Development of a novel method for single-cell whole-genome amplification to study cancer heterogeneity.

All the cells in a human body have arisen from one single cell, the zygote. Throughout development cells divide, make fate decisions, and transition between different cell types and cell states. Every cell division the genetic code is copied and passed on to the daughter cells. While this is under tight control, mistakes do occur. In most cases these mutations have no effect, but they can have severe consequences as is exemplified by the birth of children with a de novo genetic disorder or the development of cancer. Cancer develops by accumulating genetic mutations and chromosomal alterations which gives rise to distinct clonal populations. The resulting intra-tumour genetic heterogeneity is associated with an increased risk of drug resistance, metastasis and relapse. In many studies the genome of a tumour is charted by sequencing the DNA that is extracted from millions of cells. Although these bulk methods provide important information, inferring rare but important subclonal populations remains difficult. For the characterization of rare subclones or rare cancer cells, single-cell genomics is required. With the development of single-cell multi-omic methods it also became possible to investigate multiple omic layers of the same cell. For example, single-cell genome-and-transcriptome sequencing (G&T-seq) allows researchers to study both the genome as well as the transcriptome of the same cell, and hence, the consequence of genomic alterations to cellular phenotypes.

At the start of my doctoral training, single-cell DNA sequencing (scDNA-seq) required preamplification of the genome followed by additional fragmentation and sequencing library preparation which resulted in high production costs and low processing throughput. But more alarming was that preamplification, or whole-genome amplification (WGA), introduces biases. While computational methods can reduce some of these biases, imperfections remain and can cause detection of false copy number changes (CNVs) as well as false single nucleotide variants (SNVs) in single-cell genomes. Because of this, the main goal of this dissertation was to develop novel single-cell DNA sequencing methods that improved accuracy, throughput, and cost-effectiveness compared to the existing methods. To this aim we developed a method based on single-cell genomic tagmentation (Gtag), whereby WGA is omitted and whereby DNA sequencing libraries are prepared directly from single-cell lysates. While these fragments resulting from tagmentation are still amplified, their copies can be easily identified and collapsed into unique sequencing reads without redundant overlap and which are evenly distributed along the genome. We implemented Gtag into genome-and-transcriptome sequencing (Gtag&T) and we benchmarked the newly developed methods against conventional single-cell G(&T) sequencing technologies.

We found that Gtag requires less hands-on time, increases the throughput while lowering costs, and demonstrates improved coverage uniformity compared to conventional methods. Because the most commonly used transposase adapters result in the loss of 50% of the tagmented DNA fragments we developed a novel adapter to load onto the Tn5 transposase. This Y-shaped adapter prevents that a single-stranded linear DNA molecule receives the same adapter sequence on both its 5’ and 3’ end which bypasses loss of symmetrical molecules. Although we were able to successfully sequence the DNA of single cells using these novel transposase complexes (saGtagY) we experienced low mappability and a decrease in library complexity compared to Gtag.

Next, we applied G(tag)&T-seq to 930 nuclei extracted from fresh frozen rare breast cancer samples and to 341 cells from a melanoma xenograft model. In both datasets we found that Gtag, besides being less noisy and having less positional bias compared to a conventional scDNA-seq method, enables in silico construction of highly accurate subclonal pseudo-bulk genomes by clustering and merging single cells based on their DNA copy number. These in turn enable more precise breakpoint detection. At the single-cell level, we detect in the melanoma xenograft model focal amplifications that show highly heterogenous patterns in terms of structure, size, and number of copies both per cell and between the different subclones. Moreover, our data suggests differential cell state plasticity and treatment response between cancer subclones as well as a subclonal enrichment of the focal amplicons.

Furthermore, by analyzing 39 blastomeres from 8 human IVF embryos we show that Gtag can be used for the detection of aneuploidies and unbalanced translocations, which may be further developed for clinical application.

Taken together, we show that genomic tagmentation-based methods are an outstanding technology for analyzing the DNA of single cells, especially when performing low-coverage sequencing. While it is applicable to different fields we envision it contributes most when highly heterogeneous tissues, like cancer, are analysed and when a high throughput and high accuracy is required.