gbx2 (Xenopus laevis, Q91907) review notes

Project: NEURAL_CREST_ORIGINS, Tier 2 (neural plate border specifiers).
Reviewed 2026-10-05 by claude-code.

Identity and paralog/homeolog situation

• Q91907 (GBX2_XENLA, 340 aa) is the von Bubnoff 1996 Xgbx-2 cDNA; UniProt
  cross-references it to Xenbase gbx2.2.L (XB-GENE-866652). It is the only
  reviewed X. laevis Gbx entry.
• X. laevis carries four gbx2 copies (two paralogous genes, gbx2.1 and gbx2.2,
  each with L/S homeologs) plus gbx1.S. Pairwise global alignment against Q91907
  (BLOSUM62, run 2026-10-05 in scratch, sequences from UniProt REST):
  A0ACK5WZN6 gbx2.2.L 339/340 identical; A6H8J9 / A0AA58EVU7 gbx2.1.L 333/340;
  Q90YY2 ("Gbx2b", gbx2.S/gbx2.1.S) and A0A8J0PTF7 324 identical (336 aa);
  A0A1L8EPH5 gbx2.2.S 316/340.
• Tour et al. 2001 found two X. laevis Gbx2 genes PMID:11684099 and PMID:11684099. Q90YY2 is the "Gbx2b" entry and differs from Q91907 at
  about this many positions, so Tour's Xgbx2a is most likely the Q91907 lineage.
  This is an inference from sequence, not a stated mapping.
• The Wnt-target paper (Li 2009) names "Gbx2" without a paralog. Yokota 2017
  refers to it as gbx2.2 PMID:28695961,
  which matches Q91907's Xenbase gene.
• No X. laevis or X. tropicalis Gbx entry carries any experimental GO
  annotation. All 13 Q91907 rows are IBA/IEA/TAS. The Xenopus Gbx2 literature
  (Li 2009, Steventon 2012, Tour 2002) is uncurated in GO.

Expression

• PMID:8652408
• PMID:8652408
• Early gbx2 sits between Otx2 and Xcad2 domains PMID:11850185
• BMP timing: PMID:37489332 (after BMP inhibition), so gbx2 is
  in a BMP-sensitive, temporally ordered head-patterning sequence.
• In the pre-placodal region (PPR) gbx2 and otx2 form an abutting boundary:
  PMID:22564795

Upstream: a direct Wnt target

• PMID:19736322
• PMID:19736322
• Reactome R-XLA-9835542 models exactly this step (tcf7l1:ctnnb1 binding the gbx2
  promoter). Note: this is regulation of gbx2, not an activity of Gbx2.
• Conserved in zebrafish via the paralog gbx1 PMID:19341460

Molecular activity

• Homeodomain (residues 239-298, PROSITE PS50071); InterPro GBX-1/2 family
  IPR042982; PANTHER PTHR24334:SF3.
• Repressor in early ectoderm, frog data: PMID:11850185
• Repressor in the PPR, frog/chick: PMID:22564795
• But an activator for otic genes: PMID:22564795 and PMID:22564795. So GO:0000981 (generic RNAPII TF) remains the right core
  MF, and GO:0001227 is added as NEW for the well-evidenced repressor mode.
  Mouse Gbx2 carries GO:0001228 activator by IDA (PMID:23144817); I did not add
  GO:0001228 for frog, since the frog activator evidence is indirect (EnR
  phenocopy only).
• Repression mechanism (cell culture plus medaka, not frog): eh1-like motif binds
  Groucho/TLE PMID:17060451 and PMID:17060451. Not used for a frog
  NEW GO:0001222, because the protein tested was not Xenopus Gbx2 and the cached
  text is abstract-only. Raised as a question.

Network position: neural plate border specifier

• Li 2009 (abstract only in cache, PMC2808295; Europe PMC full-text fetch failed
  with HTTP 500 on 2026-10-05; the falcon deep research quotes the full text):
• PMID:19736322
• PMID:19736322
• PMID:19736322
• PMID:19736322
• Deep research (full text): two MOs inhibit snail2, rescued by MO-resistant
  Gbx2; Pax3 or Msx1 RNA rescues Gbx2 MO but Gbx2 does not rescue Pax3/Msx1
  loss; at the doses used regional neural plate markers (otx2, en2) were
  largely unchanged. That last point means the NC effect is a neural-fold
  (border) patterning role, separable from neural plate AP patterning.
• Field consensus that Gbx2 is a border specifier, not a crest specifier:
  PMID:24360906 and PMID:23509273.
• Gbx2 also turns up among Pax3+Zic1-induced genes in animal caps
  PMID:24360908, so there is feedback within the border module, but Li 2009 places Gbx2
  upstream of Pax3.

Layer placement: border specifier. It is a direct Wnt target, acts upstream
of pax3/msx1 and in parallel with Zic1, and works by giving the neural fold a
posterior identity, which separates crest (posterior border) from anterior
placodal (six1) fate. It is not a crest specifier: it does not activate the crest
programme alone, and it needs Zic1.

NC-branch term decision: GO:0014029 (not GO:0014036)

• GO:0014029 neural crest formation (QuickGO definition): "The formation of the
  specialized region of ectoderm between the neural ectoderm (neural plate) and
  non-neural ectoderm."; GO:0014036 is part_of GO:0014034, which is part_of
  GO:0014029.
• Gbx2 acts on the border region as a whole: it positions the crest/placode
  split along the AP axis and acts upstream of the other border genes. By the
  pending project convention (specifiers narrowed to GO:0014036; non-specifiers
  and border genes keep GO:0014029), GO:0014029 is the right level. GO:0014036
  would claim Gbx2 makes cells autonomously crest, which Li 2009 contradicts:
  Gbx2 activity "is dependent on the interaction with Zic1".
• Added as NEW (IMP, PMID:19736322) because Q91907 has no NC term at all.

Comparator checks (QuickGO, 2026-10-05)

1. All Gbx accessions in human, mouse, rat, chick, zebrafish, X. laevis and
   X. tropicalis (30 accessions from UniProt REST query
   (gene:gbx2 OR gene:gbx1 OR gene:gbx2.1 OR gene:gbx2.2 OR ...) AND (organism_id:8355 OR 8364 OR 9606 OR 10090 OR 10116 OR 7955 OR 9031)), then:
   https://www.ebi.ac.uk/QuickGO/services/annotation/downloadSearch?geneProductId=<30 ids>&goId=GO:0014029&goUsage=descendants&goUsageRelationships=is_a,part_of,occurs_in
   gives 0 rows. With goId=GO:0014032 / GO:0014033 / GO:0001755 descendants:
   mouse Gbx2 P48031 carries GO:0001755 neural crest cell migration by IMP
   (PMID:15996652, PMID:19700621, PMID:23144817), and rat G3V8J6 by ISO.
   So the Gbx family is not systematically excluded from the NC branch. The
   frog gap reflects uncurated literature (Q91907 has no experimental rows), not
   a convention.
2. Same-layer peers, exact GO:0014029 query on Q645N4 (pax3-a), O73689 (zic1),
   Q90Z12 (hes4-a): hes4-a 4x IMP; pax3-a IMP + IGI; zic1 4x IMP + 1 IEA.
   pax3-a and zic1 additionally carry GO:0014034 (IMP/IGI). None carries
   GO:0014036. GO:0014029 for gbx2 matches its border-specifier peers.
3. Neural plate pattern terms: GO:0060897 neural plate regionalization has a
   single non-electronic annotation in all of GO (zebrafish bmp7b), and no Gbx
   carries GO:0060896/GO:0060897. Not used.
4. GO:0030917 midbrain-hindbrain boundary development (exact): mouse Gbx2 IMP
   (PMID:16651541); X. laevis irx1-a IMP (PMID:11923198), also a frog MHB
   homeobox gene; plus zebrafish fgf8a, her5 and others. Mouse Gbx2 and zebrafish
   gbx1 also carry GO:0021555 MHB morphogenesis (IMP). Supports a NEW
   GO:0030917 for frog gbx2.
5. GO:0043049 otic placode formation (exact): used for zebrafish otic TFs (foxi1,
   dlx3b/4b, msx1a/b) and X. laevis lsm14a-a; mouse Gbx2 carries GO:0042472
   inner ear morphogenesis (IMP). Supports a NEW GO:0043049 for frog gbx2.

Midbrain-hindbrain boundary / posterior neural plate

• PMID:11850185
• PMID:11850185
• PMID:11744364
• PMID:9707329. This is gain of function only; the adhesion link is
  indirect.
• Frog evidence here is mostly gain of function plus an inducible antimorph. The
  necessity evidence comes from mouse Gbx2 (human GBX2 review, PMID:10490024,
  PMID:21266408). Synthesized: MHB development is a conserved core role.

Placodes

• Otic: PMID:22564795;
  PMID:22564795; PMID:22564795
• Later cell behaviour: PMID:27659690.
  This is an engineered EnR fusion; no GO term is proposed for it.

Evolution / outgroups

• Amphioxus has a single Gbx with a posterior domain abutting Otx PMID:16687133 and
  PMID:16687133. The Otx/Gbx AP-positioning function is
  ancestral to chordates; organizer properties are vertebrate.
• PMID:18836256. Gbx is a positional
  (AP) gene in amphioxus; I found no report of amphioxus Gbx being specifically
  border-restricted.
• Ciona has lost Gbx PMID:12736825. The
  tunicate border/"rudimentary NC" (a9.49) therefore works without Gbx, which
  argues that Gbx2's NC role is not a requirement of the ancestral border
  programme. Caveat: Ciona has lost many genes.
• Lamprey: gbx2 shows NPB dynamics like Xenopus PMID:39060477. In the same analysis myc,
  pax3, msx1 and zic1 had these dynamics in Xenopus but not lamprey. Gbx2 at the
  border is therefore at least as old as crown vertebrates. PubMed eutils for
  Gbx[tiab] AND lamprey[tiab] returned 0 hits, so no lamprey functional data.
• Interpretation: Gbx protein activity (posterior homeodomain repressor, Otx
  antagonist) is chordate-ancestral. Its use to split the border into posterior
  crest versus anterior placode looks like a vertebrate deployment of an
  ancestral AP-patterning gene. This is an "ancestral activity, new context"
  case, like Id/SoxE in Tier 1, except that the gene was already in the
  ectoderm.

Decisions summary

• All MF IBA/IEA rows ACCEPT except GO:0003677 (MODIFY to GO:0000977, as in
  human GBX2).
• GO:0006355 MODIFY to GO:0006357 (as in human).
• GO:0051960 IBA/IEA: KEEP_AS_NON_CORE (as in human; true but broad; the
  specific processes are added as NEW).
• Location rows ACCEPT.
• NEW: GO:0014029 (IMP PMID:19736322), GO:0001227 (IMP PMID:11850185 plus
  PMID:22564795), GO:0030917 (IMP PMID:11850185), GO:0043049 (IMP
  PMID:22564795).
• Not added: GO:0001228 (frog activator evidence indirect), GO:0001222
  (TLE binding shown with non-frog protein), GO:0060897 (essentially unused
  term), GO:0014036 (wrong layer), cell adhesion or migration terms (indirect).

Homeolog caveat

All NEW rows sit on Q91907 (gbx2.2.L). The morpholinos (Li 2009) and Tour's
Xgbx2a constructs cannot be mapped to a single copy from the abstracts, and the
four X. laevis gbx2 copies are 93-99.7% identical. The same evidence plausibly
covers gbx2.2.S and the gbx2.1 pair. Not fixed by assertion.