ARHGAP36: the arginine finger UniProt annotates, and the residue that is there

Every number below is produced by analyze_arhgap36.py in this directory.
Re-run it to regenerate this file; do not hand-edit.

uv run --no-project --with "biopython>=1.85" python analyze_arhgap36.py
uv run --no-project --with "biopython>=1.85" python analyze_arhgap36.py --self-test

1. What the UniProt entry says about itself

The entry therefore annotates a catalytic arginine at 258 and carries T there, while asserting GTPase-activator activity on ECO:0000250. Both statements are rule- or similarity-derived; neither is experimental.

2. Is that Site in register?

A ProRule Site is placed by profile alignment and can be misplaced, so the position is re-derived from controls that carry their own annotated arginine finger. Each control's finger is projected onto ARHGAP36 Rho-GAP domain to Rho-GAP domain.

Distinct positions the controls project onto: [258]. Unanimous: yes. Agrees with UniProt's own Site: yes.

Method control — every control projected onto every other control must recover a known arginine finger:

3. Could the arginine simply have moved?

The only result that would rescue the annotation is an arginine displaced by a residue or two. A window of ±10 residues around position 258 was scanned.

4. Does the sequence record agree with the published residue?

Three records name this residue in three different numberings. Rather than assume which isoform each belongs to, the site is recomputed under the canonical sequence and under every UniProt splice variant lying wholly upstream of it, and each published number is looked for among the results — against the record of the species that paper worked in.

4a. RHG36_HUMAN (Q6ZRI8), site 258

• T227 (PMID:33999959, human ARHGAP36) is reproduced by exactly one numbering: VSP_021358.

4b. RHG36_MOUSE (B1AUC7), site 246

• T246 (PMID:25024229, mouse Arhgap36) is reproduced by exactly one numbering: (canonical).

Both published numberings fall out of the splice-variant arithmetic. The threonine call therefore rests on three records that could each have disagreed — a profile-based UniProt annotation and two papers' own construct numbering — and on none of them alone.

4c. Is the substitution shared with the mouse ortholog?

The IBD under review sits at PANTHER node PTN000973894, a Eumetazoa-level node. Whether the substitution is a human-lineage quirk or a property of the subfamily the node propagates into turns on whether the mouse ortholog shares it. Both proteins are in PANTHER subfamily PTHR12635:SF8.

The ortholog's own Site is held to the same register test as the query's:

Human site 258 projects onto mouse position 246; mouse site 246 projects back onto human 258. Substitution shared: yes.

So the arginine was already gone in the last common ancestor of mouse and human — far below the Eumetazoan node the IBD is placed at, and therefore inside the clade that node propagates GO:0005096 into. The node placement, not the transfer mechanics, is what the review has to argue with.

5. Is the loss ARHGAP36's, or the family's?

Across all reviewed human proteins carrying the PROSITE RhoGAP profile PS50238:

Entries whose annotated arginine-finger position does not hold an arginine:

The query is among them: yes. The decoupling control (RHGBB_HUMAN) is not — it keeps its arginine and is nonetheless experimentally GAP-dead, so a retained arginine is not evidence of activity. The inference this report supports runs only in the other direction.

6. Catalysis versus binding

UniProt records an experimental RAC1 interaction via ARHGAP36's Rho-GAP domain. Losing the catalytic arginine does not by itself remove the binding surface, so the two are counted separately. The GAP:GTPase interface is taken from PDB 1TX4 (p50RhoGAP:RhoA:GDP:AlF4), as every GAP residue with an atom within 4.5 A of the GTPase or of the nucleotide/metal/fluoride.

Chain roles are proved, not assumed: chain A is 99.5% identical to Q07960 and chain B is 99.4% identical to P61586. The calculation recovers the control's own annotated arginine finger (282) among its contacts, as it must.

Projection in register (control's finger lands on ARHGAP36's own Site 258): yes.

8 of 25 projected interface positions (32%) are identical in ARHGAP36. A percentage alone has no scale, so the same interface was projected onto RHG06_MOUSE (O54834), the PAINT/IBD seed, experimentally GAP-active — the member of this family whose experiments are the basis of the IBD under review. It scores 9/25 (36%).

So the GTPase-contacting surface is conserved in ARHGAP36 to a degree comparable to the family member that does catalyse, while the one residue that performs the catalysis is not conserved. That is the configuration in which a domain can still engage a GTPase without accelerating its hydrolysis — the shape OCRL_HUMAN is documented to have. This is a statement about what the structure permits, not a measurement of binding: the binding claim rests on the curated experiment, not on this table.