Hypothesis: the haem and iron dependency of KMT2A-rearranged lines is monocytic physiology, not fusion biology — and the panel cannot tell us either way
also called monocytic confound hypothesis · haem dependency lineage hypothesis
This project found that fusion-carrying leukaemia cells need iron-and-haem genes more than other leukaemia cells do. Almost all those cells are monocytes, whose normal job is recycling iron, so the finding may be about the cell rather than the fusion — and the comparison group that was supposed to settle it turns out to contain the wrong cells.
The same thing, for a scientist
`ana-depmap-kmt2a-selective-scan-full` placed four iron and haem genes among the top selective requirements of KMT2A-rearranged leukaemia lines — ALAD rank 14 of 17,728, IREB2 15, UROS 41, ALAS1 47. `ana-depmap-haem-lineage-control` excluded the erythroid/megakaryocytic explanation (0 of 11 rearranged lines erythroid or megakaryocytic; 8 such lines in the comparison group) but recorded the monocytic explanation as untestable at the available n. This hypothesis states the monocytic explanation explicitly so it can be attacked, and adds a finding that changes its status from 'untested' to 'untestable within DepMap': the four-line monocytic control group contains a PICALM-MLLT10 (CALM-AF10) line whose transformation runs through the same DOT1L/HOXA axis, an NPM1c/DNMT3A line that is itself menin-inhibitor-sensitive, and two FAB M4 myelomonocytic rather than M5 monocytic lines, leaving **one** usable comparator; and the rearranged monocytic arm is **5 independent patients, not 6**, because MOLM-13 and MOLM-14 come from the same person. The resolution therefore has to be isogenic, which is `exp-monocytic-isogenic-haem-dependency`.
The three competing explanations, stated so they can be separated
H1 — fusion. The KMT2A fusion creates or amplifies a requirement for haem biosynthesis and iron regulation. Prediction: acute degradation of the fusion abolishes the dependency before lineage identity moves.
H2 — lineage (this hypothesis). Monocytic cells require these genes as part of normal mononuclear-phagocyte physiology, and KMT2A-rearranged AML lines are monocytic, so the dependency tracks the cell not the lesion. Prediction: the dependency is present in monocytic non-rearranged cells, scales with monocytic maturation, and is unaffected by fusion degradation.
H3 — stress. Haem-pathway genes, HMOX1 in particular, are inducible through NFE2L2/Nrf2 in AML independently of lineage or of TP53 status, so the dependency could be a general stress-response feature that the rearranged panel expresses more strongly for unrelated reasons. Prediction: the dependency tracks oxidative-stress markers rather than either lineage score or fusion status.
The graph has treated this as a two-way question. It is a three-way one, and H3 has never been addressed at all.
Why the existing control cannot decide it
The control design was sound: exclude the rearranged lines entirely, split the remaining non-rearranged leukaemia lines by monocytic identity, and ask whether the haem genes separate them. If they do, H2 wins. It returned no significance (p 0.19-0.88) with the GAPDH control itself moving at p = 0.031.
ana-kmt2a-r-line-panel-monocytic-audit shows why that result carries almost no information:
- U-937 carries PICALM-MLLT10, whose transformation is DOT1L-dependent exactly as MLL-AF10's is, and which drives HOXA5/9/10 and MEIS1 to levels indistinguishable from MLL-translocated leukaemia. It is a positive sitting in the control group.
- OCI-AML-3 carries NPM1 p.Trp288Cysfs*12 with DNMT3A p.Arg882Cys, a genotype that is itself menin-inhibitor sensitive.
- OCI-AML-3 and MUTZ-8 are FAB M4 myelomonocytic, not M5 monocytic.
- Mono-Mac-1 is the only clean comparator, and it shares an individual with Mono-Mac-6, constraining any expansion of the group.
On the other side, MOLM-13 and MOLM-14 are the same patient, so the rearranged monocytic arm is five patients.
A permutation test comparing one line against thirty-five has no power to reject H2 and no power to accept it. The honest status is undetermined, and the reason is not sample size in the usual sense but that the required control cells barely exist in any public panel.
Why it matters more than a ranking dispute
If H2 is right, the four haem genes are not targets: an inhibitor would hit normal monocytes and tissue macrophages at the same dose, removing erythrophagocytosis, iron recycling and a large part of innate immune defence, in a population of infants already immunosuppressed by therapy. The therapeutic index would be close to zero, and the project would have spent effort on a lineage artefact.
If H1 is right, the target is real but still difficult, because haem biosynthesis is a housekeeping pathway shared with erythroid precursors and hepatocytes, and the predictable first toxicity is a sideroblastic-type anaemia.
If H3 is right, both readings are wrong and the interesting biology is an oxidative-stress dependency that would be approached completely differently.
The cost of distinguishing them is one isogenic degron experiment with a paired lineage readout. The cost of not distinguishing them is the whole downstream programme.
| statement | The selective requirement for ALAD, IREB2, UROS and ALAS1 in KMT2A-rearranged leukaemia cell lines is a property of monocytic-lineage physiology rather than of the fusion, and the DepMap panel cannot distinguish the two because the required monocytic non-rearranged control group contains one usable line after removing a HOXA-axis fusion line, a menin-sensitive NPM1c line and two myelomonocytic rather than monocytic lines. |
|---|---|
| origin | data-pattern |
| prior probability | 0.4 |
| decision value | Determines whether four DepMap hits are targets or artefacts. If lineage-driven, a haem-pathway inhibitor would remove erythrophagocytosis, iron recycling and innate immune defence from normal monocytes and macrophages at the same dose, giving a therapeutic index near zero in infants already immunosuppressed by therapy. |
predictions
[ "Acute degradation of the KMT2A fusion does NOT abolish haem-gene dependency within 24 hours, before the monocytic signature score has moved.", "Mono-Mac-1, monocytic and non-rearranged and free of a HOXA-axis fusion, shows the same haem-gene dependency as the KMT2A-rearranged monocytic lines.", "PMA-induced monocytic differentiation of THP-1 increases haem-gene dependency in proportion to the monocytic maturation score, independently of fusion status.", "Re-running the original non-rearranged split with U-937, OCI-AML-3 and MUTZ-8 reassigned does not rescue significance, because the remaining group is too small — i.e. the design fails for a structural reason, not a statistical one.", "Haem FLUX measured by stable-isotope labelling is higher in monocytic than in non-monocytic leukaemia cells regardless of fusion status; if flux does not differ while transcript does, the transcriptional signature is not reporting metabolic reality." ]
experiments
[ "exp-monocytic-isogenic-haem-dependency" ]
analyses
[ "ana-kmt2a-r-line-panel-monocytic-audit" ]
falsifiers
[ "Haem-gene dependency falls within 24 hours of acute fusion degradation, before the monocytic signature moves — this would establish the dependency as fusion-driven and retire the hypothesis outright.", "Mono-Mac-1 shows no haem-gene dependency while the rearranged monocytic lines do, with matched monocytic maturation scores.", "PMA-driven monocytic differentiation of THP-1 does not raise haem-gene dependency, removing the proposed mechanism by which lineage would create it.", "Direct measurement shows the monocyte-macrophage lineage runs high haem CATABOLIC but not high haem SYNTHETIC flux, which would mean monocytic physiology cannot explain a dependency on ALAD, UROS and ALAS1 — the internal objection recorded in E2 counter-evidence, and the cheapest single measurement that could kill the hypothesis." ]
competing hypotheses
[
{
"label": "H1 fusion-driven",
"discriminating_prediction": "dependency falls on acute fusion degradation before lineage moves"
},
{
"label": "H2 lineage-driven (this hypothesis)",
"discriminating_prediction": "dependency present without a fusion and scaling with monocytic maturation"
},
{
"label": "H3 stress-driven via NFE2L2/HMOX1",
"discriminating_prediction": "dependency tracks oxidative-stress markers rather than lineage score or fusion status; never previously considered in this graph"
}
]panel facts
{
"rearranged_monocytic_patients": 5,
"rearranged_monocytic_lines": 6,
"clean_control_lines": 1,
"haem_gene_ranks_of_17728": {
"ALAD": 14,
"IREB2": 15,
"UROS": 41,
"ALAS1": 47
}
}- If this hypothesis is right, an intervention built on the haem finding would be on-target against normal mononuclear phagocytes, and the toxicity would be indistinguishable from the intended effect.
- If it is wrong, the target is still a housekeeping biosynthetic pathway shared with erythroid precursors and hepatocytes, so a positive result creates a toxicity problem rather than removing one.
- The audit that motivates this hypothesis would, if applied consistently, shrink the control arm of several existing comparisons in this graph and reduce the apparent strength of previously reported selective dependencies.
- Because fusion degradation itself moves lineage, the isogenic experiment designed to separate H1 from H2 can return an uninterpretable answer, and the graph should plan for that outcome rather than assume a verdict.
- Does the monocyte-macrophage lineage run high haem BIOSYNTHETIC flux, or only high catabolic flux? The hypothesis depends on the first and only the second is documented, so this is the single measurement most likely to settle the matter cheaply.
- Has H3, the NFE2L2/HMOX1 oxidative-stress explanation, ever been considered for this finding? It has not been considered in this graph, and it predicts a different experiment from either of the other two.
- Are there monocytic, non-KMT2A-rearranged, non-HOXA-axis, non-NPM1c leukaemia lines in any public panel outside DepMap? If not, the between-line design is permanently unavailable and every future attempt will fail the same way.
- How many other selective-dependency claims in this graph rest on a control group screened only for KMT2A rearrangement and not for pathway-equivalent lesions?
- E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
Monocytic AML carries a distinct metabolic and apoptotic programme — loss of BCL2, dependence on MCL1 to mediate oxidative phosphorylation — and responses to venetoclax-azacitidine track developmental stage, with monocytic subpopulations selected at relapse; monocytic identity is therefore a metabolically consequential cell state and not a cosmetic classification.
Monocytic Subclones Confer Resistance to Venetoclax-Based Therapy in Patients with Acute Myeloid Leukemia · Pei S, Pollyea DA, Gustafson A, Stevens BM, Minhajuddin M, Fu R, et al., Jordan CT · Cancer Discovery 2020What this evidence cannot show
Establishes a distinct monocytic metabolic programme in AML in general; it does not measure haem biosynthesis or iron-regulatory gene dependency, so it supports the plausibility of H2 rather than H2 itself. Read: abstract only.
Evidence that argues against this (1)
- Monocytic differentiation was not clinically predictive of azacitidine/venetoclax response in an independent integrated cohort even though cultured monocytic cells were resistant in vitro, with response instead tracking the BCL2/BCL-xL/MCL1 ratio in leukaemic stem cells. If the monocytic metabolic phenotype is partly a property of cultured cells, then a dependency measured across cultured cell lines may be reporting a culture artefact rather than either lineage or fusion. Read: abstract only. Waclawiczek A et al. 2023
- E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
Haem-iron handling is a defining lineage function of the monocyte-macrophage compartment: macrophages with both Nramp1/SLC11A1 and DMT1/SLC11A2 transport disabled show severely compromised recycling of haemoglobin-derived iron, a blunted labile iron pool, less iron released and diminished upregulation of ferroportin and haem oxygenase 1 after erythrophagocytosis.
Both Nramp1 and DMT1 are necessary for efficient macrophage iron recycling · Soe-Lin S, Apte SS, Mikhael MR, Kayembe LK, Nie G, Ponka P · Experimental Hematology 2010What this evidence cannot show
Mouse macrophage line; concerns iron transport and haem catabolism, whereas three of the four DepMap hits (ALAD, UROS, ALAS1) are haem BIOSYNTHESIS genes. The lineage argument would be tighter if monocytic cells were shown to depend on haem synthesis rather than on haem breakdown, and that has not been shown. Read: abstract only.
Evidence that argues against this (1)
- This is the strongest internal objection to H2 and it is recorded here rather than buried: the monocyte-macrophage lineage is characterised by haem CATABOLISM and iron recycling, not by haem SYNTHESIS, and the DepMap hits ALAD, UROS and ALAS1 are synthetic-pathway enzymes. The cell type whose physiology would most obviously predict a haem-biosynthesis dependency is the erythroid precursor, which the lineage control has already excluded. H2 therefore requires an additional assumption — that monocytic cells run high synthetic flux as well as high catabolic flux — that no cited source establishes. Ali D et al. 2016
- E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
U-937, one of the four monocytic non-rearranged comparison lines, carries the PICALM-MLLT10 (CALM-AF10) fusion, whose transformation is abolished by Dot1l inactivation exactly as MLL-AF10's is, and whose leukaemias over-express HOXA5, HOXA9, HOXA10 and MEIS1 at levels similar to MLL-translocated acute leukaemia; the control group therefore contains a line driven through the same axis as the index group.
Abrogation of MLL-AF10 and CALM-AF10-mediated transformation through genetic inactivation or pharmacological inhibition of the H3K79 methyltransferase Dot1l · Chen L, Deshpande AJ, Banka D, Bernt KM, Dias S, Buske C, et al., Armstrong SA · Leukemia 2013What this evidence cannot show
Mouse transformation models establishing a shared DOT1L requirement, not a shared haem requirement. Whether U-937 would show the haem dependency has not been measured; its exclusion from the control arm is a precaution against a shared upstream axis. Read: abstract only.
Evidence that argues against this (1)
- The shared axis did not produce shared clinical behaviour: single-agent pinometostat reduced H3K79 methylation but gave complete remission in only 2 of 51 adults with advanced acute leukaemia, both t(11;19). A genetic dependency shared in mouse models is a weak basis for disqualifying a control line, and disqualifying U-937 costs the control arm 25 percent of its membership on an argument that has not been tested for this readout. Read: abstract only. Stein EM et al. 2018
- E4established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper
MOLM-13 and MOLM-14 were established from the peripheral blood of a single patient at relapse of FAB M5a acute monocytic leukaemia that had evolved from myelodysplastic syndrome, both expressing monocyte-specific esterase and MLL-AF9 fusion mRNA; they are therefore not independent observations in a cell-line panel.
Two acute monocytic leukemia (AML-M5a) cell lines (MOLM-13 and MOLM-14) with interclonal phenotypic heterogeneity showing MLL-AF9 fusion resulting from an occult chromosome insertion, ins(11;9)(q23;p22p23) · Matsuo Y, MacLeod RA, Uphoff CC, Drexler HG, Nishizaki C, Katayama Y, et al., Orita K · Leukemia 1997What this evidence cannot show
Both lines derive from AML that evolved from MDS and carry trisomy 8 present in the MDS phase, so their genetic background differs from de novo paediatric KMT2A-rearranged disease in ways that could themselves affect metabolic dependencies. Read: abstract only.
Evidence that argues against this (1)
- Collapsing the pair to one observation may over-correct. The founding report documents interclonal phenotypic heterogeneity between them — MOLM-13 is CD34+CD13-CD14- while MOLM-14 is CD4+CD13+CD14+ — so they sample different subclones at different points of monocytic maturation. For a dependency that varies with monocytic maturation, which is precisely what H2 proposes, the pair carries genuine information and treating it as n=1 discards it. Read: abstract only (same paper). Matsuo Y et al. 1997
a class-wide AML heme dependency is an alternative explanation to monocytic identity for the same signal, and neither has been excluded
states explicitly the alternative that analysis left open
the isogenic design that could settle it
the audit that shows the between-line design is unavailable
the lineage physiology the hypothesis invokes
- related to →The whole protein-coding genome: menin is still first, and a metabolic pathway appears (run)
the scan whose four haem hits are under question
a pathway-wide haem signal is more consistent with a lineage explanation than a narrow one, not less