Hypothesis: where a KMT2A fusion sits in the nucleus is set entirely by retained KMT2A modules, and no partner changes the compartment - so every published 'the fusion is here' result is a wild-type KMT2A result
also called compartment is retained-KMT2A · partner does not choose the compartment
Every experiment that has looked at where an MLL fusion protein sits inside the nucleus found it in the same place as the normal protein, whichever partner was attached. This proposes that is not a coincidence or a limitation of the pictures: the address really is written entirely in the MLL half, and the partner changes what the protein does once it arrives, not where it arrives.
The same thing, for a scientist
Four independent lines converge. (1) Nuclear localisation of MLL-AF6 maps to a region of the MLL N-terminus containing the AT-hook motifs, and the chimera is nuclear although AF6 is cytoplasmic [E1]. (2) Punctate nuclear distribution of chimeric ALL-1 proteins maps to two short trithorax-conserved sequences within the N-terminal ~1,100 residues [E2]. (3) A conditionally expressed 50 kDa N-terminal fragment reproduces the full localisation repertoire - punctate nucleoplasm, envelope-attached regions, nucleolar periphery, nuclear matrix, mitotic scaffold - and MLL-AF9 and MLL-AF4 show the same nucleoplasmic and nucleolar distribution, which the authors call a common target conferred by the MLL amino terminus [E3]. (4) MLL-SEPT6 is nuclear while SEPT6 alone, overexpressed in the same cells, is not [E4]. **The hypothesis is that this is complete**: no KMT2A fusion partner alters the chimera's subnuclear compartment, and the entire partner contribution is to activity at the address rather than to the address. If true, the correct reading of the imaging literature is not that the pictures failed to detect a difference but that **there is no difference to detect**, and a whole class of proposed experiments - image the fusion, compare compartments across partners - has a predictable null result. The hypothesis is unusual in this graph in that it predicts the *absence* of an effect, which makes it cheap to falsify and impossible to confirm.
What would falsify it, in order of cost
- A fractionation showing the fusion and wild-type KMT2A in different fractions of the same lysate. Cheapest, no new reagent, and the experiment this beat proposes. A difference in chromatin-bound fraction between fusion and parent is a compartment difference the imaging literature could never have seen.
- Any partner-specific compartment. Two candidates exist and neither has been tested: MLL-GPHN, whose partner is a submembranous scaffolding protein and for which no localisation experiment of any kind has been published; and a septin fusion other than SEPT6.
- Endogenous partner sequestration changing the fusion's own distribution. MLL-AF6 pulls endogenous AF6 into the nucleus; if that creates a fusion-plus-partner co-compartment, the address would be partner-dependent after all.
- A tagged endogenous knock-in showing the fusion in a compartment wild-type KMT2A does not occupy. Most decisive, most expensive, and constrained by the finding that an N-terminal tag on MLL::AF9 disturbs the very targeting being measured.
What it would mean if true
For this graph. The nuclear speckle, Pol II cluster, PML body, Polycomb body and paraspeckle entries would all be asking a question with a known answer, and their value would shift from 'is the fusion here?' to 'is KMT2A here, and does the fusion bring different machinery to the same place?'. That is a different and more tractable question.
For therapeutics. A mislocalisation strategy would be dead, because the address module is shared with the abundant wild-type protein and with KMT2B. Selectivity would have to come from the partner's activity, which is where the field already is.
For the menin-interface hypothesis. Neutral to mildly supportive. Menin and LEDGF bind the retained N-terminus, so if the address is entirely N-terminal then menin is part of the address, and a drug that breaks the menin contact is attacking the address directly rather than a downstream consequence. The hypothesis does not threaten the menin interface; it locates it.
The honest weakness
Every observation supporting this hypothesis was made with a readout that could not have detected the alternative. Immunofluorescence with a shared epitope cannot resolve fusion from wild-type, so 'the fusion is where wild-type is' is the expected observation under both models. The hypothesis is therefore currently supported only by evidence that cannot discriminate it, which is a weak position that reads as a strong one.
The constructs are the partial exception: in a transfected cell expressing only the chimera, the signal is attributable. Those experiments do support the hypothesis. But they are overexpression systems, and overexpression is the condition under which spurious compartments most readily appear.
So: plausible, convergent, cheap to falsify, and resting on a literature whose principal defect is the one the hypothesis is about.
| statement | No KMT2A fusion partner alters the chimera's subnuclear compartment; the address is set entirely by retained KMT2A N-terminal modules and the partner contributes activity at that address. |
|---|---|
| origin | literature-gap |
| prior probability | 0.55 |
predictions
[ "A five-fraction blot will show the fusion and wild-type KMT2A in the same fractions in the same proportions.", "MLL-GPHN, when finally imaged, will be nuclear and punctate like every other KMT2A chimera.", "No partner-specific compartment will be found for any of the 100-plus KMT2A partners." ]
experiments
[ "exp-five-fraction-blot-attributable-localisation-kmt2a-aff1", "exp-fusion-fate-imaging-menin-timecourse" ]
analyses
[ "ana-attribution-audit-subnuclear-localisation-kmt2a-aff1" ]
falsifiers
[ "A fractionation in which the fusion and wild-type KMT2A occupy different fractions of the same lysate.", "Any partner-specific subnuclear compartment shown in an attributable system.", "An endogenously tagged fusion occupying a compartment wild-type KMT2A does not.", "Demonstration that the HRX/ALL1-eps15 'smaller and more numerous bodies' result is partner-driven rather than expression-level-driven." ]
- If the compartment is a retained-KMT2A property, no mislocalisation therapeutic can be selective for the fusion, because the targeting module is shared with wild-type KMT2A and with KMT2B.
- Accepting this hypothesis would license skipping localisation experiments for new partners, and the one partner class that might break it - submembranous scaffolds such as gephyrin - is exactly the one that has never been tested.
- Is there any KMT2A partner class - submembranous scaffolds, cytoskeletal proteins - for which the prediction has never been tested even once?
- Does the reciprocal product, which carries the partner N-terminus and the KMT2A C-terminus, obey the opposite rule?
- Can the HRX/ALL1-eps15 result be reproduced with expression matched to endogenous level?
- E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
MLL-AF6 chimeric products localise in the nuclei despite AF6 being cytoplasmic, by immunofluorescence and cell fractionation, and the responsible region of the MLL N-terminus was found to be a region containing AT-hook motifs.
Chimeric MLL products with a Ras binding cytoplasmic protein AF6 involved in t(6;11) (q27;q23) leukemia localize in the nucleus · Joh T, et al. · Oncogene 1997What this evidence cannot show
Abstract only; no PMCID. A region assignment, and nuclear import is not distinguished from nuclear retention.
Evidence that argues against this (1)
- MLL-AF6-rearranged cells relocate endogenous AF6 into the nucleus, so the fusion does change a compartment - the partner's, not its own - which is a partner-dependent localisation effect the hypothesis must accommodate. Manara E et al. 2014
- E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
ALL-1/MLL distributes in a nuclear punctate pattern, several chimeric ALL-1 proteins expressed in transfected cells showed similar speckles, and the distribution was assigned to two main domains within the N-terminal ~1,100 residues spanning two short trithorax-conserved sequences.
Nuclear punctate distribution of ALL-1 is conferred by distinct elements at the N terminus of the protein · Yano T, et al. · Proceedings of the National Academy of Sciences of the United States of America 1997What this evidence cannot show
Transfected overexpression with deletion mapping in constructs; 'several chimeric proteins' is not an exhaustive partner survey and the partners tested are not enumerated in the abstract.
Evidence that argues against this (1)
- In the same era, HRX/ALL1-eps15 from t(1;11) localised exclusively to the nucleus in bodies smaller and more numerous than wild-type, which is a partner-associated change in the pattern and is the clearest published counter-example. Rogaia D et al. 1997
- E3moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper
A conditionally expressed 50 kDa N-terminal MLL fragment reproduced the endogenous protein's punctate nucleoplasmic distribution including envelope-attached regions and nucleolar periphery, nuclear-matrix presence and mitotic-scaffold co-localisation with topoisomerase II, and nucleoplasmic and nucleolar localisation was observed even for conditionally expressed MLL-AF9 and MLL-AF4, indicating a common target conferred by the MLL amino terminus.
The amino terminus targets the mixed lineage leukemia (MLL) protein to the nucleolus, nuclear matrix and mitotic chromosomal scaffolds · Caslini C, et al. · Leukemia 2000What this evidence cannot show
Conditional overexpression of a fragment and of chimeras in U937 and HeLa, neither carrying the translocation. Abstract-level; no PMCID. **Retrieval correction by the coordinator, 2026-09-20.** This beat logged PMID 11069025 as having no PMCID and being readable at abstract level only. It has **PMCID PMC7543881**. `fullTextXML` returns HTTP 500, but the **PMC HTML page returns 63,706 characters of stripped text**. The paper is readable and was read; the full text supplies the image-level detail *"the presence of label at the edge of the nuclear envelope"* and confirms that the word *lamin* appears nowhere in it.
Evidence that argues against this (1)
- Two fusions sharing a compartment in an overexpression system does not establish that no partner could differ; the partners tested here (AF9, AF4) are both nuclear elongation factors, which is the class least likely to differ. Ono R et al. 2005
- E4strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
MLL-SEPT6 distributed as small nuclear speckles or dots in transfected 293T while overexpressed SEPT6 alone did not shift from its physiological cytoplasmic localisation.
Dimerization of MLL fusion proteins and FLT3 activation synergize to induce multiple-lineage leukemogenesis · Ono R, et al. · The Journal of clinical investigation 2005What this evidence cannot show
Transient transfection in a non-haematopoietic line. The partner-alone control in the same figure is the design strength.
Evidence that argues against this (1)
- SEPT6 is one septin of several fused to KMT2A, and no septin fusion other than SEPT6 has been imaged, so the generalisation across the septin class is untested. Eguchi M et al. 2004
- E5moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper
The counter-example the hypothesis has to survive: wild-type HRX/ALL1 localised to both cytoplasm and nucleus with diffuse staining, perinuclear accumulation and nuclear bodies of variable size, whereas the HRX/ALL1-eps15 fusion from t(1;11) localised exclusively to the nucleus in bodies that were smaller and more numerous.
The localization of the HRX/ALL1 protein to specific nuclear subdomains is altered by fusion with its eps15 translocation partner · Rogaia D, et al. · Cancer research 1997What this evidence cannot show
Transfected overexpression; 'smaller and more numerous' is a morphological description of a punctate pattern, not a compartment assignment, and could reflect expression level rather than partner identity. No PMCID and no DOI in the PubMed record.
Evidence that argues against this (1)
- The same paper reports wild-type HRX/ALL1 in both cytoplasm and nucleus, which conflicts with every other report of an exclusively nuclear wild-type protein and suggests the system was reporting construct behaviour rather than physiology. Caslini C et al. 2000
the cheapest falsifier: fusion and wild-type in different fractions of the same lysate
rests on the 1997 mapping of the localisation determinant to retained KMT2A sequence
MLL-GPHN is the untested case that could falsify it
the 2000 conditional-expression study is the strongest single support, covering two fusions and a fragment
the hypothesis is currently supported only by evidence the audit found cannot discriminate it
if the address is entirely N-terminal, menin is part of the address rather than downstream of it
the one measured module-to-compartment link concerns a module the fusion has lost, which would make the fusion's compartment differ from wild-type