Hypothesis: CLEC12A is a cell-intrinsic dependency in KMT2A-rearranged leukaemia, not only a surface marker - and if so, antigen-negative escape is self-defeating
also called CLEC12A dependency hypothesis · marker-versus-dependency hypothesis
In one other kind of leukaemia, deleting CLEC12A killed the cells, so CLEC12A may be something the cancer needs rather than just a label on its surface. If the same is true in KMT2A leukaemia, then the way a CLEC12A therapy usually fails - the cancer drops the target - would instead make the cancer weaker.
The same thing, for a scientist
**Statement.** CLEC12A is required for the growth or survival of KMT2A-rearranged leukaemic cells, as it is for NUP98::NSD1 AML [E1], rather than being an inert surface antigen. **Why it is worth stating now.** The practical consequence is unusually sharp. KITE-222 failed in phase 1 because CLL-1-negative blasts persisted after CLL-1-positive blasts were cleared [E2]. That is the canonical failure of a **marker**. If CLEC12A is a **dependency**, the CLL-1-negative cells are not merely escapees - they are cells that have paid a fitness cost, and the therapeutic logic changes from 'bind and kill' to 'disrupt the receptor's function', which a CAR does not do. **Origin: literature-gap plus analogy.** It is generated by transporting a result across fusion families, which is the PI's standing priority, and by a bounded absence: as of 2026-09-20 no published study tests CLEC12A dependence in KMT2A-rearranged leukaemia [E3]. **Prior probability: low, and stated as such.** The strongest argument against is [E2] itself - human R/R AML blasts that are CLL-1-negative survive and proliferate in patients, which is hard to square with a general myeloid-leukaemia dependency. The hypothesis is therefore most plausible in its narrow form: a fusion-family-restricted requirement.
Predictions, in the order they should be tested
- CRISPR knockout of CLEC12A in KMT2A::MLLT3 and KMT2A::AFF1 models reduces colony formation and raises apoptosis relative to a non-targeting control, by an effect size comparable to the NUP98::NSD1 result 1.
- The effect is larger in the presence of a cooperating RAS lesion than without it, if the NUP98::NSD1 phenotype is partly an NRAS-context phenotype.
- DepMap CRISPR effect for CLEC12A is more negative in KMT2A-rearranged lines than in comparator leukaemia lines - and, per this project's standing correction, the index arm must be lineage-split (myeloid and lymphoid halves reported separately) before any selectivity claim, and raw effect reported beside any z.
- CLL-1-negative blasts persisting after CLEC12A-directed therapy show reduced clonogenicity ex vivo compared with matched CLL-1-positive blasts from the same patient.
- The knockout phenotype attenuates in NET-free culture, if the requirement runs through the receptor's known ligand, extracellular NET DNA 4.
What would kill this hypothesis
See data.falsifiers. The cheapest and most decisive is prediction 3: if CLEC12A's CRISPR effect in DepMap is flat in KMT2A-rearranged lines and indistinguishable from comparators, the cell-intrinsic version of the hypothesis is dead without any bench work. This analysis has not been run - it is written up as ana-clec12a-dependency-in-kmt2a-r-depmap and is planned, not run, so nothing in this entry should be read as a DepMap result.
| statement | CLEC12A is required for the growth or survival of KMT2A-rearranged leukaemic cells, not merely displayed on them; if true, antigen-negative escape from CLEC12A-directed therapy carries a fitness cost rather than being a free escape route. |
|---|---|
| origin | analogy |
| prior probability | 0.15 |
predictions
[ "CRISPR knockout of CLEC12A in KMT2A::MLLT3 and KMT2A::AFF1 models reduces colony formation and increases apoptosis", "the phenotype is larger with a cooperating RAS lesion than without", "DepMap CLEC12A CRISPR effect is more negative in KMT2A-rearranged lines than comparators, in BOTH the myeloid and the lymphoid half of the index arm", "CLL-1-negative blasts persisting after CLEC12A-directed therapy are less clonogenic ex vivo than matched CLL-1-positive blasts", "the knockout phenotype attenuates in NET-free culture" ]
experiments
[ "exp-clec12a-knockout-in-kmt2a-r-models" ]
analyses
[ "ana-clec12a-dependency-in-kmt2a-r-depmap" ]
falsifiers
[ "CRISPR knockout of CLEC12A in two independent KMT2A-rearranged models produces no reduction in colony formation, apoptosis or engraftment relative to non-targeting controls - the hypothesis is dead.", "DepMap CLEC12A CRISPR effect is flat (near zero) in KMT2A-rearranged lines and indistinguishable from comparator leukaemia lines in BOTH lineage halves - the cell-intrinsic version is dead without bench work.", "CLL-1-negative blasts from post-CAR patients show clonogenicity equal to or greater than matched CLL-1-positive blasts - the fitness-cost prediction, which is the whole clinical point, is dead.", "The NUP98::NSD1 phenotype fails to reproduce in a NUP98::NSD1-only background without NRAS-G12D - the founding observation was a RAS-context phenotype and the analogy has no source." ]
- **If the hypothesis is TRUE, the current clinical strategy is wrong.** A dependency calls for functional disruption, not for cytotoxic targeting that enriches antigen-negative cells. Confirming this hypothesis would be an argument against, not for, the CAR programmes.
- **If the hypothesis is FALSE, nothing is lost but the DepMap query is still worth running**, because a flat CLEC12A dependency profile is itself a useful fact for anyone reading the NUP98::NSD1 paper as a general result.
- **The analogy is doing more work than the data.** One murine model, one fusion family, one cooperating lesion. This graph has been burned before by transporting a mechanism across fusion families on the strength of a single paper.
- Does DepMap contain enough KMT2A-rearranged lines, split by lineage, to test prediction 3 with any power? This graph's own dependency work puts the KMT2A-r arm at roughly 7 myeloid to 4 lymphoid, which is thin for a lineage-split comparison.
- Is there a CLEC12A small molecule or degrader at all? If the dependency version is right, the modality that exploits it does not currently exist and the ChEMBL activity-row count for this target should be checked before anyone plans around it.
- Could CLEC12A dependency and CLEC12A-directed immunotherapy be actively antagonistic - the CAR selecting for the very cells the dependency says are unfit, and thereby producing a slower-growing but untargetable disease?
- E1moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper
CRISPR/Cas9 depletion of Clec12a in NUP98::NSD1+NRAS-G12D-immortalised murine cells increased apoptosis and reduced colony numbers in vitro, and Clec12a deletion significantly reduced leukaemic engraftment and prolonged survival in vivo.
Clec12a is required for the pathogenesis of NUP98::NSD1 AML. · Mohanty S, Charles Cano F, Gabdoulline R, Lai CK, Othman B, Sudarsanam H, et al. · Blood Adv 2025What this evidence cannot show
Abstract read in full from raw NCBI efetch XML on 2026-09-20; full text not read. Murine Clec12a; the model carries a cooperating NRAS-G12D with no reported knockout arm lacking it; effect sizes not available at abstract level. This is the sole empirical foundation of the hypothesis and it is one paper in one fusion family.
Evidence that argues against this (1)
- A single-model CRISPR phenotype in an immortalised line is the weakest form of dependency evidence, and the authors themselves write only that the data 'suggest to further explore CLEC12A as a potential target'. Nothing in the abstract rules out an off-target or clonal-selection artefact. Xiu Y et al. 2018
- E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
In the KITE-222 phase 1 study, CLL-1-negative blasts persisted in patients whose CLL-1-positive blasts had been nearly completely depleted by CLL-1-directed CAR T cells, and total blast counts did not fall.
Phase 1 Study of KITE-222, an Autologous CLL-1-Directed CAR T-cell Therapy in Patients with Relapsed/Refractory Acute Myeloid Leukemia. · Daver N, Blachly JS, Ghobadi A, Advani A, Muffly L, Garciaz S, et al. · Clin Cancer Res 2026What this evidence cannot show
Abstract read in full from raw NCBI efetch XML on 2026-09-20; full text not read. n=12, adult R/R AML, no KMT2A stratification reported. This is cited here as the principal evidence AGAINST the hypothesis, which is why it is in the evidence array rather than only in counter-evidence.
Evidence that argues against this (1)
- The persistence was observed in heavily pre-treated relapsed/refractory adult AML, where clonal selection has already occurred; a dependency present at diagnosis can be lost in a clone that has evolved through multiple therapies. So this is strong evidence against a universal dependency and weak evidence against a fusion-family-restricted one at diagnosis. Xiu Y et al. 2018
- E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.dataset
As of 2026-09-20, PubMed returns 0 records for `CLEC12A AND (KMT2A OR MLL) AND (knockout OR knockdown OR CRISPR OR dependency)` while returning 156 for `CLEC12A` alone and 9,229 for `KMT2A OR MLL` alone, and the alias-broadened query returns 1 record that title-checks as unrelated.
Q48 surface-antigen beat: bounded counted-absence sweep, PubMed E-utilities and Europe PMC REST · Q48-surface-antigens · Oncofusion knowledge graph (internal measurement, 2026-09-20) 2026What this evidence cannot show
Bounded counted absence, date 2026-09-20, tool NCBI PubMed E-utilities esearch with Europe PMC cross-check; full query strings and title-checking are recorded in `mechanism-clec12a-dependence-nup98-nsd1`. Rule 67 positive controls passed through the identical code path. Residual: 72 Europe PMC full-text co-mentions were title-checked and not read.
Evidence that argues against this (1)
- An absence of testing is not a negative result, and it is also not support for the hypothesis. It establishes only that the hypothesis is open, and a hypothesis being open is the weakest possible reason to believe it. Q48-surface-antigens 2026
- E4established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper
MICL (CLEC12A) directly recognises DNA in neutrophil extracellular traps and this interaction regulates neutrophil activation, with loss or inhibition of MICL function leading to uncontrolled NET formation through the ROS-PAD4 pathway.
Recognition and control of neutrophil extracellular trap formation by MICL. · Malamud M, Whitehead L, McIntosh A, Colella F, Roelofs AJ, Kusakabe T, et al. · Nature 2024What this evidence cannot show
Abstract read in full from raw NCBI efetch XML on 2026-09-20; full text not read. Neutrophil biology, not leukaemia; the inference that a leukaemic dependency might run through the same ligand is entirely mine and is the least supported step in this entry.
Evidence that argues against this (1)
- If the receptor's only known ligand is extracellular NET DNA, a cell-intrinsic requirement in a colony-formation assay is hard to explain, because methylcellulose colony assays are not obviously NET-rich. That argues the NUP98::NSD1 phenotype runs through something other than canonical MICL ligand binding - or that it is not a signalling phenotype at all. Xiu Y et al. 2018
the hypothesis is about this gene's role changing from marker to dependency
transported across fusion families from NUP98::NSD1 to KMT2A-r
four-arm CRISPR knockout with an in-house NUP98::NSD1 positive control
cheapest available falsification, planned not run