Proc. Nati. Acad. Sci. USA Vol. 89, pp. 2742-2746, April 1992 Evolution Sequence of Prochloron didemni atpBE and the inference of chloroplast origins (endosymbiosis/Prochlorophyta/phylogeny/ATP synthase/cyanobacteria) P. J. LOCKHART*, T. J. BEANLANDt, C. J. HOWEtt, AND A. W. D. LARKUM* *School of Biological Sciences, Macleay Building A12, University of Sydney, NSW 2006, Sydney, Australia; and tDepartment of Biochemistry, University of Cambridge, Tennis Court Road, Cambridge CB2 1QW U.K. Communicated by Hans Kornberg, December 2, 1991 s The prochlorophytes, oxygenic photosynABSTRACT thetic prokaryotes containing chlorophylls a and b, have been put forward as descended from the organisms that gave rise to chloroplasts of green plants and algae by endosymbiosis, although this has always been controversial. To assess the phylogenetic position of the prochlorophyte Prochloron didemni, we have cloned and sequenced its atpBE genes. Phylogenetic inference under a range of models gives moderate to strong support for a cyanobacterial grouping rather than a chloroplast one. Possible systematic errors in this and previous analyses of prochlorophyte sequences are discussed. w S W W A 3. A P S C P W S A C C p W A ~~~P C w A C C S W 6. 5. 4. A C P C W A S P W A range of evidence suggests that plastids originated by the uptake of oxygenic photosynthetic bacteria by nonphotosynthetic hosts (1, 2), although which extant bacteria are the descendants of the protoendosymbionts for the various plastid groups remains unclear (3). For green chloroplasts, which contain chlorophylls a and b and lack phycobiliproteins, prochlorophytes (oxygenic photosynthetic bacteria with the same pigments) have been suggested (4-6). Ultrastructural and biochemical data from prochlorophytes (4, 7, 8) are inconclusive, the only evidence even that Prochlorales is a monophyletic order coming from the immunochemical crossreactivity of the chlorophyll-a/b-binding proteins ofProchloron didemni (P. didemni) and Prochlorothrix hollandica (Px. hollandica) (9). Sequence data are available for only two of the three prochlorophyte species. For P. didemni, analyses of a 16S rRNA RNase T1 oligonucleotide catalogue (10) and complete 5S rRNA sequence (11) apparently rule out chloroplast ancestry, although in neither case is such a conclusion statistically excluded (12-14). In contrast, analysis of partial 16S rRNA (15) and rbcLS (16) sequences from Px. hollandica leads to robust trees placing it among the cyanobacteria. The original analysis of Px. hollandica D1 (psbA) sequences (17, 18) proposed a common ancestry with chloroplasts to the exclusion of cyanobacteria, but it lacked any estimation of tree robustness and did not utilize D2 (psbD) as an outgroup. Use of D2 in this way (19) suggests that the C-terminal 7-amino acid gap common to green chloroplast and Px. hollandica D1 proteins that was claimed to support a relationship between these groups (17, 18) may in fact be a primitive feature. Maximum likelihood analysis of Px. hollandica psbA data (20) by using an explicit model of amino acid substitution indicates a robust cyanobacterial affinity. We report here the sequence of P. didemni atpB and atpE genes.§ Adopting a hypothesis-testing approach to inference (20, 21), we find some strong evidence (P = 0.01) for a cyanobacterial affinity. In common with the analyses discussed, this conclusion is, however, still critically dependent on the inference models employed. FIG. 1. Unrooted topologies for step 1 testing of the relationships between the oxygenic taxa. A, Anabaena sp. PCC7120; S, Synechococcus sp. PCC6301; P, P. didemni; W, wheat (Triticum aestivum) chloroplast; C, C. reinhardtii chloroplast. All the trees are unrooted; branch lengths are arbitrary. The publication costs of this article were defrayed in part by page charge payment. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. §1734 solely to indicate this fact. *To whom reprint requests should be addressed. §The sequence reported in this paper has been deposited in the A P S 9. 8. 7. S P W S A _ S P _W C 10 C P S A P P S W C A p A S p w 12. 5 / A A 15. 14. ~~C S W \ 13. A C C C W P MATERIALS AND METHODS Cloning and Sequencing of P. didemni atpBE. P. didemni within its host ascidian Lissoclinum patella was collected from shallow water around Heron and One Tree Islands on the Great Barrier Reef, north of Gladstone, Australia. High molecular weight genomic DNA was extracted and a genomic library was constructed in A EMBL3 as described elsewhere (14). Clones hybridizing with a probe containing the Zea mays chloroplast rbcL and partial atpB genes (22) were sequenced, subclones being isolated using a gene-internal probe for the Synechococcus sp. PCC6301 atpB gene (a gift from A. L. Cozens, Laboratory of Molecular Biology, Cambridge) or synthetic oligonucleotide probes (Oligonucleotide Synthesizing Facilities, Dept. of Biochemistry, University of Cambridge, and School of Biological Sciences, Macquarie GenBank data base (accession 2742 no. M86384). Evolution: Lockhart 1 et Proc. Natl. Acad. Sci. USA 89 (1992) al. 60 ATGGTAGCGA CAACAGAAAC AACAAACATT GGTAAAATTA CCCAAATCAT CGGCCCTGTA G P V T Q I I G K I T N I T T E T M V A 2743 1080 1021 GTGCTGTCTC GGGGTTTGGC TTCTAAGGGT ATTTATCCTG CTGTAGATCC TTTAGACTCC L D A V D P S Y P S K A G I R L G V L S 1140 1081 120 61 GTGGATGCGG AGTTTCCATC TGGCAAAATG CCCCGAATCT ACAATGCCTT GAGAGTTGAA R V E Y N A L P R I G K M E F P S V D A ACCAGCACCA TGTTACAGGC GGGAATTGTG GGTGAAGACC ACTACAATAC CGCTCGTGCA A R A H Y N T G E D M L Q A G I V T S T 180 121 GGCAAAAATG CCGCCGGACA AGATGTAGCC GTAACCTGCG AGGTGCAGCA GTTGCTGGGA G L L E V D V A V T C Q A A Q G Q G K N 1200 1141 GTGCAGTCTA CCTTGCAGCG CTATAAAGAA CTGCAAGATA TTATTGCCAT TTTGGGTCTG L G L L Q D I I A I V Q S Y K E T L Q R 240 181 GACAACCAAG TACGGGCTGT TTCCATGAGC AGCACGGACG GTCTGGTGCG GGGAATGGAA G L V R G M E S T D S M V R A V S D N Q 1260 1201 GATGAATTGT CGGAAGAAGA CCGCTTGATA GTAGATCGGG CTCGGAAGGT GGAGCGTTTC E R F A R K V V D R S E E D R L I D E L 300 241 ATTACCGATA CTGGCGCACC CATTAACGTT CCTGTGGGCA AGGCTACCCT GGGTCGGATT K A T L G R I I N V P V G T G A P I T D 1320 1261 TTGTCTCAGC CTTTCTTTGT GGCGGAAGTA TTTACTGGCG CACCTGGCAA GTACGTTTCT K Y V A P S V F T G Q V A E G P F F L S 360 301 TTCAATATCT TGGGGGAACC AGTAGATAAT CAGGGTCCTG TGTATACTGC TGAAACTTCT S T P V Y T A E V N D Q G L E P I G F N 1380 1321 CTGGAAGATA CTATCAAAGG CTTCAAGATG ATTCTGTCTG GGGAATTAGA TGACCTGCCA D L P G E L D I L S F K M T I K G L E D 420 361 CCTATTCACC GAGCTGCCCC TAAATTTACC GATTTAGACA CCAAGCCCAC TGTATTTGAG T K P T V F E D L D K F T R A A P P I H 1440 1381 GAACAGGCAT TCTACTTGGT AGGAGATATT CAGGAAGCTA AGGCTAAAGC TGAAAAACTC 480 421 ACTGGGATCA AGGTTATCGA CTTGCTGACT CCCTATCGTC GCGGCGGTAA AATCGGCCTG P Y R I G L R G G K L L T K V I D T G I 1500 1441 AAGCAAGATT AAGATCCCCC TATCCCCCCT TGATCCCCCC TTGGTCCCCC CCAGTGGGGG * K Q D 540 481 TTTGGCGGTG CTGGTGTGGG CAAAACCGTT ATCATGATGG AGCTAATCAA CAATATTGCC N I A K T V I M M E L I N A G V G F G G 1560 1501 GGAAACAGGG GGGAGAAAAC AACCCCCCTT ATTTAAGGGG GGAGAAGAGG TTGGTTTGTC 600 541 ATCAACCACG GTGGAGTCTC CGTCTTCGGC GGTGTGGGAG AGCGCACTCG TGAAGGGAAT E G N E R T R V F G G V G I N H G G V S 660 601 GACCTTTACA ATGAAATGAT TGAATCGAAG GTTATTAACG CTGATAACCT CAACGAGTCT V I N A D N L N E S E S K D L Y N E M I 720 661 AAAATTGCTC TAGTTTACGG TCAGATGAAT GAACCCCCTG GTGCGAGAAT GCGGGTAGGT G A R M R V G 0 M E P P L V Y G N K I A 780 721 CTATCTGCTC TGACTATGGC TGAGTATTTC CGGGATGTGA ACAAGCAAGA TGTGTTGCTG V L L NK Q D R D V E Y F L T M A L S A 840 781 TTCATCGACA ATATTTTTCG CTTTGTTCAA GCTGGTTCTG AGGTATCTGC CCTGTTAGGT F V Q E V S A L L G A G S F I D N I F R E Q A F Y L V G D I Q E A K A K A E K L 1620 1561 AGTTACTGCT TGGTAAAAAC AAACAACAAA CAACCAATAA CAAAAAACAA ACAACAAATA 1680 1621 ACAAACAACC AAAAATGACT TTAACTTTGC GGGTAATTAC CCCAGATAAG ACAGTTTGGG T V W P D K R V I T L T L M T 1740 1681 GTTTTGACAG: ACAGGTAGGG ACGATAGTGT AGAAGAAATT GTCCTGCCCA GTACTACGGG Q V G V L T V L P S T T G E E I D D S V 1800 1741 GTCACGCTCC TCTGTTAACG GCTTTGGATA CTGGGGTGAT GCGAGTTCGT CCTGGCAAAG P R V R A L D T G V M G K L L T G H A P 1860 1801 ATTGGCAGGC GATCGCCCTC ATGGGTGGAT TTGCTGAAGT AGAGAACAAC GAGGTGAAAG E V K F A E V E N N M G G I A L D W Q A 1920 1861 TTCTAGTGAA TGGTGCGGAA GTGGGAGATA GTATCGATAA AGAAACTGCT CGCACTGAGT R T E S I D K E T A V G D G A E L V N V 900 841 CGCATGCCTT CTGCTGTGGG TTACCAGCCT ACTCTGGGTA CTGACGTGGG AGATTTGCAA D L Q P T L G T D V G Y Q S A V G R M P 960 901 GAGCGGATTA CTTCTACTAA GGAAGGTTCT ATTPCCTCTA TTCAAGCGGT TTACGTTCCT Y V P I Q A V I T S T S T K E G S E R I 1020 961 GCGGACGATT TAACCGACCC CGCTCCTGCT ACTACTTTTG CTCACTTAGA CGGTACTACG A H L D G T T T T F A P A L T D P A D D 1980 1921 TCCAACAAGC GGAACAAAAT CTCGCTCGAG CCAATCAAGG AGACAACCGC CAAGAGCTAA Q Q Q Q Q D N R E L A N E N L A R G A F 2040 1981 TTCAAGCAAC CCAAGAGTTC AAGAAAGCAA GAGCCCGCTT TCAAGCTGCT GGGGGCATGA Q A A K K A R A R F G G M I Q A T Q E F 2041 2050 CTTAAGGCAA T * FIG. 2. Sequence of the atpBE locus from P. didemni. The sequence is shown starting with the atpB initiation codon with the predicted amino acid sequence of the and E polypeptides (from positions 1-1452 and 1635-2045, respectively). University). Sequencing was carried out by the dideoxynucleotide chain-termination method (23). Data Sets for Phylogenetic Inference. The P. didemni predicted 3 and e polypeptide sequences were aligned with homologues from chloroplasts (wheat and Chlamydomonas reinhardtii), cyanobacteria (Synechococcus sp. PCC6301 and Anabaena sp. PCC7120), and Escherichia coli (all sequences from GenBank, release 67) using MSA (24). Regions where gaps greater than three residues long are shared by two or more taxa were omitted to prevent artifacts (e.g., ref. 17). Nucleotide sequences were aligned to correspond to the amino acid alignment. For phylogenetic inference the PHYLIP package (versions 3.3 and 3.4; ref. 25) was used. PROTPARS carries out amino acid parsimony based on the genetic code, and DNAPARS nucleotide parsimony scoring all substitutions equally. DNAML was used for nucleotide maximum likelihood inference. DNADIST estimates a distance matrix from nucleotide data, from which FITCH infers a phenogram using the Fitch-Margoliash least-squares criterion (26). For the models in DNADIST and DNAML, a transition/transversion ratio was estimated from the data, and probabilities in the transition matrix were derived from this and (F option) the observed population base frequences (27). To take account of different rates of substitution at the three codon sites (28), these were assigned weights based on all possible pairwise comparisons of the five oxygenic taxa. For DNAPARS, weights (W1/W2/ W3) were derived- as reciprocals of the pairwise observed substitution rates. For DNADIST and DNAML, transformed rates (K1/K2/K3) were estimated using the Kimura twoparameter equation (28). Hypotheses Tested. A two-step approach was used for parsimony and maximum likelihood methods (19), as exhaus- 2744 Evolution: Lockhart et al. Proc. Natl. Acad. Sci. USA 89 (1992) Table 1. Estimates of the parameters used in inference programs Rates Weights Transition/ Gene (K1/K2/K3) (Wl/W2/W3) F 40 transversion ratio Synechococcus 1.2 2.16:1.00:18.55 8:17:2 atpB 1.5 1.58:1.00:5.06 4:5:2 atpE Weights are rounded to whole numbers for input to DNAPARS, and transition/transversion ratios are given to one decimal place only because of variation between different codon positions. tive pairwise analysis in a single step would have required the testing of too many trees to be practicable. In step 1, 15 unrooted trees (Fig. 1, trees 1-15) were analyzed, an exhaustive test that requires no assumption on the monophyly of the green chloroplasts. Tree robustness for these programs was assessed by a site-by-site pairwise test (27) that determines whether the step (or support) difference between each tree and the best is significant. This requires that the site-by-site step (support) differences are normally distributed, an assumption for which there is empirical support (19). If the step (support) difference for two trees exceeds its standard error by a factor of 1.96 or greater, the trees are considered different (null hypothesis rejected) at P = 0.05. Step 2 of testing takes all trees not different from the best at P = 0.05 (or other appropriate level) and analyzes all possible rooted versions using E. coli as an outgroup. Use of a "purple" bacterium (as defined by Woese; ref. 29) to root the oxygenic taxa is standard (15) and justified by comparison of the photosynthetic systems of the two groups. Sequences from all six taxa were also used together in a single-step analysis without pairwise testing to verify that the best trees identified in step 2 were the best overall. For DNADIST/FITCH, robustness was determined empirically by bootstrapping, again using E. coli as an outgroup. RESULTS AND DISCUSSION Features of the Sequence. The sequence of 2050 base pairs (bp) of the P. didemni atpBE locus is shown in Fig. 2. The genes are separated by 182 bp, more than in cyanobacteria (Synechococcus sp. PCC6301, 70 bp; Anabaena sp. PCC7120, 93 bp) and in contrast to the overlapping genes of many higher plant chloroplasts (30-33). P. didemni lacks the tRNAMet gene (data not shown) found downstream of the higher plant chloroplast atpE gene (33, 34), although, like the size of the intergenic spacer, this does not suggest a cyanobacterial affinity as lack of the tRNAMet and long intergenic spacers are both primitive features (i.e., characteristic of E. coli) (35). Interestingly, the P. didemni intergenic region has four direct repeats of a 10-bp element (AACAAACAAC) containing a 7-bp sequence similar to a repetitive element in the Anabaena sp. PCC7120 atpBE intergenic region (31). The role of such elements is unclear. Phylogenetic Inference Parameters. Estimates of parameters derived from the data set and used in the inference models are shown in Table 1. All agree with reported values (28, 36, 37) and indicate a higher degree of conservation ofatpB than atpE. a.pS. Step 1 testing using atpB sequences for all programs selected trees 1-3 (Fig. 1) for rooting. For PROTPARS, this group of trees was favored over any of trees 4-15 at P = 0.05. s \w f \a 1. s1 /WA 2. Cf >c4.-K /b A d P Sb g\ C P A C P C Is FIG. 3. Topologies for step 2 testingof the relationships between the oxygenic taxa. For trees 1-3 of Fig. 1, 21 trees a-g were produced by introducing E. coli as an outgroup on the branch indicated. Thus tree la has the outgroup on the Synechococcus sp. PCC6301 branch, etc. Branch lengths are arbitrary; abbreviations are as defined in Fig. 1. Prochloron 63 | Wheat F Ch100 Chlamydomonas Anabaena E. coli FIG. 4. Phenogram from bootstrap analysis of atpB sequences using DNADIST and FITCH. The numbers indicate the number of times out of 100 that the species to the right of each node were placed in a monophyletic group. The tree is rooted using E. coli as an outgroup. Branch lengths are arbitrary. Testing the 21 rooted trees la-3g (step 2; Fig. 3) for PROTPARS gave a best tree with P. didemni not related to the chloroplasts to the exclusion of Anabaena sp. PCC7120 (Fig. 3, tree 2e). Although this tree was poorly resolved from most of the other trees tested in step 2 (suggesting no unique phylogeny is supported), it was favored significantly (P = 0.05) over the three topologies containing P. didemni ancestral to the chloroplast clade (trees la, lb, and ic). This is therefore evidence against P. didemni being descended from the chloroplast protoendosymbiont. For DNAPARS, trees 1-3 (Fig. 1) were again favored over trees 4-15 (at P = 0.15), as they were for DNAML (P = 0.05). At step 2, the best topologies for both DNAPARS and DNAML were the same (Fig. 3, tree 2a) and placed P. didemni with the cyanobacteria. However, in neither case was this conclusion significantly favored over a chloroplast affinity. Distance matrix analysis favored tree 2d, although bootstrapping (Fig. 4) suggested that the support for this topology over a chloroplast affinity is not significant (P = 0.76; 24 of the 100 trees contained a P. didemni-chloroplast clade). Therefore, in contrast to PROTPARS, analysis of the atpB nucleotide sequence does not permit a robust conclusion as to the phylogeny of P. didemni. apE. Step 1 testing using atpE sequences for PROTPARS, DNAPARS, and DNAML selected (P = 0.05) the same trees 1-3 as atpB (Fig. 1) for rooting. At step 2, rooted trees containing a Prochloron-Anabaena-Synechococcus clade were favored over any lacking such a grouping by PROTPARS (P = 0.05) and DNAPARS (P = 0.1) (Table 2). In neither case was the branch order within this bacterial clade determined. The best rooted DNAML tree was le, apparently confirming this conclusion, although under likelihood inference this tree was not signifTable 2. Results of inference from atpE sequences Evidence for a Best Prochloron-SynechococcusAnabaena clade, P value topology Program 0.05 le PROTPARS 0.1 3e DNAPARS 0.01 le DNADIST/FITCH NS le DNAML Best topologies are described in Fig. 3. No branch order within the Prochloron-Synechococcus-Anabaena clade is specified. NS, not significant. Evolution: Lockhart et al. Proc. Natl. Acad. Sci. USA 89 (1992) Prochloron 99 | Anabaena 64 Synechococcus Wheat 2745 prochlorophyte phylogenies determined to date should be treated with caution. We are grateful to David Judge, David Penny, Mike Hendy, and "Fast" Eddie Holmes for discussions; to Roger Hiller and Mike Weldon for synthesis of oligonucleotides; and to Cheryl Handford for technical assistance. This work has been supported by the Australian Research Council, the University of Sydney, the Science and Engineering Research Council, the Cambridge Philosophical Society, Corpus Christi College, Cambridge, and the Wellcome Foundation. 72 Chlamydomonas E. coli FIG. 5. Phenogram from bootstrap analysis of atpE sequences using DNADIST and FITCH. The numbers indicate the number of times out of 100 that the species to the right of each node were placed in a monophyletic group. The tree is rooted using E. coli as an outgroup. Branch lengths are arbitrary. icantly different (P = 0.05) from trees la, lb, and ic containing P. didemni-chloroplast clades. DNADIST/FITCH again favored tree le, bootstrap values (Fig. 5) strongly supporting the same Prochloron-AnabaenaSynechococcus clade (P = 0.01; only 1 of 100 replicate trees grouped the prochlorophyte specifically with the chloroplasts). This is thus strong evidence for a cyanobacterial affinity, suggesting that chlorophyll-a/b-binding polypeptides in chloroplasts and P. didemni may have arisen independently. The order within the P. didemni-cyanobacteria clade is once again not determined. Overall, atpE thus offers evidence ranging from being merely consistent with (DNAML) to very strongly in favor of (FITCH) a cyanobacterial affinity for P. didemni. This conclusion is wholly consistent with findings for atpB and the same as that drawn from analysis of the Px. hollandica 16S rRNA (15), rbcLS (16), and psbA (19, 20) sequences. Hence analyses of prochlorophyte sequence data at present provide little or no evidence for a phylogeny placing prochlorophytes as most closely related to green chloroplasts. Systematic Errors. It should be recognized, however, that a systematic error characterizes all these analyses. At present, inference from sequence data assumes explicitly or otherwise that the sequences studied have evolved with a fixed base frequency since divergence from a common ancestor (25, 38-40). Violation of this assumption is very noticeable with oxygenic taxa, whose genomes typically show a wide range of G + C contents, with chloroplast genes usually being A + T-rich and prochlorophyte sequences either intermediate in G + C content between those from plastids and cyanobacteria (e.g., atpE) or within the cyanobacterial range (e.g., psbAI, psbAII, atpB, rbcLS, and 16S rRNA) (15-18). When sequences contain such a substitutional bias, inference has been shown to be unreliable (40-42), sequences tending to group because of shared substitution processes rather than necessarily common ancestry. This problem is compounded when long edges are present (40-42) and where the radiation occurs over a relatively short time scale, both of which may apply to the oxygenic taxa diverging after the evolution of the ability to split water (43). When the data analyzed fit the inference models so poorly, tests of tree robustness become less reliable (40-42), hence the lack of robustness for much of the data presented, and even topologies with strong statistical support become suspect (40). Until methods are developed to overcome such problems, the 1. Raven, P. H. (1970) Science 169, 641-646. 2. Whatley, J. M. & Whatley, F. R. (1981) New Phytol. 87, 233-247. 3. Gray, M. W. (1989) Trends Genet. 5, 294-299. 4. Lewin, R. A. (1984) Phycologia 23, 203-208. 5. Burger-Wiersma, T., Veenhuis, M., Korthals, H. J., van de Wiel, C. C. M. & Mur, L. R. (1986) Nature (London) 320, 262-264. 6. Chisholm, S. W., Olsen, R. J., Zettler, E. R., Goericke, R., Waterbury, J. B. & Welschmeyer, N. A. (1988) Nature (London) 334, 340-342. 7. Cox, G. & Dwarte, D. M. (1981) New Phytol. 88, 427-438. 8. Golecki, J. R. & Jurgens, U. W. (1989) Arch. Microbiol. 152, 77-82. 9. Bullerjahn, G. S., Matthijs, H. C. P., Mur, L. R. & Sherman, L. A. (1990) FEMS Microbiol. Lett. 67, 99-106. 10. Seewaldt, E. & Stackebrandt, E. (1982) Nature (London) 295, 618-620. 11. Mackay, R. M., Salgado, D., Bonen, L., Stackebrandt, E. & Doolittle, W. F. (1982) Nucleic Acids Res. 10, 2%3-2970. 12. van Valen, L. M. (1982) Nature (London) 298, 493-494. 13. van den Eynde, H., de Baere, R., de Roeck, E., van de Peer, Y., Vandenberghe, A., Willekens, P. & de Wachter, R. (1988) J. Mol. Evol. 27, 126-132. 14. Lockhart, P. J., Larkum, A. W. D., Beanland, T. J. & Howe, C. J. (1989) FEBS Lett. 248, 127-130. 15. Turner, S., Burger-Wiersma, T., Giovannoni, S. J., Mur, L. R. & Pace, N. R. (1989) Nature (London) 337, 380-382. 16. Morden, C. W. & Golden, S. S. (1991) J. Mol. Evol. 32, 379-395. 17. Morden, C. W. & Golden, S. S. (1989) Nature (London) 337, 382-384. 18. Morden, C. W. & Golden, S. S. (1989) Nature (London) 339, 400. 19. Beanland, T. J. (1990) Ph.D. thesis (Univ. of Cambridge, Cambridge). 20. Kishino, H., Takashi, T. & Hasegawa, M. (1990) J. Mol. Evol. 31, 151-160. 21. Beanland, T. J. (1990) J. Theor. Biol. 145, 535-545. 22. Bedbrook, J. R., Coen, D. M., Beaton, A. R., Bogorad, L. & Rich, A. (1979) J. Biol. Chem. 245, 905-910. 23. Bankier, A. T. & Barrell, B. G. (1989) in Nucleic Acids Sequencing: A Practical Approach, eds. Howe, C. J. & Ward, E. S. (IRL, Oxford), pp. 37-78. 24. Lipman, D. J., Altschul, S. F. & Kececiolgu, J. D. (1989) Proc. Nati. Acad. Sci. USA 86, 4412-4415. 25. Felsenstein, J. (1988) Annu. Rev. Genet. 22, 521-565. 26. Fitch, W. M. & Margoliash, E. (1967) Science 155, 279-284. 27. Kishino, H. & Hasegawa, M. (1989) J. Mol. Evol. 29, 170-179. 28. Kimura, M. (1981) Proc. Natl. Acad. Sci. USA 78, 454-458. 29. Woese, C. R. (1987) Microbiol. Rev. 51, 221-271. 30. Cozens, A. L. & Walker, J. E. (1987) J. Mol. Biol. 194,359-383. 31. Curtis, S. E. (1987) J. Bacteriol. 169, 80-86. 32. Zurawski, G., Bottomley, W. & Whitfeld, P. R. (1982) Proc. Natl. Acad. Sci. USA 79, 6260-6264. 33. Shinozaki, K., Ohme, M., Tanaka, M., Wakasugi, T., Hayashida, N., Matsubayashi, T., Zaita, N., Chunwongse, J., Obokata, J., Yamaguchi-Shinozaki, K., Ohto, C., Torazawa, K., Meng, B. Y., Sugita, M., Deno, H., Kamogashira, T., Yamada, K., Kusuda, J., Takaiwa, F., Kato, A., Tohdoh, A., Shimada, H. & Sugiura, M. (1986) EMBO J. 5, 2043-2049. 34. Howe, C. J., Fearnley, I. M., Walker, J. E., Dyer, T. A. & Gray, J. C. (1985) Plant Mol. Biol. 4, 334-345. 35. Walker, J. E., Saraste, M. & Gay, N. J. (1984) Biochim. Biophys. Acta 768, 164-200. 2746 Evolution: Lockhart et al. 36. Ritland, K. & Clegg, M. T. (1987) Am. Nat. 130, S74-S100. 37. Zurawski, G. & Clegg, M. T. (1987) Annu. Rev. Plant Physiol. 38, 391-418. 38. Sidow, A. & Wilson, A. C. (1990) J. Mol. Evol. 31, 51-6. 39. Gillespie, J. H. (1986) Annu. Rev. Ecol. Syst. 17, 637-665. 40. Lockhart, P. J., Howe, C. J., Bryant, D. A., Beanland, T. J. & Larkum, A. W. D. (1992) J. Mol. Evol. 34, 153-162. Proc. Nat!. Acad. Sci. USA 89 (1992) 41. Penny, D., Hendy, M., Zimmer, E. A. & Handby, R. K. (1990) Aust. Syst. Bot. 3, 21-38. 42. Lockhart, P. J. (1990) Ph.D. thesis (Univ. of Sydney, Sydney, Australia). 43. Awramik, S. M. (1990) in Palaeobiology, A Synthesis, eds. Briggs, D. E. & Crowther, P. R. (Blackwell, Oxford), pp. 336-341.