Hypothesis: a single dormant leukemia stem cell state confers cross-resistance to menin inhibition, venetoclax and T-cell engagement at once — the common-cause failure that orthogonal combination therapy assumes away
Adding drugs that work in completely different ways is supposed to close every escape route. Engineers who build backup systems know they usually fail together for one shared reason. There may be one state these cells can enter that defeats all three approaches simultaneously.
The same thing, for a scientist
The graph's `hyp-menin-plus-engager-orthogonal-pre-emption` argues that mechanistically orthogonal partners extend durability, and lists as its own falsifier the discovery of a single mechanism conferring cross-resistance to menin inhibition, venetoclax and CD19-directed therapy simultaneously. Reliability engineering treats common-cause failure not as a residual risk but as the dominant failure mode of redundant systems, which inverts the burden of proof: the orthogonality strategy should be assumed vulnerable until a search for the common cause has been made and has failed. We nominate a specific candidate — a quiescent, primitive, stem-like state — and a specific reason to expect it to be cross-cutting: quiescence reduces dependence on active transcriptional programs (menin), shifts apoptotic priming away from BCL-2 (venetoclax), and lowers surface antigen density and immune visibility (T-cell engagement). BET-inhibitor resistance in MLL-AF9 emerging from leukaemia stem cells, cross-resistant to genetic depletion and not efflux-mediated [E1], is the closest published precedent. This is a deliberately adversarial hypothesis aimed at an existing entry in this graph.
Why this is written as an attack on another hypothesis
A knowledge graph accumulates hypotheses that support each other and rarely accumulates hypotheses designed to kill each other. The orthogonality argument is currently the graph's main strategic recommendation for combination design, it is plausible, and it has been stated with a prior of 0.55. It deserves a named adversary with its own experiment rather than a falsifier listed in its own text and never pursued.
The adversarial framing also changes what gets measured. Testing orthogonality means running combinations and measuring durability. Testing common cause means deliberately applying all three pressures in sequence and looking for survivors that share a state — a different experiment, with a different and more informative negative result.
Why quiescence is the specific candidate
The three mechanisms fail for three different reasons in the same state:
- Menin inhibition works by collapsing an actively maintained transcriptional program and by forcing differentiation. A cell not currently cycling and not currently executing that program has less to lose.
- Venetoclax works on cells primed for apoptosis at BCL-2. Quiescent and primitive cells characteristically shift toward MCL-1 and BCL-xL dependence.
- T-cell engagement requires surface antigen density and an engageable immune synapse. Primitive states carry lower CD19 and are the compartment from which lineage switch originates, and post-immunotherapy lineage switch is concentrated in KMT2A-rearranged disease with more than half of cases switching from documented MRD-negative remission 3.
That is three independent-sounding failures with one shared cause, which is the textbook shape of a beta-factor problem. Notably, the field's own data hint at it: after menin-inhibitor failure, all complete remissions occurred with venetoclax-containing regimens or menin-inhibitor switching, and yet 77% of that cohort had already been venetoclax-exposed 2 — a population in which two of the three layers had already been breached.
What would make this false, and why that matters as much
If sequential triple pressure produces survivor populations that are transcriptionally distinct from each other — menin-escapers looking one way, venetoclax-escapers another, engager-escapers a third — then there is no common cause among these three agents and the orthogonality strategy is vindicated on its own terms. That is a genuinely valuable result and would raise confidence in a combination programme that is currently justified by inference alone.
The experiment is therefore symmetric: it either identifies the shared vulnerability that caps the strategy, or removes the main theoretical objection to it. Both outcomes are worth the cost, which is unusual and is the main argument for running it.
Counter-evidence found 2026-09-18: the strong form is contradicted
This entry was written before two menin-inhibitor combination trials reported. Both cut against it directly.
- SAVE (revumenib + decitabine/cedazuridine + venetoclax, relapsed/refractory AML, n=42): composite CR 71%, CR/CRh 60%, MRD-negative in 80% of responders, median duration of CR/CRh 10.5 months — in patients with a median of two prior lines of whom 52% had already seen venetoclax.
- Frontline aza + venetoclax + revumenib (n=43, newly diagnosed, older adults): ORR 88.4%, composite CR 81.4% (88.9% in KMT2A-rearranged), no residual disease by centralised flow in all 37 patients assessed.
A single quiescent state conferring simultaneous cross-resistance to menin inhibition, BCL-2 inhibition and hypomethylating agents cannot easily coexist with those numbers, least of all with universal flow-MRD negativity in the frontline cohort and with high response in venetoclax-pretreated patients.
What survives. Three weaker claims remain live and are worth keeping: (i) durability benefit may still be sub-multiplicative relative to formal independence, which neither trial was designed to test; (ii) 13% of SAVE patients developed emergent menin-binding-site mutations, so a shared escape route does operate, it is just a genetic one rather than a stem-cell state; (iii) the flow-MRD assays used cannot see a reservoir below their detection floor, so the convergence prediction — that survivors under sequential triple pressure occupy one transcriptional state — is untested rather than refuted. That prediction, and only that prediction, is what exp-triple-pressure-common-cause-screen should now be scoped to.
confidence_overall lowered from speculative to weak on 2026-09-18 for this reason.
| statement | A single quiescent, primitive leukemia stem cell state in KMT2A-rearranged leukemia confers simultaneous cross-resistance to menin inhibition, BCL-2 inhibition and CD19-directed T-cell engagement; consequently, combinations of mechanistically orthogonal agents will show markedly less durability benefit than their independence would predict, and survivor populations under sequential triple pressure will converge on one transcriptional state rather than diverging into three. |
|---|---|
| origin | analogy |
| source analogy | analogy-engineering-fault-tree-minimal-cut-set |
| prior probability | 0.3 |
| prior reasoning | Raised above baseline by the BET precedent, in which a stem-cell state rather than a lesion was the unit of resistance, and by the shared mechanistic logic linking quiescence to all three failures. Held down by direct counter-evidence: PRC1.1-deficient menin-resistant cells are specifically venetoclax-sensitive, and in the post-failure clinical cohort venetoclax-containing regimens produced the only complete remissions despite heavy prior venetoclax exposure. On balance 0.3, with the experiment valuable in both directions. |
| adversarial to | hyp-menin-plus-engager-orthogonal-pre-emption. This hypothesis exists to pursue that hypothesis's own listed falsifier rather than leaving it unexamined. |
| decision value | Either identifies the shared vulnerability that caps combination strategy and nominates a fourth agent to target it, or removes the principal theoretical objection to a combination programme that is currently justified by inference. |
predictions
[ "Cells surviving sequential exposure to a menin inhibitor, venetoclax and a CD19 engager converge on a single transcriptional state rather than three distinct ones, by single-cell profiling.", "That state is quiescent, primitive, and shows shifted apoptotic priming away from BCL-2 by BH3 profiling.", "Pre-sorting a KMT2A::AFF1 population for the candidate state produces cells resistant to all three agents before any exposure.", "Adding an agent that targets quiescence or forces cycling restores sensitivity to all three, which would be the practical payoff.", "In clinical cohorts, combination regimens show smaller durability gains over monotherapy than independence of resistance mechanisms would predict." ]
experiments
[ "exp-triple-pressure-common-cause-screen" ]
analyses
[ "ana-menin-resistance-cell-state-signature", "ana-resistance-route-cooccurrence" ]
falsifiers
[ "Survivor populations under the three pressures are transcriptionally distinct from each other, showing no common cause among these agents.", "Cells sorted for the candidate quiescent state are sensitive to at least one of the three agents.", "PRC1.1-deficient-like venetoclax sensitivity generalises, so that menin-resistant states are reliably BCL-2-dependent rather than BCL-2-independent.", "Clinical combination data show durability gains consistent with or exceeding what independence predicts." ]
- If a common cause exists, every combination trial now running is buying triple toxicity for less than triple benefit, in children.
- The countermeasure — forcing quiescent cells to cycle, or targeting MCL-1 — hits normal haematopoietic stem cells by the same mechanism, so the therapeutic index of any fix may be worse than the problem.
- A negative result would be reassuring, which makes it the more dangerous outcome to report carelessly: absence of a common cause among three specific agents is not absence of a common cause in general, and should not be written up as vindication of orthogonality as a principle.
- Is there a surface or metabolic marker that prospectively isolates the candidate quiescent state in KMT2A::AFF1, or does it have to be defined retrospectively from survivors?
- Does the state pre-exist at diagnosis or is it induced by the first pressure applied, and does the order of the three pressures change the answer?
- If a common cause exists, is targeting it (forcing cycling, or targeting MCL-1) safe in an infant whose normal stem cell compartment shares the same state?
- E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
BET-inhibitor resistance in MLL-AF9-immortalised murine haematopoietic progenitors emerged from leukaemia stem cells, was cross-resistant to JQ1 and to genetic BET knockdown, and was not efflux- or metabolism-mediated — establishing that a stem-cell state, not a target lesion, can be the unit of resistance in this disease.
BET inhibitor resistance emerges from leukaemia stem cells · Fong CY, Gilan O, Lam EYN, et al. · Nature 2015What this evidence cannot show
Murine, a single drug class, and cross-resistance was demonstrated within that class rather than across mechanistically unrelated classes — which is precisely the extension this hypothesis proposes and does not yet have evidence for.
Evidence that argues against this (2)
- REFUTES THE STRONG FORM. The hypothesis predicts that a single quiescent stem-cell state confers simultaneous cross-resistance to menin inhibition, BCL-2 inhibition and hypomethylating pressure, so that a mechanistically orthogonal combination should deliver markedly less durability than independence predicts. In the SAVE phase I-II study the all-oral triplet of revumenib plus decitabine/cedazuridine plus venetoclax produced a composite complete remission rate of 71% and CR/CR with partial haematologic recovery of 60% in 42 patients with relapsed or refractory AML who had a median of two prior lines, with flow MRD negativity in 80% of responders and median duration of CR/CRh of 10.5 months. Critically, 52% of these patients had prior venetoclax exposure, which is the condition under which a shared cross-resistant state should have been pre-selected. A common-cause state of the kind proposed should not permit those numbers. The hypothesis is not dead in its weak form (durability benefit may still be sub-multiplicative, and 13% developed emergent menin-binding-site mutations), but the claim that orthogonal combination is defeated is contradicted by data. Issa GC et al. 2026
- Second, independent contradiction in the frontline setting. Azacitidine plus venetoclax plus revumenib in 43 newly diagnosed older adults with NPM1-mutant or KMT2A-rearranged AML gave an intention-to-treat overall response of 88.4%, composite complete remission of 81.4% (88.9% in the KMT2A-rearranged subset), no patient refractory after one to two cycles, and no measurable residual disease by centralised flow in all 37 patients evaluated. Universal MRD negativity is the specific observation a shared quiescent cross-resistant reservoir should make impossible, because that reservoir is by construction the compartment that survives all three pressures. Caveat: this is a phase I cohort of older adults, follow-up for durability is short, and MRD negativity by flow does not exclude a reservoir below assay sensitivity — which is exactly the measurement the hypothesis would need to be re-tested against. Zeidner JF et al. 2025
- E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
In 84 adults with relapsed or refractory AML after menin-inhibitor failure, 86% had been intensively pretreated and 77% were venetoclax-exposed; overall response to next therapy was 32%, and all complete remissions occurred with venetoclax-containing regimens or with menin-inhibitor switching, with median overall survival 4.4 months.
Outcomes of relapsed or refractory acute myeloid leukemia after menin inhibition failure · Multicentre retrospective consortium · Blood Advances 2026What this evidence cannot show
Retrospective and adult. That venetoclax-containing regimens still produced the only complete remissions in a 77% venetoclax-exposed cohort argues, if anything, against a simple shared-resistance state for those two agents — and should be weighed as counter-evidence.
Evidence that argues against this (1)
- The cited post-failure cohort is read here as evidence for shared cross-resistance, but its own structure argues the other way: all complete remissions in that cohort occurred with venetoclax-containing regimens or with menin-inhibitor switching, i.e. the orthogonal agents were the only things that still worked. SAVE extends this — of 42 relapsed/refractory patients, 52% venetoclax-exposed, CRc was 71% — so prior exposure to one arm of the combination does not install resistance to the combination. A common-cause account has to explain why the remaining responses concentrate in precisely the agents it predicts should be cross-blocked. Issa GC et al. 2026
- E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
Post-immunotherapy lineage switch is concentrated in KMT2A-rearranged disease (64.3% of 70 cases), occurs at a median of 1.5 months with 81.4% inside six months, and 54.3% of affected patients had achieved an MRD-negative complete remission of their B-ALL beforehand.
Project EVOLVE: an international analysis of postimmunotherapy lineage switch, an emergent form of resistance · Project EVOLVE consortium · Blood 2025What this evidence cannot show
Establishes that immune escape in this disease proceeds through a state change rather than antigen mutation; it does not show that the same state resists menin inhibition or venetoclax.
Evidence that argues against this (1)
- Lineage switch is used here as a third face of one common-cause state. Whole-genome phylogenetic reconstruction of a lineage switch after anti-CD19 cellular and antibody therapy instead showed the AML descended from the initial ALL through a genetically distinct subclone carrying its own driver mutations that were detectable before the switch emerged — extensive prior diversification followed by selection. That is a private, pre-existing genetic lineage, not a shared quiescent state that all three therapeutic pressures happen to spare. If lineage switch is clonal selection of a distinct subclone, it belongs to a different mechanistic category from menin-inhibitor and venetoclax escape and should not be pooled with them as one common cause. Caveat: n=1, and the case was KMT2A wild-type, which is not the population this hypothesis is about. Coorens THH et al. 2023
- E4strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
Loss of PRC1.1 produces menin-inhibitor-resistant cells with reduced monocytic signature that are susceptible to BCL-2 inhibition and are resensitised to menin inhibition by venetoclax, showing that at least one characterised menin-resistant state is specifically venetoclax-vulnerable.
Epigenetic regulation of noncanonical menin targets modulates menin inhibitor response in acute myeloid leukemia · Zhou X, Zhang L, Aryal S, et al. · Blood 2024What this evidence cannot show
This is direct counter-evidence to the hypothesis for one specific resistant state, and it is the main reason the prior here is moderate rather than high. A common cause would have to be a different state from the PRC1.1-deficient one.
Evidence that argues against this (1)
- Searched specifically for published evidence that PRC1.1-deficient, menin-inhibitor-resistant cells are also venetoclax-cross-resistant, which is what the common-cause hypothesis requires of them. No such report was found. The available evidence runs the opposite way: the PRC1.1-loss state is venetoclax-vulnerable in models, and clinically the menin plus venetoclax plus hypomethylating triplet achieved composite CR of 81.4% frontline with universal flow-MRD negativity. Recorded here so the next reader knows this gap was searched rather than skipped. Zeidner JF et al. 2025
SAVE triplet (revumenib+decitabine/cedazuridine+venetoclax) gave CRc 71% in R/R AML with 52% prior venetoclax exposure; frontline aza+ven+revumenib gave CRc 81.4% with universal flow-MRD negativity — strong form of the common-cause claim contradicted
orthogonal menin + BCL-2 + hypomethylating combination is highly active in exactly the pretreated population a shared quiescent cross-resistant state would protect
- related to →MEN1 drug-interface mutations — the first proof that a chromatin drug can drive Darwinian escape in patients
what survives the counter-evidence: 13% of SAVE patients developed emergent menin-binding-site mutations, so the shared escape route that does operate is genetic rather than a stem-cell state
- proposed by →Fault trees, minimal cut sets and common-cause failure — reliability engineering on a buffered oncogenic program
generated by common-cause failure analysis and the beta-factor
pursues that hypothesis's own listed falsifier as a named adversarial hypothesis
- depends on →The stem-like self-renewal state — a limited stem-cell module grafted onto a committed cell
nominates the quiescent primitive state as the candidate common cause
- related to →MYC reactivation through PRC1.1 loss — a named, druggable bypass out of menin dependence
PRC1.1-deficient cells being venetoclax-sensitive is the main counter-evidence
- related to →Lineage switch as antigen escape from CD19-directed therapy in KMT2A-rearranged leukemia
immune escape through a state transition is the third limb of the proposed common cause
convergence test across three independently derived resistant populations