Oncofusion

start / graph / concept-programmed-dna-breaks-vdj-and-csr

Explainermoderate confidence 3 of 5: One strong study, or several smaller consistent ones.Q32-break-machinery

Programmed DNA breaks - V(D)J and class switching - and the verdict on whether either one breaks KMT2A

also called programmed double-strand breaks · RAG and AID at KMT2A · immune recombinase off-target at 11q23

The immune system deliberately cuts and re-joins its own chromosomes twice in a B cell's life, and both machines are known to cut the wrong place sometimes. This entry asks whether either one is what breaks KMT2A, and the answer is: the V(D)J machine almost certainly not, the antibody-mutating enzyme perhaps, by a side door.

The same thing, for a scientist

Two programmed DNA-break systems operate in B-lineage cells: RAG1/RAG2 cleaving at recombination signal sequences during V(D)J recombination, and AID deaminating cytosine in transcribed single-stranded DNA during class switch recombination and somatic hypermutation. Both are proven causes of oncogenic translocation elsewhere - RAG-mediated recombination is the predominant driver of oncogenic rearrangement in ETV6::RUNX1 ALL [E1], and AID is required for c-myc/IgH translocations in vivo. KMT2A-rearranged leukaemia is a B-lineage disease and the question of whether either machine makes its founding break is therefore not idle. The verdicts differ. **RAG: almost certainly not.** The only cohort-scale junction analysis concluded the events were NOT produced by immune-system recombinases [E2], and title-checked PubMed searches find no report of a cryptic recombination signal sequence at the KMT2A BCR [E3]. **AID: an open side door.** Off-target AID binding at the MLLbcr initiates base-excision-repair-mediated single-strand cleavage that is converted by endonuclease G into a rearrangement-competent lesion, and APE1 inhibition protected human cord-blood HSPC from that cleavage [E4]; AID is expressed and functionally consequential in haematopoietic stem/progenitor cells outside any germinal centre [E5]. That route uses AID's chemistry without using a switch region, so it is not class switch recombination - it is AID off target. **And a 2024 finding complicates the cell-of-origin story:** in 121 KMT2A::AFF1 cases across all age classes, V(D)J recombination is proximally biased with up to 50% of samples using the most proximal IGHV gene IGHV6-1, and de novo DJH recombination is more abundant than in other BCP-ALL subtypes [E6]. The RAG machinery is therefore demonstrably active in these blasts, even though nothing implicates it in the founding break.

Go deeper
The two machines, and what this entry does not restate

The graph holds full entries for each: mechanism-vdj-recombination-programmed-breaks covers RAG's transposon ancestry, the hairpin/Artemis chemistry and the SCID phenotypes; mechanism-class-switch-recombination-programmed-breaks covers AID's deamination mechanism, the hyper-IgM phenotype and the genetic separation of the translocation pathway from switching; mechanism-meiotic-recombination-spo11-programmed-breaks covers the third case; mechanism-rag-off-target-recombination covers the ETV6::RUNX1 result in detail. This entry is the umbrella the replication-stress entry points at, and its only job is the KMT2A verdict.

RAG at KMT2A: the negative, and exactly how strong it is1 deeper

The positive case has never been argued in print. The negative rests on two things and neither is a cohort-scale modern test.

  1. The junction evidence. Reciprocal breakpoints from 14 primary t(4;11) leukaemias carried duplications, deletions and inversions of under a few hundred nucleotides, and the authors explicitly concluded these were products of error-prone DNA repair and not of immune-system recombinase activity 2. n=14, Sanger era, never repeated.
  2. The absence of a motif report. Title/abstract searches for a recombination signal sequence at this locus return zero records 3.

A trap worth recording, because it nearly produced a wrong number here. The query (MLL OR KMT2A) AND (recombination signal sequence OR cryptic RSS) returns a healthy-looking 16 records. Reading the querytranslation shows PubMed mapped recombination signal sequence onto "protein sorting signals"[MeSH] and RSS onto "webcast"[Publication Type] OR "rss"[All Fields]. The hits include transgenic sunflower and carrot interferon papers. The corrected phrase query "recombination signal sequence"[tiab] returns 0 3. A count of 16 here would have read as a literature and was noise.

But the RAG machinery is running in these cells

Muller et al. analysed 379 BCP-ALL cases by whole-transcriptome sequencing, including 48 adult, 36 paediatric and 37 infant KMT2A::AFF1 cases, and found strongly favoured usage of proximal IGHV genes, with up to 50% of KMT2A-r samples carrying a rearrangement involving IGHV6-1, the most proximal IGHV gene; a median of three distinct DNJH stems per sample; and shorter CDR3 regions than most other subtypes 6. Their reading is that both the proximal bias and the abundant de novo DJH recombination indicate a more primitive differentiation stage, because a fully randomised repertoire requires IGH locus decontraction controlled by PAX5 and WAPL.

This matters for this beat in a way the paper does not discuss. It means RAG is active in the KMT2A::AFF1 clone and is continuing to recombine after transformation - so the absence of a RAG signature at the founding junction is not explained by the cells lacking RAG. That was the comfortable explanation and it is not available. The honest statement becomes: RAG is present and working in these cells, and it still does not appear to be what cut KMT2A.

AID at KMT2A: not class switching, but AID chemistry

The AID route to KMT2A does not run through a switch region and is not class switch recombination. Gole et al. traced replication-stress-induced MLLbcr cleavage to off-target AID binding, which initiates base-excision-repair-mediated single-strand cleavage that derailed endonuclease G activity converts into a rearrangement-competent lesion; inhibiting APE1 protected both tumour cells and human cord-blood-derived haematopoietic stem/progenitor cells from MLLbcr cleavage 4.

The obvious objection - AID is a germinal-centre enzyme, and the cell of origin here is a fetal progenitor - is weaker than it looks. AID is expressed and active in haematopoietic stem/progenitor cells: Aid loss in mice expands myeloid cells and reduces erythroid progenitors with dysregulated Cebpa and Gata1, and silencing AID in human bone marrow cells skews differentiation toward the myelomonocytic lineage 5. AID expression is also documented outside germinal centres in human disease. What remains genuinely unmeasured is AID in the human fetal liver HSC/LMPP compartment, which is the specific cell this disease arises in.

Why the answer matters, in one sentence each

If RAG made the break, the ETV6::RUNX1 logic would transfer: cryptic RSS motifs would be findable, the break would be developmentally scheduled, and a RAG-timing argument would constrain the window. It does not appear to.

If AID made the break, prevention has a pharmacological node (APE1) that has already been shown to protect the right cell type 4, and the mechanism becomes partly independent of topoisomerase II. That is the single strongest argument in this graph for a preventable break, and it is developed in hyp-block-kmt2a-break-machinery-upstream.

If neither did, the remaining candidates are TOP2 (exogenous or endogenous), apoptotic nucleases, and replication stress - all of which are covered by their own entries, and none of which offers a target that could be safely occupied in a pregnant woman.

Facts and figures
query trap recorded(MLL OR KMT2A) AND (recombination signal sequence OR cryptic RSS) returns 16 records; querytranslation maps the phrase to protein sorting signals and RSS to webcast[Publication Type]. The quoted tiab form returns 0.
druggable node if aid route trueAPEX1 (APE1)
systems
[
  {
    "name": "V(D)J recombination",
    "nuclease": "RAG1/RAG2",
    "target_motif": "recombination signal sequence (12/23 spacer)",
    "stage": "pro-B and pre-B, and T-lineage equivalents",
    "kmt2a_verdict": "not implicated in the founding break; junction evidence against, no RSS reported, but RAG is demonstrably active in the established clone",
    "graph_entry": "mechanism-vdj-recombination-programmed-breaks"
  },
  {
    "name": "class switch recombination and somatic hypermutation",
    "nuclease": "AID (AICDA) plus UNG and base-excision repair",
    "target_motif": "WRCY/RGYW hotspots in transcribed single-stranded DNA; switch regions",
    "stage": "activated mature B cells, germinal centre",
    "kmt2a_verdict": "class switching itself not implicated; AID off-target activity at the MLLbcr via base-excision repair and endonuclease G IS on record",
    "graph_entry": "mechanism-class-switch-recombination-programmed-breaks"
  },
  {
    "name": "meiotic recombination",
    "nuclease": "SPO11",
    "target_motif": "PRDM9-directed hotspots",
    "stage": "meiosis",
    "kmt2a_verdict": "not applicable to a somatic break; PRDM9 allelic status in KMT2A-r infants has never been looked at",
    "graph_entry": "mechanism-meiotic-recombination-spo11-programmed-breaks"
  }
]
vdj activity in kmt2a aff1
{
  "n_cases": 121,
  "n_adult": 48,
  "n_paediatric": 36,
  "n_infant": 37,
  "proximal_ighv6_1_usage_pct_up_to": 50,
  "median_dnjh_stems_per_sample": 3,
  "cdr3_length": "shorter than most other BCP-ALL subtypes",
  "interpretation_by_authors": "more primitive differentiation stage; IGH decontraction requires PAX5 and WAPL",
  "source": "E6"
}
How this could bite us back
  • The one programmed-break route with a drug attached, AID to base-excision repair to endonuclease G, runs through APE1 - the central abasic-site endonuclease. Inhibiting it chronically would be mutagenic and myelosuppressive, which is the opposite of what a prophylactic needs to be.
  • c-NHEJ, the pathway that joins the KMT2A ends, is absolutely required for V(D)J recombination, so any attempt to block the joining step in a fetus or infant would cause severe combined immunodeficiency.
  • Concluding that RAG is not involved removes the developmental-timing argument that would otherwise bound the window in which the break can occur, which makes prevention harder to schedule, not easier.
Still open
  1. Has anyone scanned the 3401 determined KMT2A breakpoints for cryptic RSS motifs, including the nonamer-independent CCACCTCT class that RAG uses at BCL2? A negative at that scale would close the RAG question permanently, and it is a text search over sequences that already exist.
  2. Is AID protein detectable in human fetal liver HSC/LMPP? Every existing measurement is adult bone marrow or mouse. This is a flow or single-cell question on tissue that human developmental atlases already hold.
  3. Does the proximal V(D)J bias in KMT2A::AFF1 predate the fusion or follow it? If KMT2A::AFF1 itself constrains IGH locus decontraction through PAX5/WAPL, the bias is a fusion phenotype; if it reflects the cell of origin, it is a marker of the target compartment.
  4. Do the rare KMT2A-r cases with a mature pro-B phenotype and productive IGH rearrangement have different junction structures from CD10-negative fetal-type cases?
Every source behind this entry (6)
  1. E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    RAG-mediated recombination is the predominant mutational process generating oncogenic rearrangement in ETV6::RUNX1 ALL, with RSS motifs at breakpoints, non-templated junctional sequence and roughly 30-fold enrichment at promoters and enhancers of genes actively transcribed in B-cell development.

    PMID 24413735DOI
    What this evidence cannot show

    Cited here as the comparator subtype, not as a claim about KMT2A. The analysis concerns SECONDARY lesions in ETV6::RUNX1 ALL; the founding ETV6::RUNX1 translocation itself is not established as RAG-mediated by this work.

    Evidence that argues against this (1)
    • The contrast with KMT2A is the point and it is on record in the primary literature: sequencing of reciprocal breakpoints in 14 primary t(4;11) leukaemias found duplications, deletions and inversions consistent with error-prone repair, and the authors explicitly concluded the events were not produced by immune-system recombinases. The ETV6::RUNX1 prevention and detection logic therefore does not transfer to KMT2A-rearranged infant ALL. Gillert E et al. 1999
  2. E2moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    Reciprocal t(4;11) breakpoints sequenced from 14 primary leukaemias carry duplications, deletions and inversions of under a few hundred nucleotides consistent with error-prone DNA repair, and were explicitly concluded not to be the product of immune-system recombinase activity.

    PMID 10467413DOI
    What this evidence cannot show

    n=14, Sanger-era cloning, one laboratory, never repeated at cohort scale. The confidence here is deliberately 'moderate' rather than 'strong' despite the claim being widely repeated, because this is a negative resting on a small sample and on the absence of a motif nobody searched for systematically.

    Evidence that argues against this (1)
    • An exclusion supported by an absence of contrary work is weaker than an exclusion that won a contest, and no published study has ever argued the positive case for RAG at KMT2A. Separately, RAG is known to nick non-immune loci by a nonamer-INDEPENDENT mechanism - at the BCL2 minor breakpoint cluster region through a CCACCTCT motif rather than a canonical RSS - and then synapse that nick with a genuine RSS in cells. A search for canonical RSS motifs at KMT2A would miss exactly that mechanism, so 'no RSS found' is not equivalent to 'RAG was not involved'. Nambiar M et al. 2012
  3. E3moderate confidence 3 of 5: One strong study, or several smaller consistent ones.database

    Title/abstract searches run 2026-09-20 return zero records reporting a recombination signal sequence at the KMT2A locus, and the unquoted form of the query returns 16 records only because PubMed maps 'recombination signal sequence' to protein sorting signals and 'RSS' to the webcast publication type.

    What this evidence cannot show

    Title/abstract indexing, two gene symbols. A cryptic-RSS analysis reported in a supplementary table would be invisible. Singular and plural were both run and both return 0, per the standing plural rule.

    Evidence that argues against this (1)
    • A zero count over a motif NAME does not establish that the motif is absent from the sequence, only that nobody published about it under that name. Because the recombinome holds 3401 determined breakpoints, the motif question is directly answerable computationally and has not been, which the graph's own RAG entry lists as its first open question. Meyer C et al. 2023
  4. E4moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    Replication-stress-induced endonuclease G recruitment and MLLbcr cleavage are AID/base-excision-repair dependent, and APE1 inhibition protected human cord-blood-derived haematopoietic stem/progenitor cells from MLLbcr cleavage.

    PMID 28626219DOI
    What this evidence cannot show

    One laboratory, never independently replicated. The protection result is a cleavage assay, not a demonstration that fusion formation falls. Read from the abstract. The graph's `mechanism-replication-stress-endog-cleavage-kmt2a` carries the full treatment; this entry cites it only for the AID limb.

    Evidence that argues against this (1)
    • The endogenous KMT2A break has been mapped to where transcription is, not to where a deaminase 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 the authors conclude MLL's genomic instability is presumably linked to transcriptional processes. An AID/BER route competes with, rather than complements, that account. Scharf S et al. 2007
  5. E5moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    AID is expressed and functionally consequential in haematopoietic stem/progenitor cells outside any germinal centre: Aid loss in mice expands myeloid cells and reduces erythroid progenitors with dysregulated Cebpa and Gata1, and silencing AID in human bone marrow cells skews differentiation toward the myelomonocytic lineage.

    PMID 28077417DOI
    What this evidence cannot show

    Mouse knockout plus human adult bone-marrow knockdown. This is NOT a measurement of AID in human FETAL LIVER HSC/LMPP, which is the compartment the KMT2A::AFF1 fusion arises in, and that measurement remains missing. A knockdown phenotype implies expression but does not quantify it.

    Evidence that argues against this (1)
    • The same paper draws an explicit limit on how far AID's role in progenitors extends: in contrast to Tet2 loss, Aid loss does NOT enhance HSC self-renewal and does not cooperate with Flt3-ITD to induce myeloid transformation. AID being present in progenitors is therefore not evidence that it is oncogenically consequential there, and an entry using this to argue for AID-driven leukaemogenesis is over-reading it. Kunimoto H et al. 2017
  6. E6strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    Across 48 adult, 36 paediatric and 37 infant KMT2A::AFF1 BCP-ALL cases, V(D)J recombination is proximally biased with up to 50% of samples carrying a rearrangement involving the most proximal IGHV gene IGHV6-1, a median of three distinct DNJH stems per sample, and shorter CDR3 regions than most other BCP-ALL subtypes - read by the authors as indicating a more primitive differentiation stage.

    PMID 38650597DOI
    What this evidence cannot show

    Full text read via Europe PMC fullTextXML (PMC11033919, 8,878 characters - a short-format report). Whole-transcriptome sequencing, so IGHV usage is measured as EXPRESSION of rearranged alleles, not as genomic rearrangement; a silent rearrangement is invisible. Cohort is from one commercial diagnostic laboratory plus a St Jude validation set. The efetch record returns an EMPTY abstract for this PMID, which is why it would have been missed by an abstract-only screen.

    Evidence that argues against this (1)
    • Ongoing V(D)J activity in a diagnostic sample says nothing about RAG status in the cell at the moment of the founding translocation, which occurred in utero, possibly years earlier, in a progenitor that may not have expressed RAG at all. The authors themselves frame the finding as clonal evolution of IGH alleles and as a problem for IG-based MRD detection, not as a statement about leukaemia initiation. Gale KB et al. 1997
Connected to