CBFA2T3::GLIS2 supplies its own CAR-T antigen - FOLR1 came out of a list of fusion-specific genes in an isogenic human model. KMT2A-rearranged leukaemia is treated with a LINEAGE antigen it can abandon, or with one a drug has to create; nobody has built the fusion-specific surface list for KMT2A::AFF1.
also called FOLR1 CAR-T · fusion-driven antigen · the escape-proof antigen · C/G surface target discovery
In one infant leukaemia, researchers built a human model of the fusion, compared it with normal cells, and pulled out a list of surface proteins that only the fusion's cells make - then built a CAR-T cell against one of them. The leukaemia this project studies is treated with markers that belong to the cell type rather than to the fusion, which is why the disease can escape by changing cell type, and nobody has made the equivalent fusion-specific list for it.
The same thing, for a scientist
Le et al. expressed CBFA2T3::GLIS2 in human cord-blood HSPCs on an endothelial-cell coculture, recapitulated the transcriptome, morphology and immunophenotype of the human disease and produced aggressive leukaemia in xenografts, then **interrogated the transcriptome of the model cells and of primary C/G AML to identify a library of C/G-fusion-specific genes as potential therapeutic targets**, developed CAR T cells against one of them - **folate receptor alpha, FOLR1** - and showed preclinical efficacy in vitro and in xenografts [E1]. **The important part is not FOLR1. It is the pipeline**: an isogenic human model of the fusion, a differential transcriptome against the matched normal progenitor, and a surface-restricted shortlist. **Why that matters here.** The immunotherapy that works in KMT2A-rearranged B-ALL targets **CD19, a lineage antigen** - and lineage antigens are abandonable, which is why lineage switch to a myeloid phenotype under CD19-directed pressure is one of this disease's defining resistance routes. The alternative this graph records, CLEC12A, is a myeloid antigen whose expression in KMT2A-rearranged AML is reported to be **induced by menin inhibition** (`mechanism-menin-inhibition-induces-clec12a`), which makes it a **drug-created** target: doubly contingent, on continued drug exposure and on the fusion remaining menin-inhibitable, in a disease where *MEN1* interface mutations are an established clinical resistance mechanism and revumenib's CR/CRh rate in relapsed disease is 30% [E2]. **A fusion-driven antigen has neither weakness.** It cannot be abandoned without abandoning the driver, and it does not require a drug to appear. **Whether KMT2A::AFF1 has one is not known, because the experiment that produced FOLR1 has not been run on it** - this graph's own surface-antigen enumeration (`ana-surface-antigen-enumeration-q48`) lists NG2/CSPG4, CD15, HLA-DR, TdT, CD65 and CLEC12A, every one of which is a lineage or maturation marker rather than a fusion-specific transcriptional output.
Three kinds of antigen, and only one of them is escape-proof
| antigen class | example | how it is lost |
|---|---|---|
| lineage antigen | CD19 in KMT2A-rearranged B-ALL | lineage switch - the blast changes phenotype and keeps the fusion |
| drug-induced antigen | CLEC12A under menin inhibition, as this graph records | lost when the drug stops working, which in this disease it does - MEN1 interface mutation, 30% CR/CRh in relapse 2 |
| fusion-driven antigen | FOLR1 in CBFA2T3::GLIS2 1 | cannot be lost without losing the driver |
The third row is the only one where antigen loss and oncogene loss are the same event. That is the property CAR-T therapy in a plastic leukaemia most needs, and it is a property of the transcriptional output of the fusion, not of the cell type it happens to be in.
This is the constructive counterpart to a negative result in a sibling entry. analogy-npm1-alk-patients-immunise-themselves-against-the-fusion-and-kmt2a-patients-have-never-been-shown-to argues that KMT2A::AFF1 cannot supply a fusion neoantigen, because both partners are ubiquitously expressed and the only foreign sequence is the junction. A fusion-driven surface antigen is a different object and that argument does not apply to it. FOLR1 is an entirely normal self-protein; what makes it a target is that the fusion turns it on where it does not belong. KMT2A::AFF1 is a transcriptional neomorph with a large and well-characterised output programme, so the mechanism that produced FOLR1 is available to it in principle even though the neoantigen mechanism is not.
The transferable protocol, and why this project is well placed to run it
Le et al.'s route was: (1) express the fusion in human cord-blood HSPCs, (2) in a niche coculture that makes the model resemble the disease, (3) confirm the model against primary patient samples, (4) differential transcriptome to a fusion-specific gene library, (5) filter to surface proteins, (6) build the CAR 1.
Every step has a counterpart already present in this graph for KMT2A::AFF1 - human CD34+ and cord-blood fusion models, a fetal-progenitor cell-of-origin argument, primary infant samples, and an existing surface-antigen inventory. What is missing is step 4 restricted to step 5: nobody has published a KMT2A::AFF1-specific surface-gene library derived from an isogenic human model against its matched normal progenitor.
Two design points from the source are worth copying exactly. First, the comparison was against the matched normal progenitor in the same system, not against unrelated AML - which is what makes the list fusion-specific rather than lineage-specific, and is precisely where this project's existing antigen inventory falls down. Second, the endothelial niche was not decoration: the paper's own framing is that the niche promotes leukaemic transformation of the transduced cord-blood cells, so the model's fidelity - including its immunophenotype - depended on it. A KMT2A::AFF1 version run in a plain liquid culture may produce a different surface list for reasons that have nothing to do with the fusion.
Where the analogy breaks, including the part that would stop a trial
It breaks on toxicity, and the source says so itself. 'FOLR1 is also expressed in renal and pulmonary epithelium, raising concerns for toxicity that must be addressed for the clinical application of this therapy' 1. A fusion-driven antigen is escape-proof and it is not automatically tumour-restricted - the fusion turns on a normal gene, and that gene may be on somewhere the patient needs it. The escape-proof property buys durability, not safety, and those are different problems solved by different columns of a target table.
It breaks on disease biology. CBFA2T3::GLIS2 AMKL and KMT2A::AFF1 infant B-ALL share an infant age and a dismal prognosis, but the fusion architectures are unrelated - an oligomerising corepressor scaffold fused to a Gli-family DNA-binding domain versus a chromatin adaptor fused to an elongation-complex subunit - and there is no reason from first principles that a neomorphic transcriptional output must include a surface protein. The experiment could return an empty list, and that is a real possible outcome rather than a formality.
And it breaks on where the disease actually is. CBFA2T3::GLIS2 is a myeloid/megakaryoblastic disease being targeted with a CAR against a non-haematopoietic antigen, so normal marrow is spared. Most fusion-driven output in KMT2A::AFF1 that is well documented - HOXA9, MEIS1, the elongation programme - is nuclear. A surface list may be short for reasons that are structural rather than technical.
| domain | engineering |
|---|---|
| where the analogy breaks | Three ways. (1) Toxicity: FOLR1 is expressed in renal and pulmonary epithelium and the source flags this itself, so a fusion-driven antigen is durable but not necessarily safe. (2) Architecture: the two fusions are unrelated in structure and there is no principle requiring a neomorphic transcription factor to switch on a surface protein - the KMT2A::AFF1 list could be empty. (3) Compartment: the well-documented KMT2A::AFF1 output is nuclear, and the CBFA2T3::GLIS2 case benefits from targeting a non-haematopoietic antigen in a myeloid disease, which spares normal marrow in a way a B-lineage antigen would not. |
| what it predicts here | Running the Le et al. pipeline on KMT2A::AFF1 - isogenic human cord-blood or fetal-progenitor model, niche coculture, differential transcriptome against the MATCHED normal progenitor in the same system, filtered to surface proteins and cross-checked against primary infant samples - should return either a short fusion-specific surface list or an empty one. Both results are publishable and both change what this field does about lineage switch. |
| immune axis | surface antigen and CAR-T target discovery; cell type is the human cord-blood/fetal HSPC in both models |
| relationship to existing entries | fusion-cbfa2t3-glis2 already records FOLR1 and cites this paper as its E3; what it does not do is contrast the antigen CLASS with the KMT2A-r situation or extract the discovery pipeline as a transferable protocol. mechanism-menin-inhibition-induces-clec12a records the drug-induced-antigen route and calls it a target-creation claim; ana-surface-antigen-enumeration-q48 enumerates this project's antigens, all of which are lineage or maturation markers. analogy-npm1-alk-patients-immunise-themselves-against-the-fusion-and-kmt2a-patients-have-never-been-shown-to is the sibling entry: it closes the neoantigen route and this entry opens the fusion-driven-surface-antigen route, which is a different mechanism and is not touched by that argument. |
mapping
[
{
"bio": "CBFA2T3::GLIS2 drives FOLR1; a CAR against it is effective preclinically and the antigen is a transcriptional output of the fusion",
"analog": "KMT2A-rearranged B-ALL is treated with CD19, a LINEAGE antigen, which the leukaemia can abandon by lineage switch while keeping the fusion"
},
{
"bio": "the target came from a differential transcriptome of an isogenic human fusion model against its matched normal progenitor, in a niche coculture",
"analog": "no equivalent fusion-specific surface-gene library has been published for KMT2A::AFF1; this graph's antigen inventory is lineage and maturation markers"
},
{
"bio": "a fusion-driven antigen cannot be lost without losing the driver",
"analog": "the KMT2A-r alternative CLEC12A is reported to be INDUCED by menin inhibition - a drug-created target, contingent on the drug still working"
},
{
"bio": "FOLR1 is expressed in renal and pulmonary epithelium and the authors flag the toxicity risk",
"analog": "escape-proof is not the same as tumour-restricted; a fusion-driven antigen inherits the normal expression pattern of the gene the fusion switched on"
},
{
"bio": "the fusion's documented output in CBFA2T3::GLIS2 includes surface phenotype (CD56-bright, FOLR1)",
"analog": "KMT2A::AFF1's best-documented output - HOXA9, MEIS1, the elongation programme - is nuclear, so the surface list could genuinely be short"
}
]generated hypotheses
[]
antigen classes
[
{
"class": "lineage",
"example": "CD19 in KMT2A-rearranged B-ALL",
"escape": "lineage switch - documented in this disease"
},
{
"class": "drug-induced",
"example": "CLEC12A under menin inhibition",
"escape": "loss of drug activity; MEN1 interface mutation; 30% CR/CRh in relapse"
},
{
"class": "fusion-driven",
"example": "FOLR1 in CBFA2T3::GLIS2",
"escape": "requires losing the driver - but the antigen may be expressed in normal tissue"
},
{
"class": "fusion neoantigen (junction)",
"example": "none demonstrated in any fusion; anti-ALK epitopes map to the ALK moiety, not the junction",
"escape": "not applicable - route argued closed for KMT2A::AFF1 in the sibling entry"
}
]- On-target off-tumour toxicity is the named risk and it comes from the source, not from speculation: FOLR1 is expressed in renal and pulmonary epithelium. Any fusion-driven antigen inherits the normal expression pattern of whatever gene the fusion switched on, and the property that makes it escape-proof does nothing to make it tissue-restricted.
- A negative result is a real possibility. If KMT2A::AFF1's surface output is empty, this entry will have proposed an experiment whose value is entirely in closing a route - which is worth doing, but a programme should be told that in advance rather than after.
- Antigen density, not antigen presence, decides whether a CAR works. A transcriptional output detectable by RNA-seq can be far below the surface density a CAR needs, and a differential-expression list does not measure that.
- Combining a fusion-driven CAR with a menin inhibitor may be self-defeating if the antigen is part of the fusion's transcriptional output: the drug that evicts the fusion from chromatin would be expected to turn the antigen OFF. That is the exact inverse of the CLEC12A target-creation logic and it needs checking before the two are combined.
- What surface proteins does KMT2A::AFF1 itself drive, in an isogenic human model compared with its matched normal progenitor? This is the single experiment this entry exists to propose, and this graph's existing surface-antigen inventory cannot answer it because every antigen in it is a lineage or maturation marker.
- Does the fusion-driven antigen survive lineage switch in a model? The escape-proof claim is an inference. A KMT2A-r line driven to switch under CD19 pressure, assayed for whichever fusion-driven antigens exist, would test the principle directly and could be done with existing reagents once a candidate exists.
- Is the endothelial niche necessary for the surface phenotype, or only for transformation? Le et al. report that the niche was needed for the model to recapitulate immunophenotype as well as transcriptome, which means a KMT2A::AFF1 antigen screen run without a niche might produce a list that is an artefact of culture.
- For every fusion in this graph, is the fusion's transcriptional output known to include a surface protein? That is a tractable comparative table - CBFA2T3::GLIS2 supplies FOLR1 and CD56, NUP98::KDM5A and KMT2A fusions supply HOXA/MEIS1 - and it would say whether fusion-driven antigens are common or whether CBFA2T3::GLIS2 is a lucky exception.
- E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
Expression of the CBFA2T3-GLIS2 fusion in human cord-blood haematopoietic stem and progenitor cells in an endothelial-cell coculture system recapitulated the transcriptome, morphology and immunophenotype of CBFA2T3-GLIS2 AML and induced highly aggressive leukaemia in xenografts; interrogating the transcriptome of these cells and of primary CBFA2T3-GLIS2 AML identified a library of fusion-specific genes as potential therapeutic targets; CAR T cells directed against one of them, folate receptor alpha (FOLR1), showed preclinical efficacy in vitro and in xenograft models; and the authors note that FOLR1 is also expressed in renal and pulmonary epithelium, raising toxicity concerns that must be addressed before clinical application.
CBFA2T3-GLIS2 model of pediatric acute megakaryoblastic leukemia identifies FOLR1 as a CAR T cell target. · Le Q, Hadland B, Smith JL, Leonti A, Huang BJ, Ries R, Hylkema TA, Castro S, Tang TT, McKay CN, Perkins L, Pardo L, Sarthy J, Beckman AK, Williams R, Idemmili R, Furlan S, Ishida T, Call L, Srivastava S, Loeb AM, Milano F, Imren S, Morris SM, Pakiam F, Olson JM, Loken MR, Brodersen L, Riddell SR, Tarlock K, Bernstein ID, Loeb KR, Meshinchi S · J Clin Invest 2022What this evidence cannot show
Abstract read in full from the PubMed record on 2026-09-23. Preclinical only - in vitro and xenograft, no clinical data. The fusion-specific gene library is described but not enumerated in the abstract, so this entry cites the METHOD and the single named target rather than the list. The disease is paediatric acute megakaryoblastic leukaemia driven by CBFA2T3::GLIS2, not a KMT2A rearrangement and not B-lineage. The on-target off-tumour risk in kidney and lung is stated by the authors and is not hypothetical.
Evidence that argues against this (1)
- The escape-proof property claimed for a fusion-driven antigen is an inference from the fact that the fusion drives it, not a demonstrated resistance result: no CAR-T relapse series exists for FOLR1 in this disease, and antigen loss can occur through mechanisms that do not require losing the driver - epitope masking, transcriptional downregulation below the CAR's sensitivity threshold, or trogocytosis. The claim should be read as a prediction of the model, not as an observed durability advantage. Issa GC et al. 2023
- E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
In the first-in-human phase 1 trial of revumenib in relapsed or refractory KMT2A-rearranged or NPM1-mutant acute leukaemia, the complete remission or complete remission with partial haematologic recovery rate was 30% in the efficacy analysis population.
The menin inhibitor revumenib in KMT2A-rearranged or NPM1-mutant leukaemia. · Issa GC, Aldoss I, DiPersio J, Cuglievan B, Stone R, Arellano M, Thirman MJ, Patel MR, Dickens DS, Shenoy S, Shukla N, Kantarjian H, Armstrong SA, Perner F, Perry JA, Rosen G, Bagley RG, Meyers ML, Ordentlich P, Gu Y, Kumar V, Smith S, McGeehan GM, Stein EM · Nature 2023What this evidence cannot show
Abstract read in full from the PubMed record on 2026-09-23. Cited here for one narrow purpose - to quantify how contingent a menin-inhibitor-induced antigen would be in the relapsed setting where a CAR-T combination would first be tested. It is a relapsed/refractory phase 1 figure across a mixed KMT2A-rearranged and NPM1-mutant population and should not be read as the response rate in newly diagnosed infant ALL.
Evidence that argues against this (1)
- A drug-induced antigen is not disqualified by a 30% response rate: the antigen-induction effect may occur in patients who do not achieve remission, in which case the combination could work precisely where the single agent does not, and the preprint this graph records makes exactly that argument for CLEC12A. The contingency argument in this entry is therefore a risk, not a refutation, of the drug-induced antigen route. Le Q et al. 2022
- related to →CBFA2T3::GLIS2, inv(16)(p13.3q24.3)
the family that supplies its own CAR-T antigen from the fusion's transcriptional output
- related to →KMT2A::AFF1 (MLL-AF4), t(4;11)(q21;q23)
the index fusion, for which no fusion-specific surface-gene library has been built
the drug-induced antigen route this entry contrasts with a fusion-driven one
the KMT2A-r antigen whose expression is reported to require a menin inhibitor
the existing inventory, every member of which is a lineage or maturation marker rather than a fusion-specific output
sibling entry - that one closes the junction-neoantigen route, this one opens the fusion-driven surface-antigen route
runs the screen that entry proposed, and narrows its claim
narrows that entry's claim that nobody has built the fusion-specific surface list for KMT2A-r
the quantitative form of that entry's lineage-escape argument, and the correction that two fusion-driven antigens do exist