K3 / MIR1 (KSHV, HHV-8 type P isolate GK18, UniProt P90495) — curation notes

Reviewer: claude-code (CACAO audit batch 2). Date: 2026-10-29.

Deep research was unavailable for this gene. No deep-research API keys are
provisioned in this container, so just deep-research was not run and no
K3-deep-research-<provider>.md file exists. The literature below was assembled
by hand from the cached publications in publications/, from papers fetched for
this review with just fetch-pmid, and from the curated UniProt record
K3-uniprot.txt. The GONUTS wiki (gowiki.tamu.edu), which would have recorded
the CACAO students' own evidence statements, was offline at the time of
review ("This site is temporarily offline while we perform upgrades"), so the
students' reasoning could not be inspected directly; each CACAO row below was
therefore judged against the cited paper itself.

What the protein is

K3 (ORF K3, also IE1B, MIR1 — modulator of immune recognition 1) is a 333-residue
KSHV membrane protein with an N-terminal cytosolic RING-CH (RINGv) zinc finger,
two transmembrane helices, and a long cytosolic C-terminal tail. It and its ~40%
identical paralog K5/MIR2 are the prototypes of the MARCH family of membrane-bound
RING-CH E3 ubiquitin ligases PMID:23460925.

The N-terminal domain was first called a plant homeodomain but the NMR structure
(PDB 1VYX) showed it is a variant RING: PMID:15465811 member of the RING family,
with the canonical PMID:15465811
arrangement, and it is the E2-docking surface — the domain was PMID:15465811 and UbcH13.
UniProt records eight Zn-coordinating residues (Cys9, Cys12, Cys24, His26, Cys34,
Cys37, Cys50, Cys53) in the RING-CH-type zinc finger (residues 1-60).

Catalytic activity and mechanism

K3 is a bona fide E3: PMID:11756476, and the activity requires the
zinc finger — PMID:11756476 whose disruption ablates downregulation.
Unusually, MIR1/K3 can ubiquitinate non-lysine residues:
PMID:15994556 down-regulation of
MHC I lacking cytoplasmic lysines, and PMID:15994556. UniProt assigns
EC 2.3.2.36 on this basis (cysteine-to-cysteine ubiquitin transfer).

Substrates and the immune-evasion phenotype

MHC class I (the core target). K3 strips all four tested HLA allotypes from
the surface — PMID:10799607,
PMID:10799607 — and does so by
accelerating endocytosis rather than by ER retention:
PMID:10799607;
PMID:10859362. The ubiquitin mark is the internalization and sorting
signal: K3 PMID:12006494 and this provides the PMID:12006494, after which class I enters the
TSG101-dependent late endocytic pathway.

IFN-gamma receptor 1 (type II interferon, not type I). PMID:17166914, with the
functional consequence that PMID:17166914. The documented
interferon target of K3 is therefore the type II (IFN-gamma) pathway; no paper
found here shows K3 blocking type I IFN (IFNAR) signalling. This matters for the
GO:0039502 row, which the UniProt keyword mapping places on type I.

CD1d. UniProt cites PubMed:15864354 for CD1D as a K3/K5 target.

DC-SIGN (CD209) and DC-SIGNR (CLEC4M) — the CACAO paper, see below.

Timing in the real infection. K3 is not redundant with K5 but sequential:
PMID:24899205.

Localization

Two curated locations, both from PubMed:10799607 per UniProt. Ishido et al. put the
bulk near the plasma membrane — PMID:10799607 — while Coscoy
and Ganem, using a KDEL-GFP marker and calnexin, put the steady-state bulk in the ER:
PMID:10859362. These are not contradictory: K3 is made and resides largely in the
ER/pre-Golgi, and acts on its targets at or near the plasma membrane. The 2013 paper
adds that K3 engages immature, EndoH-sensitive lectin, so that
PMID:23460925.

The CACAO reference: PMID:23460925 (Lang et al. 2013, PLoS ONE) — full text available

All four CACAO rows on this gene cite this one paper. What it actually shows for
K3 (as distinct from K5):

• Fig. 2 (flow cytometry, 293 DC-SIGN / DC-SIGNR stable lines + GFP-K3):
  PMID:23460925, and
  PMID:23460925. The same
  figure records K3's effect on endogenous MHC I:
  PMID:23460925
• Fig. 3: the tyrosine-based endocytosis motif, in addition to the RING-CH
  domain, is needed; notably K3 Y/A still downregulates MHC I.
• Fig. 4 (dynasore washout, THP-1 and 293):
  PMID:23460925, but no endocytosis of DC-SIGNR was seen.
• Fig. 5A (western blot): PMID:23460925, RING-CH
  dependent.
• Fig. 6 (GST pull-down): PMID:23460925,
  preferentially the immature EndoH-sensitive form.
• Fig. 7A/B (HA-Ub pull-down, MG132/chloroquine):
  PMID:23460925 and
  PMID:23460925, with a lysosomal
  fraction.

Paralog hazard, handled. The paper's title names both proteins, and K5 dominates
the infection experiments: in the recombinant-virus arm
PMID:23460925,
PMID:23460925. But every figure
from Fig. 2 onward assays K3 in its own right, with its own wild-type, mZn, Y/A,
Y/F, P/A and DE12 constructs, so the K3 annotations from this paper are not
mis-attributed paralog data. The one caveat is scope: K3's activity on these lectins
is shown in exogenous expression, not in the context of infection.

Viral entry — what the paper does and does not show. DC-SIGN and DC-SIGNR are
entry-enhancing co-receptors for KSHV: PMID:23460925. That is a property of the host lectins,
and the paper's Fig. 1 measures it in cells expressing no viral MARCH protein at all
(Bac16 ΔK3ΔK5). K3 is on the opposite side of that transaction: it removes the
entry co-receptor. The CACAO row GO:0075509 "endocytosis involved in viral entry into
host cell" therefore reads the wrong actor into the term — it is the virion that is
endocytosed in that term's definition ("Any endocytosis that is involved in the uptake
of a virus into a host cell"), not a host receptor. K3 is, in addition, expressed
after entry (immediate-early / late lytic), so it cannot participate in uptake of
the virion that delivered it.

Intraviral interactions (PMID:16339411)

The two IPI protein binding rows come from the Uetz/Haas herpesviral yeast
two-hybrid interactome. The partners are KSHV (HHV-8 type M) proteins: P88941, an
uncharacterized 110-aa product, and P88951, the ribonucleoside-diphosphate reductase
small subunit (RIR2). The abstract is all that is cached
(full_text_available: false) and reports only the network-level result
PMID:16339411; there is no functional follow-up
for either pair, and neither partner is a known K3 ubiquitination substrate. Bare
GO:0005515 is uninformative here and no better molecular function can be inferred
from the screen, so the protein-binding policy applies (REMOVE as uninformative, which
does not assert that the Y2H interactions are false).

Summary of review decisions

CACAO failure-taxonomy note

One new-ish shape and one known one:

1. Wrong actor in a two-sided interspecies term (GO:0075509). The paper contains
   both "K3 drives endocytosis of a host receptor" and "the host receptor enhances
   virion entry". The CACAO term fuses them into "K3 is involved in endocytosis for
   viral entry", which asserts K3 does something the paper attributes to the lectin,
   and in the opposite direction (K3 destroys the co-receptor). This is adjacent to
   the catalogued "own-process term for a host-directed activity" shape, but inverted:
   here the host's own process is attributed to the symbiont. Call it actor
   substitution across the host/symbiont boundary.
2. Right pathway, wrong granularity (GO:0039504). The term is true of K3 and the
   paper does assay the relevant readout (MHC I surface levels, Fig. 2/3), but the
   paper and the wider literature support the specific child
   GO:0046776 (suppression of host antigen processing and presentation via MHC class I).

Two rows (GO:0006511 IDA, GO:0039648 IMP) are clean: right protein, right assay,
right term, correct evidence code. Across the four CACAO rows, two are accepted as
written and two need a term change — in both cases the student read the right paper
and ran the right figures, and the failure is entirely in term choice. That matches the
batch-1 conclusion recorded in projects/CACAO.md.