ABHD18 — review notes

PAINT no-IBA project review, using the affinage deep-research provider plus UniProt Q0P651,
the GOA TSV and the primary literature.

The gap

Two GO annotations, both mitochondrion. No molecular function, no biological process —
despite UniProt naming the protein "Cardiolipin-specific deacylase, mitochondrial" and carrying
a Rhea-referenced CATALYTIC ACTIVITY block (RHEA:32935).

Two independent 2025 papers characterised the enzyme. Both are now used.

Two framings of the same enzymology

This is the interesting curation problem, and it only surfaced because a PR reviewer noticed the
affinage record cited a paper I had not adjudicated.

PMID:40903572 (Nature) PMID:40378955 (JBC)
Framing the missing deacylase of the remodelling cycle a degradative lipase, homolog of yeast Cld1
ABHD18/TAZ relationship partners in a hydrolysis–reacylation cycle "ABHD18 catalyzes the breakdown of CL, whereas TAZ protects CL from degradation"
Extent of reaction CL → MLCL stepwise, past MLCL to dilyso-CL
Clinical angle Barth syndrome therapeutic target — ABHD18 loss rescues TAZ-mutant phenotypes knockdown lowers MLCL in Taz-KO myoblasts

Both describe the same chemistry. GO:0035965 cardiolipin acyl-chain remodeling is retained
because both agree ABHD18 supplies the deacylation step of the ABHD18/tafazzin pair — and because
GO has no cardiolipin catabolic process term, so the degradative reading cannot currently be
expressed at all. Filed as a suggested_question.

My first draft said "removes one acyl chain". PMID:40378955 is explicit that this is wrong:
"Rather than removing just one fatty acid, we show that ABHD18 deacylates CL further." Corrected
in the description, core_functions and the proposed term.

Why A2-type, not the generic parent

RHEA:32935 writes the product as 1'-[1,2-diacyl-sn-glycero-3-phospho],3'-[1-acyl-sn-glycero-3-phospho]-glycerol.
The chain remaining on the deacylated glycerol is at sn-1, so the chain removed was at
sn-2 — which is A2 regiochemistry. GO:0004623 A2-type glycerophospholipase activity is
therefore more precise than GO:0004620, and is used.

The GO gap

GO resolves cardiolipin metabolism at substrate level almost everywhere — synthase
(GO:0008808), CMP-forming synthase (GO:0043337), phospholipase D (GO:0035755),
dehydrogenase (GO:0160241) — but has nothing for deacylation, even though that step is half
of GO:0035965. The gap is historical: the enzyme was unidentified until 2025.

Proposed: cardiolipin deacylase activity, defined to cover the stepwise reaction rather than
only the first product.

Provider note

Affinage's prose added little here — UniProt had already curated the primary paper with a Rhea
cross-reference, and for a gene with one recent, promptly-curated paper the provider is
redundant. But an earlier draft concluded it "added nothing", and that was wrong: its
citation list contained PMID:40378955, which I had missed and which materially qualifies the
framing. A concrete case where the provider's value was in its reference list, not its
narrative.