AEBP2 is a nuclear C2H2 zinc-finger protein that acts as an accessory subunit of Polycomb repressive complex 2 (PRC2), the histone methyltransferase that mono-, di- and trimethylates lysine 27 of histone H3. Together with JARID2 it defines the PRC2.2 holo-complex, which is mutually exclusive with the PCL/PALI1-containing PRC2.1 assembly. AEBP2 docks through its C-terminal region onto the non-canonical C2 domain of SUZ12, competing there with PHF19 and repositioning that domain; its extreme C-terminus extends into the histone-H3-tail pocket of RBBP4, where its own Lys502 and Arg503 occupy the site normally used by H3 Lys4 and Arg2, so that H3K4 recognition by the complex is progressively excluded. The same C-terminal segment contributes to the nucleosome-binding activity of the assembled complex, and on H2AK119-ubiquitylated nucleosomes AEBP2 and JARID2 each contact a ubiquitin moiety and the H2A-H2B surface, coupling PRC1-deposited H2A ubiquitylation to H3K27 methylation. The gene produces developmentally regulated long and short protein forms with opposite effects on the complex. The long form carries a vertebrate-specific acidic N-terminal extension and is the form present in essentially all human tissues; complexes containing it are almost inactive as histone methyltransferases, below the activity of the core complex alone, they gain none of the DNA-binding affinity the short form confers, and in cells they lower the amount of core complex on chromatin below that of cells expressing no AEBP2 at all. The short form lacks that extension, is in humans largely confined to testis, raises the complex's affinity for DNA by roughly ninetyfold and is required for de novo silencing of Polycomb target genes during the exit from naive pluripotency. In mouse the gene is essential for embryogenesis and is expressed mainly in neural-crest-derived cells, and its loss produces a Trithorax-like phenotype with increased H3K27 methylation at Polycomb targets. The protein localises to chromatin at PRC2 target loci, including the inactive X chromosome.
Definition: A Polycomb repressive complex 2 holo-complex containing the core subunits EED, EZH1 or EZH2, SUZ12 and RBBP4 or RBBP7 together with the accessory subunits AEBP2 and JARID2, and lacking the Polycomb-like and PALI1 accessory subunits characteristic of PRC2.1.
Justification: GO:0035098 has no children at all, so the ontology cannot express the PRC2.1/PRC2.2 distinction even though it is an organising fact of current PRC2 biology: the two assemblies are mutually exclusive, use different recruitment logic, and their balance is required for correct regulation of Polycomb targets during differentiation. ComplexPortal already models the distinction, with CPX-2209, CPX-2212, CPX-2213 and CPX-2330 all named as PRC2.2 variants and all listed as AEBP2's complexes in UniProt. Without child terms, an AEBP2 or JARID2 annotation is indistinguishable from a PHF19 or MTF2 annotation, and the subcomplex-specific literature cannot be captured. A sibling PRC2.1 term would be needed at the same time.
Parent term: ESC/E(Z) complex
Supporting Evidence:
Definition: Binds to and increases the activity of a histone methyltransferase.
Justification: A specificity gap rather than a blocking one, and an asymmetry in the ontology rather than an oversight in any one annotation. GO has GO:0180000 histone methyltransferase inhibitor activity for the negative direction, and GO:0010698 acetyltransferase activator activity for the analogous positive direction on a different class of histone-modifying enzyme, but the only methyltransferase activator term is GO:0141106, restricted to tRNA methyltransferases. The consequence is visible inside this complex: EED and SUZ12 both carry the generic parent GO:0008047 enzyme activator activity for their stimulation of EZH2, and any annotation of the short isoform of AEBP2 could only be made the same way. Because the direction of AEBP2's own effect is disputed, this term is proposed for the ontology's benefit rather than for use on this gene, and no activator annotation is proposed for AEBP2 here. Note that the molecular-function route is the only one GO now offers for this kind of claim: the whole biological-process branch for regulating histone methylation has been obsoleted without replacement - GO:0016571 histone methylation, GO:0031056 regulation of histone modification, GO:0031060/0031061/0031062 regulation of histone methylation, and the H3-K27-specific GO:0061085/0061086/0061087 are all obsolete and none carries a term_replaced_by. A search of live biological_process terms returns no non-obsolete regulation-of-histone-methylation term, controlled against a search for heterochromatin formation that does return live terms. So a process-level term for restraining H3K27 methylation cannot be requested in a branch GO has deliberately dismantled, and the inhibitor and activator molecular functions are where the claim belongs.
Parent term: enzyme activator activity
Supporting Evidence:
| GO Term | Evidence | Action | Reason |
|---|---|---|---|
| GO:0035098 ESC/E(Z) complex | IBA GO_REF:0000033 | ACCEPT | Summary: Correct and core. AEBP2 is an established component of the PRC2.2 holo-complex, and this phylogenetic propagation rests on a mouse ortholog carrying three independent IDA rows plus a self-reference recording a PAINT curator's judgement that the function is core. Reason: All three WITH/FROM tokens were resolved from the GOA field rather than by hand. MGI:MGI:1338038 is mouse Aebp2 UniProtKB:Q9Z248, reviewed, 504 aa; the bare number had to be used because a query containing the inner colon returns HTTP 400, and three additional TrEMBL candidates were returned and are reported rather than hidden. Queried for its own evidence, that donor carries GO:0035098 by IDA three times over, from PMID:20064375, PMID:20064376 and PMID:31451685; the EZH2 positive control returns a non-zero from the same query pattern, so this is a real count. UniProtKB:Q6ZN18 is a self-reference, which is valid for an IBA and is recorded as NO_FAILURE_CORE rather than as circularity. PANTHER:PTN002323211 is a tree node, not a protein. Reciprocally, the node reaches 117 recipients and gives GO:0035098 to only 79 of them, so it is withholding the animal complex term from 38 recipients rather than propagating it indiscriminately. Propagation Review Root cause: NO FAILURE CORE Sources checked: MGI:MGI:1338038 · mouse Aebp2 (UniProtKB:Q9Z248, reviewed, 504 aa) SUPPORTS TRANSFER Carries GO:0035098 by IDA from three independent references (PMID:20064375, PMID:20064376, PMID:31451685). True ortholog, not a paralog. PANTHER:PTN002323211 · PANTHER tree node (117 recipients, 79 of which receive GO:0035098) SUPPORTS TRANSFER An internal tree node rather than a protein. Its reach is not indiscriminate: 38 of its 117 recipients receive GO:0006357 without GO:0035098. UniProtKB:Q6ZN18 · AEBP2 itself (self-reference) SUPPORTS TRANSFER A self-referential WITH/FROM on an IBA records a PAINT curator judging the function core for this gene. Not circular propagation. Supporting Evidence: file:human/AEBP2/AEBP2-bioinformatics/RESULTS.md | mouse Aebp2 | GO:0035098 | 6 | IDA GO:0035098 PMID:20064375; IDA GO:0035098 PMID:20064376; IDA GO:0035098 PMID:31451685 | file:human/AEBP2/AEBP2-uniprot.txt monomeric PRC2.2 (class 2) complex consisting of at least SUZ12, RBBP4, |
| GO:0035098 ESC/E(Z) complex | IEA GO_REF:0000120 | ACCEPT | Summary: The term is correct and independently established by an IDA, two IPI rows and an IBA on this gene, so this row is accepted. Its inference route is nonetheless worth recording: the named ARBA rule grants membership of a specific multiprotein complex from structural-family membership plus taxon, inspecting no interaction evidence. Reason: Every token was resolved, which is the point of a combinatorial reference: two of the three reduce to the same entity, because ensembl:ENSMUSP00000084896 is UniProtKB:Q9Z248, mouse Aebp2. So this row records one protein counted twice plus one rule, not three independent pipelines agreeing. Fetching the rule shows ARBA00089504 asserts exactly one annotation, GO:0035098, from 8 alternative condition sets, each a conjunction of FunFam identifiers and a taxon clause with no residue, interaction or assay condition anywhere in it. The set that fires on AEBP2 requires FunFam 3.30.160.60:FF:000323 AND FunFam 3.30.160.60:FF:000471 AND Eukaryota AND NOT Ascomycota, and AEBP2's own cross-references are exactly those two FunFams. Requiring two co-occurring AEBP2-specific FunFams is what keeps the rule from over-reaching, and the remaining seven condition sets key on other PRC2 subunits' folds (2.170.270.10 for the SET-domain subunits, 2.130.10.10 for the WD40 subunits), each with its own taxon restriction. Since the conclusion is true and separately demonstrated, this is documented rather than filed as a rule defect. Propagation Review Root cause: NO FAILURE CORE Failure modes: SOURCE EVIDENCE WEAK Sources checked: ARBA:ARBA00089504 · ARBA rule asserting GO:0035098 from FunFam identifiers plus taxon SOURCE WEAK OR INFERRED The rule's entire condition set is structural-family membership and taxon. It happens to reach the right answer for AEBP2 because it demands two AEBP2-specific FunFams simultaneously, but it inspects no evidence that the protein is in the complex. UniProtKB:Q9Z248 · mouse Aebp2 (reviewed, 504 aa) SUPPORTS TRANSFER Carries GO:0035098 by IDA from three independent references. ensembl:ENSMUSP00000084896 · mouse Aebp2 - the same entity as UniProtKB:Q9Z248 CIRCULAR OR REDUNDANT Not an independent witness. This Ensembl protein identifier resolves to Q6ZN18's mouse ortholog Q9Z248, the token listed alongside it, so the reference records one source twice. The transfer itself is supported; it is the appearance of independent agreement that is not. Supporting Evidence: file:human/AEBP2/AEBP2-bioinformatics/RESULTS.md | [3] (fires on Q6ZN18) | `['3.30.160.60:FF:000323', '3.30.160.60:FF:000471']` | `['Eukaryota', 'NOT Ascomycota']` | |
| GO:0035098 ESC/E(Z) complex | IPI PMID:33514705 Structures of monomeric and dimeric PRC2:EZH1 reveal flexibl... | ACCEPT | Summary: Sound per-protein evidence of complex membership from a structural study of PRC2-EZH1. Unlike the three NAS rows this reference also supplies, the GO:0035098 IPI rows are assigned per subunit and not uniformly across every entity the paper touches. Reason: The projection test distinguishes this row from its NAS neighbours. The reference annotates nine entities in total, two of them ComplexPortal complexes, and gives all nine an identical NAS term set; but GO:0035098 IPI goes to six entities and not to the seventh protein (RBBP7), which is the signature of per-subunit curation rather than of a blanket projection. UniProt independently records the same assembly. Supporting Evidence: file:human/AEBP2/AEBP2-uniprot.txt monomeric PRC2.2 (class 2) complex consisting of at least SUZ12, RBBP4, PMID:33514705 PRC2 plus AEBP2 and JARID2 |
| GO:0035098 ESC/E(Z) complex | IPI PMID:29348366 Structures of human PRC2 with its cofactors AEBP2 and JARID2... | ACCEPT | Summary: Sound per-protein evidence of complex membership from the cryo-EM study that defined the architecture of PRC2 with AEBP2 and JARID2. Reason: The same discrimination applies as for the PMID:33514705 IPI row: this reference gives an identical NAS term set to all nine entities it annotates, but GO:0035098 IPI to six entities only. The membership conclusion is also isoform-independent, which matters because the paper's own construct covers Q6ZN18 residues 209-503 only: the SUZ12-binding region 407-478 lies inside that span and is present in all three isoforms. Supporting Evidence: PMID:29348366 JARID2 and AEBP2. Both cofactors mimic the binding of histone H3 tails. JARID2, file:human/AEBP2/AEBP2-bioinformatics/RESULTS.md | 6C23 | 2018-01-24 | PMID:29348366 | 209–503 | |
| GO:0035098 ESC/E(Z) complex | IDA PMID:20075857 JARID2 regulates binding of the Polycomb repressive complex ... | ACCEPT | Summary: The oldest direct evidence in this set that AEBP2 co-purifies with PRC2 in embryonic stem cells. Accepted, with the limitation that the cached record is abstract-only and the abstract is about JARID2. Reason: The abstract of this reference does not mention AEBP2, so the assignment rests on full text I have not read; per the rule against second-guessing an experimental annotation on the strength of a title, that is not grounds to doubt it. The projection test supports the curator instead: the reference gives GO:0035098 IDA to eight PRC2 components, all assigned by UniProt, while its only functional row, GO:0048863 IMP, stays on mouse Jarid2 alone - the gene actually perturbed. A phenotype confined to the perturbed gene while the composition term spreads is per-protein curation, not a complex-level projection. Supporting Evidence: PMID:20075857 PRC2 forms a stable complex with the Jumonji- and ARID-domain-containing file:human/AEBP2/AEBP2-bioinformatics/RESULTS.md | `PMID:20075857` | 23 | 12 | 0 | `{'IntAct': 12, 'UniProt': 11}` | 1 terms × 5 entities | |
| GO:0006357 regulation of transcription by RNA polymerase II | IBA GO_REF:0000033 | ACCEPT | Summary: Correct, and correct precisely because it is general. The two donor organisms disagree on the sign of the regulation, so this direction-neutral parent is their lowest common ancestor, and it is also the right level for a gene whose two protein isoforms push transcription in opposite directions. Reason: Asking which term each donor holds rather than merely whether it holds one is what settles this row. Mouse Aebp2 (MGI:MGI:1338038 = UniProtKB:Q9Z248, reviewed) carries GO:0000122 by IMP; Drosophila jing (FB:FBgn0086655 = UniProtKB:Q7KHG2, reviewed, 1486 aa, the fly member of the AEBP2/jing family) carries GO:0000122 by IMP and GO:0045944 by IMP. The donor set is therefore split on direction, GO:0006357 is exactly its LCA, and GRANULARITY_MISMATCH is inapplicable because that failure mode requires the donors to agree. This is independently the right granularity for the human gene: the broadly expressed long isoform suppresses PRC2 while the short isoform promotes it, so no directional child is true of the gene as a whole. The committed analysis asserts the donor disagreement rather than stating it, and fails if it ever stops holding. Propagation Review Root cause: NO FAILURE CORE Sources checked: MGI:MGI:1338038 · mouse Aebp2 (UniProtKB:Q9Z248, reviewed, 504 aa) SUPPORTS TRANSFER Carries GO:0000122, a descendant of the propagated term, by IMP (PMID:10329662). FB:FBgn0086655 · Drosophila jing (UniProtKB:Q7KHG2, reviewed, 1486 aa) SUPPORTS TRANSFER Carries GO:0000122 IMP and GO:0045944 IMP - opposite directions from different references, so no citation cross-product. The fly member of the same family per UniProt's SIMILARITY line, so an orthology-strength inference is available. PANTHER:PTN002323211 · PANTHER tree node (117 recipients, all of which receive this term) SUPPORTS TRANSFER A tree node, not a protein. All 117 recipients receive GO:0006357, which is consistent with a term that is true family-wide, whereas only 79 receive the animal complex term. Supporting Evidence: file:human/AEBP2/AEBP2-bioinformatics/RESULTS.md | Drosophila jing | GO:0006357 | 4 | IMP GO:0000122 PMID:16510782; IMP GO:0045944 PMID:21061018 | PMID:41168462 Taken together, these data suggest that AEBP2S and AEBP2L perform opposing functions in regulating PRC2 localisation on chromatin: ectopic expression of AEBP2S increases the chromatin occupancy of PRC2, whereas AEBP2L reduces it. |
| GO:0000122 negative regulation of transcription by RNA polymerase II | IEA GO_REF:0000107 | KEEP AS NON CORE | Summary: Retained, but not as the gene's core process. The term is supported for the short isoform and in the PRC2.2 context, while the broadly expressed long isoform has been shown to do the opposite; and the underlying ortholog evidence is sufficiency-type over-expression, not a requirement. Reason: Both WITH/FROM tokens resolve to the same entity, mouse Aebp2 Q9Z248, so this is a single-donor transfer. That donor's GO:0000122 IMP comes from PMID:10329662, whose repression evidence is co-transfection over-expression at the aP2 promoter plus a Gal4-tethering experiment; GO's evidence codes do not distinguish sufficiency from requirement, so it is stated here that this is sufficiency. Against it, complexes containing the human protein's dominant isoform are almost inactive as methyltransferases and gain none of the DNA-binding affinity the short isoform confers, loss of that isoform raises PRC2 occupancy and H3K27me3 at target genes, and an independent group reported the long isoform activating and the short isoform repressing the same target. The term is therefore not wrong - it is one direction of a gene that is annotated at the right level by its direction-neutral parent GO:0006357 - and is kept as non-core rather than accepted as core or removed. Propagation Review Root cause: NO FAILURE NON CORE Failure modes: REGULATORY SIGN INVERSION Sources checked: UniProtKB:Q9Z248 · mouse Aebp2 (reviewed, 504 aa) SUPPORTS SOURCE BUT NOT TARGET Holds GO:0000122 by IMP from PMID:10329662, and the annotation is sound for the donor. What it cannot carry is the direction for the human gene as a whole, because that depends on which isoform is present, and in human tissues the predominant isoform restrains PRC2. ensembl:ENSMUSP00000084896 · mouse Aebp2 - the same entity as UniProtKB:Q9Z248 CIRCULAR OR REDUNDANT Resolves to Q9Z248, so this row has one donor, not two. Supporting Evidence: PMID:10329662 Moreover, a chimeric construct encoding a fusion AEBP2 protein with the Gal4 PMID:41168462 While core PRC2 alone and PRC2–AEBP2S exhibited comparable HMTase activities, PRC2–AEBP2L was almost completely inactive PMID:27317809 targeted mutation of Aebp2 unexpectedly revealed a Trithorax phenotype, normally |
| GO:0000122 negative regulation of transcription by RNA polymerase II | NAS PMID:33514705 Structures of monomeric and dimeric PRC2:EZH1 reveal flexibl... | KEEP AS NON CORE | Summary: A ComplexPortal statement of what PRC2 does, projected onto every subunit. Kept as non-core, consistent with the other two GO:0000122 rows. Reason: This reference gives an identical NAS term set - GO:0000122, GO:0031507 and GO:0005634 - to all nine entities it annotates, including the two ComplexPortal complexes themselves, so for AEBP2 it records a property of the complex rather than an observation on this protein. Unlike GO:0031507, however, the term itself has independent support on this gene through the Ensembl Compara row above, so it is retained rather than marked over-annotated. It is non-core for the direction reason given on that row. Supporting Evidence: file:human/AEBP2/AEBP2-bioinformatics/RESULTS.md | `PMID:33514705` | 38 | 9 | 2 | `{'ComplexPortal': 35, 'DisProt': 3}` | 4 terms × 5 entities | |
| GO:0000122 negative regulation of transcription by RNA polymerase II | NAS PMID:29348366 Structures of human PRC2 with its cofactors AEBP2 and JARID2... | KEEP AS NON CORE | Summary: The same ComplexPortal projection as the PMID:33514705 row, from the companion structural paper. Kept as non-core. Reason: Nine entities, two of them ComplexPortal complexes, all receiving the identical NAS term set. As with its sibling row the term has independent support on this gene elsewhere in the record, so the projection is a provenance observation rather than grounds for removal. Supporting Evidence: file:human/AEBP2/AEBP2-bioinformatics/RESULTS.md | `PMID:29348366` | 35 | 9 | 2 | `{'ComplexPortal': 35}` | 4 terms × 6 entities | |
| GO:0031507 heterochromatin formation | NAS PMID:33514705 Structures of monomeric and dimeric PRC2:EZH1 reveal flexibl... | MARK AS OVER ANNOTATED | Summary: Over-annotated for this protein. The term reaches AEBP2 only as a ComplexPortal projection of a complex-level property, the cited reference performs no assay of heterochromatin assembly, and what AEBP2-specific functional data exist point the other way for the isoform that human tissues actually express. Reason: Three independent lines. First, the projection test: this reference annotates nine entities and gives all nine - including the CPX-2212 and CPX-2330 complexes themselves - the same NAS term set, while the functional rows it carries (GO:0031491, GO:0046976, GO:0140693, all IMP) sit on EZH1 alone. So ComplexPortal projected the process terms and withheld the catalytic ones, and AEBP2's GO:0031507 is the projected kind. Second, the census across the complex: GO:0031507 is held by NAS alone on AEBP2, JARID2, MTF2, PALI1, PHF19, RBBP4, RBBP7 and SUZ12, and with IDA, ISS or IBA only on EED, EZH1 and EZH2 - the term tracks complex membership, not per-subunit evidence. Third, the definition: GO:0031507 requires a process that starts with heterochromatin nucleation, spreads, and ends with boundary formation, and neither this reference nor PMID:29348366 assays any of that; both are structural studies. Against that, the broadly expressed long isoform renders PRC2 almost inactive as a methyltransferase and its loss increases H3K27me3 at target genes, so AEBP2's own data do not support a positive contribution to heterochromatin assembly. MARK_AS_OVER_ANNOTATED rather than REMOVE because this is unmeasured for AEBP2 rather than measured and absent, and because PRC2 as a whole does establish facultative heterochromatin, so nothing here is refuted. Propagation Review Root cause: PROPAGATION BAD Failure modes: ROLE CONFLATION SOURCE EVIDENCE WEAK Sources checked: ComplexPortal:CPX-2212 · Polycomb repressive complex 2.2, EZH1-RBBP7 variant SUPPORTS SOURCE BUT NOT TARGET The complex-level statement is defensible for the complex. Projecting it onto an accessory subunit whose own biochemistry restrains that complex is what makes the row an over-annotation of AEBP2. ComplexPortal:CPX-2330 · Polycomb repressive complex 2.2, EZH1-RBBP4 variant SUPPORTS SOURCE BUT NOT TARGET Same projection. Both complexes and all seven subunits in this reference receive the identical NAS term set. Supporting Evidence: file:human/AEBP2/AEBP2-bioinformatics/RESULTS.md `GO:0031507` held by **NAS alone**: `['AEBP2', 'JARID2', 'MTF2', 'PALI1(LCOR)', 'PHF19', 'RBBP4', 'RBBP7', 'SUZ12']`. Held with other evidence: `['EED', 'EZH1', 'EZH2']`. PMID:33514705 Structures of monomeric and dimeric PRC2:EZH1 reveal flexible modules involved in chromatin compaction. PMID:41168462 While core PRC2 alone and PRC2–AEBP2S exhibited comparable HMTase activities, PRC2–AEBP2L was almost completely inactive |
| GO:0031507 heterochromatin formation | NAS PMID:29348366 Structures of human PRC2 with its cofactors AEBP2 and JARID2... | MARK AS OVER ANNOTATED | Summary: The same complex-level projection from the companion structural paper, and equally over-annotated for AEBP2. Reason: Identical shape to the PMID:33514705 row: nine entities, two of them ComplexPortal complexes, all receiving the same NAS term set; a structural study with no heterochromatin assembly assay; and no AEBP2-specific evidence for the process anywhere. The reasoning is given in full on that row and is not restated here. Propagation Review Root cause: PROPAGATION BAD Failure modes: ROLE CONFLATION SOURCE EVIDENCE WEAK Sources checked: ComplexPortal:CPX-2209 · Polycomb repressive complex 2.2, EZH2-RBBP4 variant SUPPORTS SOURCE BUT NOT TARGET Complex-level property projected onto every subunit including AEBP2. ComplexPortal:CPX-2213 · Polycomb repressive complex 2.2, EZH2-RBBP7 variant SUPPORTS SOURCE BUT NOT TARGET Complex-level property projected onto every subunit including AEBP2. Supporting Evidence: file:human/AEBP2/AEBP2-bioinformatics/RESULTS.md | `PMID:29348366` | 35 | 9 | 2 | `{'ComplexPortal': 35}` | 4 terms × 6 entities | |
| GO:0005634 nucleus | EXP PMID:29499137 Unique Structural Platforms of Suz12 Dictate Distinct Classe... | ACCEPT | Summary: Direct experimental localisation of AEBP2 to the nucleus, and the anchor that lets the other three nucleus rows stand on this gene's own data rather than on a complex-level statement. Reason: UniProt records the subcellular location with an experimental evidence tag citing this same reference. The projection test is reassuring rather than concerning here: the reference annotates four entities - SUZ12, RBBP4, JARID2 and AEBP2 - which are exactly the four components of the monomeric PRC2.2 assembly it studied, each given the term by UniProt independently. Supporting Evidence: file:human/AEBP2/AEBP2-uniprot.txt SUBCELLULAR LOCATION: Nucleus {ECO:0000269|PubMed:29499137}. file:human/AEBP2/AEBP2-bioinformatics/RESULTS.md | `PMID:29499137` | 4 | 4 | 0 | `{'UniProt': 4}` | 1 terms × 4 entities | |
| GO:0005634 nucleus | IEA GO_REF:0000044 | ACCEPT | Summary: A keyword-to-GO mapping of UniProt's own SUBCELLULAR LOCATION line, which itself carries an experimental evidence tag on this protein. Correct. Reason: The WITH/FROM token is the SubCell vocabulary term SL-0191, not a gene product, so there is no ortholog transfer to interrogate. The upstream statement is ECO:0000269|PubMed:29499137, so this row is a restatement of the same experimental localisation captured directly by the EXP row above rather than an independent inference. Supporting Evidence: file:human/AEBP2/AEBP2-uniprot.txt SUBCELLULAR LOCATION: Nucleus {ECO:0000269|PubMed:29499137}. |
| GO:0005634 nucleus | NAS PMID:33514705 Structures of monomeric and dimeric PRC2:EZH1 reveal flexibl... | ACCEPT | Summary: Part of the ComplexPortal projection, but the term is independently established on this gene by an EXP row from a different reference, so it is accepted. Reason: The row arrives with the same nine-entity identical-term-set signature as its GO:0000122 and GO:0031507 siblings, so it is complex-level in provenance. It is nonetheless accepted rather than marked over-annotated because AEBP2's nuclear localisation is directly measured: this is the case that separating a projection's terms by whether the target's own data support them is meant to identify. Supporting Evidence: file:human/AEBP2/AEBP2-uniprot.txt SUBCELLULAR LOCATION: Nucleus {ECO:0000269|PubMed:29499137}. |
| GO:0005634 nucleus | NAS PMID:29348366 Structures of human PRC2 with its cofactors AEBP2 and JARID2... | ACCEPT | Summary: The companion ComplexPortal projection of the same location term, accepted for the same reason. Reason: Complex-level in provenance, correct in substance, and independently supported on this gene by the EXP row from PMID:29499137. Supporting Evidence: file:human/AEBP2/AEBP2-uniprot.txt SUBCELLULAR LOCATION: Nucleus {ECO:0000269|PubMed:29499137}. |
| GO:0005654 nucleoplasm | IDA GO_REF:0000052 | ACCEPT | Summary: Direct immunofluorescence localisation to the nucleoplasm, consistent with the protein's chromatin association and more informative than the bare nucleus term. Reason: An antibody-based imaging observation from the Human Protein Atlas, concordant with UniProt's experimentally supported nuclear localisation and with the protein's ChIP-defined chromatin occupancy. Nothing in the record contradicts it. Supporting Evidence: file:human/AEBP2/AEBP2-uniprot.txt SUBCELLULAR LOCATION: Nucleus {ECO:0000269|PubMed:29499137}. |
| GO:0005654 nucleoplasm | TAS Reactome:R-HSA-212252 | ACCEPT | Summary: A Reactome pathway-level location assignment for the reaction in which PRC2 is recruited to chromatin. Correct but redundant with the HPA IDA row. Reason: Reactome places the whole PRC2 assembly, AEBP2 included, in the nucleoplasm for this reaction. The compartment is right and matches an independent direct observation; the term is simply less specific than chromatin would be for a reaction whose subject is chromatin recruitment. Supporting Evidence: file:human/AEBP2/AEBP2-uniprot.txt Note=Localizes to chromatin as part of the PRC2 complex. |
| GO:0005654 nucleoplasm | TAS Reactome:R-HSA-212263 | ACCEPT | Summary: The same Reactome compartment assignment for the H3K27 trimethylation reaction. Correct and redundant. Reason: Same basis as the other two Reactome rows: a pathway-level compartment for the complex in which AEBP2 sits, concordant with the HPA IDA. Supporting Evidence: file:human/AEBP2/AEBP2-uniprot.txt Note=Localizes to chromatin as part of the PRC2 complex. |
| GO:0005654 nucleoplasm | TAS Reactome:R-HSA-5638332 | ACCEPT | Summary: The third Reactome compartment row, from a reaction whose name identifies AEBP2 as part of the acting complex. Correct and redundant. Reason: This reaction is named for the PRC2 core together with AEBP2, so the assignment is explicitly about this protein's complex rather than PRC2 generically. The compartment is concordant with the direct HPA observation. Supporting Evidence: file:human/AEBP2/AEBP2-uniprot.txt Note=Localizes to chromatin as part of the PRC2 complex. |
| GO:0000785 chromatin | ISA GO_REF:0000113 | ACCEPT | Summary: Correct, and the most informative location term on the gene. Notable also for what this route did not assign: the same reference gives 97.5% of its recipients a sequence-specific DNA-binding transcription factor activity, and deliberately withheld it from AEBP2. Reason: UniProt states the chromatin association directly, as part of the PRC2 complex, with an experimental evidence tag, and it is corroborated by ChIP occupancy at PRC2 target loci. The route deserves recording because the obvious hazard for a three-zinc-finger protein is that a transcription-factor pipeline converts the fold into sequence-specific DNA binding. Measured against the whole reference rather than asserted: GO_REF:0000113 covers 727 human gene products, 709 of which (97.5%) receive GO:0000981, and AEBP2 is one of 18 given GO:0000785 alone. The other 17 are HOPX, ZFPM1, ZFPM2, NCOA1, NCOA2, NCOA3, SMAD6, SMAD7, NR0B1, NR0B2, NFX1, NFXL1, ZC3H6, ZC3H8, TFDP3, DMRTC1 and HMBOX1 - proteins whose DNA-binding-like folds serve protein contact rather than sequence recognition. The discrimination was made correctly, so this is reported as a non-confirmation of the predicted over-annotation. Supporting Evidence: file:human/AEBP2/AEBP2-uniprot.txt Note=Localizes to chromatin as part of the PRC2 complex. file:human/AEBP2/AEBP2-bioinformatics/RESULTS.md | receives `GO:0000785` chromatin **only** | 18 | 2.5% | |
| GO:0031491 nucleosome binding | IDA Q6ZN18-2 PMID:29499137 Unique Structural Platforms of Suz12 Dictate Distinct Classe... | NEW | Summary: Proposed new annotation. AEBP2 has no molecular function annotation at all in GOA, and its best-measured molecular contribution is to the nucleosome-binding activity of the assembled complex - which is why the qualifier is contributes_to rather than enables. Reason: The activity is a property of the assembled module, not of AEBP2 in isolation: AEBP2's C2B element binds the SUZ12 C2 domain and repositions it to promote nucleosome binding, and its extreme C-terminus contributes further. The isoform difference is quantitative rather than qualitative: it is strongest in the isoforms that retain the full C-terminus, isoforms 1 and 3, and isoform 2 supports the activity but roughly twofold less well. The measurements make that precise. The wild-type reconstituted module carried the last 97 residues of isoform 2 and bound mononucleosomes robustly at low micromolar affinity where the module lacking AEBP2 bound poorly; adding back the 14 residues that isoforms 1 and 3 carry and isoform 2 lacks gave a further twofold enhancement; and only an engineered deletion of the last five residues, which is not any natural isoform, abolished binding. The isoform field records Q6ZN18-2 because that is the form whose C-terminus was actually assayed, not because the activity is restricted to it. UniProt encodes the same conclusion as a feature. Note what this row deliberately does not say: the structural papers describe AEBP2 mimicking an unmodified H3 tail and displacing H3K4 from RBBP4, which is AEBP2 occupying a histone-binding pocket rather than AEBP2 binding a histone, so GO:0042393 histone binding is not proposed. Supporting Evidence: PMID:29499137 It binds to the C2 domain of Suz12 and relocates the latter to a unique position in S12R4J2A2 to promote nucleosome binding PMID:29499137 while S12R4J2A2d lost binding to mononucleosomes, S12R4J2A2a exhibited a 2-fold enhancement of nucleosome binding compared to the wild-type counterpart PMID:29499137 We disturbed the H3K4D by either deleting the last 5 residues to mimic Aebp2 from some lower eukaryotes (A2d) or adding the 14 residues found in isoforms 1 and 3 (A2a) PMID:29499137 We found that S12R4 exhibited poor nucleosome binding by itself PMID:29499137 In stark contrast, S12R4J2A2 displayed robust binding towards mononucleosomes, with a low micromolar binding affinity PMID:29499137 The Aebp2 fragment used in the crystal structure of S12R4J2A2 contains the last 97 residues of human Aebp2 isoform 2, including the C2B and H3K4D domains (Fig. 1A). file:human/AEBP2/AEBP2-uniprot.txt Important for nucleosome binding activity of the file:human/AEBP2/AEBP2-deep-research-affinage.md **molecular_activity:** GO:0003677 DNA binding, GO:0140110 transcription regulator activity, GO:0098772 molecular function regulator activity, GO:0060090 molecular adaptor activity, GO:0042393 histone binding |
| GO:0180000 histone methyltransferase inhibitor activity | IDA Q6ZN18-1 PMID:41168462 Auto-inhibition of PRC2 by the broadly expressed long isofor... | NEW | Summary: Proposed new annotation, scoped to the long isoform. Reconstituted PRC2 containing the long isoform is almost completely inactive as a histone methyltransferase, below the activity of the core complex alone, which is inhibition rather than reduced stimulation. Reason: The comparison's two arms matter and are stated: against core PRC2 alone, the complex carrying the short isoform has comparable methyltransferase activity while the complex carrying the long isoform is almost inactive. So the long isoform takes the enzyme below baseline, which matches GO:0180000's definition, binding to and reducing the activity of a histone methyltransferase. Evidence type is IDA: purified recombinant human proteins in a reconstituted assay. Requirement-type evidence agrees - loss of the long isoform increases PRC2 occupancy and H3K27me3 at target genes in embryonic stem cells, and an independent group reported the long isoform acting as an activator of a shared target gene. The precedent was measured, not assumed: GO:0180000 has 15 annotations in GOA, all EZHIP orthologs, with one experimental anchor, human EZHIP Q86X51 by IDA, so AEBP2 would be the second protein to hold the term and the first that is itself a subunit of the complex it inhibits. The isoform field records Q6ZN18-1 because that is the construct the methyltransferase experiment is reported for. A limitation worth stating rather than glossing: the HMTase result is reported for "PRC2-AEBP2L" without separating the two long isoforms, so Q6ZN18-2's methyltransferase behaviour is inferred rather than measured - the two were separately measured only in the DNA-binding arm, where they were indistinguishable. The short form Q6ZN18-3 must not carry this term. Supporting Evidence: PMID:41168462 While core PRC2 alone and PRC2–AEBP2S exhibited comparable HMTase activities, PRC2–AEBP2L was almost completely inactive PMID:41168462 In humans, there are three confirmed isoforms, which in this study are referred to as AEBP2L(iso1) (Q6ZN18-1), AEBP2L(iso2) (Q6ZN18-2) and AEBP2S (Q6ZN18-3) PMID:41168462 that AEBP2 enhances PRC2 function, we find that the widely expressed AEBP2L PMID:30923826 EZHIP/CXorf67 mimics K27M mutated oncohistones and functions as an intrinsic |
Loading supporting content…
Download this section (compressed HTML)Q: Is the absence of any DNA-binding annotation on AEBP2 deliberate? Neither human AEBP2, mouse Aebp2 nor Drosophila jing carries GO:0003677 or any descendant, verified with a positive control that returns 15 rows for human TP53 from the identical query, yet the protein's DNA binding has been measured twice - to the AE-1 element of the aP2 promoter, and to a bipartite CTT(N)15-23cagGCC motif derived from ChIP-cloned loci. PMID:19293275, the study that defined that motif, has zero annotations anywhere in GOA. If DNA binding is being withheld because both measurements are on the mouse protein and use GST fusions, that judgement is defensible and worth recording; if the reference was simply never curated, it is a gap.
Suggested experts: GO Central, MGI
Q: Should UniProt's FUNCTION and SUBUNIT lines record the isoform dependence of AEBP2's effect on PRC2? The entry currently states that AEBP2 acts as an accessory subunit of PRC2 leading to transcriptional repression, citing PMID:15225548, PMID:29499137 and PMID:31959557. PMID:41168462 reports that complexes containing the broadly expressed long isoform are almost inactive as histone methyltransferases and gain none of the DNA-binding affinity the short isoform confers, and it attributes the earlier stimulation results to use of the short isoform or of N-terminally truncated protein. Note the two arms differ: the methyltransferase result is below the core-complex baseline, whereas on naked DNA the long-isoform complexes sit at that baseline and it is the short isoform that raises affinity, so "inhibits DNA binding" holds for chromatin occupancy in cells rather than for the in vitro probe. That attribution is checkable and largely checks out: of 17 PDB entries resolving an AEBP2 chain, 15 declare an N-terminally truncated construct starting at residue 209 or later.
Suggested experts: UniProt
Q: Should a complex-level biological process be projected onto an accessory subunit that restrains the complex? AEBP2's only support for GO:0031507 heterochromatin formation is two ComplexPortal NAS rows, each of which gives an identical term set to the complex plus all seven of its subunits. Across the 12 PRC2-associated proteins surveyed, 11 carry the term (EPOP carries none) and 8 of those 11 hold it by NAS alone. Both references are structural studies that assay no step of heterochromatin assembly, and the isoform that human tissues predominantly express renders PRC2 nearly inactive.
Suggested experts: ComplexPortal, GO Central
Q: Is UniProt's GO:0003712 transcription coregulator activity for human AEBP2, listed in the entry as IEA:Ensembl, intended to reach GOA? It is absent from GOA, while mouse Aebp2 carries GO:0003712 by both IDA and IMP from PMID:10329662. Two considerations argue against simply importing it: no human PRC2 subunit carries the term, and its definition requires modulating transcription via binding to a DNA-binding transcription factor, which is not how PRC2 is recruited. Resolving the divergence one way or the other would be more useful than leaving the two records disagreeing. Separately, the entry's three IEA:UniProtKB-KW molecular function rows (GO:0003677, GO:0008270, GO:0006351) are absent from GOA because the keyword route was retired, so they are noted rather than acted on.
Suggested experts: UniProt, GO Central
Experiment: Reconstitute PRC2.2 with recombinant AEBP2 isoform 1 versus isoform 2 and measure mononucleosome binding by EMSA or a fluorescence-based assay, and H3K27 methyltransferase activity on a nucleosomal substrate. PMID:29499137 already showed that adding the 14 residues back to an isoform-2-derived construct doubles nucleosome binding, but the comparison has not been made between the two natural full-length isoforms, and isoform 2 is the product of the reference transcript.
Hypothesis: The MANE-Select protein form of human AEBP2, isoform 2, supports weaker nucleosome binding by PRC2 than isoform 1, because its VSP_034359 deletion removes 14 of the 23 residues UniProt annotates as important for that activity.
Type: in vitro reconstitution
Experiment: An isoform-resolved proteomic or immunoblot survey of human tissues using antibodies or peptides that discriminate the N-terminal extension, complementing the transcript-level CAGE data. This determines whether the PRC2-promoting arm of AEBP2 biology exists outside testis in humans at all, which decides whether GO:0000122 is a whole-gene or a tissue-restricted statement.
Hypothesis: The short protein form of AEBP2 is expressed at protein level in human tissues other than testis; or it is not, in which case the gene's PRC2-promoting activity is essentially testis-restricted in humans.
Type: isoform-resolved proteomics
Experiment: Compare complex stoichiometry, SUZ12 C2 occupancy and JARID2 incorporation between PRC2 reconstituted with the long and the short isoform, and test whether the isolated acidic region inhibits in trans. An in-trans effect would separate an autoinhibitory-module mechanism from a steric one, and would determine whether GO:0180000 is the right molecular function or whether a nucleic-acid-mimicry description is closer.
Hypothesis: The acidic N-terminal region of the long isoform inhibits PRC2 by competing with nucleic-acid or nucleosome binding surfaces rather than by disrupting complex assembly.
Type: biochemical reconstitution and mutagenesis
Experiment: ChIP-seq or CUT&RUN for endogenous AEBP2 in a human cell line, with motif analysis, alongside SUZ12 and H3K27me3. All existing occupancy and motif data for this protein are mouse, which is why no DNA-binding annotation is proposed for the human gene despite two published measurements.
Hypothesis: Human AEBP2 occupies a sequence-defined subset of PRC2 target loci, as its mouse counterpart does.
Type: genomics
What is not known — curated, literature-grounded statements of the open unknowns (the inverse of core functions).
Gap: Whether AEBP2 stimulates or inhibits PRC2 is unresolved as a statement about the gene, because the answer depends on which protein isoform is present and the two directions rest on different bodies of work.
NARROWING BIOLOGYCURATION MF_DARK
What is known: Established: the complex containing the long isoform is almost inactive as a methyltransferase, below the level of the core complex alone, and loss of that isoform raises PRC2 occupancy and H3K27me3 at target genes. Also established: earlier work reporting stimulation used the short isoform or N-terminally truncated protein, a claim that survives a census of the deposited structures. Not established: whether the inhibition result replicates outside the single 2025 study, and what the short isoform does to methyltransferase activity, which against core complex alone was comparable rather than stimulatory. The DNA-binding arm is not the ambiguous one: there the short isoform raised affinity from above 4000 nM to 45.5 nM, roughly ninetyfold, so it is strongly stimulatory.
Significance: It decides the sign of every regulatory annotation on the gene, and it is why the direction-neutral GO:0006357 is the accepted core process while GO:0000122 is kept as non-core.
Provenance (the field's own admissions):
Gap: AEBP2 has no molecular function annotation in GOA, and the measured molecular activities that could fill the gap are either isoform-specific, complex-level, or measured only on the mouse protein.
NARROWING CURATION MF_DARK
What is known: Established: it is the only one of the 12 PRC2-associated human proteins surveyed with zero molecular_function rows, against 94 for EZH2 and 5 for MTF2. Established: it contributes to the assembled complex's nucleosome binding, and the long isoform inhibits the complex's methyltransferase activity - the two annotations proposed here. Not established: whether the protein binds DNA in a way attributable to the human gene product, since both DNA-binding measurements used mouse protein, and whether a coregulator-type molecular function applies at all, given that no human PRC2 subunit carries GO:0003712.
Significance: A subunit with no molecular function annotation is invisible to any analysis that reasons over molecular function, even when its biochemistry is among the best characterised in its complex.
Provenance (the field's own admissions):
Loading supporting content…
Download this section (compressed HTML)Loading supporting content…
Download this section (compressed HTML)Loading supporting content…
Download this section (compressed HTML)Loading supporting content…
Download this section (compressed HTML)