In order to uncover any connections between transcription, intron splicing, and chromatin marks in rapidly growing fission yeast cells, we determined global Pol II occupancy using chromatin immunoprecipitation on microarray (ChIP-chip) experiments. Furthermore, we applied ChIP-chip experiments to analyze the global distributions of histone H3 and lysine 36 trimethylation of histone H3 (H3K36Me3), a modification that is enriched in the body of actively transcribed genes 14 . In addition, to verify the histone H3 occupancy and reveal genomic regions that are relatively protein free, we applied formaldehyde-assisted isolation of regulatory elements (FAIRE) 15 16 . We used the same high-density Affymetrix tiling array platform for all these genome-wide approaches (Materials and methods).

The 5' ends of genes, corresponding to the nucleosome-free regions of promoters, had high FAIRE signals in fission yeast (Figure 1). These results are consistent with the originally published results in human 15 . Figure 2 shows the average patterns for the different chromatin- and transcription-related features across intron-less and intron-containing genes. Peaks of Pol II enrichment were evident in the promoter regions of genes, reflecting the accumulation of Pol II before transcription elongation 17 18 . Moreover, these regions showed high FAIRE signals, but relative depletion of histone H3 and, even more so, for its H3K36Me3 modification (Figure 2).

Gene promoters are known to contain nucleosome-free regions 19 20 21 . Notably, we found that the 3' ends of genes, corresponding to the terminator regions, also show Pol II enrichment, low histone H3 density and high FAIRE signal (Figures 2 and 3). While the nucleosome-free regions in promoters have been well characterized, a similar depletion of nucleosomes in terminator regions is not as well defined. A recent report in budding yeast shows depletion of nucleosomes at the 3' end of transcribed genes, and this depletion is coupled to transcriptional activity 22 . Our findings are also consistent with reports in mammalian cells that describe pausing of Pol II in terminator regions 23 24 . The start and end of introns showed lower levels of H3 occupancy (Figure 2b). This pattern might result from a 'looped' arrangement of exons and introns analogous to that proposed for the human BRCA1 gene 25 . Although this exon-intron pattern is not reflected in FAIRE, the overall patterns support the notion that nucleosome density is likely the major determinant for the FAIRE signals.

We next assessed the effects of transcript levels on the observed Pol II- and chromatin-related patterns across genes. To this end, we sorted all genes with measurable expression on Affymetrix chips into decile ranked groups, with the first decile representing the 10% most highly expressed genes, and so on. Average expression values for unspliced and spliced genes were calculated for each data set and for each expression bin and plotted either relative to the values in each bin (Figure 3) to highlight the range within each expression group or on a single scale according to the range of values of the entire dataset (Figure 4) to show the absolute enrichment. This analysis revealed that gene expression levels strongly influence the Pol II- and chromatin-related patterns. Coherent differences depending on expression level group were apparent (Figure 4): the most highly expressed genes showed the highest Pol II occupancy (Figure 4a), but the lowest density of histone H3 (Figure 4b), and the highest levels of H3K36me3 modification (after correcting for nucleosome density; Figure 4c). Glover-Cutter et al. 26 made similar observations of inverse enrichment between Pol II and nucleosomes, which could reflect displacement of nucleosomes by Pol II. The Pol II patterns were also apparent at the level of highly or lowly expressed single genes (Additional file 1).

Expression level-dependent differences in Pol II promoter patterns were also apparent: only lowly expressed genes showed promoter regions with higher levels of Pol II enrichment relative to downstream exonic regions, while highly expressed genes had more Pol II in exonic regions (Figure 3a). While Pol II is known to pause on promoters of some genes 17 18 , a global and gradated relationship between gene expression level and Pol II enrichment at promoters has not been reported previously. Unlike the promoter-associated accumulation, the increased Pol II occupancy at the terminator region showed little difference with varying expression levels (Figure 3). Our data show that Pol II behavior in terminator regions is less dependent on expression level than in promoter regions. The enrichment of Pol II in the terminator region could reflect the time required for the release of Pol II from the DNA and/or interactions between promoter and terminator regions 25 27 28 .

To our surprise, intronic regions showed distinct patterns with respect to the chromatin-related features. The overall H3 occupancy was lower in introns than in surrounding exons, and it dropped even lower in exon-intron junctions at both 5' and 3' ends of introns (Figure 2b). A pattern of decreasing nucleosome occupancy at exon-intron boundaries has also been described in other organisms 29 30 . Accordingly, the FAIRE signals were substantially higher in introns than in exons, similar to the promoter and terminator regions (Figures 2b and 5a). This effect of increased FAIRE signals was not dependent on intron position within genes (Figure 6b). The differential patterns were not caused by sequence bias between introns and exons because the hybridization signals were normalized using genomic DNA signals (or input signals for ChIP-chip experiments) to correct for hybridization differences due to GC content 7 31 . Moreover, within average introns, we observed higher Pol II and FAIRE signals towards the 3' ends of introns (Figure 2b). This effect did not reflect any sequence disparity: a GC content comparison of 25-bp sequences (the length of one microarray probe) at either end of introns revealed no significant differences (p = 0.48).

Finally, we also detected significantly lower densities of the H3K36me3 modification (normalized for histone H3 density) within introns compared to surrounding exons (Figures 2b and 5c). Other papers have also reported such differential marking of introns and exons for the H3K36me3 modification in worms and humans 12 29 30 . This modification is enriched within the ORFs of transcribed genes and is catalyzed by the histone methyltransferase Set2 32 33 , which is conserved in fission yeast 14 . The H3K36me3 modification depends on the interaction of Set2 and the terminus of Pol II 34 ; it is possible that the altered transcription kinetics that we detect in intronic regions interferes with H3K36me3 marking. It has been reported that Pol II in human cells is more enriched in exons than in introns, the reverse from our data 29 . A possible explanation for this discrepancy is that, in the previous study, any signals that fall within a 400-bp window (centered on the exon) are associated with that exon. Given the small average size of human exons (approximately 200 bp), extended intronic sequences on either side of exons would have been included with the exons for the analysis. If Pol II pauses at the 3' end of introns, as indicated by the Pol II and FAIRE enrichments in the much smaller fission yeast introns (typically < 100 bp), this may not have been detected in human 29 . Of course, it is also possible that Pol II progression across exonic and intronic regions differs between fission yeast and human genes.

As described above for other gene regions, we also assessed the effects of transcript levels on the observed Pol II- and chromatin-related patterns across intronic regions (Figures 3 and 4). The introns of lowly expressed genes showed more pronounced drops in H3K36me3 modification signals relative to neighboring exons (Figure 3c). Strikingly, the relative difference in Pol II enrichments in introns compared to exons was directly related to the expression level of genes: the ratios of intronic to exonic Pol II occupancy levels increased with decreasing gene expression (Figure 7a). The same effect was evident when plotting the p-values of t-tests of the intron and exon signals for each expression bin against expression bin numbers (Figure 7b). These data, which cannot be explained by biased intron size as a function of gene expression (Additional file 2), demonstrate that with decreasing gene expression, there is decreasing H3K36me3 modification and increasing Pol II accumulation within intronic regions relative to exonic regions.

In accordance with the FAIRE signals, Pol II occupancy was also significantly higher on average in intronic regions than in exonic regions (Figures 2b and 5b). Consistent results were obtained from Pol II occupancy and quantitative real time PCR data of single genes (Additional files 1 and 3). This increased Pol II signal was not dependent on intron position within genes (Figure 6b). The average intron of spliced genes thus showed a pattern of Pol II enrichment similar to the promoter and terminator regions, raising the possibility that Pol II also accumulates in intronic regions. Notably, the Pol II and FAIRE signals increased throughout intronic regions and peaked towards the 3' ends of introns (Figure 3a, d). We therefore propose that Pol II actually accumulates at the 3' end of introns before resuming transcription. Given that Pol II accumulation was most pronounced in the most lowly expressed genes (Figures 3a and 7), any pausing seems to mostly affect genes that are poorly transcribed. Anti-sense transcription is unlikely to cause Pol II accumulation in introns as only 11 of 372 anti-sense transcripts actually overlap with introns and none reside entirely within introns 7 .

Analyzing the processivity of Pol II, it has been noted that transcription does not continuously progress at the highest possible speed 35 . Pol II enrichment in introns could be related to observations that transcriptional speed can play a role in influencing alternative splicing of transcripts 36 . We envisage two, not mutually exclusive, possibilities why Pol II is enriched on introns. First, certain chromatin remodeling factors required to displace nucleosomes could be limiting. Recent studies have noted that exons contain well positioned nucleosomes relative to introns 29 30 37 . A sudden 'road block' of nucleosomes at the end of introns might cause Pol II to slow down or pause. Alternatively, or in addition, Pol II enrichment in introns could be directly linked to co-transcriptional splicing and could reflect the time required for splicing to finish before transcription can resume. Although we have not investigated the dynamics of the Pol II enrichment, evidence exists for a kinetic link between transcription and splicing 38 , where cellular treatment to pause elongating Pol II results in increased co-transcriptional splicing. Some RNA processing or export factors are known to be associated with intronic regions 39 . Moreover, we have previously observed a global coordination between transcriptional and splicing efficiencies, with increased transcription leading to increased splicing in two genes tested 7 .

Wild-type fission yeast cells (972 h^- ) were grown in rich yeast extract media at 32°C before being harvested for all experiments at exponential phase (approximately 5 × 10⁶ cells/ml).

Chromatin immunoprecipitions were performed, in biological duplicate, as described 7 using an antibody specific for the Pol II carboxy-terminal domain (CTD) (4H8, Abcam Cambridge, UK), histone H3 (ab1791, Abcam) or H3K36me3 (ab9050, Abcam). The two Pol II ChIP-chip experiments analyzed here were those reported by 7 . The whole cell extract was prepared using a Fastprep machine with glass beads to break cells after fixation, and the resulting lysate was sonicated to an average size of approximately 150 bp using the bioruptor (3 × 5', 30 s on, 30 s off). The immunoprecipitated material and input control were amplified in two steps as described 48 . During the second step, dUTPs were added to the PCR mix for subsequent fragmentation of the products. Fragmentation and labelling of the amplified products were performed using the GeneChip^® WT Double-Stranded DNA terminal labelling kit (Affymetrix Santa Clara, CA, USA). The duplicated immunoprecipitated samples and corresponding input material were hybridized on four separate Affymetrix GeneChip^® S. pombe Tiling 1.0FR arrays. The log₂ signals of the probes on the input arrays were subtracted from the log₂ signals of the Pol II arrays and the biological replicates were averaged. The H3K36me3 signals were normalized for the histone H3 signals.

Biological triplicates of FAIRE were performed essentially as described 15 . Briefly, yeast cells were fixed with formaldehyde in medium at a final concentration of 1%. Cells were left to incubate for 10 minutes at room temperature before being spun down, washed once with water, and resuspended in the same lysis buffer as for ChIP with protease inhibitors (mini-complete EDTA free tablets, Roche Applied Science, Welwyn, UK). Cells were broken using glass beads and a Fastprep machine (20 seconds at 6.0 m/s) and then sonicated using a Bioruptor (Diagenode, Liège, Belgium) with 6 minutes total time (15 s on, 30 s off). DNA was phenol/chloroform extracted twice, and the resulting material was RNAse treated for 20 minutes before re-precipitating. The resulting DNA was then labeled according to standard Affymetrix protocols. The log₂ FAIRE signals were normalized by subtracting the average signal of three genomic DNA hybridizations to correct for GC bias.

For analyzing differences in Pol II, histone H3 and H3K36me3 ChIP or FAIRE signals in introns and exons, 25-bp Affymetrix probes were mapped back to the S. pombe genome (GeneDB). Probes where the entire 25 bp length fell within an intron or exon were classed as 'intron probes' or 'exon probes', respectively. All probes that fell entirely within the ORF of unspliced genes were used for calculating the signal of unspliced genes.

Every annotated S. pombe gene (downloaded from GeneDB) was divided into three parts, the promoter, coding, and terminator regions; in the case of spliced genes, the ORF was further divided into exons and introns. For both unspliced and spliced genes, 400-bp regions upstream of the start of the ORF and downstream of the end of the ORF were taken as the promoter and terminator, respectively. These 400-bp windows were divided into ten bins of 40 bp each, and Affymetrix probes were assigned to bins depending on where their midpoint fell (13th base pair). For the ORFs of unspliced genes, the lengths of ORFs were divided into 20 bins of equal size, with Affymetrix probes being assigned to bins based on their midpoint position. For spliced genes, each intron and exon was first divided into 10 or 20 bins of equal size, respectively, with probes assigned to bins based on their midpoint. In order to calculate an average of every exon-intron-exon junction without counting probes multiple times, the last ten bins of every upstream exon, the ten bins of every intron, and the first ten bins of every downstream exon were used to average probe signals from the various experiments. Probes falling in the first ten bins of every first exon and the last ten bins of every last exon were averaged to create the first and last ten bins for upstream and downstream exons, respectively.

Replicate gene expression data collected previously 7 from Affymetrix Yeast 2.0 Genechip^® arrays were first filtered for undetectable signal (< 1; 480 of 5,296 genes excluded) and then sorted into spliced and unspliced genes (2,218 and 2,598 genes, respectively). Lists of spliced/unspliced genes were then ranked in descending order and split into 10 equal groups (approximately 220 and 260 genes per group for spliced and unspliced genes, respectively). Average gene profile calculations were then performed as described above for genes within each expression bin.

All microarray data used have been submitted to ArrayExpress under the accession number E-TABM-946.