Downloaded from genome.cshlp.org on October 1, 2021 - Published by Cold Spring Harbor Laboratory Press 1 Blood-derived mitochondrial DNA copy number is associated with gene expression across multiple 2 tissues and is predictive for incident neurodegenerative disease 3 4 Stephanie Y. Yang1, Christina A. Castellani1, Ryan J. Longchamps1, Vamsee K. Pillalamarri1, Brian 5 O’Rourke2, Eliseo Guallar3, Dan E. Arking1,2 6 7 8 1McKusick-Nathans Department of Genetic Medicine, Johns Hopkins University School of Medicine, 9 Baltimore, MD 10 2Division of Cardiology, Department of Medicine, Johns Hopkins University School of Medicine, 11 Baltimore, MD 12 3Departments of Epidemiology and Medicine, and Welch Center for Prevention, Epidemiology, and 13 Clinical Research, Johns Hopkins University Bloomberg School of Public Health, Baltimore, MD 14 15 Email addresses: 16 [email protected] 17 [email protected] 18 [email protected] 19 [email protected] 20 [email protected] 21 [email protected] 22 [email protected] 23 24 25 26 27 28 29 30 31 32 33 34 35 *Correspondence and address for reprints to: 36 Dan E. Arking, Ph.D. 37 Johns Hopkins University School of Medicine 38 733 N Broadway Ave 39 Miller Research Building Room 459 40 Baltimore, MD 21205 41 (410) 502-4867 (ph) 42 [email protected] 1 Downloaded from genome.cshlp.org on October 1, 2021 - Published by Cold Spring Harbor Laboratory Press 43 ABSTRACT 44 Mitochondrial DNA copy number (mtDNA-CN) is a proxy for mitochondrial function and is associated 45 with aging-related diseases. However, it is unclear how mtDNA-CN measured in blood can reflect 46 diseases that primarily manifest in other tissues. Using the Genotype-Tissue Expression Project, we 47 interrogated relationships between mtDNA-CN measured in whole blood and gene expression from 48 whole blood and 47 additional tissues in 419 individuals. 49 mtDNA-CN was significantly associated with expression of 700 genes in whole blood, including nuclear 50 genes required for mtDNA replication. Significant enrichment was observed for splicing and ubiquitin- 51 mediated proteolysis pathways, as well as target genes for the mitochondrial transcription factor NRF1. 52 In non-blood tissues, there were more significantly associated genes than expected in 30 tissues, 53 suggesting that global gene expression in those tissues is correlated with blood-derived mtDNA-CN. 54 Neurodegenerative disease pathways were significantly associated in multiple tissues, and in an 55 independent dataset, the UK Biobank we observed that higher mtDNA-CN was significantly associated 56 with lower rates of both prevalent (OR=0.89, CI=0.83;0.96) and incident neurodegenerative disease 57 (HR=0.95, 95% CI= 0.91;0.98). 58 The observation that mtDNA-CN measured in blood is associated with gene expression in other tissues 59 suggests that blood-derived mtDNA-CN can reflect metabolic health across multiple tissues. 60 Identification of key pathways including splicing, RNA binding, and catalysis reinforces the importance of 61 mitochondria in maintaining cellular homeostasis. Finally, validation of the role of mtDNA CN in 62 neurodegenerative disease in a large independent cohort study solidifies the link between blood- 63 derived mtDNA-CN, altered gene expression in multiple tissues, and aging-related disease. 64 65 66 2 Downloaded from genome.cshlp.org on October 1, 2021 - Published by Cold Spring Harbor Laboratory Press 67 INTRODUCTION 68 Mitochondria perform multiple essential metabolic functions including energy production, lipid 69 metabolism, and signaling for apoptosis. Mitochondria possess circular genomes (mtDNA) that are 70 distinct from the nuclear genome. While cells typically only possess two copies of the nuclear genome, 71 they contain 100s to 1000s of mitochondria, and each individual mitochondrion can hold 2-10 copies of 72 mtDNA resulting in wide variation in mtDNA copy number (mtDNA-CN) (Wai et al. 2010). The amount of 73 mtDNA-CN also varies widely across cell types, with higher energy demand cell types typically possessing 74 higher levels of mtDNA-CN (Chabi et al. 2003; Miller et al. 2003; Kelly et al. 2012). Due to the 75 importance of mitochondria in metabolism and energy production, mitochondrial dysfunction plays a 76 role in the etiology of many human diseases (Herst et al. 2017). mtDNA-CN has been shown to be a 77 proxy for mitochondrial function, and is consequently an attractive biomarker due to its ease of 78 measurement (Malik and Czajka 2013; Castellani et al. 2020). Indeed, low levels of mtDNA-CN in 79 peripheral blood have been associated with an increased risk for a number of chronic aging-related 80 diseases including frailty, kidney disease, cardiovascular disease, heart failure, and overall mortality 81 (Ashar et al. 2015; Tin et al. 2016; Ashar et al. 2017; Huang et al. 2016). 82 Crosstalk between the mitochondrial and nuclear genomes is essential for maintaining cellular 83 homeostasis. Many essential mitochondrial proteins are encoded by the nuclear genome, and 84 expression of these nuclear genes must be modified to match mitochondrial activity. Likewise, 85 mitochondrial activity must respond to cellular energy demands. Polymorphisms in the nuclear genome 86 have been associated with changes in mitochondrial gene expression, and mitochondrial genome 87 variation has been associated with changes in nuclear gene expression, suggesting interplay between 88 the two genomes (Ali et al. 2019; Lee et al. 2017b). 89 In cancer cells, mtDNA-CN alters gene expression through modifying DNA methylation (Sun and St John 90 2018; Reznik et al. 2016). Recent work from our lab has shown that mtDNA-CN is also associated with 3 Downloaded from genome.cshlp.org on October 1, 2021 - Published by Cold Spring Harbor Laboratory Press 91 nuclear DNA methylation in noncancer settings (Castellani et al. 2019). Given that DNA methylation can 92 modify gene expression, the current study seeks to explore the potential association between blood- 93 derived mtDNA-CN and gene expression. Past work has shown that mtDNA-CN is associated with gene 94 expression of nuclear-encoded genes in lymphoblast cell lines, but this may not reflect biological 95 processes occurring in other tissues, especially after an extended culturing period (Gibbons et al. 2014). 96 Therefore, we leveraged data from the Genotype-Tissue Expression Project (GTEx), a cross-sectional 97 study with gene expression data from multiple non-diseased postmortem tissues, to examine 98 associations between mtDNA-CN and expression of both nuclear and mitochondrially-encoded genes 99 (Lonsdale et al. 2013). This study aimed to evaluate associations between blood-derived mtDNA-CN and 100 gene expression across multiple tissues, and to follow up on a novel association between 101 neurodegenerative disease and blood-derived mtDNA-CN. 102 103 RESULTS 104 Determination and validation of mtDNA-CN metric 105 mtDNA-CN estimates were generated from whole genome sequences performed on DNA derived from 106 whole blood using the ratio of mitochondrial reads to total aligned reads. As mtDNA-CN is known to be 107 affected by cell type composition, cell counts for samples with available RNA-sequencing data were 108 deconvoluted using gene expression measured in whole blood (Zhang et al. 2017; Aran et al. 2017). We 109 identified a batch effect that resulted in significantly altered mtDNA-CN for individuals sequenced prior 110 to January 2013. Therefore, only individuals sequenced after January 2013 were retained for analysis 111 (Supplemental Fig S1). After quality control, outlier filtering, and normalization of the RNA-sequencing 112 data, 419 individuals remained for analyses (see Methods). 113 To validate mtDNA-CN measurements in the filtered GTEx data, we determined the association between 114 mtDNA-CN and known correlated measures, including age, sex, and neutrophil count (Moore et al. 2018; 4 Downloaded from genome.cshlp.org on October 1, 2021 - Published by Cold Spring Harbor Laboratory Press 115 Zhang et al. 2017; Mengel-From et al. 2014). We observed a significant association with neutrophil 116 count (p=8.4e-05), with higher neutrophil count associated with lower mtDNA-CN. While not statistically 117 significant, effect size estimates between mtDNA-CN and age (p=0.18) and sex (p=0.14) were also in the 118 expected direction, with older individuals and males having lower mtDNA-CN (Supplemental Fig S2). 119 Effect sizes estimates for age and neutrophils were also consistent with prior literature (Longchamps et 120 al. 2020) (Supplemental Table S1). Based on variance explained from previous studies, the current study 121 was only powered to detect a significant effect for neutrophil count. For all downstream analyses, 122 mtDNA-CN was defined as the standardized residual from a linear regression model adjusted for age, 123 sex, cell counts estimated from RNA-seq deconvolution, ischemic time, and cohort (see Methods). 124 125 Association of mtDNA-CN derived from whole blood with gene expression in blood 126 A priori, we expect that mitochondrially encoded gene expression would be positively correlated with 127 mtDNA-CN. Likewise, multiple nuclear encoded genes are involved in the regulation of mtDNA 128 replication, and thus, expression levels of these genes are expected to be correlated with mtDNA-CN 129 (Garcia et al. 2017; Rusecka et al. 2018). We therefore evaluated the associations between mtDNA-CN 130 and expression of these two classes of genes, correcting for cohort, sample ischemic time, genotyping 131 PCs, age, race, and surrogate variables derived from RNA-sequencing data to capture known and hidden 132 confounders (Supplemental Fig S3) (Leek and Storey 2007). 133 To minimize the potential impact of outliers, we performed an inverse normal transformation on both 134 the mtDNA-CN metric and the gene expression values. To evaluate the association between mtDNA-CN 135 and mitochondrial RNA (mtRNA) levels, we used the median gene expression value calculated from 136 scaled expression values across 36 mtDNA-encoded genes that passed expression thresholds (see 137 Methods).