Oncofusion

start / graph / fusion-etv6-runx1

Fusionestablished confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.Q11-comparative

ETV6::RUNX1 (TEL-AML1), t(12;21)(p13;q22)

also called TEL-AML1 · TEL::AML1 · ETV6-RUNX1 · ETV6-CBFA2 · t(12;21) ALL

In about a quarter of children with the commonest childhood leukaemia, two chromosomes swap ends so that a gene that normally silences genes gets bolted onto a gene that normally switches blood genes on, and the hybrid silences the genes it should activate. The first half of this accident usually happens before birth, and most children who then develop the leukaemia are cured.

The same thing, for a scientist

t(12;21)(p13;q22) fuses the first five exons of ETV6 (P41212) — the SAM/pointed self-association module plus the central N-CoR-binding repression domain — to almost the entire coding region of RUNX1 (Q01196), retaining the Runt DNA-binding/CBFB-heterodimerisation domain, the transactivation domain and the C-terminal VWRPY motif [E1]. The chimera therefore binds RUNX1 consensus sites with a normal Runt domain but delivers ETV6-borne corepressor complexes instead of coactivators, converting selected RUNX1 target genes from activated to repressed, and additionally blunting SMAD3-dependent TGF-beta growth restraint in early B progenitors [E3][E4]. The fusion is generated in utero in a fetal/neonatal B-lineage progenitor and is detectable in cord blood at roughly 100-fold the lifetime incidence of the leukaemia, so it is an initiating but grossly insufficient lesion requiring postnatal secondary hits — most characteristically deletion of the second, untranslocated ETV6 allele [E2][E5]. Clinically it defines the most favourable common subtype of paediatric B-ALL (5-year EFS ~89%, OS ~97% on DFCI 95-01) with a characteristic late-relapse pattern and no fusion-directed drug [E6].

What it looks like
loading 1H9D…
PDB 1H9D ↗
shape

This is real experimental data — atoms placed by crystallography or cryo-EM, deposited by the people who did the experiment.

I have never read one of these pictures before

A protein is one long chain of amino acids that folds up. The ribbon traces that chain from one end to the other; it is not a wire, it is a simplification of where the backbone runs. Drag to rotate, scroll to zoom.

Domains are the stretches of chain that do a particular job, and they are what the coloured blocks above the picture mark out. Click one and the view flies to it. In this project the important question about any domain is usually whether the fusion still has it — the labels say lost where it does not.

Anything in yellow sticks is a bound small molecule, which is usually the drug.

Go deeper
What is kept and what is thrown away1 deeper

The break is remarkably stereotyped, which is why this fusion is a clean comparative reference point.

Kept from ETV6 (5-prime partner, exons 1-5):

  • the SAM / pointed (PNT) domain, a self-associating module that polymerises head-to-tail into an open-ended helical polymer (PDB 1JI7, 1LKY) 7;
  • the central repression domain, the surface that recruits the nuclear receptor corepressor N-CoR/NCOR1 and through it histone deacetylase activity 3.

Lost from ETV6: the ETS DNA-binding domain and the C-terminus. This matters: the fusion no longer reads ETS sites, so ETV6's own target repertoire is not what the fusion occupies.

Kept from RUNX1 (3-prime partner, essentially the whole protein): the Runt domain — which both contacts the TGYGGT consensus DNA and heterodimerises with CBFB — plus the transactivation domain and the C-terminal VWRPY motif that docks TLE/Groucho corepressors 1.

Lost from RUNX1: only the short exon-1-encoded N-terminus.

So the composite protein has RUNX1's address book and ETV6's silencing machinery. That asymmetry is the mechanism, and it is the opposite arrangement to KMT2A fusions, where the 5-prime partner supplies the address and the 3-prime partner supplies an activating elongation engine.

Why the SAM polymer is not a footnote

ETV6's SAM domain forms a genuine helical polymer, not a dimer, and polymerisation is required for ETV6-dependent transcriptional repression and for the transforming activity of ETV6-kinase fusions 7. In ETV6::RUNX1 the polymer is retained while the ETS domain is not, which predicts that the fusion decorates RUNX1 sites with an extended, multivalent corepressor scaffold rather than a single repressor molecule.

That is a structural story adjacent to — but not identical with — the FG-repeat condensates of NUP98 fusions, and the graph keeps the two mechanistically distinct: NUP98 uses low-complexity, multivalent phase separation; ETV6 uses a defined, ordered, stoichiometric polymer interface with a solved structure.

Mechanism: a corrupted transcription factor, not an enzyme

The molecular class is transcription-factor repression corruption. RUNX1 normally sits in the core-binding factor (CBF) complex with CBFB and activates myeloid and lymphoid differentiation genes; ETV6::RUNX1 occupies the same sites and recruits N-CoR/HDAC, so a subset of RUNX1 targets flip from on to off 3. This is a dominant-interfering plus neomorphic activity: it is not simply RUNX1 loss of function, because the ETV6 moiety adds a corepressor-recruiting surface that wild-type RUNX1 does not have.

A second, separable effect is on growth-factor signalling. ETV6::RUNX1 binds SMAD3 and compromises its transactivation, so expressing cells become relatively resistant to TGF-beta-mediated growth inhibition and outcompete their neighbours when TGF-beta is present; in human cord-blood progenitors the fusion expands a candidate preleukaemic population with a CD34+CD38-CD19+ phenotype 4. That gives a concrete cellular explanation for a silent, years-capable preleukaemic clone, and a concrete route by which postnatal infection-driven cytokine exposure could promote it.

There is no catalytic activity anywhere in this fusion. No kinase, no methyltransferase, no demethylase, no ligand pocket. That is precisely why it has resisted targeted therapy while BCR::ABL1 and PML::RARA did not.

Natural history: made before birth, needs a second hit

Three independent lines of evidence put the translocation prenatally: concordant leukaemia in monozygotic twins sharing a clonotypic fusion sequence, retrospective detection of the fusion in archived neonatal blood spots, and direct detection in cord blood 25.

The frequency argument is the important one. Screening unselected cord bloods finds ETV6-RUNX1 (and AML1-ETO) at roughly 100-fold the cumulative risk of the corresponding leukaemia, in cells of the matching lineage, at 1e-4 to 1e-3 of mononuclear cells — i.e. already clonally expanded 5. Twin concordance is only about 5%, and latency is years 2. So the fusion is necessary but far from sufficient, and the rate-limiting events are postnatal.

The commonest cooperating lesion is deletion of the remaining wild-type ETV6 allele, plus gains of the der(21) and of RUNX1, and an assortment of B-lineage regulator deletions. Notably CDKN2A deletion is less frequent here at diagnosis (~15%) than in most other subtypes (61% in t(9;22), 78-89% in TLX-rearranged T-ALL), showing the cooperating-lesion spectrum at presentation is subtype-specific rather than generic 8. The relapse picture is different: deletion of CDKN2A/B or CCNC rises to the majority of ETV6-RUNX1 relapses 8. Additional der(12)t(12;21) deletions and duplications affecting the reciprocal RUNX1-ETV6 product occur in ~12% of cases and were all seen in children under 10 with good outcome 9.

Clinical phenotype, age and prognosis

Disease: B-cell precursor ALL, essentially exclusively. An RT-PCR survey of paediatric AML found 0/29 positive 10; one adult AML-M5a with a FISH- and RT-PCR-confirmed ETV6::RUNX1 has since been reported, so restriction to B-lineage ALL is near-complete rather than absolute (see the 2026-09-19 correction below).

Age: peak 2-5 years, the classic common-ALL age peak; frequency falls off sharply in adolescents and the fusion is rare in adults.

Frequency: ~25% of childhood B-cell precursor ALL 19. Cytogenetically cryptic — invisible on conventional karyotype, requiring FISH, RT-PCR or RNA-seq, which is why the commonest rearrangement in paediatric cancer was only cloned in 1995 1.

Outcome: the most favourable of the common subtypes. On the prospective DFCI ALL Consortium 95-01 study, 5-year EFS was 89% vs 80% for TEL/AML1-negative B-precursor ALL, and 5-year OS 97% vs 89% 6. Modern MRD-directed regimens do better still. Two clinically important caveats: (i) in the same study's multivariable model, age and presenting white count — not fusion status — were the independent predictors, so part of the apparent benefit is confounded by this subtype's favourable age/count distribution 6; (ii) relapses, when they occur, are characteristically late, often off-therapy and years out, and can be clonally distinct, arising from the persistent preleukaemic ancestor rather than the diagnostic clone.

Druggability: the honest answer is undrugged

There is no ETV6::RUNX1-directed agent, approved or in late-phase trials. The reasons are structural:

  1. No active site. Both moieties are protein-protein and protein-DNA interaction modules. There is no ATP pocket like ABL1's and no ligand-binding pocket like RARA's.
  2. The retained RUNX1 Runt domain is the wild-type surface. Anything that blocks Runt-DNA or Runt-CBFB binding hits wild-type CBF as well, and CBF is required for definitive haematopoiesis. The available structures (PDB 1H9D, 1E50, 1IO4) are of the normal Runt-CBFB-DNA clamp, not of the fusion 11.
  3. The exploitable surfaces are all indirect: the N-CoR/HDAC axis (hence historical interest in HDAC inhibitors, none subtype-approved), the SAM polymerisation interface, and the restored-TGF-beta-sensitivity idea implied by the SMAD3 work 34.

In practice this fusion is the case study for curing a leukaemia with chemotherapy and immunotherapy rather than by drugging the driver. The current clinical agenda is de-escalation — giving these children less cytotoxic therapy without losing the 97% survival — not target discovery.

Correction, 2026-09-19: B-lineage restriction is near-complete, not strict

Withdrawn. This entry previously said, in the cooperating-lesion section, that an "RT-PCR survey of paediatric AML found 0/29 positive, consistent with strict B-lineage restriction [E10]".

Why. Jeon et al. report a 66-year-old man with AML-M5a in whom targeted RNA sequencing detected ETV6::RUNX1, confirmed by multiplex RT-PCR and by FISH showing t(12;21)(p13;q22) in 10/200 nuclei of the post-induction marrow, with the signal disappearing at morphological remission. The blasts were myeloperoxidase- and non-specific-esterase-positive and negative for CD2, CD5, CD7, CD10, CD19, CD20 and CD22, and the co-occurring sequence variants were myeloid-type (PMID 36045062, doi 10.3343/alm.2023.43.1.92). One confirmed case does not make this a myeloid fusion - it remains overwhelmingly a B-precursor ALL lesion - but it does mean the word strict was doing work the evidence cannot support.

What the entry now says. The 0/29 paediatric AML survey bounds the frequency of ETV6::RUNX1 in myeloid disease rather than excluding it, and restriction to B-lineage ALL is described as near-complete. E10's claim sentence was already hedged ("supporting restriction") and is unchanged; only the body prose was overstated.

A separate point worth recording rather than correcting: TEL-AML1-positive B-precursor ALL expresses myeloid antigens more often than fusion-negative BCP-ALL, so even within the B lineage the immunophenotype is not cleanly lymphoid.

Analogy: the same prenatal origin as infant KMT2A::AFF1, and the opposite penetrance

Closest analogue, and the axis. fusion-kmt2a-aff1, on the axis of shared cell of origin and shared timing - both fusions are generated before birth in a fetal or neonatal B-lineage progenitor, and both are recoverable from archived neonatal blood spots. On every other axis in this beat they are unalike, which is what makes the one shared axis worth isolating.

What the analogy predicts, and the test nature already ran. If a fusion is prenatal and a monochorionic twin pair shares a circulation, the clone reaches both fetuses. Concordance then measures penetrance directly: whatever fraction of co-twins develops leukaemia is the fraction for which the fusion plus the shared intrauterine environment was enough 12. That is an experiment nobody could ethically design, and the two subtypes give opposite answers - roughly 5% concordance across childhood ALL, where ETV6::RUNX1 is the largest single contributor, against a rate believed close to unity with very short latency in infant KMT2A-rearranged disease. The graph assembles the numbers in epi-twin-concordance-infant-kmt2a-leukemia.

So the axis they share and the axis they differ on are both informative:

ETV6::RUNX1KMT2A::AFF1 (infant)
originprenatalprenatal
covert clones at birth~100-fold the lifetime incidence of the leukaemia 5far closer to the disease incidence
cooperating lesionsrequired; commonest is deletion of the second, untranslocated ETV6 allele 25close to none - a quiet genome
latencyyears, with a characteristic late-relapse patternmonths
twin concordance~5% (childhood ALL overall)believed near-unity, with a published exception
outcomemost favourable common subtype, 5-year EFS ~89% 6among the worst

Where it breaks, and it breaks in the direction that matters. The contrast is usually written as categorical - one fusion needs second hits, the other does not. It is not categorical. The first published discordant infant KMT2A twin pair had a fusion-positive clone in the healthy co-twin that cleared, temporally associated with viral cytopenias (counter-evidence on 12). So the infant figure is not unity, an immune-clearance step exists in the KMT2A natural history too, and the difference between the two subtypes is a steep gradient in penetrance rather than the presence or absence of a second-hit requirement. Two further limits: twin ascertainment is biased toward concordant pairs, and the ~5% figure is for childhood ALL as a whole rather than for ETV6::RUNX1 specifically.

A second analogue on a different axis: fusion-dux4-igh. ETV6::RUNX1 and DUX4-rearranged B-ALL are the two good-prognosis, non-infant B-ALL subtypes of this beat, and in both the cooperating event is loss of function of an ETS-family transcription factor - the untranslocated ETV6 allele here, ERG in DUX4-rearranged disease. The disanalogy is developed in that entry and is the useful half: one of those ETS losses is a stochastic postnatal second hit and the other is written by the driver itself.

Facts and figures
gene5ETV6
gene3RUNX1
cytobandt(12;21)(p13;q22)
frequency noteThe adult figure is an order-of-magnitude literature estimate rather than a value verified in a single cited cohort; the paediatric 25% and the paediatric-AML 0% are cited to [E9] and [E10] respectively.
prognosisThe most favourable common paediatric B-ALL subtype: 5-year EFS 89% and OS 97% on DFCI ALL Consortium 95-01 versus 80% and 89% for fusion-negative B-precursor ALL, though fusion status was not an independent predictor once age and white-cell count were modelled, and relapses are characteristically late [E6].
uniprot5P41212
uniprot3Q01196
breakpoints noteRUNX1 exon numbering differs between the RUNX1a/b/c transcript conventions; exon 2 here corresponds to the first large coding exon containing the start of the Runt domain in the AML1b numbering used by Golub et al. Treat the number as a convention label, not a genomic coordinate.
PDB noteNo structure of the ETV6::RUNX1 fusion exists. 1H9D / 1E50 / 1IO4 are the wild-type RUNX1 Runt domain with the CBFB core domain on DNA — the retained 3-prime module, and the surface that cannot be drugged selectively because it is shared with wild-type RUNX1. 1JI7 and 1LKY are the isolated ETV6/TEL SAM domain polymer, the retained 5-prime module. Everything between these fragments, including the junction itself, is unmodelled.
druggability classundrugged
stage of hematopoiesisFetal/neonatal B-lineage progenitor; the candidate preleukaemic cell carries a CD34+CD38-CD19+ early-B phenotype and is generated in utero, with overt disease presenting years later at age 2-5 [E4][E5]
age peak2-5 y
developmental order rank6
frequency
pediatric_aml_pct0.000pediatric_aml_pct: 0adult_b_all_pct2.000adult_b_all_pct: 2pediatric_b_all_pct25.0pediatric_b_all_pct: 25
diseases
[
  "pediatric B-cell precursor ALL",
  "adolescent B-ALL (uncommon)",
  "adult B-ALL (rare)"
]
frequency — the numbers behind the chart
{
  "pediatric_b_all_pct": 25,
  "adult_b_all_pct": 2,
  "pediatric_aml_pct": 0
}
breakpoints
[
  {
    "id": "bp-etv6-exon5-runx1-exon2",
    "gene5_exon_last": 5,
    "gene3_exon_first": 2,
    "frequency_pct": 90,
    "population": "pediatric B-cell precursor ALL",
    "domains_retained_5": [
      "SAM/pointed (PNT) self-association domain",
      "central N-CoR-binding repression domain"
    ],
    "domains_lost_5": [
      "ETS DNA-binding domain",
      "ETV6 C-terminus"
    ],
    "domains_retained_3": [
      "Runt domain (DNA binding + CBFB heterodimerisation)",
      "transactivation domain",
      "C-terminal VWRPY / TLE-Groucho motif",
      "nuclear matrix targeting signal"
    ],
    "domains_lost_3": [
      "RUNX1 exon-1-encoded N-terminal segment"
    ],
    "phenotype_notes": "The dominant in-frame junction: ETV6 exons 1-5 joined to almost the entire RUNX1 coding region [E1]. A minor variant joins ETV6 exon 5 to RUNX1 exon 3. Breakpoint position has no established prognostic effect, in contrast to KMT2A, where break position is actively debated."
  }
]
PDB IDS
[
  "1H9D",
  "1E50",
  "1IO4",
  "1JI7",
  "1LKY"
]
mechanism class
[
  "tf-repression-corruption"
]
How this could bite us back
  • Any agent that blocks the Runt-CBFB interface or Runt-DNA binding will inhibit wild-type RUNX1 and CBFB too; core-binding factor is required for definitive haematopoiesis, so on-target marrow failure is the predictable toxicity. This is the mirror image of the menin story, where the drugged interface has at least a partially separable normal function.
  • De-escalation is the real clinical risk here. Because the multivariable evidence shows age and white count — not the fusion — driving outcome, treating every ETV6::RUNX1 child as low risk would under-treat older or high-count patients [E6].
  • HDAC inhibition, the obvious inference from N-CoR recruitment, is broadly toxic and not fusion-selective, because the corepressor complex is used across the transcriptome.
  • The preleukaemic clone is itself a bite-back: it predates therapy and is not fusion-addicted in the way the overt blast is, so therapy aimed at the diagnostic clone leaves the ancestor behind and the late relapse can be a genetically new leukaemia from the same fusion-positive founder.
Still open
  1. Does the retained ETV6 SAM polymer actually assemble at physiological fusion levels on chromatin in patient blasts, and if so does polymer length set the depth of repression at RUNX1 targets? Directly testable by comparing polymerisation-dead SAM point mutants in an isogenic fusion background.
  2. Late relapses in this subtype can be clonally distinct from diagnosis, implying a persistent preleukaemic reservoir that survives therapy. Can that reservoir be identified prospectively in remission marrow by fusion-plus-ancestral-genotype single-cell assays, and would eradicating it change the late-relapse rate?
  3. If the fusion's growth advantage runs through blunted TGF-beta/SMAD3 signalling, is there a pharmacological way to restore TGF-beta sensitivity selectively in fusion-positive cells, and would it be tolerable given TGF-beta's role in normal haematopoiesis?
  4. Why is this fusion so strictly B-lineage restricted when RUNX1 is required across haematopoiesis and the reciprocal RUNX1::ETV6 product arises in myeloid-competent cells?
  5. Is the ~5% concordance figure right for ETV6::RUNX1 specifically, or is it a childhood-ALL average that this beat has been quoting as a subtype property? The number matters, because the entire latency-and-penetrance contrast with infant KMT2A::AFF1 is built on it, and no source in this graph reports a fusion-stratified twin concordance rate. A registry-level recount of monozygotic twin pairs by molecular subtype - ETV6::RUNX1, DUX4-rearranged, KMT2A-rearranged - is a paper-sized question that needs no new samples.
Every source behind this entry (12)
  1. E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    t(12;21) fuses the helix-loop-helix (SAM/pointed) region of TEL/ETV6 to the DNA-binding and transactivation domains of AML1/RUNX1, was cryptic on karyotype, and was cloned from childhood pre-B ALL.

    Fusion of the TEL gene on 12p13 to the AML1 gene on 21q22 in acute lymphoblastic leukemia · Golub TR, Barker GF, Bohlander SK, Hiebert SW, Ward DC, Bray-Ward P, Morgan E, Raimondi SC, Rowley JD, Gilliland DG · Proc Natl Acad Sci U S A 1995
    PMID 7761424DOI
    What this evidence cannot show

    Two index cases only; the exon-level breakpoint consensus and the ~25% frequency figure come from the larger series that followed, not from this paper.

    Evidence that argues against this (1)
    • The cloned chimera is only part of the lesion as it presents in patients. Raynaud et al. screened 50 children with B-lineage ALL by microsatellite typing, FISH and fusion-transcript analysis: the TEL-AML1 fusion RNA was present in every t(12;21) case, but the reciprocal AML1-TEL transcript in only a subset, and deletion of the non-translocated TEL allele was among the most frequent abnormalities accompanying the translocation - in two patients confined to a subclone, i.e. acquired after it. Allelic loss of TEL is detectable in 15-22% of childhood ALL overall. Describing t(12;21) as 'fusion of TEL to AML1' therefore omits both the variable loss of the reciprocal product and the second-allele deletion that the same event sets up. Raynaud S et al. 1996
  2. E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.review

    ETV6-RUNX1 arises prenatally as a monoclonal initiating event, demonstrated by concordant clonotypic fusions in monozygotic twins and by detection in archived neonatal blood spots, with twin concordance only about 5%.

    ETV6-RUNX1 Acute Lymphoblastic Leukaemia in Identical Twins · Ford AM, Greaves M · Adv Exp Med Biol 2017
    PMID 28299660DOI
    What this evidence cannot show

    A review by the group that generated most of the primary data, so not an independent appraisal; twin studies are necessarily small, and the infection-promotion arm of the model remains a hypothesis rather than a demonstrated human mechanism.

    Evidence that argues against this (1)
    • The backtracking argument requires ETV6-RUNX1-positive cells to be present at birth at a detectable level, and the largest direct test does not find them. Lausten-Thomsen et al. screened mononuclear cells from 1417 umbilical cord bloods isolated within 24 hours of birth using a highly sensitive real-time RT-PCR: 14 samples were positive in the first PCR run at levels below 10^-5, specific hybridisation confirmed only 9, and none survived re-analysis of stored mRNA or flow-sorting of CD19+, CD8+ and CD19-/CD8- fractions (mean 18 x 10^6 cells sorted per sample). Their conclusion is that the level and/or frequency of ETV6-RUNX1-positive cells is markedly lower than previous studies suggested. Published newborn-prevalence estimates for this fusion differ by orders of magnitude between assays, so the neonatal-blood-spot premise is assay-dependent rather than settled, and the 5% twin concordance figure rests on a handful of pairs. Lausten-Thomsen U et al. 2011
  3. E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    The central region of TEL retained in TEL-AML1 binds the nuclear receptor corepressor N-CoR, and this interaction is required for TEL-mediated transcriptional repression, giving the fusion a repressive activity that wild-type AML1 lacks.

    Recruitment of the nuclear receptor corepressor N-CoR by the TEL moiety of the childhood leukemia-associated TEL-AML1 oncoprotein · Guidez F, Petrie K, Ford AM, Lu H, Bennett CA, MacGregor A, Hannemann J, Ito Y, Ghysdael J, Greaves M, Wiedemann LM, Zelent A · Blood 2000
    PMID 11001911
    What this evidence cannot show

    Co-immunoprecipitation and reporter assays in transfected cell lines with overexpressed constructs; does not establish N-CoR occupancy at endogenous RUNX1 targets in patient blasts. No DOI is registered for this 2000 Blood article, so the PubMed record is the citable anchor.

    Evidence that argues against this (1)
    • Repression of RUNX1 targets by ETV6::RUNX1 is not uniform: expression studies find both repressed and activated genes and a substantially neomorphic rather than purely repressive programme, so N-CoR recruitment cannot be the whole mechanism. Monovich AC et al. 2024
  4. E4strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    TEL-AML1 binds SMAD3 and impairs its transactivation, making early B progenitors relatively resistant to TGF-beta growth inhibition, and expands a candidate CD34+CD38-CD19+ preleukaemic population in human cord-blood progenitors.

    The TEL-AML1 leukemia fusion gene dysregulates the TGF-beta pathway in early B lineage progenitor cells · Ford AM, Palmi C, Bueno C, Hong D, Cardus P, Knight D, Cazzaniga G, Enver T, Greaves M · J Clin Invest 2009
    PMID 19287094DOI
    What this evidence cannot show

    Relies on an inducible murine pro-B line, a transgenic mouse and transduced cord-blood cells with non-physiological fusion levels; the preleukaemic-stem-cell phenotype is inferred from an in vitro growth advantage, not from serial transplantation of patient-derived preleukaemic cells.

    Evidence that argues against this (1)
    • A competing account of the same preleukaemic phenotype does not invoke TGF-beta and locates the expansion elsewhere. Tsuzuki et al. transplanted TEL-AML1-transduced bone-marrow stem cells syngeneically into mice: what accumulated was primitive c-kit-positive MULTIPOTENT progenitors plus a modest rise in myeloid colony-forming cells, not a B-restricted CD19+ compartment; in B lymphopoiesis the fusion produced an increase in early pro-B cells but a differentiation deficit beyond that stage, with reduced surrogate light-chain lambda5 and reduced IL-7 receptor expression, reduced clonogenicity in IL-7 and reduced marrow B-cell output. A model in which the fusion expands progenitors by relieving TGF-beta growth restraint is hard to reconcile with fusion-expressing progenitors that respond less well to their own growth factor. Tsuzuki S et al. 2004
  5. E5weak confidence 2 of 5: Preliminary, small, indirect, or not yet peer reviewed.paper

    TEL-AML1 is detectable in unselected cord blood at roughly 100-fold the cumulative risk of the corresponding leukaemia, in lineage-restricted and already clonally expanded cells, establishing it as insufficient on its own.

    Chromosome translocations and covert leukemic clones are generated during normal fetal development · Mori H, Colman SM, Xiao Z, Ford AM, Healy LE, Donaldson C, Hows JM, Navarrete C, Greaves M · Proc Natl Acad Sci U S A 2002
    PMID 12048236DOI
    What this evidence cannot show

    The absolute prevalence has been contested on technical grounds: nested RT-PCR at this sensitivity is vulnerable to contamination and to amplification from very few cells, and later attempts have reported lower rates. The 100-fold-excess argument survives even at substantially lower prevalence estimates, but the exact number should not be quoted as settled. Confidence lowered from 'strong' to 'weak' (coordinator, 2026-09-18, confidence-split audit). This project's own entry `epi-fusion-positive-newborns-versus-leukemia-risk` records that the Mori 2002 cord-blood frequency is **unreproduced**: Lausten-Thomsen et al. screened 1,417 cord bloods and confirmed none. Two entries carried the paper at 'strong' while two carried it at 'weak', and the graph's own position is the latter. Confidence attaches to how well a claim is supported, not to how well known a paper is.

    Evidence that argues against this (1)
    • Independent replication of the cord-blood prevalence has been inconsistent, with some groups reporting markedly lower or undetectable rates and attributing the discrepancy to assay sensitivity and contamination. We could not identify a single definitive refutation paper to cite, so this is recorded as an acknowledged methodological dispute rather than a specific counter-citation; the review below discusses the prevalence estimates. Ford AM et al. 2017
  6. E6established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    In a prospective trial, TEL/AML1-positive children had 5-year EFS 89% and OS 97% versus 80% and 89% for fusion-negative B-precursor ALL, but fusion status was not an independent predictor in multivariable analysis.

    Prospective analysis of TEL/AML1-positive patients treated on Dana-Farber Cancer Institute Consortium Protocol 95-01 · Loh ML, Goldwasser MA, Silverman LB, Poon WM, Vattikuti S, Cardoso A, Neuberg DS, Shannon KM, Sallan SE, Gilliland DG · Blood 2006
    PMID 16493009DOI
    What this evidence cannot show

    Median follow-up only 5.2 years, which systematically under-counts the late relapses characteristic of this subtype; therapy is a 1996-2000 regimen without MRD-directed stratification, so absolute numbers are not those of a contemporary protocol. The paper itself reports that fusion status lost independent significance once age and leukocyte count were modelled.

    Evidence that argues against this (1)
    • A smaller Hong Kong cohort found no significant EFS or OS advantage for TEL/AML1-positive children (EFS 60% vs 56%, OS 100% vs 72%, both non-significant) and explicitly argued against treatment de-escalation on the basis of the fusion. Tsang KS et al. 2001
  7. E7established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

    The SAM domain of TEL/ETV6 polymerises into an open-ended helical polymer, and this polymerisation is required for TEL-mediated transcriptional repression.

    What this evidence cannot show

    The structure is of the isolated SAM domain of wild-type TEL (1LKY carries a solubilising mutation), not of the ETV6::RUNX1 fusion; whether the polymer forms at physiological fusion concentrations on chromatin is unverified. Verified by fetching the RCSB entry record rather than by reading the primary paper.

    Evidence that argues against this (1)
    • Polymerisation is necessary but far from sufficient, and in the fusion it is one of three separable requirements, two of which come from the RUNX1 side. Uchida et al. tested a deletion series of TEL-AML1 on the AML1-responsive IL-3 promoter in lymphoid cells and identified three distinct domains required for repression: the TEL HLH/pointed (SAM) motif, the RUNX1 Runt homology domain, and the 74 amino acids immediately downstream of the Runt domain. Attributing the fusion's repressive polarity to the SAM polymer over-weights the ETV6 contribution, because two of the three necessary elements belong to the moiety that ETV6 is supposed merely to redirect. Uchida H et al. 1999
  8. E8strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    At diagnosis, CDKN2A deletion is less frequent in ETV6-RUNX1 ALL (about 15%) than in most other subtypes (61% in t(9;22), 78-89% in TLX-rearranged T-ALL), showing that the cooperating-lesion spectrum at presentation is subtype-specific.

    A comprehensive analysis of the CDKN2A gene in childhood acute lymphoblastic leukemia reveals genomic deletion, copy number neutral loss of heterozygosity, and association with specific cytogenetic subgroups · Sulong S, Moorman AV, Irving JAE, Strefford JC, Konn ZJ, Case MC, Minto L, Barber KE, Parker H, Wright SL, Stewart ARM, Bailey S, Bown NP, Hall AG, Harrison CJ · Blood 2008
    PMID 18838613DOI
    What this evidence cannot show

    Retrospective single-consortium cohort; detection combined FISH, SNP array and aCGH with differing resolution, and subgroup sizes for the rarer cytogenetic classes are small.

    Evidence that argues against this (1)
    • The 15% figure is a diagnosis-timepoint number and does not survive to relapse. van Delft et al. compared genome-wide copy-number profiles at presentation and at matched relapse (12-119 months later) in 21 ETV6-RUNX1 patients: deletion of CDKN2A/B or CCNC (6q16.2-3) or both rose from 38% at presentation to 76% at relapse, and the relapse clone derived from either a major or a minor presentation subclone, with FISH backtracking identifying a matching minor subclone at diagnosis about 10 years earlier. Cell-cycle-gene deletion is therefore the majority state in the clone that actually kills, so the low figure describes the sampled timepoint rather than the cooperating-lesion spectrum of the clone that causes treatment failure. van Delft FW et al. 2011
  9. E9strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    ETV6-RUNX1 occurs in 25% of childhood B-cell precursor ALL, and secondary abnormalities of the der(12)t(12;21) (deletion 8%, duplication 4% of 247 cases) occur in children under 10 with good outcome.

    Abnormalities of the der(12)t(12;21) in ETV6-RUNX1 acute lymphoblastic leukemia · Al-Shehhi H, Konn ZJ, Schwab CJ, Erhorn A, Barber KE, Wright SL, Gabriel AS, Harrison CJ, Moorman AV · Genes Chromosomes Cancer 2012
    PMID 23077088DOI
    What this evidence cannot show

    FISH and microarray on a referral cytogenetics cohort; good outcome is stated descriptively with no formal survival comparison against ETV6-RUNX1 cases lacking these der(12) changes, and the functional consequence for the reciprocal RUNX1-ETV6 product is hypothesised, not tested.

    Evidence that argues against this (1)
    • The 'good outcome' attached to secondary abnormalities is contradicted for the commonest of them, and the frequencies are far higher on FISH. Attarbaschi et al. screened 372 children on two consecutive Austrian trials, found 94 (25%) TEL/AML1-positive, and by interphase FISH detected TEL deletion in 52 (55%), trisomy 21 in 13 (14%) and an additional der(21)t(12;21) in 14 (15%). On univariate analysis, TEL/AML1-positive patients with deletion of the non-translocated TEL allele had a WORSE prognosis than those without it (P=0.034). Secondary changes in this subtype are therefore neither as rare as the 8% and 4% quoted here nor uniformly prognostically neutral. Attarbaschi A et al. 2004
  10. E10moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    TEL/AML1 was undetectable in 29 evaluable paediatric AML samples, supporting restriction of this fusion to B-lineage ALL.

    Lack of TEL/AML1 fusion in pediatric AML: further evidence for lineage specificity of TEL/AML1 · Loh ML, McLean TW, Buckley JD, Howells W, Gilliland DG, Smith FO · Leuk Res 1998
    PMID 9652733DOI
    What this evidence cannot show

    n = 29 is far too small to exclude rare myeloid occurrence; 0/29 is statistically compatible with a true frequency of up to ~10%. The lineage-restriction claim really rests on the accumulated absence of this fusion from large modern AML sequencing cohorts, not on this study alone.

    Evidence that argues against this (1)
    • A documented counter-example exists. Jeon et al. report a 66-year-old man with AML-M5a - 63% monoblasts, myeloperoxidase- and non-specific-esterase-positive, negative for CD2, CD5, CD7, CD10, CD19, CD20 and CD22 - in whom a targeted RNA-sequencing panel detected the ETV6::RUNX1 fusion, confirmed by multiplex RT-PCR and by FISH showing t(12;21)(p13;q22) in 10/200 nuclei of the post-induction marrow, with the FISH signal disappearing at morphological remission; the co-occurring sequence variants were myeloid-type. The fusion is therefore not absolutely confined to B-lineage ALL, and 0/29 in a paediatric AML survey bounds its frequency in myeloid disease rather than excluding it. Jeon SJ et al. 2023
  11. E11established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

    Crystal structures of the RUNX1 Runt domain in complex with CBFB and DNA exist (PDB 1H9D, 1E50, 1IO4) and describe a DNA-induced molecular clamp, but no structure of the ETV6::RUNX1 fusion protein has been solved.

    What this evidence cannot show

    Verified by fetching the RCSB entry record. It covers only the isolated Runt domain plus the CBFB core domain on a short DNA duplex, i.e. a fragment of the fusion's 3-prime moiety and none of its 5-prime moiety.

    Evidence that argues against this (1)
    • The constructs behind those coordinates are engineered in a way that changes the property being measured, and the engineering removes exactly what the fusion replaces. Wolf-Watz et al. showed that a Runt-domain fragment of AML1 soluble and stable enough for structural work (residues 46-185) required a double cysteine-to-serine substitution and 200 mM MgSO4, that the 45 N-terminal residues of AML1 are unstructured, and that REMOVING that N-terminal segment ENHANCES sequence-specific DNA binding; NMR titration with a 14-bp cognate duplex showed the Runt domain itself changes structure on binding. In ETV6::RUNX1 those N-terminal residues are precisely what the ETV6 moiety replaces, so the isolated Runt/CBFB/DNA 'molecular clamp' may not describe the fusion's DNA engagement even qualitatively - the missing fusion structure is compounded by an N-terminal truncation effect inside the structures that do exist. Wolf-Watz M et al. 1999
  12. E12strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.review

    Concordant leukaemia in monozygotic twins is caused by intraplacental spread of one clone through a shared monochorionic circulation rather than by shared germline risk, and the concordance rate therefore measures how much additional change the founding clone still needs: clonotypic fusion sequences are found in archived neonatal blood spots of children who later develop leukaemia, and for childhood ALL and AML the concordance data reveal a variable and often protracted latency and a requirement for further postnatal exposures or genetic events to produce clinical disease.

    Leukemia in twins: lessons in natural history. · Greaves MF, Maia AT, Wiemels JL, Ford AM · Blood 2003
    PMID 12791663DOI
    What this evidence cannot show

    A review, cited here as the map of a literature rather than as a primary measurement; the quantitative contrast this graph relies on (roughly 5% concordance across childhood ALL against a near-unity rate believed for infant KMT2A-rearranged leukaemia) is assembled in epi-twin-concordance-infant-kmt2a-leukemia from several sources and is not a single number reported in this paper. Twin pairs are a selected and small population, monochorionic placentation is required, and ascertainment of concordant pairs is biased upward relative to discordant ones.

    Evidence that argues against this (1)
    • The comparator end of the contrast is softer than it is usually quoted. Concordance of KMT2A-rearranged leukaemia in infant monozygotic twins was believed to be 100% with very short latency - the observation that makes ETV6::RUNX1 look so different - and the first published exception is a pair in which twin A presented at 9 months with MLL-ENL-positive ALL while twin B remained healthy three years later, with a clonal MLL-ENL-positive population demonstrated in twin B's marrow and blood and then cleared, temporally associated with viral-induced cytopenias. If the infant KMT2A rate is not unity, the latency and penetrance gap between the two subtypes is a gradient rather than a categorical difference, and an immune-clearance step belongs in both natural histories. Chuk MK et al. 2009
Connected to