MITO: ROBUST INFERENCE OF MITOCHONDRIAL PHYLOGENIES AND CLONES

Somatic evolution, the process by which cells acquire genetic and epigenetic changes throughout an individual’s lifetime, underlies both normal development and diseases like cancer. Single-cell lineage tracing (scLT) has emerged as a powerful approach to study these cellular dynamics, especially in primary tissues. In this scenario, mtDNA variants (MT-SNVs) have gained special attention recently, due to their low-profiling costs and compatibility with other informative cell-state modalities. This thesis introduces MiTo, an novel toolkit for MT-SNV-based scLT. MiTo provides an integrated pipelines for flexible preprocessing of scLT data, lineage inference, and interactive exploration of MT-SNV-derived phylogenies and clonal structures. MiTo integrates seamlessly with popular single-cell analysis libraries, filling a significant gap in the scLT community. To benchmark MiTo, we generated a new single-cell multi-modal dataset, with simultaneous and longitudinal profiling of gene expression, expressed MT-SNVs, and lentiviral barcode labels. We used this dataset to benchmark several tasks in MT-SNVs based scLT (MT-scLT), demonstrating superior performance of MiTo compared to state-of-the-art tools. Here, we found that informative MT-SNVs spaces may include even rare (i.e., 0.02-0.03 allelic frequency in at least 2 cells, and mean 1.2 1.5 alternative UMIs) detection events, but that statistically sound MT-SNVs genotyping methods are needed to handle the intrinsic noise of single-cell measurement, especially in high-clonal-complexity scenarios. Moreover, we show that MT-SNVs-based-phylogenies (MT-phylogenies) exhibit remarkable robustness to noise, even though the constrained number of available characters limits their resolution. Finally, by tracing clonally-enriched MT-SNVs, we show that multiple sub-clonal lineages within individual clones participate to the metastatic dissemination in Breast Cancer xenografts, implying stronger lineage-dependency of the metastatic phenotype in Breast Cancer that previously thought. In summary, this work highlights opportunities and limitations of MT-scLT, providing novel data analysis tools and benchmarking datasets, and demonstrating the power of scLT to investigate complex cellular dynamics in somatic evolution.