Oncofusion

start / graph / subtraction-afdn-n-terminus

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

AFDN (AF6): nothing is discarded from the fusion, yet this is the clearest subtraction in the set — the fusion steals wild-type AF6 out of the cytoplasm

also called MLL-AF6 AF6 sequestration · afadin nuclear sequestration RAS · subtraction without deletion

The AF6 fusion throws away almost nothing — the break happens before the first working part of the protein. But the fusion drags the cell's normal AF6 into the nucleus, and because normal AF6 was holding down a growth switch called RAS out in the cytoplasm, that switch gets stuck on.

The same thing, for a scientist

The AFDN junction lies N-terminal to RA1 (UniProt P55196 residues 39-133), so the chimera retains the entire annotated afadin architecture — RA1, RA2, FHA, DIL, PDZ and the proline-rich C-terminus — and discards only an unannotated leader of fewer than 39 residues. At the level of the fusion protein there is no discarded module and therefore no discarded-module subtraction. The subtraction is one compartment up: in MLL-AF6-rearranged cells, endogenous AF6, normally cytoplasmic and controlling RAS-GTP levels, is relocated to the nucleus, and RAS is aberrantly activated; silencing the fusion restores cytoplasmic AF6 and lowers RAS-GTP. Separately, RA1-mediated self-association is the minimal element sufficient for immortalisation and short-latency leukemogenesis — an addition. AFDN is therefore 'both', by two mechanisms that operate in different compartments.

Go deeper
What is N-terminal to the break, and does it do anything?

Almost nothing. UniProt P55196 (afadin, 1,824 aa) places RA1 at 39-133, RA2 at 246-348, FHA at 426-492, DIL at 668-908, PDZ at 1007-1093, and a proline-rich, F-actin-associated C-terminal region running to 1824. This graph records the fusion as retaining RA1 through the proline-rich tail at both the major-BCR and minor-BCR architectures, which places the junction before residue 39. The interval 1-38 carries no annotated domain and no assigned function.

So the honest answer to the question this beat was sent to ask is: for AFDN, the discarded region is a short unannotated leader, and there is no reason to think its loss does anything. If the audit had only looked at discarded domains, AFDN would have been filed as a clean addition partner and the interesting biology would have been missed entirely.

The addition: RA1 self-association

RA1 is the minimal region sufficient for MLL-AF6-mediated myeloid progenitor immortalisation and short-latency leukemogenesis in vivo 1. The important nuance is which property of RA1 does the work: point mutations that abolish AF6-Ras binding did not abrogate MLL-AF6 oncogenesis, and neither heterologous RA domains (c-Raf1, RalGDS) nor direct MLL fusion to activated RAS or RAP1 were sufficient 1. What matters is RA1's capacity for self-association — this is a constitutive-dimerisation fusion, in the same class as MLL-GAS7 and MLL-EPS15, not a Ras-signalling fusion.

Yokoyama et al. add that MLL-AF6, a fusion with a cytoplasmic protein, does not form MLL/AEP hybrid complexes as the AF4- and ENL-family fusions do, yet still constitutively recruits AEP to target chromatin by an alternative route 3. AFDN reaches the same destination as the elongation-family partners by a different mechanism.

The subtraction: nuclear sequestration of endogenous AF6 releases RAS1 deeper

'AF6 is expressed in the cytoplasm of healthy bone marrow cells and controls RAS-GTP levels. By contrast, in MLL-AF6-rearranged cells, AF6 is found localized in the nucleus, leading to aberrant activation of RAS... Silencing MLL-AF6, we restored AF6 localization in the cytoplasm, thus mediating significant reduction of RAS-GTP levels.' 2

This is a genuine loss-of-function contribution, and it satisfies the standard this beat demands: not merely that something is absent, but that the absence is doing work. The silencing-and-rescue experiment is the causal link — remove the fusion, AF6 returns to the cytoplasm, RAS-GTP falls.

But note carefully what is subtracted and from where. No domain is missing from the fusion protein. What is missing is wild-type AF6 from the cytoplasm, because the fusion (nuclear by virtue of MLL's AT-hooks 4) drags it in, presumably through the same RA1 self-association that drives transformation. The two mechanisms are plausibly the same molecular event read in two compartments, which nobody has tested.

Why this matters beyond AFDN

The retained-half framing this beat was sent to audit fails here for a reason the King's entry did not anticipate. It is not that the fusion discarded a brake; it is that the fusion relocated one. A scan of discarded modules would have returned 'nothing lost' and moved on.

That means the subtraction class needs at least three sub-types, and only the first is visible to a domain-fate analysis:

  1. Discarded-module subtraction — MLLT10 PZP.
  2. Sequestration subtraction — AFDN (endogenous AF6 pulled out of the cytoplasm), ELL (endogenous EAF1 pulled out of speckles).
  3. Allelic subtraction — KMT2A-PTD (the wild-type allele silenced); see subtraction-kmt2a-ptd-wildtype-silencing.

Sequestration is the sub-type most likely to have been systematically missed, because it leaves no trace in a sequence.

Facts and figures
partnergene-afdn
fusionfusion-kmt2a-afdn
discarded regionAFDN residues 1-38, an unannotated N-terminal leader with no assigned function
intact function of discarded regionnone known
genuinely discarded at observed breakpointstrue
discarded region functionally significantfalse
loss of discarded module shown to contribute to transformationfalse
alternative subtraction mechanismsequestration — the nuclear fusion relocates endogenous cytoplasmic AF6, removing its control of RAS-GTP; reversible by silencing the fusion
sequestration shown to cause a loss of functiontrue
sequestration shown to be required for transformationfalse
verdictboth
verdict rationaleAddition (RA1 self-association, minimal and sufficient) is established. Subtraction is real but operates at the cellular rather than the domain level (depletion of cytoplasmic AF6 and consequent RAS activation, established as fusion-dependent). Neither mechanism involves a discarded module, so a discarded-module audit alone would have scored AFDN as pure addition and been wrong.
experiment to settleExpress an MLL-AF6 variant carrying RA1 self-association-dead point mutations that still localises to the nucleus, and ask whether endogenous AF6 is still sequestered and RAS still activated. If sequestration survives loss of self-association, the two mechanisms are separable and can be scored independently; if it does not, transformation and RAS activation are one event and the subtraction is downstream of the addition.
discarded residues
{
  "start": 1,
  "end": 38,
  "uniprot": "P55196",
  "numbering": "P55196-1 (1824 aa); boundary inferred from the fusion retaining RA1 at 39-133"
}
retained modules
[
  "RA1 39-133",
  "RA2 246-348",
  "FHA 426-492",
  "DIL 668-908",
  "PDZ 1007-1093",
  "proline-rich F-actin-associated C-terminus to 1824"
]
How this could bite us back
  • If the fusion activates RAS by sequestration, a menin inhibitor that displaces the fusion from chromatin does nothing about the cytoplasmic depletion, so this partner could show transcriptional response without proliferative response.
  • Conversely, any therapy that lowers fusion protein levels would release sequestered AF6 and lower RAS-GTP — a potential biomarker of target engagement for this partner that costs nothing to measure and has never been used.
Still open
  1. Are RA1 self-association and endogenous-AF6 sequestration the same molecular event? A self-association-dead, nuclear-localised MLL-AF6 would separate them in a single experiment.
  2. KMT2A::AFDN is the worst-prognosis KMT2A-r AML subgroup in this graph and RAS-pathway comutations are the commonest cooperating lesions in KMT2A-r disease. If the fusion already activates RAS by sequestration, are RAS comutations redundant in this subgroup, or additive? The graph holds comutation entries (`comut-ras-pathway-kmt2a-r`) that could be re-cut by partner to check.
  3. Does AF6 sequestration occur in other cytoplasmic-partner MLL fusions (GAS7, EPS15, SEPT family), which would make sequestration a property of the cytoplasmic-partner class rather than of AF6?
  4. Afadin is a tumour suppressor at adherens junctions in epithelia. Is there a non-haematopoietic phenotype in KMT2A::AFDN patients that would corroborate functional depletion of cytoplasmic AF6?
Every source behind this entry (4)
  1. E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    The RA1 domain of AF6 is the minimal region sufficient for MLL-AF6-mediated myeloid progenitor immortalisation and short-latency leukemogenesis in vivo, acting through self-association: point mutations abolishing AF6-Ras binding did not abrogate oncogenesis, and heterologous RA domains or direct MLL fusion to activated RAS/RAP1 were not sufficient.

    Self-association mediated by the Ras association 1 domain of AF6 activates the oncogenic potential of MLL-AF6 · Liedtke M, Ayton PM, Somervaille TCP, Smith KS, Cleary ML · Blood 2010
    PMID 20395419DOI
    What this evidence cannot show

    Murine myeloid progenitor immortalisation and transplant models. Identifier and abstract content verified via PubMed on 2026-09-18 by a verification pass dedicated to this question.

    Evidence that argues against this (1)
    • 'Minimal region sufficient' in a retroviral immortalisation assay does not identify the rate-limiting mechanism, and MLL-AF6 turns out to run through the same H3K79 axis as the partners this beat treats as mechanistically separate. Deshpande et al. showed that H3K79me2 is converted to me1 at HOX loci as haematopoietic cells mature, that the DOT1L cofactor AF10 controls both the methyltransferase activity and the me1-to-me2 conversion, and that AF10 inactivation reverses leukaemia-associated epigenetic profiles, precludes abnormal HOXA expression and impairs the transforming ability of MLL-AF6 as well as MLL-AF9 and NUP98-NSD1. If MLL-AF6 needs endogenous, PZP-containing AF10 supplied in trans to transform, then RA1 self-association is an enabling step upstream of a shared mechanism rather than the mechanism, and this entry's 'nothing is discarded from the fusion, so the subtraction must be one compartment up' reasoning has skipped a possibility: the fusion depends on a module it never carried. Deshpande AJ et al. 2014
  2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    AF6 is cytoplasmic in healthy bone marrow cells and controls RAS-GTP levels; in MLL-AF6-rearranged cells AF6 is localised to the nucleus and RAS is aberrantly activated, and silencing MLL-AF6 restores cytoplasmic AF6 localisation and significantly reduces RAS-GTP levels.

    MLL-AF6 fusion oncogene sequesters AF6 into the nucleus to trigger RAS activation in myeloid leukemia · Manara E, Baron E, Tregnago C, Aveic S, Bisio V, Bresolin S, Masetti R, Locatelli F, Basso G, Pigazzi M · Blood 2014
    PMID 24695851DOI
    What this evidence cannot show

    Patient samples and cell lines; the silencing-rescue establishes dependence of AF6 localisation and RAS-GTP on the fusion, but not that RAS activation is required for transformation. Note the tension with E1, where Ras binding by RA1 was shown to be dispensable for oncogenesis — the two results are not incompatible (E1 concerns the fusion's own Ras binding, E2 concerns endogenous AF6's cytoplasmic Ras control) but they are frequently conflated. Identifier and abstract content verified via PubMed on 2026-09-18.

    Evidence that argues against this (1)
    • Liedtke et al. showed that abolishing AF6-Ras binding does not abrogate MLL-AF6 oncogenesis and that fusing MLL directly to activated RAS is not sufficient to transform. If RAS activation were the load-bearing consequence of AF6 sequestration, MLL-RAS should have worked. The most parsimonious synthesis is that RA1 self-association drives transformation and RAS activation is a co-occurring, possibly cooperating, consequence rather than the mechanism. Liedtke M et al. 2010
  3. E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    MLL-AF6, an MLL fusion with a cytoplasmic protein, does not form MLL/AEP hybrid complexes as AF4- and ENL-family fusions do, but nevertheless constitutively recruits AEP (AF4-family/ENL-family/P-TEFb) to target chromatin by an alternative mechanism.

    PMID 20153263DOI
    What this evidence cannot show

    The alternative recruitment mechanism is explicitly described by the authors as unknown. Verified against PubMed on 2026-09-18.

    Evidence that argues against this (1)
    • The 'unknown alternative mechanism' has a published, non-mysterious candidate that this entry does not consider: trans-recruitment through the wild-type MLL1 complex. Milne et al. mapped an MLL1 recruitment domain spanning the CXXC domain and the PHD fingers, controlled by direct interaction with the PAF elongation complex and by H3K4me2/3, and showed that MLL1 FUSION proteins - which lack the PHD fingers - require prebinding of a wild-type MLL1 complex plus CXXC recognition of DNA for stable HoxA9 association. On that account an MLL fusion with a cytoplasmic partner needs no partner-borne AEP contact at all, because the retained MLL portion docks where a wild-type MLL1 complex already sits. It is a competing explanation for the same observation and it is testable: it predicts MLL-AF6's AEP recruitment should be sensitive to loss of wild-type MLL1, which was not tested. Milne TA et al. 2010
  4. E4moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    Chimeric MLL-AF6 products localise to the nucleus, and the nuclear localisation is driven by MLL's AT-hook motifs rather than by AF6.

    Chimeric MLL products with a Ras binding cytoplasmic protein AF6 involved in t(6;11)(q27;q23) leukemia localize in the nucleus · Joh T, Yamamoto K, Kagami Y, Kakuda H, Sato T, Yamamoto T, Takahashi T, Ueda R, Kaibuchi K, Seto M · Oncogene 1997
    PMID 9349501DOI
    What this evidence cannot show

    A 1997 transfection/localisation study predating the sequestration model; included because it establishes that the nuclear destination is imposed by the MLL half, which is what makes sequestration of the endogenous partner possible. Identifier verified via PubMed on 2026-09-18.

    Evidence that argues against this (1)
    • A 1997 transfection result that the nuclear address is imposed by the MLL AT-hooks was never confirmed on endogenous protein, and the one study that did look at endogenous fusion protein found no relocation at all. Li, Frestedt and Kersey raised antibodies against AF4, detected the predicted 240-kDa MLL-AF4 fusion in every t(4;11) cell line, and reported that the fusion protein had a SIMILAR subcellular distribution to wild-type AF4, with both t(4;11) and non-t(4;11) leukaemic cells showing the same punctate nuclear staining. That is a different partner, but it is the only endogenous-protein comparison of this kind in the KMT2A literature and it returned a negative. Attributing a destination to the MLL half on the strength of overexpressed chimeras in COS-type systems is therefore not safe, and the entry should not lean on E4 as the mechanism that makes sequestration possible. Li Q et al. 1998
Connected to