AFF4 is the central scaffold of the super elongation complex (SEC), the metazoan assembly that releases promoter-proximally paused RNA polymerase II into productive elongation. Most of the 1163-residue protein is intrinsically disordered and carries short, separately mapped interaction motifs along its length: an N-terminal segment binds the cyclin T1 subunit of P-TEFb without contacting the CDK9 kinase, a central 50-residue "ELLBow" binds the C-terminal domain of ELL2, and further motifs recruit the ENL/AF9 proteins. Because these sites are non-overlapping, AFF4 can hold P-TEFb and an ELL-family elongation factor on the same polymerase, and it does so without any additional partner; this bridging is the molecular activity of the protein itself, whereas the resulting stimulation of elongation is an activity of the assembled complex. Its one folded module, the C-terminal homology domain, mediates AFF4 homodimerisation and AFF1-AFF4 heterodimerisation and carries a surface loop phosphorylated by CDK9. AFF4 acts in the nucleus on transcriptionally active euchromatin, and its loss slows elongation, deepens promoter-proximal pausing and lowers Ser2-phosphorylated polymerase II genome-wide. It is required for Sertoli-cell-dependent spermiogenesis, and it promotes both adipogenic and osteogenic differentiation of mesenchymal stem cells - the latter in the opposite direction to its paralogue AFF1. It is the shared partner of MLL/KMT2A fusion oncoproteins, and is hijacked by the HIV-1 Tat protein, which folds onto the same cyclin T1 surface to build a Tat-AFF4-P-TEFb complex that transactivates the viral promoter. Missense substitutions in its ALF homology domain that make the protein resistant to SIAH1-directed proteasomal degradation cause CHOPS syndrome, a gain-of-function developmental disorder.
Definition: Binding to a cyclin that partners a transcriptional cyclin-dependent kinase and whose abundance does not oscillate with the cell cycle, such as cyclin T1, cyclin T2, cyclin K or cyclin H.
Justification: GO has no term for binding a transcriptional cyclin, and the existing cyclin-binding term cannot be stretched to cover one. GO:0030332 cyclin binding is defined for cyclins whose levels rise steadily until mitosis and then fall abruptly to zero, which is a cell-cycle differentia that cyclin T1 does not satisfy. Usage confirms that curators have respected this rather than stretching the term: CDK9, HEXIM1 and BRD4, the three canonical cyclin T1 binders, carry zero GO:0030332 annotations each, measured against the cell-cycle kinase CDK2 as a positive control, which carries five. The consequence is that the best-characterised molecular interaction of AFF4 - a crystallographically resolved, mutationally validated contact with cyclin T1 whose interface mutants reduce transcription - cannot be expressed in GO except as bare protein binding or as the more abstract adaptor activity used in this review. The same gap affects HEXIM1, BRD4, CDK9, the AFF1 paralogue and the HIV-1 Tat literature generally.
Parent term: protein binding
Supporting Evidence:
| GO Term | Evidence | Action | Reason |
|---|---|---|---|
| GO:0006355 regulation of DNA-templated transcription | IBA GO_REF:0000033 | ACCEPT | Summary: Accepted. The family-level statement of AFF4's core activity, and the best-supported of the five IBA rows: three of its four protein donors hold their own experimental annotation to this term or a descendant. Reason: Every protein donor in the WITH/FROM field was resolved and queried for its own evidence, so the usual "the donors only carry the same family-level inference" objection is testable here and is false. Drosophila lilli (FB:FBgn0041111 = Q9VQI9, Swiss-Prot) holds GO:0006355 by IMP; mouse Aff3 (MGI:MGI:106927 = P51827) holds it by both IDA and IMP; human AFF1 (UniProtKB:P51825) holds the descendants GO:0032786 and GO:0032968 by IMP. The remaining token is the PANTHER tree node, which is the inference structure rather than an evidence source. AFF4's own human data agree independently, and support a more specific child, GO:0032968, which is proposed as a separate NEW row rather than by modifying this one - the node covering AFF4 is the whole AF4/FMR2 family and is functionally heterogeneous (AFF2's only experimental molecular function is G-quadruplex RNA binding, AFF3's is DNA-binding transcription factor activity), so a general term is the correct least common ancestor for the node and an IBA should not be made more precise than the inference that produced it. Propagation Review Root cause: NO FAILURE CORE Sources checked: PANTHER:PTN000829417 · PANTHER tree node covering the AF4/FMR2 family SUPPORTS TRANSFER A tree node, not a protein. It carries 395 annotations over 79 recipient gene products and assigns all five of its terms uniformly to all of them. FB:FBgn0041111 · Drosophila lilli (Q9VQI9, Swiss-Prot, 1673 aa) SUPPORTS TRANSFER Holds GO:0006355 by its own IMP (PMID:11171404). The single AF4/FMR2 family member in Drosophila, so it is equally the orthologue of all four human paralogues. MGI:MGI:106927 · mouse Aff3 (P51827, Swiss-Prot) SUPPORTS TRANSFER Holds GO:0006355 by IDA and IMP (PMID:25162227). A paralogue rather than an orthologue, which is legitimate for an IBA but means no ortholog-strength inference comes from this token. MGI:MGI:1100819 · mouse Aff1 (O88573, Swiss-Prot) SUPPORTS TRANSFER Holds the descendant GO:0045893 by IDA (PMID:9365243). Resolved from ten candidate accessions for this MGI id; the reviewed one was chosen and the count recorded rather than silently taking the first hit. UniProtKB:P51825 · human AFF1 SUPPORTS TRANSFER Holds the descendants GO:0032786 and GO:0032968 by IMP (PMID:41062835). Supporting Evidence: PMID:37609817 we found that AFF4 knockdown by RNA interference in human HEL cells decreased not only cellular level but also global chromatin occupancy of CTD serine 2 phosphorylated Pol II file:human/AFF4/AFF4-bioinformatics/RESULTS.md | `FB:FBgn0041111` | protein | Q9VQI9 | AFFL_DROME (lilli) | Swiss-Prot | Drosophila melanogaster | 1673 | 3 | |
| GO:0003712 transcription coregulator activity | IBA GO_REF:0000033 | ACCEPT | Summary: Accepted, and kept rather than replaced, because the single protein donor holds this exact term by its own experimental annotation. The AFF4-specific molecular function is captured by two additional NEW rows instead. Reason: Drosophila lilli holds GO:0003712 by IMP (PMID:11171404), so the propagation lands exactly at its donor's own experimental term rather than above or beside it. The term's definition requires modulation of specific gene sets via binding a DNA-binding transcription factor at a specific locus, either alone or as part of a complex, and AFF4 satisfies this at least for the MLL-target genes, where the AFF4-containing complex is recruited to specific chromatin to facilitate transcription. It is nevertheless not the most informative statement of what AFF4 does, and the two candidates that are were verified not to be redundant with it: GO:0003711 transcription elongation factor activity and GO:0003712 are SIBLINGS under GO:0140110 - both ancestor closures were fetched and neither contains the other - and GO:0030674 protein-macromolecule adaptor activity sits under GO:0060090, outside the binding branch entirely. Both are therefore proposed as additional NEW rows rather than as replacements for this one. Propagation Review Root cause: NO FAILURE CORE Sources checked: PANTHER:PTN000829417 · PANTHER tree node covering the AF4/FMR2 family SUPPORTS TRANSFER A tree node, not a protein. FB:FBgn0041111 · Drosophila lilli (Q9VQI9, Swiss-Prot) SUPPORTS TRANSFER Holds GO:0003712 by its own IMP (PMID:11171404) - the propagation lands at the donor's term, not above it. Supporting Evidence: PMID:20153263 AF4 and ENL family proteins are frequently fused with MLL, and they comprise a higher order complex (designated AEP) containing the P-TEFb transcription elongation factor. PMID:20153263 Here, we show that AEP is normally recruited to MLL-target chromatin to facilitate transcription. file:human/AFF4/AFF4-bioinformatics/RESULTS.md | elongation factor activity is NOT under coregulator activity | GO:0003711 | GO:0003712 | False | False | |
| GO:0006354 DNA-templated transcription elongation | IBA GO_REF:0000033 | ACCEPT | Summary: Accepted at the polymerase-agnostic level the family inference supports, with the polymerase II-specific child proposed separately on AFF4's own human evidence. Reason: The one protein donor, human AFF1, holds this exact term by EXP (PMID:22547686), so the chain contains real experimental evidence. Everything known about AFF4 itself is polymerase II-specific - P-TEFb phosphorylates the polymerase II CTD, the super elongation complex is defined as acting on polymerase II, and AFF4 depletion changes Ser2-phosphorylated polymerase II and polymerase II pausing - so GO:0006368 transcription elongation by RNA polymerase II is the right term for the gene. It was confirmed by fetching the ancestor closure that GO:0006368 is a descendant of GO:0006354, so this is a specificity relationship and not a sideways move. It is proposed as a NEW row rather than as a MODIFY of this one, and the reason is specific to AFF4 rather than general. The WITH/FROM field on this row is byte-identical on AFF4 and on AFF1 - PANTHER:PTN000829417 plus UniProtKB:P51825 - but on AFF1 that makes it SELF-REFERENTIAL, so it records a PAINT curator judging the function core for AFF1 itself, whereas on AFF4 the only protein named is a PARALOGUE. A paralogue-derived inference does not carry AFF4 down to the polymerase, so the specific term is put on a row bearing AFF4's own evidence code instead of being folded into the phylogenetic row. That self-referential-versus-paralogue-derived distinction is checkable from the WITH/FROM column of this gene's own committed GOA table and from AFF1's, and it is what the verdict rests on. Recorded as a cross-reference rather than as evidence: the concurrent AFF1 review, PR #2348, reaches MODIFY on the same term for this reason inverted, and the divergence is deliberate; the net GO content is the same either way, since GO:0006368 is added here as a NEW row. This term is therefore a verified ancestor of one that core_functions already records, and is not restated there. One caveat that argues the other way and is recorded rather than suppressed: the same node hands GO:0032783 to all 79 of its recipients, and that term's definition is explicitly about RNA polymerase II, so the node is not in fact behaving as though its clade were polymerase-agnostic. That is an internal inconsistency in the node's own assignments, and it is filed as a PAINT question rather than used to justify further precision here - because this review separately questions whether that complex term should reach AFF2 and AFF3 at all, and it would be circular to lean on an assignment it doubts. Propagation Review Root cause: NO FAILURE CORE Sources checked: PANTHER:PTN000829417 · PANTHER tree node covering the AF4/FMR2 family SUPPORTS TRANSFER A tree node, not a protein. UniProtKB:P51825 · human AFF1 SUPPORTS TRANSFER Holds GO:0006354 by its own EXP annotation (PMID:22547686). Supporting Evidence: PMID:37528066 disruption of AFF4 results in slow elongation and early termination in a subset of AFF4-bound active genes, whereas AFF1 deletion leads to fast elongation and transcriptional readthrough in the same subset of genes file:human/AFF4/AFF4-bioinformatics/RESULTS.md | Pol II elongation is_a DNA-templated elongation | GO:0006368 | GO:0006354 | True | True | |
| GO:0050877 nervous system process | IBA GO_REF:0000033 | KEEP AS NON CORE | Summary: Kept as non-core. Both donor experiments are learning-and-memory phenotypes on paralogues, not on an AFF4 orthologue, and AFF4's own nervous-system connection is a developmental syndrome rather than a molecular role. Reason: Both protein donors do hold their own experimental evidence, but at the descendant GO:0007611 learning or memory (verified: GO:0007611's ancestor closure contains GO:0050877), and both are paralogues - mouse Aff2, which is FMR2, the FRAXE intellectual-disability gene (PMID:11923441), and Drosophila lilli. No AFF4 orthologue appears among the seeds of any of the five IBA rows. AFF4 does have a clinical connection to the nervous system, since cognitive impairment is the C of CHOPS syndrome, and that is why the term is not removed - but it is named here as context and explicitly NOT as evidence of a molecular function, because the variants that cause it are stabilising gain-of-function substitutions and reading such a phenotype as molecular involvement is the move the GO:0032968 row on this gene refuses. What carries the verdict is this: the donors carry real experimental evidence at a descendant term, the term itself is broad and this gene is ubiquitously expressed with an embryonic-lethal null, and AFF4's only nervous-system-specific study of its own (PMID:22528490, on ghrelin-induced AMPK signalling in hypothalamic neurons) is a single unreplicated report that this review does not rely on anywhere. So the row is kept as a family-level inference and marked non-core rather than treated as AFF4-specific biology. Related trap checked and avoided: the robotic ataxia mouse frequently discussed alongside CHOPS carries a missense mutation in Aff1, not Aff4. Propagation Review Root cause: NO FAILURE NON CORE Failure modes: GRANULARITY MISMATCH Sources checked: PANTHER:PTN000829417 · PANTHER tree node covering the AF4/FMR2 family SUPPORTS TRANSFER A tree node, not a protein. FB:FBgn0041111 · Drosophila lilli (Q9VQI9, Swiss-Prot) SUPPORTS TRANSFER Holds the descendant GO:0007611 by IMP (PMID:18310460), so the propagation generalises a specific behavioural phenotype upward to the organ-system level. MGI:MGI:1202294 · mouse Aff2 / Fmr2 (O55112, Swiss-Prot) SUPPORTS TRANSFER Holds the descendant GO:0007611 by IMP (PMID:11923441). A paralogue, and the one whose loss-of-function causes an intellectual-disability phenotype in its own right, so the nervous-system content of this node is largely AFF2's. Supporting Evidence: PMID:11923441 Impaired conditioned fear and enhanced long-term potentiation in Fmr2 knock-out mice PMID:25730767 A missense mutation in the ALF homology domain of Aff1 (Af4) was reported in the robotic mouse, an ataxia mouse model created by ENU mutagenesis |
| GO:0032783 super elongation complex | IBA GO_REF:0000033 | ACCEPT | Summary: Accepted, and the single most important row on the gene - but it reaches AFF4 only by phylogenetic inference, while the human experiments that established AFF4 as the complex's scaffold have never been curated onto it. Reason: The term is unambiguously correct: UniProt calls AFF4 a key component of the super elongation complex acting as its central scaffold, and the term's own definition requires an ELL-family factor, an EAF protein and an AFF family protein. The donor chain is also sound - Drosophila lilli holds GO:0032783 by its own IPI (PMID:22195968). What is wrong is the evidence code, and the reason is a curation gap rather than a propagation defect. PMID:20159561 is titled for AFF4 and states that the complex includes AFF4 and that AFF4 is required for its stability, yet the single annotation that reference has produced anywhere in GOA is GO:0032783 IDA on AFF1; and PMID:20471948, PMID:23471103, PMID:24843025, PMID:28134250, PMID:31147444 and PMID:32128251 have produced no GO annotation at all. So AFF4 holds by IBA what its paralogue holds by IDA from an AFF4-titled paper. Accepted as correct, with the evidence-code upgrade raised as a curation request rather than fabricated here. Propagation Review Root cause: NO FAILURE CORE Sources checked: PANTHER:PTN000829417 · PANTHER tree node covering the AF4/FMR2 family SUPPORTS TRANSFER A tree node, not a protein. It gives this term uniformly to all 79 of its recipients, including AFF2 and AFF3, neither of which has been shown to be an SEC subunit - raised as a PAINT question rather than acted on here, since the term's definition admits an AFF family protein or distant relative. FB:FBgn0041111 · Drosophila lilli (Q9VQI9, Swiss-Prot) SUPPORTS TRANSFER Holds GO:0032783 by its own IPI (PMID:22195968) - the propagation lands at the donor's term. Supporting Evidence: PMID:20159561 SEC includes ELL, P-TEFb, AFF4, and several other factors. AFF4 is required for SEC stability and proper transcription by poised RNA polymerase II in metazoans. file:human/AFF4/AFF4-uniprot.txt SIMILARITY: Belongs to the AF4 family. {ECO:0000305}. file:human/AFF4/AFF4-bioinformatics/RESULTS.md | PMID:20159561 | AFF4 is a component of SEC; AFF4 required for SEC stability; knockdown lowers MLL-chimera target genes | 1 | 1 | **False** | |
| GO:0005634 nucleus | IEA GO_REF:0000120 | ACCEPT | Summary: Accepted. Correct and independently established experimentally, though the row's three tokens reduce to two witnesses rather than three. Reason: The nuclear localisation is not in doubt: UniProt records it from PubMed:12065898, the same observation appears in this gene's GOA as an EXP row, HPA's immunofluorescence gives nucleoplasm as the main location, and an MCEF-EGFP fusion localised exclusively to the nucleus. The row is retained as correct. It is worth recording that the combinatorial reference's apparent agreement between three independent methods is overstated: fetching ARBA00026330 shows that of its 1309 condition sets exactly one matches AFF4, and that set is IPR007797 AND IPR043640 AND Eukaryota, so the ARBA token and the InterPro token are the same signature evidence reached twice; and SL-0191 derives from the UniProt SUBCELLULAR LOCATION line evidenced by PubMed:12065898, the same experiment as the EXP row. Nucleus is an ancestor of the nucleoplasm and euchromatin locations carried in core_functions, so it is not restated there. Propagation Review Root cause: NO FAILURE CORE Sources checked: ARBA:ARBA00026330 · ARBA rule assigning GO:0005634 SUPPORTS TRANSFER A 1309-condition-set rule; the one set that matches AFF4 is InterPro id=IPR007797 AND InterPro id=IPR043640 AND taxon=Eukaryota. InterPro:IPR043640 · AF4/FMR2 C-terminal homology domain signature CIRCULAR OR REDUNDANT The same signature the ARBA rule is conditioned on, and it separately carries its own interpro2go mapping to GO:0005634 - one witness counted twice, not two. UniProtKB-SubCell:SL-0191 · UniProt subcellular location vocabulary term Nucleus CIRCULAR OR REDUNDANT Derives from UniProt's own SUBCELLULAR LOCATION line, which is evidenced by PubMed:12065898 - the same experiment already recorded as the EXP row. Supporting Evidence: file:human/AFF4/AFF4-uniprot.txt SUBCELLULAR LOCATION: Nucleus {ECO:0000269|PubMed:12065898}. PMID:17389929 In HeLa cells, an MCEF-EGFP fusion protein, localized exclusively to the nucleus. |
| GO:0005694 chromosome | IEA GO_REF:0000044 | ACCEPT | Summary: Accepted. Redundant with the experimental row for the same term, and independently supported by chromatin immunoprecipitation of AFF4. Reason: SL-0468 Chromosome derives from UniProt's SUBCELLULAR LOCATION statement, which is evidenced by the same PubMed:12065898 that this gene's EXP row for the same term already cites, so the row adds no independent witness. It is nevertheless correct, and better supported than either route on its own: the CHOPS study maps AFF4 across chromatin genome-wide alongside cohesin and RNAP2 by ChIP-seq. Chromosome is an ancestor of the euchromatin location carried in core_functions and is not restated there. Propagation Review Root cause: NO FAILURE CORE Sources checked: UniProtKB-SubCell:SL-0468 · UniProt subcellular location vocabulary term Chromosome CIRCULAR OR REDUNDANT Derives from the UniProt SUBCELLULAR LOCATION line evidenced by PubMed:12065898, which is also the reference for this gene's EXP row to the same term. Supporting Evidence: PMID:25730767 Transcriptome and chromatin immunoprecipitation sequencing (ChIP-seq) analyses demonstrated similar alterations of genome-wide binding of AFF4, cohesin and RNAP2 in CdLS and CHOPS syndrome. |
| GO:0010468 regulation of gene expression | IEA GO_REF:0000002 | ACCEPT | Summary: Accepted. True, and derived from a family-specific signature rather than a bare fold, but it is the least informative possible statement of what this gene does. Reason: IPR007797 is the AF4/FMR2 family signature, not a generic fold, so the mapping is well founded rather than an accident of domain naming. The predicted fold-to-activity error was tested for and is absent: AFF4 matches exactly three InterPro signatures and interpro2go maps IPR007797 to this term, IPR043640 to GO:0005634, and IPR043639 to nothing at all, so no molecular function is manufactured from any of them. The term is a verified ancestor of the GO:0006355 IBA on this gene and of the proposed GO:0032968, so it is retained as correct and not separately restated in core_functions. Propagation Review Root cause: NO FAILURE CORE Sources checked: InterPro:IPR007797 · AF4/FMR2 family InterPro signature SUPPORTS TRANSFER A family-specific signature; its interpro2go mapping is to this term only, and to no molecular function. Supporting Evidence: file:human/AFF4/AFF4-bioinformatics/RESULTS.md | IPR043639 | **nothing** | |
| GO:0005515 protein binding | IPI PMID:20153263 A higher-order complex containing AF4 and ENL family protein... | MODIFY | Summary: Modified to protein-macromolecule adaptor activity. The cyclin T1 interaction is one of the two arms of the bridge AFF4 forms, it is resolved crystallographically, and its interface mutants reduce transcription. Reason: Bare protein binding says nothing, and this partner is real: 8 IntAct records across four independent publications at MI-score 0.81, and the reference behind this row is a directed biochemical study annotating only 6 entities rather than a screen. The informative replacement is adaptor activity rather than a binding term, for two measured reasons. First, GO:0030332 cyclin binding cannot be used for cyclin T1: its definition specifies cyclins whose levels rise until mitosis and then fall abruptly, which is not a transcriptional cyclin, and usage agrees - CDK9, HEXIM1 and BRD4, the three canonical cyclin T1 binders, carry zero GO:0030332 annotations each, while the cell-cycle kinase CDK2 carries five as a positive control. Second, what AFF4 does with this site is bridge: it holds cyclin T1 through an N-terminal segment resolved in PDB 4IMY and ELL2 through a separate central segment resolved in PDB 5JW9, and it can mediate the ELL2-P-TEFb interaction with no other factor present. Note this is a change of kind rather than a specialisation - GO:0030674's ancestor closure was fetched and does not contain GO:0005515 - and it is the replacement CLAUDE.md's curation guidance names for an adapter function. The replacement term is licensed by this gene's aggregate evidence rather than by this row's own reference alone, and this is the row that sets that out for the other eight: a row's reference establishes an interaction, whereas the two separately crystallised sites - cyclin T1 at residues 2-73 and ELL2 at 301-351 - together with the demonstration that AFF4 mediates the ELL2-P-TEFb interaction with no other factor present are what establish that each interaction is one arm of a bridge. Proposed replacements: protein-macromolecule adaptor activity Supporting Evidence: PMID:23471103 AFF4 meanders over the surface of the P-TEFb cyclin T1 (CycT1) subunit but makes no stable contacts with the CDK9 kinase subunit. Interface mutations reduced CycT1 binding and AFF4-dependent transcription. PMID:20471948 Without Tat, AFF4 can mediate the ELL2-P-TEFb interaction, albeit inefficiently. file:human/AFF4/AFF4-bioinformatics/RESULTS.md | CDK9 (P50750) | 0 | - | |
| GO:0005515 protein binding | IPI PMID:20153263 A higher-order complex containing AF4 and ENL family protein... | MODIFY | Summary: Modified to protein-macromolecule adaptor activity. AF9 is one of the factors AFF4 recruits into the complex, through a short motif on its disordered axis. Reason: A genuine complex partner rather than a screen hit: the reference is a directed study annotating 6 entities, AF9 appears in 4 IntAct records across two publications, and UniProt lists MLLT3/AF9 as a constituent of the super elongation complex. The recruitment is motif-based and was mapped both in vitro and in vivo. Adaptor activity is the informative replacement, on the same reasoning as the cyclin T1 rows; no GO term exists for binding an ENL/AF9-family YEATS protein. On the licence for the replacement term: this row's reference establishes the interaction, and the aggregate structural evidence set out on the first cyclin T1 row is what establishes that the interaction is one arm of a bridge. Proposed replacements: protein-macromolecule adaptor activity Supporting Evidence: PMID:23251033 HIV-1 Tat recruits transcription elongation factors dispersed along a flexible AFF4 scaffold PMID:20153263 AF4 and ENL family proteins are frequently fused with MLL, and they comprise a higher order complex (designated AEP) containing the P-TEFb transcription elongation factor. |
| GO:0005515 protein binding | IPI PMID:20153263 A higher-order complex containing AF4 and ENL family protein... | MODIFY | Summary: Modified to protein-macromolecule adaptor activity. ENL is the best-replicated of the recruited factors, with 11 IntAct records across three publications. Reason: The same treatment as the AF9 row and for the same reasons, with stronger replication: 11 IntAct records across PMID:20153263, PMID:21729782 and PMID:33961781, at MI-score 0.76, and three distinct detection methods within the directed study alone. ENL and AF9 are alternative occupants of the same position in the complex, so a single informative molecular function covers both. On the licence for the replacement term: this row's reference establishes the interaction, and the aggregate structural evidence set out on the first cyclin T1 row is what establishes that the interaction is one arm of a bridge. Proposed replacements: protein-macromolecule adaptor activity Supporting Evidence: PMID:20153263 AF4 and ENL family proteins are frequently fused with MLL, and they comprise a higher order complex (designated AEP) containing the P-TEFb transcription elongation factor. file:human/AFF4/AFF4-bioinformatics/RESULTS.md | MLLT1 (Q03111) | 11 | 3 | 5 | 0.76 | 148 | |
| GO:0005515 protein binding | IPI PMID:21729782 Human mediator subunit MED26 functions as a docking site for... | MODIFY | Summary: Modified to protein-macromolecule adaptor activity. An independent replication of the cyclin T1 interaction, from a study whose own subject is MED26. Reason: This row is the second of four independent publications recording the AFF4-cyclin T1 interaction, which is why the partner is treated as established rather than screen-derived. The paper's own subject is MED26 rather than AFF4, so it corroborates rather than establishes; the mutational evidence that the AFF4 site on cyclin T1 is distinct from the CDK9 and Hexim1 sites comes from elsewhere and is cited here because it is what makes the interaction mechanistically specific rather than an incidental co-purification. On the licence for the replacement term: this row's reference establishes the interaction, and the aggregate structural evidence set out on the first cyclin T1 row is what establishes that the interaction is one arm of a bridge. Proposed replacements: protein-macromolecule adaptor activity Supporting Evidence: PMID:24985467 A single point mutation in cyclin T1 eliminates binding to Hexim1, Cdk9 and RNA but not to AFF4 and enforces repression of HIV transcription. PMID:21729782 Reactivation of paused Pol II correlates with recruitment of super-elongation complexes (SECs) containing ELL/EAF family members, P-TEFb, and other proteins |
| GO:0005515 protein binding | IPI PMID:21729782 Human mediator subunit MED26 functions as a docking site for... | MODIFY | Summary: Modified to protein-macromolecule adaptor activity. The second independent record of the AF9 interaction. Reason: Same partner and same replacement as the AF9 row from PMID:20153263; this row is what makes that interaction two publications rather than one. UniProt's own SUBUNIT statement lists MLLT3/AF9 among the constituents of the complex, which is consistent with both records. On the licence for the replacement term: this row's reference establishes the interaction, and the aggregate structural evidence set out on the first cyclin T1 row is what establishes that the interaction is one arm of a bridge. Proposed replacements: protein-macromolecule adaptor activity Supporting Evidence: PMID:21729782 Reactivation of paused Pol II correlates with recruitment of super-elongation complexes (SECs) containing ELL/EAF family members, P-TEFb, and other proteins PMID:23251033 HIV-1 Tat recruits transcription elongation factors dispersed along a flexible AFF4 scaffold |
| GO:0005515 protein binding | IPI PMID:21729782 Human mediator subunit MED26 functions as a docking site for... | MODIFY | Summary: Modified to protein-macromolecule adaptor activity. The second independent record of the ENL interaction. Reason: Same partner and same replacement as the ENL row from PMID:20153263. Counting distinct publications rather than IntAct records or UniProt's NbExp matters here: NbExp has been observed elsewhere to count sub-methods of a single screen, replicates within one study, and even domains of one protein, so this review counts distinct publications and distinct publication-method pairs throughout. On the licence for the replacement term: this row's reference establishes the interaction, and the aggregate structural evidence set out on the first cyclin T1 row is what establishes that the interaction is one arm of a bridge. Proposed replacements: protein-macromolecule adaptor activity Supporting Evidence: PMID:21729782 Reactivation of paused Pol II correlates with recruitment of super-elongation complexes (SECs) containing ELL/EAF family members, P-TEFb, and other proteins PMID:20153263 AF4 and ENL family proteins are frequently fused with MLL, and they comprise a higher order complex (designated AEP) containing the P-TEFb transcription elongation factor. file:human/AFF4/AFF4-bioinformatics/RESULTS.md | MLLT1 (Q03111) | 11 | 3 | 5 | 0.76 | 148 | |
| GO:0005515 protein binding | IPI PMID:22190034 Global landscape of HIV-human protein complexes. | KEEP AS NON CORE | Summary: Kept as non-core. The interaction is real and far better evidenced than the row that records it - three crystal structures resolve AFF4 and Tat in one complex - but a host-pathogen interaction is not a core function of the human gene. Reason: The GOA support on its own is weak: PMID:22190034 is an affinity-purification survey that annotates 104 entities in the reference-projection test, and IntAct holds only 4 records for this pair, all from that one study at MI-score 0.56. The interaction is nonetheless established beyond doubt by structure - AFF4 and Tat are both resolved in PDB 4OGR and 4OR5, and in 5L1Z and 6CYT with TAR RNA - and the two proteins contact each other directly on the cyclin T1 surface. So this is the opposite case from the four unreplicated two-hybrid rows on this gene: a screen row rescued by independent structural evidence, which is why it is kept rather than marked over-annotated. It is non-core because AFF4's function is cellular polymerase II elongation and Tat binding is a viral hijack of it; the functional consequence is captured instead by the proposed GO:0043923 row. Supporting Evidence: PMID:24843025 Tat and AFF4 fold on the surface of CycT1 and interact directly. file:human/AFF4/AFF4-uniprot.txt DR PDB; 4OGR; X-ray; 3.00 A; C/G/L=2-73. |
| GO:0005515 protein binding | IPI PMID:25416956 A proteome-scale map of the human interactome network. | MARK AS OVER ANNOTATED | Summary: Marked as over-annotated. A single yeast two-hybrid screen logged as three sub-methods, with no orthogonal assay and no mention in any AFF4 study. Reason: AP2B1 has exactly 3 IntAct records for this pair, all from PMID:25416956, logged as two hybrid array plus two hybrid prey pooling approach plus validated two hybrid - three sub-methods of one experiment, which is what UniProt reports as NbExp=3 - at MI-score 0.56. No orthogonal assay exists and the partner appears in no AFF4 paper. AP2B1 is a clathrin adaptor subunit acting at the plasma membrane, against a protein UniProt localises to nucleus and chromosome and whose tagged form localises exclusively to the nucleus; AP2B1 also carries 537 IntAct interaction records against AFF4's 207, so it behaves as a hub. This follows the repo's established handling of unreplicated screen rows rather than being a fresh judgement, and it is deliberately not REMOVE, since a two-hybrid positive is an observation and no experiment has contradicted it. Supporting Evidence: PMID:17389929 In HeLa cells, an MCEF-EGFP fusion protein, localized exclusively to the nucleus. file:human/AFF4/AFF4-bioinformatics/RESULTS.md | AP2B1 (P63010) | 3 | 1 | 3 | 0.56 | 537 | |
| GO:0005515 protein binding | IPI PMID:25416956 A proteome-scale map of the human interactome network. | MARK AS OVER ANNOTATED | Summary: Marked as over-annotated. The same single two-hybrid screen, and the most promiscuous partner in the set. Reason: GOLGA2 is a Golgi matrix protein, and its evidence for AFF4 is the same 3 records from the same one screen at MI-score 0.56, with the same three sub-methods and no orthogonal assay. It carries 2335 IntAct interaction records against AFF4's 207 - the largest count among the nine partners - which is the profile of a sticky hub rather than a specific partner. No AFF4 study reports a Golgi association or a Golgi partner. Supporting Evidence: file:human/AFF4/AFF4-bioinformatics/RESULTS.md | GOLGA2 (Q08379) | 3 | 1 | 3 | 0.56 | 2335 | file:human/AFF4/AFF4-uniprot.txt SUBCELLULAR LOCATION: Nucleus {ECO:0000269|PubMed:12065898}. |
| GO:0005515 protein binding | IPI PMID:25416956 A proteome-scale map of the human interactome network. | MARK AS OVER ANNOTATED | Summary: Marked as over-annotated. The same single two-hybrid screen; a cytoplasmic TNF-receptor adaptor with no reported connection to transcription elongation. Reason: TRAF2 has 3 IntAct records for this pair, all from PMID:25416956, the same three sub-methods of one screen at MI-score 0.56, and 1727 IntAct records of its own. It functions in cytoplasmic TNF-receptor signalling; no AFF4 study places AFF4 in that pathway and no orthogonal assay supports the pair. Marked over-annotated rather than removed, on the same grounds as the other three screen rows. Supporting Evidence: file:human/AFF4/AFF4-bioinformatics/RESULTS.md | TRAF2 (Q12933) | 3 | 1 | 3 | 0.56 | 1727 | |
| GO:0005515 protein binding | IPI PMID:25416956 A proteome-scale map of the human interactome network. | MARK AS OVER ANNOTATED | Summary: Marked as over-annotated. The same single two-hybrid screen; a microtubule-associated protein with no reported connection to AFF4. Reason: MTUS2 has 3 IntAct records for this pair, all from PMID:25416956, the same three sub-methods of one screen at MI-score 0.56, and 1476 IntAct records of its own. It is a microtubule-associated protein; nothing in the AFF4 literature reports a cytoskeletal partner or a cytoplasmic pool of AFF4. Marked over-annotated rather than removed. Supporting Evidence: file:human/AFF4/AFF4-bioinformatics/RESULTS.md | MTUS2 (Q5JR59) | 3 | 1 | 3 | 0.56 | 1476 | |
| GO:0005515 protein binding | IPI PMID:25416956 A proteome-scale map of the human interactome network. | MODIFY | Summary: Modified to ubiquitin protein ligase binding. This row shares its GOA support with the four rows above and must not be treated the same way - the interaction is independently demonstrated with purified proteins and is the mechanism of the CHOPS disease variants. Reason: Deciding per partner rather than per screen is what separates this row from the other four on the same reference. SIAH1's IntAct evidence is the identical HuRI triple, but PMID:22483617 measured the interaction directly with purified AFF4 and Siah1 - AFF4 accumulated on Siah1 beads while displacing ELL2 from them, and the authors describe it as a direct Siah1-AFF4 interaction - and PMID:25730767 makes it disease-relevant, since the CHOPS missense variants act by resisting SIAH1-directed proteasomal degradation. SIAH1 is an E3 ubiquitin-protein ligase, so GO:0031625 ubiquitin protein ligase binding, defined as binding to a ubiquitin protein ligase enzyme, any of the E3 proteins, is an exact and informative replacement for bare protein binding. Note the direction of traffic: AFF4 is the ligase's substrate, so this is a binding term and not a ubiquitination activity for AFF4. Proposed replacements: ubiquitin protein ligase binding Supporting Evidence: PMID:22483617 The direct Siah1-AFF4 interaction detected in these two reactions is consistent with the reported binding of Siah1 to AFF1 PMID:22483617 The addition of increasing amounts of AFF4 into the binding reactions gradually decreased the levels of ELL2 bound to Siah1, and at the same time increased the amounts of AFF4 retained on the Siah1 beads PMID:25730767 In the probands reported here, the missense mutations create a resistance to ubiquitination-dependent proteasomal degradation, resulting in excessive amounts of mutant AFF4 protein accumulation. |
| GO:0005515 protein binding | IPI PMID:28514442 Architecture of the human interactome defines protein commun... | MODIFY | Summary: Modified to protein-macromolecule adaptor activity. The third independent publication recording the cyclin T1 interaction, from a large affinity-purification map. Reason: Treated as corroboration rather than as independent establishment, because the reference-projection test for PMID:28514442 is uninformative - 3731 annotations, too many to page through without sampling, so no entity count is derived from it. Its value here is that the cyclin T1 pair recurs in a third, methodologically different dataset, which is what makes the partner set for this gene look like biology rather than one laboratory's preparation. The replacement term and its justification are as for the other cyclin T1 rows. On the licence for the replacement term: this row's reference establishes the interaction, and the aggregate structural evidence set out on the first cyclin T1 row is what establishes that the interaction is one arm of a bridge. Proposed replacements: protein-macromolecule adaptor activity Supporting Evidence: PMID:23471103 AFF4 meanders over the surface of the P-TEFb cyclin T1 (CycT1) subunit but makes no stable contacts with the CDK9 kinase subunit. Interface mutations reduced CycT1 binding and AFF4-dependent transcription. file:human/AFF4/AFF4-bioinformatics/RESULTS.md | CCNT1 (O60563) | 8 | 4 | 6 | 0.81 | 263 | |
| GO:0005515 protein binding | IPI PMID:33961781 Dual proteome-scale networks reveal cell-specific remodeling... | MODIFY | Summary: Modified to protein-macromolecule adaptor activity. The fourth independent publication recording the cyclin T1 interaction. Reason: As for the other cyclin T1 rows. The projection test for PMID:33961781 is uninformative (9514 annotations), so this row is corroboration; the substantive evidence for the interaction and for the adaptor call is the crystal structure with interface mutants and the demonstration that AFF4 can mediate the ELL2-P-TEFb interaction unaided. On the licence for the replacement term: this row's reference establishes the interaction, and the aggregate structural evidence set out on the first cyclin T1 row is what establishes that the interaction is one arm of a bridge. Proposed replacements: protein-macromolecule adaptor activity Supporting Evidence: PMID:20471948 Without Tat, AFF4 can mediate the ELL2-P-TEFb interaction, albeit inefficiently. file:human/AFF4/AFF4-uniprot.txt DR PDB; 4IMY; X-ray; 2.94 A; G/H/I=2-73. |
| GO:0005515 protein binding | IPI PMID:33961781 Dual proteome-scale networks reveal cell-specific remodeling... | MODIFY | Summary: Modified to protein-macromolecule adaptor activity. The third independent publication recording the ENL interaction. Reason: As for the other ENL rows; this row is what makes that interaction three publications rather than two. Its own reference is a proteome-scale map whose projection test is uninformative, so it corroborates rather than establishes. On the licence for the replacement term: this row's reference establishes the interaction, and the aggregate structural evidence set out on the first cyclin T1 row is what establishes that the interaction is one arm of a bridge. Proposed replacements: protein-macromolecule adaptor activity Supporting Evidence: PMID:20153263 AF4 and ENL family proteins are frequently fused with MLL, and they comprise a higher order complex (designated AEP) containing the P-TEFb transcription elongation factor. file:human/AFF4/AFF4-bioinformatics/RESULTS.md | MLLT1 (Q03111) | 11 | 3 | 5 | 0.76 | 148 | |
| GO:0000791 euchromatin | IEA GO_REF:0000107 | ACCEPT | Summary: Accepted. The functionally meaningful location, transferred from the genuine mouse orthologue, which holds it by its own IDA. Reason: The donor is mouse Aff4 (Q9ESC8, Swiss-Prot) - an orthologue, not a paralogue - and it holds GO:0000791 by IDA from PMID:22195968, so the transfer rests on a real experimental observation on the right protein. The location is the one that matters functionally: euchromatin is defined as chromatin in a transcription-competent conformation, which is where a polymerase II elongation scaffold must act, and UniProt's note records that AFF4 associates with transcriptionally active chromatin but not at snRNA genes. Independently, ChIP-seq maps AFF4 genome-wide across chromatin with cohesin and RNAP2. Propagation Review Root cause: NO FAILURE CORE Sources checked: UniProtKB:Q9ESC8 · mouse Aff4 (Swiss-Prot, 1160 aa) SUPPORTS TRANSFER The genuine orthologue, holding GO:0000791 by its own IDA (PMID:22195968). ensembl:ENSMUSP00000051479 · Ensembl protein identifier for the same mouse Aff4 entry CIRCULAR OR REDUNDANT The same protein under a second identifier, not a second witness. Supporting Evidence: PMID:25730767 Transcriptome and chromatin immunoprecipitation sequencing (ChIP-seq) analyses demonstrated similar alterations of genome-wide binding of AFF4, cohesin and RNAP2 in CdLS and CHOPS syndrome. file:human/AFF4/AFF4-bioinformatics/RESULTS.md | `UniProtKB:Q9ESC8` | AFF4_MOUSE (Aff4) | GO:0000791 IDA PMID:22195968 | |
| GO:0034976 response to endoplasmic reticulum stress | IEA GO_REF:0000107 | MARK AS OVER ANNOTATED | Summary: Marked as over-annotated. The donor is the genuine rat orthologue, so this is not a paralogue transfer; the problem is one level down, in a single IEP annotation curated from a drug study in which Aff4 was not the subject. Reason: Resolving the donor is not enough here, so its own evidence was queried too. Rat Aff4 (A0A8I6GGV9) holds GO:0034976 by exactly one annotation, an IEP - inferred from expression pattern - curated by RGD from PMID:31466050, a study of genipin in a rat acute lung injury model whose subject is the PI3K/AKT pathway and not Aff4. The reference-projection test shows that paper produces exactly one annotation on one entity in the whole of GOA, so it is a single curation from a single expression experiment rather than a bulk import. A transcript changing abundance under ER stress does not establish that the gene product participates in the ER stress response, and no human AFF4 study reports any ER-stress result. The donor entry is also unreviewed, so its protein name is an automatic label; that is a separate point from its GO annotation, which is real. Marked over-annotated rather than removed because AFF4's involvement is unmeasured, not refuted, and the Ensembl Compara pipeline behaved as documented. Propagation Review Root cause: SOURCE WEAK OR INFERRED Failure modes: SOURCE EVIDENCE WEAK CONTEXT OR TISSUE MISMATCH Sources checked: UniProtKB:A0A8I6GGV9 · rat Aff4 (unreviewed / TrEMBL, 1164 aa) SOURCE WEAK OR INFERRED The correct orthologue, but its sole evidence for this term is one IEP from a drug-intervention study in rat lung. Expression correlation in a treated tissue is not participation in the process. ensembl:ENSRNOP00000088228 · Ensembl protein identifier for the same rat Aff4 entry CIRCULAR OR REDUNDANT The same protein under a second identifier, not a second witness. Supporting Evidence: PMID:31466050 Genipin attenuates mitochondrial-dependent apoptosis, endoplasmic reticulum stress, and inflammation via the PI3K/AKT pathway in acute lung injury file:human/AFF4/AFF4-bioinformatics/RESULTS.md | `UniProtKB:A0A8I6GGV9` | A0A8I6GGV9_RAT (Aff4) *[TrEMBL]* | GO:0034976 IEP PMID:31466050 | |
| GO:0001650 fibrillar center | IDA GO_REF:0000052 | KEEP AS NON CORE | Summary: Kept as non-core. HPA's own record grades this as an additional rather than a main location, and the fibrillar centre is a polymerase I compartment while AFF4's substrate is polymerase II. Reason: The Human Protein Atlas record for ENSG00000072364 lists nucleoplasm as the subcellular main location and nucleoli fibrillar center and nuclear bodies as additional locations, which is a checkable basis for grading the three GO_REF:0000052 rows differently rather than treating them alike. The fibrillar centre is the zone of transcription of the pre-rRNA genes by polymerase I; the super elongation complex acts on polymerase II, and no AFF4 role in rRNA transcription is reported anywhere. The observation is kept, because it is a curated immunofluorescence result that I cannot re-examine and nothing contradicts it, but it is not part of the gene's core localisation and does not appear in core_functions. Supporting Evidence: PMID:37528066 AFF1 mainly binds upstream of the TSS, while AFF4 is enriched downstream of the TSS. PMID:22895430 The super elongation complex (SEC) consists of the RNA polymerase II (Pol II) elongation factors eleven-nineteen Lys-rich leukaemia (ELL) proteins, positive transcription elongation factor b (P-TEFb) and several frequent mixed lineage leukaemia (MLL) translocation partners. |
| GO:0005654 nucleoplasm | IDA GO_REF:0000052 | ACCEPT | Summary: Accepted. HPA's main location for this protein, and the compartment in which the super elongation complex acts. Reason: This is the one HPA row graded as the subcellular main location rather than an additional location, and it agrees with everything else known: UniProt records a nuclear localisation from PubMed:12065898, a tagged form localises exclusively to the nucleus, and the complex AFF4 scaffolds acts on chromatin-associated polymerase II. Eleven Reactome TAS rows carry the same term from separate polymerase II elongation reactions and share this action; they are restatements of one fact and are annotated as such. Supporting Evidence: PMID:17389929 In HeLa cells, an MCEF-EGFP fusion protein, localized exclusively to the nucleus. file:human/AFF4/AFF4-uniprot.txt SUBCELLULAR LOCATION: Nucleus {ECO:0000269|PubMed:12065898}. |
| GO:0016604 nuclear body | IDA GO_REF:0000052 | KEEP AS NON CORE | Summary: Kept as non-core. An additional HPA location, and independently corroborated - AFF4 forms nuclear puncta with FUS - but not where the elongation function is executed. Reason: HPA grades nuclear bodies as an additional location rather than the main one, and there is independent support for it: live-cell imaging shows FUS co-localising with AFF4 in nuclear puncta, which are disrupted by aliphatic alcohol treatment. So this row is corroborated rather than doubtful, which is worth stating because it distinguishes it from the fibrillar-centre row, where no literature exists. It is non-core because AFF4's elongation function is executed on gene bodies rather than in a membraneless nuclear compartment, and because the FUS study frames the puncta as a mechanism restraining AFF4 rather than as its site of action. Supporting Evidence: PMID:31238957 Live cell imaging demonstrates that FUS co-localizes with AFF4 within nuclear punctuate PMID:31238957 In this study we identify fused in sarcoma (FUS) as a partner of AFF4 in cells. |
| GO:0005634 nucleus | EXP PMID:12065898 MCEF, the newest member of the AF4 family of transcription f... | ACCEPT | Summary: Accepted. The original experimental localisation of AFF4, and the ultimate source of the UniProt statement that several other rows on this gene restate. Reason: UniProt records the nuclear localisation as experimentally evidenced by this reference, and it is corroborated independently by HPA immunofluorescence and by an exclusively nuclear MCEF-EGFP fusion. The cached copy of this paper is abstract-only, so the localisation claim is taken from UniProt's curated SUBCELLULAR LOCATION line rather than from a sentence of the paper I have not read. Nucleus is an ancestor of the nucleoplasm and euchromatin locations recorded in core_functions and is therefore not restated there. Supporting Evidence: file:human/AFF4/AFF4-uniprot.txt SUBCELLULAR LOCATION: Nucleus {ECO:0000269|PubMed:12065898}. |
| GO:0005694 chromosome | EXP PMID:12065898 MCEF, the newest member of the AF4 family of transcription f... | ACCEPT | Summary: Accepted. Chromatin association is now much better evidenced than by this reference alone, through genome-wide chromatin immunoprecipitation of AFF4. Reason: UniProt records Chromosome for AFF4 with the same experimental evidence tag as Nucleus, from this reference. The claim has since been established far more strongly: ChIP-seq maps AFF4 across chromatin genome-wide, and the CHOPS study shows its distribution changing together with cohesin and RNAP2. As with the nucleus row, the cached copy is abstract-only and the localisation is taken from the UniProt curation. Chromosome is an ancestor of the euchromatin location recorded in core_functions. Supporting Evidence: PMID:25730767 Transcriptome and chromatin immunoprecipitation sequencing (ChIP-seq) analyses demonstrated similar alterations of genome-wide binding of AFF4, cohesin and RNAP2 in CdLS and CHOPS syndrome. |
| GO:0005654 nucleoplasm | TAS Reactome:R-HSA-112379 | ACCEPT | Summary: Accepted. One of eleven Reactome rows carrying the same true term from separate reactions of the polymerase II elongation complex. Reason: Reactome places AFF4 in its RNA polymerase II elongation complex, and every reaction of that complex therefore lists AFF4 with the compartment the reaction occurs in. This reaction, the recruitment of elongation factors to form the elongation complex, is the one that most directly matches AFF4's role. The term is correct and independently supported by HPA immunofluorescence, which is why all twelve GO:0005654 rows on this gene share the ACCEPT action; the eleven Reactome rows are restatements of one fact rather than eleven observations, and that is recorded here rather than left implicit. Supporting Evidence: PMID:20159561 SEC includes ELL, P-TEFb, AFF4, and several other factors. AFF4 is required for SEC stability and proper transcription by poised RNA polymerase II in metazoans. |
| GO:0005654 nucleoplasm | TAS Reactome:R-HSA-112381 | ACCEPT | Summary: Accepted. The same compartment statement, from the Reactome reaction in which P-TEFb hyperphosphorylates the polymerase II CTD. Reason: This is the reaction whose chemistry AFF4 actually promotes, since the complex it scaffolds delivers P-TEFb to paused polymerase; AFF4 knockdown lowers Ser2-phosphorylated polymerase II genome-wide. The compartment term is correct and shared with the other eleven GO:0005654 rows on this gene. Supporting Evidence: PMID:37609817 we found that AFF4 knockdown by RNA interference in human HEL cells decreased not only cellular level but also global chromatin occupancy of CTD serine 2 phosphorylated Pol II |
| GO:0005654 nucleoplasm | TAS Reactome:R-HSA-112385 | ACCEPT | Summary: Accepted. The same compartment statement, from the nucleotide-addition reaction of the Reactome elongation complex. Reason: A pathway-participation restatement of the nucleoplasm location rather than an independent observation. Retained because the term is correct and because the Reactome model's placement of AFF4 in the elongation complex is itself well founded. Supporting Evidence: file:human/AFF4/AFF4-uniprot.txt SUBCELLULAR LOCATION: Nucleus {ECO:0000269|PubMed:12065898}. |
| GO:0005654 nucleoplasm | TAS Reactome:R-HSA-112392 | ACCEPT | Summary: Accepted. The same compartment statement, from the Reactome reaction for resumption of elongation after pausing. Reason: Of the eleven Reactome reactions this is, with the P-TEFb phosphorylation step, the one closest to AFF4's measured effect: AFF4 loss increases promoter-proximal pausing of polymerase II on thousands of genes. The compartment term is correct and shared with the other GO:0005654 rows. Supporting Evidence: PMID:37609817 we found by ChIP-seq and PRO-seq that AFF4 loss also increased promoter-proximal pause of Pol II on several hundred HS and thousands of non-HS genes |
| GO:0005654 nucleoplasm | TAS Reactome:R-HSA-112395 | ACCEPT | Summary: Accepted. The same compartment statement, from the Reactome abortive-termination reaction. Reason: A pathway-participation restatement of the nucleoplasm location. AFF4 has no reported role specific to abortive termination; the row is retained for the compartment it asserts, which is correct, and not read as evidence about that reaction. Supporting Evidence: file:human/AFF4/AFF4-uniprot.txt SUBCELLULAR LOCATION: Nucleus {ECO:0000269|PubMed:12065898}. |
| GO:0005654 nucleoplasm | TAS Reactome:R-HSA-112396 | ACCEPT | Summary: Accepted. The same compartment statement, from the Reactome transcript-separation reaction. Reason: A pathway-participation restatement of the nucleoplasm location, retained for the compartment it asserts. No AFF4-specific evidence bears on this particular reaction. Supporting Evidence: file:human/AFF4/AFF4-uniprot.txt SUBCELLULAR LOCATION: Nucleus {ECO:0000269|PubMed:12065898}. |
| GO:0005654 nucleoplasm | TAS Reactome:R-HSA-113411 | ACCEPT | Summary: Accepted. The same compartment statement, from the Reactome short-backtracking reaction. Reason: A pathway-participation restatement of the nucleoplasm location. Polymerase backtracking is mechanistically adjacent to the pausing that AFF4 relieves, but no AFF4 experiment addresses backtracking specifically, so the row is retained only for its compartment. Supporting Evidence: file:human/AFF4/AFF4-uniprot.txt SUBCELLULAR LOCATION: Nucleus {ECO:0000269|PubMed:12065898}. |
| GO:0005654 nucleoplasm | TAS Reactome:R-HSA-113412 | ACCEPT | Summary: Accepted. The same compartment statement, from the Reactome reaction for processive movement of the elongation complex. Reason: A pathway-participation restatement of the nucleoplasm location, and the reaction whose rate AFF4 affects - AFF4 disruption produces slow elongation. Retained for the compartment, which is correct. Supporting Evidence: PMID:37528066 disruption of AFF4 results in slow elongation and early termination in a subset of AFF4-bound active genes, whereas AFF1 deletion leads to fast elongation and transcriptional readthrough in the same subset of genes |
| GO:0005654 nucleoplasm | TAS Reactome:R-HSA-113413 | ACCEPT | Summary: Accepted. The same compartment statement, from the Reactome TFIIS-mediated recovery reaction. Reason: A pathway-participation restatement of the nucleoplasm location, retained for the compartment. TFIIS-mediated rescue is a different mechanism from the P-TEFb-dependent pause release AFF4 promotes, and no AFF4 evidence bears on it. Supporting Evidence: file:human/AFF4/AFF4-uniprot.txt SUBCELLULAR LOCATION: Nucleus {ECO:0000269|PubMed:12065898}. |
| GO:0005654 nucleoplasm | TAS Reactome:R-HSA-113414 | ACCEPT | Summary: Accepted. The same compartment statement, from the Reactome long-backtracking reaction. Reason: A pathway-participation restatement of the nucleoplasm location, retained for the compartment. As with the other backtracking reaction, no AFF4 experiment addresses it specifically. Supporting Evidence: file:human/AFF4/AFF4-uniprot.txt SUBCELLULAR LOCATION: Nucleus {ECO:0000269|PubMed:12065898}. |
| GO:0005654 nucleoplasm | TAS Reactome:R-HSA-113429 | ACCEPT | Summary: Accepted. The same compartment statement, from the Reactome reaction in which the elongating transcript meets a template lesion. Reason: A pathway-participation restatement of the nucleoplasm location, retained for the compartment. The last of the eleven Reactome rows; together they demonstrate how a single complex membership in a pathway database becomes eleven identical GO rows, which is why they are annotated once as a set rather than read as eleven independent observations. Supporting Evidence: file:human/AFF4/AFF4-uniprot.txt SUBCELLULAR LOCATION: Nucleus {ECO:0000269|PubMed:12065898}. |
| GO:0000791 euchromatin | ISS GO_REF:0000024 | ACCEPT | Summary: Accepted. The manual curator transfer of the same mouse observation that the automatic row also carries - two routes, one witness. Reason: This row and the GO:0000791 IEA row are the manual and automatic transfers of the same fact from the same donor, mouse Aff4, which holds the term by its own IDA from PMID:22195968. For an ISS row the WITH/FROM correctly names only the sequence-similar entity and no interaction partner, which is what it does here. The location is the functionally meaningful one for a polymerase II elongation scaffold, and UniProt's own note - that AFF4 associates with transcriptionally active chromatin but not at snRNA genes - is itself tagged as a by-similarity inference from this same mouse entry, so the human statement of this location has no independent human experiment behind it. That is a gap worth recording rather than a reason to reject a well-founded orthologue transfer. Propagation Review Root cause: NO FAILURE CORE Sources checked: UniProtKB:Q9ESC8 · mouse Aff4 (Swiss-Prot, 1160 aa) SUPPORTS TRANSFER The genuine orthologue, holding GO:0000791 by its own IDA (PMID:22195968). For an ISS row this is the correct entity type - the sequence-similar protein, not an interactor. Supporting Evidence: PMID:25730767 Transcriptome and chromatin immunoprecipitation sequencing (ChIP-seq) analyses demonstrated similar alterations of genome-wide binding of AFF4, cohesin and RNAP2 in CdLS and CHOPS syndrome. PMID:22195968 LEC subunits are highly enriched at RNA Pol II-transcribed small nuclear RNA (snRNA) genes file:human/AFF4/AFF4-bioinformatics/RESULTS.md | `UniProtKB:Q9ESC8` | AFF4_MOUSE (Aff4) | GO:0000791 IDA PMID:22195968 | |
| GO:0008023 transcription elongation factor complex | IDA PMID:22195968 The little elongation complex regulates small nuclear RNA tr... | ACCEPT | Summary: Accepted, and the reference-projection test that would have exposed a subunit-projection defect was run and came back negative. Reason: This row has the shape that produces a complex-level-evidence-attributed-to-a-subunit defect - one reference, 61 annotations over 26 entities, with this term reaching 16 of them - so both halves of the projection test were applied rather than only the entity count. First, all 16 recipients of GO:0008023 IDA from this reference are genuine elongation-complex subunits (human AFF1, AFF4, CDK9, EAF1, EAF2, ELL, ELL2, ELL3, MLLT1, MLLT3 and the Drosophila SEC and LEC proteins CycT, Eaf, Ell, Ice1, Ice2 and lilli); there is no bystander in the list. Second, and decisively, the paper's functional term does not spread: GO:0042795 snRNA transcription by RNA polymerase II IMP sits on only 7 entities - the ELL and ICE1 proteins actually perturbed - and does not reach AFF4. That is correct biology, since the paper's point is that the little elongation complex rather than the super elongation complex works at snRNA genes. So the complex membership spread with the biochemistry while the phenotype stayed with the perturbed genes, which is the benign pattern rather than the defective one. GO:0008023 is a verified ancestor of GO:0032783, so core_functions records the specific complex and not this parent. Supporting Evidence: PMID:20159561 SEC includes ELL, P-TEFb, AFF4, and several other factors. AFF4 is required for SEC stability and proper transcription by poised RNA polymerase II in metazoans. PMID:22195968 Here, we identify an SEC-like complex in Drosophila, as well as a distinct ELL-containing complex that lacks P-TEFb and other components of SEC named the "little elongation complex" (LEC). file:human/AFF4/AFF4-bioinformatics/RESULTS.md - GO:0042795 IMP [biol] → 7 entities |
| GO:0030674 protein-macromolecule adaptor activity | IPI PMID:20471948 HIV-1 Tat and host AFF4 recruit two transcription elongation... | NEW | Summary: New annotation proposed. This is the molecular function AFF4 performs on its own, as distinct from the elongation activity of the complex it builds, and it is what the nine bare protein-binding rows are groping at. Reason: AFF4 has no molecular-function annotation in GOA other than bare protein binding, which is the gap this row closes. The claim is that AFF4 itself brings two macromolecules together, and it is supported rather than inferred: the cyclin T1 site sits in residues 2-73 (PDB 4IMY, and 4OGR and 4OR5 with Tat), the ELL2 site is a separate 50-residue segment at 301-351 (PDB 5JW9), deletion of 301-400 abolishes ELL2 binding while leaving other interactions intact, and AFF4 mediates the ELL2-P-TEFb interaction with no other factor present. The distinction from the complex's own activity is deliberate and is the schema question this gene raises: bridging is AFF4's, whereas stimulating polymerase II is the complex's, so the latter is recorded separately as contributes_to. Note the ancestor closure of GO:0030674 was fetched and does not contain GO:0005515, so this is a change of kind and not a specialisation of protein binding. Supporting Evidence: PMID:20471948 Without Tat, AFF4 can mediate the ELL2-P-TEFb interaction, albeit inefficiently. PMID:20471948 Through the bridging functions of Tat and AFF4, P-TEFb and ELL2 combine to form a bifunctional elongation complex that greatly activates HIV-1 transcription. PMID:28134250 Here we report the 2.0-Å resolution crystal structure of the human ELL2 C-terminal domain bound to its 50-residue binding site on AFF4, the ELLBow. file:human/AFF4/AFF4-uniprot.txt DR PDB; 5JW9; X-ray; 2.00 A; A=301-351. |
| GO:0003711 transcription elongation factor activity | IMP PMID:37609817 AFF4 globally affects the release of paused RNA polymerase I... | NEW | Summary: New annotation proposed, with the contributes_to qualifier because this activity belongs to the super elongation complex and AFF4 supplies the scaffold rather than the chemistry. Reason: The complex AFF4 scaffolds stimulates the elongation properties of polymerase II, which is exactly what GO:0003711 asserts, and AFF4's own contribution is measurable: knockdown lowers cellular and chromatin-bound Ser2-phosphorylated polymerase II and deepens promoter-proximal pausing genome-wide, and AFF4 disruption slows elongation. The qualifier is contributes_to rather than enables because the catalytic subunit of the complex is CDK9 and AFF4 supplies no chemistry - this is the distinction the gene exists to test, and getting it wrong in either direction would misstate the evidence. The term is not redundant with the existing GO:0003712 IBA: both ancestor closures were fetched and GO:0003711 and GO:0003712 are siblings under GO:0140110, neither containing the other. It is also a consistency fix rather than a new precedent, since human AFF1 already holds GO:0003711 by IMP. Supporting Evidence: PMID:37609817 we found that AFF4 knockdown by RNA interference in human HEL cells decreased not only cellular level but also global chromatin occupancy of CTD serine 2 phosphorylated Pol II PMID:37528066 disruption of AFF4 results in slow elongation and early termination in a subset of AFF4-bound active genes, whereas AFF1 deletion leads to fast elongation and transcriptional readthrough in the same subset of genes file:human/AFF4/AFF4-bioinformatics/RESULTS.md - [MF] GO:0003711 IMP PMID:41062835 |
| GO:0006368 transcription elongation by RNA polymerase II | IMP PMID:37528066 Distinct roles of two SEC scaffold proteins, AFF1 and AFF4, ... | NEW | Summary: New annotation proposed. AFF4 has no experimental biological-process annotation at all, and this is the polymerase II-specific statement that its own human data support and that the family-level IBA cannot make. Reason: Everything measured about AFF4 is polymerase II-specific: the complex it scaffolds is defined as acting on polymerase II, its partner P-TEFb phosphorylates the polymerase II CTD, and AFF4 depletion changes polymerase II Ser2 phosphorylation, polymerase II pausing, elongation rate and termination. The existing GO:0006354 IBA is accepted as correct at the polymerase-agnostic level the family inference supports; this row adds the specific term on AFF4's own evidence. The ancestry was verified rather than assumed - GO:0006354 is in GO:0006368's ancestor closure - so the two rows are consistent and this one asserts strictly more. The evidence code is IMP because the experiments are RNA interference and gene disruption, which is a requirement result; no sufficiency claim is made from them. Supporting Evidence: PMID:37528066 AFF1 mainly binds upstream of the TSS, while AFF4 is enriched downstream of the TSS. PMID:37528066 disruption of AFF4 results in slow elongation and early termination in a subset of AFF4-bound active genes, whereas AFF1 deletion leads to fast elongation and transcriptional readthrough in the same subset of genes file:human/AFF4/AFF4-bioinformatics/RESULTS.md | Pol II elongation is_a DNA-templated elongation | GO:0006368 | GO:0006354 | True | True | |
| GO:0032968 positive regulation of transcription elongation by RNA polymerase II | IMP PMID:37609817 AFF4 globally affects the release of paused RNA polymerase I... | NEW | Summary: New annotation proposed. The direction of AFF4's effect is measured and positive, and no row in GOA records it. Reason: AFF4 loss lowers Ser2-phosphorylated polymerase II and increases promoter-proximal pausing on thousands of genes, and AFF4 disruption slows elongation - so the protein promotes rather than restrains elongation, and the sign is measured rather than assumed. The existing GO:0006355 IBA carries no direction; this row supplies it, and GO:0006355 was verified to be an ancestor of GO:0032968 so the two are consistent. Two things this row deliberately does not rest on. The CHOPS syndrome data are not used: a stabilising missense with a gain-of-function transcriptional phenotype shows that excess AFF4 dose is sufficient to activate transcription, not that AFF4 is required for elongation, and conflating the two would be the classic reading of a disease mechanism as a molecular function. And the PNUTS-PP1 dephosphorylation of AFF4 Ser-549 is not used either, because the cached copy of that paper is abstract-only and its abstract does not mention AFF4. What this row rests on is knockdown and disruption, which are requirement experiments. Supporting Evidence: PMID:37609817 we found by ChIP-seq and PRO-seq that AFF4 loss also increased promoter-proximal pause of Pol II on several hundred HS and thousands of non-HS genes PMID:37609817 we found that AFF4 knockdown by RNA interference in human HEL cells decreased not only cellular level but also global chromatin occupancy of CTD serine 2 phosphorylated Pol II file:human/AFF4/AFF4-bioinformatics/RESULTS.md | positive reg. of Pol II elongation under reg. of transcription | GO:0032968 | GO:0006355 | True | True | |
| GO:0042803 protein homodimerization activity | IDA PMID:31147444 Structure of the super-elongation complex subunit AFF4 C-ter... | NEW | Summary: New annotation proposed. Two independent crystal structures of the same C-terminal region show an AFF4 dimer, and point mutants abolish it. Reason: This is direct evidence about AFF4 itself rather than about the complex, which is why it is annotated as enables rather than contributes_to. The C-terminal homology domain mediates both an AFF4 homodimer and an AFF1-AFF4 heterodimer; a second group crystallised the same region and showed that either F1014A or Y1096A alone impairs dimer formation, and PDBe reports two AFF4 copies in the deposited assemblies of 6R80 and 6K7P. The two groups name the region differently - C-terminal homology domain and THD - but it is the same 899-1163 segment, which UniProt and InterPro record as the AF4/FMR2 C-terminal homology domain. Only the homodimerisation term is proposed; the AFF1-AFF4 heterodimer would need a separate row and its own partner evidence, and AFF4 has no AFF1 protein-binding row in GOA to attach one to. Supporting Evidence: PMID:31147444 Our analyses further revealed that AFF4-CHD mediates the formation of an AFF4 homodimer or an AFF1-AFF4 heterodimer. PMID:32128251 Mutagenesis analysis revealed that single mutations of either Phe1014 or Tyr1096 of AFF4 to file:human/AFF4/AFF4-uniprot.txt DR PDB; 6KN5; X-ray; 2.20 A; A=899-1163. |
| GO:0043923 host-mediated activation of viral transcription | IMP PMID:20471948 HIV-1 Tat and host AFF4 recruit two transcription elongation... | NEW | Summary: New annotation proposed, and this is a consistency fix rather than a request for new precedent - AFF4's own P-TEFb partners already hold this term while AFF4 does not. Reason: GO:0043923 carries 20 human annotations over 18 gene products - 17 proteins plus the CCNT1-CDK9 complex - and they include CCNT1 and CDK9 by IDA - the two subunits of the P-TEFb complex that AFF4 binds and delivers - together with SNW1 by IDA and IMP, EP300 by IDA and SMARCA4 and SMARCB1 by IMP. AFF4 holds it not at all, although it is the factor that brings ELL2 into the same complex and is resolved with Tat in the crystal structures of it. So GO already curates this class, and for this exact mechanism; the asymmetry is the argument. One caveat recorded rather than glossed: the CCNT1 and CDK9 rows are assigned by ComplexPortal from a single reference and are themselves a complex-level projection, so the non-projected precedents are SNW1, EP300 and the SWI/SNF pair. The evidence code is IMP because the AFF4 contribution was shown by manipulating AFF4 in reporter and viral transcription assays; the structures establish the interaction, not the transcriptional outcome. This row is deliberately absent from core_functions, for the same reason the Tat protein-binding row is kept as non-core: activating a viral promoter is something done to AFF4's machinery by a pathogen, not part of what the human gene is for. Supporting Evidence: PMID:20471948 Through the bridging functions of Tat and AFF4, P-TEFb and ELL2 combine to form a bifunctional elongation complex that greatly activates HIV-1 transcription. PMID:24843025 Tat and AFF4 fold on the surface of CycT1 and interact directly. file:human/AFF4/AFF4-bioinformatics/RESULTS.md | CDK9 | IDA | PMID:10866664 | ComplexPortal | |
| GO:0045600 positive regulation of fat cell differentiation | IMP PMID:36149892 AFF4 regulates cellular adipogenic differentiation via targe... | NEW | Summary: New annotation proposed. The best-evidenced endogenous human process this gene has, shown by requirement and by sufficiency and confirmed in vivo, and GOA carries no row from the paper at all. Reason: Added after review, which correctly pointed out that this review proposed a row for a viral hijack it calls non-core while leaving the best-evidenced endogenous process to a curation request. The evidence is unusually complete for this gene: knockdown in human mesenchymal stem cells and in mouse 3T3-L1 preadipocytes inhibits adipogenic differentiation, which is a requirement result; overexpression enhances adipogenesis and produces ectopic adipose tissue, which is a sufficiency result; and an adipose-specific Aff4 knockout mouse impedes adipocyte development, which tests it in vivo. Both directions are named because GO's evidence codes do not distinguish them. The mechanism is also mapped - AFF4 binds and transcriptionally activates ATG5 and ATG16L1, and depleting either abrogates adipogenesis in AFF4-overexpressing cells - so this is not a bare phenotype. The evidence code is IMP because the human arm is siRNA knockdown; the mouse knockout corroborates in another species rather than supplying the human annotation. Supporting Evidence: PMID:36149892 Knockdown of AFF4 in human mesenchymal stem cells (hMSCs) and mouse 3T3-L1 preadipocytes inhibits cellular adipogenic differentiation. PMID:36149892 Overexpression of AFF4 enhances adipogenesis and ectopic adipose tissue formation. PMID:36149892 We further generate Fabp4-cre driven adipose-specific Aff4 knockout mice and find that AFF4 deficiency impedes adipocyte development and white fat depot formation. PMID:36149892 AFF4 directly binds to autophagy-related protein ATG5 and ATG16L1, and promotes their transcription. |
| GO:0045669 positive regulation of osteoblast differentiation | IMP PMID:28955517 AFF1 and AFF4 differentially regulate the osteogenic differe... | NEW | Summary: New annotation proposed. The paralogue-discriminating counterpart of the adipogenic row - AFF4 promotes osteogenic differentiation of human MSCs while AFF1 restrains it - tested in human cells and in vivo. Reason: Added after review, together with the reference itself, because the description asserted an osteogenic role whose only source in this PR was the provider record. Reading the paper made the case stronger than expected rather than weaker: siRNA depletion of AFF4 in human MSCs inhibits osteogenic potential (requirement), overexpression enhances it, and the two are confirmed in vivo by MSC-mediated bone formation, so this is not an in vitro-only result. It is also the cleanest demonstration anywhere that AFF4 and AFF1 are not interchangeable: depleting AFF1 increases alkaline phosphatase activity and mineralisation while depleting AFF4 does the opposite, which is the differentiation counterpart of the elongation antagonism reported in PMID:37528066, and the concurrent AFF1 review independently records GO:0045668 for that gene. The mechanism is mapped as well as the phenotype: AFF4 depletion lowers ID1 mRNA and protein, overexpression raises it, AFF4 is enriched at the ID1 promoter by ChIP, and AFF4 depletion restricts BMP2-induced BRE-luciferase responses, SP7 induction and ALP activity - so the osteogenic arm runs through BMP-pathway control of ID1, the counterpart of the ATG5/ATG16L1 route in the adipogenic row. The paper's 2020 erratum, PMID:32257529, was read rather than treated as an unknown, and it does not touch this row: it corrects the second lane label of Fig. 7b, which should read siAFF4, and replaces misused alpha-tubulin images in Figs. 6b and 7b, with the authors stating it does not affect any of their original conclusions. Both panels are protein blots in the mechanism figures - the AFF1-to-DKK1 western and the AFF4-to-ID1 western - and neither is among the differentiation or bone-formation assays quoted below. Supporting Evidence: PMID:28955517 depletion of AFF4 significantly inhibits the osteogenic potential of MSCs PMID:28955517 we confirm that overexpression of AFF1 and AFF4 differentially affects osteogenic differentiation in vitro and MSC-mediated bone formation in vivo PMID:32257529 Correction to: AFF1 and AFF4 differentially regulate the osteogenic differentiation of human MSCs |
Loading supporting content…
Download this section (compressed HTML)Q: Almost none of the AFF4 functional literature has reached GO. Of twenty papers that establish something about human AFF4, fifteen have produced no GO annotation anywhere in GOA and eighteen have produced none on AFF4 itself; as a result AFF4 has no experimental biological-process annotation at all, and its only experimental molecular-function rows are six bare protein-binding IPI rows. The sharpest instance is PMID:20159561, which is titled for AFF4 and states that AFF4 is required for super elongation complex stability, and whose single annotation anywhere in GOA is GO:0032783 IDA on AFF1. Would GO Central consider a targeted curation pass over PMID:20159561, PMID:20471948, PMID:22483617, PMID:23471103, PMID:24843025, PMID:28134250, PMID:31147444, PMID:32128251, PMID:36149892, PMID:37528066 and PMID:37609817?
Suggested experts: GO Central, UniProt curators
Q: AFF4 has no protein-binding annotation for ELL2, although PMID:28134250 reports a 2.0 A co-crystal of the ELL2 C-terminal domain with a 50-residue AFF4 site and UniProt states that the interaction is direct. Every one of AFF4's IntAct AFF4-ELL2 records is spoke-expanded from co-immunoprecipitation of the complex, and every one of the nine partners GOA does export as GO:0005515 has at least one non-spoke-expanded record - so being spoke-expanded-only is sufficient to account for this particular absence. It is deliberately not claimed to be the whole export rule: seven further partners do have non-spoke-expanded records and are also absent from GOA, so some additional filter operates that this review did not identify. Either way the actionable gap is the same: the one paper that measured a binary AFF4-ELL2 interaction and crystallised it has never been curated. Should PMID:28134250 be curated to IntAct as a binary interaction, or annotated directly as IPI or IDA?
Suggested experts: IntAct curators, GO Central
Q: The PANTHER node PTN000829417 assigns GO:0032783 super elongation complex uniformly to all 79 of its recipients, which includes AFF2 and AFF3. Neither has been shown to be a subunit of that complex - AFF2's only experimental molecular function is G-quadruplex RNA binding and AFF3's is DNA-binding transcription factor activity - whereas AFF1 and AFF4 have. The term's definition does admit an AFF family protein or distant relative, so the propagation is defensible rather than plainly wrong. Should the complex-membership term be restricted to a node covering the AFF1/AFF4 clade, leaving the other family-level terms where they are?
Suggested experts: PAINT curators, GO Central
Q: Human AFF4's five phylogenetic annotations name no AFF4 orthologue among their seeds - they are seeded by Drosophila lilli, human AFF1 and mouse Aff1, Aff2 and Aff3 - even though mouse Aff4 is a co-recipient of the same node and carries its own experimental annotations. Is that a deliberate consequence of how IBD seeds are selected, or would mouse Aff4's own annotations qualify it as a seed?
Suggested experts: PAINT curators
Q: Ensembl Compara projected rat Aff4's GO:0034976 IEP onto human AFF4, but did not project mouse Aff4's GO:0007286 spermatid development IMP, which is a much stronger annotation from a knockout. The mouse row carries acts_upstream_of_or_within rather than involved_in, which is a plausible technical reason. Is qualifier-based filtering the explanation, and if so, should an IEP donor annotation be eligible for projection when a knockout-based IMP is not?
Suggested experts: Ensembl Compara maintainers, GOA
Q: GO:0030332 cyclin binding is defined for cyclins whose abundance oscillates with the cell cycle, so it cannot be used for the transcriptional cyclins, and usage shows curators have not stretched it - CDK9, HEXIM1 and BRD4 carry no GO:0030332 annotation while CDK2 carries five. Should the definition be broadened, or a sibling term for transcriptional cyclin binding be created? A large literature on P-TEFb, 7SK and HIV-1 Tat currently has no way to express its central interaction.
Suggested experts: GO ontology editors
Q: Is there any AFF4-specific evidence for the euchromatin location in human cells? Both GOA rows are transfers from mouse Aff4, and UniProt's note that AFF4 associates with transcriptionally active chromatin but not at snRNA genes is itself tagged as a by-similarity inference from the same mouse entry - so the human statement rests entirely on the mouse experiment.
Suggested experts: UniProt curators
Q: Does the fibrillar-centre localisation reported by immunofluorescence correspond to any function? AFF4 acts on RNA polymerase II, and the fibrillar centre is the polymerase I transcription zone; no AFF4 role in ribosomal RNA transcription has been reported. The annotation is kept as an observation but is not treated as core in this review.
Suggested experts: Human Protein Atlas, GO Central
Experiment: Reconstitute pause release in vitro with purified polymerase II, DSIF, NELF, P-TEFb and ELL2, and titrate wild-type AFF4 against three separation-of-function mutants - a cyclin T1 interface mutant from PDB 4IMY, the 301-400 deletion that abolishes ELL2 binding, and the F1014A dimerisation mutant. Read out Ser2 phosphorylation and run-off transcript length. The prediction that distinguishes the adaptor model from a dose model is that the two interface mutants fail while the dimerisation mutant does not.
Hypothesis: AFF4's adaptor activity - simultaneous engagement of P-TEFb and an ELL-family factor - is separable from its dimerisation and from its recruitment to chromatin, and only the adaptor function is required for pause release.
Type: in vitro reconstitution
Experiment: Compare wild-type AFF4 with the T254A, T254S and R258W variants expressed from a single genomic locus at matched protein levels, using a degron or an inducible promoter to hold abundance equal, and measure complex composition by quantitative proteomics, chromatin occupancy by ChIP-seq, and elongation rate by PRO-seq. If the mechanism is purely dose, all three variants should be indistinguishable from wild type once abundance is equalised; any residual difference would indicate that the ALF homology domain does something beyond presenting the SIAH1 degron.
Hypothesis: The CHOPS syndrome variants change AFF4 abundance and nothing else about the protein's molecular behaviour.
Type: dose-controlled variant comparison
Experiment: Purify the super elongation complex from cells in which either AFF1 or AFF4 is the only AFF scaffold present, and quantify ELL, ELL2, ELL3, MLLT1 and MLLT3 stoichiometry by absolute quantitative mass spectrometry. Pair this with ChIP-seq of each subunit in each background to test whether the upstream-versus-downstream distribution difference reflects a compositional difference or only a positional one.
Hypothesis: AFF4 and AFF1 build complexes with distinct ELL-family and ENL-family occupancy, which is why their loss produces opposite elongation phenotypes.
Type: quantitative proteomics of purified complexes
Experiment: The domain's RNA and DNA binding was suggested by fluorescence anisotropy with purified protein, which is why no nucleic-acid-binding annotation is proposed in this review. Test it in cells by CLIP and by ChIP-exo using a separation-of-function domain mutant that retains dimerisation, and ask whether any recovered signal is independent of the polymerase II machinery AFF4 travels with.
Hypothesis: The AFF4 C-terminal homology domain binds nucleic acid in cells, not only in vitro.
Type: in-cell nucleic acid crosslinking
What is not known — curated, literature-grounded statements of the open unknowns (the inverse of core functions).
Gap: AFF4 has no experimental biological-process annotation in GO, and no experimental molecular-function annotation other than bare protein binding, despite being the structurally best-characterised subunit of the super elongation complex.
NARROWING CURATION RESIDUAL_SUBGAP
What is known: What is known is extensive - nine PDB entries covering three separate interaction surfaces, interface and dimerisation mutants, and genome-wide knockdown phenotypes. What is missing is the curation, not the biology.
Significance: An unannotated gene looks uncharacterised to any downstream analysis that reads GO rather than the literature, and this gene is a drug target of interest in both leukaemia and HIV latency.
What would resolve it: Eight NEW rows are proposed here for the part of the gap that current GO terms can express; the rest needs a curation pass over the eleven references named in the first suggested question.
Provenance (the field's own admissions):
Gap: Whether AFF4's stimulation of polymerase II elongation requires anything beyond holding P-TEFb and an ELL-family factor in proximity is not known.
OPEN BIOLOGY MF_DARK
What is known: The two binding sites are mapped to residue level and their mutants reduce transcription, and CDK9 phosphorylates a loop of the C-terminal domain; but no experiment separates a pure proximity effect from an allosteric one.
Significance: It determines whether the adaptor activity annotated here is the whole molecular function or only part of it.
Provenance (the field's own admissions):
Gap: The human euchromatin localisation of AFF4 has no human experiment behind it.
OPEN CURATION CC_DARK
What is known: Mouse Aff4 holds the term by IDA and both human rows are transfers from it; UniProt's associated note is itself a by-similarity inference from the same mouse entry. Human AFF4 chromatin occupancy has been mapped by ChIP-seq, which supports chromosome and chromatin but not specifically the euchromatin subclass.
Significance: Small, but it is the location on which the functional interpretation of every other row depends.
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)