Measure the one number the whole YEATS-selectivity argument turns on: what fraction of ENL and of AF9 is in SEC, in SEC-L2, in SEC-L3 and free — with an exchange control, because the question may not be well posed
also called reader-pool partition experiment · sequential immunodepletion of the AFF scaffolds
Take a leukaemia cell, pull out each version of the machine one at a time, and count how much of the two reader proteins came out with each. Nobody has ever done this, and every argument about whether a reader-blocking drug is specific depends on the answer.
The same thing, for a scientist
Sequential immunodepletion of AFF1, AFF4, AFF2 and AFF3 from cleared nuclear lysate of a KMT2A::AFF1 line, with quantitative western for ENL/MLLT1 and AF9/MLLT3 in each depleted supernatant against the input, yielding the fraction of each reader held by each scaffold plus a residual free/other pool. **The design's non-obvious requirement is an exchange control**: post-lysis mixing of lysates from cells carrying differentially tagged scaffolds, to show that reader does not redistribute between complexes after lysis. Without it the numbers are uninterpretable, because the AF9 AHD is intrinsically disordered and exchanges between partners, and a dynamic pool would return a well-formed partition that means nothing. The outcome is scored against the pre-registered falsifiers in `hyp-q187-yeats-inhibition-is-scaffold-blind`: **>95% ENL and AF9 removed by AFF1+AFF4 depletion kills the scaffold-blindness consequence; <85% supports it; >30% redistribution in the exchange control means the question was malformed and must be replaced by a residence-time measurement.**
Design
Cells. A KMT2A::AFF1 line — SEM (infant-derived, B-lineage, the lineage-appropriate choice) with MV4;11 as a second line, because MV4;11 is the line the published YEATS-inhibitor deposits used and pairing to them makes the result comparable.
Fractionation. Nuclear extract, benzonase-treated so recovery is not DNA-bridged (the defining SEC-family purification used benzonase for the same reason).
Depletion series. Four sequential rounds, each to exhaustion (three passes over fresh beads, confirmed by absence of the bait in a fourth): anti-AFF1, anti-AFF4, anti-AFF2, anti-AFF3. Quantitative western on every supernatant and every bead eluate for MLLT1, MLLT3, CDK9, and each bait.
Readout. Fraction of input ENL, and separately of input AF9, removed at each step; residual after all four is the free-plus-other pool, which includes DOT1L-bound reader.
Order control. Run the series in reversed order (AFF3, AFF2, AFF4, AFF1) in parallel. If the partition depends on order beyond what sequential depletion arithmetic predicts, exchange is occurring and the exchange control below is the decisive one.
Exchange control, the part that is usually skipped. Lyse cells expressing tagged AFF1 and cells expressing a differently tagged AFF4 separately, mix the lysates after lysis, then deplete. Redistribution of reader into the complex it was never in, in vivo, quantifies post-lysis exchange. Above 30% and the partition is not a property of the cell.
Why this and not something else
Every quantitative form of the scaffold-blindness argument reduces to this number, and every entry in this beat says so independently:
mechanism-q187-target-specificity-is-a-scaffold-property-while-the-reader-is-sharednames it as the largest hole.ana-q187-the-scaffold-blindness-argument-does-not-survive-the-published-therapeutic-windowlists it as route 2 of three by which a composition argument fails to settle selectivity.risk-q187-yeats-inhibition-touches-an-imprinting-complex-in-an-infantdepends on it for severity.hyp-q187-yeats-inhibition-is-scaffold-blindmakes it prediction 1.
It is also the cheapest wet experiment in the beat: antibodies, beads, a gel, no animals, no new reagent synthesis, no sequencing.
What will go wrong
- Antibodies. This corpus already records that two published anti-AFF1 antibodies disagree with each other more than two patients do (
ana-q142-the-two-aff1-antibodies-disagree-more-than-the-two-patients-do). The defining SEC-family purification used Flag-tagged over-expressed AFF2 and AFF3 rather than endogenous IP, which is a strong hint that endogenous antibodies for those two are the weak link. If endogenous AFF2/AFF3 depletion cannot be shown to go to exhaustion, run the tagged-knock-in version instead and say so. - Abundance. If AFF2 and AFF3 protein are as scarce as their mRNA suggests, their depletion steps may remove a fraction of reader below the quantification limit of a western. Pre-specify the limit of detection and report the result as a bound ("under X%") rather than as a point estimate, or the experiment will produce a comfortable zero that means nothing.
- The fusion is itself a scaffold. In a KMT2A::AFF1 line the anti-AFF1 antibody pulls the fusion and wild-type AFF1 together unless the blot resolves them by size. It does — the fusion runs at about 240 kDa and wild-type AF4 at 125 and 145 kDa in t(4;11) lines — so report the fusion-bound reader fraction separately. That sub-number has never been measured either and is arguably the most interesting output of the whole experiment.
| design | sequential immunodepletion of AFF1, AFF4, AFF2 and AFF3 from benzonase-treated nuclear lysate of SEM and MV4;11, each round to exhaustion, with a reversed-order arm and a post-lysis lysate-mixing exchange control using differentially tagged scaffolds |
|---|---|
| readout | fraction of input ENL/MLLT1 and of input AF9/MLLT3 removed at each depletion step by quantitative western, with the fusion-bound fraction resolved from wild-type AFF1 by size (fusion ~240 kDa, wild-type AF4 125 and 145 kDa) |
| controls | depletion-to-exhaustion confirmed by bait absence on a fourth pass; reversed-order series; post-lysis mixing exchange control with a pre-specified 30% redistribution threshold; CDK9 as a positive co-depletion marker; pre-specified western limit of detection so a below-threshold result is reported as a bound and not as a zero |
| cost tier | low |
| timeline | 6-8 weeks assuming antibodies validate; add 3-4 months if tagged knock-ins are needed for AFF2 and AFF3 |
| scores hypothesis | hyp-q187-yeats-inhibition-is-scaffold-blind |
| decision rule | >95% of ENL and AF9 removed by AFF1+AFF4 depletion = hypothesis refuted; <85% = supported; 85-95% = ambiguous, report as such and do not resolve by choosing a side; >30% redistribution in the exchange control = MALFORMED, withdraw and replace with a residence-time question |
| shown for fusion | N/A - this is a proposed experiment; it is designed IN the index fusion (SEM) rather than imported from another system, which is unusual for this beat and deliberate |
measurements
[
{
"quantity": "decision_threshold_reader_removed_by_aff1_aff4_depletion",
"subject": "total ENL and total AF9 in a KMT2A::AFF1 nuclear lysate | sequential anti-AFF1 then anti-AFF4 depletion",
"value": 0.95,
"units": "fraction of input, refutation threshold",
"n": 2,
"universe": "SEM and MV4;11 nuclear lysate, benzonase-treated, quantitative western against input; thresholds pre-registered 2026-09-24 before any data exist",
"null": null,
"note": "Pre-registered, not measured. The support threshold is 0.85 and the band 0.85-0.95 is pre-declared ambiguous so that a mid-range result cannot be read as a verdict."
}
]- This experiment can produce a number that looks decisive and is not. A partition measured in lysate is a partition of complexes that survive lysis, and the complex most likely not to survive is the least abundant one — which is exactly SEC-L2 and SEC-L3. A low sibling fraction is therefore confounded with fragility, and the result should be reported with the depletion efficiencies and the limit of detection alongside it.
- If the answer supports the hypothesis, the immediate temptation is to treat the imprinting liability as established. It would not be: a large sibling reader fraction in a leukaemia cell says nothing about whether the reader is at an imprinted control region in a normal one, which is a separate experiment.
- The experiment is cheap enough that its main risk is being run badly and believed. Without the exchange control it will produce a clean partition whatever the underlying dynamics, and this project has a documented history of instruments that supply an answer the data did not.
- Do validated endogenous ChIP- or IP-grade antibodies exist for human AFF2 and AFF3? The defining paper raised its own antisera against His-tagged AFF2 and AFF3 fragments, which suggests commercial reagents were inadequate in 2012 and may still be.
- What fraction of ENL and AF9 is bound to DOT1L rather than to any AFF scaffold? The residual pool in this design lumps DotCom with free protein; adding an anti-DOT1L round would separate them and costs one more antibody.
- Does the partition differ between the fusion-bound and wild-type-AFF1-bound pools? The blot resolves them by size, so the answer comes free with the experiment, and nobody has ever asked how much of a leukaemic cell's reader is held by the oncoprotein itself.
- E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
The defining SEC-family purification recovered ENL and AF9 with AFF2 and AFF3 by MudPIT and size-exclusion chromatography using Flag-tagged baits in a 293 Flp-In-TRex system, with endogenous co-immunoprecipitation as confirmation, establishing that the reader is present in all three complexes but reporting no stoichiometry and no partition of the cellular reader pool.
The super elongation complex family of RNA polymerase II elongation factors: gene target specificity and transcriptional output. · Luo Z, Lin C, Guest E, Garrett AS, Mohaghegh N, Swanson S, Marshall S, Florens L, Washburn MP, Shilatifard A · Molecular and Cellular Biology 2012What this evidence cannot show
`expected_absent_terms` carries the load-bearing negative: the purification reports composition and not stoichiometry, which is exactly the gap this experiment fills. The use of Flag-tagged over-expressed AFF2 and AFF3 rather than endogenous IP is the practical warning for the design. HEK293-derived, not haematopoietic.
Evidence that argues against this (1)
- An over-expressed Flag bait distorts the very quantity this experiment wants to measure, so the source establishing that the complexes exist cannot be used to predict how much reader each holds in a normal cell. If endogenous AFF2 and AFF3 antibodies turn out not to work, the tagged-knock-in fallback reintroduces the same distortion at lower magnitude and the result must be reported as a partition in an engineered line, not in a leukaemia cell. Luo Z et al. 2012
- E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper
The AF9 AHD is intrinsically disordered, recruits multiple transcription factors by coupled folding and binding, and retains significant dynamic behaviour which the authors propose may facilitate exchange between disordered partners — which is the specific reason this experiment needs a post-lysis exchange control rather than only an order control.
Leukemia fusion target AF9 is an intrinsically disordered transcriptional regulator that recruits multiple partners via coupled folding and binding. · Leach BI, Kuntimaddi A, Schmidt CR, Cierpicki T, Johnson SA, Bushweller JH · Structure 2013What this evidence cannot show
NMR on an isolated domain with a peptide partner; the exchange proposal is the authors' interpretation of relaxation data, not a measured exchange rate between full-length complexes in a cell. It is cited here as the reason to include a control, which is the appropriate weight for it. AF9, not ENL.
Evidence that argues against this (1)
- If exchange is fast on the timescale of a lysate preparation, this experiment cannot succeed in its stated form and the correct output is the MALFORMED verdict pre-registered in the hypothesis rather than a partition number. That outcome would be a real result, but it would leave the whole beat's quantitative argument without an instrument and would require a live-cell approach — FRAP or single-molecule tracking of tagged ENL in scaffold-knockout backgrounds — that is an order of magnitude more expensive. Lu H et al. 2015
prediction 1: AFF1+AFF4 depletion removing >95% of ENL and AF9 kills the hypothesis
measures how much of the cellular reader pool SEC actually holds
ENL pool partition across the AFF scaffolds
AF9 pool partition, measured separately from ENL because the two are not interchangeable