SLC45A4 (Q5BKX6) — curation journal

Scope of the problem

SLC45A4 went from orphan to twice-deorphanized within three months of 2025, by two
independent groups that agree on the chemistry (polyamines) but disagree on the
compartment (plasma membrane vs peroxisome membrane). A third 2025 paper independently
nominates plasma-membrane polyamine handling. On top of that sits a legacy family-level
assignment as a sucrose:H+ symporter. This journal records what each line of evidence
actually shows.

1. The legacy sucrose annotation — where it really comes from

The brief's working hypothesis was that GO:0008506 (sucrose:proton symporter activity)
on human SLC45A4 is a plant-derived IBA. That is only half true, and the half that is
untrue matters. Tracing the donors:

• IBA WITH/FROM for human Q5BKX6 is
  AGI_LocusCode:AT1G09960|AT1G22710|AT1G71880|AT1G71890|AT2G02860|AT2G14670|AT5G06170|MGI:MGI:1922082|MGI:MGI:2146236|MGI:MGI:2153040|PANTHER:PTN000751434|PomBase:SPAC2F3.08|UniProtKB:Q10R54|UniProtKB:Q69JW3.
  Q10R54 = rice SUT1, Q69JW3 = rice SUT5 (both plant sucrose transporters) — but the
  three MGI: donors are mouse SLC45 paralogs, not plants.
• QuickGO for GO:0008506, taxon 10090, ECO:0000314 returns exactly three mouse genes —
  Slc45a2 (P58355), Slc45a3 (Q8K0H7) and Slc45a4 (Q0P5V9) — all IDA, all from
  PMID:25164149.
• The human ISS (GO_REF:0000024, ParkinsonsUK-UCL, 2016-03-01) has
  WITH/FROM = UniProtKB:Q0P5V9, i.e. it is a direct ortholog transfer from that mouse
  SLC45A4 IDA.

So the PAINT node placement is not a plant-to-animal leak: the curator had mammalian
experimental annotations sitting inside the clade. The problem is one node up — the
source experiment itself.

PMID:25164149 (Bartölke et al., Biochem J 2014) is a yeast heterologous expression
study: PMID:25164149
with PMID:25164149 A Km of ~5 mM for a disaccharide that
mammals never present to the cytosolic face of any membrane (dietary sucrose is hydrolysed
in the intestinal lumen by sucrase-isomaltase) is a weak basis for a physiological
molecular function.

Two 2025 papers tested sucrose directly and found nothing:

• Purified human protein, nanoDSF: PMID:40836097 — while polyamines did:
  PMID:40836097.
• Isotope tracing in cells: PMID:41266324 (they are appropriately careful that this does not rule
  out activity in the absence of glucose).

The homology premise is also thinner than the family name suggests:
PMID:40836097 and
PMID:40836097. UniProt has already added a CAUTION to Q5BKX6 on exactly
this point.

Decision: REMOVE GO:0008506 (IBA and ISS) and GO:0015770 (ISS), with
propagation_review root_cause SOURCE_WEAK_OR_INFERRED — the failure is not the
phylogenetic placement, it is that the seed IDA is a heterologous, low-affinity,
non-physiological activity that the human protein has since been shown not to exhibit.
This should propagate upstream: the mouse IDAs on Slc45a2/3/4 deserve re-examination too.

2. Compartment: plasma membrane vs peroxisome

Case for plasma membrane (PMID:40836097, Nature 2025; PMID:41075780, Cell Metab 2025)

PMID:40836097
PMID:40836097
PMID:41075780

Caveat: both are overexpression systems (eGFP fusion electroporated into mouse DRG
neurons; N2a cells; HEK293T KO + cDNA). Overexpression of a multi-pass MFS protein routinely
saturates organellar retention and spills to the surface.

Case for peroxisome membrane (PMID:41266324, Nat Commun 2025)

PMID:41266324
PMID:41266324
PMID:41266324
PMID:41266324
PMID:41266324
and biochemically, two independent peroxisome preparations:
PMID:41266324

Strength: this is endogenous protein, with an antibody validated against an HA-tagged
construct, in A549/H1299 lung-cancer lines, backed by subcellular fractionation. It is
methodologically the stronger localisation dataset. Weakness: it is one lab, in
transformed epithelial lines, and does not test neurons.

How I am reading it

These two results are not logically incompatible. The systems differ on essentially every
axis — endogenous cancer-line protein vs overexpressed neuronal protein — and dual
localisation of a polyamine carrier (surface uptake plus peroxisomal delivery to diamine
oxidase) is a coherent cell-biological picture. I therefore keep GO:0005886 (IDA,
is_active_in) rather than displacing it, and add GO:0005778 (peroxisomal membrane) as a
reviewer-proposed NEW annotation from PMID:41266324 rather than as a replacement. The
description states the disagreement as a fact about the biology, and the unresolved
question goes to suggested_questions. The cryo-EM structure (which is compartment-agnostic)
constrains the substrate, not the location.

3. Substrate: how much of the "spermine/spermidine/putrescine" IDA set is direct?

Worth being precise, because GOA carries four separate MF IDAs from PMID:40836097.

• Spermidine is the only directly radiolabelled flux: 14C-SPD uptake time course in N2a
  cells overexpressing human or mouse SLC45A4 vs empty vector.
• Spermine, putrescine and cadaverine enter via (i) nanoDSF thermal destabilisation of
  the purified protein and (ii) competition against 14C-SPD uptake:
  PMID:40836097.
  Binding + competition is good evidence of recognition; it is one inferential step short of
  demonstrated translocation for each individual amine.
• Selectivity is for polyamines but not among them:
  PMID:40836097, with no
  inhibition by L-lysine, L-ornithine or L-arginine.

All four MF IDAs are kept (ACCEPT); the nuance is recorded in each reason. The substrate
disagreement in the headlines ("polyamine" vs "putrescine") is therefore largely illusory:
putrescine is a polyamine and is the highest-affinity ligand in the Nature dataset.

A direction-of-transport wrinkle

PMID:41075780's cellular readout is not obviously an uptake signature:
PMID:41075780 and PMID:41075780. Intracellular depletion plus
extracellular accumulation reads as export of acetylated polyamines. The authors
themselves flag the evidential ceiling: PMID:41075780 GO MF terms are direction-agnostic, so this does not change any
term choice, but it is a real open question and goes to suggested_questions.

Note also that the reference title for PMID:41075780 is about SLC25A45, not SLC45A4 —
the citation is nonetheless correct: SLC45A4 is a co-equal subject of that paper
(PMID:41075780). Flagged in
reference_review so a future reader does not mistake it for a transposed PMID.

4. Pain

PMID:40836097 plus
PMID:40836097 and a UKB GWAS
association with chronic pain intensity. GO:0019233 IMP accepted. The ARBA IEA for the
same term is accepted for consistency (same-term-same-action).

5. GO:0005515 protein binding (IPI, PMID:36115835)

Eighteen partners, all PDZ-domain scaffolds (DLG1–4, GRIP1/2, MPDZ, PATJ, PDZK1, MAGI1,
LNX1/2, WHRN, NHERF4, APBA2, IL16, FRMPD2), from a quantitative PDZ-domain holdup assay.
Real, and biologically suggestive of a C-terminal PDZ-binding motif anchoring SLC45A4 at a
membrane — but protein binding itself carries no functional information, so
MARK_AS_OVER_ANNOTATED per project guidance. GO:0030165 (PDZ domain binding) would be the
informative term if a curator wanted to keep it.

6. Actions taken