ATG3 is the E2-like conjugating enzyme of the autophagy ubiquitin-like conjugation system that lipidates ATG8-family proteins (LC3s and GABARAPs). After the E1 ATG7 activates an ATG8-family protein, ATG3 receives it as a thioester on its catalytic cysteine and, together with the E3-like ATG12~ATG5-ATG16L1 complex, transfers its C-terminal glycine to phosphatidylethanolamine in the phagophore membrane. An N-terminal amphipathic helix senses membrane curvature and couples it to the catalytic core, restricting lipidation to the highly curved rim of the growing phagophore, and a noncanonical LIR motif engages LC3/GABARAP. ATG3 also acts as the E2 for its own conjugation to ATG12 on a single lysine; the ATG12-ATG3 conjugate is dispensable for starvation-induced autophagy but affects mitochondrial homeostasis and late endosome function. ATG3 is cytosolic and is recruited to phagophore membranes. The donkey ATG3 gene encodes at least two products that differ by a 28-residue segment inserted after residue 222; this entry is the 342-residue long splice form, which keeps the catalytic cysteine and the ATG12-acceptor lysine, while the 314-residue product matches human ATG3 in length and is 98.4% identical to it.
Summary: TreeGrafter transfer from PANTHER node PTN000305319 (ATG3 subfamily PTHR12866:SF2, to which donkey, human and yeast ATG3 all belong). ATG3-dependent LC3/GABARAP lipidation is one of the key steps of phagophore expansion.
Reason: ATG3 performs the lipidation step of autophagosome biogenesis (it catalyzes the transfer of ATG8-family proteins to PE), so it does part of the work of the process; human ATG3 has IMP support (PMID:33446636). The donkey protein sits in the same PANTHER subfamily as human and yeast ATG3, and both donkey splice products keep the catalytic cysteine. Whether the 28-residue insertion specific to this long splice form alters activity is untested; the annotation is accepted at the gene level.
During autophagy the enzyme Atg3 catalyzes the covalent conjugation of LC3 to the amino group of phosphatidylethanolamine (PE) lipids, which is one of the key steps in autophagosome formation.
Summary: TreeGrafter phagophore assembly site from PTN000305319.
Reason: ATG3 acts on the phagophore membrane, where its N-terminal curvature-sensing amphipathic helix targets the enzyme to the phagophore tip and confines lipidation to the target membrane; human ATG3 carries the same location by IBA.
Summary: TreeGrafter autophagy of mitochondrion from PTN000305319.
Reason: Mitophagy uses the core LC3-lipidation machinery that ATG3 supplies, so the inherited family assertion (also on human ATG3 by IBA) is reasonable, but it is one of several selective forms of autophagy rather than a distinct ATG3 function. The separate ATG12-ATG3 role in mitochondrial homeostasis (PMID:20723759) is about mitochondrial mass and cell death, not mitophagy as such, and does not change this call.
Reason: ATG3 is a soluble cytoplasmic enzyme (human IDA, PMID:11825910) that cycles between ATG7 and the membrane-bound E3; the donkey protein has no signal peptide or transmembrane segment.
Summary: TreeGrafter Atg8-family ligase activity from PTN000305319. In GO this term is_a GO:0061659 ubiquitin-like protein ligase activity (the E3 class), whereas ATG3 is the E2 that receives ATG8-family proteins from ATG7 as a thioester and transfers them to PE.
Reason: The activity is real, but GO:0141046 Atg8-family conjugating enzyme activity (is_a GO:0061650 ubiquitin-like protein conjugating enzyme activity, the E2 class) describes the thioester-relay mechanism of ATG3 correctly. Human ATG3 carries GO:0141046 by IDA (PMID:24186333, PMID:37252361), and the current PAINT annotation on human ATG3 from the same ATG3 subfamily is GO:0141046 by IBA, so this TreeGrafter row looks like an older E3-class term that the family curation has since moved to the E2-class term.
ATG7 functions as an activating E1 enzyme using ATP to form a thioester with LC3/GABARAP and transferring them to an E2 enzyme, ATG3, via a trans-thioesterification reaction.
Atg3 likely cycles back and forth between binding to Atg7 for loading with the UBL LC3 and binding to E3 to promote LC3 lipidation
GO:0019787 ubiquitin-like protein transferase activity
IEA GO_REF:0000120
MODIFY
Summary: Ubiquitin-like protein transferase activity, combined ARBA/InterPro (Atg3/Atg10 family).
Reason: True but less specific than the evidence supports. GO:0141046 Atg8-family conjugating enzyme activity, which the gene already carries through the MODIFY of the TreeGrafter ligase row, is a descendant (via ubiquitin-like protein conjugating enzyme activity) and states the core activity. ATG3 also acts as the E2 for ATG12 conjugation onto its own lysine (PMID:20723759), but PAINT did not propagate an ATG12-transfer term to the ATG3 subfamily, so that activity is not added here.
ATG12-ATG3 complex formation requires ATG7 as the E1 enzyme and ATG3 autocatalytic activity as the E2, resulting in the covalent linkage of ATG12 onto a single lysine on ATG3.
Summary: TreeGrafter nucleophagy from PTN000305319.
Reason: Inherited family assertion (human ATG3 carries the same IBA) that the core ATG8 conjugation machinery acts in autophagic degradation of nuclear material; plausible because nucleophagy depends on Atg8-family lipidation, but not a distinct ATG3 role.
Summary: TreeGrafter glycophagy from PTN000305319.
Reason: Inherited family assertion (human ATG3 carries the same IBA) that the core lipidation machinery supports autophagic glycogen delivery; plausible by the same reasoning as other selective autophagy terms, non-core.
Core Functions
E2-like enzyme that accepts ATG8-family proteins (LC3/GABARAP) from ATG7 as a thioester on its catalytic cysteine and, with the ATG12~ATG5-ATG16L1 E3, conjugates them to phosphatidylethanolamine on the phagophore membrane, driving autophagosome assembly.
ATG7 functions as an activating E1 enzyme using ATP to form a thioester with LC3/GABARAP and transferring them to an E2 enzyme, ATG3, via a trans-thioesterification reaction.
During autophagy the enzyme Atg3 catalyzes the covalent conjugation of LC3 to the amino group of phosphatidylethanolamine (PE) lipids, which is one of the key steps in autophagosome formation.
Q: Is the 342-residue ATG3 splice form, whose extra 28-residue exon is predicted in both donkey and horse genomes but absent from human, expressed as protein in equid tissues, and does the insertion (21 residues upstream of the ATG12-acceptor lysine) alter LC3 lipidation or ATG12 autoconjugation?
Suggested Experiments
Experiment: Express the 342-residue and 314-residue donkey ATG3 products in ATG3-knockout cells and compare LC3-II formation, ATG12-ATG3 conjugate formation and in vitro LC3-PE conjugation, and use RT-PCR across the extra exon to measure how much of the long transcript is made in donkey tissues.
External Prediction Reviews
These computational predictions are reviewed separately from the GOA annotation set used for this review. The assessments below are from this project and do not constitute official GO annotations or endorsement by GO/UniProt. They are not included in the existing annotation review above.
Assessment of the single ProtNLM2 GO prediction for donkey ATG3 (A0A9L0IHI0, DONKEY40 cohort), using the reviewed human ortholog Q9NT62, human ATG3 literature, and a donkey-human alignment (98.41% identity, catalytic cysteine conserved, 28-residue insertion in the donkey model). The prediction, cytoplasm, is correct but less precise than the cytosol and phagophore assembly site annotations already in donkey GOA from TreeGrafter (LSP); the ProtNLM2 evidence record cites phagophore assembly site. ProtNLM2 emitted no function text or subcellular-location statement for this accession in the snapshot.
LSP — Less precise than existing annotation Review score: 2/2
Prediction method: ProtNLM2 · Version: UniProt API snapshot 2026-10-14
Review rationale: Human ATG3 is experimentally cytoplasmic and acts as a soluble E2 that is recruited to phagophore membranes; the donkey protein is 98.4% identical with the catalytic cysteine and membrane-sensing N-terminus conserved, so cytoplasm is correct. Donkey GOA already contains cytosol and phagophore assembly site (TreeGrafter), both part of cytoplasm in GO, so the prediction is less precise than existing annotation. Training-set membership is unknown.