The early lymphocyte precursor (ELP) state — what KMT2A-rearranged infant B-ALL scores as against a developing bone marrow reference, and why it is earlier than the pro-B arrest this graph also records
also called ELP state · early lymphocyte precursor · ELP-dominated transcriptional state · hybrid myeloid-lymphoid precursor state
When you compare single leukaemia cells against a map of normal developing bone marrow, KMT2A-rearranged infant leukaemia lands at a very early lymphoid point — earlier than where its surface markers suggest — and carries myeloid features at the same time. Those non-physiological marker combinations are exactly what a targeted therapy needs to tell cancer from healthy tissue.
The same thing, for a scientist
Khabirova et al. asked where KMT2A-rearranged infant B-ALL sits developmentally by projecting single lymphoblast transcriptomes onto a developing human bone marrow reference, supported by bulk mRNA meta-analysis. The answer: the disease is **uniquely dominated by an early lymphocyte precursor (ELP) state**, whereas the less adverse NUTM1-rearranged infant ALL shows signals of later developing B cells, as do most other childhood B-ALLs [E1]. Comparing the lymphoblasts with normal ELP cells revealed **hybrid myeloid-lymphoid features, including non-physiological antigen combinations**, whose surface co-expression was validated by flow cytometry and which the authors propose as a route to cancer-specific targeting [E1]. The same paper contains a result that constrains this entire class of inference: from shared mutations in two separate leukaemias in one child — infant KMT2A-rearranged B-ALL relapsing as AML — they established that the rearrangement **occurred in very early development, before haematopoietic specification**, and concluded that **cell of origin cannot be inferred from transcriptional state** [E1]. So the ELP state is a description of where the blast sits now, explicitly not of where it came from. That distinction is the entry's main contribution, because a graph that holds both a cell-of-origin literature and a developmental-state literature will otherwise merge them.
ELP versus pro-B arrest: two entries, one cell, and the reason both are right
This graph already holds cell-state-pro-b-arrest, defined by a CD10-negative CD19-positive CD34-positive immunophenotype with NG2 expression, an immature non-expanded immunoglobulin repertoire, and IL7R/KIT co-expression. It also holds cell-type-fetal-preprob-cd10-negative as a candidate normal counterpart.
The ELP claim is earlier than that, and it is made on a different instrument. Immunophenotype places a cell by the proteins on its surface; reference projection places it by the whole transcriptome against a normal developmental atlas. The two can disagree, and here they do: a cell can display pro-B surface markers while its transcriptome scores closest to a precursor upstream of the pro-B stage.
The reconciliation the paper itself offers is the hybrid finding: the blast carries myeloid and lymphoid features simultaneously, in combinations that do not occur physiologically 1. A cell with a lymphoid surface phenotype and a partly myeloid transcriptome will project early, because the normal reference has no cell like it and the nearest match is the last point before the lineages separate.
So: pro-B by immunophenotype, ELP by transcriptome, neither by ontogeny. All three statements are in the literature and they are about different measurements.
The sentence that should govern every cell-of-origin claim in this graph
"...we established that KMT2A rearrangement occurred in very early development, before hematopoietic specification, emphasizing that cell of origin cannot be inferred from the transcriptional state." 1
The evidence is a natural experiment: one child with infant KMT2A-rearranged B-ALL who relapsed as AML, with shared mutations across the two separate leukaemias placing the common ancestor before lineage specification.
This graph holds a substantial cell-of-origin literature — fetal liver HSC, LMPP, haemogenic endothelium, CD10-negative fetal pre-pro-B — much of it argued from transcriptional similarity between blasts and a candidate normal cell. That inference is exactly what this sentence forbids. It does not make the candidates wrong; it means transcriptional resemblance is not evidence for them, and the arguments that survive are the ones built on genetics, on timing, or on modelling the translocation in a defined cell.
It also cuts the other way, usefully: if the lesion predates haematopoietic specification, then the lineage plasticity this disease shows under therapy — lineage switch under blinatumomab or CAR-T, and the B-to-AML relapse in this very patient — is not a surprising acquisition but a return to the range of a cell that was never committed.
Non-physiological antigen combinations as the therapeutic payload
The practically important part of E1 is the smallest: the hybrid myeloid-lymphoid programme produces antigen combinations that do not occur on any normal cell, and exemplar combinations were confirmed by flow cytometry 1.
That is a specificity argument of a kind this graph rarely gets. Single-antigen targeting in this disease has a known failure mode — risk-lineage-switch-cd19-directed-therapy records CD19-directed therapy driving escape into the myeloid compartment — and a combination requirement (antigen A and antigen B on the same cell) is resistant to that escape in a way single-antigen targeting is not, because the escape route requires losing both.
The caveats are real: the paper validates co-expression, not that a logic-gated therapy against such a pair is feasible, and dual-antigen CAR and bispecific engager platforms have their own toxicity and manufacturing constraints. The abstract names no specific pair, so this entry deliberately does not name one.
| state definition | The position a leukaemic blast occupies when its whole transcriptome is projected onto a normal developing bone marrow reference; for KMT2A-rearranged infant B-ALL this is an early lymphocyte precursor position, upstream of the pro-B stage its surface immunophenotype suggests |
|---|---|
| method | single lymphoblast transcriptomes projected onto a developing human bone marrow reference, plus bulk mRNA meta-analysis; flow cytometry validation of exemplar surface antigen pairs |
| contrast group | NUTM1-rearranged infant ALL, which is less adverse and shows later developing B-cell signals, as do most other childhood B-ALLs |
| hybrid feature | simultaneous myeloid and lymphoid programme with non-physiological antigen combinations; proposed as a route to cancer-specific targeting |
| ontogeny constraint | shared mutations across a B-ALL and its AML relapse in one child place the rearrangement before haematopoietic specification; the authors state cell of origin cannot be inferred from transcriptional state |
relation to other entries
{
"cell-state-pro-b-arrest": "immunophenotypic placement (CD10-negative CD19+ CD34+), later than ELP; not contradicted, measured differently",
"cell-type-fetal-preprob-cd10-negative": "a candidate normal counterpart argued from transcriptional resemblance, which is the inference E1 constrains",
"mechanism-lineage-switch-kmt2a": "a pre-specification lesion makes lineage switch a return to range rather than an acquisition"
}five integrative clusters
[ "IC1 IRX-type differentiated", "IC2 IRX-type undifferentiated (100% RAS pathway mutated, most refractory)", "IC3 HOXA-type MLLT1", "IC4 HOXA-type MLLT3", "IC5 HOXA-type AFF1" ]
atlas refs
[
{
"label": "ELP-dominated developmental state of KMT2A-rearranged infant B-ALL",
"pmid": "35288693",
"title": "Single-cell transcriptomics reveals a distinct developmental state of KMT2A-rearranged infant B-cell acute lymphoblastic leukemia.",
"authors": "Khabirova E, Jardine L, Coorens THH, et al.",
"venue": "Nat Med",
"year": 2022,
"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."
},
{
"label": "Five integrative clusters spanning B-lymphoid and hemato-endothelial stages in 84 infants",
"pmid": "36042201",
"title": "Multi-omics analysis defines highly refractory RAS burdened immature subgroup of infant acute lymphoblastic leukemia.",
"authors": "Isobe T, Takagi M, Sato-Otsubo A, et al.",
"venue": "Nat Commun",
"year": 2022,
"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."
}
]- A developmental state that is earlier and more plastic than the immunophenotype suggests is a warning about every single-antigen immunotherapy in this disease: the cell has more lineages available to it than its surface implies, and lineage switch under CD19-directed therapy is already documented.
- Non-physiological antigen combinations are specific to the blast only relative to whatever normal reference was used; an infant's tissues include fetal and neonatal cell types absent from adult atlases, so the specificity claim needs an age-matched reference before it is a safety claim.
- If cell of origin cannot be inferred from transcriptional state, then several cell-of-origin arguments in this graph rest on an inference the field's own data forbids, and the therapeutic strategies built on targeting the cell of origin inherit that weakness.
- Dual-antigen logic-gated targeting is resistant to single-antigen escape and is also substantially harder to manufacture, dose and monitor in an infant than a single CAR; the specificity gain is not free.
- Is the ELP assignment a property of the blast or of the reference? Projecting the same cells onto a second, independently built developing marrow atlas would say whether 'ELP-dominated' survives a change of reference, and no such cross-atlas test has been published.
- Does the ELP state map onto the five integrative clusters, and if so how? One study says the class is ELP-dominated and another resolves it into five developmental positions tracking fusion partner; both datasets are public and the comparison has not been made.
- Which non-physiological antigen pairs are actually co-expressed at therapeutic density on enough of the blast population, and are any absent from normal fetal and infant tissue rather than just from adult marrow? The specificity argument depends on a reference that includes the tissues an infant actually has.
- If the lesion predates haematopoietic specification, does a non-haematopoietic reservoir exist? This graph holds concept-haematopoietic-not-epithelial-endothelial; the pre-specification timing is the strongest published reason to keep asking.
- E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
Bulk mRNA meta-analysis with single lymphoblast transcriptomes projected onto a developing bone marrow reference showed that KMT2A-rearranged infant B-ALL is uniquely dominated by an early lymphocyte precursor state, whereas less adverse NUTM1-rearranged infant ALL shows signals of later developing B cells as do most other childhood B-ALLs; comparison of infant lymphoblasts with ELP cells revealed hybrid myeloid-lymphoid features including non-physiological antigen combinations whose surface co-expression was validated by flow cytometry; and analysis of shared mutations in two separate leukaemias from one child with infant KMT2A-rearranged B-ALL relapsing as AML established that the rearrangement occurred in very early development before haematopoietic specification, so that cell of origin cannot be inferred from the transcriptional state.
Single-cell transcriptomics reveals a distinct developmental state of KMT2A-rearranged infant B-cell acute lymphoblastic leukemia · Khabirova E, Jardine L, Coorens THH, Webb S, Treger TD, Engelbert J, et al., Behjati S · Nat Med 2022What this evidence cannot show
Reference projection assigns a cell to the nearest normal state in the atlas used; a cell with no normal counterpart - which this paper argues is the case here - is assigned to the closest available point, so 'ELP-dominated' is a statement about the reference as much as about the blast. The pre-specification timing rests on shared mutations in ONE child, an n of 1 natural experiment, however clean. Fusion partners within the KMT2A-rearranged group are not separated. No functional experiment tests whether the ELP-like state is required for disease maintenance. This citation was re-fetched from PubMed for this entry rather than copied from the several existing graph entries that also cite it. Read: abstract only.
Evidence that argues against this (1)
- A single developmental label for this disease is contradicted by multi-omics clustering of a larger cohort. RNA sequencing, methylation arrays, whole-exome and targeted deep sequencing of 84 infants with KMT2A-rearranged leukaemia, with single-sample and single-cell inference of haematopoietic differentiation, established FIVE robust integrative clusters with different master transcription factors, fusion partners and corresponding stages of B-lymphopoietic and early hemato-endothelial development - IRX-type differentiated, IRX-type undifferentiated, HOXA-type MLLT1, HOXA-type MLLT3 and HOXA-type AFF1 - and the most refractory cluster carried RAS pathway mutations in 100% of cases. If the disease occupies five distinguishable developmental positions that track fusion partner, then 'uniquely dominated by an ELP state' is an average across a heterogeneous set rather than a property of the class. Read: abstract only. Isobe T et al. 2022
- E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
Multi-omics clustering of 84 infants with KMT2A-rearranged leukaemia, combining RNA sequencing, methylation arrays, whole-exome and targeted deep sequencing with single-sample and single-cell inference of haematopoietic differentiation, established five robust integrative clusters with different master transcription factors, fusion partners and corresponding stages of B-lymphopoietic and early hemato-endothelial development — IRX-type differentiated, IRX-type undifferentiated, HOXA-type MLLT1, HOXA-type MLLT3 and HOXA-type AFF1 — and deep mutational analysis showed that the number of RAS pathway mutations predicts prognosis, with the most refractory subgroup possessing 100% frequency and the heaviest burden of RAS pathway mutations.
Multi-omics analysis defines highly refractory RAS burdened immature subgroup of infant acute lymphoblastic leukemia · Isobe T, Takagi M, Sato-Otsubo A, Nishimura A, Nagae G, Yamagishi C, et al., Takita J · Nat Commun 2022What this evidence cannot show
Integrative clustering imposes a chosen number of clusters and 'robust' is a statistical property of the clustering, not a demonstration that the five groups are biologically discrete; boundary cases are not reported in the abstract. Differentiation stage is inferred computationally, not measured functionally. The RAS burden result is an association within a single cohort of 84, and RAS mutations are a known late subclonal event whose detection depends on sequencing depth, so 100% in the refractory cluster is partly a statement about assay sensitivity. This citation was re-fetched from PubMed for this entry rather than copied from the several existing graph entries that also cite it. Read: abstract only.
Evidence that argues against this (1)
- A cluster structure that tracks fusion partner does not establish that the partner determines the developmental state, because fusion partner is also strongly associated with lineage, with age at presentation and with which cells were available to be transformed. The same paper that supplies the ELP reading makes the stronger structural point against reading ontogeny off any of this: the rearrangement arises before haematopoietic specification, so the developmental position a blast occupies is a state it has adopted, not a record of its origin, and clusters of adopted states can be generated by downstream selection rather than by the lesion. Read: abstract only. Khabirova E et al. 2022
KMT2A-rearranged infant B-ALL is uniquely dominated by an ELP state against a developing marrow reference, unlike NUTM1-rearranged infant ALL and most childhood B-ALL
the same cell placed by two instruments: pro-B by surface immunophenotype, ELP by whole-transcriptome projection; the hybrid myeloid-lymphoid programme reconciles them
a candidate normal counterpart argued from transcriptional resemblance — the exact inference this paper's pre-specification timing result forbids
- supports →Lineage switch in KMT2A-rearranged leukemia — the same clone changes what kind of blood cell it is
a lesion arising before haematopoietic specification makes lineage switch a return to the cell's original range rather than an acquired capability
- related to →Lineage switch as antigen escape from CD19-directed therapy in KMT2A-rearranged leukemia
non-physiological myeloid-lymphoid antigen pairs are the specificity argument against single-antigen escape, and the escape is already documented for CD19
RAS-pathway mutation burden is heaviest, at 100% frequency, in IC2 - the IRX-type undifferentiated, most refractory cluster of infant KMT2A-rearranged leukaemia