Oncofusion

start / graph / mark-h4k20me2

Mechanismstrong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.Q104-interactome-read-queue

H4K20me2 - the abundant repair-linked H4 mark, the 53BP1 docking site, and the first H4K20 state in this graph; never once measured in a KMT2A-rearranged cell

also called H4K20me2 · H4-K20me2 · histone H4 lysine 20 dimethylation · H4 Lys20 dimethyl · dimethyl H4K20

H4K20me2 is a chemical tag on histone H4 that acts as the landing pad for 53BP1, the protein that marks broken DNA for repair. Until this entry the graph had eighteen histone-mark records and not one of them was any H4K20 state, and no published study has ever measured it in a KMT2A-rearranged leukaemia cell.

The same thing, for a scientist

H4K20 methylation resolves into three states with three canonical assignments - **me1 to transcription, me2 to repair, me3 to constitutive heterochromatin** [E2]. The me2 state is the docking site for the **tandem Tudor domains of 53BP1**, and the 2006 co-crystal explains the state specificity structurally: a five-residue binding cage that accommodates a dimethyl-lysine and **excludes** a trimethyl-lysine [E1]. For seven years after the 2010 review the mark had **no known eraser at all** - the review says so in its own abstract, "there is no known demethylase of H4K20, though the search is ongoing" [E2] - and the gap was closed by the discovery of a Jumonji subfamily, defined by *C. elegans* DPY-21 and its mammalian member **RSBN1/KDM9**, that converts H4K20me2 to H4K20me1 in worms and mammals [E3]. RSBN1 is in this project's interactome read queue as an AFF4 proximity partner, which is the only reason this mark has an entry: **it arrived through the interactome, not through the chromatin literature.** Measured against this project's index gene the mark is wholly unstudied - `H4K20me2[tiab] AND KMT2A[tiab]`, `H4K20me2[tiab] AND MLL[tiab]`, `H4K20me1[tiab] AND KMT2A[tiab]` and `H4K20me1[tiab] AND MLL[tiab]` all return **parsed zeros**, and the only record in which any H4K20 spelling meets MLL in an abstract is a 2007 ASH1L paper whose conclusion is that the trithorax-family distribution does **not** resemble H4K20 [E4].

Go deeper
Why this node exists, and what it is not

This graph holds eighteen mark-* nodes: H2AK119ub, H2BK120ub, H2BK34ub, H3K27ac, H3K27me3, H3K36me2-3, H3K4me2, H3K4me3, H3K79me1, H3K79me2, H3K79me3, H3K9ac, H3K9me1, H3K9me2, H3K9me3, H3R2me2a, H3R2me2s, H4K16ac, H4R3me2a, H4R3me2s. Every one of them is on H3 or is an H4 acetyl or arginine mark. There was no H4K20 state of any kind, which means the single most abundant lysine methylation on H4 in mammalian chromatin was missing from a chromatin-mechanism graph.

This entry is deliberately NOT a claim about KMT2A. It is the substrate side of gene-rsbn1, written so that the eraser has something to point at and so that the absence is on the record as a bounded, dated absence rather than as a silence. The honest state of the H4K20-and-KMT2A question is: nobody has ever looked, in any system, under any of the spellings tried.

The three states, the writers and the reader
statecanonical association [E2]enzyme
H4K20me1transcription; cell-cycle and chromosome compactionwritten by KMT5A/SET8/PR-Set7; produced by RSBN1/DPY-21 demethylation of me2 and me3 35
H4K20me2DNA repair - 53BP1 tandem-Tudor ligand 1written by KMT5B/SUV420H1 from me1; erased by RSBN1/KDM9 356
H4K20me3constitutive heterochromatinwritten by KMT5C/SUV420H2; erased by RSBN1 to me1 in mouse germ cells 5

The reader result is the strongest single fact here and it is structural. Botuyan 2006 solved 53BP1 and S. pombe Crb2 tandem Tudor domains bound to H4K20me2 by X-ray crystallography and NMR, and showed the five-residue cage fits a dimethyl-lysine and excludes a trimethyl-lysine - so the me2/me3 distinction is not a matter of degree, it is a binary at the reader 1. Any perturbation that converts me2 to me1, which is exactly what RSBN1 does, therefore removes a 53BP1 binding site rather than weakening one.

And the eraser is recent. The 2010 review states in its abstract that no H4K20 demethylase was known 2; the Jumonji subfamily was defined in 2017 3; the mammalian activity on both me2 and me3 was reported in 2021 in mouse 5; and the first human promoter-level result is 2025, in glioma 6. The whole eraser literature is three papers.

The bounded absence, with every spelling tried

Run live 2026-09-22, each operand separately tagged (no (A OR B)[tiab] form, per rule 94), and both KMT2A symbol generations paired with every mark spelling:

querycountverdict
H4K20me2[tiab] AND KMT2A[tiab]0parsed zero, warninglist: ["No items found."]
H4K20me2[tiab] AND MLL[tiab]0parsed zero
H4K20me1[tiab] AND KMT2A[tiab]0parsed zero
H4K20me1[tiab] AND MLL[tiab]0parsed zero
H4K20[tiab] AND KMT2A[tiab]0parsed zero
H4K20[tiab] AND MLL[tiab]1PMID 17923682, and it is a NEGATIVE for H4K20
H4K20me1[tiab] AND leukemia[tiab]5non-zero; NOT read by this beat
H4K20me2[tiab] AND 53BP1[tiab]38positive control - the instrument can return a one on this mark
SUV420H1[tiab] AND H4K20[tiab]38positive control

Three things this table is careful about. First, the positive controls: 38 records on the reader pair and 38 on the writer pair prove the compact H4K20me2 operand retrieves real literature, so the KMT2A/MLL zeros are not instrument failures (rule 67). Second, the spelling decides the answer: H4K20 reaches one MLL record and H4K20me1/H4K20me2 reach none, which is the compact-mark one-way-street failure this project's registry already records for H3K9me1/2 - the compact spelling and the bare-residue spelling do not bound each other. Third, H4K20me1[tiab] AND leukemia[tiab] returns 5, and this beat did not read them; that is a live gap, not a null, and it is named in the open questions rather than counted as a zero.

And the limit that matters most. All of these bound titles and abstracts only. This project has already found one case (an H3K79me2 ChIP across five HOXA promoters, PMID 18299449) that existed solely in a supporting figure and was invisible to every abstract query. An H4K20me1 or me2 panel could be sitting in a supplementary figure of a KMT2A paper right now, and nothing here would see it.

Facts and figures
markH4K20me2
histoneH4
residueK20 (Lys-20), N-terminal tail
chemistryepsilon-N,N-dimethyl-lysine. The me2/me3 distinction is a BINARY at the reader, not a gradient: the 53BP1 tandem-Tudor cage accommodates dimethyl and excludes trimethyl [E1].
canonical assignmentDNA repair. The three-state consensus is me1 -> transcription, me2 -> repair, me3 -> constitutive heterochromatin [E2].
writers verification statusNOT INDEPENDENTLY VERIFIED BY THIS BEAT. The writer assignments come from the review [E2] and from a positive-control count, not from reading a writer paper. They are recorded because the mark node needs its writers named, and they are flagged because this project's rule 91 is that a pointer is not a source. A beat extending this node should open a KMT5B primary paper before any claim rests on the writer identity.
eraser statusERASER KNOWN SINCE 2017 AND ONLY SINCE 2017. The 2010 review states in its own abstract that no H4K20 demethylase was known [E2]; the Jumonji subfamily was defined in 2017 [E3]; mammalian me2+me3 activity was reported in mouse in 2021 [E5]; the first human promoter-level result is 2025, in glioma [E6]. The entire eraser literature is three papers and none of them is in a haematopoietic cell.
reader caveat53BP1 recruitment to double-strand breaks requires H2AK15ub in addition to H4K20me2, so H4K20me2 is necessary but not sufficient. This graph holds no H2AK15ub node, so the second half of the 53BP1 signal is unrepresented.
nomenclature noteTHE SPELLING DECIDES THE ANSWER. `H4K20[tiab] AND MLL[tiab]` returns 1 record; `H4K20me1[tiab] AND MLL[tiab]` and `H4K20me2[tiab] AND MLL[tiab]` each return 0. The compact state-specific spelling and the bare-residue spelling retrieve different sets and NEITHER BOUNDS THE OTHER - the same one-way-street failure this project's registry records for the slashed H3K9me1/2 form. Any bounded absence on this mark must run both.
relationship to kmt2a and the fusionNONE PUBLISHED, IN EITHER DIRECTION. Five bounded queries return parsed zeros [E4]. The single record in which any H4K20 spelling meets MLL in an abstract is Gregory 2007 (PMID 17923682), and it is a NEGATIVE: mammalian ASH1L occupies transcribed regions with a distribution resembling H3K4me but NOT H3K9 or H4K20, while ASH1L and MLL1 have similar distributions and recruit independently of each other. So the only datum linking the trithorax family to H4K20 says the mark does not track it - and that datum is about ASH1L, not about KMT2A, whose own H4K20 relationship has never been measured.
cell typeNo haematopoietic cell has ever been assayed for this mark in a KMT2A context. The published contexts reached by this beat are C. elegans somatic and germ cells [E3], mouse round and elongated spermatids [E5], human glioma cell lines and xenografts [E6], and in vitro peptide/Tudor-domain structures with no cell at all [E1]. NOT SEARCHED: the 5 records returned by H4K20me1[tiab] AND leukemia[tiab], which may contain a haematopoietic measurement of the me1 state. No B-lineage cell, no infant cell, no primary leukaemia sample and no KMT2A::AFF1-expressing cell appears anywhere in the H4K20 record this beat reached.
cell stateCell-cycle-coupled at the me1 state and compaction-linked at the family level [E2][E3]; the me2 state is the DNA-damage-response state via 53BP1 docking [E1]; and in mouse the erasure is a spermatid-maturation event [E5]. NOT assayed in any differentiation state, and specifically not in the differentiation-blocked state this project's fusion imposes. No cell-cycle-resolved H4K20 measurement in a haematopoietic cell was found here.
route into this graphThrough the INTERACTOME, not the chromatin literature. The mark has an entry because its eraser RSBN1 appeared in the core-complex interactome read queue as an AFF4 proximity partner. That provenance is recorded because it means the mark's presence here is NOT evidence that anyone has connected H4K20 to KMT2A biology.
localizationChromatin-wide, on nucleosomal histone H4; not a compartment-restricted mark. The me3 state is enriched at constitutive heterochromatin and the me1 state is cell-cycle-coupled [E2]; no compartment assignment is asserted here for me2 beyond 'nucleosomal'.
subnuclear localizationNOT ASSIGNED. No subnuclear body or compartment assignment for H4K20me2 was found by this beat, and none is claimed. H4K20me3 is the state associated with constitutive heterochromatin (pericentromeric and telomeric) per E2; for the me2 state specifically the honest record is that it is distributed across chromatin and no sub-nuclear address has been established. Recorded as a bounded absence, not an omission.
epigenetic readoutThis node IS the epigenetic readout: H4K20me2 as the 53BP1 docking state, its conversion to H4K20me1 by RSBN1/KDM9, and the fact that no H4K20 state has ever been measured in a KMT2A-rearranged cell.
shown for fusionNO - nobody has looked in any KMT2A fusion, in any system. Not KMT2A::AFF1, not MLL-AF9, not any other partner. Five bounded queries, five parsed zeros [E4].
absence classnever-measured
absence class noteMIGRATED 2026-09-23, mechanically and not by judgement: the class string's own first word is this token. The field previously held the whole verdict as free text and that string is preserved here verbatim because it says more than the token can: 'never-measured (not measured-and-unreported, and not unpublishable). An H4K20me1/me2 immunoblot or ChIP in SEM, RS4;11 or MV4-11 against a KMT2A-wild-type B-ALL line is a one-week experiment with commercial antibodies, and no result of that kind exists under any spelling tried.'
writers
[
  "KMT5B/SUV420H1 - converts H4K20me1 to me2 (positive-control query SUV420H1[tiab] AND H4K20[tiab] = 38 records; no individual writer paper was read by this beat, so the enzyme assignment is carried here from the review E2 and is NOT independently verified)",
  "KMT5C/SUV420H2 - me2 and me3 (same status: carried from E2, not independently verified)",
  "KMT5A/SET8/PR-Set7 - the me1 monomethylase upstream (same status)"
]
erasers
[
  "RSBN1/KDM9 (Q5VWQ0) - Jumonji subfamily defined by C. elegans DPY-21; converts H4K20me2 (and me3) to H4K20me1 [E3][E5][E6]. THIS GRAPH'S ENTRY: gene-rsbn1."
]
readers
[
  "53BP1 tandem Tudor domains - state-specific, dimethyl-in trimethyl-out, X-ray and NMR [E1]. No 53BP1 entry exists in this graph as of 2026-09-22.",
  "Crb2 (S. pombe 53BP1 counterpart) - same fold, same specificity [E1]"
]
literature bounds
{
  "queried": "2026-09-22",
  "counts": {
    "H4K20me2[tiab] AND KMT2A[tiab]": 0,
    "H4K20me2[tiab] AND MLL[tiab]": 0,
    "H4K20me1[tiab] AND KMT2A[tiab]": 0,
    "H4K20me1[tiab] AND MLL[tiab]": 0,
    "H4K20[tiab] AND KMT2A[tiab]": 0,
    "H4K20[tiab] AND MLL[tiab]": 1,
    "H4K20me2[tiab] AND 53BP1[tiab]": 38,
    "SUV420H1[tiab] AND H4K20[tiab]": 38,
    "H4K20me2[tiab] AND demethylase[tiab]": 12,
    "\"H4 lysine 20\"[tiab] AND demethylase[tiab]": 12,
    "H4K20me1[tiab] AND leukemia[tiab]": 5
  },
  "querytranslations": {
    "H4K20me2[tiab] AND KMT2A[tiab]": "\"H4K20me2\"[Title/Abstract] AND \"KMT2A\"[Title/Abstract]",
    "H4K20me2[tiab] AND MLL[tiab]": "\"H4K20me2\"[Title/Abstract] AND \"MLL\"[Title/Abstract]",
    "H4K20me1[tiab] AND MLL[tiab]": "\"H4K20me1\"[Title/Abstract] AND \"MLL\"[Title/Abstract]",
    "H4K20[tiab] AND MLL[tiab]": "\"H4K20\"[Title/Abstract] AND \"MLL\"[Title/Abstract]"
  },
  "warninglist_on_every_zero": [
    "No items found."
  ],
  "positive_controls": "38 and 38 - the compact H4K20me2 operand demonstrably returns literature, so the KMT2A/MLL zeros are results and not instrument failures",
  "not_tried": [
    "H4K20me3 paired with KMT2A or MLL",
    "KMT5A/KMT5B/KMT5C paired with KMT2A or MLL",
    "\"H4-K20\"",
    "\"lysine 20\" paired with MLL-AF4, MLL-AF9 or \"t(4;11)\""
  ],
  "bounded_by": "titles and abstracts only; supporting figures and supplementary tables NOT searched"
}
How this could bite us back
  • An H4K20 eraser inhibitor would raise H4K20me2 and me3 genome-wide, and H4K20me2 is the 53BP1 docking state. Raising it non-specifically perturbs double-strand-break repair pathway choice in every proliferating cell, including normal haematopoietic progenitors - an essential-function toxicity, not a lineage-restricted one.
  • Conversely, anything that LOWERS H4K20me2 removes 53BP1 docking sites and would be expected to shift repair away from non-homologous end joining. In a disease caused by a translocation, deliberately perturbing break-repair pathway choice in surviving cells is a plausible route to second rearrangements.
  • The me1 state produced by erasure is cell-cycle-coupled and compaction-linked. An intervention aimed at a leukaemia-specific locus would land on a mark whose function is chromosome-scale, so the off-target surface is the genome rather than a gene set.
  • This mark has no measurement in the index disease at all, so any intervention argument built on it would currently be extrapolated from C. elegans dosage compensation, mouse spermatids and human glioma. That is exactly the partner-imported evidence pattern this project documents as its commonest synthesis failure.
Still open
  1. What are the H4K20me1, me2 and me3 levels in a KMT2A::AFF1 cell line versus a KMT2A-wild-type B-ALL line? This is the cheapest unexplored mark question in the corpus: commercial antibodies, an immunoblot, and no published result exists under any spelling tried.
  2. What is in the 5 records returned by H4K20me1[tiab] AND leukemia[tiab]? This beat counted them and did not read them. If any contains a haematopoietic H4K20me1 profile, the mark stops being a blank in leukaemia and the KMT2A question becomes a comparison rather than a first measurement.
  3. Does a menin inhibitor change H4K20 methylation? This project has previously found that the broad claim 'no menin-inhibitor study reports mark X' failed for H3K27me3 because the measurement sat in a genetic arm. The same query has never been run for H4K20.
  4. Is H4K20me2 perturbed at the HOXA locus in KMT2A-rearranged cells, where H3K79me2 and H3K4me3 both are? An H4K20me2 track over HOXA9/HOXA10 would say whether the repair-linked mark is excluded from the fusion's own target domain.
  5. Does KMT2A's PHD1-3/bromodomain cassette - the module lost from every 5'-KMT2A fusion - read any H4 tail state? The ASH1L negative [E4] concerns distribution, not binding, and no H4K20 binding assay has been run on the KMT2A reader module.
Every source behind this entry (6)
  1. E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    The tandem Tudor domains of mammalian 53BP1 and fission-yeast Crb2 bind histone H4 specifically dimethylated at Lys20, and the X-ray and NMR structures reveal a five-residue binding cage that accommodates a dimethyl-lysine but excludes a trimethyl-lysine, explaining methylation-state-specific recognition of H4K20 in targeting DNA repair proteins to double-strand breaks.

    PMID 17190600DOI
    What this evidence cannot show

    Read from the abstract only; the PDB entries for the complexes were not fetched, so no structure id is asserted in data.pdb_ids. In vitro structural work on peptides and isolated Tudor domains - it establishes the binding chemistry, not the in vivo occupancy of any locus. No leukaemia, no KMT2A, no fusion. Cited as the reader assignment and for the me2-versus-me3 binary, which is the property that makes an eraser consequential.

    Evidence that argues against this (1)
    • 53BP1 recruitment to breaks is now understood to require H2AK15ub as well as H4K20me2 - a bivalent nucleosome signal - so H4K20me2 alone is necessary but not sufficient, and a change in H4K20me2 alone does not predict a change in 53BP1 recruitment. This graph already holds H2AK119ub and H2BK120ub nodes but no H2AK15ub node, so the second half of the 53BP1 signal is unrepresented here. Balakrishnan L et al. 2010
  2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.review

    The consensus assignment of the three H4K20 methylation states is H4K20me1 to transcription, H4K20me2 to repair and H4K20me3 to constitutive heterochromatin; a large number of methylases capable of methylating H4K20 had been identified and characterised, and as of 2010 there was NO known demethylase of H4K20.

    Decoding the histone H4 lysine 20 methylation mark. · Balakrishnan L, Milavetz B · Crit Rev Biochem Mol Biol 2010
    PMID 20735237DOI
    What this evidence cannot show

    THIS IS A REVIEW and is cited as a map, per the quality bar, for two things only: the three-state canonical assignment, and the dated statement that no H4K20 demethylase was known in 2010. Read from the abstract. The three-state assignment is a consensus summary and the review's own wording is 'an emerging consensus' whose 'specific details ... are now the subject of active investigation' - it is a frame, not a measurement, and no per-locus or per-cell-type claim should be taken from it. The 2010 no-eraser statement is superseded by E3 and is quoted BECAUSE it is superseded: it dates the eraser literature to after 2010.

    Evidence that argues against this (1)
    • The review's own no-demethylase statement was overturned seven years later, which is a warning about the class of claim: an absence asserted in a review is an absence of publication at a date, and this one had a seven-year half-life. Any absence in this entry, including the KMT2A zeros, should be read with the same expiry. Brejc K et al. 2017
  3. E3established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    The structure and activity of the C. elegans dosage-compensation complex subunit DPY-21 define a Jumonji demethylase subfamily that converts H4K20me2 to H4K20me1 in worms and mammals; selective inactivation of the demethylase activity eliminates H4K20me1 enrichment in somatic cells, elevates X-linked gene expression, reduces X chromosome compaction and disrupts topologically associating domain formation.

    PMID 28867287DOI
    What this evidence cannot show

    SYSTEM: C. elegans dosage compensation. The mammalian half of the claim is the abstract's own wording and this beat read the abstract only, so which mammalian enzyme was assayed and by what method is not verified here - it is the weakest link in the H4K20 eraser chain and is flagged as such in gene-rsbn1 too. Cited for the reaction and for the functional consequence that a shift in H4K20 methylation state changes higher-order chromosome structure, which is the reason the mark is worth measuring at all in a disease driven by a chromatin complex.

    Evidence that argues against this (1)
    • The same abstract reports that DPY-21 also acts on germ-cell autosomes in a complex-INDEPENDENT manner. An enzyme whose recruitment is complex-dependent in one tissue and complex-independent in another cannot have its targeting inferred from a single co-purification partner, which is exactly what the AFF4 proximity row for its mammalian member invites. Brejc K et al. 2017
  4. E4strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

    PubMed esearch, 2026-09-22, each operand separately tagged: H4K20me2[tiab] AND KMT2A[tiab] = 0, H4K20me2[tiab] AND MLL[tiab] = 0, H4K20me1[tiab] AND KMT2A[tiab] = 0, H4K20me1[tiab] AND MLL[tiab] = 0, H4K20[tiab] AND KMT2A[tiab] = 0 - all five parsed zeros - against positive controls of H4K20me2[tiab] AND 53BP1[tiab] = 38 and SUV420H1[tiab] AND H4K20[tiab] = 38; and H4K20[tiab] AND MLL[tiab] = 1, that single record being an ASH1L paper reporting that the ASH1L distribution does NOT resemble H4K20.

    What this evidence cannot show

    Every zero carries a NON-EMPTY querytranslation - e.g. `"H4K20me2"[Title/Abstract] AND "KMT2A"[Title/Abstract]` - and `warninglist.outputmessages = ["No items found."]`, this project's signature of a genuine zero rather than an instrument failure. TWO POSITIVE CONTROLS at 38 records each prove the compact operand retrieves literature, per rule 67. BOTH KMT2A SYMBOL GENERATIONS were run against every spelling, and the result differs between them for the bare form: H4K20 reaches 1 MLL record while the compact me1/me2 forms reach none, so the compact and bare spellings do NOT bound each other - the same one-way-street failure the registry records for H3K9me1/2. NAME LIST: H4K20, H4K20me1, H4K20me2, 'H4 lysine 20', SUV420H1, 53BP1, KMT2A, MLL. NOT TRIED: H4K20me3 paired with KMT2A/MLL, KMT5A/KMT5B/KMT5C paired with KMT2A/MLL, 'H4-K20', 'lysine 20' paired with MLL, MLL-AF4, MLL-AF9 or t(4;11). NOT A ZERO AND NOT READ: H4K20me1[tiab] AND leukemia[tiab] returns 5 records which this beat did not open - that is an open gap, recorded as such, not counted as an absence. BOUNDED BY TITLES AND ABSTRACTS ONLY, per rule 80: this project has already found an H3K79me2 ChIP that existed only in a supporting figure.

    Evidence that argues against this (1)
    • The five unread records in H4K20me1[tiab] AND leukemia[tiab] are the standing refutation of any strong reading of this absence. There is an H4K20me1-and-leukaemia literature; what has never been done is to look in a KMT2A-rearranged cell. Anyone quoting this entry should quote the narrow claim (no KMT2A/MLL intersection in titles or abstracts) and not the broad one (H4K20 is unstudied in leukaemia), which is false. NCBI E-utilities 2026
  5. E5moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    Mouse RSBN1 demethylates both dimethyl and trimethyl H4K20 to monomethyl H4K20, and H4K20me2 and H4K20me3 but not H4K20me1 are absent from RSBN1-positive germ cells in the seminiferous tubule.

    PMID 34185806DOI
    What this evidence cannot show

    SYSTEM: mouse spermatogenesis, immunostaining plus a demethylase assay; read from the abstract only. Not human, not haematopoietic, no fusion. The me3 activity is this paper's extension of the prior me2 assignment and this beat found no independent reproduction of it. Antibody-based mark distributions in germ cells carry the usual specificity caveat, which this graph records elsewhere for arginine marks: an anti-H4K20me2 antibody that cross-reacts with me3 would produce the same figure.

    Evidence that argues against this (1)
    • If one enzyme removes both me2 and me3 to me1, then the three canonical state assignments in E2 - repair, heterochromatin, transcription - are not independently controllable, and a single eraser activity moves chromatin from a repair-competent and a heterochromatic state into a transcription-linked one simultaneously. That is a strong claim resting on one mouse paper, and it is the part of the H4K20 eraser story most in need of replication in a somatic human cell. Balakrishnan L et al. 2010
  6. E6moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    In human glioma cells KDM9 (RSBN1) mediates demethylation of H4K20me2 at the CCND1 promoter and interacts with BRD2 to facilitate CCND1 transcription; KDM9 knockdown suppresses xenograft growth and temozolomide resistance.

    PMID 41291235DOI
    What this evidence cannot show

    SYSTEM: human glioma lines and xenografts; read from the abstract only. This is the ONLY report placing an H4K20me2 demethylation event at a named human promoter, so the claim that this mark is actively erased at specific loci in a human somatic cancer cell rests on a single 2025 paper. Not leukaemia, no KMT2A fusion, and the partner is BRD2 rather than the BRD4 that sits in this graph's elongation axis.

    Evidence that argues against this (1)
    • A promoter-level demethylation result in one tumour type with no replication is the weakest form of a mark-mechanism claim, and CCND1 is a cell-cycle gene that is downstream of many oncogenic inputs - including, in this project's own disease, the fusion itself. So the glioma mechanism does not transfer: in a KMT2A-rearranged cell, CCND1 activation is already accounted for. Cheng J et al. 2025
Connected to