The near-silent genome of infant KMT2A-r ALL — 1.3 non-silent mutations, and why that is bad news as well as good

Two independent cohorts, sequenced by different consortia on different protocols, agree.

St. Jude/Washington University PCGP 1. Paired-end whole-genome sequencing of 22 infants with KMT2A-r ALL plus matched remission material (average haploid coverage 39x), extended by targeted sequencing of 232 genes and SNP arrays in a validation cohort of 43 further infant leukemias (25 KMT2A-r) and 18 infant non-KMT2A-r cases, plus whole-exome sequencing of 20 non-infant KMT2A-r leukemias aged 7-19 years.

• Across the whole genome: a mean of 111 somatic sequence mutations per case (range 29-229) and 10 CNAs/SVs per case (range 2-26).
• Restricted to coding regions and regulatory RNAs: 8.2 events per case, of which 2.2 non-silent SNVs (range 0-4) and 6.0 CNA/SVs (range 2-19).
• Restricted to the dominant leukemic clone: a mean of 1.3 non-silent mutations per case, of which only 0.6 are expressed at the RNA level.

That last figure is the one worth remembering, because it is the number of coding lesions actually present in every leukemic cell. Only pediatric low-grade glioma — a clinically indolent tumour — has a lower somatic coding mutation count among sequenced cancers; the rate here is 2- to 180-fold lower than reported adult and pediatric cancers 1.

Interfant-99/06 3. Multi-layered genome-wide analysis of 124 infant B-ALL cases uniformly diagnosed and treated on a single protocol found an average of one genomic rearrangement and 2.5 non-silent single-nucleotide variants per case, described by the authors as the most silent mutational landscape reported so far for any sequenced pediatric cancer.

A three-patient whole-genome pilot 2 reached the same conclusion first, from a different direction: whole-genome sequencing of diagnostic blasts and matched germline from three infant KMT2A::AFF1 pro-B ALL cases found very few somatic changes of any class — copy-number, loss of heterozygosity or SNV — and concluded that only a very small number of mutations are needed to generate infant KMT2A-r leukemia.

In favour of the fusion-directed strategy. A single clonal driver with 1.3 accompanying coding lesions means: (i) the fusion is not one hit among many but the hit, so fusion-directed agents are attacking the actual dependency rather than a passenger; (ii) the pre-existing pool of resistance mutations is small, because there simply are not many mutations to draw from; (iii) there is no high background mutation rate to generate new resistance quickly, since the mutation spectrum showed no signature of a specific mutational mechanism (the commonest change being ordinary C>T/G>A transition) 1. The paper's own conclusion is explicit: because the fusion is a potent driver and the cells lack a high mutational rate, therapy directed at the fusion protein or its required cofactors is the rational approach.

Against. The same silence removes the standard oncology playbook.

• No co-target. There is no second recurrent driver to hit in combination. The only frequent co-lesion class is kinase/PI3K/RAS signalling (47% of cases), and it is largely subclonal and frequently lost at relapse, which is why the PCGP authors concluded that specifically targeting those lesions will likely provide little therapeutic benefit 1. See comut-ras-pathway-kmt2a-r.
• No synthetic-lethal partner. Synthetic lethality needs a loss-of-function lesion to exploit. Infant KMT2A-r ALL has essentially none: no recurrent TP53, no recurrent CDKN2A/B (4/47, 8.5%), no recurrent IKZF1, no PTEN 1, 3. Any vulnerability in this disease has to be a fusion-created dependency, not an inherited genetic scar. See cnv-copy-number-landscape-kmt2a-r-all.
• No mutational-burden-driven immunogenicity. A genome with 1.3 non-silent mutations generates almost no neoantigens. Whatever makes this disease responsive to CD19-directed immunotherapy, it is not tumour mutational burden.
• Biomarker vacuum. With nothing recurrent to sequence beyond the fusion itself, risk stratification in Interfant-06 still runs on age, white-cell count, prednisone response and fusion partner rather than on any co-mutation 4.

Infant KMT2A-r ALL is simultaneously one of the least mutated and one of the most refractory cancers in children. The Interfant-06 international phase III study of 651 infants reported 6-year event-free survival of 46.1% (SE 2.1) and overall survival of 58.2% (SE 2.0); in the high-risk group 6-year EFS was 20.9% (SE 3.4), and KMT2A rearrangement was the single strongest prognostic factor 4. Relapse is also fast: most relapses occur during treatment, two thirds within one year and 90% within two years of diagnosis 7.

A mutation-accumulation model cannot produce that. Three candidate explanations are on the table, and the evidence is unequal between them.

1. Epigenetic and transcriptional plasticity (best supported). Integrated genome-wide DNA methylome and transcriptome analysis of 69 de novo infant B-ALL cases treated on Interfant-99/06 found methylome signatures concentrated in chromatin states related to enhancer and transcriptional control, with coordinated hypomethylation and over-expression of the AP-1 members FOS and JUN and of RUNX factors specifically in KMT2A-r cases. Pharmacological inhibition or functional ablation of AP-1 dramatically impaired KMT2A-r leukemic growth in vitro and in patient-derived xenografts 5. This is the most direct evidence that the state space this leukemia can move through is epigenetic rather than genetic, and the authors frame it exactly that way: as predicted by its prenatal origin, the silent DNA landscape implies alternative epigenetic mechanisms substantially contribute to leukemogenesis.

2. Subclonal RAS escape (real but partly refuted as a relapse mechanism). 65% (20/31) of the activating kinase/PI3K/RAS mutations sat at MAF below 30%, and in seven matched diagnosis-relapse pairs the mutation was maintained or increased in two, decreased in one (MAF 39% to 15%) and lost in two 1. The Interfant cohort concurs: two thirds of relapsing patients lost the diagnostic RAS mutation 3. So subclonal RAS is a plausible source of clonal diversity at diagnosis, but it is demonstrably not the clone that usually comes back. Clonal-evolution analysis in two PCGP patients who relapsed three years out showed relapse arising from a minor diagnostic clone present at ~6.25%, which then acquired further mutations under treatment 1 — the mechanism is selection of a rare pre-existing clone, not acquisition of a new driver.

3. Lineage switch (rare, mechanistically important, and increasingly visible). Under lineage-directed pressure the same clone can change lineage rather than mutate. In 231 children and young adults receiving CD19-directed CAR T-cell therapy, KMT2A-r infant ALL (n = 13) achieved 2-year relapse-free survival of 67% and overall survival of 62% with no increased hazard of relapse overall (HR 0.70, 95% CI 0.21-2.90), but a higher proportion of relapses associated with myeloid lineage switch and a 3.6-fold increased risk of all-cause death (95% CI 1.04-12.75, P = 0.0434) 6. A documented single case in a t(1;11) infant traced the identical three dominant IGH/IGL clonal sequences through the original B-ALL, the lineage-switched AML and the post-transplant B-ALL relapse — i.e. reprogramming of one clone, not emergence of a genetically distinct one 8.

All three candidates share a feature: they are ways of changing phenotype without changing genotype. That is precisely what a near-silent genome forces a cancer to do, and it is the argument for why this disease is an epigenetics problem dressed as a genetics problem.

Querying the public curation of the same PCGP study (cBioPortal study all_stjude_2015, 93 sequenced samples, API retrieved 2026-09-18) returns, across all samples, the following variant counts in the kinase/PI3K/RAS genes the paper names 9: KRAS 14 mutations in 12 samples; NRAS 8 in 8; FLT3 5 in 5; PIK3CA 4 in 4; PIK3R1 3 in 3; NF1 1 in 1; PTPN11 0. The structural-variant profile contains 13 distinct KMT2A partner combinations, including KRAS::KMT2A and RAD51B::KMT2A, matching the two reciprocal in-frame products reported in the paper.

This is a useful sanity check with one caveat worth flagging: the curated variant table does not contain a PTPN11 mutation, although the paper's text names PTPN11 among the non-recurrent pathway genes in both the infant and non-infant cohorts. Whichever is right, PTPN11 in this disease rests on at most one or two cases. See comut-ptpn11-kmt2a-r.