RNA interference was first observed in petunias, when Napoli et al. (1990) [8] discovered that introduction of a pigment-producing gene under control of a powerful promoter suppressed expression of both the introduced gene and the homologous endogenous gene, a phenomenon they called “cosuppression.” Cosuppression was subsequently found to occur in many species of plants and fungi (where it was called “quelling”) and to occur at the post-transcriptional level [8–10].

Such post-transcriptional gene silencing (PTGS) was shown to be mediated by a diffusible, trans-acting molecule in both Neurospora and plants [9, 11]. The crucial 1998 discovery  by Fire et al., that injection of dsRNA—a mixture of both sense and antisense strands of the target mRNA, rather than either strand alone—into gonads of the nematode Caenorhabditis elegans resulted in extremely potent silencing unequivocably identified dsRNA as the inducer of RNA interference [12].

Short (~25 nt), antisense RNAs were first implicated in PTGS in plants [13]. Further work in Drosophila-derived in vitro systems showed that long dsRNA was processed to short interfering RNAs (siRNAs), 21–23 nt long, that could mediate target mRNA cleavage in the region corresponding to the introduced siRNA [14–16]. Current thinking about the mechanism of RNAi and components of the RNA-induced silencing complex is described more fully in the text.

In C. elegans, the demonstration that silencing could be induced by simply feeding nematodes with bacteria that had been engineered to express dsRNA homologous to the target gene, or by soaking worms in such dsRNA, drove rapid adoption of RNAi as a technology for reverse genetics at the organismal level [17–19]. Likewise, technology was quickly developed for inducing RNAi in Drosophila cell culture by bathing or transfection with dsRNA (see Armknecht et al. 2005 [20]).

Efforts to use RNAi in mammalian cells were hampered at first because of a nonspecific, interferon-mediated response to dsRNA longer than 30 bp that is seen in most mammalian cell lines. With the demonstration that 21–22 nt siRNAs, either chemically synthesized or expressed from a plasmid vector, could efficiently induce RNAi in mammalian cells without inducing the inteferon response, the door was opened for development of RNAi tools in mammalian systems [21, 22].

While some researchers were studying the phenomenon of RNAi and working to exploit its use as a tool for reverse genetics, others were trying to understand a distinct but related set of small RNAs—microRNAs (miRNAs). The first miRNAs to be discovered, lin-4 and let-7, were identified through loss-of-function mutations affecting control of postembryonic development in C. elegans [23, 24]. Other miRNAs were soon identified in C. elegans, Drosophila, and mouse by combinations of forward genetics, cDNA cloning, bioinformatics, and reverse genetics. By 2002, miRNAs were firmly established as a large class of conserved regulatory molecules in animals and plants [25–29]. miRNAs have been shown to play a role in developmental timing, cell death, cell proliferation, and oncogenesis [30–32]. This class of small RNA may represent 2–3% of the total number of genes in humans [4, 33], and estimates of miRNA target-binding sites indicate that miRNAs may play a role in regulating as many as 30% of mammalian genes [34].

A formal connection between the phenomenon of RNA interference and miRNAs was made when the same enzyme, Dicer, was shown to process both long dsRNA into siRNAs and cytoplasmic miRNA precursors (pre-miRNAs) into mature miRNAs [35–38]. Other components of the RNAi and miRNA silencing pathways have been shown to be closely related, most prominently Argonaute, which is a key component of the RNA-induced Silencing Complex (RISC). It now appears that miRNAs function in a silencing complex that is similar, if not identical, to RISC to regulate expression of target genes either through cleavage of mRNA or translational repression: if the miRNA exhibits perfect complementarity to its target mRNA, the mRNA is cleaved (typically in plants); if there is only partial complementary, translational repression occurs (typically in animals). The further discovery of endogenous small RNAs, distinct from miRNAs, that function in transcriptional silencing and genome stability has driven the adoption of the more general terms “RNA silencing” and “small RNAs” to describe the collection of related silencing pathways and their RNA guides. The term “RNA interference” (RNAi) continues to describe mRNA cleavage that is induced by a dsRNA trigger.

siRNAs and rasiRNAs
Short interfering RNAs (siRNAs) were originally identified as intermediates in the RNA interference pathway. siRNAs are 21–25 bp dsRNA with dinucleotide 3' overhangs that are formed in the cell from longer dsRNA molecules. The fully assembled RNA-induced silencing complex (RISC) contains only one strand of the siRNA, the guide strand. The guide strand is thought to provide target specificity for RISC-mediated cleavage through perfect base pairing with the mRNA target.

Long dsRNA (several hundred base pairs) can be used to artificially induce the RNAi pathway in fungi, plants, insects, and worms. In mammalian cells, dsRNA >30 bp induces a nonspecific, antiviral interferon response. However, synthetic siRNAs, ~21 bp with dinucleotide 3' overhangs, are potent triggers of RNAi in mammalian cells.

Although siRNAs are typically thought of as originating from exogenous precursors or triggers, it is perhaps not surprising that siRNAs generated from endogenous precursors have been identified. These endogenous siRNAs have been found to function in maintenance of chromatin structure and genome integrity as well as post-transcriptional gene silencing.

Endogenous siRNAs have been identified in plants, fungi, and animals. These siRNAs are derived in vivo from perfectly base-paired dsRNA precursors comprised of two distinct RNA strands. In many cases, endogenous siRNAs originate from repetitive elements within the genome, such as heterochromatic regions at centromeres and telomeres, and are therefore known as repeat-associated siRNAs (rasiRNAs). rasiRNAs appear to function, through an RNA-induced transcriptional silencing (RITS) complex, in maintaining the heterochromatic, and hence transcriptionally repressed, state of the region that encodes the rasiRNA. In Drosophila, rasiRNAs encoded by the Y-encoded Su(Ste) locus have also been implicated in trans regulation of the X-linked Stellate (Ste) locus, but it is not clear if silencing is at the transcriptional or post-transcriptional level [39]. In the protozoan Tetrahymena, endogenous siRNAs have also been shown to play a role in specifying sequences that are eliminated during macronucleus formation after sexual conjugation (Sontheimer and Carthew 2005 [4] and references within).

Endogenous nonrepetitive siRNAs have also been identified in the model plant Arabidopsis as well as in Caenorhabditis elegans [40, 41]. In some cases these endogenous siRNAs are complementary to longer, unique, coding RNAs. In Arabidopsis, nonrepetitive siRNAs have been implicated in regulation of transcript abundance through mRNA cleavage [41], indicating an RNAi mechanism similar to that triggered by exogenous siRNAs.

miRNAs
MicroRNAs (miRNAs) are a large class of evolutionarily conserved RNAs found in plants and animals. These small RNAs have been shown to play critical roles in developmental timing, hematopoietic cell differentiation, cell death, cell proliferation, and oncogenesis [30–32]. miRNAs may represent 2–3% of the total number of genes in humans [4, 33], and estimates of the number of miRNA target binding sites indicate that miRNAs may play a role in regulating as many as 30% of mammalian genes [34].

miRNAs are 19–23 nt single-stranded RNAs, originating from single-stranded precursor transcripts that are characterized by imperfectly base-paired hairpins. miRNAs function in a silencing complex that is similar, if not identical, to RISC.

In plants, miRNAs resemble siRNAs in that they exhibit perfect complementarity to binding sites in target mRNA and appear to silence gene expression through mRNA cleavage. Animal miRNAs typically exhibit partial complementarity to target binding sites in the 3' untranslated region (UTR) of target mRNA and repress translation without mRNA cleavage. It originally appeared that the process of miRNA gene silencing in animals involved reduced steady-state protein levels for the targeted gene without a reduction in the corresponding levels of mRNA; however, examples to the contrary have challenged this idea [4, 6]. miRNAs have also been implicated in mRNA turnover pathways in Drosophila and humans that use AU-rich elements (ARE), which are found in the 3' UTRs of many short-lived transcripts. miRNAs are also implicated in transcriptional silencing via hypermethylation in Arabidopsis [4].

Biogenesis of miRNAs and Endogenous siRNAs
miRNAs are synthesized in the nucleus as long (up to 1000 nt) RNA polymerase II transcripts, called pri-miRNAs, that are characterized by imperfect hairpin structures. Many pri-miRNAs arise from intergenic regions, or are in antisense orientation to known genes, indicating independent transcription units. Other genes for miRNAs are found in intronic regions and could be transcribed as part of the primary transcript for the corresponding gene [42]. A dsRNA-specific endonuclease, Drosha, in conjunction with a dsRNA-binding protein, called Pasha in Drosophila and DGCR8 in humans, processes the pri-miRNA into hairpin RNAs 70–100 nt in length, called pre-miRNAs. Pre-miRNAs are transported to the cytoplasm via an Exportin-5 dependent mechanism. There Dicer works with a dsRNA-binding partner, Loqs in Drosophila and TRBP in humans, to process the pre-miRNA into mature, single-stranded miRNA and load it into an RNA silencing complex that is similar to RISC. Loqs and TRBP are functionally analogous to the dsRNA-binding protein R2D2 which partners with Dicer to process and assemble siRNAs into RISC.

Endogenous dsRNA precursors for siRNAs could, in principle, arise from bidirectional transcription of genomic regions encoding siRNAs. However, in C. elegans, Arabidopsis, and fungi, silencing requires an RNA-dependent RNA polymerase (RdRP), raising the possibility that RdRP activity generates dsRNA from single-stranded RNA transcripts. Synthesis of dsRNA using the target mRNA itself as a template (using siRNA as a primer, or via de novo synthesis) could explain the phenomenon of “spreading” of RNAi, the production of siRNAs encoded by the target gene but not by the trigger RNA. Spreading is associated with “systemic silencing” seen in C. elegans and plants, in which silencing is inherited or is spread to distant parts of the organism. RdRP enzymes have not been found in Drosophila and mammals, and the spreading and systemic silencing phenomena are not seen in these organisms.

Many questions remain about endogenous siRNA production and function. How the cell distinguishes between normal mRNA and aberrant single-stranded RNA that is processed to dsRNA triggers for RNA silencing is unknown. Endogenous siRNA precursors are presumably transported to the cytoplasm before Dicer processing; the mechanism by which this happens is unknown, but is presumed to be similar if not identical to Exportin-5 mediated transport of pre-miRNAs. The biogenesis of endogenous siRNAs remains to be more clearly defined in all systems.