APOO / MIC26 (Q9BUR5) — N-terminal targeting sequence and paralogue architecture

Reproduce with:

cd genes/human/APOO/APOO-bioinformatics
uv run python targeting_sequence_analysis.py

All sequences and feature spans are fetched live from the UniProt REST API at run time;
nothing about the outcome is hardcoded. The script asserts that each parsed sequence length
matches the length declared on that entry's ID line, and it reports (rather than silently
drops) any panel member whose N-terminal feature is not the type the panel design assumes.

Question

UniProt annotates Q9BUR5 residues 1..25 as SIGNAL with evidence ECO:0000255, i.e. a
sequence-analysis prediction with no experimental support, and the same record still lists
Secreted, Golgi apparatus membrane and Endoplasmic reticulum membrane among its
subcellular locations. Those locations, and the historical identity of this protein as the
plasma apolipoprotein "ApoO", all trace back to a 55 kDa immunoreactive band that a later
knockout-controlled study showed to be non-specific (PMID:37279200). The paralogue
Q6UXV4 (APOOL / MIC27) — same PANTHER family PTHR14564, same complex, same membrane — is
annotated by UniProt not with a SIGNAL peptide but with a TRANSIT peptide (1..27).

So: does the MIC26 N-terminus carry the sequence signature of a secretory signal peptide, or
of a mitochondrial matrix-targeting presequence?

Design

Two control panels of human proteins, scored on their own UniProt-annotated N-terminal
segment:

Discriminating metrics, all computed from the segment sequence: net charge (K+R−D−E), number
of acidic residues, Arg per 10 residues, maximum Eisenberg hydrophobic moment over an
11-residue window (amphipathicity, the hallmark of a matrix-targeting presequence), and the
maximum Kyte–Doolittle hydropathy over a 7-residue window (the h-region, the hallmark of a
secretory signal peptide).

Result

metric mito TRANSIT mean±SD [min,max] secretory SIGNAL mean±SD [min,max] APOO/MIC26 APOOL/MIC27 closer to
net charge (K+R−D−E) 4.80±1.55 [2.00,7.00] 0.60±0.84 [−1.00,2.00] 3.00 3.00 mitochondrial (|z|=1.2 vs 2.8)
acidic residues (D+E) 0.60±1.07 [0.00,3.00] 0.30±0.48 [0.00,1.00] 0.00 0.00 mitochondrial (|z|=0.6 vs 0.6)
Arg per 10 residues 1.58±0.41 [0.80,2.27] 0.17±0.28 [0.00,0.74] 0.40 0.37 secretory (|z|=2.9 vs 0.8)
max Eisenberg hydrophobic moment (11-res) 0.61±0.16 [0.37,0.81] 0.25±0.06 [0.16,0.34] 0.60 0.31 mitochondrial (|z|=0.1 vs 5.8)
max Kyte–Doolittle h-region (7-res) 1.80±0.52 [1.16,2.79] 3.04±0.31 [2.63,3.59] 1.70 1.69 mitochondrial (|z|=0.2 vs 4.4)

MIC26 segment scored: MFKVIQRSVGPASLSLLTFKVYAAP (residues 1–25).
MIC27 segment scored: MAAIRMGKLTTMPAGLIYASVSVHAAK (residues 1–27).

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 two most informative numbers are the last two rows:

It also has zero acidic residues in 25, as presequences characteristically do.

The one metric that disagrees, and why it does not rescue the signal-peptide call

MIC26 has only one arginine in the segment (Arg/10 = 0.40), well below the mitochondrial
panel (1.58±0.41). Classic presequences are Arg-rich, so this is a genuine mismatch and is
reported as such. But the paralogue behaves identically (Arg/10 = 0.37) and UniProt
nonetheless annotates its N-terminus as a TRANSIT peptide, so within this family a low Arg
count is evidently compatible with mitochondrial targeting. MIC26's basic character is
carried by lysine instead (K3, K20 in the segment, K26/K27 immediately after it): net charge
is +3 with no acidic counter-charges. Neither MIC26 nor MIC27 has Arg at −2 or −3, the MPP
cleavage motif — consistent with these being inner-membrane proteins whose N-terminal segment
may not be processed by MPP at all.

MIC26 vs MIC27 architecture

Global BLOSUM62 alignment of Q9BUR5 (198 aa) against Q6UXV4 (268 aa): 72/197 identities in
aligned columns (36.5%). MIC27's features projected onto MIC26 coordinates:

MIC27 feature MIC27 span aligned MIC26 span MIC26 residues max KD (7-res)
TRANSIT 1..27 1..26 MFKVIQRSVGPASLSLLTFKVYAAPK +1.70
TRANSMEM 1 111..129 110..128 FFPRLGVIGFAGLIGLLLA +3.01
TRANSMEM 2 138..155 136..153 LVYPPGFMGLAASLYYPQ +1.70

MIC26's own UniProt-annotated transmembrane span is 108..128 — i.e. it coincides with the
region aligned to MIC27's TM1, and MIC26 has no annotated counterpart to MIC27's TM2. A 2026
simulation study predicts two short transmembrane helices in both MIC26 and MIC27
(PMID:42647630). The hydropathy here is not decisive either way: the MIC26 region aligned
to MIC27 TM2 scores +1.70, clearly less hydrophobic than its own TM1 (+3.01) but not polar.
This analysis neither confirms nor refutes a second MIC26 membrane-embedded helix; it is
recorded as an open question, not a result.

Conclusions

  1. The MIC26 N-terminal segment that UniProt calls a SIGNAL peptide has the composition and
    amphipathicity of a mitochondrial targeting presequence, not of a secretory signal
    peptide, and in particular lacks the hydrophobic h-region that every member of the
    secretory control panel has. This is an independent, sequence-level line of support for
    the experimental conclusion (PMID:37279200) that MIC26 is exclusively mitochondrial, and
    it identifies a likely origin of the historical "secreted apolipoprotein" identity: a
    sequence-analysis prediction (ECO:0000255) that assigned the wrong class of targeting
    peptide.
  2. The homologous N-terminus of the paralogue MIC27 is annotated by UniProt as TRANSIT, and
    scores the same way. The SIGNAL/TRANSIT asymmetry between the two family members is an
    annotation inconsistency, not a biological difference.
  3. This is a sequence-feature argument, not an experiment. It cannot by itself exclude a
    minor secreted pool; the knockout, tagged-construct and mass-spectrometry evidence in
    PMID:37279200 is what does that. It is offered as corroboration and as an explanation of
    how the prediction went wrong.

Caveats