MOF/KAT8 is required by both KMT2A::MLLT3 and NUP98::HOXA9 — which is why it is the wrong kind of convergence
also called MOF KAT8 shared requirement · convergence versus lineage requirement · H4K16ac in myeloid leukaemia
An enzyme called MOF is needed by two completely different kinds of fusion leukaemia, which sounds like a shared vulnerability worth attacking. It is not: MOF is needed by essentially every leukaemia cell we can measure, so the overlap tells us about what myeloid leukaemia needs in general, not about what these two fusions share.
The same thing, for a scientist
Two fusion families requiring the same protein is only interesting if the requirement is **greater in them than in lineage-matched leukaemia without the fusion**. MOF (KAT8) fails that test and menin (MEN1) passes it, and the contrast is the cleanest way this graph has to define what 'convergence' should mean. Conditional *Mof* deletion reduces tumour burden and prolongs survival in MLL-AF9 leukaemia; catalytically inactive MOF fails to rescue, so the **enzymatic** activity is required; and *"Mof inactivation suppressed leukemia development in an NUP98-HOXA9-driven AML model"* [E1]. **An important correction to how that sentence is usually quoted: the catalytic work was done in MLL-AF9 only.** The two HAT-dead constructs, the H4K16ac western blots, the RNA-seq and the gamma-H2AX readout are all MLL-AF9 experiments; the NUP98-HOXA9 arm received **gene-level *Kat8* deletion and nothing catalytic** [E5]. So "MOF catalytic activity is required for both MLL-AF9 and NUP98-HOXA9" is not what the paper shows, and this entry does not assert it. The authors themselves generalise rather than particularise: the activity *"may be important for multiple AML subtypes"* [E1], and the mechanism they report — loss of DNA-damage-repair gene expression, increased γH2AX foci, loss of global H4K16 acetylation [E1] — is a housekeeping requirement, not a fusion-directed one. Run against DepMap in this cycle, **KAT8 shows a KMT2A-rearranged versus other-leukaemia Cohen's d of 0.003** (means −1.4888 vs −1.4893) while being strongly required in both, against **d = −1.879 for MEN1** in the same lines [E2]. So MOF is a deep, lineage-wide requirement with no fusion specificity detectable at all, and MEN1 is a fusion-specific one.
The distinction, stated as a rule
A protein required by fusion family A and by fusion family B can be one of three things:
- A convergent fusion dependency — required by both because of what the fusions do, and less required by lineage-matched leukaemia without either fusion. This is the interesting case, and it is what makes a target better than a one-family quirk.
- A lineage requirement — required by myeloid leukaemia generally, with the two fusion families inheriting it along with everything else in the lineage.
- A cell requirement — required by essentially all proliferating cells.
The experiments that establish 1 and the experiments that establish 2 look identical if you only ever test the fusions. Mof deletion suppressing both MLL-AF9 and NUP98-HOXA9 leukaemogenesis 1 is compatible with all three, because the comparator arm — a lineage-matched leukaemia driven by neither fusion — was not part of the design.
The discriminating comparison is against lineage-matched non-fusion leukaemia, not against normal cells. That is the comparison this graph's DepMap machinery supplies, and it is why the lineage split is a standing rule here rather than a nicety.
What the DepMap numbers say
From ana-depmap-nup98-men1-power, over 54 leukaemia lines (11 KMT2A-rearranged, 42 other leukaemia) 2:
| gene | mean in KMT2A-r | mean in other leukaemia | Cohen's d | reading |
|---|---|---|---|---|
| MEN1 | −1.2629 | −0.4174 | −1.879 | strongly fusion-specific |
| KAT8 (MOF) | −1.4888 | −1.4893 | 0.003 | deeply required, zero fusion specificity |
| ACTB (control) | −0.9493 | −0.7626 | −0.490 | the per-line offset floor |
KAT8's specificity is not merely small; it is smaller than the pan-essential control's, which means it is indistinguishable from nothing. And its absolute requirement is the deepest in the panel, deeper than ACTB. That is the signature of category 2 or 3 above.
The honest limit: at n=11 versus n=42 the minimum detectable effect is roughly d ≈ 1.0, so a modest fusion-specific KAT8 difference is not excluded — only a large one is. And because there are zero NUP98-rearranged DepMap lines, the NUP98 half of the published claim could not be tested this way at all; the DepMap contribution here is the KMT2A-versus-other comparison, and the NUP98 arm rests entirely on the mouse model 1.
Why the mechanism reported is itself a warning sign
The route by which Mof loss hurts MLL-AF9 cells, as reported 1:
- downregulation of DNA damage repair genes controlled by MOF
- a significant increase in γH2AX nuclear foci
- loss of global H4K16 acetylation in the tumour genome
None of these is a statement about the fusion's target genes. A global histone-acetylation loss with a DNA-damage phenotype is what removing a broadly acting acetyltransferase does to a rapidly dividing cell. Contrast the menin-MLL1 result in the same two families, where inhibition evicts the fusion oncoprotein itself from a specific set of target genes and collapses the HOXA/MEIS programme 3. One is a mechanism aimed at the lesion; the other is a mechanism aimed at the cell.
MOF's physical association with MLL1 4 makes the confusion easy: a protein that binds MLL1 and is required by an MLL1-fusion leukaemia looks like a fusion dependency. The binding is real and the requirement is real; the specificity is what is missing.
And MOF loss does not even go through HOX
The clinching detail is in the paper's own text: "MOF loss does not lead to clear downregulation of MA9 target genes such as Hoxa9 and Hoxa cluster genes" 5.
That matters because the entire reason MOF looked like a candidate for cross-family convergence is that both KMT2A-rearranged and NUP98-rearranged leukaemia are HOXA-driven. If removing MOF kills the cells without collapsing the HOXA programme, then whatever MOF is doing is not the thing the two fusion families have in common. The proposed mechanism - DNA damage and chromosomal integrity - is orthogonal to the fusion biology.
The contrast with menin is exact. Menin-MLL1 inhibition collapses the HOXA/MEIS1 programme and evicts the fusion from chromatin; MOF loss kills the cell and leaves the programme running.
Two further practical points from the same work 5:
- The inhibitor used, MG149, is weak and non-specific (IC50 19.6-50.9 uM) and also killed Kasumi-1, U937 and K562; the authors concede it "is not particularly potent". No selective MOF inhibitor supports any of this.
- The authors themselves speculate that "it may be that... it is in fact the HAT activity of MOF that is required... and MLL1 merely functions as a scaffolding protein" - a reading that, if right, moves MOF further from rather than closer to a fusion-directed target.
What evidence would settle it
Three experiments, in increasing cost, each of which would convert this from an argument into a result:
- Three to four NUP98-rearranged leukaemia lines in DepMap. The same run that produced the KAT8 null would then produce a NUP98-versus-other-leukaemia KAT8 comparison and a NUP98 MEN1 comparison in one pass. Power analysis says about three lines suffice for a MEN1-sized effect (
ana-depmap-nup98-men1-power). - A three-arm conditional-deletion experiment in mice: Mof deleted in MLL-AF9, in NUP98-HOXA9, and in a lineage-matched non-fusion myeloid leukaemia (for example an NRAS/Trp53 or CEBPA-driven model). If the third arm dies as readily, the convergence claim is dead. This arm is the one the published study did not run.
- Dose-matched comparison of therapeutic index: the relevant quantity for a target is not whether the leukaemia needs it but whether the leukaemia needs it more than normal haematopoiesis does. For MOF, whose requirement is deep everywhere, the prior should be that the window is narrow.
The falsifier of this entry's own position is simple: a lineage-matched non-fusion myeloid leukaemia that tolerates Mof loss. That result would make MOF a genuine cross-family convergence and this entry wrong.
The generalisation this beat should carry forward
Applied to the rest of the NUP98 story: KAT6A and KAT7 have now been reported as dependencies in NUP98-rearranged AML, and KAT7 is already in this graph as a dependency of KMT2A-rearranged leukaemia. That is the same pattern as MOF — two families, one acetyltransferase, no lineage-matched comparator reported. Before that convergence is treated as a target rationale it should be run through the same test, and this entry exists so that the test is available rather than reinvented.
Menin passes the test. That is the reason this beat argues NUP98 is the second menin-dependent family rather than merely another myeloid leukaemia.
| rule | A protein required by two fusion families is a convergent fusion dependency only if it is required MORE than in lineage-matched leukaemia without either fusion. |
|---|---|
| discriminating comparator | lineage-matched non-fusion myeloid leukaemia, not normal cells |
| kat8 verdict | fails the test: Cohen's d = 0.003 between KMT2A-rearranged and other leukaemia lines, deeper absolute requirement than ACTB |
| men1 verdict | passes the test: Cohen's d = -1.879 in the same lines |
| nup98 arm testable in depmap | false |
| minimum detectable effect at current n | approximately d = 1.0 |
| falsifier of this entry | a lineage-matched non-fusion myeloid leukaemia that tolerates Mof loss |
| same pattern to watch | KAT6A/KAT7 reported as NUP98-rearranged dependencies while KAT7 is already a KMT2A-rearranged dependency, again with no lineage-matched comparator reported |
| correction to the common paraphrase | The claim 'MOF catalytic activity is required for both MLL-AF9 and NUP98-HOXA9' is not supported. The catalytic-dead rescue, H4K16ac, RNA-seq and gamma-H2AX work were MLL-AF9 only; NUP98-HOXA9 received gene-level Kat8 deletion alone. |
| kat8 does not act through hox | MOF loss does not lead to clear downregulation of Hoxa9 or Hoxa cluster genes. |
| nup98 hoxa9 kat8 deletion survival | median 117 days vs 28-39 days across five control groups (p < 0.001), gene-level deletion |
| inhibitor quality | MG149, IC50 19.6-50.9 uM, non-selective (also kills Kasumi-1, U937, K562); no selective MOF inhibitor exists |
| current nup98 myst dependency | KAT6A/MOZ and KAT7/HBO1 with BRPF1, read out on H3K23ac, not KAT8/MOF |
kat8 mechanism reported
[ "DNA damage repair gene downregulation", "increased gamma-H2AX foci", "global H4K16ac loss" ]
men1 mechanism reported
[ "eviction of MLL1 and the NUP98 fusion from specific target genes", "HOXA and MEIS1 programme collapse" ]
settling experiments
[ "three to four NUP98-rearranged lines deposited in DepMap", "three-arm conditional Mof deletion including a lineage-matched non-fusion myeloid leukaemia", "therapeutic-index comparison against normal haematopoiesis" ]
- Targeting a deep lineage-wide requirement means the therapeutic window is set by normal myelopoiesis, and MOF's DepMap requirement is deeper than beta-actin's in every leukaemia line measured - that is what a narrow window looks like before any drug exists.
- A MOF inhibitor developed on a two-fusion-family rationale would be tested in exactly the populations where its lineage requirement guarantees activity, producing a positive trial that says nothing about selectivity.
- If this entry is right, the published MOF convergence should not be used to argue that NUP98 and KMT2A families share a druggable node; using it that way would put a lineage requirement in the same sentence as menin and devalue the menin claim by association.
- Does a lineage-matched non-fusion myeloid leukaemia tolerate Mof loss? Nobody has published the arm, and it is the single result that would decide whether MOF is convergence or lineage.
- Is the same lineage-confound argument true of KAT6A and KAT7 in NUP98-rearranged AML, where the comparator arm is likewise absent?
- What sample size would exclude a modest (d = 0.3) fusion-specific KAT8 effect, and is DepMap's KMT2A-rearranged arm ever going to reach it?
- Does MOF's requirement differ between transformation and maintenance? The published experiment tests leukaemogenesis and DepMap tests maintenance, and the two disagree, which may be a real biological distinction rather than a discrepancy.
- E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
Conditional Mof deletion reduces tumour burden and prolongs survival in MLL-AF9 leukaemia, catalytically inactive MOF fails to rescue so the acetyltransferase activity is required, the mechanism involves loss of DNA-damage-repair gene expression and global H4K16 acetylation, and Mof inactivation also suppressed leukaemia development in a NUP98-HOXA9-driven AML model.
Histone Acetyltransferase Activity of MOF Is Required for MLL-AF9 Leukemogenesis · Valerio DG, et al. · Cancer Research 77(7):1753-1762 2017What this evidence cannot show
The NUP98-HOXA9 result is one sentence of the abstract and the model is a mouse leukaemogenesis assay, not an established-disease treatment assay; the two fusions were not tested under identical conditions in the same figure as far as the abstract reveals. Crucially, NO lineage-matched non-fusion myeloid leukaemia arm is reported, which is the comparison that would distinguish a fusion dependency from a myeloid one. The authors' own conclusion, that the activity 'may be important for multiple AML subtypes', is a generalisation away from fusion specificity.
Evidence that argues against this (1)
- In DepMap leukaemia lines run in this cycle, KAT8 shows a KMT2A-rearranged versus other-leukaemia Cohen's d of 0.003 while being more deeply required than the pan-essential ACTB control in both groups. A requirement with no measurable differential against lineage-matched leukaemia is a lineage or cell requirement, and this directly opposes reading the shared MLL-AF9/NUP98-HOXA9 requirement as fusion convergence. Open Targets / Broad Institute DepMap 2026
- E2moderate confidence 3 of 5: One strong study, or several smaller consistent ones.dataset
Across 54 DepMap leukaemia lines, KAT8 shows a KMT2A-rearranged versus other-leukaemia difference of +0.0006 (Cohen's d = 0.003) while MEN1 shows -0.8455 (d = -1.879), and KAT8's absolute requirement is deeper than the ACTB pan-essential control in both groups.
DepMap CRISPR (Chronos) gene effect via the Open Targets Platform GraphQL API · Open Targets / Broad Institute DepMap · Open Targets Platform; run as analyses/depmap-nup98-rearranged-men1-power in this repository 2026What this evidence cannot show
n = 11 KMT2A-rearranged versus 42 other leukaemia gives a minimum detectable effect around d = 1.0, so a modest fusion-specific KAT8 difference is not excluded, only a large one. The DepMap release version is not stated by the API. Effects are uncorrected for each line's global sensitivity offset, which the ACTB row bounds. Cell lines are not primary leukaemia and a uniform requirement in culture can differ in vivo. **The power calculation is optimistic because the n is wrong.** The eleven KMT2A-rearranged lines are **nine patients** — SEMK2 is a Cellosaurus child record of SEM and MOLM-14 of MOLM-13 — so a minimum detectable effect computed on n=11 independent samples understates the true detectable effect. See `concept-the-eleven-lines-are-nine-patients`. This is the only entry in the graph found to draw a statistical inference over the panel size; a targeted audit of all 990 entries turned up no others.
Evidence that argues against this (2)
- Established cell lines test maintenance, not transformation. The published claim is about leukaemogenesis - whether MOF is needed to MAKE the leukaemia - and a gene can be uniformly required for maintenance while being selectively required for transformation by a particular fusion. This DepMap comparison therefore cannot refute the published experiment; it constrains how the result should be generalised to a therapeutic target. Valerio DG et al. 2017
- Recomputing on nine patients rather than eleven lines widens the minimum detectable effect, so a negative result in this entry becomes weaker rather than stronger - the direction that matters, since the entry's conclusion is that the KAT8 requirement is myeloid rather than fusion-specific. A null on an overstated n is less informative than it appears. SIB Swiss Institute of Bioinformatics 2026
- E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
Menin-MLL1 inhibition in NUP98-fusion leukaemia evicts both MLL1 and the NUP98 fusion protein from a specific set of target genes and suppresses the HOXA/Meis1 programme, a lesion-directed mechanism rather than a global one.
The menin-MLL1 interaction is a molecular dependency in NUP98-rearranged AML · Heikamp EB, et al. · Blood 139(6):894-906 2022What this evidence cannot show
Mouse cell lines and PDX; the tool compound VTP50469, not a clinical agent. 'A critical set of genes' is the authors' characterisation and the selection criteria for that set were not read in this session, so the specificity of the eviction is taken from the abstract.
Evidence that argues against this (1)
- MEN1 is itself required in normal haematopoiesis and endocrine tissue, so fusion specificity in a leukaemia-versus-leukaemia comparison does not by itself establish a therapeutic window. This graph's endocrine and normal-tissue beats record the constraint directly. Syndax Pharmaceuticals 2026
- E4established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper
MOF physically associates with MLL1 and the two coordinate function, which is why a MOF requirement in MLL1-fusion leukaemia is easy to mistake for a fusion-directed dependency.
Physical association and coordinate function of the H3 K4 methyltransferase MLL1 and the H4 K16 acetyltransferase MOF · Dou Y, et al. · Cell 121(6):873-885 2005What this evidence cannot show
Biochemical association in a non-leukaemic context, twenty years old; complex composition assignments from that era have in several cases been revised. It is cited here for the association itself, not for stoichiometry or for any claim about fusion proteins.
Evidence that argues against this (1)
- A physical association with MLL1 predicts, if anything, that MOF would be MORE required in MLL1-fusion leukaemia than elsewhere. The DepMap comparison finds no such enrichment, so the association does not translate into differential requirement, and the biochemistry should not be used to infer the genetics. Open Targets / Broad Institute DepMap 2026
- E5strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
In the MOF study, the catalytic-dead rescue constructs (CoA-binding-site deletion and G327E), the H4K16ac measurements, the RNA-seq and the gamma-H2AX readout were performed in MLL-AF9 only; the NUP98-HOXA9 arm received gene-level Kat8 deletion alone (median survival 117 days versus 28-39 days across five control groups, p < 0.001); MOF loss does not clearly downregulate Hoxa9 or Hoxa cluster genes; and the MOF inhibitor used, MG149, has an IC50 of 19.6-50.9 uM and also killed Kasumi-1, U937 and K562.
Histone Acetyltransferase Activity of MOF Is Required for MLL-AF9 Leukemogenesis · Valerio DG, et al. · Cancer Research 77(7):1753-1762 2017What this evidence cannot show
These details are from the full text, read in a delegated research pass in this session rather than by the author of this entry. Incomplete Cre excision at death is reported in the in vivo arms, so surviving cells are escapers and the survival differences understate the effect of complete deletion. The Hoxa9 null result is a statement about clear downregulation and does not exclude a modest effect.
Evidence that argues against this (1)
- That the catalytic requirement was shown only in MLL-AF9 does not mean it is absent in NUP98-HOXA9; the experiment was simply not run. Treating the absence of the catalytic arm as evidence against a catalytic requirement would be the mirror of the overstatement this entry corrects. Michmerhuizen NL et al. 2025
the convergence test this entry applies to Tip60; the comparator arm is absent for both
HDAC3 fails the same convergence test, and fails it on the original paper's own comparator arms
the eraser half reaches the same verdict the writer half did: the attributable enzyme is not fusion-specific
the MOF requirement is not fusion-selective; this entry supplies the mark it is measured in
KAT8 Cohen's d 0.003, no fusion specificity
MOF catalytic activity required for both MLL-AF9 and NUP98-HOXA9 leukaemogenesis
- related to →Lost from the KMT2A half: the MOF/KAT8 contact - required in the cell, but not shown to matter to the fusion
the MLL1-MOF contact that makes the requirement look fusion-directed
- depends on →KMT2A::MLLT3 (MLL-AF9), t(9;11)(p21.3;q23.3)
Mof deletion reduces MLL-AF9 tumour burden and prolongs survival
bounds the analogy: shared requirement is not convergence without a lineage-matched comparator
- related to →KAT7 / HBO1 — an enzyme, a leukaemia-stem-cell dependency, and an inhibitor already in a crystal
same untested pattern: KAT7 a dependency in both families, no comparator arm reported