RNA polymerase III and RNA polymerase II promoter complexes are heterochromatin barriers in Saccharomyces cerevisiae (original) (raw)

EMBO J. 2001 Feb 1; 20(3): 520–531.

Rohinton T. Kamakaka

Unit on Chromatin and Transcription, NICHD/NIH, Bldg 18T, Room 106, 18 Library Drive, Bethesda, MD 20892, USA1Corresponding author e-mail: vog.hin.xileh@notnihoR

Unit on Chromatin and Transcription, NICHD/NIH, Bldg 18T, Room 106, 18 Library Drive, Bethesda, MD 20892, USA1Corresponding author e-mail: vog.hin.xileh@notnihoR

Received 2000 Aug 31; Revised 2000 Nov 17; Accepted 2000 Nov 24.

Copyright © 2001 European Molecular Biology Organization

Abstract

The chromosomes of eukaryotes are organized into structurally and functionally discrete domains. Several DNA elements have been identified that act to separate these chromatin domains. We report a detailed characterization of one of these elements, identifying it as a unique tRNA gene possessing the ability to block the spread of silent chromatin in Saccharomyces cerevisiae efficiently. Transcriptional potential of the tRNA gene is critical for barrier activity, as mutations in the tRNA promoter elements, or in extragenic loci that inhibit RNA polymerase III complex assembly, reduce barrier activity. Also, we have reconstituted the Drosophila gypsy element as a heterochromatin barrier in yeast, and have identified other yeast sequences, including the CHA1 upstream activating sequence, that function as barrier elements. Extragenic mutations in the acetyltransferase genes SAS2 and GCN5 also reduce tRNA barrier activity, and tethering of a GAL4/SAS2 fusion creates a robust barrier. We propose that silencing mediated by the Sir proteins competes with barrier element-associated chromatin remodeling activity.

Keywords: boundaries/HMR/promoters/tRNA/silencing

Introduction

All organisms regulate expression of their genome for proper development and survival. In a given mammalian cell type, ∼90% of the genome is transcriptionally silent (Allis and Gasser, 1998); therefore, a key question in eukaryotic biology is how individual cells coordinate the expression of only a fraction of their genome at specific times and within specific cell types. A large degree of this specificity is due to the proper spatial and temporal expression of distinct subsets of transcription factors that bind to gene regulatory sequences (Tjian and Maniatis, 1994). Recently, long-range chromatin structure effects on transcription have been recognized as an equally critical determinant in developmental, tissue-specific and inducible gene expression (Felsenfeld, 1996).

In the interphase nucleus, chromatin is physically and functionally organized into active and inactive domains (Lamond and Earnshaw, 1998). Euchromatin is transcriptionally active and is maintained in this open state through the action of specific promoter, enhancer and locus control region sequences (Bulger and Groudine, 1999). Heterochromatin, originally defined as chromosomal loci remaining condensed during interphase (Eissenberg et al., 1995; Weiler and Wakimoto, 1995), is generally transcriptionally repressed, but can contain embedded active genes (Wakimoto and Hearn, 1990; Sun et al., 2000). The structure and function of heterochromatin are mediated by sequences that nucleate the formation of repressor protein complexes that silence the region.

It has been proposed that chromatin boundary or insulator elements function to delimit the domains of silencer and enhancer function (Gerasimova and Corces, 1996; Geyer, 1997; Bell and Felsenfeld, 1999; Sun and Elgin, 1999). Insulators are defined as DNA sequences that block enhancer activation of a promoter when located between the two elements, preventing inappropriate gene activation. Barriers are sequences that block the spread of heterochromatin, thereby separating domains of active and inactive chromatin (Sun and Elgin, 1999).

The yeast Saccharomyces cerevisiae contains chromosomal loci that share molecular features with metazoan heterochromatin. Haploid wild-type yeast encode non-expressed copies of the mating-type specific MATa and MATα genes at the HMR and HML loci, and a variably transposed active copy of MATa or MATα at the MAT locus (Loo and Rine, 1995; Haber, 1998; Lustig, 1998; Stone and Pillus, 1998). Although identical in sequence to genes expressed at the MAT locus, the genes at HMR and HML are silenced due to the formation of a repressive chromatin structure. Flanking DNA sequences designated _HMR_-E and _HMR_-I are required for silencing. These silencer elements are composed of autonomously replicating sequences (ARS) that bind the origin recognition complex (ORC), and of sites that bind the proteins Rap1p and Abf1p. These proteins nucleate the assembly of a specific complex on the silenced DNA that contains the Sir proteins (Sir1p, Sir2p, Sir3p and Sir4p) and other factors involved in maintaining repression (Loo and Rine, 1995). Gene non-specific silencing in yeast is also observed near telomeric regions of chromosomes (Gottschling et al., 1990), and within the rDNA repeats (Smith and Boeke, 1997).

Previously, we reported the presence of barrier sequences flanking the heterochromatic HMR locus in S.cerevisiae (Donze et al., 1999). Detailed analysis of the telomere-proximal barrier shows that a specific tRNA gene is necessary and sufficient for barrier activity. tRNA genes are transcribed by RNA polymerase III (Pol III), and contain promoter elements (box A and box B, also known as internal control regions, or ICRs) within the transcribed sequences that bind essential transcription factors. Mutations that compromise the activity of Pol III transcription factors TFIIIC and TFIIIB impair barrier activity of this tRNA gene. We demonstrate that this barrier is important for proper regulation of the adjacent GIT1 gene on chromosome III, since deletion of this tRNAThr leads to _SIR4_-dependent repression of GIT1. Finally, we show that the promoter of the CHA1 gene, adjacent to _HML_-I, can function as a barrier to silencing. Analysis of these elements has led us to propose passive and active models of barrier function.

Results

The HMR right barrier element is a unique tRNAThr gene

The _HMR_-E silencer alone is sufficient to completely repress a1 gene transcription when located upstream. Inserting a DNA fragment containing a putative barrier between the _HMR_-E silencer and the a1 gene blocked the spreading of silenced chromatin and allowed expression of a1, which was assayed by patch mating (Donze et al., 1999). When transformed into a MATα hmrΔ strain, the a1 gene on the plasmid is silenced by _HMR_-E, allowing the cells to retain the α mating phenotype. However, a functional barrier inserted between _HMR_-E and the a1 gene blocked the spread of silencing, thus allowing a1 expression leading to a non-mating phenotype.

Deletion analysis of the HMR downstream barrier (Figure 1) revealed that the centrally located tRNAThr gene [designated tRNAThr1a [AGT] CR1 in the Saccharomyces Genome Database (SGD), referred to here as HMR tRNAThr] and its flanking regions were both necessary and sufficient for heterochromatin barrier activity. Plasmids that lacked the tRNA gene (pRO367, 386 and 390) showed a complete absence of barrier function, indicated by the formation and growth of diploid cells in the patches due to silencing of the a1 gene. Each test construct that contained the HMR tRNAThr gene showed little or no mating in the assay (pRO392, 366, 385, 466, 388, 391 and 387), indicating barrier activity and subsequent expression of a1. Interestingly, a minimal tRNAThr sequence of 119 bp (Figure 1, pRO387) showed a reduced level of barrier activity, as indicated by the formation of a limited number of colonies in the patch. This partial activity suggested that regions immediately flanking the HMR tRNAThr also contributed to blocking the spread of heterochromatin.

An external file that holds a picture, illustration, etc. Object name is cde032f1.jpg

Fig. 1. Deletion analysis of the HMR telomere-proximal barrier element. Fragments of the 1.0 kb HMR barrier element were cloned between _HMR_-E and the a1 gene of pRO363, transformed into ROY113 (MATα hmrΔ), and transformants tested for barrier activity in a mating assay. Barrier activity leads to expression of the a1 gene and a non-mating phenotype; growth indicates lack of barrier function. Relative +/– barrier activity was determined from multiple experiments. SGD chromosome III coordinates of the 1.0 kb fragment are listed at the ends of the scale, and each tick mark corresponds to 100 bp. Numbers to the left correspond to pRO plasmid numbers.

Deletion of HMR tRNAThr leads to partial repression of a downstream gene

We next wanted to ask whether deletion of the tRNAThr sequence would lead to the spread of silencing and whether this would affect the next downstream gene, GIT1, the promoter of which is 3.6 kb from _HMR_-I. Strains containing a precise 85 bp chromosomal deletion of the HMR tRNA gene were constructed, and GIT1 expression was assayed by RNA blot analysis. Deletion of the HMR tRNA sequence consistently showed a reduction in GIT1 mRNA levels that was SIR4 dependent (Figure 2), demonstrating that in the absence of a barrier element, the HMR domain can partially repress GIT1 expression.

An external file that holds a picture, illustration, etc. Object name is cde032f2.jpg

Fig. 2. Deletion of HMR tRNAThr from S.cerevisiae chromosome III results in partial repression of the downstream GIT1 gene. Isogenic wild-type (ROY1685) and barrierΔ (ROY1681) strains in SIR4 and sir4Δ backgrounds (ROY1679, sir4Δ; ROY1675, sir4Δ barrierΔ) were grown in YPD, and total RNA was prepared for northern analysis. GIT1 expression was normalized to ACT1.

The HMR tRNAThr barrier is expressed when located downstream of a silencer

Since our results showed that HMR tRNAThr was necessary for barrier function, we wished to determine whether the tRNA positioned downstream of the _HMR_-E silencer was active when it acted as a barrier. To assess whether the HMR tRNAThr was transcribed, we inserted a 19 bp extension at the 3′ end of the gene to distinguish its transcript from those of the other seven copies of tRNAThr present in the S.cerevisiae genome (Cherry et al., 1997; Mewes et al., 2000). The modified tRNA displayed unaltered barrier activity as compared with the wild-type gene in the mating assay (Figure 3A, tRNAThr+19). Total RNA from the same transformants used in the barrier experiment was resolved on an acrylamide gel, electroblotted and probed with an oligonucleotide complementary to the transcribed sequence of tRNAThr1a. Figure 3B shows the presence of the wild-type tRNAThr1a transcripts in all cells, and the modified transcript in cells carrying the plasmid containing the marked tRNAThr+19 gene, demonstrating that the barrier tRNA was not silenced when located downstream of the _HMR_-E silencer. Additionally, a modified Sup53 tRNALeu gene that exhibited barrier activity (Figure 4B) efficiently suppressed a lys2-801 amber mutation when located downstream of _HMR_-E (our unpublished data), which confirmed that a barrier tRNA gene downstream from _HMR_-E is not silenced.

An external file that holds a picture, illustration, etc. Object name is cde032f3.jpg

Fig. 3. The tRNAThr CR1 gene is expressed when downstream of the _HMR_-E silencer. A 19 bp extension was inserted into tRNAThr1a CR1 between the box B site and the transcription termination site to distinguish its transcript from the other seven tRNAThr1a genes in S.cerevisiae. Plasmids pRO363 (no barrier), pRO466 (HMR tRNAThr barrier) and pRO519 (tRNA+19 bp) were transformed into ROY1863 and URA+ transformants isolated. The same transformants tested in the barrier assay were grown in liquid media for RNA isolation. (A) The extended tRNAThr1a CR1+19 gene retains full barrier activity. (B) Expression of the tRNAThr+19 gene; the arrow indicates the extended transcript.

An external file that holds a picture, illustration, etc. Object name is cde032f4.jpg

Fig. 4. (A) Analysis of other RNA Pol III transcribed genes as barrier elements. The SNR6 gene (spliceosomal U6 RNA) and the RDN5 gene (5S rRNA) were cloned into pRO363 and assayed as in Figure 1. The no insert and HMR tRNA plasmids are as in Figure 1, the U6 gene plasmids are pRO461 and 462, and the 5S plasmids are pRO463 and 464. (B) Analysis of other yeast tRNA genes for barrier activity. Plasmids containing the following S.cerevisiae tRNA genes as barriers were tested as in Figure 1: HMR tRNAThr(AGT) CR1 from chromosome III (pRO466); tRNAThr (AGT) NL1 (chromosome XIV, pRO469); tRNAThr (CGT) KL (chromosome XI, pRO470); tRNAThr (TGT) GR1 (chromosome VII, pRO471); and SUP53 tRNALeu3 with (pRO465) or without (pRO467) its intervening sequence (IVS). (C) Flanking regions of tRNAThr CR1 contribute to barrier activity. The 5′ and 3′ flanking regions of tRNAThr1a CR1 and tRNAThr1a NL1 were swapped and tested for barrier function. The plasmid designated 5′-CR1/3′-CR1 is pRO466, 5′-NL1/3′-NL1 is pRO469, 5′-CR1/3′-NL1 is pRO479, and 5′-NL1/3′-CR1 is pRO481. A dimerized tRNA NL1 (pRO529) also functioned as a barrier to silencing (bottom panel).

RNA Pol III transcribed genes are not necessarily barrier elements

Since the HMR tRNAThr gene showed robust barrier activity, we asked whether other RNA Pol III promoters, those of the 5S rRNA and U6 RNA genes, displayed barrier activity. As shown in Figure 4A, neither gene was as effective as the HMR tRNAThr gene in its ability to block the spread of silencing from _HMR_-E. The U6 gene showed slight activity in one orientation (pRO461), but the barrier activity of the 5S gene was indistinguishable from the no-barrier control (pRO463 and pRO464).

Since barrier activity was not an intrinsic property of Pol III genes, we asked whether it was specific to tRNA genes. Since the HMR tRNA is a tRNAThr and there are multiple copies of tRNA with threonine anticodons in the yeast genome, we tested other tRNAThr genes for barrier activity (Figure 4B). The tRNAThr genes NL1 (chromosome XIV, pRO469) and KL (chromosome XI, pRO470) showed weak barrier activity, while tRNAThr GR1 (chromosome VII, pRO471) showed robust barrier activity comparable to the HMR tRNAThr barrier.

The SUP53 tRNALeu gene is a well characterized gene in studies on tRNA regulation; therefore, we asked whether it was capable of acting as a barrier. However, as shown in Figure 4B, the SUP53 tRNALeu gene showed no barrier activity (pRO465). Detailed inspection of its DNA sequence revealed a key difference between HMR tRNAThr and SUP53 in that the suppressor tRNA contains an intron that increases the box A–box B promoter element distance to 74 bp, compared with 32 bp in HMR tRNAThr. To test whether this larger spacing of the promoter elements affected barrier function, we deleted this intron to create a _Sup53Δ_IVS allele, with the ICR spacing reduced to 42 bp. This tRNA gene showed a dramatic increase in barrier activity (Figure 4B, pRO467).

It is interesting that tRNAThr NL1 is a weaker barrier than the HMR tRNA, as it has the identical coding sequence (and therefore identical box A and box B promoter elements and spacing), but differs in the flanking regions. To demonstrate further the importance of the flanks on barrier activity, we created hybrid tRNAThr NL1 genes containing either the upstream or downstream flank of the HMR tRNA. Figure 4C shows that substitution of either flank increases boundary activity of NL1, consistent with the results of deleting the flanks of HMR tRNAThr shown in Figure 1 (pRO387).

Furthermore, the tRNA NL1 gene, which is a weak barrier on its own, is converted to a strong barrier when this gene is duplicated in tandem (Figure 4C, bottom panel), suggesting that tRNA promoter occupancy may play a role in barrier function.

Mutations that inhibit the function of RNA Pol III transcription factors compromise the activity of the barrier tRNA

Having established that all Pol III transcribed genes were not competent for barrier function, we conducted a series of experiments to determine whether Pol III-mediated transcription was necessary for barrier activity by examining mutations in the HMR tRNAThr gene. Mutations GCC10–12/AAA in the box A sequence and C56/G in the box B sequence are known to reduce transcription of tRNA genes in vitro (Newman et al., 1983) and in vivo (Hull et al., 1994), while the GGC23–25/TTT mutation, which lies between the promoter elements, has no effect on transcription. Figure 5A shows barrier assay results of the 320 bp HMR tRNAThr fragment, which displayed full activity (pRO466), and the same fragment containing point mutations in the box A and box B promoter elements. Only the mutations known to affect transcription (GCC 10–12 AAA and C56G) resulted in a loss of barrier function, while the GGC23–25/TTT mutation had no effect. This loss of barrier activity is consistent with the severity of the transcription defects caused by these mutations in vitro (Newman et al., 1983), with C56/G showing the greatest effect.

An external file that holds a picture, illustration, etc. Object name is cde032f5.jpg

Fig. 5. (A) Barrier activity of mutant tRNAThr1a CR1 genes. Box A and box B promoter elements were mutagenized in pRO466 and tested for barrier activity. The mutant HMR tRNA barrier plasmids are pRO499 (GGC23–25/TTT), pRO498 (GCC10–12/AAA) and pRO468 (C56/G). (B) Analysis of barrier activity of tRNAThr1a CR1 in strains carrying conditional mutations in RNA Pol III factors. Upper panel, wild-type (ROY1863), rpc31-236 (ROY1861) and tfc3-G349E (ROY1862) strains were transformed with pRO363 or pRO495 (pRO363 containing the same insert as pRO391 in Figure 1, but in the opposite orientation) and tested as in Figure 1. Lower panel, wild-type (ROY1864), brf-II.6 (ROY1539) and brf-II.9 (ROY1531) strains were transformed with pRO363 or pRO495 as above.

The first step in Pol III transcription is the binding of TFIIIC to box B and box A promoter elements. This is followed by the recruitment of TFIIIB, which binds upstream of the tRNA gene. The final step in transcription is the recruitment of the polymerase subunits. Since mutations in the promoter elements affected barrier activity, we wished to determine whether mutant alleles of TFIIIB, TFIIIC or Pol III also affected barrier activity. Since each of these RNA Pol III complexes is essential for cell viability, we performed the experiments using temperature-sensitive alleles at the permissive temperature, where only subtle phenotypes were evident. However, as shown in Figure 5B, a strain mutant in a TFIIIC DNA binding component, tfc3G349E (Lefebvre et al., 1994; Arrebola et al., 1998), resulted in a subtle but consistent loss of barrier activity. Mutations in brf1 (Andrau et al., 1999), the gene coding for the 70 kDa component of TFIIIB, showed a strong reduction in barrier function, as demonstrated by a significant increase in the level of mating (Figure 5B, lower panel). The rpc31-236 mutation in a polymerase subunit (Thuillier et al., 1995) that is competent for pre-initiation complex assembly, but defective in initiation of Pol III transcription, did not show significant loss of barrier activity. While we cannot at this time conclude whether RNA Pol III transcription itself is required to block the spread of heterochromatin, our results indicated that efficient TFIIIC and TFIIIB assembly was required for barrier activity.

Effects of acetyltransferase mutations on barrier activity

It has been shown that yeast silent chromatin contains hypoacetylated histones (Braunstein et al., 1993), a feature common to heterochromatin in other eukaryotes (Jeppesen and Turner, 1993; O’Neill and Turner, 1995). Silencing at HMR is SIR2 dependent, and it has recently been shown that Sir2p possesses NAD-dependent histone deacetylase activity that is required for silencing (Imai et al., 2000; Landry et al., 2000; Smith et al., 2000). One possible mechanism of barrier activity could be the recruitment of acetyltransferase activity to counteract the effects of Sir2p-mediated deacetylation. We tested the activity of the HMR tRNAThr barrier in a series of acetyltransferase mutants and the results are shown in Figure 6A. Barrier activity is markedly reduced in strains carrying a deletion of the SAS2 gene or a disruption of GCN5, as observed by the increased mating due to repression of a1. A strain mutant for both SAS2 and GCN5 shows an even greater relative loss of activity. Deletion of SAS3 had no apparent effect on barrier activity, nor did a temperature-sensitive mutation in the essential acetyltransferase ESA1 (Clarke et al., 1999).

An external file that holds a picture, illustration, etc. Object name is cde032f6.jpg

Fig. 6. (A) Effects of mutations in acetyltransferase genes on barrier activity. Strains isogenic to ROY1864 (wild type) containing deletions of SAS3 (ROY1510) or SAS2 (ROY1528), disruption of GCN5 (ROY1544), or a conditional point mutation in ESA1 (ROY1643) were constructed and barrier activity of the HMR tRNA was tested in each. (B) Tethering of acetyltransferases or a transcriptional activator can create a barrier to heterochromatin. Plasmids expressing the GAL4 DNA binding domain (pLP493), GAL4/SAS2 (pLP646), GAL4/GCN5 (pLP871) or GAL4/VP16 (pRO541) were co-transformed with pRO486 (4 × GAL4 UAS between _HMR_-E and a1) or pRO363 (no barrier) into ROY1864, and mating assays performed as above.

To determine whether acetyltransferase activities could directly create a barrier effect, we cloned four Gal4p binding sites between _HMR_-E and the a1 gene, and tested the Gal4p sites for barrier activity in the presence of various Gal4 fusion proteins. We tested strains expressing either the Gal4p DNA binding domain alone or strains expressing Gal4–Sas2 or Gal4–Gcn5 fusion proteins. Expressing the Gal4 DNA binding domain alone did not act as a barrier, suggesting that the mere presence of a protein bound to these sites is not sufficient to stop the spread of silencing. Expression of Gal4–Sas2 created a strong barrier to the spread of silencing, which was dependent on the presence of Gal4 binding sites, while Gal4–Gcn5 created a relatively weak barrier. Similarly, tethering of the potent transcriptional activator Gal4– VP16 also created a barrier to silencing (Figure 6B). Western blot analysis showed that these fusion proteins were present at equivalent levels (our unpublished data).

Other boundaries in yeast

In addition to the silent mating loci, telomeres and the rDNA locus, there are other genes in yeast that are conditionally repressed. We were therefore interested in determining whether regions flanking these genes contained barrier elements that could stop the spread of silencing from _HMR_-E. To address this question, we looked at several genes in S.cerevisiae where transcriptional repression is known to spread from a _cis_-acting element, and used sequences flanking these repressed regions in our barrier assay. STE6 is a cell-type-specific gene that is repressed in α cells by mechanisms requiring Ssn6p and Tup1p (Herskowitz et al., 1992), and it has recently been shown that Tup1p spreads along the STE6 gene (Ducker and Simpson, 2000). A tRNAThr (KL) gene resides upstream of the STE6 gene, but this tRNA gene is a weak barrier to _HMR_-E-mediated silencing (Figure 4B). We therefore tested a 0.6 kb fragment at the 3′ flank of the STE6 gene for barrier activity, and as demonstrated in Figure 7A, this sequence was unable to block the spread of silencing from _HMR_-E.

An external file that holds a picture, illustration, etc. Object name is cde032f7.jpg

Fig. 7. (A) Other heterochromatin barrier sequences in S.cerevisiae. Plasmids containing the following S.cerevisiae sequences as barriers were tested as in Figure 1: _STE6_-3′ (pRO506); _SMK1_-3′ (pRO508); _TEF2_-UAS (pRO504); CHA1 promoter (pRO510); _CHA1_-UAS (pRO512); and the TRP1 gene (pRO262). Strain ROY1775 has a PCR-generated cha4Δ::TRP1 chromosomal deletion. CHA1 + serine plates were supplemented with 400 µM serine. (B) The Drosophila gypsy element can be reconstituted as a heterochromatin boundary in yeast. ROY1863 was transformed with pRO493, which contains eight su(Hw) binding sites between _HMR_-E and the a1 gene, and plasmids expressing su(Hw), mod(mdg4), or their respective vector controls.

SMK1 is a meiosis-specific gene that is repressed in vegetative cells by a pathway requiring SUM1 and HST1 (Xie et al., 1999). HST1 encodes a gene homologous to SIR2, and probably mediates repression by similar mechanisms, since overexpression of HST1 can partially suppress mating-type silencing defects in a sir2 strain (Brachmann et al., 1995). We were therefore interested in determining whether the region flanking the downstream end of SMK1 contained elements capable of blocking the spread of silencing. However, like the STE6 3′ region, this fragment also showed no barrier activity in our assay (Figure 7A).

We next looked at the CHA1 gene, the promoter of which is ∼2 kb downstream of the _HML_-I silencer. Moreira and Holmberg (1998) have shown that Sir4p is required for full repression of CHA1 transcription when uninduced, suggesting that silencing from _HMR_-I may spread toward the centromere at least as far as the CHA1 promoter. When serine is present in the medium, CHA1 transcription is robustly induced, suggesting that the activated promoter can overcome the effects of _SIR4_-mediated repression. Figure 7A shows that the CHA1 promoter plus UAS (_CHA1_-prm) or the UAS alone exhibited no barrier activity in media lacking serine. However, when the experiment was performed in the presence of serine, robust barrier activity was observed, comparable to the HMR tRNA. This barrier activity was dependent on the presence of the CHA4 gene (Figure 7A, CHA1-prm/cha4Δ), which encodes a C6 zinc cluster transcription factor that binds to the CHA1 UAS and is required for serine induction of CHA1 (Holmberg and Schjerling, 1996). As additional controls, we tested barrier activity of the TEF2 UAS region identified as a barrier element in another study (Bi and Broach, 1999), and the TRP1 gene as another RNA Pol II promoter. The TEF UAS was an efficient barrier in our assay (Figure 7A), while the TRP1 gene including its promoter was completely inactive as a barrier, consistent with previous results describing the silencing of TRP1 at HMR (Sussel and Shore, 1991; Donze et al., 1999).

We also tested whether a known metazoan boundary could block the spread of _HMR_-E-initiated silencing in yeast. The Drosophila gypsy insulator element contains multiple binding sites for the suppressor of hairy wing [_su(Hw)_] protein, which is required for both enhancer blocking and heterochromatin barrier functions of the element (Holdridge and Dorsett, 1991; Geyer and Corces, 1992; Roseman et al., 1993). Some of the functions of su(Hw) also require expression of the modifier of mdg4 [_mod(mdg4)_] protein (Georgiev and Gerasimova, 1989; Gerasimova et al., 1995). Kim et al. (1993) have demonstrated that the gypsy insulator does not display enhancer blocking activity in yeast, but heterochromatin barrier activity was not tested in this study. We cloned a cassette containing eight su(Hw) binding sites into our barrier test vector and performed the mating assay experiment with and without ectopic expression of su(Hw) or mod(mdg4) proteins. The results in Figure 7B show that full barrier activity was seen in strains expressing su(Hw) [with or without concomitant mod(mdg4) expression], but not in strains where only mod(mdg4) or neither protein was expressed. This result demonstrated that heterochromatin barrier activity of the gypsy insulator could be reconstituted in yeast when su(Hw) was ectopically expressed.

Discussion

The HMR right boundary is a unique and functional tRNA gene

Our functional identification of the HMR tRNAThr as the telomere-proximal boundary of the HMR domain is in perfect correlation with restriction enzyme accessibility studies of the region (Loo and Rine, 1994). Those experiments mapped the downstream end of the inaccessible HMR chromatin to a region between the _Pst_I and _Ava_II sites on either side of the HMR tRNAThr (SGD chromosome III coordinates 294054 for _Pst_I and 294360 for _Ava_II; see Figure 1). We had previously shown that the chromosomal deletion of a 1.0 kb region containing this tRNA and a Ty1 long terminal repeat led to _SIR_-dependent repression of URA3 inserted downstream of the deleted sequence (Donze et al., 1999). In this report we have demonstrated that the HMR tRNAThr gene and its immediate flanks were both necessary and sufficient to prevent the spread of silencing from the HMR locus.

tRNA genes contain regulatory sequences within the transcribed regions, called ICRs, which are the highly conserved box A and box B promoter sequences (Geiduschek and Kassavetis, 1992; Chedin et al., 1998; Kumar et al., 1998). Transcription of tRNA genes is initiated by the stepwise assembly of a large macromolecular complex, composed of the multisubunit transcription factors TFIIIC and TFIIIB, and the multisubunit RNA Pol III. TFIIIC binds directly to the box A and box B sequences, followed by recruitment of TFIIIB, which binds upstream of tRNA genes in a non-sequence-specific manner. TFIIIB binding is followed by recruitment of the catalytic RNA Pol III subunits. This assembly creates an extremely large complex of ∼1.5 MDa, consisting of 26 polypeptides, and protects ∼150 bp of DNA in DNase I footprint assays (Geiduschek and Kassavetis, 1992; Chedin et al., 1998). We have shown that mutations in the HMR tRNAThr promoter boxes, and mutations in TFIIIB and TFIIIC subunits, resulted in a loss of barrier function, suggesting that the integrity of the complex was required for efficient barrier activity. One possible mechanism of barrier activity is that a stably bound RNA Pol III pre-initiation complex acts as a large physical block to the propagation of silencing.

The assembly of a Pol III–tRNA gene complex is known to have a number of effects on both the structure of chromatin and several DNA processes. The SUP53 tRNALeu gene can repress the level of transcription of a nearby HIS3 gene (Hull et al., 1994), and Ty element retrotransposition is directed to chromosomal locations in close proximity to genes transcribed by RNA Pol III (Chalker and Sandmeyer, 1992; Kirchner et al., 1995; Devine and Boeke, 1996). Assembly of a Pol III complex on a tRNA gene can dominantly override nucleosome phasing normally induced by a nucleosome positioning element (Morse et al., 1992), and tRNA genes have also been shown to act as DNA replication fork pause sites in S.cerevisiae (Deshpande and Newlon, 1996). Each of these tRNA effects requires a fully assembled and functional RNA Pol III complex. Mutations that reduce Pol III factor binding or polymerase activity also abolish or reduce these secondary effects, suggesting that the entire complex is required for the observed phenomena. We see similar requirements for a fully functional Pol III–tRNA gene complex for barrier activity. Our demonstration that the HMR tRNAThr gene transcript is produced when cloned downstream from the _HMR_-E silencer confirms that a functional RNA Pol III complex is present at the HMR tRNAThr. The expression and mutational data presented here demonstrate that this barrier tRNA is a fully functional tRNAThr gene.

It is interesting that only a subset of Pol III transcribed genes function as boundaries, suggesting that there are unique features of this tRNA gene at the HMR locus. The requirement of the flanking sequences for full barrier function is not surprising, as several studies have shown that tRNA flanks contribute to transcriptional potential, even though the promoter elements are within the genes (Sprague et al., 1980; Dingermann et al., 1982; Raymond and Johnson, 1983; Shaw and Olson, 1984; Young et al., 1991; Ong et al., 1997). This could reflect preferential interaction of TFIIIB with flanking sequences, since TFIIIB creates an extensive footprint on the 5′ flank of tRNA genes (Huibregtse and Engelke, 1989; Kassavetis et al., 1989). Other studies have demonstrated that different tRNA genes with identical coding sequences are expressed at different levels due to specific upstream elements (Raymond et al., 1985; Ong et al., 1997). One study specifically demonstrated that the efficiency of TFIIIB loading varies with the 5′ flanking sequence (Joazeiro et al., 1996). Raymond et al. (1985) identified a pentadecanucleotide sequence (TTTCAACAAATAAGT) contiguous with the 5′ end of several copies of tRNALeu3 that is 60–80% conserved in tRNA genes that are abundantly transcribed. The HMR tRNAThr 5′ region has 67% homology to this sequence, which may partially explain the function of the 5′ flank in barrier activity. Such differences in tRNA genes may result in more stable interaction of certain gene flanks with TFIIIB and the Pol III complex, increasing the probability that Pol III complex formation efficiently competes with heterochromatin formation. The loss of barrier activity seen in the brf_1 mutant backgrounds is consistent with this hypothesis. The barrier activity of a multimerized tRNA NL1 may also be due to increasing the probability of Pol III complex formation and occupancy. An alternative possibility is that the HMR tRNAThr flanks interact with additional factors involved in barrier activity. The increase in barrier activity resulting from the deletion of the SUP53 intron suggests that promoter element spacing is also important in tRNA barrier function. This result is in accordance with studies of the transcriptional promoter strength of tRNA genes. Natural tRNA sequences have box A–_box B separations of ∼30–90 bp, but optimal expression is seen when the boxes are only 30–60 bp apart (Baker et al., 1987; Fabrizio et al., 1987).

A general theme of RNA Pol III genes as chromosomal domain boundaries is emerging. As described above, a number of effects on chromatin processes are caused by the presence of nearby tRNA genes. Recently, two other types of boundary-like activities have been reported for RNA Pol III genes. Alu elements are repetitive sequences in primate genomes, most of which contain functional Pol III promoters. (Deininger and Batzer, 1999). Willoughby et al. (2000) have demonstrated that a human Alu element flanking the keratin-18 gene can confer position-independent and copy-number-dependent transgene expression in mice. Mutation of the box B site of this Alu element abolishes this protection against position effects. In Schizosaccharomyces pombe, mapping of centromeric proteins Swi6 and Mis6 by chromatin immunoprecipitation has defined discrete domains of interaction of these two proteins within centromeric chromatin, and a pair of tRNA genes demarcate the transition zone (Partridge et al., 2000). There may be special features of chromatin assembled RNA Pol III complexes, in addition to their large size, that may prevent the propagation of a given chromatin structure.

Potential mechanisms of barrier activity

How do barrier sequences prevent the spread of heterochromatin? The RNA Pol III complex has a cumulative mass of 1.5 MDa and footprints ∼150 bp on a tRNA gene (Chedin et al., 1998; Kassavetis et al., 1998). Such a large, stably bound complex could be a physical impediment to heterochromatin spreading by disrupting the binding of SIR complex proteins. Nucleosome mapping of silenced HMR and HML domains reveals a distinctive _SIR_-dependent positioned dinucleosome array, which may be important for propagation and stability of the heterochromatic state (Weiss and Simpson, 1998; Ravindra et al., 1999). By creating a large gap in this array, a tightly bound RNA Pol III complex may passively disrupt this SIR protein-mediated nucleosomal organization and stop the propagation of heterochromatic structure (Figure 8).

An external file that holds a picture, illustration, etc. Object name is cde032f8.jpg

Fig. 8. Models of heterochromatin barrier activity. Silencer bound complexes nucleate the spread of histone deacetylation and the binding of heterochromatin-associated proteins along the silenced region. This structure is propagated until a barrier is reached. In the passive barrier model, a large multiprotein complex stably bound to DNA physically interferes with the propagation of the heterochromatic structure. In the active model, chromatin-modifying activities (e.g. acetyltransferases or nucleosome remodeling complexes) are recruited to the barrier to modify histones and perhaps other chromatin-associated factors. The modified nucleosomes would be less efficient substrates for incorporation into heterochromatin.

Acetylation of histones and other chromatin substrates plays a major role in the regulation of higher order chromatin structure and transcriptional competence of chromosomal loci (Sterner and Berger, 2000), and several histone acetyltransferases (HATs) have been identified. Hyperacetylation of histones is associated with increased general DNase I sensitivity and transcriptional potential of chromosomal domains (Hebbes et al., 1994), while hypoacetylation is a characteristic of inactive heterochromatin (Lin et al., 1989). Yeast heterochromatin is known to be hypoacetylated (Braunstein et al., 1993), and deacetylation of Lys16 of histone H4 appears to be particularly important for the maintenance of SIR protein-mediated silencing (Johnson et al., 1990). Results from several laboratories (Imai et al., 2000; Landry et al., 2000; Smith et al., 2000) suggest that Sir2p-catalyzed deacetylation of histones, and possibly other chromatin substrates, is a key activity in the maintenance of yeast heterochromatin. Our observation of an apparent loss of barrier activity in strains with mutations in the HAT genes GCN5 and SAS2 suggests another possible mechanism of barrier function: that of competing chromatin-modifying activities (Figure 8). In this model, Sir2p-mediated deacetylation is propagated from the silencer until it meets a barrier element that stably recruits one or more acetyltransferases that actively compete with the effects of silencer-recruited Sir2p. Although no acetyltransferase has yet been demonstrated to be associated with the RNA Pol III complex in yeast, HAT activity associated with human TFIIIC has been identified (Kundu et al., 1999). While this HAT has no yeast homolog, one or more of the other S.cerevisiae HATs may have redundant functions in yeast Pol III activity.

One caveat to the active model is that a null mutation of GCN5 has been shown to increase silencing at telomeres (Sun and Hampsey, 1999), and strains lacking Sas2p show a restoration of silencing at a weakened HMR locus (Reifsnyder et al., 1996; Ehrenhofer-Murray et al., 1997). Thus, the strengthening of silencing by these mutations may simply override the barrier activity in our assays. Even if this is the case, these results suggest that barrier activity may arise from an underlying competition between the chromatin remodeling and silencing activities at the interface of euchromatin and heterochromatin. However, the ability of tethered GAL4/SAS2 to block silencing (Figure 7) suggests that the recruitment of an acetytransferase complex can be sufficient for formation of a heterochromatin barrier in yeast. GAL4/VP16 expression can also create a strong barrier in our assay system, possibly due to the recruitment of chromatin-modifying factors such as Gcn5p.

The ability of the TEF2 UAS regions to function as boundaries would also fit either the physical block or the competing activity models, which are not mutually exclusive. Bi and Broach (1999) have proposed that chromatin-bound Rap1p at the TEF2 UAS may act as a barrier by creating a nucleosomal ‘hole’ that inhibits the migration of silencer nucleated Sir complex. However, since Rap1p can act as either a repressor or an activator of promoters (Shore, 1994; Morse, 2000), one can not rule out the possibility that Rap1p barrier activity may be due to recruitment of chromatin remodeling factors associated with gene activation. It is interesting that Rap1p shares this characteristic of being an activator or a repressor with CTCF (Klenova et al., 1993), a vertebrate protein recently identified as a boundary binding factor (Bell et al., 1999; Bell and Felsenfeld, 2000; Hark et al., 2000). The observed barrier effect of the su(Hw) protein in yeast may also fall into this category, as it has been reported to have properties consistent with activation (Parkhurst and Corces, 1986) and repression (Gerasimova et al., 1995; Gerasimova and Corces, 1996) of transcription.

The mechanism of CHA1 UAS barrier activity appears to suggest that disruption of the nucleosomal array alone may not be sufficient to form a barrier. In the uninduced state, the CHA1 TATA box is packaged within an array of positioned nucleosomes. This positioning in the repressed state is dependent on both Sir4p and the nucleosome remodeling complex RSC (Moreira and Holmberg, 1998, 1999). However, upstream of the CHA1 TATA box is a 250 bp region constitutively free of nucleosomes. This region encompasses the CHA1 UAS sites, which bind the Cha4p activator in both the induced and uninduced states. Upon addition of serine, CHA1 transcripts are quickly induced in a _CHA4_-dependent manner, and nucleosome positioning at CHA1 is disrupted. This situation of a repressed but constitutively bound activator is similar to that seen when _HMR_-E is positioned upstream of the HSP82 promoter (Lee and Gross, 1993; Sekinger and Gross, 1999). Also, as shown in Figure 6, expression of the Gal4p DNA binding domain is not sufficient to create a barrier at GAL4 UAS sites. Therefore, in these situations, simply creating a nucleosomal gap is not sufficient for barrier activity, so we propose that upon induction these DNA-bound factors recruit additional proteins or mediate chromatin remodeling, which leads to formation of a barrier to silencing.

In summary, we have identified the HMR downstream barrier as the tRNAThr CR1 gene, and demonstrated that a fully assembled RNA Pol III complex is required to resist the spread of heterochromatin. This and other recent studies suggest a general role for RNA Pol III genes in demarcating the genome. The observed barrier activity of HMR tRNAThr, CHA1 UAS and the TEF2 UAS suggest active mechanisms by which these elements can disrupt silenced chromatin. Elucidation of the mechanisms by which these sequences resist Sir-mediated silencing will tell us much about the interface between euchromatin and heterochromatin, and how these regions are restricted to their respective domains.

Materials and methods

Yeast strains

Saccharomyces cerevisiae W-303 background strains generated in our laboratory for this study are as follows:

ROY 113, MATα ade2-1 his3-11 leu2-3,112 trp1-1 ura3-1 hmrΔ::bgl-bcl (Donze et al., 1999);

ROY1863, MATα his3-11 leu2-3,112,112 trp1-1-1 ura3-1-1 hmrΔ::bgl-bcl;

ROY1864, MATα his3-11 leu2-3,112–3,112 lys2Δ trp1-1 ura3-1 hmrΔ::bgl-bcl;

ROY1861, MATα his3-11 leu2-3,112 trp1-1 ura3-1 rpc31–236 hmrΔ::bgl-bcl;

ROY1862, MATα his3-11 leu2-3,112 trp1-1 ura3-1 tfc3-G349E hmrΔ::bgl-bcl;

ROY1531, MATα his3-11 leu2-3,112 lys2Δ trp1-1 ura3-1 hmrΔ::bgl-bcl brf1Δ::HIS3 [brf1 II.9 _ARS CEN LEU2_];

ROY1539, MATα his3-11 leu2-3,112 lys2Δ trp1-1 ura3-1 hmrΔ::bgl-bcl brf1Δ::HIS3 [brf1 II.6 _ARS CEN LEU2_];

ROY1510, MATα his3-11 leu2-3,112 lys2Δ trp1-1 ura3-1 hmrΔ::bgl-bcl sas3Δ::HIS3;

ROY1528, MATα his3-11 leu2-3,112 lys2Δ trp1-1 ura3-1 hmrΔ::bgl-bcl sas2Δ::TRP1;

ROY1544, MATα his3-11 leu2-3,112 lys2Δ trp1-1 ura3-1 hmrΔ::bgl-bcl gcn5::TRP1;

ROY1639, MATα his3-11 leu2-3,112 lys2Δ trp1-1 ura3-1 hmrΔ::bgl-bcl sas2Δ::TRP1 gcn5::TRP1;

ROY1643, MATα his3-11 leu2-3,112 lys2Δ trp1-1 ura3-1 hmrΔ::bgl-bcl es_a1_-L327S;

ROY1675, MATα his3-11 leu2-3,112 lys2Δ trp1-1 ura3-1 sir4Δ::LEU2 tRNAThr1a CR1Δ;

ROY1679, MATα his3-11 leu2-3,112 lys2Δ trp1-1 ura3-1 sir4Δ::LEU2;

ROY1681, MATα his3-11 leu2-3,112 lys2Δ trp1-1 ura3-1 tRNAThr1a CR1Δ;

ROY1685, MATα his3-11 leu2-3,112 lys2Δ trp1-1 ura3-1;

ROY1775, MATα his3-11 leu2-3,112 trp1-1 ura3-1 hmrΔ::bgl-bcl cha4Δ::TRP1.

Plasmids

Barrier test plasmids were constructed in pRO363, which contains the _Sac_I–_Sal_I HMRΔI fragment from pRO22 (Donze et al., 1999) cloned into pRS406 (Sikorski and Hieter, 1989) and containing a synthetic polylinker with a unique _Bam_HI site cloned into the _Eco_NI site of the a2 gene. All fragments tested were cloned by PCR amplification with _Bam_HI sites in the primers. Mutant tRNAs were created using the Quik-Change Mutagenesis Kit (Stratagene), and mutations were in the same box A and box B positions as described (Newman et al., 1983) for pRO499, pRO498 and pRO468, or by inserting 19 bp between box B and the transcription termination site for pRO519. The SUP53 gene was the _Bam_HI fragment as described (Hull et al., 1994), and the intron was deleted using site-directed mutagenesis as described above. Hybrid tRNAThr1a genes used in the flanking region swap experiment were made by PCR cloning each region and joining the swapped fragments at the _Hin_dIII site at the start of the HMR tRNA sequence. Final test clones of each barrier sequence were verified by either manual or automated dideoxy DNA sequencing. SGD (Cherry et al., 1997) coordinates of each fragment tested are available on request. Designations of tRNA genes are those listed in the Munich Information Centre for Protein Sequences (MIPS) Database (Mewes et al., 2000).

Expression plasmids for su(Hw) and mod(mdg4) were constructed by first cloning the ADH1 promoter from pCF113 (ARS CEN URA3 ADH1 promoter in pRS416; gift from Catherine Fox) into pRS414 to create an ARS CEN TRP1 ADH1 promoter expression plasmid (pRO489). To construct pRO494 [_ADH1_-_su(Hw) ARS CEN URA3_], the su(Hw) cDNA was cloned into pCF113. _ADH1_-mod(mdg4) ARS CEN TRP1 (pRO497) contained the mod(mdg4) cDNA in pRO489 (cDNAs were gifts from Victor Corces and Dolan Ghosh). Eight su(Hw) binding sites were cloned into pRO490 (a LEU2 version of pRO363) to create pRO493. Plasmids pLP493, pLP646 and pLP871 were a generous gift from Lorraine Pillus.

Mating assays

Mating assays to determine barrier activity were performed as described (Donze et al., 1999) with selection on yeast minimal media plates supplemented with dextrose (YMD) plus adenine, leucine, lysine and tryptophan. In experiments using lys2Δ strains, the selection medium was YMD plus adenine, histidine, leucine and tryptophan, and in the gypsy element experiment, the selection medium was YMD plus adenine and lysine.

RNA analysis

Expression of the marked tRNA+19 gene was determined as described (Krieg et al., 1991). The blot was probed using a 32P-end-labeled 76mer oligonucleotide complementary to the tRNAThr1a coding sequence using the oligonucleotide hybridization and washing protocol supplied with the membrane (Bio-Rad Zeta Probe). Northern blot analysis was performed using the NorthernMax kit (Ambion), and blots were probed with RNA complementary to the first 600 nucleotides of either GIT1 or ACT1. Riboprobes were synthesized using Maxiscript in vitro transcription reagents (Ambion). Quantitative analysis was performed on a Molecular Dynamics Storm 820 PhosphorImager.

Acknowledgements

We thank Pierre Thuriaux, Michel Werner, David Stillman, David Engelke, Catherine Fox, Victor Corces, Lorraine Pillus, Astrid Clarke, Sandi Jacobson and Jasper Rine for strains, plasmids and valuable advice on their use. We also thank David Wasserman and Erin Schlag for automated DNA sequencing. Finally, we are grateful to members of the Kamakaka, Wasserman, Lilly and Wolffe laboratories for helpful comments and suggestions.

References

Articles from The EMBO Journal are provided here courtesy of Nature Publishing Group