How a cytoskeletal partner reaches chromatin at all — the KMT2A N-terminus drags it in, and the traffic runs both ways

The retained KMT2A N-terminus supplies nuclear localisation signals, the AT-hooks, the CXXC domain and the menin/LEDGF-binding region. Nothing about the partner is required for nuclear entry, which is why the partner census can contain proteins with no nuclear function at all without the model collapsing. The direct demonstration in a septin fusion is unambiguous: fusion protein in nuclear speckles, wild-type septin in cytoplasm, same cells, same assay 1.

This is the answer to the naive objection and it is worth stating because the objection is the first thing a reader raises. It is also, on its own, the end of the localisation story — and therefore not very interesting. The interesting part is what follows.

In MLL-AF6 leukaemia the fusion relocates endogenous AF6 into the nucleus; RAS-GTP rises; silencing the fusion restores cytoplasmic AF6 and reduces RAS activation 2. The cytoplasmic pool of a normal protein is depleted by a nuclear oncoprotein, and the cytoplasmic consequence — RAS activation — is a second, independent oncogenic input. This is a loss of function achieved by relocation rather than by deletion, and it is invisible to any analysis that only asks which domains the fusion retains.

Septins are obligate hetero-oligomers. A nuclear MLL-SEPT6 that retains the G domain and coiled coil has, in principle, exactly the surfaces needed to recruit endogenous SEPT2, SEPT7 and SEPT11 into a nuclear complex. That would predict (a) reduced cytoplasmic septin pools, (b) a cytokinesis or membrane-remodelling phenotype, and (c) a genetic dependency of the fusion on the endogenous partner septins. None of these has been looked for. It would also supply an explanation for the otherwise puzzling requirement for the SEPT6 coiled coil, which is dispensable for native filament assembly 3 but required for transformation 1.

Sept6-null mice are normal and MLL-SEPT6 disease is unchanged on that background 4, which rules out SEPT6 haploinsufficiency as a contributor but says nothing about sequestration of OTHER septins by the fusion. The Sept6-null experiment removes one septin; sequestration would remove several at once from the cytoplasm, and the septin-redundancy argument that explains the null phenotype predicts that only a multi-septin perturbation would show anything. The experiment that discriminates is imaging and fractionation of endogenous septins in fusion-expressing cells, not another knockout.

In patients, septin expression is altered in MLL-SEPT leukaemias: SEPT6 is down 1.7-fold (p=0.002) in MLL-SEPT6 disease alongside a 6.4-fold MLL reduction (p=0.008), SEPT9 shows a 4.6-fold downward trend (p=0.077) in MLL-SEPT9, and MLL-SEPT2 and MLL-SEPT6 neoplasms cluster together by septin-expression profile, apart from other AML subgroups 5. Transcript-level change is not protein relocalisation, and these data do not distinguish the two. They do establish that the septin system is perturbed in these leukaemias beyond the fused allele, which is the minimum precondition for the sequestration hypothesis to be worth testing.

1. This entry asserts nuclear localisation signals with no citation, and none exists in the form implied. The sentence "The retained KMT2A N-terminus supplies nuclear localisation signals" has carried no source. The nearest primary source is PMID 9349501 (Joh et al., Oncogene 1997), which assigns nuclear targeting by deletion to a region containing the AT-hook motifs - a region, not a mapped NLS.

KMT2A's nuclear localisation signal has never been mapped as an NLS. (MLL OR KMT2A) AND "nuclear localization signal" returns 2 records, neither of which maps an NLS in KMT2A itself. And a 1997 deletion assignment cannot separate active import from chromatin retention - a fragment that stays nuclear because it binds DNA looks identical to one carrying an import signal. This graph has repeated the NLS claim as though it were characterised.

2. The 'cytoplasmic partner' premise fails for at least two partners.

• EEN / endophilin A2 is "localized predominantly in nuclei of various cell lines of haemopoietic, fibroblast and epithelial origin, in contrast with its reported cytoplasmic localization in neurons and osteoclasts", and it shuttles (PMID 15214844). Its cytoplasmic reputation comes from neurons, not from blood cells.
• AF3p21 carries a bipartite NLS and is nuclear (PMID 11241789).

The mechanism is not wrong; its scope is narrower than stated. For a genuinely cytoplasmic partner - the septin case, with a partner-alone control in the same figure - relocalisation to the nucleus is a real and well-demonstrated phenomenon. What fails is treating the partner census as a list of cytoplasmic proteins dragged nuclear. Some partners were nuclear already, in the lineage that matters, and for those the entry's argument is answering an objection nobody needed to raise.