Oncofusion

start / graph / gene-ing2

Proteinmoderate confidence 3 of 5: One strong study, or several smaller consistent ones.Q101-writers-and-readers

ING2 - the PHD finger that reads H3K4me3 and turns it into REPRESSION, and whose founding co-crystal is filed under the MOUSE accession

also called ING2 · inhibitor of growth protein 2 · ING1L · p33ING2 · p32

ING2 grips the same chromatin tag that this leukaemia's fusion protein is built to create - but instead of switching a gene on, ING2 uses the tag to hold a switching-off machine in place. It also grips a rare fatty molecule in the nucleus, which competes for the same finger.

The same thing, for a scientist

> 🔴 **Correction, cycle 158: ING2 has NO valid B-lineage leukaemia record.** The single study that supplied one — a miR-153-3p/ING2 paper in acute lymphoblastic leukaemia cell lines — is **retracted** (PubMed PMID 31138034, article type 'Retracted Publication', verified against PubMed 2026-09-24). E7 is withdrawn and retained only as a record of what was considered. **This simplifies the entry rather than damaging it:** that item was already the lone source pointing against every other line of evidence, which uniformly treats ING2 as growth-inhibitory. The contradiction it created is now resolved by subtraction, and the honest position is that ING2's role in B-lineage leukaemia is **unstudied**, not disputed. --- ING2 (Q9H160, 280 aa, HGNC:6063) is a 280-residue protein that is essentially one PHD-type zinc finger (212-261) plus a polybasic region (264-280) and a long disordered middle [E1]. Its PHD finger is a specific, high-affinity reader of **H3K4me3** and, more weakly, H3K4me2 - and ING2 is a native subunit of the **repressive mSin3a-HDAC1** deacetylase complex, so on DNA damage its recognition of H3K4me3 **stabilises a corepressor at proliferation-gene promoters**. That paper's own framing is the useful one here: *a new mechanism by which H3K4me3 functions in ACTIVE GENE REPRESSION* [E2]. This graph treats H3K4me3 as the activating mark a KMT2A fusion deposits and that KMT2A's own PHD3 reads; ING2 is the documented counter-example, and it had no entry. The same PHD finger is also the first identified nuclear phosphoinositide receptor, binding PtdIns5P in vivo - which is how `gene-pip4k2a`, a childhood-ALL susceptibility locus, reached for it [E3]. Two database defects are recorded here in both directions: **UniProt cross-references ZERO PDB entries for Q9H160, and an RCSB accession query on Q9H160 returns nothing - while PDB 2G6Q is the 2006 crystal structure of the ING2 PHD finger bound to an H3K4me3 peptide, mapped by SIFTS to Q9ESK4, the MOUSE orthologue at 281 aa, one residue from the human protein** [E4].

What it looks like
ING2280 aa
PHD-type zinc fingerPHD-type zinc finger · residues 212–261PBR PBR (polybasic region) · residues 264–2800100200
loading AF-Q9H160-F1…
AlphaFold AF-Q9H160-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
An H3K4me3 reader that represses - which is the opposite of what this graph assumes the mark does

2 (Nature 2006, Shi X et al., Gozani lab) established the ING-family PHD fingers as a class of H3K4me3 effector domains, and made the functional point that matters here:

  • The ING PHD domains are specific and highly robust binding modules for H3K4me3 and H3K4me2.
  • ING2 is a native subunit of a repressive mSin3a-HDAC1 histone deacetylase complex and binds the trimethylated species with high affinity.
  • On DNA damage, ING2's PHD recognition of H3K4me3 stabilises the mSin3a-HDAC1 complex at the promoters of proliferation genes.
  • ING2 modulates cellular responses to genotoxic insult, and those functions depend critically on the H3K4me3 interaction.

Why that is load-bearing for this project. This graph's model of a KMT2A fusion is a machine that deposits and sustains H3K4me3 and H3K79me2 at HOXA and MEIS1, with H3K4me3 as the activating mark - mark-h3k4me3, interface-kmt2a-phd3-h3k4me3, exp-bivalency-cuttag-kmt2a-aff1. It also holds an explicit open absence, absence-no-experiment-tests-cxxc-reader-competition-at-a-kmt2a-fusion-target, about competition for the fusion's DNA address. ING2 is the analogous competition one layer up - competition for the MARK rather than for the DNA - and the reader that wins can invert the transcriptional output of the same mark. No experiment anywhere has put ING2 and a KMT2A fusion in the same cell.

Bounded: ING2[tiab] AND H3K4me3[tiab] = 23, ING2[tiab] AND Sin3[tiab] = 3, ING2[tiab] AND KMT2A[tiab] = 0, ING2[tiab] AND MLL[tiab] = 1 and that one record is a WDR5/fetal-globin paper, not an ING2 paper 5.

The lipid ligand, and the entry that reached for it

The same PHD finger is the first identified nuclear phosphoinositide receptor: ING2's PHD binds phosphatidylinositol 5-phosphate (PtdIns5P) in vivo, and that interaction regulates ING2's ability to activate p53 and p53-dependent apoptosis 3. UniProt places the responsiveness in the polybasic region at 264-280, immediately C-terminal to the PHD finger 1, so the module has two ligands and an adjacent basic patch that senses one of them.

gene-pip4k2a is in this graph because the 10p12 childhood-ALL susceptibility locus contains it, and it reached ING2 because PIP4K2A consumes nuclear PtdIns5P - the exact ligand. That entry records the trap that made the link invisible: PIP4K2A[tiab] AND ING2[tiab] = 0, and the connection is only visible in the lipid's vocabulary (PtdIns5P[tiab] AND ING2[tiab] = 5). This entry does not restate that reasoning; it is the ING2 half of it, and it adds the mechanistic consequence gene-pip4k2a could not state without an ING2 entry: if PtdIns5P and H3K4me3 engage the same finger, then a lipid kinase at a leukaemia risk locus is one step from setting how much of an H3K4me3 reader is available to compete with KMT2A's PHD3. Whether the two ligands actually compete is not established by either paper and is written here as an open question.

Two database defects, in both directions, on the same structure

This is the cleanest instrument finding of the beat and it was found by trying to fill pdb_ids.

queryresult (2026-09-22)
UniProt Q9H160 uniProtKBCrossReferences where database = PDB0 entries
RCSB accession-index query on Q9H160HTTP 204, no content - zero hits
RCSB entry 2G6QCrystal structure of ING2 PHD finger in complex with H3K4Me3 peptide, released 2006-07-11
2G6Q entity 1 SIFTS referenceQ9ESK4 - ING2_MOUSE, 281 aa
RCSB full-text search for ING23 hits, ranked 9M4S (Arabidopsis thaliana ING2 PHD + H3K4me3, B3H615, ING2_ARATH), then 2G6Q, then 1WES (also Q9ESK4)

So the founding co-crystal for the module this entry is about is unreachable from the human accession in either database, and a name-based search returns a plant protein first. Three separate traps stack:

  1. UniProt under-reports: zero PDB cross-references for a protein whose own DOMAIN comment says its PHD-type zinc finger mediates H3K4me3 binding.
  2. The structure is filed under the orthologue: human ING2 is 280 aa and mouse Ing2 is 281 aa - one residue apart. verify-accessions.py asserts accession plus length, so it cannot discriminate human from mouse ING2, and a cross-species accession swap would pass silently. This is the rule-66 length-discrimination failure in a cross-species rather than cross-paralogue form, which the collision registry does not currently carry a row for.
  3. The symbol is shared across kingdoms: ING2 resolves to an Arabidopsis PHD protein with its own H3K4me3 co-crystal, which outranks the mammalian ones in full-text search.

pdb_ids in this entry is therefore recorded as an empty list with a note, not as 2G6Q, because 2G6Q is not a structure of the human protein - and the note is the finding.

Localization, cell type, and the one ALL paper, which points the wrong way

Subcellular: Nucleus, one location value, ECO:0000269 PubMed:12859901 - and that citation is the phosphoinositide-receptor paper 1. Rule 91 in its commonest form: the coded localisation citation is a paper about something else that happened to contain a nuclear observation. Subnuclear: NO COMPARTMENT ASSIGNED, recorded as an absence.

Tissue: UniProt records ING2 as widely expressed, higher in colon-cancer tumour than in normal colon 1. DepMap: not a dependency anywhere - mean gene effect +0.0029 over 1,258 lines, 0.4% of lines depleted, lymphoid tissue -0.0284 over 97 lines 6. In this project's own KMT2A scan ING2 sits at rank 15,154 of 17,728 (85.5th percentile, anti-selective) on a difference of +0.0555 Chronos units between two arms that are both at zero - which is noise ranked, and is recorded only so nobody quotes the percentile.

The one leukaemia-specific record is a B-lineage one and it inverts the protein's own name. ING2[tiab] AND leukemia[tiab] = 2; one is an AIRE review, and the other reports that in acute lymphoblastic leukaemia cell lines miR-153-3p is decreased, its overexpression reduces proliferation, migration and invasion, ING2 is a direct miR-153-3p target, and ING2 overexpression REVERSES that suppression - i.e. ING2 behaves as pro-proliferative in ALL, not as an inhibitor of growth 7. It is a weak paper in a low-tier journal with unnamed cell lines, no patient material and no KMT2A status, and it is recorded because it is the only B-lineage datum that exists and because it disagrees with the tumour-suppressor framing every other source uses.

Facts and figures
uniprotQ9H160
uniprot IDING2_HUMAN
gene symbolING2
length AA280
hgnc IDHGNC:6063
ensembl gene IDENSG00000168556
alphafold IDAF-Q9H160-F1
accession assertionQ9H160 asserted explicitly. gene_exact:ING2 AND organism_id:9606 AND reviewed:true returns exactly ONE reviewed human record (2026-09-22), 280 aa, with ING1L as the gene synonym. IDENTIFIER TRAP, MEASURED: mouse Ing2 is Q9ESK4 at 281 aa - ONE residue from the human protein - so verify-accessions.py cannot discriminate the species, and the only ING2 PHD/H3K4me3 co-crystal is filed under the mouse accession. A third organism shares the symbol: B3H615 is ING2_ARATH, an Arabidopsis PHD protein, 262 aa, which outranks both mammalian entries in an RCSB full-text search for ING2.
PDB noteDELIBERATELY EMPTY, and the emptiness is a finding. UniProt Q9H160 cross-references ZERO PDB entries and an RCSB accession-index query on Q9H160 returns HTTP 204 (no hits), 2026-09-22. The founding structure of this module - PDB 2G6Q, 'Crystal structure of ING2 PHD finger in complex with H3K4Me3 peptide', released 2006-07-11 - has its polymer entity mapped by SIFTS to Q9ESK4, the MOUSE orthologue, as does 1WES ('Solution structure of PHD domain in inhibitor of growth family, member 1-like'). 2G6Q is therefore NOT listed here as a structure of the human protein. An RCSB full-text search for ING2 returns 9M4S (Arabidopsis ING2 PHD with H3K4me3) ranked FIRST. Whether the SIFTS mouse assignment is itself an error was NOT resolved - the deposited sequence was not compared with either orthologue - so no claim is made about which database is wrong, only that the human accession reaches no structure in either.
domain architecture noteA 280-residue protein that is essentially one reader module: a long disordered region (122-204), the PHD-type zinc finger at 212-261 that mediates H3K4me3 binding, and a polybasic region at 264-280 that is responsive to phosphoinositides including PtdIns5P, with a second disordered stretch at 258-280 overlapping the PBR. No catalytic domain of any kind - ING2 is a reader and a complex subunit, not an enzyme.
enzyme classNOT AN ENZYME. ING2 is a PHD-finger reader and a subunit of an mSin3A/HDAC corepressor complex; the catalytic activity in that complex belongs to HDAC1 and HDAC2.
localizationNUCLEUS, one location value on UniProt Q9H160 (fetched 2026-09-22), coded ECO:0000269 PubMed:12859901. Rule 91 in its commonest form: that citation is the PHD-finger-as-phosphoinositide-receptor paper, not a localisation study, so the code records that a curator saw a nuclear observation inside a paper about something else. No cytoplasmic value, no by-similarity value and no localisation note.
localization evidence codeECO:0000269, one citation (PubMed:12859901), and that paper WAS read at abstract level for this entry - it is about nuclear phosphoinositide binding by the PHD finger and does contain an in vivo nuclear context, so the code is defensible, but it is not a localisation study and no founding ING2 localisation paper was identified.
subnuclear localizationNO COMPARTMENT ASSIGNED, recorded as an absence rather than inferred. No nucleolar, speckle, polycomb-body or heterochromatin assignment appears in UniProt or in any source read for this entry, and none of the three primary papers read here reports an imaging experiment resolving ING2 within the nucleus. REOPENING CONDITION: ING2 immunofluorescence or chromatin fractionation in a t(4;11) line, ideally alongside H3K4me3 CUT&RUN, which is also the experiment gene-pip4k2a asks for from the lipid side.
localization shown for fusionNO - nobody has looked in any fusion. ING2[tiab] AND KMT2A[tiab] = 0 and AND MLL[tiab] = 1 (off topic), so there is no ING2 experiment of any kind in a KMT2A-rearranged cell, let alone a localisation one.
epigenetic readoutREADER, and the mark it reads is this project's central activating mark - which it converts into REPRESSION. ING2's PHD-type zinc finger (212-261) is a specific, high-affinity binding module for H3K4me3 and, more weakly, H3K4me2; because ING2 is a native subunit of the repressive mSin3a-HDAC1 deacetylase complex, H3K4me3 recognition STABILISES A COREPRESSOR at proliferation-gene promoters on DNA damage, which the founding paper frames as a mechanism by which H3K4me3 functions in ACTIVE GENE REPRESSION (PMID 16728974). ING2 is not a writer and not an eraser; the catalysis in its complex is HDAC1/HDAC2 deacetylation, so the second-order readout is loss of histone acetylation at the promoters it is held to. A SECOND LIGAND competes for the same region of the protein: the polybasic region at 264-280, immediately C-terminal to the PHD finger, is responsive to nuclear PtdIns5P, and PtdIns5P binding by the ING2 PHD regulates p53-dependent apoptosis (PMID 12859901) - which is how gene-pip4k2a, a childhood-ALL susceptibility locus, reaches this protein. BOUNDED, queries and translations read back 2026-09-22: ING2[tiab] AND H3K4me3[tiab] = 23, AND Sin3[tiab] = 3, AND PHD[tiab] = 34, AND KMT2A[tiab] = 0 (re-run, fully parsed translation with 'No items found.'), AND MLL[tiab] = 1 and off topic. NO experiment has placed ING2 and a KMT2A fusion in the same cell, and no H3K4me3-reader-competition experiment at a fusion target exists - this graph already holds that gap as absence-no-experiment-tests-cxxc-reader-competition-at-a-kmt2a-fusion-target, one layer down at the DNA address rather than at the mark.
epigenetic mark namedH3K4me3 (primary ligand, high affinity) and H3K4me2 (weaker) - READ, not written. Downstream consequence of the complex it carries: loss of nucleosomal histone acetylation by HDAC1/HDAC2. The competing non-histone ligand is the lipid PtdIns5P.
cell typeWidely expressed per UniProt, higher in colon-cancer tumour than in normal colon. Measured systems in the sources read here: human cancer cell lines under genotoxic stress (the H3K4me3-reader work), in vitro and in vivo phosphoinositide-binding systems (the PHD-receptor work), and unnamed acute lymphoblastic leukaemia cell lines (the one B-lineage record). DepMap: 1,258 lines mean +0.0029; lymphoid tissue n = 97 mean -0.0284. NO measurement in an infant t(4;11) sample, none in fetal liver, none in LMPP or pro-B, and no ING2 protein measurement in any primary leukaemia material.
cell stateDNA DAMAGE is the documented state dependence and it is the whole of it: ING2's H3K4me3-dependent stabilisation of mSin3a-HDAC1 at proliferation-gene promoters occurs IN RESPONSE TO DNA DAMAGE, and ING2's modulation of responses to genotoxic insult depends on that interaction (PMID 16728974). The p53-dependent apoptotic arm is likewise a stress-response state (PMID 12859901). No proliferative, quiescent, differentiating or niche state has been examined, and this graph holds several DNA-damage-adjacent states (exp-replication-stress-readout-kmt2a-r, exp-hr-competence-under-menin-inhibition-in-kmt2a-rearranged-blasts) in which ING2 has never been measured.
shown for fusionNO - nobody has looked in any fusion. Named per rule 86: H3K4me3 reading and corepressor stabilisation shown in human cancer cell lines under DNA damage; PtdIns5P binding shown in vitro and in vivo; a pro-proliferative role reported in unnamed ALL cell lines. No KMT2A fusion of any partner, in any lineage, in any species.
absence classliterature-absence-bounded
absence class noteMIGRATED 2026-09-23, mechanically and not by judgement: the class string itself begins 'literature'. 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: 'literature - bounded with counts, id checks, querytranslations and controls; plus one STRUCTURAL absence bounded against two databases (see pdb_note and E4).'
PDB IDS
[]
domains
[
  {
    "name": "PHD-type zinc finger",
    "start": 212,
    "end": 261,
    "pfam": null
  },
  {
    "name": "PBR (polybasic region)",
    "start": 264,
    "end": 280,
    "pfam": null
  }
]
complex membership
{
  "msin3a_like": "SIN3A + HDAC1 + HDAC2 + RBBP4/RbAp48 + RBBP7/RbAp46 + SAP30 + ING2",
  "note": "RBBP7 is also the subunit that PRMT5's H3R2me2s EXCLUDES from the H3 tail (see gene-prmt5), so this corepressor and that arginine mark meet at the same protein. No ING2-containing complex has been reported to include KMT2A, a KMT2A fusion or menin."
}
depmap
{
  "source": "DepMap CRISPR (Chronos) via Open Targets Platform GraphQL, fetched 2026-09-22",
  "release": "NOT STATED BY THE API - unverified (rule 83)",
  "is_essential": false,
  "n_screens": 1258,
  "mean_gene_effect": 0.0029,
  "fraction_lines_depleted_below_minus_0_5": 0.004,
  "lymphoid_tissue": {
    "n_lines": 97,
    "mean_gene_effect": -0.0284
  }
}
depmap kmt2a selectivity
{
  "chronos_mean_kmt2a_rearranged": 0.044,
  "chronos_mean_other_leukaemia": -0.0116,
  "difference": 0.0555,
  "pan_essentiality_all_lines": 0.0028,
  "selectivity_rank": 15154,
  "rank_denominator": 17728,
  "percentile": 85.481,
  "direction": "NOT A DEPENDENCY IN EITHER ARM - both means are at zero, so the 85th percentile ranks noise and must never be quoted as anti-selectivity. Recorded only to close off that misreading. Note also that a dropout screen cannot detect loss of a growth INHIBITOR, so this figure carries no information about ING2's function in either direction.",
  "provenance": "Joined 2026-09-22 from analyses/depmap-kmt2a-selective-scan-full (generated 2026-09-19T10:12:06Z), 11 KMT2A-rearranged and 43 other leukaemia lines; the scan records NO DepMap release (rule 83)."
}
reached for by
[
  "gene-pip4k2a - PIP4K2A consumes nuclear PtdIns5P, and PtdIns5P is the ligand bound by the PHD finger of ING2; that entry names ING2 thirty times and is the only entry in the corpus that mentions it at all"
]
How this could bite us back
  • There is no intervention here to bite back at, and that is the honest statement: ING2 has no inhibitor, no degrader, no probe compound identified in this entry, and no clinical programme. It is written as a mechanism node and as a reading hazard, not as a target.
  • If ING2 were ever targeted, the direction is against the project: it is a growth INHIBITOR by name, annotation and mechanism, holding an HDAC corepressor at proliferation-gene promoters. Removing it would be expected to release proliferation, and the only ALL-specific paper - which points the other way - is too weak to overturn that.
  • The reader-competition rationale can fail on domain loss rather than on biology: PHD1-3 are lost from essentially every KMT2A fusion, so any argument built on readers competing with KMT2A's PHD3 applies to the wild-type allele, not to the chimera.
  • An assay risk rather than a therapeutic one, and it is the reason this entry exists: pdb_ids, structure-based modelling and any Boltz-style co-folding of the ING2 PHD with H3K4me3 will silently return nothing from the human accession, because the only co-crystal is indexed under the mouse orthologue at 281 aa - one residue from the human protein, inside this project's length-discrimination limit.
  • Paralogue buffering: five ING proteins share the PHD H3K4me3-reader module, so any single-gene perturbation of ING2 may be compensated, and the DepMap mean of +0.0029 is exactly what that looks like.
Still open
  1. ING2 in B-lineage leukaemia is now an empty literature, not a contested one. Is that worth a bounded absence entry? It reads on H3K4me3, the mark this fusion's retained CXXC-adjacent machinery is built around, so an unstudied repressive reader of that mark is a real gap.
  2. Do PtdIns5P and H3K4me3 actually compete for the ING2 PHD finger and its adjacent polybasic region? The two ligands were characterised three years apart by overlapping groups and no paper read here tests them against each other - yet the whole PIP4K2A-to-chromatin link in this graph depends on the answer.
  3. Is ING2 present at KMT2A::AFF1 target promoters, where H3K4me3 is high? If an H3K4me3-reading corepressor is recruited to the fusion's own mark, the fusion would be building the binding site for a brake on itself - a self-limiting circuit nobody has looked for.
  4. Which H3K4me3 reader wins at a fusion target: KMT2A's own PHD3, ING2, or something else? PHD1-3 are lost in essentially every KMT2A fusion, so the reader competition at the mark may be entirely between non-KMT2A proteins - and this graph's existing reader-competition absence is written about the DNA address, not the mark.
  5. Does ING2's DNA-damage dependence matter in a disease treated with anthracyclines and etoposide? Its documented function only engages after genotoxic insult, and every KMT2A-rearranged treatment protocol in this graph delivers exactly that.
  6. What is ING2's direction in B-lineage leukaemia? One weak microRNA study says pro-proliferative, the protein's name and every other annotation say growth-inhibitory, and no primary leukaemia material has ever been examined.
  7. Has any human ING2 structure been deposited? The question could not be answered from either database by accession, and resolving it would also settle whether UniProt's zero PDB cross-references is a curation gap or the correct answer.
Every source behind this entry (7)
  1. E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

    ING2 (Q9H160, 280 aa) is a component of an mSin3A-like corepressor complex consisting of at least SIN3A, HDAC1, HDAC2, RBBP4, RBBP7, SAP30 and ING2; it interacts with H3K4me3 and to a lesser extent H3K4me2 through a PHD-type zinc finger at 212-261; its polybasic region at 264-280 is responsive to phosphoinositides including PtdIns5P; it is located in the Nucleus; and it is widely expressed with higher expression in colon-cancer tumour than in normal colon.

    What this evidence cannot show

    A database record, and a thin one: the FUNCTION comment is three hedged sentences ('Seems to be involved in p53/TP53 activation', 'probably involved in deacetylation'), the SUBUNIT block carries no inline citations at all, and the single Nucleus location value is coded to PubMed:12859901 - a paper about the PHD finger as a phosphoinositide receptor rather than a localisation study (rule 91). The record also cross-references ZERO PDB entries, which E4 shows to be wrong. The accession is asserted explicitly: human ING2 is 280 aa and mouse Ing2 (Q9ESK4) is 281 aa, so a length readback cannot confirm the species.

    Evidence that argues against this (1)
    • The record's corepressor framing is not the only role assigned to this protein, and the alternative is a positive one that the same record hedges: ING2 'seems to be involved in p53/TP53 activation and p53/TP53-dependent apoptotic pathways, probably by enhancing acetylation of p53'. So UniProt has ING2 both holding an HDAC complex on chromatin and enhancing acetylation of p53 - deacetylase-associated and acetylation-promoting in the same comment block - and neither statement carries an experimental evidence code in the FUNCTION text. An entry that read only the SUBUNIT block would describe a corepressor subunit; one that read only FUNCTION would describe a p53 coactivator. UniProt Consortium 2026
  2. E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    The PHD domains of the ING family are a class of methylated-H3K4 effector domains that bind H3K4me3 and H3K4me2 specifically and robustly; ING2, a native subunit of a repressive mSin3a-HDAC1 histone deacetylase complex, binds the trimethylated species with high affinity, and in response to DNA damage ING2's PHD recognition of H3K4me3 stabilises the mSin3a-HDAC1 complex at the promoters of proliferation genes - establishing a mechanism by which H3K4me3 functions in ACTIVE GENE REPRESSION - with ING2's modulation of responses to genotoxic insult critically dependent on that H3K4me3 interaction.

    ING2 PHD domain links histone H3 lysine 4 methylation to active gene repression. · Shi X, Hong T, Walter KL, Ewalt M, Michishita E, Hung T, Carney D, Pena P, Lan F, Kaadige MR, Lacoste N, Cayrou C, Davrazou F, Saha A, Cairns BR, Ayer DE, Kutateladze TG, Shi Y, Cote J, Chua KF, Gozani O · Nature 2006
    PMID 16728974DOI
    What this evidence cannot show

    The system is human cancer cell lines under genotoxic stress; there is no haematopoietic arm, no fusion and no leukaemia anywhere in this paper. 'Proliferation genes' is the paper's own category and is not a defined locus set, so this cannot be mapped onto KMT2A-fusion target genes without an experiment. The claim of repression is specific to the DNA-damage condition - the paper does not assert that H3K4me3 is repressive in general - and reading it that way would be exactly the over-generalisation this entry is trying to avoid. Abstract read in full 2026-09-22; full text not read.

    Evidence that argues against this (1)
    • A reader with high in-vitro affinity for a mark need not compete for it in a cell, and in this graph the KMT2A side of that competition is mostly moot: KMT2A's own H3K4me3 reader is PHD3, and PHD1-3 are LOST in essentially every KMT2A fusion (subtraction-kmt2a-phd3-bromo-cassette, interface-kmt2a-phd3-h3k4me3). So an H3K4me3-reader-competition argument applies to the wild-type allele and to the fusion's DOWNSTREAM readers, not to the chimera's own chromatin engagement - and if the fusion does not read the mark it writes, ING2 is competing with something else entirely. Oncofusion Knowledge Graph 2026
  3. E3established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    The PHD finger of ING2 is a nuclear phosphoinositide receptor: the PHD fingers of ING2 and other nuclear proteins bind phosphatidylinositol phosphates in vitro including the rare species PtdIns5P, the ING2 PHD finger interacts with PtdIns5P in vivo, and that interaction regulates ING2's ability to activate p53 and p53-dependent apoptosis.

    The PHD finger of the chromatin-associated protein ING2 functions as a nuclear phosphoinositide receptor. · Gozani O, Karuman P, Jones DR, Ivanov D, Cha J, Lugovskoy AA, Baird CL, Zhu H, Field SJ, Lessnick SL, Villasenor J, Mehrotra B, Chen J, Rao VR, Brugge JS, Ferguson CG, Payrastre B, Myszka DG, Cantley LC, Wagner G, Divecha N, Prestwich GD, Yuan J · Cell 2003
    PMID 12859901DOI
    What this evidence cannot show

    This paper predates the H3K4me3-reader result by three years and does NOT test whether the two ligands compete for the same surface - the lipid responsiveness is placed by UniProt in the polybasic region at 264-280, immediately C-terminal to the PHD finger at 212-261, so 'the same finger binds both' is a simplification this entry should not make. There is no leukaemia, no haematopoietic cell and no KMT2A in this paper. It is already cited once in this graph, in gene-pip4k2a evidence[5]; this entry carries it for the ING2 side of the same fact and does not restate that entry's reasoning.

    Evidence that argues against this (1)
    • The link from a lipid kinase to an H3K4me3 reader is a metabolite-level inference and the intervening step has never been measured in this or any related system. gene-pip4k2a records its own bounded absences on exactly this point - PIP4K2A[tiab] AND ING2[tiab] = 0, AND histone[tiab] = 0, AND KMT2A[tiab] = 0 - so nobody has shown that changing PIP4K2A activity changes nuclear PtdIns5P, ING2 occupancy or any H3K4me3-dependent output in a haematopoietic cell. Two real facts joined by a shared small molecule is a hypothesis, and this entry treats it as one. Oncofusion Knowledge Graph 2026
  4. E4established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

    The founding co-crystal of the ING2 PHD finger with an H3K4me3 peptide is unreachable from the human accession in both major structural resources: UniProt Q9H160 cross-references ZERO PDB entries and an RCSB accession-index query on Q9H160 returns HTTP 204 with no hits, while PDB 2G6Q is titled 'Crystal structure of ING2 PHD finger in complex with H3K4Me3 peptide', was released 2006-07-11, and its polymer entity is mapped by SIFTS to Q9ESK4 - ING2_MOUSE, 281 aa, one residue longer than the 280-aa human protein; an RCSB full-text search for ING2 returns three entries, ranked 9M4S (Arabidopsis thaliana ING2 PHD with H3K4me3, B3H615) first, then 2G6Q, then 1WES (also Q9ESK4).

    What this evidence cannot show

    This is a statement about two databases' indexing on one date, not about the science: 2G6Q is a real structure of a real ING2 PHD finger and the mouse and human PHD fingers are near-identical, so nothing about the biology is in doubt. The SIFTS mapping to the mouse accession may itself be correct - the deposited construct may genuinely be the mouse protein - in which case the defect is only UniProt's zero PDB cross-references for the human entry plus the absence of any human ING2 structure at all. That distinction was NOT resolved here: the deposited sequence was not compared with either orthologue, so this entry does not claim the SIFTS assignment is an error, only that the human accession reaches no structure.

    Evidence that argues against this (1)
    • The opposite reading is that both databases are correct and the entry's own expectation was the defect. If no human ING2 structure has ever been deposited, then UniProt reporting zero PDB cross-references is accurate reporting rather than under-reporting, and the RCSB 204 is the right answer to the question asked. What survives either way, and is the part worth carrying, is the retrieval hazard: a beat filling pdb_ids from the human accession gets an empty field for a module with a famous 2006 co-crystal, and a beat filling it from a name search gets an Arabidopsis protein - so the honest value of pdb_ids is an empty list with a note, which is what this entry records. RCSB PDB 2026
  5. E5strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

    ING2 has essentially no KMT2A literature: ING2[tiab] = 133, ING2[tiab] AND H3K4me3[tiab] = 23, ING2[tiab] AND PHD[tiab] = 34, ING2[tiab] AND Sin3[tiab] = 3, ING2[tiab] AND p53[tiab] = 31, ING2[tiab] AND leukemia[tiab] = 2, ING2[tiab] AND KMT2A[tiab] = 0 and ING2[tiab] AND MLL[tiab] = 1, and that single MLL record is a WDR5 fetal-globin-silencing paper rather than an ING2 study.

    What this evidence cannot show

    Titles and abstracts only; rule 80 applies and these counts cannot see a supporting figure. The KMT2A zero was re-run and returned a fully parsed translation - '"ING2"[Title/Abstract] AND "KMT2A"[Title/Abstract]' - together with warninglist.outputmessages = ['No items found.'], which is the signature of a genuine compound-query zero rather than the empty-translation instrument failure of registry row 123. Instrument controls through the identical code path: H3K79me2[tiab] = 224 positive, TMEM103[tiab] = 0 with the term echoed back unparsed. Registry row 124 was checked and does NOT rescue anything here: AND MLL returns 1 and it is off topic, so both operands are genuinely empty.

    Evidence that argues against this (1)
    • A gene-symbol query is the wrong instrument for this particular protein twice over, and gene-pip4k2a already demonstrated the pattern from the other side: the ING2-to-PIP4K2A link is invisible to every PIP4K2A query and visible only in the LIPID's vocabulary (PtdIns5P[tiab] AND ING2[tiab] = 5). By the same logic the ING2-to-KMT2A link, if it exists, would be written in the MARK's vocabulary rather than either gene's - and H3K4me3[tiab] alone is a four-figure literature that no beat can screen. A counted zero on two gene symbols does not bound a mark-level relationship. NCBI PubMed 2026
  6. E6strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

    ING2 is not a dependency in any DepMap arm: mean gene effect +0.0029 across 1,258 screened cell lines with 0.4% of lines depleted below -0.5, and a lymphoid-tissue mean of -0.0284 over 97 lines.

    What this evidence cannot show

    Rule 83: the API does not state the DepMap release; 1,258 is ING2's own screen count. A mean of +0.0029 is indistinguishable from zero, so in this project's KMT2A scan the rank of 15,154 of 17,728 ranks noise and must never be quoted as anti-selectivity - the arms are +0.044 (KMT2A-rearranged) and -0.0116 (other leukaemia), both zero. A viability screen is also the wrong assay for a tumour suppressor: loss-of-function of a growth inhibitor should make cells grow BETTER, which a dropout screen scores as no effect or a mild positive, and ING2's positive mean is consistent with that without being evidence for it.

    Evidence that argues against this (1)
    • ING2 is one of five ING paralogues (ING1 to ING5) sharing the PHD H3K4me3-reader module, and ING1 additionally shares a first-exon/alternative-transcript relationship with ING2's own alias ING1L. A single-gene CRISPR knockout in a family whose members share a reader domain is the textbook setup for paralogue buffering, which this graph already holds as a standing hypothesis - so 'not a dependency' here is uninformative about whether H3K4me3-reader function matters, and only a combined perturbation would test it. Shi X et al. 2006
  7. E7speculative confidence 1 of 5: Our own inference or analogy. Not shown to be true.paper

    [RETRACTED SOURCE - THIS CLAIM IS WITHDRAWN] The only B-lineage leukaemia record for ING2 assigns it a pro-proliferative role that contradicts its name: in acute lymphoblastic leukaemia cell lines miR-153-3p expression is decreased, its overexpression reduces proliferation, migration and invasion, ING2 is validated as a direct miR-153-3p target by luciferase reporter and western blot, and ING2 overexpression REVERSES the suppressive effects of miR-153-3p.

    miR-153-3p Suppresses Inhibitor of Growth Protein 2 Expression to Function as Tumor Suppressor in Acute Lymphoblastic Leukemia. · Jiang J, Liu Y, Zhao Y, Tian F, Wang G · Technology in Cancer Research and Treatment 2019
    PMID 31138034DOI
    What this evidence cannot show

    RETRACTED. PubMed lists PMID 31138034 (Jiang J et al., Technol Cancer Res Treat 2019;18:1533033819852990, doi 10.1177/1533033819852990) with the article type 'Retracted Publication', verified against PubMed on 2026-09-24. A retracted source does not support a claim at any confidence, so this item is retained only as a record of what was considered and must not be cited onward. Weak on every axis and graded accordingly. The abstract names no cell line, reports no patient material, states no KMT2A or cytogenetic status, and the readouts are CCK-8, wound healing and Transwell invasion - migration and invasion being assays of doubtful meaning for a leukaemia. It is a single-microRNA study in a low-impact journal, the mechanism is inferred from a reporter assay, and no chromatin or H3K4me3 measurement appears anywhere in it. It is recorded solely because it is the ONLY acute-lymphoblastic-leukaemia record for ING2 and because it disagrees with every other source's direction; nothing should be built on it.

    Evidence that argues against this (2)
    • The paper is RETRACTED (PubMed article type 'Retracted Publication', PMID 31138034, verified 2026-09-24). This is not a weakness to be graded down - it is a disqualification. The correct consequence is stated in the entry body: ING2 has NO valid B-lineage leukaemia record, which is a cleaner and more useful statement than the contested one this item used to make.
    • Every other source treats ING2 as growth-inhibitory - its name, its family name, its p53-activating and apoptosis-promoting annotation, and its role as a corepressor subunit that shuts down proliferation-gene promoters on DNA damage. A single microRNA-target study whose direction opposes all of that, in unnamed cell lines with invasion assays, is more likely to be reporting an artefact of the reporter system or a cell-line idiosyncrasy than a lineage-specific inversion of the protein's function. The honest position is that ING2's direction in B-lineage leukaemia is UNKNOWN, not that it is pro-proliferative. Shi X et al. 2006
Connected to