"Dynamic transcriptome of Schizosaccharomyces pombe shown by RNA-DNA hybrid mapping".

Natalie Dutrow ^{1, 2}, David A Nix ^{3, 4}, Derick Holt ^{1, 2}, Brett Milash ⁴, Brian Dalley ⁵, Erick Westbroek ^{1, 2}, Timothy J Parnell ^{1, 2}, and Bradley R Cairns ^{1, 2}

¹ Department of Oncological Sciences, University of Utah School of Medicine, Salt Lake City, Utah 84132, USA.
² Howard Hughes Medical Institute and Huntsman Cancer Institute, University of Utah School of Medicine, Salt Lake City, Utah 84112, USA.
³ Research Informatics, Huntsman Cancer Institute, Salt Lake City 84112, USA.
⁴ Bioinformatics Core Facility, Huntsman Cancer Institute, University of Utah School of Medicine, Salt Lake City 84112, USA.
⁵ Microarray Core Facility, Huntsman Cancer Institute, University of Utah School of Medicine, Salt Lake City 84112, USA.

Correspondence to: Bradley R Cairns ^{1, 2}   e-mail: brad.cairns@hci.utah.edu 

We have determined the high-resolution strand-specific transcriptome of the fission yeast S. pombe under multiple growth conditions using a novel RNA-DNA hybridization mapping (HybMap) technique. HybMap uses an antibody against an RNA-DNA hybrid to detect RNA molecules hybridized to a high-density DNA oligonucleotide tiling microarray. HybMap showed exceptional dynamic range and reproducibility, and allowed us to identify strand-specific coding, noncoding and structural RNAs, as well as previously unknown RNAs conserved in distant yeast species. Notably, we found that virtually the entire euchromatic genome (including intergenics) is transcribed, with heterochromatin dampening intergenic transcription. We identified features including large numbers of condition-specific noncoding RNAs, extensive antisense transcription, new properties of antisense transcripts and induced divergent transcription. Furthermore, our HybMap data informed the efficiency and locations of RNA splicing genome-wide. Finally, we observed strand-specific transcription islands around tRNAs at heterochromatin boundaries inside centromeres. Here, we discuss these new features in terms of organism fitness and transcriptome evolution.

Supplementary Information:
http://www.nature.com/ng/journal/v40/n8/suppinfo/ng.196_S1.html

Recently, genome-wide approaches to monitor transcription have revealed several surprising features of the eukaryotic transcriptome, including extensive intergenic transcription, antisense transcription and a notable number of noncoding RNAs (ncRNA) 1, 2, 3, 4, 5, 6. Furthermore, it has been found that, in some cases, it is the act of transcription (or the attendant chromatin modifications) that is important, rather than the ncRNA produced 7. Together, these studies have initiated a new era in defining genetic elements and generated high interest in understanding the purpose and function of ncRNAs in the genome. Genome-wide transcriptome mapping has emerged as an important approach to help understand transcriptional dynamics and transcription-chromatin relationships, and to identify candidate functional ncRNAs.

Traditional transcript profiling approaches are often fraught with bias because of difficulties in labeling the isolated RNAs for detection on microarray formats 8. To circumvent these limitations, we made use of a previously characterized RNA-DNA hybrid antibody that has been used to quantify RNA levels of structured and small RNAs in an array format 9. Here, we adapted and extended this approach, creating a technique for acquiring a strand-specific transcriptome of an entire genome, termed HybMap, that is not susceptible to many of the common technical limitations of traditional methods. To apply HybMap, we chose the fission yeast Schizosaccharomyces pombe, an important model organism for understanding transcription and chromosomal biology with a highly compact genome. Basic elements of the S. pombe genome include a total of 13.8 Mb of DNA on three chromosomes, 20 kb of mitochondrial DNA, and fewer than 5,000 protein-coding genes (1 per ~2.5 kb). Detailed features of the genome will be discussed in the context of the transcriptome.

Our work had four goals: (i) to develop a technique for deriving an accurate strand-specific transcriptome of a genome at high resolution and low cost, (ii) to apply this technique to derive the S. pombe transcriptome, (iii) to uncover new candidate ORFs and ncRNAs for future studies and (iv) to reveal new features of transcription that inform our understanding of the relationship between transcription and chromatin.

Results:

HybMap RNA-DNA hybrid mapping

To create a high-resolution and strand-specific array, we tiled both strands of the entire S. pombe genome with 458,566 60-mer DNA oligonucleotide 'probes' (55-base resolution, 5-base overlap). All the 60-mer probes on the top (forward) strand had a precise 60-mer reverse complement probe on the bottom (reverse) strand, allowing strand-specific transcription to be monitored. We also included 500 60-mer control probes from the zebrafish (Danio rerio) genome not present in the S. pombe genome for array normalization and to provide a nonspecific hybridization background level of signal useful for comparison to true S. pombe probes; the median level of the zebrafish control probes will hereafter be termed 'background' in the text and will be depicted as a dotted line in all figures.

The HybMap technique can utilize total RNA isolated without poly(A) selection, reverse transcription, RNA (or DNA) amplification or nucleic acid labeling; thus omitting steps that might strongly bias the RNA pool and/or disfavor the isolation of small noncoding or structured RNAs. Purified total RNA was hybridized to the tiling array, and then incubated with a monoclonal antibody (S9.6) that recognizes an RNA-DNA hybrid. We then incubated arrays with a secondary antibody labeled with the fluorophore Cy3, allowing quantitation (Supplementary Methods online). The S9.6 antibody to RNA-DNA hybrids has been used previously for transcript detection in an array format 9, 10. S9.6 has negligible sequence specificity 10 and does not show a significant bias for GC content (Supplementary Fig. 1 online). S9.6 recognizes a hybrid of ~15 bp 9, a result we have independently verified by analyzing signal from tRNAs with different registers of overlap with array probes (Supplementary Fig. 2 online). S9.6 is highly sensitive to mismatch: one mismatched base pair in a stretch of 15 bp reduces signal by 80-fold, and a second mismatch reduces ~20,000 fold (Supplementary Fig. 3 and Supplementary Table 1 online).

Evidence for extensive transcription in S. pombe

We purified total RNA from two separate cultures (biological replicates) of a prototrophic strain grown in rich medium (poly(A) RNA presented later) and carried out HybMap (see Methods). HybMap provided ~7,600-fold range of signal intensity (12.9 on a log₂ scale) and virtually all the probes on the array showed mean signal intensity well above the background control probes (Supplementary Fig. 4 online). Notably, 99.6% of all smoothed 2-probe windows had a signal above background, providing initial evidence for widespread transcription of the euchromatic S. pombe genome. Widespread transcription in other genomes has been reported previously 11, 12; however, our use of internal control probes and a strand-specific approach allows for a direct and quantitative comparison genome-wide. We note that HybMap, like all array formats, does not directly measure transcription rate, but rather provides a steady-state RNA measurement.

Transcription in relation to a defined baseline

Determining whether a gene is activated or silenced requires a baseline standard of reference. The vast majority of intergenic regions (heterochromatin regions excepted) seemed to be transcribed at a very similar level, ~12-fold above background (Fig. 1a). To accurately quantify this basal level, we determined the median value of transcription of all intergenic regions, following their trimming on either end by 240 bp to limit the impact of numerous 5' and 3' UTR extensions (discussed below). This trimmed intergenic value is hereafter referred to as the 'baseline' and is depicted by a solid line in the figures; the baseline is 12-fold above the background. Notably, signal intensity generally deflects from baseline along the physical map in accordance with annotated genes and their strand specificity.

Figure 1: Extensive transcription of the S. pombe genome.

Figure 1 : Extensive transcription of the S. pombe genome.

(a) S. pombe probes were categorized according to genomic feature and classified into sense and antisense probes on the basis of gene orientation.

(b) The percentage of the genome that is transcribed (y axis) at different signal thresholds (x axis), shown for two different units of measurement, genes or base pairs.

In accordance with extensive transcription, 87% of 5,049 genes, 79% of interrogated base pairs (25,180,884) and 75% of oligonucleotide probes were expressed at or above the baseline level (Fig. 1b). On the basis of signal intensity, we calculated that the average annotated gene is transcribed approx12-fold above baseline and 149-fold above background. Median signals for different gene classes are shown in Figure 1a. For Pol II genes, the highest signals, 630-fold above baseline and ~7,600-fold above background, mapped to ribosomal protein genes.

Transcriptional features of euchromatin

To illustrate the high-resolution features of euchromatin, we show the HybMap signal intensity of two genomic regions alongside current annotation (Fig. 2a,b). One region has a relatively simple transcriptional architecture; most genes reside in tandem orientation on one strand (bottom/reverse), and the data generally conforms to current annotation, with UTR designations excepted (Fig. 2a). The HypMap results of the second region deviate considerably from annotation in several locations, with, for example, attribution of sno52 to the incorrect strand, longer untranslated RNA (UTR) extensions, and the appearance of antisense transcription (Fig. 2b). We will analyze elements from both of these regions in detail throughout the manuscript to illustrate concepts.

Figure 2: Features of transcription and splicing in euchromatin.

Figure 2 : Features of transcription and splicing in euchromatin.

High-resolution views of different loci are depicted. The physical map is presented in a strand-specific manner, with currently annotated exons (gray boxes) and introns (gray lines). RNA signal intensities for probes along the top/forward (blue) and bottom/reverse (orange) strands are provided (probe centered). The background (dashed line) and the baseline (solid line) are shown for each strand.

(a) A locus that generally conforms to current annotation. Brackets denote regions where signal drops below baseline on only one strand, near the TSS.

(b) A locus that deviates from current annotation at several locations. Unannotated UTRs and antisense transcription are evident.

(c) A typical locus, where signal intensity drops in accordance with annotated introns.

(d) snoR54 signal reflects its transcription and processing from an intron within gua1. snoRNAs are highly stable and typically accumulate to levels well beyond their host RNAs.

(e) tRNA genes are multicopy and highly transcribed, resulting in a saturated signal. We note that because of the high sequence conservation among tRNAs, their signal cannot be uniquely attributed.

Features of UTRs and RNA processing

As in other organisms 1, 3, we commonly observed the extension of 5'- and especially 3'-UTRs for Pol II genes beyond current annotation: 65% have extensions beyond their annotated 5' end, and 87% have extensions beyond their annotated 3' end; the median 5' UTR is 81 bp and the median 3' UTR is 231 bp. Additional features, GO TERM classification and discussion of UTRs (including overlapping convergent 3' UTRs) are provided in Supplementary Table 2, Supplementary Figure 5 and Supplementary Note online.

According to current annotation, introns reside in 43% of genes and average ~81 bases 13. Thus, we interrogated most introns by one probe fully within the intron-exon boundary (intron probes) and two probes that overlap the boundary (boundary probes). Our data verified the notion that splicing is efficient in S. pombe, as intron probes showed a mean 32-fold decrease in signal intensity relative to exon probes (Fig. 2c). The current annotation of predicted introns seemed largely accurate; 94% of annotated introns examined showed a pronounced decrease in signal intensity. However, of the remaining 6%, 5% were expressed at a level similar to that of the surrounding gene, suggesting either inefficient splicing or possible misannotation. Notably, the remaining 1% showed a signal intensity greater than eightfold that observed at the surrounding gene (Fig. 2d), including three potential ncRNAs residing within introns of annotated protein-coding genes, which we verified by strand-specific qPCR—rps 3, SPBC1734.10c and SPAC6F6.03c (Supplementary Fig. 6 and Supplementary Table 3 online). Unique transcripts processed from introns are rare; only four are annotated in S. pombe, and all are snoRNAs (Fig. 2d). As mature snoRNAs are structured and lack a poly(A) tail, they are highly underrepresented in other transcriptome formats 3, whereas they are exceptionally prominent features in the HybMap datasets. Finally, tRNAs are structured RNAs that provide exceptionally high signal (Fig. 2e), although we note that their similarity precludes their precise mapping.

The abundance of non-annotated transcripts

We observed large numbers of non-annotated transcription units (fragments), termed 'transfrags' 12. For example, the recently identified 14, 15 telomerase ncRNA (TER1) was easily observed in our dataset (Fig. 3a). We identified transfrags on the basis of the following criteria: (i) two or more consecutive probes with a pseudo median value at least 3.5-fold above baseline and (ii) separation by at least 180 bp from any annotated element. Using these thresholds, we defined 7,093 previously unknown transfrags in rich medium, with increasing thresholds yielding moderately lower transfrag numbers (Fig. 3b). Of these, 38% reside in intergenic regions (Fig. 3c) and 35% are antisense to a known transcript (Fig. 3d). The remainder reside in intergenic–UTR boundaries and are a mixture of independent transfrags and long unannotated UTRs. Twenty transfrags were tested by both RT-PCR (Supplementary Table 3) and strand-specific quantitative PCR, and all were verified (Supplementary Fig. 6). We then analyzed the genome-wide occupancy profile of RNA Pol II (ref. 16) and found that most highly transcribed transfrags, including SPAC6F6.03c (noted above) show levels of Pol II well above the genome average (Fig. 3c–e and data not shown). Taken together, HybMap and the Pol II ChIP data reveal large numbers of potential new transcripts.

Figure 3: Quantitation of transcription fragments and potential noncoding RNAs.

Figure 3 : Quantitation of transcription fragments and potential noncoding RNAs.

Transfrags are quantified on the basis of threshold criteria. New transfrags identified by HybMap are denoted by hatched boxes.

(a) Recently identified ncRNA TER1 is observed as a transfrag in our dataset.

(b) Genome-wide analyses yield similar high numbers of transfrags in each condition.

(c–e) Data from poly(A)-enriched RNA (overlaid blue line) and a Pol II ChIP (green line) provide evidence for the existence of potential ncRNAs. Transfrags are found in three general genomic environments: in intergenic regions (c), in the antisense strand of known genes (d) and in intronic regions (e).

Transcriptomes for alternative growth and stress conditions

We then derived transcriptomes for alternative growth and stress conditions: minimal medium, heat stress and DNA damage (methyl methanesulfonate, MMS). As expected, GO analyses showed changes in gene classes that related to the conditions imposed (Supplementary Tables 4, 5, 6, 7, 8, 9 online) 17. Here, we discuss only key general features, showing the signal intensity from rich medium in a bar format and that from the alternative condition as a superimposed colored line (Fig. 4). For heat stress, the heat shock gene hsp16 is moderately transcribed at 30 °C but upregulated more than tenfold during heat stress (Fig. 4a). Likewise, genes for thiamine biosynthesis (thi4) and phenazine metabolism (ase1) are upregulated in minimal medium or DNA damage, respectively (Fig. 4c,e). Full datasets are available at the National Center for Biotechnology Information Gene Expression Omnibus (NCBI GEO) database (see Accession codes section in Methods).

Figure 4: Transcriptomes and loci derived from alternative growth conditions.

Figure 4 : Transcriptomes and loci derived from alternative growth conditions.

Signals from alternative growth conditions (colored line) are overlaid on standard conditions (bars).

(a–f) Under heat stress (a,b), minimal medium (c,d) or MMS treatment (e,f); the expression of particular genes (a,c,e) and transfrags (b,d,f) is induced.

Transfrags upregulated during alternative growth conditions

To identify candidate functional ncRNAs for each growth condition, we defined the transfrags present in each condition that increase or decrease substantially relative to their values in rich medium, termed 'diffrags' (differential transfrags). To quantify diffrags, we first selected particular transfrags—those upregulated >3.5-fold from baseline in their initial condition and present in intergenic regions >180 bp away from an annotated gene (total ~7,200, Fig. 3b). We then determined (for each condition) the percentage of these transfrags whose signal intensity differed more than twofold from their levels in rich medium. By these criteria, we found that the percentage of transfrags that are 'diffrags' is 24% in heat stress, 12% in minimal media and 15% in the presence of MMS. Thus, a substantial fraction of the non-annotated transcriptome is altered when growth conditions are changed. Examples of notable diffrags include an ncRNA arising on an antisense strand during heat stress (Fig. 4b), and two arising from intergenic regions near a 5' end of a gene, either in minimal medium or MMS (Fig. 4d,f).

Conservation of transcribed loci

We then determined whether RNAs newly identified through HybMap are conserved in distantly related yeast species. We selected twenty newly identified RNAs that spanned multiple probes (typically >5 probes) and were expressed >10-fold above baseline, and then interrogated two recently sequenced distant relatives of S. pombe, S. japonicus and S. octosporus. Using a stringent cut-off (E value of 10^-4), we found that 11 of these RNAs were conserved in S. octosporus, and 3 in the more distantly related S. japonicus. This analysis suggests that HybMap is indeed revealing new functional transcripts. To determine the potential for these RNAs to encode proteins, we carried out a 3-frame translation and searched for ORFs with a codon bias specific for S. pombe; 8 of the 20 RNAs revealed such an ORF. Finally, our intersection of the conservation and ORF searches identified two RNAs (transfrags Q and AA) that share both codon bias and sequence conservation within their predicted longest ORF (discussed further below and in Supplementary Table 10 online). Thus, HybMap identifies transfrags with significant potential for function, either as ncRNAs or small proteins.

Properties of antisense transcripts

Antisense transcription occurs at particular loci in the mammalian genome 18 and in the S. pombe 19 genome. Here, we characterize its extent genome-wide, its properties and its source (Fig. 1a and Fig. 5). First, we observed a significant correlation between signals from sense-antisense probe pairs (+0.595, P value <2 X 10^-16), and between antisense signal and RNA Pol II occupancy (+0.50, P value <2 X 10^-16). Sense-antisense correlation is especially clear in dynamic transcriptome comparisons: genes upregulated >fourfold during heat shock show greatly increased antisense transcription, and repressed genes show a substantial decrease (Fig. 5a). Of note, antisense transcription is slightly negatively correlated with histone H3 occupancy (–0.20; P value <2 X 10^-16) 16, but slightly positively correlated with H3K36me3 (+0.21; P value <2 X 10^-16), a mark found on Pol II-transcribed regions. Together, these results raise the possibility that areas that are highly transcribed and slightly histone deficient are susceptible to increased antisense transcription; that is, that strong sense transcription begets antisense transcription.

Figure 5: Features of antisense transcription.

Figure 5 : Features of antisense transcription.

(a) Correlation of sense and antisense transcription. The 432 genes that differ at least fourfold in heat shock from standard conditions are plotted. For each gene, the median change for the sense strand was plotted against the antisense strand.

(b,c) Signal in standard conditions (bars) overlaid with poly(A)-enriched data (overlaid blue line). Antisense transcription may be generally divided into two classes comprised of genes with negligible polyadenylated antisense transcripts (the vast majority) (b) and rare genes with polyadenylated antisense transcripts (c).

Notably, our examination of poly(A)-selected RNA samples showed that antisense transcripts were highly reduced (Fig. 5b and Fig. 6a), suggesting that most antisense transcripts lack polyadenylation. Furthermore, the deflection of signal from background at genes was often greater, and the distinction between annotated genes was clearer as a result of the lack of antisense signal (Fig. 6a). However, we did find a few locations where antisense signal is high along the entire gene in our poly(A) datasets, suggesting that certain antisense transcripts, including those derived from genes encoding certain transcription factors, are indeed polyadenylated (Fig. 5c and Supplementary Table 11 online). Additional features of antisense transcripts at the different RNA PolI/II/III classes are described in the Supplementary Note.

Figure 6: Poly(A) RNA shows divergent transcription and previously unknown RNA transcripts.

Figure 6 : Poly(A) RNA shows divergent transcription and previously unknown RNA transcripts.

(a) We depict the locus in Figure 2b overlaid with poly(A) data (solid purple line). Brackets denote three short transcripts diverging from hsp60, cip2 and SPAC630.07c. Notably, these transcripts would not be distinguishable in total RNA, as a result of the prevalence of nonadenylated antisense.

(b,c) A polyadenylated transcript clearly corresponds to each of the two transfrags most conserved among the related yeast species.

Induced divergent transcription

A particularly marked feature of our datasets was the prevalence of 'induced divergent transcription', in which transcripts were found on the opposite strand directed away from the promoter of a highly transcribed annotated gene. For example, during heat stress, a transcript diverges away from the hsp16 promoter on the opposite strand, generating an antisense transcript that encompasses most of the flanking gene (Fig. 4a); this was confirmed by strand-specific qPCR (data not shown). Another confirmed instance is the cnl2-thi4 locus in minimal medium (Fig. 4c and data not shown).

Previously unknown divergent poly(A) RNAs were also common in our static poly(A) datasets, suggesting that divergent transcription is a feature of basal transcription at many loci. For example, we found that poly(A) transcripts diverge from hsp60, cip2 and SPAC630.07c (Fig. 6a). Two of these divergent transcripts, hsp60 and cip2, were tested and verified by strand-specific qPCR (data not shown). These unannotated poly(A) transcripts are obscured in the total RNA dataset as a result of the high levels of antisense (which is generally not polyadenylated), but are obvious transfrags in the poly(A)-selected RNA dataset (Fig. 6a). In addition, we found a clear strand-specific polyadenylated transcript corresponding to the two transcripts (transfrag Q and AA, Supplementary Table 10) that are highly conserved among the three related yeast species and that share an S. pombe codon bias (Fig. 6b,c). The total RNA and poly(A)-selected RNA datasets are thus useful for uncovering different aspects of the transcriptome.

A strand-specific drop in signal before the TSS

One feature we commonly observed at tandem genes is a reduction in signal intensity of sense probes just before the true transcription start site (TSS). A clear example of this is shown in Figure 2a, where 'dips' are seen selectively in the sense strand just adjacent to the 5' end (note that these genes are on the bottom/reverse strand). Although common, it is not a constant feature. At present, we do not understand the basis for this common and curious feature.

Intergenic transcription measures telomeric chromatin

The S. pombe genome contains three types of heterochromatic regions: telomeres, centromeres and the silent mating locus. In yeasts, repressive telomeric heterochromatin propagates from the telomere end (which bears repetitive elements) into the chromosome arm, and is opposed by chromatin modifications associated with transcription 19, 20, by transcription itself, or by Pol III genes (TFIIIC binding sites) 21. In S. pombe, heterochromatic regions bear H3K9me (ref. 22). A notable feature of our data is the extent to which intergenic transcription gradually falls from baseline levels to near background levels as examination progresses from the gene-rich arm toward the telomere (Fig. 7a); intergenic transcript signals drop on both strands in synchrony as H3K9me is encountered. We suggest that the level of opportunistic transcription within intergenic regions serves as a proxy measurement for the extent of chromatin silencing.

Figure 7: Transcriptional features of heterochromatic loci and flanking regions.

Figure 7 : Transcriptional features of heterochromatic loci and flanking regions.

Signals derived from the top/forward and bottom/reverse strands are overlaid, to enable direct comparison. A colored solid line (underneath) indicates the locations of unique (green) and non-unique (purple) probes.

(a) Transcriptional repression near telomeres. The physical map of the subteleomere of chromosome 1 is accompanied by maps of the heterochromatic mark H3K9me2 (ChIP data, black) and a mark correlated with transcribed ORFs, H3K36me3 (ChIP data, green). Notably, as the signal from H3K9me2 increases, intergenic transcription approaches the background level.

(b) Transcription within the centromere of chromosome 1. Notably, both strands are transcribed in the regions flanking the dh1 repeats. The opposite strand flanking the highly expressed tRNAs is transcribed. Results for chromosomes 2 and 3 are provided in Supplementary Figure 7.

Transcriptional features of centromeres

The centromeres of S. pombe are complex. For brevity, we will separate them into four general regions: (i) a partially unique central region that bears high levels of the histone H3 variant CENPA 23, 24, flanked by (ii) an innermost repeat that bears sets of tRNAs, which may help form a boundary 21, 25 between the central region and (iii) regions of repetitive DNA elements (dg, dhI, others), which are silenced through an RNAi-dependent mechanism (involving H3K9me and histone deacetylation) and are typically flanked by (iv) tRNAs of various numbers (chr. 1, right side excepted) present at the boundary between the silent repeat region (heterochromatin) and expressed genes 26, 27, 28, 29, 30, 31. With repetitive elements, the signal represents a class-average sum, and in certain repeat regions, there is little if any transcription when copy number is considered. However, certain sections of the centromere (that is, those flanking dh1) are clearly transcribed and have observable signal from both strands, consistent with earlier reports 32, and our data helps inform the locations and magnitude of that transcription (Fig. 7b and Supplementary Fig. 7 online). Results for an additional heterochromatic region (the MAT locus, Supplementary Fig. 8 online) as well as the rDNA locus (Supplementary Fig. 9 online) are also presented.

tRNAs and strand-specific transcription in heterochromatin

One notable feature of the centromere datasets is the effect of tRNAs residing in the imr repeats. We note that signal from tRNAs themselves cannot be uniquely attributed to a map location, given their near identity. The tRNA transcript itself, however, is not the feature of interest. Rather, we observed that the region surrounding the tRNA is selectively transcribed on the strand opposite the tRNA (Fig. 7b, Supplementary Fig. 7 and data not shown). In most cases, these flanking sequences are unique, allowing definitive attribution. Furthermore, this pronounced strand-specific effect is not generally observed with tRNAs in euchromatin (Fig. 2a,e). Thus, by an unknown mechanism, either Pol III transcription or machinery seems to create an 'island' within heterochromatin that enables opportunistic transcription on the strand opposite the tRNA.

Discussion:

Here, a method for high-density strand-specific transcriptome profiling is developed and applied to S. pombe. We find that HybMap is a simple, highly reproducible, accurate and relatively inexpensive technique that should be widely applicable to other organisms. As HybMap uses total RNA without amplification or labeling, structured and non-polyadenylated RNAs are retained and clearly detected. Although our studies were done with a single array platform (Agilent Technologies) the simplicity of the technique should allow its application to other array platforms, although we have not tested this directly. Our use of a large set of internal control probes (from zebrafish) enabled effective normalization of arrays and provided a measure of the nonspecific background signal, a feature important for distinguishing noise from true transcription. The sensitivity and wide dynamic range of the technique allowed us to examine new properties of intergenic and antisense transcription, as well as other features of the transcriptome.

First, we provide extensive information useful for improving the annotation of the genome, including large numbers of previously unknown ncRNAs at various threshold criteria. Of particular interest are the selective changes at non-annotated loci (diffrags) in alternative growth conditions. A challenge for future studies is determining what proportion of these transcripts contributes to general or condition-dependent fitness. Here, our separate datasets for both total RNA and poly(A)-selected RNA should prove useful for determining whether these transcripts are more likely new coding or ncRNAs.

We provide evidence for extensive transcription of intergenic regions in S. pombe. Although we have not evaluated the purpose for this transcription, we hypothesize several possible uses. First, this transcription is likely antagonistic to repressive chromatin and may help in chromatin dynamics by helping to increase histone turnover, thus providing searchable space for transcription factors and other DNA-binding proteins. Alternatively, it might be a mechanism for locating DNA damage, with RNA polymerase recruiting factors for transcription-coupled repair to damaged sites 33. Finally, we emphasize its possible utility in genome evolution: with opportunistic transcription allowing the cell to survey genome sequence space for new useful coding or ncRNAs and thereby transform opportunistic transcription into functional transcription. By this mechanism, an ncRNA or small ORF is provided the opportunity to evolve (through mutations or insertions, for example) from one of minor benefit to a more important contributor. Transcripts derived from intergenics were not generally polyadenylated, but may still be at levels sufficient for generating low levels of peptides whose composition and abundance could be modified through mutation and selection. A particularly notable feature was basal and induced divergent transcription at certain strong promoters, which were often polyadenylated (Fig. 6a). One interpretation is that these transcripts might arise from a misregulation of directional fidelity, with basal factors (such as TBP) and RNA polymerase simply assembling and initiating in the 'wrong' direction, perhaps as a result of the lack of sequence stringency for TBP binding. However, we speculate that this process might provide a means for evolving coordinated expression of coding or ncRNAs; through mutation, rearrangement or insertions, a more useful divergent promoter and transcript can evolve from this initial version. Thus, a transcriptome that might initially seem wasteful and inaccurate may, in fact, be a vehicle to assist in adaptation.

We observe extensive antisense transcription in genes, which is shown to generally lack polyadenylation. Thus, antisense transcription may originate without the full use of the factors (basal and chromatin) for site-specific promoter initiation and processivity, although this remains to be formally tested. We observed increased antisense transcription at highly transcribed and histone-deficient genes, attributes consistent with an opportunistic origin in open chromatin. At present, it is not clear whether antisense transcription within genes has a functional purpose in S. pombe. We note that both sense and antisense transcription are very low (near background) from intergenics within heterochromatin (telomeres; Fig. 7a). We suggest that intergenic transcription levels serve as a proxy measurement of the chromatin state of a large region, with polymerized heterochromatin preventing opportunistic initiation.

One key issue is how opportunistic sense and antisense transcription is insulated from the S. pombe RNAi pathway, as double-stranded RNAs can lead to siRNA silencing 32, 34, 35, 36. First, as HybMap does not measure the level of cellular dsRNA 37, it remains unknown whether the complementary sense/antisense transcripts are indeed double-stranded in vivo. However, the positive feedback loop for siRNA production is known to utilize nascent poly(A) RNA, whereas our data shows that the vast majority of antisense transcripts lack polyadenylation. Thus, a lack of polyadenylation may serve to insulate opportunistic dsRNA from entering the RNAi pathway.

One particularly notable feature was a sense strand-specific 'dip' adjacent to the 5' end, most commonly observed at tandem genes. One speculative possibility (among many) is that a DNA loop may form between an enhancer and the pre-initiation complex, and that the enhancer-binding protein could capture an opportunistic RNA polymerase traveling from the upstream gene and then pass Pol II directly to the pre-initiation complex, thus preventing it from passing through and transcribing the loop region in one particular direction. An additional strand-specific feature was the presence of 'islands' of transcription on the antisense strand around tRNAs in heterochromatin. Here, it will be of interest to determine how this occurs and whether this transcription helps form the boundary between the very different chromatin of the central region (bearing CENPA) and the pericentric repeat region. This feature may be mechanistically similar to the previously demonstrated transcription near heterochromatin boundary elements (TFIIIC binding sites) at the MAT locus 21. These are some of the many prominent and unexpected features of the S. pombe transcriptome worthy of exploration in future studies.

Methods:

Yeast strains and treatment conditions.

Yeast growth, manipulations and molecular biology were done according to standard protocols. We used a prototrophic Schizosaccharomyces pombe h90 strain for all experiments. For rich medium, the strain was grown on yeast extract with glucose (2%) supplemented with histidine, uracil, leucine and adenine at 32 °C to an optical density at 600 nm of 1.0, unless otherwise noted. We harvested the cells and, after mechanical pulverization, we stored the cell material at -80 °C. For the heat shock transcriptome, cells were cultured to an optical density at 600 nm of 1.0 and transferred to flasks that were preheated in a 40 °C water bath. Cells were incubated, with shaking, for 15 min and harvested. For the minimal medium transcriptome, cells were cultured as described in media lacking amino acids and containing 2% glucose. For the methyl methanesulfonate transcriptome, cells were cultured to an optical density at 600 nm of 0.6 and treated with MMS (0.02%) for 1 h. All samples were grown and collected in duplicate.

Nucleic acid purification.

We extracted total RNA using the Ambion mirVana kit and isolated poly(A) RNA with the Qiagen Oligotex PolyA Purification kit.

Design of tiling array.

We tiled the entire S. pombe genome, including the mitochondrial genome, with 60-mer oligos every 55 base pairs. We generated the reverse compliment of each probe on the forward strand in order to tile the reverse strand for a total of 458,566 probes. In addition, we added 500 probes from a zebrafish array with no homology to S. pombe as background controls. The zebrafish probes have the same median GC content (37%) and melting temperature (71 °C) as the S. pombe probes. We aligned each probe to the genome using the BLAST algorithm to acquire copy number information. The set of two 244K arrays was manufactured by Agilent Technologies.

Array hybridization and antibody detection.

RNA samples (15 mug total RNA or 1 mug poly(A) RNA) were fragmented and combined with Agilent Hi-RPM GE Hybridization kit reagents. We then hybridized these samples to the array slides at 65 °C for 17 h. The slides were washed in 6times SSPE, 0.005% N-lauroylsarcosine for 1 min at 20 °C and in 0.06times SSPE, 0.005% N-lauroylsarcosine for 1 min at 31 °C. We then incubated the slides with 0.025 mg/ml primary monoclonal mouse antibody to RNA-DNA hybrid complexes, available in a hybridoma cell line from ATCC (HB-8730). The antibody was diluted in 500 mul of SB+BSA (100 mM MES, pH 6.6, 1 M NaCl, 0.05% Tween 20, and 1 mg/ml BSA). The 60 min incubation is at 25 °C. After this incubation, we washed the slides four times for 3 min in NSWB (6times SSPE, 0.01% Tween 20). The secondary Cy3-labeled goat antibody to mouse (KPL, 078-18-061) was used at a concentration of 3 mug/ml in 500 mul of SB+BSA. We incubated the slides with the secondary antibody for 60 min at 25 °C. Finally, we washed the slides four times for 3 min in NSWB at room temperature and scanned them in an Agilent G2505B Microarray Scanner using extended dynamic range software. Feature extraction of the scanned images was done using Agilent Feature Extraction Software version 9.5.1.

Computational analytical methods.

All of the software used in this analysis is open source and available from the TiMAT2 project site. We obtained S. pombe annotation and genomic sequence from the Wellcome Trust Sanger Institute (referred to here as the April 2007 build). These files, as well as the low- and high-level analysis files, can be downloaded from our bioserver. See below for relevant URLs.

Transcriptome analysis.

For static maps, processing of the Agilent microarray transcriptome data was done in three basic steps: data normalization, sliding window summaries and enriched region identification. For each condition (growth and or RNA fraction) and strand, we extracted the median unadjusted signal intensities from the cy3 channel. We then mapped probes to the April 2007 build and retained those with <100 exact matches. Two biological replicate samples were isolated for each condition, and between replicate pairs the datasets were highly similar (heat stress, r = 0.96; minimal medium, r = 0.96; DNA damage, r = 0.97). We subject the paired datasets for each condition to quantile normalization 38 and two different ways of scaling. In the first, we scaled the data such that the median of the zebrafish control probes was one. In the second, we scaled the data such that the median of the trimmed intergenic region probes was one. Trimmed intergenic probes were defined as those >240 bp (4 probes) from any known annotation. We calculated probe level 'oligo' summaries by taking the log₂ of the mean of the two biological replica probes for each condition. Window level summaries were generated by identifying windows of 60 bp containing two or more oligo start positions and calculating a log₂ pseudo median on the associated values. This window summary score was assigned to the center position of the window "Pse" or represented as heat map "PseHM" data. Lastly, extended regions of high scoring windows, called "intervals," were identified by merging windows that exceed a set threshold and abut or overlap. We picked several different thresholds on the basis of the relative transcription above the trimmed intergenic or zebrafish control median value.

Dynamic difference maps.

To identify regions of change between different RNA fractions (total and poly(A)) and different growth conditions (rich, minimal, DNA damage and heat shock), we used a similar approach as that used in generating the static maps. The two biological replicas for each condition were quantile normalized and scaled such that the median of the zebrafish controls was one. We calculated probe level summaries by taking the log₂ ratio between the mean treatment and mean control. The control data was defined as either the RNA derived from the rich medium condition or the total RNA. We calculated window level summaries by identifying 60-bp windows that contain 2 or more oligos. These windows were scored by first calculating all the relative difference pairs between the treatment and the control replica probes, and second, by calculating the pseudo median of these relative difference pairs. In some cases, the pseudo median relative difference scores were converted to log₂ (ratios). Lastly, overlapping high- or low-scoring windows were merged into intervals and ranked by their best window score.

Strand-specific quantitative PCR analysis.

See Supplementary Methods.
http://www.nature.com/ng/journal/v40/n8/suppinfo/ng.196_S1.html

URLs:

TiMAT2 project site, http://sourceforge.net/projects/timat2; S. pombe annotation, ftp://ftp.sanger.ac.uk/pub/yeast/pombe/GFF (S.pombe.gff 3/16/07); S. pombe genome sequence, ftp://ftp.sanger.ac.uk/pub/yeast/pombe/Chromosome_contigs (chromosomeX.contig 4/23/07);

Supplementary datasets, http://bioserver.hci.utah.edu/SupplementalPaperInfo/2008/Dutrow_NatGen_PombeTranscriptome and http://bioserver.hci.utah.edu:8080/DAS2/das2; IGB, http://bioserver.hci.utah.edu/BioInfo/index.php/Software:IGB.

Accession codes.

NCBI GEO: S. pombe transcriptome microarray data have been deposited with accession number GSE11619.

Note: Supplementary information is available on the Nature Genetics website.
http://www.nature.com/ng/journal/v40/n8/suppinfo/ng.196_S1.html

Author contributions:

N.D., D.A.N., and B.R.C.; system design and experimental approaches. B.D.; array method optimization. D.A.N., N.D., B.M., T.J.P., D.H., and B.R.C.; array design, data analysis methods and data analysis. E.W.; feature computation. N.D., D.A.N., and D.H.; figures. B.R.C., N.D. and D.A.N. wrote the manuscript.

Acknowledgments:

We thank S. Leppla (US National Institutes of Health) for generously providing the S9.6 antibody and Bob Schackmann (University of Utah) for oligo synthesis. The work was supported by the Howard Hughes Medical Institute (B.R.C., D.H., T.J.P., and supplies), US National Institutes of Health Genetics Training Grant T32 GM007464 (N.D.), the Huntsman Cancer Institute (D.A.N.), and CA24014 (for core facilities).

Received 8 April 2008; Accepted 12 June 2008; Published online 20 July 2008.

References:

   1. Birney, E. et al. Identification and analysis of functional elements in 1% of the human genome by the ENCODE pilot project. Nature 447, 799–816 (2007).

   2. Okazaki, Y. et al. Analysis of the mouse transcriptome based on functional annotation of 60,770 full-length cDNAs. Nature 420, 563–573 (2002).

   3. David, L. et al. A high-resolution map of transcription in the yeast genome. Proc. Natl. Acad. Sci. USA 103, 5320–5325 (2006).

   4. Washietl, S., Hofacker, I.L., Lukasser, M., Huttenhofer, A. & Stadler, P.F. Mapping of conserved RNA secondary structures predicts thousands of functional noncoding RNAs in the human genome. Nat. Biotechnol. 23, 1383–1390 (2005).

   5. Kampa, D. et al. Novel RNAs identified from an in-depth analysis of the transcriptome of human chromosomes 21 and 22. Genome Res. 14, 331–342 (2004).

   6. Steigele, S., Huber, W., Stocsits, C., Stadler, P.F. & Nieselt, K. Comparative analysis of structured RNAs in S. cerevisiae indicates a multitude of different functions. BMC Biol. 5, 25 (2007).

   7. Struhl, K. Transcriptional noise and the fidelity of initiation by RNA polymerase II. Nat. Struct. Mol. Biol. 14, 103–105 (2007).

   8. Perocchi, F., Xu, Z., Clauder-Munster, S. & Steinmetz, L.M. Antisense artifacts in transcriptome microarray experiments are resolved by actinomycin D. Nucleic Acids Res. 35, e128 (2007).

   9. Hu, Z., Zhang, A., Storz, G., Gottesman, S. & Leppla, S.H. An antibody-based microarray assay for small RNA detection. Nucleic Acids Res. 34, e52 (2006).

  10. Boguslawski, S.J. et al. Characterization of monoclonal antibody to DNA.RNA and its application to immunodetection of hybrids. J. Immunol. Methods 89, 123–130 (1986).

  11. Kapranov, P., Willingham, A.T. & Gingeras, T.R. Genome-wide transcription and the implications for genomic organization. Nat. Rev. Genet. 8, 413–423 (2007).

  12. Kapranov, P. et al. RNA maps reveal new RNA classes and a possible function for pervasive transcription. Science 316, 1484–1488 (2007).

  13. Wood, V. et al. The genome sequence of Schizosaccharomyces pombe. Nature 415, 871–880 (2002).

  14. Leonardi, J., Box, J.A., Bunch, J.T. & Baumann, P. TER1, the RNA subunit of fission yeast telomerase. Nat. Struct. Mol. Biol. 15, 26–33 (2008).

  15. Webb, C.J. & Zakian, V.A. Identification and characterization of the Schizosaccharomyces pombe TER1 telomerase RNA. Nat. Struct. Mol. Biol. 15, 34–42 (2008).

  16. Gordon, M. et al. Genome-wide dynamics of SAPHIRE, an essential complex for gene activation and chromatin boundaries. Mol. Cell. Biol. 27, 4058–4069 (2007).

  17. Chen, D. et al. Global transcriptional responses of fission yeast to environmental stress. Mol. Biol. Cell 14, 214–229 (2003).

  18. Katayama, S. et al. Antisense transcription in the mammalian transcriptome. Science 309, 1564–1566 (2005).

  19. Nicolas, E. et al. Distinct roles of HDAC complexes in promoter silencing, antisense suppression and DNA damage protection. Nat. Struct. Mol. Biol. 14, 372–380 (2007).

  20. Wiren, M. et al. Genomewide analysis of nucleosome density histone acetylation and HDAC function in fission yeast. EMBO J. 24, 2906–2918 (2005).

  21. Noma, K., Cam, H.P., Maraia, R.J. & Grewal, S.I. A role for TFIIIC transcription factor complex in genome organization. Cell 125, 859–872 (2006).

  22. Volpe, T.A. et al. Regulation of heterochromatic silencing and histone H3 lysine-9 methylation by RNAi. Science 297, 1833–1837 (2002).

  23. Allshire, R.C., Javerzat, J.P., Redhead, N.J. & Cranston, G. Position effect variegation at fission yeast centromeres. Cell 76, 157–169 (1994).

  24. Takahashi, K. et al. A low copy number central sequence with strict symmetry and unusual chromatin structure in fission yeast centromere. Mol. Biol. Cell 3, 819–835 (1992).

  25. Scott, K.C., White, C.V. & Willard, H.F. An RNA polymerase III-dependent heterochromatin barrier at fission yeast centromere 1. PLoS ONE 2, e1099 (2007).

  26. Baum, M., Ngan, V.K. & Clarke, L. The centromeric K-type repeat and the central core are together sufficient to establish a functional Schizosaccharomyces pombe centromere. Mol. Biol. Cell 5, 747–761 (1994).

  27. Partridge, J.F., Scott, K.S., Bannister, A.J., Kouzarides, T. & Allshire, R.C. cis-acting DNA from fission yeast centromeres mediates histone H3 methylation and recruitment of silencing factors and cohesin to an ectopic site. Curr. Biol. 12, 1652–1660 (2002).

  28. Steiner, N.C., Hahnenberger, K.M. & Clarke, L. Centromeres of the fission yeast Schizosaccharomyces pombe are highly variable genetic loci. Mol. Cell. Biol. 13, 4578–4587 (1993).

  29. Clarke, L., Amstutz, H., Fishel, B. & Carbon, J. Analysis of centromeric DNA in the fission yeast Schizosaccharomyces pombe. Proc. Natl. Acad. Sci. USA 83, 8253–8257 (1986).

  30. Nakaseko, Y., Kinoshita, N. & Yanagida, M. A novel sequence common to the centromere regions of Schizosaccharomyces pombe chromosomes. Nucleic Acids Res. 15, 4705–4715 (1987).

  31. Nakaseko, Y., Adachi, Y., Funahashi, S., Niwa, O. & Yanagida, M. Chromosome walking shows a highly homologous repetitive sequence present in all the centromere regions of fission yeast. EMBO J. 5, 1011–1021 (1986).

  32. Motamedi, M.R. et al. Two RNAi complexes, RITS and RDRC, physically interact and localize to noncoding centromeric RNAs. Cell 119, 789–802 (2004).

  33. Lindsey-Boltz, L.A. & Sancar, A. RNA polymerase: the most specific damage recognition protein in cellular responses to DNA damage? Proc. Natl. Acad. Sci. USA 104, 13213–13214 (2007).

  34. Cam, H.P. et al. Comprehensive analysis of heterochromatin- and RNAi-mediated epigenetic control of the fission yeast genome. Nat. Genet. 37, 809–819 (2005).

  35. Verdel, A. et al. RNAi-mediated targeting of heterochromatin by the RITS complex. Science 303, 672–676 (2004).

  36. Reinhart, B.J. & Bartel, D.P. Small RNAs correspond to centromere heterochromatic repeats. Science 297, 1831 (2002).

  37. Kato, H. et al. RNA polymerase II is required for RNAi-dependent heterochromatin assembly. Science 309, 467–469 (2005).

  38. Bolstad, B.M., Irizarry, R.A., Astrand, M. & Speed, T.P. A comparison of normalization methods for high density oligonucleotide array data based on variance and bias. Bioinformatics 19, 185–193 (2003). 

This new technique of Natalie Dutrow , David Nix, Derick Holt , Brett Milash , Brian Dalley, Erick Westbroek , Timothy Parnell, and Bradley Cairns is designed for analyzing intact DNA-RNA hybrids at sites of active gene transcription, and offers both markedly higher resolution and the ability to distinguish each DNA strand at a particular gene locus. Most if not all of the DNA sequences of extended euchromatin appear to be undergoing active transcription to RNA. Protein-coding RNAs, non-coding RNAs, transfer RNAs, microRNAs, and both transcription start sites and and RNA splicing sites can be detected.  Some gene loci appear to produce paired sense-antisense RNA doublets, and on occasion these paired RNA-RNA  doublets involve coding RNAs and/or transfer RNAs. Telomere sites, centromere sites, and heterochromatin-transition sites can also be analyzed. This work has been done on yeast species, and is now being extended to more complex species.
 

Nature Genetics 40, 935 - 936 (August, 2008) doi:10.1038/ng0808-935
http://www.nature.com/ng/journal/v40/n8/abs/ng0808-935.html
http://www.nature.com/ng/journal/v40/n8/full/ng0808-935.html

"Mapping the strand-specific transcriptome of fission yeast",

Thomas R Gingeras

Thomas R. Gingeras is at the Department of Functional Genomics, Cold Spring Harbor Laboratory, Cold Spring Harbor, New York, 11724, USA.
      e-mail: tom_gingeras@affymetrix.com

Pervasive genome-wide transcription is widespread in eukaryotic cells, but key features of the transcriptome have yet to be fully characterized. A new study using antibody-based detection of RNA-DNA duplexes on tiling arrays now reveals a complex, strand-specific transcriptional world in fission yeast.

The biochemical evidence for pervasive genome-wide transcription has been well established for many organisms 1, 2. These and many other studies point to a transcriptional organization for many genomes that can be characterized as highly interleaved (Fig. 1) 3. However, at present, it is still unclear what the biological roles of these previously unannotated transcripts are. One step toward designing useful experiments is to investigate whether the unannotated transcribed regions share characteristics that would allow for hypotheses to be formulated and genetically tested. Dutrow et al., reporting on page 977 of this issue 4, provide such a dataset in their analysis of the fission yeast Schizosaccharomyces pombe transcriptome.

Figure 1: Three levels of resolution showing pervasive interleaved transcription.

Figure 1 : Three levels of resolution showing pervasive interleaved transcription.
The clustering of genes in the upper portion of the figure cloaks the overlapping transcription of protein-coding and noncoding transcripts observed within and between genic regions, as depicted in the middle panel. These transcripts also point to multiple regulatory regions (triangles and circles) that are positioned within genes and present on opposite strands. The ultimate fate of some of these long transcripts (for example, promoter-associated long RNAs; PALRs) is to provide short RNAs such as microRNAs, promoter-associated short RNAs (PASRs) and termini-associated short RNAs (TASRs).
(Gene Clusters: Purple), (Transcription Start Sites (TSS): Green.), (Primary Transcripts (PT): Blue).

Detecting duplexes

Consistent with data obtained from studies of human 5, mouse 6 and fly 7, the authors find that virtually all of the euchromatic genome is transcribed. However, two things of note distinguish this observation from other genome-wide transcription studies. The first involves the technical approach used by the authors to obtain their transcriptome maps. Specifically, the authors used an antibody (S9.6) raised against RNA-DNA duplexes to identify duplex regions formed on DNA probes that were part of whole-genome tiling arrays. This allowed for the strand-specific designation of each of the detected transcribed regions without the use of reverse transcriptase or RNA-DNA amplification techniques, which can have issues associated with the production of only single-stranded products. The results using this technical approach are notable in both their range of detection (7,600-fold) and their specificity (reduction in hybridization signal of 80- and 20,000-fold with one and two base mismatches, respectively). However, given that these results are dependent upon the use of an immunological approach, there is a concern as to whether these results will be reproducibly achieved by other laboratories using even slightly different labeling protocols, antibody preparations from the hybridoma cell line or detection array systems. Similar concerns have plagued chromatin immunoprecipitation studies. Nevertheless, the results achieved using this approach provide a fresh alternative to achieve strand-specific RNA maps using tiling arrays.

The second notable aspect of this study focuses on the characterization of previously unannotated transcribed genomic regions. One class of unannotated transcription in S. pombe noted by Dutrow et al. involves the detection of widespread antisense transcription. The authors indicate that the detection of antisense transcription in S. pombe was less prevalent when polyadenylated (poly(A)+) RNA was mapped compared to total RNA. The authors interpret this to be consistent with the possibility of the antisense transcripts being poly(A)- or possessing reduced length polyadenylation. Dutrow et al. suggest that the cause of this antisense transcription is opportunistic on the basis of the negative correlation with histone H3 occupancy and the positive correlation of coordinated expression of sense-antisense transcription during conditions of gene expression changes (for example, heat shock) and the presence of histone H3 lysine 36 trimethylation (H3K36me3). However, the conclusion that S. pombe antisense transcription is in large measure composed of poly(A)- RNA implies that such transcripts would likely have a rapid turnover rate and be relatively short-lived. If correct, this observation would represent a significant difference compared to that observed in mouse and human cells 3, 8, in which most of the detected antisense transcription is detected in poly(A)+ RNA samples and contains reasonable length polyadenylation, as revealed by cDNA sequencing. Direct empirical determination of the polyadenylation state of most antisense transcripts in S. pombe would be straightforward and would not only confirm these observations but also provide possible insights as to why such a marked contrast is observed between fission yeast and higher eukaryotic cells.

Opportunistic or deterministic?

Another characteristic of antisense transcripts observed by Dutrow et al. involves the previously uncharacterized transcribed regions flanking the tRNAs residing in the imr repeats found in the heterochromatic centromeric regions. The regions flanking the tRNA genes seem to be transcribed on the antisense strand relative to the tRNA gene. Such flanking antisense transcription is reported not to be observed at tRNA genes found in euchromatin regions. Again, the authors see this antisense transcription as the result of opportunistic conditions set up by the directed transcription of the tRNA genes. Extending this line of thought, the authors conclude that the genome-wide baseline transcription observed along intergenic regions is also opportunistic and indicative of the chromatin state of these regions.

Although initially attractive, this explanation places the role of transcription of a large portion of a genome as a passive and baseline condition in the cell. Such a promiscuous role for transcription is troubling for two reasons. First, as indicated by the authors themselves, the RNA detected in their studies reflects a steady state condition in the cells, and thus, these molecules are not likely to be short-lived. This is especially the case when the same transcribed regions are observed at multiple time points during development or in response to external stimuli, as seen in this study. These nontransient RNAs within cells are thus likely to be immediately associated with a diverse collection of RNA-binding proteins. The roles of these proteins are substantial but their abundance in a cell is not. An organizational strategy that uses opportunism on such a global scale greatly increases the requirement for regulatory complexity to discern the products of opportunism from determinism so as to judiciously use the limited RNA binding–protein resources of the cell and sets up conditions for creating transcripts that have the same regulatory signals as transcripts created in a deterministic fashion.

One of the results from the pilot ENCODE studies may point to a less opportunistic reason for the synthesis of such transcripts 1. An analysis of ENCODE regions aligned for 23 mammals and 5 other vertebrates showed significant enrichment for short islands of conservation within transcripts of unknown function (TUFs), despite absence of detectable conservation when the enrichment was averaged across the whole length of each transcribed region. Thus, long intergenic or antisense transcripts can be made in a directed and regulated fashion in order to provide short functional RNAs (for example, microRNA primary transcripts) or allow for the rapid evolution of sense-antisense transcript pairs (Fig. 1). It seems likely that additional studies will be undertaken involving transcripts originating from unannotated regions to determine whether stable short RNAs are also found mapping to these same regions and which are enriched in evolutionary conserved sequences, as has been observed in human cell lines 3.

Competing interests statement:
The author declares competing financial interests.
Declaration: The author is a consultant and former employee of Affymetrix, Inc.

References

   1. Birney, E. et al. Nature 447, 799–816 (2007). | Article |

   2. Gingeras, T.R. Genome Res. 17, 682–690 (2007). | Article |

   3. Kapranov, P. et al. Science 316, 1484–1488 (2007). | Article |

   4. Dutrow, N. et al. Nat. Genet. 40, 977–986 (2008). | Article |

   5. Efroni, S. et al. Cell Stem Cell 2, 437–447 (2008). | Article |

   6. Katayama, S. et al. Science 309, 1564–1566 (2005). | Article |

   7. Manak, J.R. et al. Nat. Genet. 38, 1151–1158 (2006). | Article |

   8. Carninci, P. et al. Science 309, 1559–1563 (2005). | Article |

1. Schwartz JC, Younger ST, Nguyen N-B, Hardy DB, Monia BP, Corey DR, and Janowski BA,
"Antisense transcripts are targets for activating small RNAs".

2. Frenster JH, and Hovsepian JA, "Models of  Embryonic RNA Initiating and Reverting Adult Neoplasms".

3. Ogawa Y, Sun BK, and Lee JT,
"Intersection of the RNA Interference and X-Inactivation Pathways".

4. Place RF, Li L-C, Pookot D, Noonan EJ, and Dahiya R, "MicroRNA-373 induces expression of genes with complementary promoter sequences".

5. Borel C, Gagnebin M, Gehrig C, Kriventseva EV, Zdobnov EM, and Antonarakis SE,
"Mapping of Small RNAs in the Human ENCODE Regions".

6. Zhu X, Ling J, Zhang L, Pi W, Wu M, and Tuan D, "A facilitated tracking and transcription mechanism of long-range enhancer function".

7. Frenster JH, and Hovsepian JA, "DNase-I Ultrastructural Probe Sites and Kissing Chromosomes".

8. Han J, Kim D, and Morris KV, "Promoter-associated RNA is required for RNA-directed transcriptional gene silencing in human cells".

Links to Euchromatin Activator RNA Reviews:
Links to Euchromatin Activator RNA Research:
Links to Ultrastructural Probes of DNase I-Sensitive Sites:
Links to RNA as a Therapeutic Agent:
Links to Hodgkin Lymphoma Immuno-Pathology:
Links to Activated T-Lymphocyte Immunotherapy:
Links to Medical Systems Biology:
Links to Selective Gene Transcription:
Links to RNA-Induced Epigenetics:
Links to RNA-Induced Embryogenesis:
Links to RNA and Biological Causality:
Links to Reprogramming and Neoplasia:

A Brief History of Activator RNA:

"Ultrastructural Probes of Active DNA Sites, and the RNA Activators of DNA". (PowerPoint Presentation).