Large gene overlaps and tRNA processing in the compact mitochondrial genome of the crustacean Armadillidium vulgare

Abstract

A faithful expression of the mitochondrial DNA is crucial for cell survival. Animal mitochondrial DNA (mtDNA) presents a highly compact gene organization. The typical 16.5 kbp animal mtDNA encodes 13 proteins, 2 rRNAs and 22 tRNAs. In the backyard pillbug Armadillidium vulgare, the rather small 13.9 kbp mtDNA encodes the same set of proteins and rRNAs as compared to animal kingdom mtDNA, but seems to harbor an incomplete set of tRNA genes. Here, we first confirm the expression of 13 tRNA genes in this mtDNA. Then we show the extensive repair of a truncated tRNA, the expression of tRNA involved in large gene overlaps and of tRNA genes partially or fully integrated within protein-coding genes in either direct or opposite orientation. Under selective pressure, overlaps between genes have been likely favored for strong genome size reduction. Our study underlines the existence of unknown biochemical mechanisms for the complete gene expression of A. vulgare mtDNA, and of co-evolutionary processes to keep overlapping genes functional in a compacted mitochondrial genome.

Introduction

Metazoan mitochondrial DNA (mtDNA) is highly compact. Its classical content is 37 genes comprising 2 rRNA (rRNA) genes, 13 genes coding for proteins of the respiratory chain, 22 tRNA (tRNA) genes and a D-loop regulatory region, the major noncoding part of the genome. The coding sequences of mitochondrial genes are usually separated by at most a few nucleotides.¹ The genome is transcribed as 3 large polycistronic RNAs. Many mRNA (mRNA) coding sequences are immediately adjacent at their 3′ end to tRNA genes and according to the tRNA punctuation model, primary transcripts are processed by precise endonucleolytic cleavages by RNases P and Z before and after each tRNA to release mature rRNAs and mRNAs.^2-4 In addition, several mRNAs excised from the precursor molecules end with U or UA at their 3′-termini, and require polyadenylation to create a functional stop codon.⁵ The 5′ extremity of mature mRNAs usually begins with the initiation codon and there is usually no 5′ UTR. However, even if the punctuation model is often used in the mt metazoan world, it is known since the sequence of the human mtDNA has been obtained that this mechanism cannot be applied in all cases (for a review, see ref. 6). For instance, in human mitochondria, 4 mRNA sites must be cleaved following tRNA-independent processes.^4,7 Furthermore, in silico analyses allowed to speculate that overlapping sequences may code simultaneously for tRNAs and mRNAs in many metazoan species, but without any direct demonstration so far.⁸ A major reason why this question remains open is the difficulty to precisely annotate and localize expressed tRNA genes in animal mtDNA.

Indeed, while variations in protein-coding and rRNA genes number are rare, the number of identified tRNA genes varies in almost 16% of arthropod mitochondria (e.g. ^9-11). Several reasons can explain this extensive variation: increase can be due to changes in the genetic code or a consequence of tRNA gene duplication, while decrease is generally due to the loss of mt tRNA genes during evolution, compensated by the import of nucleus-encoded tRNAs, a classical phenomenon in mitochondria.^12,13 Alternatively, variation in tRNA gene number in annotated genomes may also be the consequence of defects of tRNA search algorithms that identify false positive tRNA or conversely fail to detect them. Metazoan mt tRNAs often appear degenerated, without the complete cloverleaf secondary structure: absence of the highly conserved D-loop and/or T-loop and even complete arms renders their detection problematic.^14,15 As a matter of fact, in addition to the classical tool tRNAscan-SE¹⁶ that detects cloverleaf structures, other more specialized tools such as ARWEN¹⁷ and MiTFi¹⁴ were developed to significantly improve metazoan mt tRNA gene identification. Moreover, tRNA genes can also be missed because of extensive RNA editing that modifies the RNA sequence as compared to the genomic DNA sequence. In mitochondria, several editing processes have been observed: 3′- replacement editing (mainly due to polyadenylation) in various metazoans, 5′-replacement editing in protists and chytridomycete fungi and insertion editing in slime molds (e.g.,^18,19). Finally, heteroplasmic mtDNA may allow for the presence of two tRNA genes with different anticodons at the same genome position, an example of “hidden” tRNA recently described in the mtDNA of Oniscidea.^20-22 Unfortunately, in most cases, experimental data to validate or invalidate the expression of tRNA candidates are mostly unavailable yet.

A correct expression of the mtDNA is essential for life and its perturbation may cause multiple syndromes. The importance of RNA maturation in the regulation of metazoan mt gene expression is far from being fully understood (for a review, see refs. 6, 23). Here, we focused our work on the identification and expression of overlapping genes generated during the evolution of metazoan mitochondrial genomes. For that purpose we choose an appropriate model organism, the isopod crustacean Armadillidium vulgare. With a size of 13,939 bp, its mtDNA is among the smallest in arthropods,²⁴ and the smallest described in crustacean malacostraca (average malacostraca mtDNA size of 15,747 bp, calculated from 136 species with complete sequences available on NCBI in April 2015). Although it codes for the usual set of 13 protein-coding genes and two rRNA subunits, only a partial set of mitochondrial tRNA genes was identified by computational means.^20,25 This partial set of tRNA genes was however not sufficient to account for a size reduction of 1.8 kbp in comparison to average malacostraca mt genome. Additional size reduction is due to shortness of the control region,²⁴ but also to putative gene overlaps, implicating tRNA genes. Based on these observations, we first reanalyzed the tRNA gene content and experimentally validated the tRNA candidates. Then we demonstrated that a tRNA expressed from a single truncated mt gene and overlapping with surrounding genes, is extensively repaired at its 3′ end. Likewise we showed that many expressed tRNA genes overlap with other genes. Extreme cases were encountered such as a large overlap between two tRNA genes transcribed in the same direction or tRNA genes either partially or fully inserted into a protein-coding gene in direct or opposite orientation. The data presented here support the existence of novel but still unknown biochemical mechanisms crucial for animal mt genes expression. From an evolutionary point of view, two remarkable consequences are worth to note. First, under selective pressure for genome size reduction, gene overlaps and dual-coding sequences may allow mtDNA genome compaction while avoiding the loss of tRNA genes. Second, this implies a strong co-evolution of both overlapping gene sequences and/or the development of repair system to keep the two genes functional. Finally, the existence of large gene overlaps may represent a powerful marker to perform phylogenetic studies on metazoan mtDNA.

Results

Mitochondrial tRNA gene expression in A. vulgare

Using tRNAscan-SE²⁶ and/or ARWEN¹⁷ software, Marcadé et al²⁰ and Kilpert et al²⁵ reported 9 and 13 tRNA genes from the A. vulgare mtDNA, respectively. Using the recently developed MiTFi pipeline under standard condition,¹⁴ 10 tRNA genes (E-value cutoff of E <0.001) were recovered (updated NCBI A. vulgare genome EF643519.3). A common set of tRNA genes exists between the three annotations but also discrepancies and several tRNA genes are missing (Table S1). It was therefore difficult to assess through in silico analysis, the real set of tRNAs expressed from A. vulgare mt genes. To address this question, in the presence of a very limited amount of total RNA available, the expression of each putative tRNA gene was analyzed by RT-PCR (or circularized RT-PCR (cRT-PCR) when required); PCR products were cloned and sequenced. Only 13 mt tRNA genes were shown expressed (Fig. 1; Table S1). Among them, we confirmed the expression of both tRNAs valine and alanine from the previously identified valine/alanine alloacceptor tRNA gene,²¹ half of the clones analyzed being valine GTA anticodon and the others alanine GCA anticodon.

Figure 1.

Representation of A. vulgare mtDNA map with confirmed expressed genes. Sense of the arrows indicates sense of transcription. Genes coding for proteins of the respiratory chain, for rRNAs and for tRNAs are represented by black, dark gray and gray arrows respectively. Below the map, images of ethidium bromide stained gels of RT-PCR products demonstrating the expression of the tRNA genes are presented. Experiments were designed with pairs of primers specific for each tRNA (Supplementary Table 1). The presence (+) or absence (−) of reverse transcriptase (RT) during the cDNA synthesis is indicated. Amino acids specifying tRNAs are represented by the one-letter code.

Post-transcriptional maturation of tRNA^His sequence

The identification of the tRNA^His gene was particularly difficult. First because its secondary structure (Fig. 2A) presents six mismatches in the acceptor stem, which is unexpected for a functional tRNA, and second because this gene was potentially overlapping with two other genes, tRNA^Phe and nad4 (Fig. 3A). As the sequence of the specific reverse transcriptase-dependent product amplified by RT-PCR showed the expression of this gene (Fig. 1), we decided to determine the 5′- and 3′- extremities of the mature tRNA^His. For this purpose, total A. vulgare RNAs were circularized. A specific product was then amplified by RT-PCR, cloned, sequenced, and a cloverleaf structure of the mature tRNA^His deduced (Fig. 2D). In the mature tRNA^His, the base-pairing of the acceptor stem is fully restored thanks to nucleotide replacement at the 3′-end of the tRNA. In addition, the presence of a CCA triplet at the 3′ end and of an additional G_-1 guanosine residue at the 5′ end, a nearly universal feature for aminoacylation by histidyl-tRNA synthetase^27-29 confirms that A. vulgare mt tRNA^His encodes a functional tRNA. Out of 95 clones, 87 sequences correspond to the fully repaired tRNA^His while 8 represent intermediate products (Fig. S1). From this analysis, we propose the following steps of A. vulgare mt tRNA^His biogenesis (Fig. S2). The primary transcript is processed at the 5′- end by the mt RNase P, then a RNase Z activity may cleave unusually between the T stem and the acceptor stem. Although no processing intermediates have been found, we cannot exclude a classical cleavage of RNase Z followed by the involvement of a 3′- to 5′- exonuclease to remove the unpaired nucleotides. The 3′-end of the tRNA is then repaired. Two ways to restore base-pairing in the stem are possible. Either the truncated sequence is repaired by a trial and error process involving nucleotides elongation/degradation until the sequence is correct (a tRNA repair process demonstrated in metazoan mitochondria³⁰), or the 5′-end of the acceptor stem is used as a template for synthesis of the 3′-end by a RNA-dependent RNA polymerase (RdRp) (as suggested in centipede mitochondria³¹). No tRNA processing intermediates showing mis-incorporated nucleotides were found, thus the action of the second repair process appears to be more likely. The two last nucleotides before the CCA-end are two As. The A₇₂ pairs with the U₊₁ while the A₇₃ is the discriminator nucleotide. To add these As, three enzymatic activities are possible, either a RdRp, a polyA polymerase, or the mt tRNA nucleotidyltransferase that classically adds the CCA-termini.³⁰ Finally, the G_-1 is incorporated thanks to a tRNA^His-dependent guanylyl transferase. Indeed, ESTs encoding a mt homolog of Thg1p, the cytosolic S. cerevisiae tRNA^His-dependent guanylyl transferase,³² were retrieved from A. vulgare EST libraries (Fig. S3). As in human and mouse, the A. vulgare enzyme possesses a putative N-terminal mt targeting sequence and this mt enzyme likely adds the essential G_-1 residue.

Figure 2.

The biogenesis of A. vulgare mt tRNA^His requires repair at both the 5′- and 3′-ends. (A) Inferred tRNA secondary structure of the tRNA^His primary transcript deduced from the genomic sequence. The primers P1/P2 used for cRT-PCR analyses are depicted by black arrows, their precise positions are presented on Table S2. (B) Total nucleic acids were analyzed by cRT-PCR using primers P1 and P2. After circularization in the presence of T4 RNA ligase, cDNA was synthezized in the presence of primer P1. The image shows the ethidium bromide stained gel of the PCR product amplified using primers P1 and P2. The lane marked M shows the migration of the DNA ladder. (C) Sample sequence of one of the clones obtained from the cRT-PCR product shown in (B). The 3′- and 5′- extremities of the tRNA molecule are presented. The two extremities are separated by a vertical arrow. The CCA extremity is underlined in bold. Note that primers used for tRNA-His sequencing (Table S2) were designed from the online reference genome sequence (GenBank accession EF643519.3), which presents one nucleotide polymorphism in the variable loop (indicated by an asterisk). This did not affect the amplification of cDNA and the demonstration of post-transcriptional modification. (D) Inferred secondary structure of the mature tRNA^His deduced from the sequence of the cloned cRT-PCR product. Nucleotides differing between the genomic sequence and the RNA sequence are indicated in red. (E) Comparison between the sequence of the genomic DNA region coding for the A.vulgare mt tRNA^His and the sequence deduced from the corresponding RNA amplified by cRT-PCR.

Figure 3.

Overlapping A. vulgare mt tRNA genes. Double stranded DNA regions coding for overlapping genes are presented. tRNA genes are under gray backgrounds. Horizontal arrows indicate the sense of transcription on RNA sequences. Stars indicate UAA stop codons or U residues requiring 2 As residues to generate them. For nad4 and nad6, RNA sequences not present in vivo are in italics and UAA stop codons not used in vivo are underlined. (A) overlaps nad4/tRNA^His and tRNA^His/tRNA^Phe, (B) absence of overlap between nad6/tRNA^Ser, (C) overlap tRNA^Val/Ala/nad1, (D) overlaps tRNA^Cys/tRNA^Tyr and tRNA^Tyr/cox1, (E) potential overlap cytb/tRNA^Thr, depending on the start codon used (AUA or GUG) (F) overlap cox1/tRNA^Lys.

Multiple gene overlaps in the A. vulgare mt genome

Complete sequencing of A. vulgare mt genome revealed potential large overlaps between genes,^20,25 with no functional demonstration. We further investigated the precise location of each expressed tRNA gene with regard to the other A. vulgare mt genes. To do so, we first determined the unknown extremities of the two rRNAs (Fig. S4) and the 3′ extremity of mRNAs (Fig. 3 and Fig. S5). It is to note that while the 3′ extremities of the two rRNAs are homogeneous in size and are polyadenylated in vivo, their 5′ extremities are rather heterogeneous. The 5′ extremity of the large rRNA ends within a stretch of 10 A residues (except for one sequence) while for the small rRNA, two types of 5′ extremity were found, one type about 30 nucleotides longer than the other. In both cases, whether this reflects in vivo situation, instability of the 5′ extremity or a cloning artifact is unknown.

With the exception of tRNA^Gln, tRNA^Trp and tRNA^Ser, the other tRNA genes may overlap with other tRNA or mRNA genes. From the in silico analysis, the tRNA^His gene seems to overlap with two adjacent genes. On the opposite strand, tRNA^His overlaps with tRNA^Phe over 19 nt (Fig. 3A). As mt tRNA^His and tRNA^Phe are encoded on different strands, their processing from two independent primary transcripts is easily achieved. In the same orientation of transcription, tRNA^His shows an apparent minimum overlap with nad4 over 17 nt, while considering the UAA codon as a stop codon. However, after sequencing the exact 3′- end of nad4 mRNA (Fig. S5), we show that an alternative and truncated stop codon is polyadenylated (addition of two A residues to complete a UAA stop codon) (Fig. 3A), a common phenomenon in metazoan.⁶ This reduces the size of the gene and avoids overlap between nad4 and tRNA^His genes.

As shown for tRNA^His, in silico analysis suggested that tRNA^Ser overlaps with nad6 (Fig. 3B). However, this was invalidated after the determination of the 3′- end of nad6 mRNA, showing a shortened version of the gene (with a stop codon generated by addition of a polyA tail), thus excluding any overlap. By contrast, the 65 nt alloacceptor tRNA^Val/Ala gene fully overlaps with nad1 gene (in antisense orientation) while the overlap was expected to be of “only” 26 nt from in silico analysis when considering the UAA stop codon as the 3′ extremity of the mRNA (Fig. 3C). Even more surprising, tRNA^Tyr overlaps with two genes, the tRNA^Cys on the same strand, over 22 nt (Fig. 3D and Fig. S6), and cox1 on its opposite strand, over 3 nt. The tRNA^Cys gene is completely included in the nad2 gene. The two genes are transcribed in opposite direction and the nad2 mRNA is 78 nt longer than predicted in silico. The tRNA^Thr and cytb genes, transcribed on the same strand (Fig. 3E), may share 26 nt if we consider codon AUA as a start codon. However, if the codon GUG is used to initiate translation, a situation already described for mt gene expression, (e.g.^33,34) then there is no overlap and the processing of tRNA^Thr will release the 5′ extremity of the cytb mRNA. As we failed in determining the 5′ extremity of cytb mRNA, none of these two possibilities can be excluded. Finally the 60 nt long tRNA^Lys is fully integrated into the coding sequence of cox1 (Fig. 3F). This latter observation was validated by the identification of the 3′ extremity of cox1 mRNA, where polyadenylation creates a stop codon (Fig. 4A). Moreover, the tRNA^Lys gene is inserted into a conserved region of cox1 (Figure S7). Finally, the expression of a mature tRNA^Lys was validated by cRT-PCR followed by cloning and sequencing of the PCR product (Figs. 4B and 4C).

Figure 4.

tRNA^Lys gene lies within cox1 gene. (A) Sample sequence of one of the clones obtained from the analysis of the 3′ extremity of cox1 mRNA. The polyA tail found at the 3′ extremity is indicated by a vertical arrow and the generated stop codon underlined. This confirms the complete expression of cox1. (B) Sample sequence of one of the clones obtained from the cRT-PCR product amplified in the presence of primers specific of A. vulgare mitochondrial tRNA^Lys. The 3′- and 5′- extremities of the tRNA molecule are separated by a vertical arrow. The post-transcriptionally added CCA extremity is underlined. (C) Deduced cloverleaf structure of the A. vulgare mitochondrial tRNA^Lys.

Discussion

A. vulgare mt genome encodes a limited set of tRNA genes

A. vulgare mtDNA was shown to encode a limited set of 13 tRNA genes. Among them, degenerated tRNA genes were identified: several genes have reduced D- or T-arms, tRNA^Phe, tRNA^Pro , tRNA^Trp and tRNA^Gln have no T-arm and tRNA^Cys has no D-arm.^20,25 This is in agreement with the multiple observations of tRNA genes with short D-and/or T-arms or even lacking at least one of them in metazoan mitochondria,^14,35 one way toward genome size reduction. The extreme situation such as the existence of armless mt tRNA genes in nematodes Enoplea^15,36 has not been observed here.

Although we cannot exclude that very short, post-transcriptionally modified and/or “bizarre” tRNA genes escaped our detection, it is unlikely that a full set of functional tRNAs is encoded by the mt genome of A. vulgare, where up to 9 mitochondria-encoded tRNAs are likely missing. Indeed, many mtDNAs have now lost at least few tRNA genes and the absence of an apparently full set of mt tRNA genes is rather frequent.¹⁰ For instance, in Cnidaria or Chaetognatha, nearly all mt tRNA genes have been lost (e.g.,^14,37). In seven other isopod mt genomes, various numbers (from 1 to 8) of missing tRNA genes were observed.²⁵ The missing genes are likely compensated by mitochondrial import of nucleus-encoded tRNAs, a widespread phenomenon in evolutionary divergent organisms.¹⁰ The loss of tRNA genes in A. vulgare mtDNA may have happened together with the reduction of the genome size and of the occurrence of multiple rearrangements within the mtDNA of isopod crustaceans.^20,25

A. vulgare mt tRNA^His is extensively repaired

A second strategy to cope with genome size reduction and to maintain functional tRNA genes is to repair truncated tRNAs expressed from degenerated genes. A. vulgare mt tRNA^His is extensively repaired at its 3′ extremity in order to restore amino acid acceptor stem base-pairing, and classically a G_-1 is added at the 5′ end to allow the recognition by the mt Histidyl-tRNA synthetase. This is an extreme case where both ends need to be edited. Integrity of the 3′- extremity of tRNAs is essential for their functionality and truncated mt metazoan tRNAs lacking up to 6 nt at their 3′- end were previously shown to be repaired.³⁸ In most cases, repairs likely involve the tRNA nucleotidyltransferase and/or a polyA polymerase. However, a few observations are consistent with the existence of another type of polymerases able to add nucleotides to the 3′-end of tRNAs.^30,39,40 The data presented here strongly suggests that a RdRp is involved in the synthesis of the 3′-end using the 5′-end of the tRNA^His acceptor stem as a template.

Overlapping genes in the A. vulgare mt genome

In highly compact metazoan mtDNAs, most mRNA coding sequences are immediately adjacent at their 3′ end to tRNA genes and according to the tRNA punctuation model (Fig. 5A), primary transcripts are processed by precise endonucleolytic cleavages by RNases P and Z before and after each tRNA to release mature mRNAs.^2-4 These mRNAs possess no significant 5′UTR and several of them require polyadenylation to create a functional stop codon.⁵ As described above, the tRNA punctuation model can be applied to several A. vulgare genes (Fig. 5A; e.g. nad3 with the alloaceptor tRNA^Val/Ala gene).

Figure 5.

RNA processing of overlapping mitochondrial genes. (A) In most cases of RNA processing of animal mtDNA, the DNA is transcribed into polycistronic RNAs. mRNAs are punctuated by tRNAs. The tRNAs are processed by RNases P and Z, thus excising mRNAs.^2-4 When mRNAs are incomplete, polyadenylation is used to create a stop codon, and mRNA is translated.⁵ In A. vulgare, this tRNA punctuation model is valid for part of the RNAs, including the tRNA^Val-Ala (V-A in the gray box), adjacent of the nad3 gene. In cases of gene overlaps, possible additional models are as follow. (B) When there is overlap between a tRNA and the 5′-end of the coding sequence of a mRNA, either cleavage occurs unusually releasing a full-length mRNA and a truncated tRNA that is further repaired or the tRNA is classically processed, thus excising an incomplete mRNA (ΔmRNA) that can be either translated thanks to an unusual initiation codon²⁵ or degraded. This is the case of tRNA^Thr (T) which overlaps with the gene cytb. (C) When a tRNA is integrated within the coding sequence of a mRNA, alternative processing events are required to get both RNA molecules functional. If a tRNA cloverleaf structure is recognized by the tRNA processing machinery, a mature tRNA is generated and truncated mRNA enters the degradation pathway. If an alternative folding of the RNA molecule does not allow the cleavage by RNAses P and Z, the mRNA is engaged in the translation process. This is the case of tRNA^Lys (K) embedded in the coding sequence of cox1. (D) When two tRNAs oriented in the same direction overlap, two alternative tRNA foldings may allow the processing of two tRNAs. The sequence common to both tRNAs is schematically presented in dark gray, and the sequences specific to tRNA₁ and tRNA₂ in black and pale gray respectively. This is the case for the overlapping tRNA^Cys (C) and tRNA^Tyr (Y). Stars represent initiation codon. To note, the other post-transcriptional modifications (CCA addition, nucleotides modification) required to get functional tRNAs are not presented here. Examples of A.vulgare genes where each potential model can be applied are presented under gray background.

The presence of overlapping genes represents another way to shorten a genome. Short overlaps of 1 to 3 nt have been identified, mainly in silico, in several metazoan mtDNAs.¹⁴ In A. vulgare mtDNA, potentially only three short overlaps exist. Cox1 and tRNA^Tyr genes are in opposite direction and overlap over 3 nt. A 2 nt overlap between tRNA^Pro and nad4L and a 6 nt overlap between tRNA^Met and nad2 may also exist (Fig. S8). Alternatively, if other downstream initiation codons are used, there will be no overlap. By contrast, the presence of potential larger overlaps between genes of various arthropods, including A. vulgare, has been postulated with no demonstration yet.^25,41-44 Here we show that, indeed, the other overlaps are larger, from 19 to 65 nt, which is much more than usually described. When two overlapping genes are on different strands (e.g., the alloacceptor tRNA^Val/Ala gene and nad1), cleaving the two independent primary transcripts will directly generate functional RNAs. By contrast, when the two overlapping genes are on the same strands (e.g., tRNA^Cys and tRNA^Tyr; tRNA^Lys within cox1), the ways to produce functional RNAs are less obvious. Several non-exclusive processes are possible (Fig. 5). In the case of overlap between 2 tRNAs (e.g., tRNA^Cys and tRNA^Tyr), processing implicating alternative secondary structure of the primary transcript appears to be the most plausible explanation (Fig. 5D). Possible alternative processing has already been reported for the squid Loligo bleekeri tRNA^Tyr and tRNA^Cys but in that case it involves only a one nt overlap,⁴⁰ while here the 22 nt overlap likely implies stronger constraints. If the overlap is between a tRNA and a mRNA (Fig. 5B), classical processing of the tRNA precursor will release a truncated mRNA and a non classical codon can be used to initiate translation, leading to a slightly shorter protein. The use of alternative start codons in isopod mt mRNAs has already been reported.²⁵ Otherwise, if processing occurs at the expected initiation codon of the mRNA, a truncated tRNA is generated, and repair would imply the re-synthesis of more than one third of the complete tRNA, which is rather unlikely but cannot be excluded as observed for tRNA^His that requires extensive repair. Alternative processing might be possible for the production of either a complete mRNA or a complete tRNA (Fig. 5C), a mechanism that would need to be elucidated. This last explanation is the most likely for the tRNA^Lys gene that is fully integrated into cox1 gene, as we confirmed the complete expression of both genes. In all potential alternative processing pathways proposed here, protein factors are likely implicated and need to be characterized.

From an evolutionary point of view, rather than to represent an exception, the situation found in A. vulgare mitochondria is likely more widespread than previously thought. Based on computational analysis, the presence of mitochondrial sequences that simultaneously code for tRNAs and proteins has already been envisaged in nematodes, arthropods or onychophora but was not demonstrated yet.⁸ Large overlaps between protein coding genes were also reported in cnidarians and vertebrates.^7,45,46 When looking for bizarre tRNA genes expressed from mtDNA, the investigation is usually oriented toward intergenic regions. Here, we show the necessity to screen for tRNA genes all along the DNA sequence. This is particularly true for genomes where tRNA genes are apparently missing and we can speculate that the number of tRNA genes that do not follow the tRNA punctuation model will rapidly increase once we will start to look for them.

The use of overlapping genes for genome size reduction seems generally restricted to tRNA genes. Indeed, whether they are on the same or on the opposite strand, overlaps require a strong co-evolution between gene sequences. Such constraint is more likely to be too stringent for protein-coding genes overlaps. In contrast, tRNA gene sequences evolve rapidly and recent evidence showed that a simple secondary structure coupled with a specific anticodon may be sufficient to generate a functional tRNA,^47,48 or that incomplete cloverleaf structure may be repaired post-transcriptionally (e.g., tRNA^His in this study).

Considering animal mt genome expression and regulation, the existence of a few animal mitochondrial mRNAs not delimited by tRNA genes and called tRNA-less RNA precursors, shows that their release cannot rely solely on the processing of tRNAs by RNase P and RNase Z. Indeed, the RNA binding protein GRSF1 (G-rich sequence factor 1) was recently shown to be involved in the processing of such primary transcripts.⁴⁹ Here, the existence of tRNA genes partially or fully integrated within protein coding genes transcribed in the same direction represents another situation where the tRNA punctuation model of RNA processing described in metazoan mitochondria^2,3 presents limitations. Furthermore, we also showed that A. vulgare nad1 mRNA does not end at the level of the stop codon but rather terminates at the level of the antisense sequence of the alloaceptor tRNA^Ala/Val gene sequence. This suggests that the antisense sequence of a true tRNA can be folded into a tRNA-like structure and can be recognized as a processing site to release the 3′ end of a mRNA. This intriguing but open question will need to be addressed in the future.

In addition to the tRNA punctuation model described as a major way for the release of functional RNAs, other mechanisms were early suspected to account for complete processing of animal mt primary transcripts. Our data strongly support the existence of still unknown alternative pathways to generate functional mRNAs and tRNAs in metazoan mitochondria. The importance and extent of alternative processing models such as the ones we propose here is likely more widespread than initially thought in evolutionary divergent metazoans. The pathways and protein factors involved in mitochondrial gene expression and regulation appear more complex than previously estimated and their characterization is a challenge for the future.

Materials and Methods

Total RNA extract

Five A. vulgare females from BF 2739 strain (BF) were used as starting material. The isofemale lines (i.e., a lineage of individuals all descended from a single female) are reared in the laboratory and are known to be asymbiotic for Wolbachia (i.e., A. vulgare endocellular symbiont) avoiding contamination. Total RNA was extracted from gonads, fat tissues, and nervous system using a Trizol protocol, as previously described.⁵⁰ From five A. vulgare females, not more than 5 µg of total nucleic acid can be obtained.

Cloning of reverse transcription (RT)-PCR products

A. vulgare total nucleic acid was first treated with DNASE RQ1 according to manufacturer's intructions (Promega). The DNase-treated RNA sample was then used as substrate for RT-PCR amplification using relevant pairs of primers, cloned into pGEM-T Easy (Promega), as described in ⁵¹ and sequenced.

Cloning of circular RT-PCR products

Circular reverse transcription-PCR (cRT-PCR) was used to determine 5′- and 3′- termini of tRNA^His essentially as described in.⁵¹ Briefly, total RNA extracted from A. vulgare was incubated with 40U of T4 RNA ligase (New England Biolabs) in the supplied buffer supplemented with 2U of RNase inhibitor and in a total volume of 25 µl. Following circularization, all steps of RT-PCR were classically carried out using a relevant pair of primers. PCR products were cloned into pGEM-T Easy and sequenced.

Determination of 5′- and 3′-ends of rRNAs and mRNAs

Sequence determination was mostly performed as described in.³⁶ Using T4 RNA ligase, total RNA was fused to a tagging DNA oligonucleotide that carried a ribonucleotide at the phosphorylated 5′-end and a 3′-end blocked by a 3′-3′- linked T. To determine RNA 3′-end, cDNA, synthesized by hybridizing RT primer to the tagging oligonucleotide, was then amplified in the presence of the RT primer and a second primer specific to the gene of interest. All steps of RT-PCR and cloning of PCR products were carried out as described above. To determine RNA 5′-end, the cDNA synthesized above was used in a tailing reaction with terminal deoxynucleotidyl-transferase and dCTP according to the manufacturer (Invitrogen). The reaction product was amplified by PCR using a C-tail primer and a second primer specific to the gene of interest. Cloning of PCR products were carried out as described above.

EST Library

EST data were extracted from the EST libraries previously performed in A. vulgare by SSH method.⁵²

Miscellaneous

Oligonucleotides were synthesized by Eurofins. Oligonucleotide sequences used in this study are available as supplemental information (Tables S2 and S3).

Funding Statement

This work has been supported by Center National de la Recherche Scientifique (CNRS) in association with the University of Strasbourg. This work has been published under the framework of the LABEX (ANR-11-LABX-0057_MITOCROSS) and benefits from a funding from the state managed by the French National Research Agency as part of the Investiments for the future program. V.D. has been supported by the région Poitou-Charentes.

Disclosure of Potential Conflicts of Interest

No potential conflicts of interest were disclosed.

Acknowledgments

Philippe Giegé, Anthony Gobert, Eric Westhof are acknowledged for critical reading of the manuscript and Myriam Badawi for informatic help. The authors wish to thank the membership of the ANR-2010-BLANC-170101 “ImmunSymbArt” for providing us the nuclear EST sequence.

Supplemental Material

Supplemental Material may be downloaded here: publisher's website