EIF2B5 encodes the catalytic epsilon subunit of eIF2B. GDP release from eIF2 allows its reuse in cytoplasmic translation initiation. Native human factor preparations, human complex reconstitution and disease-associated human epsilon variants support this mechanism; the rat epsilon-alone assay provides a separate, explicitly nonhuman line of catalytic evidence. These preparations should not be described as an isolated-human-epsilon assay (PMID:9139680(https://pubmed.ncbi.nlm.nih.gov/9139680/), PMID:11323413(https://pubmed.ncbi.nlm.nih.gov/11323413/), PMID:15060152(https://pubmed.ncbi.nlm.nih.gov/15060152/), PMID:29599213(https://pubmed.ncbi.nlm.nih.gov/29599213/)).
The official displayed human product is Q13144, 721 residues. No alternative-product block or natural isoform suffix is introduced. The 6K72 deposited epsilon sequence matches this reference, but sequence identity alone does not establish contact geometry or isolated activity. The initially incomplete clone described in PMID:8688466 is a historical experimental limitation, not a natural alternative product.
Cytosol is supported by the target Human Protein Atlas page at enhanced location-specific reliability and by the literal human Reactome Q13144 entity. HPA064370 was recorded in MCF-7, PC-3 and U2OS, and HPA069303 in HaCaT, RT-4 and U2OS. This permits refinement of broad cytoplasm annotations while preserving the original source context. Human epsilon–eGFP nuclear exclusion in BHK cells and positive endogenous mouse nuclear staining are distinct observations; neither implies exclusive localization in every condition (HPA, PMID:12426392(https://pubmed.ncbi.nlm.nih.gov/12426392/), PMID:12388085(https://pubmed.ncbi.nlm.nih.gov/12388085/)).
Heat and glucose responses remain contextual functions. The rat Q64350 heat donor has direct experimental grounding in PMID:9341116; this supports the conserved broad response claim without making epsilon a thermosensor. The unavailable operative heat-specific details in PMID:15723074 and PMID:12499492 remain reference-access limitations. Glucose-stimulated whole-complex activity in rat pancreatic islets is likewise not a human glucose-receptor assay (PMID:8567668(https://pubmed.ncbi.nlm.nih.gov/8567668/)).
Positive clinical evidence is retained without equating tissue damage with a direct developmental step. The 93 isolated premature-ovarian-failure subjects in PMID:15507143 were negative for the screened variants; the positive ovarioleukodystrophy cases are in PMID:12707859. Clinical myelin pathology and oligodendrocyte density do not alone demonstrate epsilon-mediated sheath formation or lineage progression. Human glial-progenitor RNAi and GFAP induction support an astrocyte context, with the inaccessible operative details kept explicit (PMID:15217090(https://pubmed.ncbi.nlm.nih.gov/15217090/), PMID:15723074(https://pubmed.ncbi.nlm.nih.gov/15723074/)).
The rat epsilon phosphorylation experiments link inhibition of translation-supporting activity to cell death. Resistant mutants protect Rat1/PC12 cells under the tested perturbations; this does not establish an active wild-type epsilon proapoptotic step. The positive-apoptosis annotation is therefore over-specific, rather than refuted solely by a lack of wild-type-overexpression effect (PMID:11756553(https://pubmed.ncbi.nlm.nih.gov/11756553/)).
Binding evidence is assessed by its demonstrated functional information. The engineered human epsilon peptide/GSK3B record is supplemented by full-length and truncated rat epsilon/GSK3 docking experiments. The STX11, DYNC1H1 and KLK6 generic binding assertions are retained as non-core positive interactions, without promoting partner class into an independently established functional role. STX11 has matching binary-screen and cellular-association records; DYNC1H1/KLK6 have literal curated positive pairs despite unavailable complete operative Methods. Gamma association in the two-subunit expression context remains distinct from delta recovery with an assembled complex. Reference sequence checksums are not experimental insert sequences.
Three source assertions remain unresolved: both myelination rows and T-cell-receptor signaling. The broad ER-stress and peptide-hormone responses are retained as non-core on independent mammalian evidence, while their original patient/donor experiment limitations remain explicit. Imported identities and qualifiers are preserved. The dispositions are 25 ACCEPT, 9 MODIFY, 17 KEEP_AS_NON_CORE, 3 UNDECIDED and 4 MARK_AS_OVER_ANNOTATED, with no new annotation or publication quotation. Positive screen interactions remain non-core without unsupported partner-class refinements. The source guide distinguishes original-text access, abstract-only findings and database evidence.
See source evidence, original locators and the portable results. The computation reproduces literal records and joins; it does not replace the biological assessment.
The glucose-response wording is corrected against the original abstract of PMID:8567668(https://pubmed.ncbi.nlm.nih.gov/8567668/), cached lines 48–52: glucose stimulated eIF2B exchange activity, whereas mannoheptulose did not activate it. The abstract does not establish inhibition of the glucose response. Both contextual glucose-response decisions remain non-core. The annotation reason, propagation comments, reference assessment and original-locator finding now agree on this boundary.
The ovarian-follicle annotation retains its process-scope MAO judgment. A separate probable source-reference query asks UniProt to check which primary evidence was intended for the imported IMP/PMID:15507143 assertion. The paper's Results are a negative screen for selected variants; its Discussion cites earlier positive cases in PMID:12707859 (reference 9). Neither a replacement reference nor absence of ovarian disease involvement is asserted. The query is recorded as CHECK and has not been reported upstream. The relevant original Results and Discussion are at cached lines 94–108.
The PMID:14566705 myelination row remains undecided because clinical/MRI pathology and Table 1 do not resolve whether epsilon contributes to myelin formation beyond supplying translation in affected cells. That uncertainty is independent of the other myelination row's access limitation. Reading a full clinical study does not itself settle a more specific process attribution; positive clinical evidence is retained.
An explicit reference to the unchanged UniProt record now records FUNCTION, SUBUNIT, ACTIVITY REGULATION, INTERACTION and PTM coverage. The source guide separates experimental evidence codes from the Ser-544/DYRK2 by-similarity statement. This corroboration does not relabel engineered peptides as native-protein assays.
No annotation actions, imported source identities or natural-product assertions change. Existing supported generic-binding decisions and source-specific uncertainty remain intact. This follow-up adds precise source locators and reference-only citations, with no new publication quotations or modified source caches, under the campaign's standing instructions.