AIGR TreeGrafter Function-Inference Stress Test — Final Report

Target: TAS1R3 (UniProt A0A7L3N407), Oreotrochilus melanogaster (Andean hillstar, a hummingbird; NCBITaxon:689266, organism code 9AVES)
Seed hypothesis under test: TAS1R3 has sweet taste receptor activity (GO:0033041) — propagated by TreeGrafter/PANTHER, evidence type IEA, original reference GO_REF:0000118.

Summary

The TreeGrafter/PANTHER graft that placed A0A7L3N407 into subfamily PTHR24061:SF435 (Taste receptor type 1 member 3, TAS1R3) is correct at the subfamily level. The protein is a genuine, intact class C (family 3) G-protein-coupled receptor with the canonical taste-receptor architecture: an extracellular Venus flytrap domain, a 9-cysteine linker, and a seven-transmembrane bundle. It is 95.1% identical to the functionally characterized Calypte anna T1R3. Thus the sequence was not mis-placed within its superfamily (failure mode #3 excluded), and it is not a pseudo-enzyme — indeed it is not an enzyme at all, so there is no catalytic active site to degrade (failure mode #2 is categorically inapplicable).

The error the pipeline makes is one of granularity and biological context — failure mode #1. "Sweet taste receptor activity" is an emergent property of the T1R2+T1R3 heterodimer, not an intrinsic, autonomous molecular function of the TAS1R3 subunit. TAS1R3 is the obligate shared co-receptor of the T1R family: it partners with T1R2 to make a sweet receptor, or with T1R1 to make an umami/amino-acid receptor, and its ligand specificity is set by the partner subunit. The principal sweetener-binding pocket lies in T1R2, not T1R3. The pipeline's own output betrays this: the same run annotated this single protein with both GO:0033041 (sweet) and GO:0050917 (umami perception) — the classic signature of annotating a shared co-receptor with the function of every complex it joins.

In this organism the over-generalization is compounded into a near-impossibility. A UniProt proteome query shows the hummingbird possesses TAS1R1 and TAS1R3 but no TAS1R2 (zero TAS1R2 across all of Trochilidae), so the canonical sweet receptor cannot physically assemble. Hummingbirds nevertheless strongly prefer sugar because the ancestral T1R1–T1R3 umami receptor was evolutionarily repurposed into a carbohydrate sensor (Baldwin et al. 2014, the paper underlying the GO_REF context). The physiological percept ("sweet") is preserved, but the molecular machine responsible is different from the one GO:0033041 defines. The recommended curation lead is therefore generalize / retain-with-qualification — generalize to taste receptor activity (GO:0008527), keep the parallel umami term, and flag the mammalian→bird organism-specific carryover for curator verification.

Executive Judgment

Verdict: PARTIALLY SUPPORTED / TOO GENERAL (qualified).

Failure mode: #1 — Granularity / family-vs-subfamily, specifically a shared co-receptor subunit mislabeled with the emergent function of the complex it participates in — compounded by an organism-specific caveat.

• Not failure mode #2 (pseudo-enzyme): TAS1R3 is a GPCR, not an enzyme; there are no catalytic/metal-binding residues to lose. The fold is intact.
• Not failure mode #3 (within-superfamily mis-placement): the graft onto PANTHER SF435 (TAS1R3) is correct; 95% identity to the characterized Calypte anna T1R3 confirms it.

Single most decisive piece of evidence: a UniProt proteome query for Oreotrochilus melanogaster (taxon 689266) returns TAS1R1 (A0A7L3NBT4) and TAS1R3 (A0A7L3N407) but NO TAS1R2, and a family-wide query across Trochilidae (taxon 9242) returns zero TAS1R2 entries. The sweet-specific partner subunit required to build the canonical GO:0033041 sweet receptor (T1R2+T1R3) is physically absent from this animal, so the only assemblable T1R complex is T1R1+T1R3, the repurposed umami→carbohydrate receptor (PMID: 25146290). Reinforcing this, the same TreeGrafter run annotated the single TAS1R3 protein with both GO:0033041 (sweet) and GO:0050917 (umami) — direct internal evidence that TAS1R3 is a promiscuous co-receptor, not a sweet-specific receptor.

Independent Family / Function Assignment

From our own analysis (domain architecture + orthology + pairwise alignment), independent of the propagated term:

• Identity: Class C (family 3) GPCR, Taste receptor type 1 member 3 (TAS1R3 / T1R3) — the obligate shared co-receptor subunit of the T1R family. Architecture confirmed: extracellular Venus flytrap domain (ANF_receptor, PF01094, ~471 aa), 9-cysteine linker (NCD3G, PF07562; 10 Cys in the 481–560 linker), and 7-transmembrane bundle (7tm_3, PF00003).
• Nearest characterized homolog: Calypte anna (Anna's hummingbird) T1R3 (UniProt A0A088DCH0), 95.1% identical — the exact species in which Baldwin et al. experimentally showed T1R1/T1R3 functions as a sugar receptor. The specific function it carries: co-receptor subunit of the hummingbird T1R1–T1R3 carbohydrate (umami-derived) receptor.
• Function I would assign: TAS1R3 is a taste receptor co-receptor subunit that, within a heterodimer, participates in sweet and/or umami transduction. The most defensible intrinsic molecular-function term is GO:0008527 (taste receptor activity) (with GO:0004930 G-protein-coupled receptor activity already applicable). GO:0033041 is acceptable only as a complex-subunit annotation, and in this species is additionally unrealizable.

Granularity relative to seed: the seed is a sibling/partner-dependent complex term; the honest independent assignment is the more general taste receptor activity (GO:0008527), retaining the parallel umami term.

Key Findings

Finding 1 — TAS1R3 is a genuine class C GPCR co-receptor subunit, not a sweet-specific receptor

A0A7L3N407 is an 832-amino-acid protein correctly grafted to PANTHER subfamily PTHR24061:SF435 "Taste receptor type 1 member 3." Its domain architecture is the canonical T1R class C GPCR architecture: the extracellular ANF_receptor / Venus flytrap domain (PF01094) that binds ligand and mediates dimerization; the 9-cysteine NCD3G linker (PF07562), with 10 cysteines observed in the 481–560 region; and the 7tm_3 (PF00003) seven-transmembrane class C bundle. This is a textbook, intact T1R receptor — there is no enzymatic active site to degenerate, which categorically rules out the pseudo-enzyme failure mode.

Critically, the literature defines sweet taste receptor activity as a property of the T1R2+T1R3 heterodimer (PMID: 31553164: "Sweet taste receptor, a heterodimer belonging to the class C G-protein coupled receptor (GPCR) family and composed of the T1R2 and T1R3 subunits"), and molecular-dynamics work localizes the principal sweetener-binding Venus flytrap pocket to T1R2, "rather than T1R3 region" (PMID: 31308429). TAS1R3 therefore contributes dimerization and G-protein coupling, not sweet-ligand specificity — it is the shared obligate co-receptor whose downstream specificity is set by its partner. The fact that the same TreeGrafter run propagated both the sweet (GO:0033041) and umami (GO:0050917) terms to this one protein is itself diagnostic: a single molecule cannot be both "the sweet receptor" and "the umami receptor"; both are complex-level functions of two different heterodimers that share T1R3.

Finding 2 — In hummingbirds the canonical sweet receptor is absent; sugar is sensed by a repurposed T1R1–T1R3 receptor

Baldwin et al. 2014 (Science) — the very paper (PMID:25146290) cited in the GO_REF:0000118 decision context — report "the widespread absence in birds of an essential subunit (T1R2) of the only known vertebrate sweet receptor," and demonstrate by receptor-expression and behavioral assays that "the ancestral umami receptor (the T1R1-T1R3 heterodimer) was repurposed in hummingbirds to function as a carbohydrate receptor." Oreotrochilus melanogaster is a hummingbird (family Trochilidae). The mechanistic basis for sweet/sugar perception in this lineage is therefore structurally distinct from the mammalian T1R2+T1R3 sweet receptor around which GO:0033041 is defined. Annotating this bird's T1R3 with "sweet taste receptor activity" imports a mammalian complex that cannot assemble here.

Finding 3 — Proteome query empirically confirms TAS1R2 is absent while TAS1R1 + TAS1R3 are present

A UniProt proteome query for taxon 689266 (Oreotrochilus melanogaster) returns TAS1R3 (A0A7L3N407, 832 aa) and TAS1R1 (A0A7L3NBT4, 572 aa) — but no TAS1R2. Broadening to the whole hummingbird family (Trochilidae, taxon 9242) returns zero TAS1R2 entries, while TAS1R1 (e.g., Calypte anna A0A088DBT6) and TAS1R3 (Calypte anna A0A088DCH0) are present. This is a direct, organism-specific empirical reproduction of the bird-wide T1R2 loss reported by Baldwin et al. The only assemblable T1R heterodimer in this hummingbird is T1R1/T1R3 — the umami-derived carbohydrate receptor — so the canonical sweet receptor simply cannot form.

Finding 4 — The query T1R3 is 95% identical to the functionally characterized Calypte anna T1R3

A global Needleman–Wunsch alignment (executed in this work, numpy implementation) established orthology and correct placement:

The ~95% identity to Calypte anna T1R3 — the species where T1R1/T1R3 was shown to drive sugar-taste behavior "in captive and wild hummingbirds" (PMID: 25146290) — establishes that the query is the same functionally validated co-receptor, not a diverged paralog, and excludes a degenerate/pseudogene fold. The ~42% identity to human T1R3 is consistent with a genuine, well-conserved ortholog across the bird–mammal split, confirming the PANTHER graft onto SF435 is correct.

Mechanistic Model / Interpretation

The T1R taste-receptor system is a combinatorial heterodimer system built on one shared subunit:

  MAMMALS (e.g., human)                    HUMMINGBIRDS (Oreotrochilus)
  ---------------------                    ----------------------------

   T1R2 ──┐                                         T1R2  ✗ (GENE LOST across all birds)
  ├──► SWEET receptor  (GO:0033041)
   T1R3 ──┘  (sweetener pocket in T1R2 VFD)

   T1R1 ──┐                                         T1R1 ──┐
  ├──► UMAMI receptor  (GO:0050917)                ├──► repurposed CARBOHYDRATE /
   T1R3 ──┘                                         T1R3 ──┘    "sweet-sensing" receptor
                                                      (Baldwin et al. 2014)

   T1R3 = shared obligate CO-RECEPTOR — specificity is set by its PARTNER, not by T1R3 itself

The pipeline correctly identified the subunit (TAS1R3 → PANTHER SF435) but then propagated the complex-level function down to the subunit. Because T1R3 is shared between two complexes, the pipeline emitted both sweet and umami terms for the same protein — a logical signature of co-receptor annotation inflation.

Layered on top is an organism-specific problem that makes the sweet term not just over-general but physically unrealizable: this hummingbird (and all surveyed birds) lacks the T1R2 gene entirely. Without T1R2, no T1R2+T1R3 dimer can form, so there is no molecular machine in this animal with "sweet taste receptor activity" as GO defines it. Yet hummingbirds obviously detect and prefer sugar — because evolution re-tuned the ligand-binding pocket of the T1R1–T1R3 umami dimer to respond to carbohydrates. The physiological outcome ("sweet perception") is preserved; the molecular identity of the responsible receptor is different from the GO:0033041 definition.

The correct reading of A0A7L3N407's molecular function is therefore: (1) it is a bona fide T1R3 co-receptor subunit (graft correct → not mode #3); (2) it is an intact GPCR, not a degraded enzyme (→ not mode #2); (3) the propagated term is the emergent function of a heterodimer, too specific/mis-contextualized for the isolated subunit, and in this species additionally unassemblable (→ failure mode #1).

Active-Site / Placement Analysis

There is no catalytic active site to test — TAS1R3 is a class C GPCR, not an enzyme — so the catalytic-residue arm of the stress test is not applicable, and the pseudo-enzyme failure mode (#2) is categorically ruled out. The analytically meaningful equivalents are (a) structural integrity of the receptor machinery, (b) subfamily placement, and (c) functional-partner availability, all computed below.

(a) Receptor architecture integrity (UniProt/Pfam features):

(b) Subfamily / orthology placement (provenance: global Needleman–Wunsch alignment, numpy, executed this work):

(c) Functional-partner availability (provenance: UniProt proteome query, taxa 689266 and 9242):

Explicit statement: Catalytic residues — not applicable (no enzyme). Receptor fold and co-receptor subunit — intact and correctly placed. The decisive deficiency is absence of the T1R2 partner, which makes the GO:0033041 complex unassemblable in this organism.

Evidence Base (Evidence Matrix)

How the literature supports or challenges the findings: PMID: 31553164 and PMID: 31308429 jointly establish that sweet activity is a heterodimer property with the ligand pocket in T1R2 — challenging the notion that the isolated TAS1R3 subunit "has" sweet activity. PMID: 25146290 (the GO_REF-context paper) supplies the decisive organism-specific mechanism: T1R2 is lost in birds and sugar is sensed via a repurposed T1R1–T1R3 receptor. Our own computational work (proteome query + alignment) reproduces the T1R2 loss in this exact species and confirms the query is the same co-receptor those studies characterized.

Limitations and Knowledge Gaps

• Proteome completeness / absence-of-evidence. The TAS1R2-absence conclusion rests on current UniProt proteome content. It is strongly corroborated by Baldwin et al.'s bird-wide survey, but a database absence is formally weaker than a verified syntenic pseudogene at the TAS1R2 locus in the O. melanogaster assembly.
• Sequence quality. The UniProt entry shows fragment-like/ambiguous features (one ambiguous residue), so the predicted TM-helix count (5 of the expected 7) reflects incomplete prediction, not evidence of structural degeneration.
• No direct functional assay on this exact species. Functional inference for A0A7L3N407 is by 95% homology to Calypte anna, not by assay of the Oreotrochilus protein itself. Baldwin et al. characterized other hummingbird species.
• Tuning residues not individually mapped. We did not localize the specific VFD substitutions in the hummingbird T1R1–T1R3 pocket that confer carbohydrate responsiveness; doing so would upgrade the umami/carbohydrate assignment from homology to mechanism.
• GO modeling nuance. Whether curation policy permits "contributes_to"/complex-subunit framing for co-receptors affects whether GO:0033041 can be retained-with-qualification versus replaced.

Proposed Follow-up Experiments / Actions

1. Syntenic genome check for the TAS1R2 locus in the O. melanogaster assembly — confirm gene loss / pseudogenization directly. Decisive for the "canonical sweet receptor is impossible here" claim and upgrades the absence-in-DB caveat.
2. Heterologous co-expression functional assay. Express this TAS1R3 with conspecific T1R1 (and attempt T1R2+T1R3) in a calcium-reporter cell line; test sugars vs. amino acids to reproduce the Baldwin et al. assay in this species.
3. Map the known hummingbird sweet-tuning residues onto this T1R1/T1R3 sequence to confirm the carbohydrate-response machinery is present and identify the responsible substitutions.
4. Curation action (lead requiring verification): generalize GO:0033041 → GO:0008527 (taste receptor activity) as the intrinsic MF; retain the co-propagated GO:0050917 (umami) in parallel; if retaining GO:0033041 at all, mark it as a complex-subunit / "contributes_to" annotation and flag the mammalian→bird organism-specific carryover. Avoid "protein binding" as the fallback; the correct nuance is co-receptor subunit of a heterodimeric taste receptor.

GO Curation Implications

Lead action (requires curator verification): GENERALIZE / RETAIN-WITH-QUALIFICATION — do not treat as an autonomous sweet-receptor claim.

1. GO:0033041 (sweet taste receptor activity): do not keep as an unqualified intrinsic MF of the isolated subunit — it is a heterodimer-emergent function and, in this hummingbird, is unrealizable because TAS1R2 is absent. At most, keep as a complex-subunit annotation with explicit qualification.
2. Preferred robust annotation: generalize to GO:0008527 (taste receptor activity) (with GO:0004930 G-protein-coupled receptor activity as applicable) — the correct granularity for a shared co-receptor whose specificity is partner-conferred.
3. Companion term: retain the co-propagated GO:0050917 (umami perception) in parallel; its presence is the key signal that TAS1R3 is a shared co-receptor, and in hummingbirds this is the ancestral route now tuned to carbohydrates.
4. Avoid "protein binding" as the final recommendation.

Bottom line

The graft correctly identifies the protein as TAS1R3, but GO:0033041 overstates the case: TAS1R3 is a shared, partner-dependent co-receptor, and in this hummingbird the canonical sweet (T1R2/T1R3) receptor does not even exist — sweet/sugar is sensed through a repurposed T1R1/T1R3 umami receptor. Curate as generalize / retain-with-qualification, keeping the parallel umami annotation.

Artifacts

• OpenScientist final report
• OpenScientist final report