Getting to the root of high-risk leukemias Free

In this issue of Blood, Chen et al employ single-cell genomics to characterize a hematopoietic stem and progenitor cell (HSPC)-like subpopulation in primary pediatric leukemia samples associated with resistance to treatment and poor overall survival.¹ 

Leukemia-initiating cells (LICs), also often referred to as leukemia stem cells (LSCs), were first identified in the 1990s in pioneering experiments that demonstrated that only a tiny fraction of leukemia cells were able to initiate malignancy when transplanted into immunodeficient mice.² These and subsequent experiments revealed that LICs (LSCs) also have stemlike properties of differentiation and proliferation, evoking a model in which cancer stem cells sit atop a hierarchy of tumor cells and disproportionately contribute to long-term tumor growth and spread (see figure). Subsequent work has suggested that stemlike properties of these HSPC-like cells, including cellular plasticity and reversible quiescence, imbue these cells with critical roles in therapy resistance and relapse.³ 

Many open questions remain regarding HSPC-like cells, including the extent to which these cells represent a static population vs a more transient state and whether these cells are the major source of chemotherapy resistance.^4,5 Furthermore, to date, there have been limited data on the heterogeneous phenotypic properties of HSPC-like cells, both within and across patients. The rarity of HSPC-like cells, which have been estimated to be present at a frequency ranging from 1 in 1000 to 1 in 1 million cells, has made these questions difficult to answer.^2,6 For this reason, deep single-cell phenotyping is an essential tool to identify and characterize these rare cells in human leukemia samples.^2,6 

Here, Chen et al generated single-cell RNA-sequencing data for 649 599 cells and single-cell ATAC-sequencing data for 767 122 cells across 96 pediatric bone marrow and peripheral blood samples spanning diverse blood cancers including acute myeloid leukemia, B-cell acute lymphoblastic leukemia, T-cell acute lymphoblastic leukemia (T-ALL), and mixed-phenotype acute leukemia, including varied genomic subtypes with distinct driver mutations. The authors distinguished leukemic cells from healthy cells using a clustering-based approach complemented by somatic mutation and copy number variation inference. The authors then projected these cells onto trajectories derived from normal hematopoiesis, identifying leukemic blasts, which are HSPC- and multipotent progenitor–like, as well as blasts that more closely resemble lineage-committed progenitors (lineage-like). Interestingly, when patient cells were transplanted into immunodeficient mice, leukemias with more HSPC-like blasts did not engraft at a higher rate than lineage-like blasts, but HSPC-like blasts maintained a more stemlike phenotype postengraftment, as assessed by expression of stem cell and differentiated B-cell progenitor markers. Notably, HSPC-like blasts were greatly enriched for transcriptional signatures implicated in aggressive forms of leukemia, including a 17-gene stemness score previously associated with relapse risk, and a MECOM transcriptional network involved in HSC maintenance, which has been shown to be coopted in high-risk forms of acute myeloid leukemia (AML).^7,8 

Although there is a body of evidence linking cancer stem cells to therapy resistance in AML, some work has challenged this idea.^5,9 Here, Chen et al asked whether the identified HSPC-like cell states were associated with therapy resistance, finding that a higher HSPC-like fraction was positively associated with measurable residual disease level after induction therapy in AML and T-ALL samples and that HSPC-like cells were more resistant to vincristine and daunorubicin in vitro. To validate the identified signatures in external cohorts, the authors estimated scores for HSPC-like signatures in deconvoluted bulk RNA-seq data from the Therapeutically Applicable Research to Generate Effective Treatments (TARGET) AML and B-ALL cohorts and the Gabriella Miller Kids First T-ALL X01 cohort, finding a strong association between HSPC score and overall survival. This striking finding suggests a potential use for HSPC transcriptomic signatures in clinical prognostication, but prospective clinical trials will be necessary to validate this approach, and integrating RNA-seq into clinical use may face cost and workflow barriers. The authors went on to prioritize HOXA3/5/9, AP-1, and CEBPA as core transcription factors associated with HSPC-like cells and nominated potential drug targets among the regulatory networks associated with these factors, validating some of these targets as preferentially lethal to HSPC-like blasts isolated from early T-cell precursor ALL patient-derived xenografts.

Collectively, this study provides a rich data set and deeply characterized phenotypic profiling of HSPC-like cells in primary pediatric leukemias, identifying stemlike signatures that are linked to therapy resistance and overall survival. This work adds to a growing body of evidence that stemlike states are important for leukemic initiation, growth, and resistance to conventional therapies. Future work will be necessary to identify which cells among the HSPC-like subsets represent true LICs. Moreover, the transcriptional drivers underlying the acquisition or maintenance of these stemlike programs remain to be defined. Critical open questions include the overall hierarchy by which HSPC-like leukemia cells are organized and how rare stemlike cells contribute to more prevalent lineagelike leukemia blasts. Delineating these cellular relationships and the contributions of stemlike cells to the overall leukemic hierarchy both at tumor initiation and upon relapse will require lineage tracing in primary and posttreatment leukemia samples using emerging methods to identify cellular relationships in primary human cells.¹⁰ The work from Chen et al provides a wonderful starting point to consider the complex heterogeneity underlying the roots of high-risk leukemias.

Conflict-of-interest disclosure: The authors declare no competing financial interests.