Oncofusion

start / graph / mechanism-tf-repression-corruption

Mechanismstrong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.Q11-comparative

Mechanism class 2 — transcription-factor corruption (RUNX1::RUNX1T1, PML::RARA, ETV6::RUNX1, CBFB::MYH11) — a balance, not a pure dominant negative

also called dominant-negative TF class · corepressor-tethering class · CBF corruption

Here the fusion keeps the part of a master switch that binds DNA but replaces the part that turns genes on with a part that shuts them off, so development stalls. It is not simply that the normal switch is disabled — these leukaemias still need a working copy of it, and die without one.

The same thing, for a scientist

In this class the fusion retains the DNA-binding or heterodimerisation module of a lineage master transcription factor and acquires a repression module — an NCoR/SMRT/HDAC-recruiting domain (RUNX1T1/ETO nervy homology regions, PML, ETV6 repression domain) or a sequestering scaffold (the MYH11 coiled-coil). The fusion occupies the wild-type factor's cognate sites and converts activation into repression, or drags the obligate cofactor out of productive complexes. The phenotype presents as a block of a specific differentiation step rather than a proliferative drive, and the fusion is usually insufficient on its own, requiring cooperating lesions (KIT, RAS, FLT3, secondary deletions). **The 'dominant-negative' shorthand for this class is withdrawn.** t(8;21) and inv(16) cells are addicted to NATIVE RUNX1: knocking it down kills them, knocking down AML1-ETO rescues them, and an active RUNX1 allele is preserved in these patients while RUNX1 is frequently inactivated in other leukaemias [E6]. RUNX1 is likewise required for the survival of AML1-ETO and MLL-AF9 cells, via BCL2, with RUNX2 able to compensate [E7]. The pathology is a maintained BALANCE between fusion and native factor competing for shared sites, and the therapeutic implication is the reverse of the intuitive one — the authors of both papers propose that inhibiting, not restoring, RUNX1 activity is the tractable direction.

Go deeper
Four variations on jamming a master switch1 deeper

RUNX1::RUNX1T1 (AML1-ETO, t(8;21)): keeps the RUNX1 Runt domain, so it still binds RUNX-motif DNA and still heterodimerises with CBFB, but replaces the RUNX1 transactivation domain with ETO, which recruits corepressors. Critically, the ETO nervy homology region 2 (NHR2) makes the fusion oligomerise; NHR2-mediated oligomerisation is required for the differentiation block, and expressing a competing NHR2 peptide dissolves the high-molecular-weight complex, restores RUNX1 target gene transcription and kills t(8;21) cells 12.

CBFB::MYH11 (inv(16)): keeps the CBFB Runt-binding surface and adds the smooth-muscle myosin coiled-coil. It outcompetes wild-type CBFB for RUNX1 and deregulates RUNX1 activity. AI-10-49, a small molecule that binds CBFB-SMMHC and disrupts its binding to RUNX1, restores RUNX1 transcriptional activity, kills primary inv(16) blasts and delays leukemia in mice 3. This is the single clearest proof that a fusion-partner interaction surface in this class can be drugged directly.

PML::RARA (t(15;17)): keeps the RARA DNA-binding and ligand-binding domains and adds the PML RBCC/coiled-coil, which forces homodimerisation. Forcing RARA into a homodimer is by itself enough to prime mice for APL-like leukemia, which localises the oncogenic essence of the fusion to the dimerisation module rather than to any unique PML function 4. Uniquely in this class, the retained ligand-binding pocket makes the fusion directly ligandable (see concept-why-pml-rara-became-curable).

ETV6::RUNX1 (t(12;21)): keeps the ETV6 repression/SAM region fused to nearly all of RUNX1, converting RUNX1 from an activator into a repressor at RUNX1 sites. It arises in utero, is detectable in neonatal blood spots years before disease, and needs a postnatal second hit — the twin studies are the cleanest human demonstration that this class creates a silent preleukemic clone rather than leukemia 5.

The common structural signature: an obligate self-association or partner-binding surface

Three of the four (ETO NHR2, PML coiled-coil, MYH11 coiled-coil) work through a forced oligomerisation or sequestration interface contributed by the 3-prime partner. That is a recurring, rather specific structural motif, and in two of the three there is already chemical or peptide matter that engages it 23. It is the strongest argument in the whole atlas that 'undruggable transcription factor' is a statement about pockets, not about mechanism: the druggable feature is the protein-protein interface the fusion cannot do without.

Therapeutic implication — stated plainly

Partly drugged, unevenly. PML::RARA is curable (ATRA plus arsenic) and BCR::ABL1 sits in a different class, so within this class only PML::RARA has delivered. CBFB::MYH11 has a validated chemical probe (AI-10-49) that has not reached registration. RUNX1::RUNX1T1 has peptide-level proof of principle at NHR2 and no clinical molecule. ETV6::RUNX1 has no targeted agent at all, and does not need one urgently because chemotherapy cures roughly 90% of these children.

Why the gap: repression-type fusions are removed by restoring a transcriptional program, which requires either destroying the fusion (ATRA/arsenic, only possible because PML::RARA is degradable) or breaking an interface (AI-10-49, NHR2 peptides). Neither is an enzyme-inhibitor problem, so the class has been starved of the medicinal chemistry that kinase fusions attracted. The lesson for KMT2A::AFF1 is that the successful examples in this class all found an obligate interface, exactly the lab's working hypothesis, and that the interface was not always on the fusion itself.

Correction (cycle 33): 'dominant-negative' withdrawn for the core-binding-factor members

The class was named and described as a dominant-negative block. For its two core-binding-factor members that is wrong, on evidence published in 2013.

  • Withdrawn: the claim that the fusion's pathology is dominant-negative inactivation of the native factor. Kasumi-1 (t(8;21)) and ME-1 (inv(16)) cells require wild-type RUNX1 to survive; RUNX1 knockdown attenuates the mitotic checkpoint and causes apoptosis; knocking down AML1-ETO in those cells rescues them; and an active RUNX1 allele is usually preserved in t(8;21)/inv(16) patients while RUNX1 is frequently inactivated in other leukaemias 6. Independently, RUNX1 is required for AML1-ETO and MLL-AF9 cell survival through BCL2, RUNX2 compensates, and combined Runx1/Cbfb deletion blocks MLL-AF9 leukaemia in mice 7.
  • Replaced with: a competitive-balance model. Fusion and native factor occupy common genomic sites and the malignant state depends on the ratio, not on the absence of the native factor.
  • Consequence for the beat. 'Restore the corrupted factor' is the intuition behind AI-10-49 and behind the broader obligate-partner-interface argument. Both cited papers propose the opposite therapeutic direction. An interface drug at the CBF interface is tuning a set point in a system that needs native RUNX1, which is a narrower and more fragile therapeutic window than 'relieve a blockade' implies.
  • Unaffected: the PML::RARA row and the requirement for cooperating lesions.
  • Corrected 2026-09-19: the sentence 'The RUNX1-addiction result has not been tested in ETV6::RUNX1' is withdrawn. Wray et al. tested it directly 8. They characterised the ETV6-RUNX1 regulome and found the fusion works mainly by competition for RUNX1 binding sites plus transcriptional repression - which is the competitive-balance model above, arrived at independently in the third member of the class - and then showed that in frank ETV6-RUNX1 leukaemia, knockdown of RUNX1 or of CBFB kills the cells, reproduced chemically with an allosteric CBFB inhibitor. They report that RUNX1 addiction extends to other genetic subtypes of paediatric and adult B-ALL while sparing normal haematopoiesis. So the ETV6::RUNX1 row is not unaffected: it is now the third member of the class where 'dominant-negative' is the wrong description, and the therapeutic direction is again tuning a set point rather than relieving a blockade. If RUNX1 addiction generalises across B-ALL subtypes, it is also not a fusion-specific vulnerability, which weakens the argument that it flows from the corruption mechanism at all.
Facts and figures
mechanism classtf-repression-corruption
invariant moduleretained DNA-binding or obligate-heterodimerisation module of a lineage master TF (RUNX1 Runt domain, CBFB Runt-binding surface, RARA DBD/LBD, RUNX1 via ETV6)
variable moduleacquired repression or forced-oligomerisation module (ETO NHR1-4, PML RBCC coiled-coil, MYH11 coiled-coil, ETV6 repression domain)
druggable nodethe partner-supplied oligomerisation or sequestration interface (ETO NHR2, CBFB-SMMHC:RUNX1, PML-driven dimer), plus degradation of the fusion where the partner makes it degradable
druggability classemerging
why tractable where it isThree of four members depend on a partner-supplied coiled-coil or oligomerisation interface, which is a discrete protein-protein interface rather than a diffuse surface, and one member additionally retains a nuclear-receptor ligand pocket.
why mostly undruggedRemoving a repressive fusion requires restoring a program, which needs either fusion degradation or interface disruption rather than enzyme inhibition, and the class has attracted far less medicinal chemistry than kinase fusions.
member fusions
[
  "fusion-runx1-runx1t1",
  "fusion-cbfb-myh11",
  "fusion-pml-rara",
  "fusion-etv6-runx1",
  "fusion-tcf3-hlf",
  "fusion-cbfa2t3-glis2"
]
drugs yielded
[
  "ATRA and arsenic trioxide for PML::RARA"
]
drugs failed or stalled
[
  "AI-10-49 (CBFB-SMMHC:RUNX1 interface, preclinical only)",
  "NHR2-interfering peptides (proof of concept only)",
  "HDAC inhibitors as class-wide de-repressors"
]
dominant negative framing withdrawn
{
  "scope": "core-binding-factor members (RUNX1::RUNX1T1, CBFB::MYH11)",
  "evidence": "t(8;21) and inv(16) cells require native RUNX1; RUNX1 knockdown kills them and AML1-ETO knockdown rescues; active RUNX1 allele preserved in patients",
  "pmids": [
    "24055056",
    "23979164"
  ],
  "replacement_model": "competitive balance between fusion and native factor at shared sites; malignant state depends on the ratio",
  "therapeutic_direction": "both source papers propose INHIBITING rather than restoring RUNX1 activity in fusion-driven AML",
  "untested_in": [
    "ETV6::RUNX1",
    "PML::RARA"
  ]
}
How this could bite us back
  • De-repressing a master transcription factor in a large blast burden risks a differentiation-syndrome-like on-mechanism toxicity, as seen with both ATRA and menin inhibitors.
  • These fusions are usually insufficient alone, so blocking the fusion may leave the cooperating lesion (KIT, RAS, FLT3) driving residual disease.
  • RUNX1 and CBFB are required for normal definitive haematopoiesis, so restoring or manipulating CBF activity has a narrow therapeutic window in a child mid-development.
  • If these leukaemias need native RUNX1 to survive, a drug that restores native RUNX1 activity may be pushing the ratio in the direction the leukaemia already wants at low doses, with benefit only above a threshold that has never been defined in patients.
  • Inhibiting core binding factor is deliberate, systemic, pharmacological RUNX1 haploinsufficiency - and the human phenotype of constitutional RUNX1 haploinsufficiency is a lifelong platelet defect plus predisposition to AML. The on-target risk of this strategy is not cytopenia, it is leukaemogenesis in the host, and it will not appear on the timescale of the experiments that report normal haematopoiesis being spared.
  • In adult mice, Runx1 loss produces a myeloproliferative phenotype with expanded myeloid progenitors and extramedullary haematopoiesis. A RUNX1 inhibitor would therefore be expected to expand precisely the compartment from which a second leukaemia arises.
  • Roughly half of t(8;21) patients carry an activating KIT mutation acquired after the fusion, and it is the KIT clone that disappears in remission while the fusion persists. Any trial reading out early response to a RUNX1- or fusion-directed agent risks attributing to the fusion axis a response that belongs to the kinase co-lesion, or missing a fusion-axis benefit that only shows at the level of eradication.
Still open
  1. Is the ETO NHR2 tetramer interface tractable to a drug-like small molecule rather than a peptide, and does its buried surface area fall inside the range that has yielded PPI drugs elsewhere?
  2. Why did AI-10-49 stall after a Science-level result, and is the obstacle chemistry, pharmacokinetics or insufficient commercial pull for a chemo-curable subtype?
  3. Does KMT2A::AFF1 contain any analogous forced-oligomerisation interface (for example through the AFF1 ALF domain or KMT2A dimerisation) that could be attacked the same way?
  4. Does the RUNX1-addiction result extend to ETV6::RUNX1, where the same native factor is the 3-prime partner rather than the 5-prime one? Nobody has run the equivalent knockdown.
  5. Is there a therapeutic window between the RUNX1 dose a t(8;21) or inv(16) blast needs and the dose normal megakaryopoiesis needs, given that constitutional 50% loss already causes a human platelet disorder - and has anyone measured that window rather than assuming it from short-term knockdown?
Every source behind this entry (8)
  1. E1moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    Oligomerisation through the ETO NHR2 domain is essential for RUNX1/ETO oncogenic activity, and a competing NHR2-containing peptide disrupts the high-molecular-weight complex, restores target gene transcription, reverts the differentiation block and kills t(8;21) cells.

    PMID 17332359DOI
    What this evidence cannot show

    Intracellular expression of a 128-residue peptide in cell lines, not a drug-like molecule and not an in vivo efficacy study; the effect size in primary patient blasts was not established.

    Evidence that argues against this (1)
    • The class name describes a dominant negative, and for the two core-binding-factor members that is wrong. t(8;21) Kasumi-1 and inv(16) ME-1 cells REQUIRE wild-type RUNX1 to survive: RUNX1 knockdown attenuates the mitotic checkpoint and causes apoptosis, knocking down AML1-ETO in the RUNX1-knockdown cells RESCUES them, and an active RUNX1 allele is usually preserved in t(8;21) and inv(16) patients while RUNX1 is frequently inactivated in other leukaemias. The pathology is a maintained balance between fusion and native factor at shared genomic sites, not a simple blockade of the native factor. Ben-Ami O et al. 2013
  2. E2weak confidence 2 of 5: Preliminary, small, indirect, or not yet peer reviewed.paper

    A cell-penetrating TAT-NHR2 polypeptide delivered exogenously reduced proliferation and increased apoptosis in the t(8;21) line Kasumi-1, with specificity shown against an unrelated BCR oligomerisation-domain control.

    PMID 23865046DOI
    What this evidence cannot show

    A single cell line, recombinant TAT-fusion delivery with well-known artefacts, no animal data, and a low-impact venue; treat as proof of concept for the interface only, not for the modality.

    Evidence that argues against this (1)
    • The delivery claim rests on an assay class with a documented artefact that inflates exactly this readout. TAT and other arginine-rich cell-penetrating peptides appeared for years to enter cells by an energy-independent route; that appearance came from redistribution of the peptide and its cargo during the fixation step used for fluorescence microscopy, and from strong ionic interaction with the cell surface that overestimates the cell-associated signal. Uptake is now established to be endocytic and energy-dependent, which means the delivered peptide is largely endosomal rather than free in the nucleoplasm where an NHR2 competitor would have to act. An antiproliferative effect in Kasumi-1 is compatible with very little of the peptide reaching the target; this claim is correctly marked 'weak' and the reason it is weak is the delivery assay, not the biology. Vives E et al. 2003
  3. E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    AI-10-49 selectively binds CBFB-SMMHC, disrupts its binding to RUNX1, restores RUNX1 transcriptional activity, selectively kills primary inv(16) AML blasts and delays leukemia progression in mice.

    PMID 25678665DOI
    What this evidence cannot show

    Leukemia was delayed, not eradicated; AI-10-49 has pharmacokinetic limitations and has not advanced to a registration trial, so clinical tractability of this interface remains unproven.

    Evidence that argues against this (1)
    • The same correction applies to the therapeutic logic behind AI-10-49. 'Restoring RUNX1 transcriptional activity' reads as relieving a dominant-negative block, but inv(16) cells need native RUNX1 and die without it. Independently, RUNX1 is required for the survival of AML1-ETO and MLL-AF9 cells, the survival effect is mediated by BCL2, RUNX2 can compensate for RUNX1 loss, and the authors conclude that INHIBITING rather than enhancing RUNX1 activity may be the promising strategy in fusion-driven AML -- the opposite direction of travel from 'restore the corrupted factor'. Goyama S et al. 2013
  4. E4strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    Forcing RARA into a homodimer is sufficient to prime mice for APL-like leukemia, localising the oncogenic activity of PML::RARA to the partner-supplied dimerisation function.

    Forced retinoic acid receptor alpha homodimers prime mice for APL-like leukemia · Sternsdorf T, Phan VT, Maunakea ML, et al. · Cancer Cell 2006
    PMID 16473276DOI
    What this evidence cannot show

    Engineered dimerisation constructs in mice with long latency; the model primes for rather than fully reproduces human APL, so PML-specific contributions (nuclear-body disruption, degradation sensitivity) are not excluded.

    Evidence that argues against this (1)
    • PML::RARA's own repression arm is escapable at the drug site: PML B2-domain missense mutations confer clinical arsenic resistance while leaving the fusion intact (2/15 relapsed APL patients), so the partner-supplied module that this class treats as the stable, function-critical element is a mutable drug surface like any other. Goto E et al. 2011
  5. E5strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    ETV6::RUNX1 genomic fusion sequences are detectable in neonatal blood spots of children who later develop ALL at 2-5 years, showing an in-utero first hit that is insufficient for leukemia without a postnatal second event.

    Prenatal origin of acute lymphoblastic leukaemia in children · Wiemels JL, Cazzaniga G, Daniotti M, et al. · Lancet 1999
    PMID 10551495DOI
    What this evidence cannot show

    Small n (six informative of nine plus a twin pair), clone-specific PCR on Guthrie cards with negative results uninformative rather than truly negative.

    Evidence that argues against this (1)
    • The 'in-utero first hit, insufficient without a second event' reading depends on the fusion being common in healthy newborns, and that premise failed to replicate. Screening 1417 umbilical cord blood samples isolated within 24 hours of birth with a highly sensitive real-time RT-PCR, 14 samples were positive on first-run PCR at levels below 10^-5 and 9 survived specific hybridisation - but on repeat analysis of stored mRNA and of flow-sorted CD19+, CD8+ and double-negative fractions (mean 18 million cells sorted per sample) NOT ONE was positive. The authors conclude the level and/or frequency of ETV6-RUNX1-positive cells is markedly lower than previous studies suggested. Backtracking a fusion from the blood spot of a child who did go on to develop ALL is a different and much safer claim than the population statement it is usually paired with, and only the former is supported here. Lausten-Thomsen U et al. 2010
  6. E6strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    t(8;21) and inv(16) leukaemia cells are addicted to native RUNX1: RUNX1 knockdown causes apoptosis, knockdown of AML1-ETO rescues it, and an active RUNX1 allele is usually preserved in these patients.

    Addiction of t(8;21) and inv(16) acute myeloid leukemia to native RUNX1. · Ben-Ami O, Friedman D, Leshkowitz D, et al. · Cell Reports 2013
    PMID 24055056DOI
    What this evidence cannot show

    Abstract only. Two cell lines plus a clinical-genetics inference from allele preservation; not tested in primary cells or in vivo in that paper. Independently supported by Goyama 2013 (PMID 23979164) in AML1-ETO and MLL-AF9 cells.

    Evidence that argues against this (1)
    • Addiction to native RUNX1 is measured in the laboratory; what the patients' cells are actually running on at presentation is often something else. In 54 t(8;21) cases, 26 (48.1%) carried an activating KIT mutation, and allele-specific PCR ordered the two events: at presentation most leukaemic cells carried both, whereas in complete remission AML1-ETO was still detectable and the KIT mutation was not. Two consequences for this claim. First, the proliferative dependency in nearly half of these patients is a kinase lesion acquired after the fusion, which is separately druggable and is the lesion that clears with therapy. Second, the fusion persists in morphological remission, so a strategy aimed at the fusion-RUNX1 axis is aimed at the component that remission does NOT remove - which may be the right target for eradication but is not what the RUNX1-knockdown lethality experiments model. Wang YY et al. 2005
  7. E7strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    RUNX1 is required for the survival of AML1-ETO and MLL-AF9 leukaemia cells through BCL2, RUNX2 can compensate for its loss, combined Runx1/Cbfb deletion inhibits MLL-AF9 leukaemia in mice, and the authors propose inhibiting rather than enhancing RUNX1 activity in fusion-driven AML.

    Transcription factor RUNX1 promotes survival of acute myeloid leukemia cells. · Goyama S, Schibler J, Cunningham L, et al. · Journal of Clinical Investigation 2013
    PMID 23979164DOI
    What this evidence cannot show

    Abstract only. Human cord blood transduction models plus mouse genetics; RUNX1 overexpression conversely inhibits normal cord blood growth, so the effect is dose-dependent and a therapeutic window is not demonstrated. Directly relevant to KMT2A-r as well as to core-binding-factor disease.

    Evidence that argues against this (1)
    • The therapeutic proposal in this claim - inhibit rather than enhance RUNX1 activity - has a directly relevant in vivo answer that runs the other way. Conditional deletion of Runx1 in ADULT mice (as opposed to the embryonic-lethal germline null) does not abolish haematopoiesis, but it produces exactly the phenotypes a therapy would have to accept: lineage-specific blocks in B- and T-cell maturation, pronounced inhibition of common lymphoid progenitor production, inefficient platelet production, and a MILD MYELOPROLIFERATIVE phenotype with increased peripheral neutrophils, expanded myeloid progenitors and extramedullary haematopoiesis. Systemic RUNX1 inhibition therefore reproduces, in normal marrow, a pre-malignant myeloid expansion. The RUNX2 compensation reported at [E7] is part of the same picture: a target whose loss is buffered in the tumour by a paralogue and unbuffered in the host is the wrong way round. Growney JD et al. 2005
  8. E8strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    Integrating ChIP-seq and RNA-seq, ETV6-RUNX1 functions primarily through competition for RUNX1 binding sites and transcriptional repression; in frank leukaemia, knockdown of RUNX1 or CBFB causes cell death, recapitulated by an allosteric CBFB inhibitor, and this RUNX1 addiction extends to other genetic subtypes of paediatric and adult B-ALL while sparing normal haematopoiesis.

    Regulome analysis in B-acute lymphoblastic leukemia exposes Core Binding Factor addiction as a therapeutic vulnerability · Wray JP, Deltcheva EM, Boiers C, Richardson SE, Chhetri JB, Brown J, Gagrica S, Guo Y, Illendula A, Martens JHA, Stunnenberg HG, Bushweller JH, Nimmo R, Enver T · Nature Communications 2022
    PMID 36411286DOI
    What this evidence cannot show

    Cell lines, knock-in hPSC-derived B-lineage cells and a tool compound rather than patient material or an in vivo therapeutic study; 'addiction' is measured as cell death on knockdown, and because the same dependency is reported in other B-ALL subtypes it is not specific to the fusion. Read: abstract only.

    Evidence that argues against this (1)
    • 'While sparing normal haematopoiesis' is the load-bearing half of this claim and human genetics contradicts it at the relevant dose. Familial platelet disorder with predisposition to AML is caused by heterozygous loss of one RUNX1 allele - nonsense mutations, intragenic deletion, or missense at R166 and R201 - co-segregating with disease across six pedigrees; marrow or blood from affected individuals showed decreased megakaryocyte colony formation, demonstrating that RUNX1 dosage, not just RUNX1 absence, affects normal megakaryopoiesis. A 50% reduction in RUNX1 function in an otherwise normal human causes a lifelong platelet defect AND predisposes to leukaemia. A core-binding-factor inhibitor is pharmacological haploinsufficiency; the sparing observed in a short-term experiment is not the relevant timescale, and the on-target risk is leukaemogenic rather than merely cytopenic. Song WJ et al. 1999
Connected to