Host-Symbiont Cospeciation of Termite-Gut Cellulolytic Protists of the Genera Teranympha and Eucomonympha and their Treponema Endosymbionts

Cellulolytic flagellated protists inhabit the hindgut of termites. They are unique and essential to termites and related wood-feeding cockroaches, enabling host feeding on cellulosic matter. Protists of two genera in the family Teranymphidae (phylum Parabasalia), Eucomonympha and Teranympha, are phylogenetically closely related and harbor intracellular endosymbiotic bacteria from the genus Treponema. In order to obtain a clearer understanding of the evolutionary history of this triplex symbiotic relationship, the molecular phylogenies of the three symbiotic partners, the Teranymphidae protists, their Treponema endosymbionts, and their host termites, were inferred and compared. Strong congruence was observed in the tree topologies of all interacting partners, implying their cospeciating relationships. In contrast, the coevolutionary relationship between the Eucomonympha protists and their endosymbionts was more complex, and evidence of incongruence against cospeciating relationships suggested frequent host switches of the endosymbionts, possibly because multiple Eucomonympha species are present in the same gut community. Similarities in the 16S rRNA and gyrB gene sequences of the endosymbionts were higher among Teranympha spp. (>99.25% and >97.2%, respectively), whereas those between Teranympha and Eucomonympha were lower (<97.1% and <91.9%, respectively). In addition, the endosymbionts of Teranympha spp. formed a phylogenetic clade distinct from those of Eucomonympha spp. Therefore, the endosymbiont species of Teranympha spp., designated here as “Candidatus Treponema teratonymphae”, needs to be classified as a species distinct from the endosymbiont species of Eucomonympha spp.

Numerous insects have established intimate mutualistic relationships with microorganisms, which play key roles in host adaptation to various environments (16,40). This symbiotic relationship with microorganisms has a profound impact on the adaptation of termites to their xylophagous, wood-feeding lifestyle, and also contributes to the expansion of the termite niche in terrestrial ecosystems. Cellulolytic flagellated protists in the termite hindgut are unique to termites and related wood-feeding cockroaches, and are essential for host feeding on cellulosic matter (2,31).
The family Rhinotermitidae includes the global pest genera of termites, such as Coptotermes, Reticulitermes, and Heterotermes (20). In this termite family, cellulolytic protists from the genus Pseudotrichonympha (phylum Parabasalia) are widely distributed as gut symbionts (17). The cells of Pseudotrichonympha protists harbor endosymbiotic bacteria of a unique phylogenetic lineage within the order Bacteroidales (22). The members of this triplex symbiotic system appear to have cospeciated during their evolution (25), possibly because of the crucial roles played by symbionts in termite nutrition (13,33). Various examples of species-specific symbiotic relationships between protists and bacteria have been identified in the termite gut, and include ectosymbioses on the protist cell surface (29); many ectosymbionts are Bacteroidales and Spirochaetes members (14,21,23). As expected for endo-symbioses, most, if not all, have cospeciated (15,25,39,41), with some exceptions (10); ectosymbiotic relationships appear to be less strictly cospeciated than endosymbiosis (26).
The composition of the gut protist fauna is specific to the host species, and is suggested to reflect the host's phylogeny (18). However, in the family Rhinotermitidae, termites from the genus Reticulitermes typically lack Pseudotrichonympha protists. The gut protist composition in this genus is unique within the Rhinotermitidae, and is similar to that of the phylogenetically distant genus Hodotermopsis from the family Archotermopsidae, an early branching group of termites (1). These two termite genera are often distributed sympatrically, and the Japanese species Reticulitermes speratus or R. amamianus, and Hodotermopsis sjostedti frequently coexist in the same fallen log (18). A previous study reported that the protist fauna of a Hodotermopsis termite may have been laterally transferred to Reticulitermes termites (18). Hodotermopsis and Reticulitermes termites are both characterized by the rich species diversity of gut protists, and commonly harbor members of the family Teranymphidae (phylum Parabasalia) as large-size cellulolytic protists. Two genera of Teranymphidae, Eucomonympha and Teranympha (5), are phylogenetically closely related and inhabit H. sjostedti and Reticulitermes termites, respectively (27).
Eucomonympha and Teranympha protists both harbor bacterial endosymbionts (32). In H. sjostedti, a rod-shaped endosymbiont of Eucomonympha sp. was shown to be a spirochete species from the genus Treponema and was designated as "Candidatus Treponema intracellularis" (32). Recent biochemical and single-cell genome analyses revealed that this endosymbiotic species plays crucial roles in termite nutrition, e.g., reductive acetogenesis from H 2 plus CO 2 and nitrogen fixation (32). The former provides acetate as a major energy source for termites, while the latter provides nitrogen sources, which are poor in the termite diet (3). These dual functions of the endosymbiont as well as the cellulolytic ability of the protist may enhance the adaptability of the host termite to xylophagy, and are also responsible for the stability of the protist in the gut. In the gut of R. speratus, the endosymbiont species of Teranympha mirabilis was shown to be closely related to the endosymbiont of Eucomonympha sp. (32). Nevertheless, the distribution and phylogenetic relationships of the endosymbiotic Treponema among Teranymphidae protists have not yet been investigated.
In the present study, we identified the endosymbionts of Teranymphidae protists in Reticulitermes and H. sjostedti termites. The molecular phylogenies of the three symbiotic partners, Teranymphidae protists, their host termites, and their Treponema endosymbionts, were inferred and compared in order to obtain a clearer understanding of the evolutionary history of this triplex symbiotic relationship.

Data collection
Protist samples investigated in the present study and their host termites are listed in Table 1. All termite samples were collected in Japan (Fig. S1) and maintained in plastic boxes before use. Living termites were used as the material in the phylogenetic analyses of insects, protists, and endosymbiotic bacteria.
Termite DNA was extracted from the head and legs, and used as a template for PCR, as described previously (25). Mitochondrial COI, COII, and 16S rRNA genes were amplified, and used directly for DNA sequencing, as described previously (25). Teranymphidae protists were consistently found in the hindgut flora of the respective termites and were easily recognizable based on their morphological characteristics (6). The protist cells in the hindgut suspension of each termite were isolated manually and washed extensively under a microscope equipped with a micromanipulator (Cell Tram; Eppendorf, Hamburg, Germany), as described elsewhere (25). Each cell showing a typical morphology was isolated and was used as a template for isothermal whole-genome amplification (WGA), as previously described (27); amplified genomic DNA was used as a template for PCR. Regarding the expressed small subunit ribosomal RNA (SSU rRNA), 20 cells were directly used as a template for first-strand cDNA synthesis with the primer PZR1 (25,27) and the resulting cDNA was used for PCR amplification.
The protist gene encoding SSU rRNA, and bacterial 16S rRNA and gyrB genes were amplified using the WGA sample prepared from the same isolated cell as a template. The three genes were amplified by PCR, using the high fidelity DNA polymerase KOD Plus Neo (Toyobo, Osaka, Japan), with the previously described primers (25). The amplified products were separated by agarose gel electrophoresis, purified, and then cloned using the Zero Blunt pCR4-TOPO PCR cloning kit for sequencing (Invitrogen, Carlsbad, CA, USA) and Competent Quick DH5α (Toyobo). Clones containing inserts of the expected size were picked and partially sequenced, and the complete DNA sequence of each representative clone was obtained. In each sample, 3-9 clones were analyzed for the Teranympha SSU rRNA gene, while 1-4 clones were analyzed for the Eucomonympha SSU rRNA gene. Most of the clone sequences of Eucomonympha spp. were almost identical to either one of the three phylotype sequences identified in the whole gut community of H. sjostedti (unpublished data).
Eight clones were typically analyzed for the bacterial 16S rRNA and gyrB genes in each WGA sample; the partial DNA sequences of clones were sorted into phylotypes at >98.5% (16S rRNA gene) and >99.4% (gyrB gene) sequence similarities. The 16S rRNA gene sequences that were distantly related to those of endosymbiotic Treponema were excluded from further analyses. In the case of gyrB, the phylotype sequences sharing high sequence similarities with the Treponema sequences were phylogenetically analyzed (Fig.  S2), and only the sequences identified as the endosymbiotic Treponema species were concatenated for the phylogenetic analysis. The sequences obtained in this study have been deposited in the DNA Databank of Japan and the accession numbers are shown in Tables S1 and S2.

Phylogenetic analyses
The DNA sequences obtained in the present study and publicly available sequences were aligned using ClustalW2 (19) and alignments were refined manually. Ambiguously aligned positions were omitted from the subsequent phylogenetic inference analysis. An appropriate model of sequence evolution was selected using the program Jmodeltest 2.4 (8). In the tree from each dataset, a maximum likelihood (ML) analysis was conducted in RAxML 8.2 (38) using the GTRGAMMAI model. Bootstrap values were obtained from 10,000 replicates for the SSU rRNA gene. A Bayesian analysis was performed in MrBayes 3.2.1 (34) using the GTR+I+Γ model. The starting tree was random, and four simultaneous Markov chains in duplicate were run for 10,000,000 generations. Log likelihoods stabilized well before 2,500,000 generations, and the remaining generations were used to measure Bayesian posterior probabilities.
The alignments of bacterial 16S rRNA and gyrB gene sequences were concatenated manually, and the ML tree was estimated in RAxML using the GTRGAMMAI model. Parameters and branch lengths were individually optimized for each partition, and bootstrap values were obtained from 10,000 replicates. The GTR+I+Γ model was employed in the Bayesian analysis, and parameters and branch lengths were individually optimized for each partition. The starting tree was random, and four simultaneous Markov chains were run in duplicate for 10,000,000 generations. As described above, log likelihoods stabilized well before 2,500,000 generations, and the remaining generations were used to measure Bayesian posterior probabilities.
Three mitochondrial genes (mt16S rRNA, COI, and COII) were used in host termite phylogeny analyses, as described previously (25). The ML analysis was performed in RAxML using the GTRGAMMA model. Parameters and branch lengths were individually optimized for each partition, and bootstrap values were obtained from 10,000 replicates. A Bayesian analysis was performed in MrBayes using the HKY+Γ model, as described above.

Cophylogenetic analyses
Cophylogenetic analyses were conducted using five host Reticulitermes spp. and 12 samples of the protists and their endosymbionts. Tree topologies reconstructed using the ML method for each phylogenetic analysis were used in the analysis of the congruence of host and symbiont phylogenies. The Jane 4 program with 100 generations and a population size of 100 was used (7). The analysis also tested the significance of topological congruence between the hosts and symbionts. Cophylogeny mapping in Jane 4 employed heuristics to find solutions that minimized the overall cost of a historical reconstruction. The default event costs were as follows: 0 for a codivergence event, 1 for duplication and host switch events, and 2 for loss events. TreeMap (Charleston, M. TREEMAP 3.0 beta. Available at http://sites.google.com/site/cophylogeny) was used to construct the tanglegram for hosts and symbionts.

Fluorescence in situ hybridization (FISH)
FISH was performed to detect endosymbionts, as described previously (21,24,35). Briefly, termite gut contents were fixed in 4% paraformaldehyde; fixed cells were then spotted onto a silane-coated glass slide (Matsunami Glass, Osaka, Japan). After dehydration in ethanol, the slides were incubated with the hybridization solution (0.9 M NaCl and 0.1 M Tris-HCl) containing fluorescently-labeled probes at 48°C for 2 h. Specimens were washed for 20 min in a washing buffer (0.2 M NaCl and 0.1 M Tris-HCl) at 48°C. They were then mounted using a Fluoro-Keeper antifade reagent (Nacalai Tesque, Kyoto, Japan) and observed under an Olympus epifluorescence microscope; BX-63 (Olympus, Tokyo, Japan). Probes for the intracellular Treponema species and general eubacterial probes (EUB338) were described previously (32). The 5ʹ-termini of the probes were labeled with 6-carboxyfluorescein (6-FAM) or Texas red.

Host termites
Five Reticulitermes termites that harbored Teranympha protists in their guts were analyzed ( Table 1). The distribution of these termites and sampling points are shown in Fig. S1. Reticulitermes termites have distinct geographical distributions in Japan, and H. sjostedti distributes sympatrically with R. speratus and R. amamianus. The phylogenetic relationship between these host termites was inferred from the concatenated data of the three mitochondrial genes of 16S rRNA, and cytochrome oxidase subunits I and II. In the ML tree ( Fig.  1A), each sequence of the five termite species was closely related to that of the corresponding species in the database, and every species formed a robust monophyletic group. Although the order of their divergence was not fully resolved because of poor bootstrap support in some nodes, the tree topology was congruent with that reported previously (9,28).

Teranympha and Eucomonympha protists
The molecular phylogeny of the gut symbiotic protists of the genera Teranympha and Eucomonympha was inferred based on their SSU rRNA gene sequences (Fig. 1B). Each Reticulitermes termite harbored a morphologically unique Teranympha species, and most of the sequences obtained from two independent single protist cells from each termite host typically shared high sequence similarity (>98.0%). However, two sequences sharing 88.0% sequence similarity in R. amamianus (Ra2Tera1bSSU3 and Ra2Tera1bSSU6) were obtained from one single-cell sample of Teranympha; two sequences (Ra2Tera1cSSU2 and Ra2Tera1cSSU1) sharing 90.0% similarity were obtained from the other sample. Among these four, Ra2Tera1bSSU3 and Ra2Tera1bSSU2 shared more than 98% sequence similarity. We examined the transcribed rRNA sequences by cloning the RT-PCR products from a pool of Teranympha cells in this termite; the sequences of fifteen clones were almost identical to one another, and to Ra2Tera1bSSU3 and Ra2Tera1cSSU2 (>98%). Teranympha sp. in R. amamianus appear to have harbored at least two copies of the SSU rRNA gene in the genome, only one of which was functional. A highly similar sequence was obtained from the two single-cell samples of protists in R. kanmonensis and R. yaeyamanus (two samples per termite); however, the distantly related sequences, RkTera1xSSU3 and RyTera1BSSU4, respectively, were also obtained and appeared to be the same as that in R. amamianus. Consequently, we analyzed the phylogenetic relationships of the protists with sequences of the presumed functional genes only, and the tree topology obtained was used for subsequent comparisons (Fig. S4 and Table S2). Based on these data, Teranympha species in R. okinawanus and R. amamianus were grouped together, and this group was a sister to the group of Teranympha species in R. kanmonensis and R. yaeyamanus. T. mirabilis from R. speratus formed a clade together with the other Teranympha species, but branched out most basally.
In the case of the Eucomonympha protists, the sequences from seven single-cell samples were clustered into three sequence groups (A, B, and C in Fig. 1B), although group C was weakly supported. The clones in each group showed high sequence similarities (more than 99%), except for those in group C (94.3%). These sequences were detected in our analyses of hundreds of clones of the SSU rRNA gene amplified by RT-PCR from the whole gut community of H. sjostedti as abundant clones (unpublished data), suggesting that the three sequence groups identified herein are functional and represent major populations of Eucomonympha spp. in the gut of H. sjostedti. These sequences of Eucomonympha protists formed a sister group to Teranympha species from the five Reticulitermes termites. The currently described Eucomonympha lineages are paraphyletic (30). As reported previously (4,30), Eucomonympha imla in the wood-feeding cockroach Cryptocercus punctulatus is distantly related to Eucomonympha spp. in H. sjostedti, and the genus of the latter species needs to be reclassified.

Endosymbiotic bacteria
The 16S rRNA gene sequences of the endosymbiotic bacteria were obtained from each single-cell sample of the Teranympha and Eucomonympha species. Only a single phylotype affiliated with the genus Treponema was detected in each sample. Most of the sequences shared more than 99.5% identity, except for one sample from which a sequence sharing 98.5% identity with others (RkTera1B) was obtained. The representative sequences derived from the same termite species shared high sequence similarities (99.6-100%), while the sequences from other termite species shared lower sequence identities than those from conspecies. As previously reported  (32), sequences of the endosymbionts from both protist genera formed a phylogenetic cluster distinct from the ectosymbiont lineages of termite-gut protists in the termite Treponema cluster II (14,32). The endosymbionts of Teranympha spp. formed a monophyletic clade, which was clearly separated from the clade formed by the endosymbionts of Eucomonympha spp. (Fig. S2).
We also obtained sequences of the gyrB gene of the endosymbionts. Since the 16S rRNA gene sequences derived from two single-cell samples of Teranympha spp. shared high similarity, only one single-cell sample for each host Teranympha termite species was analyzed. Although many of the sequences obtained were closely related to the genomic sequence of the Treponema endosymbiont of Eucomonympha sp., several distantly related sequences were also obtained; some of these were affiliated with the genus Treponema and others were related to those encoded by bacteria from the family Opitutaceae (phylum Verrucomicrobia) or the genus Ureaplasma (phylum Tenericutes), both of which are the intracellular bacteria of Eukaryotes. However, except for the Treponema-like sequences, frame-shift insertions/deletions or stop codons were detected in these sequences. Since 16S rRNA pseudogenes have been reported in the intranuclear Verrucomicrobia symbionts of their host protist's genome, and the related intranuclear bacteria are widely distributed in termite-gut protists (36), these gyrB pseudogenes appear to be derived from intranuclear symbionts. As in the case of phylogenetic analyses based on the 16S rRNA gene, the Treponema endosymbionts formed a monophyletic cluster, and two clades comprising Teranympha and Eucomonympha endosymbionts were detected (Fig. S2).
In both phylogenetic trees, the branching orders within each clade of the Teranympha and Eucomonympha endosymbionts were poorly resolved because of their high sequence similarities ( Table 2). In the case of the endosymbionts of Eucomonympha spp., their phylogenetic relationships did not appear to correlate with the host protist's phylogeny; the endosymbiont sequences from the same Eucomonympha species were not always grouped together. In order to overcome the poor phylogenetic resolution of the single marker gene analysis, the concatenated sequences of 16S rRNA and gyrB genes were used in the phylogenetic analysis (Fig. 1C). Although the supporting values at each node were not significant, the Teranympha endosymbionts formed a monophyletic clade that was a sister of, but clearly separated from the Eucomonympha endosymbionts.

Topology-based analyses of host-symbiont relationships
The host and symbiont phylogenies reconstructed by the ML method were used to assess their phylogenetic congruence (Fig. 2). The tanglegram and relationships among the host Reticulitermes termites and Teranympha protists revealed a significant coevolutionary congruence, although the tree topologies did not perfectly match (Fig. 2A). The evolutionary event that was incongruent in the two trees occurred at a weakly supported node. We then used Jane 4 (7) to generate  coevolutionary events with the lowest cost sets (Fig. 2B). A significant global cost (P=0.02) was observed, which was 3 for cospeciation; 0 for duplication; 1 for host-switch; 1 for loss; and 0 for failure to diverge between the host termites and Teranympha protists (Table 3).
Reconciling the phylogenies of host protists and their endosymbionts indicated a significant topological congruence (P=0.04), with seven codivergence events out of a possible 11, four host switches, and three losses (Table 3). Similar to the relationship between the host termites and their gut protists, the tree topologies of the Teranympha protists and their endosymbionts were largely congruent with three codivergence events out of a possible four (Fig. 2C and D), and incongruence was observed at a weakly supported node. In contrast, the coevolutionary relationship between the Eucomonympha protists and their endosymbionts was complex, with three host switch events (Fig. 2C and D). These incongruence events occurred at strongly supported nodes of the endosymbiont tree. The endosymbiont of the host protist group C (Hs2EC-c) was closely related to the endosymbionts of the two host protists in group A (Hs1EA-a and Hs2EA-b), and the endosymbionts of the host protists in groups B and C (Hs3EB-a and Hs3EC-c, respectively) were also closely related (see also Fig. 1C).

In situ identification of Teranympha endosymbionts and their morphology
A previously used oligonucleotide probe (IIC-637) targeting the 16S rRNA of the Treponema endosymbionts of Eucomonympha spp. was used to detect the endosymbionts of Teranympha spp. by FISH because, as shown above, the targeted sequence was conserved (32). This probe annealed to the rod-shaped endosymbiotic bacteria of Teranympha protists in the gut of all Reticulitermes termites examined (Fig. 3A, D, G, J, M, and Q). The same rod-shaped endosymbionts were detected during simultaneous hybridization of a universal bacterium probe (Fig. 3B, E, H, K, N, and R). Although the universal probe also detected ectosymbiotic bacteria, e.g., spirochetes attached at the posterior end of Teranympha cells (Fig. 3R), both probes simultaneously detected all bacteria in the protist cytoplasm. Most of the Teranympha cells harbored endosymbiotic bacteria that were homogeneously and densely packed in the cytoplasm, except for the anterior end of the cell, the rostrum (Fig. 3A, D, G, J, M, and Q). The endosymbionts of T. mirabilis in the gut of R. speratus (Fig. 3M, N, O, P, Q, R, S, T, and U) appeared to be more densely packed than Teranympha spp. in other termites. The Teranympha protists in R. yaeyamanus appeared to be colonized to a lesser  extent, with the endosymbionts being restricted to the anterior of the cells, except for the rostrum (Fig. 3J, K, and L). Transmission electron microscopy of ultrathin sections of T. mirabilis revealed the presence of numerous endosymbiont cells in the cytoplasm of T. mirabilis (Fig. 3U), as previously reported (32). The endosymbionts were 0.42±0.07 μm in width (mean±standard deviation) and 2.09±0.29 μm in length (n=9), and were longer than the Eucomonympha endosymbionts (1.33±0.31 μm×0.41±0.05 μm) (32). Endosymbiont cells possessed a single cell membrane and were surrounded by an external membrane that may have been derived from the host protist, as in the case of the endosymbionts of Eucomonympha sp. (Fig. 3U).

Discussion
Cophylogenetic analyses of gut protists from the family Teranymphidae, and their endosymbionts in five Reticulitermes termites and H. sjostedti revealed their possible cospeciating evolutionary relationships, particularly between the triplex symbiotic partners of Teranympha protists, their endosymbionts, and their host termites. Incongruence in evidence against a completely cospeciating relationship was observed, but occurred at nodes showing poor resolution, i.e., for the relationship between R. yaeyamanus and R. kanmonensis, or between their symbionts. Therefore, we consider cospeciation to be the most plausible evolutionary scenario, albeit with the possibility of deviations.
A similar cospeciating relationship between the triplex symbiotic partners was reported for Pseudotrichonympha protists, their Bacteroidales endosymbionts, and their host termites from the family Rhinotermitidae (except for the genus Reticulitermes) (25). As previously discussed, the strict vertical transmission of gut symbionts is likely warranted by the proctodeal feeding behavior that characterizes the sociality of termites (31).
The termite gut fauna is discarded at every molting, but is acquired from nestmates; it is also carried over to a newly founded colony by the primary termite reproductives (31). Since endosymbionts are abundant in the protist cell, the cell is stably transmitted during protist cell division. The crucial roles played by the protists and their endosymbionts in their host termites, i.e., cellulose decomposition by the former and nutrient supply by the latter, appear to have had a significant impact on the maintenance of these symbiotic partners and their mutual adaptation.
In contrast to Pseudotrichonympha, which are widely distributed among termites from the family Rhinotermitidae, the Teranympha protists have only been found in Reticulitermes termites and are exclusively limited to the Asian species in this genus; however, the composition of other protist species in all Reticulitermes termites appears to be similar (17). In the present study, we confirmed this close phylogenetic relationship, which suggests a possible common ancestor of Teranympha spp. and Eucomonympha spp. in the H. sjostedti termite. This observation also supports the previously proposed horizontal transfer of the gut fauna between Reticulitermes and Hodotermopsis termites (18). If this is the case, the absence of Teranympha likely represents a secondary loss after fauna transfer, with other members of the gut fauna possibly complementing the symbiotic roles of Teranympha and its endosymbionts. Furthermore, the Treponema endosymbionts of Teranympha spp. and Eucomonympha spp. form a closely related sister group. The results of our analyses imply that an ancestor of both protist groups acquired the endosymbiont before fauna transfer and that, after this transfer, the protists and their endosymbionts were vertically inherited together over a long evolutionary time period and cospeciated according to the species differentiation of the host termites.
In Teranympha and Pseudotrichonympha cases, only a single protist species appeared to be present in the gut microbial community of a single host termite species. This may affect the cospeciating evolutionary process because the spatial separation of host protist lineages creates the potential for accepting endosymbionts; hence, the opportunity for a host switch does not exist. In contrast, Eucomonympha spp. are sympatric and present simultaneously within the gut community of a single host termite. Although cospeciation appears to be a general rule for the Teranympha and Eucomonympha Panels O and S show differential interference contrast micrographs. A merged image of panels M and N is shown in panel P; that of panels Q and R is shown in panel T. A transmission electron micrograph of the endosymbionts in a T. mirabilis cell is shown in panel U. The amorphous yellow signals in panels A, D, G, and J are autofluorescence signals, mostly originating from wood particles incorporated into the protist cells by phagocytosis. Bars correspond to 50 μm in panels C, F, I, and L; 10 μm in panels P and T; and 1 μm in panel U.
protists and their endosymbionts, as shown by the overall significance in the cophylogeny analysis (Table 3), the phylogenies of Eucomonympha spp. and their endosymbionts are not always congruent. For example, the endosymbiont phylotype Hs2EC-c from group-C protists was more closely related to those from group-A protists than to Hs3EC-c. Furthermore, the endosymbiont phylotypes Hs3EB-a and Hs3EC-c from protist groups-B and -C, respectively, were very closely related, and the endosymbiont phylotype Hs2EA-a from group-A protists was distantly related to the other endosymbionts from group-A protists. These examples of incongruence in evidence against cospeciation suggest more frequent host switches of endosymbionts in Eucomonympha spp. than in Teranympha spp., possibly because multiple protist species are present in the same gut community. A similar host switchlike relationship was reported for Trichonympha sp. and its endosymbiont "Ca. Endomicrobium trichonymphae" in the gut of H. sjostedti, with multiple Trichonympha species being present in the gut of this termite (15). However, this host switch-like relationship is the only exception among the strict cospeciating relationships of Trichonympha protists and Endomicrobium endosymbionts. The genomes of the endosymbionts of Trichonympha and Pseudotrichonympha are small in size, likely as a consequence of reductive evolution (12,13). These endosymbiont species with reduced genomes may be more dependent on the host protists than the endosymbionts of Eucomonympha, and, thus, their opportunity for a host switch may be reduced. In contrast, the genome size of the endosymbiont of Eucomonympha was not as severely reduced, although its gene content, as evaluated by a cluster of orthologous group analysis, was similar to the genomes of the Trichonympha and Pseudotrichonympha endosymbionts; for example, the relative abundance of cell motility genes is universally lower, and that of the translation and coenzyme metabolism genes is universally higher in the genomes of the endosymbionts of termite-gut protists than in cultured treponemes (32). The acquisition of the endosymbiont by Eucomonympha protists appears to be a recent evolutionary event, and the endosymbiont species is at an initial developmental stage of adaptive evolution. The endosymbiont of Eucomonympha may be more independent of the host protists than the other two, and may be able to switch the host lineages more frequently.
In the present study, we confirmed that Treponema endosymbionts are widely and commonly distributed among the Teranympha and Eucomonympha protists examined to date. As discussed above and previously (32), Treponema endosymbionts are monophyletic and appear to have originated from an ectosymbiont lineage of termite-gut protists. Nevertheless, the phylogenetic analyses revealed that the endosymbionts of Teranympha spp. clearly form a distinct clade from those of Eucomonympha spp. The sequence similarity of the 16S rRNA gene of the endosymbionts of Teranympha species (intra-genus) was greater than 99.25%, whereas the sequence similarity of the endosymbionts between Eucomonympha and Teranympha species (inter-genera) was less than 97.1% (Table 2). It is generally accepted that bacteria sharing less than 98.7% 16S rRNA gene sequence similarity need to be classified into distinct species because this sequence similarity corresponds to a DNA reassociation value less than 70%, i.e., a threshold of bacterial species delineation (37). The shared sequence similarity of the gyeB gene of the endosymbionts was also high among Teranympha spp., but low between Teranympha and Eucomonympha protists ( Table 2). Based on these findings, we propose classifying the endosymbiont species of Teranympha into a species distinct from 'Ca. T. intracellularis', the endosymbiont of Eucomonympha spp. (32). Therefore, we propose a novel species "Candidatus Treponema teratonymphae" for Teranympha endosymbionts. Although we used the popular original name of the genus "Teranympha" in the present study, its erroneous orthography was pointed out earlier (Teratonympha is the proper name) (11), and we adopted the proper form for the nomenclature of the Treponema endosymbiont species.
Description of "Candidatus Treponema teratonymphae" Te.ra.to.nym'phae. N.L. gen. n. teratonymphae, of Teratonympha, referring to the generic name of the host protist. Cells are rods, 1.64-2.39 μm×0.32-0.51 μm in size (average 2.09±0.29 μm×0.41±0.07 μm). They lack flagella and are surrounded by two membranes; the outer membrane is presumably the host-derived membrane. The bacterium is specifically found in the cytoplasm of the parabasalian protist Teranympha (or its amended Teratonympha) spp., in the hindgut of Reticulitermes spp. termites. Thus far uncultured, but forms a monophyletic group based on sequence analyses of the 16S rRNA and gyrB genes (DDBJ accession numbers: LC276704-LC276733).