Genome-wide analysis of plant nat-siRNAs reveals insights into their distribution, biogenesis and function

Background: Many eukaryotic genomes encode cis-natural antisense transcripts (cis-NATs). Sense and antisense transcripts may form double-stranded RNAs that are processed by the RNA interference machinery into small interfering RNAs (siRNAs). A few so-called nat-siRNAs have been reported in plants, mammals, Drosophila, and yeasts. However, many questions remain regarding the features and biogenesis of nat-siRNAs. Results: Through deep sequencing, we identified more than 17,000 unique siRNAs corresponding to cis-NATs from biotic and abiotic stress-challenged Arabidopsis thaliana and 56,000 from abiotic stress-treated rice. These siRNAs were enriched in the overlapping regions of NATs and exhibited either site-specific or distributed patterns, often with strand bias. Out of 1,439 and 767 cis-NAT pairs identified in Arabidopsis and rice, respectively, 84 and 119 could generate at least 10 siRNAs per million reads from the overlapping regions. Among them, 16 cis-NAT pairs from Arabidopsis and 34 from rice gave rise to nat-siRNAs exclusively in the overlap regions. Genetic analysis showed that the overlapping double-stranded RNAs could be processed by Dicer-like 1 (DCL1) and/or DCL3. The DCL3-dependent nat-siRNAs were also dependent on RNA-dependent RNA polymerase 2 (RDR2) and plant-specific RNA polymerase IV (PoIIV), whereas only a fraction of DCL1-dependent nat-siRNAs was RDR- and PoIIV-dependent. Furthermore, the levels of some nat-siRNAs were regulated by specific biotic or abiotic stress conditions in Arabidopsis and rice. Conclusions: Our results suggest that nat-siRNAs display distinct distribution patterns and are generated by DCL1 and/or DCL3. Our analysis further supported the existence of nat-siRNAs in plants and advanced our understanding of their characteristics.