Human PNPLA3 / adiponutrin / iPLA2-epsilon. Patatin-like phospholipase domain-containing
protein 3; closest paralogue of PNPLA2/ATGL. Single-pass type II membrane protein that
partitions between ER/membranes and the lipid droplet (LD) surface. Catalytic dyad
Ser47 (nucleophile) / Asp166 (proton acceptor) in the PNPLA domain (UniProt Q9NST1).
The common missense variant I148M (rs738409) is the single strongest common genetic risk
factor for steatotic liver disease. Three incompatible mechanistic models are currently in
print, and GOA carries annotations that are traceable to two mutually exclusive
biochemical claims. This file records what each primary source actually reports.
The founding biochemistry. Three iPLA2 family members (adiponutrin/PNPLA3, TTS-2.2, GS2)
expressed in Sf9 cells and affinity purified.
PMID:15364929
PMID:15364929
Note what the abstract does not report: no phospholipase A2 assay result for these three
proteins. The "iPLA2-epsilon"/"phospholipase A2" naming is a family-membership label based on
the dual GXGXXG + GXSXG signature motifs, not a measured PLA2 rate. GOA nevertheless carries
GO:0004623 A2-type glycerophospholipase activity IDA from this paper, and UniProt carries
EC 3.1.1.4 with ECO:0000269|PubMed:15364929. Flagged below as a probable name-driven
over-annotation.
PMID:21878620
PMID:21878620
PMID:21878620
PMID:21878620
Full text, results section heading:
PMID:21878620
Curation problem. GOA carries GO:0003841 1-acylglycerol-3-phosphate O-acyltransferase
activity with evidence code EXP and PMID:21878620 as the reference, with a plain enables
qualifier (no NOT). That paper reports the opposite: no detectable LPAAT activity, with
CGI-58 as a positive control. The annotation is traceable to UniProt's EC 2.3.1.51 /
RHEA:19709 catalytic-activity block, which tags both PubMed:21878620 and PubMed:22560221 as
ECO:0000269 evidence; only the latter reports a positive result. Either the annotation should
carry a NOT qualifier or the reference is wrong. Recorded in the review and in
suggested_questions; the term itself is kept (see below) because a different paper does
report the activity.
So two groups, using purified/immunoisolated protein, reach opposite conclusions on the same
reaction, and they also disagree on the direction of the I148M effect (Huang/He: loss of
hydrolase; Kumari: gain of acyltransferase). A protein cannot be the physiologically relevant
catalyst of both the hydrolysis and the synthesis of the same acyl linkage; at most one of
these is the reaction that matters in vivo. GOA currently carries both, which is why the
lipase and acyltransferase branches are adjudicated differently below rather than all
accepted at face value.
PMID:24511104
A cell-based readout that is compatible with either a lipase or a transacylase mechanism; it
constrains the net metabolic effect rather than the elementary reaction.
Johnson et al. 2024, Nat Commun.
PMID:38844467
This paper explicitly frames the open question:
PMID:38844467
Wang et al. 2025, J Hepatol.
PMID:39550037
PMID:39550037
PMID:39550037
PMID:39550037
Crucially, the same paper is the first to show that PNPLA3's own lipase activity is
ABHD5-stimulated, i.e. it does not deny that PNPLA3 is a lipase:
PMID:39550037
PMID:39550037
The genetic argument against pure LOF, from the same discussion: Pnpla3-/- mice do not
develop steatosis even on high-carbohydrate diets, whereas 148M knock-in mice do, and a
catalytically dead S47A knock-in phenocopies 148M only when the protein is expressed.
Sherman et al. 2025, Cell Rep (abstract only in cache).
PMID:41046517
PMID:41046517
Teskey et al. 2025, J Biol Chem.
PMID:39814233
PMID:39814233
Mikaeeli & Cohen 2025, J Hepatol, is a dedicated editorial titled "Loss or gain of function:
The functional complexity of the PNPLA3 I148M variant" (PMID:39892821). PubMed holds no
abstract for it, so it is cited as evidence that the dispute is recognised rather than quoted.
Is PNPLA3's physiological molecular function catalytic at all? The honest answer is that
it is both catalytic and non-catalytic, and the non-catalytic arm is the one that carries
the disease signal.
Triacylglycerol lipase activity (GO:0004806) is real and is kept as a core molecular
function. Four independent groups measured TG hydrolysis with purified protein
(Jenkins 2004, He 2010, Huang 2011, Wang 2025), it is ABHD5-stimulated like ATGL's, and
Johnson 2024 gives it a substrate preference (polyunsaturated TG) and an in vivo readout.
What is disputed is not the activity but its quantitative importance: Pnpla3-null mice
have no hepatic phenotype, so the lipase is dispensable in mouse liver under standard
conditions.
The ABHD5-competition arm is not represented in GOA at all, and it should be. All
three 2024-25 models converge on ABHD5/CGI-58 as the node. Wang 2025 shows WT PNPLA3 —
not only 148M — binds ABHD5 preferentially over ATGL and competes with ATGL for it; 148M
differs by escaping degradation and accumulating ~40x on LDs, so the same molecular
activity is simply present in much greater amount. The molecular function is therefore
sequestration of a lipase co-activator, i.e. inhibition of ATGL. GO:0140311 protein
sequestering activity is the closest existing MF term and is used, together with the
downstream BP GO:0010897 negative regulation of triglyceride catabolic process; both are
added as action: NEW annotations and as a second core function. Neither GO:0140311 nor
GO:0055102 lipase inhibitor activity captures "sequesters the co-activator of a lipase",
so a child term lipase activator sequestering activity is proposed.
The acyltransferase branch is demoted, not deleted. GO:0003841,
GO:0042171 and GO:0016411 rest on Kumari 2012 (plus Reactome TAS) and are directly
contradicted by Huang 2011 with an appropriate positive control. Marked
MARK_AS_OVER_ANNOTATED throughout, with the PMID:21878620 provenance error called out.
The transacylation terms GO:0051264/GO:0051265 are different and are ACCEPTed: those are
the CoA-independent mono-olein/diolein transacylase reactions that Jenkins 2004 measured
directly, and CoA-independent transacylation is chemically compatible with the same Ser47
nucleophile that performs hydrolysis, so it does not contradict the lipase activity the way
the acyl-CoA-dependent LPAAT claim does.
A2-type glycerophospholipase (GO:0004623) is marked over-annotated. It is a
family-name inference; Huang 2011 tested phospholipids and found little or no activity.
Rodent-orthology BP transfers (GO:0009744 response to sucrose, GO:0032869 cellular
response to insulin stimulus, GO:0050872 white fat cell differentiation, GO:1905243
response to T3) describe transcriptional regulation of PNPLA3 by nutritional state, not
processes PNPLA3 carries out; they are kept as non-core, except GO:0050872 white fat cell
differentiation, which is marked over-annotated (PNPLA3 is induced during adipogenesis, but
there is no evidence it drives it). UniProt records the same:
"By changes in energy balance: down-regulated following very low-calorie diet, whereas
refeeding elevates the mRNA level" (Q9NST1 INDUCTION). Demoted to non-core.
GO:0003841 EXP annotation citing PMID:21878620 a missing NOT qualifier, or a