AIGR TreeGrafter Function-Inference Stress Test — Final Report

Target: Heterocephalus glaber (naked mole-rat) Cgas — UniProt A0AAX6RS70 (554 aa)
Seed hypothesis (TreeGrafter/PANTHER, IEA, GO_REF:0000118): Cgas has 2',3'-cyclic GMP-AMP synthase activity (GO:0061501)
Verdict: SUPPORTED — the propagated term is correct and already at maximal specificity. RETAIN.

Summary

The naked mole-rat protein A0AAX6RS70 ("Cgas") was electronically annotated with 2',3'-cyclic GMP-AMP synthase activity (GO:0061501) by the TreeGrafter/PANTHER phylogenetic pipeline. The task was to independently judge whether this protein directly performs that function, and — because automated propagation can silently fail — to actively test three characteristic failure modes: (1) granularity (the term being a broad family function rather than this protein's specific subfamily activity), (2) pseudo-enzyme status (a conserved fold with a degenerate active site), and (3) within-superfamily mis-placement (grafting onto a structurally related but functionally distinct neighbor of the shared Mab-21 nucleotidyltransferase fold).

Across three iterations of analysis — domain-architecture inspection, global pairwise alignment to characterized human and mouse cGAS, residue-level active-site mapping, and a paralog-discrimination panel across the Mab-21 superfamily — all three failure modes were explicitly tested and excluded. The protein was grafted to the cGAS-specific subfamily node (PANTHER PTHR10656:SF35, "CYCLIC GMP-AMP SYNTHASE"), not the broad family node, so there is no granularity problem. Its complete catalytic apparatus — the three-acidic-residue Mg²⁺-coordinating triad and the zinc thumb — is conserved with zero substitutions, insertions, or deletions, so it is not a pseudo-enzyme. And its nearest characterized neighbor, by a wide margin, is cGAS itself (≈57% global identity to both human and mouse cGAS), ruling out paralog mis-placement.

The conclusion is that the seed hypothesis is SUPPORTED: A0AAX6RS70 is a bona fide mammalian cGAS ortholog that directly performs the 2',3'-cGAMP synthase reaction (ATP + GTP → 2',3'-cGAMP). GO:0061501 should be retained at its current granularity; no generalization, sibling-replacement, or removal is warranted. The only residual limitation is the absence of a direct biochemical assay in H. glaber — all evidence is sequence/structure-based orthology inference, which is strong but formally inferential.

Key Findings

Finding 1 — A0AAX6RS70 is a genuine cGAS ortholog with fully intact catalytic machinery

The UniProt record for A0AAX6RS70 (554 aa, Heterocephalus glaber) carries the submission name "Cyclic GMP-AMP synthase isoform X1" and the canonical cGAS domain signature: Pfam PF03281 (Mab-21) plus PF20266 (Mab-21_C), and InterPro IPR046903 (Mab-21-like nucleotidyltransferase, residues 246–420) plus IPR046906 (HhH / H2TH-like, residues 437–541). This is the two-lobed architecture of mammalian cGAS: an N-terminal nucleotidyltransferase (NTase) catalytic lobe fused to a C-terminal Mab-21 helix-hairpin-helix / zinc-thumb DNA-binding lobe.

Crucially, PANTHER placed the protein at PTHR10656:SF35 = "CYCLIC GMP-AMP SYNTHASE" — a subfamily-level node specific to cGAS, not the broad pan-Mab-21 family node. The GO propagation therefore originated from the cGAS-specific clade, which immediately removes the most common granularity failure: the term inherited is the subfamily function, not a generic family function.

A Needleman–Wunsch global alignment (BLOSUM62) of A0AAX6RS70 against human CGAS (Q8N884) gives 57.2% whole-chain identity (297/519 aligned positions) and 76.9% identity across the catalytic core (human positions ~212–421). The whole-chain divergence is concentrated in the intrinsically disordered, low-complexity N-terminus (roughly residues 28–197), which is not part of the catalytic machinery and is expected to vary across species. Within the catalytic core, every functional residue is conserved:

There are no substitutions, insertions, or deletions at any active-site, catalytic, or metal-binding position, and the spacing between residues is preserved. The E–D–D triad coordinates the two divalent metal ions (Mg²⁺, or Mn²⁺) required for the two-metal-ion nucleotidyl-transfer chemistry, and the C₃H₁ zinc thumb stabilizes the activation loop and mediates DNA binding. A protein that conserves this entire apparatus in a mammal performs the mammal-specific 2',3'-cGAMP reaction. This excludes the pseudo-enzyme failure mode.

The canonical reaction is firmly established in the primary literature: cGAS is activated by cytosolic double-stranded DNA and uses ATP and GTP to synthesize the cyclic dinucleotide second messenger 2',3'-cGAMP (PMID: 42810574; PMID: 41806557). This reaction corresponds precisely to GO:0061501.

Finding 2 — The sequence is cGAS, not a Mab-21 family paralog (rules out within-superfamily mis-placement)

The Mab-21 superfamily contains several paralogs (MAB21L1, MAB21L2, MAB21L3, MB21D2) that share the NTase/Mab-21 fold but lack cGAS catalytic activity. To exclude the third failure mode — grafting onto a structurally related but functionally distinct neighbor — we computed the global Needleman–Wunsch percent identity of A0AAX6RS70 against each characterized superfamily member:

The nearest neighbors are the two cGAS orthologs, and the margin over the next-closest paralog (MAB21L3) is large — ~12 percentage points, and the MAB21L3 match spans only 258 positions versus full-length coverage for cGAS. Over the next tier (MAB21L1/2) the margin is ~23 percentage points. Decisively, the cGAS-diagnostic zinc thumb (His422/Cys428/Cys429/Cys436) is present in A0AAX6RS70 — this motif is a hallmark of cGAS and is absent from the MAB21L1/2 paralogs that lack cGAS activity. The combination of (i) best global match to cGAS, (ii) full-length coverage of the catalytic core, and (iii) the cGAS-specific zinc thumb jointly excludes any within-superfamily mis-placement.

Mechanistic Model / Interpretation

The logical structure of the stress test — and how each failure mode was tested and excluded — is summarized below.

       TreeGrafter grafts A0AAX6RS70 → PANTHER PTHR10656:SF35
          propagates GO:0061501 (IEA)
                     |
  -------------------------------------------------------
  |                          |                          |
  (1) GRANULARITY?           (2) PSEUDO-ENZYME?         (3) MIS-PLACEMENT?
  Broad family node, so      Fold intact but active     Grafted onto a
  term is too general?       site degenerate?           different Mab-21
                                        paralog, not cGAS?
  |                          |                          |
  Node = SF35, cGAS-         E257/D259/D351 triad       Best global match =
  SPECIFIC subfamily.        intact; zinc thumb         cGAS (57.2%), >12 pts
  No child term exists;      H422/C428/C429/C436        above nearest paralog;
  term already maximally     intact; 0 indels at        cGAS-specific zinc
  specific.                  active site.               thumb present.
  |                          |                          |
     NOT A PROBLEM              NOT A PROBLEM              NOT A PROBLEM
  -------------------------------------------------------
                     |
       VERDICT: SUPPORTED — RETAIN GO:0061501

The domain architecture of cGAS (and of A0AAX6RS70):

 N-term (disordered, divergent)   NTase catalytic lobe        Mab-21 / zinc-thumb DNA-binding lobe
 |----------~245 aa-------------|--IPR046903 (246–420)-----|--IPR046906 HhH/H2TH (437–541)--|
                   Mg2+ triad:                 Zinc thumb:
                   E257, D259, D351            H422, C428, C429, C436

Biologically this is fully coherent. cGAS is the principal cytosolic dsDNA sensor of the cGAS–STING innate-immune axis. On binding dsDNA it undergoes allosteric activation and, using ATP and GTP with two catalytic metal ions, synthesizes 2',3'-cGAMP — the cyclic dinucleotide that binds STING and triggers type-I interferon signaling. The defining feature of the mammalian enzyme is the mixed 2'-5'/3'-5' phosphodiester linkage of its product (hence "2',3'-cGAMP"), which is exactly what GO:0061501 encodes. The naked mole-rat ortholog preserves every residue that executes this chemistry, and its sequence divergence is confined to the disordered, non-catalytic N-terminus, which is typical of cross-species cGAS comparison. Nothing in the sequence suggests a functional pivot to a different activity.

Evidence Base

How the literature supports the call. PMID: 42810574 and PMID: 41806557 jointly establish that the canonical, defining activity of mammalian cGAS is dsDNA-activated synthesis of 2',3'-cGAMP from ATP and GTP — i.e., exactly GO:0061501. PMID: 32814054 confirms the two-metal-ion mechanism that the conserved E257/D259/D351 triad serves. PMID: 28940468 provides a careful qualifier: in vitro, the 2',3'-cGAMP product can be minor relative to linear di-nucleotides — but this concerns catalytic efficiency and product ratio, not the identity of the enzyme or the validity of assigning GO:0061501. None of the reviewed literature challenges the assignment.

Active-Site / Placement Analysis (Provenance)

Method: Global pairwise Needleman–Wunsch alignment (BLOSUM62) of A0AAX6RS70 against human cGAS (Q8N884) and against a panel of characterized Mab-21 superfamily paralogs (Q8C6L5, Q13394, Q9Y586, Q8N816, Q8IYB1). Active-site positions were mapped from human cGAS numbering to the naked mole-rat sequence via the alignment. Sequences retrieved from UniProt REST; alignment executed in-session.

Catalytic / metal-binding residue conservation — verdict: FULLY INTACT

Subfamily placement — verdict: cGAS is the nearest characterized neighbor (see Finding 2 table). Catalytic-core identity to human cGAS is 76.9%; whole-chain 57.2% (297/519), with divergence confined to the disordered N-terminus.

Explicit statement: The catalytic and metal-binding residues are intact and correctly spaced, and the protein's nearest characterized neighbor is cGAS. Both independent tests converge on an active cGAS enzyme. No pseudo-enzyme signature; no paralog mis-placement.

GO Curation Implications

Recommended curator action (lead requiring verification): RETAIN GO:0061501 (2',3'-cyclic GMP-AMP synthase activity).

• The term is correct and at the right granularity — same granularity as the seed; it is a leaf-level molecular-function term encoding the diagnostic mixed 2'-5'/3'-5' linkage, so no more-specific child exists ("make-more-specific" not applicable).
• Do not generalize to a parent nucleotidyltransferase term — that would discard well-supported specificity.
• Do not replace with a sibling term — there is no evidence of a diverged or sibling activity; the paralog panel excludes MAB21L-type functions.
• Do not remove — the active site is intact and the ortholog assignment is unambiguous.
• Treat as a core molecular-function annotation for cGAS.

Because the current evidence code is IEA (electronic, GO_REF:0000118), the annotation remains inference-based. The convergent active-site conservation and paralog-distance evidence substantially raise confidence and could justify a curator-reviewed phylogenetic evidence note; a direct assay would be required to upgrade to an experimental code (IDA). "Protein binding" is explicitly not an appropriate fallback.

Limitations and Knowledge Gaps

1. No direct biochemical assay in H. glaber. All conclusions rest on orthology plus active-site conservation. Conserved residues strongly predict, but do not prove, catalytic competence and the exact 2',3' product regiochemistry.
2. Alignment-based residue mapping. Active-site positions were transferred from human cGAS numbering via global alignment rather than from an experimental naked mole-rat structure. An AlphaFold model superposition would add geometric confirmation but was not run.
3. Product-ratio nuance. Human-cGAS kinetics (PMID: 28940468) note that 2',3'-cGAMP can be a minor product relative to linear nucleotides in vitro. This concerns output proportions, not enzyme identity or the validity of GO:0061501.
4. Database carry-over. UniProt submission name and PANTHER/InterPro calls are themselves automated; however they are mutually consistent and corroborated by the independent active-site alignment, so paralog confusion (Mab21L1/2 or cGLR) is ruled out by the cGAS-diagnostic residue set.
5. Species-specific innate-immune biology. Naked mole-rats have notable immune and longevity peculiarities; whether NMR cGAS has altered regulation (DNA affinity, STING coupling) is unknown — relevant to downstream biological-process annotation but outside the scope of this molecular-function term.
6. Isoform scope. The record is "isoform X1"; alternative isoforms were not individually assessed.

Proposed Follow-up Experiments / Actions

1. Direct activity assay (definitive). Express recombinant NMR Cgas, activate with dsDNA, incubate with ATP/GTP + Mg²⁺, and detect 2',3'-cGAMP by LC-MS/MS (with linkage analysis to confirm the 2'-5'/3'-5' bond) or via a STING-reporter activation readout. This is the single most efficient experiment to convert IEA → IDA and settle any residual doubt.
2. AlphaFold structural superposition. Superpose the AlphaFold model of A0AAX6RS70 on a human cGAS–DNA–nucleotide complex to confirm the 3D spacing of E257/D259/D351 and the zinc thumb, complementing the sequence-level conservation.
3. Phylogenetic tree reconstruction. Build a maximum-likelihood tree across the Mab-21 superfamily including the query, to visually confirm it falls inside the cGAS clade rather than at an internal paralog-splitting node — independent corroboration of the PANTHER SF35 graft.
4. Curator note. Record the active-site conservation table and paralog-distance panel as provenance supporting retention of GO:0061501, documenting that all three TreeGrafter failure modes were explicitly tested and excluded.

Conclusion

The TreeGrafter/PANTHER propagation of GO:0061501 to naked mole-rat Cgas (A0AAX6RS70) is correct. The protein is a genuine mammalian cGAS ortholog grafted to the cGAS-specific subfamily node (PTHR10656:SF35), with the complete catalytic Mg²⁺-coordinating triad (E257/D259/D351 = human E225/D227/D319) and zinc thumb (H422/C428/C429/C436 = human H390/C396/C397/C404) perfectly conserved, and with cGAS as its unambiguous nearest characterized neighbor (~57% global identity, ~12–23 points above the next paralog). None of the three characteristic failure modes — granularity, pseudo-enzyme, or within-superfamily mis-placement — applies. Curation action: RETAIN at current granularity.

Provenance: UniProt REST (A0AAX6RS70, Q8N884, Q8C6L5, Q13394, Q9Y586, Q8N816, Q8IYB1); custom Needleman–Wunsch alignment (BLOSUM62) executed in-session; PubMed PMIDs 41806557, 42810574, 28940468, 32814054, 41738955.

Artifacts

• OpenScientist final report
• OpenScientist final report