Discussion
We have successfully assembled ∼14.1 kb of mitochondrial genome sequences for two additional species of terrestrial isopods, Trachelipus rathkei and Cylisticus convexus, using second-generation sequencing approaches. Two lines of evidence suggest that these assemblies are complete: (i) the termini of the mitochondrial monomer sequence in A. vulgare possess telomeric hairpin sequences, and these features are present in both of our assemblies; and (ii) these assemblies are both longer than the complete mitochondrial sequence in A. vulgare (13939 bp) (Doublet et al. 2013). Moreover, we find the full set of protein-coding and ribosomal RNA genes typically found in animal mitochondrial genomes. The relative ordering of these genes is identical between the two species (Figure 1). This is not surprising given that mitochondrial gene order is also conserved across most other isopods that have been studied, with a few exceptions (Kilpert et al. 2012).
Despite conservation in the arrangement of protein-coding and ribosomal RNA genes, the tRNA genes provide some surprising evolutionary insights. Most notably, we have discovered multiple heteroplasmic sites in tRNA genes that are also common to both species (Table 1, Table 2, Figure 2). Although one of these sites was previously known (Marcadé et al. 2007; Doublet et al. 2008), two are novel. The fact that all three of these shared heteroplasmies alter anticodon sequences, allowing these loci to function as dual tRNA genes, seems unlikely to be a chance occurrence. Although it is unclear exactly how long ago T. rathkei and C. convexus diverged, they are classified in different families, and the relatively low sequence identity of their mitochondrial genomes suggests they shared a common ancestor at least millions of years ago. Strong selection therefore seems likely to have maintained the crucial dual function of these genes for a very long time, and the atypical organization of the mitochondrial genome probably facilitates the evolution of such stable heteroplasmic sites. Conversely, selection to maintain the heteroplasmic sites probably also maintains the unusual linear/circular architecture of these mitochondrial genomes.
In contrast to the conserved heteroplasmic tRNA loci, evolutionary turnovers in the locations of other tRNA genes appear to be more common. One possible explanation for this apparent contradiction is that tRNA genes are normally evolutionarily labile, but that the dual heteroplasmic state represents a sort of evolutionary “trap.” However, we also recognize that the annotations themselves may be imperfect or incomplete. For instance, some tRNA genes were only detected by one of the two software packages we used, and some tRNA genes appear to be missing altogether (e.g., tRNAThr gene in T. rathkei). While it is possible that these tRNA genes are truly absent from these mitochondrial genomes, we cannot rule out the alternative possibility that they were simply missed by the annotation software. tRNA genes with unusual secondary structures, or subject to substantial RNA editing, for instance, might be especially difficult to detect; some of the identified genes lack the D and/or T arm (Figure S5). Although this is not especially unusual—mitochondrial tRNA genes missing arms have been noted in other lineages (e.g., Masta and Boore 2008), and editing of tRNA molecules is suspected in the isopod Ligia oceanica (Kilpert and Podsiadlowski 2006)—it is still possible that additional tRNA genes with reduced structures were present but not successfully identified.
In any case, these were not the only heteroplasmic sites detected. In C. convexus, one additional polymorphism was found (site 6998) but is probably an artifact because it occurs in a homopolymer sequence, which are known to be error-prone in 454 data. In T. rathkei, we found evidence of three others in the terminal palindromes at the ends of the mitochondrial genome (sites 62, 73, and 14107). Two additional heteroplasmic sites in T. rathkei (10074 and 11244) were polymorphic, found in only one of the two individuals we sequenced. One of these sites (10074) was in one of the few apparently nonfunctional areas of the mitochondrial genome, whereas the other (11244) was a frameshift-causing 1-bp deletion in the cytochrome b gene, almost certainly resulting in a nonfunctional product. Although such a mutation in an essential mitochondrial gene would normally be expected to be deleterious, the presence of another functional copy of the gene likely mitigates those effects. Interestingly, the loss-of-function allele was also detected at a very low frequency of 4% in the other individual we sequenced. Although this low-frequency heteroplasmy might be explained by sequencing or alignment artifacts, an intriguing alternative possibility is that gene conversion coupled with selection may be acting at this site to maintain a functional cytochrome b sequence.
The relative rarity of fixed heteroplasmic sites in these mitochondrial genomes (3–6 sites out of ∼14 kb) suggests that gene conversion is a potent homogenizing force in isopod mitochondrial genomes. Concerted evolution has also been observed in duplicated regions in mitochondrial genomes in other lineages (Ogoh and Ohmiya 2007; Tatarenkov and Avise 2007). However, a high frequency of gene conversion would raise the question of how heteroplasmic tRNA genes are maintained over such long time scales. Of course, as mentioned, selection is likely to play a role. We also cannot discount the possibility that these sites somehow escape gene conversion, for example, if they play a mechanistic role in the gene conversion process. Another possibility is that they may be secondarily edited after gene conversion events. These hypotheses are not mutually exclusive, and could even act together.
Further study is clearly needed on the processes influencing mitochondrial variation and evolution in this lineage, including on the mechanisms underlying gene conversion and recombination, on mutation rates and the frequency of gene conversion events, and on the strength of selection on different types of variants. For instance, further haplotype phasing within individuals could shed light on the tempo of gene conversion; the presence of all possible haplotypes within each individual would suggest rapid and dynamic recombination of mitochondrial monomer units within individuals. Preliminarily, there appear to be just two distinct haplotypes across two closely linked heteroplasmic sites near one end of the T. rathkei mitochondrial DNA sequence (Figure S2), but haplotype phasing at the other sites was impossible with our short-read data, even when taking paired read information into account. Similarly, in C. convexus, just two haplotypes are found in equal numbers across the heteroplasmies at sites 11755 and 12160 (Figure S4). Studying such sites across multiple individuals in a family would answer important questions about mechanisms of transmission. For example, while the tRNAAla/tRNAVal heteroplasmy was stably transmitted in one study in A. vulgare (Doublet et al. 2008), repeating such studies in families with multiple heteroplasmic sites, especially combined with haplotype phase information, would confirm pure maternal cytoplasmic inheritance and stable transmission of multiple heteroplasmic sites. Finally, examining many individuals across multiple polymorphic populations, combined with modeling studies, would shed light on the microevolutionary dynamics of these sites (e.g., frequency of gene conversion and strength of selection). Plummeting sequencing costs, combined with increasing throughput and read lengths, should make sequencing whole mitochondrial genomes in many individuals feasible, providing rich opportunities to answer these questions in the future.