id: P82987
gene_symbol: ADAMTSL3
aliases:
- Punctin-2
- ADAMTSL-3
- KIAA1233
product_type: PROTEIN
status: COMPLETE
taxon:
  id: NCBITaxon:9606
  label: Homo sapiens
description: 'ADAMTSL3, also called punctin-2, is a large secreted glycoprotein of the extracellular matrix.
  Despite the family name it is not a proteinase: it is built from two arrays of thrombospondin type-1
  repeats flanking three immunoglobulin-like C2-type domains, with an ADAMTS-type cysteine-rich region
  and a C-terminal PLAC domain, and it lacks both the catalytic metalloprotease domain and the disintegrin-like
  domain that define the ADAMTS proteinases. It carries a cleaved signal peptide and extensive N-linked
  glycosylation, and is deposited into the matrix as a roughly 210-kDa glycoprotein.


  Its molecular role is that of a matricellular adaptor within the fibrillin microfibril network. It binds
  the N-terminal, assembly-competent half of fibrillin-1 at a site deleted in Weill-Marchesani syndrome,
  binds the non-catalytic C-terminal region of the proteinase ADAMTS10 with nanomolar affinity, and binds
  the C-terminal domains of latent TGF-beta binding protein 1 - the two matrix proteins that between them
  sequester the large latent TGF-beta complex. Consistent with that placement, ADAMTSL3 restrains TGF-beta
  signalling: its loss in mice raises active TGF-beta and SMAD2 phosphorylation in the pressure-overloaded
  heart, while raising it in human cardiac fibroblasts lowers SMAD2 phosphorylation, myofibroblast differentiation
  and collagen synthesis. In the nervous system it acts through the receptor DCC to set hippocampal synapse
  density and GABAergic synapse function, and parvalbumin interneurons require it cell-autonomously to
  build and maintain perineuronal nets, a role conserved to the nematode orthologue Ce-Punctin/MADD-4,
  which specifies cholinergic versus GABAergic postsynaptic identity through UNC-40/DCC.


  The protein is broadly expressed - in epithelia of colon, skin, breast, prostate, liver and pancreas,
  and in endothelium, smooth muscle, fibroblasts, cortical and ganglionic neurons and cardiac myocytes.
  Common variants at the ADAMTSL3 locus are among the reproducible genome-wide association signals for
  adult height and lean body mass, and the gene has been proposed as a candidate at schizophrenia.'
alternative_products:
- name: '1'
  id: P82987-1
- name: '2'
  id: P82987-2
  sequence_note: VSP_037810
references:
- id: GO_REF:0000002
  title: Gene Ontology annotation through association of InterPro records with GO terms
  findings: []
  reference_review:
    relevance: MEDIUM
    correctness: VERIFIED
    review_notes: The InterPro2GO mapping used is IPR013273 (ADAMTS/ADAMTS-like) to GO:0030198. The signature
      is family-level and covers catalytic and non-catalytic members alike, so the mapping is a family
      inference; it is nonetheless independently supported for this gene.
- id: GO_REF:0000033
  title: Annotation inferences using phylogenetic trees
  findings: []
  reference_review:
    relevance: HIGH
    correctness: VERIFIED
    review_notes: PAINT node PTN000347317 in PTHR13723. All 16 protein donors on the single IBA row were
      resolved and each carries its own experimental annotation to the propagated term or a descendant.
- id: PMID:14667842
  title: ADAMTSL-3/punctin-2, a novel glycoprotein in extracellular matrix related to the ADAMTS family
    of metalloproteases.
  findings:
  - statement: ADAMTSL3/punctin-2 lacks both the metalloprotease and the disintegrin-like domain of the
      ADAMTS proteinases while preserving the rest of the ADAMTS domain order.
    supporting_text: These multi-domain proteins lack both a protease domain and a disintegrin-like domain
      but are remarkably similar in their domain organization to the ADAMTS proteases, hence the name
      ADAMTS-like.
    full_text_unavailable: true
  - statement: Punctin-2 is secreted as a 210-kDa glycoprotein that is deposited in the extracellular
      matrix of transfected cells - the direct experimental basis for UniProt's ECO:0000269 subcellular
      location.
    supporting_text: Using these and a monoclonal antibody to a C-terminal myc tag, we show that in transfected
      COS-7 cells, punctin-2 is expressed as a 210-kDa glycoprotein that is located in the extracellular
      matrix.
    full_text_unavailable: true
  - statement: The TSR arrays are interrupted by three immunoglobulin-like repeats, an architecture not
      seen in the ADAMTS proteinases.
    supporting_text: Punctin-2 contains 13 TSRs arranged in two arrays separated by a region containing
      three immunoglobulin-like repeats.
    full_text_unavailable: true
  reference_review:
    relevance: HIGH
    correctness: VERIFIED
    review_notes: 'The defining cloning and characterisation paper. Note the TSR count differs between
      methods: 13 here, 10 PROSITE domains and 12 SMART repeats in the current UniProt entry.'
- id: PMID:17597111
  title: ADAMTSL3/punctin-2, a gene frequently mutated in colorectal tumors, is widely expressed in normal
    and malignant epithelial cells, vascular endothelial cells and other cell types, and its mRNA is reduced
    in colon cancer.
  findings:
  - statement: ADAMTSL3 protein is broadly distributed across human epithelia, vasculature, fibroblasts,
      neurons and cardiac myocytes - the basis for treating the mouse cardiac and neuronal phenotypes
      as relevant to the human protein.
    supporting_text: ADAMTSL3 is expressed in epithelial cells of the colon, fallopian tube, skin, breast,
      prostate, epididymis, liver, pancreatic islets and bile ducts, as well as by vascular endothelial
      cells, smooth muscle cells, fibroblasts, cortical and ganglionic neurons and cardiac myocytes.
    full_text_unavailable: true
  reference_review:
    relevance: MEDIUM
    correctness: VERIFIED
    review_notes: Expression survey by validated IHC and in situ hybridisation. Used only for tissue distribution;
      the colon-cancer mRNA reduction is not used to support any term.
- id: PMID:22242013
  title: Microenvironmental regulation by fibrillin-1.
  findings:
  - statement: The C-terminal region of ADAMTSL-3 binds the N-terminal half of fibrillin-1 by SPR; ADAMTSL-1
      does not, so the interaction is selective within the family.
    supporting_text: Similar to ADAMTSL-6 [24], ADAMTSL-2, -3, and papilin polypeptides interacted with
      the N-terminal half of fibrillin-1, while ADAMTSL-1 did not.
  - statement: The Weill-Marchesani three-domain deletion in fibrillin-1 abolishes ADAMTSL binding, mapping
      the site.
    supporting_text: ADAMTSL-2, -3, and -6 and papilin polypeptides did not bind to recombinant fibrillin-1
      polypeptides with the WMS three-domain deletion (Figure 5a and Table S1).
  - statement: ADAMTSL-3 binds the non-catalytic C-terminal region of ADAMTS-10 with a KD of 2 nM, and
      this too is selective within the family.
    supporting_text: SPR also showed that the C-terminal end of ADAMTS-10 interacted with the C-terminal
      end of ADAMTSL-3 with high binding affinity (KD = 2 nM) (Figure 5c).
  - statement: ADAMTSL-3 binds the C-terminal domains of LTBP-1 but not LTBP-4, connecting it to latent
      TGF-beta sequestration.
    supporting_text: Binding between ADAMTSL-3 and the C-terminal domains of LTBP-1 was also detected,
      but neither ADAMTSL-2 nor -3 interacted with LTBP-4.
  - statement: No colocalisation of ADAMTSL-3 with fibrillin-1 was tested, because ADAMTSL3-specific antibodies
      did not exist. This is the reason the elastic-fiber annotation cannot be treated as an established
      location.
    supporting_text: Since antibodies specific for ADAMTSL-2 and -3 are not yet available, we were unable
      to determine whether these proteins also colocalize with fibrillin-1 in skin and whether these are
      also reduced in WMΔ mutant mice.
  - statement: 'The Methods contradict the Results on the species of the ADAMTSL-3 reagent: the Results
      call it human, the Methods describe a RIKEN clone plus mouse lung cDNA.'
    supporting_text: Constructs for ADAMTSL3 were made using a clone (RIKEN) and mouse lung cDNA.
  reference_review:
    relevance: HIGH
    correctness: VERIFIED
    review_notes: 'The single most informative paper for ADAMTSL3''s molecular function, and entirely
      absent from its GOA. Verified against the PMC full text. One internal inconsistency, recorded above
      and raised as a question rather than silently resolved: Results say the ADAMTSL-3 polypeptide was
      human, Methods say the construct came from a RIKEN clone and mouse lung cDNA.'
- id: PMID:23962539
  title: Elastic fibres in health and disease.
  findings:
  - statement: Elastic fibres consist of a cross-linked elastin core within a fibrillin microfibril template
      - the structure that makes ADAMTSL3's fibrillin-1 binding a plausible basis for the elastic-fiber
      annotation, without demonstrating it.
    supporting_text: They are laid down during development, and comprise a cross-linked elastin core within
      a template of fibrillin-based microfibrils.
    full_text_unavailable: true
  reference_review:
    relevance: MEDIUM
    correctness: UNVERIFIED
    review_notes: 'A review, cached abstract-only, and the abstract does not mention ADAMTSL3, so the
      statement supporting the annotation is in text I cannot read. Reference-scope check: GO_Central
      used it for 66 TAS annotations over 62 entities, but selectively within this family - ADAMTSL3,
      ADAMTSL4, ADAMTSL5 and THSD4 in, ADAMTSL1, ADAMTSL2 and PAPLN out - so it is a real traceable statement
      rather than a blanket projection. Marked UNVERIFIED rather than VERIFIED because the supporting
      sentence itself has not been read.'
- id: PMID:25416956
  title: A proteome-scale map of the human interactome network.
  findings:
  - statement: HI-II-14 is a systematic proteome-scale binary interactome screen, not a targeted study
      of ADAMTSL3; five of ADAMTSL3's GO:0005515 rows come from it.
    supporting_text: While currently available information is highly biased and only covers a relatively
      small portion of the proteome, our systematic map appears strikingly more homogeneous
  reference_review:
    relevance: LOW
    correctness: VERIFIED
    review_notes: 'Correctly cited by IntAct as the source of five Y2H pairs. Low relevance to ADAMTSL3''s
      function: all five partners are keratin-family or nuclear/cytosolic proteins that a secreted matrix
      protein does not encounter, and none has been followed up.'
- id: PMID:32296183
  title: A reference map of the human binary protein interactome.
  findings:
  - statement: HuRI is a systematic Y2H map of the human proteome; ten of ADAMTSL3's GO:0005515 rows come
      from it, and it shares the CCSB ORFeome and search space with HI-II-14, so overlap between the two
      is not independent replication.
    supporting_text: Specifically, yeast two-hybrid (Y2H) represents the only binary PPI assay that can
      be operated at sufficient throughput to systematically screen the human proteome for binary PPIs.
  reference_review:
    relevance: LOW
    correctness: VERIFIED
    review_notes: 'Correctly cited by IntAct. Same limitation as PMID:25416956: the partners recovered
      for ADAMTSL3 are dominated by high-degree keratin-associated preys.'
- id: PMID:36539599
  title: ADAMTSL3 knock-out mice develop cardiac dysfunction and dilatation with increased TGFβ signalling
    after pressure overload.
  findings:
  - statement: Overexpression of human ADAMTSL3 in human foetal cardiac fibroblasts lowers phospho-SMAD2
      and active TGF-beta - the only human-cell functional experiment on this protein.
    supporting_text: Immunoblotting revealed reduced pSMAD (Fig. 4h) and reduced active TGFβ (Fig. 4i),
      indicating reduced TGFβ signalling.
  - statement: The gain-of-function experiment used adenoviral overexpression, so the appropriate evidence
      code is IMP rather than IDA.
    supporting_text: For further mechanistic insights, we overexpressed full-length ADAMTSL3 (L3) and
      a vehicle control adenovirus (veh) in cultures of human foetal CFBs (hfCFBs)
  - statement: Loss and gain of function agree in sign, which is what makes the negative-regulation direction
      safe.
    supporting_text: Taken together, the loss-of-function and gain-of-function studies suggest ADAMTSL3
      as an inhibitor of TGFβ signalling in the heart in vivo and in cardiac fibroblasts in vitro.
  - statement: Adamtsl3-null mouse hearts show raised active TGF-beta and cardiac fibroblast activation
      after aortic banding.
    supporting_text: We found that Adamtsl3 knock-out mice develop exacerbated cardiac dysfunction and
      dilatation with increased mortality, and hearts show increased TGFβ activity and CFB activation
      after pressure overload by aortic banding.
  - statement: 'The authors state the family-level premise independently: ADAMTSLs lack a catalytic domain.'
    supporting_text: ADAMTSLs are structurally related to the ADAMTS ECM proteases12, but lack a catalytic
      domain, leaving their biological function largely unknown.
  reference_review:
    relevance: HIGH
    correctness: VERIFIED
    review_notes: Verified against the PMC full text. Loss-of-function is mouse; the gain-of-function
      is human protein in human cells, which is what carries the human annotation. Not in ADAMTSL3's GOA.
- id: PMID:37572323
  title: Adamtsl3 mediates DCC signaling to selectively promote GABAergic synapse function.
  findings:
  - statement: Mouse conditional deletion of Adamtsl3 in neurons reduces DCC protein and the density of
      glutamatergic and GABAergic synapses.
    supporting_text: We demonstrate that early post-natal deletion of Adamtsl3 in neurons impairs DCC
      protein expression, causing reduced density of both glutamatergic and GABAergic synapses.
    full_text_unavailable: true
  - statement: Adamtsl3 acts as a hippocampal synapse organiser through the DCC receptor.
    supporting_text: Here, we identify that the secreted protein Adamtsl3 functions as critical hippocampal
      synapse organizer acting through the transmembrane receptor DCC (deleted in colorectal cancer).
    full_text_unavailable: true
  reference_review:
    relevance: HIGH
    correctness: VERIFIED
    review_notes: Cached abstract-only. Mouse conditional knockouts, so it supports an ISS rather than
      an IMP annotation for the human gene.
- id: PMID:42277231
  title: Schizophrenia-associated glycoprotein Adamtsl3 regulates perineuronal net formation, maintenance,
    and adult cortical plasticity.
  findings:
  - statement: Adamtsl3 is required cell-autonomously by parvalbumin interneurons for perineuronal net
      integrity - independent evidence that this protein organises a specialised extracellular matrix.
    supporting_text: Here, we identify the schizophrenia-associated glycoprotein Adamtsl3 as a PV+ cell-autonomous
      regulator of PNN integrity.
    full_text_unavailable: true
  - statement: 'The proposed mechanism runs through MMP9: Adamtsl3 loss raises MMP9 and the perineuronal
      net defect is rescued by MMP9 inhibition.'
    supporting_text: Mechanistically, Adamtsl3 modulates matrix metalloprotease-9 (MMP9) activity, and
      Adamtsl3 deletion results in elevated MMP9 levels, PNN reduction, decreased Otx2 uptake, and heightened
      oxidative stress in PV+ cells.
    full_text_unavailable: true
  reference_review:
    relevance: HIGH
    correctness: VERIFIED
    review_notes: '2026 paper, cached abstract-only, mouse. Cited as independent corroboration of extracellular
      matrix organisation rather than as the basis for an MMP-inhibitor annotation: the abstract reports
      elevated MMP9 levels as well as activity, so whether ADAMTSL3 inhibits the enzyme or limits its
      abundance is not resolved from the abstract.'
- id: file:human/ADAMTSL3/ADAMTSL3-deep-research-affinage.md
  title: Affinage mechanistic annotation for ADAMTSL3 (human)
  findings:
  - statement: 'Machine-generated record. Its one directly usable contribution is a corpus-level negative:
      no biochemical mechanism has been published for the effect of ADAMTSL3 on DCC. Independently confirmed
      here by searching the ADAMTSL3 literature.'
    supporting_text: The biochemical mechanism by which this non-catalytic ECM protein controls DCC receptor
      levels has not been characterized in the available corpus.
  reference_review:
    relevance: LOW
    correctness: LOW_QUALITY
    review_notes: 'gates_passed: False. The tripped gate is a suspected symbol collision raised because
      the narrative opens on C. elegans; on inspection that is a false positive, since the narrative correctly
      identifies human ADAMTSL3/punctin-2 and is discussing the worm orthologue Ce-Punctin/madd-4. Every
      claim taken from the record was nonetheless re-verified against the cited PMIDs, and four defects
      were found. (1) A mis-joined citation: the record attributes a synaptic-organiser function for human
      ADAMTSL3 to PMID:14667842 plus PMID:24896188, but the first is a 2003 ECM cloning paper with no
      synaptic content and the second is about the worm protein - each constituent statement is defensible
      and the join is not. (2) A finding sourced to bioRxiv with an em dash in the PMID column; it has
      since appeared as PMID:42497245 but supports no GO term. (3) PMID:18474779 is over-read as showing
      an antiangiogenic role for ADAMTSL-3 when it shows only that the transcript is raised in glaucoma
      astrocytes. (4) It misses the two most informative papers for this gene, PMID:22242013 and PMID:36539599,
      so its statement that no mechanism has been characterised is a statement about its own corpus. Used
      here only for that corpus-level negative, and only after confirming it independently.'
- id: PMID:24896188
  title: C. elegans Punctin specifies cholinergic versus GABAergic identity of postsynaptic domains.
  findings:
  - statement: The nematode orthologue is a secreted synaptic organiser that assigns neurotransmitter
      identity to postsynaptic domains, establishing that the synaptic role is ancestral.
    supporting_text: C. elegans Punctin specifies cholinergic versus GABAergic identity of postsynaptic
      domains.
    full_text_unavailable: true
  reference_review:
    relevance: MEDIUM
    correctness: VERIFIED
    review_notes: About the worm protein, not human ADAMTSL3, and MADD-4 is the co-orthologue of ADAMTSL1
      and ADAMTSL3 - so it supports conservation of the role, not a claim about the human gene. The affinage
      record cites this paper together with PMID:14667842 to assert a synaptic-organiser function for
      human ADAMTSL3; neither paper makes that claim.
- id: PMID:22014523
  title: MADD-4 is a secreted cue required for midline-oriented guidance in Caenorhabditis elegans.
  findings:
  - statement: MADD-4 signals through UNC-40/DCC, the same receptor that mediates Adamtsl3's effect in
      the mouse hippocampus - so the receptor pairing, not just the protein, is conserved.
    supporting_text: MADD-4's activity is dependent on UNC-40/DCC, a netrin receptor, which functions
      cell-autonomously to direct membrane extension.
    full_text_unavailable: true
  - statement: MADD-4 is the co-orthologue of both ADAMTSL1 and ADAMTSL3, so worm phenotypes cannot be
      assigned to either human paralogue alone.
    supporting_text: The biological role of MADD-4 orthologs, including ADAMTSL1 and 3 in mammals, is
      unknown.
    full_text_unavailable: true
  reference_review:
    relevance: MEDIUM
    correctness: VERIFIED
    review_notes: Worm genetics. Cited for the conserved MADD-4/UNC-40-DCC axis and for the explicit statement
      that MADD-4 has two mammalian orthologues.
- id: PMID:18391952
  title: Genome-wide association analysis identifies 20 loci that influence adult height.
  findings:
  - statement: ADAMTSL3 is one of the reproducible genome-wide association loci for adult height, and
      was interpreted by the authors as an extracellular matrix gene. This motivates the growth-plate
      experiment but supports no GO term on its own.
    supporting_text: The genes implicate a number of biological pathways and processes in the normal determination
      of human height, including Hedgehog signaling (IHH, HHIP, PTCH1), basic cell cycle regulation (CDK6,
      one of the cyclin-dependent kinases implicated in cell cycle progression13), extracellular matrix
      (ADAMTSL3 and EFEMP1) and chromatin rearrangement and polycomb proteins (HMGA2 and SCMH1).
  reference_review:
    relevance: LOW
    correctness: VERIFIED
    review_notes: A locus association, not a statement about the protein's activity. Recorded because
      it is why the gene is studied; deliberately not used to support any annotation.
- id: PMID:21239144
  title: 'ADAMTSL3 as a candidate gene for schizophrenia: gene sequencing and ultra-high density association
    analysis by imputation.'
  findings:
  - statement: The schizophrenia association at ADAMTSL3 is refined but not resolved; the paper's own
      conclusion is that the gene remains a candidate.
    supporting_text: Our study confirms ADAMTSL3 as a candidate for further investigation in schizophrenia,
      using the variants identified here.
    full_text_unavailable: true
  reference_review:
    relevance: LOW
    correctness: VERIFIED
    review_notes: Genetic association only, correctly described by its own abstract as a candidate rather
      than an established gene. Cited for context behind the two mouse neuronal papers; supports no GO
      term.
- id: PMID:32266537
  title: Genome-wide CRISPR knockout screens identify ADAMTSL3 and PTEN genes as suppressors of HCC proliferation
    and metastasis, respectively.
  findings:
  - statement: ADAMTSL3 emerged from a genome-wide CRISPR screen in one hepatocellular carcinoma line
      as a suppressor of proliferation. Single screen, single line, no mechanism.
    supporting_text: Moreover, knocking out either the ADAMTSL3 or PTEN genes promoted either the proliferation
      or metastasis of HCC cells, respectively.
    full_text_unavailable: true
  reference_review:
    relevance: LOW
    correctness: LOW_QUALITY
    review_notes: The ADAMTSL3 result rests on sgRNAs appearing twice in one genome-wide screen in one
      cell line, with no mechanistic follow-up and no independent replication. Cited so the claim is on
      record and explicitly not annotated.
- id: PMID:18474779
  title: Antiangiogenic characteristics of astrocytes from optic nerve heads with primary open-angle glaucoma.
  findings:
  - statement: ADAMTSL-3 transcript is raised in optic-nerve-head astrocytes from glaucoma donors. The
      reduced tube formation reported in the same paper is a property of the astrocyte populations, not
      of ADAMTSL3, which was never perturbed.
    supporting_text: The ONH astrocytes from donors with POAG decreased expression of proangiogenic factors
      (vascular endothelial growth factor C and platelet-derived growth factor A) and increased expression
      of antiangiogenic factors (collagen XVIII and ADAMTSL-3) when compared with normal ONH astrocytes.
    full_text_unavailable: true
  reference_review:
    relevance: LOW
    correctness: MISCITED
    review_notes: 'Cited by the affinage deep-research record as showing an antiangiogenic role for ADAMTSL-3.
      It shows no such thing: ADAMTSL-3 is one of two transcripts raised in glaucoma astrocytes, and the
      coculture tube-formation assay compares whole astrocyte populations with no ADAMTSL3 gain or loss
      of function. Correlation presented as function.'
existing_annotations:
- term:
    id: GO:0031012
    label: extracellular matrix
  evidence_type: IBA
  original_reference_id: GO_REF:0000033
  qualifier: is_active_in
  supporting_entities:
  - FB:FBgn0003137
  - FB:FBgn0032252
  - MGI:MGI:109249
  - MGI:MGI:1347346
  - MGI:MGI:1347356
  - MGI:MGI:1913798
  - MGI:MGI:1925044
  - MGI:MGI:2386139
  - MGI:MGI:2389008
  - MGI:MGI:2672033
  - PANTHER:PTN000347317
  - RGD:621241
  - UniProtKB:Q6ZMM2
  - UniProtKB:Q8TE56
  - UniProtKB:Q9H324
  - UniProtKB:Q9P2N4
  - WB:WBGene00003242
  review:
    summary: 'Correct, well founded, and the best-supported statement in this gene''s record. Accepted
      as core.


      Resolving the WITH/FROM field settles it rather than merely making it plausible. The seventeen tokens
      are sixteen protein donors plus the PANTHER node PTN000347317, and every one of the sixteen resolves:
      Drosophila papilin and lonely heart, C. elegans mig-6/papilin, mouse Adamts1, Adamts2, Adamts7,
      Adamtsl2, Adamtsl4, Adamtsl5, Papln and Thsd4, rat Adamts1, and human ADAMTS9, ADAMTS10, ADAMTS17
      and ADAMTSL5. Asking what each donor carries for this term in its own right, rather than assuming,
      returns sixteen out of sixteen with their own IDA or HDA to GO:0031012 or a descendant. There is
      no self-referential token, so none of the support is circular. Two of the sixteen (Drosophila loh
      and mouse Adamtsl5) are unreviewed TrEMBL entries, which weakens their names as evidence but not
      their curated annotations.


      The donors are not, however, in the same place. Their experimental annotations are spread across
      GO:0005604 basement membrane (both papilins, mouse Adamts1), GO:0005614 interstitial matrix (mouse
      Adamtsl4), GO:0001527 microfibril (mouse Thsd4) and GO:0031012 itself. So the general term is the
      genuine least common ancestor of a heterogeneous donor set, and refining it downward would mean
      picking one donor''s compartment over another''s. No specificity upgrade is warranted, and the row
      is not less precise than its donors.


      ADAMTSL3 also has its own direct evidence for this location, which the propagation happens to agree
      with: punctin-2 expressed in COS-7 cells is deposited in the extracellular matrix, and UniProt records
      the subcellular location with ECO:0000269 against that paper. That experimental annotation has never
      been transferred into GOA, so the IBA is currently doing work that an IDA should be doing alongside
      it.'
    action: ACCEPT
    reason: Sixteen of sixteen protein donors carry their own experimental annotation to this term or
      a descendant, with no self-referential token; the donor set is heterogeneous across basement membrane,
      interstitial matrix and microfibril, so GO:0031012 is the correct LCA rather than an under-specified
      term; and the target has independent direct evidence for the same location from PMID:14667842.
    supported_by:
    - reference_id: file:human/ADAMTSL3/ADAMTSL3-bioinformatics/RESULTS.md
      supporting_text: '**16 of 16** carry their own experimental annotation to GO:0031012 or a descendant.'
    - reference_id: file:human/ADAMTSL3/ADAMTSL3-bioinformatics/RESULTS.md
      supporting_text: the donor set is heterogeneous in *which* matrix compartment it occupies, so the
        general parent is the least common ancestor rather than a curator failing to be specific.
    - reference_id: PMID:14667842
      supporting_text: Using these and a monoclonal antibody to a C-terminal myc tag, we show that in
        transfected COS-7 cells, punctin-2 is expressed as a 210-kDa glycoprotein that is located in the
        extracellular matrix.
      full_text_unavailable: true
    knowledge_gaps:
    - gap_statement: UniProt records ADAMTSL3's extracellular matrix localisation with ECO:0000269|PubMed:14667842,
        but no corresponding IDA row exists in GOA, so the only localisation annotation this gene carries
        for its best-established compartment is a phylogenetic inference.
      boundary: 'The experiment itself is not in doubt: a 210-kDa glycoprotein deposited in the matrix
        of transfected COS-7 cells, detected with two independent polyclonal antisera and a C-terminal
        myc tag.'
      gap_kind:
      - CURATION
      status: OPEN
      significance: A direct experimental localisation that exists in UniProt but not in GOA is invisible
        to any analysis that reads GOA, and it makes the gene look more computationally annotated than
        it is.
      resolution: Transfer the UniProt ECO:0000269 subcellular location to GOA as an IDA row.
      provenance:
      - reference_id: file:human/ADAMTSL3/ADAMTSL3-uniprot.txt
        supporting_text: 'CC   -!- SUBCELLULAR LOCATION: Secreted, extracellular space, extracellular'
    propagation_review:
      root_cause: NO_FAILURE_CORE
      source_entities:
      - source_id: PANTHER:PTN000347317
        source_label: PTHR13723 root node (ADAMTS and ADAMTS-like)
        source_status: SUPPORTS_TRANSFER
        comment: 'An internal tree node, not a protein. It carries four IBD annotations in PANTHER''s
          IBD.gaf: this term, GO:0030198, GO:0004222 metalloendopeptidase activity and GO:0006508 proteolysis.
          Only this one reaches ADAMTSL3; see the suggested question about where the catalytic-loss NOT
          is attached.'
      - source_id: FB:FBgn0003137
        source_label: Drosophila Ppn, papilin (Q868Z9, Swiss-Prot)
        source_status: SUPPORTS_TRANSFER
        comment: 'Two independent IDA annotations to GO:0005604 basement membrane (PMID:11076767, PMID:3320045).
          Ambiguous cross-reference: two TrEMBL/Swiss-Prot candidates, the reviewed one taken.'
      - source_id: FB:FBgn0032252
        source_label: Drosophila loh, lonely heart (Q9VKV3, TrEMBL)
        source_status: SUPPORTS_TRANSFER
        comment: IDA to GO:0031012 (PMID:23874219). Unreviewed entry, so its protein name is an automatic
          label; the curated annotation is nonetheless experimental. Four candidate accessions for this
          gene.
      - source_id: MGI:MGI:109249
        source_label: mouse Adamts1 (P97857, Swiss-Prot)
        source_status: SUPPORTS_TRANSFER
        comment: 'Three IDA annotations across GO:0005604 and GO:0031012. Resolved via the bare numeric
          MGI id; the doubled MGI:MGI: form returns HTTP 400.'
      - source_id: MGI:MGI:1347346
        source_label: mouse Adamts7 (Q68SA9, Swiss-Prot)
        source_status: SUPPORTS_TRANSFER
        comment: IDA to GO:0031012 (PMID:15192113).
      - source_id: MGI:MGI:1347356
        source_label: mouse Adamts2 (Q8C9W3, Swiss-Prot)
        source_status: SUPPORTS_TRANSFER
        comment: IDA to GO:0031012 (PMID:16556917).
      - source_id: MGI:MGI:1913798
        source_label: mouse Adamtsl5 (D3Z689, TrEMBL)
        source_status: SUPPORTS_TRANSFER
        comment: HDA and IDA to GO:0031012 (PMID:22159717, PMID:23010571). Unreviewed entry.
      - source_id: MGI:MGI:1925044
        source_label: mouse Adamtsl2 (Q7TSK7, Swiss-Prot)
        source_status: SUPPORTS_TRANSFER
        comment: IDA to GO:0031012 (PMID:17509843). This is the one family member for which PAINT has
          curated the loss of catalytic activity.
      - source_id: MGI:MGI:2386139
        source_label: mouse Papln, papilin (Q9EPX2, Swiss-Prot)
        source_status: SUPPORTS_TRANSFER
        comment: HDA to GO:0031012 and IDA to GO:0005604 basement membrane (PMID:18757743).
      - source_id: MGI:MGI:2389008
        source_label: mouse Adamtsl4 (Q80T21, Swiss-Prot)
        source_status: SUPPORTS_TRANSFER
        comment: IDA to GO:0005614 interstitial matrix and to GO:0031012 (PMID:18757743).
      - source_id: MGI:MGI:2672033
        source_label: mouse Thsd4 / ADAMTSL6 (Q3UTY6, Swiss-Prot)
        source_status: SUPPORTS_TRANSFER
        comment: IDA to GO:0001527 microfibril and to GO:0031012 (PMID:19940141). The most specific compartment
          in the donor set.
      - source_id: RGD:621241
        source_label: rat Adamts1 (Q9WUQ1, Swiss-Prot)
        source_status: SUPPORTS_TRANSFER
        comment: IDA to GO:0031012 (PMID:15777654).
      - source_id: UniProtKB:Q6ZMM2
        source_label: human ADAMTSL5 (Swiss-Prot)
        source_status: SUPPORTS_TRANSFER
        comment: IDA to GO:0031012 (PMID:23010571).
      - source_id: UniProtKB:Q8TE56
        source_label: human ADAMTS17 (Swiss-Prot)
        source_status: SUPPORTS_TRANSFER
        comment: IDA to GO:0031012 (PMID:28176809).
      - source_id: UniProtKB:Q9H324
        source_label: human ADAMTS10 (Swiss-Prot)
        source_status: SUPPORTS_TRANSFER
        comment: IDA to GO:0031012 (PMID:21402694). Also a direct high-affinity binding partner of ADAMTSL3
          (KD 2 nM, PMID:22242013), so this donor is functionally as well as phylogenetically connected.
      - source_id: UniProtKB:Q9P2N4
        source_label: human ADAMTS9 (Swiss-Prot)
        source_status: SUPPORTS_TRANSFER
        comment: IDA to GO:0031012 (PMID:12514189).
      - source_id: WB:WBGene00003242
        source_label: C. elegans mig-6, papilin (O76840, Swiss-Prot)
        source_status: SUPPORTS_TRANSFER
        comment: 'IDA to GO:0005604 basement membrane (PMID:19297413). Worth flagging for anyone re-deriving
          this row: this token does not resolve through UniProt''s xref:wormbase- index, which is keyed
          on transcript ids, and it is NOT madd-4 (WBGene00009958) even though madd-4 is the worm orthologue
          that dominates this gene''s neural literature.'
- term:
    id: GO:0030198
    label: extracellular matrix organization
  evidence_type: IEA
  original_reference_id: GO_REF:0000002
  qualifier: involved_in
  supporting_entities:
  - InterPro:IPR013273
  review:
    summary: 'Accepted, and it is better founded than its provenance suggests.


      As delivered this is a family-level signature mapping: IPR013273 is the ADAMTS/ADAMTS-like PRINTS
      signature, which matches catalytic ADAMTS proteinases and non-catalytic ADAMTS-like proteins alike,
      so on its own it says only that ADAMTSL3 belongs to a family that remodels matrix. For a protein
      with no catalytic domain that would be a thin argument, since the family''s best-known route to
      matrix organisation is proteolysis, which ADAMTSL3 cannot perform.


      Three independent lines of evidence nonetheless support the term for this protein specifically.
      It binds the N-terminal, assembly-competent half of fibrillin-1 by surface plasmon resonance, and
      the Weill-Marchesani three-domain deletion in fibrillin-1 abolishes that binding, so the interaction
      maps to a defined region of a microfibril scaffold protein. Adamtsl3 knock-out mice show altered
      matrix output and TGF-beta activity in the pressure-overloaded heart, and ADAMTSL3 overexpression
      in human cardiac fibroblasts lowers collagen synthesis and myofibroblast differentiation. And in
      the brain, loss of Adamtsl3 reduces perineuronal nets, which are a specialised extracellular matrix,
      in a cell-autonomous manner in parvalbumin interneurons.


      So the term is right, and the interesting point is that the reasoning behind the automatic assignment
      is not the reasoning that justifies it. Kept general rather than refined: the three lines of evidence
      point at three different matrices (microfibril, cardiac interstitium, perineuronal net), which is
      exactly the situation in which the parent term is the correct call.'
    action: ACCEPT
    reason: 'A family-level InterPro signature term that independent gene-specific evidence happens to
      support: fibrillin-1 binding by SPR with a mapped binding site, altered cardiac matrix output in
      Adamtsl3-null mice with a human-cell gain-of-function counterpart, and cell-autonomous loss of perineuronal
      nets in Adamtsl3-null parvalbumin interneurons. Left general because those three lines implicate
      three different matrices.'
    additional_reference_ids:
    - PMID:22242013
    - PMID:36539599
    - PMID:42277231
    supported_by:
    - reference_id: file:human/ADAMTSL3/ADAMTSL3-uniprot.txt
      supporting_text: DR   InterPro; IPR013273; ADAMTS/ADAMTS-like.
    - reference_id: PMID:22242013
      supporting_text: ADAMTSL-2, -3, and -6 and papilin polypeptides did not bind to recombinant fibrillin-1
        polypeptides with the WMS three-domain deletion (Figure 5a and Table S1).
    - reference_id: PMID:42277231
      supporting_text: Here, we identify the schizophrenia-associated glycoprotein Adamtsl3 as a PV+ cell-autonomous
        regulator of PNN integrity.
      full_text_unavailable: true
- term:
    id: GO:0005515
    label: protein binding
  evidence_type: IPI
  original_reference_id: PMID:25416956
  qualifier: enables
  supporting_entities:
  - UniProtKB:P0C7H8
  review:
    summary: 'Over-annotated. KRTAP2-3 (P0C7H8, 128 aa) is a cysteine-rich hair-cortex keratin-associated
      protein. The pair was reported only in HI-II-14, the CCSB proteome-scale binary interactome map
      (PMID:25416956), and KRTAP2-3 has 10 distinct partners in IntAct - the one partner in this set that
      is not a high-degree prey.


      All fifteen GO:0005515 rows on this gene come from two systematic yeast two-hybrid interactome maps
      and nothing else. Querying IntAct directly for P82987 returns the same pairs logged under three
      method names - two hybrid array, two hybrid prey pooling approach and validated two hybrid - which
      are sub-methods of one Y2H pipeline rather than three independent assays, so the NbExp values in
      the UniProt INTERACTION block count Y2H variants and not orthogonal confirmation. Disaggregating
      IntAct by pair rather than by gene - which is the only way the question can be answered, because
      ADAMTSL3''s gene-level record also covers 17 partners that are not in GOA - shows that 0 of these
      13 pairs carries any non-Y2H detection method. The anti-tag co-immunoprecipitation and BioID methods
      present in ADAMTSL3''s IntAct record belong entirely to those other 17 partners, from separate publications
      that are not the source of any GOA row reviewed here.


      The partner set behaves like screen background rather than like biology. Nine of the thirteen partners
      are hair keratins or keratin-associated proteins and a tenth, CYSRT1, is a cornified-envelope protein
      embedded in the same interaction network; the remaining three are cytosolic or nuclear. The median
      number of distinct IntAct partners across the thirteen is 188 against 30 for ADAMTSL3 itself, and
      six of them exceed 180, so most of this set consists of high-degree promiscuous preys.


      There is also a compartment problem that the assay cannot see. ADAMTSL3 has a cleaved signal peptide
      and is secreted into the extracellular matrix; hair keratin-associated proteins are intracellular
      constituents of the hair cortex, GLRX3 is cytosolic and MDFI is nuclear and cytoplasmic. In a two-hybrid
      assay both proteins are expressed in the yeast nucleus, where a signal peptide is inert, so the
      pairing is never tested in a compartment where it could occur.


      Marked over-annotated rather than removed. These are real observations recorded correctly by IntAct
      curators from published screens, and a screen hit can turn out to matter; what it does not do is
      license a functional claim, and bare protein binding conveys none in any case.'
    action: MARK_AS_OVER_ANNOTATED
    reason: Unreplicated systematic yeast two-hybrid hit. KRTAP2-3 is a cysteine-rich hair-cortex keratin-associated
      protein and carries 10 distinct IntAct partners against ADAMTSL3's 30. Disaggregating IntAct per
      pair rather than per gene shows this pair, like all 13, carries no non-Y2H detection method; it
      has no orthogonal assay, no follow-up in the ADAMTSL3 literature, and ADAMTSL3 is a secreted matrix
      protein that does not share a compartment with it in vivo. Bare protein binding conveys no function
      even where the pairing is real.
    supported_by:
    - reference_id: file:human/ADAMTSL3/ADAMTSL3-bioinformatics/RESULTS.md
      supporting_text: ADAMTSL3 itself has 30 distinct IntAct partners; the median for its 13 annotated
        partners is 188.
    - reference_id: file:human/ADAMTSL3/ADAMTSL3-bioinformatics/RESULTS.md
      supporting_text: '**GOA partners with any non-Y2H detection method: 0 of 13** - every one of these
        pairs rests on yeast two-hybrid sub-methods alone, with no orthogonal assay.'
    - reference_id: file:human/ADAMTSL3/ADAMTSL3-uniprot.txt
      supporting_text: FT   SIGNAL          1..26
- term:
    id: GO:0005515
    label: protein binding
  evidence_type: IPI
  original_reference_id: PMID:25416956
  qualifier: enables
  supporting_entities:
  - UniProtKB:P60410
  review:
    summary: 'Over-annotated. KRTAP10-8 (P60410, 259 aa) is a cysteine-rich hair-cortex keratin-associated
      protein. The pair was reported only in HI-II-14, the CCSB proteome-scale binary interactome map
      (PMID:25416956), and KRTAP10-8 has 415 distinct partners in IntAct - an extreme interactome hub.


      All fifteen GO:0005515 rows on this gene come from two systematic yeast two-hybrid interactome maps
      and nothing else. Querying IntAct directly for P82987 returns the same pairs logged under three
      method names - two hybrid array, two hybrid prey pooling approach and validated two hybrid - which
      are sub-methods of one Y2H pipeline rather than three independent assays, so the NbExp values in
      the UniProt INTERACTION block count Y2H variants and not orthogonal confirmation. Disaggregating
      IntAct by pair rather than by gene - which is the only way the question can be answered, because
      ADAMTSL3''s gene-level record also covers 17 partners that are not in GOA - shows that 0 of these
      13 pairs carries any non-Y2H detection method. The anti-tag co-immunoprecipitation and BioID methods
      present in ADAMTSL3''s IntAct record belong entirely to those other 17 partners, from separate publications
      that are not the source of any GOA row reviewed here.


      The partner set behaves like screen background rather than like biology. Nine of the thirteen partners
      are hair keratins or keratin-associated proteins and a tenth, CYSRT1, is a cornified-envelope protein
      embedded in the same interaction network; the remaining three are cytosolic or nuclear. The median
      number of distinct IntAct partners across the thirteen is 188 against 30 for ADAMTSL3 itself, and
      six of them exceed 180, so most of this set consists of high-degree promiscuous preys.


      There is also a compartment problem that the assay cannot see. ADAMTSL3 has a cleaved signal peptide
      and is secreted into the extracellular matrix; hair keratin-associated proteins are intracellular
      constituents of the hair cortex, GLRX3 is cytosolic and MDFI is nuclear and cytoplasmic. In a two-hybrid
      assay both proteins are expressed in the yeast nucleus, where a signal peptide is inert, so the
      pairing is never tested in a compartment where it could occur.


      Marked over-annotated rather than removed. These are real observations recorded correctly by IntAct
      curators from published screens, and a screen hit can turn out to matter; what it does not do is
      license a functional claim, and bare protein binding conveys none in any case.'
    action: MARK_AS_OVER_ANNOTATED
    reason: Unreplicated systematic yeast two-hybrid hit. KRTAP10-8 is a cysteine-rich hair-cortex keratin-associated
      protein and carries 415 distinct IntAct partners against ADAMTSL3's 30. Disaggregating IntAct per
      pair rather than per gene shows this pair, like all 13, carries no non-Y2H detection method; it
      has no orthogonal assay, no follow-up in the ADAMTSL3 literature, and ADAMTSL3 is a secreted matrix
      protein that does not share a compartment with it in vivo. Bare protein binding conveys no function
      even where the pairing is real. Seen in both HI-II-14 and HuRI, but those two maps share the CCSB
      ORFeome and search space, so the repeat is not independent replication - and KRTAP10-8 is one of
      the highest-degree preys in the set.
    supported_by:
    - reference_id: file:human/ADAMTSL3/ADAMTSL3-bioinformatics/RESULTS.md
      supporting_text: ADAMTSL3 itself has 30 distinct IntAct partners; the median for its 13 annotated
        partners is 188.
    - reference_id: file:human/ADAMTSL3/ADAMTSL3-bioinformatics/RESULTS.md
      supporting_text: '**GOA partners with any non-Y2H detection method: 0 of 13** - every one of these
        pairs rests on yeast two-hybrid sub-methods alone, with no orthogonal assay.'
    - reference_id: file:human/ADAMTSL3/ADAMTSL3-uniprot.txt
      supporting_text: FT   SIGNAL          1..26
- term:
    id: GO:0005515
    label: protein binding
  evidence_type: IPI
  original_reference_id: PMID:25416956
  qualifier: enables
  supporting_entities:
  - UniProtKB:Q6A162
  review:
    summary: 'Over-annotated. KRT40 (Q6A162, 431 aa) is a type I hair-follicle intermediate-filament keratin.
      The pair was reported only in HI-II-14, the CCSB proteome-scale binary interactome map (PMID:25416956),
      and KRT40 has 448 distinct partners in IntAct - an extreme interactome hub.


      All fifteen GO:0005515 rows on this gene come from two systematic yeast two-hybrid interactome maps
      and nothing else. Querying IntAct directly for P82987 returns the same pairs logged under three
      method names - two hybrid array, two hybrid prey pooling approach and validated two hybrid - which
      are sub-methods of one Y2H pipeline rather than three independent assays, so the NbExp values in
      the UniProt INTERACTION block count Y2H variants and not orthogonal confirmation. Disaggregating
      IntAct by pair rather than by gene - which is the only way the question can be answered, because
      ADAMTSL3''s gene-level record also covers 17 partners that are not in GOA - shows that 0 of these
      13 pairs carries any non-Y2H detection method. The anti-tag co-immunoprecipitation and BioID methods
      present in ADAMTSL3''s IntAct record belong entirely to those other 17 partners, from separate publications
      that are not the source of any GOA row reviewed here.


      The partner set behaves like screen background rather than like biology. Nine of the thirteen partners
      are hair keratins or keratin-associated proteins and a tenth, CYSRT1, is a cornified-envelope protein
      embedded in the same interaction network; the remaining three are cytosolic or nuclear. The median
      number of distinct IntAct partners across the thirteen is 188 against 30 for ADAMTSL3 itself, and
      six of them exceed 180, so most of this set consists of high-degree promiscuous preys.


      There is also a compartment problem that the assay cannot see. ADAMTSL3 has a cleaved signal peptide
      and is secreted into the extracellular matrix; hair keratin-associated proteins are intracellular
      constituents of the hair cortex, GLRX3 is cytosolic and MDFI is nuclear and cytoplasmic. In a two-hybrid
      assay both proteins are expressed in the yeast nucleus, where a signal peptide is inert, so the
      pairing is never tested in a compartment where it could occur.


      Marked over-annotated rather than removed. These are real observations recorded correctly by IntAct
      curators from published screens, and a screen hit can turn out to matter; what it does not do is
      license a functional claim, and bare protein binding conveys none in any case.'
    action: MARK_AS_OVER_ANNOTATED
    reason: Unreplicated systematic yeast two-hybrid hit. KRT40 is a type I hair-follicle intermediate-filament
      keratin and carries 448 distinct IntAct partners against ADAMTSL3's 30. Disaggregating IntAct per
      pair rather than per gene shows this pair, like all 13, carries no non-Y2H detection method; it
      has no orthogonal assay, no follow-up in the ADAMTSL3 literature, and ADAMTSL3 is a secreted matrix
      protein that does not share a compartment with it in vivo. Bare protein binding conveys no function
      even where the pairing is real.
    supported_by:
    - reference_id: file:human/ADAMTSL3/ADAMTSL3-bioinformatics/RESULTS.md
      supporting_text: ADAMTSL3 itself has 30 distinct IntAct partners; the median for its 13 annotated
        partners is 188.
    - reference_id: file:human/ADAMTSL3/ADAMTSL3-bioinformatics/RESULTS.md
      supporting_text: '**GOA partners with any non-Y2H detection method: 0 of 13** - every one of these
        pairs rests on yeast two-hybrid sub-methods alone, with no orthogonal assay.'
    - reference_id: file:human/ADAMTSL3/ADAMTSL3-uniprot.txt
      supporting_text: FT   SIGNAL          1..26
- term:
    id: GO:0005515
    label: protein binding
  evidence_type: IPI
  original_reference_id: PMID:25416956
  qualifier: enables
  supporting_entities:
  - UniProtKB:Q7Z3S9
  review:
    summary: 'Over-annotated. NOTCH2NLA (Q7Z3S9, 236 aa) is a hominid-specific partial NOTCH2 duplicate
      acting on NOTCH2 in the secretory pathway. The pair was reported only in HI-II-14, the CCSB proteome-scale
      binary interactome map (PMID:25416956), and NOTCH2NLA has 276 distinct partners in IntAct - an extreme
      interactome hub.


      All fifteen GO:0005515 rows on this gene come from two systematic yeast two-hybrid interactome maps
      and nothing else. Querying IntAct directly for P82987 returns the same pairs logged under three
      method names - two hybrid array, two hybrid prey pooling approach and validated two hybrid - which
      are sub-methods of one Y2H pipeline rather than three independent assays, so the NbExp values in
      the UniProt INTERACTION block count Y2H variants and not orthogonal confirmation. Disaggregating
      IntAct by pair rather than by gene - which is the only way the question can be answered, because
      ADAMTSL3''s gene-level record also covers 17 partners that are not in GOA - shows that 0 of these
      13 pairs carries any non-Y2H detection method. The anti-tag co-immunoprecipitation and BioID methods
      present in ADAMTSL3''s IntAct record belong entirely to those other 17 partners, from separate publications
      that are not the source of any GOA row reviewed here.


      The partner set behaves like screen background rather than like biology. Nine of the thirteen partners
      are hair keratins or keratin-associated proteins and a tenth, CYSRT1, is a cornified-envelope protein
      embedded in the same interaction network; the remaining three are cytosolic or nuclear. The median
      number of distinct IntAct partners across the thirteen is 188 against 30 for ADAMTSL3 itself, and
      six of them exceed 180, so most of this set consists of high-degree promiscuous preys.


      There is also a compartment problem that the assay cannot see. ADAMTSL3 has a cleaved signal peptide
      and is secreted into the extracellular matrix; hair keratin-associated proteins are intracellular
      constituents of the hair cortex, GLRX3 is cytosolic and MDFI is nuclear and cytoplasmic. In a two-hybrid
      assay both proteins are expressed in the yeast nucleus, where a signal peptide is inert, so the
      pairing is never tested in a compartment where it could occur.


      Marked over-annotated rather than removed. These are real observations recorded correctly by IntAct
      curators from published screens, and a screen hit can turn out to matter; what it does not do is
      license a functional claim, and bare protein binding conveys none in any case.'
    action: MARK_AS_OVER_ANNOTATED
    reason: Unreplicated systematic yeast two-hybrid hit. NOTCH2NLA is a hominid-specific partial NOTCH2
      duplicate acting on NOTCH2 in the secretory pathway and carries 276 distinct IntAct partners against
      ADAMTSL3's 30. Disaggregating IntAct per pair rather than per gene shows this pair, like all 13,
      carries no non-Y2H detection method; it has no orthogonal assay, no follow-up in the ADAMTSL3 literature,
      and ADAMTSL3 is a secreted matrix protein that does not share a compartment with it in vivo. Bare
      protein binding conveys no function even where the pairing is real.
    supported_by:
    - reference_id: file:human/ADAMTSL3/ADAMTSL3-bioinformatics/RESULTS.md
      supporting_text: ADAMTSL3 itself has 30 distinct IntAct partners; the median for its 13 annotated
        partners is 188.
    - reference_id: file:human/ADAMTSL3/ADAMTSL3-bioinformatics/RESULTS.md
      supporting_text: '**GOA partners with any non-Y2H detection method: 0 of 13** - every one of these
        pairs rests on yeast two-hybrid sub-methods alone, with no orthogonal assay.'
    - reference_id: file:human/ADAMTSL3/ADAMTSL3-uniprot.txt
      supporting_text: FT   SIGNAL          1..26
- term:
    id: GO:0005515
    label: protein binding
  evidence_type: IPI
  original_reference_id: PMID:25416956
  qualifier: enables
  supporting_entities:
  - UniProtKB:Q99750
  review:
    summary: 'Over-annotated. MDFI (Q99750, 246 aa) is a nuclear and cytoplasmic inhibitor of MyoD-family
      bHLH transcription factors. The pair was reported only in HI-II-14, the CCSB proteome-scale binary
      interactome map (PMID:25416956), and MDFI has 483 distinct partners in IntAct - an extreme interactome
      hub.


      All fifteen GO:0005515 rows on this gene come from two systematic yeast two-hybrid interactome maps
      and nothing else. Querying IntAct directly for P82987 returns the same pairs logged under three
      method names - two hybrid array, two hybrid prey pooling approach and validated two hybrid - which
      are sub-methods of one Y2H pipeline rather than three independent assays, so the NbExp values in
      the UniProt INTERACTION block count Y2H variants and not orthogonal confirmation. Disaggregating
      IntAct by pair rather than by gene - which is the only way the question can be answered, because
      ADAMTSL3''s gene-level record also covers 17 partners that are not in GOA - shows that 0 of these
      13 pairs carries any non-Y2H detection method. The anti-tag co-immunoprecipitation and BioID methods
      present in ADAMTSL3''s IntAct record belong entirely to those other 17 partners, from separate publications
      that are not the source of any GOA row reviewed here.


      The partner set behaves like screen background rather than like biology. Nine of the thirteen partners
      are hair keratins or keratin-associated proteins and a tenth, CYSRT1, is a cornified-envelope protein
      embedded in the same interaction network; the remaining three are cytosolic or nuclear. The median
      number of distinct IntAct partners across the thirteen is 188 against 30 for ADAMTSL3 itself, and
      six of them exceed 180, so most of this set consists of high-degree promiscuous preys.


      There is also a compartment problem that the assay cannot see. ADAMTSL3 has a cleaved signal peptide
      and is secreted into the extracellular matrix; hair keratin-associated proteins are intracellular
      constituents of the hair cortex, GLRX3 is cytosolic and MDFI is nuclear and cytoplasmic. In a two-hybrid
      assay both proteins are expressed in the yeast nucleus, where a signal peptide is inert, so the
      pairing is never tested in a compartment where it could occur.


      Marked over-annotated rather than removed. These are real observations recorded correctly by IntAct
      curators from published screens, and a screen hit can turn out to matter; what it does not do is
      license a functional claim, and bare protein binding conveys none in any case.'
    action: MARK_AS_OVER_ANNOTATED
    reason: Unreplicated systematic yeast two-hybrid hit. MDFI is a nuclear and cytoplasmic inhibitor
      of MyoD-family bHLH transcription factors and carries 483 distinct IntAct partners against ADAMTSL3's
      30. Disaggregating IntAct per pair rather than per gene shows this pair, like all 13, carries no
      non-Y2H detection method; it has no orthogonal assay, no follow-up in the ADAMTSL3 literature, and
      ADAMTSL3 is a secreted matrix protein that does not share a compartment with it in vivo. Bare protein
      binding conveys no function even where the pairing is real. Seen in both HI-II-14 and HuRI, but
      those two maps share the CCSB ORFeome and search space, so the repeat is not independent replication
      - and MDFI is one of the highest-degree preys in the set.
    supported_by:
    - reference_id: file:human/ADAMTSL3/ADAMTSL3-bioinformatics/RESULTS.md
      supporting_text: ADAMTSL3 itself has 30 distinct IntAct partners; the median for its 13 annotated
        partners is 188.
    - reference_id: file:human/ADAMTSL3/ADAMTSL3-bioinformatics/RESULTS.md
      supporting_text: '**GOA partners with any non-Y2H detection method: 0 of 13** - every one of these
        pairs rests on yeast two-hybrid sub-methods alone, with no orthogonal assay.'
    - reference_id: file:human/ADAMTSL3/ADAMTSL3-uniprot.txt
      supporting_text: FT   SIGNAL          1..26
- term:
    id: GO:0005515
    label: protein binding
  evidence_type: IPI
  original_reference_id: PMID:32296183
  qualifier: enables
  supporting_entities:
  - UniProtKB:A8MQ03
  review:
    summary: 'Over-annotated. CYSRT1 (A8MQ03, 144 aa) is a cornified-envelope protein of the keratin/KRTAP
      interaction network. The pair was reported only in HuRI, the CCSB human reference binary interactome
      map (PMID:32296183), and CYSRT1 has 516 distinct partners in IntAct - an extreme interactome hub.


      All fifteen GO:0005515 rows on this gene come from two systematic yeast two-hybrid interactome maps
      and nothing else. Querying IntAct directly for P82987 returns the same pairs logged under three
      method names - two hybrid array, two hybrid prey pooling approach and validated two hybrid - which
      are sub-methods of one Y2H pipeline rather than three independent assays, so the NbExp values in
      the UniProt INTERACTION block count Y2H variants and not orthogonal confirmation. Disaggregating
      IntAct by pair rather than by gene - which is the only way the question can be answered, because
      ADAMTSL3''s gene-level record also covers 17 partners that are not in GOA - shows that 0 of these
      13 pairs carries any non-Y2H detection method. The anti-tag co-immunoprecipitation and BioID methods
      present in ADAMTSL3''s IntAct record belong entirely to those other 17 partners, from separate publications
      that are not the source of any GOA row reviewed here.


      The partner set behaves like screen background rather than like biology. Nine of the thirteen partners
      are hair keratins or keratin-associated proteins and a tenth, CYSRT1, is a cornified-envelope protein
      embedded in the same interaction network; the remaining three are cytosolic or nuclear. The median
      number of distinct IntAct partners across the thirteen is 188 against 30 for ADAMTSL3 itself, and
      six of them exceed 180, so most of this set consists of high-degree promiscuous preys.


      There is also a compartment problem that the assay cannot see. ADAMTSL3 has a cleaved signal peptide
      and is secreted into the extracellular matrix; hair keratin-associated proteins are intracellular
      constituents of the hair cortex, GLRX3 is cytosolic and MDFI is nuclear and cytoplasmic. In a two-hybrid
      assay both proteins are expressed in the yeast nucleus, where a signal peptide is inert, so the
      pairing is never tested in a compartment where it could occur.


      Marked over-annotated rather than removed. These are real observations recorded correctly by IntAct
      curators from published screens, and a screen hit can turn out to matter; what it does not do is
      license a functional claim, and bare protein binding conveys none in any case.'
    action: MARK_AS_OVER_ANNOTATED
    reason: Unreplicated systematic yeast two-hybrid hit. CYSRT1 is a cornified-envelope protein of the
      keratin/KRTAP interaction network and carries 516 distinct IntAct partners against ADAMTSL3's 30.
      Disaggregating IntAct per pair rather than per gene shows this pair, like all 13, carries no non-Y2H
      detection method; it has no orthogonal assay, no follow-up in the ADAMTSL3 literature, and ADAMTSL3
      is a secreted matrix protein that does not share a compartment with it in vivo. Bare protein binding
      conveys no function even where the pairing is real.
    supported_by:
    - reference_id: file:human/ADAMTSL3/ADAMTSL3-bioinformatics/RESULTS.md
      supporting_text: ADAMTSL3 itself has 30 distinct IntAct partners; the median for its 13 annotated
        partners is 188.
    - reference_id: file:human/ADAMTSL3/ADAMTSL3-bioinformatics/RESULTS.md
      supporting_text: '**GOA partners with any non-Y2H detection method: 0 of 13** - every one of these
        pairs rests on yeast two-hybrid sub-methods alone, with no orthogonal assay.'
    - reference_id: file:human/ADAMTSL3/ADAMTSL3-uniprot.txt
      supporting_text: FT   SIGNAL          1..26
- term:
    id: GO:0005515
    label: protein binding
  evidence_type: IPI
  original_reference_id: PMID:32296183
  qualifier: enables
  supporting_entities:
  - UniProtKB:O76003
  review:
    summary: 'Over-annotated. GLRX3 (O76003, 335 aa) is a cytosolic monothiol glutaredoxin and cytosolic
      iron-sulfur assembly factor. The pair was reported only in HuRI, the CCSB human reference binary
      interactome map (PMID:32296183), and GLRX3 has 190 distinct partners in IntAct - an extreme interactome
      hub.


      All fifteen GO:0005515 rows on this gene come from two systematic yeast two-hybrid interactome maps
      and nothing else. Querying IntAct directly for P82987 returns the same pairs logged under three
      method names - two hybrid array, two hybrid prey pooling approach and validated two hybrid - which
      are sub-methods of one Y2H pipeline rather than three independent assays, so the NbExp values in
      the UniProt INTERACTION block count Y2H variants and not orthogonal confirmation. Disaggregating
      IntAct by pair rather than by gene - which is the only way the question can be answered, because
      ADAMTSL3''s gene-level record also covers 17 partners that are not in GOA - shows that 0 of these
      13 pairs carries any non-Y2H detection method. The anti-tag co-immunoprecipitation and BioID methods
      present in ADAMTSL3''s IntAct record belong entirely to those other 17 partners, from separate publications
      that are not the source of any GOA row reviewed here.


      The partner set behaves like screen background rather than like biology. Nine of the thirteen partners
      are hair keratins or keratin-associated proteins and a tenth, CYSRT1, is a cornified-envelope protein
      embedded in the same interaction network; the remaining three are cytosolic or nuclear. The median
      number of distinct IntAct partners across the thirteen is 188 against 30 for ADAMTSL3 itself, and
      six of them exceed 180, so most of this set consists of high-degree promiscuous preys.


      There is also a compartment problem that the assay cannot see. ADAMTSL3 has a cleaved signal peptide
      and is secreted into the extracellular matrix; hair keratin-associated proteins are intracellular
      constituents of the hair cortex, GLRX3 is cytosolic and MDFI is nuclear and cytoplasmic. In a two-hybrid
      assay both proteins are expressed in the yeast nucleus, where a signal peptide is inert, so the
      pairing is never tested in a compartment where it could occur.


      Marked over-annotated rather than removed. These are real observations recorded correctly by IntAct
      curators from published screens, and a screen hit can turn out to matter; what it does not do is
      license a functional claim, and bare protein binding conveys none in any case.'
    action: MARK_AS_OVER_ANNOTATED
    reason: Unreplicated systematic yeast two-hybrid hit. GLRX3 is a cytosolic monothiol glutaredoxin
      and cytosolic iron-sulfur assembly factor and carries 190 distinct IntAct partners against ADAMTSL3's
      30. Disaggregating IntAct per pair rather than per gene shows this pair, like all 13, carries no
      non-Y2H detection method; it has no orthogonal assay, no follow-up in the ADAMTSL3 literature, and
      ADAMTSL3 is a secreted matrix protein that does not share a compartment with it in vivo. Bare protein
      binding conveys no function even where the pairing is real.
    supported_by:
    - reference_id: file:human/ADAMTSL3/ADAMTSL3-bioinformatics/RESULTS.md
      supporting_text: ADAMTSL3 itself has 30 distinct IntAct partners; the median for its 13 annotated
        partners is 188.
    - reference_id: file:human/ADAMTSL3/ADAMTSL3-bioinformatics/RESULTS.md
      supporting_text: '**GOA partners with any non-Y2H detection method: 0 of 13** - every one of these
        pairs rests on yeast two-hybrid sub-methods alone, with no orthogonal assay.'
    - reference_id: file:human/ADAMTSL3/ADAMTSL3-uniprot.txt
      supporting_text: FT   SIGNAL          1..26
- term:
    id: GO:0005515
    label: protein binding
  evidence_type: IPI
  original_reference_id: PMID:32296183
  qualifier: enables
  supporting_entities:
  - UniProtKB:P60328
  review:
    summary: 'Over-annotated. KRTAP12-3 (P60328, 96 aa) is a cysteine-rich hair-cortex keratin-associated
      protein. The pair was reported only in HuRI, the CCSB human reference binary interactome map (PMID:32296183),
      and KRTAP12-3 has 141 distinct partners in IntAct - a moderately connected prey.


      All fifteen GO:0005515 rows on this gene come from two systematic yeast two-hybrid interactome maps
      and nothing else. Querying IntAct directly for P82987 returns the same pairs logged under three
      method names - two hybrid array, two hybrid prey pooling approach and validated two hybrid - which
      are sub-methods of one Y2H pipeline rather than three independent assays, so the NbExp values in
      the UniProt INTERACTION block count Y2H variants and not orthogonal confirmation. Disaggregating
      IntAct by pair rather than by gene - which is the only way the question can be answered, because
      ADAMTSL3''s gene-level record also covers 17 partners that are not in GOA - shows that 0 of these
      13 pairs carries any non-Y2H detection method. The anti-tag co-immunoprecipitation and BioID methods
      present in ADAMTSL3''s IntAct record belong entirely to those other 17 partners, from separate publications
      that are not the source of any GOA row reviewed here.


      The partner set behaves like screen background rather than like biology. Nine of the thirteen partners
      are hair keratins or keratin-associated proteins and a tenth, CYSRT1, is a cornified-envelope protein
      embedded in the same interaction network; the remaining three are cytosolic or nuclear. The median
      number of distinct IntAct partners across the thirteen is 188 against 30 for ADAMTSL3 itself, and
      six of them exceed 180, so most of this set consists of high-degree promiscuous preys.


      There is also a compartment problem that the assay cannot see. ADAMTSL3 has a cleaved signal peptide
      and is secreted into the extracellular matrix; hair keratin-associated proteins are intracellular
      constituents of the hair cortex, GLRX3 is cytosolic and MDFI is nuclear and cytoplasmic. In a two-hybrid
      assay both proteins are expressed in the yeast nucleus, where a signal peptide is inert, so the
      pairing is never tested in a compartment where it could occur.


      Marked over-annotated rather than removed. These are real observations recorded correctly by IntAct
      curators from published screens, and a screen hit can turn out to matter; what it does not do is
      license a functional claim, and bare protein binding conveys none in any case.'
    action: MARK_AS_OVER_ANNOTATED
    reason: Unreplicated systematic yeast two-hybrid hit. KRTAP12-3 is a cysteine-rich hair-cortex keratin-associated
      protein and carries 141 distinct IntAct partners against ADAMTSL3's 30. Disaggregating IntAct per
      pair rather than per gene shows this pair, like all 13, carries no non-Y2H detection method; it
      has no orthogonal assay, no follow-up in the ADAMTSL3 literature, and ADAMTSL3 is a secreted matrix
      protein that does not share a compartment with it in vivo. Bare protein binding conveys no function
      even where the pairing is real.
    supported_by:
    - reference_id: file:human/ADAMTSL3/ADAMTSL3-bioinformatics/RESULTS.md
      supporting_text: ADAMTSL3 itself has 30 distinct IntAct partners; the median for its 13 annotated
        partners is 188.
    - reference_id: file:human/ADAMTSL3/ADAMTSL3-bioinformatics/RESULTS.md
      supporting_text: '**GOA partners with any non-Y2H detection method: 0 of 13** - every one of these
        pairs rests on yeast two-hybrid sub-methods alone, with no orthogonal assay.'
    - reference_id: file:human/ADAMTSL3/ADAMTSL3-uniprot.txt
      supporting_text: FT   SIGNAL          1..26
- term:
    id: GO:0005515
    label: protein binding
  evidence_type: IPI
  original_reference_id: PMID:32296183
  qualifier: enables
  supporting_entities:
  - UniProtKB:P60371
  review:
    summary: 'Over-annotated. KRTAP10-6 (P60371, 365 aa) is a cysteine-rich hair-cortex keratin-associated
      protein. The pair was reported only in HuRI, the CCSB human reference binary interactome map (PMID:32296183),
      and KRTAP10-6 has 77 distinct partners in IntAct - a moderately connected prey.


      All fifteen GO:0005515 rows on this gene come from two systematic yeast two-hybrid interactome maps
      and nothing else. Querying IntAct directly for P82987 returns the same pairs logged under three
      method names - two hybrid array, two hybrid prey pooling approach and validated two hybrid - which
      are sub-methods of one Y2H pipeline rather than three independent assays, so the NbExp values in
      the UniProt INTERACTION block count Y2H variants and not orthogonal confirmation. Disaggregating
      IntAct by pair rather than by gene - which is the only way the question can be answered, because
      ADAMTSL3''s gene-level record also covers 17 partners that are not in GOA - shows that 0 of these
      13 pairs carries any non-Y2H detection method. The anti-tag co-immunoprecipitation and BioID methods
      present in ADAMTSL3''s IntAct record belong entirely to those other 17 partners, from separate publications
      that are not the source of any GOA row reviewed here.


      The partner set behaves like screen background rather than like biology. Nine of the thirteen partners
      are hair keratins or keratin-associated proteins and a tenth, CYSRT1, is a cornified-envelope protein
      embedded in the same interaction network; the remaining three are cytosolic or nuclear. The median
      number of distinct IntAct partners across the thirteen is 188 against 30 for ADAMTSL3 itself, and
      six of them exceed 180, so most of this set consists of high-degree promiscuous preys.


      There is also a compartment problem that the assay cannot see. ADAMTSL3 has a cleaved signal peptide
      and is secreted into the extracellular matrix; hair keratin-associated proteins are intracellular
      constituents of the hair cortex, GLRX3 is cytosolic and MDFI is nuclear and cytoplasmic. In a two-hybrid
      assay both proteins are expressed in the yeast nucleus, where a signal peptide is inert, so the
      pairing is never tested in a compartment where it could occur.


      Marked over-annotated rather than removed. These are real observations recorded correctly by IntAct
      curators from published screens, and a screen hit can turn out to matter; what it does not do is
      license a functional claim, and bare protein binding conveys none in any case.'
    action: MARK_AS_OVER_ANNOTATED
    reason: Unreplicated systematic yeast two-hybrid hit. KRTAP10-6 is a cysteine-rich hair-cortex keratin-associated
      protein and carries 77 distinct IntAct partners against ADAMTSL3's 30. Disaggregating IntAct per
      pair rather than per gene shows this pair, like all 13, carries no non-Y2H detection method; it
      has no orthogonal assay, no follow-up in the ADAMTSL3 literature, and ADAMTSL3 is a secreted matrix
      protein that does not share a compartment with it in vivo. Bare protein binding conveys no function
      even where the pairing is real.
    supported_by:
    - reference_id: file:human/ADAMTSL3/ADAMTSL3-bioinformatics/RESULTS.md
      supporting_text: ADAMTSL3 itself has 30 distinct IntAct partners; the median for its 13 annotated
        partners is 188.
    - reference_id: file:human/ADAMTSL3/ADAMTSL3-bioinformatics/RESULTS.md
      supporting_text: '**GOA partners with any non-Y2H detection method: 0 of 13** - every one of these
        pairs rests on yeast two-hybrid sub-methods alone, with no orthogonal assay.'
    - reference_id: file:human/ADAMTSL3/ADAMTSL3-uniprot.txt
      supporting_text: FT   SIGNAL          1..26
- term:
    id: GO:0005515
    label: protein binding
  evidence_type: IPI
  original_reference_id: PMID:32296183
  qualifier: enables
  supporting_entities:
  - UniProtKB:P60410
  review:
    summary: 'Over-annotated. KRTAP10-8 (P60410, 259 aa) is a cysteine-rich hair-cortex keratin-associated
      protein. The pair was reported only in HuRI, the CCSB human reference binary interactome map (PMID:32296183),
      and KRTAP10-8 has 415 distinct partners in IntAct - an extreme interactome hub.


      All fifteen GO:0005515 rows on this gene come from two systematic yeast two-hybrid interactome maps
      and nothing else. Querying IntAct directly for P82987 returns the same pairs logged under three
      method names - two hybrid array, two hybrid prey pooling approach and validated two hybrid - which
      are sub-methods of one Y2H pipeline rather than three independent assays, so the NbExp values in
      the UniProt INTERACTION block count Y2H variants and not orthogonal confirmation. Disaggregating
      IntAct by pair rather than by gene - which is the only way the question can be answered, because
      ADAMTSL3''s gene-level record also covers 17 partners that are not in GOA - shows that 0 of these
      13 pairs carries any non-Y2H detection method. The anti-tag co-immunoprecipitation and BioID methods
      present in ADAMTSL3''s IntAct record belong entirely to those other 17 partners, from separate publications
      that are not the source of any GOA row reviewed here.


      The partner set behaves like screen background rather than like biology. Nine of the thirteen partners
      are hair keratins or keratin-associated proteins and a tenth, CYSRT1, is a cornified-envelope protein
      embedded in the same interaction network; the remaining three are cytosolic or nuclear. The median
      number of distinct IntAct partners across the thirteen is 188 against 30 for ADAMTSL3 itself, and
      six of them exceed 180, so most of this set consists of high-degree promiscuous preys.


      There is also a compartment problem that the assay cannot see. ADAMTSL3 has a cleaved signal peptide
      and is secreted into the extracellular matrix; hair keratin-associated proteins are intracellular
      constituents of the hair cortex, GLRX3 is cytosolic and MDFI is nuclear and cytoplasmic. In a two-hybrid
      assay both proteins are expressed in the yeast nucleus, where a signal peptide is inert, so the
      pairing is never tested in a compartment where it could occur.


      Marked over-annotated rather than removed. These are real observations recorded correctly by IntAct
      curators from published screens, and a screen hit can turn out to matter; what it does not do is
      license a functional claim, and bare protein binding conveys none in any case.'
    action: MARK_AS_OVER_ANNOTATED
    reason: Unreplicated systematic yeast two-hybrid hit. KRTAP10-8 is a cysteine-rich hair-cortex keratin-associated
      protein and carries 415 distinct IntAct partners against ADAMTSL3's 30. Disaggregating IntAct per
      pair rather than per gene shows this pair, like all 13, carries no non-Y2H detection method; it
      has no orthogonal assay, no follow-up in the ADAMTSL3 literature, and ADAMTSL3 is a secreted matrix
      protein that does not share a compartment with it in vivo. Bare protein binding conveys no function
      even where the pairing is real. Seen in both HI-II-14 and HuRI, but those two maps share the CCSB
      ORFeome and search space, so the repeat is not independent replication - and KRTAP10-8 is one of
      the highest-degree preys in the set.
    supported_by:
    - reference_id: file:human/ADAMTSL3/ADAMTSL3-bioinformatics/RESULTS.md
      supporting_text: ADAMTSL3 itself has 30 distinct IntAct partners; the median for its 13 annotated
        partners is 188.
    - reference_id: file:human/ADAMTSL3/ADAMTSL3-bioinformatics/RESULTS.md
      supporting_text: '**GOA partners with any non-Y2H detection method: 0 of 13** - every one of these
        pairs rests on yeast two-hybrid sub-methods alone, with no orthogonal assay.'
    - reference_id: file:human/ADAMTSL3/ADAMTSL3-uniprot.txt
      supporting_text: FT   SIGNAL          1..26
- term:
    id: GO:0005515
    label: protein binding
  evidence_type: IPI
  original_reference_id: PMID:32296183
  qualifier: enables
  supporting_entities:
  - UniProtKB:Q07627
  review:
    summary: 'Over-annotated. KRTAP1-1 (Q07627, 177 aa) is a cysteine-rich hair-cortex keratin-associated
      protein. The pair was reported only in HuRI, the CCSB human reference binary interactome map (PMID:32296183),
      and KRTAP1-1 has 188 distinct partners in IntAct - an extreme interactome hub.


      All fifteen GO:0005515 rows on this gene come from two systematic yeast two-hybrid interactome maps
      and nothing else. Querying IntAct directly for P82987 returns the same pairs logged under three
      method names - two hybrid array, two hybrid prey pooling approach and validated two hybrid - which
      are sub-methods of one Y2H pipeline rather than three independent assays, so the NbExp values in
      the UniProt INTERACTION block count Y2H variants and not orthogonal confirmation. Disaggregating
      IntAct by pair rather than by gene - which is the only way the question can be answered, because
      ADAMTSL3''s gene-level record also covers 17 partners that are not in GOA - shows that 0 of these
      13 pairs carries any non-Y2H detection method. The anti-tag co-immunoprecipitation and BioID methods
      present in ADAMTSL3''s IntAct record belong entirely to those other 17 partners, from separate publications
      that are not the source of any GOA row reviewed here.


      The partner set behaves like screen background rather than like biology. Nine of the thirteen partners
      are hair keratins or keratin-associated proteins and a tenth, CYSRT1, is a cornified-envelope protein
      embedded in the same interaction network; the remaining three are cytosolic or nuclear. The median
      number of distinct IntAct partners across the thirteen is 188 against 30 for ADAMTSL3 itself, and
      six of them exceed 180, so most of this set consists of high-degree promiscuous preys.


      There is also a compartment problem that the assay cannot see. ADAMTSL3 has a cleaved signal peptide
      and is secreted into the extracellular matrix; hair keratin-associated proteins are intracellular
      constituents of the hair cortex, GLRX3 is cytosolic and MDFI is nuclear and cytoplasmic. In a two-hybrid
      assay both proteins are expressed in the yeast nucleus, where a signal peptide is inert, so the
      pairing is never tested in a compartment where it could occur.


      Marked over-annotated rather than removed. These are real observations recorded correctly by IntAct
      curators from published screens, and a screen hit can turn out to matter; what it does not do is
      license a functional claim, and bare protein binding conveys none in any case.'
    action: MARK_AS_OVER_ANNOTATED
    reason: Unreplicated systematic yeast two-hybrid hit. KRTAP1-1 is a cysteine-rich hair-cortex keratin-associated
      protein and carries 188 distinct IntAct partners against ADAMTSL3's 30. Disaggregating IntAct per
      pair rather than per gene shows this pair, like all 13, carries no non-Y2H detection method; it
      has no orthogonal assay, no follow-up in the ADAMTSL3 literature, and ADAMTSL3 is a secreted matrix
      protein that does not share a compartment with it in vivo. Bare protein binding conveys no function
      even where the pairing is real.
    supported_by:
    - reference_id: file:human/ADAMTSL3/ADAMTSL3-bioinformatics/RESULTS.md
      supporting_text: ADAMTSL3 itself has 30 distinct IntAct partners; the median for its 13 annotated
        partners is 188.
    - reference_id: file:human/ADAMTSL3/ADAMTSL3-bioinformatics/RESULTS.md
      supporting_text: '**GOA partners with any non-Y2H detection method: 0 of 13** - every one of these
        pairs rests on yeast two-hybrid sub-methods alone, with no orthogonal assay.'
    - reference_id: file:human/ADAMTSL3/ADAMTSL3-uniprot.txt
      supporting_text: FT   SIGNAL          1..26
- term:
    id: GO:0005515
    label: protein binding
  evidence_type: IPI
  original_reference_id: PMID:32296183
  qualifier: enables
  supporting_entities:
  - UniProtKB:Q6L8G8
  review:
    summary: 'Over-annotated. KRTAP5-7 (Q6L8G8, 165 aa) is a cysteine-rich hair-cortex keratin-associated
      protein. The pair was reported only in HuRI, the CCSB human reference binary interactome map (PMID:32296183),
      and KRTAP5-7 has 61 distinct partners in IntAct - a moderately connected prey.


      All fifteen GO:0005515 rows on this gene come from two systematic yeast two-hybrid interactome maps
      and nothing else. Querying IntAct directly for P82987 returns the same pairs logged under three
      method names - two hybrid array, two hybrid prey pooling approach and validated two hybrid - which
      are sub-methods of one Y2H pipeline rather than three independent assays, so the NbExp values in
      the UniProt INTERACTION block count Y2H variants and not orthogonal confirmation. Disaggregating
      IntAct by pair rather than by gene - which is the only way the question can be answered, because
      ADAMTSL3''s gene-level record also covers 17 partners that are not in GOA - shows that 0 of these
      13 pairs carries any non-Y2H detection method. The anti-tag co-immunoprecipitation and BioID methods
      present in ADAMTSL3''s IntAct record belong entirely to those other 17 partners, from separate publications
      that are not the source of any GOA row reviewed here.


      The partner set behaves like screen background rather than like biology. Nine of the thirteen partners
      are hair keratins or keratin-associated proteins and a tenth, CYSRT1, is a cornified-envelope protein
      embedded in the same interaction network; the remaining three are cytosolic or nuclear. The median
      number of distinct IntAct partners across the thirteen is 188 against 30 for ADAMTSL3 itself, and
      six of them exceed 180, so most of this set consists of high-degree promiscuous preys.


      There is also a compartment problem that the assay cannot see. ADAMTSL3 has a cleaved signal peptide
      and is secreted into the extracellular matrix; hair keratin-associated proteins are intracellular
      constituents of the hair cortex, GLRX3 is cytosolic and MDFI is nuclear and cytoplasmic. In a two-hybrid
      assay both proteins are expressed in the yeast nucleus, where a signal peptide is inert, so the
      pairing is never tested in a compartment where it could occur.


      Marked over-annotated rather than removed. These are real observations recorded correctly by IntAct
      curators from published screens, and a screen hit can turn out to matter; what it does not do is
      license a functional claim, and bare protein binding conveys none in any case.'
    action: MARK_AS_OVER_ANNOTATED
    reason: Unreplicated systematic yeast two-hybrid hit. KRTAP5-7 is a cysteine-rich hair-cortex keratin-associated
      protein and carries 61 distinct IntAct partners against ADAMTSL3's 30. Disaggregating IntAct per
      pair rather than per gene shows this pair, like all 13, carries no non-Y2H detection method; it
      has no orthogonal assay, no follow-up in the ADAMTSL3 literature, and ADAMTSL3 is a secreted matrix
      protein that does not share a compartment with it in vivo. Bare protein binding conveys no function
      even where the pairing is real.
    supported_by:
    - reference_id: file:human/ADAMTSL3/ADAMTSL3-bioinformatics/RESULTS.md
      supporting_text: ADAMTSL3 itself has 30 distinct IntAct partners; the median for its 13 annotated
        partners is 188.
    - reference_id: file:human/ADAMTSL3/ADAMTSL3-bioinformatics/RESULTS.md
      supporting_text: '**GOA partners with any non-Y2H detection method: 0 of 13** - every one of these
        pairs rests on yeast two-hybrid sub-methods alone, with no orthogonal assay.'
    - reference_id: file:human/ADAMTSL3/ADAMTSL3-uniprot.txt
      supporting_text: FT   SIGNAL          1..26
- term:
    id: GO:0005515
    label: protein binding
  evidence_type: IPI
  original_reference_id: PMID:32296183
  qualifier: enables
  supporting_entities:
  - UniProtKB:Q99750
  review:
    summary: 'Over-annotated. MDFI (Q99750, 246 aa) is a nuclear and cytoplasmic inhibitor of MyoD-family
      bHLH transcription factors. The pair was reported only in HuRI, the CCSB human reference binary
      interactome map (PMID:32296183), and MDFI has 483 distinct partners in IntAct - an extreme interactome
      hub.


      All fifteen GO:0005515 rows on this gene come from two systematic yeast two-hybrid interactome maps
      and nothing else. Querying IntAct directly for P82987 returns the same pairs logged under three
      method names - two hybrid array, two hybrid prey pooling approach and validated two hybrid - which
      are sub-methods of one Y2H pipeline rather than three independent assays, so the NbExp values in
      the UniProt INTERACTION block count Y2H variants and not orthogonal confirmation. Disaggregating
      IntAct by pair rather than by gene - which is the only way the question can be answered, because
      ADAMTSL3''s gene-level record also covers 17 partners that are not in GOA - shows that 0 of these
      13 pairs carries any non-Y2H detection method. The anti-tag co-immunoprecipitation and BioID methods
      present in ADAMTSL3''s IntAct record belong entirely to those other 17 partners, from separate publications
      that are not the source of any GOA row reviewed here.


      The partner set behaves like screen background rather than like biology. Nine of the thirteen partners
      are hair keratins or keratin-associated proteins and a tenth, CYSRT1, is a cornified-envelope protein
      embedded in the same interaction network; the remaining three are cytosolic or nuclear. The median
      number of distinct IntAct partners across the thirteen is 188 against 30 for ADAMTSL3 itself, and
      six of them exceed 180, so most of this set consists of high-degree promiscuous preys.


      There is also a compartment problem that the assay cannot see. ADAMTSL3 has a cleaved signal peptide
      and is secreted into the extracellular matrix; hair keratin-associated proteins are intracellular
      constituents of the hair cortex, GLRX3 is cytosolic and MDFI is nuclear and cytoplasmic. In a two-hybrid
      assay both proteins are expressed in the yeast nucleus, where a signal peptide is inert, so the
      pairing is never tested in a compartment where it could occur.


      Marked over-annotated rather than removed. These are real observations recorded correctly by IntAct
      curators from published screens, and a screen hit can turn out to matter; what it does not do is
      license a functional claim, and bare protein binding conveys none in any case.'
    action: MARK_AS_OVER_ANNOTATED
    reason: Unreplicated systematic yeast two-hybrid hit. MDFI is a nuclear and cytoplasmic inhibitor
      of MyoD-family bHLH transcription factors and carries 483 distinct IntAct partners against ADAMTSL3's
      30. Disaggregating IntAct per pair rather than per gene shows this pair, like all 13, carries no
      non-Y2H detection method; it has no orthogonal assay, no follow-up in the ADAMTSL3 literature, and
      ADAMTSL3 is a secreted matrix protein that does not share a compartment with it in vivo. Bare protein
      binding conveys no function even where the pairing is real. Seen in both HI-II-14 and HuRI, but
      those two maps share the CCSB ORFeome and search space, so the repeat is not independent replication
      - and MDFI is one of the highest-degree preys in the set.
    supported_by:
    - reference_id: file:human/ADAMTSL3/ADAMTSL3-bioinformatics/RESULTS.md
      supporting_text: ADAMTSL3 itself has 30 distinct IntAct partners; the median for its 13 annotated
        partners is 188.
    - reference_id: file:human/ADAMTSL3/ADAMTSL3-bioinformatics/RESULTS.md
      supporting_text: '**GOA partners with any non-Y2H detection method: 0 of 13** - every one of these
        pairs rests on yeast two-hybrid sub-methods alone, with no orthogonal assay.'
    - reference_id: file:human/ADAMTSL3/ADAMTSL3-uniprot.txt
      supporting_text: FT   SIGNAL          1..26
- term:
    id: GO:0005515
    label: protein binding
  evidence_type: IPI
  original_reference_id: PMID:32296183
  qualifier: enables
  supporting_entities:
  - UniProtKB:Q9BYR7
  review:
    summary: 'Over-annotated. KRTAP3-2 (Q9BYR7, 98 aa) is a cysteine-rich hair-cortex keratin-associated
      protein. The pair was reported only in HuRI, the CCSB human reference binary interactome map (PMID:32296183),
      and KRTAP3-2 has 84 distinct partners in IntAct - a moderately connected prey.


      All fifteen GO:0005515 rows on this gene come from two systematic yeast two-hybrid interactome maps
      and nothing else. Querying IntAct directly for P82987 returns the same pairs logged under three
      method names - two hybrid array, two hybrid prey pooling approach and validated two hybrid - which
      are sub-methods of one Y2H pipeline rather than three independent assays, so the NbExp values in
      the UniProt INTERACTION block count Y2H variants and not orthogonal confirmation. Disaggregating
      IntAct by pair rather than by gene - which is the only way the question can be answered, because
      ADAMTSL3''s gene-level record also covers 17 partners that are not in GOA - shows that 0 of these
      13 pairs carries any non-Y2H detection method. The anti-tag co-immunoprecipitation and BioID methods
      present in ADAMTSL3''s IntAct record belong entirely to those other 17 partners, from separate publications
      that are not the source of any GOA row reviewed here.


      The partner set behaves like screen background rather than like biology. Nine of the thirteen partners
      are hair keratins or keratin-associated proteins and a tenth, CYSRT1, is a cornified-envelope protein
      embedded in the same interaction network; the remaining three are cytosolic or nuclear. The median
      number of distinct IntAct partners across the thirteen is 188 against 30 for ADAMTSL3 itself, and
      six of them exceed 180, so most of this set consists of high-degree promiscuous preys.


      There is also a compartment problem that the assay cannot see. ADAMTSL3 has a cleaved signal peptide
      and is secreted into the extracellular matrix; hair keratin-associated proteins are intracellular
      constituents of the hair cortex, GLRX3 is cytosolic and MDFI is nuclear and cytoplasmic. In a two-hybrid
      assay both proteins are expressed in the yeast nucleus, where a signal peptide is inert, so the
      pairing is never tested in a compartment where it could occur.


      Marked over-annotated rather than removed. These are real observations recorded correctly by IntAct
      curators from published screens, and a screen hit can turn out to matter; what it does not do is
      license a functional claim, and bare protein binding conveys none in any case.'
    action: MARK_AS_OVER_ANNOTATED
    reason: Unreplicated systematic yeast two-hybrid hit. KRTAP3-2 is a cysteine-rich hair-cortex keratin-associated
      protein and carries 84 distinct IntAct partners against ADAMTSL3's 30. Disaggregating IntAct per
      pair rather than per gene shows this pair, like all 13, carries no non-Y2H detection method; it
      has no orthogonal assay, no follow-up in the ADAMTSL3 literature, and ADAMTSL3 is a secreted matrix
      protein that does not share a compartment with it in vivo. Bare protein binding conveys no function
      even where the pairing is real.
    supported_by:
    - reference_id: file:human/ADAMTSL3/ADAMTSL3-bioinformatics/RESULTS.md
      supporting_text: ADAMTSL3 itself has 30 distinct IntAct partners; the median for its 13 annotated
        partners is 188.
    - reference_id: file:human/ADAMTSL3/ADAMTSL3-bioinformatics/RESULTS.md
      supporting_text: '**GOA partners with any non-Y2H detection method: 0 of 13** - every one of these
        pairs rests on yeast two-hybrid sub-methods alone, with no orthogonal assay.'
    - reference_id: file:human/ADAMTSL3/ADAMTSL3-uniprot.txt
      supporting_text: FT   SIGNAL          1..26
- term:
    id: GO:0005515
    label: protein binding
  evidence_type: IPI
  original_reference_id: PMID:32296183
  qualifier: enables
  supporting_entities:
  - UniProtKB:Q9BYR9
  review:
    summary: 'Over-annotated. KRTAP2-4 (Q9BYR9, 128 aa) is a cysteine-rich hair-cortex keratin-associated
      protein. The pair was reported only in HuRI, the CCSB human reference binary interactome map (PMID:32296183),
      and KRTAP2-4 has 89 distinct partners in IntAct - a moderately connected prey.


      All fifteen GO:0005515 rows on this gene come from two systematic yeast two-hybrid interactome maps
      and nothing else. Querying IntAct directly for P82987 returns the same pairs logged under three
      method names - two hybrid array, two hybrid prey pooling approach and validated two hybrid - which
      are sub-methods of one Y2H pipeline rather than three independent assays, so the NbExp values in
      the UniProt INTERACTION block count Y2H variants and not orthogonal confirmation. Disaggregating
      IntAct by pair rather than by gene - which is the only way the question can be answered, because
      ADAMTSL3''s gene-level record also covers 17 partners that are not in GOA - shows that 0 of these
      13 pairs carries any non-Y2H detection method. The anti-tag co-immunoprecipitation and BioID methods
      present in ADAMTSL3''s IntAct record belong entirely to those other 17 partners, from separate publications
      that are not the source of any GOA row reviewed here.


      The partner set behaves like screen background rather than like biology. Nine of the thirteen partners
      are hair keratins or keratin-associated proteins and a tenth, CYSRT1, is a cornified-envelope protein
      embedded in the same interaction network; the remaining three are cytosolic or nuclear. The median
      number of distinct IntAct partners across the thirteen is 188 against 30 for ADAMTSL3 itself, and
      six of them exceed 180, so most of this set consists of high-degree promiscuous preys.


      There is also a compartment problem that the assay cannot see. ADAMTSL3 has a cleaved signal peptide
      and is secreted into the extracellular matrix; hair keratin-associated proteins are intracellular
      constituents of the hair cortex, GLRX3 is cytosolic and MDFI is nuclear and cytoplasmic. In a two-hybrid
      assay both proteins are expressed in the yeast nucleus, where a signal peptide is inert, so the
      pairing is never tested in a compartment where it could occur.


      Marked over-annotated rather than removed. These are real observations recorded correctly by IntAct
      curators from published screens, and a screen hit can turn out to matter; what it does not do is
      license a functional claim, and bare protein binding conveys none in any case.'
    action: MARK_AS_OVER_ANNOTATED
    reason: Unreplicated systematic yeast two-hybrid hit. KRTAP2-4 is a cysteine-rich hair-cortex keratin-associated
      protein and carries 89 distinct IntAct partners against ADAMTSL3's 30. Disaggregating IntAct per
      pair rather than per gene shows this pair, like all 13, carries no non-Y2H detection method; it
      has no orthogonal assay, no follow-up in the ADAMTSL3 literature, and ADAMTSL3 is a secreted matrix
      protein that does not share a compartment with it in vivo. Bare protein binding conveys no function
      even where the pairing is real.
    supported_by:
    - reference_id: file:human/ADAMTSL3/ADAMTSL3-bioinformatics/RESULTS.md
      supporting_text: ADAMTSL3 itself has 30 distinct IntAct partners; the median for its 13 annotated
        partners is 188.
    - reference_id: file:human/ADAMTSL3/ADAMTSL3-bioinformatics/RESULTS.md
      supporting_text: '**GOA partners with any non-Y2H detection method: 0 of 13** - every one of these
        pairs rests on yeast two-hybrid sub-methods alone, with no orthogonal assay.'
    - reference_id: file:human/ADAMTSL3/ADAMTSL3-uniprot.txt
      supporting_text: FT   SIGNAL          1..26
- term:
    id: GO:0071953
    label: elastic fiber
  evidence_type: TAS
  original_reference_id: PMID:23962539
  qualifier: located_in
  review:
    summary: 'Kept, but as a non-core location, because no one has ever localised ADAMTSL3 protein to
      an elastic fibre.


      The source is a specialist review of elastic-fibre biology, not primary data, and the cached record
      is abstract-only. Its abstract does not mention ADAMTSL3, so the supporting statement is in text
      I cannot read.


      I therefore tested whether this is a bulk projection. Querying QuickGO by reference returns 66 annotations
      over 62 distinct entities, all TAS and all from GO_Central, split across four terms - so it is indeed
      a reference-driven curation of an elastic-fibre proteome. But it is not an indiscriminate family
      sweep, and that is what decides the action: the GO:0071953 set takes ADAMTSL3, ADAMTSL4, ADAMTSL5
      and THSD4 with their mouse orthologues while leaving out ADAMTSL1, ADAMTSL2 and PAPLN. A curator
      applying the term by family membership would have taken all seven. The selection therefore tracks
      the review''s content, which makes this a genuine traceable author statement rather than an artefact.


      Against that, the biochemistry stops short of a location. ADAMTSL3 binds the N-terminal half of
      fibrillin-1 and the C-terminal domains of LTBP-1, which is why the association is plausible - but
      the paper that established the binding says in as many words that the colocalisation experiment
      could not be done for want of antibodies, and no later work has done it. The gap is therefore not
      an oversight in this review; it is the state of the field.


      Kept as non-core rather than removed: overruling a specialist curator''s traceable statement on
      the strength of a full text I have not read would be exactly the wrong call, and the term is consistent
      with everything that is known. Not accepted as core because a location for which no localisation
      experiment exists should not be presented as established.'
    action: KEEP_AS_NON_CORE
    reason: A traceable author statement from a specialist elastic-fibre review, applied selectively within
      the ADAMTSL family rather than by family membership, and consistent with ADAMTSL3's demonstrated
      binding to fibrillin-1 and LTBP-1. But no published experiment localises ADAMTSL3 protein to an
      elastic fibre or a microfibril - the study that established the fibrillin-1 binding states that
      the colocalisation could not be attempted because antibodies did not exist - so it is retained without
      being treated as an established core location.
    additional_reference_ids:
    - PMID:22242013
    supported_by:
    - reference_id: PMID:23962539
      supporting_text: They are laid down during development, and comprise a cross-linked elastin core
        within a template of fibrillin-based microfibrils.
      full_text_unavailable: true
    - reference_id: PMID:22242013
      supporting_text: Since antibodies specific for ADAMTSL-2 and -3 are not yet available, we were unable
        to determine whether these proteins also colocalize with fibrillin-1 in skin and whether these
        are also reduced in WMΔ mutant mice.
    knowledge_gaps:
    - gap_statement: ADAMTSL3 has never been localised in tissue. There is no published immunolocalisation
        of the endogenous protein to microfibrils, elastic fibres, or any other defined matrix structure.
      boundary: The binding side is solid - SPR against fibrillin-1 and LTBP-1 with mapped sites and negative
        controls - and the protein is unambiguously secreted into matrix in transfected cells.
      gap_kind:
      - BIOLOGY
      dark_aspect: CC_DARK
      status: OPEN
      significance: Whether ADAMTSL3 is a microfibril component or a soluble matrix protein that binds
        fibrillin-1 transiently changes what it can be doing to latent TGF-beta, and the elastic-fibre
        annotation currently rests on the more committal of those two readings.
      resolution: Raise ADAMTSL3-specific antibodies and perform immunofluorescence and immunoelectron
        microscopy on skin, aorta and myocardium, with fibrillin-1 co-staining and an Adamtsl3-null control
        for specificity.
      provenance:
      - reference_id: PMID:22242013
        supporting_text: Since antibodies specific for ADAMTSL-2 and -3 are not yet available, we were
          unable to determine whether these proteins also colocalize with fibrillin-1 in skin
- term:
    id: GO:0050840
    label: extracellular matrix binding
  evidence_type: ISS
  original_reference_id: PMID:22242013
  qualifier: enables
  supporting_entities:
  - UniProtKB:G3UXC7
  review:
    summary: 'Proposed. This is the gene''s best-characterised molecular activity and it is entirely absent
      from GOA, which currently gives ADAMTSL3 no molecular function beyond bare protein binding to a
      set of two-hybrid preys.


      Surface plasmon resonance shows that the C-terminal end of ADAMTSL-3 binds the N-terminal half of
      fibrillin-1, that the Weill-Marchesani three-domain deletion in fibrillin-1 abolishes that binding,
      that the same ADAMTSL-3 fragment binds the non-catalytic C-terminal region of ADAMTS-10 with a KD
      of 2 nM, and that it binds the C-terminal domains of LTBP-1. Two negative controls in the same panel
      show the binding is selective rather than sticky: ADAMTSL-1 does not bind fibrillin-1 or ADAMTS-10,
      and ADAMTSL-3 does not bind LTBP-4.


      Fibrillin-1 and LTBP-1 are both extracellular matrix components, so GO:0050840 states what was measured.
      The same term was chosen independently for the same biology in this repository''s ADAMTSL4 review,
      which keeps the two paralogues consistent.


      Deliberately not the child term GO:0050436 microfibril binding, which is defined as binding to a
      microfibril - a supramolecular assembly. What was assayed was binding to soluble recombinant fibrillin-1
      polypeptides, which is binding to a microfibril component, not to a microfibril; and the two other
      measured partners, LTBP-1 and ADAMTS-10, are not microfibrils at all, so the child term would cover
      only part of the evidence while asserting more than any of it. Whether ADAMTSL3 engages assembled
      microfibrils is precisely the untested question recorded against the elastic-fiber row.


      Coded ISS rather than IPI because the species of the ADAMTSL-3 reagent is not resolvable from the
      paper. The Results describe the panel as recombinant human ADAMTSL-1, -2 and -3, but the Methods
      say the ADAMTSL3 constructs were made from a RIKEN clone and mouse lung cDNA - and say the same,
      explicitly, for ADAMTSL2, which the Results sentence also calls human. Only ADAMTSL1 is unambiguously
      human in the Methods. A curator should upgrade this to IPI, with UniProtKB:P35555 (human FBN1) as
      the WITH/FROM interactor, if the authors confirm the construct was human; until then the conservative
      code is the honest one. Note the WITH/FROM here carries only UniProtKB:G3UXC7, the mouse orthologue:
      for ISS the field must hold the sequence-similar entity from which the inference is drawn, and the
      binding partner belongs there only under IPI.'
    action: NEW
    reason: ADAMTSL3 binds fibrillin-1 and LTBP-1, both extracellular matrix proteins, by surface plasmon
      resonance with mapped sites and selective negative controls, yet GOA records no molecular function
      for this gene other than bare protein binding to two-hybrid preys. ISS rather than IPI because the
      paper contradicts itself about whether the ADAMTSL-3 polypeptide was human or mouse.
    supported_by:
    - reference_id: PMID:22242013
      supporting_text: Similar to ADAMTSL-6 [24], ADAMTSL-2, -3, and papilin polypeptides interacted with
        the N-terminal half of fibrillin-1, while ADAMTSL-1 did not.
    - reference_id: PMID:22242013
      supporting_text: SPR also showed that the C-terminal end of ADAMTS-10 interacted with the C-terminal
        end of ADAMTSL-3 with high binding affinity (KD = 2 nM) (Figure 5c).
    - reference_id: PMID:22242013
      supporting_text: Binding between ADAMTSL-3 and the C-terminal domains of LTBP-1 was also detected,
        but neither ADAMTSL-2 nor -3 interacted with LTBP-4.
    - reference_id: PMID:22242013
      supporting_text: Constructs for ADAMTSL3 were made using a clone (RIKEN) and mouse lung cDNA.
- term:
    id: GO:0030512
    label: negative regulation of transforming growth factor beta receptor signaling pathway
  evidence_type: IMP
  original_reference_id: PMID:36539599
  qualifier: involved_in
  review:
    summary: 'Proposed. Loss-of-function and gain-of-function agree in sign, and the gain-of-function
      half was done on human cells with the human protein.


      Adenoviral overexpression of full-length ADAMTSL3 in cultured human foetal cardiac fibroblasts reduced
      phospho-SMAD2 and active TGF-beta, and reduced the large latent complex components LAP and LTBP1
      in both cell and matrix lysates; a 84-gene TGF-beta pathway array showed TGFB1, TGFB2, SMAD3, SMAD4
      and TGFBR1 down and the inhibitory SMAD6 up. In the reciprocal direction, Adamtsl3-null mouse hearts
      show raised active TGF-beta and phospho-SMAD2 after aortic banding.


      Coded IMP because adenoviral overexpression is a perturbation, not an assay of the protein at native
      levels.


      The mechanism is consistent with the binding data: LTBP-1 and fibrillin-1 are the two proteins that
      hold the large latent TGF-beta complex in the matrix, and ADAMTSL3 binds both. What is not yet shown
      is whether the reduction in signalling follows from sequestering latent complex or from the transcriptional
      changes, so the term is proposed at the level of the pathway rather than as a specific molecular
      mechanism.


      The cardiac phenotypes themselves are deliberately not proposed as GO terms. They are mouse, and
      they are responses to experimental pressure overload rather than normal physiology; the reviewable
      claim is the regulation of the signalling pathway, which has human-cell support.'
    action: NEW
    reason: Human ADAMTSL3 overexpressed in human cardiac fibroblasts reduces phospho-SMAD2, active TGF-beta
      and large-latent-complex components, and Adamtsl3-null mouse hearts show the reciprocal increase
      after pressure overload. GOA carries no biological-process annotation for this gene beyond a family-level
      InterPro term, and this is the best-supported process claim in its literature.
    supported_by:
    - reference_id: PMID:36539599
      supporting_text: For further mechanistic insights, we overexpressed full-length ADAMTSL3 (L3) and
        a vehicle control adenovirus (veh) in cultures of human foetal CFBs (hfCFBs)
    - reference_id: PMID:36539599
      supporting_text: Immunoblotting revealed reduced pSMAD (Fig. 4h) and reduced active TGFβ (Fig. 4i),
        indicating reduced TGFβ signalling.
    - reference_id: PMID:36539599
      supporting_text: Taken together, the loss-of-function and gain-of-function studies suggest ADAMTSL3
        as an inhibitor of TGFβ signalling in the heart in vivo and in cardiac fibroblasts in vitro.
- term:
    id: GO:0050808
    label: synapse organization
  evidence_type: ISS
  original_reference_id: PMID:37572323
  qualifier: involved_in
  supporting_entities:
  - UniProtKB:G3UXC7
  review:
    summary: 'Proposed, on mouse conditional knockouts, and coded ISS for that reason.


      Early post-natal neuronal deletion of Adamtsl3 in mouse reduces DCC protein and the density of both
      glutamatergic and GABAergic synapses; adult deletion in GABAergic or glutamatergic neurons leaves
      DCC-Netrin-1 function at glutamatergic synapses intact but controls DCC signalling at GABAergic
      synapses, and the Adamtsl3-DCC unit is required for activity-dependent adaptation there. A second
      mouse study finds Adamtsl3 is required cell-autonomously in parvalbumin interneurons for perineuronal
      net formation and maintenance, with adult deletion reactivating juvenile ocular-dominance plasticity.


      The orthology behind the inference is unusually good for a mouse-to-human transfer. The C. elegans
      orthologue Ce-Punctin/MADD-4 is itself a secreted synaptic organiser that specifies whether a postsynaptic
      domain is cholinergic or GABAergic, and it acts through UNC-40/DCC - the same receptor as in the
      mouse hippocampus. So the receptor and the synaptic role are conserved from nematode to mouse, which
      is a stronger basis than sequence similarity alone. Human ADAMTSL3 protein is present in cortical
      and ganglionic neurons.


      One caveat is recorded rather than glossed: MADD-4 is the orthologue of both ADAMTSL1 and ADAMTSL3,
      so the worm data are shared between the two human paralogues and cannot be assigned to ADAMTSL3
      alone. The mouse conditional knockouts can, and they are what the annotation rests on; the worm
      work is cited as evidence that the DCC-dependent synaptic role is ancestral, not as evidence about
      the human gene.


      Held as non-core, and deliberately not carried into core_functions: the direct evidence is entirely
      non-human and the molecular activity underlying the effect on DCC is unknown, which is recorded
      below as an MF_DARK knowledge gap. An earlier draft of this review listed the neuronal role among
      the core functions while this row called it non-core; the two passages rested on the same premise
      and reached opposite conclusions, so the core_functions entry has been removed and this row is the
      single statement of the position. The biology is not lost: it is described in the gene description,
      set out in full here, and its matrix component is carried by GO:0030198 in core function 1.'
    action: NEW
    reason: Mouse conditional knockouts show Adamtsl3 is required for hippocampal synapse density and
      for DCC signalling at GABAergic synapses, and a second mouse study shows a cell-autonomous requirement
      in parvalbumin interneurons for perineuronal nets. The DCC-dependent synaptic role is conserved
      to the C. elegans orthologue, and human ADAMTSL3 is expressed in cortical and ganglionic neurons.
      Coded ISS because no human experiment exists, and proposed as non-core for the same reason.
    additional_reference_ids:
    - PMID:42277231
    - PMID:24896188
    - PMID:22014523
    - PMID:17597111
    knowledge_gaps:
    - gap_statement: No molecular activity is known for ADAMTSL3 in the nervous system. How a secreted
        matrix glycoprotein raises DCC protein levels, and whether it does so by binding DCC, by binding
        something that presents DCC, or by altering the matrix in which DCC sits, is unestablished.
      boundary: The genetic requirement is firm - conditional deletion in mouse reduces DCC protein and
        synapse density, and the effect is cell-autonomous in parvalbumin interneurons in an independent
        study.
      gap_kind:
      - BIOLOGY
      dark_aspect: MF_DARK
      status: OPEN
      significance: Without a molecular activity the neural role cannot be given a molecular-function
        term, and it is the reason this row is proposed as a process annotation only.
      resolution: Test direct ADAMTSL3-DCC binding by SPR with purified ectodomains, and ask whether the
        fibrillin-1-binding C-terminal region is sufficient for the synaptic effect in a rescue experiment.
      provenance:
      - reference_id: file:human/ADAMTSL3/ADAMTSL3-deep-research-affinage.md
        supporting_text: The biochemical mechanism by which this non-catalytic ECM protein controls DCC
          receptor levels has not been characterized in the available corpus.
    supported_by:
    - reference_id: PMID:37572323
      supporting_text: We demonstrate that early post-natal deletion of Adamtsl3 in neurons impairs DCC
        protein expression, causing reduced density of both glutamatergic and GABAergic synapses.
      full_text_unavailable: true
    - reference_id: PMID:22014523
      supporting_text: MADD-4's activity is dependent on UNC-40/DCC, a netrin receptor, which functions
        cell-autonomously to direct membrane extension.
      full_text_unavailable: true
    - reference_id: PMID:17597111
      supporting_text: ADAMTSL3 is expressed in epithelial cells of the colon, fallopian tube, skin, breast,
        prostate, epididymis, liver, pancreatic islets and bile ducts, as well as by vascular endothelial
        cells, smooth muscle cells, fibroblasts, cortical and ganglionic neurons and cardiac myocytes.
      full_text_unavailable: true
    - reference_id: file:human/ADAMTSL3/ADAMTSL3-deep-research-affinage.md
      supporting_text: The biochemical mechanism by which this non-catalytic ECM protein controls DCC
        receptor levels has not been characterized in the available corpus.
- term:
    id: GO:0002020
    label: protease binding
  evidence_type: ISS
  original_reference_id: PMID:22242013
  qualifier: enables
  supporting_entities:
  - UniProtKB:G3UXC7
  review:
    summary: 'Proposed. The tightest measured interaction this protein has is with a protease, and no
      term currently records it.


      Surface plasmon resonance gives a KD of 2 nM between the C-terminal end of ADAMTSL-3 and the C-terminal
      end of ADAMTS-10, a secreted metalloproteinase mutated in recessive Weill-Marchesani syndrome. GO:0002020
      is defined as binding to a protease or a peptidase, so the term states exactly that. It is selective
      within the family in the same experiment: neither ADAMTSL-1 nor ADAMTSL-2 bound ADAMTS-10, which
      the authors read as ADAMTS enzymes partnering with specific ADAMTSL proteins.


      Worth being clear about what this does and does not claim. The region of ADAMTS-10 involved is its
      non-catalytic C-terminal Tsp1-repeat region, not the catalytic domain, so this is not evidence of
      protease inhibition, and no inhibitor term is proposed. Whether ADAMTSL3 modulates ADAMTS-10 activity
      or merely co-occupies fibrillin-1 with it is untested, and is one of the suggested experiments.


      Coded ISS for the same reason as the GO:0050840 row: the paper''s Results call the ADAMTSL-3 reagent
      human while its Methods describe a RIKEN clone and mouse lung cDNA. On an IPI upgrade the WITH/FROM
      interactor would be UniProtKB:Q9H324 (human ADAMTS10). This row is part of the same matricellular
      adaptor function as GO:0050840 rather than a separate one, which is why core function 1 describes
      the ADAMTS-10 binding but carries GO:0050840 as its single molecular_function term.


      Family precedent: this repository''s merged ADAMTSL4 review also carries GO:0002020, so the two
      paralogues stay consistent.'
    action: NEW
    reason: 'ADAMTSL3 binds the non-catalytic C-terminal region of the metalloproteinase ADAMTS10 with
      a KD of 2 nM by SPR, selectively within the family, and no GO term currently records the tightest
      measured interaction this protein has. ISS rather than IPI because the source paper contradicts
      itself about whether the ADAMTSL-3 polypeptide was human or mouse. No inhibitor term is proposed:
      the contact is with the Tsp1 region, not the catalytic domain.'
    supported_by:
    - reference_id: PMID:22242013
      supporting_text: SPR also showed that the C-terminal end of ADAMTS-10 interacted with the C-terminal
        end of ADAMTSL-3 with high binding affinity (KD = 2 nM) (Figure 5c).
    - reference_id: PMID:22242013
      supporting_text: However, neither ADAMTSL-2 nor -1 bound to ADAMTS-10, indicating that ADAMTS enzymes
        may partner only with specific ADAMTSL proteins.
    - reference_id: PMID:22242013
      supporting_text: The C-terminal recombinant ADAMTS-10 polypeptide used in the SPR studies represents
        the noncatalytic region of ADAMTS-10, a region composed primarily of Tsp1 repeats (Figure S3).
core_functions:
- description: Secreted matricellular adaptor of the fibrillin-1 microfibril network. The C-terminal region
    of ADAMTSL3 binds the N-terminal, assembly-competent half of fibrillin-1 at the site removed by the
    Weill-Marchesani three-domain deletion, binds the non-catalytic C-terminal region of ADAMTS10 with
    a KD of 2 nM, and binds the C-terminal domains of LTBP-1; the same panel shows it does not bind LTBP-4,
    and that ADAMTSL1 binds neither fibrillin-1 nor ADAMTS10, so the interactions are selective. Because
    fibrillin-1 and LTBP-1 are the proteins that tether the large latent TGF-beta complex to the matrix,
    this places ADAMTSL3 at the point where microfibril architecture and TGF-beta availability meet.
  molecular_function:
    id: GO:0050840
    label: extracellular matrix binding
  directly_involved_in:
  - id: GO:0030198
    label: extracellular matrix organization
  - id: GO:0030512
    label: negative regulation of transforming growth factor beta receptor signaling pathway
  locations:
  - id: GO:0031012
    label: extracellular matrix
  supported_by:
  - reference_id: PMID:22242013
    supporting_text: Similar to ADAMTSL-6 [24], ADAMTSL-2, -3, and papilin polypeptides interacted with
      the N-terminal half of fibrillin-1, while ADAMTSL-1 did not.
  - reference_id: PMID:22242013
    supporting_text: SPR also showed that the C-terminal end of ADAMTS-10 interacted with the C-terminal
      end of ADAMTSL-3 with high binding affinity (KD = 2 nM) (Figure 5c).
  - reference_id: PMID:14667842
    supporting_text: Using these and a monoclonal antibody to a C-terminal myc tag, we show that in transfected
      COS-7 cells, punctin-2 is expressed as a 210-kDa glycoprotein that is located in the extracellular
      matrix.
    full_text_unavailable: true
- description: Negative regulator of TGF-beta signalling in the extracellular matrix. Overexpression of
    human ADAMTSL3 in human foetal cardiac fibroblasts lowers phospho-SMAD2 and active TGF-beta and depletes
    the large latent complex components LAP and LTBP1 from both cell and matrix fractions, with coordinate
    downregulation of TGFB1, TGFB2, SMAD3, SMAD4 and TGFBR1 and upregulation of the inhibitory SMAD6;
    Adamtsl3-null mouse hearts show the reciprocal rise in active TGF-beta and phospho-SMAD2 after pressure
    overload, with worse dilatation and increased mortality. Whether the restraint operates by sequestering
    latent complex through the LTBP-1 and fibrillin-1 interactions, or through the transcriptional changes,
    is not resolved.
  directly_involved_in:
  - id: GO:0030512
    label: negative regulation of transforming growth factor beta receptor signaling pathway
  locations:
  - id: GO:0031012
    label: extracellular matrix
  supported_by:
  - reference_id: PMID:36539599
    supporting_text: Immunoblotting revealed reduced pSMAD (Fig. 4h) and reduced active TGFβ (Fig. 4i),
      indicating reduced TGFβ signalling.
  - reference_id: PMID:36539599
    supporting_text: Taken together, the loss-of-function and gain-of-function studies suggest ADAMTSL3
      as an inhibitor of TGFβ signalling in the heart in vivo and in cardiac fibroblasts in vitro.
suggested_questions:
- question: 'PAINT has already made the key-residue judgement that the ADAMTS catalytic domain is lost
    in this family, but has attached it to the wrong node. In PTHR13723, the root node PTN000347317 carries
    GO:0004222 metalloendopeptidase activity and GO:0006508 proteolysis as IBD, seeded exclusively by
    genuine ADAMTS proteinases; node PTN002673039 carries NOT GO:0004222 by IKR and NOT GO:0006508 by
    IRD. Streaming the PANTHER leaf GAF shows PTN002673039 projects onto exactly 22 leaves, every one
    an ADAMTSL2 orthologue. The identical loss of the metalloprotease and disintegrin domains is shared
    by ADAMTSL1, ADAMTSL3, ADAMTSL4, ADAMTSL5, PAPLN, THSD4 and the invertebrate members madd-4, loh and
    papilin, none of which carries the NOT. Could the IKR be moved or duplicated to the ancestral node
    of the non-catalytic ADAMTSL clade? That single edit would state the catalytic loss for six further
    human genes. I have not proposed a gene-level NOT GO:0004222 row for ADAMTSL3 alone, because the defect
    is clade-wide and fixing one leaf would leave the other five untouched while duplicating machinery
    PAINT already has in the right form - but a leaf-level NOT, inferred from the absent metalloprotease
    and disintegrin domains, would be the correct fallback if the node cannot be re-annotated. Related
    observation for the same curators: the NOT GO:0006508 IRD row at PTN002673039 produces no leaf annotation
    at all, not even on ADAMTSL2.'
  experts:
  - GO Central / PAINT curators
  - PANTHER
- question: Within PTHR13723, why does GO:0030198 extracellular matrix organization reach ADAMTSL2, ADAMTSL4
    and THSD4 from node PTN000347317 but not ADAMTSL3 or ADAMTSL5, and why do human ADAMTSL1 and PAPLN
    receive no PAINT annotation at all when their mouse and fly orthologues do? The split does not track
    any biological criterion I can find - ADAMTSL3 has direct fibrillin-1 binding data and a matrix phenotype
    in two independent mouse knockouts, which is more than several of the members that do receive the
    term.
  experts:
  - GO Central / PAINT curators
- question: PMID:22242013 contradicts itself about the species of the ADAMTSL-3 reagent. The Results describe
    recombinant human ADAMTSL-1, -2 and -3, while the Methods state that the ADAMTSL3 constructs were
    made from a RIKEN clone and mouse lung cDNA, and say the same explicitly for ADAMTSL2. Which is correct?
    The answer decides whether the fibrillin-1, ADAMTS-10 and LTBP-1 interactions should be annotated
    to human ADAMTSL3 as IPI or as ISS from the mouse orthologue.
  experts:
  - Authors of PMID:22242013 (Sakai / Apte laboratories)
- question: UniProt records ADAMTSL3's extracellular matrix localisation with ECO:0000269|PubMed:14667842,
    but GOA has no corresponding IDA row, so the only localisation annotation the gene carries is a phylogenetic
    inference. Can the UniProt experimental localisation be propagated into GOA?
  experts:
  - UniProt-GOA
- question: ADAMTSL3 binds fibrillin-1 and LTBP-1 and restrains TGF-beta signalling, and the ADAMTSL2
    and ADAMTS10 mutations that cause geleophysic dysplasia and Weill-Marchesani syndrome act on the same
    microfibril axis - yet no Mendelian disorder has been attributed to ADAMTSL3, only common-variant
    associations with height and lean mass. Is that because loss of ADAMTSL3 is tolerated, or because
    the phenotype has not been looked for? Are there biallelic loss-of-function carriers in large sequencing
    cohorts, and what are their stature and cardiac phenotypes?
  experts:
  - Human genetics of the fibrillinopathies
  - gnomAD / large cohort sequencing
suggested_experiments:
- hypothesis: ADAMTSL3 is a component of fibrillin microfibrils rather than a soluble matrix protein that
    binds fibrillin-1 transiently.
  description: Raise and validate ADAMTSL3-specific antibodies against a region absent from the other
    ADAMTSLs, using Adamtsl3-null tissue as the specificity control - the reagent gap has blocked this
    question since 2012. Then perform immunofluorescence and immunoelectron microscopy on skin, aorta
    and myocardium with fibrillin-1 co-staining, and test whether ADAMTSL3 co-fractionates with microfibrils
    extracted by bacterial collagenase digestion.
  experiment_type: immunolocalisation and biochemical fractionation
- hypothesis: ADAMTSL3 restrains TGF-beta signalling by retaining the large latent complex on microfibrils,
    not by altering transcription of the pathway.
  description: Separate the two mechanisms in the human cardiac fibroblast system. Compare full-length
    ADAMTSL3 with a C-terminal fragment carrying the fibrillin-1 and LTBP-1 binding activity but nothing
    else, and with a construct in which those binding surfaces are disrupted; read out phospho-SMAD2 and
    active TGF-beta against matrix-bound versus soluble LTBP1 and LAP. If sequestration is the mechanism,
    the binding-competent fragment should suffice and the effect should require an assembled fibrillin
    scaffold - testable by repeating in FBN1-null fibroblasts.
  experiment_type: structure-function and epistasis in human cells
- hypothesis: ADAMTSL3 and ADAMTS10 act as a complex on fibrillin-1, as the 2 nM binding affinity suggests.
  description: 'Test whether ADAMTSL3 modulates ADAMTS10''s activity or its microfibril association: reconstitute
    fibrillin-1 microfibril assembly in vitro and in cell culture with and without ADAMTSL3, and ask whether
    ADAMTSL3 alters ADAMTS10 binding to fibrillin-1, its zymogen processing, or its proteolytic activity.
    A non-catalytic family member that binds a proteinase of the same family at nanomolar affinity is
    a candidate regulator of that proteinase, which no GO term currently records.'
  experiment_type: in vitro reconstitution
- hypothesis: The mouse Adamtsl3 neuronal phenotypes reflect a function conserved in human ADAMTSL3.
  description: 'Do the experiment that would license an experimental rather than an inferred annotation
    for the human gene: delete or knock down ADAMTSL3 in human iPSC-derived cortical neuron and parvalbumin-interneuron
    cultures and assay DCC protein levels, synapse density and perineuronal net formation, with rescue
    by wild-type human ADAMTSL3. Include the reported schizophrenia-associated variant rs950169 in the
    rescue panel.'
  experiment_type: human iPSC neuronal model
- hypothesis: The height and lean-mass association at the ADAMTSL3 locus acts through TGF-beta availability
    in the growth plate.
  description: Measure growth-plate architecture, chondrocyte proliferation and phospho-SMAD2 in Adamtsl3-null
    mice against littermate controls, and test whether long-bone growth retardation is rescued by TGF-beta
    pathway inhibition. The Weill-Marchesani fibrillin-1 mouse, which loses the ADAMTSL binding site,
    already shows retardation of long bone growth, so the axis is testable genetically.
  experiment_type: mouse skeletal phenotyping