Oncofusion

start / graph / mechanism-different-breaks-different-biology

Mechanismmoderate confidence 3 of 5: One strong study, or several smaller consistent ones.Q2-fusion-atlas

Do different breaks give different biology? A verdict on partner identity vs breakpoint position

also called breakpoint-phenotype correlation · partner vs breakpoint · KMT2A genotype-phenotype

Yes for the partner gene: which gene KMT2A joins strongly predicts whether the child gets lymphoid or myeloid leukaemia, how old they are, and how they do. Much weaker for where inside KMT2A the break lands: that changes the gene-expression flavour and correlates with age, but the evidence that it changes survival rests on about thirty patients.

The same thing, for a scientist

The two axes must be separated. **Partner identity** is a strong, replicated determinant of lineage and age: in 2182 KMT2A-r ALL cases AFF1 accounts for 56.5%, MLLT1 18.5% and MLLT3 11.8%, whereas in 1116 AML cases MLLT3 accounts for 30.4%, MLLT10 18.7%, KMT2A-PTD 10.7%, ELL 10.1% and AFDN 8.1%, with AFF1 almost absent from AML (0.8%) [E1]. T-ALL is dominated by MLLT1 (43/123) and MLLT10 (40/123) [E1]. Adult ALL is 82.7% AFF1 [E1]. **Breakpoint position** is a weaker, partly confounded determinant: it correlates with age within a partner (infant KMT2A::AFF1 breaks in intron 11, adult in introns 9-10), with lineage at the cohort level (AML prefers intron 9, infant ALL prefers intron 11 up to ~6 months), and with the HOXA transcriptional signature (intron 9-10 = HOXA-high, exon 11+ = HOXA-low) [E1][E2]. The one prognostic claim - intron 11 breaks do worse - rests on 30 breakpoint-typed patients in a single retrospective cohort and is explicitly not confirmed by the 3401-patient recombinome [E1][E3]. Crucially, breakpoint position does **not** change which named domain survives on the direct fusion protein: every major-BCR break yields the same AT-hook/MBM/CXXC oncoprotein. Where it demonstrably changes protein content is the **reciprocal** product [E1].

Go deeper
Axis 1: partner identity drives lineage - strong evidence1 deeper

From 3401 genomically characterised patients 1:

PartnerALL (n=2182)AML (n=1116)Signature niche
AFF11233 (56.5%)9 (0.8%)infant + adult B-ALL; 82.7% of adult KMT2A-r ALL
MLLT1 (ENL)404 (18.5%)47 (4.2%)infant B-ALL; top T-ALL partner (43/123)
MLLT3 (AF9)258 (11.8%)339 (30.4%)the true bilineage partner; top AML partner
MLLT10 (AF10)80 (3.7%)209 (18.7%)AML and T-ALL (40/123)
ELL3 in infant ALL113 (10.1%)almost pure AML
AFDN (AF6)46 (2.1%)91 (8.1%)AML-leaning; minor-BCR T-ALL cases
KMT2A-PTD1 (0.05%)120 (10.7%)adult AML (25.6% of adult KMT2A-r AML)

The partner-lineage effect is enormous and reproducible: AFF1 is ~70-fold enriched in ALL over AML, ELL is essentially an AML partner, and MLLT3 is the one partner that is genuinely common in both. This is the axis on which the field's phenotype predictions rest, and it is solid.

MPAL sits between: 61/3401 cases (1.8%), 35 of them infants, distributed across AFF1, MLLT1, MLLT3, MLLT10, AFDN and others - i.e. no single partner owns the mixed phenotype 1.

Lineage is not fixed by the partner alone

A CRISPR/Cas9-engineered t(4;11) in primary human HSPCs - the same fusion pair as infant pro-B ALL - retained the plasticity to differentiate into either the myeloid or the lymphoid lineage 4. So the partner biases lineage; the cell of origin and its environment finish the job. That is consistent with the clinical existence of KMT2A-r MPAL and of lineage switching under therapy.

Axis 2: breakpoint position - real but weaker, and confounded1 deeper

What is solid. Breakpoint position differs significantly between partners and age groups in a 3401-patient cohort 1:

  • KMT2A::MLLT1 and KMT2A::SEPTIN6 prefer intron 11; most other partners prefer intron 9 or upstream.
  • Infant KMT2A::AFF1 breaks predominantly in intron 11; adult KMT2A::AFF1 shifts to introns 9-10.
  • Paediatric KMT2A::MLLT10 and KMT2A::ELL shift towards intron 9; KMT2A::AFDN breaks mostly in intron 9.
  • KMT2A::MLLT6 shifts to introns 8-9; KMT2A::MLLT11 to introns 9-10.
  • Therapy-induced leukaemia shifts significantly to intron 11 - but for KMT2A::ELL only, not for AFF1, MLLT3 or MLLT10.
  • ALL switches from an intron-11 preference (<~6 months) to an intron-9 preference thereafter; AML prefers intron 9 throughout.
  • Meyer et al. count 10 partner subgroups skewed to exons 11-13 and 25 skewed to exon 9-intron 10, and conclude that certain fusion genes have a selective preference for distinct breakpoints.

What is soft. The mechanism and the prognosis.

  • HOXA: intron 9/exon 10/intron 10 breaks associate with a HOXA-high signature, exon 11 and downstream with HOXA-low 12. But breakpoint and age are tightly confounded in t(4;11), and the primary study that established the split used a small microarray cohort with n=3 normal references 2.
  • Outcome: the only direct evidence is Emerenciano et al., where breakpoints were typed in 30 KMT2A-r cases out of 545 children and poor-outcome patients preferentially broke in intron 11 3. Age <=6 months was independently and much more strongly prognostic in the same study (P<0.001). The recombinome authors state plainly that whether the breakpoint shifts affect outcome 'is yet unclear' 1.

Verdict: evidence is thin. There is no prospective, adequately powered, age- and partner-adjusted analysis of KMT2A breakpoint position versus survival anywhere in the literature we could find. This is the beat's biggest gap.

Exon 9 vs exon 11 junctions: what actually differs

It is worth being blunt, because this is where intuition misleads.

Same on the direct fusion. A break in intron 9 and a break in intron 11 both lie 5' of PHD1. Both direct fusion proteins therefore contain AT-hooks 1-3, the SNL/repression region with MBM1 and MBM2, the LEDGF-binding region and CXXC, and both lack PHD1-3, the bromodomain, ePHD4, FYRN/FYRC, the TAD, the WIN motif and SET 1. The difference in the direct protein is ~2 exons of intervening sequence, not a functional module.

Different on the reciprocal fusion. PHD1-PHD2-ePHD3 is encoded by KMT2A exons 11-16. A break upstream of exon 11 delivers an intact PHD1 to the reciprocal partner::KMT2A protein; a break within or downstream of exon 11 truncates PHD1 mid-domain, which is reported to alter its folding and to impair CYP33/PPIE binding and PHD1 homodimerisation 1. Since CYP33 binding to PHD3 is the switch that recruits BMI1/HDAC/CBX4/KDM5B repressors to CXXC, an intron-11 break removes that brake from the reciprocal product.

Consequence for the lab. If breakpoint position genuinely changes prognosis, the mechanism most likely runs through the reciprocal protein or through the transcriptional signature, not through the direct oncoprotein. That is a sharp, falsifiable prediction and it re-centres the reciprocal product as a target.

Where evidence is thin - stated explicitly
  1. No prognosis-by-breakpoint study of adequate size. n=30 typed patients, retrospective, single country, unadjusted for age. Everything downstream of this claim is provisional.
  2. No menin-inhibitor response data stratified by breakpoint. Revumenib/ziftomenib trials enrol by KMT2A-r status (FISH/karyotype/RNA), not by genomic breakpoint, so this analysis has never been done even though the samples exist.
  3. No breakpoint-matched HOXA analysis. To separate age from breakpoint you need infants and non-infants with the same partner and the same breakpoint. Published cohorts are too small.
  4. Minor-BCR biology is essentially unstudied. 47 patients total; no functional model of a PHD-retaining KMT2A fusion exists.
  5. No independent replication of the recombinome breakpoint distributions. One diagnostic centre generated essentially the entire world dataset. Referral bias cannot be excluded.
  6. The reciprocal-outcome claim is unverified at source. We could not resolve the two primary clinical papers asserting better outcome with reciprocal expression.
Facts and figures
mechanism classgenotype-phenotype correlation
axis 1 partner identity
{
  "strength": "strong, replicated",
  "all_partner_shares_pct": {
    "AFF1": 56.5,
    "MLLT1": 18.5,
    "MLLT3": 11.8,
    "MLLT10": 3.7,
    "AFDN": 2.1,
    "EPS15": 1.6,
    "USP2": 0.9
  },
  "aml_partner_shares_pct": {
    "MLLT3": 30.4,
    "MLLT10": 18.7,
    "KMT2A-PTD": 10.7,
    "ELL": 10.1,
    "AFDN": 8.1,
    "MLLT1": 4.2,
    "MLLT11": 2.4,
    "SEPTIN6": 1.7,
    "MLLT6": 1.6,
    "EPS15": 1.3,
    "SEPTIN9": 1.3,
    "AFF1": 0.8
  },
  "t_all_partners": {
    "n_total": 123,
    "MLLT1": 43,
    "MLLT10": 40
  },
  "mpal": {
    "n": 61,
    "pct_of_cohort": 1.8,
    "infants": 35,
    "paediatric": 16,
    "adult": 10
  },
  "population": "3401 KMT2A-r acute leukaemia patients (DCAL 2003-2022)"
}
axis 2 breakpoint position
{
  "strength": "moderate for age/lineage/HOXA correlation; weak for prognosis",
  "intron_11_preferring_partners": [
    "MLLT1",
    "SEPTIN6",
    "infant AFF1",
    "paediatric and adult ELL",
    "therapy-induced leukaemia overall"
  ],
  "intron_9_preferring_partners": [
    "AFDN",
    "adult AFF1",
    "paediatric MLLT10",
    "paediatric ELL",
    "MLLT11 (9-10)",
    "MLLT6 (8-9)",
    "AML overall"
  ],
  "hoxa_high_breaks": "exon 9 / intron 9 / exon 10 / intron 10",
  "hoxa_low_breaks": "exon 11 and downstream",
  "age_switch_point_all": "~6 months (intron 11 preference before, intron 9 after)",
  "prognostic_claim": "intron 11 worse; cut point intron10/exon11",
  "prognostic_evidence_n": 30,
  "domain_difference_on_direct_fusion": "none - all major-BCR breaks yield the same AT-hook/MBM1-2/CXXC oncoprotein",
  "domain_difference_on_reciprocal_fusion": "intact PHD1 (upstream of exon 11) vs truncated, mis-folded PHD1 with impaired CYP33 binding and homodimerisation (within/downstream of exon 11)"
}
evidence gaps
[
  "no prospective, age- and partner-adjusted breakpoint-versus-survival analysis exists",
  "no menin-inhibitor response data stratified by KMT2A breakpoint",
  "no breakpoint-matched infant vs non-infant HOXA comparison",
  "no functional model of a PHD-retaining minor-BCR KMT2A fusion",
  "no independent replication of the recombinome breakpoint distributions outside one diagnostic centre"
]
How this could bite us back
  • If lineage is set by cell of origin and environment rather than by the fusion (as the CRISPR t(4;11) plasticity result suggests), then lineage switching under lineage-directed therapy - blinatumomab, CAR-T - is not a rare accident but the expected behaviour of a plastic cell, and is already documented in KMT2A-r disease.
  • Stratifying trials by partner gene alone mixes HOXA-high and HOXA-low, PHD-retaining and PHD-lacking, reciprocal-positive and reciprocal-negative disease. Real efficacy differences between these strata would be invisible and would appear instead as unexplained heterogeneity in response.
  • A one-size menin inhibitor would most plausibly miss: minor-BCR PHD-retaining fusions, reciprocal-driven chromatin states, and HOXA-low infant disease if HOXA collapse is the relevant pharmacodynamic mechanism.
Still open
  1. Does KMT2A breakpoint position predict survival after adjustment for age at diagnosis, WBC and partner gene? Requires breakpoint typing on an Interfant-06 or COG-scale prospective cohort - the single highest-value missing analysis in this beat.
  2. Does breakpoint position predict menin-inhibitor response? The banked samples from revumenib and ziftomenib trials could answer this retrospectively today.
  3. If the only protein-level difference between an intron-9 and an intron-11 break is on the reciprocal fusion, does reciprocal-negative disease (e.g. spliced KMT2A::MLLT10) lose the breakpoint-position effect entirely? That would be a clean test of the mechanism.
  4. KMT2A::MLLT3 is the one partner common in both ALL and AML. Does breakpoint position (not cell of origin) segregate its lineage? Meyer's data do not report an ALL-vs-AML breakpoint split within MLLT3, and this is directly answerable from the existing DCAL database.
Every source behind this entry (5)
  1. E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    Partner identity strongly determines lineage and age distribution in 3401 KMT2A-r patients, and breakpoint position differs significantly by partner, age and therapy-relatedness, associating with HOXA signature; whether breakpoint position affects outcome is stated to be unclear.

    The KMT2A recombinome of acute leukemias in 2023. · Meyer C, Larghero P, Almeida Lopes B, et al. · Leukemia 2023
    PMID 37019990DOI
    What this evidence cannot show

    Single-centre referral cohort (DCAL Frankfurt) accruing 2003-2022 from centres that chose to send samples; partner frequencies are therefore not population incidences. Pre-screening method varied (cytogenetics, FISH, RT-PCR, RNA-seq) and 2702/3401 were typed by LD-PCR (~95% discovery) versus 696 by NGS (~100%). Subgroup breakpoint comparisons are chi-square tests against the cohort mean with no reported multiple-testing correction across ~13 partners x 3 age classes.

    Evidence that argues against this (1)
    • The 2013 edition of the same series (n=1590) reported the same seven dominant partners, so the partner spectrum is internally stable - but that is replication within one centre, not independent replication. Meyer C et al. 2013
  2. E2moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    Paediatric t(4;11) ALL splits into infants with low HOXA expression and MLL breakpoints in introns 11/12 versus non-infants with high HOXA expression and breakpoints in introns 9/10.

    Two independent gene signatures in pediatric t(4;11) acute lymphoblastic leukemia patients · Trentin L, Giordan M, Dingermann T, Basso G, Te Kronnie G, Marschalek R · European Journal of Haematology 2009
    PMID 19558506DOI
    What this evidence cannot show

    Small microarray cohort, pairwise normalisation against only 3 normal bone-marrow samples, and age is confounded with breakpoint by construction. Same senior author as the recombinome series, so not an independent line.

    Evidence that argues against this (2)
    • The HOXA split may be the partner speaking, not the breakpoint. KMT2A::SEPTIN6, a partner the recombinome places preferentially in intron 11, actively SUPPRESSES HOXA7 and HOXA9 in the only cell-line model of that fusion. Since partner identity and breakpoint preference are correlated across the cohort, an infant HOXA-low / non-infant HOXA-high contrast built on breakpoint position alone cannot separate a breakpoint effect from a partner-mix effect, and the microarray cohort here is too small to stratify by partner. Nagel S et al. 2026
    • High HOXA expression is not specific to any KMT2A breakpoint, or to KMT2A at all. CALM-AF10 (PICALM::MLLT10) T-ALL, which has no KMT2A break, overexpresses HOXA5, HOXA9, HOXA10 and MEIS1 at levels comparable to MLL-translocated acute leukaemia. HOXA status is therefore a readout of the effector complex the fusion assembles, and using it as the phenotype that different breaks produce assumes the specificity the comparison is meant to demonstrate. Dik WA et al. 2005
  3. E3weak confidence 2 of 5: Preliminary, small, indirect, or not yet peer reviewed.paper

    Among 545 children <=24 months with acute leukaemia, poor-outcome KMT2A-rearranged patients preferentially displayed breakpoints within MLL intron 11; but breakpoint location was determined in only 30 cases and age <=6 months was independently prognostic (P<0.001).

    The distribution of MLL breakpoints correlates with outcome in infant acute leukaemia · Emerenciano M, Meyer C, Mansur MB, Marschalek R, Pombo-de-Oliveira MS · British Journal of Haematology 2013
    PMID 23432364DOI
    What this evidence cannot show

    n=30 breakpoint-typed patients; retrospective; single-country cohort (Brazil) with 60.5% overall survival on a treatment backbone that differs from current Interfant/COG protocols; breakpoint effect not adjusted for age, WBC or partner gene; authors describe it as first evidence. Not replicated prospectively.

    Evidence that argues against this (1)
    • The largest available dataset (n=3401) declines to confirm a breakpoint-outcome relationship, writing that whether these findings have an impact on clinical outcome is yet unclear. Meyer C et al. 2023
  4. E4moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    A CRISPR/Cas9-engineered t(4;11) in primary human haematopoietic stem/precursor cells retains plasticity to differentiate into either the myeloid or the lymphoid lineage.

    PMID 41145673DOI
    What this evidence cannot show

    In vitro differentiation assay over 5-6 months with IL-7 supplementation; no in vivo leukaemia readout, so this speaks to lineage potential of a translocated cell rather than to the lineage of the eventual leukaemia. Single laboratory, very recent.

    Evidence that argues against this (1)
    • Lineage potential measured in a dish does not predict the leukaemia, or whether there is one. A germline Mll-AF4 knock-in mouse showed the same mixed lymphoid/myeloid deregulation - bone-marrow cells forming unique mixed pro-B/myeloid colonies in IL-7 plus Flt3 ligand - and then went on to develop mostly B-cell LYMPHOMAS after prolonged latency, prompting the authors to conclude that Mll-AF4-induced lymphoid/myeloid deregulation alone is insufficient to produce malignancy. A 5-6 month in vitro differentiation assay is compatible with no leukaemia and with the wrong disease. Chen W et al. 2006
  5. E5strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    Retention of PHD fingers in an MLL fusion protein blocks transformation, establishing that the domain boundary at the 3' end of the BCR is functionally decisive.

    The PHD fingers of MLL block MLL fusion protein-mediated transformation · Muntean AG, Giannola D, Udager AM, Hess JL · Blood 2008
    PMID 18796627DOI
    What this evidence cannot show

    Murine retroviral immortalisation assay with MLL-AF9 only.

    Evidence that argues against this (1)
    • The boundary is decisive at PHD3, which is nowhere near the 3' end of the BCR. Inserting PHD1 and PHD2 into MLL-ENL had no effect at all on haematopoietic stem cell immortalisation; only PHD3, alone or inside the full cassette, blocked it, by restoring CYP33 and HDAC1 recruitment and converting the fusion from a constitutive transactivator to a repressor. PHD3 is encoded by exons 14-16, so this experiment does not make the intron 10 / exon 11 boundary functionally decisive - it makes a boundary three exons further 3' decisive, and every major-BCR break is 5' of both. Chen J et al. 2008
Connected to