Hypothesis: KMT2A and AFF1 break where they do because both breakpoint regions are CTCF/cohesin loop anchors sitting in the same nuclear hub
also called shared loop anchor hub model · anchor-anchor translocation model for t(4;11)
Chromosomes are held in loops, pinned at anchor points, and those pins are where DNA most often snaps. This hypothesis says KMT2A and its commonest partner are both pinned, both pinned in the same part of the nucleus, and that is the whole explanation for the fusion.
The same thing, for a scientist
The hypothesis combines two measured facts and one unmeasured link. Measured: CTCF/cohesin-bound loop anchors accumulate TOP2B-mediated DSBs, polymorphisms that redistribute CTCF/cohesin move the cleavage sites, and loop-anchor DSBs fall at breakpoint clusters commonly translocated in cancer [E1]; TOP2 DSBs are enriched at transcriptionally active loop anchors and genes recurrently translocating to drive leukaemias are enriched there, with MLL fusions arising at recurrent hotspots dependent on both TOP2 isoforms and on elongation [E2]. Unmeasured: that the 8.3 kb KMT2A BCR and the AFF1 breakpoint region are each a named loop anchor in a haematopoietic progenitor, and that the two anchors sit in a shared hub. The only direct proximity datum is 2-3% of transcribing KMT2A alleles near AF9 or AF4 in transformed lines [E3]. The hypothesis is attractive because it would explain position AND partner choice with one mechanism, and it is dangerous for the same reason: a single mechanism that explains everything is usually explaining the thing that is easiest to measure. The strongest argument against it is that recurrent oncogenic translocations track break FREQUENCY rather than contact frequency when both are measured in the same system [E4], and the second strongest is that loop-anchor breakage is reported as largely cell-type-independent [E1], which would leave the fusion's fetal-progenitor specificity entirely unexplained.
The chain, with each link's status1 deeper
- Loop anchors are DSB hotspots. Measured, genome-wide, twice 12.
- The KMT2A BCR is a loop anchor in a haematopoietic progenitor. Not measured. A PubMed search for KMT2A breakpoints with CTCF, loop anchors or cohesin returns zero records 5.
- The AFF1 breakpoint region is a loop anchor. Not measured, and the AFF1 side has no structural characterisation at all comparable to KMT2A's - the graph's
breakpoint-kmt2a-bcr-architecturelists this as an open question and the only relevant primary work is a 2001 halo-mapping study 6. - The two anchors are in a shared hub. Not measured. 2-3% colocalisation of transcribing alleles in transformed lines is the nearest datum 3.
- Breaking both anchors in the same hub yields t(4;11). Would follow from 1-4 plus ordinary end joining; nothing additional is required.
The hypothesis is therefore three unmeasured links deep, and every one of them is measurable with public data or with one experiment.
Why link 2 could be checked this week
CTCF and cohesin ChIP-seq and Hi-C loop calls exist for cord-blood CD34+ cells and for numerous haematopoietic lines. Overlaying the KMT2A BCR coordinates on those tracks is a bedtools intersect. If the BCR does not overlap a CTCF/cohesin peak or a called loop anchor in ANY haematopoietic cell type, link 2 fails and the hypothesis is dead without further work. This is specified as part of ana-repeat-feature-enrichment-kmt2a-breakpoints, and it is the cheapest falsification available in this whole beat.
The uncomfortable implication if it is true
This graph already notes the discomfort in mechanism-3d-genome-proximity-partner-choice and it is worth restating sharply here. If the fusion pairs two genes that were already co-located because both are highly transcribed elongation genes at loop bases, then the oncoprotein assembles machinery that was partly pre-assembled in the normal nucleus. KMT2A and AFF1 would have been functional neighbours before they were fused. That is consistent with the menin-interface model but it weakens one of its implicit premises: that the fusion creates an abnormal association. On this hypothesis the fusion makes a pre-existing association covalent and constitutive, which is a different kind of lesion and could have a different kind of vulnerability.
What kills it
Three results, in ascending cost.
- The BCR is not at an anchor. A public-data overlay showing no CTCF/cohesin peak or called loop anchor overlapping the KMT2A BCR in any haematopoietic cell type. Kills link 2.
- Cohesin depletion does nothing. A RAD21 degron in a human progenitor-like cell, etoposide challenge, inverse PCR readout: if the frequency and the within-BCR position of KMT2A rearrangement are unchanged, the anchor is not load-bearing even if it exists. This is the decisive experiment and it is not expensive.
- Partner rank does not track anchor strength or contact. Specified in
ana-kmt2a-partner-rank-vs-contact-frequency. If the seven commonest partners do not rank by contact or anchor strength in any haematopoietic cell type, the hypothesis explains position at best and not partner choice.
A fourth, softer result would demote rather than kill it: if contact frequency in fetal liver progenitors is no higher than in adult cells (exp-hic-kmt2a-aff1-proximity-fetal-liver), the hypothesis survives as an account of position but stops explaining why this is an infant disease.
Correction: the absence was false, and the paper that settles it splits this hypothesis
E5 recorded that PubMed "returns zero records linking KMT2A breakpoints to CTCF ... or to loop anchors and cohesin". It returns eighteen, and the relevant one was read in full (PMC8511143, 76,628 characters, reached by the PMC HTML route after fullTextXML returned HTTP 500 and efetch db=pmc returned HTTP 429 - had the reader stopped at route 2 this correction would not exist).
It does not confirm this hypothesis. It splits it into three.
Supported. In normal cells of KMT2A-rearranged AML patients, the KMT2A intron 10 / exon 11 region breaks 2-10x more than in non-fusion AML individuals, group average 3.2x; 404 CTCF, 517 RAD21, 256 SMC3 and 304 TOP2-sensitive sites are preferentially break-prone; and low-dose etoposide raises breakage at those regions only in the KMT2A-rearranged patients [E-atkin].
Not addressed. Whether the KMT2A breakpoint cluster region is itself a named CTCF or cohesin anchor. That is a bedtools intersect against deposited tracks and it remains unpublished - which is the narrower true version of what E5 was trying to say.
Weakly contradicted. Among the 34 regions carrying all four features, no partner gene of any of the three patients appeared, and only AFDN's 3' end - not its breakpoint region - carried a break-prone CTCF site. The authors invoke the break-frequency-dominant model that this hypothesis already names as its rival.
Design limits, stated because the effect is the headline: n = 3 per arm, adult, treated remission samples, and the comparator is other AML rather than healthy donors - so it cannot separate inherited susceptibility from a sequela of the disease or its treatment.
| statement | The KMT2A and AFF1 breakpoint cluster regions are each CTCF/cohesin-bound chromatin loop anchors in haematopoietic progenitors, the two anchors occupy a shared nuclear hub, and t(4;11) arises when both are broken in the same hub and mis-joined. |
|---|---|
| origin | literature-gap |
| prior probability | 0.2 |
| distinguishing alternative | break-frequency-dominant model (Hakim 2012, E4): the KMT2A BCR is simply a high-break-rate site and partner choice follows from partner break rate plus selection on the fusion product, with geometry setting only a background |
predictions
[ "The 8.3 kb KMT2A BCR overlaps a CTCF/cohesin peak and a called loop anchor in cord-blood or fetal CD34+ cells", "The AFF1 breakpoint region likewise overlaps an anchor", "Trans contact between the two regions exceeds that of expression- and distance-matched control locus pairs", "Acute cohesin (RAD21) depletion reduces etoposide-induced KMT2A rearrangement frequency and shifts its position within the BCR", "The seven commonest partner loci rank by anchor strength or contact frequency in the same order as their clinical frequency" ]
experiments
[ "exp-hic-kmt2a-aff1-proximity-fetal-liver", "exp-human-cd34-kmt2a-breakage-assay" ]
analyses
[ "ana-repeat-feature-enrichment-kmt2a-breakpoints", "ana-kmt2a-partner-rank-vs-contact-frequency" ]
falsifiers
[ "The KMT2A BCR overlaps no CTCF/cohesin peak and no called loop anchor in any haematopoietic cell type with public data - kills the hypothesis at link 2 for the cost of a bedtools intersect", "Acute RAD21 degron leaves both the frequency and the within-BCR position of etoposide-induced KMT2A rearrangement unchanged - the anchor exists but is not load-bearing", "Partner loci show no rank relation between anchor strength or contact frequency and clinical partner frequency - the hypothesis then explains position at best, not partner choice", "Junction analysis shows KMT2A partners are dominated by loci that are NOT at loop anchors" ]
- Nothing in this hypothesis is druggable. CTCF and cohesin build the regulatory architecture of every cell; perturbing them to prevent a rare break would be catastrophic and cohesin loss is itself a documented route to myeloid neoplasia.
- If the hypothesis is true, KMT2A and AFF1 were co-located and co-regulated before the fusion, which means the oncoprotein assembles a partly pre-existing machine - a fusion-specific interface may then be harder to find than the menin-interface programme assumes.
- Because loop-anchor breakage is reported as cell-type-independent, a true positive here would explain the break's position and leave the disease's cell-type and age specificity entirely unexplained, which is the part clinicians care about.
- Does the KMT2A BCR overlap a called loop anchor in any haematopoietic cell type? Unpublished, and answerable from deposited data today.
- Is there any structural characterisation of the AFF1 breakpoint region later than 2001? The KMT2A side has been mapped repeatedly and the partner side has not, which makes every two-sided model in this literature half-evidenced.
- If KMT2A and AFF1 were already functional neighbours in the normal progenitor nucleus, does the fusion create a new association or make an old one covalent? The answer changes what kind of vulnerability the fusion creates.
- Would a RAD21 degron in a human CD34+ system be tolerated long enough to read out rearrangement frequency, given cohesin's role in proliferation? This is the feasibility crux of the decisive experiment.
measures the shared-hub link of the hypothesis
- depends on →Chromatin loop anchors are DNA-break hotspots - and the two papers that established it disagree about why
the hypothesis is the locus-level application of loop-anchor fragility
supplies the concrete structural mechanism that entry's loop-anchor section gestures at
predicts the BCR overlaps a CTCF/cohesin anchor; falsified by a public-data overlay if it does not
requires the AFF1 breakpoint region to be an anchor; no structural characterisation of it since 2001
tests the hypothesis's partner-choice prediction and can falsify it
no CTCF/cohesin measurement exists at AFF1, so the hypothesis is only half-testable from prior literature
corrects that entry's counted absence: PMID 34405474 links KMT2A breakage to CTCF/cohesin sites in patients, and finds no partner-side anchor
the acute RAD21 degron that hypothesis names as both prediction and falsifier
the loop-anchor hypothesis a STAG2 track would sharpen
the closest published single-nucleus test of proximity-explains-outcome; the proximity half holds and the lineage half came back negative