Werner BMC Biology 2013, 11:31 http://www.biomedcentral.com/1741-7007/11/31 COMMENTARY Open Access Biological functions of natural antisense transcripts Andreas Werner Abstract The most common meaning of the term 'antisense tran- script' refers to a protein coding sense transcript and a In theory, the human genome is large enough to fully processed (capped, polyadenylated) antisense RNA keep its roughly 20,000 genes well separated. In with complementarity in exonic regions (Figure 1). The practice, genes are clustered; even more puzzling, in key issue of whether the antisense transcripts act as ex- many cases both DNA strands of a protein coding quisitely specific gene regulators or are simply transcrip- gene are transcribed. The resulting natural antisense tional waste that a cell has learned to live with is still a transcripts can be a blessing and curse, as many matter of controversy [2]. However, a study published in appreciate, or simply transcriptional trash, as others BMC Genomics reports conservation of natural antisense believe. Widespread evolutionary conservation, as transcripts at a large scale between human, rat and mouse, recently demonstrated, is a good indicator for which strongly suggests that there is biological sense to potential biological functions of natural antisense having antisense transcripts [3]. transcripts. See research article: http://www.biomedcentral.com/ Regulatory antisense transcripts 1471-2164/14/243 The most convincing way to demonstrate the biologi- cal significance of an antisense transcript is to interfere with its expression and demonstrate phenotypic conse- By the mid-1980s antisense transcription in mammalian quences or altered expression levels of the sense tran- genomes had already been described by a few isolated script. This approach has been pursued to investigate a reports. Despite the recognized regulatory potential of still limited but increasing number of antisense tran- complementary RNA, antisense transcription in mam- scripts. The best characterized examples include Airn, mals was long perceived as a biological oddity. This Kcnq1ot1 and Tsix, antisense transcripts involved in par- perception started to change at the beginning of the gen- ental imprinting (Airn, Kcnq1ot1) and X chromosome omic era when pioneering data mining efforts and the inactivation (Tsix). Expression of Airn, Kcnq1ot1 and sequencing of large cDNA datasets revealed significant Tsix induces allele-specific chromatin changes and numbers of antisense transcripts. It is now accepted that eventually leads to the silencing of the cognate sense a significant proportion of genomic loci in mammalian transcript. genomes - for example, 40% in human and 72% in Parental imprinting and X chromosome inactivation are mouse - are transcribed in both directions [1]. The dis- epigenetic phenomena prominently observed in mammals crepancy between human and mouse is more a reflec- but absent or mechanistically distinct in other animals tion of the fact that antisense transcription has been such as birds or insects [4]. In human and mouse, studied in much greater detail in mouse than in human however, there are well documented examples of anti- rather than a 'real' biological difference. sense-induced chromatin modifications, suggesting that Sense and antisense transcript pairs come in various antisense RNA-guided gene silencing could well have forms depending on the exact location of the comple- broader significance. For example, a rare form of inherited mentary overlap and the processing of the transcripts. α-thalassemia is caused by ectopic expression of an an- tisense transcript. It originates from the constitutively ac- tive LUC7 gene, which is brought into close vicinity of Correspondence: [email protected] HBA2 LUC7 RNA Interest Group, Institute for Cell and Molecular Biosciences, Newcastle the gene by a gene deletion. The transcript University, Framlington Place, Newcastle upon Tyne, NE2 4HH, UK (antisense to HBA2) triggers the methylation of the HBA2 © 2013 Werner; licensee BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. Werner BMC Biology 2013, 11:31 Page 2 of 3 http://www.biomedcentral.com/1741-7007/11/31 Interesting examples include the bi-directionally tran- Antisense Convergent α sense/antisense scribed genes encoding HIF-1 (hypoxia-induced factor α β Sense transcripts 1 )and -secretase, where the antisense transcripts have a stabilizing effect on the protein coding sense transcript. Antisense Divergent This is achieved by either blocking an RNA destabilizing sense/antisense motif in HIF-1α mRNA or competing for a microRNA site Sense transcripts in β-secretase mRNA [7,8]. Such RNA-masking could well be of general significance since many sense-antisense pairs Antisense Non over-lapping overlap at their 3’ ends, where both stability motifs and sense/antisense Sense transcripts microRNA binding sites are predominantly situated. An unresolved problem regarding stoichiometric in- Figure 1. Schematic arrangements of transcripts from bi- teractions between the complementary transcripts is directionally transcribed genes (sense exons are in red, antisense exons are in blue). represented by the observation that antisense RNAs are generally expressed at much lower levels than protein coding transcripts. Alternatively, sense/antisense hybrids promoter, the silencing of the protein coding sense gene could be processed by the RNA interference pathway and, therefore, a reduction of α-hemoglobin [5]. Another and lead to transcriptional or post-transcriptional effects. antisense transcript has been linked in mouse to hete- As attractive as such a scenario would appear, there is lit- rochromatin formation and concomitant silencing of the tle evidence for endogenous small interfering RNAs de- tumor suppressor gene p15, associated with a variety of rived from complementary sense/antisense transcripts cancers [6]. (genic endo-siRNAs). Large scale sequencing experiments The fact that sense and antisense transcripts share com- have revealed some genic endo-siRNAs and also a specific plementary regions, as well as extensive indirect evidence gene, Slc34a1 (encoding a sodium/phosphate transport that sense and antisense transcripts may be co-expressed system), has been shown to produce them in selected tis- in the same cell, suggests the formation of sense-antisense sues [9]. However, a link between RNA interference and RNA hybrids. Such an assumption comes, however, with a antisense transcription seems to represent an exception in considerable reservation: double-stranded RNA molecules somatic cells and only to occur in restricted cell popula- in the cytoplasm are a signal of viral infection that acti- tions (Figure 2). vates protein kinase R and the interferon pathway, eventu- To summarize, a growing number of antisense tran- ally triggering an immune response. How this might be scripts with an established function have been identified avoided is unclear, but despite this concern, a growing but still many questions remain concerning their mecha- number of antisense transcripts have been shown to re- nisms of action. Progress is rather slow because of for- gulate the expression of their cognate sense transcripts midable experimental difficulties inherent to antisense through mechanisms based on RNA-RNA interaction. research. To start, most of the antisense transcripts are Antisense Chromatin Sense modifications CH3 DNA methylation CpG Histone modifications RNA masking miRNA RNase Antisense Protection from degradation Sense Dicer Endo-siRNAs Antisense RNA interference, Sense RNA interference gene silencing Figure 2. Established cellular mechanisms related to the transcription of natural antisense transcripts. Top panel: over-expression of an antisense transcript (in blue) causes the modification and concomitant silencing of the sense promoter. The exact mechanism of how the repressive chromatin marks are established is yet unknown. Middle panel: in RNA masking, the antisense transcript directly interacts with the sense transcript (in red) and occludes regulatory sequences. Bottom panel: in RNA interference, sense and antisense transcripts hybridize and the double-stranded RNA is further processed by the RNA interference-linked enzymatic machinery. This may lead to post-transcriptional or transcriptional gene silencing. endo-siRNA, endogenous small interfering RNA; miRNA, microRNA. Werner BMC Biology 2013, 11:31 Page 3 of 3 http://www.biomedcentral.com/1741-7007/11/31 expressed at very low levels and are difficult to detect transposons are thought to add a creative touch to learn- on northern blots or by in situ hybridization. Amplifica- ing, and in testis they help an organism to quickly adapt tion then carries the risk of losing the orientation and to a changing environment. Retrotransposition, however, swapping sense for antisense. More significantly, the also comes with the danger of randomly damaging vital intricate relation between sense and antisense transcrip- genes. An admittedly speculative but highly intriguing tion means that experimental perturbation of one tran- hypothesis suggests that natural antisense transcripts script inevitably interferes with the expression of the may quality control the transcriptional output after an other. Whether documented changes in expression