Does a menin inhibitor create a BMI1 dependency? A two-arm knockout-under-drug experiment in KMT2A::AFF1 cells, with the CDKN2A readout that decides the mechanism
also called BMI1 x revumenib synthetic lethality test · PCGF4/PCGF1 knockout under menin inhibition
Knock out BMI1 in KMT2A-fusion leukemia cells and nothing happens. The test is whether that changes once a menin inhibitor has switched off HOXA9, which is thought to be doing BMI1's job. Measure the brake gene directly, so that a positive result comes with its explanation rather than without one.
The same thing, for a scientist
A 2x2 genetic-by-pharmacological experiment in human KMT2A::AFF1 B-lineage lines (SEM, RS4;11): BMI1 knockout or non-targeting control, crossed with a menin inhibitor or vehicle. Primary readout is relative fitness; the mechanistic readout is CDKN2A/p16INK4a transcript and protein, measured in every arm, so that a phenotype arrives with or without its proposed cause rather than being inferred. Two arms are added because the hypothesis has two obvious ways of being right in spirit and wrong in detail: **PCGF1 alongside BMI1**, because this project treats PRC1.1 rather than canonical PRC1 as the central polycomb arm here, and an **isogenic CDKN2A-null derivative**, because the mechanism predicts the effect should vanish there. A serial-replating arm is required, not optional: BMI1's published phenotype is self-renewal, and live DepMap shows a 21-day proliferation assay gives BMI1 a gene effect of -0.0103 across 57 leukaemia lines, so a growth curve alone has a near-zero prior of showing anything in either arm [E1].
Design
Cells. SEM and RS4;11, both human t(4;11) KMT2A::AFF1 B-lineage lines. A KMT2A-germline B-ALL line (e.g. REH) as the specificity control, because the hypothesis is specifically that the fusion's HOXA output is the buffer.
Genetic arm. CRISPR knockout of BMI1, of PCGF1, and both; non-targeting sgRNA control. Cas9 cutting controls at a neutral locus to separate cutting toxicity from gene loss, which matters because the expected baseline effect is near zero and a small artefact would be the whole signal.
Pharmacological arm. A menin inhibitor at a concentration that collapses the HOXA/MEIS1 programme without killing the vehicle-arm cells over the assay window - dose set in a pilot by MEIS1 qPCR, not by viability, so the two arms are matched on target engagement rather than on cell number.
Readouts, in the order they decide things.
- CDKN2A/p16INK4a transcript and protein, all four arms. This is the first readout and it can kill the hypothesis before the fitness data exist: if neither BMI1 loss nor menin inhibition derepresses the locus at baseline, the axis is not running in these cells.
- Relative fitness / competitive growth, 14 and 21 days.
- Serial replating, three rounds, because self-renewal is the published phenotype.
- Senescence markers (SA-beta-gal, cell size), because derepressing CDKN2A is a senescence programme and a fitness drop without death is a materially different result from killing.
Isogenic control. A CDKN2A-null derivative of SEM. The mechanism predicts the interaction disappears; if it survives, the phenotype is real and the explanation is wrong.
What each outcome means, decided in advance
| result | reading |
|---|---|
| BMI1 KO neutral in both arms, in growth AND replating | hypothesis dead as stated; check PCGF1 arm before abandoning the polycomb direction |
| BMI1 KO neutral alone, costly under menin inhibitor, CDKN2A derepressed synergistically, effect lost in CDKN2A-null | hypothesis confirmed with mechanism |
| BMI1 KO costly under menin inhibitor but CDKN2A unmoved, or effect survives in CDKN2A-null | real interaction, wrong mechanism - consistent with the published finding that Ink4a/Arf deletion only partially rescues Bmi1-null LSCs |
| PCGF1 KO shows the effect and BMI1 KO does not | the hypothesis is right about polycomb and wrong about the subunit; this is the most likely partial outcome and it collides with resistance-prc1-1-loss-menin-inhibitor-nup98, which records PRC1.1 LOSS as a menin-inhibitor resistance route - a direction conflict that would have to be resolved before any therapeutic reading |
| Effect equally large in the KMT2A-germline control | not fusion-specific; the interaction is about polycomb and menin generally and the whole framing changes |
Writing the readings down before the experiment is the point. This project has published entries on a screen difference and then failed its own null in the next cycle; the fix is to state the interpretation in advance and compute the null first.
Cost, and why this one is cheap
All reagents exist: SEM and RS4;11 are standard, CRISPR knockout of two genes is routine, a menin inhibitor is commercially available as a tool compound, and every readout is a qPCR, a western, a growth curve or a plate count. There is no animal arm in the first pass, deliberately - the published BMI1 requirement is an in vivo requirement, and the honest consequence is that a negative in vitro result does not close the question. It does, however, decide whether an animal experiment is worth designing.
Estimated cost tier: low. Estimated timeline: eight to twelve weeks including the CDKN2A-null derivative, which is the long pole.
Do readout 1 first and alone. A p16INK4a western across a menin-inhibitor time-course in untransduced SEM cells is about one week and tests the hypothesis's premise before anything is spent on knockouts.
| design | 2x2 genetic-by-pharmacological: CRISPR knockout of BMI1 (PCGF4), of PCGF1, or both, versus non-targeting sgRNA; crossed with a menin inhibitor versus vehicle. Cell lines SEM and RS4;11 (human t(4;11) KMT2A::AFF1, B-lineage), with a KMT2A-germline B-ALL line as specificity control and an isogenic CDKN2A-null SEM derivative as the mechanism control. Menin-inhibitor concentration set in a pilot by MEIS1 qPCR (target engagement), not by viability, so the arms are matched on engagement. |
|---|---|
| readout | PRIMARY MECHANISM: CDKN2A transcript and p16INK4a protein in all arms. PRIMARY PHENOTYPE: competitive fitness at 14 and 21 days. REQUIRED THIRD: serial replating, three rounds. FOURTH: senescence markers (SA-beta-gal, cell size), because derepressing CDKN2A is a senescence programme and arrest is not death. |
| controls | non-targeting sgRNA; Cas9 cutting at a neutral locus (essential, because the expected baseline effect is near zero and a cutting artefact would be the entire signal); vehicle arm; KMT2A-germline line; isogenic CDKN2A-null derivative |
| cost tier | low |
| timeline | 8-12 weeks including the CDKN2A-null derivative; the premise test alone (p16INK4a western across a menin-inhibitor time-course in untransduced SEM) is ~1 week and should be run first |
| stage gate | If neither BMI1 loss nor menin inhibition derepresses CDKN2A in these cells, stop - the axis is not operating and no fitness arm is worth running. |
| prespecified readings | recorded in the sections of this entry, before execution, because this project has published on a screen difference and failed its own null in the next cycle |
| tests hypothesis | hyp-menin-inhibition-creates-a-bmi1-dependency |
| status | planned |
- A positive result identifies a genetic dependency, not a drug. No BMI1 inhibitor of credible clinical standing was assessed by this beat, so a confirmed synthetic lethality would still need chemical matter.
- The intended mechanism is CDKN2A derepression, which induces senescence rather than death - so the experiment could 'succeed' by producing arrested, secretory, surviving blasts, which is why the senescence readout is in the design rather than optional.
- If PCGF1 rather than BMI1 carries the effect, the therapeutic reading inverts: this graph records PRC1.1 LOSS as a menin-inhibitor resistance mechanism, so inhibiting PRC1.1 therapeutically would be pushing in the direction resistance already goes.
- A dependency created by a drug exists only while the drug is on. Any schedule with a trough restores the HOXA buffer, so a constant-exposure in vitro result overstates what an intermittent clinical schedule would deliver.
- Normal HSCs need BMI1 too. If the combination works, its therapeutic index depends on normal HSCs not experiencing the same HOXA collapse - an assumption this experiment does not test and that a normal-CD34+ arm would.
- CLOSED this cycle: the name-list omission was real and was fixed. The counted zero was re-run with SNDX-50469, SNDX50469, MI-3454, MI-463, DS-1594, JNJ-75276617 and BMF-219 added; the menin-inhibitor operand rose from 355 to 363 and the BMI1 cross stayed at 0, with positive controls of 18 and 9 through the identical shape. The absence stands.
- Is SEM the right line, given it is from a five-year-old rather than an infant? Every in vitro argument in this beat rests on non-infant t(4;11) lines because no infant-derived line with CRISPR data exists - ledger priority 16.
- Should the first pass include an in vivo arm after all? The published BMI1 phenotype is an in vivo LSC phenotype, so a negative in vitro result is weak evidence; the argument for staying in vitro is cost, not inference.
- E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database
Fetched live on 2026-09-20, DepMap Chronos gene effect for BMI1 is +0.0585 across 1,258 lines and -0.0103 across 57 leukaemia-annotated lines, with 1.43% of lines at or below -0.5 - so a standard proliferation-based knockout screen has a near-zero prior of detecting a BMI1 effect, which is why this design requires a serial-replating arm rather than a growth curve alone.
DepMap CRISPR (Chronos) gene effect for BMI1 (ENSG00000168283), retrieved via the Open Targets Platform GraphQL API · Broad Institute DepMap; Open Targets Platform; query run by this project 2026-09-20 · Database query 2026What this evidence cannot show
Untreated lines only; DepMap has no menin-inhibitor arm, which is the whole reason this experiment is needed. 21-day fitness in immortalised lines, no infant-derived line, and the leukaemia subset is a coarse text match on the screen's own diseaseFromSource field rather than curated lineage.
Evidence that argues against this (1)
- If BMI1's requirement is entirely a self-renewal requirement, then a near-zero proliferation baseline is uninformative about what the drug arm will do, and a null result in the growth readout would not be evidence against the hypothesis. Lessard J et al. 2003
- E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database
No published study combines a menin inhibitor with BMI1. A PubMed title/abstract query crossing a TWENTY-term menin-inhibitor name list with BMI1, BMI-1 and PCGF4 returns 0 records, against 363 for the menin-inhibitor list alone and 3,926 for the BMI1 list alone, with positive controls through the identical query shape of 18 (x HOXA9), 9 (x H3K4me3) and 6 (x polycomb/PRC1/PRC2), and 2 for x PCGF1/PRC1.1. The bare-operand check menin[tiab] AND BMI1[tiab] returns 1 record, PMID 34446068, which title-checks to a prostate-cancer study of MEN1 silencing and is off-topic. The same list crossed with CDK6 returns 6, and with PBX3/PBX returns 2 and with HOXA10 returns 1.
PubMed E-utilities esearch, counted absence with operand controls, positive control and title check · NCBI PubMed; queries run by this project 2026-09-20 · Literature index 2026What this evidence cannot show
Bounded by the name lists actually tried, and the menin list was EXTENDED after a first pass to close a known omission. MENIN INHIBITORS, all [tiab]: 'menin inhibitor', 'menin inhibitors', 'menin inhibition', revumenib, ziftomenib, bleximenib, enzomenib, VTP-50469, VTP50469, MI-503, MI-538, MI-2-2, SNDX-5613, SNDX-50469, SNDX50469, MI-3454, MI-463, DS-1594, JNJ-75276617, BMF-219. BMI1: BMI1, BMI-1, PCGF4 [tiab]. The zero was 0 under both the 13-term and the 20-term list, and the operand count rose only from 355 to 363, so the extension changed the denominator and not the answer. Title/abstract only - a paper that measures BMI1 in a figure without naming it in the abstract is invisible to this query, which is the standard limit of a tiab absence and is NOT closed here.
Evidence that argues against this (1)
- The same menin-inhibitor name list crossed with CDK6 returns 6 records including a study reporting that menin inhibition attenuates CDK6 and synergises with abemaciclib - so the instrument reaches combination studies perfectly well when they exist, which strengthens rather than weakens the BMI1 zero. Fiskus W et al. 2022
- E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
In murine MLL-AF9, Hoxa9 regulates the p16(Ink4a)/p19(ARF) locus as Bmi1 does and substitutes for it: MLL-AF9 readily transforms Bmi1(-/-) cells while non-Hox-activating fusions trigger senescence there, and either Bmi1 or Hoxa9 alone restores leukaemic stem cell establishment in Bmi1(-/-)Hoxa9(-/-) mice - which is the buffering relationship this experiment tests pharmacologically.
Functional crosstalk between Bmi1 and MLL/Hoxa9 axis in establishment of normal hematopoietic and leukemic stem cells · Smith LL, Yeung J, Zeisig BB, Popov N, Huijbers I, Barnes J, Wilson AJ, Taskesen E, Delwel R, Gil J, Van Lohuizen M, So CW · Cell Stem Cell 8:649-662 2011What this evidence cannot show
Abstract read this cycle. MURINE MLL-AF9 MYELOID disease with germline knockouts and a transformation/initiation readout. This experiment extrapolates twice - from germline absence to pharmacological collapse, and from murine myeloid initiation to human B-lineage maintenance - and either step could fail. That is why readout 1 tests the premise before anything else is spent.
Evidence that argues against this (1)
- A second 2011 study found Bmi1 indispensable for MLL-AF9 leukaemia in vivo despite Hoxa9 being present, which if correct means the buffer is incomplete at baseline and the experiment may find a smaller unmasked effect than the design is powered for. Yuan J et al. 2011
2x2 knockout-by-drug with a CDKN2A mechanism readout and a serial-replating arm