Reviewed the eleven direct contexts from six whole PMID groups. See the source guide, complete source-keyed decisions and product-context joins. The proposal retains two integrin-activation annotations, refines lipid binding to phospholipid binding, and keeps seven positive cellular/location contexts as non-core. The exosome context remains undecided because its exact FERMT3 table entry was not recovered.
The key experimental distinctions are human patient-cell biology versus mouse-protein rescue in PMID:19234463; isolated F1-loop phospholipid binding versus a hypothetical full-length membrane model in PMID:27101375; and comparative MDS platelet proteomics versus selective FERMT3 causality in PMID:23382103. The membrane-proteome thesis provides positive related primary target detections, but is not the final published supplement. These limits remain explicit in each whole-reference assessment.
Both curated splice-product records, all 31 imported source objects, all reference identities/findings, and the twenty unowned database reviews are preserved. No new annotation or publication quotation was added. The source-specific uncertainty is retained without inferring curator error from access failures.
The genuine intake Falcon request failed authentication, and its configured fallback failed for quota. No automated research report was created or invented. This journal and source guide are authored evidence records. Core functions and a standalone gene description are reserved for the integrated review.
Kindlin-3 acts on the cytoplasmic side of integrin adhesion machinery. Human patient platelets and leukocytes retain integrin abundance but lose activation-dependent antibody/ligand binding, aggregation and spreading; KINDLIN3 rescue and knockdown connect these phenotypes to the gene (PMID:19234460(../../../publications/PMID_19234460.md), Figures 1–2 and 4, main Results and rescue Methods). Agonist-induced P-selectin and several upstream responses persist, distinguishing the activation defect from loss of all platelet signaling. Independent human FERMT3 lesions and deficient B-cell contacts/migration support the same mechanism, but the rescue reagent is explicitly mouse EGFP-KINDLIN3 in human recipient cells (PMID:19234463(../../../publications/PMID_19234463.md), Figures 2–4 and construct Methods). Neither study compares the two current human splice products.
The molecular interaction has ortholog experimental support beyond loss-of-function phenotypes. The mouse platelet study assays recombinant kindlin-3 F3 domain with integrin tails and distinguishes that direct interaction from platelet-lysate pull-downs (PMID:18278053(../../../publications/PMID_18278053.md), Figure 3c–g and GST-fusion Methods). Mouse beta2-tail association is mapped with motif controls in platelet lysate; that assay is not a purified full-length human affinity measurement (PMID:19234461(../../../publications/PMID_19234461.md), Figure 1d–e). The latter study separates defective neutrophil firm adhesion/arrest from preserved selectin-mediated rolling (Figures 3–4). Integrin binding and activation therefore form the central molecular/process pair; cell-type-specific aggregation, migration and spreading remain positive context rather than independent catalytic activities.
Adhesion-site location has distinct experimental resolutions. Tagged kindlin-3 accumulates at adherent contacts in rescued human patient cells (PMID:19234460(../../../publications/PMID_19234460.md), Figure 4e). Mouse hematopoietic cells show podosome-associated endogenous staining, while EGFP-kindlin-3 colocalizes with vinculin in the actin-surrounding ring of dendritic-cell podosomes (PMID:16876785(../../../publications/PMID_16876785.md), Kindlin-3 localization Results and Figure 5F). Diffuse ectopic localization in fibroblasts is a different cell context. The inspected ontology has no strict is_a/part_of path from podosome to cell-substrate junction; neither label alone is used to discard the other location. These structures are not asserted as protein-containing complexes.
The isolated Leu142–Gln224 F1 loop interacts preferentially with POPS and POPG vesicles by NMR; POPC effects are weaker (PMID:27101375(../../../publications/PMID_27101375.md), expression/vesicle/NMR Methods and Figure 3). Phospholipid binding captures both tested acidic lipids more precisely than generic lipid binding. This is a measured physical interaction, but the full-length membrane-orientation model in Figure 6 is explicitly hypothetical. It does not establish PH-domain phosphoinositide specificity. The printed EK21 peptide also has an extra initial E relative to its stated interval; the construct table preserves the mismatch without inventing a natural product. A single integrin-binding core is retained, with the measured lipid interaction described as a mechanistic contribution and full-length reconstitution proposed to test it.
The MDS platelet study identifies URP2_HUMAN/Kindlin-3 among several reduced proteins; its target appears in Table I, spot 6 (PMID:23382103(../../../publications/PMID_23382103.md)). Main Methods and Figures 5–8 distinguish talin-specific FRET, receptor activation, spreading and aggregation measurements. Multiple proteins change, so the MDS phenotype is not assigned uniquely to FERMT3. Independent human FERMT3-deficiency evidence supports retaining the platelet role while preserving the comparative-proteomics limitation.
Membrane association and extracellular compartments require separate evidence. A related first-author thesis contains two human YTS membrane-fraction Kindlin-3 target rows, but its larger dataset is not treated as the final published supplement for PMID:19946888(../../../publications/PMID_19946888.md). Together with independent contact localization and phospholipid interaction, it corroborates a broad membrane claim. For PMID:20458337(../../../publications/PMID_20458337.md), original Methods/Results establish human RN B-cell exosomes and distinguish whole-exosome proteins from an MHC-II-associated subset, but the exact FERMT3 supplemental row remains unread. That specific exosome annotation stays UNDECIDED.
The actual Reactome reaction R-HSA-481007 includes Q86UX7 in alpha-granule input and extracellular output sets. The exact FERMT3 result in its platelet-releasate source, PMID:23640497(../../../publications/PMID_23640497.md), could not be inspected. A release measurement would not by itself distinguish luminal storage from vesicle release, shedding or another route. Both source-specific locations remain UNDECIDED; intracellular integrin activity and preserved P-selectin responses neither prove nor refute these locations. Cargo membership does not establish that kindlin-3 performs exocytosis.
The two natural products and their four-residue difference are preserved in the product map. Tagged proteins, isolated domains and patient-associated truncated bands are not substituted for product-specific assays. The direct and database guides give original locators and access limits. The PAINT evidence establishes literal inherited assertions and sequence matches, while an unavailable tree leaves the target-to-node path independently unverified. No publication quotations or new annotation rows are added.
The long and short kindlin-3 products now have a direct functional comparison. Lee et al. (PMID:42574321, Blood, online 10 August 2026; DOI 10.1182/blood.2026034274) tested sequence-confirmed full-length tagged human proteins in FERMT3-deficient neutrophil-like HL-60 cells. Both rescue beta2-integrin activation and arrest. The long IPRR-containing form shows greater recruitment at the plasma-membrane contact zone and greater arrest in this setting (Figures 1–2; Methods, manuscript lines 136–171; Results 175–235). Figure 2 does not show a significant difference in spread area. The lipid co-sedimentation comparison uses isolated PH domains, residues 348–479 (Figure 3; Results 236–246), not purified full-length proteins. Human constructs in mouse fetal-liver HSPCs are a separate complementation assay. The transcript associations during G-CSF exposure and transplantation do not demonstrate that one splice form causes clinical recovery. These results support explicit functional records for both natural splice products while preserving their original UniProt mappings.
The broad membrane annotation can be refined at the gene level to plasma membrane. The 2026 qDF/TIRF observations are direct cellular evidence, supported by human B-cell contact-site localization in PMID:19234460 Figure 4e. The isolated F1-loop experiments in PMID:27101375 provide a complementary lipid-binding mechanism. The exact final NK-cell proteomics target join for PMID:19946888 remains unverified; the related author thesis remains corroborating evidence rather than a substitute for the published supplement. The refinement does not change that source limitation or assign the NK signal to a natural splice product.
UniProt's isoform-2 statement deserves a precise source correction check. Its FUNCTION evidence list names PMID:19064721/19234460/19234463, whereas reference RN5 specifically links possible isoform-2 function to Wang et al., PMID:18280249. The accessed 2008 abstract and original cloning sections concern a 663-residue URP2SF product: overexpression suppresses the NF-kappaB reporter and induces apoptosis; knockdown activates the reporter and attenuates TNF-alpha-induced apoptosis. Thus the abstract does not support a blanket description as a repressor of apoptosis. Complete functional Results/controls/figures were not available, so this is a bounded attribution/direction question, not a manufactured GO source defect or a claim of four-residue-specific causality. The 2009 clinical originals establish integrin-activation defects and rescue; PMID:19234463 uses mouse EGFP-KINDLIN3 in human recipient B cells. They must not be recast as a comparative human natural-splice NF-kappaB assay.
The generic question and proposed experiment asking whether the two forms differ in membrane recruitment or integrin activation have been superseded by the 2026 comparison. The remaining scoped question concerns how the reported URP2SF NF-kappaB/apoptosis effects relate to the endogenous products. Earlier journal entries remain historical accounts of the evidence available when written.
The Reactome platelet-release model places FERMT3 in an alpha-granule-lumen input pool (entity 8862993, within set 481003). A granule lumen is separated from the cytosol by a membrane. FERMT3's known integrin-tail interactions occur on the cytoplasmic side, and its UniProt record contains no SIGNAL or TRANSMEM feature. This leaves entry into a luminal pool unexplained. Feature absence is not a direct localization-negative result, and no alternative FERMT3 entry mechanism has been established in the sources examined. The accessible first page of PMID:23640497 describes release inferred from depletion after platelet stimulation; it does not resolve membrane sidedness. Because the FERMT3 table and complete compartment evidence remain unread, the lumen annotation remains UNDECIDED. Preserved P-selectin responses in PMID:19234460 address platelet activation and do not settle FERMT3's location.
The 2026-10-26 QuickGO query restricted to is_a and part_of confirms that neither podosome (GO:0002102) nor cell-substrate junction (GO:0030055) is a strict ancestor of the other. Both locations remain supported in their respective cellular contexts: the podosome ring in the mouse localization study (PMID:16876785, Results and Figure 5F), and adherent human patient-cell contact sites in PMID:19234460, Figure 4e. This ontology result does not imply that the structures are biologically unrelated.
Annotation decisions now live only in the canonical review. The direct and database sidecars retain their source keys, row indices and evidence locators. Repeated generic policy language has been shortened while preserving the source-specific access limits. The F1-loop lipid-binding evidence remains described in the core mechanism; isolated-domain assays do not by themselves establish an additional full-length molecular activity.