Molecular Biology

RNA Interference

Double-stranded RNA → Dicer → Argonaute → silenced mRNA

RNA interference (RNAi) is a conserved gene-silencing pathway that destroys messenger RNA matched by short guide RNAs. Long double-stranded RNA is chopped by the RNase III enzyme Dicer into 21–23 nucleotide small interfering RNAs (siRNAs); one strand of each duplex loads into the Argonaute protein at the heart of the RNA-induced silencing complex (RISC). RISC scans cytoplasmic mRNAs, finds those with sequence complementary to its loaded guide, and slices them. Fire and Mello showed in 1998 that injecting dsRNA into C. elegans silences the matching gene — they shared the 2006 Nobel Prize, only eight years later. Patisiran (Onpattro, 2018) was the first FDA-approved siRNA drug.

  • DiscoverersFire & Mello (1998, Nobel 2006)
  • Dicer cuts to21–23 nt siRNA duplexes
  • EffectorArgonaute (AGO2 in mammals)
  • Cleavage siteBetween guide nt 10 and 11
  • First siRNA drugPatisiran, 2018 (TTR amyloidosis)
  • Conserved acrossPlants, fungi, invertebrates, mammals

Interactive visualization

Press play, or step through manually. The visualization is yours to drive — try it before reading on.

Open visualization fullscreen ↗

Watch the 60-second explainer

A condensed visual walkthrough — narrated, captioned, under a minute.

The RNAi pathway, step by step

  1. Trigger. Long dsRNA enters the cytoplasm — from viral replication intermediates, transposon read-through, or experimentally injected hairpins.
  2. Dicing. Dicer (RNase III, with two active sites) measures from the 3' end and cuts every ~22 nt, producing siRNA duplexes with 2-nucleotide 3' overhangs and 5' phosphates.
  3. RISC loading. The duplex is loaded into AGO2 with help from Dicer and TRBP. The 5' end of one strand sits in the MID pocket; thermodynamic asymmetry — the strand with the less stable 5' end — is favored as the guide.
  4. Passenger ejection. AGO2 nicks the passenger strand at position 10–11, then unwinds and discards it. Mature RISC contains AGO + guide only.
  5. Target search. The seed region (guide nt 2–8) is pre-organized in an A-form helix and scans 3' UTRs and ORFs by transient base-pairing.
  6. Slicing. When complementarity extends through the central region, the PIWI fold cleaves the target between nt 10 and 11 of the guide. The fragments are degraded by XRN1 and the exosome.
  7. Recycling. AGO releases cleaved fragments and finds another transcript — catalytic, not stoichiometric.

Pathway diagram

  long dsRNA  (viral / hairpin / shRNA)
       │
       ▼  Dicer (RNase III)
       │
       ▼  21–23 nt siRNA duplex  (2-nt 3' overhangs)
       │
       ▼  RISC loading complex (AGO2 + Dicer + TRBP)
       │
       ▼  passenger strand sliced & ejected
       │
       ▼  mature RISC  (AGO2 + guide)
       │
       ▼  target mRNA scan  (seed nt 2–8 first)
       │
       ▼  perfect complementarity → cleave between guide nt 10 / 11
       │
       ▼  XRN1 + exosome degrade fragments

siRNA vs miRNA vs piRNA — three small-RNA pathways

siRNAmiRNApiRNA
Length21–23 nt21–23 nt24–32 nt
SourceLong dsRNA (viral, transgene, shRNA)Genomic hairpin transcriptsSingle-stranded transcripts from piRNA clusters
ProcessingDicer onlyDrosha (nucleus) → Dicer (cytoplasm)Dicer-independent; ping-pong amplification
Effector proteinAGO2 (Argonaute clade)AGO1–4 (Argonaute clade)PIWI clade (PIWIL1–4 in mammals)
Complementarity to targetFull (perfect match)Partial (seed match in 3' UTR)Mostly perfect, against transposons
MechanismSlice (endonucleolytic)Translational repression + mRNA decaySlice + transcriptional silencing via H3K9me
Tissue distributionMost cells (low endogenous)All tissues, ~2,500 in humansGerm line and embryonic stem cells
Primary biological roleAntiviral defense (plants, invertebrates)Fine-tuning gene expressionTransposon suppression, genome integrity

Real-world applications

  • Patisiran (Onpattro, 2018). First FDA-approved siRNA drug. Lipid nanoparticles deliver siRNA against transthyretin (TTR) to hepatocytes, treating hereditary ATTR amyloidosis. The 2018 approval validated 20 years of RNAi drug development.
  • GalNAc-siRNAs. Triantennary N-acetylgalactosamine binds the asialoglycoprotein receptor (~500,000/hepatocyte). Subcutaneous, no nanoparticle needed. Givosiran (acute hepatic porphyria, 2019), lumasiran (primary hyperoxaluria, 2020), inclisiran (LDL-C lowering, 2020), vutrisiran (ATTR, 2022), nedosiran (hyperoxaluria, 2023). Inclisiran dosed twice yearly halves LDL on top of statins.
  • Functional genomics. Genome-wide siRNA libraries (Dharmacon, Sigma MISSION) enable unbiased loss-of-function screens. CRISPR has overtaken much of this, but RNAi still wins where partial knockdown is more informative than total knockout.
  • Plant antiviral defense. Plants use RNAi as their primary innate antiviral system. Engineered crops express dsRNA against pest insects (Western corn rootworm; Smartstax Pro, 2017) — the first commercial RNAi-based pesticide.
  • Research workhorse. Transient siRNA transfection knocks down a target in 24–48 hours, far faster than CRISPR knockout. Standard for confirming on-target activity of any candidate gene.

Variants and related pathways

  • shRNA (short hairpin RNA). A hairpin transcript transcribed from a U6 or H1 promoter, processed by Dicer into siRNA — gives stable knockdown via lentiviral integration.
  • miRNA-mimic and antagomir. Synthetic dsRNA that loads RISC like an endogenous miRNA, or chemically modified antisense oligo that sequesters one. Used to manipulate miRNA networks.
  • Endo-siRNAs. Naturally occurring siRNAs from genome-encoded inverted repeats; abundant in oocytes, scarce in somatic cells.
  • Plant 24 nt siRNAs. RDR2 + DCL3 produce siRNAs that direct DNA methylation (RNA-directed DNA methylation, RdDM) — RNAi acting on chromatin.
  • Quelling and meiotic silencing. Fungal RNAi flavors discovered before the Fire/Mello paper, retroactively classified.
  • Bacterial CRISPR. A non-RNAi small-RNA system that uses Cas proteins (not Argonaute) and DNA targets. Mechanistically distinct, conceptually parallel.

Pitfalls and design considerations

  • Off-target seed effects. Any siRNA whose seed (nt 2–8) matches 3' UTRs of unintended transcripts will silence them like a miRNA. The original Jackson and Linsley 2003 paper showed this dominates apparent siRNA off-targets. Standard mitigation: 2'-O-methyl modification at position 2.
  • Innate immune activation. dsRNA >30 nt activates PKR and OAS; unmodified siRNA can trigger TLR7/8 (5'-UGUGU motif) and RIG-I (5'-triphosphate). Use chemically modified, blunt-end siRNAs to minimize.
  • Saturation of endogenous machinery. Overexpressed shRNAs can saturate exportin-5 and AGO, displacing miRNAs and causing toxicity. Grimm et al. (Nature 2006) reported lethal liver injury in mice from AAV-shRNA overload.
  • Delivery is the bottleneck. Naked siRNA is filtered by kidneys (half-life under 5 min) and digested by RNase A. LNP and GalNAc solved this for liver; brain, lung, and tumor delivery remain open problems.
  • Knockdown ≠ knockout. RNAi rarely gives 100% silencing. Residual protein from a stable transcript can persist for days. Confirm with western blot and CRISPR knockout when interpretation depends on full loss of function.
  • SMA cautionary tale. RNAi was once proposed for spinal muscular atrophy; in the end an antisense oligonucleotide (Spinraza) and a small molecule (Risdiplam) — both modulators of SMN2 splicing rather than RNAi — won approval. RNAi is one tool, not the only one.

Frequently asked questions

How was RNAi discovered?

Andrew Fire and Craig Mello injected sense RNA, antisense RNA, or both into C. elegans (1998). Sense and antisense alone gave weak silencing; the double-stranded mixture silenced the matching gene completely and could spread between cells. They published in Nature (Feb 1998) and won the Nobel Prize in 2006 — among the fastest discovery-to-Nobel runs ever recorded. Earlier hints came from co-suppression in petunias (Napoli, 1990) and quelling in Neurospora — RNAi was reinventing itself across kingdoms.

What's inside the RISC complex?

The minimal silencing engine is one Argonaute protein loaded with one ~22 nt guide RNA. Argonaute has four domains: N, PAZ (3' guide-end anchor), MID (5' phosphate pocket), and PIWI (RNase H-like fold that does the slicing). Mammals have four Argonautes (AGO1–4); only AGO2 has slicer activity. Loading requires Dicer plus TRBP and a conformational rearrangement that ejects the passenger strand. Mature RISC is remarkably stable and can turn over thousands of mRNAs.

What's the difference between siRNA and miRNA?

Both are ~22 nt and both load Argonaute, but they originate and act differently. siRNAs come from long perfectly base-paired dsRNA (often viral or transgenic); they bind their target with full complementarity and cleave it. miRNAs come from genome-encoded hairpins processed by Drosha and then Dicer; they bind partial complementary sites — usually in the 3' UTR — and trigger translational repression and mRNA decay rather than slicing. One miRNA regulates hundreds of mRNAs; one siRNA targets one transcript precisely.

How is RNAi used as a therapy?

Synthetic siRNAs target disease-causing transcripts. Delivery is the hard part: naked siRNA is degraded in serum and doesn't cross membranes. Patisiran (Onpattro, 2018) uses lipid nanoparticles to silence transthyretin in liver, treating hereditary amyloidosis. GalNAc conjugation drives liver-specific delivery in givosiran (porphyria), inclisiran (cholesterol), lumasiran (oxalosis), and vutrisiran (amyloidosis). Six approved siRNA drugs exist as of 2025.

What are off-target effects?

siRNAs can silence unintended transcripts by partial complementarity in the "seed" region (positions 2–8 of the guide), behaving like a miRNA. They can also trigger innate immunity through TLR3, TLR7, RIG-I, and PKR sensors that recognize dsRNA, causing interferon induction. Chemical modifications (2'-O-methyl, 2'-fluoro, phosphorothioate backbones) reduce both off-targeting and immune activation.

Does RNAi happen in mammals naturally?

Yes — but mostly as miRNA-mediated repression, not viral defense like in plants and invertebrates. Mammals replaced antiviral RNAi with the interferon system. Endogenous siRNAs do exist (in oocytes from pseudogene-derived dsRNA), and the Argonaute machinery still slices when given perfect matches, but the predominant cellular role of mammalian Ago is miRNA function.