Oncofusion

start / graph / gene-brca1

Genemoderate confidence 3 of 5: One strong study, or several smaller consistent ones.Q32-brca1-hr-licensing

BRCA1 (P38398, 1,863 aa) - named in 19 entries of this graph and never written: the HR mediator that CTCF and the H3K9 machinery are now reported to license at a break, and the only cascade member with a human therapy-related-leukaemia signal after topoisomerase II therapy

also called BRCA1 · RNF53 · P38398 · BRCA1_HUMAN · Breast cancer type 1 susceptibility protein · BRCA1/BARD1 heterodimer component

BRCA1 is the famous breast-cancer gene, and what it actually does is run the accurate repair of broken DNA. A 2026 paper shows the silencing machinery this graph has been arguing about is what brings BRCA1 to a break, which matters because the drug this graph proposes giving to children would switch that machinery off during chemotherapy.

The same thing, for a scientist

**Identity.** P38398, `BRCA1_HUMAN`, **1,863 aa**, primary gene name BRCA1, single UniProt gene synonym RNF53, 33 PDB entries. RING-type zinc finger 24-65, a PALB2-interaction region 1397-1424, and tandem BRCT domains at **1642-1736** and **1756-1855** [E1]. `gene_exact:BRCA1 AND organism_id:9606 AND reviewed:true` returned **exactly one** reviewed record and its `gene_primary` reads back BRCA1 (rules 62/65 asserted, not assumed). **Why it enters this graph now.** BRCA1/BARD1 is the obligate heterodimer at the **HR licensing point, the DNA end resection step**, and a 2026 *Nucleic Acids Research* paper places it directly downstream of the two arms this graph has been arguing about for several cycles. In U2OS and HEK-293T cells, **CTCF recruits HP1-gamma, KAP1, SUV39H1, SETDB1 and KDM5A, and that repressive-chromatin assembly is what enables rapid accumulation of BRCA1/BARD1 at DSBs**; CTCF depletion abrogates it with loss of H3K9 methylation and H3K4 demethylation at breaks; and the consequence is abrogated EXO1/DNA2 recruitment, defective end resection and dysfunctional HR [E2]. The same paper distinguishes a **CTCF-dependent 'early' BRCA1/BARD1** pool, which is what recruits SMARCAD1 and EXO1, from a **CTCF-independent 'late'** pool that is dispensable for resection and acts post-resection [E2]. **The published counter-position, from the same field.** The founding CTCF-HR paper concluded the opposite topology: CTCF **acts downstream of BRCA1** in the HR pathway and associates with BRCA2 in a PARylation-dependent manner [E3]. The 2026 paper's early/late distinction is the reconciliation it offers, and it is offered by the paper that needs it. **The human signal, which is the part that bears on an intervention.** Two BRCA1 germline-mutated hereditary breast-ovarian carcinoma patients developed therapy-related acute leukaemia after **topoisomerase II-directed therapy**, one a t-APL carrying a RARA intron 2 breakpoint at chr17:40,347,487 consistent with TOP2-inhibition breakpoint clustering; the authors argue coincidence is unlikely because a previously reported case carried the *identical* BRCA1 mutation after the same primary disease [E4]. At population scale, **10 of 47 (21%)** breast-cancer survivors who developed therapy-related leukaemia carried a deleterious germline mutation in a susceptibility gene - BRCA1 3 (6%), BRCA2 2 (4%), TP53 3 (6%), CHEK2 1, PALB2 1 [E5]. **Neither result concerns KMT2A**, and [E5] reports no non-TRL comparison group, so this is a direction and not an effect size. **Dependency.** Mean gene effect **-0.489** across **1,258** screens, 44.0% of lines below -0.5, 4.7% below -1.0; in the 147 haematopoietic and lymphoid screens the mean is shallower at **-0.281** with 17.7% below -0.5 [E6]. So BRCA1 is a moderate, broad dependency that is *less* deep in blood lines than in the panel as a whole - the opposite of what a leukaemia-selective target looks like. **And nobody has ever put it at this graph's locus.** No published work localises BRCA1 at *KMT2A*, its breakpoint cluster region or its exon-11 CTCF element, in any cell type - but unlike SUV39H1, the archive **can** be asked: ENCODE holds **7 released BRCA1 TF ChIP-seq experiments including K562 and GM12878**, so this is `measured-not-analysed` rather than `never-measured` [E7, and `absence-no-resection-factor-localised-at-the-kmt2a-locus`].

What it looks like
Breast cancer type 1 susceptibility protein1,863 aa
RING-type zinc finger · residues 24–65Interaction with PALB2 · residues 1397–1424BRCTBRCT 1 · residues 1642–1736BRCTBRCT 2 · residues 1756–185505001,0001,500

No three-dimensional model is available for this protein, so the map above is the whole picture. That is itself informative: AlphaFold does not model chains this long, and the parts that have been crystallised are held under the entries for the individual domains.

Go deeper
Why a gene named in 19 entries had no entry, and what the hole cost

BRCA1 appears in nineteen node files of this graph and had no entry of its own until this beat. That is not a catalogue gap. Without an entry there was nowhere to record three things that change how two existing arguments read.

  1. The H3K9 arm and the CTCF arm converge on BRCA1. mark-h3k9me3 and mechanism-h3k9me1-2-ctcf-balance-gates-kmt2a-rearrangement predict opposite signs for the same intervention. 2 gives both a shared downstream node: CTCF recruits the H3K9 machinery, and the H3K9 machinery licenses BRCA1/BARD1. hyp-the-writer-reaches-the-kmt2a-anchor-only-through-ctcf-and-only-after-the-break already states that topology; what was missing was an entry for the thing at the bottom of it.
  2. The graph's proposed prophylaxis acts on this node. drug-unc0642 is proposed for prophylactic G9a/GLP inhibition alongside anthracycline or etoposide. 2 tested UNC-0638, chaetocin and BIX-01294 and reports that they disturb the immediate accumulation of H3K9me3, HP1-gamma and BRCA1/BARD1 at DSBs. See risk-an-h3k9-axis-inhibitor-given-with-a-top2-poison-removes-a-repair-licence.
  3. There is a human observation in the same shape. BRCA1 germline mutation plus topoisomerase II-directed therapy has produced therapy-related acute leukaemia with a TOP2-type breakpoint 4, and germline repair-gene mutations are found in 21% of therapy-related leukaemias after breast cancer 5. A drug that phenocopies partial BRCA1 loss for the duration of etoposide exposure is intervening on a state with an observed clinical outcome.
Early BRCA1 and late BRCA1 are two different objects, and only one of them is in this cascade

2's most useful structural claim is a split. The classical route - ATM-dependent RNF8/RNF168 ubiquitination, RAP80 recognition, ABRAXAS1-BRCA1 recruitment through the BRCT phospho-pocket - produces a pool of BRCA1/BARD1 the paper calls late. The CTCF/HP1-gamma/H3K9me route produces an early pool.

The experiments that separate them: ATM inhibition with Ku-55933 removes late BRCA1 while early BRCA1 remains, and ATM inhibition alone has no effect on end resection although it markedly reduces HR efficiency; concurrent ATM inhibition with CTCF or HP1-gamma depletion does not worsen resection or HR beyond CTCF or HP1-gamma depletion alone. The authors read the non-additivity as late BRCA1/BARD1 being dispensable for resection itself and acting at a post-resection step 2.

Two consequences for this graph. UniProt's own coded statement about where BRCA1 goes at a break names ABRAXAS1 and the BRCA1-A complex - that is the late pool, so the database annotation is about the arm this cascade is not. And an experiment that measures 'BRCA1 foci' without separating the two pools cannot test the hypothesis this beat is about.

What is NOT established here
  • No haematopoietic cell. Every mechanistic experiment is U2OS or HEK-293T with engineered breaks 23.
  • No KMT2A locus. BRCA1[tiab] AND KMT2A[tiab] returns 5, BRCA1[tiab] AND MLL[tiab] returns 8, BRCA1[tiab] AND 11q23[tiab] returns 4, and title-checking all of them finds zero that place BRCA1 at this locus. Counts and querytranslations are recorded in absence-no-resection-factor-localised-at-the-kmt2a-locus.
  • No fusion. And the one controlled test of whether a t(4;11) product changes DSB repair was negative on both reporters.
  • A rival topology is on record, from the paper that founded the CTCF-HR link 3.
Facts and figures
uniprotP38398
uniprot entry nameBRCA1_HUMAN
gene symbolBRCA1
gene primary readbackBRCA1
protein nameBreast cancer type 1 susceptibility protein
length AA1863
n PDB entries33
how the accession was resolvedgene_exact:BRCA1 AND organism_id:9606 AND reviewed:true, fetched live 2026-09-22, returned EXACTLY ONE reviewed human record; gene_primary read back as BRCA1 and length asserted at 1,863 aa so verify-accessions.py can catch a swap. No PDB polymer entity was fetched, so no construct, span or resolution claim is made from the 33 deposits.
identifier trap checkedRule 88 registry read run BEFORE the first query, over d['collisions'] (177 rows), case-sensitive: the only row matching 'BRCA1' is the BACH1 row, where BACH1 is an alias of BOTH O14867 (bZIP transcription factor, 736 aa) and Q9BX63 (BRIP1/FANCJ, 1,249 aa, 'BRCA1-associated C-terminal helicase 1'). That row is about BACH1, not about BRCA1, and BRCA1 itself has no row. A NEW trap found in this beat and reported for a row: BARD1[tiab] AND MLL[tiab] returns PMID 19704708, 'BARD1 may be renamed ROW1 because it functions mainly as a REPRESSOR OF WUSCHEL1' - a PLANT gene where both BARD1 and MLL mean something else entirely.
localizationNucleus [ECO:0000269, PubMed:15133502, 17525340, 20160719, 21144835, 26778126, 9528852]; Chromosome [ECO:0000269 x3, PubMed:23269703, 25472942, 26778126]; Cytoplasm [ECO:0000269, PubMed:20160719 and 8972225] - the cytoplasmic pool is condition-specific, described as translocation during UV-induced apoptosis. Rule 91 applies: these are pointers, and the location most relevant to this beat is stated in an UNCODED note.
subnuclear localizationUniProt's uncoded note is the relevant statement: BRCA1 'localizes at sites of DNA damage at double-strand breaks (DSBs); recruitment to DNA damage sites is mediated by ABRAXAS1 and the BRCA1-A complex (PubMed:26778126)'. That is the ATM/RNF8/RNF168/RAP80-dependent route - which in [E2]'s terms is the LATE pool. The compartment [E2] adds is a CTCF- and HP1-gamma-dependent EARLY pool at the same physical site, reached through H3K9-methylated chromatin rather than through ubiquitin. NONE of this is at the KMT2A locus, and there is no imaging of BRCA1 relative to any nuclear compartment in a KMT2A-rearranged cell.
localization shown for fusionfalse
cell typeNo KMT2A::AFF1 cell, no infant material, no B-lineage progenitor. The cascade evidence is U2OS osteosarcoma reporter lines (AsiSI-ER, ER-mCherry-LacI-FokI-DD, DR-GFP, EJ5-GFP) and HEK-293T [E2]; the founding CTCF-HR paper is also non-haematopoietic [E3]; the human leukaemia evidence is ADULT therapy-related leukaemia after breast or ovarian cancer therapy [E4, E5]. The DepMap panel includes 147 haematopoietic and lymphoid screens [E6] and the ENCODE experiments include K562 (CML blast crisis) and GM12878 (EBV-transformed B-lymphoblastoid) [E7] - neither of which is KMT2A-rearranged.
cell stateTWO states gate everything about this protein and neither is the index disease's state. (1) **CELL CYCLE.** HR requires a sister chromatid, so BRCA1/BARD1-dependent resection is S/G2-restricted; BARD1's own UniProt note says the BRCA1/BARD1 foci form 'during S phase of the cell cycle'. The putative cell of origin of infant t(4;11) is a fetal progenitor, and a quiescent progenitor may not run this cascade at all. (2) **DNA DAMAGE.** The CTCF-dependent early pool is a minutes-scale transient event [E2], and the older KAP1/HP1/SUV39H1 loading is stripped off again when ATM phosphorylates KAP1. So BRCA1 occupancy at any locus is a function of damage state and of cell-cycle phase, and a resting-state measurement measures neither.
epigenetic readoutBRCA1 is neither a writer nor an eraser here; it is a chromatin-state-DEPENDENT EFFECTOR, and that dependency is the content. Its early recruitment requires H3K9me2/me3 at the break (written by SUV39H1/SETDB1, scaffolded by KAP1, read by HP1-gamma) and H3K4 DEMETHYLATION at the break (by KDM5A); inhibitors of H3K9 methylation - UNC-0638, chaetocin and BIX-01294 - disturb the immediate accumulation of H3K9me3, HP1-gamma and BRCA1/BARD1 but NOT of CTCF [E2].
mark read indirectlyH3K9me2/H3K9me3 via HP1-gamma; the H3K4me0 state via KDM5A demethylation
shown for fusionNO - nobody has looked, in any fusion. Neither [E2] nor [E3] involves a fusion protein or a leukaemia; the two human leukaemia sources [E4, E5] are therapy-related leukaemias with no KMT2A rearrangement reported in the abstracts, and one of them is a PML-RARA t-APL. The single controlled test in this graph of whether a KMT2A fusion changes DSB repair found NO difference from wild type for either t(4;11) product on both HR and NHEJ reporters (`concept-neither-t411-product-changes-dna-damage-repair`), so a fusion-driven BRCA1 phenotype is not merely unmeasured - it has been looked for and not found in one system.
archive bound 2026 09 22ENCODE released TF ChIP-seq, per target label: BRCA1 7 experiments (GM12878, H1, HeLa-S3, HepG2, K562). EXO1, DNA2, SUV39H1 and RBBP8 return HTTP 404 on both the filtered search and the bare target record, i.e. no ENCODE target record exists at all. **BARD1 has been removed from that list, corrected 2026-09-22** - /targets/BARD1-human/ returns 200 and its 404 was a property of the assay_title filter. CBX3 4 (GM12878, HCT116, K562), CBX1 2 (HepG2, K562), CBX5 4, KDM5A 3 (A549, H1, HepG2 - no blood line), CTCF 502, TRIM28 5, SETDB1 5.
absence classmeasured-not-analysed
absence class reasoningFor BRCA1 specifically the archive route is OPEN - two blood-lineage ENCODE ChIP-seq datasets exist and have never been read at the KMT2A locus - so closing this absence needs compute and no wet work. That is a different and much cheaper task than the SUV39H1 case in `absence-no-h3k9-writer-or-reader-localised-at-the-kmt2a-locus`, which is never-measured.
domains
RING-type zinc finger24.0RING-type zinc finger: 24Interaction with PALB21,397Interaction with PALB2: 1397BRCT 11,642BRCT 1: 1642BRCT 21,756BRCT 2: 1756
gene synonyms in uniprot
[
  "RNF53"
]
domains — the numbers behind the chart
[
  {
    "name": "RING-type zinc finger",
    "start": 24,
    "end": 65
  },
  {
    "name": "Interaction with PALB2",
    "start": 1397,
    "end": 1424
  },
  {
    "name": "BRCT 1",
    "start": 1642,
    "end": 1736
  },
  {
    "name": "BRCT 2",
    "start": 1756,
    "end": 1855
  }
]
depmap
{
  "n_screens_with_gene_effect": 1258,
  "mean_gene_effect": -0.4888,
  "median_gene_effect": -0.4622,
  "min_gene_effect": -1.8413,
  "max_gene_effect": 0.3641,
  "frac_below_minus_0_5": 0.4404,
  "frac_below_minus_1_0": 0.0469,
  "n_blood_screens": 147,
  "blood_tissue_blocks": [
    "lymphoid tissue",
    "hematopoietic tissue"
  ],
  "blood_mean_gene_effect": -0.2808,
  "blood_frac_below_minus_0_5": 0.1769,
  "n_tissue_blocks_flattened": 29,
  "open_targets_data_version": {
    "year": "26",
    "month": "06",
    "iteration": null
  },
  "depmap_release": null,
  "rule_83": "the Open Targets depMapEssentiality payload carries NO DepMap release identifier. What is recorded is the Open Targets data version and it is not a substitute. The screen count is gene-specific and must not be read as a corpus denominator.",
  "rule_84": "no median-split proportion is reported. The fractions are against absolute Chronos thresholds, and no claim is made that BRCA1 falls in any half of any ranking.",
  "live_zero_screen_control": "POLR2A was fetched in the SAME call and returned 0 tissue blocks and 0 screens with its approvedSymbol reading back as POLR2A - so an empty payload here would have looked like 'no dependency' when it means 'no data'. The control fired, so the BRCA1 numbers come from a populated row.",
  "reading": "moderate and broad, and SHALLOWER in blood (-0.281) than in the panel (-0.489). Nothing about this figure supports BRCA1 as a leukaemia target; it is recorded because the cascade cannot be perturbed without knowing the fitness cost."
}
How this could bite us back
  • BRCA1 is not a target here - it is the thing an intervention elsewhere in the cascade would remove. Any argument that reaches BRCA1 through CTCF or the H3K9 writers is an argument for NOT giving that intervention during genotoxic therapy, not for a new drug.
  • CTCF, the top of the cascade, is a near-total pan-essential in DepMap - mean -1.826, 99.76% of 1,258 lines below -0.5 - so the upstream node cannot be pharmacologically modulated for benefit, only damaged.
  • The prophylaxis this graph proposes (drug-unc0642, G9a/GLP) acts on the middle of this cascade, and the sibling compound UNC-0638 is reported in [E2] to block BRCA1/BARD1 accumulation at breaks. Removing an HR licence while TOP2 breaks are being made in a child on curative therapy is the specific liability.
  • Conversely, if this graph ever proposed ENHANCING the cascade to protect a locus: the BRCA1-A complex RESTRICTS resection to prevent excessive resection, so more BRCA1 is not monotonically better repair and over-resection is its own mutagenic outcome.
Still open
  1. ANSWERED IN CYCLE 63, AND THE ANSWER IS NO. All 13 released ENCODE BRCA1, CBX3 and CBX1 ChIP-seq tracks were read at seven KMT2A intervals: ZERO called IDR peaks in 13 of 13, from peak sets holding 280-26,789 peaks genome-wide, and the BCR core reads 0.566-1.235 fold over the genome-wide median of same-width tiles. The exon-11 element looked enriched in two BRCA1 tracks (2.05x, top 3.5% in HeLa-S3) and fell to 0.34-1.25x against 6,000 CTCF anchors of the same cell type, below the anchor median in 10 of 13. A conservative-IDR CTCF peak IS on the element in all six biosamples, so the anchor is there and the resection factors are not. See ana-q35-brca1-cbx3-and-cbx1-are-not-at-the-kmt2a-breakpoint-cluster-in-any-encode-cell. Resting state only - the damage state remains unmeasured in any cell type. Original text: Is BRCA1 present at the KMT2A exon-11 CTCF element in an undamaged blood cell? ENCODE's K562 and GM12878 BRCA1 ChIP-seq can answer this with compute and no wet work, and nobody has looked.
  2. Does a break at the KMT2A breakpoint cluster region recruit early (CTCF/HP1-gamma-dependent) BRCA1/BARD1 in a haematopoietic progenitor, and is that recruitment lost under UNC0642? No damage-state BRCA1 measurement exists at this locus in any cell type.
  3. Is the BRCA1-plus-TOP2-poison therapy-related leukaemia signal enriched for 11q23 rearrangements specifically, or is it partner-agnostic illegitimate recombination? The therapy-related-leukaemia registries that produced [E5] hold the karyotypes and the question has not been asked in those words.
  4. Does the early/late BRCA1 split reproduce outside the group that proposed it? It reconciles [E2] with [E3], and a reconciliation offered by one of the two parties needs independent replication before this graph builds on it.
Every source behind this entry (7)
  1. E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

    BRCA1 is UniProt P38398, entry name BRCA1_HUMAN, 1,863 amino acids, primary gene name BRCA1 with the single gene synonym RNF53; the feature table annotates a RING-type zinc finger at 24-65, a PALB2-interaction region at 1397-1424 and tandem BRCT domains at 1642-1736 and 1756-1855, and the record cross-references 33 PDB entries; coded subcellular locations are Nucleus (six ECO:0000269 citations), Chromosome (three) and Cytoplasm, with an uncoded note stating that BRCA1 localizes at sites of DNA damage at double-strand breaks and that recruitment to damage sites is mediated by ABRAXAS1 and the BRCA1-A complex.

    What this evidence cannot show

    A database record, fetched 2026-09-22 with gene_exact:BRCA1 AND organism_id:9606 AND reviewed:true returning exactly one reviewed record. Rule 91 applies to every coded location: ECO:0000269 means a curator read an experiment in that paper, not that the paper is about BRCA1's localisation. No PDB polymer entity was fetched, so nothing is claimed about what the 33 deposits cover.

    Evidence that argues against this (1)
    • UniProt's coded locations name no double-strand-break compartment at all; the DSB statement is in an UNCODED note. A filter on ECO:0000269 would keep 'Nucleus' and discard the only sentence in the record that is about this beat - the same failure rule 91 documents for KMT2A's fragment comments. UniProt Consortium 2026
  2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    In U2OS and HEK-293T cells, CTCF recruits the repressive chromatin proteins HP1-gamma, KAP1, SUV39H1, SETDB1 and KDM5A, thereby enabling rapid accumulation of BRCA1/BARD1 at DSBs; CTCF depletion abrogates their enrichment with loss of H3K9 methylation and H3K4 demethylation at DSBs and defective rapid BRCA1/BARD1 recruitment; depriving any one repressive component, or deregulating H3K9 methylation/H3K4 demethylation, compromises the immediate accumulation of the others and of BRCA1/BARD1; consequently EXO1/DNA2 recruitment is abrogated, causally linked to defective DNA end resection and dysfunctional HR. Read in the full text: CTCF associates with BARD1 but rarely BRCA1 without damage and 50 micromolar etoposide for 1 h increases both; CTCF interacts preferentially with HP1-gamma, modestly with HP1-alpha and weakly with HP1-beta; the HP1-binding PxVxL motif of BARD1 is at residues 569-573 and the PEELI mutant of it largely loses binding to the chromoshadow domain of ALL HP1 subtypes; the inhibitors UNC-0638, chaetocin and BIX-01294 disturb the immediate accumulation of H3K9me3, HP1-gamma and BRCA1/BARD1 but not of CTCF at DSBs; and ATM inhibition alone does not affect end resection while markedly reducing HR efficiency, with CTCF or HP1-gamma co-depletion giving no further loss, which the authors read as late BRCA1/BARD1 being dispensable for resection and acting post-resection.

    CTCF primes repressive chromatin for rapid BRCA1/BARD1 recruitment to drive homologous recombination repair of DNA double-strand breaks. · Hwang SY, Choi D, Choi MJ, Jo HR, Jeong JH, Kang MA, Lee JS · Nucleic Acids Research 2026
    PMID 42475683DOI
    What this evidence cannot show

    System bounds, read off the methods and not inferred: U2OS osteosarcoma reporter lines (AsiSI-ER, ER-mCherry-LacI-FokI-DD, DR-GFP, EJ5-GFP) and HEK-293T. Damage modalities are FokI-induced clustered breaks at a Lac operator array, AsiSI restriction, laser micro-irradiation, phleomycin, and etoposide at 50 micromolar for 1 h used for the co-immunoprecipitations - so a TOP2 poison IS used in this paper, for the biochemistry rather than for the recruitment imaging or the HR reporter. NO haematopoietic cell, no fusion protein, no KMT2A locus. Received 2025-04-29, accepted 2026-06-29. The UNC-0638/chaetocin/BIX-01294 result is in Supplementary Fig. S5E-G, i.e. a supporting panel, invisible to any title or abstract query. Chaetocin is a nonspecific histone lysine methyltransferase inhibitor and a thioredoxin-reductase-1 inhibitor (see gene-suv39h1 E14/E15), so it adds little independent evidence; UNC-0638 and BIX-01294 are the interpretable arms. DOI extracted from PubmedData/ArticleIdList only - the loose .//ArticleId path on this record returns 191 id elements from the reference list.

    Evidence that argues against this (1)
    • The founding CTCF-HR paper concluded the OPPOSITE topology: CTCF acts DOWNSTREAM of BRCA1 in the HR pathway and associates with BRCA2 in a PARylation-dependent manner, enhancing BRCA2 recruitment, while not influencing 53BP1 or LIGIV. The 2026 paper's early/late BRCA1 split is the reconciliation, and it is proposed by the paper that requires it rather than by an independent group. Hilmi K et al. 2017
  3. E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    CTCF is recruited to double-strand breaks through its zinc finger domain independently of PARP-1-catalysed PARylation, ensures proper DSB repair kinetics after gamma-irradiation, and its loss compromises HR-mediated repair with hypersensitivity to DNA-damaging agents and to PARP inhibitors; mechanistically the authors place CTCF DOWNSTREAM of BRCA1 in the HR pathway and show it associates with BRCA2 in a PARylation-dependent manner, enhancing BRCA2 recruitment to DSBs, while not influencing recruitment of the NHEJ proteins 53BP1 or LIGIV.

    CTCF facilitates DNA double-strand break repair by enhancing homologous recombination repair. · Hilmi K, Jangal M, Marques M, Zhao T, Saad A, et al. · Science Advances 2017
    PMID 28560323DOI
    What this evidence cannot show

    Abstract read; full text not fetched in this beat. This is the paper establishing the CTCF-HR link on which [E2] builds, and its stated topology conflicts with [E2]'s. Neither paper is in a haematopoietic cell and neither touches the KMT2A locus.

    Evidence that argues against this (1)
    • [E2] contradicts this paper's ordering and offers early-versus-late BRCA1 as the resolution; until an independent group reproduces the split, the direction of the CTCF-BRCA1 arrow is contested in the primary literature. Hwang SY et al. 2026
  4. E4weak confidence 2 of 5: Preliminary, small, indirect, or not yet peer reviewed.paper

    In two hereditary breast-ovarian carcinoma patients carrying germline BRCA1 mutations, therapy-related acute leukaemia developed following topoisomerase II-directed therapy - one therapy-related acute promyelocytic leukaemia whose genomic characterisation revealed a RARA intron 2 breakpoint at chr17:40,347,487, corroborating breakpoint clustering in t-APL after topoisomerase II inhibition, and one therapy-related ALL; the germline BRCA1 mutation in the t-APL case was in exon 8 (c.512dupT) and the t-ALL case carried c.68_69delAG (187delAG), identical to a previously reported case after treatment of the same primary disease, which the authors argue makes chance unlikely.

    Characterization of therapy-related acute leukemia in hereditary breast-ovarian carcinoma patients: role of BRCA1 mutation and topoisomerase II-directed therapy. · Bagal B, Kumar R, Gaur T, Talreja V, Bonda A, Patkar N, Shetty D, Kowtal P, Subramanian PG, Gupta S, Sarin R, Hasan SK · Medical Oncology 2020
    PMID 32277283DOI
    What this evidence cannot show

    n = 2, a case characterisation with no comparison group and no incidence denominator, in ADULTS after breast or ovarian cancer therapy. Neither leukaemia is KMT2A-rearranged in the abstract - one is a PML-RARA t-APL - so this is evidence about BRCA1 plus a TOP2 poison producing a TOP2-type illegitimate recombination, not about KMT2A specifically. The authors' own framing is that it 'warrants further studies'. Germline BRCA1 loss is also not the same perturbation as transient pharmacological loss of an upstream chromatin licence.

    Evidence that argues against this (1)
    • The same literature contains therapy-related myeloid neoplasms after PARP inhibitors in germline BRCA1 carriers, where the leukaemogenic exposure is a synthetic-lethal agent rather than a TOP2 poison - so germline BRCA1 status is associated with therapy-related leukaemia across drug classes and the TOP2 specificity in [E4] is not isolated by design. Tariq H et al. 2023
  5. E5moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    Among breast-cancer survivors who developed therapy-related leukaemia after cytotoxic therapy, inherited mutations in cancer susceptibility genes are common: of 47 subjects with available germline DNA screened on a comprehensive breast-cancer susceptibility panel, 10 (21%) carried a deleterious inherited mutation - BRCA1 in 3 (6%), BRCA2 in 2 (4%), TP53 in 3 (6%), CHEK2 in 1 (2%) and PALB2 in 1 (2%); 19 of 88 survivors with therapy-related leukaemia (22%) had an additional primary cancer and 40 of 70 with an available family history (57%) had a close relative with breast, ovarian or pancreatic cancer.

    PMID 26641009DOI
    What this evidence cannot show

    A single-institution registry series, n = 47 sequenced of 88 identified, in ADULT breast-cancer survivors. The abstract reports NO non-TRL comparison group and no population base rate, so the 21% cannot be converted into a relative risk here - rule 84 applies and this is a proportion without its null. Five different genes with 1-3 carriers each; no gene-level inference is defensible. The cytotoxic exposures are not resolved to TOP2 poisons in the abstract.

    Evidence that argues against this (1)
    • The same series identified TP53 in as many subjects as BRCA1 (3 each), and TP53 is not an HR gene - so the association is with cancer-susceptibility genotype broadly rather than with homologous recombination specifically, which weakens any reading of this series as evidence about the resection cascade. Churpek JE et al. 2016
  6. E6strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

    BRCA1 is a moderate, broad CRISPR dependency and is shallower in blood lines than in the panel: flattening every tissue block of the Open Targets depMapEssentiality payload gives 1,258 screens with a gene effect, mean -0.4888, median -0.4622, minimum -1.8413, maximum 0.3641, with 44.0% of lines below -0.5 and 4.7% below -1.0; restricted to the 147 screens in the hematopoietic and lymphoid tissue blocks the mean is -0.2808 with 17.7% below -0.5. Fetched in the same call: BARD1 mean -0.5742 over 1,258 screens (58.2% below -0.5, blood mean -0.4533), CBX3 -0.0817, CBX1 -0.3494, CBX5 -0.0830, KDM5A -0.1185, EXO1 -0.1273, and CTCF -1.8260 with 99.76% of lines below -0.5 and 96.9% below -1.0.

    Open Targets Platform GraphQL API, target.depMapEssentiality with all tissue blocks flattened, plus meta.dataVersion · Open Targets Platform; Broad Institute DepMap · Open Targets Platform 2026
    What this evidence cannot show

    Rule 83: the payload names NO DepMap release; the Open Targets data version is year 26, month 06, iteration null, and that is not a release identifier. Screen counts are gene-specific and must not be read as a corpus denominator - DNA2 in the same call returns only 80 screens across 17 blocks, a sixteen-fold smaller denominator, which is why its mean is not comparable with these (see `ana-q32-depmap-across-the-hr-licensing-cascade`). Rule 84: no median-split proportion is reported. The live zero-screen control fired - POLR2A returned 0 blocks and 0 screens in the same call with a passing approvedSymbol readback. A CRISPR knockout is a total, permanent loss over a two-week screen and is not a model of transient pharmacological inhibition.

    Evidence that argues against this (1)
    • A gene-effect score cannot distinguish 'dispensable' from 'buffered by a redundant pathway', and for a repair gene the screen is run in the absence of exogenous genotoxic stress - precisely the condition under which an HR gene is least needed. The shallow blood mean is therefore uninformative about what BRCA1 loss costs a progenitor during etoposide exposure. Open Targets Platform; Broad Institute DepMap 2026
  7. E7strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

    The archive route for BRCA1 occupancy at any locus is OPEN, unlike that for SUV39H1: ENCODE holds 7 released TF ChIP-seq experiments for target label BRCA1, in GM12878, H1, HeLa-S3, HepG2 and K562 - two of them blood-lineage cells. In the same census, CBX3 has 4 (GM12878, HCT116, K562), CBX1 2 (HepG2, K562), CBX5 4, KDM5A 3 (A549, H1, HepG2 - no blood line), CTCF 502, TRIM28 5 and SETDB1 5, while EXO1, DNA2, SUV39H1 and RBBP8 return HTTP 404 on both endpoints, meaning no ENCODE target record exists for them at all. **BARD1 was listed here and is removed, corrected 2026-09-22**: bare target record 200, filtered search 404, the same pattern as RAD50.

    What this evidence cannot show

    Counted 2026-09-22 on the released TF ChIP-seq facet only; other assay titles and other archives (GEO individual depositions, ChIP-Atlas, ReMap) were NOT swept, so this bounds ENCODE and not the archive world. A 404 on the facet is read here as 'no target record', which is the honest reading of the API's behaviour but is not the same as a searched zero. K562 is CML blast crisis and GM12878 an EBV-transformed B-lymphoblastoid line; neither is KMT2A-rearranged and neither is a fetal progenitor.

    Evidence that argues against this (1)
    • An existing ChIP-seq dataset is not automatically able to answer this question: BRCA1 occupancy in an undamaged cultured line measures the resting state, and the cascade in [E2] is damage-conditional and minutes-scale. An ENCODE re-analysis at the KMT2A locus could return a clean, well-powered zero that is uninformative about the hypothesis, exactly as Q108's resting-state H3K9me3 null was. Hwang SY et al. 2026
Connected to