Reviewed 2026-10-10. Human CDH2 encodes N-cadherin (UniProt P19022), a classical transmembrane cadherin whose extracellular region mediates calcium-dependent adhesion and whose cytoplasmic tail binds catenins. The review preserves all 114 seeded assertions, their original evidence, qualifiers, partners and isoform fields, and both seeded products. There are no NEW annotations. The core molecular activities are cadherin binding involved in cell-cell adhesion (GO:0098641), beta-catenin binding (GO:0008013), and calcium ion binding (GO:0005509).
The adhesion refinement replaces the existing generic cadherin-binding annotation; it does not add an overlapping NEW row. GO:0098641 captures the cadherin-mediated trans interaction more precisely than the broader cell-cell adhesion mediator term. The primary developmental-disease study tests wild-type and variant N-cadherin in L-cell aggregation and mixed-cell trans-adhesion assays, with expression and cell-surface controls. Its abstract reports that variants “impair the adhesive activity of N-cadherin” (PMID:31585109(https://pubmed.ncbi.nlm.nih.gov/31585109/)). The selected full Results and Figure 3 were also inspected through the primary PMC article. Calcium was present in the assay; changes to calcium-binding pockets were modelled, rather than established by direct variant calcium-affinity measurements. The main text does not establish construct species, and supplementary methods were not accessible. The phenotype is therefore evidence for N-cadherin-mediated adhesion, with this construct limitation retained. Human leukemia-cell adhesion to N-cadherin-coated wells provides independent human-cell support at abstract scope (PMID:19101069(https://pubmed.ncbi.nlm.nih.gov/19101069/)).
Beta-catenin binding is independently grounded in endogenous N-cadherin association in human HEK293 cells and the effects of beta-catenin armadillo-domain deletion, read in selected full Results and Methods (PMID:17679699(https://pubmed.ncbi.nlm.nih.gov/17679699/)). Human 293T co-immunoprecipitation in PMID:18948590 and U373 cadherin/catenin complexes in PMID:17098867 corroborate that role. The latter study used pan-cadherin antibodies and estimated that most cadherin in that system was N-cadherin; it is not a purified binary interaction assay. N-cadherin cytoplasmic GST fusions were actually included in the beta-catenin controls of the otherwise VE-cadherin-focused PMID:17057644. Selected full figures were inspected, avoiding a conclusion based only on the article title. Precise construct species for that experiment was not independently resolved.
Plakoglobin association has separate support from human HT-1080 co-immunoprecipitation and deletion mapping at abstract scope (PMID:7650039(https://pubmed.ncbi.nlm.nih.gov/7650039/)). That abstract maps plakoglobin regions that associate with N-cadherin and alpha-catenin separately; it does not settle direct CDH2–alpha-catenin binding. The original alpha-catenin IPI is consequently UNDECIDED, without alleging an erroneous curator attribution.
The bounded source-field extract preserves exact source queries, response hashes, assay fields and annotation joins. Four positive IntAct records exactly match the original human pairs: p120-catenin and beta-catenin from PMID:21357690, and both pairs from PMID:33961781. The former records identify SH-SY5Y co-immunoprecipitation; the latter identify HEK293T bait-prey associations expanded from complexes. They do not establish every pair as isolated binary binding. The beta-catenin rows can be refined using the independent target experiments above; supported p120 associations remain non-core.
The authentic publisher source-data workbook for PMID:35922511(https://www.nature.com/articles/s41586-022-05028-x) identifies CDH2 P19022–FCER2 P06734 in Figure1_network row 172, with screening score 2.970475, and Figure2_network row 126. Selected full Methods establish the recombinant extracellular-domain screening context. This exact network record supports retaining the generic binding annotation as non-core. The heatmap sheet contains zero values for both orientations; those entries are preserved in the extract and are not presented as positive measurements. A target-specific SPR trace was not inspected, and the modelling sheet's affinity field is not asserted to be an independently measured affinity. No immune-cell process is newly assigned from this association.
The exact CDH2 supplementary rows for PMID:17510365, PMID:35271311 and PMID:36950384 were not recovered. Their beta-catenin-binding refinements follow the original GOA WITH/FROM partner P35222 and independently established target-specific binding evidence. The original screen rows were not independently inspected; no screen join is invented. The CDH2–connexin43 assertion from PMID:29366904 remains UNDECIDED because the accessible abstract distinguishes mouse RPE immunoprecipitation and human ARPE-19 transport experiments without resolving the exact human pair.
Supported generic interactions without a more informative established molecular activity are retained as non-core under the fixed project instruction. Functions of binding partners, including protease, kinase and phosphatase activities, are not transferred to CDH2.
The exact Human Protein Atlas CDH2 subcellular page reports supported plasma-membrane and cell-junction staining, including U-251MG/U2OS with CAB000141 and THP-1/U2OS with CAB080169. Human adhesion and catenin-binding studies independently support activity at the membrane and its cytoplasmic face. The cytoplasm is_active_in assertion is refined to adherens junction: the tail/catenin activity occurs on the intracellular face of the membrane-associated junction, and GO:0005737 excludes the plasma membrane. This differs from the separately retained non-core IDA located_in observation of intracellular immunoreactivity. The distinction concerns the asserted relation and location, not a ranking of IBA below IDA.
The full flotillin study was inspected through indexed primary article text. Its N-cadherin experiments use mouse C2C12 cells, whereas the principal human MCF7 experiments test E-cadherin. The species of the N-cadherin-GFP construct, referred back to Taulet 2009, was not resolved. Human lamellipodial and cortical-actin assertions remain UNDECIDED; general human junctional localization is independently supported (PMID:24046456(https://pubmed.ncbi.nlm.nih.gov/24046456/)). Colocalization is not converted into direct actin binding.
The exact CDH2 peptide records underlying focal-adhesion PMID:21423176 and extracellular-matrix PMID:27559042 localization were not recovered; those assertions remain UNDECIDED. The original Reactome cache was read at identity/summary scope. A current authentic reaction 8952289 ContentService retrieval now resolves its complex 8952321, substrate set 8956956 and CDH2 entity 8956761.1. The entity explicitly spans P19022 residues 160–906, including transmembrane 725–745 and cytoplasmic 746–906. It is not a soluble propeptide or isolated extracellular fragment. The source ER-lumen assignment is therefore refined to ER membrane (GO:0005789), retaining luminal exposure of the N-terminal domain and its substrate context. The bounded source extract preserves the exact API fields, URLs and hashes; the normal cache is unchanged. Other membrane TAS annotations are retained as core on independent human evidence while their original source provenance remains intact.
The source extract also traces the current mouse P15116 records for 50 comparative-source rows. This is a source audit, not a reconstruction of historical PAINT IBD placements; neither short donor lists nor target self-donors are treated as weak or circular evidence. Mouse cardiac, neuronal and stem-cell findings support conserved specialized roles without being relabelled as direct human experiments. The full mouse MIN6/betaTC6 and neural-crest co-culture experiments in PMID:26750727 include N-cadherin neutralization/knockdown and cell-clustering effects. Adult murine stem-cell anchorage and MMP24-dependent shedding in PMID:24952463 support niche maintenance at abstract scope, but the exact glial-differentiation experiment was unavailable; both glial-differentiation rows remain UNDECIDED. Mouse desmoplakin perturbation in PMID:25496840 supports the context of N-cadherin-containing growth-front complexes, not an isolated human axon-growth mechanism. PMID:34702855 distinguishes human family genetics, synthetic-peptide furin cleavage, and mouse H150Y knock-in synaptic phenotypes. CDH2 is the furin substrate, not the protease.
The current mouse P15116 RNA-binding donor is independently confirmed in the MGI J:241317 source records. The complete primary study and recovered publisher supplements measure transcript regulation in mouse embryonic stem cells. Supplementary Note S5, page 8, and Table S1, Targeted Loci row 28, show Chd2 transcript changes after linc2025 promoter deletion; Table S3 supplies transcript-capture probes. These are RNA measurements, not evidence of Cdh2 protein binding to RNA. Independent inspection of all three workbooks found no Cdh2/P15116/cadherin target record. With this previously missing evidence recovered, the human electronic transfer is MARK_AS_OVER_ANNOTATED for its source support. The original curator-to-assay mapping remains unresolved; no Chd2-to-Cdh2 naming error is asserted, nor is RNA binding declared biologically impossible. Exact MGI fields, selected workbook cells, URLs and response hashes are in the bounded source extract.
Mixed cardiac adhering junctions associated with the area composita in mouse embryos justify retaining the source desmosome context without making N-cadherin a member of the desmosomal-cadherin family (selected full PMID:15479741). The kinase-binding source PMID:23392350 was read at abstract scope; its 2024 erratum was not independently resolved. No catalytic activity is assigned to CDH2. The muscle-stretch TAS remains non-core: PMID:30568254 explicitly places the N-cadherin/beta-catenin complex in mechanosensing and measures its abundance under cardiac pressure overload. This preserves the author-stated complex context. Its gelsolin/PI3K interventions do not isolate CDH2 stretch-to-signal conversion, so the annotation is not recast as IDA or a standalone sensor core. Absence of a direct CDH2 assay alone does not negate a TAS assertion.
Normal gene fetching generated the GOA, UniProt and seed files. Existing main-branch publication and Reactome caches were reused exactly; missing sources were fetched through the normal publication/Reactome tooling. Fresh downloads were not used to overwrite pre-existing source files. Machine-fetched titles replaced four seed TODO titles, with reference identifiers preserved. Reference assessments distinguish abstract-only access, selected full-text reading, exact target records and unresolved supplementary evidence; a cache's full_text_available flag alone is not treated as proof that a target assay was read.
A genuine Falcon research attempt and the explicit fallback each timed out without producing a research report. This document records the manual primary-source review; no provider report was fabricated. Annotation and core-function review workflows were applied, with an independent bounded peer of the direct adhesion/core interpretation. All source assertions and products remain available for audit, including unresolved ones. Repeated quotations were avoided; brief anchors in the YAML and notes are counted together by source.
The location refinements preserve the original annotations and qualifiers. Calcium coordination is represented as the protein’s own molecular activity, rather than a contributes_to assertion about another catalytic subunit. The original structural and variant-assay limitations remain explicit. Source-tracing questions belong in these notes; the structured questions retain biological uncertainties about catenin contact, isoform differences and mechanotransduction. UNDECIDED denotes unresolved evidence or interpretation and is not restricted to complete failure to retrieve a publication; access to a main text does not automatically resolve its target-specific supplementary data.
Round 4 of projects/INVOLVED_IN_PRECOMPOSITION.md (2026-11-07): the core-function molecular function now uses the genus GO:0045296 cadherin binding instead of the precomposed GO:0098641; the adhesion context is carried by GO:0044331 cell-cell adhesion mediated by cadherin in the core functions. The cadherin-binding IBA row is accepted as it stands rather than refined, so the earlier statements above that the core activity is GO:0098641 and that the adhesion refinement replaces the generic cadherin-binding annotation are superseded.