AI Gene Review — Focused Hypothesis Report

K9IMD0 (TRLF_DESRO), Desmodus rotundus lactotransferrin / draculin

Focus: function_assignment — "lactotransferrin functions and draculin specialization"

Executive Judgment

Verdict: PARTIALLY SUPPORTED / largely OVER-ANNOTATED (family carry-over).

K9IMD0 is unambiguously draculin, and draculin has been explicitly established
as "a mutated version of the lactotransferrin scaffold" (PMID:23748026). The seed
hypothesis asks whether this protein still directly holds a battery of inherited
transferrin/lactoferrin GO terms independent of its anticoagulant specialization.
The evidence splits sharply:

• Experimentally supported for THIS protein (retain, core): secretion /
  extracellular region (GO:0005576, IDA) and negative regulation of blood
  coagulation (GO:0030195, IDA) — FXa/FIXa inhibition (PMID:9795244, PMID:10556567).
  Anticoagulation is the primary, demonstrated function.
• Structurally defensible but not assayed in bat (retain only as ISS/IEA with a
  species caveat): metal-ion binding (GO:0046872) — the N-lobe iron site is fully
  conserved; and serine-type peptidase (GO:0008236) — the nucleophile Ser is
  conserved but one essential catalytic residue is altered (Lys→Arg).
• Weak / electronic carry-over (generalize, flag non-core, or remove pending
  evidence): iron ion transport (GO:0006826/0006811) — the C-lobe iron site is
  broken (Y454→D452) so bilobal transport is degraded; antibacterial humoral
  response / defense-to-bacterium / immune-system-process (GO:0019731/0042742/0002376);
  and the endosome / recycling-endosome / plasma-membrane locations
  (GO:0005769/0055037/0005886), which derive from mammalian lactoferrin receptor
  recycling and have no support for a secreted salivary bat protein.

Most important caveat: every non-anticoagulant term is IEA:TreeGrafter or a
family keyword — none rests on a direct assay of draculin. The seed's framing is
correct that "anticoagulant specialization does not exclude inherited functions,"
but the molecular evidence shows active degradation of two ancestral active sites
(C-lobe iron Tyr; catalytic Lys of the protease dyad), so inheritance cannot be
assumed and should be curated conservatively.

Evidence Matrix

GO Curation Implications (leads — require curator verification)

See K9IMD0_GO_decision_table.csv (computed provenance). Summary:

• RETAIN (core, experimental): GO:0005576 extracellular region (CC, IDA);
  GO:0030195 negative regulation of blood coagulation (BP, IDA). These are the
  protein's demonstrated identity and primary function.
• RETAIN as ISS/IEA with explicit "not verified in D. rotundus" caveat, treat
  as non-core: GO:0046872 metal ion binding (N-lobe site intact — a bona fide
  structural feature); GO:0008236 serine-type peptidase (nucleophile Ser conserved).
• GENERALIZE / flag non-core / candidate for removal pending evidence:
  GO:0006826 & GO:0006811 iron/ion transport (C-lobe site broken; no bat transport
  assay; a secreted protein without the receptor-recycling context); GO:0008233
  peptidase, GO:0016787 hydrolase, GO:0006508 proteolysis (downstream/too-broad
  parents of an unverified activity); GO:0019731/0042742/0002376
  antibacterial/defense/immune (family carry-over + transcriptomic grouping only).
• REMOVE or down-weight to peripheral-IEA: GO:0005769 early endosome,
  GO:0055037 recycling endosome, GO:0005886 plasma membrane — these describe the
  mammalian lactoferrin-receptor recycling itinerary and are not supported for a
  secreted salivary bat protein. (The seed's point that "missing TM topology does
  not exclude peripheral association" is fair, but there is no positive evidence
  of a membrane/endosomal pool of draculin.)

We avoid recommending "protein binding" as a terminal term; the informative core
terms are the anticoagulation BP and extracellular CC.

Mechanistic Scope

• Immediate molecular function (demonstrated): noncompetitive tight-binding
  inhibition of activated coagulation factors Xa and IXa (Ki ≈ 15 nM), requiring
  correct glycosylation. This is the direct gene-product activity.
• Inherited scaffold features (molecular, partly degraded): a bilobal
  transferrin fold with an intact N-lobe metal-binding pocket and a broken C-lobe
  pocket; a conserved protease nucleophile Ser but altered catalytic Lys.
• Downstream / inferred-only (not direct activity): iron transport (a
  systemic/cellular process needing receptor recycling), antibacterial response,
  and endosomal/plasma-membrane localization — all inferred from human lactoferrin
  biology, not observed for draculin.

Conflicts and Alternatives

• Database carry-over (main risk): TreeGrafter/PANTHER propagation and
  keyword→GO mapping assign the full lactoferrin repertoire to a protein the same
  authors call "mutated." Direct measurements exist only for anticoagulation.
• Organism/paralog specificity: all protease/antibacterial/iron-transport
  measurements are on human lactoferrin (PMID:12535064/9770539/11163480), not
  draculin. Two catalytic-site substitutions (C-lobe iron Tyr; protease Lys) argue
  against naive transfer.
• Interpretation of the cationic N-terminus: draculin retains a lactoferricin-like
  basic N-terminal stretch (net +9), so a peptide-based antimicrobial activity is
  plausible — a genuine alternative route to GO:0019731 that does not depend on
  the protease dyad. This remains untested.

Knowledge Gaps

1. Does draculin bind iron? Checked: sequence/structure only (N-lobe intact,
   C-lobe Tyr lost). Matters because it gates GO:0046872/0006826. Resolve with a
   spectrophotometric Fe³⁺-saturation / urea-gel iron-binding assay on recombinant
   or salivary draculin.
2. Does draculin have protease activity? Checked: dyad mapping (Ser conserved,
   Lys→Arg). Matters for GO:0008236/0008233/0016787/0006508 and mechanistically for
   antibacterial. Resolve with an H. influenzae IgA1-protease/Hap cleavage assay
   ± serine-protease inhibitors, as in PMID:9770539.
3. Is draculin antibacterial? Checked: only transcriptomic grouping
   (PMID:23411029). Resolve with MIC/killing assays vs Gram+/– bacteria.
4. Any endosomal/plasma-membrane pool? Checked: none found; IEA only. Resolve
   with cell-surface receptor-binding / uptake assays in a relevant cell type.

Discriminating Tests

• Iron binding: UV-vis (465 nm holo-transferrin band) and urea-PAGE mobility
  shift on apo vs Fe-loaded recombinant draculin; compare N-lobe-only vs full-length.
• Protease: recombinant draculin (and a draculin-R92K back-mutant) against
  IgA1-protease/Hap substrates; serine-protease-inhibitor panel.
• Antibacterial: killing/MIC assays and a draculin N-terminal peptide
  (lactoferricin-analog) test to separate peptide- vs protease-based activity.
• Localization: anti-draculin surface staining / receptor pulldown to test any
  membrane/endosomal association.
• Comparative genomics: map the C-lobe Y→D and protease K→R substitutions across
  bat lactoferrins to test whether they are draculin-lineage specialization vs
  species polymorphism.

Curation Leads (require curator verification)

Candidate references to attach (with exact snippets to verify):
- PMID:23748026 — verify: "…a very large sequence stretch of draculin and thus
established that it is a mutated version of the lactotransferrin scaffold"
(identity of K9IMD0 = draculin; supports functional divergence).
- PMID:12535064 — verify: "mutation of either Ser259 or Lys73 results in a dramatic
decrease in proteolysis" (basis of the serine-protease term; note Lys→Arg in bat).
- PMID:9770539 — verify: "…localized to the N-lobe of the bilobed lactoferrin
molecule and were inhibited by serine protease inhibitors…" (protease↔antibacterial link).
- PMID:9795244 / PMID:10556567 — anticoagulant core function (FIXa/FXa inhibition).
- PMID:23411029 — lactotransferrin among accessory-gland antimicrobials (localization/expression, review-of-glands level).

Candidate action changes:
- Keep GO:0030195 and GO:0005576 as core/experimental.
- Downgrade GO:0006826/0006811, GO:0008233/0016787/0006508, GO:0019731/0042742/0002376
to non-core, IEA/ISS with caveats; consider removing the CC endosome/PM terms
(GO:0005769/0055037/0005886) unless localization evidence is produced.
- Where retained (GO:0046872 metal binding; GO:0008236 serine peptidase), annotate
with an explicit note that the site is structurally present but one essential
residue diverges and the activity is unverified in D. rotundus.

Suggested curator questions: Is species-blinded IEA propagation appropriate for a
protein the source literature calls "mutated"? Should the C-lobe iron-site loss and
protease Lys→Arg be captured as sequence-feature caveats on the MF terms?

Limitations

All non-anticoagulant conclusions rest on sequence/structure inference and
human-lactoferrin literature; no functional assay of draculin's iron binding,
protease, or antibacterial activity exists. Alignment-based ligand mapping (72%
identity to P02788) is robust for conserved columns but a single-substitution
functional prediction (e.g., Lys→Arg) is a hypothesis, not a measurement.

Artifacts

• OpenScientist final report
• OpenScientist final report