This review concerns human BMP4 (UniProt P12644). The normal seed contains 211 raw GOA rows representing 209 distinct source assertions, including six NOT assertions, no isoform-specific assertion and no alternative products. All source terms, evidence codes, qualifiers, original references and supporting entities are preserved. Two repeated projected rows differ only in source date; the original TSV remains intact. There are 60 reference entries: 50 PMIDs, seven GO references and three Reactome reactions.
The normal just deep-research-falcon human BMP4 --fallback perplexity-lite attempt failed before research output because both providers could not resolve the pinned deep-research-client package in the isolated offline tool cache. These are manually authored research notes. No provider-generated research file or provider findings are claimed.
All 50 cached Abstract sections and all three Reactome summaries were inspected; 49 PMIDs have substantive abstracts and PMID:18076106 is citation-only. Selected full-text passages were inspected for PMIDs:14749725, 17850284, 18436533, 18548003, 18823971, 20559569, 22529382, 23610558, 33308444 and 9389648. This is not a claim to have read all complete articles, figure images or supplements. The source-reading ledger in the review preparation records each cache and selected passage separately.
Existing canonical publication caches were preserved. In particular, PMID:18548003 has an authenticated full-text archive variant, used for selected Results, while its pre-existing canonical cache is abstract-only. PMIDs:17850284 and 22529382 have existing full-text caches richer than the incoming abstract variants. Full-text flags describe cache availability, not a claim that every relevant experiment appears in the extracted text. PMID:10748465 is indexed as a Letter and has an experimental abstract; its machine document category is not a reason to discard it. The R-NUL Reactome reaction explicitly pairs rat Grem1 with human BMP4 despite its Homo sapiens summary header.
BMP4 is a secreted TGF-beta-superfamily growth factor. Its precursor has an N-terminal signal peptide and is proteolytically processed to a disulfide-linked mature ligand. The ligand binds type I/type II BMP receptor complexes and initiates receptor-mediated SMAD1/5/8 signaling. The receptors phosphorylate signaling proteins, and downstream SMAD complexes perform nuclear regulatory work; BMP4 is neither a kinase nor a DNA- or RNA-processing enzyme. Its extracellular ligand function unifies many context-dependent outcomes in proliferation, survival, migration and differentiation.
The single core-function synthesis records extracellular growth-factor activity in BMP signaling. BMP4's experimentally supported chemoattractant action in periureteral mesenchyme is retained as a specific non-core activity rather than a second universal function. Tissue-development terms remain contextual outcomes rather than separate core molecular activities. No NEW assertion was needed or proposed.
In the engineered reporter experiments of PMID:9389648(https://pubmed.ncbi.nlm.nih.gov/9389648/), BMP4 stimulation does not activate GAL4-SMAD constructs in SMAD4-deficient SW480.7 cells without rescue by full-length SMAD4 or its C domain. The positive evidence is that rescue, not the preceding statement of reporter inactivity. Separate nuclear-localization assays also use activated receptor cotransfection; these requirements remain explicit in the corresponding annotation reasons.
Both inspected production GO-CAMs, 66e382fb00001232 and 646ff70100002308, place BMP4 cytokine activity in the extracellular region and BMP signaling. They distinguish BMP4 from receptor catalytic activities and downstream SMAD transcription-factor activities. The source models and index were not changed. PAINT assertions were treated as ancestral functional-inheritance judgments; BMP4 appearing among supporting experimental descendants is expected, not circular.
The official AmiGO definitions were checked for extracellular matrix binding, co-receptor binding and growth factor activity. The first two refinements describe BMP4 binding its partners, not the partner's own activity.
PMID:14749725(https://pubmed.ncbi.nlm.nih.gov/14749725/) tests recombinant human BMP4 in mouse C2C12 cells and human mesenchymal stem cells. Myotube inhibition supports negative regulation of myoblast differentiation; the broader striated-muscle-tissue term is refined to that measured endpoint. Osteogenic markers and calcium deposition depend on culture conditions and are enhanced by SB431542. The two distinct mineralization source assertions are retained with their co-treatment context. PMID:18436533(https://pubmed.ncbi.nlm.nih.gov/18436533/) independently reports BMP4-dose-dependent calcium deposits in primary human mesenchymal cultures. BMP2-specific SCUBE3 cartilage experiments were not silently reused as BMP4 cartilage evidence.
PMID:18823971(https://pubmed.ncbi.nlm.nih.gov/18823971/) explicitly uses recombinant human BMP4 in frog animal-cap experiments, supporting the epidermal/fate-choice effects in that assay. PMID:12631064(https://pubmed.ncbi.nlm.nih.gov/12631064/) includes a modified Boyden-chamber assay and bead-induced periureteral mesenchymal accumulation, which directly support chemoattractant activity. Its metanephric and periureteral effects differ. PMID:12141440(https://pubmed.ncbi.nlm.nih.gov/12141440/) directly tests BMP4 inhibition of branching and nephrogenesis in mouse explants, but the abstract does not distinguish every S-shaped/comma-shaped-body endpoint; those exact assignments remain undecided.
BMP4-stimulated endothelial proliferation and migration in PMID:17992660(https://pubmed.ncbi.nlm.nih.gov/17992660/) are compatible with inhibitory pulmonary artery smooth-muscle responses in PMIDs:11502704 and 17600318. The latter's positive-proliferation source row remains UNDECIDED because its precise condition was inaccessible; an abstract-level discrepancy is insufficient for removal. Pancreatic epithelial-colony formation in PMID:11865031(https://pubmed.ncbi.nlm.nih.gov/11865031/) supports a developmental response but does not by itself resolve the separate proliferation endpoint. Likewise, endothelial sprouting is not automatically proof of the precise cell-division term.
PMID:17850284(https://pubmed.ncbi.nlm.nih.gov/17850284/) includes BMP4 immunostaining despite emphasizing SHH in its abstract. The IEP smooth-muscle annotation is retained as expression-based support, with correlation rather than BMP4 perturbation explicit. The accessible PMID:22529382(https://pubmed.ncbi.nlm.nih.gov/22529382/) Discussion reports BMP4-induced TEAD4 nuclear localization but leaves the mechanism unresolved; import-specific regulation remains UNDECIDED. BMP4-dependent downstream protein association in PMID:23610558(https://pubmed.ncbi.nlm.nih.gov/23610558/) is retained without assigning BMP4 a nuclear bridging function.
The many specialized mouse transfers were considered individually. Broad pathway or organ-development evidence was not used to automatically accept exact valve, coronary-vessel, tendon, deltoid-tuberosity, glomerular-capillary or extrinsic-apoptotic-pathway claims. Conversely, an unresolved donor lookup was not used to remove an assertion. Relevant target-specific corroboration supports retention where the same biological process is demonstrated, with exact donor-edge limits stated.
The five cardiac mouse-transfer NOT assertions concern muscular septum, right ventricle, cushion EMT, cushion apoptosis and cardiac jelly. All five preserve their NOT flags, ISS evidence, GO_REF:0000024 and P21275 supporting entity, and remain UNDECIDED. Independent bounded lookups did not verify their exact mouse GO-to-PMID edges.
Selected Results from PMID:18924235(https://pmc.ncbi.nlm.nih.gov/articles/PMC2728547/) are concrete candidate evidence: a Mef2c-AHF conditional model showed no apparent muscular-septum/right-ventricle phenotype, comparable E10.5 cardiac jelly, no initial EMT difference in the reported E9.5 explant assay, and no proximal-outflow TUNEL difference at E10.5/E12.5. These are stage-, tissue- and assay-limited observations, not universal absence of BMP4 function. The EMT result was reported without displayed data. No donor edge was inferred or original source reference replaced merely because this paper is relevant.
The sixth NOT, PMID:18548003(https://pubmed.ncbi.nlm.nih.gov/18548003/), is retained as non-core in the tested miR-21 context. Selected full-text experiments use time courses, alpha-amanitin, promoter reporters and CMV-driven pri-miR-21 to distinguish processing from transcription. SMAD1/5, p68 and DROSHA-associated experiments support increased pri-to-pre processing; DROSHA performs the RNA-processing step. This does not mean BMP4 cannot regulate any miRNA transcription: PMID:22988237(https://pubmed.ncbi.nlm.nih.gov/22988237/) reports repression of the miR-302–367 cluster. Reduced miR-494 abundance in PMID:28124060(https://pubmed.ncbi.nlm.nih.gov/28124060/) alone does not resolve transcription versus processing/stability, so its transcription-specific row remains undecided.
The draft has 32 ACCEPT, 96 KEEP_AS_NON_CORE, seven MODIFY and 74 UNDECIDED decisions, with no NEW or REMOVE assertions. One core function records extracellular growth-factor activity in BMP signaling. Nine short exact cached quotations support selected binding, chemoattraction and ligand-signaling decisions; no source contributes more than 25 quoted words across all occurrences. Existing references and source evidence remain present. UNDECIDED records a precise access or assay limit, not a presumption that the original curator was wrong.
The focused validation retains six expected advisories: two concern supported generic protein-binding rows retained under the task instruction, and four concern different actions for the same term with different source evidence. The latter involve osteoblast differentiation, co-receptor binding, positive DNA-templated transcription and positive RNA-polymerase-II transcription. Independently supported capacity does not make every source-specific experiment verified, so those decisions were not homogenized. The status remains DRAFT.
The two previously uncached contextual papers, PMID:18924235(https://pubmed.ncbi.nlm.nih.gov/18924235/) and PMID:36825302(https://pubmed.ncbi.nlm.nih.gov/36825302/), now have authenticated normal publication caches. Their machine-fetched titles have been added to the structured references, bringing the total to 62 entries: 52 PMIDs, seven GO references and three Reactome reactions. Cache metadata records PMID:18924235: full_text_available=true; PMID:36825302: full_text_available=true. Cache availability does not claim that all text, figures or supplements were read; the selected evidence and assay limits recorded above remain the basis of the judgments.
PMID:18924235 is linked as additional corroboration for the five retained outflow-tract/membranous-septum transfers. The earlier sentence describing it as an uncached manual lead is superseded by this cache closure. Its mouse conditional experiments do not reconstruct the exact GO donor edges or establish a direct human experiment. The five cardiac NOT transfers and the more specific unresolved valve/cushion assignments remain UNDECIDED. PMID:36825302 continues to constrain physiological extrapolation of fibrillin binding in the examined kidney/lung settings; it does not negate the purified-fragment binding result. All 209 actions, six NOT flags, original source assertions, nine quotations and the single core function are unchanged.
This followup changes eight decisions using the source evidence already identified in the initial review. It supersedes the earlier UNDECIDED statements for the three DAF-4 rows and five cardiac NOT rows; the original notes remain as provenance. There are now 32 ACCEPT, 101 KEEP_AS_NON_CORE, 10 MODIFY and 66 UNDECIDED decisions across the same 209 source objects. All six negations and the single core function remain intact.
The cached abstract of PMID:8413626(https://pubmed.ncbi.nlm.nih.gov/8413626/) explicitly establishes human BMP4 binding to DAF-4 expressed in monkey COS cells. NCBI Gene 175781 independently maps G8JY02, P50488 and Q8ITX8 to daf-4 products. Rows 54–56 therefore refine the existing interaction to GO:0070700 BMP receptor binding. The exact historical construct-to-accession mapping remains unresolved, as recorded in the existing question; it does not prevent the molecular-function decision. Every original partner accession is preserved. Only the cached abstract was read for this paper; no full-paper or original construct audit is claimed.
Selected cached Results and Methods from PMID:18924235(https://pubmed.ncbi.nlm.nih.gov/18924235/) support retaining the five cardiac NOT transfers as context-specific non-core assertions. Mef2c-AHF-Cre conditionally removes Bmp4 where its expression overlaps the anterior-heart-field lineage, particularly outflow myocardium; the pharyngeal endoderm retains expression. Each revised row points to this paper without replacing its original GO reference, ISS evidence or mouse P21275 donor.
These findings support the transferred assertions within the stated model, stage and assay limits. They do not establish universal BMP4 dispensability, direct human negative experiments, or an independently reconstructed GO donor evidence edge. The cached primary text was inspected for these observations and their methods; original figure images and supplements were not newly audited. The separate miR-21 transcription NOT row 198 is unchanged. Supported antagonist-binding rows 46 and 57 remain non-core under the standing task instruction; their source evidence, actions and rationales are unchanged.