Chemotherapy selects for myeloid subclones in multiple subtypes of B-cell acute lymphoblastic leukemia

Abstract Lineage infidelity, manifested as either lineage drift or lineage switch, has emerged as a mechanism of resistance to B-cell–targeted therapies. While lineage switch has been most extensively documented in KMT2A-rearranged (KMT2A-r) B-cell acute lymphoblastic leukemia (B-ALL), lineage switches have also been reported in non-KMT2A-r subtypes under the pressure of B-cell-targeted therapy. We anticipate that altered expression of B-cell antigens (including full lineage switches) will become more common as CD19 targeted T-cell engagers such as blinatumomab are broadly incorporated into frontline therapies. We and others have previously demonstrated the presence of small subclones with stem cell/myeloid transcriptional signature in B-ALL at initial diagnosis using single cell multi-omics, suggesting a pre-existing reservoir of cells that are not committed to a B-cell fate. To explore the presence and functional significance of lineage infidelity in B-ALL, we performed comprehensive multiomic profiling—including single-cell RNA sequencing (scRNA-seq), single-cell ATAC-seq (scATAC-seq), single-cell DNA-seq (scDNA-seq), CITE-seq, and high-dimensional spectral flow cytometry on a cohort of standard- and high-risk paired pediatric B-ALL samples collected at diagnosis and at the end of induction therapy. Our integrated analysis uncovered the following key insights: Evidence of CD19 surface protein negative stem cell/myeloid leukemogenic subclones: Prior studies identified stem cell/myeloid subclones using single cell transcriptomic methods, which are inferior in assessing surface protein expression of several antigens, including CD19, due to high drop-out. In this study we validated the existence of this stem cell/myeloid-like population by detecting known leukemia-defining mutations such as KRAS, PAX5, and CREBBP within this population, confirming their clonal relationship to the bulk leukemia, and we confirm their surface protein marker profile using CITE-Seq and spectral flowcytometry. The stem/myeloid-like population showed reduced or absent expression of canonical B-cell markers such as CD19, CD22, and CD79a, both at the transcript and protein levels. Unexpectedly, this phenotype was not restricted to KMT2A-r ALL but was also observed in other subtypes such as Ph-like, Ph+, and iAMP21. To assess the functional potential of these lineage-infidel subclones, we sorted CD19⁻ blasts and transplanted them into immunodeficient NSG mice. These cells were successfully engrafted and gave rise to leukemia. This is the first study to provide evidence of lineage plasticity by integrating surface protein marker detection with transcriptional profiling and functional assays documenting the leukemogenicity of these clones. Stem cell/myeloid-like blasts enriched at minimal residual disease (MRD): Analysis of paired initial diagnosis and MRD samples documented enrichment of stem cell / myeloid clones at MRD. Again, this was not confined to KMT2A-r samples but also seen in several patients with other high-risk genotypes. Integration with drug profiling suggests increase resistance of the myeloid subclones to standard ALL chemotherapy agents. These findings suggests that in a subset of patients ALL Induction chemotherapy selects for a residual CD19-negative blast population that may jeopardize the efficacy of subsequent B-cell directed immunotherapy such as blinatumomab. Collectively, our findings have two key implications: (1) our data challenges current risk stratifications that rely on B-lineage flowcytometry to detect residual blasts at the end of Induction. Patients with a shift in lineage phenotype may potentially be misclassified when using standard flow cytometry based on lymphoid markers only. Lineage infidelity may also explain the occasionally observed discordant MRD assessments when using both flowcytometry and molecular methods. (2) Our findings have implications for treatment. Flowcytometry for stem cell / myeloid markers at MRD may identify a subgroup of patients that benefits less from blinatumomab. In addition, we identified surface protein markers associated with the CD19-negative population that may serve as alternative targets for novel immunotherapies.