Lineage Tracing in Humans Enabled by Mitochondrial Mutations and Single-Cell Genomics

https://www.cell.com/cell/pdf/S0092-8674(19)30055-8.pdf

Leif S. Ludwig, Caleb A. Lareau, Jacob C. Ulirsch, Elena Christian, Christoph Muus, Lauren H. Li, Karin Pelka, Will Ge, Yaara Oren, Alison Brack, Travis Law, Christopher Rodman, Jonathan H. Chen, Genevieve M. Boland, Nir Hacohen, Orit Rozenblatt-Rosen, Martin J. Aryee, Jason D. Buenrostro, Aviv Regev, and Vijay G. Sankaran

INTRODUCTION

• Lineage tracing involves inferring the developmental history of an organism, with respect to its ancestors. Since single cells divide and proliferate, an emerging field is the inference of lineages of single-cells.
• In model organisms, this can be achieved through engineered genetic labels and single-cell RNA sequencing. These two approaches cannot be used together in humans, because of the genetic manipulations required to tag cells with heritable marks.
• Therefore, to date lineage tracing studies in humans have relied on the detection of naturally occurring somatic mutations in the nuclear genome. However, these mutations have high error rates and their detection is costly and difficult to perform at scale.
• The mitochondrial genome provides an attractive target for inferring cellular lineages for several reasons: 
• MtDNA is large enough to show substantial levels of variation
• It is short enough to be cost-effective for targetted sequencing: 18,000 mitochondrial genomes (17k bases) can be sequenced at 100-fold coverage for the same cost as a single nuclear genome (3.2bn bases) at 10-fold coverage. 
• Its mutation rate is reported to be 10-100 times larger than the nuclear genome.
• MtDNA is held in high copy number per cell (100-1000s), therefore less amplification is necessary.
• Mutations in mtDNA often reach a variant allele fraction of ~100% due to partitioning noise, random genetic drift, and faster replication relative to nuclear DNA.
• Existing methods (ATAC-seq and single-cell RNA-seq) can be used to detect mtDNA sequences and genetic variation.

MAIN FACTS OF THE PAPER

• The authors established 65 individual sub-clonal populations, over 8 generation, in an immortalised cell line. They derived subclones (populations of cells derived from a single cell) from the parental colony at each generation, and performed bulk mitochondria individual cells’ l genome sequencing  through ATAC-seq. The authors used high-confidence mtDNA mutations to reconstruct clonal relations between the subpopulations, allowing them to predict the most recent common ancestor with >80% accuracy (See Fig 1C, 1E and 1F).
• Since mtDNA is almost entirely transcribed, the authors hypothesized that single-cell RNA-seq would also be able to detect heteroplasmic mutations in mtDNA. The authors tested 6 protocols and found that full-length scRNA-seq methods showed better coverage of the mitochondrial genome than 3'-end-directed methods, with Smart-Seq2 having the best performance. 
• The authors performed whole-genome sequencing and single-cell RNA-seq simultaneously for single cells using SIDR, finding that several mutations were highly heteroplasmic in RNA, but not present in the genome, suggesting: RNA editing, transcription errors or technical errors in sc-RNA seq (Fig 2B). 
• To investigate inter- and intra-individual heterogeneity in mtDNA mutations, the authors analysed bulk RNA-seq data from 8.8k samples, spanning 49 tissues from at least 25 donors, as well as 426 donors with at least 10 tissues (GTEx project). 
• The authors found 2.7k mutations that were tissue-specific within an individual donor at a minimum of 3% heteroplasmy 
• Typically, ~25% of total mRNA originates from the mitochondrial genome across tissues, although this can be much larger in tissues such as the brain and heart. Tissues with a large proportion of mitochondrial mRNA tend to show very large variability -- see Fig 4B.
• Mitochondrial mutations around 10% are not uncommon across the whole mitochondrial genome (Fig 4D) and somatic mtDNA mutations with levels as low as 5% heteroplasmy can be stably propagated and serve as clonal markers in primary human cells.
• Every tissue had at least one tissue-specific mtDNA mutation across all individual donors, which likely arose via somatic mutation in a tissue-specific manner
• The authors used primary hematopoietic stem cells from two individual donors, and found that the mtDNA mutation profile separates single cells according to their donor of origin, as well as their single-cell-derived colony of origin via highly heteroplasmic mtDNA mutations.
• The authors performed bulk ATAC-seq and scRNA-seq on cells from colorectal adenocarcinoma primary tumor resection. Upon sequencing 238 cells, the authors found 12 distinct clusters of mtDNA mutations, suggesting clonal heterogeneity. 
• The authors provide an improved mutation detection framework, where mutation are first identified through bulk sequencing, and then called in scRNA-seq data. 

CONCLUSION AND OBSERVATIONS

• A potential limitation of inferring cell lineage from mtDNA sequence data comes from horizontal transfer of mtDNA between cells. However, the authors show that horizontal transfer would have to be relatively large to confound their analysis.
• Mapping the phenotypic impact of such genotypic diversity remains an open challenge.
• The authors use techniques for which reads mapping to the  mitochondrial genome are usually considered an unwanted by-product. Using assays focussed on the mitochondrial genomescan reduce costs and increase coverage.