APOOL (MIC27, Q6UXV4) — review notes

Human APOOL, HGNC:24009, X chromosome, UniProt Q6UXV4 (MIC27_HUMAN, 268 aa, sequence
version 1, entry version 154 of 02-SEP-2026). Accession checked: the fetched record is
ID MIC27_HUMAN with GN Name=APOOL; Synonyms=CXorf33, FAM121A, MIC27;, so this is the
intended protein and not a merged-accession redirect. Secondary accessions Q3KNU7, Q5H9D1.

Despite the name, this is not a plasma apolipoprotein. UniProt: DE RecName: Full=MICOS complex subunit MIC27; AltName: Full=Apolipoprotein O-like;. The name is a 2003 secreted-protein
discovery-initiative artefact (RN [1], PMID:12975309, "The secreted protein discovery initiative
(SPDI) ... to identify novel human secreted and transmembrane proteins"), and it accounts for two of
the three bad cellular-component rows. The third, SAM complex, is an independent complexome
conflation and has nothing to do with the name.

1. What the protein is

PANTHER family PTHR14564 "MICOS COMPLEX SUBUNIT MIC26 / MIC27 FAMILY MEMBER" (from
DR PANTHER; PTHR14564;), Pfam PF09769 (ApoO), InterPro IPR019166 / IPR033182. The family
holds two mammalian paralogues, MIC26 (APOO, Q9BUR5, 198 aa) and MIC27 (APOOL, Q6UXV4, 268 aa);
PANTHER's representative-member list for the family gives one member per invertebrate genome it
covers (C. elegans moma-1 Q21154, D. melanogaster Mic26-27 / CG5903 Q9VEY5), and the fly protein
is treated in the literature as the single MIC26/MIC27 orthologue. Huynen et al. place the pair late: PMID:26477565

Topology (UniProt FT, evidence ECO:0000255 except where noted): TRANSIT 1–27 (mitochondrion),
TOPO_DOM 28–110 intermembrane, TRANSMEM 111–129, TOPO_DOM 130–137 matrix, TRANSMEM 138–155,
TOPO_DOM 156–268 intermembrane, disordered 187–268, phosphoserine at 204
(ECO:0007744|PubMed:24275569). Weber et al. saw the same architecture: PMID:23704930.

2. Localization — settled, and settled against the name

Weber et al. 2013 did the whole ladder in HeLa/143B cells. Immunofluorescence plus subcellular
fractionation: PMID:23704930 Alkaline carbonate extraction:
PMID:23704930 — i.e. integral or
tightly membrane-associated. Protease protection with graded digitonin: PMID:23704930

Lubeck et al. 2023 closed the secreted-isoform question for the whole family, in four human cell
lines with four antibodies plus tagged constructs: PMID:37279200 and PMID:37279200 The 55 kDa
"secreted apolipoprotein O" band that motivated the family name turned out to be a non-specific
antibody signal.

This matters because GOA carries two Reactome TAS rows placing APOOL in the extracellular region
and the platelet alpha granule lumen. I traced them: Reactome still lists Q6UXV4 among the 27
participants of Platelet degranulation (R-HSA-114608), with entities R-HSA-8863003 (platelet alpha
granule lumen) and R-HSA-8862979 (extracellular region); the reaction's only PubMed-backed literature
reference is Coppinger et al. 2004, a thrombin-releasate shotgun proteomics screen: PMID:14630798 That is a 2004 bulk list, nine years before anyone knew the protein
was mitochondrial. Reactome itself now also carries Q6UXV4 as R-HSA-8949612 "APOOL (MIC27)" in the
mitochondrial inner membrane under Cristae formation (R-HSA-8949613), so the platelet placement is
a stale duplicate rather than a claim Reactome is defending.

Reference-projection check (scripted, QuickGO paginated): Reactome:R-HSA-481007 supports 134
annotations over 67 distinct gene products, exactly two terms each — GO:0005576 ×67 and
GO:0031093 ×67. So both APOOL rows are one undifferentiated projection of a pathway membership,
carrying no APOOL-specific observation at all.

3. Complex membership

4. Molecular function: cardiolipin binding, and the hole in GOA

The one molecular activity ever demonstrated for this protein: PMID:23704930 — recombinant, affinity-purified,
GST non-fusion control negative on every lipid, triplicate lipid strips.

GOA has no molecular function for APOOL except three bare GO:0005515 protein binding IPI rows.
GO:1901612 cardiolipin binding exists and is not obsolete (QuickGO: "Binding to cardiolipin",
isObsolete: false). Neither GOA nor the UniProt DR GO block carries it. This is the single
clearest gap in the record and is the one NEW row I add.

Mechanistic corroboration (computational, 2026): PMID:42647630 and PMID:42647630 This is AlphaFold3 + coarse-grained/atomistic MD, not
experiment, and I treat it as such.

Bioinformatics (this review). APOOL-bioinformatics/loop_conservation.py tests the "conserved
positive loop motif" claim against the family alignment, using live UniProt sequences, lengths
asserted against interpro/panther/PTHR14564/PTHR14564-entries.csv, and UniProt's own TM
boundaries. The human inter-TM loop is residues 130–137 = RKGSKFKK, five basic residues, no
acidic residue, net +5. Projected over the nine PTHR14564 representative members plus the fly PAINT
seed Q9VEY5, 10/10 carry ≥2 K/R and a net positive charge in the aligned block, with human
MIC26 at RGSKIKK (+4) and the alignment placing APOOL R130 opposite MIC26 R129. So APOOL retains —
indeed maximises — the family's lipid-facing basic element: there is no residue-level argument for
divergence, and the retention supports cardiolipin binding as a property of MIC27 itself rather than
of the complex. Limits are stated in APOOL-bioinformatics/RESULTS.md: this is conservation and
charge, not a binding demonstration, and no CL-binding-deficient APOOL mutant has been reported.

The three protein binding IPI rows (PMID:25764979) resolve to MIC60/IMMT (Q16891), MIC10/MICOS10
(Q5TGZ0) and MIC26/APOO (Q9BUR5) — all verified against UniProt. They are real, but they are
intra-complex contacts already stated by part_of GO:0061617, and protein binding carries no
functional information. There is no informative MF to promote them to: the mechanistic claim that
would justify one — that MIC27 stabilises MIC10 oligomers — is yeast work (see §6), and the
human genetics contradicts transferring it.

5. The knockdown/knockout evidence, and the one real conflict in the literature

Three human perturbation studies, and they do not all agree.

  1. Weber 2013, miRNA knockdown, HeLa/143B. PMID:23704930, plus grossly reduced
    basal OCR. Overexpression gave the reciprocal perturbation (abstract: "Overexpression of APOOL led
    to fragmentation of mitochondria, a reduced basal oxygen consumption rate, and altered cristae
    morphology"). Both directions are perturbation-of-level experiments — IMP-grade, not IDA, and the
    overexpression arm is ectopic.
  2. Ott 2015, inducible shRNA knockdown, HeLa — negative. PMID:25781180, and their headline
    conclusion PMID:25781180 They note the disagreement with Weber themselves.
  3. Anand 2020, CRISPR knockouts and the double knockout, HAP1 — resolves it. This is the study
    that reconciles 1 and 2, because it counts crista junctions rather than scoring morphology, and
    because it removes the confound both earlier studies had: MIC26 and MIC27 regulate each other's
    levels, so a single depletion is partially compensated. PMID:32788226 and PMID:32788226. The onion-stack MICOS signature appears only in the
    DKO; the single MIC27 KO gives PMID:32788226.

So: MIC27 loss does reduce crista junctions in human cells, Ott's negative result reflects the
insensitivity of qualitative morphology scoring under paralogue compensation, and cristae formation
(GO:0042407) is properly asserted for this gene on human experimental grounds — which GOA currently
supports only by IBA, InterPro IEA and a ComplexPortal NAS.

Two things Anand 2020 shows that constrain how far to go:

6. Yeast and fly work — context, not human fact

A large part of the MIC27 mechanistic literature is S. cerevisiae and must not be restated as human
biology (Anand 2020 directly contradicts the transfer of the scaffolding claim):

The human-versus-yeast split is the reason I do not promote the protein binding IPIs to an
oligomer-stabilisation MF.

7. The two IBA rows

Both come from PANTHER node PTN001803267 in PTHR14564. The local PAINT slice
(interpro/panther/PTHR14564/PTHR14564-paint.tsv) has exactly two IBD rows, both at that node, and
no IRD or IKR anywhere in the family:

node term aspect evidence seeds
PTN001803267 GO:0061617 MICOS complex C IBD FB:FBgn0038400, UniProtKB:Q6UXV4, UniProtKB:Q9BUR5
PTN001803267 GO:0042407 cristae formation P IBD UniProtKB:Q9BUR5

Donors resolved:

Both transfers land inside the clade that inherited the function: APOOL is one of the two human
members of the family, it is experimentally in MICOS, and §4 shows it retains the family's basic
inter-TM loop. Note the asymmetry worth recording — the cristae-formation node is seeded by
MIC26 alone, so that IBA is, phylogenetically, MIC26's experiment inherited by MIC27; the
MICOS-complex node is seeded by both human paralogues plus the fly single-copy member. Human
MIC27-specific evidence for the inherited process has since arrived independently (Anand 2020,
§5), which is the strongest possible outcome for an IBA review.

8. What affinage missed / got right

Affinage trust gates cleared (✓ APOOL: trust gates clear), self_evaluation_pairwise: win,
faith_pct: 100.0, and the record describes the correct protein throughout — no symbol collision.
Its narrative and dated-findings table are accurate as far as they go, and it surfaced the yeast
mechanism papers, the MIC13 dependency and the DKO study that the seeded review file alone would
never have reached.

What it missed, and why it matters here:

Its mechanism_profile proposes GO:0008289 lipid binding and GO:0005198 structural molecule
activity. I did not import either: GO:1901612 cardiolipin binding is the specific child the data
actually support, and "structural molecule activity" is contradicted for human MIC27 by Anand 2020's
dispensability result.

Its citation list contains one non-numeric id, PMID:bio_10.1101_2025.05.20.655052 (a bioRxiv
preprint on MIC27 APEX2 proximity labelling). Not citable as a PMID and not used.

9. Review decisions (27 GOA rows + 1 NEW)

10. Open questions

Recorded as knowledge_gaps in the review: (i) no cardiolipin-binding-deficient APOOL mutant exists,
so the binding activity has never been tied to any cellular phenotype; (ii) whether the OXPHOS
supercomplex/ATP-synthase destabilisation in the MIC26/MIC27 DKO is a direct role or a downstream
consequence of reduced cardiolipin and lost crista junctions is explicitly unresolved by the authors;
(iii) MIC26 and MIC27 are reciprocally regulated post-transcriptionally and the mechanism is unknown,
which is what made every single-gene depletion study before 2020 hard to interpret.