The KMT2A-fusion causal chain: six links, one measured ordering, and two arms that are parallel rather than sequential

Abstract read 2026-09-20. Murine myeloid progenitors, engineered constructs, 2004 methodology; a transformation assay, not a measurement of occupancy kinetics. This establishes that link 1 is REQUIRED. It says nothing about what link 1 hands to link 2, and no experiment in this chain measures that hand-off.

• The same targeting module is present in wild-type KMT2A, which is not leukaemogenic, so CXXC binding is necessary and by itself carries no oncogenic information. Work on MOZ-rearranged leukaemia shows wild-type KMT2A performing the identical unmethylated-CpG promoter activation through AEP/SEC, DOT1L and p300/CBP, i.e. the fusion inherits a normal circuit rather than creating one. Miyamoto R et al. 2020

Abstract read 2026-09-20. Murine myeloid models with conditional Men1 alleles. The requirement is genetic and acute. The POSITIONAL claim that menin holds the fusion at its target loci is a separate proposition and is the weaker one (see E4).

• Menin is required, but requirement is not specificity: KMT2B carries an equivalent N-terminal menin motif and cannot be made into a transforming fusion, so the menin contact cannot be what makes KMT2A oncogenic. Bach C et al. 2009

Abstract read 2026-09-20. Biochemistry plus transformation assays in murine cells. Establishes a three-way requirement; it does not establish that the module acts as a POSITIONAL clamp on the fusion, which is the form in which the chain usually states it.

• In MLL-transformed progenitors, LEDGF knockdown DECREASES wild-type MLL occupancy and INCREASES MLL-fusion occupancy, and an MLL-AF9 F129A mutant unable to bind LEDGF still associates with MLL target genes - LEDGF is required for retention of the wild-type protein, not of the fusion. Zhu L et al. 2016

**Full text read 2026-09-20** via efetch db=pmc on PMC4930721 (Europe PMC fullTextXML returned HTTP 500; brief rule on the four routes), 61,451 characters after tag-stripping then entity-unescaping. The result is in the body, not the abstract - the abstract is about ASH1L and H3K36me2 - which is why it is easy to miss and why this entry quotes it. Verbatim: *"a missense mutant of MLL-AF9 (F129A) that cannot interact with LEDGF retained an ability to associate with MLL target genes"* and *"Decreased occupancies of WT MLL and increased occupancies of MLL fusion proteins were also observed for MLL-AF10 and MLL-ENL in LEDGF knockdown cells"*. This is the most important qualification on link 2 in this chain and it is usually omitted: the menin/LEDGF module is functionally required for leukaemia while being POSITIONALLY dispensable for the fusion, so the familiar clamp picture is an inference from the wild-type protein transplanted onto the fusion. It replicates across three fusions within one paper and one laboratory; it has not been repeated by an independent group, and not in a human KMT2A::AFF1 line.

• The menin contact is nonetheless the target of the only approved drug in this disease, and clinical resistance arises through MEN1 mutations at the drug-menin interface, so whatever the positional mechanism, disrupting this contact does collapse the programme in patients. Perner F et al. 2023

Abstract read 2026-09-20; the fetched record carries no PMCID, so full text was not obtained. 'Mll2' here is KMT2B (Q9UMN6, 2,715 aa), not KMT2D - resolved by the authors' description of it as the closest relative of MLL, per brief rule 66. Engineered constructs in murine cells; a negative transformation result in one assay system with one set of constructs. This is the load-bearing argument that the menin interface is a requirement and not the determinant.

• KMT2B is not irrelevant to this disease: Mll2 rather than Mll1 is reported to sustain established MLL-rearranged AML in trans, so the paralogue matters - just not as a fusion backbone. Chen Y et al. 2017

Abstract read 2026-09-20. Biochemical purification plus knockdown in human cells. Establishes co-association and requirement. It does not order SEC recruitment relative to menin tethering or to H3K79 deposition, and no published experiment does.

• AFF1-SEC and AFF4-SEC are distinct complexes with largely non-overlapping gene target specificity, so 'the SEC' in this chain is at least two different assemblies and a result obtained on one does not transfer to the other. Lu H et al. 2015

Abstract read 2026-09-20. Reconstitution from purified components, so the negative concerns a stable biochemical association and not co-occupancy in cells. It is nonetheless the result that breaks the chain metaphor: the elongation arm and the methylation arm are separate assemblies on the same partner surface, not consecutive links.

• The AF9 AHD binds AF4, DOT1L, BCOR and CBX8 at one mutually exclusive site, so the two arms are not merely distinct - they compete directly for the retained partner surface, which makes a serial ordering of them incoherent rather than merely unmeasured. Schmidt CR et al. 2020

Abstract read 2026-09-20. Murine MLL-AF9, myeloid. The selectivity is the strong part. That H3K79me2 MAINTAINS the programme rather than merely marking it is supported by the Dot1l loss experiment, but the causal ordering between mark loss and transcriptional loss is not resolved by it.

• Pinometostat reduced H3K79 methylation as intended in 51 adults and produced only two complete remissions, both in t(11;19), so removing the mark pharmacologically does not reproduce the genetic result, and the link from mark to programme is weaker in humans than in mice. Stein EM et al. 2018

Abstract read 2026-09-20. This is the ONLY link in the chain measured on an AFF1 backbone in a lymphoid context, and it is still a mouse. Every other step with in-vivo evidence has it in MLL-AF9 or MLL-ENL myeloid models.

• Non-canonical H3K79me2-dependent pathways promote survival of KMT2A-rearranged cells, with internally calibrated ChIP showing paradoxical H3K4me3 increases and H3K27me3 loss at pinometostat-downregulated loci, so the mark's output is not a single programme-maintenance function. Richter WF et al. 2021

Abstract read 2026-09-20. Human lines and primary samples; a maintenance result. shRNA is partial depletion, so it bounds the programme's dose sensitivity only loosely.

• Hoxa9-null Mll-AF9 knock-in mice develop myeloid leukaemia with similar incidence and time course to Hoxa9-competent mice, only with a more immature phenotype - HOXA9 is required for maintenance of established human disease and dispensable for initiation in mouse. Kumar AR et al. 2004

Abstract read 2026-09-20. Germline knockout, so developmental compensation cannot be excluded, and the result concerns initiation rather than maintenance. It is nonetheless the clearest statement that link 5 is a buffered module and not a single-gene bottleneck.

• In human KMT2A-rearranged leukaemia HOXA9 depletion kills selectively, so the mouse-initiation result must not be read as 'HOXA9 does not matter' - the two experiments ask different questions of different species at different disease stages. Faber J et al. 2009

Abstract read 2026-09-20. Murine, myeloid, retroviral MLL-AF9. The terminal state of the chain in the index human disease - infant B-lineage KMT2A::AFF1 - is not established by this or by any equivalent experiment.

• Within paediatric t(4;11) there are two subgroups - infants with LOW HOXA expression and breakpoints in introns 11/12, and non-infants with high HOXA - so a single terminal programme cannot be assumed across the population this chain claims to describe. Trentin L et al. 2009

Abstract read 2026-09-20. Murine, myeloid, transgene-level expression. This is the strongest single argument that the chain is causal rather than correlative, and it is a statement about complete genetic removal, not about partial pharmacological occupancy.

• Heavily pretreated KMT2A::AFF1 ALL samples were markedly less menin-inhibitor-sensitive than earlier samples from the same patients despite no prior exposure, with transcriptomics showing sustained on-target effect, so reversibility on complete removal does not predict reversibility under a drug. Mahdavi L et al. 2026

Abstract read 2026-09-20. This is the ONLY measurement of temporal ordering anywhere in this chain, and it orders exactly one pair of events (elongation before chromatin) for exactly one fusion (MLL::AF9) in one cell system. No degron model of KMT2A::AFF1 exists, so the ordering has never been tested on the index fusion.

• Complete protein removal is not partial pharmacological occupancy; clinical material shows resistance arising downstream of sustained target engagement, so the kinetics of degradation do not transfer to the kinetics of inhibition. Mahdavi L et al. 2026

Abstract read 2026-09-20. Single-arm, relapsed/refractory, heavily pretreated, mixed AML and ALL and mixed ages including adults. This is the clinical endpoint of the chain and it is the number any mechanistic account has to be consistent with: complete remission in roughly one in five, not in most.

• Somatic MEN1 mutations at the drug-menin interface mediate clinical resistance and are recovered by an unbiased base-editor screen, so a substantial part of the residual disease is on-pathway escape rather than a failure of the mechanism. Perner F et al. 2023

Abstract read 2026-09-20. Screens in cell lines, not in the index disease. The consequence for this chain is large: an effect of a menin inhibitor can arise through WILD-TYPE KMT2A/B and the polycomb balance rather than through the fusion, and no published menin-inhibitor experiment separates the two. The paper's own H3K27me3 CUT&Tag is in the GENETIC arm (MEN1 knockout versus control: total H3K27me3 unchanged, occupancy reduced at bivalent non-Menin-bound promoters); its VTP50469 arm carries Menin ChIP-seq and RNA-seq and no H3K27me3. Full text read via Europe PMC fullTextXML (PMC7614190) on 2026-09-20. The correct bounded statement is that no H3K27me3 measurement under a menin INHIBITOR has been reported, not that none exists at all; see ana-h3k27me3-under-menin-inhibition-the-absence-is-narrower-than-stated.

• Menin-inhibitor compounds engage menin-MLL1, menin-MLL2 and other menin interactions, yet Men1 deletion in MLL-AF9 cells produces consequences distinct from Mll1;Mll2 co-deletion, so the wild-type route is real but is not the whole of the drug's action either. Chen Y et al. 2019

Abstract read 2026-09-20. This inverts part of the chain's therapeutic arm: the silencing that follows the drug is an ACTIVE deposition by a host polycomb complex, not simply the decay of what the fusion had installed. Cell-line and mouse work; not measured in infant KMT2A::AFF1 material.

• Loss of PRC1.1 subunits such as PCGF1 or BCOR causes menin-inhibitor RESISTANCE by reactivating non-canonical menin targets including MYC - the same complex read as a resistance determinant rather than as an effector, and the two readings have not been reconciled in one system. Zhou X et al. 2024

Abstract read 2026-09-20. AML models; the infant B-lineage index disease is not represented. Together with E17 this means the drug's output is partly executed by a complex belonging to the host cell, so a patient's polycomb state is part of the mechanism.

• The DOT1L/PRC1.1 account places PRC1.1 downstream of H3K79 loss as the effector of silencing, whereas this paper places it as the gatekeeper whose loss confers resistance; both cannot be the complete description and no experiment tests them against each other. Neville D et al. 2026

Abstract read 2026-09-20. Murine and human AML models. The importance for the chain is structural: link 2 is drawn as a fixed protein-protein interface and is in fact a function of nucleotide supply and ribosome biogenesis, which no version of this chain represents. This graph's metabolism beat records an alternative, SAMHD1-centred account in which IMPDH sensitivity is not KMT2A-specific; see concept-metabolic-dependency-consequence-or-independent-kmt2a.

• Menin-inhibitor compounds also engage menin-MLL1 and menin-MLL2 and other menin interactions, so a treatment that lowers menin protein is not equivalent to one that blocks the fusion's motif, and the two are conflated when both are described as acting on link 2. Chen Y et al. 2019

Abstract read 2026-09-20. Expression-array era, modest single-centre cohort, not replicated in a modern cohort within this graph's citations. If it holds, link 5 of this chain is absent or strongly attenuated in the youngest infants - the group the index disease is defined by.

• HOXA9 depletion kills KMT2A-rearranged human lines selectively and HOXA transcripts are the field's default pharmacodynamic readout, so a subgroup with low HOXA is either running a different mechanism or reflecting a measurement artefact, and nothing published distinguishes those. Faber J et al. 2009

Abstract read 2026-09-20. Cell lines only, short assays, no in-vivo or primary-sample arm. It is the cleanest available demonstration that only links 2 and 4 are both required and fusion-selective, while link 3 is mechanistically central and built from common-essential machinery.

• A menin inhibitor combined with a DOT1L inhibitor suppresses fusion targets and MYC targets more profoundly than either alone while sparing normal haematopoiesis, so the two selective links are not independent and their individual selectivity does not predict the combination's. Dafflon C et al. 2017

Abstract read 2026-09-20. Murine transformation assays plus biochemistry. The AFDN contrast is the useful control: constitutive elongation-complex capture is what the common partners supply, and at least one recurrent partner reaches a similar endpoint by another route.

• Not every KMT2A partner is an elongation factor and the chain as written has no place for the non-elongation partners; the SEC family itself splits into AFF1 and AFF4 complexes with non-overlapping targets, so 'the fusion hijacks the SEC' names a family rather than a mechanism. Lu H et al. 2015