Co-option of lineage plasticity as a hallmark of multipotent acute leukemias

Abstract

Leukemias are classified by hematopoietic lineage and genetic alterations. Mixed-lineage and biphenotypic leukemias have long challenged diagnostic classification because of the coexpression of markers of distinct lineages. Recent studies have revealed previously unappreciated multilineage potential in a subset of B-cell acute lymphoblastic leukemias (B-ALL) as well as leukemias with markers of myeloid and T-cell differentiation with shared genetic features variably classified as acute myeloid leukemia, early T-cell progenitor ALL, or T/myeloid mixed-phenotype acute leukemias. We propose that co-option of stem cell plasticity programs can be used to classify these as multipotent acute leukemias (MAL). Based on the urgent need for improved diagnostic and therapeutic strategies, we review the latest evidence and propose new ways to diagnose and treat MAL as distinct types of acute leukemias.

Introduction

Hematopoiesis is a physiologically regulated process that ensures the continuous production of the diverse cell types required for immunity, hemostasis, oxygen delivery, and other blood functions. Original models considered hematopoiesis as a hierarchy, with self-renewing and multipotent hematopoietic stem cells (HSCs) at the apex, ensuring a lifelong blood supply by differentiation into all mature blood cell types. According to this model, HSCs undergo a stepwise, tightly regulated differentiation process, with HSCs giving rise to multipotent progenitors, which then differentiate into common progenitors to subsequently generate the full range of hematopoietic cells of myeloid, erythroid, and lymphoid lineages.¹ However, recent studies using single-cell transcriptomics (single-cell RNA sequencing [scRNA-seq]), single-cell chromatin accessibility (single-cell assay for transposase-accessible chromatin with sequencing), and lineage tracing under steady state conditions in vivo have challenged this sequential paradigm. Instead, stem cell proliferation and differentiation are now viewed as continuous and dynamic processes in which some progenitors exhibit early fate bias, leading to long-term production of blood cells by lineage-restricted progenitor cells, whereas others maintain multilineage potential even when differentiating into phenotypically identifiable lymphoid and myeloid cells (Figure 1).2, 3, 4, 5, 6, 7, 8

Figure 1.

Schematic model of normal hematopoiesis, illustrating how divergent blood cell lineages share common progenitor cells. Normal hematopoiesis proceeds from hematopoietic stem cells (HSC), proceeding to multipotent progenitor (MPP), multilymphoid progenitor (MLP), and common lymphoid progenitor (CLP) cells, which can give rise to differentiating B-cells, T-cells, and myeloid cells. BCR, B-cell receptor; LMPP, Lymphoid-primed Multipotent Progenitors; MEP, Megakaryocytic-Erythroid Progenitor; MLP, multilymphoid progenitor; MPP, multipotent progenitor; SP, single-positive. The figures were created with BioRender.com. Gutierrez A. and Kentsis A. (2026) https://BioRender.com/m3atqll.

Blood cancers disrupt this physiologic process, as observed in acute leukemias, which are marked by the clonal expansion of transformed hematopoietic progenitor cells. The precise developmental stages resembled by the malignant clones strongly influence treatment selection and often predict clinical response, as inferred in human patient specimens and corroborated in genetically engineered mouse models in vivo.9, 10, 11, 12 The clinical approach to selection of appropriate therapies for individual patients with acute leukemia can be classified into 3 general approaches: (1) drugs targeting a clonal oncogenic mutation, (2) therapies that exploit therapeutic vulnerabilities of specific lineages, or (3) nonspecific cytotoxic therapies with activity in the relevant lineage. Although drugs targeting tumor-selective oncogenic mutations are ideal, monotherapy with these agents is rarely curative, and we lack such drugs for most patients with acute leukemias. Thus, the selection of treatment regimens with curative intent relies heavily on drugs expected to have activity in the relevant cell lineage, highlighting the importance of accurate diagnostic classification.

In clinical medicine, the diagnostic classification of leukemic blasts has traditionally depended on detailed assessments of morphological and immunophenotypic features to gauge their differentiation status and lineage assignment relative to normal hematopoietic cells. However, current diagnostic approaches rely on a fixed set of phenotypic markers, a few of which are considered lineage defining: CD3 for the T-cell lineage, CD19 in combination with at least 1 additional B-cell marker (CD10, CD22, and CD79) for the B-cell lineage, and either myeloperoxidase or at least 2 monocytic markers (CD11c, CD14, CD36, CD64, and lysozyme) define the myeloid cell lineage.¹³ However, these markers often lack the precision needed to distinguish distinct cell states within a given lineage, such as the multiple differentiation states shown in Figure 1, for example, thereby complicating accurate leukemia diagnosis.

Current classification systems have been augmented by the inclusion of specific genetic lesions. Contemporary systems established by the World Health Organization (WHO Classification of Tumours; fifth edition) and the International Consensus Classification, both released in 2022, use frameworks combining morphological, immunophenotypic, and genetic information to define diagnostic disease groups. In practice, this means that, beyond assessing cell surface markers and cytochemical features, clinicians now routinely screen for recurrent genetic aberrations, including point mutations, chromosomal abnormalities, and fusion genes. For example, in acute myeloid leukemia (AML), the presence of mutations in genes such as NPM1, FLT3, or CEBPA as well as chromosomal abnormalities such as t(8;21) or inv(16) establishes defined AML subtypes and provides important prognostic and therapeutic information. Similarly, in acute lymphoblastic leukemia (ALL), genetic alterations such as the BCR::ABL1 fusion (the Philadelphia chromosome) or KMT2A (MLL) gene rearrangements (KMT2Ar) are important for disease stratification and therapy selection.

Although integrating immunophenotypic and genetic data has enhanced diagnostic precision for many leukemias, current approaches are not always sufficient. This challenge is especially pronounced in acute leukemias of ambiguous lineage, whose cell surface marker expression pattern precludes categorization into 1 specific lineage, especially when these lack a disease-defining genetic lesion. The WHO fourth edition introduced “acute leukemias of ambiguous lineage” to better characterize entities that blur myeloid and lymphoid lineages traditionally considered distinct.¹⁴ These include acute undifferentiated leukemias, which lack expression of any lineage-defining marker per the WHO fifth edition, and mixed-phenotype acute leukemias (MPAL), which express markers considered to define >1 hematopoietic lineage.¹³

Despite the aforementioned improvements, some biologically related leukemias are currently divided across separate diagnostic entities owing to their variable expression of markers considered to be lineage defining. One example is acute leukemias with shared T-lymphoid and myeloid features that can express myeloperoxidase (myeloid defining) and/or CD3 (T-lymphoid defining), leading to their variable classification as AML, T-ALL (often the early T-cell progenitor [ETP] subset of T-ALL¹²), or MPAL. This led to proposals to unify them as acute myeloid/T-lymphoblastic leukemia,¹⁵ which could be considered a subset of acute leukemias of ambiguous lineage, instead of the separate AML, T-ALL, and MPAL lineages that are assigned to these cases within the WHO fifth edition classification. Recent insights into the biology of MPAL have prompted the WHO fifth edition classification to include genetically defined subsets of MPAL based on BCR::ABL1 or KMT2A rearrangements.¹³ However, further diagnostic improvements are needed to incorporate the continuously dynamic nature of hematopoietic differentiation into diagnostic systems and therapy stratification.

Technological scRNA-seq advances have enabled the simultaneous measurement of the expression of thousands of genes in individual cells. This high-resolution multimarker approach facilitates the identification of diverse hematopoietic cell types while capturing dynamic transcriptional states and activated gene expression programs.¹⁶ When applied to leukemic specimens, these techniques have revealed significant transcriptional heterogeneity, uncovering cell populations associated with distinct phenotypic states, some resembling normal stem and multipotent progenitors with multilineage potential and others aligning with more committed progenitors exhibiting restricted differentiation capacity.¹⁷

Furthermore, single-cell studies conducted on fetal, pediatric, and adult hematopoietic tissues, together with data from purified hematopoietic cell populations, have led to the development of increasingly comprehensive atlases of normal blood development. These maps capture a wide spectrum of hematopoietic cell differentiation by delineating stage-specific transcriptional and epigenetic programs that regulate developmental cellular plasticity and lineage commitment. Single-cell profiles of leukemic cells can be projected onto these atlases to infer the corresponding developmental states of malignant cells, and to identify aberrant cell states that deviate from normal differentiation trajectories and lineages. Indeed, the detection of phenotypically multipotent leukemic cells in clinical specimens has been linked to disease relapse and therapy resistance.18, 19, 20, 21, 22

Here, we examine recent studies of high-risk lymphoid and myeloid leukemias with a focus on cellular heterogeneity and lineage plasticity, and we explore how leukemic blasts, despite being traditionally classified by fixed markers of differentiation, often retain multipotent features reminiscent of normal hematopoietic stem and progenitor cells (HSPCs). The coexpression of stem, myeloid, and lymphoid markers in high-risk leukemias underscores the dynamic nature of malignant transformation and its implications for therapy resistance. We propose that understanding the co-option of multipotent differentiation programs should lead to their diagnostic classification as MAL. In turn, this should be useful in refining current diagnostic criteria and developing improved strategies for treatment stratification in future clinical trials.

Defining MAL

Despite diverse genetic and phenotypic manifestations, certain leukemia subtypes share common multipotent states that can promote lineage plasticity and alter therapeutic responses. Multipotency refers to the ability of stem and progenitor cells to give rise to >1 lineage within a specific tissue or developmental system. It is formally defined using functional experiments (eg, colony-forming unit assays, transplant assays, and genetic barcoding or inducible labeling), which are difficult to incorporate into clinical diagnostics, but can also be inferred using computational assays (transcriptomic or epigenetic features),23, 24, 25 which can be incorporated into contemporary molecular diagnostics.

B-ALL with multipotent-like features

Although the origin of B-ALL has been hypothesized to derive from progenitors already committed to the B-lymphoid cell lineage, recent studies demonstrated that its origin can be more primitive and involve stem and early lymphoid progenitor cells (multilymphoid progenitor, common lymphoid progenitor [CLP], and pre–pro-B cells).²⁰ This is indeed the case for B-ALL with ZNF384 gene rearrangements, alterations commonly also found in subsets of B/myeloid MPAL.^26,27 Recent single-cell transcriptomic studies have revealed that leukemic cells with ZNF384 rearrangements (ZNF384-r) retain transcriptional signatures that closely resemble normal HSCs.²⁰

Moreover, B-ALL subtypes harboring DUX4 rearrangements (DUX4-r), KMT2A rearrangements (KMT2A-r), and BCR::ABL1 have been found to exhibit at least 2 distinct cellular transcriptional states: 1 characterized by a conventional predominance of committed lymphoid progenitors at the pro-B stage and another displaying a distinct enrichment of more immature lymphoid progenitors that retain capacity to differentiate toward the myeloid lineage. This latter subset of B-ALL, enriched in multipotent-like cells, shows alteration-specific parallels with B/myeloid MPAL; for instance, B-ALL with ZNF384-r closely resembles ZNF384-r MPAL, and similar patterns are observed for both KMT2A-r– and BCR::ABL1-driven cases. These findings suggest that such leukemias may originate from a common hematopoietic precursor with multilineage potential, representing a continuous spectrum of closely related leukemias that we propose should be termed MAL.

Interestingly, these multipotent B-ALL subtypes exhibit upregulation of myeloid progenitor programs with overexpression of the myeloid transcription factor CEBPA and downregulation of B-cell lineage restriction programs. Despite their transcriptomic features of multipotency, these entities are currently classified as B-ALL or MPAL based on variable morphological and immunophenotypic features.¹³ Notably, in pediatric B-ALL, the presence of leukemic blasts with a multipotent phenotype correlates with inferior clinical outcomes,²⁰ underscoring that developmental heterogeneity can foster treatment resistance. Indeed, single-cell multiomics analyses of primary B-ALL specimens demonstrated that leukemia cells with a pre–pro-B phenotype are resistant to asparaginase, whereas those exhibiting a pro-B–like state are markedly sensitive.²⁸ This study also pinpointed B-cell lymphoma 2 (BCL2) as a critical mediator of resistance in pre–pro-B–like cells and its pharmacological inhibition using venetoclax substantially enhanced asparaginase efficacy both in vitro and in vivo. We propose that the inference of multipotency using specific diagnostic assays that can guide treatment decision-making will help refine the current molecular classification of B-ALL and may be used to improve risk stratification.

T-ALL with phenotypic progenitor cells

Studies of normal hematopoiesis have also helped to delineate the developmental origins of T-ALL, which currently includes at least 15 subtypes with distinct genomic features, gene expression patterns, immunophenotypic states, and clinical outcomes.²⁹ Using multiple approaches, including scRNA-seq, chromatin accessibility profiling, and proteomics, Xu et al recently described the presence of a bone marrow progenitor–like (BMP-like) cell population in ETP ALL, near-ETP ALL, and non-ETP T-ALL specimens that is associated with treatment resistance, transcriptional plasticity, and stem cell–like features.³⁰ This subpopulation can represent <5% of the leukemic cells at diagnosis, making it difficult to detect using traditional bulk assays whose ability to detect intratumoral heterogeneity is limited. BMP-like cells share phenotypic features with hematopoietic progenitor cells, including the expression of HOXA9, MEF2C, and RUNX1 genes associated with self-renewal and multipotency, and absence of mutations in the NOTCH1 pathway signaling, stereotypically observed in other subsets of T-ALL with more differentiated leukemic cells.

Likewise, using integrative gene expression profiling, chromatin accessibility assays, and conformational genome mapping, Wang et al recently identified ZMIZ1, a protein inhibitor of activated STAT (PIAS)-like transcriptional coactivator, as a master regulator of a stem-like transcriptional network in subsets of ETP-ALL.³¹ ZMIZ1 was previously reported to be overexpressed in ETP-ALLs relative to normal hematopoietic progenitors³² and more recently was found to bind stem cell super enhancers that regulate key oncogenes in BMP-like ETP-ALL, including MYCN, BCL2, MEF2C, and MYB. ZMIZ1 depletion in normal and malignant ETPs reduces population growth, decreases myeloid potential, and promotes NOTCH1-driven T-cell differentiation. Conversely, ZMIZ1 overexpression enhanced ETP expansion and opposed T-cell differentiation, supporting its role in maintaining an undifferentiated, stem-like leukemia state. Intriguingly, BMP-like blasts are also present in myeloid and mixed-phenotype leukemias. Clinically, patients with a high burden of BMP-like cells at diagnosis exhibit worse overall survival and an increased risk of relapse,³⁰ suggesting that incorporating multipotency markers into clinical risk stratification will improve our ability to deliver risk-adapted therapy. Thus, as with subsets of B-ALL, subsets of T-ALL should be more accurately classified as MAL.

AMLs with multipotency features

AMLs encompass a heterogeneous group of diseases with diverse morphologies, immunophenotypes, genetic alterations, and clinical presentations. Genetic aberrations shape AML biology, lead to distortion of normal differentiation, and activate unique transcriptional programs that cooperate with secondary lesions to establish biologically distinct AML subtypes. The advent of scRNA-seq technologies has transformed our ability to study AML at an unprecedented resolution and has provided further insights into its heterogeneity and cellular hierarchies.³³ Early scRNA-seq studies revealed that AML does not represent a homogeneous expansion of a single malignant population but instead spans a wide spectrum of transcriptional differentiation states. Leukemic cells can mirror various stages of hematopoietic differentiation, ranging from primitive stem and progenitor-like cells to more differentiated monocyte or erythroid-like phenotypes. These analyses revealed distinct enrichment patterns of AML differentiation states across major classification frameworks of AML heterogeneity, including the WHO fifth edition classification and the International Consensus Classification 2022. For example, KMT2A-r and CBFB::MYH11 were linked to early stem/progenitor states, whereas NPM1 mutations aligned with monocytic and dendritic cell states. TP53-mutated and complex-karyotype AML showed strong enrichment of early and late erythroid states, which was more pronounced when both TP53 mutation and complex karyotype were coexisting. Interestingly, samples with RUNX1 mutations were found to be associated with enrichment for early lymphoid-like signatures. This association was stronger in cases with biallelic RUNX1 mutations as compared with those with monoallelic mutations, suggesting a dose-dependent effect of RUNX1 loss on lineage plasticity.²¹ Given that RUNX1 is a master regulator of hematopoietic cell differentiation, its inactivation may lead to aberrant lineage priming. Stem and early progenitor-like states have been demonstrated to impair treatment responses. Ng et al³⁴ developed a stem cell score (LSC17 score) comprising a 17-gene expression signature derived from functionally validated leukemic stem cell–enriched populations. This score identifies patients with a high-stemness transcriptional program who have inferior survival outcomes even after adjusting for established cytogenetic and molecular risk groups. This score correlates with minimal residual disease (MRD) persistence and predicts resistance to both conventional chemotherapy and novel targeted therapies.^35,36 The concept that multilineage potential can inform therapeutic responses in AML has recently been bolstered further. In a reanalysis of ex vivo drug sensitivity data, lymphoid differentiation was associated with increased susceptibility to approved agents such as sorafenib and midostaurin. Clinically, AML differentiation states inform clinical outcomes, and stem states define high-risk subtypes of AML, demanding refinement of the current risk stratification models to better predict patient outcomes and guide therapeutic decisions.

Collectively, these studies suggest that transcriptional mechanisms governing normal HSPC development are co-opted in distinct forms of aggressive leukemias, regardless of their conventional classification as B-ALL, T-ALL, or AML. This raises the possibility that certain leukemias may originate from a common progenitor state, breaking the boundaries between traditionally defined myeloid and lymphoid malignancies. Instead, they exhibit MPAL-like features, suggesting that they may be more accurately defined as MAL.

Importantly, the observed lineage plasticity in distinct high-risk lymphoid and myeloid leukemias is consistent with the current models of normal hematopoiesis. In particular, progenitor cells, such as those in the thymus destined for T-cell differentiation, retain latent myeloid potential.³⁷ Despite receiving strong signals promoting T-cell differentiation from the thymic microenvironment, these progenitor cells continue to express combinations of transcription factors and immunophenotypic markers associated with both T-cell and myeloid lineages. Likewise, early B-cell progenitors also exhibit latent myeloid cell potential. During normal hematopoiesis, progenitor cells at the CLP and pre–pro-B stages express a shared set of transcription factors that regulate B-cell differentiation, such as E2A and EBF1. However, these cells can also transiently coexpress myeloid-associated factors such as SPI1 (PU.1) or CCAAT-enhancer binding protein (C/EBP) factors, reflecting their origin in progenitors with shared myeloid and lymphoid potential. Under typical conditions, the bone marrow and thymus microenvironments reinforce B-cell and T-cell commitment, respectively, while suppressing myeloid development.^38,39 Thus, multipotency and shared lineage features of specific high-risk MAL can reflect malignant transformation of normal multipotent progenitor cells and/or dedifferentiation and acquisition of multipotency in response to particular oncogenic mutations.

Lineage switching in MAL

Leukemias with multipotent cell populations exhibit greater lineage plasticity, allowing for lineage switching (change in the leukemic immunophenotype from one hematopoietic lineage to another) in response to therapeutic pressure, with relapsed leukemias transitioning from lymphoid to myeloid lineage or vice versa. For instance, lineage switching from ALL at diagnosis to AML at relapse has been described for subtypes of ALL with enrichment of multipotent-like cells, such as BCR::ABL1, KMT2A, DUX4-r, and ZNF384-r leukemias after either induction chemotherapy or CD19-directed immunotherapies.40, 41, 42, 43, 44, 45, 46 Chen et al, using a single-cell assay for transposase-accessible chromatin with sequencing and RNA-seq in samples from infants aged <6 months with B-ALL KMT2A-r, demonstrated that leukemic cells undergoing lymphoid-to-myeloid lineage switching exhibit significant heterogeneity with coexpression of both B-lymphoid and myeloid lineage genes and increased populations of phenotypic HSPCs expressing KMT2A-r.¹⁸ Transplantation of these HSC-like cells into immunodeficient mice resulted in leukemias with more myeloid immunophenotypes, demonstrating that early leukemic hematopoietic progenitors in these patients retain the capacity to differentiate along multiple lineages. In particular, leukemias arising from early HSPCs appear to retain epigenetic and transcriptional programs that enable bidirectional lineage plasticity. Such plasticity allows for dynamic shifts in lineage identity, which can emerge under the selective pressures imposed by chemotherapy or immunotherapy.^46,47 Molecular profiling of relapsed specimens frequently reveals clonal continuity with the original disease, indicating that lineage switch does not reflect the emergence of a second primary malignancy, but rather an evolution of the initial clone.

In addition to multipotent progenitors capable of differentiating toward >1 lineage under therapeutic pressure, other mechanisms have been hypothesized to explain lineage plasticity, including rewiring of epigenetic and transcriptional networks, clonal selection, and changes induced by the tumor microenvironment. KMT2A::AFF1 leukemias can undergo a lymphoid-to-myeloid lineage switch via changes in chromatin and rewiring of gene regulatory networks upon CHD4 gene abnormalities.⁴⁰ This demonstrates that epigenetic dysregulation can impair lineage fidelity, enabling leukemic cells to transition from lymphoid to myeloid programs under therapeutic pressure. Clonal selection can also favor expansion of preexisting myeloid-primed subclones,⁴⁸ whereas therapy-induced stress and genetic instability, particularly in the context of TP53 mutations, may facilitate cellular reprogramming or dedifferentiation. For example, TP53 mutations and complex karyotypes have also been detected in patients who had a lymphoid-to-myeloid switch under the selective pressure of B-cell–directed therapies.⁴⁹ Moreover, the tumor microenvironment, by releasing cytokines and metabolic factors, can promote differentiation of leukemic cells, affecting disease initiation, progression, and therapy response. For instance, BCR::ABL1-rearranged B-ALL can be reprogrammed into macrophage-like cells that express myeloid markers when exposed to myeloid-inducing cytokines (interleukin-3, macrophage colony-stimulating factor, and granulocyte-macrophage colony-stimulating factor).⁵⁰

Myeloid-to-lymphoid lineage switching is also associated with adverse outcomes in AML.^44,51 Comparative analysis of AML specimens obtained at diagnosis and relapse showed that relapsed AML exhibited less mature transcriptional states, with the appearance of B-lymphoid-like features.¹⁹ For example, specimens with KMT2A-r and RUNX1::RUNX1T1 presented with myelomonocytic-like cells at diagnosis but exhibited CLP-like to pre–B-cell–like lymphoid cells upon relapse. Relapsed cells expressed B-lymphoid markers, including CD79A and CD19. Importantly, in these cases, conventional immunophenotyping failed to detect B-lymphoid cells at diagnosis, but scRNA-seq enabled the detection of cells with B-lymphoid programs, enforcing the need for more sensitive diagnostic approaches for MAL.

Lineage switching poses a particular challenge for therapies that exploit lineage-selective vulnerabilities. Moreover, lineage switching can also lead to the evasion of diagnostic detection focused on fixed immunophenotypes, such as, for example, using flow cytometry with limited markers for the detection of MRD. Moreover, lineage switches can arise from distinct stages within the lymphoid developmental hierarchy, potentially impairing MRD assessment based on sequencing of immunoglobulin receptor gene rearrangements. Indeed, when lineage switches originate from an earlier progenitor stage of leukemogenesis, the leukemic population responsible for the switch may lack the B-cell receptor or T-cell receptor gene rearrangements that were detectable in the bulk population at initial diagnosis. For example, in a pediatric patient diagnosed with ALL positive for KMT2A-r and rearranged immunoglobulin heavy chain (IgH-r), lineage switch to AML was accompanied by loss of the IgH-r, despite persistence of the KMT2A-r mutation.⁴⁸ Therefore, the absence of an immunoglobulin gene rearrangement in the lineage switch clone does not necessarily exclude a direct clonal relationship between the original and lineage-switched leukemic cell populations, necessitating alternative strategies to monitor MRD. The persistence of the hallmark fusion oncogene or somatic driver mutation across both disease phases supports the hypothesis that the B-ALL and AML derive from the same ancestral clone. Given the growing recognition of lineage plasticity in leukemia and its clinical impact, improved diagnostic and therapeutic strategies are needed to target key regulators of lineage commitment and developmental stability.

Future directions to improve the diagnosis and therapy of multipotency in leukemia

Current therapies for leukemias of ambiguous lineage result in meaningful rates of durable remissions and 5-year leukemia-free survival for both children and adults.^52,53 However, distinct leukemia subsets continue to have inferior outcomes with current therapies, necessitating improved approaches. The existence of hematopoietic progenitor cells with both lymphoid and myeloid potential expands the spectrum of MAL, which may reflect the induction of oncogenic mutations in multipotent HSPCs. For example, most immature intrathymic T-cell progenitors retain myeloid potential until they commit to the T-cell lineage via B-cell lymphoma/leukemia 11B–dependent processes.54, 55, 56 Moreover, lineage tracing studies have shown that these immature progenitors can differentiate into macrophages and neutrophils that harbor T-cell receptor gene rearrangements in vivo.^57,58

In addition, ectopic expression of Myc and Bcl2 in early double-negative T-cell progenitors in mice has been found to induce acute leukemias with variable expression of both myeloid and T-lymphoid markers, thereby indicating that multipotent early T-cell progenitors can act as the cellular origin of MAL.⁵⁹ Alternatively, MAL can be induced by aberrant transdifferentiation in response to specific oncogenic mutations. For example, inactivating mutations in RUNX1 within hematopoietic progenitors lead to abnormal myeloid differentiation,⁶⁰ and KMT2A-r can trigger reactivation of a self-renewal gene expression program in committed progenitors.⁶¹ We posit that mediators of multilineage differentiation potential and/or oncogenic transdifferentiation could represent rational targets for therapy of MAL (Figure 2).

Figure 2.

Shared pathophysiology of MAL. Schematic of normal hematopoiesis (left), involving multipotent progenitor cells, which share molecular properties with multipotent cell–like leukemia blasts, contributing to MAL in conventionally diagnosed ALL, AML, and MPAL. This can give rise to lineage switching and inferior outcomes with current clinical therapies, which may be targeted using selective inhibitors (right). Ly, lymphoid; My, myeloid. The figures were created with BioRender.com. Gutierrez A. and Kentsis A. (2026) https://BioRender.com/m3atqll.

One major transcriptional driver of multipotency in MAL is KMT2A-r mutations. Recent efforts have revealed that these and other acute leukemias with distinct immunophenotypes but similar dysregulated HOX gene expression, including NUP98- and UBTF-rearranged or NPM1-mutant leukemias, require a specific interaction of the KMT2A:menin chromatin remodeling complex.62, 63, 64, 65, 66 As a result, pharmacologic menin:KMT2A inhibitors can downregulate oncogenic gene expression, including stem and myeloid cell progenitor self-renewal programs, leading to leukemia cell differentiation and death. Menin:KMT2A inhibitors are under active clinical investigation to define optimal combination therapies for HOX-dysregulated myeloid leukemias.⁶⁷ To the extent that KMT2A:menin–dependent gene expression may contribute to the maintenance of myeloid and stem cell gene expression programs in MAL with specific molecular mutations present, investigation of menin inhibitors may represent a rational approach to elimination or therapeutic reprogramming of MAL cells (Figure 2).

Oncogenic KMT2A rearrangements also induce the expression of MEF2C, a member of the MADS family of transcription factors that regulates hematopoietic self-renewal and differentiation, which is also expressed in subsets of leukemias without KMT2A rearrangements. Rearrangements leading to overexpression of MEF2 family members, including MEF2B, MEF2C, and MEF2D, are recurrent in refractory lymphoblastic leukemias and lymphomas. Depletion of MEF2C blocks self-renewal of KMT2A-r mixed lineage leukemias in vivo.⁶¹

Importantly, aberrant MEF2C expression and activation are observed in diverse subtypes of human AML and are associated with chemotherapy resistance. Indeed, AML relapse–associated lineage switching is associated with increased expression of MEF2C.¹⁹ Because MEF2C phosphorylation is regulated by distinct kinases, including salt-inducible kinase (SIK) effectors SIK2/3, treatment of MEF2C-activated AMLs with SIK2/3-selective kinase inhibitors can overcome cytarabine resistance in vitro and block leukemia progression in vivo.^68,69 Likewise, enforced MEF2C expression in T-cell progenitor cells induces biphenotypic ETP leukemias, in which treatment with SIK inhibitors promotes differentiation into the T-cell lineage and sensitizes cells to prednisolone treatment.⁷⁰ Thus, targeting aberrant MEF2C activation may represent a rational approach to therapy of MAL (Figure 2).

Independent of the specific genetic driver alteration, additional factors that control hematopoietic self-renewal and differentiation appear to be co-opted and may constitute rational therapeutic targets in MAL. MYB, a transcription factor critical for hematopoietic progenitor proliferation and differentiation, is frequently aberrantly activated in multipotent leukemias through diverse mechanisms. In leukemias driven by KMT2A rearrangements, MYB is activated because of direct transcriptional induction mediated by fusion oncoproteins, resulting in persistent MYB expression and maintenance of multipotent, progenitor-like states.⁷¹ MYB is also aberrantly co-opted in diverse non–KMT2A-rearranged subtypes of AML, in which MYB assembles an aberrantly organized transcriptional coactivator complex, leading to the induction of leukemogenic gene expression and blockade of normal hematopoietic differentiation.⁷² In subtypes of T-ALL, MYB can be activated through oncogenic mutations of noncoding regulatory elements.⁷³ In forms of B-ALL, MYB is induced by oncogenes such as BCR-ABL1, cooperates with lineage-controlling transcription factors including PAX5 and EBF1, and is required for aberrant self-renewal and resistance to apoptosis of leukemia cells (Figure 2).

Critically, MYB exhibits specific oncogenic functions in leukemia cells that can be separated from its requirements for normal blood development. This led to the development of pharmacologic MYB mimetic inhibitors that selectively interfere with the oncogenic MYB:CBP/P300 transcriptional coactivator complex in leukemia but not normal blood progenitor cells.⁷⁴ Thus, clinical-grade MYB mimetic inhibitors are currently being explored for clinical trials for patients, which may represent a rational approach to MAL, given their reliance on stem and myeloid progenitor gene expression dysregulation, which frequently involves MYB. Likewise, MYB-selective small-molecule splicing inhibitors are currently being investigated in ongoing clinical trials for AML, with potential for expansion to other acute leukemias (ClinicalTrials.gov identifiers: NCT06297941 and NCT06462183; Figure 2).

In addition, lineage-controlling transcription factors important for stem and progenitor cell function control cellular apoptosis via the expression of specific BH3 factors, suggesting that their pharmacologic targeting can also be explored for MAL therapy. For instance, both MYB and MEF2C control the expression of BCL2, a key antiapoptotic protein that blocks the release of cytochrome c from mitochondria, thereby blocking activation of apoptosis.⁶⁸ The BCL2-selective inhibitor venetoclax can induce a rapid onset of apoptosis of leukemia cells, including those from MAL. Resistance to mitochondrial apoptosis is a common feature among therapy-resistant leukemic cells, particularly those thought to originate from HSPCs, as documented in conventional lymphoid and myeloid leukemias.⁷⁵ Thus, incorporation of venetoclax should be explored in investigational clinical trials of improved combination therapies for relapsed or refractory MAL (Figure 2).

Finally, effective therapeutic strategies for the targeting and elimination of MAL will require improved approaches to their diagnosis. In current clinical practice, therapy decisions are predominantly guided by lineage-defining immunophenotypic classification of leukemia cells and the presence of specific genetic mutations. However, recent evidence indicates that multipotent HSPC-like populations play crucial roles in chemotherapy resistance and disease relapse.^19,76,77 Thus, precise identification of these multipotent leukemia-initiating populations will be essential for diagnosing them clinically and optimizing precise therapeutic interventions (Figure 2). In particular, improved diagnostic modalities with the requisite resolution and sensitivity for detection of MAL cells can be combined with response-based risk stratification, in which patients with inadequate response to conventional therapies or disease relapse can be nominated for investigational assays to detect MAL.

High-resolution immunophenotypic strategies, including multiparametric flow cytometry and mass cytometry, offer scalable approaches with the requisite sensitivity and specificity for the detection of leukemia cell heterogeneity at single-cell resolution for diagnostic decisions. These technologies can precisely distinguish rare multipotent HSPC-like cells from more committed leukemia progenitors, facilitating improved monitoring of disease burden and MRD. However, the establishment of specific immunophenotypic marker combinations will require new prospective studies before they can be used in investigational clinical trials and patient care. For practical implementation, these technologies must be integrated into standardized clinical workflows. The development of reference immunophenotypic atlases, derived from well-annotated patient cohorts and linked to genomic and clinical data, would serve as essential resources for interpretation and benchmarking. To ensure reproducibility and clinical applicability, flow and mass cytometry assays would need to be validated across institutions with harmonized gating strategies, compensation settings, and instrument calibration protocols. Additionally, bioinformatics pipelines for automated gating, clustering, and visualization would need to be used to reduce subjectivity and enhance sensitivity in detecting MAL subpopulations.

Additionally, high-throughput gene expression profiling methods, including both bulk and scRNA-seq, can provide extensive detection of the gene expression mechanisms governing leukemia cell stemness and lineage plasticity, and consequently diagnostic markers for MAL definition and clinical detection. However, implementation of these approaches for clinical use will require improvements in their practical accessibility, robustness, and cost. We would also recommend a future effort led by the WHO classification group to establish specific diagnostic criteria for MAL. These are important directions for future work in the field. By leveraging improved understanding of the molecular determinants of MAL, we can define specific MAL prognostic biomarkers, thereby enhancing diagnostic accuracy, predicting therapeutic response, and identifying patients with a need for improved targeted therapies. Together, the integration of molecular phenotypic, cell composition, and genetic features should substantially refine leukemia diagnosis and guide more effective personalized therapeutic strategies for patients with multipotent and other refractory acute leukemias.

Conflict-of-interest disclosure: A.G. is a consultant and scientific advisory board member for Attivare Therapeutics. A.K. is a consultant for Novartis, Rgenta, Blueprint, Syndax, and Sellas, some of which are developing investigational drugs for leukemia therapy.