AIGR Deep Research Report: AP3M2 (P53677) — "Clathrin and Golgi Contexts" Function-Assignment Hypothesis OpenScientist openscientist-autonomous 6 artifacts 2026-09-21T03:44:12.041910

AIGR Deep Research Report: AP3M2 (P53677) — "Clathrin and Golgi Contexts" Function-Assignment Hypothesis

Target gene: AP3M2 (mu3B AP-3 subunit), Homo sapiens (NCBITaxon:9606)
UniProt: P53677
Focus type: function_assignment
Hypothesis slug: clathrin-and-golgi-contexts
Source: genes/human/AP3M2/AP3M2-ai-review.yaml (free-text)


Summary

The seed hypothesis bundles six distinct claims about human AP3M2/mu3B: (1) trans-Golgi localization, (2) Golgi-to-vacuole/lysosome trafficking, (3) clathrin-adaptor-complex membership, (4) clathrin-cargo adaptor activity, (5) AP-3-dependent clathrin cargo loading, and (6) AP-3-complex binding. The research objective is to evaluate whether the gene product directly holds each of these functions, separating genuine mu3B activity from paralog over-annotation, organism-specific attributes, and clathrin claims that are propagated by phylogenetic inference rather than experiment.

Overall verdict: partially supported. When each claim is evaluated separately, three cluster as supported (retain), three as over-annotations (remove/generalize/qualify), and localization as real-but-organism-weighted. AP3M2/mu3B is a genuine constitutive AP-3 medium subunit whose directly assayed molecular function is YXXΦ tyrosine-based sorting-signal (cargo) recognition (PMID 9748267), acting in AP-3-mediated endosome/Golgi-to-lysosome/vacuole trafficking (PMID 15051738, 11588176, 39705307) and neuronal endosome-derived vesicle budding. Those claims hold. In contrast, the clathrin-specific annotations — clathrin adaptor complex (GO:0030131), clathrin-cargo adaptor activity (GO:0035615), and the "clathrin-coated" qualifier of GO:0035654 — are over-annotations: UniProt's curated record states AP-3 is "not clathrin-associated," GO's own definitions hedge on AP-3/clathrin, and the best modern mechanism data (human 9C5B / PMID 39705307 and yeast PMID 34688652) show an Arf1-driven, self-polymerizing, clathrin-free coat.

Two structural caveats are decisive for the curator. First, the clathrin-cargo adaptor MF term (GO:0035615) has experimental (IDA) support in this family only for AP2M1; on AP3M2 — and on AP3M1, AP1M1, and even the clathrin-independent AP4M1 — it is IBA-only phylogenetic carry-over across the PTHR10529 family, the textbook signature of paralog over-annotation. Second, the 9C5B cryo-EM structure invoked by the seed is the paralog AP3M1 (Q9Y2T2), not AP3M2; the two mu paralogs are 84.2% identical, which supports transfer of cargo-contact residues but creates no clathrin evidence for mu3B. trans-Golgi localization is real but largely a yeast-derived attribute; the demonstrated human/neuronal mu3B site is the tubular sorting endosome.


Executive Judgment

Verdict: Partially supported — with specific clathrin terms flagged for removal or qualification.

Evaluated claim-by-claim:

  1. Supported (retain): AP-3-complex membership; the directly assayed cargo-selection activity (YXXΦ tyrosine-based sorting-signal recognition); and AP-3-mediated Golgi/endosome-to-vacuole/lysosome trafficking.
  2. Over-annotated (remove / generalize / add "NOT clathrin"): clathrin adaptor complex (GO:0030131), clathrin-cargo adaptor activity (GO:0035615), and the "clathrin-coated" flavor of GO:0035654.
  3. Real but organism-weighted (qualify): trans-Golgi localization — genuine but yeast-weighted; the demonstrated human/neuronal mu3B site is the tubular sorting endosome en route to lysosomes/lysosome-related organelles (LROs).

Most important caveat: the strongest "pro-clathrin" primary evidence (PMID 9545220) is an in-vitro beta3-appendage/clathrin binding result on the beta3 subunit — not mu3B — explicitly hedged ("AP-3 function in protein sorting may depend on clathrin"). Absence of a human-specific mu3B assay is not itself refutation, but the affirmative clathrin annotations are not carried by any AP3M2-specific experimental evidence.


Key Findings

Finding 1 — The clathrin-flavored GO terms conflict with the curated "AP-3 is not clathrin-associated" statement and with GO's own hedged definitions

UniProt P53677's curated FUNCTION describes AP-3 as "an adaptor-related complex which is not clathrin-associated." This is not an oversight — GO's own definitions are explicitly hedged for AP-3:

The only primary evidence pulling toward a clathrin link is (a) in-vitro beta3-appendage/clathrin binding (PMID 9545220), hedged as "AP-3 function in protein sorting may depend on clathrin," and (b) partial colocalization (PMID 15051738): "AP-3 colocalizes with clathrin, but to a lesser extent than does AP-1." Both are weak and indirect, and neither concerns mu3B. This motivates flagging the clathrin CC/MF terms as over-annotations.

Finding 2 — Human AP-3 (including mu3B) acts principally at tubular sorting endosomes en route to lysosomes/LROs; "trans-Golgi" is largely a yeast-derived attribute

Immuno-EM in human cells (PMID 15051738) localizes AP-3 to budding profiles from a tubular endosomal compartment, proposing "AP-3 defines a novel pathway by which lysosomal membrane proteins are transported from tubular sorting endosomes to lysosomes." The neuronal AP-3 form — which uses the mu3B/beta3B subunits — produces synaptic vesicles from endosomes in vitro (PMID 11588176: "only the neuronal form of AP-3 can produce synaptic vesicles from endosomes in vitro").

By contrast, the explicit trans-Golgi cargo-collection role is demonstrated in yeast AP-3 reconstitution (PMID 34688652), and the recent human AP-3 cryo-EM mechanism paper (PMID 39705307) describes cargo sorting "from endosomes to lysosomes" via an Arf1-driven, self-polymerizing coat without invoking clathrin. UniProt still annotates Golgi apparatus (IEA) and notes AP-3 is "associated with the Golgi region as well as more peripheral structures." So trans-Golgi localization is real but should be qualified as organism-weighted; the demonstrated human/neuronal mu3B site is the tubular sorting endosome. The founding AP-3 paper (PMID 8666661) does place the complex at "non-clathrin-coated buds and vesicles associated with the TGN" — supporting a TGN role but explicitly as a non-clathrin coat.

Finding 3 — The directly assayed molecular activity of mu3B/AP3M2 is tyrosine-based (YXXΦ) sorting-signal recognition, not clathrin bridging

Ohno et al. (PMID 9748267) directly assayed mu3A and mu3B (AP3M1/AP3M2) in a combinatorial yeast two-hybrid screen and showed each medium subunit binds distinct-but-overlapping sets of YXXΦ tyrosine-based sorting signals ("each medium subunit favored specific sets of residues at the X and O positions"). This is the direct cargo-recognition function of the mu subunit and the correct, informative molecular-function description for AP3M2 — cargo/signal binding, not clathrin binding.

In-vivo cross-linking/MS of AP-3 (PMID 19010779) co-isolated AP-3 with BLOC-1, BLOC-2, HOPS, clathrin, and PI4KIIalpha — documenting AP-3 subunit/complex associations. Because mu3B is a constitutive component of the AP-3 heterotetramer (UniProt SUBUNIT), the MF term GO:0035651 "AP-3 adaptor complex binding" describes subunit-within-complex association rather than binding of AP-3 as an external ligand. The seed correctly cautions against presuming the GO binding definition requires an external ligand; mu3B is part of AP-3, so "AP-3 complex binding" is best interpreted as structural membership.

Finding 4 — The 9C5B human AP-3 structure uses mu3A (AP3M1), not mu3B (AP3M2); the paralogs are 84% identical

RCSB PDB 9C5B (from PMID 39705307) polymer entities map to Arf1 (P84077), AP3D1 delta (O14617), AP3M1 mu-1 (Q9Y2T2), AP3S1 sigma (Q92572), LAMP1 cargo (P11279), and AP3B1 beta (O00203) — i.e., the medium subunit in the structure is AP3M1, NOT AP3M2/P53677. A pairwise Needleman–Wunsch alignment (P53677 vs Q9Y2T2, both 418 aa, ungapped) gives 352/418 = 84.2% identity. At AP3M2 residue 405 the alignment shows Ile opposite AP3M1 Val (context m2 'KPFKGIKYMTK' vs m1 'KPFKGVKYVTK'), a conservative V→I substitution matching the seed's claim. Critically, the structure contains a YXXΦ cargo (LAMP1) and Arf1 but no clathrin — reinforcing that the modern mechanistic picture of AP-3 is a clathrin-free, Arf1-nucleated coat, and that any cargo-contact-residue transfer from mu3A to mu3B supports cargo recognition, not a clathrin role.

Finding 5 — Clathrin association of AP-3 is genuinely mixed, not established

The AP-3 discovery paper (Simpson et al., PMID 8666661) found "p47 is not enriched in preparations of clathrin-coated vesicles" and that the complex "is a component of a novel type of coat" at "non-clathrin-coated buds and vesicles associated with the TGN." Conversely, in melanocytes Theos et al. (PMID 16162817) report that "AP-3 and AP-1 localize... primarily to clathrin-coated buds on tubular early endosomes" and recognize the tyrosinase dileucine signal, functioning partially redundantly with AP-1. In HPS2 (AP-3-deficient) cells, "blockage of the clathrin-mediated endocytic pathway... did not reverse the inhibited virus assembly" (PMID 22875976), indicating AP-3 function is separable from clathrin-mediated endocytosis. Together with UniProt's curated "not clathrin-associated" statement and the clathrin-free human/yeast AP-3 structures, the balance of evidence is that clathrin association is cell-type/condition-specific at best, not a defining, constitutive property warranting a core GO term.

Finding 6 — GO:0035615 (clathrin-cargo adaptor activity) has experimental support only for AP2M1 and is IBA-only carry-over on AP3M2

A QuickGO annotation query for GO:0035615 across human AP mu subunits shows the term's evidence base is decisive:

Gene UniProt GO:0035615 evidence codes
AP2M1 Q96CW1 IDA (ECO:0000314) + TAS (ECO:0000304) + IBA
AP1M1 Q9BXS5 IBA only (ECO:0000318)
AP3M1 Q9Y2T2 IBA only
AP3M2 P53677 IBA only
AP4M1 O00189 IBA only

Only AP2M1 — the genuine clathrin-mediated-endocytosis adaptor — has experimental (IDA) grounding for GO:0035615. AP4M1 carries the term by IBA despite AP-4 being an established clathrin-INDEPENDENT complex (Hirst 2011, PMID 22022230; Barois 2005, PMID 15377281). This demonstrates that GO:0035615 is propagated across the PTHR10529 family by phylogenetic inference irrespective of actual clathrin dependence — the classic signature of paralog/family over-annotation. On AP3M2 the term is therefore an AP-2-anchored IBA carry-over, not AP-3-specific evidence.


Mechanistic Model / Interpretation

The AP-3 complex is a heterotetramer (delta/AP3D1, beta3/AP3B1 or beta3B/AP3B2, mu3/AP3M1 or mu3B/AP3M2, sigma3/AP3S1). Its subunit-level division of labor is well established:

             AP-3 heterotetramer
   ┌──────────────┬──────────────┬──────────────┬──────────────┐
   │  delta        │  beta3/3B     │  mu3/mu3B     │  sigma3       │
   │ (AP3D1)       │ (AP3B1/AP3B2) │ (AP3M1/AP3M2) │ (AP3S1)       │
   │ dileucine     │ appendage:    │  YXXΦ tyrosine │ small-subunit │
   │ signal / SNARE│ in-vitro      │  SORTING-SIGNAL│ structural    │
   │ (VAMP7)       │ clathrin bind │  RECOGNITION   │               │
   └──────────────┴──────────────┴──────────────┴──────────────┘
    │                           │
    │ recruited by Arf1-GTP + PI-lipids (BFA-sensitive)
    ▼                           ▼
   Tubular SORTING ENDOSOME  ──►  Lysosome / LRO (melanosome, synaptic vesicle)
(major human/neuronal site)
   trans-Golgi network  ──►  vacuole   (major yeast site; human = "Golgi region")

   COAT = Arf1-driven, self-polymerizing, CLATHRIN-FREE (9C5B; yeast recon.)
   Clathrin colocalization = partial, cell-type-specific (melanocytes), NOT core

Where the seed's claims land on this model:

Seed claim Verdict Basis
AP-3-complex membership Supported mu3B is a constitutive subunit (UniProt SUBUNIT)
Cargo-adaptor / cargo selection Supported (as YXXΦ recognition) Direct assay PMID 9748267
Golgi-to-vacuole/lysosome trafficking Supported (endosome-to-lysosome in human) PMID 15051738, 11588176, 39705307
trans-Golgi localization Qualified — real but yeast-weighted PMID 34688652 (yeast), 8666661 (TGN, non-clathrin)
Clathrin-adaptor-complex membership (CC) Over-annotated UniProt "not clathrin-associated"; clathrin-free structures
Clathrin-cargo adaptor activity (MF, GO:0035615) Over-annotated (IBA carry-over) IDA only on AP2M1; IBA on AP3M2
AP-3-dependent clathrin cargo loading Unresolved / not AP-3-specific Only in-vitro beta3 binding (PMID 9545220), hedged

The key mechanistic distinction the seed asks for — general cargo selection vs. specifically clathrin-dependent endocytic assembly — resolves cleanly: mu3B performs general (tyrosine-signal) cargo selection within an Arf1-nucleated AP-3 coat, and there is no direct evidence that it participates in clathrin-dependent endocytic vesicle assembly (GO:0035615's actual definition, and AP-2's job).


Evidence Base / Evidence Matrix

Citation Evidence type Supports / Refutes / Qualifies / Competing Claim tested Key finding Context Confidence & limitations
PMID 9748267 Direct assay (Y2H) Supports cargo selection; refutes clathrin-bridging as the mu activity mu3B molecular function mu3A/mu3B each recognize distinct/overlapping YXXΦ tyrosine sorting signals Combinatorial yeast two-hybrid, mu3A & mu3B High for cargo recognition; in-vitro binding
PMID 39705307 Structural (cryo-EM) Supports clathrin-free coat; refutes core clathrin role AP-3 coat mechanism Arf1-driven, self-polymerizing coat with LAMP1 cargo; no clathrin Human AP-3 (structure uses AP3M1, not AP3M2) High for mechanism; paralog mismatch (84.2% id)
PMID 34688652 Reconstitution Qualifies trans-Golgi (yeast); supports clathrin-free budding Golgi-to-vacuole cargo collection Yeast AP-3 collects cargo at Golgi without clathrin S. cerevisiae in-vitro High for yeast; not direct human AP3M2 evidence
PMID 15051738 Localization (immuno-EM) Qualifies localization & clathrin AP-3 site in human cells AP-3 buds from tubular sorting endosomes; clathrin colocalization "less than AP-1" Human cells High for endosome route; weak clathrin colocalization
PMID 11588176 Direct assay (in-vitro budding) Supports mu3B trafficking role Neuronal AP-3 (mu3B) function Neuronal AP-3 produces synaptic vesicles from endosomes Neuronal AP-3 form High; endosome-derived, not TGN
PMID 9545220 Interaction (in-vitro) Competing / weakly supports clathrin AP-3–clathrin link beta3 appendage binds clathrin; "may depend on clathrin" In-vitro + colocalization Low; beta3 not mu3B; hedged
PMID 8666661 Biochemistry + localization Refutes clathrin-coat identity; supports TGN AP-3 coat identity p47/mu3 complex is a "novel," non-clathrin coat at TGN buds Rat/neuronal, founding paper High; foundational
PMID 16162817 Localization + function Competing (pro-clathrin in melanocytes) AP-3/clathrin in melanocytes AP-3 at clathrin-coated buds; recognizes dileucine signal; redundant with AP-1 Melanocytes Moderate; cell-type-specific; dileucine (delta), not YXXΦ (mu)
PMID 22875976 Mutant phenotype Refutes clathrin-dependence of AP-3 function AP-3 vs clathrin endocytosis Blocking clathrin endocytosis did not reverse AP-3-deficient phenotype HPS2 (AP3B1-null) human fibroblasts High for separability
PMID 19010779 Interaction (XL-MS, IDA) Supports complex associations AP-3 interactome AP-3 co-isolates with BLOC-1/2, HOPS, clathrin, PI4KIIα Mouse (Q8R2R9) in-vivo XL/MS Moderate; clathrin co-isolation ≠ clathrin coat
PMID 22022230 Primary/review Supports family clathrin-independence AP-4/AP-5 clathrin status AP-4/AP-5 do not associate with clathrin; AP-3 basal complex HeLa Moderate; orientation for carry-over
PMID 15377281 Localization Supports family clathrin-independence context AP-4 clathrin status AP-4 studied as distinct from clathrin machinery MDCK/MelJuSo Moderate
UniProt P53677 (curated) Database Refutes clathrin association AP-3 clathrin status "adaptor-related complex which is not clathrin-associated" Curated record Curated; high weight
QuickGO GO:0035615 Database (computed) Refutes AP3M2 clathrin-cargo term Evidence base of term IDA only on AP2M1; IBA-only on AP3M2/AP3M1/AP1M1/AP4M1 Human AP mu subunits High; direct annotation query

GO Curation Implications

Lead recommendations (require curator verification):

  1. GO:0035615 — clathrin-cargo adaptor activity (MF): Recommend REMOVE or DOWNGRADE on AP3M2. Evidence base is IBA-only phylogenetic carry-over from AP2M1 across PTHR10529; the definition ("bringing together a cargo protein with clathrin… formation of endocytic vesicles") describes AP-2's plasma-membrane endocytic role, not AP-3's endosome/Golgi-to-lysosome sorting. The same term sits by IBA on the clathrin-independent AP4M1, confirming propagation is definitional, not evidential. Replace with a cargo-recognition MF reflecting the direct assay — a "cargo receptor / sorting-signal binding"–style term capturing YXXΦ tyrosine-based signal recognition (PMID 9748267), which is more informative than bare "protein binding."

  2. GO:0030131 — clathrin adaptor complex (CC): Recommend REMOVE or replace with GO:0030123 (AP-3 adaptor complex). AP-3 is not constitutively a clathrin adaptor complex; UniProt curates "not clathrin-associated" and the modern coat is Arf1-driven/clathrin-free. GO:0030123 is the correct, specific CC.

  3. GO:0035654 "clathrin-coated" qualifier: Recommend GENERALIZE / drop the clathrin qualifier. The cargo-loading role is real; the "clathrin-coated" flavor is not supported for AP-3.

  4. AP-3 complex (CC, GO:0030123) and AP-3-mediated Golgi/endosome-to-lysosome/vacuole transport (BP): RETAIN. Well supported; qualify localization to emphasize the tubular sorting endosome → lysosome/LRO route in human cells, with trans-Golgi as an organism-weighted (yeast) attribute.

  5. GO:0035651 — AP-3 adaptor complex binding (MF): RETAIN as subunit-within-complex membership, noting it describes structural membership (mu3B is constitutive), not binding of AP-3 as an external ligand; the mouse IDA (PMID 19010779) is XL-MS of complex associations.

Net: evidence supports MF = cargo/sorting-signal recognition (not clathrin-cargo adaptor), CC = AP-3 adaptor complex (not clathrin adaptor complex), and BP = AP-3-mediated endosome/Golgi-to-lysosome/LRO transport. The clathrin-specific terms should be treated as non-core over-annotations.


Mechanistic Scope

The immediate molecular function under test is mu3B/AP3M2's cargo-recognition activity within the AP-3 heterotetramer. Directly assayed, this is YXXΦ tyrosine-based sorting-signal binding (PMID 9748267) — a sequence-specific protein–protein recognition that selects transmembrane cargo for inclusion into AP-3 transport carriers. This is a genuine direct gene-product activity.

Distinguished from that direct activity:
- Downstream phenotypes: lysosome/LRO biogenesis defects, hypopigmentation (melanosome cargo mis-sorting), HPS2, and impaired HIV-1 assembly are loss-of-function consequences at the pathway/organism level, not mu3B molecular functions.
- Coat assembly / vesicle budding is a complex-level property driven by Arf1 and the beta3/delta subunits; mu3B contributes cargo selection, not coat polymerization.
- Clathrin binding (where reported) is a property of the beta3 appendage in vitro (PMID 9545220), not of mu3B, and is not required for AP-3 function (PMID 22875976).


Conflicts and Alternatives

  1. Paralog confusion (most important): the seed's cited 9C5B structure is AP3M1 (mu3A), not AP3M2 (mu3B). Cargo-contact-residue transfer across the 84.2%-identical paralogs is reasonable for cargo binding but generates no clathrin evidence for mu3B.
  2. Family over-annotation: GO:0035615 is IBA-propagated across PTHR10529 (AP1/AP2/AP3/AP4 mu subunits); only AP2M1 has experimental support. This is the mechanism by which AP3M2 inherited a clathrin term it does not merit.
  3. Organism-specific differences: trans-Golgi cargo collection is best demonstrated in yeast (PMID 34688652); human/neuronal mu3B acts at tubular sorting endosomes (PMID 15051738, 11588176). GO's definitions explicitly note AP-3/clathrin differs "in some organisms."
  4. Cell-type-specific pro-clathrin data: melanocyte imaging (PMID 16162817) places AP-3 at clathrin-coated buds and shows dileucine (delta-subunit) recognition — not the mu3B YXXΦ activity. Real but context-specific and subunit-mismatched.
  5. Interaction ≠ coat: XL-MS co-isolation of clathrin with AP-3 (PMID 19010779) documents proximity/association, not that AP-3 forms a clathrin coat; clathrin blockade does not abolish AP-3 function (PMID 22875976).

Limitations and Knowledge Gaps

Gap What was checked Why it matters What would resolve it
No human AP3M2-specific localization/function assay Human data are AP-3-complex-level or use mu3A structure Curation is for AP3M2 specifically mu3B-tagged immuno-EM / knock-in localization in human cells
trans-Golgi vs endosome primacy in human PMID 15051738 (endosome), 8666661 (TGN, yeast/neuronal) Determines CC localization terms Quantitative colocalization of AP3M2 with TGN vs endosome markers
Direct mu3B–clathrin test absent Only beta3 in-vitro binding (PMID 9545220) Core question of the hypothesis Direct mu3B/clathrin binding assay (though mu is definitionally a cargo-recognition subunit)
PMID 9545220 / 11588176 full text unread Verified snippets; full methods not re-derived Strength of strongest pro-clathrin claim Full-text read of PMID 9545220 methods
PMID 15051738 full text partially unchecked Snippet verified Whether TGN action is excluded or just secondary Full-text read
Functional redundancy mu3A/mu3B Paralogs 84.2% identical, overlapping YXXΦ specificity Whether AP3M2 terms should mirror AP3M1 Isoform-specific knockdown/rescue with cargo panels

Discriminating Tests

  1. Direct mu3B cargo panel vs clathrin binding: extend PMID 9748267-style Y2H or use ITC/SPR with purified mu3B against a YXXΦ cargo panel and against clathrin terminal domain — expect strong YXXΦ binding, no direct clathrin binding.
  2. AP3M2-specific localization: endogenous-tag AP3M2 in human cells (neuronal + non-neuronal) and quantify colocalization with TGN46 (trans-Golgi) vs EEA1/tubular-endosome and LAMP1 markers.
  3. Clathrin-dependence of AP3M2 cargo delivery: knock down clathrin heavy chain and test whether AP3M2-dependent LAMP1/LRO cargo delivery is impaired — prior HPS2 data (PMID 22875976) predict little effect.
  4. Cryo-EM of an AP3M2-containing (neuronal beta3B/mu3B) coat: determine whether the mu3B coat, like the mu3A 9C5B structure, is clathrin-free and Arf1-nucleated.
  5. Cross-family GO audit: confirm via QuickGO/PAINT that GO:0035615 IBA on AP3M2/AP4M1 traces to the AP2M1 experimental annotation, justifying an annotation-review request.

Proposed Follow-up Actions / Curation Leads (require curator verification)

Candidate action changes:
- Remove/downgrade GO:0035615 (clathrin-cargo adaptor activity) on AP3M2 as IBA-only family carry-over from AP2M1; document rationale citing QuickGO evidence codes.
- Replace clathrin adaptor complex (GO:0030131) CC with AP-3 adaptor complex (GO:0030123).
- Generalize GO:0035654 by dropping the "clathrin-coated" qualifier for AP-3.
- Retain AP-3 complex membership, AP-3-mediated Golgi/endosome-to-lysosome/vacuole transport (BP), and GO:0035651 (AP-3 adaptor complex binding, as structural membership).
- Add/strengthen an informative MF for tyrosine-based (YXXΦ) sorting-signal / cargo recognition grounded in PMID 9748267 (avoid bare "protein binding").
- Qualify trans-Golgi CC as organism-weighted; emphasize tubular sorting endosome → lysosome/LRO in human.

Candidate references with exact snippets to verify:
- PMID 9748267 — "interact with the medium (mu) subunits of clathrin adaptor (AP) complexes" (direct mu3B YXXΦ assay).
- PMID 9545220 — "AP-3 function in protein sorting may depend on clathrin"; "AP-3 is involved in signal-mediated protein sorting to endosomal-lysosomal organelles."
- PMID 15051738 — "AP-3 colocalizes with clathrin, but to a lesser extent than does AP-1."
- PMID 8666661 — "p47 is not enriched in preparations of clathrin-coated vesicles"; "non-clathrin-coated buds and vesicles associated with the TGN."
- PMID 22875976 — "Blockage of the clathrin-mediated endocytic pathway... did not reverse the inhibited virus assembly."
- UniProt P53677 FUNCTION — "adaptor-related complex which is not clathrin-associated."

Suggested curator questions:
- Does any AP3M2 clathrin term trace to experimental evidence, or only IBA/ISS from AP2M1?
- Should trans-Golgi CC be retained for the human protein given endosome-primacy data?
- Is GO:0035651 best interpreted as structural membership rather than external-ligand binding?

Suggested experiments: the five Discriminating Tests above, prioritizing (1) direct mu3B cargo-vs-clathrin binding and (2) AP3M2-specific human localization.


Bottom Line

AP3M2/mu3B is a bona fide constitutive AP-3 medium subunit whose directly assayed function is YXXΦ tyrosine-based cargo-signal recognition, acting in AP-3-mediated endosome/Golgi-to-lysosome/vacuole/LRO trafficking and neuronal endosome-derived vesicle budding. Those claims (AP-3 membership, cargo selection, Golgi-to-vacuole/lysosome transport) are supported. The clathrin-specific annotations — clathrin adaptor complex (GO:0030131), clathrin-cargo adaptor activity (GO:0035615), and the "clathrin-coated" qualifier — are over-annotations: UniProt curates AP-3 as "not clathrin-associated," GO's own definitions hedge, GO:0035615 has experimental support only for AP2M1 (IBA-only on AP3M2 and even on clathrin-independent AP4M1), and the best modern mechanism data (clathrin-free Arf1-driven coat; note 9C5B is the paralog AP3M1) confirm no core clathrin role. trans-Golgi localization is real but yeast-weighted; the demonstrated human/neuronal mu3B site is the tubular sorting endosome.

Artifacts