The reviewed human entry is O60885 (BRD4; HGNC:13575), with a 1,362-residue canonical protein. The normal seed contains 81 distinct source assertions from 82 GOA rows. One RelA interaction occurs twice with provider/date differences only. All 81 machine source objects are preserved. There are no negated or isoform-specific GOA rows.
The three named products remain A (Brd4L, Long), O60885-1; C (Brd4S, Short), O60885-2; and B, O60885-3. Experimental “isoform B” chromatin-insulation evidence is kept distinct from the long P-TEFb-binding product; names are not silently remapped by numerical order. BRD4-NUTM1 fusion experiments do not establish every function of native BRD4. Secondary accessions O60433, Q4G0X8, Q86YS8 and Q96PD3 and observed RefSeq identifiers are identity metadata, not additional accepted gene symbols. HUNK1 and HUNKI are distinct recorded aliases.
Source recovery imported three exact normal primary files and 21 previously absent normal source files (19 PubMed caches and two Reactome records), following the normal fetch failure and authenticated original workflow artifact. Eleven pre-existing source variants were preserved, and neither PTHR22880 export was installed. Full-text flags describe the normal fetcher's output, not a claim that every experiment or supplement has been read. All 30 source PubMed abstracts and both Reactome summaries were read. Specific Results/Methods/captions were read for the decisions below; complete figures, raw data and all screen tables were not independently audited.
The standard Falcon research wrapper could not first resolve its offline tool package. The documented override to the already installed requested client version then attempted Falcon and its perplexity-lite fallback once; both failed with DNS errors and created no provider output. This journal contains manual research, not provider-generated research.
BRD4 recognizes acetylated chromatin and selected acetylated transcription factors, and recruits transcriptional regulators. The long product recruits P-TEFb; CDK9 supplies the kinase activity attributed to P-TEFb in these experiments. The ET-domain route recruits additional partners and can support transcription independently of P-TEFb. Chromatin reading and transcription-factor coactivation are represented as two cooperating core units. A single histone-reader core captures the coupled multi-mark mechanism; the specific accepted source annotations retain mark-level detail without duplicating one mechanism as a parent and six child cores.
The paper on late-mitotic P-TEFb recruitment supports transcriptional regulation and elongation, while antibody perturbation in HeLa cells supports contextual G2/M regulation. The latter is retained as non-core, without assigning BRD4 a cell-cycle kinase activity. The chromatin-remodeling IBA is retained as a noncatalytic recruitment/organization contribution: BRD4 associates with remodeling partners and the studied isoform B recruits condensin II. Neither observation makes BRD4 an ATPase. The source's mouse Th17 ISS donor is preserved, but its exact lineage experiment remains unresolved.
Primary sources: late-mitotic recruitment, ET-domain effectors, chromatin insulation, anti-pause enhancers.
The human CD4+ T-cell source uses ChIP-seq to identify BRD4 occupancy at promoters and enhancers, relating occupancy to acetylated histones and Pol II transcription. The exact GO definition of transcription cis-regulatory region binding requires DNA-sequence binding. The source establishes chromatin association rather than direct sequence recognition by BRD4, so that source row is refined to chromatin binding, which is already supported independently. This does not assert that BRD4 can never contact DNA. The p53 study assigns sequence recognition to p53 and distinguishes it from regulated BRD4-p53 interaction.
Sources: human T-cell occupancy, GO:0000976, phospho-switch study. The official local GO mirror used for definitions is dated 2026-03-25; successful live AmiGO confirmation was obtained for NF-kappaB binding, not for every term.
The standing project curation instruction retains biologically supported generic protein binding as KEEP_AS_NON_CORE when no evidence-backed specific replacement is established. This is an intentional task-specific departure from the annotation-reviewer skill's informational-exclusion recommendation, not a change to global policy or external maintainer approval. Unread or unresolved target evidence remains UNDECIDED.
Three RelA-source rows are refined to NF-kappaB binding, supported by acetyl-Lys310 recognition and cellular interaction. TWIST1 binding is refined to bHLH transcription factor binding, with its acetylated motif and the tumor-cell context retained. GATA4 co-IP supports the more specific RNA polymerase II DNA-binding transcription factor binding term, without claiming purified binary contact. Viral E2, ET-domain partners and mixed P-TEFb/JMJD6 associations remain scoped to their observed roles. Correct generic interactions are not deleted merely for low informativeness.
Some screen rows remain unresolved because individual target tables or assays were not accessible. Independently established BRD4-JMJD6 and BRD4-E2 interactions are retained with explicit curator deference; corroboration from another paper does not certify an unread original screen edge. FAM90A1, SPDYE4 and SH3-screen target limits are preserved. The SPOP source explicitly distinguishes WT binding from loss with F133L; SPOP substrate-adaptor binding does not confer E3 ligase activity on BRD4.
The fuller authenticated ET-domain source was read as supplemental evidence without replacing its canonical abstract-only cache. Likewise, the SPOP-BRD4 statement comes from a Results passage in the fuller authenticated PMID:35512704 variant; the preserved canonical partial body omits that passage. No quotation from that supplemental passage is placed in the YAML. The ET-domain source supports endogenous and tagged co-IP, but not purified binary interactions for every partner. One NSD3 co-IP uses a short isoform; it is not treated as catalytic. The canonical GBAF source itself describes BRD4 association with GLTSCR1 and selected complex proteins. Complex association is not equivalent to a universal stable complex subunit assignment.
Sources: RelA recognition, RelA maintenance, Twist, JMJD6 structure, GBAF, cardiac interactomes.
Independent biochemical and recruitment evidence supports H4-reader function. The six specific reader rows from the Cornelia de Lange-like syndrome paper require finer distinctions. The inspected text supports H3K9ac and H3K27ac association, mutant chromatin occupancy and the reader mechanism; human isolated bromodomain preparations are distinguished from mouse cellular Y430C/Y431C experiments. No exclusive mark specificity or unmeasured human full-length affinity is asserted.
Four individual H4-mark results remain UNDECIDED because the exact Fig.2A array display was inaccessible. The H4K8ac reagent in Methods alone is not a positive result. These access limits do not negate the independent broad H4-reader studies. The two indexed official publisher correction notices concern corresponding authorship and consortium placement; neither reports an experimental correction or retraction.
Sources: reader study, accepted manuscript, 2018 correction, 2019 correction. Access included normal source text and indexed publisher correction paragraphs; direct figure/supplement inspection was not successful.
The 2012 BRD4 kinase report contains substantive positive biochemical evidence, including recombinant/in-gel and bacterial-expression controls. The 2021 follow-up adds positive mouse-protein assays; it shares investigators with the earlier work and is not independent-laboratory replication. An independent 2014 study found no phosphorylation using human BRD4 residues 1–698 expressed in E. coli or Sf21, while allowing preparation or activation requirements. Thus the source kinase IBA and CTD-kinase IMP remain UNDECIDED. The disputed catalysis is not dismissed merely because BRD4 lacks a conventional kinase domain or recruits CDK9. Physical CTD binding remains supported separately. The PAINT node and donor experiments were not reconstructed, and a self-source is not labeled circular.
The reported 2016 HAT activity also has substantive human/bacterial/in-gel controls. It is recorded as an unresolved question rather than a new HAT annotation or settled catalytic core. Recruitment or activation of p300, and BRD4-NUT fusion effects, must not be confused with intrinsic native BRD4 catalysis. A 2024 GEO update corrects sample/data associations; this is not a retraction or automatic biochemical refutation.
Sources and actual access: 2012 kinase — canonical abstract and indexed primary Results/captions; 2021 kinase — indexed primary Results/Discussion/captions; 2014 negative human-fragment experiment — indexed negative-assay paragraph and Fig.8 caption; 2016 HAT — indexed selected Results/Methods/captions; GEO update — official record. Direct PMC openings encountered access challenges. These are bounded primary-web readings, not normal caches, complete paper/image audits, raw purity-MS analysis or new source imports. The YAML cites normal cached sources only.
All 81 source assertions are assessed: 38 ACCEPT, 24 KEEP_AS_NON_CORE, six MODIFY and 13 UNDECIDED. No original source assertion is removed and no NEW annotation is added. Three products and all original reference identifiers/titles are preserved. Eleven distinct short supporting quotations are exact canonical substrings; two are reused in the core synthesis, yielding thirteen quote fields and at most 24 quoted words per source across the candidate. Full normal candidate validation, including terms, references and GOA, passed with 21 warnings and no errors. Nineteen warnings reflect the standing instruction to retain supported generic binding. One records the absence of a structured PAINT assessment for the unresolved kinase inference; no ancestral-node audit or failure classification is invented. The remaining warning is for the broad histone-reader core not appearing as an exact source term: existing child reader activities already cover this synthesis, so no redundant NEW assertion is added. DRAFT records these warnings and evidence limits rather than incomplete enumeration. Canonical application and publication remain separate from this completed candidate assessment.
Three additional primary papers are now available in the normal publication cache. This entry supersedes the earlier access statement for those three papers; the 2012 kinase cache remains abstract-only. The source assertions, all 81 actions, three products and two reader/coactivator cores remain unchanged. Both kinase assertions remain UNDECIDED because the reported experiments conflict under different species, construct and preparation conditions.
The original abstract explicitly reports direct CTD Ser2 phosphorylation [PMID:22509028 BRD4 is an atypical kinase that phosphorylates serine2 of the RNA polymerase II carboxy-terminal domain., "directly phosphorylates its serine 2"]. The independent negative experiment used human BRD4 residues 1–698 expressed in E. coli or Sf21 [PMID:24860166 Brd4 activates P-TEFb for RNA polymerase II CTD phosphorylation., "no phosphorylation activity was seen"]. Its Methods and negative-result paragraph were inspected in the normal XML cache. The authors allow additional preparation or activation requirements; the result is not a demonstration that every human BRD4 form lacks catalytic activity.
The later positive study uses mouse full-length BRD4 and truncations [PMID:34688663 The intrinsic kinase activity of BRD4 spans its BD2-B-BID domains., "WT and tΔC (1–730) efficiently phosphorylated full-length GST-CTD"]. The inspected Results and Methods distinguish Sf9 preparations, smaller bacterial fragments, and substrate-binding requirements. It shares investigators with the 2012 paper, so it is not independent-laboratory replication. These differences remain unresolved rather than being collapsed into either universal kinase activity or its absence. No PAINT node reconstruction, raw-image analysis or clean separation-of-function point-mutant result is claimed. Physical CTD binding and P-TEFb recruitment remain separate supported mechanisms. Primary links: 2012, 2014, 2021.
The HAT question now identifies the 2016 source [PMID:27159561 BRD4 is a histone acetyltransferase that evicts nucleosomes from chromatin., "BRD4 acetylated all four core histones"]. The inspected normal-cache Results and Methods include human and mouse preparations, bacterial expression, solution and in-gel assays, mutant controls and nucleosome experiments. These are substantive catalytic evidence, not an inference from colocalization or an argument based on domain architecture. Their scope still leaves endogenous human intrinsic catalysis versus recruited p300 activity open; no NEW HAT assertion or catalytic core is added. Primary study.
The official 2017 correction reports that Figure 2c blots had been duplicated from Figure 2b and that the HTML/PDF versions were corrected. This concerns the figure used in the mutant histone-binding control, so it is disclosed as an experimental-figure correction rather than an authorship correction. The publisher notice text was read; corrected images and raw purity-MS data were not independently audited. The notice is supplementary linked context, not a fabricated normal cache or a retraction claim.
The two kinase papers are cited with short verbatim anchors in both unresolved annotation decisions. Counting repetitions in this entry and the YAML, the new kinase anchors use 15 words per source for the 2012 and 2014 papers and 24 for the 2021 paper. The HAT anchor uses six words. All earlier quotations and historical notes remain intact. The narrower biochemical observations are recorded separately from the disputed general catalytic attribution.
The scoped follow-up passed independent scientific review and was applied to the canonical review after checking exact preimages. Full normal candidate and canonical validation passed without errors, retaining the same 21 expected warnings and DRAFT status. The standard codex/gpt-6 EDIT history was scaffolded and validated. All 81 source assertions and actions, three products, two cores, earlier quotations, raw sources, publication caches and prior history records were preserved. The three additional primary caches were imported separately before this application; this application changed no source cache. Earlier access and candidate-stage entries remain as historical records.