Enhancer-promoter contacts at KMT2A-fusion target loci — what the fusion uses, what it does not, and what survives the drug
also called H3K79me2/3 enhancer elements · KEE · enhancer-promoter looping MLL-AF4 · PROM1/CD133 KEE
Genes are switched on by distant DNA switches that physically touch the gene's start. In KMT2A-fusion leukemia a specific chemical mark keeps some of those touches in place, and removing the mark breaks the contact and the gene goes quiet. But blocking other proteins stops the gene being read while leaving the contacts intact, and at other genes switching an enhancer on actually moves it further away - so touching and switching-on are not the same thing, in either direction.
The same thing, for a scientist
The Milne group has defined a class of enhancers in MLL-AF4 leukemia carrying H3K79 di/trimethylation, which they term KEEs (H3K79me2/3 enhancer elements). DOT1L inhibition with EPZ-5676 removes H3K79me2/3 and specifically reduces chromatin accessibility, histone acetylation and transcription-factor binding at KEEs but not at non-KEE enhancers, and abolishes enhancer-promoter interactions at a subset of KEEs [E1]. The best-worked single locus is PROM1/CD133: MLL-AF4 binds directly, an intragenic KEE drives activation, and enhancer-promoter interactions between PROM1 and the neighbouring TAPT1 produce co-regulated dual-locus expression; in non-expressing cells the locus is PRC2-bound and the interactions are lost [E2]. Two findings cut against the simple super-enhancer story: MLL-AF4 spreading sites are **distinct from** super-enhancers and it is spreading, not super-enhancer status, that predicts DOT1L-inhibitor sensitivity [E3]; and BET inhibition or BRD4 degradation strongly reduces transcription while leaving enhancer-promoter contacts largely intact, so contact maintenance and transcriptional activation are separable [E4]. **Standing caution (added cycle 35):** whether contact frequency is causal or consequential is unresolved and the relationship is not even monotone. Enhancer activation at Shh is accompanied by *decreased* enhancer-promoter proximity [E1 counter], and live imaging of Sox2 finds transcriptional bursting with no association to enhancer proximity at all [E2 counter]. Combined with E4's finding that contacts persist when transcription is switched off, contact frequency is a poor read-out of regulatory state in both directions, and no experiment in this entry establishes that a KEE contact is causal for the transcription it accompanies.
The one intervention that actually breaks the contact1 deeper
Of the epigenetic drugs used in this disease, DOT1L inhibition is the only one shown to dismantle enhancer-promoter contacts at fusion target loci. Godfrey and colleagues treated MLL-AF4 cells with EPZ-5676 and found the effect was enhancer-class-specific: accessibility, acetylation and transcription-factor binding fell at KEEs and not at non-KEE enhancers, and enhancer-promoter interactions were lost at a subset of KEEs 1.
That gives a mechanistic reason why pinometostat behaves differently from a BET inhibitor, and a reason to revisit DOT1L inhibition in combination rather than as a failed monotherapy: it is the agent that changes the wiring, not just the output.
PROM1/CD133 as the worked example, and why it matters clinically
PROM1 encodes CD133, a pan-cancer stem-cell marker and an actual antibody/CAR target. In MLL-AF4 cells the fusion binds the locus directly, an intragenic KEE drives its activation, and there are increased enhancer-promoter interactions between PROM1 and the nearby gene TAPT1, whose expression correlates with PROM1 across leukemias. In cells that do not express CD133, the PROM1 locus is PRC2-bound and TAPT1 expression is reduced, partly through loss of interaction with the PROM1 locus 2.
The clinical hook is that CD133 surface expression in KMT2A-r ALL is an architectural readout. If a DOT1L inhibitor collapses the KEE and switches CD133 off, a CD133-directed immunotherapy given in combination could lose its antigen. That is an unforced error the field could make, and it follows directly from published data.
Two results that spoil the tidy story (and are the most useful things here)
Super-enhancers are the wrong frame. Kerry and colleagues showed MLL-AF4 binding spreads into the gene body at a subset of targets, with spreading Menin binding, increased H3K79me2/3, disrupted H3K36me3 and unmethylated CpG in the body — and that these spreading sites are distinct from super-enhancers. Spreading-gene expression, not super-enhancer status, is what is preferentially downregulated by DOT1L inhibitors 3. Any model that maps KMT2A-r onto the BRD4/super-enhancer paradigm imported from other cancers is importing the wrong paradigm.
Contacts and transcription are separable. BET inhibition or pharmacological BRD4 degradation, and even dissolving phase condensates, strongly reduce transcription with very little impact on enhancer-promoter interactions; BRD4 and Mediator leave the enhancers but the loops stay 4. The authors attribute residual loop maintenance to other activator complexes and, at some sites, CTCF and cohesin.
Put together with the boundary work in genome-ctcf-cohesin-hoxa-tad-boundary, this produces the single most therapeutically uncomfortable idea in this beat: a successful transcriptional shutdown may leave the physical architecture of the leukemic program fully intact, poised to restart. Nobody is measuring loop persistence in menin-inhibitor-treated patients.
A third result, added cycle 35, that spoils it further: the arrow can point the other way. At Shh, enhancer activation during differentiation is accompanied by a measurable decrease in enhancer-promoter proximity, reproducible with a synthetic activator, and live-cell imaging of Sox2 and its control region shows transcriptional bursting with no association to enhancer proximity whatsoever [E1, E2 counter-evidence]. Put next to E4, the field has now reported: contact without transcription, transcription without contact, and transcription with reduced contact. A Capture-C difference after a drug is therefore a description of what happened to the conformation, not an explanation of what happened to the gene, and this entry should not be read as establishing that KEE contacts cause KEE-gene transcription.
And the clinical test of the spreading model is in. Pinometostat in 51 adults with advanced acute leukaemia reduced H3K79 methylation as intended and produced 2 complete remissions, both in t(11;19) [E3 counter-evidence]. The mark can be removed in patients; the disease mostly does not care. Any argument for revisiting DOT1L inhibition in combination has to start from that, not from the cell-line mechanism.
| element class | H3K79me2/3 enhancer element (KEE) |
|---|---|
| defining mark | H3K79me2/me3, written by DOT1L |
| genome build note | Element coordinates are reported in the source papers' browser tracks; we have not re-derived hg38 intervals and therefore assert none. |
worked loci
[
{
"locus": "PROM1 (CD133)",
"element": "intragenic KEE",
"partner_locus": "TAPT1",
"repressed_state": "PRC2-bound",
"refs": [
"E2"
]
},
{
"locus": "MLL-AF4 spreading targets",
"element": "gene-body spreading with Menin co-spreading",
"distinct_from": "super-enhancers",
"refs": [
"E3"
]
}
]drug effects on contacts
[
{
"agent": "EPZ-5676 (pinometostat, DOT1L inhibitor)",
"effect_on_transcription": "reduced at KEE/spreading genes",
"effect_on_contacts": "lost at a subset of KEEs",
"refs": [
"E1"
]
},
{
"agent": "BET inhibitor / BRD4 degrader",
"effect_on_transcription": "strongly reduced",
"effect_on_contacts": "largely retained",
"refs": [
"E4"
]
},
{
"agent": "menin inhibitor",
"effect_on_transcription": "reduced (other beats)",
"effect_on_contacts": "not measured in any published study we could find",
"refs": []
}
]contact causality status
{
"verdict": "unresolved; contact frequency is not established as causal for transcription at any KMT2A-fusion locus",
"contact_without_transcription": "BET inhibition/BRD4 degradation (E4, PMID 33431820)",
"transcription_without_contact": "Sox2 live imaging (PMID 31124784)",
"activation_with_reduced_contact": "Shh enhancer activation (PMID 31494034)",
"clinical_test_of_the_dot1l_prediction": "pinometostat phase 1, 2/51 CR despite H3K79me reduction (PMID 29724899)"
}- Transcriptional silencing without architectural dismantling leaves a poised program; relapse could be transcriptional restart rather than clonal selection, which would look like drug failure but need a different fix.
- DOT1L inhibition collapses a stem-cell surface antigen (CD133), so combining epigenetic and antigen-directed therapy could be actively antagonistic.
- H3K79 methylation and KEEs exist in normal cells too, so DOT1L inhibition is not fusion-selective at the level of enhancer biology; selectivity comes from the spreading phenotype, which is a quantitative not qualitative difference.
- Importing the super-enhancer/BET paradigm from other cancers into KMT2A-r has already been shown to be the wrong frame; strategy built on it inherits the error.
- If contact frequency is consequential rather than causal, a therapeutic programme built on 'breaking the loop' is aimed at a read-out rather than a mechanism, and would fail in exactly the way pinometostat monotherapy did.
- Do enhancer-promoter contacts at KMT2A-fusion targets survive menin inhibition? If they do, architecture is a resistance reservoir and a Capture-C time course on revumenib-treated cells would show it. This is the single cheapest high-value experiment in this beat.
- Would a DOT1L inhibitor given with a CD133-directed immunotherapy extinguish the antigen it is aimed at? The PROM1 KEE data predict exactly that antagonism.
- Are KEEs present at the same loci in KMT2A::AFF1 infant B-ALL primary blasts as in the cell lines where they were defined?
- What holds the loops up after BRD4 is degraded? The authors suggest CTCF and cohesin at some sites; identifying the residual loop-maintaining factor would name a target whose removal could be additive with menin inhibition.
- Is any enhancer-promoter contact at a KMT2A-fusion target locus causal for its transcription? The discriminating experiment is a forced-proximity or CTCF-anchor-insertion perturbation that changes contact without changing chromatin state, and it has not been done in this disease.
- E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
H3K79me2/3-marked enhancer elements (KEEs) exist in multiple cell types; in MLL-AF4 leukemia cells DOT1L inhibition with EPZ-5676 reduces chromatin accessibility, histone acetylation and transcription-factor binding specifically at KEEs and abolishes enhancer-promoter interactions at a subset of them.
DOT1L inhibition reveals a distinct subset of enhancers dependent on H3K79 methylation · Godfrey L, Crump NT, Thorne R, Lau IJ, Repapi E, Dimou D, Smith AL, Harman JR, Telenius JM, Oudelaar AM, Downes DJ, Vyas P, Hughes JR, Milne TA · Nature Communications 2019What this evidence cannot show
Cell-line based (MLL-AF4 lines); 'a subset of KEEs' is explicitly a subset, so loss of enhancer-promoter interaction is not a general consequence of DOT1L inhibition. EPZ-5676 has pleiotropic effects over the multi-day treatment required to deplete H3K79me2/3. Added cycle 35: the paper shows contact loss co-occurring with transcriptional loss and cannot separate cause from consequence; published counter-examples exist in which enhancer activation reduces proximity.
Evidence that argues against this (1)
- The direction of the contact/activity relationship is not fixed. At the Shh locus, enhancer activation during ES-to-neural-progenitor differentiation is accompanied by a *decrease* in enhancer-promoter spatial proximity by super-resolution 3D-FISH and 3C, reproducible by synthetic enhancer activation; the authors conclude that models of enhancer-promoter communication must encompass conformations other than looping. So 'DOT1L inhibition abolished enhancer-promoter interactions at a subset of KEEs' is compatible with contact being a consequence of the transcriptional state rather than its cause, and the KEE data do not discriminate between the two. Benabdallah NS et al. 2019
- E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
MLL-AF4 binds the PROM1 locus directly and activates CD133 through an intragenic H3K79me2/3 enhancer element, with increased enhancer-promoter interactions between PROM1 and the neighbouring TAPT1; in non-expressing cells the locus is PRC2-bound and TAPT1 expression falls partly through loss of that interaction.
H3K79me2/3 controls enhancer-promoter interactions and activation of the pan-cancer stem cell marker PROM1/CD133 in MLL-AF4 leukemia cells · Godfrey L, Crump NT, O'Byrne S, Lau IJ, Rice S, Harman JR, Jackson T, Elliott N, Buck G, Connor C, Thorne R, Knapp DJHF, Heidenreich O, Vyas P, Menendez P, Inglott S, Ancliff P, Geng H, Roberts I, Roy A, Milne TA · Leukemia 2021What this evidence cannot show
One locus, largely in cell lines with supporting patient expression correlation. Volume/issue is Leukemia 35(1):90-106 with online publication in April 2020; cite the 2021 issue year to avoid a mismatch. The therapeutic inference about CD133-directed immunotherapy is ours, not the authors'. Added cycle 35: contact-frequency increases in fixed-cell population assays are not evidence of causal proximity; live-imaging work at other loci finds bursting uncoupled from enhancer proximity.
Evidence that argues against this (1)
- Live-cell imaging of Sox2 and its essential control region found no evidence of enhanced spatial proximity between enhancer and promoter, limited spatial dynamics over tens of minutes, and no association between transcriptional bursting and enhancer proximity. Applied to PROM1/TAPT1: an increase in measured contact frequency in a fixed-cell population assay is not evidence that physical proximity drives co-regulation, and the dual-locus co-expression could equally reflect a shared chromatin state that both loci occupy. Alexander JM et al. 2019
- E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
MLL-AF4 binding spreads into the gene body at a subset of targets with spreading Menin binding and increased H3K79me2/3; these spreading sites are distinct from super-enhancers, and spreading-gene expression is preferentially downregulated by DOT1L inhibition.
MLL-AF4 Spreading Identifies Binding Sites that Are Distinct from Super-Enhancers and that Govern Sensitivity to DOT1L Inhibition in Leukemia · Kerry J, Godfrey L, Repapi E, Tapia M, Blackledge NP, Ma H, Ballabio E, O'Byrne S, Ponthan F, Heidenreich O, Roy A, Roberts I, Konopleva M, Klose RJ, Geng H, Milne TA · Cell Reports 2017What this evidence cannot show
Defined in MLL-AF4 cell lines; 'spreading' is an operational definition from ChIP-seq signal distribution and its threshold choice affects which genes qualify. The distinction from super-enhancers depends on the super-enhancer calling method, which is itself contested. Added cycle 35: the model's clinical prediction was tested - pinometostat achieved target engagement in phase 1 with only modest activity (2/51 CR), so spreading status does not identify a load-bearing dependency in patients.
Evidence that argues against this (1)
- The spreading model's therapeutic prediction was tested and largely failed. In the phase 1 study of pinometostat in 51 adults with advanced acute leukaemia, mostly MLL-r, target engagement was achieved (H3K79 methylation was reduced) but clinical activity was modest - 2 complete remissions, both in t(11;19). Collapsing the H3K79me2/3 enhancer/spreading class in patients therefore does not collapse the disease, which is the strongest available argument that spreading and KEE status describe a real but non-load-bearing feature of the programme. Stein EM et al. 2018
- E4strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
BET inhibition or pharmacological BRD4 degradation, and dissolution of phase condensates, strongly reduce transcription but have very little effect on enhancer-promoter interactions, showing that transcriptional activation and maintenance of enhancer-promoter contacts are separable events.
BET inhibition disrupts transcription but retains enhancer-promoter contact · Crump NT, Ballabio E, Godfrey L, Thorne R, Repapi E, Kerry J, Tapia M, Hua P, Lagerholm C, Filippakopoulos P, Davies JOJ, Milne TA · Nature Communications 2021What this evidence cannot show
Acute pharmacological timescales; contacts might decay with longer exposure. Contact measurement is by Capture-C/Tri-C at selected loci, so a genome-wide statement is an extrapolation. Condensate dissolution used 1,6-hexanediol, a blunt reagent with known off-target effects on chromatin.
Evidence that argues against this (1)
- A standing caution from this project about over-reading 3D contact data, now attached here. Hakim et al. measured nuclear contact frequency and DNA-break rate simultaneously in mouse B lymphocytes and found that translocations track contact frequency only in the absence of targeted damage; for *recurrent* rearrangements it is the rate of break formation, not proximity, that governs location and frequency. Contact-frequency maps are a weak causal instrument in this system generally: for translocation they are dominated by break rate, and per Benabdallah and Alexander, for transcription they can move in either direction or not at all. Hakim O et al. 2012
- related to →CTCF/cohesin boundaries and the HOXA topological domain — the 3D scaffold the fusion program runs on
Counter-evidence (Benabdallah 2019, PMID 31494034): at Shh, enhancer activation is accompanied by DECREASED enhancer-promoter proximity, reproducible with a synthetic activator - so loss of contact after DOT1L inhibition is compatible with contact being consequential rather than causal
Counter-evidence (Alexander 2019, PMID 31124784): live-cell imaging of Sox2 and its control region shows no enhanced enhancer-promoter proximity and no association between transcriptional bursting and proximity, so a fixed-cell contact-frequency increase is not evidence of causal looping
- ← refutesPinometostat (EPZ-5676) — the DOT1L inhibitor, and the most instructive failure in this field
Counter-evidence (Stein 2018, PMID 29724899): pinometostat phase 1 in 51 adults with advanced acute leukaemia reduced H3K79 methylation as intended but produced only 2 complete remissions, both t(11;19) - removing the KEE/spreading mark in patients does not collapse the disease, so the spreading model describes a real but non-load-bearing feature
Counter-evidence (Hakim 2012, PMID 22314321): nuclear contact frequency governs only non-targeted rearrangements, while recurrent translocations track DNA-break rate - a standing caution in this project that 3D contact maps are a weak causal instrument, now applied to enhancer-promoter contacts as well as to translocation
both concern co-transcriptional structure at the fusion's target loci and would be measured together
BENC contacts the MYC promoter across 1.7 Mb; the inv(16) study confirms the physical connection by enhancer-interaction analysis
arm 2 asks whether the myeloid enhancer-promoter contact exists in B-lineage cells