Oncofusion

start / graph / exp-cohesin-degron-break-map-in-a-human-b-precursor-at-kmt2a

Experimentspeculative confidence 1 of 5: Our own inference or analogy. Not shown to be true.Q107-break-machinery-deposition

Move the cohesin degron into a cell where the element actually breaks: RAD21-AID in Nalm6 or CD34+, sBLISS plus both END-seq variants, with the fusion readout in the same cells

also called the Nalm6 RAD21 degron · premise (i) in the right cell type · the experiment the HCT116 mismatch demands

Repeat the one cohesin-removal experiment that exists, but in a blood-precursor cell rather than a bowel-cancer cell, because that is where the KMT2A spot actually breaks. Count both the breaks and the rearrangements in the same cells, at several timepoints.

The same thing, for a scientist

The deposited data establish that the *KMT2A* exon-11 CTCF/cohesin element is a top-0.05%-of-chr11 etoposide DSB site in Nalm6, primary CD34+ marrow, K562 and TK6, and **not** a hotspot in HCT116 - the only human cell in which cohesin has ever been degraded with a DSB readout. **So the single highest-value experiment in this dispute is the existing one, relocated.** Design: introduce a **RAD21-mAID degron** (or use an existing STAG2/RAD21 degron) into **Nalm6** - already the best-characterised human B-precursor line in this literature, with a deposited END-seq baseline at this element [E1] - and, if resources allow, into mobilised **CD34+** cells by mRNA-delivered degron or by acute RAD21 knockdown. Then, in the same cells and the same experiment: **(1)** sBLISS or END-seq +/- auxin, with **both** exonuclease variants (ExoT-only and ExoVII+ExoT) to separate trapped TOP2 cleavage complexes from irreversible DSBs - a pairing Canela ran 18 times in mouse and never in human [E2]; **(2)** a **time course with wash-out** (0.5, 2.5, 7.5 h plus release) so the readout reports persistence with a known lifetime rather than a single snapshot; **(3)** the **fusion readout in the same wells** - 4-color C-Fusion 3D for *MLL-AF4/AF6/AF9/ENL*, or better, a junction-level assay; **(4)** CTCF and RAD21 ChIP to confirm occupancy loss at this element specifically, not just globally. **This is the experiment `exp-ctcf-degron-breaks-and-rearrangements-same-cells-kmt2a` describes, with the cell type fixed and the assay-pairing specified.**

Go deeper
Why Nalm6 and not HCT116, in one line each
criterionNalm6HCT116
element is an etoposide DSB hotspotyes, 99.948th chr11 percentileno, 94.0th with etoposide and 92.9th without
lineageB-cell precursor - the lineage of the index diseasecolorectal epithelium
deposited baseline at this elementyes, GSM2635574/5 END-seq, ETO and NT, hg19yes, but it is the negative
cohesin degron existsno - this is the workyes (Natsume 2016 RAD21-mAID)
gene-editing tractabilitygood; widely used for HDR and knock-inexcellent

And a serious alternative worth costing in parallel: put the readout in HCT116's RAD21-AID line and ask why the element does not break there. If the answer is KMT2A is not transcribed in G1-arrested HCT116, that is a one-track ENCODE lookup and it converts the HCT116 negative from a confound into a confirmation of Gothe's transcription requirement - which would be a result in itself and would cost nothing.

Pre-registered readings, so the experiment cannot be reinterpreted afterwards

Let B = element DSB signal (persistence), T = trapped-TOP2cc signal (ExoVII+ExoT minus ExoT), F = MLL-AF4 fusion frequency, all on auxin versus no auxin in the same clone.

BTFreading
downdowndownpremise (i) confirmed and premise (ii) refuted. Cohesin loading is what makes the element cleavable, and losing it is protective. Gray 2023's FISH result would then need re-examination - most likely as a distinct, copy-number mechanism rather than a translocation one
downdownupthe triad resolves the way hyp-ctcf-occupancy-splits-break-formation-from-rearrangement-fate predicts: cohesin makes breaks and also makes them heal locally. Fewer breaks, worse outcomes. This is the most interesting outcome and the one the hypothesis was written for
flatflatupoccupancy acts only on fate; premise (i) is false and Gray's protection is a repair-channel effect. Also the cleanest possible support for Gray
downflatanycohesin changes lifetime, not cleavage. Premise (i) as usually stated is false and the whole dispute is about repair. This outcome is invisible to any design without the ExoT pairing, which is why the pairing is not optional
upupupcohesin is protective against cleavage - nobody predicts this, and it would mean the loop anchor shields rather than exposes

The design is only decisive if the +/- auxin comparison is within one clone. The deposited HCT116 arm compares genotypes (WT parental versus RAD21-AID clone) with no minus-auxin control, and that single design choice is why its 5.5-fold difference cannot be interpreted 2.

Cost and the cheap first step

Tier: medium-high. A RAD21-mAID knock-in in Nalm6 is a few months of standard work (OsTIR1 line, biallelic mAID tagging, auxin dose-response, RAD21 western and ChIP validation). sBLISS is published with an in-suspension protocol from the same group as Gothe (PMID 33139954, Bouwman/Crosetto), so the readout is transferable. C-Fusion 3D requires the imaging pipeline and probe sets from the Roukos lab. Realistically this is a collaboration, not a solo experiment - the reagents are split across the Nussenzweig, Roukos and Crosetto groups.

Do this first, and it costs a day. Read the element in the tracks that already exist but were not read here:

  1. Gray 2023's four CTCF bigWigs in GSE210480, to settle whether Gray's protective site and this element are one object. Now tractable - see ana-pybigtools-reopens-the-bigwig-route-refused-in-cycle-57, and note the hg19+100,000 padded chrom.sizes.
  2. An HCT116 transcription track (ENCODE RNA-seq or Pol II ChIP) at KMT2A, to test the transcription explanation for the HCT116 negative.
  3. The Gothe CD34+ .bed.gz break lists (145-160 MB per etoposide replicate), to place the break at base resolution relative to the CTCF motif.
  4. A CTCF PWM scan of the 550-bp footprint, for motif orientation - a desk job that decides whether this is the left or right side of a loop.

If all four come back consistent, the degron experiment is worth commissioning. If (1) shows Gray's site is canonical exon 10, the dispute dissolves into a two-element story and this experiment becomes far less urgent.

Facts and figures
designRAD21-mAID degron in Nalm6 (human B-cell-precursor ALL), and if feasible acute RAD21 depletion in primary CD34+ cells. +/- auxin WITHIN ONE CLONE. Readouts in the same experiment: (1) sBLISS or END-seq with BOTH exonuclease variants (ExoT-only and ExoVII+ExoT) to separate trapped TOP2ccs from DSBs; (2) a time course with wash-out at 0.5, 2.5 and 7.5 h plus release; (3) MLL-AF4/AF6/AF9/ENL fusion frequency in the same wells by 4-color C-Fusion 3D or a junction-level assay; (4) CTCF and RAD21 ChIP to confirm occupancy loss at GRCh37 chr11:118,359,150-118,359,700 specifically.
readoutelement DSB persistence (B), trapped-TOP2cc signal (T) and fusion frequency (F), each on auxin versus no auxin, at the 550 bp footprint and genome-wide
controlsminus-auxin same clone; parental Nalm6; DMSO instead of etoposide; a cohesin-independent break site as a positive etoposide control; the H19/IGF2 ICR as a CTCF-occupancy positive control; the GAPDH promoter as a CTCF negative
cost tierhigh
timeline6-12 months for the knock-in and validation, plus 3 months for the readouts; realistically a collaboration across the Nussenzweig (END-seq), Crosetto (sBLISS) and Roukos (C-Fusion 3D) groups
why not hct116the element is not an etoposide DSB hotspot there (94.0th chr11 percentile with etoposide, 92.9th without), so the phenotype the degron is meant to perturb is absent
cheap first steps
[
  "Read Gray 2023's four CTCF bigWigs in GSE210480 at this element to test whether Gray's protective site is this element or canonical exon 10 - beware the hg19+100,000 padded chrom.sizes",
  "Look up an ENCODE HCT116 RNA-seq or Pol II track at KMT2A to test whether the HCT116 negative is a transcription fact",
  "Download the two Gothe CD34+ etoposide .bed.gz break lists (145-160 MB) for base-resolution break placement",
  "Scan a CTCF PWM over the 550 bp footprint for motif orientation"
]
How this could bite us back
  • If the outcome is 'fewer breaks, more fusions', then any intervention that reduces breakage at this element - a TOP2 inhibitor holiday, a cohesin destabiliser - would raise leukaemogenic risk while looking protective on the assay everyone runs. That is the single most consequential thing this experiment could find and it is the outcome the hypothesis predicts.
  • Acute cohesin depletion collapses loop domains genome-wide and shuts down a large fraction of transcription, so an element-level effect will be entangled with a global transcription collapse - and since Gothe's result is that breakage scales with transcription, a fall in element breakage could be entirely secondary. The design needs a nascent-transcription arm at KMT2A or it will not be able to separate these.
Still open
  1. Does a RAD21 or STAG2 degron Nalm6 line already exist in any published resource? A single reagent search could remove most of the cost, and this beat did not run one.
  2. Is sBLISS sensitive enough at n = 2 to detect a two-fold change at one 550 bp element, or does the design need the genome-wide loop-anchor aggregate as its primary endpoint with the element as a named secondary? Gothe's own replicate variation at this element (16.4 versus 21.3 raw; 99.658th versus 99.796th percentile) suggests the rank is stable and the raw value is not.
  3. Would the mouse Smc3f/f+CRE and Waplf/f+CRE END-seq libraries already in GSE129524 answer premise (i) at the mouse Kmt2a orthologue for the cost of one day's reading? The blocker is locating the mouse element, which cannot be done by exon number.
Every source behind this entry (2)
  1. E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.dataset

    Nalm6, a human B-cell-precursor acute lymphoblastic leukaemia line, has a deposited END-seq etoposide/untreated pair (GEO GSM2635574 and GSM2635575, hg19) in which the KMT2A exon-11 CTCF/cohesin element sits at the 99.948th percentile of 918,411 chr11 147-bp bins with etoposide (475 bins above it, 10.77x the file's genome mean) against the 95.948th percentile untreated (1.81x), with the signal maximum at GRCh37 chr11:118,359,390 - within 2 bp of the RAD21 occupancy summit measured in HCT116.

    GEO GSE99194 samples GSM2635574 and GSM2635575 - deposited END-seq bigWig tracks read over HTTP range requests with pybigtools 0.3.0 on 2026-09-22, with the chr11 147-bp-bin null computed from the same files · Canela A, Maman Y, Jung S, Wong N, Callen E, Day A, Kieffer-Kwon KR, Pekowska A, Zhang H, Rao SSP, Huang SC, Mckinnon PJ, Aplan PD, Pommier Y, Aiden EL, Casellas R, Nussenzweig A · NCBI GEO (data accompanying Cell 170:507-521.e18) 2017
    PMID 28735753DOI
    What this evidence cannot show

    n = 1 for each arm; no replicate. The etoposide dose and duration for the Nalm6 arm are not stated in the sample's own fields. Nalm6 carries an ETV6::PDGFRB rearrangement, is from a 19-year-old, and is not KMT2A-rearranged, so it is a lineage proxy for an infant disease. The 2 bp agreement with the HCT116 RAD21 summit is within the ~25 bp honest precision of both maxima and should not be read as base-level coincidence. The percentile null is chr11-only and includes a pericentromeric artefact region.

    Evidence that argues against this (1)
    • CD34+ primary bone-marrow progenitors are arguably the better host for this experiment than any cell line - Gothe's CD34+ arm has two replicates and the element sits at the 99.796th and 99.658th percentiles there - and they are the cell compartment a fetal-origin model actually cares about. The objection is tractability: a biallelic degron knock-in in primary CD34+ cells is not realistic, so the experiment there would have to be acute knockdown or protein degradation by a different route, which is a weaker perturbation. That trade-off, not the biology, is why Nalm6 is proposed first. Gothe HJ et al. 2019
  2. E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.dataset

    The existing human cohesin-degron break map cannot be interpreted because of three specific design choices recorded in its own deposit: the contrast is between genotypes (GSM3714903 genotype WT versus GSM3714904 genotype 'RAD21 AID') rather than +/- auxin within one clone, the only untreated library (GSM3714905) is from the WT genotype so the depleted arm has no baseline, and only the ExoVII+ExoT END-seq variant was run for the human samples while every mouse condition received both that and the ExoT-only variant that 'detect[s] only DSBs'; the same series additionally contains an 0.5-to-12.5 h etoposide time course and a wash-out reversibility series in mouse pre-B cells only.

    What this evidence cannot show

    Bounds the deposit. A minus-auxin RAD21-AID library, replicates, and an ExoT-only HCT116 library could all have been generated and not deposited; the published paper was not read for this entry. The genotype field 'WT' for GSM3714903 is the submitter's and could conceivably describe the RAD21-AID clone without auxin, in which case the first objection falls - but the field says WT and the growth protocol describes auxin as applied, so the plain reading is parental.

    Evidence that argues against this (1)
    • The mouse arms of the same series are much better designed - Smc3f/f+CRE against a WT control and Waplf/f+CRE against Waplf/f, both directions of cohesin residence, with the full ExoT pairing available elsewhere in the series. So an alternative to a new human experiment is to read the mouse Kmt2a locus in GSM3714899-GSM3714902, which costs a day rather than a year. The obstacle is rule-5 grade: mouse Kmt2a exon numbering is a fourth frame and Canela's own 'near exon 10'/'at exon 12' references are mouse and unreconciled, so the mouse element must be located by sequence homology and CTCF occupancy, not by exon number. NCBI Gene Expression Omnibus 2019
Connected to