Hypothesis: tubulin/microtubule association is a recurrent, unnamed property of the KMT2A partners whose transforming module has no other assignment
also called microtubule-templating hypothesis · tubulin module hypothesis
Two KMT2A partners that supposedly have nothing in common turn out to share one thing: they both stick to the cell's microtubule tracks. The idea here is that this is not a coincidence — that binding to a long polymer, rather than simply pairing up, may be how some of these partners force KMT2A into the clusters that cause leukaemia.
The same thing, for a scientist
Speculative, and offered as a pattern worth testing rather than a claim. The minimal transforming module of gephyrin in MLL-GPHN is described by its discoverers as a **15-amino-acid tubulin-binding domain**, which is necessary and sufficient and which also confers oligomerisation capacity on MLL [E1]. Independently, proteomic analysis of MLLT11/Mllt11 — the 90-residue partner with no annotated domain of any kind — found it **associates with stabilised microtubules**, with Mllt11 loss affecting microtubule staining in callosal axons [E2]. A third partner class, the septins, is the fourth cytoskeletal system and the KMT2A-septin literature is explicitly framed around cytoskeletal function [E3]. The hypothesis: for the subset of KMT2A partners with no elongation module and no canonical oligomerisation module, **tubulin/microtubule association is the shared biochemical property, and the oligomerising effect on MLL is templating onto a polymer rather than dimerisation**. Prior probability is set low (0.15) and the strongest existing counter-argument is already in the graph: synthetic dimerisation modules — which bind no cytoskeletal polymer at all — fully substitute for the EPS15 and GAS7 coiled coils and for MLL oligomerisation generally [E4][E5]. If proximity alone suffices, tubulin binding is at most a route to proximity and not a mechanism in its own right.
The pattern, stated conservatively
Three observations, none of which was made with the others in mind.
- GPHN. Eguchi et al. reduced the transforming activity of MLL-GPHN to a 15-amino-acid tubulin-binding domain, necessary and sufficient in vitro and in vivo, and reported that the same domain confers oligomerisation capacity on MLL 1. The module is named for tubulin binding, not for oligomerisation; the oligomerisation was the inferred consequence.
- MLLT11. The partner with the least to offer any model — 90 residues, one annotated motif, no domain, no structure — turns out on proteomics to associate with stabilised microtubules in neurons, and its loss changes microtubule staining in axons 2. Nobody has asked whether AF1q binds microtubules in a haematopoietic cell, and the question has never been connected to the fusion.
- The septins. The KMT2A-septin partners are the fourth cytoskeletal system, and the review literature frames their leukaemic role around cytoskeletal function 3. Septin-microtubule association is well described in cell biology generally, though this entry does not cite a primary demonstration of it and therefore does not use the septins as evidence, only as context.
What would make this a mechanism rather than a coincidence is the difference between dimerising and templating. A coiled coil makes two copies. A polymer-binding module can string many copies along a filament. Those are different objects with different consequences for how much MLL is concentrated where, and the field has assumed the first without measuring it for any of these fusions.
The strongest argument against, taken seriously
The synthetic-dimeriser result is the problem, and it is decisive if taken at face value. So and colleagues showed that homodimerisation of MLL by synthetic dimerisation modules mimics bona fide MLL fusion proteins, producing Hox gene activation and enhanced progenitor self-renewal 4, and Martin and colleagues independently showed that induced dimerisation of a truncated MLL immortalises bone marrow and upregulates Hoxa7, Hoxa9 and Meis1 5. A synthetic dimeriser binds no microtubule. If two copies of MLL brought together by an inert chemical module are sufficient, then the specific chemistry of the partner's module — tubulin binding included — is incidental.
Three responses, offered as what would have to be true rather than as rebuttals:
- Sufficiency is not necessity. A synthetic dimeriser being sufficient does not establish that the real fusions work the same way, and Martin et al. showed in the same paper that MLL-AF9 is not dimerised in vivo 5 — so at least one common fusion does not use the mechanism its own field's founding experiment demonstrates.
- Different modules may give different diseases. Nothing in the graph compares the latency, penetrance or lineage of a synthetic-dimeriser leukaemia against a genuine MLL-GPHN or MLL-GAS7 leukaemia side by side. If templating produced a distinguishable disease, the sufficiency result would be compatible with a real mechanistic difference.
- GAS7 pushes the same way. Its self-assembly determinant turned out to be an F-BAR surface forming two-dimensional sheets, not a dimerising coiled coil 6 — which is already an instance of a KMT2A partner whose oligomerisation is lattice-like rather than pairwise, arrived at from a completely different direction.
The honest position is that the prior is low and the hypothesis is cheap to test.
What this would change if true
Little for the common fusions and a great deal for this beat's partners.
If the transforming module of a subset of KMT2A fusions is a polymer-binding surface, then the oligomeric species is not a dimer and an inhibitor designed against a dimer interface has no target. It would also predict that these leukaemias are sensitive to microtubule-state perturbation in a way the elongation-partner fusions are not — a testable and, given the existing clinical pharmacopoeia of tubulin-targeting agents, unusually actionable prediction.
It would not touch the menin-interface hypothesis either way. Menin binds the retained KMT2A N-terminus, which every fusion in this beat keeps; how the C-terminal partner oligomerises is orthogonal to whether the N-terminus needs menin. That independence is worth stating explicitly, because a reader could easily take a partner-side mechanism as competing with a KMT2A-side one when the two are compatible.
| statement | For the subset of KMT2A fusion partners that supply no elongation module and no canonical oligomerisation module, tubulin/microtubule association is a shared and under-recognised biochemical property, and the oligomerising effect on MLL is templating onto a polymer rather than pairwise dimerisation. |
|---|---|
| origin | data-pattern |
| prior probability | 0.15 |
| status of the pattern | Two independent observations (GPHN, MLLT11) plus context from a third class (septins). No source has ever connected them, and no experiment in the list above has been performed. |
predictions
[ "Purified recombinant AF1q binds microtubules in a co-sedimentation assay.", "The gephyrin 15-residue module's tubulin binding and its oligomerisation activity can be separated by point mutation, and transformation tracks one of them and not the other.", "MLL-GPHN and, if it transforms, KMT2A::MLLT11 show a microtubule-associated cytoplasmic pool in addition to any nuclear pool.", "Leukaemias driven by these partners are more sensitive to microtubule-stabilising or -destabilising agents than leukaemias driven by elongation partners, at matched cytotoxic exposure.", "The oligomeric species formed by MLL-GPHN is larger than a dimer by size-exclusion chromatography or mass photometry." ]
experiments
[ "exp-af1q-oligomeric-state-and-fusion-transformation" ]
analyses
[ "ana-kmt2a-deferred-partner-enumeration" ]
falsifiers
[ "Purified AF1q does not co-sediment with microtubules. This removes the MLLT11 leg and leaves the pattern resting on GPHN alone, which is one paper.", "A gephyrin 15-mer point mutant that abolishes tubulin binding while retaining self-association still transforms. This would show the tubulin binding is incidental and the oligomerisation is the operative property.", "MLL-GPHN is shown by size-exclusion chromatography or mass photometry to be a simple dimer. Templating predicts a larger and more heterogeneous species.", "A synthetic dimeriser produces a leukaemia indistinguishable from MLL-GPHN in latency, penetrance, lineage and transcriptional programme in a head-to-head comparison. That would make pure proximity a complete account and leave nothing for a polymer to explain.", "KMT2A::MLLT11 is shown not to transform at all, which would remove the partner that makes this pattern interesting in the first place." ]
- If this were true it would suggest microtubule-targeting agents as a partner-selective strategy — and that inference should be resisted hard. Vinca alkaloids and taxanes are already used in paediatric leukaemia protocols, so any apparent partner-specific benefit would be extremely difficult to attribute, and a false positive here would misdirect a trial.
- The hypothesis predicts a cytoplasmic pool for fusion proteins that everyone assumes are nuclear. If that pool exists and is functional, nuclear-targeted interventions including menin inhibitors would be acting on only part of the oncoprotein population — a resistance mechanism nobody is looking for.
- A templating mechanism would mean the oligomerisation interface is a polymer-binding surface, which is a far harder target than a coiled-coil interface and is shared with essential host proteins.
- Does purified AF1q bind microtubules? One co-sedimentation assay on protein that already exists in refolded recombinant form would decide whether this hypothesis has two legs or one.
- Can the gephyrin 15-residue module's tubulin binding be separated from its self-association by mutation? This is the single experiment that would convert the hypothesis from a pattern into a mechanism or kill it outright.
- What is the oligomeric state of any KMT2A fusion protein? Remarkably, no KMT2A fusion has had its stoichiometry measured — 'dimerisation' is a description inherited from the modules, not from the proteins.
- If the shared property is 'binds a cytoskeletal polymer' rather than specifically tubulin, is the hypothesis still non-trivial? GAS7 binds actin and membranes, the septins are their own polymer system, and at that breadth the claim risks becoming unfalsifiable.
- E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
A small 15-amino-acid TUBULIN-BINDING domain of GPHN is necessary and sufficient for MLL-GPHN to transform haematopoietic progenitors in vitro and in vivo despite gephyrin lacking transcriptional activity, and this domain also confers oligomerisation capacity on MLL protein.
The small oligomerization domain of gephyrin converts MLL to an oncogene. · Eguchi M, Eguchi-Ishimae M, Greaves M · Blood 2004What this evidence cannot show
READ FROM THE ABSTRACT ONLY; the full text is subscription-only with no PMC record and both publisher mirrors returned HTTP 403. Whether the domain's tubulin binding was separated experimentally from its oligomerisation is therefore UNKNOWN to this entry, and that separation is the crux of the hypothesis. The domain's residue numbers are not asserted anywhere. One laboratory, no independent replication found in 22 years.
Evidence that argues against this (1)
- The paper's own title calls it 'the small OLIGOMERIZATION domain', and the abstract presents oligomerisation as the mechanistic conclusion while 'tubulin-binding' is used descriptively to identify the region. Reading the naming as evidence that tubulin binding is the operative property inverts the authors' own emphasis, and nothing in the abstract shows the two activities were separated. So CW et al. 2003
- E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
Proteomic analysis revealed that Mllt11 (AF1q) associates with stabilised microtubules, and Mllt11 loss affected microtubule staining in callosal axons; Mllt11 is detected in varicosities, growth cones and soma of cultured primary cortical neurons and is required for neuritogenesis and migration of upper-layer cortical projection neurons.
Mllt11 Regulates Migration and Neurite Outgrowth of Cortical Projection Neurons during Development. · Stanton-Turcotte D, Hsu K, Moore SA, Yamada M, Fawcett JP, Iulianella A · J Neurosci 2022What this evidence cannot show
Mouse developing neocortex, not haematopoietic cells and not the fusion. The microtubule link is a proteomic co-association plus an immunostaining phenotype, not a demonstrated direct binding; no binding site, stoichiometry or affinity is reported. Whether human AF1q binds microtubules in a haematopoietic cell is entirely untested.
Evidence that argues against this (1)
- The same paper reports Mllt11 expression as exclusive to developing neurons, which the haematopoietic literature contradicts — AF1q mRNA is high in thymus and present in every leukaemic cell line tested. A microtubule association demonstrated in a cell type whose expression claim does not transfer is weak ground for a claim about leukaemic blasts. Tse W et al. 1995
- E3moderate confidence 3 of 5: One strong study, or several smaller consistent ones.review
Five septin genes were known as MLL fusion partners by 2011, producing chimeras in which the MLL N terminus is fused in frame to almost the entire septin open reading frame, and the review literature frames their possible leukaemic mechanism around septin roles in cytokinesis, membrane remodelling and compartmentalisation.
MLL-SEPTIN gene fusions in hematological malignancies. · Cerveira N, Bizarro S, Teixeira MR · Biol Chem 2011What this evidence cannot show
A review, cited here only as CONTEXT for the cytoskeletal framing of this partner class — not as evidence that septins bind microtubules, which this entry does not claim and for which no primary source was fetched. The review's mechanistic speculation was speculation in 2011 and remains untested.
Evidence that argues against this (1)
- The only functional test performed on any KMT2A-septin fusion argues directly against the partner's cytoskeletal function mattering: Sept6-null mice develop normally and the disease induced by MLL-SEPT6 is unchanged on that background, so the wild-type septin's cellular role is dispensable even though its coiled coil is required within the fusion. Ono R et al. 2005
- E4strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
Homodimerisation of MLL by synthetic dimerisation modules mimics bona fide MLL fusion proteins, resulting in Hox gene activation and enhanced self-renewal of haematopoietic progenitors, and the coiled-coil oligomerisation domains of AF1p/Eps15 and GAS7 are necessary and sufficient for transformation by the respective MLL fusions.
Dimerization contributes to oncogenic activation of MLL chimeras in acute leukemias. · So CW, Lin M, Ayton PM, Chen EH, Cleary ML · Cancer Cell 2003What this evidence cannot show
Read from the abstract; the full text is subscription-only and cell.com returned HTTP 403, so the identity of the synthetic dimerisation module, the construct boundaries and the quantitative comparison against the genuine fusions are NOT verified. This is the evidence that most strongly opposes the hypothesis and it is cited at full strength for that reason.
Evidence that argues against this (1)
- GAS7, one of the two partners in this very result, was later shown to self-assemble through F-BAR hydrophilic loops into two-dimensional membrane-templated SHEETS rather than as a coiled-coil dimer — so a partner whose module the sufficiency experiment classed as 'a coiled coil' in fact oligomerises in a lattice-like way, which is the shape the hypothesis proposes. Hanawa-Suetsugu K et al. 2019
- E5strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
Induced dimerisation of a truncated MLL immortalises bone marrow and imposes a reversible block on myeloid differentiation with upregulation of Hoxa7, Hoxa9 and Meis1, and dimerised MLL binds with higher affinity than undimerised MLL to a CpG island within the Hoxa9 locus; however, MLL-AF9 is not dimerised in vivo.
Dimerization of MLL fusion proteins immortalizes hematopoietic cells. · Martin ME, Milne TA, Bloyer S, Galoian K, Shen W, Gibbs D, et al. · Cancer Cell 2003What this evidence cannot show
Read from the abstract. The 'MLL-AF9 is not dimerised in vivo' statement is a negative biochemical result whose assay is not described in the abstract and has not been inspected for this entry; it carries a lot of weight in this argument and deserves the scepticism due to any unexamined negative.
Evidence that argues against this (1)
- This result cuts against the hypothesis on its main axis: a chemically induced dimer, with no cytoskeletal polymer involved, reproduces the transcriptional and self-renewal phenotype of the genuine fusions. If pure proximity is sufficient, tubulin binding can be at most one way to achieve it. So CW et al. 2003
- E6strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
GAS7's F-BAR domain possesses unique hydrophilic loops that permit two-dimensional sheet-like assembly on flat membranes, in contrast to most BAR-domain proteins which form helical submicron-diameter filaments; the same assemblies are observed on phagocytic cups and on liposomes, and mutating the loops abolishes both membrane localisation and phagocytosis.
Phagocytosis is mediated by two-dimensional assemblies of the F-BAR protein GAS7. · Hanawa-Suetsugu K, Itoh Y, Ab Fatah M, Nishimura T, Takemura K, Takeshita K, et al. · Nat Commun 2019What this evidence cannot show
Purified protein on membranes and macrophage imaging. The assemblies are MEMBRANE-templated, not microtubule-templated, so this supports the general idea that a KMT2A partner can oligomerise as a lattice while being neutral or unhelpful on the specific tubulin claim. It is cited as a partial analogy and labelled as such.
Evidence that argues against this (1)
- GAS7 is an actin-cytoskeleton and membrane protein, not a microtubule protein, so using it to support a tubulin-specific hypothesis stretches the pattern; if the shared property is 'binds a cytoskeletal polymer' rather than 'binds tubulin', the hypothesis becomes much weaker because nearly any cytoplasmic partner would qualify. Chao CC et al. 2005
the minimal transforming module is named a tubulin-binding domain by its discoverers
Mllt11 associates with stabilised microtubules by proteomics in cortical neurons
proposes templating on a polymer rather than pairwise dimerisation; the synthetic-dimeriser result is the strongest argument against it
AF1q oligomeric state is a stated falsifier; the transformation arm decides whether the pattern is worth pursuing