Gpdh3 (DROME) notes

2026-10-21 — Full re-review, independent of the ProtNLM2 predictions

ProtNLM2 output was not opened or used.

The benchmark accession E1JIT1 is Gpdh3-PB (1291 aa), one of three FlyBase products of
Gpdh3/CG43343 (FBgn0263048): PA = Q8IN06 (1304 aa), PB = E1JIT1 (1291 aa), PC = Q9VD20 (1303 aa)
[file:DROME/Gpdh3/Gpdh3-flybase.txt "Gpdh3-PB FBpp0297799 144.8 1291 4.81 E1JIT1 NP_001163685 ACZ94981"].
None of the three is Swiss-Prot reviewed. PANTHER uses Q8IN06 (PA) as the gene's tree leaf, and the
gene-level IBA/ISS annotations (cytosol ISS, IBA GO:0047952 and GO:0006072) sit on Q8IN06, not E1JIT1.
PB is a genuine alternative isoform, not a fragment. Compared with PA it replaces the 14-residue
segment EFPLFTAIHLICQS (PA 318-331) by a single D, right after the Pfam C-terminal GPDH domain
(194..312). The rest is identical, including the whole ~950-residue C-terminal extension and every
GPDH-core residue examined [file:DROME/Gpdh3/Gpdh3-bioinformatics/RESULTS-rereview.md].

PANTHER PTHR11728:SF8 ("GLYCEROL-3-PHOSPHATE DEHYDROGENASE [NAD(+)]-RELATED"), the same subfamily
as fly Gpdh1 (P13706), fly Gpdh2 (Q9W1U3), CG34149, yeast GPD1/GPD2 and others.
PANTHER tree (v19, filtered to D. melanogaster): leaves Gpdh1 PTN000198085 (branch 0.179), Gpdh2
PTN000198083 (0.606), Gpdh3 PTN002271176 (0.791), and CG34149 PTN001712215 (2.0) under PTN000198082.
PAINT placed NOT (IRD) annotations for GO:0005829, GO:0047952 and GO:0006072 on the CG34149 leaf
only. Gpdh3 keeps the positive IBAs from PTN000919759 (GO:0047952) and PTN000198006
(GO:0006072, GO:0005829). I read this as a deliberate PAINT decision and did not REMOVE anything on
the basis of family placement.
Bao & Friedrich 2020 list Gpdh2 and Gpdh3 as testis-biased, long-branch paralogs of the
muscle-specific cytosolic Gpdh1 PMID:33083121.
This fits Bgee's top tissue, early-mid elongation-stage spermatids.

GPD1 K204 (UniProt ACT_SITE, proton acceptor) corresponds to T205 in E1JIT1. The catalytic-loop
window is GALKN in GPD1 and Gpdh1 but GTLTD in Gpdh3.
The NAD-binding loop GSGNWG reads GAEGWG (middle Gly to Glu). This is conserved in every long
Drosophilid Gpdh3 ortholog recovered, so it is ancestral to the Gpdh3 lineage.
The catalytic Lys is retained in obscura-group, willistoni and lebanonensis orthologs, and lost
(T/S/N) in the melanogaster group. In D. melanogaster it is T in all cDNA- and genome-derived
entries, so it is not a gene-model error.
Conclusion: canonical GPDH activity is ancestral to the lineage but doubtful in D. melanogaster.
All catalytic and process annotations are now UNDECIDED. The direction-specific catabolic process
and the UniPathway-derived glycerophospholipid term are MARK_AS_OVER_ANNOTATED. No core function
is asserted.

GO:0141152 ACCEPT -> UNDECIDED. GO:0006072 ACCEPT -> UNDECIDED. GO:0016616 MODIFY -> UNDECIDED.
GO:0005975 MODIFY -> UNDECIDED. GO:0051287 KEEP_AS_NON_CORE -> UNDECIDED.
GO:0046168 UNDECIDED -> MARK_AS_OVER_ANNOTATED. GO:0006650 UNDECIDED -> MARK_AS_OVER_ANNOTATED.
GO:0042803 stays UNDECIDED.
The previous description claimed "supporting retained ... interconversion chemistry" from domain
presence alone. It has been rewritten.

Europe PMC search for Gpdh3/CG31169/CG43343 found no paper characterizing the protein. The falcon
deep research agrees and warns that historical "GPDH-3" is an isozyme of the classical Gpdh1 locus
(Bewley 1989, Wilanowski 1998), so those data were not used.

Relative: E1JIT1 is an isoform (PB). The other products, PA (Q8IN06) and PC (Q9VD20), are TrEMBL
entries with the same GPDH core and the same divergent residues. Q8IN06 is the PANTHER/GOA gene
representative, but no product of the gene is a reviewed canonical entry, and this folder already
covers the gene. The closest experimentally characterized relative is the in-subfamily paralog
Gpdh1 (P13706), SF8, the PTN000198082 clade; it has no review in the repo. No ortholog of
Gpdh3 outside Drosophilidae is characterized.
Modules: modules/glycerol_phosphate_shuttle.yaml and modules/glycerol_metabolism.yaml use PTHR11728
for the cytosolic NAD-GPDH step. Gpdh3 should not be a representative of that step.
Family review interpro/panther/PTHR11728/PTHR11728-review.yaml uses E1JIT1 as the SF8
representative and states that "Every subfamily is an NAD-dependent glycerol-3-phosphate
dehydrogenase". The subfamily placement is not in dispute: Gpdh3 is SF8 in both UniProt and the
PANTHER tree. But the family-wide scope for GO:0141152 does not hold for D. melanogaster Gpdh3, and
E1JIT1 is a poor representative for SF8; Gpdh1 P13706 would be better.