No second driver means no second target - the quiet genome forecloses the combination logic that works in every other leukaemia
also called no cooperating lesion to co-target · why 'add a second targeted agent' does not transfer to infant KMT2A-r ALL · the load-bearing negative of the quiet genome
In most leukaemias, when one targeted drug stops working you add a second one aimed at another mutation the cancer carries. Infant KMT2A leukaemia barely has another mutation to aim at, so that strategy has nothing to attach to. Any combination in this disease has to be chosen from mechanism rather than from the patient's genome.
The same thing, for a scientist
This entry states a negative that follows from the rest of the beat and that this graph has never written down plainly. **The premise.** Infant KMT2A-rearranged ALL carries a mean of 1.3 non-silent mutations in the dominant clone [E1]; the only recurrently point-mutated genes are KRAS and NRAS [E2]; two thirds of those lesions are subclonal and a substantial fraction are gone at relapse [E1]; the recurrent copy-number lesions - PAX5 in 23%, CDKN2A/B in 8.5% - are **tumour-suppressor losses, which are not druggable** [E1]; and TP53, DNMT3A and IKZF1, the adult partners, are not present at diagnosis [E3][E4]. **The consequence.** The combination logic that drives modern leukaemia therapy - identify the co-mutation, add the agent that hits it, as with FLT3 inhibitors in NPM1-mutant AML or TKIs plus chemotherapy in Ph-positive ALL - **has no substrate here**. There is no second clonal driver, so there is nothing for a second targeted agent to be selected on. **And the one attempt to run the logic anyway produced the predicted result.** The best-supported co-targeting experiment in KMT2A-rearranged disease combined the menin inhibitor revumenib with the MEK inhibitor selumetinib in RAS-mutant KMT2A-rearranged AML. It was synergistic in vitro and reduced leukaemia burden in PDX models in vivo, and it **did not improve overall survival over menin inhibition alone** [E5]. The authors' own conclusion is that the combination will require further optimisation. **What this does not say.** It does not say combinations are futile in this disease. It says combinations here must be chosen on **mechanism** - the elongation complex, the differentiation block, apoptotic priming, the immune compartment - rather than on **genotype**, and that any protocol proposing to stratify patients for a combination by co-mutation status will find almost nobody to stratify.
The inventory, and what is left after it
Everything the graph now holds about co-lesions in infant KMT2A-rearranged ALL, sorted by whether a drug could act on it:
| lesion | frequency | druggable? | why not |
|---|---|---|---|
| KRAS mutation | recurrent, 4/22 WGS; subclonal in ~2/3 | indirectly, via MEK | subclonal target; MEK sensitivity ex vivo does not track genotype |
| NRAS mutation | recurrent, 2/22 WGS; resistant clones at 0.06-0.34% | indirectly, via MEK | same, plus the reservoir alleles are G13D/G12S, not the mutant-selective chemistry's targets |
| PTPN11 mutation | non-recurrent | yes, SHP2 inhibitors exist | no open paediatric route; the one leukaemia protocol was withdrawn for 'Sponsor decision.' |
| NF1 loss | 1 homozygous deletion in 22 | no | tumour suppressor; nothing restores it |
| PAX5 focal loss | 5/22 (23%) - the commonest | no | monoallelic transcription-factor loss |
| CDKN2A/B deletion | 4/47 (8.5%) | CDK4/6 inhibition in principle | under one in ten patients, clonality unmeasured, untested in this disease |
| TP53 | not reported at diagnosis | no | and absent anyway |
| IKZF1 | not reported at diagnosis | no | and the available pharmacology degrades it further |
| DNMT3A | not reported at diagnosis | no | adult lesion |
The commonest recurrent second lesion in this disease is an undruggable monoallelic tumour-suppressor loss, and the only druggable ones are subclonal. That is the whole argument in one line.
Four consequences that follow, in order of confidence
1. Genotype-selected combination trials cannot be powered in this disease. With a modal disease incidence already in the tens of cases per year in a large trial group, selecting on a co-mutation present in 8-25% of patients, some of it subclonal, leaves a stratum too small to randomise. This is an arithmetic consequence, not a biological claim.
2. The pressure moves to mechanism-chosen combinations. Venetoclax, DOT1L inhibition, immunotherapy and chemotherapy backbones are chosen against the fusion's own biology rather than a second genotype. Those are the combinations this graph should evaluate, and the criterion for them is not 'does the patient carry the target' but 'does the fusion create the dependency'.
3. Resistance must be pre-empted rather than genotyped. In other leukaemias the co-mutation landscape at relapse tells you what to add. Here the relapse landscape is TP53 and IKZF1 alterations carrying a chemotherapy-exposure signature 4 - i.e. the relapse genotype is partly a product of the treatment, and in the worst prognostic group, very early relapse, it is absent altogether (9% versus 79%). There is no genotype to read.
4. The negative cuts against the graph's own optimism too. 'One target, no second driver' is usually written here as a reason menin inhibition should work well. It is equally a reason that when menin inhibition fails there will be no genomically indicated partner - and the two documented resistance findings in KMT2A::AFF1 point at KMT2C/D and at unexplained upfront resistance in pretreated samples, with emergent RAS and TP53 comutations explicitly not sufficient to cause resistance in vitro 6.
| claim type | load-bearing negative |
|---|---|
| statement | Infant KMT2A-rearranged ALL has no recurrent clonal second driver, so the genotype-selected combination logic used in other leukaemias has no substrate; combinations here must be chosen on mechanism, not on co-mutation status. |
druggability inventory
[
{
"lesion": "KRAS mutation",
"frequency": "4/22 WGS, subclonal in ~2/3",
"druggable": "indirect (MEK)",
"blocker": "subclonal target; ex-vivo MEK sensitivity does not track RAS genotype"
},
{
"lesion": "NRAS mutation",
"frequency": "2/22 WGS; resistant clones 0.06-0.34%",
"druggable": "indirect (MEK)",
"blocker": "reservoir alleles G13D/G12S are outside mutant-selective RAS chemistry"
},
{
"lesion": "PTPN11 mutation",
"frequency": "non-recurrent",
"druggable": "yes (SHP2)",
"blocker": "no open paediatric route; leukaemia protocol withdrawn for a sponsor decision"
},
{
"lesion": "NF1 loss",
"frequency": "1/22 homozygous deletion",
"druggable": "no",
"blocker": "tumour suppressor"
},
{
"lesion": "PAX5 focal loss",
"frequency": "5/22 (23%) - commonest recurrent second lesion",
"druggable": "no",
"blocker": "monoallelic transcription-factor loss"
},
{
"lesion": "CDKN2A/B deletion",
"frequency": "4/47 (8.5%)",
"druggable": "in principle (CDK4/6)",
"blocker": "under one in ten patients, clonality unmeasured, untested in this disease"
},
{
"lesion": "TP53 / IKZF1 / DNMT3A",
"frequency": "not reported at diagnosis",
"druggable": "no",
"blocker": "absent, and undruggable if present"
}
]tested once
{
"combination": "revumenib + selumetinib in KMT2A-r RAS-mutant AML",
"in_vitro": "synergistic",
"in_vivo_burden": "reduced",
"in_vivo_survival": "not improved over menin inhibition alone",
"source_pmid": "42085603"
}- The absence of a second driver is usually presented as favourable for menin inhibition. It is equally a statement that when menin inhibition fails there is no genomically indicated partner to add, and the two published resistance investigations in KMT2A::AFF1 point at epigenomic changes rather than at any mutation.
- The one co-targeting experiment run in the setting most favourable to it - clonal RAS mutations in KMT2A-rearranged AML - reduced burden and did not extend survival. A strategy that fails where conditions are best should not be assumed to work where they are worse.
- A genotype-selected combination trial in this disease would be selecting on lesions present in 8-25% of patients, partly subclonal, in a disease of a few hundred cases a year worldwide. The trial cannot be built even if the biology were right.
- If combinations must be chosen on mechanism, the selection burden moves from the patient's genome to the strength of the preclinical model - and this graph already documents that the index model for this disease, MLL-AF4 in cord-blood CD34+ cells, does not make leukaemia.
- Is there a mechanism-chosen combination whose benefit has been shown to be independent of co-mutation status in this disease? MEK inhibition plus prednisolone synergised in RAS-wild-type cells, which is the shape of such a result and has never been followed up.
- Does the foreclosure lift at relapse, where TP53 and IKZF1 alterations are present in 79% of early relapses? If so, the combination window is at relapse and the targets are undruggable ones, which is a worse answer than it first appears.
- What would a combination trial in this disease stratify on, if not genotype? Ex-vivo drug sensitivity is the obvious candidate and the infrastructure for it exists; no protocol in this disease uses it as an allocation variable.
- Would an adult KMT2A-rearranged AML trial of a genotype-selected combination succeed where an infant ALL one cannot, purely because the co-mutations there are clonal? That is a testable difference between two diseases sharing a fusion.
- E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
Infant MLL-rearranged ALL carries a mean of 1.3 non-silent mutations in the predominant leukaemic clone, an average of 1.0 copy-number alteration per case whose commonest recurrent targets are PAX5 (5 of 22 cases) and CDKN2A/CDKN2B (3 of 22, 4/47 overall), and 65% of its activating kinase/PI3K/RAS mutations sit below 30% mutant allele frequency with pathway mutations frequently absent at relapse.
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
All figures conditioned as described in mechanism-low-mutation-burden-is-depth-bounded: 39x WGS, dominant-clone conditioning, point mutations counted separately from copy number, 22-case discovery cohort. The inventory this entry builds is therefore an inventory of what one study plus its validation arm could see. Full text read via efetch db=pmc (PMC4553269).
Evidence that argues against this (1)
- Duplex sequencing finds 62% of cancer-associated mutations in this disease below standard NGS resolution, so the inventory is a clonal inventory. It does not change the argument - a sub-1% lesion is a worse combination target than a subclonal one, not a better one - but it means 'the disease has almost no mutations' is the wrong premise even though the conclusion holds. Pilheden M et al. 2022
- E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
In 124 uniformly treated infant B-ALL cases, recurrent mutations were found exclusively in K-RAS and N-RAS, were subclonal, and were frequently lost at relapse, within the most silent mutational landscape reported for any sequenced paediatric cancer.
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
Abstract read. 'Exclusively' is a statement about which genes recurred; the same abstract reports a larger number of non-recurrent non-silent mutations, none of which constitutes a target because a non-recurrent lesion cannot support a stratified trial. Sequencing depth not given.
Evidence that argues against this (1)
- A non-recurrent mutation can still be a target in a genuinely personalised setting - one patient, one drug - so 'no recurrent second driver' forecloses stratified combination trials specifically, not individualised treatment. Agraz-Doblas A et al. 2019
- E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
Across nearly 1,000 COG paediatric AML participants, mutations in DNMT3A and TP53, common in adults, were conspicuously absent from virtually all paediatric cases.
The molecular landscape of pediatric acute myeloid leukemia reveals recurrent structural alterations and age-specific mutational interactions. · Bolouri H, Farrar JE, Triche T, et al. · Nature Medicine 2018What this evidence cannot show
Paediatric AML, not infant ALL, and the authors' qualitative wording rather than a count. It is used here to support the claim that the adult co-mutation set does not transfer to children, which is the form of the statement it supports. Abstract read.
Evidence that argues against this (1)
- The same study found NRAS, KRAS, FLT3, CBL, GATA2 and WT1 mutations frequent in paediatric AML, so paediatric leukaemia in general is not co-mutation-poor; the foreclosure argued here is specific to infant KMT2A-rearranged ALL and should not be generalised to paediatric AML including its KMT2A-rearranged subset. Bolouri H et al. 2018
- E4strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
In relapsed infant and childhood KMT2A-rearranged acute leukaemia, alterations in drug-response genes, most commonly TP53 and IKZF1, were enriched in early relapse ALL (79%) with a chemotherapy-exposure mutational signature, but present in only 9% of very early relapse ALL and absent as a signature from AML relapse.
The genomic landscape of relapsed infant and childhood KMT2A-rearranged acute leukemia. · Ahlgren L, Pilheden M, Sturesson H, et al. · Nature Communications 2025What this evidence cannot show
36 relapsed patients split into timing subgroups, so subgroup counts are small. The category is the authors' curated 'drug-response genes' rather than TP53 and IKZF1 alone. Abstract read; full text not retrieved for this entry.
Evidence that argues against this (1)
- That TP53 and IKZF1 alterations ARE present in 79% of early relapse means the relapsed disease does eventually acquire co-lesions, so the foreclosure argued in this entry applies to diagnosis and weakens at relapse - which is the moment a combination would most likely be used. Ahlgren L et al. 2025
- E5moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper
Combining the MEK1/2 inhibitor selumetinib with the menin inhibitor revumenib in KMT2A-rearranged RAS-pathway-mutant AML produced synergistic loss of viability, cell-cycle arrest, apoptosis and Myc-target downregulation in cell lines and PDX-derived cells, and reduced leukaemia burden 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, not infant ALL - and the RAS lesions in AML are more often clonal, so this is the combination logic tested in the setting MOST favourable to it, which is why its survival result carries the weight it does here. A negative survival result in PDX models is not a clinical result and the authors frame it as requiring optimisation, not as a refutation. Abstract read; full text not retrieved for this entry.
Evidence that argues against this (1)
- A burden reduction without a survival benefit can reflect dosing, schedule or model limitations rather than the strategy, and the authors explicitly propose optimisation rather than abandonment. Scheidegger N et al. 2026
- E6moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper
In patient-derived xenografts from two patients with multiply relapsed KMT2A::AFF1 ALL and no prior menin-inhibitor exposure, later samples were markedly less menin-inhibitor sensitive than earlier samples from the same patients despite sustained on-target transcriptional effect, and emergent RAS-pathway and TP53 comutations were individually not sufficient to induce menin-inhibitor resistance in vitro, with KMT2C/D downregulation implicated instead in one patient.
Upfront menin-inhibitor resistance in multiply pretreated leukemias. · Mahdavi L, Alikarami F, Goodrow H, et al. · Experimental Hematology 2026What this evidence cannot show
Two patients for the serial-resistance arm. The KMT2C/D mechanism is offered for one patient. 'Not sufficient in vitro' is a negative result in an engineered system. The paper's practical recommendation - use menin inhibitors early, before genomic and epigenomic evolution - is a reasonable inference from a small series and is not a trial result. Abstract read; full text not retrieved for this entry.
Evidence that argues against this (1)
- If resistance in pretreated disease is epigenomic rather than genomic, then the absence of a cooperating lesion to co-target is not the limiting problem it appears here - the limiting problem is that the relevant state is not readable from a genotype at all. Mahdavi L et al. 2026
the therapeutic negative rests on the audited burden claim, not the unqualified one
no genomically indicated combination partner exists when menin inhibition fails in this disease
the load-bearing negative: genotype-selected combination logic has no substrate here
the mechanism-over-genotype conclusion follows from this risk