ADAM17 (TACE) — curation notes

Provenance note. Provider deep research was unavailable for this gene: the Falcon
provider returned 402 Payment Required and perplexity is not configured in this
checkout. No *-deep-research-<provider>.md file was created. The synthesis below is
my own, built from the UniProt record (ADAM17-uniprot.txt), the cached publications
in publications/ (every PMID cited in the GOA-seeded review), and targeted
literature retrieval (PubMed E-utilities; one additional paper, PMID:20676056, was
fetched with just fetch-pmid). Every assertion below carries an inline citation
with a verbatim quote from a cached publication.

1. What the protein is

ADAM17 (TACE, TNF-alpha converting enzyme; UniProt P78536; EC 3.4.24.86) is a type I
transmembrane zinc metalloprotease of the adamalysin/ADAM family. Domain order from the
UniProt feature table: signal peptide, inhibitory pro-domain, catalytic metalloprotease
domain (HExxHxxGxxH zinc site), disintegrin domain, membrane-proximal (MPD/CANDIS)
region, transmembrane segment, cytoplasmic tail ending in a group-I PDZ-binding motif.
It was cloned twice in 1997 as the protease that converts the 26-kDa membrane TNF
precursor to the soluble cytokine PMID:9034190; PMID:9034191.

Zinc dependence and cofactor requirement are recorded in UniProt (Binds 1 zinc ion per subunit), and the pro-domain is removed by a furin-type convertase in the secretory
pathway, which is the prerequisite for activity in this protease subfamily
PMID:9920899 — for ADAM17 itself the same furin-dependent maturation in the
Golgi/raft compartment was shown directly PMID:17010968.

2. Core activity: regulated ectodomain shedding

The single core molecular function is metalloendopeptidase activity exerted on the
juxtamembrane stalks of transmembrane proteins — i.e. ectodomain shedding at the
plasma membrane. UniProt summarises the breadth: "Transmembrane metalloprotease which
mediates the ectodomain shedding of a myriad of transmembrane proteins including
adhesion proteins, growth factor precursors and cytokines important for inflammation
and immunity". A 2024 paper puts a number on it PMID:38771644.

Substrates with direct, cited experimental support in the cached set:

• TNF (pro-TNF-alpha) — the founding substrate PMID:9034190;
  the cleavage site is recorded in the UniProt catalytic-activity line
  (Pro-Leu-Ala-Gln-Ala-|-Val-Arg-Ser-Ser-Ser).
• EGFR ligands (amphiregulin, TGF-alpha, HB-EGF, epiregulin, epigen) —
  PMID:12743035; PMID:29897336.
• IL6R — PMID:16034137
• KIT (c-Kit) — PMID:14625290.
• GP5 (platelet glycoprotein V) — PMID:15691827.
• FCGR3A/CD16A on NK cells, with the cleavage site mapped between Ala195 and Val196
  PMID:24337742.
• MICA (NKG2D ligand; tumour immune escape) PMID:18676862.
• CXCL16 PMID:18373975
• VASN (vasorin) — the mechanism by which ADAM17 loss raises TGF-beta signalling
  PMID:21170088.
• ACE2 — ADAM17 competes with TMPRSS2 for the receptor
  PMID:24227843.
• TREM2 — relevant to the Alzheimer/microglia literature; FTD-associated TREM2
  missense variants "abolish shedding by ADAM proteases" PMID:24990881.
• NOTCH1 at the S2 site, in vitro and glycosylation-sensitive PMID:24226769.
• DCC/Frazzled in commissural axon guidance PMID:36476876

3. How the activity is controlled (the part that matters for GO review)

ADAM17 is better described as the catalytic subunit of a sheddase complex than as a
free-standing protease. Three control layers are well documented:

1. iRhom (RHBDF1/RHBDF2) dependence. ER exit, surface delivery and activation all
   require an iRhom PMID:22246777, and the iRhom tail is the platform that
   transduces activating stimuli PMID:29897333.
   FRMD8/iTAP stabilises the complex at the surface PMID:29897336.
2. Membrane microenvironment. Shedding activity partitions into cholesterol-rich
   rafts PMID:17010968, and cholesterol efflux to HDL mobilises it PMID:17786981. Tetraspanins (CD9, CD81, TSPAN8) tune substrate
   selectivity PMID:36078095
3. Inhibitors and interactors. TIMP3 is the physiological inhibitor and its complex
   with the catalytic domain has been crystallised PMID:18638486. Integrin alpha5beta1 binds
   the disintegrin domain and restrains catalysis
   PMID:30455686. The cytoplasmic PDZ motif recruits PTPH1/PTPN3 and SAP97/DLG1, both
   of which modulate levels or shedding output PMID:12207026;
   PMID:12668732

This matters for review decisions: most of ADAM17's signalling annotations are
consequences of releasing a ligand or receptor ectodomain, not of ADAM17 performing a
step inside the downstream cascade. Where a signalling-pathway term is supported by
ADAM17 cleaving the pathway's own activating ligand (EGFR ligands → EGFR pathway; TNF →
TNF pathway; Notch S2 cleavage → Notch pathway), the participation test is met because
ADAM17 catalyses the activating step. Where the term describes a cascade downstream of
that release (PI3K/AKT signal transduction, G1/S transition, cell proliferation/growth),
ADAM17 is an upstream requirement rather than a participant, and those rows are best
kept as non-core or marked over-annotated.

4. ADAM17 vs ADAM10 as the APP alpha-secretase

This is the question the dismech Alzheimer entry depends on, and the evidence is
asymmetric:

• ADAM17 is the regulated (phorbol-ester-stimulated) alpha-secretase. The original
  knockout study is explicit PMID:9774383 An overexpression/RNAi comparison in glioblastoma cells
  placed ADAM9, ADAM10 and ADAM17 all in the alpha-secretase category PMID:12535668, but that design
  cannot separate constitutive from stimulated activity.
• ADAM10, not ADAM17, is the constitutive alpha-secretase in neurons. The decisive
  knockdown series found PMID:20676056 while ADAM17 was required specifically for the
  stimulated arm PMID:20676056; the same paper notes that ADAM17-deficient fibroblasts show no
  altered constitutive alpha-cleavage PMID:20676056 and leaves the physiological relevance
  of ADAM17-mediated APP cleavage explicitly open PMID:20676056.

Reading for the muscarinic M1 node. The pharmacology that connects M1 signalling to
non-amyloidogenic APP processing runs through PKC-stimulated, i.e. regulated,
alpha-secretase — which is the arm ADAM17 demonstrably serves, and the arm ADAM10 is
dispensable for. So naming ADAM17 as the alpha-secretase effector of M1/PKC-stimulated
shedding is defensible; naming it as "the alpha-secretase" without that qualifier is
not, because the constitutive neuronal activity is ADAM10's. The strength of evidence is
moderate for the regulated arm in cell lines and weak-to-absent for a physiological
GPCR-driven APP cleavage in neurons (PMID:20676056's own caveat above). Note also that
the same asymmetry recurs for Notch: ADAM10 serves ligand-induced physiological
cleavage, and a 2024 paper is blunt that PMID:38771644

5. Loss of function phenotypes

• Mouse TaceΔZn/ΔZn animals are hypermetabolic and lean PMID:18687778, which is the (mouse) basis of the negative-regulation-of-thermogenesis
  annotation.
• A human CANDIS-domain variant causes autosomal-dominant hypotrichosis with woolly
  hair through enhanced ubiquitin-mediated degradation of the mutant protein
  PMID:38771644 — note this is a property of the mutant protein, which is why the
  derived ubiquitin binding molecular-function annotation from this paper is
  questionable (see the review row).
• Conditional deletion in mouse spinal cord reduces commissural midline crossing
  PMID:36476876

6. Review decisions taken (summary of reasoning patterns)

• GO:0004222 metalloendopeptidase activity and GO:0006509 membrane protein ectodomain proteolysis — accepted throughout; these are the core function and are supported by
  many independent IDA/IMP rows.
• GO:0005515 protein binding — removed per the curation guideline (uninformative),
  except where the cited paper supports a more informative function, in which case the
  row is MODIFYed (e.g. the TIMP3 structure paper → peptidase inhibitor binding;
  the integrin papers already carry integrin binding).
• Downstream signalling/proliferation terms (GO:0043491, GO:1900087,
  GO:0008284, GO:0030307) — kept as non-core or marked over-annotated: the ADAM17
  step is the ligand release, not a step inside the cascade.
• ISS rows transferred from the mouse gene paper PMID:10433800 (germinal centre
  formation, B/T cell differentiation, spleen development, mast cell apoptosis,
  xenobiotic response, cell motility) — these are mouse knockout phenotypes propagated
  by sequence similarity; the cited reference is a cDNA/gene-structure paper that
  contains none of those data, so each is kept as non-core (the underlying mouse
  biology is real) rather than accepted as a core function.
• No NEW annotations were proposed. The candidates considered and rejected:
  GO:0050870 positive regulation of T cell activation, GO:0002687 positive regulation of leukocyte migration and amyloid-pathway terms beyond the existing
  GO:0042987. In each case ADAM17's contribution is the release of a ligand already
  annotated for the downstream process, which fails the participation test, and the
  comparator check (other sheddases such as ADAM10, BACE1 and MMP14 are likewise not
  annotated to the downstream immune-activation terms their products drive) indicates a
  convention rather than a gap.