ZEB2 level sets how easily a KMT2A-rearranged blast can change lineage under CD19-directed pressure
also called ZEB2 and lineage switch · lineage fidelity hypothesis
When children with KMT2A-fusion leukemia are treated with therapies aimed at a B-cell marker, the leukemia sometimes escapes by becoming a different kind of blood cell entirely. ZEB2 is a protein whose known job is to stop blood cells from switching on the wrong identity, so this asks whether how much ZEB2 a blast has decides how easily it can make that escape.
The same thing, for a scientist
ZEB2 is a transcriptional repressor that restrains premature expression of mature haematopoietic lineage programs in HSPCs and, with ZEB1, enforces lineage fidelity [E1]. Lineage switch, predominantly B-to-myeloid, is the characteristic escape from CD19-directed therapy in KMT2A-rearranged leukemia, and this graph already holds that it most likely reflects competitive release of a pre-existing clone rather than de novo transdifferentiation. The hypothesis is that **ZEB2 dosage is one of the parameters that sets how permissive a blast is to that switch** — high ZEB2 holding lineage identity closed, low ZEB2 opening it. The origin of the hypothesis is a data pattern, not a paper: ZEB2 ranked 11 of 4,589 in this project's unbiased KMT2A-selectivity scan [E2], which is a coincidence with its lineage-fidelity role rather than evidence for this claim. **Prior probability is set low (0.15) and the hypothesis is speculative.**
Where this comes from, stated honestly
Two facts that this project holds independently, placed next to each other:
- ZEB2 'acts predominantly as a transcriptional repressor involved in restraining mature hematopoietic lineage gene expression programs from being expressed too early in HSPCs', with ZEB1 fine-tuning that repression 'to ensure hematopoietic lineage fidelity' 1.
- Lineage switch is the characteristic route of escape from CD19-directed therapy in KMT2A-rearranged leukemia (
risk-lineage-switch-cd19-directed-therapy,mechanism-lineage-switch-kmt2a).
And one that prompted looking: ZEB2 ranked 11 of 4,589 in an unbiased dependency scan of KMT2A-rearranged versus other leukemia lines 2.
Nothing read in this cycle connects ZEB2 to a lineage switch in any patient or model. No paper was found that measures ZEB2 across a switch, and the hypothesis is therefore an analogy between two literatures rather than an extrapolation from one. That is why the prior is 0.15 and not higher.
What it predicts, and where it would break
The prediction that matters is directional and awkward: if ZEB2 enforces lineage fidelity, then reducing ZEB2 should make switching easier, not harder. That makes ZEB2 a poor drug target under this hypothesis even though it is a real dependency — suppressing it could promote the very escape the field is trying to prevent.
That awkwardness is a feature. A hypothesis that only ever recommends what was already wanted is not doing work. This one, if true, argues against pursuing a gene that a dependency scan just nominated.
It also makes a prediction the graph can test cheaply against data it already contemplates: in paired diagnosis and post-switch samples, ZEB2 should be lower in the switched population, and the pre-existing myeloid-primed subclone identified by exp-diagnostic-subclone-archaeology-lineage-switch should be ZEB2-low at diagnosis.
The competing explanation that must be excluded first
This graph's existing position, hyp-lineage-switch-is-competitive-release-of-preexisting-clone, is that lineage switch reflects outgrowth of a clone that was already there. If that is right, then any ZEB2 difference observed after a switch may simply be a property of the clone that won, not a cause of the switching. Observing ZEB2-low cells after a switch is therefore consistent with both models and discriminates neither.
The discriminating experiment has to manipulate ZEB2 and then apply the pressure, not measure ZEB2 after the pressure. That is stated in the falsifiers.
| statement | In KMT2A-rearranged B-lineage leukemia, the level of ZEB2 in a blast sets how permissive that blast is to lineage switching under CD19-directed pressure: ZEB2-low cells switch more readily, ZEB2-high cells hold B-lineage identity. |
|---|---|
| origin | data-pattern |
| prior probability | 0.15 |
| entered graph via | followed from the scan hit gene-zeb2; the hypothesis is this Queen's, not the literature's |
predictions
[ "ZEB2 transcript and protein are lower in post-switch myeloid blasts than in the paired diagnostic B-lineage blasts from the same patient.", "At diagnosis, the myeloid-primed minor subclone detectable by subclone archaeology is ZEB2-low relative to the dominant B-lineage population.", "ZEB2 knockdown in a KMT2A::AFF1 B-lineage model increases the frequency of myeloid-marker-positive cells under blinatumomab or CD19 CAR-T pressure, relative to control.", "ZEB2 overexpression in the same model reduces that frequency.", "ZEB2 dependence in DepMap is stronger in myeloid-annotated than B-lineage-annotated KMT2A-rearranged lines, because the published functional evidence is myeloid." ]
experiments
[ "exp-diagnostic-subclone-archaeology-lineage-switch", "exp-menin-inhibitor-lineage-destination" ]
analyses
[ "ana-lineage-switch-signature-scoring", "ana-depmap-lineage-conditioned" ]
falsifiers
[ "ZEB2 is unchanged or higher in post-switch myeloid blasts compared with paired diagnostic samples. That kills the directional core of the hypothesis.", "ZEB2 knockdown in a KMT2A::AFF1 B-lineage model under CD19-directed pressure does not increase myeloid-marker-positive escape relative to control. Manipulation with no effect is the decisive negative, because the observational predictions are all confounded by competitive release.", "The myeloid-primed diagnostic subclone is ZEB2-high rather than ZEB2-low.", "ZEB2 dependence in KMT2A-rearranged lines disappears entirely under lineage conditioning, in which case there is no KMT2A-specific phenomenon to explain." ]
- If this hypothesis is right, the therapeutic implication is the opposite of the one a dependency scan invites: suppressing ZEB2 would loosen lineage identity and could accelerate the escape route that already defeats CD19-directed therapy. A selective dependency that you must not drug is a real category and this may be an instance of it.
- The hypothesis is built on an analogy between normal-progenitor lineage fidelity and leukemic immunophenotype switching. Those may be different phenomena with a shared vocabulary, and the graph has previously been misled by exactly that kind of borrowed word.
- Every observational prediction here is confounded by this graph's own competitive-release model. Only the manipulation experiments discriminate, and they are the expensive ones.
- Does any existing paired diagnosis/lineage-switch dataset already have ZEB2 expression in it? Answering this may cost nothing but a re-analysis of published data, and it should be checked before any new experiment is designed.
- If ZEB2 loss promotes switching, does menin inhibition change ZEB2? A menin inhibitor that lowered ZEB2 would be creating the permissive state at the same time as it applies differentiation pressure, which would be an unwanted interaction worth knowing about before combination trials.
- Is there a ZEB1 compensation? ZEB1 fine-tunes the same repression, so a ZEB2-only manipulation may be buffered, and the buffering is exactly the failure mode `hyp-paralogue-buffering-hides-dependencies` predicts.
- E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
ZEB2 acts predominantly as a transcriptional repressor restraining mature haematopoietic lineage gene expression programs from being expressed too early in HSPCs, with ZEB1 fine-tuning that repression to ensure haematopoietic lineage fidelity; combined inducible inactivation causes acute bone-marrow failure with decreased differentiation into all lineages, and Zeb2 inactivation enhances survival in secondary MLL-AF9 AML models.
Interplay between the EMT transcription factors ZEB1 and ZEB2 regulates hematopoietic stem and progenitor cell differentiation and hematopoietic lineage fidelity · Wang J, Farkas C, Benyoucef A, Carmichael C, Haigh K, Wong N, Huylebroeck D, Stemmler MP, Brabletz S, Brabletz T, Nefzger CM, Goossens S, Berx G, Polo JM, Haigh JJ · PLoS Biology 19:e3001394 2021What this evidence cannot show
Read as abstract only. Mouse genetics in normal haematopoiesis and in MLL-AF9 AML. 'Lineage fidelity' here means orderly lineage commitment in normal HSPCs, which is not the same phenomenon as a leukemic blast changing immunophenotype under therapeutic pressure. The step from one to the other is the entire content of this hypothesis and is not supported by this paper.
Evidence that argues against this (1)
- If lineage switch is competitive release of a pre-existing clone, as this graph's own hypothesis holds, then lineage fidelity in a normal progenitor may have nothing to do with it and the whole analogy is misplaced. Li H et al. 2017
- E2weak confidence 2 of 5: Preliminary, small, indirect, or not yet peer reviewed.dataset
ZEB2 ranks 11 of 4,589 protein-coding genes by KMT2A-rearranged selectivity in this project's unbiased DepMap scan (difference -0.464) and is not pan-essential (-0.358), which is what prompted the search that produced this hypothesis.
DepMap CRISPR (Chronos) gene effect via the Open Targets Platform GraphQL API; analyses/depmap-kmt2a-selective-scan/ · Broad Institute DepMap; Cellosaurus (SIB); Open Targets Platform; analysis run by this project · Databases and derived analysis 2026What this evidence cannot show
Our own unreviewed computation on 11 lines, none infant-derived, covering 24% of the protein-coding genome. A dependency rank says nothing whatever about lineage switching; it is recorded here only as the honest origin of the idea, and it is deliberately marked weak so that nobody later mistakes it for support.
Evidence that argues against this (1)
- The published ZEB2 dependency is an AML-wide one that also holds in t(8;21) and t(15;17), so even the rank's KMT2A-specific framing is doubtful. De Conti G et al. 2020
the hypothesis rests on ZEB2's documented lineage-fidelity role
- related to →Lineage switch in KMT2A-rearranged leukemia — the same clone changes what kind of blood cell it is
proposes a molecular parameter that would set permissiveness to the switch
- ← testsExperiment: go back to the diagnostic sample and look for the myeloid clone before it existed
paired diagnosis/switch samples would test whether the myeloid-primed subclone is ZEB2-low
the competing explanation: if the switch is outgrowth of a pre-existing clone, a ZEB2 difference is a property of the winner, not a cause