TNFRSF21 (DR6) review notes

Why this gene was selected

DR6 is an orphan TNF-receptor-superfamily member whose most famous claimed function - being the
receptor for an N-terminal fragment of APP, and thereby driving developmental axon pruning and
neuron death - has come apart in two separate steps: the founding paper was retracted in 2024, and
a 2026 eLife study failed to reproduce a separate, later claim that DR6 drives Wallerian
degeneration. The curation question is how much of GOA is actually exposed.

What GOA actually carries (checked before judging)

TNFRSF21-goa.tsv was read column by column. Findings that constrain the whole review:

• The only molecular function on the gene is bare GO:0005515 protein binding, four IntAct
  rows: PMID:23559013 with UniProtKB:P08138 (NGFR/p75NTR), PMID:32296183 with
  UniProtKB:O43765 (SGTA), and PMID:35922511 with UniProtKB:P05067 (APP) and
  UniProtKB:Q6UXB8 (PI16). So an APP-binding annotation does exist, but only as
  uninformative protein binding; there is no "APP receptor", "amyloid-beta precursor protein
  binding" or death-receptor MF anywhere on the gene.
• There is no NOT qualifier anywhere in the file - every row is a positive assertion.
• There is no axon-degeneration or axon-pruning term on the gene at all. GO has no
  axon degeneration or axon pruning term (the nearest is GO:0016322 neuron remodeling, which
  TNFRSF21 does not carry). The closest thing in GOA is GO:0007413 axonal fasciculation, and
  GO:0051402 neuron apoptotic process.
• No GO annotation on human or mouse TNFRSF21 derives from the retracted paper. Checking the
  mouse ortholog (Q9EPU5) in QuickGO, which is the ISS/IEA donor for most of the human BP set,
  the experimental sources are PMID:11485735 (T cells), PMID:12515813 (B cells),
  PMID:21725297 (oligodendrocytes/myelination) and PMID:23559013 (neuron apoptotic process).
  PMID:19225519 (Nikolaev et al.) appears nowhere.

That last point is the single most important finding of this review: the exposure to the retraction
is essentially zero, and the exposure to the Wallerian refutation is limited to one propagated IEA.

The retraction

The founding APP/DR6 paper is retracted:
PMID:38110576, for image duplication and statistical
errors plus superseded conclusions
PMID:38110576 and
PMID:38110576.

The retraction is partial in substance. The authors explicitly maintain the core genetic claim:
PMID:38110576. That maintenance rests on a separate, non-retracted paper
which repeated the in vivo genetics:
PMID:24806670 and PMID:24806670, while correcting the mechanism
PMID:24806670.

So: developmental retinal axon pruning in vivo survives the retraction via PMID:24806670; the
beta-secretase/N-APP mechanistic model does not.

The Wallerian refutation

The 2017 claim being tested:
PMID:28285993
and PMID:28285993

The 2026 non-replication:
PMID:41891813
PMID:41891813
PMID:41891813
PMID:41891813

This is a strong non-replication: two independent knockout lines, one of them the original line,
both in vivo and in the in vitro assay format the original used. It is nonetheless a single
laboratory's non-replication with no response yet from the original authors.

What the refutation does and does not cover

Covered: injury-induced Wallerian degeneration of peripheral axons after axotomy, Schwann cell
injury responses, and the associated in vitro axotomy assay. Also, by the authors' own framing,
the therapeutic rationale that runs through WD:
PMID:41891813

Not covered:
- Developmental axon pruning in the CNS (retinocollicular pruning in PMID:24806670). Different
process, different context, different assay. Wallerian degeneration is injury-triggered
disintegration of a severed distal axon; developmental pruning is a trophic-cue-driven,
receptor-mediated removal of a branch on an intact neuron. A null result in one says nothing
about the other, and PMID:24806670 has not been independently re-tested.
- Trophic-deprivation-induced degeneration, which the original 2009 work also assayed.
- Aβ-induced neuron death, an independent line: PMID:23559013 and PMID:23559013 This is the
actual evidential basis for the GO:0051402 neuron apoptotic process annotation.
- The immune and oligodendrocyte roles, which are entirely separate literatures (below).

The well-supported functions, which are unaffected

T cells: PMID:11485735 and PMID:11485735, with Th2 skewing
PMID:11485735

B cells: PMID:12515813 and PMID:12515813

Oligodendrocytes: PMID:21725297, PMID:21725297 and
PMID:21725297

Localisation: PMID:19654028

Curation position taken

Nothing is removed. The refutation and the retraction land on claims that GOA never encoded.

• All plasma-membrane and immune-regulatory terms: ACCEPT. Grounded in two independent mouse
  knockout papers that the recent disputes do not touch.
• GO:0051402 neuron apoptotic process (IBA + IEA): ACCEPT, with the provenance spelled out in
  reason. Its mouse source is PMID:23559013 (Aβ/p75NTR), not the retracted paper, and the eLife
  study addressed axon degeneration, not neuronal apoptosis. Kept at ACCEPT rather than demoted
  partly because demoting an IBA would require a structured propagation_review that this review
  cannot honestly supply - the PAINT tree was not inspected - and partly because the annotation is
  in fact experimentally grounded in the ortholog.
• GO:0097252, GO:0048713, GO:0031642: ACCEPT. Mi et al. 2011 is independent of both
  disputes.
• GO:0042552 myelination: MODIFY to GO:0031642 negative regulation of myelination. DR6 is a
  negative regulator; the unqualified parent misstates the direction of the effect, and the gene
  already carries the correct child.
• GO:0032693 / GO:0032696 / GO:0032714 (negative regulation of IL-10 / IL-13 / IL-5
  production): KEEP_AS_NON_CORE. These are readouts of the Th2 skewing seen in DR6-null CD4 T
  cells, i.e. downstream consequences of the T-cell regulatory role, not separate functions.
• GO:0050852 T cell receptor signaling pathway: KEEP_AS_NON_CORE. DR6 restrains the
  proliferative outcome of TCR stimulation; it is not a component of the TCR signalling pathway.
• GO:0006915 apoptotic process (IMP, PMID:22761420): KEEP_AS_NON_CORE. Real but derived from
  ectopic overexpression PMID:22761420, with a Bax requirement PMID:22761420; overexpression of a
  death-domain receptor causing apoptosis is weak evidence for a physiological role, and the term is
  maximally general.
• GO:0007413 axonal fasciculation (IEA, Ensembl transfer from rat D3ZF92):
  MARK_AS_OVER_ANNOTATED. The rat source is an IMP from PMID:25898930, a study of prion-peptide
  (PrP106-126)-induced axonal degeneration in cultured rat spinal neurons. GO:0007413 is
  defined as "The collection of axons into a bundle of rods, known as a fascicle", which that paper
  does not address; and the underlying phenomenon - DR6-dependent axon degeneration - is precisely
  what PMID:41891813 failed to reproduce in mouse. The rat experimental annotation is not
  overruled here (the curator saw the full text and this review has not); the human IEA that
  propagates it is flagged.
• GO:0071356 cellular response to tumor necrosis factor (IDA, PMID:19654028): UNDECIDED. That
  paper is a study of N-/O-glycosylation, palmitoylation and lipid-raft targeting of DR6
  PMID:19654028; the cached record is abstract-only and describes no TNF-stimulation
  experiment. DR6 has no known TNF-family ligand and is routinely described as orphan
  PMID:41891813. Flagged rather than removed, per the rule against overruling an experimental
  annotation from an abstract.
• GO:0005515 protein binding x3: MARK_AS_OVER_ANNOTATED per project guidance. Noted in the
  reason that one of these rows is the APP interaction, redetected independently of the retracted
  work in a systematic extracellular-interactome screen (PMID:35922511), and that a specific
  APP-binding MF term would be more informative than bare protein binding.

Loose ends and honest uncertainty

• Whether DR6 is genuinely an APP receptor is not settled here. The physical interaction has
  independent support (the IntAct row from PMID:35922511) and the retraction notice affirms it, but
  the affirmation comes from the retracting authors, and the structural model of the interaction was
  one of the things retracted.
• Developmental CNS pruning rests on PMID:24806670, which has not been independently replicated and
  shares authorship with the retracted paper. It is not refuted; it is untested by others.
• GO:0002250 adaptive immune response and GO:0006959 humoral immune response are accepted but are
  grouping-level terms; the informative statements are the negative-regulation children.

One annotation added

GO:0004888 transmembrane signaling receptor activity is proposed as a NEW annotation (ISS from
the mouse ortholog). Without it the gene has no molecular function at all once bare GO:0005515 is
set aside as uninformative. GO:0005035 death receptor activity was considered and rejected: it
requires combining with an extracellular death ligand, and DR6 has none identified. This is also the
molecular function carried in core_functions.