A1BG (alpha-1B-glycoprotein) — review notes

Reviewed as part of the PAINT human no-IBA project, using the affinage deep-research
provider (A1BG-deep-research-affinage.md) plus UniProt P04217, the GOA TSV, and the
primary literature.

What the protein is

A1BG is an abundant secreted plasma glycoprotein of unknown catalytic function. The
mature chain (residues 22–495) follows a cleaved signal peptide and is built from five
tandem immunoglobulin-like domains; there is no transmembrane segment anywhere in the
protein and UniProt lists SUBCELLULAR LOCATION: Secreted with keyword Secreted
[file:human/A1BG/A1BG-uniprot.txt, "SUBCELLULAR LOCATION: Secreted."].

The complete protein sequence was determined directly from plasma-purified protein
[PMID:3458201, "The complete amino acid sequence has been determined for alpha
1B-glycoprotein"], which also established the five-domain internal duplication and the
immunoglobulin-superfamily relationship [PMID:3458201, "consists of five repeating
structural domains, each containing about 95 amino acids and one disulfide bond"].

Why the IBA annotations are wrong for this gene

This is the central curation finding for A1BG.

PANTHER places A1BG in PTHR11738 ("MHC class I NK cell receptor"), the LILR/KIR
leukocyte-receptor family (subfamily PTHR11738:SF184), and InterPro assigns
IPR016332 A1B_glyco/leuk_Ig-like_rcpt and IPR050412 Ig-like_Receptors_ImmuneReg
[file:human/A1BG/A1BG-uniprot.txt, "PANTHER; PTHR11738; MHC CLASS I NK CELL RECEPTOR; 1."].
The other members of that family are type-I transmembrane immune receptors. A1BG is the
divergent, soluble, secreted member — it kept the Ig ectodomain repeats but has no
transmembrane anchor and no cytoplasmic signalling tail.

The three GO_Central IBA annotations therefore propagate receptor biology that A1BG
physically cannot perform:

The first two are straightforward PROPAGATION_BAD: the source annotations are correct
for the transmembrane receptors they came from, but the term must not transfer to the
secreted member.

The growth-hormone IBA is a source problem, not a transfer problem

I initially assumed GO:0060396 was another family over-propagation, but tracing the
WITH/FROM field disproves that. It cites PANTHER:PTN000200788 and MGI:MGI:2152878 —
and MGI:2152878 is mouse A1bg itself, i.e. the true ortholog, which carries an MGI
IMP to the same term. So the transfer is mechanically legitimate; the defect is in the
source annotation.

The mouse IMP rests on PMID:16723264, which isolated A1bg ("cDNA #5") in a subtractive
screen for transcripts induced in the liver of GH-transgenic mice, and found it absent from
GH-deficient dwarf mice, GH-antagonist mice, and GH-receptor knockouts
[PMID:16723264, "These findings suggest that induction of mRNA #5 in the liver requires a
continuous pattern of GH secretion and an intact GH-GH receptor-signaling complex."].

That demonstrates A1bg is a transcriptional target downstream of the GH axis. GO:0060396
is defined as the molecular signals generated as a consequence of the GH receptor binding
growth hormone — a GH-inducible secreted liver product is not part of that cascade. This is
a ROLE_CONFLATION at the source (response-to-X read as participates-in-X-pathway). The
defensible term for the mouse data would be a response term such as GO:0060416 response to growth hormone; there is no human evidence for even that, so the human IBA is removed
rather than modified. Flagged as a suggested_question so the mouse source annotation can
be re-curated upstream.

All three IBAs are therefore recommended for REMOVE, each with a propagation_review
recording the root cause. None of these is second-guessing a human experimental curator.

What A1BG actually does

A1BG works through its Ig-like domains as a plasma binding/sequestering partner for
CAP-superfamily (CRISP) proteins.

• CRISP3. CRISP3 circulates in plasma bound to A1BG; the complex is 1:1, noncovalent,
  electrostatically held, with a nanomolar dissociation constant
  [PMID:15461460, "We demonstrate that CRISP-3 is a specific and high-affinity ligand of A1BG
  with a dissociation constant in the nanomolar range as evidenced by surface plasmon
  resonance."]. UniProt records this as the sole curated SUBUNIT interaction
  [file:human/A1BG/A1BG-uniprot.txt, "Interacts with CRISP3."]. The same paper notes that the
  analogous complexes between snake-venom toxins and A1BG-like opossum plasma proteins
  neutralise toxin activity — i.e. this protein family is a plasma sequestering/neutralising
  system.
• CRISP2 / CAP proteins. Co-expression of A1BG with CAP proteins abolished their sterol
  export function and A1BG inhibits their sterol binding in vitro
  [PMID:39433128, "Coexpression of A1BG with CAP proteins abolished their sterol export
  function in yeast and their interaction inhibits sterol-binding in vitro."]. The contact is
  mapped to the third of the five Ig-like domains and requires Mg²⁺
  [PMID:39433128, "We map the interaction between A1BG and CRISP2 to the third of five
  repeated immunoglobulin-like domains within A1BG."].

This is a coherent molecular function — binding a partner protein so as to block that
partner's own ligand-binding/export activity — and maps onto GO:0140311 protein sequestering activity ("Binding to a protein to prevent it from interacting with other partners or to
inhibit its localization to the area of the cell or complex where it is active"). It is
proposed as a NEW annotation because GOA currently gives A1BG no molecular function at
all other than the incorrect receptor IBA.

Other reported activities (recorded, not annotated as core)

• NAMPT stabilisation / chemoresistance. Adipocyte-secreted A1BG binds NAMPT and stabilises
  it, raising NAD⁺ and boosting PARP1-dependent DNA repair, driving cisplatin resistance in
  osteosarcoma [PMID:40560034, "Further investigation revealed a direct interaction between
  A1BG and NAMPT, leading to the stabilization of NAMPT and an increased NAD+ production."].
  Single-paper, cancer-context; noted in suggested_questions rather than annotated.
• Female-specific cardiac role. Cardiomyocyte-restricted A1bg knockout gives dilated
  cardiomyopathy in female but not male mice [PMID:40270023, "The glycoprotein A1BG has emerged
  as a female-specific regulator of cardiac structure and integrity"]. Mouse phenotype; not
  used for a human GO process annotation here.

Localisation calls

GO:0005576 extracellular region is the core location and is supported by direct isolation
from plasma (IDA), by UniProt SubCell, and by several independent proteomics datasets. The
granule-lumen terms (GO:0034774, GO:1904813, GO:0031093) come from Reactome
degranulation reactions and the exosome/blood-microparticle terms from HDA proteomics; all are
consistent with an abundant secreted plasma protein but are peripheral, so they are kept as
non-core.

Affinage assessment

The affinage report passed its trust gates at fetch time (accession P04217 matching the local
UniProt record, no non-human organism token; it carries no self_evaluation_pairwise score) and its narrative — secreted Ig-domain plasma protein acting through
protein–protein interactions with CRISP2 and NAMPT — agrees with the primary literature. Its
own mechanism_profile grounding (GO:0098772 molecular function regulator activity,
GO:0005576 extracellular region) is correct but coarse; GO:0140311 was chosen instead as
the specific descendant supported by the CRISP2/CRISP3 experiments. Notably the affinage
narrative independently contains nothing about membrane receptors or growth-hormone
signalling, corroborating the IBA removals.

Limitation worth recording: affinage gave no signal at all about the incorrect IBAs — it
simply does not mention them. The IBA root-cause analysis above required going outside the
deep-research report, to the GOA WITH/FROM field, the mouse ortholog's own GO record, and
the source paper. Affinage is useful for establishing what a gene does; it does not audit
what GO already says.

Full IBA re-review, 2026-09-20

All 18 rows (17 source assertions and one pre-existing NEW MF) were re-reviewed. All original source identities, evidence codes and qualifiers are retained. The exact current P04217 leaf PTN002482657 descends from membrane/receptor IBD PTN002621170 and GH IBD PTN000200788 in PTHR11738; extracted lineage and response hash are retained in the project audit.

The previous exclusive-soluble argument missed a direct experiment: [PMID:40270023, Fig. 4D, "A1BG is associated with the outer surface of atrial CMs."]. This is mouse A1BG/tropomyosin co-immunostaining, not a human IDA result, but it removes the supposed impossibility of membrane association. Plasma membrane REMOVE becomes KEEP_AS_NON_CORE; receptor and GH pathway REMOVE become UNDECIDED. No-TM architecture raises a mechanistic question about receptor capacity, not a prohibition of all membrane activity. The full text behind the mouse GH IMP remains unavailable, so its abstract cannot justify SOURCE_BAD. A focused OpenScientist hypothesis was launched after a negative exact-target cache check.

PMID:39433128 full Results establish selective CRISP2 sterol-binding inhibition and >50% reduction of reconstructed yeast sterol export, with preserved fatty-acid activity; A1BG itself does not bind cholesterol sulfate. Retain the existing protein-sequestering MF proposal independently of the receptor judgment. Do not import venom neutralization by opossum proteins as a human core function. PMID:40560034 Fig. 5 concerns cell-associated NAMPT turnover, not specifically secreted NAMPT; co-IP and colocalization are association evidence. Reactome granule and proteomics locations remain curated contextual pools, without speculative contamination claims. The Affinage report was fully read and its omissions are not negative evidence.

Focused OpenScientist incorporation, 2026-09-20

The completed focused report, HTML/PDF and all computational artifacts were assessed independently. No action changes: retain plasma membrane KEEP_AS_NON_CORE and receptor/GH UNDECIDED. The report's categorical negatives exceed its evidence: it misses the explicit mouse outer-surface experiment, uses family membership without tracing the topology, and calls cell-associated NAMPT stabilization extracellular/direct binding. Its canonical hydropathy maximum reproduces (0.9842), but this does not exclude peripheral association or receptor-complex function. The full-GH-paper access claim has no independently verifiable supplied source beyond abstract-derived statements. The GH expression-versus-participation question remains a primary-full-text/curator follow-up, not a reason to launch the same report again.

See critical assessment and reproducible hydropathy check. Findings-specific report reviews cite the correcting primary experiments. The existing CRISP2 sequestration proposal and qualified core remain supported by the full 39433128 assays; no new annotation, negation, source mutation or additional biological exclusivity claim was introduced.

Recovery PR localization follow-up (2026-09-22)

Preserve superseded donor-specific analysis where present, replace rebutted report passages with actual supporting evidence, and clarify location or process scope. PAINT rows support inherited assertions as phylogenetic judgments, not direct target experiments.