SPG21 (maspardin) review notes

Why this gene was selected

SPG21 is a case where the protein's fold asserts more than the evidence supports. Maspardin carries
an alpha/beta hydrolase (ABHD) domain, is a member of the AB hydrolase superfamily in UniProt, and sits
in Pfam PF00561 (Abhydrolase_1) - yet in twenty-five years no catalytic activity has been reported for
it, and a 2026 J Cell Biol paper still opens by calling it a protein of unknown function:
PMID:41400694 and
PMID:41400694. GOA's molecular-function content is a single
GO:0005515 protein binding from ten interactome screens, plus a GO:0042609 CD4 receptor binding
pair from 2001.

The catalytic-residue question, checked rather than assumed

The nucleophile elbow is intact. In Q9NZD8 the sequence at residues 107-111 is G-A-S-L-G, the canonical
G-x-S-x-G motif, with Ser109 at the nucleophile position; UniProt annotates the AB hydrolase-1 domain at
residues 87-159. So this is not a case of a degenerate fold.

The rest of the triad is not. Four independent sequence analyses, spanning the original disease paper to
a 2025 structural-prediction analysis, all report that the acid and histidine partners of the triad are
absent:

The original CD4 paper had already framed the domain as non-catalytic and repurposed:
PMID:11113139

What is still missing, and why I have not written "pseudoenzyme" as a finding. No assay has ever been
run. Every statement above is a sequence or structure inference, and the authors of the mouse-knockout
paper say so explicitly: PMID:20661613. An invertebrate comparative study hedges the same way:
PMID:22729480. The convergence of four independent
analyses is strong, and it is reinforced by the fact that the retained Ser109 sits inside the recurrent
pathogenic missense allele p.Ala108Pro, whose effect is protein misfolding and degradation rather than loss
of a catalysis - but converging predictions are not a negative experiment. This goes in
suggested_questions and knowledge_gaps, not into core_functions as an asserted pseudoenzyme call.

What maspardin does do: RAB7A effector

Two 2025-2026 papers, from independent laboratories, converge on the same molecular description.

Bonifacino group: PMID:40833810 and PMID:40833810, with the specificity result
PMID:40833810.

Boonen group: PMID:41400694 and PMID:41400694, with the mechanism running through RAB7
positioning: PMID:41400694 and PMID:41400694.

So the honest molecular function is small GTPase binding (GO:0031267; GO merged "Rab GTPase binding"
into it) - specifically binding GTP-loaded RAB7A - and the honest statement of what remains dark is the
effector output: nobody knows what maspardin does to RAB7A once bound, or how binding keeps RAB7A on
retromer/TBC1D5-positive late endosomes rather than on lysosomes.

The CD4 annotation

GO:0042609 CD4 receptor binding is a real result: a two-hybrid screen with the CD4 cytoplasmic tail
recovered ACP33, and the interaction was mapped to CD4's C-terminal hydrophobic residues
PMID:11113139. It is not overturned,
and I have not removed it. But it is not the conserved core function: maspardin is ubiquitously expressed,
is conserved to insects and beyond (CD4 is vertebrate-specific), the disease is neurological rather than
immunological, and the IBA row is seeded solely by this human gene's own IPI (WITH/FROM
PANTHER:PTN002698091|UniProtKB:Q9NZD8), so the phylogenetic assertion projects a vertebrate-specific
partner across a family that predates it. Marked non-core with that reasoning stated.

Curation position taken

Open question for experts

Has anyone ever assayed maspardin for hydrolase activity? The pseudoenzyme call has been repeated for
twenty-two years on sequence grounds alone, and the retained, mutation-hit nucleophile elbow is exactly
the configuration in which a fold-retaining protein sometimes turns out to retain residual activity.