Human AP4E1 Q9UPM8, HGNC:573: 25 source assertions and two alternative products were initialized by the exact Seed9 normal fetch. The initial research entries below precede annotation review; their pending reading tasks are superseded by the later closure and authored-decision sections. All 25 assertions have now been reviewed. The configured normal deep-research attempt failed; its redacted log and result are separate.
PMID10436028 canonical full text was read at targeted cloning, immunofluorescence and Discussion paragraphs. Human epsilon cDNA came from human testis IMAGE1031294 (residues35–455) and a human heart library; assembled1108-aa historical accession AF155156 differs from current1137-aa canonical length and is not assigned to a current isoform without sequence reconciliation. The mouse EST was a probe, not evidence that the epsilon construct was mouse. Original beta4 was mouse; mu4 and sigma4 were human. Rat1-host human beta4-tag imaging does not automatically establish human epsilon localization; endogenous epsilon antibody staining and HeLa contexts must be kept distinct. The paper describes nonclathrin carriers at the TGN and mu4 recognition of tyrosine cargo motifs, not direct epsilon cargo recognition.
PMID26542808 canonical abstract and actual Seed9 candidate abstract were read, and primary identity independently searched. Tepsin has distinct beta4- and epsilon-ear motifs, with the S[AV]F[SA]FLN motif binding epsilon; bivalent accessory recruitment is a plausible source-specific refinement. Full actual Methods and construct species remain to be read. Direct PMC access was a CAPTCHA and three targeted searches gave unrelated results; none supplies evidence.
PMID32073997 canonical record is abstract-only, while the exact Seed9 quarantined candidate has full text. Both actual abstracts and primary identity were read. HOOK1/2 directly engage mu4/AP4M1 in the described complex; whole-AP4/FHF affinity association does not automatically establish direct epsilon-to-motor binding. Full candidate Methods and relevant Results remain to be read separately; canonical cache is preserved.
PMID10066790 canonical and candidate abstracts were read. Canonical PMID10436028 full cache is preserved even though the new candidate is abstract-only. All relevant Reactome and gene annotation evidence still needs systematic adjudication.
The indexed original PMC article supplied targeted GST-fusion, ITC and cell Methods, epsilon-binding Results/Figure6 legend, and cell recruitment/Discussion text. Direct opening still showed a CAPTCHA. Purified epsilon ear840–1137 binds the tepsin515–525 peptide by ITC (Kd2.5±0.5 micromolar); GST and motif mutations provide specificity controls. Y2H separates the epsilon ear from its trunk/hinge. Cell coassociation and TGN recruitment require both tepsin motifs; proposed intercomplex crosslinking remains a model. The depositor reagent record identifies this GST-ear insert as human AP4E1, with silent codon changes. These sections establish a specific accessory interaction; they do not show direct epsilon cargo recognition. No figure-pixel or complete supplement reading is claimed. Canonical PMID26542808 remains abstract-only.
PMID11707398 targeted Results, deletion and switch-mutant experiments and construct Methods were read from the canonical full text. Epsilon cDNA was RT-PCR amplified from HeLa RNA. Y2H and in-vitro translation/coprecipitation compare GTP-locked Q71L and GDP-preferring T31N ARF1, N-terminal truncation and epsilon fragments. Epsilon1–727 binds activated ARF1 whereas the complementary727–1135 fragment does not; switch I/II mutations disrupt association. The epsilon1–138 negative construct still binds sigma4, providing an expression/function control. These results support an ARF1-effector binding role, not GTP hydrolysis or nucleotide exchange by epsilon. The historical fragment numbering is retained without assigning a current isoform.
PMID41565640 identity and title were independently confirmed against PubMed. Canonical full-text targeted plasmid/protein Methods, structural Results/Figures1–2, interface-mutant/liposome Results/Figure4, bead-recruitment Results/Figure5, cellular Results/Figure6 and corresponding assay Methods were read. All four AP-4 subunits and ARF1 were amplified from a human cDNA library. Sf9 is the expression host; bovine albumin is a control reagent, not the epsilon construct. The structural core uses epsilon1–612 and lacks its tepsin-binding ear. Epsilon L132E/L133E perturb ARF1 association, and L133E reduces recruitment to ARF1-decorated liposomes. Streptavidin/ARF1 control surfaces distinguish protein-mediated recruitment from autonomous lipid binding. Synergistic recruitment by ARF1 and an ATG9A peptide is a whole-complex property; cellular rescue uses AP4B1 knockout and beta4 mutants, not an epsilon knockout. Low-resolution structures do not resolve every side chain or bound nucleotide. These targeted sections were read; no complete supplement or figure-pixel inspection is claimed.
The exact Seed9 quarantined full-text candidate for PMID32073997 was read independently of its preserved abstract-only canonical cache. Targeted Results, Figure2/3 legends, Discussion, human DNA-construct and pull-down Methods establish human epsilon/tepsin Y2H as a positive control. Human epsilon TSF affinity purification and beta4 immunoprecipitation establish FHF complex association with negative myrlysin/GFP controls; direct HOOK1/2 binding maps to mu4. Epsilon must not inherit mu4's direct motor-adaptor interaction. Epsilon and tepsin constructs are explicitly human; rodent AP-1/2/3 constructs occur in comparator assays.
Both cached Reactome events R-HSA-5229111 and R-HSA-5229132 were read. Their APP-recognition narrative identifies the cytoplasmic APP tail and mu4; the event's endosome-lumen destination is not an experiment placing epsilon in the lumen. The official endosome-lumen definition denotes the enclosed volume. Actual human biochemical/structural evidence places AP-4 on the cytoplasmic face. PMID20230749's canonical abstract was read, not its full original body; it reports APP cytoplasmic-tail recognition by mu4 and cargo transport to endosomes.
The official cargo-adaptor term is under protein-macromolecule adaptor activity and part of vesicle-mediated transport. AP-4 lacks clathrin; epsilon contributes to the complex's activity rather than independently recognizing the mu4 cargo motif. The current small-GTPase-binding term includes former ADP-ribosylation-factor binding GO:0030306, so an obsolete specific identifier must not be introduced. No AP4E1/Q9UPM8 entry was found in the local GO-CAM index. The two source IBA nodes are preserved; no target phylogeny or ancestral placement has yet been independently reconstructed, and self-evidence in the AP-4-complex IBA is not circularity.
All 25 source assertions and both alternative products are preserved. The broad endosome inference remains UNDECIDED because TGN enrichment does not exclude a transient pool. The endosome-lumen TAS is removed on the distinct cytoplasmic-coat topology argument. Human epsilon-to-tepsin recruitment is refined to an accessory-adaptor function with the epsilon-containing core supplying the other side of the bridge. The two screen-specific pair assertions remain uncertain. One contributed cargo-adaptor core is recorded; no NEW annotation or downstream biological process is added. The direct ARF1-binding evidence is described within this mechanism without relabeling the TEPSIN source rows.
Six original protein-targeting, transport and localization assertions are now ACCEPTed. Their protein-cargo information was lost by the earlier lateral replacement with the already-present vesicle-mediated transport term. The broad assertions correctly describe AP-4 participation, and the existing core retains the transport mechanism.
The four TEPSIN rows are distinguished by what their own experiments support. The two uninspected screen pairs are retained as independently corroborated non-core interactions, with their exact supplementary-table and control limits explicit. The human epsilon–tepsin Y2H positive control in PMID:32073997(https://pubmed.ncbi.nlm.nih.gov/32073997/) likewise retains its observed interaction. Only PMID:26542808(https://pubmed.ncbi.nlm.nih.gov/26542808/) carries the accessory-adaptor refinement: this paper measures motif-dependent binding and recruitment of tepsin to AP-4. Evidence from that paper is not silently assigned to the other three original IPI sources. The parent adaptor term describes accessory recruitment, whereas the inherited cargo-adaptor activity concerns the assembled heterotetramer and mu4 cargo recognition. No new redundant ancestor is proposed.
One NEW molecular function, GO:0031267 small GTPase binding, records direct epsilon–ARF1 recognition. Its original evidence is PMID:11707398(https://pubmed.ncbi.nlm.nih.gov/11707398/), whose targeted interaction Results, DNA-construct Methods and in-vitro binding Methods were reread. Human epsilon was amplified from HeLa RNA; yeast is the Y2H host. The assays compare Q71L/T31N nucleotide-state mutants, epsilon trunk/ear fragments and ARF1 switch-I/II substitutions. In-vitro translated material is not described as purified full-length protein. The epsilon 1–138 negative fragment retains sigma4 binding. PMID:41565640(https://pubmed.ncbi.nlm.nih.gov/41565640/) supplies independent recombinant human AP-4 interface evidence; its beta4-knockout rescue is not relabelled as an epsilon knockout. The term’s current definition and ARF synonym were verified. This specific interaction concerns ARF1, distinct from existing generic TEPSIN partners. No NEW biological process, nucleotide-exchange activity or GTP hydrolysis is proposed.
Official PubMed identities were independently verified for PMID:11707398, PMID:10436028 and PMID:32073997; reference-review flags now reflect the verified identities and the bounded primary reading already documented above. No cache availability flag or primary-source byte changed. Earlier authored decisions are historical and superseded by this section. Independent prospective consultation required the source-specific restrictions adopted here. Final authored validation and consultation are recorded separately.
The two TEPSIN screen records (PMID:26496610 and PMID:33961781) now remain UNDECIDED. Their precise pair entries, reagents and controls have not been inspected. Independent evidence for a real epsilon–tepsin partnership does not settle those source experiments, and the separate recruitment quotation has been removed from these rows. This is an access/scope limit, not a claim that either screen disproves the interaction.
The human Y2H interaction in PMID:32073997, Results and Figure 3 was directly inspected, along with the human-reagent Methods. The actual source states: “Control experiments showed interactions of tepsin with AP-4 ε and AP-4 β4”. This is a positive epsilon–tepsin pair result; the paper's direct HOOK interface is assigned to mu4. The observed pair remains KEEP_AS_NON_CORE. The reviewer requested removal solely because the generic term is uninformative, but the explicit user criteria permit experimental REMOVE for genuinely contradicted functions, not genericity alone. A same-source adaptor bridge was not demonstrated by this Y2H control, so no mechanism is substituted. The exact result is attached through these linked notes because the immutable normal cache is abstract-only.
The PMID:11707398 original Results, Figure 5C and construct Methods were rechecked. In-vitro-translated epsilon trunk material co-precipitates with the activated ARF1 comparator but not the GDP-preferring comparator; deletion and switch-mutant controls delimit the interaction. These assays are not described as a purified full-length reconstitution. The NEW small-GTPase-binding row and its existing core now carry the positive binding-result quotation alongside the separately identified human HeLa-RNA construct provenance. The two duplicate reference findings have been replaced by separate result and provenance findings. The authored NEW receives qualifier: enables; all 25 original source assertions and their qualifiers, both products, the prior NEW function/evidence, and both core functions remain otherwise intact.
This follow-up adds no molecular function, process or localization claim. The actual PMID:26542808 recruitment-based refinement remains unchanged. Published histories and all raw source caches are preserved.
The source review's optional suggestion to restore an ARF1-effector finding is adopted. The actual PMID:11707398 Results, Figure 6 describes loss of epsilon interaction after ARF1 switch-region substitutions, in addition to the separately recorded nucleotide-state comparison and human construct provenance. This distinct finding supports the already recorded ARF1-binding function; it adds no annotation, process, core function or isoform assertion.
The removal threshold used here comes from the user's session-supplied AGENTS.md instructions, not solely from repository CLAUDE.md. That action enum defines REMOVE as an annotation unlikely to be correct from the combined evidence, and the supplied experimental-removal guidance identifies genuinely contradicted functions. The local annotation-reviewer skill recommends a broader uninformativeness-based removal policy. Our interpretation of the user-provided criteria retains the directly inspected human epsilon–tepsin positive-control interaction as KEEP_AS_NON_CORE, outside the core-function synthesis. This is an intentional difference from the local skill recommendation; the current automated reviewer still requests REMOVE. No contradiction of the actual interaction, or same-source adaptor bridge, has been established. The two uninspected screen assertions remain UNDECIDED.
All 25 original source assertions, the existing NEW, both alternative products and both core functions remain unchanged. The existing abstract-only PMID:32073997 cache is preserved; the separate primary-body reading and note-mediated quotation remain explicitly identified above.