Intrapatient Complex Transcriptional Responses to Conventional Chemotherapy in Relapsed Acute Myeloid Leukemia Revealed By Single Cell RNA-Seq Analysis

Resistance to anthracycline and cytarabine based conventional chemotherapy often occurs and results in extremely poor prognosis in patients with acute myeloid leukemia (AML). Although chemotherapy resistance is the most critical clinical problem, the mechanisms by which AML confers resistance to conventional chemotherapy are not yet fully understood. In this study, we investigated the key mechanisms of chemotherapy resistance through single cell RNA-sequencing analysis using paired bone marrow AML cells longitudinally collected from two AML-MRC patients at diagnosis and relapse after anthracycline-based chemotherapy. AML blasts were sorted by CD45/SSC gating and subjected to single cell RNA-seq analysis. Single cell RNA-seq was performed using 10x Genomics' Chromium System. Mean estimated number of cells per sample was 3.403 (2,731-4,200) and median detected genes per cell ranged 3,030 to 3,918 among four samples. Data collected from paired samples were combined in following analysis. Transcriptome based clustering following UMAP dimensionality reduction distinguished 5 and 9 cluster groups in each paired sample. Chemotherapy sensitive cluster groups dominant at diagnosis and chemotherapy resistant cluster groups dominant at relapse were clearly divided. In each paired sample, a few AML cells at diagnosis were allocated to chemotherapy resistant cluster groups. This suggested that transcriptionally identifiable less frequent cells resistant to chemotherapy existed at diagnosis and may expand during and/or after chemotherapy maintaining its transcriptional features. Next, to determine whether these transcriptional features are correlated with DNA mutation profiles, we labeled DNA mutation status to each cell and compared frequencies of mutation. As far as we detected, AML recurrent mutations such as DNMT3A ^R882C and TP53 missense mutation were not related to chemotherapy resistant cluster groups, although this method was relatively limited by the nature of RNA-seq-based mutation detection. Then we sought to determine transcriptional features of resistant clones. Gene set enrichment analysis identified some gene groups such as E2F signaling pathway, MYC signaling pathway, hedgehog signaling pathway and TNFA signaling pathway as transcriptional signatures related to emergence after chemotherapy. Analysis of known hematopoietic differentiation gene signatures showed distinct differentiation profiles in each cluster groups, whereas resistant cluster groups were not necessarily related to hematopoietic stem cell signatures. Intrapatient variations of transcriptional signatures among the resistant cluster groups were detected, which indicated that accurate detection of transcriptional features related to chemotherapy resistance may be difficult by using bulk RNA-seq method. As for other cluster groups which were not dominant both at diagnosis and relapse, these cluster groups hardly changed its frequencies between at diagnosis and relapse, which suggested less proliferative leukemia cells persisted during chemotherapy and have various transcriptional features although whether these persisting cells contribute to relapse was unclear. Since enriched transcriptional signatures in resistant cluster groups were not consistent between the two patients, further analysis using samples collected from more patients would be needed to determine common critical chemotherapy resistant transcriptional signature. In conclusion, our analysis suggested that a transcriptionally identifiable small fraction of cells showing gene signatures related to chemotherapy resistance at diagnosis may expand during chemotherapy and revealed intrapatient transcriptional complexity of response to chemotherapy, which cannot be uncovered by bulk RNA-sequencing.