Oncofusion

start / graph / gene-dnmt1

Proteinstrong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.Q79-cxxc-readers

DNMT1 - 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

also called DNMT1 · CXXC-type zinc finger protein 9 · MCMT · DNA MTase HsaI · M.HsaI

DNMT1 copies the chemical tags on DNA after it is duplicated, so it is the enzyme that creates the tagged CG letters the KMT2A fusion cannot read. Oddly, it carries the same untagged-CG reader as the fusion - and uses it as a brake on itself. It is also the only borrowed version of that reader that lets an MLL fusion still cause leukaemia.

The same thing, for a scientist

DNMT1 (P26358, 1,616 aa) carries a CXXC-type zinc finger at **646-692** between its RFTS domain (400-533) and its tandem BAH domains (755-880, 972-1100), upstream of the C5 methyltransferase domain (1139-1599) [E1]. Song et al. solved mouse and human DNMT1 bound to unmethylated-CpG DNA and showed the mechanism: **the CXXC binds unmethylated CpG specifically and positions the CXXC-BAH1 linker between the DNA and the active site, preventing de novo methylation** - an autoinhibitory device that ensures only hemimethylated CpG is methylated [E2]. So DNMT1's CXXC reads the same ligand as KMT2A's and uses it for the opposite purpose: **to refuse to act.** DNMT1 is the family's antagonist in a second and larger sense - it maintains the 5mC that every other CXXC reader, including the fusion, requires to be absent. **And it is the one heterologous CXXC domain that works in the fusion.** In the MLL-AF9 swap, MLL(DNMT1 CXXC)-AF9 gave robust in-vitro colony formation and in-vivo leukaemogenesis comparable to MLL-AF9, while CGBP and MBD1 domains abrogated both [E3]. What distinguished them was which CpG residues were protected from methylation, which prevented H3K9me3 and enabled Hoxa9 expression [E3]. A 2025 report adds a protein-level link: **decitabine covalently degrades DNMT1, and in every ALL cell line in which DNMT1 was degraded, KMT2A protein was also diminished** - only those cells showed biological effects, and the effect was synergistic with menin inhibition [E4]. In this project's own unbiased scan DNMT1 is at **rank 16,949 of 17,728, the 95.6th percentile** - KMT2A-rearranged lines are *less* dependent on DNMT1 than other leukaemia lines (-0.572 versus -0.675), against a pan-essentiality of -0.596 [E5].

What it looks like
DNA (cytosine-5)-methyltransferase 11,616 aa
RFTS (replication foci targeting sequence)RFTS (replication foci targeting sequence) · residues 400–533CXXC (CpG-binding zinc finger) · residues 646–692BAH (bromo-adjacent homology) 1BAH (bromo-adjacent homology) 1 · residues 755–880BAH (bromo-adjacent homology) 2BAH (bromo-adjacent homology) 2 · residues 972–1100SAM-dependent MTase C5-typeSAM-dependent MTase C5-type · residues 1139–159905001,0001,500
loading AF-P26358-F1…
AlphaFold AF-P26358-F1 ↗
colourshape

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.

very high (90+)confident (70–90)low (50–70)very low (<50)
I have never read one of these pictures before

A protein is one long chain of amino acids that folds up. The ribbon traces that chain from one end to the other; it is not a wire, it is a simplification of where the backbone runs. Drag to rotate, scroll to zoom.

Domains are the stretches of chain that do a particular job, and they are what the coloured blocks above the picture mark out. Click one and the view flies to it. In this project the important question about any domain is usually whether the fusion still has it — the labels say lost where it does not.

Anything in yellow sticks is a bound small molecule, which is usually the drug.

Go deeper
A reader used as a brake

The DNMT1 structure is the family's clearest demonstration that a CXXC domain is a module, not a function. Bound to unmethylated CpG, DNMT1's CXXC pulls the CXXC-BAH1 linker into the path between the DNA and the catalytic site, occluding the substrate; a loop from BAH2 holds the target-recognition domain retracted 2. The consequence is negative catalysis: encountering an unmethylated CpG, DNMT1 reads it and declines to methylate it, which is what makes it a maintenance rather than a de novo methyltransferase.

Every other CXXC protein in the census uses the same ligand to recruit an activity. DNMT1 uses it to suppress one. Any pharmacological argument that treats 'CXXC domain' as a target class is therefore arguing about a module whose sign is set by what it is attached to.

Why the swap result is the most important fact in this beat

Risner et al. replaced the MLL CXXC domain in MLL-AF9 with three heterologous domains 3:

substituted domainbinds unmethylated DNA in vitrocolony formation and leukaemogenesis
MLL (wild type)yesyes
DNMT1 CXXCyesyes, comparable to MLL-AF9
CGBP/CXXC1 CXXCyesabrogated
MBD1 CXXCnoabrogated
MBD1 MBDnoabrogated

All five constructs still targeted the Hoxa9 locus. What separated the functional from the non-functional was which CpG residues were protected from methylation, and protection prevented H3K9me3 and permitted Hoxa9 expression 3.

Two consequences the graph should carry:

  1. DNA binding is necessary and not sufficient. CGBP binds and fails. The address is not the whole job.
  2. The functional variable is a chemical state of the DNA, not a protein-protein contact. That is a statement about the methylome, and it is the only rigorous support in this literature for the idea that the fusion's address depends on the cell's methylation landscape. It supports the brief's premise - but through DNMT1 and protection-from-methylation, not through TET2 co-mutation, which the data in gene-tet2 rules out.
The dependency direction, and why it is not a base rate

Rule 84 in this project's brief warns that 'in the anti-selective half' is a null, because half of every gene set is there by construction. DNMT1 is not in the anti-selective half, it is at the 95.6th percentile of 17,728 genes - in the most anti-selective 4.4%. KMT2A-rearranged lines are measurably less dependent on DNMT1 than other leukaemia lines: -0.572 against -0.675, difference +0.103 5.

Read with the pan-essentiality of -0.596, DNMT1 is a broadly required gene that KMT2A-rearranged lines need somewhat less than their comparators. That is the sign a model would predict if the fusion's programme is already running on hypomethylated CpG islands and depends less on the maintenance machinery - but it is 11 lines against 43, myeloid-enriched, with no infant line, and this project has documented that a z-scaled difference in a screen tracks comparator narrowness. It is a suggestive sign, not a result, and the decisive experiment is the decitabine one below, which was done in ALL lines and points the other way.

Decitabine: DNMT1 degradation takes KMT2A protein with it

Brock et al. report that decitabine, which covalently binds and degrades DNMT1, also diminished KMT2A protein in every ALL cell line that experienced DNMT1 degradation, and only cells with reduced KMT2A protein showed biological effects; KMT2A wild-type and rearranged cells arrested in G2 and G1 respectively, and the effect acted synergistically with menin inhibition 4.

Their stated rationale is structural similarity within the DNA-binding domains of DNMT1 and KMT2A - i.e. the CXXC domain. If that is the mechanism, it is a demonstration that the shared fold is pharmacologically reachable on both proteins at once, which is exactly what the phase-2 CXXC dossier's selectivity gate was written to avoid, arriving as a therapeutic opportunity rather than a liability. It also means a CXXC-family cross-reactivity is already being administered to patients as a licensed drug.

Caution: this is one 2025 report, the mechanism linking DNMT1 degradation to KMT2A loss is inferred rather than shown, and decitabine has many off-DNMT1 effects. This graph holds no decitabine or azacitidine entry at all, which is a gap this beat is flagging rather than filling.

Facts and figures
uniprotP26358
gene symbolDNMT1
protein nameDNA (cytosine-5)-methyltransferase 1
length AA1616
alphafold IDAF-P26358-F1
accession resolutionUniProt REST gene_exact:DNMT1 AND organism_id:9606 AND reviewed:true, run 2026-09-21; one row; primary gene name read back as DNMT1 and length 1616 from the same fetched record.
functionmaintenance DNA methyltransferase; its CXXC reads unmethylated CpG and autoinhibits catalysis, restricting the enzyme to hemimethylated substrate
classificationthe family's ANTAGONIST - it writes the 5mC that excludes every other CXXC reader, and its own CXXC is a brake rather than a recruiter; and the ONLY heterologous CXXC domain that supports MLL-AF9 leukaemogenesis
shown for fusionMLL-AF9 in mouse, as the one POSITIVE heterologous CXXC substitution - colony formation and leukaemogenesis comparable to MLL-AF9; separately, decitabine-induced DNMT1 degradation coincides with KMT2A protein loss in human ALL lines including KMT2A-rearranged ones
cell type evidencecrystallography of mouse and human protein; murine bone-marrow progenitors and transplantation (the swap); human ALL cell lines and xenografts (decitabine); DepMap leukaemia lines. No infant primary material.
epigenetic readout5mC maintenance; CpG protection at Hoxa9 in the swap work; H3K9me3 as the downstream consequence of failed protection
pharmacology noteDecitabine and azacitidine act through DNMT1. THIS GRAPH HOLDS NO ENTRY FOR EITHER, which is a gap flagged by this beat and not filled by it - a drug entry needs label, trial and resistance data this beat did not gather.
subnuclear localizationReplication foci in S phase and chromatin in G2 and M, both experimental in UniProt; the RFTS domain is the replication-focus targeting element. Never imaged relative to a KMT2A fusion.
cell typeCrystallography of mouse and human protein; murine bone-marrow progenitors and transplantation (the CXXC swap); human ALL cell lines and xenografts including KMT2A-rearranged ones (decitabine); DepMap leukaemia lines. No infant primary material.
domains
RFTS (replication foci ta…400.0RFTS (replication foci targeting sequence): 400CXXC (CpG-binding zinc fi…646.0CXXC (CpG-binding zinc finger): 646BAH (bromo-adjacent homol…755.0BAH (bromo-adjacent homology) 1: 755BAH (bromo-adjacent homol…972.0BAH (bromo-adjacent homology) 2: 972SAM-dependent MTase C5-ty…1,139SAM-dependent MTase C5-type: 1139
domains — the numbers behind the chart
[
  {
    "name": "RFTS (replication foci targeting sequence)",
    "start": 400,
    "end": 533,
    "pfam": "PF12047"
  },
  {
    "name": "CXXC (CpG-binding zinc finger)",
    "start": 646,
    "end": 692,
    "pfam": "PF02008"
  },
  {
    "name": "BAH (bromo-adjacent homology) 1",
    "start": 755,
    "end": 880,
    "pfam": "PF01426"
  },
  {
    "name": "BAH (bromo-adjacent homology) 2",
    "start": 972,
    "end": 1100,
    "pfam": "PF01426"
  },
  {
    "name": "SAM-dependent MTase C5-type",
    "start": 1139,
    "end": 1599,
    "pfam": "PF00145"
  }
]
localization
{
  "subcellular": "nucleus (ECO:0000269) and chromosome (ECO:0000269)",
  "subnuclear": "associates with DNA replication sites in S phase and with chromatin in G2 and M (UniProt, experimental)",
  "localization_shown_for_fusion": "NO - nobody has imaged DNMT1 relative to a KMT2A fusion"
}
essentiality
{
  "depmap_scan": "analyses/depmap-kmt2a-selective-scan-full",
  "difference_kmt2a_r_minus_other": 0.1031,
  "mean_kmt2a_rearranged": -0.5723,
  "mean_other_leukemia": -0.6754,
  "pan_essentiality_all_lines": -0.5957,
  "n_all_lines": 1208,
  "rank_of_17728": 16949,
  "percentile": 95.606,
  "interpretation": "broadly required; KMT2A-rearranged lines need it LESS than other leukaemia lines, in the most anti-selective 4.4% of the scan",
  "release_note": "Open Targets dataVersion 26.06, read 2026-09-21"
}
How this could bite us back
  • A CXXC-directed molecule that hits DNMT1 would raise unmethylated CpG genome-wide, which OPENS more address for the fusion. The intended effect and the off-target effect push in opposite directions on the same variable.
  • DNMT1 is broadly required (-0.596 across 1,208 lines) and is essential in normal proliferating tissue. A paediatric agent with DNMT1 cross-reactivity carries a general proliferative toxicity, and the graph has no entry recording what decitabine's paediatric toxicity profile actually is.
  • The DNMT1 CXXC substitution result means a functional CXXC domain for this fusion is not rare. If a resistance mutation in the KMT2A CXXC ever mattered, the fusion has a large sequence space of functional alternatives in its own genome to teach us what tolerated variation looks like.
Still open
  1. Why does the DNMT1 CXXC domain support MLL-AF9 leukaemogenesis when the CGBP domain does not, given that both bind unmethylated DNA? The paper's answer is CpG protection from methylation; the affinity, residence time and sequence preference of the two domains have never been compared side by side in the fusion context.
  2. Is the decitabine-induced loss of KMT2A protein a CXXC-mediated off-target of the drug, or a downstream consequence? A DNMT1 catalytic-dead or CXXC-mutant knock-in would separate them, and if it is CXXC-mediated it is the first evidence that this fold is drugged in the clinic already.
  3. KMT2A-rearranged lines are less dependent on DNMT1 than other leukaemia lines, yet decitabine arrests them. Are the CRISPR and pharmacological results measuring the same thing?
  4. Should this graph carry decitabine and azacitidine entries? The synergy with menin inhibition reported here is a clinically actionable combination with no entry anywhere in the corpus.
Every source behind this entry (6)
  1. E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

    DNMT1 (P26358, 1,616 aa) carries a CXXC-type zinc finger at 646-692, an RFTS domain at 400-533, tandem BAH domains at 755-880 and 972-1100 and a C5-type methyltransferase domain at 1139-1599, and its curated DOMAIN comment states the CXXC binds unmethylated CpG and positions the autoinhibitory linker.

    What this evidence cannot show

    Fetched 2026-09-21; accession resolved by gene_exact:DNMT1 AND organism_id:9606 AND reviewed:true, one row, primary gene name read back as DNMT1 and length 1616 from the same record. Three isoforms are annotated; the coordinates are canonical.

    Evidence that argues against this (1)
    • The RFTS domain is also autoinhibitory, occluding the active site until it binds ubiquitinated H3 - so DNMT1 has at least two independent autoinhibitory devices and attributing substrate discrimination to the CXXC alone oversimplifies it. UniProt Consortium 2026
  2. E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    Crystal structures of mouse and human DNMT1 bound to unmethylated-CpG DNA show the CXXC binding unmethylated CpG specifically and positioning the CXXC-BAH1 linker between the DNA and the active site, preventing de novo methylation and ensuring only hemimethylated CpG is methylated.

    PMID 21163962DOI
    What this evidence cannot show

    Crystal structures of truncated constructs with short DNA duplexes. An autoinhibited state in a crystal is a state the protein can adopt; the in-vivo balance between inhibited and active conformations, and how the RFTS domain and H3 ubiquitination shift it, is not settled by the structure.

    Evidence that argues against this (1)
    • A comprehensive specificity analysis of DNMT1 finds the enzyme's accuracy is set by several overlapping mechanisms, so the CXXC autoinhibition is one contributor to fidelity rather than the determinant. Adam S et al. 2023
  3. E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    MLL(DNMT1 CXXC)-AF9 shows robust in-vitro colony-forming activity and in-vivo leukaemogenesis comparable to MLL-AF9, while the CGBP and MBD1 CXXC domains and the MBD1 MBD abrogate both; all substituted fusions still target Hoxa9, and what distinguished them was which CpG residues were protected from methylation, preventing H3K9me3 and enabling Hoxa9 expression.

    Functional specificity of CpG DNA-binding CXXC domains in mixed lineage leukemia. · Risner LE, Kuntimaddi A, et al. · J Biol Chem 2013
    PMID 23990460DOI
    What this evidence cannot show

    Murine MLL-AF9 myeloid system, retroviral transduction of primary bone-marrow progenitors, transplantation. Three heterologous domains of the eleven available. The methylation-protection mechanism is read out at one locus. Nothing human, B-lineage or KMT2A::AFF1.

    Evidence that argues against this (1)
    • CGBP/CXXC1's isolated domain binds unmethylated DNA and the fusion carrying it still reaches Hoxa9, yet it does not transform - so DNA binding and locus targeting are both insufficient and the DNMT1 result may turn on an affinity or kinetic property not measured. Risner LE et al. 2013
  4. E4moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    Decitabine covalently binds and degrades DNMT1, and in ALL cell lines and xenografts KMT2A protein was diminished in every line that experienced DNMT1 degradation; only cells with reduced KMT2A protein showed biological effects, KMT2A wild-type and rearranged cells arrested in G2 and G1 respectively, and the effect acted synergistically with menin inhibition.

    PMID 39754404DOI
    What this evidence cannot show

    Human ALL cell lines and xenograft models. The causal link from DNMT1 degradation to KMT2A loss is inferred from co-occurrence and from the structural similarity of the DNA-binding domains, not demonstrated by a separation-of-function mutant. Decitabine has substantial DNMT1-independent effects including DNA damage. Abstract read this session; full text not read.

    Evidence that argues against this (1)
    • Decitabine's demethylating action raises DNMT1 and KMT2A mRNA in the same experiment while KMT2A protein falls, so transcript and protein move in opposite directions and a simpler explanation - global stress-induced turnover - is not excluded. Brock L et al. 2025
  5. E5moderate confidence 3 of 5: One strong study, or several smaller consistent ones.dataset

    KMT2A-rearranged cell lines are LESS dependent on DNMT1 than other leukaemia lines: mean effect -0.572 versus -0.675, difference +0.103, rank 16,949 of 17,728 (95.6th percentile), against a pan-essentiality of -0.596 across 1,208 lines.

    Genome-wide KMT2A-rearranged selectivity scan, 17,728 genes (DepMap CRISPR Chronos via the Open Targets Platform GraphQL API) · Oncofusion knowledge-graph project · analyses/depmap-kmt2a-selective-scan-full/results.json, generated 2026-09-19; Open Targets dataVersion year 26 month 06, re-read from the API on 2026-09-21 2026
    What this evidence cannot show

    The scan compares 11 Cellosaurus-annotated KMT2A-rearranged lines against 43 other leukaemia lines; none of the 11 is infant-derived and the arm is myeloid-enriched, so a difference mixes fusion status with lineage. Both the difference and the pan-essentiality across all 1,208 lines are quoted, because a gene required everywhere is not a selective dependency. Open Targets dataVersion 26.06; DepMap's own release identifier is not returned by that endpoint. Rule 84 applies in reverse here: 'in the anti-selective half' would be a null, but the 95.6th percentile is in the most anti-selective 4.4% and is not produced by construction. The scan's statistic is a raw difference, not a z, so the narrowness artefact this project documented does not apply.

    Evidence that argues against this (1)
    • The decitabine study reports that KMT2A-rearranged ALL cells respond to DNMT1 degradation with G1 arrest, which is not obviously compatible with reduced dependency on DNMT1; cell-line CRISPR and pharmacological degradation are different perturbations. Brock L et al. 2025
  6. E6strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    DNMT1's CXXC domain was measured alongside every other human CXXC domain in the systematic family comparison of DNA-binding selectivity.

    PMID 29276034DOI
    What this evidence cannot show

    Isolated domains, short duplexes, in vitro; DNMT1's domain in isolation is stripped of the linker-and-BAH machinery that gives it its function.

    Evidence that argues against this (1)
    • The whole point of DNMT1's CXXC is the conformation it imposes on the adjacent linker, which an isolated-domain binding assay cannot report. Song J et al. 2011
Connected to