Oncofusion

start / graph / gene-ncoa6

Genemoderate confidence 3 of 5: One strong study, or several smaller consistent ones.Q44-kmt2-family

NCOA6 (ASC-2) - the subunit that gives MLL3 and MLL4 their target genes, and a recorded near-absence in leukaemia

also called ASC-2 · ASC2 · RAP250 · AIB3 · PRIP · TRBP · KIAA0181 · ASCOM (the complex)

NCOA6 is the part of the MLL3/MLL4 machine that decides which genes it goes to. Despite being essential to how that machine works, almost nobody has asked what it does in leukaemia - this entry records that gap with the exact searches, rather than inventing a role for it.

The same thing, for a scientist

NCOA6/ASC-2 (UniProt Q14686, 2,063 aa) is the coactivator that gives the complex its historical name, ASCOM (ASC-2 complex). It associates with MLL3 and MLL4, and the founding genetic demonstration is a specificity result: mice expressing a SET-domain mutant of MLL3 share phenotypes with isogenic ASC2+/- mice; expression and H3K4 trimethylation of the retinoic-acid-receptor target *RARbeta2* are impaired in ASC-2-null MEFs and in MEFs with siRNA against both MLL3 and MLL4; MLL3 and MLL4 are in **distinct** ASC-2-containing complexes rather than one common complex; and they are recruited to *RARbeta2* by ASC-2 - while *RARbeta2* expression is **intact in MEFs devoid of menin**, a component of the MLL1 and MLL2 complexes [E1]. That last clause is the clean branch separation this graph needs and it is 20 years old. **In leukaemia, NCOA6 is essentially unstudied**: `(NCOA6 OR ASC-2 OR RAP250 OR AIB3 OR PRIP) AND (leukemia OR leukaemia)` returns 13 records, title-checked, and none reports an NCOA6 experiment in a leukaemia model; `ASCOM AND (leukemia OR leukaemia OR MLL OR KMT2)` returns 3 [E2]. This entry is written as a bounded absence with the queries recorded, because a padded roster is worse than a recorded gap.

Go deeper
The one substantive result, and what it settles

1 is a target-specificity experiment and it answers a question this graph keeps asking in other forms: what decides where a COMPASS-family enzyme goes?

The answer for the MLL3/MLL4 branch is the coactivator. ASC-2 recruits MLL3 and MLL4 to RARbeta2; removing ASC-2 or both enzymes impairs both expression and H3K4me3 there; a SET-domain point mutant of MLL3 phenocopies ASC-2 haploinsufficiency. And the menin control is explicit - the same target gene is unaffected in menin-null MEFs, because menin belongs to the other branch.

So the two branches of this family are targeted by different adaptors: menin and LEDGF for KMT2A/KMT2B, ASC-2/NCOA6 (and PTIP/PA1) for KMT2C/KMT2D. A menin inhibitor can therefore only reach the MLL3/4 arm through chromatin competition, never through a shared adaptor - which is exactly the mechanism resistance-mll34-utx-tumour-suppressor-failure proposes, and this is independent 2006 evidence for the premise that entry needs.

The finding that MLL3 and MLL4 sit in distinct ASC-2 complexes rather than one is also worth carrying: the field routinely writes 'MLL3/4' as a unit and this graph has inherited that habit.

The bounded absence, stated properly

Queries run 2026-09-20 with querytranslation inspected 2:

querycountwhat the titles are
(NCOA6 OR ASC-2 OR RAP250 OR AIB3 OR PRIP) AND (leukemia OR leukaemia)13KMT2C in GI cancer, ETV6 reviews, MPAL genetics, optical genome mapping, an AIEOP-BFM workflow paper, oestrogen/GREB1, UGT1A1, Pygo2 glioma, MAFA/MAFB islets, bile-acid transporters, breast MLL3 mutations, the 2006 specificity paper, and a 2000 Tax paper that matched on 'ascorbate'
ASCOM AND (leukemia OR leukaemia OR MLL OR KMT2)3two of the above plus a breast-cancer MLL3 mutation paper

Not one is an NCOA6 experiment in a leukaemia system. The 13 include two pure retrieval artefacts - the alias PRIP and the word ascorbate - which is the alias-precision failure mode this project logged for NELFE's RD.

What would change this: NCOA6 appearing in a leukaemia dependency screen, in a KMT2A-fusion interactome, or in a menin-inhibitor resistance screen. None of the three has been reported. This is recorded as a gap, not as evidence that NCOA6 does not matter - nobody has looked.

Facts and figures
uniprotQ14686
length AA2063
gene primary verifiedNCOA6
accession resolutiongene_exact:NCOA6 AND organism_id:9606 AND reviewed:true returned exactly one row; gene_primary equals NCOA6; length 2063 asserted. NOTE the one-character difference from KMT2D's O14686 (5537 aa), a subunit of the same complex
complex branchMLL3/MLL4 COMPASS-like; the complex's historical name ASCOM derives from this protein
roletarget-gene specificity determinant - recruits MLL3 and MLL4 to nuclear-receptor target genes
menin branch contrastRARbeta2 expression is intact in menin-null MEFs, establishing that the menin branch and the ASC-2 branch have separate adaptors
leukaemia evidencenone found; bounded absence recorded with queries
pubmed counts 2026 09 20
{
  "(NCOA6 OR ASC-2 OR RAP250 OR AIB3 OR PRIP) AND (leukemia OR leukaemia)": 13,
  "ASCOM AND (leukemia OR leukaemia OR MLL OR KMT2)": 3
}
aliases searched
[
  "NCOA6",
  "ASC-2",
  "RAP250",
  "AIB3",
  "PRIP",
  "ASCOM"
]
aliases not searched
[
  "TRBP",
  "KIAA0181"
]
depmap kmt2a selectivity
{
  "chronos_mean_kmt2a_rearranged": -0.0661,
  "chronos_mean_other_leukaemia": -0.1287,
  "difference": 0.0626,
  "pan_essentiality_all_lines": -0.2015,
  "selectivity_rank": 15601,
  "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
  • NCOA6 is a broad nuclear-receptor coactivator, so anything that perturbs it reaches retinoic-acid, thyroid, PPAR and oestrogen signalling at once. There is no plausible selective intervention here and none has been attempted.
  • A recorded absence invites someone to fill it badly. The honest state of this gene is 'unstudied in leukaemia', and the risk is that a future cycle reads the ASCOM biology, finds it mechanistically attractive, and writes a role the literature does not support.
Still open
  1. Does NCOA6 appear in any leukaemia dependency screen? This is a database question, not an experiment, and it has not been asked in this project. If NCOA6 scored in DepMap's leukaemia lines it would convert a recorded gap into a target.
  2. Is NCOA6 in a KMT2A-fusion interactome? This graph has catalogued dozens of fusion interactome studies. None names NCOA6, and the reason may be that the fusion has no reason to touch the other branch - which would be a useful negative if anyone stated it.
  3. Do MLL3 and MLL4 really sit in distinct ASC-2 complexes? The 2006 paper says so and the field has written 'MLL3/4' as a unit ever since, including in this graph. If they are distinct, every MLL3/4 result should be asked which enzyme it belongs to.
  4. Would NCOA6 loss confer menin-inhibitor resistance the way KMT2C/KMT2D loss is proposed to? If the mechanism is failure of the MLL3/4 activating arm, then losing the adaptor that targets that arm should do the same thing, and nobody has tested the prediction.
Every source behind this entry (3)
  1. E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    ASC-2 (NCOA6) associates with the H3K4 methyltransferases MLL3 and MLL4; mice expressing a SET-domain mutant of MLL3 share phenotypes with isogenic ASC2+/- mice; expression and H3K4 trimethylation of the RAR target gene RARbeta2 are impaired in ASC-2-null MEFs and in MEFs expressing siRNAs against both MLL3 and MLL4; MLL3 and MLL4 are found in distinct ASC-2-containing complexes rather than a common one and are recruited to RARbeta2 by ASC-2; and RARbeta2 expression is intact in MEFs devoid of menin, a component of the MLL1 and MLL2 complexes.

    Coactivator as a target gene specificity determinant for histone H3 lysine 4 methyltransferases. · Lee S, Lee DK, Dou Y, et al. · Proc Natl Acad Sci U S A 2006
    PMID 17021013DOI
    What this evidence cannot show

    Abstract fetched and read 2026-09-20; full text (PMC1622834) not obtained in this beat. Mouse embryonic fibroblasts and a single nuclear-receptor target gene, RARbeta2 - a locus chosen because ASC-2 is a nuclear-receptor coactivator, which makes the specificity result strongest in exactly the context where it is least surprising. 2006 methodology: siRNA knockdown and locus-specific ChIP, not genome-wide. **The paper's 'MLL2' is stated to be a menin-containing complex and is therefore KMT2B on this graph's reading, but the paper gives no accession and the identification is inference from the MLL1/MLL2 versus MLL3/MLL4 grouping it uses.**

    Evidence that argues against this (1)
    • A later quantitative survey of all six human SET1/MLL complexes assigns PTIP and PA1 as the distinguishing MLL3/4 subunits and does not identify NCOA6 among the stoichiometric interactors it reports, so ASC-2's status as the defining MLL3/4 subunit rests on the earlier targeted work rather than on the unbiased measurement. van Nuland R et al. 2013
  2. E2moderate confidence 3 of 5: One strong study, or several smaller consistent ones.dataset

    PubMed run 2026-09-20 with querytranslation inspected: `(NCOA6 OR ASC-2 OR RAP250 OR AIB3 OR PRIP) AND (leukemia OR leukaemia)` returns 13 records and `ASCOM AND (leukemia OR leukaemia OR MLL OR KMT2)` returns 3; all were title-checked and none reports an NCOA6 experiment in a leukaemia model, with two of the 13 being retrieval artefacts of the aliases PRIP and of the word 'ascorbate'. UniProt reviewed human record Q14686 gives gene_primary NCOA6, 2,063 aa.

    What this evidence cannot show

    A counted absence is only as good as its symbol set. The aliases run were NCOA6, ASC-2, RAP250, AIB3 and PRIP; TRBP and KIAA0181 were not run separately and TRBP in particular collides with the RNA-binding protein TARBP2, which would inflate rather than shrink a count. The complex form ASCOM was run per rule 64 and is a coined acronym with low contamination. The repository was grepped before recording this and NCOA6 appears in no entry in this graph. **Absence of published study is not evidence of biological irrelevance.**

    Evidence that argues against this (1)
    • The decisive paper for the KMT2B entry in this beat was unreachable from any KMT2B symbol query because it framed everything at the complex level. The same mechanism would hide an NCOA6 leukaemia result inside a paper that says 'MLL3/4 complex' or 'COMPASS-like', and the phrase query for those terms is so contaminated by plant literature that it gives weak assurance. van Nuland R et al. 2013
  3. E3established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

    UniProt reviewed human records fetched 2026-09-20 give NCOA6 as Q14686, 2,063 aa, and KMT2D as O14686, 5,537 aa - accessions differing by a single character, for two proteins that are subunits of the same MLL3/MLL4 complex and therefore co-occur in exactly the contexts where a transposition would not look wrong.

    What this evidence cannot show

    gene_exact queries for both symbols returned exactly one row each, with gene_primary matching the queried symbol. The near-collision is an observation about the two strings, not a documented error anywhere; it is recorded because this project has had ten identifier-from-recall incidents and this is a plausible eleventh. The length check - 2,063 against 5,537 - separates them decisively, which is why both are asserted.

    Evidence that argues against this (1)
    • This graph's logged accession traps have all been synonym or paralogue matches returned by a query, not transcription errors between similar strings, so this is a hypothetical hazard rather than a demonstrated one. Lee S et al. 2006
Connected to