Oncofusion

start / graph / cell-type-erythroid-progenitor

Cell typemoderate confidence 3 of 5: One strong study, or several smaller consistent ones.Q73-vocabulary

Erythroid progenitor - the committed red-cell precursor, distinct from the bipotent MEP that two entries had to borrow, and the compartment that pays for any haem-pathway or fetal-globin intervention

also called erythroid precursor · CFU-E · proerythroblast · committed erythroid progenitor · erythroblast

Erythroid progenitors are the marrow cells committed to becoming red blood cells; they are the stage after the shared red-cell/platelet ancestor. This graph needed a node for them because two of its genes have their only measured blood-cell requirement here, and because anything that attacks haem or globin production in a leukaemia will hit these cells too.

The same thing, for a scientist

Q69 recorded that `cell-type-mep` had been reused for HEXIM1 and EAF1 and was an approximation: MEP is the bipotent megakaryocyte-erythroid progenitor, and both results are about committed erythropoiesis. The evidence that earns this node its place is specific. HEXIM1 is an essential transcription regulator during **human** erythropoiesis - it controls erythroid gene expression and function, enforces RNAPII pausing at cell-cycle checkpoint genes while raising RNAPII occupancy at cycle-progression genes, and its overexpression shifts RNAPII and GATA1 from the beta-globin to the gamma-globin locus, raising **fetal globin** and the lncRNA BGLT3 [E1]. In zebrafish, deleting eaf1 or eaf2 - Super Elongation Complex components - reduces erythropoiesis and hypoxia tolerance, with reduced scl, lmo2 and gata1a expression, increased H3K27me3 specifically at the gata1a promoter and reduced TCF4 occupancy at scl and lmo2 [E2]. Two SEC-associated factors, two species, one compartment. The compartment also matters to this disease from the therapeutic side: haem biosynthesis is reported as a selective dependency in AML with inhibition collapsing mitochondrial Complex IV and inducing cuproptosis [E3], and every erythroid progenitor in the marrow runs the same pathway at maximum flux. Finally, a 2025 single-cell reference atlas of human haematopoiesis mapping 1.2 million cells from 318 leukaemia samples found unexpected AML cell states **resembling erythroid progenitors** that were prognostic within normal-karyotype AML [E4], so 'erythroid progenitor' is now a state leukaemic cells are scored against, not only a normal cell type.

Go deeper
Why MEP was the wrong node, and what changes

cell-type-mep is the bipotent megakaryocyte-erythroid progenitor. Both results that were attached to it are about cells already committed to the red lineage and already executing the erythroid transcriptional programme - HEXIM1's work is measured on globin loci with GATA1 redistribution 1, and the zebrafish result turns on gata1a, scl and lmo2 2. Attaching a globin-locus result to a bipotent progenitor overstates how early in the hierarchy the effect sits.

This entry does not delete that relationship: MEP is upstream and the link is real. What it does is give the committed compartment its own term so the two claims can be attached where they were measured.

The fetal-globin result, and why it is not a digression in this graph

HEXIM1 overexpression shifts RNAPII and GATA1 off beta-globin and onto gamma-globin, raising fetal globin and BGLT3, and GATA1 is the determinant of whether HEXIM1 represses or activates a given gene: genes gaining both HEXIM1 and GATA1 get more RNAPII and more expression, genes gaining HEXIM1 but losing GATA1 get more pausing and less 1.

This graph's central developmental claim is fetal programme retention - that infant KMT2A-rearranged disease holds onto a fetal transcriptional identity. Here is a second, unrelated system in which a P-TEFb regulator flips a locus from its adult to its fetal output. It is not evidence about the leukaemia and must not be cited as such; it is a mechanistic precedent showing that pause-release control alone can select between fetal and adult programmes at one locus. gene-hexim1 also carries the epidermal-progenitor result in which the AFF1-scaffolded SEC with inactive CDK9 holds a progenitor state - the same protein, the same kind of switch, a third tissue.

The erythroid compartment as the bill for this project's metabolic ideas

De novo haem biosynthesis has been reported as a selective dependency in AML, with inhibition of the pathway collapsing mitochondrial Complex IV, dysregulating the copper-chaperone system and inducing cuproptosis, and with glycolysis synthetic-lethal with haem biosynthesis 3.

Erythroid progenitors synthesise haem at a rate no other cell in the body approaches, because each one is filling itself with haemoglobin. A haem-pathway inhibitor given to a child with a cytopenic marrow after induction has an obvious, predictable, on-target toxicity in exactly this compartment, and none of the cited work measures it. cell-state-oxphos-dependence already raises this as a bite-back with no cell-type node to attach it to; that node now exists.

Erythroid identity as a leukaemic state, not only a normal cell type

A reference atlas of human haematopoiesis built from 263,159 single-cell transcriptomes spanning 55 cellular states, used to map more than 1.2 million cells from 318 leukaemia samples, found 12 recurrent patterns of aberrant differentiation in AML - including unexpected AML cell states resembling lymphoid and erythroid progenitors that were prognostic within normal-karyotype AML independently of genomic classification, with genotype-to-phenotype associations mapped for more than 45 genetic drivers 4.

Two consequences for this graph. Erythroid-progenitor identity is a scoreable leukaemic state, so 'is there an erythroid-like fraction in KMT2A-rearranged disease?' is now an answerable question against a public reference. And the same atlas maps differentiation landscapes by driver, which is the natural place to look for what KMT2A rearrangement does to hierarchy position - this beat did not read that paper's KMT2A results and does not claim them.

Facts and figures
ontologyCL:0000038
ontology noteCL:0000038 is 'erythroid progenitor cell'. CL:0000558 (CFU-E / colony-forming unit erythroid) and CL:0000547 (proerythroblast) are the finer terms. cell-type-mep holds CL:0000050 (megakaryocyte-erythroid progenitor), which is the BIPOTENT upstream cell and is what HEXIM1 and EAF1 had to borrow.
position in hierarchyHSC -> MPP -> CMP -> MEP (bipotent) -> committed erythroid progenitor (BFU-E, CFU-E) -> proerythroblast -> erythroblast -> reticulocyte
why this node existsQ69 reported that cell-type-mep had been reused for HEXIM1 and EAF1 and was an approximation, and declined to create a 52nd node outside its remit. Both results are about committed erythropoiesis with globin and gata1a readouts, not about a bipotent progenitor.
why this compartment matters therapeuticallyErythroid progenitors run de novo haem biosynthesis at a rate no other cell approaches. A haem-pathway inhibitor, proposed as a selective AML dependency (PMID 41265435), has its most predictable on-target toxicity here - and the reported direction of the leukaemic finding, that AML cells DOWNREGULATE the pathway, makes the normal compartment the more exposed one. cell-state-oxphos-dependence raises this as a bite-back and previously had no cell-type node to attach it to.
shown for fusionNO - nobody has looked. No study has measured the erythroid compartment, erythroid-like states or globin switching in KMT2A-rearranged leukaemia. A 2025 single-cell atlas now makes erythroid-progenitor identity a scoreable leukaemic state in AML (PMID 40294241), so the question is answerable against a public reference and has not been asked for this disease.
markers
[
  "CD71 (TFRC) high",
  "GYPA/CD235a acquired with maturation",
  "KIT+ early, lost with maturation",
  "CD41- and CD42b- (distinguishes from the megakaryocyte arm of MEP)",
  "GATA1 and KLF1 as the lineage transcription factors"
]
genes in this graph with a measured requirement here
[
  {
    "gene": "gene-hexim1",
    "requirement": "essential transcription regulator during HUMAN erythropoiesis; overexpression shifts RNAPII and GATA1 from beta- to gamma-globin, raising fetal globin and BGLT3",
    "pmid": "37738561",
    "system": "human erythroid cells"
  },
  {
    "gene": "gene-eaf1",
    "requirement": "eaf1 or eaf2 deletion reduces erythropoiesis and hypoxia tolerance via gata1a, scl and lmo2, with H3K27me3 gain specifically at the gata1a promoter",
    "pmid": "37002435",
    "system": "zebrafish embryos - held at weak confidence",
    "retraction_warning": "PMID 23717633 (Eafs and erythroid fate, PLoS One 2013) is flagged RETRACTED and overlaps this claim's Wnt arm. It is not cited anywhere in this graph and must not be."
  }
]
How this could bite us back
  • A haem-pathway inhibitor in a child with a cytopenic marrow after induction attacks the compartment that consumes most of the body's haem output. The predicted toxicity is transfusion-dependent anaemia, and no cited study measures erythropoiesis under the intervention.
  • The leukaemic cells in the haem story DOWNREGULATE the pathway while erythroid progenitors run it flat out, so the therapeutic index may be inverted relative to the intuitive reading of 'a selective AML dependency'.
  • Cuproptosis depends on copper handling, and copper status in infants is nutritionally and genetically variable, so a death mechanism keyed to a trace metal carries a pharmacodynamic confounder concentrated in the same patients.
  • Fetal globin induction is a therapeutic goal in a different disease. Any agent that raises fetal globin as a side effect - and a P-TEFb-directed agent might - will produce a haematological change that is beneficial in sickle cell disease and simply confusing in a leukaemia trial's laboratory monitoring.
Still open
  1. Is there an erythroid-like fraction in KMT2A-rearranged leukaemia? A public reference atlas of 55 human haematopoietic states now makes this a scoreable question against 318 mapped leukaemia samples, and nobody has asked it for this disease. The analysis is a projection, not an experiment.
  2. Does the HEXIM1 fetal-globin switch have anything to say about fetal programme retention in infant KMT2A-rearranged disease? Both are cases of a fetal transcriptional output being selected over an adult one, one by pause-release control at a single locus and the other genome-wide by a fusion. The shared element is P-TEFb, and nobody has asked whether the two mechanisms touch.
  3. What is the erythroid on-target cost of a haem-biosynthesis inhibitor? The AML dependency work does not measure erythropoiesis, and the compartment at highest flux is the one the drug would hit hardest. A colony assay on normal human CD34-positive cells under the same inhibitor is the missing control.
  4. Why do two Super Elongation Complex-associated factors move erythropoiesis in opposite directions - EAF1/2 loss reducing it, HEXIM1 gain promoting it? Either the SEC's role in this compartment is not a single axis, or one of the two results does not generalise across species.
Every source behind this entry (4)
  1. E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    IN HUMAN ERYTHROPOIESIS, HEXIM1 is an essential transcription regulator: it controls erythroid gene expression and function, overexpression promotes erythroid proliferation and fetal globin expression, it regulates proliferation by enforcing RNAPII pausing at cell-cycle checkpoint genes while increasing RNAPII occupancy at cycle-progression genes, and genome-wide profiling shows the most dramatic changes at the beta-globin loci with loss of RNAPII and GATA1 at beta-globin and gain at gamma-globin, increasing fetal globin and the lncRNA BGLT3; GATA1 co-occupancy determines whether HEXIM1 represses or activates a gene.

    HEXIM1 is an essential transcription regulator during human erythropoiesis. · Lv X, Murphy K, Murphy Z, Getman M, Rahman N, Nakamura Y, et al., Steiner LA · Blood 2023
    PMID 37738561DOI
    What this evidence cannot show

    Human erythroid cells, which is a strength, but the fetal-globin result comes from HEXIM1 OVEREXPRESSION, a gain-of-function manipulation whose physiological relevance is not established by the same experiment. No leukaemia and no KMT2A-rearranged cell appears anywhere in it. Already cited in gene-hexim1. Read: abstract only.

    Evidence that argues against this (1)
    • HEXIM1 is not a KMT2A-selective dependency and sits at the opposite extreme of this project's own scan - as recorded in gene-hexim1, Chronos -0.2533 in the KMT2A-rearranged arm against -0.3829 in other leukaemias, difference +0.1296, rank 17,307 of 17,728, one of the deepest anti-selective placements recorded here. That is mechanistically coherent, since HEXIM1 INHIBITS P-TEFb and a leukaemia driven by P-TEFb recruitment should need the inhibitor less - but it means the erythroid requirement documented here has no read-across to the disease, and an erythroid result for this gene should not be imported into a leukaemia argument. Internal analysis, not a published source, and the arm means carry no computable interval. Oncofusion Knowledge Graph coordinator 2026
  2. E2weak confidence 2 of 5: Preliminary, small, indirect, or not yet peer reviewed.paper

    IN ZEBRAFISH EMBRYOS, deletion of the ELL-associated factors eaf1 or eaf2 - Super Elongation Complex components - reduces hypoxia tolerance and reduces erythropoiesis, with significantly decreased expression of the erythroid transcriptional regulators scl, lmo2 and gata1a at 24 and 72 hours post fertilisation, a significant increase in H3K27me3 enrichment at the gata1a promoter but not at the scl or lmo2 promoters, and reduced phospho-beta-catenin with reduced TCF4 binding at the scl and lmo2 promoters at 24 hpf.

    PMID 37002435DOI
    What this evidence cannot show

    Zebrafish embryonic haematopoiesis - not mammalian, not human, not leukaemia, and the furthest-from-human system attached to any cell node in this graph. Held at WEAK confidence deliberately and not treated as replicated, because the most obvious prior paper linking Eaf proteins to erythroid fate (PMID 23717633, 'Eafs control erythroid cell fate by regulating c-myb expression through Wnt signaling', PLoS One 2013) is flagged RETRACTED by NCBI publication type and its claim overlaps the Wnt arm of this one. That retracted paper is not cited anywhere in this graph and must not be. Already cited in gene-eaf1. Read: abstract only.

    Evidence that argues against this (1)
    • The strongest human evidence in this node runs through a different Super Elongation Complex protein in the opposite direction on the same compartment: HEXIM1, the INHIBITOR of P-TEFb, is required for human erythropoiesis and its overexpression promotes erythroid proliferation. EAF1/2 loss reducing erythropoiesis and HEXIM1 gain promoting it cannot both be read as 'more SEC activity, more erythropoiesis'. The SEC's relationship to this compartment is therefore not a single-direction one, and a zebrafish loss-of-function does not establish it. Read: abstract only. Lv X et al. 2023
  3. E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    IN ACUTE MYELOID LEUKAEMIA, integrated analysis of mouse models, human cell lines and primary patient samples identifies de novo haem biosynthesis as a selective dependency, underpinned by a propensity of AML cells and especially leukaemic stem cells to downregulate haem biosynthesis enzymes; inhibiting those enzymes collapses mitochondrial Complex IV, dysregulates the copper-chaperone system and induces cuproptosis, with glycolysis among the synthetic-lethal partners.

    Inhibition of heme biosynthesis triggers cuproptosis in acute myeloid leukemia. · Lewis AC, Gruber E, Franich R, Armstrong J, Kelly MJ, Opazo CM, et al., Kats LM · Cell 2026
    PMID 41265435DOI
    What this evidence cannot show

    AML as a class with no KMT2A stratification reported in the abstract, and nothing about lymphoid disease. It is cited in this node not for the leukaemia claim - cell-state-oxphos-dependence already holds that - but to identify the normal compartment that would pay the on-target cost, which this paper does not measure. Read: abstract only.

    Evidence that argues against this (1)
    • The direction of the haem finding is counterintuitive and complicates the toxicity inference: AML cells and especially leukaemic stem cells DOWNREGULATE haem biosynthesis enzymes, and it is that downregulation which promotes self-renewal. If leukaemic cells run the pathway at low flux while erythroid progenitors run it at maximum, then a pathway inhibitor has a worse therapeutic index in this compartment than a naive reading suggests - the normal cell is the one most exposed. Read: abstract only. Lewis AC et al. 2026
  4. E4strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    A REFERENCE ATLAS of human haematopoiesis built from 263,159 single-cell transcriptomes spanning 55 cellular states, used to map more than 1.2 million cells across 318 leukaemia samples, revealed 12 recurrent patterns of aberrant differentiation in AML including unexpected AML cell states resembling lymphoid and ERYTHROID progenitors that were prognostic within normal-karyotype AML independently of genomic classification, with genotype-to-phenotype associations mapped for more than 45 genetic drivers.

    Single-cell Transcriptional Atlas of Human Hematopoiesis Reveals Genetic and Hierarchy-Based Determinants of Aberrant AML Differentiation. · Zeng AGX, Iacobucci I, Shah S, Mitchell A, Wong G, Bansal S, et al., Dick JE · Blood Cancer Discov 2025
    PMID 40294241DOI
    What this evidence cannot show

    AML, not ALL, so the erythroid-like states are described in myeloid disease; this beat did not read the paper and makes no claim about what it reports for KMT2A-rearranged samples specifically. 'Resembling erythroid progenitors' is a transcriptional-similarity assignment against a reference, not a demonstration of erythroid function. Read: abstract only. Not previously cited in this graph.

    Evidence that argues against this (1)
    • A cell state named by mapping to a reference atlas is a similarity score, and this graph has already been caught by that class of inference in the other direction: a shared transcriptomic signature between a normal fetal progenitor and infant ALL blasts does not establish cell of origin, because a fusion can instruct the phenotype it wants in whatever cell it lands in. An 'erythroid-like' AML state may reflect a programme the driver imposes rather than a hierarchy position the cell occupies. Read: abstract only. Lin S et al. 2016
Connected to