AIGR TreeGrafter Function-Inference Stress Test — PP_2320 (Q88KH0)

Target: PP_2320 / Q88KH0, Pseudomonas putida KT2440 (NCBITaxon:160488)
Seed hypothesis (TreeGrafter / PANTHER, IEA, GO_REF:0000118): PP_2320 has peptidoglycan L,D-transpeptidase activity (GO:0071972)
Focus: function_assignment · function-hypothesis-go-0071972

Executive Judgment

Verdict: SUPPORTED — retain GO:0071972.

The automated phylogenetic propagation is correct in this case. None of the three
characteristic failure modes applies:

• (1) Granularity / family-vs-subfamily: The protein maps to the single-domain
  ErfK/YbiS/YnhG group of the L,D-transpeptidase family (PANTHER PTHR30582:SF24).
  The family-level molecular function term GO:0071972 is the correct level of
  granularity — see caveat below on the exact sub-reaction.
• (2) Pseudo-enzyme / loss of activity: The active site is intact. The
  catalytic His193 and the catalytic nucleophile Cys209 are both conserved and
  correctly spaced; Cys209 is in fact the only cysteine in the entire protein,
  so the catalytic thiol is unambiguously present.
• (3) Within-superfamily mis-placement: The sequence belongs squarely to the
  YkuD / L,D-transpeptidase fold and subfamily, not a structurally related but
  functionally distinct neighbor. Pfam PF03734, PROSITE PS52029, CDD cd16913
  (YkuD_like), and Gene3D 2.40.440.10 all agree.

Single most decisive piece of evidence: By Smith-Waterman alignment (live
UniProt sequences), PP_2320's catalytic dyad residues map one-to-one onto the
experimentally annotated active sites of characterized E. coli L,D-transpeptidases
— His193→YbiS His194 / ErfK His191, and Cys209→YbiS Cys210 / ErfK Cys207 — at
37–44% sequence identity. A degenerate pseudo-enzyme would show loss or mis-spacing
of exactly these residues; it does not.

The sole residual uncertainty is not about whether the molecular function term
is right, but about which sub-reaction the enzyme performs (lipoprotein anchoring
vs. 3→3 peptidoglycan cross-linking). That distinction is sequence-ambiguous here
and should not be over-specified. The evidence remains computational (IEA): no
direct enzymatic assay of PP_2320 itself exists.

Independent Family/Function Assignment

From first-principles analysis of the sequence, domain architecture, and active-site
conservation — independent of the propagated annotation — PP_2320 is a secreted
(periplasmic) L,D-transpeptidase of the ErfK/YbiS/YnhG class.

• Most likely specific molecular function: peptidoglycan L,D-transpeptidase
  activity — cysteine-nucleophile catalysis forming L,D (3→3) linkages in/attached to
  peptidoglycan.
• Characterized homolog the assignment is based on: E. coli YbiS/LdtB
  (UniProt P0AAX8; 43.5% identity, nearest characterized neighbor), closely followed
  by YnhG/LdtE (P76193; 41.6%) and ErfK/LdtA (P39176; 36.9%).
• GO term I would assign: GO:0071972 (peptidoglycan L,D-transpeptidase
  activity) — same granularity as the seed term. I would not descend to a more
  specific child term (e.g., a dedicated Braun's-lipoprotein-anchoring term vs. a
  3→3 cross-linking term), because the two nearest characterized homologs carry
  different sub-roles and are near-equidistant (43.5% vs. 41.6%), making the exact
  sub-reaction undeterminable from sequence alone.

Conclusion on granularity: seed term = assigned term = GO:0071972. Retain.

Key Findings

Finding 1 — PP_2320 is a catalytically intact L,D-transpeptidase-family protein

Independent primary-evidence analysis of UniProt Q88KH0 (325 aa) establishes a
complete, canonical L,D-transpeptidase architecture and an intact active site.

Domain architecture. The protein comprises (i) an N-terminal Sec signal peptide
(residues ~1–25) directing periplasmic/secreted localization; (ii) a LysM
peptidoglycan-binding domain (CDD cd00118 / InterPro IPR018392), which tethers the
enzyme to its peptidoglycan substrate; and (iii) a C-terminal L,D-transpeptidase
(YkuD) catalytic domain spanning approximately residues 98–233 (Pfam PF03734,
PROSITE PS52029, CDD cd16913 "YkuD_like", Gene3D 2.40.440.10, InterPro
IPR005490). This is the textbook domain layout of a periplasmic L,D-transpeptidase.

Catalytic residues intact and correctly spaced. The catalytic machinery is fully
present: His193 (proton donor/acceptor, in the local context "PGYLIHGSNKK") and
the canonical S-H-G-C motif at positions 206–209 with Cys209 as the catalytic
nucleophile (local context "RTSHGCFRMFN"). Critically, Cys209 is the only cysteine
in the entire protein — so the catalytic thiol nucleophile is unambiguously conserved
and cannot be confused with any other cysteine.

Structural confidence. AlphaFold model AF-Q88KH0 has a global pLDDT of 84.4,
with 63.7% of residues in the very-high-confidence band; the catalytic domain is
well-folded in the model, consistent with a competent enzyme rather than a disordered
pseudo-enzyme.

Family placement. PANTHER assigns the protein to PTHR30582 (L,D-transpeptidase
family), subfamily SF24 (ErfK/SrfK-related). The L,D-transpeptidase family uses a
catalytic cysteine nucleophile to form L,D (3→3) peptidoglycan cross-links, the exact
activity described by GO:0071972
(PMID: 30514506).

Finding 2 — Nearest characterized homologs are E. coli ErfK/YbiS with 1:1 catalytic-residue correspondence

Smith-Waterman alignment (BLOSUM62, gap penalty −8; sequences fetched live from
UniProt) against characterized L,D-transpeptidases gives:

In both alignments, the query's catalytic residues map one-to-one onto the
targets' experimentally annotated active-site residues:

• query His193 → ErfK His191 / YbiS His194 (target residue = H, confirmed)
• query Cys209 → ErfK Cys207 / YbiS Cys210 (target residue = C, confirmed)

Both targets belong to the same PANTHER subfamily as PP_2320 (PTHR30582:SF24,
ErfK/SrfK-related). This is the strongest single discriminator against the
"pseudo-enzyme" and "mis-placement" failure modes: the catalytic dyad is not merely
present but is positionally homologous to residues with direct experimental support.
Provenance: provenance/PP_2320_alignment.py.

Finding 3 — Single-domain ErfK/YbiS/YnhG group; MF robust, exact sub-reaction sequence-ambiguous

Extending the alignment panel across both functional sub-clades of the family:

PP_2320 clearly belongs to the single-domain ErfK/YbiS/YnhG group (top three
hits), and in all three the catalytic His193 and Cys209 align one-to-one to the
targets' annotated active sites (His194/Cys210, His193/Cys209, His191/Cys207
respectively). However, the two closest hits carry different sub-roles —
YbiS/LdtB anchors Braun's lipoprotein to peptidoglycan, while YnhG/LdtE forms 3→3
cross-links — and they are near-equidistant (43.5% vs. 41.6%). Consequently, the
family-level molecular function (GO:0071972) is robustly supported, but the precise
chemical sub-reaction cannot be resolved from sequence alone and should not be
over-specified. Provenance: provenance/PP_2320_subclade.py.

Mechanistic Model / Interpretation

The convergent lines of evidence assemble into a coherent, internally consistent model
of a functional periplasmic L,D-transpeptidase:

   N ──[ Sec signal ]──[ LysM PG-binding ]──────[ YkuD / L,D-TPase catalytic ]────── C
(~1–25)          (CDD cd00118)            (~98–233; PF03734, PS52029)
secretion        substrate tether         His193 ···· Cys209 (sole Cys)
                                   proton shuttle   nucleophile
                                        │
                                        ▼
         Cys-thioester acyl-enzyme intermediate on a
         peptidoglycan stem peptide (meso-DAP, position 3)
                                        │
                                        ▼
    ┌───────────────────────────────────────────────────┐
    │  L,D (3→3) transpeptidation — one of two sub-roles: │
    │   • 3→3 peptidoglycan cross-link (YnhG/LdtE-like)   │
    │   • Braun's Lpp anchoring to PG (YbiS/ErfK-like)    │
    │  ← sequence cannot discriminate (43.5% vs 41.6%)    │
    └───────────────────────────────────────────────────┘

Why the propagation succeeded here. TreeGrafter grafted PP_2320 onto a node whose
diagnostic feature — the YkuD catalytic domain with its Cys-His dyad — is precisely the
feature PP_2320 retains. Unlike the three classic propagation failures, there is no
granularity mismatch (the term is at the correct family level), no active-site decay
(the dyad is intact and uniquely identifiable), and no superfamily mis-placement (every
independent domain database agrees on YkuD/PF03734). The LysM substrate-binding module
and Sec signal peptide further corroborate that this is a bona fide periplasmic
peptidoglycan-remodeling enzyme operating in the compartment where L,D-transpeptidation
occurs.

The one honest caveat. The L,D-transpeptidase family has functionally diverged into
at least two sub-activities that share an identical catalytic mechanism and fold:
3→3 cross-link formation and lipoprotein anchoring. PP_2320's two nearest characterized
neighbors sit on opposite sides of this divide at nearly identical sequence identity.
The shared molecular-function umbrella term GO:0071972 correctly captures what we can
assert; descending to a sub-reaction-specific term would overreach the evidence.

Evidence Base / Matrix

Note: no paper reports a direct enzymatic assay of PP_2320/Q88KH0 itself; all direct-assay
evidence is on homologous family members.

How the key papers bear on the findings.
PMID: 30514506 establishes the shared family
mechanism — "Formation of 3 → 3 crosslinks in the peptidoglycan layer of M. tuberculosis
is catalyzed by l,d-transpeptidases" — anchoring the inference that PP_2320's conserved
Cys-His dyad performs the GO:0071972 activity.
PMID: 39294212 and
PMID: 31024745 provide direct biochemical and
crystallographic characterization of a representative family member (Mtb Ldt2), confirming
the enzyme class is genuinely catalytic.
PMID: 33941679 qualifies the picture by showing
that LDT paralogs (LdtF) can act on Braun's lipoprotein rather than form cross-links,
which is exactly why the sub-reaction of PP_2320 must be treated as ambiguous.

Active-Site / Placement Analysis

Catalytic residues: INTACT.

Subfamily placement: PANTHER PTHR30582:SF24 (ErfK/SrfK-related), single-domain
ErfK/YbiS/YnhG group. Independent confirmation from Pfam (PF03734), PROSITE (PS52029),
CDD (cd16913 YkuD_like), Gene3D (2.40.440.10), InterPro (IPR005490) — all consistent.

Conclusion: The catalytic dyad is present, correctly spaced, and positionally
homologous to experimentally annotated active sites of characterized family members.
This decisively rules out the pseudo-enzyme and superfamily-misplacement failure modes.
Provenance scripts: provenance/PP_2320_alignment.py, provenance/PP_2320_subclade.py.

GO Curation Implications

Recommended curation action: RETAIN GO:0071972 (peptidoglycan L,D-transpeptidase
activity) at the family level — lead requiring curator verification.

• Do NOT remove. The active site is intact; there is no evidence of pseudo-enzyme
  status.
• Do NOT generalize. The term is already at the appropriate family granularity; a
  broader parent would lose information.
• Do NOT make more specific. Although the enzyme clearly performs L,D-transpeptidation,
  the exact sub-reaction (3→3 cross-linking vs. Braun's-lipoprotein anchoring) is
  sequence-ambiguous — the two nearest characterized homologs carry different sub-roles
  at near-identical identity (43.5% vs. 41.6%). Descending to a sub-reaction-specific
  child term would exceed the evidence.
• Do NOT replace with a sibling term. No evidence points to a neighboring,
  functionally distinct activity.
• Evidence tier: keep as IEA / computational. There is no direct assay or mutant
  phenotype for PP_2320 itself; all catalytic evidence is by homology and in-silico
  active-site conservation. This supports retaining the term but not upgrading the
  evidence code to an experimental tier.
• Not "protein binding." Avoided as a recommendation, per instructions; the
  catalytic MF term is well justified here.

The accompanying cellular-component interpretation (periplasmic/secreted, via the Sec
signal peptide and LysM module) and biological-process context (peptidoglycan remodeling
/ cell wall organization) are consistent and could be curated alongside the MF term if
not already present.

Limitations and Knowledge Gaps

• No direct experimental data for PP_2320 itself. There is no enzymatic assay,
  experimental structure, or mutant phenotype for Q88KH0; all catalytic evidence is
  homology-based (IEA). This is sufficient to retain the term but not to upgrade the
  evidence tier.
• Sub-reaction ambiguity. The two nearest characterized homologs (YbiS/LdtB, 43.5%;
  YnhG/LdtE, 41.6%) carry different sub-roles, so sequence cannot discriminate whether
  PP_2320 primarily forms 3→3 cross-links or anchors a lipoprotein.
• Organism-specific envelope differences. P. putida does not share the identical
  Braun's-lipoprotein configuration of E. coli, complicating direct transfer of the
  "Lpp-anchoring" sub-role; this is unresolved by sequence alone.
• Structural evidence is predicted, not experimental. AlphaFold confidence is good
  globally (pLDDT 84.4) but is a model.
• Family can harbor hydrolase paralogs. LdtF-type paralogs cleave rather than form
  bonds (PMID: 33941679); PP_2320's intact
  transpeptidase dyad argues against this, but it is a formal caveat.

Proposed Follow-up Experiments / Actions

In priority order, the most efficient discriminating analyses are:

1. In-vitro activity assay of recombinant PP_2320 with (a) a peptidoglycan-mimetic
   donor/acceptor to test 3→3 cross-linking and (b) a Braun's-lipoprotein peptide to test
   anchoring — directly resolves the sub-reaction and upgrades the evidence tier.
2. Cys209→Ala active-site mutant + activity assay to confirm the nucleophile is
   catalytically required (distinguishes true enzyme from pseudo-enzyme).
3. ΔPP_2320 phenotyping in P. putida KT2440: muropeptide HPLC/MS to quantify 3→3
   cross-link abundance, plus β-lactam/carbapenem susceptibility shifts, to place the
   enzyme in its native cell-wall-remodeling role.
4. Experimental structure (X-ray / cryo-EM) to confirm the dyad geometry and
   substrate channel predicted by AlphaFold.

Immediate curation action: mark GO:0071972 as retained (verified by independent
active-site analysis), keep IEA evidence, annotate the sub-reaction as undetermined,
and flag follow-up assay #1 as the decisive experiment.

Summary (for busy readers)

PP_2320 (Q88KH0) from Pseudomonas putida KT2440 is a bona fide, catalytically intact
L,D-transpeptidase, and the TreeGrafter/PANTHER-propagated term GO:0071972
(peptidoglycan L,D-transpeptidase activity) is SUPPORTED and should be RETAINED. The
protein has the canonical architecture — Sec signal peptide, LysM peptidoglycan-binding
domain, and a YkuD catalytic domain (Pfam PF03734 / PROSITE PS52029) — and its catalytic
dyad (His193 and the sole-cysteine nucleophile Cys209) is fully conserved and correctly
spaced. Smith-Waterman alignment to characterized E. coli enzymes (YbiS/LdtB 43.5%,
YnhG/LdtE 41.6%, ErfK/LdtA 36.9% identity) maps these catalytic residues one-to-one onto
experimentally annotated active sites.

Systematic testing for the three characteristic propagation-failure modes returns
negative on all three: the term is at the correct granularity (family level), the
active site is intact (no pseudo-enzyme signal), and the sequence sits firmly in the
YkuD/L,D-transpeptidase subfamily (no superfamily mis-placement). The only residual
uncertainty is the precise sub-reaction — 3→3 peptidoglycan cross-linking versus Braun's
lipoprotein anchoring — which is sequence-ambiguous because PP_2320's two nearest
characterized homologs carry different sub-roles at nearly equal identity. The family-level
molecular function term correctly captures what the evidence supports; it should not be
made more specific.

All evidence is computational / by-homology (IEA): no direct assay of PP_2320 exists. The
recommended curation action is therefore to retain GO:0071972 at the family level, keep
the IEA evidence tier, and not over-specify the sub-reaction pending an in-vitro activity
assay or a ΔPP_2320 muropeptide-profiling experiment.

Artifacts

• OpenScientist final report
• OpenScientist final report