TRAF3 (Q13114) — curation notes

Working journal for the TRAF3 GO annotation review. Provenance is recorded inline as
[PMID:xxxx "verbatim supporting text"]. Quotes are taken from the cached records in
publications/; where only an abstract is cached this is noted.

1. Identity and architecture

• UniProt Q13114, TRAF3_HUMAN, 568 aa, HGNC:12033. EC 2.3.2.27 (RING-type E3 ubiquitin
  transferase) with ECO:0000269|PubMed:25847972.
• Historical names reflect its discovery as a CD40-tail binder: CRAF1, CD40BP/CAP-1, and
  LAP1 (LMP1-associated protein 1).
• Domain layout (UniProt): N-terminal RING-type zinc finger, several TRAF-type zinc fingers,
  a coiled-coil (TRAF-N) region, and a C-terminal MATH/TRAF-C domain.
  DOMAIN: The MATH/TRAF domain binds to receptor cytoplasmic domains.
DOMAIN: The Ring-type zinc finger domain is required for its function in down-regulation of NFKB2 proteolytic processing.
• Two annotated isoforms: Q13114-1 (canonical) and Q13114-2 (VSP_040040). No isoform-specific
  GO annotations are present in GOA, and none of the literature reviewed here assigns a
  distinct function to isoform 2.

The two original cloning papers already establish the adaptor architecture:

• PMID:7527023 (abstract only)
• PMID:7530216 (abstract only)

Quaternary structure: TRAF3 homotrimerises and also forms heterotrimers with TRAF2/TRAF5.
PMID:15383523 (abstract only)

2. Two opposite-sign core roles

TRAF3 is unusual among TRAFs in being a negative regulator of one NF-κB branch and a
positive regulator of the type I interferon branch. Both are strongly supported.

2.1 Constitutive brake on non-canonical NF-κB (NIK/MAP3K14 turnover)

This is the best-established TRAF3 function and it is absent from the current GOA set.

• TRAF3 binds NIK and targets it for proteasomal degradation:
  PMID:15084608 (abstract only)
• The brake operates inside a TRAF2–TRAF3–cIAP1/2 assembly, with TRAF3 supplying NIK
  recruitment and TRAF2 supplying the cIAPs:
  PMID:18997794
  Note the same paper shows the genetic epistasis: the lethality of TRAF3 deficiency in mice could be rescued by a single NIK gene.
• The complementary study reaches the same architecture from the TRAF2 side
  (PMID:18997792, "Nonredundant and complementary functions of TRAF2 and TRAF3 in a
  ubiquitination cascade that activates NIK-dependent alternative NF-kappaB signaling").
• TRAF3 also restrains canonical NF-κB downstream of LTβR, and its loss de-represses
  the non-canonical pathway autonomously:
  PMID:20185819
• Loss-of-function of TRAF3 is a recurrent, selected event in multiple myeloma, exactly as
  expected for a constitutive brake on this pathway (PMID:17692805, "Promiscuous mutations
  activate the noncanonical NF-kappaB pathway in multiple myeloma").
• UniProt states the same: Also acts as a constitutive negative regulator of the alternative NF-kappa-B pathway, which controls B-cell survival and lymphoid organ development. and
  Promotes ubiquitination and proteasomal degradation of MAP3K14.

Curation consequence: propose GO:1901223 negative regulation of non-canonical NF-kappaB
signal transduction and GO:0043161 proteasome-mediated ubiquitin-dependent protein
catabolic process as NEW annotations. The existing GO:0030162 (regulation of proteolysis,
IMP from PMID:20185819) is the vague shadow of this.

2.2 Positive, non-redundant driver of type I interferon induction

• TRAF3 is required for IFN induction by several TLRs and by cytosolic virus sensing, and it
  bridges TLR adaptors to the IRF kinases:
  PMID:16306936 (abstract only)
• The division of labour with TRAF6 is explicit, and gives the TRIF/TBK1 mechanism:
  PMID:16306937 (abstract only)
• On the RLR arm, TRAF3 is recruited to mitochondrial MAVS through a TRAF-interacting motif,
  and this contact is required:
  PMID:16858409
  and PMID:21200404
• Human genetics confirms non-redundancy in vivo:
  PMID:20832341
  UniProt records this as Immunodeficiency 132A (MIM:614849), with a second, distinct
  dominant disorder IMD132B (MIM:621096).
  Note the mechanism is dominant-negative, not haploinsufficiency: the paper's own
  results section is headed "The TRAF3 mutant allele is dominant-negative", and patient
  cells retain PMID:20832341 -
  well below the ~50% a null allele would leave - because R118W destabilises the product of
  the wild-type allele through heterotrimer formation
  PMID:20832341.

2.3 The ubiquitin switch that separates the two outputs

The sign of TRAF3's output is set by which chain type it carries:

PMID:19898473

Upstream and downstream of that switch:

• cIAP1/2 place the degradative marks: PMID:20097753
• DUBA removes the activating K63 chains: PMID:17991829 Note this abstract also
  states directly: TRAF3 is an E3 ubiquitin ligase that preferentially assembled lysine-63-linked polyubiquitin chains.
• PTPN22 promotes the activating marks: PMID:23871208
• DDX3 scaffolds two successive K63 waves on TRAF3 downstream of MAVS:
  PMID:27980081
• Triad3A/RNF216 imposes the K48 off-switch: PMID:19893624
• NEDD4L and FBXO11 add further, non-degradative marks
  (PMID:33608556 K29-linked; PMID:36897010 K63-linked, NEDD8-dependent).
• ATP1B1 and TRIM35 are positive regulators acting through TRAF3 ubiquitination
  (PMID:34011520; PMID:32562145).

Most of this is TRAF3 as substrate, not TRAF3 as enzyme, and should not be mistaken for
TRAF3 molecular function during curation.

2.4 TRAF3 as enzyme: direct substrates

The clearest direct-substrate evidence is ASC:

PMID:25847972 (abstract only; this is the EC 2.3.2.27 evidence in UniProt)

Plus TRAF3 self-ubiquitination (PMID:19898473, above) and the NIK degradation route
(§2.1). Supports GO:0070534 protein K63-linked ubiquitination as a NEW annotation.

3. Receptors and localisation

• Receptor engagement is via the MATH/TRAF-C domain, to CD40 (the founding interaction,
  PMID:7527023, PMID:7530216), LTβR, BAFF-R, BCMA, OX40, CD30, EDAR, and the EBV oncoprotein
  LMP1. Fn14/TweakR also recruits it: PMID:11728344 (abstract only)
• LIGHT–LTβR death signalling is TRAF3-dependent: PMID:10799510 (abstract only; dominant-negative evidence, so an indirect,
  context-restricted claim on apoptosis).
• Endosomal pool on the TLR4/TRAM route: PMID:18222170 (abstract only)
• Mitochondrial pool: via MAVS (PMID:16858409, PMID:21200404) and co-localisation with TRIM35
  (PMID:32562145). UniProt: Mitochondrion {ECO:0000269|PubMed:32562145} and Co-localized to mitochondria with TRIM35.
• A K33-linked mark connects TRAF3 to vesicle trafficking:
  PMID:27438768 This is real but narrow (bladder
  epithelial cells, bacterial expulsion) — non-core.

4. Notes on specific GOA rows

Points where the review departs from GOA, with the reasoning:

• GO:0008063 "Toll signaling pathway" (IEA, InterPro2GO). The GO definition is the
  Drosophila Toll receptor pathway: "The series of molecular signals initiated by an
  extracellular ligand binding to the receptor Toll on the surface of a target cell". Human
  TRAF3 acts in TLR pathways; MODIFY to GO:0002224.
• GO:0060337 "type I interferon-mediated signaling pathway" (NAS, PMID:25847972). Defined
  as signalling initiated by type I interferon binding to its receptor. TRAF3 acts upstream,
  on IFN production; the correct term (GO:0032481) is already annotated. REMOVE.
• GO:0016020 "membrane" (IEA, ARBA). TRAF3 has no TM segment; it is a cytosolic protein
  transiently recruited to membrane-proximal complexes, which GO:0009898 and GO:0010008
  already capture precisely. REMOVE.
• GO:0005739 "mitochondrion" IDA anchored to PMID:19898473. The cached full text of
  Tseng et al. (73 kB, full_text_available: true) contains no occurrence of "mitochond".
  The term is nevertheless correct and independently supported (PMID:32562145 EXP;
  PMID:16858409 IC via MAVS), so the term is accepted and the citation discrepancy is recorded
  in that row's reason rather than by deleting an experimental annotation whose figures and
  supplement are not in the cache.
• GO:0004842 "ubiquitin-protein transferase activity" (2× Reactome TAS, 1× IDA). Parent of
  GO:0061630, which is annotated from the same evidence. MODIFY to GO:0061630.
• GO:0007166 "cell surface receptor signaling pathway" (IBA) is left as ACCEPT, not
  narrowed. The first draft of this review proposed GO:0023035 (CD40 signaling pathway) as
  a replacement. That is wrong for an IBA: the row's WITH set is
  FB:FBgn0265464|MGI:...|RGD:...|UniProtKB:O00463|UniProtKB:Q12933|UniProtKB:Q9Y4K3, i.e.
  it spans Drosophila and TRAF2/TRAF5/TRAF6, and CD40 is vertebrate-specific — so a CD40
  pathway term cannot be what is conserved across that PANTHER clade. The broad term is the
  phylogenetically honest one here. The CD40-specific proposal is instead attached to the
  human-experimental GO:0007165 TAS row anchored to PMID:7530216, where it belongs.
• GO:0031996 "thioesterase binding" / GO:0031625 "ubiquitin protein ligase binding"
  (IPI, PMID:11279055). That paper is a bioinformatics-driven survey in which the isolated
  TRAF domains of MUL/TRIM37 and USP7 bound all six TRAFs in vitro
  ["MUL and USP7 are capable of binding in vitro via their TDs to all of the previously
  identified TRAF family proteins (TRAF1, TRAF2, TRAF3, TRAF4, TRAF5, and TRAF6)"], i.e. a
  pan-TRAF, non-selective interaction. Kept, but non-core. (Independently, TRAF3 genuinely
  binds ubiquitin thioesterases OTUB1/OTUB2/OTUD5 (= DUBA), so the term itself is not wrong.)
• GO:1902554 "serine/threonine protein kinase complex" (NAS, PMID:24622840). From the
  STING–TRAF3–TBK1 complex. TRAF3 is a transient adaptor within a signalling complex that
  contains a kinase, not a stoichiometric subunit of a kinase complex. Marked over-annotated.
• GO:0006915 "apoptotic process" (TAS, PMID:10799510). Rests on a dominant-negative TRAF3
  blocking LIGHT-induced death; TRAF3 is upstream signalling, not part of the apoptotic
  execution programme. Marked over-annotated (the regulatory parent GO:0042981 is kept as
  non-core).
• GO:0005515 "protein binding" (25 IPI rows). Uninformative per project guidance; the
  informative content is preserved by GO:0005164, GO:0035591, GO:0019901, GO:0019903 and the
  proposed K63-ubiquitination terms.

5. Gaps

Added to the review as NEW annotations (terms absent from GOA entirely):

Recommended as proposed_replacement_terms on MODIFY rows (the term is already
represented in GOA, but by a parent that is too general or unsigned):

Proposed as a new ontology term (nothing suitable exists): a
lymphotoxin beta receptor signaling pathway term, as a sibling of GO:0023035 under
GO:0007166. Checked against OLS — GO has lymphotoxin terms only for the ligand and its
production (GO:0032641, GO:0032681, GO:0062048), none for LTBR-initiated signalling.

6. Open questions

• Is the "negative regulator of canonical NF-κB / MAPK" role a direct TRAF3 activity or
  purely a consequence of NIK-level and cIAP-level effects? The LTβR data (PMID:20185819)
  argue for a receptor-complex-composition mechanism (TRAF3 excludes TRAF2/IKK1), which would
  be a genuine adaptor function. This is why GO:0043122 is kept as the unsigned parent rather
  than being narrowed to GO:0043124 (negative regulation): signing the term would assert a
  mechanism the current evidence does not distinguish. Suggested experiment 1 in the review
  is designed to settle it.
• Isoform 2 (Q13114-2) has no functional data at all in the reviewed literature.
• Which substrates besides ASC does TRAF3 ubiquitinate directly? Most published "TRAF3
  ubiquitination" work is TRAF3-as-substrate.