Oncofusion

start / graph / fusion-kmt2a-ell

Fusionstrong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.Q11-comparative

KMT2A::ELL (MLL-ELL), t(11;19)(q23;p13.1)

also called MLL-ELL · MLL::ELL · KMT2A-ELL · MLL/ELL · t(11;19)(q23;p13.1)

Here the broken KMT2A gene joins ELL, a genuine component of the machinery that makes RNA polymerase stop pausing and run. But when researchers cut ELL apart, the speed-up part turned out not to be what causes the leukemia - the part that grabs a partner protein called EAF1 is. It is mostly a myeloid leukemia partner, notably common in infant myeloid leukemia and in leukemias caused by previous chemotherapy.

The same thing, for a scientist

KMT2A::ELL accounts for 119/3401 (3.5%) of genomically characterised KMT2A rearrangements and is myeloid-restricted: ~10.1% of KMT2A-rearranged AML, 15.2% of infant AML and 11.6% of adult AML KMT2A-r cases [E1]. Its breakpoint behaviour is unusual and diagnostically informative - paediatric cases prefer KMT2A intron 9 while adult cases prefer intron 11, and that adult shift is specifically attributable to therapy-induced leukaemias rather than to age itself [E1]. ELL is mechanistically distinct from every other partner here: it is not a recruiter of the elongation machinery but a **direct RNA polymerase II elongation factor**, identified as such because it suppresses transient pausing and increases the catalytic elongation rate of Pol II in vitro [E2]. It is also a core structural subunit of the super elongation complex, binding the AFF1/AFF4 scaffold through its occludin-homology domain and operating with P-TEFb on chromatin [E3][E4]. It is tempting to conclude that ELL supplies the enzymatic accelerator of transcription itself, and this entry previously said so. **The structure-function data say otherwise.** The fusion carries ELL 46-621 (exon 1, aa 1-45, is excluded) [E7]; deleting ELL's elongation domain from the fusion does not abolish immortalisation, whereas deleting its EAF1-interaction domain does, and a heterologous MLL-EAF1 fusion causes AML with shorter latency than MLL-ELL [E7]. So ELL behaves like the other 11q23 partners after all - it recruits an AF4-family-like acidic activation domain (EAF1's) - and the elongation activity, though genuinely present, is not the transforming function. t(11;19)(q23;p13.1) was not among the independently prognostic subgroups in paediatric KMT2A-r AML, so it sits in the intermediate band around the cohort 5-year EFS of 44% +/- 5% [E5]. Menin inhibition applies partner-agnostically [E6].

What it looks like
loading 2DOA…
PDB 2DOA ↗
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 lost on each side1 deeper

Kept from KMT2A: AT-hooks 1-3, SNL1/2, MBM1/MBM2, the LEDGF/PSIP1-binding region, CXXC. Lost: PHD1-3, bromodomain, PHD4, FYRN/FYRC, TAD, WIN, SET 1. The canonical partner-agnostic half.

Kept from ELL: amino acids 46 to 621 of 621 — that is, everything except exon 1 7. Concretely this includes:

  • the occludin-homology (OCEL) domain, aa 507-617, through which ELL binds the AFF1/AFF4 SEC scaffold 4 and, overlapping it, the minimal EAF1-interaction domain aa 508-621 7;
  • most of the elongation-stimulatory domain, which Luo et al. map to aa 1-373 and which the fusion therefore keeps as aa 46-373 7.

Lost from ELL: exon 1 only, amino acids 1-45 7. This is the N-terminal EAF1/EAF2-interaction domain 8 — not, as an earlier version of this entry stated, a Pol II/DSIF-contacting region. The consequence is specific and interesting: ELL binds EAF1 at two separate sites (N-terminal and C-terminal) but EAF2 only at the N-terminal one, so MLL-ELL retains EAF1 binding and loses EAF2 binding altogether 8.

The correction that matters most. Listing the elongation-stimulatory region as retained is a sequence fact and it is correct. But it is not the transforming element. Luo et al. deleted the whole elongation domain from the fusion (MLL-ELL380-621) and it still immortalised myeloid progenitors, while the EAF1-interaction domain was critical; MLL fused to the minimal EAF1-interaction domain alone caused AML in transplanted mice, and a heterologous MLL-EAF1 fusion immortalised progenitors and caused AML with shorter latency than MLL-ELL 7. So the ELL half of this chimera works by recruiting an AF4-family-like transcriptional activation domain, exactly like the other 11q23 partners — not by donating polymerase kinetics.

What survives of the 'distinctive feature' claim. Wild-type ELL is a genuine elongation factor: purified ELL suppresses transient pausing and raises the catalytic elongation rate 2. That remains true of the protein. What the structure-function data remove is the inference that this is why the fusion transforms.

Therapy-related leukemia and the breakpoint signature

KMT2A::ELL has an unusual age-breakpoint pattern: paediatric cases shift towards KMT2A intron 9, whereas adult cases shift towards intron 11 - and the recombinome analysis attributes the adult intron-11 shift specifically to therapy-induced cases rather than to adult age per se 1. Intron 11 is the most feature-dense fragile element in KMT2A, carrying topoisomerase-II binding and cleavage sites, a DNase-I hypersensitive site, a Pol II binding site and an apoptotic cleavage site - exactly the substrate expected after topoisomerase-II-directed chemotherapy. KMT2A::ELL is therefore one of the cleanest illustrations in the family that the breakpoint remembers the mutagen, and that a partner's apparent age distribution can be an artefact of treatment history.

Mechanism: an elongation factor is fused in, but the elongation activity is not what transforms

Read this section with the structure-function correction above in hand. Wild-type ELL really is an RNA polymerase II elongation factor 2 and really is a core SEC subunit 34. But Luo et al. showed the elongation domain is dispensable for MLL-ELL transformation and the EAF1-interaction domain is essential 7, so the tidy story that this fusion works by donating polymerase kinetics is not supported. It works, like AFF1/LAF4/AF5q31 fusions, by recruiting an acidic transcriptional activation domain — here EAF1's, borrowed through the retained OCEL module.

Menin/LEDGF docking and CXXC targeting deliver the chimera to CpG-island promoters, as for all major-BCR fusions. What the ELL half then contributes is an OCEL module that binds EAF1 and the AFF1/AFF4 scaffold - a recruitment surface, not a catalytic one.

In its normal role ELL sits at the heart of the super elongation complex alongside the AFF1/AFF4 scaffold and P-TEFb (CDK9-cyclin T1) 3. The complex is recruited to chromatin and connected to Pol II through the PAF complex, which physically links SEC to polymerase on chromatin 3. Within SEC, ELL binds AFF4 through an occludin-homology domain; the paralogue ELL2 has been crystallised bound to an AFF4 peptide (PDB 5JW9) and captured inside an active Pol II-DSIF-ELL2-EAF1 elongation complex (7OKX), so the architecture of this arm of the machine is structurally defined in a way that the AFF1 arm is not 4.

An earlier version of this entry concluded here that KMT2A::ELL is the fusion for which 'elongation hijack' is most literally true, with no intermediate recruitment step to disrupt. That conclusion is withdrawn. There is an intermediate recruitment step - the OCEL-mediated EAF1 contact - and it is the step the genetics says matters 7. The therapeutic implication flips from discouraging to conventional: this partner poses the same acidic-activation-domain recruitment problem as AFF1, rather than a uniquely unapproachable catalytic one.

The handle: the ELL OCEL/occludin domain (aa 507-617) is a defined, structurally characterised protein-protein interface against AFF4 (5JW9) 4 and, overlapping it, the genetically essential EAF1-interaction domain (aa 508-621) 7. It is the best-evidenced partner-specific target on this fusion. Two caveats stand: disrupting it would dismantle SEC in normal cells as well, and MLL-EAF1 transforms without ELL at all 7, so the recruited activation domain is substitutable in principle.

Clinical phenotype, age and prognosis

Lineage skew: myeloid, and essentially only myeloid. ~10.1% of KMT2A-rearranged AML, with negligible representation in ALL 1. Alongside MLLT3, MLLT10 and AFDN, ELL is one of the four partners that make up the KMT2A-r AML landscape.

Age skew: 15.2% of infant AML and 11.6% of adult AML KMT2A-r cases 1 - present across the age range, with the adult cases enriched for therapy-related disease.

Morphology: typically AML with monocytic features; therapy-related cases frequently follow topoisomerase-II inhibitor exposure with a short latency.

Prognosis: intermediate, by exclusion. In the 756-patient international paediatric KMT2A-r AML series the independently prognostic subgroups were t(1;11) (favourable, HR 0.1), t(6;11) (HR 2.2), t(10;11)(p12;q23) (HR 1.5) and t(10;11)(p11.2;q23) (HR 2.5); t(11;19)(q23;p13.1) was not identified as independently prognostic, placing it near the cohort 5-year EFS of 44% +/- 5% 5. This is a negative inference from a study not powered for this subgroup, and should be treated as such rather than as a positive finding of intermediate risk.

Contrast within the family. AFF1 is the lymphoid infant partner; MLLT3 the commonest myeloid partner; MLLT10 and AFDN the adverse myeloid partners; ELL the myeloid partner most associated with prior chemotherapy. The comparative point is that even among the four myeloid partners, the prognostic spread is wide and partner-determined.

Druggability and therapeutic status

Menin inhibition: clinically validated for the class, partner-agnostic (CR+CRh 22.8%, ORR 63.2% in KMT2A-translocated R/R acute leukemia) 6. This is the only clinically available targeted option.

No partner-specific target of the DOT1L type exists, because ELL is not a recruiter. The nearest equivalents are:

  • CDK9 / P-TEFb inhibition. SEC's catalytic subunit is CDK9, and ELL's function is coupled to it. CDK9 inhibitors exist but are broadly cytotoxic; targeting the general elongation kinase to treat an elongation-factor fusion is close to targeting transcription itself.
  • The ELL OCEL/occludin-domain interface (aa 507-617) — structurally defined against AFF4 (5JW9) and, in the same region, genetically validated: Luo et al. showed the overlapping minimal EAF1-interaction domain (aa 508-621) is essential for MLL-ELL immortalisation and leukemogenesis 7. This is the one partner-specific handle on this fusion that has both a structure and a loss-of-function requirement behind it. Undrugged. Disruption would destabilise SEC generally rather than selectively affecting the fusion, since the same interface operates in normal cells.
  • BRD4/BET inhibition, which acts upstream on Pol II pause release. Non-selective for this fusion.

Honest status: undrugged at the partner level, emerging only via the shared menin node. KMT2A::ELL is the clearest example in the family of a fusion whose partner contribution is functionally essential but pharmacologically unapproachable, because the function it contributes is the housekeeping activity of the transcription machine.

Dispute settled: the elongation domain is retained but dispensable, and the lost N-terminus was mislabelled

Dispute settled 2026-09-18; the objection partly held and the entry has been corrected. An audit beat made two claims. Both were checked against Luo RT, Lavau C, Du C, Simone F, Polak PE, Kawamata S, Thirman MJ, The elongation domain of ELL is dispensable but its ELL-associated factor 1 interaction domain is essential for MLL-ELL-induced leukemogenesis, Mol Cell Biol 2001;21(16):5678-87, PMID 11463848 (abstract from PubMed; body text read from the PMC deposit, PMC87288).

Claim 1 — 'Luo 2001 showed the elongation-stimulatory region is dispensable for transformation.' CONFIRMED as stated, but it does not make the old entry wrong. Luo et al. deleted the elongation domain (ELL aa 1-373) from the fusion and MLL-ELL380-621 retained immortalising capacity 7. However, retained in the fusion and required for transformation are different propositions. The entry listed the region under domains_retained_3, a sequence claim, and that claim is true — ELL 46-621 includes aa 46-373. The entry's real fault was elsewhere: its prose implied the elongation activity was the mechanism. Both the list and the prose now carry the dispensability result explicitly.

Claim 2 — 'the discarded N-terminus is mislabelled.' CONFIRMED, and this was a genuine error. The entry described the lost ELL segment as 'the Pol II/DSIF-contacting region in the wild-type elongation complex'. Nothing supports that. The lost segment is ELL exon 1, aa 1-45 7, and its documented identity is the N-terminal EAF1/EAF2-interaction domain, which Simone et al. showed is disrupted by the translocation so that MLL-ELL keeps an EAF1 site but not an EAF2 site 8. The mislabel appears to have been back-inferred from PDB 7OKX, a Pol II-DSIF-ELL2-EAF1 complex; ELL2 is the paralogue, its construct there is full-length, and the ordered ELL2 element in that architecture is the C-terminal occludin/OCEL module, not residues 1-45. The label has been replaced.

A third thing found while checking, which the entry had missed. MLL-ELL exerts a dominant effect on wild-type ELL biology: transfection of MLL-ELL delocalises endogenous EAF1 from nuclear speckles into diffuse nucleoplasm, and EAF1 speckles are undetectable in MLL-ELL-derived leukemic lines 9. That is a subtraction the fusion imposes in trans, separate from anything it gains.

Analogy: an AF4-class fusion assembled in trans

Closest analogue, and the axis. fusion-kmt2a-aff1, on the axis of shared mechanism class - recruitment of an AFF/AF4-family acidic transactivation domain to KMT2A targets. The pairing is not obvious, because ELL and AFF1 look like different kinds of protein: AFF1 is the SEC scaffold and ELL is a catalytic elongation factor. The structure-function data collapse the difference. ELL's elongation domain is dispensable for immortalisation while its EAF1-interaction domain is essential 7, and EAF1's own activation domain is homologous to the activation domains of AF4, LAF4 and AF5q31 - that is, to the AFF family, of which AFF1 is the index member 9.

So the two fusions deliver the same class of module and differ in where the module comes from:

  • KMT2A::AFF1 carries the AFF-family activation domain in cis, inside the oncoprotein.
  • KMT2A::ELL recruits an AFF-like activation domain in trans, on EAF1, using ELL as the adaptor.

What the analogy predicts, and the test that has been run. If the transforming module really is the AF4-like activation domain, then handing it to KMT2A directly should work at least as well as recruiting it. It does, and the direction of the difference is the informative part: MLL-EAF1 immortalises myeloid progenitors like MLL-ELL and induces AML with a shorter latency than MLL-ELL, whereas MLL fused only to the minimal EAF1-interaction domain of ELL gives AML with a longer latency 7. Recruitment is a less efficient way of getting the same module to the same place than carrying it. That is a genuine, tested, quantitative prediction of the analogy, and it is the reason this entry no longer says ELL supplies the enzymatic accelerator of transcription.

Where it breaks, and this is the interesting half: mechanism class does not carry lineage. The two fusions are in the same mechanistic class and produce opposite diseases.

KMT2A::AFF1KMT2A::ELL
lineageB-lineage, overwhelminglymyeloid-restricted
index populationinfant B-ALL, the disease this graph is about~10.1% of KMT2A-rearranged AML, 15.2% of infant AML 1
adult breakpoint behaviour-shift to KMT2A intron 11, attributable to therapy-induced leukaemia rather than to age 1
how the AFF-class AD arrivesin cisin trans, on EAF1

If an AFF-family activation domain recruited to KMT2A targets were sufficient to specify infant B-lineage disease, KMT2A::ELL should produce it. It does not. Whatever sets the lineage of KMT2A::AFF1 is therefore not the class of activation domain the partner supplies, and the candidates left standing are the cell of origin, the fetal developmental window, and something partner-specific that is not shared with EAF1.

Two further limits worth stating plainly. First, the EAF1 evidence is murine retroviral transformation throughout; no human t(11;19)(q23;p13.1) primary sample has been shown to depend on EAF1. Second, the delocalisation imaging that first linked ELL to EAF1 is as compatible with sequestration as with recruitment, and it is the MLL-EAF1 construct - not the imaging - that settles the direction (second counter-evidence note on 9).

Facts and figures
gene5KMT2A
gene3ELL
cytobandt(11;19)(q23;p13.1)
cytoband noteMust be distinguished from t(11;19)(q23;p13.3), which is KMT2A::MLLT1 - a different partner on the same chromosome arm with a lymphoid rather than myeloid phenotype. Band-level cytogenetics alone is not reliable for this discrimination.
prognosisIntermediate by exclusion: t(11;19)(q23;p13.1) was not identified as an independently prognostic subgroup in 756 children with KMT2A-rearranged AML, where overall 5-year EFS was 44% +/- 5% and the adverse subgroups were t(6;11) and t(10;11). No ELL-specific survival estimate exists.
uniprot5Q03164
uniprot3P55199
uniprot5 verifiedQ03164 = KMT2A_HUMAN, 3969 aa (UniProt REST, checked 2026-09-18)
uniprot3 verifiedP55199 = ELL_HUMAN, gene ELL, RNA polymerase II elongation factor ELL, 621 aa, Homo sapiens (UniProt REST, checked 2026-09-18)
exon numbering noteKMT2A exon numbers follow the Meyer/Marschalek DCAL consensus. The ELL exon number is the commonly reported junction and should be treated as indicative rather than exact.
PDB noteNo structure of the KMT2A::ELL fusion protein exists. 2DOA = solution structure of a helical domain of human ELL1 (P55199), the actual fusion partner, and the only ELL1 entry in the PDB; 5JW9 = crystal structure of the ELL2 occludin domain bound to an AFF4 peptide, i.e. the ELL:SEC-scaffold interface, by paralogue homology; 7OKX = active Pol II-DSIF-ELL2-EAF1 elongation complex, showing where an ELL-family protein sits on the polymerase; 7UJ4 and 4GQ6 = menin with revumenib and with an MLL peptide, the partner-agnostic drugged node. All verified against the RCSB entry API on 2026-09-18.
druggability classundrugged
druggability noteUndrugged at the partner level, but NOT for the reason this entry previously gave. The old note said ELL's contribution is the housekeeping elongation activity of RNA polymerase II itself, so there is no DOT1L-style recruitment step to interrupt. Luo et al. 2001 (PMID 11463848) refute that: the elongation domain is dispensable for transformation and the EAF1-interaction domain (aa 508-621, overlapping the OCEL/occludin domain 507-617) is essential. There IS a recruitment step, and it is the one that matters. It remains undrugged, and the general-transcription candidates (CDK9, BET) remain non-selective; the fusion is also covered by the partner-agnostic menin node. Corrected 2026-09-18.
stage of hematopoiesisbone-marrow haematopoietic stem cell or myeloid progenitor; in therapy-related cases a mature-marrow HSC that acquired the translocation after topoisomerase-II-inhibitor exposure, i.e. a postnatal, fully adult cell of origin
age peakbimodal across life: infant AML (15.2% of KMT2A-r infant AML) and adult AML, with the adult cases enriched for therapy-related disease
developmental order rank6
developmental order noteRanked 6 of 10. Later than the infant lymphoid and infant myeloid partners because a large share of ELL cases are therapy-related and arise in a postnatal adult bone-marrow HSC, which is the most mature stem-cell origin in the KMT2A set. The infant-AML subset would rank earlier, so this single integer averages two distinct origins.
ell structure function noteCorrected 2026-09-18 against PMID 11463848 (Luo et al., MCB 2001) and PMID 12446457 (Simone et al., Blood 2002). ELL residues retained in MLL-ELL: 46-621. Lost: aa 1-45 (exon 1) = the N-terminal EAF1/EAF2-interaction domain, NOT a Pol II/DSIF-contacting region as this entry previously stated. Elongation domain = aa 1-373, retained as 46-373, dispensable for transformation. Minimal EAF1-interaction domain = aa 508-621, overlapping the UniProt OCEL/occludin domain 507-617, essential for transformation. Net EAF outcome: EAF1 binding retained, EAF2 binding lost.
frequency
all_kmt2a_r_acute_leukemi…3.500all_kmt2a_r_acute_leukemia_pct: 3.5kmt2a_r_aml_pct10.1kmt2a_r_aml_pct: 10.1kmt2a_r_adult_aml_pct11.6kmt2a_r_adult_aml_pct: 11.6kmt2a_r_infant_aml_pct15.2kmt2a_r_infant_aml_pct: 15.2n_cases_dcal_cohort119.0n_cases_dcal_cohort: 119
diseases
[
  "paediatric AML",
  "infant AML",
  "adult AML",
  "therapy-related AML"
]
frequency — the numbers behind the chart
{
  "all_kmt2a_r_acute_leukemia_pct": 3.5,
  "kmt2a_r_aml_pct": 10.1,
  "kmt2a_r_infant_aml_pct": 15.2,
  "kmt2a_r_adult_aml_pct": 11.6,
  "n_cases_dcal_cohort": 119
}
breakpoints
[
  {
    "id": "bp-kmt2a-intron-9-ell",
    "gene5_exon_last": 9,
    "gene3_exon_first": 2,
    "frequency_pct": 37.2,
    "population": "paediatric KMT2A::ELL preferentially breaks in intron 9; frequency_pct is the share of all KMT2A breaks in BCR subregion A (exon 9 + intron 9), not the share of ELL fusions",
    "domains_retained_5": [
      "MBM1",
      "MBM2",
      "AT-hook 1",
      "AT-hook 2",
      "AT-hook 3",
      "SNL1/2",
      "CXXC"
    ],
    "domains_lost_5": [
      "PHD1",
      "PHD2",
      "PHD3",
      "Bromodomain",
      "PHD4",
      "FYRN",
      "TAD",
      "FYRC",
      "WIN",
      "SET"
    ],
    "domains_retained_3": [
      "occludin-homology / OCEL domain, aa 507-617 (UniProt P55199) — binds the AFF1/AFF4 SEC scaffold AND EAF1; overlaps the minimal EAF1-interaction domain aa 508-621, which Luo et al. 2001 showed is ESSENTIAL for MLL-ELL immortalisation and leukemogenesis",
      "ELL elongation-stimulatory region (ELL aa 1-373, retained in the fusion as aa 46-373) — RETAINED as a sequence fact, but DISPENSABLE for transformation: MLL-ELL380-621, which deletes the domain outright, still immortalises myeloid progenitors (Luo et al. 2001, PMID 11463848)",
      "ELL C-terminus to residue 621 (the same segment as the OCEL/EAF1 module above, listed separately in earlier versions of this entry)"
    ],
    "domains_lost_3": [
      "ELL exon 1, residues 1-45 — the N-terminal EAF1/EAF2-interaction domain. CORRECTED 2026-09-18: this was previously labelled 'Pol II/DSIF-contacting region in the wild-type elongation complex', which no source supports. Luo et al. 2001 (PMID 11463848) state that ELL exon 1 spans aa 1-45 and is excluded because the genomic breaks lie 3' of it, so the fusion carries ELL 46-621; Simone et al. 2002 (PMID 12446457) identify the ELL N-terminus as a protein-protein interaction domain binding both EAF1 and EAF2 and state that it is disrupted by the fusion, so MLL-ELL retains an EAF1 site but loses EAF2 binding."
    ],
    "phenotype_notes": "HOXA-high group; de novo paediatric and infant AML. gene3_exon_first = 2 is confirmed by Luo et al. 2001 (PMID 11463848): 'the genomic breaks within ELL in patients with (11;19)(q23;p13.1) translocations occur 3' of the first exon of ELL', exon 1 spanning aa 1-45, and the chimeric protein contains ELL amino acids 46 to 621. The retained KMT2A half also contains the LEDGF/PSIP1-binding region, which has no lane in the parent gene-kmt2a domain map and is therefore recorded here rather than in the domain arrays.",
    "gene3_residue_range_retained": "ELL 46-621 of 621 aa (Luo et al. 2001, PMID 11463848, full text via PMC87288)"
  },
  {
    "id": "bp-kmt2a-intron-11-ell-therapy",
    "gene5_exon_last": 11,
    "gene3_exon_first": 2,
    "frequency_pct": 39.5,
    "population": "adult KMT2A::ELL, where the intron-11 preference is attributable specifically to therapy-induced cases; frequency_pct is the share of all KMT2A breaks in BCR subregion C",
    "domains_retained_5": [
      "MBM1",
      "MBM2",
      "AT-hook 1",
      "AT-hook 2",
      "AT-hook 3",
      "SNL1/2",
      "CXXC"
    ],
    "domains_lost_5": [
      "PHD1",
      "PHD2",
      "PHD3",
      "Bromodomain",
      "PHD4",
      "FYRN",
      "TAD",
      "FYRC",
      "WIN",
      "SET"
    ],
    "domains_retained_3": [
      "occludin-homology / OCEL domain, aa 507-617 (UniProt P55199) — binds the AFF1/AFF4 SEC scaffold AND EAF1; overlaps the minimal EAF1-interaction domain aa 508-621, which Luo et al. 2001 showed is ESSENTIAL for MLL-ELL immortalisation and leukemogenesis",
      "ELL elongation-stimulatory region (ELL aa 1-373, retained in the fusion as aa 46-373) — RETAINED as a sequence fact, but DISPENSABLE for transformation: MLL-ELL380-621, which deletes the domain outright, still immortalises myeloid progenitors (Luo et al. 2001, PMID 11463848)",
      "ELL C-terminus to residue 621 (the same segment as the OCEL/EAF1 module above, listed separately in earlier versions of this entry)"
    ],
    "domains_lost_3": [
      "ELL exon 1, residues 1-45 — the N-terminal EAF1/EAF2-interaction domain. CORRECTED 2026-09-18: this was previously labelled 'Pol II/DSIF-contacting region in the wild-type elongation complex', which no source supports. Luo et al. 2001 (PMID 11463848) state that ELL exon 1 spans aa 1-45 and is excluded because the genomic breaks lie 3' of it, so the fusion carries ELL 46-621; Simone et al. 2002 (PMID 12446457) identify the ELL N-terminus as a protein-protein interaction domain binding both EAF1 and EAF2 and state that it is disrupted by the fusion, so MLL-ELL retains an EAF1 site but loses EAF2 binding."
    ],
    "phenotype_notes": "Therapy-related AML following topoisomerase-II-directed chemotherapy; intron 11 carries topoisomerase-II binding and cleavage sites, a DNase-I hypersensitive site, a Pol II binding site and an apoptotic cleavage site, so the breakpoint records the mutagen. The retained KMT2A half also contains the LEDGF/PSIP1-binding region, recorded here rather than in the domain arrays because the parent gene-kmt2a domain map has no corresponding lane.",
    "gene3_residue_range_retained": "ELL 46-621 of 621 aa (Luo et al. 2001, PMID 11463848, full text via PMC87288)"
  }
]
PDB IDS
[
  "2DOA",
  "5JW9",
  "7OKX",
  "7UJ4",
  "4GQ6"
]
mechanism class
[
  "chromatin-elongation-hijack",
  "activation-domain-recruitment"
]
drugs
[
  "drug-revumenib",
  "drug-ziftomenib"
]
key interactors
[
  "gene-aff1",
  "gene-aff4",
  "gene-cdk9",
  "gene-ell",
  "gene-hoxa9",
  "gene-meis1",
  "gene-men1",
  "gene-paf1",
  "gene-psip1"
]
trials
[
  "NCT04065399"
]
How this could bite us back
  • Resistance: menin-inhibitor resistance via MEN1 binding-site mutation applies identically, because the drugged surface is on the shared KMT2A half.
  • Essential-function toxicity: every candidate partner-directed target here - CDK9/P-TEFb, the ELL:AFF4 interface, BET proteins - is a component of general transcription. Inhibiting them is not fusion-selective and carries the marrow, gut and cardiac toxicity expected of global transcriptional shutdown.
  • Compensation: ELL has paralogues (ELL2, ELL3) that also bind AFF4 and function in elongation, so an ELL1-selective agent would likely be buffered; and EAF1/EAF2 modulate ELL activity, adding another layer of adaptation.
  • Therapy-induction paradox: a large fraction of adult KMT2A::ELL disease is caused by prior chemotherapy. Any intensification of topoisomerase-II-directed therapy to treat other leukemias will generate more of this fusion - a genuine iatrogenic feedback loop that a targeted-therapy programme should count as a cost.
  • Delivery and diagnosis: t(11;19)(q23;p13.1) and t(11;19)(q23;p13.3) are cytogenetically adjacent but biologically opposite (myeloid ELL versus lymphoid MLLT1). Misassignment at the band level would send a patient into the wrong biology, and partner-specific strategies depend on molecular confirmation.
  • The partner-specific handle here is the ELL OCEL/EAF1 interface, and Luo et al.'s own result argues it is replaceable rather than unique: MLL-EAF1 transforms at least as well as MLL-ELL [E7]. A ligand that blocks ELL:EAF1 would be blocking one instance of a recruitment step that several different acidic activation domains can supply, which is the same redundancy problem the AFF1/LAF4/AF5q31 family poses.
Still open
  1. MLL-ELL loses EAF2 binding while keeping EAF1 binding [E8], and a heterologous MLL-EAF2 fusion immortalises progenitors. Is losing the EAF2 contact a functionally meaningful subtraction, or is EAF1 simply sufficient? A cis-resupply experiment - MLL-ELL with the N-terminal EAF interaction domain restored in frame - would separate them.
  2. Luo et al. found MLL fused to the minimal EAF1-interaction domain of ELL caused AML with LONGER latency than full MLL-ELL, while MLL-EAF1 caused it with SHORTER latency [E7]. Something in ELL 46-507 slows disease relative to EAF1 and speeds it relative to the minimal domain. Nobody has explained that ordering, and it is the sharpest unexplained observation about this fusion.
  3. If ELL's contribution is the general elongation activity of RNA polymerase II, is KMT2A::ELL leukemia more or less dependent on menin than the recruitment-based fusions? A partner-stratified menin-inhibitor response analysis would answer this and is a direct prediction: fusions whose partner supplies its own effector should be no less menin-dependent, because menin controls targeting rather than firing.
  4. Is the ELL:AFF4 occludin-domain interface (structurally defined in the ELL2 paralogue, 5JW9) druggable, and would blocking it be tolerated given that the same interface operates in every normal cell?
  5. The adult intron-11 shift in KMT2A::ELL is attributed to therapy-induced cases on small numbers. Does a larger therapy-related AML cohort confirm that this partner is specifically selected by topoisomerase-II-inhibitor exposure, and if so does that reflect ELL locus fragility, 3D proximity during the DNA-damage response, or selection at the cellular level?
  6. Why is ELL myeloid-restricted when its SEC partner AFF1 is lymphoid-restricted, given that both proteins work in the same complex? This is the sharpest test available of whether lineage is set by the partner locus's expression in the cell of origin rather than by the protein's function.
  7. Is EAF1 required for KMT2A::ELL leukaemia in a human cell, and is the requirement menin-dependent? Everything placing KMT2A::ELL in the AF4 mechanism class is murine retroviral structure-function from 2001. A CRISPR dropout of EAF1 and EAF2 in a human t(11;19)(q23;p13.1) line or PDX, scored alongside AFF1 dropout in a KMT2A::AFF1 line, would test the analogy directly - and the prediction that distinguishes it from the SEC-scaffold model is that EAF1 loss should phenocopy menin inhibition in KMT2A::ELL cells and not in KMT2A::AFF1 cells.
Every source behind this entry (9)
  1. E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    KMT2A::ELL accounts for 119/3401 (3.5%) of genomically characterised KMT2A rearrangements, about 10.1% of KMT2A-rearranged AML, 15.2% of infant AML and 11.6% of adult AML KMT2A-r cases; paediatric cases prefer KMT2A intron 9 while the adult shift to intron 11 is attributable specifically to therapy-induced leukaemias.

    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

    Referral-based single-centre cohort, and with only 119 ELL cases the age- and therapy-stratified breakpoint claims rest on small subgroups. Therapy-induced status is recorded from referring-centre clinical annotation, not independently adjudicated. Percentages are of cases with an identified partner.

    Evidence that argues against this (1)
    • These are shares of the recombinome series rather than population frequencies. The series is what a reference-laboratory network accumulated from samples sent to it for KMT2A breakpoint characterisation between 2003 and 2022, so its denominator is referrals, not a population. It is also a long-tailed distribution: of 107 in-frame KMT2A fusions catalogued, 37 are recurrent and 63 have been seen only once, and the seven commonest partners plus partial tandem duplications account for more than 90% of recombinations. Percentages from it are shares of a referral series and should be quoted that way, especially alongside percentages from population-based registries or from prospective trial cohorts, which are computed on different denominators. Meyer C et al. 2023
  2. E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    ELL is a genuine RNA polymerase II elongation factor: purified ELL suppresses transient pausing by Pol II and increases the catalytic rate of transcript elongation.

    An RNA polymerase II elongation factor encoded by the human ELL gene · Shilatifard A, Lane WS, Jackson KW, Conaway RC, Conaway JW · Science 1996
    PMID 8596958DOI
    What this evidence cannot show

    In vitro reconstituted transcription with purified factors; ELL's behaviour in isolation differs from its behaviour in the holo-ELL complex, where associated proteins (EAF1/EAF2) suppress an inhibitory activity ELL shows alone. The biochemistry therefore establishes that ELL acts directly on polymerase, not the precise net effect inside a cell or inside a KMT2A fusion.

    Evidence that argues against this (1)
    • The purified-ELL elongation activity is a biochemical property measured on isolated components, and in cells ELL does not arrive on chromatin by itself. ELL is delivered as a subunit of the AFF1/AFF4-nucleated super elongation complex, and it is knockdown of the AFF4 SCAFFOLD, not of ELL, that reduces KMT2A-chimera target gene expression. So the step from 'ELL stimulates elongation in vitro' to 'the fusion works because it delivers an elongation factor' passes through a complex whose limiting component appears to be the scaffold. The biochemistry is not in doubt; its use as the causal account of the fusion is. Lin C et al. 2010
  3. E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    ELL is a core subunit of the super elongation complex together with the AFF1/AFF4 scaffold and P-TEFb, and the PAF complex physically connects SEC to RNA polymerase II on chromatin.

    Human Polymerase-Associated Factor complex (PAFc) connects the Super Elongation Complex (SEC) to RNA polymerase II on chromatin · He N, Chan CK, Sobhian B, et al. · Proceedings of the National Academy of Sciences USA 2011
    PMID 21873227DOI
    What this evidence cannot show

    Biochemical and cell-line work centred on HIV Tat transactivation as well as MLL fusions, so the SEC-PAFc connection is established for SEC generally rather than specifically for a KMT2A::ELL chimera.

    Evidence that argues against this (1)
    • SEC-plus-PAF is the right description for this partner and is not the class mechanism, which matters because this entry is part of a comparative set. The cytoplasmic KMT2A partners AF1p/EPS15 and GAS7 have no elongation function at all, and their coiled-coil oligomerisation domains alone are necessary AND sufficient for transformation by the corresponding fusions; synthetic dimerisation modules with no biology reproduce Hox activation and enhanced progenitor self-renewal. Forced self-association covers every partner, SEC recruitment covers this one. So CW et al. 2003
  4. E4strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

    ELL-family proteins bind the AFF4 scaffold through an occludin-homology domain and are integral to the active Pol II elongation complex, as shown by the ELL2 occludin domain-AFF4 peptide crystal structure and by a Pol II-DSIF-ELL2-EAF1 elongation complex structure.

    What this evidence cannot show

    Both structures are of ELL2 (UniProt O00472), the paralogue, not ELL1 (P55199) which is the KMT2A fusion partner; the only ELL1 structure in the PDB is 2DOA, a solution structure of an isolated helical domain. The architecture is therefore transferred by homology. Entries verified against the RCSB entry API on 2026-09-18.

    Evidence that argues against this (2)
    • Both structures are of the PARALOGUE, and one of them is composite. Checked directly against the RCSB entry records on 2026-09-19: 5JW9 is an X-ray structure at 2.00 A with two polymer entities, titled 'The crystal structure of ELL2 occludin domain and AFF4 peptide', whose primary citation is about the structural basis for ELL2 and AFF4 activation of HIV-1 PROVIRAL transcription; 7OKX is a cryo-EM structure at 3.3 A with 18 polymer entities and is explicitly labelled a COMPOSITE structure of Pol II-DSIF(SPT5-KOW5)-ELL2-EAF1, i.e. assembled from more than one map rather than observed as one particle. Neither contains ELL1, which is the protein fused to KMT2A in t(11;19)(q23;p13.1), and neither contains any KMT2A sequence. The occludin-homology/AFF4 interaction is well supported; what has been solved is the ELL2 version of it in an HIV transcription context, and a composite model of a complex in which ELL2 stands in for the family. RCSB Protein Data Bank entry record 2021
    • The companion structure has the same substitution. 5JW9 is ELL2's occludin domain bound to an AFF4 peptide at 2.00 A, and its primary citation frames the work as the structural basis for ELL2 and AFF4 activation of HIV-1 proviral transcription. Using it to support a statement about how ELL-family proteins engage AFF4 is reasonable; using it as a structure of the KMT2A::ELL interface is not, and the entry should say which it means. Checked against the RCSB entry record on 2026-09-19. RCSB Protein Data Bank entry record 2017
  5. E5moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    In 756 children with 11q23/KMT2A-rearranged AML the independently prognostic subgroups were t(1;11) (HR 0.1), t(6;11) (HR 2.2) and the two t(10;11) subgroups (HR 1.5 and 2.5); t(11;19)(q23;p13.1) was not among them, and overall 5-year EFS was 44% +/- 5%.

    PMID 19528532DOI
    What this evidence cannot show

    This is a negative inference: absence from the list of independently prognostic subgroups in a study not powered for ELL is not evidence of intermediate risk. Retrospective pooling across 11 groups and many protocols; karyotype-based assignment cannot distinguish t(11;19)(q23;p13.1) from p13.3 reliably in all laboratories. Pre-dates menin inhibitors.

    Evidence that argues against this (1)
    • Absence from a prognostic-hazard table is absence on one axis only, and there is a clinical signal on another that points the opposite way for this partner. In the phase 1 study of the DOT1L inhibitor pinometostat in 51 adults with advanced acute leukaemia, exactly two patients achieved complete remission, and BOTH had t(11;19). The only single-agent complete remissions ever produced by a DOT1L inhibitor in this disease came from this cytogenetic group. A chemotherapy-era hazard ratio that finds t(11;19)(q23;p13.1) unremarkable does not license the inference that the partner is biologically unremarkable. Stein EM et al. 2018
  6. E6established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    Revumenib was studied and approved in KMT2A-translocated relapsed/refractory acute leukemia without restriction by partner gene, achieving CR+CRh 22.8% and ORR 63.2%.

    Menin Inhibition With Revumenib for KMT2A-Rearranged Relapsed or Refractory Acute Leukemia (AUGMENT-101) · Issa GC, Aldoss I, Thirman MJ, et al. · Journal of Clinical Oncology 2025
    PMID 39121437DOINCT04065399
    What this evidence cannot show

    No partner-stratified efficacy analysis; activity in KMT2A::ELL specifically is inferred from the shared retained MBM module and the partner-agnostic eligibility criteria, and with only ~3.5% of KMT2A-r cases being ELL the trial is very unlikely to have contained many.

    Evidence that argues against this (1)
    • 'Without restriction by partner gene' describes the label, not a demonstration of equal benefit, and the number quoted carries no duration. In the NPM1-mutant cohort of the same trial the median duration of CR+CRh was 4.7 months and only 5 of 30 responders reached transplant. For a partner-level entry the relevant caution is arithmetic: 57 efficacy-evaluable patients spread across a partner distribution in which 63 of 107 partners have been seen once cannot support per-partner efficacy estimates in either direction, and the absence of a partner effect in that trial is an absence of power as much as an absence of effect. Arellano ML et al. 2025
  7. E7strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    The MLL-ELL chimeric protein contains ELL amino acids 46 to 621, exon 1 (aa 1-45) being excluded because the genomic breaks occur 3' of it; the ELL elongation domain (aa 1-373) is dispensable for myeloid immortalisation (MLL-ELL380-621 still transforms) whereas the minimal EAF1-interaction domain (aa 508-621) is essential, MLL fused to that minimal domain causes AML in transplanted mice, and a heterologous MLL-EAF1 fusion immortalises progenitors and causes AML with shorter latency than MLL-ELL.

    The elongation domain of ELL is dispensable but its ELL-associated factor 1 interaction domain is essential for MLL-ELL-induced leukemogenesis · Luo RT, Lavau C, Du C, Simone F, Polak PE, Kawamata S, Thirman MJ · Molecular and Cellular Biology 21(16):5678-87 2001
    PMID 11463848DOI
    What this evidence cannot show

    Abstract read from PubMed; body text and all residue numbers read from the PMC deposit (PMC87288) rather than the typeset PDF, so the residue boundaries (46-621, 1-45, 1-373, 380-621, 508-621, 515-621) are transcribed from that text and were not independently re-derived from the figures. Retroviral overexpression of constructs in mouse myeloid progenitors plus bone-marrow transplant; 'dispensable' means the deletion construct still immortalised in that assay, not that elongation activity contributes nothing in a patient blast. Deletion rather than point mutants, so a partial or structural contribution of the elongation domain is not excluded. Constructs are MLL-ELL, not the endogenous locus.

    Evidence that argues against this (1)
    • The same group's in-vivo data temper the clean dichotomy: MLL fused to the minimal EAF1-interaction domain alone caused AML with LONGER latency than full-length MLL-ELL, so the rest of ELL - including the elongation domain - contributes something to disease kinetics even if it is not required. Luo RT et al. 2001
  8. E8strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    ELL contains a protein-protein interaction domain within its amino terminus that binds both EAF1 and EAF2, and this amino-terminal interaction domain is disrupted in the formation of the MLL-ELL fusion protein, so MLL-ELL retains an interaction domain for EAF1 (at the ELL carboxy-terminus) but not for EAF2.

    ELL-associated factor 2 (EAF2), a functional homolog of EAF1 with alternative ELL binding properties · Simone F, Luo RT, Polak PE, Kaberlein JJ, Thirman MJ · Blood 101(6):2355-62 2002
    PMID 12446457DOI
    What this evidence cannot show

    ABSTRACT ONLY - full text not obtained, so the abstract gives no residue boundaries for the N-terminal EAF1/EAF2-interaction domain; the identification of that domain with ELL exon 1 (aa 1-45) is an inference from combining this paper with Luo et al. [E7], not a statement either paper makes in the text read. Co-immunoprecipitation and confocal colocalisation in cell lines plus retroviral progenitor immortalisation, i.e. overexpression systems.

    Evidence that argues against this (1)
    • Loss of the EAF-binding domain does not leave MLL-ELL without elongation machinery, which is how this observation is often read. KMT2A chimeras including MLL-ELL co-purify the AFF1/AFF4-nucleated super elongation complex with P-TEFb, and AFF4 knockdown reduces chimera target-gene expression - so the fusion reaches the elongation apparatus through the scaffold rather than through ELL's own N-terminal partners. The disrupted EAF interaction is better read as a change in which complex the retained ELL fragment joins than as a loss of function. Lin C et al. 2010
  9. E9strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    EAF1 contains a serine/aspartate/glutamate-rich region homologous to the transcriptional activation domains of three other MLL fusion partners - AF4, LAF4 and AF5q31 - and a transcriptional activation domain has been identified in that region. Endogenous EAF1 and ELL colocalise in a distinct nuclear speckled pattern; transfection of MLL-ELL delocalises EAF1 from speckles to a diffuse nucleoplasmic distribution, and EAF1 speckles are not detectable in leukemic cell lines derived from mice transplanted with MLL-ELL-transduced bone marrow.

    EAF1, a novel ELL-associated factor that is delocalized by expression of the MLL-ELL fusion protein · Simone F, Polak PE, Kaberlein JJ, Luo RT, Levitan DA, Thirman MJ · Blood 98(1):201-9 2001
    PMID 11418481DOI
    What this evidence cannot show

    ABSTRACT ONLY - full text not obtained. Confocal immunofluorescence after transient transfection of MLL-ELL, i.e. an overexpression readout, plus murine MLL-ELL leukemic lines. 'Speckles not detected' is a negative imaging result and is sensitive to antibody and fixation. The functional consequence of EAF1 delocalisation was not established. The AF4/LAF4/AF5q31 homology added to this claim on 2026-09-21 is a database sequence comparison plus a reporter transactivation assay, not a structural or biochemical demonstration that the EAF1 and AFF-family activation domains engage the same machinery; the abstract was re-fetched to take it. AF5q31 is now AFF4 and LAF4 is AFF3, so the homology statement is with the AFF family generally.

    Evidence that argues against this (2)
    • A nuclear speckled pattern that becomes diffuse on expression of a second protein is a concentration-dependent partitioning readout, and elongation-factor bodies are now known to behave that way. ENL and its intrinsically disordered region are sufficient to initiate liquid droplet formation of the super elongation complex in unmodified cells, AFF4 fluidises those droplets, and the droplets form and disperse within minutes of a stimulus; fusion of ENL to MLL boosts SEC phase separation. Against that background, delocalisation of EAF1 on MLL-ELL transfection is consistent with redistribution of a partitioning protein between competing assemblies at altered stoichiometry, which is a different claim from EAF1 being functionally displaced. The 2001 experiment cannot distinguish the two, and the absence of EAF1 speckles in leukaemic cells is an imaging observation whose sample size and expression-level controls are not stated in this entry. Guo C et al. 2020
    • The homology plus delocalisation result is equally consistent with a DOMINANT-NEGATIVE reading - MLL-ELL sequestering EAF1 away from normal ELL - which is the opposite of productive recruitment of an AFF-class activation domain to KMT2A target genes, and the imaging cannot separate them. The structure-function study settles the direction by building the recruited module in rather than imaging it: a heterologous MLL-EAF1 fusion immortalised myeloid progenitors as MLL-ELL does and induced AML with a SHORTER latency than MLL-ELL, while MLL fused to the minimal EAF1-interaction domain of ELL gave AML with a LONGER latency. A pure sequestration model predicts neither observation, because MLL-EAF1 sequesters nothing. Luo RT et al. 2001
Connected to