Eighteen free-living litostomateans from the subclasses Rhynchostomatia and Haptoria were sampled in a variety of habitats around the world (Table 1). Species were identified using live observation, protargol impregnation, and SEM (Foissner, 1991). Seven to two hundred cells were picked with a micropipette, washed at least twice in sterile spring water to remove contaminants, and transferred into 180 μl ATL or EB buffer (Qiagen, Hildesheim, Germany). Samples were stored at +1 to +3 °C pending DNA extraction.

Genomic DNA was extracted according to the methods described in Vd’ačný et al. (2011a,b). The ITS1-5.8S rRNA-ITS2 region and the first two domains of the 28S rRNA gene of all species except for three (see below) were amplified by the polymerase chain reaction (PCR) using the ITS-F (5′-GTAGGTGAACCTGCGGAAGGATCATTA-3′) and LO-R (5′-GCTATCCTGAGRGAAACTTCG-3′) primers that were complementary to the conserved regions of the 3′ end of the 18S rRNA gene and the 3′ end of the D2 domain of the 28S rRNA gene (Miao et al., 2008; Pawlowski, 2000). PCR conditions and cycling parameters followed our previous protocol (Vd’ačný et al., 2011a,b). The PCR products were cloned into the vector plasmid pCR 2.1 using the TOPO TA Cloning kit (Invitrogen, Carlsbad, CA, USA). Plasmids were sequenced bi-directionally using M13 forward and reverse primers supplied with the kit at Beckman Coulter Genomics (Danvers, MA, USA).

The 18S rDNA, ITS1-5.8S rRNA-ITS2 region of Balantidion pellucidum, Cultellothrix lionotiformis, and Microdileptus microstoma were amplified using primers Euk A (5′-AACCTGGTTGATCCTGCCAGT-3′) and reverse primer 5′-TTGGTCCGTGTTTCAAGACG-3′ (Jerome and Lynn, 1996; Medlin et al., 1988). The ITS1-5.8S rRNA-ITS2 region for these taxa was directly sequenced with primer 1280 F (5′-TGCATGGCCGTTCTTAGTTGGTG-3′) and the reverse amplification primer (Wylezich et al., 2002) at Sequetech Corporation (Mountainview, CA, USA).

Obtained sequences were imported into Chromas ver. 2.33 (Technelysium Pty Ltd.) to check for data quality and trim the 5′ and 3′ ends. The consensus sequences, based on sequences from 2 to 12 clones (Table 1), were created in BioEdit ver. 7.0.5.2 (Hall, 1999) with an inclusion threshold frequency of 80% identity.

The boundaries of the ITS1 and 5′ end of the 5.8S rRNA gene were identified by comparison with the sequences available in GenBank via the Rfam database available on the web page http://www.sanger.ac.uk/Software/Rfam/ (Griffiths-Jones et al., 2005). The boundaries of ITS2 were determined using the ITS2 Annotation feature of the ITS2 database (http://its2.bioapps.biozentrum.uni-wuerzburg.de) (Koetschan et al., 2010), which recognizes the ITS2 proximal stem, i.e., a hybridized 5.8S-28S rRNA fragment forming a characteristic approximately 15 bp-long imperfect helix (Keller et al., 2009).

The consensus sequences of the litostomatean ITS1 and ITS2 regions were aligned by the ClustalW algorithm implemented in the program Clustal X ver. 2.0.12 (Larkin et al., 2007) with arbitrarily chosen parameters (8.0 for gap opening penalty and 6.0 for gap extension penalty). The consensus ITS1 and ITS2 secondary structures were predicted from the ITS1 and ITS2 alignment, respectively, using the RNAalifold WebServer (http://rna.tbi.univie.ac.at/cgi-bin/RNAalifold.cgi) with default options (Bernhart et al., 2008; Gruber et al., 2008). With the guidance of these consensus structures, the secondary structures of ITS sequences were predicted with Mfold (http://mfold.bioinfo.rpi.edu/cgi-bin/rna-forum1.cgi) by screening for thermodynamically optimal and suboptimal secondary structures using the default values (Zuker, 2003). Results for all studied litostomatean species were compared to reveal the folding pattern common to all of them. Subsequently, we have established conserved structural models for litostomateans in order to reveal evidence of homology useful for aiding alignments. The base frequency at each position and mutual information of base-paired regions were calculated using the program RNA Structure Logo (http://www.cbc.dtu.dk/~gorodkin/appl/slogo.html) (Gorodkin et al., 1997). For each structural domain, the number of base pairs, unpaired bases in bulges and interior loops were investigated and compared. If different nucleotides occurred at the same position of an ITS structural domain, they were encoded in motifs according to the NCBI webpage (www.ncbi.nlm.nih.gov/blast/fasta.shtml).

To determine the genealogical relationships among litostomateans, we analyzed eleven alignments (Table 2) comprising up to 23 in-group taxa (Table 3). Outgroup sequences were not included because it was not possible to obtain an unambiguous alignment with either armophorean and/or spirotrichean sequences. The litostomatean 18S rRNA sequences were aligned in the ARB-package (Ludwig et al., 2004), while Clustal X (Larkin et al., 2007) was used to align the ITS sequences. The resulting alignments were manually edited in the program BioEdit (Hall, 1999), according to the secondary structural features of the 18S rRNA molecule as predicted by the ARB-package, and the ITS molecules proposed in this study. Ambiguously aligned and hyper-variable regions were eliminated.

All alignments were analyzed with three phylogenetic methods: Bayesian inference (BI), maximum likelihood (ML), and maximum parsimony (MP). Bayesian inference analyses were performed with the program MrBayes (Ronquist and Huelsenbeck, 2003), using the nucleotide substitution models determined by the program jModelTest ver. 0.1.1 (Posada, 2008) under the Akaike Information Criterion for each alignment (Table 4). Four simultaneous MCMC chains were run for 5,000,000 generations with trees sampled every 1000 generations. The first 25% of sampled trees were considered as burn-in and discarded prior tree reconstruction. A 50% majority rule consensus of the remaining trees was used to calculate posterior probabilities (PP) of the branching pattern. The ML analyses were computed on the CIPRES Portal (http://www.phylo.org), using RAxML with settings as described in Stamatakis et al. (2008). The MP trees were constructed using PAUP^* ver. 4.0b8 with randomly added species (10 replications) and tree bisection–reconnection (TBR) branch-swapping algorithm in effect (Swofford, 2003). The reliability of the ML and MP trees was tested by the bootstrap approach, using 1000 pseudoreplicates and a heuristic search algorithm. Support values from all tree-building methods were annotated onto the tree topology having the best log-likelihood score.

Constrained analyses were carried out on the 18S + ITS1 + 5.8S + ITS2 + 5′ end 28S dataset in which the family Dileptidae was forced to be monophyletic, and on the 18S + ITS1 + 5.8S + ITS2 dataset where a sister relationship was forced for the subclasses Haptoria and Trichostomatia. Constrained trees were built in PAUP^*, using ML criterion and heuristic search with TBR swapping algorithm and 10 random sequence addition replicates. The site-wise likelihoods for the best unconstrained ML trees and all constrained trees were calculated in PAUP^* under the substitution models with parameters as suggested by jModeltest (see above and Table 4). The reliability of the constrained trees was analyzed through the approximately unbiased test (AU) and the Shimodaira–Hasegawa test (SH) implemented in the CONSEL software package (Shimodaira and Hasegawa, 2001).

The ITS1 sequences vary slightly in length from 105 nt in Protospathidium muscicola to 114 nt in Monomacrocaryon terrenum. In contrast, the GC content spans a wide range, from 17.1% in Apobryophyllum schmidingeri to 29.2% in Dileptus costaricanus (Table 5).

The litostomatean ITS1 molecules are less conserved than the ITS2 molecules, making prediction of their secondary structures less consistent. Therefore, we follow the secondary structure models of the ITS1 proposed by Ponce-Gordo et al. (2011) for Balantidium coli and several other litostomateans. All studied litostomatean taxa share a similar basic pattern of ITS1 secondary structure, showing an open internal loop radiating four helices (Fig. 2).

Helix I is the most consistent structural feature of the ITS1, with high base pairing conservation and low variation in length (19–21 nt). In the rhynchostomatians, helix I displays a highly conserved motif at the basal (5′-UUAA vs. UUAA-3′) and terminal (5′-CAA vs. UUG-3′) portions, which are separated by a 5′-AAC-3′ bulge situated on the 3′ side of helix I (Figs. 2 and 3). Nucleotide composition of the haptorian helix I is more variable than that of the rhynchostomatians (for comparison, see Fig. 3). Helix II is comparatively short in the rhynchostomatians (10–15 nt), presenting a very conserved motif 5′-CUU vs. GAA-3′ in the basal portion. On the other hand, helix II of the haptorians is almost two times longer (22–27 nt) and shows a relatively high variation in base pairing. Helix III is the most variable constituent of the litostomatean ITS1, with respect to its length (11–26 nt) and nucleotide composition. Helix IV is also highly variable in length (15–33 nt) and possesses a bulge in several haptorians (Arcuospathidium sp. and Semispathidium sp.) and the majority of the rhynchostomatians (Apodileptus visscheri rhabdoplites, Monomacrocaryon terrenum, Rimaleptus mucronatus, and Trachelius ovum). Helix IV has the following motif 5′-UAACU vs. GGUUG-3′ in the rhynchostomatians, while 5′-AACC(U)A(U) vs. UGGUU-3′ in the haptorians. Frequencies of bases at each position and mutual information in base-pair regions in helices I and IV are shown in the RNA structure logos (Fig. 3). The estimated thermodynamic energy of putative secondary structures of the litostomatean ITS1 is on average −10.2 kcal/mol at 37 °C.

There are zero to four GU appositions in putative secondary structure of the ITS1 transcripts (Table 5). Although these pairings are less stable than the Watson–Crick complementarities, they still retain the RNA helical structure and hence support our prediction of secondary structure.

The ITS2 sequences have an average length of 107 nt with a range of 105 nt (Arcuospathidium sp.) to 109 nt (Cultellothrix lionotiformis). In contrast, the GC content varies greatly from 22.4% in Protospathidium muscicola to 37.7% in Enchelys gasterosteus (Table 6).

The consensus putative secondary structure model of the ITS2 is shown in Fig. 2. Its main features include an internal loop bearing three helices of unequal length. Helix I is the shortest and most variable helix of the ITS2 molecule with a range of 6–19 nt. Helix II displays a highly conserved motif 5′-GURAGAGA vs. YCUCUYAU-3′ in its basal portion (Fig. 3) and varies in length from 19 to 21 nt. However, Balantidion pellucidum, as a sole exception, has an extremely elongated (31 nt) helix II which also displays a bulge near its basal portion. The terminal loop of helix II invariably comprises 5 nt (5′-AYHWU-3′) in all haptorians and 4 nt (5′-WAAV-3′) in all rhynchostomatians (except for Monomacrocaryon terrenum, containing 6 nt: 5′-AGACUU-3′). Helix III is longest and has one to four bulges containing on average five unpaired bases. The length of helix III spans a wide range of 36–62 nt, averaging 47 nt. In 11 out of 18 species studied, helix III presents a highly conserved motif 5′-AGCAGUCACA vs. UGUGAGCU-3′ in its distal portion (Fig. 3). The terminal loop of helix III invariably includes 4 nt in all rhynchostomatians and in most haptorians (5′-YHHU-3′), while 6 nt in the rest of the haptorians (5′-UUHRUU-3′). The frequencies of bases at each position and mutual information in base-pair regions in helices II and III are shown in the RNA structure logos (Fig. 3). The estimated thermodynamic energy of putative secondary structures of the litostomatean ITS2 molecules spans a range of −29.10 kcal/mol to −19.47 kcal/mol at 37 °C, with an average of −23.30 kcal/mol at 37 °C.

Compensatory base changes have been revealed within the proposed putative secondary structure of the litostomatean ITS2 transcripts. Specifically, there are at most four GU appositions. One occurs invariably in the highly conserved region of helix II and zero to three are present in helix III, corroborating the proposed model.

To determine the genealogical relationships among the litostomateans and to reconstruct their evolutionary history, we carried out 33 phylogenetic analyses based on 11 alignments using three different algorithms (Table 2). First, we performed 21 single-locus analyses on litostomatean sequences of molecules that are either involved in rRNA biogenesis (ITS1 and ITS2) or code some components for the small (18S rRNA) and large (5.8S rRNA) ribosomal subunit. Given the overall congruence between the single-locus tree topologies, though sometimes with weak support for some nodes, we compiled the following concatenated alignments: ITS1 + 5.8S + ITS2; 18S + ITS1 + 5.8S + ITS2; and 18S + ITS1 + 5.8S + ITS2 + 5′ end 28S. This second multiple-locus set of analyses resulted, as expected, in similar tree topologies as those obtained with the single-locus approach. However, most nodes were generally better supported statistically in the multiple-locus phylogenies (Table 7).

Out of 33 phylogenetic analyses performed, 31 have consistently recovered a fundamental bifurcation of the class Litostomatea into two lineages (Table 7 and Figs 4–6). The first clade is designated as subclass Rhynchostomatia. This unites free-living litostomateans having a proboscis that carries a complex oral ciliature comprising a circumoral kinety, a perioral kinety and many preoral kineties (Fig. 1A–C). The second cluster includes the free-living subclass Haptoria nesting the endocommensal subclass Trichostomatia. Members of both subclasses lack a proboscis (i.e., they have an apical oral apparatus) and have a comparatively simple oral ciliature including only a circumoral kinety and/or oralized somatic kinetids (Fig. 1D, E, and H).

As concerns the internal relationships among the rhynchostomatians, ITS1 and ITS2 sequences considered separately were not able to unravel a well-supported branching pattern, very likely due to the mutational saturation. However, analyses of the 5.8S, 18S, and all concatenated datasets provided a well-resolved phylogeny of the rhynchostomatians. Specifically, Trachelius ovum represents a separate branch, the order Tracheliida, which is characterized by a curious lateral fossa, a dikinetidal circumoral kinety and an ordinary three-rowed dorsal brush. All other rhynchostomatians, i.e., the order Dileptida, are depicted as a sister group of T. ovum (Figs. 4–6). By contrast to the Tracheliida, the Dileptida lack a lateral fossa and are characterized by a hybrid circumoral kinety (i.e., composed of dikinetids in proboscis and monokinetids around oral opening) and a staggered two- or multi-rowed dorsal brush. The relationships at family level of the order Dileptida were not consistently recovered. The family Dimacrocaryonidae (represented here by Microdileptus microstoma, Monomacrocaryon terrenum and Rimaleptus mucronatus; however, M. microstoma was included only in the 5.8S, ITS2, and 18S datasets, as we were not able to obtain its full ITS1 sequence), which typically has one or two macronuclear nodules and a single micronucleus, was found as monophyletic in 21 out of 33 analyses (Table 7). The family Dileptidae (represented here by Apodileptus visscheri rhabdoplites, Dileptus costaricanus and Pseudomonilicaryon fraterculum), which is defined by having at least four macronuclear nodules and many micronuclei, was depicted as monophyletic only in the 5.8S, 18S, and 18S + ITS1 + 5.8S + ITS2 analyses (Table 7 and Fig. 5). In the other analyses, P. fraterculum did not group with D. costaricanus and A. visscheri rhabdoplites, but was placed basally within the order Dileptida, however, with low support (Figs. 4 and 6). Nevertheless, the monophyly of the family Dileptidae could not be rejected by the approximately unbiased (P-value = 0.480) and the Shimodaira–Hasegawa (P-value = 0.475) statistical topology tests carried out on the largest 18S + ITS1 + 5.8S + ITS2 + 5′ end 28S dataset.

The internal relationships within the Haptoria–Trichostomatia clade were not well resolved in any dataset, suggesting one or several radiation event(s) and/or undersampling of haptorian and trichostomatian genera. However, we have identified several common branching patterns (Table 7 and Figs. 4–6). (1) The orders Spathidiida and Haptorida always grouped together. (2) Two spathidiids, Apobryophyllum schimidingeri and Cultellothrix lionotiformis, consistently formed a strongly supported monophylum. These two species, in contrast to all other spathidiids included in the analyses, display a dorsal brush on the left (vs. dorsal) side of the body. (3) Two traditional haptorids, Balantidion pellucidum and Enchelys gasterosteus, consistently clustered together with strong support. Unlike all haptorians in the dataset, they share oralized somatic monokinetids. (4) The subclass Trichostomatia was monophyletic and usually branched within the order Spathidiida or was placed in the basal polytomy of the free-living haptorians. The approximately unbiased (P-value = 0.348) and Shimodaira–Hasegawa (P-value = 0.276) topology tests were not able to significantly reject the sister relationship of the Trichostomatia and Haptoria for the 18S + ITS1 + 5.8S + ITS2 dataset. However, this could be caused by the strong undersampling of trichostomatian sequences in the dataset.

The internal transcribed spacer 1 (ITS1) lies between the regions coding for 18S rRNA and 5.8S rRNA gene (Maroteaux et al., 1985). Some structural elements of the ITS1 could play a role in the cleavage of the pre-18S rRNA molecule at the ITS1 site (van Nues et al., 1994). This process is controlled to some extent by the interaction of the 18S rRNA precursor with some small nucleolar RNAs, indicating a minor function of the ITS1 in the rRNA processing. This, in turn, permits large variations both in the primary sequence and in the secondary structure of the ITS1 gene between various groups of organisms (Ferreira-Cerca, 2008; Ponce-Gordo et al., 2011).

In general, the putative secondary structure of the ITS1 transcripts includes an open loop with a different number of helices, ranging from three in peritrich ciliates (Sun et al., 2010) to seven in flatworms (von der Schulenburg et al., 1999), and up to nine in plants (Kan et al., 2007). In the free-living litostomatean ciliates, our results and those of Ponce-Gordo et al. (2011) indicate that there are four helices radiating from a 5′–3′ common loop, matching the ciliate ITS1 secondary structure models proposed by Hoshina (2010) and Sun et al. (2010). However, in the endocommensal litostomateans, there is an extra helix E1 which separates helix I from the other three helices (Ponce-Gordo et al., 2011). This different pattern was explained by the minor role of the ITS1 in the rRNA processing, enabling large variations also in the secondary structure (see above). On the other hand, helix I is the most conserved feature in the secondary structure of the litostomatean ITS1 transcript, with respect to its length and nucleotide composition (Ponce-Gordo et al., 2011; Table 5 and Fig. 3). This suggests that helix I could play an important role in the pre-18S rRNA molecule cleavage at the 5′ site of the ITS1.

The internal transcribed spacer 2 (ITS2) lies between the 5.8S and 28S rRNA region and is important in the maturation processes of both the 5.8S and 28S rRNA molecules (Ferreira-Cerca, 2008; Maroteaux et al., 1985). This causes the primary and secondary structure of the ITS2 transcript to be rather conserved in eukaryotes (Coleman, 2003; Schultz et al., 2005; Wolf et al., 2005). In ciliates, two models were recognized: a “ring model” with a common loop radiating three to four helices (Coleman, 2005; Miao et al., 2008; Ponce-Gordo et al., 2011) and a “hairpin model” in which the common loop is started and closed by helix I and bears only helices II and III; helix IV, if present, is located between helix I and the 28S rRNA region (Sun et al., 2010). Côté et al. (2002) proposed a dynamic conformational model for the role of ITS2 processing: initial formation of the ring structure may be required for essential, early events in processing complex assembly and followed by an induced transition to the hairpin strucuture that facilitates subsequent processing events. This dynamic model explains very well the existence of both the hairpin model predicted for our litostomatean ITS2 transcripts, and the ring model proposed for the litostomatean ITS sequences by Ponce-Gordo et al. (2011). Both litostostomatean ITS2 models match very well in the structure and motifs of helices II and III, but differ in helix I which is very likely the dynamic constituent of the ITS2 transcripts, enabling switching from the ring to the hairpin pattern.

The litostomatean ITS2 molecules differ from those of oligohymenophorean ciliates by lack of the pyrimidine–pyrimidine bulge in helix II (Coleman, 2005; Miao et al., 2008; Sun et al., 2010), as already recognized by Ponce-Gordo et al. (2011). Spirotrichean ciliates also lack the pyrimidine–pyrimidine bulge in helix II which is designated as helix A (Gao et al., 2010; Weisse et al., 2008; Yi et al., 2008). This supports the 18S rRNA gene megaclassification of intramacronucleate ciliates in that litostomateans, armophoreans and spirotricheans form a super-clade which is sister to ventrate ciliates including oligohymenophoreans (Cavalier-Smith, 2004; Vd’ačný et al., 2010). On the other hand, spirotrichean helix III, which is designated as helix B, consistently displays a multi-branch pattern not occurring in litostomateans (Gao et al., 2010; Ponce-Gordo et al., 2011; Weisse et al., 2008; Yi et al., 2008; present study). Thus, the secondary structures of the ITS2 molecules may help to unravel the deep evolutionary history of intramacronucleate ciliates, whose relationships are still poorly understood.

The ITS region is frequently utilized for phylogenetic analyses at the genus and species levels in ciliates (e.g., Ponce-Gordo et al., 2011; Stoeck et al., 2007; Weisse et al., 2008; Yi et al., 2008). However, application of the ITS region in unraveling phylogeny at higher taxonomic levels was previously limited by uncertainties in alignment due to excessive insertions, deletions and mutational saturations. Coleman (2003) argued that the secondary structure of the ITS transcripts provides the key to solve this problem. Indeed, the present and previous studies (Coleman, 2005; Miao et al., 2008; Sun et al., 2010) on various ciliate groups document that ITS molecules have an appropriate signal for untangling relationships not only at species level but also at higher taxonomic ranks, when their secondary structure information is utilized to aid alignment. Thus, once secondary structure has been established for the litostomatean ITS transcripts, their sequences can be aligned across broader taxonomic levels including subclasses and orders. With multiple loci of a few hundred nucleotides, one can analyze relationships among the litostomateans from the subspecies to the subclass rank.

It is important to note that insertions and deletions within of the ITS sequences seem to be also an important phylogenetic marker, resulting in various sequence lengths. Specifically, spirotricheans possess the longest ITS transcripts, while litostomateans display the shortest ones, having 35–100 nt fewer than other ciliates (Coleman, 2005; Hoshina, 2010; Miao et al., 2008; Ponce-Gordo et al., 2011; Stoeck et al., 2007; Sun et al., 2010; Weisse et al., 2008; Yi et al., 2008; present study). Similarly, the litostomatean 18S rRNA gene is shorter in comparison with other ciliates due to the deletions in helices 23–1, 23–8, 23–9, and the absence of the entire helix 23–5 (Leipe et al., 1994; Strüder-Kypke et al., 2006; Vd’ačný et al., 2011a,b; Wright and Lynn, 1997a,b; Wright et al., 1997). The comparatively short 18S rRNA, ITS1 and ITS2 sequences in all litostomatean ciliates indicate that one or several deletion events occurred in the rRNA gene region of their last common ancestor.

To test whether 18S rRNA gene provides a suitable proxy for reconstruction of the litostomatean evolutionary history, we used the ITS region sequences. Although these are localized on the same transcriptional unit as 18S and 28S rRNA genes, only 5.8S rRNA molecule is incorporated into the ribosome, while ITS1 and ITS2 are removed from the transcript and consequently degraded (e.g., Retèl and Planta, 1967). This very likely causes that ITS sequences are subject to reasonably mild functional constraints which, in turn, enables a comparatively frequent occurrence of insertions and deletions as well as a preponderance of nucleotide sites that would evolve essentially neutrally (Álvarez and Wendel, 2003). Thus, in relation to functionality, it can be assumed that ITS sequences are not the same “unit of selection” as rRNA genes. Consequently, ITS1 and ITS2 could be used as a rather independent test whether 18S rRNA molecules elucidate the litostomatean phylogeny properly. Because ITS1 and ITS2 phylogenies are congruent with 18S rRNA gene trees, we find 18S rRNA molecules as an effective tool for unraveling the litostomatean evolutionary history. This is also supported by the highest values of tree indices for the 18S rRNA gene dataset, i.e., consistency index (CI), consistency index excluding uninformative characters (CIex), retention index (RI), and rescaled consistency index (RC) show the best fit of the 18S rRNA data to the inferred phylogenetic trees (Table 2).

The monophyletic origin of the class Litostomatea was consistently supported by five strong morphological apomorphies and the 18S rRNA gene (for review, see Vd’ačný et al., 2011a). However, the morphology-based phylogenetic relationships among the litostomateans conflict with the 18S rRNA gene phylogenies, especially, in that (1) the Litostomatea are not subdivided into the free-living haptorians (including also rhynchostomatians) and the endocommensal trichostomatians (Foissner and Foissner, 1988; Jankowski, 2007; Lynn, 2008; Lynn and Small, 2002), but into the free-living rhynchostomatians and the free-living haptorians including the endocommensal trichostomatians (Vd’ačný et al., 2011a,b); (2) rhynchostomatians, the crown litostomateans with the most complex morphology and ontogenesis, do not represent highly derived spathidiids (Vd’ačný and Foissner, 2008, 2009; Xu and Foissner, 2005), but are classified at the base of the Litostomatea as a sister group of all other litostomateans (Vd’ačný et al., 2011a,b); (3) simple polar enchelyine haptorids without circumoral kinety have not been found as an ancestral group of the litostomateans, but rather are nested within the spathidiid cluster (Foissner, 1984; Foissner and Foissner, 1985; Vd’ačný et al., 2011a); (4) spathidiids and haptorids have been not recovered as separate monophyla, but are mixed together usually without any clear morphological connection (Foissner and Foissner, 1988; Jankowski, 2007; Lynn, 2008; Vd’ačný et al., 2011a); and (5) the endocommensal trichostomatians are shown originating from rapacious terrestrial spathidiids, such as Epispathidium papilliferum or Protospathidium muscicola (Foissner and Foissner, 1985; Gao et al., 2008; Pan et al., 2010; Strüder-Kypke et al., 2006, 2007; Vd’ačný et al., 2010, 2011a,b). In our previous studies, we were able to overcome most of these conflicting issues by suggesting body polarization and simplification of the oral apparatus as the main evolutionary trends in the Litostomatea (for details, see Vd’ačný et al., 2010, 2011a,b and below). Indeed, our proposed reconstruction of the deep litostomatean evolutionary history is in a good agreement with single-locus trees inferred from 5.8S rRNA, ITS1, and ITS2 as well as with multiple-locus trees inferred from the whole ITS region and its concatenation with the 18S rRNA gene and the first two domains of the 28S rRNA gene.

Body polarization, i.e., apicalization of the oral opening, has been revealed as one of the main evolutionary trends in the litostomateans (see above). Most intramacronucleate ciliates including armophoreans, the supposed sister group of the litostomateans, have a ventral oral opening and a complex oral ciliature (Cavalier-Smith, 2004; Jankowski, 2007; Lynn, 2008; Vd’ačný et al., 2010). Within the Litostomatea, only rhynchostomatians display this plesiomorphic body organization (Vd’ačný et al., 2010, 2011a,b). Therefore, Vd’ačný et al. (2011a) used the morphology of the rhynchostomatians and armophoreans to hypothesize that the last common ancestor of the Litostomatea had a keyhole-shaped oral bulge with the oral opening situated in the widened posterior part (Fig. 1A and B). This is the only site of the rhynchostomatian oral bulge that can open and ingest prey (Dragesco, 1962; Vd’ačný and Foissner, 2012; Visscher, 1923; Fig. 1C). In contrast, the oral opening of the stemline of the Haptoria and Trichostomatia was very likely located apically and in the centre of the oral bulge (Vd’ačný et al., 2011a; Fig. 1D and E). This pattern possibly evolved by reduction of the proboscis causing only the posterior portion of the oral bulge, which surrounds the ancestral oral opening, to remain and become located in the anterior body pole (Fig. 7). Even when the oral bulge extends far posteriorly (e.g., in bryophyllids and pleurostomatids), only its apical portion opens during feeding like in simple polar haptorians (Dragesco, 1962; Foissner and Lei, 2004; Foissner and Xu, 2007; Foissner et al., 1995, 1999, 2002; Guhl and Hausmann, 2008; Maupas, 1883; Woodruff and Spencer, 1921, 1922; Fig. 1G, I, and J). As concerns the trichostomatians, the polar position of the oral opening was maintained in the archistomatids, such as Wolskana, Didesmis, and Alloiozona, while in more derived vestibuliferids and isotrichids the opening sunk into an anterior vestibulum or was displaced posteriorly (Grain, 1966; Lynn, 2008; Lynn and Small, 2002; Vd’ačný et al., 2011a). To summarize, the oral opening is located far subapically in the rhynchostomatians, while apically in all haptorians and basal trichostomatians. We find this differing location of the oral opening as the key morphological trait, explaining the deep split of the Litostomatea into Rhynchostomatia and Haptoria + Trichostomatia.