Oncofusion

start / graph / exp-q149-paired-rna-and-oncoprotein-occupancy-in-a-single-kmt2a-eps15-blast

Experimentmoderate confidence 3 of 5: One strong study, or several smaller consistent ones.Q149-irx1-expression-across-partners

Measure IRX1 transcript and KMT2A-N oncoprotein occupancy in the SAME KMT2A::EPS15 blast - the one cell where expression-without-occupancy is most likely and has never been paired

also called q149 paired EPS15 experiment · the cheapest decisive IRX1 experiment

Take one child's leukaemia cells with a KMT2A-EPS15 fusion and, from the same sample, measure both whether IRX1 is switched on and whether the fusion protein is sitting on it. Every existing measurement does one or the other in different children, which is why the question is still open.

The same thing, for a scientist

**The design is deliberately narrow: one fusion partner, one aliquot, two assays, and a pre-specified 2x2 outcome.** **Why EPS15 rather than any other non-AFF1 partner.** It is the only partner where the two existing assays point in opposite directions with real magnitude: **2 of 3 KMT2A::EPS15 infant ALL cases carry IRX1 at 10.11 and 11.07 log2** (cohort AFF1 maximum 11.37) [E1], while **the only KMT2A::EPS15 patient with occupancy data has IRX1 at 0.109x and 0.047x of an average promoter** across two samples of that same patient, with HOXA at 5.74x and 2.94x [E2]. MLLT1 and MLLT3 agree across both assays (off and unoccupied), so they cannot discriminate. **The material.** One or more viably banked or fresh diagnostic KMT2A::EPS15 blast samples - t(1;11)(p32;q23) - split from a single cell suspension. **The two assays must come from the same suspension**, because the whole point is to remove the across-patient inference. EPS15 is roughly 1-2% of infant KMT2A-r ALL, so this needs a biobank request rather than a prospective accrual: St Jude, COG/Interfant banks, or Texas Children's and Fred Hutch, which are the institutions already represented in GSE252378. **The readouts.** - **Transcript:** RNA-seq with an exogenous spike-in (ERCC), plus a targeted RT-qPCR for IRX1 and HOXA9 so the call does not rest on one pipeline. Report IRX1 in absolute spike-in-normalised units, not a within-sample share. - **Occupancy:** AutoCUT&RUN with **both** KMT2A-N antibodies (the GSE252378 protocol, so the numbers are comparable to the 0-of-18), IgG, and a **spike-in calibration** - because `ana-q147-under-revumenib-hoxa-occupancy-collapses-and-the-irx1-question-becomes-unanswerable` shows that within-sample shares become uninterpretable exactly when the global signal moves. - **Purity:** blast percentage by flow on the same suspension. **This is not optional.** The two IRX1-high EPS15 array samples have abnormal clones in 4 of 24 and 8 of 20 metaphases, so a bulk IRX1 signal from a low-blast sample is the leading benign explanation and must be excluded. - **Sorted arm if blasts are below 70%:** sort CD19+CD10+ blasts and repeat both assays on the sorted fraction. **The pre-specified 2x2, and what each cell means:** | IRX1 transcript | KMT2A-N at IRX1 | reading | |---|---|---| | **high** | **absent** | **branch B confirmed in one cell.** The partner controls oncoprotein ACCESS to an already-active locus. This is the result that makes `hyp-q149-irx1-status-is-set-by-the-cell-of-origin...` the better model. | | high | present | the 0-of-18 was a sampling accident; EPS15 belongs with AFF1 and the address is not AFF1-specific | | low | absent | branch A. IRX1 is simply off in this blast and the two IRX1-high array cases were purity artefacts or a different state | | low | present | the fusion sits at a silent promoter - poised, and the most mechanistically interesting single-cell result available |

Go deeper
Powering, and the honest statement about n

This experiment is not powered and does not need to be. It is a paired existence test: a single blast that is IRX1-transcript-high with no oncoprotein at IRX1 establishes that the state exists, which no across-cohort comparison can. One informative sample is the deliverable.

What it cannot do: estimate how often that state occurs. For a frequency, n per partner would have to be at least 5 by this beat's own pre-registered rule, and KMT2A::EPS15 at 1-2% of infant KMT2A-r ALL makes 5 cases a multi-institution request over years.

So the design is: 1-3 EPS15 samples for the existence test, and the same protocol applied opportunistically to any non-AFF1 KMT2A-r blast that is IRX1-high on a rapid RT-qPCR screen. The screen is the cheap part; it costs one qPCR per banked sample and would tell a biobank which vials are worth the CUT&RUN.

A cheaper first step that is nearly free. Before requesting any material, ask the GSE252378 depositors whether R3595/R3652 had RNA taken at the same draw. If any of their 24 patients has banked RNA, the paired measurement may already exist in a freezer. This corpus has repeatedly found that the expensive experiment was a supplementary table.

Controls, and the ones this project has learned to require
  1. A HOXA positive control in the same libraries. The 0-of-18 is credible because HOXA was occupied in 17 of those 18. Any new occupancy library must show HOXA occupancy or it cannot report an IRX1 zero.
  2. A width-matched cannot-be-occupied locus panel, not IgG alone - the floor rule this corpus established in concept-q147-a-width-matched-nonexpressed-locus-panel-is-a-usable-floor-where-igg-is-not.
  3. A transcript floor that is NOT a background gene panel. This beat showed that a ten-gene silent-tissue panel is not silent - three of its members sit above IRX1's cohort median in bulk AML - so use the ERCC spike-in, not a housekeeping-style floor. See concept-q149-the-on-off-verdict-tracked-the-threshold-so-the-binary-was-measuring-the-ruler.
  4. A KMT2A::AFF1 IRX1-high blast run in parallel as the positive control for the paired design itself. Without it, an all-negative result cannot be distinguished from a protocol that does not work on banked material.
  5. Report both assays' calls before looking at either. The 2x2 above is the pre-registration; write it down before the first library.
Facts and figures
designsingle-aliquot paired assay on primary KMT2A::EPS15 t(1;11)(p32;q23) infant ALL blasts: spike-in-normalised RNA-seq plus IRX1/HOXA9 RT-qPCR, AND AutoCUT&RUN with both KMT2A-N antibodies, IgG and spike-in calibration, on material split from ONE cell suspension
readoutpre-specified 2x2 of IRX1 transcript high/low against KMT2A-N present/absent at the IRX1 promoter and gene body
controlsHOXA occupancy in the same libraries; width-matched cannot-be-occupied locus panel rather than IgG alone; ERCC spike-in rather than a background gene panel as the transcript floor; blast percentage by flow on the same suspension; a KMT2A::AFF1 IRX1-high blast run in parallel as the positive control for the paired protocol
cost tiermed
timeline3-9 months once material is obtained; the biobank request is the rate-limiting step
n required1-3 informative samples - this is a paired EXISTENCE test and is not powered for a frequency
material sourcebanked diagnostic KMT2A::EPS15 infant ALL blasts; the institutions already represented in GSE252378 are St Jude, Texas Children's, Fred Hutch and Seattle Children's
cheaper first stepask the GSE252378 depositors whether RNA was banked at the same draw for any of their 24 patients; the paired measurement may already exist
screening stepone IRX1 RT-qPCR per banked non-AFF1 KMT2A-r vial to find which are IRX1-high before committing to CUT&RUN
absence classnever-measured
measurements
[
  {
    "quantity": "n_informative_samples_required",
    "subject": "paired IRX1 RNA + occupancy existence test in KMT2A::EPS15",
    "value": 1,
    "units": "samples",
    "n": null,
    "universe": "a paired existence test; one IRX1-high oncoprotein-absent blast settles branch B",
    "null": null,
    "note": "for a FREQUENCY the pre-registered n>=5 per partner would apply, and EPS15 at 1-2% of infant KMT2A-r ALL makes that a multi-year multi-institution request"
  },
  {
    "quantity": "prior_that_the_state_exists",
    "subject": "IRX1 transcript high with no KMT2A-N at IRX1, in one non-AFF1 blast",
    "value": 0.5,
    "units": null,
    "n": null,
    "universe": "author's prior: 2 of 3 EPS15 cases are IRX1-high in one cohort and 0 of 1 EPS15 patients is IRX1-occupied in another; purity is the leading alternative explanation",
    "null": null,
    "note": "deliberately set at a coin flip - the point of the experiment is that nothing currently distinguishes the two"
  }
]
How this could bite us back
  • If the result is IRX1-high with no oncoprotein, then IRX1 in that blast is not maintained by the fusion, and menin inhibition would not touch it. An IRX1-driven programme that survives the drug this field is built around is a resistance route nobody is watching.
  • If the result is IRX1-low, the two IRX1-high array cases were probably purity artefacts - and every bulk-expression-derived IRX1 subgroup claim in this literature, including the 21-of-48 AFF1 rate this beat reports, inherits the same doubt.
Still open
  1. How many viably banked KMT2A::EPS15 infant ALL samples exist anywhere? Nothing in this corpus records a census of them and the experiment's feasibility turns entirely on that number.
  2. Would the depositors of GSE252378 share blast percentages for their 27 samples? That single table would let the existing 0-of-18 be re-read against purity, at no experimental cost.
  3. Should the same paired design be run first on a KMT2A::MLLT10 or MLLT3 FAB-M7 AML blast instead? IRX1 is ON in 65% of those, so IRX1-high material is far easier to find there - at the cost of leaving the B-lineage question untouched.
Every source behind this entry (2)
  1. E1weak confidence 2 of 5: Preliminary, small, indirect, or not yet peer reviewed.dataset

    Two of three KMT2A::EPS15 infant ALL cases in GSE68720 carry IRX1 transcript at 11.07 and 10.11 log2 against per-sample floors of 6.75 and 6.41, comparable to the highest KMT2A::AFF1 case in the same cohort at 11.37.

    PMID 22210879DOI
    What this evidence cannot show

    n=3, single array probe, no blast count, and one of the two positive samples has its abnormal clone in only 4 of 24 metaphases. This is the motivation for the experiment, not a result the experiment would confirm.

    Evidence that argues against this (1)
    • Three KMT2A-germline infant ALL cases in the same cohort are also IRX1-ON, one at 10.37 - so two IRX1-high EPS15 cases may be two draws from an infant-ALL background rather than anything partner-related, and the experiment might find nothing because there was nothing there. Kang H et al. 2012
  2. E2moderate confidence 3 of 5: One strong study, or several smaller consistent ones.dataset

    The only KMT2A::EPS15 patient with deposited oncoprotein occupancy data has IRX1 within-sample shares of 0.109x and 0.047x of an average promoter across two samples, against HOXA shares of 5.74x and 2.94x in the same libraries.

    GSE252378 - AutoCUT&RUN profiling of KMT2A-rearranged paediatric leukaemia patient samples · GEO depositors, series GSE252378 · NCBI Gene Expression Omnibus 2024
    What this evidence cannot show

    Read from analyses/q147-hoxa-irx1-across-partners/results_gene_scale147.json rather than recomputed from bigWigs in this beat. R3595 and R3652 are recorded in q147 as the same patient sampled twice, so this is one patient. R3652's HOXA at 2.94x is at the weak end of the deposit.

    Evidence that argues against this (1)
    • One patient is not a bounded absence for a partner. The EPS15 arm's rule-of-three bound is meaningless at n=1 and the 0-of-18 pooled bound draws its strength from AFDN, AF10, AF9 and ENL, not from EPS15. So the tension this experiment is built on may simply be that nobody has looked at EPS15 properly - which is an argument FOR running it and not an argument that the result is predictable. GEO depositors et al. 2024
Connected to