Replication stress, AID, base-excision repair and endonuclease G cut the KMT2A locus
also called EndoG MLLbcr cleavage · replication stress MLL rearrangement · AID-BER-EndoG axis
When a cell's DNA-copying machinery stalls, a normally mitochondrial nuclease called endonuclease G moves into the nucleus and cuts the KMT2A gene at a specific hotspot. Blocking one of the repair enzymes upstream of it (APE1) protected human cord-blood stem cells from that cut, which makes this the most concrete preventive target in the whole fusion-genesis story.
The same thing, for a scientist
Wiesmuller and colleagues defined a replication-stress-triggered, TOP2-independent route to KMT2A breakage. The ~0.4 kb hotspot at the KMT2A intron 11 / exon 12 boundary, placed into an EGFP-based recombination reporter, enhanced both spontaneous and genotoxin-induced recombination in multiple human cell types. Endonuclease G (EndoG), an apoptotic nuclease, is essential for this MLLbcr-specific recombination after replication stress: replication stress drives nuclear accumulation of EndoG, DNA binding, and EndoG-dependent chromosomal cleavage at the MLLbcr, with additional dependence on ATM signalling to the H2B monoubiquitinase RNF20. A follow-up siRNA screen identified ten further factors and connected activation-induced cytidine deaminase (AID) to EndoG via base-excision repair: replication-stress-induced EndoG recruitment and cleavage are AID/BER dependent, and inhibition of APE1 protected both tumour cells and human cord-blood-derived haematopoietic stem/progenitor cells from MLLbcr cleavage. This is the strongest candidate mechanism for de novo, non-chemotherapy KMT2A breakage.
The hotspot is intrinsically recombinogenic
The MLLbcr hotspot is not just where breaks are found after the fact. Inserted into a reporter construct in a neutral genomic context, the 0.4 kb intron 11 / exon 12 element increased both spontaneous and genotoxin-induced recombination in several human cell types 1. That is a sequence-and-structure-intrinsic property, transferable out of its native locus.
The cascade: replication stress to AID to BER to EndoG1 deeper
Replication stress (modelled with aphidicolin) drives nuclear accumulation of EndoG and EndoG-dependent cleavage of the chromosomal MLLbcr; this also requires ATM signalling to RNF20, the H2B monoubiquitinase that relaxes chromatin 1. The upstream trigger was then traced to off-target binding of AID at the MLLbcr, which initiates base-excision-repair-mediated single-strand cleavage; derailed EndoG activity downstream converts this into the rearrangement-competent lesion 2. Note that this brings AID, a programmed mutator of the adaptive immune system, into the mechanism of a leukaemia that arises in a pro-B cell (see concept-programmed-dna-breaks-vdj-and-csr).
APE1 inhibition protects human cord-blood HSPC
Inhibition of the core BER factor APE1 (APEX1) protected against MLLbcr cleavage not only in tumour cells but in human cord-blood-derived haematopoietic stem/progenitor cells, which are the closest accessible surrogate for the cell of origin 2. This is the single most actionable finding in this beat: a pharmacologically tractable node whose inhibition demonstrably reduces cleavage at the exact locus in the right cell type. It is also the basis of hyp-block-kmt2a-break-machinery-upstream.
Why this fits de novo infant disease better than TOP2 poisons do
Three reasons. (1) It is TOP2-independent, so it does not require an exogenous topoisomerase poison. (2) Its hotspot is the intron 11 / exon 12 boundary. CORRECTED 2026-09-19. This clause previously read "in the region where de novo rather than therapy-related breakpoints concentrate [E3]", and that is withdrawn. The intron 11 / exon 12 boundary lies in the telomeric half of the BCR, and Scharf et al. report that infant and therapy-related patients alike cluster preferentially in the telomeric portion (hot spot SCII) — infant disease being de novo. So the hotspot's position does not distinguish de novo from therapy-related breakage, and the de-novo-equals-centromeric generalisation drawn from the small 1996 pre-NGS cohort in 3 does not hold for the de novo subgroup this graph is about. See the counter-evidence on 3. (3) Replication stress is a normal feature of the rapidly expanding fetal haematopoietic compartment, so the trigger is endogenous. None of this is proof that it happens in a human fetus, and it has not been demonstrated there.
| mechanism class | replication-stress-triggered nucleolytic cleavage |
|---|---|
| hotspot | KMT2A intron 11 / exon 12 boundary, ~0.4 kb |
| top2 dependence | independent |
| druggable node | APEX1 (APE1) inhibition protected human cord-blood HSPC from MLLbcr cleavage |
effectors
[ "ENDOG", "AICDA (AID)", "APEX1 (APE1)", "ATM", "RNF20" ]
aetiology fit
{
"de_novo_infant_kmt2a_aff1": "moderate",
"therapy_related_kmt2a_r": "moderate"
}- APE1 is the central abasic-site endonuclease of base-excision repair and is essential; chronic inhibition would be mutagenic and myelosuppressive. Any preventive use is implausible and it is better treated as a mechanistic probe.
- EndoG loss-of-function would impair apoptotic DNA degradation and mitochondrial DNA turnover, with unknown developmental consequences.
- If the same AID/BER machinery also generates the secondary lesions in other childhood ALL subtypes, blocking it could shift rather than remove leukaemia risk.
- Is AID expressed at all in the human fetal liver HSC/LMPP that gives rise to KMT2A::AFF1 infant ALL? If not, the AID arm of this model cannot apply to the de novo disease and the mechanism needs a different initiator.
- Does APE1 inhibition reduce actual KMT2A::AFF1 fusion formation, not just cleavage, in a humanised fetal progenitor system?
- Has anyone replicated the EndoG dependence outside the Wiesmuller laboratory? I could not find an independent replication.
- E1moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper
The MLLbcr hotspot enhances spontaneous and genotoxin-induced recombination, and replication stress causes EndoG-dependent cleavage of the chromosomal MLLbcr requiring ATM-RNF20 signalling.
Endonuclease G initiates DNA rearrangements at the MLL breakpoint cluster upon replication stress. · Gole B, Baumann C, Mian E, Ireno CI, Wiesmuller L · Oncogene 2015What this evidence cannot show
Reporter-based recombination readout in HeLa and other lines, aphidicolin-induced replication stress at experimental doses, single laboratory. Reporter recombination is not the same event as a genuine reciprocal translocation to AFF1.
Evidence that argues against this (1)
- A reporter measures recombination at a transplanted sequence; the endogenous locus has been mapped, and the breaks fall where transcription is. Etoposide-induced double-strand breaks across the MLL breakpoint cluster region colocalise with the RNA polymerase II binding site and the transcription initiation region of a gene-internal MLL promoter, and NOT with a nearby topoisomerase II consensus sequence; the authors conclude MLL's genomic instability is presumably linked to transcriptional processes. An EGFP recombination substrate in a neutral genomic context removes exactly the variable — native transcription through the hotspot — that the locus-level break map identifies as the operative one, so 'transferable out of its native locus' is a statement about sequence-intrinsic recombinogenicity, not a demonstration that the native break is replication-stress-driven. Scharf S et al. 2007
- E2moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper
Replication-stress-induced EndoG recruitment and MLLbcr cleavage are AID/base-excision-repair dependent, and APE1 inhibition protects human cord-blood haematopoietic stem/progenitor cells from MLLbcr cleavage.
Base excision repair proteins couple activation-induced cytidine deaminase and endonuclease G during replication stress-induced MLL destabilization · Gole B, Mian E, Rall M, Wiesmuller L · Leukemia 2018What this evidence cannot show
Same laboratory as E1, so not independently replicated. siRNA screen hits require orthogonal validation. The APE1 protection result is a cleavage assay, not a demonstration that fusion formation is reduced. AID expression in human fetal liver progenitors is itself not well established.
Evidence that argues against this (1)
- The competing apoptotic-nuclease model attributes the same cleavage to caspase-activated nucleases downstream of apoptotic signalling rather than to replication stress; the two models overlap in effector (EndoG is an apoptotic nuclease) but differ in trigger. Betti CJ et al. 2001
- E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
De novo KMT2A breakpoints concentrate in the centromeric half of the BCR, distinct from the telomeric TOP2-rich half favoured in therapy-related disease.
Distribution of 11q23 breakpoints within the MLL breakpoint cluster region in de novo acute leukemia and in treatment-related acute myeloid leukemia: correlation with scaffold attachment regions and topoisomerase II consensus binding sites · Broeker PL, Super HG, Thirman MJ, et al. · Blood 1996What this evidence cannot show
Small n and pre-NGS resolution; the exact alignment of the EndoG hotspot with the de novo breakpoint peak has not been formally tested in a large modern cohort.
Evidence that argues against this (1)
- CONTRADICTED FOR INFANTS, and this entry's section prose has been corrected accordingly (see 'Why this fits de novo infant disease better than TOP2 poisons do'). Scharf et al. state that 'infant AND therapy-related acute leukaemia patients display chromosomal breakpoints preferentially clustered in the TELOMERIC portion of the MLL breakpoint cluster region (SCII)'. Infant leukaemia is de novo, so a de-novo-equals-centromeric generalisation drawn from a small 1996 pre-NGS cohort does not hold for the de novo subgroup this graph is about; and the EndoG hotspot at the intron 11 / exon 12 boundary sits in the telomeric half, i.e. in the region shared by infant and therapy-related breakpoints rather than in one that distinguishes them. Read from the abstract of PMID 16983345, not the full text. Scharf S et al. 2007
supplies the AID limb the replication-stress entry points at, plus evidence AID is expressed in HSPC
the APE1 protection result is the hypothesis's only real mechanistic asset
the same phrase covers two different literatures: breakage at the KMT2A bcr before the fusion exists, versus dependency of the cell that already carries it
ATM signalling to RNF20 is required for replication-stress-induced cleavage of the KMT2A breakpoint cluster region
- ← co-occurs withApoptotic nucleases cut KMT2A, and cells that survive the attempt can carry a translocation
two proposed endogenous sources of the double-strand break that starts this disease; they compete as explanations and should be read together
any proposed break mechanism has to explain the position of both cluster regions, not just the major one
Counter-evidence (Scharf 2007, PMID 16983345): INFANT and therapy-related patients alike cluster in the TELOMERIC portion of the MLL BCR (SCII), so the EndoG intron 11/exon 12 hotspot does not distinguish de novo from therapy-related breakage; the de-novo-equals-centromeric clause in this entry was corrected on 2026-09-19