NSUN7 (Q8NE18) — review notes

Summary of the position taken

NSUN7 is the pseudoenzyme of the NSUN (NOL1/NOP2/Sun) m5C RNA methyltransferase family.
It retains the canonical NSUN fold and the sequence motifs that the family uses for catalysis,
but it has lost the key SAM-contacting aspartate of motif IV, does not bind SAM, and its loss
does not change RNA m5C levels in the tissue where it is actually expressed. Its demonstrable
biochemical activity is RNA (chiefly mRNA) binding, and its physiological role is
post-transcriptional regulation in elongated spermatids supporting sperm flagellum assembly
and progressive motility.

Consequently the only molecular-function annotation GOA carries — GO:0008168 methyltransferase activity, IEA from InterPro IPR001678 — is argued down (REMOVE). This is exactly the case
CLAUDE.md permits: an electronic, family-signature-derived inference that direct biochemistry
on the protein itself contradicts. UniProt itself has already added a CAUTION to the same effect
(see below).

Catalytic inactivity — the direct evidence

Li et al. 2025 (Nat Commun) is the decisive study. They measured SAM binding by ITC on purified
protein and measured m5C transcriptome-wide in knockout testis:

Independent genetic corroboration from a different laboratory, using a catalytic-site point
mutant rather than a null: Guseva et al. 2025 (Biochimie) knocked in the putative catalytic
cysteine substitution C382A and saw no phenotype at all:

This is a strong design: if the phenotype were caused by loss of methyl transfer, a catalytic
mutant should phenocopy the null. It does not. The same paper reports structural modelling
showing a re-arranged catalytic site with an abnormally long motif IV–motif VI cysteine distance.

UniProt has already acted on this (genes/human/NSUN7/NSUN7-uniprot.txt):

CAUTION: In contrast to other NSUN family members, NSUN7 lacks RNA cytosine C5-methyltransferase
activity, due to sequence variations in S-adenosyl-L-methionine(SAM)-binding residues.
Substitution of the conserved SAM-binding Asp-373 with Leu-373 abolishes SAM binding and likely
renders the enzyme catalytically inactive.

The UniProt keyword list also already carries RNA-binding but no Methyltransferase/Transferase
keyword. The stale piece is the InterPro2GO mapping (IPR001678 → GO:0008168), which fires on
the fold, not on the catalytic residues.

What NSUN7 actually does

Guseva et al. 2025 (RNA) did UV CLIP-seq from testis and RNA-seq of the knockout:

Notably, the same paper's own bisulfite sequencing failed to find NSUN7-dependent methylation
that tracked with binding:

Li et al. independently observed RNA-dependence of NSUN7 protein stability:

So mRNA binding (GO:0003729) is the informative molecular function to put in its place —
and it is exactly the sort of replacement CLAUDE.md asks for when removing an uninformative or
wrong MF term.

An unresolved conflict in the direction of the effect

The two groups disagree about the sign of NSUN7's post-transcriptional effect:

Li et al. are explicit that they cannot separate cause from effect, since the knockout also has
gross structural defects that could secondarily perturb mRNA storage and stability. For this
reason I have not asserted GO:0061157 mRNA destabilization (or an mRNA-stabilisation term)
in core_functions; only the binding activity and the flagellar assembly process, both of which
both groups agree on. The direction is recorded as a suggested_question.

A third study (proteomics of sorted spermatid populations) favours a protein-complex-stabilising
model rather than an RNA-turnover model:

Localisation and phenotype (basis of the ISS annotations)

The human ISS annotations are transferred from mouse Nsun7 (UniProtKB:Q14AW5) and are well grounded:

The original ENU forward-genetic identification:

Human genetic association with asthenozoospermia:

Weighing the counter-claims that NSUN7 deposits m5C

Three recent disease papers assert NSUN7-catalysed m5C. All three are knockdown/knockout plus
MeRIP-or-bisulfite correlation studies: none of them tests catalysis directly (no SAM binding,
no purified-enzyme in vitro methylation, no catalytic-dead rescue).

A further tissue-expression problem: NSUN7 is strongly testis-restricted
(PMID:41381527),
yet these studies place it as a dominant m5C writer in kidney, ovary and brain tumours.

Weighting. Evidence that directly interrogates the catalytic mechanism (ITC SAM binding on
purified protein; a catalytic-residue knock-in mouse with no phenotype; transcriptome-wide m5C
mass spectrometry in the knockout) beats evidence that assumes catalysis and measures a
downstream correlate. The three positive claims are therefore recorded but do not rescue
GO:0008168.

Annotation decisions

Term Evidence Action Why
GO:0008168 methyltransferase activity IEA (InterPro IPR001678) REMOVE No SAM binding (ITC, mouse and human); catalytic-site mutant mouse is normal; KO does not change m5C; UniProt CAUTION agrees. Electronic fold-based inference, contradicted by direct biochemistry.
GO:0031514 motile cilium IEA ACCEPT Sperm flagellum is a motile cilium; the more specific GO:0036126 is also present.
GO:0036126 sperm flagellum ISS ACCEPT Core location.
GO:0005737 cytoplasm (x2) IEA, ISS ACCEPT Correct, if general.
GO:0030317 flagellated sperm motility ISS ACCEPT Core process.
GO:0120316 sperm flagellum assembly ISS ACCEPT Core process.

New MF proposed in core_functions: GO:0003729 mRNA binding.

Reusable principle applied here

A family signature is not a catalytic assay. Where an IEA molecular function comes from a
domain/family signature (InterPro2GO, PANTHER family) and direct experiment on the protein shows
the catalytic requirements are not met — cofactor not bound, catalytic residue substituted, no
change in product on loss of function — the IEA should be removed and replaced by whatever the
protein is actually shown to do. Do not leave the catalytic term in place merely because the
protein belongs to an enzyme family.