Parallel evolution of histophagy in ciliates of the genus Tetrahymena

Background Species of Tetrahymena were grouped into three complexes based on morphological and life history traits: the pyriformis complex of microstomatous forms; the patula complex of microstome-macrostome transformers; and the rostrata complex of facultative and obligate histophages. We tested whether these three complexes are paraphyletic using the complete sequence of the small subunit rDNA (SSrDNA). Results In addition to the 16 species of Tetrahymena whose SSrDNA sequences are known, we sequenced the complete SSrDNA from the following histophagous Tetrahymena species; Tetrahymena bergeri, Tetrahymena mobilis, Tetrahymena rostrata, and Tetrahymena setosa as well as the macrostome species Tetrahymena vorax. We also included a ciliate tentatively identified as Lambornella sp., a parasite of the mosquito Aedes sp. We confirmed earlier results using SSrDNA, which showed two distinct clusters of Tetrahymena species: the australis group and borealis group. The genetic distances among Tetrahymena are in general very small. However, all nodes were supported by high bootstrap values. With the exception of T. bergeri and T. corlissi, which are both histophagous and group as sister species, all other histophagous Tetrahymena species are most closely related to a bacterivorous species. Furthermore, Lambornella sp. and T. empidokyrea, both mosquito parasites, are sister species, although there is a considerable genetic distance between them. Conclusions There has been parallel evolution of histophagy in the genus Tetrahymena and the three classical species complexes are paraphyletic. As the genus Lambornella arises within the Tetrahymena clade, it is not likely a defensible one.


Background
All species of the genus Tetrahymena are morphologically very similar. As such, ecological, morphological, biochemical, and molecular features have been used over the years in attempts to classify them. The earliest classi-fications were based on morphological and ecological data. Czapik [1] regarded the presence or absence of a caudal cilium as an important character. Later, Corliss [2] distinguished three morphological species complexes: the pyriformis complex with smaller, bacterivorous species and fewer somatic kinetics; the rostrata complex with larger parasitic or histophagous species, more somatic kinetics, and the ability to form resting cysts; and the patula complex with species that undergo microstome-macrostome transformation. Within the complexes, particularly the pyriformis complex, species are distinguishable by mating capacity and/or isozyme patterns [3][4][5]. Finally, Corliss [6] suggested another approach based on the degree of parasitism. Since, the Tetrahymena species are free-living, as well as facultative and obligate parasites, Corliss [6] suggested an evolutionary lineage from free-living species, considering T. pyriformis to be the basal species, to facultative parasites, and then to obligate parasites.
More recently, gene sequences of ribosomal DNA (rD-NA) and histones have been used to determine relationships among Tetrahymena species. Phylogenies based on these sequences revealed that there is little divergence between the Tetrahymena species [7][8][9][10][11][12]. The 5.8S rDNA sequence is too short and consists of relatively conserved regions, which make it difficult to resolve the phylogenetic relationships among the species of a complex [13,14]. Partial regions of the large subunit ribosomal DNA (LSrDNA) that have been sequenced are identical for some species [9,10]. The small subunit ribosomal DNA (SSrDNA) is longer and there are sufficient differences between the sequences to infer a stable topology for most Tetrahymena species, with the exception of several species of the australis group that share identical or almost identical sequences [12]. Despite the high degree of relatedness, the tree topologies inferred from these analyses are consistent and well supported, separating the species into two branches -the australis group and the borealis group. According to the data inferred from 5S, 5.8S, and 23S rRNA, the species within the genus Tetrahymena were clustered into six ribosets (i.e., sets of species with similar sequences in the regions studied [11]) and later molecular analyses by Nanney et al. [10] generally confirmed these groupings. Riboset C corresponds to the australis group while ribosets A1, A2, and B include members of the borealis group.
In their LSrDNA analysis of several Tetrahymena species, Nanney et al. [10] demonstrated that histophagous and macrostome species grouped within clades of bacterivorous species, and they concluded therefore that macrostomy and histophagy arose by parallel evolution. By sequencing and analyzing the SSrDNA of more histophagous species, we further tested whether histophagy evolved several times independently within the genus Tetrahymena.
The SSrDNA sequences of all investigated Tetrahymena species differ only in 69 positions, which are located in the variable regions V1-V9 of the SSrRNA molecule Additional file 1 Fig. 1). Over half the variable positions are found in variable regions V2 and V4 (Fig. 1). The sequence of the histophagous T. mobilis is identical to those of the two microstome species, T. tropicalis and T.

Figure 1
Secondary structure model of the small subunit ribosomal RNA molecule of Tetrahymena bergeri. The model was constructed with the RNA Viz program [42] and shows 65 variable sites (in red). The additional 4 variable sites are missing in the Tetrahymena bergeri sequence (cf. Additional file 1:Table1.xls).
furgasoni, while the histophagous T. setosa shares an identical SSrDNA sequence with the bacterivorous T. pyriformis Additional file 2. Tetrahymena rostrata shows only one mismatch in its sequence to T. canadensis and T. borealis. The SSrRNA sequences of the latter two species are identical as are those of T. hyperangularis and T. pigmentosa Additional file 1 Additional file 2.
Tetrahymena bergeri, which was described by Roque et al. [15] and which has been regarded as doubtful species, has a unique SSrDNA sequence that differs in 9 nucleotide positions from its sister species T. corlissi.  3). The evolutionary distances within the australis group and borealis group are very small Additional file 2. Between the species of these two main groups, however, the distances are larger. Although the two main groups and some of the branches therein were very well supported by bootstrap values (Fig. 2) and ML support values (Fig.  3), other relationships within the genus Tetrahymena remain unresolved. This is demonstrated by the MP analysis, which computed 21 equally parsimonious trees that all supported the two major groups, but differed in the placement of the species within the clusters. The consensus tree of the MP analysis could only resolve three clusters: the australis group, the borealis group, and the T. empidokyrea/Lambornella sp. pair. All other branches were collapsed (tree not shown). Both distance analyses (LS and NJ) computed stable and comparable trees with high bootstrap support for the australis group and A distance tree for tetrahymenid ciliates inferred from small subunit ribosomal DNA sequences. The tree was derived from evolutionary distances produced by the Kimura-2parameter correction model [35]. The numbers at the nodes represent the bootstrap percentages of 1,000 for the least squares method (LS [36]) followed by the bootstrap values for the neighbor joining method (NJ) of Saitou and Nei [37]. Evolutionary distance is represented by the branch length separating the species. The scale bar corresponds to 5 substitutions per 100 nucleotide positions.

Figure 3
A maximum likelihood tree inferred from small subunit ribosomal DNA gene sequences of tetrahymenid ciliates. The tree was constructed using quartet puzzling. The numbers are support values for the internal branches while the branch lengths reflect maximum likelihood estimates of genetic distance [40].
the borealis group and sufficient bootstrap support to place T. empidokyrea and Lambornella sp. within the australis group (Fig. 2). The ML analysis also showed three clusters, with T. empidokyrea and Lambornella branching basal to all other Tetrahymena species (Fig.  3).
In all analyses, T. bergeri and T. corlissi branched basal within the borealis group (Fig. 2, 3). The other relationships within the borealis group, however, have to be regarded as unresolved. The bootstrap support for a closer relationship of the T. pyriformis and T. rostrata Tetrahymena bergeri is confirmed as a valid species, closely related to T. corlissi but only distantly related to T. rostrata, which it resembles morphologically ( Fig. 2, 3). The three newly-sequenced species, T. rostrata, T. setosa, and T. mobilis, each group together with a bacterivorous species, relationships supported by high bootstrap values (Fig. 2, 3).
The two Tetrahymena species isolated from mosquitoes, Lambornella sp. and T. empidokyrea, grouped together within the Tetrahymena clade. The evolutionary distance that separates Lambornella sp. and T. empidokyrea (i.e., d = 0.0042) is within the range of those separating other valid species Additional file 2.
If the different modes of nutrition are traced on the phylogenetic tree, it is evident that the two macrostome species T. vorax and T. patula are interspersed among the microstome species and that most of the histophagous species have a microstome species as sister group (Fig. 4, 5). To address the question which nutritional strategy can be regarded as ancestral, we performed character tracing with two different assumptions. In the first tree (Fig. 4), bacterivory is considered to be ancestral, and it shows under these circumstances that histophagy would have evolved five times independently within the genus Tetrahymena and that macrostomy would have evolved twice from bacterivory. If histophagy is considered to be the ancestral nutritional strategy (Fig. 5), bacterivory would have evolved five times within the genus and macrostomy would have evolved once from a bacterivorous and once from a histophagous ancestor. In both models five parallel steps are necessary to construct the topology.

Figure 4
Character state distribution of nutritional life history strategy of Tetrahymena species. The tree topology was derived by maximum parsimony analysis [38] and characters were traced using MacClade [41]. Bacterivory was assumed to be ancestral.

Figure 5
Character state distribution of nutritional life history strategy of Tetrahymena species. The tree topology was derived by maximum parsimony analysis [38] and characters were traced using MacClade [41]. Histophagy was assumed to be ancestral.

Discussion
Several phylogenies of the genus Tetrahymena have been constructed, based on various molecules like 5S and 5.8S rRNA [13,14], SSrRNA [12], LSrRNA [9][10][11], and histone H3II/H4II [7]. They are consistent in their topologies and separate two main clusters within the genus Tetrahymena: the australis and the borealis group. The first group is homogenous and consists of species of the riboset C as defined by Preparata et al. [11] (= T. australis group). The second group is more heterogeneous and comprises both ribosets A (A1: T. thermophila group and A2: T. tropicalis/borealis group) as well as riboset B (T. pyriformis group). In the analyses of Preparata et al. [11], the two macrostome species T. paravorax and T. caudata did not group with any of the ribosets but branched basal to all other Tetrahymena species. As in the analyses of Nanney et al. [9,10], our trees depict the australis group as a separate branch, coinciding with riboset C. The parasitic species T. empidokyrea and Lambornella sp. grouped basal within this clade. The topologies within the borealis group are in agreement with the SSrRNA tree of Sogin et al. [12]. However, the branching pattern is not highly supported by the bootstrap values, since the genetic distances for the SSrRNA within those main clusters are small, even for Tetrahymena species. In comparison with the results of Brunk et al. [7] who sequenced the histone H3II/H4II regions, our tree topologies show only minor differences. In their analysis, T. borealis and T. rostrata were closely related to T. tropicalis whereas in our analyses, T. borealis and T. rostrata were always grouped with the T. pyriformis group (=riboset B).
The three newly sequenced species T. mobilis, T. setosa, and T. rostrata each grouped with a bacterivorous species as the closest relative. Moreover, the sequences of T. mobilis and T. setosa are identical with their sister species and T. rostrata shows only one mismatch to T. borealis, a pattern that appears in other sequenced molecules as well. Nanney et al. [10] found the partial LSrRNA sequences of T. pyriformis identical to that of T. setosa while the partial LSrRNA sequence of T. canadensis (identical to T. borealis in their SSrDNA) was identical to that of T. rostrata. Sogin et al. [12] identified identical SSrRNA sequences for several Tetrahymena species. Our results increase the number of Tetrahymena species that share identical SSrDNA sequences. However, those species that are identical in their rDNA sequence can be distinguished morphologically or isozymically. Since Tetrahymena species are polymorphic for many isozymic traits, some of the species were defined on the basis of their specific isozymic characteristics [3,5]. However, the data derived from isozyme studies cannot be reliably used for the construction of phylogenetic trees within the genus Tetrahymena. If the rDNA trees are compared to the tree inferred from isozymic characters, a general accordance is achieved, but the species of ribosets A1 and A2 are scattered throughout the isozymic dendrogram [9].
Tetrahymena bergeri is confirmed as a valid species with a unique SSrDNA sequence and life cycle. Tetrahymena bergeri is closely related to T. corlissi, but several morphological and biological characteristics distinguish them from each other [15,16]. The main differences are the rostrum of T. bergeri, the infraciliature on the apical part of the cell, the oral infraciliature, the location of the pores of the contractile vacuole, and the resting cyst, which has not been observed for T. bergeri. Tetrahymena rostrata resembles T. bergeri morphologically [15], but based on life cycle features and such morphological characters as the polar basal body complex and minor differences in the buccal structures, Lynn [16] recognized them as two different species. Our results reveal that there is a large genetic distance between them, supported by high bootstrap values of the branching pattern. Thus, the rostrata complex as defined by Corliss [2] is shown to be a paraphyletic assemblage of species with a convergent life cycle but not a close genetic relationship.
The Lambornella species, presumably derived from the tree-hole Aedes mosquito, grouped with T. empidokyrea within the clade of Tetrahymena species. There is enough evolutionary distance between Lambornella sp. and T. empidokyrea to separate them as two different species. Another species of Lambornella, L. clarki, grouped within the genus Tetrahymena in an analysis of the D2 domain of the LSrDNA, much closer to other Tetrahymena species than T. paravorax [10]. In fact, Nanney et al. [10] showed that Lambornella clarki had a close relationship to T. corlissi. However, the Lambornella species in our analysis grouped with a different major cluster from T. corlissi, and this had high bootstrap support. Since we were unable to culture and stain our Lambornella species, it might have been a contaminant Tetrahymena from the tree-hole habitat. In our analysis T. corlissi is the sister species to T. bergeri and these two species grouped basal to the ribosets A and B (i.e., the thermophila-borealis/tropicalis and pyriformis groups). Since Lambornella sp. and T. empidokyrea also branched basally within the riboset C (i.e., T. australis group), this might explain the affinity of L. clarki and T. corlissi observed in the analysis of Nanney et al. [10], which included a different set of Tetrahymena species. Additionally, the method used by Nanney et al. [10] is most reliable for closely related species (i.e., within a species group), but shows limitations for more distantly related taxa. Taxonomically, another mosquito-parasitizing hymenostome species, Lambornella stegomyiae Keilin, 1921, had been assigned to the genus Tetrahyme-na as T. stegomyiae [17]. Corliss & Coats [18] transferred it back to the genus Lambornella when they described a second species, L. clarki. The main generic character separating Lambornella from Tetrahymena is the cuticular cyst of Lambornella from which it invades the haemocoel of its larval host. Our data show that the genetic distances between Lambornella sp. and the Tetrahymena species are in most cases smaller than the ones between the mosquito parasite T. empidokyrea and other species of Tetrahymena (cf. Fig. 2). Further analyses must be performed to test these placements of Lambornella within the genus Tetrahymena. If they prove to be correct, the genus Lambornella has to be regarded as invalid.
Our results support the claim that histophagy has evolved within the genus Tetrahymena several times independently. If the mode of food uptake is traced on the phylogenetic distance tree, it is evident that the two macrostome species are interspersed among the microstome species. This confirms the LSrRNA trees of Nanney et al. [9,10] in which the macrostome species are interspersed among the bacteria-feeding species while the other two macrostome species -T. paravorax and T. caudatagrouped basal to all Tetrahymena species. Most of the histophagous species we studied have a microstome species as sister taxon. Therefore, the three complexes -pyriformis, patula, and rostrata -must be regarded paraphyletic.
Based on our genetic distance tree we traced the character of food uptake under two different assumptions: the first assumption was that bacterivory was ancestral; and the second assumption was that histophagy was ancestral. Under both assumptions, there would have been five parallel steps necessary to construct the topology. The two macrostome species that we studied both grouped among the more derived Tetrahymena species; therefore, we did not assume macrostomy to be ancestral. Macrostome Tetrahymena species can be bacterivorous and morphologically similar to other bacterivorous Tetrahymena under certain conditions, but they are able to rearrange and enlarge their buccal ciliature and subsequently live as carnivores (see [19]).
Histophagous species have a rather complex life cycle, often with cyst formation and some morphological transformation. The histophagous species are either mostly free-living (i.e., T. bergeri, T. mobilis, T. setosa), facultatively parasitic (i.e., T. corlissi in invertebrates and lower vertebrates, T. rostrata in invertebrates), or apparently obligate parasites of mosquitoes (i.e., T. empidokyrea, Lambornella sp.). Hill [20] made the observation that Tetrahymena species must be highly derived based on their loss of biosynthetic abilities: they require 10 amino acids, 6 vitamins, guanine, and uracil; they have no urea cycle enzymes; and they probably make neither sterols, glutathione, nor carbamylphosphate. Could the genus have evolved from a Tetrahymena-like ancestor that was histophagous and that reverted to bacterivory or did histophagy emerge numerous times within the genus whose ancestor was bacterivorous? Since character state distributions for these two scenarios require an equal number of steps, we can make no certain conclusion at this time. However, we prefer the first model, which suggests the parallel evolution of histophagy from a bacterivorous ancestor. As different potential invertebrate and vertebrate host species evolved, it is possible that parallel evolution of histophagy within the genus Tetrahymena occurred to exploit these new habitats. An intriguing nutritional correlation is recorded by Hill [20]. He noted that, compared to bacterivorous pyriformis species, sterol is required for growth of T. corlissi, T. setifera (=T. setosa), and T. paravorax while phospholipid additions aid the growth of these fastidious species, T. corlissi, T. limacis, T. patula, and T. vorax. Could it be that sterol and phospholipid dependence evolved as these species exploited histophagy and macrostomy as nutritional life history strategies?
Our study demonstrates that the tetrahymenine ciliates have diversified as genetically isolated gene pools that are adapted to a variety of distinctive ecological niches. This rapid diversification may have occurred within the last 158 million years, since the late Jurassic [21]. Based on the genetical diversity within the genus Tetrahymena, the hypothesis that protist diversity is limited [22,23] must be questioned. It is obvious that morphological similarity does not reflect genetical identity nor does it necessarily reflect the ecological niche: morphologically similar Tetrahymena species may be genetically very different and may be either bacterivorous or histophagous. Further research on the genetic diversity within and between species of ciliates is needed to determine how widespread is this disconnect between morphology and genetics.

Conclusions
Within the genus Tetrahymena, two main clusters can be separated by molecular phylogenetic analyses: the australis and the borealis group. Generally, genetic distances for the SSrDNA among the species within those two clusters are very small. Our results distinguish Tetrahymena bergeri from any other Tetrahymena species. The other three newly sequenced, histophagous species T. mobilis, T. setosa, and T. rostrata each group with a bacterivorous species as the closest relative and show identical or almost identical sequences to their sister species. Thus, the rostrata complex of histophagous Tetrahymena is shown to be a paraphyletic assemblage of species with a convergent life cycle but not a close genetic relationship. This supports the model of parallel evolution of histophagy from a bacterivorous ancestor within the genus Tetrahymena, triggered by the evolution of different potential invertebrate and vertebrate host species.

Source of the species strains and culturing
The histophages, Tetrahymena rostrata (strain ID-3, ATCC #30770) and Tetrahymena setosa (strain HZ-1, ATCC #30782), and the macrostome Tetrahymena vorax (strain V2S, ATCC #30421) were obtained from ATCC (American Type Culture Collection, Manassas, VA, USA). The histophage Tetrahymena bergeri was obtained by D. Lynn from the culture collection at the Université de Clermont-Ferrand in 1976, and has been maintained in our laboratory since then. A culture of the histophage Tetrahymena mobilis was a gift from W. Foissner, Salzburg, Austria. The species was described as Saprophilus mobilis (Kahl, 1926). In an reinvestigation, Foissner W & Schiftner U (in prep.) found that it belongs to the genus Tetrahymena (W. Foissner, pers. comm.). All species except T. mobilis were cultured in proteose peptone-yeast extract medium (1.25 g/l dextrose anhydrous, 5 g/l proteose peptone, 5 g/l yeast extract) with a biweekly transfer. Tetrahymena mobilis was cultured in spring water with fragmented mealworms (Tenebrio molitor) as food source. The species, tentatively identified as a Lambornella species, was isolated from a sample derived from a tree-hole Aedes mosquito and provided to us by the laboratory of J. 0. Washburn and J. R. Anderson (University of California, Berkeley).

DNA extraction and sequencing
Lambornella sp., T. bergeri, T. mobilis, T. rostrata, and T. setosa were harvested by centrifugation, and washed in TE buffer (10 mM Tris base, 1 mM EDTA, pH 8.0). DNA extraction followed the standard protocol of Sambrook, Fritsch and Maniatis [24]. The DNA extraction of T. vorax followed the protocol of Walsh, Metzger and Higuchi [25]. One ml of the culture was centrifuged and the supernatant was discarded. Then, 200 µl of 5% Chelex ® 100 (Sigma, Oakville, ON, Canada) were added to the pellet. The mixture was vortexed, and incubated for 30 min in a waterbath at 56°C. Then, the mix was boiled for 8 min at 100°C and vortexed again. Finally, the sample was centrifuged at 16,000 g for 3 min in an Eppendorf Microcentrifuge 5415C, and 15 µl of this template were used for the subsequent PCR reaction. The PCR amplification of the SSrRNA genes was performed in a PTC-100™ thermal cycler (MJ Research Inc., Watertown, MA) or in a Perkin-Elmer GeneAmp 2400 thermal cycler (PE Applied Biosystems, Mississauga, ON, Canada). The SSrDNA of T. vorax was amplified using the in-ternal forward primer 82F (5'-GAAACTGCGAATGGCTC-3' [26]) and the Medlin B reverse primer (5'-TGATCCTTCTGCAGGTTCACCTAC-3' [27]). For all other species, the universal eukaryotic Medlin A forward primer (5'-AACCTGGTTGATCCT-GCCAGT-3' [27]) and the reverse primer 5'-TTGGTC-CGTGTTTCAAGACG-3' [8] were used in the PCR reactions. The SSrDNA of Lambornella sp. was subsequently cloned and sequenced following previously described methods [28]. PCR products were purified using the GeneClean kit (BIO/CAN, Mississauga, ON, Canada). They were sequenced in both directions using an ABI Prism 377 Automated DNA Sequencer (Applied Biosystems Inc., Foster City, CA), using dye terminator and Taq FS with three to four forward and four reverse internal universal SSrRNA primers [26].
The alignment of the sequences was performed with the Dedicated Comparative Sequence Editor (DCSE) program [33] and further refined by considering secondary structural features of the SSrRNA molecule. Genetic distances were calculated with the DNADIST program of the PHYLIP package, ver. 3.51c [34] based on the Kimura 2-parameter model [35]. The programs FITCH (Fitch-Margoliash least squares method [36]) and NEIGHBOR (neighbor-joining method [37]) of this package were used to construct distance trees. A maximum parsimony analysis was performed with PAUP * , ver. 4.0 [38]. Both parsimony and distance data were bootstrap resampled 640 times (FITCH) and 1,000 times (NEIGHBOR, PAUP) [39] respectively. PUZZLE, ver. 4.0.2 (maximum likelihood method [40]) was used to construct a maximum likelihood tree with support values for the internal branches and maximum likelihood branch lengths.
Out of the most parsimonious trees we chose the one that showed basically the same topology as the distance tree and imported it into MacClade ver. 3.0 [41] in order to

Additional material
Additional file 1