Mitochondrial genome copy number variation across tissues in mice and humans

Abstract

The mammalian mitochondrial genome (mtDNA) is multicopy and its copy number (mtCN) varies widely across tissues, in development and in disease. Here, we systematically catalog this variation by assaying mtCN in 52 human tissues across 952 donors (10,499 samples from the Genotype-Tissue Expression project) and 20 murine tissues using qPCR, capturing 50- and 200-fold variation, respectively. We also estimate per cell mtCN across 173 human cell lines from the Cancer Cell Line Encyclopedia using whole-genome sequencing data and observe >50-fold variation. We then leverage the vast amount of genomics data available for these repositories to credential our resource and uncover mtDNA-related biology. Using already existing proteomics data, we show that variation in mtCN can be predicted by variation in TFAM, histone, and mitochondrial ribosome protein abundance. We also integrate mtCN estimates with the CRISPR gene dependency measurements to find that cell lines with high mtCN are resistant to loss of GPX4, a glutathione phospholipid hydroperoxidase. Our resource captures variation in mtCN across mammalian tissues and should be broadly useful to the research community.

Keywords: GPX4; TFAM; histone; mitochondrial ribosome; mtDNA.

Fig. 1.

MtDNA copy number variation across human and murine tissues. (A) Absolute mtDNA copy number (mtCN) per diploid nuclear genome measured by qPCR in 10,449 human donor-tissue pairs from GTEx. (B) mtCN measured by qPCR in 20 murine tissues and normalized to the tissue with the lowest mtCN. Error bars represent SE from biological replicates (n = 3 kidney, n = 4 all other tissues). (C) Scatter plot of human and murine mtCNs for tissues common across “A” and “B.”

Fig. 2.

MtDNA copy number analysis across cancer cell lines integrated with protein expression and gene dependency scores. (A) mtCN calculated from whole-genome sequencing coverage in 173 cell lines from the CCLE. All estimates are divided by that of the cell line with the lowest mtCN. (B) Scatter plot of mtCN calculated in A (converted to percentile) vs. mtCN percentiles predicted by a stepwise linear regression model. (C) MtCN percentiles were predicted in all 21 human and 19 murine tissues for which protein levels were available. Scatter plots of the measured mtCN from Fig. 1 (converted to percentile) vs. mtCN percentiles predicted by the stepwise linear regression model in (B). Error bars represent SD around the mean mtCN percentile across all donors for each tissue. (D) Spearman correlation between mtCN and Chronos scores of 17,928 gene knockouts across all 139 CCLE cell lines with joint data. (E) Scatter plots of GPX4 Chronos score vs. mtCN rank across lineages and within the two lineages with a significant correlation.