Introduction

Over the last decade mitochondrial (mt) genomes have become a major data sources for comparative genomics (Boore 1999). These small genomes, typically less than 16,000 bp, in combination with a generally conserved gene complement and rapid rate of nucleotide substitution, provide an ideal system for a wide range of comparative studies that have furthered our understanding of genome evolution. One question has received more attention than any other: the frequency and distribution of mt genome rearrangements. The discovery of novel gene arrangements has been reported with considerable interest (e.g. Shao et al. 2001; Covacin et al. 2006). The relative frequency of rearrangements is known to vary between lineages, and considerable effort has been devoted to documenting which lineages have the highest rates of genome rearrangement. For example, most vertebrate species share a common mt gene order except for groups such as amphibians (Mueller and Boore 2005; Fonseca et al. 2006) and birds (Mindell et al. 1998). However rearrangements within these groups still appear to be uncommon. In contrast, rearrangements appear much more common in invertebrates and variability in gene order can be found within groups at many taxonomic levels (Dowton et al. 2002; Hassanin et al. 2005).

Beyond the absolute number of genome rearrangements within a taxonomic lineage is the question of whether certain types of gene rearrangement are more common than others. Genome rearrangements can be characterized in several aspects: (1) the types of genes rearranged, tRNAs only (termed minor rearrangements) or also protein-coding and rRNA genes (major rearrangements); (2) whether genes are translocated along the same strand or inverted between coding strands; and (3) the localization of the rearrangement (between local or distant gene blocks) (Dowton et al. 2002). A given gene rearrangement may exhibit one or more of these aspects. In addition, the likelihood of each type of rearrangement would vary according to different mechanistic models of how genome arrangements occur. Accurate knowledge of the frequency of the different types of rearrangements is thus useful in assessing the validity of these models. For example, the incidence of genome inversions has been used as evidence for the occurrence of intra-mitochondrial recombination, because inversions could not be explained by the alternative replication-slippage-based models (Dowton and Campbell 2001). An estimate of the frequency of gene inversions could thus give an indication of the frequency of recombination.

Secondly, there is the influence that a rearrangement itself may play on the evolution and function of the mt genes. The substitution rates of mitochondrial genes have been correlated with their physical position within the genome (Saccone et al. 1999; Faith and Pollock 2003). There is a mutational bias in mt genomes favoring As and Cs on the leading replicative strand. This bias is due to the deamination of As and Cs on the lagging-strand during replication leading to the accumulation of Gs and Ts on the lagging strand and their complementary pairs As and Cs on the leading strand (Reyes et al. 1998). The strength of the mutational bias is related to the length of time each gene spends in a single-stranded state during the replication or transcription cycles (Tanaka and Ozawa 1994; Francino and Ochman 1997; Hassanin et al. 2005). Rearrangements could theoretically alter these patterns by shifting genes to sites with a lower mutational background, or shifting between strands and thus reversing the mutational bias. Another possibility is that the origin of replication itself could be shifted or inverted, thus altering the mutational dynamics of the entire genome. It is unknown over what time scales these mutational effects would persist. Therefore, studies of rearrangements could potentially be informative as to the relative strengths of the various mutational effects operating within the mt genome: gene position versus strand biases, synonymous versus nonsynonymous etc. Rearrangements could also lead genes to violate the models used in phylogenetic or molecular dating approaches; however, such effects have never been examined in groups in which rearrangements are known to occur (Rubinoff et al. 2006).

Functionally, rearrangements would likely have an impact on mt gene transcription and translation. Mitochondrial genomes are transcribed as polycistronic pre-mRNAs which are processed by endonucleases which recognize the specific secondary structures of tRNA genes (Ojala et al. 1980, 1981), or possibly stem-loop structures in the 3′ end of those protein-coding genes which are not flanked by tRNAs (Kim et al. 2006; Fenn et al. 2007). Such mechanisms suggest that to ensure proper translation protein-coding genes would be limited to rearrangements which place them adjacent to tRNAs, otherwise post-transcriptional processing would fail. Genome rearrangements could also potentially be constrained by the signaling regions used to identify initiation and termination of transcription. The traditional model of mt genome transcription has emphasized the transcription of two full genome-sized polycistronic pre-mRNAs (Taanman 1999). However, recently a model was proposed where genes encoded on the same strand, and located between initiation and termination sites, are transcribed as a block considerably smaller than full genome size (Roberti et al. 2003, 2006). Rearrangements that moved genes from within their transcription block would therefore result in a loss or serious reduction in the rate of transcription. Examination of rearranged genomes may yield evidence toward which genome features are important as signaling sites or reveal novel methods of genome function that predisposed them to rearrange by releasing the genome from such constraints.

Comparative evolutionary studies of mt genome rearrangements are therefore necessary to understand their role in genome evolution and to accurately identify the mechanisms responsible for rearrangements. In choosing study groups for additional attention it is necessary to examine examples in which significant genome rearrangements have occurred within the group rather than in the group’s common ancestor (genome variability as opposed to genome synapomorphies). Ideally a wide variety of the different rearrangement types (minor and major, translocations and inversions, local and distant) would also occur within the study group to allow the simultaneous examination of multiple classes of rearrangement types. Under these criteria, lice (Insecta: Phthiraptera) appear to be an ideal group for comparative mt genomics. Rearrangements have been identified in only 11 of 29 insect orders examined (Cameron et al. 2006). However, major rearrangements or those involving inversions and rearrangements between distant gene blocks are limited to the four hemipteroid orders (Hemiptera, Thysanoptera, Psocoptera and Phthiraptera) and appear to be most frequent in lice. Each of the two lice that have been sequenced exhibit unique gene orders, and share only a single gene boundary in common with each other and only three (Campanulotes) or two (Heterodoxus) gene boundaries in common with the ancestral insect arrangement (Shao et al. 2001; Covacin et al. 2006). Comparisons between the two louse mt genomes are, however, complicated by the fact that each represents a different louse suborder, Amblycera for Heterodoxus and Ischnocera for Campanulotes. Recent phylogenetic studies also suggest that lice are not monophyletic and that parasitism may have evolved independently in these two suborders (Johnson et al. 2004). In so far as parasitism has been proposed as a potentially significant factor in mt genome rearrangement (Castro et al. 2002), comparisons of additional lice within each suborder are necessary to understand the evolution of mt genome rearrangement in the broadest sense. In this study we sequenced the mt genome of a second species of Ischnocera, Bothriometopus macrocnemis, to document the prevalence of additional rearrangements within this suborder and to examine the influence of rearrangements on mt genome evolution.

Materials and Methods

Specimen Collection

Specimens of Bothriometopus macrocnemis were collected from the Crested Screamer, Chauna torquata (Aves: Ahminidae) in Cordoba, Argentina on 25 April 2001, preserved in 100% ethanol, and stored at −80°C until used for DNA extraction. A voucher consisting of a slide mount of the exoskeleton cleared following DNA extraction was deposited in the Brigham Young University Life Science Museum collection, accession number IGC-PH65.

PCR Amplification and Sequencing

Whole genomic DNA was extracted from an individual louse specimen with the DNeasy Tissue kit (QIAGEN, Valencia, California, USA). Initial polymerase chain reaction (PCR) amplifications included a partial section of the cox1 gene (approx 400 bp) using the primers L6625 and H7005 (Hafner et al. 1994) and the region rns to rnl (approx. 3,000 bp) using the degenerate primers GON1 and GON2 designed previously from the rns and rnl genes of lice (Covacin et al. 2006). These sequences were used to design specific primers to amplify the remaining mt genome by long PCR: rnl to cox1 (approx. 8,000 bp) and cox1 to rns (approx. 6,000 bp). Primer sequence and location for each long PCR is listed in Table 1. Within each long PCR product the full, double-stranded sequence was determined by primer walking (primers available from SLC upon request). Long PCRs were performed using Elongase (Invitrogen, Carlsbad, California) with the following cycling conditions: 92°C for 2 min; 40 cycles of 92°C for 30 sec, 50°C for 30 sec, 68°C for 12 min; and a final extension step of 68°C for 20 min. Sequencing was performed using ABI BigDye ver. 3 dye terminator sequencing technology and run on ABI 3770 or ABI 3740 capillary sequencers. Cycle sequencing conditions were 28 cycles of 94°C for 10 sec, 50°C for 5 sec, 60°C for 4 min.

Table 1 Primers, sequence and location for long polymerase chain reactions
Full size table

Analysis and Annotation

Raw sequence files were proof read and assembled into contigs in Sequencher version 4.6 and 4.7 (GeneCodes Corporation). Transfer RNA analysis was conducted using tRNAscan-SE (Lowe and Eddy 1997) using invertebrate mitochondrial codon predictors and a cove score cut-off of 1. Reading frames between tRNAs were found in Sequencher and identified using translated basic logical alignment search tool (BLAST) searches (blastx) (Altschul et al. 1997) as implemented at the National Centre for Biotechnology Information (NCBI) website (http://www.ncbi.nlm.nih.gov/). Codon usage was calculated using MEGA ver. 3 (Kumar et al. 2004). Homology of tRNAs was determined in the first instance by anticodon sequence and checked by phylogenetic analysis with the tRNAs from Triatoma (Hemiptera), lepidopsocid RS-2001 (Psocoptera), and Campanulotes (Phthiraptera). All tRNA isotypes for the four species were aligned using their secondary structure (as predicted in tRNAScan-SE) with length-variable loop regions and the anticodon loop removed and phylogenies inferred on the most conserved portions of the genes. The resulting dataset was analyzed using Neighbour Joining in PAUP ver. 4b10 (Swofford 2002) and MrBayes version 3.1 (Ronquist and Huelsenbeck 2003). Alignments of the protein-coding genes were made in MEGA ver. 3 for the three louse species with Triatoma and lepidopsocid RS-2001 as outgroups. Hydrophobicity profiles were calculated in Genious version 2.5.4 (Drummond et al. 2006). Secondary structures between adjacent protein-coding genes and within the repetitive nontranslated regions were predicted using Mfold version 3.2 (Zuker 2003).

Results and Discussion

Genome Organization and Structure

The mt genome of Bothriometopus macrocnemis is circular and 15,564 bp in size, making it the largest louse mt genome yet sequenced (GenBank accession number EU183542). The genome encodes 40 genes, 36 of the 37 genes common to most metazoan mt genomes plus three additional copies of the tRNA-Trp, and an additional copy of the tRNA-Val. The gene for the tRNA-Ala is lacking. Additionally, there are two putative nontranslated regions each consisting of a 5′ nonrepetitive section plus two complete and one partial repeat units. Each region includes an open reading frame that BLASTs with low significance to the cob gene of the pseudococcid bug Ferrisia (Baumann and Baumann 2005). There is, however, no significant sequence similarity at either the DNA or amino acid level between these ORFs and the cob gene from Bothriometopus, suggesting that these regions are not pseudogenes and are unrelated to the full-length cob gene. They are the first record of nontranslated, repetitive sequence in a louse mt genome. There are also 25 intergene spacer regions ranging in size from 1 to 83 bp, 14 of which are over 20 bp in size. This is considerably more spacer regions than are found in most insect mitochondrial genomes, particularly the other louse genomes, which are especially compact. Collectively the spacer and nontranslated regions are responsible for increased size of the Bothriometopus mt genome, which is approximately 750 bp larger than the next largest louse genome.

All 40 genes of the Bothriometopus mt genome are transcribed on the same strand, a condition not typical in arthropods. Similarly, in the ischnoceran louse Campanulotes, 36 of 37 genes are transcribed on the same strand (Covacin et al. 2006). The only other arthropod known to have all genes on the same strand is the copepod Tigriopus (Machida et al. 2002; Burton et al., unpublished data). More broadly, a mt genome with all genes on the same strand is found in many metazoan phyla including Acanthocephala (Steinauer et al. 2005), Annelida (Boore and Brown 2000), Cnidaria (Medina et al. 2006; Brugler and France 2007), Mollusca (Hoffmann et al. 1992), Nematoda (He et al. 2005; but this is not the ancestral condition for the phylum, see Lavrov and Brown 2001), Platyhelminthes (Littlewood et al. 2006), Terebrantia (Helfenbein et al. 2001) and Urochordata (Yokobori et al. 1999, 2005). Indeed, genes transcribed on both strands only appears to be the ancestral state within Ecdysozoa (Webster et al. 2006), Vertebrata (Boore 1999) and several basal metazoan groups such as Porifera (Lavrov and Lang 2005a) and Placozoa (Signorovitch et al. 2007). It is unclear what, if any, significance the secondary adoption of single-stranded transcription in lice provides. Use of a single polycistronic transcript for mRNA, tRNA, and rRNA production could conceivably be more energetically efficient than using two or more, but this does not explain why such large groups as arthropods and vertebrates almost exclusively use at least two transcripts for genes distributed across both strands.

The mt genome arrangement of Bothriometopus differs radically from those of both other louse species and from the inferred ancestral insect arrangement (Fig. 1; Table 2). Only four gene boundaries and three gene blocks are shared between Bothriometopus and the ancestral insect: G-nad3, atp8-atp6, and nad4-H-nad5. Two of these boundaries are shared with the pigeon louse Campanulotes: G-nad3 and atp8-atp6. However, no derived gene arrangements are shared between these two ischnoceran species. The gene boundary atp8-atp6 is the only one shared with wallaby louse Heterodoxus, an amblyceran. A member of the closest relatives of parasitic lice, Psocoptera (Lepidopsocidae) shares none of its derived gene arrangements with Bothriometopus or with any of the other louse mt genomes. Even if the arrangement of the tRNA genes is ignored and only the relative positions of the protein-coding genes and rRNAs are considered, no additional arrangements are shared between Bothriometopus and Campanulotes. In addition, only a single extra arrangement, nad3-rnl, is shared by Bothriometopus and Heterodoxus, although this gene block is inverted between the two genomes. Such a lack of conserved gene blocks precludes the accurate reconstruction of the rearrangement events that gave rise to the extant mt genomes of lice. Additional louse mt genomes are needed to split up the long evolutionary distances between the current exemplar species and allow more accurate interpretation of their genomic history.

Fig. 1
figure 1

Maps of the mitochondrial genomes of Bothriometopus, the other louse species Heterodoxus and Campanulotes, a psocopteran, and the ancestral insect. Each genome has been linearized at an arbitrary point, the 5′ end of cox1. Gene names are the standard abbreviations used in this paper; tRNA genes are indicated by the single-letter IUPAC-IUB abbreviation for their corresponding amino acid, duplicated tRNA genes are numbered as in the text, pseudo-tRNA genes are hatched and indicated by a lower-case p; intergenic spacers are cross hatched and numbered as in the text, putative control regions are in black; NTRR: nontranslated repeat region. Genome orientation for each species is 5′ to 3′ on the majority strand and genes coded on the minority strand are underlined. Shared gene boundaries are indicated by brackets and letter coded: A: atp8/atp6; B: tRNA-G/nad3; C: nad4-tRNA-H-nad5; D: nad4L-nad4

Full size image
Table 2 Summary of the mitogenome of Bothriometopus
Full size table

Nucleotide Composition

The nucleotide composition of the coding strand of the Bothriometopus mt genome is: A: 32.1% (5,001), C 13.7% (2,135), G 15.5% (2,410) and T 38.7% (5,018). Combined GC content was thus 29.2%, in the middle of the range for insect mt genomes (Fenn et al. 2007) and comparable to the previously sequenced louse mt genomes: Campanulotes 29.9% and Heterodoxus 20.7%. There is almost no nucleotide skew (sensu Perna and Kocher 1995); A-skew is –0.002 and C-skew is –0.06. This is comparable to Heterodoxus, A-skew –0.02 and C-skew 0.01, but in marked contrast to the other ischnoceran Campanulotes, A-skew –0.25 and C-skew –0.38. Nucleotide frequency and skew statistics calculated for the entire genome do not take into account the distinction between synonymous and nonsynonymous mutations. Mutations at the third codon position of fourfold degenerate sites, those coding for the amino acids A, G, L(CTN), P, R, S(TCN), S(AGN), T and V, are selectively neutral and nucleotide composition at these sites is more indicative of background mutational pressures on nucleotide bias and skew (Kimura 1983). Table 3 shows nucleotide composition, A-skew, and C-skew for Bothriometopus and Campanulotes in which all 13 protein-coding genes are transcribed from a single strand, plus results for each strand for Heterodoxus, a lepidopsocid (Order Psocoptera), and Triatoma (Order Hemiptera) in which genes are transcribed on both strands. These results confirm that the mt genome of Bothriometopus, while strongly AT-biased, is essentially unskewed; there are approximately equal numbers of each of the complementary nucleotides, A:T, G:C. This finding is unusual because mt genomes typically demonstrate pronounced strand asymmetry in the nucleotide skew (Tanaka and Ozawa 1994; Reyes et al. 1998; Hassanin et al. 2005; Cameron and Whiting 2007). One strand is typically A and C rich and the other T and G rich, as is clearly demonstrated by the mt genome of Triatoma, and to a lesser extent by the lepidopsocid, in which the majority strand shows pronounced A and C skew and the minority strand significant T and G skew. In contrast, each of the three lice differs from this pattern. Campanulotes has pronounced reverse skew: strong T and G skew on the majority strand. Heterodoxus appears to be largely unskewed, although interestingly both the majority and minority strands have comparable A and G skews. This is unexpected as skews are usually complementary, i.e., an A skew on one strand is balanced by T skew on the other, as nucleotide biases have typically been attributed to strand-specific effects related to replication or transcription (Francino and Ochman 1997). Two models of mt genome replication have been proposed to explain the strand specificity of nucleotide biases. Under the strand-displacement model (Clayton 1982), replication of one strand lags the other. The lagging strand exists in a single-stranded state for much of the replication cycle and is exposed to potential mutations for a longer period than the leading strand, and so is expected to accumulate A and C nucleotides (Sancar and Sancar 1988; Reyes et al. 1998). In the strand-coupled model (Yang et al. 2002), the two strands remain double stranded at all times, and strand-specific nucleotide biases are attributed to the incorporation of RNA intermediates during the replication cycle. However, this model has only been demonstrated in mammals (Bogenhagen and Clayton 2003). These mutational effects are therefore likely to affect nucleotide skew on a strand-wide basis, so the absence of an asymmetric strand effect in Heterodoxus and Bothriometopus is surprising. Hassanin et al. (2005) attributed this lack of asymmetry in Heterodoxus to a recent reversal in the direction of transcription. Following a reversal, strand asymmetry should be eroded and eventually reversed entirely, as was demonstrated for several other invertebrate genomes with reversed skew (e.g., the mollusc Katharina). An absence of asymmetry was proposed to be a transient effect and in mt genomes lacking skew, insufficient time had passed to allow opposite asymmetry to become established. However, this does not explain the absence of skew in Bothriometopus, while a member of the same suborder, Campanulotes, shows such skew.

Table 3 Strand asymmetry in nucleotide composition at fourfold degenerate sites
Full size table

Other possibilities to explain the absence of skew in the mt genome are easily conceived. Genome rearrangements that include gene inversions are expected to homogenize any general pattern of asymmetry. Flipping a gene’s transcription from one strand to another would result in our calculations in Table 3 being based on inappropriate groupings, i.e., some of the genes on the majority strand may retain nucleotide skew accumulated when the gene was located on the minority strand. This would suggest that the organization of the Bothriometopus genome, in which all the genes are transcribed on the same strand, is of evolutionarily recent origin. As inversions appear to be common in louse mt genome this is a potentially confounding factor in examinations of nucleotide skew. Calculating skew for each gene for the various louse species (Supplementary Table S1), the genes do not separate into two groups, one A and C skewed and the other T and G skewed. This pattern, if it occurred, could have been interpreted as demonstrating a division between genes historically located on one strand or the other, and the proper strand assignment now obscured by gene inversions and rearrangements (see Fonseca et al. 2006 for an example of this approach in amphibians). Of the Bothriometopus genes, only cob demonstrates both A and C skew, although it is extremely weak; cox1 and nad1 demonstrate both T and G skew but again this is not strong. In Heterodoxus, nad1 and nad6 demonstrate A and C skew, and cox1 and cob T and G skew, but the G-skew of cob is the only strong skew in this genome. In contrast, in Campanulotes, all genes except cox3 have pronounced T and G skew, and the A skew in cox3 is marginal and its G skew is very strong. These results suggest that the genome-wide calculations (Table 3) are closely matched by calculations based on single genes (Table S1) and the absence of asymmetry in Bothriometopus and Heterodoxus is not the result of recent gene inversions that have homogenized any nucleotide biases between the strands.

A further factor that has been proposed to account for nucleotide skew is the process of transcription, which also results in the nontranscribed strand spending a significant period in a single-stranded state, with the same expected pattern of A and C biased mutation on the nontranscribed strand (Hassanin et al. 2005). Transcription related effects on one strand should nevertheless be counter-balanced by transcription on the other because the entire mt genome is transcribed in a single polycistronic mRNA (Ojala et al. 1980, 1981). Full transcription of each strand would expose each complementary strand in turn to a similar period in the single-stranded state. An exception would be genomes in which all genes are coded on one strand and have ceased transcription of the other strand as a result, such as Bothriometopus. It should be noted that several of the genomes which Hassanin et al. (2005) found to have T and G skew such as Tigriopus and Katharina and so concluded to have undergone reversals of the direction of replication also have all genes transcribed on a single strand and so the reversed asymmetry could alternatively be attributed to a transcription effect. Transcription in such a genome could either reinforce the strand asymmetry produced by replication, if the transcribed strand is also the leading strand in replication, or counterbalance it, if the transcribed strand is the lagging strand. Thus, the absence of strand asymmetry in Bothriometopus could be explained if replication of this mt genome proceeds with the noncoding strand leading. Transcription may also explain the strong T and G skews in the coding strand of Campanulotes because transcription-induced asymmetry would produce A and C skew on the nontranscribed strand. However, 36 of the 37 mt genes are transcribed on the same strand in Campanulotes. It would seem grossly inefficient to transcribe the entire minority strand of the Campanulotes mt genome simply for a single tRNA gene, but the signaling sites used to initiate transcription are unknown for lice and have not been studied in insects for any groups other than Drosophila (e.g. Matsushima et al. 2004, 2005; Roberti et al. 2003, 2006).

Under the consensus model of transcription, the initiation of transcription (IT) sites for each strand are located in the control region plus an additional IT located immediately upstream of the rns gene which in concert with a termination site (mTERM) located immediately downstream of the nd1 gene in Drosophila or rnl in vertebrates (Taanman 1999) allows transcription of the rRNA genes independently of, and at a heightened rate relative to, the rest of the genome. Three transcripts are thus produced, a complete transcript of the entire mt genome for each strand plus one for the ribosomal gene block. Roberti et al. (2003, 2006) have recently proposed an alternative model of transcription in Drosophila in which each of the four major blocks of genes coded on the same strand have unique IT sites upstream of their coding region. They suggest that mTERM sites downstream of each coding block act as transcription attenuators, terminating some but not all transcription through these sites, producing a mix of partial and complete genome transcripts. This model accounts for the excess of sense over anti-sense mRNAs (Berthier et al. 1986) within mitochondria and suggests that transcription of just a targeted coding region is possible. A similar mechanism in Campanulotes could supply the single tRNA-Gln, which is coded on the minority strand, without the seemingly wasteful production of a transcript for the entire mt genome. The tRNA-Gln in Campanulotes is, however, not located adjacent to either of the two noncoding regions identified as potential control regions due to the presence of stem-loop structures (Covacin et al. 2006). This finding indicates that either the partial transcript is much larger than just the tRNA gene, the stem-loop structures are 3,035 and 8,325 bp from tRNA-Gln, respectively, or that other stem-loop structures within adjacent tRNAs or the rns gene function as IT sites in this genome.

Nucleotide skew and strand asymmetry in louse mt genomes thus probably reflect the wide range of genomic effects that can alter how the underlying patterns of mutation are reflected in the nucleotide composition of the genome. Reversals of the directions of replication, gene inversion, gene rearrangements and the pattern of transcription likely all contribute to complicated patterns of nucleotide skew. The sequencing of additional louse genomes will allow comparisons between much more closely related taxa and allow these factors to be teased apart. In conjunction with molecular dating analyses it could be possible to begin to determine the rates of substitution induced by each of these strand-specific mutational biases and thus infer which are most responsible for the overall shaping of mt genome nucleotide composition.

Transfer RNA Genes

Twenty-one of the 22 tRNA genes found in most metazoan mt genomes were found in Bothriometopus (Fig. 2). The tRNA-Ala was not found by tRNA-ScanSE and attempts to fold by hand a tRNA that coded for alanine from the nontranslated portions of the genome failed. In addition two tRNA-like pseudogenes were found between tRNA-Asn and tRNA-Thr and between tRNA-Ser(UCN) and cox1. In both regions a cruciform secondary structure resembling a tRNA could be found; however, neither could form a functional anticodon loop. The tRNA-Pseudo1 had an additional base in the anticodon loop, seven additional base pairs between the DHU and anticodon loops and two additional bases in the extra arm. The tRNA-Pseudo2 was missing one base from the anticodon loop and had a shortened DHU arm, two bases rather than the normal three or four. Due to the altered anticodon loops it is impossible to assign a tRNA isotype to these genes. Additionally, the stems of each did not show significant sequence similarity to any of the other tRNA genes, so it is impossible to determine homologies for these pseudogenes.

Fig. 2
figure 2

tRNA secondary structures from Bothriometopus. Structural features are listed on tRNA-Arg at the top left. tRNAs are labelled with the abbreviations of their corresponding amino acids. Sequences are given 5′ to 3′ in the direction of transcription; anticodon sequences are in bold; hydrogen bonds indicated by a thick dash. Duplicated and pseudo-tRNAs are labelled as in the order they occur in the genome as for Fig. 1

Full size image

Nineteen of the 21 tRNA genes used the same anticodon sequences as are found in the majority of insect species. The tRNA-Lys and tRNA-Ser(AGN) had the anticodons UUU and UCU rather than the usual CUU and GCU respectively. In Campanulotes these same tRNAs have the same divergent anticodon sequences as Bothriometopus and the remaining 20 tRNAs possess the same anticodon sequences as most insects. The same divergent anticodons are also found for tRNA-Lys in the Heterodoxus and Pediculus mt genomes; the remaining tRNAs have the insect ancestral sequence (Shao et al. 2001; Pittendrigh et al., unpublished data). These changes are consistent with the phylogeny of lice: tRNA-Lys (CUU → UUU) occurred in the common ancestor of Amblycera and Ischnocera and tRNA-Ser(AGN) (GCU → UCU) in the common ancestor of Ischnocera (Johnson et al. 2004). Such changes may constitute “rare genomic changes” which are diagnostic for the respective groups (Rokas and Holland 2000).

While there are four copies of tRNA-Trp in the Bothriometopus mt genome, there are only two types. Copies 1, 3, and 4 (relative gene order as per Fig. 1) share almost identical sequences differing only at the 3′ discriminator nucleotide (an A in copies 1 and 3 vs a G in copy 4). Copy 2 differs significantly from the other copies. Both copies of tRNA-Val have identical sequences. A phylogeny of the tRNA genes was constructed to test to homology of the two tRNA-Trp types. This approach has previously been used to identify the homology of duplicated tRNAs in wasp (Dowton and Austin 1999) and sponge mt genomes (Lavrov and Lang 2005b). These trees indicated homology between the Bothriometopus tRNA-Trp2 and the tRNA-Trp genes from Triatoma and the lepidopsocid and between Bothriometopus tRNA-Trp1,3,4 and Bothriometopus tRNA-Glu. This suggests that the duplicated copies of tRNA-Trp in the Bothriometopus mt genome were ultimately derived from tRNA-Glu by tRNA remolding (Higgs et al. 2003; Rawlings et al. 2003) or tRNA recruitment (Lavrov and Lang 2005b) as substitutions to the anticodon loop which change the apparent identity of a tRNA gene, have been variously termed. This result is somewhat unexpected as it would require, minimally, the anticodon mutations UUC → UCA. Two anticodon substitutions seems excessive given how conservative the anticodon loop is across insect mt tRNAs (i.e., 19 of 22 genes with the identical sequences across insects and the only recorded differences in the first or wobble nucleotide). This suggests that the Bothriometopus tRNA-Trp1,3,4 genes are probably nonfunctional but this would be difficult to test.

These are the first instances of tRNA duplications in a louse mt genome although this has been found previously in calliphorid blowflies (Lessinger et al. 2004) and in Hymenoptera (Dowton et al. 2003; Castro et al. 2006). In both these examples, however, the tRNAs are tandemly duplicated and the copies are thus physically closely located. In Bothriometopus they are distributed throughout the genome; 4,916 bp separate tRNA-Trp1 and tRNA-Trp3 1,740 bp tRNA-Trp3 and tRNA-Trp4, and 8,731 bp tRNA-Trp4 and tRNA-Trp1. The two copies of tRNA-Val are separated by 2,606 bp. Sequence similarity between the different copies of these tRNAs may be due to their recent duplication or may be maintained over longer time frames by concerted evolution. Under the widely accepted duplication-loss models of genome rearrangement (Moritz and Brown 1987; Macey et al. 1997) it would be unusual to find such large distances between the duplicated genes (cf. the local tRNA duplications in wasps found by Dowton et al. (2003)) as it would be considered unlikely that the loss of the intervening genes would occur faster than point mutations would accumulate within the duplicated tRNAs. The existence of widely separated tRNAs with identical sequences could thus be evidence that the duplication occurred through a recombination mediated process which could move portions of the mt genome to any other point within the genome (Dowton and Campbell 2001). Alternatively sequence similarity between duplicated tRNAs, having arisen through a duplication-loss process, could be maintained through concerted evolution, itself a recombination mediated process. Concerted evolution of nontranslated portions of the mt genome has been frequently recorded (e.g., Shao et al. 2004; Ogoh and Ohmiya 2007) but confined to duplicated control regions. It has been proposed that genomes with two control regions are able to replicate/transcribe faster than those with a single control region. If this is the case, concerted evolution would likely be necessary so that both control regions are recognizable to the replication or transcription machinery (Shao and Barker 2003). Concerted evolution between duplicated tRNAs has not previously been recorded, however tRNAs have been implicated as signaling sites for mt genome replication in vertebrates and such a role could drive their concerted evolution (Taanman 1999).

Protein-Coding Genes

Eleven of the 13 protein-coding genes were identified by comparison with other insect mt genomes on GenBank using Blastx searches. The genes atp6 and nad4l were identified by comparison of amino acid sequence similarity and of hydrophobicity plots between putative open reading frame regions and copies of these genes that have been identified from the mt genomes of related insects. Codon composition of the protein-coding genes was similar to that found in other louse and hemipteroid mt genomes (Table 4). In the Bothriometopus mt genome, 14% of codons were G+C rich and 44.8% A+T rich. A+T rich codons are defined as those with an A or a T at the first and second codon position (Asn, Ile, Leu(UUR), Lys, Met, Phe, Tyr); similarly G+C rich codons are those with a G or a C in the first and second positions(Ala, Arg, Gly, Pro) (Foster et al. 1997). This is marginally lower than that of other lice: Campanulotes 13.6% G+C rich and 46.4% A+T rich; Heterodoxus 11.1% G+C rich and 55.5% A+T rich; and members of closely related orders such as Psocoptera, 13.8% G+C rich and 56% A+T rich, and Triatoma, 15.7% G+C rich and 45% A+T rich. This is consistent with the overall nucleotide composition and bias of the Bothriometopus mt genome, which was less extreme than in other closely related species.

Table 4 Amino acid percentage usage reported for each gene in the Bothriometopus (B) mt genome. Total percentage amino acid usage is compared against Campanulotes (C), Heterodoxus (H), lepidopsocid RS-2001 (L), and Triatoma (T)
Full size table

Recently, Kim et al. (2006) and Fenn et al. (2007) have examined mt protein-coding genes for potential secondary structures that may serve as recognition sites for mRNA cleavage. In the ancestral insect mt genome arrangement, 9 of the 13 protein-coding genes are flanked on their 3′ end by tRNA genes. Ojala et al. (1980, 1981) have proposed that endonucleasis of the polycistronic pre-mRNAs at the tRNA stem liberates mature mRNAs for the protein-coding genes. The remaining four protein-coding genes are flanked at their 3′ end by other protein-coding genes and the reading frames of the two proteins frequently overlaps. Fenn et al. (2007) have demonstrated stem-loop structures at each of these gene boundaries that may act as binding sites for endonucleases. Highly rearranged mt genomes such as those of lice similarly contain gene boundaries between protein-coding genes for which secondary structure-mediated cleavage may play a role in mRNA processing. Several of these gene boundaries are retained from the ancestral insect mt genome arrangement, e.g., atp8-atp6 found in each of the three louse species, but others are novel gene boundaries not found in other groups, e.g., nad5-nad1 in Bothriometopus. Correct processing of these polycistronic pre-mRNAs would be necessary for the correct translation into mature proteins so we examined the potential of each protein-coding gene to form similar stem-loop structures.

Stem-loop structures were inferred at the 3′ end of each of the protein-coding gene to protein-coding gene boundaries in each of the three species of lice (Fig. 3): nad5–nad1 and atp8–atp6 in Bothriometopus; nad6–rns, cox1–cob–cox2, atp8–atp6 and cox3–nad4 in Campanulotes; and nad1–cob, nad4–nad4l and atp6–atp8 in Heterodoxus. There does not appear to be a consistent pattern to these secondary structures. The stem portion varied from as little as 4 (nad1 in Heterodoxus) to as many as 23 (atp8 in Campanulotes) paired bases. The loop portion varied from as few as 3 (atp8 in Bothriometopus) to 64 (nad6 in Campanulotes) nucleotides in size. Finally the position of the stop codon varied from entirely within the stem (nad1 in Heterodoxus) to two bases down stream of the stem (atp8 in Bothriometopus). There was even large variability between structures inferred for atp8, a gene which abuts atp6 in each of the three louse mt genomes, and which one would expect to demonstrate some degree of similarity due to gene homology. The stems inferred for atp8 varied from 9 bases long in Heterodoxus to 23 in Campanulotes and the loops from 3 to 14 bp in size. This variability is however consistent with the low level of amino acid similarity in the 3′ end of the translated atp8 protein across lice. Such variability confounds attempts to discern what consistent features within the secondary structures are recognized by the endonucleases. The endonucleasis of mt pre-tRNAs is mediated by two enzymes: RNase P which cleaves the 5′ end of the molecule and RNase Z which cleaves the 3′ (Dubrovsky et al. 2004). The consistent clover-leaf secondary structure of tRNAs plays a role in this specificity (Frank and Pace 1998). Because the stem-loop region includes a coding portion of the mRNA, it is probable that RNase P does not play a role in the processing of the mt pre-mRNAs. RNase P is very sensitive to even single base substitutions in the stems of substrate tRNAs (Levinger et al. 1995, 1998), let alone the huge variations in stem length which we found inferred here for the mRNAs, which is consistent with RNase P not recognizing these structures. RNase Z is less sensitive to substrate variations (Levinger et al. 1995, 1998). However, it is unclear whether this enzyme or another endonuclease is responsible for RNA processing of the protein-coding genes, as enzymatic activity of this type has not previously been investigated for even model insect species such as Drosophila, let alone lice.

Fig. 3
figure 3

mRNA secondary structure stem-loops found at the 3′ end of protein-coding genes. A: Protein-coding genes abutting protein-coding genes in Bothriometopus. B: Protein-coding genes abutting protein-coding genes in Campanulotes. C: Protein-coding genes abutting protein-coding genes in Heterodoxus. D: Protein-coding genes abutting tRNA genes in Bothriometopus. Sequences are given 5′ to 3′ upper sequence looping to the lower; hydrogen bonds are indicated by a thick dash; the number of bases in the loop region is indicated on the right hand end of each structure; the stop codon is indicated by a dashed box

Full size image

Messenger RNA processing is also a potential limitation on mt genome rearrangements. The absence of an appropriate stem-loop structure in some of the protein-coding genes could result in mt genome arrangements in which those genes that are not flanked by tRNA genes are impossible to process by the post-transcriptional machinery. Such genome arrangements could therefore be lethal. Under such a scenario, protein-coding genes could only rearrange as units with their flanking tRNA genes. This however is not supported by the genome arrangements of the three louse species in which there is little conservation of the gene boundaries between protein-coding genes and tRNAs. Of the nine protein-coding gene to protein-coding gene boundaries described from lice (Fig. 3), only four are found in other insect groups. The remaining genes have either developed novel stem-loop structures as a compensatory mechanism for the absence of a 3′ tRNA gene or such structures are widespread even in protein-coding genes that are flanked by tRNA genes. Analysis of the 13 protein-coding genes within Bothriometopus found stem-loop structures at the 3′ end of all mt genes except cox1 and nad4 which were similar to those found in the protein-coding gene flanked genes. In contrast, analysis of the protein-coding genes of Triatoma, whose mt genome retains the ancestral insect arrangement, revealed stem-loops in the 4 genes flanked at their 3′end by protein-coding genes as expected, but of those which are flanked by tRNAs only cox1 and cob possessed such structures. These comparisons suggest that the development of 3′ end stem-loops in lice may have provided these genes with the potential to rearrange, because genome arrangements that separated them from 3′ flanking tRNA genes would no longer result in a loss of proper mRNA processing in these genomes. These phenomena, however, require much additional investigation. Until the secondary structures targeted by endonucleases and the specificity of those endonucleases are better understood, it will be impossible to discriminate between genuine structures involved in signaling and random structures without any evolutionary or genomic significance.

Putative Control Regions

The control region of the Bothriometopus mt genome is difficult to discern due to the larger than normal number of intergenic spacer regions. Each of the three large intergenic spacer regions, plus each of the nontranslated repeat regions can form the stem-loop structures typically associated with the origins of replication/transcription (Taanman 1999; Saito et al. 2005) (Fig. 4). Stem-loop 1, between tRNA-Glu and tRNA-Val1, is located 10 bp into the spacer region and consists of an 11 bp stem and 8 bp loop. Stem-loop 2, between nad4l and tRNA-Leu(UUR) and located 22 bp into the spacer, has an 11 bp stem and 3 bp loop. Two stem-loop structures were found between tRNA-Trp4 and tRNA-Tyr each consisting of an 8 bp stem and 3 bp loop, the first located 5 bp into the spacer and the second 6 bp from the end of the spacer. There is only limited sequence similarity between any of these stem-loop structures and if they do represent duplicated control regions then they are not undergoing concerted evolution.

Fig. 4
figure 4

Putative control regions within the mitochondrial genome of Bothriometopus. A: Intergenic spacer 1, between tRNA-Glu and tRNA-Val1. B: Intergenic spacer 2, between nad4l and tRNA-Leu(UUR). C: Intergenic spacer 3, between tRNA-Trp3 and tRNA-Tyr. D: 5′ Nonrepetitive section of the nontranslated repetitive regions (NTRR). E: Repeat unit of the nontranslated repeat regions. Unpaired, nonstem regions are indicated by a black line with the size in bp listed. Differences between the partial and complete repeat units, and between the two NTRRs are indicated by box delimiting the regions that are absent in certain copies of the repeats

Full size image

Additionally each of the nontranslated repeat regions are also capable of forming stem-loop structures. Nontranslated repeat region B, between tRNA-Thr and tRNA-Ser(AGN), is the larger of the two and has a more complete secondary structure. The 5′ nonrepetitive region forms a complicated structure (Fig. 4) with a 7 bp stem between the 5′ and 3′ ends (stem 1), and a large loop comprising most of the region from which two additional stem-loops are formed (stems 2 and 3). The largest portion of the loop, the 13 bp between stem 3 and stem 1, is primarily composed of an 11 bp poly-A stretch. The 5′ nonrepetitive region is 27 bp shorter in nontranslated repeat region A than in nontranslated repeat region B and this corresponds to a loss of the 5′ sides of stems 1 and 2 and stem 3 is the only stem-loop which can be formed. The repeat units within each nontranslated repeat region also form stem-loop structures a long 18 bp stem (stem 4) and short 6 bp stem (stem 5). The two complete repeat units each consists of both stems 4 and 5, the partial repeat unit B3 consists only of stem 4 and the partial repeat unit A3 is 14bp shorter than B3 which corresponds to a loss of all but the uppermost 6 bp of stem 4. The second repeat unit of both nontranslated repeat regions, repeats A2 and B2, has a single substitution A → T in the 3 bp loop of stem 4. This suggests that the sequence identity between corresponding repeat units, A1 to B1 etc, is due to recent duplication of the entire nontranslated repeat region rather than duplication of repeat units independently within each location. Further the absence of 27 bp at the 5′ end and 14 bp at the 3′ end in nontranslated repeat region A, which correspond to portions of stem-loops found in nontranslated repeat region B, suggests that the nontranslated repeat region A is an imperfect copy of nontranslated repeat region B, rather than that nontranslated repeat region B was formed by a duplication of nontranslated repeat region A into an existing intergenic spacer region and that the additional bases in nontranslated repeat region B are unrelated to its structure or potential function. The structure of the nontranslated repeat regions is the most similar to that found in the control regions of other insects (Lewis et al. 1994; Zhang and Hewitt 1997) in that it is composed of a stem-loop structure followed by a poly-A or poly-T stretch (the 5′ nonrepetitive region) followed by a series of large repeat units with internal secondary structures (the repetitive region). In contrast, the putative control regions found in other lice consist simply of a stem-loop structure without the poly-A/ T region or repetitive sequence regions.

There are thus five potential control regions in the nontranslated portions of the mt genome, plus the two sets of duplicated tRNAs which appear to be undergoing concerted evolution. Any one of these could potentially be an origin of replication, the initiation or termination sites for transcription or these roles could be independent and occur at different sites in the mt genome, or each of these functions could be simultaneously performed by multiple sites. The mt genome’s origin of replication has only been experimentally mapped for a small number of insects (Goddard and Woolstenholme 1980; Saito et al. 2005), but no louse species or close relative has been examined. While stem-loop structures and associated poly-A/T regions are a common feature of the origin of replication, the size of the structures varies considerably and the poly-A/T region, located variously up or down stream of the stem-loop, forms the loop region itself or is absent altogether in Orthoptera (Saito et al. 2005). Given such variability, inferences based on sequence analysis alone cannot be regarded as completely accurate. Previously all major noncoding portions of the louse mt genomes have been interpreted as potential control regions. Heterodoxus was posited to have two control regions and Campanulotes three (Shao et al. 2001; Covacin et al. 2006). Following these precedents Bothriometopus would be interpreted to have five. However, there does not appear to be much justification for such functional interpretations of the potential to form secondary structures, particularly as secondary structures are almost ubiquitous in louse mt genomes occurring in most protein-coding genes, all tRNA and rRNA genes, as well as most nontranslated regions. Additional studies, both functional (Saito et al. 2005) and comparative (Brehm et al. 2001), will be necessary to properly interpret the putative mt control regions of lice.

Conclusion

Louse mt genomes display amazing variability in genome arrangements, nucleotide composition, biases and skew, tRNA duplication and evolution, the mRNA structure of the protein-coding genes, and putative control regions. The degree to which phenomena such as strand-specific nucleotide skew or tRNA anticodon sequences, which over the remainder of the insects do not vary, show variability within lice is remarkable. Lice therefore have the potential to illuminate many aspects of mt genome biology, and the major impediment to this understanding is a lack of mt genomes covering the phylogenetic diversity of lice. Such additional data, coupled with the understanding of the nuclear encoded, mitochondrially functional genes that will be made available by the impending release of the nuclear genome of the human body louse, Pediuclus humanus, (Pittendrigh et al. 2006) will make lice a potent model system for understanding the evolution of the mt genome.