Compared to their counterparts in animals, the mitochondrial (mt) genomes of angiosperms exhibit a number of unique features. However, unravelling their evolution is hindered by the few completed genomes, of which are essentially Sanger sequenced. While next-generation sequencing technologies have revolutionized chloroplast genome sequencing, they are just beginning to be applied to angiosperm mt genomes. Chloroplast genomes of grasses (Poaceae) have undergone episodic evolution and the evolutionary rate was suggested to be correlated between chloroplast and mt genomes in Poaceae. It is interesting to investigate whether correlated rate change also occurred in grass mt genomes as expected under lineage effects. A time-calibrated phylogenetic tree is needed to examine rate change.
We determined a largely completed mt genome from a bamboo, Ferrocalamus rimosivaginus (Poaceae), through Illumina sequencing of total DNA. With combination of de novo and reference-guided assembly, 39.5-fold coverage Illumina reads were finally assembled into scaffolds totalling 432,839 bp. The assembled genome contains nearly the same genes as the completed mt genomes in Poaceae. For examining evolutionary rate in grass mt genomes, we reconstructed a phylogenetic tree including 22 taxa based on 31 mt genes. The topology of the well-resolved tree was almost identical to that inferred from chloroplast genome with only minor difference. The inconsistency possibly derived from long branch attraction in mtDNA tree. By calculating absolute substitution rates, we found significant rate change (∼4-fold) in mt genome before and after the diversification of Poaceae both in synonymous and nonsynonymous terms. Furthermore, the rate change was correlated with that of chloroplast genomes in grasses.
Citation: Ma P-F, Guo Z-H, Li D-Z (2012) Rapid Sequencing of the Bamboo Mitochondrial Genome Using Illumina Technology and Parallel Episodic Evolution of Organelle Genomes in Grasses. PLoS ONE 7(1): e30297. https://doi.org/10.1371/journal.pone.0030297
Editor: Jonathan H. Badger, J. Craig Venter Institute, United States of America
Received: August 24, 2011; Accepted: December 13, 2011; Published: January 17, 2012
Copyright: © 2012 Ma et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Funding: This project is supported by National Natural Science Foundation of China (no. 31170204), the Chinese Academy of Sciences under the knowledge innovation program (grants KSCX-YW-N-029 and KSCX2-YW-N-067), and Yunnan Provincial Government through an innovation team program. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: The authors have declared that no competing interests exist.
Next-generation sequencing that is not only high-throughput but also low-cost has already revolutionized approaches for genome sequencing and is now becoming ‘now-generation’ sequencing . Recently complete or nearly complete chloroplast (cp) genomes of plants have been successfully recovered by Illumina sequencing or 454 pyrosequencing from total DNA containing cpDNA as well as nuclear and mitochondrial (mt) DNA –. Compared to conventional approaches for cp genome sequencing involving purification or PCR amplification of cpDNA , these methods are more simple and effective –. On the contrary, the next-generation sequencing technologies are just beginning to be applied to sequencing of mt genomes of plants and only a few examples of plant mt genome next-generation sequencing have been published so far –. The 21 available mt genomes of higher plants (Table 1) are essentially obtained by Sanger sequencing, involving extracting or PCR amplification of mtDNA and library construction prior to sequencing. This time- and labour-intensive sequencing approach to some extent limits the sequencing of plant mt genomes. Recently, we have sequenced six bamboo cp genomes from total DNA that was enriched for cpDNA using Illumina sequencing . Meanwhile, we found a large number of sequence reads in the total reads being of mtDNA origin. Could the whole or largely completed mt genome be assembled from these reads at the same time? And if so, this would be a rapid and efficient approach to sequence angiosperm mt genomes and more sequenced genomes will be helpful in understanding the extraordinary evolutionary history of angiosperm mt genomes.
The mt genomes of angiosperms exhibit a number of unique features, which distinguish them from their counterparts in animals or other organisms. These features include expanded genome size, frequent structure rearrangement via recombination, ongoing gene loss and transfer to the nuclear genome, uptake of cpDNA and nuclear DNA, and a generally low rate of molecular evolution –. Among them, the slow mt sequence evolution is probably the most prominent and has long been appreciated , , . However, recent studies based on one or a few genes have identified several cases of rate acceleration in mt genomes of certain angiosperm lineages –, and some of these rate increases are temporary with rates returning to normally low levels after acceleration , . These changes in evolutionary rate are mostly restricted to mt genome as expected under locus-specific effects without correlated rate changes in cp and/or nuclear genes –. Nevertheless, a few studies have reported parallel rate changes in mt and cp genomes as expected under lineage effects , . Among them a widely cited case demonstrated a faster rate of synonymous substitutions, which was correlated across cp, mt and nuclear loci in grasses relative to palm . In addition, Zhong et al.  found that episodic rate acceleration of cp genomes occurred in the ancestral grasses and then the rate reverted to the slow rate typical of most monocot species in the descendant lineages. Here, we are interested to investigate whether the similar pattern of rate change occurred in mt genomes during grass evolution.
A well-supported phylogenetic tree that is time calibrated is necessary to examine absolute substitution rates. Although the backbone phylogeny of angiosperms has been established – and the major phylogenetic relationships within the grass family (Poaceae) are resolved with the whole family divided into several basal lineages plus two major lineages (the BEP clade and the PACMAD clade, core Poaceae) , –, these studies have relied mainly on the cpDNA sequences. However, the mtDNA sequences are also useful for reconstructing phylogeny of angiosperms, especially at deep level , . It would be informative to use signals from the mt genome to evaluate independently these relationships derived from the cp genome. In addition, there are numerous studies that could provide reliable estimate of divergence times during the grass evolution , –. Well-resolved phylogenetic relationships and reliable calibration in combination would make it feasible to examine the rate change in mt genomes of grasses.
Here we demonstrated a new approach for sequencing angiosperm mt genome using Illumina sequencing-by-synthesis technology and determined a largely completed mt genome from a bamboo, Ferrocalamus rimosivaginus T. H. Wen (Poaceae). A phylogenetic tree including 22 taxa whose mt genomes have been sequenced was reconstructed based on the sequences of 31 mt genes and the topology was nearly congruent with that inferred from the cp genome. By examining the evolutionary rate of grass mt genomes along the time-calibrated tree, we found a parallel rate change in mt and cp genomes of grasses.
Illumina Sequencing, Genome Assembly and PCR Validation
The template DNA for the F. rimosivaginus cp genome sequencing, which was extracted by a rapid and simple procedure from fresh leaves, in fact contained mtDNA as well . We employed a whole-genome shogun sequencing strategy and one paired-end library for Illumina sequencing was constructed with insert size of about 500 bp. The Illumina system produced 1,594,119 usable paired-end (73 bp and 75 bp) reads in one run for genome assembly . These reads were a mixture of reads derived from cp, mt and nuclear genomes.
We assembled the raw reads using the software SOAPdenovo  with optimal parameters. All the assembled scaffolds and contigs larger than 100 bp were first mapped to the reference mt genomes from nine species of the grass family (Table 1), resulting in 68 mapped contigs and scaffolds. Subsequently we searched these 68 contigs and scaffolds for sequences with significant identity (≥90%) to the cp genome of F. rimosivaginus, and the aligned 3 scaffolds and 35 contigs were removed to avoid the impact of the cpDNA reads on our assembly. In addition, a 3,399 bp sequence located at one end of a scaffold and related to the remaining sequence by an estimated 386 bp length gap was deleted from this scaffold for the same reason. At this point, we obtained 16 scaffolds and 14 contigs with an N50 size of 49.6 kb, achieving a total length of 431.7 kb (Table 2, initial assembly) (see details in Table S1). The average sequencing depth was 39.5× (114,891 mt-derived paired-end reads and 7.2% of total), and there was a relatively narrow variation in the 16 scaffolds sequencing coverage, ranging from 31.9× to 46.5× with a median value of 39.3×. Although these scaffolds contained internal gaps as paired-end information was used to join contigs into scaffolds, the total number of gaps was only 24 with a mean estimated size of 151.5 bp (Table 2).
We further joined the initially assembled contigs and scaffolds into larger scaffolds using the sequence overlap information between them and synteny between the assembled sequences and the reference genomes. This procedure combined 4 scaffolds and 13 contigs in total. All the linkages between them were successfully confirmed by PCR amplification with conventional Sanger sequencing. To close the intra-scaffold gaps, we designed PCR primers and Sanger sequenced the amplified regions. In sum, 23 gaps (an assumed 17 bp gap was proved to be zero-length by PCR analysis and thus not accounted) were closed with a mean size of 172.8 bp which was slightly larger than the estimated size, thereby validating the linkages between regions spanning gaps within scaffolds. The final assembly had 13 scaffolds and one contig with an N50 size of 53.1 kb and a total length of 432.8 kb (Table 2, see details in Table S1), achieving an average sequencing depth of 39.4×.
Validation of linkages and closing gaps by conventional sequencing altogether generated 23,053 bp sequences, of which 16,910 bp sequences could be directly compared to the assembly for accuracy (Table S2). In this comparison, we found 21 mismatched sites distributed in 5 Sanger reads, three of which had low sequencing quality scores. Furthermore, we PCR-amplified and resequenced 16 randomly chosen regions surrounding putative genome rearrangements in comparison with the bamboo Bambusa oldhamii mt genome (EU365401). All the regions were successfully recovered and only one nucleotide substitution was observed in 10,837 bp resequenced regions (Table S2). In total, we tested 27,747 bp sequences by conventional sequencing, validating the accuracy of sequencing and assembly of our mt genome. Only 22 nucleotide substitution errors were found and the error rate was 0.079%, or 0.018% without accounting the errors associated with low Sanger sequence quality. This rate was close to the 0.037–0.056% error rates with the next-generation sequencings reported before , .
The assembled sequence amounted to 432,839 bp distributed in 13 scaffolds and one contig. Since no estimated size existed for the F. rimosivaginus mt genome in previous studies, we evaluated the degree of sequence completion by comparing the assembly size to the average size (484,329 bp) of mt genomes from three closely related species B. oldhamii, Oryza sativa (NC_011033)  and Triticum aestivum (NC_007579) . Based on this comparison, we could assume that the F. rimosivaginus mt genome has been largely assembled and 14 gaps of unknown size remained while mt genome was considered as a circular molecule . The contig and scaffold sequences were deposited in GenBank under Accession Numbers JQ235166 to JQ235179.
The largely completed F. rimosivaginus mt genome has a GC content of 44.1% (Table 3), which is close to the median value of fully sequenced angiosperm mt genomes. As in other angiosperms , , most of the F. rimosivaginus mt genome sequences are noncoding sequences. The coding and intron sequences comprise only 8.9% (38,642 bp) and 5.7% (24,730 bp) of the total length, respectively, including 34 protein, 19 tRNA, and 3 rRNA genes (Table 3). These annotated genes are scattered throughout 12 of the 13 assembled scaffolds. Eight protein genes contain 22 groupII introns in all, 6 of which are trans-spliced.
The F. rimosivaginus mt genome has nearly the same coding capability as those of other grasses (Table 4). All respiratory genes except for sdh3 and sdh4 are present in the genome, in agreement with the suggested frequent losses of these two genes during angiosperm evolution , . The other 15 frequently lost genes are all ribosomal protein coding genes , , among which five genes rpl2, rps8, rps10, rps11, and rps14 are absent or appear to be pseudogenes in the sequenced mt genome. Of this group, the rpl2 gene is the most intact with just a single 165 bp insertion compared to annotated rpl2 gene in other grass mt genomes. Although the insertion does not alter the downstream reading frame, there is a premature stop codon in the sequence of the rpl2 gene and thus it may be not functional. Like other grass mt genomes , a nearly full-length rpl14 pseudogene is retained in the genome and the open reading frame is disrupted by several frameshift mutations. The other three genes have little or no remnant in the genome. The F. rimosivaginus mt genome does not have the translational capacity to recognize all the 61 sense codons even if assuming that U in the third codon/first anticodon wobble position can recognize all bases in the mtDNA, encoding only 14 of the 20 amino acids. Additionally, nearly half of the identified tRNA genes (9 out of 19) are of cp origin, and one of these genes, trnS (gga), has two copies in the genome. All three ribosomal RNA genes (rrn5, rrn18, and rrn26) are present in the mt genome of F. rimosivaginus as other grasses.
Phylogenetic analyses were performed on a 22-taxon (Table 1), 31 mt genes of 28,728 aligned nucleotide positions using maximum likelihood (ML) method with 3 different partitioning strategies: unpartition, partitioning the data set by gene or codon position. The same tree topology and similar bootstrap support (BS) values were obtained regardless of the partitioning strategies (Figure 1A and Figure S1). To evaluate the influence of different methods imposed on phylogenetic reconstruction, maximum parsimony (MP) and Bayesian inference (BI) were also used. The BI analysis generated the same phylogenetic relationships within angiosperms as those inferred from ML and all nodes of the tree received a posterior probability of 1.0 (Figure 1B). The MP analysis resulted in a single most parsimonious tree with a length of 9,977, a consistency index (CI) of 0.64 (excluding uninformative characters), and a retention index (RI) of 0.85 (Figure 1C). The MP tree differed from the ML tree only in the placement of Vitis vinifera, and overall BS values were slightly lower than those of the ML tree. Furthermore, there were four nodes in the MP tree only receiving weak support (BS = 59% to 67%), two of which involved in the placement of V. vinifera. Below, we would focus on phylogenetic relationships with an emphasis on the ML topology.
The ML tree (A), the BI tree (B) and the MP tree (C) based on 31 mitochondrial (mt) genes. The topology of the chloroplast tree (D) is constrained by previous studies , , . Numbers at nodes indicate bootstrap support (BS) values ≥50% or BI posterior probabilities. Branch lengths of the ML tree calculated through RAxML analysis, and correspond to scale bar (in units of substitutions/site).
The ML topology was well supported and phylogenetic resolutions were increased relative to a previous analysis based on only four mt genes with dense taxon sampling , although the taxon sampling was very different between these two studies and comparing them was not straightforward. Phylogenetic relationships inferred from mtDNA were congruent with those based on cpDNA (Figure 1A and 1D) , –,  with the exception of one topological difference. Within Poaceae, the three subfamilies Bambusoideae (B. oldhamii and F. rimosivaginus), Ehrhartoideae (O. sativa and O. rufipogon), and Pooideae (T. aestivum) formed the BEP clade which had a sister relationship to the PACMAD clade (the other five grasses in the tree belong to Panicoideae) in all the previous phylogenetic analyses of the cpDNA sequences , –. However, ML analysis of mtDNA sequences did not support the monophyly of the BEP clade but a sister relationship of Ehrhartoideae+Panicoideae with 100% bootstrap support, although this sister relationship only received 63% BS support in the MP tree. Furthermore, two relative long branches (the Ehrhartoideae branch and the Panicoideae branch) separated by the short internode (Figure 1A) implied that this sister relationship might be an artifact of phylogeny reconstruction due to long branch attraction (LBA) , .
To detect the LBA artifact, we first performed phylogenetic analyses at the amino acid level of the total 31 genes. The BS value for Ehrhartoideae+Panicoideae decreased dramatically from 100% to 58% in the ML tree (Figure 2A), while BS values for other nodes within Poaceae did not decrease proportionally but instead remained high (BS = 84% to 100%). On the other hand, the BEP clade was weakly supported (BS = 52%) as a monophyletic group in the MP tree (Figure 2B). We further partitioned the 31 genes into fast- and slow-evolving ones according to their substitution rates. The genes with lower substitution rate than the average of all the 31 genes (0.060 substitutions/site) were considered as slow-evolving genes and the rest as fast-evolving ones (see Materials and Methods for detail). If Ehrhartoideae+Panicoideae were an LBA artifact, then support for this group would be favored by the partitions of fast-evolving genes, whereas it would be minimized in partitions of slow-evolving genes. Consistent with our hypothesis, analyses of the concatenated 12 fast-evolving genes supported Ehrhartoideae+Panicoideae with 100% BS value in the ML tree (Figure 3A), and BS value for this group in the MP tree increased from 63% based on all the 31 genes to 71% (Figure 3B and Figure 1C). Furthermore, the monophyly of the BEP clade (BS = 32%) was recovered in ML analysis of the concatenated 19 slow-evolving genes (Figure 3C), and the relationships of the BEP clade within Poaceae were not resolved in MP analysis of this data (Figure 3D).
The ML tree (A) and the MP tree (B). Numbers associated with branches are bootstrap support (BS) values. Oryza and related BS values indicated in bold.
The ML tree (A) and the MP tree (B) inferred from the concatenated 12 fast-evolving mitochondrial (mt) genes. The ML tree (C) and the MP tree (D) inferred from the concatenated 19 slow-evolving mt genes. Numbers associated with branches are bootstrap support (BS) values. Oryza and related BS values indicated in bold.
Rare genomic changes such as gene losses, arrangements of genes, insertions and deletions of introns are less prone to homoplasy than nucleotide sequences and have become an alternative approach for phylogenetic studies , . Two gene losses, rpl5 and rps19, were restricted to the mt genomes of Panicoideae and pseudogene of rps14 was only retained in the mt genomes of the BEP clade (Table 4). These rare genomic changes supported the monophyly of the BEP clade as well.
Pattern of Rate Change in Grass Mitochondrial Genomes
The long branch leading to Poaceae implied that the mt genomes of this family may undergo an elevated evolutionary rate (Figure 1A), just like the cp genomes of this family , , . To quantify the evolutionary rate in grass mt genomes, we calculated absolute substitution rates (R) in substitutions per site per billion years (SSB) as described by Parkinson et al.  in mtDNA tree with the placement of Oryza constrained by Figure 1D. The values of R for branches involving Poaceae were calculated by dividing the branch length by the length of time for that branch. Divergence times were based on estimates in previous studies , –, with separation of monocots/eudicots and origin of core Poaceae setting at 135 Myr and 65 Myr, respectively (other divergence times in Table S3). Because there was no reliable estimate of divergence times within genera Oryza and Zea, R was averaged among them from the terminal species to the common ancestor of Oryza/Triticum or Zea/Tripsacum.
Figure 4 illustrated that the evolutionary rate of mt genome changed many times during grass evolution. Comparison of the fastest and slowest lineages showed that the rate varied by a factor of 17. The elevation of rate occurred on the common branch of Poaceae after they diverged from eudicots, and the majority of rates after diversification of Poaceae were ∼4-fold slower than that on the branch leading to Poaceae. For example, the rate along the lineages from monocots/eudicots separation to O. sativa changed from 0.52 to 0.05 to 0.12 SSB. This pattern of rate change resembled that observed in grass cp genomes . Another notable feature of Figure 4 was the lower evolutionary rate in the lineages of Bambusoideae compared to the other three subfamilies in Poaceae. To examine the impact of divergence times on demonstrating rate change, we also applied much older divergence times 212 Myr and 72 Myr used in estimating evolutionary rate in grass cp genomes  for monocots/eudicots separation and origin of core Poaceae, respectively. Nevertheless, the same tendency for rate change in mt genome was obtained with these calibration points (Figure S2).
Values above each branch indicate absolute substitution rate (R) in substitutions per site per billion years (SSB) for that branch. Among them, 0.12±0.01 and 0.11±0.01 above and below the branch to Oryza are the mean values along the lineage from the common ancestor of Oryza/Triticum to O. sativa and O. rufipogon, respectively, while 0.21±0.02 above the branch to Zea is the average value of R along the lineage from the common ancestor of Zea/Tripsacum to three Zea species.
To check the rate decrease of mt genome after diversification of Poaceae in more detail, we partitioned the substitutions into synonymous and nonsynonymous ones. The same pattern of rate change as that of total substitutions was observed in both types of substitutions (Figure S3). And nonsynonymous/synonymous rate ratio (ω) on the branch leading to Poaceae was 0.42, indicating overall purifying selection operating on these genes during grass evolution. However, we did not exclude the RNA editing sites of mt genes in calculating ω and the existence of RNA editing sites can bias the estimate of it .
The evolutionary rates above were estimated on the concatenated 31-gene data set. However, different genes of mt genome in the same plant lineage could have various evolutionary rates , . To explore the rate change in individual genes, the ratio of R before diversification of Poaceae to that on the line from the common ancestor of Poaceae to O. sativa for each gene was calculated (Figure 5). Among them two genes, nad4L and nad7, had no nucleotide substitutions during the period from Poaceae origin to O. sativa and thus the ratios for them were artificially given the average value 6.46 for all the other genes. All the genes except for nad6 showed the same pattern of rate change of elevated rate before diversification of Poaceae with subsequent slow-down after diversification (Figure 5) and this pattern was consistent with that based on analysis of combined genes. Furthermore, the rate change exhibited limit of variation between genes and the majority of ratios had a value around 5.00.
Due to their high-throughput and low-cost, next-generation sequencing technologies have greatly improved the approaches for genome sequencing. For angiosperm organelle genomes, however, they have been largely restricted to sequencing of cp genomes until now. Using next-generation sequencing platforms for mt genome sequencing has only recently been explored –. Furthermore, only a few angiosperms have had their mt genomes sequenced, and more completed genomes are necessary to study the evolution of angiosperm mt genomes. Here we present a largely completed mt genome from the bamboo F. rimosivaginus mainly based on Illumina sequencing, providing the demonstration of the feasibility for sequencing angiosperm mt genomes with Illumina sequencing technique. With successful sequencing of the mt genomes using Illumina and 454 sequencing technologies – it has become evident that high-throughput next-generation sequencers could hold promise for the angiosperm mt genomes sequencing in the near future. The Illumina sequence reads for the other five bamboos in  are under analysis and similar results are obtained. This rapid and effective approach for the bamboo F. rimosivaginus mt genome sequencing could be an alternative to the established methods for angiosperm mt genomes sequencing.
Feasibility of Illumina Sequencing of Angiosperm Mitochondrial Genomes
Given the highly variable sizes of angiosperm mt genomes , , , we could only deduce that the F. rimosivaginus mt genome were largely completed based on mean size of mt genomes from three closely related grasses. Nevertheless, our assembly appears to be 100% complete with regard to gene content, as the housekeeping genes shared by the reference mt genomes are all identified from our assembly. Additionally, the draft genome is of relatively high quality with a N50 size of 53.1 kb and only 14 gaps remained to be finished to complete the genome. However, it should be noted that the availability of closely related reference genomes in the grass family and presumed less large repeats in bamboo mt genomes  may have made it relatively easier to assemble than some other angiosperm mt genomes.
In compared to angiosperm cp genomes, it would be much difficult to assemble mt genome from Illumina short reads because of larger genome size, more repetitive sequences and frequent genome rearrangements. For assembly of the F. rimosivaginus mt genome, there are three probable reasons that could explain why we have not obtained a single, full-length genome sequence. First, the raw reads just do not represent complete coverage of the genome. The sequencing depth could vary throughout the genome due to GC content or other factors . However, in light of the relatively high 39.5× average sequencing depth, we argue that the raw reads may have covered the whole genome. Second, it is failure to retrieve certain assembled contigs and/or scaffolds whose sequences are actually derived from the F. rimosivaginus mt genome during assembly. However, the availability of several closely related reference genomes and the low threshold used in mapping contigs and scaffolds (see details in Materials and Methods) would dramatically reduce this possibility. In fact, about 17.9% of the assembled sequences have no similarity to the sequences of any other sequenced mt genomes of plants as well as the sequences in the NCBI non-redundant nucleotide and protein databases (data not shown). Third, the incorporated sequences from cpDNA of F. rimosivaginus in the mt genome were excluded from the assembly. The reported uptakes of cpDNA sequences by angiosperm mt genomes constitute 1.1–11.5% of the genome size , . These cp-derived sequences are very likely to affect our assembly and thus we removed assembled sequences with significant sequence identity to the cp genome of F. rimosivaginus. This is a potential disadvantage of our approach for mt genome sequencing. Among three explanations above, the last one may contribute mostly to the incompleteness of our assembled genome. In the future, modifying the method used in DNA extraction to reduce the proportion of cpDNA in the total DNA could solve this problem. Advances in next-generation sequencing technologies, such as increase in length and number of sequence reads (more coverage) and paired-end sequencing with larger insert size, will also improve the assembly.
Gene Content of the F. rimosivaginus Mitochondrial Genome
The mt genome of F. rimosivaginus is the second sequenced mt genome from a bamboo species in Poaceae. It encodes roughly the same protein genes as mt genomes of other grasses, with the loss of sdh3, sdh4 and several ribosomal protein genes which are all prone to loss during angiosperm evolution . An rpl14 pseudogene is retained in the genome in good agreement with survival of the pseudogene in grasses reported before . Like other grass mt genomes, the F. rimosivaginus mt genome does not encode the full set of tRNAs necessary to recognize all codons, and nearly half of them are cp-derived tRNAs. The mt genome contains the identical gene set for tRNAs of mt origin among the ten sequenced grasses, while their genomes do not share the same tRNAs of cp origin, indicating an ongoing process of acquisition and loss of cp-derived tRNAs.
Parallel Episodic Evolution of Organelle Genomes
At present, phylogenies of angiosperms are essentially reconstructed from the cpDNA sequences. If the cp and mt genomes are both strictly maternally inherited in angiosperms, we would expect them to contain identical phylogenetic signals. As expected, the relationships within our reconstructed mtDNA tree were largely congruent with phylogenetic analyses of the cpDNA sequences , –, . The well-supported topology confirms the utility of mtDNA for phylogenetic reconstruction ,  and mt phylogenomics may be useful for resolving some difficult angiosperm phylogenies. In spite of overall congruence, the angiosperm mtDNA tree is in obvious contrast to cpDNA tree in the monophyly of the BEP clade. Using more conserved protein sequences of mt genes, the monophyly or significant decrease in support for non-monophyly of the BEP clade was obtained in mtDNA tree. Furthermore, the fast-evolving genes were more favored than the slow-evolving genes in non-monophyly of the BEP clade. Genome-level features of mt genome that are less subjective to homoplasy also support the monophyly of the BEP clade. In summary, these results suggest that the sister relationship of Ehrhartoideae+Panicoideae inferred from mtDNA is most likely an LBA artifact. The fast-evolving genes which have high different evolutionary rates among different lineages within Poaceae may largely contribute to the LBA artifact , . In addition, the taxon sampling in our phylogenetic analyses was very sparse. Thus, LBA due to poor taxon sampling ,  may also result in this phylogenetic inconsistency.
Unlike prior studies based on a few mt genes –, we used nearly all the protein genes of mt genome to demonstrate episodic evolution of mt genome in grasses. Moreover, all the examined genes except for nad6 had the same pattern of rate change. Although completed mt genomes of other monocots are not available to break the long, poorly sampled branch leading to Poaceae, evolutionary rate averaging along the branch is still faster than those after diversification of Poaceae. However, the range of rate variation is much smaller than those recently documented –. And in contrast to rate variations being restricted to synonymous substitutions in these studies, nonsynonymous substitutions in grass mt genomes also show the corresponding rate change. The factors that are responsible for this rate change are very likely to be different from those with locus effect proposed before, such as the efficiency of mtDNA repair. Furthermore, that the episodic evolution is correlated between the mt genome and the cp genome in grasses is consistent with lineage effects. The same underlying factors for the rate change may simultaneously operate on the mt and cp genomes of grasses. However, we could not confirm the rate acceleration of mt genome only occurring in the ancestral grasses like that of cp genome with the present data. Completed mt genomes from taxa closely related to Poaceae are required to resolve this question in further studies.
Materials and Methods
Mitochondrial DNA Isolation, Genome Sequencing and Assembly
Total DNA enriched for cpDNA was extracted and sequenced using the protocols described in . The raw short reads were firstly assembled using SOAPdenovo  with K = 31 bp and scaffolding contigs with a minimum size of 100 bp. All the assembled scaffolds and contigs (≥100 bp) were mapped to the sequenced mt genomes in Poaceae (Table 1) using BLASTN searches from NCBI with default parameters. Those contigs and scaffolds whose query coverage was greater than 40% were retrieved and then they were used to search against the NCBI non-redundant nucleotide and protein databases with BLASTN (http://blast.ncbi.nlm.nih.gov/). These without significant identity to sequences from the cp and/or nuclear genomes were deemed as sequences derived from mtDNA. In this procedure, we removed 3 scaffolds and 35 contigs which could be aligned to reference mt genomes but supposed to be of cpDNA origin as they had significant sequences identity (≥90%) to the mt genomes as well as the F. rimosivaginus cp genome. We realigned all the raw reads onto the assembled sequences using software Bowtie  with −v = 3. The aligned paired-end reads were used to determine the sequencing depth. A second round of assembly was carried out on the initially assembled contigs and scaffolds. By using the information of overlap of ≥12 bp between contigs and scaffolds and synteny between assembled and reference genomes, we further joined them into larger scaffolds.
PCR-Based Genome Finishing and Validation
Following the steps above, the candidate linkages of the contigs and/or scaffolds were validated by PCR analysis. We designed primers according to the nucleotide sequences surrounding the linkage and in all 12 primer pairs were used. To close the gaps inside the scaffolds, 20 primer pairs were also designed. Furthermore, we randomly chose 16 regions in the assembly for resequencing using PCR. All the primer sequences were in Table S4. PCR products were sequenced by ABI 3730xl genetic analyzer using standard protocols. Sanger sequences and the assembly were aligned using MEGA 4.0  to determine if there were any differences.
A preliminary annotation was carried out by mapping BLASTN hits employing known mt genes of grasses as queries and subsequently, by testing for consistency of the open reading frame. The exact gene and exon boundaries were determined by alignment of homologous genes from B. oldhamii. We also used tRNAscan-SE 1.21  to corroborate tRNA boundaries identified by BLASTN. The sequences of identified tRNA genes were BLAST searched against the cp genome of F. rimosivaginus to detect cp-derived tRNA genes.
We extracted nucleotide sequences for all protein genes from 22 seed plant mt genomes (Table 1). After excluding several genes (sdh3, sdh4, rpl2, rpl6, rps1, rps2, rps8, rps10, rps11, rps14, and rps19) that are missing from many of the sequenced genomes, 31 genes were retained. Each gene was aligned separately with MEGA 4.0 constrained by its amino acid sequence. Alignments of nucleotide sequences were then manually adjusted, and ambiguously aligned regions were excluded from the analysis. We did not exclude RNA editing sites as several studies suggested that they were not a problem in phylogenetic reconstruction, especially with large sequence data sets , , , . Finally, individual genes were concatenated, and the resulted alignment consisted of 28,728 nucleotides.
ML analyses of the 31-gene data set were performed with RAxML 7.0.4  using three partitioning strategies: unpartitioned, partitioned the data based on gene region, and partitioned the protein genes by each codon position. All ML analyses used the rapid BS algorithm implemented in RAxML 7.0.4 with 1,000 replicates and the general time-reversible (GTR) model of evolution with among-site rate variation. For BI, we used software MrBayes 3.1.2  with GTR+G+I model. The run started with a random tree, default priors, and four Markov chains, totalling 1 million generations with sampling trees every 100th generation. When convergence was obtained, a consensus tree was calculated after omitting the first 25% of trees as burn-in. For MP, we used PAUP*4.0b10  to implement heuristic searches that consisted of TBR branch swapping, starting from 1,000 trees built by random taxon stepwise addition, and multrees option in effect. Non-parametric bootstrap analysis was conducted under 200 replicates with TBR branch swapping from 100 random taxon addition staring trees. To detect long branch attraction, phylogenetic trees were constructed from the amino acid sequences of 31 genes with both ML and MP analysis conducted like before. ML analysis of protein sequences was performed with the Dayhoff matrix of amino acid substitutions and a discrete gamma distribution with four rate categories. The substitution rates (in number of substitutions per site) of the 31 genes in the 22 sampled species were calculated by MEGA 4.0 under the model of Kimura 2-Parameter, ranging from 0.027 to 0.125 with a mean of 0.060. Nineteen of the 31 genes (nad1, nad2, nad3, nad4, nad4L, nad5, nad6, nad7, nad9, cob, cox1, cox3, ccmB, ccmC, rpl16, rps7, rps12, rps13, and mttB) with substitution rates lower than the average were concatenated to form the 19 slow-evolving genes data and the rest (cox2, atp1, atp4, atp6, atp8, atp9, ccmFC, ccmFN, rpl5, rps3, rps4, and matR) formed the 12 fast-evolving genes data. ML and MP analyses were also performed on the concatenated 12 fast-evolving genes and 19 slow-evolving genes in the same way.
Estimation of Absolute Substitution Rates
Absolute substitution rates were calculated for the Poaceae lineages using method that has been described before . Briefly, branch lengths that represent the number of substitutions per site were determined for concatenation of 31 genes using codeml in PAML v4.4  with topologically constrained tree. Codon frequencies were computed by using F3×4 method and separate ω ratios were estimated for each branch. Relationships within the tree were based on Figure 1D. Divergence times used (Table S3) were taken from previous studies , –. Then absolute substitution rate per branch was calculated by dividing the branch length by the length of time for that branch. Standard errors were determined as in . Absolute substitution rates in terms of nonsynonymous and synonymous substitutions were calculated in the same way. For 30 individual genes (rpl5 was excluded for loss in Panicoideae), the absolute substitution rate was analyzed in the same manner as that of concatenated genes and then the ratio of absolute substitution rate before diversification of Poaceae to from origin of the common ancestor of Poaceae to O. sativa was calculated.
Phylogenetic trees as determined by RAxML based on the 31 mitochondrial genes, under the following partitioning schemes: A) partitioned by gene; B) partitioned by codon position. Numbers at nodes indicate bootstrap support (BS) values.
Rate changes in grass mitochondrial genes during evolution with divergence times 212 Myr and 72 Myr for monocots/eudicots separation and origin of core Poaceae, respectively.
Rates of nonsynonymous (A) and synonymous (B) substitutions changes in grass mitochondrial genes during evolution.
Lists of assembled initial and final contigs and scaffolds.
Comparison of the assembly and Sanger sequences.
Divergence times used in calculating absolute substitution rates of grass mitochondrial genes.
We are deeply indebted to Yun-Jie Zhang and Jun-Bo Yang for assistances in experiments, to Drs. Chun-Xia Zeng, Yu-Xiao Zhang, and Jin-Mei Lu for helpful discussions.
Conceived and designed the experiments: D-ZL. Performed the experiments: P-FM. Analyzed the data: P-FM Z-HG. Contributed reagents/materials/analysis tools: P-FM Z-HG. Wrote the paper: P-FM Z-HG D-ZL.
- 1. Neafsey DE, Haas BJ (2011) ‘Next-generation’ sequencing becomes ‘now-generation’. Genome Biol 12: 303.
- 2. Atherton RA, McComish BJ, Shepherd LD, Berry LA, Albert NW, et al. (2010) Whole genome sequencing of enriched chloroplast DNA using the Illumina GAII platform. Plant Methods 6: 22.
- 3. Nock CJ, Waters DL, Edwards MA, Bowen SG, Rice N, et al. (2011) Chloroplast genome sequences from total DNA for plant identification. Plant Biotechnol J 9: 328–333.
- 4. Wolf PG, Der JP, Duffy AM, Davidson JB, Grusz AL, et al. (2011) The evolution of chloroplast genes and genomes in ferns. Plant Mol Biol 76: 251–261.
- 5. Jansen RK, Raubeson LA, Boore JL, dePamphilis CW, Chumley TW, et al. (2005) Methods for obtaining and analyzing whole chloroplast genome sequences. Methods Enzymol 395: 348–384.
- 6. Fujii S, Kazama T, Yamada M, Toriyama K (2010) Discovery of global genomic re-organization based on comparison of two newly sequenced rice mitochondrial genomes with cytoplasmic male sterility-related genes. BMC Genomics 11: 209.
- 7. Rodriguez-Moreno L, Gonzalez V, Benjak A, Marti MC, Puigdomenech P, et al. (2011) Determination of the melon chloroplast and mitochondrial genome sequences reveals that the largest reported mitochondrial genome in plants contains a significant amount of DNA having a nuclear origin. BMC Genomics 12: 424.
- 8. Davila JI, Arrieta-Montiel MP, Wamboldt Y, Cao J, Hagmann J, et al. (2011) Double-strand break repair processes drive evolution of the mitochondrial genome in Arabidopsis. BMC Biol 9: 64.
- 9. Zhang YJ, Ma MF, Li DZ (2011) High-throughput sequencing of six bamboo chloroplast genomes: phylogenetic implications for temperate woody bamboos (Poaceae: Bambusoideae). PLoS ONE 6: e20596.
- 10. Palmer JD, Adams KL, Cho Y, Parkinson CL, Qiu YL, et al. (2000) Dynamic evolution of plant mitochondrial genomes: mobile genes and introns and highly variable mutation rates. Proc Natl Acad Sci U S A 97: 6960–6966.
- 11. Adams KL, Qiu YL, Stoutemyer M, Palmer JD (2002) Punctuated evolution of mitochondrial gene content: high and variable rates of mitochondrial gene loss and transfer to the nucleus during angiosperm evolution. Proc Natl Acad Sci U S A 99: 9905–9912.
- 12. Kubo T, Newton KJ (2008) Angiosperm mitochondrial genomes and mutations. Mitochondrion 8: 5–14.
- 13. Palmer JD, Herbon LA (1988) Plant mitochondrial DNA evolves rapidly in structure, but slowly in sequence. J Mol Evol 28: 87–97.
- 14. Wolfe KH, Li WH, Sharp PM (1987) Rates of nucleotide substitution vary greatly among plant mitochondrial, chloroplast, and nuclear DNAs. Proc Natl Acad Sci U S A 84: 9054–9058.
- 15. Sloan DB, Barr CM, Olson MS, Keller SR, Taylor DR (2008) Evolutionary rate variation at multiple levels of biological organization in plant mitochondrial DNA. Mol Biol Evol 25: 243.
- 16. Mower JP, Touzet P, Gummow JS, Delph LF, Palmer JD (2007) Extensive variation in synonymous substitution rates in mitochondrial genes of seed plants. BMC Evol Biol 7: 135.
- 17. Cho Y, Mower JP, Qiu YL, Palmer JD (2004) Mitochondrial substitution rates are extraordinarily elevated and variable in a genus of flowering plants. Proc Natl Acad Sci U S A 101: 17741–17746.
- 18. Parkinson CL, Mower JP, Qiu YL, Shirk AJ, Song K, et al. (2005) Multiple major increases and decreases in mitochondrial substitution rates in the plant family Geraniaceae. BMC Evol Biol 5: 73.
- 19. Sloan DB, Oxelman B, Rautenberg A, Taylor DR (2009) Phylogenetic analysis of mitochondrial substitution rate variation in the angiosperm tribe Sileneae. BMC Evol Biol 9: 260.
- 20. Soria-Hernanz DF, Braverman JM, Hamilton MB (2008) Parallel rate heterogeneity in chloroplast and mitochondrial genomes of Brazil nut trees (Lecythidaceae) is consistent with lineage effects. Mol Biol Evol 25: 1282–1296.
- 21. Eyre-Walker A, Gaut BS (1997) Correlated rates of synonymous site evolution across plant genomes. Mol Biol Evol 14: 455–460.
- 22. Zhong B, Yonezawa T, Zhong Y, Hasegawa M (2009) Episodic evolution and adaptation of chloroplast genomes in ancestral grasses. PLoS One 4: e5297.
- 23. Jansen RK, Cai Z, Raubeson LA, Daniell H, Depamphilis CW, et al. (2007) Analysis of 81 genes from 64 plastid genomes resolves relationships in angiosperms and identifies genome-scale evolutionary patterns. Proc Natl Acad Sci U S A 104: 19369–19374.
- 24. Moore MJ, Bell CD, Soltis PS, Soltis DE (2007) Using plastid genome-scale data to resolve enigmatic relationships among basal angiosperms. Proc Natl Acad Sci U S A 104: 19363–19368.
- 25. Moore MJ, Soltis PS, Bell CD, Burleigh JG, Soltis DE (2010) Phylogenetic analysis of 83 plastid genes further resolves the early diversification of eudicots. Proc Natl Acad Sci U S A 107: 4623–4628.
- 26. Zhang W (2000) Phylogeny of the grass family (Poaceae) from rpl16 intron sequence data. Mol Phylogenet Evol 15: 135–146.
- 27. Bouchenak-Khelladi Y, Salamin N, Savolainen V, Forest F, Bank M, et al. (2008) Large multi-gene phylogenetic trees of the grasses (Poaceae): progress towards complete tribal and generic level sampling. Mol Phylogenet Evol 47: 488–505.
- 28. Grass Phylogeny Working G (2001) Phylogeny and subfamilial classification of the grasses (Poaceae). Ann Mo Bot Gard 88: 373–457.
- 29. Qiu YL, Li L, Wang B, Xue JY, Hendry TA, et al. (2010) Angiosperm phylogeny inferred from sequences of four mitochondrial genes. J Syst Evol 48: 391–425.
- 30. Qiu YL, Li L, Hendry TA, Li R, Taylor DW, et al. (2006) Reconstructing the basal angiosperm phylogeny: evaluating information content of mitochondrial genes. Taxon 55: 837–856.
- 31. Chaw SM, Chang CC, Chen HL, Li WH (2004) Dating the monocot-dicot divergence and the origin of core eudicots using whole chloroplast genomes. J Mol Evol 58: 424–441.
- 32. Piperno DR, Sues HD (2005) Paleontology. Dinosaurs dined on grass. Science 310: 1126–1128.
- 33. Vicentini A, Barber JC, Aliscioni SS, Giussani LM, Kellogg EA (2008) The age of the grasses and clusters of origins of C4 photosynthesis. Global Change Biol 14: 2963–2977.
- 34. Bouchenak-Khelladi Y, Verboom GA, Savolainen V, Hodkinson TR (2010) Biogeography of the grasses (Poaceae): a phylogenetic approach to reveal evolutionary history in geographical space and geological time. Bot J Linn Soc 162: 543–557.
- 35. Li R, Zhu H, Ruan J, Qian W, Fang X, et al. (2010) De novo assembly of human genomes with massively parallel short read sequencing. Genome Res 20: 265–272.
- 36. Moore MJ, Dhingra A, Soltis PS, Shaw R, Farmerie WG, et al. (2006) Rapid and accurate pyrosequencing of angiosperm plastid genomes. BMC Plant Biol 6: 17.
- 37. Cronn R, Liston A, Parks M, Gernandt DS, Shen R, et al. (2008) Multiplex sequencing of plant chloroplast genomes using Solexa sequencing-by-synthesis technology. Nucleic Acids Res 36: e122.
- 38. Notsu Y, Masood S, Nishikawa T, Kubo N, Akiduki G, et al. (2002) The complete sequence of the rice (Oryza sativa L.) mitochondrial genome: frequent DNA sequence acquisition and loss during the evolution of flowering plants. Mol Genet Genomics 268: 434–445.
- 39. Ogihara Y, Yamazaki Y, Murai K, Kanno A, Terachi T, et al. (2005) Structural dynamics of cereal mitochondrial genomes as revealed by complete nucleotide sequencing of the wheat mitochondrial genome. Nucleic Acids Res 33: 6235–6250.
- 40. Mower JP, Bonen L (2009) Ribosomal protein L10 is encoded in the mitochondrial genome of many land plants and green algae. BMC Evol Biol 9: 265.
- 41. Ong HC, Palmer JD (2006) Pervasive survival of expressed mitochondrial rps14 pseudogenes in grasses and their relatives for 80 million years following three functional transfers to the nucleus. BMC Evol Biol 6: 55.
- 42. Kellogg EA, Birchler JA (1993) Linking phylogeny and genetics: Zea mays as a tool for phylogenetic studies. Syst Biol 42: 415.
- 43. Bergsten J (2005) A review of long-branch attraction. Cladistics 21: 163–193.
- 44. Soltis DE, Albert VA, Savolainen V, Hilu K, Qiu YL, et al. (2004) Genome-scale data, angiosperm relationships, and ‘ending incongruence’: a cautionary tale in phylogenetics. Trends Plant Sci 9: 477–483.
- 45. Boore JL, Fuerstenberg SI (2008) Beyond linear sequence comparisons: the use of genome-level characters for phylogenetic reconstruction. Philos Trans R Soc Lond B Biol Sci 363: 1445–1451.
- 46. Boore JL (2006) The use of genome-level characters for phylogenetic reconstruction. Trends Ecol Evol 21: 439–446.
- 47. Gaut BS, Muse SV, Clark WD, Clegg MT (1992) Relative rates of nucleotide substitution at the rbcL locus of monocotyledonous plants. J Mol Evol 35: 292–303.
- 48. Lu MZ, Szmidt AE, Wang XR (1998) RNA editing in gymnosperms and its impact on the evolution of the mitochondrial coxI gene. Plant Mol Biol 37: 225–234.
- 49. Alverson AJ, Wei X, Rice DW, Stern DB, Barry K, et al. (2010) Insights into the evolution of mitochondrial genome size from complete sequences of Citrullus lanatus and Cucurbita pepo (Cucurbitaceae). Mol Biol Evol 27: 1436–1448.
- 50. Hsu CL, Mullin BC (1989) Physical characterization of mitochondrial DNA from cotton. Plant Mol Biol 13: 467–468.
- 51. Alverson AJ, Zhuo S, Rice DW, Sloan DB, Palmer JD (2011) The mitochondrial genome of the legume Vigna radiata and the analysis of recombination across short mitochondrial repeats. PLoS One 6: e16404.
- 52. Bentley DR, Balasubramanian S, Swerdlow HP, Smith GP, Milton J, et al. (2008) Accurate whole human genome sequencing using reversible terminator chemistry. Nature 456: 53–59.
- 53. Goremykin VV, Salamini F, Velasco R, Viola R (2009) Mitochondrial DNA of Vitis vinifera and the issue of rampant horizontal gene transfer. Mol Biol Evol 26: 99–110.
- 54. Lockhart P, Novis P, Milligan BG, Riden J, Rambau A, et al. (2006) Heterotachy and tree building: A case study with plastids and eubacteria. Mol Biol Evol 23: 40–45.
- 55. Wu CS, Wang YN, Hsu CY, Lin CP, Chaw SM (2011) Loss of different inverted repeat copies from the chloroplast genomes of Pinaceae and Cupressophytes and influence of heterotachy on the evaluation of gemnosperm phylogeny. Genome Biol Evol.
- 56. Langmead B, Trapnell C, Pop M, Salzberg SL (2009) Ultrafast and memory-efficient alignment of short DNA sequences to the human genome. Genome Biol 10: R25.
- 57. Tamura K, Dudley J, Nei M, Kumar S (2007) MEGA4: Molecular Evolutionary Genetics Analysis (MEGA) software version 4.0. Mol Biol Evol 24: 1596–1599.
- 58. Schattner P, Brooks AN, Lowe TM (2005) The tRNAscan-SE, snoscan and snoGPS web servers for the detection of tRNAs and snoRNAs. Nucleic Acids Res 33: W686–689.
- 59. Bowe LM, dePamphilis CW (1996) Effects of RNA editing and gene processing on phylogenetic reconstruction. Mol Biol Evol 13: 1159–1166.
- 60. Picardi E, Quagliariello C (2008) Is plant mitochondrial RNA editing a source of phylogenetic incongruence? An answer from in silico and in vivo data sets. BMC Bioinformatics 9: S14.
- 61. Stamatakis A (2006) RAxML-VI-HPC: maximum likelihood-based phylogenetic analyses with thousands of taxa and mixed models. Bioinformatics 22: 2688–2690.
- 62. Ronquist F, Huelsenbeck JP (2003) MrBayes 3: Bayesian phylogenetic inference under mixed models. Bioinformatics 19: 1572–1574.
- 63. Swofford D (2002) PAUP* 4.0: phylogenetic analysis using parsimony: Sinauer Associates, Sunderland, Mass.
- 64. Yang Z (2007) PAML 4: phylogenetic analysis by maximum likelihood. Mol Biol Evol 24: 1586–1591.
- 65. Chaw SM, Shih AC, Wang D, Wu YW, Liu SM, et al. (2008) The mitochondrial genome of the gymnosperm Cycas taitungensis contains a novel family of short interspersed elements, Bpu sequences, and abundant RNA editing sites. Mol Biol Evol 25: 603–615.
- 66. Clifton SW, Minx P, Fauron CM, Gibson M, Allen JO, et al. (2004) Sequence and comparative analysis of the maize NB mitochondrial genome. Plant Physiol 136: 3486–3503.
- 67. Unseld M, Marienfeld JR, Brandt P, Brennicke A (1997) The mitochondrial genome of Arabidopsis thaliana contains 57 genes in 366,924 nucleotides. Nat Genet 15: 57–61.
- 68. Kubo T, Nishizawa S, Sugawara A, Itchoda N, Estiati A, et al. (2000) The complete nucleotide sequence of the mitochondrial genome of sugar beet (Beta vulgaris L.) reveals a novel gene for tRNA(Cys)(GCA). Nucleic Acids Res 28: 2571–2576.
- 69. Handa H (2003) The complete nucleotide sequence and RNA editing content of the mitochondrial genome of rapeseed (Brassica napus L.): comparative analysis of the mitochondrial genomes of rapeseed and Arabidopsis thaliana. Nucleic Acids Res 31: 5907–5916.
- 70. Sugiyama Y, Watase Y, Nagase M, Makita N, Yagura S, et al. (2005) The complete nucleotide sequence and multipartite organization of the tobacco mitochondrial genome: comparative analysis of mitochondrial genomes in higher plants. Mol Genet Genomics 272: 603–615.
- 71. Sloan DB, Alverson AJ, Storchova H, Palmer JD, Taylor DR, et al. (2010) Extensive loss of translational genes in the structurally dynamic mitochondrial genome of the angiosperm Silene latifolia. BMC Evol Biol 10: 274.