MPEG1/perforin-2 is a MACPF-superfamily pore former with a single, well-supported molecular
function (GO:0022829 wide pore channel activity, IDA) that is deployed in two very different
biological settings: killing of bacteria inside phagolysosomes, and rupture of antigen-containing
endocytic compartments so that antigen reaches the cytosol for cross-presentation. A 2026 EMBO J
paper (PMID:42618800) argues that the form of the protein that makes the pore differs between the
two settings, which raises the curation question of whether this is one molecular function used
twice, or two mechanistically distinct activities.
The cryo-EM/AFM study that underpins the GOA GO:0022829 IDA established both the prepore and the
pore, and the acid trigger:
The same paper is the origin of the "TM anchor orients the pore away from the host membrane" model
that PMID:42618800 now challenges:
Independent morphological evidence for pores on the bacterial surface:
UniProt (Q2M385) records the mechanism as an acid-triggered prepore-to-pore transition in which two
alpha-helical regions refold into transmembrane beta-hairpins (TMH1/TMH2) forming a giant
beta-barrel, citing PubMed:31537793.
The delivery route is vesicular, not secretory:
PMID:26402460
Discovered by genetic screening in cross-presenting DCs:
PMID:37347855
PMID:37347855
That 2023 paper proposed that maturation releases the pore-forming domain:
PMID:37347855
An independent 2026 study reproduces the cross-presentation requirement, though it attributes it to
different proximal steps (uptake and control of acidification rather than escape per se):
PMID:41951589
PMID:41951589
PMID:42618800 (Laub, Chatterjee & Kozik, EMBO J 2026 - same lab as PMID:37347855) revises the 2023
model of its own group:
Status of the evidence: this is abstract-only in the cache (full_text_available: false), it is a
single paper, and it has no published rebuttal. It does not dispute that MPEG1 forms pores, nor
that it is required for cross-presentation; both of those are independently replicated. What it
disputes is (a) the topological model in which the TM anchor merely aims the pore at bacteria, and
(b) the assumption - inherited from the acid-activation result of PMID:31537793 - that acidification
is required for pore formation in every setting.
Note that the two claims pull in the same direction and are mutually consistent: if the protein
stays anchored, its pore is necessarily made in the host compartment membrane, and a DC phagosome
that is deliberately kept non-acidic cannot supply the acid trigger, so a different trigger must
exist. UniProt already records the released-ectodomain model as an inference from the mouse ortholog
("cleaved by trypsin in proximity of the helical transmembrane domain releases the ectodomain into
the lysosomal lumen to orient the pore-forming domain toward the endogenous membranes",
ECO:0000250|UniProtKB:A1L314) and as a separate "processed form" chain; PMID:42618800 argues this is
not the species that does the work in cross-presentation.
One core molecular function, not two. GO:0022829 covers the act of building an oligomeric
transmembrane pore; that act is the same chemistry (MACPF prepore -> beta-barrel) in both settings.
What PMID:42618800 changes is which molecular species oligomerises (full-length anchored vs
released ectodomain), which membrane is perforated (host endocytic membrane vs engulfed bacterium)
and what triggers it (pH-independent vs acid-triggered). Those are differences in regulation,
topology and target - none of which GO's molecular-function branch distinguishes, and none of which
would be captured by splitting GO:0022829 into two entries. Splitting on a single, abstract-only
paper would also assert more confidence than the evidence carries. The two deployments are instead
separated in directly_involved_in (antibacterial defence terms vs the cross-presentation terms)
and the divergence is recorded in the reason fields of GO:0022829 and GO:0061474 and in
suggested_questions.
No annotation is removed on the strength of PMID:42618800. GO:0061474 phagolysosome membrane
(is_active_in, IDA) is if anything strengthened by it: a pore made by anchored full-length
protein is unambiguously active in the compartment membrane. GO:0005576 extracellular region is
kept as non-core - it reflects the secreted, TM-less splice isoform perforin-2b
PMID:28705375 and PMID:28705375 - not the pore-forming species.
Ontology observation (not acted on). GO:0022829 is defined as "the energy-independent
facilitated diffusion of propanediol through a large pore, un-gated channel. Examples include gap
junctions ... and porins", which is a poor fit for an acid-gated lytic MACPF beta-barrel; GO:0140911
pore-forming activity ("a protein is inserted into the membrane of another cell where it forms
transmembrane pores") fits the antibacterial deployment far better. GO:0022829 is nonetheless the
term GO uses consistently across this protein class (PRF1, GSDMD and MPEG1 all carry it), so it is
accepted here rather than replaced for MPEG1 alone; adding GO:0140911 - as PRF1 already has via
InterPro - is raised as a recommendation instead. Neither term covers pore formation in the cell's
own endomembrane, which is what the cross-presentation deployment does.
GO:0055085 transmembrane transport is an automatic inter-ontology inference from GO:0022829GO:0002478 is the parent of GO:0042590; both are carried by ISS from the mouse ortholog. Kept,