Oncofusion

start / graph / hyp-fusion-continuously-required-not-hit-and-run

Hypothesisstrong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.Q1b-genesis2

Hypothesis: KMT2A::AFF1 is continuously required, not a hit-and-run initiator whose job is done once the epigenome is set

also called hit-and-run oncogenesis · continuous dependency · is the fusion still needed

One worry about drugging the fusion is that it might only be needed at the very beginning, like a match that lights a fire and is then thrown away. If that were true, removing it later would not put the fire out. The evidence says the fusion is more like the gas supply than the match.

The same thing, for a scientist

A hit-and-run model would say that the fusion installs a self-sustaining epigenetic state (H3K79me2 domains, HOXA9/MEIS1 output, a bivalent-to-active chromatin configuration) and is thereafter dispensable. Under that model every interface-directed drug in this project fails, however well it engages its target. The hypothesis stated here is the opposite: the fusion is continuously load-bearing, and the leukaemic state decays when it is removed. Three converging lines support it. First, degron and tet-off fusion-removal experiments collapse the program and regress established leukaemia (`mechanism-fusion-dependency-reversibility-kinetics`). Second, transient siRNA knockdown of MLL-AF4 in t(4;11) lines reduced clonogenicity, induced caspase-3-dependent apoptosis, lowered HOXA7, HOXA9 and MEIS1, reduced CD133 with evidence of differentiation, and reduced leukaemic engraftment in xenografts [E1] — with transient reagents, so the dependency cannot be an artefact of permanent genetic ablation. Third, the fusion is truncal and retained at relapse rather than discarded once the state is established (`mechanism-fusion-clonal-stability-diagnosis-to-relapse`). The hypothesis is not yet closed because the decisive experiment, endogenous-locus degron removal of KMT2A::AFF1 specifically in a human t(4;11) context with long washout, has not been published.

Go deeper
What a hit-and-run result would look like1 deeper

Removal of the fusion would produce an initial transcriptional dip followed by recovery, with H3K79me2 domains and HOXA cluster expression re-establishing themselves in the absence of the fusion, and xenograft engraftment preserved. Nothing like this has been reported for KMT2A fusions, but the absence is partly an absence of the right experiment: most reversibility data come from MLL::AF9 rather than KMT2A::AFF1, and washout periods are short relative to the half-life of a stable chromatin domain.

The partial-hit-and-run variant, which is the realistic worry

A softer and more dangerous version: the fusion is required to establish the state and also to maintain it, but only above some occupancy threshold, so a drug that achieves 70% interface disruption produces no phenotype at all. That is a dose-response question, not a mechanism question, and it is testable by degron titration. It is also the version most consistent with the clinically observed decoupling in which menin inhibitors preserve target engagement but lose efficacy in heavily pretreated KMT2A::AFF1 samples.

Why this matters for the menin hypothesis specifically

Menin-interface disruption does not remove the fusion; it removes the fusion from chromatin. If the leukaemic state is maintained by the fusion's continued presence at target loci, displacement is equivalent to removal and the hypothesis holds. If the state is maintained by something the fusion installed and then left behind (a DNA methylation configuration, a stable H3K79me2 domain, a rewired enhancer contact), displacement fails while degradation might still work. The mechanism-kmt2a-r-dna-methylation-signature and mechanism-dot1l-h3k79-methylation entries hold the candidate memory carriers; this hypothesis is the frame that decides between displacement and degradation as the correct modality.

Facts and figures
statementThe leukaemic transcriptional and self-renewal program of KMT2A::AFF1 requires the continuous presence and chromatin occupancy of the fusion protein; removing or displacing it at any point after establishment causes the program to decay rather than persist autonomously.
originliterature-gap
prior probability0.8
predictions
[
  "Endogenous-locus degron depletion of KMT2A::AFF1 in a t(4;11) line causes loss of HOXA9/MEIS1 output that does NOT recover over a 14-day washout while the degron is maintained.",
  "H3K79me2 domains at HOXA decay with a measurable half-life after fusion removal rather than being maintained.",
  "Fractional degradation produces a graded, not thresholded, loss of program output; if a threshold exists, it is above 50% degradation.",
  "Menin-interface disruption phenocopies fusion degradation at the transcriptional level in a sensitive line, but diverges in lines with an installed DNA-methylation memory."
]
experiments
[
  "exp-degron-washout-recovery-kmt2a-aff1"
]
analyses
[]
falsifiers
[
  "HOXA9/MEIS1 expression and self-renewal recover to baseline during continued degron-mediated fusion depletion, i.e. the program re-establishes without the fusion.",
  "Established xenograft leukaemia continues to propagate after complete, verified fusion degradation.",
  "H3K79me2 domains at fusion target loci are stable over multiple cell divisions after fusion removal with no loss of transcription."
]
How this could bite us back
  • If the hypothesis is wrong even partially, interface disruption fails while the target is fully engaged, and the clinical signature of that failure looks exactly like resistance, which would send the field hunting for resistance mutations that do not exist.
  • A continuously required driver is also a continuously present antigenic and functional load; chronic suppression rather than eradication means indefinite dosing in infants, with all the developmental toxicity that implies.
Still open
  1. What is the actual washout duration needed to exclude hit-and-run? Every published fusion-removal experiment reads out within days. A 3-4 week degron hold is the minimum that would be persuasive and nobody has run it.
  2. Is displacement from chromatin (menin inhibition) truly equivalent to degradation? A head-to-head in the same line, with matched occupancy loss measured by CUT&RUN, would tell the project whether to invest in degraders as a hedge.
  3. If a threshold exists, what fraction of fusion must be removed? This single number determines whether a partial-occupancy small molecule can ever be sufficient.
Every source behind this entry (3)
  1. E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    Transient siRNA depletion of MLL-AF4 in t(4;11) human leukaemia cells reduces proliferation and clonogenicity, induces caspase-3-dependent apoptosis, lowers HOXA7/HOXA9/MEIS1 and CD133, and reduces engraftment in xenografted SCID mice.

    Targeting MLL-AF4 with short interfering RNAs inhibits clonogenicity and engraftment of t(4;11)-positive human leukemic cells · Thomas M, Gessner A, Vornlocher HP, Hadwiger P, Greil J, Heidenreich O · Blood 2005
    PMID 16046533DOI
    What this evidence cannot show

    Two cell lines (SEM, RS4;11) plus xenograft; siRNA transfection before transplant rather than treatment of established disease, so it tests whether the fusion is needed to initiate engraftment more cleanly than whether it is needed to maintain an established leukaemia. Off-target siRNA effects are possible although the HOXA/MEIS1 readout is mechanistically coherent.

    Evidence that argues against this (1)
    • Hit-and-run is not hypothetical in this lineage. Green et al. transiently expressed Bcl6 in murine haematopoietic stem/progenitor cells and obtained mature B-cell lymphomas that lacked Bcl6 expression and lacked Bcl6 target-gene repression, while carrying epigenetic changes conserved from the HSPC stage through to the mature B cell; they reached this design from the human observation that gain of 3q27.2 (containing BCL6) associates with adverse outcome in DLBCL without altering BCL6 transcript levels or target-gene repression. This is the exact model this hypothesis denies, demonstrated for a transcription-factor oncogene, in B-lineage, initiated in HSPCs. Transient siRNA knockdown (E1) cannot distinguish the two models, because a transient reagent tests whether the fusion is needed during the window of knockdown, not whether an already-installed state would persist without it. Read: abstract only. Green MR et al. 2014
  2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    The fusion remains the dominant clonal lesion at relapse while cooperating mutations are lost, which is inconsistent with the fusion being discarded after initiation.

    The landscape of somatic mutations in infant MLL-rearranged acute lymphoblastic leukemias · Andersson AK, Ma J, Wang J, Chen X, Gedman AL, Dang J, et al. · Nature Genetics 2015
    PMID 25730765DOI
    What this evidence cannot show

    Retention of a genomic lesion does not by itself prove continued functional requirement; a passenger can be retained. This is supporting, not decisive, evidence.

    Evidence that argues against this (1)
    • Retention of the fusion does not imply that the fusion is determining the cell state, and there is a direct demonstration in an MLL-fusion system. Liu et al. took transplantable MLL-AF9 murine leukaemia cells carrying doxycycline-inducible Yamanaka factors and efficiently converted them into iPS cells that formed teratomas and produced chimeras; RNA-seq showed reversible global expression patterns between the interchangeable leukaemic and pluripotent states on inactivation or reactivation of MLL-AF9. The fusion-bearing genome is therefore compatible with a completely non-leukaemic state, and the leukaemic programme is a function of cellular context as much as of fusion presence. That undercuts the inference from genomic retention at relapse (E2) to continuous functional requirement. Read: abstract only. Liu Y et al. 2014
  3. E3moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    A fusion-positive clone can persist in a healthy individual without producing disease, showing that fusion presence alone is not the same as an active leukaemic state.

    PMID 19411627DOI
    What this evidence cannot show

    Single case; cuts both ways, since it shows the fusion is not sufficient for the leukaemic state as well as showing the state is not autonomous of the cell context.

    Evidence that argues against this (1)
    • The twin clearance case speaks to establishment, not maintenance, and the maintenance question has a positive answer in the MLL family. Horton et al. built a Tet-regulatable MLL-ENL system in primary murine haematopoietic progenitors and generated immortalised myeloid lines that depend on continued MLL-ENL expression for survival and proliferation: on doxycycline the cells terminally differentiate or die, multiple Hoxa genes fall, and the authors showed the Hox reduction is specific to loss of MLL-ENL rather than a consequence of differentiation. So once the immortalised state exists, an MLL fusion IS continuously load-bearing. This is the strongest published support for this hypothesis and is recorded here because it is the direct rebuttal to reading E3 as evidence for autonomy. It is a murine, retroviral, MLL-ENL myeloid system, not human t(4;11) B-ALL, which is exactly the gap exp-degron-washout-recovery-kmt2a-aff1 exists to close. Read: abstract only. Horton SJ et al. 2005
Connected to