Mitochondrial mutations in Caenorhabditis elegans show signatures of oxidative damage and an AT-bias

Abstract

Rapid mutation rates are typical of mitochondrial genomes (mtDNAs) in animals, but it is not clear why. The difficulty of obtaining measurements of mtDNA mutation that are not biased by natural selection has stymied efforts to distinguish between competing hypotheses about the causes of high mtDNA mutation rates. Several studies which have measured mtDNA mutations in nematodes have yielded small datasets with conflicting conclusions about the relative abundance of different substitution classes (i.e., the mutation spectrum). We therefore leveraged Duplex Sequencing, a high-fidelity DNA sequencing technique, to characterize de novo mtDNA mutations in Caenorhabditis elegans. This approach detected nearly an order of magnitude more mtDNA mutations than documented in any previous nematode mutation study. Despite an existing extreme AT bias in the C. elegans mtDNA (75.6% AT), we found that a significant majority of mutations increase genomic AT content. Compared to some prior studies in nematodes and other animals, the mutation spectrum reported here contains an abundance of CG→AT transversions, supporting the hypothesis that oxidative damage may be a driver of mtDNA mutations in nematodes. Furthermore, we found an excess of G→T and C→T changes on the coding DNA strand relative to the template strand, consistent with increased exposure to oxidative damage. Analysis of the distribution of mutations across the mtDNA revealed significant variation among protein-coding genes and as well as among neighboring nucleotides. This high-resolution view of mitochondrial mutations in C. elegans highlights the value of this system for understanding relationships among oxidative damage, replication error, and mtDNA mutation.

Keywords: Duplex Sequencing; cytosine deamination; low-frequency variant; metazoan mtDNA; mitochondrial mutation; mutation accumulation; mutation spectra; oxidative damage; oxidized guanine; replication error.

Figure 1

The C. elegans mtDNA mutation spectrum is dominated by mutations that increase AT content and exhibits strand asymmetry. (A) Variation in the frequency of mutations across six substitution classes. CG→AT transversions and CG→TA transitions are the most abundant substitution types. Each point on the plot represents one of the nine replicates assayed. SNV frequencies were calculated as the number of sites in a replicate with a mutation normalized by the coverage of the corresponding base pair type. For example, the CG→AT mutation frequency shows the CG→AT event count divided by GC coverage for each replicate. A floor was applied to frequencies below 3e-08, which approaches the error threshold of Duplex Sequencing (Wu et al. 2020). (B) Strand asymmetry of mutations in C. elegans mtDNA. Both CG→AT and CG→TA substitutions show significant strand asymmetry (one-way ANOVA, P-values noted on figure), with the G→T and C→T changes (underlined in red) occurring predominately on the forward (F) strand, which in C. elegans mtDNA is synonymous with the heavy-strand and for genic regions also the coding-strand. Mutation frequencies were calculated as the average of the nine replicates and were normalized by the sequencing coverage of each base type on the F-strand. For example, the G→T mutation frequency shows all G→T events on the F-strand divided by G coverage on the F-strand, and the C→A mutation frequency shows all C→A events on the F-strand divided by C coverage on the F-strand.

Figure 2

The distribution of mutations across the mitochondrial genome. (A) Map of the C. elegans mtDNA and summary of Duplex Sequencing data. The outermost track depicts the gene order and type, with protein-coding (CDS) genes shown in tan, rRNA genes shown in red, tRNA genes shown in blue, and intergenic regions shown in pink. The next track in from the outside depicts the total mutation counts in 50-bp windows, with deletions, insertions and SNVs colored differently according to the key at the top of the track. The next track in from the outside (red histogram) depicts the cumulative (sum of nine replicates) DCS coverage in 50-bp windows, with a scale bar included at the top of the track. The innermost track shows the relative fraction of each base type in 50-bp windows, with colors specified by the key in the figure center. The figure was generated with Circos v0.69-8 (Krzywinski et al. 2009). (B) Variation in SNV frequencies by genomic region. Note that low intergenic coverage resulted in the detection of only a single intergenic substitution (in replicate 3 b). The other replicates had intergenic substitution counts of zero, but also had extremely low intergenic coverage, making comparisons that include intergenic SNV frequency low powered (see main text). No significant variation is observed among CDS, tRNA, or rRNA genes when the intergenic region is excluded. (C) Significant variation in SNV frequencies across the 12 protein-coding genes was observed for only the two most common substitution classes (CG→AT transversions and CG→TA transitions). See Supplementary Figure S4 for the gene specific mutation frequencies of all substitution classes.

Figure 3

Variation in AT→NN (top panel) and CG→NN (bottom panel) mutation frequency across different trinucleotide contexts, where NN refers to any other base-pair. Significant variation was seen in the trinucleotide contexts for CG→AT transversions and CG→TA transitions (one-way ANOVA, P-values in figure legend).

Figure 4

Simulation of mutations to derive null expectation of the ratio of nonsynonymous: synonymous substitutions (NS: S). The observed ratio of NS: S (marked with the vertical red line) is 1.3-fold higher than the null expectation generated from 10,000 simulations where the observed number and type of coding sequence mutations were randomly mapped onto the coding sequence, though this difference is not significant. The two-tailed P-value was derived by dividing twice the number of simulated NS: S ratios with values greater than the observed NS: S ratio by the total number of simulations (378*2/10000).

Figure 5

Indels are predominately found at homopolymers in C. elegans mtDNA. Here, indels that were shared between replicates were treated as independent events. See Supplementary Figure S5 for counts that assume indels shared between replicates are the result of common ancestry. Indels that are expansions or contractions of existing homopolymers are shown in red, while those that are not associated with homopolymers are shown in tan.