|Home | About | Journals | Submit | Contact Us | Français|
Users may view, print, copy, download and text and data- mine the content in such documents, for the purposes of academic research, subject always to the full Conditions of use: http://www.nature.com/authors/editorial_policies/license.html#terms
Eukaryotic transcriptomes are characterized by widespread transcription of non-coding and antisense RNAs1–3, which is linked to key chromosomal processes, such as chromatin remodeling, gene regulation, and heterochromatin assembly4–7. However, these transcripts can be deleterious, and their accumulation is suppressed by several mechanisms including degradation by the nuclear exosome8,9. The mechanisms by which cells differentiate coding RNAs from transcripts targeted for degradation are not clear. Here we show that the variant histone H2A.Z, which is loaded preferentially at the 5' ends of genes by the Swr1 complex containing a JmjC domain protein, mediates suppression of antisense transcripts in the fission yeast Schizosaccharomyces pombe genome. H2A.Z is partially redundant in this regard with the Clr4/Suv39h-containing heterochromatin silencing complex that is also distributed at euchromatic loci, and with RNAi component Argonaute (Ago1). Loss of Clr4 or Ago1 alone has little effect on antisense transcript levels, but cells lacking either of these factors and H2A.Z show markedly increased levels of antisense RNAs that are normally degraded by the exosome. These analyses suggest that in addition to performing other functions, H2A.Z is a component of a genome indexing mechanism that cooperates with heterochromatin and RNAi factors to suppress read-through antisense transcripts.
Histones are a major structural component of chromatin, critical for packaging and regulation of eukaryotic genomes. Specialized chromatin can be differentiated by local incorporation of variant histones into a subset of nucleosomes10. A conserved variant histone, H2A.Z has been linked to diverse chromosomal processes10, but its specific regulatory function(s) are unclear. Although H2A.Z is dispensable for viability in S. cerevisiae, it is essential in Drosophila and mice, where it is linked to both transcriptional activation and repression10,11. Evidence suggests that metazoan H2A.Z localizes to heterochromatic loci and its depletion causes defective chromosome segregation12.
In this study, we used S. pombe to explore H2A.Z functions. S. pombe genome contains large blocks of heterochromatin, whose assembly requires histone H3 lysine 9 (H3K9)-specific methyltransferase Clr4/Suv39h, Swi6/HP1 family proteins, and the RNAi machinery7. Deletion of pht1, the gene encoding H2A.Z causes segregation defects13. However, it is not known whether this phenotype is linked to defective heterochromatin.
H2A.Z deposition is catalyzed by the Swr1 complex in S. cerevisiae10. To establish if a similar complex deposits H2A.Z in S. pombe, we purified factors associated with H2A.Z and Swr1 homolog. These analyses identified an analogous complex in S. pombe, which also contained a JmjC domain protein, Msc1 (Supplementary Fig.1)14. Msc1, while dispensable for complex assembly (Supplementary Fig. 1), affected chromatin association of Swr1 (Fig. 1a). Both Swr1 and Msc1 were required for H2A.Z deposition at various sites, though Δswr1 was more defective (Fig.1b; Supplementary Fig. 2a–d).
ChIP-chip analysis showed a non-random distribution of H2A.Z across the genome (Supplementary Fig. 2c). H2A.Z is relatively depleted throughout heterochromatic domains enriched for methylated H3K9 (H3K9me) and lacking H3K4me15 (Supplementary Figs. 3a). At telomeres, H3K9me is distributed across a ~100 kb domain (Fig. 1c), beyond the heterochromatin domain previously defined using a less sensitive method15. H2A.Z is depleted in this entire H3K9me-coated domain (Fig. 1c). Δpht1 causes a slight increase in silencing at pericentromeric region but H3K9me distribution at heterochromatic loci is not severely altered (Supplementary Fig. 4a). At euchromatic loci, H2A.Z localizes preferentially in intergenic regions (Supplementary Fig. 3b) with maximum enrichment occurring ~200 bp upstream of translation start sites (Fig. 1d). This pattern is similar in other organisms16–21, suggesting that this is a general feature of H2A.Z deposition. Comparison of H2A.Z and RNA polymerase II (Pol II) distributions revealed that S. pombe H2A.Z is preferentially enriched at repressed or weakly expressed genes (Supplementary Fig. 3c), as in S. cerevisiae16–18.
To determine whether H2A.Z affects transcription, we monitored transcript levels from both strands using tiling arrays. Consistent with reports that most of the S. pombe euchromatin is transcribed1,2, we detected widespread sense and antisense transcripts in wild-type cells. Δpht1 caused only modest changes in sense transcription. The affected loci included subtelomeric genes such as tlh2 sharing homology to centromeric repeats (Supplementary Fig. 4b). Silencing of tlh2 and its paralogs requires heterochromatin machinery7 (Supplementary Fig. 4b), butH2A.Z, undetected in our assays, may also contribute to tlh2 repression.
Δpht1 caused a disproportionate increase in antisense transcripts at many (~5–8%) euchromatic loci (Fig. 1e–g; Supplementary Figs. 5), as confirmed by strand-specific RT-PCR (Fig. 1f). More than 90% of antisense RNAs mapped to convergent genes transcribed from opposite DNA strands. These antisense RNAs appear to result from read-through transcription (Fig. 1g; Supplementary Figs. 6; see below). Since heterochromatin factors have been implicated in control of antisense transcripts22, including read-through transcripts at three convergent loci23, we asked whether H2A.Z interacts genetically with heterochromatin machinery. When Δpht1 was combined with mutant alleles of clr4 or rik1, components of Clr4-containing ClrC complex7, the resultant double mutants displayed severe growth defects and a large, synergistic increase in antisense RNAs at >20% of genes (Fig. 2a; Supplementary Figs. 5, 7a–b). Consistent with ClrC directly participating in antisense suppression, Rik1 was found at the convergent loci (Supplementary Fig. 7c).
ClrC collaborates with RNAi factors including an Argonaute (Ago1)-containing RITS complex24, to silence heterochromatic repeats7. We asked whether Ago1 contributes to antisense suppression. Combining Δago1 with Δpht1 resulted in synergistic upregulation of antisense transcripts, as in Δclr4Δpht1 (Fig. 2a). These data are consistent with RNAi-dependent heterochromatin assembly, especially Swi6/HP1 bound to H3K9me, recruiting cohesin to mediate transcription termination23. However, antisense RNAs did not accumulate extensively in Δswi6 cells and the synergistic increase in antisense RNAs observed in Δclr4Δpht1 mutant was not observed in the Δswi6Δpht1 mutant (Fig. 2a). Thus, ClrC and Ago1 contribute to antisense suppression via novel mechanism(s).
To gain insight into the roles of H2A.Z and heterochromatin/RNAi in antisense suppression, we examined common features of antisense RNAs in different mutants. Similar to Δpht1, most loci affected in Δclr4Δpht1 and Δago1Δpht1 double mutants were convergent genes, including loci affected in the single mutants. The average profile of relative changes between mutant versus wild-type clearly demonstrates synergistic interactions of H2A.Z with Clr4 and Ago1 in suppressing antisense transcripts (Fig. 2c). The increase in antisense transcripts was not restricted to a specific region, but occurred uniformly throughout the gene body (Fig. 2c), consistent with read-through transcripts traversing adjacent loci (Supplementary Fig. 6). This is also illustrated in the average profile of antisense signal intensities (Fig. 2d). Examination of individual loci recapitulated the accumulation of ORF-spanning read-through antisense transcripts in double mutants (Fig. 2b).
Northern analyses confirmed that antisense RNAs observed in Δpht1, Δclr4 and Δclr4Δpht1 mutants correspond to read-through transcripts rather than new initiation events. Strand-specific probes for antisense transcripts at cyp7 or srb4 hybridized to long RNAs that accumulated in mutant backgrounds (Fig. 3a; Supplementary Fig. 8). The length and direction of antisense RNA, as determined by oligo-targeted RNase H cleavage, was consistent with read-through transcripts originating at the promoter of the adjacent convergent gene. Long read-through transcripts were also detected with probes used to detect sense transcripts at the c16H5.04-cyp7 and c29A3.19-c18e5.02c gene pairs (Fig. 3a and Supplementary Fig. 9a). Importantly, Δpht1 and Δpht1Δclr4 mutants showed elevated levels of antisense transcripts, but sense transcripts were not upregulated (Fig. 3a). Thus, H2A.Z and ClrC appear to specifically suppress read-through antisense transcripts.
The accumulation of antisense RNAs could result from defects in transcriptional termination and/or processing. Although termination failure cannot be completely excluded, levels of nascent antisense transcripts, as determined by transcription run-on assay, were not considerably affected in Δpht1 or Δpht1Δclr4 mutant (Supplementary Fig. 9b). We hypothesized that H2A.Z and heterochromatin factors might selectively promote degradation of antisense RNAs. Indeed, aberrant transcripts including antisense RNAs are known to be degraded by the 3' to 5' exonuclease activity of nuclear exosome8,9,22, which also processes heterochromatic repeat transcripts targeted by ClrC22,25. Deletion of exosome subunit rrp6 led to an antisense profile closely resembling that of Δclr4Δpht1, with read-through antisense RNA covering entire ORFs at convergent genes (Fig. 3b). Striking similarities in levels and patterns of antisense RNAs in Δclr4Δpht1 and Δrrp6 mutants were observed at individual loci (Fig. 3c). Similar analyses revealed a remarkable correlation of antisense RNA upregulation between rik1pht1 or ago1pht1 mutants and rrp6 mutant (Fig. 4a). We conclude that read-through transcripts are produced in wild type cells, but are processed by the exosome in a manner dependent upon H2A.Z, ClrC and Ago1.
The histone deacetylase (HDAC) Clr6 complex suppresses antisense transcripts initiating from cryptic promoters via a mechanism involving H3K36 methyltransferase Set222,26. To compare distributions of antisense transcripts, we comprehensively profiled and correlated distributions of antisense RNAs in different mutants (Fig. 4b). Δset1Δpht1 mutant shows severe synthetic growth defects similar to Δclr4Δpht127 but did not exhibit cumulative upregulation of antisense RNAs. Correlation coefficients from pairwise comparisons of antisense profiles were used to perform hierarchical clustering (Fig. 4b), in which mutants were sorted according to similarities in their antisense profiles. Double mutants lacking Pht1 and Ago1 or ClrC subunits clearly belong to a distinct cluster that associated closely with Δrrp6. This cluster was separated from a second cluster that included Δset2 and Clr6 complex mutants (Fig. 4b). Therefore, the effect of H2A.Z/ClrC on antisense RNAs is distinct from that of the HDAC Clr6.
This study implicates H2A.Z and heterochromatin factors in suppression of potentially deleterious antisense RNAs. H2A.Z nucleosomes might directly obstruct Pol II progression, or facilitate loading of factors involved in structural organization of chromosome, which in turn promotes RNA degradation by stalling Pol II. Alternatively, H2A.Z might signal to Pol II-associated exosome28 that transcription has escaped its natural termination and therefore produced an aberrant transcript (Fig. 4c). H2A.Z and ClrC (Rik1 subunit resembles cleavage and polyadenylation factor CPSF-A7) may be components of a RNA quality control mechanism, which stimulates exosome activity via exosome cofactors8. Indeed, the loss of Cid14 subunit of TRAMP, implicated in exosome stimulation8 causes accumulation of antisense RNAs (Fig. 4b), and the Drosophila homolog of H2A.Z mediates targeting of mRNA processing factors29. In addition to ClrC, Ago1 also collaborates with H2A.Z to suppress antisense RNAs. Drawing from analyses at centromeric repeats7, Ago1 and ClrC might cooperate not only to process antisense RNAs but also to influence local chromatin to block antisense transcription23. Exosome might contribute to degradation of RNAs targeted by RNAi/ClrC in a manner similar to mRNAs targeted by RISC30. Future studies will explore the interplay between chromatin and RNA processing factors in suppressing antisense RNAs.
Considering that H2A.Z localizes at the 5'end of genes in most eukaryotes16–21 and that heterochromatin factors are found in euchromatic regions of different species7, the involvement of H2A.Z and heterochromatin machinery in antisense suppression might be conserved in other species. Defects in RNA processing could result in genome instability and might be partially responsible for chromosome segregation defects observed in H2A.Z and heterochromatin-defective mutants7,13, similar to those seen in other mutants showing upregulation of antisense RNAs22.
ChIP and ChIP-chip were performed as described previously15, except that hybridizations were performed as per manufacturer (Agilent) recommendation. Antibodies used were: FLAG (M2; Sigma), H3K9me2 (Abcam), Pol II (8WG16; Covance), and Myc (Santa Cruz). For transcriptome analyses, total RNA was extracted with Master Pure Yeast RNA Purification kit (Epicenter) and 5 ug of RNA was reverse transcribed and labeled with Superscript Indirect cDNA labeling kit (Invitrogen). Data analysis was performed as described in Methods online. Transcription run-on was performed as described previously23. Standard methods were used to perform Northern analysis. RT-PCR experiments were performed with one-step RT-PCR kit (Qiagen).
Extracts were prepared from cells expressing H2A.Z-FLAG, Swr1–3xFLAG or Msc1–3xFLAG. Specifically, exponentially growing cells were resuspended in 2xHC buffer (200 mM HEPES [pH 7.4], 250 mM KCl, 2 mM EDTA, and 20% glycerol) supplemented with protease inhibitors, frozen and disrupted. Extracts were spun at 100 000 × g, pre-cleared with IgG and FLAG tagged proteins recovered by incubation with anti-FLAG M2 slurry (Sigma) for 4 hours, washed 12 times with 1xHC buffer and eluted with FLAG peptide.
Purified proteins were separated by SDS-PAGE and stained using SimpleBlue® (Invitrogen). Excised gel pieces were subjected to in-gel digestion using trypsin. Peptides extracted from the gel pieces were desalted using a C18 ZipTip (Millipore, Bedford, MA), lyophilized and resuspended in 16 μL of 0.1% trifluoroacetic acid (TFA) prior to analysis by microcapillary reversed-phase liquid chromatography-tandem mass spectrometry (LC-MS2). Each sample (6 μl) was loaded onto an Agilent 1100 nano-capillary HPLC system (Agilent Technologies, Palo Alto, CA) equipped with a 10 cm integrated nanoRPLC-electrospray ionization (ESI) emitter column (made in house), coupled online with a linear ion-trap (IT) mass spectrometer (LTQ XP, ThermoElectron, San Jose, CA). After sample injection, a 20 min. wash with 98% mobile phase A (0.1% formic acid) was applied and peptides were eluted using a linear gradient of 2% mobile phase B (0.1% formic acid in acetonitrile) to 42% mobile phase B within 40 min. at a constant flow rate of 200μl/min. The seven most intense molecular ions in the MS scan were sequentially selected for collision-induced dissociation (CID) using normalized collision energy of 35%. The mass spectra were acquired over a mass-to-charge (m/z) range of 350–1800. The ESI source capillary voltage and temperature were maintained at 1.5 kV and 200 °C, respectively. The MS data were searched against the UniProt S. pombe database downloaded from the European Bioinformatics Institute website (http://www.ebi.ac.uk/integr8) using SEQUEST (ThermoElectron) operating on a 10 node Beowulf parallel virtual machine computer cluster (Dell, Inc., Round Rock, TX). Up to two missed tryptic cleavage sites and oxidation of methionyl residues were allowed during the database search. Thresholds for legitimate peptide identifications were set as follows: minimum delta correlation (DCn) ≥ 0.1 and charge state dependent cross correlation scores (Xcorr) ≥ 2.0 for [M+H]1+ ≥ 2.5 for [M+2H]2+ and ≥ 3.0 for [M+3H]3+.
Cells expressing H2A.Z-FLAG were resuspended in NIB buffer (0.25 M sucrose, 15 mM Pipes [pH 6.8], 60 mM KCl, 15 mM NaCl, 5 mM MgCl2, 1 mM CaCl2, 0.8% Triton X-100) supplemented with protease inhibitors, 2 mM ZnSO4, 10 ng/ml trichostatin A, and homogenized with glass beads. Cell pellets were collected by centrifugation at 11,000 × g and histones extracted with 0.4 M H2SO4 and precipitated with trichloroacetic acid. H2A.Z was detected by Western analysis with FLAG M2 antibody.
Chromatin immunoprecipitation was preformed as described previously15. Briefly, exponentially growing cells (5×108) were fixed in 3% paraformaldehyde and chromatin was subsequently crosslinked by treatment of cells with 10 mM dimethyl adipimidate. Cells were washed with phosphate buffered saline, resuspended in lysis buffer (50 mM Hepes/KOH [pH 7.5], 140 mM NaCl, 1 mM EDTA, 1% Triton X-100, 0.1% DOC) and homogenized with glass beads. Chromatin was sheared by sonication to fragments of 500–1000 bp, precleared with protein A slurry and immunoprecipitated with 2–4 μg of antibody. Immunoprecipitated chromatin was recovered by incubation with protein A or protein G slurry, washed extensively and reverse crosslinked by incubation at 65°C. Immunoprecipitated and DNA isolated from whole cell extract (WCE) were analyzed by multiplex PCR or subjected to microarray-based ChIP-chip analysis.
For ChIP-chip, ChIP and WCE DNA samples were amplified using random-primed PCR and resulting samples were labeled with either Cy5 (ChIP) or Cy3 (WCE) dyes. Equal amounts of ChIP and WCE labeled DNA were hybridized to a custom Agilent 4×44K S. pombe oligonucleotide array containing 43,987 probes of 60mer oligos, which represent alternately the plus and minus strands, tiling a large part of chromosome II at 50bp resolution. This array design also contains randomly spotted oligonucleotides that have no homology to the S. pombe genome but serve as controls for determining background signals. Hybridizations were performed according to Agilent recommended protocol. Microarrays were scanned using Agilent scanner (type-G2505B) with extended dynamic range, and data were extracted with Agilent feature extraction software (CHIP-v1_95_May07 protocol). Processed signals were normalized using combined rank consistency filtering with LOWES intensity normalization. PValueLogRatio that measures the significance of signal log ratio over systematic noise level of an individual spot were calculated by feature extraction program and were used for subsequent analysis. Enrichment values were calculated as a ratio of Cy5 processed signal/ Cy3 processed signal. To reduce random background noise inherent in microarray data, we subjected the data to a limited sliding-window average filter and smoothed the data with sliding weighted average of three neighboring probes.
Total RNA was isolated using Master Pure Yeast RNA purification kit (Epicenter). Invitrogen RiboMinus Yeast kit was used to partly deplete ribosomal RNA from samples before the reverse transcription (RT) step. RT was performed using random hexamer and anchored oligodT primer mix. 3 ug of RNA was labeled with Superscript Indirect cDNA labeling kit (Invitrogen). Labeled cDNA samples from mutant (Cy5) and wild type (Cy3) were mixed and hybridized to a high-density microarray, as described above. Data were extracted with Agilent Feature Extraction Software (GE2-v5_95_Feb07 protocol) and processed using combined rank consistency filtering with LOWES intensity normalization. Expression ratios were calculated as Cy5processed signal/Cy3processed signal. Ratios with non-significant P values (PValueLogRatio≥0.05) were set to 1. Background signal was estimated as a median processed signal of 152 oligonucleotides with no homology to S. pombe genome, randomly spotted on array. Due to the column purification step in cDNA labeling method, very short transcripts (< 100–150 nucleotides), such as those described around the promoter regions in the rrp6 mutant strains in S. cerevisiae and other systems, might not be recovered. However, analysis of our data reveals a specific increase in short transcripts upstream of promoter regions in particular in Δrrp6 cells, which is compatible with the recently reported results.
To calculate average gene profiles, 842 ORFs (region Chr2:1000000-Chr2:3049820) were aligned at 5' end or double aligned at 5' and 3' ends (see text). For double alignment, genes shorter than 1,650 bp were aligned without stretching. Genes longer than 1,650 bp were subjected to the following method: 14 probes from each end were included in dataset without any change while extra probes in the middle of gene body were linearly compressed into 5 points. Extremely short genes (<200 bp) were excluded from analysis. Intergenic regions upstream and downstream of ORFs were aligned according to translational start and stop sites, respectively. Probes represented in neighbor ORF were excluded. Intergenic probes within 1,000 bp region of both adjacent ORFs were included twice (once for each adjoined ORF), but weighted according to their distance from the ORFs. Repetitive probes (>80% homology with two or more chromosomal localization) were excluded from the averaging process. Expression profiles of genes (expression signal or expression ratio) were calculated similarly to the ChIP profiles with the exception that probes with multiple characteristics were weighted fully in the average.
Median antisense ratios (mutant/wt) for 842 genes (region Chr2:1000000-Chr2:3049820 – excluded the heterochromatic regions and repetitive probes) were calculated for each expression dataset. Linear regression analysis was carried out between datasets, and Pearson-correlation index was calculated. r values were converted into color codes and plotted as a matrix. We repeated the analysis using the non-parametric Spearman's rank correlation coefficient, and observed very similar results (data not shown). Cluster analysis (“average” method in hclust() R function) was carried out for Euclidean distance of the r values between mutant strains.
TRO was conducted essentially as described previously23. Fifty ml of exponentially growing culture was cooled and harvested at OD595nm~ 0.2, washed with ice-cold TMN buffer (10 mM Tris/HCl pH 7.5, 5 mM MgCl2, 100 mM NaCl) and resuspended in ice-cold water and supplemented with 10% N-lauryl sarcosine sulfate to a final concentration of 0.6%. Cells were incubated on ice for 25 min, pelleted and resuspended in 60 μl of 2.5x transcription buffer (50 mM Tris/HCl pH 7.5, 80 mM MgCl2, 500 mM KCl, 5 mM DTT). Run was initiated by addition of 0.5 mM rATP, rCTP, rGTP supplemented with 200 μCi of α-P32UTP and incubated at 30°C for 2 min. Reaction was stopped by addition of 1ml of ice-cold TMN, and cells were pelleted. RNA was isolated as for Northern Blot analysis, precipitated and before hybridization treated with 0.15 M NaOH on ice for 5 min, neutralized, mixed with hybridization solution (10 mM Tris/HCl pH 7.5, 82.5 μg/ml of yeast tRNA, 10x Denhardt's solution, 0.3 M NaCl, 10mM EDTA, 1% SDS, 0.5% BSA), heat-denatured and hybridized to 10 μg of RNA probes slot-blotted to BrightStarR-Plus membrane at 65°C for 48 hours. After hybridization, membranes were washed twice at 65°C in 2xSCC supplemented with 1% SDS for 30 min.
Total RNA was extracted from exponentially growing cells and Northern blot analysis conducted using the NorthernMax™ kit (Ambion). Probes were synthesized by in vitro transcription with α-P32-UTP using the T7 MAXIscript kit (Ambion).
We thank K. Noma for strain constructions, D. McPheeters for run-on protocol, M. Lichten, F. Reyes-Turcu, E. Bartlett and H. Cam for comments on the manuscript, P. Fitzgerald for designing microarrays, and D. Venzon for advise with data analyses. This research was supported by the Intramural Research Program and under Contract N01-CO-12400 of the National Institutes of Health, National Cancer Institute.
Full Methods are available in the online version of the paper.
Supplementary Information is linked to the online version of the paper.