A low-abundance class of Dicer-dependent siRNAs produced from a variety of features in C. elegans

Canonical small interfering RNAs (siRNAs) are processed from double-stranded RNA (dsRNA) by Dicer and associate with Argonautes to direct RNA silencing. In Caenorhabditis elegans, 22G-RNAs and 26G-RNAs are often referred to as siRNAs but display distinct characteristics. For example, 22G-RNAs do not originate from dsRNA and do not depend on Dicer, whereas 26G-RNAs require Dicer but derive from an atypical RNA duplex and are produced exclusively antisense to their messenger RNA (mRNA) templates. To identify canonical siRNAs in C. elegans, we first characterized the siRNAs produced via the exogenous RNA interference (RNAi) pathway. During RNAi, dsRNA is processed into ∼23 nt duplexes with ∼2 nt, 3'-overhangs, ultimately yielding siRNAs devoid of 5'G-containing sequences that bind with high affinity to the Argonaute RDE-1, but also to the microRNA (miRNA) pathway Argonaute, ALG-1. Using these characteristics, we searched for their endogenous counterparts and identified thousands of endogenous loci representing dozens of unique elements that give rise to mostly low to moderate levels of siRNAs, called 23H-RNAs. These loci include repetitive elements, putative coding genes, pseudogenes, noncoding RNAs, and unannotated features, many of which adopt hairpin (hp) structures reminiscent of the hpRNA/RNAi pathway in flies and mice. RDE-1 competes with other Argonautes for binding to 23H-RNAs. When RDE-1 is depleted, these siRNAs are enriched in ALG-1 and ALG-2 complexes. Our results expand the known repertoire of C. elegans small RNAs and their Argonaute interactors, and demonstrate that key features of the endogenous siRNA pathway are relatively unchanged in animals.