Preview

PDF, English - main document Download (4MB) | Terms of use

Abstract

Success rates of cancer treatment have continually increased as new targeted therapies have been approved and stratification of patients has developed. Much work remains however, as only 13% of US patients responded to immune checkpoint inhibitor therapies in 2018, and a mere 4.9% responded to targeted therapies. To further increase treatment success, systematic methods for quickly, cheaply and accurately determining new patient-specific drug combinations are essential. Drug combinations provide a unique opportunity to quickly develop new treatments, increase treatment efficacy and prevent relapse. This will eventually allow treatments to be chosen which reflect the complexity of the cancer itself, by targeting multiple clones or targetable aberrations within a cancer. One promising approach for more complete patient stratification is via drug-perturbation screens directly on patient cancer cells, in which patient-specific drug sensitivity is determined without requiring any biomarker information. Such screens can help doctors to choose between available treatments and could even suggest new combinations of those. In addition, data from perturbation screens are essential to build more accurate mathematical models of drug response and identify new biomarkers which indicate drug sensitivity. One area where perturbation screens would be incredibly useful, is in the prescription of Immune Checkpoint Inhibitors (ICIs). In the first part of this thesis, work towards this goal is presented. A robust workflow was established for selection and activation of cytotoxic T-lymphocytes (CTLs), and subsequent induction of CTL-mediated target cell death. Two different methods were tested to enable quantification of CTL activity and the outlook for development of such ICI screens is discussed. Another important focus of personalized medicine research is the development of high-throughput screens of targeted therapies with high-content readouts. Unfortunately, such approaches usually require large numbers of cells, so are incompatible with screening directly on patient cells due to the low number of cells which are typically available. Use of droplet microfluidics allows very significant reduction in volume per sample, and therefore cell numbers required. Transcriptomics is an ideal candidate for a high-content readout, as it has proven to be one of the techniques with the best predictive capacity for patient stratification. In the second part of this thesis, I build on previous work in the group which used droplet microfluidics and combinatorial DNA barcoding to identify cell line-specific drug combinations via a transcriptomic readout. In the previous workflow, the drugs which could be used were severely limited due to exchange of hydrophobic drugs between droplets. To address this issue and bring the workflow closer to one which could feasibly be used to inform decisions made in the clinic, I adapted and integrated different microfluidic components, ensuring a robust and rapid workflow. Specifically, in order to perform experiments in plugs rather than surfactant-stabilized droplets, I further developed a microfluidic device for consistent injection of reagents into microfluidic plugs. I also identified a detergent which was compatible with plugs, as standard cell lysis agents led to cross-contamination of contents between microfluidic plugs. Next I validated the workflow by developing a qPCR-based approach for quantifying differential expression of target genes, by incorporating qPCR barcodes during reverse transcription. Finally, I ran a preliminary experiment to treat target cells in microfluidic plugs, then perform reverse transcription and sequencing to determine whether differences in gene expression could be identified in the barcoded cDNA.