EPHB4 encodes an active ephrin-B receptor tyrosine kinase. The displayed human product is a 987-residue membrane receptor with an extracellular ephrin-binding region and a cytoplasmic kinase domain. Human receptor activation by ephrin-B2 is supported by the operative experiments in PMID:27400125(https://pmc.ncbi.nlm.nih.gov/articles/PMC4966301/), while purified human EPHB4 residues 588–987 phosphorylate protein substrate in PMID:34431498(https://pmc.ncbi.nlm.nih.gov/articles/PMC8454701/), Figure 4. The kinase reagent is not a full-length receptor. These complementary assays support the specific transmembrane-ephrin receptor activity, GO:0005005.
EphB4–ephrin-B2 contact is bidirectional. Receptor activation and ephrin-B2 reverse signaling both affect endothelial behavior. In PMID:16424904(https://www.embopress.org/doi/full/10.1038/sj.emboj.7600949), endothelial full-length and artificial kinase-truncated EphB4 change vessel architecture and sprouting. The results support angiogenesis without reducing every effect to receptor forward catalysis. Human HUVEC and MCF-7 experiments in PMID:22555806(https://www.tandfonline.com/doi/full/10.4161/cbt.20080) show opposite ERK responses to the same receptor stimulus, with different downstream dependencies. A single universal positive or negative ERK-regulation claim would lose this context.
Human vascular and lymphatic variants can impair receptor activation, abundance or localization. PMID:30578106(https://pmc.ncbi.nlm.nih.gov/articles/PMC10292091/), Figure 4 and STAR Methods, tests HA-tagged mouse Ephb4 in human HEK293T, including Ras-GAP association. This supports a conserved GAP-binding role with an explicit preparation boundary. It does not establish that EPHB4 itself catalyzes GTP hydrolysis. Clinical vascular phenotypes are not interchangeable with direct measurements of every angiogenic step.
Generic binding rows are refined only where a supported partner role exists: ephrin-B2 binding contributes to the receptor activity; RASA1 association is GAP binding; the EPHA5/EPHA6/EPHA7 AP-MS associations fit Eph-receptor binding. The latter assays establish complex association, not a purified binary interface, trans-cell orientation or ligand agonism. Actual target rows and methods are listed in the source guide.
PMID:26481148(https://pmc.ncbi.nlm.nih.gov/articles/PMC4643364/) also reports erythropoietin binding and receptor-dependent activation in selected human tumor-cell models. Its recombinant binding and receptor-knockdown controls justify a contextual receptor-activity refinement. EPO–EphB4 apparent affinity is weaker than the EPOR comparison, recombinant boundaries are incompletely specified in the main Methods, and these experiments do not establish a general physiological erythropoiesis role. This activity is not promoted into a second core function. Current PubMed metadata had no linked correction or retraction; that check does not establish agreement throughout subsequent literature.
All four source product identifiers remain distinct. Independent official FASTA sequences exactly match reconstruction from the UniProt VSP edits: P54760-1, 987 residues; P54760-2, 516; P54760-3, 306; P54760-4, 414. The shorter products lack the displayed transmembrane helix and kinase region. PMID:18593464(https://pmc.ncbi.nlm.nih.gov/articles/PMC2575619/), Table 1 and Additional file 1, identifies their cloned transcripts, but the ten functionally tested ASVs in Table 2 exclude these EPHB4 variants. Their secretion, ligand affinity and physiological roles remain unestablished. Engineered soluble EphB4 constructs do not supply that missing product-specific evidence.
Reactome explicitly models human EPHB4 extracellular and cytosolic cleavage fragments. Those are separate from the natural splice products. The recovered original cleavage papers PMID:18713744(https://pmc.ncbi.nlm.nih.gov/articles/PMC2570862/) and PMID:17428795(https://pmc.ncbi.nlm.nih.gov/articles/PMC4005067/) directly test EphB2, with variable family susceptibility discussed. The EPHB4 fragment annotations remain undecided pending target-specific support. This is an experimental-transfer uncertainty, not a claim that the target is absent from the model.
The single core captures transmembrane-ephrin receptor activity at the plasma membrane and the ephrin receptor signaling pathway, linked to the already accepted angiogenesis role. The contextual erythropoietin response remains outside that core.
All 76 imported contexts were manually considered. The completed review has 46 ACCEPT, 22 MODIFY and 8 UNDECIDED decisions, with no NEW annotations. The eight uncertain contexts are sprouting-associated cell migration, two heart-morphogenesis transfers, two cleavage-fragment compartments, cell adhesion, target-specific exosome detection, and the unlocated 2025 EPHB4–EFNB2 AP-MS observation. The exact heart experiments and exosome Supplemental Table 2 were not recovered. Inaccessible older papers were not rejected because their titles emphasize another target or because the cache is incomplete.
The PAINT target path, ancestral assertions and mouse experimental donor source were checked. No target-outside-clade or circularity claim is warranted. Broad kinase, ephrin-receptor, signaling and membrane terms are refined to established specific functions or locations while keeping the original machine-fetched fields intact. Plasma-membrane rows attached to Reactome events assert receptor location; accepting that location does not validate every downstream reaction named by the event.
The review follows the campaign's supported-binding and no-new-publication-quotation instructions. Source caches, their metadata and all imported source fields are preserved. Falcon and the perplexity-lite fallback were genuinely attempted but failed; no provider-named research output was fabricated. The original-paper reading and source-chain work summarized here supply the manual research. The source guide records actual access and distinguishes separately inspected figure images from captions.
Normal whole-gene validation passed on 2026-10-24, including schema, reference, GOA and ontology-term checks, with zero curation advisories. The normal status updater computes COMPLETE, with no remaining status mismatch. Eight source-specific UNDECIDED decisions remain explicit scientific uncertainties; COMPLETE is the validator workflow status, not a claim that those questions are resolved. All 34 normal publication caches, all 76 imported source objects and all four source product objects are preserved.
The 2025 AP-MS interaction remains source-specifically UNDECIDED because the exact EPHB4–EFNB2 record was not recovered. Its absence from a targeted search is not evidence against the established receptor–ligand function. The erythropoietin refinement now leads with the positive binding, receptor-knockdown and signaling results from Figures 1–3 of PMID:26481148, while retaining the experimental context and affinity limits.
PMID:16424904 was available as publisher main text despite its abstract-only normal cache. Main Results and operative Methods were read, as well as figure captions; “captions only” refers to figure inspection, not to article access. The reference assessment now states this explicitly. Its ectopic viral constructs target mouse endothelium, including an artificial 1–611 truncation; neither a particular construct species where unstated nor equivalence with natural short human isoforms is inferred. The existing angiogenesis annotation is now reflected in the core, retaining the forward/reverse signaling distinction. No source fields, source caches, product records or publication quotations were changed.