Oncofusion

start / graph / target-rna-targeting-fusion-transcript

Targetmoderate confidence 3 of 5: One strong study, or several smaller consistent ones.Q1b-genesis2

RNA-targeting the fusion transcript: verdict THE LEAST HOPELESS of the three fusion-level routes, blocked by delivery rather than by biology

also called siRNA against MLL-AF4 · antisense oligonucleotide fusion junction · Cas13 · fusion transcript knockdown

The fusion makes a hybrid message that exists in no healthy cell, so a drug that destroys just that message would be as specific as anything in medicine. It works in the dish and in mice. The reason it is not a treatment is that we cannot reliably get such drugs into leukaemia cells in the body.

The same thing, for a scientist

Of the three fusion-level intervention routes the PI asked about, this is the only one where the biology is proven and the obstacle is purely pharmaceutical. The fusion junction is a sequence that exists in no normal transcriptome, giving a specificity ceiling no small molecule can match. Proof of concept is old and solid: transient siRNA against MLL-AF4 in t(4;11) human lines impaired proliferation and clonogenicity, induced caspase-3-dependent apoptosis with reduced BCL-XL, lowered HOXA7, HOXA9 and MEIS1, reduced the stem marker CD133 with evidence of differentiation, and reduced leukaemia-associated morbidity and mortality in xenografted SCID mice, which the authors interpreted as an effect on leukaemia-initiating cells [E1]. Two things stand between that and a drug. **Delivery** to haematopoietic cells in vivo is the unsolved problem of the whole oligonucleotide field; the approved successes are hepatocyte-directed via GalNAc-ASGPR, an address that blood cells do not have. **Junction heterogeneity**: KMT2A breakpoints are distributed across an 8.3 kb region and the partner is one of more than a hundred genes, so a junction-specific reagent is patient-specific by construction, which collides with the economics of a disease affecting ~100 US infants a year. The honest verdict is 'blocked, but blocked on something the wider field is actively solving', which is not true of prevention or of editing.

Go deeper
Why the specificity argument is stronger here than anywhere else in the project1 deeper

A menin inhibitor blocks a pocket on a wild-type protein that has essential functions in normal tissue (gene-men1, interface-menin-kmt2a-mbm1), which is the origin of every on-target toxicity concern in the therapeutics beat. A junction-directed oligonucleotide has no wild-type counterpart to hit. The one caveat is the trans-splicing and read-through literature (mechanism-trans-splicing-chimeric-rna-normal-tissue, mechanism-readthrough-transcription-chimeras-normal): chimeric transcripts do occur in normal tissue, so 'no normal cell makes this sequence' should be verified for the specific junction rather than assumed. For a genomic KMT2A::AFF1 junction with patient-specific intronic sequence retained, the assumption is safe; for an exon-exon junction shared across patients, it is an assumption that should be tested.

Three modalities, ranked

ASO (gapmer, RNase H-mediated). Chemically mature, clinically validated in other diseases, and the only one of the three with an approved-drug precedent in an oligonucleotide class. Junction-spanning gapmers are straightforward to design. Delivery to marrow is the gap.

siRNA. The published proof of concept in this exact disease 1. Same delivery gap, plus a conjugate landscape (GalNAc) that is liver-directed.

Cas13 / RNA-targeting CRISPR. Highest theoretical specificity and the ability to target without RNase H chemistry, but requires protein delivery and carries collateral-cleavage concerns; least mature. It is a research tool here, not a candidate.

Relationship to degraders

Protein degradation is the sibling strategy and is already covered in this graph by target-phase2-degrader-modality-dossier. The two differ in where they act and in what they need: a degrader needs a binder to the fusion protein, which is hard because KMT2A::AFF1 is largely intrinsically disordered outside its structured domains; an RNA agent needs no binder at all, only a sequence. That asymmetry is why RNA targeting deserves standing separately rather than being folded into the degrader dossier, and why the two should be pursued as alternatives rather than as variants of one idea.

Facts and figures
modalityRNA-targeting (ASO gapmer, siRNA, Cas13)
verdictLEAST HOPELESS of prevention / early detection / correction; blocked on delivery, not on biology
targetfusion-kmt2a-aff1
specificity ceilingthe fusion junction has no wild-type counterpart, giving a specificity no small molecule targeting a wild-type partner protein can match
tractable next stepdesign exon-exon junction gapmers for the handful of recurrent KMT2A::AFF1 junctions that cover most patients, and test them against the marrow-delivery problem with existing non-hepatic LNP formulations
blocking failures
[
  "no validated systemic delivery of oligonucleotides to haematopoietic cells; approved conjugates are hepatocyte-directed",
  "junction heterogeneity across an 8.3 kb KMT2A breakpoint cluster region and >100 partners makes genomic-junction reagents patient-specific",
  "patient-specific reagents collide with the economics of ~100 US cases per year"
]
How this could bite us back
  • An oligonucleotide that is patient-specific has no regulatory or commercial path at this disease's incidence, so the modality could be scientifically correct and still never reach a child; the n-of-1 trial design entries in this graph exist for exactly this situation.
  • If a chimeric transcript with the same junction exists in any normal tissue, the specificity argument collapses and the agent acquires an unpredictable toxicity.
  • Partial knockdown selects for cells with higher fusion expression, which is a resistance route that small-molecule interface disruption does not obviously have.
Still open
  1. How many distinct exon-exon KMT2A::AFF1 junctions would a reagent set need to cover 80% of patients? The KMT2A recombinome data needed to answer this already exist in the graph's breakpoint entries, and the answer decides whether this is a patient-specific or an off-the-shelf modality. It is a one-afternoon analysis and nobody in this project has done it.
  2. Is there any LNP or conjugate formulation with demonstrated uptake by leukaemic blasts in marrow? The field's attention is on hepatocytes and, increasingly, on T cells; leukaemic blasts have not been a target of formulation development.
  3. Would partial knockdown suffice? This is the same threshold question as for degron depletion (`hyp-fusion-continuously-required-not-hit-and-run`), and a single titration experiment would serve both.
  4. Could an RNA agent be combined with a menin inhibitor to attack the same node by two mechanisms, and would that be orthogonal or merely redundant? Given the low mutational burden, two independent routes to the same target may be the only combination available.
Every source behind this entry (3)
  1. E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    Transient siRNA knockdown of MLL-AF4 in t(4;11) human leukaemia cells reduces proliferation and clonogenicity, induces caspase-3-dependent apoptosis with reduced BCL-XL, lowers HOXA7/HOXA9/MEIS1 and CD133, and reduces leukaemia-associated morbidity and mortality in xenografted SCID mice.

    Targeting MLL-AF4 with short interfering RNAs inhibits clonogenicity and engraftment of t(4;11)-positive human leukemic cells · Thomas M, Gessner A, Vornlocher HP, Hadwiger P, Greil J, Heidenreich O · Blood 2005
    PMID 16046533DOI
    What this evidence cannot show

    Two cell lines (SEM, RS4;11); siRNA delivered by transfection ex vivo before transplantation rather than systemically to established disease, so the xenograft result speaks to leukaemia initiation more cleanly than to treatment of established leukaemia. No systemic delivery was attempted, which is precisely the unsolved step.

    Evidence that argues against this (4)
    • Even perfect transcript elimination has a pharmacodynamic floor set by protein half-life, and for this fusion that floor is unusually high. Liu et al. showed that wild-type MLL is degraded in cell-cycle-specific windows by SCF(Skp2) and APC(Cdc20), that degradation initiates within the N-terminal approximately 1400 residues retained by every fusion, and that the prevalent MLL fusions tested resist that degradation. A gapmer or siRNA that abolishes the fusion transcript therefore acts on the fastest step while the slowest step is untouched, and the E1 siRNA experiments (transfection before transplantation rather than treatment of established disease) could not have revealed this. It is a specific reason to expect an RNA agent against KMT2A::AFF1 to act more slowly and shallowly in vivo than the ex vivo data suggest, and an argument for measuring fusion PROTEIN rather than transcript as the pharmacodynamic endpoint. Read: abstract only. Liu H et al. 2007
    • If the modality chosen is Cas13 rather than a gapmer or siRNA, the specificity ceiling this entry claims does not exist. Ai et al. showed by co-transfection in Drosophila and human cells that the off-target effects of RxCas13d, the most widely used Cas13 effector, can be as strong as the on-target knockdown; the extent scales with target RNA expression level and persists even when enzyme and guide levels are lowered. PspCas13b was more specific but still not clean, and the magnitude varied with cell type and target. A perfectly unique junction guide does not help, because the damage is done to bystander RNAs after the enzyme is activated, not through mis-targeting. Read: abstract only. Ai Y et al. 2022
    • Independent confirmation of the same failure mode in human cells. Shi et al. found that RfxCas13d (CasRx) causes collateral transcriptome destruction when targeting abundant reporter or endogenous RNAs, with a consequent proliferation defect in the targeted cells, and noted that the initial studies failed to detect this in eukaryotic cells, which is how the specificity claim entered the field unchallenged. The second-order point matters here: they showed the collateral activity can be deliberately harnessed to deplete a cell population defined by a marker RNA, so in a leukaemia the bystander toxicity might be repurposed as the killing mechanism rather than avoided, at the cost of abandoning the junction-specificity argument that motivated the modality in the first place. Read: abstract only. Shi P et al. 2023
    • Counter to the counter, recorded so the entry is not left pessimistic by default. Tong et al. built a dual-fluorescence reporter for collateral activity, screened over 200 engineered variants, and isolated high-fidelity Cas13d and Cas13X variants that retain wild-type on-target knockdown with markedly reduced collateral activity, no detectable transcriptome-wide off-targets, no Cas13-induced growth arrest, and no detectable collateral damage in transgenic mice or in AAV-mediated somatic targeting. The Cas13 collateral problem is therefore engineerable and largely engineered away; it is a reason to specify the variant in any proposal, not a reason to exclude the modality. Read: abstract only. Tong H et al. 2023
  2. E2moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    Cancer-associated fusion transcripts can be produced in normal cells without a genomic rearrangement, so junction specificity should be verified rather than assumed for any exon-exon junction reagent.

    A chimeric RNA characteristic of rhabdomyosarcoma in normal myogenesis process · Yuan H, Qin F, Movassagh M, Park H, Golden W, Xie Z, Zhang P, Sklar J, Li H · Cancer Discovery 2013
    PMID 24089019DOI
    What this evidence cannot show

    A different fusion in a different lineage; no trans-spliced KMT2A::AFF1 transcript has been reported in normal tissue. Cited as a reason to check, not as evidence that the problem exists here.

    Evidence that argues against this (1)
    • Someone has in fact looked for a trans-spliced version of this exact fusion, and did not find one. Yamamoto et al. applied nested PCR to 65 normal cord blood samples and detected no MLL or AF-4 chimeric mRNA in any of them. That is reassuring for a KMT2A::AFF1 junction gapmer. The same study is a warning about the other end of the reagent: an in-frame AF-4 (AFF1) exon 11 to exon 4 or 5 fusion transcript was present in 20 of 65 healthy cord bloods (31%) and in 4 of 6 t(4;11) leukaemic samples, while absent from 21 peripheral blood samples and from fetal fibroblasts. A reagent whose complementarity extends into AFF1 exons 4-5, or across that internal junction, would hit an abundant transcript in normal fetal blood. Read: abstract only. Yamamoto S et al. 1998
  3. E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    Infant KMT2A-rearranged ALL carries one of the lowest somatic mutation frequencies of any sequenced cancer - whole-genome, exome, RNA and targeted sequencing of 65 infants and 20 older children - so the fusion is the near-sole driver, which is the premise an agent eliminating the fusion transcript depends on. REWORDED 2026-09-20: this item previously claimed 'removing the fusion collapses the leukaemic programme', which this paper does not show - it is a mutational landscape study and reports no fusion-removal experiment. The caveat already said so; the claim now says what the citation actually supports, per concept-the-confidence-grading-rule, which grades the claim rather than the source. The collapse evidence itself lives in mechanism-fusion-dependency-reversibility-kinetics and in E1.

    The landscape of somatic mutations in infant MLL-rearranged acute lymphoblastic leukemias · Andersson AK, Ma J, Wang J, Chen X, Gedman AL, Dang J, et al. · Nature Genetics 2015
    PMID 25730765DOI
    What this evidence cannot show

    This paper establishes that the fusion is the near-sole driver rather than directly demonstrating programme collapse; the collapse evidence is in mechanism-fusion-dependency-reversibility-kinetics and in E1.

    Evidence that argues against this (1)
    • The premise that removing the fusion collapses the programme is not safe for a transcription-factor oncogene in the haematopoietic system. Green et al. transiently expressed Bcl6 in murine haematopoietic stem/progenitor cells and obtained mature B-cell lymphomas that LACKED Bcl6 expression and lacked Bcl6 target-gene repression, while carrying epigenetic changes conserved from the HSPC stage through to the mature B cell: an explicit hit-and-run mechanism in a B-lineage malignancy. If KMT2A::AFF1 behaved that way in even a subset of cases, an agent that eliminated the fusion transcript would leave the installed state intact and the specificity advantage of RNA targeting would buy nothing. This is a different oncoprotein in a mouse model, and the KMT2A case is genuinely open (see hyp-fusion-continuously-required-not-hit-and-run). Read: abstract only. Green MR et al. 2014
Connected to