Science Wiki

RNA Interference (RNAi), also known historically in plants and fungi as post-transcriptional gene silencing (PTGS) or quelling, is a sequence-specific, evolutionarily conserved cellular mechanism in eukaryotes by which short non-coding RNA molecules—typically 20 to 30 nucleotides in length—inhibit gene expression by directing the endonucleolytic cleavage, destabilization, or translational repression of complementary target messenger RNA (mRNA) transcripts, or by nucleating repressive chromatin modifications in the cell nucleus. Elucidated mechanistically in 1998 by Andrew Fire and Craig Mello in the nematode *Caenorhabditis elegans* (work recognized with the 2006 Nobel Prize in Physiology or Medicine), RNA interference evolved in early eukaryotes both as an innate intracellular immune system against double-stranded RNA (dsRNA) viruses and transposable elements and as a master developmental network that post-transcriptionally fine-tunes over 60% of all human protein-coding genes.

The canonical trigger of RNA interference is the presence of double-stranded RNA (dsRNA) or imperfectly paired stem-loop hairpin RNA precursors inside the cell. Eukaryotic cells express three major classes of small silencing RNAs distinguished by their biogenesis, precursor structure, and Argonaute partner:

The MicroRNA (miRNA) Biogenesis Pathway

Endogenous microRNAs (miRNAs)—first discovered by Victor Ambros, Rosalind Lee, and Gary Ruvkun in 1993 (*lin-4* and *let-7* in *C. elegans*, awarded the 2024 Nobel Prize in Physiology or Medicine)—are transcribed in the nucleus primarily by RNA Polymerase II (either from dedicated non-coding miRNA genes or from the spliced introns of host protein-coding genes, termed "mirtrons") as long capped and polyadenylated primary transcripts called pri-miRNAs:

1. Nuclear Cropping by the Microprocessor Complex: Inside the nucleus, the RNase III endonuclease Drosha, complexed with two molecules of its double-stranded RNA-binding cofactor DGCR8 (Pasha), recognizes the basal junction of the pri-miRNA stem-loop (~11 bp from the single-stranded/double-stranded RNA junction) and cleaves both strands to release a ~60–70 nucleotide hairpin intermediate called the precursor miRNA (pre-miRNA) bearing a characteristic 2-nucleotide 3' overhang.
2. Nuclear Export: The nuclear export receptor Exportin-5, in a complex with Ran-GTP, specifically recognizes the 2-nt 3' overhang and double-stranded stem of the pre-miRNA and transports it through the nuclear pore complex into the cytoplasm.
3. Cytoplasmic Dicing: In the cytoplasm, the pre-miRNA (or a long exogenous dsRNA in the siRNA pathway) is bound by Dicer, a ~220 kDa multidomain RNase III endonuclease Complexed with TRBP (or PACT).

Dicer as a Molecular Ruler

Crystallographic and cryo-EM structures invented by Jennifer Doudna and colleagues revealed that Dicer functions as an atomic molecular ruler:

• Dicer's PAZ (Piwi–Argonaute–Zwille) domain possesses a specialized hydrophobic pocket that binds the 2-nucleotide 3' overhang and 5'-monophosphate at the open end of the dsRNA or pre-miRNA duplex.
• Separated from the PAZ domain by a rigid, positively charged flat protein platform measuring exactly ~65 Å in length (corresponding to the helical pitch of 21 to 22 base pairs of A-form dsRNA) sit two intramolecular RNase III catalytic domains (RNase IIIa and RNase IIIb).
• Each RNase III domain coordinates two catalytic Mg²⁺ ions and cleaves one phosphodiester backbone strand of the RNA duplex staggered by two base pairs, excising the terminal loop and liberating a mature 21–23 nucleotide RNA duplex possessing 5'-monophosphate (5'-PO₄²⁻) groups and 2-nucleotide 3'-hydroxyl (3'-OH) overhangs at both ends.

A naked 21-nt siRNA or miRNA duplex cannot silence genes on its own; it must be loaded into a multiprotein effector machine called the RNA-Induced Silencing Complex (RISC), whose catalytic core is a member of the Argonaute (AGO) protein superfamily (in humans, AGO1, AGO2, AGO3, and AGO4).

Asymmetric Guide Strand Selection

Using the Hsp70/Hsp90 chaperone machinery and ATP hydrolysis, the double-stranded 21-nt RNA duplex is loaded into the central cleft of an Argonaute protein. Only one of the two strands—termed the guide strand (antisense to the target mRNA)—is retained to form active holo-RISC, while the opposite passenger strand (sense strand) is cleaved or unwound and degraded.

How does RISC decide which of the two strands becomes the guide strand? Strand selection is governed by a strict thermodynamic asymmetry rule: the protein sensors inspect the base-pairing stability (ΔG°) of the first 1 to 4 base pairs at both ends of the 21-nt duplex. The strand whose 5' end lies at the thermodynamically less stable (more AU-rich or mismatched, easier-to-unwind) end of the duplex is preferentially selected as the guide strand, anchored tightly into the MID (Middle) domain of Argonaute via its 5'-monophosphate group, while the 3' end is held by the PAZ domain.

Slicing (AGO2 PIWI Domain) vs. Seed-Mediated Deadenylation

Once the passenger strand is ejected, the single-stranded guide RNA sits pre-organized inside Argonaute in a helical A-form conformation along nucleotides 2 through 8 from its 5' end—the critical seed region—reducing the entropic penalty of scanning and hybridizing with cytoplasmic mRNAs. The post-transcriptional fate of the target mRNA depends on the degree of base-pairing complementarity between the guide RNA and the target mRNA:

1. Perfect Complementarity → Endonucleolytic Slicing (siRNA Mode): When a guide strand (especially loaded into human AGO2, the primary catalytically competent "Slicer" Argonaute in mammals, or plant AGO1) base-pairs with perfect or near-perfect complementarity across the central region (nucleotides 9–12) of the target mRNA, the target phosphodiester bond located directly opposite guide nucleotides 10 and 11 is positioned into the PIWI domain of AGO2. The PIWI domain adopts an RNase H-like fold containing a catalytic DEDH (Asp-Glu-Asp-His) tetrad that coordinates two Mg²⁺ ions. AGO2 slices the target mRNA backbone in a single cleavage event, leaving a 5' fragment (lacking a poly(A) tail, rapidly degraded 3' → 5' by the cytoplasmic exosome) and a 3' fragment (lacking a 5' 7-methylguanosine cap, rapidly degraded 5' → 3' by the exoribonuclease XRN1). The intact RISC complex then recycles catalytically to cleave dozens of additional mRNA copies.
2. Partial "Seed" Complementarity → Translational Repression and Deadenylation (Animal miRNA Mode): In animals, endogenous miRNAs rarely share full complementarity with their targets; instead, they base-pair perfectly along the seed region (nucleotides 2–8)—typically within the 3' untranslated region (3' UTR) of target mRNAs—while exhibiting central bulges at positions 9–12 that physically prevent AGO2 slicing. Instead, Argonaute recruits the scaffolding protein TNRC6 (GW182), which in turn recruits the PAN2–PAN3 and CCR4–NOT deadenylase complexes to strip off the mRNA's 3' poly(A) tail, inhibits eIF4E/eIF4G cap-dependent translation initiation at the ribosome, and recruits the DCP1–DCP2 decapping complex to trigger 5' → 3' mRNA decay inside cytoplasmic P-bodies.

In plants, fungi, nematodes (*C. elegans*), and insects (*Drosophila*), RNAi serves as the primary adaptive antiviral immune system: viral replicative dsRNA intermediates are diced into 21-nt viral siRNAs (vsiRNAs) that destroy viral transcripts. Furthermore, plants, fungi, and *C. elegans* possess host RNA-dependent RNA polymerases (RdRPs) that use primary siRNAs as primers on target mRNAs to synthesize secondary dsRNAs, amplifying the silencing signal and spreading it systemically throughout the organism (via plasmodesmata in plants or the SID-1 transmembrane dsRNA channel in *C. elegans*).

Mammals and other vertebrates lost cellular RdRP genes during evolution and evolved the protein-based Type I Interferon (IFN) innate immune system: introducing long dsRNA (>30 base pairs) into a mammalian somatic cell triggers cytosolic pattern-recognition receptors (PKR, RIG-I, MDA5, and OAS/RNase L), which phosphorylate eIF2α to shut down all cellular protein synthesis and trigger apoptosis. In 2001, Thomas Tuschl and colleagues made the breakthrough discovery that chemically synthesized 21-nucleotide siRNA duplexes with 2-nt 3' overhangs completely bypass the mammalian PKR/interferon response while directly entering endogenous Dicer/AGO2 RISC to silence any chosen human gene.

Today, synthetic RNAi has become a major clinical pharmacology modality. By chemically stabilizing siRNAs (replacing 2'-OH groups with 2'-O-methyl and 2'-fluoro modifications and adding phosphorothioate backbone linkages to resist serum RNases) and conjugating them to trivalent N-acetylgalactosamine (GalNAc) ligands that bind the hepatocyte Asialoglycoprotein Receptor (ASGPR), FDA-approved siRNA drugs—beginning with patisiran (2018, for hereditary transthyretin amyloidosis) and inclisiran (targeting *PCSK9* mRNA to lower LDL cholesterol with just two injections per year)—achieve durable, specific gene knockdown in human patients.

Why did introducing long dsRNA into mammalian cells initially fail to silence single genes?

In nematodes, insects, and plants, long dsRNA (300–800 bp) is cleanly diced into specific 21-nt siRNAs without toxicity. In mammalian somatic cells, however, dsRNA longer than 30 base pairs is recognized as a viral infection by Protein Kinase R (PKR) and 2'-5'-oligoadenylate synthetase (OAS), triggering a global interferon response, eIF2α phosphorylation, complete shutdown of all translation, and cell death. Using short 21–23 nt synthetic siRNAs bypasses PKR while engaging AGO2 directly.

How do microRNAs (miRNAs) differ from small interfering RNAs (siRNAs)?

Although both are 21–23 nt small RNAs processed by Dicer and loaded into Argonaute, siRNAs derive from long, perfectly base-paired dsRNA, require 100% sequence complementarity to slice a single specific target mRNA via AGO2, and act as precise molecular scalpels. Conversely, miRNAs are transcribed from endogenous genome hairpin loci, processed by Drosha and Dicer, and base-pair imperfectly via a 7-nucleotide seed region (nt 2–8), allowing a single miRNA to translationally repress and deadenylate hundreds of different cellular mRNAs simultaneously.

How does RNAi differ from CRISPR-Cas9 gene editing?

CRISPR-Cas9 operates in the cell nucleus at the DNA level, introducing permanent, heritable double-strand breaks to knock out or edit a gene sequence in the chromosome (gene knockout). RNA interference (RNAi) operates primarily in the cytoplasm at the mRNA level, catalytically destroying transcripts after transcription while leaving the underlying chromosomal DNA untouched—producing a reversible, dose-dependent gene knockdown ideal for drugging essential genes or transient liver targets.