AHNAK2 is a 5795-residue human protein built from a short N-terminal segment carrying a single PDZ domain (residues 112-193), a very large central region of ~165-residue low-complexity repeats that occupies roughly two thirds of the chain, and a C-terminal segment of about 100 kDa. It has no catalytic domain, no nucleotide- or metal-binding site and no enzymatic activity of any kind; its architecture is that of a giant submembrane scaffold. The one molecular property measured directly is homodimerisation: the crystal structure of its PDZ domain is an intertwined, domain-swapped dimer in which a six-stranded antiparallel beta sheet is assembled from alternating strands of the two chains, with a preformed peptide-binding pocket occupied by a class III ligand. The same architecture is seen in the PDZ domain of periaxin, which is AHNAK2's closest structural relative. In cells AHNAK2 is cytosolic and plasma-membrane-associated. In mouse skeletal muscle, antibodies specific for AHNAK2 place it in the costameric network, co-distributing with vinculin at the sites that couple sarcomeric Z-discs laterally to the sarcolemma and the extracellular matrix, and it associates with dysferlin. Outside muscle, AHNAK2 is required for the stress-induced, non-classical export of fibroblast growth factor 1: on heat shock it associates with FGF1 and both move into the submembrane, F-actin-rich fraction, and loss of AHNAK2 largely abolishes stress-induced FGF1 release without affecting FGF2 release or Notch-inhibition-induced FGF1 release. Loss-of-function variants segregate with autosomal recessive Charcot-Marie-Tooth disease, and AHNAK2 expression is repeatedly associated with invasive behaviour in epithelial tumours. Beyond homodimerisation, the biochemical activity through which AHNAK2 performs any of these roles has not been established.
| GO Term | Evidence | Action | Reason |
|---|---|---|---|
| GO:0005515 protein binding | IPI PMID:17185750 AHNAK, a novel component of the dysferlin protein complex, r... | KEEP AS NON CORE | Summary: Dysferlin. Kept, but as non-core and with the paralogue caveat stated: the abstract of the cited paper reports only AHNAK, while UniProt's AHNAK2 entry carries the interaction at ECO:0000269 from a curator who read the full text. Reason: The cited paper's abstract is about AHNAK throughout, and human AHNAK received the reciprocal row from the same reference, so the first impression is of a paralogue transfer. It is not. UniProt's AHNAK2 entry states that AHNAK2 interacts with DYSF and that the interaction is direct and Ca(2+)-independent, tagged ECO:0000269|PubMed:17185750 - a manual experimental assertion made specifically on Q8IVF2 by a curator with access to the full text, which is subscription-only (Europe PMC reports inEPMC:N, hasPDF:N). QuickGO by reference shows the curation was deliberate rather than copied: reciprocal pairs were built on both sides, DYSF-with-Q8IVF2 and AHNAK2-with-O75923. Per the campaign rule I do not second-guess an experimental annotation whose full text I have not read. Kept as non-core because a bare "protein binding" states no molecular function, and because the assay bears on muscle physiology rather than on AHNAK2's one measured molecular property. The partner accession resolves to canonical Swiss-Prot DYSF (O75923, 2080 aa) - no TrEMBL or ORFeome substitution. Supporting Evidence: PMID:17185750 By coimmunoprecipitation coupled with mass spectrometry, we demonstrate that AHNAK interacts with dysferlin. |
| GO:0005515 protein binding | IPI PMID:17185750 AHNAK, a novel component of the dysferlin protein complex, r... | UNDECIDED | Summary: Myoferlin. The abstract attributes this interaction explicitly to AHNAK, and UniProt's AHNAK2 SUBUNIT comment records DYSF but not MYOF. The full text is inaccessible, so this is UNDECIDED rather than removed. Reason: This row is weaker than its DYSF sibling and the difference is checkable. The abstract says the myoferlin interaction was with the carboxyterminal domain of AHNAK, naming the paralogue; and UniProt's AHNAK2 SUBUNIT comment, written by a curator who did read the full text, records the DYSF interaction at ECO:0000269 and says nothing about MYOF. So a curator with the evidence in front of them carried one partner across to AHNAK2 and not the other, while GOA carries both. That asymmetry is a real signal but it is not a refutation, and the paper is subscription-only. CLAUDE.md is explicit that UNDECIDED is the correct action when the relevant publication cannot be accessed, and that REMOVE is never appropriate for an experimental annotation on the strength of a cached abstract. Partner accession checked: Q9NZM1 is canonical Swiss-Prot MYOF, 2061 aa. Supporting Evidence: PMID:17185750 As expected, the N-terminal domain of myoferlin also interacts with the carboxyterminal domain of AHNAK. |
| GO:0005634 nucleus | IBA GO_REF:0000033 | MARK AS OVER ANNOTATED | Summary: Three of the four protein donors are periaxin and the fourth is AHNAK; the AHNAK2 orthologue is absent from this donor set. In the donors the nuclear pool is a small N-terminal isoform, and the row sits on the full-length chain with no isoform qualifier. Reason: Resolving the WITH/FROM set gives mouse Prx, rat Prx, human PRX and mouse Ahnak. Mouse Ahnak2 (MGI:2144831), which is on this gene's cytoplasm row, is not on this one, so no orthologue-strength inference is available here at all. Each donor does carry its own experimental evidence for nucleus, so this is not a case of weak sources - the propagation itself is the problem. In AHNAK and in periaxin the nuclear pool is a small N-terminal PDZ-containing isoform that shuttles from the cytoplasm, not the giant chain; UniProt annotates no such isoform for AHNAK2, whose only short isoform (Q8IVF2-2, VSP_031550, residues 5003-5795) is C-terminal and lacks the PDZ domain entirely. The qualifier compounds it: is_active_in asserts that AHNAK2's activity occurs in the nucleus, and no activity of AHNAK2 has been localised anywhere. Human Protein Atlas immunofluorescence for AHNAK2, at Enhanced reliability, calls cytosol and plasma membrane and no nuclear compartment. That is a measurement declining to confirm rather than proof of absence, which is why this is marked over-annotated rather than removed. The discriminator against AHNAK, whose identical IBA row is defensible, is that human AHNAK independently holds GO:0005634 by ISS from its own mouse orthologue (PMID:21940993) - an ortholog-strength, same-gene transfer made by a curator reading the paper. AHNAK2 has no such transfer, and neither AHNAK2 nor AHNAK holds the term by its own experimental evidence. Propagation Review Root cause: PROPAGATION BAD Failure modes: WRONG ORTHOLOG OR PARALOG COMPARTMENT OR COMPLEX MISMATCH Sources checked: MGI:MGI:108176 · Mus musculus Prx, periaxin (O55103, Swiss-Prot, 1391 aa; this MGI id has 5 UniProt hits) SUPPORTS SOURCE BUT NOT TARGET Holds GO:0005634 by EXP (PMID:10671475) and IDA (PMID:15282162), so the source annotation is sound. Periaxin shares only the PDZ domain with AHNAK2 and its nuclear pool is a small isoform. MGI:MGI:1316648 · Mus musculus Ahnak (E9Q616, TrEMBL - no Swiss-Prot entry exists, 5656 aa) SUPPORTS SOURCE BUT NOT TARGET Holds GO:0005634 by IDA from PMID:21940993, where the nuclear species is explicitly the small 17-kDa AHNAK isoform rather than the 700-kDa chain. PANTHER:PTN001156025 · PANTHER tree node, not a protein NOT RELEVANT The inference structure rather than an evidence source. Its reach spans three PTHR23348 subfamilies - SF37 AHNAK2, SF41 AHNAK, SF42 periaxin. RGD:619960 · Rattus norvegicus Prx, periaxin (Q63425, Swiss-Prot, 1383 aa; 3 UniProt hits) SUPPORTS SOURCE BUT NOT TARGET Holds GO:0005634 by EXP, so the source annotation is sound - the same status as its two periaxin siblings on this row. Worth noting separately that its EXP is PMID:10671475, the same paper as mouse Prx's, so it adds a species rather than an independent characterisation; that is a weight observation, not a circular chain. UniProtKB:Q9BXM0 · Homo sapiens PRX, periaxin (Swiss-Prot, 1461 aa) SUPPORTS SOURCE BUT NOT TARGET Holds GO:0005634 by IDA from PMID:24633211. Third periaxin in a four-donor set, so this term is overwhelmingly a periaxin property. Supporting Evidence: PMID:21940993 These proteins interact in the cytoplasm, but the small AHNAK is also present in the nucleus. PMID:21940993 A small 17-kDa isoform of Periaxin similarly traffics between the cytoplasm and the nucleus to regulate mRNA splicing. file:genes/human/AHNAK2/AHNAK2-bioinformatics/RESULTS.md **The AHNAK2 ortholog is absent from two of the three IBA donor sets.** |
| GO:0005634 nucleus | IEA GO_REF:0000044 | MARK AS OVER ANNOTATED | Summary: A mechanical mapping of UniProt's sole SUBCELLULAR LOCATION line for AHNAK2, which is a bare ECO:0000250 "by similarity" with no source accession. Nothing measured stands behind it. Reason: GO_REF:0000044 converts whatever UniProt's SUBCELLULAR LOCATION comment says into a GO row, so the question reduces to what that comment carries. For AHNAK2 it is Nucleus with a bare ECO:0000250 and nothing else - the weakest statement UniProt makes, and here it does not even name the entry it was transferred from. Tracing the presumed donor does not help. Surveying the SUBCELLULAR LOCATION comment of all seven entries in this gene's donor set, every experimentally tagged nuclear location belongs to periaxin - human PRX at ECO:0000269|PubMed:24633211, mouse Prx and rat Prx both at ECO:0000269|PubMed:10671475 - while the three AHNAK-branch entries carry, respectively, a bare ECO:0000250 with no accession (AHNAK2), no evidence tag at all (human AHNAK), and an automatic machine rule (ECO:0000256|ARBA:ARBA00004123, mouse Ahnak). So there is no curated experimental nuclear location anywhere in the AHNAK branch of this family. GO separately holds a nucleus IDA on mouse Ahnak from PMID:21940993, which UniProt's unreviewed mouse entry has not absorbed; that IDA is on the small nuclear isoform. Against it, the one human AHNAK2 localisation measurement in GOA is the HPA immunofluorescence pair on this same record, Enhanced reliability, calling cytosol as main location and plasma membrane as additional, with no nuclear compartment among the calls. This is unsupported rather than refuted, so over-annotated rather than removed; the durable fix is upstream, in sourcing or withdrawing the by-similarity statement, and that is raised in suggested_questions. |
| GO:0005737 cytoplasm | IBA GO_REF:0000033 | KEEP AS NON CORE | Summary: The one IBA row whose donor set includes the AHNAK2 orthologue, and a genuinely heterogeneous set, so the general parent is the correct least-common-ancestor call. Kept as non-core because cytoplasm is true and uninformative. Reason: Seven resolvable donors, all with their own experimental evidence for cytoplasm or a descendant, and unlike the nucleus and splicing rows this set includes mouse Ahnak2 (MGI:2144831), the actual orthologue, plus a self-referential UniProtKB:Q8IVF2 that records a PAINT curator's own judgement for this gene and is valid rather than circular. The donors do not agree on a compartment - periaxins are curated cytoplasmic in Schwann cells, mouse Ahnak and mouse Ahnak2 both hold GO:0043034 costamere by IDA, and mouse Ahnak2 additionally holds GO:0030018 Z disc and GO:0030659 cytoplasmic vesicle membrane - so cytoplasm really is the least-common-ancestor of this donor set and there is no granularity defect to fix. I considered MODIFY down to costamere on the ACRV1 pattern, where an IBA lands above its donors' specific terms, and rejected it: narrowing this row would mean picking one donor's compartment over three others'. Costamere is proposed additively as its own row instead. Non-core because cytoplasm carries no information about what AHNAK2 does. Propagation Review Root cause: NO FAILURE NON CORE Sources checked: MGI:MGI:108176 · Mus musculus Prx, periaxin (O55103, Swiss-Prot) SUPPORTS TRANSFER Holds GO:0005737 by EXP (PMID:10671475, PMID:9488714) and IDA (PMID:15282162). MGI:MGI:1316648 · Mus musculus Ahnak (E9Q616, TrEMBL - no Swiss-Prot entry exists) SUPPORTS TRANSFER Holds GO:0005829 cytosol by IDA and GO:0043034 costamere by IDA (PMID:20833135), so its evidence sits at descendants of cytoplasm. MGI:MGI:2144831 · Mus musculus Ahnak2 (A0A7N9VR94, TrEMBL - no Swiss-Prot entry exists, 3501 aa; this MGI id has 3 UniProt hits) SUPPORTS TRANSFER The orthologue, and the only one in any of this gene's IBA donor sets. Holds GO:0043034 costamere (IDA, PMID:20833135), GO:0030018 Z disc and GO:0030659 cytoplasmic vesicle membrane (both IDA, PMID:15007166). PANTHER:PTN001156025 · PANTHER tree node, not a protein NOT RELEVANT The inference structure rather than an evidence source. RGD:619960 · Rattus norvegicus Prx, periaxin (Q63425, Swiss-Prot) SUPPORTS TRANSFER Holds GO:0005737 by EXP from PMID:10671475 and PMID:8155317. UniProtKB:Q09666 · Homo sapiens AHNAK, the paralogue (Swiss-Prot, 5890 aa) SUPPORTS TRANSFER Holds GO:0005737 by IDA (PMID:22057634). A paralogue donor is legitimate for IBA; it is flagged because AHNAK is the donor the other two IBA rows principally depend on. UniProtKB:Q8IVF2 · AHNAK2 itself - self-referential WITH/FROM SUPPORTS TRANSFER A self-referential IBA records the PAINT curator's own judgement for this gene and is valid, not circular propagation. UniProtKB:Q9BXM0 · Homo sapiens PRX, periaxin (Swiss-Prot) SUPPORTS TRANSFER Holds GO:0005737 by IDA (PMID:24633211). |
| GO:0005737 cytoplasm | IEA GO_REF:0000117 | KEEP AS NON CORE | Summary: An ARBA rule row. The term is right and independently supported by the HPA IDA rows, but the rule's own published condition sets cannot be shown to fire on AHNAK2. Reason: Kept because cytoplasm is correct for AHNAK2 and is independently established by the HPA cytosol IDA on this same record, so nothing turns on the rule. The provenance is nonetheless unresolved and worth recording: ARBA00026971 has 2388 condition sets, conditioned only on InterPro, PANTHER, FunFam and taxon, and not one of them mentions any of AHNAK2's six signatures (IPR001478, IPR036034, IPR052082, PTHR23348, PTHR23348:SF37, Gene3D 2.30.42.10). The checking script self-tests first by constructing a true positive from the rule's own conditions, so this is a result rather than a broken matcher. Either the REST view of the rule is incomplete or the annotation predates a change in its conditions. Reported, not guessed at, and not used to argue for or against the term. |
| GO:0005737 cytoplasm | NAS PMID:17185750 AHNAK, a novel component of the dysferlin protein complex, r... | KEEP AS NON CORE | Summary: A non-traceable author statement taken from the AHNAK/dysferlin paper. The term is right for AHNAK2 on other grounds, but this row is one of five the paper contributed to AHNAK2 while its abstract discusses only AHNAK. Reason: Cytoplasm is correct for AHNAK2 and is separately supported by the HPA cytosol IDA, so there is nothing to remove. What is worth recording is the pattern this row belongs to: QuickGO by reference shows PMID:17185750 produced 19 annotations over 7 entities, and human AHNAK and human AHNAK2 received the same trio of NAS localisation terms from it - cytoplasm, T-tubule and sarcolemma - while the dysferlin rows from the same paper are IDA. The evidence code is the curator's own signal that these are statements rather than experiments. Non-core: an unqualified cytoplasm says nothing about AHNAK2's function. Supporting Evidence: PMID:17185750 Moreover, in regenerating rat muscle, dysferlin and AHNAK showed a marked increase and cytoplasmic localization |
| GO:0005829 cytosol | IDA GO_REF:0000052 | ACCEPT | Summary: One of only two GOA rows on this gene that rests on a measurement of the human AHNAK2 protein. Accepted. Reason: Human Protein Atlas immunofluorescence, curated under GO_REF:0000052, with HPA reliability Enhanced and cytosol as the main location call. This is antibody-validated imaging of the human protein in human cells, and it is concordant with everything else measured on AHNAK2: the FGF1 study reports a diffuse cytoplasmic distribution at normal temperature before heat shock redistributes the protein to the submembrane cytoskeletal fraction. It is also the evidence that weighs against the by-similarity nucleus, since HPA's vocabulary includes nuclear compartments and none was called. |
| GO:0005886 plasma membrane | IDA GO_REF:0000052 | ACCEPT | Summary: The second of the two rows resting on a measurement of human AHNAK2, and the one that matches its biology as a submembrane scaffold. Accepted. Reason: Human Protein Atlas immunofluorescence, Enhanced reliability, plasma membrane as an additional location. Independently corroborated in three systems that used different reagents: AHNAK2-specific antibodies place the protein in the costameric network directly beneath the sarcolemma of mouse skeletal muscle; a tagged C-terminal fragment colocalises with submembrane F-actin after heat shock; and the PDZ structure paper frames both periaxin and AHNAK2 as scaffolds linking the extracellular matrix to the cytoskeletal network, which is a submembrane role. This is the correct general term for that, and the muscle-specific refinement is proposed separately as costamere rather than substituted here. Supporting Evidence: PMID:25560297 In heat-shocked cells, FGF1 and the C-terminal fragment of AHNAK2 colocalized with F-actin in the vicinity of the cell membrane. PMID:24675079 Tight homodimerization may be central to the scaffolding functions of PRX and AHNAK2 in molecular complexes linking the extracellular matrix to the cytoskeletal network. |
| GO:0005886 plasma membrane | IEA GO_REF:0000117 | ACCEPT | Summary: ARBA duplicate of the HPA plasma membrane IDA on the same record. Correct term, redundant row, and like its sibling ARBA row its condition set cannot be shown to fire on AHNAK2. Reason: The term is right and is already established by IDA on this record, so this row adds no new information but nothing argues against it. ARBA00027801 has 686 condition sets over the same four condition types as its sibling rule, and none mentions any AHNAK2 signature, so the same unresolved provenance applies and is recorded in full on the cytoplasm ARBA row. Nothing in this review depends on the rule either way; accepted for consistency with the IDA row for the same term. |
| GO:0030315 T-tubule | NAS PMID:17185750 AHNAK, a novel component of the dysferlin protein complex, r... | MARK AS OVER ANNOTATED | Summary: The cited paper's T-tubule observation is about AHNAK and dysferlin. The independent AHNAK2 evidence is a pan-AHNAK antibody plus a cosedimentation inference in cardiomyocytes, and the only study using AHNAK2-specific antibodies found no T-tubule signal in skeletal muscle. Reason: Three separate things had to be checked before settling on over-annotated rather than removed. First, provenance: the row is NAS from a paper whose abstract places dysferlin and AHNAK, not AHNAK2, at the sarcolemmal membrane and T-tubules, and human AHNAK received the identical NAS row from the same reference. Second, the contrary evidence: the one study that used antibodies specific for AHNAK2 reports the costameric network and no T-tubule signal, with its own costameric result acting as the positive control that the reagent works. Third, and this is why REMOVE is not earned: AHNAK2 does have independent T-tubule evidence, from the AHNAK2 discovery paper, in cardiomyocytes rather than skeletal muscle. That evidence is weak - the abstract itself says the AHNAKs only "appear to localize" and that the T-tubule proximity "raise the possibility" of a role, the localisation came from cosedimentation with a dihydropyridine-receptor vesicle fraction rather than from an AHNAK2-specific reagent, and the paper's discussion attributes the AHNAK2 signal to residual staining in AHNAK1-null mice - but it is a real observation in a tissue the contrary study did not examine. Over-annotated is the calibrated call: not wrong, but resting on no AHNAK2-specific reagent, in a different tissue from the one the citing reference describes. Supporting Evidence: PMID:17185750 In normal skeletal muscle, dysferlin and AHNAK colocalize at the sarcolemmal membrane and T-tubules. PMID:20833135 In contrast, no AHNAK expression was detected in the T-tubule system. PMID:15007166 The localization of the AHNAKs in close proximity to transverse tubule membranes and Z-band regions of cardiac sarcomeres raise the possibility that they might be involved in regulating excitation/contraction coupling of cardiomyocytes |
| GO:0042383 sarcolemma | NAS PMID:17185750 AHNAK, a novel component of the dysferlin protein complex, r... | KEEP AS NON CORE | Summary: Same non-traceable provenance as its T-tubule sibling, but the term itself survives independent AHNAK2-specific evidence: costameres are sub-sarcolemmal structures and the AHNAK2-specific antibody study places AHNAK2 there. Reason: The citing reference supports this for AHNAK rather than AHNAK2, exactly as for the T-tubule row, but the two rows part company on what independent evidence says. AHNAK2-specific antibodies place the protein in the costameric network of mouse skeletal muscle, co-distributing with vinculin; costameres are the sub-sarcolemmal arrays that couple Z-discs laterally to the sarcolemma, so a sarcolemmal assignment for AHNAK2 is corroborated rather than contradicted, and the HPA plasma membrane IDA on this record agrees at the general level. Note that GO does not place GO:0043034 costamere under GO:0042383 sarcolemma - costamere sits under cytoplasm, myofibril and contractile muscle fiber - so the costamere row proposed below is an addition rather than a refinement of this one, and both should stand. Non-core not because the compartment is muscle-specific - costamere is equally muscle-specific and is core - but because this particular row carries none of the AHNAK2-specific evidence: it is a non-traceable statement from a paper about AHNAK, and sarcolemma is the general compartment that the costamere row states precisely and with an AHNAK2-specific reagent behind it. Supporting Evidence: PMID:20833135 Co-localization of AHNAK1 and AHNAK2 with vinculin clearly demonstrates that both proteins are components of the costameric network. |
| GO:0043484 regulation of RNA splicing | IBA GO_REF:0000033 | MARK AS OVER ANNOTATED | Summary: The clearest family-level over-reach on this gene. Two donors; one holds the term only by IBA, and the other is the paralogue AHNAK, whose evidence is a cis, locus-autonomous feedback loop mediated by a small nuclear isoform AHNAK2 is not known to make. Reason: The donor set is the smallest of the three IBA rows: mouse Prx and mouse Ahnak. Querying each donor's own evidence rather than stopping at its identity is what settles it. Mouse Prx holds GO:0043484 by IBA alone, so it is part of the same propagation and adds no independent support. That leaves exactly one donor with real evidence, mouse Ahnak, by IDA from PMID:21940993 - and the mechanism that paper demonstrates is cis and locus-autonomous: a small nuclear isoform establishing a positive feedback loop on the splicing of its own locus, with the same shown separately for a 17-kDa isoform of periaxin on the periaxin locus. Transferring that to AHNAK2 asserts something specific and untested, namely that AHNAK2 also produces a small nuclear isoform that feeds back on the AHNAK2 locus. Two things keep this at over-annotated rather than removed. The PAINT call is reasonable rather than careless: the same authors state that the gene architecture is shared with AHNAK2, and two of the three family members with that architecture do it, which is what IBA exists to capture. And the counter-argument is an absence - UniProt annotates no small N-terminal isoform for AHNAK2, its only short isoform Q8IVF2-2 being residues 5003-5795 with no PDZ domain - and an unannotated transcript is not an absent one. The decisive context is what the curator who read the paper actually did: PMID:21940993 produced 17 annotations over exactly two entities, and the transfer they chose was mouse Ahnak IDA to human AHNAK ISS - one species step within one gene. They did not extend it to AHNAK2 or to PRX. PAINT then gave the term to all three human PTHR23348 members from a byte-identical WITH/FROM set, and none of the three holds it by its own experimental evidence. That ISS is also the discriminator against AHNAK, whose identical IBA row is defensible precisely because AHNAK has the same-gene transfer and AHNAK2 does not. The deciding experiment is named in suggested_experiments. Propagation Review Root cause: PROPAGATION BAD Failure modes: WRONG ORTHOLOG OR PARALOG CIRCULAR PROPAGATION Sources checked: MGI:MGI:108176 · Mus musculus Prx, periaxin (O55103, Swiss-Prot) CIRCULAR OR REDUNDANT Holds GO:0043484 by IBA only - a recipient of the same propagation, not an independent source. Periaxin's own splicing-regulation result in PMID:21940993 has not been curated as an experimental annotation. MGI:MGI:1316648 · Mus musculus Ahnak (E9Q616, TrEMBL - no Swiss-Prot entry exists) SUPPORTS SOURCE BUT NOT TARGET The only donor with its own evidence - GO:0043484 by IDA from PMID:21940993. The annotation is sound for Ahnak; what it records is feedback on the Ahnak locus itself, by a small nuclear isoform. PANTHER:PTN001156025 · PANTHER tree node, not a protein NOT RELEVANT The inference structure rather than an evidence source. The AHNAK2 orthologue MGI:2144831 is on this gene's cytoplasm row but not on this one. Supporting Evidence: PMID:21940993 During muscle differentiation the small AHNAK is strongly increased, thereby establishing a positive feedback loop to regulate mRNA splicing of its own locus. PMID:21940993 This genetic structure is shared by AHNAK2 and Periaxin, which share a common ancestor. |
| GO:0042803 protein homodimerization activity | IDA PMID:24675079 Periaxin and AHNAK nucleoprotein 2 form intertwined homodime... | NEW | Summary: Proposed. A 1.75 A crystal structure of AHNAK2's PDZ domain shows an intertwined, domain-swapped homodimer, curated by UniProt at ECO:0000269, and the paper produced no GO annotation on any protein. Reason: This is the best-measured molecular property of human AHNAK2 and GO records none of it. PDB 4CN0 resolves residues 108-203 as two chains, A and B, forming a six-stranded antiparallel beta sheet in which every other strand comes from a different chain, and UniProt states it on the AHNAK2 entry as a homodimer via the PDZ domain with ECO:0000269|PubMed:24675079. Querying QuickGO by reference returns zero annotations for PMID:24675079 across all of GOA, and none of AHNAK2, PRX or AHNAK holds GO:0042803, GO:0046982 or GO:0051260, so this is a coverage gap rather than a curatorial judgement. It also gives the gene its first informative molecular function: AHNAK2's GOA molecular function content is otherwise two bare GO:0005515 rows, which state nothing. GO:0042803 is a descendant of GO:0042802 identical protein binding and of GO:0046983 protein dimerization activity, both under GO:0005515, so it refines rather than conflicts with what is already there. Supporting Evidence: PMID:24675079 The unique intertwined, domain-swapped dimers provide a structural basis for the homodimerization of both proteins. PMID:24675079 We present the crystal structures of the PDZ-like domains from PRX and its homologue AHNAK nucleoprotein 2 (AHNAK2). |
| GO:0043034 costamere | ISS PMID:20833135 AHNAK1 and AHNAK2 are costameric proteins: AHNAK1 affects tr... | NEW | Summary: Proposed. AHNAK2-specific antibodies place the mouse orthologue in the costameric network with vinculin, and the mouse entry holds GO:0043034 by IDA from this paper while human AHNAK2 has nothing from it. Reason: This is the only study that has imaged AHNAK2 in muscle with a reagent specific for AHNAK2 rather than for the family, and it is the source of the one localisation this gene has that is both specific and measured. Mouse Ahnak2 (A0A7N9VR94) holds GO:0043034 by IDA from PMID:20833135, and mouse Ahnak holds it by IDA from the same paper, yet QuickGO by reference shows the paper produced three annotations over two entities and none on human AHNAK2. Human AHNAK has picked costamere up by ISS (from PMID:21940993); human AHNAK2 has not, which looks like a coverage gap rather than a decision, since the costamere paper is the one study in which an AHNAK2-specific reagent was used and AHNAK2 was one of its two subjects. ISS rather than IDA because the experiment is in mouse skeletal muscle, with mouse Ahnak2 as the supporting entity. Note that GO does not place costamere under sarcolemma - its ancestors are cytoplasm, myofibril and contractile muscle fiber - so this is additive to the existing sarcolemma row rather than a replacement, and it is also a descendant of the GO:0005737 cytoplasm the IBA already gives, which is why that row was kept rather than narrowed. Supporting Evidence: PMID:20833135 We used specific AHNAK1 and AHNAK2 antibodies to analyzed the detailed localization of both proteins in mouse skeletal muscle. PMID:20833135 Co-localization of AHNAK1 and AHNAK2 with vinculin clearly demonstrates that both proteins are components of the costameric network. |
| GO:0009306 protein secretion | ISS PMID:25560297 AHNAK2 Participates in the Stress-Induced Nonclassical FGF1 ... | NEW | Summary: Proposed. Depleting AHNAK2 nearly abolishes stress-induced FGF1 export while leaving FGF2 export and Notch-induced FGF1 release intact. ISS because the knockdown was of the mouse protein. Reason: A specific, well-controlled loss-of-function result that produced no GO annotation anywhere - QuickGO returns zero annotations for PMID:25560297 across all of GOA. The specificity controls are what make it worth annotating: AHNAK2 depletion did not affect spontaneous FGF2 export or the Notch-inhibition route of FGF1 release, so this is not a general secretory defect. The evidence code is decided by the paper's methods: the shRNA constructs suppress mouse AHNAK2 in NIH 3T3 cells, so for the human protein this is ISS with the mouse entry as supporting entity, not IMP. GO:0009306 rather than GO:0050714 positive regulation of protein secretion because AHNAK2 physically associates with the cargo and co-translocates with it, which is participation rather than regulation; GO has no unconventional- or non-classical-secretion term, and the curated machinery of this pathway (S100A13, SYT1) carries exocytosis-branch terms rather than anything FGF-specific. One caveat on the evidence code, stated rather than hidden: the supporting entity, mouse Ahnak2 (A0A7N9VR94), carries no GO:0009306 annotation of its own for this ISS to inherit - the mouse experiment has never been curated. The entity is therefore the sequence-similarity anchor for the transfer rather than the holder of a donor annotation, and this proposal implicitly proposes the mouse annotation too. Supporting Evidence: PMID:25560297 Depletion of AHNAK2 resulted in a drastic decrease of stress-induced FGF1 export but did not affect spontaneous FGF2 export and FGF1 release induced by the inhibition of Notch signaling. PMID:25560297 Genetic constructs pGFP-V-RS expressing shRNAs that suppress the expression of mouse AHNAK2 were obtained from Origene (Rockville, MD). |
| GO:0017134 fibroblast growth factor binding | IPI PMID:25560297 AHNAK2 Participates in the Stress-Induced Nonclassical FGF1 ... | NEW | Summary: Proposed, with its limitations stated. A heat-shock-dependent association between FGF1 and AHNAK2 was found by immunoprecipitation mass spectrometry and confirmed by reciprocal co-IP, but on an overexpressed tagged C-terminal fragment. Reason: Proposed because it is an informative molecular function where the alternative is another bare GO:0005515, and because the association was found by unbiased IP-MS and then confirmed reciprocally with anti-V5 and anti-HA antibodies. The limitations belong in the record rather than in a silent downgrade: the confirmation used an adenovirally overexpressed V5-tagged C-terminal 483-residue fragment of human AHNAK2 in mouse NIH 3T3 cells, because the authors state that commercially available anti-AHNAK2 antibodies failed to detect the protein reliably. So the interaction is established for a fragment under overexpression, the association is conditional on heat stress, and no affinity or stoichiometry has been measured. If a reviewer judges that too thin for a molecular function, the fallback is to drop this row and let GO:0009306 carry the finding; I have proposed it because an FGF-specific binding term is checkable whereas a bare protein binding row would not be. Supporting Evidence: PMID:25560297 The heat shock-dependent association of FGF1 with the large protein AHNAK2 was observed. PMID:25560297 commercially available anti-AHNAK2 antibodies failed to reliably detect AHNAK2 both in immunoblotting and immunofluorescence |
Loading supporting content…
Download this section (compressed HTML)Q: UniProt's only SUBCELLULAR LOCATION for AHNAK2 is Nucleus carrying a bare ECO:0000250 with no source accession, and the presumed donor - human AHNAK - carries an untagged Nucleus of its own. Can UniProt source this statement or withdraw it? It is the sole basis of the GO_REF:0000044 nucleus row, while HPA immunofluorescence at Enhanced reliability calls cytosol and plasma membrane with no nuclear compartment.
Q: PMID:24675079 solved the PDZ domains of periaxin and AHNAK2 at 1.75 A and reported the same domain-swapped homodimer for both, and UniProt curates it on both entries at ECO:0000269 - yet the paper has zero GO annotations in all of GOA, and neither PRX nor AHNAK2 holds any dimerisation term. This review proposes GO:0042803 for AHNAK2; PRX has the identical claim outstanding.
Suggested experts: Petri Kursula, Hongmin Han
Q: PAINT node PTN001156025 gives AHNAK2, AHNAK and PRX the same three IBA terms from byte-identical WITH/FROM sets, treating three different PTHR23348 subfamilies as interchangeable when their only shared domain is 28.4% (AHNAK2 vs AHNAK) to 56.8% (AHNAK2 vs PRX) identical. The AHNAK2 orthologue (MGI:2144831) is in only one of the three donor sets, none of the three human members holds GO:0043484 by its own experimental evidence, and only PRX holds GO:0005634 by its own IDA. Meanwhile mouse Ahnak2's IDA-backed terms - costamere, Z disc, cytoplasmic vesicle membrane - have not propagated to human AHNAK2 at all. A periaxin-specific node already exists for GO:0032287, so the machinery for finer placement is there; should the shared node's term set be rebalanced toward what each subfamily's own orthologues carry?
Q: Several papers state that AHNAK2 binds directly to periaxin, including PMID:31011849, which uses it to explain a Charcot-Marie-Tooth phenotype. I cannot find a primary measurement of an AHNAK2-PRX heterodimer. The structural paper usually cited in this context reports domain-swapped homodimers of each protein separately. Is there a primary source for the heterotypic interaction, or has a two-homodimer paper been read as a heterodimer paper?
Suggested experts: Petri Kursula, Marina L Kennerson
Q: GO has no unconventional- or non-classical-secretion term, and GO:0009306 protein secretion has no fibroblast-growth-factor child even though it has ten cargo-named children including Wnt protein secretion, BMP secretion and insulin secretion. Is GO:0090269 fibroblast growth factor production - whose definition covers appearance due to "biosynthesis or secretion" - intended to cover machinery components of stress-induced FGF1 export such as AHNAK2, S100A13 and SYT1, or is a GO:0009306 child needed? I have annotated to the undifferentiated parent rather than guess.
Suggested experts: Igor Prudovsky
Q: ARBA00026971 and ARBA00027801 supply AHNAK2's cytoplasm and plasma membrane IEA rows, but across 2388 and 686 condition sets neither rule's published conditions mention any of AHNAK2's InterPro, PANTHER or FunFam signatures. Is the REST view of these rules incomplete, or were the annotations applied under earlier conditions?
Experiment: 5'-RACE and long-read RNA-seq across the AHNAK2 locus, then immunoblotting of nuclear and cytoplasmic fractions with an antibody raised against residues 1-200. This single experiment decides both nucleus rows and the regulation-of-RNA-splicing row: the mechanism transferred from AHNAK requires such an isoform, and UniProt annotates none for AHNAK2, whose only short isoform is C-terminal.
Hypothesis: AHNAK2 produces a small N-terminal, PDZ-containing isoform analogous to the 17-kDa AHNAK and periaxin species, and that isoform enters the nucleus.
Type: transcript mapping and subcellular fractionation
Experiment: Deplete AHNAK2 in differentiating myotubes and quantify the ratio of long to short AHNAK2 transcripts by isoform-specific RT-qPCR, with AHNAK as the positive control. A negative result would retire GO:0043484 on AHNAK2 rather than leaving it marked over-annotated.
Hypothesis: AHNAK2 regulates splicing at its own locus, as AHNAK does at AHNAK.
Type: knockdown plus isoform-specific RT-qPCR
Experiment: SEC-MALS and native mass spectrometry on the purified 108-203 fragment across a concentration series, plus crosslinking of endogenous AHNAK2 in cells followed by immunoblotting for a dimeric species.
Hypothesis: The domain-swapped PDZ dimer exists at native concentrations and is not a crystallisation artefact.
Type: biophysics and crosslinking
Experiment: Screen a C-terminal peptide array or a proteomic peptide-phage display library against the purified domain, then test the top hits by holdup assay. This would convert AHNAK2's scaffold role from an architectural inference into a measured molecular function.
Hypothesis: The preformed AHNAK2 PDZ pocket binds a physiological C-terminal peptide ligand.
Type: peptide binding screen
Experiment: CRISPR knockout of AHNAK2 in a human cell line that exports FGF1 under heat stress, with rescue by full-length AHNAK2 and by the C-terminal fragment separately. This would upgrade the ISS to direct human evidence.
Hypothesis: Full-length human AHNAK2 is required for stress-induced FGF1 export, and the C-terminal region is sufficient.
Type: knockout and rescue
Experiment: Image AHNAK2 in Ahnak1-null mouse cardiomyocytes and skeletal fibres side by side using a validated AHNAK2-specific antibody, with the dihydropyridine receptor and vinculin as compartment markers.
Hypothesis: The cardiac T-tubule signal attributed to AHNAK2 is real and tissue-specific rather than an artefact of pan-AHNAK reagents.
Type: comparative immunofluorescence
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)