Genetically controlled mtDNA deletions prevent ROS damage by arresting oxidative phosphorylation

Simon Stenberg; Jing Li; Arne B Gjuvsland; Karl Persson; Erik Demitz-Helin; Carles González Peña; Jia-Xing Yue; Ciaran Gilchrist; Timmy Ärengård; Payam Ghiaci; Lisa Larsson-Berglund; Martin Zackrisson; Silvana Smits; Johan Hallin; Johanna L Höög; Mikael Molin; Gianni Liti; Stig W Omholt; Jonas Warringer

doi:10.7554/eLife.76095

. 2022 Jul 8;11:e76095. doi: 10.7554/eLife.76095

Genetically controlled mtDNA deletions prevent ROS damage by arresting oxidative phosphorylation

Simon Stenberg ^1,², Jing Li ^3,⁴, Arne B Gjuvsland ¹, Karl Persson ², Erik Demitz-Helin ², Carles González Peña ², Jia-Xing Yue ^3,⁴, Ciaran Gilchrist ², Timmy Ärengård ², Payam Ghiaci ², Lisa Larsson-Berglund ², Martin Zackrisson ², Silvana Smits ², Johan Hallin ², Johanna L Höög ², Mikael Molin ^2,⁵, Gianni Liti ⁴, Stig W Omholt ^6,^✉, Jonas Warringer ^2,^✉

Editors: Jan Gruber⁷, Jessica K Tyler⁸

PMCID: PMC9427111 PMID: 35801695

Abstract

Deletion of mitochondrial DNA in eukaryotes is currently attributed to rare accidental events associated with mitochondrial replication or repair of double-strand breaks. We report the discovery that yeast cells arrest harmful intramitochondrial superoxide production by shutting down respiration through genetically controlled deletion of mitochondrial oxidative phosphorylation genes. We show that this process critically involves the antioxidant enzyme superoxide dismutase 2 and two-way mitochondrial-nuclear communication through Rtg2 and Rtg3. While mitochondrial DNA homeostasis is rapidly restored after cessation of a short-term superoxide stress, long-term stress causes maladaptive persistence of the deletion process, leading to complete annihilation of the cellular pool of intact mitochondrial genomes and irrevocable loss of respiratory ability. This shows that oxidative stress-induced mitochondrial impairment may be under strict regulatory control. If the results extend to human cells, the results may prove to be of etiological as well as therapeutic importance with regard to age-related mitochondrial impairment and disease.

Research organism: S. cerevisiae

Introduction

Mitochondrial impairment is strongly associated with aging (Sun et al., 2016) and the pathogenesis of age-related human diseases, including Alzheimer’s disease (Hu et al., 2017), Parkinson’s disease (Ammal Kaidery and Thomas, 2018), the deterioration of skeletal and cardiac muscle (Hepple, 2016), and macular degeneration (Hyttinen et al., 2018). Mitochondrial DNA (mtDNA) deletions are perceived to contribute markedly to this impairment (Krishnan et al., 2008). The general conception is that mtDNA deletions are deleterious events related to either faulty mtDNA replication or the mis-repair of mtDNA of double-strand breaks (Nissanka et al., 2019). However, considering that mtDNA deletions are prone to cripple oxidative phosphorylation (Fontana and Gahlon, 2020) (OXPHOS), and thus turn off the production of intramitochondrial superoxide anion (O₂^∙−), it is conceivable that mtDNA deletion is also under genetic control. The main rationale is that compared to a mitophagic response (Bess et al., 2012; Palikaras and Tavernarakis, 2014; Sedlackova and Korolchuk, 2019; Gustafsson and Dorn, 2019; Ng et al., 2021), it would serve as a less costly defense mechanism against an abrupt increase in reactive oxygen species (ROS) inundating the primary antioxidant defenses (Sies et al., 2017; Shpilka and Haynes, 2018). The disclosure of an additional genetically controlled defense layer against O₂^∙− damage, situated between the primary antioxidant defenses and mitophagy, would bring a fresh perspective to what causes mitochondrial impairment and how it can be mitigated. This motivated us to search for the existence of such a regulatory layer in wild budding yeast (Saccharomyces cerevisiae).

Results

Paraquat impairs cell growth through mitochondrial O₂^∙− production

Domestication has systematically enhanced fermentative and reduced respiratory asexual growth (De Chiara et al., 2020), with common lab strains harboring multiple defects in mitochondrial respiratory biology (Gaisne et al., 1999). To avoid possible confounding effects of domestication we therefore chose to work with the wild strain YPS128 (Liti et al., 2009). We exposed haploid YPS128 cell populations expanding clonally on glucose to the mitochondrial O₂^∙− generator and redox cycler paraquat (N,N-dimethyl-4–4′-bipiridinium dichloride) (Cochemé and Murphy, 2008). Paraquat generates O₂^∙− by passing electrons from OXPHOS complex III and mitochondrial NADPH dehydrogenases to O₂ (Cochemé and Murphy, 2008; Castello et al., 2007), a mode of O₂^∙− generation that is a good proxy for the in vivo situation (Zou et al., 2017).

We titrated the paraquat dose to cause a 2.5-3 fold increase in cell doubling time (Figure 1—figure supplement 1A). At the chosen concentration (400 µg/mL), key mitochondrial oxidative stress response genes, copper/zinc dependent O₂^∙− dismutase (Cu/ZnSOD, Sod1), manganese-dependent O₂^∙− dismutase (MnSOD, Sod2) and mitochondrial cytochrome C peroxidase (Ccp1), increased their transcript levels 2–12-fold in early lag-phase (Figure 1—figure supplement 1B). The cells maintained the elevated expression of these antioxidant transcripts throughout the exponential growth phase and the following two growth cycles without causing any detectable reduction in the cell doubling time. Assuming that this expression increase reflects mobilization of the whole repertoire of primary antioxidant defenses, the lack of growth improvement shows that the paraquat-induced O₂^∙− production was well above the reach of these defenses, while still allowing for cellular function.

Addition of vitamin C, an antioxidant that accepts electrons from PQ⁺, the free radical (or ‘damaging’) state of paraquat (Sendra et al., 1999), caused the growth of paraquat exposed wild type cells, as well as paraquat exposed sod2Δ cells, to be on par with that of unexposed wild type cells (Figure 1—figure supplement 1C, D). This suggested that a paraquat concentration of 400 µg/mL impaired cell growth entirely through its effect on O₂^∙− production.

Swift adaptation to increased O₂^∙− production

We then used a high throughput growth platform (Zackrisson et al., 2016) to observe how 96 asexually reproducing yeast cell populations (colonies) on solid agar medium adapted to the chosen paraquat dose in terms of change in cell doubling time. The evolution experiment was run for 50 growth cycles, from lag to stationary phase (Figure 1—figure supplement 1E), with each cycle lasting 72 hr. The number of cells in each colony doubled 2.5–6 x in each cycle, and over the 50 growth cycles the populations doubled in size ~240 times, on average. Neglecting cell deaths and assuming synchronous cell divisions, this corresponds to ~240 cell generations. To provide a comparative data set, we similarly exposed a total of 672 cell populations to seven other stressors not explicitly challenging mitochondrial function over 50 growth cycles (Supplementary file 1). All 96 cell populations exposed to paraquat adapted much faster than every other cell population exposed to any other stressor (Figure 1A). Between 4 and 10 cell generations, populations on average reduced their doubling time by 106 min. Assuming the minimum achievable cell doubling time to be that of the wild type before exposure to paraquat (a mean of 93 min), this corresponded to 49.3% of the maximum possible reduction. Thenceforth, adaptation entered a second phase where the reduction in cell doubling time progressed much slower until it plateaued after 75 generations at 72.6% of the maximum possible reduction (mean) (Figure 1A).

Figure 1. — (A) Mean temporal adaptive response to paraquat and seven other stressors. y-axis shows log₂ fold reduction in cell doubling time (h) from pre-stress, adjusting for plate, position and pre-culture effects. 96 populations for each stressor (n=6). Shade: S.E.M. (B) Loss of the acquired adaptation as a function of number of cell generations after release from the selection pressure. Colored lines: mean of 96 populations (each measured at n=5). Shade: S.E.M. The populations were released from stress after reaching 70–90% of their endpoint (**t₅₀**) adaptation. (C) The difference in cell doubling time (h) in a no-stress environment between 96 populations (each measured at n=5) having achieved 70–90% of their endpoint adaptation to paraquat, arsenic and glycine, respectively, and the founder population. The difference reflects the selective advantage of losing the acquired adaptations when the populations are no longer exposed to stress. p-values: one-sided t-test. Error bars: S.E.M. See also Figure 1—figure supplements 1 and 2.

Figure 1—source data 1. Doubling time data of 96 populations adapted to each of eight different environments over G generations; doubling times are in the respective selection environment.
Data are shown in Figure 1A and Figure 6D.

elife-76095-fig1-data1.xlsx^{(1.1MB, xlsx)}

Figure 1—source data 2. Difference in doubling time in absence of stress and in the respective selection environment, for adapted populations having achieved 70-90% of their final adaptation.
Data are shown in Figure 1C.

elife-76095-fig1-data2.xlsx^{(44.1KB, xlsx)}

Figure 1—source data 3. Doubling time data of 96 populations adapted to paraquat, arsenic and glycine over *G_s* generations and then released from selection for G_r generations; doubling times are in paraquat, arsenic, and glycine, respectively.
Data are shown in Figure 1B.

elife-76095-fig1-data3.xlsx^{(235.1KB, xlsx)}

Figure 1—figure supplement 1. — (A) Mean temporal adaptive response to paraquat and seven other stressors. y-axis shows log₂ fold reduction in cell doubling time (h) from pre-stress, adjusting for plate, position and pre-culture effects. 96 populations for each stressor (n=6). Shade: S.E.M. (B) Loss of the acquired adaptation as a function of number of cell generations after release from the selection pressure. Colored lines: mean of 96 populations (each measured at n=5). Shade: S.E.M. The populations were released from stress after reaching 70–90% of their endpoint (**t₅₀**) adaptation. (C) The difference in cell doubling time (h) in a no-stress environment between 96 populations (each measured at n=5) having achieved 70–90% of their endpoint adaptation to paraquat, arsenic and glycine, respectively, and the founder population. The difference reflects the selective advantage of losing the acquired adaptations when the populations are no longer exposed to stress. p-values: one-sided t-test. Error bars: S.E.M. See also Figure 1—figure supplements 1 and 2.

Figure 1—source data 1. Doubling time data of 96 populations adapted to each of eight different environments over G generations; doubling times are in the respective selection environment.
Data are shown in Figure 1A and Figure 6D.

elife-76095-fig1-data1.xlsx^{(1.1MB, xlsx)}

Figure 1—source data 2. Difference in doubling time in absence of stress and in the respective selection environment, for adapted populations having achieved 70-90% of their final adaptation.
Data are shown in Figure 1C.

elife-76095-fig1-data2.xlsx^{(44.1KB, xlsx)}

Figure 1—source data 3. Doubling time data of 96 populations adapted to paraquat, arsenic and glycine over *G_s* generations and then released from selection for G_r generations; doubling times are in paraquat, arsenic, and glycine, respectively.
Data are shown in Figure 1B.

elife-76095-fig1-data3.xlsx^{(235.1KB, xlsx)}

We used a numerical model of the adaptation process to test if the extraordinarily fast paraquat adaptation could reasonably be accounted for by cell populations accumulating loss-of-function point mutations or chromosome duplications, both of which have previously been linked to fast Darwinian adaptation in experimental yeast populations exposed to selection (Voordeckers and Verstrepen, 2015). We combined population genetics and population dynamics theory with existing data on yeast mutation rates and our experimental measures of effect sizes of loss-of-function mutations and aneuploidies (Gjuvsland et al., 2016). The numerical model was completely unable to reproduce the observed extraordinarily swift response to paraquat, indicating that nuclear mutations were unlikely to explain the phenomenon (Figure 1—figure supplement 2).

We then tested experimentally whether a Darwinian adaptive process driven by selection of new mutations could account for the observed paraquat adaptation in a stress-release experiment. To this end, we exposed new cell populations to paraquat over many consecutive growth cycles. After each growth cycle, a fraction of the adapting cells was placed in a paraquat-free medium for 1–10 growth cycles before being exposed to paraquat once more. The rationale being that if the adaptation is due to accumulation of random mutations, loss of the adaptation would progress gradually and take many growth cycles. All 96 cell populations retained their acquired tolerance to paraquat (mean reduction in cell doubling time: 106 min) for only 1–3 growth cycles before abruptly losing it (Figure 1B). When employing the same experimental procedure to 96 cell populations from each of the two other environments to which adaptation was also fast (arsenic and glycine), we found that despite the presence of a much stronger Darwinian counterselection (Figure 1C), these populations lost their acquired adaptations more slowly and gradually (Figure 1B). Thus, while a Darwinian mutation/selection-based adaptive process could potentially explain the data for seven of the eight tested stressors, the paraquat adaptation could hardly be reconciled with such a process.

Mitophagy is not responsible for the swift first adaptation phase

As the fast adaptation was unlikely to be a result of canonical mutation/selection dynamics, we went on to investigate whether a mitophagic response was responsible. Mitochondrial fragmentation is a well-documented prelude to canonical mitophagy (Sprenger and Langer, 2019). We therefore first assayed mitochondrial morphology before, during and after paraquat exposure by confocal and electron microscopy. In both cases, paraquat caused a rapid shift (<5 hr) from a tubular to a fragmented mitochondrial organization (Figure 2A and B, Figure 2—figure supplement 1). After removing the paraquat stress we observed a rapid reversal (<5 hr) back to a tubular organization (Figure 2B). These results are consistent with the notion that mitochondrial O₂^∙− generation influences the mitochondrial fission and fusion dynamics (Frank et al., 2012; Hung et al., 2018; Sprenger and Langer, 2019). Nevertheless, the cumulative mitochondrial volume remained near pre-stress levels with at the most a marginal reduction after 77–79 hr of paraquat exposure (Figure 2A and B). Most importantly, cell populations (n=16) lacking Atg32, a key component of canonical mitophagy (Liu and Okamoto, 2021), adapted to paraquat over ~80 generations as wild-type populations (Figure 2C). This led us to conclude that the initial swift adaptation to paraquat did not depend on canonical mitophagy.

Figure 2. — (A) EM microscopy of cells before (panel 1) and after short-term stress (panel 2, red arrowheads mark representative mitochondria). Panel 3 shows the number of mitochondria per imaged cell (left) and the imaged cell area occupied by mitochondrial area (%) (right), used as proxy for mitochondrial volume. Error bars: S.E.M. (n=100 cells). p-values: Welch two-sided t-test. (B) Confocal microscopy of cells with a Cox4-GFP mitochondrial tag. Color: z-dimension (yellow = front, purple = back; 18 slices). *Lower left diagram:* Number of mitochondria per imaged cell. *Lower right diagram:* Mean sum of mitochondrial volume as a fraction of cell volume. Error bars: S.E.M. (n=473–910 cells). p-values: Welch two-sided t-test (C) Adaptation of *atg32Δ* and wild-type populations to paraquat. Shade: S.E.M. (n=16 populations, each measured at n=6). See also Figure 2—figure supplement 1.

Figure 2—source data 1. Mitochondrial and cell area quantified, based on electron microscopy micrographs.
Data are shown in Figure 2A.

elife-76095-fig2-data1.xlsx^{(1.1MB, xlsx)}

Figure 2—source data 2. Number of mitochondria quantified, based on electron microscopy micrographs.
Data are shown in Figure 2A.

elife-76095-fig2-data2.xlsx^{(38.7KB, xlsx)}

Figure 2—source data 3. Number and volume of mitochondria of cells segmented from confocal microscopy micrographs of cells exposed to paraquat.
Data are shown in Figure 2B and in Figure 2—figure supplement 1.

elife-76095-fig2-data3.xlsx^{(137.9KB, xlsx)}

Figure 2—source data 4. Doubling time data of wild type and populations adapting to paraquat over generations G; doubling times are in paraquat.
Data are shown in Figure 2C.

elife-76095-fig2-data4.xlsx^{(53.5KB, xlsx)}

(A) EM microscopy of cells before (panel 1) and after short-term stress (panel 2, red arrowheads mark representative mitochondria). Panel 3 shows the number of mitochondria per imaged cell (left) and the imaged cell area occupied by mitochondrial area (%) (right), used as proxy for mitochondrial volume. Error bars: S.E.M. (n=100 cells). p-values: Welch two-sided t-test. (B) Confocal microscopy of cells with a Cox4-GFP mitochondrial tag. Color: z-dimension (yellow = front, purple = back; 18 slices). *Lower left diagram:* Number of mitochondria per imaged cell. *Lower right diagram:* Mean sum of mitochondrial volume as a fraction of cell volume. Error bars: S.E.M. (n=473–910 cells). p-values: Welch two-sided t-test (C) Adaptation of *atg32Δ* and wild-type populations to paraquat. Shade: S.E.M. (n=16 populations, each measured at n=6). See also Figure 2—figure supplement 1.

Figure 2—source data 1. Mitochondrial and cell area quantified, based on electron microscopy micrographs.
Data are shown in Figure 2A.

elife-76095-fig2-data1.xlsx^{(1.1MB, xlsx)}

Figure 2—source data 2. Number of mitochondria quantified, based on electron microscopy micrographs.
Data are shown in Figure 2A.

elife-76095-fig2-data2.xlsx^{(38.7KB, xlsx)}

Figure 2—source data 3. Number and volume of mitochondria of cells segmented from confocal microscopy micrographs of cells exposed to paraquat.
Data are shown in Figure 2B and in Figure 2—figure supplement 1.

elife-76095-fig2-data3.xlsx^{(137.9KB, xlsx)}

Figure 2—source data 4. Doubling time data of wild type and populations adapting to paraquat over generations G; doubling times are in paraquat.
Data are shown in Figure 2C.

elife-76095-fig2-data4.xlsx^{(53.5KB, xlsx)}

mtDNA segmental deletions cause the swift adaptation to paraquat

After excluding canonical mitophagy we considered mtDNA copy number variation as a possible explanation for the swift adaptation. We first used short read sequencing to measure the mean coverage of the mitochondrial genome in five of the 96 yeast cell populations adapting to paraquat (Figure 3A). In all five populations, we found the mean mtDNA copy number to decrease dramatically during the early phase of paraquat adaptation. This led us to use qPCR to track changes in the copy numbers of individual mtDNA genes in nine paraquat-adapting cell populations. We found that all nine cell populations lost copy numbers of some, but not all, mtDNA-encoded genes during the early adaptation phase (Figure 3B, Figure 3—figure supplement 1). As adjacent genes were lost concomitantly, and to the same extent, the observed loss in copy numbers implied that the early adaptation phase was associated with deletion of entire mtDNA segments. In addition, the qPCR data showed that the lost segments were unevenly distributed across the mitochondrial genome: all nine cell populations lost one or more segments within the mtDNA region spanning COX1 to VAR1, while a few also lost the 21 S rRNA and COX2 rapidly thereafter. The lost segments also contained almost all of the mitochondrial tRNAs. The retained mtDNA segments, which in all nine cases encompassed COX3-RPM1 and 15 S rRNA, remained at near founder lever throughout the early adaptation phase. Since the mtDNA coverage prior to paraquat adaptation was perfectly even (Figure 3—figure supplement 2A), the observed mtDNA loss was clearly induced by paraquat.

Figure 3. — (A) mtDNA copy number change (left y-axis, red line, median coverage relative to the haploid nuclear genome) during paraquat adaptation (right y-axis, green line, n=6) for 5 populations (panels). Shade: S.E.M. (B) mtDNA deletions associate with the paraquat adaptation. *Circle:* mtDNA (77 kb) before exposure to paraquat. Genes, origins of replication and position (kb) are indicated. *Coloured fields:* mtDNA deletions with concerted copy number change. *Diagrams:* mtDNA copy number change (left y-axis, purple line) of individual mtDNA genes during adaptation (right y-axis, green line) in population A7. Shade: S.E.M. (n=3). (C) The early-phase paraquat adaptation coincides with the loss of respiratory (glycerol) growth. Shade: S.E.M. 96 populations, each measured at n=5. Broken line indicates no growth (cell doubling time >24 h). (D) Recovery of the copy number of deleted mtDNA (right y-axis, red line) after release from 6 generations of paraquat exposure coincides with loss of the early-phase paraquat adaptation (left y-axis, green line, shade = S.E.M (n=15)) in populations A7, A8, and B12. (E) Doubling time (h) of wild type and *mip1Δ* cell populations in paraquat. Error bars: S.E.M. (n=191). p-values: Welch two-sided t-test. See also Figure 3—figure supplements 1–3.

Figure 3—source data 1. Doubling time data of 96 populations adapted to paraquat for G generations exposed; doubling times are in paraquat and respiratory media (glycerol).
Data are shown in Figure 3C.

elife-76095-fig3-data1.xlsx^{(56.9KB, xlsx)}

Figure 3—source data 2. Mean log₂ coverage of 1 kb windows spanning the mitochondrial genome of five sequenced paraquat adapting populations over generations G of selection.
Data are shown in Figure 3A.

elife-76095-fig3-data2.xlsx^{(218.6KB, xlsx)}

Figure 3—source data 3. Mean log₂ coverage of 1 kb windows spanning the mitochondrial genome of sequenced populations adapting to paraquat and then released from this selection; data is given as a function of generations G of relaxation of selection.
Data are shown in Figure 3D and in Figure 3—figure supplement 2C.

elife-76095-fig3-data3.xlsx^{(179.1KB, xlsx)}

Figure 3—source data 4. qPCR data for mitochondrial DNA genes and nuclear DNA controls over generations of paraquat adaptation.
Data are shown in Figure 3B and Figure 3—figure supplement 1.

elife-76095-fig3-data4.xlsx^{(104.9KB, xlsx)}

(A) mtDNA copy number change (left y-axis, red line, median coverage relative to the haploid nuclear genome) during paraquat adaptation (right y-axis, green line, n=6) for 5 populations (panels). Shade: S.E.M. (B) mtDNA deletions associate with the paraquat adaptation. *Circle:* mtDNA (77 kb) before exposure to paraquat. Genes, origins of replication and position (kb) are indicated. *Coloured fields:* mtDNA deletions with concerted copy number change. *Diagrams:* mtDNA copy number change (left y-axis, purple line) of individual mtDNA genes during adaptation (right y-axis, green line) in population A7. Shade: S.E.M. (n=3). (C) The early-phase paraquat adaptation coincides with the loss of respiratory (glycerol) growth. Shade: S.E.M. 96 populations, each measured at n=5. Broken line indicates no growth (cell doubling time >24 h). (D) Recovery of the copy number of deleted mtDNA (right y-axis, red line) after release from 6 generations of paraquat exposure coincides with loss of the early-phase paraquat adaptation (left y-axis, green line, shade = S.E.M (n=15)) in populations A7, A8, and B12. (E) Doubling time (h) of wild type and *mip1Δ* cell populations in paraquat. Error bars: S.E.M. (n=191). p-values: Welch two-sided t-test. See also Figure 3—figure supplements 1–3.

Figure 3—source data 1. Doubling time data of 96 populations adapted to paraquat for G generations exposed; doubling times are in paraquat and respiratory media (glycerol).
Data are shown in Figure 3C.

elife-76095-fig3-data1.xlsx^{(56.9KB, xlsx)}

Figure 3—source data 2. Mean log₂ coverage of 1 kb windows spanning the mitochondrial genome of five sequenced paraquat adapting populations over generations G of selection.
Data are shown in Figure 3A.

elife-76095-fig3-data2.xlsx^{(218.6KB, xlsx)}

Figure 3—source data 3. Mean log₂ coverage of 1 kb windows spanning the mitochondrial genome of sequenced populations adapting to paraquat and then released from this selection; data is given as a function of generations G of relaxation of selection.
Data are shown in Figure 3D and in Figure 3—figure supplement 2C.

elife-76095-fig3-data3.xlsx^{(179.1KB, xlsx)}

Figure 3—source data 4. qPCR data for mitochondrial DNA genes and nuclear DNA controls over generations of paraquat adaptation.
Data are shown in Figure 3B and Figure 3—figure supplement 1.

elife-76095-fig3-data4.xlsx^{(104.9KB, xlsx)}

As even small mtDNA deletions are prone to cripple oxidative phosphorylation (OXPHOS) (Fontana and Gahlon, 2020), we predicted that the observed loss of multiple protein, rRNA and tRNA functions had caused a substantial loss of OXPHOS. We therefore tested the respiratory growth capacity of the adapted populations by growing them on glycerol and found that the early paraquat adaptation coincided almost perfectly with loss of respiratory growth (Figure 3C). Similarly, the rapid loss of the acquired adaptations after removal of the O₂^∙− stress coincided with the restoration of the capacity for respiratory growth (Figure 3—figure supplement 2B). To ensure that this restoration was associated with restoration of the mtDNA pool, we repeated the stress release experiment on the five sequenced populations and tracked, by short-read sequencing, the change in copy numbers of the lost mtDNA segments over the course of the experiment. As expected, the five populations quickly restored their capacity for respiratory growth after removal of paraquat, and this restoration coincided with the reestablishment of the wild type mtDNA profile and the loss of the acquired paraquat adaptation (Figure 3D, Figure 3—figure supplement 2B, C).

Together with the observation that ρ⁰ yeast cells devoid of mtDNA devoid of OXPHOS growing on a normal glucose medium show a twofold decrease in cellular O₂^∙− production (Reddi and Culotta, 2013), the data led us to hypothesize that the reduction of paraquat toxicity, leading to a substantial reduction in cell doubling time during the first adaptation phase, was closely associated with the loss of OXPHOS caused by loss of critical mtDNA segments.

We therefore predicted that cells already deprived of mtDNA would be preadapted to paraquat, i.e. their growth rate would be on par with that of cells that had just gone through the initial swift adaptation. To test this, we deleted the sole mitochondrial DNA polymerase Mip1 (Pol γ homolog) (Lodi et al., 2015), and cultivated the resulting ρ⁰ cell populations (n=12) in the presence and absence of paraquat. In the absence of paraquat, the growth of mip1Δ cells was slower than that of the wild type cells (mean doubling time, mip1Δ=3.54 hr, WT = 1.55 hr). However, they did not show the pronounced petite phenotype (very small colonies) that characterizes mip1Δ lab strains, and which is caused by defects in the mitochondrial amino acid biosynthetic machinery (Vowinckel et al., 2021). Thus, the ρ⁰ state was not likely to have spurred major compensatory cellular reconfigurations (Veatch et al., 2009) that could possibly have caused a response to paraquat distinctly different from that of the wild type. The mip1Δ cells were clearly preadapted to paraquat as their growth was equivalent to that of wild-type populations after these had undergone nine generations of adaptation to paraquat (Figure 3E). Together with the finding that the adaptation rate of mip1Δ populations exposed to paraquat was on par with the adaptation rate of wild-type cells in their second phase of adaptation (Figure 3—figure supplement 3A), this implies that loss of OXPHOS due to loss of mtDNA segments was indeed the most important factor underlying the first adaptation phase (Figure 1A).

To further test the validity of the mip1Δ results in different genetic backgrounds, we exposed the >4.700 gene knockout strains in the lab strain BY4741 yeast deletion collection to paraquat. Genes encoding mitochondrial proteins were overrepresented in the 100 most paraquat-resistant deletion strains (32.3% vs 18.6%, permutation test, p=0.00314). We also considered all gene deletion strains recently reported to have a low mtDNA copy number (mean < 1 copy per cell) (Puddu et al., 2019; Grant et al., 1997). As a group, these strains convert a cell doubling time defect in absence of paraquat (mean 16 min slower than deletion strains with a normal mtDNA copy number of 5–35 copies, p=3.05*10^–6) into a cell doubling time advantage in presence of paraquat (mean 13 min faster than deletion strains with a normal mtDNA copy number of 5–35 copies, p=0.045). In addition to being fully consistent with the mip1Δ data, these results demonstrate the considerable phenotypic penetrance of mtDNA loss across a range of genetic backgrounds and perturbed cellular physiologies.

Contemplating the possibility that mtDNA loss does not drive paraquat resistance by crippling OXPHOS, but causes enhanced cellular exclusion or inactivation of paraquat by some unknown mechanism specific to paraquat, we next exposed wild type populations, at different stages of paraquat adaptation, to menadione. As menadione is a mitochondrial O₂^∙− generator that is structurally distinct from paraquat (Fukui et al., 2012), paraquat adapted cells should in this case not be preadapted to menadione. However, they were strongly preadapted to menadione, and this preadaptation became manifest during the early paraquat adaptation, that is concomitant with the mtDNA loss (Figure 3—figure supplement 3B).

The deletion of mtDNA segments requires SOD2

Together, the above results strongly implied that mtDNA deletion was the predominant mechanism underlying the first adaptation phase. However, the data did not allow any firm judgement of whether the deletions were under regulatory control or were just due to unspecific paraquat-induced oxidative damage. We reasoned that if mtDNA deletion was part of a deliberate regulatory scheme for handling supraphysiological O₂^∙− production, then deletion of genes coding for proteins being key for this regulation would delay the adaptation. On the other hand, if the mtDNA deletions were just due to paraquat-induced unspecific oxidative damage there would be no such delay. Because paraquat exposure induces supraphysiological production of O₂^∙−, we tested these mutually exclusive outcomes by deleting the two key actors in mitochondrial redox sensing and signaling, the O₂^∙− dismutases Sod1 and Sod2 (Zou et al., 2017; Reddi and Culotta, 2013).

Cells lacking either one of the O₂^∙− dismutases showed virtually no growth when exposed to the original paraquat dose (400 μg/mL) (Figure 4—figure supplement 1A), demonstrating the importance of enzymatic O₂^∙− dismutation. The kinetics of adaptation depends heavily on the strength of selection (Couce and Tenaillon, 2015) in the sense that the adaptation rate is positively correlated with the stress level as long as the selection pressure is not overwhelming. In the case of paraquat, it was recently shown that 335 single gene deletion strains adapted near exactly as fast as predicted by the level of stress they experienced (Persson et al., 2022). In order for the stress level, that is the cell doubling time, of the mutant cells to be comparable with that of the wild type at 400 μg/mL, and thus provide a relevant deletion test, we reduced the paraquat concentration to 12.5 μg/mL. At this concentration, we found that the sod1Δ populations, in terms of reduction in cell doubling time, adapted as the wild-type populations over 10 growth cycles, while the sod2Δ populations barely showed any adaptive response (Figure 4A).

(A) Doubling time (h) in wild type (400 µg/mL paraquat: green), *sod2Δ* (12.5 µg/mL paraquat: red; 50 µg/mL paraquat: yellow) and *sod1Δ* (12.5 µg/mL paraquat: blue) cell populations adapting to equivalent stress levels. (B) mtDNA change in *sod2Δ* cell populations 2E, 4E, 9C, and 3C adapting to 50 µg/mL paraquat. *Circle:* mtDNA (77 kb) before exposure to paraquat. Genes, origins of replication and nucleotide positions (kb) are indicated. *Colored fields:* mtDNA deletions with concerted copy number change. *Diagrams:* mtDNA copy number change left y-axis, purple line, (n=2) of individual mtDNA genes during the adaptation (right y-axis, green line). Shade: S.E.M. (C) Mean growth of wild type (n=480; green), *mip1Δ* (n=288; red), *sod2Δ* (n=96; purple), and *sod2Δmip1Δ* (n=384; blue) cell populations in the presence of 400 µg/mL (left) and 50 µg/mL (right) paraquat. See also Figure 4—figure supplement 1.

Figure 4—source data 1. Doubling time data of wild type, *sod2∆, and sod1∆* populations adapting to paraquat; doubling times are in paraquat.
Data are shown in Figure 4A and B and in Figure 4—figure supplement 1B.

elife-76095-fig4-data1.xlsx^{(43KB, xlsx)}

Figure 4—source data 2. Growth curves of populations of *mip1∆sod2∆, mip1∆, sod2∆,* and wild type exposed to paraquat.
Data are shown in Figure 4C.

elife-76095-fig4-data2.xlsx^{(115.5KB, xlsx)}

Figure 4—source data 3. qPCR data for mitochondrial DNA genes and nuclear DNA controls in *sod2Δ* populations over generations of paraquat adaptation.
Data are shown in Figure 4B and in Figure 4—figure supplement 1B.

elife-76095-fig4-data3.xlsx^{(91.3KB, xlsx)}

To exclude that the lack of any adaptive response in the sod2Δ populations at 12.5 ug/mL was because the actual stress level was lower than anticipated such that the mtDNA deletion response was not triggered, we exposed eight sod2Δ populations to a paraquat concentration of 50 μg/mL. Four populations still showed very marginal adaptation and we confirmed by qPCR that their mtDNA gene copy number remained at, or near, pre-stress levels (Figure 4B). The remaining four populations had a very delayed adaptation that coincided with different, single mtDNA deletions (Figure 4B, Figure 4—figure supplement 1B). To exclude that the delay in adaptive response was due to a lower selection for mtDNA deletions shutting down OXPHOS in the sod2Δ populations, we deleted Mip1 in the sod2Δ background. In 50 μg/mL of paraquat there was a marked reduction in doubling time in sod2Δmip1Δ populations relative to sod2Δ populations. The difference was similar to the one between mip1Δ populations and wild type populations in 400 μg/mL of paraquat (Figure 4C). This implies that the delayed adaptation in sod2Δ strains was because the mtDNA deletions causing shutdown of OXPHOS emerged at a much lower rate than in the wild type.

While Sod2 is located in the mitochondrial matrix, Sod1 is located in the cytosol and mitochondrial intermembrane space. The fact that sod1Δ populations adapted as the wild type suggests that the location of the redox signaling involved in the triggering of the mtDNA deletion response is the mitochondrial matrix. Regardless of which signals and processes within the matrix that are responsible for inducing and propagating down-stream effects, the above results strongly suggest that without Sod2 the cells appear incapable of launching the speedy mtDNA deletion process we observe in the wild type.

The mtDNA editing process requires anterograde mito-nuclear communication

In budding yeast, deletion of mtDNA alters the expression of a multitude of genes resulting in increased glycolytic production of ATP (Epstein et al., 2001), dubbed the retrograde response. Upon mitochondrial OXPHOS dysfunction, the cytosolic protein Rtg2 (Sekito et al., 2000) causes the transcription factors Rtg1 and Rtg3 to translocate from the cytosol to the nucleus where they together activate retrograde transcription (Rothermel et al., 1995; Rothermel et al., 1997). This mechanism is the most well-documented channel in yeast for communicating mitochondrial dysfunction to the nucleus, and in particular mtDNA deletion (Guaragnella et al., 2018). However, activation of retrograde transcription has also been reported to cause mtDNA deletions (Farooq et al., 2013). We therefore probed whether the retrograde response is required for paraquat-induced mtDNA deletions by exposing rtg2Δ and rtg3Δ cell populations (n=16) to paraquat for 80 generations. In both cases, all cell populations failed to adapt (Figure 5A), and their capacity for respiratory growth was virtually unperturbed at the end of the experiment (Figure 5B).

Figure 5. — (A) Doubling time adaptation of 18 wild type, *rtg2Δ*, *rtg3Δ,* and *mip1Δ* cell populations to 400 µg/mL of paraquat. Shade: S.E.M. Each population type measured at n=4. (B) Respiratory (glycerol) growth of wild type, *rtg2Δ*, *rtg3Δ,* and *mip1Δ* cell populations, before (left) and after (right) 70–78 generations of paraquat adaptation. Shade: S.E.M (n=72–144 populations, each measured at n=1).

Figure 5—source data 1. Doubling time data of wild type, *rtg2∆, rtg3∆,* and *mip1Δ* populations adapting to paraquat; doubling times are in paraquat.
Data are shown in Figure 5A.

elife-76095-fig5-data1.xlsx^{(53.2KB, xlsx)}

Figure 5—source data 2. Growth curves of wild type, *rtg2∆, rtg3∆,* and *mip1Δ,* adapted and not adapted to paraquat; doubling times are in respiratory media (glycerol).
Data are shown in Figure 5B.

elife-76095-fig5-data2.xlsx^{(178.3KB, xlsx)}

Importantly, the growth of the rtg2Δ and rtg3Δ populations was similar to that of the wild type in the presence as well as the absence of paraquat. This implies that the lack of these proteins did not cause reduced growth by affecting the cellular growth physiology, such as anaplerosis. Considering the paraquat concentration used, the lack of adaptation after 80 generations of selection and the retainment of OXPHOS, the O₂^∙− production was likely on par with the production in the wild type immediately after exposure to paraquat. This puts a very restrictive upper bound on the frequency of mtDNA deletions in the wild type that are caused by unspecific oxidative damage due to paraquat-induced increase in O₂^∙− production. Thus, the data strongly support the notions that the mtDNA deletion process is under regulatory control and that this regulation is dependent on a two-way mito-nuclear communication facilitated by Rtg2 and Rtg3.

Sustained mtDNA deletion causes irrevocable mitochondrial impairment

We found that 44 of the 96 populations had completely lost their capacity for restoring the pool of intact mtDNA genomes back to pre-stress levels after 24 generations of paraquat exposure, and after 242 generations all did (Figure 6—figure supplement 1). Moreover, in the five sequenced cell populations, the capacity to restore the copy numbers of intact mtDNA and the respiratory growth after removal of stress was lost between 15 and 42 generations of paraquat exposure (Figure 6A). Together, these data suggested that the deletion of mtDNA genes continued after the first adaptation phase, ultimately leading to a complete loss of intact mtDNA genomes. We therefore assayed the mtDNA loss in 44 sequenced endpoint (t₅₀) populations. Out of these, 25 populations had almost completely lost their entire 77 kb mtDNA (97–99%), retaining only small (<1 kb) segments (Figure 6B, Figure 6—figure supplement 2A, B), in line with previous reports (Fangman et al., 1989). 18 endpoint populations remained in a rho negative (ρ^-) state, retaining 6–34 kb mtDNA segments with copy numbers somewhat above the original founder levels (mean: 20% increase). The mtDNA genes in the region spanning COX1 to VAR1 were lost in all, or almost all, 18 populations, while COX3-RPM1 and 15 S rRNA were retained in most populations, confirming the deletion bias observed in the early adaptation phase (Figure 6C). The coverage across the retained mtDNA segments was even in each of these 18 ρ^- endpoint populations, the most likely explanation being that it reflected a single continuous stretch of mtDNA (Figure 6—figure supplement 2A). To confirm this, we sequenced two end point populations (D1 and A7) with longer Nanopore reads. As expected, we found that many reads spanned the entire set of retained segments (Figure 6—figure supplement 3). This speaks strongly against a model where oxidative damage destabilizes the mtDNA and where parts of it are retained as different small fragments in different mitochondria and cells. The alignment data also suggested that some retained mtDNA segments persisted as linear tandem amplifications. To confirm this, we performed a PCR directed outwards across the ends of the retained mtDNA segments. This would produce a product only if two copies of the mtDNA segments were located next to each other on the same mtDNA molecule (Figure 6—figure supplement 2C). The PCR resulted in the expected product in the D1 endpoint population, showing that at least some of the mtDNA molecules in this population must therefore have been tandemly amplified, in line with the standard configuration of intact mtDNA in yeast cells (Maleszka et al., 1991).

Figure 6. — (A) Paraquat adapting cell populations (G=generations of exposure to paraquat) ultimately lose their capacity to recover respiratory (glycerol) growth (right y-axis, purple line, log₂ doubling time relative to founder) and the loss coincides with the genetic fixation of the paraquat adaptation (left y-axis, green line). Shade: S.E.M. of 5 populations, each measured at n=5. (B) All but one (ρ⁺) sequenced cell population adapted to long-term paraquat stress (t₅₀) retain only small (6–30 kb; ρ^-) or very small (<2 kb, ρ^--) mtDNA segments. *Panels:* Representative populations. y-axis: mtDNA copy number (median coverage in 1 kb windows relative to haploid nuclear genome). Gene positions are indicated. (C) Number of ρ^- populations after long-term paraquat exposure (t₅₀) in which the specified mtDNA gene was lost. (D) The ρ^-- populations became less fit than the ρ^- populations during a long-term exposure to paraquat. See also Figure 6—figure supplements 1–5.

Figure 6—source data 1. Mean log₂ coverage of 1 kb windows spanning the mitochondrial genome of each sequenced paraquat adapted endpoint population.
Data are shown in Figure 6B and C, Figure 6—figure supplement 2A and Figure 6—figure supplement 4B.

elife-76095-fig6-data1.xlsx^{(162.1KB, xlsx)}

Figure 6—figure supplement 1. — (A) Paraquat adapting cell populations (G=generations of exposure to paraquat) ultimately lose their capacity to recover respiratory (glycerol) growth (right y-axis, purple line, log₂ doubling time relative to founder) and the loss coincides with the genetic fixation of the paraquat adaptation (left y-axis, green line). Shade: S.E.M. of 5 populations, each measured at n=5. (B) All but one (ρ⁺) sequenced cell population adapted to long-term paraquat stress (t₅₀) retain only small (6–30 kb; ρ^-) or very small (<2 kb, ρ^--) mtDNA segments. *Panels:* Representative populations. y-axis: mtDNA copy number (median coverage in 1 kb windows relative to haploid nuclear genome). Gene positions are indicated. (C) Number of ρ^- populations after long-term paraquat exposure (t₅₀) in which the specified mtDNA gene was lost. (D) The ρ^-- populations became less fit than the ρ^- populations during a long-term exposure to paraquat. See also Figure 6—figure supplements 1–5.

Figure 6—source data 1. Mean log₂ coverage of 1 kb windows spanning the mitochondrial genome of each sequenced paraquat adapted endpoint population.
Data are shown in Figure 6B and C, Figure 6—figure supplement 2A and Figure 6—figure supplement 4B.

elife-76095-fig6-data1.xlsx^{(162.1KB, xlsx)}

To assess whether there was an adaptive advantage associated with the sustained depletion of mtDNA, we compared the doubling times of the ρ^- populations still retaining 6–34 kb mtDNA segments with those that had lost almost all their mtDNA. Intriguingly, the latter group consistently grew slower on paraquat than the former (Figure 6D). This suggests that the sustained mtDNA depletion under long-term stress, leading to irreversible loss of intact mtDNA genomes, is an artifactual response driven by prolonged induction of a regulatory mechanism dimensioned by natural selection to handle O₂^∙− stress it can successfully deal with before the pool of intact mtDNA genomes disappears.

Chromosome duplications explain the second adaptation phase

Wild type cells realized 23.3% of their adaptation potential in the second adaptation phase (Figure 1A). This may, for example, be adaptation to the effects of mtDNA loss as such (Figure 6—figure supplement 4A), or to paraquat-induced O₂^∙− production not dependent on the presence of an intact mtDNA pool (e.g. through association with the cytosolic Yno1, Rinnerthaler et al., 2012). In any case, the much slower doubling time reduction characterizing this second phase suggested that a Darwinian mutation/selection process was involved (Figure 1—figure supplement 1).

To search for nuclear genome changes that could explain the second adaptation phase we analyzed the sequence data of 44 random endpoint (t₅₀) populations. We did not find evidence for adaptive point mutations, that is mutations rarely occurred in the same gene across populations or coincided with growth improvement (Figure 6—figure supplement 4B, C). However, all but four endpoint populations carried extra copies of chromosomes II (n=29), III (n=21) and/or V (n=16) (Figure 6—figure supplement 5A) at near fixation (mean p: 0.97). In the five sequenced populations for which we had time-resolved data, these chromosome gains appeared after the early and very swift O₂^∙− adaptation phase (Figure 6—figure supplement 5B). To assess their contribution to the second phase of adaptation, we crossed clones carrying the individual aneuploidies back to wild type cells over three consecutive meiotic generations. We then compared the tolerance for paraquat, and the capacity for respiratory growth, in offspring with and without extra chromosomes. Part of the paraquat resistance and part of the loss of respiratory growth co-segregated with chromosomes II and V aneuploidies. The duplications of chromosome II and V caused a reduction in respiratory growth relative to the wild type (cell doubling times: 5.7 hr (chr II), 4.9 hr (chr V), 3.3 hr (WT)) (Figure 6—figure supplement 5C). They also reduced the cell doubling time during paraquat exposure by 31 and 38 min, respectively (Figure 6—figure supplement 5D). Duplication of chromosome III caused no apparent reduction in cell doubling time. Assuming an additive phenotypic effect of the chromosome II and V duplications, the cell doubling time would be reduced by 69 min. Together with the effect from the mtDNA deletions disrupting OXPHOS function, this would correspond to cells realizing 81% of the possible adaptation. As the cells actually realized on average 72.6%, an approximately additive phenotypic effect of these duplications appears to fully explain the second phase of adaptation in populations having both duplications genetically fixed.

Discussion

As argued above, the lack of adaptation of the sod2Δ, rtg2Δ, and rtg3Δ populations are hard to reconcile with the operation of random O₂^∙− induced mtDNA deletions (e.g. caused by genomic instability) and subsequent selection of cells possessing OXPHOS-impaired mitochondria. Instead, the data support the notion that the observed deletion of mtDNA segments causing loss of OXPHOS activity is under regulatory control. We think it is apt to denote this regulatory process as ‘mtDNA editing’ as it alters the mtDNA content to suit a particular purpose (Merriam-Webster’s Coll Thes, 2021). However, the fidelity of the mtDNA deletion mechanism in terms of which mtDNA segments are lost in the first adaptation phase does seem to be moderate. Still, there was a clear preference for deletions within the COX1-VAR1 region, and the segment containing COX3, RPM1, and 15 S RNA were not deleted in this phase, and more rarely also in the second adaptation phase.

Any mtDNA deletion that removes an enzymatic function required for electron transfer at, or before, the point of electron leakage to oxygen, such as the COB-encoded cytochrome B in the cytochrome c reductase complex, or any of the rRNA or tRNA genes that are essential to expression of these functions, would be sufficient to interrupt the electron transfer to oxygen. Thus, the moderate specificity of the mtDNA deletion mechanism makes sense from an evolutionary point of view: natural selection would not be able to increase the fidelity of the mtDNA editing program beyond the point where no further adaptation is achieved. But it should be noted that all our paraquat adapting cell populations lost either COB itself, or genes required for COB expression, during the first adaptation phase. It cannot be excluded that deletions of some individual genes, e.g. of COX1, may cause increased O₂^∙− production through mechanisms driven by paraquat or by excessive natural leakage of electrons to oxygen. But such single gene deletions cannot be present in our material for the simple reason that because of their negative effect on growth rate they would be exposed to purifying selection. That is, the gain in growth rate during the first adaptation phase is so huge that cells with a reduced growth rate due to increased O₂^∙− production would not be able to propagate.

The speedy restoration of the wild type mtDNA pool after release from paraquat is most probably due to proliferation of heteroplasmic cells that become homoplasmic through recovery of their wild type mtDNA profile. However, the lack of selective advantage for the wild type mtDNA (Figure 1B and C) implies that this recovery is not caused by a Darwinian process based on differences in cell doubling time. Thus, either the replication of the mtDNA genomes possessing deletions has to stop (Jakubke et al., 2021; Zhang et al., 2019; Chen et al., 2020), or the standing pool of these mtDNA genomes specifically has to be removed by some sort of targeted mitophagy (Twig et al., 2008; Ban et al., 2017) . Both processes may also be simultaneously operative. Regardless of the finer details of how this restoration is orchestrated, one would expect that it is under homeostatic control.

Our data bring fresh perspectives to the table concerning (i) the relationship between stress-induced mitochondrial fragmentation and canonical mitophagy (Zorov et al., 2019), (ii) under which conditions do mitochondria deprived of OXPHOS genes produce more O₂^∙− due to increased electron leakage (Aerts et al., 2017), (iii) whether clonal expansion is the main mechanism underlying the propagation of mitochondria containing deletions (Nido et al., 2018), and (iv) under which conditions, and to which degree, does the main retrograde response to mtDNA deletion in yeast mediate two-way mito-nuclear communication (Guaragnella et al., 2018). They also support the emerging notion that selective mitophagy is an important mechanism for local mitochondrial repair (Gustafsson and Dorn, 2019), and that selective mitophagy may be deliberately repressed while the cell experiences O₂^∙− stress.

Our results strongly suggest that there is an additional genetically controlled defense layer against ROS induced damage in budding yeast, situated between the primary antioxidant defenses and mitophagy. Budding yeast and humans share a range of evolutionary conserved mechanisms concerning respiratory chain biology and mitochondrial quality control, including antioxidant enzymes and key proteins regulating mitophagy (Montava-Garriga and Ganley, 2020; Kumar and Reichert, 2021; Barrientos, 2003). As our experimental yeast strain has a fully intact OXPHOS system very similar to higher eukaryotes, it is reasonable to expect that important features of the disclosed mtDNA editing mechanism may also be evolutionary conserved. In light of this, the available data on the cancer therapeutics doxorubicin and cisplatin suggest that our main results may extend to post-mitotic cells. They both accumulate in mitochondria (Kalyanaraman et al., 2002; Genc et al., 2014), where they induce O₂^∙− production through redox cycling (Malhi et al., 2012; Song et al., 2017). Their long-term administration cause oxidative injury to a variety of cells and tissues (Song et al., 2017; Songbo et al., 2019; Moruno-Manchon et al., 2018; Ren et al., 2019), and they both increase the frequency of mtDNA deletions (Genc et al., 2014; Adachi et al., 1993). Arrestment of OXPHOS activity by mtDNA editing, followed by homeostatic restoration of the pool of intact mitochondrial genomes, may therefore possibly be a generic eukaryotic adaptation to obviate a costlier mitophagic response in a variety of situations. If so, the biomedical implications in terms of the etiology of age-related disease and therapeutic opportunities appear to be noteworthy.

Materials and methods

Yeast cells

Founder strain

We used a single, haploid clone of the S. cerevisiae strain YPS128 (MATα ura3::NatMX-barcode ho::HYGMX) as the background genotype. YPS128 is a wild, oak isolate with a North American genome composition (Liti et al., 2009) whose respiratory capacity has not been impaired by domestication (De Chiara et al., 2020). In contrast to common lab-strains, it carries neither HAP1 defects, impairing mitochondrial regulation, nor MIP1 defects that lead to spontaneous mtDNA loss (Gaisne et al., 1999).

Deletion strains

We generated a ρ⁰ YPS128 strain, lacking all mtDNA, by deleting MIP1. We also constructed gene deletion strains lacking SOD2, SOD1, CCP1, ATG32, YAP1, RTG2 and RTG3. Genotypes: YPS128, MATα, ura3::natMX-barcode, ho::hygMXΔ, genex::kanMX, with each target deleted from start to stop codon. We also constructed a mip1Δsod2Δ double deletion mutant as sod2::kanMX, mip1::URA3. For all strains, two to three independent clones, verified by PCR to carry the deletion cassette and to lack the deleted gene at the target locus, were isolated and retained. The independent clones were used as replicates in all experiments. For yeast deletion collection experiments, we used the haploid BY4741 single gene deletion collection (MATa;his3Δ1;leu2Δ0;met15Δ0;ura3Δ0; genex::kanMX) (Giaever et al., 2002), which was cultivated in absence and presence of each stressor. Collection size: n=4580, each cultivated at n=6. We report the doubling time data for this collection in Data S7. Primers are reported in Source data 1.

Aneuploidic strains

Cells with and without one extra chromosome II, III or V were generated by repeatedly (3 x) backcrossing clones from endpoint (t₅₀) populations carrying chromosome duplications to founder clones of the opposite mating type (MATa ho::HYGMX). Each backcross was done on YPD (Yeast Peptone Dextrose) medium using haploids verified by qPCR to retain the chromosome duplication. Diploid hybrids were selected after three days of growth on solid minimum media (0.675% Yeast Nitrogen Base (CYN2210, ForMedium), 2% (w/v) D-Glucose, pH = 6–6.5 (NaOH), 2.5% agar) medium. They were let to sporulate overnight on solid 1% potassium acetate sporulation medium to generate recombined haploids. These were genotyped at the URA and ho locus and ura- MATα haploids were passed on to the next round of backcrossing. After three rounds of backcrossing, we selected ura- MATα haploids with (n=2 clones) and without (n=2 clones) the chromosome duplication of interest and estimated their respective growth rates (n=6) in a completely randomized design on the media of interest. We compared the cell doubling time of clones with and without the respective chromosome duplication. Primers are reported in Source data 1.

Cox4-EGFP fusion strain

To construct Cox4-EGFP fusion as a reporter for mitochondrial morphology, founder cells were transformed with PCR fragments of EGFP amplified from pYM27, with flanking regions homologous to COX4. Downstream of COX4 we inserted kanMX as selection marker during transformation. Cox4 localizes to the mitochondrial inner membrane (Zhu et al., 2019). Primers are reported in Source data 1.

Cell cultivation media

Except where otherwise stated, yeast strains were cultivated on a Complete Supplement Mixture medium (CSM medium; hereafter: “Background medium”) composed of 0.14% Yeast Nitrogen Base (CYN2210, ForMedium), 0.50% NH₄SO₄, 0.077% Complete Supplement Mixture (CSM, DCS0019, ForMedium), 2.0% (w/w) glucose, pH set to 5.80 with 1.0% (w/v) succinic acid and 0.6% (w/v) NaOH. For solid medium cultivations, 2.0% (w/v) agar was added. For pre-cultures to glycine, isoleucine, citrulline and tryptophan selection environments, we modified the background medium to avoid confounding growth on the stored nitrogen (Gutiérrez et al., 2016) by replacing CSM by 20 mg/L uracil and by reducing the NH₄SO₄ concentration (30 mg N/L). Simple modifications of the background medium were made to generate four of the eight stressor environments:+0.8 µg/mL rapamycin,+400 µg/mL paraquat (methylviologen; N,N-dimethyl-4–4′-bipiridinium dichloride),+3 mM arsenic ([As III]; NaAs₂O₃),+62.5 mg/L citric acid. To generate the four other stressor environments, we replaced NH₄SO₄ in the background medium with 30 mg N/L of L-glycine, L-isoleucine, L-citrulline or L-tryptophan, together with 20 mg/L uracil. For the respiratory growth experiments, we replaced 2% glucose with 2% glycerol. For the menadione growth experiments, we added 0.25 mM menadione to the background medium. For the vitamin C experiment, we added 180 mM of ascorbic acid (vitamin C) to background medium with and without paraquat. We cast all solid plates 10–15 hr prior to use in PlusPlates (Singer Instruments, UK), on a level surface, by pouring 50 mL of selection medium in the same upper right corner of each plate. We removed excess liquid by drying plates in a laminar airflow in a sterile environment. We stored cells at –80 °C in 20% glycerol and cultivated them at 30 °C. Populations were subsampled and transferred to and from plates using robotics (ROTOR HDA, Singer Instruments Ltd, UK), at the indicated transfer format.

Experimental evolution of cells

We single streaked and then expanded a single haploid YPS128 clone to moderate colony size (~2 million cells), sampled the colony (~50,000 cells) and expanded the sample until stationary phase (~2 million cells; 36 hr) in 5 mL of background medium. A subsample of these founder cells were stored. We poured a sample of the stationary phase culture on top of a solid plate (background medium) and allowed the lawn of cells to grow, again until stationary phase (72 hr). We then repeatedly sampled the lawn using 384 short pin pads to generate eight solid plates with 1,152 colonies each. These colonies served as pre-cultures (t_-1) to the first selection cycle of each of the eight selection environments. We expanded these pre-cultures on background, or nitrogen background, medium until stationary phase (~2 million cells; 72 hr). We transferred samples of the pre-culture with 384 short pin pads to experimental plates to generate the 1,152 populations to be evolved in each selection environments (Supplementary file 1). We then cycled all 8 × 1152 populations through 50 rounds of expansion until stationary phase (72 hr), subsampling and transfer to fresh plates, to produce t₁ to t₅₀. We evolved many (n=24–192, see figure legends) mip1Δ, rtg2Δ, rtg3Δ, sod1∆, sod2∆, and atg32Δ cell populations in a similar design, over a varying number of growth cycles. We consistently interleaved several wild type cell populations to serve as controls on the same plates.

Establishing and cultivating frozen chronological records of cell populations

In parallel to the sampling of colonies for transfer to fresh plates, we systematically sampled a large subset of populations in each environment to generate a frozen chronological record of their evolution. For each of the eight selection environments, we systematically sampled (1,536 short pin pads) the same 96 populations at the end of growth cycles 0, 1, 2, 3, 4, 5, 7, 9, 12, 15, 20, 25, 30, 35, 40, 45, and 50, to generate a dense chronological adaptation record of 768 populations. We transferred the samples to a liquid selection medium (100 μL), expanded the populations until stationary phase (72 hr), added 100 μL of glycerol (final concentration: 20% (w/w)) and stored them at –80 °C. We thawed and re-suspended these frozen stocks, and transferred cells (96 short pin pads) to a solid background, or a nitrogen background medium. To generate a randomized design, we used the randint function in the Python package NumPy (version 1.15.4). We pre-cultivated cells until stationary phase (72 hr), sampled and transferred pre-cultures (1,536 short pin pads) to selection environments plates, interleaving (384 short pin pads) 384 separately pre-cultivated, wild type, founder controls among the evolving populations on each plate. We cultivated all 1536 cells populations until stationary phase (72 hr), while tracking their growth and adaption as described below. Using the same design, we also established and cultivated a frozen chronological record of paraquat-exposed mip1Δ, rtg2Δ, rtg3Δ, sod1∆, sod2∆, and atg32Δ cell populations.

We performed three distinct release-from-selection experiments, using the frozen chronological records as start point. First, we thawed, re-suspended, sampled and transferred t₀, t₁, t₂, t₃, t₄, t₅, t₇, and t₅₀ samples of the 96 frozen paraquat-adapting populations to no stress solid medium plates. We evolved these populations over ten growth cycles (~84 generations) on no stress plates, sampled each population at the end of each growth cycle and stored samples at –80 °C (as above) to create a chronological record of samples first adapted to paraquat for different time-periods, and then released from the paraquat selection, again for different time periods. We thawed, re-suspended, sampled, randomized and pre-cultivated (no stress) this second chronological record, and sampled and transferred stationary phase cells to paraquat selection plates. Second, to compare the kinetics of loss of paraquat adaptive gains to that of populations adapting to other challenges characterized by fast adaptation, we repeated (3x) the above selection relaxation experiment, including also those adapting to arsenic and glycine. We selected the time point in the chronological record where the populations had achieved 70–90% of their endpoint adaptation. We then thawed, re-suspended and sampled these stocks, expanded revived cells under relaxed selection for 10 growth cycles and created a frozen chronological record, which was revived, randomized, pre-cultivated and cultivated in the original stress, as above (n=5). For the glycine-adapting populations a nitrogen-limited background medium was used. Third, to compare the kinetics of loss of paraquat adaptation to that of the restoration of wild type mtDNA and of respiratory growth, we again repeated the release-from-paraquat experiment, but only for the five sequenced paraquat-adapting populations (A7, A8, B12, B5, and B8). Procedures were as above, but we replicated the experiment for each sample 3x and assayed both paraquat and respiratory (2% glycerol) growth at n=5 (randomization) for each replicate.

Tracking cell growth and adaptation

Counting cells in growing populations

We assayed the growth of cell populations in all experiments using the Scan-o-matic system (Zackrisson et al., 2016), version 1.5.7 (https://github.com/Scan-o-Matic/scanomatic.git; Zackrisson, 2019). Cultivation plates were maintained undisturbed and without lids for the duration of the experiment (72 h) in high-quality desktop scanners (Epson Perfection V800 PHOTO scanners, Epson Corporation, UK) standing inside dark, temperature (30.0 C) and moisture controlled thermostatic cabinets with air circulation. We imaged plates at 20 min intervals using transmissive scanning at 600 dpi, identified the position of colonies and extracted intensities for pixels included in, and outside, each colony. For each colony, we estimated its sum pixel intensity as well as the median pixel intensity of the local background, subtracted the latter from the former and converted the remaining cell-associated pixel intensity to cell counts by using a pre-established calibration function, which had been obtained by estimating cell numbers using both spectrometry and flow cytometry. We smoothed and quality controlled growth curves, rejecting approximately 0.3% of growth curves as erroneous while being blinded to sample identities (for details, see Zackrisson et al., 2016). To allow direct visual comparison of growth curves of different samples while accounting for confounding effects from initial population size differences, we adjusted growth curves shown in figures in the y-dimension. We applied the function $N_{t, a d j u s t e d} = 2^{l o g_{2} (N_{t}) / [l o g_{2} (N_{t}) / l o g_{2} (m e d i a n (N_{0}))]}$ to the mean growth curves to be visualized in figures, where N₀ is the mean initial population size across replicates, N_t is the mean population size at time t across replicates, and the median(N₀) is the median of the mean N₀ of the samples to be visualized together.

Cell doubling time and adaptation

We extracted the cell doubling time, D, from expanding cell populations. We used the 384 fixed spatial controls introduced at every fourth position to account for systematic doubling time variations within and across plates. By interpolating across the log₂(D) values of the 384 measured controls (see Zackrisson et al., 2016) we estimated the log₂(D) value a control colony would have had in each position. From the log₂(D) value for each colony we then subtracted the corresponding log₂(D) control value, thereby obtaining a normalized, relative log₂ doubling time, log₂(D)_norm. When relevant, we also adjusted the log₂(D)_norm value for the bias associated with spatial controls having a slightly different pre-cultivation history than evolving populations, by use of the equation log₂(D)_adj = log₂(D_t)_norm - log₂(D₀)_norm, where the subscripts t and 0 refer to the growth cycle number. In some cases, we converted D_norm back to a doubling time in hours while maintaining the normalization in order to ease interpretation. This measure is denoted Doubling time in figures, and set equal to 2^DnormD_{control, grand}, where D_{control, grand} is the grand mean of the raw doubling times of all controls run in a particular experimental series.

Counting cell generations

We estimated the number of cell generations for any missing growth cycle by interpolating the values estimated for the two adjacent growth cycles. For each cell population, the total number of cell generations was calculated by summing over all growth cycles.

Maximum possible reduction in cell doubling time

We estimated how much of the maximum possible reduction in cell doubling time the paraquat adapting populations had achieved at a given generation number by comparing their cell doubling times with that of the founder population growing on normal medium, assuming that the latter represented a lower boundary for what was physiologically possible.

RNA sequencing to measure SOD1, SOD2 and CCP1 expression

Wild type cell populations were pre-cultivated for two consecutive 72 hr growth cycles on no stress background medium, sampled and transferred to background medium w. and w/o 400 µg/mL paraquat (as above). We exposed cells to paraquat for three growth cycles and then removed the paraquat for one additional growth cycle. We sampled cell populations: (i) immediately (10–15 s) after transfer (paraquat cycle 1 and 2), after 0.75 hr (paraquat cycle 1 and 2), 1.5 hr (paraquat cycle 1), 5 hr (paraquat cycle 1 and 2), 20 hr (paraquat cycle 1), and 25 hr (paraquat cycle 1). Cells to be harvested at early time-points (<5 hr after transfer) were cultivated in a 6,144 colony format, otherwise we used a 1,536 format. All samples corresponding to the same growth cycle were cultivated in parallel. To generate one replicate of one sample, we harvested all colonies on a plate by pouring 5 mL of liquid medium, w. or w/o paraquat, on top of the solid medium and scrapping off colonies with a sterile plastic rake into this liquid medium. The cells were pelleted at 12,000 G (2 min in 4 °C), re-suspended in RNAlater (Sigma Aldrich R0901) and stored at 4 °C. We extracted RNA from all the stored samples in parallel, first diluting the RNAlater solution with an equal volume of PBS and then pelleting cells at 5000 G (5 min, 4 °C). Cells were lysed by adding 600 µL of acid washed 0.5 mm beads and subsequent homogenization in a FastPrep homogenizer (three rounds at 40 s at 6 m s^–1 separated by 1 min on ice). RNA quality was determined using a Tapestation 2200 and Nanodrop (threshold; ABS_260/280 > 2.2 and RINe > 8). RNA sequencing was performed at SciLife (Stockholm, Sweden) using the Illumina TruSeq Stranded mRNA kit and a NovaSeq 6000 S4. RNA reads were checked for contamination using FastQ Screen (Wingett and Andrews, 2018). Filtered reads were aligned to the YPS128 reference genome using STAR (Dobin et al., 2013), and optical duplicates were marked with Picard-tools. The abundance of the SOD1, SOD2 and CCP1 transcripts was quantified with featureCounts from the subread package across all samples (Liao et al., 2014). We normalized their read counts as fragments per kilobases per million reads, using the DESeq2 package for R (Love et al., 2014). We estimated significant differences compared to no stress at t₀ using Wald tests and Benjamini-Hochberg FDR correction, with a cut-off of q<0.05. The normalized read counts for SOD1, SOD2, and CCP1 are reported.

DNA sequencing of evolving cell populations

Long read (PacBio) sequencing of the YPS128 founder strain

The total genomic DNA was extracted from a founder population cultivated overnight in background medium, using a standard phenol-chloroform protocol. We sequenced the genome on a PacBio RS II instrument using the P4-C2 chemistry. Additional PacBio sequencing data of the same YPS128 genotype were incorporated from and older assembly (Yue et al., 2017). A total of 9 SMRT cells were used to produce 1352628 reads, corresponding to approximately 205x genome coverage. We ran the de novo assembly using the hierarchical assembly protocol RS_HGAP_Assembly3.3 with an expected genome size of 12 Mb. Data were deposited at Sequencing Read Archive (SRA), accession number PRJNA622836.

Very long read (Oxford nanopore) sequencing

To exclude confounding effects of very early mtDNA changes, i.e. during freezing, thawing and the first round of paraquat cultivation, we thawed and single streaked frozen cells from founder (A7 position), A7 t₅₀ and D1 t₅₀ populations. We isolated and expanded one clone from each population and cultivated these in the presence of paraquat until stationary phase. DNA was extracted using Qiagen Genomic-tip 100 /G DNA extraction kit. Libraries for Oxford Nanopore sequencing were prepared using 1D Native barcoding genomic DNA with the EXP-NBD104 and SQK-LSK108kit. The flow cell version was FLO-MIN106, and the raw nanopore reads were basecalled by guppy (v2.1.3) with a minimal quality score cutoff of 5 (options: --qscore_filtering --min_qscore 5). For all basecalled reads that passed the quality filter, demultiplexing was further performed by guppy with the help of the guppy_reads_classifier.pl from LRSDAY (v1.3.1) (Yue and Liti, 2018). The de-multiplexed reads were processed by LRSDAY (v1.3.1) for adapter trimming, reads down sampling (down sampled to 50x coverage), de novo assembly, assembly polishing, assembly scaffolding, and dotplot visualisation. We deposited data at Sequencing Read Archive (SRA), accession number PRJNA622836.

Resequencing of adapted populations and populations released from selection

We thawed and subsampled frozen chronological record populations and cultivated cells in liquid medium in presence of paraquat overnight (24 hr). DNA was extracted using a modified protocol of the Epicentre MasterPure Yeast DNA Purification Kit. Pool sequencing was performed at SciLife (Stockholm, Sweden), using Illumina HiSeq2500, 2 × 126 bp. Libraries were prepared using the Nextera XT kit to accommodate the low DNA yield from small cultures. At least two founder controls were included in each flow cell.

Calling de novo point mutations

Sequenced reads were quality-trimmed and nextera transposase sequences were removed with TrimGalore (v.0.3.8). Reads were mapped to the YPS128 pacbio assembly (see above) using BWA MEM (v.0.7.7-r441). PCR and optical duplicates were flagged using Picard-tools (v.1.109 [1716]). Base alignment quality scores were calculated using samtools calmd (v.0.1.18 [r982:295]) and variants were called using Freebayes (v0.9.14–8-g1618f7e). All alleles were reported regardless of frequency or genotype model. Variants were annotated using SnpEFF (v.3.6c). Variants below a quality score of 20 and variants present in the sequenced founder samples were filtered out. Data were deposited at Sequencing Read Archive (SRA), accession number PRJNA622836.

Calling aneuploidies

Aneuploidies were called using a sliding, non-overlapping 200 bp window coverage of reads mapped. Reads with a MAPQ of <1 were not counted. The window coverage ratio was calculated as $l o g_{2} (k_{w} / w_{f o u n d e r, i})$ , where $w_{i}$ is the depth of coverage of mapped reads in each 200 bp window, $w_{f o u n d e r}$ is the depth of coverage of each i in a founder sequenced in the same flow cell, and $k = \sum_{i = 1}^{G} D_{f o u n d e r} / \sum_{i = 1}^{G} D_{s a m p l e}$ , where G is the YPS128 genome size and D is the depth of coverage for each nucleotide. Aneuploidies were called by determining the median log₂ window coverage for each chromosome.

Calling mtDNA copy number change

mtDNA copy number was calculated for each sample using a sliding, non-overlapping window of 1 kB. The mtDNA copy number relative to the euploid nuclear genome was calculated for each window as: Inline graphic $l o g_{2} (W_{i} / W_{m e d i a n, e u p l o i d})$ , where $W_{m e d i a n, e u p l o i d}$ is the median of all 1 kB windows of the nuclear genome, excluding chromosomes with detected aneuploidies. We estimated the median absolute number of mtDNA molecules across all windows, assuming one copy of the nuclear genome and no sequencing bias for mitochondrial DNA, as $2^{m e d i a n (l o g_{2} (W_{i} / W_{m e d i a n, e u p o l o i d}))}$ .

Numerical model of evolving cell populations

Cell population parameters

To generate simulated adaptation trajectories based on empirical effect sizes and mutation rates of point mutations and aneuploidies, we used an individual-based model implemented in Python (Gjuvsland et al., 2016). We repeated each simulation 1152x. We started from a haploid, isogenic founder population that was subsampled at the end of each growth cycle to found the next cultivation cycle. The population parameters were population size at the start of each growth cycle (N), the number of cell divisions before subsampling in each growth cycle (M_t) and the total number of growth cycles (n=50 cycles). When the total population size reached 2^MtN cells, N cells were sub-sampled randomly to found the next cycle. N was set to equal the approximate mean across all empirical sub-samplings. M_t was set to equal the mean (across populations) empirical measure in each growth cycle t. Each cell divided 12x before it died. Mating, meiosis, sporulation or ploidy change were not included, and there was no population structure.

Mutation effect sizes

We estimated the mutation effect sizes empirically. To estimate gene loss-of-function mutation effect sizes underlying simulations shown in Figure 1D, we used the haploid BY4741 single gene deletion collection (MATa;his3Δ1;leu2Δ0;met15Δ0;ura3Δ0; genex::kanMX) (Giaever et al., 2002), as above. To estimate chromosome duplication effect sizes, we used the duplications of chromosome II, III, V, X, and XVI constructed by backcrossing, as above. We reconstructed the chromosome duplications IV, VI, VIII, IX, XI, XII, XIII, XIV, and XV as in Zebrowski and Kaback, 2008. We genetically modified the founder clone genotype to match the his3Δ (complete deletion by transformation with pSH47) and can1::STE2pr-HIS3 genotype of the aneuploidic construct, as described in Zebrowski and Kaback, 2008, and used these as controls. Duplications of I and VII could not be obtained by either methods, despite repeated tries. Strains carrying duplications were cultivated in absence and presence of each stressor (n=9) and doubling times, D, extracted.

Mutation rate parameters

All cells began as identical, haploid founder cells. Cells had 4947 nuclear encoded protein genes, and 16 chromosomes specified by the sequenced reference genome (R64-1-1). Cells had no mitochondrial genome. Essential genes were not included. Cells independently and randomly acquired nuclear genome mutations as chromosome duplications and point mutations in protein coding genes at the end of each cell division. Mutation rates were constant, equal for all genomes, for all chromosomes and for all nucleotide sites. Chromosomes and nucleotide sites were only allowed to mutate once. Sites on new chromosomes did not mutate. Chromosome duplications occurred at rate of µ=4.85*10^–5 duplications/cell division. Point mutations occurred at a rate µ=0.33*10⁻⁹ point mutations/bp/division (Zhu et al., 2014).

Mutation effect size parameters

We tracked the mutations of each cell, its reproductive age, and its cell division time. Mutations and cell division time were passed to daughter cells. Cells began at a cell division time equal to the founder population doubling time. Change in cell division time was affected by mutations only, and mutations only affected cell division time. Because chromosome duplication and loss-of-gene function point mutations are the most common drivers of adaptive evolution in experiments (Chevereau et al., 2015), we assumed these to be the only sources of change in cell division time. We estimated the cell division effect size of chromosome duplications as described above. We estimated the cell division effect size of point mutations by downloading the SIFT yeast database (http://sift-db.bii.a-star. edu.sg/public/Saccharomyces_cerevisiae/EF4.74/) and extracting all possible stop gain base changes and nonsynonymous mutations, with attached SIFT scores (Vaser et al., 2016). All stop gain base changes and all nonsynonymous mutations with a SIFT score <0.05 affected cell division time with an effect size equal to the population doubling time effect of the corresponding gene deletion. Reproductive age did not affect cell division time and there were no cell-cell interactions. We assumed that the cell doubling time under stress at any instance could not become shorter than the measured mean founder cell doubling time in absence of stress. We implemented the well documented principle of diminishing return of mutations with increasing fitness (Chou et al., 2011; Khan et al., 2011), by letting a mutation m define the cell division time, D_m, obtained from the equation, $D_{m} = k (D_{G} - D_{f o u n d e r, n o s t r e s s}) + D_{f o u n d e r, n o s t r e s s}$ where D_G is the cell division time of the genotype before the mutation occurred, and $k = m a x ((2^{D_{r}} D_{f o u n d e r, s t r e s s} - D_{f o u n d e r, n o s t r e s s}) / (D_{f o u n d e r, s t r e s s} - D_{f o u n d e r, n o s t r e s s}), 0)$ Inline graphic . D_{founder, stress} is the measured mean doubling time of the founder in presence of stress, and D_r is the estimated doubling time effect size of the mutation assuming no epistasis. No other form of epistasis was included.

Quantitative PCR of mtDNA genes in evolving cell populations

To track the copy number dynamics of mtDNA genes in evolving populations, we performed quantitative PCR (qPCR) on the frozen chronological samples from a subset of populations (A4, A7, A8, A9, B5, B8, B12, D1). The samples were revived in 3 mL liquid background media supplemented with 400 µg/mL paraquat, expanded to stationary phase (72 hr). DNA was extracted from harvested cells using a MasterPure Yeast DNA purification kit (Epicentre), as per the manufacturer’s instructions. Primers were designed for each of the protein and rRNA encoding mtDNA genes and for one nuclear control: CDC5. The small size and extreme AT richness of the RNA subunit of RNase P (RPM1) prevented the design of working PCR primers for this mtDNA gene. We ran qPCR for duplicates of the entire frozen chronological record of one mtDNA gene in a single run, together with CDC5 controls. The qPCR was performed using iTaq Universal SYBR Green Supermix (total volume: 20 µL) and run on a Bio-Rad CFX Connect using Bio-Rad Hard-Shell PCR 96-well thin-wall plates sealed with adhesive transparent film. The PCR protocol was: initial denaturation (95 C; 15 min) followed by 45 cycles of: denaturation (95 C; 15 s), anneal (60 C; 30 s), extension (72 C; 30 s), and a melting curve analysis. We quantified PCR products at the annealing step of each cycle due to the low melting temperature of all PCR products, which follows from extremely low GC% of the mtDNA. The relative copy number was calculated as $l o g_{2} (2^{- (C t_{m t D N A} - C t_{C D C 5}) - (C t_{m t D N A} - C t_{C D C 5})_{0}})$ .

We capped all Ct values at 30. We also set Ct values to 30 for rare (5.6%) sample replicates where the COX1 primer pair produced non-PCR based background signals.

Light and fluorescence microscopy of evolving cells

We performed light and fluorescence microscopy on DNA (DAPI) stained ρ⁺ (founder, WT), ρ^- (population A7 at t₅₀), ρ^-- (population B8 at t₅₀) and ρ⁰ (mip1Δ; lacking the mitochondrial DNA polymerase) cells to validate that ρ^-- cells contained mitochondrial DNA. We cultivated cell populations overnight in liquid background medium, diluted pre-cultures in fresh media to OD₆₀₀=0.3 and incubated with agitation in 15 mL growth tubes until OD₆₀₀=0.6. Cells were pelleted by centrifugation at 3,000 G for 1.5 min. The supernatant was discarded and cells were suspended in 1 mL of ice cold 70% ethanol. Cells were fixated by incubation in RT for 5 min. The cells were then pelleted again by centrifugation, washed in non-ionic water and re-suspended in PBS (Fisher Bioreagents BP2944-100) solution. DNA was stained with DAPI (D3571, Merck) at 50 ng/mL just before imaging. Images were acquired using a Zeiss Axio Observer Z1 Inverted microscope with Plan-Apochromat 100 x/1.40 Oil DIC M27 objective and AxioCam MR R3 camera.

Electron microscopy of evolving cells

We assayed mitochondrial dynamics by electron microscopy before paraquat stress, after 5 hr (~1 cell doubling) of paraquat stress, during long-term paraquat stress (A7, t₅₀) and 5 hr after release from long-term paraquat stress (A7, t₅₀). Frozen stocks were revived by transfer to solid background medium (with or without paraquat) and cells were cultivated for 5 hr (1 population doubling). Cells were harvested by rinsing the plate with 5 mL liquid background media (with or without paraquat), suspended by stirring with a plastic spreader and pelleted by centrifugation at 600 G for 1.5 min. Pelleted cells were frozen under high pressure using a Wohlwend Compact 03 (M. Wohlwend GmbH, Sennwald, Switzerland). Freeze substitution was performed in a Leica EM AFS2 (Leica Microsystems, Vienna, Austria) by incubating the cells with 2% uranyl acetate dissolved in 10% methanol and 90% acetone for 1 hr at –90 °C (Hawes et al., 2007). Freeze substituted cells were washed (2 x) in 100% acetone and the temperature was raised 2.9 C/hr to –50 °C. Cell pellets were broken into smaller pieces to improve resin infiltration. Infiltration of cells was performed using a ladder of Lowicryl HM20 (Polysciences, Warrington, PA) diluted in decreasing acetone concentrations (1:4, 2:3, 1:1, 4:1) followed by three changes in pure Lowicryl. Each step lasted 2 hr. The resin was polymerized with UV light, first for 72 hr at –50 °C and then for 24 hr at room temperature. Resin embedded blocks were sectioned in 70 nm ultra-thin sections using a Reichert-Jung Ultracut E Ultramicrotome (C. Reichert, Vienna, Austria) equipped with an ultra 45° diamond knife (Diatome, Biel, Switzerland). Sections were collected on copper grids coated with 1% formvar and stained with 2% uranyl acetate and Reynold’s lead citrate. Stained sections were imaged at 120 kV using a Tecnai T12 microscope (FEI Co., Eindhoven, The Netherlands) and a Ceta CMOS16 camera. The IMOD package (Kremer et al., 1996) was used for quantification of the two-dimensional area covered by cells, as a proxy for cell volume, and of the two-dimensional area covered by mitochondria, as a proxy for mitochondrial volume, in 100 cell sections per sample. To validate quantifications, a large subset of images was analyzed by blind-test by a second person. Conflicting quantifications were discarded.

Confocal microscopy of evolving cells

We also tracked the mitochondrial dynamics by confocal microscopy of fluorescently labelled (Cox4-EGFP) mitochondria in cells before exposure to paraquat, after 7 hr (1 doubling) of paraquat stress, after 79 hr (1 growth cycle +1 doubling) of stress and 7 hr after release from 144 hr (2 growth cycles) of paraquat exposure. We isolated three transformants and ensured that their respiratory growth was normal by spot-assays on glycerol medium. Transformants were cultivated with or without paraquat for 0 hr or 72 hr in liquid medium, the stationary phase cultures were diluted in fresh media with or without paraquat to OD₆₀₀=0.3 and then incubated with agitation until OD₆₀₀=0.6 (exponential phase). The cells were then pelleted by centrifugation at 3000 G for 1.5 min and washed in MQ water, centrifuged and suspended in PBS. Cells were fixated by incubation at room temperature with 3.7% formaldehyde, washed 3 x in PBS, suspended in ProLong Diamond mounting media and directly mounted on slides and imaged in the microscope. Z-stacks of cells were acquired using a Zeiss Axio Observer LSM 700 inverted confocal microscopy with a Plan-Apochromat 63 x/1.40 Oil DIC M27. The signal was averaged between 4 frames to reduce noise. Z-stacks were color-coded according to Z-dimension (slice) using Temporal-Color Code in Fiji/ImageJ v. 1.52. We pre-processed images using the difference of Gaussians (DoG) to enhance the objects to be measured, that is either cells or mitochondria. To measure cells, we calculated the 3D gradient of the DoG filtered images, using triangle algorithm. To measure mitochondria, we calculated the 3D median filter of the DoG filtered images before segmentation. We segmented images using Otsu’s method, separating touching objects processed using seed-assisted watershed algorithm. The used seeds were suggested automatically, with a human blinded to sample identities correcting for mistakes. The final quantification of shape descriptors for each cell and the mitochondria within were done in MATLAB, with the relevant code and readme files available at: https://github.com/CamachoDejay/SStenberg_3Dyeast_tools, Stenberg, 2022 copy archived at swh:1:rev:a047daf337fa05f75f7cb3affb498ed70b6d7703.

Data and materials availability

Sequence data that support the findings of this study have been deposited in Sequencing Read Archive (SRA) with the accession codes PRJNA622836. The growth phenotyping code can be found at https://github.com/Scan-o-Matic/scanomatic.git; Zackrisson, 2019, the simulation code at https://github.com/HelstVadsom/GenomeAdaptation.git; Vadsom, 2017 and the imaging code at https://github.com/CamachoDejay/SStenberg_3Dyeast_tools, Stenberg, 2022 copy archived at swh:1:rev:a047daf337fa05f75f7cb3affb498ed70b6d7703. The authors declare that all other source data supporting the findings of this study are available at https://data.mendeley.com/datasets/mvx7t7rw2d. All unique strains and stored populations generated in this study are available from the Lead Contact without restriction.

Acknowledgements

We thank Lars-Göran Ottosson for help and advice with strain construction and design of adaptation experiment, Olga Kourtchenko for help with designing nitrogen-limited environments, and Tom Kirkwood for instrumental comments to an earlier version of this paper. The authors acknowledge support from the National Genomics Infrastructure in Stockholm funded by Science for Life Laboratory, the Knut and Alice Wallenberg Foundation and the Swedish Research Council, and SNIC/Uppsala Multidisciplinary Center for Advanced Computational Science for assistance with massively parallel sequencing and access to the UPPMAX computational infrastructure. The authors acknowledge PacBio sequencing technical support from the Norwegian Sequencing Centre. We acknowledge the Centre for Cellular Imaging at the University of Gothenburg and the National Microscopy Infrastructure (VR-RFI 2016–00968) for assistance with the confocal microscopy.

Funding Statement

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Contributor Information

Stig W Omholt, Email: Stig.omholt@ntnu.no.

Jonas Warringer, Email: jonas.warringer@cmb.gu.se.

Jan Gruber, Yale-NUS College, Singapore.

Jessica K Tyler, Weill Cornell Medicine, United States.

Funding Information

This paper was supported by the following grants:

Vetenskapsrådet 2014-6547 to Jonas Warringer.
Vetenskapsrådet 2014-4605 to Jonas Warringer.
Vetenskapsrådet 2015-05427 to Mikael Molin.
Vetenskapsrådet 2018-03638 to Mikael Molin.
Vetenskapsrådet 2018-03453 to Johanna L Höög.
Cancerfonden 2017-778 to Mikael Molin.
Norges Forskningsråd 178901/V30 to Stig W Omholt.
Norges Forskningsråd 222364/F20 to Stig W Omholt.
Agence Nationale de la Recherche ANR-11-LABX-0028-01 to Gianni Liti.
Agence Nationale de la Recherche ANR-13-BSV6-0006-01 to Gianni Liti.
Agence Nationale de la Recherche ANR-15-IDEX-01 to Gianni Liti.
Agence Nationale de la Recherche ANR-16-CE12-0019 to Gianni Liti.
Agence Nationale de la Recherche ANR-18-CE12-0004 to Gianni Liti.
Human Frontiers Science Program LT000182/2019-L to Johan Hallin.

Additional information

Competing interests

No competing interests declared.

Author contributions

Resources, Data curation, Software, Formal analysis, Validation, Investigation, Visualization, Methodology, Writing – original draft, Writing – review and editing.

Formal analysis, Visualization, Methodology, Writing – review and editing.

Conceptualization, Data curation, Software, Formal analysis, Funding acquisition, Validation, Methodology, Project administration, Writing – review and editing.

Formal analysis, Investigation, Visualization, Writing – review and editing.

Formal analysis, Investigation, Writing – review and editing.

Formal analysis, Investigation, Visualization.

Formal analysis, Investigation, Writing – review and editing.

Software, Formal analysis, Writing – review and editing.

Formal analysis, Investigation.

Formal analysis, Validation, Investigation, Methodology.

Resources, Software, Methodology.

Formal analysis, Investigation, Visualization, Methodology.

Formal analysis, Funding acquisition, Methodology, Writing – review and editing.

Supervision, Funding acquisition, Methodology, Writing – review and editing.

Conceptualization, Supervision, Funding acquisition, Writing – original draft, Writing – review and editing.

Conceptualization, Supervision, Funding acquisition, Writing – review and editing.

Conceptualization, Supervision, Funding acquisition, Writing – original draft, Project administration, Writing – review and editing.

Conceptualization, Supervision, Funding acquisition, Visualization, Methodology, Writing – original draft, Project administration, Writing – review and editing.

Additional files

Supplementary file 1. Description of stressor environments used as selection pressures.

elife-76095-supp1.docx^{(13.3KB, docx)}

Transparent reporting form

elife-76095-transrepform1.docx^{(245.4KB, docx)}

Source data 1. Primers used for strain construction and qPCR.

elife-76095-data1.xlsx^{(28.6KB, xlsx)}

Data availability

Sequence data that support the findings of this study have been deposited in Sequencing Read Archive (SRA) with the accession codes PRJNA622836. The growth phenotyping code can be found at https://github.com/Scan-o-Matic/scanomatic.git, the simulation code at https://github.com/HelstVadsom/GenomeAdaptation.git and the imaging code at https://github.com/CamachoDejay/SStenberg_3Dyeast_tools copy archived at swh:1:rev:a047daf337fa05f75f7cb3affb498ed70b6d7703. The authors declare that all other data supporting the findings of this study are available in the article and at https://doi.org/10.17632/mvx7t7rw2d.1.

The following previously published dataset was used:

Warringer J. 2020. Chronic superoxide distress causes irreversible loss of mtDNA segments. NCBI BioProject. PRJNA622836

References

Adachi K, Fujiura Y, Mayumi F, Nozuhara A, Sugiu Y, Sakanashi T, Hidaka T, Toshima H. A deletion of mitochondrial DNA in murine doxorubicin-induced cardiotoxicity. Biochemical and Biophysical Research Communications. 1993;195:945–951. doi: 10.1006/bbrc.1993.2135. [DOI] [PubMed] [Google Scholar]
Aerts L, Morais VA, Aerts L. Parkinson’s Disease. Elsevier; 2017. Electron transport chain; pp. 41–75. [Google Scholar]
Ammal Kaidery N, Thomas B. Current perspective of mitochondrial biology in Parkinson’s disease. Neurochemistry International. 2018;117:91–113. doi: 10.1016/j.neuint.2018.03.001. [DOI] [PMC free article] [PubMed] [Google Scholar]
Ban T, Ishihara T, Kohno H, Saita S, Ichimura A, Maenaka K, Oka T, Mihara K, Ishihara N. Molecular basis of selective mitochondrial fusion by heterotypic action between OPA1 and cardiolipin. Nature Cell Biology. 2017;19:856–863. doi: 10.1038/ncb3560. [DOI] [PubMed] [Google Scholar]
Barrientos A. Yeast models of human mitochondrial diseases. IUBMB Life. 2003;55:83–95. doi: 10.1002/tbmb.718540876. [DOI] [PubMed] [Google Scholar]
Bess AS, Crocker TL, Ryde IT, Meyer JN. Mitochondrial dynamics and autophagy aid in removal of persistent mitochondrial DNA damage in Caenorhabditis elegans. Nucleic Acids Research. 2012;40:7916–7931. doi: 10.1093/nar/gks532. [DOI] [PMC free article] [PubMed] [Google Scholar]
Castello PR, Drechsel DA, Patel M. Mitochondria are a major source of paraquat-induced reactive oxygen species production in the brain. The Journal of Biological Chemistry. 2007;282:14186–14193. doi: 10.1074/jbc.M700827200. [DOI] [PMC free article] [PubMed] [Google Scholar]
Chen Z, Wang Z-H, Zhang G, Bleck CKE, Chung DJ, Madison GP, Lindberg E, Combs C, Balaban RS, Xu H. Mitochondrial DNA segregation and replication restrict the transmission of detrimental mutation. The Journal of Cell Biology. 2020;219:e201905160. doi: 10.1083/jcb.201905160. [DOI] [PMC free article] [PubMed] [Google Scholar]
Chevereau G, Dravecká M, Batur T, Guvenek A, Ayhan DH, Toprak E, Bollenbach T. Quantifying the determinants of evolutionary dynamics leading to drug resistance. PLOS Biology. 2015;13:11. doi: 10.1371/journal.pbio.1002299. [DOI] [PMC free article] [PubMed] [Google Scholar]
Chou HH, Chiu HC, Delaney NF, Segrè D, Marx CJ. Diminishing returns epistasis among beneficial mutations decelerates adaptation. Science. 2011;332:1190–1192. doi: 10.1126/science.1203799. [DOI] [PMC free article] [PubMed] [Google Scholar]
Cochemé HM, Murphy MP. Complex I is the major site of mitochondrial superoxide production by paraquat. The Journal of Biological Chemistry. 2008;283:1786–1798. doi: 10.1074/jbc.M708597200. [DOI] [PubMed] [Google Scholar]
Couce A, Tenaillon OA. The rule of declining adaptability in microbial evolution experiments. Frontiers in Genetics. 2015;6:99. doi: 10.3389/fgene.2015.00099. [DOI] [PMC free article] [PubMed] [Google Scholar]
De Chiara M, Barré B, Persson K, Chioma AO, Irizar A, Schacherer J, Warringer J, Liti G. Domestication Reprogrammed the Budding Yeast Life Cycle. bioRxiv. 2020 doi: 10.1101/2020.02.08.939314. [DOI] [PubMed]
Dobin A, Davis CA, Schlesinger F, Drenkow J, Zaleski C, Jha S, Batut P, Chaisson M, Gingeras TR. STAR: ultrafast universal RNA-seq aligner. Bioinformatics. 2013;29:15–21. doi: 10.1093/bioinformatics/bts635. [DOI] [PMC free article] [PubMed] [Google Scholar]
Epstein CB, Waddle JA, Hale W, Davé V, Thornton J, Macatee TL, Garner HR, Butow RA. Genome-wide responses to mitochondrial dysfunction. Molecular Biology of the Cell. 2001;12:297–308. doi: 10.1091/mbc.12.2.297. [DOI] [PMC free article] [PubMed] [Google Scholar]
Fangman WL, Henly JW, Churchill G, Brewer BJ. Stable maintenance of a 35-base-pair yeast mitochondrial genome. Molecular and Cellular Biology. 1989;9:1917–1921. doi: 10.1128/mcb.9.5.1917-1921.1989. [DOI] [PMC free article] [PubMed] [Google Scholar]
Farooq MA, Pracheil TM, Dong Z, Xiao F, Liu Z. Mitochondrial DNA instability in cells lacking aconitase correlates with iron citrate toxicity. Oxidative Medicine and Cellular Longevity. 2013;2013:493536. doi: 10.1155/2013/493536. [DOI] [PMC free article] [PubMed] [Google Scholar]
Fontana GA, Gahlon HL. Mechanisms of replication and repair in mitochondrial DNA deletion formation. Nucleic Acids Research. 2020;48:11244–11258. doi: 10.1093/nar/gkaa804. [DOI] [PMC free article] [PubMed] [Google Scholar]
Frank M, Duvezin-Caubet S, Koob S, Occhipinti A, Jagasia R, Petcherski A, Ruonala MO, Priault M, Salin B, Reichert AS. Mitophagy is triggered by mild oxidative stress in a mitochondrial fission dependent manner. Biochimica et Biophysica Acta. 2012;1823:2297–2310. doi: 10.1016/j.bbamcr.2012.08.007. [DOI] [PubMed] [Google Scholar]
Fukui M, Choi HJ, Zhu BT. Rapid generation of mitochondrial superoxide induces mitochondrion-dependent but caspase-independent cell death in hippocampal neuronal cells that morphologically resembles necroptosis. Toxicology and Applied Pharmacology. 2012;262:156–166. doi: 10.1016/j.taap.2012.04.030. [DOI] [PMC free article] [PubMed] [Google Scholar]
Gaisne M, Bécam AM, Verdière J, Herbert CJ. A “natural” mutation in Saccharomyces cerevisiae strains derived from S288c affects the complex regulatory gene HAP1 (CYP1) Current Genetics. 1999;36:195–200. doi: 10.1007/s002940050490. [DOI] [PubMed] [Google Scholar]
Genc G, Okuyucu A, Meydan BC, Yavuz O, Nisbet O, Hokelek M, Bedir A, Ozkaya O. Effect of free creatine therapy on cisplatin-induced renal damage. Renal Failure. 2014;36:1108–1113. doi: 10.3109/0886022X.2014.917576. [DOI] [PubMed] [Google Scholar]
Giaever G, Chu AM, Ni L, Connelly C, Riles L, Véronneau S, Dow S, Lucau-Danila A, Anderson K, André B, Arkin AP, Astromoff A, El-Bakkoury M, Bangham R, Benito R, Brachat S, Campanaro S, Curtiss M, Davis K, Deutschbauer A, Entian K-D, Flaherty P, Foury F, Garfinkel DJ, Gerstein M, Gotte D, Güldener U, Hegemann JH, Hempel S, Herman Z, Jaramillo DF, Kelly DE, Kelly SL, Kötter P, LaBonte D, Lamb DC, Lan N, Liang H, Liao H, Liu L, Luo C, Lussier M, Mao R, Menard P, Ooi SL, Revuelta JL, Roberts CJ, Rose M, Ross-Macdonald P, Scherens B, Schimmack G, Shafer B, Shoemaker DD, Sookhai-Mahadeo S, Storms RK, Strathern JN, Valle G, Voet M, Volckaert G, Wang C, Ward TR, Wilhelmy J, Winzeler EA, Yang Y, Yen G, Youngman E, Yu K, Bussey H, Boeke JD, Snyder M, Philippsen P, Davis RW, Johnston M. Functional profiling of the Saccharomyces cerevisiae genome. Nature. 2002;418:387–391. doi: 10.1038/nature00935. [DOI] [PubMed] [Google Scholar]
Gjuvsland AB, Zörgö E, Samy JK, Stenberg S, Demirsoy IH, Roque F, Maciaszczyk-Dziubinska E, Migocka M, Alonso-Perez E, Zackrisson M, Wysocki R, Tamás MJ, Jonassen I, Omholt SW, Warringer J. Disentangling genetic and epigenetic determinants of ultrafast adaptation. Molecular Systems Biology. 2016;12:12. doi: 10.15252/msb.20166951. [DOI] [PMC free article] [PubMed] [Google Scholar]
Grant CM, MacIver FH, Dawes IW. Mitochondrial function is required for resistance to oxidative stress in the yeast Saccharomyces cerevisiae. FEBS Letters. 1997;410:219–222. doi: 10.1016/s0014-5793(97)00592-9. [DOI] [PubMed] [Google Scholar]
Guaragnella N, Coyne LP, Chen XJ, Giannattasio S. Mitochondria-cytosol-nucleus crosstalk: learning from Saccharomyces cerevisiae. FEMS Yeast Research. 2018;18:e88. doi: 10.1093/femsyr/foy088. [DOI] [PMC free article] [PubMed] [Google Scholar]
Gustafsson ÅB, Dorn GW. Evolving and expanding the roles of mitophagy as a homeostatic and pathogenic process. Physiological Reviews. 2019;99:853–892. doi: 10.1152/physrev.00005.2018. [DOI] [PMC free article] [PubMed] [Google Scholar]
Gutiérrez A, Sancho M, Beltran G, Guillamon JM, Warringer J. Replenishment and mobilization of intracellular nitrogen pools decouples wine yeast nitrogen uptake from growth. Applied Microbiology and Biotechnology. 2016;100:3255–3265. doi: 10.1007/s00253-015-7273-y. [DOI] [PubMed] [Google Scholar]
Hawes P, Netherton CL, Mueller M, Wileman T, Monaghan P. Rapid freeze-substitution preserves membranes in high-pressure frozen tissue culture cells. Journal of Microscopy. 2007;226:182–189. doi: 10.1111/j.1365-2818.2007.01767.x. [DOI] [PubMed] [Google Scholar]
Hepple RT. Impact of aging on mitochondrial function in cardiac and skeletal muscle. Free Radical Biology & Medicine. 2016;98:177–186. doi: 10.1016/j.freeradbiomed.2016.03.017. [DOI] [PubMed] [Google Scholar]
Hu H, Tan CC, Tan L, Yu JT. A mitocentric view of alzheimer’s disease. Molecular Neurobiology. 2017;54:6046–6060. doi: 10.1007/s12035-016-0117-7. [DOI] [PubMed] [Google Scholar]
Hung CHL, Cheng SSY, Cheung YT, Wuwongse S, Zhang NQ, Ho YS, Lee SMY, Chang RCC. A reciprocal relationship between reactive oxygen species and mitochondrial dynamics in neurodegeneration. Redox Biology. 2018;14:7–19. doi: 10.1016/j.redox.2017.08.010. [DOI] [PMC free article] [PubMed] [Google Scholar]
Hyttinen JMT, Viiri J, Kaarniranta K, Błasiak J. Mitochondrial quality control in AMD: does mitophagy play a pivotal role? Cellular and Molecular Life Sciences. 2018;75:2991–3008. doi: 10.1007/s00018-018-2843-7. [DOI] [PMC free article] [PubMed] [Google Scholar]
Jakubke C, Roussou R, Maiser A, Schug C, Thoma F, Bunk D, Hörl D, Leonhardt H, Walter P, Klecker T, Osman C. Cristae-dependent quality control of the mitochondrial genome. Science Advances. 2021;7:eabi8886. doi: 10.1126/sciadv.abi8886. [DOI] [PMC free article] [PubMed] [Google Scholar]
Kalyanaraman B, Joseph J, Kalivendi S, Wang S, Konorev E, Kotamraju S. Doxorubicin-induced apoptosis: implications in cardiotoxicity. Molecular and Cellular Biochemistry. 2002;234–235:119–124. [PubMed] [Google Scholar]
Khan AI, Dinh DM, Schneider D, Lenski RE, Cooper TF. Negative epistasis between beneficial mutations in an evolving bacterial population. Science. 2011;332:1193–1196. doi: 10.1126/science.1203801. [DOI] [PubMed] [Google Scholar]
Kremer JR, Mastronarde DN, McIntosh JR. Computer visualization of three-dimensional image data using IMOD. Journal of Structural Biology. 1996;116:71–76. doi: 10.1006/jsbi.1996.0013. [DOI] [PubMed] [Google Scholar]
Krishnan KJ, Reeve AK, Samuels DC, Chinnery PF, Blackwood JK, Taylor RW, Wanrooij S, Spelbrink JN, Lightowlers RN, Turnbull DM. What causes mitochondrial DNA deletions in human cells? Nature Genetics. 2008;40:275–279. doi: 10.1038/ng.f.94. [DOI] [PubMed] [Google Scholar]
Kumar R, Reichert AS. Common principles and specific mechanisms of mitophagy from yeast to humans. International Journal of Molecular Sciences. 2021;22:4363. doi: 10.3390/ijms22094363. [DOI] [PMC free article] [PubMed] [Google Scholar]
Liao Y, Smyth GK, Shi W. featureCounts: an efficient general purpose program for assigning sequence reads to genomic features. Bioinformatics. 2014;30:923–930. doi: 10.1093/bioinformatics/btt656. [DOI] [PubMed] [Google Scholar]
Liti G, Carter DM, Moses AM, Warringer J, Parts L, James SA, Davey RP, Roberts IN, Burt A, Koufopanou V, Tsai IJ, Bergman CM, Bensasson D, O’Kelly MJT, van Oudenaarden A, Barton DBH, Bailes E, Nguyen AN, Jones M, Quail MA, Goodhead I, Sims S, Smith F, Blomberg A, Durbin R, Louis EJ. Population genomics of domestic and wild yeasts. Nature. 2009;458:337–341. doi: 10.1038/nature07743. [DOI] [PMC free article] [PubMed] [Google Scholar]
Liu Y, Okamoto K. Regulatory mechanisms of mitophagy in yeast. Biochimica et Biophysica Acta. General Subjects. 2021;1865:129858. doi: 10.1016/j.bbagen.2021.129858. [DOI] [PubMed] [Google Scholar]
Lodi T, Dallabona C, Nolli C, Goffrini P, Donnini C, Baruffini E. DNA polymerase γ and disease: what we have learned from yeast. Frontiers in Genetics. 2015;6:106. doi: 10.3389/fgene.2015.00106. [DOI] [PMC free article] [PubMed] [Google Scholar]
Love MI, Huber W, Anders S. Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2. Genome Biology. 2014;15:12. doi: 10.1186/s13059-014-0550-8. [DOI] [PMC free article] [PubMed] [Google Scholar]
Maleszka R, Skelly PJ, Clark-Walker GD. Rolling circle replication of DNA in yeast mitochondria. The EMBO Journal. 1991;10:3923–3929. doi: 10.1002/j.1460-2075.1991.tb04962.x. [DOI] [PMC free article] [PubMed] [Google Scholar]
Malhi SS, Budhiraja A, Arora S, Chaudhari KR, Nepali K, Kumar R, Sohi H, Murthy RSR. Intracellular delivery of redox cycler-doxorubicin to the mitochondria of cancer cell by folate receptor targeted mitocancerotropic liposomes. International Journal of Pharmaceutics. 2012;432:63–74. doi: 10.1016/j.ijpharm.2012.04.030. [DOI] [PubMed] [Google Scholar]
Merriam-Webster’s Coll Thes CT. Merriam-Webster’s Coll Thes. 2021. [September 2, 2021]. https://unabridged.merriam-webster.com/thesaurus/edit
Montava-Garriga L, Ganley IG. Outstanding questions in mitophagy: what we do and do not know. Journal of Molecular Biology. 2020;432:206–230. doi: 10.1016/j.jmb.2019.06.032. [DOI] [PubMed] [Google Scholar]
Moruno-Manchon JF, Uzor N-E, Kesler SR, Wefel JS, Townley DM, Nagaraja AS, Pradeep S, Mangala LS, Sood AK, Tsvetkov AS. Peroxisomes contribute to oxidative stress in neurons during doxorubicin-based chemotherapy. Molecular and Cellular Neurosciences. 2018;86:65–71. doi: 10.1016/j.mcn.2017.11.014. [DOI] [PMC free article] [PubMed] [Google Scholar]
Ng MYW, Wai T, Simonsen A. Quality control of the mitochondrion. Developmental Cell. 2021;56:881–905. doi: 10.1016/j.devcel.2021.02.009. [DOI] [PubMed] [Google Scholar]
Nido GS, Dölle C, Flønes I, Tuppen HA, Alves G, Tysnes OB, Haugarvoll K, Tzoulis C. Ultradeep mapping of neuronal mitochondrial deletions in Parkinson’s disease. Neurobiology of Aging. 2018;63:120–127. doi: 10.1016/j.neurobiolaging.2017.10.024. [DOI] [PubMed] [Google Scholar]
Nissanka N, Minczuk M, Moraes CT. Mechanisms of mitochondrial dna deletion formation. Trends in Genetics. 2019;35:235–244. doi: 10.1016/j.tig.2019.01.001. [DOI] [PubMed] [Google Scholar]
Palikaras K, Tavernarakis N. Mitochondrial homeostasis: the interplay between mitophagy and mitochondrial biogenesis. Experimental Gerontology. 2014;56:182–188. doi: 10.1016/j.exger.2014.01.021. [DOI] [PubMed] [Google Scholar]
Persson K, Stenberg S, Tamás MJ, Warringer J. Adaptation of the Yeast Gene Knockout Collection Is Near-Perfectly Predicted by Fitness and Diminishing Return Epistasis. bioRxiv. 2022 doi: 10.1101/2022.03.25.485769. [DOI] [PMC free article] [PubMed]
Puddu F, Herzog M, Selivanova A, Wang S, Zhu J, Klein-Lavi S, Gordon M, Meirman R, Millan-Zambrano G, Ayestaran I, Salguero I, Sharan R, Li R, Kupiec M, Jackson SP. Genome architecture and stability in the Saccharomyces cerevisiae knockout collection. Nature. 2019;573:416–420. doi: 10.1038/s41586-019-1549-9. [DOI] [PMC free article] [PubMed] [Google Scholar]
Reddi AR, Culotta VC. SOD1 integrates signals from oxygen and glucose to repress respiration. Cell. 2013;152:224–235. doi: 10.1016/j.cell.2012.11.046. [DOI] [PMC free article] [PubMed] [Google Scholar]
Ren X, Keeney JTR, Miriyala S, Noel T, Powell DK, Chaiswing L, Bondada S, St Clair DK, Butterfield DA. The triangle of death of neurons: Oxidative damage, mitochondrial dysfunction, and loss of choline-containing biomolecules in brains of mice treated with doxorubicin, Advanced insights into mechanisms of chemotherapy induced cognitive impairment (“chemobrain”) involving TNF-α. Free Radical Biology & Medicine. 2019;134:1–8. doi: 10.1016/j.freeradbiomed.2018.12.029. [DOI] [PMC free article] [PubMed] [Google Scholar]
Rinnerthaler M, Büttner S, Laun P, Heeren G, Felder TK, Klinger H, Weinberger M, Stolze K, Grousl T, Hasek J, Benada O, Frydlova I, Klocker A, Simon-Nobbe B, Jansko B, Breitenbach-Koller H, Eisenberg T, Gourlay CW, Madeo F, Burhans WC, Breitenbach M. Yno1p/aim14p, a NADPH-oxidase ortholog, controls extramitochondrial reactive oxygen species generation, apoptosis, and actin cable formation in yeast. PNAS. 2012;109:8658–8663. doi: 10.1073/pnas.1201629109. [DOI] [PMC free article] [PubMed] [Google Scholar]
Rothermel BA, Shyjan AW, Etheredge JL, Butow RA. Transactivation by Rtg1p, a basic helix-loop-helix protein that functions in communication between mitochondria and the nucleus in yeast. The Journal of Biological Chemistry. 1995;270:29476–29482. doi: 10.1074/jbc.270.49.29476. [DOI] [PubMed] [Google Scholar]
Rothermel BA, Thornton JL, Butow RA. Rtg3p, a basic helix-loop-helix/leucine zipper protein that functions in mitochondrial-induced changes in gene expression, contains independent activation domains. The Journal of Biological Chemistry. 1997;272:19801–19807. doi: 10.1074/jbc.272.32.19801. [DOI] [PubMed] [Google Scholar]
Sedlackova L, Korolchuk VI. Mitochondrial quality control as a key determinant of cell survival. Biochimica et Biophysica Acta. Molecular Cell Research. 2019;1866:575–587. doi: 10.1016/j.bbamcr.2018.12.012. [DOI] [PubMed] [Google Scholar]
Sekito T, Thornton J, Butow RA. Mitochondria-to-nuclear signaling is regulated by the subcellular localization of the transcription factors Rtg1p and Rtg3p. Molecular Biology of the Cell. 2000;11:2103–2115. doi: 10.1091/mbc.11.6.2103. [DOI] [PMC free article] [PubMed] [Google Scholar]
Sendra B, Panadero S, Gómez-Hens A. Selective kinetic determination of paraquat using long-wavelength fluorescence detection. Journal of Agricultural and Food Chemistry. 1999;47:3733–3737. doi: 10.1021/jf990072d. [DOI] [PubMed] [Google Scholar]
Shpilka T, Haynes CM. The mitochondrial UPR: mechanisms, physiological functions and implications in ageing. Nature Reviews. Molecular Cell Biology. 2018;19:109–120. doi: 10.1038/nrm.2017.110. [DOI] [PubMed] [Google Scholar]
Sies H, Berndt C, Jones DP. Oxidative Stress. Annual Review of Biochemistry. 2017;86:715–748. doi: 10.1146/annurev-biochem-061516-045037. [DOI] [PubMed] [Google Scholar]
Song Z, Chang H, Han N, Liu Z, Liu Y, Wang H, Shao J, Wang Z, Gao H, Yin J. He-Wei granules (HWKL) combat cisplatin-induced nephrotoxicity and myelosuppression in rats by inhibiting oxidative stress, inflammatory cytokines and apoptosis. RSC Advances. 2017;7:19794–19807. doi: 10.1039/C7RA02830J. [DOI] [PMC free article] [PubMed] [Google Scholar]
Songbo M, Lang H, Xinyong C, Bin X, Ping Z, Liang S. Oxidative stress injury in doxorubicin-induced cardiotoxicity. Toxicology Letters. 2019;307:41–48. doi: 10.1016/j.toxlet.2019.02.013. [DOI] [PubMed] [Google Scholar]
Sprenger HG, Langer T. The good and the bad of mitochondrial breakups. Trends in Cell Biology. 2019;29:888–900. doi: 10.1016/j.tcb.2019.08.003. [DOI] [PubMed] [Google Scholar]
Stenberg S. Simon stenberg 3D yeast tools. swh:1:rev:a047daf337fa05f75f7cb3affb498ed70b6d7703Software Heritage. 2022 https://archive.softwareheritage.org/swh:1:dir:895f1538a0d631c088bb3ec6260151856787c4a6;origin=https://github.com/CamachoDejay/SStenberg_3Dyeast_tools;visit=swh:1:snp:ac5183da9e781260fa588c16a510bafe4091683b;anchor=swh:1:rev:a047daf337fa05f75f7cb3affb498ed70b6d7703
Sun N, Youle RJ, Finkel T. The mitochondrial basis of aging. Molecular Cell. 2016;61:654–666. doi: 10.1016/j.molcel.2016.01.028. [DOI] [PMC free article] [PubMed] [Google Scholar]
Twig G, Elorza A, Molina AJA, Mohamed H, Wikstrom JD, Walzer G, Stiles L, Haigh SE, Katz S, Las G, Alroy J, Wu M, Py BF, Yuan J, Deeney JT, Corkey BE, Shirihai OS. Fission and selective fusion govern mitochondrial segregation and elimination by autophagy. The EMBO Journal. 2008;27:433–446. doi: 10.1038/sj.emboj.7601963. [DOI] [PMC free article] [PubMed] [Google Scholar]
Vadsom H. GenomeAdaptation. swh:1:rev:d9589fad41a6d0557b039dd78cb035c7e79c46c2GitHub. 2017 https://github.com/HelstVadsom/GenomeAdaptation
Vaser R, Adusumalli S, Leng SN, Sikic M, Ng PC. SIFT missense predictions for genomes. Nature Protocols. 2016;11:1–9. doi: 10.1038/nprot.2015.123. [DOI] [PubMed] [Google Scholar]
Veatch JR, McMurray MA, Nelson ZW, Gottschling DE. Mitochondrial dysfunction leads to nuclear genome instability via an iron-sulfur cluster defect. Cell. 2009;137:1247–1258. doi: 10.1016/j.cell.2009.04.014. [DOI] [PMC free article] [PubMed] [Google Scholar]
Voordeckers K, Verstrepen KJ. Experimental evolution of the model eukaryote Saccharomyces cerevisiae yields insight into the molecular mechanisms underlying adaptation. Current Opinion in Microbiology. 2015;28:1–9. doi: 10.1016/j.mib.2015.06.018. [DOI] [PubMed] [Google Scholar]
Vowinckel J, Hartl J, Marx H, Kerick M, Runggatscher K, Keller MA, Mülleder M, Day J, Weber M, Rinnerthaler M, Yu JSL, Aulakh SK, Lehmann A, Mattanovich D, Timmermann B, Zhang N, Dunn CD, MacRae JI, Breitenbach M, Ralser M. The metabolic growth limitations of petite cells lacking the mitochondrial genome. Nature Metabolism. 2021;3:1521–1535. doi: 10.1038/s42255-021-00477-6. [DOI] [PMC free article] [PubMed] [Google Scholar]
Wingett SW, Andrews S. FastQ Screen: A tool for multi-genome mapping and quality control. F1000Research. 2018;7:1338. doi: 10.12688/f1000research.15931.2. [DOI] [PMC free article] [PubMed] [Google Scholar]
Yue J-X, Li J, Aigrain L, Hallin J, Persson K, Oliver K, Bergström A, Coupland P, Warringer J, Lagomarsino MC, Fischer G, Durbin R, Liti G. Contrasting evolutionary genome dynamics between domesticated and wild yeasts. Nature Genetics. 2017;49:913–924. doi: 10.1038/ng.3847. [DOI] [PMC free article] [PubMed] [Google Scholar]
Yue JX, Liti G. Long-read sequencing data analysis for yeasts. Nature Protocols. 2018;13:1213–1231. doi: 10.1038/nprot.2018.025. [DOI] [PubMed] [Google Scholar]
Zackrisson M, Hallin J, Ottosson LG, Dahl P, Fernandez-Parada E, Ländström E, Fernandez-Ricaud L, Kaferle P, Skyman A, Stenberg S, Omholt S, Petrovič U, Warringer J, Blomberg A. Scan-o-matic: High-resolution microbial phenomics at a massive scale. G3: Genes, Genomes, Genetics. 2016;6:3003–3014. doi: 10.1534/g3.116.032342. [DOI] [PMC free article] [PubMed] [Google Scholar]
Zackrisson M. Scan-o-Matic/scanomatic. swh:1:rev:27649ae392cec469008ea2c269ef877c1b2031c8GitHub. 2019 https://github.com/Scan-o-Matic/scanomatic
Zebrowski DC, Kaback DB. A simple method for isolating disomic strains of Saccharomyces cerevisiae. Yeast. 2008;25:321–326. doi: 10.1002/yea.1590. [DOI] [PMC free article] [PubMed] [Google Scholar]
Zhang Y, Wang Z-H, Liu Y, Chen Y, Sun N, Gucek M, Zhang F, Xu H. PINK1 inhibits local protein synthesis to limit transmission of deleterious mitochondrial DNA mutations. Molecular Cell. 2019;73:1127–1137. doi: 10.1016/j.molcel.2019.01.013. [DOI] [PMC free article] [PubMed] [Google Scholar]
Zhu YO, Siegal ML, Hall DW, Petrov DA. Precise estimates of mutation rate and spectrum in yeast. PNAS. 2014;111:E2310–E2318. doi: 10.1073/pnas.1323011111. [DOI] [PMC free article] [PubMed] [Google Scholar]
Zhu J, Zhang ZT, Tang SW, Zhao BS, Li H, Song JZ, Li D, Xie Z, Glick B, Emr SD. A validated set of fluorescent-protein-based markers for major organelles in yeast (Saccharomyces cerevisiae) MBio. 2019;10:19. doi: 10.1128/mBio.01691-19. [DOI] [PMC free article] [PubMed] [Google Scholar]
Zorov DB, Vorobjev IA, Popkov VA, Babenko VA, Zorova LD, Pevzner IB, Silachev DN, Zorov SD, Andrianova NV, Plotnikov EY. Lessons from the discovery of mitochondrial fragmentation (fission): A review and update. Cells. 2019;8:E175. doi: 10.3390/cells8020175. [DOI] [PMC free article] [PubMed] [Google Scholar]
Zou X, Ratti BA, O’Brien JG, Lautenschlager SO, Gius DR, Bonini MG, Zhu Y. Manganese superoxide dismutase (SOD2): is there a center in the universe of mitochondrial redox signaling? Journal of Bioenergetics and Biomembranes. 2017;49:325–333. doi: 10.1007/s10863-017-9718-8. [DOI] [PubMed] [Google Scholar]

eLife. doi: 10.7554/eLife.76095.sa0

Editor's evaluation

Jan Gruber ¹

Stenberg et al., explore how cells adapt to mitochondrial oxidative stress using yeast as their model system. The authors propose that reversible loss of mtDNA leads to reduced ETC function and diminished free radical production and that this represents an evolved survival mechanism. The idea that reversible loss of mtDNA may be an adaptive response under genetic control that is triggered to permit survival under adverse conditions where oxidative stress is elevated is novel and potentially important, especially given critical questions on how chronic or acute oxidative stress may contributes to loss of mtDNA integrity and mitochondrial dysfunction.

eLife. doi: 10.7554/eLife.76095.sa1

Decision letter

Editor: Jan Gruber¹

Reviewed by: Michel B Toledano²

Our editorial process produces two outputs: (i) public reviews designed to be posted alongside the preprint for the benefit of readers; (ii) feedback on the manuscript for the authors, including requests for revisions, shown below. We also include an acceptance summary that explains what the editors found interesting or important about the work.

Decision letter after peer review:

Thank you for submitting your article "Genetically controlled mtDNA editing prevents ROS damage by arresting oxidative phosphorylation" for consideration by eLife. Your article has been reviewed by 2 peer reviewers, and the evaluation has been overseen by a Reviewing Editor and Jessica Tyler as the Senior Editor. The following individual involved in review of your submission has agreed to reveal their identity: Michel B Toledano (Reviewer #1).

The reviewers have discussed their reviews with one another, and the Reviewing Editor has drafted this to help you prepare a revised submission.

Essential revisions:

Comments regarding validity of the proposed mechanism:

1) The reviewers comment that the key conclusions of the paper depend on interpretation of data that is correlative in nature and are concerned that no direct evidence is provided e.g. for the claim that mtDNA loss CAUSES specific adaptation to paraquat. Would you be able to demonstrate that the mtDNA mutation specifically reduces ROS production and that cells with these mutations are therefore able to grow in the presence of PQ.

2) The reviewers point out that yeast mitochondria produce superoxide mainly in the bc1 complex and that loss of Cox1 would be expected to increase electron leak at the bc1 complex. Please address this concern with specific reference to yeast mtDNA mutations and ROS production.

3) To expand on the above concern, an alternative explanation was suggested based on work from Prof. Alexander Tzagoloff, showing that destabilization of mtDNA may lead to irreversible mtDNA loss. The reviewer suggests that under such conditions, a percentage of cells may become rho-minus by amplifying small segments of mtDNA into tandem repeats. This reviewer is concerned that this may be what was detected by qPCR. Consistent with this notion would also be your observation that the copy number of some mtDNA segments is increased above levels seen in the wild type genome.

4) Also, a small fraction of the cells in such a population would be expected to be heteroplasmic. The reviewer suggests that such cells can then rapidly become rho plus by recovering the wild-type mtDNA and thus become homoplasmic and respiratory competent when paraquat is withdrawn. The review suggests that this scenario might result in observations similar to what you report but in this scenario the mtDNA would represent the retained mtDNA from only a few percent of the original cell population and would not have much to do with the majority of cells in the population. We would invite you to specifically address / provide further evidence for why you think that the above scenario (points 3 and 4) cannot explain the data that you obtained.

5) Related to this alternative explanation, One reviewer suggests also carefully reading and discussing some of the early literature of mtDNA genetics in yeast to re-evaluate your claims. Specifically, the reviewer suggests to discuss existing publications showing that increased oxidative stress is a condition for inducing rho- cells in which a specific mtDNA segment is amplified into tandem repeats (see full reviewer comment for details).

6) The authors observed that loss of mtDNA from mip1 cells makes them resistant to paraquat. In fact, loss of mtDNA itself causes a crisis of cell survival followed by adaptation (see Cell. 2009 Jun 26;137(7):1247-58. doi: 10.1016/j.cell.2009.04.014). mip1 cells may simply have reduced paraquat uptake or epigenetic changes to resist superoxide. Thus, mtDNA deletion as a specific adaptive process is unfounded.

7) Given these concerns, one of the reviewers felt that describing paraquat-induced mtDNA mutation as a regulatory "gene editing" program is inappropriate. You may want to address this concern, either by changing the term of arguing why you feel that it is appropriate here.

8) Finally, a fundamental concern is that budding yeast is normally anaerobic and that physiological implication of mtDNA mutations therefore would be quite different from those in mammalian cells. The reviewer suggests that more care should be taken when interpreting your data in terms of their implications to mitophagy, cancer therapy and aging-related diseases in mammals.

Further comments:

9) Figure 5 – loss of rtg2 and rtg3 may affect anaplerosis thereby reducing adaptive growth. It does not necessarily involve an antioxidant mechanism.

10) The data on chromosome duplication (Figure S9) are again just correlative rather than causative to reduced adaptive potential.

11) Figure 2 – Mitochondrial morphology changes in function of culture medium and growth stage. The data are meaningless if no vigorous controls for these parameters are in place.

12) Figure S3B – I am not convinced that the signals correspond to mtDNA. Where are the nuclei in these cells?

13) The discussion contains too many speculations and unfunded claims that are not relevant to the reported data.

14) In the abstract, the authors claim that a regulatory circuitry underlies mtDNA "editing". Where is the "circuitry" that acts on mtDNA?

15) The paper suffers from its style, which is very elliptic, the use of complicated, long sentences, the use of terms such as growth cycles, generations etc… that have not been clearly defined at start.

16) Page 4, and S1 legends says that " we see PQ causes doubling time to increase", but where do we have to see that? It is not clear how the growth rate is calculated? How are experiments performed on solid, liquid medium? The figure only shows gene expression data? What is a growth cycle and what is its length? What do you mean by 240 generations? What is the length of one generation in hours? What is a population, and what is the difference with "clonally reproducing cell populations? At best a picture of the plates used to monitor growth should be shown for one to understand how it is done. How do you calculate that 106 min doubling time reduction equals 49.3 % of the maximum possible reduction? And what is this maximum? Figure 1B is confusing: it is understandable that cells are exposed to PQ enough to adapt, then grown without PQ, and then again with PQ, but over the generations shown in the picture, do one not expect to see adaptation after a few "cycles"?

17) Page 5. How can we compare in parallel mitochondrial morphology and growth if the metrics used in the two experiments are different?

18) Page 6: “cells retained the mtDNA segments that were not lost at near…” revise the grammar of this sentence.

19) Page 7. The experiment described in S6A cannot be used to rule out signaling by H2O2: adding 3 mM H2O2 to cells that have already adapted to PQ, whether or not by use of an H2O2 signal, amounts to a severe H2O2 stress, exacerbated by the lack of a functional respiratory chain (petite cells are more sensitive to H2O2, relative to WT). One way of tackling this question would be to see whether adaptive doses of H2O2 (100-300 microM) prior to exposure to paraquat would speed up growth adaptation or not (cross adaptations have been described in the past). Similarly, the WT PQ adaptative response of cells lacking Yap1 or other antioxidants does not prove anything: signaling by H2O2 is mostly localized in confined areas, and this should persist even in a Yap1 mutant.

20) Page 8. The need of SOD2 for PQ adaptation to occur is not really convincing because of the sickness of SOD mutants in general. Further, it shows that there is no adaptation at 12 microG/mL PQ, but then adaptation occurs at a higher dose, but slower, relative to WT. What is the point authors want to make? That SOD by dismutation of the superoxide anion produces H2O2 needed for signaling ? But, authors already ruled out the need for H2O2 to signal adaptation? Please don’t be too peremptory in your conclusion on this experiment. In addition, it is hard to follow the writer: “in four populations, the copy number…” then “two of these fail to adapt” then “the remaining four populations”, but which ones? Lastly the text of Figure 4c indicates 12.5 mG/mL, but the figure 50?

21) Page 10: “the sustentation of mtDNA deletion” is complicated, rather the occurrence of mtDNA deletions.

22) Page 11. It says “the adaptation to on- or off-target effects of PQ”: clarify. Is the duplication event fixed or reversible?

We invite you to address the above points and to provide stronger support for your key claims – either by providing additional data, further interpreting the current datasets or by reference to the published literature.

Reviewer #2 (Recommendations for the authors):

The key data leading to the main conclusion are presented in Figure 3, 4 and S6. mtDNA loss seems to correlate with adaptation and mtDNA loss from the mip1 mutant seems to make cells resistant to paraquat. Figure S6D shows delayed reduction in cell doubling time that coincided with acquisition of a single mtDNA deletion in 4 out of 8 sod2 cultures. Loss of mtDNA in sod2 mip1 double mutant seems to increase adaptation for cell growth. This led to the interpretation that delayed adaptation to superoxide production is due to lower rate of mtDNA deletions. The author concluded that (1) loss of OXPHOS is the predominant mechanism responsible for adaptation; (2) Sod2 plays a role in initiating mtDNA deletion that seems to help adaptation, (3) mtDNA deletion and loss of OXPHOS is under regulatory control by “mtDNA editing”. I have numerous concerns with these conclusions.

(1) These data just show a correlation between mtDNA loss and adaptation. mtDNA is vulnerable to oxidative stress and mtDNA loss after paraquat treatment is not unexpected. No direct evidence is shown to support the idea that mtDNA loss CAUSES specific adaptation to paraquat. The authors observed that loss of mtDNA from mip1 cells makes them resistant to paraquat. In fact, loss of mtDNA itself causes a crisis of cell survival followed by adaptation (see Cell. 2009 Jun 26;137(7):1247-58. Doi: 10.1016/j.cell.2009.04.014). mip1 cells may simply have reduced paraquat uptake or epigenetic changes to resist superoxide. Thus, mtDNA deletion as a specific adaptive process is unfounded.

(2) mtDNA instability and deletions can be caused by many mutations in mtDNA. The use of the term “mtDNA editing” is inappropriate.

(3) Throughout the manuscript, the authors did not consider the dynamics of mtDNA mutations in yeast. Increased oxidative stress is a condition for inducing rho- cells in which a specific mtDNA segment is amplified into tandem repeats. Rho- mtDNAs are typically 1-2 kb in length, and can be rapidly segregated into homoplasmy within 6 generations. As such, the qPCR data for the detection of mtDNA deletions may not reflect the genetic status of most cells in cell populations. The “retained” mtDNA segments may just reflect for formation of rho- mtDNA in a cell population in which most cells are rho-zero. I am surprised that the authors can recover rho- genomes of 6 and 34 kb. Some cells are called rho-is odd. The authors should read the early literature of mtDNA genetics in yeast, please.

eLife. 2022 Jul 8;11:e76095. doi: 10.7554/eLife.76095.sa2

Author response

Essential revisions:

Comments regarding validity of the proposed mechanism:

1) The reviewers comment that the key conclusions of the paper depend on interpretation of data that is correlative in nature and are concerned that no direct evidence is provided e.g. for the claim that mtDNA loss CAUSES specific adaptation to paraquat. Would you be able to demonstrate that the mtDNA mutation specifically reduces ROS production and that cells with these mutations are therefore able to grow in the presence of PQ.

The instability of the molecule makes separating O₂^•− production and breakdown very challenging. Measuring O₂^•− levels would be insufficient to address the causality issue as these would decrease in paraquat adapted cells, even if paraquat adaptation was due to paraquat exclusion or inactivation. However, we maintain that we already have provided compelling evidence for the conclusion that mtDNA loss causes specific adaptation to paraquat.

One of the most efficient ways to test a claim about a causal relation is to do an intervention. Our mip1Δ experiment is such an intervention (please see the justification of the representativeness of the mip1Δ results in our response to comment 6). Removal of Mip1—the sole mitochondrial DNA polymerase—is the gold standard to validate effects of mtDNA loss/reduction in yeast because in contrast to other proteins associated with mtDNA maintenance, and in contrast to ethidium bromide treatment, Mip1 is not known to affect cellular biology through mechanisms other than those mediated by the mtDNA loss. There is absolutely no support for the possibility that Mip1 would be responsible for paraquat exclusion or inactivation, neither in the literature nor in our data. If the mtDNA deletions had nothing to do with the observed adaptation and were just a side effect due to random oxidative mtDNA damage, then mip1Δ cells would not be preadapted to paraquat. But in the original manuscript, we showed that mip1Δ cells are indeed preadapted to paraquat.

The mip1Δ intervention demonstrates that cells lacking OXPHOS activity are much more capable to grow in the presence of paraquat than wild-type cells. Since the most prominent effect of paraquat is that it causes enhanced O₂^•− production in OXPHOS-active mitochondria by redox-cycling, it follows that the O₂^•− production is lower in mip1Δ cells than in wild type cells when they are exposed to the same paraquat concentration. This explains the mip1Δ cells’ resistance to paraquat. Indeed, it has been extensively documented that the natural leakage of electrons to oxygen is 2-fold lower in fermenting rho0 yeast cells (1). Now, one may object that the total lack of mtDNA in mip1Δ cells cannot be directly compared with the more modest mtDNA deletions we observe in the early, but still most prominent, adaptation phase. However, even small deletions in the mtDNA tend to cripple the OXPHOS system (e.g. (2)). And our mtDNA segmental deletions cover several protein, rRNA and tRNA encoding genes that are absolutely required for the expression or function of OXPHOS. Moreover, paraquat-mediated shuffling of electrons occurs at complex III, with cytochrome B having the critical enzymatic role (3). Cytochrome B is encoded in the mtDNA gene COB, and as all our paraquat adapting cell populations lose either COB itself, or genes required for COB expression, it follows that paraquat dependent O₂^•− production is perturbed in our paraquat adapting cell populations.

As a response to the concerns of the reviewers, we exposed paraquat-adapted populations to menadione and found them to be much more resistant than wild type cells (Figure 3—figure supplement 3B). Like paraquat, menadione enhances mitochondrial O₂^•− production(4, 5). But the compound is structurally completely distinct from paraquat, and if an export or inactivation mechanism was responsible for the adaptation to paraquat, it is unlikely that menadione would be exported or inactivated through exactly the same mechanism. We therefore think it is highly legitimate to claim that the preadaptation of paraquat-adapted cells to menadione is because menadione causes much less ROS production in these cells than in wild-type cells. As the source for the ROS production is the OXPHOS system, it follows that the OXPHOS system has to a large degree become deactivated in cells adapted to paraquat. With reference to the paragraph above, the absolutely most straightforward explanation of this deactivation is that it is caused by mtDNA deletion.

We also added vitamin C to paraquat exposed cells. Addition of vitamin C, an antioxidant that accepts electrons from PQ⁺, the free radical (or ‘damaging’) state of paraquat(6), caused the doubling time of not only paraquat exposed wild type cells, but also of paraquat exposed sod2Δ cells, to be on par with that of unexposed wild type cells (Figure 1—figure supplement 1C, D). Thus, the paraquat concentration of 400 µg/mL impaired cell growth entirely through its effect on O₂^•− production, and the doubling time adaptation observed must be understood solely as an adaptation to the elevated O₂^•− production.

Finally, we measured the growth of all viable deletion strains in the BY4741 deletion collection in the presence and absence of paraquat. We found a strong tendency for the loss of mitochondrial proteins to lead to paraquat resistance. Moreover, gene deletion strains reported to have low mtDNA copy numbers tended to grow slower than the wild type in absence of paraquat, but better than the wild-type in presence of paraquat (7). While these deletion strains are less suitable as controls than mip1Δ, because of their many varied functions that affect ROS resistance and growth independently of their effect on mtDNA, the results are still fully in tune with the mip1Δ and menadione results, supporting the conclusion that “mtDNA loss CAUSES specific adaptation to paraquat”.

Thus, we find it difficult to acknowledge the possibility that mtDNA deletions leading to shut-down of OXPHOS activity, and consequently shut-down of both natural and paraquat-induced mitochondrial O₂^•− production, is just correlated with the observed paraquat adaptation.

However, the intervention data and the other data discussed above establish only that paraquat causes mtDNA deletion and that this mtDNA deletion causes the adaptation to paraquat. They do not allow any conclusion about whether this is due to unspecific mitochondrial damage or deliberate regulation. We deal with this issue below.

We have renamed the section “mtDNA loss drives the first adaptation phase” to “mtDNA segmental deletions cause the swift adaptation to paraquat”, included new data and analyses, and expanded the text considerably to make our reasoning more explicit.

2) The reviewers point out that yeast mitochondria produce superoxide mainly in the bc1 complex and that loss of Cox1 would be expected to increase electron leak at the bc1 complex. Please address this concern with specific reference to yeast mtDNA mutations and ROS production.

The lost segments cover several genes encoding proteins, rRNA and tRNA. When we talked about the COX1-VAR1 region which was shown in our plots, we were explicitly talking about the genes COX1, COB, ATP6, ATP8, OLI1 and VAR1. And due to a lack of well-functioning qPCR probes (the tRNA genes are too short to be targeted by qPCR directly) we were implicitly talking about tRNA genes in between these. The observed loss of multiple mtDNA segments containing genes essential for the expression and function of OXPHOS, and the observed loss of respiratory capacity accompanying mtDNA deletions, hardly open for any other interpretation than what we proposed, i.e. the mtDNA deletions cause a reduced OXPHOS activity and electron flow through the ETC in the affected cells. Without the latter, and, in particular, without the enzymatic cytochrome B function in complex III, paraquat is not capable of producing O₂^•− through redox-cycling.

We cannot exclude that deletions of some individual genes in complex IV, e.g. of COX1 which codes for the enzymatic function of complex IV, may cause increased O₂^•− production through mechanisms driven by natural leakage of electrons to oxygen, or by paraquat. But such single gene deletions are not present in our material, and cannot be, for the simple reason that they would be totally wiped out by natural selection because of their negative effect on growth rate. The adaptive gain in growth rate during the first adaptation phase is huge, which implies that cells with a reduced growth rate due to increased O₂^•− production would not be able to propagate in a population. In other words, after the first adaptation phase, every cell population produces less O₂^•− than what they did immediately after becoming exposed to paraquat because the vast majority of mitochondrial genomes possess mtDNA deletions that have effectively shut down OXPHOS activity.

In the section “mtDNA segmental deletions cause the swift adaptation to paraquat” we now mention tRNAs specifically, and we have included a paragraph under Discussion explaining why we will not find mtDNA deletions causing enhanced O₂^•− production.

3) To expand on the above concern, an alternative explanation was suggested based on work from Prof. Alexander Tzagoloff, showing that destabilization of mtDNA may lead to irreversible mtDNA loss. The reviewer suggests that under such conditions, a percentage of cells may become rho-minus by amplifying small segments of mtDNA into tandem repeats. This reviewer is concerned that this may be what was detected by qPCR. Consistent with this notion would also be your observation that the copy number of some mtDNA segments is increased above levels seen in the wild type genome.

We believe the reviewer envisions a paraquat-induced mtDNA destabilization process that is exclusively driven by unspecific oxidative mtDNA damage and which is accompanied by natural selection for cells with disrupted mtDNAs causing reduced OXPHOS activity. The resulting fragmentation of the mtDNA genome, with cells amplifying different small fragments into tandem repeats, which are retained and inherited through unknown mechanisms, would in this model be what we detect with qPCR and the sequence data. This explanation can in principle account for the intervention data and the other data alluded to above. However, it is strongly contradicted by the sequence data, the qPCR data, the Rtg2/3 data and the data on respiratory ability.

As shown in Figure 6—figure supplement 2A, sequence coverage across both retained and deleted regions, is remarkably even. This is fully consistent with segmental deletions, but highly inconsistent with fragmentation and amplification of different small segments to various extents. We also visually inspected sequence data for five populations (A7, D1, A11, A5 and D9) using the Integrated Sequence Viewer (https://software.broadinstitute.org/software/igv/), which specifically marks tandem amplifications based on changes in the read orientation of read-pairs. We found tandem amplifications of small fragments to be exceedingly rare and similar in number and size to what we observed in these populations before paraquat exposure. Moreover, the qPCR data for one of our large segments was in fact constituted by several very small qPCR probes complementary to small sections of DNA in adjacent genes along the segment. And it is extremely unlikely that several qPCR probes will show the same mtDNA copy number, and the same change in mtDNA copy number unless they are part of the same continuous mtDNA segment. In the same vein, mtDNA fragmentation and amplification of small segments would not be expected to be consistent with the qPCR data either, as it would be highly unlikely that our small qPCR probes would, by coincidence, correspond to the small mtDNA fragments that have been amplified.

To further demonstrate that amplification of many small mtDNA fragments in different cells can be ruled out, we isolated single cells from two paraquat adapted populations, clonally expanded these and sequenced their genomes with long-read nanopore sequencing. With this sequencing technology many of the reads have a length that approaches the length of the retained segments. The read alignment shows that most of the long reads span the entire, or almost the entire segment, i.e. they correspond to one, intact mtDNA molecule and not many small fragments (Figure 6—figure supplement 3). In contrast, despite the dynamic nature of mtDNA, we rarely observed tandem amplifications of smaller mtDNA segments.

As the majority of the yeast mtDNA is normally mostly composed of linear, tandemly duplicated mtDNA molecules, we probed whether the retained long mtDNA segments existed as tandem duplications in very long mtDNA molecules as well. We did this by letting the PCR be directed outwards from segment ends, across a potential breakpoint (Figure 6—figure supplement 2C). The PCR data show that at least in one cell population, a tandem arrangement of the retained segment exist. And the existence of tandem duplications is also supported by the read alignments (Figure 6—figure supplement 3). Although this result is a side-point, we think it is worthwhile to point it out, as it is entirely in line with normal yeast mtDNA biology.

We exposed rtg2Δ and rtg3Δ cell populations (n=16) to paraquat (400 μg/mL) for 80 generations. In both cases, all cell populations failed to adapt (Figure 5A), and their capacity for respiratory growth was virtually unperturbed at the end of the experiment (Figure 5B). Importantly, the growth of the rtg2Δ and rtg3Δ populations was similar to that of the wild type in the presence, as well as the absence, of paraquat. This implies that the lack of these proteins did not cause reduced adaptive growth by affecting the cellular growth physiology, such as anaplerosis. Considering the paraquat concentration used, i.e. rtg2Δ and rtg3Δ cells experiencing the same paraquat stress as wild type cells, the lack of adaptation after 80 generations of selection and the retainment of OXPHOS function, the O₂^•− production was arguably, throughout the 80 generations, on par with the production In the wild type immediately after exposure to paraquat. This puts a very restrictive upper bound on the frequency of mtDNA deletions that are caused by unspecific oxidative damage due to the increase in O₂^•− production following paraquat exposure. It follows that unspecific oxidative damage is incapable of explaining the swiftness of the mtDNA deletion process.

Moreover, our data on respiratory ability clearly show that we are not dealing with a small percentage of cells having become rho-minus. The vast majority of cells must experience an almost complete deprivation of OXPHOS ability during the initial very swift adaptation phase in order for these data to make sense.

An increase in copy numbers of segments of the mtDNA are frequently encountered in connection with clonal expansion. However, due to the above, we find it much more likely that the temporary increase we observe in our data, and its disappearance during restoration back to wild-type levels, is part of the homeostatic regulatory response (as pointed out in the Discussion).

We have now expanded the text under the sections “mtDNA segmental deletions cause the swift adaptation to paraquat” and “The mtDNA deletion process requires mito-nuclear communication” to highlight that the above model is not capable of explaining the data. However, despite some effort, we have not been able to identify papers by A. Tzagoloff that should be cited in this connection. The provision of some further coordinates would be highly appreciated.

4) Also, a small fraction of the cells in such a population would be expected to be heteroplasmic. The reviewer suggests that such cells can then rapidly become rho plus by recovering the wild-type mtDNA and thus become homoplasmic and respiratory competent when paraquat is withdrawn. The review suggests that this scenario might result in observations similar to what you report but in this scenario the mtDNA would represent the retained mtDNA from only a few percent of the original cell population and would not have much to do with the majority of cells in the population. We would invite you to specifically address / provide further evidence for why you think that the above scenario (points 3 and 4) cannot explain the data that you obtained.

Concerning the statement: “Also, a small fraction of the cells in such a population would be expected to be heteroplasmic. The reviewer suggests that such cells can then rapidly become rho plus by recovering the wild-type mtDNA and thus become homoplasmic and respiratory competent when paraquat is withdrawn.” This ‘point 4’ scenario does indeed explain parts of the data concerning the recovery after short-term paraquat stress, and it was part of the explanation we proposed in the original manuscript to account for the rapid restoration of wildtype mtDNA copy numbers in populations released from short-term paraquat stress (see end of section “Sustained mtDNA deletion causes.” in the original manuscript). However, because the data tell us that there is no Darwinian selection pressure favoring the removal of mtDNA genomes causing loss of OXPHOS activity, this scenario does not account for the speed at which this restoration occurs, i.e. it does not explain why there is not a sustained replication of non-intact mtDNA genomes after release from paraquat. In fact, according to the current clonal expansion theory, one would expect that mtDNA genomes possessing deletions would outcompete the wild-type mtDNA genomes. But this is contrary to what we observe.

More specifically, the mtDNA genomes possessing deletions causing cessation of OXPHOS activity are faithfully replicated until the cells are released from paraquat. After the release, the heteroplasmic cells “then rapidly become rho plus by recovering the wild-type mtDNA and thus become homoplasmic and respiratory competent”. But in order for this to happen with the speed we observe, either the replication of the mtDNA genomes possessing deletions has to stop or the standing pool of these mtDNA genomes has to be removed by some sort of mitophagy. Both processes may also be simultaneously operative. In any case, it implies that a considerable amount of regulation has to be introduced in order to explain our experimental data. Such a regulatory machinery is well supported by recent publications, and we now expand on this in the discussion.

If the mtDNA deletions are just due to unspecific oxidative damage then, according to the vast literature on the topic, one would expect that while the cells are exposed to paraquat, the mitophagic machinery would be activated to remove these mitochondria. But it is not. And after release from paraquat, one would expect, according to current clonal expansion theory, that mtDNA genomes possessing deletions would outcompete the wildtype mtDNA genomes in individual cells. But they do not. Instead they disappear remarkably fast. Thus, one may claim that the existing literature implies that the combined ‘points 3 and 4’ explanation is internally inconsistent, as well as being in conflict with the experimental data by not acknowledging any sort of regulatory control of the replication and maintenance of non-intact mtDNA genomes as a function of paraquat exposure.

So, if there is no deliberate regulation involved, why do two genes documented to play a key role in mito-nuclear communication, in particular in connection with mtDNA deletions, have such an impact on the adaptation to paraquat? The fragmentation hypothesis suggested under point 3 is not in conflict with the existence of signaling from the mitochondria to the nucleus as a result of mtDNA deletion. But how does it explain that the supposed random mtDNA deletion process due to oxidative damage is hampered in rtg2Δ and rtg3Δ populations? The Rtg2/3 signaling from the mitochondria to the nucleus activates a whole range of responses dedicated to compensate for the anticipated damage and restore cellular integrity. Thus, a direct prediction by the «point 3 fragmentation» hypothesis is that if there is a cellular response from deletion of Rtg2 or Rtg3, it should be opposite of what we actually observe, i.e. more damage, more fragmentation and more mtDNA deletion. We therefore think that in this case the «point 3 fragmentation» hypothesis is in direct conflict with the data.

In conclusion: Yes, the ‘points 3 and 4’ scenario is capable of explaining some of the experimental data we provided. But we think it is fair to demand that an alternative explanatory scenario is, at least to some degree, able to account for all the data we presented in the original manuscript and not just a subset. In any case, it should not be in conflict with any of the data, unless it is justified that these data can be neglected. However, the ‘points 3 and 4’ scenario is in direct conflict with the Rtg2/3 results, the qPCR and sequence data. And these can hardly be dismissed as irrelevant. In addition, the ‘points 3 and 4’ scenario is incapable of explaining the dynamics of non-intact mtDNA genomes during and after paraquat exposure. Thus, we think it is fair to claim that this scenario does not represent a valid alternative explanatory scheme.

In addition to the revisions mentioned under the previous point, we revised one of the paragraphs under Discussion to make some of the above reasoning clearer, and we contrast our explanatory scenario with the one proposed above, using their mutually exclusive predictions.

5) Related to this alternative explanation, One reviewer suggests also carefully reading and discussing some of the early literature of mtDNA genetics in yeast to re-evaluate your claims. Specifically, the reviewer suggests to discuss existing publications showing that increased oxidative stress is a condition for inducing rho- cells in which a specific mtDNA segment is amplified into tandem repeats (see full reviewer comment for details).

As explicated above, our data do not support the amplification of tandem repeats of small mtDNA fragments. The newly generated long read Nanopore data show that some of the mtDNA molecules in our paraquat adapted populations exist as linear, tandem duplications of the entire mtDNA segment, as discussed below. But such a duplicated linear arrangement is the normal state for yeast mtDNA and is not in any way in conflict with the argumentation in the paper. So, we do not see the need to re-evaluate our claims. But we have revisited the early literature, and we are incapable to see how this literature contradicts the existence of an active mtDNA deletion process. We may have overlooked something, but to find this information it needs to be pointed out to us.

We have now included to two older references:

– Fangman WL, Henly JW, Churchill G, Brewer BJ (1989) Stable maintenance of a 35-base-pair yeast mitochondrial genome. Mol Cell Biol 9(5):1917–1921.

– Maleszka R, Skelly PJ, Clark-Walker GD (1991) Rolling circle replication of DNA in yeast mitochondria. EMBO J 10(12):3923–9.

6) The authors observed that loss of mtDNA from mip1 cells makes them resistant to paraquat. In fact, loss of mtDNA itself causes a crisis of cell survival followed by adaptation (see Cell. 2009 Jun 26;137(7):1247-58. doi: 10.1016/j.cell.2009.04.014). mip1 cells may simply have reduced paraquat uptake or epigenetic changes to resist superoxide. Thus, mtDNA deletion as a specific adaptive process is unfounded.

We interpret this comment to say that mip1Δ cells will undergo dramatic cellular reconfiguration to adapt to the loss of mtDNA even when growing on a rich glucose medium, and that this reconfiguration causes reduced paraquat uptake or epigenetic changes to resist paraquat. We have already dealt with this in our reply to Comment 1 above. But we want to restate that to the best of our knowledge there is absolutely no evidence in the literature supporting that mip1Δ, or mtDNA loss, would be responsible for paraquat exclusion or inactivation. The cited Cell paper does not provide such a backing. Moreover, it describes a mechanism for mtDNA loss in response to replicative aging, an inherently multifactorial process, and any crisis of cell survival would then need to be considered in the context of the other aging factors also accumulating.

Moreover, in terms of cell division time, our mip1Δ populations do by no means indicate that they have undergone a crisis of cell survival. Yes, their doubling time on glucose without paraquat is higher than that of the wild type (3.5. vs. 1.5h), so they certainly suffer from the total lack of mtDNA. But this slowing-down of growth is hardly large enough to justify a claim about “a crisis of cell survival”. We constructed and generated this deletion strain in the lab. Considering the short time span (20-30 cell generations) that passed from the isolation of mip1Δ clones to the onset of paraquat experiments, it seems quite unlikely that a purely Darwinian process of mutation and selection would be able to create and fix solutions that dramatically change the physiological configuration of mip1Δ cells. As mentioned above, paraquat exposure tends to suppress the growth defects of rho0 deletion strains, which instead enjoy a growth advantage compared to the wild type. Thus, the argument that the preadaptation of the mip1Δ cells to paraquat is because of effects other than their complete lack of an active ETC system is hard to reconcile with the data, as well as our knowledge of how paraquat acts.

Moreover, our respiratory capacity experiments demonstrate that a large fraction of paraquat adapted populations was totally stripped of OXPHOS activity shortly after the first adaptation phase. This is consistent with the conception that their large mtDNA deletions prevent the expression of a functional OXPHOS. Thus, the cells in these populations do indeed lack an intact ETC system, and this, in turn, impairs both paraquat-induced O₂^•− production at complex III, as well as base-line O₂^•− production. The fact that these cells homeostatically maintain the copy numbers of undeleted mtDNA, while at the same time growing dramatically better than the wild type, implies that the concerns attached to the mip1Δ cells do not apply here. Still, their behavior is fully consistent with the behaviors of the mip1Δ cells and the rho0 deletion strains. And, still fully in line with the mip1Δ results, the growth rate of every single paraquat adapted cell population becomes much less perturbed by the ROS-inducer menadione than that of wild type populations. One can of course claim that this preadaptation of the paraquat-adapted cells to menadione is not caused by their demonstrated lack of OXPHOS activity due to mtDNA deletions, but that each and every one of them developed independently other mechanisms that prevent uptake of menadione. However, as we find absolutely no support for this idea in the sequence data, we cannot see any biological justification for why the most straightforward explanation should be discarded.

With all due respect, we think our responses to points 1, 2, 3, 4 and 6 show very clearly that the statement “mtDNA deletion as a specific adaptive process is unfounded” is not justified. Instead, we think our comments on points 1-6 fully justify that there is ample reason for claiming that mtDNA deletion drives the adaptation to supraphysiological mitochondrial oxidative stress. Given how beneficial these mtDNA deletions are in terms of fitness, and how rapidly they emerge and disappear, the most straight-forward conclusion is that the mtDNA deletion represents an evolutionary favored defense mechanism over which the cell has some level of regulatory control. And this is very strongly supported by the results showing that the process depends on Sod2, Rtg2 and Rtg3.

We now deliberately deal with this concern regarding mip1 cells under the section “mtDNA segmental deletions cause the swift adaptation to paraquat”.

7) Given these concerns, one of the reviewers felt that describing paraquat-induced mtDNA mutation as a regulatory "gene editing" program is inappropriate. You may want to address this concern, either by changing the term of arguing why you feel that it is appropriate here.

According to Merriam-Webster Unabridged one of the meanings of the transitive verb edit is “to alter, adapt, or refine especially to bring about conformity to a standard or to suit a particular purpose». As the mtDNA deletions cause cessation of paraquat-induced supraphysiological mitochondrial O₂^•− production and that the deletion process depends on the retrograde pathway, we think we are fully justified to use the term gene-editing in the sense that the cells deliberately alter their mtDNA profile to suit a particular purpose.

Justification of the term is made more explicit in the Discussion, and now the term is introduced for the first time in the Discussion. In line with this, we have also replaced “edit” with “deletion” in the title.

8) Finally, a fundamental concern is that budding yeast is normally anaerobic and that physiological implication of mtDNA mutations therefore would be quite different from those in mammalian cells. The reviewer suggests that more care should be taken when interpreting your data in terms of their implications to mitophagy, cancer therapy and aging-related diseases in mammals.

We agree that the evolutionary distance between yeast and humans should have been given greater consideration. However, we intentionally used a wild yeast strain for our experiments. The perception that yeast is very strongly oriented towards fermentative metabolism derives from experiments on domesticated strains. Recent comparisons of wild to domesticated yeast show that domestication has led to a loss of respiratory growth capacity in domesticated lineages and a gain of fermentative growth capacity (8). Extreme orientation towards fermentative growth may thus constitute a relatively recent yeast adaptation to man-made niches with highly concentrated sugar, while a more extensive aerobic respiration, and concomitant higher O₂^•− production, is a natural part of the biology of wild and ancestral yeast lineages. The physiological implications of mtDNA deletion in the context of aerobic respiration is thus not likely to be very different from those of mammalian cells.

It is therefore entirely reasonable to be open for the possibility that the disclosed mtDNA editing system emerged in the shared ancestor of yeast and men, and that it has been retained in both lineages – as is the case for core components of both the antioxidant defense (e.g. Sod1, Sod2), and mitophagy(9–14).

We have made these points clearer in the text, while at the same time being less conclusive on whether our results extend to mammals.

Further comments:

9) Figure 5 – loss of rtg2 and rtg3 may affect anaplerosis thereby reducing adaptive growth. It does not necessarily involve an antioxidant mechanism.

Indeed, rtg∆ strains have a glutamate auxotrophic phenotype (15, 16). The RTG pathway is believed to be activated by glutamate starvation, and the activations result in the induction of Cit1, Cit2, Aco1, Idh1 and Idh2 expression. These enzymes catalyze the first three steps from the Krebs cycle, generating glutamate precursors (17). However, our experiments were performed on a synthetic complete medium, in which a very substantial excess of glutamate and other amino acids and nucleotides is added externally. This suppresses and vastly reduces the need for glutamate, and for anaplerosis more generally. Moreover, we would expect anaplerosis to be important in both the presence and absence of paraquat, and perhaps more important in absence of paraquat when growth and metabolic needs are generally higher. However, the growth of rtg2Δ and rtg3Δ remained as for wild-type cells in both the presence and absence of paraquat. Thus, any anaplerotic effect is clearly within bounds the cells can handle well. Moreover, the fact that cells missing Rtg2 and Rtg3 not only fail to adapt to paraquat, but also retain normal respiratory growth, shows that the mtDNA remains functional in these cells. Clearly, if paraquat was just causing unspecific oxidative damage of mtDNA leading to mtDNA deletion, then one would not observe this mtDNA retainment even if the deletion of Rtg2 and Rtg3 just affected anaplerosis and caused reduced growth. So, this explanation can hardly be reconciled with the Rtg2/3 results.

Thus, we think the most parsimonious explanation of the Rtg2/3 results is that these two proteins are critical for the observed rapid emergence of mtDNA deletions by being mediators of two-way mito-nuclear communication being part of the deliberate regulatory response to excess mitochondrial O₂^•− production. We can, at this stage, not discount completely that glutamate deficiency, e.g. by very extensive oxidative damage to glutamate uptake mechanisms, plays a role in activating RTG under ROS exposure. However, whether glutamate deficiency or a pure ROS signal activate the RTG pathway has no bearing on the conclusion that the RTG pathway is necessary for the mtDNA deletions that drive the O₂^•− adaptation. Hence, speculation along these lines is premature in this paper.

We have addressed this concern in the section “The mtDNA deletion process requires mito-nuclear communication”.

10) The data on chromosome duplication (Figure S9) are again just correlative rather than causative to reduced adaptive potential.

If we only had observed the emergence of chromosome duplications in the later stages of adaptation, then this objection would be fully justified. But, as reported in the original manuscript, we validated a causal effect through an intervention experiment. We backcrossed adapted cells with chromosome duplications to unadapted founder cells over three consecutive meiotic generations to generate highly recombined gametes with and without chromosome duplications (i.e. the intervention). As a part of the paraquat tolerance and a minor part of the loss of respiratory growth co-segregated with two of the chromosome duplications across these three meiotic generations, the data are clearly not just correlative.

We cannot completely exclude that undetected mutations contributing to these effects were located on the duplicated chromosomes. However, we detected no genes that were mutated in more than two populations and no mutations that coincided in time with the early paraquat adaptation. We also ensured that the cells used for backcrossing were chosen from populations containing no, or very few, detectable mutations. Due to this, we think the backcrossing experiment counts as a genuine causality test.

We have slightly revised the text in the section “Chromosome duplications explain the second adaptation phase” to make it more transparent that the data are based on an intervention allowing causal inference.

11) Figure 2 – Mitochondrial morphology changes in function of culture medium and growth stage. The data are meaningless if no vigorous controls for these parameters are in place.

We performed the sampling as diligently as possible, making sure to take cells cultivated with and without paraquat in the exponential growth stage, corresponding to 1 to 1.5 population doubling (or 5-7h). We determined the timing of the exponential growth stage by following the growth in real-time of undisturbed populations cultivated in parallel. At the time of sampling, all cell populations were far from entering into stationary phase, which is associated with a mitochondrial morphological shift. Mitochondrial fragmentation due to oxidative stress has been observed several times in various organisms, in mitotic as well as postmitotic cell types. So, the observed fragmentation is fully in line with what was to be expected.

Minor changes of the text in the section “Mitophagy is not responsible for the swift first adaptation phase”.

12) Figure S3B – I am not convinced that the signals correspond to mtDNA. Where are the nuclei in these cells?

The extent to which DAPI stains the mitochondrial and nuclear genomes depends on staining time, as well as the type of fixation (ethanol or formaldehyde) – due to the general preference of DAPI for AT-rich regions, which are enriched in the mtDNA. In the case of fomer Figure S3B, and in contrast to in former Figure S9B (now Figure 6—figure supplement 2B), the experimental design was intended to maximize the differences between mtDNA and nuclear staining. However, we acknowledge that the staining protocol may cause unnecessary ambiguity.

We have removed the former Figure S3B, as it represents a side-point that has little bearing on the central message of the paper.

13) The discussion contains too many speculations and unfunded claims that are not relevant to the reported data.

Without a more specific reference to what is unfunded and what is not relevant to the reported data, we are not able to act on this concern. In fact, in our opinion, the Discussion contained no unfunded claims and a Discussion section should allow authors to place their results in what they conceive as the relevant context, as long as it is backed up with references to the literature.

The Discussion has been somewhat expanded in order to meet some of the specific concerns of the reviewers.

14) In the abstract, the authors claim that a regulatory circuitry underlies mtDNA "editing". Where is the "circuitry" that acts on mtDNA?

The existence of a regulatory circuitry is implicated almost per definition by the Sod2/Rtg2/Rtg3 results showing that there is a two-way mito-nuclear communication / interaction that depends on the retrograde pathway. We note that Rtg2 and Rtg3 have no other known cellular functions besides “regulation”. The results showing that mitophagy is repressed as long as the cells are exposed to paraquat, and that this repression is lost after release from paraquat, must also be part of this regulatory circuitry. So, we think it is justified to allude to the existence of regulation that involves Sod2/Rtg2/Rtg3. Admittedly, however, we have not explored the role of other components of the retrograde pathway, i.e. of Rtg1, Mks1, Bmh1/2, Grr1 and Lst8. And we agree that we do not need to use the term “circuitry” to make the point.

We now avoid the use of the term circuitry in the text.

15) The paper suffers from its style, which is very elliptic, the use of complicated, long sentences, the use of terms such as growth cycles, generations etc… that have not been clearly defined at start.

We acknowledge that we should have given more attention to defining key concepts in the text and not just in Methods, and that we in some cases sacrificed clarity for conciseness.

Key concepts are now defined also in text. And in several cases we have expanded and simplified the text in order to be less elliptic.

16) Page 4, and S1 legends says that " we see PQ causes doubling time to increase", but where do we have to see that? It is not clear how the growth rate is calculated?

We believe most readers to be sufficiently familiar with the growth rate concept to be able to digest the results without a mathematical description of its extraction, which is instead given in the Methods.

How are experiments performed on solid, liquid medium? The figure only shows gene expression data?

The growth dose-response of wildtype cells to paraquat was misplaced and shown in former Figure S7. We now display it in Figure1—figure supplement 1A, where it is shown together with the expression data.

What is a growth cycle and what is its length? What do you mean by 240 generations?

We now describe this in the Results which reads:

“We then used a high throughput growth platform (18) to observe how 96 asexually reproducing yeast cell populations (colonies) on solid agar medium adapted to the chosen paraquat dose in terms of change in cell doubling time over approximately 50 cycles of growth from lag to stationary phase (Figure1—figure supplement 1E), each cycle lasting about 72 h. Cell numbers in each colony doubled 2.5-6x in each cycle, and over the 50 growth cycles the populations doubled in size ~240 times, on average. Neglecting cell deaths and assuming synchronous cell divisions, this corresponds to ~240 cell generations.”

What is the length of one generation in hours? What is a population, and what is the difference with "clonally reproducing cell populations?

We have now replaced “clonally” with “asexually” to highlight that these are yeast cell populations incapable of sex and dividing mitotically. Further on in the manuscript, we simplify by abbreviating the longer phrase as “populations” or “cell populations” and we believe there should be little cause for confusion. Sexual reproduction is only used in one experiment, to backcross adapted clones to wildtypes, and then this is explicitly stated.

At best a picture of the plates used to monitor growth should be shown for one to understand how it is done.

We now show a schematic of the workflow in Figure1—figure supplement 1E. Images of colony scans, depicting how colonies are arrayed on plates, is extensively shown in the published methods paper by Zackrisson et al., 2016, which we refer to in many places. Moreover, this type of colony array is now quite standard in yeast genetics, and an intrinsic property of experiments based on the Singer RoToR, or equivalant, robots. Thus, we don’t think it is necessary to show such images again.

How do you calculate that 106 min doubling time reduction equals 49.3 % of the maximum possible reduction? And what is this maximum?

We now describe this with the sentence:

“Assuming the minimum achievable cell doubling time to be that of the wild type before exposure to paraquat (a mean of 93 min), this corresponded to 49.3% of the maximum possible reduction.”

Figure 1B is confusing: it is understandable that cells are exposed to PQ enough to adapt, then grown without PQ, and then again with PQ, but over the generations shown in the picture, do one not expect to see adaptation after a few "cycles"?

We realize that Figure 1B is hard to digest and we now describe the figure, and its interpretation, more exhaustively in both text and figure legends. The former reads:

“We then tested experimentally whether a Darwinian adaptive process driven by selection of new mutations could account for the observed paraquat adaptation in a stress-release experiment. To this end, we exposed new cell populations to paraquat over many consecutive growth cycles. After each growth cycle, a fraction of the adapting cells was placed in a paraquat-free medium for 1 to 10 growth cycles before being exposed to paraquat once more. The rationale being that if the adaptation is due to accumulation of random mutations, loss of the adaptation would progress gradually and take many growth cycles. All 96 cell populations retained their acquired tolerance to paraquat (mean reduction in cell doubling time: 106 min) for only 1-3 growth cycles before abruptly losing it (Figure 1B). When employing the same experimental procedure to 96 cell populations from each of the two other environments to which adaptation was also fast (arsenic and glycine), we found that despite the presence of a much stronger Darwinian counterselection (Figure 1C), these populations lost their acquired adaptations more slowly and gradually (Figure 1B). Thus, while a Darwinian mutation/selection-based adaptive process could potentially explain the data for seven of the eight tested stressors, the paraquat adaptation could hardly be reconciled with such a process.”

In a classical Darwinian scenario, the loss of the paraquat adaptation during growth in absence of paraquat, emerge as selection against nuclear gene variants underlying this paraquat adaptation and for other variants. To generate the observed quite dramatic drop of mean cell doubling time in paraquat, over just a few generations of Darwinian evolution in absence of paraquat, there would need to be a truly massive disadvantage of the gene variants that caused the paraquat adaptation. But, in Figure 1C we show that there is no disadvantage at all – a paraquat adapted population grows as fast in absence of paraquat as it did before its paraquat adaptation. This gives very compelling support for the conclusion that the recovery from paraquat adaptation, and implicitly also the paraquat adaptation itself, is not due to Darwinian selection on nuclear gene variants – but to something else. We show this else to be regulated deletions of the mtDNA.

The text has been revised as shown by the above quotes.

While we originally defined and described these concepts in the Methods, we agree that also covering these aspects in the Results makes reading more convenient for the reader.

17) Page 5. How can we compare in parallel mitochondrial morphology and growth if the metrics used in the two experiments are different?

Electron microscopy was performed exactly as in the growth experiments – on cells cultivated on solid growth medium. The confocal microscopy was done on cells cultivated in liquid. As the results from the electron and the confocal microscopy data are very similar, the liquid/solid cultivation medium distinction appears irrelevant for the conclusions drawn. Moreover, in the confocal microscopy, we estimated mitochondrial volume by stacking confocal micrographs of sectioned cells. However, stacking of electron micrographs into a complete 3-dimensional image of cells, is time consuming and the sample size would therefore be very limited. We consequently used the 2-dimensional information present in single electron micrographs of thin sections to extract measures of the cell area occupied by mitochondria in cells. The use of the 2D area covered by organelles, and mitochondrial area specifically, as proxy for volume is standard in the field (19–21). The electron microscopy and the confocal microscopy show similar result: that the size of the mitochondrial network does not change substantially but it is re-arranged from a tubular to fragmented organization following paraquat exposure. That two distinct methods, and metrics, support this conclusion increases our confidence in its validity.

We have now described the experiment design better in Methods.

18) Page 6: "cells retained the mtDNA segments that were not lost at near…" revise the grammar of this sentence.

Revised.

19) Page 7. The experiment described in S6A cannot be used to rule out signaling by H2O2: adding 3 mM H2O2 to cells that have already adapted to PQ, whether or not by use of an H2O2 signal, amounts to a severe H2O2 stress, exacerbated by the lack of a functional respiratory chain (petite cells are more sensitive to H2O2, relative to WT). One way of tackling this question would be to see whether adaptive doses of H2O2 (100-300 microM) prior to exposure to paraquat would speed up growth adaptation or not (cross adaptations have been described in the past). Similarly, the WT PQ adaptative response of cells lacking Yap1 or other antioxidants does not prove anything: signaling by H2O2 is mostly localized in confined areas, and this should persist even in a Yap1 mutant.

We acknowledge that we should have been more careful when addressing this topic, and that our data do not allow us to rule out H2O2 signaling.

We have completely revised the section “The deletion of mtDNA segments requires SOD2”.

20) Page 8. The need of SOD2 for PQ adaptation to occur is not really convincing because of the sickness of SOD mutants in general. Further, it shows that there is no adaptation at 12 microG/mL PQ, but then adaptation occurs at a higher dose, but slower, relative to WT. What is the point authors want to make? That SOD by dismutation of the superoxide anion produces H2O2 needed for signaling? But, authors already ruled out the need for H2O2 to signal adaptation? Please don't be too peremptory in your conclusion on this experiment. In addition, it is hard to follow the writer: "in four populations, the copy number…" then "two of these fail to adapt" then "the remaining four populations", but which ones? Lastly the text of Figure 4c indicates 12.5 mG/mL, but the figure 50?

Yes, sod2Δ showed an expected enhanced sensitivity to paraquat, which forced us to reduce the paraquat concentration to achieve an initial fitness reduction similar to the initial fitness reduction of the wild type. But we had to do exactly the same for sod1Δ, showing that the two deletions have a similar negative effect on growth rate compared to the wild type. The deletion of SOD1 had no effect on paraquat adaptation, though, and we think this is the best negative control for a sod2Δ specific effect one can achieve. Thus, if one discards our sod2Δ results with reference to the sickness of these cells, then one should at the same time explain why there is such a striking contrast with the equally sick sod1Δ cells. We are not aware of any data providing such an explanation.

We acknowledge that we should have done a better job when presenting these data. But the point we wanted to make was the following: The reason for why we did not get any adaptive response in sod2Δ at 12.5 ug/mL could be that the actual stress level relative to the one the wild type experienced at 400 ug/mL was lower than anticipated – for some unknown reason. We therefore increased the paraquat concentration to 50 ug/mL to check out this possibility. We did not want to go beyond this concentration as we assumed this would cause too high O₂^•− stress compared to what the wild type experienced. In this case, we did get a very much delayed response in four out of eight populations. In all of these four cases we observed a single large mtDNA deletion that clearly caused the adaptive response. It can very well be that increasing the concentration somewhat more would have caused that all eight populations had shown an adaptive response. But given the Sod1 results, we would still be left with the conclusion that Sod2 is indeed crucial for the swift deletion of mtDNA segments. The fact that we do get a very postponed deletion in the sod2Δ case, and not at all in the rtg2Δ and rtg3Δ cases, strongly suggests that the observed deletions are not due to accidental oxidative damage. One possible explanation is that when the O₂^•− concentration in the matrix gets above a certain level, its auto-dismutation rate may become high enough to trigger a H2O2 signal that induces mtDNA deletion. We have not elaborated on this idea in the paper though. In the absence of more conclusive data, we prefer to remain agnostic as to whether Sod2 has a signaling function that is separate from its enzymatic activity or H2O2 accumulating in the mitochondrial matrix serves as a signal.

We have corrected the detected typo in the figure legend. As mentioned, we have completely revised the section “The deletion of mtDNA segments requires SOD2”, and this revision was also guided by the very helpful concern raised in this comment.

21) Page 10: "the sustentation of mtDNA deletion" is complicated, rather the occurrence of mtDNA deletions.

The word was carefully chosen as we wanted to emphasize that we refer to the sustained activity of an active cellular process that results in the continuous emergence mtDNA deletions. The term occurrence does not capture this effect.

22) Page 11. It says "the adaptation to on- or off-target effects of PQ": clarify.

We acknowledge that it is not necessary to make this distinction in order to convey the main message in the actual paragraph.

We have revised the paragraph such that it does not make this distinction any longer.

Is the duplication event fixed or reversible?

The duplications are fixed in nearly all populations, and we are sorry for not clearly informing about this.

The text now reads:

“However, all but four endpoint populations carried extra copies of chromosome II (n=29), III (n=21) and/or V (n=16) (Figure 6—figure supplement 3A) at fixation or near fixation (mean frequency: 0.97)”

References

1. A. R. Reddi, V. C. Culotta, SOD1 integrates signals from oxygen and glucose to repress respiration. Cell. 152, 224–35 (2013).

2. G. A. Fontana, H. L. Gahlon, Mechanisms of replication and repair in mitochondrial DNA deletion formation. Nucleic Acids Res. 48, 11244–11258 (2020).

3. P. R. Castello, D. A. Drechsel, M. Patel, Mitochondria are a major source of paraquat-induced reactive oxygen species production in the brain. J. Biol. Chem. 282, 14186–93 (2007).

4. F. A. V. Castro, D. Mariani, A. D. Panek, E. C. A. Eleutherio, M. D. Pereira, Cytotoxicity Mechanism of Two Naphthoquinones (Menadione and Plumbagin) in Saccharomyces cerevisiae. PLoS One. 3, e3999 (2008).

5. G. Loor, J. Kondapalli, J. M. Schriewer, N. S. Chandel, T. L. Vanden Hoek, P. T. Schumacker, Menadione triggers cell death through ROS-dependent mechanisms involving PARP activation without requiring apoptosis. Free Radic. Biol. Med. 49, 1925–36 (2010).

6. B. Sendra, S. Panadero, A. Gómez-Hens, Selective Kinetic Determination of Paraquat Using Long-Wavelength Fluorescence Detection. J. Agric. Food Chem. 47, 3733–3737 (1999).

7. F. Puddu, M. Herzog, A. Selivanova, S. Wang, J. Zhu, S. Klein-Lavi, M. Gordon, R. Meirman, G. Millan-Zambrano, I. Ayestaran, I. Salguero, R. Sharan, R. Li, M. Kupiec, S. P. Jackson, Genome architecture and stability in the Saccharomyces cerevisiae knockout collection. Nature. 573, 416–420 (2019).

8. M. De Chiara, B. P. Barré, K. Persson, A. Irizar, C. Vischioni, S. Khaiwal, S. Stenberg, O. C. Amadi, G. Žun, K. Doberšek, C. Taccioli, J. Schacherer, U. Petrovič, J. Warringer, G. Liti, Domestication reprogrammed the budding yeast life cycle. Nat. Ecol. Evol. 6, 448–460 (2022).

9. J. Rutter, A. L. Hughes, Power2: The power of yeast genetics applied to the powerhouse of the cell. Trends Endocrinol. Metab. 26, 59–68 (2015).

10. M. B. Toledano, A. Delaunay, B. Biteau, D. Spector, D. Azevedo, (2003; http://link.springer.com/10.1007/3-540-45611-2_6), pp. 241–303.

11. G. Farrugia, R. Balzan, Oxidative Stress and Programmed Cell Death in Yeast. Front. Oncol. 2 (2012), doi:10.3389/fonc.2012.00064.

12. G. G. Perrone, S.-X. Tan, I. W. Dawes, Reactive oxygen species and yeast apoptosis. Biochim. Biophys. Acta – Mol. Cell Res. 1783, 1354–1368 (2008).

13. D. J. Jamieson, Oxidative stress responses of the yeast Saccharomyces cerevisiae. Yeast. 14, 1511–27 (1998).

14. P. Moradas-Ferreira, V. Costa, Adaptive response of the yeast Saccharomyces cerevisiae to reactive oxygen species: defences, damage and death. Redox Rep. 5, 277–285 (2000).

15. Y. Jia, B. Rothermel, J. Thornton, R. A. Butow, A basic helix-loop-helix-leucine zipper transcription complex in yeast functions in a signaling pathway from mitochondria to the nucleus. Mol. Cell. Biol. 17, 1110–1117 (1997).

16. X. Liao, R. A. Butow, RTG1 and RTG2: Two yeast genes required for a novel path of communication from mitochondria to the nucleus. Cell. 72, 61–71 (1993).

17. M. T. McCammon, C. B. Epstein, B. Przybyla-Zawislak, L. McAlister-Henn, R. A. Butow, Global transcription analysis of Krebs tricarboxylic acid cycle mutants reveals an alternating pattern of gene expression and effects on hypoxic and oxidative genes. Mol. Biol. Cell. 14, 958–72 (2003).

18. M. Zackrisson, J. Hallin, L.-G. Ottosson, P. Dahl, E. Fernandez-Parada, E. Ländström, L. Fernandez-Ricaud, P. Kaferle, A. Skyman, S. Stenberg, S. Omholt, U. Petrovič, J. Warringer, A. Blomberg, Scan-o-matic: High-Resolution Microbial Phenomics at a Massive Scale. G3 (Bethesda). 6, 3003–14 (2016).

19. P. H. Reis-Barbosa, J. J. Carvalho, M. del Sol, C. A. Mandarim-de-Lacerda, Commentary on mitochondrial stereology in transmission electron microscopy. Int. J. Morphol. 38, 26–29 (2020).

20. A. V Loud, A quantitative stereological description of the ultrastructure of normal rat liver parenchymal cells. J. Cell Biol. 37, 27–46 (1968).

21. A. A. Gerencser, C. Chinopoulos, M. J. Birket, M. Jastroch, C. Vitelli, D. G. Nicholls, M. D. Brand, Quantitative measurement of mitochondrial membrane potential in cultured cells: calcium-induced de- and hyperpolarization of neuronal mitochondria. J. Physiol. 590, 2845–71 (2012).

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Data Citations

Warringer J. 2020. Chronic superoxide distress causes irreversible loss of mtDNA segments. NCBI BioProject. PRJNA622836

Supplementary Materials

Figure 1—source data 1. Doubling time data of 96 populations adapted to each of eight different environments over G generations; doubling times are in the respective selection environment.

Data are shown in Figure 1A and Figure 6D.

elife-76095-fig1-data1.xlsx^{(1.1MB, xlsx)}

Figure 1—source data 2. Difference in doubling time in absence of stress and in the respective selection environment, for adapted populations having achieved 70-90% of their final adaptation.

Data are shown in Figure 1C.

elife-76095-fig1-data2.xlsx^{(44.1KB, xlsx)}

Figure 1—source data 3. Doubling time data of 96 populations adapted to paraquat, arsenic and glycine over G_s generations and then released from selection for G_r generations; doubling times are in paraquat, arsenic, and glycine, respectively.

Data are shown in Figure 1B.

elife-76095-fig1-data3.xlsx^{(235.1KB, xlsx)}

Figure 1—figure supplement 1—source data 1. FPKM data of selected oxidative defense genes, obtained from RNA-sequencing of cells exposed to paraquat.

Data are shown in Figure 1—figure supplement 1.

elife-76095-fig1-figsupp1-data1.xlsx^{(54.3KB, xlsx)}

Figure 1—figure supplement 1—source data 2. Doubling time data of WT, mip1Δ and sod2Δ populations, with and without paraquat and with and without vitamin C.

Data are shown in Figure 1—figure supplement 1.

elife-76095-fig1-figsupp1-data2.xlsx^{(222.2KB, xlsx)}

Figure 1—figure supplement 1—source data 3. Doubling time data of WT in different concentrations of paraquat.

Data are shown in Figure 1—figure supplement 1A.

elife-76095-fig1-figsupp1-data3.xlsx^{(42.2KB, xlsx)}

Figure 1—figure supplement 2—source data 1. Doubling time data for the BY4741 single gene deletion collection under paraquat exposure; used as input for simulations in Figure 1—figure supplement 2.

elife-76095-fig1-figsupp2-data1.xlsx^{(5.1MB, xlsx)}

Figure 1—figure supplement 2—source data 2. Doubling time data of disomic strains growing in paraquat; used as input for simulations in Figure 1—figure supplement 2.

elife-76095-fig1-figsupp2-data2.xlsx^{(52.7KB, xlsx)}

Figure 2—source data 1. Mitochondrial and cell area quantified, based on electron microscopy micrographs.

Data are shown in Figure 2A.

elife-76095-fig2-data1.xlsx^{(1.1MB, xlsx)}

Figure 2—source data 2. Number of mitochondria quantified, based on electron microscopy micrographs.

Data are shown in Figure 2A.

elife-76095-fig2-data2.xlsx^{(38.7KB, xlsx)}

Figure 2—source data 3. Number and volume of mitochondria of cells segmented from confocal microscopy micrographs of cells exposed to paraquat.

Data are shown in Figure 2B and in Figure 2—figure supplement 1.

elife-76095-fig2-data3.xlsx^{(137.9KB, xlsx)}

Figure 2—source data 4. Doubling time data of wild type and populations adapting to paraquat over generations G; doubling times are in paraquat.

Data are shown in Figure 2C.

elife-76095-fig2-data4.xlsx^{(53.5KB, xlsx)}

Figure 3—source data 1. Doubling time data of 96 populations adapted to paraquat for G generations exposed; doubling times are in paraquat and respiratory media (glycerol).

Data are shown in Figure 3C.

elife-76095-fig3-data1.xlsx^{(56.9KB, xlsx)}

Figure 3—source data 2. Mean log₂ coverage of 1 kb windows spanning the mitochondrial genome of five sequenced paraquat adapting populations over generations G of selection.

Data are shown in Figure 3A.

elife-76095-fig3-data2.xlsx^{(218.6KB, xlsx)}

Figure 3—source data 3. Mean log₂ coverage of 1 kb windows spanning the mitochondrial genome of sequenced populations adapting to paraquat and then released from this selection; data is given as a function of generations G of relaxation of selection.

Data are shown in Figure 3D and in Figure 3—figure supplement 2C.

elife-76095-fig3-data3.xlsx^{(179.1KB, xlsx)}

Figure 3—source data 4. qPCR data for mitochondrial DNA genes and nuclear DNA controls over generations of paraquat adaptation.

Data are shown in Figure 3B and Figure 3—figure supplement 1.

elife-76095-fig3-data4.xlsx^{(104.9KB, xlsx)}

Figure 3—figure supplement 2—source data 1. Doubling time data of populations adapted to paraquat for G_s generations and then released from selection for G_r generations; doubling times are in paraquat and respiratory media (glycerol).

Data are shown in Figure 3—figure supplement 2B and Figure 6A.

elife-76095-fig3-figsupp2-data1.xlsx^{(43.1KB, xlsx)}

Figure 3—figure supplement 3—source data 1. Doubling time data of 96 populations adapted over G generations to paraquat; doubling times are in 0 and 0.25 mM of menadione.

Data are shown in Figure 3—figure supplement 3B.

elife-76095-fig3-figsupp3-data1.xlsx^{(9.9KB, xlsx)}

Figure 4—source data 1. Doubling time data of wild type, sod2∆, and sod1∆ populations adapting to paraquat; doubling times are in paraquat.

Data are shown in Figure 4A and B and in Figure 4—figure supplement 1B.

elife-76095-fig4-data1.xlsx^{(43KB, xlsx)}

Figure 4—source data 2. Growth curves of populations of mip1∆sod2∆, mip1∆, sod2∆, and wild type exposed to paraquat.

Data are shown in Figure 4C.

elife-76095-fig4-data2.xlsx^{(115.5KB, xlsx)}

Figure 4—source data 3. qPCR data for mitochondrial DNA genes and nuclear DNA controls in sod2Δ populations over generations of paraquat adaptation.

Data are shown in Figure 4B and in Figure 4—figure supplement 1B.

elife-76095-fig4-data3.xlsx^{(91.3KB, xlsx)}

Figure 4—figure supplement 1—source data 1. Mean growth curves of wild type, sod1Δ, and sod2Δ and wild type exposed to different concentrations of paraquat.

Data are shown in Figure 4—figure supplement 1A.

elife-76095-fig4-figsupp1-data1.xlsx^{(252.9KB, xlsx)}

Figure 5—source data 1. Doubling time data of wild type, rtg2∆, rtg3∆, and mip1Δ populations adapting to paraquat; doubling times are in paraquat.

Data are shown in Figure 5A.

elife-76095-fig5-data1.xlsx^{(53.2KB, xlsx)}

Figure 5—source data 2. Growth curves of wild type, rtg2∆, rtg3∆, and mip1Δ, adapted and not adapted to paraquat; doubling times are in respiratory media (glycerol).

Data are shown in Figure 5B.

elife-76095-fig5-data2.xlsx^{(178.3KB, xlsx)}

Figure 6—source data 1. Mean log₂ coverage of 1 kb windows spanning the mitochondrial genome of each sequenced paraquat adapted endpoint population.

Data are shown in Figure 6B and C, Figure 6—figure supplement 2A and Figure 6—figure supplement 4B.

elife-76095-fig6-data1.xlsx^{(162.1KB, xlsx)}

Figure 6—figure supplement 1—source data 1. Doubling time data of 96 populations adapted to paraquat for G_s generations, followed by release from this selection over G_r generations; doubling times on paraquat.

Data are shown in Figure 6—figure supplement 1.

elife-76095-fig6-figsupp1-data1.xlsx^{(610.4KB, xlsx)}

Figure 6—figure supplement 4—source data 1. Small indels and SNPs called in sequenced paraquat adapted endpoint populations.

Data are shown in Figure 6—figure supplement 4B.

elife-76095-fig6-figsupp4-data1.xlsx^{(89.4KB, xlsx)}

Figure 6—figure supplement 4—source data 2. Doubling time data of mip1Δ cells grown in stress, and of wild type cells grown in no stress.

Data are shown in Figure 6—figure supplement 4A.

elife-76095-fig6-figsupp4-data2.xlsx^{(35.2KB, xlsx)}

Figure 6—figure supplement 5—source data 1. Mean log₂ coverage for each chromosome in each sequenced paraquat adapted endpoint 1 population.

Data are shown in Figure 6—figure supplement 5A and in Figure 6—figure supplement 4B.

elife-76095-fig6-figsupp5-data1.xlsx^{(94.2KB, xlsx)}

Figure 6—figure supplement 5—source data 2. Mean log2 coverage for each chromosome in five sequenced paraquat adapting populations over generations G of selection.

Data are shown in Figure 6—figure supplement 5B.

elife-76095-fig6-figsupp5-data2.xlsx^{(55.5KB, xlsx)}

Supplementary file 1. Description of stressor environments used as selection pressures.

elife-76095-supp1.docx^{(13.3KB, docx)}

Transparent reporting form

elife-76095-transrepform1.docx^{(245.4KB, docx)}

Source data 1. Primers used for strain construction and qPCR.

elife-76095-data1.xlsx^{(28.6KB, xlsx)}

Data Availability Statement

Sequence data that support the findings of this study have been deposited in Sequencing Read Archive (SRA) with the accession codes PRJNA622836. The growth phenotyping code can be found at https://github.com/Scan-o-Matic/scanomatic.git, the simulation code at https://github.com/HelstVadsom/GenomeAdaptation.git and the imaging code at https://github.com/CamachoDejay/SStenberg_3Dyeast_tools copy archived at swh:1:rev:a047daf337fa05f75f7cb3affb498ed70b6d7703. The authors declare that all other data supporting the findings of this study are available in the article and at https://doi.org/10.17632/mvx7t7rw2d.1.

The following previously published dataset was used:

Warringer J. 2020. Chronic superoxide distress causes irreversible loss of mtDNA segments. NCBI BioProject. PRJNA622836

[bib1] Adachi K, Fujiura Y, Mayumi F, Nozuhara A, Sugiu Y, Sakanashi T, Hidaka T, Toshima H. A deletion of mitochondrial DNA in murine doxorubicin-induced cardiotoxicity. Biochemical and Biophysical Research Communications. 1993;195:945–951. doi: 10.1006/bbrc.1993.2135. [DOI] [PubMed] [Google Scholar]

[bib2] Aerts L, Morais VA, Aerts L. Parkinson’s Disease. Elsevier; 2017. Electron transport chain; pp. 41–75. [Google Scholar]

[bib3] Ammal Kaidery N, Thomas B. Current perspective of mitochondrial biology in Parkinson’s disease. Neurochemistry International. 2018;117:91–113. doi: 10.1016/j.neuint.2018.03.001. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib4] Ban T, Ishihara T, Kohno H, Saita S, Ichimura A, Maenaka K, Oka T, Mihara K, Ishihara N. Molecular basis of selective mitochondrial fusion by heterotypic action between OPA1 and cardiolipin. Nature Cell Biology. 2017;19:856–863. doi: 10.1038/ncb3560. [DOI] [PubMed] [Google Scholar]

[bib5] Barrientos A. Yeast models of human mitochondrial diseases. IUBMB Life. 2003;55:83–95. doi: 10.1002/tbmb.718540876. [DOI] [PubMed] [Google Scholar]

[bib6] Bess AS, Crocker TL, Ryde IT, Meyer JN. Mitochondrial dynamics and autophagy aid in removal of persistent mitochondrial DNA damage in Caenorhabditis elegans. Nucleic Acids Research. 2012;40:7916–7931. doi: 10.1093/nar/gks532. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib7] Castello PR, Drechsel DA, Patel M. Mitochondria are a major source of paraquat-induced reactive oxygen species production in the brain. The Journal of Biological Chemistry. 2007;282:14186–14193. doi: 10.1074/jbc.M700827200. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib8] Chen Z, Wang Z-H, Zhang G, Bleck CKE, Chung DJ, Madison GP, Lindberg E, Combs C, Balaban RS, Xu H. Mitochondrial DNA segregation and replication restrict the transmission of detrimental mutation. The Journal of Cell Biology. 2020;219:e201905160. doi: 10.1083/jcb.201905160. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib9] Chevereau G, Dravecká M, Batur T, Guvenek A, Ayhan DH, Toprak E, Bollenbach T. Quantifying the determinants of evolutionary dynamics leading to drug resistance. PLOS Biology. 2015;13:11. doi: 10.1371/journal.pbio.1002299. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib10] Chou HH, Chiu HC, Delaney NF, Segrè D, Marx CJ. Diminishing returns epistasis among beneficial mutations decelerates adaptation. Science. 2011;332:1190–1192. doi: 10.1126/science.1203799. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib11] Cochemé HM, Murphy MP. Complex I is the major site of mitochondrial superoxide production by paraquat. The Journal of Biological Chemistry. 2008;283:1786–1798. doi: 10.1074/jbc.M708597200. [DOI] [PubMed] [Google Scholar]

[bib12] Couce A, Tenaillon OA. The rule of declining adaptability in microbial evolution experiments. Frontiers in Genetics. 2015;6:99. doi: 10.3389/fgene.2015.00099. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib13] De Chiara M, Barré B, Persson K, Chioma AO, Irizar A, Schacherer J, Warringer J, Liti G. Domestication Reprogrammed the Budding Yeast Life Cycle. bioRxiv. 2020 doi: 10.1101/2020.02.08.939314. [DOI] [PubMed]

[bib14] Dobin A, Davis CA, Schlesinger F, Drenkow J, Zaleski C, Jha S, Batut P, Chaisson M, Gingeras TR. STAR: ultrafast universal RNA-seq aligner. Bioinformatics. 2013;29:15–21. doi: 10.1093/bioinformatics/bts635. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib15] Epstein CB, Waddle JA, Hale W, Davé V, Thornton J, Macatee TL, Garner HR, Butow RA. Genome-wide responses to mitochondrial dysfunction. Molecular Biology of the Cell. 2001;12:297–308. doi: 10.1091/mbc.12.2.297. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib16] Fangman WL, Henly JW, Churchill G, Brewer BJ. Stable maintenance of a 35-base-pair yeast mitochondrial genome. Molecular and Cellular Biology. 1989;9:1917–1921. doi: 10.1128/mcb.9.5.1917-1921.1989. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib17] Farooq MA, Pracheil TM, Dong Z, Xiao F, Liu Z. Mitochondrial DNA instability in cells lacking aconitase correlates with iron citrate toxicity. Oxidative Medicine and Cellular Longevity. 2013;2013:493536. doi: 10.1155/2013/493536. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib18] Fontana GA, Gahlon HL. Mechanisms of replication and repair in mitochondrial DNA deletion formation. Nucleic Acids Research. 2020;48:11244–11258. doi: 10.1093/nar/gkaa804. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib19] Frank M, Duvezin-Caubet S, Koob S, Occhipinti A, Jagasia R, Petcherski A, Ruonala MO, Priault M, Salin B, Reichert AS. Mitophagy is triggered by mild oxidative stress in a mitochondrial fission dependent manner. Biochimica et Biophysica Acta. 2012;1823:2297–2310. doi: 10.1016/j.bbamcr.2012.08.007. [DOI] [PubMed] [Google Scholar]

[bib20] Fukui M, Choi HJ, Zhu BT. Rapid generation of mitochondrial superoxide induces mitochondrion-dependent but caspase-independent cell death in hippocampal neuronal cells that morphologically resembles necroptosis. Toxicology and Applied Pharmacology. 2012;262:156–166. doi: 10.1016/j.taap.2012.04.030. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib21] Gaisne M, Bécam AM, Verdière J, Herbert CJ. A “natural” mutation in Saccharomyces cerevisiae strains derived from S288c affects the complex regulatory gene HAP1 (CYP1) Current Genetics. 1999;36:195–200. doi: 10.1007/s002940050490. [DOI] [PubMed] [Google Scholar]

[bib22] Genc G, Okuyucu A, Meydan BC, Yavuz O, Nisbet O, Hokelek M, Bedir A, Ozkaya O. Effect of free creatine therapy on cisplatin-induced renal damage. Renal Failure. 2014;36:1108–1113. doi: 10.3109/0886022X.2014.917576. [DOI] [PubMed] [Google Scholar]

[bib23] Giaever G, Chu AM, Ni L, Connelly C, Riles L, Véronneau S, Dow S, Lucau-Danila A, Anderson K, André B, Arkin AP, Astromoff A, El-Bakkoury M, Bangham R, Benito R, Brachat S, Campanaro S, Curtiss M, Davis K, Deutschbauer A, Entian K-D, Flaherty P, Foury F, Garfinkel DJ, Gerstein M, Gotte D, Güldener U, Hegemann JH, Hempel S, Herman Z, Jaramillo DF, Kelly DE, Kelly SL, Kötter P, LaBonte D, Lamb DC, Lan N, Liang H, Liao H, Liu L, Luo C, Lussier M, Mao R, Menard P, Ooi SL, Revuelta JL, Roberts CJ, Rose M, Ross-Macdonald P, Scherens B, Schimmack G, Shafer B, Shoemaker DD, Sookhai-Mahadeo S, Storms RK, Strathern JN, Valle G, Voet M, Volckaert G, Wang C, Ward TR, Wilhelmy J, Winzeler EA, Yang Y, Yen G, Youngman E, Yu K, Bussey H, Boeke JD, Snyder M, Philippsen P, Davis RW, Johnston M. Functional profiling of the Saccharomyces cerevisiae genome. Nature. 2002;418:387–391. doi: 10.1038/nature00935. [DOI] [PubMed] [Google Scholar]

[bib24] Gjuvsland AB, Zörgö E, Samy JK, Stenberg S, Demirsoy IH, Roque F, Maciaszczyk-Dziubinska E, Migocka M, Alonso-Perez E, Zackrisson M, Wysocki R, Tamás MJ, Jonassen I, Omholt SW, Warringer J. Disentangling genetic and epigenetic determinants of ultrafast adaptation. Molecular Systems Biology. 2016;12:12. doi: 10.15252/msb.20166951. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib25] Grant CM, MacIver FH, Dawes IW. Mitochondrial function is required for resistance to oxidative stress in the yeast Saccharomyces cerevisiae. FEBS Letters. 1997;410:219–222. doi: 10.1016/s0014-5793(97)00592-9. [DOI] [PubMed] [Google Scholar]

[bib26] Guaragnella N, Coyne LP, Chen XJ, Giannattasio S. Mitochondria-cytosol-nucleus crosstalk: learning from Saccharomyces cerevisiae. FEMS Yeast Research. 2018;18:e88. doi: 10.1093/femsyr/foy088. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib27] Gustafsson ÅB, Dorn GW. Evolving and expanding the roles of mitophagy as a homeostatic and pathogenic process. Physiological Reviews. 2019;99:853–892. doi: 10.1152/physrev.00005.2018. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib28] Gutiérrez A, Sancho M, Beltran G, Guillamon JM, Warringer J. Replenishment and mobilization of intracellular nitrogen pools decouples wine yeast nitrogen uptake from growth. Applied Microbiology and Biotechnology. 2016;100:3255–3265. doi: 10.1007/s00253-015-7273-y. [DOI] [PubMed] [Google Scholar]

[bib29] Hawes P, Netherton CL, Mueller M, Wileman T, Monaghan P. Rapid freeze-substitution preserves membranes in high-pressure frozen tissue culture cells. Journal of Microscopy. 2007;226:182–189. doi: 10.1111/j.1365-2818.2007.01767.x. [DOI] [PubMed] [Google Scholar]

[bib30] Hepple RT. Impact of aging on mitochondrial function in cardiac and skeletal muscle. Free Radical Biology & Medicine. 2016;98:177–186. doi: 10.1016/j.freeradbiomed.2016.03.017. [DOI] [PubMed] [Google Scholar]

[bib31] Hu H, Tan CC, Tan L, Yu JT. A mitocentric view of alzheimer’s disease. Molecular Neurobiology. 2017;54:6046–6060. doi: 10.1007/s12035-016-0117-7. [DOI] [PubMed] [Google Scholar]

[bib32] Hung CHL, Cheng SSY, Cheung YT, Wuwongse S, Zhang NQ, Ho YS, Lee SMY, Chang RCC. A reciprocal relationship between reactive oxygen species and mitochondrial dynamics in neurodegeneration. Redox Biology. 2018;14:7–19. doi: 10.1016/j.redox.2017.08.010. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib33] Hyttinen JMT, Viiri J, Kaarniranta K, Błasiak J. Mitochondrial quality control in AMD: does mitophagy play a pivotal role? Cellular and Molecular Life Sciences. 2018;75:2991–3008. doi: 10.1007/s00018-018-2843-7. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib34] Jakubke C, Roussou R, Maiser A, Schug C, Thoma F, Bunk D, Hörl D, Leonhardt H, Walter P, Klecker T, Osman C. Cristae-dependent quality control of the mitochondrial genome. Science Advances. 2021;7:eabi8886. doi: 10.1126/sciadv.abi8886. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib35] Kalyanaraman B, Joseph J, Kalivendi S, Wang S, Konorev E, Kotamraju S. Doxorubicin-induced apoptosis: implications in cardiotoxicity. Molecular and Cellular Biochemistry. 2002;234–235:119–124. [PubMed] [Google Scholar]

[bib36] Khan AI, Dinh DM, Schneider D, Lenski RE, Cooper TF. Negative epistasis between beneficial mutations in an evolving bacterial population. Science. 2011;332:1193–1196. doi: 10.1126/science.1203801. [DOI] [PubMed] [Google Scholar]

[bib37] Kremer JR, Mastronarde DN, McIntosh JR. Computer visualization of three-dimensional image data using IMOD. Journal of Structural Biology. 1996;116:71–76. doi: 10.1006/jsbi.1996.0013. [DOI] [PubMed] [Google Scholar]

[bib38] Krishnan KJ, Reeve AK, Samuels DC, Chinnery PF, Blackwood JK, Taylor RW, Wanrooij S, Spelbrink JN, Lightowlers RN, Turnbull DM. What causes mitochondrial DNA deletions in human cells? Nature Genetics. 2008;40:275–279. doi: 10.1038/ng.f.94. [DOI] [PubMed] [Google Scholar]

[bib39] Kumar R, Reichert AS. Common principles and specific mechanisms of mitophagy from yeast to humans. International Journal of Molecular Sciences. 2021;22:4363. doi: 10.3390/ijms22094363. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib40] Liao Y, Smyth GK, Shi W. featureCounts: an efficient general purpose program for assigning sequence reads to genomic features. Bioinformatics. 2014;30:923–930. doi: 10.1093/bioinformatics/btt656. [DOI] [PubMed] [Google Scholar]

[bib41] Liti G, Carter DM, Moses AM, Warringer J, Parts L, James SA, Davey RP, Roberts IN, Burt A, Koufopanou V, Tsai IJ, Bergman CM, Bensasson D, O’Kelly MJT, van Oudenaarden A, Barton DBH, Bailes E, Nguyen AN, Jones M, Quail MA, Goodhead I, Sims S, Smith F, Blomberg A, Durbin R, Louis EJ. Population genomics of domestic and wild yeasts. Nature. 2009;458:337–341. doi: 10.1038/nature07743. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib42] Liu Y, Okamoto K. Regulatory mechanisms of mitophagy in yeast. Biochimica et Biophysica Acta. General Subjects. 2021;1865:129858. doi: 10.1016/j.bbagen.2021.129858. [DOI] [PubMed] [Google Scholar]

[bib43] Lodi T, Dallabona C, Nolli C, Goffrini P, Donnini C, Baruffini E. DNA polymerase γ and disease: what we have learned from yeast. Frontiers in Genetics. 2015;6:106. doi: 10.3389/fgene.2015.00106. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib44] Love MI, Huber W, Anders S. Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2. Genome Biology. 2014;15:12. doi: 10.1186/s13059-014-0550-8. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib45] Maleszka R, Skelly PJ, Clark-Walker GD. Rolling circle replication of DNA in yeast mitochondria. The EMBO Journal. 1991;10:3923–3929. doi: 10.1002/j.1460-2075.1991.tb04962.x. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib46] Malhi SS, Budhiraja A, Arora S, Chaudhari KR, Nepali K, Kumar R, Sohi H, Murthy RSR. Intracellular delivery of redox cycler-doxorubicin to the mitochondria of cancer cell by folate receptor targeted mitocancerotropic liposomes. International Journal of Pharmaceutics. 2012;432:63–74. doi: 10.1016/j.ijpharm.2012.04.030. [DOI] [PubMed] [Google Scholar]

[bib47] Merriam-Webster’s Coll Thes CT. Merriam-Webster’s Coll Thes. 2021. [September 2, 2021]. https://unabridged.merriam-webster.com/thesaurus/edit

[bib48] Montava-Garriga L, Ganley IG. Outstanding questions in mitophagy: what we do and do not know. Journal of Molecular Biology. 2020;432:206–230. doi: 10.1016/j.jmb.2019.06.032. [DOI] [PubMed] [Google Scholar]

[bib49] Moruno-Manchon JF, Uzor N-E, Kesler SR, Wefel JS, Townley DM, Nagaraja AS, Pradeep S, Mangala LS, Sood AK, Tsvetkov AS. Peroxisomes contribute to oxidative stress in neurons during doxorubicin-based chemotherapy. Molecular and Cellular Neurosciences. 2018;86:65–71. doi: 10.1016/j.mcn.2017.11.014. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib50] Ng MYW, Wai T, Simonsen A. Quality control of the mitochondrion. Developmental Cell. 2021;56:881–905. doi: 10.1016/j.devcel.2021.02.009. [DOI] [PubMed] [Google Scholar]

[bib51] Nido GS, Dölle C, Flønes I, Tuppen HA, Alves G, Tysnes OB, Haugarvoll K, Tzoulis C. Ultradeep mapping of neuronal mitochondrial deletions in Parkinson’s disease. Neurobiology of Aging. 2018;63:120–127. doi: 10.1016/j.neurobiolaging.2017.10.024. [DOI] [PubMed] [Google Scholar]

[bib52] Nissanka N, Minczuk M, Moraes CT. Mechanisms of mitochondrial dna deletion formation. Trends in Genetics. 2019;35:235–244. doi: 10.1016/j.tig.2019.01.001. [DOI] [PubMed] [Google Scholar]

[bib53] Palikaras K, Tavernarakis N. Mitochondrial homeostasis: the interplay between mitophagy and mitochondrial biogenesis. Experimental Gerontology. 2014;56:182–188. doi: 10.1016/j.exger.2014.01.021. [DOI] [PubMed] [Google Scholar]

[bib54] Persson K, Stenberg S, Tamás MJ, Warringer J. Adaptation of the Yeast Gene Knockout Collection Is Near-Perfectly Predicted by Fitness and Diminishing Return Epistasis. bioRxiv. 2022 doi: 10.1101/2022.03.25.485769. [DOI] [PMC free article] [PubMed]

[bib55] Puddu F, Herzog M, Selivanova A, Wang S, Zhu J, Klein-Lavi S, Gordon M, Meirman R, Millan-Zambrano G, Ayestaran I, Salguero I, Sharan R, Li R, Kupiec M, Jackson SP. Genome architecture and stability in the Saccharomyces cerevisiae knockout collection. Nature. 2019;573:416–420. doi: 10.1038/s41586-019-1549-9. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib56] Reddi AR, Culotta VC. SOD1 integrates signals from oxygen and glucose to repress respiration. Cell. 2013;152:224–235. doi: 10.1016/j.cell.2012.11.046. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib57] Ren X, Keeney JTR, Miriyala S, Noel T, Powell DK, Chaiswing L, Bondada S, St Clair DK, Butterfield DA. The triangle of death of neurons: Oxidative damage, mitochondrial dysfunction, and loss of choline-containing biomolecules in brains of mice treated with doxorubicin, Advanced insights into mechanisms of chemotherapy induced cognitive impairment (“chemobrain”) involving TNF-α. Free Radical Biology & Medicine. 2019;134:1–8. doi: 10.1016/j.freeradbiomed.2018.12.029. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib58] Rinnerthaler M, Büttner S, Laun P, Heeren G, Felder TK, Klinger H, Weinberger M, Stolze K, Grousl T, Hasek J, Benada O, Frydlova I, Klocker A, Simon-Nobbe B, Jansko B, Breitenbach-Koller H, Eisenberg T, Gourlay CW, Madeo F, Burhans WC, Breitenbach M. Yno1p/aim14p, a NADPH-oxidase ortholog, controls extramitochondrial reactive oxygen species generation, apoptosis, and actin cable formation in yeast. PNAS. 2012;109:8658–8663. doi: 10.1073/pnas.1201629109. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib59] Rothermel BA, Shyjan AW, Etheredge JL, Butow RA. Transactivation by Rtg1p, a basic helix-loop-helix protein that functions in communication between mitochondria and the nucleus in yeast. The Journal of Biological Chemistry. 1995;270:29476–29482. doi: 10.1074/jbc.270.49.29476. [DOI] [PubMed] [Google Scholar]

[bib60] Rothermel BA, Thornton JL, Butow RA. Rtg3p, a basic helix-loop-helix/leucine zipper protein that functions in mitochondrial-induced changes in gene expression, contains independent activation domains. The Journal of Biological Chemistry. 1997;272:19801–19807. doi: 10.1074/jbc.272.32.19801. [DOI] [PubMed] [Google Scholar]

[bib61] Sedlackova L, Korolchuk VI. Mitochondrial quality control as a key determinant of cell survival. Biochimica et Biophysica Acta. Molecular Cell Research. 2019;1866:575–587. doi: 10.1016/j.bbamcr.2018.12.012. [DOI] [PubMed] [Google Scholar]

[bib62] Sekito T, Thornton J, Butow RA. Mitochondria-to-nuclear signaling is regulated by the subcellular localization of the transcription factors Rtg1p and Rtg3p. Molecular Biology of the Cell. 2000;11:2103–2115. doi: 10.1091/mbc.11.6.2103. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib63] Sendra B, Panadero S, Gómez-Hens A. Selective kinetic determination of paraquat using long-wavelength fluorescence detection. Journal of Agricultural and Food Chemistry. 1999;47:3733–3737. doi: 10.1021/jf990072d. [DOI] [PubMed] [Google Scholar]

[bib64] Shpilka T, Haynes CM. The mitochondrial UPR: mechanisms, physiological functions and implications in ageing. Nature Reviews. Molecular Cell Biology. 2018;19:109–120. doi: 10.1038/nrm.2017.110. [DOI] [PubMed] [Google Scholar]

[bib65] Sies H, Berndt C, Jones DP. Oxidative Stress. Annual Review of Biochemistry. 2017;86:715–748. doi: 10.1146/annurev-biochem-061516-045037. [DOI] [PubMed] [Google Scholar]

[bib66] Song Z, Chang H, Han N, Liu Z, Liu Y, Wang H, Shao J, Wang Z, Gao H, Yin J. He-Wei granules (HWKL) combat cisplatin-induced nephrotoxicity and myelosuppression in rats by inhibiting oxidative stress, inflammatory cytokines and apoptosis. RSC Advances. 2017;7:19794–19807. doi: 10.1039/C7RA02830J. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib67] Songbo M, Lang H, Xinyong C, Bin X, Ping Z, Liang S. Oxidative stress injury in doxorubicin-induced cardiotoxicity. Toxicology Letters. 2019;307:41–48. doi: 10.1016/j.toxlet.2019.02.013. [DOI] [PubMed] [Google Scholar]

[bib68] Sprenger HG, Langer T. The good and the bad of mitochondrial breakups. Trends in Cell Biology. 2019;29:888–900. doi: 10.1016/j.tcb.2019.08.003. [DOI] [PubMed] [Google Scholar]

[bib69] Stenberg S. Simon stenberg 3D yeast tools. swh:1:rev:a047daf337fa05f75f7cb3affb498ed70b6d7703Software Heritage. 2022 https://archive.softwareheritage.org/swh:1:dir:895f1538a0d631c088bb3ec6260151856787c4a6;origin=https://github.com/CamachoDejay/SStenberg_3Dyeast_tools;visit=swh:1:snp:ac5183da9e781260fa588c16a510bafe4091683b;anchor=swh:1:rev:a047daf337fa05f75f7cb3affb498ed70b6d7703

[bib70] Sun N, Youle RJ, Finkel T. The mitochondrial basis of aging. Molecular Cell. 2016;61:654–666. doi: 10.1016/j.molcel.2016.01.028. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib71] Twig G, Elorza A, Molina AJA, Mohamed H, Wikstrom JD, Walzer G, Stiles L, Haigh SE, Katz S, Las G, Alroy J, Wu M, Py BF, Yuan J, Deeney JT, Corkey BE, Shirihai OS. Fission and selective fusion govern mitochondrial segregation and elimination by autophagy. The EMBO Journal. 2008;27:433–446. doi: 10.1038/sj.emboj.7601963. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib72] Vadsom H. GenomeAdaptation. swh:1:rev:d9589fad41a6d0557b039dd78cb035c7e79c46c2GitHub. 2017 https://github.com/HelstVadsom/GenomeAdaptation

[bib73] Vaser R, Adusumalli S, Leng SN, Sikic M, Ng PC. SIFT missense predictions for genomes. Nature Protocols. 2016;11:1–9. doi: 10.1038/nprot.2015.123. [DOI] [PubMed] [Google Scholar]

[bib74] Veatch JR, McMurray MA, Nelson ZW, Gottschling DE. Mitochondrial dysfunction leads to nuclear genome instability via an iron-sulfur cluster defect. Cell. 2009;137:1247–1258. doi: 10.1016/j.cell.2009.04.014. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib75] Voordeckers K, Verstrepen KJ. Experimental evolution of the model eukaryote Saccharomyces cerevisiae yields insight into the molecular mechanisms underlying adaptation. Current Opinion in Microbiology. 2015;28:1–9. doi: 10.1016/j.mib.2015.06.018. [DOI] [PubMed] [Google Scholar]

[bib76] Vowinckel J, Hartl J, Marx H, Kerick M, Runggatscher K, Keller MA, Mülleder M, Day J, Weber M, Rinnerthaler M, Yu JSL, Aulakh SK, Lehmann A, Mattanovich D, Timmermann B, Zhang N, Dunn CD, MacRae JI, Breitenbach M, Ralser M. The metabolic growth limitations of petite cells lacking the mitochondrial genome. Nature Metabolism. 2021;3:1521–1535. doi: 10.1038/s42255-021-00477-6. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib77] Wingett SW, Andrews S. FastQ Screen: A tool for multi-genome mapping and quality control. F1000Research. 2018;7:1338. doi: 10.12688/f1000research.15931.2. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib78] Yue J-X, Li J, Aigrain L, Hallin J, Persson K, Oliver K, Bergström A, Coupland P, Warringer J, Lagomarsino MC, Fischer G, Durbin R, Liti G. Contrasting evolutionary genome dynamics between domesticated and wild yeasts. Nature Genetics. 2017;49:913–924. doi: 10.1038/ng.3847. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib79] Yue JX, Liti G. Long-read sequencing data analysis for yeasts. Nature Protocols. 2018;13:1213–1231. doi: 10.1038/nprot.2018.025. [DOI] [PubMed] [Google Scholar]

[bib80] Zackrisson M, Hallin J, Ottosson LG, Dahl P, Fernandez-Parada E, Ländström E, Fernandez-Ricaud L, Kaferle P, Skyman A, Stenberg S, Omholt S, Petrovič U, Warringer J, Blomberg A. Scan-o-matic: High-resolution microbial phenomics at a massive scale. G3: Genes, Genomes, Genetics. 2016;6:3003–3014. doi: 10.1534/g3.116.032342. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib81] Zackrisson M. Scan-o-Matic/scanomatic. swh:1:rev:27649ae392cec469008ea2c269ef877c1b2031c8GitHub. 2019 https://github.com/Scan-o-Matic/scanomatic

[bib82] Zebrowski DC, Kaback DB. A simple method for isolating disomic strains of Saccharomyces cerevisiae. Yeast. 2008;25:321–326. doi: 10.1002/yea.1590. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib83] Zhang Y, Wang Z-H, Liu Y, Chen Y, Sun N, Gucek M, Zhang F, Xu H. PINK1 inhibits local protein synthesis to limit transmission of deleterious mitochondrial DNA mutations. Molecular Cell. 2019;73:1127–1137. doi: 10.1016/j.molcel.2019.01.013. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib84] Zhu YO, Siegal ML, Hall DW, Petrov DA. Precise estimates of mutation rate and spectrum in yeast. PNAS. 2014;111:E2310–E2318. doi: 10.1073/pnas.1323011111. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib85] Zhu J, Zhang ZT, Tang SW, Zhao BS, Li H, Song JZ, Li D, Xie Z, Glick B, Emr SD. A validated set of fluorescent-protein-based markers for major organelles in yeast (Saccharomyces cerevisiae) MBio. 2019;10:19. doi: 10.1128/mBio.01691-19. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib86] Zorov DB, Vorobjev IA, Popkov VA, Babenko VA, Zorova LD, Pevzner IB, Silachev DN, Zorov SD, Andrianova NV, Plotnikov EY. Lessons from the discovery of mitochondrial fragmentation (fission): A review and update. Cells. 2019;8:E175. doi: 10.3390/cells8020175. [DOI] [PMC free article] [PubMed] [Google Scholar]

[bib87] Zou X, Ratti BA, O’Brien JG, Lautenschlager SO, Gius DR, Bonini MG, Zhu Y. Manganese superoxide dismutase (SOD2): is there a center in the universe of mitochondrial redox signaling? Journal of Bioenergetics and Biomembranes. 2017;49:325–333. doi: 10.1007/s10863-017-9718-8. [DOI] [PubMed] [Google Scholar]

PERMALINK

Genetically controlled mtDNA deletions prevent ROS damage by arresting oxidative phosphorylation

Simon Stenberg

Jing Li

Arne B Gjuvsland

Karl Persson

Erik Demitz-Helin

Carles González Peña

Jia-Xing Yue

Ciaran Gilchrist

Timmy Ärengård

Payam Ghiaci

Lisa Larsson-Berglund

Martin Zackrisson

Silvana Smits

Johan Hallin

Johanna L Höög

Mikael Molin

Gianni Liti

Stig W Omholt

Jonas Warringer

Roles

Abstract

Introduction

Results

Paraquat impairs cell growth through mitochondrial O2∙− production

Swift adaptation to increased O2∙− production

Figure 1. Distinct adaptation to paraquat.

Figure 1—figure supplement 1. Titration of the paraquat (PQ) dose and design of adaptation experiment.

Figure 1—figure supplement 2. Comparison of predicted paraquat adaptation with the experimental data.

Mitophagy is not responsible for the swift first adaptation phase

Figure 2. Mitochondrial fragmentation precedes the swift adaptation to paraquat (PQ).

Figure 2—figure supplement 1. Paraquat (PQ) stress leads to rapid mitochondrial fragmentation.

mtDNA segmental deletions cause the swift adaptation to paraquat

Figure 3. mtDNA editing causes the early adaptation to paraquat.

Figure 3—figure supplement 1. Editing of mtDNA during the early adaptation to paraquat.

Figure 3—figure supplement 2. Homeostatic restoration of mtDNA copy numbers and ability for respiration after release from paraquat (PQ) stress.

Figure 3—figure supplement 3. Control of paraquat resistance through mtDNA deletions.

The deletion of mtDNA segments requires SOD2

Figure 4. The mtDNA editing critically involves Sod2.

Figure 4—figure supplement 1. The mtDNA deletion process critically involves Sod2, but not Sod1.

The mtDNA editing process requires anterograde mito-nuclear communication

Figure 5. The mtDNA editing critically involves anterograde mito-nuclear communication.

Sustained mtDNA deletion causes irrevocable mitochondrial impairment

Figure 6. Chronic exposure to paraquat causes irreversible mitochondrial impairment by sustained mtDNA editing.

Figure 6—figure supplement 1. Long-term exposure to paraquat causes genetic fixation of adaptation.

Figure 6—figure supplement 2. mtDNA loss during long-term exposure to paraquat.

Figure 6—figure supplement 3. Dot-plot mapping of long mtDNA reads from Nanopore sequencing of clones from the D1 and A7 paraquat adapted populations.

Figure 6—figure supplement 4. Nuclear genome evolution during long-term exposure to paraquat (PQ).

Figure 6—figure supplement 5. Chromosome duplications explain the second phase of adaptation to paraquat.

Chromosome duplications explain the second adaptation phase

Discussion

Materials and methods

Yeast cells

Founder strain

Deletion strains

Aneuploidic strains

Cox4-EGFP fusion strain

Cell cultivation media

Experimental evolution of cells

Establishing and cultivating frozen chronological records of cell populations

Tracking cell growth and adaptation

Counting cells in growing populations

Cell doubling time and adaptation

Counting cell generations

Maximum possible reduction in cell doubling time

RNA sequencing to measure SOD1, SOD2 and CCP1 expression

DNA sequencing of evolving cell populations

Long read (PacBio) sequencing of the YPS128 founder strain

Very long read (Oxford nanopore) sequencing

Resequencing of adapted populations and populations released from selection

Calling de novo point mutations

Calling aneuploidies

Calling mtDNA copy number change

Numerical model of evolving cell populations

Cell population parameters

Mutation effect sizes

Mutation rate parameters

Mutation effect size parameters

Quantitative PCR of mtDNA genes in evolving cell populations

Paraquat impairs cell growth through mitochondrial O₂^∙− production

Swift adaptation to increased O₂^∙− production