An RNA Polymerase III-Dependent Heterochromatin Barrier at Fission Yeast Centromere 1

Heterochromatin formation involves the nucleation and spreading of structural and epigenetic features along the chromatin fiber. Chromatin barriers and associated proteins counteract the spreading of heterochromatin, thereby restricting it to specific regions of the genome. We have performed gene expression studies and chromatin immunoprecipitation on strains in which native centromere sequences have been mutated to study the mechanism by which a tRNAAlanine gene barrier (cen1 tDNAAla) blocks the spread of pericentromeric heterochromatin at the centromere of chromosome 1 (cen1) in the fission yeast, Schizosaccharomyces pombe. Within the centromere, barrier activity is a general property of tDNAs and, unlike previously characterized barriers, requires the association of both transcription factor IIIC and RNA Polymerase III. Although the cen1 tDNAAla gene is actively transcribed, barrier activity is independent of transcriptional orientation. These findings provide experimental evidence for the involvement of a fully assembled RNA polymerase III transcription complex in defining independent structural and functional domains at a eukaryotic centromere.


INTRODUCTION
Eukaryotic genomes are packaged into two main categories of chromatin that can determine the behavior of the underlying DNA sequence. Euchromatin is typically found in gene rich regions of the genome that are accessible to factors involved in various biological processes including transcription, replication and recombination. In contrast, regions of heterochromatin are generally gene poor, and confer transcriptional repression to inserted reporter genes [1]. Heterochromatin also is required for proper chromosome segregation [2,3]. Hypoacetylation of histones and methylation of histone H3 at lysine 9 are distinguishing marks of heterochromatin in many eukaryotic genomes, including the fission yeast Schizosaccharomyces pombe (S. pombe) [4]. In addition to the general categories of heterochromatin and euchromatin, a third type of specialized chromatin exists at centromeres. Centromeric chromatin contains blocks of canonical nucleosomes, methylated on H3 lysine 4, interspersed with blocks of nucleosomes containing the histone H3 variant, cenH3, a protein that provides the structural and functional foundation of all active kinetochores [5].
A key feature of heterochromatin is its ability to spread in cis, causing epigenetic silencing of an otherwise euchromatic gene [6]. The genomic and/or epigenetic features at the confluence of discrete chromatin domains remain poorly understood; however, at some loci a specialized class of DNA element, known as a chromatin barrier [7], plays an active role in demarcating the different chromatin states. Chromatin barriers restrict heterochromatin to specific genomic regions and fall within a broader class of elements called insulators [8]. The range of proteins associated with chromatin barriers is incompletely defined, although increasing evidence suggests that barrier activity correlates with the recruitment of histone acetylase activity and/or the assembly of a transcription complex [9,10,11,12]. Accordingly, many insulators are coincident with the promoters of genes [13, 14,15] We recently defined a novel chromatin barrier element in the fission yeast genome that partitions centromere 1 (cen1) chromatin into structurally distinct domains of pericentromeric heterochromatin and centromeric chromatin [16]. The absence of this barrier results in both propagation of pericentromeric heterochromatin beyond its normal boundary into centromeric chromatin, as well as defects in chromosome segregation during meiosis. Barrier activity is dependent upon an intact transfer RNA alanine gene (cen1 tDNA Ala ) that is transcribed from its endogenous, centromeric location [16]. In this study, we further characterize the properties of this novel cen1 tDNA Ala barrier, which differ in several aspects from previously described fission yeast barriers.

RESULTS
Barrier activity is a general feature of centromeric tDNAs tDNAs in the fission yeast genome range in size from 71-102 bp and are characterized by highly conserved internal control elements. To determine whether barrier activity is a general property of tDNAs or if the cen1 tDNA Ala has additional sequence features that convey barrier activity, we engineered strains in which cen1 tDNA Ala (including 40 bp upstream sequence and 25 bp downstream) was replaced with tDNAs that encode two different fission yeast tRNA isotypes . The presence of cen1 tDNA Ala blocks the spread of heterochromatin and permits gene expression from a centromere proximal ura4 + reporter gene (at a moderate level of 27% of wildtype ura4 + activity), as observed previously ( [16] ; Fig 1). In the absence of these sequences, however, the spread of pericentromeric heterochromatin results in a significant reduction of ura4 + reporter gene expression to only ,6% of wild-type levels. In comparison, replacement of cen1 tDNA Ala with non-centromeric tDNA Glu or tDNA Ile restored reporter gene expression (35% and 44%, respectively), suggesting that the ability to block the spread of heterochromatin at the centromere is a general property of tDNAs in fission yeast.
RNA polymerase III and TFIIIC associate with the tDNA barrier Transcription of tDNAs by RNA polymerase III (Pol III) involves the multi-step assembly of transcription factors into a pre-initiation complex that recruits Pol III [17] . Two highly conserved internal control regions, the A and B boxes, together form a specific binding site for the multisubunit transcription factor IIIC (TFIIIC). Promoter bound TFIIIC provides an interaction platform for the productive assembly of TFIIIB at a TATA box immediately upstream of tDNAs [17,18]. TFIIIB next recruits Pol III and transcription proceeds through a facilitated re-initiation pathway that involves polymerase recapture after transcription termination [19]. Previous work has indicated that cen1 tDNA Ala and its surrounding sequences are sensitive to micrococcal nuclease digestion [20] and correspond to DNAse I hypersensitive sites [21]. Furthermore, as we have shown previously, cen1 tDNA Ala is transcribed and both intact TATA and A box regions are required for robust barrier activity [16]. These findings together suggest that Pol III and its associated transcription complex gain access to the tDNA barrier at the centromere, thus forming a large stable DNA-bound complex. Genome-wide chromatin immunoprecipitation (ChIP on microarray) previously demonstrated low, but measurable levels of association of both TFIIIC and Pol III with cen1 tDNAs [22]. We confirmed this observation by ChIP analysis over a 1.6 kb region surrounding tDNA Ala . These studies revealed that both Sfc6, a TFIIIC subunit, and Rpc130, a Pol III subunit [17], were enriched at cen1 tDNA Ala 28-and 3-fold, respectively, over a noncentromeric control locus ( Figure 2A).
We next investigated whether components of TFIIIC and Pol III are present at centromeres where barrier activity has been compromised by mutation of either the A box internal control element or the upstream TATA box ( Figure 2B, [16]). Using specific primers to distinguish the modified barrier, which also bears the ura4 + reporter gene, from the wild-type tDNA Ala on the inverted repeat of cen1, we confirmed that both Sfc6 and Rpc130 were present at this locus. However, association of Rpc130 was essentially abolished in strains carrying the mutant barrier ( Figure 2B, right panel). Enrichment of Sfc6 was also investigated in these strains. The A box mutation displayed ,49% reduction in Sfc6, whereas the level of enrichment in TATA box mutants was not changed ( Figure 2B, middle panel). Importantly, both mutations have reduced levels of ura4 + expression (9.4% and 18.5% of wildtype ura4 + expression, respectively; Figure 2C left panel), and, therefore, reduced barrier activity. Taken together, these data demonstrate that Sfc6 association is not sufficient to counter the spread of pericentromeric heterochromatin [22] and suggest that robust cen1 tDNA Ala barrier activity requires the assembly of a full Pol III transcription complex.
Barrier activity is independent of tDNA orientation Transcriptional interference between Pol III and RNA polymerase II (Pol II) genes has been described in Saccharomyces cerevisiae. This tDNA ''position effect'' has been observed mostly with selected artificial constructions, in which Pol II-transcribed reporter genes were found to be inhibited 2-to 60-fold by a neighboring tDNA [23,24]. However modest tDNA position effects have also been reported to operate at native chromosomal loci, particularly when the Pol II-transcribed gene is less than 1000 bp from a tDNA [25,26,27]. Assembly of pericentromeric heterochromatin in fission yeast requires Pol II transcription of the outer repeats [28], followed by processing of the transcripts by the RNAi machinery [1]. Compellingly, siRNAs within 1kb of tDNA Ala have been identified [29], although due to both the small size of siRNAs and the sequence redundancy of fission yeast centromeres, their origin cannot be confirmed. These observations suggest the possibility that barrier activity results from a position effect that represses Pol II transcription of repetitive DNA, thus weakening local pericentromeric heterochromatin assembly at tDNA Ala . Thus, we hypothesized that barrier activity may be dependent upon the orientation of transcription of tDNA Ala . As a control, we first determined whether ura4 + reporter gene expression is influenced by cen1 tDNA Ala in the absence of heterochromatin. In strains that lack Clr4, the histone methyltransferase [30] required for pericentromeric heterochromatin assembly [31,32], expression of the centromeric ura4 + reporter gene is ,95% (+/26%) of chromosomal ura4 + , demonstrating that the Pol III complex has no significant positive or negative position effects on the ura4 + reporter gene when in a euchromatic environment. Next, we engineered strains in which the direction of both cen1 tDNA Ala and ura4 + transcription was reversed, and assayed reporter gene expression in these otherwise wild-type strains. Our results indicate that barrier activity is independent of tDNA transcriptional orientation (Table 1).

DISCUSSION
Transitions between discrete chromatin types have been studied in many organisms, from yeast to vertebrates [8,33]. In fission yeast, three structurally and genomically distinct types of boundaries have been described ( Figure 3A-C). At telomeres, chromatin domains are fluid and broad and are believed to depend on the balance between opposing effects of histone modifying activities and/or histone variants [29,34], rather than on specific genomic sequences that serve as chromatin boundary elements. A second type of boundary is defined by inverted repeats at the centromere. This class of barrier is transcribed by Pol II [29] and, like the vertebrate 59HS4 bglobin barrier [35], is enriched in active chromatin modifications such as histone acetylation and H3K4me [36]. These euchromatic modifications may act as a chain terminator for the propagation process that generates methylated pericentromeric heterochromatin [12,37]. A third type of boundary flanks the silent mating type cassettes. These inverted repeats contain several Pol III B Box motifs, and barrier activity requires the association of TFIIIC. It is important to note that B box barriers do not associate with Pol III and have been proposed to function by forming chromatin loops and/or by altering the nuclear localization of domains of chromatin, since dispersed sites of TFIIIC association coalesce at the nuclear periphery [22].
Here, we demonstrate a fourth type of barrier in fission yeast ( Figure 3D), using strains in which native tDNA barrier sequences at cen1 have been altered. Centromeric tDNA barriers require the assembly of a fully functional Pol III promoter complex. This is exemplified most strongly by strains in which the A box internal control element has been mutated ( Figure 2). In this case, Sfc6 (TFIIIC) is recruited, presumably through the unaltered B box sequences; however, it is not sufficient to counter the spread of pericentromeric heterochromatin. Barrier activity is observed only when the Pol III complex is recruited to the centromere. Thus, centromeric tDNA barriers are mechanistically distinct from previously described heterochromatin barriers in fission yeast.
Pol III is a nuclear enzyme that has been specialized to produce small non-translated RNAs in great abundance [38]. While our studies demonstrate that barrier activity is independent of the orientation of tDNA transcription, they do not address whether transcription is required for barrier activity or whether complex formation, in the absence of transcription, is sufficient. Mutations that impair the enzymatic activity of the Pol III complex, but not its assembly, have not been identified, and the development of such mutations represents an important area of future work. The fully assembled Pol III complex is at least 1.3 MDa [38], footprints nearly 150 bp along the chromatin fiber [39,40], preventing nucleosome placement [41,42,43]. Based on the data presented here, we hypothesize that the Pol III complex at cen1 likely behaves as a chain terminator (and hence a barrier) to the nucleosomedependent propagation of pericentromeric heterochromatin.
Cen1 tDNA Ala barrier activity may also result from the cooperative effects of more than one mechanism [12,37]. Interestingly, two histone demethylases (lsd1 + and lsd2 + ) associate with cen1 tDNA Ala , and the deletion of lsd1 results in the propagation of pericentromeric heterochromatin beyond tDNA Ala [44], suggesting that histone demethylases contribute to barrier activity by locally regulating histone methylation levels. Moreover, transcriptionally active tDNAs in S. cerevisiae localize to the nucleolus [45,46], raising the possibility that barrier activity correlates with nuclear position.
tDNA barriers have also been demonstrated in the budding yeast S. cerevisiae, where silencing is mediated by Sir proteins. However, not all budding yeast tDNAs are capable of barrier activity [47], due to differing affinities of TFIIIB for tDNA upstream sequences, the association of the HMG-like protein Nhp6 [48], as well as the distance between the internal control elements [47]. In contrast, we have shown here that other fission yeast tDNA sequences do retain barrier activity in our centromere specific assay (Figure 1). Notably, Pol III transcription complex assembly differs significantly between the budding and fission yeasts: fission yeast tDNAs have an upstream TATA element, centered at position 230 that participates in direct recruitment of TFIIIB [18]. Indeed, TATA motifs can be identified at all three tDNAs shown to have barrier activity in this study. Intriguingly, however, a significant difference in the strength of barrier activity was observed between strains containing tDNA Ala and tDNA Ile (p,0.001) (Fig 1). We suggest that more potent barrier activity may arise either as a consequence of variable TFIIIB affinity for different TATA boxes or because of a 23 bp increase in the tDNA Ile constructs between heterochromatin nucleating DNA at cen1 and the ura4 + reporter gene. It is noteworthy that a single tDNA flanking the left side of cen1, between pericentromeric heterochromatin and euchromatin, lacks barrier activity, despite Strains contained the ura4 + reporter gene at centromere 1 in both wild type and clr4 2 backgrounds. b Transcriptional orientation of tDNA Ala : N, native (antisense strand); R, reverse (sense strand). c Transcriptional orientation of the ura4 + reporter gene: F, forward (sense strand); R, Reverse (antisense strand). d ura4 + transcript levels were measured as in Figure 1. its association with both TFIIIC and Pol III [22], suggesting that the centromere provides a favorable environment for tDNA barrier activity.
The domain organization at each of the three fission yeast centromeres is similar and involves transitions among three types of chromatin: heterochromatin, euchromatin and centromeric chromatin. Previous genome-wide mapping of covalently modified histones showed an abrupt transition from marks associated with pericentromeric heterochromatin to those associated with centromeric chromatin [29]. Both chromatin types are essential for centromere activity [49], and mis-segregation events are observed in the absence [50,51] or mislocalization [16] of heterochromatin. Compellingly, each chromatin transition correlates with the presence of a pair or cluster of tDNAs [29], including seven different tRNA isotypes [20]. We speculated previously that the neighboring tDNAs perform an essential barrier function at fission yeast centromeres, supported by our observation of chromsosome missegregation during meiosis in the absence of tDNA Ala [16]. Thus, unlike conventional barriers that ensure appropriate euchromatic gene expression in euchromatin (Figure 3; A-C), cen tDNA barriers may ensure proper centromere assembly and function by protecting the central core domain of centromeric chromatin from the distinct structural changes and epigenetic features associated with pericentromeric heterochromatin.

Plasmid DNAs
Plasmids SM353 and SM349 [16] contain 1.78 kb of imr1 sequence, including the cen1 tDNA Ala and a ura4 + reporter gene at the unique HindIII site. These plasmids were modified by site directed mutagenesis to introduce unique ClaI (245) and EcoRI(+20) sites flanking tDNA Ala . tDNA Glu , tDNA Ile , and reverse tDNA Ala were amplified from genomic DNA with primers containing either EcoRI or ClaI restriction sites. The resulting PCR products were purified, digested with EcoRI and ClaI, and ligated with vector that had been digested with EcoRI and ClaI to liberate tDNA Ala . Resulting plasmids were purified, sequenced and digested with Not1 and KpnI to generate a 3.5 kb product used for yeast transformation.

Fission yeast strains
The genotypes for S. pombe strains used in this study are listed in Table 2. Media were prepared according to standard procedures [52]. All transformations were performed as in [16]. At least three independent transformed strains were established from each construct. Strains were crossed 2-3 times to KFY 3/4 before further analyses.

Real-Time RT-PCR
Yeast were grown in YES to 5610 6 cells/ml at 32uC. cDNA was prepared by oligo dT primed RT-PCR (Invitrogen). Real-time PCR was performed in the presence of SYBR Green on a Bio-Rad iCycler with the following primer pairs for ura4 + : 5026, 59-TGATATGACCCCAAGAAGCA-39; 5027, 59-AAAAACTGG-TGGCCTTAGGT-39 and act1 + : 7453, 59-AATCCAACCGT-GAGAAGATGA-39; 7454, 59-ACGACCAGAGGCATACAA-AGA-39. A standard curve of at least four orders of magnitude was generated with genomic DNA isolated from the wild-type strain (ura4 + KFY2). Data were analyzed with iCycler iQ Optical System Software. Data were analyzed further only if the PCR efficiency was between 90-110% and the correlation coefficient was between 0.990 and 1.0. ura4 + levels were normalized to act1 + levels and quantified relative to the wild-type strain. At least three biological replicates were peformed for each mutant. CHIP ,2.5610 8 cells were grown to mid-log phase at 32uC, then shifted to 18uC for two hours. Cells were pelleted, fixed in 3% paraformaldehyde for 30 min, pelleted and washed twice in ice-cold PBS. Cells were resuspended in 10 ml of 10 mM dimethyladipidate, including 0.0025% DMSO and incubated at room temperature for 30 min. Cells were pelleted and washed twice in ice cold PBS. ChIP was performed as previously described [16]. 7ul of anti-Sfc6 or anti-Rpc130 antisera [18]; [17] were used for each ChIP. PCR products were quantified as described for RT-PCR. Enrichment at each query locus was calculated relative to the euchromatic act1 + value and then corrected for the ratio obtained for the input PCR.