RSC regulates nucleosome positioning at Pol II genes and density at Pol III genes

Loss of Sth1 protein results in loss of transcription

To study the impact of RSC on transcription and chromatin structure, we generated an STH1 allele (sth1^td) tagged with an N-terminal temperature-sensitive degron to enable rapid and inducible loss of the Sth1 protein (Dohmen et al, 1994). We reasoned that the loss of the catalytic subunit of RSC should inactivate enzymatic activity. For our studies, we employed two isogenic yeast strains carrying the sth1^td allele that differed only at UBR1, which encodes the protein that recognizes the degron tag. One strain contained an inducible UBR1 gene (degron strain) while the second was ubr1Δ (control strain; Labib et al, 2000). The degron strain failed to grow under degradation-inducing conditions (galactose media at 37°C), while full growth was observed for the control strain (Figure 1A). Furthermore, western blot analysis of cell lysates showed substantial loss of Sth1 protein selectively in the degron strain (Figure 1B). While the loss of Sth1 protein reduced cell growth, the effect was reversible; cells subjected to loss of Sth1 protein for 2 h and allowed to recover at room temperature showed >95% viability (data not shown). Similar results with another sth1^td allele were also recently reported (Campsteijn et al, 2007). Together, these results show the utility of the sth1^td allele in analyzing RSC function.

To determine the impact of the loss of RSC activity on new transcription, cells were grown in the presence of ³H-labeled uracil for 25 min either before or after Sth1 protein degradation. RNA was isolated and separated by PAGE. Ethidium bromide staining of the gel revealed uniform RNA levels, while autoradiograph exposure revealed the striking loss of new RNA transcription in the degron strain after Sth1 protein loss (Figure 1C). Importantly, transcription by all three RNA polymerases was affected, as tRNAs, rRNAs, and mRNAs were greatly diminished. Thus, the loss of Sth1 protein had a global impact on transcription. Whether these effects were direct or indirect presented a considerable challenge, which we addressed by multiple approaches.

Histone density increases at a Pol III locus with loss of Sth1 protein

RSC may directly impact transcription through alterations of the chromatin structure at RSC targets. If so, then direct targets might share a common chromatin feature. Notably, the largest class of RSC-occupied loci is Pol III genes, primarily tRNA genes (Ng et al, 2002). Furthermore, tRNA genes have the lowest histone occupancy genome-wide (Supplementary Figure 1; Rao et al, 2005). This inverse correlation between histone and RSC occupancy at tRNA genes raised the possibility that RSC helps to keep Pol III genes nucleosome sparse, and that the loss of RSC activity might result in a gain of histone occupancy over time.

To initially investigate RSC function at a Pol III locus, we examined SCR1, the longest Pol III gene in the genome and a known RSC target. SCR1 was previously used as a model Pol III gene to map the distribution of Pol III components over the body of the gene (Roberts et al, 2006). Histone chromatin immunoprecipitation (ChIP) was performed in the degron and control strains both before and after Sth1 protein loss. The precipitated DNA was then analyzed by quantitative PCR using six amplicons distributed across the locus (Figure 2A and B). Histone occupancy was low over the body of the SCR1 gene in both strains before Sth1 loss, consistent with our observations at tRNA genes genome-wide. After 1 h of Sth1 loss, however, we observed a small gain of histone density and a more robust gain after 2 h over the body of SCR1 in the degron strain, but not in the control strain. As an alternative measure, the chromatin structure of SCR1 was probed using micrococcal nuclease (MNase) digestion and analyzed by Southern blot. The body of the SCR1 gene yielded a detectable nucleosomal ladder of hybridization in the degron strain but not the control strain (Figure 2C). Together, these results indicated that Sth1 protein loss results in a progressive gain of nucleosomes at the SCR1 locus.

We next examined whether this progressive return of histone density was due to the loss of transcription, and whether DNA replication was additionally involved. To determine whether this nucleosomal gain was simply due to a lack of transcription, we performed histone ChIP with STH1⁺ cells during nutrient deprivation, which rapidly (within 10 min) represses Pol III transcription (Clarke et al, 1996; Harismendy et al, 2003; Roberts et al, 2003). Interestingly, histone density was not restored to SCR1 during Pol III repression (Figure 2D). Notably, the Pol III machinery is known to maintain residence at Pol III genes during Pol III repression (Roberts et al, 2006), prompting a test of whether RSC likewise maintains association. Here, we found that during repression RSC remained at the three Pol III loci tested: SCR1 and two tRNA genes (Supplementary Figure 2).

To determine whether replication-dependent chromatin assembly contributes to the restoration of histone density in the absence of RSC, we first arrested Sth1^td strains in G1 phase with alpha-factor, degraded Sth1, and then examined histone density over the SCR1 locus after 2 h (Figure 2E). Interestingly, the restoration of histone density was attenuated compared with cycling cells, indicating that replication-dependent chromatin assembly accounts for a large portion of the acquired histone density on Pol III genes lacking RSC. Taken together, our work suggests that RSC helps keep Pol III genes nucleosome-deficient, regardless of their transcriptional status, and that the absence of RSC makes Pol III loci susceptible to replication-dependent assembly into chromatin.

A genome-scale microarray examines chromatin changes at numerous RSC targets

We then determined whether our observations of histone gain at SCR1 were representative of Pol III genes in general, and whether this observation extended to Pol II genes. Toward this end, we generated a custom RSC-specific miniature microarray at relatively high resolution to interrogate a large number of target loci. The array covers 218 specific genomic locations selected by their RSC occupancy levels and genomic content (Figure 3A; Ng et al, 2002). Four general categories of loci were selected: 99 highly-occupied RSC target loci containing 155 Pol II promoters (including 56 divergent pairs), 43 loci containing Pol II promoters not occupied by RSC, 54 loci containing Pol III genes of varying RSC occupancies, and 22 other loci. Each locus, which typically consisted of an intergenic and a portion of the flanking ORFs, was tiled with 60-mer oligonucleotide probes at a mean resolution of 110 bp, which was sufficient for monitoring changes in individual nucleosomes.

As an initial test of the microarray, we performed ChIP against RSC. We assessed occupancies of two different RSC subunits, Rsc3 and Rsc8, in wild-type STH1⁺ cells, since RSC demonstrates low cross-linking efficiency (Ng et al, 2002). ChIP eluates were amplified, labeled, and hybridized to the arrays. We averaged the RSC occupancy of all probes within a 500-bp window at each Pol II promoter on the array, and plotted the distribution of RSC occupancies at all the Pol II promoters (Figure 3B). Under normal growth conditions (mock before), we observed a distinct enrichment at a large number of promoters, consistent with our expectations from our array design. Comparison of our results with published results (Ng et al, 2002) revealed a strong positive correlation (Pearson correlation coefficient=0.87). Since the array represents a small fraction of the genome, the enrichment ratios are not absolute values but rather a relative representation of the RSC occupancy for the promoters on this array. Therefore, we divided promoters into classes representing the least, moderately, and most likely occupied by RSC; for simplicity, these classes are referred to as non-occupied, moderate, and RSC-occupied, respectively (Figure 3C). Based on these criteria, 84% of the promoters predicted to be RSC occupied demonstrated occupancy, and 63% of the promoters predicted not to be occupied demonstrated lack of occupancy; the majority of the remaining promoters fell in the moderate class. These results indicate the validity of using this microarray in analyzing RSC function at its target sites in vivo.

Since analysis of our sth1^td allele requires growth in media containing galactose at 37°C, we needed to determine whether the RSC targets we identified under normal conditions remain targets under mock Sth1^td protein degradation growth conditions. Therefore, we also performed RSC ChIP under mock degradation conditions (mock after). Interestingly, we observed a redistribution of RSC occupancy (Figure 3B). Presumably, this redistribution of RSC reflects the transcriptional program changes associated with the new growth conditions. Surveying the promoters present on this array, we found fewer but still significant numbers of promoters that we could confidently identify as having RSC occupancy (Figure 3C). Notably, we observed bidirectional shifts, with some promoters gaining occupancy while others losing occupancy. While shifts in RSC occupancy at Pol II promoters occurred, tRNA genes appeared remarkably stable in RSC occupancy during mock after conditions (Figure 3D). Together, these results indicated that RSC occupancy is dynamic at Pol II promoters and stable at Pol III genes under our conditions.

Nucleosomes return to Pol III genes upon loss of Sth1 protein

To examine chromatin changes at Pol III genes, we performed ChIP against tagged H4 in the degron and control strains and hybridized the eluates to the microarray. We first present microarray results from the SCR1 locus (Figure 4A). As a measure of change in histone occupancy between the presence and absence of Sth1 protein, we plotted the ratio of the signal from the degron and control strains, such that a ratio of 1 indicates no change, >1 indicates a gain of nucleosomes in the degron strain, and <1 indicates a loss of nucleosomes. In agreement with our PCR results (Figure 2), we observed a large gain in nucleosome density over the SCR1 gene after loss of Sth1 protein, indicating a gain in nucleosome occupancy.

We next asked whether the gain of nucleosomes at SCR1 was representative of all Pol III genes. We examined as a class the general chromatin structure of loci containing a tRNA gene by identifying those microarray probes that were located within a 400-bp window centered over the tRNA gene (on average 4–5 probes). We then averaged the degron/control ratios of the probes within this window and plotted the distribution of the resulting values for all the tRNA loci (Figure 4B). We observed a dramatic shift in nucleosome occupancy at tRNA loci following the loss of Sth1 protein (P=2.59E−17), similar to what we observed at SCR1. Examination of two additional Pol III-transcribed genes, RPR1 and SNR6, also revealed general nucleosome gains, including both their coding and 5′ flanking intergenic regions (data not shown). Together, these results suggest that RSC, directly or indirectly, helps to remove nucleosomes from these loci, and that nucleosome density increases in the absence of RSC function. Furthermore, this effect appears to be a general feature at all Pol III loci.

Since an increase in nucleosome density was observed at tRNA loci, we next asked whether this would impact Pol III occupancy. We performed ChIP against the tagged Rpc82 subunit of Pol III in our sth1^td strains and analyzed the eluates on our microarray (Figure 4C). Interestingly, we observed a striking, corresponding decrease in Pol III occupancy at tRNA loci in the degron strain upon loss of Sth1 protein (P=3.66E−15). This result suggests a mutually exclusive relationship between histone and Pol III, and may help explain the loss of Pol III transcription upon loss of Sth1 protein.

Loss of Sth1 affects nucleosome positioning at Pol II promoters

We next turned our attention to Pol II promoters. Since the loss of Sth1 had a profound impact on transcription, our ability to attribute transcriptional defects directly to RSC action relied upon observing a more robust change in chromatin structure at RSC-occupied promoters than at non-occupied promoters. To test for this, we focused on those probes located within promoter regions, defined as −400 to +100 bp from the ATG. The mean ratios of the probes within each promoter region were then plotted as a distribution. At non-occupied promoters, we detected a slight histone gain upon loss of Sth1 protein (Figure 5A; P=0.005). The effects at these ‘non-occupied' promoters could either be indirect or due to low or transient RSC occupancy. However, we also observed only modest changes at the vast majority of RSC-occupied promoters (Figure 5B; P=0.014), although there were a small percentage of promoters that did show a notable gain. Therefore, regardless of classification, the majority of Pol II promoters examined did not exhibit the large general gains in histone density observed at Pol III promoters.

We then considered that RSC-associated chromatin changes at Pol II promoters might be more subtle than the gross nucleosome density changes that are detected by conventional ChIP. To examine chromatin structure at a more detailed level, we performed MNase digestion on chromatin and hybridized the liberated mononucleosomal DNA (∼150 bp) to our microarrays (Sekinger et al, 2005; Yuan et al, 2005). This method can detect either the gain or loss of a single nucleosome centered over an individual probe, and thus monitor single nucleosome events. With traditional ChIP, the fractional gain or loss of a single nucleosome may not be easily detected due to the larger fragment size (∼500 bp) and the associated summation of nucleosome occupancy. Hence, the mononucleosome technique can provide a more sensitive detection of chromatin alterations at the nucleosome level.

We anticipated that three types of chromatin events could be detected by mononucleosome hybridization: movement or sliding of a nucleosome, as inferred by the concordant gain and loss in signal of neighboring probes; single nucleosome gain; and single nucleosome loss. Examples of three candidate loci demonstrating each class of events are presented in Figure 6. In some cases, two separate events were identified within the same promoter or intergenic (Figure 6B and C). In many cases, the MNase procedure revealed a single location where nucleosome density changed, which was either largely (Figure 6A) or partly (Figure 6B) masked in the H4 ChIP data due to the nucleosome averaging effect of ChIP. In other cases, the MNase data correlated well with the H4 ChIP results (Figure 6C); these cases may represent promoters where more than one nucleosome was gained. Together, these examples illustrate the types of nucleosome events occurring in response to loss of RSC function.

As a simple measure of nucleosome events, we identified the probe in each promoter that gained the most nucleosome occupancy from the mononucleosome hybridization, termed the Most Enriched Probe (MEP), in the degron strain relative to the control strain (Figure 5C and D). The MEP could arise either from a single nucleosome gain or through a nucleosome movement where the gain is associated with the repositioning of the nucleosome completely over a probe location. Examination of the distribution of MEP ratios in non-occupied promoters revealed a slight gain in the magnitude of the MEP (P=0.001), a change similar to what was observed with the ChIP result. However, examination of the RSC-occupied promoters revealed a prominent shift in the distribution (P=2.5E−13), indicating that loss of RSC function resulted in a gain of nucleosome occupancy over at least one probe. These results suggested that single nucleosome events are much more prevalent than changes in gross nucleosome occupancy and are the primary effects due to the loss of RSC function at Pol II promoters.

Single nucleosome events generated by the loss of RSC function may impact transcription, depending upon the location of the event relative to promoter sequence elements. The TSS usually resides within the NFR and may be subject to regulation by nucleosome positioning (Yuan et al, 2005). We identified those probes that overlap a 50-bp window centered over the TSS and examined the distribution of degron/control ratios (Figure 5E and F). We observed no significant shift in the distribution of ratios for non-occupied promoters (P=0.28). However, a clear shift is noted at RSC-occupied promoters (P=1.57E−4), suggesting that one location of RSC action may be the NFR that encompasses the TSS.

Transcriptional shutoff by rpb1-1 results in nucleosome loss in the promoter

The observation of chromatin changes at the TSS prompted us to ask whether the changes we observed in our sth1^td mutant were in direct response to the loss of RSC activity or indirectly due to the loss of transcription. To help answer this question we prepared mononucleosomal DNA from an rpb1-1 mutant both at permissive and non-permissive temperature (mimicking our degron time-course protocol). This temperature-sensitive Pol II mutant abrogates new transcription within minutes of shifting to a non-permissive temperature (Nonet et al, 1987). We determined the distribution of MEP ratios (non-permissive over permissive temperature) for both RSC-occupied and non-occupied promoters (Supplementary Figure 3). Surprisingly, we observed a lower mean ratio at RSC-occupied promoters compared with non-occupied promoters (P=1.32E−6), suggesting that RSC promoters generally lose mononucleosomes upon loss of transcription in the rpb1-1 mutant. Furthermore, when the probes corresponding to the TSS were examined, no significant difference in the distribution was observed between the two promoter classes (P=0.59). These results suggested that loss of transcription in rpb1-1 strains is associated with structural changes in promoter chromatin, including a general loss of nucleosome occupancy. The mechanism underlying the loss of nucleosomes in the rpb1-1 strain is unknown, but the result nevertheless highlights the distinction between changes in promoter chromatin structure due to loss of transcription versus loss of RSC activity.

Media, genetic methods, and strains

Standard procedures were used for media preparation, transformations, integrations, and genetic analyses. Plasmid and strain constructions will be provided upon request. Strain genotypes (S288C derivatives) are listed in Table I. For degron inductions, cultures were grown in liquid YP containing 2% raffinose and 0.05 mM Cu(II)SO₄ at 27°C until OD₆₀₀ 0.4–0.6. Galactose was then added to 2% final concentration and the culture incubated for 60 min to induce UBR1. Afterwards, the temperature was shifted to 37°C for 120 min to induce Sth1 degradation. Samples were cross-linked in 1% formaldehyde for 5 min for mononucleosome analysis, 15 min for histone ChIP, and 30 min for other proteins. For G1 arrest studies, alpha-factor was added to bar1Δ degron and control strains simultaneously with galactose and the time course followed as above. Loss of BAR1 did not affect Sth1 degradation. The nutrient deprivation time-course experiment was performed as described (Roberts et al, 2003).

In vivo RNA labeling

Cells were grown following the degron time course above. Aliquots (2 ml) of culture were incubated with 165 μCi of ³H-labeled uracil for 25 min at each time point. Total RNA was immediately harvested using acid phenol extraction. Messenger RNA was enriched by one round of poly-A selection on oligo-dT beads. Enriched RNA was separated by denaturing polyacrylamide gel electrophoresis, followed by either staining with ethidium bromide or electrophoretic transfer to nylon membrane. The membrane was soaked in Amplify (Amersham) and exposed to film.

Microarray design and analyses

The microarray format was based on Agilent Technologies' 8-pack 22K format. Details on the custom microarray design, ChIP, mononucleosome preparation, hybridization, and data processing are available in Supplementary data. Microarray data are provided as Supplementary data and are present at the NCBI GEO repository under experiment GSE8862.

Genomic and statistical analyses

Promoters were defined as −400 to +100 relative to the translation start site. Regions containing tRNA genes were defined as −150 to +250 relative to the start site. Probes within or overlapping these regions were selected for analysis. TSSs were collected from published sources (Zhang and Dietrich, 2005; Miura et al, 2006). The most commonly used TSS position or, failing that, the position closest to the ATG was selected and used in the analysis. P-values between data sets before and after Sth1 degradation were determined by two-tailed, equal-variance Student's t-test. Distribution histograms for H4 ChIP, MEP, and Pol III ChIP were normalized so that the mean for the Before curve was 1.