A three-question triage to detect subtraction mechanisms, applied to every KMT2A partner in the graph
also called subtraction triage · loss-mechanism screen
A short checklist that can be run on every fusion partner already in the database, to catch the kinds of hidden losses that the project's current way of looking would miss.
The same thing, for a scientist
A desk analysis, not an experiment. For every KMT2A partner entry in the graph, answer three questions the existing per-fusion template never asks: what is discarded and is it functional; does the fusion relocate any endogenous protein; is the untranslocated allele of either partner silenced. Outputs a per-partner subtraction score with an explicit 'not determinable' category, and a list of the partners for which the answer requires only a literature search rather than a new experiment.
The three questions
- Discarded-module. What does the fusion protein discard at the observed breakpoints, and does the discarded region have a documented function in the intact protein? Two sub-answers are required and are routinely conflated: is it absent (a sequence fact, cheap) and does its absence do anything (needs an add-back or an equivalent).
- Sequestration. Does the fusion relocate any endogenous protein — most obviously its own wild-type counterpart — and does that relocation remove a function from where it was needed? This question would have caught AFDN and ELL, and the current template does not ask it.
- Allelic. Is the untranslocated allele of KMT2A, or of the partner, transcriptionally silenced in patient blasts?
Each partner is then scored addition / subtraction / both / not determinable, with the sub-type named where subtraction applies.
Result of the first pass (this beat, seven partners)1 deeper
| Partner | Q1 discarded-module | Q2 sequestration | Q3 allelic | Verdict |
|---|---|---|---|---|
| MLLT10 | PZP, absent and causal (in CALM-AF10) | not asked | not asked | subtraction |
| MLLT3 | YEATS, absent, causality untested | not asked | not asked | not determinable |
| MLLT1 | YEATS retained in 84% and required | not asked | not asked | addition |
| ELL | N-terminal EAF1/EAF2 site absent, causality untested | endogenous EAF1 delocalised from speckles | not asked | both |
| AFDN | nothing functional discarded | endogenous AF6 pulled from cytoplasm, RAS released | not asked | both |
| EPS15 | EH domains, retention disputed | not asked | not asked | addition |
| KMT2A-PTD | nothing discarded | not asked | wild-type allele silenced, restoration lethal | both |
The column that pays is Q2. It was asked for two partners and returned a positive for both, and it had never been asked at all before this beat. Q3 was asked for one partner and returned a positive. Neither is visible to the domain-fate analysis the project has been running.
What the table over-claims
Two of the three positives in the table have a published alternative reading, and the table as written does not show it.
Q2 / AFDN. The AF6 RA1 domain is the minimal region SUFFICIENT for MLL-AF6 immortalisation and short-latency leukaemogenesis, Ras-binding-dead substitutions do not abrogate oncogenesis, heterologous RA domains and direct Ras fusions do not substitute, and RA1 works by self-association [E1 counter, Liedtke 2010]. On that account MLL-AF6 transforms by constitutive MLL oligomerisation - an addition - and the nuclear relocation of endogenous AF6 is a by-product. The verdict for AFDN should read 'both, with the sequestration component not shown to be required', which is what the hypothesis entry says and what this table flattens.
Q3 / KMT2A-PTD. Mll(PTD/-) mice die at birth; HoxA up-regulation and increased CFU-GM occur in both Mll(PTD/-) and Mll(PTD/WT) fetal liver, which the authors read as a DOMINANT gain of function; and the increase in pluripotent CFU-GEMM progenitors occurs only when wild-type Mll is present, i.e. the PTD needs the wild-type allele in that compartment [E2 counter, Dorrance 2008]. The allelic sub-type therefore rests on an observation in patient blasts whose causal reading is contested in the mouse.
The methodological point of this beat survives - Q2 and Q3 had never been asked, and asking them surfaced real phenomena. What does not survive is the framing that asking them returned two clean positives. It returned two phenomena with contested causal status, which is a weaker and more accurate claim.
What to run next, ordered by cost
Free, from existing graph content and literature: ask Q2 and Q3 for the partners where they were never asked — MLLT10, MLLT3, MLLT1, EPS15, and the SEPT-family and GAS7 partners not covered in this beat. The cytoplasmic-partner fusions are the highest-prior candidates for Q2 because sequestration requires a normally non-nuclear endogenous counterpart.
Cheap, from existing data: allele-specific expression analysis of KMT2A in translocation-driven KMT2A-r blasts, using the transcriptomic resources already catalogued in this graph (ana-phase2-dataset-target-gdc, ana-phase2-dataset-stjude-cloud). This is the single highest-value unrun analysis identified by this beat, because a positive would extend the allelic sub-type from one lesion to the whole class.
Expensive: the in-cis add-backs (exp-pzp-cis-resupply-kmt2a-mllt10, exp-yeats-cis-resupply-kmt2a-mllt3).
| script note | No script. This is a structured literature and graph-content triage, run by hand against the per-partner entries; it is recorded as an analysis so the next cycle can extend the table rather than rediscover the questions. |
|---|---|
| status | run |
| result summary | Seven partners triaged. Verdicts: subtraction 1 (MLLT10), both 3 (ELL, AFDN, KMT2A-PTD), addition 2 (MLLT1, EPS15), not determinable 1 (MLLT3), plus AFF1 confirmed addition as the control. The sequestration question (Q2) was asked for the first time in this beat, was asked of only two partners, and surfaced a documented phenomenon in both. CAVEAT ADDED 2026-09-19: the causal status of both of the headline positives is contested in the primary literature. For AFDN, the AF6 RA1 domain is sufficient for transformation and acts by MLL self-association, with Ras binding dispensable (Liedtke 2010, PMID 20395419), so sequestration may be a by-product rather than the mechanism. For KMT2A-PTD, Mll(PTD/-) mice die at birth and HoxA up-regulation is dominant rather than recessive, with the PTD requiring wild-type Mll in the pluripotent-progenitor compartment (Dorrance 2008, PMID 18617636). The finding that the questions had never been asked stands; the finding that they returned clean positives does not. |
| highest value next analysis | Allele-specific expression of KMT2A in translocation-driven KMT2A-r blasts from existing transcriptomic cohorts, to test whether the allelic sub-type extends beyond PTD |
inputs
[ "fusion-kmt2a-aff1", "fusion-kmt2a-mllt1", "fusion-kmt2a-mllt3", "fusion-kmt2a-mllt10", "fusion-kmt2a-ell", "fusion-kmt2a-afdn", "protein-eps15-domains" ]
not yet triaged
[ "SEPT family partners", "GAS7", "MLLT11 (AF1q)", "MLLT4 variants", "EPS15 for Q2 and Q3", "MLLT10 for Q2 and Q3" ]
verdict confidence
{
"MLLT10": "discarded-module shown causal in CALM-AF10 only; an addition mechanism (OM-LZ/DOT1L) already suffices",
"AFDN": "sequestration documented; requirement for transformation contradicted by Liedtke 2010",
"KMT2A-PTD": "allelic silencing documented in patient blasts; recessive reading contradicted by Dorrance 2008",
"ELL": "sequestration documented, causality untested",
"MLLT3": "not determinable",
"MLLT1": "addition",
"EPS15": "addition"
}- A triage that produces 'not determinable' for a third of partners is honest but hard to act on. The risk is that 'not determinable' is read as 'no' by whoever uses the table next, which would reinstate exactly the framing error this beat exists to correct.
- The triage's own headline - 'Q2 was asked twice and returned positive twice' - is the kind of small-n pattern the project is supposed to distrust. Two positives out of two questions, both with published alternative explanations, is weaker evidence for a new question class than the table's framing implies.
- Should the sequestration question be asked of every fusion whose 3' partner is normally cytoplasmic, as a standing rule rather than case by case?
- How many of this graph's existing per-fusion dossiers would change verdict if all three questions were applied retroactively? The first pass changed two of seven from 'addition' to 'both'.
- Does the triage need a fourth column for 'shown to be required for transformation' as opposed to 'documented'? Both Q2 and Q3 positives are documented phenomena whose necessity is contested (Liedtke 2010; Dorrance 2008), and the current table has no place to record that distinction.
- E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
Sequestration of an endogenous partner protein by an MLL fusion is a documented phenomenon with at least two independent instances: MLL-AF6 relocating endogenous AF6 from cytoplasm to nucleus with RAS activation, and MLL-ELL delocalising endogenous EAF1 from nuclear speckles.
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 2014What this evidence cannot show
The second instance (EAF1) comes from Simone et al., Blood 2001, PMID 11418481, cited in `subtraction-ell-n-terminal-eaf-site`. Two instances make a pattern worth screening for, not an established class. Verified against PubMed on 2026-09-18.
Evidence that argues against this (1)
- The AFDN/AF6 sequestration positive - the single result that made question Q2 look productive - has a published alternative explanation that is an ADDITION, not a loss. Liedtke et al. identified the AF6 RA1 domain as the MINIMAL REGION SUFFICIENT for MLL-AF6-mediated myeloid progenitor immortalisation in vitro and short-latency leukemogenesis in vivo; single amino-acid substitutions known to abolish the AF6-Ras interaction did NOT abrogate MLL-AF6 oncogenesis; fusing MLL to heterologous RA domains of c-Raf1 or RalGDS, or directly to constitutively active K-RAS, H-RAS or RAP1, was not sufficient; and RA1 was shown to mediate efficient SELF-ASSOCIATION, leading the authors to conclude that constitutive MLL self-association is a more common pathogenic mechanism than previously appreciated. On that account the transforming event is oligomerisation of the MLL moiety, and the relocation of endogenous AF6 to the nucleus is a side effect of expressing a nuclear AF6 fragment rather than the mechanism. Both observations can be true; only one of them is shown to be required for transformation, and it is not the sequestration. Liedtke M et al. 2010
- E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
Transcriptional silencing of the wild-type KMT2A allele occurs in MLL-PTD AML and its pharmacological reversal is associated with selective cell death, establishing the allelic sub-type as a real mechanism at least once.
The MLL partial tandem duplication: evidence for recessive gain-of-function in acute myeloid leukemia identifies a novel patient subgroup for molecular-targeted therapy · Whitman SP, Liu S, Vukosavljevic T, Rush LJ, Yu L, Liu C, Klisovic MI, Maharry K, Guimond M, Strout MP, Becknell B, Dorrance A, Klisovic RB, Plass C, Bloomfield CD, Marcucci G, Caligiuri MA · Blood 2005What this evidence cannot show
Shown for PTD only; whether it generalises to translocation-driven disease is the analysis this entry proposes. Verified against PubMed on 2026-09-18.
Evidence that argues against this (1)
- The 'recessive gain-of-function' reading of MLL-PTD, which is the sole basis of the allelic sub-type, is contradicted in the genetically cleanest test of it. Dorrance et al. generated mice expressing Mll-PTD in the ABSENCE of wild-type Mll and compared them with Mll(PTD/WT). Mll(PTD/-) mice die at birth, unlike Mll(PTD/WT), Mll(WT/-) and Mll(WT/WT) animals, so removing the wild-type allele is not simply permissive. Both Mll(PTD/-) and Mll(PTD/WT) fetal liver cells showed increased HoxA expression and increased CFU-GM, which the authors read as the PTD acting in a DOMINANT gain-of-function fashion; and only Mll(PTD/WT) cells showed increased pluripotent CFU-GEMM progenitors, i.e. in selected tissues the PTD REQUIRES cooperation with wild-type Mll to produce the abnormality. Wild-type-allele silencing in patient blasts is a real observation, but the inference that the silencing is what makes the lesion oncogenic does not follow, and in the compartment that matters most for stemness the wild-type allele appears to be needed rather than dispensable. Dorrance AM et al. 2008
Counter-evidence (Liedtke 2010, PMID 20395419): the AF6 RA1 domain is the minimal region sufficient for MLL-AF6 immortalisation and short-latency leukaemogenesis, Ras-binding-dead substitutions do not abrogate oncogenesis, and RA1 acts by MLL self-association - so the Q2 sequestration positive is documented but not shown to be the transforming mechanism
Counter-evidence (Dorrance 2008, PMID 18617636): Mll(PTD/-) mice die at birth, HoxA up-regulation is dominant with or without the wild-type allele, and increased pluripotent CFU-GEMM requires wild-type Mll - so the recessive gain-of-function reading behind the Q3 positive is contested
the desk analysis that applies the three questions across partners
the sequestration question, asked for the first time, returned positive for AFDN
the allelic question, asked for the first time, returned positive for PTD
the allele-specific expression analysis this triage recommends next would run on these cohorts
the triage table that produced the per-partner verdicts