Abstract
Artificial microRNA (amiRNA) technology has allowed researchers to direct efficient silencing of specific transcripts using as few as 21 nucleotides (nt). However, not all the artificially designed amiRNA constructs result in selection of the intended ~21-nt guide strand amiRNA. Selection of the miRNA guide strand from the mature miRNA duplex has been studied in detail in human and insect systems, but not so much for plants. Here, we compared a nuclear-replicating DNA viral vector (tomato mottle virus, ToMoV, based), a cytoplasmic-replicating RNA viral vector (tobacco mosaic virus, TMV, based), and a non-viral binary vector to express amiRNAs in plants. We then used deep sequencing and mutational analysis and show that when the structural factors caused by base mismatches in the mature amiRNA duplex were excluded, the nucleotide composition of the mature amiRNA region determined the guide strand selection. We found that the strand with excess purines was preferentially selected as the guide strand and the artificial miRNAs that had no mismatches in the amiRNA duplex were predominantly loaded into AGO2 instead of loading into AGO1 like the majority of the plant endogenous miRNAs. By performing assays for target effects, we also showed that only when the intended strand was selected as the guide strand and showed AGO loading, the amiRNA could provide the expected RNAi effects. Thus, by removing mismatches in the mature amiRNA duplex and designing the intended guide strand to contain excess purines provide better control of the guide strand selection of amiRNAs for functional RNAi effects.