ARHGAP23 is a large (1491-residue) multidomain Rho GTPase-activating protein built on the PDZ - pleckstrin-homology - RhoGAP architecture it shares with its closest paralog ARHGAP21. Its Rho-GAP domain is intact: the catalytic arginine finger is present at position 942, in register with the arginine resolved in the p50RhoGAP:RhoA transition-state structure, and at the structurally defined GAP:GTPase interface the protein is residue-for-residue almost identical to the experimentally active ARHGAP21. Cellular assays place its activity on RhoA and on Rac1 but not on Cdc42, although the two substrates have been measured in different systems and one study that saw RhoA suppression saw no Rac1 activation, so the Rac1 arm is cell-type dependent rather than settled; no assay on purified protein has been reported. ARHGAP23 acts where cells contact something. In migrating cells it is held by the planar cell polarity protein Prickle1 at the non-protrusive lateral cortex that flanks active lamellipodia, and losing either partner raises GTP-bound RhoA, increases myosin light-chain phosphorylation and disorganizes the actomyosin network - a spatial restriction of RhoA that confines where a cell can protrude. In keratinocytes it is bound by the junctional scaffold plakophilin 4, which lowers its association with RhoA and thereby licenses cortical RhoA activity and the cortical actin ring; when plakophilin 4 is absent, ARHGAP23 suppresses cortical RhoA from the cytoplasm and stress fibres form instead. It is also enriched at focal adhesions, with a pericentric distribution within each adhesion, and in endothelial cells it is one of a small set of proteins that bind VE-cadherin. Its reported cellular consequences - adhesion maturation, dendritic spine maturation, cell shape fluctuation during migration - are all downstream of the same job: locally inactivating Rho-family GTPases at a membrane where a junction, an adhesion or a polarity cue has specified that contractility should be kept off.
| GO Term | Evidence | Action | Reason |
|---|---|---|---|
| GO:0005096 GTPase activator activity | IBA GO_REF:0000033 | ACCEPT | Summary: The core molecular function, and a well-founded IBA. The node's named donor is ARHGAP21, which carries its own experimental GO:0005096 IDA, and ARHGAP23 has since acquired direct cellular evidence of its own, so the phylogenetic inference and the target-specific data agree. Reason: Three things had to hold for this to be more than a domain-presence guess, and all three do. First, the donor is real: genes/human/ARHGAP21/ARHGAP21-goa.tsv carries "enables GO:0005096 IDA PMID:15793564", so the node was placed on experimental descendant evidence rather than on similarity. Second, the catalytic machinery is intact in the target - the committed analysis maps ARHGAP1's structurally resolved arginine finger (R282, PDB 1TX4) onto ARHGAP23 R942 reciprocally, and finds ARHGAP23 identical to the active paralog at 22 of the 25 positions that contact RhoA or the transition state in that structure. Third, and decisively, ARHGAP23 no longer needs the inference: three independent cellular studies measure GAP activity directly. The residue argument is deliberately not overstated. Retention of the arginine is not evidence of activity - one of the two known GAP-dead controls in the same analysis passes the same test, and ARHGAP11B, reviewed earlier in this batch, keeps its arginine while two IDA rows record that it has no activity. The residue result removes an objection; it does not make the case. The case is made by the experiments. Propagation Review Root cause: NO FAILURE CORE Sources checked: PANTHER:PTN008592632 · PANTHER ancestral node for the ARHGAP21/ARHGAP23 PDZ-PH-RhoGAP clade SUPPORTS TRANSFER ARHGAP21 receives its own IBA from this same node, which is the expected pattern for a node seeded by that gene's experimental annotation rather than a sign of circularity. UniProtKB:Q5T5U3 · human ARHGAP21, the closest paralog and the node's experimental grounding SUPPORTS TRANSFER Carries enables GO:0005096 by IDA from PMID:15793564. Its measured substrate there is Cdc42, which is not ARHGAP23's measured profile - but GO:0005096 asserts no substrate, so the transfer of this term is unaffected. The substrate divergence is recorded as a knowledge gap rather than as a propagation failure. Supporting Evidence: file:human/ARHGAP23/ARHGAP23-bioinformatics/RESULTS.md | RHG01_HUMAN (Q07960, resolved in 1TX4) | 282R | 942 | **R** | yes | yes | **yes** | PMID:27226243 Moreover, overexpressing Arhgap23 in MDA/shPk1 cells suppressed the elevation of active RhoA to a similar level as treatment with the active RhoA inhibitor, C3 transferase (Supplementary Fig. 3d). PMID:28114311 A Rac FRET biosensor [39] confirmed that shRNA against ARHGAP23 increased Rac1 activity to levels indistinguishable from constitutively active Raichu Rac V12 (Fig 5D and 5E), demonstrating that ARHGAP23 functions as a novel Rac1 GAP in adhesion maturation of migratory CHO.K1 cells as well as synapse maturation in neurons. |
| GO:0005096 GTPase activator activity | TAS Reactome:R-HSA-8981637 | ACCEPT | Summary: A set-level Reactome projection that happens to be correct here. The reaction's catalyst is the defined entity "GTPase activator activity of RHOA GAPs", so every member receives this term; ARHGAP23's membership is sourced in Reactome's own summary to the FRET specificity screen, which scores it significantly active on RhoA. Reason: This is the same annotation mechanism that produced a false positive for ARHGAP11B earlier in this batch, so it was checked the same way rather than trusted or dismissed by pattern. The difference is that Reactome names a source for this member - "ARHGAP23 (Müller et al. 2020)" - and that source, PMID:32203420, measures a significant RhoA-directed effect in cells (normalised delta-R/R0 -0.364, p = 2.9e-05 in the Fig. 1b Source Data). The set membership is therefore evidence-based, not homology-based, and nothing in the record contradicts it. Accepted rather than modified because there is no better term to move to: the Rho- and Rac-specific GAP activity terms were merged into GO:0005096 and survive only as narrow synonyms, so the RHOA specificity that is the entire content of this reaction cannot be carried into the annotation. That loss is recorded as an ontology knowledge gap rather than as a defect in this row. Propagation Review Root cause: NO FAILURE CORE Sources checked: Reactome:R-HSA-8981637 · Reactome reaction "RHOA GAPs stimulate RHOA GTPase activity" SUPPORTS TRANSFER A bulk projection over a defined catalyst set, which is a weak mechanism in general. It is sound for this member because Reactome's summary cites a specific experimental source for including ARHGAP23, and that source measured RhoA. Supporting Evidence: Reactome:R-HSA-8981637 ARHGAP23 (Müller et al. 2020) PMID:27226243 Following siRNA-mediated knockdown of either Pk1 or Arhgap21/23, the level of GTP-bound, active RhoA was significantly enhanced (Fig. 4a), and active Rac1 was either unchanged or decreased, respectively |
| GO:0005096 GTPase activator activity | TAS Reactome:R-HSA-9013144 | ACCEPT | Summary: The RAC1 counterpart of the previous row, from the equivalent RAC1 GAP catalyst set. Reactome sources ARHGAP23's membership to the Rac FRET biosensor study PMID:28114311 and to the family-wide screen, both of which report Rac1-directed activity. Reason: Kept for the same reason as the RHOA row: the membership is sourced rather than inferred from the domain, and two independent cellular biosensor experiments support it. The evidence is nonetheless weaker than for RhoA, and this review says so rather than letting the two identical GO rows imply equal confidence. PMID:27226243 measured Rac1 in the same experiment in which it measured RhoA and found active Rac1 decreased, not increased, on ARHGAP21/23 knockdown - the opposite of a Rac1-GAP expectation. That does not license removing the term, because two biosensor studies in other cell types report the activity directly and a decrease could be an indirect consequence of altered RhoA-Rac1 crosstalk; it does mean the Rac1 arm is plausibly cell-type dependent. As with the RHOA row, no more specific molecular-function term exists to move to, so the RAC1 specificity recorded in the reaction is invisible in the resulting annotation - which is also why this row and the previous one are indistinguishable in GOA despite describing different substrates. Propagation Review Root cause: NO FAILURE CORE Failure modes: GRANULARITY MISMATCH Sources checked: Reactome:R-HSA-9013144 · Reactome reaction "RAC1 GAPs stimulate RAC1 GTPase activity" SUPPORTS TRANSFER Membership sourced in Reactome's summary to "Martin Vilchez et al. 2017", which is PMID:28114311, the Rac FRET biosensor study, and to the family-wide screen. Flagged GRANULARITY_MISMATCH because the term emitted cannot express the RAC1 specificity that distinguishes this reaction from R-HSA-8981637, not because the transfer is wrong. Supporting Evidence: Reactome:R-HSA-9013144 ARHGAP23 (Martin Vilchez et al. 2017; Müller PMID:28114311 A Rac FRET biosensor [39] confirmed that shRNA against ARHGAP23 increased Rac1 activity to levels indistinguishable from constitutively active Raichu Rac V12 (Fig 5D and 5E), demonstrating that ARHGAP23 functions as a novel Rac1 GAP in adhesion maturation of migratory CHO.K1 cells as well as synapse maturation in neurons. |
| GO:0005829 cytosol | TAS Reactome:R-HSA-8981637 | KEEP AS NON CORE | Summary: The compartment attached to the RHOA GAP reaction, projected onto every member. A cytosolic pool of ARHGAP23 is real and is functionally implicated in the keratinocyte work, but it is not where the protein does most of its reported work, and this row is not evidence of localization in its own right. Reason: The annotation is defensible rather than earned. It carries no localization experiment: it is the compartment label on a reaction, given to all twenty-odd members of the catalyst set. Independently of that, a cytoplasmic pool is supported - in PKP4-knockout keratinocytes ARHGAP23 depletion abolishes stress fibres, which the authors read as an active cytoplasmic ARHGAP23 suppressing cortical RhoA. So the term is kept. Marked non-core because the protein's function is positional. Every study that has looked places the functionally relevant pool at a membrane: the Prickle1-dependent lateral cortex, the cell cortex in keratinocytes, focal adhesions, and the VE-cadherin complex in endothelium. A record that says "cytosol" and nothing else would describe the least informative thing about this protein, which is why cortex and focal adhesion are proposed as new annotations below. Propagation Review Root cause: NO FAILURE NON CORE Sources checked: Reactome:R-HSA-8981637 · Reactome reaction "RHOA GAPs stimulate RHOA GTPase activity" SUPPORTS TRANSFER The compartment is a property of the modelled reaction, not an observation about this protein. Retained because independent experimental work supports a cytoplasmic pool. Supporting Evidence: PMID:38970683 In PKP4-KO cells, ARHGAP23 depletion correlated with a loss of stress fibers, suggesting that ARHGAP23 is active in the PKP4-KO cell cytoplasm to reduce cortical RhoA activity and suppress cortical ring formation. file:human/ARHGAP23/ARHGAP23-deep-research-affinage.md ARHGAP23 is a multidomain RhoGAP-family protein that functions as a cytoplasmic suppressor of RhoA activity at cell-cell junctions |
| GO:0005829 cytosol | TAS Reactome:R-HSA-9013144 | KEEP AS NON CORE | Summary: A duplicate of the preceding row, arising because ARHGAP23 is a member of two Reactome catalyst sets that are both annotated to the cytosol. It adds no independent information. Reason: Identical in substance to the GO:0005829 row from R-HSA-8981637 and kept on the same grounds. It is recorded separately because GOA does, and flagging it as redundant rather than silently treating the two as one keeps the count of independent localization evidence honest: there is one cytosol claim here, from one modelling decision, not two. Propagation Review Root cause: EVIDENCE CIRCULAR OR REDUNDANT Failure modes: SOURCE EVIDENCE WEAK Sources checked: Reactome:R-HSA-9013144 · Reactome reaction "RAC1 GAPs stimulate RAC1 GTPase activity" CIRCULAR OR REDUNDANT Emits the same compartment term as R-HSA-8981637 for the same protein. Redundant with that row rather than wrong; neither is an observation of this protein's localization. Supporting Evidence: PMID:38970683 In PKP4-KO cells, ARHGAP23 depletion correlated with a loss of stress fibers, suggesting that ARHGAP23 is active in the PKP4-KO cell cytoplasm to reduce cortical RhoA activity and suppress cortical ring formation. |
| GO:0007165 signal transduction | IEA GO_REF:0000002 | MODIFY | Summary: The InterPro2GO mapping from the RhoGAP domain signature IPR000198. It is not merely vague - it is on the wrong branch. A GAP does not transduce a signal; it terminates one, which makes it a regulator of signal transduction rather than a participant in it. Reason: This is a genuine branch difference rather than a granularity complaint, and that was checked rather than assumed: GO:0035024 and GO:0035021 are both descendants of GO:0051056 regulation of small GTPase mediated signal transduction and neither is a descendant of GO:0007165, confirmed against api.geneontology.org and OLS4. So the replacement is not a specialization of the existing term; it moves the assertion from "participates in signal transduction" to "negatively regulates Rho/Rac signal transduction", which is what the evidence supports. The direction is not inferred from the domain. PMID:27226243 shows that removing ARHGAP23 raises GTP-bound RhoA and that adding it back lowers RhoA to the level achieved by the inhibitor C3 transferase; PMID:28114311 shows that removing it raises Rac1 to constitutively-active levels. Both replacement terms are proposed because both substrates have direct cellular evidence, with the caveat recorded elsewhere in this review that the Rac1 evidence is less consistent across cell types than the RhoA evidence. A curator who wanted to be maximally conservative could take GO:0035024 alone. Propagation Review Root cause: TERM SCOPING PROBLEM Failure modes: ROLE CONFLATION GRANULARITY MISMATCH Sources checked: InterPro:IPR000198 · RhoGAP domain InterPro signature SUPPORTS SOURCE BUT NOT TARGET The signature genuinely identifies a RhoGAP domain in ARHGAP23; the defect is in the term the mapping emits, which casts a negative regulator of a pathway as a participant in it. This is a property of the InterPro2GO mapping and applies to every protein carrying this signature, not to ARHGAP23 specifically. Proposed replacements: negative regulation of Rho protein signal transduction negative regulation of Rac protein signal transduction Supporting Evidence: PMID:27226243 Moreover, overexpressing Arhgap23 in MDA/shPk1 cells suppressed the elevation of active RhoA to a similar level as treatment with the active RhoA inhibitor, C3 transferase (Supplementary Fig. 3d). PMID:28114311 A Rac FRET biosensor [39] confirmed that shRNA against ARHGAP23 increased Rac1 activity to levels indistinguishable from constitutively active Raichu Rac V12 (Fig 5D and 5E), demonstrating that ARHGAP23 functions as a novel Rac1 GAP in adhesion maturation of migratory CHO.K1 cells as well as synapse maturation in neurons. |
| GO:0051056 regulation of small GTPase mediated signal transduction | TAS Reactome:R-HSA-9012999 | MODIFY | Summary: True, correctly placed on the regulation branch, and too general. It is a pathway-membership projection: every participant in Reactome's RHO GTPase cycle receives this term, so it records that ARHGAP23 is in the cycle rather than what it does there. Reason: Unlike the GO:0007165 row above, there is nothing wrong with this term - the replacements are its descendants, verified on two independent services. The case for specializing is that three cellular studies establish both the family of GTPase and the sign of the regulation, and the process branch, unlike the molecular-function branch, still has terms that can carry that. Leaving the annotation at GO:0051056 discards information that is available and checkable. This is the one place in the record where the substrate distinction survives. The three GO:0005096 rows are indistinguishable even though they descend from a RHOA reaction, a RAC1 reaction and a phylogenetic node whose experimental grounding is a Cdc42 GAP, because the specific GAP-activity terms no longer exist. GO:0035024 and GO:0035021 therefore carry the only machine-readable statement of substrate this gene can have. Propagation Review Root cause: TERM SCOPING PROBLEM Failure modes: GRANULARITY MISMATCH Sources checked: Reactome:R-HSA-9012999 · Reactome pathway "RHO GTPase cycle" SUPPORTS TRANSFER A pathway-membership projection distributed over every participant. Correct for this gene; uninformative because it cannot distinguish a GAP from a GEF, a GDI or the GTPase itself. Proposed replacements: negative regulation of Rho protein signal transduction negative regulation of Rac protein signal transduction Supporting Evidence: PMID:27226243 Following siRNA-mediated knockdown of either Pk1 or Arhgap21/23, the level of GTP-bound, active RhoA was significantly enhanced (Fig. 4a), and active Rac1 was either unchanged or decreased, respectively PMID:38970683 In PKP4-KO cells, ARHGAP23 depletion correlated with a loss of stress fibers, suggesting that ARHGAP23 is active in the PKP4-KO cell cytoplasm to reduce cortical RhoA activity and suppress cortical ring formation. |
| GO:0070062 extracellular exosome | HDA PMID:19199708 Proteomic analysis of human parotid gland exosomes by multid... | MARK AS OVER ANNOTATED | Summary: A single identification in a bulk MudPIT inventory of human parotid-gland exosomes. Nothing about ARHGAP23's biology predicts or requires an extracellular location, and no study has followed the observation up. Reason: Not removed, because the detection is a real measurement and the annotation honestly carries the HDA code that says so. Marked over-annotated because "located_in extracellular exosome" reads as a statement about where the protein resides and functions, which this evidence does not support. ARHGAP23 has no signal peptide and no transmembrane segment in its UniProt record, is a 1491-residue protein whose every characterized pool is on the inner face of a membrane or in the cytoplasm, and the study reports no enrichment, no orthogonal validation and no secretion route for it. The generic caution about large mass-spectrometry inventories applies with particular force to a protein that is Tdark in Pharos and has no orthogonal localization data outside the handful of papers reviewed here. Retaining it as over-annotated keeps the observation visible to anyone who later finds a reason for it, without letting it stand alongside the experimentally grounded localizations proposed below. |
| GO:0005938 cell cortex | IDA PMID:27226243 A lateral signalling pathway coordinates shape volatility du... | NEW | Summary: Proposed new annotation. ARHGAP23 is localized to the cell cortex, specifically the non-protrusive lateral cortex flanking active lamellipodia, and its presence there requires the planar cell polarity protein Prickle1. Reason: This is the localization the gene's mechanism actually depends on, and it is missing from the record entirely - GOA currently offers only cytosol, twice, from a Reactome compartment label. Two independent studies in different cell types support it. In MDA-MB-231 cells ARHGAP23 is at the lateral cortex and disappears from the periphery when Prickle1 is knocked down, which is a positional dependency rather than a static image. In keratinocytes overexpressed ARHGAP23 is at the cell cortex, where the authors interpret it as inactivating RhoA to prevent cortical ring formation. GO:0005938 rather than a plasma-membrane term because what is observed is the region just beneath the membrane, and because the functional claim is about a cortical actomyosin network rather than about membrane insertion - the protein has no transmembrane segment. The Prickle1 dependence is recorded in the reason rather than as a separate term because GO's localization terms cannot express what recruits a protein. Supporting Evidence: PMID:27226243 In migrating cells, we find that Pk1 and Arhgap21/23 are located at non-protrusive membranes that are lateral to active protrusions. PMID:27226243 Moreover, on Pk1 knockdown, Arhgap23 was absent from the periphery following ACM treatment (Fig. 2g) PMID:38970683 overexpressed ARHGAP23 partially localized at the cell cortex where it could inactivate RhoA to prevent cortical actin ring formation and led to increased stress fibers in WT cells |
| GO:0005925 focal adhesion | IDA PMID:32203420 Systems analysis of RhoGEF and RhoGAP regulatory proteins re... | NEW | Summary: Proposed new annotation. ARHGAP23 is one of the focal-adhesion-enriched RhoGEF/RhoGAP regulators in a TIRF localization screen, with a distinctive pericentric distribution - concentrated at the rim of each adhesion and lower at its centre. Reason: The primary evidence is the TIRF screen in PMID:32203420, whose Supplementary Table 4 records for ARHGAP23 "Focal adhesion localization found by TIRF imaging (GAP-deficient arginine finger mutant), pericentric", and whose figure places it among the FA-enriched regulators. Two caveats are recorded rather than hidden. The imaged construct was a point mutant, so this is the localization of a catalytically-modified protein; and that mutant, R986K, does not in fact target the annotated arginine finger (see the bioinformatics results), which if anything makes it more likely that the construct behaves like the wild-type protein, but is a reason not to describe the construct as the paper does. Independent support comes from function rather than imaging: in PMID:28114311, ARHGAP23 knockdown in CHO.K1 cells produces nascent adhesions and its overexpression produces enlarged, mature adhesions, which places the protein in the adhesion's regulatory neighbourhood by perturbation. The two lines are different in kind, which is why the annotation is proposed on the imaging evidence and the perturbation evidence supports the separate process term below. Supporting Evidence: file:human/ARHGAP23/ARHGAP23-bioinformatics/RESULTS.md Focal adhesion localization found by TIRF imaging (GAP-deficient arginine finger mutant), pericentric. PMID:28114311 Conversely, cells expressing GFP-tagged ARHGAP23 exhibited increased adhesion maturation |
| GO:0051893 regulation of focal adhesion assembly | IMP PMID:28114311 RhoGTPase Regulators Orchestrate Distinct Stages of Synaptic... | NEW | Summary: Proposed new annotation. Loss and gain of ARHGAP23 move focal adhesions in opposite directions: knockdown yields nascent adhesions, overexpression yields mature ones. Reason: Proposed at the unsigned "regulation of" level deliberately, even though the reported direction is consistent, because the mechanism does not sit comfortably with the sign. RhoA promotes focal adhesion maturation, so a RhoA GAP that increases adhesion maturation on overexpression is doing something indirect; the authors attribute it to Rac1 inactivation relieving the Rac-Rho antagonism. Asserting positive regulation would encode a directional claim that rests on one cell line and on a mechanism the same paper explains as a knock-on effect. The evidence is nonetheless real and quantified, both directions were tested in the same system, and the adhesion phenotype is independently echoed by the focal-adhesion localization in PMID:32203420 and by the altered focal adhesion dynamics reported on ARHGAP21/23 knockdown in PMID:27226243. A curator with access to further cell types may well be able to sign it. Supporting Evidence: PMID:28114311 Conversely, cells expressing GFP-tagged ARHGAP23 exhibited increased adhesion maturation file:human/ARHGAP23/ARHGAP23-bioinformatics/RESULTS.md Focal adhesion localization found by TIRF imaging (GAP-deficient arginine finger mutant), pericentric. |
Loading supporting content…
Download this section (compressed HTML)Q: GO:0005100 "Rho GTPase activator activity" is obsolete and merged into GO:0005096, whose narrow synonyms now include the Rho, Rac, Ras, Rab, Ran, Ral, Rap, ARF and Sar GAP activities, and which has no is_a children. Was the intention that substrate specificity move to the process branch, where GO:0035024 and GO:0035021 still exist? If so, the convention is not discoverable from the annotations: for ARHGAP23, three GO:0005096 rows derived from a RHOA reaction, a RAC1 reaction and a Cdc42-grounded phylogenetic node are textually identical, and the specificity is recoverable only by reading the Reactome reaction names.
Q: Should the InterPro2GO mapping from IPR000198 (RhoGAP domain) emit GO:0007165 signal transduction? A GAP terminates Rho signalling rather than transducing it, and GO:0035024 is not a descendant of GO:0007165, so the mapping places every RhoGAP-domain protein on the participant branch rather than the regulator branch. This affects the whole family, not just ARHGAP23.
Q: Is ARHGAP23's Rac1-directed activity cell-type dependent? Two cellular biosensor studies report it (PMID:28114311 in CHO.K1 and neurons, PMID:32203420 in the family-wide screen), but PMID:27226243 measured Rac1 in MDA-MB-231 cells in the same experiment that showed RhoA elevation on ARHGAP21/23 knockdown and found active Rac1 decreased rather than increased. Is the decrease an indirect consequence of Rho-Rac antagonism, or does ARHGAP23 act only on RhoA in that cell type?
Q: Which domain targets ARHGAP23 to its several membrane addresses? The protein is recruited to the lateral cortex by Prickle1, is found in complex with plakophilin 4, is enriched at focal adhesions, and binds VE-cadherin, yet its PDZ domain has never been assayed - it failed amplification for the human PDZome. Plakophilin 4 has a PDZ-binding motif, which makes the PDZ domain the obvious candidate, but that has not been tested.
Q: What PMID:28114311 measures for ARHGAP23 is focal adhesion *maturation* - knockdown gives nascent adhesions, overexpression gives enlarged ones - but the nearest GO term is GO:0051893 regulation of focal adhesion *assembly*, whose only neighbours are the signed forms and GO:0048041. Is maturation intended to fall under assembly, or is this the same expressivity gap as the missing substrate-specific GAP activity terms, one branch over? The annotation proposed here uses GO:0051893 because it is the closest available, not because the fit is exact.
Q: Does anything distinguish ARHGAP23 from ARHGAP21 functionally beyond expression pattern? They share the PDZ-PH-RhoGAP architecture, are treated as a redundant pair in the Prickle1 work, and are annotated from the same PANTHER node. The one clean discrimination in the literature points the other way from the usual assumption: plakophilin 4 co-purifies ARHGAP23 but not ARHGAP21, and the ARHGAP21 experimental grounding is a Cdc42 GAP activity that ARHGAP23's own screen scores negative.
Experiment: Purify the isolated ARHGAP23 Rho-GAP domain (residues 905-1097) and measure single-turnover GTP hydrolysis against RhoA, Rac1 and Cdc42, with the R942A or R942K mutant as the negative control and a purified ARHGAP1 domain as the positive control. This is the assay the field has never done for this protein - the systematic in-vitro survey of the family omitted it - and it would simultaneously establish the intrinsic substrate preference and whether R942 is the catalytic residue.
Experiment: Express R942K and R986K side by side in the cellular FRET biosensor assay used in PMID:32203420, reading out RhoA and Rac1. This directly resolves the discrepancy between the residue the literature treats as ARHGAP23's arginine finger and the one UniProt and the transition-state structure identify, and it is a small addition to an assay that already works for this protein.
Experiment: Delete or mutate the PDZ domain (residues 71-155) and ask which of ARHGAP23's localizations survive: the Prickle1-dependent lateral cortex, the plakophilin-4-associated pool, and the focal-adhesion pool. Since the domain could not be amplified for the human PDZome, a domain-swap with the ARHGAP21 PDZ domain would be an informative parallel arm - it would test whether the one documented difference between the paralogs, plakophilin 4 binding, is PDZ-encoded.
Experiment: Repeat the RhoA and Rac1 pulldowns of PMID:27226243 in MDA-MB-231 cells with ARHGAP23 silenced alone rather than together with ARHGAP21, and in parallel in a cell type where the Rac1 result was positive. The published experiments knock the two paralogs down together, so the conflicting Rac1 result cannot currently be attributed to either protein.
Experiment: Generate an Arhgap23 knockout mouse or, more cheaply, a conditional lymphatic-endothelial knockout, given that Arhgap23 is a FOXC1/FOXC2 target reduced in lymphatic-endothelial Foxc1;Foxc2 double knockouts. No loss-of-function organismal phenotype exists for this gene in any species, which is why every process statement in this review rests on cell culture.
What is not known — curated, literature-grounded statements of the open unknowns (the inverse of core functions).
Gap: No GO molecular-function term can record which GTPase ARHGAP23 acts on. GO:0005100 "Rho GTPase activator activity" is obsolete and merged into GO:0005096, which now carries "Rho GAP activity", "Rho GTPase activator activity", "Rac GAP activity" and "Rac GTPase activator activity" as narrow synonyms and has no is_a children. The consequence is visible in this gene's own record: three GO:0005096 rows descend from a RHOA reaction, a RAC1 reaction and a phylogenetic node whose experimental grounding is a Cdc42 GAP, and all three are identical.
OPEN ONTOLOGY MF_DARK
What is known: The substrate profile itself is not unknown. A cellular FRET screen scored ARHGAP23 significantly active on RhoA and Rac1 and inactive on Cdc42, and Reactome models the RHOA and RAC1 reactions separately. The information exists, is curated upstream, and is destroyed at the point of GO annotation.
Significance: Substrate specificity is the only thing that distinguishes one of the 66 human RhoGAPs from another. A molecular-function branch that cannot express it makes every member of the family look interchangeable, which is precisely the condition under which annotation transfer between paralogs goes unchallenged - the failure mode that produced a false GAP-activity assertion for ARHGAP11B and a residue-free one for ARHGAP36 elsewhere in this batch.
What would resolve it: Either restore substrate-specific children under GO:0005096, or establish a convention for recording the GTPase as an input on the activity. The process branch already solves the problem - GO:0035024 and GO:0035021 exist and are used in this review - so the asymmetry between the two branches is the concrete thing to raise with the ontology editors.
Gap: It is not known whether ARHGAP23's catalytic arginine is required for its measured activity, and the one published point mutant described as its arginine-finger mutant does not target that residue. ARHGAP23-R986K, from the family-wide screen, mutates an arginine that aligns to ARHGAP1 R323 - inside the GAP:GTPase interface but not in contact with the transition state - while the annotated arginine finger is R942, which aligns to ARHGAP1 R282.
OPEN BIOLOGYCURATION RESIDUAL_SUBGAP
What is known: The residue itself is firmly established: R942 is UniProt's annotated Site, it is reciprocally in register with ARHGAP1's structurally resolved arginine finger, and the surrounding interface is essentially identical to that of an experimentally active paralog. What is missing is the functional test, and the mutant that exists was used for localization imaging rather than as an activity control - the screen's catalytic controls were ARHGAP11A, ARHGAP40, ARHGAP4, FAM13A and SYDE2.
Significance: Every assertion that ARHGAP23 is a GAP currently rests on cellular assays in which the protein's own catalysis was never isolated from its scaffolding and localization roles. A catalytically dead mutant that still localizes correctly is the standard way to separate the two, and for this protein the field believes it has one when it does not. The discrepancy also means a curator reading the supplementary table would record the wrong residue.
What would resolve it: An in-vitro GAP assay on the isolated Rho-GAP domain against RhoA, Rac1 and Cdc42 with R942 mutated as the negative control, and a side-by-side comparison of R942K and R986K in the cellular biosensor assay to establish which position the activity depends on.
Provenance (the field's own admissions):
Gap: No ligand is known for ARHGAP23's PDZ domain, and the domain has never been assayed. It was one of ten PDZ domains that could not be amplified for the human PDZome resource, so it is absent from the one systematic experimental survey of human PDZ binding.
OPEN BIOLOGY MF_DARK
What is known: The domain is real and is the defining feature of the ARHGAP21/ARHGAP23 architecture; UniProt places it at residues 71-155. The only published ARHGAP23 PDZ-ligand claim, a PTEN C-terminal peptide, comes from docking with no experimental validation.
Significance: Every characterized ARHGAP23 interaction is with a junctional or polarity scaffold - plakophilin 4, Prickle1, VE-cadherin, alpha-catenin - which is exactly the class of partner a PDZ domain would be expected to engage. If the PDZ domain is the targeting module, it explains how one enzyme is directed to several different membrane addresses; if it is not, something else is doing that work and has not been identified.
Provenance (the field's own admissions):
Loading supporting content…
Download this section (compressed HTML)Loading supporting content…
Download this section (compressed HTML)Loading supporting content…
Download this section (compressed HTML)Loading supporting content…
Download this section (compressed HTML)