ACRBP (Q8NEB7) — review notes

Human ACRBP / sp32 / OY-TES-1 / CT23. Acrosin-binding protein, 543 aa, testis-restricted,
chromosome 12p12-p13.

What the protein is

ACRBP is an acrosomal matrix protein of the sperm head. It is synthesised as a ~60-kDa
precursor with a cleaved signal peptide and is post-translationally processed to a 32-kDa
mature form by removal of the N-terminal half. The mature form binds the acrosin zymogen
(proacrosin, pro-ACR) and controls when that zymogen becomes active protease.

Two literatures exist and both matter:

  1. Reproductive biology — pig, mouse, guinea-pig and boar work spanning 1994–2021 that
    established the pro-ACR binding, the processing, the tyrosine phosphorylation during
    capacitation, and the knockout phenotype.
  2. Tumour immunology — ACRBP is the cancer/testis antigen OY-TES-1 / CT23, discovered
    independently and studied almost entirely as an immunogen and as a knockdown target in
    cancer lines. This literature says essentially nothing about the protein's molecular
    function, and none of it has produced a GO annotation.

The two never merged: the human gene was named and characterised as a CT antigen, and all
of its GO annotations come from ortholog transfer out of the reproductive literature.

Human-specific evidence

The only human primary paper is the cloning/CT-antigen paper.

There is no human functional or biochemical assay of ACRBP. PE 1: Evidence at protein level in UniProt reflects proteomic detection, not a functional experiment. Every functional
GO annotation on Q8NEB7 is ISS or IEA.

Caution recorded for future work: PMID:11248070 Only one reviewed human ACRBP entry
exists; the second signal is not resolved in the current literature and should not be assumed
to be a functional paralog.

The core molecular function, from the ortholog literature

So the sign of the regulation is context-dependent — it holds pro-ACR latent inside the
acrosome and accelerates its autoactivation once conditions change. MGI captured only the
positive direction: mouse Acrbp carries GO:0016504 peptidase activator activity (ISA,
PMID:8144514, with/from pig Q29016). Human ACRBP has no molecular-function annotation at
all, which is the largest single gap in its GO record — the protein is named for an activity
GO does not record for the human gene.

Domain caution: UniProt lists SUPFAM SSF100895 Kazal-type serine protease inhibitors for the
fold. ACRBP is not reported to inhibit a protease through a Kazal mechanism; it binds the
zymogen and does not bind mature acrosin at all. The fold name must not be read as an activity.

The isoform question — and why the obvious objection fails

Mouse produces two ACRBP forms by alternative splicing, and the 2016 knockout paper partitions
the phenotype between them:

Read together, these say the acrosome-assembly half of the mouse phenotype is attributable
to ACRBP-V5, described as absent from other mammals. That is a serious prima facie objection
to the human GO:0001675 acrosome assembly ISS/IEA annotations, and I set out to mark them as
over-annotated on that basis.

The bioinformatics check refuted my own hypothesis. See
file:human/ACRBP/ACRBP-bioinformatics/RESULTS.md:

One conserved intron-5 read-through event at the sequence level, then, not a rodent invention.

But the sequence level is only half the objection, and the other half survives. The 2019
paper — same laboratory, three years after the knockout study, and a paper this review already
cites for the acrosome-reaction mechanism — states the human negative outright in its
introduction: PMID:30606959. That is an RT-PCR claim about what is transcribed, and it outranks a genome
annotation on precisely that question.

So the honest conclusion is a discrepancy, not a refutation: GENCODE annotates
ENST00000536350, and the primary literature reports that human spermatogenic cells do not make
the corresponding mRNA. Nobody has looked for the protein. The two claims have never been
reconciled, and reconciling them is the point of the GO:0001675 knowledge gap in the review.

GO:0001675 acrosome assembly is nonetheless ACCEPTed, for a reason that does not depend on the
isoform argument at all: the mouse source annotation is an IMP on a knockout of the whole gene,
which ablates both forms, so what transferred is a gene-level requirement for normal acrosome
formation rather than a V5-specific activity. UniProt's curator reading of the same full text also
credits the mature form with partially contributing to acrosomal-granule assembly.

Process note, recorded because it is the kind of error worth remembering: the first draft of this
review concluded that the isoform objection "fails and the term stands", and marked PMID:23426433
DISPUTED on the strength of the GENCODE annotation. Both were wrong, and the evidence against
them was in a cached full text this review was already citing four times — found by the PR
reviewer, not by me. A genome annotation does not outrank an expression experiment. The reference
is now VERIFIED and the conclusion is scoped to the sequence level.

Consequence for curation: UniProt Q8NEB7 has no ALTERNATIVE PRODUCTS section, so there is no
human isoform identifier to put in the review's isoform: field even though the mouse function
is isoform-partitioned. This is a concrete UniProt correction to report, and it is why every
human annotation here is left isoform-agnostic.

The two pro-ACR-binding regions map onto this cleanly: human REGION 26..106 (93.8% identical
to mouse) lies inside the propeptide and inside the V5 span — it is the V5/granule-associated
site; human REGION 319..427 (73.4% identical) lies in the mature chain — it is the
ACRBP-C/zymogen-retention site.

Downstream phenotype

Together these support a role in regulating the acrosome reaction (GO:0060046), which no
ACRBP entry in any species currently carries. Proposed as a new human annotation at ISS
strength.

WITH/FROM resolution (every accession in the GOA TSV)

Accession Identity Verdict
UniProtKB:Q29016 ACRBP_PIG, Sus scrofa ACRBP true 1:1 ortholog, 81.9% identical to human; legitimate donor
UniProtKB:Q3V140 ACRBP_MOUSE, Mus musculus Acrbp true 1:1 ortholog, 75.2% identical to human; legitimate donor
MGI:MGI:1859515 mouse Acrbp same gene as Q3V140
PANTHER:PTN001085565 ACRBP node of PTHR21362 family has one member per mammalian genome; no paralog expansion
ensembl:ENSMUSP00000085632 mouse Acrbp protein same gene as Q3V140
UniProtKB-SubCell:SL-0007 "Acrosome" maps to GO:0001669
UniProtKB-SubCell:SL-0243 "Secreted" maps to GO:0005576

No paralog transfer anywhere. PANTHER PTHR21362 has 904 members across 1596 taxa but only one
reviewed protein per mammalian genome (pig, mouse, rat, guinea pig, human), so
WRONG_ORTHOLOG_OR_PARALOG is not available as a failure mode for any of these rows.

The IBA (GO_REF:0000033) is not self-referential: its with/from cites mouse and pig, not
human ACRBP. It is the PAN-GO annotation counted by UniProt's PAN-GO; Q8NEB7; 1 GO annotation
cross-reference.

Family-wide GO state (QuickGO, checked for this review)

Annotations on all five reviewed ACRBP orthologs, to establish what is and is not already
recorded before proposing anything new:

Species MF annotations Experimental (non-ISS/ISO/IBA/IEA) rows
human Q8NEB7 none GO:0005634 HDA only
mouse Q3V140 GO:0016504 (ISA), GO:0005515 (IPI) GO:0001675/0007286/0009566 IMP, GO:0001669 IDA
pig Q29016 GO:0005515 (IPI ×2, partner P08001 = pig acrosin) GO:0001669 IDA, GO:0002080 IDA, GO:0005576 IDA+EXP, GO:0048240 TAS
rat Q6AY33 none none — all 16 rows are ISS/ISO/IBA/IEA
guinea pig Q60485 none none — all 10 rows are ISS/IEA

So the entire experimental base for this family is mouse and pig, and four of the five entries
have no molecular function at all. Direct queries confirm that GO:0035375 zymogen binding,
GO:0061135 endopeptidase regulator activity, GO:0097341 zymogen inhibition and GO:0060046
regulation of acrosome reaction each return zero annotations across all five orthologs — so
all four proposed rows are genuinely new to the family, not duplicates of something already
recorded elsewhere in it.

Two further points from that table. The mouse GO:0005515 IPI (PMID:11164898) cites Q08279
(guinea-pig IGF2) and Q6UW60 (human PCSK4) as partners, neither of which is proacrosin; PCSK4 is
the convertase of PMID:22357636, and the IGF2 entry looks like a with/from artefact. Neither was
used here. And the single pig GO:0002080 IDA has propagated by ISS
(GO_REF:0000024, with/from UniProtKB:Q29016) to four entries — human Q8NEB7, mouse Q3V140,
rat Q6AY33 and guinea pig Q60485 — plus a further rat ISO row (GO_REF:0000121, with/from
RGD:14255780). So the MODIFY argued below is not a one-gene correction: one
resolution-limited immunofluorescence call is currently the sole basis for acrosomal-membrane
localisation across the whole reviewed family.

Annotations that do not survive review

GO:0002080 acrosomal membrane (ISS from pig Q29016) — traced to pig's own
GO:0002080 ... IDA PMID:15955892, an indirect-immunofluorescence localisation. Light
microscopy of the acrosome cannot separate the acrosomal matrix from the acrosomal membrane.
The protein has no basis for residing in a bilayer: all five reviewed orthologs have a cleaved
signal peptide and no transmembrane, intramembrane or lipidation feature, and sp32 was
originally purified as a soluble protein from acid extracts of ejaculated sperm
PMID:8144514. GO:0043159 acrosomal matrix is both more accurate and more
informative, and is the term the primary literature actually uses
PMID:11248070.
→ MODIFY to GO:0043159.

GO:0005634 nucleus (HDA, PMID:21630459) — from a human sperm-nucleus proteome. The
authors claim high purity, PMID:21630459 but ACRBP is a
signal-peptide-bearing secretory-pathway protein and the acrosome physically caps the sperm
nucleus; carry-over is the parsimonious reading. The dataset is also broad —
PMID:21630459 and it reports "zinc fingers and transcription
factors, so far not known to be associated with the sperm chromatin". The same PMID has been
handled as an over-annotation for other genes in this repository (AAAS, ACADS, ACADM).
→ MARK_AS_OVER_ANNOTATED, not REMOVE: the peptide detection itself is not in dispute.

Process notes