AIGR Hypothesis Review — S. pombe atg2 (O94649) lipid transfer activity (GO:0120013)
Focus type: computational_prediction Prediction under review: GO-GPT (via BioReason-Pro) predicts lipid transfer activity (GO:0120013) for the Schizosaccharomyces pombe autophagy protein atg2 (UniProt O94649).
Executive Judgment
Verdict: SUPPORTED (with one specificity caveat).
The prediction that atg2 has lipid transfer activity is biologically correct and rests on unusually strong, organism-matched evidence. atg2 is the founding member of the Atg2 bridge-like lipid-transfer-protein (LTP) family, and the very protein used to establish Atg2 lipid-transfer activity in the seminal biochemical study was the S. pombe ortholog itself (Osawa et al. 2019, 30911189). The UniProt record for O94649 confirms the diagnostic domain architecture: a Chorein N-terminal domain (residues 26–121), InterPro ATG2 (IPR026849), and Pfam ATG2_CAD (PF13329), on a long (1646-aa) rod-like scaffold typical of bridge-like LTPs. In vivo phospholipid flux from ER to the isolation membrane and cryo-EM visualization of lipids filling the internal hydrophobic cavity have since corroborated the activity in orthologs.
Most important caveat (specificity, not correctness): GO:0120013 is a high-level parent term. PomBase already annotates O94649 with a more specific child, triglyceride transfer activity (GO:0140344, TAS), and with intermembrane phospholipid transfer (GO:0120010, BP, TAS). The directly assayed activity is phospholipid transfer, which maps to the more specific MF child phospholipid transfer activity (GO:0120014). Thus the prediction is not wrong — it is simply less precise than existing curated knowledge. The seed's "lipid transfer vs. tethering/scaffold" dichotomy is a false binary: atg2 does both (it also carries EXP-supported protein-membrane adaptor activity, GO:0043495), with lipid transfer being the defining catalytic-like molecular function.
Evidence Matrix
| Citation | Evidence type | Supports/Refutes | Claim tested | Key finding | Context | Confidence & limitations |
|---|---|---|---|---|---|---|
| Osawa et al. 2019 (30911189) | Direct in vitro assay + mutant phenotype | Supports | atg2 directly transfers phospholipids | S. pombe Atg2 N-terminal region has a lipid-transfer hydrophobic cavity accommodating phospholipid acyl chains; bridges curved liposomes and transfers phospholipid in vitro; transfer-impairing mutations block autophagosome formation in vivo | S. pombe Atg2, reconstituted liposomes | High. Exact target protein/organism. In vitro reconstitution. |
| UniProt/InterPro (O94649; IPR026849, PF13329, Chorein-N 26–121) | Structural/evolutionary (computational) | Supports | atg2 has Atg2-family LTP architecture | 1646-aa rod protein with Chorein N-terminal domain, ATG2 InterPro, ATG2_CAD Pfam, PANTHER PTHR13190 | Database record | High. Domain assignment is diagnostic of the bridge-like LTP fold. |
| AlphaFold AF-O94649-F1 (v6) | Structural (computational, this study) | Supports | atg2 forms an elongated bridge-like rod | End-to-end Cα distance 209 Å (~21 nm); anisotropy 12.5; Rg 63 Å; Chorein-N (26–121) pLDDT 78.8 | AlphaFold model, full 1646 aa | Medium-high. Predicted model (global pLDDT 59); rod length matches an ER–phagophore contact-site bridge. |
| McEwan & Ryan 2022 (34783437) | Review | Supports | Chorein-N ATG2 proteins are lipid transporters | ATG2/VPS13 Chorein-N proteins form molecular bridges transporting lipids ER→phagophore | Review synthesis | Medium (review-level orientation). |
| Hao et al. 2026 (41805856) | In vivo probe assay | Supports | Atg2 transfers phospholipid in vivo | R18 dye tracing shows phospholipid transfer from ER to isolation membrane via Atg2 during autophagy | Yeast, in vivo | High for pathway; ortholog (S. cerevisiae) context. |
| Ramirez et al. 2026 (42162239) | Structural (cryo-EM) + MD | Supports | Atg2 cavity holds/transfers lipids | Lipid densities fill Atg2's internal hydrophobic cavity along its full length; complex promotes lipid transfer into phagophore | Yeast Atg2–Atg18 | High for mechanism; ortholog context. |
| Zheng et al. 2025 (40128367) | Regulation/mechanism | Supports | ATG2 is a lipid transfer protein | ATG2A described as a rod-like LTP transporting phospholipids ER→phagophore; S-palmitoylation regulates it | Human ATG2A, cells | High for family activity; human ortholog. |
| PomBase/UniProt existing GO (GO:0140344 TAS; GO:0120010 TAS; GO:0043495 EXP) | Database/curation | Qualifies | Most-specific supported MF | atg2 already curated with triglyceride transfer activity + intermembrane phospholipid transfer + membrane-adaptor activity | Database | High. Shows prediction is redundant with a deeper child term. |
| Valverde et al. 2019 (30952800) | Direct in vitro assay + mutant | Supports (convergent) | ATG2 family transfers lipid | Human ATG2A binds tens of glycerophospholipids at once and transfers lipids robustly in vitro; N-terminal fragment is necessary and sufficient to rescue autophagosome biogenesis | Human ATG2A, in vitro/cells | High. Independent lab + ortholog, same conclusion, same year as 30911189. |
| Ghanbarpour/Valverde et al. 2021 (33850023) | Model/mechanism | Qualifies | LTP + scramblase partnership | ATG2 (LTP) partners with scramblases TMEM41B/VMP1/ATG9 for membrane expansion | Model, in vitro assays | Medium. Frames tethering/scramblase roles as complementary, not competing. |
| Wang et al. 2001 (11382760) | Mutant phenotype | Qualifies (context) | Atg2 required for autophagy/Cvt/pexophagy | Atg2 is a peripheral membrane protein essential for sequestering-vesicle formation | S. cerevisiae | Medium; establishes membrane association, predates LTP model. |
GO Curation Implications
Lead (requires curator verification):
- The MF prediction lipid transfer activity (GO:0120013) is biologically supported and should not be rejected. However, because it is a parent term and PomBase already annotates the more specific children, the recommended curation action is to generalize the prediction to the most-specific experimentally supported MF term rather than add the generic parent:
- Best-supported MF: phospholipid transfer activity (GO:0120014) — directly assayed for the S. pombe protein (30911189, EXP/IDA-grade). This is a more informative annotation than GO:0120013.
- Existing triglyceride transfer activity (GO:0140344, TAS) is also a valid child; a curator may reconcile which lipid-species child(ren) are best evidenced.
- Retain the companion BP annotation intermembrane phospholipid transfer (GO:0120010) and CC phagophore/phagophore assembly site annotations.
- Do not replace lipid transfer with "protein binding" or with tethering alone; the adaptor/tethering role (GO:0043495, EXP) is real but complementary, not a substitute for the lipid-transfer MF.
Net: the prediction is a correct-but-generic lead. Curation outcome = retain the concept, make it more specific (GO:0120014), avoid redundant parent annotation.
Mechanistic Scope
Immediate molecular function tested: direct, non-vesicular transfer of phospholipid monomers between membrane bilayers via an elongated Chorein-N/VPS13-like hydrophobic groove. This is a bona fide molecular activity of the atg2 gene product (demonstrated by cell-free reconstitution with purified S. pombe Atg2), distinct from: - Downstream phenotype: autophagosome/phagophore biogenesis and autophagic flux (BP/CC consequences). - Complementary role: membrane tethering/adaptor activity bridging ER exit sites and the phagophore rim (positions the LTP; enables but is not the transfer chemistry itself).
The lipid-transfer activity is thus the core molecular function, not an inference from loss of function.
Conflicts and Alternatives
- Paralog/ortholog confusion: None problematic. Much mechanistic detail comes from S. cerevisiae Atg2 and human ATG2A/ATG2B, but the defining biochemistry was done on the S. pombe protein itself, so cross-species carry-over is not the basis of the call.
- In-vitro-only concern: Initially the activity was in vitro (30911189); this has since been supported in vivo (41805856) and structurally (42162239), reducing the artifact concern.
- "Tethering-only" alternative: Older work framed Atg2 as a peripheral membrane/tethering protein (11382760). This is not a true conflict — modern data show tethering and lipid transfer coexist in one rod-like molecule. The scramblase-partnership model (33850023) explicitly casts ATG2 as the lipid transfer component working with separate scramblases, reinforcing rather than replacing the LTP assignment.
- No refuting evidence found. Targeted literature searches for a "tethering-without-transfer" or disputed-lipid-transfer view returned nothing. Instead, an independent group demonstrated the same activity in the human ortholog ATG2A (30952800) in the same year as the S. pombe assay — two labs, two orthologs, convergent conclusion. The lipid-transfer assignment is a robust cross-species consensus, not a single-paper or single-species claim.
- Lipid-species specificity: A genuine open question is which lipids atg2 preferentially transfers (phospholipid vs. triglyceride vs. broad). PomBase's triglyceride-transfer annotation vs. the phospholipid-transfer assays is a specificity nuance for curators, not a challenge to the parent-level prediction.
Knowledge Gaps
- Lipid-species preference of atg2 specifically. Checked: assays show phospholipid transfer (30911189); PomBase lists triglyceride transfer (TAS). Matters because it determines the most-specific MF child term. Resolve with headgroup-resolved in vitro transfer assays (PE/PC/PI/PS vs. neutral lipids) on purified O94649.
- Rate/directionality in the S. pombe cell. Checked: directionality/reversibility shown in budding yeast (41805856). Resolve with an S. pombe in vivo phospholipid-flux probe.
- Structure of S. pombe atg2. Checked: cryo-EM exists for other orthologs (42162239); AlphaFold model of O94649 is available but not experimentally solved. Resolve with cryo-EM of the S. pombe Atg2–Atg18 complex.
Discriminating Tests
- Headgroup-resolved liposome transfer assay with purified recombinant O94649 (± Chorein-N cavity mutations) to confirm phospholipid vs. triglyceride specificity and pin the deepest MF term.
- Cavity-mutant complementation in S. pombe atg2Δ: mutations that block in vitro transfer should block autophagosome formation (already indicated by 30911189) — separates transfer function from tethering.
- AlphaFold/structure comparison of O94649 against solved Atg2/VPS13 structures to confirm the continuous hydrophobic groove (public, low-cost confirmation of the bridge-like fold).
Curation Leads (require curator verification)
- Candidate reference to attach: PMID 30911189 (Osawa et al. 2019) — snippet: "the conserved amino-terminal region of Schizosaccharomyces pombe Atg2 includes a lipid-transfer-protein-like hydrophobic cavity that accommodates phospholipid acyl chains" and "Atg2 bridges highly curved liposomes, thereby facilitating efficient phospholipid transfer in vitro." Organism-exact, assay-grade support.
- Candidate MF term (more specific than prediction): phospholipid transfer activity GO:0120014 (child of GO:0120013), evidence code EXP/IDA via 30911189.
- Action change: treat the GO:0120013 prediction as supported but subsumed; annotate the specific child rather than the parent to avoid redundancy with existing GO:0140344/GO:0120010.
- Keep: GO:0043495 (protein-membrane adaptor activity, EXP) as a complementary MF; GO:0120010 (BP) and phagophore CC terms.
- Suggested question for curator: Which lipid-species child term(s) best reflect atg2's assayed specificity — phospholipid (assay) vs. triglyceride (current TAS)?
- Suggested experiment: headgroup-resolved transfer assay on purified O94649 (see Discriminating Tests).
Provenance
Computational checks run (see execute_code outputs): (1) UniProt O94649 domain/feature and existing-GO retrieval — confirmed Chorein-N domain (26–121), IPR026849, PF13329, and existing lipid-transfer-related GO annotations; (2) QuickGO ontology relationship check — confirmed GO:0120013 is an is_a ancestor of both GO:0140344 and GO:0120014; (3) AlphaFold model AF-O94649-F1 (v6) Cα geometry — end-to-end 209 Å, anisotropy 12.5, Chorein-N pLDDT 78.8, confirming an elongated bridge-like rod. No results were fabricated; all API calls returned live data.
Artifact files (in artifacts/): atg2_evidence_matrix.csv, atg2_GO_decision_table.csv, O94649_structural_metrics.csv.