CDC42 is a Rho-family GTP-controlled signaling switch. Its intrinsic GTP hydrolysis and nucleotide-dependent effector recognition connect spatial signals to actin organization and cell polarity. These are properties of CDC42 itself: kinase, GEF, GAP and actin-nucleation activities belong to other proteins in the associated machinery. The core uses GO:0003925, whose current definition includes small monomeric G proteins. The core separately records GBD/CRIB effector recognition and uses the specific positive-filopodium and epithelial apical/basal-polarity processes; their broader ancestors remain in the annotation review.
The primary WASP study explains how CDC42 binding releases an autoinhibited region, “enabling its interaction with the actin regulatory machinery” (PMID:10724160(https://pubmed.ncbi.nlm.nih.gov/10724160/), abstract). This supplies a mechanism for actin regulation without making CDC42 an actin polymerase. Recombinant Cdc42 homodimerization and intrinsic hydrolysis are directly described in PMID:9748241(https://pubmed.ncbi.nlm.nih.gov/9748241/); nucleotide-loading reporters in other papers are not themselves purified hydrolysis assays. Human PAR6 proteins bind active CDC42 and participate in polarity complexes (PMID:11260256(https://pubmed.ncbi.nlm.nih.gov/11260256/)); the Rac-specific ruffling experiment in that abstract is not transferred to CDC42.
Selected full Results in human Caco-2 cells show that CDC42 depletion disrupts junctional and apical/basal markers; constitutive activation also perturbs junctions. Spatial control, rather than simply more activity, is important (PMID:22891260(https://pubmed.ncbi.nlm.nih.gov/22891260/), Figures 1 and 7). Human endothelial experiments connect fibronectin/NRP1/ABL1 signaling to CDC42 activation and inhibitor-sensitive protrusions (PMID:26051942(https://pubmed.ncbi.nlm.nih.gov/26051942/), selected Figure 4 Results). These support actin and polarity as core outputs. Their vascular, adhesion and stress-fiber contexts remain non-core.
The exact two seeded alternative-product records are preserved, including legacy secondary accessions. The UniProt record identifies different C termini in the placental and brain products. Experiments on palmitoylated GAP43/paralemmin motifs are not direct assays of CDC42 brain-product lipidation (PMID:14978216(https://pubmed.ncbi.nlm.nih.gov/14978216/)). Endogenous product targeting remains a useful experimental question.
The standing project curation instruction explicitly permits “KEEP_AS_NON_CORE” for supported generic binding when no more informative MF is established. This governs supported associations here. Inaccessible exact assertions remain UNDECIDED. Neither repeated database records nor later reuse of the same screen is independent biological replication.
The durable interaction evidence extract contains exact pair, taxon, method, feature and source identifiers, record URLs, query-response hashes, and the three OpenCell dataset entries. It preserves source-linked curation and processed data as traceable evidence; it is not independent replication or a raw-data reanalysis. Long copied source comments are omitted. JSON is retained as the format of this machine-derived retrieval receipt, with response hashes and literal database fields, rather than as a new authored biological model.
The independent screen consultation checked all 37 assertions across PMIDs 20936779, 25416956, 28514442, 31980649, 32296183, 32707033, 33961781 and 35271311. Thirty-four have exact source-linked human IntAct pairs; three have exact official OpenCell dataset entries. Fifteen kinase-partner assertions are refined to protein kinase binding after checking partner identities; 22 remain generic non-core associations. HuRI's CDC42EP1 Q00587-2 and HDAC7 Q8WUI4-6 fragment 699–974 are retained. OpenCell used an endogenous N-terminal CDC42 tag and protein-group prey identification. Missing mutation fields do not prove a wild-type sequence or a particular nucleotide state; AP-MS does not establish a direct interface or kinase activation.
The separate low-throughput audit joined exact accession pairs to source-linked IntAct records and read relevant primary abstracts or selected full assays. The audit preserves activating, dominant-negative and interface-mutant comparisons, foreign effectors, cross-species partners and partner isoforms. PAK/PRKCI/MARK4/LRRK2/MLK3/MRCK association supports kinase binding, not CDC42 kinase catalysis. GBD refinement is used where WASP-domain experiments were actually resolved, including PMIDs 15361624, 16293614 and 9660763.
Selected direct source checks include the following:
The original PAK3 binding-domain capture experiment in PMID:19039103 resolves the source kinase-binding assertion: early infected HeLa-cell CDC42 is recovered, with later loss following VopS action. This does not give CDC42 AMP-transferase activity. Official ACS Supplementary Table S1 for PMID:20873783 lists the full 191-residue CDC42 Y2H prey, and Figure S1C shows endogenous NEK6/CDC42 overlap in HEK293 cells. The actual panel was viewed; it has no centrosomal marker. Kinase binding is retained, while precise centrosome and complex membership claims await the relevant main-text assays. Selected full PMID:21566117 Results independently identify CDC42 in SopB affinity-capture/SILAC preparations; exact A0A0H3NA16 maps to SopB in the assayed SL1344 strain.
TRE17/USP6: selected primary Methods/Results in PMID:12612085(https://pubmed.ncbi.nlm.nih.gov/12612085/) support preferential association of activated human CDC42 with TRE17, interpreted as an indirect effector complex. PMID:15509780(https://pubmed.ncbi.nlm.nih.gov/15509780/) identifies the earlier form as catalytically inactive TRE17(onco). The historical thioesterase name is legitimate, but an active-enzyme class cannot be inferred for that tested product. The exact thioesterase-binding assertion remains UNDECIDED; no modern isoform accession or automatic generic replacement is invented.
BNI1/Bin1: PMID:9082982 and IntAct IM-20287-1/-3 have a material unresolved conflict. The experimental descriptions name yeast Bni1p, but the matched Q9VEX9 accession currently identifies Drosophila Bin1/SAP18. Full original construct Methods were not recovered. An accession match alone does not settle biology; the row remains UNDECIDED without a definitive misattribution claim.
The ubiquitin-ligase assertion is removed because the actual selected experiments fail to establish the stated catalytic function. PMID:21435037(https://pubmed.ncbi.nlm.nih.gov/21435037/), selected full Results and Figures 2 and 8, shows CDC42-dependent enhancement of SopB ubiquitination in yeast, while a CDC42-binding-deficient SopB mutant remains ubiquitinated in HeLa infection. It does not demonstrate CDC42-catalyzed ubiquitin transfer. This judgment is based on the experiments, not a title or a presumed incompatibility with GTPase activity. The same paper's bacterial replication/yeast-growth effects are not relabeled as positive regulation of human cell growth. Exact SptP/SopE2 comparisons remain unresolved.
PMID:16380373(https://pubmed.ncbi.nlm.nih.gov/16380373/), selected full Figure 4 Results, examines endogenous CDC42 Golgi localization and brefeldin-A redistribution. It supports Golgi apparatus rather than membership of the eight-subunit COG complex. The distinct Golgi-membrane ISS assertion retains its rat Q8CFN2 donor.
The storage-vacuole inference has an authentic mouse donor: PMID:21828338(https://pubmed.ncbi.nlm.nih.gov/21828338/), full Figure 1C/Methods, detects Cdc42 in the VAMP2-marked MIN6 insulin-granule fraction. Secretory granule is the supported refinement. Fractionation does not resolve membrane versus lumen, and this remains an orthology-based human assertion.
The exact focal-adhesion and extracellular-vesicle target tables in PMIDs 21423176, 19056867, 19199708, 20458337 and 23533145 were not inspected. Existing proteomic assertions remain UNDECIDED; intracellular signaling is not a reason to reject an extracellular-vesicle finding. The remaining source-specific NEK6 compartment/complex, macrophage-differentiation, Wnt planar-polarity and substantia-nigra-development claims also retain their stated gaps. Proteomic presence is not automatically a developmental mechanism. An abstract's omission is not evidence of a curator error.
Activity-qualified PAINT annotations retain their ancestral inference. Target entries in WITH/FROM are valid descendant evidence, and donor counts are not used to judge confidence. Reactome records supply pathway/compartment provenance; summaries are not new independent imaging experiments. Selected long event summaries were read for the relevant location context, not exhaustively reconstructed into entire causal models.
Normal intake produced 285 GOA lines and 284 seeded assertions. The normal deduplicator merged the same CDC42–SopB assertion from two providers, differing in assignment/date fields; both original TSV lines remain intact. All seeded source fields, products and reference identities are preserved. Failed title retrieval placeholders are replaced by titles from the authentic normal caches. Two additional references support the donor/partner interpretation above; no NEW annotation is added.
The first normal fetch encountered a silent first-page failure and produced an empty seed; it is preserved separately and was superseded by a successful fresh normal fetch. All 98 original PMID caches and 49 Reactome caches were recovered through ordinary retrieval or exact reuse of existing caches. Normal Falcon and explicit perplexity-lite fallback each reached the bounded 90-second timeout and produced no report. These are manual notes, not a provider report.
Access distinguishes cached abstracts, selected full sections and exact source-linked target records. Some full primary sections were read through indexed publisher/PMC or article-mirror text even when the unchanged normal cache is abstract-only. Figure legends and text were read where stated; figure pixels, raw measurements and complete supplementary datasets were not implicitly inspected. The separate packet records exact cache hashes, source receipts, consultation scope, preservation checks, aggregate quotation counts and normal validation results.
The review prompted a source-access clarification, a more complete core synthesis and explicit limits for optional suggestions. All 284 original annotation objects retain their source fields and actions. The two core activities now connect G-protein switching to epithelial apical/basal polarity and GBD-domain recognition to positive regulation of filopodium assembly. Broader actin regulation is explained in prose; the polarity and filopodium-regulation ancestor terms are not repeated as separate core processes. This adds no new annotation.
Golgi source and relation. The author-uploaded primary article for PMID:16380373, Results adjoining Figure 4C-D, explicitly reports “redistribution of Cdc42 from the perinuclear region to the cytosol” after brefeldin A in HeLa cells. Figure 4E-G compares CDC42 with p50RhoGAP. Its abstract focuses on endosomes, but that does not erase the CDC42 experiment. The normal cache remains abstract-only; this is an externally accessed selected full Results/Methods read, not a rewritten or newly complete cache. Thus the reference remains VERIFIED for this bounded claim. The COG-to-Golgi proposal explicitly requires located_in; original part_of stays in the source record only. The separate Golgi-membrane ISS assertion retains its rat donor. The independent PMID:15361624 abstract anchor is now present in row 212, with its MEF/3T3 imaging scope stated.
GBD evidence. The author-hosted full PMID:15361624 PDF, Figure 1, identifies WASP residues 201-321 and the “CRIB domain essential for Cdc42 binding”; the H246D/H249D controls disable binding. This supports the source-specific GBD refinement independently of a generic protein-interaction label. Human-neutrophil extracts are assayed separately from the MEF/3T3 imaging. For PMID:9660763(https://pubmed.ncbi.nlm.nih.gov/9660763/), the abstract itself reports “The fragments bind with submicromolar affinities in a GTP-dependent manner”; the flanking sequence, not just the short CRIB motif, strengthens binding. For PMID:16293614(https://pubmed.ncbi.nlm.nih.gov/16293614/), the abstract describes autoinhibition-dependent GBD phosphorylation/dephosphorylation and states that “this effect is largely reversed by Cdc42”. These latter refinements do not depend on an unrecorded full-text read. The existing short PMID:10724160 quote is used once in the GBD core, and its separate earlier notes excerpt is included in the aggregate quotation budget.
Secretory granules. The externally accessed PMID:21828338 primary article, Figure 1C, defines “SG, storage granule pool”. Its Results and fractionation Methods place the comparison in mouse MIN6 cells and identify VAMP2 as the granule marker. CDC42 is detected alongside βPix in that fraction; human-islet expression is a separate experiment. Secretory granule remains the proposed orthology-based compartment, without a membrane-versus-lumen claim.
NEK6 association. The official ACS supplement for PMID:20873783, Table S1, lists “CDC42” with “AAH03682.1” and “1-191/ 191”. The previously inspected Figure S1C and its legend concern endogenous CDC42/NEK6 overlap in HEK293 cells. This establishes the screen target and supports the kinase-binding decision. That panel has no centrosomal marker; neither shared staining nor the article title settles the separate centrosome or protein-complex assertion. Those two remain UNDECIDED.
Ubiquitination. The PMID:21435037 abstract says binding-deficient SopB proteins “were ubiquitinated upon translocation in mammalian cells”. Selected full primary Results/Figures 2 and 8 distinguish enhancement in yeast coexpression from retained SopB ubiquitination in HeLa infection. These are not CDC42 ubiquitin-transfer catalytic assays. The source-scoped REMOVE is retained, with an accessible abstract anchor and the full-section locator; the partial source access does not establish every other assertion from this paper.
Optional localization and PAK1 suggestions. Full selected PMID:11035016 Results and Figure 4 legend confirm Myc-CDC42 staining in NIH-3T3 fibroblasts, but the exact filopodial signal/construct mapping for the source IDA remains unresolved here. Regulation of protrusion formation does not by itself prove residence inside a protrusion. Conversely, the active-state reporter in PMID:15361624 is not a total-protein localization assay, so its filopodial activity pattern is not used to reject location. Row 222 stays UNDECIDED. For the two PAK1 rows, the reasons now distinguish a named Q13153 WITH partner from an inspected source-specific assay: the bounded IntAct query returned no target record for PMIDs 19596003 or 21167153. This absence is not biological counterevidence.
The closing description now states the observed alternative C-terminal sequences without implying that endogenous product targeting was experimentally resolved in this audit. Optional blanket VERIFIED assessments that did not distinguish individual source checks have been removed; source-specific assessments above, all reference identities/titles, annotation rationales and access flags remain available. A missing optional reference assessment is not a new negative judgment. Source caches and the interaction receipt are unchanged. HTML is regenerated from this exact YAML and these notes.