1810Mol. Biol. Evol. 18(9):1810–1822. 2001᭧ 2001 by the Society for Molecular Biology and Evolution. ISSN: 0737-4038Accelerated Evolution of Functional Plastid rRNA and Elongation FactorGenes Due to Reduced Protein Synthetic Load After the Loss ofPhotosynthesis in the Chlorophyte Alga PolytomaDawne Vernon,*1 Robin R. Gutell,† Jamie J. Cannone,† Robert W. Rumpf,*2and C. William Birky Jr.*‡*Department of Molecular Genetics, Ohio State University; †Institute of Cellular and Molecular Biology, University of Texasat Austin; and ‡Department of Ecology and Evolutionary Biology and Graduate Interdisciplinary Program in Genetics,University of ArizonaPolytoma obtusum and Polytoma uvella are members of a clade of nonphotosynthetic chlorophyte algae closelyrelated to Chlamydomonas humicola and other photosynthetic members of the Chlamydomonadaceae. Descendedfrom a nonphotosynthetic mutant, these obligate heterotrophs retain a plastid (leucoplast) with a functional proteinsynthetic system, and a plastid genome (lpDNA) with functional genes encoding proteins required for transcriptionand translation. Comparative studies of the evolution of genes in chloroplasts and leucoplasts can identify modesof selection acting on the plastid genome. Two plastid genes—rrn16, encoding the plastid small-subunit rRNA, andtufA, encoding elongation factor Tu—retain their functions in protein synthesis after the loss of photosynthesis intwo nonphotosynthetic Polytoma clades but show a substantially accelerated rate of base substitution in the P. uvellaclade. The accelerated evolution of tufA is due, at least partly, to relaxed codon bias favoring codons that can beread without wobble, mainly in three amino acids. Selection for these codons may be relaxed because leucoplastsare required to synthesize fewer protein molecules per unit time than are chloroplasts (reduced protein syntheticload) and thus require a lower rate of synthesis of elongation factor Tu. Relaxed selection due to a lower proteinsynthetic load is also a plausible explanation for the accelerated rate of evolution of rrn16, but the available dataare insufﬁcient to test the hypothesis for this gene. The tufA and rrn16 genes in Polytoma oviforme, the sole memberof a second nonphotosynthetic clade, are also functional but show no sign of relaxed selection.IntroductionNonphotosynthetic land plants and algae serve as abasis for interesting natural experiments on the evolution-ary consequence of the loss of a signiﬁcant cell function.After losing the ability to do photosynthesis, nonphoto-synthetic species use various alternative carbon sources,with the plants becoming parasitic on other plants, whilethe algae take up complex organic molecules from theirenvironment. Recognized by their lack of chlorophyll,these nongreen organisms have unique plastid (‘‘leuco-plast’’) genomes. The evolutionary consequences of theloss of photosynthesis can be studied by comparing theleucoplast genomes of nonphotosynthetic species with thechloroplast genomes of their closest photosynthetic rela-tives. The majority of the genes in the chloroplasts of pho-tosynthetic green algae and land plants encode proteinsrequired for photosynthesis or gene expression (transcrip-tion and translation [Gillham 1994]; for recent data fromcomplete chloroplast genome sequences, see Turmel, Otis,and Lemieux , Lemieux, Otis, and Turmel ,and the NCBI chloroplast genome page at http://www.ncbi.nlm.nih.gov:80/PMGifs/Genomes/plastidstax.html). These two functions account for 32 and 16 genes,1 Present address: National Institute of Standards and Technology,DNA Technologies Group, Gaithersburg, Maryland.2 Present address: LabBook.com, Inc., Columbus, Ohio.Key words: Polytoma, Chlamydomonas, chloroplast rRNA gene,chloroplast elongation factor gene, substitution rate, codon bias.Address for correspondence and reprints: C. William Birky Jr.,Department of Ecology and Evolutionary Biology, Biological SciencesWest, University of Arizona, Tucson, Arizona 85721. E-mail:firstname.lastname@example.org, in Chlamydomonas reinhardtii, which alsoencodes a minimal set of rRNA and tRNA genes (http://www.biology.duke.edu/chlamygenome/chloro.html).Genes encoding proteins with other functions, as wellas unidentiﬁed open reading frames, are found in sometaxa; C. reinhardtii has 15 of these. There is strong in-direct evidence that at least one of these genes, yet tobe identiﬁed, codes for a protein that has an essentialnonphotosynthetic (ENP) function (Gillham 1994, pp.83–86).Investigations of leucoplast genome structure andgene sequences of nonphotosynthetic organisms havebeen limited to several parasitic angiosperm families(Scrophulareae and Orobanchaceae; e.g., dePamphilisand Palmer 1990; Wimpee, Morgan, and Wrobel 1992a,1992b; Nickrent, Duff, and Konings 1997; Wolfe anddePamphilis 1998; Young and dePamphilis 2000) andthe euglenoid alga Astasia (Siemeister, Buchholz, andHachtel 1990; Siemeister and Hachtel 1990a, 1990b).The leucoplasts of these nonphotosynthetic species dif-fer from the chloroplasts of their close green relativesin numerous features. Morphologically, they are char-acterized by reduction or elimination of the thylakoidmembranes. They all retain leucoplast ribosomes andleucoplast DNA. However, their leucoplast genomes areoften reduced in size and complexity compared withchloroplast genomes in photosynthetic relatives.These leucoplast genomes allow investigation ofthe effects of selection on rates of evolution. When theability to do photosynthesis is lost, the photosyntheticand photorespiration genes lose their function and con-sequently are no longer subject to selection; they areexpected to become pseudogenes and can be lost entire-ly. In contrast, the leucoplast genes needed for transcrip-
Accelerated Evolution of Leucoplast Genes 1811FIG. 1.—Cladogram showing the relationships of the taxa used inthis study based on Rrn18 sequences (Rumpf et al. 1996; unpublisheddata).tion and translation of the leucoplast genome probablyremain functional and subject to at least some degree ofselection because they are needed to transcribe andtranslate the ENP genes. The leucoplast genomes of thebeech root parasite Epifagus virginiana (Orobancha-ceae), the oak parasite Conopholis americana (Oroban-chaceae), and the euglenoid Astasia longa display thesecharacteristics. Most expression genes are still presentin the leucoplast genomes and appear intact, and someleucoplast RNA and protein products have been dem-onstrated (Siemeister, Buchholz, and Hachtel 1990; Sie-meister and Hachtel 1990a, 1990b; Wimpee, Morgan,and Wrobel 1992; Wolfe, Morden, and Palmer 1992).Conversely, many leucoplast genes that coded for pho-tosynthetic proteins appear nonfunctional in key do-mains, are grossly truncated, or are absent from theseleucoplast genomes. Although most expression genesare selectively retained, the leucoplast genome in Epi-fagus is missing more than a dozen expression genes(tRNA genes, ribosomal protein genes, and all fourRNA polymerase subunit genes; Morden et al. 1991),all four RNA polymerase genes are pseudogenes inLathraea (Lusson, Delavault, and Thalouarn 1998), andthe leucoplast genome in Conopholis is apparently miss-ing several leucoplast tRNA genes (Wimpee, Morgan,and Wrobel 1992b). Presumably, these tRNAs and pro-teins are imported from the cytoplasm at a rate that maybe too low for detection but is sufﬁcient for protein syn-thesis in leucoplasts.The tempo of evolution has also changed in theleucoplast genomes of these nonphotosynthetic species.Most, but not all, of the apparently still-functional genesanalyzed in leucoplasts show an increased rate of nu-cleotide substitution compared with rates in their greenrelatives. Most of the rate increases found in functionalleucoplast genes in Astasia and Epifagus are in the rangeof 1.5-fold to 8-fold, while rrn16 genes in Epifagus andConopholis show 40-fold rate increases (references inResults and Discussion). This suggests that selection onthese genes has been relaxed, even though it has notbeen eliminated completely.To test the generality of these results, we extendedthe analysis of the sequence and evolution of leucoplasttranslation genes to the nonphotosynthetic chlorophytealgae, separated from the euglenoid plastid and fromland plants by over 400 Myr of evolution and large dif-ferences in physiology and habitats. Phylogenetic anal-yses of Rrn18 sequences show that the members of thenonphotosynthetic genus Polytoma belong to two dif-ferent lineages within the clade that includes all Chla-mydomonas species as well as a number of other pho-tosynthetic genera and another nonphotosynthetic clade,Polytomella (Rumpf et al. 1996; unpublished data).Many species of Chlamydomonas are facultative auxo-trophs, capable of utilizing acetate as their sole carbonand energy source, and nonphotosynthetic mutants arereadily isolated in C. reinhardtii (Harris 1989). It is as-sumed that Polytoma species arose as nonphotosyntheticmutants of facultative auxotrophs similar to the extantChlamydomonas. The single large cup-shaped leucoplastin Polytoma does not have thylakoid membranes but stillcontains ribosomes, DNA (lpDNA), rRNA, and storedstarch granules (Lang 1963; Scherbel, Behn, and Arnold1974; Siu, Chiang, and Swift 1976; Vernon-Kipp, Kuhl,and Birky 1989). Polytoma is sensitive to inhibitors ofchloroplast protein synthesis, which is additional evi-dence that the leucoplast is synthesizing at least oneprotein that is essential for auxotrophic growth and re-production (Scherbel, Behn, and Arnold 1974).Although these data strongly suggest that Polytomaretains a functional leucoplast expression system, theleucoplast genes involved have not been identiﬁed anddemonstrated to be functional. We sequenced the rrn16gene and the tufA gene (encoding the plastid elongationfactor Tu) from two representatives of the Polytomauvella clade (P. uvella 964 and Polytoma obtusumDH1), plus Polytoma oviforme, which is the sole mem-ber of the second Polytoma lineage. For comparison,these genes were also sequenced from two closely re-lated photosynthetic relatives (Chlamydomonas humi-cola SAG 11-9 and Chlamydomonas dysosmos UTEX2399). The evolutionary relationships of these strainsand some others involved in the analysis are shown inﬁgure 1. This cladogram agrees with phylogenetic anal-yses of the rrn16 and tufA genes (ﬁgs. 2 and 3). Rela-tive-rate tests showed increased substitution rates inrrn16 and tufA, compared with green relatives, in thetwo P. uvella species. However, sequence analysesshowed that the genes were subject to selection andtherefore functional. The increase in substitution ratewas greater at sites subject to less stringent selection,implicating a partial relaxation of selection. The tufAgene of P. obtusum showed a large reduction in codonpreference, suggesting that the relaxed selection is dueat least partly to a reduced load of protein synthesis.This was proposed earlier for the increased substitutionrates in Epifagus (Wolfe et al. 1992), but alternative ex-planations were not ruled out. We observed no increasein the substitution rate in the branch leading to P. ovi-forme, suggesting that photosynthesis was lost more re-cently in this lineage. This is the ﬁrst analysis of themolecular evolutionary consequences of the loss of pho-tosynthesis in a chlorophyte alga.Materials and MethodsOrganismsPolytoma uvella (UTEX 964) and P. oviforme(SAG 62-27) were obtained from the University of Tex-
1812 Vernon et al.FIG. 2.—Neighbor-joining tree of rrn16 sequences. Branch lengths in percentages of substitutions are shown above the branches; belowthe branches are the lengths in the most parsimonious tree, which had an identical topology.FIG. 3.—Neighbor-joining tree of tufA sequences. Branch lengths in percentages of substitutions are shown above the branches; below thebranches are the lengths in the most parsimonious tree and the maximum-likelihood tree, which had identical topologies.as Culture Collection of Algae and from Sammlung vonAlgenkulturen Gottingen, respectively. P. obtusum (des-ignated strain DH1 by us) was obtained from David Her-rin at the University of Texas at Austin; it originallycame from Luigi Provasoli’s collection at Yale. All cul-tures were subcloned once or twice and grown in Po-lytomella medium.Chlamydomonas humicola UTEX 225 and C. dy-sosmos UTEX 2399, from the University of Texas Cul-ture Collection of Algae, were combined under the spe-cies name Chlamydomonas applanata Pringsheim basedon morphology and autolysin cross-reactions (Ettl 1976)and identity of the nuclear Rrn18 gene sequences (Gor-don et al. 1995). Consistent with this, we found no sub-stitution differences and only one insertion or deletiondifference in their chloroplast rrn16 sequences, whilethe tufA sequences showed two synonymous differencesand one nonsynonymous differences and no insertionsor deletions. Consequently, we included only the C.humicola sequences in the analyses described here.DNA PreparationThe rrn16 gene of P. uvella was cloned. Whole-cell DNA was isolated with a lysis method designed toyield high-molecular-weight chloroplast DNA, modiﬁedfrom Grant, Gillham, and Boynton (1980) as describedin Vernon (1996). This DNA was fractionated in aCsClϩbisbenzimide equilibrium gradient. The top bandin the gradient was identiﬁed as leucoplast DNA bySouthern hybridization with an rrn16 probe and dot blothybridization with a tufA probe. The C. reinhardtiicpDNA probes were provided by Elizabeth Harris (DukeUniversity) and Jeffrey Palmer (Indiana University). Thetop band was used to prepare a HindIII library clonedin pBluescript. DNA obtained from these clones by al-kaline lysis plasmid minipreps (Sambrook, Fritsch, andManiatis 1989) was electrophoresed, and Southern blotson GeneScreen Plus were hybridized with the cpDNAprobes. A clone containing the rrn16 gene in a 6.2-kbinsert was identiﬁed and puriﬁed for sequencing with aGeneClean kit (Bio 101). All other new sequences usedin this study were of genes ampliﬁed from partially pu-riﬁed whole-cell DNA isolated from CTAB lysates of1L algal cultures.Polymerase Chain Reaction AmpliﬁcationsPrimers located near the ends of the rrn16 and tufAgenes were used to obtain DNA templates for sequenc-ing. The 5Ј and 3Ј primers for rrn16 were A-17 (5Ј-GTTTGATCCTGGCTCAC-3Ј) and 5005-15 (3Ј-CA-TGTGTGGCGGGCA-5Ј). The 5Ј and 3Ј degenerateprimers for all but one of the tufA genes were 1F (5Ј-GGDCAYGTTGAYCAYGG-3Ј) and 5R (3Ј-TGA-CANCCRCGRCCRCA-5Ј). Primer 5R did not amplifytufA from P. obtusum, so the 3Ј ends of the tufA genesfrom the other chlamydomonad species were inspectedfor conserved areas, and an alternative 3Ј primer(1130R: 3Ј-CCRATACGGDCCACTRGC-5Ј) was de-signed and used, located 100 bases farther 5Ј of the orig-inal 3Ј primer 5R. This ampliﬁed a tufA fragment fromP. obtusum that was approximately 100 bases shorterthan the other chlamydomonad sequences. The rrn16
Accelerated Evolution of Leucoplast Genes 1813ampliﬁcation products sequenced were about 1.3 kblong, except for P. uvella, which was about 1.6 kb long;the tufA ampliﬁcation products were about 1.1 kb long,except for P. obtusum, which was about 1.0 kb long.Optimal ampliﬁcation conditions were determined foreach gene empirically; multiple separate ampliﬁcationswere performed and pooled, then puriﬁed usingGeneClean.SequencingBoth strands of all genes were sequenced manuallyusing a modiﬁed dsDNA Cycle Sequencing kit (LifeTechnologies). Most internal primers for sequencingwere obtained from Paul Fuerst for the rrn16 gene andfrom Jeffrey Palmer for the tufA gene; additional inter-nal primers in conserved regions were designed to ﬁllgaps in sequence coverage.Alignment of rrn16 SequencesThe ﬁve new sequences for the study reported here(P. uvella, P. obtusum, P. oviforme, C. humicola, andC. dysosmos) were initially aligned using CLUSTAL Win SeqApp (Gilbert 1992) to match the rrn16 primarystructure alignment in the Ribosomal Database Project(Maidak et al. 1994). The alignment was further reﬁnedby comparison with 70 publicly available plastid rrn16sequences using a SUN Microsystems workstation withthe alignment editor AE2 (developed by T. Macke,Scripps Research Institute, San Diego, Calif., and avail-able at http://www.cme.msu.edu/RDP/html/index.html).Sequences were initially aligned for maximum primarystructure similarity; then, all positions associated withthe comparatively inferred base pairs were checked toassure that these base-paired positions were properlyaligned. The ﬁnal alignment (with a complete list ofspecies and numerous chloroplast and Polytoma SSUrRNA secondary-structure diagrams) is available in thesupplement (on the MBE web site) as GenBank ﬁles(ﬁg. 3c in the supplement); a subset of sequences usedfor phylogenetic analysis is shown in less detail in se-quential format (ﬁg. 6 in the supplement) and in inter-leaved Pretty Print format (ﬁg. 7 in the supplement).Alignment of tufA SequencesTo assist alignment of tufA sequences, C. Delwicheand J. Palmer at Indiana University provided their align-ment, with 18 eubacterial, 8 cyanobacterial, 26 algal,and 4 land plant sequences (array described in Delwiche,Kuhsel, and Palmer 1995). The Polytoma and Chlamy-domonas sequences were aligned to various subsets ofthis array using DNA sequences but were inﬂuenced bythe resulting amino acid alignment. One thousand ﬁfty-three base pairs of the tufA gene were aligned (85% ofthe coding region), leaving out the ﬁrst 72 5Ј positionsand the last 96 3Ј positions for lack of data in some orall species. The complete alignment is available in thesupplement (ﬁg. 5); a subset of sequences is shown inless detail in sequential format (ﬁg. 6 in the supplement)and in interleaved Pretty Print format (ﬁg. 8 in thesupplement).Phylogenetic AnalysesGene trees for were produced with PAUP* (Swof-ford 1998) and PHYLIP, version 3.56 (Felsenstein1993). Sequence differences were corrected for multiplehits using the Jukes-Cantor one-parameter model (Jukesand Cantor 1969); otherwise, all analyses used defaultsettings. Before a set of sequences was subjected to phy-logenetic analysis or relative-rate tests, sites that weremissing in one or more species were removed from allsequences.Relative-Rate TestsRelative-rate tests (Sarich and Wilson 1973; Wuand Li 1985) were performed to detect differences be-tween rates of nucleotide (or amino acid) substitution inthe three Polytoma species studied, compared with greenspecies. Each relative-rate test involved a Polytoma iso-late (nongreen, N), its closest photosynthetic relative(green, G), and a photosynthetic outgroup species (O).The test parameters were KON and KOG, the estimatednumbers of base substitutions per site occurring alongthe lineages leading from the outgroup to the nongreenPolytoma and to the green ingroup, respectively. Theestimated numbers of substitutions per site (K) were ob-tained by correcting the observed sequence differencesper site for multiple hits with the Jukes-Cantor modelimplemented in the MEGA sequence analysis package,PHYLIP, or PAUP*. Two other correction methods wereused for comparison: the Kimura (1980) two-parametermethod, which allows different rates of transition versustransversion, and the Tamura (1992) method, which usesinformation about GϩC content as well as separate tran-sition and transversion rates, again using MEGA. Allthree correction methods added approximately the samenumber of unobserved substitutions (data not shown),so the Jukes-Cantor method was used for the relative-rate test because it had the smallest variance. KON andKOG were related to the evolutionary rates EON and EOGalong the nongreen and green lineages by KON ϭ EONTand KOG ϭ EOGT, where T is the time since divergenceof the two lineages and was, of course, the same forboth lineages. Any rate differences between green lin-eages and the nongreen lineages can be expressed as thedifference between these two numbers of substitutions(KON Ϫ KOG ϭ [EON Ϫ EOG]T). The signiﬁcance of ratedifferences was evaluated as in Muse and Weir (1992).A second method was also used to separate ob-served sequence differences into rates along differentgreen or nongreen lineages, employing phylogeneticsoftware. Gene trees in which the observed substitutionswere apportioned to the various branches of the tree byphylogenetic algorithms provided the inferred substitu-tions on each green or nongreen branch. The appor-tioned substitutions from a nongreen Polytoma speciesand from its nearest green relative (ingroup) to theirnearest ancestral node, KAN and KAG, respectively, wereused to calculate the ratio KAN/KAG or the difference KAN
1814 Vernon et al.Table 1Pairwise Numbers of Substitutions per Site Among rrn16 GenesChlamydomonasreinhardtiiChlamydomonasmoewusiiPolytomaoviformeChlamydomonashumicolaPolytomauvellaPolytomaobtusumChlorella . . . . . . . . . . . . . . . . . . . . . .Chlamydomonas reinhardtii. . . . . . .Chlamydomonas moewusii . . . . . . . .Polytoma oviforme . . . . . . . . . . . . . .Chlamydomonas humicola . . . . . . . .Polytoma uvella . . . . . . . . . . . . . . . .0.162 0.1850.1380.1710.1230.1060.1700.1190.1250.0950.2270.2120.2130.1940.1520.2200.1930.2000.1660.1420.054NOTE.—Estimated number of substitutions per site (sequence divergence) ϭ sequence differences per site corrected for multiple hits by the Jukes-Cantor method.Table 2Relative-Rate Tests on rrn16 Based on Neighbor-Joining, Maximum-Parsimony, and Maximum-Likelihood TreeBranch Lengths and on Pairwise Numbers of SubstitutionsNONGREENSPECIESGREENSPECIESKAN/KAGNeighborJoiningMaximumParsimonyMaximumLikelihoodOUTGROUPSPECIES KON ϪKOGPolytoma uvella 964Polytoma obtusumPolytoma oviformeChlamydomonas humicolaC. humicolaChlamydomonas moewusii3.322.890.674.344.030.503.052.680.65Chlamydomonas moewusiiChlamydomonas reinhardtiiC. moewusiiC. reinhardtiiChlamydomonas humicolaC. reinhardtii0.088*0.093**0.075**0.074**Ϫ0.030Ϫ0.015* Signiﬁcant at the 1% level.** Signiﬁcant at the 0.1% level.Ϫ KAG. The difference divided by KAG, i.e., (KAN ϪKAG)/KAG, can be used to compare the magnitudes ofthe rate increases along two different nonphotosyntheticlineages with different ingroups.Codon usage and the amount of codon usage biasin tufA were also investigated in P. obtusum, P. ovifor-me, C. humicola, and C. reinhardtii. All gaps were re-moved from the aligned sequences of these four species,leaving 349 codons. DNA Strider was used to calculatecodon usage in these sequences. Relative synonymouscodon usage (RSCU) was calculated for each codon us-ing MEGA. RSCU is the ratio of the observed frequencyof a particular codon to the expected frequency of thatcodon calculated on the assumption that all codons areused equally frequently; an RSCU value signiﬁcantlydifferent from 1 is evidence of biased codon usage(Sharp and Li 1987). A Pascal program provided byBrian Morton was used to calculate the codon bias index(CBI) and the codon adaptation index (CAI). The CBIis an overall measure of codon bias for the entire gene(Morton 1993); the CBI ranges from 0 (no codon biasin the gene) to 1 (maximum codon bias). The CAI ismeasure of bias in the use of a codon relative to its usein a reference set of highly expressed genes (Sharp andLi 1987).Results and DiscussionThe Evolutionary Rate of rrn16 is Accelerated in P.uvella and P. obtusum but not in P. oviformeFigure 2 shows the neighbor-joining tree of rrn16sequences; the topology of the most parsimonious treefrom an exhaustive maximum parsimony search is iden-tical. This tree is compatible with the trees of the nuclearRrn18 gene (ﬁg. 1). Above each line in ﬁgure 2 is thelength of the branch in the Neighbor-Joining tree in per-centage of substitutions; below each line is the lengthof the same branch in the parsimony tree. The treeshows a strong acceleration of substitution rate alongthe branches leading to the nonphotosynthetic P. uvellalineage.We used the distances on the tree in ﬁgure 2 tocalculate the ratio of substitution rates on nonphotosyn-thetic and photosynthetic lineages, as well as the differ-ence between nonphotosynthetic and photosyntheticrates. We also used the method of Wu and Li (1985) forrelative-rate tests based on corrected frequencies of pair-wise substitutions (table 1). Table 2 shows the results ofthese relative-rate calculations. The tests for P. uvellaand P. obtusum used C. humicola as their closest relativeand Chlamydomonas moewusii or C. reinhardtii as theoutgroup; the test for Polytoma oviforme used C. moe-wusii as the closest green relative and C. humicola orC. reinhardtii as the outgroup. All relative-rates testsshowed signiﬁcantly increased substitution rates in thebranch leading to P. obtusum versus the branch leadingto the ingroup (C. humicola), and an even greater rateincrease was seen in P. uvella. Figure 2 shows that thebranch leading from the common ancestor of the P.uvella clade and C. humicola to the common ancestorof P. uvella and P. obtusum is longer, i.e., has moresubstitutions, than the branch leading to C. humicola.This shows that the acceleration began in the commonancestor of the P. uvella clade, as expected. As a con-trol, we performed a relative-rate test on the nuclearRrn18 gene of P. uvella (not shown). The test showeda small increase in this nongreen species, but it was not
Accelerated Evolution of Leucoplast Genes 1815Table 3Pairwise Numbers of Substitutions per Site Among tufA GenesORGANISMSSEQUENCE DIVERGENCEAll Sites 3rd Position1st ϩ 2ndPositionsPolytoma obtusum–Chlamydomonas reinhardtii. . . .Chlamydomonas humicola–C. reinhardtii . . . . . . . . .P. obtusum–C. humicola . . . . . . . . . . . . . . . . . . . . . . .Polytoma oviforme–C. humicola. . . . . . . . . . . . . . . . .P. oviforme–C. reinhardtii . . . . . . . . . . . . . . . . . . . . .0.253340.133490.226190.155010.159880.748050.310780.622760.432020.401080.091120.058430.086120.047320.06324NOTE.—Estimated number of substitutions per site (sequence divergence) ϭ sequence differences per site corrected for multiple hits by the Jukes-Cantor method.Table 4Relative-Rate Tests on tufA Based on Tree Branch Lengths and Pairwise Distance MatricesNONGREENSPECIESGREENSPECIESKAN/KAGParsimonyNeighborJoiningMaximumLikelihood OUTGROUPKON Ϫ KOGAll Sites3rdPosition1st ϩ 2ndPositionsPolytoma obtusumPolytoma oviformeChlamydomonas humicolaChlamydomonas reinhardtii3.231.218.104.22.1681.34C. reinhardtiiC. humicola0.1998**0.02150.4408*0.11670.0278*Ϫ0.00951* Signiﬁcant at the 1% level.** Signiﬁcant at the 0.1% level.statistically signiﬁcant. No rate increase was seen in thebranch leading to P. oviforme.The Evolutionary Rate of tufA is Accelerated in P.obtusum but not in P. oviformeSequences of tufA are available for C. humicola, P.obtusum, C. reinhardtii, P. oviforme, and a number ofgreen algae outside of the Chlamydomonadaceae. Ofthese, Codium is the closest relative, but when we usedit as an outgroup, all three tree-making algorithmsgrouped Codium with P. obtusum, presumably due tolong-branch attraction. We therefore used only the fourChlamydomonadaceae, with C. reinhardtii serving asthe outgroup for C. humicola and P. obtusum, and C.humicola serving as the outgroup for C. reinhardtii andP. oviforme. The tufA sequences of these four specieshave 1,014 sites in common. The topology of the neigh-bor-joining tree of these genes (ﬁg. 2) is consistent withthe Rrn18 tree (ﬁg. 1). However, long-branch attractionwas still a problem with the parsimony and maximum-likelihood algorithms, which favored the tree that placedP. obtusum with C. reinhardtii. The correct parsimonytree (the one with the same topology as the Neighbor-Joining tree and all trees involving rrn16 or Rrn18) wasthe least parsimonious and had the lowest likelihoodscores, although not by much. The branch lengths forthe correct trees from all three algorithms are shown inﬁgure 2. In every case, the branch leading to P. obtusumis much longer than that leading to C. humicola, whileP. oviforme shows no acceleration. Table 3 shows theestimated pairwise numbers of substitutions amongthese species; the branch leading to P. obtusum is ac-celerated in all three trees (parsimony, Neighbor-Join-ing, and maximum likelihood).We performed relative-rate tests of the evolution oftufA in P. obtusum and P. oviforme, using the pairwisedifferences with the Jukes-Cantor correction for multiplehits (table 3). In addition to calculating relative rates ofnucleotide substitutions for all aligned sites, we com-pared ﬁrst ϩ second codon positions with third codonpositions. The results are shown in table 4; all testsshowed a signiﬁcantly higher substitution rate in thebranch leading to the nonphotosynthetic P. obtusum thanin the branch leading to the photosynthetic ingroup, C.humicola. The increase was greater in the third codonpositions than in the ﬁrst and second positions. No sig-niﬁcant difference was found between the branches lead-ing to P. oviforme and C. reinhardtii, in agreement withthe data from rrn16.The Plastid tufA Genes in Polytoma RemainFunctional After the Loss of PhotosynthesisOne possible explanation for the accelerated evo-lution of rrn16 and tufA is that the genes became non-functional in the nonphotosynthetic lineages. This is un-likely, given the evidence that they remain functional innonphotosynthetic land plants. We found additional ev-idence that tufA remained subject to selection, and hencefunctional, in the Polytoma lineages:1. There are no premature stop codons in the entiregene. This could be because no stop mutations oc-curred since the loss of photosynthesis or becausethey were eliminated by selection. For the P. uvellaclade, we estimated the probability of no stop mu-tations occurring as follows: First, we assumed thata truncation would not inactivate the protein if it oc-curred between the carboxyl terminal of the proteinand the 14th amino acid, since the ﬁrst 13 aminoacids are not involved in intermolecular bonding inEscherichia coli (Kawashima et al. 1996). In the re-mainder of the protein, we found 108 codons thatwere one substitution away from being stop codons.As described above, we know that more synonymous
1816 Vernon et al.substitutions occurred along the branch leading to P.obtusum than along the branch leading to C. humi-cola from their common ancestor. There were 0.2235extra substitutions per site in third codon positions,which must have occurred after photosynthesis waslost; this is also an estimate of the number of muta-tions per site. We found 81 sense codons in the tufAgene of P. obtusum that could have become a stopcodon as a result of one kind of substitution (e.g.,UCG to UAG); the expected number of such substi-tutions in the absence of selection was 81 ϫ 0.2235ϫ 1/3 ϭ 6.034. We found 27 sense codons that couldhave become stop codons as a result of either of twokinds of substitutions (e.g., UAC to UAA or UAG);the expected number of such substitutions was 27 ϫ0.2235 ϫ 2/3 ϭ 4.023. Consequently, the expectednumber of premature stop codons in the absence ofselection was 10.057, and from the Poisson distri-bution the probability of ﬁnding no premature stopcodons was eϪ10.057 ϭ 4.3 ϫ 10Ϫ5. We conclude thatthe tufA gene of P. obtusum must have been underselection that eliminated genes with premature stopcodons most or all of the time since the loss ofphotosynthesis.Additional evidence was obtained using a tufA se-quence obtained from P. uvella by Nedelcu (2001)using the UTEX stock without subcloning. Wealigned 999 bp, or 333 complete codons, of P. uvellaand P. obtusum. The sequences of these species dif-fered by 0.07892 synonymous substitutions per site,all of which must have occurred since they divergedfrom a common ancestor, after photosynthesis waslost. We used parsimony to reconstruct 321 codonsof the sequence of their most recent common ances-tor. This sequence contained 72 codons which couldhave become stop codons if they had incurred singlespeciﬁc mutations. The expected number of suchsubstitutions in the absence of selection was 72 ϫ0.07892 ϫ 1/3 ϭ 1.894. The ancestral sequence alsocontained 28 codons that could have become stopcodons as a result of either of two kinds of substi-tutions; the expected number of such substitutionswas 28 ϫ 0.07892 ϫ 2/3 ϭ 1.473. Consequently, theexpected number of premature stop codons in the ab-sence of selection is 3.367, and from the Poisson dis-tribution the probability of ﬁnding no premature stopcodons was eϪ3.367 ϭ 0.0345.2. All of the amino acid substitutions that occurred (out-side of the hypervariable region discussed below) inP. oviforme must be compatible with the normalfunction of tufA, because each of the substituted ami-no acids can be found at a comparable position in atleast one functional algal, cyanobacterial, or nonpho-tosynthetic bacterial gene in the alignment array ofDelwiche, Kuhsel, and Palmer (1995). The same istrue of all but six amino acid substitutions seen in P.obtusum. Both Polytoma sequences contain only con-servative amino acid substitutions, except for somenonconservative substitutions on the surface of theEF-Tu protein of P. obtusum. None of these aminoacid substitutions are likely to change the folding ofthe EF-Tu protein.3. Nucleotide substitution rates in the tufA sequences atﬁrst and second codon positions are much lower thanthe rates at third positions (table 3), a difference thatcan only be due to selection.4. Relative to all other species, the tufA sequences fromPolytoma and Chlamydomonas contain numerousamino acid substitutions, insertions, and deletions inthe hypervariable region. Despite the variability inthis region, we believe that it is compatible withfunctionality of the protein in both Polytoma speciesfor the following reasons. First, the hypervariable re-gion of P. oviforme is identical in length to that ofC. reinhardtii, and nearly identical in sequence, withonly two conserved amino acid differences betweenthe two species. The hypervariable region in P. ob-tusum differs from that in C. reinhardtii in 13 sub-stitutions and 3 gaps. However, this region is alsohypervariable and unusually long in functional EF-Tu proteins from C. reinhardtii (Baldauf and Palmer1990), C. humicola, and C. dysosmos, which are pho-tosynthetic and therefore have functional EF-Tu pro-teins. Second, the hypervariable region is on the out-side surface of the protein in functional domain 3,where amino acid changes or extra amino acidswould probably not affect the conformational chang-es that occur during catalysis (Berthtold et al. 1993),especially since the amino acid composition of thehypervariable region is even more hydrophilic in P.obtusum than in C. humicola and C. reinhardtii.Moreover, the face of domain 3 that interacts withother EF-Tu molecules (Kawashima et al. 1996) andwith the acceptor stem or T stem of the tRNA (Nis-sen et al. 1995) is opposite the hypervariable region.The Plastid rrn16 Genes in Polytoma RemainFunctional After the Loss of PhotosynthesisThe growth of P. uvella and of another member ofthe same clade, P. uvella 62-3 ϭ P. mirum, is inhibitedby the antibiotics erythromycin, streptomycin, and spec-tinomycin at 800, 400, and 50 mg/ml, respectively (datanot shown). Scherbel, Behn, and Arnold (1974) previ-ously found that growth of P. mirum is inhibited bystreptomycin. This antibiotic is known to inhibit chlo-roplast protein synthesis in C. reinhardtii at similar orlower (erythromycin) concentrations (Harris 1989).Spectinomycin sensitivity is especially interesting: it hasno known side effects, and Chlamydomonas mutants re-sistant to high concentrations have mutations only in therrn16 gene. These data suggest that Polytoma, likeChlamydomonas, synthesizes at least one essential pro-tein on plastid ribosomes which contain functional 16SrRNA molecules.Consistent with the antibiotic studies, an analysis ofthe primary and secondary structures of the 16S rRNAmolecules, inferred from the rn16 sequences, strongly sup-ports functionality of the molecules. Here we present onlya summary; the complete analysis is included with thesecondary-structure ﬁgures and sequence alignments in the
Accelerated Evolution of Leucoplast Genes 1817FIG. 4.—Relative synonymous codon usage values of codons inPolytoma obtusum (open bars) compared with Chlamydomonas hum-icola (ﬁlled bars). Separate graphs are shown for amino acids that aresixfold-, fourfold-, and twofold-degenerate. Stars represent codons forwhich the complementary anticodons are found in tRNAs encoded inplant chloroplast genomes.supplement and at the web site http://www.rna.icmb.utexas.edu/PUBLICATIONS/BIRKY/.1. The primary and secondary structures for the threePolytoma rRNA sequences contain all of the struc-tural elements present in the chloroplast rRNAs thatare functional, and, apart from a few insertions dis-cussed below, all of the nucleotide positions in thePolytoma sequences correspond to all of the positionspresent in these 70 functional 16S rRNA chloroplastsequences (ﬁgs. 1, 2, 3a, and 3d–f in the supplementand at our web site mentioned above; see also Gutell1994).2. Differences between the Polytoma sequences occurat positions that also vary in the functional SSUrRNAs in the nuclear genes of the Eucarya, Bacteria,and Archaea and in chloroplast and mitochondrialgenes (http://www.rna.icmb.utexas.edu/RDBMS/,http://www.rna.icmb.utexas.edu/CSI/BPFREQ/16S-MODEL-BP/, and ﬁgs. 2, 3a, and 3c in the sup-plement and at our web site mentioned above). Con-versely, positions that are conserved in the three phy-logenetic domains plus chloroplasts and mitochon-dria are also conserved in the Polytoma sequences.3. Base pairs were predicted with comparative sequenceanalysis (Gutell 1996; http://www.rna.icmb.utexas.edu/METHODS/); two aligned positions that changecoordinately are considered possible base pairs.These base pairs are highly conserved in the Poly-toma SSU rRNA and are consistent with function-ality. There are 70 base-paired positions at which thePolytoma sequences differ from the chloroplast con-sensus: of these, 24 are compensatory changes; 10involve an A·U or G·C interchange to a G·U basepair; and only 12 change an A·U, G·C, or G·U pairto a noncanonical pair (ﬁgs. 2 and 3 at our web sitementioned above). The number of noncanonical basepairs in the Polytoma SSU rRNA is approximatelythe same as in other rRNAs that are known to beactive in protein synthesis (unpublished data).4. A comparison of the sequences of P. obtusum and P.uvella reveals functional and structural constraintsacting on these 16S rRNA sequences since their com-mon ancestor; because this ancestor was nonphoto-synthetic, all of the differences between the two Po-lytoma isolates arose in nonphotosynthetic lineages.Among the 55 variable paired positions, 30 are in-volved in a covariation at base pairs predicted withcomparative sequence analysis (ﬁg. 4 in the supple-ment and at our web site mentioned above). Of thesepositions, 12 involve an exchange between Watson-Crick base pairs, while three involve a Watson-Crickbase pair with something else. Of the 25 paired po-sitions with a single change, 14 exchanged betweenAU or GC and GU. Since our covariation algorithmssearch for simultaneous changes regardless of thebase pair types, the fact that AU and GC are the mostfrequent pairs identiﬁed at positions that covary in-dicates that these genes are still undergoing positiveselection. In addition, several of the noncanonical co-variation base pairs have been experimentally veri-ﬁed (reviewed in Gutell 1996), suggesting that all ofthese covarying bases could be base-paired in thesecondary structure.5. Substitution rates were higher at paired sites than atunpaired sites, presumably because single substitu-tions at paired sites are detrimental but pairs of com-pensating changes are neutral (data not shown).There are three large uncharacteristic insertions inthe Polytoma sequences. The P. oviforme gene has anincompletely sequenced insertion larger than 438 bp be-tween positions 427 and 428 (E. coli numbering). Po-lytoma obtusum and P. uvella have insertions of 223 bpand Ͼ596 bp, respectively, between positions 746 and747. Finally, P. uvella has a second insertion of 85 ntbetween (E. coli) positions 138 and 140. All of theseinsertions occur in regions of the 16S rRNA that arevariable in sequence and not considered to be function-ally important. The ﬁrst, between positions 138 and 140(E. coli numbering) contains 85 extra nucleotides in P.
1818 Vernon et al.uvella; 75 of these are A or U. We believe that thisinsertion may be an internal transcribed spacer that isexcised after transcription for two reasons. First, thisextra sequence has a very high AU content, a charac-teristic of internal transcribed spacer (ITS) sequences.Second, an ITS has been found in the bacterium Cam-pylobacter sputorum 16S rRNA at position 220 (E. colinumbering; VanCamp et al. 1993; accession numberX67775), which is directly across the helix from ourpresumed ITS. The two other insertions occur at theends of helices and may be either introns or ITSs.Relaxed Selection Is at Least Partly Responsible forthe Accelerated EvolutionThe preceding sections show that the rrn16 andtufA genes in Polytoma remained subject to selection,and hence functional, long after the loss of photosyn-thesis. Although we cannot exclude the possibility thatthese genes lost their function recently, they were prob-ably still functional when we tested their antibiotic sen-sitivity. Consequently the increased rate of base pairsubstitution in the tufA and rrn16 genes in the P. uvellalineage is not due to complete loss of function. Wetherefore considered four other possible explanations fortheir accelerated evolution:1. An increased mutation rate would increase the rateof substitution, provided, of course, that the ﬁxationprobability was not reduced by the same factor. How-ever, an increase in mutation rate cannot explain ourdata because it would affect different functional re-gions of a gene to the same extent; in contrast, wefound that the substitution rate in the rrn16 gene wasgreater in regions coding for the stems of the rRNAthan in regions coding for the loops, and in tufA therewas a much greater increase in the synonymous ratethan in the nonsynonymous rate. These observationsshow that an increased mutation rate is not the solecause of the accelerated evolution in rrn16 and tufAbut do not rule it out as a contributing factor.2. There could be an increased level of adaptive or pos-itive selection by ﬁxation of mutations that were for-merly neutral or detrimental but are now advanta-geous. This explanation is unlikely because of thelarge number of such mutations that would be re-quired to explain the data. For example, parsimonyanalysis (data not shown) suggests that about 152base pair substitutions occurred in the rrn16 genealong the lineage leading to P. uvella from the mostrecent common ancestor of P. uvella and C. humi-cola, while only 35 occurred in the C. humicola lin-eage. The nonphotosynthetic lineage thus incurred152 Ϫ 35 ϭ 117 more substitutions that would haveto be adaptive; this is 77% of all of the substitutionsafter photosynthesis was lost. Moreover, it is proba-bly not a complete explanation because it is not likelyto increase the synonymous substitution rate. How-ever, it is possible that positive selection contributedin part to the increase in substitution rate.3. A reduced effective population size (Ne) would makenatural selection against detrimental mutations lesseffective. By making some mutations that were for-merly detrimental effectively neutral, it would in-crease the overall ﬁxation probability of new muta-tions. A decrease in the frequency of recombinationof plastid genes would reduce Ne as a result of theHill-Robertson effect (Birky and Walsh 1988), butthere is no reason to suspect this in nonphotosyn-thetic lineages. A more plausible reason for a de-crease in Ne is that an obligate heterotroph has fewerniches available to it, which might reduce Ne by de-creasing the total population size or increasing thevariance in offspring number. In any event, a reducedNe cannot explain our data by itself, because the ratioKAN/KAG was greater for kinds of substitutions sub-ject to weak selection (third positions in tufA; stemsin rrn16) than for those subject to relatively strongselection (ﬁrst and second positions in tufA; loops inrrn16). A theoretical analysis of the combined effectsof directional selection and drift, using Kimura’s(1957) equation for the ﬁxation probability of a mu-tation, showed that a decrease in effective populationsize by itself would cause KAN/KAG to be larger, notsmaller, for strongly selected detrimental mutationsthan for relatively weakly selected mutations (unpub-lished data). In addition, a reduced Ne would affectnuclear genes as well as organelle genes, whereas weobserved no signiﬁcant increase in substitution ratein the nuclear Rrn18 genes of P. obtusum and P.uvella. We cannot rule out the possibility that someof the observed acceleration in evolutionary rate isdue to a reduction in effective population size toosmall to be detected in the nuclear gene, but thiscannot be a complete explanation.4. Relaxed selection against detrimental mutations is aplausible explanation for the increase in substitutionrate in nonphotosynthetic lineages. This would re-quire that some mutations that were detrimental in achloroplast are neutral or effectively neutral in a leu-coplast and thus more likely to be ﬁxed. The rrn16and tufA genes are involved in translation of plastidproteins, and mutations in these genes are likely tobe detrimental if they directly or indirectly reduce therate or accuracy of protein synthesis. Selection wouldbe relaxed if heterotrophic algae could tolerate eithera lower rate of plastid protein synthesis or a higherproportion of amino acid substitutions in plastid pro-teins due to mistranslation. At least the ﬁrst of theseconditions is very likely to be met. Most of the chlo-roplast protein-coding genes code for proteins in-volved in photosynthesis, and all mutations affectingthese genes are selectively neutral after photosynthe-sis is lost in the ancestor of a nonphotosyntheticclade; as a result, they will accumulate mutations thatblock transcription or translation or will be deletedentirely. Leucoplasts thus have to translate only abouttwo-thirds as many proteins. Among the genes thatdo not have to be expressed in heterotrophic algae orplants is rbcL, which encodes one of the most abun-dant proteins in photosynthetic organisms, ribulosebisphosphate carboxylase. A number of photosyn-
Accelerated Evolution of Leucoplast Genes 1819Table 5Codon Adaptation Index (CAI) and Codon Bias Index(CBI) for tufA in Chlamydomonas and Polytoma SpeciesOrganism CAI(Cre) CAI(Cum) CBIChlamydomonas reinhardtii. . . .Polytoma oviforme . . . . . . . . . . .Chlamydomonas humicola . . . . .Polytoma obtusum. . . . . . . . . . . .0.4870.3590.4320.220.770.6880.6870.370.7750.7740.7220.455NOTE.—CAI(Cre) is calculated relative to the codon usage in C. reinhardtiipsbA gene; CAI(Cum) is calculated relative to the codon usage in the psbA genesof C. reinhardtii, Porphyra purpurea, Odontella sinensis, and Cyanophora par-adoxa.thetic genes show high codon bias characteristic ofhighly expressed genes.We hypothesized that the loss of photosynthesis inPolytoma permitted a lower rate of synthesis of elon-gation factor Tu because these heterotrophic organismscan tolerate a lower overall rate of protein synthesis thantheir photosynthetic relatives. This hypothesis predictsthat the tufA gene will show less codon bias in P. ob-tusum than in C. humicola, and probably less codon biasthan in other Chlamydomonas species. Codon bias isgenerally greater in highly expressed genes, probablybecause selection favors codons that are read by moreabundant tRNAs (e.g., Morton 1993, 1996; Sharp et al.1995). We calculated two measures of codon bias forthe tufA genes of the Chlamydomonadaceae. The CBImeasures the overall level of codon bias in a gene andranges from 0 to 1; the CAI (Sharp and Li 1987) mea-sures codon bias for a gene relative to one or more high-ly expressed genes that show high codon bias. Two ref-erence databases of codon use in a highly expressedplastid gene were provided by Brian Morton: the psbAgene of C. reinhardtii and the combined psbA genes ofC. reinhardtii, Porphyra purpurea, Cyanophora para-doxa, and Odontella sinensis. The CBI and both CAIvalues were lower for P. obtusum than for the two Chla-mydomonas species and P. oviforme (table 5).These data show that codon bias in tufA is reducedin P. obtusum compared with its closest photosyntheticrelative, C. humicola. The change in codon bias mightbe due to a change in relative mutation rates, e.g., tofavor AT pairs over GC. To test this possibility, we cal-culated expected numbers of the four codons in all four-fold-degenerate amino acids and in the fourfold-degen-erate codon set of sixfold-degenerate amino acids, as-suming that the expected proportions of A, T, G, and Cwere the averages of those found in the third positionsof all codons. The deviation of observed numbers fromthese expected numbers can be summarized by the chi-square value, which was much greater for C. humicola(64.3) than for P. obtusum (46.2). We also found thatthe percentages of AϩT were not signiﬁcantly differentin the ﬁrst ϩ second codon positions in C. humicola andP. obtusum (53.0% vs. 52.6% respectively; P k 0.05by Fisher’s exact test) but were signiﬁcantly different inthe third codon positions (60.5% vs. 75.9%, respective-ly; P K 0.001). Since mutation rates do not differ sys-tematically between codon positions, we conclude thatthe third-position base composition difference betweenChlamydomonas and Polytoma is due to selection onsynonymous codons rather than to mutation pressure.The reduced codon bias in Polytoma is presumably dueto relaxed selection against mutations that substituteless-used codons for those that are most used in the C.humicola lineage.A more detailed comparison of codon bias in P.obtusum and C. humicola suggests that the bias that isrelaxed in the nonphotosynthetic species is largely apreference for codons that can be read without wobble(a mismatched base pair between tRNA and the mRNAthird codon position). Several plant genomes have beencompletely sequenced and found to have a set of 29tRNAs. Gene sequences of 15 of these have beenmapped on the chloroplast genome of C. humicola (E.Boudreau, personal communication), and no tRNAgenes have been identiﬁed in any green alga other thanthose known from plant chloroplasts, so it is likely thatthese algae have the same set of chloroplast tRNAs asdo plants. Many codons require wobble to be read bythis set of tRNAs, and several require ‘‘superwobble,’’in which one tRNA reads A, G, C, or U in the thirdcodon position. We calculated the RSCU for all degen-erate amino acids (those using more than one codon) inP. obtusum and C. humicola. Figure 4 shows that muchof the difference in codon bias is in the sixfold-degen-erate amino acids leucine, serine, and arginine, whichpreferentially use codons that can be read without wob-ble. This preference is much greater for C. humicolathan for P. obtusum (chi-square test; 0.01 Ͼ P Ͼ 0.001).Similar analyses for fourfold- and twofold-degenerateamino acids show no signiﬁcant difference betweenthese species; both species show preference for no wob-ble over wobble in the twofold-degenerate amino acids(P ഠ 0.9) and for no wobble or normal wobble oversuperwobble in those fourfold-degenerate amino acidsthat require superwobble (0.8 Ͼ P Ͼ 0.7).In the case of rrn16, one can imagine that the re-duced load of protein synthesis in the leucoplast hasresulted in less intense selection for translation rate orﬁdelity. Alternatively, there may be a greater tolerancefor mutations that affect the rate of transcription of thegene, of posttranscriptional processing, or of assemblyof the ribosome. Relaxed selection for translational rateor ﬁdelity might result in decreased stability of the sec-ondary structure of the small-subunit rRNA. Comparedwith C. humicola, both P. uvella and P. obtusum hadfewer GC pairs and more AU pairs, with GC/AU ratiosof 1.80, 1.40, and 1.34, respectively. These numbers arecompatible with the hypothesis that the loss of photo-synthesis has been accompanied by a relaxation of se-lection for helix stability in the leucoplast small-subunitrRNA molecule. However, it is also compatible with adecrease in the GϩC content of the genome as a whole.A more deﬁnitive test of the hypothesis will requirecomparisons of secondary-structure stability in a largersample of nonphotosynthetic species and their closephotosynthetic relatives, using the nuclear small-subunitrRNA molecule as a control, together with data on theGϩC contents of the genomes.
1820 Vernon et al.The bulk of the evidence suggests that the in-creased rate of base pair substitution in the tufA andrrn16 genes of Polytoma is due at least in part to relaxedselection, such that mutations that reduce the rate orﬁdelity of translation are less likely to be eliminated.Rigorous tests of this hypothesis will require more de-tailed analyses of a larger number of sequences of theseand other plastid expression genes.Polytoma oviforme May Have Lost PhotosynthesisMore RecentlyThe tufA gene of P. oviforme shows codon biassimilar to that of several close green relatives, and nei-ther the rrn16 gene nor the tufA gene shows an increasedsubstitution rate in the lineage leading to P. oviformerelative to green lineages. In contrast, the expressiongene sequences that we sampled from the P. uvella cladeall show signiﬁcant rate increases compared with greenrelatives, and tufA from P. obtusum (the only protein-coding gene we obtained from the P. uvella clade)shows minimal codon bias. These data are consistentwith the hypothesis that the P. uvella clade may haveresulted from a more ancient loss of photosynthesis thanthe P. oviforme lineage, and only the P. uvella clade hasbeen evolving without photosynthesis long enough forthe consequences of relaxed selection to be evident.Accelerated Evolution of Leucoplast Expression Genesin Other OrganismsSmaller increases in evolutionary rates have beendemonstrated in the leucoplast expression genes rrn16,rrn23, and tufA, as well as in the rbcL gene, of theheterotrophic euglenoid alga Astasia longa (Siemeister,Buchholz, and Hachtel 1990; Siemeister and Hachtel1990a, 1990b). An increase in the substitution rate ofthe rrn16 genes of some nonphotosynthetic angiospermswas reported by Nickrent, Duff, and Konings (1997),although it was not veriﬁed by relative-rate tests. Theincrease was accompanied by an increase in AϩT con-tent in the genes and consequently by an increase indestabilizing AU pairs in the rRNA, such as we ob-served. The authors did not compare the rates in stemsand loops. Nickrent and Starr (1994) also reported anincreased substitution rate in the nuclear Rrn18 genes ofa different set of species of holoparasitic angiospermsand presented evidence that the increase could not beexplained by a decrease in generation time or effectivepopulation size. In contrast, we observed no increase inthe evolutionary rate of Rrn18 in Polytoma (Rumpf etal. 1996) or Polytomella (unpublished data). The holo-parasitic angiosperm Epifagus virginiana was subjectedto an intensive analysis of evolutionary rates in leuco-plast genes; rate increases were detected in rrn16 andrrn23 and the pooled data for 17 tRNAs and 15 ribo-somal proteins (Wolfe et al. 1992). In the ribosomal pro-teins, accelerated evolution was seen in both nonsynon-ymous and synonymous substitutions; in contrast to ourresults, the increase was greater for nonsynonymoussubstitutions. Epifagus also differed from Polytoma inshowing no change in codon bias (Morden et al. 1991).Wolfe et al. (1992) proposed that the increase in non-synonymous substitutions was probably due to relaxedselection for both the rate and ﬁdelity of protein syn-thesis, while the increase in synonymous substitutionsprobably resulted from an increased mutation rate. How-ever, alternative explanations were not ruled out. In par-ticular, both the synonymous and the nonsynonymousrate increase might be entirely due to a decreased effec-tive population size, which under some circumstancescan increase the rate of detrimental substitutions morewhen they are under stronger selection (unpublisheddata). dePamphilis, Young, and Wolfe (1997) found thatthe substitution rate of the ribosomal protein gene rps2was signiﬁcantly accelerated in some, but not all, hol-oparasitic (nonphotosynthetic) angiosperm lineages inthe Orobanchaceae and Sdrophulariaceae. In some cas-es, there was a signiﬁcant increase in synonymous butnot nonsynonymous substitutions, and in some othercases, the reverse held. dePamphilis, Young, and Wolfe(1997) proposed that increases in nonsynonymous sub-stitution rates are probably due to reduced functionalconstraint when abundant photosynthetic proteins do nothave to be synthesized. The increases in synonymoussubstitutions were attributed to increased mutation rates.However, no formal analyses of the data were given insupport of these conclusions.The independent losses of photosynthesis in a num-ber of different plants and algae provide biologists witha series of natural experiments in which selection hasbeen reduced or eliminated to different extents in geneswith different functions. Future comparative studies ofleucoplast expression and photosynthetic gene sequenc-es from Polytoma and Polytomella will help to unravelthe roles of mutation and selection in the molecular evo-lution of plastid genes. The data to date suggest thataccelerated evolution of plastid expression genes is ageneral consequence of the loss of photosynthesis andthat the rate of evolution of expression genes is stronglyinﬂuenced by the rate of protein synthesis that they mustsustain.Supplementary MaterialThe DNA sequences have been deposited inGenBank under accession numbers AF352839 (P. ob-tusum tufA), AF352840 (P. oviforme tufA), AF352838(C. humicola tufA), AF397587 (P. obtusum rrn16),AF397589 (P. uvella rrn16), AF397588 (P. oviformerrn16), AF397590 (C. reinhardtii rrn16), andAF397586 (C. humicola rrn16). Additional supplemen-tary materials on the Molecular Biology and Evolutionweb site include tufA and rrn16 sequence alignmentsin sequential GenBank format and in interleaved PrettyPrint format, and the complete argument and additionaldata showing that the rrn16 gene is functional inPolytoma.AcknowledgmentsWe are grateful to Brian Morton for supplying hisMacintosh Pascal program and reference databases forcalculating measures of codon bias, and to Aurora Ned-
Accelerated Evolution of Leucoplast Genes 1821elcu for allowing us to analyze her tufA sequence fromP. uvella before it was published. Sally Otto analyzedKimura’s (1957) equation to verify that reducing Ne af-fects strongly selected sites more than weakly selectedsites. Primers were kindly provided by Paul Fuerst andJeffrey Palmer. The large tufA database was kindly pro-vided by Charles Delwiche. Other members of the Birkylab at Ohio State University, notably Pamela Mackowskiand Stacy Seibert, assisted in numerous ways. We aregrateful to Jennifer Wernergreen, Aurora Nedelcu, andtwo anonymous reviewers for helpful comments on ear-lier manuscripts. D.V. submitted the data and a prelim-inary analysis in a thesis in partial fulﬁllment of therequirements for the Ph.D. degree at the Ohio State Uni-versity. This work was supported in part by researchgrants from NIH (GM34094) and NSF (BSR-9107069)to C.W.B. and from NIH (GM48207) to R.R.G., and byfunds from the Ohio State University, the University ofArizona, and the University of Texas.LITERATURE CITEDBALDAUF, S. L., and J. D. PALMER. 1990. Evolutionary transferof the chloroplast tufA gene to the nucleus. Nature 344:262–265.BERTHTOLD, H., L. RESHETNIKOVA, C. O. A. REISER, N. K.SCHIRMER, M. SLPRINZL, and R. HILGENFELD. 1993. Crystalstructure of active elongation factor Tu reveals major do-main rearrangements. Nature 365:126–132.BIRKY, C. W. JR., and J. B. WALSH. 1988. Effects of linkageon rates of molecular evolution. Proc. Natl. Acad. Sci. USA85:6414–6418.DELWICHE, C., M. KUHSEL, and J. D. PALMER. 1995. Phylo-genetic analysis of tufA sequences indicates a cyanobacter-ial origin of all plastids. Mol. Phylogenet. Evol. 4:110–128.DEPAMPHILIS, C. W., and J. D. PALMER. 1990. Loss of photo-synthetic and chlororespiratory genes from the plastid ge-nome of a parasitic ﬂowering plant. Nature 348:337–339.DEPAMPHILIS, C. W., N. D. YOUNG, and A. D. WOLFE. 1997.Evolution of plastid gene rps2 in a lineage of hemiparasiticand holoparasitic plants: many losses of photosynthesis andcomplex patterns of rate variation. Proc. Natl. Acad. Sci.USA 94:7367–7372.ETTL, H. 1976. Die Gattung Chlamydomonas Ehrenberg. Beih.Nova Hedwigia 49:1–1122.FELSENSTEIN, J. 1993. PHYLIP (phylogeny inference package).Version 3.56. Distributed by the author, Department of Ge-netics, University of Washington, Seattle.GILBERT, D. G. 1992. SeqApp, a biological sequence editorand analysis program for Macintosh computers. Availablevia gopher or anonymous ftp to ftp.biol.indiana.edu.GILLHAM, N. W. 1994. Organelle genes and genomes. OxfordUniversity Press, New York.GORDON, J., R. RUMPF, S. L. SHANK, D. VERNON, and C. W.BIRKY JR. 1995. Sequences of the rrn18 genes of Chla-mydomonas humicola and C. dysosmos are identical, inagreement with their combination in the species C. applan-ata (Chlorophyta). J. Phycol. 31:312–313.GRANT, D. M., N. W. GILLHAM, and J. E. BOYNTON. 1980.Inheritance of chloroplast DNA in Chlamydomonas rein-hardtii. Proc. Natl. Acad. Sci. USA 77:6067–6071.GUTELL, R. R. 1994. Collection of small subunit (16S- and16S-like) ribosomal RNA structures: 1994. Nucleic AcidsRes. 22:3502–3507.———. 1996. Comparative sequence analysis and the structureof 16S and 23S rRNA. Pp. 111–128 in R. A. ZIMMERMANand A. E. DAHLBERG, eds. Ribosomal RNA. Structure, evo-lution, processing, and function in protein biosynthesis.CRC Press, New York.HARRIS, E. H. 1989. The Chlamydomonas sourcebook. Aca-demic Press, New York.JUKES, T. H., and C. R. CANTOR. 1969. Evolution of proteinmolecules. Pp. 21–123 in H. N. MUNRO, eds. Mammalianprotein metabolism. Academic Press, New York.KAWASHIMA, T., C. BERTHET-COLOMINAS, M. WULFF, S. CU-SACK, and R. LEBERMAN. 1996. The structure of the E. coliEF-Tu/EF-Ts complex at 2.5 A resolution. Nature 379:511–518.KIMURA, M. 1957. Some problems of stochastic processes ingenetics. Ann. Math. Stat. 28:882–901.———. 1980. A simple method for estimating evolutionaryrates of base substitutions through comparative studies ofnucleotide sequences. J. Mol. Evol. 16:111–120.LANG, N. J. 1963. Electron-microscopic demonstration of plas-tids in Polytoma. J. Protozool. 10:333–339.LEMIEUX, C., C. OTIS, and M. TURMEL. 2000. Ancestral chlo-roplast genome in Mesostigma viride reveals an earlybranch of green plant evolution. Nature 249:649–652.LUSSON, N. A., P. M. DELAVAULT, and P. A. THALOUARN.1998. The rbcL gene from the non-photosynthetic parasiteLathraea clandestina is not transcribed by a plastid-encodedRNA polymerase. Curr. Genet. 34:212–215.MAIDAK, B. L., N. LARSEN, M. J. MCCAUGHEY, R. OVERBEEK,G. J. OLSEN, K. FOGEL, J. BLANDY, and C. R. WOESE. 1994.The Ribosomal Database Project. Nucleic Acids Res. 22:3485–3487.MORDEN, C. W., K. H. WOLFE, C. W. DEPAMPHILIS, and J. W.PALMER. 1991. Plastid translation and transcription genesin a non-photosynthetic plant: intact, missing and pseudogenes. EMBO J. 10:3281–3288.MORTON, B. R. 1993. Chloroplast DNA codon use: evidencefor selection at the psbA locus based on tRNA availability.J. Mol. Evol. 37:273–280.———. 1996. Selection on the codon bias of Chlamydomonasreinhardtii chloroplast genes and the plant psbA gene. J.Mol. Evol. 43:28–31.MUSE, S. V., and B. S. WEIR. 1992. Testing for equality ofevolutionary rates. Genetics 132:269–276.NEDELCU, A. M. 2001. Complex patterns of plastid 16S rRNAgene evolution in nonphotosynthetic green algae. J. Mol.Evol. (in press).NICKRENT, D. L., R. J. DUFF, and D. A. M. KONINGS. 1997.Structural analyses of plastid-derived 16S rRNAs in holo-parasitic angiosperms. Plant Mol. Biol 34:731–743.NICKRENT, D. L., and E. M. STARR. 1994. High rates of nu-cleotide substitution in nuclear small-subunit (18S) rDNAfrom holoparasitic ﬂowering plants. J. Mol. Evol. 39:62–70.NISSEN, P., M. KJELDGAARD, S. THIRUP, G. POLEKHINA, L.RESHETNIKOVA, B. F. C. CLARK, and J. NYBORG. 1995.Crystal structure of the ternary complex of Phe-tRNA, EF-Tu, and a GTP analog. Science 270:1464–1472.RUMPF, R., D. VERNON, D. SCHREIBER, and C. W. BIRKY JR.1996. Evolutionary consequences of the loss of photosyn-thesis in Chlamydomonadaceae: phylogenetic analysis ofRrn18 (18S rDNA) in 13 Polytoma strains (Chlorophyta).J. Phycol. 32:119–126.SAMBROOK, J., E. F. FRITSCH, and T. MANIATIS. 1989. Molec-ular cloning: a laboratory manual. Cold Spring Harbor Lab-oratory Press, Cold Spring Harbor, N.Y.
1822 Vernon et al.SARICH, V. M., and A. C. WILSON. 1973. Generation time andgenomic evolution in primates. Science 179:1144–1147.SCHERBEL, G., W. BEHN, and C. G. ARNOLD. 1974. Untersu-chungen zur genetischen Funktion des farblosen Plastidenvon Polytoma mirum. Arch. Microbiol. 96:205–222.SHARP, P. M., M. AVEROF, A. T. LLOYD, G. MATASSI, and J.F. PEDEN. 1995. DNA sequence evolution: the sounds ofsilence. Philos. Trans. R. Soc. Lond. B Biol. Sci. 349:241–247.SHARP, P. M., and W.-H. LI. 1987. The codon adaptation in-dex—a measure of directional synonymous codon usagebias, and its potential applications. Nucleic Acids Res. 15:1281–1295.SIEMEISTER, G., C. BUCHHOLZ, and W. HACHTEL. 1990. Genesfor the plastid elongation factor Tu and ribosomal proteinS7 and six tRNA genes on the 73 kb DNA from Astasialonga that resembles the chloroplast DNA of Euglena. Mol.Gen. Genet. 220:425–432.SIEMEISTER, G., and W. HACHTEL. 1990a. Organization andnucleotide sequence of ribosomal RNA genes on a circular73 kbp DNA from the colourless ﬂagellate Astasia longa.Curr. Genet. 17:433–438.———. 1990b. Structure and expression of a gene encodingthe large subunit of ribulose-1,5-bisphosphate carboxylase(rbcL) in the colourless euglenoid ﬂagellate Astasia longa.Plant Mol. Biol. 14:825–833.SIU, C.-H., K.-S. CHIANG, and H. SWIFT. 1976. Characteriza-tion of cytoplasmic and nuclear genomes in the colorlessalga Polytoma. III. Ribosomal RNA cistrons of the nucleusand leucoplast. J. Cell Biol. 69:383–392.SWOFFORD, D. L. 1998. PAUP*: phylogenetic analysis usingparsimony (*and many other methods). Version 4. Sinauer,Sunderland, Mass.TAMURA, K. 1992. Estimation of the number of nucleotide sub-stitutions when there are strong transition-transversion andGϩC content biases. Mol. Biol. Evol. 9:678–687.TURMEL, M., C. OTIS, and C. LEMIEUX. 1999. The completechloroplast DNA sequence of the green alga Nephroselmisolivacea: insights into the architecture of ancestral chloro-plast genomes. Proc. Natl. Acad. Sci. USA 96:10248–10253.VANCAMP, G., Y. VAN DE PEER, J. M. NEEFS, P. VANDAMME,and R. DEWACHTER. 1993. Presence of internal transcribedspacers in the 16S and 23S ribosomal RNA genes of Cam-pylobacter. Syst. Appl. Microbiol. 16:361–368.VERNON, D. 1996. Evolutionary consequences of the loss ofphotosynthesis in the nonphotosynthetic chlorophyte algaPolytoma. Dissertation, Ohio State University, Columbus.VERNON-KIPP, D., S. A. KUHL, and C. W. J. BIRKY. 1989.Molecular evolution of Polytoma, a non-green chlorophyte.Pp. 284–286 in C. T. BOYER, J. C. SHANNON, and R. C.HARDISON, eds. Physiology, biochemistry, and genetics ofnongreen plastids. American Society of Plant Physiologists,Rockville, Md.WIMPEE, C. F., R. MORGAN, and R. L. WROBEL. 1992a. Anaberrant plastid ribosomal RNA gene cluster in the rootparasite Conopholis americana. Plant Mol. Biol. 18:275–285.———. 1992b. Loss of transfer RNA genes from the plastid16S-23S ribosomal RNA gene spacer in a parasitic plant.Curr. Genet. 21:417–422.WOLFE, A. D., and C. W. DEPAMPHILIS. 1998. The effect ofrelaxed functional constraints on the photosynthetic generbcL in photosynthetic and nonphotosynthetic parasiticplants. Mol. Biol. Evol. 15:1243–1258.WOLFE, K. H., C. W. MORDEN, S. C. EMS, and J. D. PALMER.1992. Rapid evolution of the plastid translational apparatusin a nonphotosynthetic plant: loss or accelerated sequenceevolution of tRNA and ribosomal protein genes. J. Mol.Evol. 35:304–317.WOLFE, K. H., C. W. MORDEN, and J. D. PALMER. 1992. Func-tion and evolution of a minimal plastid genome from a non-photosynthetic parasitic plant. Proc. Natl. Acad. Sci. USA89:10648–10652.WU, C.-I., and W.-H. LI. 1985. Evidence for higher rates ofnucleotide substitution in rodents than in man. Proc. Natl.Acad. Sci. USA 82:1741–1745.YOUNG, D. Y., and C. W. DEPAMPHILIS. 2000. Purifying selec-tion detected in the plastid gene matK and ﬂanking ribo-zyme regions within a group II intron of nonphotosyntheticplants. Mol. Biol. Evol. 17:1933–1941.DAVID RAND, reviewing editorAccepted May 30, 2001