Haemogenic endothelium — the vessel-wall cell that gives birth to blood
also called hemogenic endothelium · HE · haemogenic endothelial cell · hemogenic endothelial progenitor · HEP · HSC-primed HEC · endothelial precursor of haematopoiesis
Haemogenic endothelium is a rare cell that looks and behaves like part of a blood-vessel wall but can peel away and turn into a blood stem cell. It exists for only a few days in the embryo, mostly in the floor of the main artery, and it is where the entire blood system comes from.
The same thing, for a scientist
Haemogenic endothelium (HE) is a transient, spatially restricted endothelial population competent to undergo the endothelial-to-haematopoietic transition (GO:0098508) and generate definitive haematopoietic progenitors and HSCs. It is phenotypically endothelial — CD34+, CDH5/VE-cadherin+, TEK/TIE2+, PECAM1+, KDR+, CD45-negative, CD43/SPN-negative — and is distinguished from non-haemogenic endothelium by RUNX1 expression and, in human embryos, by an arterial signature with RUNX1/MYB/ANGPT1 upregulation and CD44 positivity within the CD34+CD45- endothelial fraction [E3]. In human pluripotent-stem-cell differentiation the corresponding progenitor is defined as VE-cadherin+CD73-CD235a/CD43- and arises from VE-cadherin-negative KDR-bright APLNR+ PDGFRA-low/negative hematovascular mesodermal precursors [E4]. RUNX1 is required specifically in VE-cadherin+ endothelium for intra-arterial cluster and HSC formation and dispensable once Vav1 is expressed [E2], and the transition has been imaged directly in the ventral mouse aorta [E1]. Anatomically HE activity maps to the AGM/dorsal aorta (first and most potent in human, CS10-14), the yolk sac, vitelline and umbilical arteries and the placenta, with subsequent fetal liver colonisation and expansion [E3][E6][E7]. It is a plausible but unproven cell of origin for prenatally acquired KMT2A fusions, and the only direct evidence that MLL-AF4 acts at this stage is human ESC work [E8][E9].
What it is, and why it deserves its own node1 deeper
Most cells in a blood vessel stay blood vessel. A small, short-lived subset does not: it dissolves its tight junctions, loses apical-basal polarity, rounds up and delaminates into the vessel lumen as a blood cell. That subset is haemogenic endothelium.
The process is formally named in the Gene Ontology — GO:0098508, endothelial to hematopoietic transition: "the generation of hematopoietic stem cells from hemogenic endothelial cells by a process that includes tight-junction dissolution and loss of cell polarity followed by delamination from the endothelium."
HE matters to this project for one reason: KMT2A::AFF1 is acquired prenatally, and HE is one of the few cells that exists in the right place at the right time to be the cell that acquires it. Whether it actually is remains open — see the open questions — but it cannot be dismissed, and the graph needs it as a first-class cell type rather than a footnote inside a hematopoiesis summary.
Ontology status: there is no Cell Ontology term for it
This was checked directly against the EBI Ontology Lookup Service (OLS4) over the Cell Ontology. Queries for "hemogenic", "hemogenic endothelial cell", "hemogenic endothelium" and "haemogenic" return no CL class; the only CL hit is the unrelated CL:0000037 (haematopoietic stem cell), which matches on definition text. A full listing of the 332 CL classes matching "endothelial" contains no hemogenic/haemogenic term.
So the data.ontology field is deliberately omitted rather than filled with a guess. The nearest existing terms, which are not synonyms and should not be substituted:
GO:0098508— endothelial to hematopoietic transition (the process, not the cell)CL:0002418— hemangioblast: "a pluripotent cell in the yolk sac that can give rise to mesenchymal cells including erythrocytes and endothelial cells". This is upstream of HE and distinct: Choi et al. showed hemangioblasts, which form endothelium and primitive blood, originate from more immature APLNR+PDGFRA+ mesoderm, whereas haemogenic endothelial progenitors come from VE-cadherin-negative KDR-bright APLNR+PDGFRA-low/negative hematovascular mesodermal precursors 4.CL:0002544— aortic endothelial cell;CL:0000071— blood vessel endothelial cell. Anatomically overlapping, functionally not equivalent.MeSH:D055018— Hemangioblasts (the MeSH thesaurus term that "hemogenic endothelium" text-matches to).
The absence of a CL term is itself worth recording: any single-cell atlas that annotates HE is doing so with an ad-hoc label, which makes cross-atlas comparison unreliable.
Markers — and the two that actually discriminate2 deeper
Positive / endothelial-shared. CD34, CDH5 (VE-cadherin), TEK (TIE2), PECAM1 (CD31), KDR (VEGFR2), CD105/ENG.
Positive / haemogenic-defining. RUNX1 — the single most important marker, because it is what separates haemogenic from non-haemogenic endothelium functionally as well as phenotypically 2. In human embryos, HSC-primed haemogenic endothelial cells also show an unambiguous arterial endothelial signature with upregulation of RUNX1, MYB and ANGPT1, and transiently over-represent EMCN, PROCR and RUNX1T1 as arterial ECs make the hemogenic fate choice 3. MYB, GATA2, GFI1 and SOX17 sit in the same regulatory neighbourhood.
Negative — this is the part people get wrong.
- PTPRC (CD45): negative to low. HE has not yet turned on the pan-leukocyte phosphatase.
- SPN (CD43, leukosialin): negative on HE, acquired at the transition. CD43 is detected on all emerging clonogenic haematopoietic progenitors before CD45 appears, and it reliably separates the haematopoietic CD34+ population from CD34+**CD43-**CD31+KDR+ endothelial and CD34+CD43-CD31-KDR- mesenchymal cells 5. CD43 acquisition is therefore the cleanest single-marker readout that the EHT has fired.
Enrichment strategies actually used.
- Human embryo (in vivo): subcategorising CD34+CD45- endothelial cells by CD44 positivity enriched HSC-primed HECs more than 10-fold 3. This is the most practical published in vivo sorting handle.
- Human PSC differentiation (in vitro): haemogenic endothelial progenitors capable of generating definitive haematopoietic cells are identified precisely by a VE-cadherin+ CD73- CD235a/CD43- phenotype, which discriminates true HEPs from VE-cadherin+CD73+ non-HEPs and from VE-cadherin+CD235a+CD41a- early haematopoietic cells 4.
Two molecularly distinct human HEC populations, not one
The human single-cell data complicate the textbook picture in a way that matters for timing arguments. Zeng et al. captured HSC-primed HECs at Carnegie stage 12-14 with a clear arterial identity, and separately detected another temporally and molecularly distinct intra-embryonic HEC population, mainly at the earlier CS10, which lacks the arterial feature 3.
So "haemogenic endothelium" is at least two cell states with different timings and different transcriptional programmes. If a fusion acquired at CS10 lands in the non-arterial HEC and a fusion acquired at CS13 lands in the arterial HSC-primed HEC, those are different cells of origin with different accessible enhancer landscapes — and therefore potentially different diseases. Nobody has tested this.
Where haemogenic competence comes from
HE is not the top of the hierarchy. Competence is endowed earlier, in extra-embryonic mesoderm: transient expression of the T-box transcription factor Eomesodermin (Eomes) governs haemogenic competency of yolk-sac mesoderm, by regulating accessibility of the enhancers that SCL/TAL1 uses, and Eomes is essential for normal development of RUNX1+ haemogenic endothelium. Eomes loss blocks blood-progenitor formation but not specification of FLK-1+ haematoendothelial progenitors 9.
That is a useful structural point for this project: haemogenic competence is set at the chromatin-accessibility level, upstream of RUNX1, before the cell looks like endothelium at all. A KMT2A fusion arriving in such a cell would be reading a landscape that was configured by Eomes/SCL, not by the B-lineage factors that shape the eventual leukemia.
Where and when — the anatomy and the clock
Mouse. HSCs are first detected in the dorsal aorta from embryonic day 10.5. Haematopoietic cell clusters sit attached to the aortic endothelium, and time-lapse confocal imaging of the deep-lying aorta captured HSC (Sca1+, c-Kit+, CD41+) emergence directly from ventral aortic haemogenic endothelial cells 1. RUNX1 is required in the VE-cadherin+ endothelium of the aorta, vitelline and umbilical arteries, yolk sac and placenta for intra-arterial cluster, progenitor and HSC formation 2.
Human. Highly potent HSCs emerge first in the AGM region, specifically the dorsal aorta, and only later appear in the yolk sac, liver and placenta. AGM-region cells transplanted into immunodeficient mice gave long-term high-level multilineage repopulation, and a single AGM-derived HSC generated at least 300 daughter HSCs in a primary recipient, disseminating through the whole recipient marrow and remaining retransplantable 6. Transcriptomically, the endothelial-to-HSC transition in human embryos is captured at Carnegie stage 12-14 (roughly 4-5 weeks of gestation) 3.
Placenta. The human placenta contains haematopoietic progenitors and HSCs throughout development including at term; placenta-derived stromal lines are pericyte-like and support human haematopoiesis, and haematopoietic cells sit in direct contact with pericytes/perivascular cells in placental sections 7. Whether the placenta generates blood de novo from HE or only harbours and expands it is not resolved by that study.
Then the liver. After emergence, HSCs colonise the fetal liver, which becomes the major site of expansion for the rest of gestation before the shift to bone marrow around birth. The fetal liver HSC — not the HE cell — is the compartment most often proposed as the actual KMT2A::AFF1 cell of origin, and it has its own node (cell-type-fetal-liver-hsc). HE is the stage before that.
The window is short. HE exists for days, is rare within the endothelium, and disappears. That is simultaneously why it is a plausible origin for a prenatal-only fusion and why it is nearly impossible to interrogate in patients.
Can a KMT2A fusion act at this stage? What is actually shown
Be careful here, because this is the section most likely to be over-read.
What is shown. In human embryonic stem cell differentiation:
- MLL-AF4 enhances the specification of hemogenic precursors from hESCs but strongly impairs further haematopoietic commitment in favour of an endothelial cell fate, with transcriptional priming toward hemogenic precursors prone to endothelial maturation and upregulation of vascular-endothelial master genes. And critically: MLL-AF4 was not sufficient to transform hESC-derived haematopoietic cells 8.
- MLL-AF4 plus the reciprocal AF4-MLL strongly promotes the emergence of hemato-endothelial precursors, both endothelial- and hemogenic-primed, which specify into more haematopoietic and more endothelial committed cells, without changes in survival or proliferation; H3K79me3 was specifically enriched over HOXA cluster genes in the double-fusion differentiating cells 9.
What is not shown. No study has demonstrated a KMT2A fusion arising in, or acting on, a bona fide in vivo haemogenic endothelial cell — human or mouse. No patient-derived evidence places the fusion in an endothelial or HE compartment. The hESC results are enforced-expression in vitro phenotypes at a branch point that resembles the hemato-endothelial transition; they establish that the fusion can perturb this decision, not that it does in a patient.
The most useful inference. These two papers together say the fusion is transcriptionally competent at the hemato-endothelial branch point and affects lineage output there, but is insufficient to transform at that stage. That is an argument that the cell of origin is likely downstream of HE — a fetal HSC or lympho-myeloid progenitor with a permissive landscape — while still allowing that the translocation itself could have occurred earlier. Those are separate questions and the literature routinely conflates them.
| location | aorta-gonad-mesonephros region / ventral floor of the dorsal aorta (primary and most potent in human) → also yolk sac, vitelline and umbilical arteries, and placenta → progeny colonise and expand in fetal liver, then bone marrow around birth |
|---|---|
| process ontology | GO:0098508 (endothelial to hematopoietic transition) |
| ontology note | No Cell Ontology (CL) class for haemogenic endothelium / hemogenic endothelial cell exists in the OLS4 index as retrieved on 2026-09-18; the data.ontology field is therefore omitted rather than populated with an incorrect id. |
| lifespan | transient; days. Rare within the endothelium and absent after the HSC-generating window closes |
markers
[
{
"marker": "CD34",
"gene": "CD34",
"status": "positive",
"note": "shared with endothelium and HSPC; cannot discriminate lineage alone"
},
{
"marker": "VE-cadherin",
"gene": "CDH5",
"status": "positive",
"note": "defining endothelial marker; Runx1 requirement is mapped to this compartment (E2)"
},
{
"marker": "TIE2",
"gene": "TEK",
"status": "positive",
"note": "angiopoietin receptor; shared with HSC"
},
{
"marker": "CD31",
"gene": "PECAM1",
"status": "positive"
},
{
"marker": "VEGFR2 / KDR",
"gene": "KDR",
"status": "positive",
"note": "KDR-bright in the immediate mesodermal precursor (E4)"
},
{
"marker": "RUNX1",
"gene": "RUNX1",
"status": "positive",
"note": "the haemogenic-defining factor; functionally required at the transition (E2)"
},
{
"marker": "MYB",
"gene": "MYB",
"status": "positive",
"note": "upregulated in human HSC-primed HECs (E3)"
},
{
"marker": "ANGPT1",
"gene": "ANGPT1",
"status": "positive",
"note": "upregulated in human HSC-primed HECs (E3)"
},
{
"marker": "CD44",
"gene": "CD44",
"status": "positive",
"note": "enriches human HSC-primed HECs >10-fold within CD34+CD45- endothelium (E3)"
},
{
"marker": "EMCN",
"gene": "EMCN",
"status": "transient",
"note": "transiently over-represented on hemogenic fate choice (E3)"
},
{
"marker": "PROCR",
"gene": "PROCR",
"status": "transient",
"note": "transiently over-represented on hemogenic fate choice (E3)"
},
{
"marker": "RUNX1T1",
"gene": "RUNX1T1",
"status": "transient",
"note": "transiently over-represented on hemogenic fate choice (E3)"
},
{
"marker": "CD45",
"gene": "PTPRC",
"status": "negative to low",
"note": "acquired after the transition (E5)"
},
{
"marker": "CD43 / leukosialin",
"gene": "SPN",
"status": "negative on HE; acquired at the transition",
"note": "earliest reliable haematopoietic discriminator, precedes CD45 (E5)"
},
{
"marker": "CD73",
"gene": "NT5E",
"status": "negative",
"note": "CD73-negativity discriminates true hPSC-derived HEPs from non-haemogenic endothelium (E4)"
},
{
"marker": "CD235a / glycophorin A",
"gene": "GYPA",
"status": "negative",
"note": "part of the VE-cadherin+CD73-CD235a/CD43- HEP definition (E4)"
},
{
"marker": "PDGFRA",
"gene": "PDGFRA",
"status": "low to negative in the immediate precursor",
"note": "distinguishes hematovascular mesodermal precursors from PDGFRA+ hemangioblast-forming mesoderm (E4)"
}
]immunophenotype shorthand
{
"human_embryo_in_vivo": "CD34+ CD45- CD44+ (RUNX1+, arterial signature) — enriches HSC-primed HECs >10-fold (E3)",
"human_psc_in_vitro": "VE-cadherin+ CD73- CD235a/CD43- (E4)",
"immediate_precursor": "VE-cadherin-negative KDR-bright APLNR+ PDGFRA-low/negative hematovascular mesodermal precursor (E4)",
"post_transition": "CD34+ CD43+ CD45-/+ (E5)"
}anatomical sites
[
{
"site": "dorsal aorta (AGM), ventral floor",
"species": "mouse and human",
"timing": "mouse from E10.5; human Carnegie stage 12-14 for HSC-primed HECs, with a distinct non-arterial HEC population at CS10",
"role": "definitive HSC generation; first and most potent in human",
"evidence": [
"E1",
"E3",
"E6"
]
},
{
"site": "yolk sac",
"species": "mouse and human",
"role": "Runx1-dependent haemogenic endothelium; haemogenic competence set by Eomes in yolk-sac mesoderm",
"evidence": [
"E2",
"E10"
]
},
{
"site": "vitelline and umbilical arteries",
"species": "mouse",
"role": "Runx1-dependent intra-arterial cluster formation",
"evidence": [
"E2"
]
},
{
"site": "placenta",
"species": "mouse and human",
"role": "harbours HSPCs throughout development including at term, with pericyte-like supportive stroma; de novo generation vs colonisation unresolved",
"evidence": [
"E2",
"E7"
]
},
{
"site": "fetal liver",
"species": "human",
"role": "downstream colonisation and expansion site, not a site of EHT; see cell-type-fetal-liver-hsc",
"evidence": [
"E6"
]
}
]related ontology terms
[
{
"id": "GO:0098508",
"label": "endothelial to hematopoietic transition",
"relation": "the process this cell type performs"
},
{
"id": "CL:0002418",
"label": "hemangioblast",
"relation": "upstream and distinct — yolk-sac pluripotent precursor, arises from PDGFRA+ mesoderm (E4)"
},
{
"id": "CL:0002544",
"label": "aortic endothelial cell",
"relation": "anatomical parent; most aortic endothelium is not haemogenic"
},
{
"id": "CL:0000071",
"label": "blood vessel endothelial cell",
"relation": "broader parent class"
},
{
"id": "CL:0000037",
"label": "hematopoietic stem cell",
"relation": "the immediate product of the transition"
}
]atlas refs — the numbers behind the chart
[
{
"name": "Human embryonic endothelial-to-HSC transition scRNA-seq landscape (CS10-CS17)",
"source": "Zeng et al., Cell Research 2019",
"pmid": "31501518",
"note": "defines HSC-primed HECs, the CD44 enrichment strategy and the aortic niche components",
"title": "Tracing the first hematopoietic stem cell generation in human embryo by single-cell RNA sequencing.",
"authors": "Zeng Y, He J, Bai Z, et al.",
"venue": "Cell Res",
"year": 2019,
"title_provenance": "Title/authors/venue backfilled from PubMed esummary on 2026-09-20 because this citation stored a bare PMID with no title, which made it unscoreable by verify-citations.py. BACKFILLING IS NOT VERIFICATION: it confirms the PMID resolves to a real paper, not that this is the right paper for the claim. That still requires reading."
},
{
"name": "Human PSC hemato-endothelial differentiation phenotyping",
"source": "Choi et al., Cell Reports 2012",
"pmid": "22981233",
"note": "VE-cadherin+CD73-CD235a/CD43- HEP definition and the HVMP precursor",
"title": "Identification of the hemogenic endothelial progenitor and its direct precursor in human pluripotent stem cell differentiation cultures.",
"authors": "Choi KD, Vodyanik MA, Togarrati PP, et al.",
"venue": "Cell Rep",
"year": 2012,
"title_provenance": "Title/authors/venue backfilled from PubMed esummary on 2026-09-20 because this citation stored a bare PMID with no title, which made it unscoreable by verify-citations.py. BACKFILLING IS NOT VERIFICATION: it confirms the PMID resolves to a real paper, not that this is the right paper for the claim. That still requires reading."
},
{
"name": "hESC/OP9 haematopoietic commitment marker series",
"source": "Vodyanik et al., Blood 2006",
"pmid": "16757688",
"note": "CD43-before-CD45 ordering",
"title": "Leukosialin (CD43) defines hematopoietic progenitors in human embryonic stem cell differentiation cultures.",
"authors": "Vodyanik MA, Thomson JA, Slukvin II",
"venue": "Blood",
"year": 2006,
"title_provenance": "Title/authors/venue backfilled from PubMed esummary on 2026-09-20 because this citation stored a bare PMID with no title, which made it unscoreable by verify-citations.py. BACKFILLING IS NOT VERIFICATION: it confirms the PMID resolves to a real paper, not that this is the right paper for the claim. That still requires reading."
},
{
"name": "Mouse E8.5 gastrulation/organogenesis atlas used for Eomes haemogenic competence analysis",
"source": "Harland et al., Nature Cell Biology 2021",
"pmid": "33420489",
"title": "The T-box transcription factor Eomesodermin governs haemogenic competence of yolk sac mesodermal progenitors.",
"authors": "Harland LTG, Simon CS, Senft AD, et al.",
"venue": "Nat Cell Biol",
"year": 2021,
"title_provenance": "Title/authors/venue backfilled from PubMed esummary on 2026-09-20 because this citation stored a bare PMID with no title, which made it unscoreable by verify-citations.py. BACKFILLING IS NOT VERIFICATION: it confirms the PMID resolves to a real paper, not that this is the right paper for the claim. That still requires reading."
}
]kmt2a relevance
{
"plausible_cell_of_origin": true,
"demonstrated_in_vivo": false,
"what_is_shown": "MLL-AF4 alone enhances hemogenic precursor specification but skews maturation to endothelium and does not transform (E8); MLL-AF4 plus AF4-MLL expands both hemogenic- and endothelial-primed hemato-endothelial precursors with HOXA H3K79me3 enrichment (E9). Both are hESC in vitro models.",
"what_is_not_shown": "No KMT2A fusion has been shown to arise in, or transform, a bona fide in vivo haemogenic endothelial cell, in any species; no patient data place the fusion in an endothelial or HE compartment.",
"inference": "The insufficiency result (E8) argues the transforming cell of origin is downstream of HE — most likely a fetal liver HSC or lympho-myeloid progenitor — while leaving open that the translocation event itself could occur at or before the HE stage. Origin of the lesion and origin of the leukemia are separate questions."
}- If the cell of origin is a transient embryonic cell that no longer exists by the time of diagnosis, then no therapy can target it. The origin cell is an explanatory and preventive concept, not a therapeutic one, and it should not be sold as a target.
- Any therapeutic that exploits an HE-like or hemato-endothelial program risks hitting adult endothelium and vascular homeostasis, since the shared markers (CDH5, TEK, KDR, CD34) are all expressed on normal adult vasculature. That is a bleeding, hypertension and wound-healing toxicity profile.
- The hESC evidence cuts against the lab as much as for it: MLL-AF4 alone was explicitly not sufficient to transform at this stage [E8]. A model built on an HE origin has to explain what the second event is, and if the answer is 'a later cellular context', then the HE stage may be irrelevant to therapy.
- Marker-based conclusions here are fragile. CD34+ alone cannot distinguish blood from vessel, and much of the historical 'endothelial precursor of leukemia' literature rests on exactly that ambiguity. Building on it risks inheriting a decades-old artefact.
- Is there any Cell Ontology-level consensus definition of haemogenic endothelium? There is currently no CL term [E11], so every atlas annotates it ad hoc and cross-dataset comparison is unreliable. Proposing a CL class is a small, concrete, citable contribution the lab could make.
- Human embryos contain at least two molecularly distinct HEC populations — an arterial HSC-primed one at CS12-14 and a non-arterial one at CS10 [E3]. Would a KMT2A fusion arriving in each produce the same disease? Their enhancer landscapes differ, so the prediction is no, and this has never been tested.
- Does the placenta generate blood from its own haemogenic endothelium, or only harbour and expand AGM-derived HSCs [E7]? This matters because the placenta is the one embryonic haematopoietic site that is routinely available as human tissue at term.
- Can a KMT2A::AFF1 knock-in be expressed specifically in CDH5+RUNX1+ haemogenic endothelium (rather than constitutively or in committed progenitors) and does it then produce leukemia? Existing mouse models express the fusion in or downstream of committed compartments, so the HE-stage experiment has not been done.
- If haemogenic competence is set by Eomes-dependent enhancer accessibility upstream of RUNX1 [E10], does the accessible landscape at that stage contain the KMT2A::AFF1 target repertoire at all? A single ATAC/H3K79me2 comparison between HE and fetal liver HSC would say whether HE is even a readable context for this fusion.
- E1established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper
Time-lapse confocal imaging of the mouse embryonic dorsal aorta shows dynamic de novo emergence of phenotypically defined HSCs (Sca1+, c-Kit+, CD41+) directly from ventral aortic haemogenic endothelial cells; HSCs are first detected in the dorsal aorta from E10.5, with haematopoietic clusters attached to the aortic endothelium.
In vivo imaging of haematopoietic cells emerging from the mouse aortic endothelium · Boisset JC, van Cappellen W, Andrieu-Soler C, Galjart N, Dzierzak E, Robin C · Nature 2010What this evidence cannot show
Mouse; required a new dissection procedure to expose the deep aorta, so the imaged tissue is an explant preparation rather than a wholly intact embryo. Emerging cells were identified by surface phenotype in the imaging field, not by transplanting the individual imaged cell, so 'HSC' here is phenotypic rather than functionally proven per cell. The equivalent human event has never been imaged.
Evidence that argues against this (1)
- Haemogenic endothelium imaged in the ventral dorsal aorta is not the only haemogenic endothelium. Genetic fate mapping of HE subsets in mouse shows a separate, earlier wave of HE in the vitelline and umbilical arteries between E8.5 and E9.5 that generates fetal-restricted HSPCs which are transient, distinct from erythro-myeloid progenitors, contribute extensively to fetal lympho-myelopoiesis and decline postnatally — i.e. they precede and are not the definitive HSCs the aortic imaging captures. Any claim that 'the fusion acts in haemogenic endothelium' therefore does not specify a single cell. Read: abstract only. Barone C et al. 2026
- E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper
Runx1 activity in vascular-endothelial-cadherin-positive endothelial cells is essential for intra-arterial cluster, haematopoietic progenitor and HSC formation in the aorta, vitelline and umbilical arteries, yolk sac and placenta, and is not required in cells already expressing Vav1, placing the requirement exactly at the endothelial-to-haematopoietic transition.
Runx1 is required for the endothelial to haematopoietic cell transition but not thereafter · Chen MJ, Yokomizo T, Zeigler BM, Dzierzak E, Speck NA · Nature 2009What this evidence cannot show
Mouse conditional genetics; the conclusion is bounded by VE-cadherin-Cre and Vav1-Cre specificity and by deletion timing/efficiency. It shows Runx1 is required in a VE-cadherin-expressing compartment, not that every definitive HSC transits an endothelial state.
Evidence that argues against this (1)
- This is normal developmental haematopoiesis in mouse, and the beat uses it as background for where a human fusion might arise. Runx1 requirement in VE-cadherin-positive endothelium constrains the ontogeny of HSCs; it says nothing about whether a KMT2A rearrangement occurring in that compartment would produce the disease, which is the question the entry exists to address. Oncofusion knowledge graph (internal analysis) 2026
- E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
Single-cell RNA sequencing of human embryos captured HSC-primed haemogenic endothelial cells at Carnegie stage 12-14 with an unambiguous arterial endothelial signature and upregulation of RUNX1, MYB and ANGPT1; subcategorising CD34+CD45- endothelial cells by CD44 positivity enriched HECs more than 10-fold; EMCN, PROCR and RUNX1T1 were transiently over-represented on hemogenic fate choice; and a second temporally and molecularly distinct intra-embryonic HEC population lacking arterial features was detected mainly at the earlier CS10.
Tracing the first hematopoietic stem cell generation in human embryo by single-cell RNA sequencing · Zeng Y, He J, Bai Z, Li Z, Gong Y, Liu C, Ni Y, Du J, Ma C, Bian L, Lan Y, Liu B · Cell Research 2019What this evidence cannot show
Very small numbers of extremely scarce human embryonic specimens; 'HSC-primed HEC' is a transcriptomic definition and the functional repopulating capacity of the sorted CD44+ fraction is not established at single-cell resolution. Carnegie staging of human embryonic material carries real assignment uncertainty. Marker enrichment figures are cohort-specific and may not transfer to other staging or dissociation protocols.
Evidence that argues against this (2)
- The operational definition of a haemogenic endothelial cell is not settled and the field says so in print: this study opens by stating that study of haemogenic specification 'is hampered by a lack of consensus in the timing of hemogenic endothelium analysis and the full hematopoietic potential of this population', and then shows that the haemogenic potential of sorted endothelial populations declines sharply across the course of differentiation, so the same sort gate returns a different cell depending on when it is applied. A CS12-14 arterial HEC signature is a snapshot of one window, not a stable cell identity. Read: abstract only. Garcia-Alegria E et al. 2018
- A transcriptional state does not establish a cell of origin. Sequencing two clonally related leukaemias from one child (infant KMT2A-r B-ALL relapsing as AML) placed the KMT2A rearrangement before haematopoietic specification, and the authors conclude explicitly that 'cell of origin cannot be inferred from the transcriptional state'. That cuts both ways here: it weakens transcriptome-based arguments for a haemogenic-endothelial origin as much as it weakens the pro-B and LMPP arguments. Read: abstract only. Khabirova E et al. 2022
- E4strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
Haemogenic endothelial progenitors capable of generating definitive haematopoietic cells can be obtained from human pluripotent stem cells and are identified precisely by a VE-cadherin+CD73-CD235a/CD43- phenotype, discriminating them from VE-cadherin+CD73+ non-haemogenic endothelium and from VE-cadherin+CD235a+CD41a- early haematopoietic cells; they arise directly from VE-cadherin-negative KDR-bright APLNR+PDGFRA-low/negative hematovascular mesodermal precursors, whereas hemangioblasts originate from more immature APLNR+PDGFRA+ mesoderm.
Identification of the hemogenic endothelial progenitor and its direct precursor in human pluripotent stem cell differentiation cultures · Choi KD, Vodyanik MA, Togarrati PP, Suknuntha K, Kumar A, Samarjeet F, Probasco MD, Tian S, Stewart R, Thomson JA, Slukvin II · Cell Reports 2012What this evidence cannot show
An in vitro hPSC differentiation system. Marker boundaries defined in culture do not automatically map onto the human embryo, and hPSC-derived 'definitive' haematopoietic output is well known to fall short of genuine long-term repopulating HSCs. CD73 in particular is a culture-derived discriminator with no established in vivo counterpart for this purpose.
Evidence that argues against this (1)
- The VE-cadherin+CD73-CD235a/CD43- gate is a functional enrichment, not a pure population, and the same hESC system gives a different answer depending on when the sort is done: haemogenic potential falls sharply across differentiation, and both primitive and definitive haematopoietic progenitors emerge from the earliest-specified haemogenic endothelium rather than from separate later populations. Equating this sorted fraction with the in vivo human embryonic HE is therefore an assumption. Read: abstract only. Garcia-Alegria E et al. 2018
- E5strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
CD43 (leukosialin/SPN) is detected on all emerging clonogenic haematopoietic progenitors before CD45 expression and reliably separates the haematopoietic CD34+ population from CD34+CD43-CD31+KDR+ endothelial cells and CD34+CD43-CD31-KDR- mesenchymal cells; acquisition of CD45 by CD34+CD43+CD45-Lin- cells accompanies progressive myeloid commitment and loss of B-lymphoid potential.
Leukosialin (CD43) defines hematopoietic progenitors in human embryonic stem cell differentiation cultures · Vodyanik MA, Thomson JA, Slukvin II · Blood 2006What this evidence cannot show
hESC/OP9 co-culture, not embryonic tissue. CD43 is not part of routine clinical haematopoietic flow panels, so this marker logic is a research tool. The 'before CD45' ordering is inferred from population snapshots rather than single-cell lineage tracing.
Evidence that argues against this (1)
- CD43 as a separator was established in human embryonic stem cell differentiation cultures, which model the transition rather than observe it in an embryo. Marker-based separation in a directed differentiation is calibrated to that culture, and "all emerging clonogenic progenitors" is bounded by the clonogenic assay used. Oncofusion knowledge graph (internal analysis) 2026
- E6strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
Human HSCs emerge first in the aorta-gonad-mesonephros region, specifically in the dorsal aorta, and only later appear in the yolk sac, liver and placenta; AGM-region cells provide long-term high-level multilineage repopulation in immunodeficient mice, and a single AGM-derived HSC generates at least 300 daughter HSCs in a primary recipient which disseminate throughout the recipient marrow and are retransplantable.
Highly potent human hematopoietic stem cells first emerge in the intraembryonic aorta-gonad-mesonephros region · Ivanovs A, Rybtsov S, Welch L, Anderson RA, Turner ML, Medvinsky A · Journal of Experimental Medicine 2011What this evidence cannot show
Extremely scarce human embryonic material, so n is very small and the temporal ordering of sites rests on few specimens. Xenotransplantation into immunodeficient mice under-reports human HSC activity for some populations, so 'only later appear' could reflect assay sensitivity at other sites rather than true absence. This paper concerns HSCs, not HE itself; HE is inferred as the upstream source.
Evidence that argues against this (1)
- The AGM-first ordering is about definitive HSCs with long-term repopulating activity; it does not capture the whole output of haemogenic endothelium. Fate mapping shows an extra-aortic HE wave in the vitelline and umbilical arteries at E8.5-E9.5 producing fetal-restricted lympho-myeloid progenitors that never register as HSCs in a repopulation assay. For an infant leukaemia that arises and presents within months of birth, a fetal-restricted progenitor wave is at least as plausible a substrate as a long-term HSC, and the AGM-centred account does not address it. Read: abstract only. Barone C et al. 2026
- E7strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
The human placenta contains haematopoietic progenitors and HSCs throughout development including at term; placenta-derived stromal lines are pericyte-like and support human haematopoiesis, and placental haematopoietic cells localise in close contact with pericytes/perivascular cells.
Human placenta is a potent hematopoietic niche containing hematopoietic stem and progenitor cells throughout development · Robin C, Bollerot K, Mendes S, Haak E, Crisan M, Cerisoli F, Lauw I, Kaimakis P, Jorna R, Vermeulen M, Kayser M, van der Linden R, Imanirad P, Verstegen M, Nawaz-Yousaf H, Papazian N, Steegers E, Cupedo T, Dzierzak E · Cell Stem Cell 2009What this evidence cannot show
Demonstrates presence and niche support, not de novo generation: it cannot distinguish placental HE-derived HSCs from HSCs that emerged in the AGM and colonised the placenta. Maternal-fetal cell admixture is a real confounder in term placental tissue.
Evidence that argues against this (1)
- Placental haematopoiesis is a site claim and the supporting evidence is largely associative — progenitors present, stromal lines supportive, cells in close contact with vessels. Presence at a site does not establish generation at that site, since circulating progenitors lodge in vascular niches throughout development. Oncofusion knowledge graph (internal analysis) 2026
- E8moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper
MLL-AF4 expressed in human embryonic stem cells enhances the specification of hemogenic precursors but strongly impairs further hematopoietic commitment in favour of an endothelial cell fate, primes cells transcriptionally toward hemogenic precursors prone to endothelial maturation, and is not sufficient to transform hESC-derived hematopoietic cells.
A human ESC model for MLL-AF4 leukemic fusion gene reveals an impaired early hematopoietic-endothelial specification · Bueno C, Montes R, Melen GJ, Ramos-Mejia V, Real PJ, Ayllón V, Sanchez L, Ligero G, Gutierrez-Aranda I, Fernández AF, Fraga MF, Moreno-Gimeno I, Burks D, Plaza-Calonge MdelC, Rodríguez-Manzaneque JC, Menendez P · Cell Research 2012What this evidence cannot show
Enforced constitutive fusion expression in an in vitro hESC differentiation system with no niche and no immune context; the 'hemogenic precursor' here is a culture-defined population, not a verified in vivo haemogenic endothelial cell. No leukemia results, and the authors explicitly report the fusion is not transforming in this system — which argues the true cell of origin lies downstream of this stage.
Evidence that argues against this (2)
- The same laboratory's own in vivo work undercuts the inference from this hESC result. Enforced MLL-AF4 in human cord-blood CD34+ HSPCs enhanced repopulation, homing and clonogenicity after intra-bone-marrow transplantation but was 'insufficient to initiate leukemogenesis on its own'; the authors themselves raised that either further hits are required or the cells tested were not the appropriate target. A skew in specification in a dish is not evidence that transformation happens in that cell. Read: abstract only. Montes R et al. 2011
- Cell-of-origin experiments of this shape routinely fail to discriminate. Prospectively isolated HSCs and committed granulocyte/macrophage progenitors transduced with the same MLL fusion were both efficiently immortalised and produced AML with similar latency and — crucially — an indistinguishable immunophenotype and expression profile, arrested at the same late myelomonocytic stage. The starting population did not determine the disease. Read: abstract only. Cozzio A et al. 2003
- E9moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper
Co-expression of MLL-AF4 and the reciprocal AF4-MLL in human embryonic differentiation strongly promotes emergence of hemato-endothelial precursors, both endothelial- and hemogenic-primed, which specify into significantly higher numbers of both hematopoietic and endothelial committed cells without altered survival or proliferation, with H3K79me3 enrichment specifically over HOXA cluster genes.
Enhanced hemato-endothelial specification during human embryonic differentiation through developmental cooperation between MLL-AF4 and AF4-MLL fusions · Bueno C, Calero-Nieto FJ, Wang X, Valdés-Mas R, Gutiérrez-Agüera F, Roca-Ho H, Ayllon V, Real PJ, Arambilet D, Espinosa L, Torres-Ruiz R, Agraz-Doblas A, Varela I, de Boer J, Bigas A, Gottgens B, Marschalek R, Menendez P · Haematologica 2019What this evidence cannot show
Same hESC-model limitations as E8. The reciprocal AF4-MLL fusion is expressed in only about half of t(4;11) patients, so the cooperativity applies at best to a subset. Precursor emergence and lineage output are the readouts; no transformation or leukemia is demonstrated.
Evidence that argues against this (2)
- Conditional Mll-AF4 driven by B-cell or T-cell promoter Cre still produced exclusively B-lineage neoplasia, leading the authors to conclude that 'the MLL-AF4 fusion protein does not have a mandatory role in multi-potent haematopoietic stem cells to cause cancer'. If the fusion does not require a multipotent — let alone a pre-haematopoietic — cell, then enhanced hemato-endothelial specification in a hESC system is a phenotype of the system, not a requirement of the disease. Read: abstract only. Metzler M et al. 2006
- Enhanced specification is not transformation, and no KMT2A-fusion model reproduces the disease from this route: this paper states plainly that KMT2A fusions 'fail to phenocopy short latency leukaemogenesis in mammalian models, suggesting additional factors are necessary for transformation'. It also shows that losing a different KMT2 family member (KMT2C) blocks hemogenic endothelium specification outright in hPSCs, so hemato-endothelial output in these systems is sensitive to background genotype as well as to the fusion. Read: abstract only. Maurya S et al. 2022
- E10strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
Haemogenic competence is endowed upstream of haemogenic endothelium: transient expression of the T-box transcription factor Eomesodermin governs haemogenic competency of yolk-sac extra-embryonic mesodermal progenitors by regulating accessibility of enhancers used by SCL/TAL1, and Eomes is essential for normal development of Runx1-positive haemogenic endothelium, with Eomes loss blocking blood-progenitor formation but not specification of Flk-1-positive haematoendothelial progenitors.
The T-box transcription factor Eomesodermin governs haemogenic competence of yolk sac mesodermal progenitors · Harland LTG, Simon CS, Senft AD, Costello I, Greder L, Imaz-Rosshandler I, Göttgens B, Marioni JC, Bikoff EK, Porcher C, de Bruijn MFTR, Robertson EJ · Nature Cell Biology 2021What this evidence cannot show
Mouse embryonic stem cell model plus mouse knockouts, focused on yolk-sac/extra-embryonic mesoderm and primitive erythropoiesis; extrapolation to the intra-embryonic AGM haemogenic endothelium that produces definitive HSCs is not direct. Eomes has pleiotropic roles in mesoderm and endoderm specification, which complicates attribution.
Evidence that argues against this (1)
- Transient Eomes expression governing haemogenic competency is a mouse yolk-sac result about extra-embryonic mesoderm, two developmental steps upstream of anything this beat can act on. It also pushes the relevant window earlier than the first trimester origin the beat elsewhere favours, which is a tension the entry does not resolve. Oncofusion knowledge graph (internal analysis) 2026
- E11established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database
The endothelial-to-haematopoietic transition is a defined biological process in the Gene Ontology (GO:0098508): the generation of haematopoietic stem cells from haemogenic endothelial cells by a process that includes tight-junction dissolution and loss of cell polarity followed by delamination from the endothelium; the Cell Ontology contains no class for haemogenic endothelium, the nearest terms being CL:0002418 hemangioblast, CL:0002544 aortic endothelial cell and CL:0000071 blood vessel endothelial cell.
Gene Ontology and Cell Ontology term records retrieved via the EBI Ontology Lookup Service (OLS4) API · Gene Ontology Consortium; Cell Ontology; EMBL-EBI Ontology Lookup Service · Databases 2026What this evidence cannot show
A negative result from an ontology search: it shows no CL class matched 'hemogenic', 'haemogenic', 'hemogenic endothelial cell' or 'hemogenic endothelium' in the OLS4 index at the time of retrieval (including a wildcard query and a full scan of the 332 CL classes matching 'endothelial'). OLS indexing lags upstream ontology releases, so a term could exist in a newer CL release not yet indexed, or under an unexpected label. Re-check before publishing the claim that no CL term exists.
Evidence that argues against this (1)
- A Gene Ontology biological-process term is a curated definition assembled from published descriptions. GO:0098508 records that the community recognises the transition and names its features; it is not independent evidence that the process occurs as described in human, and GO definitions are revised between releases. OBO Foundry / EMBL-EBI OLS 2026
competing candidate cells of origin for prenatally acquired KMT2A::AFF1; no published experiment discriminates between them, and transduction-and-transplant assays cannot, because the same MLL fusion in stem and in committed progenitor populations yields indistinguishable disease
- ← arises inKMT2A::AFF1 (MLL-AF4), t(4;11)(q21;q23)
possible site of the translocation event rather than of transformation; MLL-AF4 in hESC skews hemato-endothelial specification but does not transform
upstream via the endothelial-to-haematopoietic transition, before fetal liver colonisation
- ← related toIs KMT2A::AFF1 leukemia epithelial or endothelial? No — it is haematopoietic. But the question is a good one
why the endothelial question is worth asking at all
jund is required for the endothelial-to-haematopoietic transition in zebrafish; without it the haemogenic endothelial cell retains an enhanced endothelial identity. Developmental cell type, zebrafish, no postnatal counterpart
REQUIRED IN: Aml1-null mouse embryos lack definitive haematopoiesis entirely while primitive yolk-sac erythropoiesis is normal, and the defect is cell-intrinsic by chimera analysis - the requirement sits at the endothelial-to-haematopoietic transition (murine germline knockout; PMID 8565077)