Asymmetric Positioning of Nucleosomes and Directional Establishment of Transcriptionally Silent Chromatin by Saccharomyces cerevisiae Silencers

In Saccharomyces cerevisiae, silencers flanking the HML andHMR loci consist of various combinations of binding sites forAbf1p, Rap1p, and the origin recognition complex (ORC) thatserve to recruit the Sir silencing complex, thereby initiatingthe establishment of transcriptionally silent chromatin. Therehave been seemingly conflicting reports concerning whether silencersfunction in an orientation-dependent or -independent manner,and what determines the directionality of a silencer has notbeen explored. We demonstrate that chromatin plays a key rolein determining the potency and directionality of silencers.We show that nucleosomes are asymmetrically distributed aroundthe HML-I or HMR-E silencer so that a nucleosome is positionedclose to the Abf1p side but not the ORC side of the silencer.This coincides with preferential association of Sir proteinsand transcriptional silencing on the Abf1p side of the silencer.Elimination of the asymmetry in nucleosome positioning at asilencer leads to comparable silencing on both sides. Asymmetricnucleosome positioning in the immediate vicinity of a silenceris independent of its orientation and genomic context, indicatingthat it is the inherent property of the silencer. Moreover,it is also independent of the Sir complex and thus precedesthe formation of silent chromatin. Finally, we demonstrate thatasymmetric positioning of nucleosomes and directional silencingby a silencer depend on ORC and Abf1p. We conclude that theHML-I and HMR-E silencers promote asymmetric positioning ofnucleosomes, leading to unequal potentials of transcriptionalsilencing on their sides and, hence, directional silencing.

INTRODUCTION

Top

Introduction

The establishment of condensed and transcriptionally silentheterochromatin in the eukaryotic genome is achieved via aninitiation process that recruits silencing/repressor complexesto specific nucleation sequences, followed by their propagationalong the chromatin. A silencing complex usually contains ahistone-modifying enzyme(s) responsible for yielding heterochromatin-specific"histone codes" and structural proteins that recognize thesecodes and bind the modified histones with high affinity (13).In the yeast Saccharomyces cerevisiae, transcriptional silencingof the HML and HMR loci as well as regions near the telomeresis mediated by a silent chromatin that shares many similaritieswith metazoan heterochromatin at the molecular level (13). Silentchromatin at the HM loci is established by combined actionsof cis-acting DNA elements and trans-acting proteins (32). Thecis-acting elements are small specialized sequences, known asthe E and I silencers, that flank the HM loci (Fig. 1A). Eachsilencer is composed of a unique combination of binding sitesfor proteins Abf1p, Rap1p, and ORC (for "origin recognitioncomplex for DNA replication") (Fig. 1A), whose binding is requiredfor silencing (32). The trans-acting factors include silencer-bindingproteins, histones, and Sir1p through Sir4p. Sir2p is an NAD-dependenthistone deacetylase that is believed to be responsible for thereduction of histone acetylation in silent chromatin (25, 32).Sir2p through Sir4p interact with each other and form the Sircomplex that can interact with the silencer-binding proteins,as well as with histones H3 and H4. Moreover, the Sir complexhas a higher affinity for histones H3 and H4 with reduced acetylation,which is important for the establishment and maintenance ofsilent chromatin (8, 19, 32).

View larger version (63K):
[in this window]
[in a new window]
FIG. 1. Unidirectional silencing by the HMR-E and HML-I silencers. (A) HML and HMR on chromosome III in S. cerevisiae (not drawn to scale). HML-E and -I are shown as filled arrows with white letters, and HMR-E and -I are shown as open arrows with black letters. Note the directions of silencers are drawn pointing toward the inside of HML or HMR but are not necessarily the functional directions. The sites for Abf1p, Rap1p, and ACS (ARS consensus sequence) for ORC recognition in silencers are indicated by stippled, filled, and open rectangles, respectively. They are variants of motifs 5'-RTCRYNNNNNACG-3' (for Abf1p), 5'-ACACCCRYACAYM-3' (Rap1p), and 5'-WTTTAYRTTTW-3' (ORC) (M, A/C; N, A/C/T/G; R, A/G; Y, C/T; W, A/T) (6, 18, 29). Arrows above the factor-binding sites correspond to the 5' $->$ 3' directions of these consensus sequences. The ${alpha}$ 1 and ${alpha}$ 2 genes at HML and a1 and a2 genes at HMR are also indicated. CEN, centromere. Tandem arrowheads indicate telomeric repeats. (B) HMR-E and HML-I promote Sir-dependent directional silencing. Left, modified HML and HMR loci in strains 1 to 8 and 1s to 8s. [ ${Delta}$ ] indicates the deletion of HMR-I. Cells of each strain were grown to late log phase, and serial 10-fold dilutions were spotted on test plates and allowed to grow for 3 days. SC, synthetic complete medium. FOA, SC plus 1 mg/ml of 5-fluoroorotic acid (FOA).

The silencer-binding proteins are believed to promote the initiationof silencing by recruiting the Sir complex through physicalinteractions (32). Rap1p interacts with Sir3p and Sir4p, whereasORC interacts with Sir1p, which binds Sir4p. The Sir complexassociated with the silencer deacetylates histones in adjacentnucleosomes that then bind additional Sir complexes. The nucleosome-boundSir complex then deacetylates the neighboring nucleosome, whichin turn binds a new Sir complex. Interactions between Sir complexesalso help recruit the incoming Sir complex (19, 31). In thismanner, the Sir complex is thought to promote its own stepwise(nucleosome-by-nucleosome) propagation along a continuous arrayof nucleosomes, thereby establishing an extended region of silentchromatin (13). This model is supported by our finding thatdisrupting the regularity of nucleosomes by nucleosome-excludingstructures hinders the spread of silent chromatin (3).

Although yeast silencers promote silencing by a common mechanism,they exhibit different efficiencies of silencing. The HML-E,HML-I, and HMR-E silencers have the ability to initiate silencingon their own, whereas HMR-I only plays an auxiliary role insilencing at the HMR locus (5, 22, 30, 32). There is evidencethat HMR-E is stronger than HML-E, which is stronger than HML-I,in silencing (34). This likely reflects the structural differencesamong the silencers. For instance, the binding sites for Abf1p,Rap1p, or ORC in the silencers are actually distinct variantsof a consensus sequence (see the legend to Fig. 1A) that mayhave different affinities for the corresponding protein (4,7). Moreover, the arrangement of the factor-binding sites ineach silencer concerning the distance and intervening sequencebetween any pair of adjacent sites is unique. These factorsmay collectively determine the efficiency of a specific silencerin recruiting Sir proteins, thereby determining its strengthof silencing.

The model for silencer function discussed above implies thatinitiation of silencing is bidirectional. This is consistentwith an early report that HMR-E could function in either orientationto silence the a1 gene at the HMR locus (5). However, equalsilencing on the two sides of HMR-E was observed in the presenceof HMR-I that could also contribute to silencing at HMR (5,30). In the absence of HMR-I, silencing by HMR-E was detectablystronger on its Abf1p side than the ORC side (5). Moreover,HMR-E ectopically inserted at the MAT locus or the structurallysimilar HML-I in its native location also preferentially silencesgenes on its Abf1p side (2, 34). Therefore, the HMR-E or HML-Isilencer per se seems more active in promoting silencing onits Abf1p side. However, such a directionality may be maskedby other silencing elements in the genomic context (5, 43).Moreover, the observed directionality in a specific experimentis also dependent on the sensitivity/resolution of the assayused (43).

In this report, we have further examined the directionalityof silencing by HMR-E both in its native HMR location and inthe context of HML, using three different reporter genes, URA3,a1, and ${alpha}$ 1, respectively. Our results provide further supportfor the notion that for both HMR-E and HML-I silencing is generallyrobust on the Abf1p side but diminished on the ORC side. Importantly,we then show that this is correlated with asymmetric positioningof nucleosomes on the two sides of the silencer. Specifically,a nucleosome is stably positioned close to the Abf1p site butnot the ORC site of the silencer. Elimination of the asymmetryin nucleosome positioning at a silencer leads to comparablesilencing on both sides. Asymmetric positioning of nucleosomesin the immediate vicinity of a silencer is not dependent onits orientation or genomic context and is therefore the inherentproperty of the silencer. It is also independent of Sir proteinsand therefore precedes the formation of silent chromatin. Wealso present evidence that ORC and Abf1p play key roles in asymmetricchromatin organization and directional silencing by HMR-E.

MATERIALS AND METHODS

Top

Materials and Methods

Plasmids and strains. The sequences of "full-length" HML-I and HMR-E shown in Fig.1 to 7 were those defined in the original deletion analysesof HML and HMR (1, 11). Specifically, HML-I and HMR-E correspondto coordinates 14555 to 14839 and 291245 to 291559 of chromosomeIII, respectively. Every silencer contains a core element consistingof binding sites for the silencer-binding proteins and flankingsequences, as illustrated in Fig. 1A and 4. HMR-E' (291395 to291559) and HMR-E'' (291245 to 291490) are shorter versionsof HMR-E deleted for the 3' flanking sequences of the autonomousreplication sequence (ARS) consensus sequence (ACS) and theAbf1p site, respectively (Fig. 6B and 7A).

View larger version (71K):
[in this window]
[in a new window]
FIG. 7. Abf1p is involved in chromatin organization and transcriptional silencing by HMR-E. (A) Effects of inverting or mutating Abf1p sites in HMR-E on silencing. The modified HML in strains 3, 4, and 19 to 24 are shown on the left. The original and modified HMR-E silencers in these strains are highlighted. See the text for descriptions. Growth phenotypes of the strains are shown on the right. (B) Chromatin mapping of strain 23 compared to that of strain 22. MNase digestion and indirect end labeling were carried out as described in the legend to Fig. 4A. (C) Chromatin mapping of strain 24 compared to that of strain 22. The mutated Abf1p site is marked with a cross.

View larger version (80K):
[in this window]
[in a new window]
FIG. 4. Asymmetric positioning of nucleosomes around silencers. DNA isolated from MNase-treated chromatin in each strain was digested with AvaII and then subjected to gel electrophoresis and Southern blotting using a probe abutting the AvaII site (A). The positions of the silencer and URA3 in each strain are shown. Filled ellipse, inferred stably positioned nucleosome. Dashed ellipse, lack of stably positioned nucleosome. N1 to N4, MNase-treated naked genomic DNA from strains 1 to 4, respectively. See the text for detailed descriptions.

View larger version (84K):
[in this window]
[in a new window]
FIG. 6. Role of ORC in asymmetric nucleosome positioning and directional silencing by HMR-E. (A) Configurations of chromatin at silencers and the replication origin ARS1. Illustrated is a summary of data presented in Fig. 4 and 5 as well as that presented by Lipford and Bell (20). Abf1p and Rap1p sites are drawn as stippled and filled arrows, and ACS is drawn as an open arrow. B1 and B2 elements in ARS1 are indicated. Filled ellipse, stably positioned nucleosome. Dashed ellipse, absence of stably positioned nucleosome. (B) Inverting or deleting ACS in HMR-E alters its directionality and strength in silencing. The modified HML in strains 3, 4, and 13 to 18 are shown on the left. The original and modified HMR-E silencers in these strains are highlighted. See the text for more description. Growth phenotypes of the strains are shown on the right. Deletion of ACS is marked with a cross. (C) Inverting ACS in HMR-E' leads to chromatin reorganization. Shown are results of chromatin mapping in strains 13, 13s, 14, 14s, 16, 16s, 17, and 17s. MNase digestion and indirect end labeling was done as described in the legend to Fig. 4A. (D) Direct comparison of chromatin at HMR-E'm in strains 14 and 14s to that at HMR-E' in strains 16 and 16s. (E and F) Effect of ACS deletion on chromatin structure at HMR-E. MNase digestion and indirect end labeling in strains 13, 15, 16, and 18 were carried out as described in the legend to Fig. 4A.

Plasmid pUC26 was made by inserting the BamHI-HML-BamHI fragment(coordinates 9666 to 16263 of chromosome III) into pUC19. A1.1-kb URA3 sequence (2) was inserted at the EcoRV site (15411)of pUC26 to make pYZ1. The HpaI-HML-I-HindIII sequence (coordinates14555 to 14839) in pYZ1 was replaced by a HindIII site to makepYZ1m. HML-I was inserted at the HindIII site of pYZ1m in theopposite direction (compared to that in pYZ1), resulting inpYZ2. The HMR-E sequence in either direction was inserted atthe HindIII site of pYZ1m, resulting in pYZ3 and pYZ4. HMR-E'(Fig. 6B) was inserted in either orientation at the HindIIIsite of pYZ1m, resulting in pYZ13 and pYZ16. HMR-E'm (Fig. 6B)was generated by PCR and inserted in either orientation at theHindIII site of pYZ1m, resulting in pYZ14 and pYZ17. ACS inpYZ13/16 was replaced by the SpeI site, resulting in pYZ15/18.HMR-E" (Fig. 7A) was inserted in either orientation at the HindIIIsite of pYZ1m, resulting in pYZ19 and pYZ22. HMR-E"a (Fig. 7A)was inserted in either orientation at the HindIII site of pYZ1m,resulting in pYZ20 and pYZ23. HMR-E"b (Fig. 7A) was insertedin either orientation at the HindIII site of pYZ1m, resultingin pYZ21 and pYZ24. Plasmid pRS416-HMR was made by insertingHindIII-HMR-HindIII(289227-294210) into pRS416. pYZ5 was madeby inserting URA3 at the SpeI site (291110) of pRS416-HMR. HMR-Ein pYZ5 was replaced by a BamHI site, resulting in pYZ5m. Aninverted HMR-E was inserted at the BamHI site of pYZ5m, makingpYZ6. URA3 was inserted at the XbaI site (291731) of pRS416-HMR,making plasmid pYZ7m. The HMR-I silencer was deleted from pYZ7m,resulting in pYZ7. HMR-E in pYZ7 was replaced by a BamHI site,resulting in pYZ8m. An inverted HMR-E was inserted at the BamHIsite of pYZ8m, making pYZ8. Plasmids pYZ9 and pYZ10 were derivedfrom pYZ5 and pYZ6, respectively, by deleting HMR-I. PlasmidspYZ11 and pYZ12 were derived from pYZ3 and pYZ4, respectively,by deleting HML-E. pMB21 is an integration plasmid containingSIR3-TRP1 and SUP4-o (2).

The names and relevant genotypes of yeast strains are listedin Table 1. Strains 1s to 4s and 11s to 24s were made by transformingY729 to Ura⁺ with BamHI-digested pYZ1 to pYZ4 and pYZ11 to pYZ24,respectively. Strains 5s to 8s were made by transforming Y729to Ura⁺ with HindIII-digested pYZ5 to pYZ8, respectively. Thesestrains were rendered SIR3⁺ by integrating pMB21 at TRP1 inthe genome, resulting in strains 1 to 8 and 11 to 24. Strains9 and 10 were made by transforming W303 ${alpha}$ to Ura⁺ with HindIII-digestedpYZ9 and pYZ10, respectively. SIR3 in strains 9 and 10 was replacedby kanMX, resulting in strains 9s and 10s, respectively. SIR1,SIR2, and SIR4 in strain 4 were replaced by kanMX, resultingin strains 4sir1, 4sir2, and 4sir4, respectively. Strains 3'and 4' were made by transforming strains 3 and 4 to geneticinresistant with a PCR-produced fragment encoding 9-myc linkedto kanMX embedded in a sequence spanning the 3' region of theSIR3 open reading frame. The relevant genotypes of all strainswere confirmed by Southern blotting and/or PCR.

View this table:
[in this window]
[in a new window]
TABLE 1. Yeast strains

Quantitative reverse transcription-PCR (qRT-PCR). Total RNA was isolated from cells grown to an optical densityof about 0.6 and diluted to a concentration of 1 µg/ml.The RNA sample was treated with DNase I at room temperaturefor 40 min. DNase I was then heat inactivated at 70°C for10 min. The sample was then used as template for RT-PCR withthe SuperScript III One-Step RT-PCR System with Platinum TaqDNA Polymerase (Invitrogen) in a 25-µl reaction. Fivefoldserial dilutions used in PCR were the following. For measuringACT1 expression, 0.05 and 0.25 µl of RNA were used ina 25-cycle PCR. For URA3, HMRa1, and HML ${alpha}$ 1, 0.5 and 2.5 µlwere used in a 32-cycle PCR. PCR products were fractionatedon 2.0% agarose gels. The intensity of each band was determinedusing NIH Image. No-RT control PCR was performed by using PlatinumTaq DNA polymerase instead of the RT/Platinum Taq Mix.

ChIP. Chromatin immunoprecipitation (ChIP) was carried out as previouslydescribed (10). The sequences of PCR primers used in ChIP areavailable upon request.

Chromatin mapping by MNase digestion and indirect end labeling. Chromatin analysis was carried out as described previously (3).About 2 x 10⁸ permeabilized spheroplasts were treated with MNaseat 75, 150, and 300 U/ml, respectively, at 37°C for 5 min.Genomic (naked) DNA from untreated cells was digested with MNaseat 7.5 U/ml.

RESULTS

Top

Results

Directional silencing by S. cerevisiae silencers. To further examine the directionality of silencing by HMR-E,we tested its function in the context of HML-I. HMR-E in eitherorientation was used to replace HML-I and tested for its abilityto silence a URA3 gene inserted nearby (Fig. 1B, strains 3 and4). The URA3 product can convert 5-fluoroorotic acid (FOA) intoa toxic metabolite, thus, its expression renders cells sensitiveto FOA. Therefore, URA3 silencing could be measured by FOA resistanceof the cell. URA3 inserted to the right (centromere-proximal)side of HML-I was not silenced (Fig. 1B, lack of growth of strain1 on FOA medium). However, it was silenced when HML-I was inverted(Fig. 1B, strain 2), confirming that HML-I acts in a directionalfashion (2). Strain 3 was sensitive, whereas strain 4 was resistant,to FOA (Fig. 1B, strains 3 and 4 on FOA). Therefore, like HML-I,HMR-E in the context of HML-I also preferentially initiatessilencing on its Abf1p side.

We also reexamined the directionality of HMR-E in its nativelocation using the URA3 gene as the reporter for silencing.URA3 was not silenced when inserted to the left (centromereproximal) of the resident HMR-E silencer (Fig. 1B, strain 5)but was silenced when the direction of HMR-E was inverted (Fig.1B, compare strains 6 and 5). On the other hand, URA3 was silencedwhen inserted to the right of the resident HMR-E silencer inthe absence of HMR-I but was not silenced when HMR-E was inverted(Fig. 1B, compare strains 7 and 8). These results demonstratethat HMR-E also preferentially initiates silencing on its Abf1pside in its endogenous context when URA3 is used as the reporter.Note that silencing of URA3 expression in strains 1 to 8 andother strains used in this work was SIR dependent, as it wascompletely abolished by deleting SIR3 (Fig. 1B, compare strains1s to 8s with 1 to 8, respectively, on FOA; data not shown).

To directly compare the directionality of HMR-E in silencingURA3 and that in silencing the native HMRa genes, we made strains9 and 10, in which HMR-E was in opposite directions (Fig. 2A).These MAT ${alpha}$ strains both have URA3 inserted to the left of HMRand have HMR-I deleted (Fig. 2A) and would therefore allow simultaneousmeasurement of silencing of URA3 and HMRa by HMR-E alone. Theabundance of the transcripts of URA3 and HMRa as well as thecontrol ACT1 were measured by quantitative RT-PCR in strains9 and 10, as well as their sir3^– derivatives 9s and 10s.Three independent experiments were performed, and a representativegel picture is shown in Fig. 2B. Data from all three experimentswere quantified, and the mean of the relative abundance of URA3or a1 in each strain (together with standard deviations) isgraphed in Fig. 2C. The abundance of URA3 in the SIR3⁺ strain10 was only $~$ 1/5 of that in the sir3^– strain 10s, demonstratingthat HMR-E silences URA3 inserted on its Abf1p side. On theother hand, the level of URA3 in strain 9 was not significantlydifferent from that in strain 9s, demonstrating that HMR-E failsto silence URA3 inserted on its ORC side.

View larger version (46K):
[in this window]
[in a new window]
FIG. 2. Analyses of the directionality of HMR-E silencing using qRT-PCR. (A) Modified HMR loci in strains 9 and 10. [ ${Delta}$ ] indicates the deletion of HMR-I. (B) Measurement of the abundance of URA3 and HMRa1 transcripts in strains 9, 10, 9s, and 10s by qRT-PCR. Three independent experiments were performed, and the products were analyzed by agarose gel electrophoresis. Shown are results from one of these experiments. In each lane, the template concentration for URA3 and HMRa1 was the same, but that for ACT1 was 10-fold lower (see Materials and Methods). The template concentration in the even-numbered lanes was one-fifth of that in the odd-numbered ones. The RNA template for the no-RT control was of strain 9s. (C) Quantification and analysis of qRT-PCR data. Each open or shaded bar represents the ratio of intensity of URA3 or HMRa1 product over that of the ACT1 product in a strain. The means of data from all three experiments together with standard deviations are presented. (D) Modified HML loci in strains 11 and 12. [ ${Delta}$ ] indicates the deletion of HML-E. (E) Measurement of the abundance of URA3 and HML ${alpha}$ 1 transcripts in strains 11, 12, 11s, and 12s by qRT-PCR. In each lane, the template concentration for URA3 and HML ${alpha}$ 1 was the same but was 10-fold lower for ACT1. The RNA template for the no-RT control was of strain 11s. (F) Quantification of qRT-PCR data. The ratio of intensity of URA3 or HML ${alpha}$ 1 product over that of the ACT1 product in each strain is listed in the table.

The HMRa1 message was barely detectable in strain 9 but wasabundant in strain 9s (Fig. 2C, compare shaded bars 9 and 9s),demonstrating that HMRa is fully silenced by HMR-E. In strain10, HMRa1 was detectable, and its abundance was $~$ 1/3 of thatin the sir3^– strain 10s (Fig. 2C, compare shaded bars10 and 10s). These data indicate that silencing of HMRa is partialin strain 10 compared to that in strain 9 (Fig. 2C, compareshaded bars 10 and 9), and therefore HMR-E preferentially silencesHMRa on its Abf1p side. This is consistent with Brand et al.'searly S1 mapping result on a1 expression, showing that althoughHMR-E in either orientation could silence the a1 gene, the efficiencyof silencing is nevertheless orientation dependent (a1 transcriptwas not detected when HMR-E was orientated toward a1 but waspresent, albeit with low abundance, when HMR-E was in the oppositedirection) (5).

We also examined whether HMR-E ectopically placed at HML silencedHML ${alpha}$ in an orientation-dependent manner. The MATa strains 11and 12 were deleted for HML-E, and they had HML-I replaced byHMR-E in opposite directions and URA3 inserted to the rightof HML (Fig. 2D). We examined the expression of URA3 and HML ${alpha}$ 1in strains 11 and 12 as well as their sir3^– derivatives,11s and 12s, simultaneously by qRT-PCR. Note that when HML ${alpha}$ isderepressed in a MATa strain (e.g., by disrupting the SIR genes),the a1 and ${alpha}$ 2 proteins form a complex that negatively regulatesthe expression of ${alpha}$ 1 and ${alpha}$ 2 (16, 26, 35). However, ${alpha}$ 1 mRNA is detectablein such a strain, especially when methods with high sensitivitylike qRT-PCR are used (9, 23, 40). Matecic et al. has demonstratedthat the transcriptional state of HML could be examined quantitativelyby measuring the abundance of HML ${alpha}$ 1 message (23). As shown inFig. 2E and F, URA3 expression in strain 12 was about one-fifthof that in 12s and was therefore silenced by HMR-E in strain12. On the other hand, URA3 was not silenced by HMR-E in strain11 (Fig. 2E and F, compare 11 and 11s). HML ${alpha}$ 1 in strain 12 wasabout one-fifth of that in strain 12s, whereas it was not detectablein strain 11 (Fig. 2E and F, compare 12 to 12s and 11 to 11s).Therefore, while HML ${alpha}$ was completely silenced by HMR-E in strain11, it was partially silenced in strain 12. These results supportthe notion that silencing by HMR-E is stronger on its Abf1pside and is therefore directional.

Directional establishment of transcriptionally silent chromatin by the HMR-E silencer. To test whether directional silencing by a silencer is correlatedwith directional formation of silent chromatin, we used chromatinimmunoprecipitation (ChIP) to examine the association of Sir3pwith HMR-E in either orientation as well as adjacent sequences.We tagged Sir3p with the myc epitope in strains 3 and 4 (Fig.1B), resulting in 3' and 4', respectively (Fig. 3A, left). Thishad no detectable effect on the silencing of URA3 or the directionalityof HMR-E (Fig. 1B and 3A, compare 3' and 4' to 3 and 4, respectively).ChIP was performed with a myc antibody ( ${alpha}$ -myc). The abundanceof sequences a (corresponding specifically to HMR-E in the contextof HML-I), b to d (Fig. 3A, left), a control sequence spanningthe HMR-I silencer, and another control from the ACT1 locusin the immunoprecipitated chromatin fragments were measuredby PCR. All the sequences measured are between 250 to 350 bpin length. Three independent ChIP experiments were performed,and a representative gel picture was presented (Fig. 3B). Theintensity of each band was quantified and normalized againstinput control. The relative abundance of Sir3p-myc associatedwith a sequence in a strain was estimated as the ratio of theintensity of the corresponding fragment over that of the HMR-Icontrol (taken as 1.0) in the same strain. The mean of datafrom all three experiments (together with the standard deviations)is graphed in Fig. 3C. For both strains 3' and 4', there waslittle or no Sir3p-myc associated with the ACT1 sequence (Fig.3C, ACT1), which was consistent with the fact that ACT1 is normallynot subject to transcriptional silencing and not associatedwith Sir proteins. On the other hand, the abundance of Sir3p-mycat sequence a was high in both strains 3' and 4' (Fig. 3C, a),indicating that HMR-E in either orientation is able to recruitSir3p-myc. In strain 4', the Sir3p-myc abundance at b to d wasalso high (Fig. 3C, b to d, shaded bars), which was indicativeof Sir3p-myc spreading from HMR-E toward URA3. On the otherhand, Sir3p-myc abundance at b to d was significantly less abundantin 3' than in 4' (Fig. 3C, compare strain 3' and 4' at b tod), indicating that there was reduced spreading of Sir3p-mycfrom HMR-E, which was in accord with diminished silencing ofURA3 in strain 3'. These results demonstrate that HMR-E initiatesthe spread of Sir proteins preferentially on its Abf1p side,which is consistent with previous data obtained by Loo and Rineindicating that Sir-dependent silent chromatin extends moreefficiently on the Abf1p side than the ACS side of HMR-E (21).

View larger version (35K):
[in this window]
[in a new window]
FIG. 3. Directional spread of Sir3p promoted by HMR-E. (A) Myc-tagged Sir3p is functional in silencing. Strains 3' and 4' were identical to 3 and 4, respectively, except that nine tandem copies of a sequence encoding the myc tag were fused in frame to the 3' end of the SIR3 open reading frame. Sequences a to d, tested by PCR in ChIP, are indicated (roughly proportional). Growth phenotypes of strains 3' and 4' are shown. (B) Analysis of Sir3p-myc abundance near HMR-E by ChIP. The abundance of a to d as well as control sequences from HMR-I and ACT1 was measured by PCR before (input) and after ( ${alpha}$ -myc) chromatin IP. No Ab, samples from mock ChIP without using antibody. Data from three independent ChIP experiments were quantified and plotted in panel C. See the text for more descriptions.

Unidirectional silencing by HML-I or HMR-E coincides with an asymmetric distribution of nucleosomes. The directional nature of HML-I or HMR-E function might resultfrom the fact that it consists of an asymmetric DNA sequence,and perhaps more importantly that its factor-binding sites arearranged in a directional fashion: ORC-Rap1p-Abf1p (Fig. 1A).How might the sequence/structural asymmetry of a silencer betranslated into functional asymmetry? Propagation of Sir complexesalong the chromatin depends on their association with nucleosomesand presumably requires a continuous array of closely positionednucleosomes (3, 32). Given that Abf1p, Rap1p, and ORC all havethe ability to modulate local chromatin structure upon bindingto DNA (3, 20, 41, 42), we envisioned that nucleosomes mightbe organized differently (asymmetrically) on the flanks of HML-Iand HMR-E, resulting in unequal potentials for the spread ofSir proteins on the two sides. To test this hypothesis, we mappednucleosomes around HML-I and HMR-E using micrococcal nuclease(MNase) digestion and indirect end labeling (33). MNase introducesdouble-stranded breaks mainly in chromosomal DNA not assembledinto nucleosomes (i.e., linker DNA), and therefore the presenceof a nucleosome could be inferred from an MNase-resistant regionof about 147 bp (the length of DNA in a nucleosome) borderedon both sides by MNase-sensitive sites (33). To determine theputative positions of nucleosomes, the chromatin is subjectedto partial digestion with MNase, yielding a series of fragmentsassociated with various numbers of nucleosomes. DNAs from thesefragments are then examined by indirect end labeling to revealthe positions of MNase cleavage sites relative to a referencerestriction site, thereby allowing a tentative assignment ofsequences protected by nucleosomes in the region of interest(33).

We first mapped chromatin around HML-I in its original orientationin strain 1. DNA purified from MNase-treated chromatin was digestedwith AvaII 300 bp downstream from HML-I and was then subjectedto electrophoresis and Southern blotting using a probe correspondingto a 200-bp sequence abutting the AvaII site (Fig. 4A, bar labeled"probe"). We found that the HML-I core (consisting of the bindingsites for Abf1p, Rap1p, and ORC) coincided with three MNase-sensitivesites that were likely caused by the binding of Abf1p, Rap1p,and ORC (Fig. 4A, lanes 1 to 3). Notably, the patterns of MNasecleavage on the two sides of the core sequence were different.Specifically, the region adjacent to ACS (ORC recognition site)contained an MNase-hypersensitive region (labeled by a dot)and two moderately sensitive sites that were indicative of theabsence of a stably positioned nucleosome (Fig. 4A, comparelanes 1 to 3 with N1; the dashed ellipse indicates the lackof a stably positioned nucleosome). In contrast, on the Abf1pside there were two MNase-sensitive sites bracketing a regionof $~$ 150 bp that was relatively resistant to MNase cleavage, apattern indicative of the presence of a positioned nucleosome(Fig. 4A, lanes 1 to 3, filled ellipse labeled 1'). Therefore,the organization of nucleosomes around HML-I in strain 1 wasasymmetric, and the absence of a stably positioned nucleosomenear ACS coincided with the lack of URA3 silencing on this sideof HML-I, as shown in Fig. 1B. To examine if the Sir complexwas involved in the asymmetric positioning of nucleosomes atHML-I, we disrupted SIR3 in strain 1, making 1s. The MNase cleavagepattern around HML-I in strain 1s was similar to that in 1 (Fig.4A, compare lanes 4 to 6 with 1 to 3), indicating that an intactSir complex is not required for asymmetric nucleosome positioningat HML-I.

We next mapped chromatin around the inverted HML-I in strain2. The pattern of MNase digestion within the inverted HML-Isequence mirrored that of HML-I in its original direction, indicatingthat the structure of HML-I was not affected by its inversion(Fig. 4A, compare lanes 10 to 12 with 1 to 3). Strikingly, thepattern of nucleosome positioning on either side of the invertedHML-I in strain 2 was dramatically different from that of HML-Iin strain 1 (Fig. 4A, compare lanes 10 to 12 with 1 to 3). Onthe Abf1p side, three consecutive nucleosomes could be inferred(Fig. 4A, right, filled ellipses labeled 1 to 3), which is consistentwith data of chromatin mapping on the Abf1p side of native HML-Iobtained by Weiss and Simpson (39). On the other hand, no stablenucleosomes were located immediately adjacent to ACS (Fig. 4A,right, the dashed ellipse). Therefore, flipping HML-I led toan apparent inversion of the pattern of nucleosome positioningaround the silencer core (Fig. 4A, compare lanes 10 to 12 with1 to 3; note the apparent twofold rotational symmetry betweensites a and a', between the two dashed ellipses, and betweenfilled ellipses 1 and 1'). The presence of nucleosomes 1 to3 on the Abf1p side of HML-I coincided with robust URA3 silencingon this side (Fig. 1B and 4A, strain 2). To examine if the Sircomplex played any role in forming the asymmetric structureof chromatin around the inverted HML-I, we deleted SIR3 in strain2, making 2s. In strain 2s, site b marking the border of nucleosome3 disappeared, and certain sites within the region spanningnucleosomes 2 and 3 became sensitive to MNase (Fig. 4A, comparelanes 7 to 9 with 10 to 12). Therefore, it could be inferredthat nucleosomes 2 and 3 were no longer stably positioned therein strain 2s. This suggested that the positioning of nucleosomes2 and 3 was the result of the spread of the Sir complex in strain2. On the other hand, nucleosome 1 remained positioned therein 2s (Fig. 4A, compare lanes 7 to 9 to 10 to 12; note thatband a, marking the border of the putative nucleosome 1, wasretained). Similar results regarding the effect of SIR3 deletionon nucleosomes near the Abf1p side of HML-I was obtained byWeiss and Simpson (39). In summary, we conclude that asymmetricpositioning of nucleosomes on the Abf1p and ORC sides of HML-Iis independent of the direction of the silencer and of intactSir complexes.

HMR-E is similar to HML-I in that it also contains binding sitesfor Abf1p, Rap1p, and ORC in the same order (Fig. 1A). We examinedif HMR-E organized chromatin in a similar way as HML-I by mappingchromatin around HMR-E in strains 3 and 4 as well as their sir3^–derivatives, 3s and 4s. The distribution of MNase-sensitivesites around HMR-E in strain 3 was reminiscent of that aroundHML-I in strain 1 (compare lanes 1 to 3 in Fig. 4B with 1 to3 in A). A stable nucleosome was inferred to be positioned nearthe Abf1p site but not the ORC site of HMR-E (Fig. 4B, left,filled ellipse 1' and dashed ellipse). The pattern of nucleosomepositioning was largely unchanged by the deletion of SIR3 (Fig.4B, compare lanes 4 to 6 with 1 to 3), which was similar tothe situation regarding HML-I (Fig. 4A, lanes 1 to 6). Chromatinconfiguration around inverted HMR-E in strain 4 was again verysimilar to that around inverted HML-I in strain 2 (compare Fig.4A to B). Three consecutive nucleosomes (labeled 1 to 3) werepositioned near the Abf1p sites (Fig. 4B, lanes 10 to 12), apattern similar to chromatin mapping data of native HMR-E obtainedby Ravindra et al. (28). There were no stably positioned nucleosomesadjacent to ACS (lanes 10 to 12 in Fig. 4B). SIR3 deletion hadno significant effect on the positioning of nucleosome 1 (Fig.4B, compare lanes 7 to 9 with 10 to 12; note the presence ofsite a, inferred as the border of nucleosome 1 in lanes 7 to9). On the other hand, site b, inferred as the border of nucleosome3, became blurred, indicating that nucleosomes 3 and 2 wereno longer as stably positioned in strain 4s as in 4 (Fig. 4B,lanes 7 to 9), which was again similar to the situation concerningHML-I (Fig. 4A, lanes 7 to 9). Nucleosomes 2 and 3 were likelypositioned/stabilized by the Sir complex during the formationof silent chromatin on the Abf1p side of HMR-E in strain 4.In summary, these results demonstrate that, similar to HML-I,HMR-E also asymmetrically positions nucleosomes in an orientation-and Sir complex-independent fashion, and a stably positionednucleosome coincides with the presence of Sir3p and robust silencingon the Abf1p side (Fig. 1B and 3C, strains 4 and 4'), whereasthe lack of a positioned nucleosome adjacent to ACS coincideswith the absence of silencing on the ACS side (Fig. 1B and 3C,strains 3 and 3').

Nucleosome positioning in the immediate vicinity of a silencer is independent of the Sir proteins. We have shown above that asymmetric nucleosome positioning inthe immediate vicinity of a silencer was not significantly affectedby the deletion of SIR3 (Fig. 4A and B). However, the Sir2p/Sir4pcomplex can be recruited to silencers in a Sir3p-independentmanner (15). This may be mediated by Rap1p-Sir4p interactionand/or Sir1p-Sir4p interaction (32). To test whether Sir2p/Sir4pas well as Sir1p plays a role in nucleosome positioning at silencers,we deleted SIR1, SIR2, and SIR4, respectively, from strain 4,resulting in strains 4sir1, 4sir2, and 4sir4 (Fig. 4C). We mappednucleosomes near HMR-E in strains 4 (SIR⁺), 4sir1 (sir1^–),4sir2 (sir2^–), 4s (sir3^–), and 4sir4 (sir4^–).The profile of MNase-sensitive sites in strain 4s indicatedthat deletion of SIR3 had no effect on the positioning of nucleosome1 near the Abf1p site and the exclusion of a nucleosome adjacentto the ORC site of HMR-E (Fig. 4C, compare lanes 10 to 12 to1 to 3; see also lanes 7 to 9 in B). The profiles of MNase-sensitivesites in strains 4sir2 (sir2^–) and 4sir4 (sir4^–)were both very similar to that in strain 4s (sir3^–) (Fig.4C, compare lanes 7 to 9 and 13 to 15 to lanes 10 to 12), whichsuggested that like Sir3p, Sir2p and Sir4p also did not playa significant role in nucleosome positioning on either sideof HMR-E. The pattern of MNase-sensitive sites in strain 4sir1(sir1^–) was nearly identical to that in strain 4 (SIR⁺)(Fig. 4C, compare 4 to 6 to lanes 1 to 3), demonstrating thatSir1p was not involved in nucleosome positioning at HMR-E. Insummary, the above data demonstrate that asymmetric nucleosomepositioning in the immediate vicinity of HMR-E is generallyindependent of Sir proteins.

Asymmetric nucleosome positioning by a silencer is not dependent on its genomic context. In the above experiments on chromatin organization around thesilencers, HMR-E was placed in the ectopic position of HML-I(Fig. 4B). While this enabled us to directly compare the propertiesof HML-I and HMR-E in the same genomic context, it also raisedthe question of whether the observed configuration of chromatinaround HMR-E was the same as that in its native location. Moreover,URA3 inserted near HMR-E or HML-I (Fig. 4) might contributeto or alter the chromatin structure at the silencers. To addressthese issues, we mapped chromatin around HML-I and HMR-E intheir unmodified native locations in strain Y728 (SIR⁺) andits sir3^– derivative, Y729 (the parent of all the strainsused in this work). As shown in Fig. 5, the pattern of nucleosomepositioning around HML-I or HMR-E in its unmodified genomiccontext in strain Y728 (SIR⁺) resembled that around its counterpartin strain 1 or 3, respectively (compare lanes 1 to 3 in Fig.5 to 1 to 3 in Fig. 4A, and 10 to 12 as well as 16 to 18 inFig. 5 to 1 to 3 in Fig. 4B; note the filled and dashed ellipses).SIR3 deletion had little or no effect on the arrangement ofnucleosomes in the vicinity of the native HML-I and HMR-E silencers(Fig. 5, compare lanes 4 to 6 to lanes 1 to 3, 10 to 12 to lanes7 to 9, and 16 to 18 with 13 to 15, respectively). These resultssuggest that the ability of HML-I or HMR-E to asymmetricallyorganize chromatin in its immediate vicinity is an intrinsicproperty that does not depend on its genomic context.

View larger version (52K):
[in this window]
[in a new window]
FIG. 5. Chromatin organization around the native HML-I and HMR-E silencers. DNA from chromatin of strain Y728 (SIR3⁺) or Y729 (sir3^–) treated with MNase was digested with AvaII (lanes 1 to 6), XbaI (7 to 12), and BsrGI (13 to 18), respectively. Relevant DNA fragments were revealed by probes a (lanes 1 to 6), b (7 to 12), and c (13 to 18), which are 200 bp in length, and they are indicated as black bars near the AvaII, XbaI, and BsrGI sites, respectively, in the diagrams of HML and HMR (top). The structures of the silencers are shown on the left of the three panels. N, naked DNA control. Other symbols are as described in the legend to Fig. 4.

ORC plays a critical role in asymmetric chromatin organization and unidirectional silencing by the HMR-E silencer. The above results demonstrated that a stably positioned nucleosomeon the ACS side in HMR-E or HML-I was lacking, which was correlatedwith a deficiency in silencing on this side of the silencer(summarized in Fig. 6A). It was possible that ORC bound to thesilencer prevented nucleosome positioning near the 3' end ofACS (the direction of ACS is taken as 5'-WTTTAYRTTTW-3'). ORCis also involved in the initiation of DNA replication at ARSs(autonomous replication sequences) that are composed of theconserved ACS and Abf1p site as well as the poorly conservedB1 and B2 elements (Fig. 6A, ARS1). Interestingly, a uniquechromatin configuration is maintained around ARS1, with twonucleosomes stably positioned near both the Abf1p-binding siteand the ACS sequence, respectively (Fig. 6A, ARS1) (20). Thisconfiguration is important for ARS function, and Abf1p and ORCare responsible for its creation/maintenance (20, 37). Takentogether, the features of chromatin organization at silencersand ARS1 (Fig. 6A) suggest that ORC allows the positioning ofa nucleosome near the 5' end, but hinders nucleosome positioningnear the 3' end, of ACS. If this were the case, then invertingACS in HMR-E should lead to the positioning of a nucleosomenear the ACS side of the silencer, which is predicted to allowURA3 silencing to occur on this side.

To test this idea, we set out to precisely invert ACS in HMR-Ein strains 3 and 4 (Fig. 1B). For the convenience of cloning,we employed a shortened version of HMR-E, named HMR-E', thatlacked the flanking sequence of ACS in the HMR-E silencer (Fig.6B, left, HMR-E' versus HMR-E). We first made strains 13 and16 that were identical to 3 and 4, respectively, except havingHMR-E' in place of HMR-E in opposite directions (Fig. 6B, left).URA3 in strain 13, like that in 3, was not silenced by HMR-E'(Fig. 6B, compare 13 and 3), while silencing in strain 16 wascomparable to that in 4 (Fig. 6B, compare 16 to 4). Therefore,deleting the flanking sequence of ACS from HMR-E had no detectableeffect on its strength and directionality of silencing. Moreover,although the deletion caused some subtle changes in chromatin,it did not alter the overall pattern of asymmetric nucleosomepositioning around the silencer (Fig. 6C and 4B, compare strains3 and 3s to 13 and 13s, as well as 4 and 4s to 16 and 16S, respectively).For example, the MNase-hypersensitive site indicated by a dotin strain 13, 13s, 16, or 16s (Fig. 6C) was closer to ACS inHMR-E' compared to that in 3, 3s, 4, or 4s (Fig. 4B), but thisdid not affect the asymmetric nature of nucleosome positioning(Fig. 6C, 13, 13s, 16, and 16s). Therefore, HMR-E' was to alarge extent structurally and functionally equivalent to the"full length" HMR-E.

We then precisely inverted ACS in HMR-E' in strain 13, resultingin HMR-E'm in strain 14 (Fig. 6B, left). Remarkably, we nowobserved relatively robust growth of strain 14 on FOA (Fig.6B), indicating that HMR-E'm, in contrast to HMR-E' in strain13, was able to significantly silence URA3. Therefore, the directionof ACS in HMR-E' is important for silencing. We demonstratedthat nucleosome positioning around HMR-E'm in strain 14 wasmarkedly different from that around HMR-E' in 13 (Fig. 6C, compare14 and 13). Notably, there was a sharp decrease in the intensityof the band denoted by a dot in lanes 7 and 8 relative to thatin lanes 1 and 2 in Fig. 6C. Moreover, sites a and b, inferredas the boundaries of nucleosomes 2 and 3 in strain 16, werealso present in strain 14, albeit with reduced intensity (Fig.6C and D, compare lanes 7 and 8 to 15 and 16). As a consequence,the distribution of MNase-sensitive sites on the ACS side ofHMR-E'm in strain 14 started to resemble that on the Abf1p sideof HMR-E' in strain 16 (Fig. 6D, compare 14 and 16). We interpretedthese data as indicative of the presence of nucleosomes 1 to3 near ACS of HMR-E'm in strain 14 (Fig. 6C and D, lanes 7 and8, filled ellipses 1 to 3) that were less stably positionedthan those in strain 16 (Fig. 6C and D, lanes 15 and 16, filledellipses 1 to 3). This might be the reason why URA3 silencingin strain 14 was weaker than that in 16 (Fig. 6B, compare 14and 16), in light of the correlation between nucleosome positioningnear a particular side of a silencer and preferential silencingon this side (see Fig. 6A). We noted that in strain 14s, thesir3^– derivative of 14, the MNase cleavage site denotedby a dot was weaker than that in 13 and 13s but stronger thanthat in 14 (Fig. 6C, compare lanes 5 and 6 to 1 to 4 as wellas 7 and 8). This, together with the fact that site a, inferredas the border of nucleosome 1, was present in strain 14s (Fig.6C, lanes 5 and 6), indicated that nucleosome 1 was positionedin strain 14s, albeit with reduced stability compared to thatof 14. In other words, the Sir complex further stabilized nucleosome1 (and induced the positioning of nucleosomes 2 and 3) in strain14 (Fig. 6C, lanes 7 and 8), which was similar to the Sir-dependentstabilization/positioning of nucleosomes 1 to 3 in 16 (Fig.6C, compare lanes 15 and 16 to 13 and 14). The above resultsdemonstrated that inverting ACS in HMR-E' led to the positioningof a nucleosome near the ACS site (Fig. 6C, nucleosome 1 instrains 14 and 14s), which coincided with the occurrence ofsignificant silencing on the ACS side of HMR-E'm (Fig. 6B, strain14).

We were also interested in whether ACS inversion had any effecton silencing on the Abf1p side of HMR-E'. We precisely invertedACS in HMR-E' in strain 16, resulting in 17 (Fig. 6B, left).Silencing of URA3 in 17 was weaker than that in strain 16 (Fig.6B, compare 17 to 16). Therefore, inverting ACS reduced silencingon the Abf1p side of HMR-E'. We also mapped chromatin aroundHMR-E'm in strain 17 and its sir3^– derivative, 17s. Thedistribution of MNase-sensitive sites on the Abf1p side of HMR-E'min strain 17 or 17s was similar to that in 16 or 16S (Fig. 6C,compare 9 and 10 to 15 and 16, as well as 11 and 12 to 13 and14). This was consistent with the fact that significant silencingremained on the Abf1p side of HMR-E'm in strain 17 (Fig. 6B,17).

On the basis that URA3 was silenced in strain 14 but not 13,one could conclude that ACS in its native orientation apparentlyserves as an obstacle to the initiation of silencing, presumablypromoted by the rest of the HMR-E silencer consisting of theAbf1p and Rap1p-bdining sites. If this were the case, then deletingACS from HMR-E' in strain 13 should lead to URA3 silencing.To test this proposal, we replaced the 11-bp ACS consensus sequencein HMR-E' with the 6-bp restriction site for SpeI in strains13 and 16, resulting in HMR-E'n in strains 15 and 18, respectively(Fig. 6B). HMR-E'n in strain 15 promoted significant silencingof URA3 (Fig. 6B, compare strain 15 to 13). This is consistentwith the notion that ACS in HMR-E' serves as an inhibitor ofthe silencing function of the rest of the silencer consistingof the Abf1p and Rap1p sites (Fig. 6B, lack of silencing instrain 13). Moreover, we showed that deleting ACS allowed anucleosome (labeled 1) to be positioned close to HMR-E'n instrain 15 (Fig. 6E, compare lanes 17 and 18 to 1 and 2; notethe disappearance of a salient MNase site, labeled by a dotin lanes 17 and 18). The positioning of nucleosomes 2 and 3in 15 likely reflected the spreading of Sir proteins (Fig. 6E).ACS deletion in the inverted HMR-E'n in strain 18 led to a moderatedecrease in URA3 silencing (Fig. 6B, compare 18 to 16) but didnot significantly affect nucleosome positioning on the Abf1pside of the silencer (Fig. 6F, compare lanes 19 and 20 to 15and 16). Therefore, ORC at HMR-E' in strain 16 helps silencingon the Abf1p side (Fig. 6B) likely by increasing the local concentrationof Sir proteins, as it can recruit Sir2p/Sir4p via Sir1p.

The above results provided direct evidence for a role of ORCin the establishment of asymmetric nucleosome positioning anddirectional silencing by HMR-E. We think that at HMR-E, ORChelps the initiation of silencing on the Abf1p side by recruitingSir proteins but at the same time hinders silencing on the ORCside by preventing the positioning of a nucleosome. When ACSis inverted or deleted from HMR-E', the resultant silencer (HMR-E'mor HMR-E'n) becomes bidirectional (Fig. 6B, compare 14 to 17and 15 to 18). Therefore, ORC plays a critical role in makingHMR-E function in a unidirectional fashion.

Role of Abf1p in nucleosome positioning and transcriptional silencing. Of the three silencer-binding proteins, Abf1p is the least studiedconcerning its role in the initiation of silencing. A comparisonof the configurations of chromatin in the vicinities of HMR-E,HML-I, and ARS1 revealed that a nucleosome was always positionednear the Abf1p site, which coincided with strong silencing onthis side of the silencers (Fig. 6A). Like ACS, the sequenceof the Abf1p-binding site is also asymmetric (see the legendto Fig. 1A), and its direction might also be important for itsfunction in silencers. We therefore examined the effect of invertingthe Abf1p site in HMR-E on nucleosome positioning and silencing.Again, for the convenience of cloning, we used a shortened versionof HMR-E, HMR-E", that lacked the flanking sequence of the Abf1psite in HMR-E (Fig. 7A, compare HMR-E" to HMR-E). Strains 19and 22 were identical to 3 and 4, respectively, except theybore HMR-E" in place of HMR-E in opposite directions (Fig. 7A,left). URA3 silencing in strains 19 and 22 was comparable tothat in 3 and 4, respectively (Fig. 7A, compare 19 and 22 to3 and 4, respectively). Therefore, HMR-E" was equivalent toHMR-E regarding the potency and directionality of silencing.The pattern of chromatin organization around HMR-E" was alsosimilar to that around HMR-E, despite the fact that there weresome changes in the distribution of MNase sites (Fig. 4B and7B, compare 22 to 4; note the inferred presence of nucleosomes1 to 3 in both strains 4 and 22). We then precisely invertedthe Abf1p site in HMR-E" in strains 19 and 22, resulting in20 and 23, respectively (Fig. 7A, left). URA3 silencing in strains20 and 23 was comparable to that in 19 and 22, respectively(Fig. 7A), demonstrating that inverting the Abf1p site has noeffect on silencing by HMR-E". Consistently, chromatin aroundHMR-E"a in strain 23 was similar to that around HMR-E" in strain22 (Fig. 7C, compare 22 and 23). Therefore, the direction ofthe Abf1p site was not important for silencing by HMR-E". Alongthis line, it is noteworthy that the orientation of the Abf1psite in an ARS is also not important for its function as a replicationorigin (38).

It was possible that the reason why the direction of the Abf1psite was not important for HMR-E" function was that Abf1p playedno role in silencing. To test this hypothesis, we mutated theAbf1p site (ATCATaaaatACG $->$ ATgATaaaattgG; the conserved nucleotidesare capitalized, and mutated nucleotides are italicized) inHMR-E" in strain 22, making 24 (Fig. 7A). Previous studies havedemonstrated that such a mutant site would not bind Abf1p invitro or in vivo (17). URA3 silencing in strain 24 was significantlyreduced compared to that in 22 (Fig. 7A, compare 24 to 22),demonstrating that an intact Abf1p site was necessary for efficientsilencing by HMR-E". Chromatin configuration in the vicinityof HMR-E"b in strain 24 was markedly different from that inthe vicinity of HMR-E" in strain 22 (Fig. 7C, compare 24 and22). For example, sites a and c present in strain 22 were diminishedin strain 24, whereas a' appeared in strain 24 (Fig. 7C). Itcould be inferred that putative nucleosomes 1' to 3' were positionedat or near the mutated Abf1p site in strain 24 (Fig. 7C). However,compared to nucleosomes 1 to 3 in strain 22, 1' to 3' were lessclosely spaced (Fig. 7C, compare 24 and 22; note that the inferrednucleosomes 1' to 3' spanned a longer sequence than 1 to 3).This was correlated with reduced silencing in strain 24 comparedto that in strain 22 (Fig. 7A). Based on these results, we proposethat Abf1p contributes to silencer function by, at least inpart, positioning nucleosomes, which is independent of the orientationof its binding site. This is reminiscent of its role in theinitiation of DNA replication at ARS1 (20).

DISCUSSION

Top

Discussion

The silencers in S. cerevisiae present an excellent model systemfor studying the mechanisms of the establishment of transcriptionallysilent chromatin in eukaryotes (13, 32). In this work we haveinvestigated the directionality of silencer function and howit is determined. For both HML-I and HMR-E, which are structurallysimilar, silencing was found to preferentially occur on theAbf1p sides of the silencers (Fig. 6A). This is in accord withdata from previous studies demonstrating unequal silencing onthe two sides of HML-I or HMR-E (2, 34). A comparison of dataobtained here and in previous studies indicates that the differencein silencing efficiency between the two sides of HMR-E may varydepending on the assay employed (2, 5, 34, 43). This is becausethe apparent strength of silencing on either side of a silencermay be affected by nonsilencer factors, including its genomiclocation (43). Different genomic loci may contain distinct setsof elements that could influence silencer function (e.g., othersilencers and protosilencers) (12, 27, 43). We have recentlyshown that a protosilencer located on the ORC side of HMR-Ecould enhance the efficiency of silencing on the ORC side toa level that is comparable to that on the Abf1p side, therebymasking the directionality of the silencer (43). Along thisline, the presence of HMR-I at HMR likely masked the directionalityof HMR-E in silencing HMRa observed in an early experiment doneby Brand et al. (5). Consistently, silencing of a1 by HMR-Ealone was stronger on the Abf1p side than the ORC side (5) (Fig.2). The efficiency of silencing on either side of a silenceris also dependent on the promoter strength of the reporter gene(36). For example, we found that on the ORC side of HMR-E, URA3was completely depressed, whereas a1 or ${alpha}$ 1 was still silenced,albeit partially (Fig. 2).

Chromatin plays an important role in the regulation of cellulartransactions involving DNA. It has been well established thatcorrect nucleosome positioning at promoters and replicationorigins is crucial for the initiation of transcription and replication,respectively. We have demonstrated in this work that chromatinalso plays a key role in determining the directionality of asilencer. We found that nucleosomes are asymmetrically positionedaround the HMR-E and HML-I silencers. Specifically, a nucleosomeis stably positioned close to the Abf1p site but not the ORCrecognition site of the silencers, which coincides with robustsilencing on the Abf1p side and reduced silencing on the ORCside (Fig. 6A). Importantly, asymmetric nucleosome positioningat a silencer is largely independent of its orientation andgenomic location. It is also independent of the Sir proteinsand therefore happens before the establishment of silent chromatin.We propose that the asymmetry in chromatin structure at a silencerdetermines its directional nature of silencing. Specifically,we envision that the nucleosome positioned close to the Abf1pside of the silencer allows or facilitates the Sir complex tostart spreading on this side, whereas the lack of a stably positionednucleosome adjacent to the ORC site reduces the efficiency ofinitiation of Sir propagation on this side. This hypothesisis based on the fact that nucleosomes are both the substratesfor the histone deacetylase activity of the Sir complex andalso the platform for Sir recruitment, according to the currentview of the nucleosome-by-nucleosome propagation of Sir complexesalong chromatin (3, 13, 32). In support of this contention,we have demonstrated that Sir3p preferentially associates withsequences on the Abf1p side of the HMR-E silencer (Fig. 3).

How is asymmetric nucleosome positioning established at silencers?To answer this question, we turned to known examples of chromatinorganization at gene promoters and replication origins for clues.At promoters, transcriptional activators can induce changesin chromatin structure via their own nucleosome disrupting/positioningactivities and/or by recruiting chromatin remodeling complexes.Of particular interest is chromatin reorganization by Rap1pand Abf1p at the upstream activating sequences of the genesthey regulate (18, 24). Rap1p or Abf1p is believed to facilitatethe access of primary transcriptional activators such as Gcn4pto their binding sites located nearby by disturbing nucleosomesbearing its binding sites, thereby "opening" chromatin (41,42). Note that by preventing nucleosome formation on the sequencecontaining its binding site, Abf1p/Rap1p may help define theboundary of an array of nucleosomes, thereby serving to positionnucleosomes. Chromatin also plays an important role in the initiationof DNA replication. The ARS sequences consisting of ACS, B1,B2, and Abf1p sites are free of nucleosomes, whereas two nucleosomesare stably positioned near the Abf1p and ORC sites of ARS (Fig.6A) (20). This pattern of nucleosome arrangement is accomplishedby combined actions of Abf1p and ORC, and its perturbation affectsthe formation of the prereplication complex and disrupts replicationinitiation (20).

In light of the roles of Abf1p, Rap1p, and ORC in chromatinorganization at gene promoters and replication origins, thesefactors may also be responsible for the establishment of asymmetricnucleosome positioning at silencers. A comparison of chromatinconfigurations around silencers and ARS1 suggests that ORC allowsthe positioning of a nucleosome near the 3' end of ACS but excludesone near the 5' end (Fig. 6A). In accordance with this, we showedthat inverting ACS in HMR-E so that it was orientated towardthe inside of the silencer led to the positioning of a nucleosomenear the ACS side of the silencer and a concomitant significantincrease in silencing on this side. As a result, there appearedto be symmetric positioning of nucleosomes and comparable silencingon the two sides of the modified HMR-E. On the other hand, inversionof ACS in an ARS disrupts ARS function (14), likely becauseit disrupts the nucleosome normally positioned near ACS outsideof ARS (see Fig. 6A) that is important for ARS function. Theseresults suggest that ACS (via ORC) asymmetrically positionsnucleosomes on its sides, which is likely the reason why itsdirection is important for its function. We have also foundthat removing ACS from HMR-E led to symmetric positioning ofnucleosomes and comparable silencing on both sides, which furtherdemonstrated that ORC in its native orientation serves as anobstacle to silencing on the ACS side.

The fact that the Abf1p site in a silencer or ARS is associatedwith a stably positioned nucleosome nearby led us to proposethat Abf1p binding is responsible for nucleosome positioning.Consistent with this hypothesis, we demonstrated that mutatingthe Abf1p-binding site in HMR-E caused clear changes in nucleosomepositioning as well as a significant reduction in silencingon the Abf1p side of the silencer. However, inverting the directionof the Abf1p site in HMR-E caused little or no change in chromatinconfiguration or silencing. Therefore, we think that Abf1p canposition nucleosomes on both sides of its site. We concludethat both ORC and Abf1p play major roles in promoting asymmetricpositioning of nucleosomes on the two sides of HMR-E or HML-I,which determines the directional nature of silencing by thesilencer.

In summary, results presented in this work demonstrate thata yeast silencer (via the silencer-binding proteins) promotesasymmetric positioning of nucleosomes, which is independentof its orientation or genomic location. We think such a structuralasymmetry results in distinct potentials for the initiationof transcriptional silencing on the two sides of the silencer.

ACKNOWLEDGMENTS

We thank Lars Olsen and Hengping Xu for assistance and ScottHolmes for useful discussion.

This work was supported by NIH grant GM62484 to X.B.

FOOTNOTES

* Corresponding author. Mailing address: Department of Biology, University of Rochester, Rochester, NY 14627. Phone: (585) 275-6922. Fax: (585) 275-2070. E-mail: xinbi{at}mail.rochester.edu.

Published ahead of print on 14 August 2006.

These authors contributed equally.

REFERENCES

Top