Oncofusion

start / graph / disease-kmt2a-r-aml

Diseasestrong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.Q7-clinical

KMT2A-rearranged acute myeloid leukaemia in children and infants

also called 11q23/MLL-rearranged AML · MLL-rearranged AML · KMT2A-r AML

The same broken KMT2A gene that causes leukaemia of the B-cell lineage in babies can instead produce myeloid leukaemia, which fills the marrow with a different kind of immature cell. Which partner gene KMT2A joins matters enormously here: across children with 11q23-rearranged myeloid leukaemia, five-year event-free survival ranges from about 11% to about 92% depending on the partner.

The same thing, for a scientist

KMT2A rearrangement is the defining lesion in roughly 40-60% of infant AML and ~15-20% of paediatric AML overall, and forms its own WHO/ICC entity. Unlike KMT2A-rearranged ALL, the partner gene is strongly prognostic: an international retrospective study of 756 children with 11q23/KMT2A-rearranged AML found 5-year EFS of 44% overall but subgroup EFS spanning 11 +/- 5% to 92 +/- 5%, with t(1;11), t(6;11), t(10;11)(p12;q23) and t(10;11)(p11.2;q23) independent adverse predictors. Partner usage differs from ALL: MLLT3 (t(9;11)), MLLT10 and ELL dominate rather than AFF1. Counterintuitively, within KMT2A-rearranged AML infants do better than older children, the reverse of the pattern in KMT2A-germline AML and the reverse of the pattern in ALL. RAS-pathway co-mutation is common (KRAS ~20-26%) and adverse.

Go deeper
Same broken gene, different disease1 deeper

KMT2A-rearranged AML shares the mechanism of KMT2A-rearranged ALL - a menin-dependent chromatin complex locking in a HOXA9/MEIS1 self-renewal programme - but the clinical object is different in four ways.

KMT2A-r infant ALLKMT2A-r AML
Dominant partnerAFF1 (~48-53%)MLLT3 (t(9;11)), then MLLT10, ELL 14
Does the partner change prognosis?Modestly and inconsistentlyYes, decisively: 5-year EFS 11% to 92% across subgroups 1
Age effectYounger is much worseWithin KMT2A-r AML, infants do better than older children (EFS p = 0.049, OS p = 0.032) 2
Typical co-mutationRAS pathway, FLT3RAS pathway; KRAS 20.3% in infants, 26.1% in children, adverse 2

The partner-dependence is the most important clinical fact on this page, and it is the reason the international group concluded that 'screening for these translocation partners is needed for accurate treatment stratification at diagnosis' 1.

Why the age effect inverts

In a 2025 cohort of 239 paediatric KMT2A-rearranged AML cases (59 infants, 180 older children), infants had significantly better EFS and OS than older children - while in the KMT2A-germline AML comparison group of the same study infants did significantly worse 2. Part of the explanation is composition: KMT2A::MLLT4/AFDN, which carries a poor prognosis, was detected only in older children, and KMT2A::MLLT3 was also more frequent in children than infants 2. This is a warning against reasoning from 'infant leukaemia is worse' as a general law; the age effect is a property of a specific lesion in a specific lineage, not of infancy.

What it looks like at the bedside1 deeper

Presentation overlaps heavily with infant ALL - marrow failure (pallor, bruising, infection), organomegaly, high counts - but with a myeloid accent. Extramedullary disease is more prominent: myeloid sarcoma, gum and skin infiltration, and in infants leukaemia cutis. In the congenital/neonatal leukaemia literature, which is predominantly myeloid, cutaneous infiltrates were reported in about two thirds of cases (67.8%), hepatosplenomegaly in 47.5%, CSF blasts in 25.4%, hyperleukocytosis in 62.7% with a mean WBC of 68.5 x 10^9/L, and spontaneous remission in 11.9% 3. That last figure has no counterpart in infant ALL and is a real, if rare, phenomenon in neonatal myeloid disease.

Diagnostic consequence

Because lineage assignment can be ambiguous at presentation - myeloid antigens on a B-lineage blast, or a mixed phenotype - the diagnostic pathway is the same one described in concept-flow-cytometry-immunophenotype-kmt2a-r and concept-fish-split-signal-kmt2a, and split-signal FISH answers the question 'is KMT2A broken?' independently of the lineage call. Given the partner-dependence of prognosis in AML, the partner must then be identified, not just the break - which is the job of concept-molecular-confirmation-fusion-transcript.

Why this matters to a menin-directed programme

KMT2A-rearranged AML, together with NPM1-mutant AML, is where menin inhibition was first shown to work clinically, and it is a much larger population than infant ALL. A platform argument that starts from 160 US infants a year becomes tractable when the same mechanism covers paediatric and adult KMT2A-rearranged AML as well. The counterweight is that partner-dependent biology is real here (5-year EFS 11% to 92% across partners 1), so a single menin inhibitor may not behave identically across partners, and RAS-pathway co-mutation - present in a fifth to a quarter of cases 2 - offers an obvious escape route that does not require touching the menin-KMT2A interface at all.

Facts and figures
icd10C92.0
who entityAML with KMT2A rearrangement
kmt2a fraction infant aml PCT
[
  40,
  60
]
dominant partners
[
  "MLLT3 (t(9;11)(p21;q23))",
  "MLLT10",
  "ELL",
  "MLLT1",
  "AFDN/MLLT4 (older children only)"
]
survival
[
  {
    "cohort": "International 11q23/KMT2A-r paediatric AML (n=756)",
    "years": 5,
    "efs_pct": 44,
    "subgroup_efs_range_pct": [
      11,
      92
    ]
  },
  {
    "cohort": "Paediatric KMT2A-r AML, infants vs children (n=239)",
    "note": "infants significantly better EFS (p=0.049) and OS (p=0.032) than older children"
  }
]
adverse partners
[
  "t(1;11)",
  "t(6;11)",
  "t(10;11)(p12;q23)",
  "t(10;11)(p11.2;q23)",
  "KMT2A::MLLT4/AFDN"
]
co mutations
{
  "KRAS_infants_pct": 20.3,
  "KRAS_children_pct": 26.1,
  "prognostic": "adverse in childhood KMT2A-r AML"
}
extramedullary
[
  "myeloid sarcoma",
  "leukaemia cutis",
  "gum infiltration",
  "CNS"
]
How this could bite us back
  • Partner-dependent prognosis means a single menin inhibitor tested in a pooled KMT2A-rearranged AML population could succeed or fail depending on which partners happened to enrol.
  • RAS-pathway co-mutation gives the cell a proliferative input that does not run through the menin-KMT2A interface, so it is an escape route available at diagnosis rather than one that has to be acquired.
  • Lineage plasticity runs in both directions: a myeloid-phenotype KMT2A-rearranged leukaemia can arise as a switch from a B-lineage disease under CD19-directed therapy, so a 'KMT2A-r AML' diagnosis may be a treated ALL in disguise, with different prior therapy exposure and different resistance.
  • Anthracycline-heavy AML backbones mean that any survivor of KMT2A-r AML carries a higher cardiac late-effect burden than an ALL survivor, raising the bar for what a replacement therapy must achieve.
Still open
  1. Why does the KMT2A partner determine prognosis so strongly in AML (5-year EFS 11% to 92%) but much more weakly in infant ALL? Is the partner's own transcriptional contribution more consequential in a myeloid chromatin context, or is the ALL signal masked by the dominant age effect?
  2. Do menin inhibitors show partner-dependent depth or durability of response in KMT2A-rearranged AML, and if so does that track the Balgobind prognostic subgroups?
  3. RAS-pathway mutation is present in a fifth to a quarter of KMT2A-rearranged AML; does it confer primary resistance to menin inhibition, and does adding MEK inhibition restore sensitivity in patients as it appears to preclinically?
Every source behind this entry (4)
  1. E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    In an international retrospective study of 756 children with 11q23/KMT2A-rearranged AML, 5-year event-free survival was 44% overall but ranged from 11 +/- 5% to 92 +/- 5% across translocation-partner subgroups, and t(1;11), t(6;11), t(10;11)(p12;q23) and t(10;11)(p11.2;q23) were independent prognostic predictors, leading the authors to conclude that partner screening is needed for stratification at diagnosis.

    PMID 19528532DOI
    What this evidence cannot show

    Retrospective, pooled across many national groups and treatment eras, so subgroup survival reflects heterogeneous therapy as well as biology. The extreme subgroups are small, which is why their confidence bounds are wide; a 92 +/- 5% figure resting on few patients should not be read as a reliable point estimate. The study predates current risk-adapted paediatric AML protocols and menin inhibitors.

    Evidence that argues against this (1)
    • A range from 11 +/- 5% to 92 +/- 5% across partner subgroups in a retrospective international series is the signature of small per-partner numbers, and the extremes are the groups most likely to regress toward the mean when the study is repeated. The series also spans decades and many national protocols, so partner subgroup and treatment era are confounded. Blood Cells, Molecules and Diseases 2010
  2. E2moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    In 239 paediatric KMT2A-rearranged AML cases (59 infants, 24.7%; 180 children, 75.3%), infants had significantly better EFS and OS than older children (p = 0.049 and p = 0.032), the opposite of the pattern in the 538 KMT2A-germline cases; KMT2A::MLLT4 was detected only in children and carried a poor prognosis; KMT2A::MLLT3 and KMT2A::MLLT4 were more frequent in children than infants; KRAS mutation was present in 20.3% of infants and 26.1% of children and was adverse in childhood KMT2A-r AML; and age, WBC count, KMT2A::MLLT4 and KRAS mutation were independent adverse factors.

    PMID 41126738
    What this evidence cannot show

    A retrospective genomic cohort; the infant subgroup is 59 patients, so the age comparison rests on borderline p values (0.049, 0.032) that would not survive multiple-testing correction. Partner composition differs between the age groups, so the age effect and the partner effect are partly confounded and the paper's own multivariate model includes both.

    Evidence that argues against this (1)
    • The direction of the age effect here is opposite to that in KMT2A-rearranged ALL, where infancy and especially age under 90 days is strongly adverse, so neither result should be generalised across lineages. source 2021
  3. E3moderate confidence 3 of 5: One strong study, or several smaller consistent ones.book

    KMT2A rearrangement is present in approximately 40-60% of infants with AML; infant leukaemias account for about 10% of paediatric AML; in reported congenital leukaemia, cutaneous infiltrates occurred in 67.8%, hepatosplenomegaly in 47.5%, CSF blasts in 25.4%, hyperleukocytosis in 62.7% with mean WBC 68.5 x 10^9/L, anaemia in 35.7% and spontaneous remission in 11.9%; and KMT2A partner usage in AML is MLLT3 22%, MLLT10 27% and ELL 17%.

    Infant Leukemia (StatPearls) · StatPearls Publishing · NCBI Bookshelf 2024
    What this evidence cannot show

    A tertiary educational source, not a primary cohort; the congenital-leukaemia percentages come from pooled published case reports, which are subject to severe publication bias toward striking presentations (leukaemia cutis, spontaneous remission), so they overstate the frequency of visible features. Use them for the shape of neonatal myeloid presentation, not as population frequencies. Partner percentages differ from other series.

    Evidence that argues against this (1)
    • The congenital-leukaemia literature is a collection of individually reported cases, not a cohort. Roughly 175 cases assembled from the world literature over decades carries publication bias toward the visually striking and the unusual, so every frequency derived from that denominator is a frequency among CASES SOMEONE CHOSE TO PUBLISH. The cutaneous figure quoted here, 67.8%, disagrees with the 25-30% that this graph's own leukaemia-cutis entry establishes from a review of the same literature. Both cannot be right, and the discrepancy is the measure of how unstable frequencies from that source are. Blood Cells, Molecules and Diseases 2010
  4. E4strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    KMT2A partner frequencies differ between lineages: in infant ALL the dominant partner is AFF1 (~48-53%), whereas paediatric KMT2A-rearranged AML is dominated by MLLT3, MLLT10 and ELL.

    PMID 38634141DOI
    What this evidence cannot show

    This citation establishes the ALL side of the comparison directly (AFF1 53.3% of 60 typed KMT2A-r infants) and the AML side only by contrast with the AML series cited at [E1] and [E3]; the two sides come from different cohorts and eras, so the comparison is indicative rather than a matched analysis.

    Evidence that argues against this (1)
    • Partner frequencies differ between lineages and they also differ between ASSAYS. Infant ALL cohorts have been characterised largely by targeted RT-PCR panels built around AFF1, while paediatric AML has more often been karyotyped and more recently RNA-sequenced; a panel that asks about AFF1 first will report AFF1 most. How much of the lineage difference survives a single uniform assay is an open question. Leukemia 2001
Connected to