Endoglin is an extracellular ligand-binding coreceptor anchored at the plasma membrane. Human receptor-domain and reporter experiments place ENG in TGF-beta receptor systems, while direct BMP9/BMP10 binding and human endothelial signaling studies support a mechanistic role in BMP receptor signaling. The signaling receptors supply kinase activity; ENG does not become a kinase because receptor phosphorylation or SMAD activation depends on it. PMID:12015308(https://pubmed.ncbi.nlm.nih.gov/12015308/), PMID:21737454(https://pubmed.ncbi.nlm.nih.gov/21737454/), PMID:23300529(https://pubmed.ncbi.nlm.nih.gov/23300529/), PMID:28564608(https://pubmed.ncbi.nlm.nih.gov/28564608/)
The orphan region provides the directly resolved BMP9 interface, and the ZP region participates in covalent dimerization. A solved soluble ENG–BMP9 complex does not establish the complete membrane receptor assembly or demonstrate the proposed ligand-handoff step. Earlier truncation-based Cys350 interpretations must remain distinct from the later C350–C382 intramolecular disulfide structure. Soluble engineered ectodomains can sequester ligand and inhibit signaling; this does not reverse the function of membrane ENG. PMID:21737454(https://pubmed.ncbi.nlm.nih.gov/21737454/), PMID:22347366(https://pubmed.ncbi.nlm.nih.gov/22347366/), PMID:28564608(https://pubmed.ncbi.nlm.nih.gov/28564608/)
TGF-beta and activin/BMP7 experiments often examine receptor systems using affinity labeling, chemical crosslinking and immunoprecipitation. They should not be silently represented as isolated binary affinities. Three imported annotations explicitly state contributes_to, and that contribution is evaluated on its own terms. PMID:1326540(https://pubmed.ncbi.nlm.nih.gov/1326540/), PMID:9872992(https://pubmed.ncbi.nlm.nih.gov/9872992/)
The official natural products are P17813-1/Long, 658 residues, and P17813-2/Short, 625 residues. Short replaces long-product residues 619–658 with EYPRPPQ through VSP_004233. Their shared extracellular region does not establish equivalent cytoplasmic signaling. Signal peptide 1–25, processing of the mature protein, soluble shed forms, Fc/His fusions and engineered deletions are not additional natural splice products. Historical cDNAs and constructs remain unassigned to a current product where exact sequences were unavailable. UniProt P17813, PMID:8370410(https://pubmed.ncbi.nlm.nih.gov/8370410/), PMID:18974388(https://pubmed.ncbi.nlm.nih.gov/18974388/)
The porcine donor is P37176. Its historical 643-residue clone is not assumed to be identical to the current 653-residue record. Human ENG expressed in monkey, mouse or rat hosts remains a human construct in a different cellular system; native mouse knockout and human transgene experiments are recorded separately. PMID:8294451(https://pubmed.ncbi.nlm.nih.gov/8294451/), PMID:12015308(https://pubmed.ncbi.nlm.nih.gov/12015308/), PMID:17628518(https://pubmed.ncbi.nlm.nih.gov/17628518/)
All 104 frozen source objects have proposed decisions: 29 ACCEPT, 59 KEEP_AS_NON_CORE, four MODIFY, two MARK_AS_OVER_ANNOTATED and ten UNDECIDED. The generic receptor activity is refined to coreceptor activity, GO:0015026. Eleven supported generic protein-binding rows remain non-core. Three are refined using independently established BMP, type I TGF-beta receptor and type II TGF-beta receptor binding; this follows the campaign exception for evidence-backed specific terms. No new GO annotation or publication quotation is introduced.
The two glycan-recognition assignments from the 44G4-antigen study exceed its inspected experiment. Lectin affinity, galactose competition and glycosidase treatment characterize sugars carried by ENG; the lectin is the recognizing molecule. This preserves the positive glycosylation findings. The separate porcine glycosaminoglycan claim remains unresolved. PMID:3262645(https://pubmed.ncbi.nlm.nih.gov/3262645/), PMID:8294451(https://pubmed.ncbi.nlm.nih.gov/8294451/)
The three PAINT annotations have authentic ancestral IBD matches on the human target path, with a target MSA sequence matching all 658 official residues. A short donor list or the target's occurrence in its own evidence list is not a reason to reject them. The vasculogenesis uncertainty instead concerns the developmental step: two inspected mouse originals preserve early endothelial differentiation and primitive vascular networks while disrupting subsequent remodeling and branching. The human HHT original needed to settle its particular process annotations remains inaccessible. This supports uncertainty about the exact claim, not universal absence of ENG participation or a false ancestral placement. PMID:10348742(https://pubmed.ncbi.nlm.nih.gov/10348742/), PMID:10562296(https://pubmed.ncbi.nlm.nih.gov/10562296/), PMID:7894484(https://pubmed.ncbi.nlm.nih.gov/7894484/)
Five imported NOT annotations remain literal. Four have bounded early-development or gain-of-function evidence; they do not assert a negative in every tissue, developmental stage or ligand context. The negative cushion-transition assertion remains unresolved because its precise original assay was not recovered. The related positive cushion-formation transition term has a different definition and is not mechanically treated as its logical opposite. PMID:17628518(https://pubmed.ncbi.nlm.nih.gov/17628518/), PMID:19703439(https://pubmed.ncbi.nlm.nih.gov/19703439/), PMID:12941632(https://pubmed.ncbi.nlm.nih.gov/12941632/)
Most vascular, smooth-muscle and cardiac developmental outcomes are retained as non-core contexts. Trabeculation may include secondary hypoxia. Outflow septation failure occurred in combined human-ENG overexpression and Smad4 loss in mouse neural crest. Neither creates a new autonomous ENG molecular activity. Precise repression, dorsal-aorta and negative cushion-transition assertions from the inaccessible 2003 original remain uncertain. PMID:17628518(https://pubmed.ncbi.nlm.nih.gov/17628518/), PMID:19703439(https://pubmed.ncbi.nlm.nih.gov/19703439/), PMID:12941632(https://pubmed.ncbi.nlm.nih.gov/12941632/)
The ten unresolved zero-based source indices are 2, 3, 4, 40, 55, 56, 60, 67, 68 and 93. Source dictionaries, five negations, three contribution qualifiers and two natural products are preserved. The 106 raw records project to 105 normal TSV rows and 104 seeded objects; two documented collapse groups retain their original dates, tissues and input extensions in the reproducible archive.
The source guide records each selected publication's actual access, experiment and limits. Separately recovered original text does not rewrite an abstract-only normal cache flag. No complete-paper, figure-pixel, supplementary-file or exact-construct assessment is implied beyond the recorded scope. Provider deep research failed with access/quota responses; these notes are authored review notes, not a fabricated provider research output.
The frozen UniProt SUBUNIT text reverses TGFB2/TGFB3 relative to the inspected 1992 paper. Correct imported TGFB1 and TGFB3 source rows remain intact. This is retained as a record-text question, with no source defect invented on those correct annotations and no upstream message sent. PMID:1326540(https://pubmed.ncbi.nlm.nih.gov/1326540/)
The offline source-context analysis reproduces identities, products, raw source joins, PAINT placement and rule/localization applicability. It does not derive scientific annotation decisions from lookup tables.
The unresolved PAINT vasculogenesis decision additionally records the inspected ancestral node in structured propagation metadata. The prose retains the source-specific distinction between the assessed human glycan source and the unverified porcine source. Structured consistency-exception metadata is not used because the applicable rule does not consume it. The propagation metadata expresses the existing judgment; no action or source field is changed.
Direct BMP9/BMP10 recognition is now represented by a BMP-binding core unit alongside coreceptor activity. The refined BMP9 row retains its original mature-ligand partner. The two receptor-binding refinements retain the unread PMID:15702480 citation and explicitly use independent PMID:12015308 association evidence. Construct-specific affinities and shared extracellular domains do not establish equivalent natural-tail signaling. PMID:21737454(https://doi.org/10.1074/jbc.M111.260133), PMID:22347366(https://doi.org/10.1371/journal.pone.0029948), PMID:28564608(https://doi.org/10.1016/j.celrep.2017.05.011)
The reference-level availability flags were authored during this review's packaging by copying normal-cache availability; they were absent from the seeded review. They are optional reviewer fields, not immutable GOA or publication-cache metadata. Eleven flags are removed where operative original text was recovered separately. Eleven genuinely unavailable-reference flags remain. The source guide retains the exact partial-read and supplement limits, and normal publication caches remain unchanged. Flag omission does not certify a complete-paper or figure-pixel review.
The seven challenged developmental and transcription-regulation annotations remain non-core with positive experimental rationales and explicit attribution limits. Regulatory process terms permit an upstream signaling contribution. The cardiac expression assays, hypoxia evidence and compound neural-crest genotype are preserved; no myocardium-autonomous mechanism, direct ENG transcription machinery or normal human dosage requirement is claimed. The offline source archive remains unchanged, and no publication quotations or NEW annotations are added.