The days-long experiment that would close or confirm the imprinting liability: allele-specific XIST and MEG3 readout in cells treated with a YEATS antagonist
also called allele-specific expression under SGC-iMLLT · q187 imprinting safety assay
Treat cells with the reader-blocking drug and check whether genes that are supposed to be silent on one copy stay silent. The same readout was used to show that removing AFF3 breaks this, so the assay already exists and has never been pointed at a drug.
The same thing, for a scientist
Treat a female human cell line with informative heterozygous SNPs at *XIST* and at the *DLK1-DIO3* locus with SGC-iMLLT, SR-0813 or TDI-11055 at a concentration with demonstrated cellular target engagement, and read **allele-specific** expression at the XIST DMR-controlled locus and at the MEG3 polycistron by allele-discriminating RT-PCR or amplicon sequencing. **This is the same endpoint by which AFF3 knockdown was shown to de-repress the inactive XIST allele in terminally differentiated cells** [E1], so the assay is published, validated and directly comparable. The positive control is AFF3 knockdown in the same cells; the drug arms are the test. **A negative closes the liability in `risk-q187-yeats-inhibition-touches-an-imprinting-complex-in-an-infant` far more cheaply than arguing about it**, and a positive is the first non-haematopoietic safety signal for a target class whose published window rests entirely on two blood endpoints [E2].
Design
| element | choice | why |
|---|---|---|
| cells | a female human line, terminally differentiated or differentiable, with informative heterozygous SNPs in the XIST transcript | matches the system in which AFF3 knockdown gave the effect 1 |
| test arms | SGC-iMLLT, SR-0813 and TDI-11055, each at a concentration with published cellular target engagement | three chemotypes, so a shared effect is unlikely to be one compound's off-target |
| positive control | AFF3 siRNA/shRNA in the same cells | reproduces the published effect and proves the readout can fire here |
| negative control | vehicle; plus a chemically matched inactive analogue if one exists in the SGC series | separates pocket occupancy from scaffold chemistry |
| primary readout | allele-specific RT-PCR or amplicon sequencing of XIST, plus MEG3 at the Dlk1-Dio3 locus | XIST is the published endpoint; MEG3 tests the second AFF3-dependent locus |
| secondary readout | DNA methylation at the XIST DMR by bisulfite amplicon | AFF3 binding is methylation-dependent and feeds back on methylation level, so the mark is the slower half of the phenotype |
| duration | short (24-72 h) and extended (7-14 d) arms | a monoallelic state may need time to drift; a short-only design could return a false negative |
Pre-registration. Declare before running: the effect size that counts (this entry proposes a shift in allelic ratio of at least the magnitude the AFF3-knockdown control produces in the same experiment, i.e. the control sets the scale rather than an imported threshold), and that a null in the short arm alone does not close the liability.
Why this rather than the obvious alternative
The obvious alternative is ENL/AF9 ChIP at the imprinted control regions — and that is the more mechanistically decisive experiment, which is why risk-q187-... lists it second. This one is placed first for two reasons.
It is functional rather than correlative. A reader could be present at a DMR and irrelevant; an allelic ratio that moves under a drug is the outcome anyone would actually care about.
A negative here is worth more than a negative ChIP. A ChIP negative could be an antibody or depth failure. An allele-specific expression negative, with a working AFF3-knockdown positive control in the same plate, says the phenotype does not occur at that exposure — and that is what a safety argument needs.
The free version, worth running first because it costs nothing, is item 3 of the risk entry's mitigation ladder: intersect the public YEATS-inhibitor RNA-seq (GSE185091, GSE162592) with an imprinted-gene list. It cannot give a clean negative — those are AML lines and bulk RNA-seq cannot see allele-specific effects at all — but a positive would make this wet experiment urgent rather than precautionary.
| design | female human cell line with informative heterozygous SNPs at XIST and DLK1-DIO3, treated with SGC-iMLLT, SR-0813 or TDI-11055 at target-engaging concentrations, short (24-72 h) and extended (7-14 d) arms, against vehicle and an AFF3-knockdown positive control |
|---|---|
| readout | allele-specific RT-PCR or amplicon sequencing of XIST and MEG3; secondary bisulfite amplicon methylation at the XIST DMR; cellular thermal shift on ENL in the same cells as a target-engagement control |
| controls | AFF3 knockdown positive control in the same plate (sets the effect-size scale rather than importing a threshold); vehicle; chemically matched inactive analogue if available; three chemotypes so a shared effect is not one compound's off-target; ENL CETSA to confirm engagement at the concentrations used in THESE cells rather than in AML lines |
| cost tier | low |
| timeline | 4-6 weeks for the short arm; 10-12 weeks including the extended arm and methylation readout |
| closes | risk-q187-yeats-inhibition-touches-an-imprinting-complex-in-an-infant |
| scores hypothesis | hyp-q187-yeats-inhibition-is-scaffold-blind |
| prerequisite free analysis | intersect GSE185091 and GSE162592 YEATS-inhibitor RNA-seq with an imprinted-gene list first; positive makes this urgent, negative is uninformative (AML lines, bulk RNA-seq cannot see allele-specific effects) |
| shown for fusion | N/A - proposed. Deliberately NOT in a fusion line: the question is about normal tissue, so a leukaemia line would be the wrong system, which is unusual for this project and is the point. |
measurements
[
{
"quantity": "effect_size_that_counts",
"subject": "shift in XIST allelic expression ratio under a YEATS antagonist, scaled to the AFF3-knockdown arm in the same experiment",
"value": 1,
"units": "fraction of the AFF3-knockdown effect",
"n": null,
"universe": "pre-registered 2026-09-24 before any data exist; the internal positive control sets the scale deliberately, because no external threshold for 'a meaningful imprinting perturbation' exists and importing one would make the verdict track the threshold",
"null": 0,
"note": "A drug effect at or above the knockdown effect is a clear positive; below 25% of it with a working control is the negative that closes the liability; 25-100% is pre-declared ambiguous."
}
]- A negative result here will be quoted as 'YEATS inhibitors do not disturb imprinting'. It would support only 'at these exposures, in this cell type, over this duration, the XIST and MEG3 allelic ratios did not move'. Imprinting is established in the germline and early embryo, and an adult or paediatric somatic cell assay cannot address exposure during those windows at all.
- Running three chemotypes protects against one compound's off-target and creates a multiple-comparison problem nobody will adjust for. Pre-specify that a positive requires concordance across at least two chemotypes, or the experiment will generate one significant arm and an argument.
- The experiment is cheap and the liability is speculative, which is exactly the configuration in which a cheap experiment gets run, returns a null, and is used to close a question it was never powered to close. The extended arm and the internal positive control are the two design elements that prevent that, and they are the two most likely to be cut.
- Which cell line has usable informative SNPs at both XIST and MEG3 and is female and differentiable? The published AFF3 work used terminally differentiated human cells; identifying the exact line from the full text would remove the main design uncertainty and this beat did not read it.
- Is there a chemically matched inactive analogue in the SGC-iMLLT series? An inactive congener is the strongest negative control available for a reader-pocket compound and the probe paper is where to look.
- Would a bulk RNA-seq arm add anything? Probably not for imprinting, which needs allelic resolution, but it would catch a broad transcriptional effect in a non-haematopoietic cell, which nobody has ever looked at for this class either.
- E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
AFF3, identified as the central factor of SEC-L3, binds the differentially methylated region downstream of the XIST promoter in a DNA-methylation-dependent manner, and knockdown of AFF3 de-represses the inactive allele of XIST in terminally differentiated cells — establishing both that the liability's mechanism is plausible and that a validated allele-specific readout for it already exists.
AFF3-DNA methylation interplay in maintaining the mono-allelic expression pattern of XIST in terminally differentiated cells. · Zhang Y, Wang C, Liu X, Yang Q, Ji H, Yang M, Xu M, Zhou Y, Xie W, Luo Z, Lin C · Journal of Molecular Cell Biology 2019What this evidence cannot show
Knockdown, not pharmacology - the perturbations are different in kind, since removing AFF3 removes the scaffold and everything it carries while a YEATS ligand removes one subunit's histone contact and leaves the assembly bound. `expected_absent_terms` records that the paper measures no reader and no compound. One laboratory, and the same laboratory that defined SEC-L3.
Evidence that argues against this (1)
- Because AFF3 knockdown and YEATS-pocket occupancy are different perturbations, a negative in this experiment does NOT show that SEC-L3 is uninvolved in XIST regulation - it shows that blocking the reader's histone contact is insufficient to move the allelic ratio. That is the right answer for a safety question and the wrong answer for a mechanistic one, and the entry should not be cited as having tested SEC-L3's role. Wan L et al. 2017
- E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper
Multiple YEATS-pocket antagonists with cellular target engagement exist and are available for this design — SGC-iMLLT (PDB 6HT1 co-crystal, selective over YEATS2/4 and bromodomains), SR-0813 (ENL/AF9 YEATS inhibitor, IC50 25 nM, with a first-in-class ENL PROTAC SR-1114) and the orally bioavailable TDI-11055, whose on-target action is validated by a CRISPR-Cas9-recovered ENL resistance mutation — while every published safety endpoint for the class is haematopoietic.
Chemical Inhibition of ENL/AF9 YEATS Domains in Acute Leukemia. · Garnar-Wortzel L, Bishop TR, Kitamura S, Milosevich N, Asiaban JN, Zhang X, Zheng Q, Chen E, Ramos AR, Ackerman CJ, Hampton EN, Chatterjee AK, Young TS, Hull MV, Sharpless KB, Cravatt BF, Wolan DW, Erb MA · ACS Central Science 2021What this evidence cannot show
This item spans three papers and the citation block carries only the SR-0813 one (PMID 34079898); SGC-iMLLT is PMID 30288907 and TDI-11055 is PMID 36053276, both cited in full in `absence-q187-no-yeats-inhibitor-has-been-tested-against-sec-l2-or-sec-l3`. All three are AML-context reagents and none has been characterised in a non-haematopoietic cell. Abstracts only.
Evidence that argues against this (1)
- None of the three compounds has a published exposure-response in a non-haematopoietic cell, so 'a concentration with demonstrated cellular target engagement' is defined only in AML lines. Running them in a fibroblast or lymphoblastoid line at an AML-derived concentration risks either under-dosing (false negative) or a concentration with off-target activity (false positive), and the design should include a target-engagement readout - a cellular thermal shift on ENL - in the cells actually used. Liu Y et al. 2022
functional test of the imprinting limb
a negative with a working AFF3-knockdown control closes the liability