![[Feedback]](/icons/banner/feedbackACT.gif){kind=icon}
![[For Subscribers]](/icons/banner/subscriberACT.gif){kind=icon}
![[Archive]](/icons/banner/archiveACT.gif){kind=icon}
![[Search]](/icons/banner/searchACT.gif){kind=icon}
![[Contents]](/icons/banner/tocACT.gif){kind=icon}
|
|
|
|
|
||
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Vol. 13, Issue 7, 2207-2222, July 2002
Department of Molecular and Cell Biology, University of California, Berkeley, Berkeley, California 94720
Submitted March 8, 2002; Revised March 29, 2002; Accepted April 12, 2002  |
ABSTRACT |
|---|
 TOP
 ABSTRACT
 INTRODUCTION
MATERIALS AND METHODS
RESULTS
DISCUSSION
REFERENCES
|
|---|
In Saccharomyces cerevisiae, silencing at the HM loci depends on Sir proteins, which are structural components of silenced chromatin. To explore the structure and assembly of silenced chromatin, the associations of Sir proteins with sequences across the HMR locus were examined by chromatin immunoprecipitation. In wild-type cells, Sir2p, Sir3p, and Sir4p were spread throughout and coincident with the silenced region at HMR. Sir1p, in contrast, associated only with the HMR-E silencer, consistent with its role in establishment but not maintenance of silencing. Sir4p was required for the association of other Sir proteins with silencers. In contrast, in the absence of Sir2p or Sir3p, partial assemblies of Sir proteins could form at silencers, where Sir protein assembly began. Spreading across HMR required Sir2p and Sir3p, as well as the deacetylase activity of Sir2p. These data support a model for the spreading of silenced chromatin involving cycles of nucleosome deacetylation by Sir2p followed by recruitment of additional Sir2p, Sir3p, and Sir4p to the newly deacetylated nucleosome. This model suggests mechanisms for boundary formation, and for maintenance and inheritance of silenced chromatin. The principles are generalizable to other types of heritable chromatin states.
| |
INTRODUCTION |
|---|
|
TOP
ABSTRACT
INTRODUCTION
MATERIALS AND METHODS
RESULTS
DISCUSSION
REFERENCES
|
|---|
Regional repression, or silencing, involves the formation of a distinct, long-range chromatin structure that blocks transcription of genes within the silenced domain. The silenced state can be epigenetically inherited, implying that some aspect of the structure of silenced chromatin can template its own replication. Regional repression is an important means by which eukaryotic cells regulate gene expression. However, many aspects of the formation and function of repressive chromatin remain poorly understood.
The silencing of the mating type loci in Saccharomyces
cerevisiae is mediated by silencers, known as E and
I, that flank the two mating type loci, HMR and
HML. These silencers consist of binding sites for at least
two of three DNA binding proteins, ORC, Rap1p, and Abf1p. A working
model for the assembly of silenced chromatin at the HM loci
involves two steps (Hecht et al., 1995
; Braunstein et
al., 1996; Grunstein, 1998; Lustig, 1998; Moazed, 2001b). First,
the four Sir proteins bind to the silencer binding proteins at the
silencer. Then, Sir2p, Sir3p, and Sir4p spread in both directions from
the silencers. This model is derived from several lines of evidence.
For example, Sir1p binds Orc1p (Triolo and Sternglanz, 1996) and Sir3p
and Sir4p bind Rap1p (Moretti et al., 1994; Cockell et
al., 1995; Moretti and Shore, 2001). Also, physical interactions
between Sir1p and Sir4p (Triolo and Sternglanz, 1996), Sir2p and Sir4p
(Moazed et al., 1997; Ghidelli et al., 2001), and
Sir3p and Sir4p (Hecht et al., 1996) have been observed.
Sir3p and Sir4p also bind to the tails of histones H3 and H4 (Hecht
et al., 1995), and this interaction may enable the spreading
of Sir3p and Sir4p. Sir2p is assumed to spread along with Sir3p and
Sir4p via its interaction with Sir4p. Sir2p, Sir3p, and Sir4p also
associate with Rap1 proteins bound at telomeric repeats and extend
several kilobase pairs beyond these telomeric sequences
(Strahl-Bolsinger et al., 1997; Lieb et al.,
2001), supporting the notion that Sir2p, Sir3p, and Sir4p do spread
from sites of nucleation.
The N-terminal tails of histones H3 and H4 are hypoacetylated at the
HM loci (Braunstein et al., 1993; Suka et
al., 2001). Sir2p is an NAD+-dependent
deacetylase (Imai et al., 2000; Landry et al.,
2000; Smith et al., 2000), and is thought to deacetylate
histones at silenced loci. The hypoacetylated nucleosomes, together
with the Sir proteins, are thought to form an ordered, compact
structure that is restrictive to transcription. In support of this
model, silenced regions are less accessible to restriction nucleases (Loo and Rine, 1994) and display ordered, regularly spaced nucleosomes (Weiss and Simpson, 1998; Ravindra et al., 1999). This
structure, and hence the silenced region, is limited in extent by the
presence of "boundary" elements flanking the HM loci (Bi
et al., 1999; Donze et al., 1999).
After silenced chromatin is assembled, it must be stably
maintained during cell growth and inherited in both daughters upon cell
division. Strains bearing some silencer mutations or sir1
are defective in the establishment of silencing and are composed of
mixed populations of silenced and unsilenced cells (Pillus and Rine,
1989; Mahoney et al., 1991). Single-cell assays of these strains demonstrate that once silencing is established, the silenced chromatin is stably maintained during cell growth and inherited in both
daughters upon cell division. Thus, the mechanisms of maintenance and
inheritance of silenced chromatin differ from the mechanism of
establishment. In contrast to Sir1p, the other Sir proteins are
required for both the establishment and the maintenance of silenced chromatin.
Although this model for the assembly of silenced chromatin is
widely described (Hecht et al., 1995; Braunstein et
al., 1996; Grunstein, 1998; Lustig, 1998; Moazed, 2001b), it masks
several unresolved issues. One issue is the source of specificity in
silencing. Individually, ORC, Rap1p, and Abf1p bind hundreds of
locations in the genome, yet silencing is restricted to HMR,
HML, telomeres, and rDNA. Additionally, it is unclear how
the Sir proteins can spread over several kilobase pairs of varied
sequence, binding to the tails of histones H3 and H4 in nucleosomes,
yet not associate indiscriminately with DNA throughout the genome.
Furthermore, it is not known why four different Sir proteins are needed
when one is sufficient to provide deacetylase activity. These issues motivated a comprehensive study of the associations of Sir proteins with silent loci.
| |
MATERIALS AND METHODS |
|---|
|
TOP
ABSTRACT
INTRODUCTION
MATERIALS AND METHODS
RESULTS
DISCUSSION
REFERENCES
|
|---|
Yeast Strains and Plasmids
Strains used in this study were all derived from W303-1a (Table
1). JRY7131 was described previously
(Kirchmaier and Rine, 2001). The
sir1::LEU2,
sir1::TRP1,
sir2::LEU2,
sir2::TRP1, sir3::LEU2,
sir3::TRP1,
sir4::LEU2, and
sir4::TRP1 alleles were complete deletions of
the open reading frames generated by one-step gene conversion. The
SIR1-3xHA allele was derived from CFY416 (Gardner and Fox,
2001). The LEU2::sir2-N345A (Imai et al., 2000), hmr-e** (Axelrod and Rine, 1991), and
hmr 331-324, 274-256 (hmr-e
)
(Brand et al., 1987) alleles were described previously.
Plasmid pCF448 expresses SIR1-3xHA in pRS316 (Gardner and
Fox, 2001), and plasmid pJR1811 contains GAL4DBD-SIR1 fusion
expressed from the MET3 promoter in pRS313 (Fox et
al., 1997).
|
Chromatin Immunoprecipitation
Chromatin immunoprecipitations were performed as described
previously (Rusche and Rine, 2001) using 10 OD equivalents of cells. DNA was sheared by sonication to an average size of 500-800 base pairs
in all experiments. Each experiment was repeated at least once,
isogenic duplicate strains were used to repeat some of the experiments,
and results were uniformly reproducible. Antibodies were 4 µl of
serum from rabbits inoculated with recombinant Sir proteins (rabbit
2931, LacZ-Sir2 fusion protein; rabbit 2934, LacZ-Sir3 fusion protein;
rabbit 2913, C-terminal 46% of Sir4p; Axelrod 1991), 1 mg/ml rabbit
polyclonal anti-di-acetyl-histone H3 (K9 and K14; 06-599; Upstate
Biotechnology, Lake Placid, NY), or rabbit polyclonal
anti-hemagglutinin (HA) tag (06-831; Upstate Biotechnology). The
oligonucleotides used for polymerase chain reaction (PCR) are described
in Table 2.
|
Immunoblots
Logarithmically growing cells were suspended in SDS-PAGE sample
buffer (25 mM Tris-HCl, pH 6.8, 2.5% SDS, 2.5% glycerol, 0.01% bromphenol blue, 1.25% 
-mercaptoethanol) containing 0.1 mM
N-tosyl-L-phenylalanine chloromethyl
ketone, 1 mM phenylmethylsulfonyl fluoride, 5 µg/ml chymostatin, 2 µg/ml pepstatin A, 1 mM benzamidine, and 1× Complete protease
inhibitors (Roche Applied Science, Indianapolis, IN) plus glass
beads and vortexed for 3 min followed by incubation at 94°C for 1 min. Samples were separated on a 7.5% acrylamide gel and transferred
to nitrocellulose. Protein blots were probed using antibodies against
Sir2p (1:1000; described above), Sir3p (1:1000; described above), Sir4p
(1:500; sc-6671; Santa Cruz Biotechnology, Santa Cruz, CA), tubulin
(1:40,000, B206; Weinstein and Solomon, 1990), or phosphoglycerate
kinase (1:250; A-6457; Molecular Probes, Eugene, OR).
Immunoblotting was not successful with the rabbit polyclonal antibody against Sir4p (Axelrod, 1991).
RNA Blots
Total RNA was isolated from logarithmically growing cells as
described previously (Schmitt et al., 1990). RNA was
separated on 1.2% agarose-formaldehyde gels and transferred to Hybond
XL membranes (Amersham Biosciences, Piscataway, NJ; Sambrook et
al., 1989). To monitor silencing at HMR, the level of
a1 mRNA relative to that of SCR1 mRNA was
measured as described previously (Kirchmaier and Rine, 2001).
| |
RESULTS |
|---|
|
TOP
ABSTRACT
INTRODUCTION
MATERIALS AND METHODS
RESULTS
DISCUSSION
REFERENCES
|
|---|
This study revealed new dimensions to the formation of silenced chromatin. We present evidence first that partial assemblies of Sir proteins could form at the HMR-E silencer and then that Sir2p, Sir3p, and Sir4p and deacetylated histones were distributed throughout the silenced domain. Building on these observations, we turn to the requirements for Sir protein spreading and the link between spreading and deacetylation. The partial assemblies of Sir proteins, which formed at the silencer, did not spread away from the silencer efficiently. Thus, the assembly and spreading steps of silenced chromatin could be separated. Moreover, the role of the deacetylase activity of Sir2p was restricted to the spreading step. Finally, we describe the assembly of Sir proteins at other silencers and the molecular explanation for why the HMR-I silencer is not sufficient to cause silencing. These discoveries suggest a mechanism for spreading of chromatin proteins, the principles of which may be applicable to other types of inherited chromatin states.
Partial Assemblies of Sir Proteins Could Form at HMR-E Silencer
The first step in the formation of silenced chromatin is predicted
to be the association of Sir proteins with the silencer-binding proteins ORC, Rap1p, and Abf1p. However, it is not known whether the
association of individual Sir proteins with the silencer can occur
independently of other Sir proteins. Therefore, the association of each
of the four Sir proteins with the HMR-E silencer was
systematically examined in the presence or absence of other Sir
proteins by chromatin immunoprecipitation. In these experiments,
Sir1-HAp (Gardner and Fox, 2001) was immunoprecipitated with anti-HA
tag antibodies, and the other Sir proteins were immunoprecipitated with
antibodies raised against recombinant Sir proteins (Axelrod, 1991).
Immunoprecipitated DNA was simultaneously amplified by PCR for the
HMR-E silencer and the SSC1 promoter, a gene
whose transcription is not controlled by Sir proteins. To interpret the
data reliably, the input DNA was also amplified to determine the ratio
of the two PCR products when the templates are present in a 1:1 ratio.
Twofold serial dilutions of the starting material were performed to
verify that the PCR yield was sensitive to the amount of starting DNA.
Each immunoprecipitation was repeated at least once, and often three or
four times.
As expected, in wild-type cells, all four Sir proteins associated
preferentially with HMR-E relative to the SSC1
negative control (Figure 1A, WT lanes;
compare with input lanes). This silencer DNA coprecipitated
specifically with the Sir proteins, because HMR-E DNA was
not preferentially immunoprecipitated with anti-Sir antibodies from
cells in which the corresponding Sir genes were deleted or were not
tagged (Figure 1A, third lanes).
|
Sir1p has been proposed to bind Orc1p at the silencer independently of
other Sir proteins and then, through its association with Sir4p, to
increase the likelihood of the other Sir proteins assembling at the
silencer (Chien et al., 1993; Triolo and Sternglanz, 1996;
Gardner et al., 1999). Indeed, the association of Sir1-HAp with HMR-E was not noticeably affected in sir2
or sir3 cells and slightly, but reproducibly, reduced in
sir4 cells (Figure 1A, panel 1). A longer exposure of
this gel (Figure 1A, panel 2) demonstrates that the SSC1
internal control was recovered at similar levels from each
immunoprecipitation. For brevity, longer exposures are not shown for
the subsequent gels.
Deletion of Sir2p, Sir3p, or Sir4p results in loss of silencing at
HM loci and, because the cells express both a and 
genes, a 105-fold reduction in mating
efficiency (Rine and Herskowitz, 1987). Therefore, stable silenced
chromatin would not be expected in these cells, implying that in the
absence of Sir2p, Sir3p, or Sir4p, the remaining two proteins might not
associate with silent loci (see also Strahl-Bolsinger et
al., 1997; Sekinger and Gross, 2001). Indeed, in a
sir4 strain, Sir2p and Sir3p did not detectably associate
with HMR-E (Figure 1A, panels 3 and 4, sir4
lanes). Immunoblot analysis revealed that both Sir2p and
Sir3p were expressed efficiently in sir4 strains (Figure
1B; our unpublished data), establishing that their absence from
the silencer did not reflect a dramatic reduction in abundance. Thus,
Sir4p was crucial for the association of Sir2p and Sir3p.
In contrast to the sir4 strain, in sir2 and
sir3 strains, the remaining two Sir proteins did
associate with HMR-E, albeit at a reproducibly lower level
than in wild-type strains (Figure 1A, panel 3, sir3
lanes; panel 4, sir2 lanes; and panel 5, sir2 and sir3 lanes). Therefore, partial
assemblies of Sir proteins could form at the HMR-E silencer.
The spreading of these partial assemblies is explored below.
Deletion of Sir1p results in two metastable populations of cells,
silenced and not silenced (Pillus and Rine, 1989). Therefore, in a
sir1 strain, the remaining Sir proteins would
be expected to associate with the silencer in the
fraction of the cells that are silenced. In the MAT
sir1 strain used in this study, 33-42% of the HMR
loci were silenced (determined by mating efficiency; our unpublished
data). Therefore, roughly one-third as much HMR-E DNA
was expected to precipitate compared with the wild-type strain. Indeed,
Sir2p, Sir3p, and Sir4p associated with HMR-E with
approximately one-half to one-quarter the efficiency observed in
wild-type cells (Figure 1A, panels 3-5; the intensity of the first
lane of the sir1 sample [1X] falls between the
intensities of the second [1/2X] and third [1/4X] lanes of the
wild-type sample). Similar results were obtained all four times that
wild-type and sir1 cells were compared, and longer
exposures indicate that similar levels of the internal control,
SSC1, were recovered (our unpublished data). Thus,
the fraction of sir1 cells in which the other Sir proteins were associated with the silencer was similar to the fraction
of sir1 cells in which HMR was silenced,
consistent with Sir1p promoting the loading of the other Sir proteins
onto silencers.
To explore whether Sir1p, Sir2p, and Sir3p had overlapping or distinct
roles in stabilizing the Sir protein complex at HMR-E, double mutant strains were constructed. In a sir2
sir3 strain, the level of Sir4p associated with the
silencer was no less than the level in sir2 or
sir3 strains (Figure 1A, panel 5; our unpublished data), consistent with Sir2p and Sir3p stabilizing the
association of Sir4p with the silencer in a similar way. In contrast,
sir1 sir2 or sir1
sir3 mutant strains lost all residual association of the
remaining Sir proteins with HMR-E (Figure 1A, panels 3-5), implying that Sir1p acts in a different way than Sir2p or Sir3p in the
assembly of Sir proteins at the silencer.
Sir2p, Sir3p, and Sir4p Were Associated with DNA throughout the Silenced Domain
The limits of the silenced domain created by Sir proteins at
HMR have been determined by nuclease protection and
insertion of reporter genes at various positions (Loo and Rine, 1994;
Donze et al., 1999). The silenced domain is continuous
between the HMR-E and HMR-I silencers and extends
in both directions beyond the silencers, ending at boundary elements on
either side (Donze et al., 1999; Donze and Kamakaka, 2001).
To determine whether Sir proteins were present throughout
HMR and coincident with the silenced domain, DNA
coimmunoprecipitated with Sir2p, Sir3p, or Sir4p from wild-type cells
was analyzed by PCR for the presence of 12 different regions spanning
and extending beyond the HMR locus (Figure
2A, top). The ratios of each region
relative to SSC1 were compared in the input and
immunoprecipitated samples to determine whether that region exhibited
enhanced association with Sir proteins. In addition, because the same
amount of input or immunoprecipitated DNA was used to amplify each
region, the relative amounts of these regions in the immunoprecipitated
sample could be determined by comparing the ratios of the
immunoprecipitated sample to the input sample for each PCR product.
|
Sir2p, Sir3p, and Sir4p preferentially associated with sequences at the HMR-E and HMR-I silencers and at two regions between the silencers compared with SSC1 (Figure 2B, panels 1-3, HMR-E, X-Ya, Ya-Z1, and HMR-I). Furthermore, these Sir proteins were equally associated with all four regions, supporting the notion that Sir2p, Sir3p, and Sir4p were evenly distributed between the silencers. This distribution pattern of Sir proteins in wild-type cells is compared with the distribution of Sir proteins in mutant cells below.
To determine whether the distribution of Sir proteins matched the
previously defined silenced domain, sequences near the boundaries of
the silenced domain were examined. The telomere-proximal boundary of
the silenced domain at HMR is a tRNA gene (Donze and
Kamakaka, 2001). In the strain used here, an additional 440 base pairs
sequence corresponding to a Ty1 long terminal repeat (LTR) was
discovered between HMR-I and the tRNA gene (Figure 2A, top).
This Ty1 LTR did not block the spread of Sir proteins, because Sir2p,
Sir3p, and Sir4p associated with regions overlapping the Ty1 LTR
(Figure 2B, panels 1-3, regions 5 and 6). The boundary element at the tRNA gene did limit the spread of Sir2p, Sir3p, and Sir4p, because the
associations of these Sir proteins with regions 20 base pairs and 1000 base pairs beyond the tRNA gene (regions 7 and 8) were greatly reduced
relative to HMR-I.
The centromere-proximal boundary maps to a Ty1 LTR ~900 base pairs
outside of HMR-E, but it may not be a discrete element (Kamakaka and Donze, personal communication). Furthermore, although the
silenced domain extends beyond the silencers in either direction, the
composition of the chromatin may not be identical on the two sides of a
silencer. For example, a URA3 gene inserted with its promoter only 200 base pairs 5' of HMR-E and well within the
defined silenced domain (Figure 2A, top) is silenced one-tenth as well as a URA3 gene inserted 3' of HMR-E within the
a2 open reading frame (Donze et al., 1999).
Additionally, ordered nucleosomes, whose positioning depends on Sir
protein function, are present between the silencers but not beyond
HMR-E (Ravindra et al., 1999). In agreement with
these observations, the associations of Sir2p, Sir3p, and Sir4p were
diminished immediately outside of HMR-E and then tapered off
gradually, terminating in the vicinity of the Ty1 LTR (Figure 2B,
panels 1-3, regions 1-4, compare intensities of products from input
and immunoprecipitation samples). In summary, Sir2p, Sir3p, and
Sir4p were present at high levels between the silencers and were also
present up to, but not beyond, the boundaries of the silenced domain.
Association of Sir Proteins Correlated with a Reduction in Acetylated Histones
The presence at HMR of histones H3 and H4 with
hypoacetylated tails (Braunstein et al., 1993; Suka et
al., 2001), together with the histone deacetylase activity of
Sir2p, implied that Sir2p was the enzyme responsible for deacetylating
the tails of histones H3 and H4. If so, the extent of the
hypoacetylated domain at HMR should coincide with the
distribution of Sir2p. Indeed, the regions with high association of
Sir2p were less abundant in samples immunoprecipitated using antibodies
against diacetylated histone H3 (Figure 2B, panel 4). A similar
analysis examining a subset of these regions revealed a correlation
between hypoacetylation of histone H4 and the association of Sir2p (our
unpublished data). Furthermore, in cells expressing a
catalytically inactive Sir2p, Sir2-N345Ap (Imai et al.
2000), histone H3 in the vicinity of HMR-E was acetylated
(Figure 2C) despite the presence of the Sir2-N345Ap, Sir3p, and Sir4p
at HMR-E (Figure 4C). Therefore, Sir2p was likely the
deacetylase that acted at HMR.
Sir1p Was Restricted to the Silencer
Because Sir1p is required for the establishment but not the
maintenance of silencing, Sir1p has been thought to act only at the
silencer. However, recent results suggest that Sir1p spreads throughout
the silenced domain (Gardner and Fox, 2001). Therefore, the
distribution of Sir1-HAp was examined across HMR. Unlike the other Sir proteins (Figure 2B), Sir1-HAp was not associated equally with both silencers and the two regions between but instead was associated primarily with the HMR-E silencer (Figure
3A).
|
Due to the heterogeneity in size of sheared DNA in chromatin immunoprecipitation assays, this experiment could not distinguish between low-efficiency spreading of Sir1-HAp and coprecipitation of large DNA fragments with Sir1-HAp cross-linked to the silencer. Therefore, the ability of Sir1p to spread was reinvestigated using a strain in which the HMR-E silencer was replaced with a synthetic silencer, consisting of Rap1p and Abf1p binding sites as well as four Gal4p binding sites in place of the ORC binding site (Figure 2A, bottom). This strain expressed both a Gal4-Sir1 fusion protein, which binds to the synthetic silencer and is required for silencing, and Sir1-HAp, which cannot bind to this synthetic silencer lacking ORC binding sites. If additional Sir1p molecules were recruited to the HMR locus beyond those that function at the silencer, then Sir1-HAp would be associated with the HMR locus. However, Sir1-HAp was not detectably associated either with the synthetic silencer or internal to HMR (Figure 3B). In contrast, Sir1-HAp did associate with the HML-E silencer, presumably through ORC bound to HML-E, indicating that the immunoprecipitation was successful. Gal4-Sir1p mediated silencing at HMR, because transcription of a1 mRNA was repressed (Figure 3C) and Sir2p, Sir3p, and Sir4p were associated with and spread across HMR (our unpublished data). Thus, Sir1p acted primarily at the silencer and little, if any, Sir1p spread from the HMR-E silencer.
Spreading of Sir Proteins Required Sir3p and Deacetylase Activity of Sir2p
The ability of some Sir proteins to associate with the
HMR-E silencer in the absence of others (Figure 1A) raised
the question of whether Sir proteins require one another to spread. For
example, because Sir3p and Sir4p bind to both the HMR-E
silencer and the tails of histones H3 and H4 in the absence of Sir2p
(Figure 1A; Hecht et al., 1995), Sir3p and Sir4p might
spread in the absence of Sir2p. Therefore, the distributions of Sir2p,
Sir3p, and Sir4p across HMR were examined in
sir2 and sir3 strains. In contrast to the
wild-type strain (Figure 2B), these Sir proteins associated primarily
with the HMR-E silencer in these mutants (Figure
4A). To facilitate the comparison of Sir
protein distributions in wild-type and sir
cells, DNA immunoprecipitated from these strains was also analyzed by
simultaneous amplification of the HMR-E silencer and an
internal region, Ya-Z1 (Figure 4B). The amount of template DNA was
adjusted to maintain the HMR-E product at a constant level.
The internal region was relatively less abundant in samples from
sir2 or sir3 cells compared with wild-type
samples, confirming the inefficient spreading. Thus, Sir2p, Sir3p, and
Sir4p were mutually dependent on one another for stable association
with nonsilencer DNA. The chromatin immunoprecipitation assay cannot
reveal whether transient, low-affinity interactions occur between Sir
proteins and chromatin at HMR. However, even if such
interactions do occur, they are not stable and thus no spreading was
detected. The precipitation of some small amounts of the internal
regions X-Ya and Ya-Z1 was likely due to limitations in shearing the
DNA before immunoprecipitation, as seen above with Sir1-HAp. The
greatly reduced association of Sir2p, Sir3p, and Sir4p with
HMR-I was surprising because both the
HMR-E and HMR-I silencers had
previously been thought to be sites at which Sir proteins assemble (see
below).
|
There are at least two ways in which Sir2p may facilitate the spreading
of Sir3p and Sir4p. First, Sir2p could be an essential structural
component of the silenced chromatin, without which the chromatin is not
stable. Alternatively, its deacetylase activity may be required for
Sir3p and Sir4p to spread. To distinguish between these alternatives,
the spreading of Sir2p, Sir3p, and Sir4p was examined in strains
expressing a catalytically inactive point mutant of Sir2p, Sir2-N345Ap.
This mutant protein is expressed efficiently and is likely to be
structurally intact (Imai et al., 2000; Min et
al., 2001). The catalytically inactive Sir2p associated with
HMR-E (Figure 4C, panel 1), therefore, deacetylation at the silencer was not required for loading. However, mutant Sir2p did not
spread efficiently. Likewise, in this strain, Sir3p and Sir4p also
failed to spread efficiently (Figure 4C, panels 2 and 3). Therefore,
the histone deacetylase activity of Sir2p was required for the stable
spreading of Sir2p, Sir3p, and Sir4p.
The separation of Sir protein loading and spreading also enabled
further investigation of the role of Sir1p in establishing silencing.
The experiments described above demonstrated that Sir1p facilitated the
association of Sir proteins with the HMR-E silencer (Figure
1A). In principle, Sir1p might also facilitate establishment by
activating the spreading of Sir proteins from the silencer, perhaps by
inducing a conformational change or catalyzing a posttranslational modification on one of the other silencing proteins. However, no
reduction in the spreading of Sir2p, Sir3p, or Sir4p was detected in
sir1 compared with wild-type cells either by examining
each region of HMR individually or by examining the relative
proportions of sequences at the HMR-E silencer and an
internal region, Ya-Z1 (our unpublished data). Therefore, Sir1p
acted at the nucleation step but not at the spreading step of silenced
chromatin formation.
In summary, efficient spreading of Sir2p, Sir3p, and Sir4p required the presence of all three proteins at the silencer and a catalytically active Sir2p but did not require Sir1p. Furthermore, when spreading was blocked by the absence of one protein, the maximal association of Sir proteins was at the HMR-E silencer. Therefore, the HMR-E silencer was a nucleation site for the assembly of Sir proteins and was the site from which spreading occurred.
HMR-I Silencer Did Not Efficiently Nucleate Assembly of Sir Proteins
The HMR-I silencer cannot silence
independently of the HMR-E silencer (Brand et
al., 1985). Sir1p did not efficiently associate with
HMR-I (Figure 3A) and in sir2,
sir2-N345A, and sir3 strains, the remaining
Sir proteins did not associate with HMR-I, whereas they did
associate with HMR-E (Figure 4, A and C). These observations implied that the HMR-I silencer was not sufficient for
silencing because it could not efficiently nucleate the assembly of Sir proteins. In such a scenario, Sir proteins would only arrive at the
HMR-I silencer after their loading and spreading from
HMR-E. To test this idea, the association of Sir2p, Sir3p,
and Sir4p with HMR-I was examined in strains bearing
mutations (e**) (Axelrod and Rine, 1991), or deletions
(e; Brand et al., 1987), in the Rap1p and
Abf1p binding sites of the HMR-E silencer. These mutations
prevented the efficient binding of Sir2p, Sir3p, and Sir4p to
HMR-E and reduced the mating efficiency to 3 × 103 (e**) or 3 × 104 (e) relative to wild-type
(our unpublished data). In contrast to a strain bearing a
wild-type HMR-E silencer, in strains bearing mutations in
the HMR-E silencer, the association of Sir2p, Sir3p, and
Sir4p with HMR-I was significantly reduced (Figure
5, panels 1, 3, and 5, compare lanes 3 and 4 with 5-8). These Sir proteins associated with the
HML-E silencer at wild-type levels in these strains (Figure
5, panels 2, 4, and 6), indicating that the immunoprecipitation was
successful. Therefore, Sir2p, Sir3p, and Sir4p did not associate with
HMR-I unless HMR-E was functional, implying that
these Sir proteins spread from HMR-E to HMR-I
rather than initially assembling at HMR-I.
|
A Synthetic Silencer Had Similar Requirements for Loading and Spreading
Many of the key experiments exploring the mechanism of silencing
have used simplified synthetic silencers in place of HMR-E. For example, to explore the spreading of Sir1p (Figure 3B), a synthetic
silencer to which Gal4-Sir1p can be tethered was used (Figure 2A,
bottom). To determine whether such a silencer behaves similarly to the
natural HMR-E silencer, chromatin immunoprecipitation experiments were conducted in cells expressing or not expressing Gal4-Sir1p (Figure 6A). In this case,
MAT served as the negative control locus. The association of
Sir2p, Sir3p, and Sir4p with the synthetic silencer required the
expression of Gal4-Sir1p (Figure 6A, wild-type column, top gel of each
trio), consistent with Sir1p playing a critical role in the
establishment of silencing by facilitating the assembly of a Sir
protein complex at the silencer.
|
As at HMR-E, the other Sir proteins associated with
the synthetic silencer in the absence of Sir2p or Sir3p (Figure 6A,
sir2 and sir3 column, top gels; compare the
ratios of the HMR-Ga1SS to MAT PCR products from mutants
with wild-type), but required Sir4p (Figure 6A, sir4
column, top gels). Spreading, as measured by association with the
a1 gene at HMR, did not occur unless all Sir
proteins were present (Figure 6A, sir2 and
sir3 columns, middle gels). The lack of Sir2p, Sir3p, or
Sir4p resulted in loss of silencing (Figure 6B), as expected.
Therefore, the interdependencies for loading and spreading were similar
at the wild-type and synthetic silencers.
Loading of Sir Proteins at HML-E and HML-I Silencers Had Similar Requirements to HMR-E
To determine whether the architecture and assembly of silenced
chromatin at HML silencers is similar to HMR-E,
the association of Sir2p, Sir3p, and Sir4p with both HML
silencers was examined. As at the HMR-E silencer,
association of these Sir proteins with both HML silencers
required Sir4p but not Sir2p or Sir3p (Figures 6A, bottom gels, and 7).
Also, as at HMR, the association of Sir proteins at the
HML silencers was reduced in a sir1 strain
(Figure 7). The same dilutions of the
same samples were used in both Figures 1A and 7, revealing that in a
sir1 strain there was relatively more Sir2p, Sir3p, and
Sir4p at HMR-E than at either HML silencer. Therefore, the requirements for assembly of the Sir protein complex were similar at all of the silencers examined, with the exception of
HMR-I, which did not efficiently nucleate the assembly of
Sir proteins. The only detectable difference among the other silencers was the degree of dependence on Sir1p to recruit the remaining Sir
proteins.
|
| |
DISCUSSION |
|---|
|
TOP
ABSTRACT
INTRODUCTION
MATERIALS AND METHODS
RESULTS
DISCUSSION
REFERENCES
|
|---|
This study established that partial assemblies of Sir proteins
could form at the silencers but could not spread efficiently from the
silencers. Thus, the assembly and spreading steps of silenced chromatin
formation were separable, thereby enabling a more precise determination
of the step at which a protein or activity acts. For example, Sir1p
facilitated assembly of Sir proteins at a silencer, whereas the
deacetylase activity of Sir2p was required for spreading. The
separation of the assembly and spreading steps also revealed that Sir
proteins assembled at the HMR-E but not the HMR-I
silencer. Previous studies concluded that deletion of one Sir protein
resulted in the loss of all Sir proteins at the silent loci. However,
those studies used primers that either amplified regions that did not
include the silencers (Strahl-Bolsinger et al., 1997) or
only examined sir4 cells (Sekinger and Gross, 2001).
Model for Spreading of Sir Proteins
The efficient spreading of Sir2p, Sir3p, and Sir4p at
HMR required Sir2p and Sir3p (Figures 4A and 6A) and the
deacetylase activity of Sir2p (Figure 4C). Furthermore, this study
revealed that deacetylation was carried out by Sir2p (Figure 2C) and
was an integral part of the spreading process and not a later step that
occurred after Sir2p had spread. These findings suggest that the
decrease in the association of catalytically defective Sir2p with
telomeres and rDNA (Tanny et al., 1999) probably resulted from the disruption of spreading. Although Sir2 protein levels were
slightly reduced in the absence of Sir3p (Figure 1B; our unpublished
data), it is most likely that the disruption of Sir2p's ability
to spread resulted from the absence of Sir3p rather than the reduction
in Sir2p because silencing, and hence spreading, occurs in diploid
cells heterozygous for null alleles of SIR2 (Rine and
Herskowitz, 1987) as well as on HMR-bearing multicopy plasmids in strains expressing wild-type levels of Sir proteins (Abraham et al., 1984).
The requirement for a deacetylase in the spreading process inspires a
"sequential deacetylation" model for the propagation of silenced
chromatin along the DNA (Figure 8A). In
this model, the assembly of a Sir protein complex at the silencer
brings Sir2p in juxtaposition to the neighboring nucleosome, which,
like most nucleosomes in S. cerevisiae (Waterborg, 2000),
presumably has acetylated histone tails. Sir2p then deacetylates these
tails, creating new high-affinity binding sites for Sir3p and Sir4p, which bind hypoacetylated tails of histones H3 and H4 more tightly than
acetylated tails (Hecht et al., 1995). Sir3p and Sir4p then bind this newly deacetylated nucleosome adjacent to the silencer, bringing another Sir2p molecule. This new Sir2p deacetylates the next
nucleosome, allowing further association of Sir3p and Sir4p. Finally,
interactions among the Sir proteins result in the formation of an
ordered, compact structure. For simplicity, this last step is not
represented in the figure. This model borrows from and extends earlier
models for spreading (Braunstein et al., 1996; Grunstein,
1998; Moazed, 2001a).
|
This sequential deacetylation model has two key features. The first is the use of Sir3p and Sir4p as an adaptor to link Sir2p to chromatin. This feature solves the fundamental problem of restricting silencing to specific loci. Sir2p relies on the Sir3p-Sir4p adaptor to bring it into proximity to its substrate. However, Sir3p and Sir4p themselves cannot stably associate with nucleosomes in the absence of the Sir2p deacetylase, which generates a binding site for Sir4p and Sir3p on histone tails. Thus, the Sir proteins are mutually dependent on one another for their stable association with nucleosomes, preventing silenced chromatin formation from initiating at sites other than those with the means to recruit Sir proteins.
The second key feature of the sequential deacetylation model is that
Sir2p must act stoichiometrically rather than processively. Moreover,
Sir2p has low deacetylase activity in vitro compared with its paralog
Hst2p (Landry et al., 2000; Smith et al., 2000). Sir2p may be maximally active only when assembled with other Sir proteins into silenced chromatin and brought into proximity to its
substrate. These features probably prevent Sir2p from indiscriminately deacetylating nucleosomes throughout the genome. This sequential deacetylation model does not exclude the possibility that the Sir2p
deacetylase might have additional functions, such as using the energy
released by NAD+ hydrolysis to promote spreading
(Moazed, 2001b).
The inability of Sir3p to spread from the HMR-E silencer in
the absence of Sir2p (Figures 4A and 6A) seems to conflict with the
reported ability of Sir3p, when overexpressed, to form an "extended" silenced chromatin at telomeres that is reported to contain Sir3p but not Sir2p or Sir4p (Strahl-Bolsinger et
al., 1997). This extension of telomeric silencing upon
overexpression of Sir3p does require both Sir2p and Sir4p (Renauld
et al., 1993), adding to the puzzle of how Sir3p would
spread farther than Sir2p or Sir4p. Perhaps Sir2p and Sir4p are present
in extended chromatin but are not accessible to the antibodies due to
the excess Sir3p.
The ability to promote the spreading of chromatin-associated proteins
may be a general property of the Sir2p family of deacetylases. For
example, Sir2p may facilitate a similar spreading process at the rDNA,
but use adaptor proteins other than Sir3p and Sir4p. Similarly, Hst1p,
a paralog of Sir2p, may promote the spreading of Sum1-1p.
SUM1-1 is a change-of-function mutation that leads to
silencing of the HM loci in the absence of Sir proteins
(Klar et al., 1985; Laurenson and Rine, 1991). Sum1-1p
associates with Hst1p (Rusche and Rine, 2001; Sutton et al.,
2001), much as Sir4p associates with Sir2p, and Sum1-1p spreads across
HMR much as the Sir proteins do (Rusche and Rine, 2001).
Interestingly, Sum1-1p does not spread when Hst1p is absent (Rusche and
Rine, unpublished data), implying that, like Sir2p, Hst1p regulates the
spreading of chromatin-associated proteins.
Protein-Protein Interactions at the Silencer
Past models of the architecture of Sir proteins at the silencer have been inferred primarily from pairwise physical and genetic interactions between individual components of the complex. This study, using chromatin immunoprecipitation, identified some of the protein-protein interactions required for the structure and implied an order of assembly, as outlined below. The interdependencies of Sir proteins to associate with a silencer were examined at HMR-E, HML-E, HML-I, and a synthetic silencer (Figures 1A, 6A, and 7). Importantly, although these silencers have different combinations of ORC, Rap1p, and Abf1p binding sites, all displayed the same requirements for the loading of Sir proteins. Sir4p was essential for loading Sir2p and Sir3p, Sir1p either improved the efficiency of or was essential for loading the other Sir proteins, and Sir2p and Sir3p were less important. Sir1p could associate with the HMR-E silencer independently of any individual other Sir protein.
In the simplest model for the architecture of the Sir protein complex
at HMR-E (Figure 8A, top line), Sir1p binds independently of
other Sir proteins to Orc1p (Triolo and Sternglanz, 1996; Gardner et al., 1999), and perhaps Rap1p as well (Chien et
al., 1993). The association of Sir1p with the silencer was
reproducibly reduced upon deletion of Sir4p (Figure 1A), indicating
that this association may be stabilized by, although not completely
dependent on, contact with Sir4p, consistent with two-hybrid
observations (Triolo and Sternglanz, 1996).
In this model, Sir4p is central, making contacts with all three other
Sir proteins as well as Rap1p. This placement was consistent with the
deletion of Sir4p causing the most severe reductions in association of
the other Sir proteins with silencers (Figures 1A, 6A, and 7). Sir4p
associates with the silencer through Sir1p and Rap1p, and Sir3p
associates through Rap1p. Although pairwise interactions of Sir4p or
Sir3p with Rap1p are observed in vitro, we speculate that in the cell,
pairwise interactions between Sir4p or Sir3p and Rap1p or Sir1p are
individually inadequate for a stable association. Interactions among at
least three of these proteins may be required to form a stable complex.
This requirement, together with the low probability of a three-way
collision, can explain why individual Rap1p or ORC binding sites are
not silencers, because both DNA binding proteins would be needed to
stabilize interactions among the Sir proteins. In support of this
speculation, Sir3p does not associate with Rap1p at the silencer in the
absence of Sir4p (Figures 1A, 6A, and 7). In contrast, Sir4p does
associate with the silencer-bound Rap1p in the absence of Sir3p, but
not if Sir1p is also absent (Figures 1A, 6A, and 7). In this scenario, Sir1p facilitates the assembly of the Sir protein complex by anchoring Sir4p at the silencer, allowing it to bind without Sir3p. Sir3p may
also associate with Abf1p (Dhillon and Kamakaka, 2000). However, this
association is unstable in the absence of Sir4p (Figure 1A and 6A).
Finally, the interaction between Sir2p and Sir4p (Moazed et
al., 1997; Ghidelli et al., 2001) suggests that Sir4p
brings Sir2p to the silencer.
The reduced recovery of HMR-E silencer DNA in
immunoprecipitations from sir2 or sir3
cells compared with wild-type cells probably reflects the absence of
spreading to neighboring nucleosomes. With fewer proteins bound per
locus, fewer epitopes would be available for antibody binding.
Consistent with this conclusion, the associations of Sir3p and Sir4p
with the HMR-E silencer were reduced nearly as much in
sir2-N345A as in sir2 cells (our unpublished data).
The four endogenous silencers at the HM loci differ in
strength and composition of binding sites. Of the four, the
HMR-I silencer is the only one that cannot establish
silencing independently (Brand et al., 1985). Interestingly,
HMR-I was also the only silencer that did not associate with
Sir1p (Figures 1A and 3; our unpublished data) and lacked the
ability to recruit other Sir proteins efficiently (Figure 5).
Consistent with these observations, Sir-dependent nuclease protection
at HMR is abolished when the HMR-E silencer is
deleted, even although the HMR-I silencer remains (Loo and Rine, 1994). Thus, the ability of a silencer to recruit Sir proteins correlated with the ability to establish silencing.
Boundaries of the Silent Domain at HMR
This study established that Sir2p, Sir3p, and Sir4p were found
throughout the silenced domain at HMR but not outside its
previously defined borders (see also Lieb et al., 2001). The
association of Sir2p, Sir3p, and Sir4p with sequences throughout the
silenced domain is the most compelling evidence for these Sir proteins being key structural components of silenced chromatin. Additionally, Sir2p, Sir3p, and Sir4p did not seem to spread from the
HMR-E silencer equally in both directions. DNA sequences
between the silencers were much more abundant in the immunoprecipitated
samples than were sequences outside of HMR-E yet still
within the silenced domain. The reduced immunoprecipitation of regions
outside the silencers but within the silenced domain was interesting
and could result from Sir proteins either being associated in only a
fraction of the cells or being present in all cells but at a reduced
density. This result and others (Shei and Broach, 1995; Donze et
al., 1999; Ravindra et al., 1999) imply that silenced
chromatin is not equivalent on the two sides of the HMR-E
silencer. The origin and significance of this pattern awaits further study.
The sequential deacetylation model outlined above can explain how
heterochromatin boundaries function (Figure 8B). The model implies that
silenced chromatin must spread sequentially from nucleosome to
nucleosome due to the limited range of the Sir2p deacetylase.
Therefore, disrupting the chain of deacetylatable nucleosomes would
interrupt the spread of Sir proteins (see also Bi and Broach, 2001).
For example, a gap could be created in the nucleosome array by a
DNA-binding protein that displaces histones. Similarly, a localized
histone acetyltransferase, as is observed at some promoters, could
effectively limit the spread of silenced chromatin by acetylating
histones more effectively than Sir2p can deacetylate them. In fact, DNA
binding proteins and tethered acetyltransferases do act as barrier
elements (Bi and Broach, 1999; Fourel et al., 1999; Donze
and Kamakaka, 2001). Interestingly, the distribution of Sum1-1p at
HMR is virtually identical to that of Sir2p, Sir3p, and
Sir4p (Rusche and Rine, unpublished data), implying that boundary
elements are not specific to a particular type of silencing complex.
Model for Maintenance and Epigenetic Inheritance of a Chromatin Structure
Once silenced chromatin has been established, this chromatin
structure is maintained during cell growth and inherited upon cell
division (Pillus and Rine, 1989; Mahoney et al., 1991). Both of these phenomena could occur through the sequential deacetylation mechanism of spreading. Maintenance would occur by the spreading of
Sir2p, Sir3, and Sir4p back into gaps in which silencing has been disrupted.
The model presented above for the epigenetic inheritance of silenced
chromatin (Figure 8C; see also Braunstein et al., 1996) is
rooted in how nucleosomes are affected by replication. Passage of a
replication fork causes the two H2A-H2B dimers to disassociate from
both the H3-H4 tetramer and the DNA. The H3-H4 tetramers remain
associated with DNA (Kimura and Cook, 2001) and are randomly distributed to the two sister molecules (Jackson and Chalkley, 1985).
Thus, if an epigenetic mark were associated with the histones, it would
logically be on H3 and/or H4. At the silent loci, DNA replication would
distribute H3-H4 tetramers with hypoacetylated tails to both sister
chromatids. Sir proteins may remain associated with these histone tails
or may quickly reassociate with the tails after passage of the
replication fork. After replication the resulting nucleosome array
would then be completed by the incorporation of newly synthesized and
acetylated histones. However, these nucleosomes with acetylated tails
would be adjacent to the Sir2p deacetylase along with Sir3p and Sir4p
bound to the old H3-H4 tetramers. Sir2p would then deacetylate
these tails, creating new binding sites for Sir3p and Sir4p, restoring
the silenced state to the underlying gene. Cooperativity between the
nucleosome-bound Sir complexes, silencer binding proteins, and unbound
Sir proteins may help direct Sir3p and Sir4p to those newly
deacetylated histones in heterochromatin rather than deacetylated
histones elsewhere in the genome. In fact, silencer binding proteins
probably stabilize the chromatin structure (Cheng and Gartenberg,
2000), and it is likely that in the absence of a silencer, silenced
chromatin may initially be inherited but is not stable long enough to
be detected (Holmes and Broach, 1996).
Perspective
The mechanisms for spreading and inheritance outlined above could
apply to multiple types of epigenetically heritable states, including
other types of regional repression and centromere inheritance (see also
Jenuwein, 2001). There are two fundamental features required to permit
both the spreading of a chromatin state from the site of initiation and
the inheritance of that chromatin state: a mark that is inherited on
both duplexes following replication and the ability of that mark to
recruit an enzyme or complex that makes an additional mark. In the
Sir-based mechanism described above, the hypoacetylated H3-H4 tetramer
is the mark, providing a high-affinity binding site for the Sir
complex, which, in turn, creates new marks. In the case of mating-type
silencing in Schizosaccharoyces pombe or position effect
variegation in Drosophila, histone H3 methylated at lysine 9 is a mark, which is recognized and bound by a chromodomain-containing
protein, in partnership with a histone methyltransferase capable of
adding additional marks (Jenuwein, 2001). Finally, in many organisms
centromeres seem to be epigenetically inherited and are marked by the
presence of an H3-like protein, CENP-A (Sullivan et al.,
2001). The inheritance of this mark from one generation to the next
predicts that there is a CENP-A chaperone that assembles new
CENP-A-containing nucleosomes near existing ones.
| |
ACKNOWLEDGMENTS |
|---|
We thank Catherine Fox and Leonard Guarente for strains and plasmids; Rohinton Kamakaka and David Donze for communicating unpublished results; members of the laboratory for helpful discussions; and Michael Botchan, Alexa Franco, Paul Kaufmann, Michael Kobor, Judith Sharp, and reviewers for helpful comments and suggestions. This work was supported by postdoctoral fellowships from the Damon Runyon-Walter Winchell Cancer Research Fund (to L.N.R.), the National Institutes of Health (F32GM19392, to A.L.K.), and American Cancer Society (PF-01-126-01-MBC, to A.L.K.), and a grant from the National Institutes of Health (GM-31105, to J.R.). Core support was provided by National Institute of Environmental Health Sciences Mutagenesis Center grant E50 1896.
| |
FOOTNOTES |
|---|
* Corresponding author. E-mail address: jrine{at}uclink4.berkeley.edu.
Article published online ahead of print. Mol. Biol. Cell 10.1091/mbc.E02-03-0175. Article and publication date are at www.molbiolcell.org/cgi/doi/10.1091/mbc.E02-03-0175.
| |
REFERENCES |
|---|
|
TOP
ABSTRACT
INTRODUCTION
MATERIALS AND METHODS
RESULTS
DISCUSSION
REFERENCES
|
|---|
) had samples immunoprecipitated from strains lacking Sir2p
(panel 1), Sir3p (panel 2), Sir4p (panel 3), or represent a mock
precipitation with no antibody from SIR cells (panel 4).
1/19,000 of the input DNA, 1/250 or 1/500 (panels 1 and 2), 1/125 or
1/250 (panel 3), and 1/500 or 1/1000 (panel 4) of immunoprecipitated
DNA was analyzed. The primers used to amplify region 6 generated a
product of the predicted size (middle band) as well as a larger product
of unknown identity, presumably involving another Ty1 LTR. (C)
Chromatin immunoprecipitations from MATa SIR2,
sir2-N345A, or sir2
