Experiment: titrate wild-type KMT2A up and down in one leukemia cell and watch what happens to the fusion's grip on chromatin
also called the non-monotonic allele-dosage curve · how much wild-type KMT2A does the fusion want?
Two good papers disagree: one says the leukemia needs the healthy copy of KMT2A, the other says adding more of it pushes the leukemic protein off the genes it controls. Dial the healthy copy up and down in one cell and measure both, and you find out whether there is an amount that is bad for the leukemia.
The same thing, for a scientist
A degron-tagged endogenous wild-type *KMT2A* allele in a KMT2A::AFF1 line (SEM or RS4;11), titrated across roughly a ten-fold range in both directions - auxin-driven depletion downward, and CKII inhibition or a taspase1-site knock-in upward - with spike-in-normalised CUT&RUN for the fusion at its own loci as the primary readout. The experiment exists because two verified results in this graph point opposite ways: the wild-type Mll allele is genetically **required** for MLL-AF9 leukemogenesis and maintenance [E1], and **increasing** wild-type MLL1 occupancy displaces MLL-AFF1, AFF4 and ELL2 from chromatin and prolongs survival [E2]. Both cannot be linear. If the curve is non-monotonic, its shape names a therapeutic window on the one allele in this disease that no drug programme currently targets.
The design
Cells. SEM (KMT2A::AFF1) as primary, RS4;11 as replicate; an MLL-AF9 murine line for the in-vivo arm.
The handle. Knock a degron (mAID or dTAG) into the wild-type KMT2A allele only - distinguishable because the fusion allele's 3' end is AFF1. That gives graded downward titration with auxin or dTAG ligand dose. Upward titration uses two orthogonal routes so the result does not depend on one drug: (i) CX-4945 at sub-growth-inhibitory doses, which stabilises full-length MLL1 by blocking CKII phosphorylation near the taspase1 site 2; (ii) a knock-in of the non-cleavable D-to-A mutations at 2666 and 2718 in the wild-type allele, which is the clean genetic version and has no pharmacology to argue about.
Primary readout. Spike-in-normalised CUT&RUN for the fusion, using an AFF1 C-terminal antibody so that only the fusion is pulled, at HOXA9, MEIS1 and genome-wide. Spike-in normalisation is not optional: a global occupancy shift normalises away otherwise.
Secondary readouts. Wild-type KMT2A occupancy (C-terminal antibody, wild-type only); AFF4 and ELL2 occupancy, since those are what moved in the published result; HOXA9/MEIS1 expression; and proliferation.
The measurement that matters. Plot fusion occupancy against wild-type KMT2A protein abundance across the whole titration. The hypothesis is that the curve has an interior maximum.
What each shape would mean
- Interior maximum (non-monotonic). Both published results are right and the leukemic cell sits near an optimum. Either direction far enough is therapeutic, which is an unusual and valuable thing to be able to say about a target, and it immediately predicts that wild-type KMT2A copy number or expression should be a response biomarker for CKII inhibition.
- Monotonically decreasing (more wild-type, less fusion). The competition model is general; Thiel's requirement is about establishment rather than maintenance, and the therapeutic direction is simply 'up'.
- Monotonically increasing (more wild-type, more fusion). The prebinding model dominates; CX-4945's benefit in mice is not running through MLL1 at all, and both
hyp-fusion-is-exempt-from-restraint-not-free-of-itandhyp-kmt2a-half-loses-abundance-control-not-activity-controllose an important pillar. - Flat. Occupancy is not competitively determined, and the locus-level framing that both of those hypotheses depend on is wrong.
Every one of the four outcomes is informative, which is the main argument for running it.
| design | Degron knock-in (mAID or dTAG) at the wild-type KMT2A allele of SEM and RS4;11, distinguishable from the fusion allele by its 3' sequence, giving graded downward titration; upward titration by two orthogonal routes - sub-growth-inhibitory CX-4945, and a non-cleavable D2666A/D2718A knock-in of the wild-type allele. Roughly ten-fold range in each direction, at least five dose points. |
|---|---|
| readout | Primary: spike-in-normalised CUT&RUN for the fusion (AFF1 C-terminal antibody) at HOXA9, MEIS1 and genome-wide, plotted against wild-type KMT2A protein abundance. Secondary: wild-type KMT2A occupancy (C-terminal antibody), AFF4 and ELL2 occupancy, HOXA9/MEIS1 expression, proliferation. |
| controls | Spike-in chromatin in every CUT&RUN (without it a global occupancy shift is invisible); a non-KMT2A-rearranged line carried through the same titration; the enantiomer/inactive-analogue control for CX-4945 where available; and the genetic non-cleavable arm as the orthogonal check on every pharmacological result. |
| cost tier | med |
| timeline | 6-9 months |
discriminating outcomes
{
"non_monotonic": "both published results hold; the leukemic cell sits near an optimum and wild-type KMT2A dosage becomes a candidate response biomarker",
"decreasing": "competition dominates; the therapeutic direction is to raise wild-type KMT2A",
"increasing": "prebinding dominates; CX-4945's in-vivo benefit is not via MLL1 and two hypotheses in this beat lose a pillar",
"flat": "occupancy is not competitively determined and the locus-level framing is wrong"
}tests hypotheses
[ "hyp-fusion-is-exempt-from-restraint-not-free-of-it", "hyp-kmt2a-half-loses-abundance-control-not-activity-control" ]
- A therapeutic strategy built on raising wild-type KMT2A is defeated by loss or silencing of the wild-type allele, and has a hard ceiling set by how much protein can be stabilised. It is also, uniquely among strategies in this disease, one where under-dosing could plausibly help the leukemia.
- If the curve is non-monotonic, clinical trials that do not measure wild-type KMT2A dosage will average patients on both sides of the optimum and may show no effect for a drug that works well on one side.
- Can a degron be knocked into the wild-type KMT2A allele without also tagging the fusion allele? The two share the entire 5' region, so the tag must go at the 3' end, which places it beside FYRC and SET - the same tagging problem `exp-cyp33-induction-allele-resolved-cutrun` runs into.
- Is wild-type KMT2A abundance in these lines already near an optimum, so that the interior maximum is too flat to resolve? A pilot with three dose points would answer that cheaply before committing to the full design.
- Does the curve differ by partner? KMT2A::AFF1 and KMT2A::MLLT3 recruit different elongation machinery and may compete with wild-type KMT2A differently, which would make allele dosage a partner-specific biomarker rather than a general one.
- E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
The wild-type Mll allele is required for MLL-AF9-induced leukemogenesis and for maintenance of MLL-AF9-transformed cells; menin recruits both wild-type MLL and the fusion to HOX loci, and wild-type MLL controls a distinct component of the fusion-induced histone methylation.
MLL-AF9-induced leukemogenesis requires coexpression of the wild-type Mll allele · Thiel AT, Blessington P, Zou T, Feather D, Wu X, Yan J, Zhang H, Liu Z, Ernst P, Koretzky GA, Hua X · Cancer Cell 17(2):148-159 2010What this evidence cannot show
ABSTRACT ONLY. Mouse genetics in one fusion and in a transplantation model; no human allele-dosage data exist. 'Required' is binary in that design, which is exactly why a titration is needed.
Evidence that argues against this (1)
- The premise that two good papers disagree may be a false conflict, and if it is, the experiment answers a question nobody asked. The requirement result is a mouse MLL-AF9 leukemogenesis and maintenance phenotype; the displacement result is a chromatin-occupancy measurement on human MLL-AFF1. Different species, different partner, different lineage and - crucially - different dependent variables. A wild-type allele can be necessary for transformation AND, when pushed above its normal level, compete the fusion off chromatin, with no contradiction whatever. The design should state which of those it is testing. Cancer Cell 2010
- E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
Stabilising wild-type MLL1 - by taspase1 knockout or by CKII inhibition with CX-4945, which blocks the phosphorylation licensing cleavage - raises its chromatin occupancy and displaces MLL-AFF1, AFF4 and ELL2 from chromatin in SEM cells with reduced target-gene mRNA, and delays progression and prolongs survival in an MLL-AF9 mouse model.
Regulation of MLL/COMPASS stability through its proteolytic cleavage by taspase1 as a possible approach for clinical therapy of leukemia · Zhao Z, Wang L, Volk AG, Birch NW, Stoltz KL, Bartom ET, Marshall SA, Rendleman EJ, Nestler CM, Shilati J, Schiltz GE, Crispino JD, Shilatifard A · Genes & Development 33(1-2):61-74 2019What this evidence cannot show
FULL TEXT read via PMC6317322. CX-4945 is not MLL1-selective; the genetic arm (taspase1 knockout) was done in 293T/MCF7/HCT116 rather than in leukemia cells, and the SEM experiments are pharmacological. Disentangling the two is a stated purpose of this experiment.
Evidence that argues against this (1)
- Neither arm of the proposed dial is a clean manipulation of wild-type MLL1. CX-4945 inhibits casein kinase II, an enzyme with a very large substrate set, so displacement of MLL-AFF1 after CX-4945 is attributable to CKII signalling generally and not specifically to the phosphorylation that licenses taspase1 cleavage. Taspase1 knockout has the same problem from the other side: taspase1 has substrates besides MLL. Any dosage experiment built on these two tools measures the tools as much as the allele, and needs a degron or allelic-series control that moves MLL1 protein level directly. Genes & Development 2019
- E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
MLL1 fusion proteins lack the PHD fingers and require prebinding of a wild-type MLL1 complex, together with CXXC-domain recognition of DNA, for stable HoxA9 association.
Multiple interactions recruit MLL1 and MLL1 fusion proteins to the HOXA9 locus in leukemogenesis · Milne TA, Kim J, Wang GG, Stadler SC, Basrur V, Whitcomb SJ, Wang Z, Ruthenburg AJ, Elenitoba-Johnson KSJ, Roeder RG, Allis CD · Molecular Cell 38(6):853-863 2010What this evidence cannot show
ABSTRACT ONLY. One locus, recruitment rather than transformation as the endpoint, and 'prebinding' is the authors' model. It is the mechanistic reason to expect the low end of the titration to reduce fusion occupancy rather than increase it.
Evidence that argues against this (1)
- This is the mechanistic reason the proposed dosage experiment is hard to interpret rather than a premise supporting it: if the fusion requires prebound wild-type complex for stable association, then raising or lowering wild-type MLL1 changes the fusion's own binding platform, and occupancy changes cannot be attributed to competition versus cooperation. Molecular Cell 2010
- ← related toETV6::RUNX1 (TEL-AML1), t(12;21)(p13;q22)
Counter-evidence: Attarbaschi 2004 (PMID 15356655) found TEL deletion in 55% of TEL/AML1+ children and a WORSE prognosis with it (P=0.034), against the entry's good-outcome framing of secondary changes
- ← testsHypothesis: what the KMT2A half loses is control of its own abundance, not control of its activity
[tested_by] maps the competition arm of the hypothesis
[tested_by] the dosage curve that replaces the wild-type-null comparator P2 assumed
[measures] fusion chromatin occupancy as a function of wild-type KMT2A abundance
[perturbs] degron-titrated wild-type allele, plus non-cleavable knock-in for the upward arm
[references] the pharmacological route for the upward titration, with its own genetic control
[requires] the competition confound means this design needs the dosage curve to interpret a symmetric occupancy change