The Large Mitochondrial Genome of Symbiodinium minutum Reveals Conserved Noncoding Sequences between Dinoflagellates and Apicomplexans.

Even though mitochondrial genomes, which characterize eukaryotic cells, were first discovered more than 50 years ago, mitochondrial genomics remains an important topic in molecular biology and genome sciences. The Phylum Alveolata comprises three major groups (ciliates, apicomplexans, and dinoflagellates), the mitochondrial genomes of which have diverged widely. Even though the gene content of dinoflagellate mitochondrial genomes is reportedly comparable to that of apicomplexans, the highly fragmented and rearranged genome structures of dinoflagellates have frustrated whole genomic analysis. Consequently, noncoding sequences and gene arrangements of dinoflagellate mitochondrial genomes have not been well characterized. Here we report that the continuous assembled genome (∼326 kb) of the dinoflagellate, Symbiodinium minutum, is AT-rich (∼64.3%) and that it contains three protein-coding genes. Based upon in silico analysis, the remaining 99% of the genome comprises transcriptomic noncoding sequences. RNA edited sites and unique, possible start and stop codons clarify conserved regions among dinoflagellates. Our massive transcriptome analysis shows that almost all regions of the genome are transcribed, including 27 possible fragmented ribosomal RNA genes and 12 uncharacterized small RNAs that are similar to mitochondrial RNA genes of the malarial parasite, Plasmodium falciparum. Gene map comparisons show that gene order is only slightly conserved between S. minutum and P. falciparum. However, small RNAs and intergenic sequences share sequence similarities with P. falciparum, suggesting that the function of noncoding sequences has been preserved despite development of very different genome structures.

Introduction

Mitochondrial (mt) genomes are considered characteristic of eukaryotic cells (Lang et al. 1999; Gray et al. 2004). Although eukaryotic mt genomes are believed to have arisen from alpha-proteobacteria, extant eukaryotes possess either linear or circularized mtDNA with varied and reduced gene content (Lang et al. 1999). For example, most metazoan mt genomes are 13–20 kb, compact, circular molecules, encoding 12–13 proteins, 24–25 transfer RNAs (tRNAs), and 2 ribosomal RNAs (rRNAs). On the other hand, linear mtDNAs with terminal repeats (putative telomeres) have also been found in many species, such as the yeast, Candida, and the ciliate, Tetrahymena (Lang et al. 1999; Rycovska et al. 2004).

The ciliates, Paramecium aurelia, Tetrahymena pyriformis, and Tetrahymena thermophile, have linear mt genomes of 40–47 kb, which contain approximately 50 genes (Burger et al. 2000; Gray et al. 2004). In contrast, only three protein-coding genes (cox1 [cytochrome oxidase subunit I], cox3 [cytochrome oxidase subunit III], and cob [cytochrome b]) and fragmented rRNAs (LSU, large subunit; SSU, small subunit) have been identified in mt genomes of apicomplexans and dinoflagellates (Feagin 1992; Nash et al. 2008; Vaidya and Mather 2009). Recent work on the mt genome of the malaria parasite, Plasmodium falciparum, found additional fragmented rRNAs and uncharacterized small RNAs (Feagin et al. 2012). Diverse linear mt genomes have been reported in apicomplexans (Vaidya and Mather 2009). For example, Plasmodium has mt genomes in tandemly repeated arrays with a unit length of approximately 6 kb. On the other hand, Babesia and Theileria have monomeric mt genomes (Hikosaka et al. 2012). Previous work on dinoflagellate mt genomes has suggested complex organization, with extensive recombined and fragmented gene copies (Waller and Jackson 2009). Fragmented mt genomes and/or transcripts have been reported in at least 25 dinoflagellate taxa (table 1). The foregoing studies have confirmed that dinoflagellate mtDNA includes cox1, cox3, cob and fragmented rRNAs, and have detailed unusual mRNA characteristics (reviewed by Nash et al. 2008; Waller and Jackson 2009). Extensive RNA editing of the three protein-coding genes (Lin et al. 2002; reviewed in Lin et al. 2008) and trans-splicing of cox3 have been reported (Jackson et al. 2007; Imanian et al. 2012; Jackson and Waller 2013). However, transcripts from the basal dinoflagellates, Hematodinium sp. and Oxyyrhis marina, did not show RNA editing (Slamovits et al. 2007; Jackson et al. 2012). Trans-splicing of cox3 was not found in O. marina (Slamovits et al. 2007). Losses of canonical start and stop codons have also been suggested (Norman and Gray 1997; Jackson et al. 2012; reviewed in Nash et al. 2008). On the other hand, analyses of noncoding sequences have been frustrated by high recombination rates in these genomes (Patron et al. 2005; reviewed in Waller and Jackson 2009). In addition, some reports have suggested that the total dinoflagellate mt genome size is likely to be large (Waller and Jackson 2009; Shoguchi et al. 2013), and the dinoflagellate mt genome is thought to be one of the most complex (Nash et al. 2008). For example, it is estimated that 85% of the mt genome in Amphidinium carterae is noncoding (Nash et al. 2007). Although inverted repeat (IR) elements in intergenic regions have been reported, functions of these elements are unknown (Waller and Jackson 2009). Thus, it has been assumed that each alveolate lineage developed different mt genomic structure (Slamovits et al. 2007). Interestingly, recently reported mt genomes of colponemids, an early alveolate lineage, suggest that the ancestral alveolate genome encoded a typical mt gene set (Janouškovec et al. 2013).

Table 1

Summary of the Papers Reporting mt Genomes and/or Transcriptomes in Dinoflagellates

Species name	References
Akashiwo sanguinea	Zhang et al. (2008)
Alexandrium catenella	Kamikawa et al. (2007, 2009)a
Alexandrium tamarense	Zhang and Lin (2005); Zhang et al. (2008)
Amphidinium carterae	Nash et al. (2007)a; Jackson and Waller (2013)
Crypthecodinium cohnii	Norman and Gray (1997, 2001)a; Lin et al. (2002); Jackson et al. (2007)
Dinophysis acuminata	Zhang et al. (2008)
Durinskia baltica	Imanian and Keeling (2007); Imanian et al. (2012)a
Gonyaulax polyedra	Chaput et al. (2002)
Hematodinium sp.	Jackson et al. (2012)a
Heterocapsa triquetra	Patron et al. (2005)
Karenia brevis	Zhang et al. (2008)
Karlodinium micrum	Zhang and Lin (2005); Jackson et al. (2007)a
Kryptoperidinium foliaceum	Imanian and Keeling (2007); Imanian et al. (2012)a
Oxyrrhis marina	Slamovits et al. (2007)a
Pfiesteria piscicida	Lin et al. (2002); Zhang and Lin (2002)
Pfiesteria shumwayae	Zhang and Lin (2005)
Prorocentrum cassubicum	Zhang et al. (2008)
Prorocentrum micans	Zhang and Lin (2005); Zhang et al. (2008)
Prorocentrum minimum	Lin et al. (2002)
Protoceratium reticulatum	Zhang et al. (2008)
Pseudopfiesteria shumwayae	Zhang et al. (2008)
Scrippsiella sp.	Zhang et al. (2008)
Scrippsiella sweeneyae	Zhang et al. (2008)
Symbiodinium microadriacticum	Zhang and Lin (2005); Zhang et al. (2008)
Symbiodinium sp.	Zhang and Lin (2005); Zhang et al. (2008); Jackson and Waller (2013)

amt genome sequences were reported.

Our previous work on the endosymbiotic dinoflagellate, Symbiodinium minutum, has confirmed the presence of unusual nuclear (Shoguchi et al. 2013) and plastid genomes (Mungpakdee et al. 2014). In addition, this species may possess high mt genome copy numbers (Shoguchi et al. 2013). In this study, by analyzing the wealth of sequence data, we characterized the Symbiodinium mt genome and transcriptomes, including many noncoding sequences, and we compared them with mt genomes of Plasmodium and dinoflagellates. Assembly of fragmented DNA in general is technically difficult, but physical link information from fosmid end sequencing greatly aided mt genome assembly. Our analysis reveals conserved, noncoding sequences during myzozoan (apicomplexans and dinoflagellates) mt genome evolution. In addition, Symbiodinium is a large genus, classified into nine major clades (Coffroth and Santos 2005; Pochon et al. 2014); therefore, the complete Symbiodinium mt genome will be an important resource to study populations and environmental adaptations using genomic approaches (Shinzato et al. 2014).

Results and Discussion

The De Novo Assembled mt Genome of S. minutum

To reconstruct the mt genome of S. minutum, 20 analyses using only high coverage illumina paired-end reads (DNAseq) were performed (see also Materials and Methods). Two candidate mt contigs having more than 100× read coverage were obtained (19,577 and 291,416 bp) (accession numbers: LC002801 and LC002802) by 49-kmer assembly. Physical link information from fosmid paired-end sequences (FPESs) confirmed contig structures from computational sequence assembly (fig. 1A). In addition, joining of the 3′-end of the approximately 19-kb contig and the 5′-end of the approximately 291-kb contig was supported by FPES. BLAST (Basic Local Alignment Search Tool) searches showed that the approximately 19-kb contig contains the cox1 gene. The approximately 291-kb contig contains cob, cox3, and fragments of the LSU rRNA gene. Gene locations are explained in detail hereafter. Comparisons between the two contigs and the S. minutum genome assembly v1.0, using mapped FPES, showed that only scaffold 7473 (length: 15,538 bp) from genome assembly v1.0 (Shoguchi et al. 2013) was joined to the approximately 291-kb contig by more than 80 FPESs (fig. 1A). This suggested that nearly 40 kb of mtDNA had been identified. Estimation of the lengths of the two gaps was difficult. Accordingly, two bases (NN) were arbitrarily added between the two contigs and between the 291-kb contig and scaffold 7473 (see also fig. 1A). Comparison of the assembled mt genome with FPESs implies the presence of multiple recombinant mtDNA fragments, but our analysis suggests that S. minutum has a continuous mt genome of approximately 326 kb. Only simple repeats with fewer than 8 bp (∼1.49%) and low complexity (∼0.23%) were found in the mt genome assembly. The 49-bp repeats, which might be relevant to the assembly process, occurred fewer than four times in the approximately 326 kb.

F

A mitochondrial genome and transcripts in S. minutum. (A) The assembled mt genome of S. minutum showing the high copy number. Arrows show two contigs and one scaffold (scaffold 7473), which are joined by paired-end sequences of fosmid clones and are labeled “NN” because of indeterminate distances. These constitute a scaffold of 326,535 kb. The upper graph indicates the high coverage of illumina reads that were mapped onto the scaffold. The lower plot shows clone coverage by fosmid paired-end mapping, partially supporting the accuracy of the assembly. (B) Transcriptomes from mtDNA and possible ends. The S. minutum mt genome with predicted genes is shown in upper region. Genes above or below the line indicate the transcription direction. Protein-coding genes are in red. Detailed gene map information is shown in supplementary figure S4, Supplementary Material online, and table 3. The upper graph shows coverage of RNAseq reads from illumina libraries that are enriched RNAs with polyA sequences. High coverage reads are found on cox1, cox3, fragment E of the ribosomal LSU, and cob. Only reads with poly A or T (more than four) are shown on middle graph, suggesting polyadenylated transcripts and potential 3′-ends. The lower graph displays reads from the TSS library, which is enriched RNA with 5′ cap structures, indicating the presence of multiple 5′-ends.

Table 3

Predicted Genes in Mitochondrial Genome of Symbiodinium minutum

Gene	Subunit Order	Predicted Location	Orientation to Scaffold	Similarity to Plasmodium falciparum Gene (%)a
cox1		5809–7248	+	916/1,441 (63)
cox3		186587–187332	+	405/771 (52)
cob		197602–198718	+	688/1131 (60)
SSUA	S4	177279–177354	+	57/80 (71)
SSUB	S6	236311–236394	+	48/86 (55)
SSUD	S10	176902–176959	−	37/63 (58)
SSUE	S11	221699–221724	+	19/26 (73)
SSUF	S12	170409–170456	+	31/48 (64)
LSUA	L1	105339–105493	−	89/158 (56)
LSUB	L3	38813–38831	−	17/19 (89)
LSUC	L4	73563–73580	−	17/18 (94)
LSUD	L8	222761–222836	+	55/76 (72)
LSUE	L9	193381–193573	+	149/195 (76)
LSUF	L11	279828–279907	+	55/80 (68)
LSUG	L12	278801–278900	−	74/100 (74)
RNA1	L6	317063–317147	+	54/88 (61)
RNA2	L2	60688–60729	+	26/42 (61)
RNA3	L7	34514–34593	−	45/81 (55)
RNA4		220439–220506	−	39/68 (57)
RNA5	S9	138204–138280	+	48/80 (60)
RNA6	L15	14596–14626	−	27/33 (81)
RNA7		56199–56266	+	53/69 (76)
RNA8	S5	106227–106279	−	30/53 (56)
RNA9	S8	281819–281866	+	34/50 (68)
RNA10	L13	217165–217255	+	59/92 (64)
RNA11	L5	73434–73479	−	29/46 (63)
RNA12	S2	191852–191892	+	30/41 (73)
RNA13	L10	256591–256614	−	17/24 (70)
RNA14	S1	31151–31177	+	21/27 (77)
RNA15		253421–253447	−	19/27 (70)
RNA16		51961–51991	−	21/31 (67)
RNA17	S3	76949–76985	+	24/37 (64)
RNA18	L14	300291–300312	+	17/22 (77)
RNA19	S7	308063–308089	−	21/27 (77)
RNA20		242197–242225	+	20/29 (68)
RNA21		122845–122864	−	16/20 (80)
RNA22		57665–57699	+	24/35 (68)
RNA23t		186459–186487	−	20/29 (68)
RNA24t		23626–23665	−	27/40 (67)
RNA25t		302697–302717	+	16/21 (76)
RNA26t		186374–186415	−	29/43 (67)
RNA27t		278662–278712	−	36/52 (69)

Note.—Gene names and subunit order refer to Feagin et al. (2012). L and S indicate LSU and SSU, respectively.

aGenes are from P. falciparum M76611 (Feagin et al. 2012). Alignments are shown in supplementary figure S3, Supplementary Material online.

Transcriptomes of Symbiodinium mt Coding Genes

RNAseq reads were mapped onto the continuous genome (fig. 1B), revealing high coverage of cox1, cob, cox3, and the fragmented LSU gene (fig. 1B). Mapped data indicated the possibility of polycistronic expression. Mapping of reads with polyA or T in the 5′ sequence showed four major peaks for three protein-coding genes and the fragmented LSU. The highest peak is likely to be from the fragmented LSU gene, suggesting high expression and enhanced polyadenylation during RNA processing. Reads mapped from the transcription start site (TSS) library showed high coverage of multiple sites, suggesting multiple 5′ cleavage sites and transcripts with modified 5′-phosphate groups (fig. 1B). Symbiodinium minutum mt transcripts did not show evidence of RNA processing, such as 5′ oligo (U) caps of O. marina mt transcripts (Slamovits et al. 2007).

Edited RNA sites for transcripts of cox1, cob, and cox3 were investigated using comparisons between assembled genomes and transcripts. A to G editing was found in 61% of the 72 sites, showing conservation between dinoflagellates (table 2; Lin et al. 2008). In addition, patterns of RNA editing-mediated amino acid substitutions correspond to previous report about another species (supplementary fig. S1, Supplementary Material online; Lin et al. 2008).

Table 2

RNA Editing Types in Three Mitochondrial Genes of Symbiodinium minutum

Gene	Transcriptome ID	No. of Edits (%)a	Editing Type							No. of Amino Acid Substitutions (%)
			A/G	G/A	C/U	U/C	G/C	U/G	A/C
cox1	symbB1.comp234_c0_seq1	29/1,455 (2.0)	18	0	3	4	2	1	1	24/485 (4.9)
cox3	symbB1.comp4_c1_seq1, symbB1.EST_k37c20_2341	24/774 (3.1)	18	1	1	4	0	0	0	23/258 (8.9)b
cob	symbB1.EST_k37c20_4808	19/1,062 (1.8)	8	0	4	4	2	1	0	19/354 (5.4)b

aEdits in predicted coding sequences were counted.

bIncluding a signal from stop codon. For cox3, adenylated sequences may be used as stop signals.

Another unusual feature of dinoflagellate mt genes is the lack of canonical start and stop codons to direct the initiation and termination of translation (Norman and Gray 2001; Jackson et al. 2012; reviewed in Nash et al. 2008). We have characterized start and stop codons of S. minutum mt genes using manual alignments between genomic and transcriptomic sequences (supplementary fig. S2, Supplementary Material online). We found AUA (Ile) and AUU (Ile) at the 5′-end of cox1 and AUU (Ile) in cox3. They are also candidates for start codons as mt genes in both ciliates and apicomplexans use AUA and AUU for this purpose (Feagin 1992; Edqvist et al. 2000). The cob gene in S. minutum contained both canonical start and stop codons; cox3 contained a canonical stop codon resulting from polyA addition (supplementary fig. S2, Supplementary Material online), as reported in other dinoflagellates (Waller and Jackson 2009). cox1 does not contain a stop codon (supplementary fig. S2, Supplementary Material online).

Noncoding RNA Genes and Gene Map

In the apicomplexan P. falciparum, 39 RNA genes, including fragmented rRNA LSUs (15), SSUs (12), and uncharacterized small RNAs (12), have been identified (Feagin et al. 2012). These rRNA fragments are not arranged linearly, but synteny was conserved in Plasmodium (Vaidya and Mather 2009). It is suggested that this fragmentation occurred in the common ancestor of apicomplexans and dinoflagellates (Slamovits et al. 2007; Jackson et al. 2012). To predict fragmented rRNAs in the S. minutum mt genome, the most similar regions from P. falciparum (Feagin et al. 2012) were surveyed and aligned (table 3 and supplementary fig. S3, Supplementary Material online). Their alignments with interspersed regions of the S. minutum mt genome showed more than 50% similarity, corresponding to RNAseq reads from TSS libraries, and indicating the presence of multiple RNAs (fig. 1B). Comparisons of secondary structures for aligned sequences of RNA genes showed that the majority of predicted genes have stem-loop structures (supplementary fig. S3, Supplementary Material online). Thus, the assembled S. minutum genome contains orthologs to genes in the P. falciparum mt genome (table 3 and supplementary fig. S3, Supplementary Material online).

tRNA genes were not found in the S. minutum mt genome using tRNA scan. So far no studies of dinoflagellate or apicomplexan mtDNAs have identified any tRNA genes, suggesting that tRNAs have been imported from the nuclear genome, as was reported for the apicomplexan, Toxoplasma gondii (Esseiva et al. 2004).

Interestingly, two LSU fragments, L4 and L5, map onto neighboring regions of the S. minutum mt genome with fewer than 100 bp between them (table 3 and supplementary fig. S4, Supplementary Material online). The L4–L5 arrangement in S. minutum, corresponding to the continuous large rRNA sequence order, appears to be evolutionarily conserved. Secondary structure prediction for L4 and L5 sequences yields a very stable, double-stranded form; however, the predicted structure varies depending on which secondary structure prediction program is used (supplementary fig. S3 and , Supplementary Material online). Genes with unknown functions, RNA 23 t and RNA 26 t, are close, separated by only an approximately 40-bp intergenic sequence. Conserved, fragmented LSUs and SSUs may be cleaved accurately by small RNAs, such as RNA 23 t and RNA 26 t.

Comparisons of mt gene arrangements between Symbiodinium and Plasmodium showed only one microsyntenic region, which has the same gene arrangement on S12 and S10 (fig. 2 and supplementary fig. S4, Supplementary Material online), suggesting that the fragmentation occurred in the common ancestor of apicomplexans and dinoflagellates and that genome rearrangements in these lineages were very frequent. Thus, this basic information is valuable for possible functional analysis of dinoflagellate mt genomes.

F

Mitochondrial gene order comparisons between S. minutum and P. falciparum. Genes from the S. minutum mt genome (∼326 kb) to the upper are joined to those of P. falciparum. The gene order of S10 and S12 was the same in mt genomes of both P. falciparum and S. minutum (aqua lines), showing minimal conservation of gene order. Sequence similarities from intergenic regions of the P. falciparum mt genome are indicated by orange lines. Details for the S. minutum mt genome map are shown in supplementary figure S4, Supplementary Material online.

Unknown Noncoding Regions and Possible Expansion in the Dinoflagellate Lineage

Our analysis confirms that noncoding sequences of the Symbiodinium mt genome have been expanded, raising the question as to where the expanded sequences originated. Enormous expansion of intergenic content in the mt genomes of seed plants (200–2,900 kb) has been reported (Mower et al. 2012), and repeated proliferation of “selfish” DNA has contributed overwhelmingly to these expansions (Chaw et al. 2008). Highly repetitive sequences were not found in the mt genome of S. minutum. In addition, when compared with other dinoflagellate mt sequences, the S. minutum mt sequences suggest additional RNA fragments or pseudogenes (supplementary fig. S5, Supplementary Material online), but these were not identified.

To find conserved secondary structures of potential RNA genes, each chopped 300-bp sequence from the mt genome was employed as a query sequence to perform RNA homology searches using Infernal (Nawrocki and Eddy 2013). Twenty-two sequences showed similarity to reported sequences in the Rfam database, including microRNAs and LSUs (Nawrocki et al. 2015) and suggest the presence of unknown RNA genes (supplementary table S1, Supplementary Material online). Unexpectedly, possible secondary structures for these genes included a stem of more than 20 bp (supplementary fig. S6, Supplementary Material online). Although we did not find sequence similarities to reported small, IRs of dinoflagellate mtDNA (reviewed in Waller and Jackson 2009) in the S. minutum mtDNA, the result suggests that structures comprising IRs are conserved characters among dinoflagellates mtDNAs.

Transcriptional control of the alveolate mt genome is not clear (Gray et al. 2004; Waller and Jackson 2009). Our RNAseq reads support the possibility of polycistronic expression. To detect conserved intergenic regions, similarities to the six intergenic sequences of P. falciparum and mtDNA sequences of dinoflagellates were surveyed. Interestingly, our analysis showed that intergenic sequences of P. falciparum have similarities to the mt genome of S. minutum at the same level as comparisons between rRNA sequences. (fig. 2 and supplementary fig. S7, Supplementary Material online).

It is very interesting to examine how organelle genomes of dinoflagellates evolved different structures, given the large variety of structures between plastid and mt genomes of Symbiodinium. Although the plastid genome has undergone reconfiguration to a compact DNA minicircle (∼1.8–3.0 kb) with its own regulatory regions (Mungpakdee et al. 2014), the mt genome appears greatly expanded and fragmented.

Materials and Methods

Genome Assembly

Genomic DNA sequences from cloned and cultured S. minutum Mf1.05 b.01 (clade type: B1) were obtained previously (Shoguchi et al. 2013). To obtain longer mt contigs, using Velvet (version 1.2.08) software (Zerbino and Birney 2008) for illumina paired-end reads (6.8 Gb from a polymerase chain reaction-free library, accession number: DRX003100), calculations of 20 patterns were performed on a combination of ten kmer parameters (kmer size: 27, 31, 35, 37, 41, 43, 45, 47, 49, and 51) using two coverage threshold values (>50 and >100). The 151,553 FPESs with approximately 2.1× coverage of the whole genome (Hattori et al. 2000; Shoguchi et al. 2013) (deposited in DNA Data Bank of Japan, accession numbers: GA453877–GA605429) were mapped onto mt contigs using BLASTN and relationships between contigs were examined.

The assembled mt genome was also compared with FPESs and assembly version 1 scaffolds from contigs based on Roche 454 reads (GenBank ID: BASF00000000.1; see also http://marinegenomics.oist.jp/genomes/gallery, last accessed September 1, 2014) (Koyanagi et al. 2013; Shoguchi et al. 2013).

Transcriptome Mapping

Transcriptome reads deposited at DRP000944 were mapped onto the assembled mt genome using Bowtie 2 (version 2.1.0) software (Langmead and Salzberg 2012). Detection of RNA editing for cob, cox1, and cox3 was basically performed in the same manner as for plastid transcripts (Mungpakdee et al. 2014). Differences between DNA and RNA were detected by aligning transcriptome contigs to a scaffold. RNAseq reads were mapped onto mt transcriptome contigs using TopHat (Trapnell et al. 2009) and accuracy was confirmed. Reads from a TSS library (Yamashita et al. 2011; Shoguchi et al. 2013) were mapped using Bowtie 2, as described in Mungpakdee et al. (2014).

Data Analysis Software and Sequences Used for Comparisons

Repeats in the mt genome assembly were detected using RepeatMasker in default mode (http://www.repeatmasker.org). tRNAscan-SE (with default parameters in organellar mode) (Schattner et al. 2005) was used to find tRNA genes in the mt genome and transcriptome contigs. Sequence alignments between Symbiodinium and Plasmodium were performed using GENETYX-MAC version 17 and BLASTN. RNAfold (http://rna.tbi.univie.ac.at/cgi-bin/RNAfold.cgi; Zuker and Stiegler 1981) and CentroidFold (http://www.ncrna.org/centroidfold/; Sato et al. 2009) with initial settings were used for prediction of RNA secondary structure. We prepared chopped 300-bp sequences from the mt genome with 100-bp overlap sequences. RNA homology searches for each of the 300-bp sequences were performed using infernal (INFERence of RNA ALignment) using default parameters (Nawrocki and Eddy 2013). The probability cutoff value (E value) was set at 0.001.

mt genome sequences used for comparisons have the following accession numbers: “HE610722–HE610773” for Hematodinium sp., “AB265207–AB265210 and AB374233–AB374251” for Alexandrium catenella, “JX001584–JX001600” for Durinskia baltica, “EF442995–EF443047 and AM773790–AM773803” for Karlodinium micrum, “JX001601–JX001608” for Kryptoperidinium foliaceum, “EF680822–EF680839” for O. marina, “M76611” for P. falciparum, “KF651061” for Alveolata sp. 1 JJ-2013 (colponemid-like Peru), which is a new species, Acavomonas peruviana (Tikhonenkov et al. 2014), and “AF396436” for T. thermophile.

Supplementary Material

Supplementary figures S1–S7 and table S1 are available at Genome Biology and Evolution online (http://www.gbe.oxfordjournals.org/).