Anti-ELL antibodies carry a confound the other partners do not: two paralogues and a protein whose abundance is actively regulated
also called anti-ELL attribution · ELL1 ELL2 ELL3 cross-reactivity · ELL stability confound
ELL, the KMT2A partner in t(11;19)(q23;p13.1), has two close relatives in the same cell and its own amount is turned up and down by other proteins. So an ELL antibody in a KMT2A::ELL leukemia is not just seeing the normal protein plus the fusion - the background it is being compared against is itself moving.
The same thing, for a scientist
KMT2A::ELL retains the ELL C-terminal occludin-homology domain, so anti-ELL-C reagents bind the fusion and wild-type ELL alike - the standard shared-epitope problem. Two features make ELL worse than AFF1 or MLLT1. **First, paralogues.** ELL (ELL1) has the paralogues ELL2 and ELL3, all SEC-associated; an antibody's specificity against the paralogues is a reagent property that is rarely reported and never controlled for in this disease, and a cross-reactive antibody adds species that have nothing to do with the fusion. **Second, regulated abundance.** Wild-type ELL levels are set by an acetylation-ubiquitylation switch: p300-mediated site-specific acetylation stabilises ELL, HDAC3-mediated deacetylation destabilises it through Siah1-dependent polyubiquitylation, and DBC1 competes with HDAC3 for the same sites [E1]. ELL also partitions dynamically between the Little Elongation Complex and the SEC under the control of K355 acetylation, moving with mitogen and genotoxic signalling [E2]. So the wild-type pool an ELL antibody is reading is not a constant background: it changes with cell state, with drug treatment and with DNA damage. Any experiment that compares an ELL signal before and after a perturbation is measuring the fusion plus a moving wild-type term. On the favourable side, 10 of 119 ELL fusions are reciprocal-negative spliced fusions [E3], giving a small natural comparator arm.
Why regulated abundance is an attribution problem and not just noise
In the AFF1 and MLLT1 cases the wild-type partner is a fixed, if large, background term; subtracting it is hard but the target does not move. For ELL the background is a function of the perturbation being tested.
Concretely: genotoxic stress drives ATM-mediated ELL phosphorylation and Sin3A/HDAC1-dependent deacetylation of K355, shifting ELL from SEC into LEC 2. Induction chemotherapy is genotoxic stress. So an anti-ELL ChIP or co-IP comparing a KMT2A::ELL sample before and after chemotherapy would show a change in ELL-associated signal that has a complete explanation in wild-type ELL biology and requires no statement about the fusion at all.
The same logic applies to p300 and HDAC inhibitors, both of which are in clinical use and both of which move wild-type ELL abundance directly.
What would fix it
- A paralogue-specificity control. Blot the antibody against ELL2 and ELL3 knockouts or overexpression. This is a one-day experiment that is not in the KMT2A::ELL literature this beat searched.
- A wild-type ELL loading term. Any ELL measurement in a fusion-bearing cell should report the wild-type band alongside the fusion band, so the reader can see whether the background moved. The size dimension makes this free on a blot and impossible in ChIP.
- The reciprocal-negative spliced-fusion arm. Ten of 119 - small, but it holds the partner constant.
- The junction peptide. ELL is the partner where a mass-spectrometry readout is most valuable, because it is the one whose antibody background is least stable.
| reagent class | antibodies against the ELL C-terminal (retained) occludin-homology region |
|---|---|
| why it matters clinically | induction chemotherapy is genotoxic stress, which moves wild-type ELL partitioning - so a before/after ELL measurement in a patient confounds the fusion with a chemotherapy effect on the background |
species recognised
[ "wild-type ELL (ELL1)", "KMT2A::ELL fusion", "potentially ELL2 and ELL3 if the antibody cross-reacts - untested in this disease" ]
distinctive confounds
[ "two paralogues (ELL2, ELL3) with no published specificity control in KMT2A::ELL work", "wild-type abundance actively regulated by p300/HDAC3/DBC1/Siah1", "wild-type LEC-versus-SEC partitioning moves with mitogen and genotoxic signalling" ]
reciprocal negative fraction
{
"n": 10,
"total": 119
}- If wild-type ELL partitioning moves with genotoxic stress, the scheduling of a targeted agent relative to induction chemotherapy could change any ELL-based pharmacodynamic readout in either direction without the drug doing anything - the same scheduling confound this graph already records for AF9 S395.
- HDAC inhibitors and p300 inhibitors both move wild-type ELL abundance directly, so any combination trial pairing them with a KMT2A-directed agent has a built-in biomarker artefact.
- Do the anti-ELL antibodies used in KMT2A::ELL work cross-react with ELL2 and ELL3? No specificity control against the paralogues appears anywhere this beat searched, and it is a one-day experiment.
- Is the ELL portion retained in KMT2A::ELL a substrate for the Siah1-dependent degradation that controls wild-type ELL? If so the fusion's half-life is set by an acetylation switch nobody has measured on it; if not, fusion and wild-type respond oppositely to the same drugs.
- E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
Wild-type ELL abundance is actively regulated: p300-mediated acetylation stabilises it, HDAC3-mediated deacetylation destabilises it via Siah1-dependent polyubiquitylation, and DBC1 competes with HDAC3 for the same binding sites.
DBC1, p300, HDAC3, and Siah1 coordinately regulate ELL stability and function for expression of its target genes. · Basu S, Barad M, Yadav D, et al. · Proc Natl Acad Sci U S A 2020What this evidence cannot show
Worked out in non-leukaemic human cell systems with a focus on glucose-metabolism target genes; whether the same switch operates at the same amplitude in a KMT2A::ELL blast has not been shown. The attribution point does not depend on the amplitude being large, only on the background not being constant.
Evidence that argues against this (1)
- If the same acetylation-ubiquitylation switch acts on the ELL portion retained in the fusion, then fusion and wild-type abundance may move together, in which case a ratio measurement would be more stable than either alone - the opposite of the concern raised here. Nobody has tested whether the fusion is a Siah1 substrate. Nandy A et al. 2026
- E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
ELL partitions dynamically between the Little Elongation Complex and the Super Elongation Complex under the control of K355 acetylation, with mitogen signalling favouring the SEC form and genotoxic stress driving ATM-dependent phosphorylation and deacetylation that favours the LEC form.
Dynamic partitioning of a critical elongation factor between LEC and SEC regulates cellular snRNA and proliferation-related mRNA transcription. · Nandy A, Dalui S, Talukdar P, et al. · Nat Commun 2026What this evidence cannot show
Worked out for wild-type ELL, not for KMT2A::ELL. K355 lies in the ELL N-terminal region; whether it is retained in the fusion depends on the breakpoint and was not resolved here, so the direct relevance is to the wild-type background term rather than to the fusion itself.
Evidence that argues against this (1)
- If K355 is lost from the fusion, the fusion's ELL portion is not subject to this switch and the two species respond differently to the same signal - which makes the background term move independently of the fusion and strengthens rather than weakens the attribution concern. Basu S et al. 2020
- E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
Ten of 119 KMT2A::ELL fusions are spliced fusions in which no reciprocal transcript is expressed, providing a small reciprocal-negative comparator arm with the partner held constant.
The KMT2A recombinome of acute leukemias in 2023. · Meyer C, Larghero P, Almeida Lopes B, et al. · Leukemia 2023What this evidence cannot show
Read verbatim from PMC10169636. Ten cases is too small for most comparisons and the paper notes that, unlike MLLT10 and some others, ELL fusions were not clearly linked to therapy-induced leukaemia - a separate point but a reminder that ELL cases are heterogeneous in origin.
Evidence that argues against this (1)
- The equivalent MLLT1 arm is 241 of 460, an order of magnitude larger, so if a reciprocal-negative comparison is to be run at all it should be run in MLLT1 and the ELL arm is at best a replication cohort. Meyer C et al. 2023
wild-type ELL abundance and complex partitioning are actively regulated, so the background term moves with the perturbation
- related to →KMT2A::ELL (MLL-ELL), t(11;19)(q23;p13.1)
shared C-terminal epitope plus paralogue cross-reactivity, untested in this disease
- related to →Anti-ENL antibodies are the cleanest partner reagent in this disease and sit on top of the worst confound
the MLLT1 reciprocal-negative arm is 241/460 against ELL's 10/119, so ELL is at best a replication cohort
the same ELL/ELL2 conflation, in the structural literature