Chromatin insulators are defined as transcriptionally neutral elements that prevent negative or positive influence from extending across chromatin to a promoter. Here we show that yeast subtelomeric anti‐silencing regions behave as boundaries to telomere‐driven silencing and also allow discontinuous propagation of silent chromatin. These two facets of insulator activity, boundary and silencing discontinuity, can be recapitulated by tethering various transcription activation domains to tandem sites on DNA. Importantly, we show that these insulator activities do not involve direct transcriptional activation of the reporter promoter. These findings predict that certain promoters behave as insulators and partition genomes in functionally independent domains.
Silencing is a form of repression that extends to large chromosomal segments and affects many types of DNA transactions, including transcription, recombination, repair and transposition. In Saccharomyces cerevisiae, it is found at silent mating‐type loci (HM), within rDNA repeats and in telomere proximity (TPE, telomere position effect). Silencing is generally associated with an overall loss of DNA accessibility, histone hypoacetylation, a regular spacing of the nucleosomal array and a shift in DNA topology (Lustig, 1998). All of these properties suggest that silent chromatin is organized into a ‘closed‐up’, relatively uniform nucleosomal array. Silent chromatin appears to nucleate at cis‐acting elements known as silencers and to propagate along the chromosome from this initiation site (Locke et al., 1988; Renauld et al., 1993). However, recent observations demonstrate discontinuity in the ‘spreading’ of silencing, which probably involves relay elements (Fourel et al., 1999; Talbert and Henikoff, 2000). This rather argues in favour of a ‘coalescence’ model of silencing in which silencers and relay elements form complex interactions to lead to an apparent spreading of silent chromatin (Talbert and Henikoff, 2000).
Chromosomal regions can be protected from silencing, and also from the influence of positive DNA elements such as enhancers, by cis‐acting elements called insulators (Sun and Elgin, 1999). They have been described in a variety of higher eukaryotes, and more recently in yeast (Bi and Broach, 1999; Donze et al., 1999; Fourel et al., 1999). Insulators have been uncovered at the borders of both active and inactive chromosome segments, in agreement with a role in delimiting independent chromatin domains (Donze et al., 1999; Saitoh et al., 2000). Abutting the telomeric arrays of S. cerevisiae are short sequences that can protect a reporter gene placed on their centromere side from TPE (Fourel et al., 1999). Accordingly, they have been considered as insulators, although their enhancer‐blocking capacity was not addressed. They are called STARs (for subtelomeric anti‐silencing regions) and are part of longer repeated elements belonging either to the X‐ or Y′‐family (Louis, 1995). STARs have been proposed to delimit a very terminal telomeric compartment protected from influences of the rest of the genome (Fourel et al., 1999). However, STARs do not affect the interaction of telomeres with subtelomeric silencing elements naturally found on their centromere side, which relay silent chromatin propagation into more internal regions of the chromosome (Fourel et al., 1999; Pryde and Louis, 1999). The Tbf1p and Reb1p proteins probably mediate STAR activity. Indeed, both X‐ and Y′‐STAR harbour multiple binding sites for Tbf1p and Reb1p, and multimers of oligonucleotides containing either Tbf1p‐ or Reb1p‐binding sites display anti‐silencing properties (Fourel et al., 1999).
Amazingly little is known about the molecular mechanism(s) by which insulators function, and no single model satisfies all observations (Sun and Elgin, 1999). Yeast offers unique possibilities as an amenable system to screen for either the expression or the silencing of a variety of reporter genes. We therefore characterized STARs more thoroughly and devised two assays that define the insulator function in S. cerevisiae. Tbf1p and Reb1p were found to recapitulate these properties, as well as different transcription activation domains upon tethering to DNA. However, insulator activity clearly did not involve transcriptional activation of reporter genes.
Results and Discussion
STAR as boundary elements
A strain was first generated that carries a STAR derived from a subtelomeric Y′ element (Y′‐STAR) (Fourel et al., 1999) placed between two reporter genes, URA3 and TRP1, at a fragmented telomere VII‐L as shown in Figure 1A (GF98). A 120 bp fragment containing four tandem binding sites for the transcription factor Gal4p (4×UASg) was also interposed between the two genes, and a strain was constructed that carried only this fragment, as a control (Figure 1A, GF97). The 4×UASg fragment behaves neutrally with regard to silencing in this setting (not shown). The fraction of URA3 OFF colonies, as assessed by growth on synthetic medium containing 5‐fluoro‐orotic acid (5‐FOA) (Boeke et al., 1984), barely decreased upon Y′‐STAR insertion (Figure 1A, compare lines 1 and 4). In contrast, TRP1 was efficiently protected from silencing, as revealed by an increased number of colonies on a synthetic complete medium lacking tryptophan (SC‐W) (Figure 1A, compare lines 2 and 5). These results have two important implications. First, telomere silencing capacity per se is not altered by a neighbouring Y′‐STAR. Secondly, only a gene located in a distal position, outside the STAR/telomere path, is efficiently protected from silencing. Furthermore, not a single colony emerged on SC‐W+5‐FOA from GF97 cultures (Figure 1A, line 3), showing that the 0.1% fraction of colonies in which TRP1 is ON corresponds to colonies in which URA3 is also expressed. In contrast, the presence of a Y′‐STAR between TRP1 and URA3 in strain GF98 allowed expression of TRP1 in ∼1% of the URA3 OFF cells (Figure 1A, compare lines 4 and 6). Thus, a Y′‐STAR can apparently constrain the propagation of silent chromatin to a telomere‐proximal domain and decouple the silencing status of adjacent genes. It thus phenotypically behaves as a boundary to the spreading of silencing, as schematized in Figure 1B, like insulators described in higher eukaryotes. Similar results were obtained at the silent mating type locus HML (see Supplementary data).
STARs permit discontinuous silencing
We showed previously that when two STARs bracket a reporter gene, its expression is no longer influenced by surrounding silencing elements (Fourel et al., 1999). For instance, in strain GF61 (Figure 1C), TRP1 is protected by STARs not only from telomere‐driven silencing, but also from the influence of the subtelomeric silencing element core X, which induces downstream URA3 silencing in 50% of cells (Figure 1C, lines 7 and 8). In addition, expression of the two genes appears decoupled as all the URA3 OFF cells have the TRP1 gene ON (Figure 1C, lines 8 and 9). Strikingly, this decoupling was also observed when only a STAR, but not a 4×UASg fragment was inserted between URA3 and TRP1 (Figure 1C, lines 3 and 6). In addition, TPE strengthening conferred by deletion of the RIF1 gene (Hardy et al., 1992) emphasized this effect (Figure 1C, lines 13–15). Thus, synergy between the two STARs bracketing TRP1 in strain GF60Δrif1 essentially maintained its expression while allowing predominant repression of URA3. Analogous discontinuity in the ‘spreading’ of silencing was recently reported in Drosophila (Talbert and Henikoff, 2000).
Altogether, these experiments strongly suggest that two distant STARs cannot only organize an insulated domain, as previously demonstrated for boundary elements in different organisms (Pikaart et al., 1998; Sun and Elgin, 1999), but can do so without seriously affecting efficient propagation of silent chromatin beyond this domain, as summarized in Figure 1D. The capacity to generate an apparent discontinuity in silent chromatin propagation, together with the boundary activity, operationally define the STAR insulator functions.
Tbf1p and Reb1p fulfil insulator functions
We then devised Gal4‐based tethering assays in order to identify proteins that can recapitulate the insulator functions of STARs (strains GF97 and GF100Δrif1; Figure 2). The Tbf1p and Reb1p proteins probably mediate the activities of natural STAR elements (Fourel et al., 1999). Chimeras were therefore constructed in which the DNA binding domains of Reb1p and Tbf1p are replaced by that of Gal4p (Gal4p DNA binding domain or GBD). Expression of GBD alone had very little influence on the overall silencing pattern when four binding sites for Gal4p (4×UASg) are located in the silent region (strain GF97; compare Figures 1A and Figure 2A, lines 1–3). In contrast, expression of the Tbf1p– or Reb1p–GBD chimeras imparted a prominent boundary activity to the 4×UASg fragment, but had no effect on reporter gene expression in the absence of UASg. It is even more pronounced than that of Y′‐STAR in the case of Reb1p with 20% of the URA3 OFF colonies expressing the distal TRP1 gene (compare Figure 2A, lines 4, 6, 7 and 9 with Figure 1A, lines 4 and 6). Only a very modest decrease in the fraction of URA3 OFF cells occurred upon expression of the Tbf1p and Reb1p chimeras in GF97, which is comparable to that observed in the presence of an intervening STAR in GF98. Thus, tethering the N‐terminal moiety of Tbf1p and Reb1p mimics the boundary effects of a Y′‐STAR. In addition, expression of the Tbf1p and Reb1p chimeras in a strain deleted for RIF1 (GF100Δrif1) induced an apparent discontinuity in the propagation of silent chromatin (Figure 2B). While expressing the Tbf1p–GBD protein did not significantly modify the ratio of TRP1 ON and URA3 OFF cells as compared with the expression of GBD alone, this thoroughly decoupled the two reporter genes as all the TRP1 ON cells are now also URA3 OFF (Figure 2B, compare lanes 1–3 and 4–6). In addition, expression of the Reb1p–GBD protein induced a silencing pattern superposable on that of a strain carrying an X‐STAR instead of the UASg sequences (Figure 2B, compare lanes 7–9 and Figure 1C, lanes 19–21). This shows that the tethering of Reb1p is also able to provoke a discontinuous silencing effect. Altogether, these results confirm the capacity of Tbf1p and Reb1p to recapitulate STAR activities, and demonstrate that their Myb‐related DNA binding domains are not directly involved or at least can be swapped with that of Gal4p. They further validate the use of the boundary and discontinuity tethering assays to address the capacity of a variety of GBD chimera to recapitulate insulator activities.
Transcription factors as potent insulators
Because Reb1p is thus far known as a transcription factor and Tbf1p binding sites are found in the upstream regions of several yeast genes (Koering et al., 2000 and references therein), we wondered whether independently characterized transcriptional activation domains would behave as STAR insulators. A chimera containing the full‐length proline‐rich domain of CTF displayed a strong boundary activity upon expression in GF97, allowing ∼10% of the URA OFF colonies to grow on SC‐W+5‐FOA while only mildly affecting URA3 silencing (Figure 3, lines 10–12). This figure decreased to 2% upon expression of derivatives carrying either only the very short, 13 amino acid, transforming growth factor‐β‐responsive domain (TRD) (Alevizopoulos et al., 1995) or its complementary deletion mutant (Figure 3, lines 15 and 18). In contrast, essentially all the URA3 OFF colonies also had the TRP1 gene repressed upon expression of the GBD control proteins (Figure 3, lanes 3 and 30) or of the fusions carrying the glutamine‐rich activation domains of Oct1, Oct2 and Sp1 (Figure 3, lanes 6 and 9, and not shown). We also assayed the acidic transcriptional activation domain of the breast cancer protein BRCA1. This chimera displayed the strongest boundary activity among all those assayed (Figure 3, lines 31–33), whereas derivatives having partially or totally lost BRCA1 chromatin remodelling capacity (Hu et al., 1999) still proved active, but less impressively (Figure 3, lines 36 and 39, and not shown). In addition, expression of a hybrid containing the acidic activation domain of VP16 essentially abolished any silencing of both URA3 and TRP1 (Figure 3, lines 19 and 20). We therefore could not really address decoupling capacity in this context. However, truncated versions harbouring either the N‐ or C‐terminal portion of VP16 activation domain, although still largely affecting URA3 silencing, allowed 5 and 1% of the remaining 5‐FOA‐resistant colonies to grow on SC‐W+5‐FOA, respectively. Of note, none of these chimeras had any effect on reporter gene expression in the absence of UASg. The principal conclusion from these experiments is that the boundary activity of STARs can be recapitulated by tethering transcription activation domains to tandem UASg sites. In addition, proline‐rich and acidic activation domains, but not glutamine‐rich activation domains, function efficiently.
We also tested the ability of the same GBD chimera to induce a discontinuous silencing effect as assayed for Tbf1p and Reb1p chimeras in strain GF100Δrif1 (Figure 2B). Chimeras that displayed boundary capacity (Figure 3) also recapitulated silencing discontinuity and the order of efficiency was similar in both assays (see Supplementary data). These data strongly support the view that the boundary and discontinuity assays of insulator functions actually reflect related molecular phenomena.
Insulator function does not correlate with promoter transactivation
We then addressed whether the activities of reconstituted STAR sequences may in fact rely on direct transcriptional activation of the reporter genes. In a first approach, the capacity of the GBD chimeras to directly activate the GAL1 promoter was measured. There was seemingly no general correlation between the insulator activities and the cis‐activation capacity displayed by the various chimeras (not shown). In particular, the Tbf1p and Reb1p chimeras had no effect.
We further reasoned that if the boundary property imparted to UASg by tethered transcription activation domains is specific to silent chromatin, then reporter genes should not be influenced by these elements in any manner upon abrogation of silencing. Reporter transcript levels were therefore analysed in GF97′ (which has a modified telomere VII‐L identical to GF97, but in a different genetic background; see Methods) upon deletion of the SIR4 gene (Figure 4A). The levels of TRP1 transcripts normalized to CDC28 used as an internal loading control are represented in Figure 4B. In SIR4 cells, the relative TRP1 mRNA levels paralleled the proportion of colonies growing out from the same culture on SC‐W medium, although RNA analysis appears less sensitive (Figure 4B, odd lines, compare with Figures 2 and 3). In contrast, in the absence of silencing, the steady‐state level of TRP1 mRNA appeared elevated and approximately uniform irrespective of the GBD chimera expressed (Figure 4B, even lines). Therefore, none of these proteins seems capable of directly activating the TRP1 promoter from the distant UASg tethering sites in GF97′, in the absence of silencing. We conducted the same type of analysis in a strain in which URA3 is separated from telomere repeats by UASg sites (GF95; Figure 4C and D). The URA3 mRNA level was raised by less than one log upon expression of Reb1N and CTF chimeras, and by one and a half log upon expression of VP16 chimera, confirming the anti‐silencing properties of these fusion proteins. Importantly, no significant variation in URA3 mRNA levels was observed upon expression of GBD or of the various chimeras in cells deleted for SIR4 (Figure 4D, compare even lines and line 9). Thus, the various transcriptional activation domains here tethered through GBD appear unable to activate either TRP1 or URA3 promoter in these experimental settings. Of note, analysis of URA3 mRNA levels in the absence of silencing in GF97′ confirmed that the VP16 chimera can activate the URA3 promoter from a proximal targeting site (not shown), which probably accounts for the apparent retrograde anti‐silencing effect observed with VP16 chimera and its derivatives (see Figure 3).
One may object that a low, hardly observable increase in transcription might suffice to account for apparent boundary effects. We therefore decided to analyse strains in which a prominent anti‐silencing effect could be observed. URA3 appeared fully protected from E silencer‐driven silencing at the HML locus upon interposing natural STARs or reconstituted, particularly potent STARs made of Tbf1p‐ and Reb1p‐binding sites (Fourel et al., 1999; see also Supplementary data). When yeasts were grown on synthetic complete medium (SC), URA3 levels were approximately the same for all the strains analysed, except for the GF20 control strain in which URA3 expression was one log below (Figure 5A and the right part of B, odd lines). This strain does not contain any STAR sequence at HML, and is indeed the only one that displayed significant silencing of URA3 (Figure 5B, left part, line 5). In the absence of silencer, even potent STARs did not alter the levels of URA3 transcripts. Of note, the position of the URA3 transcription initiation site was not modified either by proximal STARs (see Supplementary data).
One may further object that in this experiment, a putative transcription activation capacity of STARs may have been conceivably masked by limitations on the maximal transcription initiation rate at the URA3 promoter. We therefore grew exactly the same strains on a medium lacking uracil, which activates the URA3 promoter ∼2‐fold (Figure 5, compare SC and SC‐U). Again, stimulated URA3 levels were uniform among all the strains analysed, except for GF20, for which growth on SC‐U seemingly resulted in a partial alleviation of silencing and/or the selection of cells expressing a threshold URA3 level (Figure 5A and B right panel, compare SC and SC‐U, lines 17 and 18). Overall, we conclude that STARs capable of fully protecting reporter genes from silencing at HML had absolutely no influence on the transcription of the reporter gene in the absence of silencing.
Altogether, these findings are in agreement with a model in which the insulator effects cannot be accounted for by direct, classical stimulation of the reporter promoters, but rather by phenomena specifically affecting silent chromatin. Alternatively, direct transcriptional activation of reporter genes through direct contacts between the promoter and remote transactivators may be envisaged to require functional silent chromatin and would thus not be observed in Δsir4 strains or at the HML locus deprived of silencers. Silent chromatin was indeed shown previously to bring together distant sequences and may thus facilitate the formation of an intervening loop (Strahl‐Bolsinger et al., 1997). However, the establishment and/or the maintenance of the loop owing to silent chromatin is not expected to be compatible with the efficient transcription of the underlying gene in subtelomeric constructs, and this model thus appears rather unlikely.
Collectively, these results demonstrate that a variety of transcription activation domains tethered to tandem DNA sites can reconstitute bona fide insulators in S. cerevisiae. Thus, the structure of insulators does not follow strict organization, nor do they necessarily involve specialized proteins. The fact that many different transcription factors operate implies that distinct, possibly redundant mechanisms can underlie insulator activity. Candidates are all the modifications of chromatin and/or of the higher order organization of the nucleus that are incompatible with silencing. Our results further suggest that promoters may behave as insulators and partition genomes in functionally independent domains.
Most of the yeast strains described in this study contain a modified VII‐L telomere, are derivatives of W303‐1a and were obtained following standard genetic manipulations as described in Fourel et al. (1999). The variegated expression of URA3 and TRP1 was monitored as described in Fourel et al. (1999). GF95 and GF97′ were constructed in an FYBL2‐5D background (MATα URA3‐Δ851 TRP1Δ63 leu2Δ1) in order to allow RNase protection analysis of URA3 and TRP1 reporter transcripts, and further carried a HIS3 deletion. Deletions in the RIF1 and HIS3 genes were obtained by replacement of the full‐length open reading frames (ORFs) by the KanMX cassette. The SIR4 gene disruption used pJR276 (Kimmerly and Rine, 1987). Yeast strains harbouring a modified HML locus have an S150‐2B background (Mata, leu2‐3‐112, ura3‐52 trp1289 his3Δ gal2 gal4:LEU2) and were obtained as described in Fourel et al. (1999).
All the DNA fragments assembled to generate telomeric constructs are described in Fourel et al. (1999). Fragments obtained by HiFi PCR (Boehringer Manheim) of S. cerevisiae genomic DNA and corresponding to the 405 N‐terminal residues of Tbf1p and Reb1p were introduced in GBTh to generate Tbf1p and Reb1p chimeras. GBTh itself was contructed by replacing the TRP1 selection marker of GBT9 (Stratagene) by HIS3 and harbours Gal4p residues 1–147 expressed from the ADH1 promoter. Chimeras labeled GBD′ contain Gal4p residues 1–93 fused to the indicated protein domains and are expressed from the ADH1 promoter on a pBRHAC‐derived low‐copy vector (Künzler et al., 1994). They were kindly provided by W. Schaffner, except for GBD′–TRD and GBD′–CTFN, which were PCR generated using GBD′–CTF as a template. The BRCA1 chimeras harbour Gal4p residues 1–94 (GBD°) and are expressed from the CUP1 promoter of constructs stably integrated at the LEU2 chromosomal locus, in the presence of 100 μM CuSO4 (Huet al., 1999). Expression of the Gal4p chimera was checked by immunobloting using an anti‐Gal4 antibody (Santa Cruz).
Supplementary Figure 1
Supplementary Figure 2
Supplementary Figure 3
Supplementary Figure 4
We thank E. Louis, W. Schaffner and M. Charbonneau for providing the FYBL2‐5D strain, the GBD′ series of plasmids and pSFcdc28, respectively. This work was supported by La Ligue Nationale contre le Cancer (E.G. laboratory), the National Institute of Health (R.L. laboratory) and the Swiss National Science Foundation (N.M. laboratory). Correspondence and requests for materials should be addressed to G.F.
- Copyright © 2001 European Molecular Biology Organization