APOO / MIC26 (Q9BUR5) — review notes

Human APOO, HGNC:28727, X-linked (Xp22.11), UniProt Q9BUR5 = MIC26_HUMAN, 198 aa,
PANTHER PTHR14564 ("MICOS COMPLEX SUBUNIT MIC26 / MIC27 FAMILY MEMBER"), Pfam PF09769
(ApoO), InterPro IPR019166 / IPR033182. Accession verified: the fetched record's ID line is
ID MIC26_HUMAN Reviewed; 198 AA. — the expected protein, not a merged
accession pointing elsewhere.

Two names, one protein, two incompatible identities in the literature: the plasma
apolipoprotein "ApoO" of 2006, and the MICOS subunit MIC26 of 2013 onwards. Almost every
curation question on this gene turns on which of the two the annotation rests on.

1. The two identities, and how the field resolved them

1.1 The apolipoprotein identity (2006)

APOO was named from a dog obesity cardiac-transcriptome screen and characterised as a secreted
chondroitin-sulfate proteoglycan found in lipoprotein fractions
PMID:16956892,
glycosylated
PMID:16956892,
and functional in cholesterol handling
PMID:16956892.
Everything in that paper about secretion, glycosylation and lipoprotein association was read
off a 55 kDa immunoreactive band, against a recombinant protein of 22 kDa.

1.2 The MICOS identity (2013-2015)

Complexome profiling of mammalian mitochondria placed APOO and APOOL in the Mitofilin/MINOS
(now MICOS) complex PMID:23704930, and three 2015 papers established MIC26 as a bona fide
subunit:

• Koob et al. PMID:25764979,
  with the functional consequence
  PMID:25764979.
• Ott et al., who renamed it Mic23 and showed import into the >1 MDa MIB complex
  PMID:25781180.
• Guarani et al., in whose MICOS proteomics MIC26 is degraded when the complex falls apart
  PMID:25997101.

Koob et al. tried to reconcile the two identities by proposing three coexisting forms
PMID:25764979.
That three-form model is the direct source of the Secreted, Golgi apparatus membrane and
Endoplasmic reticulum membrane lines still in the UniProt record, and hence of six GOA rows:
three IDAs cited to this paper and the three GO_REF:0000044 IEAs generated from those UniProt
lines.

1.3 The three-form model was retracted in substance in 2023

Lubeck et al. — from the same laboratory, with Anand and Reichert co-authors on both the
2015 and the 2023 paper — went back with the controls the earlier work lacked:

PMID:37279200

The 55 kDa band survives knockdown with three siRNAs, survives CRISPR knockout in four human
cell lines (HEK293, HepG2, HeLa, HAP1), is not produced by GFP- or myc-tagged MIC26, is
unaffected by mutating the predicted glycosylation sites
PMID:37279200,
and contains no MIC26 peptides
PMID:37279200.
Their conclusion:

PMID:37279200

They also note the negative external check that had been available all along
PMID:37279200.

The tell was already visible in the 2015 companion data paper, unremarked
PMID:26217777
— an antibody signal that does not go down when the gene is knocked down is, in hindsight, the
definition of non-specific.

Curation consequence. Eight GOA rows assert a secretory/ER/Golgi identity: GO:0005576 x4,
GO:0000139 x2 and GO:0005789 x2 — the six above plus the GO:0005576 EXP on PMID:16956892
and the GO:0005576 ISS from pig. All eight trace to the 55 kDa band and all eight are marked
REMOVE (the ninth REMOVE in this review is the unrelated SAM-complex row, §4). This is not second-guessing a curator from an
abstract: it is following the primary literature to the paper, by the original authors, that
withdrew the observation using knockouts.

1.4 Sequence-level corroboration (this review's bioinformatics)

UniProt's SIGNAL 1..25 on Q9BUR5 carries ECO:0000255 — prediction only. The paralogue
Q6UXV4/MIC27, same family, same complex, same membrane, is annotated TRANSIT 1..27 instead.
Scored against panels of ten human TRANSIT-annotated and ten human SIGNAL-annotated proteins
(the latter weighted towards plasma apolipoproteins), the MIC26 1-25 segment behaves like a
presequence, not a signal peptide
[file:human/APOO/APOO-bioinformatics/RESULTS.md "On 4 of the 5 metrics the MIC26 1–25 segment sits closer to the mitochondrial-presequence panel than to the secretory-signal-peptide panel."].
The decisive feature is the missing hydrophobic core
[file:human/APOO/APOO-bioinformatics/RESULTS.md "Every one of the ten signal peptides in the panel reaches a 7-residue"]
— MIC26 reaches only +1.70 on that scale, below the entire secretory panel, while its
hydrophobic moment (0.602) is at the mitochondrial panel mean (0.61) and 5.8 SD from the
secretory mean.

Honest caveat recorded in RESULTS.md: MIC26 has only one Arg in the segment, unlike classic
Arg-rich presequences. But MIC27 has the same Arg deficit and UniProt still calls its
N-terminus a transit peptide, so within this family the Arg metric does not separate the
classes. Neither protein has Arg at -2/-3.

This is corroboration, not proof — the knockout/MS evidence in PMID:37279200 is what settles
it. What the sequence analysis adds is a mechanism for the historical error: a
sequence-analysis SIGNAL call on an N-terminus that is really a mitochondrial targeting
sequence, on a protein whose amphipathic helices made "apolipoprotein" a plausible name.

2. What MIC26 actually does

2.1 Position in the complex

MICOS has seven mammalian subunits in two subcomplexes bridged by MIC13: MIC60-MIC19-MIC25 and
MIC10-MIC26-MIC27. MIC26 sits in the MIC10 subcomplex. The MICOS complex connects through
MIC60 to SAM50 in the outer membrane to form the ~2 MDa MIB complex, and complexome profiling
resolves the assembly ladder
PMID:26477565.
MIC26 and MIC27 are the family's late additions in evolution
PMID:26477565,
which is exactly what PANTHER encodes: PTHR14564 is named "MICOS COMPLEX SUBUNIT MIC26 /
MIC27 FAMILY MEMBER" and every one of the nine UniProt-reviewed members in the committed
PTHR14564-entries.csv is a MIC26 or MIC27 orthologue.

2.2 Crista junction formation — core, but partially redundant with MIC27

Three human datasets, with a real tension between them:

• Koob 2015, siRNA in HeLa: significant reduction in CJ number [PMID:25764979, quoted above].
• Ott 2015, shRNA in HEK293: no effect
  PMID:25781180.
• Anand 2020, CRISPR knockout in HAP1 — the cleanest design, single and double knockouts side
  by side: PMID:32788226
  with the double knockout worse than either single.

The knockout data break the tie in favour of a real but redundant contribution: MIC26 alone is
dispensable for complex integrity
PMID:32788226
yet its loss still measurably costs crista junctions. Human genetics agrees: a missense
I117T
PMID:32439808
and a C-terminal nonsense E178*
PMID:37649161
both cause mitochondrial disease through cristae disruption.

So GO:0042407 cristae formation is core for MIC26 specifically, not merely inherited from the
complex — the single-gene knockout phenotype is the evidence that distinguishes the two.

2.3 Cardiolipin: regulate, yes; bind, not demonstrated for MIC26

This is the easiest place on this gene to over-annotate, because the paralogue really does bind
cardiolipin and the two are routinely discussed together.

• MIC27/APOOL binds cardiolipin in vitro PMID:23704930.
  That is APOOL, a different gene.
• Anand 2020 restates the attribution unambiguously
  PMID:32788226.
• What MIC26 does do is set cardiolipin levels, and it does so in its own single knockout
  PMID:32788226,
  with cardiolipin causally upstream of the OXPHOS phenotype
  PMID:32788226.
  Koob 2015 independently linked MIC26 levels to tafazzin.
• A 2026 simulation study predicts that Mic26 (human sequence: they model residues up to 198)
  recruits cardiolipin at a positive loop
  PMID:42647630,
  but that is coarse-grained molecular dynamics on an AlphaFold model
  PMID:42647630,
  not a binding assay.

No GO:1901612 cardiolipin binding annotation is proposed for APOO. One NEW row is
proposed: GO:1900208 regulation of cardiolipin metabolic process, IMP, on the human
single-knockout lipidomics.

2.4 MIC10 oligomer regulation — yeast, so far

PMID:29733859
That is Saccharomyces cerevisiae Mic26. The human counterpart is the reciprocal protein-level
regulation seen by Koob and Anand (knocking out one raises the other, post-transcriptionally),
which is not the same claim. Not annotated on human APOO.

2.5 Mitochondrial morphology — MIC26-specific, mechanism unknown

MIC26 knockout fragments mitochondria in human HAP1 cells and MIC27 knockout does not
PMID:32788226,
with fusion machinery unaffected. The authors decline to interpret it
PMID:32788226.
Mouse macrophage knockout reaches the same phenotype by a different route
PMID:37995600.
Recorded as a knowledge gap and a suggested experiment rather than a GO annotation: it is a
reproducible steady-state morphology phenotype whose mechanistic relation to MIC26's structural
role is exactly what the primary authors say is unknown.

2.6 Physiology (non-core)

Transgenic human APOO in mouse heart promotes lipotoxic cardiomyopathy PMID:24743151;
adipocyte-specific knockout impairs BAT thermogenesis
PMID:37088120;
macrophage-specific knockout increases efferocytosis PMID:37995600. These are mouse,
tissue-specific, downstream of the cristae/OXPHOS defect, and none is annotated in the current
GOA set. They are not proposed as new human annotations.

3. The two IBA rows

Family PAINT slice fetched to interpro/panther/PTHR14564/PTHR14564-paint.tsv. The family has
one PAINT node carrying two IBD annotations — exactly the two IBAs on APOO:

Donors resolved:

• FB:FBgn0038400 -> UniProt Q9VEY5, Drosophila melanogaster Mic26-27 (CG5903,
  dApoO; TrEMBL, unreviewed). The fly has a single pre-duplication gene covering both paralogues.
  QuickGO shows Q9VEY5 carries GO:0061617 IMP from PMID:32439808 and IGI from
  PMID:38253420 — experimental grounding on the fly side.
• UniProtKB:Q6UXV4 = human APOOL/MIC27. QuickGO shows two independent IDA
  GO:0061617 annotations (PMID:25781180, PMID:25997101).
• UniProtKB:Q9BUR5 = APOO itself. The target appearing in its own WITH/FROM is expected:
  its four direct IDA GO:0061617 rows and its IMP GO:0042407 row are among the descendant
  evidences the PAINT curator weighed when placing the IBD. Not circular, and not to be marked
  CIRCULAR_OR_REDUNDANT.

Node placement: all nine UniProt-reviewed members in PTHR14564-entries.csv are MIC26 or
MIC27 orthologues (human/mouse/bovine MIC26; human/mouse/bovine/orangutan/chicken MIC27;
C. elegans moma-1). Both human paralogues and the single fly gene receive IBA from
PTN001803267 (checked in QuickGO), so the node sits at or near the family root, before the
MIC26/MIC27 duplication.
APOO is inside the inheriting clade with its own experimental evidence for both terms, and
there is no IRD/IKR anywhere in the slice. Both IBAs: NO_FAILURE_CORE.

The cristae-formation IBD is seeded by a single gene — human APOO itself. Per
projects/IBA_REVIEW.md, a short donor list is not weak evidence; the claim is about where the
function arose, and the curator had the whole tree. Nothing in the family contradicts it, and
the fly and worm members are cristae-shaping proteins.

4. The SAM complex row is a complexome-profiling projection

GO:0001401 SAM complex, HDA, PMID:26477565. Queried QuickGO by
reference=PMID:26477565&goId=GO:0001401: 12 entities, resolved individually —
MIC10 (Q5TGZ0), MIC13 (Q5XKP0), MIC19 (Q9NX63), MIC25 (Q9BRQ6), MIC26 (Q9BUR5), MIC27 (Q6UXV4),
MIC60 (Q16891), SAMM50 (Q9Y512), MTX1 (Q13505), MTX2 (O75431), MTX3 (Q5HYI7), HSPA9 (P38646).
That is the whole MIB complex, not the SAM complex: everything that co-migrated in one
complexome profile received the SAM term. The SAM complex is defined as "A large complex of the
mitochondrial outer membrane"; MIC26 is a single-pass inner-membrane protein. The same
reference already produces the correct GO:0140275 MIB complex row for APOO, so nothing is
lost by removing the SAM row.

5. GO term definitions consulted (QuickGO /ontology/go/terms/<id>/complete)

• GO:0042407 cristae formation: "The assembly of cristae, the inwards folds of the inner
  mitochondrial membrane." Descendant of GO:0007007 (ancestor list confirmed via
  /ancestors?relations=is_a,part_of). Not obsolete; no secondary ids.
• GO:0061617 MICOS complex: names MIC26 among the S. cerevisiae subunits in its own
  definition.
• GO:0007007 inner mitochondrial membrane organization: the parent of cristae formation —
  true here, but strictly less informative than the IMP-supported child, hence MODIFY.
• GO:0001401 SAM complex: "A large complex of the mitochondrial outer membrane…".
• GO:0140275 MIB complex: "…components of the MICOS complex in the inner mitochondrial
  membrane, the SAM complex in the outer membrane, a conserved DNAJ protein (human DNAJC11) and
  Metaxin 1." Exactly right.
• GO:0044284 mitochondrial crista junction; GO:0005743; GO:0005739; GO:0005576;
  GO:0000139; GO:0005789; GO:0005198; GO:0180020; GO:1900208 — all checked, none
  obsolete.

InterPro2GO (GO_REF:0000002): IPR019166 "MICOS complex subunit MIC26/MIC27" and IPR033182
"MICOS complex subunit MIC26/MIC27, animal" both map to exactly GO:0042407 + GO:0061617.
Correct family, correct terms.

ARBA (GO_REF:0000120): ARBA00026962 asserts GO:0005739; it has 380 condition sets, and
condition set 26 is InterPro id IPR019166 + taxon Eukaryota. APOO satisfies it.

6. The affinage record

self_evaluation_pairwise: win, faith_pct: 100.0, trust gates clear (.affinage.log:
"APOO: trust gates clear"). The narrative is about the right protein — MIC26, Q9BUR5 — with no
symbol collision, and it correctly foregrounds the 2023 refutation of the secreted isoform,
which is the single most important fact about this gene's annotations. Substantively strong.

Two defects found:

1. A wrong PMID, used twice. Affinage cites PMID:26217776 alongside PMID:25764979 for the
   MIC26 crista-junction and MIC26/MIC27-antagonism findings, labelling it Biochimica et
   biophysica acta. PMID:26217776 is in fact "Mass spectrometry analysis of K63-ubiquitinated
   targets in response to oxidative stress" (Silva & Vogel, Data Brief 2015) — an unrelated
   yeast ubiquitination dataset. The intended paper is almost certainly PMID:26217777,
   "Data supporting the role of the non-glycosylated isoform of MIC26 in determining cristae
   morphology" (Koob, Barrera, Anand, Reichert, Data Brief 2015) — one digit away, correct
   authors, correct content. Verified by fetching both. Nothing in this review is cited to
   26217776.
2. Imported GO grounding not reused. Affinage's mechanism_profile proposes
   GO:0005198 structural molecule activity and GO:0005739 mitochondrion. Both were
   re-derived here from the primary literature and term definitions rather than imported; they
   happen to agree, which is reassuring but was not assumed.

What affinage missed (found by independent PubMed E-utilities searches on APOO AND mitochondria, MIC26 MICOS, apolipoprotein O cristae — Europe PMC REST was unavailable):

• PMID:23704930 (Weber 2013) — the paper that identified APOO in the Mitofilin/MINOS complex
  by complexome profiling in the first place, and the source of the cardiolipin-binding claim
  that belongs to APOOL rather than APOO. Its absence is precisely the failure mode the brief
  warns about: the decisive paper is titled for the paralogue.
• PMID:26477565 (Huynen 2016) — the origin of three GOA rows (MICOS, MIB and the erroneous
  SAM complex) and of the opisthokont-restriction fact that explains the PANTHER family.
• PMID:42647630 (2026, Sci Adv) — the current structural model of the MIC10 subcomplex and
  the only source making any cardiolipin-recruitment claim about MIC26 itself.
• PMID:26217777 — the intended target of the wrong citation above.
• PMID:25781180 and PMID:25997101 — both cited by GOA, both absent from the affinage
  citation list, and PMID:25781180 is the one paper reporting a negative cristae result for
  MIC26 depletion.

7. Reconciliation checks run

• GOA rows ↔ YAML entries: 35 ↔ 35, matched on
  (term id, evidence code, reference, normalized WITH/FROM), scripted, no leftovers either side.
• Every supporting_text verified as a verbatim substring of its cache file with
  checkquotes.py (which also checks file: references, unlike CI).
• Interaction partners resolved against UniProt: Q16891 = IMMT/MIC60 (758 aa), Q5TGZ0 =
  MICOS10/MIC10 (78 aa), Q6UXV4 = APOOL/MIC27 (268 aa). ISS donor F1SPZ9 = Sus scrofa APOO,
  TrEMBL, 166 aa (shorter than the 198 aa human protein); its own GO:0005576 is an HDA
  from the same PMID:22261194 porcine ECM proteomics run, i.e. the human ISS is transferred
  from a single high-throughput detection in an ECM-enriched cardiac fraction — a preparation
  in which abundant mitochondrial proteins are a standard contaminant, from a tissue that is
  ~30% mitochondria by volume. SOURCE_WEAK_OR_INFERRED.
• Ortholog-side check: UniProt's mouse entry Q9DCZ4 (Mic26) already lists only
  "Mitochondrion inner membrane" as its subcellular location, while the human entry still lists
  Secreted/Golgi/ER. The correction has reached one species' record and not the other.

8. Open items flagged for UniProt / GOA

• Q9BUR5 SUBCELLULAR LOCATION still carries Secreted, Golgi apparatus membrane and
  Endoplasmic reticulum membrane with ECO:0000269|PubMed:16956892 / PubMed:25764979,
  and the PTM line still carries the chondroitin-sulfate O-glycosylation. PMID:37279200
  withdraws all of it. While those lines stand, GO_REF:0000044 will keep regenerating the
  three IEA rows removed here.
• SIGNAL 1..25 on Q9BUR5 versus TRANSIT 1..27 on the paralogue Q6UXV4 is an internal
  inconsistency in the family (see §1.4).