Introduction

Several algal groups are known to also include leukoplast-bearing taxa, i.e., genera that lost photosynthesis and returned to a heterotrophic life but retained their plastids. Among these groups are the Euglenophyceae with Euglena longa (Gockel and Hachtel 2000; Marin et al. 2003), the Viridiplantae with, e.g., Polytoma and Prototheca in the chlorophytes (Rumpf et al. 1996; Nedelcu 2001) and several holoparasitic lineages in the streptophytes (e.g., the plant families Orobanchaceae/Scrophulariaceae, Rafflesiaceae, and Balanophoraceae [Nickrent and Starr 1994; dePamphilis et al. 1997]), the Heterokontophyta with several genera, e.g., Pteridomonas and Ciliophrys (Sekiguchi et al. 2002), and the Cryptophyceae (Sepsenwol 1973). In the Apicomplexa, the photoautotrophic lineages are presumably completely extinct, i.e., only the leukoplast-bearing taxa survived and changed to a parasitic life (reviewed in Wilson 2002). Thus, leukoplastidy occurs in algae with plastids derived from both primary and secondary endosymbioses. In Viridiplantae, Rhodophyta, and Glaucocystophyta a phagotrophic eukaryote ingested a cyanobacterium as prey and reduced it to a plastid with two surrounding membranes. In Chlorarachniophyceae, Euglenophyceae, most photoautotroph dinoflagellates, Haptophyta, Heterokontophyta, and Cryptophyceae, a phagotrophic eukaryote engulfed a eukaryote—a red or green alga—and reduced it to a complex (secondary) plastid with more than two envelope membranes (reviewed in McFadden 2001).

Cryptophyceae occur ubiquitously in freshwater, brackish, and marine habitats as biflagellate photosynthetic unicells and are pigmented brown, red, or blue. They acquired their complex plastid from an ingested red alga (Douglas et al. 2001). The Cryptophyceae is one of two known groups of organisms, in which the nucleus of the endosymbiont was retained as a nucleomorph with a highly reduced genome in the periplastidial space between the outer and the inner pairs of plastid membranes (Douglas et al. 2001). In the other group—the Chlorarachniophyceae—the complex plastid derived from a green alga (McFadden et al. 1994a, Archibald et al. 2003). The Cryptophyceae retained the accessory pigment phycoerythrin from the ancestral red algal light harvesting complex, which was transferred into the lumina of the thylakoids (Gantt et al. 1971; Apt et al. 1995). The phycoerythrin evolved into seven different types, three red phycoerythrins and four blue “phycocyanins,” resulting in the astounding diversity of colors in the cryptophytes (Hill and Rowan 1989; Glazer and Wedemayer 1995). The type of biliprotein was shown to be congruent with clades in phylogenetic trees inferred from nuclear or nucleomorph SSU rDNA (Marin et al. 1998; Hoef-Emden et al. 2002).

Although exclusively restricted to freshwater habitats, the genus Cryptomonas is probably the most abundant of all cryptophyte genera (Gervais 1998; Hoef-Emden and Melkonian 2003). The photoautotrophic members of the genus have brownish cells characterized by the presence of a purple phycoerythrin with an absorption maximum at 566 nm and a cell invagination with an elongated slit-like opening lined with extrusomes (i.e., the furrow-gullet system [Hill 1991]).

A few colorless genera have been assigned to the cryptophytes in the past, but only two (namely, Goniomonas and “Chilomonas”) were confirmed to belong to this group by molecular phylogenetic analyses. Goniomonas is an aplastidic phagotroph which branched off basally and, thus, may have diverged prior to the acquisition of the plastid (McFadden et al. 1994b). “Chilomonas,” on the other hand, contains a leukoplast with starch synthesis in the periplastidial compartment and resembles in cell shape and ultrastructure a campylomorph of Cryptomonas (i.e., cells with a furrow–gullet system, periplast with a sheet-like inner periplast component, and flagellar root system with a keeled rhizostyle [Roberts et al. 1981; Grim and Staehelin 1984]). Indeed “Chilomonas” was shown to be a colorless representative of Cryptomonas in molecular phylogenetic analyses of nuclear and nucleomorph SSU rDNA (Marin et al. 1998; Hoef-Emden et al. 2002; Hoef-Emden and Melkonian 2003). In previous phylogenetic analyses, only two—always monophyletic—colorless Cryptomonas strains were included (Hoef-Emden and Melkonian 2003).

However, in this study the polyphyly of colorless Cryptomonas strains is shown in phylogenetic analyses with a larger number of taxa. Molecular phylogenetic analyses were performed on DNA sequences of the nuclear as well as the nucleomorph ribosomal RNA operons (17 new nucleomorph SSU rDNA sequences and 15 new nuclear ITS2 and partial LSU rDNA sequences). In addition, the differences in branch lengths and in A+T content between nuclear and nucleomorph sequences were examined to provide an indication of evolutionary rates.

Materials and Methods

Maintenance of Algal Cultures

Colorless Cryptomonas strains were obtained from the Culture Collection of Algae and Protozoa (CCAP, UK), the Algal Culture Collection Melkonian at the University of Cologne (M, Germany), and the Sammlung von Algenkulturen at the University of Göttingen (SAG, Germany). Strains M1634 and M2452 were isolated by the author from field material and submitted to the Algal Culture Collection Melkonian (for a list of newly sequenced strains, see Table 1; a complete list of strains with culture collection source, origin, and isolator is given in Supplementary Material, Table S1). All strains were grown at 15°C. Photoautotrophic strains were grown in WARIS-H freshwater medium (Kies 1967; McFadden and Melkonian 1986) and subjected to a 14/10-h light/dark cycle. Except for strains CCAP 977/2a, M1634, and M2452, all heterotrophic strains were grown in the dark in biphasic soil/water medium with a small lentil or one-eighth of a pea for supply of organic substances. Strains CCAP 977/2a, M1634, and M2452 were grown in Waris-H with 1 ml of bacterial growth medium/L (stock solution: 1.6 g Bacto peptone, 0.2 g glucose, 0.2 g meat extract, 0.2 g yeast extract in 200 ml deionized water). Strains that are not available in culture collections can be obtained from the author upon request.

Table 1 List of newly sequenced strains with accession numbers
Full size table

DNA Extraction, PCR, and Bidirectional Sequencing

Total genomic DNA was isolated with the DNeasy Plant Mini Kit according to the manufacturer’s protocol (Qiagen, Hilden, Germany). Specific combinations of primers (see Tables S2 and S3 in Supplementary Material) were used to amplify either nucleomorph SSU rDNA sequences or combined products of the nuclear internal transcribed spacer 2 with the first 800 bp of the 5′-terminus of the nuclear LSU rDNA by PCR (Saiki et al. 1988). PCRs and bidirectional sequencing were performed as previously described (Hoef-Emden and Melkonian 2003).

Phylogenetic Analyses

The sequences of the nucleomorph SSU rDNA and of the nuclear ITS2 and partial LSU rDNA were aligned with the alignment editor SeaView (Galtier et al. 1996) and refined manually according to secondary structure (for accession numbers of new sequences, see Table 1; a complete list of strains with accession numbers is given in Table S1 of Supplementary Material; alignments deposited in the EMBL-Align database under accession nos. ALIGN_000701 and ALIGN_000702). Phylogenetic analyses were performed with six different data sets (rooted nucleomorph SSU rDNA, 1489 positions; unrooted nucleomorph SSU rDNA, 1564 positions; ITS2 only, 286 positions; LSU only, 883 positions; ITS2 combined with LSU [termed ITSLSU], 1169 positions; concatenated data set [termed Concat; comprising nucleomorph SSU rDNA, nuclear ITS2, and LSU], 2733 positions). All data sets, except for the rooted nucleomorph SSU rDNA data set (48 taxa), had identical taxon sampling consisting of 41 strains. At least two photoautotrophic representatives, if available, for each of the different clades found in the genus Cryptomonas (for species designations of clades, see Hoef-Emden and Melkonian 2003) and additional sequences from colorless strains were chosen (see Table 1). The outgroup for the rooted nucleomorph SSU rDNA data set included seven strains assigned to other cryptophyte genera (Chroomonas sp. strain M1312, accession no. AJ420678; Chroomonas sp. strain M1318, accession no. AJ420679; Chroomonas sp. strain M1481, accession no. AJ420680; Chroomonas sp. strain M1703, accession no. AJ420681; Chroomonas sp. strain SAG B980-1, accession no. AJ420677; Hemiselmis rufescens strain CCMP 439, accession no. AJ420690; Hemiselmis virescens strain CCMP 443, accession no. AJ420691). Nonalignable regions were excluded prior to the analyses. All data sets passed the test for homogeneity of base frequencies across taxa.

Modeltest 3.06 was used to infer the evolutionary model best fitting the data for each data set (Posada and Crandall 1998) (Table 2). The test of homogeneity of base frequencies across taxa, maximum likelihood, maximum parsimony, and distance analyses, and the corresponding bootstrap analyses were performed with Paup 4.0b10 (portable version running under Solaris 8 or Debian GNU/Linux 3.0 [Swofford 2002]). Bayesian analyses were done with MrBayes 3.0B4 (Ronquist and Huelsenbeck 2003). TreeView 1.66 for Mac was used to convert tree files into graphics (Page 1996).

Table 2 Support for clades in the different unrooted data sets
Full size table

Distance/neighbor-joining analyses were performed using the maximum likelihood estimators inferred from the hierarchical likelihood-ratio tests by Modeltest 3.06. Also, these parameters were used for heuristic searches under the maximum likelihood criterion. Fifty percent majority rule consensus trees under the maximum parsimony and the distance criterion were determined from 1000 bootstrap replicates. For maximum likelihood bootstrap analyses, 200 subsamples were used.

Bayesian analyses were run for at least 3.5 million generations; trees were sampled every 100 generations; burn-in was determined according to the sump-plot output. The two combined data sets were subdivided into two (ITSLSU) or three (Concat) partitions corresponding to nuclear ITS2, partial nuclear LSU, and nucleomorph SSU rDNA. All parameters except for tree topology (base frequencies, rate matrix, proportion of invariable sites/covarion, and gamma distribution) were unlinked (i.e., calculated separately for each DNA partition).

For a likelihood summation analysis (Yang 1996; Bapteste et al. 2002), the 1000 best likelihood trees were saved for all five unrooted data sets. All tree topologies were combined to a large treefile consisting of 5000 trees. The log likelihood scores for the nuclear ITS2, the nuclear partial LSU rDNA, and the nucleomorph SSU rDNA were determined for all 5000 trees. The three scorefiles were then imported into a spreadsheet program and the sum of the log likelihoods was computed for each of the 5000 trees. The best tree was determined by sorting the tree topologies according to their sums of log likelihood.

Approximately Unbiased and WeightedShimodaira–Hasegawa Tests

To test tree topologies with the approximately unbiased and weighted Shimodaira–Hasegawa (2001) tests (program Consel 0.1h), maximum likelihood trees inferred from all unrooted data sets were combined to a treefile with six trees, including two trees from the ITS2 data set (NM, ITS2a, ITS2b, LSU, ITSLSU, Concat). In addition, constraint trees forcing the colorless strains to be monophyletic were calculated for each data set and added to the treefile. Hence, the first six tree topologies were the same for each data set, whereas the last tree(s)—the constraint tree(s)—differed in topology. The ITS2 data set yielded three constraint trees that were included in the tests. For each data set, the total log likelihoods and sitewise log likelihoods of the tree topologies were computed with Paup and then subjected to approximately unbiased and weighted Shimodaira–Hasegawa tests.

Secondary Structure of Selected Parts of theNucleomorph SSU rRNA

Secondary structures of two loops of the variable region V4 in the nucleomorph SSU rRNA were inferred using the mfold server (Zuker 2003; http://www.bioinfo.rpi.edu/applications/mfold/old/rna/). A complete secondary structure of the nucleomorph SSU rRNA of strain CCAP 979/61 is shown in Hoef-Emden and Melkonian 2003.

Results

Molecular Phylogenetic Analyses

Seven highly supported clades were unambiguously recovered in the concatenated data set in the analyses: C. curvata, C. marssonii, C. ovata, C. pyrenoidifera, C. tetrapyrenoidosa, LB, and NoPyr (LB—long-branch; NoPyr—small-celled Cryptomonas strains without pyrenoids; Fig. 1). The clades C. curvata, C. marssonii, C. ovata, C. tetrapyrenoidosa, and NoPyr were also significantly supported in the ITS2, LSU, ITSLSU, and NM phylogenies (topologies not shown; evolutionary models and support values for clades listed in Table 2).

Figure 1
figure 1

Unrooted maximum likelihood tree of concatenated sequences of the nucleomorph SSU rDNA, nuclear ITS2, and partial nuclear LSU rDNA. Seven clades were recovered. Three colorless lineages were found in two clades, LB and NoPyr. The combination of sequences with differing evolutionary rates resulted in a leveling of branch lengths. Interrupted outline: Clade LB not supported in ITS2, LSU, and ITSLSU phylogenies. Support values in order from left to right: maximum parsimony bootstrap/neighbor-joining bootstrap/maximum likelihood bootstrap/posterior probability. Evolutionary model: GTR + I + Γ. −lnL = 16,220.1. Strain designations in boldface, colorless strains; scale bar, substitutions per site.

Full size image

In the clades LB and NoPyr, colorless strains were found. C. paramaecium within the LB clade consisted exclusively of leukoplast-bearing strains that were subdivided into three subclades (C. paramaecium subclades A, B, and C; Fig. 1), whereas the basally diverging strains of clade LB were photosynthetic (C. borealis, C. gyropyrenoidosa, and C. lundii; Fig. 1). In the ITS2 phylogeny, clade LB was not present. Also, no support for it was found in the LSU and ITSLSU phylogenies (Table 2).

Clade NoPyr also contained colorless as well as photosynthetic strains (Fig. 1). The colorless strain SAG 977-2f was always part of the highly supported subclade C in clade NoPyr, positioned as a sister to the photosynthetic strain M2088 and with another photosynthetic strain branching off basally (strain CCAC 0109; Fig. 1 and Table 2). A second colorless strain, M1634, either was not associated with a subclade or clustered with moderate to low support with subclade B (Fig. 1 and Table 2).

In the rooted nucleomorph SSU rDNA phylogeny, the root inserted between clade NoPyr and all other ingroup clades, but without resolution of the basal branching order (not shown). Also, the resolution of the internal branching order did not improve in the unrooted analyses.

Increased evolutionary rates were found in the colorless strains, but also in the closely related photosynthetic taxa of the mixed clades (Fig. 1). However, evolutionary rates in the data sets differed markedly. In the nucleomorph SSU rDNA trees, the most extreme branch lengths were found in clade LB with C. paramaecium and the photosynthetic C. borealis, C. gyropyrenoidosa, and C. lundii strains, whereas in the mixed photosynthetic/colorless clade NoPyr, only a slight increase in evolutionary rates was found (not shown). In the nuclear phylogenies, accelerated evolutionary rates were most prominent in clade NoPyr. In clade LB, pronounced evolutionary rates were found in single strains (ITS2, strains C. gyropyrenoidosa and C. lundii; LSU, C. paramaecium; topologies not shown). Thus, in the concatenated data set of the nuclear ITS2, partial nuclear LSU rDNA, and nucleomorph SSU rDNA, the differing evolutionary rates of the three data sets resulted in a tree with long branches in clade LB, predominantly due to the phylogenetic signal in the nucleomorph SSU rDNA, as well as in the nuclear DNA sequences (for the NoPyr clade; Fig. 1).

The best tree according to the likelihood summation method was tree 78 of the 1000 likelihood trees inferred from the concatenated data set. It showed the same topology as the maximum likelihood tree depicted in Fig. 1 and, thus, was not included in the approximately unbiased and weighted Shimodaira–Hasegawa tests and is not discussed further. The 23 best trees according to likelihood summation ranking were trees that also resulted from the concatenated data set.

Approximately Unbiased (AU) and WeightedShimodaira–Hasegawa (wSH) Tests

Of all tested tree topologies, only the one inferred from the concatenated sequences of nuclear ITS2, nuclear partial LSU rDNA, and nucleomorph SSU rDNA was not rejected as significantly worse. However, in all data sets the constraint trees with a monophyletic clade of colorless strains were rejected as significantly worse; mostly they were ranked as the worst trees in the test (Table 3). In most data sets, as expected, the tree topology derived from the tested data set was the best, except for the AU and wSH test of the ITS2 data set. No log likelihood difference was present between the trees ITS2a and ITSLSU, but the tree inferred from the ITSLSU data set was ranked in top position (Table 3). When tested with the Concat and the nucleomorph SSU rDNA data set, both nuclear ITS2 topologies were even worse than the constraint trees (Table 3).

Table 3 Results of the Approximately Unbiased (AU) and weighted Shimodaira–Hasegawa (WSH) tests (log likelihood differences and probabilities)
Full size table

A+T Content

The nucleomorph SSU rDNA showed an obvious bias toward ran increased A+T content in the highly variable regions (Table 4 and Fig. 2). The complete nucleomorph SSU rDNA sequences including nonalignable regions showed a mean A+T content of 58%, whereas the mean A+T content for the complete sequences of the combined nuclear ITS2 and partial LSU rDNA was only 47%. The nucleomorph SSU rDNA data set also failed the test for homogeneity of base frequencies among taxa when using the complete sequences. An exclusion of the nonalignable regions from the nucleomorph SSU rDNA (i.e., testing the data set used for the phylogenetic analyses) resulted in a decrease in the mean A+T content to 56%, and the data set passed the test (p = 1.0).

Table 4 A+T content in the highly variable region V4 of the nucleomorph SSU rDNA in selected strains
Full size table
Figure 2
figure 2

Secondary structures of the variable region V4 in the nucleomorph SSU rRNA from four strains. The region was nonalignable and, thus, was excluded from the phylogenetic analyses. For a better comparison of loop sizes, the numbered positions refer to the excised sequences only, not to their absolute position in the complete sequences. Inset: Schematic presentation of the secondary structure of the nucleomorph SSU rRNA of strain CCAP 979/61 (Hoef-Emden and Melkonian 2003). The variable region V4 is shaded in gray.

Full size image

The most extreme bias toward T was found in strains of clade LB with T contents greater than 30% (highest total amount of T in the colorless strain CCAC 0056, 32.5%; A+T, 61%).

In Fig. 2, four secondary structures inferred from a part of the nucleomorph SSU rRNA—the highly variable region V4—are depicted. The A+T contents found in the nucleomorph SSU rDNA sequences of representative strains are listed in Table 4. The highest A+T bias in this part of the nucleomorph SSU rDNA was found in C. lundii (strain CCAC 0107), with A+T contents of up to 87% (Fig. 2 and Table 4). In the rRNA secondary structure of region V4, the 5′-helical parts of both loops of C. lundii were dominated by U homopolymers (Fig. 2), whereas the 3′-helical part consisted predominantly of A homopolymers. In the C. paramaecium strain CCAC 0056, U also accumulated in the 5′-helical parts of both loops and A dominated the 3′-helical parts (Fig. 2). In the colorless NoPyr strain M1634, A and U were relatively evenly distributed between the 5′- and the 3′-parts of the helices, whereas in the C. curvata strain CCAP 979/62 a bias of U to the 5′-helical and of A to the 3′-helical part of the loops was also present (Fig. 2). The increases in U and A homopolymers in the strains CCAC 0107 and CCAC 0056 were accompanied by extended lengths of the 5′-proximal loops. The 5′-proximal loops of both strains were almost double the size of the loops of the strains M1634 and CCAP 979/62 (CCAC 0107, 115 nucleotides [nt], and CCAC 0056, 109 nt, compared to M1634, 55 nt, and CCAP 979/62, 56 nt; Fig. 2). The shorter 5′-distal loop, however, differed. It was longest in C. lundii strain CCAC 0107 (63 nt), two-thirds of the size in C. paramaecium strain CCAC 0056 (37 nt), and of medium length in M1634 (51 nt) and CCAP 979/62 (46 nt).

Discussion

Multiple Losses of Photosynthesis

In this study, phylogenetic analyses of nucleomorph SSU rDNA as well as nuclear ITS2 and nuclear partial LSU rDNA consistently resulted in trees showing three origins of colorless Cryptomonas strains in two different clades. This result remained stable irrespective of the method of analysis. In addition, constraint trees that were forced to include a monophyletic clade of colorless strains were always rejected as significantly worse in all testing scenarios of the approximately unbiased and weighted Shimodaira–Hasegawa tests.

In higher plants, multiple losses of photosynthesis have also occurred, e.g., in the parasitic Scrophulariaceae/Orobanchaceae lineage, which contains colorless holo- as well as photosynthetically active hemiparasitic members (dePamphilis et al. 1997). The colorless holoparasitic lineages were polyphyletic and originated from hemiparasitic ancestors (dePamphilis et al. 1997). In the green algae, the colorless taxa Polytoma, Prototheca, and Helicosporidium were found to be closely related to known photosynthetic genera (Chlamydomonas and Chlorella) belonging to the Chlorophyceae and Trebouxiophyceae respectively (Rumpf et al. 1996; Nedelcu 2001; Tartar et al. 2003). In euglenophytes, permanent loss of photosynthesis was reported in Euglena gracilis, a close relative of the colorless Euglena longa, possibly caused by spontaneous mutations (e.g., Pringsheim and Pringsheim 1952; Marin et al. 2003).

However, in contrast to the holoparasites of the Orobanchaceae/Scrophulariaceae or most colorless chlorophyte lineages (Rumpf et al. 1996; Nedelcu 2001), colorless Cryptomonas were formerly assigned to one genus only (“Chilomonas”) and were never suspected to be polyphyletic. This finding is, thus, new and unexpected for the cryptophytes.

Possible Causes for Accelerated Evolutionary Rates

Accelerated evolutionary rates have often been found to be associated with changes in the mode of nutrition and/or obligate asexual reproduction. Moran (1996) used eight different genes to compare substitution rates and A+T content of free-living vs. endosymbiotic bacteria of the genus Buchnera. In the majority of the tested endosymbiotic bacterial lineages, she found accelerated evolutionary rates and overall A+T contents were higher in the endosymbionts. Lynch (1997) performed tests on nucleotide substitution rates of tRNAs from nuclear and organelle genomes and found evidence for increased evolutionary rates in the organelle tRNAs. An accelerated substitution rate and, in organelles and endosymbionts, an A+T bias were also found in plants and algae that had switched completely to a heterotrophic mode of nutrition and that had lost photosynthesis (Nickrent and Starr 1994; Rumpf et al. 1996; dePamphilis et al. 1997; Nedelcu 2001; Tartar et al. 2003).

Several possible causes for an increase in evolutionary rates related to a change of life style were discussed in previous studies. These include relaxed selective constraints (dePamphilis and Palmer 1990), population subdivision, genetic bottlenecks in small populations (Charlesworth et al. 1993), and asexuality (Lynch and Blanchard 1998). The process of accumulating slightly deleterious mutations, referred to as “Muller’s ratchet,” was explained by a lack of recombination during obligate asexual reproduction, but also was reported to occur in small isolated populations (Muller 1964; Charlesworth et al. 1993; Charlesworth and Charlesworth 1997). Rocha and Danchin (2002) proposed as an explanation for A+T-biased endosymbiotic genomes a competition of host cell and endosymbiont for metabolic resources resulting in a decreased availability of the more costly nucleotides, guanine and cytosine, for the endosymbiont.

In this study, the colorless as well as the closely related photosynthetic Cryptomonas strains of clades NoPyr and LB showed accelerated evolutionary rates in their rDNA operons. However, the degrees of divergence differed between nuclear and nucleomorph sequences. In the nucleomorph SSU rDNA, the strains of clade LB had the longest branches, whereas in the nuclear ITS2 and the nuclear partial LSU rDNA, clades NoPyr (especially subclade C) and single taxa of clade LB (ITS2, the two photosynthetic strains C. gyropyrenoidosa and C. lundii; LSU, C. paramaecium) showed increased evolutionary rates. In addition, the nucleomorph SSU rDNA showed a tendency toward an accumulation of A and T in the variable regions.

In the plastid-bearing cryptophytes, the situation is more complex than in endosymbiotic bacteria, chlorophytes, or higher plants, since four genomes are present in a cell (Douglas et al. 2001). In SSU rDNA phylogenies, the nucleomorph sequences (representing the nucleus of the former red algal endosymbiont) displayed an acceleration in substitution rates in analyses that included all eukaryotic crown groups as well as in analyses which addressed the relationships to the red algae (e.g., Van de Peer et al. 2000; Hoef-Emden et al. 2002). However, with respect to the nuclear sequences, differences in branch lengths were less pronounced in the cryptophytes (Hoef-Emden et al. 2002). Also, the completely sequenced nucleomorph genome of the photoautotrophic cryptophyte, Guillardia theta, revealed A+T contents of 75% in coding regions and local increases of up to 90% A+T in noncoding and spacer regions (Maier et al. 2000). In molecular phylogenies of plastid-encoded genes, however, the branches of the cryptophytes showed only average branch lengths (Oliveira and Bhattacharya 2000; Yoon et al. 2002).

What Are Possible Explanations for the Complex Patterns of Substitution Rates in the Cryptophyceae?

The complex plastid of cryptophytes is like a matryoshka (Russian doll): A reduced cyanobacterium resides inside a reduced red alga that resides inside a cryptophyte host. Thus, increased substitution rates in the different genomes may be considered as a result of combinations of nested evolution and coevolution caused by the consecutive endosymbiotic events. A hypothesis of how the changes in selective forces during evolution to a colorless Cryptomonas may have affected the different genomes (excluding the mitochondrial genome) is schematically shown in Fig. 3.

Figure 3
figure 3

Hypothetical evolutionary pathway of the evolution to photosynthesis loss in Cryptomonas. The first two steps depicted are the steps common to the ancestor of all plastid-bearing cryptophytes: establishment of the secondary endosymbiosis and reduction of the red alga to a complex plastid. The last two steps address the evolution of a photosynthetic to a colorless Cryptomonas (see Discussion). Dark gray, altered selective forces; light gray, unchanged selective forces. Nu, nucleus of host cell; Nm, nucleus of red alga or nucleomorph; Pt, rhodoplast/rhodoplast genome

Full size image

The genomes of free-living cyanobacteria usually comprise several million base pairs and several thousand genes (e.g., Kaneko and Tabata 1997; Kaneko et al. 2001), but when the cyanobacterial ancestor(s) of the extant three primary plastid lineages adapted to an endosymbiotic life, the genomes were extremely reduced and the of A+T content increased (step one in Fig. 3). Evolutionary rates were only slightly increased in the basal branch of the plastid lineages in a plastid SSU rDNA phylogeny (Turner et al. 1999). However, compared to primary plastids, additional genomic size reduction and increase in A+T content apparently were low in most secondary plastids studied (Hallick et al. 1993; Kowallik et al. 1995; Douglas A and Penny 1999), and a plastid SSU rDNA phylogeny showed differences in evolutionary rates among the plastid sequences without congruence to primary or secondary endosymbiosis (Turner et al. 1999). Possibly, the DNA in complex plastids was less affected by a secondary endosymbiosis, because it had already been subjected to genome size reduction and altered selective forces during establishment of the primary endosymbiosis (step two in Fig. 3). Unlike the plastid genomes, the nuclei of the endosymbionts presumably experienced during secondary endosymbiosis for the first time the altered conditions mentioned above as potential causes for increased substitution rates. In almost all algal lineages with complex plastids, the endosymbiont nuclei disappeared completely. However, those of the cryptophytes were retained as a nucleomorph, albeit with a reduced genome size, increased evolutionary rates, and an A+T bias similar to the genomes of endosymbiotic bacteria and organelles (Maier et al. 2000; Douglas et al. 2001), (step two in Fig. 3).

From Phototrophy to Heterotrophy

Whereas a secondary endosymbiosis seemingly does not necessarily affect the plastid genome of a complex plastid with respect to evolutionary rates and A+T bias, this does not apply to changes from a phototrophic to a heterotrophic life style. A massive loss of photosynthesis-related genes was reported for several leukoplast-bearing algae/plants and the leukoplast genomes were approximately only half the size of the genomes of related photosynthetic plastids (e.g., Wolfe et al. 1992; Gockel and Hachtel 2000). In the apicoplast (a complex plastid) genome of the malaria-causing parasite Plasmodium falciparum, predominantly genes for maintenance of gene expression were retained: genes for elongation factor Tu, ribosomal RNAs, ribosomal proteins, tRNAs, and three RNA polymerases. However, the genome size was only half the size of the leukoplast genomes of Epifagus virginiana or Euglena longa (Wilson et al. 1996). In almost all leukoplast genomes, accelerated evolutionary rates and increased A+T contents were found compared to photosynthetic relatives (Wolfe et al. 1992; Wilson et al. 1996; dePamphilis et al. 1997; Nedelcu 2001; Vernon et al. 2001; Euglena longa: B. Marin, personal communication). The accelerated mutation rates found in the leukoplasts were explained by relaxed selective constraints due to a reduced load of protein synthesis (e.g., dePamphilis et al. 1997; Vernon et al. 2001).

In the colorless Cryptomonas strain SAG 977-2a, rbcL and psbA genes could be amplified by PCR (Yoon et al. 2002). However, since the plastid or the nucleomorph genomes of Cryptomonas or its leukoplast-bearing forms have not been sequenced yet, the degree of size reduction of the plastid genome in the colorless Cryptomonas lineages is unknown. According to the results of this study, the SSU rDNA of the nucleomorph reacted upon photosynthesis loss similarly to plastid genes, with an acceleration of evolutionary rate and an accumulation of A and T in addition to an already high A+T content. Possibly, nucleomorph and plastid genomes were equally affected by the loss of photosynthesis (46 photosynthesis-related genes were found in the plastid genome of the cryptophyte Guillardia theta [Douglas and Penny 1999]), and hence, they reacted as an evolutionary unit to changes in selective pressures (step four in Fig. 3).

Photosynthesis loss apparently is not always accompanied by an acceleration of evolutionary rate. dePamphilis et al. (1997) explained differences in evolutionary rates among holoparasitic flowering plants with lineage-specific relaxed constraints. However, differing evolutionary rates may also be indicative of different times of divergence. In this study, the three colorless Cryptomonas lineages showed differences in acceleration of evolutionary rates. The still unknown degrees of reduction in the leukoplast genomes as well as the evolutionary pathways that resulted in the loss of photosynthesis may, thus, be unique for each of the three colorless Cryptomonas lineages.

However, accelerated evolutionary rates were found not only in plastid, but also in nuclear genes, although usually without an increased A+T content (Nickrent and Starr 1994; Rumpf et al. 1996; Nedelcu 2001; Hoef-Emden et al. 2002; this study). Thus, it may be possible that the loss of photosynthesis also relaxes selective constraints on nuclear genomes. On the other hand, accelerated evolutionary rates were also observed in photosynthetic sisters of colorless taxa in nuclear as well as in plastid genes (dePamphilis et al. 1997; this study). Hence, dePamphilis et al. (1997) suggested that loss of photosynthesis was a consequence of preceding evolutionary changes.

Probably, one prerequisite for a successful change to a heterotrophic lifestyle is mixotrophy. Often photosynthetic taxa are also facultative heterotrophs (Jones 2000). In several Cryptomonas species bacterivory was found (e.g., Tranvik et al. 1989; Urabe et al. 2000). Thus, the required biochemical pathways and transporters necessary for obligate heterotrophy may be present in at least some of the photosynthetic Cryptomonas lineages.

The initial causes of accelerated evolutionary rates in the leukoplast-bearing taxa and their close photosynthetic relatives can only be speculated about. One possible explanation may be loss of sexual reproduction. Conclusive evidence for sexual reproduction in Cryptomonas is still missing. Wawrik (1969) observed a fusion of Cryptomonas cells in a field sample. However, in the cryptophyte genera Proteomonas (Hill and Wetherbee 1986) and Cryptomonas (Hoef-Emden and Melkonian 2003), evidence for dimorphism was found and discussed as possibly related to a yet unkown sexual life cycle. This hypothesis was supported by observation of different DNA contents in the two morphotypes (Hill and Wetherbee 1986). Whereas in several Cryptomonas clades both morphotypes—cryptomorph and campylomorph—were found (C. curvata, C. pyrenoidifera, and clade NoPyr), clade LB seemingly comprises only one morphotype (campylomorph [Hoef-Emden and Melkonian 2003]). If the hypothesized sexual reproduction was lost in this clade, the loss of recombination could account for the increased evolutionary rates in clade LB (i.e., a Muller’s ratchet effect; see above). These mutations could occur randomly also in nucleus-encoded genes for nucleomorph or plastid products and, thus, may even cause a loss of photosynthesis by importing defective gene products into the plastid, hence, impairing its function and relaxing selective constraints (steps three and four in Fig. 3). However, in clade NoPyr both morphotypes are present. Thus, the initial cause(s) triggering the acceleration of evolutionary rates may also be different.

Whatever the causes for the increased evolutionary rates are, in clade NoPyr as well as in clade LB, evolution had already accelerated in a photosynthetic ancestor to the colorless taxa, indicating that successive mutations in the host as well as organelle genomes preceded the loss of photosynthesis.

Supplementary Material

s00239-004-0089-5supp.pdf

Table S1: Complete list of strains included in the analyses with accession numbers of sequences, origin, and isolator. Table S2: Sequences of PCR and sequencing primers. Table S3: Combinations of primers for PCR and bidirectional sequencing.