Chromatin insulators separate active from repressed chromatin domains. In yeast the RNA pol III transcription machinery bound to tRNA genes function with histone acetylases and chromatin remodelers to restrict the spread of heterochromatin. Our results collectively demonstrate that binding of TFIIIC is necessary for insulation but binding of TFIIIB along with TFIIIC likely improves the probability of complex formation at an insulator. Insulation by this transcription factor occurs in the absence of RNA polymerase III or polymerase II but requires specific histone acetylases and chromatin remodelers. This analysis identifies a minimal set of factors required for insulation.
THE eukaryotic genome can be divided into two chromatin states. Heterochromatin, which in most metazoans constitutes the vast majority of the genome, is condensed and contains primarily repetitive DNA sequences while euchromatin is accessible chromatin that is gene rich. Euchromatic and heterochromatic domains form a mosaic along the chromosome and often these functionally competing chromatin states reside adjacent to one another.
Gene activity in eukaryotes occurs within the context of these chromatin domains and is regulated by DNA sequence elements. Enhancers and locus control regions (LCRs) positively regulate genes while silencers repress genes. Typically these regulatory elements function within specific chromatin domains and contribute toward the formation of these domains. Enhancers and LCRs bind various combinations of transcription factors, which in turn recruit accessory proteins such as histone modifiers, chromatin remodelers and histone subtypes to open chromatin domains thereby generating a euchromatic state that is amenable to stable gene activation. Silencers on the other hand bind specific factors and recruit histone modifiers and repressor proteins that spread and encompass DNA sequences into a condensed state that is inaccessible to various enzymatic probes.
The euchromatic and heterochromatic domains are separated from one another by DNA regulatory elements called insulators (Bi and Broach 2001; Fourel et al. 2004; Valenzuela and Kamakaka 2006). Insulators are integral to proper gene regulation and have many of the same properties as promoters and occasionally are promoters of genes (Donze and Kamakaka 2001; Fourel et al. 2002; Bartkuhn et al. 2009). Insulators bind various transcription factors and these factors recruit chromatin-modifying activities to delineate chromatin domains.
Yeast tRNA insulators:
Protein translation requires tRNAs encoded by tRNA genes. In eukaryotes, tRNA genes utilize specific multisubunit transcription factors TFIIIC and TFIIIB to mediate synthesis of tRNA by RNA pol III (Geiduschek and Kassavetis 2001). Transcription of a tRNA gene involves the binding of TFIIIC to conserved intragenic promoter elements called box A and box B. TFIIIC binding leads to recruitment of TFIIIB to a ∼50-bp AT-rich region upstream of the start site of transcription. TFIIIB recruitment to the gene results in the recruitment of RNA pol III and transcription of the gene (Schramm and Hernandez 2002). These motifs within and upstream of a tRNA gene affect its transcription efficiency and function.
In eukaryotes the tRNA genes are a special form of repetitive DNA, present as multiple copies in the genome and are either arranged as small clusters or individual dispersed copies throughout the genome (Percudani et al. 1997). Since tRNA genes are dispersed throughout the genome this leads to questions of whether their location influences other chromosomal processes. The presence of a tRNA gene results in replication pausing at the gene (Deshpande and Newlon 1996), nucleosome positioning immediately adjacent to the gene (Morse et al. 1992), and preferential retroviral integration immediately upstream of the gene (Kirchner et al. 1995; Devine and Boeke 1996). Furthermore, in the yeast Saccharomyces cerevisiae and Schizosaccharomyces pombe, tRNA genes have been found to function as chromatin insulators restricting the spread of heterochromatin (Donze et al. 1999; Donze and Kamakaka 2001; Noma et al. 2006; Scott et al. 2006).
Silenced chromatin domains in the yeast S. cerevisiae and S. pombe are distinct, use different repressor proteins, and are formed via different mechanisms (Dhillon and Kamakaka 2002; Grewal and Moazed 2003). Silencing in S. cerevisiae is mediated by the Sir proteins while in S. pombe, silencing initiation utilizes the RNAi machinery and involves the association of Swi6p repressor protein with histones. Interestingly, in both systems, tRNA genes are able to block the spread of silencing. A tRNAThr gene located adjacent to the silenced HMR locus in S. cerevisiae blocks the spread of silencing, and deleting this gene leads to an increased spread of Sir-mediated silencing (Donze et al. 1999; Donze and Kamakaka 2001). Similarly in S. pombe, clusters of tRNA genes mediate barrier activity restricting the spread of Swi6p containing heterochromatin (Noma et al. 2006; Scott et al. 2006). In both organisms, data indicate that the transcription factors that bind the tRNA genes are required for insulation. In S. cerevisiae, mutations in TFIIIC and TFIIIB but not pol III affect tRNA-mediated insulation (Donze and Kamakaka 2001), while in S. pombe TFIIIC binding has been shown to be required for insulation (Noma et al. 2006).
Besides acting as insulators to the spread of silenced chromatin, TFIIIC bound elements also act as insulators to repressed chromatin (Simms et al. 2004, 2008; Valenzuela et al. 2006) suggesting that tRNA genes help partition the genome into functionally distinct domains.
The sequences of tRNA genes are highly conserved and the substitution rate of sequences within functional tRNA genes is less than that observed at protein encoding genes suggesting a strong selection (Withers et al. 2006). Pseudogenes are DNA sequences that show homology with a known functional gene but where mutations have rendered the gene product inactive. In tRNA pseudogenes, high rates of nucleotide substitutions are observed during the death and degeneration of the genes (Withers et al. 2006). Several tRNA relics have been identified in S. cerevisiae and S. pombe. Genomewide mapping of the transcription factors TFIIIC, TFIIIB, and RNA pol III identified sites in S. cerevisiae that bind either TFIIIC alone or TFIIIC and TFIIIB but not RNA pol III, suggesting ancient genes (Harismendy et al. 2003; Roberts et al. 2003; Moqtaderi and Struhl 2004). Nine such sites called Extra TFIIIC (ETC) loci have been identified in S. cerevisiae. Similarly in S. pombe TFIIIC binding sites devoid of RNA pol III have also been reported called chromosome-organizing clamp (COC) loci (Noma et al. 2006). Interestingly most of the COC sites are located adjacent to RNA pol II transcribed genes and may aid in regulating these genes.
To investigate the molecular mechanism by which RNA pol III transcription factors function in insulation we decided to use ETC loci to determine the minimal factor requirement for tRNA promoter-mediated insulation in S. cerevisiae. We find that merely recruiting TFIIIC to a DNA element will insulate a gene, but the presence of TFIIIB improves the ability of TFIIIC to insulate. Efficient insulation by TFIIIC does not require either RNA pol II or RNA pol III, but mutations in specific histone modifiers and remodelers affects TFIIIC-mediated insulation, thus delineating the key components required for insulation.
MATERIALS AND METHODS
Deletions and integrations were performed by homologous recombination, using PCR products amplified with the Expand High Fidelity PCR System (Roche) and standard lithium acetate high-efficiency transformation procedure (Ito et al. 1983). In strain LOU162 (HMR-pUC18∷URA3 sir2Δ), the 1.2-kb wild-type barrier element located to the right of the HMR locus was deleted and replaced by a 1.2-kb DNA fragment derived from pRS. The URA3 cassette was then integrated at the pRS sequences, 590 bp downstream from the HMR-I silencer. B box-containing elements were PCR amplified and used to replace the URA3 cassette at pRS sequences in strain LOU162. After transformation, FOA-resistant colonies were analyzed by PCR and crossed to obtain the final phenotypes.
Cells were grown in YPD to an OD600 of 2.0, and cross-linked with 1% HCHO for 10–15 min followed by neutralization of the HCHO with 0.125 m glycine. Cells were washed with PBS and FA-140 and lysed in FA-140 with protease inhibitors using glass beads and a bead beater.
Lysed cells were sonicated twice, first with a Diagenode Bioruptor and then with a Branson cup-horn sonicator. This sequential sonication was important as it resulted in DNA with an average length of 300 bp. The sonicated chromatin solution was centrifuged to remove insoluble cellular debris.
Immunoprecipitation reactions were performed with the desired antibodies and protein A/G beads overnight. H3, Rpc40-HAp, Tfc1-HAp, and Pol II were chromatin immunoprecipitated using Rabbit polyclonal antibodies—ab1791 against the C terminus of H3 (Abcam), the Anti-HA.11 antibody (from Babco), or the Pol II monoclonal antibody 8WG16 (from Covance), in combination with protein G plus/protein A agarose (Calbiochem). Protein Bdp1-TAP was immunoprecipitated with IgG-sepharose 6 Fast Flow (Amersham Biosciences). Beads were washed sequentially with buffer FA-140, FA-500, LiCl/Det, and finally TE.
The immunoprecipitated chromatin was eluted off the beads by the addition of 10% Chelex-100 and incubated at 100° for 10 min. Proteinase K was added to the mixture and incubated at 55° for 30 min followed by a second incubation at 100° for 10 min. The DNA bound to the Chelex beads was eluted off of the beads with water.
The input DNA and immunoprecipitated DNA were quantitated using the PicoGreen dsDNA quantitation reagent (Molecular Probes) and a Perkin Elmer Viktor3 Fluorescence spectrophotometer. (Lambda DNA was used to construct a standard graph).
Equal amounts of immunoprecipitated DNA and input DNA (usually between 50 pg and 200 pg) were used for real-time PCR analyses.
Real-time PCR-based amplification of the DNA was performed using specific primers. All primer pairs were initially screened (on average we tested three pairs of primers for each PCR fragment) for the absence of primer dimers or cross-hybridization. Furthermore, only primer pairs with similar amplification efficiencies were used.
Quantitative chromatin immunoprecipitation analysis was performed using a Corbett Life Science Rotor Gene 6000 machine. The detection dye used was SYBR Green (2× Immomix from Bioline and Platinum SYBR Green from Invitrogen). Real-time PCR was carried out as follows: 95° for 5 min (1 cycle), 95° for 15 sec, 53–58° for 20 sec, and 68–70° for 20 sec (45 cycles). The fold difference between immunoprecipitated material (IP) and total input sample for each qPCR amplified region was calculated as described in Litt et al. (2001), following the formula IP/Input = (2InputCt − IPCt).
All of the graphs shown represent the mean values and standard errors of at least two independent cross-linked samples with each sample being immunoprecipitated twice with the same antibody with the exception of RNA pol II in Figures 1 and 2.
Mapping the transcription machinery:
To study the contribution of various tRNA bound factors in insulator function we initially quantitatively mapped the distribution of various subunits of TFIIIC, TFIIIB, and RNA pol III to the native tRNA boundary at HMR. The tRNA gene present at the HMR boundary also has sequence homologs located elsewhere in the genome. We therefore devised oligonucleotides that would allow us to quantitatively map the distribution of transcription factors specifically across a 1-kb region centered on the HMR-tRNA insulator. We used a control probe for all our ChIP analysis located 500 bp from TEL6R in an intergenic region. There are no binding sites for any of the RNA pol III transcription factors in this region of the genome.
To map the distribution of various subunits of the tRNA transcription machinery across the region we used strains containing tagged proteins Tfc1, Bdp1, and Rpc40 (Roberts et al. 2003). Tfc1 is an essential subunit of the TFIIIC complex while Bdp1 is an essential subunit of TFIIIB and Rpc40 is an essential subunit of RNA pol III and RNA pol I. These three subunits were used as representatives of the three complexes required for tRNA transcription. The tagged strains (ROY 3931 and ROY4548) had no significant growth defects and behaved similarly to an untagged strain (ROY1685), suggesting that the tags had not severely compromised their essential functions though the Bdp1-tagged strain dominantly recruited RNA pol II (see supporting information, Figure S1).
Mapping the distribution of Tfc1 across the silenced HMR domain (Figure 1A) demonstrated that this protein had a unique binding site adjacent to HMR at the tRNATHR gene. Using nested probes we found increased binding of TFIIIC in regions upstream of the tRNA and at the tRNA but not downstream of the gene (compare probes IV, IV′, and V). A similar binding profile was observed for Bdp1 with increased binding upstream of the gene (Figure 1B). The binding for Rpc40 (subunit of RNA pol III) was present both upstream as well as downstream of the gene (Figure 1C). These profiles are consistent with the observations that the transcription activators for the tRNA bind the 5′ and coding regions of the gene while pol III transcribes across the entire gene.
Having mapped the distribution of RNA pol III transcription machinery at the HMR tRNA insulator, we next mapped these proteins at other loci. We mapped these factors at specific tRNA genes as well as the U6 gene, which is also transcribed by RNA pol III (Figure 2). At transcribed tRNA and U6 genes, our quantitative data indicated that the levels of RNA pol III and Bdp1 were approximately equivalent at all loci compared. In contrast the levels of TFIIIC varied from locus to locus. There was an elevated level of Tfc1 at tRNATHR-GR1 and tRNATHR-KL while tRNATHR-NL1 and SUP53 had a twofold reduction in levels of Tfc1 and the U6 gene had a very significant reduction in the level of Tfc1 compared to the HMR tRNA.
All ETC loci bind TFIIIC and some of these loci also bind TFIIIB but none of these loci bind RNA pol III (Roberts et al. 2003; Moqtaderi and Struhl 2004). These loci are scattered throughout the yeast genome. We next mapped the RNA pol III factors at a few ETC loci that bind the transcription factors TFIIIB and/or TFIIIC but do not recruit RNA pol III (Figure 2). We analyzed three different ETC loci: we investigated binding at a previously identified pseudo tRNAARG gene located on chromosome VII between genes TIM21 and RPL26B (Roberts et al. 2003). In this article we will refer to this locus as ETC9. We also investigated two previously identified loci on chromosome XV—one located in the coding region of PPM2, which we will refer to as PPM2-ETC and the second located between genes PPM2 and ARG8 called ETC2 (Roberts et al. 2003; Moqtaderi and Struhl 2004). All three loci have sequences with homology to the B box present at all RNA pol III transcribed genes and in close proximity to these B-box sequences is a sequence with homology to a consensus A box.
Upon mapping RNA pol III subunit Rpc40 at these loci (Figure 2C) we found that unlike the tRNA genes, ETC2 and PPM2-ETC did not bind RNA pol III to any significant level. ETC9 bound significant amounts of Bdp1 though the levels were lower compared to the functional tRNA genes (Figure 2B). The other ETC loci did not bind any detectable Bdp1. ETC2 and ETC9 both bound Tfc1 while the PPM2-ETC locus did not bind any pol III factors (Figure 2A). These results confirm the previous observations that ETC2 and ETC9 are genuine ETC loci (Moqtaderi and Struhl 2004) but PPM2 is not a functional ETC locus in the W-303 strain background used in this study.
In S. pombe, RNA pol II is found at the IR insulators along with TFIIIC (Noma et al. 2006) and we therefore mapped RNA pol II at the HMR tRNA using a monoclonal antibody (8WG16) specific for the unmodified large subunit of RNA pol II (Figure 1D). This analysis demonstrated the absence of the large subunit of RNA pol II across the entire HMR domain and we also did not observe any RNA pol II at or near the tRNA gene. We also tested the distribution of RNA pol II at other tRNA genes as well as at the three ETC loci (Figure 2D). All tRNA genes except SUP53 had extremely low levels of RNA pol II protein in their immediate vicinity. There were significant levels of RNA pol II near SUP53 presumably because this gene is located in a very short intergenic region immediately adjacent to a RNA pol II regulatory region. None of the ETC loci bound significant amounts of RNA pol II, suggesting that pol II recruitment to ETC elements does not occur in S. cerevisiae. This obviously does not rule out the possibility that ETC loci facilitate or regulate pol II-mediated transcription of neighboring genes.
Our results on the distribution of tagged pol III factors are consistent with previously published results on the distribution of pol III factors (Harismendy et al. 2003; Roberts et al. 2003; Moqtaderi and Struhl 2004), demonstrating that the presence of the tags does not affect the distribution of these proteins. However, interestingly, when we mapped RNA pol II in the Bdp1/Rpc40 tagged strain we surprisingly observed robust levels of RNA pol II immediately upstream of the tRNA genes tested (Figure S1, A and data not shown). This binding was present at all the tRNA genes tested and was lost when a internal promoter of a tRNA gene was deleted, demonstrating that pol II was being recruited by these tagged proteins (Figure S1, A). In contrast to the tagged Bdp1 strain (ROY4548), in the tagged Tfc1 strain (ROY3931) we did not observe any significant recruitment of RNA pol II (Figure S1, B). In future, care will need to be taken in the use of tagged proteins.
ETC loci function as insulators:
Specific tRNA genes function to block the spread of heterochromatin and previous experiments had demonstrated that mutations in TFIIIC and TFIIIB weakened insulation but mutations in RNA pol III had no effect, suggesting that transcription was not necessary for insulation (Donze and Kamakaka 2001). We decided to take advantage of the differential binding of pol III factors at ETC loci to determine the minimal DNA binding transcription factors necessary for insulation. A modified HMR locus was created by the insertion of a 300-bp fragment containing a tRNA gene or ETC loci between the HMR-E silencer and the MATa1 gene. In strains where a functional insulator was present, silencing emanating from the silencer would be blocked and the reporter gene would be insulated and be active. If the insulator were not able to function, then the gene would be repressed. These strains were assayed for barrier activity using a mating assay. If the DNA fragment had barrier activity, then the strain would be a nonmater and conversely, if barrier activity was absent, then the silent domain would spread to repress the MATa1 reporter gene and the strain would mate. This analysis (Figure 3A) indicated that tRNATHR-KL, tRNATHR-NL1, SUP53, and the U6 gene were unable to block the spread of silencing while only tRNATHR-GR1 and the HMR-tRNATHR were able to block the spread of silencing. These data also indicated that the steady state levels of Tfc1 at specific sequences did not correlate with the ability of those sequences to block the spread of silencing since the tRNATHR-KL gene on chromosome XI, which had the same levels of Tfc1 bound (see Figure 2) as tRNATHR-GR1 on chromosome VII and the HMR tRNATHR on chromosome III, was not as efficient in blocking the spread of silencing when inserted between the HMR-E silencer and the a1 reporter gene.
We next analyzed the ability of the ETC loci to function as insulators (Figure 3A). ETC2 and ETC9 were both able to block the spread of silencing while PPM2-ETC was not able to block the spread of silencing. Both ETC2 and ETC9 contained lower levels of Tfc1 than tRNATHR-KL, tRNATHR- NL1 or SUP53, indicating that the absolute levels of TFIIIC binding at their native sites in the genome are not a good indicator of whether an element will function as an insulator. These results collectively also suggest that Tfc1-bound DNA elements might be necessary for insulation but were not sufficient for insulation since tRNATHR-KL and tRNATHR -NL1, and SUP53 bound Tfc1 but were not able to block the spread of silencing.
The HMR tRNATHR gene that functions as an insulator does so in either orientation (Donze and Kamakaka 2001). We therefore decided to test whether the ETC loci could also function in both orientations or whether they were orientation dependent (Figure 3B). Using the same mating assay (coupled with serial dilutions) we found that ETC9 functioned as a barrier in both orientations but interestingly ETC2 only functioned in one orientation while PPM2-ETC was still not able to function as an insulator in either orientation. The reason for the orientation dependence of ETC2 is currently not known.
ETC multimerization improves barrier function:
The ETC fragments tested were not fully functional as insulators. We therefore asked if multimerization of the ETC9 fragments would improve their ability to function as insulators. We generated constructs with one, two, or three tandem copies of ETC9 loci inserted between the silencer and the reporter gene. Mating assays indicated that three copies of ETC9 inserted between the silencer and the reporter gene resulted in slightly better insulation (Figure 3C). We also multimerized PPM2-ETC, which was unable to block silencing when present as a single copy. Inserting two or three copies between the silencer and the reporter also did not result in barrier activity, further demonstrating that ETC-PPM2 was not a bona fide ETC locus.
Since ETC-PPM2 was unable to block silencing and since ETC2 functioned only in one orientation, we decided to generate a construct with a tandem array of consensus TFIIIC binding sites. In S. pombe, these sites recruit TFIIIC but not RNA pol III and the three TFIIIC binding sites (B boxes) present in a tandem array have been shown to function as insulators and restrict the spread of Swi6p-repressed heterochromatin in this organism (Noma et al. 2006). We therefore decided to ask whether TFIIIC binding alone could block the spread of Sir proteins in S. cerevisiae. We inserted three consensus B-box binding sites in a tandem array between the silencer and the reporter gene and monitored expression of the reporter gene (Figure 3C). Our mating data showed that unlike S. pombe, these three TFIIIC B-box sites were not sufficient to significantly block the spread of silencing. We then asked whether increasing the number of TFIIIC binding sites would lead to insulation. We inserted either 9 or 12 B-box sites and monitored insulation. Upon inserting either 9 or 12 sites, we now observed insulation, suggesting that increased TFIIIC binding was able to restrict the spread of silencing.
TFIIIC function was necessary for ETC9 and B-box mediated insulation:
Our ChIP data had indicated that Tfc1 bound both ETC9 and ETC2 at their native euchromatic loci on chromosomes VII and XV, respectively (Figure 2). Furthermore the tandem array of B boxes would be expected to bind TFIIIC as well. If these elements functioned in insulation via the RNA pol III transcription factors then conditional mutations that weaken the function of these factors should reduce the ability of the tRNA/ETC loci to act as insulators. Temperature-sensitive mutations in Tfc3 have been shown to affect HMR tRNATHR-mediated insulation (Donze and Kamakaka 2001). We tested whether this mutation also led to a reduction in ETC-mediated insulation (Figure 4). At the semi-permissive temperature (30°), ETC9-mediated insulation was reduced in a tfc3 mutant but surprisingly we did not observe any reduction in insulation mediated by ETC2, suggesting that ETC2-mediated insulation was probably due to other factors and not TFIIIC or a combination of both. We also investigated the nine B-box synthetic insulators and found that mutations in Tfc3 resulted in near complete loss of insulation from the nine B-box insulators. Due to these results we focused our further analyses on ETC9 and the tandem B-box arrays.
Rpc40 is absent from ETC loci at HMR:
Since ETC9 was able to function as an insulator at HMR, we investigated whether RNA pol III was necessary for insulation. ETC9 at its native site on chromosome VII does not bind RNA pol III and is not transcribed (Roberts et al. 2003; Guffanti et al. 2006). We inquired whether the transposed ETC9 locus, adjacent to HMR on chromosome III, recruited RNA pol III. To allow a direct comparison between ETC9 and the HMR-tRNATHR in a sequence neutral environment, two strains were constructed in which a 1-kb fragment encompassing the HMR-tRNATHR and the two TY LTRs were replaced with a fragment of the same length from a pRS vector. We then inserted either 2xETC9 or the HMR-tRNATHR into the pRS sequence at approximately the same distance from HMR-I as the native HMR-tRNATHR. We compared binding of the RNA pol III subunit Rpc40 at these loci and found nearly background levels of RNA pol III at the transposed ETC9 locus relative to the HMR-tRNA insulator, suggesting that ETC9 was most likely not transcribed even though it could function as an insulator (Figure 5A). As a control we also compared the binding of the RNA pol III transcription factor Bdp1 in these two strains (Figure 5B). While Bdp1 levels were approximately twofold reduced at ETC9 compared to the HMR-tRNATHR (Figure 5B) the levels were significantly above background. These results collectively suggest that RNA pol III-dependent transcription is unlikely to be required for insulation.
We next investigated binding of the RNA pol III factors at the three B boxes and ETC2 when these loci were integrated in place of the tRNA insulator at HMR on chromosome III. In sir2Δ strains there were significant levels of binding of Tfc1 at both ETC2 and the 3xB-box elements but very negligible binding of Bdp1 (Figure 6, A and B). However in strains containing Sir2 and silencing, binding of Tfc1 was dramatically reduced at both elements. These results suggest that the presence of the Sir proteins blocks recruitment of TFIIIC to these elements and these data help explain why these elements were not able to function as insulators in a TFIIIC-dependent manner (see Figure 4). Furthermore the results obtained with the three B boxes (Figure 3C and Figure 6A) suggested that insulator function was most likely a result of direct competition between Sir protein spreading and TFIIIC binding at the insulator, thus suggesting that a key step in the formation of an insulator is the stable binding of TFIIIC.
TFIIIC-mediated insulation can be overcome by increased Sir proteins:
Since ETC9 appeared to insulate genes in the absence of transcription but required TFIIIC, we hypothesized that insulation was a consequence of competition between TFIIIC binding and Sir protein spreading. This would predict that overexpressing Sir proteins should overcome the insulator.
To test this hypothesis we generated three reporter strains. In each strain the HMR-E silencer repressed the MATa1 gene at HMR on chromosome III. One strain had the HMR tRNATHR insulator placed between the silencer and the reporter gene, the second strain had 3xETC9 loci inserted between the silencer and the reporter while the third strain had no insulator. In the absence of the insulator, the reporter gene was silenced, but in the presence of either the tRNA or ETC9 loci, silencing was blocked and the gene was active (Figure 6C). We then assayed expression of the reporter in these strains following overexpression of Sir3. Sir3 overexpression significantly overcame both the tRNA and ETC9 insulators, resulting in repression of the reporter. These results strongly support our model that insulation is a consequence of a competition between TFIIIC binding and Sir protein spreading.
Specific acetylases and remodelers are required for ETC-mediated insulation:
At the native HMR tRNATHR barrier, the silenced domain is restricted from spreading by the tRNA gene in conjunction with the action of specific histone acetylases and chromatin remodelers. Sas2, Eaf3, Gcn5, Isw2, and Rsc (Donze and Kamakaka 2001; Jambunathan et al. 2005; Oki and Kamakaka 2005) are required for efficient restriction of the silenced domain while Sas3, Hat1, and Swr1 are not. We wished to know whether the same acetylases and remodelers were utilized by the ETC loci to restrict the spread of the silenced domain or whether these loci utilized different enzyme complexes. We used various mutant strains lacking specific histone acetylases and chromatin remodelers and asked whether in the absence of these enzymes insulation mediated by the ETC locus was affected (Figure 7). We tested mutations in Hat1, Sas2, and Eaf3 as representatives of histone acetylase subunits and we tested mutations in Isw2 and Rsc2 as representatives of chromatin remodelers. Most of the mutants affected insulation mediated by ETC9 and the 9x B-box insulators to some degree. Loss of Sas2, Eaf3, Isw2, or Rsc2 led to reduction in insulation while mutations in Hat1 and Sas3 (data not shown) had no effect. Isw2 and Eaf3 mutants had a weak effect on the tRNA-mediated insulation but had a larger effect on ETC9 and 9x B-box-mediated insulation. This may reflect the fact that the ETC insulators are weaker than the tRNA and loss of these enzymes weakens these insulators further.
RSC is required for transcription factor loading at the tRNA:
While histone acetylases such as Sas2 and Eaf3 are not specifically recruited to tRNA genes and function in insulation independently of the tRNA (Oki and Kamakaka 2005), the Rsc chromatin remodeler localizes to tRNA genes (Damelin et al. 2002; Ng et al. 2002) where it evicts histones (Parnell et al. 2008) and mutants in RSC affect tRNA-mediated insulation (Jambunathan et al. 2005). We therefore decided to investigate the role of RSC in insulation.
We initially determined the phenotype of RSC mutants on the native HMR-tRNA boundary. We generated two isogenic MATα strains that differed at the HMR boundary. In both strains, the native promoter of the MATa1 gene at HMR was deleted (HMRa Δ p) and a functionally active MATa1 gene under its own promoter was inserted in the intergenic region between the insulator and the GIT1 gene. In the wild-type strain the tRNA insulator was intact while in the barrier-deleted strain, the tRNA gene was deleted. The MATα strains were monitored for expression of the MATa1 reporter gene by mating assays. When the reporter is repressed, the strains should mate with a tester strain, form diploids, and grow on selective plates but when the reporter is active the strain should be unable to form diploids and should not grow on selective plates. Our analyses indicated that as predicted, in the wild-type strain the MATa1 gene residing outside the HMR boundary was fully active, while in the tRNA delete strain there was no apparent insulation and the reporter gene was almost fully repressed (Figure 8A). In contrast, in a rsc2Δ strain we observed significant loss of tRNA-mediated insulation as manifested by repression of the reporter gene thus confirming the previous results (shown in Figure 7).
Mutations in RSC phenotypically weaken barrier activity mediated by the tRNA at HMR. We therefore asked whether the distribution of Sir3p was altered in cells deficient for these activities. We mapped the distribution of Sir3p in a wild-type strain and a strain lacking Rsc2 (Figure 8B). Quantitative ChIP experiments indicated that the lack of Rsc2 led to a reproducible twofold increase in the levels of Sir3p outside of the HMR tRNATHR barrier. These results suggested that RSC directly or indirectly functioned to regulate Sir protein spreading at HMR. These results were also consistent with the observation that loss of RSC led to partial loss of tRNA-mediated insulation (see Figure 8A).
The HMR tRNATHR gene has been shown to reside in a histone-depleted region (Oki and Kamakaka 2005) and RSC is involved in evicting nucleosomes from chromatin (Saha et al. 2006). It was therefore possible that RSC was required to evict nucleosomes at the HMR tRNA insulator and loss of RSC would then be predicted to result in the reformation of a nucleosome at the tRNA insulator. We quantitatively mapped the distribution of histone H3 in a wild-type strain (ROY4562) and in a strain lacking Rsc2 (ROY4563) (Figure 8C). Consistent with previous data (Oki and Kamakaka 2005), histones were depleted at the HMR tRNA in the wild-type strain. In a rsc2Δ strain, histone H3 levels increased at the tRNA insulator, suggesting that RSC played a direct or indirect role in nucleosome eviction at the tRNA. HMR-I is normally nucleosome free due to binding of silencer proteins and in a Rsc2 mutant the levels of H3 at HMR-I were unchanged, demonstrating that this RSC-dependent H3 eviction effect was restricted to the tRNA insulator.
We next mapped the distribution of RNA pol III transcription factors Tfc1, Bdp1, and Rpc40 in wild-type strains (ROY4562 and ROY 4548) and in strains lacking Rsc2 (ROY4563 and ROY4564). Our data demonstrated that occupancy of all three factors was reduced in a Rsc2 mutant compared to the wild-type strain. While the reduction was small, we observed the same effects with all three RNA pol III transcription complexes and these effects were observed in different strains, suggesting that these effects are significant. The fact that we did not observe large changes in Tfc1, Bdp1, or H3 occupancy at the tRNA in a rsc2Δ strain was consistent with our phenotypic data, demonstrating that loss of Rsc2 resulted in only a partial loss of insulator activity.
RSC was also required to evict nucleosomes at ETC9-HMR:
To investigate the mechanism of RSC-mediated insulation at the ETC loci we first mapped histone occupancy at the ETC9 locus on chromosome VII in the presence and absence of Rsc2. We used probes at the promoters of TIM21 and RPL26B genes that flank ETC9 as well as a probe at the ETC9 locus itself. Using qPCR we found that in wild-type strains histone H3 occupancy was reduced at all three probes on chromosome VII compared to control probes near the telomeres of chromosome VIR (Figure 9A). In the absence of Rsc2, we did not observe any significant change in histone occupancy at these sites, indicating that histone eviction/sliding at this site was either not dependent on the RSC complex or was redundant with other chromatin remodelers/modifiers.
We then mapped histone occupancy at the ETC9 locus when this locus was transposed and inserted adjacent to HMR on chromosome III. If the mechanism of insulation were conserved between the tRNA and ETC9 then we would expect nucleosome loss at the HMR-ETC9 insulator at this site. We mapped histone H3 distribution with several probes at the HMR-I silencer as well as probes adjacent to HMR-ETC9 in the presence and absence of Rsc2 (Figure 9B). Histone H3 was depleted at the HMR-I silencer as was expected due to the binding of ORC and Abf1 to the silencer. Furthermore there was reduced H3 binding at the HMR-ETC9 insulator. However the H3 levels at HMR-ETC9 were not identical to that observed for ETC9 at its native site on chromosome VII. Interestingly upon loss of Rsc2, histone levels increased almost twofold at the HMR-ETC9 locus but not at the control telomeric loci, suggesting that RSC was directly or indirectly involved in evicting nucleosomes from the insulator. HMR-I is normally nucleosome free due to binding of silencer proteins and in a Rsc2 mutant the levels of H3 at HMR-I were unchanged, demonstrating that this RSC dependent H3 eviction effect was restricted to the TFIIIC-dependent ETC9 insulator.
Heterochromatin domains are formed by the recruitment of silencing complexes to silencers followed by the propagation of these complexes along the nucleosomal filament (Rusche et al. 2003). The extent of heterochromatin spreading is dependent upon the amount of the silencing complexes but DNA elements can actively block the spread of silencing complexes by creating local regions of chromatin that are refractory to the binding and spreading of repressor proteins (Oki and Kamakaka 2005). These DNA elements are often referred to as barriers. At the native HMR domain a tRNA gene promoter functions to block the spread of silencing (Donze et al. 1999; Donze and Kamakaka 2001). tRNA genes are also barriers to heterochromatin in S. pombe even though heterochromatin in S. pombe is different from silenced chromatin in S. cerevisiae (Noma et al. 2006; Scott et al. 2006). Given that this property of tRNA genes is conserved we set out to identify a minimal set of factors that are required for tRNA gene-mediated barrier function in S. cerevisiae.
TFIIIC is necessary for insulation:
Our study showed that 9x B boxes alone were able to function as an insulator and TFIIIC was able to block the spread of silencing in the absence of TFIIIB or RNA pol III. However, this insulation was not robust and multiple binding sites for TFIIIC were required for efficient insulation. This analysis does demonstrate that TFIIIC binding to a DNA element, in the absence of TFIIIB or RNA pol III, can insulate genes from repression. A very recent report also demonstrates that a single ETC locus that only recruited TFIIIC (ETC4) is also able to function as an insulator (Simms et al. 2008). While this report did not quantitatively measure binding of the various factors it is consistent with our conclusions that TFIIIC binding to a DNA sequence can result in insulation even in the absence of TFIIIB or RNA pol III.
Our result also suggested that occupancy of the TFIIIC binding sites was dynamic and in direct competition with the spreading Sir repressors. In the presence of three B boxes, Sir proteins displaced or prevented TFIIIC from binding but nine B boxes were able to insulate presumably because the probability of binding of TFIIIC to chromatin at any particular time was greatly increased. A similar situation was previously observed with tRNA genes. A tRNA gene (tRNATHR-NL1) that was not effective as an insulator in one copy, became an effective insulator when two copies were inserted (Donze and Kamakaka 2001). Furthermore, overexpressing the Sir proteins partially overcame an effective barrier (tRNA and ETC9) and these data collectively suggest that dynamic competition between TFIIIC binding and Sir repressor protein spreading determines the extent to which silenced domains spread.
While the ability to recruit TFIIIC is necessary for a DNA element to function as an insulator it is not sufficient. We showed that in the absence of the Sir proteins, TFIIIC could bind to three B boxes, but in the presence of the Sir proteins this binding was significantly reduced and the 3x B boxes were unable to insulate a reporter gene. Similarly we showed that some tRNATHR genes efficiently recruited TFIIIC and TFIIIB at their original sites in euchromatin, but when these genes were moved to the silenced HMR domain, they were unable to block the spread of silencing emanating from the HMR-E silencer. DNA sequences upstream and downstream of various tRNA genes have been shown to affect the stability of binding of TFIIIC and TFIIIB (Sprague et al. 1980; Dingermann et al. 1982; Raymond and Johnson 1983; Shaw and Olson 1984; Raymond et al. 1985; Joazeiro et al. 1996; Ong et al. 1997; Donze and Kamakaka 2001; Giuliodori et al. 2003). It is likely and probable that flanking sequences adjacent to TFIIIC binding sites and the internal promoter sequences that are recognized by TFIIIC play a role in determining which TFIIIC-bound promoter elements can function as insulators.
TFIIIC-mediated transcription is not necessary for insulation:
We showed that ETC9 binds TFIIIC and TFIIIB but not RNA pol III and still functioned as an insulator. The amount of RNA pol III present at ETC9 when this locus was transposed to HMR was minimal and close to background. The ETC9 data demonstrate that robust transcription and/or generation of a tRNA was unlikely to play a role in insulation. Consistent with this conclusion is the observation that there is no RNA pol III at the native ETC9 locus (Roberts et al. 2003) and this ETC9 locus is not transcribed (Guffanti et al. 2006). Our observation that 9x B boxes also function as insulators further demonstrates that transcription is not necessary for insulation since the B boxes alone do not recruit any RNA pol III (Noma et al. 2006). Finally this conclusion is also consistent with our previous report (Donze and Kamakaka 2001) that mutations in TFIIIC affect tRNA-mediated insulation but mutations in RNA pol III had no effect.
Results in S. pombe have shown that TFIIIC binding sites at the IR insulator do not recruit RNA pol III but are necessary for RNA pol II recruitment (Noma et al. 2006). It has also been shown that these RNA pol II molecules generate noncoding transcripts at the S. pombe IR insulator. While at the IR elements TFIIIC recruits pol II, TFIIIC binding alone is sufficient for insulation. We do not observe any RNA pol II recruitment to ETC loci or tRNA genes in S. cerevisiae, suggesting that TFIIIC-mediated pol II recruitment is also not necessary for insulation in S. cerevisiae.
TFIIIC-mediated barrier function utilizes chromatin remodelers:
In addition to the DNA bound transcription factors we also found that mutations in specific histone acetylases and chromatin remodelers led to a loss of insulation. We tested several mutants in histone acetylases and chromatin remodelers. Mutants in Sas2, Eaf3, Isw2, and Rsc2 all directly or indirectly affected insulation mediated by ETC9 and the 9x B boxes while mutations in Hat1 or Sas3 (data not shown) had no effect. These same mutants also affected tRNA-mediated insulation, suggesting a commonality in the factors that affect the distribution of Sir proteins at HMR.
The tRNA insulator is present in a histone-depleted region of chromatin (Oki and Kamakaka 2005). Our current data suggest that RSC is directly or indirectly required to evict these histones at the tRNA insulator. In a Rsc2 mutant, histone occupancy at the tRNA was increased, resulting in reduced binding of the RNA pol III factors. Our current results also suggest that similar mechanisms operate at ETC9-mediated insulation at HMR. Additionally a nucleosome-free region can by itself block the spread of silencing (Bi et al. 2004), suggesting that the creation of a nucleosome-free region may be an important step in insulation.
Genomewide mapping data demonstrate that histones are depleted at tRNA genes (Pokholok et al. 2005; Lee et al. 2007; Whitehouse et al. 2007; Mavrich et al. 2008; Parnell et al. 2008) and recent observations demonstrate that histone eviction around most tRNA genes is reduced in an RSC mutant (Parnell et al. 2008). Furthermore, a large number of RSC binding sites are located at RNA pol III transcribed genes (Damelin et al. 2002; Ng et al. 2002) and studies have shown that RSC interacts with subunits of the RNA polymerase machinery (Soutourina et al. 2006), suggesting a mechanism by which TFIIIC insulators may function. In this scenario, nucleosome eviction, dependent upon the RSC complex, would be required for TFIIIC binding and insulation.
An important point that needs to be borne in mind is that while RSC affected both insulation and nucleosome occupancy at the tRNA, our data are unable to unequivocally indicate whether RSC-mediated insulation is due solely to nucleosome eviction by RSC. It is entirely possible that part of the effect of RSC on insulation is due to alterations in the distribution and spreading of Sir proteins in the nucleus.
Our results collectively demonstrate that binding of TFIIIC is necessary for insulation but binding of TFIIIB along with TFIIIC likely improves the probability of complex formation at an insulator. Insulation critically depends on the ability of factors to bind stably to DNA in competition with the spreading Sir proteins. Furthermore, histone eviction mediated by RSC and possibly other enzymes may play a role in restricting the spread of silenced chromatin.
We thank Brad Cairns and David Donze for specific strains and plasmids and members of the lab for comments and criticisms during the progression of this research. This work was supported by a grant from the National Institutes of Health to R.T.K. (GM078068).
Supporting information is available online at http://www.genetics.org/cgi/content/full/genetics.109.106203/DC1.
Communicating editor: M. Hampsey
- Received June 11, 2009.
- Accepted July 5, 2009.
- Copyright © 2009 by the Genetics Society of America