Introduction: Distinguishing Possible Mechanisms of Codon Reassignment Knight et al. 2001a Yokobori et al. 2001 Santos et al. 2004 Crick 1968 Sengupta and Higgs 2005 Codon disappearance (CD) mechanism Osawa and Jukes (1989 1995 Ambiguous intermediate (AI) mechanism Schultz and Yarus (1994 1996 Unassigned codon (UC) mechanism Sengupta and Higgs 2005 Yokobori et al. 2001 Compensatory change mechanism Kimura (1985) Higgs 1998 2000 Savill et al. 2001 Sengupta and Higgs 2005 2001a 2001b Swire et al. (2005) 1 2 3 4 Fig. 1. Phylogeny of fungi and related species derived from mitochondrial proteins.  Fig. 2. Phylogeny of plants and algae derived from mitochondrial proteins.  Fig. 3. Phylogeny of alveolates, stramenopiles, and haptophytes according to published sources.  Fig. 4. Phylogeny of metazoa according to published sources.  Reassignments That Can Be Explained by Codon Disappearance Reassignments of UGA from Stop to Trp Knight et al. (2001a) Monosiga Amoebidium 1 Lang et al. 2002 Acanthamoeba 1 Burger et al. (1995) Crinipellis Schizophyllum 1 Penicillium 1 Yarrowia Knight et al. (2001a) Acanthamoeba 1 Dictyostelium Schizosaccharomyces Chondrus Porphyra 2 Burger et al. (1999) Pedinomonas 2 Turmel et al. (1999) Phaecocystis Isochrisis Hayashi-Ishimaru et al. (1997) Emiliana Sanchez-Puerta et al. 2004 3 Paramecium, Tetrahymena Plasmodium Plasmodium Cafeteria 3 Skeletonema Thalassiosira 3 Ehara et al. (2000) Trypanosoma, Leishmania Inagaki et al. (1998) 1 Amoebidium Crinipellis Schizophyllum Alfonzo et al. 1999 Table 1. Codon usage data relevant to reassignments of Stop codons UGA and UAG  UGA is tRNA-Trp anticodon Codon usage UGA UAA UAG 1   Amoebidium castellanii Trp a 78 28 12   Dictyostelium discoideum Stop CCA 2 30 8   Monosiga brevicolis Trp TCA 124 28 4   Homo sapiens b Trp TCA 92 8 3   Rhizophidium Stop CCA 5 32 c   Spizellomyces punctatus Stop CCA 12 20 c   Hyaloraphidium curvatum Not used CCA 0 16 2   Monoblepharella JEL 15 Stop CCA 1 20 5   Harpochytrium JEL105 Not used CCA 0 12 2   Harpochytrium JEL94 Not used CCA 0 13 1   Allomyces macrogynus Not used CCA 0 16 14   Mortierella verticillata Rare CCA e 21 4   Rhizopus oryzae Not used CCA 0 20 4   Crinipellis perniciosa Trp a 124 73 16   Schizophyllum commune Trp a 25 19 1   Penicillium marneffei Trp TCA 61 16 1   Hypocrea jecorina Trp TCA 89 14 5   Schizosaccharomyces japonicus Not used CCA 0 6 1   Schizosaccharomyces octosporus Not used CCA 0 8 0   Schizosaccharomyces pombe Rare CCA e 7 0   Yarrowia lipolytica Trp TCA 57 12 2   Candida stellata Trp TCA 34 8 0   Candida albicans Trp TCA 49 5 8   Saccharomyces cerevisiae Trp TCA 124 19 0 2   Malawimonas jakobiformis Stop CCA 1 46 2   Cyanidioschyzon merolae Stop CCA 2 29 3   Chondrus crispus Trp TCA 101 22 3   Porphyra purpurea Trp TCA 117 27 4   Chaetosphaeridium globosum Stop CCA 7 30 9   Chara vulgaris Stop CCA 8 30 8   Prototheca wickerhamii Not used CCA 0 35 1   Pseudoendoclonium akinetum Stop CCA 11 40 21   Pedinomonas minor Trp TCA 62 11 0   Scenedesmus obliquus e Stop CCA 1 2 b   Chlamydomonas eugametos Not used CCA 0 12 2   Chlamydomonas reinhardtii Not used CCA 0 6 2 3   Emiliana huxleyi Trp UCA 73 19 2   Rhodomonas salina Stop CCA 1 34 9   Naegleria gruberi Not used CCA 0 37 9   Plasmodium reichenowi Not used None 0 3 0   Plasmodium falciparum Not used None 0 3 0   Paramecium aurelia Trp TCA 83 29 17   Tetrahymena pyriformis Trp TCA 228 44 0   Tetrahymena thermophila Trp TCA 228 45 0   Caferteria roenbergensis Trp TCA 190 32 2   Phytophthora infestans Stop CCA 1 39 0   Saprolegnia ferax Not used CCA 0 42 1   Chrysodidymus synuroides Not used CCA 0 34 3   Ochramonas danica Not used CCA 0 30 14   Laminaria digitata Stop CCA 4 29 6   Pylaiella littoralis Stop CCA 7 38 7 a b Homo sapiens c d e S. obliquus Sengupta and Higgs 2005 1 Allomyces Rhizopus Cyanidioschyzon Chondrus Porphyra Scenedesmus Chlamydomonas Pedinomonas Plasmodium Paramecium Tetrahymena Plasmodium A mutation pressure from GC to AU, which is implicated in the disappearance of UGA, will also tend to cause rapid mutations from UGG Trp codons to UGA after the Trp tRNA gains the ability to decode UGA. This is one reason that reversal of the change is unlikely. There are often around 100 UGAs in genomes where the reassignment has occurred, and it would be very difficult for this large number to disappear by chance because this would act against the mutation pressure. A second reason is that the reassignment to Trp would be associated with the loss of function of the release factor that originally interacted with the UGA. A reversal would also require regaining of the function of the release factor. 1 Dictyostelium Schizosaccharomyces. 1 Seif et al. (2005) Mortierella Schizosaccharomyces The Probability of Disappearance of UGA Codons P dis Swire et al. (2005) A ,  C ,  G U f UGA f UAA f UAG f UGA f UAA f UAG \documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym} 
\usepackage{amsfonts} 
\usepackage{amssymb} 
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$ {f_{UGA} \over f_{UAA}} = {f_{UAG} \over f_{UAA}} = {\pi_{G} \over \pi _{A}}. $$\end{document} \documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym} 
\usepackage{amsfonts} 
\usepackage{amssymb} 
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$ f_{UGA} = f_{UAG} = \pi _{G} / ( 2\pi _{G} + \pi _{A} ) . $\end{document} C  =  G A  =  U Urbina et al. 2006 N stop \documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym} 
\usepackage{amsfonts} 
\usepackage{amssymb} 
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$ P_{dis} = (1 - f_{UGA})^{N_{stop}} $\end{document} N stop f UGA 2 N stop 1 N stop Monosiga P dis Amoebidium G A Tetrahymena P dis Paramecium P dis P dis Schizophyllum Crinipellis N stop N stop Crinipellis Moniliophthora perniciosa N stop P dis Table 2. Estimates of stop codon disappearance probabilities Case Species N stop %G %A f codon P dis UGA CD i Monosiga brevicolis 32 2.85 47.52 5.36 0.17 ii Acanthamoeba castellanii 40 9.89 26.35 21.44 6.4E-05 iii Crinipellis perniciosa 89 9.19 32.84 17.94 2.3E-08 iii Schizophyllum commune 20 3.48 42.87 6.99 0.23 iv Penicillium marneffei 17 2.38 35.64 5.89 0.36 iv Hypocrea jecorina 19 4.68 40.14 9.46 0.15 v Yarrowia lipolytica 14 1.55 49.01 2.97 0.65 v Candida stellata 8 0.94 41.57 2.17 0.84 vi Chondrus crispus 25 6.08 36.61 12.46 3.6E-02 vi Porphyra purpurea 31 10.98 35.80 19.01 1.4E-03 vii Pedinomonas minor 11 1.60 21.31 6.52 0.48 viii Emiliana huxleyi 21 9.26 36.50 16.83 2.1E-02 ix Paramecium aurelia 46 14.87 19.60 30.14 6.8E-08 ix Tetrahymena pyriformis 44 3.22 42.27 6.62 4.9E-02 ix Tetrahymena thermophila 45 2.63 40.46 5.75 7.0E-02 x Cafeteria roenbergensis 32 6.17 32.56 13.74 6.6E-03 UAG CD i Rhizophidium 37 3.33 34.05 8.18 4.3E-02 i Spizellomyces punctatus 32 6.03 30.69 14.10 7.7E-03 ii Scenedesmus obliquus 3 6.64 35.73 13.55 0.65 P dis P dis −5 Acathamoeba P dis −5 P dis −11 P dis −2 Chondrus P dis −3 Porphyra P dis 2 Reassignments of UAG Stop Codons Rhyzophidium Spizellomyces 1 Laforest et al. 1997 Scenedesmus 2 Hayashi-Ishimaru et al. 1996 Kück et al. 2000 1 Hyaloraphidium Pedinomonas Scenedesmus Asahara et al. 1993 Scenedesmus Kück et al. 2000 Scenedesmus Hayashi-Ishimaru et al. 1996 2 Scenedesmus Scenedesmus 2 Sense Codon Reassignments Linked to Codon Disappearance 3 3 Table 3. Codon usage in some Fungi lineages: intronic ORFs excluded (E) or included (I)  Leu codons Arg codons Ile and Met Codons Frequency at FFD sites CUN UUR CGN AGR AUU AUC AUA AUG %U %C %A %G S. japonicus 79 198 7 32 133 40 32 48 76.0 3.4 19.2 1.3 S. octosporus 68 236 2 34 161 34 d 57 61.1 1.9 34.8 2.2 S. pombe 53 192 7 33 113 39 49 51 56.4 1.3 40.5 1.8 Y. lipolytica 44 618 c 75 174 87 277 119 48.3 1.1 49.0 1.5 C. stellata 3 279 12 29 123 8 156 54 57.3 0.2 41.6 0.9 C. albicans 132 397 47 26 119 81 229 100 55.4 4.8 37.7 2.1 C. parapsilosis 66 547 39 45 303 32 193 117 68.6 0.7 29.8 0.9 C. parapsilosis 137 728 60 102 410 49 299 143 65.3 3.0 29.0 2.7 P. canadensis 25 714 18 67 274 18 562 105 49.2 0.7 49.4 0.7 P. canadensis 27 746 20 74 298 20 586 109 50.0 0.9 48.4 0.6 A. gossypii a 291 c 40 215 7 d,e 34 57.2 0.0 42.8 0.0 K. lactis b 286 c 48 213 16 d 63 44.0 1.6 53.3 1.1 K. lactis b 312 c 55 256 16 d 65 43.4 2.6 52.1 1.9 K. thermotolerans a 304 2 44 204 17 d 56 47.7 0.9 51.0 0.4 K. thermotolerans a 440 10 72 298 23 d 78 48.3 2.1 47.6 2.0 C. glabrata a 294 c 45 207 21 d,e 73 46.7 0.6 52.0 0.7 C. glabrata a 415 c 60 318 25 d,e 78 48.6 0.9 49.8 0.7 S. cerevisiae a 333 7 49 239 31 d,e 73 48.6 2.3 47.1 2.0 S. castellii a 274 c 40 203 7 d,e 56 47.5 1.6 49.9 1.0 S. servazzii a 300 c 46 218 11 d,e 70 37.2 0.9 59.5 2.4 a b c d e 3 P. canadensis 3 Sibler et al. 1981 Osawa et al. 1990 K. lactis Asahara et.al. 1993 3 A. gossypii 1 Sengupta and Higgs 2005 P. canadensis K. thermotolerans S. castellii K. thermotolerans S. castellii P. canadensis Sibler et al. (1981) Osawa et al. (1990) One possibility that cannot entirely be ruled out from the codon usage data is that the changes in the tRNA-Leu(UAG) occurred when the CUN codons were very rare but not entirely absent. The changes might have been such as to immediately cause the tRNA to be charged by Thr, or might have happened more gradually, so that the same tRNA could be ambiguously charged by Leu and Thr during the changeover period. We would then have to count this as an example of the AI mechanism. Nevertheless, the low CUN number resulting from mutation pressure is clearly a major factor in this case, and we therefore feel comfortable in classifying it under the CD mechanism. The tRNA gene has undergone considerable modification, including an unusual insertion that makes the anticodon loop larger than the standard seven bases. In our opinion, it is unlikely that such a large change could have happened while the tRNA remained simultaneously functional for both amino acids. 3 3 C. glabrata S. cerevisiae K. thermotolerans S. cerevisiae K. thermotolerans P. canadensis 1 1 A. gossypii S. cerevisiae K. thermotolerans K. lactis 6 3 Y. lipolytica Kerscher et al. 2001 S. pombe C. stellata Y. lipolytica The Probability of Disappearance of CUN and CGN Codons \documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym} 
\usepackage{amsfonts} 
\usepackage{amssymb} 
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$ \eqalign {{f_{CUX} \over f_{CUG}} & = {\pi _{X} \over \pi _{G}} \quad ({\rm for} \,X = A,\,C\,or\,U), \cr {f_{UUG} \over f_{CUG}} &= {f_{UUA} \over f_{CUA}} = {\pi _{U} \over \pi _{C}}} $$\end{document} \documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym} 
\usepackage{amsfonts} 
\usepackage{amssymb} 
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$ f_{CUN} = \pi _{C} / (\pi _{C} + \pi_{U} (\pi _{A} + \pi_{G})) $$\end{document} \documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym} 
\usepackage{amsfonts} 
\usepackage{amssymb} 
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$ P_{dis} = (1 - f_{CUN})^{N_{Leu}} $\end{document} N Leu 3 A. gossypii f CUN P dis A. gossypii 3 P dis Bullerwell et al. 2003 Talla et al. 2005 3 A. gossypii K. lactis K. thermotolerans S. cerevisiae C. stellata P dis C. glabrata 3 f CUN P dis 415 −7 C. glabrata A. gossypii 3 Swire et al. (2005) 2005 P dis −427 P dis \documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym} 
\usepackage{amsfonts} 
\usepackage{amssymb} 
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$ f_{CGN} = \pi _{C} / (\pi _{C} + \pi _{A} (\pi _{A} + \pi _{G})) $\end{document} \documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym} 
\usepackage{amsfonts} 
\usepackage{amssymb} 
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$ P_{dis} = (1 - f_{CGN})^{N_{Arg}} $\end{document} A. gossypii N Arg N Leu Y. lipolytica P dis 75 Osawa and Jukes (1989) Reassignments That Cannot Be Explained by Codon Disappearance Reassignment of AUA from Ile to Met Is Initiated by Loss of a tRNA-Ile Gene E. coli Muramatsu et al. 1988 2 Tomita et al. 1999b In order to locate and understand the cases of AUA codon reassignment in mitochondria, it is crucial to establish which genomes contain the tRNA-Ile(CAU) gene. This is complicated by the fact that some genomes contain three tRNAs with CAU anticodons (a Met initiator, a Met elongator. and an Ile), and the Ile tRNA is often mistakenly annotated as Met. To uncover misannotations, we constructed a phylogeny of all the tRNAs with CAU anticodon from all the fungi genomes in our data set. The genes fell into three groups that could be reliably identified. Supplementary Table S3 lists the position of the three tRNAs in each genome and, hence, shows in which genomes there have been gene deletions. A. gossypii 3 A. gossypii 1 C. albicans, C. parapsiplopsis P. canadensis P. canadensis Yokobori et al. (2001) Kluyveromyces Talla et al. (2005) K. thermotolerans VAR1 K. lactis K. thermotolerans C. glabrata Koszul et al. (2003) C. glabrata S. cerevisiae C. glabrata S. cerevisiae C. glabrata A. gossypii Saccharomyces \documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym} 
\usepackage{amsfonts} 
\usepackage{amssymb} 
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$ f_{AUA} = \pi _{A}/ (\pi _{U} + \pi _{C} + \pi _{A}) $\end{document} \documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym} 
\usepackage{amsfonts} 
\usepackage{amssymb} 
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$ P_{dis} = (1 - f_{AUA})^{N_{Ile}} $\end{document} C. glabrata (I) P dis 343 −104 Kluyveromyces Saccharomyces Schizosaccharomyces octosporus S. pombe S. japonicus 3 S. octosporus Schizosaccharomyces S. octosporus Bullerwell et al. (2003) S. octosporus C. albicans, C. parapsilosis Y. lipolytica Amoebidium Monosiga Burger et al. 2003 Lavrov et al. 2005 4 Telford et al. 2000 Yokobori et al. (2001) Drosophila 5 Tomita et al. 1999b 6 N 6 Kuchino et al. 1987 after 4 4 Table 4. Codon usage in some metazoan lineages  Ile and Met codons Ser and Arg codons Asn and Lys codons Frequency at FFD sites AUU AUC AUA AUG AGU AGC AGA AGG AAU AAC AAA AAG %U %C %A %G Axinella corrugate 222 45 205 130 74 23 b 17 107 14  86 28 35.9 11.7 30.6 21.8 Geodia neptuni 207 18 218 118 79 15 b 15 104 18  89 24 40.7 7.1 35.7 16.4 Metridium senile 190 59 110 138 58 21 b 12 84 31  79 22 42.5 17.5 25.0 15.0 Acropora tenuis 182 44 115 113 68 11 b 16 90 15  72 31 43.5 11.5 17.6 27.4 Limulus polyphemus 241 107 a 43 22 8  66 12 103 44  66 19 35.7 17.1 38.8 8.4 Daphnia pulex 187 93 a 52 49 20  68 e 75 43  48 41 36.2 20.2 24.9 18.7 Drosophila melanogaster 355 16 a 13 30 0  74 e 193 10  81 5 50.5 2.4 43.6 3.6 Caenorhabditis elegans 257 23 a 44 61 6 126 39 139 12  95 14 50.2 4.1 37.5 8.2 Trichinella spriralis 136 114 a 96 46 20  71 34 68 80  65 34 29.6 15.2 42.2 13.0 Katharina tunicate 226 49 a 45 46 36  88 35 134 47  79 30 43.1 12.3 32.0 12.5 Lumbricus terrestris 194 105 a 64 20 14  63 13 70 66  70 22 28.2 24.3 36.9 10.5 Terebratulina retusa 137 147 a 34 6 23  78 18 46 67  80 19 22.6 35.3 34.4 7.7 Fasciola hepatica 119 8  39 96 92 8  7 41 50 2 f 44 68.2 5.4 6.9 19.4 Schistosoma mansoni 134 15 155 123 102 10  62 50 54 4 f 60 52.8 4.3 23.3 19.5 Taenia crassiceps 174 2 126 101 104 1  43 29 89 5 f 49 64.3 1.4 19.9 14.5 Paracentrotus lividus 146 53 165 102 11 14  66 15 44 50 f 54 25.0 22.9 42.6 9.6 Asterina pectinifera 129 66 178 78 22 23  54 19 40 58 f 48 28.8 25.1 35.4 10.6 Balanoglossus carnosus 97 148  69 77 18 35  19 e 20 98  0 45 22.4 37.8 29.1 10.6 Saccoglossus kowalewskii 185 72  96 70 34 22  4 e 60 68  47 12 40.7 23.0 31.2 5.1 Halocynthia roretzi 170 12 a 60 125 4 c 88 78 5  29 42 53.0 3.6 26.1 17.2 Cionia intestinalis 297 19 a 34 73 7 c 37 143 16 103 22 56.5 4.8 33.3 5.4 Branchiostoma lanceolatum 186 47 a 63 78 30  12 e 86 26  40 34 37.7 9.1 32.8 20.5 Branchiostoma floridae 187 46 a 63 73 33  12 e 87 24  40 33 38.5 8.5 33.0 20.0 Epigonichthys maldivensis 159 70 a 59 49 35  31 e 70 40  37 35 34.7 13.6 29.4 22.3 Myxine glutinosa 210 154 a 44 23 28 d  1 69 77 102 15 29.2 27.4 34.9 8.5 Homo sapiens 124 196 a 40 14 39 d  1 32 132  85 10 14.5 40.4 38.7 6.4 a b c d e f 2 Knight et al. [2001a] 1 Yokobori et al. [2001] Telford et al. (2000) Ehara et al. 1997 Reassignments Involving the AGR Block Are Initiated by Loss of a tRNA-Arg Gene 4 4 Yokobori et al. 2001 Drosophila Drosophila Drosophila 4 Daphnia Limulus 4 C. elegans T. spiralis 4 Matsuyama et al. 1998 Tomita et al. 1998 7 Gissi et al. 2004 4 Spruyt et al. (1998) Branchiostoma lanceolatum B. lanceolatum B. floridae Boore et al. 1999 Epigonichthys maldivensis Nohara et al. 2005 B. lanceolatum B. floridae Knight et al. 2001a Reassignment of AAA from Lys to Asn May Proceed Via an Ambiguous Intermediate Balanoglossus 4 Saccoglossus Morris et al. 1999 Tomita et al. 1999a Yokobori et al. 2001 Castrasena et al. 1998 Balanoglossus Castrasena et al. 1998 Balanoglossus Saccoglossus Alternatively, it is possible that the changes occurred via the UC mechanism. If the tRNA-Lys mutation happened first, this would leave the tRNA-Asn able to pair inefficiently with the AAA, and subsequent changes to the tRNA-Asn would allow it to recognize AAA more easily. This argument is analogous to the case of the AUA:Ile to Met change or the AGR:Arg to Ser. However, in those two cases the loss of function is the deletion of the original tRNA for the codon in question, which is irreversible. In the AAA case, the loss of function is just the mutation of U to C in the anticodon. This would be a deleterious mutation that could easily reverse, so it is difficult to see why the change would go to fixation. This makes the UC mechanism seem less plausible for AAA. We conclude, therefore, that of all the reassignments considered in this paper, these two examples of AAA reassignment are the best candidates for the AI mechanism. Balanoglossus Balanoglossus Schizosaccharomyces octosporus S. octosporus Balanoglossus Balanoglossus Yokobori et al. 2001 Balanoglossus Introduction of New Stop Codons Thraustochytrium aureum 3 T. aureum Scenedesmus obliquus Kück et al. 2000 S. obliquus 5 S. obliquus P. minor C. eugametos C. reinhardtii Table 5. Codon usage in some green algae  Leu codons Ile Ser codons Thr codons Glu codons Arg codons FFD sites UUA UUG AUA UCA UCG ACA ACG GAA GAG AGA AGG %A %G Pedinomonas minor 411 84 115 53 8 27 0 58 11 35 2 21.3 1.6 Scenedesmus obliquus 0 384 0 13 0 111 5 73 14 0 93 35.7 6.6 Chlamydomonas eugametos 353 38 176 80 3 98 6 53 4 23 0 48.5 11.7 Chlamydomonas reinhardtii 0 251 0 0 0 0 0 0 55 0 0 26.1 3.7 Note S. obliquus C. reinhardtii S. obliquus Kück et al. (2000) Nedelcu et al. (2000) 5 S. obliquus T. aureum Nedelcu et al. (2000) would S. obliquus S. obliquus 5 S. obliquus Importation of tRNAs from the Cytoplasm to the Mitochondria Acanthamoeba Dictyostelium Pedinomonas Chlamydomonas Naegleria Acanthamoeba Dictyostelium Naegleria In all the above groups, at least four tRNAs are absent, and it is clear that the remaining tRNAs are insufficient, thus there is no doubt that import must occur. However, there are also many other groups where only a small number of tRNAs are missing. We have already discussed several cases above where the loss of a tRNA leads to the corresponding codon becoming unassigned or reassigned to a new amino acid. These cases leave clear signals in the codon usage patterns. This means that there is no import of a replacement tRNA from the cytoplasm in the cases discussed in previous sections. In contrast, we now discuss several cases where a small number of tRNAs are absent from the mitochondrial genomes but there is no change in the genetic code or unusual codon usage pattern. In these cases it appears that import of one or a few specific tRNAs is occurring. Reclinomonas Mesostigma, Scenedesmus Saks et al. 1998 Cafeteria, Thraustochytrium, Chrysodidymus, Ochromonas, Laminaria, Pylaiella Phytophthora Saprolegnia 3 1 Supplementary Table S1 lists several other genomes where just one or two tRNAs are missing. It is possible that tRNA import occurs or that there are unknown tRNA modifications that compensate for these individual losses. However, we consider this as somewhat uncertain in cases where only one genome is known with the missing gene. This could simply be a failure to locate the gene on the genome or a problem of misannotation. Chlamydomonas reinhardtii C. eugametos 5 Denovan-Wright et al. 1998 C. eugametos C. reinhardtii C. reinhardtii C. eugametos C. reinhardtii Scenedesmus obliquus 2 S. obliquus S. obliquus C. reinhardtii C. reinhardtii S. obliquus C. eugametos Discussion Distinguishing the Mechanisms Sengupta and Higgs 2005 In cases where we have determined that the codon did not disappear, the central question is to distinguish between UC and AI mechanisms. The UC mechanism is defined by the fact that the loss occurs before the gain, whereas in the AI mechanism, the gain occurs before the loss. Although this distinction is obvious in simulations, it can be more difficult in real cases, as we did not observe the order of events. Nevertheless, we gave several examples where we feel a reliable classification of UC can be made. In particular, the reassignments of AUA from Ile to Met and AGR from Arg to Ser are both associated with the deletion of a tRNA from the genome. The loss event is an irreversible gene deletion that leaves the organism in a deleterious state. The gain event is then positively selected in response to this loss. In contrast, in the case of the reassignment of AAA from Lys to Asn, the loss event is a mutation in the anticodon of the tRNA-Lys. This is reversible, so selection would favor mutation back to the original state, which seems more likely to occur than making a codon reassignment. The gain in this case is probably due to the cessation of the Q modification process, which could occur due to deleterious mutations in the modifying enzyme or disruption of the transport of the enzyme to the mitochondria. These changes seem less easily reversible than a single mutation in the anticodon. Therefore we argue that the gain occurred first and that the loss occurred in response. This makes the AAA reassignment the most likely example of the AI mechanism. The cases where both loss and gain seem easily reversible are the most difficult to classify, and both AI and UC scenarios can be proposed. (These arguments are based on the assumption that the ambiguous and unassigned states are deleterious with respect to the original code. We deal with the alternative suggestion that ambiguous translation can be positively selected later in the Discussion.) We just argued that the AUA Ile-to-Met case is initiated by deletion of the tRNA-Ile(CAU) and that this change can be reliably classed as UC. Nevertheless, it is possible that ambiguous translation plays a role in this reassignment in the following sense. After the deletion, the AUA codon would be translated inefficiently by the tRNA-Ile(GAU), but it is possible that there might also be some interaction with the tRNA-Met(CAU). When the tRNA-Met becomes modified, it definitely gains the ability to translate AUA. So, ambiguous translation of AUA as both Ile and Met at some points in this process is not unlikely. Before the original tRNA-Ile(CAU) is deleted, any slight ability of the other two tRNAs to translate AUA would be irrelevant. Also, after the modification of the tRNA-Met, any slight ability of the tRNA-Ile(GAU) to translate AUA would be irrelevant. The ambiguity, if it existed, would only be relevant when there is a competition between two poorly adapted tRNAs. Despite all this, the gain in function of the tRNA-Met only occurs after the loss of the tRNA-Ile(CAU), so this reassignment counts as UC not AI. We recommend the use of the term AI only when ambiguous translation occurs as a result of a gain occurring before a loss, and where competition occurs between two well-adapted tRNAs. after 6 Table 6. Summary of mechanisms of codon reassignment in mitochondria Codon reassignment No. of times Can this be explained by GC→AU mutation pressure? Change in no. of tRNAs Is mispairing important? Reassignment mechanism UAG: Stop → Leu 2 Yes. G → A at 3rd position. +1 No CD UAG: Stop → Ala 1 Yes. G → A at 3rd position +1 No CD UGA: Stop → Trp 12 Yes. G → A at 2nd position. 0 Possibly. CA at 3rd position. CD CGU/CGC/CGA/CGC: Arg → unassigned 5 Yes. C → A at 1st position. –1 No CD CUU/CUC/CUA/CUG: Leu → Thr 1 Yes. C → U at 1st position. 0 No CD CUU/CUC/CUA/CUG: Thr → unassigned 1 No –1 No CD AUA: Ile → Met or unassigned a No –1 Yes. GA at 3rd position. UC AAA: Lys → Asn 2 No 0 Yes. GA at 3rd position. AI AAA: Lys → unassigned 1 No 0 Possibly. GA at 3rd position. UC or AI AGA/AGG: Arg → Ser 1 No –1 Yes. GA at 3rd position. UC AGA/AGG: Ser → Stop 1 No 0 No b AGA/AGG: Ser → Gly 1 No +1 No b UUA: Leu → Stop 1 No 0 No UC or AI UCA: Ser → Stop 1 No 0 No UC or AI Note a S. octosporus b A slow response to a tRNA deletion occurs for both the tRNA-Ile(CAU) and the tRNA-Arg (UCU), which are deleted at roughly the same time in the ancestral Bilateria. This would leave AUA inefficiently translated as Ile and AGR inefficiently translated as Ser. Both these situations remain in some phyla today. However, in other groups, AUA has been captured by Met, and AGR either has become a useful Ser codon (due to mutation of the tRNA-Ser) or has been reassigned to Gly or Stop. These secondary changes may have occurred considerably later than the deletions of the original two tRNAs. compensatory change Kimura 1985 Higgs 1998 2000 Sengupta and Higgs 2005 Comparison with Previous Surveys Knight et al. (2001b) Swire et al. (2005) Knight et al. (2001b) 6 Knight et al. (2001b) Swire et al. (2005) Schultz and Yarus (1994) Knight et al. (2001b) 6 3 4 Bacillus subtilis Lovett et al. 1991 Matsugi et al. 1998 6 S. octosporus Kluyveromyces 3 4 Andersson and Kurland 1991 1995 Sengupta and Higgs 2005 6 6 Swire et al. (2005) Sengupta and Higgs 2005 Santos et al. 1999 Swire et al. (2005) Sengupta and Higgs (2005) Swire et al. (2005) Candida Santos et al. 1999 2004 Massey et al. (2006) Candida Kim et al. (2000) E. coli Freeland et al. 2003 Ardell and Sella 2002 tRNA Evolution 6 Blanchard and Lynch 2000 Trypanosoma brucei Leishmania tarentolae Simpson et al. 1989 Hancock and Hajduk 1990 Schneider and Marechal-Drouard 2000 Giegé et al. 1998 Laforest et al. 1997 Conclusions We have given arguments above as to which mechanisms seem most likely in each of the codon reassignments in mitochondrial genomes. We have shown that the many reassignments of stop codons to sense codons are readily explained by CD, given the biased base composition of most mitochondrial genomes and the small total number of occurrences of stop codons in these genomes. Disappearance of sense codons is more difficult because the total number of codons for the corresponding amino acid is large. A very strong mutational bias is required for sense codons to disappear. However, in one group of yeast species, an extreme bias against C does exist, and we argue that the reassignment of CUN and CGN codons in these species is attributable to CD. In the other examples of sense codon reassignments, the mutational bias is in the wrong direction for causing CD, and the probability of disappearance is negligible. Where the codon does not disappear, we have emphasized the important distinction between the UC and AI mechanisms. The case for the UC mechanism is most clear when the reassignment is associated with a tRNA deletion. We then argue that the deletion initiated the process and the codon reassignment occurred as a response to this. The UC mechanism does not rely on selection for reducing genome length, but if such selection were significant, this would increase the likelihood of this mechanism. Many nonessential tRNAs have been deleted during mitochondrial genome evolution and these did not initiate codon reassignments because the original code was still functional after the deletion. However, this makes it clear that chance tRNA deletion is a relatively common event. We also observed several cases where a good argument can be made for the AI mechanism, i.e., where the reassignment arose because the codon first became ambiguous. In other cases, scenarios for both AI and UC seemed equally plausible, and it is difficult to distinguish them after the event. We see these genetic code changes as chance events, rather than as changes governed by positive selection. Disappearance is a chance event that occurs under drift when there is strong mutational bias in base frequencies. If a change in tRNAs or release factors happens to occur while a codon is absent, then a codon reassignment can occur. However, this is a chance event: the codon frequency could drift back to a higher level without any reassignment occurring. Our interpretation of reassignments via UC and AI mechanisms is that they too are initiated by chance events, such as the deletion of a tRNA gene or a change in the process of base modification in an anticodon. These changes are probably slightly deleterious, but efficient functioning of the translation system can be restored by making the codon reassignment. The view that AI states are driven by positive selection seems unlikely to us at present. The origin of the canonical code is outside the scope of the present paper. However, we emphasize that the situation in codon reassignments in modern organisms is different from that during the early evolution of the canonical code, where positive selection probably had an important role. We conclude that our gain-loss framework is suitable as a description of the real codon reassignment events. It emphasizes that there are several mechanisms that are alternatives within a larger picture, and that it 
s not always profitable to discuss these mechanisms as though they were mutually exclusive. These mechanisms can and do occur in nature, and one mechanism is not sufficient to explain all cases. Electronic Supplementary Material Supplementary material