IRX1 status is set by the cell of origin, and the fusion partner predicts it only because different partners transform different cells - so the partner supplies no IRX1 address
also called q149 cell-of-origin hypothesis for IRX1 · the partner is a proxy for the target cell
Maybe IRX1 is not switched on by the fusion at all. Maybe it is already on or off in whichever blood cell the fusion happens to land in, and the partner matters only because different partners tend to arise in different cells.
The same thing, for a scientist
**The statement.** IRX1 transcription is a property of the cell the fusion arose in, not of the fusion. KMT2A::AFF1 is associated with IRX1 because it arises in a cell (or developmental window) where the IRX1 locus is already permissive; KMT2A::MLLT3 and KMT2A::MLLT10 are not, because they arise in myelomonocytic progenitors where it is not. On this reading, **`hyp-q147-the-irx1-address-requires-the-aff1-partner-not-the-menin-interface` is explaining a correlation that has nothing to do with the partner's protein.** **Why this became worth writing this cycle, rather than being an idle alternative.** Three measurements, none available before: 1. In paediatric AML, IRX1 expression tracks **FAB morphology** and not KMT2A status: 13 of 20 KMT2A-r M7 versus **54 of 97 non-KMT2A-r M7**, and 2 of 96 KMT2A-r M5 versus 3 of 52 non-KMT2A-r M5. **Within a cell type, the fusion adds nothing** [E1]. 2. In infant ALL, the non-AFF1 KMT2A-r IRX1 rate (3 of 26) is **indistinguishable from KMT2A-germline infant ALL** (3 of 17, Fisher two-sided p = 0.666) [E2]. **Without AFF1, having the fusion changes nothing.** 3. Nagel's myeloid TALE-code places IRX1's normal expression in **megakaryocyte-erythroid progenitors**, which is precisely the lineage where the AML signal sits [E3]. The expression is where the *normal* programme is, not where a novel programme would be. **The rival it must beat** is the address model: the AFF1 C-terminus recruits the complex to IRX1, and expression follows. The address model predicts occupancy causes expression; this hypothesis predicts expression permits occupancy. **What separates them is the direction of the arrow in one cell**, and neither an expression cohort nor an occupancy cohort can see it. That is why the experiments below are all paired or perturbational.
Predictions that differ from the address model
| observation | address model predicts | cell-of-origin model predicts |
|---|---|---|
| IRX1 in a KMT2A-germline infant ALL | should be rare, since nothing addresses it | should occur at the same rate as in non-AFF1 KMT2A-r - and it does, 3 of 17 vs 3 of 26 |
| IRX1 in KMT2A-r M7 AML vs non-KMT2A-r M7 | higher with the fusion | the same - and it is, 65.0% vs 55.7% |
| AFF1 C-terminus swapped onto KMT2A-MLLT3 in a myeloid background | IRX1 occupancy appears | nothing happens; the locus is not permissive |
| the same swap in an infant pro-B background | IRX1 occupancy appears | IRX1 occupancy appears only if the recipient cell is already IRX1-permissive, i.e. it will look partner-dependent in a pro-B and partner-independent in a fibroblast |
| an IRX1-high AFF1 blast under a menin inhibitor | IRX1 falls with HOXA | IRX1 persists after the oncoprotein is evicted |
| a KMT2A::AFF1 line grown from an IRX1-negative background | IRX1 turns on | IRX1 stays off - and all four CCLE AFF1 lines are IRX1-zero |
The last row is already partly observed and it favours this hypothesis. MV4-11, NALM-19, RS4;11 and SEM carry the index fusion and IRX1 at 0.0000-0.0053 RPKM. If the AFF1 partner supplied an IRX1 address, four independent AFF1 lines with no IRX1 at all is an awkward fact.
And the fifth row is the sharpest and is answerable from deposited data, except that ana-q147-under-revumenib-hoxa-occupancy-collapses-and-the-irx1-question-becomes-unanswerable records that the only menin-inhibitor-treated samples in GSE252378 lose all occupancy, so the within-sample share flattens and the question cannot be asked without a spike-in.
What this hypothesis does NOT claim
- It does not claim the AFF1 partner is irrelevant to anything. HOXA occupancy is partner-invariant (17 of 18 non-AFF1 samples) and IRX1 occupancy is not, and that asymmetry is real whatever explains it.
- It does not claim IRX1 is non-functional.
finding-irx1-binds-the-mll-af4-complex-without-displacing-meninrecords a direct physical result; a cell-of-origin origin for IRX1 expression is compatible with IRX1 protein then doing something. - It does not resolve
dispute-is-irx1-expressed-in-normal-b-cell-development, and it would be strengthened by either side of it. If Nagel is right that IRX1 is normal in pro-B cells, this hypothesis has its permissive cell. If Kuhn and Symeonidou are right that it is aberrant, the hypothesis needs a different permissive state - an infant-specific or fetal window rather than a standard pro-B. Both are live and neither is assumed here.
| statement | IRX1 transcription in KMT2A-rearranged leukaemia is a property of the cell of origin; the fusion partner predicts IRX1 status only because different partners transform different cells, and supplies no IRX1 address. |
|---|---|
| origin | data-pattern |
| prior probability | 0.35 |
| absence class | measured-not-analysed |
predictions
[ "Within any single cell type or FAB stratum, KMT2A-rearranged and non-KMT2A-rearranged cases have the same IRX1 rate. ALREADY OBSERVED in FAB M7 (65.0% vs 55.7%) and FAB M5 (2.1% vs 5.8%) in aml_stjude_2024.", "Non-AFF1 KMT2A-r infant ALL has the same IRX1 rate as KMT2A-germline infant ALL. ALREADY OBSERVED, 3 of 26 vs 3 of 17, Fisher two-sided p = 0.666 - but badly underpowered.", "A KMT2A::AFF1 cell line derived from an IRX1-negative background stays IRX1-negative. ALREADY OBSERVED in all four CCLE AFF1 lines at 0.0000-0.0053 RPKM.", "Evicting the oncoprotein from an IRX1-high AFF1 blast with a menin inhibitor leaves IRX1 transcript standing while HOXA falls. NOT YET RUN; GSE252378's menin-inhibitor samples lose all occupancy and no spike-in exists.", "Swapping the AFF1 C-terminus onto KMT2A-MLLT3 in a MYELOID background produces NO IRX1 occupancy, while the same swap in an infant pro-B background does. NOT YET RUN." ]
experiments
[ "exp-q149-paired-rna-and-oncoprotein-occupancy-in-a-single-kmt2a-eps15-blast", "exp-q147-swap-the-aff1-c-terminus-onto-kmt2a-af9-and-score-irx1-occupancy", "exp-q147-paired-irx1-rna-and-oncoprotein-occupancy-in-non-aff1-kmt2a-r-blasts" ]
analyses
[ "ana-q149-irx1-in-kmt2a-r-paediatric-aml-is-a-megakaryoblastic-lineage-property-not-a-fusion-property", "ana-q149-irx1-transcript-is-aff1-enriched-in-infant-all-and-is-not-absent-outside-aff1" ]
falsifiers
[ "Within FAB M7 paediatric AML, KMT2A-rearranged cases showing an IRX1-positive rate at least 25 percentage points ABOVE non-KMT2A-rearranged M7 cases, at n >= 30 per arm. Current values are 65.0% (13/20) and 55.7% (54/97), a difference of +9.3 points at n=20, which does not fire and is underpowered to.", "IRX1 transcript FALLING by at least 2-fold within 72 hours of menin-inhibitor exposure in a primary IRX1-high KMT2A::AFF1 infant blast, with HOXA9 falling in the same sample and an exogenous spike-in normalising both. A fall would mean the oncoprotein maintains IRX1 rather than merely finding it open.", "Introducing KMT2A::AFF1 into an IRX1-negative human infant pro-B cell and observing IRX1 transcript rise above 2-fold over the parental within 14 days. That would show the fusion CAN switch IRX1 on in a cell where it was off, which this hypothesis forbids.", "A KMT2A::AFF1 cell line that is IRX1-POSITIVE. All four in CCLE are at 0.0000-0.0053 RPKM; one positive line would not kill the hypothesis but a line that becomes positive on acquiring the fusion would." ]
measurements
[
{
"quantity": "prior_probability",
"subject": "IRX1 status is set by the cell of origin rather than by the partner",
"value": 0.35,
"units": null,
"n": null,
"universe": "author's prior after the FAB stratification and the germline-control comparison; the rival address model retains most of the remaining mass",
"null": 0.5,
"note": "set below the address model because HOXA-vs-IRX1 partner asymmetry in occupancy remains unexplained by this hypothesis"
},
{
"quantity": "difference_in_IRX1_rate_within_FAB_M7",
"subject": "KMT2A-r minus non-KMT2A-r, paediatric AML FAB M7",
"value": 0.093,
"units": null,
"n": 117,
"universe": "aml_stjude_2024, 20 KMT2A-r M7 vs 97 non-KMT2A-r M7, IRX1 log2>0",
"null": 0,
"note": "the falsifier threshold is +0.25 at n>=30 per arm; this is +0.093 at n=20 and does not fire"
}
]- If IRX1 status is cell-of-origin rather than fusion-driven, then no fusion-directed therapy - menin inhibition included - will move it, and the iALL-IRX1 subgroup's poor outcome would not be addressable by the drug class this whole project is built around. That is the most consequential version of this hypothesis being true.
- Conversely if it is false and the address model holds, an IRX1-directed strategy inherits every resistance mechanism the menin axis already has, because the address depends on the same complex.
- What IS the cell of origin difference between KMT2A::AFF1 and KMT2A::MLLT3 infant leukaemia, stated as a measurable chromatin state at the IRX1 locus rather than as a lineage label?
- If IRX1 is permissive-then-occupied rather than occupied-then-expressed, what opens it? Nothing in this corpus records an accessibility or methylation measurement at the IRX1 promoter in any KMT2A-r sample.
- Does this hypothesis survive in B-lineage at all, where there is no FAB-equivalent stratifier? The infant ALL support rests entirely on a 17-case germline control arm.
- E1moderate confidence 3 of 5: One strong study, or several smaller consistent ones.dataset
In paediatric AML, IRX1 expression rate is 13 of 20 in KMT2A-rearranged FAB-M7 cases and 54 of 97 in non-KMT2A-rearranged FAB-M7 cases, and 2 of 96 versus 3 of 52 in FAB M5 - so within a morphology the rearrangement does not change IRX1 status.
A new genomic framework to categorize pediatric acute myeloid leukemia · Umeda M, Ma J, Westover T, et al. · Nat Genet 56:281-293 2024What this evidence cannot show
FAB morphology is a proxy for cell of origin, not a measurement of it, and it is a phenotype of the transformed cell rather than of the ancestor. This is AML, and the index fusion is B-lineage with n=1 in this cohort. Stratified n are modest.
Evidence that argues against this (1)
- If the fusion CAUSES the megakaryoblastic phenotype, then stratifying on FAB conditions on a descendant of the fusion and removes the very effect being tested. On that reading the within-stratum equality is an over-control rather than a refutation of the address model. The argument survives only because the non-KMT2A-r M7 arm - which has no fusion to cause anything - has the same IRX1 rate. Umeda M et al. 2024
- E2moderate confidence 3 of 5: One strong study, or several smaller consistent ones.dataset
In 97 infant ALL cases the IRX1 expression rate is 3 of 26 in non-AFF1 KMT2A-rearranged cases and 3 of 17 in KMT2A-germline cases in the same cohort on the same array, Fisher exact two-sided p = 0.666.
Gene expression profiles predictive of outcome and age in infant acute lymphoblastic leukemia: a Children's Oncology Group study · Kang H, Wilson CS, Harvey RC, et al. · Blood 119:1872-1881 2012What this evidence cannot show
A p of 0.666 at n=26 versus 17 is an absence of evidence for a difference, not evidence of equality - the comparison is badly underpowered and the intervals overlap almost entirely (Wilson 4.0-29.0% vs roughly 6-39%). It cannot exclude a real difference of moderate size. Array, thresholded calls, no purity correction.
Evidence that argues against this (1)
- The AFF1 rate is 43.8% against the same germline null of 17.6%, and Fisher two-sided p there is 0.0798 - i.e. the AFF1-versus-germline contrast is also not significant at n=48 vs 17, while AFF1-versus-non-AFF1 is (p=0.0047). So the germline control arm is too small to anchor either comparison cleanly, and this hypothesis's strongest infant-ALL support rests on the weakest arm in the cohort. Kang H et al. 2012
- E3moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper
The myeloid TALE-code reports the exclusive expression of IRX1 in megakaryocyte-erythroid progenitors within normal myelopoiesis, which is the same lineage in which IRX1 expression concentrates in KMT2A-rearranged paediatric AML.
The Hematopoietic TALE-Code Shows Normal Activity of IRX1 in Myeloid Progenitors and Reveals Ectopic Expression of IRX3 and IRX5 in Acute Myeloid Leukemia · Nagel S, Pommerenke C, Meyer C, MacLeod RAF · Int J Mol Sci 23:3192 2022What this evidence cannot show
DO NOT CITE FROM THE TITLE - the title reads as though IRX1 were not aberrant in AML while the abstract states 'aberrant activity of IRX1 in addition to IRX3 and IRX5'. The MEP placement is from public profiling data and cell lines, not primary sorted human progenitors in this paper. A normal-lineage expression pattern matching a leukaemic one is consistent with a cell-of-origin explanation and does not establish it.
Evidence that argues against this (1)
- The same paper reports IRX1 activity in AML as ABERRANT, i.e. not simply the normal MEP programme carried forward. If the authors are right that it is aberrant, the coincidence of lineage is not enough and something is switching IRX1 on that the normal programme does not explain - which is the address model's opening. Nagel S et al. 2022
the within-FAB equality is this hypothesis first prediction and it is already observed
the non-AFF1 rate matching the KMT2A-germline control is the second prediction
four KMT2A::AFF1 cell lines carrying the index fusion with IRX1 at zero
the rival: on this reading the partner supplies no address and the correlation is cell-of-origin
a sibling confound hypothesis - lineage there, cell of origin here
- depends on →Is IRX1 expressed in normal blood at all? Two of the three papers say no, one says exclusively yes
either side of the dispute changes which permissive cell this hypothesis needs; neither is assumed
the paired existence test
the swap in a myeloid versus a pro-B background discriminates the two models
without AFF1, having the fusion does not change IRX1 status relative to fusion-free infant ALL