Localizing hotspots of antisense transcription.

Analysis of the transcriptome by computational and experimental methods has established that sense-antisense transcriptional units are a common phenomenon. Although the regulatory potential of antisense transcripts has been experimentally verified in a number of studies, the biological importance of sense-antisense regulation of gene expression is still a matter of debate. Here, we report the identification of sequence features that are associated with antisense transcription. We show that the sequence composition of the first exon and the 5'end of the first intron of many human genes is similar to the sequence composition observed in promoter regions as measured by the density of known transcription regulatory motifs. Cloned intron-derived fragments were found to possess bidirectional promoter activity. In agreement with the reported abundance of antisense transcripts overlapping the 5'UTR, mapping of the 5'ends of antisense transcripts to the corresponding sense transcripts revealed that the first exon and the 5'end of the first intron are hotspots of antisense transcription as measured by the number of antisense transcription start sites per unit sequence. CpG dinucleotide suppression that is typically weak in non-methylated promoter regions is similarly weakened upstream as well as downstream of the first exon. In support of antisense transcripts playing a regulatory role, we find that 5'UTRs and first exons of genes with overlapping antisense transcripts are significantly longer than the genomic average. Interestingly, a similar size distribution of 5'UTRs and first exons is observed for genes silenced by CpG island methylation in human cancer.