Treating a subclonal lesion is selection pressure applied to the clone you did not aim at
also called the subclonal target problem · why a 15% RAS mutation is a worse target than no RAS mutation · genotype-directed therapy against a minority clone
If only a fraction of a leukaemia's cells carry the mutation a drug is aimed at, the drug kills that fraction and leaves the rest - and the rest is most of the disease. In infant KMT2A leukaemia the drug-targetable mutations are usually in a minority of cells, which makes aiming at them a way of clearing the field for the cells nobody targeted.
The same thing, for a scientist
This is the structural risk that the whole cooperating-lesions beat produces, stated once so it can be cited rather than re-derived. **The setup.** In infant KMT2A-rearranged ALL, **65% (20/31) of the activating kinase/PI3K/RAS mutations sit below 30% mutant allele frequency** at 92% mean tumour purity, and in 6 of 8 cases carrying two or more such mutations, each was below 30% - meaning the lesions occupy different subclones [E1]. Below that, duplex sequencing finds drug-resistant clones at **0.06-0.34% of diagnostic cells** that expand at relapse [E2]. **The risk.** A genotype-directed agent selected on a subclonal lesion exerts differential pressure. The clone carrying the target is killed preferentially; the clone that does not carry it is not. The result is a leukaemia enriched for cells the drug was never chosen against, with the genotype that prompted the treatment now absent - which reads clinically as a response followed by a genotypically different relapse. **And the documented case is not hypothetical.** After conventional chemotherapy, the clones that expanded at relapse in this disease were clones present at diagnosis **below the detection floor of the assay used to characterise the patient** [E2]. The mechanism by which therapy selects a minority clone in this leukaemia is already on record; the only untested part is whether a targeted agent does it faster. **The counter-consideration that keeps this honest.** A subclonal lesion that confers a functional advantage - dissemination, glucocorticoid resistance - may be worth suppressing even if it is not the bulk of the disease, and RAS-mutant infant blasts are more glucocorticoid-resistant in vitro [E3]. Suppressing a resistant minority during induction is a coherent strategy. The risk is specific to treating a subclonal lesion as though it were the disease's driver, not to touching it at all.
Where this risk is live in current practice and current proposals
MEK inhibition selected on RAS genotype. The proposal most often made in this disease. Two facts undercut the genotype-selection half of it independently: the target is subclonal in about two thirds of patients 1, and ex-vivo MEK sensitivity does not track RAS genotype - RAS-wild-type samples were both sensitive and resistant, with no downstream over-activation to explain either 3. So the selection variable is neither clonal nor predictive.
Menin inhibition plus a RAS-pathway agent. The one in-vivo test, in KMT2A-rearranged RAS-mutant AML where the RAS lesions are more often clonal, reduced leukaemia burden and did not extend survival 4. That is what a combination looks like when it kills the targeted subpopulation without changing what causes death.
SHP2 inhibition. Doubly blocked: the PTPN11 lesion is non-recurrent in this disease, and the one registered leukaemia protocol carrying a SHP2 inhibitor was withdrawn for 'Sponsor decision.' at an 18-year age floor.
MRD interpretation. The subtler version of the same risk. If a patient is monitored on a cooperating mutation rather than on the fusion, a falling variant allele fraction can mean the targeted subclone is being cleared while the leukaemia is not - a reassuring number produced by the exact mechanism that should worry the clinician. The fusion is clonal and stable and is the correct MRD marker in this disease.
Mitigation, in order of practicality
- Never select a targeted agent on a lesion without measuring its clonality in that patient. A mutation report that gives a variant allele fraction and a tumour purity supports this; one that gives a present/absent call does not. This costs nothing and is not standard.
- Keep MRD on the fusion. The KMT2A fusion is present in every leukaemic cell and stable through relapse, and in adult KMT2A-rearranged BCP-ALL fusion-based MRD was shown to be highly reliable where immunoglobulin/T-cell-receptor markers failed in about a third of patients.
- Where a subclonal target is treated anyway, pre-specify a clone-resolved readout.
exp-duplex-clone-tracking-under-menin-inhibitionis the design; without it, a response and a selection event are indistinguishable. - Prefer mechanism-chosen combinations over genotype-chosen ones in this disease, which is the argument of
mechanism-quiet-genome-forecloses-cooperating-lesion-combinations. A combination whose benefit does not depend on a co-mutation cannot select against it - and MEK inhibition's synergy with prednisolone, which occurred in RAS-wild-type cells too, is the existing candidate.
| kind | resistance |
|---|---|
| severity | high in principle, unmeasured in practice: the selection mechanism is documented for chemotherapy in this disease and has never been measured under a targeted agent |
| mechanism | differential killing of a lesion-bearing minority clone leaves the lesion-free majority, producing a response followed by a genotypically different relapse; the genotype that prompted treatment is absent at relapse, which is routinely read as the lesion having been unimportant rather than as evidence of selection |
| mitigation | measure clonality (variant allele fraction with tumour purity) before selecting any genotype-directed agent; keep MRD on the fusion rather than on a cooperating mutation; pre-specify a clone-resolved readout wherever a subclonal target is treated; prefer combinations whose benefit does not depend on the co-mutation being present |
affects
[ "drug-revumenib", "comut-ras-pathway-kmt2a-r" ]
quantitative basis
{
"subclonal_fraction": "65% (20/31) of activating kinase/PI3K/RAS mutations below 30% MAF at 92% purity",
"separate_subclones": "6 of 8 cases with two or more pathway mutations had each below 30%",
"documented_selection": "resistant clones at 0.06-0.34% of diagnostic cells expanded at relapse after conventional chemotherapy"
}- This entry is itself an argument against a class of interventions, and it could be wrong in the direction that costs patients: if a subclonal RAS lesion confers glucocorticoid resistance, refusing to target it because it is subclonal preserves exactly the cells induction fails to kill.
- The mitigation of keeping MRD on the fusion is sound and has a cost: fusion MRD is blind to clonal composition, so a protocol that follows this advice will not be able to detect the selection event the advice exists to prevent.
- The risk is easiest to state and hardest to act on in infants, where the alternative to a genotype-directed agent is more cytotoxic chemotherapy, which is itself the therapy documented to have selected minority clones in this disease.
- If clone-resolved measurement becomes the precondition for genotype-directed therapy, the practical effect in a rare infant disease may simply be that no targeted agent is used at all, which is a decision with its own mortality.
- Has any trial in any leukaemia measured clone-level allele frequency before and after a targeted agent aimed at a subclonal lesion? If the answer is no across oncology, this risk is universally assumed and nowhere measured.
- At what clonal fraction does a genotype-directed agent stop being a reasonable choice? Nothing in the literature offers a threshold, and clinicians act on present/absent calls.
- Does suppressing a glucocorticoid-resistant RAS-mutant minority during induction improve outcome even though the clone is not the driver? That is the strongest argument against treating this risk as decisive, and it is testable.
- Would fusion-based MRD detect a selection event at all? It measures total leukaemic burden and is blind to composition, so the assay this entry recommends for reliability is by construction blind to the risk it describes.
- E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
In infant MLL-rearranged ALL, 65% (20 of 31) of activating kinase/PI3K/RAS mutations had mutant allele frequency below 30% despite 92% mean tumour purity, and in 6 of the 8 cases carrying two or more pathway mutations each mutation was below 30%, indicating occupancy of different subclones.
The landscape of somatic mutations in infant MLL-rearranged acute lymphoblastic leukemias. · Andersson AK, Ma J, Wang J, et al. · Nature Genetics 2015What this evidence cannot show
Derived from 31 mutations across 22 cases, with deep-capture confirmation on five cases at an average of 437x. Mutant allele frequency is a proxy for clonal fraction that assumes heterozygosity and two copies; a mutation on an amplified or lost allele breaks the inference. The '6 of 8' inference that two pathway mutations sit in different subclones is a reasonable reading of two sub-30% fractions but is not a single-cell demonstration. Full text read via efetch db=pmc (PMC4553269).
Evidence that argues against this (1)
- A mutant allele frequency below 30% is also produced by a clonal mutation in a sample with lower purity than reported, or by copy-number change at the locus; without single-cell or allele-specific copy-number data, subclonality is inferred rather than observed. Andersson AK et al. 2015
- E2moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper
In infant and childhood KMT2A-rearranged ALL, drug-resistant clones carrying NRAS G13D or G12S or FLT3 D835H, comprising 0.06% to 0.34% of diagnostic cells, expanded at relapse, and in longitudinal samples the relapse clone persisted as a minor subclone from diagnosis and through treatment before expanding during the last month of disease.
Duplex Sequencing Uncovers Recurrent Low-frequency Cancer-associated Mutations in Infant and Childhood KMT2A-rearranged Acute Leukemia. · Pilheden M, Ahlgren L, Hyrenius-Wittsten A, et al. · HemaSphere 2022What this evidence cannot show
27 patients, of whom 5 contributed longitudinal and 6 paired relapse material. The treatment involved was conventional chemotherapy, so this documents therapy-driven selection of a minority clone in this disease without establishing that a targeted agent behaves the same way. Abstract read; full text not retrieved for this entry.
Evidence that argues against this (1)
- Expansion of a pre-existing minor clone under chemotherapy is the expected behaviour of any cytotoxic therapy and does not by itself show that GENOTYPE-DIRECTED therapy carries extra risk; the specific claim that a targeted agent accelerates the process is untested in this disease. Pilheden M et al. 2022
- E3moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper
MEK inhibitors impaired primary RAS-mutant KMT2A-rearranged infant ALL cells in vitro and synergised with prednisolone in both RAS-mutant and RAS-wild-type cells, while among RAS-wild-type samples both sensitive and resistant cases were found with no apparent downstream over-activation to explain the difference; separately, RAS-mutated infants were significantly more resistant to glucocorticoids in vitro.
MEK inhibition is a promising therapeutic strategy for MLL-rearranged infant acute lymphoblastic leukemia patients carrying RAS mutations. · Kerstjens M, Driessen EM, Willekes M, et al. · Oncotarget 2017What this evidence cannot show
Ex-vivo cytotoxicity on small numbers of primary samples with no in-vivo arm. The glucocorticoid-resistance observation is from the earlier Interfant cohort by overlapping authors rather than from this paper's own data, and is cited here as context for why suppressing a resistant minority during induction could still be rational. PubMed gives the year as 2017 where several sibling entries in this graph record 2016 for the same DOI.
Evidence that argues against this (1)
- If MEK inhibition's benefit comes from prednisolone synergy that does not require a RAS mutation, then the drug should not be selected on genotype at all - which removes the subclonal-target risk by removing the genotype selection, and is the cleanest available mitigation. Kerstjens M et al. 2017
- E4moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper
Revumenib combined with selumetinib in KMT2A-rearranged RAS-pathway-mutant AML was synergistic in vitro and reduced leukaemia burden in patient-derived xenograft models in vivo, but did not improve overall survival compared with menin inhibition alone.
Combining menin and MEK inhibition to target poor prognosis KMT2A-rearranged RAS pathway-mutant acute myeloid leukemia. · Scheidegger N, Schneider C, Alexe G, et al. · Blood Advances 2026What this evidence cannot show
AML rather than infant ALL, PDX models rather than patients, and the study does not report clone-level allele frequencies under treatment - so it demonstrates burden reduction without survival benefit and does not demonstrate that the mechanism was clonal selection. Abstract read; full text not retrieved for this entry.
Evidence that argues against this (1)
- A burden reduction without survival benefit is equally consistent with inadequate dosing or schedule, which the authors themselves propose, so it should not be presented as evidence of the selection mechanism this entry describes. Scheidegger N et al. 2026
the only design that distinguishes a response from a selection event
the structural risk that follows from 65% of these lesions sitting below 30% MAF
why a 15% KRAS mutation is a worse target than no KRAS mutation
the mechanism-over-genotype conclusion follows from this risk
pre-existing minor clones and the burden at which escape is expected