LONP-1 and ATFS-1 sustain deleterious heteroplasmy by promoting mtDNA replication in dysfunctional mitochondria.

The accumulation of deleterious mitochondrial DNA (∆mtDNA) causes inherited mitochondrial diseases and ageing-associated decline in mitochondrial functions such as oxidative phosphorylation. Following mitochondrial perturbations, the bZIP protein ATFS-1 induces a transcriptional programme to restore mitochondrial function. Paradoxically, ATFS-1 is also required to maintain ∆mtDNAs in heteroplasmic worms. The mechanism by which ATFS-1 promotes ∆mtDNA accumulation relative to wild-type mtDNAs is unclear. Here we show that ATFS-1 accumulates in dysfunctional mitochondria. ATFS-1 is absent in healthy mitochondria owing to degradation by the mtDNA-bound protease LONP-1, which results in the nearly exclusive association between ATFS-1 and ∆mtDNAs in heteroplasmic worms. Moreover, we demonstrate that mitochondrial ATFS-1 promotes the binding of the mtDNA replicative polymerase (POLG) to ∆mtDNAs. Interestingly, inhibition of the mtDNA-bound protease LONP-1 increased ATFS-1 and POLG binding to wild-type mtDNAs. LONP-1 inhibition in Caenorhabditis elegans and human cybrid cells improved the heteroplasmy ratio and restored oxidative phosphorylation. Our findings suggest that ATFS-1 promotes mtDNA replication in dysfunctional mitochondria by promoting POLG-mtDNA binding, which is antagonized by LONP-1.

INTRODUCTION

Mitochondria provide numerous metabolic functions including being the site of energy production via oxidative phosphorylation (OXPHOS). Most of the ~1200 proteins comprising the mitochondrial proteome are encoded by nuclear genes and are imported into each mitochondrion following synthesis on cytosolic ribosomes[1]. However, 13 essential OXPHOS components alongside the tRNAs and rRNAs required for their synthesis are encoded by mitochondrial genomes (mtDNAs), which reside in the mitochondrial matrix. Each mitochondrion harbors at least one mtDNA, and most cells harbor 100s-1,000s of mtDNAs.

Mitochondrial function declines as cells age which is accelerated in multiple diseases including Parkinson’s Disease. A variety of mitochondrial diseases are caused by inherited mutations that impair OXPHOS function. The disease causing mutations can occur in genes required for OXPHOS encoded by either the nuclear genome, or mtDNAs, which acquire mutations at a significantly higher rate[2]. Single nucleotide variants and deletions are associated with inherited mitochondrial diseases affecting ~1:4000 individuals[3]. Because of the high number of mtDNAs per cell, a single mutant mtDNA has little impact. To cause the OXPHOS dysfunction that underlies mitochondrial diseases, the mutant mtDNA must accumulate to ~60% of the total cellular mtDNAs. The mixture of mutant mtDNAs and wildtype mtDNAs is known as heteroplasmy. Studies using mitochondrial-targeted nucleases that specifically cleave mutant mtDNAs suggest that a relatively modest reduction in the percentage of ΔmtDNAs is sufficient to improve mitochondrial function[4, 5].

The initial mtDNA mutation or deletion likely occurs because of an error in mtDNA replication[6]. Two mechanisms are thought to contribute to the “clonal expansion” of the ΔmtDNA. In dividing cells, non-selective genetic drift can disproportionately increase either genome. An alternative model suggests that large mtDNA deletions allow for quicker replication simply because these genomes are smaller[6, 7]. Consistent with both models, inhibition of the replicative mtDNA polymerase POLG causes preferential depletion of mutant mtDNAs[8,9]. However, the underlying mechanism(s) that confer a replicative advantage and drive the clonal amplification of ΔmtDNAs to a high enough percentage to cause OXPHOS defects in mitochondrial diseases[10-13], aging[14] and Parkinson’s Disease[15] remain unresolved.

Previously, the bZIP protein ATFS-1 was found to be required to maintain deleterious heteroplasmy in C. elegans[9, 16]. ATFS-1 harbors both a mitochondrial targeting sequence (MTS) and a nuclear localization sequence (NLS) (Fig. 1a) and regulates a transcriptional program known as the mitochondrial unfolded protein response (UPRmt)[17]. Under basal conditions, the majority of ATFS-1 is imported into mitochondria, where it is degraded by the protease LONP-1. Mitochondrial dysfunction reduces mitochondrial import capacity, resulting in a percentage of ATFS-1 trafficking to the nucleus, where it activates a transcriptional program to recover mitochondrial function[18]. Importantly, both nuclear and mitochondrial accumulation of ATFS-1 are required for development during mitochondrial dysfunction[9, 17]. However, the function of ATFS-1 within mitochondria is unclear.

Fig. 1 |

OXPHOS dysfunction increases mtDNAs through ATFS-1.

a, ATFS-1/UPRmt signaling schematic in healthy cells. b, Comparison of wildtype and uaDf5 deletion (ΔmtDNA) mtDNAs. c, Photomicrographs of hsp-6::gfp and hsp-6::gfp;uaDf5 worms (Scale bar 0.1 mm). Representative images from four biological repeats. d, Quantification of total mtDNA in homoplasmic wildtype, uaDf5 worms and uaDf5 worms raised on atfs-1(RNAi). n = 3, biologically independent samples. *P = 0.0176, **P = 0.0034. e, Photomicrographs of hsp-6::gfp worms raised on control (ctrl), cco-1 or cyc-1(RNAi). Representative images from four biological repeats. f, Photomicrographs of wildtype, isp-1(qm150) and clk-1(qm30);hsp-6::gfp worms. (Scale bar 0.1 mm). Representative images from four biological repeats. g, Quantification of total mtDNA in homoplasmic wildtype and atfs-1(null) worms raised on control(RNAi), cco-1(RNAi), or cyc-1(RNAi). n = 3, biologically independent samples. *P = 0.0128, **P = 0.002, ***P = 0.0005, ****P < 0.0001. h, Quantification of mtDNA in wildtype, isp-1(qm150), clk-1(qm30) or nduf-2.1(fc21) mutant worms raised on control(RNAi) or atfs-1(RNAi). n = 3 (WT, isp-1(qm150), isp-1(qm150) atfs-1(RNAi)), biologically independent samples. ***P = 0.0001 and 0.0005 (WT vs. isp-1(qm150) and isp-1(qm150) vs. isp-1(qm150) atfs-1(RNAi)). n = 4 (WT, clk-1(qm30)), n = 3 (clk-1(qm30) atfs-1(RNAi)), biologically independent samples. *P = 0.0443, **P = 0.0016 (WT, clk-1(qm30) and clk-1(qm30) atfs-1(RNAi)). n = 3 (WT, nduf-2.1(fc21) and nduf-2.1(fc21) vs. nduf-2.1(fc21) atfs-1(RNAi), biologically independent samples. *P = 0.0151 and 0.011 (WT vs. nduf-2.1(fc21) and nduf-2.1(fc21) vs. nduf-2.1(fc21) atfs-1(RNAi)). Each biologically independent sample contained 40–60 animals and every dot stands for averaged value from 3 technical replicates in d,g and h; One-way ANOVA was used in d,g and h; data shown represent mean ± S.E.M. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

Here, we report that the maintenance of deleterious heteroplasmy requires the accumulation of ATFS-1 within dysfunctional mitochondria. In a heteroplasmic C. elegans strain, ATFS-1 binds predominantly to ΔmtDNAs. Moreover, the replicative polymerase POLG also binds predominantly to ΔmtDNAs. Lastly, we demonstrate that the mitochondrial protease LONP-1, which degrades ATFS-1 in functional mitochondria[17, 19], is required to establish the enriched interaction between ΔmtDNAs, ATFS-1 and POLG in heteroplasmic worms. Our findings in C. elegans are conserved in cultured human cells as inhibition of LONP1 by siRNA or the drug CDDO[20], improves heteroplasmy and rescues OXPHOS function in heteroplasmic cells.

RESULTS

OXPHOS perturbations increase mtDNA quantity

OXPHOS proteins are encoded by genes located within the mitochondria and the nucleus. The heteroplasmic C. elegans strain uaDf5 harbors approximately 40% wildtype mtDNAs and 60% ΔmtDNAs that lack four essential OXPHOS protein coding genes (Fig. 1b)[21]. uaDf5 worms have impaired respiration[9], constitutive UPRmt activation, as determined by increased expression of the hsp-6::gfp reporter[9, 16] (Fig. 1c), and increased total mtDNAs (Fig. 1d)[21].

To examine the impact of OXPHOS perturbation on mtDNA content, we impaired several OXPHOS components in wildtype homoplasmic worms. As expected, worms raised on cco-1(RNAi) (complex IV) or cyc-1(RNAi) (cytochrome c) had increased hsp-6::gfp activation, as did isp-1(qm150) (complex III) or clk-1(qm30) (ubiquinone biosynthesis) mutant strains (Fig. 1e,f). Intriguingly, each of the OXPHOS perturbations resulted in increased mtDNA content as determine by qPCR (Fig. 1g,h and Extended Data Fig. 1a), which is consistent with previous reports[22, 23]. Importantly, the increase in mtDNAs caused by OXPHOS perturbation was impaired in atfs-1(null) worms that lack the entire atfs-1 open reading frame[24] (Fig. 1g) as well as in worms raised on atfs-1(RNAi) (Fig. 1h). Similarly, total mtDNAs in heteroplasmic worms was also reduced when raised on atfs-1(RNAi) (Fig. 1d). Combined, these findings indicate that the increased mtDNA content in both homoplasmic and heteroplasmic worms caused by OXPHOS perturbation requires atfs-1.

Extended Data Fig. 1

OXPHOS dysfunction increases mtDNAs

a, Quantification of total mtDNA in wildtype and nduf-7(et19) worms. n = 3, biologically independent samples (Each sample contains 40–60 animals; every dot stands for averaged value from 3 technical replicates; data shown represent mean ± S.E.M.). **P = 0.0015, Two-tailed Student’s t test. b, POLG immunoblot of wildtype worms following fractionation into total lysate (T), post-mitochondrial supernatant (S), and mitochondrial pellet (M). Tubulin (Tub) and the OXPHOS protein (NDUFS3) serve as loading controls. Representative immunoblots from four biological repeats. c, POLG immunoblot of lysates from wildtype worms raised on control(RNAi) or polg(RNAi). Tubulin (Tub) serves as a loading control. Representative immunoblots from seven biological repeats.

POLG-mtDNA binding during OXPHOS dysfunction requires ATFS-1

We next sought to gain insight into the mechanism by which mtDNAs are increased during OXPHOS perturbation. We previously found that ATFS-1 accumulates within mitochondria when mitochondrial function is perturbed by inhibiting the mitochondrial protease SPG-7 or by raising worms in the presence of ethidium bromide, which impairs mtDNA replication[17, 19]. Intriguingly, ChIP-sequencing indicated that ATFS-1 binds mtDNAs during mitochondrial dysfunction within the non-coding region (NCR)[17], which in mammals contains sequence elements that regulate mtDNA replication[25]. Here, we found that ATFS-1 also accumulates in the mitochondrial fraction of worms raised on cco-1(RNAi) (Fig. 2a). Furthermore, ATFS-1::GFP also accumulates within mitochondria raised on cco-1(RNAi), while TMRE staining was decreased (Fig. 2b) indicative of impaired OXPHOS. Lastly, ATFS-1 ChIP followed by qPCR indicated that ATFS-1 interacts with mtDNA when worms are raised on cco-1(RNAi) (Fig. 2c). Combined these data indicate that ATFS-1 accumulates within mitochondria and interacts with mtDNAs when OXPHOS is impaired.

Fig. 2 |

Defective OXPHOS impedes the degradation of ATFS-1 and facilitates mitochondrial-localized ATFS-1 and POLG binding to mtDNAs.

a, ATFS-1 immunoblots of wildtype worms raised on control or cco-1(RNAi) after fractionation into total lysate (T), post-mitochondrial supernatant (S), and mitochondrial pellet (M). Tubulin (Tub) and the OXPHOS component (NDUFS3) are used as loading controls. Arrow is mitochondrial-localized ATFS-1. Representative immunoblots from three biological repeats. b, atfs-1::atfs-1::gfp transgenic animals raised on control(RNAi) or cco-1(RNAi) with TMRE staining (Scale bar 5μm). Representative images from three biological repeats. c, Quantification of mtDNA following ATFS-1 ChIP-mtDNA in homoplasmic wildtype worms and homoplasmic atfs-1(null) worms raised on control or cco-1(RNAi). n = 4, biologically independent samples. *P = 0.0121. d, POLG immunoblots of wildtype worms raised on control or cco-1(RNAi) after mitochondrial fractionation. Representative immunoblots from three biological repeats. e, Quantification of total mtDNA following POLG ChIP-mtDNA in wildtype or atfs-1(null) homoplasmic worms. n = 3, biologically independent samples. ****P < 0.0001. f, Quantification of mtDNA following ATFS-1 ChIP-mtDNA in homoplasmic atfs-1 worms and homoplasmic atfs-1(null) worms raised on control or cco-1(RNAi). n = 3, biologically independent samples. *P = 0.0363. g, Quantification of total mtDNA following POLG ChIP-mtDNA in homoplasmic atfs-1 and atfs-1(null) worms raised on control or cco-1(RNAi). n = 3, biologically independent samples. *P = 0.0113. h, Quantification of total mtDNA in atfs-1 homoplasmic worms raised on control(RNAi) or cco-1(RNAi). n = 3, biologically independent samples. **P = 0.0033. i, Quantification of total mtDNA in wildtype, clk-1(qm30) and clk-1(qm30);atfs-1 homoplasmic wildtype worms. n = 4, biologically independent samples. **P = 0.0081, ***P = 0.0004, One-way ANOVA. Each biologically independent sample contained about 150,000 animals in c,e-g; contained 40–60 animals in h,i; every dot stands for averaged value from 3 technical replicates); Two-tailed Student’s t test was used in c and e-i; data shown represent mean ± S.E.M.

To further explore the relationship between ATFS-1 accumulation and the mtDNA increase during OXPHOS perturbation, we generated POLG antibodies which detected a ~120 KD band that co-fractionated with the OXPHOS protein NDUFS3 (Extended Data Fig. 1b) and was depleted by polg(RNAi) (Extended Data Fig. 1c). Similar to ATFS-1, POLG protein levels increased when raised on cco-1(RNAi) (Fig. 2d), and POLG interacted with more mtDNAs (Fig. 2e). To determine if the increased POLG-mtDNA interaction required ATFS-1, we performed POLG ChIP-mtDNA in wildtype and atfs-1(null) worms raised on cco-1(RNAi). Interestingly, the increased POLG-mtDNA interaction was impaired in atfs-1(null) worms, suggesting that ATFS-1 is required for increased POLG-mtDNA binding during OXPHOS dysfunction (Fig. 2e).

As atfs-1 is required for the increase in polg mRNA during mitochondrial dysfunction[9, 19], we sought to determine if the increased POLG-mtDNA binding required nuclear-localized ATFS-1. We previously generated the atfs-1 strain in which the NLS within ATFS-1 was impaired via genome editing (Extended Data Fig. 2a,b)[18]. Importantly, the atfs-1 allele impaired hsp-6 mRNA accumulation when raised on spg-7(RNAi) (Extended Data Fig. 2c). Furthermore, the atfs-1 mutation also impaired hsp-6::gfp induction in atfs-1(et18) worms, which constitutively activates UPRmt due to a mutation that reduces its import into mitochondria[18, 26] (Extended Data Fig. 2d–f). Induction of hsp-6 and polg mRNA was also impaired in atfs-1 worms when raised on cco-1(RNAi) (Extended Data Fig. 2g,h). Unlike in wildtype worms, POLG protein was not increased in atfs-1 worms indicating that atfs-1 worms were unable to regulate nuclear transcription during OXPHOS perturbation (Extended Data Fig. 2h,i). Importantly, ATFS-1nuc(−) accumulated within mitochondria similarly to wildtype ATFS-1 upon LONP-1 inhibition (Extended Data Fig. 2j) indicating the protein was expressed and processed similarly to wildtype ATFS-1. Lastly, ATFS-1nuc(−) bound similar amounts of mtDNA as wildtype ATFS-1 when raised on cco-1(RNAi) as determined by ChIP (Fig. 2c,f).

Extended Data Fig. 2

atfs-1-dependent transcription is impaired in atfs-1 worms.

a, Schematic highlighting the R (Arginine) to A (Alanine) substitution to impair the nuclear localization sequence (NLS) within ATFS-1 yielding ATFS-1nuc(−) confirmed by Sanger sequencing. b, UPRmt signaling schematic highlighting the ATFS-1nuc(−) with an impaired NLS. c, Expression level of hsp-6 mRNA in wildtype and atfs-1 worms raised on control(RNAi) or spg-7(RNAi) examined by qRT-PCR. n = 3, biologically independent samples. **P = 0.0028, One-way ANOVA. d-e, Photomicrographs of wildtype, atfs-1(et18) and atfs-1(et18);hsp-6::gfp worms (Scale bar 0.1 mm) (d); Quantification of fluorescence pixel intensity in wildtype (n = 117; Max: 12.51; Min: 1.87; Median: 5.12), atfs-1(et18) (n = 74; Max: 75.251; Min: 15.790; Median: 32.021) and atfs-1(et18) strains (n = 121; Max: 15.100; Min: 1.55; Median: 5.43). Box & whiskers plots Min to Max. ****P < 0.0001, One-way ANOVA. n means the number of sampling areas. Average pixel intensity signals were calculated from sampling areas at each condition in biological triplicates (e). f, hsp-6 mRNA expression in wildtype, atfs-1(et18) or atfs-1(et18) worms examined by qRT-PCR. n = 4 (wildtype), n = 5 (atfs-1(et18)), n = 3 (atfs-1(et18)), biologically independent samples. *P = 0.0114 (wildtype vs. atfs-1(et18)), *P = 0.0407 (atfs-1(et18) vs. atfs-1(et18)), one-way ANOVA. g, hsp-6 mRNA expression in wildtype and atfs-1 worms raised on control(RNAi) or cco-1(RNAi) examined by qRT-PCR. n = 3, biologically independent samples. ***P = 0.0004, one-way ANOVA. h, polg mRNA expression in atfs-1 worms raised on control(RNAi) or cco-1(RNAi) examined by qRT-PCR. n = 3, biologically independent samples. Two-tailed Student’s t test. i, POLG immunoblots of lysates from wildtype, atfs-1 and atfs-1(null) worms raised on control or cco-1(RNAi). Representative immunoblots from four biological repeats. j, Immunoblots of lysates from wildtype and atfs-1 worms raised on control or lonp-1(RNAi). ATFS-1 or ATFS-1nuc(−) are indicated with an arrowhead. Representative immunoblots from four biological repeats. In c, f-h, each dot represents the average from 3 technical replicates; data shown represent mean ± S.E.M.

To determine if the nuclear function of ATFS-1 is required for POLG to bind mtDNAs during OXPHOS perturbation, we examined the POLG-mtDNA interaction in atfs-1(null) and atfs-1 worms raised on control or cco-1(RNAi). Strikingly, significantly more POLG was bound to mtDNA in atfs-1 worms relative to atfs-1(null) worms (Fig. 2g), suggesting that the atfs-1-dependent increase in POLG-mtDNA binding during OXPHOS perturbation does not require the nuclear activity of ATFS-1. mtDNA content also increased in atfs-1 worms upon OXPHOS perturbation caused by cco-1(RNAi) (Fig. 2h) as well as in clk-1(qm30) worms (Fig. 2i). Combined, these results suggest that the accumulation of ATFS-1 within mitochondria during OXPHOS perturbation is required to increase POLG-mtDNA binding and mtDNA content.

ATFS-1 degradation by LONP-1impairs mtDNA propagation

We next sought to determine the events leading to the accumulation of ATFS-1 within mitochondria upon OXPHOS perturbation. The ATP-dependent protease LONP-1 degrades the majority of ATFS-1 once the bZIP protein has been imported into the mitochondrial matrix[19]. LONP-1 is an ATP-dependent protease known to recognize and degrade mitochondrial proteins damaged by reactive oxygen species (ROS)[27]. LONP-1 has also been shown to interact with mtDNA in diverse species[28, 29], and regulate mtDNA replication[30, 31].

To further examine the interaction between LONP-1 and mtDNA in C. elegans, we generated a strain in which the C-terminus of LONP-1 was tagged with the FLAG epitope via genome editing (Fig. 3a and Extended Data Fig. 3a). Introduction of the FLAG epitope did not impair worm development or cause UPRmt activation, suggesting it did not adversely affect LONP-1 function (Extended Data Fig. 3b,c). As expected, LONP-1 interacted with mtDNA in C. elegans as determined by LONP-1FLAG ChIP-mtDNA qPCR (Fig. 3b). We next examined where in mtDNA LONP-1FLAG binds in wildtype homoplasmic worms. LONP-1FLAG ChIP-seq indicated that the protease was enriched at several G-rich sites throughout mtDNA (Fig. 3c), but was especially enriched within the NCR (Fig. 3d). Interestingly, the strongest LONP-1FLAG peak within the NCR overlapped with the ATFS-1 binding site (Fig. 3d and Extended Data Fig. 3d) suggesting a potential role in mediating the ATFS-1-mtDNA and POLG-mtDNA interactions.

Fig. 3 |

LONP-1 limits ATFS-1 binding to wildtype mtDNAs and impairs replication.

a, FLAG immunoblots of LONP-1FLAG worms following fractionation into total lysate (T), post-mitochondrial supernatant (S), and mitochondrial pellet (M). Tubulin (Tub) and the OXPHOS component NDUFS3 are loading controls. Representative immunoblots from four biological repeats. b, Quantification of mtDNA from homoplasmic LONP-1FLAG worms following ChIP-mtDNA using FLAG or control (Mock) antibody. n = 3, biologically independent samples. **P = 0.0018. c, LONP-1 consensus binding motif within mtDNA. d, ChIP-seq profile of mtDNA from homoplasmic LONP-1FLAG worms raised on control(RNAi) using FLAG antibody (red). e, Quantification of total mtDNA following ATFS-1 ChIP-mtDNA in wildtype or atfs-1(null) homoplasmic worms raised on control(RNAi) or lonp-1(RNAi). n = 3, biologically independent samples. *P = 0.0132. f, Quantification of mtDNA in homoplasmic wildtype and atfs-1(null) worms raised on control(RNAi) or lonp-1(RNAi). n = 4, biologically independent samples. *P < 0.05. g-h, atfs-1::atfs-1::gfp transgenic animals raised on control(RNAi) or lonp-1(RNAi) with TMRE staining (Scale bar 5μm) (g); Percentage of mitochondria with co-localized TMRE and ATFS-1::GFP. n = 10 ctrl(RNAi) and n = 11 lonp-1(RNAi), biologically independent samples. ****P < 0.0001 (h). i, Quantification of total mtDNA following POLG ChIP-mtDNA in wildtype homoplasmic worms raised on control(RNAi) or lonp-1(RNAi). n = 3, biologically independent samples. **P = 0.0012. j, Schematic of the relationship between LONP-1 activity, mitochondrial ATFS-1 accumulation and mtDNA replication. Each biologically independent sample contained about 150,000 worms in b,e,i; each sample contained 40–60 animals in f; every dot stands for averaged value from 3 technical replicates in b,e,f and i; Two-tailed Student’s t test was used; data shown represent mean ± S.E.M.

Extended Data Fig. 3

LONP-1 inhibition promotes mtDNA content via ATFS-1

a, FLAG immunoblots of lysates from wildtype and LONP-1FLAG wildtype worms. Tubulin (Tub) serves as a loading control. Representative immunoblots from four biological repeats. b, Images of wildtype or LONP-1FLAG worms 48 hours after synchronization indicating worms expressing LONP-1FLAG at the endogenous locus develop normally (Scale bar 1 mm). Representative images from four biological repeats. c, Fluorescent photomicrographs of wildtype hsp-6::gfp or lonp-1;hsp-6::gfp worms 48 hours after synchronization indicating worms expressing LONP-1FLAG do not cause UPRmt activation (Scale bar 0.05 mm). Representative images from four biological repeats. d, Schematic of the putative ATFS-1 and LONP-1 binding sites within the mtDNA non-coding region (NCR) highlighting the proximity of both sites (~200 base pairs). e, POLG Immunoblots of lysates from wildtype worms raised on control or lonp-1(RNAi). Representative images from four biological repeats. f, Total mtDNA quantification in wildtype homoplasmic atfs-1 worms raised on control(RNAi) or lonp-1(RNAi). n = 5, biologically independent samples. ***P = 0.0004, Two-tailed Student’s t test). g, Total mtDNA quantification in wildtype homoplasmic atfs-1 worms raised on control(RNAi) or lonp-1(RNAi). n = 3, biologically independent samples. Two-tailed Student’s t test. In f and g, each biologically independent sample contained 40–60 animals; every dot stands for averaged value from 3 technical replicates; data shown represent mean ± S.E.M. *p<0.05, **p<0.01, ****p<0.0001.

We next examined the impact of lonp-1(RNAi) on mtDNA accumulation in homoplasmic wildtype worms. When raised on lonp-1(RNAi), the binding of ATFS-1 to mtDNA increased by ~5-fold (Fig. 3e) which correlated with an increase in mtDNA content (Fig. 3f). ATFS-1::GFP also accumulates within mitochondria raised on lonp-1(RNAi) (Fig. 3g,h). Importantly, exposure to lonp-1(RNAi) also increased the binding of POLG to mtDNA (Fig. 3i), consistent with increased POLG accumulation (Extended Data Fig. 3e). Moreover, LONP-1 inhibition increased mtDNA content in atfs-1 worms (Extended Data Fig. 3f), but not in atfs-1(null) worms (Fig. 3f).

To examine the role of mitochondrial-localized ATFS-1 in increasing mtDNA quantity, we sought to generate a strain lacking the MTS. Unfortunately, we were unable to establish a stable atfs-1 line via genome editing. Thus, we generated atfs-1 worms. Importantly, mtDNA copy number was not increased in atfs-1 worms raised on lonp-1(RNAi) (Extended Data Fig. 3g) These findings support a role for mitochondrial-localized ATFS-1 in promoting mtDNA replication which is impaired by LONP-1-dependent degradation (Fig. 3j).

Nuclear ATFS-1 is not required to maintain heteroplasmy

We previously found that ATFS-1 is required to maintain ΔmtDNAs[9, 16] in a heteroplasmic worm strain using atfs-1(RNAi) and a hypomorphic atfs-1 allele. Here, we crossed the atfs-1(null) allele into uaDf5 heteroplasmic worms. Impressively, atfs-1(null) worms were unable to maintain any ΔmtDNAs (Fig. 4a). The loss of ΔmtDNAs in the absence of atfs-1 may be due to increased mitophagy of mitochondria harboring ΔmtDNAs[16, 32], decreased replication of ΔmtDNAs, or a combination. To examine the role of mitophagy, we generated heteroplasmic strains lacking the mitophagy component Parkin (PDR-1 in worms) (Extended Data Fig. 4a). As expected, pdr-1-deficient worms had increased ΔmtDNAs relative to wildtype worms consistent with mitophagy limiting the accumulation of ΔmtDNAs[9, 16]. atfs-1(null);pdr-1(tm598) worms also had significantly less ΔmtDNAs than pdr-1(tm598) worms, indicating that ATFS-1 promotes heteroplasmy via a mechanism independent of mitophagy (Fig. 4a).

Fig. 4 |

ATFS-1 and POLG primarily interact with ΔmtDNAs in heteroplasmic worms.

a, ΔmtDNA quantification as determined by qPCR in uaDf5 worms, atfs-1(null);uaDf5, pdr-1(tm598);uaDf5 and atfs-1(null);pdr-1(tm598);uaDf5 worms. n = 4 (uaDf5 and atfs-1(null);pdr-1(tm598);uaDf5) and n = 3 (atfs-1(null);uaDf5 and pdr-1(tm598);uaDf5) biologically independent samples. **P < 0.01 and ****P < 0.0001, One-way ANOVA. b, Images of TMRE-stained micrographs of heteroplasmic (ΔmtDNA) worms raised on control(RNAi), or wildtype worms raised on control or spg-7(RNAi). Scale bar, 10 μM. Representative images from four biological repeats. c, Immunoblots of wildtype worms raised on control or spg-7(RNAi) and heteroplasmic (ΔmtDNA) worms raised on control(RNAi) after fractionation into total lysate (T), post-mitochondrial supernatant (S), and mitochondrial pellet (M). Tubulin (Tub) and the OXPHOS component (NDUFS3) are used as loading controls. Arrow is mitochondrial-localized ATFS-1. Representative immunoblots from three biological repeats. d, atfs-1::atfs-1::gfp or atfs-1::atfs-1::gfp;uaDf5 transgenic animals with TMRE staining (Scale bar 5μm). Representative images from six biological repeats. e, Quantification of total mtDNA following ATFS-1 ChIP-mtDNA in homoplasmic wildtype or ΔmtDNA worms. n = 3 wildtype and n = 5 uaDf5 biologically independent samples. *P = 0.0226, two-tailed Student’s t test. f, Workflow of ATFS-1 or POLG ChIP-mtDNA and quantification of wildtype mtDNA and ΔmtDNA in heteroplasmic worms. g, Quantification of wildtype mtDNA and ΔmtDNA by qPCR following ATFS-1 ChIP-mtDNA in heteroplasmic worms. Post-lysis/Input ΔmtDNA ratio was 60%. n = 4, biologically independent samples. h, Quantification of wildtype mtDNA and ΔmtDNA by qPCR following POLG ChIP-mtDNA in heteroplasmic worms. Post-lysis/Input ΔmtDNA ratio was 54%. n = 3, biologically independent samples. i, Quantification of wildtype mtDNA and ΔmtDNA following TFAM IP-mtDNA in heteroplasmic worms. Post-lysis/Input ΔmtDNA ratio was 53%. n = 4, biologically independent samples. Each biologically independent sample contained 40–60 animals in a; each biologically independent sample contained about 150,000 worms in e,g-i; every dot stands for averaged value from 3 technical replicates in a,e and g-i; data shown represent mean ± S.E.M.

Extended Data Fig. 4

Mitochondrial ATFS-1 is required to maintain ΔmtDNA in heteroplasmic worms.

a, Crossing strategy of atfs-1(null);pdr-1(tm598);uaDf5 strain. b. TMRE quantification of heteroplasmic (ΔmtDNA) worms raised on control(RNAi) (n = 475; Max: 1.052; Min: 0.21; Median: 0.618), or wildtype worms raised on control (n = 232; Max: 1.318; Min: 0.725; Median: 0.995) or spg-7(RNAi) (n = 114; Max: 0.798; Min: 0.134; Median: 0.402). Box & whiskers plots Min to Max. n means the number of sampling areas. Average pixel intensity signals were calculated from sampling areas at each condition in biological triplicates. c,d, Photomicrographs of uaDf5 and atfs-1;uaDf5;hsp-6::gfp worms (Scale bar 0.1 mm) (c); Quantification of fluorescence pixel intensity in uaDf5 (n = 199; Max: 30.89; Min: 3.430; Median:11.590) and atfs-1;uaDf5;hsp-6::gfp (n = 234; Max: 15.640; Min: 2.540; Median: 6.915). Box & whiskers plots Min to Max. n means the number of sampling areas. Average pixel intensity signals were calculated from sampling areas at each condition in biological triplicates (d). e, ΔmtDNA quantification as determined by qPCR in heteroplasmic uaDf5 worms, atfs-1(null);uaDf5 worms and atfs-1;uaDf5 worms. n = 3, biologically independent samples. f, ΔmtDNA quantification as determined by qPCR in heteroplasmic atfs-1;uaDf5 worms and atfs-1;uaDf5. n = 3 (atfs-1;uaDf5) and n = 4 (atfs-1;uaDf5), biologically independent samples. ***P = 0.0007. g, Quantification of total mtDNA following POLG ChIP-mtDNA in homoplasmic wildtype or uaDf5 worms. n = 4 (wildtype) and n = 3 (uaDf5), biologically independent samples. *P = 0.0229. In e and f, each biologically independent sample contained 40–60 animals; in g, each biologically independent sample contained about 150,000 animals; each dot stands for averaged value from 3 technical replicates in f,g; Two-tailed Student’s t test was used in d, f and g, One-way ANOVA was used in b; data shown represent mean ± S.E.M. *p<0.05, **p<0.01, ****p<0.0001.

As OXPHOS perturbation increased mtDNA content in homoplasmic wildtype worms via ATFS-1, we hypothesized that OXPHOS dysfunction may contribute to the increased mtDNA content in heteroplasmic worms (Fig. 1d). We previously found that heteroplasmic worms consumed less oxygen than wildtype worms[9]. To further evaluate mitochondrial function, we examined mitochondrial membrane potential by staining with TMRE. As expected, TMRE staining was decreased in heteroplasmic worms relative to wildtype worms, but stronger than in cco-1(RNAi)- or spg-7(RNAi)-treated worms consistent with intermediate OXPHOS function (Fig. 2b, 4b and Extended Data Fig. 4b). Importantly, ATFS-1 also accumulated within the mitochondrial fraction of heteroplasmic worms, as in worms raised on spg-7(RNAi) or cco-1(RNAi) (Fig. 4c, also see Fig. 2a). Similar to homoplasmic worms raised on cco-1(RNAi), ATFS-1::GFP accumulated within mitochondria with reduced TMRE staining in heteroplasmic worms. However, numerous mitochondria were TMRE positive, but the majority of those mitochondria lacked ATFS-1::GFP consistent with ATFS-1 being degraded in functional mitochondria (Fig. 4d). Combined, these results suggest that ATFS-1 accumulates within dysfunctional mitochondria caused by impairment of either a nuclear-encoded OXPHOS component, a nuclear-encoded mitochondrial protease, or in worms harboring ΔmtDNAs.

We next examined if maintenance of ΔmtDNAs required the nuclear activity of ATFS-1. Consistent with impaired nuclear activity, hsp-6::gfp was not increased in heteroplasmic atfs-1 worms (Extended Data Fig. 4c,d). Impressively, unlike atfs-1(null) worms, atfs-1 worms were able to maintain ΔmtDNAs although fewer ΔmtDNAs than wildtype atfs-1 worms (Extended Data Fig. 4e). To examine the role of mitochondrial-localized ATFS-1 during heteroplasmy, we crossed the atfs-1 allele into heteroplasmic worms. atfs-1 worms harbored fewer ΔmtDNAs than atfs-1 worms emphasizing the requirement for mitochondrial localized ATFS-1 to maintain heteroplasmy (Extended Data Fig. 4f).

Consistent with the accumulation of ATFS-1 in dysfunctional mitochondria (Fig. 2a and 4b–d), ATFS-1 bound more total mtDNAs in heteroplasmic worms than in wildtype homoplasmic worms as determined by ChIP-mtDNA (Fig. 4e). Similarly, POLG also interacted with more total mtDNAs in heteroplasmic worms than in wildtype homoplasmic worms suggesting increased mtDNA replication (Extended Data Fig. 4g). Combined, these findings suggest a role for mitochondrial-localized ATFS-1 in maintaining deleterious mtDNA heteroplasmy.

ATFS-1 and POLG primarily bind ΔmtDNAs

We next sought to determine if ATFS-1 differentially interacted with each genome. The interaction between ATFS-1 and wildtype mtDNAs or ΔmtDNAs was evaluated via qPCR or 3D digital PCR following ATFS-1 ChIP (Fig. 4f)[9]. As before, qPCR of mtDNA from heteroplasmic whole worm lysate indicated the strain harbored ~60% ΔmtDNAs and ~40% wildtype mtDNAs (Fig. 1b and Extended Data Fig. 5a). qPCR following ATFS-1 ChIP indicated that of the mtDNAs that interacted with ATFS-1, 90% were ΔmtDNAs and 10% were wildtype mtDNAs, indicating that ATFS-1 is significantly enriched on ΔmtDNAs (Fig. 4g and Extended Data Fig. 5b,c).

Extended Data Fig. 5

ATFS-1 and POLG primarily interact with ΔmtDNAs in heteroplasmic worms.

a, Overview of the qPCR strategy to quantify the ΔmtDNA percentage in heteroplasmic worms or heteroplasmic cells. Plasmids containing a sequence specific to the ΔmtDNA or wildtype mtDNA were created[20]. Standard curves were generated using the indicated concentration of each plasmid harboring sequences specific to either wildtype or ΔmtDNAs. Both PCR reactions were carried out simultaneously in the same qPCR machine. b, Scatter plots (b) and results (c) of 3D digital PCR quantification of wildtype mtDNA and ΔmtDNA following ATFS-1 ChIP-mtDNA in heteroplasmic uaDf5 worms. n = 4, biologically independent samples. d-e, Scatter plots (d) and results (e) of 3D digital PCR quantification of wildtype mtDNA and ΔmtDNA following POLG ChIP-mtDNA in heteroplasmic uaDf5 worms. n = 4, biologically independent samples. f, HMG-5/TFAM immunoblot of wildtype worms following fractionation into total lysate (T), post-mitochondrial supernatant (S), and mitochondrial pellet (M). Tubulin (Tub) and the OXPHOS component (NDUFS3) serve as loading controls. Representative immunoblots from two biological repeats. g, HMG-5/TFAM immunoblots of lysates from wildtype worms raised on control or hmg-5/tfam(RNAi). Tubulin (Tub) serves as a loading control. Representative immunoblots from three biological repeats. Each biologically independent sample contained 150,000 animals in c,e; data shown represent mean ± S.E.M.

Previously, we found that inhibition of POLG caused depletion of ΔmtDNAs in heteroplasmic worms relative to wildtype mtDNAs[9] similar to findings in a Drosophila heteroplasmy model[8]. Both findings suggest increased replication of ΔmtDNAs in heteroplasmic cells. To further explore the relationship between ATFS-1 and mtDNA replication, we performed ChIP-mtDNA using the POLG antibody. Interestingly, POLG ChIP indicated that the replicative polymerase also interacted with ~90% ΔmtDNAs and 10% wildtype mtDNAs (Fig. 4h and Extended Data Fig. 5d,e), similarly to ATFS-1 (Fig. 4g). As a control, we generated antibodies to the mtDNA packaging protein HMG-5 (TFAM in mammals) (Extended Data Fig. 5f,g), which interacts with mtDNAs independent of replication[33]. In contrast to ATFS-1 and POLG, the percentage of ΔmtDNAs bound to HMG-5 reflected the percentage within the whole worm lysate (Fig. 4i). Combined, these data indicate that ATFS-1 and a component of the replisome are enriched on ΔmtDNAs in heteroplasmic worms, consistent with the mutant mtDNA having a replicative advantage.

The protease LONP-1 is required to maintain heteroplasmy

Because LONP-1 binds mtDNA and its proteolytic activity limits ATFS-1 accumulation in functional mitochondria[19] (Fig. 3d,e), we examined the role of LONP-1 in heteroplasmy maintenance. We generated antibodies to C. elegans LONP-1 that recognized a ~130 KD band that was reduced when raised on lonp-1(RNAi) (Extended Data Fig. 6a). Via ChIP-qPCR, we found that LONP-1 binds ~60 fold more mtDNAs than ATFS-1 in wildtype worms (Extended Data Fig. 6b). In heteroplasmic worms, LONP-1 bound similar percentages of wildtype and ΔmtDNAs suggesting that LONP-1 interacts with mtDNAs independent of mitochondrial dysfunction (Fig. 5a and Extended Data Fig. 6c). Combined, these data suggest that LONP-1 is constitutively bound to mtDNAs and heteroplasmy is not maintained by uneven mtDNA binding by the protease.

Extended Data Fig. 6

Inhibition of LONP-1 improves the deleterious heteroplasmy ratio

a, LONP-1 immunoblots of lysates from wildtype worms raised on control(RNAi) or lonp-1(RNAi). Tubulin (Tub) serves as a loading control. Representative immunoblots from four biological repeats. b, ChIP-mtDNA using ATFS-1 or LONP-1 antibodies in wildtype worms followed by quantification of total mtDNA. n = 3, biologically independent samples. **P = 0.0042. c, ChIP-mtDNA using LONP-1 antibodies in wildtype or heteroplasmic worms followed by quantification of total mtDNA (both wildtype and ΔmtDNA). n = 3, biologically independent samples. d, ΔmtDNA quantification in atfs-1;uaDf5 worms raised on control(RNAi) or lonp-1(RNAi). n = 3, biologically independent samples. *P = 0.0168. e, The brood size of heteroplasmic worms raised on control or lonp-1(RNAi). n = 9 worms. f, ΔmtDNA and wildtype mtDNA quantification following HMG-5/TFAM ChIP-mtDNA in uaDf5 heteroplasmic worms raised on lonp-1(RNAi) indicating that the binding of HMG-5/TFAM to wildtype mtDNAs or ΔmtDNAs is similar the input ratio. n = 4, biologically independent samples. g, wildtype mtDNA quantification in uaDf5 heteroplasmic worms raised on control(RNAi) or cco-1(RNAi). n = 3, biologically independent samples. **P = 0.0075. h, wildtype mtDNA quantification in uaDf5 or clk-1(qm30);uaDf5 heteroplasmic worms. n = 3, biologically independent samples. **P = 0.0029. In b,c and f, Each biologically independent sample contained 150,000 animals; in d,g,h each biologically independent sample contained 40–60 animals; every dot stands for averaged value from 3 technical replicates in b-d and f-h; Two-tailed Student’s t test was used; data shown represent mean ± S.E.M.

Fig. 5 |

LONP-1 is required to maintain heteroplasmy.

a, ΔmtDNA and wildtype mtDNA quantification by qPCR following LONP-1 ChIP-mtDNA in heteroplasmic worms. Post-lysis/Input ΔmtDNA ratio was 59%. n = 3, biologically independent samples. b, Quantification of wildtype mtDNA by qPCR in heteroplasmic worms raised on control(RNAi) or lonp-1(RNAi). n = 3, biologically independent samples. **P = 0.0041. c, Quantification of ΔmtDNA by qPCR in heteroplasmic uaDf5 worms raised on control(RNAi) or lonp-1(RNAi). n = 3, biologically independent samples. ****P < 0.0001. d, ΔmtDNA and wild-type mtDNA quantification by qPCR following ATFS-1 ChIP-mtDNA in heteroplasmic worms raised on control(RNAi) or lonp-1(RNAi). n = 4, biologically independent samples. e-f, atfs-1::atfs-1::gfp;uaDf5 transgenic animals raised on control(RNAi) or lonp-1(RNAi) with TMRE staining (Scale bar 5μm) (e); Percentage of mitochondria with co-localized TMRE and ATFS-1::GFP. n = 8 uaDf5 ctrl(RNAi) and n = 9 uaDf5 lonp-1(RNAi), biologically independent samples (f). g, ΔmtDNA and wild-type mtDNA quantification by qPCR following POLG ChIP-mtDNA in heteroplasmic worms raised on lonp-1(RNAi). Post-lysis/Input ΔmtDNA ratio was 25%. n = 4, biologically independent samples. Each biologically independent sample contained about 150,000 animals in a,d,g; each biological replicate contained 40–60 animals in b,c; every dot stands for averaged value from 3 technical replicates in a-d and g; Two-tailed Student’s t test was used in b,c and f; data shown represent mean ± S.E.M.

We next examined the effect of inhibiting LONP-1 on heteroplasmy. Interestingly, lonp-1 inhibition via RNAi increased wildtype mtDNAs (Fig. 5b) and reduced ΔmtDNAs, which improved the heteroplasmy ratio from 59% ΔmtDNAs to 25% (Fig. 5c). Similar results were obtained in atfs-1 worms upon LONP-1 inhibition (Extended Data Fig. 6d). Importantly, lonp-1(RNAi) did not reduce the brood size of heteroplasmic worms (Extended Data Fig. 6e), suggesting that lonp-1(RNAi) did not select against embryos with dysfunctional mitochondria due to high levels of ΔmtDNAs.

We next examined the interaction between ATFS-1 and mtDNAs in heteroplasmic worms upon lonp-1 inhibition. Interestingly, when raised on lonp-1(RNAi), the percentage of ΔmtDNAs and wildtype mtDNAs bound to ATFS-1 nearly reflected the heteroplasmy percentage within the whole worm lysate (Fig. 5d). Consistent with the increased binding of ATFS-1 to wildtype mtDNAs, lonp-1(RNAi) increased ATFS-1::GFP accumulation within functional mitochondria (Fig. 5e,f). Thus, the mtDNA-bound protease is required to establish the enriched interaction between ATFS-1 and ΔmtDNAs.

In addition to increasing the percentage of wildtype mtDNAs bound by ATFS-1, lonp-1(RNAi) also increased the percentage of POLG that interacted with wildtype mtDNAs (Fig. 5g, also see Fig. 4h). However, lonp-1(RNAi) did not alter the percentage of HMG-5/TFAM bound to ΔmtDNAs, consistent with HMG-5 interacting with all mtDNAs independent of replication (Extended Data Fig. 6f). Consistent with ATFS-1-dependent mtDNA replication occurring in dysfunctional mitochondria, when heteroplasmic worms were raised on cco-1(RNAi) or crossed into the clk-1 mutant strain, wildtype mtDNAs were also specifically increased (Extended Data Fig. 6g,h).

Combined, these findings suggest that LONP-1-mediated proteolysis antagonizes the ability of mitochondrial ATFS-1 and POLG to stimulate mtDNA replication. LONP-1 activity may be compromised within mitochondrial compartments harboring ΔmtDNAs, leading to ATFS-1 and POLG accumulation and mtDNA replication. We propose that globally inhibiting LONP-1 promotes ATFS-1-mediated mtDNA replication throughout the cell, not just in compartments enriched in ΔmtDNAs, consequently leading to a recovery of wildtype mtDNAs.

LONP1 inhibition improves heteroplasmy in cybrid cells

Lastly, we examined if the role of LONP1 in maintaining ΔmtDNAs is conserved in mammals by examining two human heteroplasmic cybrid cell lines[34]. While cybrid cells are often used as models of deleterious mtDNA heteroplasmy, it is important to note that cybrid cells are cancer cells in which patient-derived heteroplasmic mtDNAs were introduced by cell fusion[34]. One cybrid line harbors a single nucleotide transition (COXI G6930A) that introduces a premature stop codon in the cytochrome c oxidase subunit I gene, which was isolated from a patient with a multisystem mitochondrial disorder (Fig. 6a)[35]. We also examined a cybrid line harboring a 4977 base pair deletion known as the “common deletion” which removes multiple OXPHOS genes and is associated with Kearns-Sayre Syndrome (KSS), progressive external ophthalmoplegia, cancer and aging (Fig. 6a)[36-39]. We first examined the impact of LONP1 siRNA on heteroplasmy in the KSS cybrid line. Similar to inhibition of C. elegans LONP-1, inhibition of human LONP1 by siRNA for 4 days (Fig. 6b), resulted in a 1.5-fold increase of wildtype mtDNAs (Fig. 6c) while KSS mtDNAs were decreased ~2 fold (Fig. 6d), resulting in a shift in the heteroplasmy ratio from 57.5% to 25.6% (Fig. 6d).

Fig. 6 |

LONP1 inhibition improves heteroplasmy in cybrid cells.

a, Schematic comparing human wildtype, KSS deletion (ΔmtDNA) and CoxI G6930A mtDNAs. b, LONP1 immunoblots from KSS heteroplasmic cells treated with hLONP1 or control (NC) siRNA. Tubulin (Tub) serves as a loading control. Representative immunoblots from three biological repeats. c, WT mtDNA quantification in KSS cells treated with control or hLONP1 siRNA. n = 4, biologically independent samples. **P = 0.0021. d, Quantification of KSS ΔmtDNA in cells treated with control or hLONP1 siRNA. n = 4, biologically independent samples. *P = 0.0185. e, Chemical structure of CDDO. f, Quantification of G6930A mtDNA percentage following treatment with DMSO, 0.1 μM CDDO, or 0.25 μM CDDO at the indicated time points up to 130 days. n = 3, biologically independent samples. *P < 0.05. g, ΔmtDNA quantification in KSS heteroplasmic cells treated with DMSO, 0.1 μM CDDO, or 0.25 μM for 4 or 13 weeks. h, Cell viability of 143b (WT) and CoxI G6930A cells exposed to the indicated concentration of CDDO for 72 hours. n = 3 in 143b(WT) cell or CoxI G6930A cell, biologically independent samples. Every dot stands for averaged value from 3 technical replicates in c,d; Two-tailed Student’s t test was used in c,d and f; data shown in c, d and h represent mean ± S.E.M.

To further investigate the impact of LONP1 protease activity on heteroplasmy, we used the LONP1 inhibitor CDDO (2-cyano-3,12-dioxo-oleana-1,9(11)-dien-28-oic acid), also known as Bardoxolone (Fig. 6e)[20,40]. As determined by deep sequencing, the CoxI G6930A cybrid line initially harbored ~90% G6930A mutant mtDNAs and ~10% wildtype mtDNAs (Fig. 6f). Interestingly, incubation with 0.1 μM or 0.25 μM CDDO for 3 weeks resulted in depletion of the CoxI G6930A mtDNA from 86% to 68% and 72%, respectively, with a concomitant increase in wildtype mtDNAs (Fig. 6f). Furthermore, continuous incubation with 0.1μM or 0.25 μM CDDO for 18.5 weeks further decreased the heteroplasmic ratios from ~90% to ~47% and 62%, respectively (Fig. 6f and Extended Data Fig. 7a). Similar results were obtained when the KSS cybrid cells were incubated with CDDO. As determined by qPCR, KSS cells initially harbored ~50% ΔmtDNAs. Incubation with 0.1 μM or 0.25 μM CDDO for 4 weeks depleted ΔmtDNAs to 19.5% and 20.3 % (Fig. 6g) and incubation for 13 weeks further depleted ΔmtDNAs to 18.6% and 10.7%, respectively (Fig. 6g). Importantly, neither 0.1 μM nor 0.25 μM CDDO affected basal respiration in homoplasmic 143b cells (Extended Data Fig. 7b). Furthermore, neither used CDDO concentration impaired the viability of CoxI G6930A or KSS cells (Fig. 6h and Extended Data Fig. 7c) suggesting that the improved heteroplasmy ratio following CDDO exposure was not due to selection against cells with high levels of mutant mtDNAs. However, it remains possible that cell death may occur during longer exposures.

Extended Data Fig. 7

Pharmacological inhibition of LONP1 improves heteroplasmy and OXPHOS function in heteroplasmic cybrid cells

a, Mutant (G6930A) mtDNA ratio confirmation by sanger sequencing in CoxI G6930A cells treated by CDDO. b, Oxygen consumption rates (OCR) of 143B (wildtype) cells treated with DMSO (ctrl), 0.1 μM or 0.25 μM CDDO for 3 days. n = 22 (ctrl) and n = 24 (0.1 μM and 0.25 μM CDDO), biologically independent samples. c, Cell viability of 143b (WT) and KSS ΔmtDNA cells exposed to various concentrations of CDDO for 72 hours. n = 3, biologically independent samples. d, Basal respiration of KSS heteroplasmic cells treated with DMSO (ctrl), 0.1 μM or 0.25 μM CDDO for 4 or 13 weeks. n = 14 (ctrl) and n = 16 (0.1 μM and 0.25 μM CDDO), biologically independent samples. ****P <0.0001, Two-tailed Student’s t test. Data shown represent mean ± S.E.M.

Last, we determined the impact of the CDDO-dependent shifts in heteroplasmy on OXPHOS. Impressively, incubation of KSS cells with 0.1 μM or 0.25 μM CDDO for 4 or 13 weeks resulted in significant increases in basal respiration suggesting improved OXPHOS (Extended Data Fig. 7d). The improved heteroplasmy caused by CDDO in CoxI G6930A cells (Fig. 7a) also resulted in increased basal respiration and maximal respiratory capacity (Fig. 7b–d). For example, 3 weeks exposure to 0.1 μM CDDO increased basal oxygen consumption ~2-fold, while exposure for 18.5 weeks improved basal oxygen consumption over 3-fold (Fig. 7b,c). Taken together, these findings suggest that inhibition of LONP1 improves deleterious heteroplasmy and recovers mitochondrial respiration. Notably, these phenotypes are independent of mtDNA length, as maintenance of mutant mtDNAs with either a large deletion or a single base pair substitution require LONP1 function.

Fig. 7 |

Pharmacological inhibition of LONP1 improves heteroplasmy and OXPHOS function in cybrid cells.

a, Percentage of CoxI G6930A mtDNA in cells treated with DMSO, 0.1 μM or 0.25 μM CDDO for 3, 5, 10 or 18.5 weeks. n = 3, biologically independent samples. These data are extracted from Fig. 6f. b, Oxygen consumption rates (OCR) of CoxI G6930A cells treated with DMSO, 0.1 μM or 0.25 μM CDDO for 3, 5, 10 or 18.5 weeks. c, Quantification of basal respiration. In 3 weeks, n = 10 (ctrl), n = 16 (0.1 μM) and n = 16 (0.25 μM), biologically independent samples. **P = 0.0028, ****P <0.0001. In 5 weeks, n = 24 (ctrl), n = 24 (0.1 μM) and n = 22 (0.25 μM), biologically independent samples. ****P < 0.0001. In 10 weeks, n = 16 (ctrl), n = 10 (0.1 μM) and n = 15 (0.25 μM), biologically independent samples. ****P < 0.0001. In 18.5 weeks, n = 16 (ctrl, 0.1 μM and 0.25 μM), biologically independent samples. ****P < 0.0001. d, Quantification of respiratory capacity. In 3 weeks, n = 10 (ctrl), n = 16 (0.1 μM) and n = 16 (0.25 μM), biologically independent samples. **P = 0.0086, ***P = 0.0002. In 5 weeks, n = 24 (ctrl), n = 24 (0.1 μM) and n = 22 (0.25 μM), biologically independent samples. ****P < 0.0001. In 10 weeks, n = 16 (ctrl), n = 10 (0.1 μM) and n = 15 (0.25 μM), biologically independent samples. ****P < 0.0001. In 18.5 weeks, n = 16 (ctrl, 0.1 μM and 0.25 μM), biologically independent samples. ****P < 0.0001. Every dot stands for averaged value from 4 technical replicates in c,d; Two-tailed Student’s t test was used in c,d; data shown in b-d represent mean ± S.E.M..

Discussion

The underlying mechanisms that govern heteroplasmy dynamics are largely unknown, however it has been proposed that the mutant genomes have a selective advantage. Previously, we and others reported that heteroplasmy requires atfs-1[9, 16]. In wildtype homoplasmic worms raised under normal conditions, ATFS-1 is imported into mitochondria and degraded by the protease LONP-1[19]. However, during mitochondrial dysfunction, a percentage of ATFS-1 accumulates within mitochondria and binds mtDNA[17]. Here, we report that mitochondrial-localized ATFS-1 promotes the accumulation of mtDNAs upon OXPHOS perturbation. Importantly, both ATFS-1 and POLG binding to mtDNA was increased upon OXPHOS perturbation. Consistent with this model, the nuclear activity of ATFS-1 is not required for the increased mtDNAs caused by OXPHOS perturbation in homoplasmic worms. Collectively, these results suggest that mitochondrial accumulation of ATFS-1 promotes recruitment of the mitochondrial replisome to mtDNA during OXPHOS dysfunction.

ATFS-1 is also required for the increased quantity of mtDNAs in heteroplasmic worms. The uaDf5 strain harbors approximately 60% ΔmtDNAs and 40% wildtype mtDNAs. However, ATFS-1 and POLG interact with 9-fold more ΔmtDNAs than wildtype mtDNAs suggesting the increase in total mtDNAs is due to increased ΔmtDNA replication. As both mtDNAs harbor the ATFS-1 binding site, how is the specificity attained? We find that heteroplasmic worms harbor distinct populations of functional and dysfunctional mitochondria. And, ATFS-1 primarily accumulates within the dysfunctional population. Importantly, LONP-1 inhibition results in ATFS-1 accumulation in all mitochondria suggesting that the replicative advantage is due to LONP-1-dependent degradation of ATFS-1 in functional mitochondria, which likely harbor wildtype mtDNAs. Somatic cells in C. elegans harbor between one and three mtDNAs[7]. Thus, we speculate that the dysfunctional mitochondria in which ATFS-1 accumulates harbor all, or nearly all ΔmtDNAs which impair OXPHOS and LONP-1, resulting in ATFS-1 accumulation.

Because LONP-1 is an ATP-dependent protease that degrades proteins damaged by ROS[27], mitochondrial dysfunction may impede degradation of ATFS-1 by LONP-1, resulting in recruitment of POLG to mtDNA. We propose that this mechanism evolved to coordinate mtDNA replication with expansion of the mitochondrial network during normal cell growth or recovery from mitochondrial dysfunction. However, if compartmental dysfunction is caused by an enrichment of ΔmtDNAs, they are inadvertently, but preferentially, replicated. Our data suggests that inhibiting LONP1 throughout the mitochondrial network negates this preferential replication, leading to a reduction in the heteroplasmic ratio and recovery of mitochondrial function.

There are currently no FDA-approved treatments for diseases caused by mutant mtDNAs. However, the mTOR inhibitor rapamycin improves heteroplasmy in cybrid cells by increasing autophagy[41]. Of note, inhibition of several TORC1 components also inhibits atfs-1-dependent mitochondrial biogenesis in C. elegans[18] suggesting TORC1 may function upstream of ATFS-1 in maintaining heteroplasmy. Here, we report that inhibition of LONP1 through siRNA-mediated knockdown or the small-molecule inhibitor CDDO reduces ΔmtDNA abundance in cybrid cells harboring patient-derived mutant mtDNAs. Moreover, this decrease in ΔmtDNAs was accompanied by improved mitochondrial respiration suggesting LONP1 inhibition may represent a therapeutic strategy for diseases caused by mutant mtDNAs.

Method

Materials and Methods

Worm strains

The reporter strain hsp-6::gfp for visualizing UPRmt activation was previously described[24]. N2 (wildtype), and ΔmtDNA (or uaDf5) strains were obtained from the Caenorhabditis Genetics Center (Minneapolis, Minnesota). The atfs-1(et18) strain was a gift from Mark Pilon. The atfs-1(null), or atfs-1(cmh15), strain was generated via CRISPR-Cas9 in wildtype worms as previously described[24]. The crRNAs (Integrated DNA Technologies) were co-injected with purified Cas9 protein, tracrRNA (Integrated DNA Technologies), and the dpy-10 co-injection marker as described[42]. atfs-1 was introduced into both wildtype worms and the hsp-6::gfp reporter strain via CRISPR-Cas9[18] (crRNAs and replacement sequence listed in Supplementary Table 1). lonp-1FLAG was introduced into both wildtype worms and the hsp-6::gfp reporter strain via CRISPR-Cas9. atfs-1 was introduced into to atfs-1 worms to generate the atfs-1 strain. Each strain was outcrossed at least 5 times. Unless otherwise noted, all worms were harvested between the late L3 and early L4 stages. All strains were maintained at 20°C.

C. elegans mtDNA and human cybrid cell KSS mtDNA quantification

L4 wildtype or uaDf5 worms were placed on agar plates seeded with control(RNAi) or RNAi specific to the described OXPHOS genes and the F1 generation was harvested at the L4 stage. Wildtype mtDNA or ΔmtDNA quantification was performed using qPCR-based methods as described previously[9]. 40–60 worms were harvested in 35 μl of lysis buffer (50 mM KCl, 10 mM Tris-HCl (pH 8.3), 2.5 mM MgCl2, 0.45% NP-40, 0.45% Tween 20, 0.01% gelatin, with freshly added 200 μg/ml proteinase K) and frozen at −80°C for 20 min prior to lysis at 65°C for 80 min. Relative quantification was used for determining the fold changes in mtDNA between samples. 1 μl of lysate was used in each triplicate qPCR reaction. qPCR was performed using the iQ™ SYBR® Green Supermix and the Biorad qPCR CFX96™(Bio-Rad Laboratories). Primers that specifically amplify wildtype or ΔmtDNA are listed in Supplementary Table 1, as are primers that amplify both wildtype and ΔmtDNA (Total mtDNAs). Primers that amplify a non-coding region near the nuclear-encoded ges-1 gene were used as an internal control for normalization (Supplementary Table 1).

For human patient fibroblast cell lines, wildtype and ΔKSS primers were used to detect wildtype mtDNA or ΔKSS mtDNA. Primers that amplify a sequence within the B2M (Human β2 myoglobin) gene were used as an internal control for normalization. Absolute quantification was also performed to determine the percentage or ratio of KSS ΔmtDNA relative to total mtDNA (KSS ΔmtDNA and wildtype mtDNA) as previously described[9]. Primers that specifically amplify wildtype or ΔmtDNA are listed in Supplementary Table 1. Standard curves for each qPCR primer set were generated using purified plasmids individually containing approximately 1 kb of the mtDNA fragments specific for each primer set.

Chromatin immunoprecipitation (ChIP)

ChIP assays for ATFS-1 and LONP-1FLAG were performed as previously described[17]. Synchronized worms were cultured in liquid and harvested at early L4 stage by sucrose flotation. The worms were lysed via Teflon homogenizer in cold PBS with protease inhibitors (Roche). Cross-linking of DNA and protein was performed by treating the worms with 1.85% formaldehyde with protease inhibitors for 15 min. Glycine was added to a final concentration of 125 mM and incubated for 5 min at room temperature to quench the formaldehyde. The pellets were resuspended twice in cold PBS with protease inhibitors. Samples were sonicated in a Bioruptor (Diagenode) for 15 min at 4°C on high intensity (30s on and 30s off). Samples were transferred to microfuge tubes and spun at 15,000*g for 15 min at 4°C. The supernatant was precleaned with pre-blocked ChIP-grade Pierce™ magnetic protein A/G beads (Thermo Scientific) and then incubated with Monoclonal ANTI-FLAG® M2 antibody (Sigma, F1804) or Mouse mAb IgG1 Isotype Control (Cell Signaling Technology, G3A1) rotating overnight at 4°C. The antibody-DNA complex was precipitated with protein A/G magnetic beads or protein A sepharose beads (Invitrogen). After washing, the crosslinks were reversed by incubation at 65°C overnight. The samples were then treated with RNaseA at 37°C for 1.5 hour followed by proteinase K at 55°C for 2 hours. Lastly, the immunoprecipitated and input DNA were purified with ChIP DNA Clean & Concentrator (Zymo Research, D5205) and used as templates for qPCR or next generation sequencing.

mtDNA-immunoprecipitation (mtDNA-ChIP) and mtDNA quantification

mtDNA-immunoprecipitation assays were performed similarly to the previously described ATFS-1 ChIP assay as described[17]. Synchronized worms were cultured in liquid and harvested at the L4 stage by sucrose flotation. The worms were lysed via Teflon homogenizer in cold PBS with protease inhibitors (Roche). Cross-linking of DNA and protein was performed by treating the worms with 1.85% formaldehyde along with protease inhibitors for 20 min at room temperature. Glycine was added to a final concentration of 125 mM and incubated for 5 min at room temperature to quench the formaldehyde. The pellets were washed twice in cold PBS with protease inhibitor. Samples were resuspended in 500 μl FA buffer + 0.1% sarkosyl* + protease/phosphatase inhibitors and transferred to new microfuge tubes and spun at 15,000*g for 15 min at 4°C. The lysates were not sonicated so that wildtype and ΔmtDNA could be quantified by qPCR. The supernatant was precleaned with pre-blocked ChIP-grade Pierce™ magnetic protein A/G beads (Thermo Scientific) and then incubated with the described antibodies rotating overnight at 4°C. The antibody-mtDNA complex was precipitated with protein A/G magnetic beads (Thermo Scientific) (LONP-1FLAG) or protein A sepharose beads (Invitrogen) for ATFS-1, POLG, TFAM or LONP-1 antibodies. Sonicated salmon sperm DNA was used to block non-specific DNA binding on beads). Beads were collected and washed as previously described[43]. After washing, the crosslinks were reversed by incubation at 250 mM NaCl 65°C overnight. The samples were then treated with RNaseA at 37°C for 1.5 hours and then proteinase K at 55°C for 2 hours. Lastly, the samples were purified with ChIP DNA Clean & Concentrator (Zymo Research, D5205) and used as templates for qPCR. The results were normalized by input and either non-specific rabbit IgG or mouse IgG1 was used as a negative control. Primers that amplify wildtype, ΔmtDNA, or all mtDNAs (Total mtDNAs) are in Supplementary Table 1.

ChIP-seq analysis

The DNA fragments were sequenced using MiSeq at the University of Massachusetts Medical School Deep Sequencing Core. The quality of the raw sequencing data was first evaluated with fastqc (0.11.5). (https://www.bioinformatics.babraham.ac.uk/projects/fastqc/), and then mapped to the C. elegans genome (ce10 from UC Santa Cruz) by Burrows-Wheeler Aligner (BWA MEM, BWA version 0.7.15) algorithm with the standard default settings[44]. Duplicate reads were removed using picard tools v1.96 (https://broadinstitute.github.io/picard/). Peaks were determined using MACS version 2.1[45] with the no-model parameter. Input was used as a control for peak-calling. Both narrow Peaks and broad Peaks were called, and the bigwig files were generated with the signal as fold enrichment by macs2 following the procedure at https://github.com/taoliu/MACS/wiki/Build-Signal-Track. The final set of peaks was determined if the difference in intensity values of control sample and input had a significance level of p-value < 0.01. IGV was used to view the peaks and signals. To identify candidate LONP-1 interacting motifs, the regions that were highly enriched were used as input for MEME (http://meme.sdsc.edu). MEME was run using the parameters minw=8, maxw=25, in two modes (zoops & anr) and the significant motifs (E-value >= 1e-01). A background model is used by MEME to calculate the log likelihood ratio and statistical significance of the motif.

Target site SNP frequency analysis in CoxI G6930A cells by deep sequencing

Library construction for deep sequencing was modified from our previous report[46]. Following CDDO treatment, cells were harvested at different time points and genomic DNA extracted. Briefly, regions flanking the CoxI G6930A site were PCR amplified using locus-specific primers bearing tails complementary to the Truseq adapters as described previously[47]. 50–100 ng input genomic DNA (mtDNA included) was PCR amplified with Phusion High Fidelity DNA Polymerase (New England Biolabs):(98°C, 15 s; 67°C 25 s; 72°C 18 s) × 30 cycles. 1 μl of each PCR reaction was amplified with barcoded primers to reconstitute the TruSeq adaptors using the Phusion High Fidelity DNA Polymerase (New England Biolabs): (98°C,15 s; 61°C, 25 s; 72°C, 18 s) × 9 cycles. Equal amounts of the products were pooled and gel purified. The purified library was deep sequenced using a paired-end 150 bp Illumina MiSeq run.

MiSeq data analysis for editing at target sites or off-target sites was performed using a suite of Unix-based-software tools. First, the quality of the paired-end sequencing reads (R1 and R2 fastq files) was assessed using FastQC (http://www.bioinformatics.babraham.ac.uk/projects/fastqc/). Raw paired-end reads were combined using paired end read merger (PEAR)[48] to generate single merged high-quality full-length reads. Reads were then filtered by quality (using Filter FASTQC[47]) to remove those with a mean PHRED quality score under 30 and a minimum per base score under 24. Each group of reads was then aligned to a corresponding reference sequence using BWA (version 0.7.5) and SAMtools (version0.1.19). To determine background SNP or sequencing errors, all reads from each negative control replicate were combined and aligned, as described above. Background SNP types and frequencies were then cataloged in a text output format at each base using bam-readcount (https://github.com/genome/bam-readcount). For each drug treatment group, the average background SNP frequencies (based on SNP type, position and frequency) of the triplicate negative control group were subtracted to obtain the accurate SNP frequencies.

QuantStudio 3D Digital PCR

The detailed method was described previously[49]. All primers and probes were ordered from IDT (Coralville, Iowa). The 3D digital PCR was used according to the manufacturer’s protocol. QuantStudio™ 3D Digital PCR Master Mix v2 and individual QuantStudio 3D digital PCR 20K Chip kit v2 were purchased from Thermo scientific (Applied Biosystems, Waltham, MA). Prepared sample mix was loaded into ChIP using QuantStudio 3D Digital PCR Chip Loader (Thermo scientific). Chip PCR amplification was performed in a ProFlex PCR System (96 °C for 10 min; 39 cycles of 60 °C for 2 min and 98 °C for 30 sec; and 60 °C for 2 min). After amplification, each chip was loaded into the QuantStudio 3D Digital reader. Data were analyzed using QuantStudio 3D Analysis Suite (Thermo scientific). All biological repeats were performed in at least triplicate.

ΔmtDNA primers F: TTGCTTTTTCTTTATATGTTTTG; R: TTTATTTAATTTGGTTAAACAAGAGGT. ΔmtDNA probes 5’ 6-FAM/ZEN/3’ IBFQ: /56-FAM/AGGATCGTA/ZEN/ACATTTTATTTTTTTGCTTTA/3IABkFQ/.

Wildtype primers F: GCTTTTTCTTTATATGTTTTGTG.

R: TCACCTTCAGAAAAATCAAATGG wildtype mtDNA probes 5’ HEX/ZEN/3’ IBFQ: /5HEX/AATTATAGT/ZEN/AATTGCTGAACTTAACCGGGC/3IABkFQ/

RNA isolation and qRT-PCR

Total RNA was isolated from worm pellets using the TRIzol™ Reagent (Invitrogen). cDNA was then synthesized from total RNA using the iScript cDNA Synthesis Kit (Bio-Rad). qPCR was performed to determine the expression levels of the indicated genes using iQ™ SYBR GREEN supermix (Bio-Rad). Primer sequences are listed in Supplementary Table 1. Relative expression of target genes was normalized to the control. Fold changes in gene expression were calculated using the comparative CtΔΔCt method as previously described[17].

Chemicals and Antibodies

CDDO (Cayman Chemicals Cat No 81035). ATFS-1 polyclonal antibodies were generated and validated previously[19]. Polyclonal antibodies were generated to amino acid amino acids 1054–1072 of C. elegans POLG and subsequently affinity purified by Thermo Fisher Scientific Inc. Polyclonal antibodies were generated to amino acid amino acids 191–204 of C. elegans HMG-5 (TFAM) and subsequently affinity purified by Thermo Fisher Scientific Inc. Polyclonal antibodies were generated to amino acid amino acids 953–971of C. elegans LONP-1 and subsequently affinity purified by Thermo Fisher Scientific Inc. Monoclonal anti-FLAG® M2 antibody (Sigma, Cat # F1804), α-tubulin (Sigma), NDUFS3 (NUO-2 in C. elegans, complex I, Abcam). Supplementary Table 2.

Cell Culture

The KSS cell line was a gift from Carlos Moraes[36, 37]. The CoxI C6930A cell line was a gift from Giovanni Manfredi[35]. Cells were cultured in DMEM (4mM L-glutamine, 4.5 g/L glucose; Gibco, Thermo Fisher Scientific) plus 10% FBS with 1% pen-strep. Total cellular mtDNA was prepared as described[50]. Cells were incubated continuously in the described concentration of CDDO for the indicated number of days. The cells were sub-cultured prior to confluence every 48 hours.

Cell Viability

At the indicated time points, cells were stained with trypan blue[51] and quantified with an automated cell counter TC-20Tm (Bio-Rad). The results are an average of three independent assays.

siRNA

Cells were grown in 6-well plates and siRNAs were transfected with Lipofectamine RNAiMAX (Thermo Fisher Scientific Cat No 13778150) following the manufacturer’s instructions. Human LONP1 siRNA was purchased from Dharmacon (L-003979-00-0005).

Respiration Assays

For mitochondrial respiration assays, oxygen consumption rate (OCR) was measured using a Seahorse Extracellular Flux Analyzer XFe96 (Seahorse Biosciences) as described[50]. 14,000 CoxI G6930A cells were seeded per well with fresh medium. OCR was measured using the Cell MitoStress Kit (as described by the manufacture). 180 μl of XF-Media was added to each well and then the plates were subjected to analysis following sequential introduction of 1.5 μM oligomycin, 1.0 μM FCCP and 0.5 μM rotenone/antimycin as indicated. Data is normalized to total protein as determined by the BCA protein assay.

Western Blots and Mitochondrial Fractionation

Mitochondrial fractionation and western blots were performed as described previously[17]. Synchronized L4 worms were harvested from liquid culture by sucrose flotation. The worm pellets were approximately 150 μl. The worms were washed two times in mitochondria buffer (70 mM sucrose, 1 mM EGTA, 210 mM sorbitol, 10 mM MOPS [pH 7.4])[52] and resuspended in 300 μl of mitochondrial buffer. The worms were then homogenized with a dounce homogenizer at 4°C. 600 μl of mitochondria buffer was added and the samples were centrifuged at 450 × g for 10min at 4°C to pellet debris. The supernatant was collected and centrifuged at 550 × g for 10min at 4°C to pellet the remaining debris. The supernatant was collected and 100 μl was saved as the Total fraction. The rest of the supernatant was centrifuged at 9500 × g for 10min at 4°C to pellet the mitochondria. The supernatant was saved as the Cytosolic fraction. The mitochondrial pellet was washed once in mitochondria buffer, resuspended in 1.5 ml mitochondrial buffer and incubated on ice for 15 min. The sample was centrifuged at 500 × g for 10min at 4°C. The supernatant was collected and again centrifuged at 9500 × g for 10min at 4°C. The pellet was collected and resuspended in 100 μl mitochondria buffer and saved as the Mitochondrial fraction.

Imaging and Fluorescence Quantification

Whole worm images were obtained using either a Zeiss AxioCam MRm camera mounted on a Zeiss Imager Z2 microscope or a Zeiss M2BIO dissecting scope as described[24]. TMRE staining was performed by synchronizing and raising worms on plates previously soaked with S-Basal buffer containing DMSO, or final concentration 100 μM TMRE (Sigma, Cat No 87917). Prior to imaging, the TMRE-stained worms were transferred to plates seeded with control(RNAi) bacteria for 3 h to remove TMRE-containing bacteria from the digestive tract. Images were acquired using identical exposure times with a ZEISS LSM800 microscope with Airyscan. TMRE staining analysis was performed as described[53, 54]. In short, the average pixel intensity values were calculated by sampling images of different worms. The average pixel intensity for each animal was calculated using ImageJ (http://rsb.info.nih.gov/ij/). The fluorescence pixel intensity was quantified by using the threshold-adjusted images from worms at each condition in biological triplicates. Mean values were compared using Student’s t test or one-way (ANOVA) variance analysis followed by the post-hoc Tukey’s test where appropriate.

ATFS-1::GFP transgenic worms were stained with TMRE as described above. Co-localization of TMRE stain and ATFS-1::GFP was determined by object-based colocalization analysis as described[55]. The amount of co-localization was calculated for each sample. Statistical analysis was performed using the Prism software package (GraphPad Software).

Statistics and reproducibility

All data are reported as mean ± SEM. Significance was accepted at P< 0.05. Asterisks denote corresponding statistical significance *P < 0.05; **P < 0.01; ***P < 0.001 and ****P < 0.0001. No statistical method was used to predetermine the sample size. Results were analyzed using Student’s t test with a two-tailed distribution or one-way ANOVA (for multiple comparison) where appropriate, using GraphPad Prism software with corrected P values < 0.05 considered significant. For multiple comparisons, P values were adjusted using the Tukey’s post hoc. The statistical analyses for oxygen consumption rate (OCR) and ChIP-mtDNA assays were all performed using two-sided Student’s t-tests; no adjustments were applied. The statistical tests performed and definition of “n” numbers in this study are indicated in the figure legends. For Figs. 1d,1g–1h, 2c,2e–2i, 3b, 3e–3f, 3h–3i, 4a,4e, 4g–4i, 5a–5d, 5f–5g, 6c,6d, 6g–6h, 7a–7d, and Extended Data Figs. 1a, 2c, 2f–h, 3f–g, 4e–g, 5c, 5e, 6b–h ,7b–d, “n” means biologically independent samples. For C. elegans western blotting and gene expression, each sample within each biological replicate corresponds to a sample pooled from 3000–5000 animals. For the immunoblotting data, each blot (Figs. 2a,2d,3a,4c,6b and Extended Data Figs. 1b–c,2i–j, 3e,5g) was repeated at least three times and each bot (Extended Data Fig. 5f) was repeated twice with similar results. In Figs.1c,1e,1f,2b,3g,4b,4d; 5e, and Extended Data Figs.2d,3b,3c,4c, these experiments were independently repeated at least three times. For cell culture western blotting, each sample within each biological replicate corresponds to one well from a tissue culture plate. For Figs. 7b–d and Extended Data Figs.7d, “n” means biological replicates and each sample within each biological replicate corresponds to a sample pooled from 14000 cells in Fig. 7b–d, 18000 cells in Extended Data Fig. 7d (4 weeks) and 10000 cells (13 weeks). The deep sequencing to quantify heteroplasmy of the CoxI G6930A mtDNA following exposure to CDDO was double-blinded. The researchers involved in the experiments were not completely blinded during sample obtainment or data analysis.