Oncofusion

start / graph / subtraction-mllt3-yeats

Mechanismmoderate confidence 3 of 5: One strong study, or several smaller consistent ones.Q11b-subtraction

MLLT3 (AF9): the YEATS acetyl-lysine reader is discarded in ~92% of MLL-AF9 patients — same shape as PZP, but the causal test has never been run

also called AF9 YEATS loss · MLL-AF9 YEATS exclusion · gain by subtraction MLLT3

Normal AF9 has a reader that recognises chemically tagged histones and steers the DOT1L enzyme to the right genes. In the great majority of MLL-AF9 leukemias the fusion keeps the part that grabs DOT1L and throws that reader away — the same pattern as the AF10 partner, except nobody has yet tested whether throwing it away is what makes the cell cancerous.

The same thing, for a scientist

AF9 (MLLT3, P42568) carries an N-terminal YEATS domain at residues 1-138 that binds H3K9ac (and more weakly H3K27ac/H3K18ac) and is required for chromatin recruitment of DOT1L. Approximately 92% of MLL-AF9 patients have breakpoints at or after AF9 residue 376, excluding YEATS while retaining the ANC1 homology domain (AHD) that docks DOT1L, AF4/AFF1 and the AEP/SEC machinery. The architecture is structurally identical to the MLLT10 case: retain the recruiter, discard the targeting reader. Unlike MLLT10, however, no in-cis add-back experiment has been reported, so the verdict here is that the sequence fact is established and the causal claim is not.

Go deeper
What the discarded region does in the intact protein

The AF9 YEATS domain (UniProt P42568, residues 1-138) binds strongly to H3K9ac and, less strongly, to H3K27ac and H3K18ac, using a serine-lined aromatic sandwiching cage rather than a bromodomain-type pocket. ChIP-seq shows strong genome-wide colocalisation of AF9 with H3K9ac, and that colocalisation is important for the chromatin recruitment of the H3K79 methyltransferase DOT1L 1.

Read that last clause carefully, because it is the crux. Intact AF9 does the same job as intact AF10: an N-terminal reader tells a C-terminal DOT1L-docking module where to act. The chemistry differs (acetyl-lysine recognition versus unmodified-H3K27 recognition) but the architecture is the same.

Is it genuinely discarded at the observed breakpoints? Yes, in about 92% of patients

Approximately 92% of MLL-AF9 leukemia patients harbour breakpoints at or after amino acid 376 of AF9, excluding the YEATS domain but retaining the AHD 2. Hu and Muntean put it more bluntly still: patients with a t(9;11) 'almost exclusively exclude the YEATS domain from resultant MLL-AF9 fusion proteins' 2.

This graph's own breakpoint records agree: both bp-kmt2a-intron-9-mllt3 and bp-kmt2a-intron-11-mllt3-therapy list the YEATS domain in domains_lost_3 and the AHD plus the AF9 central IDR in domains_retained_3. Residue 376 sits inside the long disordered region (UniProt 138-475), well downstream of the YEATS fold, so the exclusion is clean rather than a partial truncation.

Does the loss contribute to transformation? Not determinable from current evidence1 deeper

This is where the MLLT10 parallel stops, and saying so plainly is the point of this entry.

What is established: the domain is absent 2, and in the intact protein that domain is what couples DOT1L recruitment to acetylated chromatin 1.

What is not established: that the absence is doing any oncogenic work. Three alternative readings survive the current evidence.

  1. Incidental. MLL supplies its own chromatin targeting (CXXC, AT-hooks, menin/LEDGF tethering). An AF9 reader that steers to H3K9ac may simply be redundant once MLL is driving location, in which case its loss is a by-product of where the DNA happened to break, not a mechanism.
  2. Permissive rather than causal. Retaining YEATS might not suppress transformation but might merely fail to help, in which case selection for post-376 breaks reflects something else entirely — for example, which AF9 introns are physically accessible to the recombination machinery.
  3. Genuinely constraining, as for AF10. The discarded reader normally restricts DOT1L to acetylated loci, and removing it releases DOT1L to wherever MLL goes.

Only reading 3 is a subtraction mechanism, and distinguishing it requires the add-back. Wan et al. supply a relevant and slightly awkward data point: wild-type ENL, but not its paralogue AF9, is required for AML disease maintenance 3 — so AF9's reader function is, in at least one careful CRISPR comparison, the less load-bearing of the two. That cuts against, but does not refute, reading 3.

The experiment that would settle it

exp-yeats-cis-resupply-kmt2a-mllt3: build MLL-AF9 constructs that restore the AF9 YEATS domain in cis, in three flavours — wild-type YEATS, an aromatic-cage point mutant that folds but cannot read acetyl-lysine, and no YEATS (the patient-like control) — then score myeloid progenitor immortalisation, Hoxa9/Meis1 output, H3K79me2 distribution and leukemia latency in transplant.

The design is deliberately the exact analogue of Klein et al.'s CALM-AF10 PZP add-back [see subtraction-mllt10-pzp], including the reader-dead point mutant, which is the control that separates 'restoring a fold' from 'restoring a chromatin-reading constraint'. A published open question from the field asks exactly the neighbouring question: 'Why is the YEATS domain differentially retained in leukemia patients harboring MLL-ENL vs MLL-AF9 fusion proteins?' 2.

Facts and figures
partnergene-mllt3
fusionfusion-kmt2a-mllt3
discarded regionYEATS domain (acyl-lysine reader)
retained oncogenic moduleAHD (ANC1 homology domain), binds AFF1/AFF4, DOT1L, CBX8, BCoR; plus the AF9 central IDR
intact function of discarded regionBinds H3K9ac (weaker H3K27ac/H3K18ac) via a serine-lined aromatic sandwiching cage; colocalises with H3K9ac genome-wide and is important for chromatin recruitment of DOT1L
genuinely discarded at observed breakpointstrue
loss shown to contribute to transformationfalse
verdictnot determinable from current evidence
verdict rationaleThe sequence fact is firm and the intact-protein function is firm, but no in-cis add-back has been reported and a paralogue-level CRISPR comparison argues AF9's reader is the less load-bearing of the two YEATS proteins. Absent the add-back, 'incidental deletion' and 'deleted brake' are both live.
experiment to settleexp-yeats-cis-resupply-kmt2a-mllt3
discarded residues
{
  "start": 1,
  "end": 138,
  "uniprot": "P42568",
  "numbering": "P42568-1 (568 aa)"
}
breakpoint position in partner
{
  "residue_at_or_after": 376,
  "patient_fraction_pct": 92,
  "source": "Hu & Muntean 2023 (PMID 37295550)"
}
How this could bite us back
  • If the YEATS exclusion turns out to be incidental, an entire line of reasoning in this beat collapses for the commonest myeloid partner, and the subtraction class shrinks back towards being an MLLT10 curiosity.
  • YEATS-domain inhibitors (for example SGC-iMLLT) target the reader that MLL-AF9 has already discarded. For MLL-AF9 the only YEATS left in the cell is wild-type AF9 and wild-type ENL, so a YEATS inhibitor in this genotype is not hitting the oncoprotein at all — a mismatch worth stating before anyone proposes it as partner-matched therapy.
Still open
  1. Does restoring the AF9 YEATS domain in cis to MLL-AF9 suppress myeloid progenitor immortalisation and lengthen leukemia latency, as PZP re-supply does for CALM-AF10? An aromatic-cage point mutant is the control that makes the answer interpretable.
  2. Why do AF9 breakpoints cluster after residue 376 while ENL breakpoints cluster before residue 5? Selection for a functional consequence and a purely structural constraint on where the DNA breaks predict different things, and nobody has separated them.
  3. Is there a subset of MLL-AF9 patients (the residual ~8%) whose breakpoints retain YEATS, and do they behave differently clinically? That natural experiment would answer the causal question in humans without any new model system.
  4. Does H3K9ac distribution differ between MLL-AF9 and MLL-ENL leukemias in a way consistent with one fusion reading acetyl-lysine and the other not?
Every source behind this entry (3)
  1. E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    The AF9 YEATS domain binds strongly to H3K9ac and, to a lesser extent, H3K27ac and H3K18ac, using a serine-lined aromatic sandwiching cage; AF9 and H3K9ac strongly colocalise genome-wide, and that colocalisation is important for chromatin recruitment of the H3K79 methyltransferase DOT1L.

    AF9 YEATS domain links histone acetylation to DOT1L-mediated H3K79 methylation · Li Y, Wen H, Xi Y, Tanaka K, Wang H, Peng D, Ren Y, Jin Q, Dent SYR, Li W, Li H, Shi X · Cell 2014
    PMID 25417107DOI
    What this evidence cannot show

    Work on intact AF9, not on the MLL-AF9 fusion. The DOT1L-recruitment claim is from ChIP-seq plus perturbation in cell lines, not from a fusion transformation assay. Verified against PubMed on 2026-09-18.

    Evidence that argues against this (1)
    • The acetyl-reading description of the discarded module is incomplete in a way that changes what the add-back experiment would be testing. Li et al. showed the AF9 YEATS domain binds crotonyllysine with selectively higher affinity than acetyllysine, using an extended aromatic sandwiching cage whose specificity comes from pi-aromatic and hydrophobic interactions conserved among YEATS domains but not bromodomains, and reported that AF9 co-localises with crotonylated histone H3 and regulates gene expression in a YEATS-dependent manner in cells. So what MLL-AF9 discards is an acyl-lysine reader with a broader and differently-weighted ligand set than H3K9ac, the genomic prediction should be written against H3K9cr as well, and any 'binding-dead' cage mutant abolishes both activities at once and so cannot attribute a phenotype to either. Li Y et al. 2016
  2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.review

    Approximately 92% of MLL-AF9 leukemia patients harbour breakpoints at or after amino acid 376 of AF9, excluding the YEATS domain but retaining the AHD, and t(9;11) patients almost exclusively exclude the YEATS domain from the resulting fusion; the review poses as an open question why YEATS is differentially retained in MLL-ENL versus MLL-AF9.

    PMID 37295550DOI
    What this evidence cannot show

    A review, used here as a map of patient breakpoint distribution; the 92% figure is the authors' summary of patient breakpoint data rather than a primary cohort analysis in this paper. It is included because the specific comparative statistic (AF9 vs ENL retention) is not available elsewhere in this form. Full text fetched and the sentences quoted above read verbatim on 2026-09-18; PMID and DOI verified against PubMed the same day.

    Evidence that argues against this (1)
    • The 92% figure is a review's summary of reported breakpoints and has not been reproduced on a systematically typed series. Meyer et al. determined genomic breakpoints in 2,345 acute leukaemia patients, catalogued 135 distinct MLL rearrangements with 94 partners characterised at the molecular level, and showed that the MLL-side breakpoint position shifts with age and lineage; only nine fusions account for more than 90% of rearrangements. Two consequences for this entry: the AF9-side denominator behind '92%' is not stated anywhere it can be audited, and a fusion-level percentage that is stable in aggregate can hide systematic variation by age - which for a beat about infant disease is the variation that matters. Meyer C et al. 2018
  3. E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    CRISPR-Cas9 depletion shows that the YEATS-domain protein ENL, but not its paralogue AF9, is required for disease maintenance in acute myeloid leukaemia.

    ENL links histone acetylation to oncogenic gene expression in acute myeloid leukaemia · Wan L, Wen H, Li Y, Lyu J, Xi Y, Hoshii T, Joseph JK, Wang X, Loh YE, Erb MA, Souza AL, Bradner JE, Shen L, Li W, Li H, Allis CD, Armstrong SA, Shi X · Nature 2017
    PMID 28241141DOI
    What this evidence cannot show

    Concerns wild-type ENL and wild-type AF9 in AML cell lines, not the fusion proteins. Dispensability of wild-type AF9 in those lines does not establish that AF9 YEATS would be dispensable if restored to the fusion in cis; a domain can be redundant in trans and constraining in cis. Verified against PubMed on 2026-09-18.

    Evidence that argues against this (1)
    • If wild-type AF9 is dispensable for AML maintenance while ENL is not, the simplest reading of the MLL-AF9 breakpoint distribution is that AF9 YEATS is a weak functional element whose exclusion carries little information, rather than a deleted brake. That reading predicts a null result from the in-cis add-back, which is exactly why the experiment is worth running. Wan L et al. 2017
Connected to