Decision letter: The multi-tissue landscape of somatic mtDNA mutations indicates tissue-specific accumulation and removal in aging

Abstract

Article Figures and data Abstract Editor's evaluation Introduction Results Discussion Materials and methods Data availability References Decision letter Author response Article and author information Metrics Abstract Accumulation of somatic mutations in the mitochondrial genome (mtDNA) has long been proposed as a possible mechanism of mitochondrial and tissue dysfunction that occurs during aging. A thorough characterization of age-associated mtDNA somatic mutations has been hampered by the limited ability to detect low-frequency mutations. Here, we used Duplex Sequencing on eight tissues of an aged mouse cohort to detect >89,000 independent somatic mtDNA mutations and show significant tissue-specific increases during aging across all tissues examined which did not correlate with mitochondrial content and tissue function. G→A/C→T substitutions, indicative of replication errors and/or cytidine deamination, were the predominant mutation type across all tissues and increased with age, whereas G→T/C→A substitutions, indicative of oxidative damage, were the second most common mutation type, but did not increase with age regardless of tissue. We also show that clonal expansions of mtDNA mutations with age is tissue- and mutation type-dependent. Unexpectedly, mutations associated with oxidative damage rarely formed clones in any tissue and were significantly reduced in the hearts and kidneys of aged mice treated at late age with elamipretide or nicotinamide mononucleotide. Thus, the lack of accumulation of oxidative damage-linked mutations with age suggests a life-long dynamic clearance of either the oxidative lesions or mtDNA genomes harboring oxidative damage. Editor's evaluation Using the most accurate deep sequencing technology, duplex sequencing, these authors have detected over 89,000 independent somatic mtDNA mutations representing the largest catalog of somatic mtDNA point mutations during aging in a single study. The analysis of these mutations provides compelling evidence to dismiss the idea that reactive oxygen species are a driver of mtDNA mutagenesis, but suggests that ROS may be tissue dependent. These results should provide a fundamental understanding of mitochondrial DNA mutagenesis in aging that should appeal to a broad audience. The novel discovery is the significant presence of transversion mutations (C>A/G>T and C>G/G>C), which previously were assumed almost nonexistent. Moreover, the study finds that, unlike conventional mtDNA mutations, these transversions are not involved in clonal expansion and do not accumulate with age; their relative presence varies very significantly between tissues and can be affected by drug interventions. https://doi.org/10.7554/eLife.83395.sa0 Decision letter Reviews on Sciety eLife's review process Introduction Genetic instability is a hallmark of aging (López-Otín et al., 2013). A mechanistic link between somatic mutations and age-related diseases such as cancer is clear, but their importance in other aging phenotypes, long hypothesized, is poorly understood (Zhang and Vijg, 2018). Recent surveys of non-diseased somatic tissues have shown that mutations are pervasive in the nuclear genome (nDNA), increase with age, and vary considerably between tissues (Abascal et al., 2021; Li et al., 2021). Additionally, these nDNA mutations commonly occur in cancer-associated genes, show evidence of selection and clonal expansion, and may play important roles in tissue regeneration and tumor suppression (Colom et al., 2020; Martincorena et al., 2018; Martincorena et al., 2017; Martincorena et al., 2015; Zhu et al., 2019). Collectively, these studies indicate a growing realization that somatic mutagenesis and clonal dynamics are likely an important determinant of human health during aging. While the accumulation of somatic mutations in the mitochondrial genome (mtDNA) with age has long been documented, the specific nature of their occurrence, and the consequences for aging, have remained unclear (reviewed in Sanchez-Contreras and Kennedy, 2022). In vertebrates, mtDNA is a maternally inherited ~16–17 kb circular DNA molecule encoding 37 genes: 13 essential polypeptides of the electron transport chain (ETC), two ribosomal RNA genes, and 22 tRNAs. Mitochondria are involved in a broad range of crucial processes, including ATP generation via oxidative phosphorylation (OXPHOS), calcium homeostasis, iron-sulfur cluster biogenesis, regulation of apoptosis, and the biosynthesis of a wide variety of small molecules (Kowaltowski, 2000). These processes rely on mitochondria such that disruption of the genetic information encoded in mtDNA by mutation leads to dysfunction of these important processes and subsequently induces disease (Wallace, 1999). Unlike nDNA, mtDNA replication is largely independent of the cell cycle. The higher level of mtDNA replication, the absence of several cellular DNA repair pathways, and the lack of protection from histones results in mtDNA mutation rates ~100–1000× higher than that of nDNA (Khrapko et al., 1997; Marcelino and Thilly, 1999). Moreover, due to the coding density of mtDNA being higher than nDNA (~91% vs. ~1%), the probability that a mutation disrupts protein function is greater. Observational studies have shown that the genetic instability of mtDNA in somatic cells is a fundamental phenotype of aging and may be involved in the pathogenesis of several diseases (reviewed in Larsson, 2010). Collectively, studies examining endogenous mtDNA mutations have shown low levels of G→T/C→A mutations and a preponderance of G→A/C→T and T→C/A→G transitions. This has been interpreted as being contrary to free radical theories of aging by suggesting that reactive oxygen species (ROS) are not the primary driver of mutagenesis in mtDNA (Arbeithuber et al., 2020; Ju et al., 2014; Kennedy et al., 2013; Williams et al., 2013; Zheng et al., 2006). Other notable patterns include an over-abundance of mutations in the mitochondrial Control Region (mCR), an unusual strand bias, a mutational gradient in transition mutations, and a unique trinucleotide mutational signature (Ju et al., 2014; Kennedy et al., 2013; Sanchez-Contreras et al., 2021; Wei et al., 2019). However, while the presence of somatic mtDNA mutations is well documented, a clear causative role in aging remains controversial (reviewed in Sanchez-Contreras and Kennedy, 2022). One reason for this controversy stems from a poor understanding of when, where, and how somatic mtDNA mutations arise during the normal aging process. Most conclusions regarding the accumulation of mtDNA mutations during aging are based on a limited number of experimental models and tissue types, with data largely focused on brain and muscle due to their perceived sensitivity to mitochondrial dysfunction. Only a small number of pan-tissue surveys have been performed (Li et al., 2021; Ma et al., 2018; Samuels et al., 2013). Importantly, most of these prior studies made use of either ‘clone and sequence’ or conventional next-generation sequencing (NGS) to detect mutations. These approaches are technically limited in their ability to detect heteroplasmy below a variant allele fraction (VAF) of 1–2% (reviewed in Salk et al., 2018). The advent of ultrahigh-accuracy sequencing methods has shown that most heteroplasmies are present far below this analytical threshold (Arbeithuber et al., 2020; Kennedy et al., 2013). As such, determining the burden of somatic mtDNA mutations in the context of normal aging lags well behind the efforts focused on nDNA. This is especially pertinent given the heterogeneous nature of tissue decline during aging. Like the nDNA, somatic mutations in mtDNA have been proposed to be under selection (Suen et al., 2010). Cells have evolved several mitochondrial quality control pathways such as removal of damaged mitochondria by mitophagy and fusion/fission to maintain a healthy mitochondrial pool (Youle and Narendra, 2011). The formation and expression of deleterious mtDNA mutations is hypothesized to lead to a loss of mitochondrial membrane potential, mitochondrial dysfunction, and induction of mitophagy. This is a potential mechanism by which cells prevent mtDNA mutations from reaching a phenotypic threshold capable of altering cell homeostasis (Rossignol et al., 2003; Rossignol et al., 1999). Evidence for involvement of quality control machinery in removing somatic mtDNA mutations has been contradictory, with some indicating a clear role for mitophagy and fission/fusion, while other evidence indicates no effect (Chen et al., 2010; Chen et al., 2015; Pickrell et al., 2015; Suen et al., 2010). Thus, the role, if any, of the mitochondrial quality control pathways in targeting mtDNA mutations for removal remains unclear. We and others have previously identified a mitochondrially targeted synthetic peptide, elamipretide (Elam; previously referred to as SS-31 and Bendavia), and the NADH precursor nicotinamide mononucleotide (NMN) as interventions that restore mitochondrial function and tissue homeostasis late in life (reviewed in Yoshino et al., 2018, and Obi et al., 2022). The specific mechanism(s) by which these two compounds ameliorate age-related mitochondrial dysfunction differ. Elam interacts directly with the inner mitochondrial membrane and membrane-associated proteins, stabilizing the mitochondrial ultrastructure and influencing cardiolipin-dependent protein interactions to improve ETC function leading to reduced oxidant production, preservation of membrane potential, and enhanced ATP production (Campbell et al., 2019; Mitchell et al., 2020; Zhang et al., 2020). In contrast, NMN is an NAD+ precursor molecule and acts by elevating NAD+ levels and providing additional substrate for mitochondrial ATP generation (Guan et al., 2017; Martin et al., 2017; Yoshino et al., 2011). Neither intervention is expected to directly alter mtDNA repair mechanisms. Therefore, we sought to test whether these interventions would reduce the prevalence of mtDNA mutations in aged tissues because of their demonstrated ability to improve mitochondrial structure and/or function. We first addressed the relative dearth of high-accuracy data regarding age-related accumulation of mtDNA in mice across multiple tissue types. To that end, we used ultraaccurate Duplex Sequencing (Duplex-Seq) to identify organ-specific mtDNA mutation burden in heart, skeletal muscle, eye, kidney, liver, and brain in naturally aged mice (Kennedy et al., 2014; Schmitt et al., 2012). Intra-animal comparison allowed us to determine whether mtDNA mutation rates differ between organs while still accounting for inter-animal variation. Our findings point to the accumulation of somatic mtDNA mutations being a dynamic and highly tissue-specific process that can be modulated by one or more cellular pathways amenable to small molecule intervention. Results To study the effects of aging on the accumulation of somatic mtDNA mutations across tissues, we used Duplex-Seq to obtain high-accuracy variant information across the entire mtDNA. We examined six different organ systems (heart, kidney, liver, skeletal muscle, brain, and eye) at two different ages (young = 4.5 months; N=5 and old = 26 months; N=6). These two age groups were chosen for their representation of the two extremes of the adult mouse lifespan while mitigating potential confounders related to development, sexual maturation, and survival selection at more advanced ages. These tissues vary on their dependence of mitochondria function and OXPHOS (Fernández-Vizarra et al., 2011). To minimize variation of cell type substructure within tissues between animals, care was taken to isolate similar regions of each organ, as described in the Materials and methods section. In total, we sequenced over 27.9 billion high-accuracy bases, corresponding to a grand mean post-consensus depth of 10,125× for all samples with reasonably uniform coverage among experimental groups and mice, with the exception of the OriL (5160–5191) and several masked regions with high G/C content and/or repetitive sequences (Figure 1—figure supplement 1 and Figure 1—figure supplement 2, Figure 1—figure supplement 2; Supplementary file 1—Table 5). We observed a combined total of 77,017 single-nucleotide variants (SNVs) and 12,031 small insertion/deletions (In/Dels) (≲15 bp in size) across all tissue, age, and intervention groups. Collectively, these data represent the largest collection of somatic mtDNA point mutations obtained in a single study to date and is second only to Lujan et al. in terms of overall In/Del counts (Lujan et al., 2012). A summary of the data for each sample is reported in Supplementary file 1—Table 2 and Supplementary file 2. Frequency of somatic mtDNA mutations increase with age and is tissue-specific To better understand the effects of aging on somatic mtDNA mutations across tissues, we determined the frequency of both SNVs and small In/Dels (≲15 bp) in aged mice. To minimize the contribution of mtDNA mutations that could be either maternally inherited or early clonal expansions established in development, we limited our analysis to mutations occurring at or below a VAF of 1%. In young mice, an initial comparison of the frequency of mtDNA SNVs revealed a mutation frequency on the order of ~1 × 10–6, with low variability between tissues (Figure 1A). Kidney and liver were notable exceptions, exhibiting significantly higher SNV frequencies compared to the other tissues in the young cohort (Figure 1A and B). With age, we observed significant increases in SNV frequency in all tissues we surveyed (Figure 1A). Moreover, mutation frequencies varied considerably between tissues in the aged cohort, with kidney having the highest SNV frequency (6.60±0.56 × 10–6) and heart having the lowest (1.74±0.16 × 10–6) (Figure 1A and C). The observed changes in frequency with age or tissue type did not correlate with differences in mtDNA copy number, as the mtDNA:nDNA ratio did not change with age (Figure 1—figure supplement 3A, B; Supplementary file 1—Table 3). In/Dels were approximately 10-fold less prevalent than SNVs in young mice, with a mean frequency of ~1.5 × 10–7, and virtually no differences between tissues (Figure 1D and E). Like SNVs, In/Dels increased with age in most tissues we surveyed and did not correlate with copy number, but unlike SNVs, they did not significantly differ between tissue types, likely due to the high variability between samples (Figure 1D and F; Figure 1—figure supplement 3C). Figure 1 with 4 supplements see all Download asset Open asset Frequency of somatic mitochondrial genome (mtDNA) mutations increase with age and is tissue-specific. (A) The frequency by which single-nucleotide variants (SNVs) were detected in all sequenced bases in either young (~5-months of age) or old (26-months of age) tissues arranged from highest to lowest SNV frequency in aged mice. (B) The frequency by which DNA insertions or deletions (In/Del) of any size are detected within all sequenced bases either young (~5-months of age) or old (26-months of age) tissues. For (A) and (B), significance between young and old within a tissue was determined by Welch’s t-test. *0.01 < p < 0.05, **0.001 < p < 0.01, ***0.0001 < p < 0.001, ****p<0.0001; error bars = standard deviation of individual data points shown. (C–D) Heatmaps of one-way ANOVA with Tukey’s HSD for significant differences of SNV frequencies between tissues, within either young (C) or old (D) age groups. (E–F) Heatmaps of one-way ANOVA with Tukey’s HSD for significant differences of In/Del frequencies between tissues, within either young (E) or old (F) age groups. Due to mutation burdens being tissue-specific, we considered whether these differences could be driven by variation in the contribution of mitochondrial mutations in leukocytes of circulating blood. To determine this, we analyzed Duplex-Seq in a small subset of tissues from aged mice perfused with PBS to remove the blood. Duplex-Seq of mtDNA from blood collected prior to the perfusion showed that in aged mice, the average frequency of SNV in blood was 3.05±0.15 × 10–6, comparable to the frequency detected in aged hippocampus. Comparisons of perfused (no/low blood) to non-perfused tissues from liver, kidney, skeletal muscle, hippocampus, and cerebellum (the retina, retinal pigmented epithelium [RPE]/choroid, and heart were not sequenced), showed no significant difference in the frequency of SNV mutations (Figure 1—figure supplement 4). Thus, our mutation profiles are likely driven primarily by organ-specific cell types. Although little is known about the kinetics of somatic mtDNA mutation accumulation during aging, they have been reported to increase exponentially during aging in mice (Vermulst et al., 2007). Both this current study and a prior study by Arbeithuber et al. report only two time points each (4.5 months vs. 26 months and 20 days vs. 10 months, respectively), making it impossible to confirm exponential increase in either study (Arbeithuber et al., 2020). However, the combination of our data with the previously published data by Arebiethuber et al. indicates a linear increase in overall mutation frequencies across the lifespan in the three tissue types common to both studies (brain, muscle, and liver). This indicates a likely constant ‘clock-like’ accumulation analogous to what is seen in the nDNA (Abascal et al., 2021; Alexandrov et al., 2015; Arbeithuber et al., 2020; Figure 2). Together, these data demonstrate that mtDNA mutations accumulate at tissue-specific rates during aging and indicate use of a single tissue source to draw broad organism level conclusions regarding the interaction between mtDNA mutations and aging is not scientifically supported. Figure 2 Download asset Open asset Somatic single-nucleotide variant (SNV) mutations increase linearly with age. Linear regression of total SNV mutation frequency vs. age in (A) skeletal muscle, (B) brain, and (C) liver. Black = data from Arbeithuber et al.; purple = data from this study; shaded area = 95% confidence interval of linear regression. Mutation spectra of somatic mtDNA mutations demonstrate tissue-specific distribution of mutation types Previous work by us and others indicates that somatic mtDNA mutations are strongly biased toward transitions (i.e. G→A/C→T and T→C/A→G), with low levels of transversions (Ameur et al., 2011; Arbeithuber et al., 2020; Ju et al., 2014; Kennedy et al., 2013; Pickrell et al., 2015; Williams et al., 2013). Moreover, due to their low prevalence, transversions associated with oxidative lesions (i.e. G→T/C→A and G→C/C→G) have been largely discounted as contributing to age-associated mtDNA mutagenesis (Arbeithuber et al., 2020; Hoekstra et al., 2016; Itsara et al., 2014; Kauppila et al., 2018; Kennedy et al., 2013; Zheng et al., 2006). However, these findings are based on a limited number of tissue types, specifically muscle and brain. Given the wide range of SNV frequencies and known metabolic activities of the tissues we assayed, we examined the mutational spectra for each tissue. Our data show that the overall bias toward G→A/C→T transitions remains broadly true for most tissues, but the extent of this bias varies considerably, with kidney and heart being the notable extremes (Figure 3A). In agreement with prior studies, a single mutation class, G→A/C→T, is the most abundant mutation type and accounts for more than 50% all mutations in most young tissues (Figure 3—figure supplement 1). In contrast, ROS-linked G→T/C→A and G→C/C→G mutations exhibited substantial variation in the level of mutations between tissues. In the central nervous system (CNS) tissues (hippocampus, cerebellum, retina), G→T/C→A and G→C/C→G, combined, accounted for an average of 23% of the total mutation burden (retina = 18%, hippocampus 18%, cerebellum 33%) (Figure 3—figure supplement 1). These data are consistent with prior Duplex-Seq-based studies that focused on neural tissues (Arbeithuber et al., 2020; Hoekstra et al., 2016; Kennedy et al., 2013). In contrast, skeletal muscle and heart in young animals have a relatively high frequency of ROS-linked mutations, with 43% and 66% of all mutations, respectively, resulting from these two types of mutations. This suggests that ROS is a greater source of mtDNA mutagenesis earlier in life and is tissue-dependent. Figure 3 with 3 supplements see all Download asset Open asset Mutation spectra of somatic mitochondrial genome (mtDNA) mutations demonstrate tissue-specific distribution of mutation types. (A) Single-nucleotide variant (SNV) frequency by mutation type for young (~5-months of age) tissues shows that replication/deamination-linked G→A/C→T mutations largely dictate overall SNV mutation burden and predominate in all young tissues except heart. Tissues of the central nervous system: eye retina, brain hippocampus, and brain cerebellum have the lowest frequencies of G→T/C→A and G→C/C→G transversions whereas they are highest in kidney and heart. Heatmaps show adjusted p-value from one-way ANOVA with Tukey’s HSD for significant differences of SNV frequencies between young tissues within each mutation class. (B) SNV frequency by mutation type for old (26-month-old) tissues shows age-specific changes to mutation spectra. Heatmaps show one-way ANOVA with Tukey’s HSD for significant differences of SNV frequencies between old tissues within each mutation class. (C) Fold-change of frequency from young to old age calculated for each tissue and spectra and shown with log2 scaling. Heatmap shows whether fold-change values of old relative to young mice are significantly different from fold-change 0 (no change). K=kidney; L=liver; RC = retinal pigmented epithelium (RPE)/choroid; R=retina; Hi = hippocampus; C=cerebellum; M=skeletal muscle; He = heart. In comparison to the young tissues, mutation loads became more weighted toward transitions across the aged tissues we surveyed (Figure 3B). Significant differences between tissues within mutation classes also became more pronounced (Figure 3B, heatmaps). The fold-increase in most mutation types were remarkably uniform despite significant differences in SNV frequency between them (Figure 3C). Aging led to an average 3.2-fold increase of G→A/C→T and T→C/A→G transitions. Similarly, a significant 2.4-fold increase of T→A /A→T transversions was also observed (Figure 3C). A→C /T→G mutations were not evaluated due to their extreme paucity. In contrast to the other mutation types, G→T/C→A and G→C/C→G mutations did not significantly increase with age (Figure 3C). The mtDNA has been documented to accumulate 8-oxo-dG in a tissue-specific manner (Hamilton et al., 2001). In addition, manipulations and high temperatures during library construction can further increase DNA damage (Ahn and Lee, 2019). It has been noted that DNA damage can, if present at sufficiently high levels, give rise to apparent G→T/C→A mutations in Duplex-Seq data, resulting from shearing induced DNA damage (Abascal et al., 2021; Xiong et al., 2022). We typically control for this class of artifacts by clipping the post-consensus read. However, artifacts have been documented to occur up to 30 cycles into DNA that is highly degraded (Xiong et al., 2022). To investigate the potential for DNA damage to explain the presence of G→T/C→A and G→C/C→G mutations, we performed two analyses. First, we examined the single-strand consensus sequence (SSCS) data that comprise the final high-accuracy duplex consensus sequence (DCS). Apparent variants in SSCS have sequencer-based errors removed and are comprised of both real DNA mutations and PCR-derived errors. Most of the PCR-derived errors are the result of base misincorporation across from DNA damage events, such as 8-oxo-dG and cytidine deamination to uracil (Schmitt et al., 2012). Our data show tissues significantly vary in the frequencies of G→T/C→A and G→C/C→G mutations. Therefore, if ROS-linked mutations were the result of fixation of DNA damage during PCR, then we would expect that the frequency of ROS-linked variants in the SSCS data should mirror the variability that is seen the final DCS data. Plotting the G→T and C→A SSCS frequency reveals that signal from ROS damage is uniform across all tissue types (Figure 3—figure supplement 2). Therefore, in order for false variants in the SSCS data to result in false DCS variants, the rate at which this occurred would need to be tissue-specific. Importantly, we randomized the library preparation and sequencing steps across tissues and ages to avoid batch effects, making it difficult to explain how tissue-level effects would occur. Second, we treated total DNA purified from aged skeletal muscle with formamidopyrimidine DNA glycosylase (Fpg) immediately after sonication. Fpg recognizes damaged purines, including 2,6-diamino-4-hydroxy-5-formamidopyrimidine and 8-oxo-dG, to generate an apurinic (AP) site that it then cleaves via its AP lyase activity, leaving a one nucleotide DNA gap. This effectively prevents the damaged DNA strand from PCR amplifying during library construction, thus preventing fragmented DNA from being able to form a final duplex consensus. Encouragingly, we observed no significant difference in our Fpg-treated and non-treated skeletal muscle samples, indicating that shearing-induced artifacts are unlikely to account for our observations (Figure 3—figure supplement 3). Taken together, our data suggest that ROS damage to the mtDNA is unlikely to explain our observations of tissue-related variability in G→T/C→A and G→C/C→G mutations. Clonal expansion of somatic mtDNA mutations is tissue-specific Mutagenesis has been described as an irreversible process that results in increasing levels of mutations in a population over time, termed ‘Muller’s ratchet’ (Felsenstein, 1974; Muller, 1964). Consequently, absent any compensatory mechanisms, mutations should increase during life. In the case of mtDNA, this should appear as an increase in the burden of apparent heteroplasmies (or clones) within a tissue over time. Importantly, because mtDNA replicates independently of nDNA, apparent heteroplasmies can increase during aging even in the absence of substantial cell proliferation. Moreover, mitochondria are subject to surveillance by mitophagy, which may affect the age-dependent mutational dynamics in tissue-specific ways (Pickles et al., 2018). Expansion of mtDNA mutations sufficient to warrant the term ‘clonal’ has been documented in human tissues, but the prevalence of this phenomenon remains poorly documented in mice (Greaves et al., 2014; Greaves et al., 2010; Greaves et al., 2006; Nekhaeva et al., 2002). The set of tissues we examined comprise a range of varying proliferative and replicative potentials, with heart, brain, and retina being limited, while many kidney and liver cell types proliferate throughout life. Therefore, we sought to determine the tissue-specific burden and dynamics of age-related clonal expansion of mtDNA mutations. We defined a ‘heteroplasmic clone’ as a variant supported by three or more error-corrected reads and then calculated both the frequency and percentage of total mutations corresponding to these clones. We observed considerable tissue-specific variation in the effects of age on the presence of heteroplasmic clones, with all tissues exhibiting a significant increase in the frequency of total heteroplasmic clones with age (Figure 4A and C). Clones in all tissues were distributed relatively uniformly across the mtDNA coding region, but with a striking clustering of variants in the mCR (Figure 4C, green region), consistent with prior reports (Arbeithuber et al., 2020; Kennedy et al., 2013; Sanchez-Contreras et al., 2021). Figure 4 with 1 supplement see all Download asset Open asset Clonal expansion of somatic mitochondrial genome (mtDNA) mutations is tissue-specific. (A) Frequency of mtDNA clones detected in each tissue shows an increase in detection of clones with age in all tissues. Note that y-axes are set for each tissue (N=5 for young; N=6 for old, error = ± standard deviation). (B) Percentage of total mutations found in heteroplasmic clones for each tissues shows that only kidney, liver, and retina have significant increases in relative ‘clonality’ with age. For (A) and (B), significance between young and old within a tissue was determined by Welch’s t-test. *0.01 < p < 0.05, **0.001 < p < 0.01, ***0.0001 < p < 0.001, ****p<0.0001; error bars = standard deviation of individual data points shown. (C) Lollipop plots show the mtDNA genomic location of clonal hetroplasmic mutations in young (top row, blue markers, n=5) and old (bottom row, orange markers, n=6) for each tissue type. Orange = rRNA; dark blue = tRNA; purple = protein coding; green = OriL or mitochondrial control region (mCR). The mutation composition of the clones varied between tissues. In the RPE/choroid, brain, skeletal muscle, and heart, the relative percentage of SNVs found as heteroplasmic clones did not change with age (~2–3% of total SNV). In contrast, kidney, liver, and retina exhibited a disproportionate increase in the number of clonally expanded variants with age. In old kidney, the percentage of SNVs detected as heteroplasmic clones increased to 10.4%, while clones in liver increased from ~1% of SNVs in young to 5.6% in old (Figure 4B). In retina, the percentage of SNVs detected as clones increased from 3.6% in young to 5.6% in old mice. In kidney and liver, the expansion of mtDNA mutations was pervasive across the genome and suggestive of a relationship to the high proliferative and regenerative capacity of these tissues. Retina, however, is a post-mitotic tissue and displayed a very different pattern, with the age-associated increase in clonality being attributed almost entire