AIGR Gene Hypothesis Deep Research — Final Report

Target: Drosophila melanogaster rdgBbeta (UniProt Q9U9P7)

Hypothesis: Phosphatidylcholine transporter activity (GO:0008525)

Focus type: computational_prediction (BioReason-Pro SFT)

Summary

The BioReason-Pro SFT model predicts phosphatidylcholine transporter activity
(GO:0008525) for the Drosophila melanogaster protein rdgBbeta (Q9U9P7). After
three iterations of literature review, sequence/orthology analysis, and a
family-wide GO annotation audit, this prediction is REFUTED. rdgBbeta is a
single-domain Class II phosphatidylinositol transfer protein (PITP) — the
Drosophila ortholog of human PITPNC1 (the "vibrator"/RdgBβ clade) — and the
direct biochemistry of this clade shows it binds and transfers
phosphatidylinositol (PI) and phosphatidic acid (PA) but "hardly binds
phosphatidylcholine" (PMID: 22822086).

Phosphatidylcholine transfer is instead a defining property of a different
branch of the family: the Class I PITPs (PITPα/PITPβ) with dual PI/PC
specificity, and the multidomain/Sec14-type transfer proteins. A systematic
audit of GO:0008525 across human, mouse, rat, fly, and yeast PITPs found direct
experimental (IDA) support only in those PC-competent branches (PITPNA, PITPNB,
Drosophila multidomain rdgB, yeast SEC14) — and no experimental support
anywhere in the Class II clade. rdgBbeta itself carries no GO:0008525
annotation; its curated molecular functions are PI binding (GO:0035091, IBA)
and PI transfer activity (GO:0008526, ISS), which are the accurate terms.

The prediction is best explained as paralog / substrate misassignment: a
PC-transfer trait that is genuine for Class I and multidomain paralogs has been
incorrectly propagated onto the Class II member. The main caveat is that the
decisive "no PC" assay was performed on the mammalian ortholog PITPNC1, not on
the Drosophila protein directly, so the refutation rests on strong orthology
(61.6% identity to PITPNC1 vs 42.7% to Class I PITPNA) plus concordant clade-level
biochemistry rather than a fly-specific PC-transfer assay. Even so, no evidence
supports PC transport and direct evidence argues against it. Curator lead:
reject GO:0008525; retain GO:0008526 (PI transfer) as the core molecular
function.

Key Findings

Finding 1 — RdgBβ/PITPNC1 binds and transfers PI and phosphatidic acid, but hardly binds phosphatidylcholine

Direct in vitro lipid binding and transfer assays on the Class II PITP
RdgBβ (PITPNC1) demonstrate that, besides phosphatidylinositol, the protein
binds and transfers phosphatidic acid (PA) and is recovered pre-loaded with
PA / phosphatidylglycerol when expressed in E. coli — but it "hardly binds
phosphatidylcholine" (PMID: 22822086,
Garner et al. 2012). This is the single most decisive piece of evidence against
the seed hypothesis, because it assays exactly the activity in question (lipid
transfer specificity) on exactly the clade to which rdgBbeta belongs.

The result is corroborated by an independent Class II ortholog: recombinant
mouse mM-rdgBβ1 "shows the specific binding activity to phosphatidylinositol
but not to other phospholipids" (PMID: 12562526).
Two independent Class II proteins, characterized in two laboratories, converge on
the same conclusion: PI (and, for PITPNC1, PA) is the physiological cargo, and PC
is not a bound/transferred ligand.

For contrast, PC binding is a real property of the Class I PITPs: in
Class I PITPα/β "the preferred lipid that can occupy the site can be either
phosphatidylinositol (PI) or phosphatidylcholine (PC)"
(PMID: 10358925). This dual PI/PC
specificity — quantified elsewhere as an ~16-fold preference for PI over PC with
measurable PC transfer (PMID: 3651458) —
is precisely the trait a PC-transporter GO term captures, and it is a Class I /
Sec14-type feature, not a Class II feature. The prediction thus maps a real
biochemical activity onto the wrong branch of the family.

Finding 2 — Drosophila rdgBbeta (Q9U9P7) is a single-domain Class II PITP orthologous to human PITPNC1

UniProt Q9U9P7 is annotated PITC1_DROME, "Cytoplasmic phosphatidylinositol
transfer protein 1", a 273-residue protein consisting of a single PITP
domain (Pfam PF02121 IP_trans; InterPro IPR001666 PI_transfer; PANTHER
PTHR10658:SF54). It lacks the additional regulatory/membrane-targeting domains
(FFAT, DDHD, LNS2, transmembrane region) that characterize the large multidomain
RdgBα/Nir-type proteins.

Global pairwise Needleman–Wunsch alignments computed in this investigation place
rdgBbeta firmly in the Class II clade: 61.6% identity to human PITPNC1
(Class II) versus only 42.7% identity to human PITPNA (Class I). This
~19-point identity gap is a clear orthology signal — rdgBbeta is the fly
ortholog of the mammalian "vibrator"/RdgBβ protein PITPNC1, not of the
PC-competent Class I PITPs.

The single-domain Class II identity is confirmed in the review literature:
"Three members of the PITP family (PITPα, PITPβ, and RdgBβ (retinal degeneration
type B) alt. name PITPNC1) are present as single domain proteins"
(PMID: 23086419). Critically, the
current UniProt GO annotation for Q9U9P7 already reflects the correct
biochemistry: phosphatidylinositol binding (GO:0035091, IBA) and
phosphatidylinositol transfer activity (GO:0008526, ISS) — and no
GO:0008525 is present. The predicted PC-transporter term would therefore be an
addition that contradicts the existing, more accurate PI-transfer annotation.

Finding 3 — GO:0008525 has direct evidence only for Class I / multidomain PITP paralogs, never for the Class II PITPNC1/rdgBbeta clade

A systematic QuickGO/GOA enumeration of GO:0008525 (phosphatidylcholine
transporter activity) across human, mouse, rat, fly, and yeast PITP-family
proteins reveals a sharp phylogenetic pattern. Direct experimental (IDA)
support for the PC-transporter term exists only for the PC-competent branches:

• Class I PITPs: human PITPNA (Q00169) and PITPNB (P48739) and their rodent
  orthologs;
• Drosophila multidomain rdgB (P43125, IDA);
• Yeast SEC14 (P24280, IDA) — the founding dual PI/PC transfer protein, whose
  in vitro exchange of PI or PC is textbook (PMID: 15299870).

By contrast, the entire Class II clade carries a single GO:0008525
annotation — rat Pitpnc1 (A0A8I6A182), and that annotation is ISO only
(inferred electronically from an ortholog), with no IDA/EXP support anywhere
in the clade. Drosophila rdgBbeta (Q9U9P7) carries no GO:0008525 annotation
at all. Within Drosophila specifically, PC-transporter activity is annotated
to the paralogs rdgB (IDA) and vib/vibrator (IBA) — not to rdgBbeta.

This is the fingerprint of paralog over-annotation. The one Class II GO:0008525
record in existence is itself a non-experimental electronic inference that most
plausibly originated by carry-over from a Class I / multidomain paralog. A model
trained on such annotation landscapes can readily reproduce the same
misassignment for rdgBbeta. (Provenance: computed enumeration saved to
/tmp/go0008525_pitp_family.csv.)

Mechanistic Model / Interpretation

The PITP family splits into functionally distinct branches, and the seed
hypothesis fails because it ignores that split. The molecular function under test
is monomeric, headgroup-specific lipid transfer by a soluble PITP-domain protein:
extract a lipid, shield its headgroup inside a hydrophobic cavity, and deposit it
at an acceptor membrane. The discriminating variable is headgroup specificity
(PC vs PI/PA).

PITP family map and PC-transfer competence

         PITP superfamily
               |
-----------------------------------------------
|                      |                        |
   Class I PITPα/β     Multidomain RdgBα         Class II  <-- rdgBbeta (Q9U9P7)
   PITPNA / PITPNB     rdgB / SEC14 / Nir2       PITPNC1 / vibrator
|                      |                        |
   PI + PC transfer     PI + PC transfer          PI + PA transfer
   GO:0008525 (IDA)     GO:0008525 (IDA)          "hardly binds PC"
                                  GO:0008526 (PI transfer)
                                  << NO GO:0008525 >>

   Prediction path (WRONG):  Class I / multidomain PC-transfer trait
             ─────propagated across paralogs────►  rdgBbeta

Immediate molecular function. The evidence indicates the direct molecular
activity of Q9U9P7 is PI transfer (GO:0008526) and, by clade analogy with
PITPNC1, PA binding/transfer — with PC essentially excluded from the binding
pocket. PC transfer is therefore not this gene product's activity; it is the
activity of sister paralogs. Downstream roles of PITPs (phosphoinositide
signaling, photoreceptor/neural signal transduction) are pathway consequences,
not the primary molecular activity, and do not implicate PC transport.

Why the model likely erred. Three compounding factors: (1) frequency /
paralog bias — most experimentally characterized PITPs (Class I, SEC14) are
PC-competent, so "PC transporter" is the family-common label; (2) shared PITP
domain — the single conserved fold (PF02121) is common to PC-competent and
PC-incompetent members, so domain-level features do not discriminate; (3)
annotation carry-over — the one existing Class II GO:0008525 record is an ISO
electronic inference, providing a spurious training signal. None of these reflect
the measured biochemistry of the Class II clade.

Evidence Base / Evidence Matrix

GO Curation Implications

Leads (require curator verification):

1. Do NOT add GO:0008525 (phosphatidylcholine transporter activity) to
   rdgBbeta (Q9U9P7). The computational prediction is contradicted by direct
   assays on the clade ("hardly binds phosphatidylcholine",
   PMID: 22822086) and by the absence
   of any experimental PC-transporter evidence in the entire Class II clade.
   Recommend reject the prediction.

2. Retain the existing, more accurate MF annotations:

3. GO:0008526 — phosphatidylinositol transfer activity (currently ISS): the
   correct, clade-consistent molecular function and core MF.

4. GO:0035091 — phosphatidylinositol binding (currently IBA): consistent
   with clade biochemistry.

5. Consider (lead) a phosphatidic-acid–related MF as a more informative
   activity than PC transfer, given that PITPNC1 binds/transfers PA
   (PMID: 22822086). Any such term
   should be applied only with an ISS/ISO evidence code from PITPNC1 and flagged,
   since the direct PA assay was on the mammalian ortholog; Drosophila-specific
   evidence is lacking, so treat as non-core for DROME for now.

6. Treat GO:0008525 as non-core for this gene and flag the single existing
   Class II ISO GO:0008525 record (rat Pitpnc1) as a candidate paralog-carry-over
   error.

The evidence supports an MF decision (lipid-transfer specificity), not a BP or
CC change. "Protein binding" is explicitly not recommended — the informative
term (PI transfer, GO:0008526) is already supported.

Mechanistic Scope

The immediate function tested is headgroup-specific monomeric lipid transfer.
The seed prediction (PC transporter) conflates the Class II RdgBβ pocket
specificity (PI/PA) with the Class I PITPα/β dual PI/PC specificity. Everything
downstream — phosphoinositide signaling, secretory/Golgi function attributed to
some RdgB proteins, photoreceptor/neural signal transduction — is a pathway or
developmental consequence, not the primary molecular activity, and none of it
implicates PC transport by rdgBbeta. Importantly, rdgBbeta (small, cytosolic,
PITPNC1-type) must not be confused with rdgB-alpha (large multidomain
PITPNM/Nir-type); they are distinct paralogs with distinct biology, and it is the
alpha/multidomain and Class I paralogs — not rdgBbeta — that carry the direct
PC-transfer evidence.

Conflicts and Alternatives

• Paralog confusion (primary alternative and likely source of the prediction).
  PC-transporter activity is genuine for Class I PITPα/β
  (PMID: 10358925), for multidomain
  Drosophila rdgB (IDA), and for yeast SEC14
  (PMID: 15299870). These are
  paralogs of rdgBbeta, not rdgBbeta itself. The model most plausibly
  generalized a family-common PC-transfer trait onto a Class II member that lacks
  it.
• Isoform / organism-specific gaps. The strongest "no PC" assay
  (PMID: 22822086) used mammalian
  PITPNC1; the mouse Class II assay (PMID: 12562526)
  used mM-rdgBβ1. No published in vitro PC-transfer assay on the Drosophila Q9U9P7
  protein itself was located. The refutation rests on strong orthology plus
  concordant clade biochemistry rather than a fly-specific assay.
• Quantitative "hardly" vs absolute exclusion. "Hardly binds PC" is a low but
  non-zero statement; under artificial vesicle conditions trace PC association is
  conceivable. This does not justify a core PC-transporter MF annotation,
  which implies a physiological transport function.
• Database carry-over. The single Class II GO:0008525 record (rat Pitpnc1,
  ISO) is electronic and unverified; it should not be treated as independent
  support and may share the same error origin as the prediction.

Limitations and Knowledge Gaps

1. No Drosophila-protein-specific PC-transfer assay or functional phenotype
   paper. Checked: PubMed for CG17818/rdgBbeta biochemistry and
   photoreceptor/PLC function; GOA. Why it matters: specificity is inferred from
   the mammalian ortholog. Resolver: in vitro PI/PC/PA transfer assay with
   purified recombinant Drosophila rdgBbeta. (No experimental PDB exists for
   Q9U9P7, but a high-confidence AlphaFold model is available, global pLDDT ≈ 91.5,
   so structural pocket analysis is immediately feasible.)
2. PA-transfer relevance for the fly protein is inferred, not measured.
   Checked: PMID 22822086 (PITPNC1). Why it matters: a PA-related MF would be
   the most informative substitute for the rejected PC term. Resolver: PA
   binding/transfer assay on Q9U9P7 or structural pocket analysis.
3. Provenance of the single Class II ISO GO:0008525 record. Checked:
   QuickGO/GOA enumeration. Why it matters: if it seeded the prediction, it is a
   correctable pipeline artifact. Resolver: trace the ISO "with/from" source
   protein.
4. Provenance of the BioReason-Pro prediction (whether it used Class I
   features) is unverified; matters for judging systematic paralog bias.

Discriminating Tests / Proposed Follow-up Experiments

1. In vitro fluorescent-lipid transfer assay (pyrene-PI vs pyrene-PC vs
   NBD-PA) with purified recombinant DROME rdgBbeta — the single most decisive
   experiment. Expectation if refuted: robust PI (and PA) transfer, negligible PC
   transfer, mirroring PMID: 22822086.
2. Mass-spectrometric lipidomics of the co-purified ligand (as done for
   PITPNC1, which co-purified PA/PG). Absence of PC in the bound pool directly
   counters the prediction.
3. Structural pocket comparison (AlphaFold model of Q9U9P7 vs Class I PITPα
   holo-structures vs PITPNC1): identify headgroup-coordinating residues and test
   whether the pocket can accommodate the PC choline headgroup.
4. Rescue/complementation: does rdgBbeta substitute for Class I PITP
   PC-dependent functions? A negative result supports non-PC specificity.
5. Annotation-provenance audit: trace the rat Pitpnc1 ISO GO:0008525 record's
   source and any BioReason-Pro training exposure to Class I / multidomain PITP
   PC-transporter annotations, to confirm the paralog-carry-over mechanism.

Curation Leads (require curator verification)

• Action change: Reject/omit GO:0008525 for rdgBbeta; do not import the
  BioReason-Pro SFT prediction.
• Retain: GO:0008526 (phosphatidylinositol transfer activity, ISS) and
  GO:0035091 (phosphatidylinositol binding, IBA) as the accurate MF core.
• Candidate new/replacement term (lead): a phosphatidic-acid
  binding/transfer MF term supported by ISS/ISO from PITPNC1
  (PMID: 22822086), clearly flagged
  as ortholog-inferred and non-core for DROME.
• Candidate reference + snippet to verify: PMID: 22822086
  — "besides PI, RdgBβ binds and transfers phosphatidic acid (PA) but hardly binds
  phosphatidylcholine."
• Candidate reference + snippet to verify: PMID: 12562526
  — "The recombinant mM-rdgBbeta1 protein shows the specific binding activity to
  phosphatidylinositol but not to other phospholipids."
• Paralog-source reference: PMID: 10358925
  — "The preferred lipid that can occupy the site can be either phosphatidylinositol
  (PI) or phosphatidylcholine (PC)" (establishes PC transfer as a Class I trait).
• Suggested curator question: Was GO:0008525 propagated from Class I /
  multidomain PITP paralogs (rdgB, vib, PITPNA/B), and should the single Class II
  ISO record (rat Pitpnc1) also be flagged?
• Suggested experiment: direct DROME rdgBbeta PI/PC/PA transfer assay.

Conclusion

The prediction of phosphatidylcholine transporter activity (GO:0008525) for
Drosophila rdgBbeta (Q9U9P7) is REFUTED. rdgBbeta is a single-domain
Class II PITP orthologous to human PITPNC1 (61.6% identity vs 42.7% to Class I
PITPNA). Direct biochemical characterization of this clade shows it binds and
transfers phosphatidylinositol and phosphatidic acid but "hardly binds
phosphatidylcholine" (PMID: 22822086);
PC transfer is a Class I / multidomain (Sec14-type) property. The prediction is a
paralog / substrate misassignment, reinforced by the fact that GO:0008525 has
direct (IDA) support only in Class I and multidomain paralogs and none in the
Class II clade. Curators should reject GO:0008525 and retain the more accurate
phosphatidylinositol transfer activity (GO:0008526).

Provenance: UniProt Q9U9P7 JSON parse and Needleman–Wunsch identity computation
were executed in this investigation (rdgBbeta vs human PITPNC1 = 61.6%; vs PITPNA
= 42.7%). GO:0008525 family enumeration saved to /tmp/go0008525_pitp_family.csv.
Literature retrieved via PubMed.

Artifacts

• OpenScientist final report
• OpenScientist final report