Oncofusion

start / graph / gene-ell

Proteinstrong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.Q5b-interactome

ELL — rank 7 on selectivity and the deepest common-essential in the top ten, with one NMR fragment to its name

also called ELL1 · Elongation factor for RNA polymerase II · MEN · PPP1R68 · RNA polymerase II elongation factor ELL

ELL is the part of the super elongation complex that actually speeds up RNA polymerase. KMT2A-rearranged leukemia cells need it a bit more than other leukemias do, but every cell needs it enormously — and despite being central to the whole complex, the only human ELL structure ever solved is one small piece floating on its own.

The same thing, for a scientist

ELL (P55199, 621 aa) is the catalytic-adjacent elongation stimulator of the super elongation complex: it increases the catalytic rate of RNA polymerase II by suppressing transient pausing, and it enters the SEC by docking its C-terminal occludin-like domain onto the AFF-family 'ELLBow'. It sits at an uncomfortable intersection in this interactome. It is **rank 7 on selectivity** (selectivity_vs_leukemia -0.177, selectivity_vs_all_lines -0.163) and the mean Chronos in KMT2A::AFF1 lines is -1.537, the deepest of any line subset we computed for it. But it is also DepMap **common-essential**, with a mean Chronos of -1.302 across 1,208 lines and 94% of lines depleted — the deepest common-essential gene in the interactome top ten. The selective increment is real and small; the baseline requirement is enormous. Structurally it is close to a void: **2DOA**, an isolated NMR helical domain (ELL 200-292), is the only human ELL structure in the PDB. There is no ELL-AFF1 structure, no ELL-AFF4 structure, and no ELL in any Pol II elongation complex.

loading 2DOA…
PDB 2DOAThe only human ELL structure in existence
This single small helical piece, floating on its own, is everything the world has structurally for a protein that almost every human cell needs.
Tell me more about this picture

PDB 2DOA, NMR solution structure, ELL residues 200-292 — 93 of 621 residues, no partner, no polymerase.

Tell me more about this picture

ELL is DepMap common-essential in 94% of 1,208 cancer cell lines. There is no other structure of it.

Tell me more about this picture

The corresponding fragment of the paralogue ELL2 (2E5N, also 200-292) has also been solved, so the field has solved the same non-interface piece twice.

Tell me more about this picture

The C-terminal occludin-like domain that actually docks onto the AFF-family ELLBow has been solved for ELL2 (5JW9) and never for ELL.

loading 7OKX…
PDB 7OKXPol II-DSIF-ELL2-EAF1 active elongation complex — the paralogue, not ELL
A real picture of an ELL-family protein speeding up RNA polymerase — except it is ELL2, the version leukemia cells do not need.
Tell me more about this picture

PDB 7OKX, cryo-EM at 3.30 A, with 7OKY at 4.14 A; ELL2 residues 1-640 and EAF1 1-268 on a Pol II-DSIF elongation complex (PMID 34265249).

Tell me more about this picture

This is the mechanistic basis for 'allosteric transcription stimulation by the SEC'. Every application of it to ELL is a paralogue extrapolation.

Tell me more about this picture

The inversion to remember: ELL2 has the structures and no dependency (selectivity -0.010); ELL has the dependency (-0.177, rank 7) and one NMR fragment.

Tell me more about this picture

The complex is reconstituted in vitro and contains no KMT2A fusion protein.

What it looks like
RNA polymerase II elongation factor ELL621 aa
Occludin-homology / ELL_N region (helical domain)Occludin-homology / ELL_N region (helical domain) · residues 200–292 · the only structurally characterised part of human ELL; NMR, PDB 2DOA; not the AFF-binding module0200400600
loading AF-P55199-F1…
AlphaFold AF-P55199-F1 ↗
colourshape

This is a prediction, not a measurement. No one has seen this protein in this shape. The colours say how much the software trusts itself, and that is worth reading before anything else — the orange stretches are places the model is effectively saying this part has no fixed shape.

very high (90+)confident (70–90)low (50–70)very low (<50)
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
The tension, stated as bluntly as the numbers allow2 deeper

ELL is simultaneously one of the most selectively required genes in KMT2A-rearranged lines and one of the least drug-able-with-an-acceptable-index 1:

mean Chronos (1,208)frac. depletedleukemiaKMT2A-rKMT2A::AFF1selectivity vs leukemiacommon-essential
ELL-1.3020.94-1.288-1.465-1.537-0.177 (rank 7)yes
ELL2-0.0640.01-0.033-0.043+0.003-0.010no
CDK9-1.5360.993-1.610-1.687-1.552-0.077yes
AFF4-0.0560.004-0.085+0.026+0.024+0.111no

Read the ELL row twice. The first reading is encouraging: KMT2A::AFF1 lines need ELL more than KMT2A-r lines generally (-1.537 vs -1.465), which in turn need it more than leukemia generally (-1.288), which in turn need it more than the average cancer line (-1.302 is close, so the leukemia increment is small). The gradient is monotone in the right direction and puts ELL at rank 7 on our selectivity list.

The second reading is the one that decides the project. 94% of 1,208 cell lines are depleted, and the magnitude of the baseline requirement (-1.302) is nine times the size of the selective increment (-0.177). Even taking the selectivity at face value — and at n=10 KMT2A-r lines, -0.177 is only just outside our ±0.15 noise band — a systemic ELL inhibitor at a dose sufficient to matter in the tumour would be depleting essentially every proliferating cell in the body. This is a therapeutic-index wall, and it is a biology problem, not a chemistry problem. No amount of potency or selectivity for ELL over ELL2 fixes it.

Why is ELL so much more essential than ELL2? Almost certainly because ELL carries the general Pol II elongation-stimulating function in most cell types, while ELL2 is the tissue-restricted paralogue (HPA calls ELL2 'immune cell enhanced'; ELL is 'detected in all' with low tissue specificity) 1. The paralogues are not redundant in the direction that would help us: knocking out ELL is not buffered by ELL2.

ELL2 is the control that proves the point

ELL2 (O00472) is the closest possible negative control for an ELL-directed hypothesis, and it fails to be a dependency in every way ELL succeeds: mean Chronos -0.064, 1% of lines depleted, KMT2A-r mean -0.043, KMT2A::AFF1 mean +0.003, selectivity_vs_leukemia -0.010 — indistinguishable from zero 1.

This is the inversion worth recording. ELL2 has the structures; ELL has the dependency. ELL2 appears in 7OKX (cryo-EM, 3.30 Å) and 7OKY (4.14 Å) as part of a genuine Pol II-DSIF-ELL2-EAF1 active elongation complex 3 — an actual mechanistic picture of an ELL-family protein stimulating transcription on a polymerase. ELL itself has nothing of the kind. So the field's structural understanding of ELL-family function rests entirely on the paralogue that KMT2A-rearranged cells demonstrably do not need.

gnomAD agrees that ELL matters, at a moderate level

pLI 0.9999, LOEUF 0.417, 18 observed loss-of-function variants against 64 expected, missense z 1.06 1. Human heterozygous loss of ELL is selected against, though the constraint is mid-range rather than extreme (compare AFF4 at 0.166, RBBP5 at 0.146). The low missense z (1.06) is notable: ELL tolerates missense change far better than it tolerates truncation, which is the signature of a protein whose function is carried by a few short modules plus bulk scaffolding rather than by a precisely tuned active site. That is consistent with the structural picture below, and it is mildly bad news for small-molecule inhibition — there may be no single pocket whose occupancy switches the protein off.

The structural void: one NMR fragment for a protein at the centre of the complex1 deeper

This is the finding to record for the graph. Across the whole PDB, the only human ELL structure is 2DOA: an NMR solution structure of an isolated helical domain spanning ELL residues 200-292 2. That is 93 residues out of 621, with no partner, no nucleic acid and no polymerase.

What does not exist:

  • No ELL-AFF1 structure. The interaction that puts ELL into the KMT2A::AFF1 complex has never been solved.
  • No ELL-AFF4 structure. The AFF4 ELLBow has been solved bound to ELL2 (5JW9, 2.00 Å) but never to ELL.
  • No ELL in a Pol II elongation complex. 7OKX/7OKY use ELL2.
  • No ELL-EAF structure. The activation module has been solved as ELL2-EAF1, not ELL-EAF1.

For a protein that is DepMap common-essential in 94% of human cancer cell lines, that is remarkable, and it is not a quirk of curation — it is the reason there is no ELL chemistry. You cannot design against an interface nobody has seen. Every mechanistic statement about ELL in the SEC literature is, structurally, an extrapolation from ELL2.

The single interesting consequence: 2DOA's 200-292 region is the same span as 2E5N, the NMR structure of the ELL2 ELL_N2 domain (ELL2 200-292) 2. So the one fragment we have of each paralogue is the same fragment, and it is not the fragment that binds AFF-family proteins (that is the C-terminal occludin-like domain, ELL2 519-640). The field has solved, twice, the piece that is not the interface.

What to do with a structurally dark essential gene

Three honest options, in order of cost:

  1. Solve the ELL C-terminal occludin-like domain bound to the AFF1 ELLBow-equivalent segment. The ELL2/AFF4 pair crystallised at 2.00 Å (5JW9), so the ELL/AFF1 pair is not obviously intractable, and it is the single missing structure that the KMT2A::AFF1 story most needs. It would also test whether the ELL and ELL2 occludin domains present distinguishable surfaces — the prerequisite for any paralogue-selective agent.
  2. Use ELL as a tool, not a target. A degron-tagged ELL in KMT2A::AFF1 lines would separate the acute transcriptional consequence of losing ELL from the 21-day proliferation phenotype, and would establish whether the fusion's target genes are more ELL-sensitive than the housekeeping transcriptome. That is the experiment that would tell us whether the -0.177 selectivity has a mechanism behind it or is n=10 noise.
  3. Do not pursue ELL as a direct therapeutic target. A 0.94 depleted fraction across all lineages is disqualifying on its own.
Chemical matter: three rows, all of them one artefact

ChEMBL v34 lists an ELL single-protein target, CHEMBL5725004 ('RNA polymerase II elongation factor ELL'), with 3 activity rows at pChEMBL ≥ 6 and no curated mechanism-of-action drug. The best value is a Kd of 62 nM — for molibresib (CHEMBL1232461), a BET bromodomain inhibitor 4.

That is the molibresib artefact in its purest form. Molibresib is not an ELL ligand; it captures BRD4 and co-enriches the entire super elongation complex, so ELL shows up in the affinity-capture readout at an apparent 62 nM. The same single chemoproteomics document is responsible for the 'potent' rows on AFF1 (13 nM), AFF4, PAF1 (65 nM), SETD1A and ASH2L. Six genes in this interactome would be scored as ligandable on the strength of one BET inhibitor's pulldown.

ELL has no chemical matter at all. Combined with no structure of any ELL interface, ELL is currently the least tractable deep dependency in the cluster — which is exactly why the paralogue-selectivity question in section 2 is worth the cost of a crystal.

Expression and per-line detail

HPA: RNA low tissue specificity, detected in all tissues, blood-cell annotation low immune cell specificity, detected in many, protein-level 'group enriched', nucleoplasmic 1. There is no tissue-restriction to exploit — consistent with ELL carrying the general elongation function.

Per-line Chronos across the ten KMT2A-rearranged lines is uniformly deep and shows no fusion-partner pattern that survives inspection: THP-1 (MLLT3) -2.030, MV4-11 (AFF1) -1.936, NOMO-1 (MLLT3) -1.646, SEM (AFF1) -1.613, RS4;11 (AFF1) -1.400, HB11;19 (MLLT1) -1.394, SHI-1 (AFDN) -1.296, MOLM-13 (MLLT3) -1.283, SEMK2 (AFF1) -1.198, MOLM-14 (MLLT3) -0.855 1. The KMT2A::AFF1 lines span -1.936 to -1.198, i.e. the AFF1-fused group is not tight; the -1.537 group mean is driven substantially by MV4-11.

Epigenetics: ELL has no mark-level data at all, and the paralogue that does is not the fusion partner

ELL places no mark. Its function is kinetic - raising the catalytic rate of Pol II by suppressing transient pausing - and it has no chromatin-modifying or chromatin-reading module. The useful content of its epigenetic column is therefore a measured absence plus one complex-level result, and a sharp contrast with ELL2.

The complex-level result, which is more relevant than it looks. The transforming 90-amino-acid C-terminal domain of AF9, immunopurified from myeloblastic M1 cells, associates with Enl, Af4, Laf4, Af5q31, ELL and Af10; the complex - termed elongation assisting proteins (EAP) - also contains P-TEFb and the H3K79 methyltransferase Dot1L. Myeloid cells transformed by MLL fusions show higher levels and a broader distribution of EAP components at Hoxa9 and Meis1; inhibiting P-TEFb and inhibiting Dot1l each reduce Hoxa9 and Meis1 activation; and EAP, which normally dissociates from those loci during differentiation, has its dissociation prevented in the presence of MLL fusion proteins E6. So ELL sits in a DOT1L-containing assembly at the two loci this tranche amended first, and the fusion's effect is described as a failure to let go.

Note the component list, because it settles something on a sibling entry. EAP contains Af10. The gene-aff1 amendment in this tranche records a contradiction - Bitoun's Af4/Af9/Enl/Af10 complex mediates H3K79 methylation by recruiting Dot1, while reconstituted AF9·AF4·P-TEFb does not interact with DOT1 - and offers the reconciliation that AF10, the annotated DOT1L binder, is present in the positive complexes and absent from the reconstitution. EAP is a third DOT1L-containing, AF4-containing, Af10-containing purification, which is consistent with that reading. It remains an inference: no experiment here removes Af10 and asks whether Dot1L stays.

The measured absence, which is the honest headline. ELL[tiab] AND H3K79[tiab] returns exactly one record, and it is the reconstitution paper - so there is no ELL-specific H3K79 measurement in the indexed literature at all. ELL[tiab] AND histone[tiab] AND elongation[tiab] returns 7, and title-checking finds SEC and PAFc reviews, Drosophila heat-shock work, a fission-yeast Epe1 paper and a zebrafish Eaf study - no ELL mark-level result. Both counts were title-checked rather than taken raw, because ELL is a three-letter token and short word-like strings inflate counts in both directions (this project has measured 57-fold inflation for RD as a NELFE alias).

And the paralogue contrast, which this entry already makes on dependency and structure, extends to chromatin - inverted. ELL2, not ELL, has the only locus-level H3K4/H3K79 data in the family: ELL2 knockdown alters both marks on the IgH gene and impairs MLL and Dot1L occupancy there, in mouse B and plasma cells. ELL2, not ELL, is the paralogue in the solved Pol II elongation complex and in the AFF4-bound occludin structure. ELL has the dependency, the fusion and the leukaemia; ELL2 has the structures and the chromatin data. Every part of this entry now says the same thing from a different angle, and the epigenetic column is the sharpest version of it: the protein that is actually fused to KMT2A is the one nobody has measured a histone mark for.

Facts and figures
uniprotP55199
gene symbolELL
length AA621
protein nameRNA polymerase II elongation factor ELL
alphafold IDAF-P55199-F1
alphafold verifiedtrue
PDB verified against rcsbtrue
subnuclear localizationCajal bodies (coilin-positive) by the marker-controlled experiment, described as nuclear speckles in the same group's earlier unmarked images; both are carried by UniProt and the contradiction is unresolved
shown for fusionYES - MLL-ELL, t(11;19)(q23;p13.1): the fusion DISPERSES the compartment. EAF1 goes diffuse in transfected HeLa and is undetectable as foci in murine MLL-ELL leukaemia cells, p80 coilin is delocalised, and nuclear EAF1 and coilin are reduced on fractionation (PMID 11418481, 12686606). Not shown for KMT2A::AFF1 or any B-lineage fusion.
epigenetic readoutNO MARK, AND NO LOCUS-LEVEL MARK DATA EITHER - the contrast with its paralogue is the entry. ELL writes, erases and reads nothing: it increases the catalytic rate of Pol II by suppressing transient pausing and carries no SET, JmjC, HAT, HDAC or reader module. Its only chromatin-level evidence is COMPLEX MEMBERSHIP plus OCCUPANCY BREADTH, not a mark. The transforming 90-amino-acid C-terminal domain of AF9, immunopurified from myeloblastic M1 cells, associates with Enl, Af4, Laf4, Af5q31, ELL and Af10, and the resulting complex - 'elongation assisting proteins', EAP - also contains P-TEFb (Cdk9/cyclin T1/T2) AND the H3K79 methyltransferase DOT1L; myeloid cells transformed by MLL fusions show HIGHER LEVELS AND A BROADER DISTRIBUTION of EAP components at Hoxa9 and Meis1, EAP dissociates from those loci during differentiation, and in the presence of MLL fusion proteins its dissociation is PREVENTED; inhibition of P-TEFb and of Dot1l each reduce Hoxa9 and Meis1 activation (PMID 20854876). So ELL travels in a DOT1L-containing complex at the two loci this tranche has just written up - and NOTE THE COMPONENT LIST CONTAINS Af10, which is the same reconciliation recorded on gene-aff1 for the AFF1-DOT1L contradiction. BOUNDED: ELL[tiab] AND H3K79[tiab] returns exactly 1 record and it is the reconstitution paper (PMID 21896721), i.e. there is no ELL-specific H3K79 measurement at all; ELL[tiab] AND histone[tiab] AND elongation[tiab] returns 7, title-checked, with no ELL mark-level result among them. Precision warning: ELL is a three-letter token and a short, word-like string, so every count here was title-checked rather than taken raw. KEEP DISTINCT FROM ELL2: ELL2 - not ELL - has the only locus-level H3K4/H3K79 data in this family (PMID 21832080, IgH in mouse B and plasma cells), and ELL2 - not ELL - is the paralogue in the solved Pol II elongation complex and the solved AFF4-bound occludin domain. The fusion partner, however, is ELL. CAUSE OR CONSEQUENCE: not answerable for ELL, because no ELL perturbation has been paired with a histone-mark readout.
most depleted kmt2ar lines
THP-1-2.030THP-1: -2.03MV4-11-1.936MV4-11: -1.936NOMO-1-1.646NOMO-1: -1.646SEM-1.613SEM: -1.613
domains
[
  {
    "name": "Occludin-homology / ELL_N region (helical domain)",
    "start": 200,
    "end": 292,
    "note": "the only structurally characterised part of human ELL; NMR, PDB 2DOA; not the AFF-binding module"
  },
  {
    "name": "C-terminal occludin-like domain (AFF-family ELLBow-binding), by homology to ELL2 519-640",
    "start": null,
    "end": null,
    "note": "NEVER SOLVED for ELL. Position inferred from the ELL2 paralogue only; we do not assign residue numbers we cannot verify."
  }
]
PDB IDS
[
  "2DOA"
]
key structures
[
  {
    "pdb": "2DOA",
    "what": "ELL 200-292 helical domain, isolated",
    "method": "NMR",
    "resolution_A": null,
    "pmid": null
  }
]
structural gaps
[
  "No ELL-AFF1 structure exists — the interaction that places ELL in the KMT2A::AFF1 complex has never been solved.",
  "No ELL-AFF4 structure exists; the AFF4 ELLBow has been solved only against ELL2 (5JW9, 2.00 A).",
  "No ELL-EAF1 or ELL-EAF2 structure exists; the activation module was solved as ELL2-EAF1 (7OKX/7OKY).",
  "ELL appears in no RNA polymerase II elongation complex structure; 7OKX/7OKY use ELL2."
]
essentiality
{
  "depmap_common_essential": true,
  "chronos_mean": -1.302,
  "n_lines": 1208,
  "fraction_lines_depleted": 0.94,
  "chronos_mean_leukemia": -1.288,
  "chronos_mean_kmt2ar": -1.465,
  "chronos_mean_kmt2a_aff1": -1.537,
  "selectivity_vs_leukemia": -0.177,
  "selectivity_vs_all_lines": -0.163,
  "selectivity_rank_in_interactome": 7,
  "selectivity_interpretation": "Negative and monotone in the right direction (all lines -1.302 -> leukemia -1.288 -> KMT2A-r -1.465 -> KMT2A::AFF1 -1.537), placing ELL rank 7 on selectivity. But -0.177 is only marginally outside the +/-0.15 noise band at n=10, and it is roughly one ninth the size of the baseline requirement. Common-essential plus a -1.302 pan-cancer mean is a therapeutic-index wall, not a chemistry problem.",
  "gnomad_pLI": 0.9999,
  "gnomad_loeuf": 0.417,
  "gnomad_obs_lof": 18,
  "gnomad_exp_lof": 64,
  "gnomad_mis_z": 1.06,
  "hpa_rna_tissue_specificity": "Low tissue specificity",
  "hpa_rna_tissue_distribution": "Detected in all",
  "hpa_rna_blood_cell_specificity": "Low immune cell specificity",
  "mouse_ko": "not established in this node; no verified citation obtained this cycle"
}
paralogue contrast
{
  "gene": "ELL2",
  "uniprot": "O00472",
  "chronos_mean": -0.064,
  "fraction_lines_depleted": 0.01,
  "chronos_mean_kmt2ar": -0.043,
  "chronos_mean_kmt2a_aff1": 0.003,
  "selectivity_vs_leukemia": -0.01,
  "depmap_common_essential": false,
  "pdb_ids": [
    "7OKX",
    "7OKY",
    "5JW9",
    "2E5N"
  ],
  "note": "ELL2 has the structures (a genuine Pol II elongation complex at 3.30 A, plus the AFF4-bound occludin domain at 2.00 A) and is not a KMT2A-r dependency. ELL has the dependency and one NMR fragment. Knockout of ELL is not buffered by ELL2."
}
most depleted kmt2ar lines — the numbers behind the chart
[
  {
    "line": "THP-1",
    "fusion": "KMT2A::MLLT3",
    "chronos": -2.03
  },
  {
    "line": "MV4-11",
    "fusion": "KMT2A::AFF1",
    "chronos": -1.936
  },
  {
    "line": "NOMO-1",
    "fusion": "KMT2A::MLLT3",
    "chronos": -1.646
  },
  {
    "line": "SEM",
    "fusion": "KMT2A::AFF1",
    "chronos": -1.613
  }
]
chemical matter
{
  "chembl_target": "CHEMBL5725004",
  "chembl_pref_name": "RNA polymerase II elongation factor ELL",
  "n_activity_rows_pchembl_ge_6": 3,
  "best_activity": {
    "molecule": "CHEMBL1232461",
    "molecule_name": "molibresib",
    "type": "Kd",
    "value": 62,
    "unit": "nM",
    "assay_type": "B"
  },
  "curated_mechanism_drugs": [],
  "note": "No genuine chemical matter. All potent rows trace to molibresib, a BET bromodomain inhibitor whose apparent affinity for ELL (and for AFF1, AFF4, PAF1, SETD1A, ASH2L) is super-elongation-complex co-enrichment via BRD4 in a single chemoproteomics document, not direct binding."
}
interfaces
[
  "interface-aff4-ellbow-ell2-occludin"
]
pointer repairs
[
  "2026-09-20: interface references re-pointed to their existing type-correct nodes; the originals were never-written ids, not missing biology"
]
localization
{
  "subcellular": "nucleus (UniProt P55199, ECO:0000269 from PMID 11418481 and 12446457)",
  "subnuclear": "CONTESTED. Cajal bodies, p80-coilin-positive and marker-controlled, transcription-dependent (dispersed by actinomycin D, DRB, alpha-amanitin) (PMID 12686606); the same group described the same endogenous ELL/EAF1 foci as 'nuclear speckles' with no speckle marker (PMID 11418481, 12446457), and UniProt carries both compartments. Also Cajal and histone locus bodies at snRNA genes as part of the Little Elongation Complex (PMID 23932780).",
  "how_shown": "indirect immunofluorescence and confocal microscopy of endogenous ELL and EAF1 with in-house antisera, co-stained with rabbit 288 anti-p80 coilin; Pol II inhibitor dispersal; nuclear/cytoplasmic fractionation",
  "systems": "human HeLa (transfected) and murine MLL-ELL transplant leukaemia lines; no human leukaemia specimen",
  "contradiction": "speckle versus Cajal body, unresolved in the primary literature and unremarked anywhere; resolving experiment is one three-colour stain of ELL with coilin and with SON/SRRM2",
  "node": "localization-ell-cajal-bodies-versus-speckles-and-the-mll-ell-disruption"
}
cell type
{
  "evidence_class_key": "concept-expressed-required-maintains-three-evidence-classes",
  "identifier_note": "the alias ELL1 is safe but the query string ELL is short and word-like; the alias ELL2 was included in the bounded query deliberately and its hits were title-checked, because ELL2 is a different gene (brief: alias precision).",
  "expressed_in": {
    "class": 1,
    "summary": "ubiquitous"
  },
  "required_in": {
    "class": 2,
    "floor_compressed": true,
    "pct_all_lines_depleted": 93.9,
    "all_1258_lines_mean_gene_effect": -1.298,
    "kmt2a_rearranged_lines_mean_gene_effect": -1.481,
    "myeloid_kmt2ar_mean": -1.526,
    "blineage_kmt2ar_mean": -1.401,
    "verdict": "PAN-ESSENTIAL. Depleted below -0.5 in 93.9% of 1,258 lines with a mean of -1.298, the second strongest floor in the worklist after KAT8. No cell-type information is available from DepMap."
  },
  "context": "wild-type protein housekeeping; no fusion-context requirement measurement",
  "graph_links": []
}
cell state
{
  "maintains": {
    "class": 3,
    "states": [],
    "claim": "none established"
  },
  "bounded_query": {
    "date": "2026-09-20",
    "tool": "NCBI esearch db=pubmed",
    "query": "(ELL[tiab] OR \"ELL1\"[tiab] OR \"ELL2\"[tiab]) AND <cell-state clause> AND <leukaemia clause>",
    "count": 16,
    "on_topic_after_title_check": 0,
    "controls": {
      "gene_clause_alone": 923,
      "state_clause_alone": 96743
    },
    "what_the_16_actually_are": "twelve are measurable-residual-disease monitoring or fusion-detection methods in which MLL-ELL is an RT-PCR target; the remainder are transplantation outcome series and reviews. Not one measures a cell state that ELL maintains.",
    "reading": "ELL has never been placed in a defined cell STATE, in leukaemia or in normal haematopoiesis. Its entire leukaemia cell-state literature by keyword is the use of its own fusion transcript as a residual-disease marker - the gene as a MEASUREMENT of a state rather than a maintainer of one. Zero on-topic in titles and abstracts; one full text opened per rule 80 and the absence held.",
    "on_topic_after_opening_full_texts": 0
  },
  "full_texts_opened": [
    "34431785 (PMC8387021, 115,752 characters, opened 2026-09-20) - contains 7 colony-forming and 5 self-renewal occurrences, but MLL-ELL there is the CONSTRUCT used to map MLL's own THD2/HBO1 requirement. No cell-state function is attributed to ELL itself anywhere in it."
  ],
  "what_the_full_text_showed": "MLL-ELL is used as an immortalisation substrate in self-renewal assays; the variable under test is MLL's trithorax homology domain 2 and its HBO1/KAT7 contact, not ELL."
}
depmap kmt2a selectivity
{
  "chronos_mean_kmt2a_rearranged": -1.4808,
  "chronos_mean_other_leukaemia": -1.2275,
  "difference": -0.2533,
  "pan_essentiality_all_lines": -1.3018,
  "selectivity_rank": 153,
  "rank_denominator": 17728,
  "provenance": "Joined 2026-09-21 from analyses/depmap-kmt2a-selective-scan-full (generated 2026-09-19T10:12:06Z), the project's own genome-wide scan over 11 KMT2A-rearranged and 43 other leukaemia DepMap lines. NOT re-fetched and NOT independent of that scan: every caveat it carries applies here, including that the KMT2A-rearranged arm is a small number of lines drawn from nine individuals, that the rank denominator is 17,728 genes, and that a difference is not a dependency. Added because only 26 of 308 gene entries carried this number, which made any KMT2A-selectivity question answerable for 8% of the corpus.",
  "how_to_read": "A negative difference means MORE required in KMT2A-rearranged lines than in other leukaemia lines. A rank near 17,728 is ANTI-selective. Neither is evidence of a fusion-driven dependency on its own: check pan_essentiality_all_lines before calling anything selective, and note that a paralogue-redundant gene can score near zero for reasons that have nothing to do with this disease - see ana-the-paralogue-redundancy-excuse-is-not-tested-and-my-proxy-decided-the-answer."
}
epigenetic bounded absence
{
  "claim": "No ELL-specific histone-mark measurement exists in the title/abstract index; ELL's only chromatin-level evidence is complex membership and occupancy breadth.",
  "date_measured": "2026-09-21",
  "tool": "NCBI esearch db=pubmed (eutils), querytranslation read back from each response",
  "queries": [
    {
      "query": "ELL[tiab] AND H3K79[tiab]",
      "querytranslation": "\"ELL\"[Title/Abstract] AND \"H3K79\"[Title/Abstract]",
      "count": 1,
      "title_checked": "the single hit is PMID 21896721, the defined-complex reconstitution study, which measures protein-protein association and not a mark. 0 ELL-specific mark measurements."
    },
    {
      "query": "ELL[tiab] AND histone[tiab] AND elongation[tiab]",
      "querytranslation": "\"ELL\"[Title/Abstract] AND \"histone\"[Title/Abstract] AND \"elongation\"[Title/Abstract]",
      "count": 7,
      "title_checked": "PMIDs 37002435 (zebrafish Eaf1/2 erythropoiesis), 33533152 (fission-yeast Epe1 and ELL outside heterochromatin), 24607507 (Drosophila HDAC3 and hsp70), 21896721 (the reconstitution study), 21460034 (SEC and MLL review), 20854876 (EAP occupancy, now cited here), 17001490 (Drosophila Cdk9). NO ELL mark-level result."
    },
    {
      "query": "ELL2[tiab] AND H3K79[tiab]",
      "querytranslation": "\"ELL2\"[Title/Abstract] AND \"H3K79\"[Title/Abstract]",
      "count": 2,
      "role": "POSITIVE CONTROL THROUGH THE IDENTICAL SHAPE, and it is the paralogue: 2 records, one of which (PMID 21832080) is a genuine locus-level H3K4/H3K79 result at IgH. The instrument returns ones for an ELL-family protein; it returns none for ELL."
    }
  ],
  "precision_warning": "ELL is a THREE-LETTER TOKEN and a short, word-like string. This project has measured 57-fold count inflation for the short NELFE alias 'RD' and 15-fold for EAF2's 'U19' and 'TRAITS'. Every count above was title-checked and the checked number is what is recorded; the raw counts happened to be small, but the bound on this absence is the title-check and not the count.",
  "absence_class": "never-measured",
  "absence_class_basis": "Bounded by titles and abstracts only; supporting figures not searched (rule 80), which matters more than usual here because ELL occupancy and mark ChIPs are exactly the kind of supporting panel rule 80 was written about. The paralogue control returning a positive shows the query shape works."
}
How this could bite us back
  • Essential-function toxicity is the defining problem, not a caveat. ELL is DepMap common-essential with 94% of 1,208 lines depleted and a pan-cancer mean of -1.302; the KMT2A-r selective increment is -0.177, about one ninth as large. Any systemic agent potent enough to matter in the tumour would deplete proliferating normal tissue, and no improvement in compound quality changes that ratio.
  • There is nothing to design against. The only human ELL structure is one 93-residue NMR fragment (2DOA) that is not an interface, so a structure-based programme would begin by solving the target — and the interface-selective approach that worked for menin requires an interface picture that does not exist for ELL.
  • The paralogue does not buffer in the helpful direction. ELL2 is expressed in the same cells yet fails to rescue ELL loss (ELL2 knockout is neutral, ELL knockout is lethal), so we get the toxicity of hitting a non-redundant gene without the safety of paralogue compensation. Conversely, a pan-ELL/ELL2 inhibitor would add ELL2's normal immune-cell function to the toxicity budget for no efficacy gain.
  • Compensation at the complex level is untested and plausible: ELL docks onto the AFF-family ELLBow, and the AFF1 module inside the fusion may present a distinct affinity from AFF4's. If the fusion holds ELL more tightly, an interface inhibitor will be outcompeted precisely in the cells we care about — the avidity trap that also afflicts the menin interface.
  • Delivery: what is required is either a paralogue-selective inhibitor of an unsolved protein-protein interface or a degrader of a common-essential gene. The first has no structural starting point; the second maximises on-target toxicity in every tissue at once.
  • If ELL's contribution to the fusion programme runs through a DOT1L-containing complex whose release is blocked rather than through anything ELL does itself, then the druggable event is the failure to dissociate during differentiation - a property of the fusion, not of ELL - and inhibiting a common-essential elongation factor with 94% of lines depleted to reach it is the worst available route to that target.
Still open
  1. What does the ELL C-terminal occludin-like domain look like bound to the AFF1 ELLBow-equivalent segment? This is the single missing structure the KMT2A::AFF1 story most needs: the ELL2/AFF4 pair crystallised at 2.00 A (5JW9), so the ELL/AFF1 pair is probably tractable, and it would show whether the ELL and ELL2 occludin surfaces differ enough for a paralogue-selective agent.
  2. Does ELL's -0.177 selectivity have a mechanism, or is it n=10 noise? A degron-tagged ELL in KMT2A::AFF1 lines, reading out nascent transcription at fusion target genes versus the housekeeping transcriptome, would separate an acute fusion-specific effect from a general elongation effect. Nobody has published this.
  3. Why is ELL common-essential while ELL2 is dispensable, when both carry the ELL-family elongation activity? If the answer is expression level rather than intrinsic function, then forcing ELL2 expression should rescue ELL knockout — and if it does, ELL's deep essentiality is not a property of the protein and the index calculation changes.
  4. ELL's missense z is only 1.06 against a LOEUF of 0.417: it tolerates substitution far better than truncation. Does that mean there is no single pocket whose occupancy inactivates it, i.e. that ELL is intrinsically a degrader target rather than an inhibitor target?
  5. Is there any ELL function that is fusion-specific rather than general? Everything verified about ELL-family mechanism comes from ELL2 in a reconstituted complex containing no fusion protein, so the honest answer today is that we do not know what ELL does differently in a KMT2A::AFF1 cell.
  6. Has any histone mark ever been measured under ELL perturbation? The bounded search says no, in titles and abstracts. ELL is the KMT2A fusion partner in t(11;19)(q23;p13.1), it is common-essential, and it sits in a DOT1L-containing complex at Hoxa9 and Meis1 - and its paralogue, which is neither a fusion partner nor a dependency, is the one with the chromatin data. An H3K79me2 ChIP after acute ELL degradation in a KMT2A::ELL line is an obvious and unmade measurement.
  7. Can an Af4/Af9/Enl/ELL complex methylate H3K79 without Af10? The EAP purification contains Ell, Af4 AND Af10 alongside Dot1L, making it a third Af10-containing positive against one Af10-free reconstitution negative. Removing Af10 from EAP and asking whether Dot1L stays is the experiment that would settle the contradiction recorded on gene-aff1, and it would do so for ELL and AFF1 at once.
Every source behind this entry (7)
  1. E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

    ELL is DepMap common-essential with a mean Chronos of -1.302 across 1,208 cell lines and 94% of lines depleted; means are -1.288 across 64 leukemia lines, -1.465 across 10 KMT2A-rearranged lines and -1.537 across 4 KMT2A::AFF1 lines, giving selectivity_vs_leukemia -0.177 and selectivity_vs_all_lines -0.163; gnomAD gives pLI 0.9999, LOEUF 0.417, 18 observed versus 64 expected loss-of-function variants and missense z 1.06; HPA classifies it as low tissue specificity, detected in all tissues, low immune cell specificity. The paralogue ELL2 is not a dependency: mean Chronos -0.064, 1% of lines depleted, KMT2A::AFF1 mean +0.003, selectivity_vs_leukemia -0.010.

    What this evidence cannot show

    Means were computed by us from per-line API values rather than read from a DepMap release table, and the API does not state the release version. Only 10 KMT2A-rearranged lines (4 KMT2A::AFF1) contribute to the selective score; -0.177 is only marginally outside our declared +/-0.15 noise band and has not been bootstrapped. The KMT2A::AFF1 group mean of -1.537 spans -1.936 (MV4-11) to -1.198 (SEMK2), so it is not a tight group. Cell lines are not normal tissue, and DepMap common-essentiality in immortalised proliferating lines is not the same as a requirement in quiescent normal tissue. Whole-gene knockout removes the occludin-like SEC-docking domain and the elongation-stimulating activity together, and so overstates what an interface inhibitor would do.

    Evidence that argues against this (1)
    • The selective increment is dwarfed by the baseline: a 0.94 depleted fraction and a -1.302 pan-cancer mean mean that the therapeutic window implied by -0.177 of extra sensitivity is not a window a systemic agent could exploit. This is the single strongest argument against ELL as a direct target, and it comes from the same dataset that makes ELL look attractive on rank. Broad Institute DepMap; Open Targets Platform 2026
  2. E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database

    The only human ELL structure in the PDB is 2DOA, an NMR solution structure of an isolated helical domain spanning ELL residues 200-292; there is no ELL-AFF1 structure, no ELL-AFF4 structure, no ELL-EAF structure and no ELL in any Pol II elongation complex. The single structurally characterised fragment of the paralogue ELL2 outside the elongation complex, 2E5N (NMR, ELL2 200-292, the ELL_N2 domain), covers the same span, and neither corresponds to the C-terminal occludin-like domain (ELL2 519-640) that binds AFF-family proteins.

    What this evidence cannot show

    This is an absence-of-evidence claim scoped to what a search of the RCSB APIs returned in this session; a structure deposited but not yet indexed, or one where ELL is present as an unannotated entity, would not be caught. Residue ranges are RCSB's UniProt-mapped modelled segments, not independently curated domain boundaries. 2DOA is an isolated domain with no partner and carries no functional or affinity information.

    Evidence that argues against this (1)
    • A solution NMR structure of an isolated domain is the weakest structural evidence in this dossier class: the construct is chosen to be soluble and monomeric, the interaction partners are absent, and the model is an ensemble rather than a density map. The absence the entry lists — no ELL-AFF1, no ELL-AFF4, no ELL-EAF structure — is the substantive finding and should be the headline: every interface this protein is discussed through is unresolved. RCSB PDB 2026
  3. E3established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper

    Cryo-EM structures of an RNA polymerase II-DSIF active elongation complex containing ELL2 (residues 1-640) and EAF1 (1-268) define how an ELL-family protein and its EAF partner engage the polymerase and allosterically stimulate transcription: 7OKX at 3.30 A and 7OKY at 4.14 A. These use ELL2, not ELL.

    Allosteric transcription stimulation by RNA polymerase II super elongation complex · Chen Y, Vos SM, Dienemann C, et al. · Molecular Cell 2021
    PMID 34265249DOI
    What this evidence cannot show

    This is a reconstituted in vitro elongation complex, and the ELL-family subunit in it is ELL2 rather than ELL, so every inference it supports about ELL is a paralogue extrapolation. The mechanistic conclusion concerns general allosteric stimulation of Pol II, not anything KMT2A-fusion-specific, and the complex does not contain a KMT2A fusion protein.

    Evidence that argues against this (1)
    • The cryo-EM complex contains ELL2 and EAF1, not ELL, so this describes a paralogue. ELL and ELL2 differ in expression, in partner preference and in their SEC membership, so the architecture transfers only as a hypothesis. RCSB PDB 2026
  4. E4strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.database

    ChEMBL v34 contains no genuine chemical matter against ELL: the single-protein target CHEMBL5725004 has 3 activity rows at pChEMBL >= 6 and no curated mechanism-of-action drug, and the best row (Kd 62 nM) is the BET bromodomain inhibitor molibresib (CHEMBL1232461) from a chemoproteomics document in which the compound captures BRD4 and co-enriches the super elongation complex, making it a co-enrichment readout rather than direct binding.

    What this evidence cannot show

    Distinct-compound counts are page-local because of a 25-row API limit; with 3 rows this is immaterial. Absence from ChEMBL is not absence from the world and curation lags the literature. The molibresib measurement is not wrong, only mis-attributed: the compound genuinely enriches ELL, via BRD4.

    Evidence that argues against this (1)
    • The quoted affinity is an OFF-TARGET measurement. A chemoproteomics document reports what a compound pulled down, so a nanomolar number attributed to this protein from a BET bromodomain probe is evidence that the compound binds something else and touched this too. It is not chemical matter and should not enter a tractability count. ChEMBL, EMBL-EBI 2026
  5. E5strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    ELL enters the super elongation complex as part of the ELL/P-TEFb elongation complex whose AFF4 subunit is a shared subunit of MLL chimeras, and AFF4 is required for the stability of that ELL-containing complex.

    PMID 20159561DOI
    What this evidence cannot show

    Establishes complex membership and AFF4-dependent stability by biochemistry and knockdown; it does not establish the ELL-AFF1 contact within a KMT2A::AFF1 complex, and the associated proteomics is pan-MLL-chimera rather than KMT2A::AFF1-specific. No affinity measurement is reported.

    Evidence that argues against this (1)
    • Complex composition is not requirement. The AEP/SEC papers establish what co-purifies and what a reconstituted complex does; they do not establish that each subunit is rate-limiting in a patient's leukaemia, and this project's own dependency data separates the members by an order of magnitude. AFF4 being required for the complex's stability also means an ELL measurement inside that complex reports on AFF4 as much as on ELL. Nature 2012
  6. E6moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    ELL TRAVELS IN A DOT1L-CONTAINING COMPLEX AT HOXA9 AND MEIS1, AND THE FUSION PREVENTS ITS RELEASE: the 90-amino-acid C-terminal transforming domain of AF9, immunopurified after expression in myeloblastic M1 cells, associates with Enl, Af4, Laf4, Af5q31, Ell and Af10, and this 'elongation assisting proteins' (EAP) complex also contains the Pol II CTD kinase Cdk9/cyclin T1/T2 (P-TEFb) and the histone H3 lysine 79 methyltransferase Dot1L; myeloid cells transformed by MLL fusions show higher levels and a broader distribution of EAP components at Hoxa9 and Meis1, inhibition of P-TEFb and of Dot1l each contribute significantly to reduced Hoxa9 and Meis1 activation, EAP dissociates from those loci during induction of differentiation, and in the presence of MLL fusion proteins its dissociation is prevented.

    PMID 20854876DOI
    What this evidence cannot show

    Abstract read from the eutils record; full text not read in this tranche, and the paper is already cited once here for a different purpose (target-cdk9-inhibition-kmt2a-r). MURINE myeloblastic M1 cells and mouse MLL-AF9/MLL-ENL systems - not the index fusion and not B-lineage, so rule 73 applies. The bait is the AF9 C-terminal domain, NOT ELL and NOT a KMT2A fusion protein, so ELL's presence is co-purification with an overexpressed partner domain and establishes complex membership rather than a contact. No histone mark is measured: the ChIP-qPCR readout is occupancy of EAP components, and 'broader distribution' is an occupancy statement, not an H3K79me2 measurement. Rule 92 applies - a focused complex purification reports many partners of one bait by design.

    Evidence that argues against this (1)
    • A component list from an affinity purification is not a complex, and the reconstitution study exists precisely to make that point about these proteins: complexes purified through AF9, AF4 and ELL have overlapping components, different elongation activities and different cofactor associations that 'suggest dynamic interactions', and in defined minimal complexes DOT1 forms a stable complex with AF9 and does NOT interact with AF9-AF4-P-TEFb. So EAP's co-occurrence of Ell, Af4 and Dot1L in one pulldown is compatible with two or more separate assemblies sharing a bait, and cannot establish that ELL and DOT1L are in the same particle. The Af10 reconciliation offered in this graph's gene-aff1 entry is an inference from component lists and is not stated by either paper. Biswas D et al. 2011
  7. E7strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    THE PARALOGUE CONTRAST AT THE CHROMATIN LEVEL, AND IT IS THE OTHER ONE: ELL2 - not ELL - carries the only locus-level histone-mark evidence in this family. siRNA reduction of ELL2 in plasma cells influenced the methylations of histone H3K4 and H3K79 on the IgH gene and impaired the associations of MLL and Dot1L with that gene, and ChIP on mouse B versus plasma cells bearing an identical IgH gamma-2a gene showed H3K4 and H3K79 methylation extending farther downstream in plasma cells, past the IgH enhancer to the end of the transcribed region. No equivalent measurement exists for ELL.

    PMID 21832080DOI
    What this evidence cannot show

    Abstract read from the eutils record; full text not read. Recorded on the ELL entry specifically to keep the paralogues separate, which this beat was briefed to do: the system is MOUSE B and plasma cells in normal secretory differentiation, the protein is ELL2, and the MLL is wild-type MLL1 - none of it is evidence about ELL, about a KMT2A fusion or about leukaemia. The absence of an ELL equivalent is bounded only by title/abstract searching: ELL[tiab] AND H3K79[tiab] returns 1 record (the reconstitution paper) and ELL[tiab] AND histone[tiab] AND elongation[tiab] returns 7 with no ELL mark-level result on title-checking, and ELL is a three-letter token whose short form makes both recall and precision unreliable, so the bound is weaker than the counts suggest.

    Evidence that argues against this (1)
    • The paralogues are treated as interchangeable occupants of one SEC slot in the founding biochemistry - purification of multiple KMT2A chimeras identifies a super elongation complex containing ELL, P-TEFb and AFF4 - so the fact that a mark was measured for ELL2 and not for ELL may reflect which protein each field happened to study (plasma-cell immunoglobulin processing versus leukaemia) rather than a real functional difference. Against that, this entry's own dependency data show the paralogues are emphatically NOT interchangeable in practice: ELL is common-essential at -1.302 with 94% of lines depleted while ELL2 is -0.064 with 1%. The honest position is that the chromatin difference is untested, not established. Lin C et al. 2010
Connected to