KDM4B (JMJD2B) - the weakest of the four H3K9me3 demethylases, one of the two erasers knocked down in the only KDM4-perturbation arm in any archive that touches this mark
also called JMJD2B · JHDM3B · KIAA0876 · lysine-specific demethylase 4B
KDM4B is an enzyme that strips the strongest silencing tag off DNA packaging proteins, and it is the weakest of a family of four that do the same job. It matters here because it is one of the two enzymes switched off in the only public experiment that perturbs this tag's erasers.
The same thing, for a scientist
**Minimal entry, written because the corpus needed an edge target and had none.** `gene-kdm4b` did not exist before 2026-09-23 although KDM4B is named in five node files, and `gene-kdm4c` has existed since the target-rewrite beat. This entry is deliberately narrow: identity, enzymology, and the one thing this beat measured. **Identity.** **O94953**, `KDM4B_HUMAN`, **1,096 aa**. Gene synonyms `JMJD2B`, `JHDM3B`, `KIAA0876`. Domains from the live UniProt feature table: **JmjN 15-57, JmjC 146-309, Tudor 917-974, Tudor 975-1031** [E1]. **Enzymology, and the part that distinguishes it from its paralogues.** KDM4B specifically demethylates H3 Lys-9 and does not touch H3K4, H3K27, H3K36 or H4K20. UniProt records it as **only able to demethylate trimethylated H3K9, with a weaker activity than KDM4A, KDM4C and KDM4D** [E1]. The founding family paper established the JMJC-domain trimethyl-reversal chemistry for JMJD2 [E2]. **So of the four, KDM4B is the one whose loss should matter least on its own** - which is why the deposit this beat reads knocks it down *together with* KDM4C rather than alone, and why that arm cannot attribute any effect to either enzyme separately. **What this beat measured.** In MCF-7, KDM4B's occupancy at the seven stored *KMT2A* intervals is reported in `ana-q116-both-kdm4-erasers-are-elevated-across-the-whole-kmt2a-locus-and-neither-picks-out-the-anchor`, and the consequence of removing both KDM4B and KDM4C for H3K9me3 at the locus is reported in `ana-q116-knocking-down-both-kdm4-erasers-raises-h3k9me3-at-active-promoters-and-not-over-the-kmt2a-gene-body`. **Literature bound, stated with the query.** `KDM4B[tiab] AND KMT2A[tiab]` returns **3**; `KDM4B[tiab] AND MLL[tiab]` returns **3**; the retired alias `JMJD2B[tiab] AND MLL[tiab]` returns **3**; `KDM4B[tiab] AND leukemia[tiab]` returns **8**, against **203** for `KDM4B[tiab]` alone. **KDM4B is essentially absent from the KMT2A literature**, and that is a bibliographic fact bounded by titles and abstracts only [E3].
No three-dimensional model is available for this protein, so the map above is the whole picture. That is itself informative: AlphaFold does not model chains this long, and the parts that have been crystallised are held under the entries for the individual domains.
Why the paralogue distinction is load-bearing here and not pedantry
This graph already holds gene-kdm4c, written because KDM4C is recruited by MLL fusions through an interaction domain in the MLL 5' moiety and is required for MLL-fusion leukaemia in mouse models, with the KDM4C inhibitor SD70 preferentially suppressing MLL-fusion lines including SEM. None of that is a KDM4B result. That entry's own open question asks whether SD70 sensitivity in SEM survives rescue by re-expression of KDM4A or KDM4B, which is the experiment that would separate 'KDM4C is the target' from 'the family is the target' - and it has not been done.
KDM4B enters this project by a completely different door: it is half of the shKDM4B_shKDM4C arm of GSE139196, the only public perturbation of any H3K9me3 eraser in any archive whose readout is H3K9me3 ChIP-seq. That arm was built to study oestrogen-induced Pol II pause release at ER target promoters and has nothing to do with leukaemia; this project is reusing it.
The consequence for reading any result from that arm: a double knockdown of a weak enzyme and a strong one cannot attribute an effect to either, and a null from it cannot exonerate either. It is also not a clean KDM4-family knockdown, because KDM4A and KDM4D are untouched and KDM4A is the strongest H3K9me3 demethylase of the four.
| gene symbol | KDM4B |
|---|---|
| uniprot | O94953 |
| length AA | 1096 |
| ensembl | ENSG00000127663 |
| role | H3K9me3/me2 demethylase (JmjC family); weakest H3K9me3 demethylase of KDM4A-D by UniProt's summary. Here: one of the two erasers removed in the shKDM4B+shKDM4C arm of GSE139196. |
| accession resolution | gene_exact:KDM4B AND organism_id:9606 AND reviewed:true returned exactly ONE reviewed record (O94953, 1,096 aa), so the rule-62 several-reviewed-records trap does not fire. data/symbol-collisions.json was grepped for KDM4B before the first query and has no row; the neighbouring row for EHMT2/G9a warns that a retired alias can be the better operand, so JMJD2B was run separately and returns the same conjunction count as KDM4B. |
| eraser of that mark | H3K9me3 (and H3K9me2) |
| epigenetic readout | H3K9me3 demethylation; occupancy measured at the KMT2A locus in MCF-7 by this beat |
| cell type | measured here in MCF-7 only. No KDM4B occupancy measurement exists in any haematopoietic or KMT2A-rearranged cell. |
| cell state | hormone-deprived culture and 40 min oestradiol, in the deposit read here |
| shown for fusion | NO - nobody has looked in any KMT2A fusion. The family member with a KMT2A-fusion literature is KDM4C, not KDM4B, and gene-kdm4c's open question about paralogue rescue is precisely the experiment that would tell whether KDM4B substitutes. |
| absence class | never-measured |
| absence class note | For the specific question 'is KDM4B at the KMT2A locus in a haematopoietic cell', no instrument has been pointed at it in any archive. For MCF-7 the class was measured-not-analysed until 2026-09-23 and is now closed. |
| locus | chr19:4,969,113-5,153,595 (+), GRCh38, from GENCODE v29 gene records fetched from EBI 2026-09-23 |
| ensembl provenance | ENSG00000127663 was WRITTEN FROM RECALL and then verified against the GENCODE v29 GTF fetched this session, which is the wrong order (the project rule is fetch then write). It happened to be correct. Recorded because the order is the defect, not the outcome. |
domains — the numbers behind the chart
[
{
"name": "JmjN",
"start": 15,
"end": 57
},
{
"name": "JmjC (catalytic)",
"start": 146,
"end": 309
},
{
"name": "Tudor 1",
"start": 917,
"end": 974
},
{
"name": "Tudor 2",
"start": 975,
"end": 1031
}
]literature bound 2026 09 23
{
"KDM4B[tiab]": 203,
"KDM4B[tiab] AND KMT2A[tiab]": 3,
"KDM4B[tiab] AND MLL[tiab]": 3,
"JMJD2B[tiab] AND MLL[tiab]": 3,
"KDM4B[tiab] AND leukemia[tiab]": 8,
"bounded_by": "titles and abstracts only; supporting figures not searched"
}measured here
{
"occupancy": "ana-q116-both-kdm4-erasers-are-elevated-across-the-whole-kmt2a-locus-and-neither-picks-out-the-anchor",
"consequence_of_removal": "ana-q116-knocking-down-both-kdm4-erasers-raises-h3k9me3-at-active-promoters-and-not-over-the-kmt2a-gene-body",
"kmt2a_gene_body_percentile_of_mcf7_ctcf_anchors": [
81.32,
69.2,
83.38,
78.06
],
"exon11_percentile": [
84.78,
0,
92.26,
0
],
"verdict": "ORDINARY at the anchor; LOCUS-WIDE ELEVATED over the 90 kb gene body"
}- A KDM4 inhibitor that hits the family would remove a weak enzyme (KDM4B) and a strong one (KDM4C) at once, exactly as the only public perturbation does - so the published double-knockdown phenotype is the closest available model of a pan-KDM4 drug, and it was generated in a breast line studying oestrogen response. Nothing about its safety or its effect in a haematopoietic cell can be read off that.
- KDM4B is the paralogue with the least KMT2A literature (3 records) and the weakest enzymology. Targeting it selectively is the version of a KDM4 intervention with the least supporting evidence and, by UniProt's activity ranking, the least expected effect - so a selective KDM4B agent should be regarded as unmotivated here until the paralogue-rescue experiment is done.
- Does KDM4B substitute for KDM4C in a KMT2A-rearranged cell? gene-kdm4c's open question asks whether SD70 sensitivity in SEM survives rescue by KDM4A or KDM4B re-expression; that single experiment separates 'KDM4C is the target' from 'the family is the target' and nothing in the cited literature does it.
- Is KDM4B expressed in infant KMT2A::AFF1 blasts at all? This entry has no expression data in any leukaemia, and an eraser that is not expressed cannot be a therapeutic handle.
- What does KDM4B alone do to H3K9me3 at the KMT2A locus? The only public perturbation removes KDM4B and KDM4C together, and a double knockdown of a weak and a strong enzyme cannot attribute an effect to either.
- E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database
KDM4B is O94953, KDM4B_HUMAN, 1,096 aa, with gene synonyms JMJD2B, JHDM3B and KIAA0876; it specifically demethylates H3 Lys-9, does not demethylate H3K4, H3K27, H3K36 or H4K20, and is only able to demethylate trimethylated H3K9, with a weaker activity than KDM4A, KDM4C and KDM4D. Feature-table domains: JmjN 15-57, JmjC 146-309, Tudor 917-974 and Tudor 975-1031.
UniProtKB O94953 (KDM4B_HUMAN), Lysine-specific demethylase 4B · UniProt Consortium · UniProt 2026What this evidence cannot show
Fetched live 2026-09-23 via gene_exact:KDM4B AND organism_id:9606 AND reviewed:true, which returned exactly one record - so the rule-62 multiple-reviewed-record trap does not apply here, and data/symbol-collisions.json was grepped for KDM4B first and has no row for it. The 'weaker activity than KDM4A, KDM4C and KDM4D' statement is UniProt's own summary of in vitro comparisons and is not a measurement made in any cell type relevant to this project.
Evidence that argues against this (1)
- UniProt's relative-activity ranking is an in vitro enzymology statement. Cellular activity depends on expression, recruitment and substrate access, and a weak enzyme that is highly expressed and correctly targeted can dominate a locus. Nothing in this entry establishes that KDM4B is the minor eraser at any particular site, and this beat's own data cannot separate KDM4B from KDM4C because the only perturbation available removes both. source
- E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper
The JMJD2/KDM4 family reverses histone lysine trimethylation through a JmjC domain, the founding demonstration that trimethyl marks are enzymatically reversible.
Reversal of histone lysine trimethylation by the JMJD2 family of histone demethylases · Whetstine JR, Nottke A, Lan F, Huarte M, Smolikov S, Chen Z, Spooner E, Li E, Zhang G, Colaiacovo M, Shi Y · Cell 2006What this evidence cannot show
Abstract-level only; the full text was not read this session. This is the family paper, not a KDM4B-specific paper, and it is cited here for the chemistry rather than for any KDM4B-specific claim.
Evidence that argues against this (1)
- No paper in this evidence list establishes a role for KDM4B at the KMT2A locus, in a KMT2A-rearranged cell, or in any haematopoietic cell. The literature bound in E3 is the honest statement of what is known, and it is close to nothing. source
- E3moderate confidence 3 of 5: One strong study, or several smaller consistent ones.database
KDM4B is essentially absent from the KMT2A literature: KDM4B[tiab] AND KMT2A[tiab] returns 3 records, KDM4B[tiab] AND MLL[tiab] returns 3, JMJD2B[tiab] AND MLL[tiab] returns 3, and KDM4B[tiab] AND leukemia[tiab] returns 8, against 203 for KDM4B[tiab] alone.
PubMed esearch via scripts/pubmed_safe.py, 2026-09-23 · Q116-h3k9-axis-at-the-locus · NCBI PubMed 2026What this evidence cannot show
Bounded by TITLES AND ABSTRACTS only (rule 80): a KDM4B measurement in a supporting figure of a KMT2A paper is invisible to every one of these queries, and this project has already been caught by exactly that on an H3K79 query. Both the approved symbol and the retired alias JMJD2B were run separately, each with its own [tiab] tag (rule 94), and neither the record sets nor the counts were unioned into a single number. The three records returned by each conjunction were NOT read, so their relevance is unassessed - this is a count, not a review.
Evidence that argues against this (1)
- The paralogue KDM4C has a real KMT2A literature - it is recruited by MLL fusions and required for MLL-fusion leukaemia in mouse models - so 'the KDM4 family is absent from the KMT2A literature' would be false. The absence is specific to KDM4B, and the family-level question is live and open, which is exactly why gene-kdm4c's own open question asks whether KDM4B can rescue SD70 sensitivity. source
paralogue; the two are knocked down TOGETHER in the only public H3K9me3-eraser perturbation, so neither can be attributed an effect separately
JmjC demethylase of H3K9me3; UniProt records it as the weakest of KDM4A-D at that substrate
four KDM4B ChIP-seq tracks and one of the two shRNAs of that deposit's eraser-knockdown arm
the paralogue whose re-expression that experiment would test for rescue of SD70 sensitivity
one of the two erasers removed; its occupancy and the consequence of its loss at the KMT2A locus, measured for the first time
first KDM4B occupancy measurement at the KMT2A locus in any cell type: 79.69th percentile median over the 90 kb, ORDINARY at the anchor with a 92.3-point replicate spread