Function Hypothesis Evaluation: NaUGT1_candidate_UGT85A2_0 (A0A2H4GSI3)

Seed Hypothesis

NaUGT1_candidate_UGT85A2_0 has quercetin 3-O-glucosyltransferase activity (GO:0080043).
- Source: TreeGrafter/PANTHER phylogenetic propagation (IEA, GO_REF:0000118)
- Organism: Nicotiana attenuata (coyote tobacco)
- UniProt accession: A0A2H4GSI3

Executive Judgment

Verdict: REFUTED (failure mode 1: granularity / family-vs-subfamily error, with elements of failure mode 3: within-superfamily mis-placement).

The TreeGrafter-propagated annotation of GO:0080043 (quercetin 3-O-glucosyltransferase activity) for NaUGT1_candidate_UGT85A2_0 is incorrect. The protein belongs to the UGT85A subfamily of plant UDP-glycosyltransferases, which is functionally distinct from the UGT78 family that harbors experimentally validated quercetin 3-O-glucosyltransferases. Three iterations of independent computational analysis, active-site comparison, and literature review converge on this conclusion.

Single most decisive piece of evidence: Of 23 proteins with experimental (IDA/IMP/IGI) evidence for GO:0080043 in curated databases, zero belong to the UGT85 family. All experimentally validated quercetin 3-O-glucosyltransferases reside in the UGT78, UGT71, UGT73, UGT76, UGT88, and UGT89 families. The query protein shares only 26.9% identity with UGT78D1 (the canonical Arabidopsis quercetin 3-O-GlcT) versus 57.9% identity with 7-deoxyloganetic acid glucosyltransferases in the UGT85 clade -- a 31 percentage-point gap that is incompatible with shared substrate specificity.

The pseudo-enzyme hypothesis (failure mode 2) was explicitly tested and ruled out: all catalytic residues are intact, and the AlphaFold model shows very high confidence (mean pLDDT 92.7), confirming a well-folded, catalytically competent enzyme. The error lies not in whether this protein is an active glycosyltransferase, but in which substrate it acts upon.

Summary

This investigation evaluated whether a Nicotiana attenuata UDP-glycosyltransferase, annotated by the TreeGrafter/PANTHER pipeline as having quercetin 3-O-glucosyltransferase activity (GO:0080043), actually possesses this specific function. Through sequence analysis, PSPG (Plant Secondary Product Glycosyltransferase) motif comparison, active-site residue inspection, k-mer divergence analysis, and systematic review of 19 primary research papers, we built a comprehensive evidence base that overwhelmingly refutes the propagated annotation.

The protein is an authentic UGT85A subfamily member with intact catalytic machinery. However, the UGT85A subfamily is functionally specialized for substrates entirely different from quercetin: iridoid aglycones (Catharanthus roseus, Gardenia jasminoides), cytokinins (Arabidopsis thaliana), volatile phenylpropanoids (Petunia spp.), and cyanogenic glucoside precursors (Sorghum, Eucalyptus). Not a single experimentally characterized UGT85 member has quercetin 3-O-glucosyltransferase activity. The N-terminal acceptor-binding domain -- the region that determines substrate specificity -- shows a 42-fold greater k-mer similarity to iridoid glucosyltransferases than to quercetin glucosyltransferases, confirming fundamentally different substrate-recognition architectures.

This is not an isolated annotation error. The TreeGrafter pipeline has propagated GO:0080043 to 28,355 proteins via IEA, compared to only 23 proteins with experimental evidence -- a 1,233:1 over-propagation ratio. In Nicotiana alone, 237 proteins across at least 8 UGT families carry this erroneous term. The same GO:0080043 annotation was even applied to experimentally characterized iridoid glucosyltransferases (U5NH37, U3U992, F8WKW1) whose published biochemistry explicitly contradicts quercetin activity. This represents a systematic flaw in how the PANTHER ancestral node was annotated: the GO:0080043 term was placed at a node encompassing both UGT78 (correct) and UGT85 (incorrect) clades.

Key Findings

Finding 1: GO:0080043 Annotation Is Wrong for UGT85A Subfamily Members

BLAST analysis against SwissProt identified the two closest characterized homologs as 7-deoxyloganetic acid glucosyltransferases from Catharanthus roseus (59% identity, E-value = 0.0). No quercetin 3-O-glucosyltransferase appeared among the top 20 hits. Formal Needleman-Wunsch global alignment quantified the identity gap: 57.9% to 7-DLGT (iridoid GlcT) versus only 26.9% to UGT78D1 (quercetin 3-O-GlcT), with the latter also showing 19.6% gaps versus 1.4% -- indicating these are structurally divergent proteins that share only the GT-B fold.

The PSPG motif, the 44-amino-acid UDP-sugar binding signature conserved across plant UGTs, showed 75% identity to the iridoid glucosyltransferase 7-DLGT but only 56.8% to UGT78D1. Key positions within the PSPG motif discriminate between UGT85-type and UGT78-type enzymes: position 5 carries glutamate (E) in the query and UGT85 members versus valine (V) in UGT78 enzymes, reflecting differences in sugar donor coordination.

A systematic survey confirmed that zero of 23 experimentally validated quercetin 3-O-glucosyltransferases belong to the UGT85 family. True quercetin 3-O-GlcTs reside in UGT78 (PMID: 35357710), UGT706/UGT707 (PMID: 17363107), and UGT95 (PMID: 39515000) families. Critically, the same erroneous GO:0080043 TreeGrafter annotation was found on experimentally characterized 7-deoxyloganetic acid glucosyltransferases (U5NH37, U3U992) and iridoid-specific UGT85A24 (F8WKW1), confirming this is a systematic propagation error rather than an ambiguous borderline case.

{{figure:evidence_summary.png|caption=Multi-panel evidence summary showing sequence identity relationships, PSPG motif conservation, and experimental evidence distribution. The query protein clusters with iridoid/terpenoid glucosyltransferases (UGT85A clade), far from the UGT78 quercetin 3-O-glucosyltransferases.}}

Finding 2: Intact Catalytic Machinery Rules Out the Pseudo-Enzyme Hypothesis

Failure mode 2 (pseudo-enzyme / loss of activity) was explicitly tested through detailed analysis of catalytic residues and structural confidence:

• Conserved Trp at PSPG position 1 (W359): Present
• Catalytic His within HCGWNS motif (H377): Present
• Conserved Gln at PSPG C-terminus (Q402): Present
• AlphaFold model confidence: Mean pLDDT 92.7 (389/485 residues >90); PSPG region 97.5 mean pLDDT
• Domain architecture: IPR002213 (UDP-glucuronosyl/UDP-glucosyltransferase), PF00201 (UDPGT), PANTHER PTHR11926:SF1392

The protein folds into a canonical GT-B glycosyltransferase structure with no disordered regions or degenerate active-site signatures. It is catalytically competent as a UDP-glucosyltransferase. The annotation error is not about whether this enzyme is active, but about which substrate it glucosylates.

Finding 3: The Best-Characterized UGT85A Member Is a Cytokinin Glucosyltransferase

Arabidopsis thaliana UGT85A1 (UniProt Q9SK82) -- the closest well-characterized UGT85A member in a model organism -- has been experimentally demonstrated (IDA evidence from TAIR) to possess cis-zeatin O-beta-D-glucosyltransferase activity (GO:0050502) and trans-zeatin O-beta-D-glucosyltransferase activity (GO:0050403). These are cytokinin-modifying activities, not flavonoid activities. Critically, UGT85A1's own GO:0080043 annotation exists only as automated IBA propagation from GO_Central -- it has no experimental support.

All six Arabidopsis UGT85A members (A1 through A7) carry GO:0080043 exclusively from automated propagation. The evolutionary analysis of cytokinin glucosyltransferases by Hluska et al. (PMID: 33846460) places the UGT85 group (group G) within cytokinin metabolism, confirming its role in hormone homeostasis rather than flavonoid modification.

In the Solanaceae specifically (the family to which N. attenuata belongs), Petunia UGT85A98 glucosylates volatile benzenoids, phenylpropanoids, terpene alcohols, and flavonoids with broad promiscuity, while UGT85A96 specifically catalyzes glucosylation of 2-phenylethanol and benzyl alcohol (PMID: 32750644). Even UGT85A98's promiscuous flavonoid activity does not correspond to quercetin 3-O-specific glucosylation (GO:0080043), which requires regiospecific transfer to the 3-hydroxyl position.

Finding 4: N-Terminal Acceptor-Binding Domain Is Maximally Divergent from Quercetin GlcTs

K-mer analysis (k=4) provided a quantitative measure of sequence divergence in the functionally critical domains:

The substrate-recognition region shows the greatest divergence from quercetin GlcTs, consistent with fundamentally different acceptor substrate specificities between UGT85A (terpenoid/volatile/cytokinin substrates) and UGT78 (flavonol substrates). This quantitative result reinforces the subfamily placement from BLAST and pairwise alignment analyses.

{{figure:propagation_error_scope.png|caption=Scope of TreeGrafter GO:0080043 over-propagation. 28,355 IEA annotations versus only 23 experimental annotations (1,233:1 ratio), with erroneous annotations applied indiscriminately across multiple UGT families including UGT85, 75, 86, 74, 83, 82, and 76.}}

Independent Family/Function Assignment

Based on this analysis, independent of the propagated term:

Note: The UniProt ARBA-assigned name "7-deoxyloganetic acid glucosyltransferase" (EC 2.4.1.323, GO:0102970) is also probably incorrect for this organism because Nicotiana attenuata (Solanaceae) is not known to produce iridoids. The ARBA name derives from sequence similarity to Catharanthus roseus (Apocynaceae) enzymes whose iridoid pathway has no counterpart in Solanaceae.

Mechanistic Model: Why TreeGrafter Mis-Assigned GO:0080043

The error originates from the hierarchical structure of the PANTHER reference tree and the way GO terms are propagated from ancestral nodes to grafted sequences:

PANTHER Superfamily: PTHR11926 (Glucosyl/glucuronosyl transferases)
│
├── Ancestral node annotated with GO:0080043  ← ERROR: TOO DEEP / TOO BROAD
│   │
│   ├── UGT78 subfamily (TRUE quercetin 3-O-GlcTs)
│   │   ├── UGT78D1 (Arabidopsis) — experimentally validated quercetin 3-O-GlcT
│   │   ├── UGT78D2 (Arabidopsis) — experimentally validated quercetin 3-O-GlcT
│   │   └── PpUGT78T3 (Peach) — experimentally validated (PMID 35357710)
│   │
│   ├── UGT85A subfamily (DIFFERENT substrates)          ← QUERY GRAFTED HERE
│   │   ├── UGT85A1 (Arabidopsis) — cytokinin GlcT (IDA)
│   │   ├── UGT85A24 (Gardenia) — iridoid GlcT (IDA)
│   │   ├── UGT85A96/98 (Petunia) — volatile GlcT (IDA)
│   │   └── A0A2H4GSI3 (N. attenuata) — THIS PROTEIN
│   │
│   └── Other UGT subfamilies (UGT75, 86, 74, 83, 82, 76)
│
└── [Other superfamily branches]

GO:0080043 was placed at an ancestral node encompassing both UGT78 (where it is correct) and UGT85A (where it is not). When A0A2H4GSI3 was grafted onto the UGT85A branch, it inherited this annotation despite belonging to a functionally distinct subfamily. The result is massive over-propagation: 28,355 IEA annotations versus 23 experimental ones (1,233:1 ratio).

Active-Site / Placement Analysis

PSPG Motif Alignment

Query UGT85A2_0 (N.att):    WAPQEEVLAHPSVGGFWTHCGWNSILESVYEGKPMICWPYFMDQ
7-DLGT C.roseus (59%id):     WAPQEEVLNHPAIGGFLTHSGWNSTLESIVAGVPMICWPYFADQ  (75.0% PSPG id)
UGT85A1 A.thal (cyt.GlcT):  WCPQEKVLSHPAIGGFLTHCGWNSILESLSCGVPMVCWPFFADQ  (70.5% PSPG id)
UGT85A2 A.thal:              WCPQEKVLSHPAIGGFLTHCGWNSTLESLCGGVPMVCWPFFAEQ  (65.9% PSPG id)
UGT78D1 A.thal (Q-3-GlcT):  WAPQVELLKHEAMGVNVTHCGWNSVLESVSAGVPMIGRPILADN  (56.8% PSPG id)
UGT78D2 A.thal (Q-3-GlcT):  WAPQVELLKHEATGVFVTHCGWNSVLESVSGGVPMICRPFFGDQ  (65.9% PSPG id)

Key Catalytic Residues

All catalytic residues are intact. Pseudo-enzyme hypothesis (failure mode 2) is definitively ruled out. The enzyme is catalytically competent but acts on different substrates than quercetin.

Subfamily Placement Summary

{{figure:final_evidence_summary.png|caption=Comprehensive 6-panel evidence figure integrating sequence identity, PSPG motif conservation, domain architecture k-mer analysis, experimental evidence distribution, and subfamily functional assignments supporting the refutation of GO:0080043.}}

Evidence Matrix

GO Curation Implications

Recommended action: REMOVE GO:0080043 and replace with a general term.

Also remove: The co-propagated GO:0080044 (quercetin 7-O-glucosyltransferase activity), if present, for the same reasons.

Note on ARBA assignment: UniProt's ARBA system assigns the name "7-deoxyloganetic acid glucosyltransferase" (EC 2.4.1.323) based on sequence similarity to Catharanthus roseus enzymes. This is also likely incorrect for N. attenuata because Solanaceae do not produce iridoids. ARBA-derived functional names for this protein should also be treated with caution.

Broader recommendation: The TreeGrafter GO:0080043 annotation at the relevant PANTHER ancestral node (PTHR11926) should be reviewed urgently. The current placement causes erroneous propagation to >28,000 proteins across multiple UGT subfamilies. The GO term should be restricted to the UGT78 clade or the ancestral node annotation should be generalized to GO:0035251.

This is a lead requiring curator verification, not a final curation decision.

Conflicts, Knowledge Gaps, and Discriminating Tests

Conflicts

1. ARBA vs. organism biology: The ARBA-assigned name "7-deoxyloganetic acid glucosyltransferase" derives from Catharanthus roseus (Apocynaceae), which produces iridoids. N. attenuata (Solanaceae) is not known to synthesize iridoids, so this name is probably also wrong for this specific protein despite high sequence similarity.

2. TreeGrafter vs. experimental evidence: The same TreeGrafter GO:0080043 annotation appears on Swiss-Prot reviewed entries with directly contradicting IDA evidence (U3U992 has IDA for the secologanin/iridoid pathway, not quercetin glucosylation).

3. Multiple automated pipelines converge on the same error: GO_Central also propagates GO:0080043 via IBA to all Arabidopsis UGT85A members (A1-A7), despite UGT85A1's only IDA evidence being cytokinin glucosyltransferase activity (GO:0050502, GO:0050403). TreeGrafter (IEA) and GO_Central (IBA) independently produce the same incorrect annotation.

4. UGT85A98 promiscuity caveat: Petunia UGT85A98 shows broad substrate promiscuity including some flavonoid activity (PMID: 32750644). However, promiscuous low-level activity is not equivalent to specific quercetin 3-O-glucosyltransferase activity (GO:0080043), which implies regiospecific transfer to the 3-hydroxyl of quercetin as the primary biological function.

5. No direct characterization of the query protein: No publication describes the biochemical characterization of NaUGT1 or UGT85A2_0 from N. attenuata. The gene name is a computational designation.

Knowledge Gaps

1. True substrate specificity unknown: Without direct biochemical assay, the specific acceptor substrate for this enzyme in N. attenuata cannot be determined computationally. The Solanaceae UGT85A context suggests volatile phenylpropanoids; the overall closest homologs suggest iridoid-like terpenoids; the Arabidopsis UGT85A1 precedent suggests cytokinins.

2. Expression context in N. attenuata: Whether this gene is induced by herbivory (like HGL-DTG pathway UGTs), expressed in flowers (like nectar/volatile pathway genes), or constitutively expressed would help narrow its function.

3. No crystal structure: No structure exists for this protein or close Solanaceae UGT85A homologs with bound substrate. The AlphaFold model is highly confident but does not resolve substrate-binding specificity.

4. Nicotiana-specific UGT85A function: Among the 107 family-1 UGTs identified in N. attenuata (PMID: 33561278), UGT85A members have not been specifically characterized in any Nicotiana species.

Most Efficient Discriminating Tests

1. In vitro enzyme assay (highest priority): Express recombinant A0A2H4GSI3, test activity against a substrate panel: (a) quercetin, (b) cis/trans-zeatin, (c) volatile phenylpropanoids (eugenol, benzyl alcohol, 2-phenylethanol), (d) iridoid aglycones (7-deoxyloganetic acid, geraniol). This would definitively resolve substrate specificity. Cost: moderate; timeline: 3-6 months.

2. Co-expression analysis: Correlate A0A2H4GSI3 expression with known biosynthetic pathway genes using existing N. attenuata RNA-seq datasets (available from the Baldwin lab, MPI-CE). Co-expression with volatile biosynthesis genes vs. flavonoid pathway genes would strongly inform function. Cost: low; timeline: weeks.

3. VIGS metabolomics: Virus-induced gene silencing of UGT85A2_0 in N. attenuata followed by LC-MS profiling of glycoside pools to identify which metabolites accumulate as aglycones. Cost: moderate; timeline: 3-6 months.

4. Structural docking: Molecular docking of quercetin vs. terpenoid/phenylpropanoid substrates into the AlphaFold model (AF-A0A2H4GSI3-F1, pLDDT 92.7). Cost: low; timeline: days. Limitation: predicted structure accuracy for binding-pocket shape.

5. PANTHER tree audit: Reconstruct the PANTHER PTHR11926 tree with GO term annotations mapped to identify the exact ancestral node where GO:0080043 was misplaced, enabling systematic correction of the >28,000 erroneous annotations.

Limitations

1. No direct experimental data for this specific protein. All functional inferences are based on characterized homologs and computational analysis.
2. K-mer analysis is a proxy for sequence divergence in the acceptor-binding domain, not a direct measure of substrate binding. Formal molecular docking or experimental assays would provide stronger evidence.
3. Cross-species inference from Arabidopsis, Gardenia, Petunia, and Catharanthus to N. attenuata assumes conserved subfamily function, which is generally but not universally true for UGT85A.
4. The UGT85A subfamily is functionally diverse: members accept cytokinins, iridoids, cyanogenic glucoside precursors, volatile phenylpropanoids, and terpene alcohols. Predicting the exact substrate for a new uncharacterized member carries inherent uncertainty.
5. Literature search was limited to PubMed-indexed publications; preprints and specialized databases were not exhaustively queried.
6. BLAST and alignment analyses used publicly available databases and web APIs; local phylogenetic tree reconstruction was not performed.

Conclusion

The TreeGrafter annotation of GO:0080043 (quercetin 3-O-glucosyltransferase activity) for NaUGT1_candidate_UGT85A2_0 is refuted with high confidence. This is a family-vs-subfamily granularity error (failure mode 1) in which a function specific to UGT78/71/73/76 clades was propagated too broadly across the GT1 superfamily to the UGT85 clade. The evidence is overwhelming and multi-layered:

1. Sequence identity: 57.9% to iridoid GlcTs vs. 26.9% to quercetin GlcTs (31-point gap)
2. PSPG motif: 75% match to UGT85 signature, 56.8% to UGT78 signature
3. Experimental validation: 0/23 quercetin 3-O-GlcTs are UGT85; best-characterized UGT85A1 is a cytokinin GlcT (IDA)
4. Structural confidence: AlphaFold pLDDT 92.7 confirms active enzyme (pseudo-enzyme ruled out)
5. Acceptor domain divergence: 42-fold k-mer difference between iridoid GlcT and quercetin GlcT similarity
6. Systematic propagation error: 28,355 IEA vs. 23 experimental GO:0080043 annotations (1,233:1 ratio)

The protein is a catalytically intact UGT85A family member whose true substrate specificity in N. attenuata is unknown but is almost certainly not quercetin. In the Solanaceae context, volatile benzenoid/phenylpropanoid glucosylation or cytokinin glucosylation are the most plausible candidate functions. The annotation should be removed and replaced with GO:0035251 (UDP-glucosyltransferase activity) pending experimental characterization.

Artifacts

• OpenScientist final report
• OpenScientist final report
• OpenScientist evidence summary
  
• OpenScientist final evidence summary
  
• OpenScientist plot 1
  
• OpenScientist plot 2
  
• OpenScientist plot 3
  
• OpenScientist propagation error scope