NINJ1 review notes

Reviewer journal. Every assertion below is anchored to a verbatim quote from a cached
publication in publications/.

Why this gene is in the contested-function batch

Two things are live for NINJ1 in 2025-2026:

How it ruptures the plasma membrane - the "membrane-disk cutting" model versus the
"large pore" model. This peaked in 2023-24 and 2025-26 is mostly consolidation.
Whether membrane rupture is all it does. A 2026 paper shows PMR-deficient Ninj1
mutants retain a haematopoietic function, i.e. a separable second role. That bears
directly on the curation question of whether the 1990s cell-adhesion characterisation is
core biology or history.

The core function: plasma membrane rupture

Founding forward-genetic identification, and the key phenotype - cells still die, but they
do not disintegrate:
PMID:33472215
and PMID:33472215.

Mechanism from the same paper: PMID:33472215.

Oligomerisation is the activation step, not a by-product. An antibody that blocks it blocks
rupture: PMID:37196676.
And glycine, the classical 30-year-old cytoprotectant, turns out to act at this step:
PMID:36468682.

The mechanistic dispute: disks vs pores

Degen et al. resolved the filament and the amphipathic-helix insertion mechanism:
PMID:37198476 and
PMID:37198476.

The Wu lab reading is that the rings excise and shed membrane disks - explicitly not
a pore mechanism:
PMID:38614101 and
PMID:38614101.

The Dai lab agrees on solubilisation but keeps pores on the table, and explains why the
paralog NINJ2 cannot do this at all:
PMID:39667936.

Curation consequence: the disagreement is about how the activity is executed, not about
whether it exists. Every model is an instance of GO:0140912 membrane destabilizing activity
("Binding to a membrane and increasing its permeability"), so the dispute does not change the
action on any annotation. It is recorded in the reason and raised in suggested_questions.

The 2026 Genentech perspective (PMID:42350666, Surveying the roles of NINJ1 in plasma
membrane rupture, Deshpande I & Kayagaki N, Nat Struct Mol Biol 2026,
doi 10.1038/s41594-026-01834-3) was fetched but PubMed indexes no abstract for it and it
is not in PMC, so its content could not be retrieved. It is cited in references with
full_text_unavailable: true and correctness: UNVERIFIED, and no claim in this review
rests on it.

NINJ1 is not pyroptosis-specific

The single activity is used by many death programs. Ferroptosis, and crucially only at the
terminal step: PMID:38396301 plus
PMID:38396301.

And in 2026, with no known death program engaged at all:
PMID:42270656 and
PMID:42270656.

So GO:0141201 pyroptotic cell death, GO:0097300 programmed necrotic cell death and
GO:0097707 ferroptosis are all ACCEPTed, but each reason says explicitly that they are
contexts of one shared terminal effector, not evidence of pathway specificity.

The adhesion question: core or historical?

The 1996-97 characterisation is real experimental work, not a naming accident:
PMID:9261151 and
PMID:9261151.

The motif is still a live handle - a peptide from it binds endogenous NINJ1:
PMID:33028854.

And the adhesion activity has a clean cell-biological readout in humans:
PMID:22162058.

The decisive 2026 argument against calling it purely historical is that NINJ1 demonstrably
does non-lytic work:
PMID:42481508.

But be honest about the gap. That paper attributes the HSPC function to WNT, not to
adhesion: PMID:42481508. It does not test the adhesion motif. So the
existence of a PMR-independent function is established; its identity with the adhesion
activity is not.

Position taken: both GO:0140912 membrane destabilizing activity and GO:0098632 cell-cell adhesion mediator activity are kept as core functions, with the unbridged gap
stated in the reason and in suggested_questions rather than papered over. There is also a
structural tension worth flagging: the same extracellular α1/α2 helices that mediate rupture
are adjacent to the adhesion determinant, so the two activities may be two uses of one
surface. GO:0098631 (the parent, cell adhesion mediator activity) is MODIFYed to the
child, since what was shown is homophilic binding between cells, not adhesion to a substrate.

Other calls

GO:0001530 lipopolysaccharide binding (IDA, PMID:26677008) → KEEP_AS_NON_CORE, not
removed. The result is genuine - PMID:26677008 - but it is one lab, in transfected
HEK293T lysates, never reproduced, with no counterpart in any of the NINJ1 cryo-EM
structures, and the implicated residues (81-100) lie outside the α1/α2 region that carries
the characterised activity. The matching GO:0034145 TLR4 ISS is demoted for the same
reason, so the two stay consistent.
GO:0005576 extracellular region (IEA) → KEEP_AS_NON_CORE, not over-annotated. I
initially read this as a topology artefact, but UniProt annotates a genuine separate
"Secreted ninjurin-1" chain generated by cleavage. Checking the UniProt record before
calling it an error was the difference.
GO:0016020 membrane (IEA/InterPro) → MODIFY to GO:0005886 plasma membrane.
GO:0005515 protein binding (HuRI + Rolland interactome) → MARK_AS_OVER_ANNOTATED.
GO:0045766 positive regulation of angiogenesis → KEEP_AS_NON_CORE, with a flag: the
pro-angiogenic effect was produced by suppressing endogenous NINJ1 with the N-NAM peptide
(PMID:33028854), so the positive direction is not the most
transparent reading. Not corrected, because this is an experimental annotation whose full
text the curator read and the rat orthology transfer agrees with it; the doubt is recorded
instead.
The mouse-ISS cluster (GO:0097060 synaptic membrane, GO:0042692 muscle cell
differentiation, GO:0071474 hyperosmotic response, GO:0034113 heterotypic adhesion) →
KEEP_AS_NON_CORE. UniProt records several of these as "By similarity" only.

Proposed new term

GO has no term for plasma membrane rupture - only GO:0019835 cytolysis, which conflates
the active terminal step with lysis in general. The NINJ1 literature established exactly this
distinction and it is experimentally separable (gasdermin pores form, but rupture does not
follow, in Ninj1-null cells). Proposed with GO:0019835 as parent.