Oncofusion

start / graph / breakpoint-kmt2a-bcr-architecture

Breakpointmoderate confidence 3 of 5: One strong study, or several smaller consistent ones.Q1-fusion-genesis

The KMT2A breakpoint cluster region: an 8.3 kb stretch built to break

also called MLL bcr · KMT2A BCR · MLL breakpoint cluster region · 11q23 breakpoint cluster

Almost every KMT2A fusion snaps inside the same short 8.3 kilobase stretch of the gene, which happens to be loaded with features that make DNA easy to cut: places where the DNA is unwound and exposed, anchors that tether it to the nuclear scaffold, and preferred cutting sites for the enzyme topoisomerase II. The gene is not simply unlucky, that piece of it is structurally primed to break.

The same thing, for a scientist

KMT2A breakpoints cluster within an ~8.3 kb BamHI fragment spanning exons 5-11 in historic numbering. The BCR contains a weak scaffold-attachment region (SAR) centromeric to it, a high-affinity SAR in its telomeric half, six or more high-stringency vertebrate TOP2 consensus sites within that telomeric SAR, an in vivo drug-inducible TOP2 cleavage site near exon 9, and a strong DNase I hypersensitive site colocalising with it that also has cryptic promoter activity. De novo and therapy-related breakpoints are non-randomly distributed within the BCR (de novo skewed centromeric, t-AML skewed telomeric into the SAR/TOP2-rich half, P=0.02), implying different initiating mechanisms for the two aetiologies. Partner genes such as MLLT3/AF9 carry structurally homologous BCRs with their own colocalising TOP2 cleavage and DNase I HS sites, which is why the breakage-and-repair model generalises across the KMT2A recombinome.

Go deeper
What is actually in the 8.3 kb2 deeper

Mapping of 39 patient breakpoints against chromatin features identified (i) a low-affinity SAR just centromeric to the BCR, (ii) a high-affinity SAR inside the telomeric half, and (iii) one TOP2 consensus site adjacent to that telomeric SAR plus six inside it 1. Separately, an in vivo TOP2 cleavage site inducible by several different TOP2-targeting drugs was mapped near exon 9 in three leukaemia cell lines, colocalising with a strong DNase I hypersensitive site present in four leukaemia lines and in two lymphoblastoid lines with structurally normal KMT2A 2. The DNase HS site being present in normal cells is the important part: the fragility is a constitutive property of the locus, not a consequence of the rearrangement.

SARs and why nuclear-scaffold anchoring matters

Scaffold-attachment regions tether chromatin to the nuclear matrix. TOP2 is enriched at the matrix, so a SAR is effectively a place where DNA is held taut against a nuclease-like enzyme. That is a mechanical explanation for the enrichment of TOP2 consensus sites in the telomeric SAR 1. An independent line of work found that a predicted nuclear matrix attachment sequence and a strong predicted TOP2 binding site flank the apoptotic cleavage position adjacent to exon 12 4.

Cryptic promoter activity in the DNase HS region

The DNase I hypersensitive region in which therapy-related breakpoints cluster possesses cryptic promoter activity, which implicates transcription, not just topoisomerase occupancy, in the breakage mechanism 3. This is the hinge that connects the KMT2A BCR to the loop-anchor/transcription-fragility model (see mechanism-loop-anchor-dsb-hotspots).

De novo vs therapy-related breakpoints sit in different halves1 deeper

In 31 de novo acute leukaemias, 23/31 (74%) breakpoints mapped to the centromeric 4.57 kb of the BCR; in 8 therapy-related AMLs, 6/8 (75%) mapped to the telomeric 3.87 kb that contains the high-affinity SAR and the TOP2 consensus cluster (P = 0.02) 1. Read carefully, this is evidence that the same locus can be broken by at least two mechanistically distinct processes, a TOP2-poison-driven one in therapy-related disease and something else in de novo infant leukaemia. It is a strong argument against assuming the etoposide mechanism explains infant KMT2A::AFF1.

Contested: what independent cohorts found

Two things weaken this dichotomy and both come from groups other than the one that reported it.

  1. The split may be by age, not aetiology. In 40 de novo t(4;11) ALL patients, breakpoint distribution differed significantly between infants (<1 year) and older patients, defining two hotspot areas inside the BCR [E1 counter, Reichel 2001]. De novo KMT2A-r series are infant-heavy and therapy-related series are not, so the 1996 comparison confounds the two variables.
  2. The telomeric hotspot is reported to dominate infant leukaemia too. A review of the mechanism literature places the most frequent MLL rearrangements in a sub-kilobase telomeric region in infant acute leukaemia as well as in secondary leukaemia [E1 counter, Gole & Wiesmuller 2015].

The entry retains the 1996 numbers because they are the primary measurement, but the inference drawn from them - distinct initiating mechanisms - is not safe on breakpoint position alone. Separately, five secondary-AML patients who never received etoposides broke at the same TOP2-inducible hotspot as those who did [E2 counter, Langer 2003], which removes the simplest reading of the telomeric skew as a TOP2-poison signature.

Partner genes have the same architecture

AF9/MLLT3 at 9p22 has two patient breakpoint cluster regions; in BCR2 (introns 7-8) an in vivo TOP2 cleavage site colocalises with an in vitro DNase I hypersensitive site, and two SARs border the patient breakpoint regions. TOP2 cleavage at these sites was shown to be reversible by religation 5. So the partner is not a passive recipient: it is independently fragile by the same rules.

Whose data this is

This matters enough to state plainly, because it is the project's named standing weak spot. Three of the five evidence items here (E1, E2, E5) come from one laboratory - Janet Rowley's group in Chicago - and from one methodological pipeline of the Southern-blot era. The modern reference distribution of KMT2A breakpoints and partners likewise comes from a single referral-based diagnostic centre with 3401 patients accumulated 2003-2022 [E1 counter, Meyer 2023]; referral structure is itself a plausible source of the steep partner skew that cohort reports (7 partners plus PTD = >90% of recombinations, 63 partners seen once). No independent cohort has published a partner-stratified KMT2A intron distribution. The two datasets from other groups that do exist [Reichel 2001, Langer 2003] each disagree with the Chicago map in a specific way - an age split rather than an aetiology split, and partner breakpoints outside the published clusters. Where this entry asserts architecture, read 'as mapped once, by one group, in one assay generation'.

Facts and figures
geneKMT2A
cytoband11q23.3
bcr size kb8.3
bcr exon span historicexons 5-11
mechanism classmultiple: TOP2-mediated (t-AML-skewed), replication-stress/EndoG and apoptotic nuclease (de novo candidates); rejoining by non-homologous end joining with microhomology and small indels
reciprocal productder(4) carrying AFF1::KMT2A; usually expressed but its contribution is unsettled
source concentrationE1/E2/E5 all from the Rowley laboratory; the modern partner/breakpoint reference (Meyer 2023, PMID 37019990) is a single referral-based diagnostic centre. No independent partner-stratified KMT2A intron distribution has been published.
features
[
  {
    "feature": "weak scaffold attachment region (SAR1)",
    "position": "centromeric to BCR"
  },
  {
    "feature": "high-affinity SAR (SAR2)",
    "position": "telomeric half of BCR"
  },
  {
    "feature": "TOP2 consensus sites, high-stringency",
    "count": 7,
    "position": "6 within telomeric SAR, 1 adjacent"
  },
  {
    "feature": "in vivo drug-inducible TOP2 cleavage site",
    "position": "near exon 9"
  },
  {
    "feature": "strong DNase I hypersensitive site with cryptic promoter activity",
    "position": "near exon 9"
  },
  {
    "feature": "apoptosis-inducible nuclease cleavage site",
    "position": "adjacent to exon 12"
  },
  {
    "feature": "replication-stress / EndoG hotspot",
    "position": "intron 11 to exon 12 boundary, ~0.4 kb hotspot"
  }
]
breakpoint distribution
{
  "de_novo_centromeric_half_pct": 74,
  "therapy_related_telomeric_half_pct": 75,
  "p_value": 0.02,
  "n_de_novo": 31,
  "n_t_aml": 8,
  "contested": "Confounded with patient age (Reichel 2001, PMID 11420702, 40 de novo t(4;11) patients, two age-linked hotspots) and contradicted in a review that places the dominant infant hotspot in the telomeric sub-kilobase region (Gole & Wiesmuller 2015, PMID 26161385)."
}
How this could bite us back
  • Because the fragile features (DNase HS, cryptic promoter, SAR, TOP2 sites) are constitutive properties of a locus that normal cells need to transcribe, there is no way to harden the locus without disabling normal KMT2A regulation.
  • If the de novo / therapy-related breakpoint dichotomy does not hold, then the entry's cleanest argument against extrapolating the etoposide mechanism to infant KMT2A::AFF1 weakens - and the TOP2 model becomes harder to exclude rather than easier, which is the opposite of the comfortable conclusion.
Still open
  1. Is the KMT2A BCR DNase-hypersensitive and TOP2-occupied in human fetal liver HSC/LMPP, the actual cell of origin, or only in transformed lines? I could not find this measured anywhere.
  2. What is the corresponding structural map of the AFF1 breakpoint region? The KMT2A side is well characterised and the AFF1 side is not.
  3. Does the 0.4 kb intron 11 / exon 12 hotspot coincide with a CTCF/cohesin loop anchor in fetal progenitors?
  4. Has any group other than the Frankfurt diagnostic centre published a partner-stratified distribution of KMT2A intron breakpoints? If not, every frequency figure in this graph that is stratified by partner inherits one cohort's referral structure, and the project should say so wherever it uses those numbers.
Every source behind this entry (5)
  1. E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    KMT2A breakpoints are non-randomly distributed within the 8.3 kb BCR, with de novo breakpoints skewed centromeric and therapy-related breakpoints skewed into a telomeric high-affinity SAR containing six TOP2 consensus sites (P=0.02).

    PMID 8634439
    What this evidence cannot show

    Small n (31 de novo, 8 t-AML), Southern-blot-era mapping resolution, and the TOP2 consensus sites are computational predictions from in vitro vertebrate criteria rather than measured cleavage.

    Evidence that argues against this (3)
    • The de novo / therapy-related split is confounded with patient age, and an independent group's data support an age split rather than an aetiology split. Reichel et al. sequenced derivative chromosomes from 40 t(4;11) ALL patients - all de novo - and found significant differences in the distribution of chromosomal breakpoints between patients below and above one year of age, defining TWO hotspot areas within the MLL BCR on that basis and concluding that different age-linked mechanisms may operate. Because de novo KMT2A-r cohorts are infant-enriched and therapy-related cohorts are not, a cohort comparison of the 1996 design cannot separate aetiology from age. Reichel M et al. 2001
    • A review of the mechanism literature states the opposite of the clean centromeric/telomeric split: rearrangements involving a short TELOMERIC region of under 1 kb of MLL are the most frequently observed molecular change in INFANT acute leukaemia as well as in secondary leukaemia. If the same telomeric sub-region dominates both aetiologies, the P=0.02 dichotomy in 39 Southern-blot-era patients does not generalise, and the inference that de novo and therapy-related disease must have distinct initiating mechanisms cannot be drawn from breakpoint position alone. Gole B et al. 2015
    • Source concentration. Evidence items E1, E2 and E5 of this entry all originate from the Rowley laboratory in Chicago (Broeker/Super/Thirman/Rowley 1996; Strissel/Strick/Rowley/Zeleznik-Le 1998; Strissel/Strick/Rowley/Zeleznik-Le 2000), using the same Southern-blot and in vitro cleavage pipeline. The modern partner-frequency and breakpoint reference for the field - the KMT2A recombinome, 3401 patients analysed between 2003 and 2022 - is likewise a single referral-based diagnostic centre, in which the seven commonest partners plus PTD account for more than 90% of recombinations while 63 partners were seen exactly once. No independent cohort has published a partner-stratified KMT2A intron distribution. The architectural claims in this entry should be read as one laboratory's map corroborated in part by one other laboratory's cohort (Reichel 2001, Langer 2003), both of which disagree with it in detail. Meyer C et al. 2023
  2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    A drug-inducible in vivo TOP2 cleavage site and a strong DNase I hypersensitive site colocalise near KMT2A exon 9, and the DNase HS site is present in cells with structurally normal KMT2A.

    PMID 9808573
    What this evidence cannot show

    Leukaemia cell lines and lymphoblastoid lines, not primary fetal progenitors; cleavage was drug-induced rather than spontaneous.

    Evidence that argues against this (1)
    • The TOP2-inducible hotspot is used by patients with no TOP2-poison exposure. Langer et al. sequenced t(9;11) breakpoints in 11 children and two cell lines and found that all five patients with secondary AML who had NOT received etoposides during treatment of the primary disease carried almost identical MLL breakpoints, very close to the breakage hot spot inducible in vitro by topoisomerase II inhibitors OR by apoptotic triggers. The colocalisation of a drug-inducible TOP2 cleavage site with the breakpoint region is therefore compatible with cleavage by something else at the same accessible position. Langer T et al. 2003
  3. E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    Therapy-related KMT2A breakpoints cluster in a DNase I hypersensitive region that possesses cryptic promoter activity, implicating transcription alongside TOP2 activity.

    PMID 22615413DOI
    What this evidence cannot show

    Human cell lines with etoposide at pharmacological doses; the cryptic-promoter claim rests on reporter assays.

    Evidence that argues against this (1)
    • 'Transcription alongside TOP2 activity' is a weaker pairing than it reads, because where the breaks were actually measured at this locus, the topo II consensus was excluded. Mapping etoposide-induced double-strand breaks across the MLL breakpoint cluster region placed them at the RNA POLYMERASE II BINDING SITE and the transcription initiation region of a gene-internal MLL promoter, and explicitly NOT at a nearby topoisomerase II consensus sequence; the authors conclude that the genomic instability of the human MLL gene is 'presumably linked to transcriptional processes'. A cryptic promoter with DNase I hypersensitivity and a measured break position at Pol II are the same observation; TOP2 is the part that the break map did not support. Scharf S et al. 2007
  4. E4moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    A predicted nuclear matrix attachment sequence and a strong predicted TOP2 binding site flank the apoptosis-inducible cleavage site adjacent to KMT2A exon 12, and a 367 bp fragment of the BCR is sufficient for site-specific cleavage on an episome.

    PMID 16193084DOI
    What this evidence cannot show

    TK6 lymphoblastoid cells and episomal reporters, an artificial chromatin context; the motif calls are predictions.

    Evidence that argues against this (1)
    • Motif PREDICTIONS at leukaemia breakpoints have a poor record against measured breaks, and this entry is carrying two of them. At the paradigm prenatal fusion, long-distance inverse PCR of genomic junctions from nine patients found that the sequences near the breakpoints 'did not reveal any consistent characteristic signal sequences of the V(D)J recombinase, topoisomerase II consensus sites, or other sequence motifs associated with recombination' — the junctions instead showed generic non-homologous end-joining signatures (microhomology, small deletions and duplications). A predicted matrix-attachment region and a predicted strong TOP2 site flanking an episomal cleavage site are hypotheses about where a nuclease might cut; the patient data at the analogous locus say such predictions do not mark where nucleases did cut. Wiemels JL et al. 1999
  5. E5moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    The AF9/MLLT3 partner locus contains its own breakpoint cluster regions with a colocalising in vivo TOP2 cleavage site, DNase I HS site and flanking SARs, and TOP2 cleavage there is reversible.

    DNA structural properties of AF9 are similar to MLL and could act as recombination hot spots resulting in MLL/AF9 translocations and leukemogenesis · Strissel PL, Strick R, Tomek RJ, Roe BA, Rowley JD, Zeleznik-Le NJ · Human Molecular Genetics 2000
    PMID 10861294DOI
    What this evidence cannot show

    This is AF9/MLLT3, not AFF1. The equivalent structural dissection of the AFF1 breakpoint region is far less complete, which is a real gap for the lab's primary fusion.

    Evidence that argues against this (1)
    • The AF9/MLLT3 breakpoint cluster regions in this evidence item are not a complete map. In the same study, five of eleven t(9;11) patients had AF9 breakpoints located OUTSIDE the two previously described BCRs - in the centromeric region of intron 4, and in one case within intron 3. The structural-fragility account of the partner locus was built on the subset of breakpoints that fall inside the published clusters, so the fraction of the partner's breakpoint distribution the SAR/TOP2/DNase-HS architecture explains is smaller than the entry implies. Langer T et al. 2003
Connected to