MTRNR2L3 curation notes

Journal for the review of MTRNR2L3 (nuclear humanin-like locus). Written 2026-10-03.

MTRNR2L3 (P0CJ70, humanin-like 3, HN3) - locus-specific notes

• Chromosome 20; PE2; 24 aa MATRRFSCLLLSTSEIDLSVKRRI; 19/24 vs humanin (P3T, G5R, L12S, P19S, A24I).
• Pro19->Ser is the decisive locus-specific fact. It is one of the two humanin secretion segments
  and the position Bodzioch et al. proposed as the secretion switch
  PMID:19477263.
  So the extracellular rows, which exist only by transfer from humanin, are contradicted by the
  source paper's own criterion -> REMOVE (both IEA and ISS). Pro3 is also lost (humanin P3A abolishes
  activity), while Thr13 and Ser14 survive.
• Two clinical reports measure a serum "humanin-like 3" species by ELISA: higher in an MS subgroup
  PMID:41777515 and lower in restless legs syndrome PMID:38630326. Note the circularity trap in the second: it
  reaches MTRNR2L3 through GO enrichment on negative regulation of the execution phase of apoptosis
  PMID:38630326 - i.e. through the propagated annotation being
  reviewed. It must not be cited back as support for it.
• Interesting tension: a secreted immunoreactive species is reported, yet the peptide lacks a
  determinant humanin needs for secretion. Either the ELISA is not locus-specific, or secretion does
  not require Pro19 in this context. Flagged as a question, not resolved.
• Actions: 2 REMOVE (extracellular), 4 MARK_AS_OVER_ANNOTATED. core_functions: [].

The family, and why these seven reviews look alike

Humanin is a 24-residue peptide encoded by a small open reading frame inside the
mitochondrial 16S rRNA gene MT-RNR2, and it is the member of this family with all the
experimental evidence: cytoprotection, BAX/BID/BIM binding, FPRL1/FPR2 engagement, the
CNTFR-alpha/WSX-1/gp130 receptor complex. Bodzioch et al. 2009 (PMID:19477263, verified -
Genomics 94:247-256) found thirteen MT-RNR2-like sequences in the nuclear genome whose
reading frames would encode humanin-like peptides HN1-HN13
PMID:19477263, and showed that at least ten are transcribed and respond to
staurosporine and beta-carotene PMID:19477263.

That paper is the single primary source behind every UniProt MTRNR2L entry, and its own
title calls the functionality potential. UniProt agrees, and says so in a CAUTION line on
each entry: the active peptide is the product of the mitochondrial gene, and whether any of
the nuclear loci makes a physiologically active peptide is open. A 2022 review restates the
position PMID:35372353.

What the GOA rows actually assert

Each of MTRNR2L1-7 carries the same six rows (MTRNR2L5 has three more):

Tracing them:

• Both ISS rows name UniProtKB:Q8IVG9 = humanin as donor. Humanin's own extracellular and
  cytoplasm annotations are IDA/EXP, so the donor side is sound.
• Both IEA rows map UniProt's own SUBCELLULAR LOCATION lines, and those lines are themselves
  ECO:0000250|UniProtKB:Q8IVG9 - by similarity to humanin. So the IEA and the ISS are the same
  humanin claim, counted twice.
• Both IBA rows descend from PANTHER:PTN002141596, the humanin-family node. The MF row's
  grounding is humanin's IDA for receptor antagonist activity (PMID:15153530, FPRL1 competition
  with amyloid-beta); the BP row's co-listed donor is UniProtKB:P0CJ72 = MTRNR2L5, whose own
  evidence is the one functional assay ever run on a nuclear humanin-like sequence.

So the answer to the central question is: with one exception, every row on MTRNR2L1-7 traces to
humanin, by sequence similarity or by phylogenetic descent, and none to an observation made on
the locus itself. The exception is MTRNR2L5, where synthetic HN5 peptide was assayed
PMID:19477263.

Residues: where similarity stops being an argument

Humanin has an unusually complete mutagenesis map (UniProt Q8IVG9). Positions where a single
substitution abolishes neuroprotective activity: Pro3, Ser7, Cys8, Leu9, Leu12, Thr13, Ser14,
Pro19. Secretion needs Leu9-Leu11 and Pro19-Val20
PMID:35372353.
Bodzioch et al. reached the same two positions from the alignment
PMID:19477263.

Alignment of the seven against humanin (MAPRGFSCLLLLTSEIDLPVKRRA):

(computed in this session from the UniProt SQ records; Cys8 and Leu9-Leu11 are intact in all seven.)

This is what makes the reviews differ from one another:

• Pro19 lost (L3, L4, L7): the secretion claim is actively doubtful, by humanin's own secretion
  determinants and by Bodzioch's position-19 criterion. The two extracellular rows are REMOVEd.
• Thr13/Ser14 lost (L4: T13V; L7: T13I + S14R + S7G): the functional rows lose their premise.
  Ser14->Arg is literally one of the humanin substitutions annotated as abolishing activity, so L7's
  two functional rows are REMOVEd. L4 keeps Ser14 and its T13 change parallels the HN5 Thr13->Ile
  variant, which reduces rather than abolishes protection, so its functional rows are marked
  over-annotated instead.
• Everything else: MARK_AS_OVER_ANNOTATED. Not wrong, not demonstrated, and asserted of a gene
  product that has never been shown to exist.

Note the counterweight that stops this becoming a blanket purge: synthetic HN5, which carries the
same Leu12->Ser substitution as most of the set, is cytoprotective. So Leu12 divergence is a
caution, not a refutation, and the family-wide antiapoptotic claim is plausible - just untested
per locus.

Negative and orthogonal evidence

A lookup across all thirteen loci in CARDIoGRAMplusC4D found no significant association with
coronary artery disease and no significant expression QTL
PMID:31753007. That does not show the loci are inert, but it is the only genetics-scale
evidence covering them.

Methodological caution that applies to every clinical report on this family

Serum "humanin-like 3" and "humanin-like 10" ELISAs exist and are used
PMID:41777515, and
antibody-based detection of MTRNR2L2 has been reported
PMID:38221658. These peptides are 24-28 residues long
and match humanin at 16-22 of 24 positions. Locus-specific immunodetection has not been demonstrated,
so none of these measurements is used here as evidence that a particular locus makes a peptide.