This article appeared in a journal published by Elsevier. The attached copy is furnished to the author for internal non-commercial research and education use, including for instruction at the authors institution and sharing with colleagues. Other uses, including reproduction and distribution, or selling or licensing copies, or posting to personal, institutional or third party websites are prohibited. In most cases authors are permitted to post their version of the article (e.g. in Word or Tex form) to their personal website or institutional repository. Authors requiring further information regarding Elsevier’s archiving and manuscript policies are encouraged to visit: http://www.elsevier.com/authorsrights
Author's personal copy BioSystems 113 (2013) 165–176
Contents lists available at SciVerse ScienceDirect
BioSystems journal homepage: www.elsevier.com/locate/biosystems
Pocketknife tRNA hypothesis: Anticodons in mammal mitochondrial tRNA side-arm loops translate proteins? Hervé Seligmann a,b,∗,1 a b
National Natural History Museum Collections, The Hebrew University of Jerusalem, 91904 Jerusalem, Israel Department of Life Sciences, Ben Gurion University, 84105 Beer Sheva, Israel
a r t i c l e
i n f o
Article history: Received 17 April 2013 Received in revised form 2 July 2013 Accepted 3 July 2013 Keywords: tRNA synthetase Light strand replication origin Tesserae hypothesis Parallel coding systems Overlapping genes
a b s t r a c t Peptide elongation proceeds by tRNA anticodons recognizing mRNA codons coding for the tRNA’s cognate amino acid. Putatively, tRNAs possess three anticodons because tRNA side and anticodon-arms form similar stem-loop structures. Two lines of evidence indicate that mammal mitochondrial tRNA sidearms function as anticodons: numbers of TC-arm ‘anticodons’ matching speciﬁc cognates coevolve with that cognate’s usage in mitochondrial genomes; and predicted ‘tetragene’ numbers, genes coded by quadruplet codons (tetracodons), coevolve with numbers of expanded anticodons in D-arms, as previously observed between tetragenes and antisense tRNA expanded anticodons. Sidearms with long stems and high GC contents contribute most to tRNA sidearm-tetragene coevolution. Results are compatible with two hypothetical mechanisms for translation by side-arms: crossovers exchange anticodon- and side-arms; tRNA sidearms are excised, aminoacylated and function as isolated stem-loop hairpins (more probable for long, respectively stable branches). Isolated sidearms would resemble recently described armless ‘minimal’ tRNAs. Isolated hairpins might most parsimoniously explain observed patterns. tRNA genes templating for three, rather than one functional tRNA, compress minimal genome size. Results suggest fused tRNA halves form(ed) modern tRNAs, isolated tRNA subparts occasionally translate proteins. Results conﬁrm translational activity by antisense tRNAs, whose anticodons also coevolve with codon usages. Accounting for antisense anticodons improves results for sidearm anticodons. © 2013 Elsevier Ireland Ltd. All rights reserved.
1. Introduction Import of cytosolic tRNAs into mitochondria (Tarassov et al., 1995; Tarassov and Martin, 1996; Rubio and Hopper, 2011; Rubio and Alfonzo, 2012) by a variety of mechanisms (Schneider and Maréchal-Drouart, 2000; Sieber et al., 2011) indicates that regular mitochondrion-encoded tRNAs may not be sufﬁcient for efﬁcient mitochondrial protein translation. This is also suggested by coevolution between tRNA import and codon usages (Salinas et al., 2012) and other factors from the mitochondrial translational machinery (Schneider, 2011). Computational analyses suggest that mitochondrial genomes encode for additional tRNAs, the antisense tRNAs (Seligmann, 2010b,c, 2011b), which seem to induce also expression of frameshifted overlapping protein coding genes that include stop codons (Faure
∗ Correspondance address: Unité de Recherche sur les Maladies Infectieuses et Tropicales Émergentes, Faculté de Médecine, URMITE CNRS-IRD 198 UMR 6236, Université de la Méditerranée, Marseille, France. Tel.: +33 491 324375. E-mail address: [email protected]
1 Present address: Unité de Recherche sur les Maladies Infectieuses et Tropicales Émergentes, Faculté de Médecine, URMITE CNRS-IRD 198 UMR 6236, Université de la Méditerranée, Marseille, France. 0303-2647/$ – see front matter © 2013 Elsevier Ireland Ltd. All rights reserved. http://dx.doi.org/10.1016/j.biosystems.2013.07.004
et al., 2011; Seligmann, 2011c, Seligmann, 2012b,d), including overlapping genes encoded in the 3 -to-5 direction (Seligmann, 2012e, 2013) and overlapping genes coded by codons of four nucleotides (quadruplet codons or tetracodons, Seligmann, 2012a). The possibility that additional molecules function as mitochondrial tRNAs is examined here. The hypothesis assumes that loops from tRNA sidearms function as anticodons, in addition to regular tRNA anticodons on the tRNA anticodon arm. Several observations described here below ﬁt predictions made by this functional hypothesis. Two mechanisms are proposed to accommodate the possibility that tRNA sidearms decode codons: (a) crossover between tRNA anticodon arm and sidearms preserves the classical tRNA cloverleaf secondary structure, replacing the anticodon loop with sidearm loops; and (b) short RNA stem-loop hairpins originating from tRNAs (either by tRNA degradation, RNA processing or splicing) function as ‘minimal’ tRNAs. The main point of the hypothesis is whether these putative sidearm anticodons actually read codons. At this point, there is no direct experimental information available to conﬁrm the
Author's personal copy 166
H. Seligmann / BioSystems 113 (2013) 165–176
hypothesis. Here below, computational analyses explore whether patterns derived from evolutionary comparisons of predicted sequence properties are compatible with the working hypothesis. Comparative analyses of vertebrate mitochondrial tRNAs presented below suggest that sidearm loops function as anticodons, potential mechanisms are discussed.
1.1. Sidearm loops as anticodon loops in mitochondrial tRNAs The human mitochondrial tRNA Leu CUN has an anticodon loop containing the anticodon UAG. Its sidearms have each a loop of seven nucleotides, matching the size of regular anticodon loops, where two nucleotides ﬂank on each 5 and 3 ends the actual anticodon of three nucleotides. For this speciﬁc tRNA, the correspondence between sidearms and anticodon arm goes beyond the similarity between stem-loop hairpin secondary structures formed by each arm. The D-loop contains at the location tentatively corresponding to the anticodon the trinucleotide CAG that, if anticodon function is assumed, recognizes the codon CUG that codes for Leu. Similarly, the TC-loop contains at the location corresponding to the anticodon the trinucleotide CAA that, if anticodon function is assumed, recognizes the codon UUG that also codes for Leu (Fig. 1a). This situation of tRNA Leu CUN sidearm loops with ‘anticodons’ matching Leu is well conserved in mitochondria across vertebrate species and has probably a function. It is possible that sequences resembling the anticodon in the sidearm loops contribute to tRNA recognition by its tRNA synthetase, but the structural model of tRNA–tRNA synthetase interaction does not suggest that sidearm loops interact directly with the tRNA synthetase during aminoacylation (Tsunoda et al., 2007, see therein ﬁg. 1). The situation for tRNA Leu CUN would ﬁt crossover between anticodon-and sidearms for loaded tRNAs: the anticodons in the sidearms match the same cognate, so that crossovers between any sidearm and the anticodon arm would not cause amino acid misinsertions in proteins because these ‘new’ anticodons would recognize codons coding for the usual cognate amino acid of that tRNA. Not all sidearm loops have the size of typical anticodon loops: in humans, this is observed for the D-loop only for tRNA Leu CUN and tRNA Ser UCN, and for the TC-loop, for these two tRNAs and eight others (example in Fig. 1b). In general, in vertebrates, TCloops are about three times more frequently anticodon loop-like than D-loops. Most D-loops have less than seven nucleotides (in humans, 15 among 21, excluding tRNA Ser AGY that lacks a D-arm). These short loops are a minority among TC-loops (seven in human mitochondrial tRNAs, see for example Florentz et al. (2003), therein ﬁg. 5). The rest are sidearm loops with expanded anticodons, ﬁve for each D- and TC-loops in humans. The existence of expanded sidearm loops might have a link with the occurrence of tetracodons in mitochondria. Recent analyses have suggested the existence of overlapping protein coding genes coded by quadruplet codons (tetracodons) in mitochondria and decoded by antisense tRNAs with expanded anticodon loops (anticodon loops with more than 7 nucleotides, Seligmann, 2012a,b), corroborating a hypothesis on origins of quadruplet codons from mitochondria (Gonzalez et al., 2012). Therefore, it is possible that sidearm loops with loop sizes above seven nucleotides contribute to decoding tetracodons, the building blocks of tetragenes. This yields an explicit non-trivial prediction for the working hypothesis that sidearms function as anticodons: one expects coevolution between numbers of predicted tetragenes and numbers of sidearms with expanded loops, as was observed between numbers of antisense tRNAs with expanded anticodon loops and numbers of predicted tetragenes (Seligmann, 2012a,b). Therefore, analyses of vertebrate mitochondrial sidearms
Fig. 1. Cloverleaf secondary structure of mitochondrial tRNA Leu CUN and tRNA Gln GAA (a) and (b), respectively, for Pan troglodytes, indicating putative anticodons in sidearms. One sidearm putatively bears an expanded anticodon.
presented here are restricted to the 14 mammal species compared by Seligmann (2012a) for the analysis of tetracodon contents. 1.2. tRNA recognition by tRNA synthetases and anticodons in sidearm loops The situation where the sidearm’s putative anticodon matches the regular cognate of the tRNA occurs usually for tRNA Leu CUN (12 among 13 D-loops with seven nucleotides for the 14 mammal
Author's personal copy H. Seligmann / BioSystems 113 (2013) 165–176
species considered, and all 14 species for TC-loops from that tRNA). However, this situation is relatively rare and occurs, at lesser extents, only for tRNA Ser UCN (3 among 13 D-loops, and 12 among 14 TC-loops with seven nucleotides among the 14 mammal species examined), and the TC-loop of tRNA Leu UUR (all 12 species with TC-loops of seven nucleotides). This association between cognate and putative sideloop anticodons occurs occasionally for other tRNAs: D-loops, two cases among seven for tRNA Thr; and TC-loops, six among nine species for tRNA Met, and a single (different) mammal species for each tRNAs Asp and Pro. Including Leu and Ser, sidearm anticodon and cognate are matched in 35% of the cases for D-loops, and 27% of the cases for TC-loops, considering all tRNAs with adequate loop sizes. However, after excluding from analyses Leu and Ser, this association is observed only for 8.7% of the D-arms with adequate loop sizes and 6.6% of the TC-arms, which is not statistically significantly more than would be expected by chance. Nevertheless, a tendency for an association between regular cognate and sidearm putative anticodon occurs for three D-arms (tRNA Leu CUA, Thr and Ser UCN) and four TC-arms (tRNA Leu UUR, Met, Leu CUN, and Ser UCN). This association is also compatible with the hypothesis that sidearm loops interact with tRNA synthetases and contribute to tRNA recognition by its speciﬁc tRNA synthetase, at least in these speciﬁc tRNA species. However, the lack of such an association between sidearm putative anticodons and regular cognate amino acids in other tRNAs suggests that sidearm loops do generally not participate in tRNA recognition as additional anticodons. This is also compatible with the model of tRNA–tRNA synthetase interaction where no interaction between sidearm loops and tRNA synthetase recognition sites are predicted (Tsunoda et al., 2007). 1.3. Mechanisms for sidearm loops functioning as anticodon loops The hypothesis that anticodons in the sidearm loops corresponding to the regular cognate amino acid of a tRNA participate in tRNA recognition by tRNA synthetases is unlikely, not only because this association is weak in most tRNAs and because the threedimensional model of tRNA–tRNA synthetase interaction does not suggest extensive sidearm loop involvement, but also because it seems that mitochondria of some species totally lack sidearms (Jühling et al., 2012). Hence armless tRNAs are probably recognized by their tRNA synthetases, and function in translation. This suggests that sidearms are not indispensable for these tasks. Armless tRNAs have tRNA genes with reduced length. Hence considering the high pressures for reducing metabolic costs of protein synthesis (Akashi and Gojobori, 2002; Seligmann, 2003, 2012a,b,c,d,e; Alves and Savageau, 2005; Brocchieri and Karlin, 2005; Perlstein et al., 2007; Swire, 2007; Barton et al., 2010; Heizer et al., 2011) and DNA synthesis (Session and Larson, 1987; Chipman et al., 2001; Gregory, 2002), it is unlikely that such signiﬁcant reduction of tRNAs has not occurred more frequently than observed if sidearms have no additional function(s). These could be related to the involvement of tRNA genes in initiating light strand replication (Seligmann et al., 2006a,b; Seligmann and Krishnan, 2006; Seligmann, 2008, 2010a,b,c, 2011a,b,c), but also to the working hypothesis presented here, that sidearm loops may function as anticodons. There are two putative mechanisms that are compatible with this hypothesis. The ﬁrst conserves the usual tRNA cloverleaf secondary structure, while the second follows the armless tRNA approach. 1.3.1. Crossover between tRNA arms According to the ﬁrst mechanism, tRNA arms may break and be exchanged, in a way similar to what is known for chromosomal
crossovers. This putative mechanism where sidearm loops are exchanged with the anticodon loop would imply a different cognate amino acid for the tRNA if the sidearm loop bears an anticodon that does not recognize codons coding for the cognate amino acid, as is observed in most cases. This is possible when the tRNA recognition signals on the tRNA’s acceptor stem are compatible with the ‘new’ anticodon. This is obviously the case when the new anticodon matches the amino acid of the former, original anticodon of the tRNA, and could explain the associations described above between sidearm loop anticodons and tRNA cognates, but other cases where the cognate differs are also possible. This mechanism, in this respect, is less parsimonious than the one discussed below, but it is more parsimonious in the sense that it does not assume an unusual tRNA structure. 1.3.2. Minimal tRNAs The second putative mechanism considers that if armless tRNAs occur and regularly function in translation (Jühling et al., 2012), short stem-loop hairpins could too. Indeed, RNA templated by the mitochondrial light strand replication origin forms a stem-loop hairpin with tRNA-like properties, notably post-transcriptional 3 addition of CCA and in vivo chargeability with an amino acid (Yu et al., 2008). tRNA sidearms could function this way, as isolated aminoacylated stem-loop hairpins. At least two mechanisms might produce such minimal tRNAs from regular tRNAs. First, (t)RNA processing could produce tRNA sidearms as isolated stem-loop hairpins. Mitochondrial transcription produces two long RNAs that each match the entirety of the DNA light and heavy strands, respectively (Clayton, 2000). Functional RNAs (i.e. mRNAs, tRNAs, etc.) are processed by RNA endonucleases (Calvin and Li, 2008; Fujishima et al., 2011), enzymes that recognize secondary structures within these long RNAs, speciﬁcally sequences of tRNAs, a process called ‘tRNA punctuation’ (Ojala et al., 1981). In mitochondria, this seems mainly dependent of secondary structure formation/recognition by (t)RNA endonucleases and has little more speciﬁcity. Greater splicing site speciﬁcity apparently exists for nucleus-encoded cytosolic tRNAs (Calvin and Li, 2008). Hence in mitochondria, secondary structure-dependent RNA splicing may occur not only at tRNA extremities. Such intra-tRNA splicing produces numerous tRNA-derived RNA fragments that have speciﬁc expression and functional patterns and represent a large part of small RNAs that are not microRNAs (i.e. Lee et al., 2009). If such splicing occurs at the base of sidearms, this produces short stem-loop hairpins resembling armless tRNAs (Jühling et al., 2012), or the hairpin templated by the light strand replication origin (Yu et al., 2008). The possibility that this could occur in mitochondria is indicated by the 5 processing data for tRNAs found in the online database of the mitochondrial transcriptome (Mercer et al., 2011), where numerous 5 processing events occur within tRNA sequences, not only at its ‘ofﬁcial’ 5 extremity. Similarly, the fact that such stem-loop hairpins might be functional in translation is indicated by the occurrence of 3 -additions of CCA not only at the tRNA’s 3 -extremity (Mercer et al., 2011, see online database). This hypothesis implies an unusual tRNA structure, but avoids the difﬁculty of possessing an acceptor stem incompatible with the amino acid matching the ‘new’ anticodon. In addition, tRNA degradation (by RNA endonucleases) probably de facto produces such putatively functional hairpin subunits. Pressures for metabolic efﬁciency could select for translational function by these tRNA subunits. 1.4. Minimal tRNAs and the hypothesis of tRNA origin by fusion of two halves The armless nematode mitochondrial tRNAs suggest that armless tRNAs can function in translation. Their genome locations and
Author's personal copy 168
H. Seligmann / BioSystems 113 (2013) 165–176
sequence homologies clearly show that they derived from regular tRNAs which lost their sidearms (Jühling et al., 2012). However, conserved introns in tRNA sequences, including permuted and split tRNA genes frequently occur (Randau and Söll, 2008), suggesting that modern tRNAs evolved from the fusion of tRNA parts (Fujishima et al., 2008, 2009). Therefore, split tRNA genes and tRNAs with introns might be interpreted as a derived situation (Fujishima et al., 2010). However, considering them as essentially ancestral seems most parsimonious (Di Giulio, 2008a,b, 2009, 2012). This suggests a polyphyletic origin of tRNAs, a hypothesis that again seems to explain more parsimoniously the data than a monophyletic origin (Di Giulio, 1992, 1995, 1999, 2004, 2006, 2013). Polyphyly is also compatible with the observation that the (pseudo) codon contents of the 3 half of tRNAs matches that of the universal circular code found in protein coding genes (Arqués and Michel, 1996), while the pseudocodon contents in the 5 half does not (Michel, 2012, 2013). This approach implies that the ancestral, stem-loop hairpins that later duplicated and formed by fusion modern tRNAs (as also suggested for the evolution of 5S rRNA, Di Giulio, 2010) originally functioned in protein translation. Their general structure can be assumed to resemble that of the modern nematode armless tRNAs that putatively recreated secondarily this type of translationally active small RNAs and the aminoacylated RNA hairpin templated by the light strand replication origin (Yu et al., 2008). According to this scenario, minimal tRNAs are an evolutionary convergence with ancestral ‘half’ tRNAs. These ancestral tRNA halves are putatively homologous to the Dand TC-arms of modern tRNAs. Finding evidence for translational activity by anticodons within loops of TC- and D-arms would be a strong corroboration of the polyphyletic hypothesis of tRNA origins through fusion of 5 and 3 halves, with potential experimental conﬁrmation because these stem-loop hairpins are putatively still active in protein translation of modern mitochondria. Both crossover and minimal tRNA-like tRNA halves are not mutually exclusive mechanisms. Each mechanism suggests that loops of sidearms with long stems are more likely to function in translation, because branch length increases crossover probabilities, and stem stabilities of isolated stem-loop hairpins. Hence analyses here focus on showing that loops in sidearms function in translation, rather than detecting by which speciﬁc mechanism this occurs. Having this in mind, the genomic property that tRNA anticodon populations match the genomic usage of their corresponding codons, codon usage optimization (Ikemura, 1981,1982; Gouy and Gautier, 1982; Yamao et al., 1991; Percudani et al., 1997), is explored in this context, for mitochondrial genomes. This property where numbers of tRNA genes with speciﬁc anticodons and codon usages are correlated makes protein translation more efﬁcient, so that tRNA abundance and codon usage covary at different growth rates in E. coli (Dong et al., 1996) and for different tissues in humans (Plotkin et al., 2004; Dittmar et al., 2006). Mitochondrial genomes possess typically only 22 tRNAs, bearing one anticodon for each of the amino acid-codon families in the mitochondrial genetic code. Hence in this system, assuming classical tRNA function, no association between anticodon and codon usages exists. However, if one assumes that sidearm loops also function as anticodons, different codon families are matched by different numbers of anticodons. If coevolution between codon usages and anticodons from sidearms is detected, resembling what is known for cytosolic pools of anticodons and codons, this optimization would be evidence that loops in sidearms function in translation. If analyses verify this prediction, the hypothesis of sidearm anticodon activity should not be seen as baseless. The same rationale also holds for predicted mitochondrial tetragenes, which putatively would be translated by expanded
anticodons from tRNA sidearms. Previous analyses showed coevolution between predicted tetracoding and numbers of antisense tRNAs with expanded anticodons, indicating that tetracoding occurs: the predicted functional requirement for translating tetracodons is matched by numbers of antisense tRNAs with expanded anticodons (Seligmann, 2012a). One expects that these numbers of previously predicted tetracoding genes coevolve also with numbers of tRNA sidearms with sidearm loops with lengths greater than seven nucleotides. The evolutionary experiment presented here bears two main predictions, for regular anticodons and for expanded anticodons. These are independent and non-trivial. Hence it would be highly unlikely that analyses verify both predictions as a matter of chance, precluding alternative processes for explaining patterns described here. In that case, assuming anticodon functions by sidearm loops would be the most parsimonious explanation for both observed phenomena, coevolution of sidearm loops with codon usage and with tetragenes. Further predictions are compatible with crossover and tRNA subunit hypotheses. Including information on the length and GC contents of sidearm’s stems should generally improve the strength of the predicted coevolutionary associations if the working hypothesis is correct: crossover is more likely for longer sidearms; and longer stems with higher numbers of GC base pairs are more stable, especially if functioning as isolated hairpins. Analyses are performed on a predeﬁned set of 14 mammal mitochondrial genomes, for which predictions on tetragenes and antisense tRNAs have been previously described (Seligmann, 2012a).
2. Results and discussion 2.1. Coevolution between codon usage and sidearm anticodon within a genome Examination of sidearm loops from human mitochondrial tRNAs (see ﬁg. 5 in Florentz et al., 2003) detects 12 sidearm loops with 7 nucleotides which can putatively be considered as potential regular anticodons. Five among them (41.7%) would recognize codons coding for Leu. The second most frequently represented anticodon group matches codons coding for Ser. These are also the two most frequent amino acids in human mitochondrion-encoded genes. Hence according to the working hypothesis, one might expect that frequencies of codons coding for given amino acids correlate positively with numbers of tRNA sidearms with matching anticodons. Fig. 2 presents this association, for the mean codon usage, averaged across all 14 mammal genomes used by Seligmann (2012a) and the mean number of anticodons in loops of D-sidearms (parametric Pearson correlation coefﬁcient r = 0.596, P = 0.0017; non-parametric Spearman rank correlation coefﬁcient rs = 0.549, P = 0.0059, one tailed tests). Correlations were qualitatively similar, though weaker for numbers of anticodons in loops of TC sidearms (r = 0.184, P = 0.21; rs = 0.439, P = 0.0217, one tailed tests). Averaging across D- and TC sidearms, the correlation between codon usage and anticodon numbers is also positive as expected (r = 0.323, P = 0.07; rs = 0.56, P = 0.005, one tailed tests). Because tests on D and TC sidearms are independent, an overall, combined statistical signiﬁcance to this test can be calculated, following Fisher’s method to combine P values from different independent tests of the same hypothesis. Its statistic sums the −2 × ln(Pi) of all ps included, where i ranges from 1 to k tests (two tests in this case). This statistic follows a chisquare distribution with 2 × k degrees of freedom, and yields P = 0.0032 and P = 0.0013 for parametric and the more robust non-parametric tests, respectively. Hence this ﬁrst test conﬁrms the working hypothesis that
Author's personal copy H. Seligmann / BioSystems 113 (2013) 165–176
Fig. 2. Numbers of putative anticodons in loops of D-sidearms of all 22 tRNAs in a mitochondrial genome, averaged across 14 mammal species, as a function of the codon usage across 13 mitochondrial protein coding genes averaged across the same 14 mammal species.
putative anticodons in tRNA sidearms coevolve with usage of codons matched by these putative sidearm anticodons, suggesting that sidearm loops may function in translation as do anticodon loops. 2.2. Coevolution between codon usage and sidearm anticodons: between genome analysis The analyses described in Section 2.1 show that when comparing different codon families, those coding for frequent amino acids correspond to frequent putative sideloop anticodons minimization principles (see Seligmann, 2003), the evidence in Fig. 2 cannot be considered as very strong, though it is clearly supportive of the working hypothesis. However, numbers of putative anticodons in sidearms, for a given codon–anticodon family, also vary among mammal species. Hence coevolution between codon usage and sidearm anticodon numbers can also be tested at the level of single codon families. As compared to analyses in the previous section, such comparisons, for a given codon family, across different mammal species, represents testing the hypothesis at a microevolutionary scale. This scale is much less distant than that represented by variation among different types of amino acids, and hence is a stronger test of the hypothesis. Because there are 22 codon–anticodon families, the same principle of coevolution between microevolutionary changes in codon usages between mammal species and numbers of sidearm anticodons can be independently tested 22 times, for each D- and TC-sidearms. 2.3. No coevolution with codon usage for anticodons in loops of D-sidearms The 14 mammal species vary in numbers of loop D-sidearm anticodons for 16 among 22 codon–anticodon families. Hence 16 correlations could be calculated to test whether evolutionary changes in codon usage associate with evolutionary changes in numbers of matching anticodons in the loops of D-sidearms. Coevolution was positive for eight among 16 cases, which does not indicate any general tendency. In this sample of 14 species, these correlations, whether positive, as expected by the working hypothesis, or negative, are never statistically signiﬁcant at P < 0.05.
Fig. 3. Numbers of anticodons in loops of TC-sidearms of mitochondrial tRNAs of 14 mammal species as a function of the usage frequency (1000×) of the corresponding codon family in the 13 mitochondrion-encoded protein coding genes of the same 14 mammal species, for codon–anticodon families Leu UUR, ﬁlled symbols; Met, circles; Pro, triangles (continuous, discontinuous and dashed lines, respectively). Within each codon–anticodon analysis, each datapoint is for one of the 14 mammal species.
2.4. Coevolution with codon usage for anticodons in loops of T C-sidearms For the TC-loop, the data enable doing 16 correlation tests for different codon–anticodon families. Thirteen among sixteen are positive (81.25%), which is a statistically signiﬁcant majority of cases according to a one tailed sign test (P = 0.0053). Hence overall, the predicted positive association between codon usage and sidearm loop numbers is observed for TC-loops. There are 4 speciﬁc codon–anticodon families for which the correlation is statistically signiﬁcant, for Leu CUN, UUR, Met and Pro (r = 0.465, P = 0.047; r = 0.526, P = 0.027; r = 0.668, P = 0.0045 (see Fig. 3); r = 0.502, P = 0.034, respectively, one tailed tests). None of these correlations remains statistically signiﬁcant at P < 0.05 after applying the overconservative Bonferroni correction for the multiplicity of tests (Perneger, 1998), but applying the more realistic adjustment of Bonferroni’s correction, the BenjaminiHochberg method to control for false discovery rates (Benjamini and Hochberg, 1995) shows that only the correlation for Leu CUN is not signiﬁcant at P < 0.05 after applying this method (P = 0.058 after adjustment). This means that 3 or 4 tests are statistically signiﬁcant at P < 0.05 among 16 tests. At P < 0.05 and performing 16 tests, one expects on average 0.05 × 16 = 0.8 false positive results. The number of signiﬁcant correlations is 3.5–5 times greater than the expected number of false positives, so that one can assume that in TC-loops, positive microevolution between codon and anticodon is not a statistical artefact. Because the various tests are independent, one can also apply here Fisher’s test for combining P values, where k is the number of tests performed (16 in this case). This yields a chisquare statistic of 60.07 with 2 × k degrees of freedom, which yields P = 0.0019. Hence at all levels of analysis, overall, codons and matching anticodons in loops of TC-sidearms coevolve positively. It is interesting to note that analyses of macro-evolutionary variation between codon–anticodon families yield clear positive results for D-arms (Section 2.2), but weaker ones for TC-arms. The situation is opposite for microevolutionary levels, where variation among genomes, for the same codon–anticodon family is analyzed (Sections 2.3 and 2.4). In this case, there is no pattern for D-arm loops,
Author's personal copy 170
H. Seligmann / BioSystems 113 (2013) 165–176
Fig. 4. Numbers of predicted tetracoding genes (genes coded by codons of four nucleotides, tetracodons) as a function of the numbers of expanded anticodons in loops of D-sidearms of mitochondrial tRNAs of 14 mammal species.
but results are very clear and support the working hypothesis for TC-arm loops. At each macro- and microevolutionary levels, results are overall conﬁrming the working hypothesis, though in each case for a different sidearm. 2.5. Tetragenes and sidearm loops with expanded anticodons A previous analysis of the 14 mammal mitochondrial genomes examined here detected a number of regions, within the 13 regular mitochondrial protein coding genes, that code for overlapping genes formed by tetracodons, codons of four nucleotides. This rather surprising result was conﬁrmed by a number of independent properties (stop codon densities, GC contents, nucleotide ratios at third codon position (of the regular main frame) and codon usages (of the main frame and of the tetracoding frame)), and particularly by the observation that numbers of tetragenes coevolve with numbers of antisense tRNAs with predicted expanded anticodons (anticodons in loops with more than seven nucleotides). Presumably, tetracodons are frequently decoded by tetraanticodons (Walker and Frederick, 2008), but this process does not necessarily imply expanded anticodons. It can occur by complex ‘tRNA hopping’ and frameshifting (O’Connor et al., 1989, Barak et al., 1996). Hence one expects fewer expanded anticodons than regular anticodons. Nevertheless, the observed positive coevolution between antisense tRNAs with expanded anticodons and tetragenes is a functional test of the hypothesis that tetragenes are expressed. Numbers of antisense tRNAs with expanded anticodons are on average 2.57 ± 1.28 per mitochondrial genome (ranging from 1 to 5). Putatively, these might be completed by expanded anticodons in sidearms of regular sense tRNAs. Hence one expects also a positive coevolution between numbers of tetragenes per mammal genome and numbers of expanded anticodons in sidearms. Fig. 4 shows the expected positive association between numbers of tetragenes per mitochondrial genome and numbers of expanded loops in D-sidearms (r = 0.43, P = 0.056, one tailed test). This analysis does not take into account the contribution of expanded anticodons from antisense tRNAs, a possible cause for the weak correlation in Fig. 4. Indeed, summing for each genome the number of expanded anticodons in D-sidearms and antisense tRNAs yields a much higher correlation (r = 0.63, P = 0.0079, one tailed test).
This effect is not solely due to the known effect of expanded anticodons in antisense tRNAs. Partial correlation analyses between numbers of tetragenes and of expanded anticodons in loops of D-sidearms, that control for numbers of expanded anticodons from antisense tRNAs, increase the correlation between D-sidearm expanded anticodons and tetragenes, as compared to the simple analysis solely based on D-sidearms and yield r = 0.511 (one tailed P = 0.037, compare with correlation reported at the beginning of this paragraph). Hence expanded anticodons in loops of D-sidearms seem to contribute to decoding tetragenes, tentatively conﬁrming the working hypothesis. Similar analyses for expanded anticodons in loops of TCsidearms yield negative associations with numbers of tetragenes. This is not expected, but the fact that the correlation is relatively strong (partial correlation analysis controlling for expanded anticodons from antisense tRNAs yields r = −0.64, one tailed P = 0.0089) suggests that these also contribute, in an unexpected manner, to tetracodon decoding. Results from this section, combined with those from previous sections, suggest that loops in D-sidearms decode tetragenes, but not regular genes, while loops in TC-sidearms decode regular codons, but not tetragenes. The coevolution between anticodons in D-, but not in TC-sidearms with amino acid usages when comparing different amino acids (Fig. 2), indicates macroevolutionary adaptation for anticodons in D- but not TC-loops. The coevolution presented in this section between tetragenes and anticodons in D- (but not TC-) loops, might be in line with the rather long term adaptive variation of anticodons in D-loops, as opposed to short term adaptations for TC loops. 2.6. Contribution of sidearm lengths to sidearm anticodon function There are two plausible mechanisms by which anticodons in loops of tRNA sidearms can function in translation. A mechanism exchanging the sidearm with the anticodon arm, similar to chromosomal crossing overs, would not alter the general structure of the tRNA. This would in many cases yield an amino acid misinsertion at the level of protein synthesis, if the crossover occurred after the tRNA’s aminoacylation with the cognate matching its normal anticodon. It is not clear whether tRNA loading after crossovers would aminoacylate the amino acid matching the ‘new’ anticodon, because potentially the tRNA synthetase would recognize the ‘new’ anticodon, which could result in adequate aminoacylation of an amino acid coded by codons recognized by the ‘new’ anticodon. This is because tRNA loading does not only depend on anticodon recognition, but involves other signals on the tRNA (i.e. Pütz et al., 1991; Rudinger et al., 1992). Chromosomal crossovers between branches are proportional to the length of the branches. Hence according to the crossover scenario, sidearms with longer stems are more probable to be involved in crossovers with anticodons than shorter ones. The second mechanism is that of excised stem-loop hairpins corresponding to the isolated sidearm. Here too, one can assume that sidearm stem length increases probabilities that this hairpin might function in translation. Therefore, anticodons from sidearms as used in analyses from the previous sections might not be identical: the longer ones are more likely to contribute to codon reading than the shorter ones, independently of the mechanism assumed. Hence analyses from Sections 2.2 and 2.5 are repeated, after weighting each sidearm by its length, according to the following rationale. Anticodon stems have typically ﬁve base pairs. Hence when summing numbers of sidearms with a given type of anticodons, numbers of base pairs in their stem are divided by 5, and these stem length-weighted counts are summed, so that longer arms get greater weight, and shorter ones lower weight. If
Author's personal copy H. Seligmann / BioSystems 113 (2013) 165–176
Fig. 5. Numbers of predicted tetracoding genes (genes coded by codons of four nucleotides, tetracodons) as a function of the numbers of expanded anticodons in loops of D-sidearms of mitochondrial tRNAs of 14 mammal species. Data for hollow symbols are weighted according to stem length of the D-sidearm (circles in ﬁgure). Weighting is ‘1 for stems of ﬁve base pairs, 0.8 for stems of three base pairs, 0.6 for those of two base pairs and 1.2 for those of six base pairs. Data for ﬁlled symbols are weighted according to sidearm stem GC basepair contents.
sidearm length contributes to translational activity by anticodons in sidearm loops, the correlations presented in previous sections should generally increase. For anticodons of regular size in D-arms, this altered only four correlations. Two correlations are slightly increased, and two slightly decreased. Numbers of positive correlations increased from 8 among 14 (57.1%, one tailed P = 0.20, sign test) to 9 among 14 (64.3%, one tailed P = 0.11, sign test). Hence for D-sidearms, there is no noticeable effect for loops of seven nucleotides. The situation is similar for TC-arms, where 12 among 15 (80%, one tailed P = 0.009, sign test) correlations are positive, with and without weighting by sidearm length. Correlations between codon usages and numbers of regular anticodons weighted according to sidearm lengths are also done for the sum of anticodons in D- and TC-sidearms. Weighting increases correlations in approximately half the cases. The total number of positive correlations increases from 12 among 17 (70.6%, one tailed P = 0.036, sign test) to 14 among 17 (82.4%, one tailed P = 0.0032, sign test). Hence analyses suggest a weak positive effect of the length of the sidearm on translational activity by the putative anticodon in its loop. These analyses with numbers of anticodons weighted according to stem lengths are also done for the association between tetragene numbers and numbers of expanded anticodons in D-sidearms. Fig. 5 shows that the correlation between expanded anticodons in loops of D-arms and tetragenes increases after weighting according to stem length (r = 0.507, one tailed P = 0.032). Hence overall, analyses suggest that weighting anticodons in sidearms by their stem’s length does marginally increase their contribution to translation, whether of regular tri-, or tetracodons. These results are not incompatible with the functional hypotheses proposed. 2.7. Contribution of sidearm stability to sidearm anticodon function
compatible with both proposed mechanisms, crossovers between sidearms and the anticodon arm, and activity of isolated stem-loop hairpins, formed by excised tRNA sidearms, during RNA processing and/or degradation. Stem stability (estimated by the proportions of GC base pairs in the stem) might yield more information: translational activity by isolated stem-loop hairpins might be favoured by structural stability. It is also possible that crossovers are more probable for less stable stems, but stability might actually increase branch breakability, and crossover probabilities. Hence such tests including stem stabilities could, but would not necessarily enable to indicate the more likely mechanism among crossover and isolated sidearms. Analyses weighting anticodons according to proportions of GC basepairs in sidearm stems, similar to those weighting anticodons by sidearm stem lengths in the previous section, are done for associations between codon usages and numbers of sidearm anticodons, as well as for numbers of predicted tetragenes and sidearms with expanded anticodons. For coevolution between codon usage and numbers of D-sidearm anticodons, weighting by sidearm stem GC contents yields 17 correlations, which are positive for 9 anticodon families (52.9%, one tailed P = 0.25, sign test, no single correlation is signiﬁcant at P < 0.05). There is no tendency to increase correlations as compared to non-weighted analyses or those weighting anticodons according to stem lengths. For TC-sidearms, weighting by sidearm stem GC-contents yields 17 correlations, which are positive for 15 anticodon families (88.2%, one tailed P = 0.0006, sign test). Two single correlations are statistically signiﬁcant at P < 0.05: Leu UAA and Met CAU (r = 0.728, P = 0.0016; and r = 0.493, P = 0.037, respectively, one tailed tests). Results for TC-sidearms suggest that weighting by stem GC contents slightly improves coevolution between TCsidearm anticodons and codon usages, as compared to weighting by sidearm stem length. Analyses with numbers of anticodons weighted according to stem GC contents are also done for the association between tetragene numbers and expanded anticodons in D-sidearms. The correlation between expanded anticodons in loops of D-arms and tetragenes increases after GC-weighting (r = 0.589, one tailed P = 0.013, as compared to r = 0.507 when weighting by stem length and r = 0.443 without weighting). Hence overall, analyses suggest that weighting anticodons in sidearms by stem GC contents marginally contributes to predict their contribution to translation, whether of regular tri-, or tetracodons, and this also as compared to weighting by stem lengths. These results are not incompatible with the functional hypotheses proposed. It is not clear how crossovers are promoted by stem GC contents, but this cannot be ruled out, as stem ﬂexibility or stability may affect crossover. However, results are clearly in line with the hypothesis that isolated stem-loop hairpins consisting of excised tRNA sidearms function in translation, as longer stems with higher GC contents are more stable and would function longer in translation. The result that analyses weighting according to GC contents increase the strength of the association between expanded sidearm anticodons and tetragenes also ﬁts the hypothesis that tetracoding is an adaptation to high temperatures. At high temperatures, the increase in stabilities of codon:anticodon duplexes for quadruplet, rather than triplet, codons can be crucial for effective translation. Hence at high temperatures, high GC contents in sidearm stems is also particularly important to ensure proper secondary stem-loop structure despite high temperatures. 2.8. Translation by anticodons from antisense tRNAs
The main problem to elucidate, probably by direct experiments, is the mechanism(s) by which anticodons in loops of tRNA sidearms function in translation. Analyses suggest that sidearm length slightly contributes to this phenomenon, which is
The working hypothesis examined here assumes that protein translation occurs not solely by anticodons found in the anticodon loop of sense tRNAs, but also in loops of sense tRNA sidearms.
Author's personal copy 172
H. Seligmann / BioSystems 113 (2013) 165–176
Some computational evidence for a similar hypothesis, where anticodons in the predicted anticodon loop of antisense tRNAs function in translation, exists: anticodons of antisense tRNAs usually avoid matching stop codons (Seligmann, 2010b); their cloverleaf formation capacities are proportional to usages of matching codons (Seligmann, 2010c, Seligmann, 2012d) and contribute to explain pathogenic mutations in mitochondrial tRNAs (Seligmann, 2011b); antisense tRNAs with anticodons matching stop codons coevolve with frameshift-overlapping genes that require anticodons recognizing stop codons for translation (Faure et al., 2011; Seligmann, 2011c, 2012b,d). Therefore, the contribution of anticodons in loops of tRNA sidearms to correlations between anticodon numbers and codon usages, which reﬂect optimization of anticodons and codons for efﬁcient translation, may be confounded by anticodons of antisense tRNAs. This is also observed for associations between tetragenes and anticodons with expanded anticodons. Taking into account expanded anticodons from antisense tRNAs, analyses in a previous section show that the association between tetragenes and expanded anticodons in loops of D-arms becomes more detectable. Therefore, coevolution between numbers of anticodons in antisense tRNAs with usages of corresponding codons are explored here, to verify that these numbers of antisense anticodons coevolve with codon usages, for the 14 mammal mitochondrial genomes used here. This is an independent test of the hypothesis that antisense tRNAs function in translation, which has not been presented previously. This number of anticodons in antisense tRNAs varies among the 14 mammal species considered for 17 codon–anticodon families. Positive coevolution between predicted antisense tRNA anticodon numbers and codon usages occurs for a statistically nonsigniﬁcant majority, 10 among 17 cases (58.8%, one tailed sign test P = 0.157). At the level of speciﬁc codon–anticodon families, correlations are signiﬁcantly positive at P < 0.05 in four cases, Asn, Cys, Ile and Trp (r = 0.5497, P = 0.021; r = 0.501, P = 0.034; r = 0.743, P = 0.0012; and r = 0.56, P = 0.0187). Only the correlation for Ile remains signiﬁcant at P < 0.05 according to Bonferroni’s overconservative correction for multiple testing, but the more realistic Benjamini-Hochberg adjustment for false positives conﬁrms all four correlations as signiﬁcant. Applying Fisher’s method for combining multiple tests, the chisquare statistic is 53.5, which with 34 degrees of freedom yields P = 0.0179. Hence overall, analyses conﬁrm for anticodons of mammal mitochondrial antisense tRNAs codon–anticodon optimization for efﬁcient translation. This simple method is an additional, independent computational conﬁrmation of the hypothesis that antisense tRNAs function in translation. Note that not all bioinformatics analyses conﬁrm the existence of antisense tRNAs (Jühling et al., 2011a,b) and that direct experimental evidence is still lacking (Brzezniak et al., 2011a,b). In that context, the correlations described here (see Fig. 6 for Trp), are important evidence suggesting translational activity by antisense tRNAs.
Fig. 6. Proportion of codons coding for Trp in the 13 protein coding genes from 14 mammal species as a function of the numbers of anticodons in antisense tRNAs recognizing codons coding for Trp from the same mammal mitochondrial genomes.
observed between TC-arm loop anticodons and codon usages for codon-anticodon families for which there are on average few anticodons in antisense tRNAs, and the opposite for those for which anticodons are frequently present in antisense tRNAs (Fig. 7). The pattern in Fig. 7 suggests that a lack of detection of coevolution between anticodons in loops of TC-arms and codon usage is to a large part due to confounding effects of numbers of anticodons in antisense tRNAs. Hence coevolution for TC-arms is most apparent when antisense tRNAs do not contribute to reading that codon family. This suggests that in reality, the coevolution between codon usages and sidearm anticodons is probably greater than detected here, an interesting conﬁrmation of the working hypothesis that sidearm loops function as anticodon loops in translation. Note that correlation coefﬁcients estimating coevolution between TC-loop anticodons and codon usages are for analyses weighing anticodons according to numbers of base pairs in the TC-arm’s stem. It is interesting to remember that coevolution between codon usages and anticodon numbers is stronger
2.9. Anticodons in sense tRNA sidearms and anticodons of antisense tRNAs Above results in Section 2.5 show coevolution between anticodons in antisense tRNAs and codon usages in protein coding genes. These associations may interfere with the detection of coevolution between the same codon usages and numbers of anticodons in loops of tRNA sidearms. The latter is the main phenomenon examined here. It is possible that the detection of coevolution for sidearms is impeded by effects of anticodons from antisense tRNAs, and that this factor explains the wide variation found for directions and strengths of associations between codon usages and numbers of anticodons in sidearm loops. Indeed, strong positive coevolution is
Fig. 7. Correlation coefﬁcient estimating coevolution between numbers of anticodons in loops of TC-sidearms of mitochondrial tRNAs of 14 mammal species, weighted according to stem length of the TC-sidearm with codon usages in these mammal species, as a function of the mean number of predicted anticodons in antisense tRNAs.
Author's personal copy H. Seligmann / BioSystems 113 (2013) 165–176
for anticodons in loops of TC-arms than of D-arms (9 among 12 comparable cases, 75%, P = 0.037, one tailed sign test). Overall, coevolution for anticodons in antisense tRNAs is also stronger than for anticodons in loops of D-arms (9 among 12 comparable cases, 75%, P = 0.037, one tailed sign test), but anticodons in antisense tRNAs do not seem to coevolve more with codon usages than anticodons in loops of TC-arms (correlations more positive in 7 among 14 comparable cases). Hence translational activity by antisense tRNAs and sense tRNA sidearms might have overall comparable extents. 3. General discussion 3.1. Translation by sidearm anticodons and antisense tRNAs The hypotheses that efﬁcient mitochondrial translation requires additional tRNAs, and that some of these are to be found in the complementary strand of tRNA genes, the antisense tRNAs, are conﬁrmed by analyses that show coevolution between mitochondrial codon usages and numbers of antisense tRNAs with anticodons matching that codon family (Section 2.5, example in Fig. 6). However, it seems that additional anticodons of mitochondrial origins function in mitochondrial translation, anticodons in the loops of tRNA sidearms. Comparative analyses show that, when comparing codon usages for different codon-amino acid families, codon usages associate positively with numbers of anticodons in the loops of D-sidearms of regular sense mitochondrial tRNAs. Variation between different codon families reﬂects a very ancient process, hence this adaptation seems very ancient. Similar analyses do not detect macroevolutionary coevolution for anticodons in the loops of TC-sidearms of regular sense tRNAs. For the latter, analyses reveal microevolutionary adaptation: variation in codon usages between species, for the same codon family, coevolves with numbers of matching anticodons in loops of TC-sidearms of the mitochondrial genomes from the same species. Hence each tRNA sidearm, at least in mammal mitochondria, reacts adaptively to different pressures and at different time scales. Microevolutionary coevolution between codon usages and anticodons in loops of TC-sidearms is also modulated by numbers of anticodons in antisense tRNAs matching that codon family (Fig. 7). The latter phenomenon, of coevolution between antisense anticodons and codon usages, seems slightly stronger than for anticodons from tRNA sidearms, though this difference is far from statistical signiﬁcance. It is important to stress here that it is the difference in the strengths of these correlations that is not statistically signiﬁcant, but that each of these analyses is statistically signiﬁcant. Hence results support each phenomenon, translation by anticodons from sense tRNA sidearms, and from antisense tRNAs, but cannot indicate with conﬁdence which of the two phenomena is preponderant on the other. Analyses of coevolution between numbers of predicted tetragenes, genes coded by tetracodons, with expanded anticodons conﬁrm the principle that antisense tRNA activity is more frequent than translation by anticodons in sense tRNA sidearms: coevolution with expanded anticodons from antisense tRNAs seems slightly stronger than coevolution with expanded anticodons from loops of D-arms. 3.2. Statistical support of the results The main incoherence in this body of data is that expanded anticodons in loops of TC sidearms coevolve negatively with tetragenes. The non-randomness of this pattern indicates that an actual phenomenon occurs, but this phenomenon is not yet
understood in the frame of the working hypothesis presented here. Nevertheless, the coherence of other results, which conﬁrm each other, indicates that indeed, loops of tRNA sidearms function as anticodons in translation, including of tetragenes, despite a lack of direct experimental evidence. In this context, one should remember that three independent tests were done for each tRNA sidearm, one at macroevolutionary level (correlation between codon usages and sidearm anticodon numbers when comparing codon families and two at microevolutionary level; coevolution between codon usages and sidearm anticodon numbers for a given codon family; and coevolution between expanded sidearm loops and predicted tetragenes). Hence this enables to calculate for each sidearm the statistical support of the overall hypothesis, combining all three tests. For the ﬁrst test, macroevolutionary analyses, statistical significances are P = 0.0017 (Fig. 1) and P = 0.21 for D- and TC loops, respectively. For the two microevolutionary analyses, results of analyses weighting by sidearm stem GC contents are considered. The second test is coevolution of codon usages with sidearm anticodon numbers (considering separately each codon family). Statistical signiﬁcances of the distribution of numbers of positive correlations, as calculated by the conservative sign test, are used here P = 0.25 and P = 0.0006 for D- and TC loops, respectively. The third test is for coevolution between tetragenes and sidearm loops with expanded anticodons, for which P = 0.013 and 0.9955 for D- and TC loops, respectively. Note that P = 0.99555 is the one sided P value when expecting positive coevolution between expanded sidearm anticodons and tetragenes. The two sided P value of the observed negative correlation (P = 0.0089) is 1-P(two sided)/2 yields P = 0.99555. This means that the strong negative correlation observed for TC loops does not support the working hypothesis (P close to 1). The combined Ps for each D- and TC loops yields P = 0.0005 and P = 0.006, respectively. This high support, despite that for each sidearm, speciﬁc results are not statistically signiﬁcant, is no artefact. It results from the combination of different independent tests, which is a powerful experimental setup, based on the principle of repeatability. Hence overall support for the working hypothesis is strong separately for each D- and TC loops. Because the tests for D- and TC loops are independent, one can also combine results from both sidearms, which yields P = 0.00003. 3.3. Crossover versus isolated stem-loop hairpins? It is difﬁcult to indicate which of the proposed mechanisms, crossovers between anticodon and sidearms, and/or isolated stemloop hairpins excised from tRNAs and consisting of the tRNA’s sidearm, is the mechanism accounting best for the phenomena of coevolution between the needs of translation (codon usages, tetracodons) and anticodon properties of tRNA sidearms. The hypothesis of translational activity by isolated hairpins seems nevertheless, overall, most parsimonious, as it is implied by the hypothesis (and evidence) for a polyphyletic origin of tRNAs from two halves. It ﬁts rare observations of aminoacylated stemloop RNA hairpins corresponding to the light strand replication origin (Yu et al., 2008) and the probable translational activity by armless minimal tRNAs (Jühling et al., 2012). In addition, this mechanism does not imply amino acid misinsertions in proteins by tRNAs in which arm crossover occurred after tRNA loading by the cognate matching the regular anticodon. This problem is avoided in a small minority of cases where sidearm loops bear anticodons matching the cognate that matches also the regular anticodon. However, crossovers do not imply an unusual tRNA three-dimensional structure, provided that anticodon loops are able to interact by complementarity with the sidearm that
Author's personal copy 174
H. Seligmann / BioSystems 113 (2013) 165–176
Table 1 Stabilities of interactions between loops for sense and antisense tRNA. Columns are: 1. tRNA cognate and anticodon; 2. G between 3 and sense 5 sidearm loops for sense and 3. antisense tRNA; 4. G between sense anticodon loop and 5 sidearm loop, assuming the 3 sidearm functions as anticodon arm; 5. G between sense anticodon loop and sense 3 sidearm loop, assuming the 5 sidearm functions as anticodon arm. tRNA
Sense 3 –5
Antisense 3 –5
Ala UGC Arg UCG Asn GUU Asp GUC Cys GCA Gln UUG Glu UUC Gly UCC His GUG Ile GAU Leu UAA Leu UCN Lys UUU Met CAU Phe GAA Pro UGG Ser GCU Ser UGA Thr UGU Trp UCA Tyr GUA Val UAC
−1 1.7 −0.8 0.8 1.6 0.6 1.4 2.1 0.2 0.6 0.4 −1.6
−1.5 2.1 0 0.8 0.1 −1.5 0.8 2.2 −3.1 −0.2 1.5 −1 1.5 0.4 −1.5 −0.5 0.2 −2.9 0.5 0.1 −0.5 0.8
−0.5 1.5 −0.7 0.7 1.9 1.9 0.1 −0.1 −2.4 0.9 0.8 1.5 0.6 2.3 1.4 1 1.9
−1.7 −0.4 −1.1 0.3 0.1 −0.1 −0.2 −0.4 −0.5 −0.2 −0.2 −0.2 −1.
−0.9 −0.7 −0.2
1.2 −1 1 −1.2 −0.2 2.1 0.7 1.9
remained at its usual location after the crossover (this point is tested below). It seems probable that the reality is a mixture of mechanisms, where for some cases, anticodon activity by sidearm loops results from crossovers, but for most cases, anticodon activity by sidearm loops would be due to isolated hairpins produced by sidearm excision. It is also possible that anticodons in sidearm loops contribute to efﬁcient tRNA recognition by tRNA synthetases. They might ‘lure’ inadequate tRNA synthetases away from the anticodon, preventing tRNA loading, when sidearm anticodons do not match the regular anticodon’s cognate amino acid, or increase the probability of interactions with the correct tRNA synthetase, for tRNAs where the sidearm anticodon(s) match the cognate of the regular anticodon, such as observed for tRNAs Leu. This phenomenon is incompatible with the other phenomena suggested here. 3.4. Sidearm loop–loop tertiary structure interactions for sense and antisense tRNAs A testable point noted above is the matter of the threedimensional structure formed by tRNAs after crossovers between the anticodon arm and one of the sidearms. A major component of this three-dimensional structure is a bending of the sidearms towards each other, forming an arch, and where nucleosides in each arm’s loops interact, along regular complementarity rules, with each other (Florentz et al., 2003). This tertiary structure element is by no means a sufﬁcient condition for forming the actual L-shaped tRNA three-dimensional structure that functions in translation, but it is an important element of it. Table 1 presents the Gs of the hybridization between the different loops of human mitochondrial tRNAs, as predicted by Mfold’s two state melting (hybridization) prediction tool (Markham and Zuker, 2005, 2008; Dimitrov and Zuker, 2004). This procedure predicts at least some interactions between the regular loops of the D- and TC-sidearms for 18 among 22 sense tRNAs, and 17 among 22 antisense tRNAs. It is notable that loop–loop interactions are stronger for the antisense than the sense tRNA for 16 among 21 cases where a comparison was possible.
Hence, because regular sense tRNAs frequently have less stable interactions between sidearm loops than antisense tRNAs, it is clear that at least this criterion is met by antisense tRNAs. 3.5. Sidearm loop–loop tertiary structure interactions assuming crossovers in sense tRNAs The crossover hypothesis assumes, for the TC-arm functioning as anticodon, that crossover occurred between anticodon- and TC-arms, with the TC-arm functioning as anticodon arm. In this case, the bending interaction would occur between loops of the anticodon- and the D-arm. These interactions are more stable than those found for the regular sense tRNA (without crossover) for 9 among 20 cases. The same analysis assuming crossover between anticodon and D-arm, with the D-arm functioning as anticodon arm, the bending interaction would occur between loops of the anticodon- and the TC-arm. These interactions are more stable than those found for the regular sense tRNA (without crossover) for 17 among 22 cases, a statistically signiﬁcant majority according to a two tailed sign test (P = 0.00845). These results indicate that after crossovers between sidearms and the anticodon arms, most or all tRNAs could form the regular L-shaped three-dimensional tRNA structure, at least from the point of view of the capacity of loops to form the tertiary interaction necessary for the tRNA’s three-dimensional L-shape. These results suggest the possibility of a further mechanism by which anticodons in loops of sidearms might function in translation. The tRNA’s molecular ﬂexibility might enable to form L-shaped three-dimensional structures without crossover, where the anticodon loop, from its usual location as anticodon arm, interacts with the loop of one of the sidearms. It is possible that such interactions and associated bending are not incompatible with the L-shape structure, and that such tRNAs bent according to unusual rules could be aminoacylated and active in translation. This a priori unlikely possibility should also be considered as a viable mechanism, as long as one cannot prove that such interactions between anticodon and sidearm loops cannot form the L-shaped threedimensional tRNA structure. 3.6. Tetracodons and the origins of the mitochondrial genetic code Results present coevolution between expanded anticodons in mitochondrial sidearms and predicted tetragenes which conﬁrm earlier observations of coevolution between these tetragenes and antisense tRNAs with expanded anticodons (Seligmann, 2012a). These results are in line with the suggestion that the mitochondrial genetic code’s ancestor was based on tetracodons (Gonzalez et al., 2012). That hypothesis proposed that only a subset of the 264 potential tetracodons were used in the ancestral tetra-genetic code, the tesserae, which are ‘symmetric’ codons. It also implies that ancestral tRNAs had at least two mutually complementary (tetra)-anticodons. It is notable that these considerations parallel the observations reported here on sidearms. This not only in the sense that sidearms bear anticodons, but also because the loop–loop interactions presented in Table 1 imply complementarities between all three loops of the tRNA. These parallels with the tesserae hypothesis suggest that patterns described here could be interpreted as the rests of ancient mechanisms, and not of adaptive coevolution between sidearms and codon usage. It is important to stress here that this interpretation could match only results at macroevolutionay scale (Fig. 1). This does not exclude adaptive coevolution at microevolutionary scale. These could not be explained by the patterns reported here. Potentially, other alternative hypotheses, based on current molecular processes, could predict the patterns of microevolutionary coevolution of codon usages with sidearm loops. However,
Author's personal copy H. Seligmann / BioSystems 113 (2013) 165–176
the patterns related to tetracodons are incompatible with such alternatives. In contrast, both codon usage and tetracodon coevolutions are predicted by the working hypothesis. The combined occurrence of positive results from both predictions, matching previous results on tetracoding (Seligmann, 2012a), is strong evidence for the working hypothesis. The original hypothesis that tRNA sidearms bear anticodons seems unrealistic from the point of view of common knowledge on tRNAs. Hence, in the total lack of direct manipulative evidence, the present results at least justify to consider the working hypothesis as a viable working hypothesis purporting at explaining patterns observed for mitochondrial tRNAs. Mitochondrial tRNAs have several notable differences as compared to cytosolic tRNAs, such as the greater variation in sidearm loop sequences: more than 90% of sidearm ‘anticodons’ in cytosolic tRNAs would code for Leu or Ser and in sidearm loop sizes (most are seven nucleosides long, according to data from tRNAdb and mitotRNAdb; http://trnadb.bioinf.uni-leipzig.de/, Sprinzl and Vassilenko (2005), Jühling et al. (2011a); and to be reported elsewhere). 4. Conclusions Import of cytosolic tRNAs into mitochondria implies that mitochondrial translation is promoted by a greater number of tRNAs than those usually recognized templated by the mitochondrial genome. Coevolution between codon usages and numbers of anticodons from antisense tRNAs indicates that antisense tRNAs contribute to mitochondrial translation. In addition, analyses suggest coevolution between codon usages and numbers of anticodons from tRNA sidearms and matching a codon family, as well as between numbers of predicted tetragenes (genes coded by expanded codons of four nucleotides) and tRNA sidearm loops with expanded anticodons. Similar coevolution between numbers of expanded anticodons in antisense tRNAs and tetragenes have been previously reported (Seligmann, 2012a), patterns of covariance between the three tRNA loops concur with predictions of the tesserae hypothesis of mitochondrial genetic code origins as based on quadruplet codons (Gonzalez et al., 2012). Evidences support translational activities by antisense tRNAs and tRNA sidearm loops, and translation of tetracoded genes. GC contents of tRNA sidearms increases the contribution of a tRNA sidearm to translation, indicating that sidearm stability promotes translational activity by that sidearm. Patterns indicate that the tRNA’s D-arm loop adapted for anticodon function at a macroevolutionary level (variation between codon families, an ancient adaptation), and the tRNA’s TC-arm loop adapts for anticodon function at microevolutionary levels (variation within codon families, between mitochondrial genomes from different mammal species), a modern adaptation. Results on stem stability could suggest that tRNA sidearm translational activity originates from isolated, excised tRNA sidearms aminoacylated with amino acids matching their loop’s anticodon, but are not incompatible with a different mechanism, of crossover between tRNA anticodon arm and one of the tRNA’s sidearms. The former mechanism is in line with armless tRNAs in nematode mitochondria, split tRNA genes and deduced hypotheses for a polyphyletic origin of tRNAs by fusion between halves (Widmann et al., 2005) homologous with the tRNA’s sidearms. References Akashi, H., Gojobori, T., 2002. Metabolic efﬁciency and amino acid composition in the proteomes of Escherichia coli and Bacillus subtilis. Proc. Natl. Acad. Sci. U.S.A. 99, 3695–3700.
Alves, R., Savageau, M.A., 2005. Evidence of selection for low cognate amino acid bias in amino acid biosynthetic enzymes. Mol. Microbiol. 56, 1017–1034. Arqués, D., Michel, C.J., 1996. A complementary circular code in the protein coding genes. J. Theor. Biol. 182, 45–56. Barak, Z., Lindsley, D., Gallant, J., 1996. On the mechanism of leftward frameshifting at several hungry codons. J. Mol. Biol. 256, 676–684. Barton, M.D., Delneri, D., Oliver, S.G., Rattray, M., 2010. Evolutionary systems biology of amino acid biosynthetic cost in yeast. PloS ONE 5, e11935. Benjamini, Y., Hochberg, Y., 1995. Controlling for the false discovery rate: a practical and powerful approach to multiple testing. J. R. Stat. Soc. B 57, 289–300. Brzezniak, L.K., Bijata, M., Szczesny, R.J., Stepien, P.P., 2011a. Involvement of human ELAC2 gene product in 3 processing of mitochondrial tRNAs. RNA Biol. 8, 616–626. Brzezniak, L.K., Stepien, P.P., Bijata, M., Szczesny, R.J., 2011b. Human ELAC2 gene encodes the tRNAseZ responsible for mitochondrial 3 processing which acts on precursors already cleaved by RNAseP. RNA Biol. 8, 1–11. Brocchieri, L., Karlin, S., 2005. Protein length in eukaryotic and prokaryotic proteomes. Nucleic Acids Res. 33, 3390–3400. Calvin, K., Li, H., 2008. RNA-splicing endonuclease structure and function. Cell Mol. LifeSci. 65, 1176–1185. Chipman, A.D., Khaner, O., Haas, A., 2001. The evolution of genome size: what can belearned from anuran development? J. Exp. Zool. Mol. Dev. Evol. 291, 365–374. Clayton, D.A., 2000. Transcription and replication of mitochondrial DNA. Hum. Reprod. 15, 11–17. Di Giulio, M., 1992. On the origin of the transfer RNA molecule. J. Theor. Biol. 159, 199–214. Di Giulio, M., 1995. Was it an ancient gene codifying for a hairpin that, by means of direct duplication, gave rise to the primitive tRNA molecule? J. Theor. Biol. 177, 95–101. Di Giulio, M., 1999. The non-monophyletic origin of the tRNA molecule. J. Theor. Biol. 197, 403–414. Di Giulio, M., 2004. The origin of the tRNA molecule: implications for the origin of protein synthesis. J. Theor. Biol. 226, 89–93. Di Giulio, M., 2006. The non-monophyletic origin of the tRNA molecule and the origin of genes only after the evolutionary stage of the last universal common ancestor (LUCA). J. Theor. Biol. 240, 343–352. Di Giulio, M., 2008a. Permuted tRNA genes of Cyanidioschyzon merolae, the origin of the tRNA molecule and the root of the Eukarya domain. J. Theor. Biol. 253, 587–592. Di Giulio, M., 2008b. Transfer RNA genes in pieces are an ancestral character. EMBO Rep. 9, 820. Di Giulio, M., 2009. Formal proof that the split genes of tRNAs of nanoarchaeum equitans are an ancestral character. J. Mol. Evol. 69, 505–511. Di Giulio, M., 2010. A model of the origin of the 5S ribosomal RNA molecule. J. Mol. Evol. 71, 1–2. Di Giulio, M., 2012. The ‘recently’ split transfer RNA genes may be close to merging the two halves of the tRNA rather than having just separated them. J. Theor. Biol. 310, 1–2. Di Giulio, M., 2013. A polyphyletic model for the origin of tRNAs has more support than a monophyletic model. J. Theor. Biol. 318, 124–128. Dimitrov, R.A., Zuker, M., 2004. Prediction of hybridization and melting for doublestranded nucleic acids. Biophys. J. 87, 215–226. Dittmar, K.A., Goodenbour, J.M., Pan, T., 2006. Tissue-speciﬁc differences in human transfer RNA expression. PLoS Genet. 2, e221. Dong, H., Nilsson, L., Kurland, C.G., 1996. Co-variation of tRNA abundance and codon usage in Escherichia coli at different growth rates. J. Mol. Biol. 260, 649–663. Faure, E., Delaye, L., Tribolo, S., Levasseur, A., Seligmann, H., Barthélémy, R.-M., 2011. Probable presence of an ubiquitous cryptic mitochondrial gene on the antisense strand of the cytochrome oxidase I gene. Biol. Dir. 6, 56. Florentz, C., Sohm, B., Tryoen-Toth, P., Pütz, J., Sissler, M., 2003. Human mitochondrial tRNAs in health and disease. Cell. Mol. Life Sci. 60, 1356–1375. Fujishima, K., Sugahara, J., Tomita, M., Kanai, A., 2008. Sequence evidence in the archaeal genomes that tRNAs emerged through the combination of ancestral genes as 5 and 3 tRNA halves. PLoS ONE 3, e1622. Fujishima, K., Sugahara, J., Tomita, M., Kanai, A., 2010. Large-scale tRNA intron transposition in the archaeal order Thermoproteales represents mechanism of intron gain. Mol. Biol. Evol. 27, 2233–2243. Fujishima, K., Sugahara, J., Kikuta, K., Hirano, R., Tomita, M., Kanai, A., 2009. Trisplit tRNA is a transfer RNA made from 3 transcripts that provides insight into the evolution of fragmented tRNAs in archaea. Proc. Natl. Acad. Sci. U.S.A. 106, 283–2687. Fujishima, K., Sugahara, J., Miller, C.S., Baker, B.J., Di Giulio, M., Takesue, K., Sato, A., Tomita, M., Banﬁeld, J.F., Kanai, A., 2011. A novel three-unit tRNA splicing endonuclease found in ultrasmall Archaea possesses broad substrate speciﬁcity. Nucleic Acids Res. 39, 9695–9704. Gonzalez, D.L., Giannerini, S., Rosa, R., 2012. On the origin of the mitochondrial genetic code: towards a uniﬁed mathematical framework for the management of genetic information. Nat. Precedings, http://dx.doi.org/10.1038/npre.2012.7136.1. Gouy, M., Gautier, C., 1982. Codon usage in bacteria: correlation with gene expressivity. Nucleic Acids Res. 10, 7055–7074. Gregory, T.R., 2002. Genome size and developmental complexity. Genetica 115, 131–143. Heizer, E.M., Raymer, M.L., Krane, D.E., 2011. Amino acid biosynthetic cost and protein conservation. J. Mol. Evol. 72, 466–473.
Author's personal copy 176
H. Seligmann / BioSystems 113 (2013) 165–176
Ikemura, T., 1981. Correlation between the abundance of Escherichia coli transfer RNAs and the occurrence of the respective codons in its protein genes. A proposal for a synonymous codon choice that is optimal for the E. coli translational system. J. Mol. Biol. 151, 389–409. Ikemura, T., 1982. Correlation between the abundance of yeast transfer RNAs and the occurrence of the respective codons in protein genes. Differences in synonymous codon choice patterns of yeast and Escherichia coli with reference to the abundance of isoaccepting transfer RNAs. J. Mol. Biol. 158, 573–597. Jühling, F., Mörl, M., Hartmann, R.K., Sprinzl, M., Stadler, P.F., Pütz, J., 2011a. tRNAdb 2009: a compilation of tRNA sequences and tRNA genes. Nucleic Acids Res. 37, D159–D162. Jühling, F., Pütz, J., Bernt, M., Donath, A., Middendorf, M., Florentz, C., Stadler, P.F., 2011b. Improved systematic tRNA gene annotation allows new insights into the evolution of mitochondrial tRNA structures and into the mechanisms of mitochondrial genome rearrangements. Nucleic Acids Res. 40, 2833–2845. Jühling, F., Pütz, J., Florentz, C., Stadler, P., 2012. Armless mitochondrial tRNAs in enoplea (nematode). RNA Biol. 9, 1161–1166. Lee, Y.S., Shibata, Y., Malhotra, A., Dutta, A., 2009. A novel class of small RNAs: tRNA derived RNA fragments (tRFs). Genes Dev. 23, 2639–2649. Markham, N.R., Zuker, M., 2005. DINAMelt web server for nucleic acid melting prediction. Nucleic Acids Res. 33, W577–W581. Markham, N.R., Zuker, M., 2008. UNAFold: software for nucleic acid folding and hybridization. In: Keith, J.M. (Ed.), Bioinformatics, Volume II. Structure, Function and Applications, number 453 in Methods in Molecular Biology. Humana Press, Totowa, NJ, pp. 3–31 (Chapter 1) ISBN 978-1-60327-428-9. Mercer, T.R., Neph, S., Dinger, M.E., Crawford, J., Smith, M.A., Shearwood, A.-M.J., Haugen, E., Bracken, C.P., Rackham, O., Stamatoyannopoulos, J.A., Filipovska, A., Mattick, J.S., 2011. The human mitochondrial transcriptome. Cell 146, 645–658. Michel, C.J., 2012. Circular code motifs in transfer and 16s ribosomal RNAs: a possible translation code in genes. Comput. Biol. Chem. 37, 24–37. Michel, C.J., 2013. Circular code motifs in transfer RNAs. Comput. Biol. Chem. 45, 17–29. O’Connor, M., Gesteland, R.F., Atkins, J.F., 1989. tRNA hopping: enhancement by an expanded anticodon. EMBO J. 8, 4315–4323. Ojala, D., Montoya, J., Attardi, G., 1981. tRNA punctuation model of RNA processing in human mitochondria. Nature 290, 470–474. Percudani, R., Pavesi, A., Ottonello, S., 1997. Transfer RNA gene redundancy and translational selection in Saccharomyces cerevisiae. J. Mol. Biol. 268, 322–330. Perlstein, E.O., de Bivort, B.L., Kunes, S., 2007. Evolutionary conserved optimization of amino acid biosynthesis. J. Mol. Evol. 65, 186–196. Perneger, T.V., 1998. What is wrong with Bonferroni adjustments. Br. Med. J. 136, 1236–1238. Plotkin, J.B., Robins, H., Levine, A.J., 2004. Tissue-speciﬁc codon usage and the expression of human genes. Proc. Natl. Acad. Sci. U.S.A. 101, 12588–12591. Pütz, J., Puglisi, J.D., Florentz, C., Giegé, R., 1991. Identity elements for speciﬁc aminoacylation of yeast tRNA(Asp) by cognate aspartyl-tRNA synthetase. Science 252, 1696–1699. Randau, L., Söll, D., 2008. Transfer RNA genes in pieces. EMBO Rep. 9, 623–628. Rubio, M.A.T., Alfonzo, J.D., 2012. tRNA modiﬁcation, editing, and import in mitochondria. In: Bullerwell, C.E. (Ed.), Organelle Genetics. Springer, Berlin, pp. 359–380. Rubio, M.A.T., Hopper, A.K., 2011. Transfer RNA travels from the cytoplasm to organelles. Wiley Interdisciplinary Rev.: RNA 2, 802–817. Rudinger, J., Puglisi, J.D., Pütz, J., Florentz, C., Giegé, R., 1992. Determinant nucleotides of yeast tRNA(Asp) interact directly with aspartyl-tRNA synthetase. Proc. Natl. Acad. Sci. U.S.A. 89, 5882–5886. Salinas, T., Duby, F., Larosa, V., Coosemans, N., Bonnefoy, N., Motte, P., MaréchalDrouart, L., Remacle, C., 2012. Co-evolution of mitochondrial tRNA import and codon usage determines translational efﬁciency in the green alga Chlamydomonas. PLoS Genet. 8, e1002946. Schneider, A., 2011. Mitochondrial tRNA import and its consequences for mitochondrial translation. Annu. Rev. Biochem 80, 1033–1053. Schneider, A., Maréchal-Drouart, L., 2000. Mitochondrial tRNA import: are there distinct mechanisms? Trends Cell Biol. 10, 509–513. Seligmann, H., 2003. Cost-minimization of amino acid usage. J. Mol. Evol. 56, 151–161. Seligmann, H., 2008. Hybridization between mitochondrial heavy strand tDNA and expressed light strand tRNA modulates the function of heavy strand tDNA as light strand replication origin. J. Mol. Biol. 379, 188–199.
Seligmann, H., 2010a. Mitochondrial tRNAs as light strand replication origins: similarity between anticodon loops and the loop of the light strand replication origin predicts initiation of DNA replication. Biosystems 99, 85–93. Seligmann, H., 2010b. Avoidance of antisense antiterminator tRNA anticodons in vertebrate mitochondria. Biosystems 101, 42–50. Seligmann, H., 2010c. Undetected antisense tRNAs in mitochondria? Biol. Dir. 5, 39. Seligmann, H., 2011a. Mutation patterns due to converging mitochondrial replication and transcription increase lifespan, and cause growth rate-longevity tradeoffs. In: Seligmann, H. (Ed.), DNA Replication-Current Advances. InTech Publisher, http://www.intechopen.com/books/dna-replication-current-advances Seligmann, H., 2011b. Pathogenic mutations in antisense mitochondrial tRNAs. J. Theor. Biol. 269, 287–296. Seligmann, H., 2011c. Two genetic codes, one genome: Frameshifted primate mitochondrial genes code for additional proteins in presence of antisense antitermination tRNAs. Biosystems 105, 271–285. Seligmann, H., 2012a. Putative mitochondrial polypeptides coded by expanded quadruplet codons, decoded by antisense tRNAs with unusual anticodons. Biosystems 110, 84–106. Seligmann, H., 2012b. Overlapping genetic codes for overlapping frameshifted genes in Testudines, and Lepidochelys olivacea as a special case. Comput. Biol. Chem. 41, 18–34. Seligmann, H., 2012c. Positive and negative cognate amino acid bias affects compositions of aminoacyl-tRNA synthetases and reﬂects functional constraints on protein structure. BIO 2, 11–16. Seligmann, H., 2012d. An overlapping genetic code for frameshifted overlapping genes in Drosophila mitochondria: antisense antitermination tRNAs UAR insert serine. J. Theor. Biol. 298, 51–76. Seligmann, H., 2012e. Overlapping genes coded in the 3 -to-5 -direction in mitochondrial genes and the 3 -to-5’polymerization of non-complementary RNA by an “invertase”. J. Theor. Biol. 315, 38–52. Seligmann, H., 2013. Triplex DNA:RNA, 3 -to-5 inverted RNA and protein coding in mitochondrial genomes. J. Comput. Biol. 20, 1–12. Seligmann, H., Krishnan, N.M., 2006. Mitochondrial replication origin stability and propensity of adjacent tRNA genes to form putative replication origins increase developmental stability in lizards. J. Exp. Zool.: Mol. Dev. Evol. 306B, 433–449. Seligmann, H., Krishnan, N.M., Rao, B.J., 2006a. Possible multiple origins of replication in primate mitochondria: alternative role of tRNA sequences. J. Theor. Biol. 241, 321–332. Seligmann, H., Krishnan, N.M., Rao, B.J., 2006b. Mitochondrial tRNA sequences as unusual replication origins: pathogenic implications for Homo sapiens. J. Theor. Biol. 243, 375–385. Session, S.K., Larson, A., 1987. Developmental correlates of genome size in plethodontid salamanders and their implications for genome evolution. Evolution 41, 1239–1251. Sieber, F., Placido, A., El Farouk-Ameqrane, S., Duchene, A.M., Marechal-Drouart, L., 2011. A protein shuttle system to target RNA into mitochondria. Nuc. Acids Res. 39, e96. Sprinzl, M., Vassilenko, K.S., 2005. Compilation of tRNA sequences and sequences of tRNA genes. Nucleic Acids Res. 33, D139–D140. Swire, J., 2007. Selection on synthesis cost affects interprotein amino acid usage in all three domains of life. J. Mol. Evol. 31, 558–571. Tarassov, I.A., Entelis, N., Martin, R.P., 1995. An intact protein translocating machinery is required for mitochondrial import of a yeast cytoplasmic tRNA. J. Mol. Biol. 245, 315–323. Tarassov, I.A., Martin, R.P., 1996. Mechanisms of tRNA import into yeast mitochondria: an overview. Biochimie 78, 502–510. Tsunoda, M., Kusakabe, Y., Tanaka, N., Ohno, S., Nakamura, M., Senda, T., Moriguchi, T., Asai, N., Sekine, N., Sekine, M., Yokogawa, T., Nishikawa, K., Nakamura, K.T., 2007. Structural basis for recognition of cognate tRNA by tyrosyl-tRNA synthetase from three kingdoms. Nucleic Acids Res. 35, 4289–4300. Walker, S.E., Frederick, K., 2008. Recognition and positioning of mRNA in the ribosome by tRNAs with expanded anticodons. J. Mol. Biol. 360, 599–609. Widmann, J., Di Giulio, M., Yarus, M., Knight, R., 2005. tRNA creation by hairpin duplication. J. Mol. Evol. 61, 524–530. Yamao, F., Andachi, Y., Muto, A., Ikemura, T., Osawa, S., 1991. Levels of tRNAs in bacterial cells as affected by amino acid usage in proteins. Nucleic Acids Res. 19, 6119–6122. Yu, C.-H., Liao, J.-Y., Zhou, H., Qu, L.-H., 2008. The rat mitochondrial Ori L encodes a novel small RNA resembling an ancestral tRNA. Biochem. Biophys. Res. Commun. 372, 634–638.