- Email: [email protected]
A locus control region regulates yeast recombination
REVIEWS
T
he yeast mating-type system has provided models for many fundamental problems in genetics and molecular biology. The isolation of sterile mutants of yeast1 led to the elucidation of a G-protein-mediated signal transduction system responding to small peptide pheromones2. Studies of control of the expression of the HO gene led to the identification of SWI genes involved in general and cell-cycle-specific transcriptional activation3 and to the discovery of localization within the daughter cell of the Ash1p transcriptional repressor protein and its mRNA (Refs 4, 5). Interest in mating-type control of sex-specific genes led to the detailed analysis of the multifunctional homeodomain protein Mat␣2p (Ref. 6). The presence of unexpressed donor loci stimulated an intensive study of gene silencing, its relationship to changes in chromatin structure and the involvement of DNA replication in its establishment and maintenance7,8. Analysis of the recombination mechanism that switches one mating-type to the other has provided a wealth of information about homologous recombination induced by double-strand breaks9. Most recently, the study of how yeast decide to choose one donor locus over another has revealed a remarkable mechanism of activation and inactivation of an entire chromosome arm10–12.
Mating-type conversion
A locus control region regulates yeast recombination JAMES E. HABER ([email protected]) The yeast Saccharomyces can switch its mating type by a highly choreographed recombination event in which ‘a’ or ‘␣’ sequences at the mating-type (MAT) locus are replaced by opposite mating-type sequences copied from one of two donors, HML and HMR, located near the two ends of the same chromosome III. MAT␣ cells ‘know’ to choose HML, while MAT␣ cells preferentially recombine with HMR. Donor preference is regulated by a 250 bp recombination enhancer, that controls recombination of the entire left arm of chromosome III. Recent studies have shown how this locus-control region is turned on and off. information located in two silent loci, HML␣ and HMRa*, located roughly 200 and 100 kb away from MAT, respectively, each near one of the telomeres of chromosome III (Fig. 1). HML and HMR have a chromatin structure that prevents transcription and is highly refractory to endonuclease cleavage, including the sitespecific HO endonuclease. Recent studies demonstrate that these regions have a highly ordered, but discontinuous nucleosome array30 and are more negatively supercoiled13. Expression of the HO endonuclease triggers MAT switching by creating a double-strand break (DSB) at MAT (reviewed in Refs 9, 14, 15). Cleavage does not occur at identical recognition sites in the silent loci, which are therefore intact and can serve as donors to repair the DSB at MAT. The DSB is processed by a 5⬘ to
The conversion of one mating-type to the other involves the replacement of the ~700 bp portion of the MAT locus (Ya or Y␣) that determines the cell’s sexual identity (Fig. 1). MAT␣ cells, carrying Y␣, encode regulatory proteins Mat␣1p and Mat␣2p. Mat␣1p combines with Mcm1p to ‘turn on’ a set of ␣-specific genes, including the pheromone, MF␣, and the receptor for the opposite mating type’s pheromone, MFa (Fig. 2). Mat␣2p also combines with Mcm1p to act as a repressor to ‘turn off’ a-specific genes. MATa cells express a-specific genes, such as the pheromone MFa and several other genes needed in its processing, as well as a receptor to sense the pheromone of the opposite sex, (a) MF␣. The expression of a-specific a1 genes also depends on transcripRE Ya EL IL IR ER tional activation by the multi-funcMATa Ya HMRa ON tional Mcm1p. Cells lacking any mating-type gene also express aHO endonuclease cleavage specific genes. Mata1p is required only when cells conjugate to make (b) a MATa/MAT␣ diploid, where α2 α1 RE Ya Mata1p and Mat␣2p combine to repress haploid-specific genes, Yα HMRa OFF rendering cells non-mating and allowing meiosis to occur. To replace MAT Y sequences, the cell must remove the original FIGURE 1. Mating-type switching: switching from (a) MATa or (b) MAT␣. An HO Y region and copy the new Y endonuclease-induced double-strand break at MAT initiates gene conversion/replacement sequences from another template. of the Ya region with Y␣ sequences copied from HML␣. MAT␣ cells express Mat␣1p and These are provided in the form of Mat␣2p regulatory proteins, while MATa encodes Mata1p. HML␣ and HMRa contain complete copies of mating-type genes but are not expressed because of the silencing unexpressed copies of a and ␣ imposed through the adjacent E and I silencer sequences that organize a repressed *In fact, both HML and HMR can carry either a or ␣ information. Most wild-type cells carry HML␣ and HMRa.
chromatin structure (indicated by hatched lines). HML shares more sequences (regions W, X, Z1 and Z2) with MAT than does HMR (regions X and Z1). MATa strains preferentially recombine with HML␣, even if HMR also contains Y␣ instead of Ya. Donor preference is dependent on a 244 bp cis-acting recombination enhancer (RE). TIG AUGUST 1998 VOL. 14 NO. 8
Copyright © 1998 Elsevier Science Ltd. All rights reserved. 0168-9525/98/$19.00 PII: S0168-9525(98)01501-7
317
REVIEWS suggests that MAT switching is likely to involve both leading- and lagging-strand synthesis (A. Holmes and J.E. Haber, unpublished).
Choosing a donor: HML or HMR?
FIGURE 2. Regulation of mating-type-specific genes. Transcription of a-specific genes depends on Mcm1p, but is repressed by binding of the Mat␣2p–Mcm1p co-repressor to the same region. Expression of ␣-specific genes requires Mat␣1p and Mcm1p. Mat␣2p and Mata1p combine to turn off haploid-specific genes.
3⬘ exonuclease to create a long 3⬘-ended tail that can invade one of the donors in the Z1 region and begin copying the adjacent Y sequences that will replace the original Y region at MAT (Fig. 1). MAT switching is a surprisingly slow process, requiring an hour for completion. Switching requires a large number of recombination proteins, as well as proteins involved in DNA replication, for example, the single-strand-binding protein RPA, the ‘clamp’ protein PCNA and several DNA polymerases. Recent evidence from our laboratory (a)
All of this is complicated enough, but we still have not considered the most mysterious aspect of this process: the ability of the yeast cell to choose between the donor loci. It makes sense that MATa should somehow seek out and recombine with HML␣ rather than HMRa, so that the recombinational repair of the DSB will lead to a switch to the opposite mating type. Klar et al.16 found that donor selection is not dictated by the Ya or Y␣ content of the HM loci: a strain with reversed silent information (HMLa MATa HMR␣) still chooses HML. Weiler and Broach17 showed that replacing the entire HML region, including its silencers, with a cloned HMR locus does not change donor preference. So, it is the location of the donor, not the sequence differences between HML and HMR, that directs donor selection. There must, therefore, be one or more cis-acting sequences that activate or repress one or both donors. In fact, there appear to be two donor preference mechanisms at work, summarized in Fig. 3. MATa cells activate a donor on the left arm of chromosome III to be selected preferentially, although the other donor at HMR is available10,18,19. In contrast, MAT␣ cells inactivate donors on the left arm, making HMR essentially the only available donor. Thus a MATa cell deleted for HML can easily use HMR, but approximately a third of MAT␣ cells die when their only choice of a donor is HML (Refs 18, 19). The failure of many MAT␣ cells to use the ‘wrong’ donor occurs despite the fact that cells experiencing a DSB become arrested at a G2–M checkpoint. This should, theoretically, allow cells time to locate a donor and repair the DSB by gene conversion, and suggests that HML is indeed very inaccessible in MAT␣ cells.
A recombination enhancer
(b)
(c)
(d)
FIGURE 3. Donor preference is regulated by the recombination enhancer RE. (a) In MATa cells, a donor placed anywhere along the left arm of chromosome III is used preferentially in competition with HMR. This activation depends on the presence of the RE sequence. (b) In MAT␣ cells, the RE is inactivated and the use of HML (or other donors on the left arm) is strongly repressed, allowing HMR to be the preferred donor. Mutations within RE that prevent its inactivation result in increased use of HML. (c) Similarly, in MATa cells deleted for RE, donors on the left arm of chromosome III are inaccessible and HMR is again the preferred donor. (d) A similar mating-type-dependent difference is also seen for spontaneous recombination between a leu2-R allele that replaces HML and its adjacent silencer sequences and a leu2-K allele placed near MAT. TIG AUGUST 1998 VOL. 14 NO. 8
318
The key to understanding the mechanism of donor preference was provided by the yeast genome sequencing project, which chose chromosome III to be sequenced first20. The availability of sequence information made it possible to create terminal truncations (by including a telomere sequence) or internal deletions in order to pinpoint cis-acting elements responsible for the preference of MATa cells for HML (Ref. 10). A 2.5 kb deletion located 17 kb proximal to HML completely reverses donor preference, so that a MATa cell now uses HML only 10% of the time instead of 90%. By re-inserting subfragments of this deleted region, Wu and Haber10 defined a 700 bp ‘recombination enhancer’ (RE) sequence that controls MATa and
REVIEWS MAT␣ donor preference (Fig. 4b). (a) The 2.5 kb region has no open α2–Mcm1p reading frames, between the flankKAR4 RE YCL054w binding site ing KAR4 and YCL054w genes. In MATa cells, the RE acts not only to activate HML at its normal (b) location but also to activate either HML or HMR sequences placed at several different locations along the entire left arm of chromosome III (Refs 10, 19, 21) (Fig. 3). In each (c) A B C D case, if RE is deleted, HMR becomes the strongly preferred donor, just as α2–Mcm1p–α2 (TTT[A/G])10 in MAT␣ cells10,11,19. MAT␣ cells do binding site not activate HMR; rather they inactivate or sequester HML, leaving a-specific RNAs HMR as the only available donor. (d) The inaccessibility of HML in MAT␣ cells is also seen for donors inserted F1 HS F2 at other sites along the entire left arm of chromosome III and even on (e) the right arm, to a point about 25 kb proximal to MAT (Ref. 18). Thus the left arm is recombinationally inactive in MAT␣ and somehow α2–Mcm1p–α2 activated in MATa. The ‘cold’ state co-repressor in MAT␣, whether RE is present or not, is also the default state in a MATa cell deleted for RE. There- FIGURE 4. Analysis of RE. Deletion of a 2.5 kb region containing no open reading frames fore, RE acts to reverse the unusual results in loss of RE activity. (a) The region contains the RE as well as an Mat␣2p–Mcm1p inaccessibility of donors on the left binding site that has a minor role in regulating RE. (b) Insertion of subfragments derived from this region restores most RE activity. (c) A comparison of the same region in arm of chromosome III. S. cerevisiae and S. carlsbergensis yields a 244 bp region with four conserved subdomains. The regulation of recombination (d) The chromatin structure of RE in MATa cells shows several distinct protein footprints on the left arm of chromosome (F1 and F2) and an unusual DNase I hypersensitive (HS) region that overlaps a run of III is not specific to HO-mediated TTT[A/G] repeats. There are two weak a-specific transcripts that arise in this region. Their recombination nor to silenced donor exact end-points have not yet been mapped. (e) In MAT␣ cells, the entire RE region is sequences. This was shown by repressed by highly positioned nucleosomes that extend from a Mat␣2p–Mcm1p binding examining spontaneous recombina- site in RE across the entire interval between the two adjacent open reading frames, whose tion between a leu2-R allele inserted chromatin structure is not different in MATa or MAT␣ cells. in place of HML with a leu2-K allele placed near MAT (Fig. 3d). The rate of Leu⫹ spontaneous pers. commun.). Third, even when the HML␣ locus is recombinants arising by gene conversion between these unsilenced, by deleting the adjacent E and I silencers, two mutant alleles was 30 times higher in MATa cells and has a much more accessible chromatin structure, it than in MAT␣ (Refs 10, 21), even when the leu2-K allele fails to be used efficiently as a donor in MAT␣ cells18. is placed on another chromosome. This difference is lost If the RE does not exert a direct effect on the local when the RE is deleted, and the rate of Leu⫹ recombi- chromatin structure of the donor locus, what does it do? nation in MATa falls to the level seen in MAT␣ cells. This The idea we favor is that the RE changes the nuclearmating-type-dependent difference in Leu⫹ recombi- localization or the higher-order folding of the entire left nation depends on one of the two leu2 alleles being situ- arm of chromosome III to make it more ‘flexible’ in ated in the region normally occupied by HML; there is no locating and pairing with the recipient site in MATa such difference if leu2-R is inserted in place of HMR and cells. In this view, the chromosome arm would be leu2-K is near MAT. sequestered or immobilized (perhaps by being bound to the nuclear envelope) in such a way that HML is Mechanism of RE function unavailable in MAT␣ cells even though the chromatin The RE does not appear to act by regulating the structure at HML itself is unchanged. One observation chromatin structure of donor locus on the left arm of favoring this idea comes from measuring Leu⫹ recombichromosome III. First, there is no significant mating- nation between leu2-R located in place of HML but with type-dependent difference in mRNA levels for the LEU2 the leu2-K allele on a plasmid, so that the latter partner gene inserted in place of HML (which shows a 30-fold in the recombination reaction is presumably more recombination difference). Second, a direct exami- mobile and thus more able to ‘find’ leu2-K than would nation of the chromatin structure of the HML locus by be the same allele in the middle of an entire chromomicrococcal nuclease and DNase I digests does not some. As predicted the mating-type-dependent differreveal any significant mating-type-dependent difference ence is reduced from 30-fold to only 3-fold, supporting in chromosome structure (K. Weiss and R.T. Simpson, the idea that the constraint on recombination in MAT␣ TIG AUGUST 1998 VOL. 14 NO. 8
319
REVIEWS cells reflects a difficulty in bringing the two alleles together rather than the inherent inability of the leu2-R allele to recombine (M. Barlow, W-Y. Leung and J.E. Haber, unpublished). A complete understanding of the action of RE also needs to account for the discovery that the 700 bp RE is ‘portable,’ at least on chromosome III. RE still activates HML if the enhancer is deleted from its normal location 17 kb from the left end of the chromosome and a 700 bp RE fragment is inserted 70 kb from the end (G-F. Richard and J.E. Haber, unpublished). Moreover, in a strain carrying its normal RE, insertion of an additional RE adjacent to HMR increases its use in MATa cells to nearly 50%, from about 10% (Ref. 10). One explanation might be that all of chromosome III is unusual, and that even the right arm is somewhat ‘cold’ and thus responsive to the RE. Other chromosomes might not respond to RE.
How is RE regulated? By comparing the RE sequences of S. cerevisiae and S. carlsbergensis (which is functional in S. cerevisiae), it has been possible to narrow down the RE to 244 bp (Fig. 4c), within which are four well-conserved subdomains12. Deletion of regions A, C or D each abolishes RE activity. The most striking features of the conserved domains are ten repeats of TTT[A/G] in region D and a highly conserved binding site for the Mat␣2p–Mcm1p co-repressor in region C. The Mat␣2p–Mcm1p binding site in subdomain C appears to play a key role in repressing the activity of RE in MAT␣ cells. Mat␣2p and Mcm1p, in conjunction with Tup1p and to a lesser extent SSN6p, acts as a repressor of a-specific gene transcription22. Szeto et al.11 have recently reported that mutations of the two Mat␣2p-binding domains within this site are sufficient to alter donor preference in MAT␣, so that HML is used 50% of the time instead of 10%. This result also suggests that any a-specific gene products are unlikely to play a critical role in activating RE, because these genes should still be repressed by Mat␣2p–Mcm1p in a MAT␣ cell. However, the difference between HML use in MATa (80%) and MAT␣ (50%) in this mutant RE could still be attributed to a-specific genes. The genetic studies of how RE is regulated are strongly supported by analysis of the chromatin structure of the RE. In MAT␣ cells, the RE is covered by a very highly positioned set of nucleosomes on either side of the (occupied) Mat␣2p–Mcm1p binding site23. This ordered array extends across the entire 2.5 interval containing RE, and ends at the two flanking open reading frames. In MATa cells, the RE exhibits several distinctive footprints indicative of protein binding and a notable region of closely spaced DNase I hypersensitive (but not single stranded) sites that cover a region of TTT[A/G] repeats. In MAT␣ cells, when mutations of the Mat␣2p binding sites of the Mat␣2p–Mcm1p binding domain markedly increase the use of HML as a donor, the chromatin structure of the RE region becomes indistinguishable from that seen in MATa cells21. There is a second Mat␣2p–Mcm1p binding sequence located within the 2.5 interval between KAR4 and YCL054w (Fig. 4). Results of Wu and Haber10, and Wu et al.12 suggest that this second site is not important, as it is lacking in the 700 bp and other RE constructs inserted
into a deletion of the entire region, but Szeto et al.11 find that deleting this site increases the use of HML in MAT␣ strains. The second site might help stabilize the extended nucleosome array emanating from RE in MAT␣ cells.
Activation of RE in MATa cells With the exception of the RE, all well-characterized Mat␣2p–Mcm1p binding sites are located just upstream of a-specific genes. The RE does not contain an open reading frame, but there are two ‘sterile’ transcripts within or near the RE region that are transcribed in MATa, but not MAT␣ (Ref. 11). It is unlikely, however, that the sequence of the RNA transcript is important for RE activity, because truncations of RE that remove most of the normally transcribed sequence have full or substantial activity10,12. However, the act of transcription itself could still be the key regulatory feature. The existence of such transcripts reminds one of Xinactivation in mammals and their importance needs to be ascertained. The activation of a-specific genes depends on Mcm1p, which can bind to part of the Mat␣2p–Mcm1p binding site that also represses a-specific genes in MAT␣ cells22. Szeto et al.11 showed that the Mat␣2p–Mcm1p operator, fused to a reporter gene, will indeed activate transcription in MATa cells. Further evidence, establishing a definitive role for Mcm1p in activating RE has come from recent experiments by Wu et al.12. A 2 bp mutation that eliminates Mcm1p binding in the Mat␣2p–Mcm1p operator sequence also abolishes MATa donor preference so that HML is used only 10–20% of the time. Moreover, an Arg87Ala mutation in Mcm1p similarly reduces activation of HML in MATa cells and increases HML use in MAT␣ cells. This residue has recently been shown by X-ray crystallography to be an important contact between Mcm1p and Mat␣2p (Ref. 24). So it would appear that Mcm1p is a central player in activating RE. Whether Mcm1p has a co-activating protein is not yet known, nor is it clear that transcription per se is required. One very surprising result that has emerged from the analysis of the RE mutated in the Mcm1p binding site is that the chromatin structure of the mutant RE, even in MATa cells (where there is no Mat␣2p), has a highly organized nucleosome structure that is very similar to what is seen in normal MAT␣ cells. Apparently, other sequences within RE can organize a phased nucleosome structure in the absence of Mat␣2p–Mcm1p binding, although the repressor proteins seem to position more precisely and ‘lock in’ the repressing chromatin structure. These results might explain why a 2 bp mutation of the Mcm1p binding site has a more severe effect on HML usage than does a 25 bp deletion of most of the Mat␣2p–Mcm1p operator11,12.
Future perspectives The stage is set to delve more deeply into the remarkable properties of this system. There are really two different and fascinating phenomena here: (a) the unusual ‘coldness’ of the left arm in MAT␣ cells, or in MATa cells deleted for the RE; and (b) the reversal of this inhibition and the activation of the left arm for recombination when the RE is active. First, we need to find the cis-acting sites responsible for the unusual
TIG AUGUST 1998 VOL. 14 NO. 8
320
REVIEWS inactivity of the arm in the absence of RE. There might be several such sites along the left arm of chromosome III. Indeed a deletion of the first 40 kb of the left arm does not eliminate the ‘coldness’ of a donor located more proximally in MAT␣ cells. Second, we must identify additional trans-acting mutations that alter this process. While Mcm1p is clearly a very important part of the activation of RE, the entire 244 bp RE is more than 10 times larger than the size of an Mcm1p binding domain needed to activate transcription, and mutations in subdomains A and D are as profound as preventing Mcm1p binding. So far, only the CHL1 gene has been shown to affect donor preference25. A deletion of this gene, which also affects chromosome transmission of all yeast chromosomes, reduces donor preference in MATa switching but has no effect on MAT␣. What the connection is between chromosome stability and donor preference is not yet obvious. Now that the RE has been narrowed down to a small size it should be possible to identify DNA-binding proteins directly and to use this binding as an assay in characterizing other mutations affecting donor preference. Finally, we are left with the question of whether the elaborate mechanism(s) that Saccharomyces evolved to ensure donor preference have been adapted by other organisms to solve other developmental problems. One possible analogy is found in dosage compensation of metazoans. In Drosophila, the single X chromosome of a male is activated for transcription at apparently many sites26, which might act similarly to the RE. In Caenorhabditis, the problem is solved by a partial condensation of the two X chromosomes of the female to reduce transcription27, which might share some features of inactivation with the ‘coldness’ of yeast’s sex chromosome in MAT␣. (It is also useful to remember that an entire 100 kb chromosome arm in yeast is the length of a large gene in Drosophila!) Another way of thinking about RE is that it is a locus control region, which regulates recombination and not transcription. It will also be interesting to see what relationship RE has to locus control regions in mammals that activate gene expression and even DNA replication28,29.
18 Wu, X., Moore, J.K. and Haber, J.E. (1996) Mol. Cell. Biol. 16, 657–668 19 Wu, X., Wu, C. and Haber, J.E. (1997) Genetics 147, 399–407 20 Oliver, S.G. et al. (1992) Nature 357, 38–46 21 Wu, X. and Haber, J.E. (1995) Genes Dev. 9, 1922–1932 22 Szeto, L. and Broach, J.R. (1997) Mol. Cell. Biol. 17, 751–759 23 Weiss, K. and Simpson, R.T. (1997) EMBO J. 16, 4352–4360 24 Tan, S. and Richmond, T.S. (1998) Nature 391, 660–666 25 Weiler, K.S., Szeto, L. and Broach, J.R. (1995) Genetics 139, 1495–1510 26 Bashaw, G.J. and Baker, B.S. (1996) Curr. Opin. Genet. Dev. 6, 496–501 27 Chuang, P.T., Lieb, J.D. and Meyer, B.J. (1996) Science 274, 1736–1739 28 Capone, M. et al. (1993) EMBO J. 12, 4335–4346 29 Aladjem, M.I. et al. (1995) Science 270, 815–819
Reference added in proof 30 Weiss, K. and Simpson, R.T. (1998) Mol. Cell. Biol. (in press) J.E. Haber is in the Rosenstiel Center, the Keck Institute for Cellular Visualization and the Department of Biology, Brandeis University, Waltham, MA 02254, USA.
trends in
GENETICS Included in the September issue Chimeras and mosaics in mouse mutant analysis by J. Rossant and A. Spence Changing styles in C. elegans genetics by J. Hodgkin and R.K. Herman
References 1 Mackay, V. and Manney, T.R. (1974) Genetics 76, 255–271 2 Leberer, E., Thomas, D.Y. and Whiteway, M. (1997) Curr. Opin. Genet. Dev. 7, 59–66 3 Peterson, C.L. (1996) Curr. Opin. Genet. Dev. 6, 171–175 4 Long, R.M. et al. (1997) Science 277, 383–387 5 Takizawa, P.A. et al. (1997) Nature 389, 90–93 6 Wolberger, C. (1996) Curr. Opin. Struct. Biol. 6, 62–68 7 Loo, S. and Rine, J. (1994) Science 264, 1768–1671 8 Sherman, J.M. and Pillus, L. (1997) Trends Genet. 13, 308–813 9 Haber, J. (1995) BioEssays 17, 609–620 10 Wu, X. and Haber, J.E. (1996) Cell 87, 277–285 11 Szeto, L. et al. (1997) Genes Dev. 11, 1899–1911 12 Wu, C. et. al. (1998) Genes Dev. 12, 1726–1737 13 Bi, X. and Broach, J.R. (1997) Mol. Cell. Biol. 17, 7077–7787 14 Strathern, J.N. (1989) Genetic Recombination (Kucherlapati, R. and Smith, G.R., eds), pp. 445–464, American Society for Microbiology 15 Haber, J.E. (1992) Trends Genet. 8, 446–452 16 Klar, A.J., Hicks, J.B. and Strathern, J.N. (1982) Cell 28, 551–561 17 Weiler, K.S. and Broach, J.R. (1992) Genetics 132, 929–942
Survival with HIV infection: good luck or good breeding? by S.L. Rowland-Jones The Drosophila genome project: a progress report by G.M. Rubin Traps to catch unwary oncogenes by A-O. Hueber and G.I. Evan Selfishness and death... raison d’être of restriction, recombination and mitochondria by I. Kobayashi Who pulls the string to pattern cell division in Drosophila? by H. Skaer
TIG AUGUST 1998 VOL. 14 NO. 8
321
T
he yeast mating-type system has provided models for many fundamental problems in genetics and molecular biology. The isolation of sterile mutants of yeast1 led to the elucidation of a G-protein-mediated signal transduction system responding to small peptide pheromones2. Studies of control of the expression of the HO gene led to the identification of SWI genes involved in general and cell-cycle-specific transcriptional activation3 and to the discovery of localization within the daughter cell of the Ash1p transcriptional repressor protein and its mRNA (Refs 4, 5). Interest in mating-type control of sex-specific genes led to the detailed analysis of the multifunctional homeodomain protein Mat␣2p (Ref. 6). The presence of unexpressed donor loci stimulated an intensive study of gene silencing, its relationship to changes in chromatin structure and the involvement of DNA replication in its establishment and maintenance7,8. Analysis of the recombination mechanism that switches one mating-type to the other has provided a wealth of information about homologous recombination induced by double-strand breaks9. Most recently, the study of how yeast decide to choose one donor locus over another has revealed a remarkable mechanism of activation and inactivation of an entire chromosome arm10–12.
Mating-type conversion
A locus control region regulates yeast recombination JAMES E. HABER ([email protected]) The yeast Saccharomyces can switch its mating type by a highly choreographed recombination event in which ‘a’ or ‘␣’ sequences at the mating-type (MAT) locus are replaced by opposite mating-type sequences copied from one of two donors, HML and HMR, located near the two ends of the same chromosome III. MAT␣ cells ‘know’ to choose HML, while MAT␣ cells preferentially recombine with HMR. Donor preference is regulated by a 250 bp recombination enhancer, that controls recombination of the entire left arm of chromosome III. Recent studies have shown how this locus-control region is turned on and off. information located in two silent loci, HML␣ and HMRa*, located roughly 200 and 100 kb away from MAT, respectively, each near one of the telomeres of chromosome III (Fig. 1). HML and HMR have a chromatin structure that prevents transcription and is highly refractory to endonuclease cleavage, including the sitespecific HO endonuclease. Recent studies demonstrate that these regions have a highly ordered, but discontinuous nucleosome array30 and are more negatively supercoiled13. Expression of the HO endonuclease triggers MAT switching by creating a double-strand break (DSB) at MAT (reviewed in Refs 9, 14, 15). Cleavage does not occur at identical recognition sites in the silent loci, which are therefore intact and can serve as donors to repair the DSB at MAT. The DSB is processed by a 5⬘ to
The conversion of one mating-type to the other involves the replacement of the ~700 bp portion of the MAT locus (Ya or Y␣) that determines the cell’s sexual identity (Fig. 1). MAT␣ cells, carrying Y␣, encode regulatory proteins Mat␣1p and Mat␣2p. Mat␣1p combines with Mcm1p to ‘turn on’ a set of ␣-specific genes, including the pheromone, MF␣, and the receptor for the opposite mating type’s pheromone, MFa (Fig. 2). Mat␣2p also combines with Mcm1p to act as a repressor to ‘turn off’ a-specific genes. MATa cells express a-specific genes, such as the pheromone MFa and several other genes needed in its processing, as well as a receptor to sense the pheromone of the opposite sex, (a) MF␣. The expression of a-specific a1 genes also depends on transcripRE Ya EL IL IR ER tional activation by the multi-funcMATa Ya HMRa ON tional Mcm1p. Cells lacking any mating-type gene also express aHO endonuclease cleavage specific genes. Mata1p is required only when cells conjugate to make (b) a MATa/MAT␣ diploid, where α2 α1 RE Ya Mata1p and Mat␣2p combine to repress haploid-specific genes, Yα HMRa OFF rendering cells non-mating and allowing meiosis to occur. To replace MAT Y sequences, the cell must remove the original FIGURE 1. Mating-type switching: switching from (a) MATa or (b) MAT␣. An HO Y region and copy the new Y endonuclease-induced double-strand break at MAT initiates gene conversion/replacement sequences from another template. of the Ya region with Y␣ sequences copied from HML␣. MAT␣ cells express Mat␣1p and These are provided in the form of Mat␣2p regulatory proteins, while MATa encodes Mata1p. HML␣ and HMRa contain complete copies of mating-type genes but are not expressed because of the silencing unexpressed copies of a and ␣ imposed through the adjacent E and I silencer sequences that organize a repressed *In fact, both HML and HMR can carry either a or ␣ information. Most wild-type cells carry HML␣ and HMRa.
chromatin structure (indicated by hatched lines). HML shares more sequences (regions W, X, Z1 and Z2) with MAT than does HMR (regions X and Z1). MATa strains preferentially recombine with HML␣, even if HMR also contains Y␣ instead of Ya. Donor preference is dependent on a 244 bp cis-acting recombination enhancer (RE). TIG AUGUST 1998 VOL. 14 NO. 8
Copyright © 1998 Elsevier Science Ltd. All rights reserved. 0168-9525/98/$19.00 PII: S0168-9525(98)01501-7
317
REVIEWS suggests that MAT switching is likely to involve both leading- and lagging-strand synthesis (A. Holmes and J.E. Haber, unpublished).
Choosing a donor: HML or HMR?
FIGURE 2. Regulation of mating-type-specific genes. Transcription of a-specific genes depends on Mcm1p, but is repressed by binding of the Mat␣2p–Mcm1p co-repressor to the same region. Expression of ␣-specific genes requires Mat␣1p and Mcm1p. Mat␣2p and Mata1p combine to turn off haploid-specific genes.
3⬘ exonuclease to create a long 3⬘-ended tail that can invade one of the donors in the Z1 region and begin copying the adjacent Y sequences that will replace the original Y region at MAT (Fig. 1). MAT switching is a surprisingly slow process, requiring an hour for completion. Switching requires a large number of recombination proteins, as well as proteins involved in DNA replication, for example, the single-strand-binding protein RPA, the ‘clamp’ protein PCNA and several DNA polymerases. Recent evidence from our laboratory (a)
All of this is complicated enough, but we still have not considered the most mysterious aspect of this process: the ability of the yeast cell to choose between the donor loci. It makes sense that MATa should somehow seek out and recombine with HML␣ rather than HMRa, so that the recombinational repair of the DSB will lead to a switch to the opposite mating type. Klar et al.16 found that donor selection is not dictated by the Ya or Y␣ content of the HM loci: a strain with reversed silent information (HMLa MATa HMR␣) still chooses HML. Weiler and Broach17 showed that replacing the entire HML region, including its silencers, with a cloned HMR locus does not change donor preference. So, it is the location of the donor, not the sequence differences between HML and HMR, that directs donor selection. There must, therefore, be one or more cis-acting sequences that activate or repress one or both donors. In fact, there appear to be two donor preference mechanisms at work, summarized in Fig. 3. MATa cells activate a donor on the left arm of chromosome III to be selected preferentially, although the other donor at HMR is available10,18,19. In contrast, MAT␣ cells inactivate donors on the left arm, making HMR essentially the only available donor. Thus a MATa cell deleted for HML can easily use HMR, but approximately a third of MAT␣ cells die when their only choice of a donor is HML (Refs 18, 19). The failure of many MAT␣ cells to use the ‘wrong’ donor occurs despite the fact that cells experiencing a DSB become arrested at a G2–M checkpoint. This should, theoretically, allow cells time to locate a donor and repair the DSB by gene conversion, and suggests that HML is indeed very inaccessible in MAT␣ cells.
A recombination enhancer
(b)
(c)
(d)
FIGURE 3. Donor preference is regulated by the recombination enhancer RE. (a) In MATa cells, a donor placed anywhere along the left arm of chromosome III is used preferentially in competition with HMR. This activation depends on the presence of the RE sequence. (b) In MAT␣ cells, the RE is inactivated and the use of HML (or other donors on the left arm) is strongly repressed, allowing HMR to be the preferred donor. Mutations within RE that prevent its inactivation result in increased use of HML. (c) Similarly, in MATa cells deleted for RE, donors on the left arm of chromosome III are inaccessible and HMR is again the preferred donor. (d) A similar mating-type-dependent difference is also seen for spontaneous recombination between a leu2-R allele that replaces HML and its adjacent silencer sequences and a leu2-K allele placed near MAT. TIG AUGUST 1998 VOL. 14 NO. 8
318
The key to understanding the mechanism of donor preference was provided by the yeast genome sequencing project, which chose chromosome III to be sequenced first20. The availability of sequence information made it possible to create terminal truncations (by including a telomere sequence) or internal deletions in order to pinpoint cis-acting elements responsible for the preference of MATa cells for HML (Ref. 10). A 2.5 kb deletion located 17 kb proximal to HML completely reverses donor preference, so that a MATa cell now uses HML only 10% of the time instead of 90%. By re-inserting subfragments of this deleted region, Wu and Haber10 defined a 700 bp ‘recombination enhancer’ (RE) sequence that controls MATa and
REVIEWS MAT␣ donor preference (Fig. 4b). (a) The 2.5 kb region has no open α2–Mcm1p reading frames, between the flankKAR4 RE YCL054w binding site ing KAR4 and YCL054w genes. In MATa cells, the RE acts not only to activate HML at its normal (b) location but also to activate either HML or HMR sequences placed at several different locations along the entire left arm of chromosome III (Refs 10, 19, 21) (Fig. 3). In each (c) A B C D case, if RE is deleted, HMR becomes the strongly preferred donor, just as α2–Mcm1p–α2 (TTT[A/G])10 in MAT␣ cells10,11,19. MAT␣ cells do binding site not activate HMR; rather they inactivate or sequester HML, leaving a-specific RNAs HMR as the only available donor. (d) The inaccessibility of HML in MAT␣ cells is also seen for donors inserted F1 HS F2 at other sites along the entire left arm of chromosome III and even on (e) the right arm, to a point about 25 kb proximal to MAT (Ref. 18). Thus the left arm is recombinationally inactive in MAT␣ and somehow α2–Mcm1p–α2 activated in MATa. The ‘cold’ state co-repressor in MAT␣, whether RE is present or not, is also the default state in a MATa cell deleted for RE. There- FIGURE 4. Analysis of RE. Deletion of a 2.5 kb region containing no open reading frames fore, RE acts to reverse the unusual results in loss of RE activity. (a) The region contains the RE as well as an Mat␣2p–Mcm1p inaccessibility of donors on the left binding site that has a minor role in regulating RE. (b) Insertion of subfragments derived from this region restores most RE activity. (c) A comparison of the same region in arm of chromosome III. S. cerevisiae and S. carlsbergensis yields a 244 bp region with four conserved subdomains. The regulation of recombination (d) The chromatin structure of RE in MATa cells shows several distinct protein footprints on the left arm of chromosome (F1 and F2) and an unusual DNase I hypersensitive (HS) region that overlaps a run of III is not specific to HO-mediated TTT[A/G] repeats. There are two weak a-specific transcripts that arise in this region. Their recombination nor to silenced donor exact end-points have not yet been mapped. (e) In MAT␣ cells, the entire RE region is sequences. This was shown by repressed by highly positioned nucleosomes that extend from a Mat␣2p–Mcm1p binding examining spontaneous recombina- site in RE across the entire interval between the two adjacent open reading frames, whose tion between a leu2-R allele inserted chromatin structure is not different in MATa or MAT␣ cells. in place of HML with a leu2-K allele placed near MAT (Fig. 3d). The rate of Leu⫹ spontaneous pers. commun.). Third, even when the HML␣ locus is recombinants arising by gene conversion between these unsilenced, by deleting the adjacent E and I silencers, two mutant alleles was 30 times higher in MATa cells and has a much more accessible chromatin structure, it than in MAT␣ (Refs 10, 21), even when the leu2-K allele fails to be used efficiently as a donor in MAT␣ cells18. is placed on another chromosome. This difference is lost If the RE does not exert a direct effect on the local when the RE is deleted, and the rate of Leu⫹ recombi- chromatin structure of the donor locus, what does it do? nation in MATa falls to the level seen in MAT␣ cells. This The idea we favor is that the RE changes the nuclearmating-type-dependent difference in Leu⫹ recombi- localization or the higher-order folding of the entire left nation depends on one of the two leu2 alleles being situ- arm of chromosome III to make it more ‘flexible’ in ated in the region normally occupied by HML; there is no locating and pairing with the recipient site in MATa such difference if leu2-R is inserted in place of HMR and cells. In this view, the chromosome arm would be leu2-K is near MAT. sequestered or immobilized (perhaps by being bound to the nuclear envelope) in such a way that HML is Mechanism of RE function unavailable in MAT␣ cells even though the chromatin The RE does not appear to act by regulating the structure at HML itself is unchanged. One observation chromatin structure of donor locus on the left arm of favoring this idea comes from measuring Leu⫹ recombichromosome III. First, there is no significant mating- nation between leu2-R located in place of HML but with type-dependent difference in mRNA levels for the LEU2 the leu2-K allele on a plasmid, so that the latter partner gene inserted in place of HML (which shows a 30-fold in the recombination reaction is presumably more recombination difference). Second, a direct exami- mobile and thus more able to ‘find’ leu2-K than would nation of the chromatin structure of the HML locus by be the same allele in the middle of an entire chromomicrococcal nuclease and DNase I digests does not some. As predicted the mating-type-dependent differreveal any significant mating-type-dependent difference ence is reduced from 30-fold to only 3-fold, supporting in chromosome structure (K. Weiss and R.T. Simpson, the idea that the constraint on recombination in MAT␣ TIG AUGUST 1998 VOL. 14 NO. 8
319
REVIEWS cells reflects a difficulty in bringing the two alleles together rather than the inherent inability of the leu2-R allele to recombine (M. Barlow, W-Y. Leung and J.E. Haber, unpublished). A complete understanding of the action of RE also needs to account for the discovery that the 700 bp RE is ‘portable,’ at least on chromosome III. RE still activates HML if the enhancer is deleted from its normal location 17 kb from the left end of the chromosome and a 700 bp RE fragment is inserted 70 kb from the end (G-F. Richard and J.E. Haber, unpublished). Moreover, in a strain carrying its normal RE, insertion of an additional RE adjacent to HMR increases its use in MATa cells to nearly 50%, from about 10% (Ref. 10). One explanation might be that all of chromosome III is unusual, and that even the right arm is somewhat ‘cold’ and thus responsive to the RE. Other chromosomes might not respond to RE.
How is RE regulated? By comparing the RE sequences of S. cerevisiae and S. carlsbergensis (which is functional in S. cerevisiae), it has been possible to narrow down the RE to 244 bp (Fig. 4c), within which are four well-conserved subdomains12. Deletion of regions A, C or D each abolishes RE activity. The most striking features of the conserved domains are ten repeats of TTT[A/G] in region D and a highly conserved binding site for the Mat␣2p–Mcm1p co-repressor in region C. The Mat␣2p–Mcm1p binding site in subdomain C appears to play a key role in repressing the activity of RE in MAT␣ cells. Mat␣2p and Mcm1p, in conjunction with Tup1p and to a lesser extent SSN6p, acts as a repressor of a-specific gene transcription22. Szeto et al.11 have recently reported that mutations of the two Mat␣2p-binding domains within this site are sufficient to alter donor preference in MAT␣, so that HML is used 50% of the time instead of 10%. This result also suggests that any a-specific gene products are unlikely to play a critical role in activating RE, because these genes should still be repressed by Mat␣2p–Mcm1p in a MAT␣ cell. However, the difference between HML use in MATa (80%) and MAT␣ (50%) in this mutant RE could still be attributed to a-specific genes. The genetic studies of how RE is regulated are strongly supported by analysis of the chromatin structure of the RE. In MAT␣ cells, the RE is covered by a very highly positioned set of nucleosomes on either side of the (occupied) Mat␣2p–Mcm1p binding site23. This ordered array extends across the entire 2.5 interval containing RE, and ends at the two flanking open reading frames. In MATa cells, the RE exhibits several distinctive footprints indicative of protein binding and a notable region of closely spaced DNase I hypersensitive (but not single stranded) sites that cover a region of TTT[A/G] repeats. In MAT␣ cells, when mutations of the Mat␣2p binding sites of the Mat␣2p–Mcm1p binding domain markedly increase the use of HML as a donor, the chromatin structure of the RE region becomes indistinguishable from that seen in MATa cells21. There is a second Mat␣2p–Mcm1p binding sequence located within the 2.5 interval between KAR4 and YCL054w (Fig. 4). Results of Wu and Haber10, and Wu et al.12 suggest that this second site is not important, as it is lacking in the 700 bp and other RE constructs inserted
into a deletion of the entire region, but Szeto et al.11 find that deleting this site increases the use of HML in MAT␣ strains. The second site might help stabilize the extended nucleosome array emanating from RE in MAT␣ cells.
Activation of RE in MATa cells With the exception of the RE, all well-characterized Mat␣2p–Mcm1p binding sites are located just upstream of a-specific genes. The RE does not contain an open reading frame, but there are two ‘sterile’ transcripts within or near the RE region that are transcribed in MATa, but not MAT␣ (Ref. 11). It is unlikely, however, that the sequence of the RNA transcript is important for RE activity, because truncations of RE that remove most of the normally transcribed sequence have full or substantial activity10,12. However, the act of transcription itself could still be the key regulatory feature. The existence of such transcripts reminds one of Xinactivation in mammals and their importance needs to be ascertained. The activation of a-specific genes depends on Mcm1p, which can bind to part of the Mat␣2p–Mcm1p binding site that also represses a-specific genes in MAT␣ cells22. Szeto et al.11 showed that the Mat␣2p–Mcm1p operator, fused to a reporter gene, will indeed activate transcription in MATa cells. Further evidence, establishing a definitive role for Mcm1p in activating RE has come from recent experiments by Wu et al.12. A 2 bp mutation that eliminates Mcm1p binding in the Mat␣2p–Mcm1p operator sequence also abolishes MATa donor preference so that HML is used only 10–20% of the time. Moreover, an Arg87Ala mutation in Mcm1p similarly reduces activation of HML in MATa cells and increases HML use in MAT␣ cells. This residue has recently been shown by X-ray crystallography to be an important contact between Mcm1p and Mat␣2p (Ref. 24). So it would appear that Mcm1p is a central player in activating RE. Whether Mcm1p has a co-activating protein is not yet known, nor is it clear that transcription per se is required. One very surprising result that has emerged from the analysis of the RE mutated in the Mcm1p binding site is that the chromatin structure of the mutant RE, even in MATa cells (where there is no Mat␣2p), has a highly organized nucleosome structure that is very similar to what is seen in normal MAT␣ cells. Apparently, other sequences within RE can organize a phased nucleosome structure in the absence of Mat␣2p–Mcm1p binding, although the repressor proteins seem to position more precisely and ‘lock in’ the repressing chromatin structure. These results might explain why a 2 bp mutation of the Mcm1p binding site has a more severe effect on HML usage than does a 25 bp deletion of most of the Mat␣2p–Mcm1p operator11,12.
Future perspectives The stage is set to delve more deeply into the remarkable properties of this system. There are really two different and fascinating phenomena here: (a) the unusual ‘coldness’ of the left arm in MAT␣ cells, or in MATa cells deleted for the RE; and (b) the reversal of this inhibition and the activation of the left arm for recombination when the RE is active. First, we need to find the cis-acting sites responsible for the unusual
TIG AUGUST 1998 VOL. 14 NO. 8
320
REVIEWS inactivity of the arm in the absence of RE. There might be several such sites along the left arm of chromosome III. Indeed a deletion of the first 40 kb of the left arm does not eliminate the ‘coldness’ of a donor located more proximally in MAT␣ cells. Second, we must identify additional trans-acting mutations that alter this process. While Mcm1p is clearly a very important part of the activation of RE, the entire 244 bp RE is more than 10 times larger than the size of an Mcm1p binding domain needed to activate transcription, and mutations in subdomains A and D are as profound as preventing Mcm1p binding. So far, only the CHL1 gene has been shown to affect donor preference25. A deletion of this gene, which also affects chromosome transmission of all yeast chromosomes, reduces donor preference in MATa switching but has no effect on MAT␣. What the connection is between chromosome stability and donor preference is not yet obvious. Now that the RE has been narrowed down to a small size it should be possible to identify DNA-binding proteins directly and to use this binding as an assay in characterizing other mutations affecting donor preference. Finally, we are left with the question of whether the elaborate mechanism(s) that Saccharomyces evolved to ensure donor preference have been adapted by other organisms to solve other developmental problems. One possible analogy is found in dosage compensation of metazoans. In Drosophila, the single X chromosome of a male is activated for transcription at apparently many sites26, which might act similarly to the RE. In Caenorhabditis, the problem is solved by a partial condensation of the two X chromosomes of the female to reduce transcription27, which might share some features of inactivation with the ‘coldness’ of yeast’s sex chromosome in MAT␣. (It is also useful to remember that an entire 100 kb chromosome arm in yeast is the length of a large gene in Drosophila!) Another way of thinking about RE is that it is a locus control region, which regulates recombination and not transcription. It will also be interesting to see what relationship RE has to locus control regions in mammals that activate gene expression and even DNA replication28,29.
18 Wu, X., Moore, J.K. and Haber, J.E. (1996) Mol. Cell. Biol. 16, 657–668 19 Wu, X., Wu, C. and Haber, J.E. (1997) Genetics 147, 399–407 20 Oliver, S.G. et al. (1992) Nature 357, 38–46 21 Wu, X. and Haber, J.E. (1995) Genes Dev. 9, 1922–1932 22 Szeto, L. and Broach, J.R. (1997) Mol. Cell. Biol. 17, 751–759 23 Weiss, K. and Simpson, R.T. (1997) EMBO J. 16, 4352–4360 24 Tan, S. and Richmond, T.S. (1998) Nature 391, 660–666 25 Weiler, K.S., Szeto, L. and Broach, J.R. (1995) Genetics 139, 1495–1510 26 Bashaw, G.J. and Baker, B.S. (1996) Curr. Opin. Genet. Dev. 6, 496–501 27 Chuang, P.T., Lieb, J.D. and Meyer, B.J. (1996) Science 274, 1736–1739 28 Capone, M. et al. (1993) EMBO J. 12, 4335–4346 29 Aladjem, M.I. et al. (1995) Science 270, 815–819
Reference added in proof 30 Weiss, K. and Simpson, R.T. (1998) Mol. Cell. Biol. (in press) J.E. Haber is in the Rosenstiel Center, the Keck Institute for Cellular Visualization and the Department of Biology, Brandeis University, Waltham, MA 02254, USA.
trends in
GENETICS Included in the September issue Chimeras and mosaics in mouse mutant analysis by J. Rossant and A. Spence Changing styles in C. elegans genetics by J. Hodgkin and R.K. Herman
References 1 Mackay, V. and Manney, T.R. (1974) Genetics 76, 255–271 2 Leberer, E., Thomas, D.Y. and Whiteway, M. (1997) Curr. Opin. Genet. Dev. 7, 59–66 3 Peterson, C.L. (1996) Curr. Opin. Genet. Dev. 6, 171–175 4 Long, R.M. et al. (1997) Science 277, 383–387 5 Takizawa, P.A. et al. (1997) Nature 389, 90–93 6 Wolberger, C. (1996) Curr. Opin. Struct. Biol. 6, 62–68 7 Loo, S. and Rine, J. (1994) Science 264, 1768–1671 8 Sherman, J.M. and Pillus, L. (1997) Trends Genet. 13, 308–813 9 Haber, J. (1995) BioEssays 17, 609–620 10 Wu, X. and Haber, J.E. (1996) Cell 87, 277–285 11 Szeto, L. et al. (1997) Genes Dev. 11, 1899–1911 12 Wu, C. et. al. (1998) Genes Dev. 12, 1726–1737 13 Bi, X. and Broach, J.R. (1997) Mol. Cell. Biol. 17, 7077–7787 14 Strathern, J.N. (1989) Genetic Recombination (Kucherlapati, R. and Smith, G.R., eds), pp. 445–464, American Society for Microbiology 15 Haber, J.E. (1992) Trends Genet. 8, 446–452 16 Klar, A.J., Hicks, J.B. and Strathern, J.N. (1982) Cell 28, 551–561 17 Weiler, K.S. and Broach, J.R. (1992) Genetics 132, 929–942
Survival with HIV infection: good luck or good breeding? by S.L. Rowland-Jones The Drosophila genome project: a progress report by G.M. Rubin Traps to catch unwary oncogenes by A-O. Hueber and G.I. Evan Selfishness and death... raison d’être of restriction, recombination and mitochondria by I. Kobayashi Who pulls the string to pattern cell division in Drosophila? by H. Skaer
TIG AUGUST 1998 VOL. 14 NO. 8
321