The boundaries of the silenced HMR domain in Saccharomyces cerevisiae

Results

The silenced domain emanates bidirectionally from the silencers

The silent HMR domain (Fig. ) is refractory to digestion by various restriction endonucleases in wild-type cells but is accessible to these enzymes in sir3 mutants (Loo and Rine 1994) (see Fig. B). This inaccessible domain (pink box in Fig. ) is not limited to the region between the two silencers but extends several hundred base pairs beyond the silencers.

We employed an in vivo assay to refine the boundary of the silenced domain and to test how these boundary elements might function. A set of isogenic strains was created in which the URA3 gene was inserted at three specific sites on chromosome III, at varying distances from the silencer (Fig. ). Transcriptional repression was measured by the degree of silencing of the URA3 gene.

Placing the URA3 gene between the two silencers (∼640 bp to the right of HMR-E) led to near complete repression of the gene. When placed 475 bp to the left of HMR-E, URA3was also repressed, whereas when placed 2840 bp to the left ofHMR-E, URA3 was derepressed completely. Furthermore, transcriptional repression was SIR3 dependent (cf.SIR with sir3Δ in Fig. ). These results support the previous conclusion about the silenced domain extending beyond the silencer and provide a convenient genetic assay for factors or mutants affecting the limits of the silenced domain.

The silenced HMR domain could be doubled in size

The heterochromatic domain at HMR spans ∼3.5 kb of DNA. One hypothesis for the size of the silenced domain is that the pools of Sir proteins in the cell limit its size (Renauld et al. 1993). We tested this model by determining whether an increase in the distance between HMR-E and HMR-I would lead to concomitant derepression of HMR and loss of the silenced domain.

Strains were constructed wherein the HMR domain was increased from 3.5 to 4.5 kb or 7 kb, by inserting either the TRP1 gene alone, or TRP1 with one copy of an unrelated fragment of yeast DNA from the coding region of the HMG2 gene. Examining the expression of the TRP1 gene monitored silencing in these strains. TRP1 remained repressed even when the domain was expanded from 3.5 to 7.0 kb (Fig. A). This repression was dependent on the Sir proteins, suggesting that the size of the silent domain was not due to limiting pools of the Sir proteins, as it can be increased.

To confirm that the repression observed in vivo was due to the expansion of the silenced domain we utilized a differential restriction endonuclease accessibility assay (Loo and Rine 1994). In this assay, nuclei isolated from wild-type and sirΔ strains were partially digested with various restriction enzymes to assay the accessibility of the DNA to these enzymatic probes. Following digestion, the DNA was analyzed by DNA blot hybridization as described (Loo and Rine 1994). Mapping the expanded domain using the differential restriction endonuclease accessibility assay demonstrated that the entire expanded domain was inaccessible to the enzymes in wild-type cells but not sirΔ cells (Fig. B), confirming the observations made in vivo with expression of the TRP1gene. Interestingly, results from the restriction endonuclease assay also demonstrated that the expanded domain still terminated at the same DNA sequence as at the wild-type HMR locus (cf. Fig. B, top and bottom). These results suggest that the silent domain was restricted in space by DNA elements that flank the HMR locus.

Deletion of the right boundary leads to a spread of silencing

Because the boundaries of the silent domain were unchanged even when its size was doubled we speculated that the presence of a boundary element would restrict the spread of the silenced domain, and conversely, deletion of this element would lead to the spread of the domain.

Previous studies implicated a 500-bp region to the right ofHMR-I as a boundary of the silenced chromatin (Loo and Rine 1994). To test whether deleting the boundary element led to a spread of silenced chromatin, we generated strains containing an insertion of theURA3 gene ∼1420 bp to the right of HMR-I. The putative right boundary (∼1 kb) was either left intact or was replaced with a 1-kb fragment of unrelated DNA. Additionally, we overexpressed SIR3 in these strains to see whether this overexpression led to the expansion of the silenced domain, as has been observed at telomeres (Renauld et al. 1993). The effect of deleting this putative boundary element was analyzed by monitoring the expression of the URA3 gene (Fig. ). In the presence of an intact boundary element, the URA3 gene was insulated from the repressing chromatin and was active (Fig. ). However, when the putative element was deleted, the URA3 gene was significantly repressed, indicating that the silenced chromatin had expanded to repress the URA3 gene. The silencing of theURA3 gene was dependent on the normal mechanisms that function at HMR, as it required the presence of the Sir proteins.

Boundary elements block the spread of silenced chromatin

A key characteristic of a boundary element is the ability of these elements to confer position-independent expression on a gene by blocking the spread of silenced chromatin. We therefore asked whether the boundary element could block the spread of silencing when inserted between a silencer and a gene promoter.

In the first series of experiments (Fig. A), we inserted DNA fragments containing either the left or right boundary elements or a similarly sized stuffer fragment in both orientations between the HMR-E silencer and the MAT a 1gene at the wild-type HMR locus on a plasmid. To determine whether the putative boundary elements were capable of blocking the spread of silencing from the HMR-E silencer, we monitored expression of the MAT a 1 gene by a mating assay. In a wild-type MATα cell, the MAT a 1 gene atHMR is repressed and the MATα cell is capable of mating with a MAT a cell, giving rise to diploids. However, if the MAT a 1 gene at HMR is derepressed, theMATα cell is unable to mate and fails to form diploid colonies. Monitoring the expression of the MAT a 1 gene indicated that the gene remained repressed in all of the cases analyzed. This result indicated either that HMR-I was sufficient to silence MAT a 1, or neither boundary was capable of blocking repression from HMR-E.

Previous experiments established that silencing of HMR does not require HMR-I in strains with a fully functionalHMR-E (Brand et al. 1985). Therefore, in the second set of experiments, we deleted HMR-I such that silencing could initiate only from the HMR-E silencer. Inserting the left or right boundary elements between the HMR-E silencer and theMAT a 1 gene resulted in the expression of theMAT a 1 gene, suggesting that both elements blocked the spread of silencing (Fig. ; bottom). In contrast, the stuffer fragment was unable to block the spread of silencing. Further analysis indicated that the right boundary fragment was more proficient at mediating boundary function than was the left fragment (Fig. , bottom). Potentially, this difference in effectiveness could result from either inefficient insulator function of the left boundary, or perhaps the fragment used did not encompass the entire boundary element.

The HMR boundary element blocked the spread of telomere position effect

We tested whether the right boundary element could function to block the spread of telomere position effect. S. cerevisiaetelomeres possess silenced chromatin, which emanates from the TG1–3 repeats and spreads for ∼4.5 kb. The URA3 gene was silenced when inserted close to the telomere on chromosome 5 as reported previously (see Fig. A; Gottschling et al. 1990). We inserted the right boundary element between the telomere and theURA3 gene and analyzed URA3 expression. The presence of the boundary element in either orientation blocked the spread of silenced chromatin, leading to the expression of the URA3 gene, whereas insertion of a stuffer fragment of the same size did not (Fig. A).

The boundary element did not function in trans

In Drosophila it has been reported that insulator activity can function in trans such that the Su(Hw) protein bound to the gypsy insulator on one chromosome can act intrans to inactivate enhancers located in the other homolog. We therefore determined whether the boundary element identified in S. cerevisiae was also capable of functioning in trans in a diploid cell.

To perform this analysis, expression of the URA3 gene located adjacent to the HMR locus was monitored in three diploid strains. In one of these strains (ROY938) the right boundary element was deleted in both HMR alleles, whereas the second strain (ROY935) was heterozygous for the right boundary element, carrying one wild-type allele of HMR and one allele with the boundary deleted. In the third strain (ROY941), both alleles ofHMR–URA3 had the wild-type boundary element. If the boundary element was capable of functioning in trans, the URA3gene in strain ROY935 would be insulated from HMR and thus be active. However, as is evident in Figure B, the URA3 gene in ROY935 is significantly repressed in a manner similar to strain ROY938. These results indicate that in S. cerevisiae the boundary elements are unable to function in trans to insulate the reporter gene.

A deletion analysis of the right boundary element

Because the right boundary element identified above resides in a 1-kb fragment of DNA, we decided to perform a deletion analysis of this fragment to identify the minimum fragment that is necessary for boundary function. The results of this analysis are presented in Figure. The analysis was performed by inserting DNA fragments, derived from the 1-kb right boundary element, between theHMR-E silencer and the MAT a 1 gene and monitoring expression by the mating assay (described in Fig. ). The 1-kb right boundary fragment contains a Ty1 LTR and a tRNA gene. A systematic deletion analysis of the element from one end indicated that a DNA fragment containing just the Ty1 LTR (pRO367) had some boundary function but was not sufficient for mediating robust boundary function. In addition, the 5′ end of the LTR alone (pRO273) did not have any boundary activity on its own.

A systematic deletion from the other end of the 1-kb DNA fragment indicated that deletion of the LTR did not result in complete loss of boundary function, whereas deletion of a fragment of DNA encompassing the tRNA gene caused a significant loss of boundary element function (cf. pRO269 to pRO271). Further deletion analysis indicated that a small DNA fragment (<400 bp) lacking the LTR but encompassing the tRNA gene (pRO370) possessed significant boundary element function. These data suggest that robust boundary element function was mediated by a combination of both the LTR-containing fragment and the tRNA-containing fragment.

Extragenic mutations that affect boundary function

In Drosophila a multiprotein complex mediates insulator function. We therefore also began to look for mutations in genes that would lead to loss of boundary element function. Because the 1-kb right boundary element encompassed an LTR element, we tested mutations identified previously as being necessary for LTR function for their effects on boundary function (Fig. ). spt3,spt4, and spt8 mutants are defective in LTR function, but these mutations had no major effect on boundary function. We also analyzed boundary function in cells carrying mutations in proteins that have putative binding sites in the 1-kb boundary fragment—specificallytec1, mig1, pho4, and gcr1 (data not shown). Because Gcr1p has been shown to interact with Rap1p we also tested mutations in RAP1. However, individual mutations in these proteins did not result in loss of boundary element function. Then, we tested mutations in chromatin assembly proteins—cac1, hat1, and rlf6 and found that single mutations in these proteins also had no effect on boundary function. Insulator elements have been suggested to function by anchoring DNA loops to the chromosomal scaffold. We therefore tested several mutants defective in higher order chromosome structure—smc1, smc3 (data not shown), mcd1, pds1, andtop2. Interestingly, of these mutants, only loss of SMC1 andSMC3 disrupted boundary function significantly . It is not clear whether this effect is allele specific or strain specific as only smc1-2 but not smc1-259 affected boundary function (data not shown). These data implied a role for a structural protein of chromosomes in boundary function.

Discussion

Studies on the silenced HMR domain have shown that silencing is mediated by a repressive chromatin structure that covers the entire locus and extends beyond the silencers (Singh and Klar 1992;Loo and Rine 1994). Incidentally, the limits of the silent domain appear to coincide with previously mapped DNase I hypersensitive sites (Nasmyth 1982). Our study identified DNA segments in S. cerevisiae that restricted the spread of silenced chromatin and functioned as boundary elements of chromatin domains. We identified boundary elements by virtue of the spread of silenced chromatin in their absence and by their ability when inserted between a silencer and a reporter gene to block expression of the reporter.

The native HMR domain spans ∼3.5 kb of DNA. The relatively small size of this domain was not due to limiting amounts of the Sir proteins, as the domain could be expanded to two times its normal size and still remained repressed. These results were consistent with Ty insertions into HML that expanded that locus to 14 kb yet did not interfere with silencing (Mastrangelo et al. 1992). Biochemical analysis using a differential restriction endonuclease accessibility assay confirmed the observation that the entire expanded domain was inaccessibile to digestion by various restriction enzymes. Interestingly, in the expanded domain the boundaries of the domain remained constant with respect to the underlying DNA sequence. Taken together, these observations suggest the existence of boundary or insulator elements that restrict the spread of silenced chromatin.

We hypothesized that sequence near the border between silenced and nonsilenced domains would include a boundary element. We deleted a presumptive boundary element and discovered that neighboring reporter genes became inactive, presumably because of the spread of the silent chromatin from HMR. Previous studies of chromosomal translocations and PEV in Drosophila led to the suggestion that boundary elements limited the spread of inactive chromatin (Kellum and Elgin 1998). The results of this deletion analysis provided direct support for this model.

The boundary element interfered with silencer–promoter interactions in a position-dependent manner. The boundary appeared to function only when inserted between a silencer and a promoter. When two silencers flank a reporter, inserting the boundary element between one of these silencers and the promoter was not sufficient to insulate the promoter. The boundary element function in S. cerevisiae was position dependent with respect to the silencers and promoters. This behavior was similar to insulator elements from larger eukaryotes that are able to confer position-independent expression of the reporter gene only when they bracket the gene (Kellum and Schedl 1991; Roseman et al. 1993) and also disrupted the action of an enhancer only when positioned between the enhancer and the promoter. Our results also suggest that the primary function of the HMR boundary element was to limit the activity of silencers to specific domains of the chromosome.

Studies in Drosophila have also shown that boundary elements are not enhancer- or promoter-specific (Gdula et al. 1996). Similarly, we have found that the S. cerevisiae boundary element could insulate a reporter gene (URA3) from silenced chromatin at both the HMR locus and the chromosome VR telomeric locus. It will be of interest to determine whether any of the native telomeres are flanked by insulator elements and whether insulator elements are present at other loci in the cell.

An analysis of the sequences of the left and right boundary elements indicates that both elements contain a LTR. However, the results also indicate that the left element is not very efficient at boundary function, whereas a deletion analysis of the right element suggests that the LTR within this fragment on its own is also a poor boundary element. These results suggest that although the LTR contributes to boundary function, it is not sufficient. Further deletion analysis of the right boundary element demonstrated that deletion of the tRNA gene within this fragment resulted in a significant loss of boundary function. These data together suggest that the robust nature of the right boundary was probably due to a combined action of the Ty1 LTR and the tRNA gene, as both of these elements can function individually to varying degrees as boundary elements.

The correlation between Ty elements and boundary function, although surprising, was not unexpected, as in Drosophila thegypsy retrotransposon contains an insulator function in vivo (Corces and Geyer 1991). At this point we do not know whether all classes of Ty elements affect boundary function in S. cerevisiae.

The demonstration that a tRNA gene fragment acts as a boundary element is intriguing because it has been been demonstrated previously that a tRNA gene located at the HMR locus becomes silenced (Schnell and Rine 1986). In a second study, the insertion of a tRNA gene upstream of a RNA polymerase II (Pol II) promoter caused the repression of the Pol II promoter (Hull et al. 1994). However, we find that the insertion of a DNA fragment containing a tRNA gene results in the disruption of silenced chromatin and the consequent activation of the Pol II reporter gene (MAT a 1 or URA3). Further studies should help resolve this paradox. It has been shown previously that transposable elements preferentially integrate adjacent to tRNA genes and the integration is mediated by interactions between the proteins bound to the LTR and the tRNA locus (Voytas and Boeke 1993). The boundary function observed in our study could therefore be due to a cooperation between the proteins at the LTR and the tRNA, which could form a large multiprotein complex that would act as a barrier to the spread of heterochromatin.

The demonstration that DNA sequences flanking the HMR locus contained boundary element function led us to test mutations in proteins that affect boundary function. The 1-kb fragment of DNA that contains boundary element function contains putative binding sites for Mig1p, Gcn4p, Pho4p, and Gcr1p. The fragment of DNA also contained a Ty1 LTR that is regulated by the Spt gene products. There was, however, no detectable defect on boundary element function in spt3, spt4, spt8, mig1, pho4, rap1, tec1, and gcr1 mutants. Because some of these genes have been implicated in LTR-mediated transcription (Eisenmann et al. 1992, 1994; Swanson and Winston 1992; Laloux et al. 1994; Turkel et al. 1997), the results suggest that the boundary element function was not due to transcription from the LTR. Of the chromosome structure mutants tested, only SMC1 andSMC3 (data not shown) mutants had a significant effect on boundary function. It is not clear whether this effect is allele specific or strain specific, as the defect seen in the smc3-1strain varied from strain to strain. The Smc proteins (Koshland and Strunnikov 1996) are a family of proteins required for chromosome condensation and cohesion. Relatives of this family are involved in dosage compensation in Caenorhabditis elegans (Chuang et al. 1994), are part of the nuclear scaffold (Hirano and Mitchison 1994), and are thought to be involved in chromosome loop organization. Our demonstration that insulator function is disrupted in smc1 andsmc3 mutants is consistent with models in which insulator element function is dependent on chromosome architecture. Our results suggest that proteins involved in higher order chromosome structure might be involved in the functional delineation of the chromosome. Further studies should help clarify the connection between these proteins and boundary function.

In conclusion, we propose that the formation of large multiprotein complexes would prevent the spread of heterochromatin and thus act as a boundary element. It is possible that enzymatic activities such as those mediated by the Swi/Snf complex or acetyltransferases could be recruited by these complexes to the boundary to actively remodel the repressed chromatin, thereby preventing its repressive effects from spreading. Alternatively boundary elements could actively sequester specific domains into subregions of the nucleus where the effective concentration of activators and repressors would determine the transcription state of the gene.

Materials and methods

Yeast transformations and integrations

PCR-based integrations were performed with oligonucleotides containing sequences corresponding to the marker gene flanked by sequences corresponding to the 5′ or 3′ sequence of the site of integration. All fragments of DNA amplified by PCR utilizedTaq DNA polymerase. Plasmid-based integrations were performed with plasmid fragments isolated by gel purification. For integrations into yeast either 5 μg of a specific gel-eluted DNA fragment or 250 μl of PCR product was concentrated to 5 μl and used for a single transformation as described (Kamakaka and Rine 1998).

Serial dilutions

A single colony of yeast cells was used to innoculate 5 ml of liquid YPD or YM medium with the appropriate supplements to allow maintenance of a plasmid. The cells were grown overnight at 30°C to an A ₆₀₀ of ∼1.0. All cells were diluted to an initial concentration of 1.0 A ₆₀₀/ml in YMD medium and serially diluted 10-fold. Approximately 3 μl of each serial dilution was spotted onto appropriately supplemented plates using a cell spotter, and the cells were allowed to grow at 30°C for 48 hr prior to photography.

Patch mating

Patches of the appropriate strains were grown on YMD plates with selection for the plasmid for 1–2 days at either 23°C or 30°C. The mating potential of the cells was monitored by replica plating the patches onto selective YMD plates spread with a mating lawn (JRY19), maintaining the selection for the plasmids following mating.

Strains

Yeast strains and plasmids are given in Tables and , respectively.

Oligonucleotides

The sequences of the various oligonucleotides used in this study will be provided upon request.