Introduction

In the past, only one or a few loci could be utilized to reconstruct evolutionary patterns due to the high cost and low efficiency of the Sanger DNA sequencing method. The development of high-throughput DNA sequencing technology in recent years has provided researchers with efficient ways to obtain large amounts of the information embedded in genomes (Shendure and Ji 2008). Application of this technology has generated comparative data from large numbers of independent loci in different species (Faircloth et al. 2013). Massively parallel sequencing (MPS) technology, also known as next-generation sequencing, has become an effective tool to conduct genomic studies of non-model organisms in recent years (Hudson 2008; reviewed in Ekblom and Galindo 2011). MPS has been used for sequencing the mitochondrial genomes of nematodes for medical parasitology (Jex et al. 2008, 2009a) and mollusk taxonomy (Feldmeyer et al. 2010; Williams et al. 2014), and for sequencing chloroplast genomes for population and phylogenetic studies (Cronn et al. 2008; Doorduin et al. 2011). With a suitable barcoding system and bioinformatics tools, multiplex MPS is an efficient method for obtaining mitochondrial genome sequences from multiple specimens simultaneously (Jex et al. 2009b; Timmermans et al. 2010). However, recent published cnidarian mitochondrial genomes were obtained using Sanger sequencing coupled with a standard primer-walking protocol (Wang et al. 2013; Figueroa and Baco 2014; Lin et al. 2014; Kitahara et al. 2014), which is an expensive and time-consuming process (Ansorge et al. 1997).

Among cnidarian genera, Acropora is the most diverse of reef-building corals, representing about 30 % of the total number of scleractinian species (Wallace 1999; Veron 2000). Skeletal form and structure are crucial morphological traits in the systematics of this genus, and its species are categorized into 19 groups (Veron and Wallace 1984; Wallace and Wolstenholme 1998; Wallace 1999) with designations for identification purposes only that do not reflect taxonomic affinity (Veron and Wallace 1984). Sympatric ranges (Wallace 1999) and synchronous spawning (Harrison et al. 1984; Babcock et al. 1986) in many of these species offer high potential for hybridization in this genus (Babcock 1995; Willis et al. 1997; Hatta et al. 1999), which has been highlighted as a major contributor to the reticulate evolution of Acropora species (Hatta et al. 1999; reviewed in Willis et al. 2006). Acropora species boundaries have been studied intensively from morphological traits but rarely from a molecular perspective (van Oppen et al. 2000b, 2002; Wolstenholme et al. 2003), and efforts have been focused on reconstructing a molecular phylogeny of Acropora species using several potential genetic markers (Odorico and Miller 1997; van Oppen et al. 2001; Richards et al. 2013). Even so, their phylogeny remains unresolved due to the introgression of loci from ongoing hybridization (reviewed in Wallace and Willis 1994; reviewed in Willis et al. 2006), and accurate phylogenetic discrimination of species still represents a challenge for coral reef biologists.

Mitochondrial genes have been commonly used to study molecular relationships among scleractinian corals (Romano and Palumbi 1996; Chen et al. 2002; van Oppen et al. 2001; Fukami et al. 2000, 2008; Kitahara et al. 2010; Huang et al. 2011). However, the low evolutionary rates of mitochondrial genes and regions limit the determination of Acropora phylogeny (van Oppen et al. 1999; Chen et al. 2009). So far, the mitochondrial genome of A. tenuis is the only one been published and annotated; the length of the genome is nearly 18,300 bp, contains thirteen protein-coding genes and two ribosomal RNA genes (van Oppen et al. 2000a), and may contain valuable phylogenetic information for solving intraspecific relationships. Thus, multiplex MPS could provide a convenient and effective approach to study the molecular phylogenetics of this genus. The large amount of genomic data acquired from MPS could provide great insight into understanding the evolution and diversity of Acropora.

In this study, eleven Acropora species representing different species groups were sequenced using the Illumina Solexa platform for mitochondrial genomes. The low taxonomic sampling scheme and maternally inherited locus used in this study may not be good enough to draw a clear picture of the phylogeny of Acropora. However, our aim was to demonstrate the feasibility of sequencing scleractinian mitochondrial genomes using multiplex MPS technology for phylogenomic analysis. Our mitochondrial phylogenomic tree resolved relationships between most species except A. humilis, A. digitifera, and A. florida. We expect that this sequencing approach could be successfully applied to future cnidarian genomic studies.

Materials and methods

Acropora samples and DNA extraction

Eleven Acropora species representing ten morphological species groups (Wallace 1999) were collected from Kenting National Park and Green Island, Taiwan, and the Great Barrier Reef, Australia (Table 1). A fragment (2–3 cm) from an individual from each species was collected and preserved in a modified guanidine solution as described in Fukami et al. (2004). Total genomic DNA was extracted from the solution by standard phenol/chloroform (1:1) extraction followed by ethanol precipitation and resuspended in nuclease-free double-distilled water.

Table 1 Eleven Acropora species representing different species groups used in this study. Acropora tenuis was used as a reference for genome mapping
Full size table

Long PCR-coupled mitochondrial genome sequencing

To obtain the complete mitochondrial genome from DNA samples, two pairs of long PCR primers designed for amplifying Acropora coral mitochondrial genomes were used (Monti12Slong-A/Monti16Slong-B and Monti16Slong-A/Monti12Slong-B; Tseng et al. 2005). The mitochondrial genome was amplified as two 9-Kb amplicons: Amplicon 1 was between 10,792–1538 and amplicon 2 between 1852–11,131 (the position in the reference sequence), forming an overlap in the putative control region and a gap in the rnl region. Long PCR was carried out in a 50-μl reaction mixture containing 1-μl DNA template (30–50 ng/μl), 0.2 μM of each primer, 20 mM of each dNTP, 10X LA PCR™ buffer, 2.5 mM MgCl2, and 2.5 U TaKaRa LA Taq™ (Takara Bio Inc.). PCR was carried out with a Thermo PX2 thermal cycler under the following conditions: one cycle of denaturing at 94 °C (1 min), annealing at 98 °C (10 s), and extension at 68 °C (15 min) for 30 cycles, followed by one cycle of final extension at 72 °C (10 min). PCR products were checked with a 1 % agarose gel to verify size and single-band yield. Gel purification was applied to remove nonspecific fragments, and final products were evaluated with a NanoDrop® Spectrophotometer. For library construction, amplicons of eleven samples were sheared and bar-coded for multiplexing purposes. Multiple libraries were pooled in a single-flow cell lane and sequenced on an Illumina Solexa Hi-Seq 2000 to generate 100-bp paired-end reads.

Sequence assembly and verification of consensus sequences

Paired reads were trimmed for adaptors and indexes by using Trimmomatic (Bolger et al. 2014). For the MPS dataset, sequence assembly was performed on CLC Genomics Workbench v4.5 software (CLC Bio, Denmark), an integrated platform for visualizing and analyzing MPS data. Sequence reads of eleven individuals were trimmed and mapped to the reference genome of Acropora tenuis (Genbank Accession No.: AF338425). Prior to mapping, the reference sequence was manipulated by removing the gap at the rnl region and shuffled to generate a continuous sequence. The assembled draft sequences were edited manually by adding ‘N’ to the gap and reshuffled into the correct order. The gap was then filled by conventional sequencing methods in order to generate complete mitochondrial genome sequences. Two datasets were used for mapping the Acropora mitochondrial genomes, including total output reads and reduced reads (i.e., 0.1 million reads of each species) to test the redundancy of output reads. Mapping results were verified by Sanger sequencing of the entire mitochondrial genomes of A. yongei and A. muricata. A total of 36 primers covering the entire mitochondrial genome were modified from Lin et al. (2011). PCR products were sequenced by an Applied Biosystems 3730 DNA analyzer and examined using SeqMan and MegAlign programs from Lasergene software (DNAstar Inc., USA). The differences between MPS and Sanger sequences were identified and used for verifying final consensus sequences. Sanger sequences were chosen when there were ambiguous sites observed between the two methods.

Multiple sequence alignment and phylogenetic analysis

The final consensus sequences of the eleven samples were aligned to the A. tenuis mitochondrial genome using MEGA 5.0 (Tamura et al. 2011). The repetitive sequences of the putative control region among Acropora species were trimmed. Annotated sequences were examined for informative and private sites using the Web server DIVEIN (Deng et al. 2010), which estimated genetic variation along the mitochondrial genome sequences. The proportion of nucleotide differences (p-distance) among genes and whole mitochondrial genomes was calculated. The model of nucleotide substitution that best fitted the sequence data was first estimated using Modeltest (Posada and Crandall 1998), followed by tree construction via maximum likelihood and neighbor-joining methods using MEGA 5.0 and a Bayesian method using MrBayes (Huelsenbeck and Ronquist 2001; Ronquist and Huelsenbeck 2003). Each analysis was run with 1000 bootstrap pseudoreplicates to estimate statistical support.

Results and discussion

Massively parallel sequencing and sequence assembly

Eleven Acropora mitochondrial genomes were sequenced using multiplex MPS technology on the Illumina Solexa platform and deposited in GenBank (NC_022823-NC_022833). Our approach was comprised of long-range PCR followed a species-specific indexing system to sequencing multiple mt genomes simultaneously. The four long PCR primers designed for amplifying the mitochondrial genomes of Montipora and Anacropora are applicable in Acropora species, as the rnl and rns region are generally conserved among the family Acroporidae (Tseng et al. 2005). Thus, bridging the gap in rnl regions posted to the sequence assembly was successful and less complicated than generating three amplicons for sequencing.

In total, 75,426,014 sequence reads were obtained from one Illumina lane (Table 2). Each bar-coded sample was successfully identified from total sequence reads with pipeline analysis software. Read number per sample varied from 3.0 to 9.3 million, with an average read length of 94 bases. Sequence trimming by removing low-quality sequences, ambiguous nucleotides, and sequences of specific length generated 74,393,870 high-quality reads, with average read length reduced to 81.07–89.12 bases. Sequence assembly greatly reduced the complexity of the assembling process with the A. tenuis reference genome. In the end, there were 73,142,848 reads with an average read length of 81.66–89.24 bases mapped to the A. tenuis reference sequence, representing 98.3 % of the trimmed reads. The high volume of sequence reads contributed 276,652,018–821,944,202 total bases to the sequence assembly, resulting in 96.9–99.2 % matching, and the mean depth of coverage (DOC) for the eleven mitochondrial genomes was approximately 12,000- to 44,000-fold. The overall genome mapping percentage can be generally considered high. Nevertheless, 0.8–3.1 % sequence reads could not be mapped to the reference genome. These singletons could be artifacts of the multiplex PCR due to the possible multi-template in the total genomic DNA samples (Takahiro 2003), but did not affect sequence mapping.

Table 2 Statistics of read mapping and the final consensus sequence lengths of eleven Acropora mt genomes
Full size table

The sequencing output of each species from the pooled sample had various read counts that ranged from 4 to 14.4 %. The different proportion of output sequence reads was probably due to the unequal molarity of the DNA template. However, each sample produced abundant sequence reads with high genome coverage, providing strong statistical support for the sequence assembly. Even when we reduced the dataset to 0.1 million reads, coverage was still 500-fold and achieved high DOC values for sequence assembly, indicating the great sequencing output that can be analyzed by the Hi-seq 2000 platform. This could be optimized by increasing the sample size across multiple taxa using a large-scale indexing system (Smith et al. 2010; Tu et al. 2012), which would also increase the efficiency for individual sequencing runs.

Sequence analysis

Sanger sequencing of sample A.yongei_KTAcr4 successfully generated a complete A. yongei mitochondrial genome. However, two primers failed to amplify the putative control region of sample A.muricata_KTAcr14, so primers were redesigned to solve this (Table S1). The alignment of the MPS and Sanger sequences exhibited two main discrepancies in both samples: position 1443–1459 and 2358–2382 in A.yongei_KTAcr4 and position 1445–1457 and 2356–2380 in A.muricata_KTAcr14. The discrepancies are probably due to the noise of the sequence end in MPS as the two positions are approximately at the primer-binding sites where there was no sequence coverage. We also discovered the inconsistent length of three poly-G sites and two insertion sequences in the putative control region of A.muricata_KTAcr14. Guanine nucleotide deletions were found at positions 4564–4565, 5788, and 12,631–12,632, while insertions occurred between 12,283–12,396 and 12,593–12,614. The first insertion has been confirmed as a repeat sequence 114 bases long and the second is a short insert of 22 bases (van Oppen et al. 2002). These could be the consequence of assembly errors cause by repeated sequences (Miller et al. 2010). Because of inconsistencies between the two sequencing approaches, we Sanger-sequenced the putative control regions of twelve Acropora species to check repetitive structures and insert sequences, among which only the reference sequence A.florida_KTAcr2 and A.yongei_KTAcr4 did not contain repeat sequences at position 12,283–12,396 and only the reference sequence, and A.yongei_KTAcr4 did not have an insert sequence at position 12,593–12,614. However, another insertion of an 87-base repeat sequence was observed between 12,631 and 12,717 in A.hyacinthus_KTAcr39.

The consensus sequence verified by the Sanger method shows that the repetitive structures of the putative control region appears to be variable in Acropora species. Consequently, some samples with repeat sequences could not be assembled correctly as the repeat regions did not exist in A. tenuis. Therefore, we replaced the control region sequences generated by MPS with those generated by Sanger sequencing. With the exception of the sequence noise in primer-binding regions, the rough sequencing accuracy rates for A. muricata and A. yongei were 99.97 and 100 %, respectively. The comparison of the two sequencing approaches shows that the limitations of MPS were mainly associated with the assembly of repeat elements (i.e., homopolymers) and regions that may not exist in the reference genome.

General features of the Acropora mt genome

The circular mitochondrial genome sequences of eleven Acropora species ranged between 18,342 and 18,566 bases (Table 2), depending on the presence of the insertion sequence in the putative control region. Acropora hyacinthus had the largest mitochondrial genome due to three insertion sequences found in the putative control region. A. yongei had the smallest mitochondrial genome as no insert was observed in the putative control region. The base compositions of A, C, G, and T were similar in the twelve mitochondrial genomes: 24.9–25.1, 13.7–13.8, 24.2–24.4, and 36.8–37.0 %, respectively. There was no significant difference in codon usage for the thirteen protein-coding genes across the twelve Acropora species (Chi-square test, p > 0.05). The most frequently used codon was UUU (Phenylalanine), followed by UUA (Leucine) and GUU (Valine). The multiple sequence alignment of the twelve Acropora species displayed well-aligned sequences over most of the mitochondrial genome, except for the putative control region where repetitive structures and short insert sequences appeared. The twelve genomes have identical gene arrangement after annotation, indicating that the gene arrangement among Acropora species may be conserved. In the Scleractinia, 27 of the 29 reported complete mitochondrial genomes have identical gene order, but two cases of rearrangement are known in the genera Lophelia and Madrepora. Thus, changes in the gene arrangement of Scleractinia are relatively uncommon (Lin et al. 2014). The gene arrangements that we found in Acropora were conserved within the genus and have been categorized as Type SII gene arrangement, which is the dominant type within scleractinians (Lin et al. 2014).

There were 110 informative sites and 50 private sites in the Acropora mitochondrial genome, and the most variable region was the putative control region with 60 informative sites and 19 private sites. Other informative sites were located in cob (12 sites), nad3 (8 sites), igr3 (6 sites), nad5 (6 sites), rnl (4 sites), nad1 (3 sites), igr2 (2 sites), nad6 (2 sites), and igr1, nad2, nad4, rns, cox3, cox2, and igr13 genes, all of the latter having single sites (Table 3). A pairwise distance analysis of the mitochondrial genome gave p-distance values for the 32 loci (in decreasing order): trimmed putative control region (2.19 %), nad3 (0.61 %), igr1 (0.46 %), igr7 (0.32 %), igr2 (0.31 %), igr3 (0.24 %), cob (0.18 %), igr13 (0.16 %), nad5 (0.14 %), nad6 (0.14 %), cox2 (0.13 %), nad1 (0.11 %), igr6 (0.11 %), rns (0.09 %), nad2 (0.07 %), rnl (0.05 %), nad4 (0.05 %), cox3 (0.04 %), and cox1 (0.02 %); and the remaining thirteen loci (trnM, group I intron5, igr4, igr5, atp6, igr8, igr9, nad4L, igr10, igr11, trnW, igr12, and atp8) have 0 % of the p-distance values (Fig. 1). The overall p-distance estimated from the Acropora mitochondrial genome was 0.22 %, and the evolutionary rate based on the predicted divergence time of two geographically distinct Acropora species from Chen et al. (2009) was calculated to be approximately 0.011–0.063 % Mya−1, indicating a slow evolutionary rate of Acropora species. These results are consistent with previous studies based on a single gene or region comparisons (van Oppen et al. 1999; Fukami et al. 2000; van Oppen et al. 2001; Chen et al. 2009). Several mechanisms have been proposed to explain the extremely low substitution rate found in anthozoans, including a highly efficient DNA repair system, selection on mt DNA genes, and short divergence times (Shearer et al. 2002). These factors could be influence apparent substitution rates within anthozoan mitochondrial genomes at different levels and genetic divergences among species.

Table 3 List of SNP sites observed along the whole mt genome in the twelve Acropora species
Full size table
Fig. 1
figure 1

Pairwise p-distance (%) of the whole mt genome and different loci/genes among twelve Acropora species. The error bar represents the ±SD value

Full size image

Phylogenetic relationships among the twelve Acropora species

Previous studies suggested a much stronger phylogenetic structure in datasets, excluding repeat regions (Van Oppen et al. 2001), so we excluded repeat regions within the control region and mitochondrial genome for phylogenetic analyses. After alignment, the length of the putative control region and mitochondrial genome was 1086 nucleotides and 18,569 nucleotides, respectively. Each tree was constructed under different substitute models: the HKY+G model for the putative control region dataset and HKY+G+I model for the whole mitochondrial genome dataset. The maximum likelihood (ML), Bayesian, and neighbor-joining trees constructed based only on the control region yielded topology similar to the tree constructed from the whole mitochondrial genome, but the latter displayed higher levels of resolution (Fig. 2a, b). This indicates that the data for the entire mitochondrial genome may provide slightly more phylogenetic information than the putative control region alone. Shearer et al. (2002) first revealed slow mitochondrial sequence evolution in Anthozoa by comparing three mitochondrial genes across different taxa. Recent work suggests that the mitochondrial genomes of some scleractinians evolve 5× slower than their nuclear genomes (Chen et al. 2009) and 50–100× slower than the mitochondrial genomes of most other animals (Hellberg 2006). This feature makes it difficult to apply genetic bar coding to anthozoans (Huang et al. 2008; McFadden et al. 2011) for resolving inter-species relationships within the genus Acropora (van Oppen et al. 2001; Richards et al. 2013).

Fig. 2
figure 2

a Maximum likelihood tree analyzed based on the whole mt genomes of twelve Acropora species. Branch values indicate the bootstrap value of 1000 replicates for maximum likelihood, Bayesian inference, and neighbor-joining, respectively. Clade nomenclature in blue on the right reflects Richards et al. 2013. Clade nomenclature in black on the left reflects van Oppen et al. 2001. Species groups are given after each species’ name. b Maximum likelihood tree analyzed based on the control region of twelve Acropora species. Bootstrap values higher than 50 are shown at the nodes

Full size image

The twelve mitochondrial genomes used in this study belong to ten species groups. The ML tree revealed nine lineages among the twelve Acropora species, with eight out of nine lineages corresponding to their species groups except for the humilis and florida groups (Fig. 2a). A. tenuis and A. yongei, both belonging to the selago group, formed a monophyletic clade at the base with high supporting values concordant with previous studies (van Oppen et al. 2001; Richards et al. 2013). Acropora nasuta, A. divaricata, and A. muricata were included in clade III in the phylogenetic tree published by van Oppen et al. (2001), but our tree revealed a different pattern where A. murica clustered with A. aspera, A. florida, and A. humilis, which were in clade IV of van Oppen et al. (2001). In addition, the position of A. horrida differed from previous studies (Richards et al. 2013), instead being a basal lineage that is sister to A. hyacinthus in the present study. In addition, the mt genome does not have enough resolution to resolve the relationships among A. aspera, A. florida, and A. humilis. These three species belong to two species groups, which have close phylogenetic relationships to each other (Fukami et al. 2000; van Oppen et al. 2001; Wolstenholme 2004; Richards et al. 2013). This lack of resolution may indicate introgression or incomplete lineage sorting (van Oppen et al. 2001; Vollmer and Palumbi 2002; Wolstenholme et al. 2003; Nakajima et al. 2012; Richards et al. 2013). In our case, we cannot exclude these two evolutionary mechanisms, which may have acted on some Acropora species either simultaneously or at different times to contribute to the high similarities that we observed across Acropora mitochondrial genomes.

Conclusion

This study was not tending to resolve the phylogentic relationships among the genus Acropora which known to have reticulate evolution with low number of species represented in each morph groups. Instead, we demonstrated the feasibility of using MPS next-generation sequencing technology on scleractinian mitochondrial genomes. Comparing to traditional long PCRs with primer walking to obtain whole mitochondrial sequence, high-throughput sequencing method does not need to design species-specific primers (~20 primers) for primer walking during sequencing process, and it is more cost-efficient than traditional method. The barcoding system allowed the sequencing of multiple individuals simultaneously at lower processing costs and times while producing highly accurate sequence assemblies. Sequence errors and the difficulty in genome assembly were mainly due to homopolymeric regions and repetitive structures in the putative control region, and ambiguous sites still need to be verified by conventional Sanger sequencing. Our genome comparison and genetic analysis showed similar characteristics in mitochondrial genomes and low evolutionary rates in Acropora species. The relationships among closely related species remain unresolved because of the insufficient phylogenetic information presented in their mitochondrial genomes. Nonetheless, with more cnidarian mitochondrial genomes released, multiplex next-generation sequencing technology can provide an effective platform for molecular systematics and is practical for other cnidarian phylogenomic studies.