Focus type: function_assignment
Hypothesis slug: substrate-chain-length-binding-and-hydroxylase-chemistry
Target: Human ACOX2, UniProt Q99424, NCBITaxon:9606 (Homo sapiens)
Source: genes/human/ACOX2/ACOX2-ai-review.yaml (free-text)
Verdict: Partially supported — one core claim supported, one refuted as a term-definition defect, and two judged over-annotated. The seed hypothesis bundles four distinct claims about human ACOX2, and — as the seed explicitly requested — they must be evaluated independently. They resolve very differently. ACOX2's function as a FAD-dependent peroxisomal acyl-CoA oxidase acting on 2-methyl-branched-chain acyl-CoAs and C27 bile-acid-CoA intermediates is well supported by direct human enzymology and human deficiency/complementation studies, and this is its non-redundant core role. In contrast, the "exact GO:0033791 cholestanoyl-CoA 24-hydroxylase reaction" is not the chemistry ACOX2 performs: the enzyme dehydrogenates (desaturates) the bile-acid side chain to a 24-ene enoyl-CoA + H₂O₂ (RHEA:46728), and the 24-hydroxyl is installed downstream by a different protein, the D-bifunctional protein (DBP/HSD17B4).
The single most important message for a curator is that the GO:0033791 problem is a term-definition/annotation-mapping defect, not an absence of activity. The IDA to GO:0033791 (from PMID:27884763) captured the correct enzyme, the correct substrate class, and the correct biological step (bile-acid side-chain shortening). What is wrong is that the live logical definition of GO:0033791 — and its EC (1.17.99.3) and RHEA (15733) cross-references — describe a hydroxylation (25R substrate + H₂O + acceptor → 24R,25R-tetrahydroxy product), which differs from ACOX2's actual reaction on three independent axes: stereochemistry (25S vs 25R), oxidant (O₂ vs H₂O/acceptor), and product (24-ene enoyl-CoA + H₂O₂ vs a 24-hydroxy product). This is exactly the distinction the seed asked us to make: annotation error versus absence of a secondary activity. Here it is an annotation defect layered on top of a genuine, correctly identified enzyme.
The remaining two claims — free-fatty-acid binding (GO:0005504) and very-long-chain fatty-acid metabolic process (GO:0000038) — are IBA-only phylogenetic inferences with no ACOX2-specific direct evidence. Mechanistically ACOX2 binds acyl-CoA thioesters rather than free fatty acids, and straight-chain/VLCFA oxidation is the assigned role of the paralog ACOX1; ACOX2's own straight-chain capacity reaches only ~C16 at low efficiency, while its genuinely long substrates are branched C27 sterols, not aliphatic VLCFAs (>22-carbon tail). Both are candidate non-core annotations. None of these judgments relies on donor count, preferred substrate, or missing target assays, in accordance with the seed's constraints.
The curated UniProt Q99424 catalytic-activity block lists exclusively oxidase/desaturation reactions of the canonical acyl-CoA oxidase form — a 2,3-saturated acyl-CoA + O₂ = a (2E)-enoyl-CoA + H₂O₂ (e.g., RHEA:38959) — with FAD as the cofactor. The bile-acid reaction at the heart of the hypothesis is RHEA:46728:
(25S)-3α,7α,12α-trihydroxy-5β-cholestan-26-oyl-CoA + O₂ = (24E)-3α,7α,12α-trihydroxy-5β-cholest-24-en-26-oyl-CoA + H₂O₂
This is a dehydrogenation introducing a 24,25 double bond (a 24-ene), consuming O₂ and producing H₂O₂ — the diagnostic signature of a flavin acyl-CoA oxidase. By contrast, the GO term GO:0033791 "cholestanoyl-CoA 24-hydroxylase activity" is defined as a hydroxylation:
(25R)-3α,7α,12α-trihydroxy-5β-cholestan-26-oyl-CoA + H₂O + acceptor = (24R,25R)-3α,7α,12α,24-tetrahydroxy-5β-cholestan-26-oyl-CoA + reduced acceptor
cross-referenced to EC 1.17.99.3 / RHEA:15733. The two reactions differ on three independent chemical axes, summarized below:
| Feature | ACOX2 real reaction (RHEA:46728) | GO:0033791 definition (RHEA:15733) |
|---|---|---|
| Substrate stereochemistry | 25S | 25R |
| Oxidant / co-substrate | O₂ | H₂O + acceptor |
| Product | (24E)-enoyl-CoA + H₂O₂ | (24R,25R)-24-hydroxy tetrahydroxy-CoA + reduced acceptor |
| Chemistry | α,β-dehydrogenation (desaturation) | hydroxylation |
| Cofactor | FAD (oxidase) | acceptor-dependent hydroxylase |
The IDA supporting GO:0033791 traces to PMID:27884763(https://pubmed.ncbi.nlm.nih.gov/27884763/), which frames ACOX2 as "an acyl-CoA oxidase … involved in the shortening of C27 cholesterol derivatives to generate C24 bile acids" and shows that "THCA biotransformation into cholic acid was enhanced in cells overexpressing ACOX2." The experiment therefore demonstrated oxidase / side-chain-shortening chemistry, not a discrete hydroxylation. Notably, GO:0033791's own synonyms include "THCA-CoA oxidase activity," revealing that the term was intended to capture this enzyme entity — which is why the annotation should be read as a legacy term-definition defect (rooted in an old EC 1.17.99.3 "hydroxylase" nomenclature) rather than a wrong gene-to-function link.
The substrate scope of the human enzyme is anchored in direct biochemistry. Vanhove and colleagues (PMID:8387517(https://pubmed.ncbi.nlm.nih.gov/8387517/)) purified from human liver and kidney "a novel branched chain acyl-CoA oxidase, which oxidizes the CoA esters of 2-methyl-branched fatty acids as well as those of the bile acid intermediates" di- and trihydroxycoprostanic acids — establishing a single human enzyme with a dual substrate class (branched-chain FA-CoA + bile-acid-CoA) and no separate dedicated human THCA-CoA oxidase.
The modern substrate-specificity and deficiency study (PMID:29287774(https://pubmed.ncbi.nlm.nih.gov/29287774/)) partitions the family cleanly: "ACOX2 is the only human acyl-CoA oxidase involved in bile acid biosynthesis," whereas "ACOX1 is responsible for the oxidation of straight-chain fatty acids with different chain lengths." ACOX2-deficient fibroblasts retained normal pristanic-acid oxidation, indicating redundancy with ACOX3 for branched-chain fatty acids; bile-acid intermediate oxidation is the non-redundant ACOX2 role.
The UniProt FUNCTION annotation records that ACOX2 can oxidize straight-chain decanoyl-CoA (C10; RHEA:40179) and hexadecanoyl-CoA (C16; RHEA:40167) but only with low efficiency, alongside pristanoyl-CoA and other methyl-branched-chain fatty acyl-CoAs. The seed rightly cautions that low C10/C16 efficiency is measurable capacity, not absence — and this report honors that: the point is not that ACOX2 lacks straight-chain activity, but that its straight-chain reach extends only to ~C16 at low efficiency, which cannot satisfy the two IBA terms:
Both are therefore candidate non-core annotations — reasonable family inferences, but not representative of ACOX2's direct, demonstrated activity.
Ferdinandusse and colleagues (PMID:15769750(https://pubmed.ncbi.nlm.nih.gov/15769750/)) state that "peroxisomal beta-oxidation is an essential step in bile acid synthesis, since it is required for shortening of C27-bile acid intermediates to produce mature C24-bile acids" and that "D-Bifunctional protein (DBP) is responsible for the second and third step of this beta-oxidation process." In the classic β-oxidation cycle, ACOX2 performs only step 1 (oxidase), and the hydroxyl is introduced at step 2 (hydratase, DBP). The hydroxyl group that GO:0033791's definition attributes to a single "24-hydroxylase" is therefore the product of DBP hydration of the ACOX2-generated enoyl-CoA — a downstream event catalyzed by a different protein. This is corroborated by the rat-liver enzymology of Novikov and colleagues (PMID:7929456(https://pubmed.ncbi.nlm.nih.gov/7929456/)), which shows the first reaction is an acyl-CoA oxidase and that the 3-hydroxyacyl-CoA dehydrogenase/hydratase steps are separate enzymes (including the multifunctional protein).
A QuickGO ontology search returned no GO molecular-function term cross-referenced to RHEA:46728 (ACOX2's actual bile-acid oxidase reaction) or to RHEA:15733. The most accurate available MF parent is GO:0003997 acyl-CoA oxidase activity, whose definition — "a 2,3-saturated acyl-CoA + O₂ = a (2E)-enoyl-CoA + H₂O₂" — exactly matches ACOX2's chemistry and is already present as an IEA annotation. On the BP side, GO:0033540 (fatty acid beta-oxidation using acyl-CoA oxidase) likewise specifies the oxidase first step producing H₂O₂, matching ACOX2's role.
All four hypothesis components map onto a single, well-understood peroxisomal β-oxidation cycle. ACOX2 performs only the first step; the "hydroxylase" chemistry belongs to a different enzyme two steps later.
C27 bile-acid-CoA intermediate (e.g. (25S)-THC-26-oyl-CoA)
│
│ ACOX2 (FAD oxidase, RHEA:46728) ← the demonstrated reaction
│ acyl-CoA + O2 → (24E)-enoyl-CoA + H2O2
▼
(24E)-3α,7α,12α-trihydroxy-cholest-24-en-26-oyl-CoA
│
│ DBP / HSD17B4 (hydratase, step 2) ← installs the 24-OH
│ enoyl-CoA + H2O → 3-hydroxyacyl-CoA
▼
24-hydroxy(3-hydroxyacyl)-CoA
│
│ DBP / HSD17B4 (dehydrogenase, step 3)
▼
3-keto-acyl-CoA
│
│ SCPx thiolase (step 4): cleave C3 unit (propionyl-CoA)
▼
C24 bile acid (e.g. cholyl-CoA → cholic acid)
Reading the seed's specific chemical claims against this model:
The seed's phylogenetic scaffolding is consistent with this interpretation. PTHR10909 leaf PTN002474948 descending from a hydroxylase node (PTN008508564) and from fatty-acid-binding / long-chain-oxidation nodes (PTN000097533, PTN000097706) reflects family-level inheritance of MF/BP terms. Human ACOX2 legitimately seeds the hydroxylase IBD, but the IBA-propagated terms (fatty acid binding, VLCFA process) are precisely the ones that are over-general for the specific human enzyme, whose experimentally demonstrated activity is narrower (branched-chain + bile-acid-CoA oxidase).
| Citation | Evidence type | Relationship | Claim tested | Key finding | Context | Confidence / limitations |
|---|---|---|---|---|---|---|
| PMID:27884763(https://pubmed.ncbi.nlm.nih.gov/27884763/) | Direct assay + human mutant phenotype | Refutes literal 24-hydroxylase; supports oxidase | 24-hydroxylation vs oxidase side-chain shortening | "Acyl-CoA oxidase (ACOX2) is involved in the shortening of C27 cholesterol derivatives to generate C24 bile acids"; overexpression enhances THCA→cholic acid | Human; ACOX2-deficiency patient + overexpression cells | High for oxidase role; this is the source of the GO:0033791 IDA — supports term-definition-defect reading |
| UniProt Q99424 (RHEA:46728, RHEA:38959, RHEA:40167, RHEA:40179) | Database (reaction/cofactor curation) | Qualifies/refutes hydroxylase; supports oxidase | Exact catalyzed reaction | All reactions are FAD-dependent oxidase form (acyl-CoA + O₂ → 2E-enoyl-CoA + H₂O₂); bile-acid reaction yields (24E)-enoyl-CoA from 25S substrate | Human, curated | High; database-level but reaction is chemically explicit |
| GO:0033791 definition / EC 1.17.99.3 / RHEA:15733 | Database (ontology definition) | Competing definition | Is GO:0033791 chemistry the same as ACOX2's? | Term defined as hydroxylation (25R + H₂O/acceptor → 24-hydroxy product); differs in stereochem, oxidant, product | Ontology | High; direct inspection of live term; synonym "THCA-CoA oxidase" shows intent |
| PMID:8387517(https://pubmed.ncbi.nlm.nih.gov/8387517/) | Direct assay (enzyme purification) | Supports core function | ACOX2 substrate scope | Single human "branched chain acyl-CoA oxidase" oxidizes 2-methyl-branched FA-CoA and bile-acid intermediate CoAs | Human liver/kidney | High; foundational human biochemistry |
| PMID:29287774(https://pubmed.ncbi.nlm.nih.gov/29287774/) | Direct assay + mutant phenotype | Supports core; qualifies VLCFA/FA-binding | Is ACOX2 the bile-acid oxidase? the VLCFA oxidase? | "ACOX2 is the only human acyl-CoA oxidase involved in bile acid biosynthesis"; "ACOX1 is responsible for the oxidation of straight-chain fatty acids with different chain lengths" | Human; patient fibroblasts, redundancy with ACOX3 | High; directly assigns straight-chain/VLCFA to ACOX1 |
| PMID:15769750(https://pubmed.ncbi.nlm.nih.gov/15769750/) | Mutant phenotype (KO) + pathway | Refutes ACOX2-as-24-hydroxylase | Who installs the 24-OH? | DBP catalyzes steps 2–3 of peroxisomal β-oxidation shortening C27→C24 | Mouse L-/D-BP knockouts | High; localizes hydroxyl-installing steps to DBP |
| PMID:7929456(https://pubmed.ncbi.nlm.nih.gov/7929456/) | Direct assay (purification) | Supports step separation | Are oxidase and hydroxyacyl steps separate? | First step is acyl-CoA oxidase; 3-hydroxyacyl-CoA dehydrogenase/hydratase are distinct enzymes | Rat liver peroxisomes | High for enzymology; rat |
| QuickGO/GO MF search for RHEA:46728 | Computational (ontology lookup) | Qualifies curation | Is there a GO MF for ACOX2's exact reaction? | No GO MF maps to RHEA:46728 or RHEA:15733; nearest accurate parent is GO:0003997 | Ontology | Medium-high; negative search result |
| GO:0005504 (IBA, GO_REF:0000033), GO:0000038 (IBA) | Database (evidence-code inspection) | Qualifies (over-general) | Are FA-binding and VLCFA-process directly supported? | Both IBA-only, family inference; not backed by direct ACOX2 assays | Ontology/QuickGO | Medium-high; evidence-code based |
Literature notes:
The following are leads requiring curator verification, organized by GO aspect.
Molecular Function
Biological Process
Cellular Component
The immediate molecular function under test is a single-turnover FAD-dependent acyl-CoA oxidase reaction: abstraction of the 2,3 (α,β) hydrogen pair from a bile-acid-CoA or branched-chain acyl-CoA, transferring electrons to O₂ to form H₂O₂ and creating an enoyl-CoA double bond. For the C27 bile-acid substrate this places the double bond at the 24,25 position (24-ene), the committed first step of side-chain shortening.
Everything else in the hypothesis is either (a) a downstream pathway consequence — the 24-hydroxyl (DBP hydration), the eventual C24 bile-acid product, and cholic-acid output — or (b) a family-level generalization — free-fatty-acid binding and VLCFA metabolism. The disease manifestation (ACOX2 deficiency: persistent hypertransaminasemia, C27 accumulation, C24 bile-acid deficiency) is a loss-of-function phenotype, informative for confirming the pathway role but not itself evidence of hydroxylase chemistry. Curators should not let the loss-of-function bile-acid phenotype license a "24-hydroxylase" MF; the phenotype is equally explained by loss of the oxidase step.
All items below are leads requiring curator verification.
Candidate action changes
Candidate reference snippets to verify
Candidate GO terms
Suggested curator questions
Suggested experiments — see Discriminating Tests (¹⁸O₂ product ID, H₂O₂ stoichiometry, 25R/25S kinetics, DBP-knockout intermediate profiling, VLCFA kinetic panel, free-FA vs acyl-CoA binding).
A computed GO decision table (acox2_go_decision_table.csv) accompanies this report, listing the relevant GO terms with their current evidence codes and recommended curator leads, generated from the live UniProt/RHEA/QuickGO records and primary-literature evidence gathered across the three investigation iterations.
ACOX2's branched-chain and bile-acid acyl-CoA oxidase function is well supported and should anchor the review (GO:0003997 + bile-acid biosynthesis BP). The seed's "exact GO:0033791 cholestanoyl-CoA 24-hydroxylase reaction" is not performed by ACOX2 as the term is defined — the IDA captured the right enzyme and the right biological step but is bound to a term whose definition/EC/RHEA describes hydroxylase chemistry that belongs, in the pathway, to DBP hydration downstream. Treat this as a term-definition defect to flag, not proof of a hidden hydroxylase activity. The free-fatty-acid-binding (GO:0005504) and VLCFA-metabolic-process (GO:0000038) claims are IBA-only, over-general, and are candidate non-core annotations. None of these judgments rests on donor count, preferred substrate, or missing target assays, per the seed's constraints.