Oncofusion

start / graph / cell-state-hspc-like-blast-fraction

Cell statestrong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.Q139-single-cell

The HSPC-like blast fraction — a minority of blasts that transcriptionally resemble haematopoietic stem and progenitor cells, found in the blood of the youngest KMT2A-rearranged patients and predicting drug response and outcome across leukaemia subtypes

also called HSPC-like blasts · stem-cell-like blasts · HSPC-like population · BMP-like blasts (the T-ALL counterpart) · stem-like subpopulation

Inside one patient's leukaemia, not every cancer cell is the same: a small group of them reads, gene for gene, like a blood stem cell rather than like the B-cell the rest of the leukaemia is pretending to be. In KMT2A-rearranged leukaemia this group is found in the circulating blood of the youngest babies, and across many kinds of leukaemia the size of this group predicts how well chemotherapy works and how long the patient lives.

The same thing, for a scientist

Single-cell multiomic profiling (scRNA-seq, scATAC-seq and snmC-seq2) of 25 KMT2A-rearranged ALL patients — 18 diagnosed under one year, 7 over one year — plus 5 healthy paediatric donors identified **a haematopoietic stem and progenitor-like (HSPC-like) population in the blood of the younger patients that contains leukaemic blasts** and that forms an immunosuppressive signalling circuit with cytotoxic lymphocytes [E1]. The same group then generalised it: in a pan-leukaemia single-cell cohort, **HSPC-like blast signatures predict drug response and clinical outcome** across subtypes [E2], and the T-ALL counterpart — a bone-marrow-progenitor (BMP)-like subpopulation in a 40-case single-cell multiomic atlas — is **associated with treatment failure and poor overall survival**, predicts poor outcome across T-ALL subtypes, and is selectively vulnerable to apoptosis-inducing agents including venetoclax [E3]. A 2026 review consolidates the state across lymphoblastic and mixed-phenotype leukaemias and is explicit about what is *not* established: earlier primary-sample studies found immunophenotypically and transcriptomically HSPC-like blasts but **failed to consistently demonstrate hierarchical tumour-initiating potential in xenograft models** [E4]. So the state is real, reproducibly detectable and prognostic, and it is **not** established as a leukaemic stem cell in the functional sense. This is a cell-state entry, not a cell-type entry, for that reason. **Single-cell resolution is not a convenience here — it is the only way the fraction can be seen at all**: in bulk transcriptomes the HSPC-like cells are a minority averaged into a dominant B-lymphoblast signal, and the prognostic variable is the *proportion*, which bulk data cannot express.

Go deeper
What was measured, in which patients, and what the age axis does1 deeper

Chen et al. subjected KMT2A-r leukaemias and normal haematopoietic cells from patients of different ages to single-cell multiomics, explicitly to find age-specific differences 1. Three findings sit together and the middle one is the mechanism:

  1. Leukaemia cells from patients under 6 months have significantly increased lineage plasticity.
  2. Steroid-response pathways are downregulated in the most immature blasts from younger patients.
  3. An HSPC-like population in the blood of younger patients contains leukaemic blasts and forms an immunosuppressive circuit with cytotoxic lymphocytes.

Read together these say the youngest patients carry more of a cell that is simultaneously less differentiated, less steroid-responsive and actively suppressive of the immune effectors that immunotherapy depends on — which the authors offer as 'a compelling explanation for the ability of leukaemias in young patients to evade chemotherapy and immune-mediated control' 1. That maps directly onto the clinical fact this project already records, that younger age at diagnosis is associated with increased risk of relapse in infant KMT2A-r ALL.

The location matters and is easy to skim past: the HSPC-like population was found in the blood. That is a peripheral-blood compartment claim, not a marrow claim, and it is the compartment a diagnostic sample comes from most easily.

Why this is a cell STATE and not a cell TYPE

The 2026 review is the careful source here. HSPC-like blasts have been identified by immunophenotype and by transcriptome in ALL primary samples repeatedly, and those studies failed to consistently demonstrate hierarchical tumour-initiating potential in xenograft models 4. A population that looks like a stem cell and does not behave like one in the one assay that tests stemness is a state, and the graph should not promote it.

That is also the honest answer to the obvious next question — is this the leukaemic stem cell? — and the answer is that the field has tried and not shown it. The prognostic association is nonetheless robust and replicated in three independent cohorts and two lineages 123.

The regulatory circuitry, and where the fusion's own biology enters

The pan-leukaemia study reports the HSPC-like signature under a HOXA/AP-1/CEBPA regulatory network 2. HOXA is this project's central axis: the KMT2A fusion's canonical output is HOXA9/MEIS1, and the HOXA-high subgroup of infant KMT2A::AFF1 ALL is already a documented stratifier. If the HSPC-like fraction is HOXA-driven, then the fraction and the fusion's direct programme are the same object measured two ways, and a menin inhibitor — which collapses HOXA — should shrink the fraction. That is a prediction, it is testable on deposited data, and it is written up as a hypothesis in this beat rather than asserted here.

A counterweight worth stating in the same breath: this corpus's own reanalysis found the menin inhibitor response in two KMT2A::AFF1 lines did not touch HOXA or MEIS1 (ana-the-menin-inhibitor-response-is-nine-fold-fusion-selective-and-does-not-touch-hoxa-or-meis1-in-either-kmt2a-aff1-line). So the HOXA link cannot be assumed from the AML literature into the B-lineage index fusion, and the prediction above may fail for exactly that reason.

Epigenetics within the state

The 25-patient resource includes scATAC-seq and snmC-seq2 (single-nucleus methylcytosine) alongside the transcriptome 1, which means the HSPC-like fraction has a matched accessibility and DNA-methylation description and not only an expression one. That is unusual and it is the reason this dataset is worth reanalysing rather than only citing: a question of the form is the HSPC-like fraction's chromatin already open at myeloid enhancers before any myeloid transcript appears is answerable in the deposited data.

Separately, the T-ALL atlas showed NOTCH1 mutations additively drive blasts away from the BMP-like state 3 — i.e. a cooperating lesion changes the position on the stemness axis rather than acting orthogonally to it. The KMT2A-r analogue would be the RAS-pathway co-mutations, which this corpus already links to the refractory immature subgroup of infant ALL.

The analogous biology in other fusion families and lineages
systemthe stem-like fractionoutcome link
KMT2A-r ALL, 25 patientsHSPC-like, in blood, younger patients 1plasticity, steroid-response loss, immunosuppressive circuit
pan-leukaemia single cellHSPC-like signature 2predicts drug response and clinical outcome
T-ALL, 40 casesBMP-like 3treatment failure, poor OS; venetoclax-vulnerable
paediatric AML, 262 patients (RNA deconvolved with adult single-cell signatures)LSPC-Cycle (>25% cycling stem-like)worst prognosis, OS odds ratio 11.33 vs GMP-like 5

The paediatric AML row is a different instrument — bulk RNA deconvolved against single-cell reference signatures, not single-cell data — and it is included because it is the largest cohort and because its hierarchy assignment is fusion-dependent: CBFA2T3::GLIS2 and MYB::GATA1 concentrate in LSPC-Cycle, DEK::NUP214 and NUP98::KDM5A in the primitive group, while KMT2A::MLLT3 hierarchy differed with MECOM expression level rather than sitting in one class 5. That last clause is the interesting one for this project: the KMT2A fusion does not fix a position on the stemness axis the way the other fusions do.

Facts and figures
locationperipheral blood of KMT2A-r ALL patients diagnosed at younger ages; also bone marrow
shown for fusionKMT2A-rearranged ALL as a class (25 patients, mixed partners, 18 infants); NOT resolved to KMT2A::AFF1 specifically. The T-ALL and pan-leukaemia generalisations are not KMT2A-specific at all.
n patients25
age axisunder 6 months vs older: increased lineage plasticity in the younger group
functional stemness demonstratedfalse
markers
[
  "HSPC-like transcriptional signature (scRNA-seq derived)",
  "CD34 (as the classical HSPC surface correlate; not the defining feature in the single-cell studies)",
  "HOXA / AP-1 / CEBPA regulatory network activity",
  "loss of glucocorticoid-response pathway expression in the most immature blasts"
]
atlas refs
[
  "Human Tumor Atlas Network (HTAN) — 25 KMT2A-r ALL patients, scRNA + scATAC + snmC-seq2 (see dataset-htan-kmt2a-r-single-cell-multiome)",
  "https://github.com/tanlabcode/KMT2Ar_Paper"
]
How this could bite us back
  • If the HSPC-like fraction is where treatment failure lives, then a therapy that works on the bulk blast population and not on this fraction will produce exactly the clinical picture KMT2A-r infant ALL shows — rapid deep response followed by early relapse. Response-rate endpoints in early-phase trials would be blind to this, and MRD by flow would be partly blind to it too, because an HSPC-like blast does not carry the B-lymphoblast immunophenotype the MRD panel is gated on.
  • The immunosuppressive circuit with cytotoxic lymphocytes is a direct threat to every T-cell-engaging therapy in this disease — blinatumomab and CD19 CAR-T both need functional cytotoxic effectors, and the youngest patients, who have the most of this population, are the ones with the worst outcomes.
  • Targeting stemness in a cell that has not been shown to be functionally a stem cell risks paying the toxicity of an HSPC-directed agent for none of the benefit. Normal HSPCs share the signature by construction; that is what 'HSPC-like' means.
Still open
  1. Is the HSPC-like fraction in KMT2A::AFF1 specifically, or is it carried by the KMT2A::MLLT3 and other myeloid-leaning partners in the 25-patient cohort? The deposited HTAN data is partner-annotated and this is a reanalysis, not a new experiment.
  2. Does a menin inhibitor shrink the HSPC-like fraction, or enrich for it? Both are plausible: if the fraction is HOXA-driven it should shrink; if it is the least fusion-dependent compartment it should be the one that persists.
  3. The HSPC-like population was found in BLOOD. Is it a marrow population that has mobilised, or does it arise in the circulation? A paired marrow/blood single-cell comparison in the same patient answers it and the samples exist.
  4. Do the HSPC-like cells carry the fusion transcript? The paper says the population 'contains leukaemic blasts', which is not the same as every cell in it being leukaemic — and the immunosuppressive-circuit claim depends on which cells are which.
  5. Is the 3-fold higher venetoclax vulnerability of the T-ALL BMP-like state reproduced in the KMT2A-r HSPC-like fraction? This corpus already has cell-state-apoptotic-priming-bcl2 and drug-venetoclax; the join has not been made at single-cell resolution.
Every source behind this entry (5)
  1. E1strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    Single-cell multiomics of 25 KMT2A-rearranged ALL patients identified an HSPC-like population in the blood of younger patients that contains leukaemic blasts and forms an immunosuppressive circuit with cytotoxic lymphocytes; patients under 6 months showed significantly increased lineage plasticity and steroid-response pathways were downregulated in their most immature blasts.

    PMID 34864916DOI
    What this evidence cannot show

    25 KMT2A-r patients (18 infants <1 y, 7 older) and 5 healthy paediatric donors, with 2 patients contributing paired longitudinal lineage-switch samples — so the age comparison rests on 18 vs 7 and the <6-month plasticity claim on a subset of the 18. Partner gene is not uniformly KMT2A::AFF1; this is KMT2A-rearranged ALL, and the entry should not be read as an index-fusion-specific result. The immunosuppressive circuit is inferred from ligand-receptor expression in single-cell data, not from a functional suppression assay. Nothing in the paper is under menin inhibition.

    Evidence that argues against this (1)
    • Earlier primary-sample work identified immunophenotypically and transcriptomically HSPC-like blasts in ALL but failed to consistently demonstrate hierarchical tumour-initiating potential in xenograft models — i.e. the functional stemness that the name implies has not been shown. The 2026 review states this directly. Xu J et al. 2026
  2. E2strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    Single-cell-derived HSPC-like blast signatures predict drug response and clinical outcome across leukaemia subtypes.

    Single-cell panleukemia signatures of HSPC-like blasts predict drug response and clinical outcome · Chen C, Xu J, Sussman JH, et al. (senior author Tan K) · Blood 2025
    PMID 40089994DOI
    What this evidence cannot show

    Pan-leukaemia rather than KMT2A-specific, and from the same laboratory as E1, so it is a generalisation by the group that made the original observation rather than an independent replication. This corpus already cites it once, in cell-state-relapse-initiating-persister.

    Evidence that argues against this (1)
    • A signature that predicts outcome across subtypes may be predicting proliferative rate or blast immaturity, both of which are established prognostic variables that need no new biology. The pan-leukaemia design makes this MORE likely rather than less, because a signature that generalises across diseases with different drivers is by construction capturing something driver-independent. The paper's own multivariable handling of blast count and WBC is the thing to read before this is treated as a distinct axis. Chen C et al. 2025
  3. E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper

    In a 40-case single-cell multiomic T-ALL atlas, a bone-marrow-progenitor-like subpopulation was associated with treatment failure and poor overall survival, predicted poor outcome across T-ALL subtypes, was pushed away from by NOTCH1 mutations, and was vulnerable to apoptosis-inducing agents including venetoclax.

    PMID 39587259DOI
    What this evidence cannot show

    T-ALL, not KMT2A-rearranged B-ALL — included as the cross-lineage analogy the PI's analogy axis asks for, and overlapping in authorship with E1/E2. The venetoclax vulnerability came from in silico and in vitro drug screening, not a clinical arm.

    Evidence that argues against this (1)
    • The T-ALL result cuts against a simple stemness-is-bad model in one specific way: NOTCH1 mutation, the canonical T-ALL driver and a marker of BETTER outcome in several paediatric cohorts, drives blasts AWAY from the BMP-like state. So 'less stem-like' and 'better outcome' are confounded with 'NOTCH1-mutant' in that dataset, and the independent prognostic value of the BMP-like fraction needs the NOTCH1-wild-type stratum read separately. Xu J et al. 2025
  4. E4moderate confidence 3 of 5: One strong study, or several smaller consistent ones.review

    Across ALL and mixed-phenotype leukaemias, recent multi-omic profiling has improved the phenotypic definition of HSPC-like blasts and their association with treatment failure, relapse and lineage switch during therapy, while consistent hierarchical tumour-initiating potential in xenografts has not been demonstrated.

    HSPC-like blasts in acute lymphoblastic leukemia: biology and therapeutic opportunities · Xu J, Kennedy VA, Ma C, et al. (senior author Tan K) · Trends in Cancer 2026
    PMID 41436318DOI
    What this evidence cannot show

    A review, included as a map and for its explicit statement of the negative, not as primary evidence. Written by the same group as E1-E3, which makes its self-criticism more credible but does not make it independent.

    Evidence that argues against this (1)
    • This review is written by the group that produced E1, E2 and E3, so its framing of what HSPC-like blasts are and are not is not an independent assessment of its own work. Its value here is that it states the xenograft negative plainly, which is the direction an interested party would be least inclined to overstate - but a genuinely independent review of the same question does not currently exist. Xu J et al. 2026
  5. E5moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper

    In 262 paediatric AML patients, bulk RNA deconvolved against adult single-cell signatures defined five hierarchy subtypes; LSPC-Cycle carried the worst prognosis (OS odds ratio 11.33 vs GMP-like) and was an independent prognostic factor, and hierarchy assignment tracked fusion identity — but KMT2A::MLLT3 hierarchy differed with MECOM expression level rather than occupying one class.

    Identification of cellular hierarchy in paediatric acute myeloid leukaemia: The Japan Children's Cancer Group trial (JCCG AML-12) · Komori T, Kawai T, Okuno Y, et al. (senior author Shiba N) · British Journal of Haematology 2026
    PMID 41631723DOI
    What this evidence cannot show

    NOT single-cell data. Bulk RNA-seq deconvolved with adult single-cell reference signatures by Zeng's method, so every hierarchy call inherits whatever the adult reference cannot represent about paediatric and infant haematopoiesis — which is precisely the axis this beat is about. AML, not ALL. Included for the fusion-versus-hierarchy table and because 262 patients is by far the largest cohort available on this question.

    Evidence that argues against this (1)
    • Deconvolving bulk paediatric AML against ADULT single-cell signatures is the weakest link in this row and cuts directly against using it for infant biology: any paediatric or infant cell state with no adult counterpart cannot be assigned, and will be forced into the nearest adult class. The paper's own finding that two principal-component axes separate paediatric from adult AML is evidence that the reference is mismatched, reported by the authors as a result rather than as a limitation. Komori T et al. 2026
Connected to