Complex Transmission Patterns and Age-Related Dynamics of a Selfish mtDNA Deletion

Abstract

Despite wide-ranging implications of selfish mitochondrial DNA (mtDNA) elements for human disease and topics in evolutionary biology (e.g., speciation), the forces controlling their formation, age-related accumulation, and offspring transmission remain largely unknown. Selfish mtDNA poses a significant challenge to genome integrity, mitochondrial function, and organismal fitness. For instance, numerous human diseases are associated with mtDNA mutations; however, few genetic systems can simultaneously represent pathogenic mitochondrial genome evolution and inheritance. The nematode Caenorhabditis briggsae is one such system. Natural C. briggsae isolates harbor varying levels of a large-scale deletion affecting the mitochondrial nduo-5 gene, termed nad5Δ. A subset of these isolates contains putative compensatory mutations that may reduce the risk of deletion formation. We studied the dynamics of nad5Δ heteroplasmy levels during animal development and transmission from mothers to offspring in genetically diverse C. briggsae natural isolates. Results support previous work demonstrating that nad5Δ is a selfish element and that heteroplasmy levels of this deletion can be quite plastic, exhibiting high degrees of inter-family variability and divergence between generations. The latter is consistent with a mitochondrial bottleneck effect, and contrasts with previous findings from a laboratory-derived model uaDf5 mtDNA deletion in C. elegans. However, we also found evidence for among-isolate differences in the ability to limit nad5Δ accumulation, the pattern of which suggested that forces other than the compensatory mutations are important in protecting individuals and populations from rampant mtDNA deletion expansion over short time scales.

Introduction

With efficient natural selection, deleterious mutations are expected to be eliminated from populations. However, “selfish” or “parasitic” genetic elements such as segregation distorters exhibit transmission advantages and long-term persistence or accumulation in populations despite their negative effects on organismal fitness (Hurst and Werren 2001). Previous research in a variety of organisms has shown that mitochondrial DNA (mtDNA) genomes bearing large function-disrupting deletions can demonstrate such selfish evolutionary dynamics under certain conditions (e.g., Ma and O’Farrell 2016; Russell et al. 2018). This behavior may result from faster replication times that allow them to outcompete intact genomes, or from replication-dependent repair processes (Phillips et al. 2017).

Single large-scale mtDNA deletions account for a quarter of known mitochondrial disorders (Chinnery et al. 2000; Pitceathly et al. 2012). Such mutations can occur sporadically during early development and accumulate in affected tissues, their frequencies often increasing with age (Stewart and Chinnery 2015). Disease-causing mtDNA deletions are always found in a heteroplasmic state—both mutant and wildtype mtDNAs present within an individual—with biochemical defects manifesting beyond a certain heteroplasmy threshold (Moraes et al. 1999; Vielhaber et al. 2000; Gomez-Gomez 2017). mtDNA mutations can also be inherited (usually maternally) (Brown et al. 2006), although inheritance of large-scale deletions appears to be less common (Poulton et al. 1991), likely due to increased selection acting on deletions compared to other mutation types.

The mechanism(s) by which mtDNA deletions arise is still debated, but deleted sites are often flanked by short direct repeats, suggesting that they may result from replication slippage and subsequent elimination of intervening sequence. In nematodes, repeat-associated deletions have been attributed to intragenomic recombination events (Lunt and Hyman 1997). Alternatively, they may arise during mtDNA repair (Krishnan et al. 2008). The degree of sequence similarity and length of these repeats are important determinants of deletion formation. For instance, Albertini et al. (1982) demonstrated that altering a single base pair within a flanking repeat reduced the frequency of large deletion formation in Escherichia coli by an order of magnitude, and that shorter repeats with greater sequence homology were more predictive of deletion formation than longer repeats with less sequence homology (Mita et al. 1990).

In humans, the 4977-bp “common” mtDNA deletion occurs between the ATPase8 and ND5 genes where it is flanked by 13-bp direct repeats (Schon et al. 1989). The deletion is heteroplasmic, with clinical manifestations (i.e., several cancers and other diseases) at levels exceeding a critical threshold; it is also known to accumulate with age (Dani et al. 2003; Chen et al. 2011; Nie et al. 2013; Wallace and Chalkia 2013). This deletion is typically associated with sporadic disease phenotypes resulting from accumulation in somatic tissues; however, it has also been discovered in unfertilized human oocytes (Hsieh et al. 2002; Chan et al. 2005), suggesting that it may be occasionally inherited and/or selectively eliminated prior to fertilization. The deletion appears to experience greater selective pressure in highly proliferative tumor tissue, whereas greater accumulation is permitted in tissues with lower proliferative rates (Dani et al. 2003). Clearly, multiple forces, including ongoing deletion formation and selection operating at both individual and intra-individual levels, likely interact to determine natural mtDNA deletion frequencies.

Because few experimental models of heritable mtDNA deletions exist, much of our knowledge of such mutations still relies on human pedigrees. An exception comes from studies of Caenorhabditis briggsae nematodes (Howe and Denver 2008; Estes et al. 2011; Clark et al. 2012; Hicks et al. 2012; Phillips et al. 2015). Natural C. briggsae populations harbor varying levels of heteroplasmic deletion that eliminates 786 base pairs and dozens of highly conserved codons of the NADH-dehydrogenase-5 (nduo-5) gene, which encodes an integral component of mitochondrial electron transport chain Complex I. Importantly, the deletion arises sporadically and is also inherited maternally, making it a good model for both somatic accumulation and germline inheritance of mtDNA deletions. The deletion, originally termed nad5Δ, is thought to repeatedly form as a result of mispairing between 21-bp direct repeats present within nduo-5 and an upstream pseudogene element, Ψnad5-2, which is present in certain isolates. The pseudogene is homologous to nduo-5 and likely arose from an intra- or inter-genomic mtDNA recombination event (Raboin et al. 2009). We have found that nad5Δ conforms to the definition of a selfish element in that it increases in frequency across nematode generations under genetic drift (Howe and Denver 2008; Clark et al. 2012; Phillips et al. 2015) despite the fact that high nad5Δ heteroplasmy levels are associated with reduced isolate-specific fecundity and other phenotypes (Baer et al. 2005; Estes et al. 2011). The nad5Δ deletion exists at average isolate-specific levels ranging from 0% to ∼50%, with a maximum heteroplasmic threshold of ∼60% within adult individuals (Fig. 3 in Howe and Denver 2008); levels greater than ∼60% are presumed to be lethal (Howe et al. 2010).

Fig. 3.

Fold changes in intact nduo-5 in C. briggsae offspring relative to parental soma and germline samples. Mean log₂ fold changes of intact mtDNA nduo-5 abundance between (A) parental somatic tissues or (B) parental germline tissues and eight offspring from these parents for C− AF16 (top) and C+ PB800 (bottom) isolates. Error bars = 95% CI. For ease of visualization, families are arranged by mean log₂ fold change ratio for each tissue type, and thus cannot be compared between somatic and germline datasets. Ranges of mean relative fold changes for the somatic dataset (A) = –0.22 to 0.51 for AF16 and –0.78 to 0.77 for PB800. Ranges of mean relative fold changes for the germline data set (B) = –0.22 to 1.14 for AF16 families and –0.42 to 1.23 for PB800 families. Tukey’s HSD tests revealed multiple statistically distinguishable family groups for both isolates in each dataset (Supplementary Tables S4 and S5).

Importantly, certain C. briggsae isolates contain putative compensatory mutations (referred to hereafter as C+ isolates) that reduce the similarity of the flanking repeats, and are hypothesized to limit nad5Δ-formation events compared to isolates lacking these mutations (C− isolates). Specifically, a more recently diverged group (Clade II; Fig. 2 in Howe and Denver 2008), which includes natural isolates EG4181 and PB800, has evolved such mutations. Consistent with this idea, these C+ isolates exhibit lower overall nad5Δ accumulation across multiple generations of evolution under genetic drift conditions as compared to C− isolates (Clark et al. 2012; Phillips et al. 2015). However, it remains unclear how nad5Δ frequency changes across a single generation under typical population genetic circumstances wherein the contribution of novel deletions to the total deletion load is expected to be less than in multigenerational drift studies.

Fig. 2.

Fold changes in intact nduo-5 for C. briggsae offspring relative to parents. Mean log₂-transformed fold changes of intact mtDNA nduo-5 abundance between parents and 16 of their offspring for C− AF16 (top) and C+ PB800 (bottom) isolates. Error bars = 95% CI. A value of 0 indicates that offspring have equal proportions of intact mtDNA genomes (and, by extension, nad5Δ-bearing genomes) compared to their parent. Values >0 indicate an increase in intact mtDNA genomes (and an equivalent decrease in nad5Δ-bearing genomes) in offspring compared to parents; values <1 indicate the opposite scenario. Separate ANOVA performed on data from each isolate revealed substantial among-family variation in log₂ fold change: AF16 (F19 = 198.05, P < 0.0001); PB800 (F19 = 116.90, P < 0.0001). Tukey’s HSD tests revealed multiple statistically distinguishable family groups for both isolates (Supplementary Table S3)

It is similarly unclear how nad5Δ frequency changes with nematode age and whether the presence of compensatory mutations will influence these patterns.

To address these outstanding questions and further develop C. briggsae as an experimental model of mtDNA deletion dynamics, we quantified mtDNA deletion levels during development of C+ and C− C. briggsae isolates as well as changes during transmission from parent to offspring. The latter was achieved by tracking nad5Δ deletion levels in both whole-animal samples and tissue dissections to discern patterns of heteroplasmy in somatic versus germline tissues.

Methods

Nematode strains and whole-worm sampling

This study utilized five natural isolates of C. briggsae: AF16, ED3034, EG4181, HK105, and PB800, selected for their natural variation in nad5Δ heteroplasmy levels (Fig. 3B in Howe and Denver 2008). C+ isolates, PB800 and EG4181, were specifically selected due to their putative compensatory mutations. HK105 was chosen for its especially high nad5Δ level. Stocks were thawed and maintained at 20°C on 60-mm plates containing Nematode Growth Medium Lite (NGM-Lite; US Biological Life Sciences) treated with 200 mg/mL streptomycin sulfate (Sigma-Aldrich, St. Louis, Missouri), with a lawn of strain OP50-1 E. coli as the sole food source. Isolates were allowed to recover from freezing for two to three generations prior to sampling.

For assays of nad5Δ levels across developmental time, individual nematodes from AF16, ED3034, HK105, and PB800 were sampled from a single, age-synchronous population at each major developmental stage, including L1 (first larval), L2, L3, L4, and young adult (target n = 64 per life stage per isolate for a total of 320 sampled individuals) and transferred into 15 µL worm lysis buffer (50 mM KCl, 2.5 mM MgCl₂, 10 mM Tris, 0.45% Tween 20, 0.05% gelatin, and 60 µg/mL Proteinase K in DNase- and RNase-free H₂O) and stored at –80°C prior to DNA extraction (Supplementary Fig. S1). For inheritance assays, samples of 20 reproductive hermaphrodites of AF16, ED3034, and PB800, 10 from HK105 and EG4181, and 16 of their respective L1 offspring were collected via the same method (Supplementary Fig. S2). For deletion inheritance assays using qPCR, 20 L4 hermaphrodites of AF16 and PB800 were picked individually to new plates, allowed to lay embryos for 12 h before being transferred to fresh plates. Twelve hours later, mothers and hatched L1 offspring (16 from each mother) were picked individually into worm lysis buffer.

Sample preparation

For all samples, DNA was extracted via a freeze–thaw procedure to break the worm cuticle followed by processing in a Thermo Electron Corporation Px2 Thermal Cycler using the following program: 60°C for 1.5 h, 95°C for 15 min, cool to holding temperature of 10°C. Samples were stored at –80°C prior to PCR and qPCR assays.

PCR analysis of whole-worm nad5Δ abundance

Standard PCR reactions were performed on all sets of whole-animal samples (see Supplementary File S1 for details). The assay involves two PCR primers (Cb_mt1F positioned in the nduo-3 gene, and 58 R or 38 R in the nduo-5 gene) that flank the Ψnad5-2 pseudogene element and its associated deletion products (Supplementary File S1), generating a possibility of two differently sized bands that indicate presence or absence of the canonical deletion (Fig. 2 in Clark et al. 2012). PCR products were electrophoresed on standard 1.5% agarose gels and digitally recorded. Gel images were manually scored according to three observed banding categories, following the methods of Howe and Denver (2008), in which 1 = large band only, 2 = both bands, and 3 = small band only. These banding patterns were previously correlated with estimates of deletion-bearing genome proportions obtained from qPCR by Howe and Denver (2008) (see also Clark et al. 2012), indicating that this standard PCR-based assay can serve to estimate approximate heteroplasmy frequency of this region. Specifically, banding patterns 1, 2, and 3 were found to correspond to approximate nad5Δ heteroplasmy frequencies of 0–20%, 20–45%, and >45%, respectively.

qPCR analysis of whole-worm heteroplasmy

Quantitative real-time PCR was employed on the second set of parent–L1 offspring samples to evaluate relative fold change of intact nduo-5 in AF16 and PB800 offspring compared to parents. DNA was extracted as described above, quantified within each sample using a NanoDrop 2000 Spectrophotometer (Thermo Scientific), and then diluted to 1 ng/µL. For sample assays, each hermaphrodite parent and its corresponding offspring were assayed together on a single plate. 5 µL of the diluted template DNA was added to each 20 µL total volume reaction. All reactions also contained 3.4 µL molecular grade DNase- and RNase-free H₂O, 10 µL SsoAdvanced SYBR Green Supermix (Bio-Rad), and 0.8 µL each of the appropriate forward and reverse primers diluted to 5 µM. Each sample was assayed in duplicate using two distinct primer sets (Supplementary File S1). The first qPCR primer set (mdND5) amplified a nduo-5 gene region present only in intact genomes, and the second set (SrRNA) amplified a 16S ribosomal RNA gene region present in both deletion-bearing and intact genomes for which there is no evidence of heteroplasmy (Howe and Denver 2008). All qPCR reactions were set up in clear 96-well semi-skirted plates with optical flat caps and were conducted using a Stratagene Mx3005P thermocycler (Agilent Technologies) using the “Quantitative PCR” experiment type within the MxPro software. The qPCR reaction profile began with an initial denaturation step of 95°C for 2 min, followed by 40 cycles of denaturing at 95°C for 1 min, primer annealing at 60°C for 30 s, and sequence extension at 70°C for 35 s, followed by the standard default melt curve built into the software. Fluorescence readings (492 nm excitation, 516 nm emission) were taken at the end of each elongation step. Quantification cycle thresholds were set automatically in Agilent MxPro software using adaptive baseline, moving average, and amplification-cased threshold settings. The resulting C_t values for each sample were used to calculate fold changes in the abundance of intact nduo-5 gene in offspring relative to parent samples.

To demonstrate primer efficiencies, two independent standard curve assays were conducted for each isolate using known concentrations of template DNA prepared from samples of mixed hermaphrodite parents and their offspring—one using the nduo-5 primer pair and another using the 16S primer pair. Each of five dilutions (10, 5, 2.5, 1.25, and 0.625 ng) were processed in duplicate reactions per primer pair, and standard curves were produced by the Agilent MxPro software. R² for each curve exceeded 0.99 (Supplementary Fig. S3). Assays were validated by visualizing qPCR end products via gel electrophoresis in addition to observation of a single melting dissociation curve.

Gonadal dissection and qPCR analysis of somatic and germline heteroplasmy

To evaluate whether parent–offspring nad5Δ level relationships would be more strongly positively correlated if offspring deletion levels were compared to those in maternal gonadal tissue, where maternal germ cell mitochondria are expected to outnumber those of sperm or somatic cells (Lemire 2005), we performed tissue dissections on 10 PB800 and AF16 animals. After being allowed to lay embryos for 12 h, these reproductive hermaphrodites were dissected using two sterile 18-gauge hypodermic syringe needles to cut the worm beneath the head, causing the gonads to extrude. Samples of gonadal and somatic (worm anterior region, including the pharynx) tissues were collected separately (Supplementary Fig. S4). Additionally, eight L1 offspring from each hermaphrodite were picked individually into 15 µL of worm lysis buffer. DNA extraction and qPCR were conducted as described previously except that two families were assayed together on a single plate.

Data analysis

For PCR assays of nad5Δ level across stages of nematode development, a nominal logistic regression was performed to test the model: banding pattern = μ + strain + life stage + (strain × life stage) + ɛ. For PCR assays of nad5Δ inheritance, raw nad5Δ proportion data were analyzed using least-squares linear regressions. We further compared banding score frequencies between various life stages using Cochran–Armitage tests where deletion banding score (1, 2, or 3) was the ordinal variable.

For qPCR experiments, fold differences in nduo-5 or 16S abundance in offspring were calculated relative to maternal samples using the efficiency corrected ddC_q method (Pfaffl 2001). C_t values from the two technical replicates for each sample–primer pair were averaged. Extreme outliers resulting from failed amplification curves were removed from the data set. Primer efficiencies for all pairs were demonstrated to fall between 95.5% and 100% (Supplementary Fig. S3). We therefore employed a 2^ΔΔCt calculation using equation 3 in Pfaffl (2001) with 16S as reference and nduo-5 as target, and with E = 2. This calculation yielded fold change ratios of nduo-5 abundance in offspring compared to parents. These values were log₂ transformed prior to statistical analysis using nested ANOVA. We tested the model y = μ + isolate + family(isolate) + ɛ for each qPCR dataset. One-way ANOVA followed by Tukey’s HSD multiple comparison tests were used to quantify among-family variation within each isolate.

Results

Age-related nad5Δ dynamics

The final sample sizes for the gel banding assays and banding score modes for offspring from each isolate are provided in Supplementary Table S1. Certain sample groups (e.g., PB800 L2 samples) experienced more failed PCRs than others. Variance in banding pattern was primarily explained by isolate (χ²₃ = 978.1) and, to a lesser extent, by life stage (χ²₄ = 70.63) and their interaction (χ²₁₂ = 64.83); P < 0.0001 in each case. Comparing banding patterns between L1 and young adult samples, all isolates except for the C+ PB800 exhibited statistically significant increases in deletion-bearing genomes (Cochran–Armitage exact tests; P < 0.001). For instance, HK105 animals exhibited a mixture of banding types at L1, which became dominated by deletion-bearing types during later stages of development, while PB800’s nad5Δ levels were not detected at L1 and increased slowly across development compared to other isolates (Fig. 1). Only AF16 showed significantly increased nad5Δ heteroplasmy between L2 and L3/L4 stages (Cochran–Armitage test, z = –4.37; P < 0.0001), coinciding with the major period of mitochondrial expansion that occurs during nematode germline development (Lemire 2005). This increase was followed by a marked decrease in AF16 young adult worms (Cochran–Armitage test, z = 4.41, P < 0.0001). In summary, while nad5Δ proportions increased between L1 and young adult stages for isolates other than PB800, there was considerable variation in isolate-specific frequencies between each developmental stage.

Fig. 1.

Banding score proportions by life stage for each isolate. Proportion (left y axis) of each banding score (right y axis) across developmental life stage from L1 to young adult (YA) for four C. briggsae isolates. Lightest shades represent a score of 1 (intact mtDNA only), medium shades a score of 2 (heteroplasmic), and the darkest shades a score of 3 (deletion only). Column width indicates relative sample size across life stages

Parent–offspring nad5Δ transmission patterns

The final sample sizes for the transmission gel banding assays and banding score modes for parents and offspring from each isolate are provided in Supplementary Table S2. Once again, certain sample groups (e.g., ED3034 offspring samples) experienced more failed PCRs than others. As with the dataset described above, the baseline nad5Δ levels measured for each isolate generally agreed with previous findings (Howe and Denver 2008; Clark et al. 2012). Most C− HK105 parents had banding scores of 3 (>45% deletion according to Howe and Denver 2008); parent samples from the other four isolates exhibited variable proportions of banding scores 1 and 2 (Supplementary Fig. S5). Across all isolates, offspring nad5Δ levels showed a weak tendency to equal those of their mothers, but there was considerable plasticity in offspring nad5Δ levels with C− ED3034 parents of banding Type 1 having the strongest tendency to produce offspring with higher deletion banding scores than themselves (Supplementary Fig. S6). For instance, the 11 ED3034 parents with banding scores of 1 (large band only) produced offspring of Type 1 (32%), 2 (55%), or 3 (13%), while the five parents with banding score 2 (both bands) produced offspring of Type 1 (48%) or 2 (52%) only. C+ isolate, PB800, exhibited the lowest rates of nad5Δ offspring transmission. Indeed, all PB800 parents with banding Type 2 only produced offspring of wildtype banding Type 1. Conversely, parents of the lowest nad5Δ banding level categories measured for the other isolates, including the other C+ isolate EG4181, always produced a fraction of offspring with higher nad5Δ levels than themselves.

Quantitative real-time PCR methods were employed on a second set of parent–L1 offspring from two isolates to assess relative abundance of nad5Δ across inheritance in whole-animal samples. The final sample sizes alongside summary statistics for each family are provided in Supplementary Table S3. Mean C_t values for nduo-5 and 16S ±1 SEM were 22.6 (0.04) and 22.2 (0.04), respectively. Variation between technical replicates was low, with average standard deviations of 0.12 and average coefficients of variation of 0.005 for both primer sets. We found that log₂ fold changes spanned a similar range (between ±1) for each isolate (Fig. 2), corresponding to fold change ratios between ±2. Thus, mtDNA heteroplasmy levels diverged in both directions from parental levels for each isolate. Despite this similarity, the two isolates differed from each other in a manner consistent with the PCR assays. Variance in relative fold change was primarily explained by isolate (F₁ = 845.92, P < 0.0001) and, to a lesser extent, by family(isolate) (F₃₈ = 156.85, P < 0.0001). For AF16, mean log₂ fold change was –0.12, 95% confidence intervals (CI) (–0.18 to –0.06)—equivalent to a fold change ratio of 0.92 or an 8% reduction in intact mtDNA genomes (and equivalent increase in deletion-bearing genomes) in offspring compared to parents. For PB800, mean log₂ fold change was 0.26, 95% CI (0.19 to 0.32)—equivalent to a fold-change ratio of 1.2, or a 20% increase in intact mtDNA genomes (and an equivalent reduction in deletion-bearing genomes) between generations. Tukey’s HSD multiple comparison tests showed the existence of several statistically distinguishable family groups within each isolate (Supplementary Table S3). Twelve of the 20 PB800 families exhibited positive fold changes whose 95% CI did not overlap 0; many were very weakly positive, while several had more strongly positive values (Fig. 2). While AF16 also produced a few families with positive fold changes, 9/20 families exhibited negative fold change values; two were less than –1.0, corresponding to a >50% reduction in intact mtDNA genomes in offspring compared to the parents. By contrast, only 3/20 PB800 families exhibited negative values. It is important to note that our experiment had an unknown level of power to accurately detect low-fold changes; however, the fact that the number of target copies of nduo-5 should be high (likely never <40% of mtDNA genomes within an individual), alongside the observed low technical variation in our experiment, warrants a reasonable level of confidence in these results.

Offspring deletion heteroplasmy compared to that in maternal germline and soma

To understand whether offspring nad5Δ levels would correlate more strongly with those measured for parental germline tissue than with levels in somatic or whole tissues, qPCR assays were conducted on dissected tissues from AF16 and PB800 parents and their whole-animal offspring. Supplementary Tables S4 and S5 show final sample sizes alongside summary statistics for each family. We again measured log₂ fold change of intact nduo-5 in offspring compared to their parent, this time using parental gonadal or somatic parental samples for comparison. There was more between-isolate divergence in the ranges of relative fold change values for these datasets (Fig. 3) than for the whole-animal data reported above (Fig. 2). For the dataset utilizing parent gonadal samples, relative fold change was affected by isolate (F₁₈ = 39.24, P < 0.0001), with AF16 having slightly more positive values on average (0.26 as compared to 0.14 for PB800), and by family(isolate) (F₁= 12.27, P < 0.0006). For that utilizing parental somatic samples, there was no effect of isolate but a significant effect of family(isolate) (F₁ = 27.37, P < 0.0001). Thus, the isolate-specific differences we observed for the previous whole-animal dataset either reversed (Fig. 3A) or disappeared (Fig. 3B) here, although our sample size and statistical power was much reduced in these datasets. Overall, relative fold change was more positive for comparisons involving the maternal germline samples than those involving maternal somatic samples. For the germline dataset, mean log₂ fold change was 0.20, 95% CI (0.12 to 0.28), equivalent to a fold change ratio of 1.15, or a 15% increase in intact mtDNA genomes in offspring compared to the parental germline. For the somatic dataset, mean log₂ fold change was 0.012, 95% CI (–0.054 to 0.078), equivalent to a ∼0% fold change ratio.

Discussion

Variation in nad5Δ proportion across nematode development

The L1-stage banding patterns measured here were consistent with those reported previously (Howe and Denver 2008; Clark et al. 2012). We observed a good deal of variation in banding scores between different nematode life stages (Fig. 1), which we acknowledge may be partly due to stochastic sampling. In agreement with the hypothesis that nad5Δ is a selfish element, however, heteroplasmy levels increased between L1 and young adult stages in all isolates tested—except for the single C+ isolate, PB800, for which initial deletion proportion was also the lowest. Although these results are consistent with the divergent sequences present within PB800 having provided protection against accumulation of nad5Δ, this remains unclear since no other C+ isolates were tested in this experiment. Only one isolate, AF16, experienced a statistically significant increase in nad5Δ level coincident with the L3–L4 mitochondrial expansion (Lemire 2005)—against our initial prediction that all isolates would exhibit such an increase. The absence of sharp increases in deletion level at this critical stage may be consistent with purifying selection operating to maintain germline nad5Δ frequencies below critical thresholds, and suggest that selection occurs either postzygotically (i.e., embryos with high nad5Δ load fail to hatch) or prior to fertilization against nad5Δ-containing germ cells or gametes.

Isolate-specific differences in parent–offspring heteroplasmy relationships

Clark et al. (2012) employed the same gel banding assays utilized here to track nad5Δ levels across a 10-generation interval in C. briggsae isolates evolving under genetic drift. This study found that nad5Δ levels increased during this period in isolates containing the Ψnad5-2 pseudogene element and associated direct repeats, and particularly in those lacking putative compensatory mutations (i.e., C− isolates) (Fig. 3 in Clark et al. 2012). By tracking this deletion across a single generation under benign conditions, we found a different pattern: offspring nad5Δ levels showed a weak tendency to equal those of their parent, but diverged in both upward and downward directions for all isolates except for C+ PB800 (Supplementary Fig. S6). This is consistent with a mitochondrial genetic bottleneck effect, that is, mtDNA sampling during maternal germline development (e.g., Taylor and Turnbull 2005; Floros et al. 2018), perhaps followed by enhanced replication of deletion-bearing mtDNA genomes in some germ cells. Conversely, the results from PB800 indicate that any replication advantage nad5Δ may have had during maternal germline development (Fig. 1), or any appearances of novel nad5Δ elements, were at least partially opposed by another force during mtDNA transmission to offspring in this isolate.

Similarly, the more sensitive qPCR assays also showed that, while considerable divergence in mtDNA heteroplasmy levels between mothers and offspring was possible, isolates differed in their propensity for heteroplasmy expansion or contraction (Fig. 2, Supplementary Table S3). In agreement with the PCR assays, many families from the C− AF16 isolate exhibited negative fold changes, meaning that broods had fewer wildtype and more deletion-bearing mtDNAs than their mothers. This observation is consistent with either or both: (1) more frequent nad5Δ transmission in AF16 (i.e., relaxed selection on nad5Δ in the maternal germline) and (2) more frequent deletion formation during early embryogenesis in AF16 compared to PB800. Support for both possibilities comes from the previous finding that replicate lines of AF16, but not those of PB800, were especially prone to accumulating both higher levels of nad5Δ and new mtDNA deletions when evolved under drift conditions (i.e., in effective population sizes of 1 and 10) (Fig. 5 in Phillips et al. 2015). Conversely, the tendency of C+ PB800 families toward positive fold change values is consistent with less frequent nad5Δ transmission in PB800 and/or less frequent deletion formation compared to AF16—with the latter due to its compensatory sequence change.

That offspring of C+ isolate EG4181 were not protected from nad5Δ expansion in a manner similar to those of C+ PB800 (Supplementary Fig. S6) means that the presence or absence of compensatory sequence changes cannot be the primary determinant of nad5Δ transmission frequency. Because the putative role of the compensatory sequences is to reduce the likelihood of novel deletion formation, the existence of other forces to limit the transmission of preexisting deletions would not be surprising. The nature of any such force that might be unique to PB800 remains unclear, but could involve enhanced post-zygotic selection and/or action of mitochondrial quality control mechanisms operating within the maternal germline (see Palozzi et al. 2018 for a review). For instance, improved removal of faulty mitochondria through mitophagy (Twig et al. 2008; Mishra and Chan 2014) or during the normal period of “physiological apoptosis” that occurs during hermaphrodite gonadal development could account for the lower nad5Δ transmission rates observed for PB800. The latter process was recently shown to eliminate aberrant (i.e., multinucleate) germ cells in C. elegans (Raiders et al. 2018). Morales and Greenstein (2018) postulated that this process could also serve to selectively eliminate damaged mitochondria on the basis of morphology, known to be predictive of mitochondrial health (Ishihara et al. 2003; Okamoto and Shaw 2005; Hicks et al. 2012). Alternatively, Floros et al. (2018) found that a mtDNA bottleneck alongside a metabolic shift from glycolytic to oxidative metabolism likely exposes deleterious mtDNA mutations to selection early during human germ cell development, and Lieber et al. (2019) showed that selective removal of deleterious mtDNA genomes is promoted by organelle fragmentation in Drosophila melanogaster.

That offspring nad5Δ levels were able to diverge significantly from maternal levels in our experiments stands in stark contrast to results from similar experiments with the C. elegans uaDf5 deletion—a large-scale mtDNA deletion that removes 11 genes and exists at heteroplasmic frequencies of ∼20–80% (Tsang and Lemire 2002; Lemire 2005). uaDf5 frequencies evaluated by qPCR exhibit a near 1:1 relationship between parents and offspring (Liau et al. 2007). This difference in pattern may reflect a difference in the size of the mitochondrial bottleneck between species (i.e., indicate a tighter bottleneck promoting more rapid segregation of heteroplasmy in C. briggsae). However, Ahier et al. (2018) recently demonstrated that, although uaDf5 was more prone to accumulating in the C. elegans germline than in the soma, deletions levels were reduced in unfertilized oocytes compared to germline tissue—consistent with the action of purifying selection prior to oocyte maturation. It may also suggest that different rules apply to different types of deletions. For instance, uaDf5 may be an imperfect representative of natural mtDNA deletion dynamics as it was isolated from a chemical mutagenesis screen and has thus had little opportunity to evolve diverse compensatory mechanisms to limit its transmission. In support of this notion, the deletion induces the mitochondrial unfolded protein response (UPR^mt), which appears to protect uaDf5-containing mitochondria from selective removal by mitophagy (Lin et al. 2016). It would be interesting to know whether a different pattern exists within PB800 C. briggsae.

nad5Δ heteroplasmy in germline versus soma

Analysis of relative fold changes involving a smaller number of maternal germline and somatic tissue comparisons (Fig. 3) showed that while PB800 and AF16 mothers can produce broods with substantially higher or lower mtDNA heteroplasmy levels than themselves, many broods approximately match the heteroplasmy level of their mother. The range and variance in heteroplasmy levels measured in gonadal and somatic tissues was contracted compared to that in whole animals—a difference that may point to variance in nad5Δ levels in an untested parental tissue type (e.g., gut) present within the whole-worm maternal samples, and indicate that critical heteroplasmy thresholds differ among tissue types. Furthermore, relative fold change was more positive for comparisons involving the maternal germline samples than those involving maternal somatic samples, consistent with the existence of selective forces to check deletion expansion in offspring—again in contrast to results from C. elegans uaDf5 (Ahier et al. 2018). Alternatively, the difference in pattern between tissue types could merely be due to the reduction in sampling effort for these datasets. Even if real, the biological significance of this difference, amounting to a 15% reduction in average heteroplasmy in the germline compared to the soma, is unclear. Future work to employ ddPCR for absolute quantification of nad5Δ levels alongside functional assays would improve our understanding of how altered mtDNA deletion heteroplasmy levels translate into fitness consequences for the organism.

Conclusions

This study provided the first comprehensive assessment of transmission dynamics of a naturally-occurring mtDNA deletion across developmental and generational timescales. Quantification of nad5Δ heteroplasmy levels in parental individuals and across animal development provided support for the selfish nature of nad5Δ (Howe and Denver 2008; Howe et al. 2010; Clark et al. 2012; Phillips et al. 2015). However, our comparisons of parent and offspring heteroplasmy levels revealed new information suggesting that both genetic drift and purifying selection shape mtDNA heteroplasmy between generations. A caveat to our findings is that nuclear and mitochondrial genetic variation beyond nad5Δ is present and unaccounted for within the C. briggsae isolates tested here; thus, results are almost certainly confounded by effects of background genetic variation and phylogenetic relationship. Indeed, nuclear DNA sequence divergence between C+ isolates PB800 and ED3034 is likely to be at the heart of the observed differences in deletion transmission between these isolates, and suggests a potential role for mitonuclear interaction in defining mtDNA deletion dynamics. Future work to generate inbred lines of C. briggsae natural isolates with high or low heteroplasmy levels (Dolgin et al. 2007), or cytoplasmic-nuclear hybrid (cybrid) strains with varying nad5Δ heteroplasmy levels on uniform genetic backgrounds (Estes et al. 2011), could assist in disentangling the effects of nad5Δ variation among isolates from the effects of among-isolate nuclear variation.

Additional limitations to our study may result from using primers specific to the canonical nad5Δ deletion; we may be missing other deletions affecting the same region (Fig. 3 in Phillips et al. 2015). However, such deletions were observed previously only within C. briggsae populations evolved for multiple generations under strong drift conditions (Fig. 5 in Phillips et al. 2015; see also Howe et al. 2010). Furthermore, as the inter-family variability in heteroplasmy levels was substantial, assaying a larger group of parents and including more C+ and C− isolates might reveal a stronger signal. Finally, although we suspect that the contribution of novel deletions was low or negligible in our experiment, additional work is also needed to distinguish de novo occurrences from clonal expansion of nad5Δ; this would necessitate measuring deletion frequency among progeny of parents lacking the deletion.

Taken together, the results of this study suggest that nad5Δ frequencies in natural populations are likely defined by a complex balance between recurrent deletion formation as well as genetic drift (e.g., an mtDNA bottleneck effect) and selection occurring at both intra-individual and individual levels, all of which may be influenced by mitonuclear interplay. Specifically, we hypothesize that nad5Δ, promoted by the flanking repeat sequences, forms at some rate that is limited by the presence of the compensatory mutations in C+ isolates, and that nad5Δ levels are allowed to diverge during transmission and within the soma until counteracted by (in some cases, isolate-specific) selective mechanisms that limit accumulation beyond threshold levels.