H2B ubiquitination is upstream of the fusion's defining mark and is still not a target, because the fusion uses a general mechanism rather than a special one
also called the H2Bub null hypothesis · upstream but not selective
KMT2A-fusion leukaemia needs the H2B ubiquitin tag, but so does almost every other dividing cell, because the fusion is exploiting the ordinary transcription machinery rather than a special version of it. If that is right, no drug aimed at the tag-writing enzymes will ever have a safety margin.
The same thing, for a scientist
The fusion's H3K79me2 programme is licensed by H2BK120ub, which is written co-transcriptionally by RNF20/RNF40-WAC recruited through PAF1C on elongating Pol II. **Nothing about that machinery is fusion-specific.** The hypothesis is that the fusion achieves its selectivity by *where* it sits (menin/LEDGF-directed chromatin residence) and not by *what* it recruits at the ubiquitin layer, and that the H2Bub layer is therefore a shared, general requirement with no therapeutic window. The genome-wide DepMap evidence is consistent: the three writers rank 16,229, 9,953 and 7,779 of 17,728 on KMT2A selectivity while MEN1 ranks 1 [E1]. The hypothesis makes hard predictions and is falsifiable by experiments that are neither exotic nor expensive, which is why it is written down rather than assumed.
Predictions and what would kill this
Predictions if the hypothesis is true.
- H2Bub occupancy at fusion target genes will be proportionate to their transcription, not elevated above equally transcribed non-target genes.
- RNF20 degradation will reduce H3K79me2 at fusion targets and at other highly transcribed genes to a similar degree.
- A non-KMT2A leukaemia line will show the same sensitivity to RNF20 loss as a KMT2A-rearranged line at matched knockdown.
- Menin inhibition will reduce H2Bub at fusion targets, because the fusion's chromatin residence is what drives transcription there — i.e. H2Bub is downstream of the fusion, not a co-equal input.
Falsifiers. Any one of these kills it:
- H2Bub is enriched at fusion target genes beyond what their transcription level predicts (Wang et al. report enrichment at fusion targets but do not, in the abstract, normalise to transcription — so this is genuinely open).
- An RNF20 degron produces a dose-response window in KMT2A-rearranged lines that is absent in matched non-rearranged lines.
- The fusion, or a fusion-associated complex, is shown to recruit RNF20/RNF40 or MSL2 directly rather than through general Pol II coupling.
- RNF20 loss is selectively lethal in an infant B-ALL KMT2A::AFF1 model, the context DepMap does not contain.
| statement | H2B ubiquitination is a general, transcription-coupled requirement that the KMT2A fusion exploits without modifying, so the writers and carriers of the mark are necessary in KMT2A-rearranged leukaemia without being selectively necessary, and no therapeutic window exists at that layer. |
|---|---|
| origin | data-pattern |
| prior probability | 0.65 |
predictions
[ "H2Bub at fusion target genes is proportionate to transcription, not elevated above equally transcribed non-target genes", "RNF20 degradation reduces H3K79me2 similarly at fusion targets and at other highly transcribed genes", "Matched non-KMT2A leukaemia lines are as sensitive to RNF20 loss as KMT2A-rearranged lines", "Menin inhibition lowers H2Bub at fusion targets, placing the mark downstream of fusion chromatin residence" ]
experiments
[ "exp-rtf1-h2bub-separation-of-function-kmt2a" ]
analyses
[ "ana-depmap-h2bub-machinery-kmt2a-selectivity" ]
falsifiers
[ "H2Bub enrichment at fusion target genes that exceeds what their transcription level predicts", "An RNF20 degron dose-response window present in KMT2A-rearranged lines and absent in matched non-rearranged lines", "Demonstration that the fusion or a fusion-associated complex recruits RNF20/RNF40 or MSL2 directly rather than through general Pol II coupling", "Selective lethality of RNF20 loss in an infant B-ALL KMT2A::AFF1 model" ]
- If this hypothesis is right, the correct action is to stop proposing H2Bub writers as targets, and the graph should say so rather than leaving a plausible-sounding pathway entry that a reader will mistake for a lead.
- If it is wrong, it is wrong in the direction that matters most - an infant B-ALL-specific window that a cell-line panel cannot see - so the cost of accepting it prematurely is higher than the cost of testing it.
- Is 0.65 the right prior? It is set by the DepMap null plus the mechanistic generality argument, and it would move sharply on a single infant B-ALL experiment.
- If the hypothesis holds, does it generalise to other chromatin layers the graph treats as fusion-specific? The same reasoning - the fusion exploits general machinery from a specific location - would apply to several existing entries.
- E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.dataset
In a genome-wide DepMap scan of 17,728 genes comparing 11 KMT2A-rearranged with 43 other leukaemia lines, the H2B ubiquitination writers show no selectivity (RNF20 +0.076, rank 16,229; RNF40 +0.006, rank 9,953; WAC -0.011, rank 7,779; UBE2A -0.020; UBE2B +0.007; UBE2E1 +0.108; MSL2 -0.015), while MEN1 ranks 1 (-0.866) and DOT1L ranks 184 (-0.231).
DepMap CRISPR (Chronos) genome-wide KMT2A-selectivity scan of 17,728 protein-coding genes, 11 Cellosaurus-annotated KMT2A-rearranged leukaemia lines vs 43 other leukaemia lines · Broad Institute DepMap; Cellosaurus; Open Targets Platform; analysis run by this project (analyses/depmap-kmt2a-selective-scan-full) · Database and derived analysis 2026What this evidence cannot show
Eleven KMT2A-rearranged lines, AML-weighted, none infant-derived; CRISPR knockout compresses selectivity for pan-essential genes; redundancy is invisible to single-gene knockout. See ana-depmap-h2bub-machinery-kmt2a-selectivity for the full treatment.
Evidence that argues against this (1)
- Wang et al. found H2B ubiquitination enriched in the body of MLL-fusion target genes and Rnf20 required to maintain Dot1l-mediated H3K79 methylation at Hoxa9 and Meis1, with proliferation inhibited in vitro and in vivo. If the enrichment is disproportionate to transcription, prediction 1 of this hypothesis is already refuted and nobody has checked. Wang E et al. 2013
- E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper
PAF1C, which recruits the H2B ubiquitination machinery to elongating Pol II, interacts with MLL through sequences flanking the CxxC domain, and PAFc knockdown disrupts MLL fusion protein-mediated transcriptional activation and MLL recruitment to target loci; the machinery is thus shared between the fusion and general transcription rather than dedicated to the fusion.
The PAF complex synergizes with MLL fusion proteins at HOX loci to promote leukemogenesis. · Muntean AG, Tan J, Sitwala K, Huang Y, Bronstein J, Connelly JA, Basrur V, Elenitoba-Johnson KS, Hess JL · Cancer Cell 2010What this evidence cannot show
PMID fetched and title read back by this beat; abstract read in full, full text not read. The inference that a shared complex implies a shared requirement is this hypothesis's inference, and a shared complex can still be used in a fusion-specific way - which is precisely what the falsifiers test.
Evidence that argues against this (1)
- PAF1C interacting with MLL through sequences flanking the CxxC domain places the coupling on the RETAINED portion, which means it is present in the fusion — and that argues for a shared rather than selective requirement, supporting this hypothesis while also showing the evidence is about MLL1 rather than about the fusion specifically. Cancer Cell 2010
the writers rank 16,229, 9,953 and 7,779 while MEN1 ranks 1
the mark is necessary without being selectively necessary
separates the elongation and ubiquitination explanations of the KMT2A-PAF1C contact