Abstract
RNA molecules carry information in their primary sequence and also their secondarynstructure. Secondary structure can confer important functional information, but it is also a signalnfor an RNAi-like host epigenetic response mediated by small RNAs (smRNAs). In this study, wenused two bioinformatic methods to predict local secondary structures across features of the maize genome, focusing on small regions that had similar folding properties to pre-miRNA loci. We found miRNA-like secondary structures to be common in genes and most, but not all, superfamilies of RNA and DNA transposable elements (TEs). The miRNA-like regions mapped a higher diversity of smRNAs than regions without miRNA-like structure, explaining up to 27% of variation in smRNA mapping for some TE superfamilies. This mapping bias was more pronounced among putatively autonomous TEs relative to non-autonomous TEs. Genome-wide, miRNA-like regions were also associated with elevated methylation levels, particularly in the CHH context. Among genes, those with miRNA-like secondary structure were 1.5-fold more highly expressed, on average, than other genes. However, these genes were also more variably expressed across the 26 Nested Association Mapping founder lines, and this variability positively correlated with the number of mapping smRNAs. We conclude that local miRNA-like structures are a nearly ubiquitous feature of expressed regions of the maize genome, that they correlate with higher smRNA mapping and methylation, and that they may represent a trade-off between functional need and the potentially negative consequences of smRNA production.