TRMT61A review notes

Reviewed as a pair with TRMT6 (genes/human/TRMT6/TRMT6-notes.md). The two proteins form
one obligate heterotetrameric enzyme, so the two review files are kept consistent: TRMT61A is
the catalytic subunit, TRMT6 the non-catalytic substrate-binding subunit.

Why this gene was selected

Contested substrate class. GOA carries two molecular functions for TRMT61A:

• GO:0160107 tRNA (adenine(58)-N1)-methyltransferase activity — EXP (PMID:16043508), IBA, IEA
• GO:0061953 mRNA (adenine-N1-)-methyltransferase activity — IDA ×2
  (PMID:29072297, PMID:29107537)

The enzyme itself is not in doubt. What is contested is whether mRNA is a real substrate class,
because the transcriptome-wide m1A mapping methods that produced the mRNA assignments have
disputed specificity, and a 2026 biochemical validation paper concludes that most mapped internal
mRNA "m1A" is in fact inosine.

The tRNA m1A58 activity (not contested)

The two-subunit architecture and the activity were established in yeast and then shown to be
conserved in humans:
PMID:16043508 and
PMID:16043508

Site specificity was demonstrated directly by the A58U substrate mutant:
PMID:16043508

The crystal structure of the human complex bound to tRNA3(Lys) confirms the stoichiometry and the
division of labour:
PMID:26470919 and
PMID:26470919

The mechanism of access to a buried base is also resolved:
PMID:26470919

This is the core function, and nothing in the 2025-2026 literature challenges it.

The mRNA m1A activity (contested)

What the two IDA papers actually say

Importantly, the two papers GOA cites for GO:0061953 are themselves the sceptical,
base-resolution papers that cut the earlier antibody-based m1A maps down from thousands of sites
to a handful. They are not maximalist claims:

PMID:29072297

and the same paper's overall conclusion is that mRNA m1A is something cells avoid:
PMID:29072297

The independent base-resolution study agrees on the same narrow, structure-driven substrate class:
PMID:29107537

This is mechanistically coherent rather than surprising: the complex recognises its substrate by
structure (the T-loop, splayed open by TRMT6 as above), so an mRNA that folds into a T-loop-like
element is a legitimate, if adventitious, substrate. UniProt's own annotation records exactly this
qualified reading ("N(1) methylation takes place in tRNA T-loop-like structures of mRNAs and is
only present at low stoichiometries").

The 2026 challenge

PMID:42337368

and the paper's bottom line:
PMID:42337368

SCARPET is a discriminating method rather than a blanket debunk of epitranscriptomics — applied
to the equally controversial internal m7G, it confirmed the modification:
PMID:42337368

Two things follow that matter for how far the challenge reaches:

1. The target of the SCARPET refutation is the newly mapped sites from the evolved reverse
   transcriptase (Zhou et al. 2019), not the Safra/Li 2017 sites. The 2017 base-resolution studies
   are cited approvingly.
2. SCARPET positively confirms nuclear-encoded mRNA m1A at PRUNE1 (A58) and at the MALAT1
   lncRNA: PMID:42337368 and
   PMID:42337368

Caveat (recorded honestly): PMID:42337368 never mentions TRMT6 or TRMT61A. It does not itself
attribute the confirmed PRUNE1/MALAT1 sites to this complex. The attribution would have to come
from the 2017 papers, which state that the cytosolic sites are TRMT6/TRMT61A-dependent and lie in
tRNA T-loop-like contexts. Connecting the two is an inference, not something either paper asserts,
and it is flagged as such in suggested_questions. (ND5 is mitochondrial and is TRMT10C's, not
this complex's — PMID:29072297 is explicit about that.)

So the net effect of the 2026 work is to narrow the mRNA substrate class to a handful of
T-loop-like sites, not to abolish it.

The opposing stream: cancer mRNA-m1A papers

A separate literature continues to report abundant, functionally consequential TRMT6/TRMT61A-deposited
mRNA m1A driving tumour phenotypes:

• PMID:41103012, concluding
  PMID:41103012.
• PMID:42003777

These rest on m1A-seq / MeRIP-type mapping, i.e. precisely the class of method PMID:42337368 argues
is unreliable for internal mRNA m1A. Note also that the direction of effect is inconsistent across
this literature — CRC and HCC papers report m1A stabilising target mRNAs, while the
neuroblastoma paper reports m1A destabilising its target via YTHDF2
(PMID:40897821).
Opposite directions are not fatal on their own (reader context can differ), but combined with a
shared methodological dependency they lower the weight of this stream considerably.

A sceptical line on PMID:40897821 specifically. Its title is internally incoherent: it calls
TRMT6 an m1A methylase and simultaneously says it acts by demethylating its target —
PMID:40897821 — while the abstract describes the opposite, deposition:
"TRMT6 mediates m1A modification of SST". A writer cannot demethylate. This is at best sloppy
wording that survived peer review; it is not a reason to dismiss the data outright, but it is a
reason not to lean on this paper for any mechanistic claim, and no annotation in this review is
supported by it.

Curation position taken

• GO:0160107 tRNA (adenine(58)-N1)-methyltransferase activity (EXP, IBA, IEA) → ACCEPT, core.
  Direct biochemistry, structure, and a sound phylogenetic assertion all converge.
• GO:0061953 mRNA (adenine-N1-)-methyltransferase activity (IDA ×2) → KEEP_AS_NON_CORE,
  with the methodological dispute recorded in reason. Explicitly not REMOVE: these are
  experimental annotations whose full texts are not in the cache (full_text_available: false for
  both), CLAUDE.md forbids overruling an IDA from an abstract, and the cited papers are the
  conservative ones, not the discredited antibody maps. Non-core because even on their own account
  the sites are few and sub-stoichiometric, the substrate recognition is a structural spillover from
  tRNA recognition rather than a separate evolved function, and the 2026 biochemistry narrows the
  set further.
• GO:0006397 mRNA processing (IDA ×2, same two papers) → KEEP_AS_NON_CORE for the same reason
  and to keep the action consistent with the MF it derives from.
• GO:0006396 RNA processing (IEA/ARBA) → MODIFY to GO:0030488 tRNA methylation. Correct but
  a distant ancestor: GO:0030488 is a descendant of GO:0006396 (verified against QuickGO), and
  the gene already carries the specific term.
• GO:0005515 protein binding (IPI ×4) → MARK_AS_OVER_ANNOTATED per project guidance. The
  biologically meaningful interaction (TRMT6, UniProtKB:Q9UJA5, three of the four rows) is already
  captured far better by GO:0031515.
• Nucleus/nucleoplasm and complex terms → ACCEPT.

Terms verified

All GO ids used were checked against QuickGO (id, label, aspect, obsolescence) rather than recalled:
GO:0160107 MF, GO:0061953 MF, GO:0031515 CC, GO:0030488 BP, GO:0006396 BP, GO:0006397 BP,
GO:0005634 CC, GO:0005654 CC. Ancestry check confirming GO:0006396 ⊃ GO:0030488 was run
against the QuickGO ancestors endpoint.

All cited PMIDs were re-verified against PubMed metadata (title, journal, year) before use.