TFAP2A (AP-2alpha, P05549) review notes

2026-10-05 — initial review (claude-code)

Identity and molecular activity

• AP-2alpha is a sequence-specific transcription factor of the AP-2 family. It binds GCC(N3)GGC sites as a homo- or
  heterodimer formed through a C-terminal helix-span-helix (HSH) domain that sits next to a basic DNA-binding region.
  PMID:1998122
  PMID:7555706
• It acts mainly as an activator. Examples: IGFBP5 PMID:7559606,
  CDKN1A/p21 PMID:21084835,
  and FXN PMID:20808827.
  It needs the CITED2 and p300/CBP coactivators for this
  PMID:12586840.
• It also represses some promoters, and repression depends on DNA binding. Examples: Mn-SOD
  PMID:11278550, C/EBPalpha in preadipocytes
  PMID:9520389, and c-myc
  PMID:20066163.
  The corepressor NPM1 is recruited PMID:17318229,
  and KCTD1 and KCTD15 inhibit the activation domain PMID:19115315,
  PMID:23382213.
• MF conclusion: the core terms are GO:0001228, plus context-dependent GO:0001227, GO:0000978/0000977 cis-regulatory
  binding, and GO:0042803 homodimerization (via the HSH domain). The "pioneer" chromatin-opening activity comes from chick work
  PMID:31848212.
  GO has no pioneer-factor MF term. The ISS chromatin binding row is the closest existing term.

Network layer: border specifier AND crest specifier (reiterative)

• Expression. AP-2 is expressed in non-neural ectoderm and the neural plate border at gastrula stages, then in premigratory
  and migratory crest. Mouse: PMID:1989904.
  Chick: PMID:31848212.
  Frog: AP2a is the earliest border gene PMID:21169220.
• Gain of function (frog). It is sufficient to induce crest genes ectopically
  PMID:12511599,
  and to set up a border-like pattern in neuralized ectoderm
  PMID:21169220.
• Loss of function. Frog morphants lose the crest territory
  PMID:12511599.
  Zebrafish lockjaw disrupts specification PMID:14534133.
  The mont blanc allele does not PMID:14985255.
  This difference is explained by redundancy with tfap2c
  PMID:17258188.
  In mouse, crest induction is normal in nulls PMID:17258188,
  but crest-derived structures fail PMID:8622766,
  and crest-specific deletion confirms a cell-autonomous crest role
  PMID:14975722.
• Direct targets and partner switch (chick). TFAP2A/C heterodimers activate the border program, and TFAP2A/B heterodimers activate the specification program
  PMID:31848212.
• Human. Heterozygous TFAP2A mutations cause branchio-oculo-facial syndrome (BOFS), a neurocristopathy-like disorder
  PMID:18423521,
  PMID:21204207.
• Placement. AP-2alpha is the clearest case in the network of a gene that spans two layers. In frog it acts first as a
  border specifier, upstream of pax3, and again within the crest as a specifier
  PMID:21169220.
  It is also a non-neural ectoderm/epidermis factor, which is its ancestral deployment (see evolution below).

Choice of NC process term (project convention)

• Human TFAP2A has no NC-branch GO term in GOA at all, so any NC term is NEW.
• Under the pending project convention, GO:0014029 is kept for border-acting genes and GO:0014036 for crest specifiers.
  AP-2alpha does both. GO:0014036 is part_of GO:0014034, which is part_of GO:0014029 (checked via the QuickGO ancestors
  API: GO:0014036 has GO:0014029 as an ancestor). Proposing both would assert an ancestor plus its descendant, which
  CLAUDE.md rejects as redundant. Decision: one NEW, GO:0014029 neural crest formation. It covers the border-forming
  step (de Crozé 2011), which GO:0014036 would miss, and the specification step is part_of it. The zebrafish IMP
  GO:0014036 rows support the narrower claim in that species.
  Evidence code is ISS: human TFAP2A has no direct experiment, and frog/chick/zebrafish/mouse data are transferred by
  orthology because human is the proxy here.
• Participation test: AP-2alpha is a DNA-binding TF that directly occupies border and crest enhancers (chick CUT&RUN) and
  is sufficient to induce border and crest genes. It does part of the work; it is not a substrate or a mere requirement.

Comparator check (QuickGO, recorded 2026-10-05)

Query: QuickGO annotation/downloadSearch?geneProductId=<all UniProt accessions for tfap2a/b/c/d/e in human 9606, mouse 10090, rat 10116, zebrafish 7955, X. laevis 8355, X. tropicalis 8364, chicken 9031 (153 accessions from UniProt REST)>&goId=GO:0014029,GO:0014034,GO:0014036,GO:0014032,GO:0014033,GO:0001755,GO:0014031&goUsage=descendants.
Result (76 rows):
- Zebrafish tfap2a: GO:0014036 IMP (PMID:14534133), GO:0014036 IGI (PMID:20885782), GO:0014032 IGI (PMID:17258188,
17724731, 19439494, 19955371, 21963426, 22241841).
- Zebrafish tfap2c: GO:0014036 IGI (PMID:20885782), GO:0014032 IGI (several).
- Mouse Tfap2a (P34056): GO:0014032 IMP (PMID:8622766). Rat Tfap2a: GO:0014032 ISO.
- Human TFAP2A/B/C, X. laevis and X. tropicalis tfap2a, chicken TFAP2A: none.

So orthologs in two vertebrate MODs carry NC terms. Absence in human and frog is a coverage gap: the frog border papers
(Luo 2003, de Crozé 2011) were never curated to the unreviewed frog entries. It is not a convention. This is
the same pattern found for myc-a and pax3-a.

Evolution / outgroups

• Amphioxus has a single AP-2. It is expressed in non-neural ectoderm, not at the border or in the dorsal neural tube.
  Lamprey AP-2 is expressed in crest
  PMID:12397104,
  PMID:12397104.
• Interpretation: the ectodermal/epidermal role is ancestral (chordate). Crest expression is a vertebrate co-option
  through changed expression, as for Id and SoxE. The family expanded to five paralogs in vertebrates, and the
  border (A/C) versus crest (A/B) heterodimer partner switch fits subfunctionalisation after duplication
  PMID:17724731.
  So AP-2 is a border gene in vertebrates but not a border gene in amphioxus. That is unlike Pax3/7 and Zic, whose border
  expression is ancestral. AP-2 is therefore a candidate for the "new border input" that made the border crest-competent.

Annotation decisions (summary)

• MF core: GO:0001228 (ACCEPT), GO:0001227 (ACCEPT, context-dependent), cis-regulatory binding rows (ACCEPT).
• protein binding (15 rows): CITED2/EP300 → GO:0001223 transcription coactivator binding (MODIFY). NPM1, KCTD1 →
  GO:0001222 transcription corepressor binding (MODIFY; KCTD1 itself carries GO:0003714 IDA from PMID:19115315).
  UBC9 (x2), HT screens and MAPT → REMOVE as uninformative. MYO6 (PMID:11447109) → REMOVE. The cached full text
  describes the clathrin adaptor protein complex AP-2, not transcription factor AP-2. This is name confusion that can
  be verified from the full text in hand.
• identical protein binding (PMID:1998122) → MODIFY to GO:0042803 protein homodimerization activity.
• Developmental processes from mouse ISS and BOFS IMP (eye, ear, palate, limb, kidney, skeleton, cranial nerves):
  KEEP_AS_NON_CORE. These are crest- and ectoderm-derived outputs. Embryonic cranial skeleton morphogenesis ACCEPTED
  as the main crest output.
• Over-annotated: sensory perception of sound (BOFS deafness is structural), retina layer formation (IEP),
  cellular response to iron ion (mRNA change only), negative regulation of ROS metabolic process (indirect, via MYC),
  positive regulation of neuron apoptotic process and regulation of cell differentiation (overexpression in
  retinoblastoma lines).

Open issues

• GO has no neural plate border term. AP-2alpha is the best example of a gene that needs one, because it initiates the border.
• Pioneer factor activity has no GO MF term.
• Epidermis/non-neural ectoderm development is the ancestral role. Human/mouse evidence for TFAP2A alone is confounded by
  TFAP2C redundancy, and no paper establishing it was cached, so it was raised as a question and not added.