DNMT3A - an independent adverse factor in adult KMT2A-rearranged AML and conspicuously absent from paediatric disease, which is the whole finding

This is a prediction, not a measurement. No one has seen this protein in this shape. The colours say how much the software trusts itself, and that is worth reading before anything else — the orange stretches are places the model is effectively saying this part has no fixed shape.

• The DNMT3A literature in KMT2A-rearranged leukaemia is large and almost entirely adult. Any preclinical model or trial rationale that cites it for a paediatric indication is importing a lesion that is not there.
• Hypomethylating agents are the reflexive response to a methylation-writer story, and in this disease they would be given to patients with an intact DNMT3A, on a rationale that the gene-level evidence does not support.
• An age-split of this shape is a warning about every other co-mutation claim carried into this disease from adult AML, including the ones this graph holds: if DNMT3A and TP53 both fail to transfer, the prior for any adult co-mutation transferring should be low.

1. Is the DNMT3A prognostic association in adult KMT2A-rearranged AML independent of clonal haematopoiesis? Separating them needs variant allele fractions and remission-sample genotypes, and the HARMONY abstract reports neither.
2. What are the 45% of epigenetic-regulator mutations in older children with KMT2A-rearranged leukaemia, if not DNMT3A? The index study reports the category rate and this entry could not reach the gene-level breakdown without its supplementary tables.
3. Does the infant disease achieve, epigenetically, what an adult leukaemia achieves by mutating DNMT3A? This graph already holds methylome work in infant KMT2A-rearranged B-ALL, and the two have never been placed side by side.
4. Would a DNMT3A-mutant background alter menin-inhibitor sensitivity? Untested in any age group, and relevant because menin inhibitors are being moved into adult KMT2A-rearranged AML where DNMT3A mutation is present.

1. E1moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

   Among 185 adult KMT2A-rearranged AML patients with panel-based NGS from the HARMONY database, the independent prognostic factors for overall survival in intensively treated patients were age over 60 (HR 2.1, P = 0.001), secondary AML (HR 2.2, P = 0.043), DNMT3A mutation (HR 2.1, P = 0.047) and KRAS mutation (HR 2.0, P = 0.005).

   Rearrangements involving 11q23.3/KMT2A in adult AML: mutational landscape and prognostic implications - a HARMONY study. · Hernandez-Sanchez A, Gonzalez T, Sobas M, et al. · Leukemia 2024
   PMID 38965370DOI
   ▸What this evidence cannot show

   P = 0.047 for DNMT3A is borderline in a multivariable model with several covariates and no stated correction for multiple comparisons, so this is the weakest of the four factors listed. A registry cohort assembled across trials, adults only, AML only. The DNMT3A mutation frequency in the cohort is not given in the abstract, so this entry states the prognostic association without a frequency. Abstract read; full text not retrieved.

   ▸Evidence that argues against this (1)

   • DNMT3A mutation in adults is strongly associated with age-related clonal haematopoiesis, so an adverse prognostic association in an adult AML cohort may partly index the patient's haematopoietic age rather than a property of the leukaemia - and the same model independently found age over 60 adverse. Bolouri H et al. 2018
2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

   Across nearly 1,000 Children's Oncology Group paediatric AML participants assessed by whole-genome, targeted DNA, mRNA and microRNA sequencing and CpG methylation profiling, mutations in DNMT3A and TP53, which are common in adults, were conspicuously absent from virtually all paediatric cases, while mutations in GATA2, FLT3, CBL, MYC-ITD, NRAS, KRAS and WT1 were frequent.

   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 2018
   PMID 29227476DOI
   ▸What this evidence cannot show

   Paediatric AML as a whole rather than the KMT2A-rearranged subset, so the statement is about the age group and not specifically about this fusion class. 'Virtually all' and 'conspicuously absent' are the authors' qualifiers, not a stated count; this entry does not convert them into a number. Trial-enrolled patients only. Abstract read; full text not retrieved for this entry.

   ▸Evidence that argues against this (1)

   • This is an AML cohort, and the disease this graph centres on is infant ALL, so the strongest paediatric evidence for DNMT3A's absence comes from a different lineage; the ALL-side evidence is weaker, being silence in a single index study. Andersson AK et al. 2015
3. E3moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

   DNMT3A is not mentioned anywhere in the full text of the index whole-genome sequencing study of infant MLL-rearranged ALL, which comprised 22 infants by WGS and 47 infant MLL-rearranged cases in total; the same study reports that MLL-rearranged leukaemia in OLDER children had frequent mutations (45%) in epigenetic regulators, a category rarely mutated in infant MLL-rearranged ALL.

   The landscape of somatic mutations in infant MLL-rearranged acute lymphoblastic leukemias. · Andersson AK, Ma J, Wang J, et al. · Nature Genetics 2015
   PMID 25730765DOI
   ▸What this evidence cannot show

   A bounded absence from one study and an absence of reporting rather than a reported zero. Retrieved via efetch db=pmc for PMC4553269 on 2026-09-20, tags stripped before entity unescaping, 71,491 characters; the case-insensitive count for 'DNMT3A' was zero. The epigenetic-regulator observation is the more informative half: the study explicitly contrasts 45% epigenetic-regulator mutation in older children against rarity in infants, which places DNMT3A's absence inside a category-level finding rather than leaving it as pure silence.

   ▸Evidence that argues against this (1)

   • The 45%-in-older-children figure means epigenetic regulators ARE a real co-mutation class in KMT2A-rearranged leukaemia once the patient is old enough, so 'this disease does not co-mutate epigenetic regulators' is true only of the infant form and is routinely overstated. Andersson AK et al. 2015
4. E4moderate confidence 3 of 5: One strong study, or several smaller consistent ones.database

   A PubMed search for DNMT3A restricted to KMT2A-rearranged or MLL-rearranged disease and to infant, paediatric or childhood populations returns two records, neither of which is a co-mutation study: one on mesothelin expression in acute leukaemia and one review of epigenetic modifiers in normal and malignant haematopoiesis; the unrestricted query returns 138.

   PubMed E-utilities esearch, queries: (MLL OR KMT2A) AND DNMT3A AND (leukemia OR leukaemia); DNMT3A[tiab] AND (MLL OR KMT2A) AND (rearrang* OR translocation); DNMT3A AND (KMT2A-rearranged OR MLL-rearranged) AND (infant OR pediatric OR paediatric OR childhood) · National Library of Medicine · PubMed 2026
   ▸What this evidence cannot show

   Run 2026-09-20 with querytranslation inspected. A counted absence is only as good as the symbol set: DNMT3A has no widely used retired symbol, which helps, but the restriction terms 'KMT2A-rearranged' and 'MLL-rearranged' are hyphenated phrases and a paper describing the same patients as 't(4;11)' or '11q23' would not be retrieved. This is recorded as a bounded absence, not as proof that no such study exists.

   ▸Evidence that argues against this (1)

   • The narrower query's restriction terms are the failure mode this project has been bitten by before: a paper framing its cohort at the karyotype level rather than the gene level is unreachable from a gene-symbol query, and the pathway-or-complex-form check that guards against this has no analogue for a disease-name restriction. National Library of Medicine 2026
5. E_q93_dnmt3a_loc1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

   UniProt Q9Y6K1 carries Nucleus (ECO:0000269 PubMed:12138111, 12145218), Chromosome (ECO:0000269 PubMed:12138111) and Cytoplasm (ECO:0000269 PubMed:12145218), plus an UNCODED note that DNMT3A accumulates in the major satellite repeats at pericentric heterochromatin, and a tissue comment naming fetal tissues and peripheral blood mononuclear cells among its highest-expressing tissues.

   UniProtKB Q9Y6K1 (DNM3A_HUMAN), SUBCELLULAR LOCATION comment with evidence codes, sources and the uncoded note, plus TISSUE SPECIFICITY, fetched live 2026-09-21 · UniProt Consortium · Databases 2026
   ▸What this evidence cannot show

   A statement about the record. The provenance is inverted relative to usefulness: the specific and informative statement (major satellite repeats at pericentric heterochromatin) is uncoded and carries no source, while the generic values (Nucleus, Chromosome, Cytoplasm) are coded. An ECO:0000269 filter therefore keeps the uninformative lines - the exact failure brief rule 91 documents for KMT2A's fragment comments.

   ▸Evidence that argues against this (1)

   • Reading the flat record as simply deficient misses that it may be reporting a genuine mixture rather than failing to resolve one. If full-length DNMT3A is pericentric and DNMT3A2 is euchromatic, and most cells express some of both from two promoters, then a single antibody in a single cell type really does see nucleus plus chromosome plus a variable cytoplasmic fraction - and the honest single-line summary of that is close to what UniProt says. The defect is not that the values are wrong but that nothing on the record tells a reader the two species have OPPOSITE chromatin preferences, which is the fact that changes what any experiment would mean. Chen T et al. 2002
6. E_q93_dnmt3a_loc2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

   DNMT3A2, a small form of DNMT3A initiated from a downstream intronic promoter and lacking the N-terminal 223 (human) or 219 (mouse) residues, has similar cytosine methyltransferase activity to DNMT3A in vitro but a STRIKINGLY DIFFERENT subcellular localisation: full-length DNMT3A is CONCENTRATED ON HETEROCHROMATIN while DNMT3A2 displays a pattern suggestive of EUCHROMATIN association; DNMT3A2 is the predominant form in embryonic stem and embryonal carcinoma cells and is detectable in testis, ovary, thymus and spleen, while DNMT3A is expressed at low levels ubiquitously, and DNMT3A2 expression correlates with high de novo methylation activity.

   A novel Dnmt3a isoform produced from an alternative promoter localizes to euchromatin and its expression correlates with active de novo methylation. · Chen T, Ueda Y, Xie S, Li E · J Biol Chem 2002
   PMID 12138111DOI
   ▸What this evidence cannot show

   Mouse and human embryonic stem, embryonal carcinoma, breast and ovarian cancer cells - nothing haematopoietic, nothing leukaemic, nothing fusion-bearing (brief rule 73, and the claim names the systems). 'Suggestive of euchromatin association' is the authors' own hedge and is a pattern description rather than a co-stained assignment. The methylation-activity correlation is across cell lines and is an association, not a demonstration that the isoform causes the activity. Bounded by the abstract.

   ▸Evidence that argues against this (1)

   • The isoform result cuts against this entry's own framing in a way worth stating plainly. The entry exists to record an AGE SPLIT - DNMT3A mutation matters in adult KMT2A-rearranged AML and is essentially absent in children - and it reasons at the level of the gene. But if the two promoters produce proteins with opposite chromatin distributions, and the euchromatic one is enriched in embryonic and stem-like cells, then an infant blast and an adult blast may not be expressing the same DNMT3A, and an age split at the level of MUTATION FREQUENCY could coexist with an age split in ISOFORM USAGE that nobody has looked for. Neither this entry nor the adult cohorts it cites resolve which isoform carries R882. Kim GD et al. 2002
7. E_q93_dnmt3a_loc3moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

   PubMed:12145218, UniProt's sole experimental citation for DNMT3A's cytoplasmic localisation, is a study of complex formation and functional co-operation among the human DNA methyltransferases - antibodies against hDNMT1 pull down both de novo enzymes, and the work concerns catalytic communication between maintenance and de novo activities - and its abstract describes no localisation experiment.

   Co-operation and communication between the human maintenance and de novo DNA (cytosine-5) methyltransferases. · Kim GD, Ni J, Kelesoglu N, Roberts RJ, Pradhan S · EMBO J 2002
   PMID 12145218DOI
   ▸What this evidence cannot show

   Bounded by the abstract: a localisation panel may sit inside the paper, and this is not a claim that the cytoplasmic pool is imaginary. What is recorded is that a reader checking the pointer as brief rule 91 prescribes finds a co-immunoprecipitation and enzymology paper. This is the third instance this beat has logged of a compartment value sourced to an interaction or structural study rather than a localisation one, after NFKB1 p50 (a crystal structure) and KAT2A's STAGA purification.

   ▸Evidence that argues against this (1)

   • A cytoplasmic pool for DNMT3A is not implausible and one independent line of reasoning supports it: this same paper's finding that DNMT1 co-precipitates both de novo enzymes implies a soluble, pre-chromatin complex, and a soluble methyltransferase complex has to assemble somewhere. So the annotation may be recording a real biochemical fact reached by the wrong instrument - which is a different failure from a wrong annotation, and is the commonest form of rule-91 error this beat has found. The graph should hold the cytoplasmic value as plausible, unsourced-for-its-purpose, and untested in any haematopoietic cell. Chen T et al. 2002
8. E_q93_dnmt3a_loc4strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

   Bounded 2026-09-21: the DNMT3A gene clause returns 5,587 records and against a nine-term KMT2A/MLL fusion clause returns 119, yet the gene clause AND a localisation clause AND the fusion clause returns ZERO; and DNMT3A[tiab] AND (pericentric[tiab] OR pericentromeric[tiab]) returns 21 records of which NONE is a DNMT3A localisation study, the closest being PMID 19482874 on the paralogue DNMT3B interacting with CENP-C at centromeric regions.

   NCBI E-utilities esearch db=pubmed; querytranslation read back and stored beside every count · NCBI E-utilities · Databases 2026
   ▸What this evidence cannot show

   Titles and abstracts only; supporting figures not searched (brief rule 80). The zero is unusually well-controlled for this beat because the fusion-context operand is 119 rather than a handful - the instrument demonstrably returns large numbers for this gene in this disease, so the localisation zero is about the content of that literature and not about retrieval. Classed measured-not-analysed because the isoform question is answerable from deposited RNA-seq: DNMT3A and DNMT3A2 use different promoters, so promoter-level quantification in existing infant KMT2A-rearranged datasets would resolve which form is present without new wet work.

   ▸Evidence that argues against this (1)

   • The 119-versus-zero contrast is the sharpest thing in this entry and it should be read as a statement about the FIELD rather than about the gene. DNMT3A in leukaemia is studied as a mutation - allele burden, clonal haematopoiesis, prognosis, R882 hotspot structure - and a mutation is a genotype, which no localisation assay reports. So the brief's warning not to let a mutation paper stand in as a localisation record describes the entire 119-record literature, and the reason there is no compartment measurement is not neglect but that nobody working on this gene in this disease is asking a cell-biological question. That also means the gap will not close by itself. Chen T et al. 2002

• co-occurs with →KMT2A-rearranged acute myeloid leukaemia in children and infants

  independent adverse factor in adult KMT2A-r AML (HR 2.1); conspicuously absent in paediatric disease

• related to →TP53 - named in 84 entries of this graph, absent from the index genome study of the disease, and four different things depending on when you look

  the same age split: both common in adult KMT2A-r leukaemia and conspicuously absent from paediatric cases

• ← supportsEnumeration of candidate cooperating lesions in KMT2A-rearranged leukaemia - what was written, what was recorded as a gap, and what was read and rejected

  the bounded queries behind the DNMT3A age-split entry

• ← recruitsH4R3me2s - the repressive symmetric arginine mark that recruits DNMT3A, blocks a KMT2 enzyme's own PHD module, and is the ONLY arginine mark ever measured in a KMT2A-fusion system

  direct binding target for DNMT3A via its ADD domain containing the PHD motif; required for subsequent DNA methylation - the route from the arginine layer into DNA methylation

• ← related toDNMT1 - the antagonist: its CXXC domain reads unmethylated CpG in order to STOP methylation, and it is the only heterologous CXXC that runs an MLL-AF9 leukaemia

  Maintenance versus de novo methyltransferase; DNMT1's CXXC autoinhibition is the device that separates their jobs.

• ← related toThe methylome co-mutation table, recomputed: IDH2 up at p = 0.0004 and not p < 0.0001, IDH1 and DNMT3A flat, and the abstract disagrees with its own Table 2

  Fourth methylome gene in the same table and also flat: 16.0% vs 17.2%, recomputed p = 0.49.

• related to →Chromocentres and pericentromeric satellite heterochromatin — no KMT2A evidence, but the AT-hook argument says there should be

  FULL-LENGTH DNMT3A IS CONCENTRATED ON HETEROCHROMATIN and accumulates in the MAJOR SATELLITE REPEATS at pericentric heterochromatin, while the DNMT3A2 isoform - lacking the N-terminal 223 residues, from a downstream intronic promoter - is EUCHROMATIC (PMID 12138111). A positive, isoform-resolved subnuclear assignment; never measured in a leukaemic cell (Q93 2026-09-21)

• related to →Retained fetal programme — the infant-versus-adult stem cell difference, and the reason KMT2A::AFF1 needs no second mutation

  The EUCHROMATIC isoform is the stem-like one: DNMT3A2 is predominant in embryonic stem and embryonal carcinoma cells and detectable in thymus and spleen, while full-length pericentric DNMT3A is low-level ubiquitous (PMID 12138111). Which form an infant KMT2A::AFF1 blast expresses is unmeasured and answerable from deposited RNA-seq, because the two use different promoters (Q93 2026-09-21)

• related to →A partner's 'normal' compartment is usually read off a different tissue, and for at least two KMT2A partners that reading is wrong in blood cells

  Rule 91 in both of its forms on one record: the UNCODED note carries the informative pericentric-satellite statement while the CODED values are generic, so an ECO:0000269 filter keeps the useless lines; and the cytoplasmic value is sourced to a DNMT1/DNMT3 co-immunoprecipitation paper with no localisation experiment in its abstract (PMID 12145218) (Q93 2026-09-21)

• related to →NFKB1 (p105/p50) — the only menin-NF-kB subunit with a co-immunoprecipitation, and the arm the MLL leukaemia data points at

  Two processed-or-spliced proteins, two levels of database resolution: NFKB1 gets SEPARATE molecule-scoped location comments for p105 and p50, both coded, while DNMT3A's two promoter-driven forms with OPPOSITE chromatin preferences are flattened into one Nucleus/Chromosome/Cytoplasm triple (Q93 2026-09-21)