Probabilistic cell lineage inference from mitochondrial mutations

Tissues and organs develop from single founder cells, which give rise to distinct cell lineages that contribute to regeneration and maintenance of homeostasis in the adult. Tracing the genealogical relationships between individual cells and their gene expression signatures is an important step towards understanding how these processes are regulated in human health and disease. Here, we present mt-SCITE, a computational method for inferring the cell lineage tree based on mitochondrial mutations detected in single cells. We show that mt-SCITE outperforms existing methods on simulated data, accurately estimating the correct tree across a wide range of conditions. Against an experimentally defined ground-truth lineage from a serial-subcloning experiment, mt-SCITE recovered clonal structure with high accuracy, outperforming existing methods on clone recovery. To further validate our method, we applied mt-SCITE on in vitro expanded T-cells profiled with bulk ATAC sequencing, where the clonal relationships were determined independently from T-cell receptor sequences. We then used our method to reconstruct the division histories and transcriptional heterogeneity of clonally related CD8+ T-cells from single-cell RNA sequencing data obtained from a healthy human donor. This allowed us to track the in vivo development of the T-cell clones post vaccination and resolve subclonal lineage relationships. Our analysis revealed that T-cell clones can adopt predominantly memory-like or effector-like states, while some exhibit mixed identities. We further showed that mt-SCITE scales to large single-cell datasets with over thousands of cells, where it recovers lineage structure consistent with independent T-cell receptor annotations. Taken together, mt-SCITE robustly identifies clonally related cell populations and resolves subclonal structure within them, providing a broadly applicable approach to study cell lineage development.