Oncofusion

start / graph / ana-q141-menin-loses-half-its-promoter-signal-where-the-fusion-sat-and-nothing-where-it-did-not

Analysisstrong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.Q141-menin-at-fusion-targets

Menin loses half its promoter signal where the fusion sat, and nothing where it did not

When the cancer fusion protein is destroyed, menin falls away from exactly the genes the fusion was sitting on - about half of it goes in three hours - while menin at every other gene does not budge. The earlier finding that menin 'stays put genome-wide' was counting mostly places the fusion had never been.

The same thing, for a scientist

Partitioning the deposited per-gene TSS signal (TSS -1 kb/+3 kb, housekeeping-normalised) on HA occupancy resolves the apparent contradiction in this corpus between a genome-wide 80% MENIN peak-call retention and the authors' reported >50% MENIN loss at target genes. Median MEN1 dVHL/DMSO ratio 3 h after 500 nM dTAG-VHL is **0.503** over the authors' 200 published MLL::AF9 target genes (bootstrap 95% over genes 0.480-0.528, n=200) and **0.974** over the bottom half of the HA-occupancy ranking (0.960-0.994, n=717 of 8,672 genes surviving the authors' signal floor). The gap is 47.1 percentage points. The ladder is monotone in HA occupancy: top 1% 0.473, top 5% 0.747, top 10% 0.853, top 20% 0.909, bottom 50% 0.974 - which is the dilution mechanism made visible. The 19 'highly sensitive' target genes give 0.500, indistinguishable from the full target list.

Go deeper
What was measured, exactly

The quantity is the one the authors plot in Figure 4B: 'fold change of total reads ... at promoters (TSS -1kB/+3kB)', 3 h after 500 nM dTAG-VHL, taken from the deposited per-gene table GSE173599_TSS_gene_body_signal_Figure3.xlsx, sheet Menin TSS 3h (6,036 genes) 2. The stratification variable is per-gene HA (fusion) DMSO signal over input from GSE173599_TSS_signal_HA_ChIPs.xlsx, sheet ChIPseq_1909 (19,168 genes) 3. The occupied stratum is also defined independently by the authors' own published gene list 4, and the two definitions agree: their 202-gene list maps to 200 genes at median rank 114.5 of 17,343 in my HA ranking, the 0.7th percentile.

The number reproduces the authors' own sentence

The authors write: 'After 3 hours of dTAG-VHL treatment, more than 50% of MENIN and almost 70% of DOT1L was lost at MLL::AF9 target genes, while there were no changes in occupancy at non-MLL::AF9-bound genes' 1. My re-derivation gives 0.503 for MENIN (49.7% lost) and 0.401 for DOT1L (59.9% lost) over the full target list, rising to 0.330 for DOT1L over the 19 highly sensitive genes (67.0% lost, which is the 'almost 70%'). So their MENIN figure is exact and their DOT1L figure describes the highly-sensitive subset rather than the whole list.

Why this does not simply overturn the 80% figure

Both figures are correct on the same files. The corpus's 80.0% is a binary peak-call census; this is a quantitative signal ratio. A site that loses half its signal and is still called a peak contributes to both. See concept-q141-a-binary-peak-call-cannot-see-a-two-fold-change.

Scale and limits

No spike-in exists anywhere in GSE173599, so every number is a within-sample share, and the treated arms are scaled by a 21-gene housekeeping panel whose own raw 3 h H3K27ac ratio averages 0.809 and is forced to 1. A housekeeping normalisation DEFINES the global change to be zero, so only the DIFFERENCE between strata is measurable; a uniform genome-wide loss of any of these proteins would be invisible here. n = 1 per arm for every channel except HA, which has two independent 3 h batches. The cells are HCB1, transformed by KMT2A::MLLT3 (MLL::AF9) - NOT the index fusion of this project, KMT2A::AFF1 - and a prior beat in this corpus found that per-gene ordering transfers between MLLT3 and AFF1 lines while magnitude does not.

Facts and figures
scriptanalyses/menin-retention-by-ha-occupancy/analyze_tss.py
statusrun
preregistrationanalyses/menin-retention-by-ha-occupancy/PREREGISTRATION.md
result summaryMEN1 dVHL/DMSO TSS ratio 3 h: 0.503 at the authors' 200 MLL::AF9 target genes (95% bootstrap over genes 0.480-0.528) vs 0.974 at the bottom 50% of HA occupancy (0.960-0.994). Gap 47.1 points, monotone across the HA ladder (top1% 0.473, top5% 0.747, top10% 0.853, top20% 0.909). Pre-registered outcome A on the primary instrument.
peak width ratio noteNot applicable to this instrument - this is a signal ratio, not a peak call. The companion peak measure for MEN1 has a treated/control median peak width ratio of 0.93, below the pre-registered 1.3 re-score trigger.
inputs
[
  "GSE173599_TSS_gene_body_signal_Figure3.xlsx",
  "GSE173599_TSS_signal_HA_ChIPs.xlsx",
  "GSE173599_ChIPseq_TSS_Figure3_revisions.xlsx"
]
How this could bite us back
  • A menin inhibitor is dosed and monitored as if menin occupancy were the thing being corrected. Half the menin at fusion promoters here does not depend on the fusion at all, so a drug that removes only the fusion-dependent pool leaves a reservoir that can re-assemble.
  • If the residual pool is LEDGF-tethered, LEDGF is dispensable for steady-state haematopoiesis but essential for MLL leukaemogenesis, which makes it a target - and also makes menin-inhibitor monotherapy a partial intervention by construction.
  • Any pharmacodynamic assay that reads menin occupancy genome-wide will report a drug as inactive when it is working, because the fusion-occupied fraction of the genome is under 2% of measured genes.
Still open
  1. Is the residual 50% of menin at fusion-occupied promoters the LEDGF/PSIP1-tethered pool, the wild-type MLL1-tethered pool, or a slow off-rate of the same pool?
  2. Does the same partition hold in a KMT2A::AFF1 line, where the menin-binding motifs are identical but the elongation partner is not?
  3. Does menin signal at fusion targets keep falling between 3 h and 24 h, or does it plateau at ~50%? No MENIN ChIP exists in this deposit between 3 h and 4 d.
Every source behind this entry (4)
  1. E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    The authors report more than 50% MENIN loss and almost 70% DOT1L loss at MLL::AF9 target genes 3 h after dTAG-VHL, with no change at non-bound genes.

    MLL::AF9 degradation induces rapid changes in transcriptional elongation and subsequent loss of an active chromatin landscape · Olsen SN, Godfrey L, Healy JP, Choi YA, Kai Y, Hatton C, Perner F, Haarer EL, Nabet B, Yuan GC, Armstrong SA · Molecular Cell 82(6):1140-1155.e11 2022
    PMID 35245435DOI
    What this evidence cannot show

    ChIP-seq, n=1 per arm, no spike-in, KMT2A::MLLT3 in HCB1 cord-blood cells. The kinetic claim beside it (Figure 4A) is ChIP-qPCR at selected loci, a different assay from the genome-wide panel in Figure 4B.

    Evidence that argues against this (1)
    • The same deposit, read as a genome-wide peak-call census rather than a target-gene signal ratio, says menin retains 80% of its strongest calls - which reads as the opposite conclusion and is not. Olsen SN et al. 2022
  2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.dataset

    Recomputed from the deposited per-gene table, median MEN1 dVHL/DMSO TSS ratio is 0.503 at the authors' 200 target genes and 0.974 at the bottom half of the HA ranking.

    What this evidence cannot show

    Housekeeping-normalised, so this is a relative not an absolute quantity. The sheet is pre-filtered by the authors (2-fold over input AND >200 reads in either arm): 100% of the occupied stratum survives that filter but only 8.3% of the non-occupied stratum does, so the comparison population for 'no change elsewhere' is menin sites without a fusion, not the rest of the genome. n=1 per arm.

    Evidence that argues against this (1)
    • Menin is delivered to MLL target genes by LEDGF/PSIP1 as well as by MLL itself, so a partial menin residue after fusion loss is expected under the published model and is not by itself evidence against a fusion-dependent tether. Yokoyama A et al. 2008
  3. E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.dataset

    Per-gene HA (fusion) DMSO signal from the deposited HA table provides the stratification variable over the full 19,168-gene universe.

    What this evidence cannot show

    HA dVHL/DMSO columns in this workbook are RAW, not housekeeping-normalised, unlike every other channel; the two are therefore not on the same scale and are never compared directly here.

    Evidence that argues against this (1)
    • HA DMSO signal over input measures where the fusion is at steady state, not where it acts; a fusion that acts transiently at a promoter would be under-weighted by this stratification. Olsen SN et al. 2022
  4. E4strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.dataset

    The authors' own published MLL::AF9 target-gene lists (20 highly sensitive, 182 other) are deposited and were used as an independent definition of the occupied stratum.

    What this evidence cannot show

    The lists are the authors' construction, derived from their own PRO-seq and ChIP data in the same cells; using them is not an independent replication, only an independent definition.

    Evidence that argues against this (1)
    • The 'highly sensitive' subset was defined by transcriptional response to fusion degradation, so using it as an occupancy stratum imports a response-based selection into what is presented as an occupancy-based partition. Olsen SN et al. 2022
Connected to