Oncofusion

start / graph / gene-hdac2

Genemoderate confidence 3 of 5: One strong study, or several smaller consistent ones.Q39-acetylation-cycle

HDAC2 - the buffering half of the class I pair, and the one class I enzyme a drug result in acute leukaemia has actually been pinned to, in AML rather than in a KMT2A fusion

also called HD2 · RPD3 · YAF1 · histone deacetylase 2

HDAC2 is the near-twin of HDAC1, and the two cover for each other so completely that removing either one alone does little. It is also the only class I eraser that a drug study in acute myeloid leukaemia has singled out by name, through a chemical called chidamide.

The same thing, for a scientist

HDAC2 (Q92769, 488 aa) is a class I deacetylase sharing the Sin3, NuRD and CoREST corepressor complexes with HDAC1. Two facts define it here. **First, it is redundant with HDAC1**: only **genetic co-depletion of Hdac1 with Hdac2** was pro-apoptotic in the functional-genetic dissection of individual Hdacs across three murine malignancies, phenocopied by the HDAC1/2-selective agent RGFP233, while the individual dependency in that study belonged to Hdac3 [E1]. **Second, it carries the one named class I attribution in acute leukaemia pharmacology**: chidamide combined with cladribine causes cell-cycle arrest and apoptosis in AML **by targeting an HDAC2/c-Myc/RCC1 axis** [E2]. That second fact needs the beat's standard caution applied immediately. Chidamide (tucidinostat) is a **benzamide class I inhibitor covering HDAC1, HDAC2, HDAC3 and HDAC10** - it is not HDAC2-selective - so the paper's attribution to HDAC2 rests on downstream molecular analysis rather than on the compound's selectivity, and the disease is AML - although, correcting a first draft of this entry, the cell-line panel does include the KMT2A::AFF1 line MV4-11 and one primary sample carried MLL-AF6, with no KMT2A-versus-non-KMT2A comparison reported. This is the same structure that made every CK2 result in this graph CK2-level rather than CSNK2A2-level. HDAC2 nevertheless matters more than its own thin literature suggests, because **chidamide is the HDAC inhibitor that appears in both of this beat's menin-inhibitor combination papers** - enhancing MI-538's restoration of HDAC7 in t(4;11) infant B-ALL, and synergising with MI-3 in MLL-rearranged AML. See `drug-chidamide`.

Go deeper
The redundancy, and why it makes HDAC2 hard to study

The Matthews design depleted individual Hdacs in MLL-AF9;Nras AML, PML-RARa APL and Emu-Myc lymphoma 1. Its per-enzyme verdicts:

  • Hdac3 - selected against in all three tumour types, individually.
  • Hdac1 alone, Hdac2 alone - no reported individual effect of the kind Hdac3 produced.
  • Hdac1 + Hdac2 together - robust pro-apoptotic response, reproduced by RGFP233 1.

So HDAC2 has no individual dependency claim, and the honest statement is not 'HDAC2 does not matter' but 'a single-gene experiment cannot tell whether HDAC2 matters'. Those are different, and the difference is the whole content of gene-hdac1 and this entry.

The practical consequence reaches this graph's own instruments. A DepMap-style single-gene CRISPR score for HDAC2 measures the effect of removing HDAC2 while HDAC1 remains, which is by construction the experiment the biology defeats. This graph has a standing rule that a DepMap number is not a result until it is lineage-split; a buffered paralogue pair adds a second condition - it is not a result until the paralogue is co-perturbed - and that condition is not currently stated anywhere else in the graph.

The chidamide/HDAC2 attribution, and how far it goes

Gu et al. report that the HDAC inhibitor chidamide synergises with cladribine to cause cell-cycle arrest and apoptosis in AML by targeting the HDAC2/c-Myc/RCC1 axis 2. It is the only paper found in this beat that names a single class I enzyme as the effector of an HDAC inhibitor in acute leukaemia, and the attribution is carried by co-immunoprecipitation showing HDAC2 interacts with c-Myc and by ChIP showing c-Myc binds the RCC1 promoter - real molecular evidence, from a compound that cannot itself discriminate.

How much it establishes:

  • The molecular axis, within the AML models used. RCC1 is the RanGEF, so an HDAC2-c-Myc-RCC1 route connects deacetylation to nucleocytoplasmic transport and mitosis, which is a specific and testable claim.
  • Not the selectivity. Chidamide inhibits HDAC1, 2, 3 and 10. The attribution comes from downstream analysis, not from the compound discriminating among its targets - and the compound cannot.
  • Partly the disease. AML. The cell-line panel includes MV4-11, which is KMT2A::AFF1-rearranged, and one primary sample carried an MLL-AF6 fusion - but no KMT2A-versus-non-KMT2A comparison is reported, so the paper contains KMT2A-rearranged material without containing a KMT2A-rearranged finding. The correction matters: an earlier draft of this entry said 'no stated KMT2A rearrangement', which was wrong, and reading the full abstract rather than the title is what caught it.

The comparison this graph should make is to its own CK2 episode: CX-4945 is ATP-competitive and pan-catalytic-subunit, so every CK2 result is CK2-level rather than CSNK2A1- or CSNK2A2-level, and naming a subunit there was naming rather than measurement. An HDAC2 attribution from a class I inhibitor has the same structure. It can be right; it is not measured by the tool used.

Facts and figures
uniprotQ92769
length AA488
gene symbolHDAC2
accession resolutiongene_exact:HDAC2 AND organism_id:9606 AND reviewed:true returned exactly 1 reviewed row; primary gene name read back as HDAC2 (rule 62 applied). Note that HDAC1 is 482 aa and HDAC2 is 488 aa, so a length check cannot discriminate this pair - only the primary-name read-back can.
classclass I histone deacetylase; shares the Sin3, NuRD and CoREST corepressor complexes with HDAC1
individual dependency claimfalse
redundant partnergene-hdac1
selective tool for the pairRGFP233 (HDAC1/2-selective)
named pharmacological attributionHDAC2/c-Myc/RCC1 axis, chidamide plus cladribine, in AML - supported by HDAC2-c-Myc co-immunoprecipitation and c-Myc ChIP at the RCC1 promoter
attribution qualitydownstream molecular analysis, not compound selectivity - chidamide covers HDAC1, 2, 3 and 10
kmt2a rearrangement in that studyMV4-11 (KMT2A::AFF1) is among the cell lines and one primary sample carried MLL-AF6, but no KMT2A-versus-non-KMT2A comparison was made
relevance route to index diseasechidamide is the HDAC inhibitor used in BOTH menin-inhibitor combination papers in this beat: with MI-538 in t(4;11) infant B-ALL and with MI-3 in MLL-rearranged AML
instrument caveata single-gene DepMap or CRISPR score for HDAC2 is structurally uninformative because HDAC1 remains present; a buffered paralogue pair requires co-perturbation before any dependency statement is meaningful
clinical trial motivatedNCT05330364 (chidamide plus cladribine in AML)
essentiality
{
  "depmap_common_essential": null,
  "note": "DepMap not queried in this beat, and a single-gene query would not be interpretable here for the reason above."
}
depmap kmt2a selectivity
{
  "chronos_mean_kmt2a_rearranged": 0.1392,
  "chronos_mean_other_leukaemia": 0.0246,
  "difference": 0.1146,
  "pan_essentiality_all_lines": -0.0267,
  "selectivity_rank": 17134,
  "rank_denominator": 17728,
  "provenance": "Joined 2026-09-21 from analyses/depmap-kmt2a-selective-scan-full (generated 2026-09-19T10:12:06Z), the project's own genome-wide scan over 11 KMT2A-rearranged and 43 other leukaemia DepMap lines. NOT re-fetched and NOT independent of that scan: every caveat it carries applies here, including that the KMT2A-rearranged arm is a small number of lines drawn from nine individuals, that the rank denominator is 17,728 genes, and that a difference is not a dependency. Added because only 26 of 308 gene entries carried this number, which made any KMT2A-selectivity question answerable for 8% of the corpus.",
  "how_to_read": "A negative difference means MORE required in KMT2A-rearranged lines than in other leukaemia lines. A rank near 17,728 is ANTI-selective. Neither is evidence of a fusion-driven dependency on its own: check pan_essentiality_all_lines before calling anything selective, and note that a paralogue-redundant gene can score near zero for reasons that have nothing to do with this disease - see ana-the-paralogue-redundancy-excuse-is-not-tested-and-my-proxy-decided-the-answer."
}
How this could bite us back
  • Any therapeutic effect requiring HDAC1 and HDAC2 to be engaged together is an effect whose toxicity is set by simultaneous inhibition of three major corepressor complexes in every tissue. That is the pan-HDAC toxicity profile arriving by a different route than pan-HDAC chemistry.
  • Naming HDAC2 as the effector of a four-isoform inhibitor is the exact move this project already had to retract for CK2. If an HDAC2 rationale is built on chidamide data, the rationale is a hypothesis about which of four enzymes matters, presented with the confidence of a measurement.
  • A DepMap or CRISPR screen returning 'HDAC2 is not a dependency' would be an artefact of buffering and could be used to close off a real target. The failure is silent and would look like good evidence.
  • Chidamide is the agent connecting HDAC2 to the menin-inhibitor combinations that this graph's therapeutic hopes rest on. If the chidamide contribution turns out to be off-target or class-wide, two separate combination rationales lose their mechanistic half at once.
Still open
  1. Is the HDAC2/c-Myc/RCC1 axis present in KMT2A-rearranged cells, and does it explain any part of the chidamide effect in the two menin-inhibitor combination papers?
  2. Does HDAC1/HDAC2 co-depletion kill KMT2A-fusion cells? The experiment exists in the same study but was run in the Emu-Myc arm.
  3. Can an HDAC2-selective degrader be built where an HDAC2-selective inhibitor cannot? A degrader removes the scaffold as well as the catalysis, and for a corepressor-complex subunit those are different perturbations.
  4. How many buffered paralogue pairs in this disease are invisible to the single-gene dependency instruments this graph uses, and is there a systematic way to enumerate them?
Every source behind this entry (3)
  1. E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    Genetic co-depletion of Hdac1 with Hdac2 was pro-apoptotic in Emu-Myc lymphoma in vitro and in vivo and was phenocopied by the HDAC1/2-specific agent RGFP233, while in the same functional-genetic dissection across three murine malignancies including MLL-AF9;Nras(G12D) AML it was Hdac3-depleted cells that were selected against in competitive assays for all three tumour types.

    Functional-genetic dissection of HDAC dependencies in mouse lymphoid and myeloid malignancies. · Matthews GM, Mehdipour P, Cluse LA, Falkenberg KJ, Wang E, Roth M, et al. · Blood 126(21):2392-2403 2015
    PMID 26447190DOI
    What this evidence cannot show

    The co-depletion result is reported in Emu-Myc lymphoma, not in the MLL-AF9;Nras arm, so nothing here establishes that HDAC1/2 co-depletion kills KMT2A-fusion cells. shRNA rather than knockout. Mouse. The absence of a reported individual Hdac2 effect is an absence in the abstract, and this entry does not read it as a demonstrated null - an absence of a reported result is not a negative result.

    Evidence that argues against this (1)
    • The buffering conclusion is drawn from a study in which the individual-enzyme result that did emerge (HDAC3) was not fusion-specific either - it held in an APL and a Myc lymphoma as well as in the MLL-AF9 model. So the class I picture in this disease is one redundant pair with no individual claim and one individually-required enzyme with no fusion specificity, and neither offers a targeting rationale. Matthews GM et al. 2015
  2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    Chidamide combined with cladribine is synergistic on cell proliferation arrest, G0/G1 arrest and apoptosis in AML cell lines (U937, THP-1, MV4-11) and primary AML cells; the combination suppresses c-MYC, HDAC2 interacts with c-Myc, c-Myc binds the RCC1 promoter and RCC1 is suppressed by the combination in a c-Myc-dependent manner; high RCC1 and high HDAC2 expression are associated with poor survival in AML patients; and the combination suppressed growth and induced apoptosis in primary cells from patients including one with an MLL-AF6 fusion.

    PMID 36849955DOI
    What this evidence cannot show

    AML, and the only KMT2A-rearranged material is a SINGLE primary sample carrying MLL-AF6 - a partner this graph does not otherwise hold - among a panel that also included AML1-ETO, c-KIT, FLT3-ITD and a CMML-BP case. MV4-11 is KMT2A::AFF1-rearranged and is among the cell lines used, but no KMT2A-versus-non-KMT2A comparison is reported. Chidamide is a benzamide class I inhibitor covering HDAC1, HDAC2, HDAC3 and HDAC10; it is NOT HDAC2-selective, so the attribution to HDAC2 comes from co-immunoprecipitation and ChIP evidence that HDAC2 interacts with c-Myc, not from the compound discriminating among its targets. The survival association is a patient-cohort correlation. Abstract read in full; PMCID PMC9972767, full text not fetched. An earlier draft of this entry carried a wrong author list and wrong pagination written from recall; all fields here were copied from a single fetched PubMed record on 2026-09-20.

    Evidence that argues against this (1)
    • This graph has already recorded the general form of the attribution problem: CX-4945 is ATP-competitive and pan-catalytic-subunit, so every CK2 result is CK2-level rather than subunit-level, and attributing to CSNK2A1 rather than CSNK2A2 was naming and not measurement. Chidamide covers four HDAC isoforms, so the HDAC2 attribution rests entirely on the co-IP and ChIP evidence and not on the pharmacology - it can be right, but the compound did not measure it. HUGO Gene Nomenclature Committee; recorded in this project as rule 63 2026
  3. E3established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

    Human HDAC2 is UniProt Q92769, primary gene name HDAC2, 488 amino acids, primary protein name 'Histone deacetylase 2'.

    UniProtKB Q92769 (HDAC2_HUMAN), Histone deacetylase 2 · UniProt Consortium · UniProtKB, retrieved 2026-09-20 2026
    What this evidence cannot show

    Resolved per rule 62: gene_exact:HDAC2 AND organism_id:9606 AND reviewed:true returned exactly one reviewed row and the PRIMARY gene name was read back as HDAC2. The 482 aa HDAC1 and 488 aa HDAC2 records are close enough in length that a length check alone would not catch a swap between them, so the primary-name read-back is the only safeguard that works for this pair.

    Evidence that argues against this (1)
    • A reviewed database record fixes the canonical sequence on the date it was read and nothing more. It does not establish which isoform is expressed in KMT2A-rearranged blasts, and this beat's own findings turn on distinctions a canonical record hides - HDAC9's catalytically inactive MITR form, and the nomenclature collision in which HDAC9's protein name reads 'Histone deacetylase 7B'. Accession and length are asserted here so a swap fails the validator, not because they carry biological weight. Vega-Garcia N et al. 2018
Connected to