ABRACL is an 81-residue protein and the human member of the Costars family, a small eukaryotic family that is usually but not universally single-copy per genome (PANTHER PTHR46334 resolves into two subfamilies, and Arabidopsis carries two members) and absent from fungi. Its NMR structure is a winged helix-like fold - three alpha-helices packed against four antiparallel beta-strands - but it lacks DNA-binding activity, and the surface that in canonical winged-helix proteins accommodates the DNA backbone is instead a conserved hydrophobic groove, taken to be a protein-interaction site. ABRACL binds filamentous actin directly but weakly in vitro, and in cells it associates with the actin-severing protein cofilin-1 (CFL1), with which it colocalises at the leading edge of lamellipodia in migrating cells. On purified proteins it blunts cofilin-stimulated depolymerisation of F-actin in a dose-dependent way without displacing cofilin from filaments, so it acts on cofilin's activity rather than competing for the filament. Cells depleted of ABRACL migrate less and change shape, and in Dictyostelium loss of the ortholog Costars (cosA) slows chemotactic movement and disturbs F-actin distribution, defects that expression of the human protein reverses. The direction in which ABRACL shifts the F-/G-actin balance differs between the Dictyostelium and human systems, so the protein is best described as a modulator of actin filament dynamics whose sign is context-dependent. It is broadly expressed across human tissues, elevated in several carcinomas, and in the embryonic mammalian forebrain it marks subpallial progenitors, migrating interneurons and postmitotic neurons.
Definition: Binds to and decreases the activity of an actin severing protein.
Justification: GO already contains the positive half of this pair - GO:0000513 actin severing activator activity, defined as binding to and increasing the activity of an actin severing protein - but no inhibitor. ABRACL is a clean instance of the missing half: it binds cofilin, blunts cofilin-stimulated F-actin depolymerisation dose-dependently, and does so without preventing cofilin from binding the filament, so it is acting on the severing protein's activity rather than competing for actin. The proposed definition is the existing activator definition with the sign reversed, so the term needs no new modelling machinery. Without it, a protein whose entire known function is dampening a severing factor can be annotated only to actin filament binding, which is true of ABRACL but conveys none of the mechanism, or to nothing at all, which is the current state.
Parent term: molecular function inhibitor activity
Supporting Evidence:
| GO Term | Evidence | Action | Reason |
|---|---|---|---|
| GO:0032970 regulation of actin filament-based process | IBA GO_REF:0000033 | ACCEPT | Summary: Accepted as the core process, and deliberately left at this level of specificity. The transfer itself is as clean as phylogenetic transfers get. PANTHER PTHR46334 does contain two subfamilies, but the transfer stays inside one of them: both the donor and the target are PTHR46334:SF1, while the separate SF3 is plant-restricted (Arabidopsis Q8LBN7, Eutrema B4YYA9). So no subfamily boundary is crossed, and the WITH/FROM donor dictyBase:DDB_G0272861 is Dictyostelium cosA, which carries its own experimental annotation to this same term. More than that, the transfer has been tested directly: the human protein expressed in cosA-null Dictyostelium restores the mutant's motility and actin phenotypes. Functional complementation across that distance is a stronger warrant for an ortholog transfer than any sequence statistic, and independent human experimental work reaches the same conclusion from the other direction. The interesting question is whether to make the term more specific, and the answer is no. Three lines of evidence disagree about the SIGN of ABRACL's effect on the filamentous/globular actin balance. Deleting the Dictyostelium ortholog RAISES cytoskeleton-associated actin; depleting human ABRACL LOWERS the cellular F-/G-actin ratio; and purified human ABRACL on its own INHIBITS actin polymerisation, a result the authors themselves flag as contradicting their cellular data. Only in the presence of cofilin does ABRACL clearly favour F-actin. So GO:0030835 negative regulation of actin filament depolymerization and GO:0030837 negative regulation of actin filament polymerization are each supported by one part of the evidence and contradicted by another. The unsigned parent is not an under-annotation here; it is the strongest claim every dataset supports, and the PAINT curator was right to stop at it. Propagation Review Root cause: NO FAILURE CORE Sources checked: dictyBase:DDB_G0272861 · cosA (Dictyostelium discoideum Costars) SUPPORTS TRANSFER Donor carries an experimental annotation to the same term, and the human protein rescues the cosA-null phenotype, so the transfer has been tested rather than assumed. PANTHER:PTN000507089 · Costars ancestral node (PTHR46334) SUPPORTS TRANSFER PANTHER metadata records subfamilies: 2 for PTHR46334, so this is not the single-copy, paralog-free node an earlier draft of this review claimed. The transfer nonetheless stays inside one subfamily: the Dictyostelium cosA donor (Q558Y7) and human ABRACL (Q9P1F3) are both PTHR46334:SF1 (13 members, animals plus Dictyostelium plus most plant entries), while the separate SF3 is plant-restricted (2 members, Arabidopsis Q8LBN7 and Eutrema B4YYA9). No subfamily boundary is crossed, so the node supports the transfer - but for that reason, not for the absence of paralogs. Supporting Evidence: PMID:20940261 Expressing cosA or its human counterpart mCostars eliminated abnormalities of cosA(-) cells. PMID:20940261 cosA(-) cells exhibited changes in the actin cytoskeleton, showing aberrant distribution of F-actin in fluorescence cell staining and an increased amount of cytoskeleton-associated actin. PMID:33670794 both ABRACL-knockdown and ABRACL-knockout cells displayed lower cellular F-/G-actin ratios compared to the control cells PMID:33670794 These in vitro results appeared to contradict the above-mentioned finding of decreased F/G-actin ratios in cells lacking ABRACL file:human/ABRACL/ABRACL-deep-research-affinage.md ABRACL is a small winged helix-like protein that regulates actin cytoskeletal dynamics and cell motility, a role conserved from the Dictyostelium ortholog Costars |
| GO:0005929 cilium | IEA GO_REF:0000044 | REMOVE | Summary: Removed. Traced back, this annotation rests on nothing that was ever a localisation observation. The chain has four links. GOA maps UniProtKB-SubCell:SL-0066 mechanically, as GO_REF:0000044 is designed to do. The UniProt line it maps is flagged ECO:0000305 - a curator inference, not an experimental record - and cites PubMed:37759737. That paper is an expression study of mouse and cat embryonic telencephalon which reports no ciliary localisation of its own; its single sentence on the subject attributes the claim to somebody else's high-throughput proteomics. In the same paragraph it reports evidence pointing the other way, that neither the transcript nor the protein is present in the ventricular zone, which is precisely where the ciliated progenitors sit. The proteomics study being cited is PMID:26638075, the Gupta et al. centrosome-cilium BioID interactome, and it is the only source of physical-interaction data for ABRACL in IntAct at all: five proximity hits, one publication, one detection method. Proximity-dependent biotinylation labels whatever comes within roughly ten nanometres of the bait over hours, so its output is a neighbourhood rather than a compartment assignment, and three of the four baits (SASS6, CNTRL, DCTN1) are centriolar or dynactin proteins rather than ciliary ones. The fourth, RPGRIP1L, IS genuinely ciliary - a transition-zone protein - so the bait panel is not uniformly non-ciliary and that should be conceded. It does not rescue the annotation: proximity to one ciliary bait over hours of labelling is still a neighbourhood observation, not a localisation call. A family-level check points the same way: Costars is retained in Arabidopsis, rice, maize and Dictyostelium, none of which builds a cilium, basal body or centriole, while IFT88, IFT52, BBS1 and ARL13B are absent from every one of those proteomes. None of this proves ABRACL never enters a human cilium, and a lineage-specific role would not show up in a phyletic comparison. What it establishes is that no evidence for the annotation exists at any link in the chain, and that the conserved function of the family cannot be ciliary. An unsupported compartment assignment on an otherwise sparsely annotated protein is worse than no assignment, because it is the kind of statement downstream tools propagate. Reason: The UniProt subcellular location this IEA maps is a curator inference (ECO:0000305) whose cited paper reports no ciliary localisation and in fact places the protein outside the ciliated ventricular zone. The only primary data behind it are proximity-labelling hits from a single BioID screen with centriolar baits, which do not establish a localisation, and the Costars family is retained in four proteomes that build no cilium at all. Propagation Review Root cause: SOURCE WEAK OR INFERRED Failure modes: SOURCE EVIDENCE WEAK SOURCE MISCITATION COMPARTMENT OR COMPLEX MISMATCH Sources checked: UniProtKB-SubCell:SL-0066 · Cilium (UniProt subcellular location) SOURCE WEAK OR INFERRED The GOA row is a faithful mapping of this location term; the fault is upstream, in the ECO:0000305 statement being mapped, not in the mapping. PMID:37759737 · Troumpoukis et al. 2023, Abracl-expressing populations in embryonic telencephalon SOURCE BAD Cited by UniProt as the evidence for cilium, but it reports no ciliary localisation and states that neither Abracl mRNA nor protein is present in the ventricular zone. PMID:26638075 · Gupta et al. 2015, centrosome-cilium BioID interactome SOURCE WEAK OR INFERRED The actual primary source, two citations removed. Five proximity-labelling hits with centriolar and dynactin baits; a BioID neighbourhood is not a compartment call. Supporting Evidence: file:human/ABRACL/ABRACL-uniprot.txt CC {ECO:0000305|PubMed:37759737}. PMID:37759737 a high-throughput proteomics study in non-neuronal cell lines has shown that Abracl is associated with the primary cilia PMID:37759737 our results show that neither Abracl mRNA nor Abracl were expressed in the VZ file:human/ABRACL/ABRACL-bioinformatics/RESULTS.md Detection methods observed: proximity-dependent biotin identification. |
| GO:0030027 lamellipodium | IEA GO_REF:0000044 | ACCEPT | Summary: Accepted, and worth separating sharply from the cilium row it was generated alongside. Both are IEA mappings of UniProt subcellular locations under GO_REF:0000044, but the statements being mapped are not comparable: the cilium line is ECO:0000305, a curator inference, whereas the lamellipodium line is ECO:0000269, backed by experiment. The experiment is a good one for a localisation call. Endogenous ABRACL was stained with a specific antibody in migrating MDA-MB-231 cells fixed during wound closure, and the signal coincided with F-actin at the leading edge; ABRACL/cofilin colocalisation in the same cells was likewise concentrated at lamellipodia. This is a direct assay on the untagged protein at native levels, so the annotation would properly be IDA against PMID:33670794 rather than IEA against a location vocabulary. The recommendation is to upgrade the evidence code, not to change the term. Supporting Evidence: file:human/ABRACL/ABRACL-uniprot.txt CC -!- SUBCELLULAR LOCATION: Cell projection, lamellipodium PMID:33670794 fluorescence signals for endogenous ABRACL and F-actin were colocalized at the leading edge of lamellipodia PMID:33670794 the colocalization was particularly evident at lamellipodia |
| GO:0051015 actin filament binding | IDA PMID:33670794 Human Costars Family Protein ABRACL Modulates Actin Dynamics... | NEW | Summary: Proposed new annotation, and the first molecular function this gene would carry. ABRACL currently has none at all, which is the largest single gap in its record: a protein with a solved structure, a named interaction partner and two in vitro assays on purified protein has no MF row. The assay is direct and on the right material. GST-ABRACL was expressed in E. coli, cleaved with Factor Xa and the tag removed, so the co-sedimentation was performed on untagged protein; the pellet signal was quantified across independent experiments against a no-F-actin control. UniProt records the same conclusion in its SUBUNIT line. The annotation is offered with its limitation stated rather than buried: the authors describe the binding as very weak, and the fraction of ABRACL recovered in the pellet was small. That is a reason to record the activity honestly, not to omit it - a weak direct binding measured against a proper control is evidence, and it is the only representable molecular function this protein has. The activity that actually matters mechanistically, inhibition of a cofilin-type severing protein, has no GO term at all and is proposed as a new one below. Supporting Evidence: PMID:33670794 These results indicated that purified recombinant ABRACL bound to F-actin, although very weakly, under the in vitro experimental conditions. file:human/ABRACL/ABRACL-uniprot.txt CC -!- SUBUNIT: Weakly interacts with F-actin (PubMed:33670794). Interacts |
| GO:0030335 positive regulation of cell migration | IMP PMID:33670794 Human Costars Family Protein ABRACL Modulates Actin Dynamics... | NEW | Summary: Proposed new annotation. The migration phenotype is the best-replicated thing known about this gene and it is not represented in GO at all. The evidence is bidirectional and uses more than one perturbation. Transient overexpression increases Transwell migration of MDA-MB-231 cells; two independent shRNAs reduce it in HCT116; re-expression from an RNAi-resistant construct restores it, which excludes off-target knockdown; and five independent CRISPR knockout clones reduce it again. The Dictyostelium ortholog gives the same sign - cosA-null cells move more slowly toward cAMP - so this is not a cancer-cell-line peculiarity. One qualification belongs on the record rather than in a footnote: in MDA-MB-231 the basal migration of knockout cells was unaffected, and the defect appeared only under EGF stimulation. ABRACL is therefore a modulator of stimulated migration in at least some contexts rather than a component required for baseline motility, and the positive-regulation term rather than a participation term is the right shape for that. Supporting Evidence: PMID:33670794 All five ABRACL-knockout clones examined displayed significantly reduced migration in the Transwell assay PMID:33670794 In this cell line, depletion of ABRACL expression did not affect the basal migration activity PMID:33670794 ABRACL-knockout cells exhibited significantly decreased migration compared to the parental control cells PMID:20940261 Analysis of cell motion in cAMP gradients revealed decreased speed but wild-type-like directional persistence of cosA(-) cells, suggesting a defect in the cellular machinery for motility rather than for chemotactic orientation. |
Loading supporting content…
Download this section (compressed HTML)Q: Does ABRACL bind cofilin directly? Every line of evidence for the interaction is in cells - colocalisation, proximity ligation, co-immunoprecipitation - and the in vitro experiment that follows is functional rather than binary: ABRACL blunts cofilin's effect on filaments. A direct binary binding measurement between the two purified proteins, with an affinity, has not been reported, so it remains possible that the two meet on the filament rather than in solution.
Q: Which way does ABRACL push the actin balance, and why do the systems disagree? Deleting the Dictyostelium ortholog raises cytoskeleton-associated actin, depleting human ABRACL lowers the F-/G-actin ratio, and purified ABRACL alone inhibits polymerisation. The authors of the human study flag the contradiction themselves. Until it is resolved, no signed regulatory term can be justified.
Q: What occupies the conserved hydrophobic groove? The NMR structure identifies a groove shared with the ABRA C-terminal domain and proposes it as a protein-interaction site, but no residue in it has been mutated and no partner has been mapped onto it. Whether cofilin binds there is the obvious first test.
Q: Is there anything to the centrosome connection? The cilium annotation should go, but the underlying BioID data are real: ABRACL was labelled by SASS6, CNTRL, DCTN1 and RPGRIP1L baits. Cofilin and actin have roles at the centrosome, so a genuine centrosomal pool is not absurd - it has simply never been tested by any method other than proximity labelling.
Q: Is the nuclear pool reported in neural progenitors real, and is it functional? A 9 kDa protein diffuses through nuclear pores whether or not it is targeted, and the supporting experiment is overexpression in a neuroblastoma line. Distinguishing passive equilibration from active retention needs endogenous tagging.
Experiment: Measure the affinity of purified ABRACL for purified CFL1 by ITC or SPR in the absence of actin, then map the interface by NMR chemical-shift perturbation onto the 2L2O structure and test predicted groove residues by mutation, scoring loss of cofilin binding and loss of the inhibitory effect on depolymerisation together. A separation-of-function mutant that binds F-actin but not cofilin would settle which activity carries the cellular phenotype.
Hypothesis: ABRACL binds cofilin directly, through the conserved hydrophobic groove identified in the NMR structure.
Type: biophysics
Experiment: Use single-filament TIRF microscopy with fluorescently labelled actin and cofilin, titrating ABRACL, and measure severing events per filament length per unit time, cofilin cluster nucleation density, and elongation rate as separate observables. Bulk pyrene assays cannot distinguish these, which is why the sign of ABRACL's effect is currently ambiguous.
Hypothesis: ABRACL reduces cofilin-mediated severing rather than altering filament elongation or cofilin binding.
Type: in vitro reconstitution
Experiment: Knock a fluorescent tag into the endogenous ABRACL locus in a ciliated human line such as RPE-1, serum-starve to induce primary cilia, and image the tagged protein against ciliary and centriolar markers, alongside neural progenitors for the nuclear question. A negative result is as useful as a positive one and would close the question that the current annotation leaves open.
Hypothesis: The cilium and nucleus localisations attributed to ABRACL are artefacts of proximity labelling and of overexpression respectively.
Type: cell biology
Experiment: Perform the same F-/G-actin fractionation in both systems side by side, then repeat in Dictyostelium cells whose cofilin has been replaced with the human protein, and in human cells expressing the Dictyostelium ortholog. Because the human protein already rescues cosA-null cells, the reciprocal swap tests whether the partner rather than the Costars protein sets the direction.
Hypothesis: The opposite sign of the Dictyostelium and human actin phenotypes reflects a difference in cofilin regulation between the two systems rather than a divergence in ABRACL itself.
Type: comparative cell biology
What is not known — curated, literature-grounded statements of the open unknowns (the inverse of core functions).
Gap: GO cannot express what ABRACL does. It binds cofilin and reduces cofilin-stimulated filament disassembly without competing for the filament, which is an inhibitor activity directed at an actin-severing protein. GO carries the activator of exactly this activity, GO:0000513 actin severing activator activity, defined as binding to and increasing the activity of an actin severing protein, and no inhibitor counterpart.
OPEN ONTOLOGY MF_DARK
What is known: What is established is the effect and its independence from displacement: ABRACL blunts cofilin-stimulated pyrene F-actin depolymerisation dose-dependently, and does not prevent cofilin from co-sedimenting with F-actin. What is missing is a term, not an experiment.
Significance: This is a large part of why a protein with a solved structure, a named partner and two assays on purified protein carries no molecular function annotation at all. The gap is not specific to ABRACL: any protein that dampens a severing factor rather than sequestering actin faces the same missing branch.
What would resolve it: Create the sibling term, actin severing inhibitor activity, under GO:0140678 molecular function inhibitor activity, mirroring GO:0000513.
Provenance (the field's own admissions):
Gap: The direction in which ABRACL shifts the filamentous/globular actin balance is unresolved. Losing the Dictyostelium ortholog raises cytoskeleton-associated actin; depleting human ABRACL lowers the F-/G-actin ratio; and purified human ABRACL on its own inhibits polymerisation. No single signed statement fits all three.
OPEN BIOLOGYCURATION BP_DARK
What is known: That ABRACL modulates actin filament dynamics is established from three independent systems, and the cofilin-dependent arm of the mechanism is established on purified proteins. Only the sign of the net cellular effect is unsettled.
Significance: It determines whether the correct process term is negative regulation of actin filament depolymerization, negative regulation of actin filament polymerization, or neither, and it is the reason this review leaves the annotation at the unsigned parent rather than refining it.
What would resolve it: Single-filament TIRF microscopy with purified actin, cofilin and ABRACL at defined ratios, measuring severing frequency, elongation rate and cofilin cluster nucleation separately, rather than inferring the sign from bulk pyrene traces or from steady-state cellular fractionation.
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)