Introduction

In animals, mitochondrial genomes are typically small (15–20 kb), are circular, and encode 37 genes: 2 ribosomal RNAs, 22 tRNAs, and 13 protein coding genes (Wolstenholme 1992; Boore 1999). There are both a large and a small ribosomal RNA (rrnL and rrnS), and one tRNA for every amino acid except leucine and serine, which have two genes. The protein coding genes consist of ATP synthase subunits 6 and 8 (atp6 and atp8), cytochrome oxidase subunits 1–3 (cox1-cox3), cytochrome B (cob), and NADH dehydrogenase subunits 1–6 and 4L (nad1-6 and nad4L). Typically, animal mitochondrial genomes also have a large noncoding region that contains elements that control the replication and transcription of the genome (Wolstenholme 1992). With a few exceptions, the gene content for animal mitochondrial genomes is well conserved (Boore 1999), but gene order is more variable (Boore and Brown 1998; Moret et al. 2001).

Complete sequences of mitochondrial genomes are potentially useful for phylogenetic analysis. Gene order, DNA, and amino acid sequences can be used for phylogenetic reconstruction. Mitochondrial genome sequences are becoming prevalent in GenBank; however, many groups of organisms are unrepresented, resulting in gaps in taxon sampling for phylogenetic analysis, and gaps in our knowledge of mitochondrial genomes. This presents difficulties in the development of molecular markers for phylogenetic analysis.

Acanthocephala is a group of parasitic organisms with ambiguous phylogenetic status. Phylogenies based on both molecular and morphological data have shown that Acanthocephala is closely affiliated with Rotifera either as sister taxa or as a subtaxon (Winnepenninckx et al. 1995; Abouheif et al. 1998; Zrzavy et al. 1998; Garey et al. 1998; Herlyn et al. 2003). The placement of rotifers and acanthocephalans together in Protostomia differs among hypotheses (Fig. 1). These taxa have been placed in Platyzoa, which includes Cycliophora, Gnathostomulida, Gastrotricha, and Platyhelminthes, and is a sister taxon to Lophotrochozoa, which includes Mollusca, Annelida, Brachiopoda, Phoronida, Entoprocta, Nemertea, and Sipuncula (Giribet et al. 2000; Giribet 2002) (Fig. 1A). A second hypothesis based on morphological data groups Acanthocephala, Rotifera, Gnathostomulida, and Chaetognatha and places them in Spiralia, a large group that includes such taxa as Platyhelminthes, Mollusca, Annelida, Arthropoda, and Entoprocta (Nielsen 2001) (Fig. 1B).

Figure 1
figure 1

Phylogenetic hypotheses of metazoan relationships. Asterisks by taxon labels indicate those included in the phylogenetic analysis in this study. A Redrawn from Giribet (2002). B Redrawn from Nielsen (2001).

Full size image

To date, a mitochondrial genome has not been sequenced for any species of Acanthocephala. Platyhelminths are likely the most closely related organisms for which complete mitochondrial genomes have been sequenced; however, they are distantly related according to most phylogenetic hypotheses (Fig. 1). We determined the sequence and structure of the mitochondrial genome of an acanthocephalan species, Leptorhynchoides thecatus, and performed a phylogenetic analysis of Metazoa using amino acid sequences of the protein coding mitochondrial genes. Although many critical taxa still have not been sequenced, comparisons of mitochondrial genomes among lower Metazoa provide a new perspective on the phylogenetic placement of these rarely studied organisms. This study represents the first mitochondrial genome of an acanthocephalan to be sequenced and will allow further studies of systematics, population genetics, and genome evolution.

Materials and Methods

Specimens of Leptorhynchoides thecatus were collected from the intestines of fish from Atkinson Reservoir in Holt County, Nebraska (42°32′36′′N, 98°58′22′′W). Worms were preserved in 95% ethanol and kept at −20°C prior to extraction of DNA. Total DNA was extracted from multiple worms by proteinase K digestion and phenol/chloroform extraction (Sambrook et al. 1989) and concentrated by ethanol precipitation. Initially, 340 bp of the cox1 gene was amplified with primers from Bowles et al. (1992) using polymerase chain reaction (PCR). This sequence was used to design primers specifically oriented to amplify the remainder of the genome in one fragment. This fragment was amplified by long PCR with the GeneAmp PCR kit (Applied Biosystems) using the following protocol: 94°C for 1 min, 15 cycles at 94°C for 15 s and 60°C for 10 min, and 15 cycles at 94°C for 15 s and 60° C for 10 min, with an extension of 15 s for each cycle. Primers used were as follows: forward, 5′-TTG GTG GGT TGA CTG GTG TAA TTA TCT CTA A-3′; and reverse, 5′-GAG CTC ACA CTA AAC TAC CTA AAA CCC CAA T-3′. The PCR product was purified using the Qiax II Gel Extraction Kit (Qiagen). Approximately 1 μg of amplified mtDNA was sheared randomly into smaller fragments via sonication, These fragments were made blunt-ended with an end-repair reaction containing 1 unit Klenow DNA polymerase, 1 unit T4 DNApolymerase, and 2.5 units T4 polynucleotide kinase, in the presence of dNTPs and ATP. The entire reaction mixture was electrophoresed in 1% agarose. Fragments in the range 750 bp to 4 kb were cut out of the gel, and fragments were recovered from the gel block with the MinElute gel purification kit (Qiagen). Purified fragments were ligated into SmaI-cut phosphatased pGEM-3Z vector with the LigaFast ligation system (Promega). Recombinant plasmids were transformed into competent Escherichia coli JM109 cells, and colonies were screened via blue–white color selection. White (positive) colonies were transferred with sterile toothpicks into individual wells of 96-well blocks containing 1.2 ml LB medium. Two blocks of cultures (=192 clones) were grown overnight, and plasmids were isolated with the Montage 96 plasmid purification kit (Millipore). Plasmids were sequenced with M13 primers using the BigDye terminator kit (Applied Biosystems) and a BaseStation automated sequencer (MJ Research). Sequences were assembled with SequencherTM (Gene Codes Corporation). To ensure that proper assembly of the sequences was achieved, the resulting genome sequence was used to design eight sets of primers to be used in PCR, The primer sets covered the entire mitochondrial genome except a 184-bp region encompassing the end of nad5 and the beginning of cob. The primer sets were used for PCR of genomic DNA, and the resulting fragments were visualized on a 0.8% agarose gel and compared to standards. The entire genome sequence has been deposited with the GenBank Data Libraries under accession number AY562383.

Protein coding genes were identified by finding open reading frames and performing BLAST searches on the amino acid sequences (Altschul et al. 1997). To confirm their identity, hydropathic profile comparisons using MacVector 7.1 (Accelrys Inc.) and amino acid sequence comparisons were performed with organisms of the following species: Hymenolepis diminuta, Ascaris suum, Trichinella spiralis, Terebratalia transversa, and Triops cancriformis (Table 1). Following Helfenbein et al. (2001), if the 5′ end of the protein coding genes could not be determined through comparisons with other species, it was inferred to occur at the first possible start codon with no overlap with the proceeding gene with the exception of overlap with the 3’ end of an upstream tRNA. Genes were assumed to terminate early with an abbreviated stop codon if they overlapped with a downstream gene. Ribosomal genes were located by performing BLAST searches and the beginning and ends were inferred to be located at the end of the consecutive genes. The tRNAs were identified using the program tRNAscan-SE (Lowe and Eddy 1997) or through recognition by eye of potential secondary structures surrounding possible anticodon consensus sequences (NTXXXRN; XXX = anticodon). If no suitable tRNAs were discovered, then the restrictions on the consensus sequence were relaxed and the search was repeated. Folding of the tRNAs was accomplished by eye or as suggested by tRNAscan-SE. The complete mitochondrial sequence was searched for tandem repeats using Tandem Repeats Finder (Benson 1999; http://tandem.bu.edu/trf/trf.html).

Table 1 GenBank accession numbers of sequences used in this study
Full size table

Phylogenetic analyses using maximum likelihood (ML) and maximum parsimony (MP) optimality criteria were performed on amino acid sequences for all protein coding genes except nad4L, atp8, and atp6 (due to their lack of conservation). A broad representation of metazoan taxa with complete mitochondrial genomes available was chosen, and sequences were obtained from GenBank (Table 1). Sequences were aligned using ClustalX with the following para meters: gap opening = 2, gap extension = 0.5, and Gonnet Series protein weight matrix. The alignments of the genes were concatenated and areas of ambiguous alignment were removed so that the final alignment consisted of 1523 characters. The aligned amino acid sequences were then converted back into their original nucleotide sequences so that phylogenetic analyses could be performed on the DNA sequences as well as the protein sequences. The third codon position was not utilized in any of the analyses due to the high probability of substitution saturation at these sites.

Maximum parsimony tree searches of the DNA and amino acid sequences were performed heuristically using PAUP* 4.0b10 (Swofford 2001) with tree bisection–reconnection (TBR) branch swapping and random stepwise addition of taxa replicated 1000 times. Analysis of the DNA sequences was unweighted and the amino acid analysis employed the PROTPARS substitution matrix for mitochondrial DNA given in the PAUP* example files. This matrix gives the minimum number of amino acid replacement substitutions needed to convert one amino acid to another, using the mammalian mitochondrial code. Support for nodes was assessed using bootstrap analysis (Felsenstein 1985) using a heuristic search with TBR branch swapping and random stepwise addition of taxa with 100 pseudoreplicates.

Phylogenetic analysis using ML optimality criteria on nucleotide sequences was performed using the GTR+ Γ model, which was determined as the best fit model for the data by the likelihood ratio test implemented in MODELTEST 3.0 (Posada and Crandall 1998). ML analysis was carried out heuristically with PAUP* 4.0b10 (Swafford 2001) with 100 replications of a random stepwise addition of taxa and TBR branch swapping. Bootstrap analysis, executed as described earlier, was used to assess node support. Maximum likelihood tree searches based on the amino acid data were performed using Tree-Puzzle 5.0 (Strimmer and von Haeseler 1996) selecting parameter settings corresponding to the mtREV24 (Adachi and Hasegawa 1996) model of amino acid substitution and using 10,000 puzzling steps. Among-site rate variation was estimated with a gamma distribution, with sites grouped into eight different categories, and the shape parameter α estimated from the data by Tree-Puzzle (α = 0.66).

A topology test was used to compare the topology of the ML DNA tree to both hypotheses A and B from Fig. 1. The topologies of hypotheses A and B were obtained by performing constrained ML tree searches to find the most likely tree given the hypotheses and the data. A Shimodaira-Hasegawa (SH) (1999) test was performed on the three trees as implemented in PAUP* 4.0b10 (Swafford 2001) using reestimated log likelihoods (RELL) approximation and 1000 nonparametric bootstrap replicates.

Results and Discussion

General Features

The mitochondrial genome of L. thecatus is circular and 13,888 bp in length. This is smaller than that of most vertebrate taxa, but it is similar in size to that of many other invertebrates such as nematodes, cestodes, and brachiopods, which typically range from 13,600 to 14,500 bp (e.g., Okimoto et al. 1992; Noguchi et al. 2000; von Nickisch-Rosenegk et al. 2001; Hu et al. 2002). The circular structure and proper assembly of the genome was confirmed by genome-wide PCR amplifications from genomic DNA. All resulting PCR products were of the expected length; therefore, the circular structure was confirmed and proper alignment was expected to be achieved.

The genome contains at least 36 of the 37 genes typically found in animal mitochondrial genomes; however, an atp8 gene could not be positively identified. Two putative atp8 genes were identified, as were two possible trnK and three possible trnS genes. All genes except two alternative trnS genes occur on the same strand, which has a base composition of 44.5% T, 26.8% A, 19.9% G, and 8.4% C. The total base composition of the positive strand was similar to the base composition of the protein coding genes: 46.6% T, 24.7% A, 20.3% G, and 7.9% C. The genome also contains two significant noncoding regions that might be broken up by the putative atp8, trnK, and trnS genes. The inferred organization of the genome is listed in Table 2. Nucleotide positions are numbered starting with the beginning of the cox1 gene.

Table 2 Inferred organization of the mitochondrial genome of Leptorhynchoides thecatus: Shaded boxes indicate putative genes.
Full size table

Protein Coding Genes

BLAST searches identified all protein coding genes except nad6 and atp8. The hydropathic profile of open reading frame (ORF) 2658–3095 was very similar to those of the nad6 gene of two species of nematodes (Fig. 2); therefore, this ORF is likely nad6 of L. thecatus. Two possible open reading frames could code for atp8, ORF 3152–3262 and ORF 4314–4466. Both are similar in length to the atp8 gene of other organisms and ORF 3152–3262 directly precedes atp6, a common gene order in arthropod and dueterostome mitochondrial genomes (Boore 1999). However, the hydropathic profiles of both regions are very different from the atp8 genes of Terebratalia transversa and Mus musculus and the putative atp8 of Trichinella spiralis (Fig. 3); therefore, it is unlikely that either open reading frame codes for the atp8 gene. Definite atp8 genes have not been found in the mitochondrial genomes of either nematodes or platyhelminths (Boore 1999; Le et al. 2000; von Nickisch-Rosenegk et al. 2001).

Figure 2
figure 2

Comparisons of Hopp/Woods hydrophilicity profiles of nad6.

Full size image
Figure 3
figure 3

Hopp/Woods hydrophilicity profiles of two open reading frames (ORF) of Leptorhynchoides thecatus compared to the atp8 profiles of three other species. The identity of the gene for the profile of Trichinetta spiralis is hypothesized to be atp8, but it has not been confirmed.

Full size image

The most common initiation codon is ATG (6 genes), but ATA, GTG, ATT, and TTG are also used (Table 2), Many genes have alternative possible initiation codons, but those that best match the amino acid sequences and the hydropathic profiles of homologous genes are inferred to be correct. Seven genes and the two possible atp8 genes terminate with complete stop codons, TAA or TAG. Truncated stop codons of a single T apparently occur for cox1, nad4, nad5, cob, and nad1. Truncated stop codons are common among bilaterian mitochondrial protein coding genes, and are probably completed through posttranscriptional polyadenylation (Ojala et al. 1981).

Codon usage of all protein coding genes except the putative atp8 genes is depicted in Table 3. The third nucleotide position is heavily biased toward T (51.5%) and A (29.7%) and corresponds to the overall nucleotide bias of the coding strand.

Table 3 Codon usage within 12 mitochondrial protein coding genes of Leptorhynchoides thecatus.
Full size table

rRNA Genes

BLAST searches indicated sequence similarity between small portions of the L. thecatus mitochondrial genome and the rrnL and rrnS genes of other organisms; however, the similarity is low for rrnS. The boundaries of these genes were assumed to abut adjacent genes. The lengths of rrnL and rrnS are short, even compared to other invertebrates such as nematodes and flatworms, which have shorter mitochondrial genes than most bilaterians. The rrnL gene of L. thecatus is 925 bp, compared to 960 bp in A. suum, 958 bp in Necator americanus, 966 bp in H. diminuta, and 1105 bp in T. transverse. The length of rrnS is very short, 513 bp, compared to 701 bp in A. suum, 699 bp in N. americanus, 715 bp in H. diminuta, and 762 bp in T. transversa.

tRNA Genes

The tRNAscan algorithm detected only two tRNA genes (for amino acids R and I) within the complete mitochondrial genome sequence. Additional putative tRNA genes were detected by searching by eye for the anticodon sequences and surrounding sequences for secondary structure. A total of 24 possible tRNA genes were found in the mitochondrial genome of L. thecatus. This total includes the 22 tRNA genes typically found in animal mitochondrial genomes plus additional trnK and trnS genes. Many of the putative tRNA genes exhibit unusual structures and should be considered as hypotheses of the tRNAs present in the genome. Most of the tRNAs are missing the variable loop and TΨC arm, which are replaced by a TV replacement loop (Fig. 4). This structure is similar to that described for the tRNA genes of many nematode species (Wolstenholme et al. 1987). The tRNAs range from between 52 to 104 nt and typically contain an aminoacyl stem of 6 or 7 ntp, a DHU stem of 3 or 4 ntp, and an anticodon stem of 4 or 5 ntp. Exceptions include trnS2 (3163-3226), which has a DHU stem of only 1 ntp and trnS2 (10,481–10,542), which has an anticodon stem of only 3 ntp, The DHU loop ranged from 1 to 11 nt and the anticodon loop of all tRNAs was 7 nt.

Figure 4
figure 4

Inferred secondary structure of tRNAs from the mitochondrial genome of Leptorhynchoides thecatus. Numbers indicate the position of each putative tRNA within the mitochondrial genome.

Full size image

Noncoding Regions

There are potentially two significant noncoding regions in the genome that are 377 and 294 bp in length; however, these regions might be broken up by the putative atp8, trnK, and trnS genes (Table 2). These regions are the most likely candidates for a control region, but they do not contain any of the features found in the control regions of other organisms such as tandem repeats, a high A + T content, a stem–loop structure, or a T-stretch (Wolstenholme 1992; Zhang and Hewitt 1997). The A + T content of the first and second regions are 28.5% T, 32.2% A and 41.7% T, 32.1% A, respectively. These values are lower than those for the protein coding genes and therefore suggest that the non-coding regions are not A-T rich compared to the rest of the genome. A third significant noncoding region of 104 bp might also be present if the putative atp8 and trnS genes do not exist (Table 2). This region is 51% T and 30.7% A but does not contain any of the other aforementioned elements. None of these three regions is a good candidate for a control region, leaving the location of the signals that control replication and transcription unknown.

Phylogenetic Analysis

Parsimony searches of the DNA and amino acid data each resulted in one most parsimonious tree 10,782 and 8,242 steps long, respectively. Tree statistics for the nucleotide derived tree are: consistency index (CI) = 0.406, retention index (RI) = 0.483, rescaled consistency index (RC) = 0.196, and homoplasy index (HI) = 0.594, Statistics for the amino acid derived tree are CI = 0.589, RI = 0.532, RC = 0.313, HI = 0.411. The topologies of the parsimony and maximum likelihood trees are very similar in that the deuterostomes, arthropods, annelids, nematodes, and platyhelminths were each recovered as monophyletic taxa with high bootstrap support (Fig. 5). Additionally, the platyhelminths, nematodes, and acanthocephalans were recovered as a monophyletic taxon in all of the analyses and the bootstrap support of this clade was high (89–99%). Leptorhynchoides thecatus is the sister taxon to Platyhelminthes in trees from all analyses except the ML tree derived from DNA, which placed L. thecatus as a sister to Nematoda, but with low (58%) bootstrap support. Additionally, an SH test comparing the alternative positions of L. thecatus in this topology (as a sister to nematodes or platyhelminths) indicated that there was no difference in the two alternatives (p = 0.173). The placement of L. thecatus and Platyhelminthes as sister taxa in this analysis supports the hypotheses of Winnepenninckx et al. (1995), Giribet (2000), and Giribet et al. (2002) (Hypothesis A, Fig. 1), who placed these taxa together in Platyzoa. The relationship does not, however, conflict with the hypothesis of Nielsen (2001) (Hypothesis B, Fig. 1), who left the relationships undetermined as a polytomy. However, the sister relationship of Nematodes to the Acanthocephala and Platyhelminthes as depicted by these analyses conflicts with all of the aforementioned hypotheses. If the nodes with less than 50% bootstrap support are collapsed, the relationships of the taxa are mainly unresolved except for the aforementioned clades.

Figure 5
figure 5

Maximum likelihood phylogram from analysis of 1523 amino acids from the mitochondrial genome of 26 invertebrate taxa. Numbers below branches indicate quartet puzzling reliability values from the maximum likelihood analysis. Numbers above branches indicate bootstrap values from the parsimony analysis of nucleotide sequences. Only values for the shared nodes are shown.

Full size image

The topology of our ML DNA tree is significantly better than either hypothesis A or B (p < 0.019 and p < 0.001, respectively). The Shimodaira and Hasegawa (1999) likelihoods of the trees are 45,767.28 (ML DNA), 45,822,52 (A), and 45,911.57 (B). The main difference between all of the topologies derived from our study and either of the hypotheses presented in Fig. 1 is the placement of the nematodes and their relationships with the arthropods, platyhelminths, and acanthocephalans. Hypothesis A groups the nematodes and arthropods in Ecdysozoa, a clade of molting organisms, which are distantly related to the platyhelminths and the acanthocephalans. Nematodes and arthropods are the only representatives of Ecdysozoa in our analyses, but they show very distant relationships in all topologies. Ecdysozoa has been supported in some analyses of molecular data (Aguinaldo et al. 1997; de Rosa et al.1999; Giribet et al. 2000; Manuel et al. 2000; Giribet 2002; Copely et al. 2004) but not in others (Mushegian et al. 1998; Wang et al. 1999; Blair et al. 2002). In Hypothesis B, Ecdysozoa are not recognized, but the nematodes are distantly related to the platyhelminths and acanthocephalans. Additionally, this hypothesis groups the brachiopods with the deuterostomes, which is a relationship not supported by our analyses. Another difference between our analyses and both a priori hypotheses is the status of the mollusks. In our analyses, the mollusks were not monophyletic as they are in both a priori hypotheses. Nonmonophyly of mollusks has also occurred in other analyses using molecular data (Winnepennenckx et al. 1995; Carranza et al. 1997).

Taxon sampling could pose a problem in this analysis because mitochondrial genomes of many taxa have not been sequenced. Missing taxa cause the divergences among the taxa in phylogenetic analyses to be large and can lead to error (Hillis et al. 2003). Additionally, base composition bias could also influence the results of the phylogenetic analyses because most of the nematode, platyhelminth, and acanthocephalan genomes contained high percentages of T’s and low percentages of C’s, which could influence analyses using DNA sequences and possibly amino acid sequences (Foster et al. 1997). The results from Tree-Puzzle indicated that amino acid composition of 14 of 25 taxa differed significantly from the frequency distribution assumed in the maximum likelihood model, which could also bias the analysis. However, these 14 taxa included not only those from the nematode, platyhelminth, acanthocephalan clade, but also those from other parts of the tree. Long branch attraction could also cause the acantho cephalan, nematode, and platyhelminth clade to group together since all of these taxa have relatively long branches. These organisms may be evolving rapidly or the long branches could be an artifact of missing taxa within the analysis.

Conclusions

The mitochondrial genome of L. thecatus is unusual in the small size of the rRNA genes, the structure of the tRNAs, and the absence of a control region with signaling elements similar to other organisms. More acanthocephalan species need to be sequenced to determine whether these are features typical of the group. In L. thecatus, mitochondrial genes and their amino acid sequences generally showed low similarity to their homologues in other organisms during BLAST searches and when aligned with sequences from representative nematodes, platyhelminths, brachiopods, and arthropods. The estimated structure of the majority of the tRNAs is similar to that reported from nematodes in that a TV-replacement loop takes the place of the variable loop and TΨC arm.

Phylogenetic analyses of Metazoa using amino acid and nucleotide sequences of conserved regions of protein coding genes support the monophyly of most recognized phyla except Mollusca. However, in the analysis, Acanthocephala, Nematoda, and Platyhelminthes formed a monophyletic group, a hypothesis that has not previously been proposed. Furthermore, the data did not support the Ecdysozoa, a clade of molting organisms supported by previous studies. These results could be artifacts of missing taxa, codon bias, or long branch attraction. The utility of mitochondrial genome sequences for phylogenetic analysis will improve as sequence data become available from more “minor” phyla such as the Acanthocephala.