Addressing therapy failure and relapse is critical in improving outcomes for pediatric T-cell acute lymphoblastic leukemia (T-ALL) patients. Relapse is thought to represent a failure to effectively eliminate leukemia-initiating cells (L-ICs), which survive chemotherapy and self-renew, regenerating the full complement of leukemic cell populations. Results from mouse T-ALL models show that the double negative-3 (DN3) thymic progenitors are enriched for L-IC activity; however, the L-IC remains an undefined subset of this population. We used single cell RNA-sequencing (scRNA-seq) to profile leukemic development in vivo, with a focus on uncovering DN3 heterogeneity. Integrated UMAP clustering of scRNA-seq libraries from wild type (WT), Tal1/Lmo2 preleukemic (PL), and Tal1/Lmo2 T-ALL (Leuk) thymi revealed an expansion of DN3 progenitors in PL and Leuk samples with decreases in mature (single positive) thymocytes. Unlike WT thymocytes, where Notch1 target gene expression was restricted to DN3 cells, Notch1-Myc pathway activation was maintained throughout thymocyte development in Tal1/Lmo2 transgenic mice. Re-clustering of DN3 cells revealed the presence of dormant and proliferative DN3 cells in PL and Leuk thymus. Dormant DN3 signatures resembled quiescent, Notch1-active DN3a (pre-β-selection) thymocytes and proliferative DN3 cells were transcriptionally similar to DN3b thymocytes. Dormant mouse DN3 cells were enriched for signatures of quiescent, therapy resistant ALL patient cells and cells obtained from ALL patients at minimal residual disease (Ebinger et al., Cancer Cell 2016), suggesting that dormant leukemic DN3 cells resemble human L-ICs that mediate drug resistance and relapse. To test this, we employed in vivo nucleosome labeling to profile dormant DN3 cells throughout leukemia development and functionally evaluate their response to chemotherapy. To assess cell divisional history, WT, Tal1, Lmo2, or Tal1/Lmo2 mice harboring the tet-inducible H2B-GFP reporter allele were subjected to in vivo doxycycline pulse (labeling), followed by a two-week chase. Analysis of pulse-chased thymocytes revealed low frequencies of label retaining (GFPHI) DN3 cells in WT mice. However, label retention was significantly increased in Lmo2 DN3 thymocytes (2.9-fold), consistent with other Lmo2 models (Tremblay et al., Nat. Commun. 2018). Tal1 expression also promoted DN3 label retention (6.0-fold) and Tal1/Lmo2 co-expression resulted in further increases in DN3 GFPHI cells (25-fold), indicating that Tal1 and Lmo2 cooperate to promote DN3 dormancy. DN3 GFPHI cells were present at similar frequency in mice with overt T-ALL and in line with scRNA-seq data, GFPHI cells were observed nearly exclusively as a subset of the leukemic DN3a population. Treatment with Vincristine, DeXamethasone, and L-asparaginase (VXL) led to a further increase in GFPHI DN3 frequency in Lmo2 and Tal1 preleukemic mice (3.1- and 9.7-fold, respectively), suggesting that cell cycle restricted L-ICs may evade conventional chemotherapy and mediate relapse. We then tested whether dormancy is a feature of the L-IC in T-ALL patients. Analysis of relapsed pediatric T-ALL cells revealed that the CD7+CD1a- L-IC population (Chiu et al., Blood 2010) is cell cycle restricted compared to CD7+CD1a+ leukemic cells and that VXL treatment enriched for CD7+CD1a- cells in xenografted mice, suggesting that cell cycle restriction may be an important feature of both mouse and human T-ALL L-ICs. Analysis of transcriptomes to reveal mediators of dormancy that may be exploited to chemosensitize L-ICs will be discussed.
Read full abstract