Oncofusion

start / graph / hyp-ledgf-is-an-elongation-licence-not-a-chromatin-tether

Hypothesisspeculative confidence 1 of 5: Our own inference or analogy. Not shown to be true.Q33-ledgf-axis

Hypothesis: LEDGF licenses elongation at fusion target genes rather than anchoring the fusion, which predicts that a LEDGF agent is epistatic with BET and CDK9 inhibition and invisible in a fusion-ChIP assay

also called LEDGF licence model

We propose that LEDGF does not hold the leukaemia protein onto DNA but instead lets it switch genes on. If that is right, a LEDGF drug would look like it is doing nothing in the standard test, and would not add much to drugs that block the same switch.

The same thing, for a scientist

The graph's standing model is that menin bridges the KMT2A fusion to LEDGF, whose PWWP anchors the complex on H3K36me2/3 chromatin - the clamp. Direct occupancy measurement contradicts the fusion half of that: LEDGF knockdown **increases** fusion occupancy while reducing AF4, CDK9, BRD4, P-TEFb, Ser2P Pol II and H4K16ac at the same loci, and an IBD-binding-dead fusion still reaches its target genes. The alternative on offer here is that LEDGF's function for the fusion is a **licence for productive elongation**, delivered through the H4K16ac-BRD4-P-TEFb-SEC arm, not an address. The hypothesis is worth writing down rather than leaving implicit because it makes four predictions that differ from the clamp model's, three of them cheap, and because every drug-discovery decision on this axis - which module to target, which pharmacodynamic readout to use, which combinations to avoid - follows from which model is right.

Go deeper
Predictions that differ from the clamp model
#PredictionClamp model saysTestable by
1LEDGF depletion leaves fusion occupancy unchanged or increased while collapsing nascent transcription at the same locioccupancy fallsCUT&RUN + PRO-seq in a human KMT2A::AFF1 line
2A LEDGF-directed agent is epistatic with BET or CDK9 inhibition, not synergisticcombination should be at least additivethree-drug matrix in SEM
3An H4K16 acetyltransferase is the link between LEDGF and BRD4; depleting it phenocopies LEDGF loss at these genesno predictionKAT7/MOF knockdown plus ChIP
4Some AEP/SEC subunit carries an IBD-binding motifno predictionsequence match against the published IBM consensus, then co-IP

Prediction 2 is the one with the largest practical consequence and the lowest cost. It is also the one most likely to be run accidentally and misread: a combination that shows no synergy is normally recorded as a failed combination rather than as evidence about mechanism.

What would kill it
  • Fusion occupancy at HOXA9/MEIS1 falls on acute, deep LEDGF degradation in a human KMT2A::AFF1 line. The existing negative comes from shRNA in murine cells; a degron removing LEDGF in two hours is a fairer test and could reverse the result.
  • A LEDGF-directed agent proves synergistic with a CDK9 or BET inhibitor at fusion target genes.
  • The elongation factors turn out to fall simply because transcription fell. A transcription-inhibition control - blocking elongation directly and asking whether BRD4 and P-TEFb occupancy drops the same way - would settle it, and no published experiment includes one.
  • LEDGF is shown to contact the fusion's chromatin residence directly by live-cell single-molecule tracking, with a measurable change in fusion residence time on LEDGF loss.
Why the two models have coexisted

They have coexisted because they are supported by different assay classes. The clamp model rests on functional genetics - IBM2 point mutants abolish transformation, Psip1 knockout blocks leukaemogenesis - and those results are solid and are not in dispute here. The licence model rests on occupancy measurement. A factor can be functionally required and positionally dispensable at the same time, and that is precisely the claim.

The field has read the functional results as if they were occupancy results. That is the error this hypothesis names.

Facts and figures
statementLEDGF is functionally required by KMT2A fusions as a licence for productive elongation through the H4K16ac-BRD4-P-TEFb-SEC arm, not as a chromatin tether for the fusion protein.
origindata-pattern
prior probability0.55
predictions
[
  "acute LEDGF degradation in a human KMT2A::AFF1 line leaves fusion occupancy unchanged or increased while collapsing nascent transcription at the same loci",
  "a LEDGF-directed agent is epistatic with BET or CDK9 inhibition at fusion target genes rather than synergistic",
  "depleting an H4K16 acetyltransferase phenocopies LEDGF loss at fusion target genes",
  "at least one AEP/SEC subunit carries an IBD-binding motif"
]
experiments
[
  "exp-psip1-inducible-knockdown-established-kmt2a-aff1-xenograft",
  "exp-ledgf-occupancy-menin-inhibitor-monotherapy-panel"
]
falsifiers
[
  "fusion occupancy at HOXA9/MEIS1 falls on acute degron-mediated LEDGF removal in a human KMT2A::AFF1 line",
  "a LEDGF-directed agent is synergistic rather than epistatic with CDK9 or BET inhibition at fusion target genes",
  "BRD4 and P-TEFb occupancy falls equivalently under direct elongation blockade, showing the LEDGF-knockdown result is a downstream consequence of reduced transcription",
  "single-molecule tracking shows a measurable reduction in fusion chromatin residence time on LEDGF loss"
]
How this could bite us back
  • If the licence model is right, a LEDGF-directed agent duplicates the mechanism of BET and CDK9 inhibitors, which already have clinical data and known toxicities.
  • The standard pharmacodynamic readout for this axis - fusion ChIP - would show nothing or the wrong direction, so a working drug could be discarded as inactive.
  • The model does not rescue the target: it relocates the mechanism without making the protein more selective, more ligandable or less essential elsewhere.
Still open
  1. Would acute degradation give a different answer from chronic knockdown? Every existing measurement is a knockdown over days, which allows compensation.
  2. Does the licence model explain the IBM2 transformation defect better or worse than the clamp model? Both predict it, which is why the discriminating experiments are occupancy and combination experiments rather than transformation assays.
  3. If the licence model is right, is the right target LEDGF at all, rather than the H4K16 acetyltransferase or the SEC contact it delivers?
Every source behind this entry (4)
  1. E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    LEDGF knockdown in MLL-AF9-transformed HSPCs increased fusion occupancy while reducing AF4, CDK9, BRD4, P-TEFb, Ser2P Pol II and H4K16ac at Hoxa9 and Meis1, and an MLL-AF9 F129A mutant unable to bind LEDGF still associated with target genes; the authors describe a recruitment-independent role for LEDGF in MLL-fusion function.

    ASH1L Links Histone H3 Lysine 36 Dimethylation to MLL Leukemia. · Zhu L, Li Q, et al. · Cancer Discov 2016
    PMID 27154821DOI
    What this evidence cannot show

    Murine, in vitro, ChIP-qPCR at two loci, knockdown not degradation, no transcription-inhibition control. Read via efetch db=pmc after fullTextXML returned HTTP 500.

    Evidence that argues against this (1)
    • The conditional Psip1 knockout paper interprets its own data as corroborating that LEDGF tethers the fusion, on the basis that an F148A/F151A MLL-ENL cannot transform - a functional rather than positional argument, which is exactly the conflation this hypothesis identifies. El Ashkar S et al. 2018
  2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    An independent group reproduced the direction in a mouse ESC TetR system, seeing a slight increase in MLL(N) ChIP with Ledgf siRNA at 30% knockdown, and showed separately that MLL-AF4 and menin co-stabilise each other's binding in SEM cells.

    PMID 28076791DOI
    What this evidence cannot show

    30% knockdown; artificial recruitment platform. Open access (PMC5263239).

    Evidence that argues against this (1)
    • At 30% knockdown the experiment has little power, and the authors describe it as only partly supporting the original result. Kerry J et al. 2017
  3. E3established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    The functional genetics the clamp model rests on are solid and are not disputed: IBM2 point mutations abolish leukaemogenic transformation by MLL-AF9, and conditional Psip1 knockout blocks MLL-ENL leukaemogenesis in vivo.

    PMID 25305204DOI
    What this evidence cannot show

    Cited to state the strength of the rival evidence honestly. Transformation assays in murine progenitors.

    Evidence that argues against this (1)
    • A requirement for transformation is compatible with either model, since a factor that licenses elongation is as necessary for transformation as one that anchors the complex. Zhu L et al. 2016
  4. E4established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    The IBD recognises a conserved unstructured IBM consensus motif present in multiple cellular proteins, which makes prediction 4 - that an AEP/SEC subunit carries an IBM - a well-defined sequence question.

    PMID 26245978DOI
    What this evidence cannot show

    The motif is degenerate and phospho-regulated, so a sequence match would be a hypothesis rather than a result.

    Evidence that argues against this (1)
    • A short degenerate motif over-predicts badly; a genome-scale IBM match would return many false positives and a negative result would be more informative than a positive one. Sharma S et al. 2018
Connected to