AIGR Gene Hypothesis Deep Research — Final Report

Target: Drosophila melanogaster Egm / dACAD9 (UniProt Q5U117, CG9006)

Focus: function-assignment — "medium-chain catalysis versus complex-I assembly"

Gene: Egm (a.k.a. Enigma, CG9006), D. melanogaster · UniProt: Q5U117 (639 aa, "Complex I assembly factor Egm, mitochondrial") · Focus type: function_assignment · Hypothesis slug: medium-chain-catalysis-versus-assembly · Iterations completed: 3

Summary

The seed hypothesis proposes that Egm/dACAD9 (a) catalyzes medium-chain fatty acyl-CoA dehydrogenation (GO:0070991) and (b) participates in medium-chain fatty acid catabolism (GO:0051793), and it asks that molecular activity and physiological process be assessed separately without treating the assembly role or the absence of fly assays as negative evidence for catalysis. After three iterations combining sequence orthology, domain analysis, active-site register mapping, ortholog enzymology, and phylogenetic-provenance checks, the conclusion is that the specific "medium-chain" descriptor is over-annotated and should not be retained at that specificity — while any generic catalytic role remains plausible but unproven in the fly.

Four independent lines converge. First, Egm is unambiguously the ACAD9/VLCAD (long/very-long-chain) clade ortholog — 37.2% identity to human ACAD9 and 36.9% to VLCAD versus only 28.0% to the canonical medium-chain enzyme MCAD/ACADM — and it carries the ACAD9/ACADV-specific InterPro domain (PF21343 / IPR049448), not an MCAD signature. Second, the best-characterized ortholog (human ACAD9) has a long-chain unsaturated substrate optimum, and even that residual FAO activity is described in the literature as biologically controversial. Third, the canonical ACAD catalytic proton-acceptor glutamate maps onto a glycine (Egm Gly446) in a register-anchored alignment, so even generic dehydrogenase catalysis by the fly protein is inferential. Fourth, live PANTHER v19 maps Egm to PTHR48083:SF2 carrying only generic GO terms, so the medium-chain IBA terms cannot be traced to the served phylogenetic node — corroborating the seed's own concern about the PTN000098033 / PTHR43884 provenance.

The most important caveat is that no direct fly acyl-CoA dehydrogenase assay exists for or against any substrate. The correct curation posture is therefore not "refuted" but "over-annotated": the medium-chain-specific MF and BP terms are unsupported by any fly experiment and contradicted in direction by orthology and ortholog enzymology, whereas the well-evidenced core function is mitochondrial complex-I assembly (GO:0032981; MCIA localization GO:0160295). Absence of fly assays is not treated as evidence against catalysis — but neither is it evidence for the medium-chain specificity the annotation claims.

Executive Judgment

Verdict: Over-annotated — weakly/partially supported for "catalysis in general"; the specific "medium-chain" descriptor is not supported and should be removed, generalized, or re-anchored to the long/very-long-chain clade.

Evaluating the two claims separately:

• Any acyl-CoA dehydrogenase catalysis by Egm is plausible but unproven in the fly. Egm carries an intact FAD-binding flavoprotein fold (GO:0050660, IEA) and is the clear ortholog of human ACAD9, an experimentally validated (if in-vivo-controversial) flavoenzyme. Catalysis cannot be excluded, and the assembly role does not exclude it. This is weakly supported by orthology only.
• The "medium-chain" specificity is the weakest possible choice and is unsupported. Egm is sequence-closest to the ACAD9/VLCAD (long/very-long-chain) clade and clearly separated from the medium-chain enzyme ACADM/MCAD. The ortholog's measured optimum is long-chain. On Egm the medium-chain terms are IBA-only, whereas FlyBase's ISS transfer gave Egm the very-long-chain term (GO:0017099), not medium-chain.
• The primary, best-evidenced function is complex-I assembly (GO:0032981; IDA localization to the MCIA complex GO:0160295; fly genetics in Murari 2021). The lipid/lifespan phenotypes (Mourikis 2006) are equally explained by loss of OXPHOS/complex-I as by a direct medium-chain flux, so they do not independently establish medium-chain catabolism.
• The catalytic base glutamate appears degenerate in Egm. A register-anchored alignment places the proton-acceptor Glu column (ACAD9 Glu426 / VLCAD Glu462) onto a glycine (Egm Gly446); Egm carries no annotated active site. This weakens any direct catalysis call (caveat: the ACAD catalytic base can migrate between subfamilies, so structure-based superposition is needed to be definitive).
• Domain + provenance corroboration. InterPro assigns Egm the ACAD9/ACADV-specific C-terminal domain and no MCAD signature; live PANTHER v19 maps Egm to PTHR48083:SF2 carrying only generic GO:0016627/GO:0006635 — the served node does not carry the medium-chain terms.

Bottom line for the curator: Do not equate the assembly role or the absence of fly assays with negative evidence for catalysis — but the medium-chain specificity is a mis-transfer. If any substrate-resolved MF term is retained, it should be long-/very-long-chain (GO:0004466 / GO:0017099) rather than medium-chain (GO:0070991), kept as non-core relative to the assembly function. The parallel BP term GO:0051793 is likewise better generalized or dropped.

Key Findings

Finding 1 — Egm is the ACAD9/VLCAD (long/very-long-chain) clade ortholog, not the medium-chain (MCAD) clade

Global Needleman-Wunsch alignment of the 639-aa Egm protein against the five human acyl-CoA dehydrogenase paralogs produced a clear clade ranking:

Egm is closest to ACAD9 and VLCAD — the long/very-long-chain and complex-I-assembly clade — and separated by roughly 9 percentage points from the medium-chain enzyme ACADM/MCAD, which (with LCAD) it resembles least. This is corroborated by nomenclature and annotation: UniProt names the protein "Complex I assembly factor Egm, mitochondrial," and FlyBase assigns Egm the long-chain term GO:0017099 (very-long-chain acyl-CoA dehydrogenase activity) by ISS, whereas the medium-chain term GO:0070991 is present only by IBA. The direction of the sequence evidence therefore points away from a medium-chain assignment and toward the long/very-long-chain + assembly identity. (Caveat: 37% identity is modest for any confident specificity transfer, and the ±1 scoring NW gives approximate absolute values, though the rank order is robust.)

Finding 2 — Ortholog kinetics are long-chain, not medium-chain; the fly protein has no direct catalytic assay

The seed correctly notes that human ACAD9 was tested across C6–C12 substrates, but the primary enzymology reports a long-chain optimum. Ensenauer et al. 2005 (PMID: 16020546) state that "Purified mature ACAD-9 had maximal activity with long-chain unsaturated acyl-CoAs as substrates (C16:1-, C18:1-, C18:2-, C22:6-CoA)," concluding a role in β-oxidation of long-chain unsaturated fatty acids. Broad substrate use with a long-chain optimum means preference is not exclusivity — but the optimum itself is long-chain, not medium-chain.

The knockout study whose identifier the seed asked to verify is Schiff et al. 2015, PMID: 25721401 = PMC4424958 — title verified to match exactly ("Complex I assembly function and fatty acid oxidation enzyme activity of ACAD9 both contribute to disease severity in ACAD9 deficiency"). It states "ACAD9 also retains enzyme ACAD activity for long-chain fatty acids in vitro, but the biological relevance of this function remains controversial partly because of the tissue specificity of ACAD9 expression." So even in the human ortholog, residual FAO activity is framed as long-chain and its in-vivo significance is disputed.

For the fly protein, catalysis is entirely inferential. Mourikis et al. 2006 (PMID: 16434470; gene then named Enigma) describe "a mitochondrial protein with homology to enzymes of the beta-oxidation of fatty acids and that mutations in this locus affect lipid homeostasis" — homology-based, with lifespan-extension, oxidative-stress-resistance, and lipid-homeostasis phenotypes, not a demonstrated dehydrogenase assay. Murari et al. 2021 (PMID: 34386730) address complex-I biogenesis. No direct acyl-CoA dehydrogenase enzyme assay exists for Egm, and certainly none demonstrating medium-chain specificity.

Finding 3 — Egm lacks the canonical ACAD catalytic proton-acceptor glutamate

The ACAD mechanism requires a catalytic base glutamate that abstracts the α-proton. UniProt annotates this proton acceptor at ACAD9 Glu426, VLCAD Glu462, and MCAD Glu401. Egm (Q5U117) has NO annotated active site. A register-anchored, ungapped alignment of the conserved catalytic "F-E-G-T…IALTGL" motif — anchored by the invariant C-terminal "IALTGL" block and the upstream "RD" motif — places the catalytic-Glu column on a glycine in Egm:

ACAD9  PYERILRDTRILLIF[E]GTNEILRMYIALTGL   catalytic Glu426
VLCAD  GVERVLRDLRIFRIF[E]GTNDILRLFVALQGC   catalytic Glu462
Egm    TTELGLRDAAQLCTQ[G]ESLDTLGMFIALTGL   Gly446 at catalytic column

The Egm residue at the catalytic column is Gly446; a glutamate sits one position C-terminal (Egm447), but that column corresponds to the ACAD post-catalytic glycine, not the catalytic base. Conserved substrate/FAD-contacting residues are retained in Egm (D394 = MCAD D278; R408 = MCAD R281), so the flavoprotein pocket is partly intact, but substitution of the proton-acceptor glutamate by glycine means even generic dehydrogenase catalysis is not structurally guaranteed. (This is a computational inference from register-anchored alignment, not a structural assay; the ACAD catalytic base can migrate between subfamilies, so an AlphaFold/experimental active-site superposition would be needed to be definitive.)

Finding 4 — Domain and PANTHER v19 provenance place Egm in the ACAD9/ACADV subfamily with only generic GO

Live InterPro for Q5U117 assigns the ACAD9/VLCAD-specific domains PF21343 "ACAD9/ACADV, C-terminal domain" and IPR049448 "ACAD9/ACADV-like, C-terminal domain," alongside the generic ACAD domains PF00441/PF02770/PF02771 — and no medium-chain (MCAD)-specific signature. The live PANTHER geneinfo service (product version 19) maps Q5U117 to family PTHR48083, subfamily PTHR48083:SF2, with GO-slim MF GO:0016627 "oxidoreductase activity, acting on the CH-CH group of donors," BP GO:0006635 "fatty acid beta-oxidation," and CC GO:0005737 "cytoplasm" — none medium-chain-specific. InterPro still lists the legacy family PTHR43884 for the same protein, but the served v19 node does not carry GO:0070991 or GO:0051793. This directly matches the seed's observation that the PAINT TSV places the medium-chain terms at PTN000098033 while the served PTHR43884 v19 tree lacks that node and the exact Egm leaf — i.e., the medium-chain IBA annotations cannot be re-derived from the tree node actually served today.

Mechanistic Model / Interpretation

The evidence supports a model in which Egm is a dual-clade ACAD9-type protein whose demonstrable core function is mitochondrial complex-I assembly, with any fatty-acid-oxidation role being ancestral, long/very-long-chain-biased, and unproven in the fly.

     Egm / dACAD9 (Q5U117, CG9006)
                |
  ┌─────────────────────┴──────────────────────┐
  |                                             |
   WELL-EVIDENCED CORE                        INFERRED / CONTESTED
   Complex-I assembly (MCIA)                  FAO catalytic activity
   - UniProt: "Complex I assembly            - ortholog optimum = LONG-chain
     factor Egm, mitochondrial"                unsaturated (PMID16020546)
   - fly complex-I biogenesis                - residual FAO "controversial"
     (PMID34386730)                            (PMID25721401)
   - human ACAD9 = 1 of 14 CI                - catalytic Glu -> Gly446
     assembly factors (PMID34556413)           (no annotated active site)
   GO:0032981 / MCIA CC GO:0160295           - NO direct fly assay
                             - MEDIUM-chain = orthology-
                               contradicted, IBA-only

Assessing the two axes separately, as the seed requests, sharpens rather than rescues the medium-chain call. On the molecular-function axis, three of the four evidence lines specifically undermine medium-chain specificity: identity is closest to the long/very-long-chain clade; the ortholog's measured optimum is long-chain unsaturated; and the catalytic base is degenerate. On the biological-process axis, the only fly phenotypes (lifespan, oxidative-stress response, lipid homeostasis) are consistent with a complex-I/OXPHOS defect and general lipid dysregulation, not with a specific block in medium-chain fatty acid catabolism. The dual-function biology of ACAD9 orthologs (Schiff 2015; the mouse model PMID: 34556413 describing "a combined defect in fatty acid oxidation and oxidative phosphorylation due to a dual role") means assembly and FAO functions genuinely coexist in the family — but the FAO half, where it operates, is long-chain.

Crucially, this interpretation does not treat the assembly role or the absence of fly assays as negative evidence against catalysis. The claim is narrower and more defensible: the specific medium-chain terms are (i) not supported by any direct fly experiment, (ii) contradicted in direction by the closest experimental ortholog, and (iii) untraceable to the currently served phylogenetic node that supposedly justifies them.

Evidence Base / Evidence Matrix

GO Curation Implications (leads — require curator verification)

Molecular Function
- GO:0070991 "medium-chain acyl-CoA dehydrogenase activity" [IBA] on Egm — recommend REMOVE or REPLACE. It is (a) IBA-only, (b) specificity-mismatched to Egm's ACAD9/VLCAD clade, and (c) not the optimum even for the human ortholog. If a substrate-resolved MF is desired, generalize to GO:0003995 (acyl-CoA dehydrogenase activity) or align to the clade with GO:0004466 (long-chain) / GO:0017099 (very-long-chain). Given the degenerate catalytic base, treat any MF as non-core / uncertain.
- Retaining GO:0050660 (FAD binding, IEA) is fine; it reflects an intact flavoprotein fold consistent with either catalytic or scaffold roles.
- Do not downgrade to "protein binding"; the informative MF is the assembly-relevant GO:0030674 (protein-macromolecule adaptor activity, ISS) or a generalized ACAD term.

Biological Process
- GO:0051793 "medium-chain fatty acid catabolic process" [IBA] — recommend generalize/remove. If kept, generalize to GO:0006635 (fatty acid beta-oxidation), which already has IMP support in fly, and treat as non-core.
- GO:0032981 "mitochondrial respiratory chain complex I assembly" — retain as core BP (TAS + fly genetics + MCIA localization).

Cellular Component
- Retain GO:0160295 (MCIA complex) and GO:0005759 (mitochondrial matrix) — IDA-grade, core.

GO decision table (leads — require curator verification)

Net: the assembly branch is core and well-supported; the medium-chain catalysis/catabolism branch is a phylogenetic carry-over that should be non-core and either generalized or removed, not retained at "medium-chain" specificity.

Mechanistic Scope

The immediate molecular function being tested is FAD-dependent α,β-dehydrogenation of a medium-chain (C6–C12) fatty acyl-CoA to the 2-enoyl-CoA. Direct evidence for this specific reaction on Egm is absent (no fly assay). Nearest evidence = human ACAD9 in vitro, which favors long-chain unsaturated substrates. The catalytic base appears substituted (Gly446), further weakening a direct-catalysis interpretation.

The assembly function — chaperoning complex-I module assembly within the MCIA complex — is a distinct molecular activity (protein–protein scaffolding, not catalysis) and is the better-supported core. Downstream/derived observations that must not be conflated with the direct medium-chain claim include: fly lipid-homeostasis, lifespan extension, and oxidative-stress resistance (PMID: 16434470), all confounded by Egm's OXPHOS/complex-I role; human disease manifestations (cardiomyopathy, lactic acidosis) driven substantially by the assembly defect; and any inference drawn purely from loss of function, which cannot by itself localize activity to medium-chain substrates.

Conflicts and Alternatives

1. Specificity mismatch (clade/paralog confusion). The "medium-chain" label properly belongs to ACADM/MCAD, a distinct subfamily (28% identity to Egm). Egm sits with ACAD9/VLCAD. Propagating MCAD's flagship term across the whole PTHR family is classic paralog over-annotation / frequency bias.
2. In-vitro-only, controversial activity. Even in human, ACAD9's FAO activity is "controversial" and tissue-specific (Schiff 2015; mouse model 2021). A physiological medium-chain catabolism claim for the fly is doubly indirect.
3. Assembly vs. catalysis primacy. UniProt/FlyBase both name the protein an assembly factor first. The seed correctly warns not to treat the assembly role as negative evidence for catalysis — but the converse also holds: the lipid phenotypes are not positive evidence for direct catalysis.
4. Ancestral-node provenance break (PTN000098033) — partially confirmed. The seed flags that the PAINT TSV places both medium-chain terms at PTN000098033 while the served PTHR43884 v19 tree lacks that node/leaf. Live checks corroborate a mismatch: PANTHER v19 geneinfo maps Egm to PTHR48083:SF2 (different family than the InterPro-listed legacy PTHR43884), and that node carries only generic GO:0016627 and GO:0006635, not the medium-chain terms. No lineage was invented from the donor list; a curator should confirm against the authoritative PAINT export.

Limitations and Knowledge Gaps

Discriminating Tests / Proposed Follow-up Experiments

1. Recombinant Egm ACAD assay (ETF/DCPIP-coupled) across a C4–C18 acyl-CoA panel — directly tests whether Egm is catalytic and, if so, its chain-length optimum. Prediction under seed: measurable C6–C12 activity; prediction under alternative: negligible activity or long-chain optimum.
2. Active-site residue audit / structure — align Egm to MCAD/VLCAD/ACAD9 and test the catalytic base Glu and FAD-contacting residues; AlphaFold pocket geometry as a chain-length ruler. Loss of the catalytic Glu → assembly-scaffold (pseudo-enzyme) interpretation.
3. Separation-of-function rescue — express in Egm-null flies a catalytically-dead but assembly-competent Egm variant; if lipid/β-oxidation phenotypes persist while complex I is restored, catalysis is dispensable (phenotypes were OXPHOS-driven).
4. Acylcarnitine profiling of Egm mutants — a medium-chain acylcarnitine elevation (C6–C10) would support genuine medium-chain flux; a generalized/long-chain pattern or a pure OXPHOS signature would not.
5. Cross-clade complementation — test whether human ACAD9 (long-chain/assembly) but not human MCAD rescues Egm phenotypes — cross-clade rescue confirms functional orthology to ACAD9, not MCAD.
6. Provenance audit — re-map Egm through the exact GO/PANTHER release cited in the IBA annotation to confirm whether PTN000098033 legitimately carries the medium-chain terms for the Egm leaf.

Curation Leads (require curator verification)

• Action lead: On Q5U117, treat GO:0070991 [IBA] and GO:0051793 [IBA] as over-specified phylogenetic carry-over; recommend generalize (to GO:0003995 / GO:0006635) or remove, pending PTN000098033 re-verification. Keep the complex-I assembly branch (GO:0032981, GO:0160295) as core.
• Reference/snippet leads to verify:
• PMID: 16020546 — "Purified mature ACAD-9 had maximal activity with long-chain unsaturated acyl-CoAs as substrates (C16:1-, C18:1-, C18:2-, C22:6-CoA)."
• PMID: 25721401 (PMC4424958, verified) — "ACAD9 also retains enzyme ACAD activity for long-chain fatty acids in vitro, but the biological relevance of this function remains controversial."
• PMID: 16434470 — "Enigma encodes a mitochondrial protein with homology to enzymes of the beta-oxidation of fatty acids and that mutations in this locus affect lipid homeostasis."
• PMID: 34386730 — fly complex-I biogenesis genetics (supports assembly, not catalysis).
• Candidate replacement MF terms: GO:0004466 (long-chain) or GO:0017099 (very-long-chain), matching Egm's ACAD9/VLCAD clade, or the parent GO:0003995 — not GO:0070991.
• Suggested curator questions: (a) Is there any direct Drosophila acyl-CoA dehydrogenase assay for Egm? (b) Does the served PTHR43884 v19 tree actually support PTN000098033 for these terms? (c) Should substrate-resolved MF terms be demoted to non-core given the primacy of the assembly role?

Computational Provenance (this run)

• Needleman-Wunsch global identity, Egm (Q5U117, 639 aa) vs human paralogs (UniProt FASTA, match=+1/mismatch=−1/gap=−1): ACAD9 (Q9H845) 37.2%, VLCAD (P49748) 36.9%, SCAD (P16219) 29.3%, MCAD (P11310) 28.0%, LCAD (P28330) 27.5%. → Egm ∈ ACAD9/VLCAD clade, not MCAD clade.
• GO label resolution (QuickGO): GO:0070991 = medium-chain fatty acyl-CoA dehydrogenase activity (MF); GO:0004466 = long-chain…; GO:0017099 = very-long-chain…; GO:0032981 = mitochondrial respiratory chain complex I assembly (BP).
• UniProt GO evidence codes pulled live for Q5U117/Q9H845/P11310/P49748/P28330 (see Evidence Matrix).
• Catalytic-residue mapping: UniProt "Proton acceptor" active sites = ACAD9 Glu426, VLCAD Glu462, MCAD Glu401; Egm has none annotated. Register-anchored ungapped alignment of the catalytic motif window (anchored by invariant C-terminal IALTGL and upstream RD) maps the catalytic Glu of ACAD9/VLCAD onto Egm Gly446 (nearest Glu one residue off-register at Egm447). Conserved substrate/FAD residues in Egm: D394 (=MCAD D278), R408 (=MCAD R281).
• InterPro (live): Egm carries PF21343 / IPR049448 "ACAD9/ACADV C-terminal domain" plus generic ACAD domains; no MCAD-specific signature.
• PANTHER v19 geneinfo (live): Q5U117 → PTHR48083:SF2 with GO-slim MF GO:0016627, BP GO:0006635, CC GO:0005737; no medium-chain term.

Limitations: identities from a simple ±1 scoring NW (not BLOSUM), so absolute values are approximate but the rank order (ACAD9 ≈ VLCAD > SCAD/MCAD/LCAD) is robust. No direct fly enzyme data exist to test; the catalytic-base finding is an alignment inference pending structural confirmation; the PTHR43884 v19 / PTN000098033 discrepancy could not be fully reconstructed programmatically and is reported as an open provenance gap rather than resolved.

Bottom Line

Egm/dACAD9 (Q5U117) is an ACAD9/VLCAD-clade complex-I assembly factor. Its well-supported core function is mitochondrial complex-I assembly; any fatty-acid-oxidation activity is ancestral, long/very-long-chain-biased, contested even in the human ortholog, and structurally uncertain in the fly (degenerate catalytic base). The seed's medium-chain MF (GO:0070991) and BP (GO:0051793) terms are IBA-only, orthology-contradicted, and untraceable to the served phylogenetic node — they should be removed or generalized/re-anchored to long/very-long-chain and treated as non-core. This judgment does not treat the assembly role or the absence of fly assays as positive evidence against catalysis; it finds no support for the specific medium-chain claim.

Artifacts

• OpenScientist final report
• OpenScientist final report