Rpd3-dependent boundary formation at telomeres by removal of Sir2 substrate

Abstract

Boundaries between euchromatic and heterochromatic regions until now have been associated with chromatin-opening activities. Here, we identified an unexpected role for histone deacetylation in this process. Significantly, the histone deacetylase (HDAC) Rpd3 was necessary for boundary formation in Saccharomyces cerevisiae. rpd3Δ led to silent information regulator (SIR) spreading and repression of subtelomeric genes. In the absence of a known boundary factor, the histone acetyltransferase complex SAS-I, rpd3Δ caused inappropriate SIR spreading that was lethal to yeast cells. Notably, Rpd3 was capable of creating a boundary when targeted to heterochromatin. Our data suggest a mechanism for boundary formation whereby histone deacetylation by Rpd3 removes the substrate for the HDAC Sir2, so that Sir2 no longer can produce O-acetyl-ADP ribose (OAADPR) by consumption of NAD⁺ in the deacetylation reaction. In essence, OAADPR therefore is unavailable for binding to Sir3, preventing SIR propagation.

The functional distinction between euchromatic and heterochromatic domains within eukaryotic genomes is essential to maintain gene expression programs that drive development and differentiation in higher organisms (1). Each expression domain must maintain its identity, and thus junctions exist that separate active from inactive regions and maintain opposing transcriptional states. Barriers between chromatin states have been described in a variety of organisms (1). In Saccharomyces cerevisiae, histone acetylation by several histone acetyltransferase (HAT) complexes (2–5), H3 K79 methylation by the histone methyltransferase Dot1 (6), and H3 K4 methylation by Set1 (7) have been associated with boundary formation. Some modifications also may cooperate with chromatin remodeling activities to implement barrier function (4). Other chromatin alterations such as the complete loss of nucleosomes (4, 5) or incorporation of the histone variant H2A.Z (8) also are associated with a block to silencing.

Although some boundaries are fixed to a certain genomic position, others are characterized by a balance of opposing enzymatic activities and the competition between chromatin-opening and -condensing complexes. One such example is telomeric heterochromatin in S. cerevisiae, where histone deacetylation by the NAD⁺-dependent histone deacetylase (HDAC) Sir2 (9) is required for the repressive silent information regulator (SIR) complexes to bind to the chromatin (10). Deacetylation by Sir2 at telomeres is counteracted by the HAT complex SAS-I, which contains the MYST family HAT Sas2 and acetylates H4 K16. Thus, the competing activities of SAS-I and Sir2 create flexible boundaries between eu- and heterochromatin at telomeres via de-/acetylation of H4 K16 (11, 12).

Interestingly, the deacetylation reaction of Sir2 is distinct from that of non–NAD⁺-dependent HDACs in that it produces an unusual compound, O-acetyl-ADP ribose (OAADPR) (13), which has been proposed to influence SIR complex stability (14). Intriguingly, the Sir3 protein carries a domain that resembles the ATP binding pocket of AAA⁺ ATPases but lacks certain catalytic residues (15). It therefore has been hypothesized that this domain constitutes an OAADPR binding site (16).

SAS-I globally acetylates H4 K16 in subtelomeric regions (17). Similarly, the HDAC Rpd3 provides global chromatin deacetylation in addition to its function in local gene repression (18). Rpd3 is present in two different HDAC complexes: Rpd3(L) is targeted to gene promoters and establishes gene repression via promoter deacetylation, whereas Rpd3(S) provides deacetylation in the body of genes and prevents intragenic transcription (19, 20). The deletion of RPD3 leads to higher global acetylation levels (21). Intriguingly, Rpd3-mediated histone deacetylation also has been invoked in Hog1-dependent activation of osmosensitive genes (22), and rpd3Δ causes increased silencing in yeast as well as in Drosophila (23–25).

Here, we performed a screen for factors that become essential in the absence of the HAT Sas2. We found that deletion of RPD3 is lethal in sas2Δ cells and that excessive spreading of heterochromatic SIR complexes is responsible for the lethality. Rpd3 is necessary to restrict the SIR proteins to the telomeres, and targeting Rpd3 to normally silent chromatin creates a barrier to the spreading of SIR-dependent repression. Our data suggest a mechanism for boundary formation, in that Rpd3 effectively removes the substrate for Sir2 at the heterochromatin–euchromatin boundary. Therefore, heterochromatin spreading is stopped by the inability of Sir2 to perform histone deacetylation, to produce OAADPR, and thus to support heterochromatin spreading.

Results

Deletion of SAS2 and RPD3 Is Synthetically Lethal.

The SAS-I complex globally acetylates H4 K16 (26, 27). To search for factors that become essential in the absence of Sas2, we performed a synthetic lethal screen with sas2Δ cells. Surprisingly, we found that the deletion of RPD3 is lethal in the absence of Sas2 (Fig. 1 A and B), a finding that is in agreement with an earlier study (28). At higher temperatures, rpd3Δ sas2Δ double mutants are nonviable. Furthermore, the lethality requires the acetyl-CoA binding site as well as the atypical zinc finger of Sas2 (Fig. S1). We found that sas4Δ and sas5Δ also are synthetically lethal with rpd3Δ (Fig. 1A), showing that the whole SAS-I complex is involved in the lethality with rpd3Δ. Furthermore, the lethality depends upon the Rpd3(L), not the Rpd3(S), complex, because sas2Δ shows synthetic growth defects in the absence of the Rpd3(L) components Dep1 and Sds3 but not the Rpd3(S) components Rco1 and Eaf3 (Fig. 1B and Table S1), and the additional deletion of RCO1 does not exacerbate the growth defect of sas2Δ sds3Δ cells (Fig. 1B). However, set2Δ does not cause a growth defect in sas2Δ cells (Fig. 1C), indicating that neither Set2-dependent recruitment of Rpd3(S) (20) nor an Rpd3(S)-independent function of Set2 at telomere boundaries (29) is involved in the synthetic lethality between sas2Δ and rpd3Δ. The lethality is reflected further in an increased sensitivity of sas2Δ cells to treatment with the HDAC inhibitor trichostatin A (Fig. 1D). No other HAT is lethal in combination with rpd3Δ, and no other HDAC is lethal with sas2Δ (Table S1), indicating that the lethality between sas2Δ and rpd3Δ is specific to these two enzymes.

Fig. 1.

Synthetic lethality between the Rpd3(L) and the SAS-I complex. (A) Cells disrupted for subunits of the SAS-I complex are synthetically lethal with rpd3Δ. Tetrad dissection of crosses of sas2Δ, sas4Δ or sas5Δ with rpd3Δ isogenic W303 strains. The four spores from individual asci are aligned in vertical lines. Double mutants are marked with circles. (B) The lethality between rpd3Δ and sas2Δ is specific for the Rpd3(L) complex. sas2Δ cells with pURA3-SAS2 and additional deletions of RCO1, SDS3, or both were incubated on supplemented yeast minimal medium (YM) or 5-fluoroorotic acid (5-FOA) medium. (C) set2Δ does not cause synthetic lethality in sas2Δ. (D) sas2Δ cells are sensitive to trichostatin A (TSA). Filter discs with DMSO control (Upper row) or increasing amounts (16, 12 and 8 μg) of TSA (Lower row) were placed on lawns of 2 × 10⁵ cells of a WT or sas2Δ strain. Plates were incubated for 1 day at 30°C.

The Lethality Between sas2Δ and rpd3Δ Is Caused by Increased SIR Spreading.

SAS-I exerts a boundary function at telomeres in that it acetylates H4 K16 in subtelomeric regions, preventing the heterochromatin-like SIR complex from spreading toward more centromere-proximal regions (11, 12). We hypothesized that Rpd3 also might exert a boundary function similar to that of the SAS-I complex. The combined effect of SIR spreading by rpd3Δ and sas2Δ thus might cause increased SIR spreading to a degree that is lethal to the cells because of the repression of one or several essential subtelomeric genes. If this supposition is true, then relieving telomeric silencing by deleting one of the SIR components should abrogate the sas2Δ rpd3Δ lethality. Remarkably, we found that the deletion of SIR2, SIR3, or SIR4 completely suppresses the lethality of sas2Δ rpd3Δ, showing that the lethality between sas2Δ and rpd3Δ depends on the SIR proteins (Fig. 2A). This suppression is the result of the absence of the SIR complex at the telomeres rather than of changes in cell type or HM derepression that result from deletion of SIR2, SIR3, or SIR4. Specifically, the deletion of SIR1, which affects HM but not telomeric silencing, does not suppress the lethality between sas2Δ and rpd3Δ (Fig. 2A). Also, deletion of HMR in a MATα sas2Δ rpd3Δ sir2Δ strain, which reverses the pseudodiploid cell type, does not abrogate the viability of the strain.

Fig. 2.

The synthetic lethality between the Rpd3(L) and the SAS-I complex is caused by inappropriate SIR spreading. (A) Deletions of subunits of the telomeric SIR complex suppress the sas2Δ rpd3Δ synthetic lethality. Derivatives of an sas2Δ rpd3Δ pURA3-SAS2 strain (AEY 3923) with deletions for SIR1, SIR2, SIR3, or SIR4 were grown on minimal plates (YM, growth assay) and on 5-FOA plates to select against pURA3-SAS2 for 2 days at 30°C. (B) Mutations within the histone H3 and H4 N-termini suppress the sas2Δ rpd3Δ synthetic lethality. Alleles of H3 and H4 were introduced into an rpd3Δ sas2Δ pLYS2-SAS2 strain by plasmid shuffle (AEY3945; SI Materials and Methods for details), and the ability of the derivatives to survive in the absence of the SAS2 plasmid was tested. “wt” refers to WT copies of H3 and H4 in AEY3945. “Control” refers to a WT strain. H4 K -> Q designates H4 K5, 8, 12, and 16 Q. H3 K -> Q designates H3 K4, 9, 14, 18, 23, and 27 Q.

The binding of SIR complexes to the telomeres depends on the acetylation state of the amino-terminal histones of H3 and H4 and, in particular, on H4 K16 (10). Therefore, mutation of critical histone residues that abrogate silencing should suppress the sas2Δ rpd3Δ synthetic lethality. In fact, we observed that mutation of the acetylatable lysine residues in the tails of H3 or H4 allows growth of sas2Δ rpd3Δ strains and that a sole mutation of H4 K16R is sufficient for the suppression (Fig. 2B). Such mutations have been shown previously to abrogate silencing (Fig. S2) (30). The H4 K16R suppression may seem counterintuitive, given that mutations of lysine to arginine generally are thought to mimic the deacetylated state of the lysine residue and therefore would be expected to improve SIR binding and silencing. Contrary to this assumption, H4 K16R mutations cause a strong loss of telomeric silencing and defects in HM silencing (Fig. S2) (26, 30), showing that the K16R mutation is not equivalent to a deacetylated lysine. We propose that H4 K16R suppresses the sas2Δ rpd3Δ lethality through its derepressing effect on telomeric silencing.

To assess boundary function of Rpd3, we measured SIR levels at the telomeres by performing ChIP. The absence of Rpd3 leads to more Sir2 and Sir3 bound to telomeres and the presence of more Sir2 and Sir3 in centromere-proximal regions of the right arm of chromosome VI (Fig. 3A). The effect of rpd3Δ is distinct from that of sas2Δ, which leads to a shift of Sir2 toward centromere-proximal sequences with less Sir2 at sequences close to the telomere (11, 12), (Fig. 3A).

Fig. 3.

Deletion of RPD3 causes mislocalization of Sir2 and Sir3, gene silencing, and changes in histone acetylation in subtelomeric regions. (A) Sir2 and Sir3 binding at the right telomere of chromosome VI is shown as enrichment in ChIP experiments relative to their enrichment at the control gene SPS2. The amount of enrichment is given as a function of the distance to the telomere end in kb in strains with the indicated genotype. ChIP was performed with antibodies against myc-Sir2 and HA-Sir3. Error bars give SD (SI Materials and Methods). (B) Telomeric ADE2 is repressed in rpd3Δ strains. WT and rpd3Δ strains carrying telomeric ADE2 were grown for 2 days at 30°C. (C) Subtelomeric genes are repressed in rpd3Δ cells in a SIR-dependent fashion that correlates with Sir2 association. The upper row of diagrams shows the amount of cDNA of selected subtelomeric genes in the indicated strains as X-fold expression level relative to SPS2. Error bars give standard deviations of at least three PCR analyses from at least two independent reverse transcriptase reactions. The lower row shows ChIP analysis of Sir2 at the respective genes and is represented as in (A). (D) rpd3Δ causes changes in subtelomeric histone acetylation levels. Acetylation was measured by ChIP using antibodies against the indicated residues. Values indicate enrichment of a modification relative to enrichment of H4 (SI Materials and Methods). (E) Sir2-independent changes in subtelomeric histone acetylation by rpd3Δ. Experiments were performed as described in (D). (F) Schematic representation of the telomere VI-R with ORFs and fragments amplified in the ChIP experiments. Genes drawn above the upper line are transcribed from left to right; those below that line are transcribed in the opposite direction.

In agreement with the observed Sir spreading in rpd3Δ cells, we observed that rpd3Δ causes a loss of colony sectoring in strains with ADE2 inserted at the telomere (Fig. 3B), reflecting increased telomeric silencing (23). Furthermore, subtelomeric genes are more repressed in the absence of Rpd3, and the repression correlates with increased levels of Sir2 at these genes (Fig. 3C). The repression is abrogated by the additional deletion of SIR2. Expression analysis of the ORFs on TEL VI-R shows that IRC7 is strongly repressed in rpd3Δ and sas2Δ cells and that the repression is relieved by additional deletion of SIR2 (Fig. 3C). HXK1 and RPN12 show no differences in expression of rpd3Δ or sas2Δ, probably because they are more than 15 kb from the telomere and therefore outside the repressed domain (Fig. 3C). Importantly, although more Sir2 and Sir3 is bound at the telomeres, this increase is not the result of increased Sir2 or Sir3 expression in rpd3Δ cells, and Sir4 expression is also unaltered (Fig. S3). In summary, our data show that Rpd3 is required to restrict Sir2 and Sir3 levels and for localization to sequences closest to the telomere and that Rpd3 prevents deleterious gene repression by mislocalized SIR complexes in subtelomeric regions.

We furthermore determined how the loss of Rpd3 affected histone acetylation at the telomeres. Acetylation of lysine 16 of H4 is higher in the absence of Rpd3, and H4 K5 acetylation is increased at some, but not all, sites tested (Fig. 3D). This finding is consistent with the notion that the loss of a deacetylase causes an increase in acetylation and shows that Rpd3 directly affects histone acetylation levels in the vicinity of the telomeres. In contrast, acetylation of H4 K12 is decreased at some, but not all, sites in rpd3Δ cells. This finding is counterintuitive, because the absence of a deacetylase is expected to increase acetylation. One explanation is that the Sir2 deacetylase spreads into these regions in rpd3Δ cells and causes histone deacetylation. In agreement with this notion, the deletion of RPD3 causes increased histone acetylation levels at the telomeres in sir2Δ cells (Fig. 3E). In summary, these results show that Rpd3 performs global histone deacetylation in subtelomeric regions.

Targeted Rpd3 Establishes a Boundary at Telomeres and the HM Loci.

We next determined whether Rpd3 is sufficient to create a boundary when targeted to a normally silenced gene. When a Gal4 binding site is present between the telomere and the reporter (3), the expression of a GBD-Rpd3 fusion disrupts URA3 silencing, whereas the expression of GBD alone causes URA3 to be silenced by telomeric heterochromatin (Fig. 4A). The boundary function of GBD-Rpd3 depends on the catalytic activity of Rpd3. Notably, GBD-Rpd3–dependent boundary formation requires native Rpd3, because rpd3Δ abrogates the boundary function of GBD-Rpd3, and GBD-Rpd3 is unable to complement rpd3Δ in other silencing assays. Significantly, WT, but not catalytically inactive, RPD3 (22) can restore boundary function of GBD-Rpd3 (Fig. 4B). Furthermore, boundary activity also is observed for tethering of another Rpd3(L) component, Dep1, to the telomere (Fig. S4), suggesting that recruitment of the catalytically active Rpd3(L) complex creates a telomere boundary.

Fig. 4.

Tethered Rpd3 and Hos2 create a boundary against heterochromatin spreading. (A) Cells with URA3 inserted at TEL VII-L and with (+UAS) or without a Gal4 binding site at the telomere-proximal side (−UAS) were transformed with plasmids carrying the RPD3 or HOS2 genes fused to the GAL4 DNA binding domain (GBD-HDAC) or with the vector control (GBD). Repression of URA3 was tested by growth on plates lacking uracil and on 5-FOA plates. Serial dilutions of cells were grown for 2 days at 30°C. (B) The targeted boundary function of Rpd3 depends on its catalytic activity. Cells with a Gal4 UAS between telomeric heterochromatin and a subtelomeric URA3 reporter (as in A) were disrupted for endogenous RPD3 and transformed with RPD3 or a catalytically dead rpd3 allele (rpd3-H150:151A, referred to as “rpd3-HDAC⁻”). (C) Tethered Rpd3 disrupts silencing at HML and has insulating activity at HMR. Cells with ADE2 and URA3 inserted at HML or HMR were provided with GBD or with GBD-Rpd3. (D) Other HDACs did not display boundary function. GBD-HDAC fusions were assayed for boundary function as in A.

We next asked whether the ability of tethered Rpd3 to disrupt heterochromatin is specific to the telomeres or whether Rpd3 also is capable of stopping SIR spreading at the HM loci. To investigate this question, we used established boundary assays with the reporter genes ADE2 and URA3 inserted at HML (31) and HMR (32) (Fig. 4 C and D). At HML, GBD-Rpd3 causes derepression of both reporter genes but no detectable insulation of ADE2 (Fig. 4C). Furthermore, tethered Rpd3 is capable of insulating ADE2 from SIR-mediated silencing at the HMR locus (Fig. 4C), because it causes ADE2 derepression while maintaining URA3 repression. This effect of tethered Rpd3 is the same as that of Sas2 at the two loci (32) and shows that Rpd3 is not only a “desilencer” but can be classified as a “true barrier” factor. These results show that targeting of the HDAC Rpd3 disrupts heterochromatin spreading. This finding is surprising, because thus far only HATs and chromatin remodeling complexes, but not HDACs, are known to create boundaries (3–5).

Removal of Sir2 Substrate as a Mechanism for Boundary Formation.

A priori, our observation of a boundary function for Rpd3 is counterintuitive, because chromatin deacetylation generally is viewed as necessary for, rather than prohibitive to, SIR spreading in telomeric regions. One possibility is that deacetylation by Rpd3 is a prerequisite for the onset of another modification of residues deacetylated by Rpd3; alternatively, the boundary function of Rpd3 may influence chromatin remodeling, exchange of histone variants, or the presence of linker histones. However, the analysis of sas2Δ genetic interactions (Table S1) did not support these scenarios. Furthermore, the prevention of SIR spreading is not mediated by Bdf1 or Bdf2 (Table S1).

We next hypothesized that chromatin deacetylation per se by Rpd3 is the cause for the establishment of a chromatin boundary by removing the acetyl-lysine substrate for the HDAC Sir2. In this model, the process of deacetylation by Sir2 helps in the propagation of SIR complexes along the chromatin fiber. Consequently, this process is hindered by prior removal of acetyl-lysine residues by Rpd3.

This model predicts that targeting other HDACs to the telomeres also should create a barrier to heterochromatin, depending on their substrate specificity. We found that tethering the HDAC Hos2 (33) forms a boundary to telomeric silencing (Fig. 4A), showing that, in principle, histone deacetylation by other HDACs also can cause boundary formation. Other HDACs, however, do not display this activity (Fig. 4D), whereas targeting of Hst2 or Sir2 aids in heterochromatin formation (Fig. 4D and Fig. S4). We propose that these HDACs are unable to create a boundary because their different substrate specificities are incompatible with boundary function, although it also is possible that some of the HDACs lose activity by fusion to the Gal4 GBD.

How does the process of deacetylation by Sir2 contribute to SIR propagation? One possibility is suggested by the fact that Sir2 produces OAADPR in the deacetylation reaction (13), which binds to the SIR complex and has been proposed to be one of the driving forces in the polymerization of SIR complexes on chromatin (14). In this scenario, removal of Sir2 substrates renders Sir2 unable to produce OAADPR, thus reducing SIR propagation along the chromatin fiber and stopping heterochromatin spreading. This model predicts that a mutation in the OAADPR binding site within the SIR complex should abrogate SIR spreading and silencing. Sir3 contains a domain similar to the nucleotide (ATP) binding domain of AAA⁺ ATPases (16), making it a likely candidate region for an OAADPR binding pocket. Our modeling of Sir3 on the structure of AAA⁺ ATPases suggests that Sir3 contains an additional cavity as compared with other ATPases that may accommodate the O-acetyl-ribose moiety of OAADPR (Fig. 5A).

Fig. 5.

The putative OAADPR binding domain of Sir3 is necessary for its function in silencing. (A) Model of the AAA+ ATPase-like domain of Sir3. For a computational modeling of Sir3, the primary sequence of Sir3 AAA⁺ domain and structures of other proteins of the AAA⁺ family were subjected to a sequence-structure alignment, using Modeler 9v2 (see SI Materials and Methods for details). The model was visualized using PyMOL. The position of the putative OAADPR binding domain is marked by OAADPR. The mutations of Sir3 used in this study are indicated in red and blue. (B) Mutation of the AAA⁺ domain of Sir3 abrogates its ability to spread at telomeres. rpd3Δ sas2Δ sir3Δ pURA3-SAS2 strains carrying the indicated sir3 alleles were tested for their ability to lose the SAS2 plasmid on 5-FOA. sir3-AAA designates mutation of residues 575–577 to alanine. (C and D) sir3-Δ575–577 and sir3-AAA were deficient in telomeric and HML silencing. (E) Mutation of the AAA⁺ domain of Sir3 (Sir3-Δ578–585, designated “Sir3*”) reduces its ability to interact with Sir4 and Sir3. Fusions of the indicated proteins to the Gal4 DNA-binding (BD) and activation domain (AD) were tested for activation of the HIS3 reporter gene on medium lacking histidine. (F) Sir3Δ578–585 and Sir3Δ575–577 show reduced binding to the telomere. ChIP analysis was performed with Sir3-HA.

We tested whether a mutation in this region of Sir3 could abrogate its ability to support silencing. Importantly, alleles of SIR3 that deleted amino acids 575–577 or 578–585 in the Sir3 protein or that mutated residues 575–577 to alanine and therefore affected the putative OAADPR binding site are unable to restore the lethality in sas2Δ rpd3Δ sir3Δ cells (Fig. 5B), and they are unable to support telomeric and HM silencing (Fig. 5 C and D), indicating that they have lost functionality although they are expressed at levels comparable to endogenous Sir3 (Fig. S3C). Furthermore, sir3-Δ578–585 causes a reduction, but not complete loss, of interaction with Sir3 itself and a loss of Sir3 interaction with Sir4 (Fig. 5E), suggesting that the binding of OAADPR to Sir3 is important for SIR complex integrity. ChIP analysis of WT and the two mutant Sir3 proteins shows that the mutants display reduced binding to telomeric sequences (Fig. 5F). This finding suggests that the deletion or mutation of the OAADPR binding site causes reduced Sir3 function in heterochromatin spreading. In summary, these results show that the AAA⁺ domain of Sir3 is important for its silencing function, suggesting that OAADPR binding to Sir3 is critical for heterochromatin formation.

Discussion

Barriers between active and inactive chromatin in a variety of organisms have been associated with chromatin-activating mechanisms as well as with the attachment at nuclear pore structures (1, 31). Here, we describe the unexpected finding that an enzymatic activity generally associated with repression, the HDAC Rpd3, is necessary to prevent spreading of heterochromatic SIR proteins into euchromatin at yeast telomeres. In principle, one would expect an HDAC to aid in, rather than to prevent, the formation of heterochromatin, although Rpd3 has been implicated in gene activation previously (22, 33). We propose that the boundary function for Rpd3 described here reflects a global, untargeted (versus a targeted) role for the Rpd3(L) complex in establishing histone acetylation patterns at telomeres through a transient interaction with chromatin. Our data are thus in synchrony with previous work showing that Rpd3 antagonizes SIR propagation at telomeres (34). However, the mechanism of this boundary function thus far has remained unclear.

How can a histone deacetylase function as a boundary element? Our data suggest that histone deacetylation as such halts heterochromatin spreading, because two known HDACs display boundary activity. However, one still can speculate that other nonhistone targets also contribute to the boundary activity (34) or that Rpd3 deacetylation supports the binding of another factor that hinders SIR spreading. Rpd3 and Hos2 both have a relatively broad histone substrate range: Rpd3 deacetylates all acetylation sites in the H3 and H4 N-termini as well as H2B K11, K16, and H2A K7, and Hos2 is required for the preferential deacetylation of all sites in H3 and H4 (33, 35). Although currently we cannot distinguish which residues are important for boundary function, the broad substrate specificity suggests that the more histone residues are deacetylated by an HDAC, the more likely the HDAC is to possess boundary activity. This notion is in line with the observation that mutation of H4 K5 partially abrogates increased silencing by rpd3Δ (34). However, boundary activity must be independent of H4 K16, because the acetylation, rather than the deacetylation, of H4 K16 by SAS-I creates a boundary (11, 12). This finding suggests that a functional difference exists between acetylation of H4 K16 and other lysine residues. We propose that K16 is acetylated by SAS-I before or concomitant with chromatin deposition and that it becomes inaccessible to Sir2 deacetylation once deposited, so that it cannot be used by Sir2 for SIR propagation along the chromatin fiber. Perhaps acetylated H4 K16 is inaccessible because it alters the chromatin configuration (36), which is not changed by other modifications. Alternatively, recruitment of H2A.Z by H4 K16 acetylation may alter its accessibility (17).

By which mechanism does histone deacetylation halt heterochromatin spreading? SIR binding to chromatin requires deacetylated histone tails (10). Thus, the process of deacetylation by Sir2 itself may help in the process of productive SIR propagation along the chromatin fiber. The deacetylation of lysine residues on a nucleosome subsequently may stabilize SIR binding to that nucleosome. Consequently, this process is hindered if acetyl groups have been removed previously by Rpd3 and thus are no longer available for Sir2 deacetylation. Notably, there may be a difference between mere SIR spreading and concomitant gene silencing, because two studies showed SIR spreading on mutant histones, but this SIR spreading did not provide stable silencing (37, 38).

As an extension of our model, one attractive possibility is suggested by the fact that, in the deacetylation reaction, Sir2 produces OAADPR (13), which binds to the SIR complex and has been proposed to be one of the driving forces in the polymerization of SIR complexes on chromatin (14). In this scenario, removal of the Sir2 substrate renders Sir2 unable to produce OAADPR, thus reducing the efficiency of SIR propagation and functional silencing along the chromatin fiber and stopping heterochromatin spreading and silencing (Fig. S5). This model also suggests that OAADPR is not supplied in trans by the other cellular NAD⁺-dependent HDACs, Hst1–Hst4, in the absence of OAADPR production by Sir2. This possibility is quite conceivable, because OAADPR most likely is unstable in the cellular environment and because several nucleotide-cleaving enzymes exist in the cell.

We observed that mutations in the putative OAADPR binding site of Sir3 abrogate its silencing function and reduce its ability to interact with itself and Sir4 and spread on chromatin, thus providing experimental support for the notion that binding of OAADPR to Sir3 carries out an important function in heterochromatin formation. This notion is in agreement with a recent study that showed increased in vitro binding of the SIR complex to chromatin in the presence of OAADPR (39). Perhaps the metabolite is necessary for SIR–SIR complex interactions, which are important for propagation of heterochromatin along the chromatin fiber. However, it also is possible that these Sir3 mutations disrupt other aspects of Sir3 function.

Interestingly, a recent study reported apparently OAADPR-independent silencing by a fusion protein between the NAD⁺-independent HDAC Hos3 and Sir3 (40). One explanation to reconcile this observation with our data is that OAADPR is required for the interaction and recruitment of Sir2/Sir4 to Sir3. Thus, the direct Sir3-Hos3 fusion in the previous study might obviate the necessity for OAADPR in the recruitment of the HDAC Sir2.

In summary, in this work we have expanded the view of Rpd3′s function in the establishment of global histone acetylation patterns, in that we found the Rpd3(L) complex to restrict heterochromatin to telomeric regions. This boundary activity functions by a mechanism in which the spreading of SIR complexes along the chromatin fiber is halted by prior removal of acetyl-lysine groups on histones by Rpd3. We propose that the deacetylation reaction of Sir2 per se, the generation of OAADPR by Sir2, and the binding of the metabolite to Sir3 are essential for SIR spreading and are abrogated by the competing histone deacetylation activity of Rpd3 in subtelomeric regions (Fig. S5).

Materials and Methods

Yeast Strains and Plasmids.

Yeast strains and plasmids used in this study are listed in Tables S2 and S3, respectively. Yeast was grown and manipulated according to standard protocols (SI Materials and Methods).

Chromatin Immunoprecipitation.

ChIP was performed essentially as described, with modifications and analysis performed as described in SI Materials and Methods.

Expression Analysis.

The expression of subtelomeric genes was determined by reverse transcription followed by quantitative real-time PCR. Total RNA from 0.5 OD units of yeast cells was reverse transcribed using SuperScript III reverse transcriptase (Invitrogen) according to the manufacturer’s protocol.