TFAP2B (AP-2beta, Q92481) review notes

2026-10-08 — initial review (claude-code, project NEURAL_CREST_ORIGINS, Tier 4)

Identity and molecular activity

• AP-2beta is a sequence-specific AP-2 family transcription factor. It binds GCC(N3)GGC-type sites and activates
  AP-2-dependent promoters PMID:7555706.
• It dimerizes through the conserved C-terminal HSH domain, both with itself and with AP-2alpha
  PMID:7555706.
  Char syndrome missense mutants still dimerize but bind DNA abnormally and poison wild-type TFAP2B
  PMID:10802654,
  PMID:10802654,
  PMID:11505339.
• Coactivator: CITED2 (with p300/CBP) PMID:11694877,
  PMID:12586840.
  Corepressor-like inhibitor: KCTD1 PMID:19115315.
  SUMOylation by UBC9 lowers activity PMID:12072434.
• Repression is context-dependent: in chick, TFAP2B represses TFAP2C, probably directly
  PMID:31848212;
  in mouse limb it activates Bmp2 and represses Bmp4 PMID:21829553.
• MF conclusion: GO:0001228 (activator) is core; GO:0042803/GO:0046982 dimerization are accepted;
  CITED2 rows -> GO:0001223, KCTD1 -> GO:0001222 (consistent with the TFAP2A review). No pioneer-factor MF term exists.

Name confusion: PMID:7559606 ("AP-2B") is about an AP-2alpha isoform

• Two IDA rows (GO:0001228, GO:0000122) cite Duan & Clemmons 1995. The cached abstract uses a human AP-2 (alpha) expression
  construct and "AP-2B" as a dominant negative: PMID:7559606.
  AP-2B is an alternatively spliced product of the AP-2alpha gene (TFAP2A), not AP-2beta:
  PMID:8321221.
  AP-2beta was only cloned the month before (PMID:7555706, Sept 1995). Both rows REMOVED; the activator function is kept
  by the PMID:7555706 IDA rows. Report to the source curator.

Network layer: neural crest specifier (not a border specifier)

• Chick (Rothstein & Simoes-Costa 2020, full text). TFAP2B comes on only at the onset of specification
  PMID:31848212.
  Knockdown spares the border and disrupts specification
  PMID:31848212.
  It forms heterodimers with TFAP2A in vivo PMID:31848212
  and recruits TFAP2A to specification enhancers PMID:31848212.
  TFAP2C cannot replace it PMID:31848212,
  and premature TFAP2B drives premature specification PMID:31848212.
  Upstream: TFAP2A/C heterodimers plus induction genes activate TFAP2B; TFAP2B then represses TFAP2C (the partner switch).
• Frog: Pax3 and Zic1 activate tfap2b without new protein synthesis
  PMID:24360906.
  Note: the body of the same paper lists tfap2b among "known neural border specifiers"; the abstract calls it a crest
  specifier. The chick timing data settle the layer.
• Chick cranial circuit: Lhx5/Dmbx1 activate Tfap2b, and Tfap2b activates Ets1
  PMID:27339986.
  Sox8 + Tfap2b + Ets1 together (not singly) reprogram trunk crest to cranial, chondrogenic identity
  PMID:27339986,
  PMID:27339986.
  But Tfap2b expression itself is pan-axial in chick and skate PMID:31645763:
  what is cranial-specific is the circuit wiring, not TFAP2B.
• Mouse: TFAP2B is largely redundant with TFAP2A in crest. Single Tfap2b KO crest forms; sympathetic ganglia are reduced
  by progenitor apoptosis PMID:21539825;
  the double KO loses sympathetic and sensory ganglia PMID:21539825.
  Van Otterloo 2022 (full text) states that in mouse no AP-2alpha/beta heterodimer function is irreplaceable
  PMID:35333176.
• Zebrafish: tfap2a+tfap2b double deficiency removes crest cartilage, via ectoderm PMID:15944192;
  tfap2b crispants have fewer enteric neurons PMID:35874825.

Decision on the NC-branch term

• GO:0014036 neural crest cell fate specification, NEW (ISS, from chick PMID:31848212). Not GO:0014029: TFAP2B has no
  border role (absent at border stages; knockdown spares border markers), so the border-inclusive term would misplace it.
  Not GO:0014034: GO:0014036 is part_of GO:0014034 and the evidence is specifically on the reversible specification step
  (timing, enhancer recruitment). This matches the project convention (crest specifiers -> GO:0014036).
• Participation test: passed. TFAP2B is a DNA-binding heterodimer partner that itself occupies specification enhancers
  and redirects TFAP2A there; it does part of the work, not merely being required.
• Caveat for curators: necessity is shown only in chick. In mouse TFAP2B is redundant with TFAP2A. Redundancy does not
  negate participation, but the transfer to human rests on one species plus conserved protein activity and on human
  Char syndrome being a crest-derivative disorder PMID:10802654.
  Flagged for sign-off.

Comparator check (QuickGO, 2026-10-08)

Query: annotations to GO:0014029/GO:0014033/GO:0001755/GO:0014032 and descendants (is_a, part_of), all evidence, all taxa.
- No TFAP2B in any species (human Q92481, mouse Q61313, rat, chicken A0A8V0XPT7 etc., X. laevis tfap2b.L, X. tropicalis,
zebrafish tfap2b) carries any neural-crest-branch term.
- Paralogs: zebrafish tfap2a carries GO:0014036 (IMP, PMID:14534133; IGI PMID:20885782) and GO:0014032; zebrafish tfap2c carries
GO:0014036 (IGI, PMID:20885782) and GO:0014032; mouse Tfap2a GO:0014032 (IMP, PMID:8622766); rat Tfap2a GO:0014032 (ISO).
Human TFAP2A/C/D/E: none in GOA (TFAP2A has a NEW GO:0014029 in this repository only).
- Same-layer peers (frog sox10, sox9-a, sox8, snai2, snai1) carry GO:0014036 (see project notes).
- Interpretation: the absence on TFAP2B is explained, not a convention. The paralog-specific specification evidence is a
single 2020 chick paper, and chicken TFAP2B entries are unreviewed TrEMBL with only electronic annotation; mouse and
zebrafish single mutants do not block specification because of redundancy. Mouse Tfap2b does carry crest-derivative
terms (GO:0048485 sympathetic NS development, IGI PMID:21539825); zebrafish tfap2b carries GO:0048484 enteric NS
development (IMP, PMID:35874825) and GO:0051216 cartilage development (IGI, PMID:15944192).
- No GO-CAM model in gocams/index.tsv contains TFAP2B.

Other roles (pleiotropic, outside the crest GRN)

• Ductus arteriosus: human Char syndrome / PDA2; mouse KO ductus stays open
  PMID:21829553;
  ductal smooth muscle programme PMID:18635823.
• Limb: PMID:21829553.
• Kidney: distal tubule/collecting duct epithelial survival PMID:9271117,
  PMID:12695560. Not seen in Char patients.
• Sympathoadrenal / noradrenergic neurons: survival of progenitors (mouse) and noradrenergic differentiation in neuroblastoma
  PMID:26598443.
• Adipocyte / T2D association rows (PMID:15940393, 19325541, 16373396): genetic association plus overexpression; glucose
  metabolism and insulin secretion IMP rows judged over-annotated.
• Retinoblastoma overexpression rows (PMID:20607706): over-annotated, matching the TFAP2A review.

Evolution

• Amphioxus has a single AP-2, expressed in non-neural ectoderm, not at the border PMID:31645763.
• Lamprey Tfap2 is expressed in crest at all axial levels PMID:31645763.
• The TFAP2A/B/C paralogs come from the vertebrate genome duplications. Rothstein & Simoes-Costa read the nested
  expression as subfunctionalisation PMID:31848212.
  So the border->crest partner switch is a jawed-vertebrate (at least amniote) elaboration of a crest role that one AP-2
  already played in the vertebrate ancestor; how the partition is drawn differs between lineages (chick: non-redundant
  B for specification; mouse: A dominant, B redundant; zebrafish: A+C for induction).

Module suggestions (not edited here)

• Crest fate specification part: TFAP2A–TFAP2B heterodimer annoton (GO:0001228 / GO:0046982, part_of GO:0014036).
• Edges: Pax3+Zic1 -> tfap2b (frog, translation-independent); TFAP2A/C + border genes -> TFAP2B (chick CUT&RUN peaks in locus);
  TFAP2B -| TFAP2C (chick, likely direct); TFAP2A/B -> FOXD3 NC1, SOX9, SNAI2, SOX10 enhancers; Tfap2b -> Ets1 (chick ChIP).
• Cranial ectomesenchyme part: Sox8–Tfap2b–Ets1 combination (sufficient only together).