Oncofusion

start / graph / exp-calibrated-early-timecourse-fusion-profiling-under-menin-inhibition

Experimentspeculative confidence 1 of 5: Our own inference or analogy. Not shown to be true.Q83-fusion-fate

Experiment: spike-in-calibrated, fusion-attributable occupancy of KMT2A::AFF1 in the first 24 hours of menin inhibition, scored for gains as well as losses

also called the early timecourse · relocation test · calibrated fusion CUT&RUN under revumenib

Tag the fusion protein so we can see it specifically, give the drug, and take chromatin snapshots every few hours instead of waiting three days. Add a known amount of foreign chromatin to every sample so we can tell a real global change from a scaling artefact, and look for new sites as carefully as for lost ones.

The same thing, for a scientist

Every published chromatin measurement in this class is at 3 days or later, depth-normalised rather than spike-in calibrated, read out with an antibody that cannot separate the fusion from wild-type MLL1, and analysed for losses only [E1]. This experiment inverts all four choices at once, because each of them independently makes a redistribution invisible. **System.** SEM (infant-derived t(4;11) KMT2A::AFF1), with an HA or dTAG knock-in at the endogenous fusion allele so that attribution is by epitope rather than by subtraction - the approach already validated for MLL::AF9 [E2] and the one this graph's attribution audit identified as one of only two clean discriminators. RS4;11 as a second KMT2A::AFF1 background and ML2 as the genetics-only attribution control. **Perturbation.** Revumenib at 100 nM and 1 microM, the concentrations at which menin displacement is near-complete [E3], plus a DMSO arm and a MEN1-degron arm to separate pharmacology from menin loss. **Timepoints.** 0, 1, 2, 4, 8, 24 hours and 72 hours. The 72-hour point reproduces the published condition and anchors the new data to it; everything before 8 hours is unexplored. **Calibration.** Drosophila or yeast chromatin spike-in at a fixed cell-number ratio in every sample, so that a genuine global loss of fusion occupancy is distinguishable from a rescaling. This is the single most important design element: without it, 'the fusion stays' and 'the fusion goes' produce the same depth-normalised heat map. **Readouts, co-measured.** (i) Fusion occupancy by CUT&RUN against the tag. (ii) Menin occupancy, as the internal positive control - if menin does not leave, the experiment failed. (iii) Nascent transcription (TT-seq or PRO-seq) at the same timepoints, which is what makes the stays-but-stops-working reading testable. (iv) Fusion protein level by western with cycloheximide chase at 0 and 24 h, plus a bortezomib arm, to separate degradation from displacement [E4]. **Analysis, pre-specified.** Union peak set from all arms; differential in both directions; gains reported as a first-class result with their genomic distribution; intersection with a matched Bio-CAP or equivalent uCpG map made in the same cells; and the **residual occupancy fraction** reported as the primary quantitative endpoint.

Go deeper
What each result would mean
observation at 1-8 hreading
fusion signal falls globally, spike-in-calibrated, no gainseviction - the field's null, and it would be established for the first time
fusion signal retained at most sites, transcription collapsesstays and stops working - the reading Krivtsov's day-3 data already favour
fusion signal falls at DMSO peaks and rises at previously empty uCpGrelocation - and the new sites are the resistance substrate
fusion protein level falls, rescued by proteasome inhibitiondegradation - occupancy loss is a consequence, not the mechanism

These are mutually exclusive at the level of the primary endpoint, which is why the experiment is worth doing as one design rather than four.

Controls that decide whether the result is believable
  • Menin lane. Must show near-complete loss by 4 h at 1 microM. If it does not, the drug arm is uninterpretable and the fusion result means nothing.
  • Spike-in ratio. Reported per sample, with the calibration factor in the figure, not buried.
  • Tag control. An untagged parental line processed identically, to bound background at the tag antibody.
  • Wild-type MLL1 lane. ChIP with a C-terminal MLL antibody, which sees wild-type MLL1 and not the fusion, so the two species can be compared in the same cells under the same drug.
  • Degron arm as the acute reference. Degron-induced fusion degradation loses occupancy with no matching loss from the MLL-C ChIP and no change in H3K4me3 within three hours 2, which supplies the shape of a pure protein-loss curve for comparison.
Facts and figures
designEndogenous-allele epitope-tagged KMT2A::AFF1 in SEM (plus RS4;11 and ML2 controls); revumenib 100 nM and 1 microM, DMSO, MEN1-degron arm; timepoints 0, 1, 2, 4, 8, 24, 72 h; Drosophila chromatin spike-in on every sample
readoutTag CUT&RUN (fusion), menin CUT&RUN (positive control), MLL-C ChIP (wild-type MLL1), nascent transcription (TT-seq or PRO-seq), fusion protein level by western with cycloheximide chase and a bortezomib arm
controlsUntagged parental line; per-sample spike-in calibration factor reported; menin lane must show near-complete loss by 4 h at 1 microM; degron arm as the pure protein-loss reference
primary endpointResidual fusion occupancy fraction at baseline peaks, spike-in calibrated, per timepoint
cost tierhigh
timeline9-12 months including endogenous tagging; 4 months if run first with a shared-epitope antibody plus the ML2 genetics control
shown for fusionproposed for KMT2A::AFF1 in an infant-derived line, which no existing menin-inhibitor occupancy dataset covers
secondary endpoints
[
  "number and genomic distribution of treated-only peaks",
  "overlap of treated-only peaks with a matched uCpG map made in the same cells",
  "fusion share of all uCpG sites over time, against the 20% baseline",
  "time lag between occupancy change and nascent transcription change"
]
How this could bite us back
  • This experiment could show that the approved drug leaves the oncoprotein on chromatin while patients respond, which would be read as undermining a therapy that is producing complete remissions in relapsed paediatric leukaemia. The response data are clinical and would not be affected; the mechanistic story would be. That distinction has to be made loudly if the result comes out that way.
  • If relocation is real, the new sites are a resistance substrate that combination design has not considered, and every current combination partner was chosen on the assumption that the drug clears the address.
  • A negative result at every timepoint would strengthen the menin-interface hypothesis and would also retire a cheap explanation for non-response - after which the unexplained fraction of primary resistance has nowhere to hide except cell state and PRC1.1 status.
  • Endogenous tagging changes the protein. A tag at the fusion junction could perturb the very interface under study, and the untagged parental comparison is a viability control, not a chromatin control.
Still open
  1. Is a knock-in tag on the endogenous KMT2A::AFF1 allele in SEM feasible, given that the fusion allele and the wild-type allele share the entire 5' sequence the tag would target? An N-terminal tag labels both species; only a junction-proximal or partner-side tag is fusion-specific.
  2. How fast is menin displacement at 100 nM? Nobody has published a sub-day menin occupancy time course, so the experiment's own control lane is itself an unmeasured quantity.
  3. Should the uCpG map be made per timepoint? If the fusion protects CpG from methylation, its removal could change the substrate during the experiment, which would make a fixed baseline map the wrong denominator.
Every source behind this entry (4)
  1. E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    The entire published chromatin literature under menin inhibitors is at 3 days or later, quantified in TSS-anchored windows, depth-normalised without a spike-in on the fusion lane, and analysed for losses; two of eight datasets tabulate gains and both do so for a protein other than the fusion.

    PMID 31821784DOI
    What this evidence cannot show

    The census supporting this is ana-what-menin-inhibition-has-actually-been-measured-on-chromatin, bounded by a 61-record PubMed query plus full texts of the pivotal papers; two of the eight datasets were read at abstract level only, so the uniformity claim has that much slack.

    Evidence that argues against this (1)
    • Krivtsov did use a Drosophila chromatin spike-in for the CUT&RUN arm, so calibration is not foreign to the field; it was simply not applied to the ChIP-seq that carries the MLL1n conclusion. Krivtsov AV et al. 2019
  2. E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    Epitope-tag ChIP on an HA-FKBP12-tagged MLL::AF9 gives unambiguous fusion attribution, and degron-induced degradation reduces fusion but not wild-type MLL1 occupancy within three hours with no change in H3K4me3 - establishing both the attribution method and the timescale on which acute fusion loss is detectable.

    PMID 35245435DOI
    What this evidence cannot show

    MLL::AF9 rather than KMT2A::AFF1, and the tagged fusion is expressed from a construct rather than solely from the endogenous allele in all arms; this graph already records the distinction between a degron on the fusion allele and on an overexpressed copy. The three-hour result shows the assay's temporal resolution, not that a menin inhibitor acts on that timescale.

    Evidence that argues against this (1)
    • A knock-in tag on the endogenous KMT2A::AFF1 allele in SEM does not yet exist to my knowledge, so this experiment carries a line-construction step of several months before any data; a nanobody or degron-free approach would trade attribution quality for speed. Krivtsov AV et al. 2019
  3. E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    Revumenib at 100 nM produces near-complete and global displacement of menin from chromatin in KMT2A::AFF1 MV4;11 cells, which fixes the concentration at which the menin control lane should read out.

    MEN1 mutations mediate clinical resistance to menin inhibition. · Perner F, Stein EM, Wenge DV, et al. · Nature 2023
    PMID 36922589DOI
    What this evidence cannot show

    MV4;11 rather than SEM; the treatment duration for that ChIP-seq is not stated, so the concentration is anchored but the kinetics are not. Menin displacement at 100 nM may take longer than 4 h, in which case the early timepoints will show partial displacement and must be interpreted against the menin lane rather than against an assumption.

    Evidence that argues against this (1)
    • If menin displacement itself is slow at 100 nM, the 1-4 h timepoints measure a partially inhibited state and a relocation could be missed for the trivial reason that the perturbation has not happened yet. A 1 microM arm and a per-timepoint menin quantification are the mitigation, and the design includes both. Krivtsov AV et al. 2019
  4. E4strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    Menin inhibitors reduce menin protein levels and shorten menin half-life without changing MEN1 mRNA, through inhibitor-promoted menin-CHIP/Hsp70 ubiquitination, rescuable by proteasome or E1 inhibition - so a protein-level and proteasome-rescue arm is required to interpret any occupancy change as displacement rather than loss of protein.

    PMID 31497350
    What this evidence cannot show

    Demonstrated for menin, not for the fusion, in MV4;11 and THP-1 with the tool compound MI-503; no equivalent measurement of KMT2A fusion protein stability under a menin inhibitor exists anywhere. Am J Cancer Res has no DOI on the PubMed record; the article is available as PMC6726985.

    Evidence that argues against this (1)
    • If the fusion is not itself destabilised, the proteasome arm adds cost without information. The cheap prior test is a single western blot time course of the fusion under drug, which nobody appears to have published and which should be run first. Krivtsov AV et al. 2019
Connected to