AIGR Gene Hypothesis Deep Research — Final Report

Gene: K9IJK6 (DSPA-gamma, vampire-bat salivary plasminogen activator)
Organism: Desmodus rotundus (common vampire bat; NCBITaxon:9430)
Focus type: function_assignment
Hypothesis slug: dspa-gamma-signaling-and-cell-response-capacities
Source: genes/DESRO/K9IJK6/K9IJK6-ai-review.yaml (free-text)

Summary

The seed hypothesis asserts that K9IJK6 "retains" four inherited process functions —
PDGF receptor signaling (GO:0048008), smooth muscle cell migration (GO:0014909),
cellular response to lipid (GO:0033993/GO:0071396), and cellular response to
oxygen-containing compound (GO:1901701) — in addition to its inferred plasminogen
activation. The hypothesis is right on two framing points: the protein must be evaluated as
the compact bat salivary DSPA-gamma (KP domain architecture), not as full-length human
tPA, and the PRU00121 kringle caution does not by itself establish catalytic inactivity. On
those points the seed reasoning is sound.

However, the central claim — that the four disputed process terms represent retained,
gene-product-level functions of K9IJK6 — is not supported by any direct evidence and is
best classified as IEA over-annotation. Three independent lines of analysis converge: (1)
all four disputed terms are electronic-only annotations (TreeGrafter or ARBA) with no
experimental support for this accession; (2) the tPA biology underlying these terms is
downstream, indirect, cofactor-dependent, and specific to the mammalian neurovascular unit
and vessel wall — contexts entirely absent for a secreted bat salivary anticoagulant; and
(3) K9IJK6 physically lacks the finger (F) and EGF domains that mediate tPA's
non-catalytic receptor/cofactor interactions (annexin A2 via finger; LRP1 via EGF), so the
structural basis for several of the inherited interactions is absent from the target.

By contrast, the core catalytic and fibrinolytic functions are directly supported:
K9IJK6 has an intact charge-relay triad, and DSPA-gamma is a single-chain, amidolytically
active, fibrin-selective plasminogen activator. Therefore serine-type endopeptidase activity
(GO:0004252), fibrin-dependent plasminogen activation (GO:0031639), and extracellular region
localization (GO:0005576) should be retained, while the four disputed process terms are
candidates for removal or demotion to non-core. Verdict: partially supported /
over-annotated. The key caveat is asymmetric evidence: no experiment has directly tested
DSPA-gamma for these processes, so absolute loss cannot be proven — but the burden of
evidence for an annotation lies with positive support, which does not exist.

Executive Judgment

Verdict: Partially supported / over-annotated.

• Supported (retain): serine-type endopeptidase activity, fibrin-dependent plasminogen
  activation, extracellular region — directly grounded in primary biochemistry and sequence.
• Over-annotated (remove or demote to non-core): PDGF receptor signaling, smooth muscle
  cell migration, cellular response to lipid, cellular response to oxygen-containing
  compound — IEA-only carry-overs of full-length mammalian tPA's downstream, tissue-specific
  roles, with the mediating structural exosites absent from the target.

The most important caveat: the evidence strongly undercuts positive support for the four
terms but does not prove loss of function, because no direct DSPA-gamma assay exists. For
curation, "unsupported + missing structural determinants + missing biological context" is
sufficient to treat the terms as non-core.

Key Findings

Finding 1 — K9IJK6 is DSPA-gamma with an intact catalytic triad and directly demonstrated fibrin-dependent plasminogen activation

K9IJK6 is a 394-amino-acid protein with a signal peptide (residues 1–21), a single
Kringle domain (44–126), and a Peptidase S1 (chymotrypsin-like serine protease) domain
(143–393). The catalytic charge-relay system is annotated intact — His189 / Asp238 /
Ser345 — with the canonical GDSGGP serine-protease motif at position 343. This KP
(Kringle–Protease) architecture is the defining signature of DSPA-gamma, the most compact
of the four Desmodus rotundus salivary plasminogen activators. PANTHER places the protein
in PTHR24264:SF42 (tissue-type plasminogen activator subfamily).

The catalytic competence and fibrinolytic function of this family are established by primary
biochemistry. PMID: 7592732 reports that "All
DSPAs are single-chain molecules, displaying substantial amidolytic activity," directly
confirming that DSPA-gamma is an active serine protease and refuting any inference of global
catalytic inactivity. The same study quantifies fibrin selectivity: "The ratio of the
bimolecular rate constants of plasminogen activation in the presence of fibrin versus
fibrinogen (fibrin selectivity) of DSPA alpha 1, alpha 2, beta, gamma, and t-PA was found to
be 13,000, 6500, 250, 90, and 72, respectively." DSPA-gamma (ratio 90) is thus a bona
fide fibrin-dependent plasminogen activator, comparable to tPA (72) on this metric. The
compact architecture is corroborated by
PMID: 1937019, which states that "DSPA beta and
-gamma lack the F and F-EGF domains, respectively," matching the KP domain content of
K9IJK6.

Curation consequence: GO:0004252 (serine-type endopeptidase activity), GO:0031639
(plasminogen activation), and GO:0005576 (extracellular region) are directly supported and
should be retained. The PRU00121 caution concerns propagation of kringle disulfide
features, not the protease triad, and must not be read as evidence of catalytic loss.

Finding 2 — The PDGFR-signaling and smooth-muscle-migration terms are IEA carry-over of full-length tPA's downstream/context roles (over-annotation)

On K9IJK6, all four disputed process terms are electronic-only: GO:0048008 (PDGFR
signaling) and GO:0014909 (smooth muscle cell migration) are IEA:TreeGrafter; GO:0033993
(response to lipid) and GO:1901701 (cellular response to oxygen-containing compound) are
IEA:UniProtKB-ARBA. No EXP/IDA annotation exists for this accession for any of the four.

Crucially, the underlying tPA biology that seeds these terms is downstream, indirect, and
tissue-context-specific — not a direct enzymatic property that transfers cleanly to a bat
salivary paralog:

• PDGFR signaling is indirect. tPA does not bind or activate PDGFRα directly; it
  proteolytically activates the latent growth factor PDGF-CC, which then engages PDGFRα.
  PMID: 18568034 shows that "co-injection of
  neutralizing antibodies to PDGF-CC with tPA blocks this increased permeability, indicating
  that PDGF-CC is a downstream substrate of tPA within the neurovascular unit."

• The tPA→PDGF-CC reaction is inefficient and cofactor-dependent.
  PMID: 28725968 states that "in vitro,
  activation of PDGF-CC by tPA is very inefficient and the mechanism of PDGF-CC activation in
  the NVU is not known," and that microglial Mac-1/LRP1 cofactors are required. This
  undermines blanket propagation of a "PDGFR signaling" process to a strictly fibrin-dependent
  secreted enzyme in saliva.

• Smooth muscle migration is an ECM-proteolysis consequence.
  PMID: 16363896 shows that "controlled
  perivascular release of tissue plasminogen activator (tPA) can generate cleaved
  extracellular matrix (ECM) chemotactic gradients to guide the migration of vascular smooth
  muscle cells (SMCs)." This is a downstream, vessel-wall, context-specific effect of
  proteolysis — not a direct molecular function of the enzyme.

DSPA-gamma is a bat salivary secreted anticoagulant with no mammalian vascular or CNS
context, is strictly fibrin-dependent, and lacks the finger/EGF domains that mediate several
of tPA's non-catalytic receptor interactions. No direct DSPA-gamma assay for PDGF-CC cleavage,
PDGFR signaling, SMC migration, or lipid/oxygen responses exists.

Finding 3 — K9IJK6 physically lacks the tPA finger/EGF exosites; all three TreeGrafter terms were grafted en bloc from one PANTHER node

A domain-coordinate comparison against human tPA/PLAT (P00750, 562 aa) shows that the exact
structural elements mediating tPA's non-catalytic receptor/cofactor interactions are
physically absent from K9IJK6:

Thus the annexin-A2-binding site (via finger) and the LRP1-binding site (via EGF) —
key exosites for tPA's receptor engagement and cellular signaling — have no structural
counterpart in the target sequence.

QuickGO provenance further shows the three TreeGrafter terms are not independently evidenced.
GO:0031639 (plasminogen activation), GO:0014909 (smooth muscle cell migration), and GO:0048008
(PDGFR signaling) are all IEA, ECO:0007826, GO_REF:0000118 (PANTHER TreeGrafter), each
carrying the identical with-reference PANTHER:PTN000667065 — i.e., grafted en bloc from
a single eutherian tPA node rather than from term-specific evidence. The two ARBA terms are
GO:0071396 (cellular response to lipid, ARBA00027803) and GO:1901701 (cellular response to
oxygen-containing compound, ARBA00027483), GO_REF:0000117, ECO:0000256 — a separate
propagation channel with its own rule context. This is the signature of pathway/paralog
carry-over, not of target-specific evidence.

Finding 4 — PMID 23411029 confirms salivary identity but provides no functional assay for the disputed terms

PMID: 23411029 is Francischetti et al. 2013,
"The 'Vampirome': Transcriptome and proteome analysis of the principal and accessory
submaxillary glands of the vampire bat Desmodus rotundus" (J Proteomics). This is the
sialotranscriptome/proteome survey from which the salivary DSPA-gamma transcript
(JAA47048.1 / K9IJK6) derives. It corroborates salivary-gland expression and secreted
localization (supporting GO:0005576) but is an omics catalogue — it contains no functional
assay for smooth-muscle migration, PDGFR signaling, or lipid/oxygen responses. Targeted
PubMed searches for DSPA-gamma non-fibrinolytic/signaling assays and for direct
DSPA-vs-alpha1/tPA comparative PDGF-CC/SMC/lipid assays returned no results, consistent with
the complete absence of any direct functional test of these processes for any DSPA isoform.

{{figure:plot_2.png|caption=GO decision table with tPA/DSPA-gamma exosite (finger/EGF) domain comparison and QuickGO-verified annotation provenance. The three disputed TreeGrafter terms share a single with-reference (PANTHER:PTN000667065), and the finger/EGF exosites carrying annexin-A2 and LRP1 binding are absent from K9IJK6.}}

Mechanistic Model / Interpretation

The disagreement between the seed hypothesis and the evidence resolves cleanly once direct
molecular function is separated from downstream, context-dependent consequences of
proteolysis.

  DIRECT (transfers to DSPA-gamma)          INDIRECT / CONTEXT-SPECIFIC (does NOT transfer)
  ───────────────────────────────          ────────────────────────────────────────────────
  Catalytic triad His/Asp/Ser  ─┐
  Kringle + protease (KP)        │          tPA ─(inefficient, Mac-1/LRP1-dependent)─▶ PDGF-CC*
  Single-chain, amidolytic       ├─▶ cleaves          │
  Fibrin-selective (ratio 90)    │   plasminogen      ▼
                 │                 PDGFRα signaling → BBB permeability (brain NVU)
  ┌───────────────────────────┐  │
  │ GO:0004252 endopeptidase  │◀─┘          tPA ─(perivascular ECM proteolysis)─▶ chemotactic
  │ GO:0031639 plasminogen    │                       gradient → smooth muscle cell migration
  │            activation     │                       (vessel-wall remodeling)
  │ GO:0005576 extracellular  │
  └───────────────────────────┘          Requires: finger (annexin A2) + EGF (LRP1) exosites
RETAIN                                       → ABSENT in DSPA-gamma (KP only)
                                     → mammalian vascular/CNS context ABSENT
                                     → REMOVE / DEMOTE the four disputed terms
  * PDGF-CC activation by tPA is "very inefficient" in vitro (PMID 28725968)

DSPA-gamma is a feeding adaptation: a secreted salivary anticoagulant whose job is to
keep the host's blood meal flowing by driving fibrin-localized plasminogen activation. Its
extreme fibrin dependence and compact KP architecture are specializations for that role.
The processes the seed hypothesis wants to retain — PDGFR signaling, SMC migration, lipid and
oxygen responses — are properties inferred from mammalian tPA operating inside a living
vessel wall or brain, where tPA's finger/EGF exosites recruit cofactors (annexin A2, LRP1,
Mac-1) and where its proteolysis generates ECM chemotactic gradients and activates latent
growth factors. None of that machinery or context is present for a bat salivary enzyme lacking
those exosites. The annotations are therefore best understood as automated ortholog/paralog
carry-over (a single PANTHER node grafted en bloc, plus separate ARBA sequence rules), not
target-specific biology.

The evidence is asymmetric: it strongly undercuts positive support for the four terms but
does not prove absolute loss, because no one has directly assayed DSPA-gamma for these
activities. For curation, the absence of any direct evidence combined with the missing
structural determinants and missing biological context is sufficient to treat the terms as
non-core / over-annotated.

Evidence Base / Evidence Matrix

GO Curation Implications

Lead requiring curator verification. The evidence separates cleanly into a supported core
and an over-annotated periphery.

The four disputed terms are BP annotations propagated electronically. They are neither
anchored in a direct molecular activity of DSPA-gamma nor supported by any experiment on this
accession or any DSPA isoform. Because they derive from a single grafted PANTHER node (three
terms) plus generic ARBA rules (two terms), and because the structural exosites and biological
context they depend on are absent, they fit the classic profile of paralog/pathway
over-annotation and should not be treated as core functions. No "protein binding" fallback
is recommended; the informative, supported terms are the protease and plasminogen-activation
terms above.

Mechanistic Scope

The immediate molecular function under test is enzymatic: a chymotrypsin-like serine
protease (KP architecture) that, in a fibrin-dependent manner, cleaves plasminogen to plasmin.
That direct activity is supported.

The four disputed terms describe downstream phenotypes and pathway consequences, not direct
gene-product activities:

• PDGFR signaling is two steps removed — tPA cleaves latent PDGF-CC, and PDGF-CC (not tPA)
  engages the receptor; the cleavage is inefficient and cofactor-dependent.
• Smooth muscle cell migration is a tissue-level outcome of ECM proteolysis generating
  chemotactic gradients in the vessel wall.
• Response to lipid and cellular response to oxygen-containing compound are broad
  regulatory/environmental-response categories propagated by sequence rules, not molecular
  activities of the protein.

For a secreted bat salivary anticoagulant, none of these downstream/context-dependent processes
is a plausible core function, and none has been demonstrated.

Conflicts and Alternatives

1. Paralog/ortholog carry-over (primary alternative explanation). The three TreeGrafter
   terms share the identical with-reference PANTHER:PTN000667065 (a eutherian tPA subtree),
   indicating en-bloc grafting rather than term-specific evidence. This is the most parsimonious
   explanation for their presence and directly competes with the "retention" framing.
2. Full-length tPA vs. DSPA-gamma conflation. The disputed functions are documented for
   full-length human/mouse tPA operating in the vascular wall and brain. DSPA-gamma lacks the
   finger and EGF domains and the mammalian tissue context. The seed hypothesis itself warns
   against this conflation, and the evidence confirms the warning cuts against retention of
   the four terms.
3. Organism/context mismatch. tPA's PDGF-CC/BBB and SMC-migration roles are
   intravascular/CNS phenomena (PMID 18568034, 16363896, 23218119). A salivary secretion
   delivered into a host bite wound has no access to that machinery.
4. In-vitro-only / inefficient reaction. Even for tPA, PDGF-CC activation is "very
   inefficient" in vitro and cofactor-dependent (PMID 28725968), weakening any argument that
   the activity is a robust, transferable core function.
5. Separate ARBA propagation. The lipid/oxygen terms come from a different automated channel
   (UniProtKB-ARBA, ECO:0000256) with its own rule scope; they should be evaluated independently
   and are even more generic/weakly anchored.

No evidence was found that supports retention of the four terms as direct functions.

Limitations and Knowledge Gaps

• No direct DSPA-gamma functional assays exist for PDGF-CC cleavage, PDGFR signaling, SMC
  migration, or lipid/oxygen responses. We can demonstrate absence of positive evidence and
  absence of key structural determinants, but not experimentally proven loss of function. Why
  it matters: curators distinguish "unsupported" from "refuted"; this is the former,
  strengthened by structural and contextual arguments. Resolution: a direct comparative assay
  (below).
• PDGF-CC cleavage by DSPA-gamma has never been tested. tPA's protease domain performs the
  cleavage; DSPA-gamma has a homologous, active protease domain but different exosites and strict
  fibrin dependence. Whether the isolated catalytic domain could cleave PDGF-CC in vitro is
  unknown. Resolution: recombinant DSPA-gamma + latent PDGF-CC cleavage assay.
• Sequence-level exosite mapping was based on UniProt/domain annotations, not a solved
  DSPA-gamma structure. The finger/EGF absence is unambiguous (domains are simply not present),
  but fine-grained protease-domain surface-loop differences were not modeled.
• ARBA rule identity confirmed but rule logic not fully audited. The specific motifs
  triggering GO:0071396 and GO:1901701 were not dissected; these terms are generic and
  low-information regardless.

Discriminating Tests

1. Direct comparative plasminogen-activation panel (DSPA-gamma vs DSPA-alpha1 vs tPA).
   Recombinant expression (BHK secretion is established, PMID 1634121) followed by kinetic assays
   ± fibrin to confirm the core function and benchmark fibrin dependence — anchors the retained
   terms.
2. PDGF-CC cleavage assay. Incubate purified recombinant DSPA-gamma with latent PDGF-CC ±
   Mac-1/LRP1 cofactors; SDS-PAGE/Western for the activated fragment. A negative result would
   strongly refute GO:0048008 propagation; a positive (even if inefficient) would partially
   rehabilitate an indirect link.
3. SMC migration / ECM-degradation assay. Boyden chamber or ECM-gradient assay with
   DSPA-gamma vs tPA to test whether the bat enzyme can generate chemotactic ECM cleavage in the
   absence of finger/EGF-mediated cell-surface localization.
4. Annotation-provenance audit at scale. Confirm across the DESRO gene set how many process
   terms trace to PANTHER:PTN000667065 grafts, to identify systematic tPA-paralog
   over-annotation.
5. Structural modeling / superposition of a DSPA-gamma model against tPA (P00750) to quantify
   exosite/loop divergence around the protease active site.

Curation Leads (require curator verification)

Action changes (leads):
- Retain GO:0004252 (serine-type endopeptidase activity), GO:0031639 (plasminogen
activation), GO:0005576 (extracellular region) as core, literature-consistent annotations.
- Remove or mark as non-core / over-annotated GO:0048008 (PDGF receptor signaling), GO:0014909
(smooth muscle cell migration), GO:0033993/GO:0071396 (response to lipid), GO:1901701 (cellular
response to oxygen-containing compound). Rationale to record: IEA carry-over from
PANTHER:PTN000667065 (three terms) and ARBA rules (two terms); downstream, cofactor-dependent,
tissue-context-specific tPA biology; finger/EGF exosites absent from the target.

Candidate references with exact snippets to verify:
- PMID 7592732 — "All DSPAs are single-chain molecules, displaying substantial amidolytic
activity." and "…fibrin selectivity… of DSPA alpha 1, alpha 2, beta, gamma, and t-PA was found
to be 13,000, 6500, 250, 90, and 72, respectively." → supports core MF/BP.
- PMID 1937019 — "DSPA beta and -gamma lack the F and F-EGF domains, respectively." → supports KP
architecture / exosite loss.
- PMID 18568034 — "PDGF-CC is a downstream substrate of tPA within the neurovascular unit." →
supports demotion of GO:0048008 (indirect).
- PMID 28725968 — "in vitro, activation of PDGF-CC by tPA is very inefficient…" → supports demotion
of GO:0048008.
- PMID 16363896 — "…tPA can generate cleaved extracellular matrix (ECM) chemotactic gradients to
guide the migration of vascular smooth muscle cells (SMCs)." → supports demotion of GO:0014909
(downstream).
- PMID 23411029 — Vampirome salivary transcriptome/proteome → supports GO:0005576, salivary
identity of JAA47048.1/K9IJK6.

Suggested curator questions:
- Should IEA process terms grafted en bloc from a single tPA PANTHER node be retained on a
domain-reduced salivary paralog lacking the mediating exosites?
- Do we distinguish "unsupported/over-annotated" (this case) from "refuted" in the review action,
given no direct negative assay exists?

Suggested experiments: the DSPA-gamma vs alpha1/tPA comparative panel, PDGF-CC cleavage assay,
and SMC-migration assay described under Discriminating Tests.

{{figure:plot_1.png|caption=GO decision table, tPA-vs-DSPA-gamma domain comparison, catalytic-triad confirmation (His189/Asp238/Ser345), and annotation provenance for K9IJK6.}}

Proposed Follow-up Actions

1. Update the review to retain the protease/plasminogen-activation/extracellular core and
   flag the four disputed process terms as IEA over-annotation candidates for removal or non-core
   demotion, citing the provenance (single PANTHER:PTN000667065 graft + ARBA rules) and the
   finger/EGF exosite absence.
2. Record the exosite comparison (finger 39–81 / EGF 82–120 present in tPA, absent in
   K9IJK6) as structural justification in the review notes.
3. Escalate the systematic issue: audit other DESRO/bat salivary PA entries for the same
   PANTHER:PTN000667065 en-bloc graft to catch analogous over-annotations.
4. If experimental resolution is desired, prioritize the PDGF-CC cleavage assay with
   recombinant DSPA-gamma as the single most discriminating test.

Bottom Line

The seed hypothesis is partially supported: its insistence on treating K9IJK6 as the compact
DSPA-gamma (not tPA/desmoteplase) and on not misreading the PRU00121 kringle caution as catalytic
death is correct, and the protease/plasminogen-activation core is directly supported and should
be retained. But the specific claim that K9IJK6 retains PDGF receptor signaling, smooth muscle
cell migration, response to lipid, and cellular response to oxygen-containing compound as
gene-product functions is not supported — these are IEA-only, en-bloc-grafted
(PANTHER:PTN000667065) or generic-ARBA carry-overs of full-length mammalian tPA's downstream,
cofactor-dependent, tissue-specific roles, and the mediating finger/EGF exosites are physically
absent from the target. Verdict: partially supported / over-annotated. Recommended curation
lead: retain the protease, plasminogen-activation, and extracellular-region terms; remove or
demote the four disputed process terms to non-core pending direct DSPA-gamma assays.

Artifacts

• OpenScientist final report
• OpenScientist final report
• OpenScientist plot 1
  
• OpenScientist plot 2