AIGR TreeGrafter Function-Inference Stress Test — Final Report

Target: Pseudomonas putida KT2440 RelA (UniProt Q88MB8, gene relA, organism code PSEPK, NCBITaxon:160488)
Seed hypothesis under test: relA has GTP diphosphokinase activity (GO:0008728) — propagated as an IEA via TreeGrafter/PANTHER, original reference GO_REF:0000118.
Verdict: SUPPORTED — retain GO:0008728.

Summary

The automated phylogenetic pipeline (TreeGrafter/PANTHER) propagated GO:0008728 (GTP diphosphokinase activity) to Pseudomonas putida KT2440 RelA (Q88MB8). Our independent, primary-evidence evaluation finds this propagation correct. The protein is a canonical "long" RelA/SpoT homolog (RSH) of the RelA synthetase clade: its central (p)ppGpp-synthetase / nucleotidyltransferase domain is 58.7% identical to the experimentally validated Escherichia coli RelA synthetase, and the metal-coordinating catalytic FDDY motif is fully intact. GO:0008728 (EC 2.7.6.5) names precisely this (p)ppGpp-synthetase reaction, at the correct granularity — it is not a broad family-level surrogate, a sibling activity, or a mis-placed neighbor.

Crucially, none of the three characteristic TreeGrafter failure modes apply to the term under test. (1) Granularity: the sequence places in the RelA synthetase subfamily (PANTHER PTHR21262:SF31 "GTP PYROPHOSPHOKINASE"), and GO:0008728 is the matching specific term. (2) Pseudo-enzyme: the synthetase active site is intact and well conserved — this is not a catalytically dead fold. (3) Within-superfamily mis-placement: global alignment (50.5% identity to E. coli RelA vs 41.1% to SpoT) firmly assigns the protein to the RelA clade, the correct subfamily for this activity. Independent in-vivo genetics in the native organism corroborate the computational verdict: ∆relA/∆spoT mutants lose (p)ppGpp-dependent biofilm dispersal, which is restored by ectopic (p)ppGpp synthesis (PMID: 29273811).

One nuance sharpens — rather than undermines — the verdict. The protein's N-terminal HD hydrolase domain is degenerate: it retains the HD fold (Pfam PF13328) but has lost the catalytic His-Asp metal-binding pair, exactly like the hydrolase-dead E. coli RelA. This is the textbook architecture of a monofunctional RelA-type synthetase and has no bearing on GO:0008728. It does, however, flag that any co-annotated hydrolase term (GO:0008893) is the weaker, questionable call — a separate curator lead, outside this focus.

Key Findings

Finding 1 — Q88MB8 is a canonical RelA-type (p)ppGpp synthetase with an intact GTP-diphosphokinase active site (GO:0008728 directly supported)

The query protein has the full "long" RSH domain architecture expected of a catalytically competent (p)ppGpp synthetase: an N-terminal HD hydrolase domain (PF13328), a central (p)ppGpp-synthetase / nucleotidyltransferase domain (PF04607 RelA_SpoT; CDD cd05399 NT_Rel-Spo_like), the RelA_AH/RIS ribosome-interaction subdomains, and C-terminal TGS (PF02824) and ACT (PF13291) regulatory domains. PANTHER assigns it to subfamily PTHR21262:SF31 "GTP PYROPHOSPHOKINASE" — i.e., the synthetase subfamily, not merely the family root.

Sequence placement is unambiguous. Global Needleman–Wunsch alignment gives 50.5% identity to E. coli RelA (P0AG20), a validated (p)ppGpp synthetase, versus 41.1% to E. coli SpoT (P0AG24) — placing the query squarely in the RelA clade. Per-domain, the identity is concentrated exactly where it matters catalytically: the synthetase domain is 58.7% identical (98/167 residues) to E. coli RelA, the TGS domain 66.2%, and the ACT domain 41.3%, while the diverged N-terminal HD region is only ~34%. The synthetase catalytic core, including the conserved metal-coordinating FDDY motif (two catalytic aspartates), is intact: query ...PKEFDDYIANPK... versus E. coli RelA ...PDEFDDYVANPK....

In the native organism, this activity is realized biologically: P. putida RelA and SpoT are the (p)ppGpp synthetases, and relA/spoT mutants are defective in (p)ppGpp-dependent biofilm dispersal, with ectopic (p)ppGpp synthesis restoring the phenotype in a ∆relA ∆spoT background (PMID: 29273811). The conclusion — supported by both residue-level conservation and in-vivo phenotype — is that GO:0008728 names the protein's genuine, directly held molecular function.

Finding 2 — The HD hydrolase domain is degenerate, sharpening the verdict toward a monofunctional synthetase

The second finding distinguishes the RelA-type from the bifunctional SpoT-type architecture and localizes where (if anywhere) the broader annotation set is weak. The catalytic HD metal-binding motif of the Rel/SpoT hydrolase domain is present in the active bifunctional E. coli SpoT (...LMAALLHDVIED..., His-Asp at residue 72) but is absent in both the query and the hydrolase-dead E. coli RelA. At the homologous alignment position, the query reads ...KLDQDSLVAAVIYRSVR... — no catalytic His-Asp pair. N-terminal HD-region identity is correspondingly low and similar for both references (query vs RelA 35.6%, vs SpoT 32.9%), in stark contrast to the strongly conserved synthetase domain (58.7% vs RelA).

The interpretation is that the query retains the HD fold (Pfam PF13328 HD_4) but with a degenerate active site — the canonical RelA-type configuration of an active (p)ppGpp synthetase paired with a non-functional hydrolase. This has two consequences for curation. First, it reinforces that GO:0008728 (synthetase) is correct and intact. Second, it means that if GO:0008893 (guanosine-3',5'-bis(diphosphate) 3'-diphosphatase, the hydrolase activity) has also been propagated to this protein, that term is the one warranting scrutiny — not the term under test here.

Mechanistic Model / Interpretation

P. putida RelA catalyzes the committed step of the bacterial stringent response: synthesis of the alarmones (p)ppGpp by transfer of the β,γ-pyrophosphate of ATP onto the 3'-OH of GTP (→ pppGpp) or GDP (→ ppGpp).

   ATP  (β,γ-pyrophosphate donor)
     │
     ▼
   GTP / GDP ──► 3'-pyrophosphate transfer ──► pppGpp / ppGpp
     │
  [ Synthetase domain: PF04607 / cd05399 ]
  [ FDDY metal-binding motif INTACT — 58.7% id to E. coli RelA ]

   Domain architecture (N → C):
   ┌───────────────┬──────────────────────┬─────────────┬───────────────┬───────────────┐
   │ HD (PF13328)  │ Synthetase (PF04607) │ RelA_AH/RIS │ TGS (PF02824) │ ACT (PF13291) │
   │ DEGENERATE    │ ACTIVE               │ ribosome    │ regulatory    │ regulatory    │
   │ (no His-Asp)  │ (FDDY intact)        │ interaction │ (66.2% id)    │ (41.3% id)    │
   └───────────────┴──────────────────────┴─────────────┴───────────────┴───────────────┘

Functionally, the intact synthetase domain produces (p)ppGpp, which reprograms transcription — for example up-regulating σ54-dependent promoters such as the TOL-plasmid Pu promoter when relA is overexpressed (PMID: 10940009) — and drives physiological outputs such as biofilm dispersal (PMID: 29273811). The degenerate HD hydrolase domain means RelA, unlike bifunctional SpoT, cannot efficiently degrade (p)ppGpp, consistent with the canonical division of labor: RelA synthesizes, SpoT both synthesizes and hydrolyzes.

The TreeGrafter graft therefore succeeded for the synthetase function. The ancestral node from which GO:0008728 was propagated corresponds to the genuine, conserved catalytic activity of this protein, and the subfamily-level PANTHER placement (SF31) matches the specific GO term rather than over-generalizing to a family root.

TreeGrafter failure-mode audit

Active-Site / Placement Analysis

Subfamily placement

• PANTHER: PTHR21262 → subfamily SF31 "GTP PYROPHOSPHOKINASE."
• CDD: cd05399 NT_Rel-Spo_like (nucleotidyltransferase/synthetase core).
• Global Needleman–Wunsch identity: 50.5% to E. coli RelA (P0AG20) vs 41.1% to E. coli SpoT (P0AG24) → RelA synthetase clade.

Per-domain conservation (query vs E. coli RelA P0AG20)

Synthetase catalytic core alignment (E. coli RelA ~255–335)

Query: HIYSIWRKMQRKGLEFSQIYDVRAVRVLVPE-IRDCYTALGIVHTLWRHIPKEFDDYIANPKENGYRSLHTAVIGPEGKVLE
Eco  : HIYSIWRKMQKKNLAFDELFDVRAVR-IVAERLQDCYAALGIVHTHYRHLPDEFDDYVANPKPNGYQSIHTVVLGPGGKTVE
                                       ^^^^  FDDY catalytic/metal-binding motif

The metal-coordinating FDDY motif (two catalytic aspartates) is identical, and the surrounding active-site block is near-identical. Synthetase catalytic residues are present and correctly spaced — this is not a pseudo-enzyme for GO:0008728.

HD hydrolase site (for a different, co-annotated term — GO:0008893)

E. coli SpoT (active hydrolase):  KLDYETLMAAL L H D V I E D
Query (Q88MB8):                   KLDQDSLVAAV I Y R S V R
                              ^^^  catalytic His-Asp pair MISSING

The query retains the HD fold but the catalytic His-Asp metal-binding pair of the active SpoT-type hydrolase is absent, as in the hydrolase-dead E. coli RelA. Net: synthetase = intact; hydrolase = degenerate.

Evidence Base

• The stringent response promotes biofilm dispersal in Pseudomonas putida — PMID: 29273811. Directly supports the verdict. Verified snippet: "Mutants lacking elements of the stringent response - (p)ppGpp sythetases [RelA and SpoT] and/or DksA - were defective in biofilm dispersal. Ectopic (p)ppGpp synthesis restored biofilm dispersal in a ∆relA ∆spoT mutant." This establishes, in the native organism, that RelA functions as a (p)ppGpp synthetase — exactly the activity GO:0008728 describes.
• In vivo and in vitro effects of (p)ppGpp on the σ54 promoter Pu of the TOL plasmid of Pseudomonas putida — PMID: 10940009. Supports (indirect). Overexpression of relA — predicted to increase intracellular (p)ppGpp — produced a significant gain in Pu promoter activity, consistent with RelA being a productive (p)ppGpp synthetase with downstream transcriptional consequences.
• UniProt/InterPro domain annotation of Q88MB8 plus pairwise Needleman–Wunsch alignments to E. coli RelA (P0AG20) and SpoT (P0AG24). Supports (computational/evolutionary). Establish RelA-clade subfamily placement (PANTHER SF31) and residue-level conservation of the synthetase catalytic site; also reveal the degenerate HD hydrolase site.

A third retrieved paper (Acclimation of bacterial cell state... DmpR-dependent transcriptional activation, PMID: 32269250) proved tangential to the function assignment and is not used as support.

Evidence Matrix

GO Curation Implications

• Focus term GO:0008728 (GTP diphosphokinase activity): RETAIN. Correctly propagated, appropriately specific, active site intact, correct subfamily. An evidence-code upgrade from IEA is justified given the in-vivo support (PMID: 29273811) plus conserved catalytic residues. No generalization, specialization, or sibling replacement needed.
• Adjacent lead (not the focus term): If the co-annotated GO:0008893 (ppGpp 3'-diphosphatase/hydrolase) is present, it rests on the HD domain whose catalytic His-Asp motif is degenerate/absent in the query (matching hydrolase-dead E. coli RelA, not bifunctional SpoT). Suggested action for that sibling term: remove / treat-as-non-core, pending experimental confirmation. This is the annotation needing curator scrutiny — not GO:0008728.
• Any propagated CC term such as GO:0005886 (plasma membrane, IEA:TreeGrafter) would be a likely over-call for a cytoplasmic ribosome-associated synthetase, but it is outside this focus.
• "Protein binding" is not recommended as any final call; the molecular function is well-defined enzymatically.

Limitations and Knowledge Gaps

1. No purified-enzyme kinetics for Q88MB8 specifically. Catalytic competence is inferred from residue conservation and an in-vivo mutant phenotype, not from a direct in-vitro (p)ppGpp-synthesis assay on the P. putida protein.
2. Hydrolase status not experimentally measured in P. putida; the degenerate-HD call is computational.
3. Substrate preference (GTP vs GDP; pppGpp vs ppGpp output ratio) was not independently determined.
4. The active-site assessment is sequence/evolutionary; no solved Q88MB8 structure, mutagenesis, or independently rendered AlphaFold geometry was used in this run.
5. relA vs spoT ambiguity: P. putida encodes both; the query is closer to RelA (50.5% vs 41.1%), consistent with its gene name and a strong-synthetase/weak-hydrolase profile — no conflict for the synthetase term.

Proposed Follow-up Experiments / Actions

1. In-vitro (p)ppGpp synthetase assay on purified recombinant Q88MB8 (ATP + GTP/GDP → pppGpp/ppGpp; TLC or HPLC readout) — converts the IEA to direct-assay evidence (IDA) and definitively confirms GO:0008728.
2. Catalytic-Asp (FDDY) point-mutant complementation in a ∆relA ∆spoT background, scoring (p)ppGpp-dependent phenotypes (e.g., biofilm dispersal) — links the conserved catalytic residues to in-vivo synthetase function.
3. In-vitro (p)ppGpp hydrolysis assay — settles the GO:0008893 co-annotation question; expected negative given the degenerate HD motif.
4. Curator action: retain GO:0008728 (consider evidence upgrade); flag GO:0008893, if present, for review/removal; review any CC over-calls.

Bottom line

The TreeGrafter/PANTHER propagation of GO:0008728 (GTP diphosphokinase activity) to P. putida relA (Q88MB8) is correct and should be retained. The protein is a canonical RelA-type (p)ppGpp synthetase with an intact, conserved synthetase active site and in-vivo genetic confirmation of synthetase function; none of the three graft-failure modes (granularity, pseudo-enzyme, mis-placement) apply to the term under test.

Provenance: UniProt Q88MB8 / P0AG20 / P0AG24 retrieved programmatically; pairwise Needleman–Wunsch alignments and per-domain / catalytic-motif conservation computed in-session (code and outputs retained in the iteration log).

Artifacts

• OpenScientist final report
• OpenScientist final report