Competing model 2: the reciprocal allele is lost as the leukaemia evolves, and losing it selects for something worse
also called clonal loss model · AF4-MLL allele elimination · 40 percent lose the der(4)
A second model says every patient starts with both fusion genes and some leukaemias throw the mirror-image one away as they grow. The ones that throw it away behave worse.
The same thing, for a scientist
The Frankfurt group's stated motivation for its modelling work is to find a rational explanation for the elimination of the AF4-MLL allele in about 40% of patients, a loss they read as a clonal evolutionary process that selects for a more aggressive leukaemia [E1]. The model has a natural mechanistic fit with hit-and-run - a protein needed only at initiation is free to be lost - and it would explain the prognostic direction without requiring the reciprocal product to be protective. Its evidential problem is specific and large: **the word 'loss' describes a change over time, and every measurement is cross-sectional.** The 45% transcript figure is diagnostic material only [E2]; the 80-94% allele figures are diagnostic material only [E3][E4]. No paired diagnosis-relapse measurement of reciprocal status has been published, and this beat's searches for one returned nothing [E2, caveat]. Meanwhile the genomic re-analysis of reciprocal-negative t(4;11) patients found not deletion but rearrangement - reciprocal KMT2A alleles fused to other partners, out-of-frame joins, solitary 3'-KMT2A segments [E3] - which is a structural explanation for absence at diagnosis that requires no clonal evolution at all.
Three distinguishable claims that get written as one
(a) Some patients never generate a functional reciprocal allele, because the recombination was complex or out of frame. Directly supported 3. (b) Some patients generate one and silence it. Plausible; unmeasured. (c) Some patients generate one and lose it during progression. This is the clonal-loss model and it has no longitudinal evidence at all.
The 40% figure is compatible with all three and is routinely narrated as (c).
The experiment that would separate them, and why it has not been done
Reciprocal transcript status on paired diagnosis and relapse material from the same t(4;11) patients. The assay is one RT-PCR and the material exists in trial biobanks. The reason it has not been done is probably that MRD monitoring in this disease is keyed to the patient-specific genomic junction of the direct fusion 5, so nobody generates reciprocal-junction data after diagnosis as a by-product of routine care. A clone that lost the der(4) allele and a clone that kept it are identical to every assay in clinical use.
| kind | clonal evolution model |
|---|---|
| claim | the reciprocal allele is present initially and lost during progression, selecting a more aggressive clone |
| figure quoted | about 40% of patients lose the AF4-MLL allele |
| evidence class | cross-sectional only |
| longitudinal measurements published | 0 |
| competing structural explanation | complex, cryptic, three-way or out-of-frame recombination present from the outset |
- If reciprocal status changes between diagnosis and relapse, then stratifying menin-inhibitor therapy on diagnostic reciprocal status would misclassify the relapsed disease actually being treated - and relapse is exactly when menin inhibitors are used.
- A clonal-loss model implies treatment pressure could select for reciprocal-negative disease, which is the arm with 11.7% event-free survival. No protocol monitors for that.
- The model is narrated with a longitudinal verb and supported with cross-sectional data, which is the shape this project has repeatedly found behind confident wrong numbers.
- What is reciprocal transcript status at relapse in patients positive at diagnosis? One RT-PCR on paired biobanked material answers it and nobody has run it.
- What fraction of the 40% is explained by the structural causes already documented - complex, cryptic, three-way and out-of-frame recombination - rather than by loss?
- Does any centre run reciprocal-junction MRD? If so the longitudinal data may exist unanalysed.
- If the der(4) allele is lost, is it lost as a whole chromosome, a segment, or by silencing? The three have different detectability and different mechanisms.
- E1moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper
The elimination of the AF4-MLL allele in about 40% of patients appears to worsen disease outcome and is interpreted as a clonal process in which loss of the reciprocal allele selects for a more aggressive type of leukaemia.
The role of reciprocal fusions in MLL-r acute leukemia: studying the chromosomal translocation t(4;11). · Wilhelm A, Marschalek R · Oncogene 2021What this evidence cannot show
Stated as the motivation and interpretive frame of a modelling paper that contributes no patient data; the underlying figure is cross-sectional; 'elimination' is the authors' term for what the cited source reports as absence of expression at diagnosis.
Evidence that argues against this (1)
- Genomic analysis of t(4;11) patients lacking a detectable reciprocal allele found complex, cryptic and three-way translocations and out-of-frame joins rather than deletion of a previously present allele, so absence at diagnosis has a structural explanation that does not require clonal loss. Kowarz E et al. 2007
- E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
Reciprocal transcript status in the cohort that established the prognostic effect was measured by RT-PCR on diagnostic material in 124 infants, with 45% (19/43) of t(4;11) cases positive; recurrent RAS mutations in the same cohort were subclonal and frequently lost at relapse.
Unraveling the cellular origin and clinical prognostic markers of infant B-cell acute lymphoblastic leukemia using genome-wide analysis. · Agraz-Doblas A, Bueno C, Bashford-Rogers R, et al. · Haematologica 2019What this evidence cannot show
Diagnostic timepoint only for reciprocal status. The cohort did analyse relapse material and reports clonal change there - for RAS mutations - which shows the material and the capability existed and reciprocal status was simply not asked at the second timepoint. Bounded absence, searched 2026-09-19: '(AF4-MLL OR AF4.MLL) AND (relapse OR paired OR longitudinal)' returns 4 records, none longitudinal for reciprocal status; 'AF4-MLL AND minimal residual disease' returns 2, both methodological.
Evidence that argues against this (1)
- Because the same study demonstrates subclonal loss of RAS mutations between diagnosis and relapse, the clonal-evolution framework it supports is real in this disease - which makes the untested reciprocal case a genuinely open question rather than an implausible one. Agraz-Doblas A et al. 2019
- E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
t(4;11) patients scored as lacking the reciprocal allele carry cryptic and three-way translocations, in-frame reciprocal MLL fusions to novel partners including NFKB1 and RABGAP1L that are actively transcribed, out-of-frame reciprocal fusions to PBX1, ELF2, DSCAML1 and FXYD6, and in some cases only solitary 3'-MLL segments; the authors propose that all t(4;11) patients carry a reciprocal MLL allele.
Complex MLL rearrangements in t(4;11) leukemia patients with absent AF4.MLL fusion allele. · Kowarz E, Burmeister T, Lo Nigro L, et al. · Leukemia 2007What this evidence cannot show
Selected subgroup, n not stated in the abstract; establishes that structural explanations for absence exist without quantifying how much of the 40% they account for.
Evidence that argues against this (1)
- The same field's later framing continues to describe the phenomenon as allele elimination during clonal progression, so the structural explanation has not displaced the evolutionary one in practice. Wilhelm A et al. 2021
- E4strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
A reciprocal fusion gene is generated in 94% of KMT2A rearrangements in the Frankfurt series, with 70% of rearrangements balanced.
The reciprocal world of MLL fusions: A personal view. · Marschalek R · Biochim Biophys Acta Gene Regul Mech 2020What this evidence cannot show
Single-centre referral series presented as a personal view; a diagnostic-timepoint count.
Evidence that argues against this (1)
- Allele generation at 94% against transcript expression at 45% means the majority of the gap is already present at diagnosis, which limits how much clonal loss during therapy can be contributing. Agraz-Doblas A et al. 2019
- E5strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
Genomic breakpoint sequences from KMT2A-rearranged patients are used to monitor minimal residual disease, which the recombinome study presents as its direct bench-to-bedside translation.
The KMT2A recombinome of acute leukemias in 2023. · Meyer C, Larghero P, Almeida Lopes B, et al. · Leukemia 2023What this evidence cannot show
The abstract does not state which junction is used for MRD; the inference that it is the direct junction follows from the fact that the direct fusion is what is typed diagnostically, and was not independently confirmed for this entry.
Evidence that argues against this (1)
- If reciprocal-junction MRD assays exist in any centre, the longitudinal data this model needs may already have been generated and not analysed - which would make the absence a reporting gap rather than a measurement gap. Marschalek R 2020
a human case in which the reciprocal allele is absent by genomic deletion
Model for why the reciprocal product is absent in a large minority of patients.
A protein needed only at initiation is free to be lost later; the two models are designed to fit together.
Offers an explanation for the survival split that does not require the reciprocal product to be protective.
Whether loss has occurred cannot be separated from which assay was used.