ARTICLE IN PRESS
Journal of Theoretical Biology 243 (2006) 375–385 www.elsevier.com/locate/yjtbi
Mitochondrial tRNA sequences as unusual replication origins: Pathogenic implications for Homo sapiens Herve´ Seligmanna,, Neeraja M. Krishnanb, Basuthkar J. Raob a
Department of Evolution, Systematics and Ecology, The Hebrew University of Jerusalem, 91904, Israel Department of Biological Sciences, Tata Institute of Fundamental Research, Homi Bhabha Road, Colaba, Mumbai 400 005, India
b
Received 6 February 2006; received in revised form 10 May 2006; accepted 27 June 2006 Available online 1 July 2006
Abstract The heavy strand of vertebrate mitochondrial genomes accumulates deaminations proportionally to the time it spends single-stranded during replication. A previous study showed that the strength of genome-wide deamination gradients originating from tRNA gene’s locations increases with their capacities to form secondary structures resembling mitochondrial origins of light strand replication (OL), suggesting an alternative function for tRNA sequences. We hypothesize that this function is frequently pathogenic for those tRNA genes that normally do not form OL-like structures, because this could cause excess mutations in genome regions unadapted to tolerate them. In human mitochondrial genomes, pathogenic tRNA variants usually form less OL-like structures than non-pathogenic ones in cases where the normal non-pathogenic tRNA variant can function as OL, as evolutionary analyses reveal. For tRNAs lacking the putative OL-like functioning capacity, pathogenic variants form more OL-like secondary structures, particularly structures that might invoke bidirectional replication (true for 14 among 21 tRNA species, po0:05, sign test; significantly at po0:05 (1 tailed test) for 7 tRNA species), but not more unidirectional replication invoking structures. Accounting for the functional cloverleaf-like structure-forming capacities of tRNAs yields similar results. Rare, non-pathogenic tRNA mutants tend to form more OL-like structures than the common, nonpathogenic ones, suggesting weak directional selection also among non-pathogenic variants. The duration spent single stranded by a region of the heavy strand (DssH) during replication, estimated by integrating over all regions that can function as OL in Homo sapiens mitochondrial genomes, increases with distance of that region from the Dloop. This suggests convergence of single-strandedness during replication and transcription, and explains conserved locations of tRNA species in mitochondrial genomes and bacterial operons. These locations minimize deamination costs only in anticodons and not in other tRNA regions, during replication and transcription. Therefore, putative functioning as OLs by tRNA sequences is normal at some locations and pathogenic at others. r 2006 Elsevier Ltd. All rights reserved. Keywords: Functional duplication; Secondary structure; Substitution gradient; Mutational robustness; Cumulative error; Aging; Genome organization
1. Introduction Causal relationships between genotype and phenotype remain usually unknown. Error propagation and error accumulation are of general interest in this context, because they can be quantified at all levels of organization, and predictions can be developed on how error at one level propagates to the next level, such as in protein synthesis (Blomberg et al., 1985; Blomberg, 1990; Johansson and Blomberg, 1995). Error propagation was probably a major Corresponding author. Tel./fax: +972 2 6585876.
E-mail address:
[email protected] (H. Seligmann). 0022-5193/$ - see front matter r 2006 Elsevier Ltd. All rights reserved. doi:10.1016/j.jtbi.2006.06.028
factor influencing the first stages of origin of life (Blomberg, 1995). Estimates of error proneness at the level of the molecular machinery of cells, at least for the mitochondrion, are good predictors of developmental instability as measured at the whole organism level by the random components of bilateral asymmetry. This was true for repeatability estimates of protein synthesis (chemical stability of ribosomal RNA increases developmental stability, Seligmann, 2006) as well as of genome replication (secondary structural stability of the light strand replication origin increases developmental stability, Seligmann and Krishnan, 2006). They follow the principle that morphological asymmetry is the direct result of differences
ARTICLE IN PRESS 376
H. Seligmann et al. / Journal of Theoretical Biology 243 (2006) 375–385
in developmental rates of bilateral traits of the two sides. Variation in energetic outputs by mitochondria on the two sides, directly affects growth and differentiation, and hence, causes random differences in the resulting morphologies. These studies describe correlations between molecular and whole organism levels based on variation between different species. The present analyses show similar associations by comparing individuals from the same species (Homo sapiens). They show that repeatability of mitochondrial DNA replication associates with cumulative, developmental disorders at the whole organism level, in the form of various human pathologies associated with mutations in mitochondrial tRNAs. The corroboration at the level of population genetics of the same principles as those observed at micro-evolutionary levels in primates strengthens the hypothesized mechanisms that link molecular and macroscopic repeatabilities. It also emphasizes how the combination of methods from bioinformatics and evolutionary biology can reveal molecular mechanisms with potential medical applications. We consider that this parallel between pathogenic disorders and developmental instability is justified by numerous associations discovered between morphological inaccuracies and various disorders, including indicators of decreased fitness (Thornhill and Moller, 1997; Moller, 1997, 1999, also see Appendix 1 in supplementary material). Our hypothesized mechanism suggests that mutations in the DNA sequences coding for mitochondrial tRNAs alter their propensities to form secondary structures that resemble regular mitochondrial origins of replication of the light strand (OL), thereby altering replication patterns, and causing disorders (see sections below). A notable feature of the approach we adopt is the use of evolutionary variation among primates in order to indicate which mitochondrial tRNAs function in these species as origins of replication, and which do not. Seligmann et al. (2006) show that in the primate mitochondrial genomes, where a tRNA forms OL-like structures, a deamination gradient starting at the location of that tRNA exists, while in those primate species where the same homologous tRNA does not form such structures, no such gradient exists. (These gradients are a direct result of the directional mode of replication, see explanations below.) This correlation between structural propensity at a tRNA, and the strength of a deamination gradient across the genome starting at that tRNA, suggests coevolution. We consider that the tRNA for which such a coevolution exists, is used as OL, at least in some species and, therefore, we hypothesize that its use as OL within the frame of primate evolution is evidence that this function does not cause major disruptions. Hence, we predict that: (1) mutations that decrease the structural propensity of tRNA sequences to form origins of replication are pathogenic in those tRNA species that can normally function as OL (as indicated by coevolutionary analyses between secondary structure of tRNA and deamination gradients), and (2) mutations that increase the structural propensity of tRNA sequences to form
origins of replication are pathogenic in those tRNA species that normally cannot function as OL. This approach is an original application of the principles indicated by the Kluge–Kerfoot phenomenon (Kluge and Kerfoot, 1973; Pierce and Mitton, 1979), which predicts that a correspondence exists between variation between different species and variation within the same species. Our results highlight the advantages of integrating information from population genetics and evolutionary processes, in the same analyses. 1.1. General background on mitochondrial DNA replication In vertebrate mitochondria, replication of the ‘lagging’ light strand usually starts when the ‘leading’ heavy strand replication fork reaches a 30 base pair long stretch that forms a linear stem-loop hairpin structure, called the OL (Clayton, 1982). Mutations, mainly transitions resulting from hydrolytic deaminations (cytosine-thymine and adenine-hypoxanthine, which later gets converted to guanine) accumulate proportionally to the time spent single stranded by the heavy strand until the light strand replication fork synthesizes the complementary light strand. Therefore, gradients of increasing substitution frequencies exist across mitochondrial genomes, proportionally to the time spent single stranded by a site, which is determined by its location versus the OL (Krishnan et al., 2004a, b). This OL is flanked on either side by 2 and 3 tRNA genes, respectively, which also sometimes form OL-like structures (Seligmann et al., 2006; Seligmann and Krishnan, 2006). Evidences indicate that OL-flanking tRNA genes can also sometimes function as OLs: (1) developmental instability, as estimated by fluctuating asymmetry in scalation traits in several lizard families (Amphisbaenidae, Anguidae and Polychritidae), decreases with stability of their mitochondrial OL’s secondary structure, and also decreases with the capacity of the OL-flanking tRNA genes to fold as OL-like structures (Seligmann and Krishnan, 2006); (2) strengths of deamination gradients existing across different primate genomes are frequently proportional to the OL-like structure-forming capacities of the tRNA sequences that mark the origin of these gradients in these species (Seligmann et al., 2006); (3) suppressive mutant Neurospora spp. mitochondrial plasmids that carry insertions at the major 50 end, corresponding to a mitochondrial tRNA, suggest that these tRNAs play a role in replication of the plasmids by reverse transcription, leading to over-production of plasmid transcripts, thereby outcompeting mitochondrial DNA and impairing growth (Akins et al., 1989) and (4) the active sites of class II amino acyl tRNA synthetase and those of the gamma polymerase, the sole polymerase replicating vertebrate mitochondrial DNA (Bolden et al., 1977; Kaguni, 2004), are homologous (Carrodeguas et al., 1999, 2001; Fan et al., 1999; Carrodeguas and Bogenhagen, 2000). Strengthening these lines of evidence are also independent findings that replication of mitochondrial light strands occurs at multiple origins, although apparently independently of
ARTICLE IN PRESS H. Seligmann et al. / Journal of Theoretical Biology 243 (2006) 375–385
the location of tRNAs (Brown et al., 2005; Brown and Clayton, 2006). Based on these findings, we hypothesize that tRNA and OL might also have some level of functional homology or analogy. 1.2. Homologies of protein and nucleic acid chains in the context of replication and tRNA loading processes The modern mammalian gamma polymerase probably evolved from the prokaryotic glycyl tRNA synthetase, apparently acquired by horizontal transfer (Wolf and Koonin, 2001). Note that the subunit originating from this horizontal transfer does not exist in gamma polymerases of microbial eukaryotes, at least as indicated by analyses for yeast (Saccharomyces cerevisiae, Lucas et al., 2004). Homology between two distinct functional domains of proteins as for gamma polymerase and the (aa)-tRNA synthetase exists also for the IMP4/Brix protein superfamily, as found in the archaeon Methanothermobacter (Ng et al., 2005). These homologies also reflect similarities in the functions of the two domains in these proteins: (a) gamma polymerase recognizes a specific structure on the heavy strand of the DNA sequence to be replicated and in particular, the accessory subunit has a role in recognizing RNA primers synthesized at the replication origin. In aatRNA synthetase, the anticodon-binding domain recognizes the structure of the tRNA to be loaded with its specific amino acid; and (b) the polymerase starts synthesizing the light strand at the 50 flanking sequences of the OL’s structure (Hixson et al., 1986). This is reminiscent of tRNA synthetases that load their cognate amino acids to the acceptor stem of tRNAs, whose structure resembles that of the 50 flanking sequence of the OL. It consists of a single-stranded stretch flanking a stem region, and has a nucleotide composition similar to that of its associated anticodon (Seligmann and Amzallag, 2002). 1.3. Developmental instability and mitochondrial replication The association between low OL stability and developmental instability, as well as OL-like structure-forming capacities of tRNA genes flanking the regular OL and developmental instability (as estimated by fluctuating asymmetry, FA) suggest a link between the excess amount of mutations accumulated due to delays in initiation of light strand replication and low developmental regulation (Seligmann and Krishnan, 2006). Since deamination gradients are an inevitable result of directional replication, it is plausible that the overall genome organization is adapted in a way that the regions that accumulate most mutations are those that can most likely tolerate high mutation rates with no or little functional effects. Therefore, mutations in tRNA sequences that alter the OL-like structure-forming capacities, and thereby alter genomewide mutation accumulation patterns, probably decrease developmental stability, especially if excess mutations
377
accumulate in the more vulnerable coding regions. These processes would lead to variation among mitochondria, and among those on different sides of the organism. This variation would differentially affect energetic outputs of cells and tissues, ultimately resulting in different growth rates for bilateral body parts, causing morphological asymmetry. The observation that times spent single stranded by sites, estimated as a fraction of the genome length, do not exceed ‘1’ when calculated with the location of the regular OL, but these fractions sometimes do exceed ‘1’, while considering other tRNA gene clusters as putative OLs to estimate the time spent single stranded by a site (Seligmann et al., 2006) also suggests that when replication is not initiated at the regular OL, excess mutations result from prolonged single-strandedness. 1.4. OL, developmental instability and pathologies in Homo sapiens Formation of OL-like structures by tRNA sequences flanking the OL decreases developmental instability (Seligmann and Krishnan, 2006). Supposedly, this is because these OL-like structures increase the probability that replication starts in the region that is normally designed for replication initiation. According to this rationale, OLlike structures in most other regions of the genome should increase developmental instability, and the tendency to develop pathologies. We investigate whether human mitochondrial tRNA variants known to associate with pathologies have higher OL-like structure-forming capacities than the non-pathogenic variants for tRNA species that normally do not function as OL, and vice versa for those that can normally function as OL in primates (as described in Seligmann et al., 2006). This hypothesis is justifiable because (1) many observations suggest an association of diseases with developmental instability, (2) an association exists between developmental instability and OL-like structure-forming capacity of mitochondrial tRNAs and (3) clover-leaf foldability indices of pathogenic variants are low (Krishnan et al., 2005), which could be indicative of greater foldability into other types of structures. This could suggest an unsuspected mechanism of pathogenicity for mitochondrial tRNAs. We explore this for two types of OL structures, one putatively favoring unidirectional replication which is the normal process of replication in vertebrate mitochondria (Fig. 1A) and mimics the recognized OL (Fig. 1B), and one which might favor bi-directional replication (Fig. 1C). The functional clover-leaf structure of the tRNA is shown in Fig. 1D. 1.5. Mitochondrial diseases and putative alternative functions of tRNA molecules It is possible that part of the pathologies associated with variants in gamma polymerase (Trifunovic et al., 2004; Taylor and Turnbull, 2005), that are usually believed to be
ARTICLE IN PRESS 378
H. Seligmann et al. / Journal of Theoretical Biology 243 (2006) 375–385
Fig. 1. Alternative secondary structures of tRNA-Thr in Homo sapiens in (A) stem-loop OL-like; (B) hairpin structure formed by the regular OL; (C) bi-directional OL; and (D) clover-leaf shaped structures.
due to inaccurate DNA synthesis (Copeland et al., 2003), actually result from the relative differences in affinities of different polymerase variants to OL-like structures formed by regions that are usually used as OL, versus their affinities to such structures in other regions, such as some tRNA sequence variants. Indeed, a disproportionate amount of mitochondrial mutations associated with pathologies are in tRNAs, considering that tRNAs represent only 8.5% of mitochondrial genomes. Typically, these mutations lead to low ATP supply (for example, see James et al., 1999). Pathogenic tRNA mutants also showed lower clover-leaf foldability than rare polymorphic tRNA variants in the human population, suggesting that they have a greater
propensity to form other alternative structures (Krishnan et al., 2005). Several functional features of tRNAs can associate with various diseases, but the functional aspects of the genotype–phenotype relationship are not well understood (Jacobs, 2003). It is particularly relevant to this study that the many diseases associated with tRNA mutants have highly varying etiologies (see Appendix 2 in supplementary materials). This seems to fit with the ‘‘uncommon’’ OL hypothesis that we propose, because initiation of replication in an unusual region could cause a shift in the substitution frequency versus single-strandedness profile, but does not necessarily determine which specific mutation occurs. Hence, variants of the same tRNA species or even the same tRNA variant can cause
ARTICLE IN PRESS H. Seligmann et al. / Journal of Theoretical Biology 243 (2006) 375–385
very different diseases, according to which specific mutations (or combination of mutations) occurred. These various diseases could also reflect various mechanisms by which tRNA sequences might contribute to pathology, and the OL-like function might be one of them (Appendix 3 in supplementary materials reviews evidence for the various mechanisms and associated diseases). However, while many mitochondrial mutations associate with tumor development and propagation in cancers, there seems to exist no direct link with mutations in mitochondrial tRNAs (Czarnecka et al., 2006). This is particularly notable as cancer probabilities display the general age-associated, inexorable, cumulative progression as found for other diseases associated with mitochondrial tRNAs. 1.6. Confounding pathogenic effects of aminoacylation and OL-like function of tRNAs It is also possible that part of the pathogenic effects that seem associated with structural changes affecting aminoacylation, result from the use of a tRNA gene as OL during replication, because of the homology between tRNA synthetase and gamma polymerase (the sole replicating polymerase in vertebrate mitochondria, Bolden et al., 1977; Kaguni, 2004). Interactions between tRNA sequence and gamma polymerase might resemble those between tRNA and tRNA synthetase. Such confounding effects would explain why associations between genotypes (in any of the different genes involved) and diseases remain elusive (as an example, in polymerase gamma; Kollberg et al., 2005). In other words, these structural changes can cause both dysfunctional translation and unwanted replication processes. The cumulative nature of deaminations caused by the potential functioning of a given tRNA as OL is compatible with several complex observations: hypotheses on programmed cell death and dysfunction (Mirabella et al., 2000), where cell death would deterministically occur due to the accumulation of deaminations in a functionally crucial region, and because of the overall change in deamination gradient spectrum; the need to combine mutations from tRNA and coding sequences in order to predict some pathologies (Paul et al., 2000); organ-specific, age-related cumulations of mutations in specific regions (Murdock et al., 2000); and that the non-tRNA mutations are not part of the haplotype background (Torroni et al., 2003) and hence, are probably newly acquired. It is also compatible with the fact that some disease-associated mutations in tRNAs do not result in the metabolic dysfunctioning of the tRNA (Bornstein et al., 2002), however, dysfunction associated with the tRNA mutant might result subsequently from accumulation of mutations elsewhere in the genome. It has been observed that the same tRNA mutation can associate with different pathologies (Campos et al., 2003), can lack pathological symptoms (McFarland et al., 2004) and can show different ‘penetrancies’ of pathogenic mutations (Limongelli et al.,
379
2004). Other observations in line with our hypotheses are the high overall variability observed at both genotypic and phenotypic levels (Shanske and Wong, 2004; Wong and Boles, 2005); observations that Parkinson’s disease does not associate with unusually high overall mutation rates, but with higher localized ones (Smigrodzki et al., 2004); and particularly, the fact that most mitochondrially associated diseases are characterized by inexorable progression (Smith et al., 2004). Observations that some tRNA mutations do not cause strong decrease in protein synthesis rate (Janssen et al., 1999) might, and might not fit in this list, but our hypothesis cannot explain the pathologic dissonance in Parkinson disease between identical twins with identical mitochondrial sequences (Kosel et al., 2000). Similarly, the OL hypothesis does not explain how introduction of some pathogenic tRNA variants into yeast mitochondria caused metabolic dysfunction (Feuermann et al., 2003), because gamma polymerase in yeast does not possess the subunit homologous to tRNA synthetase. Superficially, pathogenicity of tRNA mutants in humans clashes with the fact that pathogenic tRNA mutation sites in Homo sapiens possess the same nucleotide as their homologous sites in some non–pathogenic tRNA variants of several non-human mammals, but in fact, the nonpathogenicity in other mammals is explained by compensating substitutions (Kern and Kondrashov, 2004; Kondrashov, 2005) that enable normal tRNA structure and its stability. It is possible that mutations increasing OL-like structure-forming capacities for the tRNAs that do not normally form OL-like structures, associate with some pathologies, especially those related to aging and progressive degeneration syndromes. Note that our hypothesized mechanism implies that several different pathologies can result from the same mutation in tRNA that confers it a putative OL-like function, because the exact nature and locations of cumulative deaminations are only probabilistically determined by the usage of the tRNA as OL. Our hypothesis predicts: (1) that disease-associated tRNA mutants have higher OL-like structure-forming capacities than the non-pathogenic polymorphic variants of the same tRNA species in human populations; and (2) for those tRNAs whose normal variants in primates do not function as OLs (as described by Seligmann et al., 2006), association between pathogenicity and OL-forming capacity is stronger and vice versa.
2. Materials and methods We used 1914 complete mitochondrial genomes available for Homo sapiens from NCBI GenBank (www.ncbi. nlm.nih.gov/) until December 2005. In order to avoid annotation problems for tRNAs, such as inaccurate limits (Slack et al., 2003), we extracted tRNA sequences from the GenBank genome data by tRNAScan-SE (Lowe and Eddy, 1997, http://www.genetics.wustl.edu). We reverse
ARTICLE IN PRESS 380
H. Seligmann et al. / Journal of Theoretical Biology 243 (2006) 375–385
complemented sequences for the tRNAs that are coded by the light strand. We used the DNA folding version of Mfold (Zuker, 2003) to examine all predicted alternative secondary structures at least half as stable as the most stable structure of the heavy strand sequences of the recognized OL. For heavy strand sequences of tRNAs, we recorded the percentage of alternative structures that are OL-like (see Fig. 1A), and the percentage of structures that could putatively function as bidirectional OL with replication forks proceeding in both directions (Fig. 1C). We calculated the OL-like structure-forming tendency of a tRNA gene cluster as the percentage of OL-like structures among the examined alternative structures for all tRNA sequences in that cluster. Mfold predictions are relatively inaccurate for long sequences, but accuracy seems to reach reasonable levels for short sequences, such as tRNAs (Doshi et al., 2004). In addition, the reported inaccuracies are for specific functional structures as compared to the sub-optimal secondary structures predicted by Mfold. However, these inaccuracies probably have little effect on average structural properties of tRNAs derived from the distribution of alternative secondary structures such as the percentage of alternative sequences resembling an OL such as used here. Pathogenic tRNA mutants are from Florentz et al. (2003) and from mitomap (www.mitomap.org; Brandon et al., 2005). Pathogenic variants from GenBank were defined as those variants identified by tRNAScan-SE (Lowe and Eddy, 1997), which were identical to one of those defined by the above sources. Other tRNA variants were considered as nonpathogenic (normal) polymorphic variants by default. We compared average OL-like structure-forming capacities of pathogenic variants with the other variants from the same tRNA species, using the parametric t-test statistics, and use the t-statistic to quantify the strength and direction of the association between pathogenicity and the capacity to form OL-like secondary structures by that tRNA gene.
3. Results and discussion 3.1. Inaccuracies in GenBank annotations Table 1 displays the number of pathogenic variants, the number of non-pathogenic polymorphic variants, the number of variants extracted from NCBI-GenBank data using tRNAScan-SE, and the number of variants extracted from GenBank using annotations from Genbank for each of the 22 tRNA genes in human mitochondrial genomes. The latter number is usually larger than the number from tRNAScan-SE, because of inconsistencies in annotations of genomes (Slack et al., 2003): for many genomes in GenBank, tRNAs were detected only by approximate sequence homology.
Table 1 Differences between OL-like structure forming capacities of pathogenic and non-pathogenic variants of heavy strand sequences of human mitochondrial tRNA genes tRNA
Pa No Scan NCBI tOL
Ala 2 11 11 Arg 0 11 8 Asn 3 8 7 Asp 1 14 11 Cys 1 21 21 Gln 3 19 17 Glu 1 9 8 Gly 5 14 14 His 3 12 10 Ile 13 13 9 Leu (UUR) 24 8 12 Leu (CUN) 6 14 16 Lys 13 11 10 Met 2 7 7 Phe 6 12 13 Pro 2 12 16 Ser (AGY) 2 16 16 Ser (UCN) 6 7 13 Thr 4 42 41 Trp 5 14 11 Tyr 3 8 8 Val 5 12 10
16 10 11 13 29 18 13 16 12 10 10 14 12 8 13 22 18 18 45 13 9 11
P
0.51 0.62 0.19 0.38 1.45 1.82 0.08 1.19 0.88 0.22 1.56 0.73 0.46 0.03 0.17 2.19 1.60 2.71 3.69 0.47 0.32 0.53
0.85 0.71 0.16 0.083 0.94 0.25 0.39 0.83 0.16 0.48 0.65 0.98 0.87 0.05 0.13 0.02 0.001 0.65 0.76 0.60
tBi 0.92
P 0.38
3.32 0.009 5.92 0.000 0.69 0.500 0.45 0.66 4.92 0.001 2.59 0.019 0.84 0.410 1.33 0.200 2.41 0.023 0.13 0.90 1.17 0.26 2.66 0.032 1.22 0.24 0.29 0.78 1.93 0.072 1.64 0.13 1.69 0.098 0.24 0.81 0.80 0.45 1.46 0.16
FOL
FBi
0.22 0.21 0.39 0.12 0.07 0.28 0.35 0.22 0.16 0.10 0.26 0.28 0.38 0.40 0.17 0.15 0.18 0.55 0.30 0.43 0.16 0.28
0.35 0.20 0.36 0.06 0.19 0.12 0.19 0.31 0.35 0.05 0.31 0.32 0.29 0.19 0.37 0.15 0.26 0.28 0.03 0.25 0.47 0.50
tRNA species are indicated by the abbreviation of their cognate amino acid. Those coded on the heavy strand are underlined. Column headers indicate: Pa—number of pathogenic variants; No—number of nonpathogenic variants; Scan—number of variants detected according to tRNAScan-SE in NCBI genomes; NCBI—number of tRNA variants based on alignments obtained according to annotations in GenBank (which includes variants due to inaccuracies in assigning gene limits); tOL and tBi— t-statistics of differences between unidirectional and bidirectional OL-like structure forming capacity of Pa and No; p—two tailed statistical significance of each; FOL and FBi indicate the correlation coefficients between the frequencies of non-pathogenic tRNA variants and their capacity to form OL-like secondary structures for putatively uni- and bi-directional OLs, respectively.
3.2. Pathogenicity and OL secondary structures The sixth column indicates the t-statistics of the differences between unidirectional OL-like structure-forming capacities of pathogenic and non-pathogenic tRNA variants, column seven indicates the statistical significance, p, of the t-test. Columns 8 and 9 indicate t-statistics and p for differences between pathogenic and non-pathogenic tRNA variants for putative bidirectional OL-like secondary structures. Pathogenic variants form, on average, more unidirectional OL-like structures in only 9 among 21 tRNA species, significantly only in tRNA-Ser (UCN) and have significantly less unidirectional OL-like secondary structures than non-pathogenic ones for two tRNAs that are coded by adjacent sequences on the mitochondrial genome (tRNA-Pro and tRNA-Thr). For a significant majority of tRNAs (14 among 21 cases, sign test, po0:05), the forming capacity of bidirectional OL-like structures is on an average higher for tRNA variants associated with diseases than for non-pathogenic polymorphisms of that tRNA
ARTICLE IN PRESS H. Seligmann et al. / Journal of Theoretical Biology 243 (2006) 375–385
species. This result is statistically significant at po0:05 for seven specific tRNAs (six according to two tailed tests, seven according to one-tailed tests). These results suggest that bidirectional replication, which is not the normal mitochondrial replication mechanism, is usually pathogenic. Alternatively, it is possible that the structures suggested to function as bi-directional OLs are more efficient unidirectional OLs, rather than bi-directional OLs: stem lengths of the 50 half of presumed bi-directional OLs are similar to the stem lengths of the normal, recognized OL, while the stem lengths of unidirectional OLs are much longer. This would also result in higher pathogenic levels of ‘bidirectional’ OL structures than unidirectional OL structures. These two unidirectional OLs would roughly correspond to the 50 and 30 halves of tRNAs, which putatively evolved from different origins, as suggested by the non-monophyletic hypothesis of origin of tRNAs (Di Giulio, 2005), or by duplication of a single hairpin (Widmann et al., 2005). 3.3. Cloverleaf-associated effects Note that the COVE index estimated by tRNA-scan SE correlates positively with the percentage of alternative MFold secondary structures resembling cloverleaf structures in 15 among 20 tRNAs (significantly at po0:05 for 10 tRNAs, one-tailed tests), which suggests that both algorithms, tRNA-scan SE and MFold, converge in their estimates of cloverleaf foldability. The COVE index correlates positively with the percentage of unidirectional OLs in 13 among 21 tRNAs, significantly only for tRNAAsn and tRNA-His (positive and negative correlations, respectively, two tailed tests). For percentages of bidirectional OL-like structures, correlations with COVE indices were negative in 16 among 21 tRNAs, significantly so only for tRNA-Ala. In order to account for possible colinearity effects between capacities to form OL-like structures and capacities to form regular cloverleaf structures, we tested for association between pathogenicity and OL-like structure-forming capacities after correcting for the association between OL-like structure-forming capacities and cloverleaf-forming capacities, estimated by the COVE index from tRNAScan-SE (Lowe and Eddy, 1997). This analysis, which calculates the residual OLforming capacity from the linear regression between the uncorrected OL-forming capacity (as dependent) and the COVE index (as independent, done separately for each uniand bi-directional OL-like structures), considers that these residuals reflect the OL-forming capacity independently of the cloverleaf-forming capacity of the sequence. This was done in order to take into account the potential association existing between the capacity to form OL-like structures and cloverleaf structures, indicated also by the possibility that primitive tRNAs had OL-like structures (Rodin et al., 1993; Di Giulio, 1999, 2004), and thereby, avoiding tests that would indirectly quantify the pathogenic effects of cloverleaf disruption. This is also done in order to control
381
for the possible functional homology between aminoacylation by tRNA synthetases and gamma polymerase binding, the former associated with the acceptor stem of the cloverleaf, the latter with the 50 sequence flanking the OL-like structure. Overall, results were qualitatively similar for regular analyses and those controlling for cloverleaf foldability, so that one can conclude that the capacity of tRNAs to form OL-like structures is pathogenic, independently of the cloverleaf-forming capacity of the sequence. This capacity of tRNAs to form OL-like structures most probably contributes to pathogenicity of tRNA mutants, in addition to the classical factors: disruption of cloverleaf structure, codon–anticodon interactions, interactions with tRNA synthetases, tRNA maturation, etc. 3.4. Potential selection on OL-like structure-forming capacities of mitochondrial tRNA variants in natural human populations The associations between OL-like structure-forming capacities and pathogenicity in tRNA sequences suggest that some weak directional selection might exist among non-pathogenic polymorphisms. If this is the case, there should be a correlation between OL-like structure-forming capacities of non-pathogenic tRNA variants and their occurrence in the natural human population, as estimated by the number of individuals possessing the particular variant out of the 1914 complete human mitochondrial genomes from GenBank. If the selection pressures on rare natural variants and pathogenic variants have similar causes, then the direction and strength of correlations between population frequency and OL-like structureforming capacities in natural variants should associate with the t-statistics in Table 1 (columns 6 and 8) that estimate the association between OL-like structure-forming capacities and pathogenicity. We found that the frequency of non-pathogenic tRNA variants correlates negatively with their capacity to form unidirectional OL structures in 17 among 22 tRNAs (considering all 22 tRNAs, a sign test detects a statistically significant tendency at po0:05; but at the level of each single tRNA, it is significant according to a one-tailed t-test only for tRNA-Thr, FOL in Table 1). This suggests that weak selection against unidirectional OL structures exists, at large, in the population for tRNAs, where this capacity does not clearly associate with pathologies. However, note that correlations between frequency and functional properties of polymorphic variants cannot indicate the direction of selection without measuring the change in their frequencies across generations. This is true because rare polymorphisms might be rare, either because of selection against them or because they newly arose, but still might be under positive selection. It is however clear that the direction of selection is against pathogenic mutants, and it is most likely that, generally, rare polymorphisms are not under positive selection and probably are being counter-selected. Convergence between population-level evidence indicating selection and outcome
ARTICLE IN PRESS H. Seligmann et al. / Journal of Theoretical Biology 243 (2006) 375–385
382
of functional analyses generally fit the relatively strict criteria considered necessary for deciding whether natural selection occurs on a trait. One of the main points is that a functional hypothesis such as ours on OL-like structures is necessary to suggest which variants are selectively favored and disfavored (Endler, 1986). 3.5. Coevolution between deamination gradients and OLlike secondary structures, in the context of pathogenicity of tRNAs
Association between tRNA-OL structure and pathogenicity
Fig. 2 plots the strength of the association of pathogenicity and OL-like structure-forming capacity in Homo sapiens (the t-statistics in column 5 from Table 1) as a function of the strength of the coevolution between the OLlike structure-forming capacity of a mitochondrial tRNA gene cluster and the strength of the deamination gradient originating from that cluster in primates (Seligmann et al., 2006). A negative correlation exists between the x- and yaxis indicating that the OL-like structure-forming capacity of tRNA variants associates positively with pathogenicity for tRNAs where the OL-like structure and correspondingly originating deamination gradients did not coevolve in primates (for example, tRNA-Cys). At the same time, a negative association between OL formation and pathogenicity exists in those tRNAs where such coevolution occurred in primates, and hence, putative OL-like function of these tRNA sequences is probably part of the normal molecular physiology in primates (notably the cluster formed by tRNA-Pro and tRNA-Thr) (r ¼ 0:45, p ¼ 0:022; Spearman rank correlation coefficient rs ¼ 0:38;
SerUCN#
Cys Tyr
0
His Ile Met Asn Trp, Ala
Gln
Glu
Asp LeuCUN
Phe* Lys*
SerAGY
r = -0.45, P = 0.022, Spearman rank r = -0.38, P = 0.045, 1 tailed tests
-3.7 0
Val* Gly* Pro*#
Thr*#
0.2 0.4 0.6 Coevolution between tRNA OL-structure and deamination gradient
0.8
Fig. 2. Association of OL-like structure-forming capacity with pathogenicity in Homo sapiens as a function of the correlation coefficient between the strength of deamination gradients starting at a tRNA cluster at the third codon position and the OL-like structure-forming capacity of tRNAs in that cluster in primates, a measure of coevolution between tRNA’s OL-like structure-forming capacity and usage (as reflected by deamination gradients) of that tRNA cluster as origin of replication. * indicates po0:05 for t-statistics (y-axis), # indicates po0:05 for correlation coefficients that estimate coevolution, (x-axis).
p ¼ 0:045, one-tailed tests). This correlation is much stronger if one considers only the tRNAs for which correlations between gradient strength and secondary structure propensity are significant as marked by asterisk in Fig. 2 (r ¼ 0:83, p ¼ 0:02; rs ¼ 0:87, p ¼ 0:012). Including in the correlation analyses, the tRNA for which propensity to form OL-like structures associates significantly with pathogenicity (tRNA Ser-UCN) still strengthens this trend (r ¼ 0:90, p ¼ 0:003; rs ¼ 0:92, p ¼ 0:002). Hence, pathogenicity of OL-like structureforming tRNAs in Homo sapiens is inversely proportional to the usage of normal tRNA variants in that gene cluster as putative OL, as estimated by coevolution between deamination gradients and OL-like structure-formation capacity of the sequences among the evolutionary relatives of Homo sapiens. 3.6. Tolerance of genome regions to accumulation of somatic mutations These results suggest that the locations of genes relative to that of the regions functioning as OLs, and the functional tolerance of gene sequences to mutations coevolve, also with other components of the molecular replication machinery of the mitochondrion. This complex coevolution of components of the molecular machinery along with whole genome properties is not unique: within primates itself, the structural stability of mitochondrial ribosomes, putatively inversely proportional to the frequency of ribosomal slippages on the mRNA, correlates negatively with whole genome content of off frame stop motifs within coding regions (Seligmann and Pollock, 2004). We calculated the time spent single stranded by genome regions during replication according to all the OLs we detected. This total DssH is calculated by summing the DssHs according to various tRNA clusters (supposedly functioning as OLs). This was done separately depending on whether the association of unidirectional OL-like structure and pathogenicity is positive or negative. The DssH from each tRNA cluster was weighted as the proportion of the (positive or negative) ts’ for that tRNA cluster out of the sum of the absolute values of all tstatistics (as in column 5 from Table 1). For both pathogenic and non-pathogenic tRNA sequences, total DssH increases linearly with the distance from the Dloop. This means that the potential effects of deaminations occurring during DNA replication, due to the use of various OLs, coincide with deamination gradients at the RNA level that probably form during the synthesis of the unique mitochondrial RNA transcript during transcription. This situation is likely to be adaptive, and not random, as it gives a single, simple solution to the problems due to deaminations during both DNA replication and RNA transcription. It could also explain the high level of conservation of the mitochondrial genome organization, including that of the localization of tRNA clusters. Indeed, tRNAs might also be affected by deamination gradients:
ARTICLE IN PRESS
(A-G+3) at first and second anticodon positions
the base contents at the wobble position of mitochondrial tRNA anticodons are almost all as expected after deamination (Xia, 2005). This is strengthened by our observation that a gradient exists in numbers of Gs subtracted from the numbers of As at first and second anticodon positions of tRNAs as a function of the distance of the tRNA from the Dloop (Fig. 3). This suggests that the conserved mitochondrial order of tRNAs is adapted to avoid deaminations from adenine to guanine in the tRNA’s functionally most critical part, the amino acid-specific anticodon positions, probably at both DNA replication and RNA transcription levels. Similar analyses of correlation of base compositions in other functional and structural elements of human mitochondrial tRNAs are usually not correlated with the distance from the Dloop (Fig. 4): the correlation observed for the anticodon is the strongest among all 28 tests (2 measures of base composition for 14 elements, including the two first anticodon positions). Only about half of the correlations were as expected by the deamination gradient hypothesis, which suggests that tRNA base compositions are usually not affected by deamination gradients. This analysis on other tRNA structural regions functions as a control, and suggests that the gradient observed in anticodons results from optimization of arrangement of tRNA sequences across the genome, as related to the functional disruption due to deamination. This rationale predicts that a similar tRNA order should exist in larger genomes (both at replication and transcription levels), which is indicated by our preliminary observations of tRNA order in Escherichia coli (not shown). As the general association between tRNAs’ OL-like structure-forming capacity and pathogenicity is now
Gln Phe
5
Glu
Asn
Correlation between base content and distance from Dloop
H. Seligmann et al. / Journal of Theoretical Biology 243 (2006) 375–385
383
0.7 Stem, A-G Stem, C-T Loop, A-G Loop, C-T
*
0
* anticodon
-0.7 1
3
5
7
9
11
13
Rank of the tRNA structural element Fig. 4. Correlation coefficients of excess A’s over G’s and excess T’s over C’s in different structurally and functionally homologous elements of 22 human mitochondrial tRNAs with the distance of the tRNA from the Dloop, as a function of the rank of the element. Ranking starts from the 50 end of the tRNA sequence according to which, elements ranked 1 and 12 are the 50 and 30 ends of the tRNA sequence that form the acceptor stem; 13 is the single stranded part of 30 acceptor stem; 7 is the anticodon loop. Structures from tRNA-Scan (Lowe and Eddy, 1997) were used to define homology between structural elements of tRNAs. The base composition conventions are according to the heavy strand sequence of the tRNA. The correlation coefficients for the base composition of the 2 first positions of the anticodons (data for A–G are shown in Fig. 3) also have distance rank 7. Asterisks indicate correlation coefficients that are statistically significant at po0:05, one-tailed test.
assessed, especially in tRNAs that have not been used as OL during primate evolution, the next logical step is a more detailed review of pathogenic variants, across tRNA species, which could suggest which specific types of pathologies associate, and which do not, with altered OL function.
Leu2 Trp Val
4
4. Conclusions
Asn Ile Met Ala
3
Ser4
Lys
Gly
Asp
Cys
2
Leu2
Arg
His Thr
r= -0.67 P < 0.001,2 tailed test (exponential regression model)
1
Pro
0 0
5000
10000
15000
Distance from Dloop Fig. 3. Base composition of first and second anticodon position in tRNAs as a function of distance from the Dloop of the tRNA sequence. The yaxis plots the subtraction of the number of Gs from the number of As in the 2 first anticodon positions, adding 3 to avoid negative values and zero. The base composition is according to the heavy strand sequence of the tRNA.
We found that: (1) secondary structures of tRNAs putatively compatible with bi-directional replication are usually pathogenic in human mitochondria; (2) secondary structures of tRNAs putatively compatible with unidirectional replication are pathogenic in human mitochondria when there is no evidence from the evolution of primate mitochondrial genomes that this tRNA functions as origin of light strand replication (OL), and the lack of such structures is pathogenic in tRNAs for which evolution of primate mitochondrial genomes suggest that this tRNA normally functions as OL without causing pathologies; (3) analyses of tRNA polymorphism frequencies in the human population at large, suggest that usually, rare polymorphic variants are selected against, and that this associates with their capacities to form OL-like structures; (4) the pattern of time spent single stranded (DssH) during DNA replication by different mitochondrial genes calculated using all
ARTICLE IN PRESS 384
H. Seligmann et al. / Journal of Theoretical Biology 243 (2006) 375–385
the tRNA clusters as potential OLs, converges with the pattern expected for the same regions during RNA transcription, suggesting that the potential functioning of tRNAs as OLs is organized so as to give a unique solution to the problem of deaminations during both DNA replication and RNA transcription; (5) the latter point also explains the conserved gene order of tRNAs in mitochondrial (and other) genomes, as their organization seems to minimize frequencies of deaminations in the anticodon. Appendix A. Supplementary Materials Supplementary data associated with this article can be found in the online version at doi:10.1016/j.jtbi.2006. 06.028.
References Akins, R.A., Kelley, R.L., Lambowitz, A.M., 1989. Characterization of mutant mitochondrial plasmids of Neurospora spp. that have incorporated tRNAs by reverse transcription. Mol. Cell Biol. 9, 678–691. Blomberg, C., 1990. Modeling efficiency, error propagation and the effect of error-enhancing drugs in protein-synthesis. Biomed. Biochim. Acta 49, 879–889. Blomberg, C., 1995. The role of accuracy for early stages of the origin of life. Origins Life Evol Biosphere 25, 219–226. Blomberg, C., Johansson, J., Liljenstrom, H., 1985. Error propagation in E. coli protein synthesis. J. Theor. Biol. 113, 407–423. Bolden, A., Noy, G.P., Weissbach, A., 1977. DNA-polymerase of mitochondria is a gamma-polymerase. J. Biol. Chem. 252, 3351–3356. Bornstein, B., Mas, J.A., Fernandez-Moreno, M.A., Campos, Y., Martin, M.A., del Hoyo, P., Rubio, J.C., Arenas, J., Garesse, R., 2002. The A8296G mtDNA mutation associated with several mitochondrial diseases does not cause mitochondrial dysfunction in cybrid cell lines. Hum. Mutat. 19, 234–299. Brandon, M.C., Lott, M.T., Nguyen, K.C., Spolim, S., Navathe, S.B., Baldi, P., Wallace, D.C., 2005. MITOMAP: a human mitochondrial genome database—2004 update. Nucleic Acids Res. 33 (Database Issue), D611–D613 URL: http://www.mitomap.org. Brown, T.A., Clayton, D.A., 2006. Origins and migrations in asymmetrically replicating mitochondrial DNA. Cell Cycle 5, e1–e5. Brown, T.A., Cecconi, C., Tkachuk, A.N., Bustamante, C., Clayton, D.A., 2005. Replication of mitochondrial DNA occurs by strand displacement with alternative light-strand origins, not via a strandcoupled mechanism. Gen. Dev. 19, 2466–2476. Campos, Y., Garcia, A., del Hoyo, P., Jara, P., Martin, M.A., Rubio, J.C., Berbel, A., Barbera, J.R., Ribacoba, R., Astudillo, A., Cabello, A., Ricoy, J.R., Arenas, J., 2003. Two pathogenic mutations in the mitochondrial DNA tRNA Leu(UUR) gene (T3258C and A3280G) resulting in variable clinical phenotypes. Neuromusc. Disord. 13, 416–420. Carrodeguas, J.A., Bogenhagen, D.F., 2000. Protein sequences conserved in prokaryotic aminoacyl-tRNA synthetases are important for the activity of the processivity factor of human mitochondrial DNA polymerase. Nucleic Acids Res. 28, 1237–1244. Carrodeguas, J.A., Kobayashi, R., Lim, S.E., Copeland, W.C., Bogenhagen, D.F., 1999. The accessory subunit of Xenopus laevis mitochondrial DNA polymerase gamma increases processivity of the catalytic subunit of human DNA polymerase gamma and is related to class II aminoacyl-tRNA synthetases. Mol. Cell Biol. 19, 4039–4046.
Carrodeguas, J.A., Theis, K., Bogenhagen, D.F., Kisker, C., 2001. Crystal structure and deletion analysis show that the accessory subunit of mammalian DNA polymerase gamma, Pol gamma B, functions as a homodimer. Mol. Cell 7, 43–54. Clayton, D.A., 1982. Replication of animal mitochondrial DNA. Cell 28, 693–705. Copeland, W.C., Ponamarev, M.V., Nguyen, D., Kunkel, T.A., Longley, M.J., 2003. Mutations in DNA polymerase gamma cause error prone DNA synthesis in human mitochondrial disorders. Acta Bioch. Pol. 50, 155–167. Czarnecka, A.M., Golik, P., Bartnik, E., 2006. Mitochondrial DNA mutations in human neoplasia. J. Appl. Genet. 41, 67–78. 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., 2005. 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. in press. Doshi, K.J., Cannone, J.J., Cobaugh, C.W., Gutell, R.R., 2004. Evaluation of the suitability of free-energy minimization using nearest-neighbor energy parameters for RNA secondary structure prediction. BMC Bioinformatics 5 Art. No. 105. Endler, J.A., 1986. Natural Selection in the Wild. Princeton Press, Princeton, NJ, USA. Fan, L., Sanschagrin, P.C., Kaguni, L.S., Kuhn, L.A., 1999. The accessory subunit of mtDNA polymerase shares structural homology with aminoacyl-tRNA synthetases: implications for a dual role as a primer recognition factor and processivity clamp. Proc. Natl. Acad. Sci. USA 96, 9527–9532. Feuermann, M., Francisci, S., Rinaldi, T., De Luca, C., Rohou, H., Frontali, L., Bolotin-Fukuhara, M., 2003. The yeast counterparts of human ‘MELAS’ mutations cause mitochondrial dysfunction that can be rescued by overexpression of the mitochondrial translation factor EF-Tu. EMBO Rep. 4, 53–58. Florentz, C., Sohm, B., Tryoen-Toth, P., Putz, J., Sissler, M., 2003. Human mitochondrial tRNAs in health and disease. Cell. Mol. Life Sci. 60, 1356–1375. Hixson, J.E., Wong, T.W., Clayton, D.A., 1986. Both the conserved stemloop and divergent 50 -flanking sequences are required for initiation at the human mitochondrial origin of light-strand replication. J. Biol. Chem. 261, 2384–2390. Jacobs, H.T., 2003. Disorders of mitochondrial protein synthesis. Hum. Mol. Genet. 12, R293–R301. James, A.M., Sheard, P.W., Wei, Y.H., Murphy, M.P., 1999. Decreased ATP synthesis is phenotypically expressed during increased energy demand in fibroblasts containing mitochondrial tRNA mutations— implications for neurodegenerative and mitochondrial DNA diseases. Eur. J. Biochem. 259, 462–469. Janssen, G.M.C., Maassen, J.A., van den Ouweland, J.M.W., 1999. The diabetes-associated 3243 mutation in the mitochondrial tRNA (Leu(UUR)) gene causes severe mitochondrial dysfunction without a strong decrease in protein synthesis rate. J. Biol. Chem. 274, 29744–29748. Johansson, J., Blomberg, C., 1995. A model of error propagation in the presence of an error-enhancing drug. J. Theor. Biol. 173, 1–13. Kaguni, L.S., 2004. DNA polymerase gamma, the mitochondrial replicase. Ann. Rev. Biochem. 73, 293–320. Kern, A.D., Kondrashov, F.A., 2004. Mechanisms and convergence of compensatory evolution in mammalian mitochondrial tRNAs. Nat.Genet. 36, 1207–1212. Kluge, A.G., Kerfoot, W.C., 1973. Predictability and regularity of character divergence. Am. Nat. 107, 426–442. Kollberg, G., Jansson, M., Perez-Bercoff, A., Melberg, A., Lindberg, C., Holme, E., Moslemi, A.R., Oldfors, A., 2005. Low frequency of mtDNA point mutations in patients with PEO associated with POLG1 mutations. Eur. J. Hum. Genet. 13, 463–469.
ARTICLE IN PRESS H. Seligmann et al. / Journal of Theoretical Biology 243 (2006) 375–385 Kondrashov, F.A., 2005. The analysis of monomer sequences in protein and tRNA and the manifestation of the compensation of pathogenic deviations in their evolution. Biofizika 50, 389–395. Kosel, S., Grasbon-Frodl, E.M., Hagenah, J.M., Graeber, M.B., Vieregge, P., 2000. Parkinson disease: analysis of mitochondrial DNA in monozygotic twins. Neurogenetics 2, 227–230. Krishnan, N.M., Seligmann, H., Raina, S.Z., Pollock, D.D., 2004a. Detecting gradients of asymmetry in site-specific substitutions in mitochondrial genomes. DNA Cell Biol. 23, 707–714. Krishnan, N.M., Seligmann, H., Raina, S.Z., Pollock, D.D., 2004b. Phylogenetic analyses detect site-specific perturbations in asymmetric mutation gradients. Curr. Computa. Mol. Biol. 2004, 266–267. Krishnan, N.M., Seligmann, H., Rao, B.J., 2005. Natural selection on cloverleaf-forming capacity fine-tunes the population frequency of human mitochondrial tRNA variants. In: Chatterji, D., Das, S., Nagaraja, V., Durga Rao, C., Vijayraghavan, U. (Eds.), 21st International tRNA Workshop, Bangalore, December 2005. tRNA2005, pp. 22–22. Limongelli, A., Schaefer, J., Jackson, S., Invernizzi, F., Kirino, Y., Suzuki, T., Reichmann, H., Zeviani, M., 2004. Variable penetrance of a familial progressive necrotising encephalopathy due to a novel tRNA(ile) homoplasmic mutation in the mitochondrial genome. J. Med. Genet. 41, 342–349. Lowe, T.M., Eddy, S.R., 1997. tRNAscan-SE: a program for improved detection of transfer RNA genes in genomic sequence. Nucleic Acids Res. 25, 955–964. Lucas, P., Lasserre, J.P., Plissonneau, J., Castroviejo, M., 2004. Absence of accessory subunit in the DNA polymerase gamma purified from yeast mitochondria. Mitochondrion 4, 13–20. McFarland, R., Elson, J.L., Taylor, R.W., Howell, N., Turnbull, D.M., 2004. Assigning pathogenicity to mitochondrial tRNA mutations: when ‘definitely maybe’ is not good enough. Trends Genet. 20, 591–596. Mirabella, M., Di Giovanni, S., Silvestri, G., Tonali, P., Servidei, S., 2000. Apoptosis in mitochondrial encephalomyopathies with mitochondrial DNA mutations: a potential pathogenic mechanism. Brain 123, 93–104. Moller, A.P., 1997. Developmental stability and fitness: a review. Am. Nat. 149, 916–932. Moller, A.P., 1999. Developmental stability is related to fitness. Am. Nat. 153, 556–560. Murdock, D.G., Christacos, N.C., Wallace, D.C., 2000. The age-related accumulation of a mitochondrial DNA control region mutation in muscle, but not brain, detected by a sensitive PNA-directed PCR clamping based method. Nucleic Acids Res. 28, 4350–4355. Ng, C.L., Waterman, D., Koonin, E.V., Antson, A.A., Ortiz-Lombardia, M., 2005. Crystal structure of Mil (Mth680): internal duplication and similarity between the Imp4/Brix domain and the anticodon-binding domain of class IIa aminoacyl-tRNA synthetases. EMBO Rep. 6, 140–146. Paul, R., Desnuelle, C., Pouget, J., Pellissier, J.F., Richelme, C., Monfort, M.F., Butori, C., Saunieres, A., Paquis-Flucklinger, V., 2000. Importance of searching for associated mitochondrial DNA alterations in patients with multiple deletions. Eur. J. Hum. Genet. 8, 331–338. Pierce, B.A., Mitton, J.B., 1979. Relationship of genetic-variation within and among populations—extension of the Kluge-Kerfoot phenomenon. Systemat. Zool. 28, 63–70.
385
Rodin, S., Ohno, S., Rodin, A., 1993. Transfer-RNAs with complementary anticodons—could they reflect early evolution of discriminative genetic-code adapters? Proc. Nat. Acad. Sci. USA 90, 4723–4727. Seligmann, H., 2006. Error propagation across levels of organization: from chemical stability of ribosomal RNA to developmental stability. J. Theor. Biol. in press. Seligmann, H., Amzallag, G.N., 2002. Chemical interactions between amino acid and RNA: multiplicity of the levels of specificity explains origin of the genetic code. Naturwissenschaften 89, 542–551. 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. B in press. Seligmann, H., Pollock, D.D., 2004. The ambush hypothesis: hidden stop codons prevent off frame gene reading. DNA Cell Biol. 23, 701–705. Seligmann, H., Krishnan, N.M., Rao, B.J., 2006. Possible multiple origins of replication in primate mitochondria: alternative role of tRNA sequences. J. Theor. Biol. in press. Shanske, S., Wong, L.J.C., 2004. Molecular analysis for mitochondrial DNA disorders. Mitochondrion 4, 403–415. Slack, K.E., Janke, A., Penny, D., Arnason, U., 2003. Two new avian mitochondrial genomes (penguin and goose) and a summary of bird and reptile mitogenomic features. Gene 302, 43–52. Smigrodzki, R., Parks, J., Parker, W.D., 2004. High frequency of mitochondrial complex I mutations in Parkinson’s disease and aging. Neurobiol. Aging 25, 1273–1281. Smith, P.M., Ross, G.F., Taylor, R.W., Turnbull, D.A., Lightowlers, R.N., 2004. Strategies for treating disorders of the mitochondrial genome. BBA-Bioenergetics 1659, 232–239 Sp. Iss. Taylor, R.W., Turnbull, D.M., 2005. Mitochondrial DNA mutations in human disease. Nature Rev. Genet. 6, 389–402. Thornhill, R., Moller, A.P., 1997. Developmental stability, disease and medicine. Biol. Rev. 72, 497–548. Torroni, A., Campos, Y., Rengo, C., Sellitto, D., Achilli, A., Magri, C., Semino, O., Garcia, A., Jara, P., Arenas, J., Scozzari, R., 2003. Mitochondrial DNA haplogroups do not play a role in the variable phenotypic presentation of the A3243G mutation. Am. J. Hum. Genet. 72, 1005–1012. Trifunovic, A., Wredenberg, A., Falkenberg, M., Spelbrink, J.N., Rovio, A.T., Bruder, C.E., Bohlooly, Y.M., Gidlof, S., Oldfors, A., Wibom, R., Tornell, J., Jacobs, H.T., Larsson, N.G., 2004. Premature ageing in mice expressing defective mitochondrial DNA polymerase. Nature 429, 417–423. Widmann, J., Di Giulio, M., Yarus, M., Knight, R., 2005. tRNA creation by hairpin duplication. J. Mol. Evol. 61, 524–530. Wolf, Y.I., Koonin, E.V., 2001. Origin of an animal mitochondrial DNA polymerase subunit via lineage-specific acquisition of a glycyl-tRNA synthetase from bacteria of the Thermus-Deinococcus group. Trends Genet. 17, 431–433. Wong, L.J.C., Boles, R.G., 2005. Mitochondrial DNA analysis in clinical laboratory diagnostics. Clin. Chim. Acta 354, 1–20. Xia, X., 2005. Mutation and selection on the anticodon of tRNA genes in vertebrate mitochondrial genomes. Gene 345, 13–20. Zuker, M., 2003. Mfold web server for nucleic acid folding and hybridization prediction. Nucleic Acids Res 31, 3406–3415.
