MicroRNAs

From genome to silenced mRNA

1. Transcription. RNA Pol II transcribes a long primary miRNA (pri-miRNA), often hosted in introns of protein-coding genes or in dedicated miRNA loci. The pri-miRNA folds into hairpins.
2. Cropping (nucleus). The Microprocessor complex — Drosha (RNase III) plus DGCR8 — recognizes the hairpin's basal segment and cleaves ~11 bp from the junction, releasing a ~70 nt pre-miRNA with a 2-nt 3' overhang.
3. Export. Exportin-5 with Ran-GTP transports the pre-miRNA through the nuclear pore.
4. Dicing. Dicer (with TRBP) measures from the 3' end and removes the loop, generating a ~22 nt RNA duplex with 2-nt overhangs at each end.
5. RISC loading. AGO grabs the duplex; the strand with the less stable 5' end becomes the guide. The passenger (or miRNA*) is ejected.
6. Target search. Mature miRISC scans 3' UTRs for matches to its seed (nt 2–8). Most miRNA-target interactions involve 6mer, 7mer-A1, 7mer-m8, or 8mer seed types.
7. Silencing. miRISC recruits GW182/TNRC6, which calls in the CCR4-NOT and PAN2-PAN3 deadenylases. Poly-A shortening leads to decapping (DCP1/DCP2) and 5'→3' decay (XRN1). Translation is repressed in parallel.

Pathway diagram — miRNA biogenesis

  pri-miRNA  (long Pol II transcript, hairpin structure)
       │
       ▼  Drosha + DGCR8 (nucleus)
       │
       ▼  pre-miRNA  (~70 nt, 2-nt 3' overhang)
       │
       ▼  Exportin-5 / Ran-GTP  → cytoplasm
       │
       ▼  Dicer + TRBP
       │
       ▼  miRNA / miRNA* duplex (~22 nt)
       │
       ▼  AGO1–4 loading; passenger ejected
       │
       ▼  miRISC scans 3' UTRs (seed match nt 2–8)
       │
       ▼  GW182 → CCR4-NOT → deadenylation
       │
       ▼  decapping (DCP1/DCP2) + XRN1 decay
       │      + translational repression
       │
       ▼  silenced target

Seed match types and their relative efficacy

Site type	Seed pairing (guide nt)	Extra constraint	Typical repression	Frequency in conserved sites	Example
8mer	2–8 (Watson-Crick)	+ A across from nt 1	Strongest	~10%	let-7 → HMGA2 (most-studied)
7mer-m8	2–8	None	Strong	~30%	miR-122 → many liver mRNAs
7mer-A1	2–7	+ A across from nt 1	Moderate	~25%	miR-21 → PDCD4
6mer	2–7	None	Weak	~35%	Many tissue-specific tunings
3' compensatory	Imperfect 2–8 + perfect 13–16	3' supplementary pairing	Variable	Rare	lin-4 → lin-14 (the founding case)
Centered site	4–14 contiguous	No seed required	Moderate	Rare	Few characterized examples
3' UTR vs CDS	—	UTR more effective; ribosome displaces RISC from CDS	—	—	Standard design rule

Real-world impact and examples

• let-7 and lung cancer. The let-7 family represses RAS, MYC, and HMGA2. Most non-small-cell lung cancers downregulate let-7 — restoring it in mouse models slows tumor growth. The 13-member let-7 family in humans is the textbook tumor-suppressor miRNA.
• miR-21 oncogene. Upregulated in nearly every solid tumor. Targets PTEN, PDCD4, TPM1, SPRY2 — all tumor suppressors. miR-21 is also induced by inflammation, linking chronic inflammation to cancer at the regulatory layer.
• miR-122 and hepatitis C. Liver-specific miR-122 binds HCV's 5' UTR and stabilizes the viral RNA — the rare case of a miRNA enhancing rather than repressing. Miravirsen, a locked-nucleic-acid antimiR-122, reduced viral load in phase 2 trials.
• Heart disease. miR-208a is heart-specific and required for cardiac stress response; antimiR-208 prevents cardiac hypertrophy in animal models. miR-1 and miR-133 are the most abundant muscle miRNAs.
• Circulating biomarkers. miRNAs are stable in serum (RNase-protected in exosomes and AGO complexes) and tissue-specific. Cardiac miR-208a in serum spikes within hours of myocardial infarction; miR-21 elevations correlate with several cancers. Liquid-biopsy panels are in active clinical development.
• Onpattro / Vutrisiran are not miRNA drugs. They use the RNAi pathway with synthetic siRNAs, not miRNAs. The two pathways share Argonaute but differ in design and behavior.

Variants and special cases

• Mirtrons. miRNAs spliced directly out of small introns, bypassing Drosha. Discovered in flies (Berezikov, 2007); ~150 in humans.
• Polycistronic clusters. Multiple miRNAs co-transcribed: miR-17-92 (oncomiR cluster), let-7 family, miR-23a/27a/24-2.
• Arm switching. The 5p and 3p arms of a hairpin can both be functional, with relative loading varying by tissue or development.
• isomiRs. Variant miRNA forms with shifted ends — different seeds, different targetomes.
• Mirtrons and shRNA-like loci. Boundary cases that blur the canonical pathway.
• Plant miRNAs. Bind with near-perfect complementarity in coding regions and slice their targets — closer to siRNA behavior than to mammalian miRNAs.
• Viral miRNAs. Herpesviruses (EBV, KSHV, HCMV) encode their own miRNAs that hijack host machinery; KSHV alone encodes 25.

Pitfalls in miRNA research

• Target prediction is noisy. TargetScan, miRDB, and DIANA each predict thousands of targets per miRNA; overlap is modest. Computational predictions need experimental validation (luciferase reporter, AGO-CLIP).
• Knockout effects are subtle. Most single miRNA knockouts show no overt phenotype because miRNAs act in redundant families and provide fine-tuning rather than on/off switches. The let-7 family has 13 members; deleting one is masked by the rest.
• Overexpression artifacts. Transfecting a miRNA mimic at high concentration saturates AGO and produces non-physiological seed-driven silencing of unrelated transcripts.
• 3' UTR overlooked. Many functional studies clone the CDS into reporter constructs and miss the 3' UTR where most miRNA action happens. Always include the native 3' UTR in luciferase assays.
• Serum miRNA pitfalls. Hemolysis releases red-cell miRNAs (miR-451, miR-16) and can swamp tissue-specific signals. Pre-analytical processing must be controlled.
• Confusion with siRNA. Same machinery downstream, different upstream biogenesis and targeting logic. siRNA designs cleave; miRNA designs repress. Use the right tool for the experiment.