To test the role of potential regulatory factors in mtDNA purifying selection, we took advantage of uaDf5, a 3.1-kb mtDNA deletion mutation that removes part or all of four protein-coding genes and seven tRNAs (Figure 1A; Tsang and Lemire, 2002). Given the presumably deleterious nature of this defective mtDNA, it was of interest to understand how it is stably transmitted despite active purifying selection processes. While its maintenance at high levels is attributable in part to stabilization by the mitochondrial UPR (Gitschlag et al., 2016), uaDf5 persists, albeit at lower levels, in animals lacking this activity. One possible explanation for this phenomenon is that the mutant mtDNA might be maintained in heteroplasmy with an otherwise intact mtDNA carrying a complementing mutation. We sought to test this possibility through deep sequencing of mtDNA isolated from the uaDf5-bearing strain. Comprehensive sequence analysis revealed that, in addition to the large deletion, the strain indeed carries a second mtDNA mutation, w47 (Figure 1A, Figure 1—figure supplement 1A–C). w47 is a single base pair insertion in the nduo-4 gene that causes a frameshift, predicted to result in a truncated NADH dehydrogenase 4 (ND4) protein lacking 321 residues (Figure 1—figure supplement 1D). ND4 is an essential transmembrane subunit within complex I of the mitochondrial respiratory chain (MRC), which drives NADH-oxidation-dependent transport of protons across the inner mitochondrial membrane (Lemire, 2005; van der Bliek et al., 2017; Sousa et al., 2018). While this raised the possibility of two complementing mtDNA genomes, we found that the w47 mutation is present at the same abundance as the uaDf5 deletion mutation (~75% of total mtDNA, Figure 1—figure supplement 1B) rather than that of the wildtype mtDNA, strongly suggesting that it resides on the same mtDNA genome. As this second mutation cannot explain stabilization of the defective mtDNA by trans-complementation of two deleterious mutations, other mechanisms appear to promote the stable inheritance of uaDf5.

In addition to the aberrant protein encoded by the w47 frameshift mutation in nduo-4, a second abnormal protein is encoded by the uaDf5 genome: one end of the deletion results in a fusion protein comprised of the first 185 amino acids of NADH dehydrogenase 1 (ND1, a homolog of the core MT-ND1 transmembrane subunit of complex I of the MRC; Lemire, 2005; Sousa et al., 2018; Baradaran et al., 2013), and the last 81 amino acids of mitochondrial-encoded cytochrome b (CTB-1/CYTB, a transmembrane subunit of complex III of the MRC; Lemire, 2005; Sousa et al., 2018; Song et al., 2016; Figure 1—figure supplement 1E). It is conceivable that accumulation of these two abnormal proteins – the truncated ND4 and the ND1-CYTB fusion protein resulting from w47 and uaDf5, respectively – activate the UPR^MT, which has been shown to result in clearance of mtDNA^uaDf5, dependent on the ATFS-1 transcription factor (Gitschlag et al., 2016; Lin et al., 2016).

Animals harboring mtDNA^uaDf5 are viable and fertile, presumably because they contain intact wildtype mtDNA (Tsang and Lemire, 2002). However, we found that uaDf5-bearing animals displayed a significant reduction in brood size (WT 304 ± 4.8; uaDf5 201 ± 8.6 embryos laid, p < 0.001) (Figure 1B) and a significant increase in embryonic lethality (WT 1.4 ± 0.2%; uaDf5 4.2 ± 0.7%, p < 0.001) (Figure 1C). Additionally, uaDf5 animals are slow-growing, evident in both the number of hours to reach gravidity (WT 63 ± 0.8; uaDf5 76 ± 1.4 hr at 20°C, p < 0.001) (Figure 1D) and the stage of development reached after 60 hr of feeding (WT: adult; uaDf5: mid-L4) (Figure 1E). In contrast, however, we were surprised to find that lifespan was not substantially affected (WT 14 ± 0.4; uaDf5 15 ± 0.5 days) (Figure 1F, Figure 1—figure supplement 2). Given the significant decline in the majority of fitness parameters tested, we conclude that uaDf5 is a useful tool for studying mitochondrial disease and mechanisms underlying mtDNA quality control.

During germline development in C. elegans, as many as 95% of nuclei destined to become potential oocytes are eliminated by PCD (Gumienny et al., 1999; Baum et al., 2005; Jaramillo-Lambert et al., 2007; Gartner et al., 2008). While this process has been proposed to be stochastically determined (Gumienny et al., 1999; Gartner et al., 2008), it has also been suggested that it may function to selectively remove all but the most ‘fit’ germline cells. As such, PCD could perform a role in purifying selection in the germline, wherein potential oocytes that undergo PCD are associated with higher levels of defective mtDNA. To test this hypothesis, we introduced uaDf5 into various PCD mutants and quantified abundance of the defective mtDNA by digital-droplet PCR (ddPCR; see Supplementary file 1 for list of mutants tested). In a wildtype genetic background, we found that the steady-state fractional abundance of uaDf5 in populations of 200 day 1 adults (first day of adulthood) is highly reproducible across four separate trials, demonstrating the reliability and robustness of the assay. Our analyses confirmed that mtDNA^uaDf5 constitutes the major molar fraction of mtDNA in the uaDf5-bearing strain by a nearly 3:1 ratio (Figure 2—figure supplement 1).

CED-3 in C. elegans is the major executioner caspase in the canonical PCD pathway (Conradt et al., 2016; Gartner et al., 2008; Cohen, 1997) and is required for virtually all PCD both in the germline and the soma (Figure 2—figure supplement 2 adapted from Figure 2 in Conradt et al., 2016). We found that two ced-3 mutations that strongly block PCD (Shaham et al., 1999) showed a significant increase in the ratio of defective to normal mtDNA from a molar ratio of 2.7:1 for ced-3(+) to 3.4:1 for ced-3(n717) and 4.6:1 for ced-3(n1286) (Figure 2A). This effect is attributable to an increase in mtDNA^uaDf5 in the PCD-deficient strains and not to a decrease in wildtype mtDNA^WT. Rather, we observe an increase in the abundance of mtDNA^WT in ced-3(−) mutant strains relative to ced-3(+) (1.45 × 1 0⁵) ranging from a statistically insignificant 1.1-fold increase (1.67 × 10⁵, p = 0.302) in ced-3(n717) to 2.2-fold increase (3.2 × 10⁵, p = 0.0262) in ced-3(n1286) (Figure 2A, Supplementary file 2). These ced-3(−) mutations both localize to the p15 domain of the protease portion of CED-3 (Figure 2—figure supplement 3), consistent with abolition of caspase activity. These findings implicate the CED-3 caspase and its p15 domain in mtDNA quality control. We found that one other mutation located in the p15 domain showed only a very slight increase that was not statistically significant (n2454: 2.9:1, Figure 2B). While it is unclear why this allele showed a weaker effect it is noteworthy that, unlike the other two mutations, which result in a dramatic alteration of the protein, this mutation is predicted to result in a relatively modest (ala → thr) single amino acid substitution.

A second caspase in C. elegans, CSP-1, also functions, albeit less prominently, in PCD. While loss of CSP-1 alone does not result in a strong reduction in PCD, it synergizes with loss of CED-3 both in PCD and in other caspase-dependent processes (Figure 2—figure supplement 2; Denning et al., 2013; Shaham and Shaham, 1998; Jeong et al., 2020). We found that removing CSP-1 in the csp-1(tm917) knockout mutant results in a significant increase in mtDNA^uaDf5 abundance to a molar ratio of 3.9:1 (Figure 2A, Supplementary file 2). Further, we found that this mutation enhances the effect of the ced-3(n717) mutation, increasing the mtDNA^uaDf5:mtDNA^WT molar ratio from 3.5:1 to 4.7:1 (Figure 2A, Supplementary file 2). Together these findings demonstrate that caspase activity, and possible PCD-mediated clearance, are crucial for mtDNA quality control and function in purifying selection of defective mtDNA.

We sought to further investigate a potential role for PCD in mtDNA purifying selection by evaluating the requirement for the pro-apoptotic factor CED-13, a BH3-only domain protein that acts specifically in the germline to activate PCD (Conradt et al., 2016; King et al., 2019; Schumacher et al., 2005). Consistent with a requirement for PCD in purifying selection, we found that two ced-13 alleles result in a very substantial increase in the mtDNA^uaDf5:mtDNA^WT molar ratio (sv32: 4.2:1 and tm536: 5.1:1) (Figure 2A, Supplementary file 2), supporting the notion that CED-13 promotes removal of defective mtDNA in the germline. CED-13 functions in PCD by antagonizing the function of mitochondrially localized CED-9/Bcl-2 (King et al., 2019; Schumacher et al., 2005), which normally sequesters the apoptosome factor CED-4/APAF1 at mitochondria, thereby preventing it from triggering autocatalytic conversion of the executioner caspase zymogen proCED-3 to its pro-apoptotic protease structure (Figure 2—figure supplement 2; Fairlie et al., 2006). An equivalent action is carried out in the soma by the BH3-only protein EGL-1 (Fairlie et al., 2006). n1950, a gain-of-function allele of ced-9 that blocks the interaction of EGL-1 with CED-9 at the mitochondria, results in elimination of PCD in the soma but not the germline (Conradt et al., 2016; Gumienny et al., 1999; Gartner et al., 2008; Fairlie et al., 2006). Consistent with the lack of effect of ced-9(n1950gf) on germline PCD, we found that the mtDNA^uaDf5:mtDNA^WT molar ratio was not increased in ced-9(n1950gf) mutants (2.2:1) (Figure 2B, Supplementary file 2). Thus, CED-3, CSP-1, and CED-13 are required both for germline PCD and for removal of mtDNA^uaDf5.

Cells that undergo PCD are cleared by the surrounding cells in the process of engulfment and degradation, which is implemented through a set of redundant pathways that converge on the CED-10 GTPase (Wu et al., 2017; Kinchen et al., 2005; Hochreiter-hufford and Ravichandran, 2013; Wang and Yang, 2016; Figure 2—figure supplement 2). Although this engulfment process is necessary primarily for removal of the resultant corpses, it also appears to play an active role in cell killing: inhibition of the engulfment pathway diminishes occurrence of PCD, likely through a complex feedback mechanism (Reddien and Horvitz, 2004; Hoeppner et al., 2001). Further supporting a role for PCD in purifying selection, we found that single or double mutations of several genes that promote engulfment of cell corpses result in elevated mtDNA^uaDf5:mtDNA^WT molar ratios, ranging from 3.5:1 to 5.2:1 (4.1:1 for ced-1(e1735), 4.5:1 for ced-2(e1752), 5.2:1 for the ced-1(e1735); ced-2(e1752) double mutant, 3.6:1 for ced-10(n1993), and 4.8:1 for ced-10(n3246); Figure 2A, Supplementary file 2). As we observed for the caspase mutants, the increase in relative abundance of mtDNA^uaDf5 seen in the cell corpse engulfment mutants is not associated with a significant decrease in wildtype mtDNA levels.

We tested whether generally increased germline PCD alters mtDNA^uaDf5 abundance by examining the effect of removing the caspase-related factor, CSP-2, which has been shown to play an anti-apoptotic role through inhibition of CED-3 autoactivation in the germline (Figure 2—figure supplement 2; Geng et al., 2009). We found that a loss-of-function mutation in csp-2, which elevates germline PCD, did not alter the mtDNA^uaDf5:mtDNA^WT molar ratio (2.9:1 for the csp-2(tm3077) knockout mutation) (Figure 2B). This observation does not conflict with a requirement for PCD in purifying selection: a general increase in PCD in the germline of csp-2(−) animals would not be expected per se to alter the mechanism that discriminates defective from normal mtDNAs and therefore the relative abundance of the two forms. Rather, our findings suggest that the mechanisms that recognizes and disposes of the defective mtDNA specifically requires PCD components acting in a selective, rather than general process (i.e., in those cells with the highest burden of the defective mtDNA).

As suggested above, it is conceivable that the effects described were attributable to changes in the abundance of mtDNA^WT per se. For example, if the nuclear mutations tested resulted in decreased mtDNA^WT copy numbers without affecting mtDNA^uaDf5 abundance, then the mtDNA^uaDf5:mtDNA^WT molar ratios would be skewed upward. However, while we observed modest variation in mtDNA^WT abundance across the mutant strains (Figure 2), across the strains, the variation does not generally correlate with the increased fractional abundance of mtDNA^uaDf5: while some mutants with elevated fractional abundance of the defective mtDNA showed somewhat higher levels of mtDNA^WT, others contained lower levels of intact mtDNA (Figure 2). For example, although the ced-13(tm536) mutant showed the highest mtDNA^uaDf5:mtDNA^WT molar ratio, the abundance of mtDNA^WT was also higher than in the wildtype nuclear background (2.4 × 10⁵ ± 3.09 × 10⁴ in ced-13(tm536) vs 1.45 × 10⁵ ± 2.44 × 10⁴ in wildtype p = 0.026). Further, while the ced-1(−); ced-2(−) double mutant showed among the largest relative abundance of mtDNA^uaDf5, this strain contained the lowest level of mtDNA^WT (1.15 × 10⁵ ± 3.02 × 10³, p = 0.254). Thus, increased levels of defective mtDNA in the PCD mutants are not apparently attributable to alterations in mtDNA^WT levels.

The foregoing results implicate a role for the pro-apoptotic CED-3 and CSP-1 caspases, CED-13, and the CED-1, 2, and -10 cell corpse engulfment factors in mtDNA purifying selection. However, several observations suggest that mitochondrial purifying selection may be regulated by a non-canonical cell death pathway, in contrast to the pathway that regulates normal, physiological germline PCD.

First, we found that the ced-3(n718) allele lowers, rather than elevates, the abundance of mtDNA^uaDf5 (mtDNA^uaDf5:mtDNA^WT molar ratio of 1.8:1, Figure 2C). This effect is likely to be attributable to the nature of the n718 mutation. The ced-3 mutations that result in increased mtDNA^uaDf5 levels (Figure 2A) alter the p15 domain, which is essential for active caspase function. In contrast, the n718 mutation changes a residue in the caspase activation and recruitment domain (CARD), located within the prodomain of the CED-3 zymogen, which is removed upon caspase activation and affects its activation by CED-4 (Figure 2—figure supplement 3; Huang et al., 2013; Shaham et al., 1999). While ced-3(n718) strongly compromises PCD, this mutation might not alter CED-3 caspase function in a way that interferes with its role in mtDNA purifying selection.

Our surprising finding that while CED-3 activity is required for mtDNA purifying selection, a CED-3 mutation that compromises its activation by CED-4 did not elevate mtDNA^uaDf5 levels prompted us to investigate the requirement of CED-4 in mitochondrial purifying selection. Consistent with the effect of the ced-3(n718) mutation, we found that eliminating the function of the pro-apoptotic regulator CED-4, the C. elegans ortholog of mammalian Apaf1 and the upstream activator of CED-3 in the canonical PCD pathway (Fairlie et al., 2006; Yang et al., 1998), did not result in a marked increase in the relative abundance of mtDNA^uaDf5. That is, while the mtDNA^uaDf5:mtDNA^WT molar ratio increased to 3.1:1 in the ced-4(n1894) mutant, the effect was not statistically significant. Moreover, the canonical allele ced-4(n1162) allele similarly showed no elevation in mtDNA^uaDf5 (molar ratio = 2.6:1; Figure 2B). These results suggest that CED-3 caspase functions in mitochondrial purifying selection independently of the caspase-activating factor CED-4.

Taken together, our results implicate many PCD regulatory factors, and potentially PCD, in the selective clearance of defective germline mtDNAs. Our additional observations suggest that defective mtDNAs may, in fact, trigger elevated germline PCD, resulting in the production of fewer mature gametes and progeny. Specifically, we found that the significant decrease in brood size that we observed in uaDf5-bearing animals with a wildtype nuclear background is partially suppressed by both ced-3(−) and ced-13(−) mutations, which prevent PCD (Figure 2D, Figure 2—figure supplement 4A), suggesting that elimination of PCD might rescue cells that would otherwise be fated to die as a result of accumulation of defective mtDNA. Our findings further underscore the observation that accumulation of defective mtDNA in those animals that do survive does not affect longevity, as we found that lifespan is unaltered in these PCD mutants even when the levels of mtDNA^uaDf5 are nearly doubled (Figure 2—figure supplement 4B, C).

Our findings that PCD regulators are required to reduce mtDNA^uaDf5 abundance, the central role that mitochondria play in PCD (Jeong and Seol, 2008; Estaquier et al., 2012; Bhola and Letai, 2016; Parsons and Green, 2010), the observed decline of mitochondrial health during the aging process (Park and Larsson, 2011; Larsson, 2010; Payne and Chinnery, 2015; Kauppila et al., 2017; Harman, 1992; Harman, 1956; Szczepanowska and Trifunovic, 2017; Bratic and Larsson, 2013; Ziegler et al., 2015), and the relationship between excessive PCD and the aging phenotype (Tower, 2015) led us to examine the dynamics of mtDNA^uaDf5 accumulation as worms age. We measured the fractional abundance of uaDf5 in adults at progressively increased ages spanning day 1, defined as the first day of egg-laying, through day 10. Day 1 through day 4 of adulthood encompasses the time during which nearly all self-progeny are produced. After day 4, hermaphrodite sperm become depleted and the animals transition into a post-gravid, progressively aging state (Kimble and Crittenden, 2005; Yoon et al., 2017; Angeles-Albores et al., 2017). By day 10, animals exhibit indications of advanced age. Analysis of the abundance of mtDNA^uaDf5 revealed a progressive increase throughout gravidity and post-reproductive aging, with the mtDNA^uaDf5:mtDNA^WT molar ratio increasing from 2.9:1 to 5.5:1 (Figure 3A, Figure 3—figure supplement 1A). This age-related accumulation of uaDf5 in adult worms is reminiscent of the accumulation of mtDNA mutations seen in aging mammals (Larsson, 2010; Szczepanowska and Trifunovic, 2017; Kujoth et al., 2005) and suggests that uaDf5 in C. elegans may be a useful tool for studying the role that mtDNA mutations play in aging.

Given that gametes are depleted with age, it is conceivable that the age-dependent increase in mtDNA^uaDf5 is attributable to accumulation in somatic mitochondria. To assess whether the observed age-related accumulation of mtDNA^uaDf5 occurs predominantly in the maternal germline or in somatic cells, we analyzed animals defective in germline development by taking advantage of the glp-4(bn2) mutant, which produces only a small number (~12) of germline cells compared to that in wildtype animals (~1500), with no known effect on somatic gonad development (Beanan and Strome, 1992). In contrast to the increased mtDNA^uaDf5 abundance with age seen at permissive temperature (mtDNA^uaDf5:mtDNA^WT of 1.8:1 at day 1, rising to 2.6:1 at day 4, for an overall increase by 48%), we found that glp-4(bn2) animals at the non-permissive temperature showed a slight decrease in the defective mtDNA from day 1 to 4 of adulthood (2.2:1 at day 1, dropping to 2:1 at day 4, for an overall decrease of 6.3%) (Figure 3B). These results strongly suggest that the observed age-dependent increase in mtDNA^uaDf5 abundance occurs exclusively in the germline. We found that mtDNA^uaDf5 does eventually appear to accumulate in somatic cells with age, as day 10 adults grown at the restrictive temperature showed a marked increase in the mtDNA^uaDf5:mtDNA^WT molar ratio compared to day 1 adults even in the absence of a germline (day 10 glp-4(bn2) molar ratio of 3.4:1, a 57% increase from day 1 levels) (Figure 3B). We also observed an increase in wildtype mtDNA copy number during the period of gravidity, which is likely the result of compensatory expansion in response to the replicative advantage of mtDNA^uaDf5. We conclude that the marked increase in mtDNA^uaDf5 with age during the period of fecundity occurs primarily in the germline and that the defective mtDNA accumulates in both germline and somatic tissue during post-reproductive life.

As the mtDNA is inherited strictly through the maternal germline, we posited that the age-dependent increase in the fractional abundance of germline mtDNA^uaDf5 might be transmitted to progeny animals. To test this hypothesis, we measured the molar ratio of mtDNA^uaDf5 in 200-worm populations of L1 larvae derived from day 1 to 4 adults (Figure 3C, Figure 3—figure supplement 1B). This analysis led to two key observations: (1) the abundance of mtDNA^uaDf5 is reduced during transmission between mother and offspring (average decrease ranging from 6% to 13%), presumably as a result of purifying selection, and (2) the abundance of the defective mtDNA in the offspring correlates with the age of the mothers: the progeny of older mothers contain a markedly higher mtDNA^uaDf5:mtDNA^WT molar ratio (4.2:1) than that of younger mothers (2.6:1) (Figure 3C, D). A similar trend was observed for mother-to-offspring transmission in five mutant strains with altered levels of mtDNA^uaDf5 (see below): in all cases, progeny contain lower abundance of mtDNA^uaDf5 than their mothers, and progeny of younger adults inherit a lower load of mtDNA^uaDf5 than progeny of older adults (Figure 3C). These results reveal that mtDNA quality control occurs between primordial germ cell proliferation in the female germline and L1 hatching, that is, during oocyte maturation, embryogenesis, or both.

We sought to determine whether the age-dependent accumulation of defective mtDNA is controlled by known molecular mechanisms that drive the aging program in C. elegans. The most prominent of these regulatory systems is the highly conserved insulin/IGF-1 (insulin-like growth factor-1) pathway (IIS), which performs a pivotal regulatory function in aging and longevity (Murphy and Hu, 2013; Bartke, 2008; Anisimov and Bartke, 2013; Figure 4—figure supplement 1). Abrogation of the IIS signaling pathways, for example, as a result of mutations in the gene encoding the IIS receptor (DAF-2, in C. elegans), results in marked slowing of the aging program and extension of lifespan in worms, flies, and mice (Piper et al., 2008; Kenyon et al., 1993; Tatar et al., 2001; Blüher et al., 2003). The IIS pathway also functions in a broad set of other processes including, in C. elegans, activation of two stages of developmental arrest, or diapause, at the L1 larval stage and in formation of the dispersal form, the dauer larva, as well as in the control of germline proliferation, stress resistance, fat metabolism, and neuronal/behavioral programs (Murphy and Hu, 2013). It was also reported that inhibition of the IIS pathway rescues various fitness parameters in a mtDNA mutator strain which contains a faulty mtDNA polymerase (Haroon et al., 2018), consistent with a possible role in mtDNA quality control.

We found that two mutant alleles of daf-2 that affect the kinase domain and which reduce rates of aging and increase lifespan, result in dramatically decreased mtDNA^uaDf5:mtDNA^WT molar ratios from 2.8:1 to as low as 0.3:1 (0.3:1 for daf-2(e1391); 0.8:1 for daf-2(e1370); Figure 4A; see Supplementary file 3 for a list of lifespan mutants used in the analyses). Thus, the lifespan-extending effects of daf-2 mutations are strongly correlated with diminished abundance of defective mtDNA, to the extent that it becomes the minor species of mtDNA.

The DAF-2 receptor acts by antagonizing the DAF-16/FoxO transcription factor, the major effector of IIS downstream, in response to insulin-like ligands. Thus, removal of daf-16 function reverses the lifespan-extending effects of daf-2(−) mutants. We tested whether the DAF-2 → DAF-16 pathway similarly functions in mtDNA purifying selection. We found that eliminating DAF-16 in two daf-16 mutants results in slightly increased, albeit not statistically significant, mtDNA^uaDf5:mtDNA^WT molar ratios (3.3:1 for daf-16(mu86); 3.5:1 for daf-16(mgDf50)) (Figure 4B). Further, we found that removal of DAF-16 in the daf-16(mu86) mutant suppressed the decreased mtDNA^uaDf5:mtDNA^WT molar ratios observed in two daf-2 mutants, from 0.3:1 for daf-2(e1391) to 1:1 for daf-16(mu86); daf-2(e1391) and from 0.6:1 for daf-2(e1370) to 1.8:1 for daf-16(mu86); daf-2(e1370), consistent with observations reported in a recent study (Kirby and Patel, 2021). While the daf-2(−) effect on mtDNA^uaDf5 levels is largely dependent on DAF-16, neither double mutant restored mtDNA^uaDf5 levels to those seen in animals with a fully intact IIS pathway, suggesting that other DAF-2 targets might participate in removal of defective mtDNA (Figure 4C). We found, conversely, that two mutations that reduce lifespan by eliminating the function of AAK-2 (AMP activated kinase-2), a conserved factor acting in the IIS pathway (Murphy and Hu, 2013; Haroon et al., 2018), result in elevated mtDNA^uaDf5:mtDNA^WT molar ratios as high as 4.4:1 compared to 2.8:1 in wildtype animals (3.8:1 for aak-2(ok524) and 4.4:1 for aak-2(gt33)) (Figure 4B). These findings demonstrate that alterations in the IIS pathway coordinately affect both lifespan and accumulation of defective mtDNA and that the DAF-2/DAF-16/AAK-2 axis acts similarly in both processes.

In addition to IIS, other molecular regulatory pathways independently contribute to the rate of aging. These include CLK-1, a mitochondrial hydroxylase that functions in the pathway for ubiquinone synthesis (Miyadera et al., 2002; Miyadera et al., 2001). clk-1 mutants with a wildtype mitochondrial genome have been shown to contain levels of mtDNA that are elevated by 30%, perhaps as the result of a compensatory process that increases demand on mitochondrial abundance, or the action of CLK-1 as a regulator of mtDNA abundance in response to energy availability within the cell (Kirby and Patel, 2021). As with long-lived daf-2 mutants, we found that long-lived clk-1(qm30) mutants showed a greatly diminished mtDNA^uaDf5:mtDNA^WT molar ratio of 0.3:1 (Figure 4A), comparable to that in daf-2 mutants; again, mtDNA^uaDf5 is the minor species in these animals. As the IIS pathway and CLK-1 appear to act separately in controlling lifespan, we postulated that elimination of both mechanisms might further reduce levels of the defective mtDNA. Indeed, we found that mtDNA^uaDf5 was completely eliminated in daf-2(e1391) clk-1(qm30) double mutants, revealing a strongly synergistic effect between the two age-determining systems (Figure 4A). Thus, distinct regulatory pathways for longevity modulate the abundance of mtDNA^uaDf5 by apparently different mechanisms and elimination of the two pathways abrogates its maintenance.

Our findings that the steady-state abundance of mtDNA^uaDf5 increases with maternal age and that mutants with increased lifespan show lower levels of the defective mtDNA raised the possibility that purifying selection is subject to the same control as aging clocks. To assess this potential connection, we analyzed the time-dependent rates of mtDNA^uaDf5 accumulation in animals with genetic backgrounds that alter the aging clock. Analysis of 21 different genetic backgrounds over the first 10 days of adulthood revealed that the age-dependent progressive accumulation of mtDNA^uaDf5 is a consistent phenomenon (Figure 4D). Comparison of long-lived mutants and wildtype using a linear regression model revealed a striking positive correlation between aging rate and age-dependent rate of accumulation of mtDNA^uaDf5: all long-lived mutants in either the IIS pathway or clk-1 accumulate mtDNA^uaDf5 at a substantially slower rate as they age than do wildtype animals (Figure 4E, Figure 4—figure supplement 2A, B). Conversely, we analyzed six short-lived IIS pathway mutant combinations and found that the aak-2(ok524) and daf-16(mu86) single mutants and daf-16(−);daf-2(−) double mutants all showed increased rates of mtDNA^uaDf5 accumulation (Figure 4E, Figure 4—figure supplement 2C, D). These observations suggest that in both slower- and faster-aging strains, the rate of accumulation of deleterious mtDNA is a predictor of aging rate. In the two exceptional cases, the aak-2(gt33) and daf-16(mgDf50) single mutants, we did not observe an increased accumulation rate compared to wildtype; however, the mtDNA^uaDf5 levels are consistently higher than in these two mutants than in wildtype at all stages (Figure 4E, Figure 4—figure supplement 2D) and thus diminished removal of mtDNA^uaDf5 overall correlates with decreased lifespan in these mutants as well. The greater mtDNA^uaDf5 accumulation rates seen in the short-lived animals is not attributable to the higher steady-state levels per se, as the rates of accumulation of defective mtDNA observed in the PCD mutants with higher mtDNA^uaDf5 levels show no correlation with the steady-state levels (Figure 4—figure supplement 2E–H); rather these increased rates appear specifically to be a property of the shortened lifespan mutants.

Consistent with a relationship between aging rates and accumulation of defective mtDNA, we found that the brood size is decreased and embryonic lethality is increased in short-lived daf-16(−) mutants, but not in the long-lived clk-1(−) mutant compared to those in a wildtype nuclear background (Figure 4—figure supplement 3). These results are consistent with the possibility that longevity pathways modulate fitness in part by regulating mitochondrial homeostasis.

We obtained evidence that defective mtDNA is more effectively removed in offspring of aging adults that carry an unusually high burden of mtDNA^uaDf5. The offspring of day 4 adults in those strains (‘high’ strains) with significantly higher steady-state fractional abundance of mtDNA^uaDf5 showed significantly greater rates of reduction of the defective mtDNA ranging from 24% reduction in ced-10(n1993) (fractional abundance of mtDNA^uaDf5 of 6.5:1–2.8:1) and 17.4% in ced-13(sv32) (6.4:1–3.3:1) to 14.5% in atfs-1(et15) (5.4:1–3.3:1), compared to offspring of (‘low’ strain) mothers with lower steady-state fractional abundance of the mutant mtDNA (13.5% reduction in ced-3(n2454) (3.9:1–2.8:1); 12.7% in ced-4(n1162) (3.2:1–2.5:1) versus only 3.2% in WT (4.7:1–4.2:1)). Remarkably, therefore, day 4 progeny from ‘high’ strains actually inherit a lower mtDNA^uaDf5 load than their siblings born from day 1 to 3 mothers (Figures 3C and 5A). Indeed, we found a strong correlation (r² = 0.61, p < 0.001) between the steady-state level of mtDNA^uaDf5 in mothers and the capacity for its removal between mother and progeny during day 4 of adulthood (Figure 5B). These results raise the possibility that very high levels of mtDNA^uaDf5 in older mothers activate an additional mtDNA purifying selection process such as mitophagy, perhaps independent of the UPR^MT and PCD machinery, thereby ensuring that progeny are not overloaded with defective mitochondria.

We found that uaDf5 affects brood size, embryonic lethality, and developmental rate, highlighting its use as a model for investigating mitochondrial diseases. The reduced brood size in uaDf5-bearing animals might reflect diminished germ cell proliferation, as mitochondria have been implicated in progression of germline maturation (Charmpilas and Tavernarakis, 2020; Folmes et al., 2016). Alternatively, the defective mtDNA might trigger hyperactivation of the germline PCD pathway that specifically removes germ cells with the highest burden of defective mtDNA, as suggested by our results, resulting in the survival of fewer mature oocytes. The increased embryonic lethality in the uaDf5 strain may be a consequence of a genetic bottleneck effect, leading to rapid differences in mtDNA allele frequencies (Floros et al., 2018; Zhang et al., 2018; Wai et al., 2008) and a subpopulation of oocytes containing levels of mtDNA^uaDf5 that exceed a threshold required for viability.

We were surprised to find, in contrast to a previous report (Liau et al., 2007), that mtDNA^uaDf5 did not alter lifespan. Given that mitochondrial mutations are often coupled with compensatory mutations in the nuclear genome (Zhu et al., 2019; Levin et al., 2014; Paliwal et al., 2014; Zhu et al., 2014; Meiklejohn et al., 2013), one possible explanation for this discrepancy might be that a compensatory nuclear mutation exists in the strain analyzed, diminishing the impact of the defective mitochondrial genome. We note, however, that we backcrossed the uaDf5 strain extensively to the laboratory reference strain N2 prior to performing the reported analyses. It is conceivable that although we observed no effect in the lab, uaDf5 might alter lifespan under natural conditions. Exposure to increased stress from growth in the wild might be less tolerated in animals bearing mtDNA^uaDf5, as has been observed with other mitochondrial mutants (Dingley et al., 2014), resulting in diminished lifespan.

Our results lend support to the hypothesis that germline PCD mechanisms may be used to cull germline cells with defective mtDNA (Figure 6A). Developmentally PCD and cell death in response to genotoxic stress are mediated by caspases upon activation by the C. elegans octameric apoptosome which is formed when the inhibition of CED-4 by CED-9 is disrupted through binding of BH-3-only proteins EGL-1 and CED-13 in the soma and germline, respectively (Conradt et al., 2016; Gartner et al., 2008). While somatic cell death is blocked by the ced-9(n1950) gain-of-function mutation, normal physiological germ cell death is not (Gumienny et al., 1999). In keeping with this finding, we found no significant difference in fractional abundance of the mitochondrial deletion in the ced-9(n1950) mutant background as compared to a wildtype nuclear genetic background. We posit that in response to mitochondrial genotoxic stress (increased mtDNA^uaDf5 load), the CED-3 and CSP-1 caspases are activated by the BH-3-only domain protein CED-13, thereby triggering mitochondrial purifying selection through a non-canonical germ cell death pathway, independent of CED-9 and the CED-4 apoptosome (Figure 6A).

Mutations that eliminate the function of caspases that act in PCD result in elevated mtDNA^uaDf5 levels. Two mutations affecting the p15 domain of the CED-3 caspase result in a significantly increased molar ratio of uaDf5, highlighting the importance of the p15 domain in mtDNA quality control. It is noteworthy that the two mutations that result in a more substantial effect on mtDNA^uaDf5 abundance affect CED-3 structure more dramatically: ced-3(n717) results in a splicing error, and ced-3(n1286) is a nonsense mutation (Shaham et al., 1999). In contrast, ced-3(n2454), which results in a subtle (statistically insignificant) increase in mtDNA^uaDf5, is a substitution predicted to impart a much weaker effect on the protein structure (Shaham et al., 1999). The effect of the ced-3(n718) allele, which resides in the prodomain of CED-3, suggests that that portion of the protein may act to inhibit mtDNA quality control. Taken together, our results suggest that CED-3 may carry out a specialized activity in mtDNA-activated PCD. We found that a second caspase, CSP-1, which plays a minor role in PCD (Denning et al., 2013; Shaham and Shaham, 1998), is also required in mtDNA quality control: a csp-1(−) knockout mutation results in increased abundance of mtDNA^uaDf5 and enhances the effect of ced-3(n717), suggesting that CSP-1 may play a larger role in removal of defective mtDNA than it does in other forms of PCD. An exciting possibility is that caspases act in mtDNA quality control via a mechanism that is distinct from their normal action in PCD. Such putative roles for caspases and other mitochondrial factors in non-apoptotic mitochondrial quality control may have been co-opted during metazoan evolution with the innovation of PCD.

Analysis of additional PCD components further implicates a role for germline PCD in mitochondrial purifying selection. These include CED-13, the germline-specific PCD effector, and components of the cell corpse engulfment pathway, which activate PCD likely through a complex feedback mechanism that ensures cells destined to die proceed irreversibly through the process (Reddien and Horvitz, 2004; Hoeppner et al., 2001). In all cases, mutations in these components result in increased mtDNA^uaDf5 levels. In vertebrates, mitochondrial reactive oxygen species (mtROS) trigger apoptosis via the intrinsic mitochondrial pathway (Simon et al., 2000). Interestingly, elevated mtROS promotes longevity that is in part dependent on the core cell death machinery in C. elegans, involving CED-9, CED-4, CED-3, and CED-13, but not EGL-1 (Yee et al., 2014). In that study, the authors reported that the protective effect of the cell death machinery on lifespan was independent of PCD in the soma; however, germline cell death was not characterized. Given that CED-13, and not EGL-1, is the predominant BH3-domain protein functioning in the germline (King et al., 2019) and that ablation of the germline leads to extended lifespan (Hsin and Kenyon, 1999), our findings support the possibility that germline progenitors carrying defective mitochondria selectively undergo PCD, ensuring homeostatic mtDNA copy number and health of progeny.

A striking exception to our findings was seen with mutations that eliminate the function of the pro-apoptotic regulator CED-4. Neither the ced-3(n718) mutation that disrupts the CED-3 CARD domain, which is involved in recruitment to the apoptosome by stabilizing its interaction with CED-4 (Huang et al., 2013; Dorstyn et al., 2018), nor two ced-4(−) mutations, result in increased accumulation of mtDNA^uaDf5, suggesting non-canonical, CED-4-independent activation of CED-3 in mitochondrial purifying selection. It is possible that the CARD mutation (G65R) in the ced-3(n718) mutant (Shaham et al., 1999) reduces the fraction of CED-3 in complex with the apoptosome, which might release more of the protein for its role in mitochondrial purifying selection. Interestingly, the CARD linker domain has been found to have an inhibitory effect on the pro-caspase-9 zymogen (Yuan et al., 2011). ced-3(n718) could be effectively acting as a gain-of-function allele in the process of purifying selection, reflected by the lower levels of mtDNA^uaDf5 in this mutant background.

It is noteworthy that CED-4 and its mammalian Apaf1 relatives regulate a variety of cellular functions that are unrelated to their activities in PCD. These include cell growth control influenced by DNA damage, centrosomal function and morphology, neuronal regeneration, and inhibition of viral replication (Zermati et al., 2007; Ferraro et al., 2011; Liu et al., 2006; Wang et al., 2019). In addition, as a result of differential RNA splicing, ced-4 encodes proteins with opposing activities, generating both an activator and a repressor of apoptosis (Chaudhary et al., 1998), which further complicates analysis of its action. Thus, it is conceivable that CED-4 might exert opposing effects on purifying selection, reflecting its pleiotropic activities in development and confounding an unambiguous interpretation of its action in this process.

We found that most of the increase in the fractional abundance of mtDNA^uaDf5 as worms age throughout the period of self-fertility (days 1–4) occurs in the germline. However, the relative amount of the defective mtDNA continues to increase in older animals when the germline is no longer actively proliferating (Kimble and Crittenden, 2005), suggesting that mtDNA proliferation also occurs in somatic tissues throughout the aging process. This behavior mirrors the dynamics of mutant mtDNAs observed in other organisms, including human, mouse, rat, and rhesus monkey (Larsson, 2010; Szczepanowska and Trifunovic, 2017; Kujoth et al., 2005). However, in contrast to what is observed over the relatively long lifespan of mammals, the expansion in the burden of defective mitochondrial DNA over the span of a few days in C. elegans is likely the outcome of a need to maintain wildtype levels of mitochondrial DNA that supports the energy demands over the short but intense period of gravidity, during which a single hermaphrodite produces hundreds or thousands of progeny. The replicative advantage of the mtDNA^uaDf5 deletion may impose a greater replication drive to maintain wildtype mtDNA levels, resulting in ‘runaway’ expansion of mutant mtDNA. In this scenario, the mechanisms that normally act to eliminate the defective mtDNA would be insufficient to keep up with the increased burden arising from replicative advantage. The expansion the mitochondrial deletion mutation in somatic tissues of old (day 10) post-gravid adults might be an outcome of different dynamics or even an absence of mitochondrial purifying selection mechanisms in the germline. While removal of the germline in worms results in extended lifespan (Mack et al., 2017; Libina et al., 2003), it is not clear whether, or to what extent, this increased lifespan might be attributable to accumulation of mutant mtDNA, since the lack of germline leads to a variety of cellular responses (Lapierre and Hansen, 2012; Antebi, 2013), any of which might lead to lifespan extension.

As in many other organisms, the offspring of older C. elegans mothers show increased embryonic lethality (Andux and Ellis, 2008; Scharf et al., 2021). This age-dependent lethality positively correlates with an increase in accumulation of defective mtDNA. Age-associated changes in the germline include reduced germ cell proliferation, a displaced distal tip cell, germline shrinking, reduction in oocyte production, oocyte clustering, and endomitotic oocytes (Scharf et al., 2021). We observed that gonads of mtDNA^uaDf5 mutants exhibit accelerated signs of aging (not shown) and smaller brood sizes. Interestingly, long-lived mutants are associated with reduced brood sizes (Antebi, 2007; Hekimi, 2006) and extension in the period of egg production, owing to a decrease in oocyte production, as seen in daf-2(−) mutants in which sporadic embryos are laid for as late as 50 days, a more than 10-fold increase over normal (Gems et al., 1998). This effect is also seen as a consequence of caloric restriction in Eat mutants, which show an egg-laying period that lasts up to 10 days, compared to 3–5 days in the N2 laboratory strain (Pickett and Kornfeld, 2013). One of the targets of pro-longevity cues is suppression of vitellogenin expression, a major maternal energy cost (DePina et al., 2011; Murphy et al., 2003; Perez and Lehner, 2019). While vitellogenin is largely dispensable for embryogenesis, the accumulation of vitellogenin in post-reproductive animals reaches pathological levels as a consequence of autophagy-dependent degradation of the worm’s intestinal tissues, which fuels continued massive production of yolk. The complex interaction between reproductive aging, egg production, transgenerational effects of vitellogenins, and levels of mitochondrial mutation load on life and healthspan are intriguing relationships that demand further study.

Our findings do not reveal whether, or how, aging and accumulation of defective mtDNA are causally linked. However, our findings that steady-state levels of mtDNA^uaDf5 are lowered in long-lived mutants (daf-2 (IIS pathway; Murphy and Hu, 2013) and clk-1 mitochondrial function; Branicky et al., 2000; Hekimi and Guarente, 2003) and that rates of its accumulation are strongly inversely correlated with lifespan extension through independent pathways, suggests that mtDNA purifying selection mechanisms are influenced by aging programs (Figure 6A). Further bolstering this potential link is our finding that short-lived mutants (daf-16 and aak-2, both involved in the IIS pathway; Murphy and Hu, 2013), show higher steady-state levels of mtDNA^uaDf5. That this effect is greater in aak-2 mutants than in daf-16 mutants suggests that the AAK-2 branch of the IIS pathway influences mtDNA quality control more significantly than does the DAF-16 branch. One of the substrates of AAK-2 is SKN-1/Nrf2, a multifaceted transcription factor with roles in stress response and longevity, and one of its isoforms, SKN-1a, localizes to the mitochondrial surface (Blackwell et al., 2015), raising the possibility that AAK-2 might influence clearance of defective mtDNA through SKN-1 action. It will be of interest to assess how defective mtDNA might coordinately trigger quality control and stress-response pathways.

Our observation that IIS pathway components and CLK-1 act synergistically on the mtDNA quality control machinery raises the possibility that these two distinct lifespan-regulating pathways converge on a common system for removal of defective mtDNA, such as a global mitochondrial stress response pathway. Candidates for mediating this removal process include mitochondrial fission/fusion (Ni et al., 2015; Tilokani et al., 2018; Youle and van der Bliek, 2012), mitophagy (Ashrafi and Schwarz, 2013; Youle and Narendra, 2011; Twig et al., 2008), and the UPR^MT (Rolland et al., 2019; Münch, 2018; Callegari and Dennerlein, 2018; Gitschlag et al., 2016; Nargund et al., 2012; Lin et al., 2016), all of which are known to act in mtDNA quality control, as well as modulation of the regulatory pathway for PCD, as suggested by our findings. It is possible that along with delaying germ cell replication, aging programs could also slow the replication of mtDNA thereby possibly attenuating the runaway replicative advantage of mtDNA harboring large deletions like uaDf5 thereby reducing the fractional abundance of mtDNA^uaDf5.

Analysis of newly hatched L1 larvae revealed that the relative load of mtDNA^uaDf5 is transmitted from mother to offspring, with evidence for intergenerational purifying selection (Figure 6B). This finding implies that mtDNA quality control occurs between germline stem cell expansion in the mature female germline and L1 hatching, a developmental period that spans many potential stages at which it might occur, including germline PCD, oocyte maturation, and the entirety of embryogenesis. It is conceivable that this selection process acts at multiple stages throughout this developmental window and that the decreased burden of defective mtDNA in newly hatched L1 larvae reflect the summation of a series of sequentially acting processes that incrementally enrich for healthy mtDNA.

The efficacy of intergenerational removal of mtDNA^uaDf5 increases in old mothers, including in strains lacking pro-apoptotic regulators. The clearance is particularly precipitous in strains with a very high burden of defective mtDNA as seen in the absence of CED-13, CED-10, and ATFS-1, suggesting a critical threshold beyond which a germline PCD-independent mtDNA quality control process may be triggered in these older mothers (Figure 6B). One possible explanation for this observation is that an mtDNA purifying selection mechanism that is typically inhibited by germline PCD might be activated in older mothers. This hypothesized purifying selection mechanism might be triggered by the unique cellular environment associated with aging such as increased organelle or macromolecule damage. Alternatively, the effect might be the result of an age-dependent genetic program.

While our study has uncovered new mechanisms acting in mtDNA purifying selection and its relationship to aging, it is of note that some level of mtDNA^uaDf5 is maintained in all but the most extreme conditions we have observed (e.g., in the daf-2(e1391) clk-1(qm30) double mutant, in which the defective mtDNA is extirpated). This finding raises the possibility that some degree of heteroplasmy, even with defective mtDNA, is not only tolerated, but may be adaptive by providing a degree of evolutionary plasticity. Cells might purposefully allow for limited heteroplasmy as a way of increasing genetic heterogeneity that might prove evolutionarily advantageous. Such heterogeneity may also be essential to allow mtDNA to co-evolve with changes arising in the nuclear genome. It may be that a dynamic balance between active mtDNA purifying selection, including the mechanisms identified here, and the permissibility of limited heteroplasmy, is modulated according to environmental or physiological demands.

Nematode strains were maintained on NGM plates as previously described at either 20 or 15°C for the temperature-sensitive strains (Stiernagle, 2006). Supplementary files 13 and 4 provide details of all strains used in this study. Strains without a JR designation were either provided by the CGC which is funded by NIH Office of Research Infrastructure Programs (P40 OD010440) or were obtained from the Mitani lab (strains with a FX designation or JR strains containing alleles with a tm designation were generated from Mitani lab strains) (Barstead et al., 2012).

Upon retrieval of a stock plate for a given strain, three chunks were taken from the stock plate and placed onto three separate large NGM plates to create three biological replicates (lines). Each of these lines was chunked approximately each generation to fresh large NGM plates (every 3 days if maintained at 20 or 25°C, or every 4 days if maintained at 15°C, being careful to not let the worms starve between chunks). After four generations of chunks, an egg prep was performed on each line (as described previously; Stiernagle, 2006) and left to spin in M9 overnight to synchronize the hatched L1s. The next day, each egg prep was plated onto three large seeded plates at an equal density and the worms were left to grow to day 2 adults (second day of egg laying). The day 2 adult worms were egg prepped for synchronization and left to spin in M9 buffer overnight. The next day, each egg prep was plated onto five large NGM plates at equal density. Once the worms reached day 1 of adulthood (first day of egg-laying), one of the plates was used to collect 200 adult worms by picking into 400 μl of lysis buffer, and the remaining adults on the plate were egg prepped for the collection of hatched L1 larvae in 400 μl lysis buffer the following day. The worms on the four remaining plates were transferred to a 40-μm nylon mesh filter in order to separate the adults from the progeny, and the resulting adults were resuspended in M9 and pipetted onto fresh large NGM plates. This process was repeated for the following 3 days (days 2–4 of adulthood). Day 5–10 adults were moved to fresh NGM plates every second day using a 40-μm nylon mesh filter, and the resulting day 10 adults were collected in lysis buffer.

The worm lysates were incubated at 65°C for 4 hr and then 95°C for 30 min to deactivate the proteinase K. Each lysate was diluted; 100-fold for 200 worm adult population lyses, 2-fold for 200 worm L1 population lyses, and 25-fold for individual adult lyses. 2 μl of the diluted lysate was then added to 23 μl of the ddPCR reaction mixture, which contained a primer/probe mixture and the ddPCR probe supermix with no dUTP. The primers used were:

The probes used were:

The ddPCR reactions were put on the Bio-Rad droplet generator and the resulting droplet-containing ddPCR mixtures were run on a Bio-Rad thermocycler with the following cycle parameters, with a ramp rate of 2°C/s for each step:

After thermocycling, the ddPCR reaction plate was transferred to the Bio-Rad droplet reader and the Quantasoft software was used to calculate the concentration of mtDNA^uaDf5 (FAM positive droplets) and mtDNA^WT (HEX positive droplets) in each well.

Confluent large plates were egg prepped and left to spin in M9 overnight for synchronization. The hatched L1s were plated onto large thick plates and allowed to grow to day 2 adults before being egg prepped a second time and left to spin in M9 overnight. The next morning, referred to as day 1 for lifespan determination, L1s were singled out onto small plates. Once the worms started laying eggs, they were transferred each day to a fresh small plate until egg laying ceased, after which the worms remained on the same plate unless bacterial contamination required transfer to a fresh plate. Worms were considered dead if there was no movement after being lightly prodded with a worm pick. Worms that died due to desiccation on the side of the plate were excluded from analysis.

Confluent large plates were egg prepped and left to spin in M9 overnight for synchronization. The hatched L1s were plated onto large thick plates and allowed to grow to day 2 adults before being egg prepped a second time and left to spin in M9 overnight. The next morning, L1s were singled out onto small plates. Once the worms started laying eggs, they were transferred each day to a fresh small plate until egg laying ceased. The day after transfer to a fresh plate, unhatched embryos and hatched larvae on the plate from the previous day were counted. This was done for each of the days of laying and the total of unhatched embryos and hatched larvae from all plates from a single worm were tabulated to determine total brood size. To determine embryonic lethality, the total number of unhatched embryos was divided by the total brood size. Worms that died due to desiccation on the side of the plate were excluded from analysis.

Confluent large plates were egg prepped and left to spin in M9 overnight for synchronization. The hatched L1s were plated onto large thick plates and allowed to grow to day 2 adults before being egg prepped a second time and left to spin in M9 overnight. The next morning, L1s were singled out onto small plates. The stage of the worms was assayed every 12 hr for the first 72 hr after plating. For determining the stage at 60 hr, L4 worms were divided up into three subgroups based on morphology: young-L4, mid-L4, and late-L4; otherwise, all other staged worms were not divided up into subgroups. Worms that died due to desiccation on the side of the plate were excluded from analysis.

DNA was collected from reference strain (N2) and two uaDf5-containing strains, LB138 and JR3630. Mitochondrial DNA was extracted and the DNA libraries were prepared using Nextera Kit and then sequenced using an Illumina NextSeq500. Prior to alignment, reads from fastq files were trimmed using Trimmomatic. Trimmed, pair-end reads (2 × 150) were then mapped to the C. elegans assembly reference sequence WBcel235 using Burrows-Wheeler Aligner (BWA) (Li and Durbin, 2009). Picard Tools (http://broadinstitute.github.io/picard/) was used to mark duplicate reads, and SAMtools (Li et al., 2009) was used to merge, index, and create pile-up format. VarScan (Koboldt et al., 2009) was used to call variants, and only variants with minimum coverage of 100 and a minimum variant frequency call of 0.01 were considered for analysis.

Summary statistics, analysis of variance, Mann–Whitney tests, and linear regression were calculated using R v3.4.1. The details of the statistical tests are reported in the figure legends.