Oncofusion

start / graph / gene-kdm2a

Proteinmoderate confidence 3 of 5: One strong study, or several smaller consistent ones.Q79-cxxc-readers

KDM2A - the CXXC reader that antagonises KMT2A leukaemogenesis, and displaces wild-type MLL rather than the fusion

also called KDM2A · FBXL11 · JHDM1A · CXXC-type zinc finger protein 8 · F-box/LRR-repeat protein 11 · FBL7 · Lilina

KDM2A parks on the same unmethylated CG islands as the KMT2A fusion and strips off an activating chemical mark. It slows down KMT2A-driven leukaemia - but the published experiment shows it pushes the NORMAL MLL protein off chromatin while leaving the cancer fusion where it is.

The same thing, for a scientist

KDM2A (Q9Y2K7, 1,162 aa) is an F-box/JmjC H3K36 demethylase carrying a **CXXC-type zinc finger at 564-610** and a PHD finger at 617-678, with the JmjC domain at 148-316 and the F-box at 889-936 [E1]. Blackledge et al. showed CpG islands **directly recruit** KDM2A through that ZF-CxxC domain, which preferentially recognises non-methylated CpG and is blocked by methylation, and that nucleation there removes H3K36 methylation and creates CpG-island chromatin uniquely depleted of the mark [E2]. That makes KDM2A the family member whose address is most nearly identical to the KMT2A fusion's - same reader chemistry, same element class, opposite output. **And it is the one family member with a direct experimental result in this disease, which is more interesting than 'competition'.** Zhu et al. report that KDM2A *antagonises MLL-associated leukaemogenesis*, but the occupancy measurement underneath that sentence is specific: KDM2A overexpression lowers H3K36me2 and **reduces the chromatin occupancy of LEDGF and of wild-type MLL - but not of the fusion** [E3]. So a CXXC reader that shares the address demonstrably displaces the wild-type protein and demonstrably fails to displace the oncoprotein. **That is a measured non-competition at the fusion, and it is the strongest single constraint this beat found on the contested-address hypothesis.** In DepMap KDM2A is the best-ranked CXXC-family gene for KMT2A selectivity - rank **442 of 17,728**, difference -0.156 - but it is also broadly required (pan-essentiality -0.620 across 1,208 lines), so the difference is a small selective increment on top of a large general requirement and should not be read as a KMT2A-specific dependency [E4].

What it looks like
Lysine-specific demethylase 2A1,162 aa
JmjCJmjC · residues 148–316CXXCCXXC (CpG-binding zinc finger) · residues 564–610PHD PHD (plant homeodomain) zinc finger · residues 617–678F-boF-box · residues 889–93602004006008001,000
loading AF-Q9Y2K7-F1…
AlphaFold AF-Q9Y2K7-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
What the CXXC domain does here

KDM2A's ZF-CxxC preferentially recognises non-methylated CpG DNA and binding is blocked when the CpG is methylated, which constrains KDM2A to non-methylated CpG islands; nucleation there removes H3K36 methylation and produces CpG-island chromatin uniquely depleted of that modification 2. UniProt's curated DOMAIN comment records the same, with a second experimental role - the CXXC is essential for KDM2A's ability to repress the transcriptional-activator activity of the CLOCK-BMAL1 heterodimer 1.

This is the same reader chemistry as the KMT2A CXXC (interface-kmt2a-cxxc-cpg-dna; the affinity is not a single number - ten published measurements span 44 uM to 33 nM depending on construct, tag, DNA and salt, with the nearest-to-physiological tagless value 5.7 uM at 150 mM NaCl, see ana-cxxc-dna-affinity-ladder) and it is deployed to the opposite end: KMT2A writes H3K4 methylation at CpG islands, KDM2A erases H3K36 methylation at the same element class.

The result that constrains this whole beat

This graph already holds the ASH1L/LEDGF/H3K36me2 axis in mark-h3k36me2-3 and across the ledgf-axis entries, and the KDM2A sentence sits inside it. Restated here at protein level because the consequence for the CXXC question had not been drawn:

  • ASH1L writes H3K36me2; LEDGF preferentially binds it and colocalises with MLL, ASH1L and H3K36me2 genome-wide 3.
  • KDM2A overexpression reduces H3K36me2 and reduces the occupancy of LEDGF and of WILD-TYPE MLL - and not of the fusion 3.
  • The abstract-level summary is that KDM2A antagonises MLL-associated leukaemogenesis 3.

The two statements are compatible and the difference between them is the finding. A CXXC reader sharing the fusion's element class, overexpressed, moved the wild-type protein and left the oncoprotein in place. Any model in which the twelve CXXC readers compete for one pool of CpG islands has to explain why the fusion is the one that does not lose. Candidate explanations the literature supports: the fusion has three AT-hooks and a menin-LEDGF tether in addition to the CXXC, so displacing one contact is not enough [see mechanism-fusion-dna-targeting-cxxc-athooks]; or the fusion's residence time is set by the elongation machinery it carries rather than by DNA affinity.

Bounding, honestly: this is one overexpression experiment read from an abstract and from this graph's existing record of it, in a murine MLL leukaemia context, with H3K36me2 as the manipulated variable rather than CpG occupancy. It is not a CXXC-competition experiment and nobody has done one.

Dependency, with both numbers

KDM2A ranks 442 of 17,728 on this project's unbiased KMT2A-selectivity scan, which puts it in the top 2.5% and makes it the highest-ranked CXXC-family gene other than CXXC1 (rank 414). The arm means are -0.683 in the 11 KMT2A-rearranged lines and -0.527 in the 43 other leukaemia lines, difference -0.156 4.

And the pan-essentiality is -0.620, which is most of the signal. KDM2A is required fairly broadly; the KMT2A-rearranged increment is about a quarter of the baseline requirement. Per this project's own rule about reporting both the effect and the pan-essentiality, this is not a selective dependency, it is a broadly required gene with a modest extra requirement in an arm that is myeloid-enriched and contains no infant line. The direction is nevertheless the opposite of what the antagonism result predicts - if KDM2A opposed the fusion, KMT2A-rearranged lines should need it less, not more. That tension is unresolved and is recorded as an open question rather than explained away.

Literature bounding and the alias hazard

(KDM2A OR FBXL11 OR JHDM1A) AND (leukemia OR leukaemia) in title/abstract returns 4 records, every title read 5: the KDM2-family review, a neonatal KMT2A-rearranged ALL case report where KDM2A is incidental, the ASH1L paper above, and a Sleeping Beauty screen. The 'retired symbol' rule was applied and mattered little here - FBXL11 adds 26 records overall and JHDM1A 21, against 191 for KDM2A, and none of the extra records is a leukaemia paper. So the small leukaemia literature is real and not an artefact of the symbol.

The collision to record is in the other direction: KDM2A is unambiguous, but FBXL11 and JHDM1A are the symbols under which the CXXC biology was originally published, and KDM2B publishes as FBXL10, JHDM1B and NDY1 - four symbols for a paralogue three residues of domain architecture away. Proposed as a registry row; see the memo.

Facts and figures
uniprotQ9Y2K7
gene symbolKDM2A
protein nameLysine-specific demethylase 2A
length AA1162
alphafold IDAF-Q9Y2K7-F1
accession resolutionUniProt REST gene_exact:KDM2A AND organism_id:9606 AND reviewed:true, run 2026-09-21; one row; primary gene name read back as KDM2A and length 1162 from the same fetched record.
alias hazardKDM2A[tiab] returns 191, FBXL11 26, JHDM1A 21; no extra leukaemia record is reached through the retired symbols. The paralogue KDM2B publishes under FBXL10, JHDM1B and NDY1, so a family-level query needs six symbols.
functionH3K36me2 demethylase recruited to non-methylated CpG islands by its ZF-CxxC domain; SCF substrate-recognition subunit
classificationCXXC-family CpG reader with the address closest to KMT2A's and the opposite output; a demonstrated antagonist of MLL-associated leukaemogenesis that does NOT displace the fusion
shown for fusionMLL leukaemia models in mouse, as an ANTAGONIST - and the occupancy readout is explicitly that wild-type MLL and LEDGF move while the fusion does not; nothing in human KMT2A::AFF1 or B-lineage
cell type evidencemouse and human somatic cell lines (CpG-island recruitment); murine MLL leukaemia models (antagonism); DepMap leukaemia lines. No infant material, no primary KMT2A-rearranged sample.
epigenetic readoutH3K36me2 at CpG islands; LEDGF and wild-type MLL occupancy
subnuclear localizationNucleoplasm (ECO:0000269) and chromosome (ECO:0000269); UniProt also records association with centromeres and a requirement to maintain the heterochromatic state. Nucleated at non-methylated CpG islands genome-wide (PMID 20417597). A combined KDM2A/KDM2B expression correlation with nucleolar size exists in breast carcinoma but there is no direct nucleolar localisation for KDM2A. Never imaged in any KMT2A-fusion cell.
cell typeMouse and human somatic cell lines (CpG-island recruitment); murine MLL leukaemia models (the antagonism result); DepMap leukaemia lines. No infant material and no primary KMT2A-rearranged patient sample.
domains
JmjC148.0JmjC: 148CXXC (CpG-binding zinc fi…564.0CXXC (CpG-binding zinc finger): 564PHD (plant homeodomain) z…617.0PHD (plant homeodomain) zinc finger: 617F-box889.0F-box: 889
aliases retired or alternate
[
  "FBXL11",
  "JHDM1A",
  "FBL7"
]
domains — the numbers behind the chart
[
  {
    "name": "JmjC",
    "start": 148,
    "end": 316,
    "pfam": "PF17811"
  },
  {
    "name": "CXXC (CpG-binding zinc finger)",
    "start": 564,
    "end": 610,
    "pfam": "PF02008"
  },
  {
    "name": "PHD (plant homeodomain) zinc finger",
    "start": 617,
    "end": 678,
    "pfam": "PF16866"
  },
  {
    "name": "F-box",
    "start": 889,
    "end": 936,
    "pfam": "PF12937"
  }
]
localization
{
  "subcellular": "nucleus, nucleoplasm (ECO:0000269) and chromosome (ECO:0000269)",
  "subnuclear": "nucleoplasm; also reported at centromeres and required to maintain the heterochromatic state (UniProt); a combined KDM2A/KDM2B expression correlation with nucleolar size exists in breast carcinoma but no direct nucleolar localisation for KDM2A",
  "localization_shown_for_fusion": "NO - nobody has imaged KDM2A in any KMT2A-fusion cell"
}
essentiality
{
  "depmap_scan": "analyses/depmap-kmt2a-selective-scan-full",
  "difference_kmt2a_r_minus_other": -0.1557,
  "mean_kmt2a_rearranged": -0.6832,
  "mean_other_leukemia": -0.5275,
  "pan_essentiality_all_lines": -0.6203,
  "n_all_lines": 1208,
  "rank_of_17728": 442,
  "percentile": 2.493,
  "interpretation": "broadly required with a modest extra requirement in the KMT2A-rearranged arm; NOT a selective dependency",
  "release_note": "Open Targets dataVersion 26.06, read 2026-09-21"
}
cxxc dna kd conditions
{
  "_why": "Added by king sweep 2026-09-21 (cycle 46). This entry quotes the KMT2A CXXC-DNA 44 uM. That number is a property of five choices, not of the molecule, and quoting it bare is how this graph manufactured a 1,300-fold 'discrepancy' with the menin contact.",
  "value": "44 uM",
  "construct": "isolated CXXC domain, KMT2A 1147-1203, untagged, 15N-labelled",
  "ligand": "DNA pal oligonucleotide, one CpG element",
  "method": "NMR titration",
  "buffer": "50 mM phosphate, pH 7.1, 300 mM NaCl, 2 mM DTT, 25 C",
  "nacl_mM": 300,
  "source_pmid": "20010842",
  "source_quote": "\"Determination of the dissociation constants by NMR for the CXXC domain (MLL 1147-1203) and mutants was performed by titration of 15N-labeled protein with concentrated solutions of DNA pal oligonucleotide. Experiments were carried out in 50 mM phosphate buffer (pH 7.1), 300 mM NaCl, 2 mM DTT at 25 C.\" -- Cierpicki 2010 NSMB, Methods, fetched and read from PMC2908503 on 2026-09-21. The result sentence reads \"The wild type protein yielded a Kd of 44 uM\".",
  "this_is_the_high_salt_end": "300 mM NaCl is roughly twice physiological ionic strength and SIX TIMES the 50 mM the menin-MBM1 53 nM was measured in. The same domain measured tagless at 150 mM NaCl gives 5.7 uM -- 7.7-fold tighter from salt alone.",
  "do_not_compare_naively": {
    "against": "menin-MBM1 53 nM (PMID 20961854), measured at 50 mM NaCl by FP",
    "best_matched_pair": "108-fold, not ~1,000-fold",
    "see": "ana-the-menin-affinity-comparison-was-made-at-one-sixth-the-salt"
  },
  "added": "2026-09-21"
}
How this could bite us back
  • KDM2A is the CXXC domain a KMT2A-CXXC-directed molecule is most likely to hit, because the reader chemistry is the closest match in the family. It is also broadly required across cell lines (-0.620), so cross-reactivity would carry a general toxicity rather than a tissue-restricted one.
  • If KDM2A antagonises the fusion's leukaemogenicity, inhibiting it would be the wrong direction - and a pan-CXXC molecule would do exactly that while also hitting KMT2A. The two effects oppose and nothing predicts the net.
Still open
  1. Why are KMT2A-rearranged lines slightly MORE dependent on KDM2A when KDM2A antagonises MLL-associated leukaemogenesis? A lineage-split of the 11 lines and a comparison against the KDM2B paralogue would show whether the DepMap signal is fusion-related at all.
  2. The KDM2A experiment displaced wild-type MLL and LEDGF but not the fusion. Is that because the fusion has more anchors, or because the fusion's residence time is set by elongation rather than by DNA affinity? A degron time-course on the fusion under KDM2A overexpression separates them.
  3. Does KDM2A occupy the same CpG islands as the fusion in a KMT2A-rearranged cell? There is no published ChIP pair - the island-class overlap is assumed, never measured in this disease.
Every source behind this entry (6)
  1. E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

    KDM2A (Q9Y2K7, 1,162 aa) carries a CXXC-type zinc finger at 564-610, a PHD finger at 617-678, a JmjC domain at 148-316 and an F-box at 889-936, and its curated DOMAIN comment states the CXXC preferentially recognises non-methylated CpG DNA with binding blocked by methylation.

    What this evidence cannot show

    Fetched 2026-09-21; accession resolved by gene_exact:KDM2A AND organism_id:9606 AND reviewed:true, one row, primary gene name read back as KDM2A and length 1162 from the same record. Five isoforms are annotated and the feature coordinates are for the canonical sequence; nothing here says which isoform is expressed in a blast.

    Evidence that argues against this (1)
    • UniProt also records KDM2A as 'required to maintain the heterochromatic state' and associating with centromeres and repetitive sequence, which is not obviously compatible with an address defined by non-methylated CpG islands; the protein evidently has more than one mode of chromatin engagement. UniProt Consortium 2026
  2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    CpG islands directly recruit KDM2A through a ZF-CxxC domain that preferentially recognises non-methylated CpG and whose binding is blocked by CpG methylation, and nucleation there removes H3K36 methylation to create CpG-island chromatin uniquely depleted of that mark.

    CpG islands recruit a histone H3 lysine 36 demethylase. · Blackledge NP, Zhou JC, et al. · Mol Cell 2010
    PMID 20417597DOI
    What this evidence cannot show

    Mouse and human somatic cell lines; genome-wide occupancy plus in-vitro binding. Not a leukaemia system and not a fusion system. The claim is about CpG islands as a class, so it does not establish occupancy at any specific KMT2A-fusion target.

    Evidence that argues against this (1)
    • The systematic family comparison classifies the human CXXC domains into several DNA-selectivity groups rather than one, so 'recognises non-methylated CpG' understates how different KDM2A's and KMT2A's sequence preferences may be. Xu C et al. 2018
  3. E3moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    KDM2A, an H3K36me2 demethylase and Polycomb-group silencing protein, antagonises MLL-associated leukaemogenesis; in the same study KDM2A overexpression reduces H3K36me2 and reduces chromatin occupancy of LEDGF and of wild-type MLL, but not of the fusion.

    ASH1L Links Histone H3 Lysine 36 Dimethylation to MLL Leukemia. · Zhu L, Li Q, et al. · Cancer Discov 2016
    PMID 27154821DOI
    What this evidence cannot show

    Abstract read from the eutils record this session; full text not read here. The occupancy detail - that wild-type MLL and LEDGF move and the fusion does not - is recorded in this graph's existing entries `mark-h3k36me2-3` and `mechanism-does-the-fusion-need-the-ledgf-pwwp-arm`, which were read before writing this one. The manipulation is KDM2A OVEREXPRESSION, which is not a physiological test of competition, and the variable acted on is H3K36me2, not CpG occupancy. Murine MLL leukaemia models.

    Evidence that argues against this (1)
    • If KDM2A antagonises MLL-associated leukaemogenesis, KMT2A-rearranged cell lines should be LESS dependent on it; the unbiased DepMap scan finds them slightly MORE dependent (-0.683 versus -0.527). The two observations point in opposite directions and neither is resolved. Oncofusion knowledge-graph project 2026
  4. E4moderate confidence 3 of 5: One strong study, or several smaller consistent ones.dataset

    KDM2A ranks 442 of 17,728 for KMT2A-rearranged selectivity with a difference of -0.156 between arms, but its pan-essentiality across 1,208 lines is -0.620, so most of the requirement is general rather than fusion-specific.

    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.

    Evidence that argues against this (1)
    • KDM2B is the paralogue and sits at rank 3,765 with a difference of -0.046, so the family-level signal is not reproduced by the nearest relative and single-gene knockouts in a paralogue pair are weak evidence either way. Oncofusion knowledge-graph project 2026
  5. E5moderate confidence 3 of 5: One strong study, or several smaller consistent ones.database

    The leukaemia literature naming KDM2A is four records in title and abstract across its current and two retired symbols, and only one of them is a mechanistic study in KMT2A-rearranged disease.

    PubMed E-utilities esearch record counts with querytranslation inspected and every returned title read · Oncofusion knowledge-graph project · NCBI E-utilities, queries run 2026-09-21; exact query strings given in the caveat of each item that uses this source 2026
    What this evidence cannot show

    Exact query: `(KDM2A[tiab] OR FBXL11[tiab] OR JHDM1A[tiab]) AND (leukemia[tiab] OR leukaemia[tiab])`, run 2026-09-21, 4 records, every title read. Bounded by titles and abstracts only - per rule 80, a KDM2A measurement appearing solely in a supporting figure of a leukaemia paper is invisible to this count, and that is exactly where occupancy measurements usually live. Bare-operand control: KDM2A[tiab] alone returns 191, so the instrument can return a one.

    Evidence that argues against this (1)
    • The KDM2 family review states that KDM2A and KDM2B are frequently deregulated across a broad spectrum of human cancers, so a small leukaemia-scoped title/abstract count is not evidence that the protein is unimportant in blood. Andricovich J et al. 2023
  6. E6weak confidence 2 of 5: Preliminary, small, indirect, or not yet peer reviewed.paper

    Combined KDM2A and KDM2B expression levels correlate with nucleolar size and prognosis in primary breast carcinomas, placing both paralogues in the nucleolar/ribosome-biogenesis axis.

    PMID 32901907DOI
    What this evidence cannot show

    Immunohistochemistry on breast carcinoma with nucleolar morphometry - a correlation in a different tissue and a different disease. Used here only to show the nucleolar connection is not unique to the single KDM2B report, not as evidence for anything in leukaemia.

    Evidence that argues against this (1)
    • The primary nucleolar result is for KDM2B/JHDM1B specifically, binding the transcribed region of rDNA; a combined-expression correlation cannot attribute the nucleolar phenotype to KDM2A. Frescas D et al. 2007
Connected to