MTRNR2L12 curation notes

Journal for the review of MTRNR2L12 (nuclear humanin-like locus). Written 2026-10-04.
Written as a sibling of the completed MTRNR2L1-MTRNR2L7 reviews; the shared family background is
kept consistent with those notes (and is summarised at the end of this file).

MTRNR2L12 (P0DMP1, humanin-like 12, HN12) - locus-specific notes

• HGNC:37169, MT-RNR2 like 12 (pseudogene), locus_type pseudogene, 3q11.2 (REST lookup at
  rest.genenames.org, 2026-10-04). GRCh37 position chr3:96336030-96337067 PMID:31753007.
  UniProt P0DMP1 is PE3, "Inferred from homology".
• 24 aa MAPRGFSCLLLSTSEIDLPVKRRA; 23/24 vs humanin MAPRGFSCLLLLTSEIDLPVKRRA - the single
  difference is Leu12->Ser. This is the closest of the thirteen nuclear copies to humanin, and its
  sequence is identical to that of MTRNR2L8 (both computed in this session from the UniProt SQ
  records). All of the humanin determinants are intact: Pro3, Ser7, Cys8, Leu9-Leu11, Thr13, Ser14,
  and both secretion segments Leu9-Leu11 and Pro19-Val20.
• Leu12 is a position where humanin Leu12->Ala abolishes neuroprotective activity
  (file:human/MT-RNR2__Q8IVG9/MT-RNR2__Q8IVG9-uniprot.txt), but synthetic HN5 carries the same
  Leu12->Ser and is still cytoprotective PMID:19477263, so this is a caution, not a refutation -
  the same reading as for MTRNR2L1.
• The Down-syndrome paper notes a coding polymorphism that would erase even that one difference
  PMID:25720405.
• GOA inconsistency worth recording. MTRNR2L12 has only four rows (extracellular and cytoplasm,
  each IEA + ISS). It carries no IBA rows, while MTRNR2L8 - whose peptide sequence is identical -
  carries both GO:0048019 and GO:1900118 from PANTHER:PTN002141596, as do L11 and L13, which are
  far more diverged (verified from the per-gene GOA files in this repository). The difference tracks
  PANTHER subfamily/node membership, not sequence: UniProt lists L12 under PTHR33895:SF5 (as L8 does),
  while L11 and L13 have their own subfamilies SF16 and SF17. So the family's IBA coverage is
  inconsistent in a way visible from the alignment alone. This is a defect of coverage, not a reason
  to add the terms here: the right fix is to restrict the node, not to extend it (see the shared
  background below).

Literature: the most-cited of the thirteen loci, and why that is misleading

A PubMed search for "MTRNR2L12" (E-utilities, 2026-10-04) returns 15 records - far more than any
other nuclear humanin-like locus (L11 returns 0, L13 returns 2). Nearly all are transcriptomic:
the gene surfaces as a differentially expressed transcript, a co-expression hub, or a lncRNA in
studies of periodontitis, nasal polyps, narcolepsy CSF, focal cortical dysplasia, PM2.5 exposure,
liver-cancer extracellular vesicles, and so on. Two are worth reading for this review.

1. PMID:25720405 (J Alzheimers Dis 2015), the one paper named after the locus. Microarray plus
   RQ-PCR in blood mononuclear cells of 48 adults with Down syndrome; MTRNR2L12 was the single gene
   distinguishing younger severely-affected from older unaffected participants
   PMID:25720405. Important caveats:
2. This is transcript evidence in blood, in a small sample, offered as a candidate biomarker.
3. It designed its own primers from ENST00000600213, acknowledging that no commercial assay
   exists; locus specificity against the other twelve near-identical copies and against the
   mitochondrial rRNA is not demonstrated in the paper.
4. It asserts, with no citation, that a peptide exists
   PMID:25720405.
   I could find no primary report of an HN12 peptide; this sentence should not be propagated, and
   I flag it in reference_review rather than using it as support for anything.
5. PMID:35421970 (BMC Med Genomics 2022), platelet co-expression in COVID-19. MTRNR2L12 is a hub
   gene of the magenta module
   PMID:35421970, and the paper itself notes
   PMID:35421970.
   The company it keeps is the tell: MT-ND1 and MT-ND5 are mitochondrially encoded. A nuclear copy of
   a mitochondrial rRNA segment clustering with mitochondrial transcripts in a short-read
   co-expression analysis is exactly what read mis-assignment looks like. This does not prove
   mis-assignment, but it means the module membership cannot be read as evidence about a nuclear gene
   product, and the same caution applies to the other transcriptomic hits above.
6. Genetics: the CAD lookup included the L12 region (rs78077066 P = 0.015, rs12106821 P = 0.52), and
   PMID:31753007
7. Net: despite being the most frequently named locus of the thirteen, MTRNR2L12 has no
   peptide-level evidence, no activity assay, no interaction data, and no locus-resolved
   quantification that excludes the other copies. Expression-level notoriety is not functional
   evidence.
8. Actions: 4 MARK_AS_OVER_ANNOTATED (extracellular IEA + ISS, cytoplasm IEA + ISS). The extracellular
   rows are not removed here, unlike L3/L4/L7/L11/L13, because L12 retains both humanin secretion
   segments intact - the same reasoning as MTRNR2L1. core_functions: [], no NEW terms.

The family, and why these 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, PubMed-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.
The later lookup paper gives the range as
PMID:31753007, so L11-L13 are the three loci left out - consistent with their PE3 status, and in tension
with the volume of transcriptomic literature naming L12.

UniProt states the position in a CAUTION line on each entry: the active peptide is the product of the
mitochondrial gene, and whether any nuclear locus makes a physiologically active peptide is open. A
2022 review restates it
PMID:35372353.
HGNC classes all thirteen as pseudogenes.

What the GOA rows on MTRNR2L12 assert

Both ISS rows name humanin (UniProtKB:Q8IVG9) as donor, and 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, which are themselves ECO:0000250|UniProtKB:Q8IVG9 - so IEA and ISS
state the same humanin claim twice over. Nothing on this gene traces to an observation made on
MTRNR2L12.

Residues

Humanin has an unusually complete mutagenesis map (UniProt Q8IVG9). Single substitutions that
abolish neuroprotective activity: Pro3, Ser7, Cys8, Leu9, Leu12, Thr13, Ser14 (S14->A,R,W,E,P);
secretion requires Leu9-Leu11 and Pro19-Val20. HN12 differs from humanin at Leu12 only, and the
synthetic HN5 peptide carrying the same Leu12->Ser is cytoprotective PMID:19477263. So for this
locus sequence divergence is not the argument - the missing link is simply that no peptide has ever
been shown to exist.

Methodological caution

RNA-seq and microarray signals over a nuclear copy that is 23/24 identical to a peptide encoded
inside an extremely abundant mitochondrial rRNA are not locus-resolved unless the analysis
demonstrates it, and serum ELISAs for "humanin-like" species have not been shown to be
locus-specific for any member of this family. No antibody- or short-read-based measurement is used
here as evidence that MTRNR2L12 makes a peptide.