Abstract

Bulk and single cell RNA sequencing have revolutionized biomedical research and empower researchers to quantify the global gene expression of populations and single cells to further understand the development, manifestation and the treatment of diseases like cancer. Acute myeloid leukemia (AML), a cancer of the myeloid line of blood cells, could benefit from these technologies as relapse and mortality rates remain high despite the extensive research conducted over several decades. This is partly because AML is a heterogeneous disease, differing substantially between patients and hence requiring more fine-grained classifications and specialised treatment strategies, for example by incorporating expression profiles. In addition, single cell RNA sequencing (scRNA-seq) can resolve genetic and epigenetic subclonal structures within a patient to improve understanding and treatment of AML. However, improving and adapting RNA-seq technologies is still often necessary to efficiently and reliably obtain expression profiles, especially from small or suboptimally processed samples. To this end, we developed a bulk RNA-seq protocol, which copes with the major challenges of limited sample quantities, different sample types, throughput and costs and subsequently applied this method to further understand the subclonal structures in AML. We were able to characterize a plastic cell state of AML cells that is defined by increased stemness and dormancy and could influence treatment outcome and relapse. For this, we isolated non-dividing AML cells based on a proliferation-sensitive dye from patient derived xenograft (PDX) models of two AML patients. We found that these cells have low levels of cell cycle genes confirming dormancy, and additionally had similar expression patterns to previously described dormant minimal residual disease (MRD) cells in lymphoblastic leukemia (ALL). This included high expression levels of cell adhesion molecules, potentially reflecting the persistence of dormant AML and ALL cells in the hematopoietic niche. Lastly, we could show that resting and cycling AML cells can transition between these two states, indicating that dormancy might be a general property of AML cells and not depend on particular genetic subclones. In a second project, we optimized a single cell RNA-seq technology. We used a systematic approach to evaluate experimental conditions of SCRB-seq, a powerful and efficient scRNA-seq method. Focussing on reverse transcription, arguably the most important and inefficient reaction, , we used a standardized human RNA (UHRR) and systematically tested nine different RT enzymes, several reaction enhancers and primer compositions to increase sensitivity. We found that Maxima H- showed the highest sensitivity and that molecular crowding using poylethylene glycol (PEG) could increase the efficiency of the reaction significantly. Together with several smaller changes in the workflow, primer design and PCR conditions, we developed mcSCRB-seq (molecular crowding SCRB-seq). We verified the 2.5x increase in sensitivity using mES cells in a side by side test between SCRB-seq and mcSCRB-seq, and further found mcSCRB-seq to be amongst the most sensitive methods using artificial RNA spike in molecules (ERCCS). Lastly, since method comparisons between studies suffer from missing accuracy due to batch effects and external factors, we participated in a complex scRNA-seq benchmark study aiming to provide a fair comparison between methods concerning sensitivity, accuracy and applicability for building expression atlases. In contrast to before, we found that in this particular setting, mcSCRB-seq did not perform well and ídentified fields for further improvement. In conclusion, my work described in this thesis not only contributes towards a deeper understanding of the emergence and progression of AML but also towards the development of experimental bulk and single-cell RNA sequencing methods, improving their widespread application to biomedical problems such as leukemia.