ENL is the only elongation cofactor that separates fusion-occupied promoters from wild-type KMT2A promoters by more than the H3K4me3 negative control does, in primary infant KMT2A::AFF1
Of the proteins that travel with the fusion, only ENL reliably marks the genes the fusion sits on in real patient samples. DOT1L does so only just, and no better than a general marker of active genes in two of the three samples.
The same thing, for a scientist
**Mean promoter share in the N/C-defined fusion set over the wild-type KMT2A set, per sample** [E1]: | track | p279 | p318 | p454 | |---|---:|---:|---:| | ENL R1 / R2 | 2.04 / 1.90 | 2.01 / 2.18 | 2.77 / 2.78 | | DOT1L R1 / R2 | 1.26 / 1.25 | 1.28 / 1.22 | 1.65 / 1.60 | | H3K4me3 R1 / R2 (the paper's negative control) | 0.98 / 0.97 | 1.05 / 1.06 | 1.04 / 1.06 | | IgG | 0.75 | 0.86 | 1.08 | **And by the enrichment-free version of the same question** - each track's own top 971 promoters, against a random null of 48.6 [E1]: | track | p279 | p318 | p454 | |---|---:|---:|---:| | ENL | 169 (3.5x) | 171 (3.5x) | 265 (5.5x) | | DOT1L | 118 (2.4x) | 107 (2.2x) | 169 (3.5x) | | **H3K4me3 benchmark** | **101 (2.1x)** | **122 (2.5x)** | **129 (2.6x)** | **DOT1L clears the H3K4me3 benchmark in one sample of three and is below it in p318.** ENL clears it in all three by 1.4-2.1x. The published claim is that DOT1L is significantly more enriched at target than wild-type sites in three of four samples and ENL in all of them [E2]; the ratio direction reproduces, and placing the negative control's own value beside it shows DOT1L's margin over an ordinary active promoter is small to absent. **The DOT1L tracks are also the least reproducible in the deposit.** Same-antibody replicate agreement: H3K4me3 0.972-0.979, ENL 0.629-0.785, **DOT1L 0.357 / 0.540 / 0.629**. The p279 DOT1L pair fires this beat's pre-registered U2 guard (replicate rho below 0.5), so p279 DOT1L numbers are reported and are not load-bearing. DOT1L is also the factor whose profile transfers worst between samples (cross-sample rho 0.488-0.613 against 0.709-0.767 for ENL and 0.932-0.970 for H3K4me3). **Global correlation cannot separate these factors and the site-set contrast can.** Within each patient, ENL's rank correlation with KMT2A-N is 0.783 / 0.811 / 0.872 while its correlation with H3K4me3 is 0.764 / 0.791 / 0.851 - a margin of 0.02 that is not an instrument. The pre-registered C3 'neither' branch therefore fires for the genome-wide correlation readout, and the discriminating measurement is the ONCO-versus-WT contrast above.
Why this matters for the DOT1L-inhibitor line of work
Q141 found, in a different fusion, that a DOT1L inhibitor strips menin from fusion promoters about as effectively as destroying the fusion does. That result presumes DOT1L is a structural occupant of fusion promoters. In primary KMT2A::AFF1 material, DOT1L's occupancy margin over a generic active promoter is 2.4x versus 2.1x, 2.2x versus 2.5x, and 3.5x versus 2.6x - i.e. present, weak, and not consistently above an active-promoter mark. That does not contradict Q141, which measured a perturbation rather than a baseline, but it means the baseline DOT1L localisation this project has been assuming is much less crisp in patients than in lines.
One striking single locus
In p318, the lineage-switched AML, the LYZ promoter carries DOT1L at 65.5x and ENL at 16.3x the sample's own mean promoter, with H3K4me3 at 3.0x - while in p279, the same patient's B-ALL, LYZ carries DOT1L at 0.0x and H3K4me3 at 0.09x. LYZ is not called a fusion target in p318 by the N/C criterion (N/C 1.59, below the top decile). So the strongest DOT1L promoter in the AML is a myeloid gene that the fusion is not selectively on.
| script | analyses/q142-aff1-in-primary-infants/extra.py |
|---|---|
| status | run |
| result summary | ENL clears the H3K4me3 benchmark in 3/3 states, DOT1L in 1/3; DOT1L replicate rho 0.357-0.629. |
| pre registered outcome | C1 on the site-set contrast, C3 on global correlation |
- A DOT1L inhibitor's pharmacodynamic readout in a patient will be measured on a track whose replicates agree at rho 0.36-0.63 in the best available primary dataset; the assay noise may exceed the drug effect.
- If ENL is used as the fusion-occupancy surrogate because it is the one that works, an ENL-targeting agent in the same trial makes the surrogate and the target the same molecule.
- Is DOT1L's weak margin in primary cells a reagent problem (replicate rho 0.36-0.63) or a biological one?
- Does ENL's advantage persist at enhancers, where this promoter-anchored design cannot look?
- E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.dataset
In the three primary KMT2A::AFF1 states of GSE252378, ENL is enriched 1.90-2.78-fold at N/C-defined fusion promoters over wild-type KMT2A promoters and its top promoters hit the fusion set at 3.5-5.5x the random null, while DOT1L gives 1.22-1.65 and 2.2-3.5x against an H3K4me3 negative-control benchmark of 2.1-2.6x; DOT1L replicates agree at rho 0.357-0.629.
GSE252378 - AutoCUT&RUN and AutoCUT&Tag profiles of 36 MLL-rearranged leukemia samples (171 GSMs, hg38) · Janssens DH, Henikoff S (Fred Hutchinson Cancer Center / Van Andel Institute) · NCBI Gene Expression Omnibus 2024What this evidence cannot show
Promoter windows, not peaks. Within-sample shares; no spike-in in this deposit. DOT1L and ENL were profiled by AutoCUT&Tag while KMT2A and AFF1 were AutoCUT&RUN, so cofactor-versus-fusion comparisons cross an assay boundary. p279 DOT1L fails this beat's replicate guard.
Evidence that argues against this (1)
- H3K4me3 is deposited by wild-type KMT2A and by the fusion's own activity, so it is not a clean negative control for fusion-target enrichment - using it as the benchmark may set the bar too high and understate DOT1L. The paper uses it as a negative control precisely because it does not distinguish the two site classes, but that is an argument about its ratio, not about its top-promoter overlap. Janssens DH et al. 2024
- E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper
The source paper reports DOT1L significantly more enriched at oncoprotein-target than wild-type sites in three of four samples and ENL in all samples profiled, and uses H3K4me3 as a negative control that does not distinguish them.
MLL oncoprotein levels influence leukemia lineage identities · Janssens DH, Duran M, Otto DJ, Wu W, Xu Y, Kirkey D, Mullighan CG, Yi JS, Meshinchi S, Sarthy JF, Ahmad K, Henikoff S · Nature Communications 15:9341 2024What this evidence cannot show
Read from PMC11522475. The paper's four samples are SEM plus the three patient samples; SEM has no KMT2A N/C libraries in this deposit, so the fourth sample could not be reproduced here.
Evidence that argues against this (1)
- The paper's significance statement is over its own peak intervals with its own statistics; a promoter-window reproduction that agrees in direction is not a test of the significance claim. Janssens DH et al. 2024
ENL 1.9-2.8x at fusion promoters, 3.5-5.5x null on top promoters
DOT1L clears the H3K4me3 benchmark in 1 of 3 primary states
site sets from the N/C control
- related to →Plain-English: what transcription elongation is, and why a traffic light explains this leukemia
elongation cofactors at fusion promoters in primary cells
supplies the benchmark that entry scores against