NOTCH1 (human, P46531) — Axon Guidance (GO:0007411) Function-Assignment Review

Focus: function_assignment · function-hypothesis-go-0007411
Source: genes/human/NOTCH1/NOTCH1-ai-review.yaml
Iteration: 3 of 3 (final)

Executive Judgment

Verdict: Partially supported / weakly supported — retain only as a non-core, phylogenetically-inferred BP annotation with an explicit caveat; do NOT treat as a direct, experimentally-established human function.

The seed hypothesis is correct on two points and unproven on the decisive one:

1. A genuine, guidance-specific Notch mechanism exists — but it has been demonstrated only in Drosophila. Notch controls axon growth and guidance through a noncanonical, Su(H)-independent pathway using the Abl tyrosine kinase, the adaptor Disabled, the GEF Trio, and Rac GTPase (PMID:18062953, 21246649, 29343637). This is mechanistically distinct from SLIT/ROBO and from canonical nuclear Notch, exactly as the hypothesis states, and it is separable from neuronal fate.
2. The "questionable cross-family annotation source" is real and confirmed. The human NOTCH1 → GO:0007411 record is an IBA/PAINT phylogenetic inference (ECO:0000318, GO_REF:0000033, GO_Central), not a human experiment. Its withFrom support set mixes legitimate Notch orthologs (fly Notch FBgn0264089; mouse Notch MGI:1315202/3/5) with human SLIT1/SLIT2/SLIT3 (O75093/O75094/O94813) — a distinct SLIT/ROBO ligand family. So SLIT cross-family contamination materially contributed to the propagation.
3. The decisive gap: there is no direct human/vertebrate axon-pathfinding evidence. The only documented mammalian NOTCH1 neuronal phenotype is inhibition of neurite outgrowth / neurite morphology via canonical CBF1 signaling (PMID:10465425) — a length/morphology effect the hypothesis explicitly asks to exclude from guidance evidence. Mammalian noncanonical Notch does exist in neurons but has been mapped to a different output (synaptic-vesicle protein expression, PMID:27040987). A computed STRING v12 check found no curated interaction between human NOTCH1 and the Abl/Trio/Rac module (the NOTCH1–ABL1 edge is text-mining-driven: tscore 0.596, escore 0.113, dscore 0).

Bottom line for the curator: The hypothesis's principled stance — that a bad annotation source and the absence of a human assay do not by themselves refute a conserved function — is fair. But "not refuted" is not "supported." The affirmative evidence is entirely Drosophila, the human annotation is inference contaminated by a paralogous family, and the accessible human data point away from directional guidance and toward neurite morphology/fate. The term should be retained at most as a non-core, inferred annotation with a caveat, or generalized; it should not be presented as a demonstrated core human function.

Evidence Matrix

Provenance artifacts: NOTCH1_GO0007411_evidence_matrix.csv (this run); QuickGO and STRING API calls executed live in iterations 1–2.

GO Curation Implications (leads — require curator verification)

• The evidence supports a BP term at most, not an MF. Do not downgrade to "protein binding."
• Do not present as a core function. If the review pipeline distinguishes core vs. non-core, mark axon guidance non-core for human NOTCH1.
• If a curator wants to preserve the conserved biology without over-claiming the human protein, an acceptable lead is to keep GO:0007411 only with the IBA evidence code and a note that the phylogenetic support is partly cross-family (SLIT), OR generalize to a less specific neuronal-development term that the mammalian data actually support (e.g., regulation of neuron differentiation / regulation of neuron projection development — curator to verify exact ID).
• Flag the PAINT annotation to GO_Central: the withFrom inclusion of SLIT1/2/3 in a Notch-family propagation is a likely tree/family-grouping artifact worth reporting.

Mechanistic Scope

• Immediate molecular step being tested (fly): transmembrane Notch receptor, after tyrosine phosphorylation and cleavage, physically associates with Disabled and Trio and activates Rac to remodel the growth-cone cytoskeleton — a bona fide guidance-relevant step, independent of Su(H)/CBF1 transcription (PMID:18062953, 21246649, 29343637).
• Downstream / excluded phenotypes: neuronal fate specification (canonical Notch), neurite outgrowth length and branching (mammalian PMID:10465425), survival, and disease phenotypes. These are explicitly outside the guidance claim and, for human, are the only things actually observed.
• Directionality distinction: "axon guidance" (GO:0007411) means directional pathfinding/patterning. Mammalian NOTCH1 data show a scalar effect on neurite length/morphology, not steering — an important curation distinction.

Conflicts and Alternatives

1. Cross-family (paralog/family) confusion: SLIT1/2/3 in the IBA withFrom set means part of the "axon guidance" signal came from the true SLIT/ROBO guidance ligands, not from Notch biology — classic annotation carry-over.
2. Organism-specific difference: The guidance mechanism is demonstrated in Drosophila; the mammalian ortholog's documented neuronal action (neurite inhibition via CBF1) is a different pathway output.
3. Canonical vs. noncanonical: The human annotation does not specify the noncanonical mechanism; a naive reader could conflate it with canonical Notch, which is not what the fly guidance work shows.
4. Interaction-data gap: No curated human NOTCH1–ABL1/TRIO/DAB1/RAC1 interaction; the apparent link is literature co-mention, not experiment.

Knowledge Gaps

Discriminating Tests

1. Growth-cone turning / directional pathfinding assay with human NOTCH1 gain/loss (and a Su(H)/RBPJ-uncoupled allele) to separate guidance from neurite length and fate.
2. Interaction mapping (co-IP, BioID/APEX) of human NOTCH1 with ABL1, TRIO, DAB1, RAC1 in neurons to test whether the noncanonical module is engaged.
3. Structure-function: express a human NOTCH1 carrying the Drosophila Dab-binding/tyrosine determinants vs. wild type; test rescue of a guidance phenotype.
4. Vertebrate in vivo: zebrafish/mouse notch1 conditional knockdown in a defined tract (e.g., commissural or motor axons) scored for pathfinding errors distinct from proliferation/fate.
5. Sequence/orthology: align P46531 ICD to Drosophila Notch guidance determinants; report conservation conservatively.

Curation Leads (require curator verification)

• Action lead: Change GO:0007411 from an unqualified functional claim to an IBA, non-core, caveated annotation — or generalize to a neuronal-development term the mammalian data support. Do not record axon guidance as a demonstrated core human function.
• Reference leads with exact snippets to verify:
• PMID:18062953 — "Notch directs axon growth and guidance in Drosophila via a "non-canonical", i.e. non-Su(H)-mediated, signaling pathway, characterized by association with the adaptor protein, Disabled, and Trio, an accessory factor of the Abl tyrosine kinase" (establishes fly guidance mechanism).
• PMID:21246649 — "it is the crucial Rho GTPase in "noncanonical" Notch/Abl signaling" (Rac effector step).
• PMID:10465425 — "Transfection at two days in vitro with full length Notch1 inhibited neurite outgrowth" (mammalian effect is neurite length, not guidance).
• PMID:27040987 — "a novel Notch signaling is involved in expression of synaptic proteins in postmitotic neurons" (mammalian noncanonical Notch output ≠ guidance).
• Database lead: Report to GO_Central that the NOTCH1 GO:0007411 PAINT withFrom set includes human SLIT1/2/3, indicating cross-family over-propagation.
• Suggested curator questions: (a) Is there any vertebrate in vivo axon-pathfinding phenotype for notch1 not attributable to fate/proliferation? (b) Does the review pipeline want to keep conserved-but-untested functions as IBA-only with caveats?
• Suggested experiments: items 1–3 under Discriminating Tests.

Limitations

• Three-iteration budget; PubMed queries intermittently rate-limited and some returned no hits (may under-sample vertebrate literature).
• Sequence-level conservation of the specific fly guidance determinants was not aligned in this run (flagged as a gap).
• STRING absence of an edge is suggestive, not proof, of no interaction.
• Local repository *-bioinformatics analyses were intentionally withheld and not consulted.

Artifacts

• OpenScientist NOTCH1 GO0007411 evidence matrix
• OpenScientist final report
• OpenScientist final report