Revisiting Criteria for Plant MicroRNA Annotation in the Era of Big Data

Abstract

MicroRNAs (miRNAs) are ∼21-nucleotide-long regulatory RNAs that arise from endonucleolytic processing of hairpin precursors. Many function as essential posttranscriptional regulators of target mRNAs and long noncoding RNAs. Alongside miRNAs, plants also produce large numbers of short interfering RNAs (siRNAs), which are distinguished from miRNAs primarily by their biogenesis (typically processed from long double-stranded RNA instead of single-stranded hairpins) and functions (typically via roles in transcriptional regulation instead of posttranscriptional regulation). Next-generation DNA sequencing methods have yielded extensive data sets of plant small RNAs, resulting in many miRNA annotations. However, it has become clear that many miRNA annotations are questionable. The sheer number of endogenous siRNAs compared with miRNAs has been a major factor in the erroneous annotation of siRNAs as miRNAs. Here, we provide updated criteria for the confident annotation of plant miRNAs, suitable for the era of "big data" from DNA sequencing. The updated criteria emphasize replication and the minimization of false positives, and they require next-generation sequencing of small RNAs. We argue that improved annotation systems are needed for miRNAs and all other classes of plant small RNAs. Finally, to illustrate the complexities of miRNA and siRNA annotation, we review the evolution and functions of miRNAs and siRNAs in plants.

Figure 1.

Examples of Valid and Invalid MIRNA Loci. Small RNA-seq data were from GSE105262 (Polydore and Axtell, 2017). Left: Arabidopsis (ath) MIR399b, a locus judged valid under the new criteria. Right: ath-MIR405a, a locus judged invalid under the new criteria.

Figure 2.

Comparative Performance of miRNA Annotation Software from Plants. Data are adapted from Lei and Sun (2014). In that publication, two Arabidopsis sRNA-seq data sets were analyzed with seven different programs. miRNA loci that the programs found that were not in miRBase (version 20) were designated false positives. True positives: miRNA loci found that were also annotated in miRBase. False negatives: expressed, miRBase-annotated miRNA loci not found by the tool. True negatives: miRBase-annotated miRNA loci that were not expressed and thus not found by any of the tools.

Figure 3.

Conserved miRNA Families in Land Plants Based on miRBase 21 Annotations. Annotations were binned into one of eight major taxonomic groupings. Top bar charts illustrate counts of the number of species, number of miRNA families, and number of high-confidence miRNA families in each group. The central heat map shows annotations for “conserved” miRNA families, which are defined here as those annotated in at least two of the eight groups and which have at least one high-confidence annotation. The bottom cladogram illustrates the approximate divergence times (based on Magallón et al. [2013] and Wang et al. [2009]) of the eight groups. MYBP, million years before present.