ACRBP

Summary: Accepted as the core location. This is the PAN-GO annotation counted by UniProt's "PAN-GO; Q8NEB7; 1 GO annotation based on evolutionary models" cross-reference, and it is the only annotation on this gene that uses is_active_in rather than located_in - the stronger claim that the protein does its job here, not merely that it is found here. That claim is right: everything ACRBP is known to do, it does inside the acrosome. The transfer is clean. Both gene-product donors are true 1:1 orthologs with experimental acrosome localisation of their own - mouse Acrbp (MGI:MGI:1859515 / Q3V140) carries GO:0001669 IDA on PMID:22357636, and pig ACRBP (Q29016) carries GO:0001669 IDA on PMID:8144514. The PANTHER node PTHR21362 has one reviewed member per mammalian genome (human, mouse, rat, guinea pig, pig), so there is no paralog for the annotation to have leaked across. Note that the with/from does not cite human ACRBP itself, so this is not a self-referential IBA.

Summary: Accepted. Mechanically redundant with the IBA above and with the UniProt ISS below - three pipelines independently reaching the same correct compartment - but nothing here is wrong. The with/from is the mouse ortholog plus the UniProt subcellular-location keyword SL-0007 "Acrosome", and that keyword itself derives from the experimental mouse and pig localisations rather than from prediction.

Summary: Kept, but as a non-core location. The keyword mapping is faithful - UniProt does record "Secreted" for Q8NEB7 - and the underlying biology is real rather than an artefact: the acrosome is a secretory, lysosome-derived vesicle, and its contents are discharged to the outside of the sperm during the acrosome reaction. In pig this has been observed directly, with acrosomal sp32 labelling disappearing after an induced acrosome reaction, and sp32 being recoverable from ejaculated sperm. It is non-core for two reasons. GO:0005576 is a very general term that says only "outside the cell", which for a protein whose entire working life is spent inside one organelle is the least informative thing one can say about it; and the extracellular state is terminal and transient - it is where the protein ends up after the compartment it functions in has been discharged, not where it acts.

Reason: Faithful keyword mapping of a genuine post-exocytosis fate, but an uninformative parent term describing the aftermath of ACRBP's function rather than its site of action. Retained for completeness and excluded from core_functions.locations.

Summary: Accepted, after a specific objection to it was tested and failed. The source is strong: mouse Acrbp carries GO:0001675 IMP on PMID:27303034, a knockout in which spermatids fail to form the large acrosomal granule and the acrosome ends up fragmented. The knockout removes the whole gene, so the mouse annotation is gene-level. The objection is that the 2016 paper partitions the phenotype between two mouse splice forms and assigns the granule-forming role specifically to ACRBP-V5, which the same group had described as a variant not found in other mammalian ACRBPs. If human ACRBP made nothing equivalent to ACRBP-V5, the transferred process would have no vehicle in human and this row would be an over-annotation. That was the working hypothesis for this review. The objection has two halves, and they come apart. At the level of genome sequence it fails: the human ACRBP locus carries an annotated protein-coding transcript, ACRBP-204 / ENST00000536350, whose exon chain is the first five canonical exons with the terminal exon extended 82 bp past the exon-5 donor site into intron 5 - the same architecture as the mouse ACRBP-V5 transcript Acrbp-205, whose terminal exon extends 242 bp into intron 5. The consequence on the protein is the same substitution at the equivalent position: human SLLQL to RYRKF at 315-319, mouse SLQQL to RYRKL at 312-316, the latter being exactly UniProt's VSP_051965/VSP_051966 for ACRBP-V5. So the human genome does encode a structural counterpart of ACRBP-V5, and the splicing event is not a rodent invention. At the level of expression the objection stands, and more firmly than the 2013 paper alone implies. The same laboratory states the human negative outright three years after the knockout paper: porcine, guinea pig and human spermatogenic cells produce only the Acrbp-W mRNA, while mouse makes both. That is an RT-PCR claim about what is transcribed, and it outranks a genome annotation on exactly that question. The honest summary is therefore a discrepancy rather than a refutation - GENCODE annotates a transcript the primary literature reports as absent from human spermatogenic cells, and nobody has looked for its protein. The term is still accepted, but for a reason independent of the isoform argument: the mouse source is an IMP on a knockout of the whole gene, ablating 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 credits the mature form with partially contributing to acrosomal-granule assembly. The residual uncertainty is recorded as a knowledge gap, not discharged by this row.

Summary: Kept as non-core. The mouse source (GO:0007286 IMP, PMID:27303034) is sound - ACRBP-null spermatids fail to build a normal acrosome - and the transfer to human is legitimate for the same 1:1 orthology as the rows above. It is non-core because it names the developmental window rather than the activity. Acrosome biogenesis happens during spermiogenesis, so any gene involved in building the acrosome is involved in spermatid development by construction; the term adds staging information but no mechanism, and GO:0001675 acrosome assembly already says the specific thing.

Reason: True but non-specific: the term captures the developmental context of the acrosome-assembly role rather than a distinct function. Retained; GO:0001675 carries the informative content.

Summary: Accepted. The mouse source is GO:0009566 IMP on PMID:27303034, and the mechanism behind that phenotype was subsequently pinned down: ACRBP-deficient sperm reach the oviduct with normal head morphology and normal motility but cannot gain access to unfertilized oocytes, which the authors attribute to an incomplete acrosome reaction rather than to a migration defect. In boar, antibody blocking of surface ACRBP reduces both capacitation and sperm-zona pellucida binding, independently placing the protein in the fertilization sequence. GO:0009566 is a broad term, but for this gene the breadth is not the problem it usually is: ACRBP's proacrosin-timing function exists in order to be used at fertilization, so involved_in fertilization is the honest summary process even though the more specific step - regulating the acrosome reaction - is proposed separately below rather than substituted here.

Summary: Kept as non-core, on the same reasoning as the GO_REF:0000044 row above. This is the curator-judged version of the same statement, transferred from pig Q29016, which holds GO:0005576 by IDA and by EXP on the two pig papers. The pig evidence is direct, and it is about the discharged state of the protein rather than its working compartment. Note that this exact annotation appears twice in the GOA table for this gene - same term, same evidence code, same GO_REF, same with/from, dated 2016-05-24 and 2020-02-06. The two rows are one annotation and are reviewed here once.

Reason: Genuine but uninformative and terminal: describes where the protein goes once the acrosome has been discharged, not where it functions. Duplicated in GOA; reviewed once.

Summary: Kept as non-core, identically to the GO_REF:0000107 row for this term. This is the manually-judged transfer from mouse Q3V140; the Ensembl Compara row is the automated version of the same inference. Both are defensible and both name the developmental window rather than the activity.

Reason: Same judgement as the Compara IEA row for this term: correct, contextual, and superseded in informativeness by GO:0001675 acrosome assembly.

Summary: Accepted, on the same evidence as the GO_REF:0000107 row for this term - this is the manual curator transfer, that one the Compara automation. The mouse knockout establishes the fertility requirement and PMID:30606959 localises the failure to the acrosome reaction.

Summary: Accepted, for the reasons set out at the GO_REF:0000107 row for this term: the mouse source is a whole-gene knockout IMP, so what transferred is a gene-level requirement, and the isoform-specificity objection resolves into a discrepancy rather than a refutation. Human ACRBP-204/ENST00000536350 is the same intron-5 read-through transcript at the sequence level, but PMID:30606959 reports experimentally that human spermatogenic cells make only the Acrbp-W mRNA - so the structural counterpart exists in the genome annotation while the primary literature says it is not transcribed. One additional point belongs on this manual row rather than on the automated one. UniProt's own FUNCTION comment for the mature form hedges the granule role ("Partially also contributes to the assembly of acrosomal proteins to form an acrosomal granule"), which is a fair reading: the 2016 rescue experiment shows ACRBP-V5 alone suffices, not that the mature form contributes nothing. The two annotated pro-ACR-binding regions of human ACRBP sit on opposite sides of the maturation cleavage - region 26-106 inside the removed propeptide and inside the V5-equivalent span, region 319-427 in the retained mature chain - so the granule role and the zymogen-retention role are carried by physically different parts of the precursor. That is a reason to keep the process annotation while proposing the molecular functions separately, not a reason to weaken it.

Summary: Accepted. Third pipeline reaching the correct compartment; this is the manual transfer from mouse Q3V140, whose acrosome localisation is experimental (IDA, PMID:22357636) and was also shown by immunofluorescence to concentrate in the acrosomal granules of round spermatids.

Summary: Modified to GO:0043159 acrosomal matrix. The compartment is right at organelle level and wrong at sub-organelle level, and a better term exists. Traced back, this row transfers pig Q29016's GO:0002080 located_in, whose evidence is an IDA on PMID:15955892 - indirect immunofluorescence of pig sperm. Light microscopy of an organelle a few hundred nanometres across cannot separate the acrosomal matrix from the acrosomal membrane that bounds it, so the pig observation supports "in the acrosome" but not specifically "in the membrane". That is a resolution limit of the assay, not a curation error. What settles it is that ACRBP has no way to be in a bilayer. All five reviewed orthologs carry a cleaved signal peptide and none carries a transmembrane, intramembrane or lipidation feature; no mature sequence in the family reaches the hydrophobicity a membrane-spanning helix would need. The protein is translocated into the secretory pathway and stays soluble - which is how sp32 came to be purified in the first place, from acid extracts of ejaculated sperm. located_in GO:0002080 asserts residence in the membrane itself, and nothing supports that. GO:0043159 acrosomal matrix ("a structural framework, or 'dense core' at the interior of an acrosome") is the term the primary literature actually uses for this protein, is part_of GO:0001669, and captures the functionally relevant point: ACRBP is in the dense core with the proacrosin it packages and holds inactive. The same correction probably applies at the pig source, but that is for the pig entry's curator to judge from the full text.

Reason: The organelle is correct but the sub-organelle compartment is not supported: an anchor-free, signal-peptide-bearing secretory protein cannot be located_in a membrane, and the pig IDA that this row transfers is an immunofluorescence localisation that cannot resolve matrix from membrane. GO:0043159 acrosomal matrix is both better supported and more informative, and is the description used in the primary literature.

Proposed replacements: acrosomal matrix

Summary: Marked as over-annotated. The peptide detection is not in dispute; the compartment inference drawn from it is. The evidence is a single high-throughput proteome of human sperm nuclei that identified 403 proteins. The authors report a stringent preparation, isolated to over 99.9% purity with no tail fragments, acrosome or mitochondria, but ACRBP is exactly the protein that preparation would struggle to exclude: the acrosome is a flattened vesicle apposed to the anterior face of the sperm nucleus, and in mouse ACRBP has been described as spread over the apical end of the nucleus in elongating spermatids. Detection of an acrosomal matrix protein in a sperm-nucleus fraction is most simply read as co-purification of the structure that caps the nucleus. Topology says the same thing independently. ACRBP has a cleaved signal peptide and is translocated into the secretory pathway at synthesis; it has no route to the nucleoplasm. The dataset is also permissive by the authors' own account - more than half its proteins were absent from previous sperm proteomes, and it reports transcription factors and zinc fingers not previously known to associate with sperm chromatin. This same PMID has been treated as an over-annotation for other genes in this repository.

Reason: High-throughput detection in a sperm-nucleus fraction, read as a location. A signal-peptide-bearing secretory-pathway protein cannot reside in the nucleoplasm, and the acrosome is physically apposed to the sperm nucleus, so carry-over is the parsimonious explanation. Retained rather than removed because the mass-spectrometric identification itself is credible.

Summary: Proposed new annotation, and the more important half of the largest gap in this gene's GO record: human ACRBP currently has no molecular-function annotation of any kind, even though the protein is named for what it binds. The binding is not generic and should not be recorded as protein binding. Purified pig sp32 binds the 55-, 53- and 49-kDa (pro)acrosin species and specifically does not bind the 43-kDa intermediate or the 35-kDa mature enzyme - it discriminates the zymogen and early intermediates from the finished protease. Mouse work reproduces this and localises the contact to two segments in the C-terminal region of pro-ACR, engaged by different regions of ACRBP. GO:0035375 zymogen binding (a child of GO:0019899 enzyme binding) states exactly this specificity. Evidence code is ISS, not IDA: the binding assays are on the pig and mouse proteins, and human ACRBP is 81.9% identical to pig and 75.2% to mouse over the precursor, with the C-terminal pro-ACR-binding region present in the human mature chain. MGI recorded the mouse counterpart of this activity as GO:0016504 peptidase activator activity (ISA, PMID:8144514, with/from Q29016); nothing was ever propagated to human.

Reason: Fills the complete absence of molecular-function annotation on human ACRBP with the specific, well-documented binding activity the protein is named for, at the appropriate ortholog-transfer evidence strength. GO:0035375 is used in preference to GO:0005515 protein binding because the documented specificity - zymogen and early intermediates but not the mature protease - is the informative part.

Summary: Proposed new annotation: the regulatory consequence of the zymogen binding, recorded at a sign-neutral level because the evidence genuinely points both ways. In the intact acrosome ACRBP holds proacrosin latent - exogenously expressed ACRBP-W blocks proacrosin autoactivation, and the authors state plainly that retaining the inactive status of proacrosin until acrosomal exocytosis is its major function. In vitro, at basic pH, the same protein accelerates proacrosin autoactivation and redirects the maturation pathway so that the 49-kDa intermediate accumulates instead of the 43-kDa one; mouse ACRBP-C reproduces the acceleration. So ACRBP is not an inhibitor of acrosin and not an activator of it: it sets the timing of zymogen conversion, holding it back in one context and speeding it in another. One thing the term name does not say, and which should be stated outright: what ACRBP holds is the early part of the activation pathway, and the mature endopeptidase is specifically excluded. It binds the 55- and 53-kDa proacrosins and the 49-kDa acrosin intermediate, and does not bind the 43-kDa intermediate or the 35-kDa mature acrosin. The discriminant is therefore early versus late species along the processing route rather than zymogen versus active enzyme - the paper's own title is "a binding protein specific for two proacrosins and an acrosin intermediate". That distinction cuts both ways, and both matter here. It is why GO:0035375 zymogen binding is recorded alongside this row rather than subsumed into it: the zymogen is where the interaction begins and where the physiological restraint is exerted. And it is why GO:0061135 is apt rather than strained: the 49-kDa species is an acrosin intermediate, that is, a post-autoactivation form generated from the 55- and 53-kDa proacrosins, so ACRBP does contact a processed enzyme species, which is what an enzyme-regulator term requires. That inference rests on the intermediate being downstream of autoactivation, which the source establishes; it does not rest on a measurement of the intermediate's catalytic activity, which the source does not report. Nothing here should be read as inhibition of the finished protease, which the binding set above excludes. It is also why GO:0061135 is proposed rather than its children GO:0004866 endopeptidase inhibitor activity or GO:0061133 endopeptidase activator activity. MGI chose the activator direction for mouse (GO:0016504, ISA on PMID:8144514), which captures the in vitro result but is contradicted in sign by the in vivo result from the same laboratory twenty-two years later. The unsigned parent is the strongest claim both datasets support. A trap to note: UniProt assigns ACRBP the SUPFAM Kazal-type serine protease inhibitor fold (SSF100895). There is no evidence ACRBP inhibits a protease by a Kazal mechanism - it does not bind mature acrosin at all - so the fold name must not be promoted to an activity.

Reason: Records the activity that ACRBP's zymogen binding accomplishes, without asserting a direction the evidence does not support. The interaction is stage-selective along the activation pathway - the 55-/53-kDa proacrosins and the 49-kDa acrosin intermediate are bound, the 43-kDa intermediate and the 35-kDa mature enzyme are not - so this row is paired with GO:0035375 rather than replacing it: one states what is bound, the other what the binding does to the activity that follows from it. ISS because all assays are on pig and mouse protein.

Summary: Proposed new annotation. GO already has a term for precisely the process the 2016 knockout paper describes - GO:0097341 zymogen inhibition, "any process that prevents the proteolytic processing of an inactive enzyme to an active form", with the synonym "prevention of zymogen activation" - and no ACRBP entry in any species carries it. This is the biological-process counterpart of the molecular function above, and it is the half of ACRBP's activity that operates inside the intact acrosome: keeping proacrosin from autoactivating prematurely, so that active acrosin appears only when the acrosome is discharged. Its physiological importance is visible in the PCSK4-null mouse, where ACRBP fails to be processed to its mature form and proacrosin then fails to autoactivate at all - the mature protein is required for normal proacrosin conversion. Proposed at ISS strength from mouse; the sign here is well determined in vivo, unlike the molecular-function row, because the in vitro acceleration is a property of the isolated system at basic pH rather than of the acrosome.

Reason: An exact, already-existing GO term for ACRBP's in vivo role that is unused across the whole family. Sign-specific at the process level, where the in vivo evidence is unambiguous, while the molecular-function row stays neutral.

Summary: Proposed new annotation, and the specific step that the broad GO:0009566 fertilization annotation leaves unnamed. No ACRBP entry in any species carries a term for the acrosome reaction. Two independent lines of evidence, from two species and by opposite experimental logic. Loss of function in mouse: ACRBP-deficient sperm arrive in the oviduct with normal head morphology and normal motility, yet cannot reach unfertilized oocytes, and the authors attribute the subfertility to incompleteness of the acrosome reaction rather than to impaired migration. Acute blockade in boar: antibodies against ACRBP on the sperm head reduce spontaneous acrosome reaction and reduce the ability of sperm to react to solubilized zona pellucida or to SERCA inhibition. Regulation, not participation, is the right framing: ACRBP does not execute membrane fusion, it determines whether the protease cargo is ready when fusion occurs. Proposed at ISS strength - mouse for the genetic evidence, pig for the antibody-blocking evidence, with no human experiment on either side.

Reason: Names the specific fertilization step ACRBP acts on, supported by loss-of-function evidence in mouse and antibody blockade in boar, and absent from every ACRBP entry in GO. Proposed as additive to GO:0009566 rather than as a replacement, since the broad term is separately supported by the mouse knockout.