Human canonical microRNA pathway — Microprocessor (DROSHA-DGCR8) pri-miRNA cleavage, XPO5-RAN-GTP pre-miRNA export, DICER1-TARBP2/PRKRA dicing, AGO1-4 RISC loading, and TNRC6-directed CCR4-NOT/PAN2-PAN3 deadenylation, DCP2 decapping and XRN1 decay

MicroRNAs (miRNAs) are ~22-nucleotide RNAs that guide Argonaute proteins to partially complementary sites, mostly in the 3' untranslated regions of messenger RNAs, and so repress translation and promote decay of hundreds of transcripts. Most human miRNAs are transcribed by RNA polymerase II as long primary transcripts (pri-miRNAs) that fold into stem-loop hairpins. In the nucleus the Microprocessor, made of the RNase III enzyme DROSHA and its double-stranded RNA-binding partner DGCR8, cuts the hairpin out of the pri-miRNA to release a ~60-70-nucleotide precursor (pre-miRNA) with a two-nucleotide 3' overhang. Exportin-5 (XPO5) binds the pre-miRNA together with GTP-bound RAN and carries it through the nuclear pore; GTP hydrolysis in the cytoplasm releases the cargo. The cytoplasmic RNase III enzyme DICER1, partnered by the dsRNA-binding protein TARBP2 (TRBP) or PRKRA (PACT), removes the terminal loop and leaves a ~22-base-pair miRNA duplex. The duplex is loaded into one of the four human Argonautes (AGO1-AGO4); one strand, the guide, is kept and the other is discarded, forming the miRNA-induced silencing complex (miRISC). AGO2 is the only efficient slicer and can cut fully complementary targets, but most animal miRNA targets pair only through the seed (guide nucleotides 2-7) and are silenced without cleavage. For those, AGO binds a GW182-family protein (TNRC6A, TNRC6B or TNRC6C), whose tryptophan-containing motifs recruit the CCR4-NOT deadenylase complex (through CNOT1 and CNOT9) and the PAN2-PAN3 deadenylase (through PAN3, which also binds the poly(A)-binding protein PABPC1). Deadenylation is followed by removal of the 5' cap by the DCP2-DCP1 decapping complex with its activators DDX6, PATL1, EDC3 and EDC4, and by 5'-to-3' degradation by XRN1. The same CCR4-NOT-DDX6 axis represses translation of the target. A few miRNAs bypass parts of this route: mirtrons are excised by the spliceosome instead of DROSHA, and miR-451 is matured by AGO2 slicing instead of DICER1.

Derived from the RGD Pathway Ontology diagram PW:0000808 (microRNA pathway) through the RGD_PATHWAY_ONTOLOGY project's diagram->module route. RGD curates 35 human genes to PW:0000808: ADAR ADARB1 AGO1 AGO2 AGO3 AGO4 CNOT1 CNOT10 CNOT2 CNOT3 CNOT6 CNOT6L CNOT7 CNOT8 CNOT9 DCP1B DCP2 DDX6 DGCR8 DICER1 DROSHA EDC3 EDC4 PABPC1 PAN2 PAN3 PATL1 RAN TARBP2 TNRC6A TNRC6B TNRC6C TUT4 TUT7 XPO5. Thirty-one are modelled as participants; ADAR, ADARB1, TUT4 and TUT7 are left as context (decision 4). Three genes not on the RGD list are added: PRKRA (the second DICER1 dsRBD partner, which the task brief and the DICER1 gene review both name), DCP1A (the better characterised DCP1 paralog; RGD lists only DCP1B) and XRN1 (drawn on the RGD diagram as associated with DCP1-DCP2, but absent from the human gene list). The mapped GO process GO:0035196 miRNA processing is the first module concept and an evidence entry; it is a relatedMatch, not an exact match, because the PW pathway also covers silencing, so GO:0035195 miRNA-mediated post-transcriptional gene silencing is the second concept. The diagram's sub-pathway links (PW:0000101 translation pathway, PW:0000127 RNA polymerase II transcription pathway, PW:0001073 spliceosome pathway) are neighbouring pathways and are not counted as covered by this module; transcription of pri-miRNAs by RNA polymerase II, and the translation that miRISC represses, are outside the boundary. Existing modules. No human mRNA deadenylation, CCR4-NOT or decapping module exists in modules/ (searched November 2026 for CCR4, CNOT, deadenylation, decapping, PAN2, DCP2, XRN1), so there was nothing to module_ref and parts 6-8 model these complexes directly. They are general mRNA-decay machinery recruited to miRNA targets by TNRC6, and are good candidates for a separate, reusable human "cytoplasmic mRNA deadenylation and decapping" module; when one is created, parts 6-8 should become module_ref links to it. modules/dmel_ccr4_not_complex.yaml (fly CCR4-NOT) and modules/dmel_risc_loading_complexes.yaml (fly Dicer-1/Loqs-PB loading into AGO1 and the AGO1-Gawky miRISC) model the Drosophila counterparts of parts 4-6, and modules/plant_mirna_biogenesis.yaml models the plant pathway, where a single nuclear DCL1 makes both cuts and HEN1 methylates the duplex. They are referenced in prose only: a cross-species module_ref would assert an identity between fly or plant complexes and the human ones that this module does not need. Molecular functions were taken from each protein's experimental (and, where noted, IBA) GO annotations in QuickGO (checked 2026-11-05) and from completed gene reviews where they exist (AGO1, AGO2, AGO3, AGO4, CNOT1, CNOT3, DICER1). DROSHA and DICER1: GO:0004525 ribonuclease III activity (IBA in GOA; DICER1 core function; DROSHA by the cited biochemistry). DGCR8: GO:0140517 protein-RNA adaptor activity (IDA, PMID:15574589), the term the human TRUB1 GO-CAM uses. XPO5: GO:0005049 nuclear export signal receptor activity (IBA), because its experimental pre-miRNA binding row is contributes_to (with RAN). RAN: GO:0003925 G protein activity (IDA). TARBP2: GO:0003725 double-stranded RNA binding (IBA); its pre-miRNA binding row is contributes_to. PRKRA: GO:0070883 pre-miRNA binding (IDA). AGO1-4: GO:0035198 miRNA binding (core in all four reviews); AGO2 additionally GO:0090624 endoribonuclease activity, cleaving miRNA-paired mRNA (IDA/IMP; core in the AGO2 review). TNRC6A/B/C: GO:0030674 protein-macromolecule adaptor activity (see the GOA observation below). CNOT1: GO:0140378 protein complex scaffold activity (core in the CNOT1 review). CNOT6, CNOT6L, CNOT7, CNOT8 and PAN2: GO:0004535 poly(A)-specific ribonuclease activity (EXP/IDA). PABPC1: GO:0008143 poly(A) binding (IDA). DCP2: GO:0140933 m7G-cap hydrolase activity (EXP/IDA). XRN1: GO:0004534 5'-3' RNA exonuclease activity (IDA/IMP). CNOT9, PAN3, DCP1A/B, EDC3, EDC4, DDX6 and PATL1 have no experimental MF in GOA that describes their role here (only binding terms, contributes_to rows, or an adaptor row from a different complex), so they carry a role and evidence but no GO function id, rather than a plausible-looking term the evidence does not support. CNOT2, CNOT3 and CNOT10 are listed as structural CCR4-NOT subunits without a function; the CNOT3 review's core term is GO:0180014 protein-tRNA adaptor activity, a role unrelated to this pathway. Boundary decisions. (1) Canonical route only as parts. Mirtrons (short introns released by the spliceosome and debranched, bypassing DROSHA; PMID:17964270) and the DICER1-independent miR-451 route (pre-miR-451 cut by AGO2 slicer activity, PMID:20448148) are recorded in prose and on the AGO2 slicer annoton, not as parts. Neither adds a protein to the module (mirtrons use the spliceosome and lariat-debranching enzyme, outside this boundary; miR-451 uses AGO2, already a participant), and both converge on the same export, loading and silencing steps. (2) Argonautes are a ONE_OR_MORE variant set on the slicer axis: AGO2 has two annotons (miRNA binding and GO:0090624 slicing); AGO1 (NOT|enables RNA endonuclease activity, IDA), AGO3 (slicing only with particular guides and targets) and AGO4 carry miRNA binding only. AGO3's restricted slicing is recorded on its annoton but not asserted as its role here. (3) TARBP2 and PRKRA are a ONE_OR_MORE variant set of DICER1 partners; neither is required for the cleavage chemistry, so neither is given a ribonuclease activator activity. (4) RNA editing and uridylation are context, not parts. ADAR and ADARB1 deaminate adenosines in some pri- and pre-miRNAs, which can block DROSHA or DICER1 processing or retarget a miRNA, and TUT4 and TUT7 uridylate pre-miRNAs (monouridylation of group II pre-let-7 promotes dicing; LIN28-directed oligouridylation blocks it). These are substrate- specific regulatory layers acting on the pathway's RNA intermediates, not steps every miRNA passes through; they are what RGD draws as side reactions. Such roles fit GO's regulation-of-miRNA-processing terms rather than miRNA processing itself, and modelling them as parts would make the canonical route look conditional on editing or uridylation. Their gene symbols are recorded here for the triage, and they are the obvious first members of a future "regulation of miRNA biogenesis" module together with LIN28A/B. (5) The RAN cycle (RCC1, RANGAP1, RANBP1) is general nuclear-transport machinery and is left out; RAN itself is a part because the XPO5-pre-miRNA complex does not form without RAN-GTP. (6) Deadenylation, decapping and XRN1 decay are parts, not context, because miRISC recruits the deadenylases directly through TNRC6 (PMID:21981923, PMID:21984185) and RGD draws all three as outputs of the pathway. XRN1 is optional (decision in part 8). (7) Translational repression is recorded as a process term on the TNRC6 and DDX6/PATL1 annotons and not as a separate part, because its mechanism in human cells (CCR4-NOT-DDX6 and 4E-T/4EHP, eIF4A displacement) is not settled and RGD itself calls it poorly understood. (8) The "substrate is not a participant" rule was applied to PABPC1: it is included because it does part of the work (stimulating PAN2-PAN3 and binding the TNRC6 PAM2 motif), not because it sits on the poly(A) tail being removed. Target mRNAs and the miRNAs themselves are not participants. GO-CAM and GOA observations for upstream review. Few production GO-CAMs contain these proteins (gocams/index.tsv, checked November 2026). The seed's best GO-CAM lead, gomodel:653b0ce600001283 (KSHV-ORF57 inhibition of host P-body formation), is a viral-antagonism model whose AGO2 RNA endonuclease and TNRC6A adaptor activities are linked here only as partial matches. The DGCR8 activity is grounded in gomodel:62183af000000874 (miRNA maturation by TRUB1) and the CNOT7 deadenylase in gomodel:63894f2500001378 (FMR1-CAPRIN1 condensates). No GO-CAM models the canonical human pathway itself (no model contains DROSHA, DICER1, XPO5 with pre-miRNA, PAN2, DCP2 in a miRNA context, or AGO1/3/4); this is the largest curation gap the module reveals. RAN's only GO-CAM (protein export from nucleus, PKM2 model) is unrelated. TARBP2's and PRKRA's GO-CAMs are antiviral models (PKR, IRF7), not miRNA processing. TNRC6A's only GOA row for GO:0030674 is HDA from an mRNA-interactome screen (PMID:22681889), which can establish RNA binding but not adaptor activity; it may be worth an upstream look, and a literature-based IDA from PMID:21981923 or PMID:21984185 would be the better grounding. Gene reviews exist for AGO1, AGO2, AGO3, AGO4, CNOT1, CNOT3 and DICER1; DROSHA, DGCR8, XPO5, RAN, TARBP2, PRKRA, TNRC6A-C, the other CNOT subunits, PAN2, PAN3, PABPC1, DCP1A/B, DCP2, EDC3, EDC4, DDX6, PATL1 and XRN1 have none. PANTHER family ids and PTNs are deliberately not asserted.