Epigenetics in Saccharomyces cerevisiae

OVERVIEW

The fraction of chromatin in a eukaryotic nucleus that bears active genes is termed euchromatin. This chromatin condenses in mitosis to allow chromosomal segregation and decondenses in interphase of the cell cycle to allow transcription to occur. However, some chromosomal domains were observed by cytological criteria to remain condensed in interphase, and this constitutively compacted chromatin was called heterochromatin. With the development of new techniques, molecular rather than cytological features have been used to define this portion of the genome, and heterochromatin, which is often found at centromeres and telomeres, was shown to contain many thousands of simple repeat sequences, particularly in higher eukaryotic organisms. The repeat-rich genomic DNA tends to replicate late in S phase of the cell cycle, is found clustered at the nuclear periphery or near the nucleolus, and is resistant to nuclease attack. Importantly, the characteristic chromatin structure that is formed on repeat DNA tends to spread and repress nearby genes. In the case of the fruit fly locus white, a gene that determines red eye color, epigenetic repression yields a red and white sectored eye through a phenomenon called position effect variegation (PEV). Mechanistically, PEV in flies reflects the recognition of methylated histone H3K9 by heterochromatin protein 1 (HP1), which can spread along the chromosomal arm. In Saccharomyces cerevisiae, also known as budding yeast, a distinct mechanism of heterochromatin formation has evolved, yet it achieves a very similar result.

S. cerevisiae is a microorganism commonly used for making beer and baking bread. However, unlike bacteria, it is a eukaryote. The chromosomes of budding yeast, like those of more complex eukaryotes, are bound by histones, enclosed in a nucleus and replicated from multiple origins during S phase. Still, the yeast genome is tiny with only 14 megabase pairs of genomic DNA divided among its 16 chromosomes, some not much larger than a bacteriophage genome. There are approximately 6000 genes in the yeast genome, closely packed along chromosomal arms, generally with less than 2 kb spacing between them. The vast majority of yeast genes are in an open chromatin state, meaning that they are either actively transcribed or can be rapidly induced. This, coupled with a very limited amount of simple repeat DNA, makes the detection of heterochromatin by cytological techniques very difficult in yeast.

Nonetheless, budding yeast has distinct heterochromatin-like regions adjacent to all 32 telomeres and at two silent mating loci on chromosome III (Chr III), shown using molecular tools. Transcriptional repression at telomeres and the silent mating loci can spread into adjacent DNA and repression of the silent mating loci is essential for maintaining a mating-competent haploid state. Both the subtelomeric regions and the silent mating type loci repress integrated reporter genes in a position-dependent, epigenetic manner; they replicate late in S phase and are present at the nuclear periphery. Thus, these loci bear most of the functional characteristics of heterochromatin, without having cytologically visible condensation in interphase. By exploiting the advantages afforded by the small genome of yeast and its powerful genetic and biochemical tools, many basic principles of chromatin-mediated repression that are relevant to heterochromatin in more complex organisms have been discovered. Nonetheless, silent chromatin in budding yeast is dependent on a unique set of nonhistone proteins that do not deposit nor recognize histone H3 lysine 9 methylation.

1. THE GENETIC AND MOLECULAR TOOLS OF YEAST

Yeast provides a flexible and rapid genetic system for studying cellular events. With an approximate generation time of 90 min, colonies containing millions of cells are produced after just 2 d of growth. In addition, yeast can propagate in both haploid and diploid forms, greatly facilitating genetic analysis. Like bacteria, haploid yeast cells can be mutated to produce specific nutritional requirements or auxotrophic genetic phenotypes, and recessive lethal mutations can either be maintained in haploids as conditional lethal alleles (e.g., temperature-sensitive mutants), or in heterozygotic diploids, which carry both wild-type and mutant alleles.

Extremely useful is the efficient homologous recombination system of budding yeast, which allows the alteration of any chosen chromosomal sequence at will. In addition, portions of chromosomes can be manipulated and reintroduced on plasmids that are stably maintained through cell division, thanks to short sequences that provide centromere and replication origin function. Large linear plasmids, or minichromosomes, which carry telomeric repeats to cap their ends, also propagate stably in yeast.

Yeast also has a unique advantage in the genetic analysis of histones and their roles in gene regulation: Unlike mammalian cells, which have as many as 60–70 copies of the core histone coding genes (H2A, H2B, H3, and H4), yeast contains only two copies of each of these genes. Because these two copies are functionally redundant, this has enabled the production of cells containing single histone gene copies. Deletion analysis or mutation of defined histone amino acids in these genes has uncovered specific roles for histone residues in heterochromatin and other cellular functions. By searching for suppressors of these mutant phenotypes it has also been possible to identify heterochromatin proteins that interact with the histone sites in question. The ease of generating histone mutants in yeast has also led to the systematic mutagenesis of most amino acid residues in each of the core histones (http://www.ncbi.nlm.nih.gov/pmc/articles/PMC2666297/) and analysis of their effects on genome function (Huang et al. 2009).

Budding yeast also provides powerful cellular readouts for epigenetic gene regulation, conceptually similar to PEV in flies, in which the white gene provides a visible screen for variegated gene expression (see Elgin and Reuter 2013 for more detail). A parallel phenomenon called telomere position effect (TPE) occurs near yeast telomeres. The study of TPE has been aided analogously by the use of the URA3 and ADE2 reporter genes (Fig. 1). The URA3 gene product is necessary for pyrimidine biosynthesis, and cells that do not express the Ura3 protein cannot grow on synthetic media lacking uracil. In addition, URA3 allows for a counter-selection against expression in the presence of 5-fluoroorotic acid (5-FOA), because the Ura3 gene product converts 5-FOA to 5-fluorouracil, an inhibitor of DNA synthesis that causes cell death. Thus, when URA3 is integrated near heterochromatin and the gene is repressed in some but not all cells, only the cells that silence URA3 are able to grow in the presence of 5-FOA. Conversely, if the strain lacks the strong activator of URA3, Ppr1, positive selection is possible; only URA3-expressing cells can grow on uracil-deficient plates. Thus the efficiency of URA3 repression/expression can be scored accurately in serial dilution assays (over a 1–10⁶-fold range) on plates that either contain 5-FOA, or lack uracil (Fig. 1A). Because 5-FOA is mutagenic and puts strong pressure on cells to repress the reporter gene, some conditions or yeast backgrounds favor the use of uracil-drop out plates over counter-selection on 5-FOA.

Another useful assay for epigenetic repression is based on the insertion of the reporter gene ADE2 near heterochromatin. When ADE2 is repressed, a precursor in adenine biosynthesis accumulates, turning the cell red. When ADE2 is expressed, cells are white (Fig. 1B). The epigenetic nature of repression of a subtelomeric ADE2 gene is visible within a single colony of genetically identical cells because the variegated expression of ADE2 generates white sectors in a red colony background, attributable to ADE2 repression (Fig. 1B). This reporter avoids the stress of selective pressure against cells that fail to repress ADE2, and thus the sectored phenotype illustrates both the switching rate and inheritance of the epigenetic state through mitotic division, much like the sectored white phenotype in the Drosophila melanogaster eye.

Combined with these genetic approaches, biochemical techniques for mapping epigenetic modifiers are readily applied to yeast. Large yeast cultures can be grown either synchronously or asynchronously. A battery of molecular tools, which includes transcriptome analysis and chromatin immunoprecipitation (ChIP), can be combined with multiplexed next generation sequencing for efficient whole-genome coverage. Combining this with proteomic approaches that map protein networks enables a comprehensive and quantitative comparison of gene expression, transcription factor binding, histone modification, and protein–protein interactions. Finally, a technology developed in yeast, called chromosome conformation capture, scores protein-mediated contacts between unlinked chromosomal domains to monitor long-range chromatin interactions (for more detail, see Dekker and Misteli 2014). With this broad range of tools, scientists have explored the mechanisms that regulate both the establishment of heterochromatin and its physiological roles in budding yeast for more than 30 years. Before describing these discoveries, however, we will first review the life cycle of yeast in more detail.

2. THE LIFE CYCLE OF YEAST

S. cerevisiae multiplies through mitotic division in either a haploid or a diploid state by producing a bud that enlarges and eventually separates from the mother cell (Fig. 2A). This is why it is called budding yeast. Haploid yeast cells can mate with each other (i.e., conjugate) because they exist in one of two mating types, termed a or α, reminiscent of the two sexes in mammals. Yeast cells of each mating type produce a distinct pheromone that attracts the cells of the opposite mating type: a cells produce a peptide of 12 amino acids (aa) called a factor, which binds to a membrane spanning a-factor receptor on the surface of an α cell. Conversely, α cells produce a 13 aa peptide (α-factor) that binds to the α-factor-receptor on the surface of a cells. These interactions result in the arrest of the two cell types in mid-to-late G₁ phase of the cell cycle. The arrested cells assume “shmoo”-like shapes (named after the pear-shaped Al Capp cartoon character; Fig. 2B). Shmoos of opposite mating type fuse at their tips to produce an a/α diploid cell.

In diploid cells the mating response is repressed, and cells propagate vegetatively (i.e., by mitotic division) unless they are exposed to starvation conditions. Nitrogen starvation induces a meiotic program in diploids that provokes the formation of an ascus containing four spores, two of each mating type. When nutrient levels are restored, these haploid spores grow into a or α cells that are again capable of mating to form a diploid, starting the life cycle over again.

Although haploid yeast cells in the laboratory are usually genetically constructed to be stable α or a cells, yeast in the wild switch their mating type nearly every cell cycle (Fig. 3A). Mating type switching is provoked by an endogenous cell-cycle-regulated endonuclease activity (HO) that induces a site-specific double-strand break at the MAT locus. A gene conversion event (in which the donor DNA sequence remains unchanged, but the recipient DNA is altered) transposes the opposite mating type information from one of two constitutively silent donor loci, HMLα or HMRa, to the MAT locus, in which the mating type–determining genes, a1 and a2, or α1 and α2, are expressed. Strains capable of mating type switching are called homothallic. This name reflects the fact that a single vegetative MATa cell can produce MATα progeny, and vice versa, allowing offspring to mate with each other.

In the laboratory it is useful to have strains with stable mating types, and thus laboratory yeast generally contain a mutant HO endonuclease gene (ho^–). These cells fail to induce a double-strand cleavage at the MAT locus, and therefore cannot switch mating type. These heterothallic cells are stable haploid strains of either a or α mating type, unless placed in proximity of cells of the opposite mating type, in which case the two haploid cell types will mate to form a diploid.

Importantly, the two stable haploid cell types do not require silencing for viability, yet they must repress the genes at the homothallic mating type loci, HML and HMR, to retain their ability to mate. If repression fails and cells express both sets of mating type genes at once, then a haploid will behave as if it were diploid, suppressing mating competence and generating a sterile haploid strain (Fig. 3B). The mechanisms that repress the two homothallic mating type loci, HML and HMR, have become a classic system for the study of heterochromatin-mediated repression.

3. YEAST HETEROCHROMATIN IS PRESENT AT THE SILENT HM MATING LOCI AND AT TELOMERES

The three mating type loci, HMLα, MAT, and HMRa, are located on one small chromosome, Chr III, and contain the information that determines α or a mating type in yeast. The silent loci, HMLα (∼12 kb from the left telomere) and HMRa (∼23 kb from the right telomere), are situated between short DNA elements called E and I silencers (Fig. 3B,C). In a wild-type cell, the silent cassettes are active only once copied and integrated into the MAT locus, which lacks silencer elements. The transfer of HMLα information into MATa results in an α mating type (MATα) cell, whereas the transfer of HMRa information into MAT results in the a mating type (MATa) (Fig. 3B). This shows that the promoters and genes at the HM loci are completely intact, and remain repressed because of their position between the E and I silencers. Deletion of the flanking silencer sequences indeed allows expression of the silent information, generating a nonmating state.

By scoring for haploid sterility, mutations that impair silencing at the HM loci were isolated (Rine et al. 1979). This allowed identification of the silent information regulatory proteins, Sir1, Sir2, Sir3, and Sir4, as being essential for the full repression of HM loci (Rine and Herskowitz 1987). Mutations in sir2, sir3, or sir4 caused a complete loss of mating, which is attributable to a loss of HM repression. In sir1 mutants, only a fraction of MATa cells were unable to mate. Taking advantage of the partial phenotype of sir1 deficient cells, it could be shown that the two alternative states (mating and nonmating) are heritable through successive cell divisions in genetically identical cells (Pillus and Rine 1989). This provided a clear demonstration that mating type repression displays the hallmark characteristic of epigenetically controlled repression. Later genetic studies showed that the amino termini of histones H3 and H4, repressor activator protein 1 (Rap1), and the origin recognition complex (ORC) are also components of silent mating locus heterochromatin (reviewed by Rusche et al. 2003). These latter two DNA binding factors have other essential functions in the nucleus, namely, the regulation of ribosomal protein gene expression or the initiation of DNA replication, and, thus, only “moonlight” as corepressors. Although less well studied, the same appears true for Abf1, a third silencer binding factor.

A similar position-dependent repression occurs immediately adjacent to the yeast telomeric repeat DNA (C_1–3A/TG_1-3) found at the ends of all yeast chromosomes. As mentioned above, the variegated but heritable repression of subtelomeric reporter genes such as URA3 and ADE2 is called TPE (Gottschling et al. 1990). TPE shares the HM requirement for Rap1, Sir2, Sir3, Sir4, and the histone amino termini (Aparicio et al. 1991; Thompson et al. 1994a), and the repression mechanisms have proven to be closely related. However, given that subtelomeric reporters can switch at detectable rates between silent and expressed states, unlike HM loci, telomere-proximal gene repression appears to be more similar to fly PEV (see Elgin and Reuter 2013).

4. Sir PROTEIN STRUCTURE AND EVOLUTIONARY CONSERVATION

The known chromatin binding factors that are essential for Sir mediated silencing are Sir2, Sir3, and Sir4, whereas Sir1 enhances the efficiency of repression at HM loci, but is not found at telomeres. The Sir2-3-4 proteins work as a trimeric complex with 1:1:1 stoichiometry (Cubizolles et al. 2006). Both Sir3 and Sir4 are able to bind nucleosomes and DNA independently, yet the Sir holo-complex remains a trimer when bound to nucleosomes. Moreover, Sir3 and Sir4 each have homo- and heterodimerization motifs in their carboxyl termini. Mutation or deletion of their interaction domains disrupts silencing in vivo (Murphy et al. 2003; Rudner et al. 2005; Ehrentraut et al. 2011; Oppikofer et al. 2013).

Sir protein expression levels are tightly regulated and a single extra copy of the SIR4 gene impairs repression, as does strong induction of SIR2 (Cockell et al. 1998). On the other hand, increasing levels of Sir3 protein alone extends the spreading of Sir3 along nucleosomes, and with it, transcriptional repression (Renauld et al. 1993; Hecht et al. 1996). Balanced overexpression of all three proteins greatly improves silencing efficiency at telomeres, and allows repression of reporters located in euchromatic regions that are flanked by silencer elements, which at normal Sir protein levels are not repressed (Maillet et al. 1996).

Although Sir2, Sir3, and Sir4 are equally essential for the structural integrity of the Sir complex and therefore, for both the establishment and maintenance of silent chromatin, each has a different function. Sir2 provides a nicotinamide dinucleotide (NAD)-dependent histone deacetylase activity that is essential for repression in a wild-type background (Imai et al. 2000), whereas Sir3 and Sir4 fulfill structural roles without obvious enzymatic activities. Sir3 is a member of the AAA⁺ ATPase family, which lacks ATPase activity. It is largely responsible for the specificity of Sir complex binding to nucleosomes because of its selective affinity for nucleosomes with unacetylated histone H4 lysine 16 (H4K16) and unmethylated histone H3 lysine 79 (H3K79) (Johnson et al. 1990; Altaf et al. 2007; Oppikofer et al. 2011). The sensitivity of Sir3 to these histone modifications helps restrict the binding of the Sir complex to appropriate sites.

Sir4 is the largest (152 kDa) and the least conserved of the Sir proteins, yet it forms a stable heterodimer with Sir2 (Moazed et al. 1997; Strahl-Bolsinger et al. 1997) and enhances Sir2 deacetylase activity (Tanny et al. 1999; Cubizolles et al. 2006). Structural information on the Sir4 interface indicates that most of the Sir2-interaction domain of Sir4 (aa 737–839) is buried in a pocket formed by the poorly conserved amino terminus and the carboxy-terminal catalytic domain of Sir2 (Hsu et al. 2013). The Sir3-binding domain within Sir4 is contained in the parallel coiled-coil structure at its extreme carboxyl terminus, which also serves as a binding site for other proteins (Chang et al. 2003; Rudner et al. 2005).

5. SILENT CHROMATIN IS DISTINGUISHED BY A REPRESSIVE STRUCTURE THAT SPREADS THROUGH THE ENTIRE DOMAIN

Repression of gene activity in euchromatin often requires the presence of a repressive protein or complex that recognizes a specific sequence in the promoter of a gene, thus preventing the active engagement of the transcription machinery. In contrast, heterochromatic repression occurs through a different mechanism that is not promoter specific: Repression initiates at specific nucleation sites, yet spreads continuously throughout a domain, silencing any and all promoters in the region (Fig. 5) (Brand et al. 1985; Renauld et al. 1993). The correlation of transcriptional repression with Sir binding was confirmed by the use of ChIP, which showed that Sir2, Sir3, and Sir4 proteins interact physically with chromatin throughout the subtelomeric domain of silent chromatin and spread continuously inward from the chromosomal end (Strahl-Bolsinger et al. 1997). Evidence that this induces a repressive, less accessible chromatin structure in vivo comes from other approaches. For instance, it was shown that the DNA of silenced chromatin was not methylated efficiently in yeast cells that express a bacterial dam methylase, although the enzyme readily methylated sequences outside the silent region. This suggests that heterochromatin restricts access to macromolecules like dam methyltransferase (Gottschling, 1992). Similarly, the ∼3 kb HMR locus in isolated nuclei is preferentially resistant to certain restriction endonucleases (Loo and Rine 1994), and nucleosomes were shown to be tightly positioned between two silencer elements creating nuclease resistant domains at silent, but not active, HM loci (Weiss and Simpson, 1998). The reduced accessibility of yeast silent chromatin to nucleolytic attack is also observed in vitro when Sir-nucleosome complexes are reconstituted from recombinant proteins (Martino et al. 2009).

The extent to which either yeast or metazoan heterochromatin is hypercondensed to hinder access to transcription factors sterically is less certain. Surprisingly, the repressive complex formed by the interaction of Sir proteins and histones appears to be dynamic because Sir proteins can be incorporated into HM silent chromatin even when cells are arrested at a stage in the cell cycle when heterochromatin assembly generally does not occur (Cheng and Gartenberg 2000). This may explain why Sir-bound heterochromatin can serve as a binding site for certain transcription factors (e.g., the heat shock transcriptional activator, HSF1) even in its repressed state (Sekinger and Gross 1999). Although such studies argue that heterochromatin does not simply hinder access for all nonhistone proteins, no obvious transcription occurs, and engaged RNA polymerases cannot be detected experimentally. Experiments by Chen and Widom (2005) argue that the step that is specifically prevented by yeast heterochromatin is formation of a complex of RNA polymerase II (RNA Pol II) with the promoter-binding transcription factors TFIIB and TFIIE. Consistently, a drop in RNA Pol II binding was seen in a system in which silencing was induced by the controlled expression of Sir3. Thus, silent yeast chromatin may allow turnover of Sir factors and some transcription factors, yet it selectively impedes the binding of specific elements of the basal transcription machinery, thereby blocking mRNA production.

6. DISTINCT STEPS IN HETEROCHROMATIN ASSEMBLY

The assembly of heterochromatin in budding yeast involves a series of molecular steps, starting with a site-specific nucleation step. This requires DNA recognition by a sequence-specific DNA binding factor. Next, heterochromatin spreads from the initiation site, limited by specific boundary mechanisms. A change in higher organization of the repressed chromatin then occurs, which is distinct from the simple binding of Sir factors. Finally, yeast silent chromatin is sequestered near the nuclear envelope, generating a subnuclear compartment that favors heterochromatin-mediated repression by promoting its duplication. Although the assembly of heterochromatin at telomeres varies in some aspects from its assembly at HM loci, both embody a very similar principle: the presence of specific DNA binding factors that nucleate the spread of general repressors. In both cases, the spreading requires active deacetylation by Sir2. These mechanisms are described in the following paragraphs.

7. THE CRUCIAL ROLE OF HISTONE H4K16 ACETYLATION AND ITS DEACETYLATION BY Sir2

The molecular interactions between the Sir proteins have been well characterized. Sir4 interacts strongly with Sir2 and separately with Sir3 in vivo and in vitro (Moazed et al. 1997; Strahl-Bolsinger et al. 1997; Hoppe et al. 2002). When coordinately expressed in insect cells, Sir2, Sir3, and Sir4 proteins can be isolated as a stable complex with a 1:1:1 stoichiometry (Cubizolles et al. 2006). Nonetheless, Sir3 has a special role in this process because Sir3 can form a stable extended multimer in vitro (Liou et al. 2005) and its overexpression extends the subtelomeric silent domain from its normal ∼3 kb to ∼15 kb from the telomeric end, coincident with the binding of Sir3 (Renauld et al. 1993; Hecht et al. 1996).

The platform on which the Sir complex spreads consists of nucleosomes with deacetylated histone H3 and H4 amino termini (Braunstein et al. 1996; Suka et al. 2001), and the manner in which Sir3 interacts with histones helps explain how spreading occurs (Fig. 7). Sir3 binds the deacetylated histone H4 amino terminus in a highly selective manner both in vitro and in vivo (Johnson et al. 1990; Johnson et al. 1992; Carmen et al. 2002; Yang et al. 2008a). In this regard, the most important histone region is contained in residues 16–29 of histone H4, and lysine 16 must be positively charged (unmodified or substituted by arginine) for Sir3 to bind (Johnson et al. 1990, 1992).

A cocrystal structure of the Sir3 amino-terminal BAH domain explains this specificity quite well (Armache et al. 2011). Sixteen residues in the Sir3 BAH domain interact with H4 tail residues 13 to 23 which are held in a rigid conformation, primarily through electrostatic interactions with amino acid side chains. A negatively charged binding pocket of BAH Sir3 accommodates the side chains of unmodified H4K16 and H4H18. Indeed, acetylation of H4K16 could potentially disrupt most of the electrostatic contacts in this pocket. In contrast, the Sir2-Sir4 subcomplex binds with slight preference to nucleosomes bearing an acetylated H4K16 residue, at least in the absence of NAD⁺ (Oppikofer et al. 2011). This is consistent with acetylated histone H4K16 being a preferred and crucial target for the Sir2 enzyme (Imai et al. 2000; Suka et al. 2002; Cubizolles et al. 2006). The AAA⁺ domain of Sir3 also binds unmodified nucleosomes in vitro (Ehrentraut et al. 2011) and requires that all four H4 acetylation sites (K5, K8, K12, and K16) are deacetylated for optimal binding (Carmen et al. 2002) and for telomeric silencing (Thompson et al. 1994a). Because histones are naturally deacetylated throughout telomeric heterochromatin, the Sir3 carboxyl terminus may contribute to the stability of the Sir3 complex through its interaction with the fully deacetylated histone tail.

On addition of NAD⁺, the Sir2-catalyzed deacetylation of H4K16ac generates a by-product called O-acetyl-ADP-ribose, as well as a deacetylated histone H4 tail (illustrated in Fig. 5 of Seto and Yoshida 2014). Intriguingly, not only the generation of the high-affinity binding site for Sir3, but apparently also production of the intermediate metabolite O-acetyl-ADP-ribose, enhances the affinity of Sir2-3-4 complexes for chromatin (Johnson et al. 2009; Martino et al. 2009). The generation of O-acetyl-ADP-ribose enhances the interaction of Sir3 with Sir4-Sir2 in vitro (Liou et al. 2005), favors the oligomerization of Sir proteins on nucleosomal arrays (Onishi et al. 2007), and enhances protection of the linker DNA from micrococcal nuclease digestion (Oppikofer et al. 2011). This is consistent with genetic studies showing that any mutation of histone H4K16 disrupts telomere repression, even substitution of lysine by the similarly charged amino acid arginine, or glutamine, which mimics the uncharged nature of acetylated lysine.

Interestingly, the basic region of aa 16–24 within the H4 amino terminus also promotes nucleosomal array compaction in vitro, suggesting that the acetylation state of H4K16 may regulate the higher-order folding of the nucleosomal fiber (Shogren-Knaak et al. 2006). Thus, histone H4K16 deacetylation by Sir2 may actively promote silent chromatin formation in several ways: First, a conformational change may be triggered in the Sir complex by the by-product O-acetyl-ADP-ribose; second, the affinity of the Sir complex for chromatin increases thanks to generation of a high-affinity Sir3 site; and third, even when Sir complexes are not bound, nucleosomal arrays may compact because of contact between the H4 tail and the adjacent nucleosomal face. From this it was clear that control over Sir-mediated silencing would lie in the acetylation/deacetylation cycle of histone H4.

We summarize here the different steps for the initiation and spreading of heterochromatin in an environment enriched for acetylated histone H4, which is likely to be deposited immediately after replication (Fig. 6). At telomeres, Rap1 and yKu recruit Sir4, and Sir4 forms a dimer with Sir2 to deacetylate histone H4 and H3 amino-terminal tails of nearby nucleosomes. Sir3 is recruited by its affinity for Sir4, but also for Abf1 and Rap1. The deacetylation of the histone H4 tail produces a high affinity Sir3 binding site on the nucleosome, which favors assembly of a Sir2-3-4 complex. The interactions of Sir3 with Sir4, Sir3 with the H4 tail and the nucleosomal core, and Sir4 nonspecifically with linker DNA, all seem to contribute to the stable binding of the Sir complex to the nucleosomal fiber. The action of Sir2 on adjacent acetylated nucleosomes appears to trigger the spread of the complex along adjacent histone tails. Finally, the long-range folding of the chromatin fiber may stabilize the repressed state. Most of these events are likely to be very similar at HM loci, although there the initial recruitment of Sir4 needs Rap1, Abf1, or ORC and Sir1. The question then arises, what causes Sir complex spreading to stop?

8. BARRIER FUNCTIONS: HISTONE MODIFICATIONS RESTRICT Sir COMPLEX SPREADING

Because the acetylated histone H4K16 binds Sir2-4 tightly, whereas its deacetylation by Sir2 is crucial for the spreading of heterochromatin, it is not surprising that interfering with the cycle of H4K16 acetylation/deacetylation impedes heterochromatin propagation (Kimura et al. 2002; Suka et al. 2002). The yeast histone acetyltransferase (HAT) Sas2, a member of the highly conserved MYST class of HATs, modifies K16 in bulk euchromatin. Therefore at the boundaries of silent chromatin, one expects to find acetylated histone H4K16. Accordingly, if the SAS2 gene is deleted, or if H4K16 is changed to arginine to simulate the deacetylated state, Sir3, Sir4, and Sir2 spread at low levels inward from the telomeric repeats, approximately fivefold further than in a wild-type cell. This suggests that the spreading of subtelomeric heterochromatin is controlled, at least in part, by the opposing activities of Sir2 and Sas2 on lysine 16 of H4 (Fig. 7). Limiting the global amount of Sir2 (or of the Sir2-Sir4 complex) naturally limits spread by limiting deacetylation.

At the HM loci, restricting the spread of silent chromatin is perhaps even more critical than at telomeres because genes important for growth are found on Chr III, and it is known that silencing can spread bidirectionally from silencers into flanking DNA sequence. One boundary that prevents further spreading of silencing from HMR, is a tRNA gene (Donze and Kamakaka 2001). This boundary function is likely to require the HAT activity that is associated either with transcription or the transcriptional potential of this locus. It is significant that one of these HATs is Sas2, although the H3 HAT, Gcn5, can also affect boundary function of the tRNA gene. This suggests that transcriptional activators, in general, can restrict Sir complex propagation by recruiting HATs. Consistently, in subtelomeric euchromatin regions, boundary activity has been attributed to the general transcription factors, Reb1, Tbf1, and to the acidic trans-activating domain of VP16 (Fourel et al. 1999, 2001). These factors are likely to promote the hyperacetylation of histones (Fig. 7).

Another modification that antagonizes the spread of telomeric heterochromatin is histone H3K79 methylation, which is deposited by the lysine methyltransferase Dot1 (Van Leeuwen et al. 2002; Ng et al. 2002). This histone lysine methyltransferase (KMT) was discovered in a screen for factors whose overexpression caused loss of telomeric silencing (Singer et al. 1998). However, Dot1 does not methylate H3K79 in heterochromatin itself, but instead in adjacent euchromatin and at active genes. Indeed, the artificial targeting of Dot1 to telomeric heterochromatin derepresses silencing by Sir proteins (Stulemeijer et al. 2011), most likely by reducing the affinity of Sir3 for nucleosomes.

Surprisingly, the H4 tail is required for the bulk methylation of H3K79 by Dot1. An elegant set of experiments (Altaf et al. 2007; Fingerman et al. 2007) has generated a model that helps explain the interplay between the H4 tail and the demethylated state of K79 in heterochromatin. Namely, when H4K16 is deacetylated by Sir2, a high affinity binding site for Sir3 is generated by a charged patch in the H4 tail (K16, R17, H18, R19, K20). Interestingly, Dot1 and Sir3 compete for this charged patch, although Sir3 is sensitive to H4K16 acetylation, and Dot1 is not. Thus, in deacetylated heterochromatin, Sir3 is a potent inhibitor of Dot1 binding. However, in adjacent euchromatin, in which K16 is acetylated, Dot1 preferentially binds the charge patch and methylates histone H3K79. The Sir3N-nucleosome crystal structure confirmed genetic evidence, suggesting that Sir3 interacts with H3K79 and that it is in close proximity to H4K16 (Armache et al. 2011). Both the binding of Sir3 and that of the holo-SIR complex are, in turn, weakened by histone H3K79 methylation. Again, this could be reconstituted in vitro, in binding assays between Sir3 and reconstituted nucleosomes that bear either H4K16ac or H3K79me (Oppikofer et al. 2011). Thus, the weak binding of Sir3 to H4K16ac in euchromatin favors Dot1 binding to the H4 tail and subsequent H3K79 methylation, which, in turn, weakens interaction with Sir3. This provides a good example of interhistone interactions as discussed and illustrated in Figure 12 of Allis et al. 2014.

In addition to the mechanism regulating H3K79 methylation and H4K16 acetylation near boundaries, it was reported that in euchromatin the presence of the variant histone H2A.Z and the RNA polymerase associated factor Bdf1 (Meneghini et al. 2003), as well as the tethering of DNA to nuclear pores (Ishii et al. 2002), generate boundaries that limit the spread of silent chromatin. Although the mechanisms by which these factors affect heterochromatin spreading are unknown, it is interesting to note that some inducible genes with H2A.Z-containing promoters, associate with nuclear pores on activation, and remain associated there in a manner that facilitates their re-induction (Ishii et al. 2002; Brickner and Walter 2004). Thus, boundary function may reflect a chromatin state that allows rapid recruitment of transcriptional activators, including HATs, KMTs like Dot1, or nucleosome remodelers that directly or indirectly disrupt histone interactions with heterochromatin proteins (Fig. 7).

9. A ROLE FOR THE H3 AMINO-TERMINAL TAIL IN HIGHER-ORDER CHROMATIN STRUCTURES

There is increasing evidence that the formation of heterochromatin involves a series of steps that include, but go beyond, the binding of Sir proteins. When Sir protein expression is induced artificially in G₁, Sir proteins spread from their initiation site by interacting with deacetylated H4 amino termini, but silencing is still defective (Kirchmaier and Rine 2006). Also, when the histone H3K56 acetylation site is mutated, Sir protein spreading occurs, but silencing is disrupted. Similarly, the presence of H3K56 acetylation leads to increased accessibility of Sir-bound nucleosomes to micrococcal nuclease in vitro without impairing Sir protein association (Oppikofer et al. 2011). Thus, the establishment of silencing requires not only Sir protein spreading, but also the deacetylation of H3K56, an event that enables nucleosomes to bind DNA more tightly to form a structure resistant to ectopically expressed bacterial dam methylase (Maas et al. 2006; Xu et al. 2007; Celic et al. 2008; Yang et al. 2008b).

Repression also involves the replacement of histone H3 methylated at K4 and K79 with unmethylated histone H3 (Katan-Khayakovich and Struhl 2005; Osborne et al. 2009). Whereas H3K56 deacetylation by Hst3 and Hst4, two homologs of Sir2, takes place every S phase, demethylation of H3K79 requires dilution by replication, and can take up to four cell divisions. These changes may parallel the formation of a compact, higher-order chromatin structure.

At first, the role of the H3 amino terminus in silencing appeared to be similar to that of the H4 amino terminus because the domain in each tail involved in silencing in vivo was shown to bind Sir3 and Sir4 in vitro (Hecht et al. 1995). However, although the H4 amino-terminal residues are required for the recruitment and spreading of Sir3 and the remaining Sir complex, the H3 tail was required neither for recruitment nor for spreading of the Sir proteins. Nonetheless, deletion of the H3 amino terminus or mutation of critical residues 11–15 (T–G–G–K–A) led to altered topology, increased accessibility to bacterial dam methylase expressed in yeast, and less tightly folded chromatin (Sperling and Grunstein 2009; Yu et al. 2011). Thus, although Sir proteins are clearly recruited by the H4 tail, they may subsequently interact with the H3 amino terminus to form compacted chromatin.

10. TRANS-INTERACTION OF TELOMERES, AND PERINUCLEAR ATTACHMENT OF HETEROCHROMATIN

In budding yeast, as in many lower eukaryotes, telomeres cluster together during interphase, in close association with the nuclear envelope. This clustering was initially observed as prominent foci of Rap1 and Sir proteins, which were detected above a diffuse nuclear background by immunostaining (Fig. 8). Disruption of silencing by histone H4K16 mutation, or interference in Rap1 or yKu function, led to the dispersion of the Sir proteins from these clusters (Hecht et al. 1995; Laroche et al. 1998). Later it was shown that not only telomeres, but also the silent HML and HMR loci are closely associated with telomeres at the nuclear envelope. Binding to the nuclear envelope is mediated by redundant pathways that depend either on the telomere-bound yKu factor, or on components of silent chromatin itself. Interestingly, the interactions that lead to telomere clustering can be genetically separated from telomere anchoring to the nuclear envelope, although both pathways involve Sir proteins (Ruault et al. 2011).

Within silent chromatin, the anchoring function has been assigned to the PAD of Sir4 (aa 950–1262) and its interaction with the nuclear envelope associated protein, Esc1 (Andrulis et al. 2002; Taddei et al. 2004). Sir4-Esc1 interactions tether the Sir-repressed chromatin domain at perinuclear sites distinct from pores. Even in the absence of a yKu anchoring pathway, the association of telomeres with the nuclear periphery can be achieved by Sir4-Esc1 interactions, as long as silent chromatin is formed (Hediger et al. 2002). Thus, rings of silent chromatin excised from their chromosomal context by recombination and lacking TG repeats, remain associated with perinuclear foci in a Sir-dependent manner (Gartenberg et al. 2004).

Initially, telomeres are recruited to the nuclear envelope by yKu, given that yKu-dependent tethering occurs even in the absence of silencing. This interaction with the nuclear envelope, together with interactions in trans between telomeres, generates a nuclear subcompartment that appears to sequester Sir proteins from the rest of the nucleoplasm (Fig. 9). yKu mediated anchoring is achieved either through yKu-Sir4 interaction, or through the interaction of yKu with telomerase (Schober et al. 2009). The Est1 subunit of telomerase binds specifically to an inner nuclear membrane-spanning protein, Mps3, which is a member of the conserved SUN domain family of inner nuclear envelope proteins. Intriguingly, this interaction is cell-cycle specific and mediates the link between yKu and the nuclear envelope only in S phase, possibly to maintain anchorage while subtelomeric chromatin is disrupted by the replication fork. In G₁ phase, there appears to be a secondary yKu anchoring pathway, and Sir4 may bind Mps3 through an intermediary to help tether telomeres (Bupp et al. 2007).

Both pathways of telomere anchoring, through Sir4-Esc1 and S-phase yKu-Est1-Mps3 pathway, are controlled by post-translational modifications. Sir4 and both subunits of yKu are modified by SUMO, a ubiquitin-like moiety, deposited specifically by the E3 SUMO ligase Siz2. Remarkably, loss of Siz2 led to the displacement of telomeres from the nuclear envelope (Ferreira et al. 2011). Silencing was reduced only slightly in siz2 mutants, possibly because telomeres became abnormally long due to telomerase deregulation (see Section 12). Thus, the perinuclear compartment created by telomere anchoring and silent chromatin appears to regulate telomere functions beyond silencing.

The perinuclear clustering generates a subnuclear compartment that favors silencing (Fig. 9). Evidence supporting this conclusion includes the fact that silencer-flanked HM constructs repress less efficiently when they are integrated far from telomeres (Thompson et al. 1994b; Maillet et al. 1996), and this can be reversed by artificially tethering the domain at the nuclear envelope through a targeted transmembrane factor (Andrulis et al. 1998). Importantly, the ability to improve repression by peripheral tethering (or by being placed in telomere proximity) is lost when Sir3 and Sir4 are no longer sequestered in foci (Taddei et al. 2009). Similarly, coordinate overexpression of Sir3 and Sir4 proteins ablates the positive effects of tethering. Thus, an uneven distribution of Sir proteins is the relevant feature of telomere sequestration at the nuclear envelope for chromatin-mediated repression. Given that displaced Sir proteins can repress promiscuously (Taddei et al. 2009), the sequestration of Sir foci also positively reinforces active gene expression. It is proposed that the assembly of newly replicated DNA into heterochromatin is likely to be favored when DNA is replicated in a zone rich in silencing factors.

The crucial interaction that mediates telomere–telomere clustering, even in the absence of silencing, appears to be Sir3. Whether this is mediated by Sir3 itself or ligands of Sir3 remains to be determined, yet some form of clustering can occur even in the absence of Sir2 and Sir4 (Ruault et al. 2011). Other factors also affect telomere–telomere interaction, namely, the other Sir proteins, the Ku heterodimer, Asf1, Rtt109, Esc2, the Cohibin complex, and two factors involved in ribosome biogenesis, Ebp2 and Rrs1. However, because these factors also affect heterochromatin formation, one cannot rule out that they promote clustering by promoting Sir3 recruitment to telomeres.

11. TELOMERE LOOPING

A further long-range interaction may stem from the folding back of a single telomere on itself, which may allow silent chromatin to bypass subtelomeric boundary elements and stabilize repressed chromatin at subtelomeric genes (Figs. 5 and 6). Although Rap1 binding sites are found only within the first ∼300 bp of TG repeat DNA on the end of a telomere, ChIP showed that Rap1 is associated with nucleosomes as far as ∼3 kb away from the TG repeat (Strahl-Bolsinger et al. 1997). Similarly, yKu is recovered for ∼3 kb from the chromosomal end to which it binds (Martin et al. 1999). When silencing is disrupted by mutation of SIR genes, both Rap1 and yKu are lost exclusively from the more internal subtelomeric sequences and not from the terminal TG repeats (Hecht et al. 1996; Martin et al. 1999). This was interpreted as showing that the truncated telomere folds back, enabling TG-bound Rap1 and yKu to bind Sir proteins in trans (Figs. 5 and 6).

Evidence for telomere looping comes from the work of de Bruin et al. (2001), who have exploited the inability of yeast transcriptional activators, such as Gal4, to function from a site downstream of the targeted gene. Strains were constructed in which the Gal4 upstream activating sequence (UAS) element was placed beyond the 3′ termination site of a reporter gene, and the construct was inserted either at an internal chromosomal location or near a telomere. At an internal site, this construct could not be induced by activating Gal4, but in a subtelomeric context, the Gal4 UAS could activate the promoter from a site 1.9 kb downstream of the promoter. This was Sir3-dependent, arguing that the telomeric end can fold back in the presence of Sir3, but not in its absence, to allow the Gal4 UAS to position itself proximal to the transcription start site. In this way, silent chromatin appears to promote at least a transient folding of the chromosome end.

12. VARIABLE REPRESSION AT NATURAL SUBTELOMERIC DOMAINS

We have set forth here a simplistic view of continuous silent chromatin emanating from the telomeric Rap1 binding sites, yet the situation at native telomeres is significantly more complex, largely because of the presence of natural boundary elements found in subtelomeric repeat sequences. Generally, when reporter constructs for telomeric repression are integrated, the subtelomeric repeat elements called X and Y′ at telomeres are deleted, placing the reporter gene and unique sequence immediately adjacent to TG repeats. All native telomeres, on the other hand, contain a core subtelomeric repeat element, X, which is positioned between the TG repeat and the most telomere-proximal gene, and 50%–70% of native telomeres also contain at least one copy of a larger subtelomeric element called Y′ (Fig. 10). Both X and Y′ elements contain binding sites for the transcriptional regulators Tbf1 and Reb1, which have been shown to reduce the spread of silent chromatin (Fourel et al. 1999). However, X elements also contain autonomously replicating sequence (ARS) consensus sequences, and binding sites for Abf1 and other transcription factors, which have the opposite effect: These re-initiate or boost the repression of reporters placed on the centromere proximal side of these elements. The result is one of discontinuity in silencing at many native telomeres. This adds a level of complexity to the model of continuous spreading outlined in Figure 6. Pryde and Louis (1999) have proposed that the unrepressed Y′ element loops out when it is found between two repressed domains, leading to discontinuity in silent domains without eliminating the need for nucleation and spreading from the TG repeats.

There is large variation in the efficiency of TPE at different native telomeres in budding yeast. If one inserts a reporter gene near a telomere without deleting the subtelomeric repeats, only about half of the telomeres appear to be subject to TPE (Pryde and Louis 1999). Empirically, it was shown that telomeres containing only X subtelomeric elements (rather than XY′ telomeres) are more likely to silence, possibly because the Y′ long terminal repeats bind factors that prevent Sir spreading, whereas various transcription factors in the X element (e.g., Reb1, Tbf1, and Abf1) contribute to Sir factor nucleation (Mak et al. 2009). These subtelomeric elements are particularly enriched for factors that regulate stress genes, whose effects on TPE may differ from their effects on transcription at nontelomeric loci. For example, Reb1 has boundary activity at telomeres (Fourel et al. 1999), but is a gene activator at internal sites. The binding of such factors can explain the discontinuity of Sir-mediated repression at native telomeres, yet they do not appear to affect the repression of classical reporters for TPE, which are integrated without X or Y′ elements (Renauld et al. 1993).

Interestingly, many of the genes found in subtelomeric domains are repressed by the HDAC Hda1 and the repressor Tup1 (Robyr et al. 2002), and are induced only on stress conditions (Ai et al. 2002). In general, the 267 genes that are within 20 kb of budding yeast telomeres produce roughly five times fewer mRNA molecules (average of 0.5/cell) than nontelomeric genes. Importantly, only 20 of these genes show derepression upon loss of Sir3. This leads to a logic of genome organization in which genes that are rarely or conditionally induced are found near telomeres, in which they are repressed by a non-Sir mechanism, but are adjacent to domains repressed by Sir proteins and, therefore, also tethered near the nuclear envelope.

13. INHERITANCE OF EPIGENETIC STATES

A universal characteristic of heterochromatin is that its silent state is passed from one generation to the next. This requires the reassembly of a repressive chromatin structure on daughter strands soon after replication of the DNA template. Pioneering work on the role of the cell cycle in the establishment or inheritance of the silent state was performed by Miller and Nasmyth, who studied the onset and loss of silencing with a temperature-sensitive sir3^ts mutant (Miller and Nasmyth 1984). A shift from permissive temperature to nonpermissive temperature caused silencing to be lost immediately, indicating that SIR3 was required for maintenance of the repressed state. However, in the reciprocal experiment, shifting from nonpermissive temperature to permissive temperature (SIR3+) did not lead to immediate restoration of repression; passage through the cell cycle was required. They concluded that an event in S phase was required for establishment of heritably repressed chromatin. This requirement was later shown to involve events in both S and G₂/M phases (Lau et al. 2002).

Initially it was thought that origin firing from the silencer linked ARS elements might be a critical event in the establishment or inheritance of silent chromatin, but because there was no detectable initiation from the origins flanking the HML locus, this seemed an unlikely explanation. Indeed, an experiment showing that ORC can be efficiently replaced by a targeted GDB-Sir1 fusion protein put to rest the notion that origin firing is essential for the inheritance of silent chromatin. In addition, recent experiments have shown that establishment of repression can occur on DNA that does not replicate (Kirchmaier and Rine 2001; Li et al. 2001). Nonetheless, on rings of silent chromatin that are excised from the genome either with or without silencers, Sir complex association is in continual flux, arguing that nucleation and/or stabilization provided by silencer elements at HM loci is needed to actively suppress rapid decay (Cheng and Gartenberg 2000).

What occurs in S phase to enable the propagation of silent chromatin? One candidate event could be the deacetylation of H4K16ac or H3K56ac, or else the suppression of the enzymes that deposit these modifications (Xu et al. 2007; Neumann et al. 2009). Alternatively, chaperones necessary for histone deposition (CAF1) may be required for generating repressed chromatin after replication. The deacetylation of histone H3K56 can be achieved in vitro by Sir2 family members (Xu et al. 2007; Oppikofer et al. 2011) and in vivo during late S phase by Hst3 and Hst4 (Maas et al. 2006; Celic et al. 2008; Yang et al. 2008b). These two Sir2 paralogs are nuclear enzymes whose activities are required in S phase when newly synthesized DNA must be assembled into repressed chromatin, yet they are not structural components of heterochromatin. Importantly, disruption of the two genes weakens, but does not eliminate, Sir mediated silencing (Yang et al. 2008b).

Other studies have shown that robust silencing is not achieved until telophase, well beyond the S-phase window of nucleosome assembly. It appears that prevention of the metaphase degradation of the cohesin subunit, Scc1, inhibits stable repression (Lau et al. 2002). Propagation of repression thus depends both on a critical S-phase component, and a further event that entails continual recruitment and loading of Sir proteins.

Intriguingly, when the proteins of telomeric heterochromatin were examined by ChIP in the transcriptionally OFF and ON states, the major difference found between them was the presence of H3K79 methylation at telomeric chromatin in the ON state (Kitada et al. 2012). Because Dot1 is recruited during transcription (Shahbazian et al. 2005), this suggests a positive feedback loop in which the ON state is triggered, possibly by decreased telomere length, to initiate transcription and K79 methylation. Dot1 would then be responsible for maintaining the ON state through its promotion of K79 methylation.

14. OTHER FUNCTIONS OF Sir PROTEINS AND SILENT CHROMATIN

Although the standard function of heterochromatin is the silencing of adjacent genes, a closer examination of silencing factors has uncovered a plethora of new functions that correlate with silencing or require silencing factors. Particularly in organisms in which repetitive centromeric DNA plays a crucial role in centromere function, it is clear that heterochromatin contributes to centromere and kinetochores function. Budding yeast, on the other hand, does not depend on silent chromatin for centromere function. Nonetheless, a number of other roles have been identified for silent chromatin or silent chromatin factors, and these are described in this section.

15. SUMMARY

Combined genetic, biochemical, and cytological techniques have been exploited in budding yeast to show fundamental principles at work during heterochromatin-mediated gene silencing. These principles include (1) the mechanism of initiation, spreading, and barriers to spreading of heterochromatin; (2) the balance of heterochromatin factors and their distribution within a subnuclear environment; (3) higher-order folding of heterochromatin; and (4) the effects of cell-cycle involvement in its formation. The in vivo and in vitro studies to date provide a strong mechanistic basis for our understanding of the assembly of heterochromatin from chromatin fibers in all eukaryotes. The in vitro systems currently being developed for the reconstitution of yeast heterochromatin promise to yield a structural reconstruction of Sir complexes bound to chromatin, enabling a first glimpse at higher-order chromatin folding.