AIGR Gene Hypothesis Deep Research — Final Report

Target: Gadus morhua tbc1d14 (UniProt A0A8C5FPT8)

Focus: computational_prediction — autophagosome localization (GO:0005776)

Hypothesis slug: prediction-autophagosome

Summary

The seed hypothesis proposed that ProtNLM2's prediction of autophagosome localization (GO:0005776) for the Gadus morhua protein tbc1d14 is likely a misassignment — the reasoning being that TBC1D14 is a TBC-domain Rab GTPase-activating protein (RabGAP) characterized as a recycling-endosome/ULK1-complex regulator of autophagy initiation, not a structural, autophagosome-resident protein. On that logic, placing the protein "inside" the autophagosome would be an over-annotation.

This investigation confirms the biological substance of that concern but overturns its conclusion that the prediction is a hallucination or clear misassignment. The decisive finding, obtained by directly interrogating curated GO annotations for the human ortholog TBC1D14 (Q9P2M4), is that autophagosome (GO:0005776) is an existing, manually curated annotation supported by two independent evidence lines: an IDA (Inferred from Direct Assay) annotation from PMID: 22613832 assigned by expert BHF-UCL curators, and an independent IBA (phylogenetic) annotation from GO_Central (PANTHER node PTN000538677). Because the Gadus morhua protein is a bona fide 1:1 ortholog (62.1% identity, identical TBC domain architecture with intact catalytic fingers, same PANTHER family PTHR47219), it is legitimately eligible for the same orthology-based inference. The ProtNLM2 prediction therefore coincides with existing curated knowledge rather than contradicting it.

The essential caveat is one of precision, not correctness. Autophagosome is a secondary, condition-dependent localization. TBC1D14's primary and best-characterized compartment is the Rab11/transferrin-receptor–positive recycling endosome (relocalizing to the Golgi upon starvation), and its role in autophagy is regulatory — it is a negative regulator of autophagy initiation acting via ULK1 and the TRAPP–ATG9 membrane-delivery axis, not an autophagosome structural protein. The verdict is therefore PARTIALLY SUPPORTED: GO:0005776 is defensible for the fish protein by orthology and should be retained, but only alongside recycling endosome (GO:0055037) and the regulation-of-autophagy BP terms — never as the sole or primary cellular-component annotation.

Key Findings

Finding F001 — The Gadus morhua protein is a bona fide 1:1 ortholog of human TBC1D14 with an intact TBC RabGAP domain

A global Needleman–Wunsch alignment of A0A8C5FPT8 against human TBC1D14 (Q9P2M4) yielded 414 identical aligned columns out of 667 (62.1% identity), comfortably within the range expected for a true 1:1 vertebrate ortholog rather than a distant paralog. Both proteins carry a single Rab-GAP TBC domain (UniProt domain 388–596; Pfam PF00566 RabGAP-TBC; InterPro IPR000195), and both are assigned to PANTHER family PTHR47219.

Critically, the dual-finger catalytic residues that define a functional RabGAP are conserved and positionally aligned between fish and human:

The N-terminal architecture is likewise conserved (disordered regions plus a coiled-coil at ~315–357). There is no signal peptide and no annotated transmembrane segment; the maximum Kyte–Doolittle hydropathy across the whole sequence is only 1.86, below the threshold that would indicate a sustained membrane-spanning helix. The protein is therefore a peripheral/cytosolic protein that associates with membranes through protein–protein and Rab interactions, not an integral membrane protein — exactly what is expected of a RabGAP regulator rather than a resident structural component of an organelle membrane.

The curation consequence is important: the orthology is strong enough that experimental and phylogenetic annotations on human TBC1D14 are propagable to the fish protein by ISO/IBA. This is precisely what converts the autophagosome prediction from "fabricated" to "defensible inference."

Finding F002 — TBC1D14 localizes to recycling endosomes/Golgi and negatively regulates autophagy; it is not an autophagosome structural protein

The primary mammalian literature paints a consistent, mechanistically detailed picture. In the founding study (PMID: 22613832), TBC1D14 was identified as a putative RabGAP that "colocalizes and interacts with the autophagy kinase ULK1" and, when overexpressed, "tubulates ULK1-positive recycling endosomes (REs), impairing their function and inhibiting autophagosome formation." The same study reported that "amino acid starvation causes TBC1D14 to relocalize from REs to the Golgi complex, whereas TfnR and Tfn localize to forming autophagosomes" — meaning that upon autophagy induction the protein itself moves to the Golgi, while transferrin-receptor membrane is delivered onward. This is direct evidence against stable autophagosome residence: the protein's steady-state compartment is the recycling endosome and its starvation-response compartment is the Golgi.

The follow-up mechanistic study (PMID: 26711178) described "TBC1D14, as a negative regulator of autophagy that controls delivery of membranes from RAB11-positive recycling endosomes to forming autophagosomes," acting via the TRAPP complex and ATG9 traffic. A systematic overexpression screen (PMID: 22874560) placed TBC1D14 among TBC-domain proteins that inhibit autophagy when overexpressed, acting "at an early stage during autophagosome formation" by regulating recycling-endosomal traffic — with ATG9 and ULK1 localizing to transferrin-receptor-positive recycling endosomes that are tubulated by excess TBC1D14, and RE membrane subsequently incorporated into newly forming autophagosomes.

Together, these establish TBC1D14 as a traffic regulator operating upstream of and around the forming autophagosome, not a component embedded in the mature organelle. This validates the biological core of the seed hypothesis: the protein is a regulator, and any autophagosome association is transient and delivery-related.

Finding F003 — CORRECTION: autophagosome (GO:0005776) is a curated experimental + phylogenetic GO annotation, not merely a ProtNLM2 artifact

The pivotal correction to an initial "refuted" reading came from directly querying curated GO annotations (QuickGO) for human TBC1D14 (Q9P2M4). GO:0005776 (autophagosome) is present with two independent evidence lines:

1. ECO:0000314 (IDA, Inferred from Direct Assay) from PMID: 22613832, assigned by expert curators at BHF-UCL.
2. ECO:0000318 (IBA, phylogenetic) from GO_Central, PANTHER ancestral node PTN000538677 (GO_REF:0000033).

The same protein is curated to a broader localization set that puts autophagosome in its proper context:

The Gadus morhua ortholog A0A8C5FPT8 currently carries only broad electronic CC terms (vacuole GO:0005773, cytoplasmic vesicle GO:0031410; ARBA) and belongs to PANTHER PTHR47219, making it eligible for the same IBA propagation that gave human TBC1D14 its autophagosome annotation. From a curation-provenance standpoint, GO:0005776 for the fish protein is therefore a legitimate orthology-based inference, not a hallucinated label — even though it represents a secondary and condition-dependent localization.

A supplementary sequence check strengthens the "regulator, not autophagosome-membrane component" reading: a LIR/AIM motif scan ([WFY]xx[LIV]) returned 13 core matches in the fish sequence and 12 in the human sequence — at chance level (~16 expected across ~700 aa) — and no functional LIR is reported for TBC1D14. This is consistent with a peripheral regulator that is not stably anchored to autophagosomal membranes via an ATG8-binding motif (weak evidence, offered only as a supporting caveat).

Mechanistic Model / Interpretation

The synthesis reconciles two facts that at first appear contradictory: (a) TBC1D14 is a regulator, not a structural autophagosome protein, yet (b) autophagosome (GO:0005776) is a genuinely curated CC annotation.

   NUTRIENT-RICH (steady state)                STARVATION (autophagy induction)
   ---------------------------                 --------------------------------
Rab11(+) RECYCLING ENDOSOME                 GOLGI COMPLEX
[PRIMARY localization]                      [TBC1D14 relocalizes here]
|                                            |
TBC1D14 --- ULK1 (interaction)              TBC1D14 leaves REs
RabGAP activity; tubulates REs                       |
when overexpressed                          TfnR / Tfn membrane -----> FORMING
|                                    (delivered onward)        AUTOPHAGOSOME
NEGATIVELY REGULATES                                                   (GO:0005776)
autophagosome formation                     TBC1D14 transiently/peripherally
(GO:0010507, GO:2000785)                    associated during membrane delivery
                                    -> basis of the IDA CC annotation

The model runs as follows. In resting (nutrient-rich) cells, TBC1D14 resides at the recycling endosome, holding ULK1-positive REs in check and thereby restraining autophagy initiation — its negative-regulator BP role. Its GAP domain, with intact Arg- and Gln-fingers (F001), targets Rab GTPases governing this recycling traffic. Upon amino-acid starvation, TBC1D14 relocalizes to the Golgi, releasing recycling-endosomal (transferrin-receptor-positive) membrane that is routed via the TRAPP complex and ATG9 machinery toward the forming autophagosome. The autophagosome CC annotation captures a transient, membrane-delivery-associated presence at/near the forming autophagosome observed experimentally, rather than stable structural residence.

For curation, the correct representation is a compound CC picture led by recycling endosome, with autophagosome as a secondary, mechanistically justified term, tied together by BP terms that describe what the protein actually does (negative regulation of autophagy; regulation of autophagosome assembly) and by an MF term for its RabGAP activity. Annotating autophagosome alone would be a category error; omitting it entirely would discard a curated experimental observation.

Evidence Base

How the literature bears on the verdict. The three primary/mechanistic papers (PMID: 22613832, PMID: 26711178, PMID: 22874560) unanimously frame TBC1D14 as a recycling-endosome-based regulator of autophagy. None describes it as an autophagosome structural protein. That is why the hypothesis is not fully supported. Yet the same founding paper (PMID: 22613832) is the documented source of a curated IDA autophagosome annotation, and that annotation plus its phylogenetic IBA companion are why the prediction is not refuted either.

Limitations and Knowledge Gaps

1. No experimental data in Gadus morhua. Every localization and functional result is mammalian. What was checked: QuickGO for A0A8C5FPT8 (only broad ARBA CC terms) and PubMed. Why it matters: the fish autophagosome call is transferred, not measured. Resolver: heterologous or fish-cell co-localization of tagged tbc1d14 with LC3/ATG8 versus Rab11/transferrin receptor.
2. Strength of the autophagosome IDA. The precise figure/condition in PMID: 22613832 underpinning GO:0005776 was not machine-verifiable here (abstract-level access; abstract emphasizes RE/Golgi). Why it matters: it determines whether the annotation reflects stable residence or transient colocalization. Resolver: a curator reading the full text and its imaging figures.
3. Endogenous vs. overexpression localization. Several phenotypes derive from overexpression, which can exaggerate RE tubulation and autophagy inhibition. Resolver: endogenous co-localization under fed/starved conditions.
4. Primary recycling-endosome source not re-read. GO:0055037 IDA cites PMID: 17562788; taken here from QuickGO orientation. Resolver: read that paper directly.
5. Rab substrate of the GAP domain unresolved. Even in human, TBC1D14 binds active Rab11 but is not established as its GAP; this affects MF specificity. Resolver: in-vitro RabGAP assays across candidate Rabs.
6. ProtNLM2 raw output not inspected. The confidence score and whether recycling-endosome or RabGAP terms were also emitted are unknown. Resolver: inspect the full prediction ranking.

Conflicts and Alternatives

• Seed framing vs. database record. The seed's "misassignment/hallucination" framing is directly contradicted by the curated IDA + IBA autophagosome annotations on the human ortholog — the central correction of this analysis. The tension is real but resolves toward "less precise," not "wrong."
• Internal investigation conflict. Iteration 1 concluded "refuted" from the mechanistic literature; Iteration 2 overturned this on discovering the curated annotation; Iteration 3 settled on "partially supported." The curator inherits this reconciliation.
• Precision, not fabrication. The mechanistic literature centers TBC1D14 on the recycling endosome/Golgi as a regulator; autophagosome is a co-annotation. The risk is over-emphasis of a secondary CC, not invention of a false one.
• Provenance concentration. The autophagosome IDA rests on a single paper and curation group (PMID: 22613832, BHF-UCL), whose abstract stresses RE/Golgi — so the autophagosome call depends on full-text figures.
• Paralog confusion (low risk). The 62.1% identity, shared PANTHER node (PTHR47219), conserved gene name, and conserved N-terminal coiled-coil argue for a specific 1:1 TBC1D14 ortholog, not a mis-mapped paralog.
• Organism-specific difference (unresolved). All experimental data are mammalian; the teleost transfer rests entirely on orthology.

Discriminating Tests

1. Endogenous co-localization panel (most decisive): immunofluorescence of endogenous TBC1D14 versus Rab11/transferrin receptor (recycling endosome), GM130 (Golgi), and LC3/ATG8 (autophagosome) in fed versus starved cells. Expected: strong RE overlap at steady state; Golgi shift on starvation; only transient/partial LC3 overlap — confirming autophagosome as secondary.
2. Purified-autophagosome proteomics: is TBC1D14 enriched relative to core ATG8-family proteins, or transient? Distinguishes residence from delivery-associated passage.
3. RabGAP activity assay: in-vitro GAP assay of the Gadus TBC domain against candidate Rab substrates (e.g., Rab11) to confirm MF GO:0005096/GO:0005097 and validate the conserved Arg/Gln fingers.
4. ULK1 interaction test: co-IP of Gadus tbc1d14 with the ULK1 ortholog to confirm the regulatory-complex association behind the BP terms.
5. Orthology/IBA audit: confirm A0A8C5FPT8 sits under PANTHER node PTN000538677 to formally license IBA propagation of GO:0005776, GO:0055037, and GO:2000785.

GO Curation Implications and Proposed Actions (leads — require curator verification)

Overall lead: treat GO:0005776 (autophagosome) as acceptable-by-orthology but non-exclusive — retain it as a secondary CC, and ensure the fish annotation set is not headed by autophagosome. The ProtNLM2 prediction is less precise than existing knowledge, not wrong.

Mechanistic scope reminder for the curator. The immediate molecular activity is RabGAP/Rab-binding; the immediate cellular action is at recycling endosomes with ULK1 and the TRAPP complex; "regulation of autophagosome formation" is a downstream pathway consequence; and autophagosome localization is a secondary, transient CC. Keep these separated in the annotation set.

Candidate references with exact snippets to verify:
- PMID: 22613832: "One of these putative RabGAPs, TBC1D14, colocalizes and interacts with the autophagy kinase ULK1. Overexpressed TBC1D14 tubulates ULK1-positive recycling endosomes (REs), impairing their function and inhibiting autophagosome formation."
- PMID: 22613832: "Amino acid starvation causes TBC1D14 to relocalize from REs to the Golgi complex, whereas TfnR and Tfn localize to forming autophagosomes" — verify the full-text autophagosome-localization figure behind the IDA.
- PMID: 26711178: "TBC1D14, as a negative regulator of autophagy that controls delivery of membranes from RAB11-positive recycling endosomes to forming autophagosomes".

Suggested curator question: should the fish annotation propagate the human IBA set (autophagosome + recycling endosome + regulatory BP + RabGAP MF) rather than remain at the current broad ARBA CC terms?

Provenance (computed this run)

• UniProt A0A8C5FPT8: ~709 aa, "Rab-GAP TBC domain-containing protein," gene tbc1d14; CC absent; MF = GO:0005096, GO:0031267; TBC domain 388–596 (PF00566/IPR000195).
• Needleman–Wunsch vs Q9P2M4: 414/667 identical = 62.1%; conserved Arg-finger IKLDISR (Gad 451 / Hum 466), Gln-finger YVQ (Gad 491 / Hum 506); max Kyte–Doolittle hydropathy 1.86 (no clear TM); no signal peptide.
• QuickGO Q9P2M4 CC: autophagosome (IDA PMID: 22613832 + IBA), recycling endosome (IDA PMID: 17562788 + IBA + Reactome), Golgi (IDA HPA + UniProt-SubCell), cytosol, vacuole, cytoplasmic vesicle; BP: negative regulation of autophagy (GO:0010507), regulation of autophagosome assembly (GO:2000785).
• QuickGO A0A8C5FPT8: only ARBA CC vacuole (GO:0005773) + cytoplasmic vesicle (GO:0031410); PANTHER PTHR47219.
• LIR/AIM motif scan ([WFY]xx[LIV]): 13 (Gadus) / 12 (human) core matches — at chance level (~16 expected in ~700 aa); no functional LIR reported for TBC1D14 (weak supporting caveat only).

Bottom Line

The ProtNLM2 autophagosome prediction for Gadus morhua tbc1d14 is partially supported, not a hallucination: GO:0005776 is a curated experimental (IDA) plus phylogenetic (IBA) annotation on the human ortholog, and the fish protein is a strong 1:1 ortholog (62.1% identity, PTHR47219) eligible for the same inference. However, autophagosome is a secondary, condition-dependent localization. The protein's primary compartment is the Rab11⁺ recycling endosome (and Golgi upon starvation), and its primary role is as a negative regulator of autophagy acting from those compartments. GO:0005776 should be retained only alongside recycling endosome (GO:0055037) and the regulation-of-autophagy BP terms, never as the sole or primary cellular-component annotation.

Artifacts

• OpenScientist final report
• OpenScientist final report