NAT10 review notes

Why this gene was selected

NAT10 is an epitranscriptomic "writer" whose enzyme identity is not in question but whose
substrate class is. GOA carries three cytidine-N4-acetyltransferase MF terms for the same
protein — one for 18S rRNA, one for tRNA, one for mRNA — and only the third is contested.
The task is to keep the rRNA/tRNA core intact, record the mRNA methodological dispute rather
than adjudicate it by fiat, and decide what is core.

What is not in doubt: NAT10 is the eukaryotic ac4C writer for 18S rRNA and tRNA

NAT10 is the human ortholog of bacterial TmcA / yeast Kre33(Rra1p), an ATP-dependent RNA
cytidine acetyltransferase with an N-terminal helicase/ATPase module and a GNAT
N-acetyltransferase domain (UniProt Q9H0A0: DOMAIN 558..753 "N-acetyltransferase";
BINDING 287..296 and 470 = ATP; BINDING 629..631, 636..642, 725 = acetyl-CoA).

18S rRNA, direct biochemistry:
PMID:25411247 and, for the ATP requirement,
PMID:25411247

Two rRNA sites plus the tRNA activity and its adaptor:
PMID:25653167 and
PMID:25653167

Structural context — NAT10 is a bona fide constituent of the nucleolar SSU processome
PMID:34516797

The contested part: does human mRNA carry meaningful ac4C?

Side A — yes (Oberdoerffer lab)

Antibody-based acRIP-seq, 2018:
PMID:30449621

Base-resolution RedaC:T-seq, 2022, with a position-dependent (bidirectional!) translational effect:
PMID:35679869 and
PMID:35679869

Note this second result cuts against the simple "positive regulation of translation" framing of
the 2018 paper: within one lab's own data ac4C is stimulatory in CDS and inhibitory at the
start codon.

Side B — no (Schwartz lab)

Base-resolution ac4C-seq applied across evolution finds abundant ac4C in archaeal mRNA and
essentially none in human mRNA:
PMID:32555463 and
PMID:32555463
The method itself is published as a protocol PMID:33772246, which also frames human mRNA ac4C as an open
question rather than an established fact PMID:33772246.

The exchange is on the record, back to back in Mol Cell 2024

Schwartz side PMID:38640895

Oberdoerffer side PMID:38640896 and, on the other
dataset, PMID:38640896.

Neither side retracted. As of 2026 the question is open.

The disease literature does not engage the dispute

A large and growing 2025-2026 disease literature simply assumes NAT10-mediated mRNA ac4C on named
transcripts. Examples reviewed here:
- gastric cancer / DUSP1 PMID:41956987
- aged liver regeneration / Parp10 PMID:42315153
- renal fibrosis, combining both proposed activities PMID:42592486

These papers use antibody-based ac4C-RIP and NAT10 knockdown, i.e. exactly the assay class whose
specificity is at issue, and none of them cite or address PMID:38640895. They are therefore
weight-of-numbers, not weight-of-evidence, for the mRNA substrate claim. They do corroborate that
loss of NAT10 changes the abundance and translation of specific mRNAs — which is expected on
either model, because NAT10 loss cripples 40S biogenesis.

Position taken

GO:0106162 mRNA cytidine N-acetyltransferase activity is retained (two independent IDAs, and
CLAUDE.md forbids removing an experimental annotation whose full text I have not read) but marked
non-core on every one of its four annotation rows. GOA already carries both the activity and,
implicitly through the rRNA/tRNA terms, the uncontested core. The dispute belongs in reason
and the open question in suggested_questions.

Two propagation issues found in the IBA set

1. GO:0002101 tRNA wobble cytosine modification (IBA) — WITH/FROM is
   PANTHER:PTN000100786|UniProtKB:P76562. P76562 is TMCA_ECOLI, "tRNA(Met) cytidine
   acetyltransferase TmcA" (verified at UniProt). In bacteria the acetylated base is the wobble
   base PMID:25411247, but in eukaryotes the NAT10/THUMPD1 target is C12 in the D-arm of tRNA-Ser and
   tRNA-Leu PMID:25653167. The position is different and the term is wrong for human.
   -> MODIFY to GO:0051391 tRNA acetylation.

2. GO:0106162 (IBA) — WITH/FROM is
   AGI_LocusCode:AT1G10490|PANTHER:PTN000100786|UniProtKB:Q5JHC6|UniProtKB:Q9H0A0. Q5JHC6 is
   TMCA_THEKO, the Thermococcus kodakarensis enzyme (verified at UniProt) — i.e. the archaeal
   case where transcriptome-wide mRNA ac4C is not contested at all. The same paper that
   established archaeal mRNA ac4C failed to detect it in human mRNA. Q9H0A0 appearing in its own
   WITH/FROM is correct and expected (its IDAs are among the descendant evidences behind the
   IBD), not circular. No propagation failure asserted; the annotation is simply demoted to
   non-core along with the human IDAs.

Does NAT10 have a defensible protein-acetyltransferase molecular function?

Claims exist: histones/hTERT (PubMed:14592445), SUN1/mitotic chromosome decondensation
(PubMed:17631499), alpha-tubulin PMID:19303003,
CCDC84/CENATAC K31 PMID:31722219, and now histone H3 via a GSDMC scaffold
PMID:42176271

Assessment: not defensible as a molecular function for this protein, on present evidence.
UniProt reaches the same conclusion and says why — the in vitro KAT assays were run on truncated
protein:

"A number of papers have reported some protein lysine acetyltransferase activity in vitro
(PubMed:14592445, PubMed:17631499, PubMed:19303003, PubMed:26882543, PubMed:27993683,
PubMed:30165671). However, most experiments have been performed in vitro using a protein
construct lacking the RNA-binding region at the terminus ... Recent evidence suggests that
NAT10 mainly acts as a RNA cytidine acetyltransferase in vivo (PubMed:30449621)."

and

"In addition to RNA acetyltransferase activity, also able to acetylate lysine residues of
proteins, such as histones, microtubules, p53/TP53 and MDM2, in vitro ... The relevance of the
protein lysine acetyltransferase activity is however unsure in vivo (PubMed:30449621)."

The primary paper makes the same point directly:
PMID:25411247

Note the construct that is catalytically competent for lysines (164-834) is exactly the one that
has lost the ATPase/RNA-binding module required for the RNA reaction; the reverse control — does
full-length NAT10 acetylate a protein substrate? — is the one that has not been done cleanly.
The GSDMC/H3 result (PMID:42176271) is a co-IP-plus-knockdown chromatin study, not an enzymology
experiment, and NAT10 knockdown collapses ribosome biogenesis, so a downstream/indirect route to
altered H3 acetylation is not excluded. GOA's exposure here is only the BP term
GO:0006473 protein acetylation (IDA, PMID:31722219) — there is no protein-lysine-KAT MF term
on the record, which is the right state of affairs. That BP annotation is kept but non-core.

Other annotation calls worth recording

• GO:0005697 telomerase holoenzyme complex (IDA, PMID:18082603). The paper itself distinguishes
  holoenzyme components from associated NTPases: PMID:18082603 while
  NAT10 falls under PMID:18082603. Marked over-annotated, not removed.
• GO:0032211 negative regulation of telomere maintenance via telomerase (IMP, PMID:18082603)
  rests on an overexpression phenotype: PMID:18082603 Kept, non-core, caveat recorded.
• GO:0016020 membrane (HDA, PMID:19946888, NK-cell membrane proteome) for a nucleolar protein is
  a high-throughput carry-over; marked over-annotated.
• GO:0045727 positive regulation of translation (IDA, PMID:30449621) is directionally too
  confident given PMID:35679869's own finding that 5'UTR/Kozak ac4C inhibits initiation.
  Modified to GO:0006417 regulation of translation.
• All seven bare GO:0005515 protein binding rows marked over-annotated per project guidance.

PMID verification

Every PMID cited above was resolved against PubMed (title/journal/year) via the PubMed MCP, and
each supporting_text used in the YAML is a verbatim substring of the corresponding cached
publications/PMID_*.md. P76562 and Q5JHC6 were resolved against the UniProt REST API.