Separate ncBAF from canonical BAF in KMT2A-rearranged cells: paired degron, paralogue-pair CRISPR, and complex-resolved CUT&RUN
also called Q47 decisive experiment
Three experiments that would settle whether MLL-rearranged leukaemia cells need the minority flavour of the SWI/SNF machine or the common one: degrade each part quickly and watch what happens, knock out the interchangeable pairs together, and map where each flavour sits on the genome.
The same thing, for a scientist
Designed to decide `hyp-ncbaf-not-canonical-baf-is-the-kmt2a-dependency` and, incidentally, the MYC contradiction. Three arms, in increasing cost. **Arm 1 - paired degron (low cost, 8-10 weeks).** dTAG or auxin-inducible degron on BRD9 and on SMARCA4 in matched KMT2A-rearranged lines spanning both lineages: THP-1 (KMT2A::MLLT3, myeloid), MV4-11 (KMT2A::AFF1, myeloid), SEM and RS4;11 (KMT2A::AFF1, B-lineage), with REH and a BCR::ABL1 line as non-KMT2A leukaemia comparators. Readouts at 2, 6 and 24 h: viability, c-Kit/CD11b or B-lineage equivalents, and nascent transcription. **Acute degradation rather than knockdown is the point** - it removes the days-long window in which a paralogue can be upregulated, which is the alternative explanation the whole hypothesis rests on. **Arm 2 - paralogue-pair CRISPR (medium cost).** Dual-guide knockout of SMARCA4+SMARCA2, ARID1A+ARID1B, SMARCD1+SMARCD2, DPF2+DPF1 and BICRA+BICRAL in the same panel. This is the direct test of the single biggest confound in this beat's dependency data, and no public dataset contains it. **Arm 3 - complex-resolved CUT&RUN (medium cost).** BRD9, BICRA, ARID1A, ARID2 and SMARCA4 against KMT2A-N-terminal and partner-C-terminal occupancy in one KMT2A-rearranged line, with menin-inhibitor treatment as a second condition. Answers which complex sits at the E1-E5 MYC enhancers, whether ncBAF occupancy is fusion-independent, and whether menin inhibition moves it. **Controls that are not optional**: MEN1 and KMT2A degrons as positive controls in arm 1, since the lineage-split panel establishes both as lineage-independent dependencies; and a MYC cDNA rescue in every arm-1 experiment, using a construct validated against an unrelated gene in the same line, because that is the control whose failure created the MYC contradiction.
What each arm decides
| arm | decides | falsifies the hypothesis if |
|---|---|---|
| 1 - paired degron | whether the ncBAF dependency is acute and real, or a slow-knockdown artefact | SMARCA4 degradation kills KMT2A-rearranged lines selectively while BRD9 degradation does not |
| 2 - paralogue-pair CRISPR | whether canonical BAF has a dependency that single-gene CRISPR cannot see | SMARCA4+SMARCA2 or ARID1A+ARID1B shows strong KMT2A selectivity |
| 3 - complex CUT&RUN | which complex is at the fusion's loci and at E1-E5 | cBAF, not ncBAF, is the complex at the enhancers and it overlaps fusion occupancy |
Arm 2 is the one this beat most wants run, because the empty canonical-BAF column in ana-swi-snf-lineage-split-dependency-scan is either the beat's main finding or its main artefact, and nothing currently distinguishes those.
| design | three arms: (1) dTAG/AID degron on BRD9 and SMARCA4 across six lines spanning both lineages and a non-KMT2A comparator, with MEN1 and KMT2A degrons as positive controls and a validated MYC cDNA rescue; (2) dual-guide CRISPR over five paralogue pairs; (3) CUT&RUN for BRD9, BICRA, ARID1A, ARID2, SMARCA4 against fusion occupancy, plus and minus acute menin inhibition |
|---|---|
| readout | viability and differentiation markers at 2/6/24 h; nascent transcription (TT-seq or PRO-seq) for the MYC and HOXA arms; occupancy overlap with fusion-bound loci and with the E1-E5 MYC enhancer orthologues |
| controls | MEN1 and KMT2A degrons as lineage-independent positive controls; BRD9 bromodomain-swap allele for any inhibitor arm; MYC cDNA validated against an unrelated gene in the same line; RAS-mutant and RAS-wild-type KMT2A-rearranged lines to address the NrasG12D confound in the MYC contradiction |
| cost tier | med |
| timeline | arm 1: 8-10 weeks from existing degron lines; arm 2: 12-16 weeks; arm 3: 10-12 weeks |
| known gaps | no infant-derived KMT2A-rearranged line exists; a patient-derived xenograft from an infant would be required to address the index population, and that is outside the cost tier above |
lines
[ "THP-1 (KMT2A::MLLT3, myeloid)", "MV4-11 (KMT2A::AFF1, myeloid)", "SEM (KMT2A::AFF1, B-lineage)", "RS4;11 (KMT2A::AFF1, B-lineage)", "REH (non-KMT2A B-ALL)", "a BCR::ABL1 line" ]
- Arm 1 could produce a clean ncBAF result in cell lines and still be irrelevant to infant KMT2A::AFF1 disease, because no infant line exists and the chromatin state may differ.
- A degron reports the acute consequence of losing a protein, which is not the same as the consequence of inhibiting it chronically with a drug - the clinically relevant perturbation remains untested by any of these arms.
- If arm 2 finds a strong SMARCA4+SMARCA2 dependency, the conclusion is that the existing drug is right and the beat's headline is wrong; the experiment should be run by someone willing to publish that.
- Do degron-tagged KMT2A-rearranged B-lineage lines exist in any collaborating laboratory, or would they have to be made? That determines whether arm 1 is 8 weeks or 6 months.
- Can arm 3 be done in a primary infant sample rather than a line? Occupancy in a cell line may not reproduce the infant chromatin state, and the entire beat is bounded by that gap.
- Is a BRD9 degrader with paediatric-suitable properties available for the in vivo follow-on, or would the arm stop at cell culture?
- E1moderate confidence 3 of 5: One strong study, or several smaller consistent ones.dataset
Single-gene CRISPR knockout of every canonical-BAF slot in this complex is confounded by a paralogue, while the ncBAF survivors are not covered by each other in the same data - making a paralogue-pair screen the decisive missing measurement.
SWI/SNF panel lineage-split DepMap dependency scan, 28 genes · Oncofusion knowledge graph, analyses/swi-snf-lineage-split/run.py · DepMap Chronos gene effect via Open Targets GraphQL 2026What this evidence cannot show
The confound is an inference from complex composition plus the observed pattern, not a demonstrated effect; no public combinatorial screen over these pairs exists to check it against.
Evidence that argues against this (1)
- Paralogues are not always functionally redundant - SMARCD1 and SMARCD2 behave oppositely in this very dataset - so a pair knockout could be uninformative if only one paralogue is expressed. Mashtalir N et al. 2018
- E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper
A bromodomain-swap allele of BRD9 that retains function despite a radically altered pocket confers resistance to a BRD9 inhibitor series, establishing the compound's on-target activity - a ready-made genetic control for any BRD9 arm.
Sensitivity and engineered resistance of myeloid leukemia cells to BRD9 inhibition. · Hohmann AF, Martin LJ, Minder JL, Roe JS, et al. · Nat Chem Biol 2016What this evidence cannot show
Established in AML lines not selected for KMT2A rearrangement; the swap allele is a tool for the inhibitor series in that paper and would need re-validation against a degrader.
Evidence that argues against this (1)
- A bromodomain-swap allele controls for bromodomain engagement, not for degradation of the whole protein, so it does not transfer unmodified to a degrader arm. Brien GL et al. 2018
- E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
A MYC cDNA that complements RNAi against Jmjd1c in the same system failed to rescue Smarca4 knockdown, which is why a validated-rescue control is mandatory in any experiment re-testing the MYC node.
SWI/SNF Subunits SMARCA4, SMARCD2 and DPF2 Collaborate in MLL-Rearranged Leukaemia Maintenance. · Cruickshank VA, Sroczynska P, Sankar A, Miyagi S, et al. · PLoS One 2015What this evidence cannot show
The failed rescue is a negative result and inherits the usual limits - the over-expressed MYC level may be wrong - but the internal positive control is what makes it worth designing around.
Evidence that argues against this (1)
- The opposite result, MYC rescuing Brg1 knockdown, was obtained in a line carrying NrasG12D; the design should therefore include a RAS-mutant and a RAS-wild-type KMT2A-rearranged line or it will not resolve the contradiction. Shi J et al. 2013
- E4strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
Menin inhibition is the relevant second condition because the menin-inhibitor-resistant state carries SMARCA4 and BRD4 as CRISPR co-dependencies, so ncBAF occupancy under menin inhibition is directly decision-relevant.
Overcoming menin inhibitor resistance in AML cells with combinations including BET proteins and a dual BRG1/BRM inhibitor. · Fiskus W, Mill CP, Issa GC, Piel J, et al. · Blood 2026What this evidence cannot show
The resistant state was derived in two lines; occupancy under acute menin inhibition in a sensitive line is a different question from the steady state of a derived resistant line, and the experiment should do the former.
Evidence that argues against this (1)
- A CRISPR co-dependency does not predict chromatin occupancy; the experiment could find no occupancy change and the resistance result would still stand. Fiskus W et al. 2024
the three-arm experiment designed to decide it
the arm-1 MYC rescue with a RAS-mutant and RAS-wild-type pair also resolves the MYC contradiction
arm 3 maps complex-resolved occupancy against the fusion's own targets
arm 2 is the SMARCA4+SMARCA2 double knockout this beat says is missing