. Sporulated colonies were stained with iodine as in Figure 2. (1185): PG1399; (1185)::(424)oriI: PG1526; (1185)::(126)oriI: PG1557. (B) Insertions of DNA elements in (482). DNA fragments produced as in A were used to replace (482) in mat1-P17::LEU2 swi6-115 cells. (482): PG1401; (482)::(424)oriI: PG1527; (482)::(134)oriI: PG1531; (482)::(136)oriI: PG1534; (482)::(88)oriI: PG1538; (482)::(126)oriI: PG1541.

In summary, this series of experiments indicates that the H3 homology box does not exert a significant repression on the M genes contained in the mat3-M cassette and it also rules out an effect of the sequences immediately distal to mat3-M. In contrast, one or several DNA elements located proximally, within <500 bp of the H3 box, play an important role in silencing.

Combination of cis- and trans-acting mutations:
Combining trans-acting mutations pairwise allows assignment of the silencing factors that repress transcription in the mating-type region to one of two groups. Factors whose mutations do not have a cumulative effect when combined are assigned to the same group. Thus, swi6, clr1, clr2, clr3, and clr4 belong to one group (THON et al. 1994 ) and esp1, esp2, and esp3 to the second group (THON and FRIIS 1997 ). The synergy between the swi6-115 mutation and the mat3-M flanking deletion of 482 bp suggests that the deleted element acts in a pathway other than the one mediated by swi6. If the deleted element participates in the pathway containing the esp products, two predictions should be fullfilled. One prediction is that deletions flanking mat3-M should potentiate not only mutations in swi6, but also mutations in clr1, clr2, clr3, and clr4. The second prediction is that expression of mat3-M should not be increased when the deletion of 482 bp is introduced in esp mutant cells such as the esp3-1 mutant. We have tested both predictions and found them to be true. First, combining the 482 bp deletion with mutations in clr1, clr2, clr3, or clr4 caused a strong derepression of mat3-M leading to very high levels of haploid meiosis (Figure 4). Second, combining the 482-bp deletion with esp3-1 did not enhance mat3-M expression: Mc transcripts were not detected in a (482) esp3-1 double mutant (Figure 5A). A low amount of Mc transcript was seen in RNA preparations from swi6-115 cells, and much higher amounts were seen in swi6-115 (482) and swi6-115 esp3-1 double mutants (Figure 5A), as expected from the sporulation phenotypes of these mutants (Figure 4 and data not shown).

View larger version (59K): In this window In a new window Download PPT slide   Figure 4. Cumulative effects of trans-acting mutations in swi6, clr1, clr2, clr3, or clr4 with a mat3-M flanking deletion. Sporulated colonies of mat1-P17::LEU2 cells with the indicated mutations were stained as in Figure 2. The strains were: +, +: PG445; +, swi6-115: PG1584; +, clr1-5: PG1587; +, clr2-760: PG1582; +, clr3-735: PG1579; +, clr4-681: PG1594; (482), +: PG1403; (482), swi6-115: PG1401; (482), clr1-5: PG1412; (482), clr2-760: PG1421; (482), clr3-735: PG1418; (482), clr4-681: PG1420.

View larger version (45K): In this window In a new window Download PPT slide   Figure 5. Combined effects of mutations in swi6, esp3, and cis-acting deletions near mat3-M or mat2-P. RNA was prepared from nitrogen-starved cells with the indicated mutations and used for Northern blot analysis. The Pm, Mc, and cdc2 probes are described in MATERIALS AND METHODS. (A) Effects on mat3-M. All strains in A contain the mat1-P17::LEU2 allele. (482) represents the mat3-M centromere-proximal deletion of 482 bp. Mc transcripts originating from mat3-M were detected with an M-specific probe and the blot was reprobed with a cdc2 probe to estimate the amount of RNA loaded in each lane. The cdc2 probe recognizes several transcripts (DURKACZ et al. 1986 ). The strains used were as follows: lane 1, PG445; lane 2, PG1742; lane 3, PG1584; lane 4, PG1741; lane 5, PG1403; lane 6, PG1435; lane 7, PG1401. (B) Effects on mat2-P. All strains used in this panel contain the mat1-Msmt-0 allele, which allows detection of transcripts originating from mat2-P with a P-specific probe. The blot was reprobed with the cdc2 probe as in A. BII-BII represents the BglII-BssHII deletion on the centromere-proximal side of the mat2-P cassette. The strains used were: lane 1, SP1124; lane 2, PG1174; lane 3, SP1126; lane 4, PG1063; lane 5, SP1151; lane 6, PG1165; lane 7, SP1138.

Deletion of one or more cis-acting elements located on the centromere-proximal side of the mat2-P cassette in a 1.5-kb BglII-BssHII fragment was previously reported to cause phenotypes similar to the phenotypes caused by the mat3-M flanking deletions reported here (THON et al. 1994 ). The BglII-BssHII deletion had only a small effect on its own but enhanced the effect of mutations in swi6, clr1, clr2, clr3, or clr4. We examined the amount of Pm transcript originating from mat2-P in single- and double-mutant backgrounds, as described above for the mat3-M cassette. As can be seen in Figure 5B, expression of the Pm gene was increased by combining a mutant swi6 allele with either the mat2-P cis-acting deletion or a mutant esp3 allele, but not by combining the cis-acting deletion with the mutant esp3 allele. The strongest derepression of mat2-P was caused by simultaneous impairment of the two trans-acting factors, swi6 and esp3, which also led to the strongest derepression of mat3-M (Figure 5A).

Hence, the mat2-P and mat3-M flanking elements have similar genetic interactions with trans-acting factors important for silencing: a cumulative effect with mutations in swi6, clr1, clr2, clr3, and clr4, and no cumulative effect with a mutation in esp3.

Effect of mat3-M proximal deletions on the expression of the S. pombe ura4 and ade6 genes placed near mat3:
The deletion analysis of the mat3-M flanking regions had revealed the presence of DNA sequences important for the repression of the mating-type genes contained within the cassette. We tested whether these sequences could silence other S. pombe genes placed on the centromere-distal side of mat3-M. The S. pombe ura4 gene is repressed in wild-type backgrounds when placed at an EcoRV site located ~150 bp away from the mat3-M cassette (THON and KLAR 1992 ). We introduced the ura4 gene at the EcoRV site in some of the strains with the mat3-M flanking deletions described above. As shown in Figure 6A, deleting the H3 box did not affect expression of ura4 placed at the EcoRV site near mat3-M, whereas deleting 482 nucleotides led to an increased expression of ura4. The ability of cells to form colonies on FOA was not abolished by the deletion, but growth was greatly reduced, resulting in abnormally small colonies. These observations confirmed that one or more elements contained within the deleted fragment had a repressive effect on transcription and that the repression was not specific for the M genes located within mat3-M but could affect the promoter of another gene placed at a distance >2.6 kb from the cis-acting element.

View larger version (266K): In this window In a new window Download PPT slide   Figure 6. Derepression of prototrophic markers caused by mat3-M flanking deletions. (A) Derepression of ura4. Cells containing the mat3-M(EcoRV)::ura4 allele in combination with the indicated deletions were propagated under nonselective conditions (YES). Expression of ura4 was monitored by spotting 10-fold serial dilutions of cell suspensions on medium lacking uracil (AA-ura), medium containing FOA (FOA), and complete medium (AA). Rows 1, PG1566; rows 2, PG1141; rows 3, PG447; rows 4, PG1560; rows 5, PG1550. (B) Derepression of ade6. Cells with the mat3-M(EcoRV)::ade6 allele were propagated in YES medium. Tenfold serial dilutions of cell suspensions were plated on medium containing a low concentration of adenine (YE), on medium containing no adenine (AA-ade), and on complete medium (AA). Rows 1, PG1570; rows 2, PG1141; rows 3, PG1672; rows 4, PG1671; rows 5, PG1649; rows 6, PG1650. (C) Variegation of ade6 expression in mat1-P17::LEU2 and h90 cells. The stability of the "Red on YE" and "White on YE" phenotypes of cells having their sole functional copy of ade6 near mat3-M was assayed by isolating red and white colonies, allowing the cells to divide in rich medium supplemented with adenine (YES) for ~10 generations, and replating them on YE. Cells with a stable mat1-P allele (mat1-P17::LEU2; strains 3–6) or with a switchable mat1 allele (h90; strains 7–10) were used. Rows 1–6, same as in B; h90 strains: rows 7, BP141; rows 8, PG1647; rows 9, PG1681; rows 10, PG1682.

Next, we introduced the S. pombe ade6 gene at the EcoRV site previously used to integrate ura4. As shown in Figure 6B, ade6 placed at the EcoRV site near mat3-M was repressed in wild-type cells and cells with the deletion of the H3 box. The mat3-M proximal deletions of 482 bp and 1185 bp caused an increased expression of ade6, allowing more cells to form colonies on a medium lacking adenine. In addition to their auxotrophic requirement, S. pombe ade6 mutant strains form red colonies on media with low concentrations of adenine. Cells with no deletion or the deletion of the H3 box formed red colonies, consistent with their poor ability to grow in the absence of adenine. Cells with the deletions of 482 or 1185 bp displayed a variegated phenotype and formed both light red and white colonies. This variegation indicated that the level of derepression of the ade6 gene was not the same in all cells and that both the lower and higher levels of expression could be inherited for several generations.

We noticed a difference between h⁹⁰ and mat1-P17::LEU2 cells in their ability to form white colonies on media poor in adenine (Figure 6C and data not shown). Expression of ade6 was consistently higher in mat1-P17::LEU2 colonies than it was in h⁹⁰ colonies. Expression of ura4 at that same location was similar (data not shown): ura4 was more expressed in P cells (mat1-P17::LEU2 or mat1-P) than it was in h⁹⁰ cells or M cells (mat1-Msmt-0 or mat1-M). This difference might be due to cell-type specific chromatin structures allowing a better accessibility of the mat3-M cassette in P cells. The deletions introduced near mat3-M did not diminish mating or the formation of zygotic asci in a wild-type background, indicating they did not affect mating-type switching (data not shown).

Variegation of ade6 expression:
Cells whose sole functional copy of ade6 was placed near mat3-M formed both red and white colonies when plated on a low concentration of adenine. Higher frequencies of white colonies were observed with strains that had the mat3-M flanking deletion of 482 or 1185 bp than in colonies that had no deletion or a deletion limited to the H3 homology box. However, occasional white or sectored colonies arose also in these latter strains. We assayed the stability of the two phenotypes by isolating red and white colonies from these strains (Figure 6C). The two phenotypes were found to interconvert in all strains examined. Reversion to a repressed state was observed more frequently in h⁹⁰ cells than in mat1-P17::LEU2 cells and, conversely, reversion to a derepressed state occurred more frequently in mat1-P17::LEU2 than in h⁹⁰ cells. The two largest deletions, of 482 and 1185 bp, slowed the reestablishment of repression and facilitated conversion from the repressed to the derepressed state.

Epigenetic effects caused by a cis-acting element near mat2-P:
Having observed that mat3-M flanking deletions conferred semistable repressed and derepressed phenotypes that could interconvert, we tested whether the mat2-P flanking deletion of 1.5 kb (THON et al. 1994 ) also had such a property. We examined two phenotypes of strains having a mat2-P flanking deletion: the ability of such strains to express a ura4 gene placed near mat2-P and the ability of unswitchable mat1-Msmt-0 strains with the mat2-P flanking deletion to sporulate. Cells with the (BglII-BssHII)mat2-P(XbaI)::ura4 allele express ura4 weakly; few can form colonies on medium lacking uracil and many form colonies on medium containing FOA (THON et al. 1994 ; Figure 7A). The stability of both Ura⁺ and FOA^R phenotypes was assayed by spotting cells from, respectively, Ura⁺ and FOA^R colonies onto selective plates (Figure 7A). Cells originating from Ura⁺ colonies conserved the ability to grow in the absence of uracil and grew poorly on FOA plates, indicating the derepressed state was stable for many generations. Cells originating from FOA^R colonies gave the same growth pattern as cells originating from complete medium: nearly all cells could form a colony on an FOA plate, indicating ura4 was repressed in these cells, and approximately one cell in a hundred could form a colony on medium lacking uracil, indicating the repressed state was reversible. The sporulation phenotypes, which reflect expression of the mat2-P genes, were consistent with there being fluctuations between an expressed and repressed state of the mat2-P region (Figure 7B). Most colonies grown under nonselective conditions were Spo^-, indicating that the mat2-P genes were repressed in these colonies. Occasional speckles and streaks of sporulation were observed, indicating transient derepression of the P mating-type information. On plates selecting for ura4 expression, high levels of haploid sporulation were observed and, on nonselective sporulation plates, Ura⁺ cells formed colonies containing more haploid asci than colonies originating from Ura^- cells. Hence, the P genes were expressed in a variegated manner and their expression covariegated with the expression of the ura4 gene at the XbaI site.

View larger version (160K): In this window In a new window Download PPT slide   Figure 7. Variegated phenotypes caused by a mat2-P flanking deletion. (A) Variegation of ura4 expression in the (BglII-BssHII)mat2-P(XbaI)::ura4 allele. Expression of ura4 placed at an XbaI site near mat2-P was assayed by spot tests in two strains: a strain with a mat2-P flanking deletion of ~1.5 kb [(BglII-BssHII)mat2-P(XbaI)::ura4: SP1151] and a strain with no deletion [mat2-P(XbaI)::ura4: SP1124]. Tenfold dilutions of three independent cultures of each strain were spotted (top). The stability of the derepressed and repressed states was assayed by spotting on selective media independent cultures of (BglII-BssHII)mat2-P(XbaI)::ura4 cells (SP1151) that had been propagated in the absence of uracil (middle) or in the presence of FOA (bottom). (B) Variegation of sporulation phenotype. The SP1151 cells used above contain a stable mat1-Msmt-0 allele. They were propagated in complete medium (AA) or medium lacking uracil (AA-ura) and plated on sporulation medium containing uracil (MSA) or lacking uracil (MSA-ura). Colonies were allowed to sporulate and stained with iodine to estimate the level of expression of mat2-P.

Range of action of the mat2-P and mat3-M silencing elements:
The mat2-P and mat3-M centromere-proximal silencing elements could both act at a distance on genes placed on the other side of, respectively, the mat2-P and mat3-M cassette. We tested whether these elements could act at an even greater distance by constructing strains with the ura4 gene at the XbaI site near mat2-P and the ade6 gene at the EcoRV site near mat3-M. All strains were of the M mating-type. In these strains, we tested whether deletion of the 1.5-kb mat2-P BglII-BssHII proximal fragment increased expression of ade6 near mat3-M and, conversely, whether deletions near the mat3-M cassette affected expression of ura4 at the mat2-P distal XbaI site. We found that none of the deletions tested affected expression of the distant prototrophic marker (Figure 8). Combining the deletions near mat2-P and mat3-M within the same mating-type region did not increase expression of ura4 near mat2-P or of ade6 near mat3-M.

View larger version (45K): In this window In a new window Download PPT slide   Figure 8. Effect of the mat2-P and mat3-M silencing elements at a distance and in combination. Expression of ura4 and of ade6 was monitored by propagating the indicated strains under nonselective conditions (YES) and spotting 10-fold serial dilutions of cell suspensions on the indicated media. In both the top and bottom, the two first rows of spots are controls where neither ade6 nor ura4 is in the mating-type region. In all other rows, ura4 is at the mat2-P XbaI site and ade6 at the mat3-M EcoRV site. + on the left denotes a mat2-P allele with no deletion and + on the right denotes a mat3-M allele with no deletion. Rows 1 and 7, PG1566; rows 2 and 8, PG1141; row 3, PG1595; row 4, PG1686; row 5, PG1675; row 6, PG1688; row 9, PG1596; row 10, PG1680; row 11, PG1683; row 12, PG1685.

Because deletion of the BglII-BssHII fragment near mat2-P caused variegated expression of the ura4 gene at the mat2-P distal XbaI site and deletion of either 482 or 1185 bp near mat3-M caused variegated expression of ade6 at the EcoRV site near mat3-M, we tested whether expression of the two prototrophic markers covariegated in strains that had both the BglII-BssHII deletion and either the 482- or 1185-bp deletion. First, we spotted cells with the double deletions on medium lacking adenine, medium lacking uracil, and medium lacking both uracil and adenine. Many fewer colonies formed on medium lacking both uracil and adenine than on media lacking either supplement alone, indicating ura4 and ade6 were not expressed at the same time (Figure 8). In a second experiment, (BglII-BssHII)mat2-P(XbaI)::ura4 (482)mat3-M(EcoRV)::ade6 cells and (BglII-BssHII)mat2-P(XbaI)::ura4 (1185)mat3-M(EcoRV)::ade6 cells were propagated on, respectively, medium containing FOA, medium lacking uracil, and medium lacking adenine. As previously noted, cells that had been propagated in the absence of uracil remained able to grow well in the absence of uracil upon replating, whereas cells propagated in the presence of FOA grew poorly in the absence of uracil when replated. In contrast, the ability to grow on medium lacking adenine was not influenced by the absence of uracil or presence of FOA in the prior growth medium (data not shown). Conversely, cells that had been propagated in the absence of adenine had a better efficiency of plating than cells propagated under nonselective conditions when they were plated on medium lacking adenine, but not when they were plated on medium lacking uracil (data not shown). Hence, the expression of ura4 near mat2-P and ade6 near mat3-M appeared to be regulated independently in these strains where both the mat2-P and mat3-M silencing elements were deleted.

Orientation-dependence of the mat3-M silencing element:
Silencing elements described in other systems repress transcription in either orientation (BRAND et al. 1985 ; MORI et al. 1990 ; KASSIS et al. 1991 ; FAUVARQUE and DURA 1993 ; KALLUNKI et al. 1995 ). We tested whether the element near mat3-M had such a property by reintroducing DNA fragments in a strain from which they had been removed in the orientation opposite to the wild-type. Two fragments were tested in this experiment: the centromere-proximal 424- and 126-bp fragments described above. In swi6⁺ backgrounds, the 424-bp fragment was able to restore silencing of ura4 equally well in either orientation, to a level similar to that of mat3-M(EcoRV)::ura4 in which no manipulation had been performed (Figure 9A). The 126-bp fragment also repressed ura4 expression compared with strains that had the 482-bp deletion. It repressed equally well in both orientations, but not as tightly as the 424-bp fragment or as the wild-type element (Figure 9A). We also examined the influence of the orientation of the silencing element on the expression of the M genes from mat3-M by examining sporulation in mat1-P17::LEU2 swi6-115 cells with mat3-M alleles where the 482-bp deletion had been replaced with the 424-bp fragment in either orientation or with the 126-bp fragment in either orientation. In this assay, the DNA fragments having the orientation opposite to wild-type were able to exert a repression, but not as efficiently as when placed in the wild-type orientation (Figure 9B). We conclude from this set of experiments that the silencing elements located on the proximal side of mat3-M can act in both orientations although with different efficiencies, the orientation found in the wild type giving rise to the tightest repression of mat3-M.

View larger version (119K): In this window In a new window Download PPT slide   Figure 9. Orientation dependence of mat3-M silencing element. (A) Effect of the orientation on ura4 expression. Spot tests were performed as described in previous figures with cells having the ura4 gene at the EcoRV site near mat3-M and the indicated deletions. Arrows represent mat3-M centromere-proximal DNA amplified by PCR as in Figure 3 and introduced in place of the 482-bp deletion in the wild type (rows 3 and 5) or reverse (rows 4 and 6) orientation. Rows 1, PG447; rows 2, PG1550; rows 3, PG1626; rows 4, PG1629; rows 5, PG1632; rows 6, PG1630. (B) Effect of the orientation on sporulation. Sporulated colonies of mat1-P17::LEU2 swi6-115 cells with the indicated replacements for the 482-bp fragment flanking mat3-M were stained with iodine vapors. The strains were, from top to bottom: PG1401, PG1527, PG1529, PG1541, and PG1554.

	DISCUSSION

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

We found that a DNA element adjacent to the mat3-M mating-type cassette of fission yeast participated in the repression of transcription of mat3-M and in the repression of prototrophic markers introduced near the cassette. Some of the properties of this element were similar to those of an element adjacent to the mat2-P cassette (THON et al. 1994 ). Deletion of either the mat2-P or mat3-M flanking element increased expression locally in an area that comprised the cassette and chromosomal region close to the deletion, but it did not increase expression in the entire mating-type region. The local derepressions were markedly increased by combining the cis-acting deletions with mutations in the trans-acting factors swi6, clr1, clr2, clr3, and clr4 but not by combining the cis-acting deletions with a mutation in esp3. Finally, simultaneous deletion of the mat2-P and mat3-M flanking elements in stable M cells did not increase expression further than each single deletion. We will discuss these phenotypes in the context of the current understanding of the mode of action of silencers in other organisms and in the context of models for silencing in the mating-type region of S. pombe.

Silencing and mat3-M flanking sequences:
DNA elements close to the mat3-M mating-type cassette have recognizable sequence features that suggest a role in silencing. These elements are as follows: the H3 homology box, which is also found at the mat2-P silent cassette; two S. pombe ARS consensus sequences; and two potential binding sites for the S. pombe DNA-binding proteins Mts1/Mts2. Our deletion analysis allows us to assess the likelihood that these elements participate in the repression of mat3-M.

Because it is present at mat2-P and mat3-M but not at mat1, the H3 homology box has been viewed as a potential silencing element (KELLY et al. 1988 ). A plasmid deletion analysis of the mat2-P flanking regions indicated the H3 box was not required for silencing mat2-P (EKWALL et al. 1991 ). Consistent with this previous study, we found that deleting the H3 box near mat3-M in the chromosome did not derepress the M genes, or other genes introduced near the mat3-M cassette. Because of the redundant nature of silencing, ruling out repression by H3, or any other cis-element, should be viewed with caution. However, deleting H3 did not potentiate the effect of a trans-acting mutation in the silencing factor Swi6, nor did it potentiate the effect of other mat3-M flanking deletions, further indicating that H3 is not a repressor element. Because of its position at the edge of the cassette, another possible function for H3 would be in resolution of mat1 gene conversions.

ARS elements are part of the HML and HMR silencers of S. cerevisiae where they attract ORC proteins (FOX et al. 1997A and references therein). In S. pombe, both perfect and near matches to the sequence (A/T)(A/G)TTTATTTA(A/T) are consistently found within DNA segments allowing autonomous plasmid replication (MAUNDRELL et al. 1988 ). Two 11-bp sequences matching the S. pombe ARS consensus are found on the centromere-proximal side of mat3-M, one 825 bp and the other 1525 bp upstream from the H2-H3 border (nucleotides 10082–10092 and 9372–9382, respectively, in U57841; GREWAL and KLAR 1997 ). In addition, 3 10/11 matches and 21 9/11 matches to the consensus are found within 2 kb of the cassette. The 4.2-kb HindIII genomic fragment containing mat3-M can replicate as an extrachromosomal element (G. THON, unpublished observations), supporting the notion that the putative ARS's are functional. We found that a deletion comprising the ARS consensus sequence closest to the cassette did not derepress mat3-M or flanking markers [(1185)::(424)oriImat3-M and (1185)::(424)oriImat3-M(EcoRV)::ura4 alleles; Figure 3A and data not shown]. That deletion had no effect in the wild type nor in cells partially derepressed by either a mutation in swi6 or by the deletion of 482 bp adjacent to mat3-M. A distinct possibility is that, if the ARS elements play a role in silencing, perfect or near consensus sequences can substitute for them after they are deleted. Hence, deletions larger than those introduced here, or deletion of trans-acting factors such as Abp1 or Abp2 (MURAKAMI et al. 1996 ; HALVERSON et al. 1997 ; SANCHEZ et al. 1998 ), might be required to observe an effect.

The heptamer sequence ATGACGT and the overlapping consensus TGACG(T/A)(A/C) are protein-binding sites well characterized in S. pombe because of their occurrence in the ade6-M26 allele and ability to bind the transcription factor Atf1, respectively. The M26 mutation in the ade6 gene increases meiotic recombination (GUTZ 1971 ) by creating the sequence ATGACGT (PONTICELLI et al. 1988 ; SZANKASI et al. 1988 ; SCHUCHERT et al. 1991 ). The ATGACGT heptamer also creates recombination hot spots when introduced by mutagenesis at various places and in either orientation in the ade6 gene or in the ura4 gene (FOX et al. 1997B ), and this effect on recombination depends on the chromosomal context as shown by transplacements of the ade6-M26 allele (VIRGIN et al. 1995 ). The Atf1/Pcr1 (Mts1/Mts2) heterodimer binds to the M26 heptamer in vitro and is required for hot-spot activity (WAHLS and SMITH 1994 ; KON et al. 1997 ). Atf1 binds in vitro to sequences matching the mammalian ATF-binding site TGACG(T/A)(A/C) (JONES and JONES 1989 ; TAKEDA et al. 1995 ) which overlaps with the M26 heptamer. The Atf1 and Pcr1 proteins function as transcriptional activators of genes expressed during sexual differentiation and genes expressed in response to stress (TAKEDA et al. 1995 ; KANOH et al. 1996 ; SHIOZAKI and RUSSELL 1996 ; WATANABE and YAMAMOTO 1996 ; WILKINSON et al. 1996 ). However, Mts1 and Mts2 do not increase recombination at the M26 site by increasing transcription, but rather by an unknown mechanism, indicating these proteins are multifunctional (KON et al. 1997 ). Two sequences matching the ATF/M26 consensus (ATGACGTA) are found near mat3-M, one 138 bp and one 1522 bp upstream from the H2/H3 junction. We found that an 88-bp fragment containing the consensus sequence closest to the cassette could partially restore silencing when introduced in a larger mat3-M flanking deletion [(482)::(88)oriImat3-M allele; Figure 3B], indicating that Mts1/Mts2 may have a role in silencing in addition to activating transcription and recombination.

Orientation dependence of silencing elements:
Two DNA fragments adjacent to mat3-M, a 482-bp and a 126-bp fragment, could exert a repression when placed at their natural location but in the reverse orientation compared to the wild type. In that reverse or the normal orientation, each fragment repressed the ura4 gene placed distal to mat3-M as judged by our plating assay. However, in a swi6-115 background, the wild-type orientation silenced the M genes more effectively than the reverse orientation. These different results could reflect properties of the markers used (ura4 vs. M genes) or an effect of the background (swi6⁺ vs. swi6-115). Several propositions can explain the ability of a silencer to work in both orientations, yet preferentially in one. One possibility is that the element orients the nucleation of a protein complex that propagates preferentially in one direction. Another possibility is that the position or orientation of the silencer relative to other elements that were not inverted in the experiment is important, with the wild-type arrangement allowing optimal interactions between the factors attracted to the region. Silencing elements from other organisms can also exert a repression in both orientations although with different efficiencies (BRAND et al. 1985 ; SHEI and BROACH 1995 ).

Redundancy of silencing in the mating-type region of S. pombe:
Previous observations have suggested that more than one pathway of repression acts in the mating-type region. First, mutations and deletions in the class of trans-acting factors that include swi6, rik1, clr1, clr2, clr3, and clr4 derepress transcription only partially (THON and KLAR 1992 ; EKWALL and RUUSALA 1994 ; ALLSHIRE et al. 1995 ; THON and FRIIS 1997 ). Mutations in swi6, clr1, clr2, clr3, and clr4 combined pairwise do not cause a more pronounced derepression than any single mutation, indicating these factors work in a common pathway (THON et al. 1994 ). Another class of trans-acting factors, encoded by the esp genes, acts in synergy with swi6 (THON and FRIIS 1997 ) and the four clr genes (G. THON, unpublished observations) and thereby defines a pathway of silencing parallel to the pathway mediated by swi6, clr1, clr2, clr3, and clr4. The study of cis-acting elements, in particular the characterization presented here, also points to several silencing mechanisms. Neither the 7.5-kb deletion in the K region, nor deletion of the mat2-P proximal element nor deletion of the mat3-M element fully derepresses transcription in the mating-type region, indicating these elements can partially substitute for each other (THON et al. 1994 ; THON and FRIIS 1997 ; this study). Combining trans-acting mutations with the deletions of cis-acting elements gives some clues to the possible interactions between cis- and trans-acting factors. Strains that have both the 7.5-kb deletion between mat2-P and mat3-M and a mutation in swi6, clr1, clr2, clr3, or clr4 display a partially derepressed phenotype similar to the phenotype caused by the trans-acting mutations alone, whereas strains that contain the cis-acting 7.5-kb deletion in combination with trans-acting mutations in the esp genes are strongly derepressed (THON and FRIIS 1997 ). In contrast, mat2-P or mat3-M flanking deletions have a pronounced cumulative effect with mutations in swi6, clr1, clr2, clr3, or clr4 (THON et al. 1994 ; this study), but not with the esp mutation whose effect was tested here (esp3-1). This suggests that the products of swi6, clr1, clr2, clr3, and clr4 interact with the mating-type region via the K region whereas the esp products interact with the mat2-P and mat3-M flanking elements.

Epigenetic switches between repressed and derepressed states:
A remarkable phenotype of cells in which cis-acting sequences have been deleted from the mating-type region is that they switch between two epigenetic states: one in which the mating-type region is partially derepressed and one in which a level of repression similar to the wild type is attained. This phenomenon was previously observed in cells with a 7.5-kb deletion between mat2-P and mat3-M (GREWAL and KLAR 1996 ; THON and FRIIS 1997 ). We show here that similar phenotypic variations occur when other cis-acting elements are deleted. The mat2-P genes as well as a ura4 gene placed near mat2-P were derepressed in a variegated manner in cells lacking the mat2-P flanking element. Similarly, expression of the mat3-M genes and of an ade6 marker placed near mat3-M variegated in cells with mat3-M flanking deletions. A two-step model in which silencing is established by the interactions between cis-acting elements and trans-acting factors and subsequently maintained and inherited by a different process could account for these variegated phenotypes. Similar models have been proposed for other systems, in particular for silencing of the HML and HMR mating-type cassettes in S. cerevisiae (PILLUS and RINE 1989 ; MAHONEY et al. 1991 ; SUSSEL et al. 1993 ; HOLMES and BROACH 1996 ). In such models, two types of mutations could cause switches between a repressed and derepressed epigenetic state: mutations decreasing the rate of establishment and mutations decreasing the fidelity of inheritance of the silenced state. We propose an extension to this model that accommodates the existence of two silencing pathways and indicates how the loss of each cis-acting element could diminish the rate of establishment without preventing complete silencing from occurring. In this model, protein-protein interactions could substitute for protein interactions with bona fide silencing elements to attract the components required for efficient silencing, albeit in an inefficient fashion. Cells containing only some of the cis-acting elements could remain for several generations in a partially derepressed state where the interactions occurring normally via the cis-acting elements present would have been established, but not all interactions required for full silencing. Occasionally, an