Oncofusion

start / graph / hyp-pause-axis-asymmetry-kmt2a-r

Hypothesisspeculative confidence 1 of 5: Our own inference or analogy. Not shown to be true.Q5-interactome4

KMT2A-rearranged cells need the factors retained through elongation MORE and the factors ejected at pause release LESS — a six-gene asymmetry nobody has reported

also called pause axis asymmetry · DSIF-up NELF-down hypothesis

In an unbiased scan of every human gene, the two proteins that stay with RNA polymerase once it gets moving are needed more by KMT2A-rearranged leukaemia cells, and two of the four proteins that have to be thrown off first are needed much less. If that is real rather than noise, it means these cells run with the brake already released.

The same thing, for a scientist

Six DepMap rows from the project's own genome-wide KMT2A selectivity scan line up by mechanistic role rather than by complex membership [E1]. The two **DSIF** subunits, which are retained in both the paused and the activated elongation complex, score on the more-required side (SUPT4H1 -0.2642, rank 138; SUPT5H -0.1215, rank 767). Two of the four **NELF** subunits, which are displaced from the Pol II funnel by PAF at pause release, score at the far anti-dependency extreme (NELFB +0.2650, rank 17,671; NELFA +0.2403, rank 17,650), with NELFE +0.0583 and NELFCD -0.0208 unremarkable. **Hypothesis:** KMT2A::AFF1 and related fusions establish a constitutively pause-released state at their target genes, so the cell derives less benefit from NELF and more from the elongation-competent DSIF clamp. **Prior probability: 0.2.** The pattern is suggestive and mechanistically coherent, but four of the six genes sit near the essentiality floor where difference statistics are unreliable, the NELF subunits disagree among themselves, and there is essentially no published literature on either complex in this disease [E2][E3].

Go deeper
The data, and the three ways it could be wrong
FactorRetained through pause release?DifferenceRank / 17,728Pan-essentiality
SUPT4H1 (DSIF)yes-0.2642138-0.6692
SUPT5H (DSIF)yes-0.1215767-1.5066
NELFCDno, displaced-0.02086,425-0.6208
NELFEno, displaced+0.058315,340-0.4482
NELFAno, displaced+0.240317,650-1.0384
NELFBno, displaced+0.265017,671-1.4051

All rows from 1; the retention/displacement assignment from the cryo-EM pair 45.

Three ways this dissolves.

  1. Floor artefact. SUPT5H (-1.51), NELFB (-1.41) and NELFA (-1.04) are deeply essential everywhere. A difference computed between two near-complete killing effects is compressed and unstable. Note that the two genes with the shallowest pan-essentiality — NELFE (-0.45) and NELFCD (-0.62) — are exactly the two showing no effect, which is the pattern a floor artefact produces and is the single strongest argument against this hypothesis.
  2. Within-complex disagreement. NELF is obligate and four-subunit; half of it moves and half does not.
  3. Line composition. Eleven KMT2A-rearranged lines, AML-heavy, none infant-derived, against 43 comparators.

If any of these explains the pattern, the hypothesis is dead and the entry should be marked so.

Why it is nevertheless worth stating

Two reasons.

It is cheap to kill. A single nascent-transcription dataset in a KMT2A::AFF1 line with a matched comparator gives the pausing index at fusion target genes directly. This is not a programme of work; it is one experiment, and the graph has none.

It sits on the sharp edge of an existing controversy. Okuda and Yokoyama state the standard model — 'it is hypothesized that relieving the promoter proximal pausing of RNA polymerase II is the rate-limiting step of MLL fusion-dependent transcription' — and report the opposite, that TBP loading through SL1 is the rate-limiting activity and not elongation 2. The ECAS paper reports that its MLL-relevant elongation mechanism operates without altering pause release 3. If this hypothesis is right, the fusion's effect on the pause is real but upstream and already saturated — the cells are past the pause, which is why perturbing pause release further does nothing and why TBP loading becomes rate-limiting. That would reconcile two published results the field currently leaves as a disagreement.

If it is wrong, the more likely truth is that pausing is simply not where this disease's biology lives, which is itself a useful thing for the graph to have established rather than assumed.

Facts and figures
statementKMT2A fusion proteins establish a constitutively pause-released transcriptional state at their target genes, so KMT2A-rearranged cells derive less benefit from the NELF pausing brake (which must be displaced for elongation) and more from the DSIF clamp (which is retained into productive elongation). The six-gene dependency asymmetry in the project's genome-wide scan is a consequence of that state.
origindata-pattern
prior probability0.2
what it would reconcileIf true, it explains why Okuda and Yokoyama find pause release is not rate-limiting for MLL fusion-dependent transcription and why the ECAS mechanism acts without altering pause release: in both cases the pause is already released and is therefore no longer the controlling step.
status noteFiled as a lead. Four of six genes sit near the essentiality floor and the NELF selectivity ordering tracks pan-essentiality depth, which is the leading alternative explanation and is stated in the counter-evidence on E1.
predictions
[
  "PRO-seq or NET-seq in a KMT2A::AFF1 line will show a LOWER pausing index at direct fusion target genes than at matched expression-controlled non-target genes in the same cells.",
  "Acute degradation of the fusion (degron) will RAISE the pausing index at its target genes within hours, before secondary transcriptional changes.",
  "NELF depletion will have a smaller proliferative effect in KMT2A-rearranged lines than in matched non-rearranged leukaemia lines, in a head-to-head arm-matched comparison rather than across DepMap's heterogeneous panel.",
  "The DSIF selectivity, if real, will be larger in lines with higher fusion expression.",
  "The effect will be absent or reversed in KMT2A-rearranged lines carrying a non-SEC partner (for example a septin partner), because those fusions do not recruit elongation machinery."
]
experiments
[
  "exp-pro-seq-pausing-index-kmt2a-aff1"
]
analyses
[]
falsifiers
[
  "PRO-seq shows the pausing index at KMT2A::AFF1 target genes is equal to or HIGHER than at matched non-target genes. This kills the hypothesis outright.",
  "Permutation testing, lineage conditioning or line normalisation removes the NELFA/NELFB and SUPT4H1 signals from the genome-wide scan, leaving no pattern to explain.",
  "An arm-matched NELF depletion shows KMT2A-rearranged lines are affected equally or more than comparators.",
  "The NELF selectivity ordering is shown to track pan-essentiality depth across a large set of complexes, establishing it as a general floor artefact rather than anything about NELF.",
  "Acute fusion degradation does not change the pausing index at target genes, showing the fusion does not control the pause at all."
]
How this could bite us back
  • If this hypothesis is true it is bad news for pause-directed therapeutics in this disease: a cell already past the pause has little left to gain or lose there, and CDK9 inhibition would be acting on a step that is not limiting.
  • The hypothesis makes NELF look like a gene whose loss the leukaemia tolerates. Nothing here suggests NELF is a target, and reading it that way would invert the logic.
  • The most likely harm is opportunity cost: a PRO-seq experiment run on a pattern that turns out to be an essentiality floor artefact. The falsifiers are written so that the cheap analysis (permutation and normalisation of the existing scan) is done first.
Still open
  1. What is the pausing index at KMT2A::AFF1 target genes? Nobody has measured it, and the whole hypothesis stands or falls on that one number.
  2. Does the NELF selectivity ordering track pan-essentiality depth across all complexes in the scan, not just NELF? That analysis would settle the floor-artefact question for the entire beat, not only this entry.
  3. Does a septin-partner KMT2A fusion, which recruits no elongation machinery, show the opposite asymmetry? That is the cleanest specificity control available and uses fusions this graph already carries.
  4. If the cells really are past the pause, does that predict resistance to CDK9 inhibitors, or heightened sensitivity?
Every source behind this entry (5)
  1. E1moderate confidence 3 of 5: One strong study, or several smaller consistent ones.database

    In the project's genome-wide KMT2A-rearranged selectivity scan of 17,728 genes across 1,208 lines (11 KMT2A-rearranged versus 43 other leukaemia lines), SUPT4H1 scores -0.2642 (rank 138, pan-essentiality -0.6692), SUPT5H -0.1215 (rank 767, -1.5066), NELFCD -0.0208 (rank 6,425, -0.6208), NELFE +0.0583 (rank 15,340, -0.4482), NELFA +0.2403 (rank 17,650, -1.0384) and NELFB +0.2650 (rank 17,671, -1.4051).

    Genome-wide KMT2A-rearranged selectivity scan, 17,728 genes (DepMap Chronos via Open Targets) · Broad Institute DepMap / Open Targets Platform; analyses/depmap-kmt2a-selective-scan-full 2026
    What this evidence cannot show

    All six rows read out of analyses/depmap-kmt2a-selective-scan-full/results.json on 2026-09-19 in one pass; none typed from recall. Not permutation-tested for these genes, not lineage-conditioned, not line-normalised. Four of the six sit deep enough for floor compression. The grouping into 'retained' and 'displaced' is imposed by this entry after seeing the numbers, which is exactly the circumstance in which a pattern is easiest to over-read.

    Evidence that argues against this (1)
    • The ordering within NELF tracks pan-essentiality almost perfectly: NELFB (-1.41) > NELFA (-1.04) > NELFCD (-0.62) > NELFE (-0.45) by depth, and the selectivity differences follow the same order. That is what a floor artefact looks like, and it is a better fit to the NELF data than the mechanistic story this hypothesis proposes. Vos SM et al. 2018
  2. E2moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    Okuda and Yokoyama state that 'it is hypothesized that relieving the promoter proximal pausing of RNA polymerase II is the rate-limiting step of MLL fusion-dependent transcription' and report the contrary: the key activity responsible for the oncogenic property of MLL-AEP fusion proteins is TBP loading through the SL1 complex, and not mediator recruitment or transcriptional elongation activities.

    TBP loading by AF4 through SL1 is the major rate-limiting step in MLL fusion-dependent transcription · Okuda H, Takahashi S, Takaori-Kondo A, Yokoyama A · Cell Cycle 15(20):2712-22 2016
    PMID 27564129DOI
    What this evidence cannot show

    Abstract fetched and read on 2026-09-19; the quoted sentence is verbatim from it. Full text not read. Structure-function dissection of MLL-AEP fusion constructs in murine myeloid transformation; one laboratory, one framework, and 'rate-limiting for transformation by engineered constructs' is not the same as 'rate-limiting at a native target gene in a patient blast'.

    Evidence that argues against this (1)
    • The pause-release model retains real support. ENL, an MLL fusion partner, initiates multivalent phase separation of SEC that compartmentalises P-TEFb and controls transcriptional pause release, and MLL-ENL fusion is reported to boost SEC phase separation. If the fusion increases pause-release capacity, this hypothesis's premise that pausing is already saturated could be right for the wrong reason - or simply wrong. Guo C et al. 2020
  3. E3moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    Acute perturbation of AQR/XAB2-dependent functions of the ECAS complex - an AFF4-ELL2-EAF1 elongation module joined to a U2 snRNP-associated intron binding complex, occupying promoter-proximal chromatin and distinct from P-TEFb-containing SEC - impairs productive gene-body elongation WITHOUT altering pause release, and Aqr perturbation impairs leukaemia initiation, maintenance and leukaemia stem cell function in MLL-rearranged leukaemia.

    A dual-module ECAS complex couples transcription elongation and RNA processing · Xiang Y, Wang H, Yin Z, Xiao R, Song Z, Lu R, Dai R, Ye J, Li C, Cui Y, Wang M, Lv J, Li Q, Wang F, Li Z, Zhang K, Zhang Y, Wu D, Fang P, Liang K · Molecular Cell, online ahead of print 16 September 2026 2026
    PMID 42748909DOI
    What this evidence cannot show

    Abstract fetched and read on 2026-09-19; full text not read. Very recent, not independently replicated. 'Without altering pause release' is reported for AQR perturbation in human cells, and the leukaemia phenotype is a separate mouse experiment; the two are not the same measurement in the same system.

    Evidence that argues against this (1)
  4. E4established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    DSIF is present in both the paused complex (Pol II-DSIF-NELF) and the activated complex (Pol II-DSIF-PAF-SPT6), whereas PAF displaces NELF from the Pol II funnel for pause release; SPT6 opens the DSIF RNA clamp after P-TEFb phosphorylates the Pol II CTD linker.

    Structure of activated transcription complex Pol II-DSIF-PAF-SPT6 · Vos SM, Farnung L, Boehning M, Wigge C, Linden A, Urlaub H, Cramer P · Nature 560(7720):607-612 2018
    PMID 30135578DOI
    What this evidence cannot show

    Abstract fetched and read on 2026-09-19; full text not read. Reconstituted, chromatin-free, pig Pol II. The retained-versus-displaced distinction is drawn from comparing two static structures.

    Evidence that argues against this (1)
    • The retained/displaced distinction is about a single round of transcription in vitro. A cell's requirement for a factor over days of proliferation depends on turnover, redundancy and total flux, not on whether the factor is present in one structural state, so mapping a CRISPR gene effect onto a structural transition is a long inferential leap. Vos SM et al. 2018
  5. E5moderate confidence 3 of 5: One strong study, or several smaller consistent ones.database

    As of 2026-09-19 there is essentially no published literature on either DSIF or NELF in KMT2A-rearranged leukaemia: '(MLL OR KMT2A) AND (SUPT5H OR SPT5 OR SPT5H OR DSIF)' returns 1 record, '(MLL OR KMT2A) AND (SUPT4H1 OR SPT4 OR SUPT4H)' returns 0, and '(MLL OR KMT2A) AND NELF' returns 1, a CDK9 review.

    What this evidence cannot show

    Counted-absence claims are floors. The alias sets are documented in gene-supt5h, gene-supt4h1 and protein-nelf-complex; one NELFE alias ('RD') had to be discarded for inflating a count fifty-seven-fold.

    Evidence that argues against this (1)
Connected to