The measurement that settles the CTCF dispute: degron-grade CTCF removal versus cohesin removal, with DNA breaks AND rearrangement outcome read out at the same 1 kb of KMT2A in the same cells, reported as rearrangements per break
also called the settling experiment · rearrangements per break at the exon-11 anchor · CTCF versus RAD21 degron at KMT2A
Two sets of papers disagree about whether the CTCF protein makes the KMT2A gene break or stops it rearranging, and neither set measured both things. This experiment removes the protein cleanly and counts both, in the same cells, at the same spot.
The same thing, for a scientist
The dispute adjudicated in `concept-ctcf-at-kmt2a-fragile-and-guarded-is-one-element` survives because no study measures DNA break formation and rearrangement outcome at the same element in the same cells: the fragility literature maps breaks and never reports an inherited event, and the protective paper reports FISH copy gain and break-apart and contains no double-strand-break assay at all [E1][E2][E3]. This experiment closes exactly that gap. **Auxin-inducible degron on CTCF and, in parallel, on RAD21** - the same instrument Canela et al. used for RAD21 in human HCT-116, so the two literatures share a method [E2] - in a human cell line plus a primary CD34+ arm, with and without doxorubicin or etoposide, and a CTCF re-expression rescue. **Four readouts at GRCh38 chr11:118,488,159-118,489,206, at matched timepoints:** END-seq or low-input DSBCapture for breaks, on the same scale as Canela's and Atkin's [E2][E3]; CTCF/RAD21/SMC3/TOP2B ChIP-qPCR to confirm what was removed; FISH break-apart plus locus copy number as Gray measured, at 12 h for the transient event and 72 h for the inherited one [E1]; and breakpoint-junction sequencing so an event is a sequence rather than a probe separation. **The derived quantity is the point: rearrangements per break at this element, which nobody has ever reported for any factor in any system.** The prediction that distinguishes the leading reconciliation from both simple models is *fewer breaks and more rearrangements* after CTCF loss.
Design
Arms. (1) CTCF-mAID degron, auxin versus vehicle. (2) RAD21-mAID degron, auxin versus vehicle - the arm that separates 'CTCF is the repair chaperone' from 'the whole anchor is', and the arm that makes the result commensurable with Canela et al. 2019, who used precisely this system in HCT-116 2. (3) CTCF degron plus doxycycline-inducible degron-resistant CTCF, as a rescue. (4) siCTCF, included deliberately as a bridge control to the published result so that a disagreement between siRNA and degron is visible rather than inferred 1. (5) Vehicle plus doxorubicin, and vehicle plus low-dose etoposide, crossed with arms 1-3.
Cells. A degron-tagged human line for tractability - HCT-116 is the precedented host for the RAD21 degron 2 - plus a primary human CD34+ arm, because every chromatin measurement on the protective side is currently in RPE epithelium, every directional experiment on the fragility side is in mouse B cells, MEFs or colorectal cells, and this graph's index disease is B-lineage infant leukaemia 12. If the anchor turns out to be unoccupied in CD34+ cells, that is itself the headline result and it should be published as one.
Timepoints. Matched to the phenotype's own kinetics: Gray et al. report the copy gain as transient at 12 h and inherited at 72 h 1, while END-seq after a 30-minute etoposide pulse is the standard window for TOP2-induced breaks 2. Breaks and outcome therefore need different sampling times in the same experiment, and the ratio must be computed across them with the sampling stated - the commonest way this design would fail silently.
Controls that decide whether the answer means anything.
- Viability, cell cycle and mitotic index in every arm. CTCF is near pan-essential - this project's own DepMap scan gives -1.827 across all lines - so a rearrangement counted in a dying, mis-segregating cell is uninterpretable. This is the existing objection to the published siCTCF arm and it must not be inherited.
- A control locus with a comparable CTCF+cohesin anchor and no rearrangement history, to show the effect is not genome-wide.
- Occupancy confirmation at the element, not just a western blot. Lehman et al. showed CTCF protein can fall while genomic binding is largely unchanged, so protein level is not occupancy 4.
- Junction sequences, classified. Gomez-Herreros et al. found TDP2-dependent end-joining creates a specific therapy-related translocation subclass with 4 bp of perfect junction homology 5; a change in junction class with no change in event count would be a real result that a FISH readout cannot see.
What each outcome would mean
| CTCF degron: breaks | rearrangements | ratio | reading |
|---|---|---|---|
| down | up | up | The leading reconciliation is right. Occupancy causes the break and guards its repair; both literatures are correct and the graph should hold one mechanism. |
| unchanged | up | up | The fragility model is a cohesin model. CTCF's role at this element is repair-side only; confirm that breaks fall in the RAD21 arm. |
| up | up | flat | The 48-fold CTCF+RAD21 fragility class does not apply to this element. Occupancy is protective at both steps; Gray wins outright and the loop-anchor model needs a locus-specific exception. |
| down | down | flat | Gray's siCTCF result does not survive a degron. Compare arm 4: if siCTCF reproduces the published rise and the degron does not, the published result is a knockdown artefact. |
| no occupancy change after auxin | - | - | the degron failed; do not interpret anything else. |
| nothing moves in CD34+ because the anchor is unoccupied there | - | - | the dispute is epithelium-specific, and the occupancy map in a haematopoietic progenitor is the paper. |
The RAD21 arm, read against the CTCF arm. If both degrons lower breaks but only CTCF raises rearrangements, CTCF has a repair function at this anchor separable from cohesin's structural one. If both raise rearrangements, the reconciliation generalises to the anchor rather than the protein, and it predicts that cohesin-mutant leukaemias should show altered KMT2A rearrangement rates or junction structures - a cohort question answerable from existing data.
Why this has not been done, and why it is cheap
Not because it is hard. END-seq, low-input DSBCapture, auxin degrons, FISH break-apart and junction sequencing are all published and all in routine use by the groups involved. The element is 1 kb and needs no genome-wide analysis to read out.
It has not been done because no group owns both readouts. The DSB-mapping groups do not count inherited copy gains; the amplification group does not map breaks - measured directly, the word 'DSB' and the words 'double-strand break' do not appear in the protective paper, nor do 'cohesin', 'RAD21' or 'SMC3' 1. Five papers across Cell, Molecular Cell and Genes Chromosomes Cancer leave the question open with no conflicting number between them, because the overlapping measurement was never taken.
The single most informative sub-experiment, if only one thing can be done: END-seq at this element after a CTCF degron, in a cell where the anchor is confirmed occupied. It is one arm, one readout, and it decides proposition (i) for CTCF rather than for cohesin - which is the specific hole in the fragility side's directional evidence.
| design | Auxin-inducible degron removal of CTCF and, in parallel, of RAD21, in a degron-tagged human line (HCT-116 precedented for RAD21) plus a primary human CD34+ arm, crossed with doxorubicin and low-dose etoposide, with a degron-resistant CTCF rescue arm and a deliberate siCTCF bridge control to the published result. |
|---|---|
| readout | At GRCh38 chr11:118,488,159-118,489,206, in the same cells at matched timepoints: (1) END-seq or low-input DSBCapture break mapping; (2) CTCF, RAD21, SMC3 and TOP2B ChIP-qPCR to confirm what was removed; (3) FISH break-apart plus locus copy number at 12 h and 72 h; (4) breakpoint-junction sequencing with junction-class assignment. Primary derived endpoint: REARRANGEMENTS PER BREAK at this element - a quantity never reported for any factor in any system. |
| controls | Viability, cell-cycle distribution and mitotic index in every arm, because CTCF is near pan-essential (this project's DepMap scan: -1.827 across all lines) and the published siCTCF arm is open to exactly this objection. A control CTCF+cohesin anchor with no rearrangement history, to show the effect is locus-specific rather than genome-wide. Occupancy confirmed by ChIP at the element and not inferred from western blot. Degron efficiency and kinetics reported. Timepoints for break and outcome readouts stated explicitly, because they necessarily differ and the ratio is computed across them. |
| cost tier | med |
| timeline | 6-9 months in a lab that already runs END-seq or DSBCapture and has a degron line; the CD34+ arm adds the donor material and the primary-cell optimisation and is the part that could double it. |
| minimal version | END-seq at this element after a CTCF degron in a cell where the anchor is confirmed occupied. One arm, one readout, and it decides whether proposition (i) holds for CTCF rather than only for cohesin - the specific hole in the fragility side's directional evidence. |
| why nobody has done it | No group owns both readouts. The break-mapping groups do not count inherited copy gains; the amplification group maps no breaks - the words 'DSB', 'double-strand break', 'cohesin', 'RAD21' and 'SMC3' do not appear in that paper at all. Five papers leave the question open with no conflicting number between them, because the overlapping measurement was never taken. |
| lineage note | The CD34+ arm is not decoration. Every chromatin measurement on the protective side is in RPE epithelium, every acute directional experiment on the fragility side is in mouse fibroblasts, mouse B cells or human colorectal cells, and no CTCF occupancy or breakage measurement at this element exists in any B-lineage cell, in an infant, or in a KMT2A::AFF1 context. Generalising from those systems to infant t(4;11) is this graph's commonest documented synthesis failure. |
decision table
[
{
"ctcf_degron_breaks": "down",
"rearrangements": "up",
"ratio": "up",
"reading": "formation and fate are separately controlled and occupancy does both - the leading reconciliation; both literatures are right"
},
{
"ctcf_degron_breaks": "unchanged",
"rearrangements": "up",
"ratio": "up",
"reading": "the fragility model is a cohesin model; CTCF acts repair-side only at this element"
},
{
"ctcf_degron_breaks": "up",
"rearrangements": "up",
"ratio": "flat",
"reading": "the CTCF+RAD21 fragility class does not apply here; the protective model wins outright"
},
{
"ctcf_degron_breaks": "down",
"rearrangements": "down",
"ratio": "flat",
"reading": "the published siCTCF rise does not survive a degron; compare the siCTCF bridge arm for an artefact call"
},
{
"ctcf_degron_breaks": "n/a",
"rearrangements": "n/a",
"ratio": "n/a",
"reading": "if auxin does not clear occupancy at the element, or if the anchor is unoccupied in CD34+ cells, nothing downstream is interpretable and the occupancy map is the result"
}
]tests hypotheses
[ "hyp-ctcf-occupancy-splits-break-formation-from-rearrangement-fate" ]
- This is a mechanism experiment and it has no clinical arm. Nothing in it would justify a change in anthracycline or etoposide use, and its doxorubicin arms are cell-culture exposures at doses chosen to avoid pleiotropic damage, not clinical exposures. Therapy-related AML after topoisomerase II poisons is long recognised, weighed in protocol design, and not made more or less likely by this experiment's outcome.
- A CTCF degron collapses genome-wide TAD structure within hours, so the experiment risks producing a result about a dying cell rather than about one locus. That is the same objection that weakens the published siCTCF arm, and if the viability, cell-cycle and control-locus readouts are dropped for cost, the experiment inherits the flaw it was designed to remove.
- If the answer is 'occupancy causes the break and guards the repair', the therapeutic reading is uncomfortable: raising occupancy would trade more breaks for fewer rearrangements, and the proteins involved sit near pan-essentiality. The tractable target would remain the methyltransferase, and a G9a inhibitor given prophylactically alongside chemotherapy in a child is a global-silencing intervention that would need a leukaemia-incidence endpoint, not a cytogenetic one, before it is discussed as a clinical idea.
- Does a CTCF degron line exist in a B-lineage or CD34+-competent human background, or would one have to be made? That is the practical rate limiter, not the assays.
- Can END-seq and FISH break-apart be run on aliquots of the same culture at the same timepoint well enough to make 'rearrangements per break' a within-experiment ratio rather than a cross-experiment one?
- Should a single-cell arm be included? A population ratio cannot exclude that the breaks and the rearrangements are happening in different subpopulations, which is a live alternative explanation for any ratio change.
- Is there an existing dataset that could give a partial answer for free - END-seq or DSBCapture in a published CTCF or RAD21 degron experiment, reanalysed at chr11:118,488,159-118,489,206? Canela et al. 2019's HCT-116 RAD21-degron END-seq is deposited and this beat did not look at it.
- E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
Gray et al. report that KDM3B depletion increases H3K9me1/2 and reduces CTCF occupancy at the MLL/KMT2A locus, promoting amplification and rearrangements; that depleting CTCF alone is sufficient; that doxorubicin suppresses KDM3B and CTCF protein levels and that KDM3B or CTCF overexpression rescues Dox-induced alterations; with readouts of DNA FISH copy gain and probe break-apart, locus qPCR and ChIP in immortalised near-diploid human RPE cells extended to KG1a, primary HSPCs and mice, the copy gain being transient at 12 h and inherited at 72 h - and their full text contains no occurrence of 'DSB', 'double-strand break', 'H2AX', '53BP1', 'END-seq', 'comet', 'cohesin', 'RAD21' or 'SMC3'.
Epigenetic balance ensures mechanistic control of MLL amplification and rearrangement. · Gray ZH, Chakraborty D, Duttweiler RR, Alekbaeva GD, Murphy SE, Chetal K, Ji F, Ferman BI, Honer MA, Wang Z, Myers C, Sun R, Kaniskan HU, Toma MM, Bondarenko EA, Santoro JN, Miranda C, Dillingham ME, Tang R, Gozani O, Jin J, Skorski T, Duy C, Lee H, Sadreyev RI, Whetstine JR · Cell 2023What this evidence cannot show
Full text read from PMC10591855 on 2026-09-22; the zero term counts are over the deposited body and figure legends, not the supplement, so a DSB assay reported only in a supplementary panel would be invisible. The 12 h and 72 h kinetics are as stated for the copy-gain phenotype in RPE cells and may not transfer to a degron or to CD34+ cells. FISH break-apart is a physical break detection but not a break map.
Evidence that argues against this (1)
- The siCTCF arm this experiment is designed to replace with a degron is a single knockdown of a protein whose pan-essentiality in this project's DepMap scan is -1.827 across all lines, so the published result could be a viability or mis-segregation artefact - which is why the degron arms need viability, cell-cycle and control-locus readouts or they will inherit the same objection. Canela A et al. 2019
- E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
Canela et al. 2019 mapped TOP2 cleavage genome-wide by END-seq after 30 minutes of 50 uM etoposide and showed that acute cohesin loss reduces the signal: in mouse embryonic fibroblasts with CRE-mediated SMC3 depletion the END-seq signal 'dropped precipitously genome-wide' in proportion to the cohesin lost ('27% and 26% reduction in RAD21 and END-seq levels, respectively'), and the result was confirmed in human HCT-116 cells in which RAD21 was depleted using an auxin-inducible degron, arrested in G1 with lovastatin and treated with auxin for 4 h before etoposide; and they separate TOP2 cleavage-complex trapping, which is transcription-independent and requires pre-existing cohesin binding, from its conversion into an irreversible DSB, which is accelerated about two-fold at transcribed loci and depends on proteasomal degradation and TDP2 phosphodiesterase activity.
Topoisomerase II-Induced Chromosome Breakage and Translocation Is Determined by Chromosome Architecture and Transcriptional Activity. · Canela A, Maman Y, Huang SN, Wutz G, Tang W, Zagnoli-Vieira G, Callen E, Wong N, Day A, Peters JM, Caldecott KW, Pommier Y, Nussenzweig A · Molecular Cell 2019What this evidence cannot show
Read in full text from PMC8170508 on 2026-09-22. This is cohesin, not CTCF, and it contains no MLL or Kmt2a analysis whatever - zero occurrences of either string - so its bearing on this element is by class membership. The depletion is partial (27% RAD21 loss) and the cells are G1-arrested with lovastatin, which is a specific and constrained state. HCT-116 is colorectal, not haematopoietic. This item establishes that the instrument and the direction exist, not that they were applied here.
Evidence that argues against this (1)
- Cohesin loss is not CTCF loss: CTCF depletion releases cohesin from boundary positions rather than removing it from chromatin, so an acute cohesin degron reducing TOP2 breakage does not predict what a CTCF degron does at one anchor - which is why this experiment needs both arms rather than treating the RAD21 result as a proxy. Canela A et al. 2017
- E3moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper
Atkin et al. developed a low-input adaptation of DSBCapture usable on patient samples at 2 ug (~330,000 cells) and 500 ng (~83,000 cells) of input, and applied it to normal non-blast cells from AML patients, finding breakage at the KMT2A intron 10 / exon 11 region (hg38 chr11:118,488,500-800) 2- to 10-fold higher in each KMT2A-rearranged remission patient than in each non-fusion patient with a significant 3.2-fold group average, and identifying preferentially break-prone TOP2-sensitive, CTCF, RAD21 and SMC3 sites in the fusion-driven patients, with low-dose etoposide ex vivo raising breakage at these regions in the fusion patients only.
Assessing acute myeloid leukemia susceptibility in rearrangement-driven patients by DNA breakage at topoisomerase II and CCCTC-binding factor/cohesin binding sites. · Atkin ND, Raimer HM, Wang Z, Zang C, Wang YH · Genes Chromosomes Cancer 2021What this evidence cannot show
n = 3 per arm, adult myeloid. Cited here for the METHOD's feasibility at primary-cell input scale, which is what makes a CD34+ arm realistic; the biological result is cross-sectional and performs no CTCF perturbation. The input figures are as recorded in this graph's gene-ctcf from the paper.
Evidence that argues against this (1)
- Break-mapping in primary cells at this input still averages over a population, so a rearrangement occurring in one cell in ten thousand cannot be paired with the breaks in the same cell - the rearrangements-per-break ratio this experiment reports is a population ratio and not a single-cell one, and a single-cell readout would be needed to exclude that breaks and rearrangements occur in different subpopulations. Atkin ND et al. 2021
- E4strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
Lehman et al. report that in normal patient-derived human mammary epithelial cells a subset of CTCF protein forms complexes localising to SC-35-containing nuclear speckles, that stress rapidly downregulates this species through changes in protein stability with loss from the speckles and altered CTCF-RNA interactions, that their ChIP-seq analysis indicated CTCF binding to genomic DNA is largely unchanged, and that the stress-sensitive pool and its speckle localisation are restored by inhibiting proteasome-mediated degradation.
Dynamic regulation of CTCF stability and sub-nuclear localization in response to stress. · Lehman BJ, Lopez-Diaz FJ, Santisakultarm TP, Fang L, Shokhirev MN, Diffenderfer KE, Manor U, Emerson BM · PLoS Genetics 2021What this evidence cannot show
Abstract only; ChIP-seq figures not read, so 'largely unchanged' is the authors' characterisation. Mammary epithelium and neuronal commitment, not haematopoiesis. Cited here for the design requirement that occupancy at the element be confirmed by ChIP rather than inferred from protein level.
Evidence that argues against this (1)
- A degron removes the protein rather than destabilising a pool, so this paper's dissociation of protein level from genomic binding is a reason to confirm occupancy but not a reason to expect the degron to leave binding intact; over-reading it would predict that a CTCF degron does not clear the anchor, which no degron study supports. Canela A et al. 2019
- E5strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
Gomez-Herreros et al. show that TDP2 rejoins DSBs induced during transcription-dependent TOP2 activity at the translocation hotspot MLL, that TDP2 suppresses TOP2-induced chromosome rearrangements and translocations arising during transcription, and that TDP2-dependent non-homologous end-joining is nonetheless implicated in the formation of a rare translocation subclass previously associated with therapy-related leukaemia and characterised by junction sequences with 4 bp of perfect homology.
TDP2 suppresses chromosomal translocations induced by DNA topoisomerase II during gene transcription. · Gomez-Herreros F, Zagnoli-Vieira G, Ntai I, Martinez-Macias MI, Anderson RM, Herrero-Ruiz A, Caldecott KW · Nature Communications 2017What this evidence cannot show
Abstract read in full on 2026-09-22; body and figures not read, so the magnitude of the effect, whether breakage was quantified in the same cells, and which MLL region was assayed are unknown here. System is breast cancer cells. Cited for the design requirement that junctions be sequenced and classified, and as the precedent that fate can move independently of formation at this locus.
Evidence that argues against this (1)
- The same paper shows the repair factor is protective in aggregate and causal for one junction class, so a junction-class shift with no change in event count is a possible outcome of this experiment that neither arm's primary endpoint would detect - a reason the junction sequencing is not optional. Gomez-Herreros F et al. 2017
the measurement that would settle the triad
degron CTCF versus degron RAD21, breaks and rearrangements in the same cells, reported per break
the degron target; viability and control-locus readouts required because CTCF is near pan-essential
the 1 kb of DNA at which both readouts must be taken
the TOP2 poison arm, matched to the 30-minute pulse used for END-seq
relocates that experiment from HCT116 to Nalm6 or CD34-positive cells, where the element actually breaks