Oncofusion

start / graph / mech-transcription-replication-conflict-kmt2a-r

Mechanismspeculative confidence 1 of 5: Our own inference or analogy. Not shown to be true.Q-fa-replication

Transcription–replication conflict as the candidate reason KMT2A-rearranged cells need the Fanconi pathway — a chain with two measured links and three inferred ones

also called TRC · transcription-replication conflict · R-loops in MLL-rearranged leukaemia · replication stress hypothesis

Cells copy their DNA and read it at the same time, and where those two machines meet on the same stretch of DNA they collide. The KMT2A fusion makes cells read certain very long genes much harder than normal, so the idea is that collisions happen more often there, and the Fanconi repair crew is what clears them up — which would explain why these cells need that crew. Most of this chain has not actually been measured in KMT2A leukaemia cells.

The same thing, for a scientist

The proposed chain is: (1) the KMT2A fusion recruits the super elongation complex and drives sustained, processive elongation over long domains; (2) long, highly transcribed genes are where replication forks meet transcription complexes head-on or co-directionally; (3) conflicts generate R-loops, fork stalling and under-replicated DNA; (4) the FA pathway is what protects and restarts those forks and completes replication at such loci; therefore (5) KMT2A-rearranged cells carry a constitutive replication-stress load and need the FA pathway more than other leukaemia cells do. **Links 2 and 4 are measured, in other systems.** FANCD2 is loaded at common fragile sites and accumulates at the central region of large fragile genes under replication stress [E3][E4]; UBE2T resolves transcription-replication conflicts and R-loops in large genes and promotes mitotic DNA synthesis [E5]; SLX4 with RTEL1 prevents transcription-mediated replication perturbation [E6]. **Link 1 is measured in KMT2A-rearranged cells** [E1], and R-loops are directly mapped by DRIP-seq in MLLr AML — but as a *productive* architectural feature stabilising CTCF boundaries and TAD integrity, not as a lesion [E2]. **Link 3 and link 5 are inferred and have never been measured in a KMT2A-rearranged cell.** The one piece of functional support for a replication-stress liability in this disease is that ATR inhibition induces chromosomal breakage and kills MLL-ENL AML cells p53-independently, with single-agent activity in allograft and xenograft models [E7] — a phenotype, not a mechanism.

Go deeper
The chain, link by link, with the status of each
linkclaimstatus in KMT2A-rearranged cells
1the fusion drives sustained elongation over long domainsmeasured — MLL-AF4 with PAF1 and FACT drives high-density enhancer interactions and elongation at leukaemic target loci 1
2long, highly transcribed loci are where forks meet transcriptionmeasured, but in fibroblasts and untransformed lines, not leukaemia 34
3that produces R-loops and fork stalling hereR-loops mapped by DRIP-seq in MLLr AML, but as productive architecture, not damage 2
4the FA pathway protects and restarts those forksmeasured, in fibroblasts, cancer lines and mouse germ cells 3456
5so KMT2A-rearranged cells carry a higher replication-stress loadnever measured. No published DNA-fibre, combing, fork-speed or origin-firing measurement exists for KMT2A-rearranged cells

The last row is the important one and it is the reason exp-replication-stress-readout-kmt2a-r exists. A PubMed title/abstract search for KMT2A-rearranged, MLL-rearranged, MLL-AF4 or MLL-AF9 against DNA fibre, DNA combing, fork speed, fork progression or origin firing returns one record, and it is an AML lncRNA paper that is not about KMT2A rearrangement 8. The chain is being asserted from its ends.

What the R-loop evidence in MLL-rearranged AML actually says, which is not what the hypothesis wants

This is where a careless reading would go wrong. R-loops have been mapped genome-wide in MLL-rearranged AML cells, by DRIP-seq, in a 2025 study of the METTL3–YTHDC1 axis 2. The finding is that METTL3 is transcriptionally activated by MLL, forms a complex with YTHDC1 and CTCF, and that m6A modification of architectural RNAs stabilises R-loops which are required to maintain CTCF boundary activity and TAD integrity. An earlier study found the same shape for the HOXA-locus lncRNA HOTTIP: HOTTIP-mediated R-loop formation reinforces the CTCF boundary and drives oncogene transcription, and eliminating those R-loops with targeted RNase H inhibits leukaemogenesis 9.

So in MLL-rearranged leukaemia, the documented R-loops are load-bearing. The cells are not merely tolerating them; they are using them. That is compatible with a higher steady-state R-loop burden requiring more FA-pathway maintenance — a cell that needs its R-loops still has to replicate through them — but it is emphatically not the same claim as 'R-loops are damage here', and anyone citing HOTTIP or METTL3 as evidence for a replication-stress lesion has read them backwards.

The two claims can be separated experimentally: if R-loops here are productive rather than pathological, RNase H1 overexpression should reduce proliferation of KMT2A-rearranged cells and should not rescue the viability loss caused by FA-gene knockout.

The strongest functional support, and what it is support for

ATR is the kinase that senses replication stress. Inhibiting it induces chromosomal breakage and kills mouse AML cells carrying an MLL-ENL fusion with constitutively active N-RAS, independently of p53 — which matters, because the attenuated p53 response is why these leukaemias resist genotoxic therapy. ATR inhibition had single-agent activity both in an immunocompetent allograft and against a human AML-MLL xenograft, and ATM inhibition also prolonged survival 7.

That is real evidence that MLL-driven AML carries an exploitable replication-stress liability. It is not evidence about where the stress comes from. The MLL-ENL model also carries oncogenic N-RAS, which is itself a classical driver of replication stress through unscheduled origin firing, and the paper's framing is AML generally with MLL as the hardest subtype rather than a mechanism specific to the fusion.

The other direction: replication stress in this disease has been studied, but for breakage, not dependency

There is a replication-stress literature at the KMT2A locus, and it is worth knowing that it is about something else. Replication stress destabilises the MLL breakpoint cluster region; endonuclease G initiates rearrangements there; base-excision-repair proteins couple AID and endonuclease G during replication-stress-induced MLL destabilisation 1011. All of that concerns how the translocation is made, in a cell that does not yet have a fusion. It says nothing about what a cell that already carries the fusion depends on, and the two questions are easy to conflate because the phrase 'replication stress in MLL rearrangements' covers both. This graph already holds that mechanism separately as mechanism-replication-stress-endog-cleavage-kmt2a.

What would kill this hypothesis
  • DNA fibre analysis shows no difference in fork speed, fork asymmetry or origin firing between KMT2A-rearranged and comparator leukaemia lines. The mechanism would then have no substrate.
  • The FA dependency does not track transcription. If FANCD2 foci or FA-gene dependency do not concentrate at the long, fusion-driven target loci, the conflict is not where the fusion is.
  • RNase H1 overexpression does not rescue FA-knockout lethality in KMT2A-rearranged cells. R-loops would then not be the intermediate, whatever else is.
  • ALDH2 methylation status predicts the dependency better than KMT2A status does 12. Then the mechanism is endogenous aldehyde genotoxicity and this entry is about the wrong thing entirely. This is the rival with actual experimental support in leukaemia and it is kept as its own entry, mech-aldehyde-fa-dependency-rival.
  • The dependency is on something other than the FA pathway. The obligate core subunits do not move together in DepMap, which is a problem for any mechanism that runs through the canonical pathway — see mech-fa-core-complex-heterogeneity-kmt2a.
Facts and figures
decisive experimentexp-replication-stress-readout-kmt2a-r
chain
fusion drives sustained e…1.000fusion drives sustained elongation over long domains: 1long, highly transcribed …2.000long, highly transcribed loci are conflict-prone: 2conflicts produce R-loops…3.000conflicts produce R-loops and fork stalling here: 3FA pathway protects and r…4.000FA pathway protects and restarts stalled forks: 4KMT2A-rearranged cells th…5.000KMT2A-rearranged cells therefore carry higher replication stress: 5
chain — the numbers behind the chart
[
  {
    "step": 1,
    "claim": "fusion drives sustained elongation over long domains",
    "status": "measured in KMT2A-rearranged cells",
    "evidence": "E1"
  },
  {
    "step": 2,
    "claim": "long, highly transcribed loci are conflict-prone",
    "status": "measured in other systems",
    "evidence": "E3, E4"
  },
  {
    "step": 3,
    "claim": "conflicts produce R-loops and fork stalling here",
    "status": "R-loops measured but as productive architecture",
    "evidence": "E2, E9"
  },
  {
    "step": 4,
    "claim": "FA pathway protects and restarts stalled forks",
    "status": "measured in other systems",
    "evidence": "E3, E4, E5, E6"
  },
  {
    "step": 5,
    "claim": "KMT2A-rearranged cells therefore carry higher replication stress",
    "status": "never measured",
    "evidence": "E8"
  }
]
rival mechanisms
[
  "mech-aldehyde-fa-dependency-rival"
]
How this could bite us back
  • If the FA dependency is a consequence of transcription-replication conflict, anything that increases elongation — including some proposed therapeutic manoeuvres — would deepen it, and anything that shuts the fusion programme down, including menin inhibition, would relieve it. A menin inhibitor and an FA-pathway inhibitor could therefore antagonise each other.
  • Inhibiting fork protection in a cell already carrying high replication stress raises mutation rate in the survivors, which is the standard route to resistance.
  • The mechanism predicts the dependency is proliferation-coupled, so it would be weakest in exactly the quiescent leukaemia-initiating cells that cause relapse.
Still open
  1. Do the loci a KMT2A fusion drives overlap the large fragile genes where FANCD2 accumulates? This is computable today from published MLL-AF4 ChIP-seq and published FANCD2 ChIP-seq, without any new experiment.
  2. Is the R-loop burden in KMT2A-rearranged cells higher than in comparator leukaemia lines, or merely differently placed?
  3. Does fusion knockdown reverse whatever replication phenotype exists, which would tie the stress to the fusion rather than to the lines?
  4. Why is FANCD2 — the substrate through which the core complex is supposed to act — among the weakest members of the set?
Every source behind this entry (12)
  1. E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    MLL-AF4 cooperates with PAF1 and FACT to drive high-density enhancer interactions in leukaemia, establishing that the fusion sustains an elongation-competent transcriptional state at its target loci.

    PMID 37626123DOI
    What this evidence cannot show

    Establishes elongation and enhancer architecture. It does not measure gene length preference, replication, or any conflict between the two.

    Evidence that argues against this (1)
    • Elongation being sustained does not entail collisions. Cells have well-characterised machinery for co-orienting replication and transcription and for scheduling highly transcribed genes early; a transcriptionally active locus is not automatically a conflict-prone one. Takedachi A et al. 2020
  2. E2moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    R-loops have been mapped by DRIP-seq in MLL-rearranged AML cells, where METTL3-mediated m6A modification of architectural RNAs stabilises them and they are required to maintain CTCF boundary activity and TAD integrity.

    PMID 41430607DOI
    What this evidence cannot show

    Recent single study. The R-loops are characterised at CTCF boundaries and promoters/enhancers, and no replication measurement is reported.

    Evidence that argues against this (1)
    • This is evidence AGAINST the naive form of the hypothesis. The R-loops described here are functional architecture the cells require, not lesions. A model in which R-loops are the damage that the FA pathway repairs has to explain why removing them is deleterious. Luo H et al. 2022
  3. E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    FANCD2 facilitates replication through common fragile sites, the late-replicating loci where forks characteristically stall.

    FANCD2 Facilitates Replication through Common Fragile Sites. · Madireddy A, Kosiyatrakul ST, et al. · Mol Cell 2016
    PMID 27768874DOI
    What this evidence cannot show

    Human fibroblast and lymphoblast systems with induced replication stress (aphidicolin), not leukaemia and not a fusion oncoprotein.

    Evidence that argues against this (1)
  4. E4strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    Under replication stress FANCD2 accumulates specifically at the central region of large fragile genes, tying the FA pathway to gene length rather than to cross-link damage.

    PMID 29394375DOI
    What this evidence cannot show

    Aphidicolin-induced stress in human cell lines. 'Large fragile genes' is not the same set as 'genes a KMT2A fusion drives', and the overlap has not been computed here.

    Evidence that argues against this (1)
    • The gene-length link is established for FANCD2, the substrate. It is not established for the core complex subunits that actually carry the DepMap signal, and the strongest of those, FANCF, has no independent replication-stress literature of its own. Okamoto Y et al. 2019
  5. E5strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    UBE2T — the FA pathway E2, and one of the eight genes carrying this signal — resolves transcription–replication conflicts, resolves R-loops in large genes, stabilises replication forks and promotes mitotic DNA synthesis.

    PMID 36928776DOI
    What this evidence cannot show

    Mouse primordial germ cells. A developmental proliferation defect in an embryonic germline is a long way from a leukaemia dependency, and PGCs are an unusually replication-stressed cell type to begin with.

    Evidence that argues against this (1)
    • The same protein's only reported dependency in human acute leukaemia is attributed to a different mechanism entirely — synthetic lethality with ALDH2 silencing and endogenous aldehyde damage. Yang Z et al. 2021
  6. E6moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    SLX4 acts with RTEL1 to prevent transcription-mediated perturbation of DNA replication, and SLX4 is itself one of the better-ranking FA-annotated genes in this sweep (rank 82 of 8,269 by raw effect, line-permutation p = 0.0086).

    SLX4 interacts with RTEL1 to prevent transcription-mediated DNA replication perturbations. · Takedachi A, Despras E, et al. · Nat Struct Mol Biol 2020
    PMID 32398829DOI
    What this evidence cannot show

    Human non-leukaemia systems. SLX4 has many partners and functions beyond the FA pathway, including at telomeres, so its rank is weak support for an FA-specific reading.

    Evidence that argues against this (1)
  7. E7strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    ATR inhibition induces chromosomal breakage and kills MLL-ENL-driven mouse AML cells independently of p53, with single-agent antitumour activity in an immunocompetent allograft and in a human MLL-rearranged AML xenograft; ATM inhibition also prolongs survival.

    PMID 27625305DOI
    What this evidence cannot show

    The model carries MLL-ENL plus constitutively active N-RAS, and oncogenic RAS is itself a canonical source of replication stress, so the fusion's contribution is not isolated. The paper frames the result as AML generally, MLL-driven leukaemias especially.

    Evidence that argues against this (1)
    • ATR is required by essentially every rapidly proliferating cell, and ATR inhibitors have broad single-agent activity across cancers. Sensitivity to an ATR inhibitor is weak evidence that a particular subtype carries unusual replication stress. Gole B et al. 2015
  8. E8strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

    No published study measures replication dynamics in KMT2A-rearranged cells: a PubMed title/abstract search for MLL-AF4, MLL-AF9, MLL-rearranged or KMT2A-rearranged against DNA fibre, DNA combing, fork speed, fork progression or origin firing returns one record, which on title-checking is an AML lncRNA study not about KMT2A rearrangement.

    What this evidence cannot show

    Title/abstract searching misses methods reported only in figures or supplements, and fibre assays are often described as 'DNA fibre analysis' inside a paper about something else. The absence is of a headline finding, not necessarily of the measurement.

    Evidence that argues against this (1)
    • A related search for KMT2A/MLL-rearranged against R-loops does return substantive work, so the field has the tools and has applied them to this disease — just not to replication. Fu R et al. 2025
  9. E9strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    HOTTIP-dependent R-loop formation at the HOXA locus reinforces CTCF boundary activity and TAD integrity in MLL-rearranged and NPM1-mutant AML, and targeting RNase H to eliminate those R-loops mitigates leukaemogenesis in xenografts.

    PMID 35180428DOI
    What this evidence cannot show

    HOTTIP overexpression is shared with NPM1-mutant AML, so this is a HOXA-programme finding rather than a KMT2A-rearrangement finding.

    Evidence that argues against this (1)
    • Directly against the naive replication-stress reading: removing R-loops here is therapeutic, not protective. If R-loops were the damage the FA pathway is cleaning up, RNase H targeting should have relieved a burden rather than mitigated leukaemogenesis. Okamoto Y et al. 2019
  10. E10strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    Replication stress destabilises the MLL breakpoint cluster region and endonuclease G initiates DNA rearrangements there — a replication-stress mechanism in this disease that concerns how the translocation is created, not what the rearranged cell depends on.

    PMID 25132265DOI
    What this evidence cannot show

    Concerns pre-leukaemic cells without a fusion. Included here specifically to keep the two 'replication stress in MLL' literatures apart.

    Evidence that argues against this (1)
    • Being fragile under replication stress before the translocation says nothing about being dependent on fork protection after it, and conflating the two is the commonest way this hypothesis gets overstated. Gole B et al. 2018
  11. E11moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    Base excision repair proteins couple AID and endonuclease G during replication-stress-induced MLL destabilisation, extending the breakage mechanism rather than the dependency one.

    PMID 28626219DOI
    What this evidence cannot show

    Again pre-translocation biology, in cell-line models of induced replication stress.

    Evidence that argues against this (1)
    • None found that disputes the breakage mechanism; the caveat is one of relevance rather than of correctness. Gole B et al. 2015
  12. E12strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    The rival mechanism with actual experimental support in acute leukaemia is endogenous aldehyde genotoxicity: UBE2T and FANCL are selective AML dependencies because ALDH2 is silenced by DNA hypermethylation, leaving the FA pathway as the only remaining defence.

    PMID 33893150DOI
    What this evidence cannot show

    AML against other cancer lineages, not KMT2A-rearranged against other leukaemia; ALDH2 methylation status has not been checked in the eleven lines carrying this project's signal.

    Evidence that argues against this (1)
Connected to