Oncofusion

start / graph / ana-q149-irx1-in-kmt2a-r-paediatric-aml-is-a-megakaryoblastic-lineage-property-not-a-fusion-property

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

IRX1 expression in KMT2A-rearranged paediatric AML is a FAB-M7 megakaryoblastic property, not a fusion property: 13 of 20 M7 versus 2 of 96 M5, and non-KMT2A-r M7 is 54 of 97

also called q149 IRX1 M7 AML · IRX1 megakaryoblastic KMT2A-r

In 887 children with AML, the IRX1 gene is on mainly in one kind of leukaemia - megakaryoblastic (M7) - whether or not the cell carries a KMT2A fusion. So IRX1 being on says which blood cell the leukaemia came from, not which fusion it carries.

The same thing, for a scientist

**In St Jude's 887-case paediatric AML cohort, IRX1 transcript is ON in 13 of 20 (65.0%) non-AFF1 KMT2A-rearranged FAB-M7 cases and in 2 of 96 (2.1%) FAB-M5 cases** - a 31-fold difference inside one fusion class [E1]. **And the same is true without the fusion: 54 of 97 (55.7%) non-KMT2A-rearranged M7 cases are IRX1-ON, against 3 of 52 (5.8%) non-KMT2A-rearranged M5.** The two M7 rates, 65.0% with a KMT2A rearrangement and 55.7% without, are not distinguishable at these n. **So IRX1 status in paediatric AML tracks morphology, and the fusion adds nothing detectable on top of it.** That is exactly what Nagel's myeloid TALE-code predicts: its abstract states verbatim *"the exclusive expression of IRX1 in megakaryocyte-erythroid progenitors (MEPs)"* and *"overexpression of IRX1, IRX3, and IRX5 in megakaryoblastic and myelomonocytic AML cell lines"* [E2]. **The megakaryoblastic half of that claim reproduces in primary patients. The myelomonocytic half does not** - M4 is 1 of 24 (4.2%) and M5 is 2 of 96 (2.1%) among KMT2A-r cases, and 41.8% / 5.8% among non-KMT2A-r ones, so whatever drives IRX1 in the M4 arm of a cell-line panel is not present in KMT2A-r M5 patients. **This entry withdraws a claim I made earlier in the same cycle.** See `concept-q149-a-fusion-versus-rest-expression-difference-can-be-a-morphology-composition-artefact`.

Go deeper
The full cross-tab, restricted with the population

aml_stjude_2024, mRNA-seq CPM profile, IRX1 ON = log2 value above 0 (an absolute anchor, chosen because the pre-registered floor-based verdict tracks the threshold). Rates reported only where n >= 5; smaller cells named.

FABnon-AFF1 KMT2A-rratenon-KMT2A-rrate
M713 / 2065.0%54 / 9755.7%
M02 / 728.6%8 / 3324.2%
M41 / 244.2%61 / 14641.8%
M52 / 962.1%3 / 525.8%
M10 / 70%23 / 11220.5%
M20 / 3NAMED, NOT COMPUTED71 / 15944.7%
M6none in this arm-5 / 1827.8%
M3none-3 / 4NAMED, NOT COMPUTED
never recorded2 / 1414.3%28 / 8632.6%

Three categories, not two: FAB is recorded for 787 of 887 samples; 14 of the 171 non-AFF1 KMT2A-r samples have no FAB value at all and are counted in their own row rather than scored as negative.

All 20 KMT2A-r M7 samples carry MOLECULAR_CATEGORY: KMT2Ar. The 97 non-KMT2A-r M7 samples are GLISr 25, GATA1 16, HOXr 16, NUP98r 14, Unclassified 11, RBM15::MKL1 10 and five singletons - so the M7 comparator is a mixture of megakaryoblastic drivers and is not a clean single entity.

The twenty IRX1-high non-AFF1 KMT2A-r AML samples, named

Top of the list, with the partner from the study's own structural-variant calls:

samplepartnerIRX1HOXA9MEIS1PROM1FAB
SJAML061507_R1USP26.514.55-0.908.21M0
SJAML074171_D1MLLT36.047.438.071.56M7
SJAML073686_D1MLLT105.946.819.568.31M0
SJAML030079_D1MLLT35.767.388.782.03M7
SJAML016567_D2USP25.764.924.216.74M4
SJAML040639_D1MLLT34.196.049.144.70M5

Every one of them is HOXA9-high as well (4.5-9.4 against a cohort median of 2.04). So these are not the HOXA-low/IRX1-high state of the infant ALL dichotomy - in AML the two are co-expressed. That negative is recorded in ana-q149-the-hoxa-irx1-mutual-exclusivity-is-aff1-specific-and-absent-in-mllt1.

The KMT2A::USP2 signal is the oddity: 2 of 3 USP2 cases IRX1-high, both with low MEIS1, and neither is M7. n=3, named, not computed - but USP2 already has a breakpoint entry in this corpus and no expression entry.

The single KMT2A::AFF1 case in this cohort, named

SJAML040667_D1 is the only KMT2A::AFF1 sample among the 180 KMT2A-rearranged cases (AFF1 is a B-lineage partner and this is an AML cohort). Its values: IRX1 -5.03 (OFF), IRX2 -5.77, IRX3 -5.19, IRX5 -0.73, PROM1 4.01, HOXA9 5.11, MEIS1 7.36, FAB M5.

n=1, NAMED, NOT COMPUTED. It is recorded because the pre-registration asked for the AFF1 reference arm and this is what exists: one myeloid AFF1 case, IRX1-negative. It is the reason NEITHER condition 4 of the pre-registration holds in this cohort - there is no AFF1 arm at n >= 5 anywhere in myeloid disease.

Facts and figures
cohortaml_stjude_2024, 887 paediatric AML, mRNA-seq CPM, per-sample KMT2A partner from the study's structural-variant calls
n kmt2a sv samples180
n nonAFF1 known partner171
n aff1 in this cohort1
n kmt2a sv no known partner excluded and named8
anchorIRX1 log2 CPM > 0
scriptanalyses/q149-is-irx1-expressed-in-non-aff1-kmt2a-r/anchors149.py
resultsanalyses/q149-is-irx1-expressed-in-non-aff1-kmt2a-r/results149_anchors.json
absence classmeasured-not-analysed
measurements
proportion_IRX1_gt00.021proportion_IRX1_gt0: 0.0208proportion_IRX1_gt00.117proportion_IRX1_gt0: 0.117proportion_IRX1_gt00.650proportion_IRX1_gt0: 0.65n_M7_samples117.0n_M7_samples: 117
measurements — the numbers behind the chart
[
  {
    "quantity": "proportion_IRX1_gt0",
    "subject": "non-AFF1 KMT2A-r paediatric AML, FAB M7",
    "value": 0.65,
    "units": null,
    "n": 20,
    "universe": "aml_stjude_2024 mrna_seq_cpm, IRX1 log2 > 0, FAB recorded",
    "null": 0.557,
    "note": "13 of 20; the null is the SAME morphology WITHOUT the rearrangement, 54 of 97 non-KMT2A-r M7"
  },
  {
    "quantity": "proportion_IRX1_gt0",
    "subject": "non-AFF1 KMT2A-r paediatric AML, FAB M5",
    "value": 0.0208,
    "units": null,
    "n": 96,
    "universe": "aml_stjude_2024 mrna_seq_cpm, IRX1 log2 > 0, FAB recorded",
    "null": 0.0577,
    "note": "2 of 96; the null is 3 of 52 non-KMT2A-r M5. Both arms are low - M5 is where IRX1 is off, fusion or not"
  },
  {
    "quantity": "proportion_IRX1_gt0",
    "subject": "non-AFF1 KMT2A-r paediatric AML, all FAB pooled",
    "value": 0.117,
    "units": null,
    "n": 171,
    "universe": "aml_stjude_2024 mrna_seq_cpm, IRX1 log2 > 0",
    "null": 0.3621,
    "note": "20 of 171 vs 256 of 707 non-KMT2A-r. THE POOLED FIGURE IS A COMPOSITION ARTEFACT - 96 of 171 KMT2A-r cases are M5, the lowest-IRX1 morphology in the cohort"
  },
  {
    "quantity": "n_M7_samples",
    "subject": "FAB M7 in aml_stjude_2024 by KMT2A status",
    "value": 117,
    "units": "samples",
    "n": 887,
    "universe": "aml_stjude_2024, FAB recorded for 787 of 887",
    "null": null,
    "note": "20 KMT2A-r + 97 non-KMT2A-r; the 97 are GLISr 25, GATA1 16, HOXr 16, NUP98r 14, Unclassified 11, RBM15::MKL1 10, and five singletons"
  }
]
How this could bite us back
  • An IRX1-directed intervention selected on 'KMT2A-rearranged AML' would be aiming at a gene that is off in 98% of the M5 cases that make up 56% of that population, and on in a subgroup that is defined by morphology rather than by the fusion. The correct enrolment axis here is M7, not KMT2A.
  • If IRX1 in M7 is normal megakaryocyte-erythroid programme rather than leukaemic, an agent that suppresses it would be attacking a lineage the patient needs for platelets - in a disease whose blasts are already megakaryocytic. That is an essential-function toxicity risk with a direct clinical readout.
  • IRX1 LOSS drives MYC in ML-DS (PMID 38513239). Lowering IRX1 in a megakaryoblastic leukaemia may push MYC up. The two published directions point opposite ways in the same lineage.
Still open
  1. Is the IRX1 signal in an M7 bulk sample coming from the blasts or from the normal megakaryocyte-erythroid cells beside them? A blast-count-stratified or single-cell measurement would settle it, and it is the cheapest way to destroy this finding.
  2. Why are 2 of 3 KMT2A::USP2 cases IRX1-high with LOW MEIS1 and neither of them M7? USP2 has a breakpoint entry in this corpus and no expression entry at all.
  3. Does the 16-case GATA1 subgroup of M7 in this cohort show IRX1 LOSS, as Sato et al. report for 16.2% of ML-DS? If it does, IRX1 is up in KMT2A-r M7 and down in GATA1 M7, inside one cohort, and that is a testable contrast nobody has drawn.
Every source behind this entry (3)
  1. E1moderate confidence 3 of 5: One strong study, or several smaller consistent ones.dataset

    In 887 paediatric AML cases with per-sample KMT2A fusion-partner calls and mRNA-seq, IRX1 transcript above log2 zero occurs in 13 of 20 non-AFF1 KMT2A-rearranged FAB-M7 cases and 2 of 96 FAB-M5 cases, while non-KMT2A-rearranged M7 cases are IRX1-positive at 54 of 97 - so the association is with megakaryoblastic morphology and not with the rearrangement.

    A new genomic framework to categorize pediatric acute myeloid leukemia · Umeda M, Ma J, Westover T, et al. · Nat Genet 56:281-293 2024
    PMID 38212634DOI
    What this evidence cannot show

    My analysis of their deposited data, served through the cBioPortal API; Umeda et al. do not report IRX1. The absolute anchor (log2 > 0) is not the pre-registered floor and was adopted because the floor-based verdict tracked the threshold - a reader should treat the exact rates as anchor-dependent and the M7-versus-M5 contrast as the stable part. FAB is a morphological call made locally and is missing for 14 of the 171 non-AFF1 KMT2A-r samples. The M7 comparator arm is a mixture of GLIS2, GATA1, NUP98 and HOXA drivers, not one disease. Partner calls come from the study's own structural-variant table; 8 KMT2A-SV samples have no known partner and were excluded and named.

    Evidence that argues against this (1)
    • Megakaryoblastic blasts sit next to, and can be contaminated by, the normal megakaryocyte-erythroid compartment where Nagel reports IRX1 is NORMALLY expressed. So an IRX1-positive M7 bulk sample can be reporting normal MEP transcript rather than blast transcript, and no purity or blast-count measurement was available here. This is the single most likely benign explanation for the whole M7 association and it is untested. Nagel S et al. 2022
  2. E2moderate 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, aberrant IRX1 activity in AML patient profiling data, and overexpression of IRX1, IRX3 and IRX5 in megakaryoblastic and myelomonocytic AML cell lines.

    PMID 35328612DOI
    What this evidence cannot show

    DO NOT CITE THIS PAPER FROM ITS TITLE. The title says 'Shows Normal Activity of IRX1 in Myeloid Progenitors and Reveals Ectopic Expression of IRX3 and IRX5', which reads as though IRX1 were NOT aberrant in AML. The abstract says the opposite in the same breath - 'aberrant activity of IRX1 in addition to IRX3 and IRX5' - and it is the abstract that supports this claim. The paper's primary material is public profiling data plus 100 leukaemia/lymphoma cell lines, not primary patient chromatin, and it does not stratify by fusion partner.

    Evidence that argues against this (1)
    • If IRX1 in AML is driven by GATA1/GATA2 and the megakaryocyte-erythroid axis, as that paper's own regulatory analysis suggests, then IRX1 in myeloid disease and IRX1 in t(4;11) B-ALL are two unrelated events sharing a gene name - and the M7 finding here would be evidence for the split reading rather than for a unified IRX1 programme. The 16 GATA1-category M7 cases in this cohort are the population that would test it and were not analysed separately. Nagel S et al. 2022
  3. E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    IRX1 inactivation occurs in 16.2% of 204 Down-syndrome myeloid leukaemia cases and is implicated in upregulation of the MYC pathway, so the polarity of IRX1 in paediatric megakaryoblastic disease is not settled.

    PMID 38513239DOI
    What this evidence cannot show

    Cited for polarity, not for KMT2A biology. ML-DS is a GATA1-mutant megakaryoblastic disease of Down syndrome with a different genetic background; a loss-of-function event there does not contradict an expression association here, but it does mean 'IRX1 high in megakaryoblastic leukaemia' and 'IRX1 lost in megakaryoblastic leukaemia' are both published, in different diseases, and neither has been reconciled.

    Evidence that argues against this (1)
    • This cohort contains 16 GATA1-category M7 cases, which are the closest thing in it to ML-DS, and 54 of 97 non-KMT2A-r M7 cases overall are IRX1-POSITIVE. If IRX1 were lost in GATA1-driven megakaryoblastic disease at 16.2%, the IRX1-positive majority of M7 here would sit beside it uncomfortably. The GATA1 subgroup was not scored separately and this tension is unresolved. Sato T et al. 2024
Connected to