|Home | About | Journals | Submit | Contact Us | Français|
Barrier elements that are able to block the propagation of transcriptional silencing in yeast are functionally similar to chromatin boundary/insulator elements in metazoans that delimit functional chromosomal domains. We show that the upstream activating sequences of many highly expressed ribosome protein genes and glycolytic genes exhibit barrier activity. Analyses of these barriers indicate that binding sites for transcriptional regulators Rap1p, Abf1p, Reb1p, Adr1p and Gcn4p may participate in barrier function. We also present evidence suggesting that Rap1p is directly involved in barrier activity, and its barrier function correlates with local changes in chromatin structure. We further demonstrate that tethering the transcriptional activation domain of Rap1p to DNA is sufficient to recapitulate barrier activity. Moreover, targeting the activation domain of Adr1p or Gcn4p also establishes a barrier to silencing. These results support the notion that transcriptional regulators could also participate in delimiting functional domains in the genome.
The eukaryotic genome is organized into discrete functional domains in which gene expression is either permitted or repressed. Domains permissive for gene expression are usually defined as euchromatin or active chromatin, and domains that repress gene expression heterochromatin or silent chromatin. The HML and HMR loci in Saccharomyces cerevisiae are silent chromatin domains (1). Compared to active chromatin, silent chromatin is more compact and its histone components are hypoacetylated (2). The SIR complex consisting of Sir2p through Sir4p is an integral part of yeast silent chromatin. Sir3p and Sir4p interact with the N-terminal tails of histones H3 and H4 providing the structural basis of silent chromatin (1). Sir2p was recently found to be an NAD-dependent histone deacetylase and was suggested to be responsible for histone hypoacetylation in silent chromatin (3). Silent chromatin is initiated at small cis-acting DNA elements known as the E and I silencers flanking each of the HM loci (4).
Increasing evidence indicates that histone deacetylation is key to the propagation of silent chromatin (4). In vertebrates and fission yeast, spread of silent chromatin involves a chain of events of histone H3 deacetylation → H3 methylation → binding of methylated H3 by HP1 (heterochromatin protein 1) or swi6 (4). In S.cerevisiae, there is evidence that Sir3p (hence the SIR complex) has much higher affinity to unacetylated histone H4 than acetylated H4 (5). Based on this and the fact that Sir2p is a histone deacetylase, a refined model for the propagation of silent chromatin can be proposed. In this model, Sir2p recruited to the silencer deacetylates histones in an adjacent nucleosome enabling it to bind another SIR complex with high affinity. The nucleosome-bound SIR complex in turn deacetylates the neighboring nucleosome allowing it to recruit a new SIR complex. In this manner, the SIR complex promotes its own propagation along an array of nucleosomes (4).
The fact that transcriptionally silent chromatin can encroach upon active chromatin poses the question of how interspersed domains of silent and active chromatin are demarcated. Studies of the Drosophila and vertebrate genomes demonstrated that some domains are delimited by boundary or insulator elements (6). These elements are specialized DNA sequences that act as barriers to enhancers and/or silent chromatin. One of the best characterized insulators is the chicken HS4 insulator at the β-globin locus. Histones surrounding this insulator are hyperacetylated indicating the presence of histone acetyltransferase (HAT) activity at the insulator (6). Recently, sequences that could block the spread of silent chromatin have been discovered in S.cerevisiae (7,8). These sequences (referred to as silent chromatin barriers) include sub-telomeric anti-silencing regions (STARs), the right boundary of HMR, and the upstream activating sequence (UAS) of the highly expressed TEF2 gene (9–11). The mechanisms underlying barrier functions in yeast are not known.
TEF2-UAS contains three tandem repressor activator protein 1 (Rap1p)-binding sites that are necessary and sufficient for its barrier function (11). Rap1p is an abundant sequence-specific DNA-binding protein (12). Variants of the 13 bp consensus sequence for Rap1p binding (ACACC CRYACAYM) (13) lie not only in the UASs of numerous genes but also in the silencers of the HM loci and in telomeric C1–3A repeats. Accordingly, Rap1p functions as a global regulator of transcriptional activation and repression, silencing, as well as telomere length (12). Rap1p performs these diverse functions by interacting with different factors in respective contexts. As an activator, Rap1p binds to the promoters of 362 ORFs (13) and can function via at least three mechanisms. First, Rap1p binding can ‘open up’ local chromatin structure at a promoter to help another activator to bind (14). Secondly, Rap1p bound to DNA can help Gcr1p bind to an adjacent site in the promoter through physical interaction (15). Thirdly, Rap1p is involved in the recruitment of the NuA4 HAT complex and TFIID to ribosome protein genes (RPGs) (16,17). As a silencing factor, Rap1p binds to silencers or telomeric repeats and recruits Sir3p/Sir4p through direct interactions (18). Rap1p also interacts with telomere-specific proteins in executing its role in regulating telomere length (19). It is puzzling that a protein that helps establish silent chromatin could also act as a barrier to its propagation.
In this study, we found that in addition to TEF2-UAS, UASs of many yeast genes exhibit barrier activity. Analyses of these UASs indicate that binding sites for transcriptional regulators Abf1p, Reb1p, Adr1p, Gcn4p, as well as Rap1p, may participate in barrier function. These UASs are dispersed across the genome and might play a role in defining functional chromosomal domains. Moreover, we demonstrated that targeting the activation domain of Rap1p, Adr1p or Gcn4p alone, led to barrier function. Since the activation domain of an activator was usually involved in interacting with co-activators and/or components of the transcriptional machinery, these data indicated that other factors might also be involved in barrier function.
Plasmids 1–53 were used to make strains 1–53, respectively (see Figs Figs11–6). Plasmid 1 was previously named pMB22-a that contained a HindIII–BamHI fragment of chromosome III (coordinates 14838–16263) with a URA3 gene inserted at its EcoRV site (20). Plasmid 2 was derived from 1 by inserting TEF2-UAS (–511 to –407 bp relative to the translation start codon) at the SnaBI site. Plasmids 3–38 were identical to 2 except that different fragments from various promoters (see Figs Figs11 and and2)2) were inserted at the SnaBI site. Plasmid 29 had a sequence of chromosome III (295327–295713) containing the HMR-tRNA gene (21) inserted at the SnaBI site of plasmid 1. Plasmids 39 and 40 were previously named pYXB26 and pYXB59, respectively (11). Plasmid 41 was identical to pYXB59 except bearing mutations illustrated in Figure Figure3A.3A. Plasmids 42–48 were previously described as pYXB29, 48, 28, 27, 59, 31 and 37, respectively (11). Plasmid 50 was derived from plasmid 1 by inserting two copies of a sequence bearing the consensus binding sequence of LexA (bold) (22), GGGGTACGTACTGTATGTACATACAGGATATCGGGG, at the SnaBI site. Plasmid 51 was identical to 50 except having only one copy of the LexA-binding sequence. Plasmid 52 was derived from pYXB26 (11) by inserting a SpeI-ADE2-AvrII sequence at the SpeI site. Plasmid 53 was derived from plasmid 52 by inserting TEF2-UAS (–511 to –407 bp relative to the translation start codon) at the SpeI site.
Plasmid L1 (see Fig. Fig.5A)5A) was made by inserting LEU2 into pBTM116 (23) carrying the LexA gene flanked by the promoter and terminator of ADH1. L2 through L8 were derived from L1 by fusing various sequences to LexA (see Fig. Fig.5A5A and B). Plasmid pRS425 (24) carried the 2 µm origin and LEU2.
Strain YXB76 was MATa ura3-52 leu2-3,112 ade2-1 lys1-1 his5-2 can1-100, E-HMLα-Iinverted (20). Y2047b was MATa HMRa HMLα EΔ79–113::SUP4-o IΔ242 LEU2-GAL10-FLP1 ura3-52 ade2-1 lys1-1 his5-1 can1-100 [cir0] (25). Strain 1 was made by transforming YXB76 to Ura+ with HindIII + BamHI digested plasmid 1. Strains 2–38, 50 and 51 were similarly made with corresponding plasmids, respectively (e.g. strain 2 was made with plasmid 2). Strains 39, 40 and 42–48 were previously described as YXB26, 48, 29, 48, 28, 27, 59, 31 and 37, respectively (11). Strains 41, 52 and 53 (see Figs Figs3C3C and and6)6) were made by transforming Y2047b to canavanine-resistant by BamHI-digested plasmids 41, 52 and 53, respectively. Strains 40′–48′ were sir3::URA3 derivatives of strains 40–48, respectively. The relevant genotypes of all the strains were confirmed by Southern blotting.
Quantitative mating was performed as described (11).
Rap1p was expressed in Escherichia coli BL21 (DE3) from pL3S5, a pET vector carrying the RAP1 gene downstream of a T7 promoter (26). Crude protein extracts were prepared from 40 ml cultures harvested after 3 h of IPTG induction and resuspended in 1 ml of lysis buffer (50 mM NaH2PO4, 300 mM NaCl, pH 8.0). A control extract was prepared from a culture without IPTG induction. Western blotting revealed that Rap1p protein was present in the induced extract but not the uninduced one (data not shown). DNA fragments of 26 bp in length (see Fig. Fig.3A)3A) were end-labeled with 32P-phosphate. Fifteen microliter binding reactions were prepared, each consisting of 20 000 c.p.m. radiolabeled DNA probe, 5 µg of crude protein extract, non-specific competitor DNAs (0.3 µg of yeast tRNA, 0.3 µg of salmon sperm DNA and 0.5 µg of E.coli DNA), 10 µg of BSA, 10 mM Tris–HCl (pH 8.0), 10 mM MgCl2 and 8% glycerol. The reactions were incubated at 25°C for 20 min and loaded onto non-denaturing polyacrylamide gels (4–5%). Gels were run at 40 mA for 1 h in 0.5× TBE buffer.
Cells were grown in YPR (yeast extract + bacto peptone + 2% raffinose) to early log phase. Galactose (2%) was then added to the culture to induce the expression of FLP1. After 2.5 h of incubation, nucleic acid was isolated using the glass bead method and fractionated on agarose gels with 30 µg/ml chloroquine. DNA circles were detected by Southern blotting.
We have previously shown that TEF2-UAS was able to block the spread of silencing in a silencer-blocking assay (11). Such an assay tested if a sequence had the ability to prevent a silencer from silencing a reporter gene when it was positioned between the silencer and the reporter. In this report we used a new silencer-blocking assay to identify more barrier elements. This assay employed the URA3 gene and an inverted HML-I that could silence sequences to the right of HML (20) (Fig. (Fig.1,1, strain 1). URA3 expression makes cells sensitive to 5-fluoroorotic acid (FOA) thus URA3 silencing could be measured by cell growth on FOA medium (Fig. (Fig.1,1, strain 1). As expected, TEF2-UAS and another known barrier element, HMR-tRNA gene (21) exhibited barrier activity in this assay as indicated by the lack of silencing of URA3 in strains 2 and 29 (Fig. (Fig.11).
TEF2 encoding translation elongation factor 1α is one of the most highly expressed genes in yeast (27). Other highly expressed genes include ribosome protein genes (RPGs) and genes coding for glycolytic enzymes (27). The coordinated regulation of most of these genes requires the so-called general regulatory factors Rap1p, Abf1p and Reb1p (28,29). Using the above silencer-blocking assay, we tested if the UASs of other highly expressed genes could also block URA3 silencing (Fig. (Fig.1).1). Seventeen new barrier elements were identified from a total of 26 UASs tested. These 17 elements were UASs of eight RPGs (see Fig. Fig.1,1, strains 3–10), eight glycolytic genes (strains 18–25) and the HIS3 gene (strain 26). All 17 elements except ADH2-UAS and HIS3-UAS exhibited very strong barrier activity as evidenced by the absence of URA3-silencing in strains 3–10 and 18–24 (Fig. (Fig.11).
Seven of the eight RPG-UASs that had barrier activity, contained a single or a pair of tandem Rap1p-binding sites (Fig. (Fig.1,1, strains 3–9). Three of them also contained an Abf1p or Reb1p site (strains 6, 8 and 9). One barrier RPG-UAS (strain 10) had only a predicted Abf1p site. Of the seven RPG-UASs that had no barrier activity (Fig. (Fig.1,1, 11–17), only two bore no site for Rap1p, Abf1p or Reb1p (11 and 12). The remaining five each contained one to three Rap1p sites (strains 13–17). RPL15A-UAS (17) also contained an Abf1p site. For most of the 15 RPG-UASs tested (3–17), the presence or absence of predicted Rap1p sites correlated with the presence or absence of Rap1p association in vivo as previously examined by chromatin immunoprecipitation assays (13). Exceptions were the UASs of RPS24B, RPS28A and RPL15A (13).
Of the eight UASs of glycolytic genes tested (Fig. (Fig.1,1, strains 18–25), seven exhibited strong barrier activity (18–24) and one had weaker but detectable barrier activity (strain 25). Glycolytic genes are among the most highly expressed genes and their high expression depends on the functions of Rap1p, Abf1p or Reb1p as well as Gcr1p (29,30). The seven strong glycolytic barriers all consist of one or two Rap1p sites plus one or two Reb1p or Abf1p sites (Fig. (Fig.1,1, strains 18–24). In addition, one or more Gcr1p-binding sites are found adjacent to each Rap1p site. It was proposed that the function of Rap1p was to facilitate the binding of Gcr1p to the UAS (15). Abf1p or Reb1p was shown to play a role similar to that of Rap1p in activating glycolytic genes (31–33).
Although the above data reinforced the notion that certain Rap1p-binding UASs could serve as barriers to silencing, they raised new questions about the essential components of a barrier element. For example, some of the UASs that contained only a single Rap1p site had barrier activity (e.g. RPL19B-UAS, strain 3) whereas others that contained two or three sites had no activity (e.g. RPL39, strain 16). One explanation was that sequences flanking the Rap1p site(s) in a particular UAS could positively or negatively influence barrier function. Moreover, the presence of Abf1p or Reb1p sites in 11 of the 17 new barriers made us wonder if they also participated in barrier function. We began to address these issues by analyzing three new barrier elements (Fig. (Fig.2).2). RPS10A-UAS containing an Abf1p-binding site and a pair of Rap1p sites was divided into three fragments that were tested in a silencer-blocking assay (Fig. (Fig.2,2, strains 30–32). It was clear that only the fragment containing Rap1p sites functioned as a barrier (strain 31). TPI1-UAS bearing a Reb1p site and a Rap1p site was divided into two parts with one containing the Rap1p site (Fig. (Fig.2,2, strain 34) and the other Reb1p site (strain 33). Neither of these fragments alone retained barrier activity indicating that concerted action of Reb1p and Rap1p was required for barrier activity. Duplicating the Reb1p-bearing part of TPI1-UAS didn’t restore barrier activity (strain 35). However, triplicating the Rap1p-containing part did re-create barrier function (strain 36). RPS28A-UAS bore, in addition to a Rap1p site, an Abf1p site and an adjacent T-rich region that are required for efficient transcription of RPS28A (34). A sequence of RPS28A-UAS deleted for a fragment bearing the Rap1p site lost the barrier activity (Fig. (Fig.2,2, strain 37). However, duplicating this sequence restored barrier activity (strain 38). The above results indicated again that Rap1p sites could form barriers (strains 31 and 36). On the other hand, they also suggested that synergistic actions of Abf1p sites (strain 38), Reb1p site + Rap1p site (strain 18), or Abf1p site + Rap1p site (strain 16) could all lead to barrier function. This was not very surprising since Rap1p, Abf1p and Reb1p could perform similar and sometimes interchangeable functions in gene regulation (16,35,36). Note, however, we haven’t ruled out the possibility that other factors (e.g. Gcr1p) or structure features of DNA (e.g. T-tracks) were also involved in barrier function.
It was interesting that ADH2-UAS and HIS3-UAS bearing no site for Rap1p, Abf1p or Reb1p exhibited detectable barrier activity (Fig. (Fig.1,1, strains 25 and 26). However, binding sites for other transcriptional activators existed in these UASs. The major regulator that binds to ADH2-UAS is Adr1p (37). On the other hand, Gcn4p binds to multiple sites in HIS3-UAS and activates HIS3 transcription (38). The possible involvement of Adr1p or Gcn4p in barrier function was examined below. Note Adr1p function was subject to glucose repression (37), which might explain why ADH2-UAS only exhibited limited barrier activity since glucose was used in the media used in our assay.
Although we had shown that the three Rap1p-binding sites in TEF2-UAS were necessary and sufficient for its barrier activity (11), we had not shown if binding of Rap1p to these sites was required. To address this we performed mutational analysis of TEF2-UAS. As shown in Figure Figure3A3A and B, the three Rap1p-binding sites in TEF2-UAS (designated as R1, R2 and R3) were all able to bind Rap1p in an EMSA (Fig. (Fig.3B).3B). Single or double C→A mutations were introduced into R1, R2 and R3 resulting in m1, m2 and m3, respectively (Fig. (Fig.3A).3A). The m2 and m3 sites had both lost the ability to bind Rap1p, whereas m1 still bound Rap1p (Fig. (Fig.3B),3B), which could be predicted from the specific mutations in each site (Fig. (Fig.3A,3A, compare m1, m2 and m3 to the consensus sequence). We then tested a mutated TEF2-UAS containing m1, m2 and m3 in a silencer-blocking assay using the HML-E silencer and the HMLα genes (Fig. (Fig.3C).3C). In this assay, the HML-I silencer was deleted but the HML-E silencer was sufficient to silence the HMLα genes (Fig. (Fig.3C,3C, strain 39) (11). Silencing in strain 39 (a-type) was measured by its ability to mate with an α-type strain (Fig. (Fig.3C).3C). As predicted, the wild-type TEF2-UAS blocked HMLα silencing (11) (Fig. (Fig.3C,3C, strain 40). On the other hand, the mutated TEF2-UAS containing m1, m2 and m3 had lost barrier activity (Fig. (Fig.3C,3C, compare the mating efficiencies of strains 41 and 40). Therefore, mutations that prevented Rap1p from binding two of the three sites in TEF2-UAS abolished its barrier activity, indicating that association of Rap1p with TEF2-UAS at more than one site was required for barrier function.
In eukaryotic cells, the topology of local DNA reflects the chromatin structure in which it resides. We had previously developed a DNA topology-based assay to examine chromatin structure in vivo, which involved excising a chromosomal region of interest from its genomic location as a circle using site-specific recombination (39). Using this strategy, we demonstrated that DNA in silenced chromatin was more negatively supercoiled than that in active chromatin (39). In this report we used the topology assay to address if barrier elements altered local chromatin structure as Drosophila or chicken insulators did (40,41).
Various sequences (designated X) from UASs of TEF2, AgTEF (TEF gene from Ashbya gossypii) and ADE2 were inserted between HML-E and HMLα at HMLΔI (HML locus deleted for the I silencer) flanked by the FRT sites for the Flp1p site-specific recombinase (Fig. (Fig.4A4A and B). The FLP1 gene was under the control of the inducible GAL10 promoter (39). The mating efficiency of each strain measured silencing of HMLα (Fig. (Fig.4C).4C). In strains 40, 42, 44 and 47, silencing was restricted to the left of the X insertions and HMLα was derepressed (Fig. (Fig.4C).4C). Consistently, the proportion of silent chromatin in the region flanked by the FRT sites in each of these strains was decreased, as indicated by the reduced negative supercoiling of the excised circle (11) (data not shown). Note that the above effect of a barrier on HML chromatin reflected both change in the scope of silencing caused by the barrier activity and alteration (if any) in chromatin structure at the barrier independent of the silencing state of the locus. To exclusively examine the latter, we analyzed the topology of HML DNA in a silencing-deficient background. The SIR3 gene in strains described in Figure Figure4B4B was disrupted rendering them defective in silencing (strains 40′–48′ in Fig. Fig.4D).4D). Recombination mediated by Flp1p in these strains led to the excision of a group of chromosomal circles that differed only in the X insertion. The supercoiling of these circles was analyzed as a function of the length of the insertion (Fig. (Fig.4D).4D). If an insertion simply lengthened a circle without causing abnormal changes in nucleosomal structure and/or density, the topoisomers of this circle would migrate slower relative to those of a similar circle with a smaller size. Otherwise, migration of the topoisomers would deviate from that predicted based on the size of the circle. As shown in Figure Figure4D,4D, the circle from strain 40′ bearing a 104 bp barrier element migrated faster rather than slower than the circle from strain 46′ bearing a 91 bp non-barrier element. Circles bearing other barrier elements also exhibited unexpected faster migration (compare strains 44′ to 45′, 47′ to 43′, 42′ to 48′, respectively). Therefore, these data indicated that barrier function correlated with altered topology of local DNA. This was further demonstrated by that mutations in TEF2-UAS that abolished its barrier activity also abolished its effect on DNA topology (Fig. (Fig.4D,4D, compare 41′ to 40′; note that circles from both strains had the same size). Since in our gel assay more negatively supercoiled DNA migrate slower, we concluded that every fragment with barrier activity caused a reduction in negative supercoiling of approximately one to two turns in HMLΔI DNA (Fig. (Fig.4D,4D, compare the centers of distributions of topoisomers in pairs of lanes, e.g. 40′ and 41′).
The topology of eukaryotic DNA reflects mainly the wrapping of DNA into nucleosomes (approximately one negative superhelical turn is associated with each nucleosome) (42). Therefore, removal of one nucleosome would eliminate approximately one turn. On the other hand, histone acetylation can reduce negative supercoiling associated with a nucleosome by approximately 0.2 turns (43). Therefore, acetylating five nucleosomes would reduce negative supercoiling by approximately one turn. Consequently, the linking number change of one to two brought about by TEF2-UAS can be accounted for by either the loss of one to two nucleosomes or acetylation of five to 10 nucleosomes. Further experiments are under way to clarify the nature of change in chromatin induced by a barrier.
The multifunctional Rap1p can be divided into at least four functional domains (Fig. (Fig.5A)5A) (12). We were interested in defining which of these domains might be involved in the barrier activity of Rap1p. To this end, we fused various parts of Rap1p to the bacterial LexA protein and tested if targeting any of the fusion proteins to LexA-binding sites could re-create a barrier to silencing (Fig. (Fig.5A).5A). The LexA-RAP1 fusion genes were carried on 2 µm based plasmids marked by LEU2 (designated L1–L6, Fig. Fig.5A).5A). When introduced into strain 50 in which two LexA-binding sites had been inserted between the inverted HML-I and URA3, LexA alone had no effect on URA3 silencing (Fig. (Fig.5A,5A, plasmid L1). Targeting the DNA binding domain, DNA bending domain, silencing domain, or the activating + silencing domain of Rap1p also did not affect URA3 silencing (L2–L5). On the other hand, targeting the activation domain alone recapitulated the barrier activity of Rap1p (L6). These results indicated that the activation domain of Rap1p was involved in barrier activity. The inability of the Rap1p-silencing domain to block URA3 silencing in our assay did not necessarily mean that this domain had no barrier function, since such a function might be suppressed by much stronger silencing activity associated with the silencing domain.
Was the lack of barrier function of fusion constructs L2–L5 due to a lack of expression of these proteins? To address this possibility, we examined the levels of L1–L6 proteins as well as L7 and L8 proteins described in Figure Figure5B5B by western blotting using an antibody against LexA. As shown in Figure Figure5C,5C, L5 was expressed at a level comparable to L6, and L4 was expressed at a higher level than L6. Therefore, the lack of barrier function of L4 and L6 was not the result of their lack of expression. On the other hand, the level of L2 or L3 was significantly lower than that of L6, hence the lack of barrier function of L2 and L3 might be due to their insufficient expression.
As an important control for the above targeting experiments, we showed that plasmids L1–L6 (Fig. (Fig.5A)5A) as well as L7 and L8 (Fig. (Fig.5B)5B) bearing LexA-fusion genes had no effect on URA3 silencing in strain 1 (Fig. (Fig.1)1) in which there was no binding site for LexA present between HML-I and URA3 (data not shown). Hence, the effect of the fusion proteins on URA3 silencing was the result of their targeting to specific loci, not the consequence of non-specific, indirect functions (if any) of these proteins.
We have shown above that ADH2-UAS bearing Adr1p-binding sites and HIS3-UAS bearing Gcn4p sites both had detectable barrier activity (Fig. (Fig.1,1, strains 25 and 26). Given that tethering the activation domain of Rap1p is sufficient for barrier function, we tested if targeting the activation domain of Adr1p or Gcn4p could also establish a barrier. Adr1p contains four separate activation domains (TADI–TADIV) that can contact Ada2p, Gcn5p and/or TFIIB (45). TAD III (415–467) interacts with only Gcn5p but not Ada2p or TFIIB. We fused this domain of Adr1p and the activation domain of Gcn4p (102–149) to LexA, respectively (Fig. (Fig.5B,5B, L7 and L8). Targeting these fusion proteins to two LexA-binding sites between HML-I and URA3 in strain 50 dramatically reduced URA3 silencing (Fig. (Fig.5B,5B, compare L7 and L8 to L1). Therefore, like LexA-Rap1p (627–695), both LexA-Adr1p (415–467) and LexA-Gcn4p (102–149) also had barrier activity. Barrier function of these fusion proteins was significantly decreased in strain 51 in which only one LexA site was inserted between HML-I and URA3 (compare strains 50 and 51 carrying plasmid L7), indicating that the potency of such barriers was dependent on the amount of tethered fusion proteins.
We have shown in this report that UASs of many highly expressed genes could function as barriers to the spread of transcriptional silencing. These newly identified barrier elements together with previously described barrier elements (STARs, HMR-tRNA gene, TEF2-UAS and CHA1-UAS) (9–11,21) are scattered across the yeast genome and may play a role in dividing the genome into functional domains. Analyses of the barrier UASs suggest that binding sites for transcriptional regulators Rap1p, Abf1p, Reb1p, Adr1p and Gcn4p can participate in barrier function. Barrier activity often requires concerted actions of more than one factor-binding site. Our mutational analysis of TEF2-UAS suggests a direct involvement of Rap1p in its barrier function. It is noteworthy that not all UASs of highly expressed genes have barrier activity. Many non-barrier UASs also consist of multiple regulator binding sites, therefore other unknown factors might determine whether a UAS could act as a barrier.
How Rap1p and the other transcriptional regulators act to block silencing is not clear. Since they can all act as transcriptional activators, one natural explanation is that they may overcome the silencing effect of a silencer and directly activate the reporter gene. However, it has been well documented that Rap1p and the other transcriptional regulators usually act at relatively short distances (less than ~600 bp upstream of the translation start codon) (28,29,46), but the binding sites for these regulators were >800 bp from the translation start codon in most of the silencing-blocking tests in this report (Figs (Figs11 and and2).2). In the test shown in Figure Figure6,6, TEF2-UAS blocked ADE2 silencing when it was ~2 kb downstream from the translation start codon of the gene. Moreover, Fourel et al. (44) demonstrated that the barrier function (or anti-silencing function) of an activator was independent of its role in activating transcription. Therefore, it is unlikely that the ability of a UAS to prevent gene silencing is due to its direct activation of the gene. In light of this, our results are in agreement with the notion that transcriptional regulators can also participate in defining the boundaries of functional chromosomal domains.
A few models have been proposed for the barrier functions of transcriptional regulators. One model proposed that binding of a regulator to a barrier led to the formation of a nucleosome-free region thereby disrupting a continuous nucleosome array (11). Since the SIR complexes spread by self-interaction and interacting with adjacent nucleosmes in a stepwise fashion (4), a nucleosome-free gap could present a barrier to the spreading SIR complex. This ‘nucleosome gap’ hypothesis is consistent with that certain barrier factors such as Rap1p and Reb1p could indeed prevent nucleosome formation around their binding sites (14) and the barrier function of TEF2-UAS correlated with its ability to alter local chromatin structure (this report). In fact, many insulator/boundary elements in higher organisms have also been shown to alter local chromatin structure. For instance, the chicken β-globin insulator was linked to changes in nucleosome positioning, and the Drosophila scs and scs′ boundary elements are associated with unusual chromatin structures that are hypersensitive to nucleases (40,41). However, it is not clear whether any chromatin change associated with barrier function in yeast or insulator function in higher cells is the cause or effect of the function.
Another model for barrier action proposed by R. Kamakaka and colleagues (21) proposed that barrier factors recruit chromatin modifying and/or remodeling complexes to counteract the more compact, hypoacetylated silent chromatin. In agreement with this hypothesis, targeted HATs SAS2, GCN5 and ESA1 all had barrier activity (21; Y.-H. Chiu, Q. Yu and X. Bi, unpublished results). Interestingly, Rap1p, Adr1p, Gcn4p and Abf1p share the ability to recruit HAT complexes NuA4 and/or SAGA to target genes (16,45,47–49). Therefore, these factors that are capable of binding to the newly identified barriers may function as barrier factors by recruiting HAT complexes. In fact, recruitment of HAT activity has also been suggested to be the mechanism underlying the function of the chicken β-globin insulator (6), although it was not known what HAT might be involved. Our demonstration that tethering the activation domain of Rap1p, Adr1p or Gcn4p alone recapitulates barrier activity is consistent with the above HAT model for barrier function, since at least the activation domain of Adr1p has been shown to directly interact with HAT complexes (45). However, we haven’t ruled out the possibility that the barrier function of the targeted activation domain of Rap1p, Adr1p or Gcn4p is not related to the barrier function of the native protein. Note that the ‘nucleosome gap’ model and the HAT model are not mutually exclusive since the nucleosome gap could result from destabilization of nucelosomes caused by HAT function. A recent intriguing study showed that targeting proteins of the nuclear pore complex (NPC) could also establish a boundary for silent chromatin, and that tethering to NPC was required for boundary activity (50). This led to a revisit to the looping model for the function of boundary/insulator elements (51). However, an alternative explanation was that the NPC compartment could simply be rich in transcriptional activating factors or poor in silencing factors like the SIR proteins (50).
In summary, results from this and other studies indicate that factors previously identified as positive and/or negative regulators of gene activation could also function in demarcating distinct domains of gene activity. How these factors carry out their boundary function is not resolved but recruitment of other factors (e.g. chromatin modifying complexes) is likely to be involved.
After this work was completed and submitted for publication the first time, Fourel et al. (52) reported that targeted activation domains of Rap1p and Abf1p had insulating capacity.
We are grateful to Drs Susan Gasser and David Shore for their gifts of plasmids. We thank Xiaohong Ye for assistance. This work was supported by NIH grant GM 62484 to X.B. and a start-up fund from University of Nebraska-Lincoln, USA.