ADAMTSL5

UniProt ID: Q6ZMM2
Organism: Homo sapiens
Review Status: COMPLETE
📝 Provide Detailed Feedback

Gene Description

ADAMTSL5 (ADAMTS-like protein 5; also known as THSD6) is a small secreted, N-glycosylated extracellular matrix glycoprotein of the ADAMTS superfamily. At 481 amino acids it is by far the smallest family member, and it is built from a signal peptide, a single thrombospondin type-1 (TSP1) repeat, a cysteine-rich module, a spacer module, and a C-terminal netrin-like (NTR) module joined to the spacer by a proline-rich disordered segment. Unlike the ADAMTS metalloproteinases it resembles, ADAMTSL5 has no metalloprotease or disintegrin-like domain and no proteolytic activity; it is a matrix-binding glycoprotein rather than an enzyme. The secreted protein binds heparin through its NTR module in a salt-sensitive, ionic interaction, and binds both fibrillin-1 and fibrillin-2, co-localising with fibrillin microfibrils in the extracellular matrix deposited by cultured fibroblasts; it is the first family member shown to bind both fibrillins. Its distribution is peri-cellular and baso-lateral, and part of the pool is proteolytically processed to release the C-terminal NTR-containing fragment. In mouse development the protein is broadly expressed, most prominently in musculoskeletal tissues including skeletal muscle, cartilage and bone, and in many epithelia. Whether ADAMTSL5 modulates microfibril biology is unresolved: adding it to fibroblast cultures changed neither microfibril density nor assembly. Beyond the matrix, a nine-residue peptide from its TSP1 repeat is presented by HLA-C*06:02 on melanocytes and acts as a CD8+ T cell autoantigen in psoriasis, and the gene is epigenetically activated in hepatocellular carcinoma, where it supports tumorigenicity in MET-sensitised hepatocytes.

Existing Annotations Review

GO Term Evidence Action Reason
GO:0031012 extracellular matrix
IBA
GO_REF:0000033 		ACCEPT
Summary: Phylogenetic propagation from PANTHER node PTN000347317, and an unusually well-founded one: all 16 protein donors in the WITH/FROM field carry their own experimental (IDA/HDA) annotation to this term or a descendant.
Reason: Every WITH/FROM token was resolved and each donor's own evidence for the propagated term was queried. The GOA WITH/FROM field matches the cached PAINT seed list for this node exactly (17 tokens = 16 protein donors + the node id). All 16 protein donors hold their own IDA or HDA to GO:0031012 or a descendant, so any claim that the sources carry only family-level inference is contradicted by measurement. The target itself also holds an independent IDA to this term from PMID:23010571, and ADAMTSL5 is a signal-peptide secreted protein demonstrated to reside in fibroblast extracellular matrix. No specificity upgrade is warranted: the donor set is heterogeneous, spanning four distinct locations (extracellular matrix 14, basement membrane 4, interstitial matrix 1, microfibril 1), so GO:0031012 is the correct least common ancestor and refining it would mean arbitrarily preferring one donor's compartment. Separately checked and negative - the propagation does not land above its donors (14 of 16 hold this exact term), so no downward MODIFY applies.
Propagation Review
Root cause: NO FAILURE CORE
Sources checked:
FB:FBgn0003137 · Ppn (D. melanogaster) SUPPORTS TRANSFER
Swiss-Prot Q868Z9; carries its own IDA to GO:0005604.
FB:FBgn0032252 · loh (D. melanogaster) SUPPORTS TRANSFER
TrEMBL Q9VKV3; carries its own IDA to GO:0031012.
MGI:MGI:109249 · Adamts1 (M. musculus) SUPPORTS TRANSFER
Swiss-Prot P97857; carries its own IDA to GO:0005604, GO:0031012.
MGI:MGI:1347346 · Adamts7 (M. musculus) SUPPORTS TRANSFER
Swiss-Prot Q68SA9; carries its own IDA to GO:0031012.
MGI:MGI:1347356 · Adamts2 (M. musculus) SUPPORTS TRANSFER
Swiss-Prot Q8C9W3; carries its own IDA to GO:0031012.
MGI:MGI:1913798 · Adamtsl5 (M. musculus) SUPPORTS TRANSFER
TrEMBL D3Z689; carries its own HDA/IDA to GO:0031012.
MGI:MGI:1925044 · Adamtsl2 (M. musculus) SUPPORTS TRANSFER
Swiss-Prot Q7TSK7; carries its own IDA to GO:0031012.
MGI:MGI:2386139 · Papln (M. musculus) SUPPORTS TRANSFER
Swiss-Prot Q9EPX2; carries its own HDA/IDA to GO:0005604, GO:0031012.
MGI:MGI:2389008 · Adamtsl4 (M. musculus) SUPPORTS TRANSFER
Swiss-Prot Q80T21; carries its own IDA to GO:0005614, GO:0031012.
MGI:MGI:2672033 · Thsd4 (M. musculus) SUPPORTS TRANSFER
Swiss-Prot Q3UTY6; carries its own IDA to GO:0001527, GO:0031012.
PANTHER:PTN000347317 · PANTHER tree node (ADAMTS/ADAMTS-like family) SUPPORTS TRANSFER
Internal PANTHER tree node, not a protein. The node carries four IBD terms but only GO:0031012 reached ADAMTSL5.
RGD:621241 · Adamts1 (R. norvegicus) SUPPORTS TRANSFER
Swiss-Prot Q9WUQ1; carries its own IDA to GO:0031012.
UniProtKB:Q6ZMM2 · ADAMTSL5 (H. sapiens) - the target itself CIRCULAR OR REDUNDANT
Self-referential seed. Valid: it records a PAINT curator judging this function core for the gene. ADAMTSL5 independently holds its own IDA to GO:0031012 from PMID:23010571.
UniProtKB:Q8TE56 · ADAMTS17 (H. sapiens) SUPPORTS TRANSFER
Swiss-Prot Q8TE56; carries its own IDA to GO:0031012.
UniProtKB:Q9H324 · ADAMTS10 (H. sapiens) SUPPORTS TRANSFER
Swiss-Prot Q9H324; carries its own IDA to GO:0031012.
UniProtKB:Q9P2N4 · ADAMTS9 (H. sapiens) SUPPORTS TRANSFER
Swiss-Prot Q9P2N4; carries its own IDA to GO:0031012.
WB:WBGene00003242 · mig-6 (C. elegans) SUPPORTS TRANSFER
Swiss-Prot O76840; carries its own IDA to GO:0005604.
Supporting Evidence:
PMID:23010571
co-localized with fibrillin microfibrils in the extracellular
file:genes/human/ADAMTSL5/ADAMTSL5-bioinformatics/RESULTS.md
All 16 protein donors carry their own experimental (IDA/HDA) annotation
GO:0005576 extracellular region
IEA
GO_REF:0000044 		ACCEPT
Summary: UniProt Subcellular Location keyword mapping (SL-0243, Secreted). Correct and independently confirmed experimentally.
Reason: ADAMTSL5 has a predicted signal peptide (residues 1-42) and a mature chain (43-481), and secretion was demonstrated directly: recombinant protein is recovered from the conditioned medium of three transfected cell lines. The SubCell-to-GO mapping is doing exactly what it should here, and the same term is independently supported by an IDA row on this gene.
Supporting Evidence:
PMID:23010571
Recombinant ADAMTSL5 is a secreted, N-glycosylated 60kDa
GO:0030198 extracellular matrix organization
IEA
GO_REF:0000002 		MARK AS OVER ANNOTATED
Summary: InterPro2GO transfer from IPR013273, a family signature that bundles the catalytic ADAMTS proteases together with the non-catalytic ADAMTS-like proteins. For the proteases, extracellular matrix organization follows from ECM proteolysis; ADAMTSL5 has no proteolytic activity, and the one direct test of matrix organization by ADAMTSL5 was negative.
Reason: Two independent legs. First, the source signature is too broad to carry this process. InterPro:IPR013273 is named "ADAMTS/ADAMTS-like", covers 26,580 proteins, and carries exactly one GO mapping - this term. It matches ADAMTSL5 legitimately, but the ECM-organizing role of the family's catalytic members is a consequence of their metalloprotease activity, which ADAMTSL5 does not have: UniProt flags the missing catalytic domain in a CAUTION line, the primary paper states the ADAMTSLs have no proteolytic activity, and a sequence scan finds no HExxHxxGxxHD zinc-binding signature - indeed no HExxH substring at all - in the 481-residue sequence, with no peptidase signature among its InterPro matches. This is the familiar "a domain's name is not an activity" failure moved one aspect over, into a family signature too broad to carry a process. Second, the only direct experimental test was negative: adding ADAMTSL5 to fibroblast cultures produced no consistent difference in microfibril density and no discernible effect on microfibril assembly, and direct fibronectin binding was not supported either. Not REMOVE, deliberately: that negative is a "data not shown" result from a single exogenous-protein assay, so it shows no role has been demonstrated rather than refuting one, and UniProt itself only claims the protein may play a role in modulation of fibrillin microfibrils. A third argument was available and deliberately not used - ADAMTSL5 did not receive this term by IBA while ADAMTSL2, ADAMTSL4 and THSD4 did from the same node - because that node's propagation is incoherent family-wide (ADAMTSL1 and PAPLN received nothing at all), so the absence is more likely a coverage gap than a curatorial judgement.
Propagation Review
Root cause: PROPAGATION BAD
Failure modes: FUNCTIONAL DIVERGENCE PSEUDO OR SUBACTIVITY LOSS
Sources checked:
InterPro:IPR013273 · ADAMTS/ADAMTS-like (family signature, 26580 proteins) SUPPORTS SOURCE BUT NOT TARGET
Sole GO mapping is GO:0030198. The signature spans both catalytic ADAMTS proteases and non-catalytic ADAMTS-like proteins, so the mapping is sound for the protease members but does not transfer to a member that lost catalysis.
Supporting Evidence:
PMID:23010571
In contrast to ADAMTS proteases, ADAMTSLs lack a catalytic domain
PMID:23010571
did not identify a consistent difference
file:genes/human/ADAMTSL5/ADAMTSL5-uniprot.txt
lacks the metalloprotease and disintegrin-like domains which are
GO:0005515 protein binding
IPI
PMID:32296183
A reference map of the human binary protein interactome. 		MARK AS OVER ANNOTATED
Summary: Partner CYSRT1 (A8MQ03), from the HuRI yeast two-hybrid screen. An extreme Y2H hub (517 distinct IntAct partners) that localises to the cornified envelope, topologically inaccessible to a secreted ECM protein.
Reason: Assessed per partner, not per gene. CYSRT1 has 517 distinct interaction partners in IntAct across 1670 records, against 12 for ADAMTSL5 - the signature of a promiscuous two-hybrid prey rather than a specific partner. Its only annotated location is the cornified envelope, an intracellular keratinocyte structure that a signal-peptide secreted protein (ADAMTSL5 signal peptide 1-42) never encounters. The interaction is also unreplicated: IntAct logs it under three sub-methods of the same experiment - two hybrid array, two hybrid prey pooling approach, and validated two hybrid - all at MI-score 0.56, so the NbExp=3 recorded in UniProt is one screen counted three ways, not three independent observations. No orthogonal assay exists. The accession resolves to the reviewed canonical Swiss-Prot entry at its canonical length (144 aa), so this is genuine screen noise rather than a mis-identified partner. Not REMOVE: the assay was performed and curated correctly by IntAct, it simply does not support a biologically meaningful binding function.
Supporting Evidence:
file:genes/human/ADAMTSL5/ADAMTSL5-bioinformatics/RESULTS.md
CYSRT1 (A8MQ03) | 1670 | **517** | cornified envelope | 144 aa |
GO:0005515 protein binding
IPI
PMID:32296183
A reference map of the human binary protein interactome. 		MARK AS OVER ANNOTATED
Summary: Partner KRTAP5-9 (P26371), from the same HuRI screen. A keratin-associated protein of the intracellular hair-keratin matrix, and a Y2H hub with 213 distinct partners.
Reason: Assessed per partner. KRTAP5-9 is an ultrahigh-sulfur keratin-associated protein that assembles into the intracellular keratin matrix of the hair cuticle; it has no secretory pathway and cannot meet a secreted ECM glycoprotein in vivo. It carries 213 distinct IntAct partners across 842 records, again characteristic of a cysteine-rich, sticky two-hybrid prey. As with the other two rows the interaction is logged under three sub-methods of a single experiment at MI-score 0.56, so it is unreplicated. The accession is the reviewed canonical Swiss-Prot entry at its canonical length (169 aa). Not REMOVE, for the same reason as the CYSRT1 row: the annotation records a real screen result, but the bare term conveys no function and the partner is not credible.
Supporting Evidence:
file:genes/human/ADAMTSL5/ADAMTSL5-bioinformatics/RESULTS.md
KRTAP5-9 (P26371) | 842 | **213** | intracellular hair-keratin matrix | 169 aa |
GO:0005515 protein binding
IPI
PMID:32296183
A reference map of the human binary protein interactome. 		MARK AS OVER ANNOTATED
Summary: Partner FHL5 (Q5TD97), from the same HuRI screen. A nuclear LIM-domain transcriptional co-activator of spermatids, with 108 distinct IntAct partners.
Reason: Assessed per partner. FHL5 is a four-and-a-half LIM domain protein annotated to the nucleus, specifically the nuclei of round and elongated spermatids, where it co-activates CREM. A secreted extracellular matrix glycoprotein and a nuclear transcriptional co-activator occupy mutually inaccessible compartments. FHL5 is also a substantial two-hybrid hub (108 distinct partners across 316 records), and like the other two partners it is supported only by three sub-methods of one experiment at MI-score 0.56 with no orthogonal validation. The accession is the reviewed canonical Swiss-Prot entry at its canonical length (284 aa). The three partners together cohere as a set of cysteine-rich, aggregation-prone Y2H preys rather than as any biology of ADAMTSL5.
Supporting Evidence:
file:genes/human/ADAMTSL5/ADAMTSL5-bioinformatics/RESULTS.md
FHL5 (Q5TD97) | 316 | **108** | nucleus (spermatid nuclei) | 284 aa |
GO:0071953 elastic fiber
TAS
PMID:23962539
Elastic fibres in health and disease. 		MODIFY
Summary: Bulk assignment from a review article used as the TAS source for 62 distinct entities. The gene's own experimental evidence supports the more specific GO:0001527 microfibril, which UniProt already records by IDA but GOA is missing.
Reason: PMID:23962539 is a review whose abstract never mentions ADAMTSL5, and QuickGO shows it is used as the reference for 66 annotations over 62 distinct entities, 41 of them receiving this same term - a bulk curation pass over the elastic fibre literature rather than a gene-specific traceable statement. Tellingly, the same curation from the same review assigned the more specific GO:0001527 microfibril to 15 other proteins including THSD4/ADAMTSL6, so the specific term was available and simply not chosen here. Meanwhile ADAMTSL5's own primary data demonstrates association with fibrillin microfibrils specifically, and UniProt already carries GO:0001527 microfibril by IDA for this protein - a term GOA does not have (QuickGO returns exactly 12 annotations for Q6ZMM2 and GO:0001527 is not among them). GO:0001527 is current, is not obsolete, and is a part_of child of GO:0071953, so replacing the review-derived parent with the evidence-backed child retains the parent by closure while gaining precision and aligns GOA with UniProt. The merged ADAMTSL4 review independently proposed GO:0001527 for its gene, which is useful convergent support. Note that elastic fibre is a poor fit on tissue grounds too: ADAMTSL5's strongest expression is in cartilage and bone, which are not elastic-fibre tissues.
Proposed replacements: microfibril
Supporting Evidence:
PMID:23010571
co-localized with fibrillin microfibrils in the extracellular
file:genes/human/ADAMTSL5/ADAMTSL5-uniprot.txt
DR GO; GO:0001527; C:microfibril; IDA:UniProtKB.
GO:0050436 microfibril binding
IDA
PMID:23010571
A disintegrin-like and metalloprotease domain containing thr... 		ACCEPT
Summary: Core molecular function. Direct affinity co-isolation showed binding to both fibrillin-1 and fibrillin-2 and to assembled fibrillin microfibrils - ADAMTSL5 is the first family member shown to bind both fibrillins.
Reason: Well supported by two complementary assays in the same paper: myc-tagged ADAMTSL5 co-isolated the N-terminal half of fibrillin-1 and both halves of fibrillin-2, with empty-vector controls showing no bead binding, and exogenous ADAMTSL5 added to fibroblast cultures decorated assembled fibrillin microfibrils. This is the most informative molecular function on the record and there is no more specific term available - GO has no fibrillin binding term (confirmed via OLS), and GO:0050436 is_a GO:0050840 extracellular matrix binding, so this is the appropriate level. Retained as a core function.
Supporting Evidence:
PMID:23010571
Taken together, the findings are consistent with specific binding of ADAMTSL5 to fibrillin-1 and fibrillin-2 and to their macromolecular assemblies, i.e., fibrillin microfibrils.
GO:0031012 extracellular matrix
IDA
PMID:23010571
A disintegrin-like and metalloprotease domain containing thr... 		ACCEPT
Summary: Direct demonstration that secreted ADAMTSL5 deposits into fibroblast extracellular matrix, with a non-permeabilised staining control establishing the signal is extracellular. MGI-assigned.
Reason: The localisation was shown by immunostaining of fibroblast cultures exposed to ADAMTSL5, with two controls that make the call solid: empty-vector conditioned medium gave no signal, and staining without prior fixation reproduced the pattern, establishing that the decorated structures are extracellular rather than intracellular. UniProt records the same conclusion as an experimental subcellular location. One caveat recorded but not acted on: the assay used exogenous recombinant protein rather than protein at native levels, which by a strict reading leans toward IMP rather than IDA. I have not overturned the curator's call - for a secreted protein this is the standard assay and the non-permeabilised control addresses the main artefact.
Supporting Evidence:
PMID:23010571
suggesting that the stained structures were extracellular and corresponded to fibrillin-1 containing microfibrils in ECM
GO:0005576 extracellular region
IDA
PMID:23010571
A disintegrin-like and metalloprotease domain containing thr... 		ACCEPT
Summary: Secretion demonstrated directly - recombinant ADAMTSL5 is recovered from the conditioned medium of transfected HEK293F, COS-1 and CHO-K1 cells. MGI-assigned.
Reason: Straightforward and correct, though it is the least informative of the three location rows since it is a parent of the extracellular matrix annotation the same paper supports. Secretion was reproduced in three independent cell lines and is consistent with the predicted signal peptide (1-42). Kept as non-redundant with GO:0031012 because it additionally captures the soluble medium pool, which matters for this protein: part of it is proteolytically processed to release the C-terminal NTR fragment into the medium.
Supporting Evidence:
PMID:23010571
in the medium of transfected cells
GO:0008201 heparin binding
IDA
PMID:23010571
A disintegrin-like and metalloprotease domain containing thr... 		ACCEPT
Summary: Core molecular function, mapped to a specific module. Heparin-agarose affinity chromatography showed salt-sensitive ionic binding, and only the C-terminal NTR-containing fragment was retained.
Reason: Directly demonstrated and unusually well localised within the protein. ADAMTSL5 was depleted from medium by heparin-agarose, was not eluted by 0.15 or 0.3 M NaCl but was substantially eluted by 0.5 and 1 M NaCl, establishing an ionic interaction; and of the two fragments present in conditioned medium only the C-terminal NTR-containing one was retained, while the 33 kDa N-terminal fragment flowed through. This is consistent with the general heparin-binding property of NTR modules and gives the protein a plausible mechanism for its observed cell-surface and peri-cellular distribution via heparan sulfate proteoglycans. GO:0008201 is_a GO:0005539 glycosaminoglycan binding, which is the right level: the ligand assayed was heparin specifically. Retained as a core function.
Supporting Evidence:
PMID:23010571
only the C-terminal fragment containing the NTR-module was retained by the heparin matrix
PMID:23010571
indicative of an ionic interaction between ADAMTSL5 and heparin
GO:0031012 extracellular matrix
IDA
PMID:23010571
A disintegrin-like and metalloprotease domain containing thr... 		ACCEPT
Summary: Duplicate of the MGI-assigned extracellular matrix IDA row, independently asserted by UniProt from the same paper. Same verdict.
Reason: GOA carries this localisation twice, once assigned by MGI and once by UniProt, both citing PMID:23010571. Both are reviewed here separately so that every GOA row receives a verdict, and both are accepted on the same evidence. The duplication is a GOA bookkeeping artefact of two groups curating the same paper, not a second independent observation - worth noting so the row count is not read as two lines of support. The paper additionally localises the protein to the peri-cellular and baso-lateral matrix.
Supporting Evidence:
PMID:23010571
These observations strongly suggest that ADAMTSL5 is associated with the peri-cellular and sub-cellular ECM of cells
GO:0001527 microfibril
IDA
PMID:23010571
A disintegrin-like and metalloprotease domain containing thr... 		NEW
Summary: Proposed new annotation. This is the gene's most specific and best-evidenced localisation, it is already in UniProt by IDA, and GOA does not have it at all.
Reason: GOA is missing the one localisation ADAMTSL5's own primary data establishes. UniProt's cross-reference block carries GO:0001527 C:microfibril IDA:UniProtKB, but QuickGO returns exactly 12 annotations for Q6ZMM2 and this term is not among them. The supporting experiment is direct: exogenous ADAMTSL5 added to fibroblast cultures decorated assembled fibrillin microfibrils, with an empty-vector control giving no signal and a non-permeabilised staining control establishing that the decorated structures are extracellular. Proposed as an additional row rather than relying solely on the MODIFY of the GO:0071953 TAS row, because that MODIFY alone would leave the specific term carrying the weak review-derived provenance of PMID:23962539, whereas the evidence that actually supports microfibril localisation is an IDA from PMID:23010571. Adding it here gives the UniProt/GOA discrepancy raised in suggested_questions a concrete target, and matches how the merged ADAMTSL4 review handled the same term.
Supporting Evidence:
PMID:23010571
the exogenous protein co-localized with fibrillin-1 in microfibrils
file:genes/human/ADAMTSL5/ADAMTSL5-uniprot.txt
DR GO; GO:0001527; C:microfibril; IDA:UniProtKB.

Core Functions

Binds fibrillin-1 and fibrillin-2 and decorates assembled fibrillin microfibrils in the extracellular matrix, acting as a microfibril-associated accessory glycoprotein rather than as a structural or enzymatic component of the microfibril itself.

Molecular Function:
microfibril binding
Cellular Locations:
microfibril extracellular matrix
Supporting Evidence:
  • PMID:23010571
    ADAMTSL5 is the first family member shown to bind both fibrillin-1

Binds heparin and, by extension, cell-surface heparan sulfate proteoglycans through its C-terminal netrin-like module in a salt-sensitive ionic interaction, which plausibly anchors the secreted protein in the peri-cellular matrix.

Molecular Function:
heparin binding
Cellular Locations:
extracellular region
Supporting Evidence:
  • PMID:23010571
    Affinity-chromatography demonstrated heparin-binding of ADAMTSL5

References

GO_REF:0000002
Gene Ontology annotation through association of InterPro records with GO terms
  • Maps InterPro signature matches to GO terms. For ADAMTSL5 the relevant signature is IPR013273 ADAMTS/ADAMTS-like, whose sole GO mapping is GO:0030198 extracellular matrix organization. The signature covers 26,580 proteins and spans both the catalytic ADAMTS proteases and the non-catalytic ADAMTS-like proteins.
GO_REF:0000033
Annotation inferences using phylogenetic trees
  • PAINT phylogenetic propagation. For ADAMTSL5 the single IBA row derives from PANTHER node PTN000347317 in family PTHR13723, with 16 protein donors. The GOA WITH/FROM field matches the cached PAINT seed list for that node exactly.
GO_REF:0000044
Gene Ontology annotation based on UniProtKB/Swiss-Prot Subcellular Location vocabulary mapping, accompanied by conservative changes to GO terms applied by UniProt
  • Maps the UniProt Secreted subcellular location keyword (SL-0243) to GO:0005576 extracellular region. Correct for ADAMTSL5, which has a signal peptide and experimentally demonstrated secretion.
PMID:23010571
A disintegrin-like and metalloprotease domain containing thrombospondin type 1 motif-like 5 (ADAMTSL5) is a novel fibrillin-1-, fibrillin-2-, and heparin-binding member of the ADAMTS superfamily containing a netrin-like module.
  • ADAMTSL5 is a secreted, N-glycosylated ~60 kDa glycoprotein. Recombinant protein is recovered from the conditioned medium of transfected HEK293F, COS-1 and CHO-K1 cells and deposits into the peri-cellular and baso-lateral extracellular matrix.
    "Recombinant ADAMTSL5 is a secreted, N-glycosylated 60kDa"
  • ADAMTSL5 binds heparin through its C-terminal NTR module in a salt-sensitive ionic interaction; only the NTR-containing C-terminal fragment is retained on heparin-agarose.
    "only the C-terminal fragment containing the NTR-module was retained by the heparin matrix"
  • ADAMTSL5 binds both fibrillin-1 and fibrillin-2 and co-localises with assembled fibrillin microfibrils in fibroblast extracellular matrix - the first family member shown to bind both fibrillins.
    "Taken together, the findings are consistent with specific binding of ADAMTSL5 to fibrillin-1 and fibrillin-2 and to their macromolecular assemblies, i.e., fibrillin microfibrils."
  • Critically, the same experiment found no effect of ADAMTSL5 on microfibril assembly, and direct binding to fibronectin was not supported. This is the only direct test of a matrix-organizing role for ADAMTSL5 and it was negative, though reported as data not shown.
    "did not identify a consistent difference"
  • The ADAMTS-like proteins, including ADAMTSL5, have no catalytic domain and hence no proteolytic activity - the basis for rejecting any protease-derived functional inference for this gene.
    "In contrast to ADAMTS proteases, ADAMTSLs lack a catalytic domain"
  • In mouse organogenesis ADAMTSL5 is broadly expressed, prominently in musculoskeletal tissues - skeletal muscle, cartilage and bone - and in many epithelia. Cartilage and bone are not elastic-fibre tissues.
    "Immunostaining during mouse organogenesis identified ADAMTSL5 in"
PMID:23962539
Elastic fibres in health and disease.
  • A general review of elastic fibre composition, assembly and disease. Its abstract does not mention ADAMTSL5. QuickGO shows it is used as the TAS reference for 66 annotations across 62 distinct entities, 41 of which receive GO:0071953 elastic fiber - a bulk curation pass rather than a gene-specific traceable statement. The same pass assigned the more specific GO:0001527 microfibril to 15 other proteins.
PMID:32296183
A reference map of the human binary protein interactome.
  • The HuRI systematic yeast two-hybrid screen, source of all three GO:0005515 rows for ADAMTSL5 (partners CYSRT1, KRTAP5-9 and FHL5). IntAct logs each partner under three sub-methods of the same experiment - two hybrid array, two hybrid prey pooling approach and validated two hybrid - all at MI-score 0.56, so UniProt's NbExp=3 is one screen counted three ways rather than three independent observations.
PMID:26621454
Melanocyte antigen triggers autoimmunity in human psoriasis.
  • ADAMTSL5 is an HLA-C*06:02-presented melanocytic autoantigen recognised by a Valpha3S1/Vbeta13S1 CD8+ T cell receptor from a psoriasis patient, and ADAMTSL5 stimulation induced IL-17A in CD8+ T cells from psoriasis patients only. The recognised peptide VRSRRCLRL maps to residues 67-75, inside the TSP type-1 domain.
    "we identified ADAMTS-like protein 5 (ADAMTSL5) as an"
PMID:33197513
ADAMTSL5 is an epigenetically activated gene underlying tumorigenesis and drug resistance in hepatocellular carcinoma.
  • ADAMTSL5 is activated by gene-body CpG island hypermethylation in mouse and human hepatocellular carcinoma. Depletion interfered with tumorigenic properties, while overexpression conferred tumorigenicity on pre-tumoural hepatocytes sensitised by modest MET expression.
    "ADAMTSL5 targeting interfered with tumorigenic properties of HCC"
file:genes/human/ADAMTSL5/ADAMTSL5-bioinformatics/RESULTS.md
ADAMTSL5 bioinformatics analysis - WITH/FROM resolution, family propagation census, and interaction-partner assessment
  • All 16 protein donors on the GO:0031012 IBA row carry their own experimental IDA/HDA evidence for that term or a descendant, and the donor set spans four distinct locations, making GO:0031012 the correct least common ancestor.
    "All 16 protein donors carry their own experimental (IDA/HDA) annotation"
  • Propagation from PANTHER node PTN000347317 is incoherent across the human family: GO:0031012 reached 5 of 7 non-catalytic members and GO:0030198 only 3 of 7, while human PAPLN received no IBA at all despite three of its own orthologs being seeds for GO:0031012 at that node.
    "PAPLN receives neither"
  • The three GO:0005515 partners are extreme Y2H hubs (517, 213 and 108 distinct IntAct partners against ADAMTSL5's 12) and all are topologically inaccessible to a secreted protein.
    "CYSRT1 (A8MQ03) | 1670 | **517** | cornified envelope | 144 aa |"
  • ADAMTSL5 has no zinc-binding metalloprotease motif - no HExxHxxGxxHD and no HExxH substring at all in its 481 residues - confirming the loss of catalysis independently of UniProt's CAUTION line.
    "no `HExxH` substring at all"
file:genes/human/ADAMTSL5/ADAMTSL5-deep-research-affinage.md
Affinage mechanistic annotation for ADAMTSL5 (human)
  • Provider deep-research record. Its narrative of the fibrillin, heparin, psoriasis and hepatocellular carcinoma literature is broadly accurate and every citation is a numeric PMID, but its GO grounding proposes GO:0008289 lipid binding and GO:0098772 molecular function regulator activity, neither of which has any support - the demonstrated ligand is heparin, a glycosaminoglycan. Not imported.
    "GO:0008289 lipid binding"

Suggested Questions for Experts

Q: PAINT recommendation, stated once for the whole family rather than per gene. Node PTN000347317 in PTHR13723 carries four IBD annotations, but their propagation to the eight human ADAMTSL/papilin members is incoherent: GO:0031012 reached 5 of 7 non-catalytic members (missing ADAMTSL1 and PAPLN) and GO:0030198 reached only 3 of 7 (missing ADAMTSL1, ADAMTSL3, ADAMTSL5 and PAPLN), while all four catalytic ADAMTS controls received all four terms. The sharpest case is human PAPLN, which received no IBA at all even though three of its own orthologs - Drosophila Ppn (FB:FBgn0003137), mouse Papln (MGI:MGI:2386139) and C. elegans mig-6/ppn-1 (WB:WBGene00003242) - are seeds for GO:0031012 at that very node. Human ADAMTSL1 likewise has no IBA at all despite being in the family (SF157). Is this a propagation-coverage gap rather than a set of deliberate curatorial exclusions, and can the node's terms be re-propagated across the clade?

Q: Related PAINT point, in the opposite direction. The family's loss of catalysis is modelled correctly - GO:0004222 and GO:0006508 reached all four catalytic ADAMTS controls and none of the seven ADAMTS-like members - but only ADAMTSL2 records this explicitly, via a NOT-qualified enables GO:0004222 IBA (IKR at node PTN002673039). For ADAMTSL1, ADAMTSL3, ADAMTSL4, ADAMTSL5, THSD4 and PAPLN the absence of catalytic activity is implicit, i.e. simply nothing propagated. Would placing the negation at the node ancestral to the whole non-catalytic clade make this informative loss explicit for all of them?

Q: Ontology issue. GO:0001527 microfibril is a part_of child of GO:0071953 elastic fiber, which asserts that every microfibril is part of an elastic fiber. This contradicts GO:0001527's own definition, which describes microfibrils as extracellular matrix components occurring independently or along with elastin, and contradicts the ciliary zonule - a fibrillin microfibril structure essentially devoid of elastin, and the reason FBN1 mutations cause ectopia lentis. Should the part_of be relaxed, or GO:0001527 re-parented under a fibrillin-microfibril concept independent of elastic fibre?

Q: UniProt/GOA discrepancy to reconcile. UniProt's cross-reference block carries GO:0001527 C:microfibril IDA:UniProtKB for ADAMTSL5, but GOA does not contain that annotation - QuickGO returns exactly 12 annotations for Q6ZMM2 and GO:0001527 is not among them. The gene's best-evidenced and most specific localisation is therefore missing from GOA, while a weaker review-derived TAS row to the broader parent GO:0071953 is present. Which record is stale?

Q: Does ADAMTSL5 have any demonstrable effect on fibrillin microfibril biology? The only test to date found no consistent difference in microfibril density and no discernible effect on assembly, in a single exogenous-protein assay reported as data not shown. Given that the protein binds both fibrillins and is proteolytically processed to release its NTR module, a loss-of-function experiment is needed before any extracellular matrix organization annotation is justified.

Suggested experts: Suneel S. Apte, Dieter P. Reinhardt

Q: What is the function, if any, of the released C-terminal NTR fragment? UniProt records that ADAMTSL5 is proteolytically cleaved to release an NTR-containing fragment, this fragment carries the heparin-binding activity, and it is detectable in mouse skeletal muscle, testis, kidney and heart. Is the cleavage regulated, and does the fragment have activity distinct from the full-length protein?

Suggested experts: Suneel S. Apte, Hannes L. Bader

Suggested Experiments

Experiment: Generate Adamtsl5-null mice and examine fibrillin microfibril architecture in skeletal muscle, cartilage, bone and the ciliary zonule by immunofluorescence and transmission electron microscopy. This is the loss-of-function test the field lacks, and it would settle whether any extracellular matrix organization annotation is warranted. Compare against the Adamtsl2 and Thsd4 nulls, whose microfibril phenotypes are established.

Hypothesis: If ADAMTSL5 modulates fibrillin microfibril assembly in vivo, loss of the gene will perturb microfibril architecture in the tissues where the protein is most abundant.

Type: Mouse knockout with tissue ultrastructure

Experiment: Map the fibrillin-binding site. The primary paper showed ADAMTSL5 binds the N-terminal half of fibrillin-1 and both halves of fibrillin-2, with the N-terminal half binding more strongly. Narrow this with recombinant fibrillin subfragments and surface plasmon resonance to obtain an affinity and a domain-level site, then test competition with ADAMTSL2, ADAMTSL4 and THSD4.

Hypothesis: ADAMTSL5 binds a discrete N-terminal region shared by fibrillin-1 and fibrillin-2, and competes there with other microfibril-associated ADAMTSL proteins.

Type: Surface plasmon resonance with recombinant fragments

Experiment: Identify the protease that releases the NTR module and test whether the cleavage is regulated. Combine N-terminal sequencing or terminal-amine labelling of the released fragment with a protease-inhibitor panel in transfected cells, then ask whether the free NTR fragment and full-length ADAMTSL5 differ in heparin affinity and in microfibril decoration.

Hypothesis: Release of the C-terminal NTR module is a regulated processing event that changes the protein's matrix-binding behaviour.

Type: Protease identification and N-terminomics

Experiment: Compare ADAMTSL5 deposition on wild-type CHO cells with the glycosaminoglycan-deficient pgsA-745 and pgsD-677 mutants, and after heparinase III treatment. If the NTR-heparin interaction is what anchors the protein, retention should be lost in the mutants while secretion itself is unaffected.

Hypothesis: Peri-cellular retention of secreted ADAMTSL5 depends on the NTR-heparan sulfate interaction rather than on protein-protein contacts with the matrix.

Type: Cell-surface retention in glycosaminoglycan-deficient mutants

Experiment: Test whether purified recombinant ADAMTSL5 added to the medium reproduces the changes in MET, EGFR and FGFR4 seen on knockdown, and whether the effect requires the NTR module. This would distinguish a bona fide extracellular signal from an intracellular or indirect consequence of depleting the gene, and would determine whether any signalling GO annotation is ever justified.

Hypothesis: The reported effect of ADAMTSL5 on receptor tyrosine kinase levels in hepatocellular carcinoma is a genuine extracellular, matricrine activity requiring the heparin-binding NTR module.

Type: Recombinant protein add-back with domain deletion

Deep Research

Affinage

(ADAMTSL5-deep-research-affinage.md)
Affinage mechanistic annotation for ADAMTSL5 (human) Affinage Affinage (Claude Sonnet reading pass + Opus synthesis pass) 7 citations

Affinage mechanistic annotation for ADAMTSL5 (human)

⚠️ CAUTION — trust gate(s) tripped; review before using:

  • Affinage's own head-to-head self-evaluation scored this record pairwise = tie (not win) vs the curated UniProt reference — treat the narrative with extra scepticism.

Current model (mechanistic narrative)

ADAMTSL5 is a secreted, N-glycosylated extracellular matrix glycoprotein that associates with fibrillin microfibrils and also functions as a disease-relevant signaling modulator and autoantigen [PMID:23010571, PMID:33197513, PMID:26621454]. As a matrix component, it binds both fibrillin-1 and fibrillin-2, co-localizes with fibrillin microfibrils in fibroblast cultures, and binds heparin through its C-terminal netrin-like (NTR) module, which can be proteolytically released PMID:23010571. In hepatocellular carcinoma, ADAMTSL5 sustains oncogenic receptor tyrosine kinase signaling: its depletion lowers expression and/or phosphorylation of MET, EGFR, PDGFRβ, IGF1Rβ, and FGFR4 while raising AXL, and its overexpression confers tumorigenicity to MET-sensitized hepatocytes, with its own expression linked to gene-body CpG island hypermethylation PMID:33197513. In psoriasis, ADAMTSL5 is an HLA-C06:02–presented melanocyte autoantigen whose VRSRRCLRL peptide is recognized by an autoreactive Vα3S1/Vβ13S1 CD8+ TCR, driving IL-17A responses PMID:26621454; the structural basis of this recognition is an extensive complementary electrostatic interface between negatively charged TCR residues and exposed arginines of the self-peptide and the HLA-C06:02 α1 helix PMID:37330172. This TCR is polyspecific, also responding to environmental peptides from wheat, microbiota, and pathogens, providing a route by which environmental antigens may trigger the ADAMTSL5-directed autoimmune response PMID:38524140.

Affinage mechanism profile (its own GO/Reactome grounding)

Recorded for reference. The AIGR evaluation found this grounding is coarse (collapses to general parents) and can contradict the narrative — do not import these GO ids directly; re-ground from the narrative + PMIDs.

  • molecular_activity: GO:0008289 lipid binding, GO:0098772 molecular function regulator activity
  • localization: GO:0031012 extracellular matrix, GO:0005576 extracellular region
  • pathway (Reactome): R-HSA-1474244 Extracellular matrix organization, R-HSA-168256 Immune System, R-HSA-1643685 Disease
  • partners: FBN1, FBN2
  • complexes: (none)

Dated findings (citation-anchored)

Year Confidence Finding PMIDs Journal
2015 High ADAMTSL5 was identified as an HLA-C*06:02-presented melanocytic autoantigen recognized by a Vα3S1/Vβ13S1 T cell receptor (TCR) reconstituted from an epidermal CD8+ T cell clone of a psoriasis patient. Melanocytes are the skin-specific target cells expressing ADAMTSL5, and ADAMTSL5 stimulation induced IL-17A in CD8+ T cells from psoriasis patients only. PMID:26621454 The Journal of experimental medicine
2012 High ADAMTSL5 is a secreted, N-glycosylated ~60 kDa glycoprotein that binds both fibrillin-1 and fibrillin-2 (the first ADAMTS family member shown to bind both), co-localizes with fibrillin microfibrils in the extracellular matrix of cultured fibroblasts, and binds heparin via its C-terminal netrin-like (NTR) module. Proteolytic release of the NTR module was also observed. Alternative splicing at the 5' end generates two transcripts encoding different signal peptides but the same mature protein, with differing translational efficiency. PMID:23010571 Matrix biology : journal of the International Society for Matrix Biology
2023 High The crystal structure of the psoriatic Vα3S1/Vβ13S1 TCR in complex with HLA-C06:02 presenting the ADAMTSL5 peptide (VRSRRCLRL) was determined. TCR docking involves an extensive complementary charge network between negatively charged TCR residues and exposed arginine residues from the ADAMTSL5 self-peptide and the HLA-C06:02 α1 helix. Mutagenesis and activation assays confirmed these electrostatic interactions are functionally critical. PMID:37330172 The Journal of biological chemistry
2020 High ADAMTSL5 maintains the function of key oncogenic signaling pathways in hepatocellular carcinoma (HCC). ADAMTSL5 depletion reduced expression and/or phosphorylation of receptor tyrosine kinases MET, EGFR, PDGFRβ, IGF1Rβ, and FGFR4, and increased AXL expression. Conversely, ADAMTSL5 overexpression conferred tumorigenicity to pre-tumoural hepatocytes sensitized by modest MET receptor expression. ADAMTSL5 expression correlates with gene body CpG island hypermethylation at its locus. PMID:33197513 Journal of hepatology
2024 Medium Multiple environmental peptides (from wheat, Saccharomyces cerevisiae, microbiota, tobacco, and pathogens) activate the same psoriatic Vα3S1/Vβ13S1 TCR that recognizes ADAMTSL5, as demonstrated by lymphocyte stimulation experiments. HLA-C*06:02 tetramers loaded with ADAMTSL5 or wheat peptides showed the same CD8+ T cell population can recognize both, establishing TCR polyspecificity as a mechanism by which environmental antigens may trigger the ADAMTSL5-directed autoimmune response. PMID:38524140 Frontiers in immunology
2016 Medium ADAMTSL5 protein is expressed not only in epidermal melanocytes but also in keratinocytes throughout the epidermis and in some dermal blood vessels and perivascular dermal cells in psoriatic skin, as shown by immunohistochemistry with three different antibodies. PMID:27857980 Journal of pigmentary disorders
2017 Medium ADAMTSL5 and LL37 protein levels are significantly increased in lesional psoriatic skin and are co-expressed by dendritic cells, macrophages, and some T cells in the dermis. ADAMTSL5 expression is significantly downregulated following treatment with IL-17 or TNF-α blockade, indicating that psoriasis-related cytokines feed-forward induction of ADAMTSL5. PMID:28482118 Experimental dermatology

Citations

  • PMID:23010571
  • PMID:26621454
  • PMID:27857980
  • PMID:28482118
  • PMID:33197513
  • PMID:37330172
  • PMID:38524140

📚 Additional Documentation

Notes

(ADAMTSL5-notes.md)

ADAMTSL5 (Q6ZMM2) — review notes

PAINT + affinage campaign. Branch paint/ADAMTSL5.

What the gene is

ADAMTSL5 (ADAMTS-like protein 5; synonym THSD6) is a small secreted, N-glycosylated
extracellular-matrix glycoprotein of the ADAMTS superfamily. At 481 aa it is the
smallest member of the family by a wide margin (ADAMTSL1 1762, ADAMTSL2 951, ADAMTSL3
1691, ADAMTSL4 1074, THSD4 1018, PAPLN 1278 aa). Domain architecture from the UniProt
feature table: signal peptide 1–42, mature chain 43–481, a single TSP type-1 domain
(45–97), and a C-terminal NTR module (360–479), joined by a proline-rich,
disordered segment (331–361).

The catalytic question — established, not assumed

The campaign brief flagged "the ADAMTSL proteins lack the catalytic metalloprotease
domain" as a lead to establish. It holds for ADAMTSL5, on three independent lines:

  • UniProt CAUTION: [file:human/ADAMTSL5/ADAMTSL5-uniprot.txt "lacks the metalloprotease and disintegrin-like domains which are typical of that family"]
  • The gene's own primary paper: PMID:23010571
  • Sequence scan (this review): the zinc-binding signature HExxHxxGxxHD is absent, and
    there is no HExxH substring at all in the 481-residue sequence. Its InterPro
    match set contains no peptidase/reprolysin (M12B) signature.

But the predicted annotation error did not occur. ADAMTSL5's GOA contains no
peptidase term of any kind — no GO:0004222, no GO:0008237, no GO:0006508. I
recorded this as a hypothesis that was not confirmed, not as a finding. PAINT
actually got this right: the catalytic terms at node PTN000347317 reached all four
catalytic ADAMTS controls and none of the seven non-catalytic members, and ADAMTSL2
even carries an explicit NOT|enables GO:0004222 (IKR at PTN002673039).

The shape of the error does appear one aspect over, in BP — see GO:0030198 below.

Annotation-by-annotation reasoning

12 GOA rows. The fetch-gene stub seeded only 9 — it collapsed the three
GO:0005515 partner rows into one and the two GO:0031012 IDA rows (MGI- and
UniProt-assigned) into one. Restored to 12 so every row gets its own verdict.

ACCEPT — the experimental core (5 rows, all PMID:23010571)

GO:0008201 heparin binding IDA, GO:0050436 microfibril binding IDA,
GO:0031012 extracellular matrix IDA ×2, GO:0005576 extracellular region IDA.

Heparin binding is well documented and localised to the NTR module:
PMID:23010571,
with the interaction shown to be ionic
PMID:23010571.

Microfibril binding rests on direct affinity co-isolation with both fibrillins plus
colocalisation:
PMID:23010571
GO:0050436 is_a GO:0050840 extracellular matrix binding, and there is no
"fibrillin binding" term in GO (OLS returns nothing), so GO:0050436 is the most
specific molecular function available.

Caveat noted but not acted on: these assays used recombinant/exogenous ADAMTSL5
added to fibroblast cultures rather than native protein, which by the campaign's rule
of thumb leans toward IMP. I did not overturn the curators' IDA calls — for a secreted
protein this is standard practice, and the authors ran a non-permeabilised staining
control confirming the signal was extracellular
PMID:23010571.

ACCEPT — GO:0031012 IBA, and it is unusually well founded

WITH/FROM has 17 tokens = 16 protein donors + the node PANTHER:PTN000347317. This
matches the cached PAINT seed list exactly (asserted in code). All 16 protein
donors carry their own experimental IDA/HDA for the term or a descendant, so
SOURCE_WEAK_OR_INFERRED would be contradicted by my own measurement. One token is
self-referential (UniProtKB:Q6ZMM2) — a PAINT curator judging the function core,
which is valid, hence NO_FAILURE_CORE.

The donor set spans four distinct locations — extracellular matrix (14), basement
membrane (4), interstitial matrix (1), microfibril (1). GRANULARITY_MISMATCH requires
donors to agree; they do not, so GO:0031012 is the LCA and refining it would mean
arbitrarily preferring one donor's compartment. ACCEPT with no specificity upgrade.

Also checked and negative (the ACRV1 shape): the IBA does not land above its donors —
14 of 16 hold the term itself — so no downward MODIFY is warranted.

ACCEPT — GO:0005576 IEA from UniProtKB-SubCell:SL-0243

Straightforward: a signal peptide (1–42) plus experimentally demonstrated secretion.
The SubCell→GO mapping is doing exactly what it should.

MARK_AS_OVER_ANNOTATED — GO:0030198 extracellular matrix organization IEA

Two independent legs:

  1. The source signature bundles proteases with non-proteases.
    InterPro:IPR013273 is named, literally, "ADAMTS/ADAMTS-like" (26,580 proteins) and
    carries exactly one GO mapping — GO:0030198. For the catalytic ADAMTS members, ECM
    organization follows from ECM proteolysis. ADAMTSL5 has no such activity. This is the
    familiar "a domain's name is not an activity" failure moved one aspect over: a
    family signature too broad to carry a process    .
  2. The only direct test of it was negative.
    PMID:23010571
    The abstract puts it as colocalisation "but without discernible effect on microfibril
    assembly". Direct binding to fibronectin was also not supported.

Why not REMOVE. The negative is a "data not shown" result from a single
exogenous-protein assay. It shows no role has been demonstrated; it does not refute
one. UniProt itself hedges — "May play a role in modulation of fibrillin microfibrils".
MARK_AS_OVER_ANNOTATED is the honest ceiling.

A third leg I deliberately did NOT use. ADAMTSL5 has no GO:0030198 IBA, while
ADAMTSL2/ADAMTSL4/THSD4 got one from the same node. That looks like PAINT declining to
propagate the process to ADAMTSL5 — but the propagation is incoherent family-wide (see
below), and ADAMTSL1 and PAPLN received nothing at all. So the absence is more likely
a propagation gap than a judgement, and leaning on it would be rationalising a number I
could not explain.

MARK_AS_OVER_ANNOTATED — the three GO:0005515 rows

All three from PMID:32296183 (HuRI). IntAct shows each logged under three
sub-methods of one experiment — two hybrid array + two hybrid prey pooling approach + validated two hybrid, MI-score 0.56. UniProt's NbExp=3 is therefore
one screen counted three ways — the ACRV1 finding, replicated here on a second gene.

Distinct partner counts (derived as entity sets; IntAct records are not partners):

protein records distinct partners localisation
CYSRT1 1670 517 cornified envelope
KRTAP5-9 842 213 intracellular hair-keratin matrix
FHL5 316 108 nucleus
ADAMTSL5 22 12 secreted / ECM

Every partner is topologically incompatible with a signal-peptide secreted ECM protein.
Decided per partner as the brief requires; all three independently come out the
same. All three resolve to reviewed canonical Swiss-Prot entries at canonical lengths —
no ORFeome/TrEMBL substitution of the ACRV1 kind (negative result, recorded).

I differ here from the merged ADAMTSL4 review, which used REMOVE on four
comparable Y2H GO:0005515 rows. Per this campaign's convention an unreplicated screen
hit is MARK_AS_OVER_ANNOTATED, and REMOVE is reserved for demonstrably wrong
inferences. Flagged as a cross-family inconsistency rather than silently diverging.

MODIFY — GO:0071953 elastic fiber TAS → GO:0001527 microfibril

The TAS source PMID:23962539 is a review (full_text_available: false) used as
the reference for 62 distinct entities, assigning GO:0071953 to 41 of them.
Its abstract never mentions ADAMTSL5. The same curation from the same review assigned
the more specific GO:0001527 microfibril to 15 other proteins including
THSD4/ADAMTSL6 — so the specific term was available and simply not chosen here.

Meanwhile the gene's own primary data supports microfibril association specifically,
and UniProt already records GO:0001527; C:microfibril; IDA:UniProtKB — a term
GOA does not carry (verified: QuickGO returns exactly 12 annotations for Q6ZMM2 and
GO:0001527 is not among them). GO:0001527 is current and is a part_of child of
GO:0071953, so the replacement retains the parent by closure while gaining precision.
The merged ADAMTSL4 review independently proposed GO:0001527 as a NEW term, which is
useful convergent support.

The main PAINT finding: one node, incoherent propagation

All eight human ADAMTSL/papilin proteins sit in PTHR13723. Node PTN000347317
carries four IBD terms. Who received them:

gene subfam GO:0031012 GO:0030198 GO:0004222 GO:0006508
ADAMTSL1 SF157 – – – –
ADAMTSL2 SF147 IBA IBA NOT-IBA –
ADAMTSL3 SF169 IBA – – –
ADAMTSL4 SF144 IBA IBA – –
ADAMTSL5 SF173 IBA – – –
THSD4 SF16 IBA IBA – –
PAPLN SF281 – – – –
ADAMTS1/9/10/17 catalytic IBA IBA IBA IBA

From a single node, GO:0031012 reached 5 of 7 and GO:0030198 reached 3 of 7, in no
biologically coherent pattern. The sharpest case is PAPLN, which received nothing
although three of its own orthologs are seeds for GO:0031012 at that very node —
fly Ppn (FB:FBgn0003137), mouse Papln (MGI:MGI:2386139) and worm mig-6/ppn-1
(WB:WBGene00003242). Human PAPLN and human ADAMTSL1 both have no IBA at all.

This is not the "right term, wrong node" defect (AADACL) nor the "mis-placed member"
defect (ACTL8). It is a propagation-coverage defect: the node and its term
assignments look correct, and the term simply failed to reach some descendants.

Reported once in suggested_questions, naming all affected genes, rather than repeated
per gene.

Cross-check with the sibling agents (ADAMTSL1, ADAMTSL3)

Derived independently from QuickGO rather than read off their branches:

  • ADAMTSL3 carries an GO:0031012 IBA identical to mine — same node
    PTN000347317, same 17 WITH/FROM tokens. Our two derivations should agree; if that
    review reaches a different verdict on the same row, the disagreement is the finding.
    ADAMTSL3 also shares the GO:0071953 TAS from PMID:23962539 and the InterPro
    GO:0030198 IEA — so all three of my non-ACCEPT verdicts have a direct counterpart
    there.
  • ADAMTSL1 has only 4 annotations and no IBA at all despite being in
    PTHR13723 (SF157). Its only BP row is the same InterPro GO:0030198 IEA. If that
    review reports ADAMTSL1 as having IBA support, that is a discrepancy worth resolving.

An ontology issue this surfaced

GO:0001527 microfibril is part_of GO:0071953 elastic fiber, which asserts every
microfibril is part of an elastic fiber. That contradicts GO:0001527's own
definition — "Extracellular matrix components occurring independently or along
with elastin" — and contradicts the ciliary zonule, a fibrillin-microfibril structure
essentially devoid of elastin (the reason FBN1 mutations cause ectopia lentis).
ADAMTSL5's own expression pattern is a further mismatch: cartilage and bone are not
elastic-fibre tissues
PMID:23010571

Non-GO biology (informs description, not annotations)

  • Psoriasis autoantigen. ADAMTSL5 is an HLA-C*06:02-presented melanocyte
    autoantigen PMID:26621454.
    The recognised peptide VRSRRCLRL maps to residues 67–75, inside the TSP type-1
    domain (verified against the UniProt sequence). This is a property of the peptide in
    an MHC groove, not a molecular function of the secreted protein, so it generates no
    GO annotation.
  • Hepatocellular carcinoma. PMID:33197513
    Knockdown/overexpression in a cancer context; I did not propose a signalling GO term
    from it, since the RTK effects are downstream and the mechanism connecting a secreted
    ECM protein to receptor levels is not established.

Provider record (affinage)

gates_passed: False. The tripped gate is specifically the self-evaluation pairwise
tie (self_evaluation_pairwise: tie), not a faithfulness failure — faith_pct is
100.0. All 7 citations are numeric PMIDs; no PMID:bio_* bioRxiv ids. I verified
every claim I used against the cited PMID directly and quoted the PMIDs, never the
provider's prose. The narrative is broadly accurate on this gene; its GO grounding
however proposes GO:0008289 lipid binding and GO:0098772 molecular function regulator activity, neither of which has any support — the ligand demonstrated is
heparin, a glycosaminoglycan (GO:0008201), not a lipid. Not imported.

Checks run, including the ones that came back negative

  • Retraction / erratum / expression of concern on all 9 PMIDs relied on, via PubMed
    publication types and each record's own CommentsCorrections (a Publisher
    Correction is invisible to a pubtype query): none flagged. PMID:26621454 has two
    Comment in entries — commentaries, not corrections.
  • Reference projection test: PMID:23010571 annotates only 2 entities
    (ADAMTSL5 + mouse ortholog) — gene-specific curation, not a projection. Negative.
  • Partner accession integrity: all three IPI partners reviewed, canonical, correct
    length. Negative.
  • IBA-above-donor precision defect: absent. Negative.
  • Dead-accession check: every WITH/FROM UniProt lookup returns a named entry; the
    resolver raises rather than reporting a silent zero.

Process notes worth carrying to the next gene

  • QuickGO's geneProductId rejects MOD ids (MGI:, FB:, WB:, RGD:) with
    HTTP 400 in every form. Donor evidence can only be queried through the resolved
    UniProt accession. My first script printed "NONE" for all 13 MOD tokens — a silent
    degradation reading as a null result, which would have inverted the conclusion from
    "16/16 donors have their own experimental evidence" to "almost none do".
  • UniProt does not index WBGene ids (xref:wormbase-WBGene00003242 → []); it keys
    WormBase xrefs on the CDS name. Resolve WBGene → symbol via the WormBase REST first.
  • That symbol search must use gene_exact:. Fuzzy gene:mig-6 also returns
    mig-10/mig-5/mig-14/mig-18 and puts the wrong entry first. The 16/16 count would
    otherwise have been right for the wrong reason.
  • QuickGO strips the DB prefix from WITH/FROM tokens. withFrom[].connectedXrefs[]
    returns {"db": "FB", "id": "FBgn0003137"} as two fields, so the id alone is
    FBgn0003137, not FB:FBgn0003137. Comparing that set naively against a GOA TSV
    WITH/FROM string reports 17 vs 17 tokens that are "not identical" for data that is
    in fact the same — the token must be reassembled as db + ":" + id first. I hit this
    while verifying the ADAMTSL3 cross-check; the sibling comparison itself was sound
    (both sides came from QuickGO, so both were in the stripped serialisation), but the
    same-size-yet-different result is exactly the kind of discrepancy that must be
    investigated rather than waved through.

Bioinformatics Results

(RESULTS.md)

ADAMTSL5 (Q6ZMM2) — bioinformatics support for the GO annotation review

Reproduce with:

uv run python genes/human/ADAMTSL5/ADAMTSL5-bioinformatics/resolve_withfrom.py
uv run python genes/human/ADAMTSL5/ADAMTSL5-bioinformatics/family_propagation_census.py

resolve_withfrom.py must be run first — the census reads its withfrom_resolution.json.
All numbers below are produced by those two scripts; none is hand-entered.

1. ADAMTSL5 has no metallopeptidase domain — verified three independent ways

The campaign brief flagged "ADAMTSL proteins lack the catalytic metalloprotease domain"
as a lead to establish, not a premise. It is established here:

source finding
UniProt CC CAUTION "Although strongly similar to members of the ADAMTS family it lacks the metalloprotease and disintegrin-like domains which are typical of that family."
The gene's own primary paper (PMID:23010571) "In contrast to ADAMTS proteases, ADAMTSLs lack a catalytic domain and thus have no proteolytic activity."
Sequence scan (this analysis) The zinc-binding metalloprotease signature HExxHxxGxxHD is absent, and there is no HExxH substring at all in the 481-residue sequence.
InterPro match set TSP1 repeat, netrin domain, TIMP-like OB-fold, ADAMTS spacer-1, ADAMTS cys-rich-3. No peptidase/reprolysin (M12B) signature.

The predicted failure mode did NOT occur. ADAMTSL5's GOA contains no
peptidase/metallopeptidase term of any kind — no GO:0004222, no GO:0008237, no
GO:0006508. This was a hypothesis, and it is reported here as not confirmed,
not as a finding. PAINT in fact handled the catalysis question correctly (§3).

Domain architecture (UniProt FT): signal peptide 1–42, mature chain 43–481,
TSP type-1 domain 45–97, NTR domain 360–479. At 481 aa ADAMTSL5 is by far the
smallest member of the family (others 951–1935 aa).

Incidental, verified: the psoriasis autoantigen peptide VRSRRCLRL (PMID:26621454)
occupies residues 67–75, inside the TSP type-1 domain.

2. The GO:0031012 IBA is exceptionally well supported — 16/16 donors

WITH/FROM has 17 tokens: 16 protein donors + the tree node PANTHER:PTN000347317.
The GOA field matches the cached PAINT seed list exactly (asserted in code:
goa_tokens - seeds == {"PANTHER:PTN000347317"} and seeds - goa_tokens == set()).

All 16 protein donors carry their own experimental (IDA/HDA) annotation to
GO:0031012 or a descendant. So SOURCE_WEAK_OR_INFERRED / SOURCE_EVIDENCE_WEAK
would be factually contradicted by the measurement; there is no propagation defect to
report on this row.

One token is self-referential (UniProtKB:Q6ZMM2) — a PAINT curator judging the
function core, which is valid.

The donor set is heterogeneous, so the general term is the LCA

donor location donors
GO:0031012 extracellular matrix 14
GO:0005604 basement membrane 4
GO:0005614 interstitial matrix 1
GO:0001527 microfibril 1

Four distinct locations. GRANULARITY_MISMATCH requires the donors to agree; they
do not, so GO:0031012 is the correct LCA and refining it would mean arbitrarily
preferring one donor's compartment. → ACCEPT, no specificity upgrade.

Also checked (ACRV1-style, negative): the IBA does not land above its donors — 14
of 16 donors hold the term itself — so no downward MODIFY is warranted.

Resolver gotchas measured here (each cost a cycle)

  • QuickGO's geneProductId rejects MOD ids outright (HTTP 400, "contains invalid
    values") for MGI:MGI:109249, MGI:109249, FB:FBgn0003137, FBgn0003137,
    WB:WBGene00003242, RGD:621241. It indexes UniProtKB only. A donor's own evidence
    can therefore only be queried through its resolved UniProt accession. The first
    version of the script silently printed "NONE" for every MOD token — a silent
    degradation reading as a null result    . Fixed, and unqueryable is now its own state.
  • UniProt does not index the WBGene id. xref:wormbase-WBGene00003242 returns [];
    UniProt keys WormBase xrefs on the CDS name (C37C3.6). Resolved via the WormBase
    REST (WBGene00003242 → mig-6) then a UniProt symbol search.
  • That symbol search must be gene_exact:, not gene:. Fuzzy gene:mig-6 also
    returns mig-10 / mig-5 / mig-14 / mig-18 and puts the wrong entry first — the
    size=1 trap in a new guise. With gene_exact the token resolves uniquely to
    O76840 PPN1_CAEEL papilin (Swiss-Prot, 2167 aa), which holds GO:0005604 by IDA.
    Getting 16/16 via a fuzzy head match would have been the right number for the wrong
    reason.
  • 13 of 16 tokens are multi-hit (a Swiss-Prot entry plus TrEMBL fragments). All
    candidates are reported; the reviewed entry is the one queried.
  • QuickGO strips the DB prefix from WITH/FROM tokens. withFrom[].connectedXrefs[]
    returns {"db": "FB", "id": "FBgn0003137"} as two separate fields, so comparing the
    bare id set against a GOA TSV WITH/FROM string yields 17 vs 17 tokens reported as
    "not identical"     for data that is actually the same. Reassemble as db + ":" + id
    before any set comparison.

Cross-check against the concurrently reviewed sibling ADAMTSL3

Derived from QuickGO directly rather than read off the sibling branch. ADAMTSL3
(P82987) carries a GO:0031012 IBA whose WITH/FROM token set is exactly identical
to ADAMTSL5's — same node, same 17 tokens, verified as a set equality in a single
serialisation. Two independent derivations of one node agree. ADAMTSL3 also shares the
GO:0071953 TAS from PMID:23962539 and the InterPro GO:0030198 IEA, so all three of
this review's non-ACCEPT verdicts have a direct counterpart there.

3. The real PAINT defect: one node, incoherent propagation across the family

All eight human ADAMTSL/papilin proteins are in PTHR13723, and node
PTN000347317 carries four IBD annotations (GO:0031012, GO:0030198,
GO:0004222, GO:0006508). Who actually received them:

gene subfamily GO:0031012 GO:0030198 GO:0004222 GO:0006508
ADAMTSL1 SF157 – – – –
ADAMTSL2 SF147 IBA IBA NOT-IBA –
ADAMTSL3 SF169 IBA – – –
ADAMTSL4 SF144 IBA IBA – –
ADAMTSL5 SF173 IBA – – –
THSD4 SF16 IBA IBA – –
PAPLN SF281 – – – –
ADAMTS1/9/10/17 (catalytic controls) IBA IBA IBA IBA

What PAINT got right. The catalytic terms reached all four catalytic controls and
none of the seven non-catalytic members — with ADAMTSL2 carrying an explicit
NOT|enables GO:0004222 (IKR at node PTN002673039). The family's loss of catalysis
is modelled correctly. This is the counter-example to the campaign's usual finding.

What is incoherent. From a single node, GO:0031012 reached 5 of 7 and
GO:0030198 reached 3 of 7, in no biologically coherent pattern:

  • PAPLN receives neither, although three of its own orthologs are seeds for
    GO:0031012 at that very node — fly Ppn (FB:FBgn0003137), mouse Papln
    (MGI:MGI:2386139) and worm mig-6/ppn-1 (WB:WBGene00003242). Human PAPLN has
    no IBA at all (5 annotations, all IEA/TAS).
  • ADAMTSL1 receives neither and likewise has no IBA at all (4 annotations).
  • ADAMTSL3 and ADAMTSL5 receive the location term but not the process term, while
    ADAMTSL2/ADAMTSL4/THSD4 receive both.

Consequence for this review, stated as a limit rather than as support: ADAMTSL5's
absence of a GO:0030198 IBA cannot be read as a deliberate curatorial judgement
that the process does not apply. The same gap left ADAMTSL1 and PAPLN with nothing at
all, so it is more likely a propagation failure than a decision. The case against
GO:0030198 for ADAMTSL5 therefore rests on §4 and §5 only.

4. GO:0030198 comes from a family signature that bundles proteases with non-proteases

InterPro:IPR013273 is named, literally, "ADAMTS/ADAMTS-like" (family, 26,580
proteins) and carries exactly one GO mapping: GO:0030198 extracellular matrix organization. The signature therefore lumps the catalytic ADAMTS proteases — for which
ECM organization is a direct consequence of ECM proteolysis — together with the
non-catalytic ADAMTS-like proteins, which have no such activity (§1).

This is the campaign's familiar shape one aspect over: not "a domain's name is not an
activity" (MF), but a family signature too broad to carry a process (BP).

5. The only direct test of ECM organization by ADAMTSL5 was negative

From PMID:23010571 (full text): after showing colocalisation with fibrillin
microfibrils, the authors report

"However, comparison of microfibril density in fBNL cell cultures grown in the
presence of ADAMTSL5 or vector conditioned medium, did not identify a consistent
difference (data not shown)."

and the abstract states colocalisation occurred "but without discernible effect on
microfibril assembly". Direct binding to fibronectin was also not supported.

Weight of this negative, stated honestly: it is a "data not shown" result from a
single exogenous-protein assay. It argues that no role in ECM/microfibril organization
has been demonstrated; it does not refute one. That supports
MARK_AS_OVER_ANNOTATED, and is not strong enough to earn REMOVE.

6. The three GO:0005515 rows are one Y2H screen, on three hub proteins

All three rows come from PMID:32296183 (HuRI). IntAct shows each partner logged under
three sub-methods of the same experiment — two hybrid array + two hybrid prey pooling approach + validated two hybrid, MI-score 0.56. UniProt's NbExp=3 is
therefore one screen counted three ways, replicating the ACRV1 finding on a second
gene.

Distinct-partner counts (derived as an entity set; IntAct records are not partners):

protein IntAct records distinct partners localisation length
CYSRT1 (A8MQ03) 1670 517 cornified envelope 144 aa
KRTAP5-9 (P26371) 842 213 intracellular hair-keratin matrix 169 aa
FHL5 (Q5TD97) 316 108 nucleus (spermatid nuclei) 284 aa
ADAMTSL5 (Q6ZMM2) 22 12 secreted / ECM 481 aa

Each partner is topologically incompatible with ADAMTSL5, a signal-peptide
(1–42) secreted ECM glycoprotein: it does not enter the cytoplasm, nucleus, or cornified
envelope. Decided per partner, all three come out the same way — unreplicated
screen noise → MARK_AS_OVER_ANNOTATED.

All three partner accessions resolve to reviewed canonical Swiss-Prot entries with
matching lengths — no TrEMBL/ORFeome substitution of the ACRV1 kind. Reported as a
negative.

The gene's real binding partners, FBN1 and FBN2, come from the same IntAct record
by pull down from PMID:23010571, and are captured by GO:0050436 microfibril binding.

7. GO:0071953 elastic fiber is a bulk review-based assignment

Projection test by reference (entities derived as a set, not from the annotation total):

reference annotations distinct entities per-term spread
PMID:23010571 (primary paper) 6 2 GO:0031012 2; the rest 1 each
PMID:23962539 (elastic-fibre review) 66 62 GO:0071953 41 entities; GO:0001527 15

PMID:23010571 annotates only ADAMTSL5 and its mouse ortholog — gene-specific
curation, not a projection. Negative result, reported.

PMID:23962539 is a review (publication_type: REVIEW, full_text_available: false) used as the TAS source for 62 entities, assigning GO:0071953 to 41 of
them. Its abstract never mentions ADAMTSL5. Notably the same curation from the same
review assigned the more specific GO:0001527 microfibril to 15 other proteins —
including THSD4/ADAMTSL6 — so the specific term was demonstrably available and was not
chosen for ADAMTSL5.

The specific term is missing from GOA but present in UniProt

UniProt's DR block carries GO:0001527; C:microfibril; IDA:UniProtKB, but GOA does
not have it (verified: QuickGO returns exactly 12 annotations for Q6ZMM2, matching the
GOA TSV, and GO:0001527 is not among them). So the gene's best-evidenced and most
specific localisation is absent from GOA while a weaker, broader TAS row is present.

GO:0001527 is current (not obsolete, no secondaryIds) and is a part_of child of
GO:0071953 — so replacing the TAS row with it retains the parent by closure while
gaining precision, and matches the merged ADAMTSL4 review, which independently proposed
GO:0001527 as a NEW term.

An ontology issue this surfaces

GO:0001527 microfibril is part_of GO:0071953 elastic fiber, which asserts that
every microfibril is part of an elastic fiber. That contradicts GO:0001527's own
definition — "Extracellular matrix components occurring independently or along
with elastin" — and contradicts the ciliary zonule, a fibrillin-microfibril structure
essentially devoid of elastin. Recorded in suggested_questions.

8. Checks run that came back negative (recorded so they are not re-run blind)

  • Retraction / erratum / expression-of-concern: all 9 PMIDs relied on
    (23010571, 23962539, 32296183, 26621454, 27857980, 28482118, 33197513, 37330172,
    38524140) checked via PubMed publication types and each record's own
    CommentsCorrections (a Publisher Correction is invisible to a pubtype query).
    None flagged. PMID:26621454 has two Comment in entries — commentaries, not
    corrections.
  • Partner accession integrity: all three IPI partners are reviewed canonical
    Swiss-Prot entries at their canonical lengths. No dead or truncated accessions.
  • IBA-above-donor precision defect (ACRV1 shape): absent — the IBA sits at the term
    14 of 16 donors hold.
  • Reference projection for the primary paper: absent — 2 entities.
  • Dead-accession check: every WITH/FROM UniProt lookup returns a named entry; the
    script raises rather than reporting a silent zero.

📄 View Raw YAML

 
id: Q6ZMM2
gene_symbol: ADAMTSL5
product_type: PROTEIN
status: COMPLETE
taxon:
  id: NCBITaxon:9606
  label: Homo sapiens
description: >-
  ADAMTSL5 (ADAMTS-like protein 5; also known as THSD6) is a small secreted,
  N-glycosylated extracellular matrix glycoprotein of the ADAMTS superfamily. At 481
  amino acids it is by far the smallest family member, and it is built from a signal
  peptide, a single thrombospondin type-1 (TSP1) repeat, a cysteine-rich module, a
  spacer module, and a C-terminal netrin-like (NTR) module joined to the spacer by a
  proline-rich disordered segment. Unlike the ADAMTS metalloproteinases it resembles,
  ADAMTSL5 has no metalloprotease or disintegrin-like domain and no proteolytic
  activity; it is a matrix-binding glycoprotein rather than an enzyme. The secreted
  protein binds heparin through its NTR module in a salt-sensitive, ionic interaction,
  and binds both fibrillin-1 and fibrillin-2, co-localising with fibrillin microfibrils
  in the extracellular matrix deposited by cultured fibroblasts; it is the first family
  member shown to bind both fibrillins. Its distribution is peri-cellular and
  baso-lateral, and part of the pool is proteolytically processed to release the
  C-terminal NTR-containing fragment. In mouse development the protein is broadly
  expressed, most prominently in musculoskeletal tissues including skeletal muscle,
  cartilage and bone, and in many epithelia. Whether ADAMTSL5 modulates microfibril
  biology is unresolved: adding it to fibroblast cultures changed neither microfibril
  density nor assembly. Beyond the matrix, a nine-residue peptide from its TSP1 repeat
  is presented by HLA-C*06:02 on melanocytes and acts as a CD8+ T cell autoantigen in
  psoriasis, and the gene is epigenetically activated in hepatocellular carcinoma,
  where it supports tumorigenicity in MET-sensitised hepatocytes.
alternative_products:
- name: '1'
  id: Q6ZMM2-1
- name: '2'
  id: Q6ZMM2-2
  sequence_note: VSP_053358
existing_annotations:

- term:
    id: GO:0031012
    label: extracellular matrix
  evidence_type: IBA
  original_reference_id: GO_REF:0000033
  qualifier: is_active_in
  review:
    summary: >-
      Phylogenetic propagation from PANTHER node PTN000347317, and an unusually
      well-founded one: all 16 protein donors in the WITH/FROM field carry their own
      experimental (IDA/HDA) annotation to this term or a descendant.
    action: ACCEPT
    reason: >-
      Every WITH/FROM token was resolved and each donor's own evidence for the
      propagated term was queried. The GOA WITH/FROM field matches the cached PAINT
      seed list for this node exactly (17 tokens = 16 protein donors + the node id).
      All 16 protein donors hold their own IDA or HDA to GO:0031012 or a descendant,
      so any claim that the sources carry only family-level inference is contradicted
      by measurement. The target itself also holds an independent IDA to this term
      from PMID:23010571, and ADAMTSL5 is a signal-peptide secreted protein
      demonstrated to reside in fibroblast extracellular matrix. No specificity
      upgrade is warranted: the donor set is heterogeneous, spanning four distinct
      locations (extracellular matrix 14, basement membrane 4, interstitial matrix 1,
      microfibril 1), so GO:0031012 is the correct least common ancestor and refining
      it would mean arbitrarily preferring one donor's compartment. Separately checked
      and negative - the propagation does not land above its donors (14 of 16 hold
      this exact term), so no downward MODIFY applies.
    supported_by:
    - reference_id: PMID:23010571
      supporting_text: "co-localized with fibrillin microfibrils in the extracellular"
    - reference_id: file:genes/human/ADAMTSL5/ADAMTSL5-bioinformatics/RESULTS.md
      supporting_text: "All 16 protein donors carry their own experimental (IDA/HDA) annotation"
    propagation_review:
      root_cause: NO_FAILURE_CORE
      source_entities:
      - source_id: FB:FBgn0003137
        source_label: Ppn (D. melanogaster)
        source_status: SUPPORTS_TRANSFER
        comment: >-
          Swiss-Prot Q868Z9; carries its own IDA to GO:0005604.
      - source_id: FB:FBgn0032252
        source_label: loh (D. melanogaster)
        source_status: SUPPORTS_TRANSFER
        comment: >-
          TrEMBL Q9VKV3; carries its own IDA to GO:0031012.
      - source_id: MGI:MGI:109249
        source_label: Adamts1 (M. musculus)
        source_status: SUPPORTS_TRANSFER
        comment: >-
          Swiss-Prot P97857; carries its own IDA to GO:0005604, GO:0031012.
      - source_id: MGI:MGI:1347346
        source_label: Adamts7 (M. musculus)
        source_status: SUPPORTS_TRANSFER
        comment: >-
          Swiss-Prot Q68SA9; carries its own IDA to GO:0031012.
      - source_id: MGI:MGI:1347356
        source_label: Adamts2 (M. musculus)
        source_status: SUPPORTS_TRANSFER
        comment: >-
          Swiss-Prot Q8C9W3; carries its own IDA to GO:0031012.
      - source_id: MGI:MGI:1913798
        source_label: Adamtsl5 (M. musculus)
        source_status: SUPPORTS_TRANSFER
        comment: >-
          TrEMBL D3Z689; carries its own HDA/IDA to GO:0031012.
      - source_id: MGI:MGI:1925044
        source_label: Adamtsl2 (M. musculus)
        source_status: SUPPORTS_TRANSFER
        comment: >-
          Swiss-Prot Q7TSK7; carries its own IDA to GO:0031012.
      - source_id: MGI:MGI:2386139
        source_label: Papln (M. musculus)
        source_status: SUPPORTS_TRANSFER
        comment: >-
          Swiss-Prot Q9EPX2; carries its own HDA/IDA to GO:0005604, GO:0031012.
      - source_id: MGI:MGI:2389008
        source_label: Adamtsl4 (M. musculus)
        source_status: SUPPORTS_TRANSFER
        comment: >-
          Swiss-Prot Q80T21; carries its own IDA to GO:0005614, GO:0031012.
      - source_id: MGI:MGI:2672033
        source_label: Thsd4 (M. musculus)
        source_status: SUPPORTS_TRANSFER
        comment: >-
          Swiss-Prot Q3UTY6; carries its own IDA to GO:0001527, GO:0031012.
      - source_id: PANTHER:PTN000347317
        source_label: PANTHER tree node (ADAMTS/ADAMTS-like family)
        source_status: SUPPORTS_TRANSFER
        comment: >-
          Internal PANTHER tree node, not a protein. The node carries four IBD terms but only GO:0031012 reached ADAMTSL5.
      - source_id: RGD:621241
        source_label: Adamts1 (R. norvegicus)
        source_status: SUPPORTS_TRANSFER
        comment: >-
          Swiss-Prot Q9WUQ1; carries its own IDA to GO:0031012.
      - source_id: UniProtKB:Q6ZMM2
        source_label: ADAMTSL5 (H. sapiens) - the target itself
        source_status: CIRCULAR_OR_REDUNDANT
        comment: >-
          Self-referential seed. Valid: it records a PAINT curator judging this function core for the gene. ADAMTSL5 independently holds its own IDA to GO:0031012 from PMID:23010571.
      - source_id: UniProtKB:Q8TE56
        source_label: ADAMTS17 (H. sapiens)
        source_status: SUPPORTS_TRANSFER
        comment: >-
          Swiss-Prot Q8TE56; carries its own IDA to GO:0031012.
      - source_id: UniProtKB:Q9H324
        source_label: ADAMTS10 (H. sapiens)
        source_status: SUPPORTS_TRANSFER
        comment: >-
          Swiss-Prot Q9H324; carries its own IDA to GO:0031012.
      - source_id: UniProtKB:Q9P2N4
        source_label: ADAMTS9 (H. sapiens)
        source_status: SUPPORTS_TRANSFER
        comment: >-
          Swiss-Prot Q9P2N4; carries its own IDA to GO:0031012.
      - source_id: WB:WBGene00003242
        source_label: mig-6 (C. elegans)
        source_status: SUPPORTS_TRANSFER
        comment: >-
          Swiss-Prot O76840; carries its own IDA to GO:0005604.

- term:
    id: GO:0005576
    label: extracellular region
  evidence_type: IEA
  original_reference_id: GO_REF:0000044
  qualifier: located_in
  review:
    summary: >-
      UniProt Subcellular Location keyword mapping (SL-0243, Secreted). Correct and
      independently confirmed experimentally.
    action: ACCEPT
    reason: >-
      ADAMTSL5 has a predicted signal peptide (residues 1-42) and a mature chain
      (43-481), and secretion was demonstrated directly: recombinant protein is
      recovered from the conditioned medium of three transfected cell lines. The
      SubCell-to-GO mapping is doing exactly what it should here, and the same term is
      independently supported by an IDA row on this gene.
    supported_by:
    - reference_id: PMID:23010571
      supporting_text: "Recombinant ADAMTSL5 is a secreted, N-glycosylated 60kDa"

- term:
    id: GO:0030198
    label: extracellular matrix organization
  evidence_type: IEA
  original_reference_id: GO_REF:0000002
  qualifier: involved_in
  review:
    summary: >-
      InterPro2GO transfer from IPR013273, a family signature that bundles the
      catalytic ADAMTS proteases together with the non-catalytic ADAMTS-like proteins.
      For the proteases, extracellular matrix organization follows from ECM
      proteolysis; ADAMTSL5 has no proteolytic activity, and the one direct test of
      matrix organization by ADAMTSL5 was negative.
    action: MARK_AS_OVER_ANNOTATED
    reason: >-
      Two independent legs. First, the source signature is too broad to carry this
      process. InterPro:IPR013273 is named "ADAMTS/ADAMTS-like", covers 26,580
      proteins, and carries exactly one GO mapping - this term. It matches ADAMTSL5
      legitimately, but the ECM-organizing role of the family's catalytic members is a
      consequence of their metalloprotease activity, which ADAMTSL5 does not have:
      UniProt flags the missing catalytic domain in a CAUTION line, the primary paper
      states the ADAMTSLs have no proteolytic activity, and a sequence scan finds no
      HExxHxxGxxHD zinc-binding signature - indeed no HExxH substring at all - in the
      481-residue sequence, with no peptidase signature among its InterPro matches.
      This is the familiar "a domain's name is not an activity" failure moved one
      aspect over, into a family signature too broad to carry a process. Second, the
      only direct experimental test was negative: adding ADAMTSL5 to fibroblast
      cultures produced no consistent difference in microfibril density and no
      discernible effect on microfibril assembly, and direct fibronectin binding was
      not supported either. Not REMOVE, deliberately: that negative is a "data not
      shown" result from a single exogenous-protein assay, so it shows no role has
      been demonstrated rather than refuting one, and UniProt itself only claims the
      protein may play a role in modulation of fibrillin microfibrils. A third
      argument was available and deliberately not used - ADAMTSL5 did not receive this
      term by IBA while ADAMTSL2, ADAMTSL4 and THSD4 did from the same node - because
      that node's propagation is incoherent family-wide (ADAMTSL1 and PAPLN received
      nothing at all), so the absence is more likely a coverage gap than a curatorial
      judgement.
    supported_by:
    - reference_id: PMID:23010571
      supporting_text: "In contrast to ADAMTS proteases, ADAMTSLs lack a catalytic domain"
    - reference_id: PMID:23010571
      supporting_text: "did not identify a consistent difference"
    - reference_id: file:genes/human/ADAMTSL5/ADAMTSL5-uniprot.txt
      supporting_text: "lacks the metalloprotease and disintegrin-like domains which are"
    propagation_review:
      root_cause: PROPAGATION_BAD
      failure_modes:
      - FUNCTIONAL_DIVERGENCE
      - PSEUDO_OR_SUBACTIVITY_LOSS
      source_entities:
      - source_id: InterPro:IPR013273
        source_label: ADAMTS/ADAMTS-like (family signature, 26580 proteins)
        source_status: SUPPORTS_SOURCE_BUT_NOT_TARGET
        comment: >-
          Sole GO mapping is GO:0030198. The signature spans both catalytic ADAMTS
          proteases and non-catalytic ADAMTS-like proteins, so the mapping is sound for
          the protease members but does not transfer to a member that lost catalysis.

- term:
    id: GO:0005515
    label: protein binding
  evidence_type: IPI
  original_reference_id: PMID:32296183
  qualifier: enables
  review:
    summary: >-
      Partner CYSRT1 (A8MQ03), from the HuRI yeast two-hybrid screen. An extreme Y2H
      hub (517 distinct IntAct partners) that localises to the cornified envelope,
      topologically inaccessible to a secreted ECM protein.
    action: MARK_AS_OVER_ANNOTATED
    reason: >-
      Assessed per partner, not per gene. CYSRT1 has 517 distinct interaction partners
      in IntAct across 1670 records, against 12 for ADAMTSL5 - the signature of a
      promiscuous two-hybrid prey rather than a specific partner. Its only annotated
      location is the cornified envelope, an intracellular keratinocyte structure that
      a signal-peptide secreted protein (ADAMTSL5 signal peptide 1-42) never
      encounters. The interaction is also unreplicated: IntAct logs it under three
      sub-methods of the same experiment - two hybrid array, two hybrid prey pooling
      approach, and validated two hybrid - all at MI-score 0.56, so the NbExp=3
      recorded in UniProt is one screen counted three ways, not three independent
      observations. No orthogonal assay exists. The accession resolves to the reviewed
      canonical Swiss-Prot entry at its canonical length (144 aa), so this is genuine
      screen noise rather than a mis-identified partner. Not REMOVE: the assay was
      performed and curated correctly by IntAct, it simply does not support a
      biologically meaningful binding function.
    supported_by:
    - reference_id: file:genes/human/ADAMTSL5/ADAMTSL5-bioinformatics/RESULTS.md
      supporting_text: "CYSRT1 (A8MQ03) | 1670 | **517** | cornified envelope | 144 aa |"

- term:
    id: GO:0005515
    label: protein binding
  evidence_type: IPI
  original_reference_id: PMID:32296183
  qualifier: enables
  review:
    summary: >-
      Partner KRTAP5-9 (P26371), from the same HuRI screen. A keratin-associated
      protein of the intracellular hair-keratin matrix, and a Y2H hub with 213 distinct
      partners.
    action: MARK_AS_OVER_ANNOTATED
    reason: >-
      Assessed per partner. KRTAP5-9 is an ultrahigh-sulfur keratin-associated protein
      that assembles into the intracellular keratin matrix of the hair cuticle; it has
      no secretory pathway and cannot meet a secreted ECM glycoprotein in vivo. It
      carries 213 distinct IntAct partners across 842 records, again characteristic of
      a cysteine-rich, sticky two-hybrid prey. As with the other two rows the
      interaction is logged under three sub-methods of a single experiment at MI-score
      0.56, so it is unreplicated. The accession is the reviewed canonical Swiss-Prot
      entry at its canonical length (169 aa). Not REMOVE, for the same reason as the
      CYSRT1 row: the annotation records a real screen result, but the bare term
      conveys no function and the partner is not credible.
    supported_by:
    - reference_id: file:genes/human/ADAMTSL5/ADAMTSL5-bioinformatics/RESULTS.md
      supporting_text: "KRTAP5-9 (P26371) | 842 | **213** | intracellular hair-keratin matrix | 169 aa |"

- term:
    id: GO:0005515
    label: protein binding
  evidence_type: IPI
  original_reference_id: PMID:32296183
  qualifier: enables
  review:
    summary: >-
      Partner FHL5 (Q5TD97), from the same HuRI screen. A nuclear LIM-domain
      transcriptional co-activator of spermatids, with 108 distinct IntAct partners.
    action: MARK_AS_OVER_ANNOTATED
    reason: >-
      Assessed per partner. FHL5 is a four-and-a-half LIM domain protein annotated to
      the nucleus, specifically the nuclei of round and elongated spermatids, where it
      co-activates CREM. A secreted extracellular matrix glycoprotein and a nuclear
      transcriptional co-activator occupy mutually inaccessible compartments. FHL5 is
      also a substantial two-hybrid hub (108 distinct partners across 316 records), and
      like the other two partners it is supported only by three sub-methods of one
      experiment at MI-score 0.56 with no orthogonal validation. The accession is the
      reviewed canonical Swiss-Prot entry at its canonical length (284 aa). The three
      partners together cohere as a set of cysteine-rich, aggregation-prone Y2H preys
      rather than as any biology of ADAMTSL5.
    supported_by:
    - reference_id: file:genes/human/ADAMTSL5/ADAMTSL5-bioinformatics/RESULTS.md
      supporting_text: "FHL5 (Q5TD97) | 316 | **108** | nucleus (spermatid nuclei) | 284 aa |"

- term:
    id: GO:0071953
    label: elastic fiber
  evidence_type: TAS
  original_reference_id: PMID:23962539
  qualifier: located_in
  review:
    summary: >-
      Bulk assignment from a review article used as the TAS source for 62 distinct
      entities. The gene's own experimental evidence supports the more specific
      GO:0001527 microfibril, which UniProt already records by IDA but GOA is missing.
    action: MODIFY
    reason: >-
      PMID:23962539 is a review whose abstract never mentions ADAMTSL5, and QuickGO
      shows it is used as the reference for 66 annotations over 62 distinct entities,
      41 of them receiving this same term - a bulk curation pass over the elastic fibre
      literature rather than a gene-specific traceable statement. Tellingly, the same
      curation from the same review assigned the more specific GO:0001527 microfibril
      to 15 other proteins including THSD4/ADAMTSL6, so the specific term was available
      and simply not chosen here. Meanwhile ADAMTSL5's own primary data demonstrates
      association with fibrillin microfibrils specifically, and UniProt already carries
      GO:0001527 microfibril by IDA for this protein - a term GOA does not have
      (QuickGO returns exactly 12 annotations for Q6ZMM2 and GO:0001527 is not among
      them). GO:0001527 is current, is not obsolete, and is a part_of child of
      GO:0071953, so replacing the review-derived parent with the evidence-backed child
      retains the parent by closure while gaining precision and aligns GOA with
      UniProt. The merged ADAMTSL4 review independently proposed GO:0001527 for its
      gene, which is useful convergent support. Note that elastic fibre is a poor fit
      on tissue grounds too: ADAMTSL5's strongest expression is in cartilage and bone,
      which are not elastic-fibre tissues.
    proposed_replacement_terms:
    - id: GO:0001527
      label: microfibril
    supported_by:
    - reference_id: PMID:23010571
      supporting_text: "co-localized with fibrillin microfibrils in the extracellular"
    - reference_id: file:genes/human/ADAMTSL5/ADAMTSL5-uniprot.txt
      supporting_text: "DR   GO; GO:0001527; C:microfibril; IDA:UniProtKB."

- term:
    id: GO:0050436
    label: microfibril binding
  evidence_type: IDA
  original_reference_id: PMID:23010571
  qualifier: enables
  review:
    summary: >-
      Core molecular function. Direct affinity co-isolation showed binding to both
      fibrillin-1 and fibrillin-2 and to assembled fibrillin microfibrils - ADAMTSL5 is
      the first family member shown to bind both fibrillins.
    action: ACCEPT
    reason: >-
      Well supported by two complementary assays in the same paper: myc-tagged ADAMTSL5
      co-isolated the N-terminal half of fibrillin-1 and both halves of fibrillin-2,
      with empty-vector controls showing no bead binding, and exogenous ADAMTSL5 added
      to fibroblast cultures decorated assembled fibrillin microfibrils. This is the
      most informative molecular function on the record and there is no more specific
      term available - GO has no fibrillin binding term (confirmed via OLS), and
      GO:0050436 is_a GO:0050840 extracellular matrix binding, so this is the
      appropriate level. Retained as a core function.
    supported_by:
    - reference_id: PMID:23010571
      supporting_text: "Taken together, the findings are consistent with specific binding
        of ADAMTSL5 to fibrillin-1 and fibrillin-2 and to their macromolecular assemblies,
        i.e., fibrillin microfibrils."

- term:
    id: GO:0031012
    label: extracellular matrix
  evidence_type: IDA
  original_reference_id: PMID:23010571
  qualifier: located_in
  review:
    summary: >-
      Direct demonstration that secreted ADAMTSL5 deposits into fibroblast
      extracellular matrix, with a non-permeabilised staining control establishing the
      signal is extracellular. MGI-assigned.
    action: ACCEPT
    reason: >-
      The localisation was shown by immunostaining of fibroblast cultures exposed to
      ADAMTSL5, with two controls that make the call solid: empty-vector conditioned
      medium gave no signal, and staining without prior fixation reproduced the
      pattern, establishing that the decorated structures are extracellular rather
      than intracellular. UniProt records the same conclusion as an experimental
      subcellular location. One caveat recorded but not acted on: the assay used
      exogenous recombinant protein rather than protein at native levels, which by a
      strict reading leans toward IMP rather than IDA. I have not overturned the
      curator's call - for a secreted protein this is the standard assay and the
      non-permeabilised control addresses the main artefact.
    supported_by:
    - reference_id: PMID:23010571
      supporting_text: "suggesting that the stained structures were extracellular and corresponded to fibrillin-1 containing microfibrils in ECM"

- term:
    id: GO:0005576
    label: extracellular region
  evidence_type: IDA
  original_reference_id: PMID:23010571
  qualifier: located_in
  review:
    summary: >-
      Secretion demonstrated directly - recombinant ADAMTSL5 is recovered from the
      conditioned medium of transfected HEK293F, COS-1 and CHO-K1 cells. MGI-assigned.
    action: ACCEPT
    reason: >-
      Straightforward and correct, though it is the least informative of the three
      location rows since it is a parent of the extracellular matrix annotation the
      same paper supports. Secretion was reproduced in three independent cell lines and
      is consistent with the predicted signal peptide (1-42). Kept as non-redundant
      with GO:0031012 because it additionally captures the soluble medium pool, which
      matters for this protein: part of it is proteolytically processed to release the
      C-terminal NTR fragment into the medium.
    supported_by:
    - reference_id: PMID:23010571
      supporting_text: "in the medium of transfected cells"

- term:
    id: GO:0008201
    label: heparin binding
  evidence_type: IDA
  original_reference_id: PMID:23010571
  qualifier: enables
  review:
    summary: >-
      Core molecular function, mapped to a specific module. Heparin-agarose affinity
      chromatography showed salt-sensitive ionic binding, and only the C-terminal
      NTR-containing fragment was retained.
    action: ACCEPT
    reason: >-
      Directly demonstrated and unusually well localised within the protein. ADAMTSL5
      was depleted from medium by heparin-agarose, was not eluted by 0.15 or 0.3 M NaCl
      but was substantially eluted by 0.5 and 1 M NaCl, establishing an ionic
      interaction; and of the two fragments present in conditioned medium only the
      C-terminal NTR-containing one was retained, while the 33 kDa N-terminal fragment
      flowed through. This is consistent with the general heparin-binding property of
      NTR modules and gives the protein a plausible mechanism for its observed
      cell-surface and peri-cellular distribution via heparan sulfate proteoglycans.
      GO:0008201 is_a GO:0005539 glycosaminoglycan binding, which is the right level:
      the ligand assayed was heparin specifically. Retained as a core function.
    supported_by:
    - reference_id: PMID:23010571
      supporting_text: "only the C-terminal fragment containing the NTR-module was retained by the heparin matrix"
    - reference_id: PMID:23010571
      supporting_text: "indicative of an ionic interaction between ADAMTSL5 and heparin"

- term:
    id: GO:0031012
    label: extracellular matrix
  evidence_type: IDA
  original_reference_id: PMID:23010571
  qualifier: located_in
  review:
    summary: >-
      Duplicate of the MGI-assigned extracellular matrix IDA row, independently
      asserted by UniProt from the same paper. Same verdict.
    action: ACCEPT
    reason: >-
      GOA carries this localisation twice, once assigned by MGI and once by UniProt,
      both citing PMID:23010571. Both are reviewed here separately so that every GOA
      row receives a verdict, and both are accepted on the same evidence. The
      duplication is a GOA bookkeeping artefact of two groups curating the same paper,
      not a second independent observation - worth noting so the row count is not read
      as two lines of support. The paper additionally localises the protein to the
      peri-cellular and baso-lateral matrix.
    supported_by:
    - reference_id: PMID:23010571
      supporting_text: "These observations strongly suggest that ADAMTSL5 is associated with the peri-cellular and sub-cellular ECM of cells"

# NEW proposal -- not a GOA row. The 12 rows above map 1:1 onto the 12 rows of
# ADAMTSL5-goa.tsv; this 13th entry is the annotation GOA is missing.
- term:
    id: GO:0001527
    label: microfibril
  evidence_type: IDA
  original_reference_id: PMID:23010571
  qualifier: located_in
  review:
    summary: >-
      Proposed new annotation. This is the gene's most specific and best-evidenced
      localisation, it is already in UniProt by IDA, and GOA does not have it at all.
    action: NEW
    reason: >-
      GOA is missing the one localisation ADAMTSL5's own primary data establishes.
      UniProt's cross-reference block carries GO:0001527 C:microfibril IDA:UniProtKB,
      but QuickGO returns exactly 12 annotations for Q6ZMM2 and this term is not among
      them. The supporting experiment is direct: exogenous ADAMTSL5 added to fibroblast
      cultures decorated assembled fibrillin microfibrils, with an empty-vector control
      giving no signal and a non-permeabilised staining control establishing that the
      decorated structures are extracellular. Proposed as an additional row rather than
      relying solely on the MODIFY of the GO:0071953 TAS row, because that MODIFY alone
      would leave the specific term carrying the weak review-derived provenance of
      PMID:23962539, whereas the evidence that actually supports microfibril
      localisation is an IDA from PMID:23010571. Adding it here gives the UniProt/GOA
      discrepancy raised in suggested_questions a concrete target, and matches how the
      merged ADAMTSL4 review handled the same term.
    supported_by:
    - reference_id: PMID:23010571
      supporting_text: "the exogenous protein co-localized with fibrillin-1 in microfibrils"
    - reference_id: file:genes/human/ADAMTSL5/ADAMTSL5-uniprot.txt
      supporting_text: "DR   GO; GO:0001527; C:microfibril; IDA:UniProtKB."

core_functions:
- description: >-
    Binds fibrillin-1 and fibrillin-2 and decorates assembled fibrillin microfibrils in
    the extracellular matrix, acting as a microfibril-associated accessory glycoprotein
    rather than as a structural or enzymatic component of the microfibril itself.
  molecular_function:
    id: GO:0050436
    label: microfibril binding
  locations:
  - id: GO:0001527
    label: microfibril
  - id: GO:0031012
    label: extracellular matrix
  supported_by:
  - reference_id: PMID:23010571
    supporting_text: "ADAMTSL5 is the first family member shown to bind both fibrillin-1"
- description: >-
    Binds heparin and, by extension, cell-surface heparan sulfate proteoglycans through
    its C-terminal netrin-like module in a salt-sensitive ionic interaction, which
    plausibly anchors the secreted protein in the peri-cellular matrix.
  molecular_function:
    id: GO:0008201
    label: heparin binding
  locations:
  - id: GO:0005576
    label: extracellular region
  supported_by:
  - reference_id: PMID:23010571
    supporting_text: "Affinity-chromatography demonstrated heparin-binding of ADAMTSL5"

suggested_questions:
- question: >-
    PAINT recommendation, stated once for the whole family rather than per gene. Node
    PTN000347317 in PTHR13723 carries four IBD annotations, but their propagation to
    the eight human ADAMTSL/papilin members is incoherent: GO:0031012 reached 5 of 7
    non-catalytic members (missing ADAMTSL1 and PAPLN) and GO:0030198 reached only 3 of
    7 (missing ADAMTSL1, ADAMTSL3, ADAMTSL5 and PAPLN), while all four catalytic ADAMTS
    controls received all four terms. The sharpest case is human PAPLN, which received
    no IBA at all even though three of its own orthologs - Drosophila Ppn
    (FB:FBgn0003137), mouse Papln (MGI:MGI:2386139) and C. elegans mig-6/ppn-1
    (WB:WBGene00003242) - are seeds for GO:0031012 at that very node. Human ADAMTSL1
    likewise has no IBA at all despite being in the family (SF157). Is this a
    propagation-coverage gap rather than a set of deliberate curatorial exclusions, and
    can the node's terms be re-propagated across the clade?
  experts: []
- question: >-
    Related PAINT point, in the opposite direction. The family's loss of catalysis is
    modelled correctly - GO:0004222 and GO:0006508 reached all four catalytic ADAMTS
    controls and none of the seven ADAMTS-like members - but only ADAMTSL2 records this
    explicitly, via a NOT-qualified enables GO:0004222 IBA (IKR at node PTN002673039).
    For ADAMTSL1, ADAMTSL3, ADAMTSL4, ADAMTSL5, THSD4 and PAPLN the absence of
    catalytic activity is implicit, i.e. simply nothing propagated. Would placing the
    negation at the node ancestral to the whole non-catalytic clade make this
    informative loss explicit for all of them?
  experts: []
- question: >-
    Ontology issue. GO:0001527 microfibril is a part_of child of GO:0071953 elastic
    fiber, which asserts that every microfibril is part of an elastic fiber. This
    contradicts GO:0001527's own definition, which describes microfibrils as
    extracellular matrix components occurring independently or along with elastin, and
    contradicts the ciliary zonule - a fibrillin microfibril structure essentially
    devoid of elastin, and the reason FBN1 mutations cause ectopia lentis. Should the
    part_of be relaxed, or GO:0001527 re-parented under a fibrillin-microfibril concept
    independent of elastic fibre?
  experts: []
- question: >-
    UniProt/GOA discrepancy to reconcile. UniProt's cross-reference block carries
    GO:0001527 C:microfibril IDA:UniProtKB for ADAMTSL5, but GOA does not contain that
    annotation - QuickGO returns exactly 12 annotations for Q6ZMM2 and GO:0001527 is
    not among them. The gene's best-evidenced and most specific localisation is
    therefore missing from GOA, while a weaker review-derived TAS row to the broader
    parent GO:0071953 is present. Which record is stale?
  experts: []
- question: >-
    Does ADAMTSL5 have any demonstrable effect on fibrillin microfibril biology? The
    only test to date found no consistent difference in microfibril density and no
    discernible effect on assembly, in a single exogenous-protein assay reported as
    data not shown. Given that the protein binds both fibrillins and is proteolytically
    processed to release its NTR module, a loss-of-function experiment is needed before
    any extracellular matrix organization annotation is justified.
  experts:
  - Suneel S. Apte
  - Dieter P. Reinhardt
- question: >-
    What is the function, if any, of the released C-terminal NTR fragment? UniProt
    records that ADAMTSL5 is proteolytically cleaved to release an NTR-containing
    fragment, this fragment carries the heparin-binding activity, and it is detectable
    in mouse skeletal muscle, testis, kidney and heart. Is the cleavage regulated, and
    does the fragment have activity distinct from the full-length protein?
  experts:
  - Suneel S. Apte
  - Hannes L. Bader

suggested_experiments:
- hypothesis: >-
    If ADAMTSL5 modulates fibrillin microfibril assembly in vivo, loss of the gene will
    perturb microfibril architecture in the tissues where the protein is most abundant.
  description: >-
    Generate Adamtsl5-null mice and examine fibrillin microfibril architecture in
    skeletal muscle, cartilage, bone and the ciliary zonule by immunofluorescence and
    transmission electron microscopy. This is the loss-of-function test the field
    lacks, and it would settle whether any extracellular matrix organization annotation
    is warranted. Compare against the Adamtsl2 and Thsd4 nulls, whose microfibril
    phenotypes are established.
  experiment_type: Mouse knockout with tissue ultrastructure
- hypothesis: >-
    ADAMTSL5 binds a discrete N-terminal region shared by fibrillin-1 and fibrillin-2,
    and competes there with other microfibril-associated ADAMTSL proteins.
  description: >-
    Map the fibrillin-binding site. The primary paper showed ADAMTSL5 binds the
    N-terminal half of fibrillin-1 and both halves of fibrillin-2, with the N-terminal
    half binding more strongly. Narrow this with recombinant fibrillin subfragments and
    surface plasmon resonance to obtain an affinity and a domain-level site, then test
    competition with ADAMTSL2, ADAMTSL4 and THSD4.
  experiment_type: Surface plasmon resonance with recombinant fragments
- hypothesis: >-
    Release of the C-terminal NTR module is a regulated processing event that changes
    the protein's matrix-binding behaviour.
  description: >-
    Identify the protease that releases the NTR module and test whether the cleavage is
    regulated. Combine N-terminal sequencing or terminal-amine labelling of the
    released fragment with a protease-inhibitor panel in transfected cells, then ask
    whether the free NTR fragment and full-length ADAMTSL5 differ in heparin affinity
    and in microfibril decoration.
  experiment_type: Protease identification and N-terminomics
- hypothesis: >-
    Peri-cellular retention of secreted ADAMTSL5 depends on the NTR-heparan sulfate
    interaction rather than on protein-protein contacts with the matrix.
  description: >-
    Compare ADAMTSL5 deposition on wild-type CHO cells with the
    glycosaminoglycan-deficient pgsA-745 and pgsD-677 mutants, and after heparinase III
    treatment. If the NTR-heparin interaction is what anchors the protein, retention
    should be lost in the mutants while secretion itself is unaffected.
  experiment_type: Cell-surface retention in glycosaminoglycan-deficient mutants
- hypothesis: >-
    The reported effect of ADAMTSL5 on receptor tyrosine kinase levels in hepatocellular
    carcinoma is a genuine extracellular, matricrine activity requiring the
    heparin-binding NTR module.
  description: >-
    Test whether purified recombinant ADAMTSL5 added to the medium reproduces the
    changes in MET, EGFR and FGFR4 seen on knockdown, and whether the effect requires
    the NTR module. This would distinguish a bona fide extracellular signal from an
    intracellular or indirect consequence of depleting the gene, and would determine
    whether any signalling GO annotation is ever justified.
  experiment_type: Recombinant protein add-back with domain deletion

references:
- id: GO_REF:0000002
  title: Gene Ontology annotation through association of InterPro records with GO
    terms
  findings:
  - statement: >-
      Maps InterPro signature matches to GO terms. For ADAMTSL5 the relevant signature
      is IPR013273 ADAMTS/ADAMTS-like, whose sole GO mapping is GO:0030198
      extracellular matrix organization. The signature covers 26,580 proteins and spans
      both the catalytic ADAMTS proteases and the non-catalytic ADAMTS-like proteins.
  reference_review:
    relevance: MEDIUM
    correctness: VERIFIED
    review_notes: >-
      The GO_REF itself is correctly applied - ADAMTSL5 genuinely matches IPR013273.
      The issue is the breadth of the signature, not a citation error. Verified the
      signature name, protein count and single GO mapping against the InterPro API.
- id: GO_REF:0000033
  title: Annotation inferences using phylogenetic trees
  findings:
  - statement: >-
      PAINT phylogenetic propagation. For ADAMTSL5 the single IBA row derives from
      PANTHER node PTN000347317 in family PTHR13723, with 16 protein donors. The GOA
      WITH/FROM field matches the cached PAINT seed list for that node exactly.
  reference_review:
    relevance: HIGH
    correctness: VERIFIED
    review_notes: >-
      Verified by resolving all 17 WITH/FROM tokens and confirming the seed set matches
      the cached PTHR13723 PAINT file exactly, plus the node id.
- id: GO_REF:0000044
  title: Gene Ontology annotation based on UniProtKB/Swiss-Prot Subcellular Location
    vocabulary mapping, accompanied by conservative changes to GO terms applied by
    UniProt
  findings:
  - statement: >-
      Maps the UniProt Secreted subcellular location keyword (SL-0243) to GO:0005576
      extracellular region. Correct for ADAMTSL5, which has a signal peptide and
      experimentally demonstrated secretion.
  reference_review:
    relevance: MEDIUM
    correctness: VERIFIED
- id: PMID:23010571
  title: A disintegrin-like and metalloprotease domain containing thrombospondin type
    1 motif-like 5 (ADAMTSL5) is a novel fibrillin-1-, fibrillin-2-, and heparin-binding
    member of the ADAMTS superfamily containing a netrin-like module.
  findings:
  - statement: >-
      ADAMTSL5 is a secreted, N-glycosylated ~60 kDa glycoprotein. Recombinant protein
      is recovered from the conditioned medium of transfected HEK293F, COS-1 and CHO-K1
      cells and deposits into the peri-cellular and baso-lateral extracellular matrix.
    supporting_text: "Recombinant ADAMTSL5 is a secreted, N-glycosylated 60kDa"
  - statement: >-
      ADAMTSL5 binds heparin through its C-terminal NTR module in a salt-sensitive
      ionic interaction; only the NTR-containing C-terminal fragment is retained on
      heparin-agarose.
    supporting_text: "only the C-terminal fragment containing the NTR-module was retained by the heparin matrix"
  - statement: >-
      ADAMTSL5 binds both fibrillin-1 and fibrillin-2 and co-localises with assembled
      fibrillin microfibrils in fibroblast extracellular matrix - the first family
      member shown to bind both fibrillins.
    supporting_text: "Taken together, the findings are consistent with specific binding
      of ADAMTSL5 to fibrillin-1 and fibrillin-2 and to their macromolecular assemblies,
      i.e., fibrillin microfibrils."
  - statement: >-
      Critically, the same experiment found no effect of ADAMTSL5 on microfibril
      assembly, and direct binding to fibronectin was not supported. This is the only
      direct test of a matrix-organizing role for ADAMTSL5 and it was negative, though
      reported as data not shown.
    supporting_text: "did not identify a consistent difference"
  - statement: >-
      The ADAMTS-like proteins, including ADAMTSL5, have no catalytic domain and hence
      no proteolytic activity - the basis for rejecting any protease-derived functional
      inference for this gene.
    supporting_text: "In contrast to ADAMTS proteases, ADAMTSLs lack a catalytic domain"
  - statement: >-
      In mouse organogenesis ADAMTSL5 is broadly expressed, prominently in
      musculoskeletal tissues - skeletal muscle, cartilage and bone - and in many
      epithelia. Cartilage and bone are not elastic-fibre tissues.
    supporting_text: "Immunostaining during mouse organogenesis identified ADAMTSL5 in"
  reference_review:
    relevance: HIGH
    correctness: VERIFIED
    review_notes: >-
      The single defining paper for this gene and the source of all five IDA rows. Full
      text available and read in full; all quotes verified verbatim against the cached
      PMC text. Projection test negative - QuickGO shows it annotates only 2 entities
      (ADAMTSL5 and its mouse ortholog), so it is gene-specific curation rather than a
      complex- or family-wide projection. No retraction, erratum or expression of
      concern.
- id: PMID:23962539
  title: Elastic fibres in health and disease.
  findings:
  - statement: >-
      A general review of elastic fibre composition, assembly and disease. Its abstract
      does not mention ADAMTSL5. QuickGO shows it is used as the TAS reference for 66
      annotations across 62 distinct entities, 41 of which receive GO:0071953 elastic
      fiber - a bulk curation pass rather than a gene-specific traceable statement. The
      same pass assigned the more specific GO:0001527 microfibril to 15 other proteins.
  reference_review:
    relevance: LOW
    correctness: LOW_QUALITY
    review_notes: >-
      Not miscited - the review is a legitimate source for elastic fibre composition,
      and TAS is the honest evidence code for it. But it is weak support for a
      gene-specific localisation: full text is unavailable, the abstract never names
      ADAMTSL5, and the same reference underpins 41 other elastic-fiber assignments.
      Flagged LOW_QUALITY on those grounds rather than MISCITED. Per campaign policy I
      did not REMOVE on an unverifiable full text; the row is MODIFIED to the specific
      term the gene's own IDA evidence supports.
- id: PMID:32296183
  title: A reference map of the human binary protein interactome.
  findings:
  - statement: >-
      The HuRI systematic yeast two-hybrid screen, source of all three GO:0005515 rows
      for ADAMTSL5 (partners CYSRT1, KRTAP5-9 and FHL5). IntAct logs each partner under
      three sub-methods of the same experiment - two hybrid array, two hybrid prey
      pooling approach and validated two hybrid - all at MI-score 0.56, so UniProt's
      NbExp=3 is one screen counted three ways rather than three independent
      observations.
  reference_review:
    relevance: LOW
    correctness: VERIFIED
    review_notes: >-
      Correctly cited and a high-quality resource; the problem is not the paper but
      what a single unreplicated Y2H hit supports. All three ADAMTSL5 partners are
      extreme interaction hubs (517, 213 and 108 distinct IntAct partners against
      ADAMTSL5's 12) and all three are intracellular, so none is topologically
      accessible to a secreted ECM protein. All three accessions resolve to reviewed
      canonical Swiss-Prot entries at canonical lengths - no ORFeome or TrEMBL
      substitution.
- id: PMID:26621454
  title: Melanocyte antigen triggers autoimmunity in human psoriasis.
  findings:
  - statement: >-
      ADAMTSL5 is an HLA-C*06:02-presented melanocytic autoantigen recognised by a
      Valpha3S1/Vbeta13S1 CD8+ T cell receptor from a psoriasis patient, and ADAMTSL5
      stimulation induced IL-17A in CD8+ T cells from psoriasis patients only. The
      recognised peptide VRSRRCLRL maps to residues 67-75, inside the TSP type-1
      domain.
    supporting_text: "we identified ADAMTS-like protein 5 (ADAMTSL5) as an"
  reference_review:
    relevance: MEDIUM
    correctness: VERIFIED
    review_notes: >-
      Verified against the cached full text; peptide position confirmed independently
      against the UniProt sequence. Included for the description because it is the
      gene's best-known biology, but it generates no GO annotation - being presented as
      an MHC-bound peptide is a property of a degradation product in an
      antigen-presentation pathway, not a molecular function of the secreted
      glycoprotein. The record carries two Comment in entries, which are commentaries
      rather than corrections; no retraction or erratum.
- id: PMID:33197513
  title: ADAMTSL5 is an epigenetically activated gene underlying tumorigenesis and drug
    resistance in hepatocellular carcinoma.
  findings:
  - statement: >-
      ADAMTSL5 is activated by gene-body CpG island hypermethylation in mouse and human
      hepatocellular carcinoma. Depletion interfered with tumorigenic properties, while
      overexpression conferred tumorigenicity on pre-tumoural hepatocytes sensitised by
      modest MET expression.
    supporting_text: "ADAMTSL5 targeting interfered with tumorigenic properties of HCC"
  reference_review:
    relevance: MEDIUM
    correctness: VERIFIED
    review_notes: >-
      Abstract-only in the cache. Used for the description but deliberately not used to
      propose a signalling GO term: the receptor tyrosine kinase changes are downstream
      readouts in a cancer model, and no mechanism connecting a secreted ECM
      glycoprotein to receptor levels has been established. Notably co-authored by
      Apte, who also authored the defining matrix paper. No retraction or erratum.
- id: file:genes/human/ADAMTSL5/ADAMTSL5-bioinformatics/RESULTS.md
  title: ADAMTSL5 bioinformatics analysis - WITH/FROM resolution, family propagation
    census, and interaction-partner assessment
  findings:
  - statement: >-
      All 16 protein donors on the GO:0031012 IBA row carry their own experimental
      IDA/HDA evidence for that term or a descendant, and the donor set spans four
      distinct locations, making GO:0031012 the correct least common ancestor.
    supporting_text: "All 16 protein donors carry their own experimental (IDA/HDA) annotation"
  - statement: >-
      Propagation from PANTHER node PTN000347317 is incoherent across the human family:
      GO:0031012 reached 5 of 7 non-catalytic members and GO:0030198 only 3 of 7, while
      human PAPLN received no IBA at all despite three of its own orthologs being seeds
      for GO:0031012 at that node.
    supporting_text: "PAPLN receives neither"
  - statement: >-
      The three GO:0005515 partners are extreme Y2H hubs (517, 213 and 108 distinct
      IntAct partners against ADAMTSL5's 12) and all are topologically inaccessible to
      a secreted protein.
    supporting_text: "CYSRT1 (A8MQ03) | 1670 | **517** | cornified envelope | 144 aa |"
  - statement: >-
      ADAMTSL5 has no zinc-binding metalloprotease motif - no HExxHxxGxxHD and no
      HExxH substring at all in its 481 residues - confirming the loss of catalysis
      independently of UniProt's CAUTION line.
    supporting_text: "no `HExxH` substring at all"
- id: file:genes/human/ADAMTSL5/ADAMTSL5-deep-research-affinage.md
  title: Affinage mechanistic annotation for ADAMTSL5 (human)
  findings:
  - statement: >-
      Provider deep-research record. Its narrative of the fibrillin, heparin, psoriasis
      and hepatocellular carcinoma literature is broadly accurate and every citation is
      a numeric PMID, but its GO grounding proposes GO:0008289 lipid binding and
      GO:0098772 molecular function regulator activity, neither of which has any
      support - the demonstrated ligand is heparin, a glycosaminoglycan. Not imported.
    supporting_text: "GO:0008289 lipid binding"
  reference_review:
    relevance: LOW
    correctness: LOW_QUALITY
    review_notes: >-
      gates_passed is False. The tripped gate is specifically the self-evaluation
      pairwise tie, not a faithfulness failure (faith_pct is 100.0), and no citation is
      a bioRxiv PMID:bio_* identifier. Marked LOW_QUALITY per campaign policy for a
      failed gate. Every claim used from it was re-verified against the cited PMID
      directly and is quoted from the PMID rather than from the provider prose; its GO
      grounding was rejected outright.