Cooperation and cheating are widespread evolutionary strategies. While cheating confers an advantage to individual entities within a group, competition between groups favors cooperation. Selfish or cheater mitochondrial DNA (mtDNA) proliferates within hosts while being selected against at the level of host fitness. How does environment shape cheater dynamics across different selection levels? Focusing on food availability, we address this question using heteroplasmic Caenorhabditis elegans. We find that the proliferation of selfish mtDNA within hosts depends on nutrient status stimulating mtDNA biogenesis in the developing germline. Interestingly, mtDNA biogenesis is not sufficient for this proliferation, which also requires the stress-response transcription factor FoxO/DAF-16. At the level of host fitness, FoxO/DAF-16 also prevents food scarcity from accelerating the selection against selfish mtDNA. This suggests that the ability to cope with nutrient stress can promote host tolerance of cheaters. Our study delineates environmental effects on selfish mtDNA dynamics at different levels of selection.
multilevel selectionnutrient availabilitycheatingmitochondriaheteroplasmymetabolismResearch organismC. eleganshttp://dx.doi.org/10.13039/100000057National Institute of General Medical SciencesGM123260PatelMaulik Rhttp://dx.doi.org/10.13039/100000057National Institute of General Medical Sciences1F31GM125344GitschlagBryan LThe Vanderbilt Diabetes Research and Training CenterPilot and Feasibility GrantGitschlagBryan LPatelMaulik RThe funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.Author impact statementFood supply and nutrient stress tolerance coordinately shape the multilevel selection dynamics of a mitochondrial cheater by promoting cheater persistence at both within-host and between-host levels of selection.Introduction
Life is generally organized into a hierarchy of cooperative collectives: multiple genes make up a genome, different genomes combine to form the eukaryotic cell, individual cells give rise to communities and multicellular organisms, and multicellular organisms are often organized into larger groups. New levels of organization emerge when natural selection favors cooperation and the loss of conflict between previously autonomous replicating entities, giving rise to a collective unit upon which selection can further operate (Michod et al., 2006; West et al., 2015; Queller and Strassmann, 2009). Cooperation thus underlies the evolution of larger, more complex biological systems (Fisher and Regenberg, 2019; Gulli et al., 2019; Michod et al., 2006; West et al., 2015; Hammerschmidt et al., 2014). However, because cooperators incur the near-term cost of contributing to the fitness of others for long-term benefit, cooperation creates opportunities for the emergence of selfish ‘cheater’ entities, which show up at multiple levels of the hierarchy of biological organization. One type of cheater, meiotic drive genes, facilitate their own transmission by compromising the fitness of gametes lacking them, with examples identified in plants, fungi, and animals (Bravo Núñez et al., 2018; Hammond et al., 2012; Hu et al., 2017; Larracuente and Presgraves, 2012; Schimenti, 2000). Cancer is characterized by unchecked cell proliferation and the monopolization of resources at the expense of other cells, constituting a form of cheating at the cellular level (Aktipis et al., 2015). Cheating behaviors likewise occur among many species of social animals (Riehl and Frederickson, 2016).
By benefiting from the contributions of cooperators without reciprocating, cheaters gain a fitness advantage (Aktipis et al., 2015; Dobata et al., 2009; Ghoul et al., 2014; Strassmann et al., 2000). This advantage can break down at higher levels of biological organization, which rely on cooperation at lower levels (Aktipis et al., 2015; de Vargas Roditi et al., 2013; Fiegna and Velicer, 2003; Moreno-Fenoll et al., 2017; Rainey and Rainey, 2003; Wenseleers and Ratnieks, 2004). Hence, selection can simultaneously favor different traits across the levels of the biological hierarchy, a phenomenon known as multilevel selection (de Vargas Roditi et al., 2013; Hammerschmidt et al., 2014; Shaffer et al., 2016; Takeuchi and Kaneko, 2019; Wilson and Wilson, 2007). Multilevel selection thus provides an explanation for the paradoxical coexistence of selfishness and cooperation in hierarchically structured populations.
Competition over limited resources shapes relative reproductive fitness, and hence Darwinian evolution. Accordingly, efficient resource utilization likely represents an adaptive benefit of cooperative groups (Koschwanez et al., 2013; Vanthournout et al., 2016). Interestingly, some studies have shown that resource abundance can promote public-goods cooperation, particularly in cases where resource abundance lowers the cost of cooperating (Brockhurst et al., 2008; Connelly et al., 2017; Sexton and Schuster, 2017). Conversely, resource scarcity can promote cooperation (Cao et al., 2015; Chisholm and Firtel, 2004; Koschwanez et al., 2013; Li and Purugganan, 2011; Pereda et al., 2017; Requejo and Camacho, 2011) and abundance can promote selfishness (Chen and Perc, 2014; Ducasse et al., 2015; Velicer et al., 1998). Given that cooperators and cheaters are favored by different levels of selection, taking multilevel selection into account can provide deeper insights into the relationship between resource availability and cooperator-cheater dynamics.
We sought to investigate the relationship between resource availability and multilevel selection using a mitochondrial heteroplasmy. Mitochondria cooperate with each other and with their host by supplying energy; in return, the nuclear genome supplies proteins and building blocks needed to replicate mitochondrial DNA (mtDNA). Mitochondrial organelles can contain multiple copies of mtDNA, which are usually non-recombining and can replicate throughout the cell cycle (Chatre and Ricchetti, 2013; Newlon and Fangman, 1975; Sena et al., 1975). This can give rise to a mixed (heteroplasmic) population of mtDNA variants that compete for transmission. Selfish mtDNA are those that undergo positive selection within hosts and negative selection at the level of host fitness, with examples documented in plants, fungi, and animals (Havird et al., 2019; Klucnika and Ma, 2019; Taylor et al., 2002). Hence, multilevel selection shapes the population dynamics of mitochondria (Dubie et al., 2020; Havird et al., 2019; Klucnika and Ma, 2019; Shou, 2015; Taylor et al., 2002).
How does resource availability shape the multilevel selection forces acting on selfish mtDNA? We address this using the model species Caenorhabditis elegans harboring the well-characterized heteroplasmic mutant genome uaDf5 (Figure 1A,B), hereinafter referred to as ∆mtDNA (Ahier et al., 2018; Gitschlag et al., 2016; Liau et al., 2007; Lin et al., 2016; Tsang and Lemire, 2002). This deletion mutation spans four protein-coding genes and seven tRNA genes (Figure 1B), disrupting gene expression and metabolic function. Previous work has shown that ∆mtDNA propagates at the expense of host fitness by exploiting regulatory mechanisms encoded by the host genome. For example, the copy number of the mutant genome increases in addition to—rather than at the expense of—wildtype mtDNA copy number (Gitschlag et al., 2016), suggesting that ∆mtDNA hitchhikes to higher levels by evading the host’s ability to regulate mtDNA copy number. The presence of ∆mtDNA also elicits the activation of host stress-response genes, inadvertently promoting further ∆mtDNA proliferation (Gitschlag et al., 2016; Lin et al., 2016). Here, we sought to expand the investigation of this biological cheater to include the ecologically relevant context of resource availability.
The <italic>uaDf5</italic> mutant variant (∆mtDNA) proliferates despite undermining host fitness, indicative of a cheater undergoing multilevel selection.
(A) Selfish mtDNA behaves as a biological cheater, outcompeting the cooperative wildtype mtDNA within hosts. (B) C. elegans mtDNA map showing uaDf5 deletion (dark red) in ∆mtDNA and color-coded genes: respiratory complex I (light red), complex III (yellow), complex IV (light blue), complex V (dark blue), ribosomal RNA (gray), tRNA (black), non-coding regions (thin line). (C) ∆mtDNA frequency across developmental stages of single broods from low (top, N = 94), intermediate (middle, N = 93), or high (bottom, N = 88) parental ∆mtDNA frequency (dotted lines). Mature adults were lysed at day 2 of adulthood, the same age at which the parents were lysed. (D) Basal and maximum aerobic respiration in age-synchronized L4 animals. Two-way ANOVA with Sidak’s multiple comparisons test. (E) Peak fecundity (viable progeny per hour per parent at day 2 of adulthood) binned according to the low end of the ∆mtDNA frequency distribution (below the population mean of 60%, N = 12) or the high end (above 60%, N = 12), with wildtype controls (N = 8). Brown-Forsythe and Welch ANOVA with Dunnett’s T3 multiple comparisons test. (F) Larval stage reached within 48 hr starting from age-synchronized embryos, plotted as a function of ∆mtDNA frequency. N = 35 nematodes per larval stage. Brown-Forsythe and Welch ANOVA with Dunnett’s T3 multiple comparisons test. All experiments featured in this figure used nematodes that were maintained on a diet of live OP50 E. coli at 20 °C.
Quantification of mtDNA copy number and ∆mtDNA frequency by droplet digital PCR.
(A) Map of mtDNA showing the uaDf5 deletion (∆mtDNA) and the strategy for oligonucleotide primer design for the multiplex quantification of ∆mtDNA and wildtype mtDNA simultaneously within the same reaction. Due to the deletion size, primers flanking the uaDf5 deletion (green arrows) amplify a PCR product off of the ∆mtDNA but not wildtype mtDNA template (see panel B). Likewise, primers complementary to a sequence within the region spanning the uaDf5 deletion (blue arrows) amplify a PCR product off of the wildtype mtDNA but not ∆mtDNA template. (B–C) Sample droplet digital PCR data plots showing mtDNA copy number in lysates from homoplasmic wildtype (B) and heteroplasmic (C) nematodes. Mutant frequency is determined from ∆mtDNA copy number over total copy number.
First, we isolate and measure selection on ∆mtDNA separately within individual hosts (sub-organismal) and between hosts (organismal). We then adapt the multilevel selection framework to study the effects of food availability and the physiology of nutrient stress tolerance on ∆mtDNA. Although diet and nutrient sensing govern overall mtDNA levels, the preferential proliferation of the selfish genome at the sub-organismal level depends on a key regulator of stress tolerance, namely the Forkhead box O (FoxO) transcription factor DAF-16. Diet restriction strengthens organismal selection against the selfish mtDNA, but only in the absence of DAF-16. We conclude that food availability and resilience to food scarcity govern the relative fitness of the cooperators and cheaters both within and between collectives.
ResultsAn experimental strategy to isolate selection on a selfish mitochondrial genome at different levels
Selection can act directly on individual mtDNA molecules within an organelle due to intrinsic replication advantage (Holt et al., 2014). Selection can also occur between organelles within a host cell (Lieber et al., 2019; Zhang et al., 2019), between cells within a multicellular host (Shidara et al., 2005), and finally between host organisms. By focusing on selection for mitochondrial genotype itself, we bypass the challenges facing the study of selection acting on organelles, which undergo fusion and fission dynamics and hence do not exist as discrete units. Moreover, the vast majority of mitochondrial content in the adult hermaphroditic nematode Caenorhabditis elegans is confined to the germline (Bratic et al., 2009), which exists as a contiguous syncytium of cytoplasm until the final stages of oocyte maturation (Pazdernik and Schedl, 2013). Sub-organismal selection thus predominantly reflects the biology of the female germline, where mtDNA variants compete for transmission (Figure 1A). Accordingly, here we focus on selection for mitochondrial genotype at the sub-organismal level as a single phenomenon, in addition to selection for mitochondrial genotype at the organismal level.
Using a multiplex droplet digital PCR (ddPCR) approach to quantify mitochondrial genotype (Figure 1—figure supplement 1A–C), we observed that ∆mtDNA frequency steadily rises across organismal development in a manner that depends on the initial inherited frequency of ∆mtDNA (Figure 1C), consistent with earlier work (Tsang and Lemire, 2002). The apparent upper limit of sub-organismal ∆mtDNA proliferation is indicative of the phenomenon of frequency-dependent selection, a common feature of cheater entities (Dobata and Tsuji, 2013; Dugatkin et al., 2005; Pruitt and Riechert, 2009; Riehl and Frederickson, 2016; Ross-Gillespie et al., 2007). Another important feature of cheaters is that their selection advantage tends to break down at higher levels of selection (Aktipis et al., 2015; de Vargas Roditi et al., 2013; Fiegna and Velicer, 2003; Moreno-Fenoll et al., 2017; Rainey and Rainey, 2003; Wenseleers and Ratnieks, 2004). Interestingly, although host stress-response mechanisms have previously been implicated in ∆mtDNA propagation (Gitschlag et al., 2016; Lin et al., 2016), such mechanisms do not appear to protect the host from the fitness cost incurred by harboring ∆mtDNA. On the contrary, we observed several indicators that ∆mtDNA proliferates while compromising host fitness, consistent with multilevel selection and in agreement with prior studies of this genome (Gitschlag et al., 2016; Liau et al., 2007; Lin et al., 2016). Indicators of reduced host fitness include reduced aerobic respiration (Figure 1D), in spite of elevated overall mitochondrial mass and the activation of mitochondrial stress-response mechanisms (Gitschlag et al., 2016; Lin et al., 2016). Other indicators that ∆mtDNA impacts host fitness include reduced fertility (Figure 1E) and slowed development (Figure 1F) in heteroplasmic animals. We therefore sought to quantitatively characterize the multilevel selection dynamics of ∆mtDNA.
To measure the impact of sub-organismal selection on the propagation of ∆mtDNA across generations, ∆mtDNA frequency was quantified longitudinally at successive developmental stages and across multiple parent-progeny lineages. Individual parent-progeny lineages were maintained in isolation from one another to minimize the confounding effect of organismal selection on ∆mtDNA frequency. Initially, we observed reduced ∆mtDNA frequency in embryos compared to their parents (Figure 2A), consistent with the notion of germline purifying selection (Ahier et al., 2018; Hill et al., 2014; Lieber et al., 2019; Ma et al., 2014; Stewart et al., 2008). However, ∆mtDNA proliferates across development, achieving even higher frequency on average in adult progeny than in their respective parents (Figure 2A). Moreover, the magnitude of this proliferation declines with increasing parental ∆mtDNA frequency (Figure 2B), consistent with the phenomenon of negative frequency-dependent selection. Overall, we have isolated and quantitatively measured the impact of selection at the sub-organismal level on ∆mtDNA propagation across generations.
Quantification of intergenerational changes in ∆mtDNA frequency due to selection at sub-organismal and organismal levels.
(A) ∆mtDNA frequency across parent-progeny lineages, maintained in isolation to minimize the effect of organismal selection. Each light gray line represents a single lineage consisting of a parent lysed individually followed by 3 of its progeny pooled and lysed together at each of 3 developmental time-points. Mature adults were lysed at day 2 of adulthood, the same age at which the parents were lysed, to ensure that parents and their adult progeny were age-matched. Box and whisker plots depict mean ∆mtDNA frequency and each quartile. Friedman test with Dunn’s multiple comparisons test. N = 30 lineages. (B) Shift in ∆mtDNA frequency per generation, obtained by subtracting ∆mtDNA frequency of mature adult progeny in panel (A) from ∆mtDNA frequency of the respective parent, plotted as a function of parental ∆mtDNA frequency. Red shaded region: 95% C.I. (C) Competition experiment designed to quantify organismal selection against ∆mtDNA. (D) Fraction of ∆mtDNA-carrying heteroplasmic individuals across 10 generations and eight replicate competed (red) versus non-competed (gray) lineages. Competed lineages consisted of heteroplasmic and homoplasmic wildtype individuals on the same food plates. Non-competed lineages consisted of heteroplasmic individuals only. Solid lines represent best-fit regressions across all replicate lineages. (E) Population-wide ∆mtDNA frequency across the organismal competition experiment. To isolate the change in ∆mtDNA frequency that occurs strictly due to organismal selection, we controlled for the confounding influence of sub-organismal ∆mtDNA dynamics by normalizing the total ∆mtDNA frequency of each lineage to the mean frequency of the non-competed lines at each generation. Because the non-competed lines consist entirely of heteroplasmic individuals, ∆mtDNA frequency across the non-competed lines is equal to mean sub-organismal ∆mtDNA frequency. The overall slope of the non-competing lines is therefore set to zero and the non-zero slope across the competing lines is due to the presence of wildtype animals (see panel C), allowing us to measure the effect of organismal selection by itself. Solid lines represent best-fit regressions across all replicate lineages. All experiments featured in this figure used nematodes that were maintained on a diet of live OP50 E. coli at 20 °C. Error bars: 95% C.I.
Summary statistics for frequency-dependent change in ∆mtDNA frequency at sub-organismal level and organismal selection against ∆mtDNA.
Change in ∆mtDNA frequency in competition experiments is due to organismal selection.
(A) PCR gel images reflecting individual adult nematodes sampled from the competition lines in the organism-level selection experiment. Approximately 48 adults were individually lysed from the first generation and every second generation thereafter. A single PCR band reflects Bristol strain nematodes (homoplasmic for wildtype mtDNA), whereas double bands reflect heteroplasmic animals carrying the ∆mtDNA allele. The numbers to the sides of the gel images label the replicate lineages. (B) Quantification of ∆mtDNA frequency across eight replicate competed (red) and non-competed (black) lineages. Same experiment as shown in Figure 2A but with y-axis expressing raw ∆mtDNA frequency measurements. Dark lines reflect best-fit regression across all replicate lineages. Nematodes were maintained on a diet of live OP50 E. coli at 20 °C.
To measure selection against ∆mtDNA strictly at the level of host fitness, we competed heteroplasmic animals carrying ∆mtDNA against their homoplasmic wildtype counterparts on the same food plate (Figure 2C). In parallel, we propagated non-competing control lines, which lacked wildtype animals. Consistent with organismal selection, we observed a decline in the fraction of individuals carrying ∆mtDNA across all eight replicate lineages (Figure 2D and Figure 2—figure supplement 1A). We also quantified ∆mtDNA frequency directly across all competing and non-competing lines using multiplex ddPCR, which revealed a dramatic decline in ∆mtDNA frequency across all eight competing lines (Figure 2—figure supplement 1B). This decline was not observed in the non-competing lines, which maintained ∆mtDNA frequency near 60% despite minor variation (Figure 2—figure supplement 1B). In order to isolate the effect of organismal selection on ∆mtDNA, we controlled for the confounding factors such as sub-organismal ∆mtDNA dynamics. To accomplish this, the ∆mtDNA frequencies of the competing lines were normalized to that of the non-competing lines at each generation. This effectively normalizes population frequency to individual (sub-organismal) frequency, since non-competing lines contain only ∆mtDNA-carrying individuals and thus their frequency is equal to average sub-organismal frequency. Moreover, normalizing to the non-competing lines sets the slope of ∆mtDNA frequency of those lines to zero (Figure 2E, gray lines). Then, whatever non-zero slope remains for the competing lines (Figure 2E) can be attributed to the presence of homoplasmic wildtype individuals (the only variable distinguishing the competing from non-competing lines). In conclusion, we have separately measured the effects of selection on ∆mtDNA at the sub-organismal and organismal levels, which we propose balance to allow the stable persistence of ∆mtDNA.
We next sought to investigate how resource availability affects the multilevel selection dynamics of ∆mtDNA. Nematodes were raised on food plates seeded with either a high or low concentration of E. coli (OP50 strain), which were UV-killed to prevent further bacterial growth (Figure 3—figure supplement 1A,B). Although UV-killed OP50 partially mimics diet restriction (Win et al., 2013), we found that nematodes raised on the more restricted (low concentration) diet harbored significantly lower ∆mtDNA frequency compared to those raised on a more abundant (high concentration) control diet (Figure 3A). Moreover, sub-organismal ∆mtDNA frequency is even higher in animals raised on live food (Figure 3—figure supplement 1C). Consistent with this observation, we notice a greater net shift in ∆mtDNA frequency from parent to adult progeny on live food (Figure 2A) compared to UV-killed food (Figure 3B). Despite the attenuated ∆mtDNA proliferation on UV-killed food, we still observe noteworthy dietary effects that result simply by varying the concentration of the UV-killed food. First, the initial selection against ∆mtDNA between parent and embryo was abolished by diet restriction, corresponding to a 100-fold dilution of the control diet (Figure 3B,C). Second, ∆mtDNA frequency rose from embryos to adults on a control diet, recovering from the initial purifying selection between parent and embryo (Figure 3B), but failed to do so in animals grown on a restricted diet (Figure 3C). These observations reveal complex, life-stage-specific effects of diet on ∆mtDNA dynamics: a plentiful diet selects against ∆mtDNA between parent and embryo but selects for ∆mtDNA across development.
∆mtDNA exploits nutrient supply and insulin signaling to proliferate at the sub-organismal level.
(A) ∆mtDNA frequency on restricted versus control diet. Nematodes were maintained at 20 °C on a diet of UV-killed OP50 E. coli. Mann-Whitney test. N = 8 pooled lysates of 5 age-synchronized day-4 adults each. (B–C) ∆mtDNA frequency across parent-progeny lineages raised on control (B) or restricted (C) diet. Each light gray line represents a single lineage consisting of a parent lysed individually followed by 3 of its progeny pooled and lysed together at each of 2 developmental time-points. Nematodes were maintained at 20 °C on a diet of UV-killed OP50 E. coli. Mature adults were lysed at day 2 of adulthood, the same age at which the parents were lysed, to ensure that parents and their adult progeny were age-matched. Box and whisker plots depict mean ∆mtDNA frequency and each quartile. Friedman test with Dunn’s multiple comparisons test. N = 24 lineages (control diet); N = 20 lineages (restricted diet). (D) FoxO-dependent insulin signaling cascade with C. elegans homologs in parentheses. (E) ∆mtDNA frequency among wildtype, daf-2(e1370) mutant, and daf-2(e1370);daf-16(mu86) double-mutant host genotypes, on a plentiful diet consisting of live OP50 E. coli. Kruskal-Wallis ANOVA with Dunn’s multiple comparisons test. N = 8 pooled lysates of 5 age-synchronized day-4 adults each. (F) mtDNA copy number of individuals in (E), normalized to wildtype mtDNA from wildtype nuclear background. Green and purple represent wildtype and ∆mtDNA copy number, respectively. Two-way ANOVA with Sidak’s multiple comparisons test. N = 8 pooled lysates of 5 age-synchronized day-4 adults each. (G) mtDNA copy number in homoplasmic adults of wildtype, daf-2(e1370) mutant, and daf-2(e1370);daf-16(mu86) double-mutant host genotypes, on a plentiful diet consisting of live OP50 E. coli. One-way ANOVA with Dunnett’s multiple comparisons test. N = 8 pooled lysates of 5 age-synchronized day-4 adults each. (H) mtDNA copy number in homoplasmic adults lacking ∆mtDNA of either wildtype or null daf-16(mu86) host genotype, on either daf-2 RNAi knockdown or empty-vector control conditions. Two-way ANOVA with Sidak’s multiple comparisons test. N = 8 lysates containing five pooled age-synchronized day-4 adults each. Experiments depicted in panels (E) through (H) used nematodes that were maintained at 16 °C during larval development and transferred at the L4 stage to 25 °C for adult maturation, corresponding to the permissive and restrictive temperatures for the daf-2(e1370) allele, respectively. Error bars: 95% C.I.
Diet and insulin signaling regulate mtDNA copy number and ∆mtDNA frequency.
(A) Schematic of experimental workflow for assaying the impact of dietary and pharmacological perturbations on ∆mtDNA frequency. (B) Experimental conditions under which changes in ∆mtDNA frequency were assayed. (C) The frequency of ∆mtDNA in age-synchronized, pooled adults (10 adults per sample) maintained at 20 °C on one of three food plates: UV-killed OP50 E. coli, live OP50 suspended in M9 buffer, and live OP50 suspended in LB medium. One-way ANOVA with Dunnett’s multiple comparisons test. (D) ∆mtDNA frequency among wildtype, daf-2(e1370) mutant, and daf-2(e1370);daf-16(mu86) double-mutant genotypes. One-way ANOVA with Dunnett’s multiple comparisons test. N = 8 pooled lysates of 5 age-synchronized adults each. (E) ∆mtDNA frequency in generation 1 versus generation 4, between wildtype and the temperature-sensitive allele daf-2(e1370). At each generation, embryos were transferred to the permissive temperature of 16 °C and maintained at this temperature during larval development to prevent developmental arrest; followed by transfer of L4-stage larvae to 25 °C, the restrictive temperature of the daf-2(e1370) allele, for progression to adulthood and the production of progeny for the subsequent generation. Two-way ANOVA with Sidak’s multiple comparisons test. N = 4 lysates containing five pooled age-synchronized adults each. (F) mtDNA copy number of individuals in (D), normalized to wildtype mtDNA from wildtype nuclear background. Green and purple represent wildtype and ∆mtDNA copy number, respectively. Two-way ANOVA with Sidak’s multiple comparisons test; p-values reflect comparisons of total mtDNA copy number. N = 8 pooled lysates of 5 age-synchronized adults each. (G) mtDNA copy number in homoplasmic adults of wildtype, daf-2(e1370) mutant, and daf-2(e1370);daf-16(mu86) double-mutant genotypes, raised from L4 stage to mature adulthood at 16 °C, the permissive temperature for the temperature-sensitive daf-2(e1370) allele. One-way ANOVA with Dunnett’s multiple comparisons test. N = 8 pooled lysates of 5 age-synchronized adults each. All experiments featured in this figure used nematodes that were lysed at day 4 of adulthood. For panels (D) through (G), nematodes were maintained on a diet of live OP50 E. coli. For panels (D), (F), and (G), nematodes were maintained throughout their lives at 16 °C, the permissive temperature for the daf-2(e1370) allele, providing a temperature control for Figures 3E, F and G, respectively. Error bars represent 95% C.I.
∆mtDNA exploits nutrient sensing to proliferate across development
To better understand the role of nutrient status in the cheating behavior of ∆mtDNA, we focused on sub-organismal ∆mtDNA proliferation across development. In particular, we hypothesized that the insulin-signaling pathway underlies ∆mtDNA proliferation. Insulin acts as a nutrient-dependent growth hormone and regulator of metabolic homeostasis (Figure 3D), tailoring the appropriate physiological responses to external nutrient conditions (Badisco et al., 2013; Danielsen et al., 2013; Lee and Dong, 2017; Lopez et al., 2013; Michaelson et al., 2010; Puig and Tjian, 2006; Shiojima et al., 2002; Das and Arur, 2017; Porte et al., 2005). Nematodes expressing a defective allele of the insulin receptor homolog daf-2 perceive starvation, even in presence of food. However, disrupting insulin signaling in young larvae causes the dauer phenotype, a form of developmental arrest (Gottlieb and Ruvkun, 1994). Thus, we used a temperature-sensitive daf-2 allele to conditionally inactivate insulin signaling. Animals were incubated at a permissive temperature (16 °C) to preserve insulin signaling during early larval development, thereby preventing developmental arrest. Animals were transferred to the restrictive temperature (25 °C) beginning at the last larval stage (L4) to inactivate insulin signaling during adult maturation when most mtDNA replication occurs. Compared to animals with intact insulin signaling, we observed lower ∆mtDNA frequency in animals expressing the defective daf-2 allele (Figure 3E). This difference was absent in control lines that were maintained at the permissive temperature of 16 °C (Figure 3—figure supplement 1D), suggesting that loss of insulin signaling limits ∆mtDNA proliferation. Moreover, no overall change in ∆mtDNA frequency occurred across four independent lineages of daf-2 mutants even after four consecutive generations (Figure 3—figure supplement 1E). In contrast, ∆mtDNA frequency increased substantially in wildtype controls. These data show that nutrient sensing via the insulin-signaling pathway is involved in sub-organismal proliferation of ∆mtDNA.
The insulin receptor communicates nutrient status to the cell largely through the negative regulation of the FoxO family of transcription factors (O-Sullivan et al., 2015), encoded by the gene daf-16 in C. elegans (Figure 3D). Nutrient limitation or inactivation of the receptor activates FoxO/DAF-16, resulting in altered expression of its target genes. Interestingly, deletion of daf-16 restores the proliferation of ∆mtDNA in animals defective for daf-2 function (Figure 3E). We conclude that insulin signaling promotes sub-organismal ∆mtDNA proliferation through the protein DAF-16.
How does DAF-16-dependent insulin signaling affect ∆mtDNA proliferation? The reduction of ∆mtDNA frequency by DAF-2 inactivation, and the rescue of ∆mtDNA frequency by loss of DAF-16 (Figure 3E), are almost entirely attributable to large differences in the copy number of ∆mtDNA but not of wildtype mtDNA (Figure 3F). In other words, insulin signaling promotes elevated total mtDNA copy number, perhaps as a driver of ∆mtDNA proliferation or as a consequence of it. To distinguish between these possibilities, we quantified copy number in animals lacking ∆mtDNA. Homoplasmic wildtype mtDNA copy number was quantified using the multiplex ddPCR method. To obtain relative copy number, raw mtDNA copy number across each nuclear genotype was normalized to that of the wildtype controls. In homoplasmic animals, we observed lower mtDNA copy number upon loss of insulin signaling, whether by daf-2 mutation (Figure 3G and Figure 3—figure supplement 1G) or by knockdown of daf-2 gene expression (Figure 3H), consistent with previous work in Drosophila (Wang et al., 2019). Loss of DAF-16 partially but significantly rescued copy number (Figure 3G,H and Figure 3—figure supplement 1G). Together, these data suggest that DAF-2 signaling inhibits DAF-16 to allow high mtDNA copy number, which permits sub-organismal ∆mtDNA proliferation.
How does DAF-16 suppress mtDNA copy number? This suppression could be achieved via mechanisms that result in the elimination of mitochondria, either at the organelle level through increased mitochondrial autophagy (mitophagy) or at the cellular level through increased apoptosis. Consistent with these possibilities, previous studies have identified FoxO/DAF-16 as a regulator of genes involved in autophagy and apoptosis (Murtaza et al., 2017; Webb and Brunet, 2014; Webb et al., 2016). We therefore reasoned that upon loss of insulin signaling, DAF-16 might suppress mtDNA copy number by upregulating either the destruction of mitochondrial organelles or cell death in the female germline. To test this idea, we genetically targeted PINK1/Parkin-dependent mitophagy using a deletion of pdr-1, encoding the C. elegans Parkin homolog. We genetically targeted apoptosis with a deletion of ced-3, encoding the terminator caspase in C. elegans. Disrupting either of these processes did not restore mtDNA copy number in daf-2 mutants (Figure 4A,B), nor did we observe increased germline apoptosis in daf-2 mutants (Figure 4—figure supplement 1A,B). Because mitochondrial degradation can occur in a PINK1/Parkin-independent manner (Allen et al., 2013; Di Rita et al., 2018; Hibshman et al., 2018), we therefore also tested for a potential role of mitochondrial fission, a common precursor of mitophagy, using a deletion in the gene drp-1, which encodes a dynamin-related protein important for mitochondrial fission. Although we observed an increase in mtDNA copy number when mitochondrial fission is disrupted in animals with intact insulin signaling (Figure 4C), consistent with reduced mitophagy, we did not observe a rescue of mtDNA copy number in daf-2These data suggest that the suppression of mtDNA content upon loss of insulin signaling is not mediated through the elimination of mitochondria by PINK/Parkin-dependent mitophagy, mitochondrial fission, or apoptosis.
DAF-16 activation upon loss of insulin signaling suppresses mtDNA content via regulation of germline proliferation.
(A–C) mtDNA copy number in age-synchronized adults of wildtype, temperature-sensitive daf-2(e1370) mutant, null daf-16(mu86) mutant, or double-mutant genotype. Copy number is also shown in wildtype, daf-2(e1370), and daf-2(e1370);daf-16(mu86) double-mutant adults each paired with pdr-1(gk448) (A), ced-3(ok2734) (B), or drp-1(tm1108) (C), representing loss-of-function alleles of the Parkin homologue, the terminator caspase CED-3, or dynamin-related protein, respectively. Copy number in daf-16(mu86) single-mutants is also shown. One-way ANOVA with Sidak’s multiple comparisons test. N = 8 lysates containing five pooled age-synchronized day-4 adults each. (D–E) Images (D) and quantification (E) of germline mitochondria labeled with TOMM-20::mCherry across wildtype, daf-2(e1370), daf-16(mu86), or double-mutant genotype. Each data point in (E) represents one adult in (D). One-way ANOVA with Sidak’s multiple comparisons test. (F–G) Representative images (F) and quantification (G) of DAPI-stained nuclei with mtDNA copy number across wildtype, daf-2(e1370), daf-16(mu86), or double-mutant genotype. Each gray data point represents one adult female gonad. For mtDNA copy number, N = 8 pooled lysates of 5 age-synchronized adults each. Two-way ANOVA with Sidak’s multiple comparisons test. (H) Schematic showing that upon loss of insulin signaling, FoxO/DAF-16 limits ∆mtDNA proliferation by restricting germline development. All experiments featured in this figure used nematodes that were maintained on a diet of live OP50 E. coli at 16 °C during larval development and transferred at the L4 stage to 25 °C for adult maturation, corresponding to the permissive and restrictive temperatures for the daf-2(e1370) allele, respectively. For panels (D) through (G), imaging was conducted on day-2 adults to visualize germlines at peak fecundity (Hughes et al., 2007). Error bars: 95% C.I.
Germline apoptosis in wildtype and <italic>daf-2</italic> mutant genotypes.
(A and B) Images (A) and quantification (B) of apoptosis as indicated by the engulfment of apoptotic cells by the reporter CED-1::GFP (white arrows in A), between age-synchronized adults of wildtype or temperature-sensitive daf-2(e1370) mutant genotype. Nematodes were maintained on a diet of live OP50 E. coli at 16 °C during larval development and transferred at the L4 stage to 25 °C for adult maturation, corresponding to the permissive and restrictive temperatures for the daf-2(e1370) allele, respectively, and were imaged at day 2 of adulthood. Mann-Whitney test. N = 10 age-synchronized adults. Error bars: 95% C.I.
Alternatively, DAF-16 might restrict mtDNA biogenesis. Nutrient availability and insulin signaling each promote development of the female germline (Angelo and Van Gilst, 2009; Drummond-Barbosa and Spradling, 2001; Michaelson et al., 2010; Narbonne and Roy, 2006; Shim et al., 2002), which harbors the vast majority of mtDNA in the adult nematode (Bratic et al., 2009). We observed that mitochondrial organelle quantity and mtDNA copy number are proportional to gonad size and cell count, respectively, across wildtype, daf-2 mutant, and daf-2;daf-16 double-mutants (Figure 4D–G). We therefore conclude that suppression of germline development by DAF-16 accounts for the reduced mtDNA content in insulin-signaling mutants (Figure 4H).
Because DAF-16 is required for mtDNA copy-number suppression upon loss of insulin signaling, we reasoned that DAF-16 should also be required for copy-number suppression in response to diet restriction. However, while diet restriction suppresses mtDNA copy number, this occurs independently of DAF-16 (Figure 5A). Given that ∆mtDNA frequency is sensitive to changes in total mtDNA copy number (Figure 3E–G), the effect of diet on total copy number suggests that diet might also modulate ∆mtDNA frequency independently of DAF-16. Remarkably, we only saw diet-dependent elevation in ∆mtDNA frequency when DAF-16 was present (Figure 5B). Moreover, while total mtDNA copy number and ∆mtDNA frequency each rose significantly across development on a control relative to restricted diet (Figure 5C), copy number rose by itself, with no accompanying change in ∆mtDNA frequency, in daf-16 mutants (Figure 5D). Because diet restriction and loss of DAF-16 were each found to attenuate ∆mtDNA proliferation, we conclude that nutrient abundance and DAF-16 are each necessary, but not sufficient individually, for ∆mtDNA to maintain a sub-organismal selection advantage.
The sub-organismal selection advantage of ∆mtDNA requires both nutrient abundance and DAF-16.
(A) Total mtDNA copy number in heteroplasmic individuals, wildtype versus null daf-16(mu86) host genotype, restricted versus control diet. N = 8 pooled lysates of 5 age-synchronized day-4 adults each. Two-way ANOVA with Bonferroni correction. (B) ∆mtDNA frequency of individuals in (A). Two-way ANOVA with Bonferroni correction. (C–D) Change in mtDNA copy number and ∆mtDNA frequency across development, with wildtype (C) versus null daf-16(mu86) (D) host genotype, on restricted versus control diet. Each data point represents the difference in copy number (horizontal axis) and ∆mtDNA frequency (vertical axis) between three pooled day-2 adults (age-matched to their respective parents) and three pooled embryos of the same brood. Mann-Whitney tests with Bonferroni correction. N = 22 wildtype, restricted diet; N = 24 wildtype, control diet; N = 24 daf-16(mu86), restricted diet; N = 24 daf-16(mu86), control diet. (E) Schematic showing that FoxO/DAF-16 is required in order for ∆mtDNA to take advantage of the increased mtDNA replication on an abundant diet. All experiments featured in this figure used nematodes maintained on a diet of UV-killed OP50 E. coli at 20 °C. Error bars: 95% C.I.
Nutrient status governs selection on ∆mtDNA at different levels
FoxO/DAF-16 regulates numerous genes involved in stress tolerance (Klotz et al., 2015; Martins et al., 2016; Murphy et al., 2003; Tepper et al., 2013; Webb et al., 2016) and promotes organismal survival during nutrient scarcity (Greer et al., 2007; Hibshman et al., 2017; Kramer et al., 2008). We therefore asked whether nutrient availability and DAF-16 affect selection on ∆mtDNA at both the organismal and sub-organismal levels. Sub-organismal selection was quantified as before (see Figure 2B), under restricted versus control diets, in the presence versus absence of DAF-16 (Figure 6A,B). Organismal selection was quantified under each of these same conditions, using the competition method previously described (see Figure 2C–E). In populations with wildtype DAF-16, diet restriction did not significantly affect the decline in ∆mtDNA frequency at the level of organismal selection (Figure 6C,D and Figure 6—figure supplement 1A). However, diet restriction accelerated the decline of ∆mtDNA frequency at the level of organismal selection among daf-16 mutants (Figure 6E,F and Figure 6—figure supplement 1B). These data indicate that although food scarcity can strengthen selection against ∆mtDNA at the organismal level, DAF-16 protects ∆mtDNA from this effect.
Nutrient status impacts multilevel selection dynamics of ∆mtDNA.
(A–B) Sub-organismal shift in ∆mtDNA frequency per generation, similar to Figure 2B, in wildtype (A) or null daf-16(mu86) (B) host genotype, on restricted or control diet. Adults were lysed at day 2 of adulthood, the same age at which the parents were lysed, to ensure that parents and their adult progeny were age-matched. Regressions compared using analysis of covariance. (C) Fraction of ∆mtDNA-carrying individuals at generation 8 of the competition experiment shown in (D) and Figure 6—figure supplement 1A, normalized to control-diet lines. Two-tailed Welch’s t-test. (D) Organismal selection against ∆mtDNA as measured by population-wide ∆mtDNA frequency relative to average sub-organismal (heteroplasmic) frequency, similar to Figure 2E, in competing lineages of wildtype nuclear background, maintained on restricted or control diet. To isolate the change in ∆mtDNA frequency that occurs strictly due to organismal selection, we controlled for the confounding influence of sub-organismal ∆mtDNA dynamics by normalizing overall ∆mtDNA across each population to that of the non-competed lines at each generation. Because all individuals within the non-competed lines contain ∆mtDNA, the frequency across a non-competing population is equal to the average sub-organismal ∆mtDNA. Hence, normalizing ∆mtDNA to the non-competing lines accounts for sub-organismal ∆mtDNA dynamics and reveals the decline in ∆mtDNA that occurs strictly due to selection at the level of organismal fitness. Solid lines reflect mean normalized ∆mtDNA frequency. Non-competed lines not shown for visual simplicity. Linear regression analysis. (E) Fraction of ∆mtDNA-carrying individuals with daf-16(mu86) nuclear background at generation 8 of the competition experiment shown in (F) and Figure 6—figure supplement 1B, normalized to control-diet lines. Two-tailed Welch’s t-test. (F) Organismal selection against ∆mtDNA as measured by population-wide ∆mtDNA frequency relative to average sub-organismal (heteroplasmic) frequency, similar to Figure 2E and (D), in competing lineages of daf-16(mu86) nuclear background, maintained on restricted or control diet. The ∆mtDNA frequency of each line, at each generation, was normalized to that of the non-competed lines in order to control for the confounding influence of sub-organismal ∆mtDNA dynamics, as was done in (D). Solid lines reflect mean normalized ∆mtDNA frequency. Non-competed lines not shown for visual simplicity. Linear regression analysis. (G) Schematic showing the influence of FoxO/DAF-16 on organismal selection against ∆mtDNA. Specifically, FoxO/DAF-16 protects ∆mtDNA from greater organismal selection during nutrient stress. (H) ∆mtDNA frequency, normalized to starting frequency, in non-competing lineages from the organismal competition experiment shown in (D) and (F). For comparison, dotted brown line represents data of non-competing data from the competition experiment on live OP50 E. coli (Figure 2—figure supplement 1B, gray lines). Linear regression analyses with Bonferroni correction for multiple comparisons. All experiments featured in this figure used nematodes maintained on a diet of UV-killed OP50 E. coli at 20 °C. Shaded regions show 95% C.I.
Summary statistics for frequency-dependent change in ∆mtDNA frequency at sub-organismal level, by diet and host genotype.
Summary statistics for frequency-dependent change in ∆mtDNA frequency at organismal level, by diet and host genotype.
Quantification of ∆mtDNA frequency in competition experiments to measure organismal selection under altered dietary conditions and <italic>daf-16</italic> genotype.
(A–B) Same data presented in Figure 6C–E but with y-axis representing raw ∆mtDNA frequency measurements. Dark lines reflect best-fit regression across all replicate lineages. Nematodes were maintained on diets of UV-killed OP50 E. coli at 20 °C.
Finally, we sought to integrate our observations of sub-organismal and organismal selection for each of the four conditions tested. We observed that the sub-organismal selection advantage of ∆mtDNA is compromised by diet restriction, loss of DAF-16, or both (Figures 5, 6A,B). Furthermore, diet restriction was observed to accelerate organismal selection against ∆mtDNA, but only in the absence of DAF-16 (Figure 6C–F and Figure 6—figure supplement 1). Taken together, these observations predict that the strongest net selection against ∆mtDNA occurs in populations lacking DAF-16 and experiencing food scarcity (Figure 6G), and the weakest overall selection occurs in populations with DAF-16 and experiencing food abundance, with the remaining two conditions each experiencing an intermediate strength of selection. Measuring ∆mtDNA frequency across non-competing heteroplasmic populations afforded the opportunity to test this prediction. Remarkably, this prediction is consistent with our observation (Figure 6H), even though UV-killed food compromises ∆mtDNA propagation even in the control diet (Figure 6H, compare gray to dotted brown line). Combined, our data reveal numerous ways in which diet and host genotype interact to shape the multilevel selection dynamics of a cheater genome (Figure 7).
Summary of influence of diet and nutrient stress tolerance on multilevel selection dynamics of ∆mtDNA.
At the sub-organismal level, FoxO/DAF-16 influences ∆mtDNA dynamics via two separate functions. On one hand, loss of insulin signaling results in activation of FoxO/DAF-16, which inhibits germline development (Figure 4). On the other hand, ∆mtDNA preferentially propagates by taking advantage of dietary nutrients but only when FoxO/DAF-16 is present (Figure 5), indicating that nutrient supply and FoxO/DAF-16-dependent nutrient sensing are each necessary, but not sufficient individually, for ∆mtDNA proliferation. During conditions of food scarcity, FoxO/DAF-16 partially shields ∆mtDNA from organismal selection (Figure 6), suggesting that nutrient supply and FoxO/DAF-16 promote ∆mtDNA propagation across organismal and sub-organismal selection levels.
Discussion
Multilevel selection offers a powerful explanatory framework to understand cooperator-cheater dynamics. However, investigations of multilevel selection face the challenge of trying to account for the confounding influence of selection acting at one level while estimating the strength of selection at a different level (Goodnight, 2015; Goodnight et al., 1992; Heisler and Damuth, 1987). To overcome these challenges, we developed an approach to empirically quantify selection for cheater mtDNA at the sub-organismal (within-host) level by tracking cheater frequency within isolated parent-progeny lineages. At the organismal (between-host) level, we devised competition experiments enabling us to identify and measure the change in population-wide cheater frequency that occurs strictly due to the cost that the cheater imposes on host fitness. Our methodology not only makes it possible to empirically measure selection at different levels, but it also provides a powerful experimental approach that can be applied broadly to future studies seeking mechanistic insight on cooperator-cheater dynamics in hierarchically structured populations.
At the sub-organismal level, we note two trends describing the dynamics of the cheater genome ∆mtDNA. First, ∆mtDNA frequency declines between parent and embryo (Figures 1C and 2A). This suggests germline purifying selection against deleterious mtDNA, a phenomenon observed across many species (Ahier et al., 2018; Fan et al., 2008; Hill et al., 2014; Lieber et al., 2019; Ma et al., 2014; Stewart et al., 2008). Although the molecular basis for germline purifying selection against ∆mtDNA is unknown, recent work in Drosophila has shown that mitochondrial protein synthesis in oocytes is localized around healthy mitochondria, providing a selection advantage for genomes that lack deleterious mutations (Zhang et al., 2019). Intriguingly, the same group also recently found that insulin signaling mediates purifying selection against a deleterious mtDNA variant in Drosophila eggs in a putatively FoxO-dependent manner (Wang et al., 2019), providing a potential basis by which maternal diet influences purifying selection between parent and embryo, as we observed (Figure 3B, C). Whether similar mechanisms underlie germline purifying selection against ∆mtDNA in C. elegans remains to be explored.
Following the initial decline from parent to embryo, ∆mtDNA proliferates across development in a frequency-dependent manner, a common feature of cheater entities (Dobata and Tsuji, 2013; Dugatkin et al., 2005; Pruitt and Riechert, 2009; Riehl and Frederickson, 2016; Ross-Gillespie et al., 2007). Specifically, the sub-organismal advantage of ∆mtDNA declines as its frequency approaches the range of 75–80% (Figures 1C and 2B). One possible explanation for this observation involves resource availability. Previous work has suggested that ∆mtDNA copy number increases in addition to—not at the expense of—wildtype mtDNA (Gitschlag et al., 2016). In other words, as ∆mtDNA frequency increases, so does total mtDNA copy number. The apparent upper limit observed for sub-organismal ∆mtDNA proliferation could therefore reflect a depletion of resources required for genome replication. Another possibility is the activation of policing mechanisms, a common strategy for enforcing cooperation (Özkaya et al., 2017; Riehl and Frederickson, 2016). The host genome might prevent ∆mtDNA from rising beyond a certain level by increasing the targeted degradation of underperforming organelles via mitophagy. Consistent with this possibility, deletion of the mitophagy gene pdr-1 was previously associated with an increase in ∆mtDNA frequency (Gitschlag et al., 2016; Valenci et al., 2015), suggesting that a policing mechanism encoded by the host genome limits the extent to which the cheater can proliferate.
In addition to ∆mtDNA frequency, we found temperature to be another condition that influences mitochondrial genome dynamics, with elevations in both ∆mtDNA frequency and overall mtDNA copy number at 25 °C (Figure 3E,F) compared to 16 °C (Figure 3—figure supplement 1D,F). Although the mechanistic basis for this temperature effect is unknown, one possible explanation involves the host response to stress. The presence of ∆mtDNA compromises mitochondrial function and elicits the activation of a physiological stress response that includes genes involved in mitochondrial biogenesis and protein quality control (Gitschlag et al., 2016; Lin et al., 2016). By seeking to restore mitochondrial function, the nuclear genome inadvertently promotes the propagation of ∆mtDNA in a vicious cycle (Gitschlag et al., 2016; Lin et al., 2016). Interestingly, more recent work has shown that warm temperature adversely affects mitochondrial function in adult nematodes (Gaffney et al., 2018), which raises the possibility that warm temperature could mimic the presence of ∆mtDNA in key ways that contribute to the propagation of the mutant genome. Future studies seeking to mechanistically characterize the impact of environmental stress as a modulator of ∆mtDNA proliferation may therefore benefit from considering temperature as a variable.
After measuring sub-organismal dynamics, we explored selection at the organismal level. Consistent with the predictions of multilevel selection, we quantitatively showed that selection at the sub-organismal level favors the cheater genome while selection at the organismal level favors the cooperative genome. These levels of selection appear to balance when ∆mtDNA frequency is near 60%, enabling ∆mtDNA to persist at this frequency across many generations (Tsang and Lemire, 2002). Having separately measured the effects of selection at different levels on a biological cheater, we turned to the question of how nutrient status influences selection at these different levels.
The role of nutrient status in shaping cooperator-cheater dynamics is not well understood. We first characterized the effect of maternal diet and nutrient sensing on sub-organismal cheater mtDNA dynamics. We found that nutrient abundance and insulin signaling promote mtDNA biogenesis in the germline, thereby providing the niche space for cheater proliferation. We further showed that the stress-response transcription factor FoxO/DAF-16 is necessary for the cheater to take advantage of nutrient supply and proliferate. Our findings reveal that while nutrient abundance may be necessary, it is not sufficient to promote cheater proliferation.
Interestingly, while we find that nutrient scarcity promotes the cooperative over cheater genotype, nutrient abundance is known to select for cooperation in other systems. We propose that the impact of nutrient availability depends on whether the cost or the benefit of cooperation is predominantly affected. Nutrient abundance can promote cooperation by reducing the cost of making a cooperative contribution (Brockhurst et al., 2008; Connelly et al., 2017; Sexton and Schuster, 2017). Alternatively, scarcity can increase the benefit of cooperating, a phenomenon we observe in heteroplasmy dynamics. Since the cheater genotype inflicts lower rates of respiration despite increased overall mtDNA levels (Tsang and Lemire, 2002), the cooperative wildtype genotype achieves a greater bioenergetic payoff per nutrient invested. When nutrients are scarce, a sub-organismal shift occurs in favor of the more metabolically efficient wildtype genome. In conclusion, we find that the benefit of cooperation, as indicated by cooperator-biased replication, increases during nutrient scarcity and decreases during abundance.
In addition to nutrient abundance, DAF-16 is also required for sub-organismal cheater mtDNA proliferation. This could occur through compensatory biogenesis that favors underperforming organelles, inadvertently biasing replication toward the cheater genotype. Consistent with this possibility, FoxO/DAF-16 has been identified as a regulator of genes associated with mitochondrial biogenesis (Tepper et al., 2013; Webb et al., 2016). Alternatively, DAF-16 might passively permit cheater proliferation by alleviating stress. DAF-16 up-regulates the expression of multiple genes involved in energy metabolism and antioxidant defense (Depuydt et al., 2014; Tepper et al., 2013; Webb et al., 2016). By seeking to rescue ATP synthesis and protect against cellular damage, DAF-16 may relax the sub-organismal selective pressure to maintain optimal mitochondrial function, thereby permitting the spread of deleterious mtDNA mutants.
How does resource availability affect selection on cooperators and cheaters at the level of competing groups? Note that the female germline harbors the mtDNA molecules that compete for transmission. Selection on mtDNA genotype at the organismal level can thus be viewed as a group-level phenomenon, while sub-organismal selection represents the within-group level. On one hand, if resource scarcity selects for cooperation, groups with a higher proportion of cooperators should gain an extra fitness advantage over other groups during times of scarcity. On the other hand, exposure to cheating can lead to an evolutionary arms race whereby cooperators acquire resistance to cheaters, a phenomenon observed in bacteria and social amoebae (Hollis, 2012; O'Brien et al., 2017; Khare et al., 2009). Could food scarcity select for adaptations that reduce the impact of cheaters on group fitness? We propose that DAF-16 functions as an example of this type of stress tolerance. Although diet restriction compromised the sub-organismal advantage of ∆mtDNA, it had no effect on the organismal disadvantage, provided DAF-16 is present. However, in daf-16 mutants, diet restriction intensified organismal selection against ∆mtDNA. We conclude that FoxO/DAF-16, known to prolong organismal survival during nutrient deprivation (Greer et al., 2007; Hibshman et al., 2017; Kramer et al., 2008), prevents food scarcity from subjecting the cheater to stronger organismal selection. Broadly, our findings suggest that the ability to cope with scarcity can promote group-level tolerance to cheating, inadvertently prolonging cheater persistence.
See Materials and methods: Knockdown of gene expression
Sequence-based reagent
PCR primers used in this study
This study
See Materials and methods: Genetic crosses and genotyping; Quantification of mtDNA copy number and ∆mtDNA frequency
Peptide, recombinant protein
BlpI restriction endonuclease
New England Biolabs
Cat#R0585L
See Materials and methods: Genetic crosses and genotyping
Chemical compound, drug
Isopropyl-β-D-thiogalactopyranoside
Research Products International
Cat#I56000-1
See Materials and methods: Genetic crosses and genotyping
Chemical compound, drug
4’,6-diamidino-2-phenylindole (DAPI)
Thermo Fisher Scientific
Cat#D1306
Chemical compound, drug
Paraformaldehyde
Electron Microscopy Sciences
Cat#15710
Chemical compound, drug
Levamisole
Fisher Scientific
Cat#AC187870100
Commercial assay, kit
DreamTaq Green DNA Polymerase
Thermo Fisher Scientific
Cat#EP0713
Commercial assay, kit
Seahorse XFe96 FluxPak
Agilent
Cat#102601–100
Other
Eppendorf 96-well twin.tec PCR plates
Fisher Scientific
Cat#951020303
Other
QX200 ddPCR EvaGreen Supermix
Bio-Rad
Cat#1864034
Other
Automated Droplet Generation Oil for EvaGreen
Bio-Rad
Cat#1864112
Other
DG32 Automated Droplet Generator Cartridges
Bio-Rad
Cat#1864108
Other
Droplet Reader Oil for ddPCR
Bio-Rad
Cat#1863004
Software, algorithm
QuantaSoft
Bio-Rad
Cat#1864011
Software, algorithm
Zen
Carl Zeiss Microscopy GmbH
RRID:SCR_013672
Software, algorithm
Prism eight for macOS
GraphPad Software, Inc
RRID:SCR_002798
Version 8.1.2
Software, algorithm
ImageJ
Wayne Rasband, NIH
RRID:SCR_003070
Version 1.49
Nematode culture
C. elegans strains used in this study were maintained on 60 mm standard nematode growth medium (NGM) plates seeded with live OP50 E. coli bacteria as a food source, unless otherwise indicated below. Nematode strains were incubated at 20 °C unless otherwise indicated. Age-matched nematodes were used in all experiments with the exception of the multigenerational competition experiment (see below).
Nematode lysis
To prepare nematodes for genotyping and quantification of mtDNA copy number and ∆mtDNA frequency, nematodes were lysed using the following protocol. Nematodes were transferred to sterile PCR tubes or 96-well PCR plates containing lysis buffer with 100 µg/mL proteinase K. Lysis buffer contained 50 mM KCL, 10 mM Tris pH 8.3, 2.5 mM MgCl2, 0.45% Tween 20, 0.45% NP-40 (IGEPAL), and 0.01% gelatin, in deionized H2O. Volume of lysis buffer varied by worm count: 10 µL for individual adults, pooled larvae, or pooled embryos; 20 µL for 5 or 10 pooled adults; 50 µL for pooled nematodes of mixed age (competition experiments, see below). Each tube or plate was then incubated at −80 °C for 10 min, then at 60 °C for 60 min (90 min for pooled nematodes), and then at 95 °C for 15 min to inactivate the proteinase K. Nematode lysates were then stored at −20 °C.
Genetic crosses and genotyping
To control for nuclear effects on ∆mtDNA proliferation, hermaphroditic nematodes carrying the ∆mtDNA allele uaDf5 were serially back-crossed into a male stock of the Bristol (N2) C. elegans nuclear background for six generations. To investigate the role of insulin signaling in selfish mitochondrial genome dynamics, the alleles daf-2(e1370) and daf-16(mu86) were introduced to the ∆mtDNA heteroplasmic lineage by classical genetic crosses. To investigate the mechanistic basis by which the insulin signaling pathway regulates mtDNA levels, mutant alleles affecting various putative downstream processes were genetically crossed into the insulin signaling-defective nuclear genotypes. Specifically, the parkin-dependent mitophagy-defective pdr-1(gk448), the mitochondrial fission-defective drp-1(tm1108), and the apoptosis-defective ced-3(ok2734) were each genetically combined with daf-2(e1370), both with and without the daf-16(mu86) allele. Nuclear genotype was confirmed by PCR using the following oligonucleotide primers:
The insulin signaling receptor allele daf-2(e1370) introduces a point mutation that eliminates a BlpI restriction endonuclease recognition site. Following PCR amplification, daf-2 PCR products were incubated with BlpI and New England BioLabs CutSmart buffer at 37 °C for 2 hr prior to gel electrophoresis. Fluorescent reporters used in this study were genotyped by fluorescence microscopy.
Quantification of mtDNA copy number and ∆mtDNA frequency
Quantification of mtDNA copy number and ∆mtDNA frequency was accomplished using droplet digital PCR (ddPCR). Nematodes were lysed as described above. Lysates were then diluted in nuclease-free water, with a dilution factor varying depending on nematode concentration: 20x for embryos, 200x for pooled larvae, 200x for single adults, 1000x for pooled adults, 20,000x for pooled nematodes of mixed age from the competition experiments (control diet) or 2000x for pooled nematodes of mixed age from the competition experiments (restricted diet). The lower dilution factor for the lysates collected from the restricted diet condition was due to the smaller population sizes of nematodes raised on a restricted diet, which arises from reduced fecundity under diet restriction and was reflected in the number of nematodes present in these lysates. Next, either 2 µL or 5 µL of each dilute nematode lysate was combined with 0.25 µL of a 10 µM aliquot of each of the following oligonucleotide primers:
For quantifying wildtype mtDNA:
5’-GTCCTTGTGGAATGGTTGAATTTAC-3’
5’-GTACTTAATCACGCTACAGCAGC-3’
For quantifying ∆mtDNA:
5-‘CCATCCGTGCTAGAAGACAAAG-3’
5-‘CTACAGTGCATTGACCTAGTCATC-3’
Mixtures of dilute nematode lysate and primer were combined with nuclease-free water and Bio-Rad QX200 ddPCR EvaGreen Supermix to a volume of 25 µL in Eppendorf 96-well twin.tec PCR plates. Droplet generation and PCR amplification were performed according to manufacturer protocol with an annealing temperature of 58 °C. For amplification of heteroplasmic nematode lysates, wildtype and ∆mtDNA primers were combined in the same reaction, and each droplet was scored as containing either wildtype or mutant template using the 2D amplitude (dual-wavelength) clustering plot option in the Bio-Rad QuantaSoft program.
Respiration assay
Basal and maximum oxygen consumption rates were measured using the Seahorse XFe96 Analyzer in the High Throughput Screening Facility at Vanderbilt University. One day before experimentation, each well of a 96-well sensor cartridge that comes as part of the Seahorse XFe96 FluxPak was incubated with 200 μL of the Seahorse XF Calibrant Solution. On the day of the experiment, 10–20 L4-stage animals were randomly sampled from either a stock population stably maintaining ∆mtDNA in the range of 50–80% (population mean approximately 60%), or from a wildtype control (Bristol strain). The animals were placed into each well of the cell culture microplate. Wells contained either M9 buffer or 10 μM FCCP. After calibration, 16 measurements were performed at room temperature. Measurements 12 through 16 were averaged and normalized to number of worms per well.
Fertility
To assay fertility, day-2 adult nematodes were individually transferred onto NGM plates seeded with live OP50 E. coli and incubated at 20 °C for 4 hr. The adults were then individually lysed as described above. Fertility was scored as the average number of viable progeny produced per hour during the 4 hr window, where viable progeny were identified as those that had progressed from embryos to larvae within 24 hr of being laid. The ∆mtDNA frequency of each parent was determined using ddPCR as described above. We assayed 48 individual ∆mtDNA-containing parents for the purpose of correlating ∆mtDNA frequency with fecundity using a linear regression. However, the regression was not significant. We then collapsed the heteroplasmic data points into two bins corresponding to low and high ∆mtDNA frequency (below and above the population mean of 60%, respectively) in order to compare their corresponding fecundities to that of a wildtype control. We set the sample size of the high-∆mtDNA frequency bin from N = 35 to N = 12, to match the sample size of the low-∆mtDNA frequency bin. To accomplish this, 12 samples were randomly selected to be retained and 23 samples were randomly selected to be discarded. To confirm that this did not affect our statistical analysis, this was repeated three times and statistical analysis was performed on each data set containing a randomly-selected sample of N = 12 for the high-∆mtDNA frequency bin. In each case, the presence of ∆mtDNA at >60% frequency continued to correspond to a significantly lower fertility rate than wildtype controls.
Development
The impact of ∆mtDNA levels on development was assayed by comparing ∆mtDNA frequency with developmental stage for each nematode in a population of age-synchronized larvae. To age-synchronize larvae, multiple mature heteroplasmic adults carrying ∆mtDNA in the Bristol nuclear background were transferred to an NGM plate seeded with live OP50 E. coli and allowed to lay eggs at 20 °C for 2 hr. Adults were then removed from the plate. After 48 hr, each nematode was individually lysed and its respective larval stage (L2, L3, or L4) was annotated. None of the nematodes had yet reached adulthood at this point. Embryos that failed to transition to larvae were discarded. The ∆mtDNA frequency of each larval nematode was determined using ddPCR as described above.
Sub-organismal selection assay
Sub-organismal selection for ∆mtDNA was quantified by measuring changes in ∆mtDNA frequency as a function of developmental stage, and as a function of initial (parental) ∆mtDNA frequency, within a single generation. This was accomplished using two complementary approaches. In the first approach, three individual age-synchronized parents were selected according to initial ∆mtDNA frequency (parents with low, middle, and high frequency). One age-matched (L4-stage) nematode was picked at random under a dissecting microscope from each line respectively maintained under artificial selection for low (<50%), medium (50–70%), and high (>70%) ∆mtDNA frequency. Each of these nematodes was placed onto a fresh NGM plate seeded with live OP50 E. coli and incubated for 2 days at 20 °C. Each day-2 adult was then transferred to a fresh food plate every 4 hr and allowed to lay embryos. At each 4 hr time point, approximately one third of the embryos produced were individually lysed. After 12 hr, the adults were individually lysed. A 12 hr time window for embryo production was chosen in order to generate a sufficient offspring count to allow for the establishment of single-brood frequency distributions of ∆mtDNA. The 12 hr time window was divided into 4 hr segments in order to maintain age-synchronicity, as each larva was lysed within 4 hr of being laid across the entire 12 hr period. After 2 days at 20 °C, approximately one third of the L4-stage larvae were individually lysed in the same 4 hr segments to maintain age synchronicity. After an additional 2 days at 20 °C, the remaining one third of offspring were individually lysed in 4 hr segments, as they reached the same age at which their respective parent was lysed. The ∆mtDNA frequency of each individual was determined using ddPCR as described above and a ∆mtDNA frequency distribution was generated for each offspring life stage.
In the second approach, multiple L4-stage heteroplasmic nematodes were picked at random under a dissecting microscope from the stock of nematodes carrying ∆mtDNA in the Bristol nuclear background. These larvae were transferred to a fresh food plate and incubated for 2 days at 20 °C. The day-2 adults were then segregated onto individual plates and incubated for 4 hr at 20 °C to produce age-synchronized progeny. After 4 hr, each parent was individually lysed. Three embryos from each parent were also lysed at the same time, in one pooled lysate per three same-parent embryos. After 2 days, three L4-stage larvae were pooled and lysed from each parent, similar to the lysis of embryos. After another 2 days, three adult progeny were pooled and lysed from each parent as they reached the age at which the parents were lysed. Each parent-progeny lineages was individually segregated from the rest. Since ∆mtDNA impacts fecundity, the progeny from parents on the lower end of the ∆mtDNA frequency are expected to be overrepresented in the offspring sampled from a mixed cohort of parents. Lineages were therefore segregated to ensure that the ∆mtDNA frequency from each progeny lysate was being compared with that of its own respective parent, in order to minimize the effect of organism-level selection on ∆mtDNA. In addition, progeny from each time-point were lysed in pools of three to reduce the effect of random drift on ∆mtDNA frequency. The ∆mtDNA frequency of parents and each developmental stage of progeny was determined using ddPCR as described above. For the measurement of sub-organismal selection on ∆mtDNA under nutrient-variable conditions, each parent was raised from embryo to adult under its respective dietary condition (diet restriction or control).
Experimental evolution (organismal selection)
Selection against ∆mtDNA that occurs strictly at the level of organismal fitness was measured using a competition assay. Heteroplasmic nematodes carrying ∆mtDNA in the Bristol nuclear background were combined with Bristol-strain nematodes on 10 cm NGM plates seeded with live OP50 E. coli. For the first generation, heteroplasmic and Bristol strain nematodes were age-synchronized. Age synchronization was accomplished using a bleaching protocol. Nematodes from a mixed-age stock food plate were washed off the plate and into a sterile 1.7 mL microcentrifuge tube with nuclease-free water. The water was brought to a volume of 750 µL. The volume of each tube was brought to 1 mL by adding 100 µL of 5 N NaOH and 150 µL of 6% bleach. Each nematode tube was incubated at room temperature for 10 min with light vortexing every 2 min to rupture gravid adults and release embryos. Nematode tubes were centrifuged for 1 min at 1000x g to pellet the nematode embryos. To wash the nematode pellets, the supernatant was removed and replaced with 1 mL of nuclease-free water. After a second spin for 1 min at 1000x g, the water was removed and the nematode embryos were resuspended in 100 µL M9 buffer. The resuspended embryos were then transferred to glass test tubes containing 500 µL M9 buffer and incubated overnight at room temperature on a gentle shaker to allow hatching and developmental arrest at the L1 larval stage. On the following day, a glass Pasteur pipette was used to transfer approximately equal quantities of heteroplasmic and homoplasmic-wildtype nematodes onto the 10 cm food plates. Approximately 500 nematodes were transferred to each plate. In addition to eight competition lines, eight control lines were established by transferring only heteroplasmic nematodes randomly selected from the same overnight incubation tubes onto food plates, with no homoplasmic-wildtype nematodes to compete against.
Every 3 days, the generation for each experimental line was reset. To do this, nematodes were washed off the plates using sterile M9 buffer into a sterile 1.7 mL collection tube. Approximately 500 nematodes of mixed age from each line were transferred to a fresh food plate. An additional 500 nematodes were lysed together in a single pooled lysate. Finally, 48 additional adults from each competition line were lysed individually in order to determine the fraction of heteroplasmic nematodes in each competition line at each generational time point. This experiment was continued for 10 consecutive generations.
Experimental evolution was also carried out to quantify nutrient-conditional organism-level selection. These conditions included 10 cm NGM plates seeded with a restricted or a control diet consisting of UV-killed OP50 E. coli (prepared as described below). Two iterations of this experiment were conducted, one with wildtype nuclear genotype and one with nematodes homozygous for the null daf-16(mu86) allele. Due to the smaller brood sizes among nematodes raised on a restricted diet, 200 nematodes were transferred and another 200 lysed at each generation, instead of the 500 as in the case of the experiment using a live bacterial diet. For these nutrient-conditional competition experiments, six replicate lines were propagated for each condition for a total of 8 consecutive generations. Lysis and quantification of ∆mtDNA frequency by ddPCR were performed as described above.
Diet restriction
Diet restriction was accomplished using variable dilutions of UV-inactivated OP50 E. coli bacterial lawns on NGM plates. To prepare diet-restricted food plates, 1 L of sterile 2xYT liquid microbial growth medium was inoculated with 1 mL of live OP50 E. coli (suspended in liquid LB) using a sterile serological pipette. The inoculated culture was then incubated overnight on a shaker at 37 °C. The following day, the OP50 E. coli was pelleted by centrifugation for 6 min at 3,900 rpm. The pellet was resuspended to a bacterial concentration of approximately 2 × 1010 cells/mL in sterile M9 buffer. This suspension was seeded onto NGM plates (control) or further diluted 100-fold to 2 × 108 cells/mL in sterile M9 buffer before being seeded onto NGM plates (diet restriction). Plates were incubated upright at room temperature 4 hr to allow the lawns to dry. To inhibit bacterial growth, plates were irradiated with UV radiation using a Stratagene UV Stratalinker 1800 set to 9.999 × 105 µJ/cm2. To confirm inhibition of bacterial growth, UV-treated plates were incubated overnight at 37 °C. Animals were picked at random under a dissecting microscope onto either control or diet restriction plates.
Insulin signaling inactivation
Insulin signaling was conditionally inactivated using the allele daf-2(e1370), encoding a temperature-sensitive variant of the C. elegans insulin receptor homolog. Because complete loss of insulin signaling during early larval development results in a stage of developmental arrest (dauer), age-synchronized nematodes were incubated at the permissive temperature of 16 °C until reaching the fourth and final larval stage. L4-stage larvae were then picked at random under a dissecting microscope for either transfer to the restrictive temperature of 25 °C or for continued incubation at 16 °C as a control. After 4 days of incubation, mature adults were lysed and ddPCR quantification of ∆mtDNA frequency was performed as described above. To follow up on the downstream mechanism by which insulin signaling regulates mtDNA dynamics, homoplasmic nematodes were incubated at the restrictive temperature of 25 °C and mtDNA copy number was measured using the same ddPCR primer pair that was used for quantifying the wildtype mtDNA in heteroplasmic worms.
Knockdown of gene expression
Expression knockdown of the C. elegans insulin signaling receptor homolog, daf-2, was accomplished using feeder plates. Cultures consisting of 2 mL LB and 10 µL ampicillin were inoculated with a bacterial culture obtained from Source BioScience harboring the Y55D5A_391.b (daf-2) ORF plasmid clone and incubated overnight on a shaker at 37 °C. Bacteria containing the empty plasmid vector were used to establish a control diet. The following day, 750 µL of culture was transferred to a flask containing 75 mL LB and 375 µL ampicillin and incubated 4–6 hr on a shaker at 37 °C, until OD550-600 >0.8. An additional 75 mL LB was added to the culture along with another 375 µL ampicillin and 600 µL 1 M isopropyl β-D-1-thiogalactopyranoside (IPTG) to induce expression of the small interfering RNA. Cultures were incubated another 4 hr on a shaker at 37 °C. Cultures were then centrifuged for 6 min at 3,900 rpm and the resulting bacterial pellets were each resuspended in 6 mL M9 buffer with 8 mM IPTG. Next, 250 µL of resuspension was seeded onto each NGM plate. Plates were allowed to dry at room temperature in the dark and then stored at 4 °C until use. Synchronized L4-stage nematodes were picked at random under a dissecting microscope onto either RNAi knockdown or control plates and incubated at 25 °C until day 4 of adulthood to match the conditions that were used for the daf-2 mutant allele. Day-4 adults were lysed and their mtDNA copy number was quantified using ddPCR as described above.
Live imaging
Overall mitochondrial content across the wildtype and defective insulin signaling genotypes was measured using the mitochondrial reporter TOMM-20::mCherry. Age-synchronized nematodes were incubated for 2 days from the L4 stage to mature adulthood at 25 °C, immobilized with 10 mM levamisole, and placed on the center of a 2% agarose pad on a microscope slide. Nematodes were imaged at 10x magnification using a Leica DM6000 B compound fluorescence microscope and mitochondrial fluorescence was quantified using ImageJ. Apoptosis was imaged in daf-2(e1370) mutant nematodes and wildtype controls using the CED-1::GFP reporter. Age-synchronized nematodes were incubated for 2 days from the L4 stage to mature adulthood at 25 °C before being immobilized and mounted on microscope slides as described above. Apoptotic cells were imaged using a Zeiss LSM 880 Confocal Laser Scanning microscope at 20x magnification.
Staining and imaging of germline nuclei
Nematode germline nuclei were quantified across age-synchronized mature adults homozygous for daf-2(e1370) or daf-16(mu86), as well as in double-mutants and wildtype controls. For each genotype, age-synchronized L4-stage nematodes were incubated for 2 days at 25 °C and then placed in a plate containing 3 mL of PBS with 200 µM levamisole. To dissect the nematode gonads, each adult was decapitated using two 25G × 1’ hypodermic needles in a scissor-motion under a dissecting microscope. Dissected gonads were fixed for 20 min in 3% paraformaldehyde. Fixed gonads were transferred to a glass test tube using a glass Pasteur pipette and the paraformaldehyde was replaced with PBT (PBS buffer with 0.1% Tween 20) and incubated for 15 min at room temperature. The PBT was then replaced with PBT containing 100 ng/mL 4’,6’-diamidino-2-phenylindole dihydrochloride (DAPI) and the gonads were incubated in darkness for another 15 min at room temperature. Gonads were then subjected to 3x consecutive washes, each consisting of a 1 min centrifugation at 1,000 rpm followed by replacement of the PBT. Gonads were then mounted directly onto a 2% agarose pad on the center of a microscope slide and imaged using a Zeiss LSM 880 Confocal Laser Scanning microscope at 20x magnification.
Statistical analysis
The effect of initial ∆mtDNA frequency on sub-organismal ∆mtDNA dynamics was observed in three experiments, one of which included the variables of diet and host daf-16 genotype. The effect of FoxO/DAF-16-dependent insulin signaling on mtDNA copy number was observed in five experiments using the temperature-sensitive daf-2(e1370) allele, plus once more using knockdown of daf-2 gene expression. Effect of diet restriction on sub-organismal ∆mtDNA proliferation was observed in three experiments, two of which included the variable of host daf-16 genotype. The effect of organismal selection on population-wide ∆mtDNA prevalence was observed in two separate competition experiments, one of which included the variables of diet and host daf-16 genotype. Each data point represents a biological replicate. For each experiment, sample sizes and number of animals per sample are provided in the respective figure legend. Sample sizes were chosen based on prior qualitative assessment of the impact of conditions such as diet, temperature, host age and host genotype on ∆mtDNA frequency and mtDNA copy number. For each experiment, significance was determined using the statistical test indicated in the respective figure legend, along with the indicated multiple comparisons test whenever two or more groups are compared.
Acknowledgements
We thank the members of the Patel Laboratory (James P Held, Cait S Kirby, Nikita Tsyba, Benjamin R Saunders, Cassidy A Johnson), Janet M Young, Mia T Levine, Sarah E Zanders, Harmit S Malik, and Antonis Rokas for their valuable feedback on the manuscript. Some strains were provided by the CGC, which is funded by NIH Office of Research Infrastructure Programs (P40 OD010440). This work was generously supported by R01 GM123260 (MRP), the Ruth L Kirschstein National Research Service Award Individual Predoctoral Fellowship 1F31GM125344 (BLG), and the Vanderbilt University Medical Center Diabetes Research and Training Center Pilot and Feasibility Grant. Confocal microscopy imaging was performed through the use of the Vanderbilt Cell Imaging Shared Resource (supported by NIH grants CA68485, DK20593, DK58404, DK59637 and EY08126). Quantification of mtDNA copy number and ∆mtDNA frequency was conducted with the help of the Simon A Mallal Laboratory at Vanderbilt University Medical Center.
Additional informationCompeting interests
No competing interests declared
Author contributions
Conceptualization, Data curation, Formal analysis, Funding acquisition, Validation, Investigation, Visualization, Methodology, Writing - original draft, Writing - review and editing
Formal analysis, Methodology, Writing - review and editing
Conceptualization, Resources, Data curation, Formal analysis, Supervision, Funding acquisition, Validation, Methodology, Project administration, Writing - review and editing
Additional filesData availability
All data generated or analyzed during this study are included in the manuscript and supporting files.
ReferencesAhierADaiCYTweedieABezawork-GeletaAKirmesIZurynS2018Affinity purification of cell-specific mitochondria from whole animals resolves patterns of genetic mosaicismNature Cell Biology2035236010.1038/s41556-017-0023-x29358705AktipisCABoddyAMJansenGHibnerUHochbergMEMaleyCCWilkinsonGS2015Cancer across the tree of life: cooperation and cheating in multicellularityPhilosophical Transactions of the Royal Society B: Biological Sciences3702014021910.1098/rstb.2014.021926056363AllenGFTothRJamesJGanleyIG2013Loss of iron triggers PINK1/Parkin-independent mitophagyEMBO Reports141127113510.1038/embor.2013.16824176932AngeloGVan GilstMR2009Starvation protects germline stem cells and extends reproductive longevity in C. elegansScience32695495810.1126/science.117834319713489BadiscoLVan WielendaelePVanden BroeckJ2013Eat to reproduce: a key role for the insulin signaling pathway in adult insectsFrontiers in Physiology420210.3389/fphys.2013.0020223966944BraticIHenchJHenrikssonJAntebiABürglinTRTrifunovicA2009Mitochondrial DNA level, but not active replicase, is essential for Caenorhabditis elegans developmentNucleic Acids Research371817182810.1093/nar/gkp01819181702Bravo NúñezMANuckollsNLZandersSE2018Genetic villains: killer meiotic driversTrends in Genetics3442443310.1016/j.tig.2018.02.00329499907BrockhurstMABucklingARaceyDGardnerA2008Resource supply and the evolution of public-goods cooperation in BacteriaBMC Biology62010.1186/1741-7007-6-2018479522CaoPDeyAVassalloCNWallD2015How myxobacteria cooperateJournal of Molecular Biology4273709372110.1016/j.jmb.2015.07.02226254571ChatreLRicchettiM2013Prevalent coordination of mitochondrial DNA transcription and initiation of replication with the cell cycleNucleic Acids Research413068307810.1093/nar/gkt01523345615ChenXPercM2014Excessive abundance of common resources deters social responsibilityScientific Reports4416110.1038/srep0416124561876ChisholmRLFirtelRA2004Insights into morphogenesis from a simple developmental systemNature Reviews Molecular Cell Biology553154110.1038/nrm142715232571ConnellyBDBrugerELMcKinleyPKWatersCM2017Resource abundance and the critical transition to cooperationJournal of Evolutionary Biology3075076110.1111/jeb.1303928036143DanielsenETMoellerMERewitzKF2013Nutrient signaling and developmental timing of maturationCurrent Topics in Developmental Biology105376710.1016/B978-0-12-396968-2.00002-623962838DasDArurS2017Conserved insulin signaling in the regulation of oocyte growth, development, and maturationMolecular Reproduction and Development8444445910.1002/mrd.2280628379636de Vargas RoditiLBoyleKEXavierJB2013Multilevel selection analysis of a microbial social traitMolecular Systems Biology968410.1038/msb.2013.4223959025DepuydtGXieFPetyukVASmoldersABrewerHMCampDGSmithRDBraeckmanBP2014LC-MS proteomics analysis of the insulin/IGF-1-deficient Caenorhabditis elegans daf-2(e1370) mutant reveals extensive restructuring of intermediary metabolismJournal of Proteome Research131938195610.1021/pr401081b24555535Di RitaAPeschiaroliAD AcunzoPStrobbeDHuZGruberJNygaardMLambrughiMMelinoGPapaleoEDengjelJEl AlaouiSCampanellaMDötschVRogovVVStrappazzonFCecconiF2018HUWE1 E3 ligase promotes PINK1/PARKIN-independent mitophagy by regulating AMBRA1 activation via IKKαNature Communications9375510.1038/s41467-018-05722-330217973DobataSSasakiTMoriHHasegawaEShimadaMTsujiK2009Cheater genotypes in the parthenogenetic ant Pristomyrmex punctatusProceedings. Biological Sciences27656757410.1098/rspb.2008.121518854297DobataSTsujiK2013Public goods dilemma in asexual ant societiesPNAS110160561606010.1073/pnas.130901011024046364Drummond-BarbosaDSpradlingAC2001Stem cells and their progeny respond to nutritional changes during Drosophila oogenesisDevelopmental Biology23126527810.1006/dbio.2000.013511180967DubieJJCarawayARStoutMMKatjuVBergthorssonU2020The conflict within: origin, proliferation and persistence of a spontaneously arising selfish mitochondrial genomePhilosophical Transactions of the Royal Society B: Biological Sciences3752019017410.1098/rstb.2019.017431787044DucasseHArnalAVittecoqMDaoustSPUjvariBJacquelineCTissotTEwaldPGatenbyRAKingKCBonhommeFBrodeurJRenaudFSolaryERocheBThomasF2015Cancer: an emergent property of disturbed resource-rich environments? ecology meets personalized medicineEvolutionary Applications852754010.1111/eva.1223226136819DugatkinLAPerlinMLucasJSAtlasR2005Group-beneficial traits, frequency-dependent selection and genotypic diversity: an antibiotic resistance paradigmProceedings of the Royal Society B: Biological Sciences272798310.1098/rspb.2004.291615875573FanWWaymireKGNarulaNLiPRocherCCoskunPEVannanMANarulaJMacgregorGRWallaceDC2008A mouse model of mitochondrial disease reveals germline selection against severe mtDNA mutationsScience31995896210.1126/science.114778618276892FiegnaFVelicerGJ2003Competitive fates of bacterial social parasites: persistence and self-induced extinction of Myxococcus xanthus cheatersProceedings of the Royal Society of London. Series B: Biological Sciences2701527153410.1098/rspb.2003.238712965020FisherRMRegenbergB2019Multicellular group formation in Saccharomyces cerevisiaeProceedings. Biological Sciences2862019109810.1098/rspb.2019.109831480977GaffneyCJPollardABarrattTFConstantin-TeodosiuDGreenhaffPLSzewczykNJ2018Greater loss of mitochondrial function with ageing is associated with earlier onset of Sarcopenia in C. elegansAging103382339610.18632/aging.10165430455409GhoulMGriffinASWestSA2014Toward an evolutionary definition of cheatingEvolution6831833110.1111/evo.12266GitschlagBLKirbyCSSamuelsDCGangulaRDMallalSAPatelMR2016Homeostatic responses regulate selfish mitochondrial genome dynamics in C. elegansCell Metabolism249110310.1016/j.cmet.2016.06.00827411011GoodnightCJSchwartzJMStevensL1992Contextual analysis of models of group selection, soft selection, hard selection, and the evolution of altruismThe American Naturalist14074376110.1086/285438GoodnightCJ2015Multilevel selection theory and evidence: a critique of gardner, 2015Journal of Evolutionary Biology281734174610.1111/jeb.1268526265012GottliebSRuvkunG1994daf-2, daf-16 and daf-23: genetically interacting genes controlling dauer formation in Caenorhabditis elegansGenetics1371071208056303GreerELDowlatshahiDBankoMRVillenJHoangKBlanchardDGygiSPBrunetA2007An AMPK-FOXO pathway mediates longevity induced by a novel method of dietary restriction in C. elegansCurrent Biology171646165610.1016/j.cub.2007.08.04717900900GulliJGHerronMDRatcliffWC2019Evolution of altruistic cooperation among nascent multicellular organismsEvolution731012102410.1111/evo.13727HammerschmidtKRoseCJKerrBRaineyPB2014Life cycles, fitness decoupling and the evolution of multicellularityNature515757910.1038/nature1388425373677HammondTMRehardDGXiaoHShiuPK2012Molecular dissection of Neurospora spore killer meiotic drive elementsPNAS109120931209810.1073/pnas.120326710922753473HavirdJCForsytheESWilliamsAMWerrenJHDowlingDKSloanDB2019Selfish mitonuclear conflictCurrent Biology29R496R51110.1016/j.cub.2019.03.02031163164HeislerILDamuthJ1987A method for analyzing selection in hierarchically structured populationsThe American Naturalist13058260210.1086/284732HibshmanJDDoanAEMooreBTKaplanREHungAWebsterAKBhattDPChitrakarRHirscheyMDBaughLR2017daf-16/FoxO promotes gluconeogenesis and trehalose synthesis during starvation to support survivaleLife6e3005710.7554/eLife.3005729063832HibshmanJDLeuthnerTCShobenCMelloDFSherwoodDRMeyerJNBaughLR2018Nonselective autophagy reduces mitochondrial content during starvation in Caenorhabditis elegansAmerican Journal of Physiology-Cell Physiology315C781C79210.1152/ajpcell.00109.201830133321HillJHChenZXuH2014Selective propagation of functional mitochondrial DNA during oogenesis restricts the transmission of a deleterious mitochondrial variantNature Genetics4638939210.1038/ng.292024614072HollisB2012Rapid antagonistic coevolution between strains of the social Amoeba Dictyostelium discoideumProceedings of the Royal Society B: Biological Sciences2793565357110.1098/rspb.2012.097522719037HoltIJSpeijerDKirkwoodTB2014The road to rack and ruin: selecting deleterious mitochondrial DNA variantsPhilosophical Transactions of the Royal Society B: Biological Sciences3692013045110.1098/rstb.2013.045124864317HuWJiangZDSuoFZhengJXHeWZDuLL2017A large gene family in fission yeast encodes spore killers that subvert Mendel's laweLife6e2605710.7554/eLife.2605728631610HughesSEEvasonKXiongCKornfeldK2007Genetic and pharmacological factors that influence reproductive aging in NematodesPLOS Genetics3e2510.1371/journal.pgen.003002517305431KhareASantorelliLAStrassmannJEQuellerDCKuspaAShaulskyG2009Cheater-resistance is not futileNature46198098210.1038/nature0847219794414KlotzLOSánchez-RamosCPrieto-ArroyoIUrbánekPSteinbrennerHMonsalveM2015Redox regulation of FoxO transcription factorsRedox Biology6517210.1016/j.redox.2015.06.01926184557KlucnikaAMaH2019A battle for transmission: the cooperative and selfish animal mitochondrial genomesOpen Biology918026710.1098/rsob.18026730890027KoschwanezJHFosterKRMurrayAW2013Improved use of a public good selects for the evolution of undifferentiated multicellularityeLife2e0036710.7554/eLife.0036723577233KramerJMSladeJDStaveleyBE2008Foxo is required for resistance to amino acid starvation in DrosophilaGenome5166867210.1139/G08-047LarracuenteAMPresgravesDC2012The selfish segregation distorter gene complex of Drosophila melanogasterGenetics192335310.1534/genetics.112.14139022964836LeeSDongHH2017FoxO integration of insulin signaling with glucose and lipid metabolismJournal of Endocrinology233R67R7910.1530/JOE-17-000228213398LiSIPuruggananMD2011The cooperative Amoeba: Dictyostelium as a model for social evolutionTrends in Genetics : TIG27485410.1016/j.tig.2010.11.00321167620LiauWSGonzalez-SerricchioASDeshommesCChinKLaMunyonCW2007A persistent mitochondrial deletion reduces fitness and sperm performance in Heteroplasmic populations of C. elegansBMC Genetics8810.1186/1471-2156-8-817394659LieberTJeediguntaSPPalozziJMLehmannRHurdTR2019Mitochondrial fragmentation drives selective removal of deleterious mtDNA in the germlineNature57038038410.1038/s41586-019-1213-431092924LinYFSchulzAMPellegrinoMWLuYShahamSHaynesCM2016Maintenance and propagation of a deleterious mitochondrial genome by the mitochondrial unfolded protein responseNature53341641910.1038/nature1798927135930LopezALChenJJooHJDrakeMShidateMKseibCArurS2013DAF-2 and ERK couple nutrient availability to meiotic progression during Caenorhabditis elegans oogenesisDevelopmental Cell2722724010.1016/j.devcel.2013.09.00824120884MaHXuHO'FarrellPH2014Transmission of mitochondrial mutations and action of purifying selection in Drosophila melanogasterNature Genetics4639339710.1038/ng.291924614071MartinsRLithgowGJLinkW2016Long live FOXO: unraveling the role of FOXO proteins in aging and longevityAging Cell1519620710.1111/acel.1242726643314MichaelsonDKortaDZCapuaYHubbardEJ2010Insulin signaling promotes germline proliferation in C. elegansDevelopment13767168010.1242/dev.04252320110332MichodREViossatYSolariCAHurandMNedelcuAM2006Life-history evolution and the origin of multicellularityJournal of Theoretical Biology23925727210.1016/j.jtbi.2005.08.04316288782Moreno-FenollCCavaliereMMartínez-GarcíaEPoyatosJF2017Eco-evolutionary feedbacks can rescue cooperation in microbial populationsScientific Reports74256110.1038/srep4256128211914MurphyCTMcCarrollSABargmannCIFraserAKamathRSAhringerJLiHKenyonC2003Genes that act downstream of DAF-16 to influence the lifespan of Caenorhabditis elegansNature42427728310.1038/nature0178912845331MurtazaGKhanAKRashidRMuneerSHasanSMFChenJ2017FOXO transcriptional factors and Long-Term livingOxidative Medicine and Cellular Longevity20171810.1155/2017/349428928894507NarbonnePRoyR2006Regulation of germline stem cell proliferation downstream of nutrient sensingCell Division12910.1186/1747-1028-1-2917150096NewlonCSFangmanWL1975Mitochondrial DNA synthesis in cell cycle mutants of Saccharomyces cerevisiaeCell542342810.1016/0092-8674(75)90061-61098780O'BrienSLujánAMPatersonSCantMABucklingA2017Adaptation to public goods cheats in Pseudomonas aeruginosaProceedings of the Royal Society B: Biological Sciences2842017108910.1098/rspb.2017.108928747481O-SullivanIZhangWWassermanDHLiewCWLiuJPaikJDePinhoRAStolzDBKahnCRSchwartzMWUntermanTG2015FoxO1 integrates direct and indirect effects of insulin on hepatic glucose production and glucose utilizationNature Communications6707910.1038/ncomms807925963540ÖzkayaÖXavierKBDionisioFBalbontínR2017Maintenance of microbial cooperation mediated by public goods in single- and Multiple-Trait scenariosJournal of Bacteriology199e0029710.1128/JB.00297-17PazdernikNSchedlT2013Introduction to germ cell development in Caenorhabditis elegansAdvances in Experimental Medicine and Biology75711610.1007/978-1-4614-4015-4_122872472PeredaMZurroDSantosJIBriz I GodinoIÁlvarezMCaroJGalánJM2017Emergence and evolution of cooperation under resource pressureScientific Reports74557410.1038/srep4557428362000PorteDBaskinDGSchwartzMW2005Insulin signaling in the central nervous system: a critical role in metabolic homeostasis and disease from C. elegans to humansDiabetes541264127610.2337/diabetes.54.5.126415855309PruittJNRiechertSE2009Frequency-dependent success of cheaters during foraging bouts might limit their spread within colonies of a socially polymorphic spiderEvolution632966297310.1111/j.1558-5646.2009.00771.xPuigOTjianR2006Nutrient availability and growth: regulation of insulin signaling by dFOXO/FOXO1Cell Cycle550350510.4161/cc.5.5.250116552183QuellerDCStrassmannJE2009Beyond society: the evolution of organismalityPhilosophical Transactions of the Royal Society B: Biological Sciences3643143315510.1098/rstb.2009.009519805423RaineyPBRaineyK2003Evolution of cooperation and conflict in experimental bacterial populationsNature425727410.1038/nature0190612955142RequejoRJCamachoJ2011Evolution of cooperation mediated by limiting resources: connecting resource based models and evolutionary game theoryJournal of Theoretical Biology272354110.1016/j.jtbi.2010.12.00521147130RiehlCFredericksonME2016Cheating and punishment in cooperative animal societiesPhilosophical Transactions of the Royal Society B: Biological Sciences3712015009010.1098/rstb.2015.009026729930Ross-GillespieAGardnerAWestSAGriffinAS2007Frequency dependence and cooperation: theory and a test with BacteriaThe American Naturalist17033134210.1086/51986017879185SchimentiJ2000Segregation distortion of mouse t haplotypes the molecular basis emergesTrends in Genetics : TIG1624024310.1016/s0168-9525(00)02020-510827448SenaEPWelchJWHalvorsonHOFogelS1975Nuclear and mitochondrial deoxyribonucleic acid replication during mitosis in Saccharomyces cerevisiaeJournal of Bacteriology12349750410.1128/JB.123.2.497-504.19751097413SextonDJSchusterM2017Nutrient limitation determines the fitness of cheaters in bacterial siderophore cooperationNature Communications823010.1038/s41467-017-00222-228794499ShafferZSasakiTHaneyBJanssenMPrattSCFewellJH2016The foundress's dilemma: group selection for cooperation among queens of the harvester ant, Pogonomyrmex californicusScientific Reports62982810.1038/srep2982827465430ShidaraYYamagataKKanamoriTNakanoKKwongJQManfrediGOdaHOhtaS2005Positive contribution of pathogenic mutations in the mitochondrial genome to the promotion of Cancer by prevention from apoptosisCancer Research651655166310.1158/0008-5472.CAN-04-201215753359ShimYHChunJHLeeEYPaikYK2002Role of cholesterol in germ-line development of Caenorhabditis elegansMolecular Reproduction and Development6135836610.1002/mrd.1009911835581ShiojimaIYefremashviliMLuoZKureishiYTakahashiATaoJRosenzweigAKahnCRAbelEDWalshK2002Akt signaling mediates postnatal heart growth in response to insulin and nutritional statusJournal of Biological Chemistry277376703767710.1074/jbc.M20457220012163490ShouW2015Acknowledging selection at sub-organismal levels resolves controversy on pro-cooperation mechanismseLife4e1010610.7554/eLife.1010626714105StewartJBFreyerCElsonJLWredenbergACansuZTrifunovicALarssonNG2008Strong purifying selection in transmission of mammalian mitochondrial DNAPLOS Biology6e1010.1371/journal.pbio.006001018232733StrassmannJEZhuYQuellerDC2000Altruism and social cheating in the social Amoeba Dictyostelium discoideumNature40896596710.1038/3505008711140681TakeuchiNKanekoK2019The origin of the central dogma through conflicting multilevel selectionProceedings of the Royal Society B: Biological Sciences2862019135910.1098/rspb.2019.135931575361TaylorDRZeylCCookeE2002Conflicting levels of selection in the accumulation of mitochondrial defects in Saccharomyces cerevisiaePNAS993690369410.1073/pnas.07266029911891344TepperRGAshrafJKaletskyRKleemannGMurphyCTBussemakerHJ2013PQM-1 complements DAF-16 as a key transcriptional regulator of DAF-2-mediated development and longevityCell15467669010.1016/j.cell.2013.07.00623911329TsangWYLemireBD2002Stable heteroplasmy but differential inheritance of a large mitochondrial DNA deletion in NematodesBiochemistry and Cell Biology8064565410.1139/o02-13512440704ValenciIYonaiLBar-YaacovDMishmarDBen-ZviA2015Parkin modulates heteroplasmy of truncated mtDNA in Caenorhabditis elegansMitochondrion20647010.1016/j.mito.2014.11.00125462019VanthournoutBGreveMBruunABechsgaardJOvergaardJBildeT2016Benefits of group living include increased feeding efficiency and lower mass loss during desiccation in the social and inbreeding spider Stegodyphus dumicolaFrontiers in Physiology71810.3389/fphys.2016.0001826869936VelicerGJKroosLLenskiRE1998Loss of social behaviors by Myxococcus xanthus during evolution in an unstructured habitatPNAS95123761238010.1073/pnas.95.21.123769770494WangZHLiuYChaitankarVPiroozniaMXuH2019Electron transport chain biogenesis activated by a JNK-insulin-Myc relay primes mitochondrial inheritance in DrosophilaeLife8e4930910.7554/eLife.4930931612862WebbAEKundajeABrunetA2016Characterization of the direct targets of FOXO transcription factors throughout evolutionAging Cell1567368510.1111/acel.1247927061590WebbAEBrunetA2014FOXO transcription factors: key regulators of cellular quality controlTrends in Biochemical Sciences3915916910.1016/j.tibs.2014.02.00324630600WenseleersTRatnieksFL2004Tragedy of the commons in Melipona beesProceedings. Biological Sciences271 Suppl 5S310S31210.1098/rsbl.2003.015915504003WestSAFisherRMGardnerAKiersET2015Major evolutionary transitions in individualityPNAS112101121011910.1073/pnas.142140211225964342WilsonDSWilsonEO2007Rethinking the theoretical foundation of sociobiologyThe Quarterly Review of Biology8232734810.1086/52280918217526WinMTYamamotoYMunesueSHanDHaradaSYamamotoH2013Validated liquid culture monitoring system for lifespan extension of Caenorhabditis elegans through Genetic and Dietary ManipulationsAging and Disease417818523936742ZhangYWangZHLiuYChenYSunNGucekMZhangFXuH2019PINK1 inhibits local protein synthesis to limit transmission of deleterious mitochondrial DNA mutationsMolecular Cell731127113710.1016/j.molcel.2019.01.0133077217510.7554/eLife.56686.sa1Decision letterShouWenyingReviewing EditorFred Hutchinson Cancer Research CenterUnited StatesShouWenyingReviewerFred Hutchinson Cancer Research CenterUnited StatesMaHansongReviewerUniversity of California, San FranciscoUnited StatesHaagEricReviewerUniversity of MarylandUnited States
In the interests of transparency, eLife publishes the most substantive revision requests and the accompanying author responses.
Acceptance summary:
Using a heteroplasmic strain of C. elegans, authors examined the dynamics of "cheating" mitochondrial DNA within and across a generation, and how this dynamics was affected by nutrient status. This work is elegant and quantitative, and makes a timely contribution to multi-level selection (within-host selection and between-host selection) literature.
Decision letter after peer review:
Thank you for submitting your article "Selfish mitochondria exploit nutrient status to proliferate across the different levels of selection" for consideration by eLife. Your article has been reviewed by four peer reviewers, including Wenying Shou as the Reviewing Editor and Reviewer #1, and the evaluation has been overseen by George Perry as the Senior Editor. The following individuals involved in review of your submission have agreed to reveal their identity: M Florencia Camus (#2); Hansong Ma (Reviewer #3); Eric Haag (Reviewer #4).
The reviewers have discussed the reviews with one another and the Reviewing Editor has drafted this decision to help you prepare a revised submission.
This paper needs a serious revision in writing. Overall, we feel that although interesting, you oversimplified your work so much that it ended up being very difficult to follow. Temperature and nutrient are two variables in your experiments, yet you chose to focus on the latter without explaining why (even though the former clearly had an effect). The frequency-dependent advantage of cheating mtDNA over wild type mtDNA needs to be clearly stated and emphasized to reduce confusion. See more detailed comments below.
Given the interest in the science but critical need for major changes to the presentation, the full reviewer comments are provided below for you to consider as you rework the paper.
Summary:
Glitschlag et al. examine the fate of cheating mitochondria in C. elegans as a function of environmental (nutrient and temperature) status. Mitochondria DNA (mtDNA) with a large deletion (cheating mtDNA) has a fitness advantage over normal cooperative mtDNA during organismal development, although the frequency of cheating mtDNA drops somewhat from parent to embryo. Thus, cheating mtDNA has a within-host advantage over cooperating mtDNA, especially when cheating mtDNA is initially rare. On the other hand, host with high fraction of cheating mtDNA suffers fitness cost. Thus, cheating mtDNA has a between-host disadvantage compared to cooperating mtDNA. During germline development (when mtDNA undergoes most proliferation), if nutrients are in excess and temperature is upshifted, insulin signalling through DAF-2 inhibits DAF-16, which allows mtDNA to proliferate, which in turn allows cheating mtDNA to takeover. In this case, DAF-16 inhibits mtDNA proliferation (which deters takeover by cheating mtDNA). In normal temperature, DAF-16 promotes cheating mtDNA within host in both poor and rich diets. Thus, the role of DAF-16 is complex.
Reviewer #1:
The work is interesting, and techniques for quantifying within- and between-host fitness difference between cheating and cooperating mtDNA are nice. The paper does not read smoothly especially at the DAF-16/diet/temperature part. The part of the Price equation is not needed, and seems incorrect/incomplete anyways. Feel free to save the Price equation part which, after a more rigorous analysis, can stand on its own as an independent article.
1) The Abstract does not give a clear picture of what the story is. Contents of supplementary figures are rarely given in the main text. Figure legends are also very brief.
2) Figure 1: When cheater mtDNA gets to 80%, it can no longer increase in frequency. Why is that? For the general audience, it is also useful to add a bit more detail on why cheating mtDNA declines from parent to embryo.
3) Figure 2C: I am not sure what this means. Does this (decreasing or similar %cheating mtDNA over development) contradict your Figure 1 (increasing %cheating mtDNA over development)?
4) Figure 2E: I find the figure hard to digest. I would like to see three panels plotted in the same figure for comparison: (i) within host advantage of cheating mtDNA (Figure 1D); (ii) between host disadvantage of cheating mtDNA (Figure 2—figure supplement 1B); and (iii) population level mtDNA (Figure 2E). According to Price Equation, (iii) should be a function of (i) and (ii) and variations in cheating mtDNA among offspring embryos.
5) Price equation: In the Price equation, there are two terms (as suggested in Figure 3—figure supplement 1, although note the typo: () should be ()). The first term of covariance describes how the host fecundity covaries with cooperative mtDNA in the host (and is a function of variations in cheating mtDNA among offspring embryos), and the second term of expectation describes how the cooperative mtDNA is lost to cheating mtDNA within host. Either you have misunderstood Price equation, or your writing is very confusing, especially when you sum up the sub-organismal and organismal covariance terms.6) Subsection “Insulin signaling influences sub-organismal proliferation of ∆mtDNA”, last paragraph: DAF-16 promotes cheating mtDNA in both rich and poor diet, and not just rich diet, correct? (Figure 6B). In general, the two environmental variables (temperature and diet) plus the genotype variables makes data interpretation and writing (and reading) tricky. Your interpretation has mainly focused on diet despite the apparent importance of temperature (Figure 4F). For example, what happens during restrictive diet and temperature shift? I am not necessarily asking you to do more experiments, but careful writing and illustration is critical when biology gets complex.
7) Figure 7: This should ideally be the summary figure (data can go to the supplement?). You can have schematics for different environments and genotypes. In each schematic, you can include numbers such as selection coefficient of cheating mtDNA from parent to embryo (and variations among embryos), from embryo to Adult (within host advantage, which depends on initial frequency of cheating mtDNA), and at the organismal level (between-host disadvantage). Again, I am not asking you to do more experiments to fill all entries, but putting everything together makes it easier to spot patterns/discrepancies.
Reviewer #2:
Authors examine mitochondrial proliferation dynamics in a very nice C. elegans model system. Using a heteroplasmic strain of worm, they investigate the proportion of a deleterious "ΔmtDNA" passed onto the next generation. This ΔmtDNA lacks several core mitochondrial genes as well as tRNAs, and authors demonstrate that is impacts several physiological traits. Interestingly, in developing worms they see a decrease in the frequency of the deleterious haplotype, however as the worm matures this percentage is increased. When in competition, however, the detrimental effects of the ΔmtDNA are clear, with a decrease in frequency over generations. Authors then go on to examine this dynamic across two different environments; different nutritional environments and temperatures. They find that two insulin-signalling genes (DAF-2, DAF-16) are involved in mtDNA proliferation, but this is contingent on the temperature that worms were at.
While I really enjoyed this paper, authors could tone down their language a bit. Furthermore, bits of the paper could be re-written to give a clearer picture to the reader. This is a paper with a lot of technical experiments, and it would be good for the authors describe experimental setup, and statistical inferences in the Results in a clearer way. Personally, I think it is a very timely contribution to the scientific literature, and is not only important to the mitochondrial field, but across the wider biological sciences.
Comment about statistics:
1) Are the statistical tests performed in relative or raw datapoints? If they are performed on the relative data, can authors please justify this? I find that by standardising data, differences appear to be bigger and less biologically meaningful.
2) Along these lines, it would be good for the authors to tell the reader in the Results sections about these relative values (how they were calculated, and what they mean). There is a little bit in the figure legends, but this is of great importance for biological conclusions.
Other comments:
1) Subsection “An experimental strategy to isolate selection on a selfish mitochondrial genome at different levels”, second paragraph: I think you mean Figure 1C and D in this sentence instead of Figure 1D and E?
2) Figure 1C – It would be very helpful to write on each panel "low", "medium" or "high". I wrote it on my printed copy and it made the figure easier to interpret.
3) Figure 1D – I understand what you are trying to get to with this figure, but maybe a better way to present it would be using simple boxplots, or mean ± SE (within each of their respected categories). To me it looks a bit messy, and it took me a bit to get my head around it. I'm mainly saying this because this figure conveys a very important message, and the reader would greatly benefit from a clear figure.
4) Figure 1E – Is the "shift" = mature adult offspring% – parental% ? If so, I would write that instead of "parent to adult progeny".
5) Figure 2A – The O2 consumption measurements relative to the average wildtype basal rate, right? Then say so in the plot.
6) Figure 2B – Why did the authors use very different sample sizes for the three treatments? N=8, 12, and 35 seems a bit varied.
7) “These data show that nutrient sensing via the insulin-signaling pathway regulates sub-organismal proliferation of ∆mtDNA”- A bit of a bold claim. How about, "These data show that nutrient sensing via the insulin-signalling pathway is involved in the regulation of sub-organismal proliferation of ΔmtDNA”?
8) Figure 4F – If higher ΔmtDNA entirely accounts for the higher total copy number, then why not plot the copy number of ΔmtDNA and have this plot as a supplementary?
9) Subsection “Insulin signaling influences sub-organismal proliferation of ∆mtDNA”, third paragraph: Again, rather than "these data show", "these data suggests"… I say this because deletion of daf-16 partially rescues phenotype.
10) Subsection “Insulin signaling influences sub-organismal proliferation of ∆mtDNA”, fourth paragraph: These few lines need more explanation, especially if you are showing new mutants in the figures. Please expand.
11) Figure 5E: When the y-axis label says "Relative mitochondrial content", what is it relative to?
12) Discussion: I find that the Discussion goes from an overall "big picture" paragraph straight to nutrient-sensing. Can authors discuss Figures 1-3?
Reviewer #3:
In this manuscript, Gitschlag et al. used a heteroplasmic C. elegans line carrying both wildtype and Df5-deletion mitochondrial genomes to reveal how nutrition status impacts selection on selfish mtDNA at different levels. The authors first characterized and quantified selection on the Df5 mtDNA at sub-organismal and organismal levels. They then limited nutrition by providing worms with control and restricted diets and explored how that affects the Df5 propagation at the sub-organismal level. They showed that the Df5 frequency increase from embryos to adults observed in heteroplasmic worms growing in the normal diet is abolished in worms raised on the restricted diet or worms raised in the normal diet but defective for daf-2, the insulin receptor. Deletion of daf-16 restores the Df5 increase and proliferation during adult maturation in daf-2 mutants by restoring the germline development. Interestingly, DAF-16 was shown not to be required for copy-number suppression in response to diet restriction but only required for the Df5 frequency increase when nutrition is abundant. Finally, they characterized the role of DAF-16 on selection at the organismal level and showed that diet restriction accelerates the decline of Df5 in daf-16 mutants, but not in wildtype.
Overall, the manuscript is well written, and the findings are interesting. Below, I included some of my concerns and questions:
1) The sub-organismal selection outcome is a synergistic effect of purifying selection against the Df5 at the mitochondrial and cell level, and positive selection for the Df5 due to its replicative advantage. Upon the diet restriction, the author focused on the disappearance of Df5 increase during adult maturation (Figure 4B and C). However, there is another big change – the disappearance of Df5 decrease from parents to embryos, which could be due to reduced purifying selection during germline development or increased selfish gain of Df5 at that stage. This suggests that diet restriction can affect the propagation of Df5 differently at different developmental stages. I think that it is worth including a discussion on this in the manuscript.
2) I wonder the generality of the nutrition restriction impact on the transmission of selfish mitochondrial genomes described in the Df5 heteroplasmic line. Is it possible to perform one of the key experiments (e.g. Figure 4C) on a different heteroplasmic line that carries a different type of deletion/selfish mtDNA (e.g. ND5 deletions in natural populations of C.briggsae)?
3) Figure 7E and Figure 7—figure supplement 1A showed that non-competing wildtype heteroplasmic flies show a reduced Df5 genome over generations in control and restricted diet. However, in Figure 2E and Figure 2—figure supplement 1A, such a decline was not observed in the normal diet with wildtype heteroplasmic flies. This indicates that both control and restricted diets affect the cross-generational transmission of Df5 compared to the normal diet. Please comment on this. It will also be nice if the curve for wildtype flies growing in the normal diet can be included in the same figure for easy comparison.
4) Subsection “Nutrient status governs selection on ∆mtDNA at different levels”, first paragraph: the authors showed that diet restriction accelerated the decline of Df5 frequency at the level of organismal selection among daf-16 mutants. Is this acceleration due to much-reduced fertility (e.g. caused by reduced sperm viability or oocyte production) when heteroplasmy is combined with daf-16 deficiency? For Figure 7C and D, I think it is more important to show how the fraction containing the Df5 decreases over generations rather than population/individual or population-wide percentage of Df5, because the latter two also contain the sub-organismal selection effect, especially for the daf-16 mutant.
5) I found the presentation of the data not easy to follow. For example, the population/individual calculation was not properly explained in Figure 2E or its legend. I suggest that either the Figure 2—figure supplement 1A and B to be put in Figure 2E, or the figure legend explains what exactly population-wide Df5 frequency relative to average individual heteroplasmic frequency means. Alternatively, the authors can present only Figure 2—figure supplement 1B in the main figure, which shows how the fraction of heteroplasmic flies decreases over generations. After all, we are looking at competition at the organismal level. Figure 1, Figure 2 and Figure 4E-G legends should contain information on growth conditions including food (e.g. live OP50) and temperature (20 degrees) although such information has been described in the Materials and methods and supplementary information. The presence of the above information in the legend will help readers understand experimental setups without referring to other parts of the paper and make comparisons between data from different figures much easier. The authors also assumed readers to have plenty of pre-knowledge of the Df5 heteroplasmic line. It will be good to have additional descriptions of the Df5 heteroplasmic line in the Introduction.
Reviewer #4:
This manuscript reveals a number of interesting and surprising phenomena that influence the relative and absolute abundance of a deletion-bearing mitochondrial genome (ΔmtDNA). It also ties these into a multi-level selection model in an effort to define how organismal and sub-organismal fitnesses are linked. There is much to commend here, and the writing is first-rate. However, I do see a few weakness, which are noted below.
1) The paper's title and overall text focus on the impact of nutritional status, and the signaling pathway that relays this to the cells and tissues, on the replication of the ΔmtDNA genome. However, the huge effect of temperature on even wild-type worms is not given much billing nor explored, beyond serving as an assay to examine the impact of the insulin pathway and nutrition. Is the temperature effect related to the germ line? A re-motivation of the experiments that begins by problematizing the temperature effect would produce a more straightforward paper. For example, the failure to do this makes the following sentence very hard to parse “Together, these data show that DAF-2 signaling inhibits DAF-16 to allow high mtDNA copy number, which permits sub-organismal ∆mtDNA proliferation at the warmer temperature”.
2) Like abundant food, we see that high temp. also promotes ∆mtDNA proliferation, though this requires DAF-2 function. Thus, high temp. acts like abundant food. This is unusual, given that 25°C is somewhat stressful and dauer-promoting. I'm having a hard time reconciling these two results.
3) "We therefore conclude that suppression of germline development by DAF-16 accounts for the reduced mtDNA content in insulin-signaling mutants." We thus have a bit of paradox: growth 25°C doesn't stimulate germline proliferation, yet it does increase ∆mtDNA. So, there may be two independent effects here?
4) The two pairwise comparisons in Figure 6B look qualitatively similar to me. Seems dangerous to build much of an argument around this difference.
10.7554/eLife.56686.sa2Author response
Reviewer #1:
The work is interesting, and techniques for quantifying within- and between-host fitness difference between cheating and cooperating mtDNA are nice. The paper does not read smoothly especially at the DAF-16/diet/temperature part. The part of the Price equation is not needed, and seems incorrect/incomplete anyways. Feel free to save the Price equation part which, after a more rigorous analysis, can stand on its own as an independent article.
We appreciate the reviewer’s interest in the work and the techniques, and we do intend to follow this work up by applying these techniques to additional heteroplasmies. As detailed below, we have made extensive revisions to address the concern regarding the writing surrounding DAF-16/diet/temperature results. To address the concerns regarding the Price Equation, we have eliminated this framework from our treatment of multilevel selection, focusing instead on showing how ∆mtDNA frequency itself changes due to selection at each level (within and between host organisms). As per reviewer’s suggestion, we plan to expand on the Price equation analysis and publish it as a separate article, potentially as a Research Advance in eLife.
1) The Abstract does not give a clear picture of what the story is. Contents of supplementary figures are rarely given in the main text. Figure legends are also very brief.
We have revised the Abstract for greater clarity and have revised the main text to reference the supplementary information more thoroughly. In addition, the figure legends have been expanded to provide additional relevant information such as diet, temperatures, and ages for all experiments.
2) Figure 1: When cheater mtDNA gets to 80%, it can no longer increase in frequency. Why is that? For the general audience, it is also useful to add a bit more detail on why cheating mtDNA declines from parent to embryo.
We appreciate the reviewer’s interest in these observations regarding the behavior of ∆mtDNA. Although we were initially concerned that addressing these might overly complicate an already detailed study, we agree that these are interesting observations and merit some attention. To this effect, we have addressed these topics by making the following revisions:
1) In the Results section (subsection “An experimental strategy to isolate selection on a selfish mitochondrial genome at different levels”, second paragraph), we now explicitly point out and elaborate on the observation regarding the apparent upper limit of sub-organismal ∆mtDNA proliferation. Additionally, we have expanded the Discussion section (to address this phenomenon of negative frequency-dependent selection (Discussion, third paragraph), whereby the proliferation of ∆mtDNA is attenuated as its sub-organismal frequency increases).
2) We have also expanded the Discussion section (second paragraph), devoting a paragraph to address the phenomenon of purifying selection from parent to embryo.
3) Figure 2C: I am not sure what this means. Does this (decreasing or similar %cheating mtDNA over development) contradict your Figure 1 (increasing %cheating mtDNA over development)?
We appreciate that the way we presented the data in Figure 2C may have caused confusion. We have clarified the presentation of these data in the revised manuscript. In particular, although the frequency of ∆mtDNA increases across larval development, as the reviewer notes, the presence of ∆mtDNA also impedes the rate at which larvae develop. The data in original Figure 2C (now Figure 1F) shows the larval development stage that has been reached, starting from a cohort of age-synchronized embryos. To clarify that the rate of development is negatively associated with ∆mtDNA frequency, we have reversed the axes: the developmental stage reached (vertical axis) is now plotted as a function of ∆mtDNA frequency (horizontal axis). We also reworded the vertical axis label to reflect that the dependent variable is “developmental stage reached,” and we note in the figure legend that these varying-stage larvae began as synchronized embryos, to emphasize that these data reflect an organismal fitness consequence of ∆mtDNA, rather than a sub-organismal dynamic of ∆mtDNA. Accordingly, although ∆mtDNA frequency rises within individuals across development, this rise is not sufficient to conceal the relationship between mutant frequency and the developmental stage that larvae are able to reach within 48 hours.
4) Figure 2E: I find the figure hard to digest. I would like to see three panels plotted in the same figure for comparison: (i) within host advantage of cheating mtDNA (Figure 1D); (ii) between host disadvantage of cheating mtDNA (Figure 2—figure supplement 1B); and (iii) population level mtDNA (Figure 2E). According to Price Equation, (iii) should be a function of (i) and (ii) and variations in cheating mtDNA among offspring embryos.
We appreciate that a main focal point of this paper is the way in which selection acts on ∆mtDNA at two levels, within and between hosts. Accordingly, we have made the following revisions:
1) We combined the ∆mtDNA frequency distribution data (Figure 1C in the original manuscript) and the organismal phenotype data (Figure 2A-C in the original manuscript) into the revised Figure 1. In the revised manuscript, Figure 1 is geared toward establishing ∆mtDNA as a genetic element that undergoes selection both within and between host organisms.
2) We revised Figure 2 to show the quantitative effects of selection at the sub-organismal and organismal levels on ∆mtDNA. To this effect, Figure 2 now contains the intergenerational shift in ∆mtDNA frequency due to sub-organismal heteroplasmy dynamics (panels A and B), a schematic of the organismal selection experiment (panel C), the fraction of ∆mtDNA-containing heteroplasmic hosts per generation (panel D), and the population-wide measurement of ∆mtDNA frequency during the organismal competition experiment (panel E), in accordance with how the reviewer suggested the data be presented.
We thank the reviewer for their suggestions as we think that the new organization of Figures 1 and 2 makes the story much more clear and easier to understand.
5) Price equation: In the Price equation, there are two terms (as suggested in Figure 3—figure supplement 1, although note the typo: () should be ()). The first term of covariance describes how the host fecundity covaries with cooperative mtDNA in the host (and is a function of variations in cheating mtDNA among offspring embryos), and the second term of expectation describes how the cooperative mtDNA is lost to cheating mtDNA within host. Either you have misunderstood Price equation, or your writing is very confusing, especially when you sum up the sub-organismal and organismal covariance terms.
We acknowledge the unfortunate typographical error in the original version of Figure 3—figure supplement 1; the equation should have included the covariance term E[cov(wi,∆zi)] instead of (), corresponding to the version of the Price Equation shown in equation A17 in Price, 1972. Moreover, for the sake of readability, and with the understanding that this work may be of interest to readers outside of evolutionary biology, we have taken the advice of the reviewers and have eliminated the Price Equation framework altogether from our discussion of multilevel selection (and hope to publish it as a separate follow-up article). We therefore have also eliminated the original Figure 3 and its corresponding figure supplements, and renumbered the subsequent figures.
6) Subsection “Insulin signaling influences sub-organismal proliferation of ∆mtDNA”, last paragraph: DAF-16 promotes cheating mtDNA in both rich and poor diet, and not just rich diet, correct? (Figure 6B). In general, the two environmental variables (temperature and diet) plus the genotype variables makes data interpretation and writing (and reading) tricky. Your interpretation has mainly focused on diet despite the apparent importance of temperature (Figure 4F). For example, what happens during restrictive diet and temperature shift? I am not necessarily asking you to do more experiments, but careful writing and illustration is critical when biology gets complex.
Because the effect of warmer temperature on ∆mtDNA proliferation is abolished by the loss of insulin signaling in temperature-sensitive daf-2 mutants, the variable of temperature provided an opportunity to investigate the role of insulin signaling on the heteroplasmy dynamics of ∆mtDNA. However, we appreciate the suggestion that the effect of temperature itself introduces an additional confounding influence on ∆mtDNA dynamics and warrants discussion. To this effect, we have revised our presentation of the data to address the reviewer’s concerns by making the following changes:
1) The data collected from nematodes incubated at 25°C (the restrictive temperature for the daf-2(e1370) allele) is shown in the revised Figure 3 by itself, to compare ∆mtDNA frequency and mtDNA copy number across host nuclear backgrounds under the conditions which inactivate daf-2-mediated insulin signaling.
2) The data collected from nematodes incubated at 16°C (the permissive temperature for the daf-2(e1370) allele) was moved to Figure 3—figure supplement 1.
3) The Results section has been revised (subsection “∆mtDNA exploits nutrient sensing to proliferate across development”, first paragraph) to emphasize the host genotype effect on ∆mtDNA frequency, while still acknowledging that the difference across nuclear backgrounds requires incubation at the restrictive temperature for daf-2(e1370).
4) The putative influence of temperature is now addressed in the fourth paragraph of the revised Discussion section, which is entirely devoted to this phenomenon.
7) Figure 7: This should ideally be the summary figure (data can go to the supplement). You can have schematics for different environments and genotypes. In each schematic, you can include numbers such as selection coefficient of cheating mtDNA from parent to embryo (and variations among embryos), from embryo to Adult (within host advantage, which depends on initial frequency of cheating mtDNA), and at the organismal level (between-host disadvantage). Again, I am not asking you to do more experiments to fill all entries, but putting everything together makes it easier to spot patterns/discrepancies.
We appreciate the reviewer suggestion to use a summary figure to bring greater clarity and understanding of the key results. Having eliminated Figure 3 from the original manuscript, we were able to devote Figure 6 of the revised manuscript to show the effect of diet restriction and daf-16 genotype on multilevel selection, and use Figure 7 to provide the summary schematic. Additionally, in light of the fact that the role of DAF-16 is complex, we now include schematics in the revised Figures 4-6, which highlight each observed effect of daf-16 genotype on mtDNA dynamics. To make the overall summary conclusions more apparent and to allow the reader to put everything together, the revised Figure 7 schematic is actually a composite of the schematics from the revised Figures 4-6. Finally, since we no longer discuss multilevel selection using the framework of the Price equation and covariance relationships, we now address selection on ∆mtDNA in terms of changes in mutant frequency itself, rather than converting to selection coefficients.
Reviewer #2:
[…]
Comment about statistics:
1) Are the statistical tests performed in relative or raw datapoints? If they are performed on the relative data, can authors please justify this? I find that by standardising data, differences appear to be bigger and less biologically meaningful.
Several figure panels show data that was normalized to a control condition and/or control genotype (for example the copy number data in the revised Figure 3G, H and Figure 4, and the imaging data in Figure 4). To address the reviewer concern, we performed statistical tests on the raw data, which did not affect the results of the statistical tests. For experiments involving ∆mtDNA frequency, the statistical tests were performed directly on ∆mtDNA frequency, expressed as a percentage of total mtDNA, for each sample.
2) Along these lines, it would be good for the authors to tell the reader in the Results sections about these relative values (how they were calculated, and what they mean). There is a little bit in the figure legends, but this is of great importance for biological conclusions.
We agree with this very important suggestion and have revised the Results section accordingly. This suggestion to explain how we normalize the data is especially important for the data presented in revised Figure 2E. The last paragraph of the subsection “An experimental strategy to isolate selection on a selfish mitochondrial genome at different levels” isdevoted to providing this explanation.
Other comments:
1) Subsection “An experimental strategy to isolate selection on a selfish mitochondrial genome at different levels”, second paragraph: I think you mean Figure 1C and D in this sentence instead of Figure 1D and E?
We have separated the data shown in Figure 1C and D into two separate figures, with the Figure 1D data having been moved to Figure 2 for the comparison with organismal selection, as per the recommendation of reviewer 1. We have also revised the main text to make sure the figures are appropriately cited.
2) Figure 1C – It would be very helpful to write on each panel "low", "medium" or "high". I wrote it on my printed copy and it made the figure easier to interpret.
Figure 1C has been revised as per the reviewer’s suggestion.
3) Figure 1D – I understand what you are trying to get to with this figure, but maybe a better way to present it would be using simple boxplots, or mean ± SE (within each of their respected categories). To me it looks a bit messy, and it took me a bit to get my head around it. I'm mainly saying this because this figure conveys a very important message, and the reader would greatly benefit from a clear figure.
As the reviewer suggests, we have replaced the original version of this graph with one that features box-and-whisker plots. But due to this being longitudinal data that features multiple isolated parent-progeny lineages, we also thought it would be appropriate to retain the trajectories in ∆mtDNA frequency, albeit in a more transparent and less distracting format. We reasoned that this is a way to present the data in a less confusing way without sacrificing the longitudinal nature of the data, and is consistent with the visual presentation of data in Figure 1C, D of Wohl et al., eLife, 2020 (https://elifesciences.org/articles/52702).
4) Figure 1E – Is the "shift" = mature adult offspring% – parental% ? If so, I would write that instead of "parent to adult progeny".
To address the reviewer concern, this figure (Figure 2B in the revised manuscript) has been revised so that the vertical axis label reads: “Generational shift in % ∆mtDNA (adult progeny – parental).”
5) Figure 2A – The O2 consumption measurements relative to the average wildtype basal rate, right? Then say so in the plot.
The axis in this figure (Figure 1D in the revised manuscript) now reads: “O2 consumption rate relative to wildtype basal rate.”
6) Figure 2B – Why did the authors use very different sample sizes for the three treatments? N=8, 12, and 35 seems a bit varied.
The disparity in sample sizes was because we initially collected dozens of heteroplasmic parent lysates for the purpose of correlating ∆mtDNA frequency with fecundity using a linear regression. However, the regression was not significant. We then collapsed the heteroplasmic data points into only two bins (above and below the population mean of 60%) in order to compare their corresponding fertility rates to that of a wildtype control using a Brown-Forsythe and Welch ANOVA. Although we did not find ∆mtDNA frequency to be linearly related to fertility rate, we nevertheless observed the presence of ∆mtDNA to impact fertility compared to wildtype controls. For the sake of consistency and to avoid confusion, we addressed the reviewer concern regarding the very different sample sizes. In particular, we shrank the sample size of the “60-80% ∆mtDNA” bin to N=12 by randomly assigning 12 samples for retaining and 23 samples for discarding. To confirm that this did not affect our statistical analysis, we repeated this three times and performed the statistical analysis on each data set containing a randomly-selected sample of N=12 for the high-∆mtDNA-frequency bin. In each case, the presence of ∆mtDNA at 60-80% frequency continued to correspond to a significantly lower fertility rate than wildtype controls. This information is now provided in the “Fertility” sub-section of Materials and methods.
7) “These data show that nutrient sensing via the insulin-signaling pathway regulates sub-organismal proliferation of ∆mtDNA”- A bit of a bold claim. How about, "These data show that nutrient sensing via the insulin-signalling pathway is involved in the regulation of sub-organismal proliferation of ΔmtDNA.
That statement has been revised in accordance with the reviewer’s suggestion (subsection “∆mtDNA exploits nutrient sensing to proliferate across development”, first paragraph).
8) Figure 4F – If higher ΔmtDNA entirely accounts for the higher total copy number, then why not plot the copy number of ΔmtDNA and have this plot as a supplementary?
The original Figure 4F (Figure 3F in the revised manuscript) has been updated in the following way:
1) The height of the purple data points reflects ∆mtDNA copy number rather than total (wildtype + mutant) copy number, as per the reviewer’s recommendation.
2) The 16°C temperature controls have been moved to the supplement, so that the main figure highlights the comparison of wildtype and mutant copy number across different nuclear backgrounds.
9) Subsection “Insulin signaling influences sub-organismal proliferation of ∆mtDNA”, third paragraph: Again, rather than "these data show", "these data suggests"… I say this because deletion of daf-16 partially rescues phenotype.
That statement has been revised in accordance with the reviewer’s suggestion (subsection “∆mtDNA exploits nutrient sensing to proliferate across development”, third paragraph).
10) Subsection “Insulin signaling influences sub-organismal proliferation of ∆mtDNA”, fourth paragraph: These few lines need more explanation, especially if you are showing new mutants in the figures. Please expand.
We have expanded this part of the Results section to provide a more detailed rationale for the putative roles of mitophagy, fission, and cell death in germline mtDNA content (subsection “∆mtDNA exploits nutrient sensing to proliferate across development”, fourth paragraph).
11) Figure 5E: When the y-axis label says "Relative mitochondrial content", what is it relative to?
These data reflect the quantification of mitochondrial content according to the TOMM-20::mCherry signal; to obtain the relative values, the raw signal across all genotypes was normalized to that of the wildtype nuclear genome. The axis label has been updated to clarify these details.
12) Discussion: I find that the discussion goes from an overall "big picture" paragraph straight to nutrient-sensing. Can authors discuss Figures 1-3?
We appreciate the reviewer pointing out the need to better explain the first third of the figures. The second through fifth paragraphs of the revised Discussion section have been added to more thoroughly address the multilevel selection experiments. In particular, we address the phenomenon of purifying selection from parent to embryo, the negative frequency-dependent sub-organismal selection (whereby the magnitude of ∆mtDNA proliferation was negatively associated with initial frequency), the effect of temperature on sub-organismal ∆mtDNA dynamics, and role of the balancing selection in maintaining ∆mtDNA frequency in a heteroplasmic population.
Reviewer #3:
[...]
1) The sub-organismal selection outcome is a synergistic effect of purifying selection against the Df5 at the mitochondrial and cell level, and positive selection for the Df5 due to its replicative advantage. Upon the diet restriction, the author focused on the disappearance of Df5 increase during adult maturation (Figure 4B and C). However, there is another big change – the disappearance of Df5 decrease from parents to embryos, which could be due to reduced purifying selection during germline development or increased selfish gain of Df5 at that stage. This suggests that diet restriction can affect the propagation of Df5 differently at different developmental stages. I think that it is worth including a discussion on this in the manuscript.
We have revised the manuscript to address the reviewer’s comment. In particular, the Results section now reports the effect of diet on purifying selection between parent and embryo (subsection “Nutrient availability influences sub-organismal ∆mtDNA dynamics”). In addition, the second paragraph of the revised Discussion section has been added to more thoroughly discuss the relationship between maternal diet and the observed purifying selection against ∆mtDNA between parent and embryo.
2) I wonder the generality of the nutrition restriction impact on the transmission of selfish mitochondrial genomes described in the Df5 heteroplasmic line. Is it possible to perform one of the key experiments (e.g. Figure 4C) on a different heteroplasmic line that carries a different type of deletion/selfish mtDNA (e.g. ND5 deletions in natural populations of C. briggsae)?
The reviewer raises an excellent point regarding the generality of the observations reported in the present study. We chose to focus on uaDf5 because it is an exceptionally well-characterized selfish heteroplasmic genome with respect to its organismal fitness effects and sub-organismal dynamics. In previous work, we and the Cole Haynes Lab have shown that this mutant genome proliferates, at least in part, by evading host stress-response mechanisms such as mtDNA copy-number control and mitochondrial autophagy, as well as by preferentially benefiting from mtDNA biogenesis. These processes are implicated in heteroplasmy dynamics in other systems, for example in the works of Durham et al. AJHG 2007, Suen et al. PNAS 2010, Kandul et al. Nature Communications 2016, and Chiang et al. Current Biology 2019. Although we therefore expect that the findings of the present study might be generalizable to many other heteroplasmies, we share the reviewer’s sentiment that the generality of these findings should be investigated using other heteroplasmies. To this end, we are equipped with a library of heteroplasmic nematode lines carrying mutations that affect different components of the electron transport chain, which will enable us to identify general principles between the nature of respiratory dysfunction and mutant mitochondrial genome dynamics. Given the scope of this research project, we plan to publish findings from this project in follow-up manuscript, possibly as a Research Advance in eLife.
3) Figure 7E and Figure 7—figure supplement 1A showed that non-competing wildtype heteroplasmic flies show a reduced Df5 genome over generations in control and restricted diet. However, in Figure 2E and Figure 2—figure supplement 1A, such a decline was not observed in the normal diet with wildtype heteroplasmic flies. This indicates that both control and restricted diets affect the cross-generational transmission of Df5 compared to the normal diet. Please comment on this. It will also be nice if the curve for wildtype flies growing in the normal diet can be included in the same figure for easy comparison.
The key difference between the non-competing heteroplasmic lines on live food (Figure 2 and Figure 2—figure supplement 1) and the control diet (Figure 6H in the revised manuscript) is that the OP50 bacteria used in the control diet was UV-killed, which we show has a negative effect on ∆mtDNA frequency during development (Figure 3—figure supplement 1C). We have revised the Results section to acknowledge the impact of live versus UV-killed food on the dynamics of ∆mtDNA in the non-competing lines (subsection “Nutrient availability influences sub-organismal ∆mtDNA dynamics”). We have also revised Figure 6H in accordance with the reviewer’s request.
4) Subsection “Nutrient status governs selection on ∆mtDNA at different levels”, first paragraph: the authors showed that diet restriction accelerated the decline of Df5 frequency at the level of organismal selection among daf-16 mutants. Is this acceleration due to much-reduced fertility (e.g. caused by reduced sperm viability or oocyte production) when heteroplasmy is combined with daf-16 deficiency? For Figure 7C and D, I think it is more important to show how the fraction containing the Df5 decreases over generations rather than population/individual or population-wide percentage of Df5, because the latter two also contain the sub-organismal selection effect, especially for the daf-16 mutant.
The reviewer raises an interesting question regarding the nature of the greatly reduced organismal-level fitness when diet restriction is combined with the daf-16 mutant genotype. Although we do not know the underlying basis of this accelerated selection, we note that FoxO/DAF-16 prolongs organismal survival during nutrient deprivation, as shown in the work of Greer et al., 2007, Kramer et al., 2008, and Hibshman et al., 2018. We speculate that nutrient scarcity and the metabolic dysfunction due to ∆mtDNA represent two sources of energy stress, which additively may compromise host survival and possibly development in the absence of the stress-response functions of daf-16.
With respect to the effect of organismal selection on ∆mtDNA, since every individual among the non-competing lines is a carrier of ∆mtDNA, the population-wide frequency of the non-competing lines is equal to the mean heteroplasmic (sub-organismal) frequency. We therefore normalized to sub-organismal frequency by normalizing to the population-wide frequency of the non-competing lines. Moreover, normalizing ∆mtDNA frequency to that of the non-competing lines at each generation sets the overall slope of the non-competing lines to zero. Any significant non-zero slope that remains among the competing lines can be reasonably attributed to the presence of wildtype animals, as this is the only difference between the competing and non-competing lines. Accordingly, normalizing to the frequency of the non-competing lines provides a way to control for sub-organismal selection and isolate the effect of organismal selection on ∆mtDNA frequency. To avoid confusion, we have updated the Results section to explain this method of normalization more clearly (subsection “An experimental strategy to isolate selection on a selfish mitochondrial genome at different levels”, last paragraph). The vertical axis label of Figure 2E has also been updated to clarify that population-wide ∆mtDNA frequency is normalized to the mean sub-organismal frequency. In the competition experiments involving diet and daf-16, we used the same approach for controlling for the effect of sub-organismal selection. In addition, we also added the fractions of ∆mtDNA-containing individuals at the conclusion of the competition experiments (generation 8) to the main figures, in accordance with the reviewer’s request.
5) I found the presentation of the data not easy to follow. For example, the population/individual calculation was not properly explained in Figure 2E or its legend. I suggest that either the Figure 2—figure supplement 1A and B to be put in Figure 2E, or the figure legend explains what exactly population-wide Df5 frequency relative to average individual heteroplasmic frequency means. Alternatively, the authors can present only Figure 2—figure supplement 1B in the main figure, which shows how the fraction of heteroplasmic flies decreases over generations. After all, we are looking at competition at the organismal level. Figure 1, Figure 2 and Figure 4E-G legends should contain information on growth conditions including food (e.g. live OP50) and temperature (20 degrees) although such information has been described in the Materials and methods and supplementary information. The presence of the above information in the legend will help readers understand experimental setups without referring to other parts of the paper and make comparisons between data from different figures much easier. The authors also assumed readers to have plenty of pre-knowledge of the Df5 heteroplasmic line. It will be good to have additional descriptions of the Df5 heteroplasmic line in the Introduction.
We acknowledge the complexity of the data. We have taken several steps to address reviewer concerns, including taking their suggestions. Specifically, we have expanded the figure legends to include additional information for every single experiment: diet, temperature, and ages at which animals were used for experiments. Additionally, Figure 2—figure supplement 1B from the original manuscript has been moved to the main figure. The Results section (subsection “An experimental strategy to isolate selection on a selfish mitochondrial genome at different levels”, last paragraph) and the legend accompanying Figure 2E have been revised to more thoroughly explain the rationale and the method for normalizing the population-wide ∆mtDNA frequency to that of the individual/sub-organismal frequency. Finally, the Introduction section has been expanded to provide background information on ∆mtDNA, in accordance with the reviewer’s request (Introduction, fifth paragraph).
Reviewer #4:
This manuscript reveals a number of interesting and surprising phenomena that influence the relative and absolute abundance of a deletion-bearing mitochondrial genome (∆mtDNA). It also ties these into a multi-level selection model in an effort to define how organismal and sub-organismal fitnesses are linked. There is much to commend here, and the writing is first-rate. However, I do see a few weakness, which are noted below.
1) The paper's title and overall text focus on the impact of nutritional status, and the signaling pathway that relays this to the cells and tissues, on the replication of the ∆mtDNA genome. However, the huge effect of temperature on even wild-type worms is not given much billing nor explored, beyond serving as an assay to examine the impact of the insulin pathway and nutrition. Is the temperature effect related to the germ line? A re-motivation of the experiments that begins by problematizing the temperature effect would produce a more straightforward paper. For example, the failure to do this makes the following sentence very hard to parse “Together, these data show that DAF-2 signaling inhibits DAF-16 to allow high mtDNA copy number, which permits sub-organismal ∆mtDNA proliferation at the warmer temperature”.
We made several revisions to clarify the way in which temperature was used in these experiments, and what we think underlies the effects of temperature on ∆mtDNA frequency. In particular, we now emphasize the comparison across genotype in the main figure, which contains 25°C data – the restrictive temperature that inactivates insulin signaling in the daf-2(e1370) mutants – and we moved the data corresponding to the 16°C permissive-temperature controls to the supplement. These revisions are reflected in the first paragraph of the subsection “∆mtDNA exploits nutrient sensing to proliferate across development”. In addition, the fourth paragraph of the revised Discussion section has been added to address the temperature effect.
2) Like abundant food, we see that high temp. also promotes ∆mtDNA proliferation, though this requires DAF-2 function. Thus, high temp. acts like abundant food. This is unusual, given that 25°C is somewhat stressful and dauer-promoting. I'm having a hard time reconciling these two results.
We concur that the combined variables of diet and temperature can make the data difficult to parse, and we thank the reviewer for bringing this to our attention. Although temperature and nutrient deprivation each represent a source of stress, we note that not all stressors are equivalent. In particular, whereas nutrient deprivation impedes ∆mtDNA proliferation, we and others have previously found that the stress caused by the presence of ∆mtDNA itself elicits nuclear-encoded stress responses that promote ∆mtDNA proliferation (Gitschlag et al., 2016 and Lin et al., 2016). We propose that the warmer temperature might stress the animals in some ways that are similar to the presence of ∆mtDNA, which could elicit the amplification of the mutant genome. Accordingly, we have expanded the Discussion section to elaborate on this point (fourth paragraph).
3) "We therefore conclude that suppression of germline development by DAF-16 accounts for the reduced mtDNA content in insulin-signaling mutants." We thus have a bit of paradox: growth 25°C doesn't stimulate germline proliferation, yet it does increase ∆mtDNA. So, there may be two independent effects here?
We agree with the reviewer that the data are indicative of two independent effects. We also agree that our data suggest that temperature increases ∆mtDNA proliferation independent of stimulating germline proliferation. As per the request of other reviewers, who have also noted the importance of addressing the effect of temperature, we have taken two key steps in the revised manuscript. First, we have revised the presentation of the data in Figure 3E-G to focus more on the comparison across genotypes, by moving the data from the permissive 16°C to the supplement. Second, we now devote a paragraph in the Discussion section to discuss the effect on temperature (Discussion, fourth paragraph), and speculate on the potential mechanisms that may underlie these effects. Taken together, we think that the changes we have made should help bring clarity to the presentation of the data, while at the same time acknowledging the effect of temperature.
4) The two pairwise comparisons in Figure 6B look qualitatively similar to me. Seems dangerous to build much of an argument around this difference.
We acknowledge that the difference between wildtype and daf-16 – with respect to diet-dependent ∆mtDNA proliferation – is modest. However, this difference is significant and reproducible. Importantly, we note that the plots in Figure 5C, D show that across development from embryo to adulthood, increasing copy number coincides with a shift to higher ∆mtDNA frequency; however, this rise is dependent on both food abundance (panel C, compare diets) and also daf-16 (compare panels C and D, control diet).