ACRV1 (SP-10): what the LU domain, the splice isoforms and the family do and do not license

Generated by analyze.py from the API snapshot in data/.
Rebuild the snapshot with uv run python fetch_data.py, then rerun
uv run python analyze.py. Machine-readable output: results.json.

1. ACRV1 is a low-complexity acidic tail plus a canonical LU (three-finger) domain

Member-database support for the same C-terminal domain:

The domain carries 10 cysteines at 190, 193, 202, 209, 217, 235, 239, 255, 256, 262 (spacing [3, 9, 7, 8, 18, 4, 16, 1, 6]), and the whole region N-terminal to it
contains 0 cysteines. Across the 38 LU domains in the 30 reviewed human LU-domain proteins the cysteine count distribution is {'5': 4, '6': 1, '7': 2, '8': 5, '9': 1, '10': 22, '11': 2, '12': 1} (mode 10); ACRV1 matches the modal count.

The 166-residue linker is composed almost entirely of a few residue types (S 18.7%, E 17.5%, G 13.3%, Q 7.8%, A 7.2%, H 7.2%).

Bearing on the review. The 1990 cloning paper reported that SP-10 "did not show any significant homology to other sequences" (PMID:1693291). That is no longer true: the C-terminal third is a canonical LU / three-finger-protein domain, the same fold as CD59, uPAR, SLURP1 and the snake three-finger toxins. Sections 3 and 4 test what, if anything, that buys us functionally.

2. Every abundant splice isoform keeps the LU domain; only rare ones truncate it

ACRV1 has 11 annotated isoforms, all generated by in-frame deletions. Mapping each deletion onto the LU domain (188-264, 77 aa):

LU domain intact in isoforms 1, 2, 3, 4, 5, 6, 7, 9, 10; truncated in 8, 11. Abundances are from PMID:7619499 (quantitative competitive RT-PCR, testes from four men); the remaining seven isoforms together account for <1% of SP-10 message.
All isoforms with a measured abundance retain the complete LU domain: True.

Bearing on the review. Alternative splicing in ACRV1 varies the length of the low-complexity spacer, not the folded domain, in >99% of the message. The isoform heterogeneity is therefore not evidence for isoform-specific molecular functions, and a GO annotation made on any of the abundant isoforms is an annotation on the same LU module.

3. ACRV1 is one of the few human LU-domain proteins with no GPI anchor

19 of 30 reviewed human LU-domain proteins carry a GPI-anchor lipidation site; 11 do not: ACRV1, GML, LY6G5B, LY6G5C, PATE1, PATE2, PATE3, PATE4, PINLYP, SLURP1, SLURP2.
ACRV1 has a signal peptide (True) and no lipidation feature (GPI anchor: False).

Bearing on the review. ACRV1 is not anchored, consistent with the experimental finding that SP-10 partitions as a hydrophilic, peripheral acrosomal protein released by chaotrope but not by detergent or salt (PMID:1591355), and its compartment is the lumen of a secretory organelle, topologically separate from the cytosol -- which is why its cytosolic yeast-two-hybrid partners cannot meet it in vivo.
What this count does NOT do is separate ACRV1 from the family members whose molecular functions might otherwise be transferred to it. PATE1, PATE4, SLURP1 and SLURP2 are unanchored too, and they are the acetylcholine-receptor modulators of section 5. Anchoring is therefore the wrong discriminator; the next subsection supplies the right one.

3b. Reaching a cell-surface receptor: the discriminator that does work

Classifying each member's UniProt subcellular locations as extracellular-accessible (any of Secreted, Cell membrane, Cell surface, Membrane, caveola): 29 of the 30 members carrying a location annotation qualify, and ACRV1 is the sole exception.
Every one of the 11 members carrying an acetylcholine-receptor molecular function is extracellular-accessible: True.

ACRV1's only annotated location is Cytoplasmic vesicle, secretory vesicle, acrosome, and it is extracellular-accessible: False. The members annotated to the acrosome are ACRV1, PATE4, SPACA4, TEX101 -- so PATE4 shares ACRV1's compartment, but PATE4 is also annotated as secreted, and ACRV1 is not annotated as secreted or as a cell-membrane protein at all.

4. ACRV1 has no close human paralogue inside the LU family

Subfamily-resolution signatures (CDD conserved-domain models and PANTHER families/subfamilies; the fold-level Pfam/InterPro entries are excluded because they are what defines the family in the first place) partition the 30 reviewed human LU-domain proteins into subfamilies. ACRV1's are PTHR17571, cd23628.
No other reviewed human LU-domain protein shares either model, so ACRV1 is a subfamily singleton in the human proteome.

Bearing on the review. ISS/ISO transfer to ACRV1 has to come from its orthologues (mouse Acrv1, baboon ACRV1, fox FSA-ACR.1), not from a human paralogue: it has none at subfamily resolution. Note also that the CDD model matching ACRV1 is named for SP-10 and the PATE-like proteins, so the PATE clade is its nearest structural neighbourhood -- which matters for section 5.

5. The human LU family has no shared molecular function to transfer

Of 30 reviewed human LU-domain proteins, 23 have at least one molecular-function annotation with an experimental evidence code, and 16 have one that is informative (excluding GO:0005102, GO:0005515, GO:0019899, GO:0019904, GO:0042802, GO:0048306).
Those 16 proteins between them carry 15 distinct molecular-function terms, and the most widely shared single term is held by 7 of 30 members:

ACRV1's own experimental molecular-function annotations are GO:0005515, of which informative: none.

Bearing on the review. The LU fold is a protein-interaction scaffold whose members do unrelated things -- complement regulation, urokinase receptor activity, nicotinic acetylcholine receptor modulation, laminin and integrin binding, phospholipase inhibition.
The one function with any breadth is GO:0030550 acetylcholine receptor inhibitor activity (7/30 members: LY6E, LY6G6D, LY6H, LY6S, LYNX1, LYPD1, SLURP2), and it is worth being explicit about the two members that make this a live question rather than a dismissal: PATE1 and PATE4 carry GO:0030548 acetylcholine receptor regulator activity, and the CDD model that matches ACRV1 is named for SP-10 and the PATE-like proteins. So ACRV1's nearest structural neighbours do have an assigned molecular function. PATE4 is closer still: it is annotated to the acrosome, the same compartment as ACRV1.
Transferring it would nonetheless be wrong on three counts computed above, and the first is topology, not GPI anchoring -- PATE1, PATE4, SLURP1 and SLURP2 have no anchor either (section 3), so anchoring cannot be the discriminator. What separates them is reachability: all 11 members with an acetylcholine-receptor function are annotated to a secreted or cell-membrane location where a surface receptor can be engaged, whereas ACRV1's only annotated location is the lumen of an intact secretory organelle, with no secreted or cell-membrane assignment (section 3b). PATE4 illustrates the point rather than undermining it: it shares the acrosomal location but is also secreted. Second, the term is not shared family-wide, only within one subclade, and ACRV1 belongs to neither the Ly-6/LYNX/SLURP nor the PATE PANTHER subfamily (section 4). Third, ACRV1 has no human paralogue at subfamily resolution to inherit from at all. Combined with ACRV1's own experimental record -- GO:0005515 and nothing informative -- the statement "no molecular-function term is currently justifiable for ACRV1" is a tested result rather than an assumption.