Introduction

Evolutionary biologists have been increasingly using data from entire mitochondrial (mt) genomes, including nucleotide sequence and gene arrangement, to study the evolutionary history of Metazoa (e.g., von Nickisch-Rosenegk et al. 2001; Grande et al. 2002; Ingman and Gyllensten 2003; Scouras et al. 2004; Macaulay et al. 2005). To date, mt genomes of over 620 species of Metazoa have been sequenced (see http://evogen.jgi.doe.gov/top_level/organelles.html). The vast majority of these genomes are circular, are about 15 kb long, and have 37 genes and a single large noncoding region (LNR). These 37 genes encode the large- and small-subunit rRNAs of the mt ribosome, 13 proteins for oxidative phosphorylation, and 22 tRNAs for the translation of the proteins encoded by mt genomes (Boore 1999). So the gene content of the mt genome is highly conserved in Metazoa. The arrangement of genes in the mt genome of Metazoa is generally conserved at low taxonomic levels, such as within a genus or family, but varies substantially at high levels, such as among phyla (Boore 1999). Indeed, there are more than 120 types of gene arrangement among the species of Metazoa that have been sequenced for entire mt genomes (counted from the web site above). The difference between any two types of gene arrangement can be in the position and/or the orientation of transcription of one or more genes or blocks of genes. A gene or a block of genes may change its position and/or the orientation of transcription in an mt genome by (1) translocation to a near site in the genome, (2) translocation to a distant site in the genome, (3) inversion and translocation, (4) inversion but not translocation, and (5) gene shuffling, where most genes in a genome are translocated and/or inverted.

Comparison of the arrangement of mt genes has been shown to be a powerful tool for phylogenetic studies (Boore and Brown 1998; Dowton et al. 2002). Yet we still do not know much about how the various arrangements of mt genes evolved in Metazoa. To date, four molecular mechanisms have been proposed to account for rearrangement of mt genes in Metazoa: (1) tandem duplication of a section of a genome followed by random deletion of excess genes (Moritz and Brown 1986; Boore 2000), (2) tandem duplication followed by transcription-orientation-dependent deletion of excess genes (Lavrov et al. 2002), (3) nonhomologous intragenome recombination (Lunt and Hyman 1997; Dowton and Campbell 2001), and (4) nonhomologous intergenome recombination (Boore 2000; Shao et al. 2005a). Among these four mechanisms, tandem duplication followed by random deletion is relatively well supported (Moritz and Brown 1986; Boore 2000); there is limited evidence so far for the other three mechanisms.

We recently reported that the mt genome of a chigger mite, Leptotrombidium pallidum, has a novel gene content—it has an extra gene for large-subunit rRNA, a pseudo-gene for small-subunit rRNA, and three extra LNRs, additional to the 37 genes and LNR that are typical of Metazoa (Shao et al. 2005a). Further, the arrangement of mt genes of L. pallidum differs drastically from that of the hypothetical ancestor of the arthropods. To find to what extent the novel gene content and gene arrangement occurred in Leptotrombidium, we sequenced the entire mt genomes of two more Leptotrombidium species, L. akamushi and L. deliense, and the partial mt genome of another species, L. fletcheri. We found that these three species have a type of mt genome that differs from that of L. pallidum in both gene content and gene arrangement. By comparison between Leptotrombidium species and the ancestor of the arthropods, we propose that (1) the type of mt genome present in L. pallidum evolved from the type present in the other three Leptotrombidium species, and (2) three molecular mechanisms were involved in the evolution of mt gene content and gene arrangement in Leptotrombidium species.

Materials and Methods

Specimen Collection, DNA Extraction, PCR Amplification, Genome Sequencing, and Annotation

L. akamushi, L. deliense, and L. fletcheri were from colonies reared in our laboratory. L. akamushi was originally from a Japanese grass vole, Microtus montebelli, from the Akiata Prefecture of Japan. L. deliense and L. fletcheri were originally from the former U.S. Army Medical Research Unit at the Institute for Medical Research in Kuala Lumpur, Malaysia. Genomic DNAs were extracted from five adult L. akamushi and five adult L. deliense using the method detailed by Shao et al. (2004) and one adult L. fletcheri using the DNeasy Tissue Kit (QIAGEN). The entire mt genomes of L. akamushi and L. deliense were amplified in two overlapping fragments (6.4 and 7.3 kb) by long-PCR and were then sequenced by a shotgun method. Two sections (3.1 and 1.6 kb) of the mt genome of L. fletcheri that were particularly pertinent to this study were amplified by PCR and were then sequenced by primer walking. PCR primers were (1) La-cox1F with La-nad6R, and La-nad6F with La-cox1R, for L. akamushi; (2) Ld-cox1F with Ld-cobR, and Ld-nad6F with Ld-cox1R, for L. deliense; and (3) Lf-nad1F with Lf-nad4R, and Lf-cox3F with Lf-nad5R, for L. fletcheri (Table 1). Our methods for PCR amplification, shotgun sequencing, and mt genome annotation were described by Shao et al. (2005a). The sequences of the mt genomes of L. akamushi, L. deliense, and L. fletcheri were deposited in DDBJ and GenBank under accession numbers AB191044-5 and AY973038-9.

Table 1 Primers used to amplify the mitochondrial genomes of Leptotrombidium akamushi (La), L. deliense (Ld), and L. fletcheri (Lf)
Full size table

Construction of Secondary Structures of Large- and Small-Subunit rRNAs

To identify the domains that are absent in the rRNAs of Leptotrombidium species, we constructed the secondary structures of the large- and small-subunit rRNAs of L. pallidum. We retrieved the secondary structure of the large-subunit rRNA of the hard tick, Ixodes hexagonus, from the European ribosomal RNA database (http://www.psb.ugent.be/rRNA/) and used it as a template to construct the putative secondary structure of the large-subunit rRNA of L. pallidum. We chose I. hexagonus because it was most closely related to L. pallidum among the arthropods whose secondary structures of large-subunit rRNAs were available in the database. There was no secondary structure of small-subunit rRNA of chelicerates (ticks, mites, and their kin) in the database; only those of insects and crustaceans were available. So we constructed the putative secondary structure of the small-subunit rRNA of L. pallidum using that of the fruit fly, Drosophila yakuba, as a template.

Phylogenetic Analyses of Nucleotide Sequences of Large Noncoding Regions

We inferred the phylogeny of the LNRs of the four Leptotrombidium species to discover whether the LNRs of each species had evolved independently or in concert. Nucleotide sequences of the LNRs were aligned with ClustalX (Thompson et al. 1997). For both pairwise and multiple-sequence alignments, the gap-opening penalty was 15.00, the gap-extension penalty 2.00, and the DNA weight matrix IUB. For multiple-sequence alignments, the delay-divergent sequence was set at 30% and the DNA-transition-weight was 0.50. Neighbor-joining (NJ), maximum-likelihood (ML), and maximum-parsimony (MP) trees were inferred with PAUP 4.0b10 (Swofford 2000). The general time-reversible model and the gamma-distributed rates were used in ML analysis; the instantaneous rate matrix, base frequencies, and shape of the gamma distribution were estimated by PAUP. Bootstrap analyses (1000 replicates) were done with the NJ, ML, and MP trees.

Results

Variation in Mitochondrial Genome Size, Gene Content, and Gene Arrangement Among Leptotrombidium Species

We sequenced the entire mt genomes of L. akamushi and L. deliense, and two sections of the mt genome of L. fletcheri. The mt genomes of L. akamushi and L. deliense are circular and have 13,698 and 13,731 bp, respectively. Both mt genomes have the 37 genes that are typical of Metazoa, but both genomes have two LNRs rather than one LNR, which is typical of Metazoa. These two mt genomes are substantially smaller (by 3081 and 3048 bp, respectively) than that of L. pallidum (16,779 bp) because L. pallidum has an extra gene for large-subunit rRNA (rrnL#2), a pseudo-gene for small-subunit rRNA (PrrnS), and two extra LNRs (nos. 3 and 4), compared to L. akamushi and L. deliense (Fig. 1).

Fig. 1
figure 1

The mitochondrial (mt) genomes of the hypothetical ancestor of the arthropods and the four Leptotrombidium species. The circular mt genomes were linearized at the 5′-end of cox1 to aid comparison of their gene content and gene arrangement. Two events of tandem duplication followed by random deletion of excess genes (A), followed by one event of nonhomologous intergenome recombination (B), is the simplest scenario for the evolution of the type of mt genome present in L. pallidum from the type present in L. akamushi, L. deliense, and L. fletcheri. Dark-gray boxes represent genes that changed positions relative to the ancestor of the arthropods; light-gray boxes represent genes that changed both position and orientation of transcription. The arrow line below an LNR indicates the orientation of transcription of that LNR relative to that of other LNR(s). Scissor symbols represent deletions of excess genes. The circling arrow indicates a change in the orientation of transcription from left-to-right to right-to-left. Asterisks indicate gene boundaries that are present in L. pallidum but not in the other three Leptotrombidium species. Note that LNR#3 of the hypothetical intermediate genome 1 became LNR#4 of L. pallidum after nonhomologous intergenome recombination (B). Protein-coding and rRNA genes are abbreviated atp6 and atp8 (for ATP synthase subunits 6 and 8), cox1-3 (for cytochrome c oxidase subunits 1–3), cob (for cytochrome b), nad1-6 and 4L (for NADH dehydrogenase subunits 1–6 and 4L), and rrnL and rrnS (for large- and small-subunit rRNAs). tRNA genes are shown with the single-letter abbreviations of their corresponding amino acids. The two tRNA genes for leucine are L 1 (anticodon nag) and L 2 (yaa), and those for serine are S 1 (nct) and S 2 (nga). Genes are transcribed from left to right except those that are underlined, which are transcribed from right to left.

Full size image

The arrangement of genes and LNRs in the mt genome of L. akamushi is identical to that of L. deliense. L. fletcheri apparently also has this arrangement of genes and LNRs, but we sequenced only two sections (4.7 kb long in total) of its mt genome (Fig. 1). The arrangement of mt genes and LNRs of L. akamushi and L. deliense differs substantially from that of the hypothetical ancestor of the arthropods inferred by Staton et al. (1997), since (1) 30 of the 39 gene boundaries in the mt genomes of L. akamushi and L. deliense are novel for an arthropod, (2) at least 16 of their 37 genes have changed position, and (3) another 10 genes have changed both position and orientation of transcription relative to their counterparts of the ancestor of the arthropods (Fig. 1). The arrangement of genes and LNRs of L. akamushi, L. deliense, and L. fltecher also differs from that of L. pallidum in that (1) five novel gene boundaries, which are associated with the extra rrnL, PrrnS, and two extra LNRs, in the mt genome of L. pallidum are not present in L. akamushi, L. deliense, and L. fletcheri; and (2) trnQ is between rrnS and an LNR in L. pallidum but is between trnY and rrnL in L. akamushi, L. deliense, and L. fletcheri (Fig. 1).

Reduced Sizes of Mitochondrial Genes of Leptotrombidium Species

The mt genes of L. akamushi, L. deliense, and L. fletcheri have the same or similar sizes as their counterparts in L. pallidum (Table 2). The mt genes of these four Leptotrombidium species are shorter or substantially shorter (by 0.4–30.1%; Table 2) than their counterparts in D. yakuba, which are in the middle range among those of the arthropods studied (Crease 1999; Navajas et al. 2002). Like those of L. pallidum, the inferred mt proteins of L. akamushi, L. deliense, and L. fletcheri also lack part of the domains that are present in D. yakuba (data not shown but see Shao et al. 2005a). Most of the inferred mt tRNAs of L. akamushi, L. deliense, and L. fletcheri lack stem-and-loops on either the D-arm or the T-arm and, thus, cannot form the conventional cloverleaf structures (Fig. 2). The fact that the unusual secondary structures of tRNAs we found in L. pallidum are also conserved in L. akamushi, L. deliense, and L. fletcheri provides additional evidence to our view that the mt tRNA genes of L. pallidum encode a complete set of the functional tRNAs, in spite of their unusual secondary structures (Shao et al. 2005a).

Table 2 Comparison of the sizes of mitochondrial genomes and genes of the chigger mite Leptotrombidium species with those of the fruit fly Drosophila yakuba
Full size table
Fig. 2
figure 2

Alignment of the sequences of the mitochondrial tRNA genes of Leptotrombidium akamushi (La), L. deliense (Ld), L. fletcheri (Lf), and L. pallidum (Lp). tRNA genes are named with the single-letter abbreviations of their corresponding amino acids. Nucleotides that pair at the arms are underlined. Sequences of anticodons are in boldface. Conserved nucleotides are shaded in gray.

Full size image

We misannotated the size of rrnL of L. pallidum as 1282 bp in Shao et al. (2005a). In that paper, we could not determine accurately the 5′-end of rrnL of L. pallidum because an LNR lies upstream rrnL. In L. akamushi, L. deliense, and L. fletcheri, however, trnQ lies upstream of rrnL; this allows us to determine the 5′-end of rrnL with more accuracy than in L. pallidum. We annotated the 5′-end of rrnL of L. akamushi, L. deliense, and L. fletcheri as the first nucleotide downstream of trnQ. By alignment and comparison of nucleotide sequences, we annotated that rrnL of L. pallidum is 1008 bp long, which is similar to that of L. akamushi (1014 bp), L. deliense (1024 bp), and L. fletcheri (1014 bp; Table 2). The 274 bp that we misannotated as part of rrnL of L. pallidum in Shao et al. (2005a) has no significant similarity to any mt genes; we thus annotated this 274 bp as part of the LNR that lies upstream of rrnL (see also the section below).

The average length of rrnL of the four Leptotrombidium species we studied is shorter by 23.5% than its counterpart in D. yakuba (Table 2). This is also the case for rrnS in these four Leptotrombidium species—its average length is shorter by 23.4% than its counterpart in D. yakuba. We constructed the secondary structures of the large- and small-subunit (LSU and SSU) rRNAs of L. pallidum to identify which domains of LSU and SSU rRNAs were absent due to the substantial size -reduction of rrnL and rrnS. In comparison with that of D. yakuba, the LSU rRNA of L. pallidum lacks all the helixes and nucleotides from helix D2 to the 5′-end, including D1, C1, B20, B12, B11, and B9′ (Fig. 3A). All the helixes from D2 to the 3′-end in the LSU rRNA of D. yakuba are present in L. pallidum, except G13. The SSU rRNA of L. pallidum also lacks the helixes at the 5′-end that are present in D. yakuba, including helixes 12, 8, 7, 6, 5, 4, 2, and 1 (Fig. 3B). In the middle of SSU rRNA, helix 22 is absent in L. pallidum. Only one helix each is present in L. pallidum in the section where three helixes, 24, 25, and 26, are present, and in the section where four helixes, 39, 40, 41, and 42, are present in D. yakuba. We did not construct secondary structures for the LSU and SSU rRNAs of L. akamushi, L. deliense, and L. fletcheri. However, the domains of LSU and SSU rRNAs that are absent in L. pallidum are likely to be absent in these three species too, because rrnL and rrnS of these three species have the same or similar sizes to those in L. pallidum (Table 2) and share a high nucleotide sequence similarity with those in L. pallidum (data not shown).

Fig. 3
figure 3

Putative secondary structures of the large-subunit rRNA (A) and the small-subunit rRNA (B) of Leptotrombidium pallidum. Dots indicate Watson-Crick bonds and bonds between U and G. The numbering of the helixes is after de Rijk et al. (1997) for large-subunit rRNA and after van de Peer et al. (1997) for small-subunit rRNA.

Full size image

Duplicate and Quadruple Large Noncoding Regions of Leptotrombidium Species

Both L. akamushi and L. deliense have two LNRs in their mt genomes. The two LNRs of each of these two species have similar sizes and near-identical nucleotide sequences (Tables 2 and 3) but have opposite orientations of transcription (Fig. 1). L. fletcheri also has two near-identical LNRs, which are in the same relative positions as the two LNRs of L. akamushi and L. deliense. Relative to L. akamushi, L. deliense, and L. fletcheri, L. pallidum has two extra LNRs (nos. 3 and 4; see Fig. 1). The four LNRs of L. pallidum have sections with near-identical nucleotide sequences (Table 3), but these four LNRs differ substantially in size and have opposite orientations of transcription (Table 2, Fig. 1). The two or four LNRs of a Leptotrombidium species are more similar to each other in size and nucleotide sequence than they are to the LNRs of other Leptotrombidium species (Table 3). In a similar vein, our phylogenetic analysis of the LNR sequences revealed that the two or four LNRs of each Leptotrombidium species are more closely related to each other than they are to the LNRs of other Leptotrombidium species (Fig. 4).

Table 3 Comparison of sequence similarity (%) among the large noncoding regions (LNRs) of L. akamushi (La), L. deliense (Ld), L. fletcheri (Lf), and L. pallidum (Lp)
Full size table
Fig. 4
figure 4

An unrooted neighbor-joining (NJ) tree inferred from the nucleotide sequences of the large noncoding regions (LNRs) of Leptotrombidium akamushi (La), L. deliense (Ld), L. fletcheri (Lf), and L. pallidum (Lp). The maximum-likelihood (ML) and maximum-parsimony (MP) trees (not shown) have the same topologies as the NJ tree except for the grouping among the four LNRs of L. pallidum. Percentage bootstrap support (1000 replicates) is shown near each branch. The three branches that have only one value, 100, have 100% bootstrap support from all three methods of analysis: NJ, ML, and MP.

Full size image

Discussion

Evolution of Two Types of Mitochondrial Genome in the Genus Leptotrombidium

Our results reported above revealed two types of mt genome in the genus Leptotrombidium: one is present in L. akamushi, L. deliense, and, apparently also, L. fletcheri; the other is present in L. pallidum. These two types of genome differ in genome size, gene content, and arrangement of trnQ but share reduced sizes of genes and the arrangement of all genes other than trnQ. It is unusual that species of the same genus differ in gene content and gene arrangement in their mt genomes, since gene content and gene arrangement are generally conserved at low taxonomic levels (see Boore 1999). The results above prompted us to ask two questions. First, between the two types of mt genome, which one is more likely to be ancestral to Leptotrombidium, and which one is more likely to be derived to Leptotrombidium? Second, what mechanisms of gene rearrangement were likely involved in the evolution of the mt genomes of the Leptotrombidium species from one type to another?

In order to address the first question, an outgroup is needed with which the Leptotrombidium species can be compared. In the absence of data from any outgroup species that are closely related to Leptotrombidium, we chose the hypothetical ancestor of the arthropods as an outgroup for the comparison of mt gene content and gene arrangement with Leptotrombidium species. The ancestor of the arthropods is a meaningful outgroup for such comparison, in our view, because both types of mt genome of the Leptotrombidium species can be viewed as being derived from the ancestral type of mt genome of the arthropods. Further, the gene content and gene arrangement of the ancestral type of mt genome of the arthropods remain unchanged in at least five species of the Acari studied (Shao et al. 2005a), suggesting that the ancestor of the Acari and some lineages of the Acari have the same mt gene content and gene arrangement as the ancestor of the arthropods. By comparison of the gene content and gene arrangement of the two types of mt genome of the Leptotrombidium species with those of the ancestral type of mt genome of the arthropods, we propose that it is more parsimonious to infer that the type of mt genome present in L. pallidum (with 37 genes typical of Metazoa, four LNRs, an extra rrnL, and a PrrnS; hereafter type II) evolved from the type of mt genome present in L. akamushi, L. deliense, and L. fletcheri (with 37 genes and two LNRs; hereafter type I) (Fig. 5A) than vice versa (Fig. 5B). There are three lines of evidence for our proposal. First, it requires three events of duplication to account for the evolution of two LNRs in the type I genome from one LNR in the ancestral type of mt genome of the arthropods, and then to four LNRs in the type II genome; but it requires three duplications and two deletions to account for the evolution of four LNRs in the type II genome from one LNR in the ancestral type of mt genome of the arthropods, and then to two LNRs in the type I genome (Fig. 5C). Second, it requires two duplications and two deletions to account for the evolution of the arrangement, rrnL-rrnS, in the type I genome from the arrangement, rrnS - V - rrnL, in the ancestral type of mt genome of the arthropods, and then to the arrangements, rrnL - rrnS and rrnL - PrrnS, in the type II genome; but it requires two duplications and three deletions to account for the evolution of the arrangements, rrnL - rrnS and rrnL - PrrnS, in the type II genome from the arrangement, rrnS - V - rrnL, in the ancestral type of mt genome of the arthropods, and then to the arrangement, rrnL-rrnS, in the type I genome (Fig. 5D). Third, the inference that the type II genome evolved from the type I genome does not require any assumption about the phylogeny among the four Leptotrombidium species (Fig. 5A). However, the inference that the type I genome evolved from the type II genome requires the assumption that L. akamushi, L. deliense, and L. fletcheri are more closely related to each other than either of them is to L. pallidum (Fig. 5B). Otherwise, it would be difficult to explain why L. akamushi, L. deliense, and L. fletcheri had exactly the same deletions (Figs. 5C and D). There is no evidence so far for this assumption, although it is not necessarily wrong.

Fig. 5
figure 5

Evolution of the two types of mitochondrial (mt) genomes of the Leptotrombidium species. Circles represent the ancestral type of mt genome of the arthropods, squares represent the type I genome of the Leptotrombidium species, and diamonds represent the type II genome of the Leptotrombidium species. A shows the inference that the type II genome evolved from the type I genome, and B shows the opposite inference. C and D illustrate the minimum events required to account for the evolution of the large noncoding region (LNR) and the arrangement of rrnL and rrnS, respectively, under the two inferences shown in A and B. L. Leptotrombidium; dup., duplication(s); del., deletion(s). See the legend to Fig. 1 for the abbreviation of gene names. Arrow lines indicate the direction of evolution. X symbols indicate deletions. Genes underlined and not underlined have opposite orientations of transcription in an mt genome. Genes linked by a hyphen are next to each other; genes linked by dots are separated from each other. The numbering and the order of LNRs in C and D are consistent with those in Fig. 1; so is the order of rrnL#1-rrnS and PrrnS-rrnL#2. Asterisks denote that V is not immediately downstream of rrnS in the mt genomes of Leptotrombidium species (see Fig. 1).

Full size image

On the basis of the proposal above, we were able to address the second question: What mechanisms of gene rearrangement were likely involved in the evolution of the type of mt genome present in L. pallidum from the type of mt genome present in L. akamushi, L. deliense, and L. fletcheri? Several scenarios may be speculated but the most parsimonious one, in our view, is that two mechanisms of gene rearrangement were involved: (1) tandem duplication of a section of mt genome followed by random deletion of excess genes and (2) nonhomologous intergenome recombination. First, the gain of LNR#3 between Y and rrnL#1 in L. pallidum and the translocation of Q from between Y and rrnL in L. akamushi, L. deliense, and L. fletcheri to between rrnS and LNR#1 in L. pallidum, i.e., from Y-Q-rrnL-rrnS-LNR#1 to Y-LNR#3-rrnL#1-rrnS-Q-LNR#1, were generated by two events of tandem duplication of a section of genome followed by random deletion of excess genes (Fig. 1A). Tandem duplication followed by random deletion is a well-documented mechanism for mt gene rearrangement (see Moritz and Brown 1986; Boore 2000). The change from Y-Q-rrnL-rrnS-LNR#1 to Y-LNR#3-rrnL#1-rrnS-Q-LNR#1 bears four typical characteristics of this mechanism: (1) a section of a genome (LNR) was duplicated; (2) neighbor genes (Q and rrnL - rrnS) swapped positions; (3) no changes in the orientation of transcription occurred; and (4) the section involved in the change contains LNR and, thus, is at the hot spot for tandem duplication (Boore and Brown 1998). In the case of this change, there is no need to invoke any other mechanisms, i.e., tandem duplication followed by transcription-orientation-dependent deletion of excess genes or nonhomologous intra- or intergenome recombination (see also Introduction), because there is no sign suggestive of these mechanisms, e.g., change of the orientation of transcription or distant translocation of single genes. Second, the gain of the section PrrnS-rrnL#2-LNR#4 between LNR#2 and W in L. pallidum, in addition to having LNR#3-rrnL#1-rrnS between Y and Q (genes underlined and not underlined have opposite orientations of transcription in the genome), was generated by a single event of nonhomologous intergenome recombination (Fig. 1B). The only alternative explanation is that tandem duplication generated LNR#3-rrnL#1 - rrnS-LNR#4-rrnL#2-PrrnS first and then nonhomologous intragenome recombination moved and inverted LNR#4-rrnL#2-PrrnS to be between LNR#2 and W. This explanation requires one event of tandem duplication and one event of nonhomologous intragenome recombination and, thus, is less parsimonious than the explanation of a single event of nonhomologous intergenome recombination (see also Shao et al. 2005a).

Concerted Evolution of Duplicate or Quadruple Large Noncoding Regions in Each Leptotrombidium Species

Shao et al. (2005b) showed that duplicate LNRs (also called control regions) in the mt genome of Metazoa tend to evolve in concert rather than independently. For all other species of Metazoa that have duplicate LNRs, the two LNRs in an mt genome have the same orientation of transcription. However, for Leptotrombidium species, the two or four LNRs in an mt genome have opposite orientations of transcription (Fig. 1). Yet two lines of evidence indicate that the LNRs of each Leptotrombidium species have also evolved in concert. First, the nucleotide sequences of the two or four LNRs of a Leptotrombidium species are substantially more similar to each other (92–99%; Table 3) than they are to the LNRs of other Leptotrombidium species (56–76%). Second, the two or four LNRs of each species always clustered together with substantial bootstrap support in our phylogenetic analyses (86–100%), regardless of whether the NJ, ML, or MP method was used (Fig. 4). Two mechanisms may account for the concerted evolution of duplicate LNRs: tandem duplication of a section of genome, followed by random deletion of excess genes, and gene conversion (Kumazawa et al. 1998). The former, however, cannot account for the concerted evolution of duplicate LNRs that have opposite orientations of transcription. The latter, gene conversion, can account for the concerted evolution of LNRs that have opposite orientations of transcription, such as those in the Leptotrombidium species. Gene conversion is a type of homologous recombination. For gene conversion to occur, the two LNRs of an mt genome need to be physically close to each other, and to align with each other in the same orientation of transcription, so that a Holliday structure can form (Kumazawa et al. 1998). It is known that in the fruit fly, Drosophila melanogaster, the mt genomes exist in two distinct and stable superhelical forms: one has few turns, whereas the other has many turns (Rubenstein et al. 1977). Such superhelical forms of the mt genome make it possible that a Holliday structure can form between two LNRs of an mt genome, regardless of whether the two LNRs have the same or opposite orientations of transcription.

In conclusion, we found that two types of mt genome exist in the genus Leptotrombidium: one in L. akamushi and L. deliense, and probably in L. fletcheri too; the other in L. pallidum. It is more likely that the type of mt genome present in L. pallidum (with 37 genes typical of Metazoa, four LNRs, an extra rrnL, and a PrrnS) evolved from the type of mt genome present in L. akamushi, L. deliense, and L. fletcheri (with 37 genes and two LNRs) than vice versa. The simplest scenario for this evolution involves two mechanisms: tandem duplication of a section of genome followed by random deletion of excess genes and nonhomologous intergenome recombination. A third mechanism, gene conversion, is required to account for the concerted evolution of the duplicate or quadruple LNRs that have opposite orientations of transcription in each species of Leptotrombidium. The Leptotrombidium chigger mites thus provide evidence for the action of multiple molecular mechanisms in the evolution of gene content and gene arrangement in the mt genomes of Metazoa.