“Candidatus Finniella” (Rickettsiales, Alphaproteobacteria), Novel Endosymbionts of Viridiraptorid Amoeboflagellates (Cercozoa, Rhizaria)

Axenic cultures of Viridiraptor invadens and Orciraptor agilis, which were established through single-cell sorting, were grown with the corresponding food algae in liquid culture as previously described (1). The cultures were regularly checked for bacterial contaminants with 1% (vol/vol) bacterial standard medium (0.8% peptone, 0.1% glucose, 0.1% meat extract, 0.1% yeast extract) (10) in liquid cultures and on agar plates. Neither extracellular bacteria in the liquid cultures (high-resolution differential interference contrast [DIC] and 4′,6′-diamidino-2-phenylindole [DAPI] staining) nor bacterial colonies on the ager plates could be observed. Free-swimming or gliding viridiraptorid cells were separated from algal remnants with a plankton net (48 μm), and in some strains (OrcA01 and OrcA02), subsequent cell sorting with a BD FACSVantage SE (Becton-Dickinson, Mountain View, CA, USA) with a 100-μm jet nozzle was used to collect about 10,000 cells per well for DNA extraction. DNA was extracted with the DNeasy blood and tissue kit (Qiagen GmbH, Hilden, Germany) with one modification: lysis buffer with proteinase K from the kit was added to frozen viridiraptorid cells in reaction tubes or wells of a 96-well microtiter plate. Subsequently, the samples were incubated at 65°C in a water bath or in a thermocycler and mixed by vortexing or pipetting to destroy the cells. The supernatant was processed according to the protocol of the manufacturer. The eubacterial small-subunit (SSU) rRNA gene was amplified by PCR (11) using DreamTaq DNA polymerase (Fermentas, St. Leon-Rot, Germany) and modified primers previously designed for plastids (SG1_baci and SG2_baci) as well as the primer Fin_H28 (Table 1). The primer Fin_H28 was designed with the primer-designing tool of NCBI (http://www.ncbi.nlm.nih.gov/tools/primer-blast/) and evaluated for specificity with Probe Match (https://rdp.cme.msu.edu/probematch/search.jsp) and TestProbe 3.0 (http://www.arb-silva.de/search/testprobe/). The numbers of accessions matched with both search tools are shown in Table 1. The PCR was performed with an initial denaturation step at 95°C for 180 s, followed by 30 cycles including denaturation (95°C for 45 s), annealing (60 s), and elongation (72°C for 180 s). Annealing temperatures were 55°C for the primers SG1_baci/SG2_baci and 60°C for SG1_baci/Fin_H28. For cleaning of PCR products, DNA was precipitated with isopropanol (50% isopropanol, −20°C, 1 h), centrifuged (17,000 × g, 20 min), subsequently washed in 80% ethanol, dried, and finally dissolved in distilled water. The cleaned PCR products were sequenced with the primers mentioned above using commercial ABI Sanger sequencing (Eurofins MWG Operon Sequencing Department, Martinsried, Germany), and the resulting sequences were assembled with the software AlignIR 2.0 (LI-COR Biosciences, Lincoln, NE, USA). The sequences obtained for the four strains of “Ca. Finniella” were critically proofread (electropherograms), compared to other rickettsial sequences, and trimmed at their ends according to the sequence quality.

The sequences obtained were subjected to BLAST searches (http://blast.ncbi.nlm.nih.gov/) that unambiguously indicated an affiliation with the Rickettsiales (“Ca. Paracaedibacter acanthamoebae” was the closest hit, with an identity of 89%). Therefore, an alignment comprising four sequences of “Ca. Finniella,” 93 sequences of other Rickettsiales (as traditionally defined, i.e., including Holospora, Caedibacter, and relatives, excluding “Ca. Pelagibacter”) and 78 sequences of other Alphaproteobacteria from various orders was constructed using the SILVA database (http://www.arb-silva.de/) (12, 13) and NCBI (http://www.ncbi.nlm.nih.gov/). The alignment was manually refined with the software SeaView 4.5.4 (14, 15) using information on the SSU rRNA secondary structure; i.e., certain sequences were folded on the Mfold web server (16). Different sets of taxa and selected nucleotide sites from the resulting alignment were subjected to phylogenetic analyses (maximum likelihood) using raxmlGUI (17, 18) to test the influence of taxon sampling and selection of sites. Sites were critically evaluated in terms of sequence coverage and alignment quality and were manually selected (e.g., exclusion of hypervariable regions). Finally, two sets of sequences were analyzed with maximum likelihood (ML), Bayesian inference (BI), and neighbor joining (NJ): (i) a character set of 1,299 sites, including 97 rickettsial sequences, and (ii) a character set of 1,266 sites, including all sequences mentioned above (Alphaproteobacteria). For ML analyses, 100 trees were retrieved with the PTHREADS version of RAxML 7.7.2 (17), and NJ analyses were done with the software PAUP 4.0 (19). For Bayesian inferences, the program MrBayes 3.2.2 was used without the covarion model (20, 21). The Bayesian analyses included two Markov chain Monte Carlo (MCMC) chains and 5,000,000 generations. Trees were sampled every 1,000 generations and the burn-in (1,250,000 generations) was determined by the convergence criterion. The model used was GTR+G+I. For ML and NJ, 1,000 bootstrap repetitions were performed for each analysis. The resulting support values as well as posterior probabilities of BI are displayed at the branches of the tree figures, which were designed with SeaView and Adobe Illustrator CS4 (Adobe Systems, Munich, Germany).

The sequences of “Ca. Finniella” were aligned to the global alignment of the SILVA database with the online SINA aligner (http://www.arb-silva.de/aligner/), subsequently downloaded, and added to the SILVA alignment Ref NR 99 using the ARB software (22, 23). The ARB probe design tool was used to search for probes specific for either the candidate genus (“Ca. F. inopinata” and “Ca. F. lucida”) or for one of the “Ca. Finniella” species (“Ca. F. inopinata” or “Ca. F. lucida”). The proposed probes were checked for specificity in silico with the probeCheck online database (24) and with BLAST of NCBI (http://blast.ncbi.nlm.nih.gov/). Probes with high specificity were subjected to mismatch analyses with the online tool mathFISH to determine parameters (temperature and formamide concentration) for optimal stringency of the probe (25, 26). If needed, probes were manually modified according to these results and checked again with probeCheck, BLAST, and mathFISH. Three oligonucleotide probes targeting (i) “Ca. Finniella (probe FIN93), (ii) “Ca. Finniella inopinata” (probe FININ731), and (iii) “Ca. Finniella lucida” (probe FINLU731) were synthesized and labeled with a cyanine dye (Cy3) or fluorescein amidite (6-FAM) by the biopolymer factory biomers (biomers.net GmbH, Ulm, Germany). Probes were tested with viridiraptorid cultures and uncharacterized mixtures of heterotrophic Eubacteria in various formamide (FA) concentrations (0 to 30%) using the slightly modified FISH protocols of Manz et al. (27) and Hugenholtz et al. (28) as shown below. The sequences of oligonucleotide probes used in this study are listed in Table 1 and have been deposited at probeBase (29).

Viridiraptorid cells were concentrated by gentle centrifugation (200 g, 10 min), fixed with 4% formaldehyde solution for 10 min, washed with distilled water and finally applied to gelatin-coated glass slides. After the cells air-dried, they were incubated in ethanol (96%) for about 5 min and dried again. Hybridization buffer with an appropriate concentration of FA and probe (5 ng/μl) was added to the dried samples and incubated for 2 to 3 h at 47°C in a moist chamber. Subsequently, samples were rinsed with and incubated in washing buffer for 45 min at 48°C, rinsed in distilled water for 30 to 60 s, and air-dried. Slides were covered with DAPI solution (1 μg/ml), incubated for 5 to 10 min, rinsed with distilled water, air-dried, and finally mounted with medium containing 85% glycerol, 14% phosphate-buffered saline (PBS), and 1% 1,4-diazabicyclo[2.2.2]octane (DABCO). The processed cells were examined and documented by confocal laser scanning microscopy (CLSM) using the Leica TCS SPE system and Leica LCS software (Leica Microsystems, Wetzlar, Germany). The obtained CLSM data were processed with the Java image processing program ImageJ (http://rsbweb.nih.gov/ij/) in terms of pseudocolors, contrast, merging, and Z projection. For transmission electron microscopy (TEM), viridiraptorid cells were chemically fixed, dehydrated and, embedded in Epon as previously described (2). Ultrathin sections (60 nm) were examined with a CM 10 transmission electron microscope (FEI Europe, Eindhoven, The Netherlands). Micrographs were taken with the Orius SC200W TEM charge-coupled device (CCD) camera and DigitalMicrograph software (Gatan Inc., Pleasanton, CA) and processed in terms of brightness and contrast with Adobe Photoshop CS4 (Adobe Systems, Munich, Germany).

The sequences determined here were deposited in GenBank and are available under the accession numbers shown in Table 2.

Intracellular bacteria could be detected in four viridiraptorid strains by DAPI staining and FISH with the probes EUB338 (30) and FIN93 (Table 2), whereas five other viridiraptorid strains (VirI01, VirI03 to VirI05, and OrcA03) showed negative results (Fig. 1A to F). Numerous rod-shaped bacteria about 1 to 1.5 μm in length were scattered throughout the cytoplasm but were not detected in the nucleus of the host (Fig. 1B and C). The bacteria occurred in all life history stages of Viridiraptor invadens and Orciraptor agilis in large numbers: free-swimming, starving cells of Viridiraptor invadens (strain VirI06) contained about 50 bacterial cells (mean, 54; range, 33 to 121; n = 20), whereas starving cells of Orciraptor agilis (strain OrcA01) contained about 30 bacteria (mean, 30; range, 17 to 41; n = 20). Postphagocytotic cells of Viridiraptor invadens (strain VirI06) located within the cells of the prey alga Zygnema pseudogedeanum contained about 60 bacteria on average (mean, 57; range, 40 to 82; n = 17) (Fig. 1G and H). Divisional stages of bacteria could be frequently observed (Fig. 1H, arrowheads). With the specific FISH probes FININ731 and FINLU731, it was possible to discriminate between two distinct bacterial populations which specifically resided in the viridiraptorid species: “Ca. Finniella lucida” in Orciraptor agilis and “Ca. Finniella inopinata” in Viridiraptor invadens. As predicted by the mismatch analysis using mathFISH, the discriminatory power of the probes, which differed in only a single nucleotide, increased with the formamide concentration. A clear-cut separation of the signals from both probes was achieved with ≥25% FA and a hybridization temperature of 47°C (Fig. 1I to M).

The cells of “Ca. Finniella inopinata” (previously detected by TEM in Viridiraptor invadens strain VirI02 [2]) measured about 1 to 1.7 by 0.3 to 0.4 μm and displayed two discernible membranes, the outer of which was covered by a layer of electron-dense particles (see Fig. 1G in reference 2). However, the quality of preservation was considerably better in O. agilis, and thus we focused here on the yet-unknown ultrastructural characteristics of “Ca. Finniella lucida.” As shown by fluorescence microscopy, the cells of “Ca. F. lucida” were scattered throughout the cytoplasm (Fig. 2A; arrows), were rod shaped, and measured up to 1.6 μm in length (dividing cells). In cross section, the cells appeared roundish or slightly ellipsoid and were about 450 nm (mean, 446 nm; range, 372 to 476 nm; n = 10) in width (Fig. 2A and B). The cytoplasm of “Ca. F. lucida” was surrounded by two wrinkled membranes displaying the characteristic trilaminar structure of lipid bilayers (Fig. 2C and D, arrowheads). The periplasmic space of these two bacterial membranes was very narrow, often hardly discernible. Furthermore, the cells were surrounded by a smooth, sometimes discontinuous, electron-dense envelope or layer which never displayed a trilaminar structure and was well separated from the two (bacterial) membranes by an electron-translucent gap of about 24 nm (range, 18 to 30 nm) (Fig. 2B to E, asterisks). Dividing bacteria were characterized by a central constriction (Fig. 2E).

We obtained four 16S rRNA gene sequences of endosymbiotic bacteria of viridiraptorids, two from Viridiraptor invadens (strains VirI02 and VirI06) and two from Orciraptor agilis (strains OrcA01 and OrcA02) (Table 2). The sequences of the bacteria found in V. invadens (= “Ca. F. inopinata”) were identical, whereas the sequences obtained from O. agilis (= “Ca. F. lucida”) differed in two nucleotides. The difference between the sequences of “Ca. F. inopinata” and “Ca. F. lucida” was 10 nucleotides (out of 1,429), corresponding to a 16S rRNA sequence identity of 99.3%. The tree in Fig. 3 presents the result of sequence comparisons of the obtained sequences in the context of 93 rickettsial sequences. The phylogeny recovered six deep-branching clades displaying moderate (73/1.0/96 [where the first and last values are bootstrap percentages and the middle value is posterior probability]) or high (≥97/1.0/100) support: Anaplasmataceae, Holosporaceae, Rickettsiaceae, “Ca. Midichloriaceae,” the Caedibacter/Nucleicultrix clade (clade C/N), and a so-far-unnamed clade containing “Ca. Odyssella,” “Ca. Paracaedibacter,” “Ca. Captivus,” and “Ca. Finniella” along with several environmental sequences. The latter clade, for which we propose the name “Ca. Paracaedibacteraceae” (see below), again comprised at least six mostly well-supported subclades. These subclades either contain the candidate taxa mentioned above or lack any phenotypically characterized representatives (clades P1 and P2) (Fig. 3). The subclade containing the sequences determined during this study together with the environmental sequences deposited under accession numbers GQ453073 and AB636944 is defined here as the genus “Ca. Finniella.” All currently known members of this clade can be specifically targeted with the FISH probe FIN93. Depending on the host species (Viridiraptor invadens versus Orciraptor agilis), the four new sequences of “Ca. Finniella” grouped into two very closely related, but distinct lineages that correspond to the candidate species “Ca. Finniella inopinata” and “Ca. Finniella lucida.” The clades most closely related to “Ca. Finniella” (clade P1 and P2) exclusively comprise environmental sequences, which originated from various geographically distant freshwater and soil ecosystems and the digestive systems of metazoa (Fig. 3).

Sequence comparisons of the extended alignment revealed that the Anaplasmataceae, Holosporaceae, Rickettsiaceae, “Ca. Midichloriaceae,” and “Ca. Paracaedibacteraceae” are also well or moderately supported by all methods (≥89/1.0/99) in a broad alphaproteobacterial context (Fig. 4). Clade C/N was recovered only by ML and NJ (84/−/76), since BI placed “Ca. Nucleicultrix” basal to Holosporaceae (though with no support) (see Fig. S2 in the supplemental material). Interfamily relationships could be resolved only for Rickettsiaceae, Anaplasmataceae, and “Ca. Midichloriaceae,” which formed a well-supported clade (Rickettsiales I), whereas the relationships of the other rickettsial clades remained unclear. Thus, the order Holosporales sensu Ferla et al. (31), which was based on sequences of Caedibacter caryophilus and “Ca. Odyssella thessalonicensis,” could not be recovered (conflicting topologies with no support) (Fig. 4; also, see Fig. S1 and S2 in the supplemental material). Since the sister relationship of clade C/N to Holospora, “Ca. Gortzia,” and “Ca. Paraholospora” lacks convincing support in both analyses (Fig. 3 and 4), we restrict the Holosporaceae to the three last-mentioned genera. As known from previous 16S rRNA gene phylogenies “Ca. Pelagibacterales” (clade SAR11) was sister to the Rickettsiales I (56/0.99/—), which was discussed as an artifact caused by sequence composition bias (for example, see references 32 and 33).

Since the novel Rickettsiales could not be grown in pure culture, i.e., outside the host cell, they do not fulfil the requirements of the Bacteriological Code (1990 Revision) for the description of valid prokaryote taxa (34, 35). Therefore, we follow the recommendations of Murray and Stackebrandt (36) and describe provisional taxa using the “Candidatus” status, namely, two candidate species and the corresponding candidate genus and family.

“Candidatus Paracaedibacteraceae” ([Rickettsiales, Alphaproteobacteria] not cultivated [NC]; not applicable [NA]; rod [R]; basis of assignment, phylogeny [see below]; symbiotic [S] [various protists, e.g., Acanthamoeba, Petalomonas, Orciraptor, Viridiraptor, cytoplasm]; aerobic; mesophilic [M]). Etymology: The name “Paracaedibacteraceae” derives from the name of the included candidate genus “Ca. Paracaedibacter” by adding the family suffix “-aceae.” Additional information: Rod-shaped cells with two membranes, host-derived (symbiosomal) membrane not clearly observed, probably absent; cells scattered throughout the cytoplasm of various protists including representatives from Rhizaria, Excavata, and Amoebozoa; vegetative bacterial cells known, no other life cycle stages observed; phylogenetically defined as a well-supported clade containing sequences of “Ca. Captivus,” “Ca. Finniella,” “Ca. Odyssella” and “Ca. Paracaedibacter” but excluding sequences of Caedibacter, Holospora, “Ca. Gortzia,” and “Ca. Paraholospora.” Proposed type genus: “Ca. Paracaedibacter.” Taxa currently included: “Ca. Captivus,” “Ca. Finniella,” “Ca. Odyssella,” “Ca. Paracaedibacter.”

“Candidatus Finniella” ([Rickettsiales, Alphaproteobacteria] NC; NA; R; nucleic acid sequence [NAS] [GenBank numbers KT343634, KT343635, KT343636, and KT343637], oligonucleotide sequence complementary to unique region of 16S rRNA is 5′-ACCCGTCTGCCACCTAAGTA-3′ [probe FIN93]; S [Viridiraptor and Orciraptor {Cercozoa}, cytoplasm]; Aer.; M). Etymology: finnia, -ae, f [Latin], Finland; -ella [Latin], diminutive suffix (feminine). The name “Finniella” refers to the lovely “Finnish spirit” that was present during the lab work on the novel Rickettsiales. Additional information: Numerous (up to several dozen) rod-shaped cells colonizing the cytoplasm of viridiraptorid amoeboflagellates (Cercozoa, Rhizaria); vegetative cells about 1 to 1.5 μm in length, no other life cycle stages observed; two membranes and a distinct halo discernible in TEM micrographs; phylogenetically defined to contain sequences of the candidate species “Ca. Finniella lucida” and “Ca. Finniella inopinata” but excluding the sequences of clade P1 (accession numbers AM991220, EF667926, GQ500786, HQ420131, JF917200, and JQ684365). Proposed type candidate species: “Ca. Finniella lucida.” Taxa currently included: “Ca. Finniella inopinata,” “Ca. Finniella lucida.”

“Candidatus Finniella lucida” ([Rickettsiales, Alphaproteobacteria] NC; NA; R; NAS [GenBank numbers KT343634 and KT343635], oligonucleotide sequence complementary to unique region of 16S rRNA is 5′-AGTACCGAGCCAGAAAACCGC-3′ [probe FINLU731]; S [Orciraptor agilis {Cercozoa}, cytoplasm]; Aer.; M). Etymology: lucidus, -a, -um (Latin), clear, bright, shining (figuratively: clear, perspicuous, lucid). The word “lucida” refers to the clear appearance of the cells in the TEM micrographs, which led to the discovery of “Ca. Finniella.” Additional information: About 20 to 40 cells colonizing the cytoplasm of Orciraptor agilis (Viridiraptoridae, Cercozoa, Rhizaria); phylogenetically defined to contain the sequences of strains FinLu01 (GenBank number KT343634) and FinLu02 (KT343635) but not those of “Ca. Finniella inopinata” (KT343636 and KT343637); can be discriminated from “Ca. Finniella inopinata” by fluorescence in situ hybridization with the probe FINLU731. Reference strain: FinLu01 (inhabiting Orciraptor agilis strain OrcA01). Locality of reference strain: Ditch with rotting leaves, Sphagnum sp., and delicate Mougeotia sp. in the spring fen near Neuenhähnen, Waldbröl, Germany; 50°50′27.91″N 7°32′8.40″E. Reference material (EM blocks): NHMUK 2013.7.1.5 to NHMUK 2013.7.1.8 (deposited at the National History Museum, London, United Kingdom).

“Candidatus Finniella inopinata” ([Rickettsiales, Alphaproteobacteria] NC; NA; R; NAS [GenBank numbers KT343636 and KT343637], oligonucleotide sequence complementary to unique region of 16S rRNA is 5′-AGTATCGAGCCAGAAAACCGC-3′ [probe FININ731]; S [Viridiraptor invadens {Cercozoa}, cytoplasm]; Aer.; M). Etymology: inopinatus, -a, -um [Latin], unexpected, unforeseen, surprising. The word “inopinata” refers to the late identification of the unexpected bacteria in the TEM preparations of Viridiraptor invadens. Additional information: About 30 to 120 cells colonizing the cytoplasm of Viridiraptor invadens (Viridiraptoridae, Cercozoa, Rhizaria); phylogenetically defined to contain the sequences of strains FinIn01 (GenBank number KT343637) and FinIn02 (KT343636), but not those of “Ca. Finniella lucida” (KT343634 and KT343635); can be discriminated from “Ca. Finniella lucida” and other Rickettsiales by fluorescence in situ hybridization with the probe FININ731. Reference strain: FinIn01 (inhabiting Viridiraptor invadens strain VirI02). Locality of reference strain: Myriophyllum basin in the Botanical Garden of the University of Bonn, Bonn, Germany; 50°43′28.31″N 7°5′35.64″E. Reference material (EM blocks): NHMUK 2013.7.1.1 to NHMUK 2013.7.1.4 (deposited at the National History Museum, London, United Kingdom).

The phylogenetic analyses and the FISH experiments proved that “Ca. Finniella” belongs to a deep-branching and well-supported clade within the Rickettsiales. This clade, here described as “Ca. Paracaedibacteraceae,” also includes “Ca. Odyssella,” “Ca. Captivus,” and “Ca. Paracaedibacter” and is characterized by at least four endosymbioses with phylogenetically distant groups of protists, namely, Amoebozoa, Excavata, and Rhizaria. “Ca. Odyssella” and “Ca. Paracaedibacter” colonize Acanthamoeba spp. (37–39), “Ca. Captivus” was found in unidentified protists that inhabit acid mine drainage (40), and the yet-undescribed Ric2 endosymbionts reside in Petalomonas sphagnophila (41).

The phylogenetic position of “Ca. Paracaedibacteraceae” within the Rickettsiales is still uncertain. Our phylogenetic analyses corroborated the close relationship of Anaplasmataceae, Rickettsiaceae, and “Ca. Midichloriaceae,” but the relationships of the other rickettsial lineages could not be resolved with the 16S rRNA gene. Previous phylogenetic analyses of Alphaproteobacteria using concatenated SSU and large-subunit rRNA gene sequences suggested a nonmonophyletic relationship of Caedibacter plus “Ca. Odyssella” and the rest of the Rickettsiales (Anaplasmataceae, Rickettsiaceae, and “Ca. Midichloriaceae”), leading to the introduction of the new order Holosporales for Caedibacter and “Ca. Odyssella” (31, 42). However, the taxon sampling of the orphan Rickettsiales (here defined as Holosporaceae, “Ca. Paracaedibacteraceae,” and clade C/N) was very limited in this study (two sequences). Since our phylogeny did not support the monophyly of Holosporaceae, “Ca. Paracaedibacteraceae,” and clade C/N, we refrain from adopting the ordinal name Holosporales for this assemblage and use the traditional definition of Rickettsiales (i.e., including Caedibacter, Holospora, and “Ca. Odyssella”) (43), pending further analyses.

In most multigene studies, members of the Holosporaceae, “Ca. Paracaedibacteraceae,” and clade C/N were absent or poorly represented (e.g., only “Ca. Odyssella thessalonicensis”), preventing clear conclusions about the interrelationships of these lineages and the monophyly of the Rickettsiales sensu lato (44–48). Recently, phylogenetic analyses and gene order patterns of newly generated genomic data from additional orphan Rickettsiales (Caedibacter acanthamoebae, two strains of “Ca. Paracaedibacter,” and the “NHP bacterium”) supported a monophyletic origin of Holosporaceae, clade C/N, and “Ca. Paracaedibacteraceae” as well as of the Rickettsiales sensu lato (33). In future, the genomic investigation of additional orphan Rickettsiales may help to resolve the controversial phylogeny of the Rickettsiales sensu lato and also to elucidate the origin of mitochondria (33).

Ultrastructural data for “Ca. Finniella” revealed a Gram-negative-type cell morphology (two bacterial membranes separated by a periplasmic space) and a halo-like translucent zone that surrounds the cell, both matching the characteristics of other Rickettsiales (for example, see references 37 to 39 and 49 to 53). However, “Ca. Finniella” also exhibits a third, electron-dense layer that was especially distinct in “Ca. F. lucida.” Since this layer lacks the trilaminar pattern of biological membranes, it is still uncertain whether it represents a host-derived membrane, a proteinaceous layer, or a preparation artifact. To set our ultrastructural observations in an evolutionary context, we analyzed published data on rickettsial ultrastructure in the context of the present phylogenetic tree (Fig. 3). Most of the rickettsial taxa do not display a third, host-derived membrane, i.e., the cells are located free in the host cytoplasm. Only two major rickettsial lineages, namely, Anaplasmataceae and “Ca. Midichloriaceae,” comprise members bound by additional (foreign) membranes. In Wolbachia species (Anaplasmataceae), a third, tight-fitting membrane can be clearly discerned (54–57), whereas the pathogenic representatives of the Anaplasmataceae (Anaplasma, Ehrlichia, and Neorickettsia) reside and multiply in phagosomes, leading to morula-like assemblages of bacteria that are surrounded by a single host-derived membrane (58). Members of “Ca. Midichloriaceae” display different patterns: “Ca. Midichloria mitochondrii” lives in the periplasmic compartment of host mitochondria and is surrounded by the outer mitochondrial membrane (59, 60), whereas other members of “Ca. Midichloriaceae” (“Ca. Anadelfobacter,” “Ca. Cyrtobacter,” and a bacterium from Acanthamoeba strain UWC8) lack a visible third membrane and instead display a translucent halo (49, 53). However, another endosymbiont of Acanthamoeba (strain UWC36), which was recently described as “Ca. Jidaibacter acanthamoeba,” is the sister strain to the endosymbiont of UWC8 and colonizes host-derived vacuoles (61). The discrepancy between the data on the Acanthamoeba endosymbionts (UWC8 and UWC36) suggests that ultrastructural data from chemically fixed material should be interpreted with caution. The cells of “Ca. Lariskella” and the endosymbiont of Eutreptiella, which also belong to “Ca. Midichloriaceae,” are clearly surrounded by one or two additional membranes (62, 63). Considering the phylogenetic position of “Ca. Finniella” (separate from Anaplasmataceae and “Ca. Midichloriaceae”) and the ultrastructural data available for its closest known relatives, “Ca. Odyssella” and “Ca. Paracaedibacter” (37–39), it seems likely that the electron-dense outer layer surrounding “Ca. Finniella” does not represent a host-derived membrane. This assumption is furthermore supported by the fact, that “Ca. Finniella” neither displays a pathogenic life cycle with endocytosis and host cell lysis (as in Anaplasmataceae) nor enters specific intracellular compartments (as in “Ca. Midichloriaceae”). Instead, “Ca. Finniella” is scattered throughout the host cytoplasm and seems to have a well-balanced relationship with the viridiraptorid host cell as observed in other Rickettsiales that live free in the host cytoplasm. However, the varying ultrastructural appearance of closely related rickettsial species (“Ca. Finniella” included) suggests that fixation artifacts may play a role. Therefore, a sound comparison of rickettsial ultrastructure would require data from cryofixed samples, as are available for the endosymbiont MIDORIKO of Carteria (“Ca. Megaira,” Rickettsiaceae) (64).

Rickettsiales are obligate intracellular bacteria that colonize a broad range of eukaryotic hosts, including various arthropod and mammalian cells, leeches, nematodes, and cnidarians (9, 65). Besides metazoan hosts, diverse protists harbor rickettsial endosymbionts as well (5, 9, 65). Previous analyses of molecular sequence data as well as our present results reveal that the rickettsial phylogeny is not congruent with the phylogeny of the host organisms, indicating a frequent lateral transfer of rickettsial endosymbionts between phylogenetically distant hosts (9, 65). For pathogenic representatives such as Rickettsia, Anaplasma, and Ehrlichia, it is well established that these bacteria cycle between hosts on a regular basis (e.g., between an arthropod vector and a mammalian host), whereas most of the protist-colonizing Rickettsiales have been detected in single natural host species only. However, there is some evidence from cross-infection experiments that, e.g., “Ca. Nucleicultrix” can infect and grow in different protist hosts (42). Therefore, some protist-colonizing Rickettsiales might have a broader host spectrum, as suggested by the current yet limited data on phenotypically characterized Rickettsiales.

In light of the current phylogenies, two groups of protists seem to be of particular relevance as hosts for Rickettsiales, namely, ciliates (Alveolata) and Acanthamoeba species (Amoebozoa). Members of the ciliate genera Paramecium and Euplotes are especially well known to serve as hosts for Holospora, Caedibacter, and “Ca. Megaira” strains (51, 66, 67). Acanthamoeba species also harbor rickettsial representatives from diverse phylogenetic groups such as “Ca. Midichloriaceae” (“Ca. Jidaibacter”), “Ca. Paracaedibacteraceae” (“Ca. Paracaedibacter,” “Ca. Odyssella,” and several yet-undescribed endosymbionts) and clade C/N (Caedibacter). Considering the diversity of hosts, protist groups other than ciliates and Acanthamoeba are surprisingly underrepresented. However, this imbalance could be also caused by biased sampling. Acanthamoeba species are very common and abundant in freshwater and soil ecosystems (68), and due to their medical relevance, these amoebae are intensely studied (69). Similarly, ciliates, especially Paramecium species, have long been recognized as hosts for Rickettsiales (for example, see reference 70): the ciliate-colonizing genera Holospora and Caedibacter display complex life cycles and toxic cell inclusions (71, 72).

In fact, more and more reports on novel, phylogenetically diverse protistan hosts of Rickettsiales are accumulating. The amoebozoon Hartmannella was found to be the natural host of “Ca. Nucleicultrix” (clade C/N) (42). Eutreptiella and Petalomonas, two distantly related Euglenozoa (Excavata), harbor midichloriacean and paracaedibacteracean endosymbionts, respectively (41, 62). A Rickettsia species was found in the cytoplasm of the naked, filose amoeba Nuclearia pattersonii (Nucleariidae, Opisthokonta) (73), and representatives of the “Ca. Megaira” clade (Rickettsiaceae) colonize volvocalean green algae such as Carteria, Pleodorina, and Volvox (Chlorophyta, Viridiplantae) (74–76). There is also one report on “Ca. Megaira” in Haplosporidium, an intracellular, osmotrophic parasite of molluscs that belongs to the phylum Cercozoa (Rhizaria) (77). In light of the large diversity of phagotrophic (often bacterivorous) Cercozoa with flagellate or amoeboid morphology (78–81), it is remarkable that the only evidence for a rickettsial endosymbiont in a rhizarian host so far comes from a plasmodial parasite, which lost its ability to feed by phagocytosis and lives inside other eukaryotic cells. Although endosymbiotic bacteria have been occasionally found in phagotrophic Cercozoa such as Cercomonas sp. during ultrastructural studies (82), these endosymbionts still await molecular characterization. Therefore, the “Ca. Finniella”-viridiraptorid relationship represents evidence of a rickettsial member that colonizes free-living, phagotrophic Cercozoa.

Protist-Rickettsiales symbioses can be of different nature. Many Rickettsiales seem to establish a stable coexistence with their protist hosts (e.g., “Ca. Anadelfobacter,” “Ca. Megaira,” “Ca. Cryproprodotis,” “Ca. Cyrtobacter,” and “Ca. Paracaedibacter”). In contrast, members of the Holosporaceae and clade C/N display special life history stages to escape the host cell and to infect new host cells (Holospora spp.), confer killer traits to the host cell (Caedibacter spp.), or consume and finally lyse the host cell (“Ca. Nucleicultrix”). So far, neither special life history stages nor adverse effects on viridiraptorid cells (e.g., uncontrolled bacterial growth or cell death) could be observed in “Ca. Finniella,” suggesting that the “Ca. Finniella”-viridiraptorid symbiosis is a well-balanced relationship. The phylogenetic affiliation of “Ca. Finniella” and the fact that a host-free cultivation was unsuccessful indicate that this candidate genus comprises obligate intracellular bacteria. Since most of the viridiraptorid strains, which are genetically very similar or identical to each other (1), did not contain bacterial endosymbionts, it seems likely that the “Ca. Finniella”-viridiraptorid symbiosis is not obligatory for the viridiraptorid hosts. The endosymbionts undoubtedly use viridiraptorid resources for intracellular growth and multiplication, but the possibility that they also provide benefits to their hosts cannot be excluded. Compared with others, the bacterium-harboring strains VirI02 and VirI06 of Viridiraptor invadens are very aggressive and grow especially fast (S. Hess, unpublished observation). Future experiments involving the establishment of bacterium-free viridiraptorid strains and comparison of their viability and multiplication rates to their bacterium-harboring counterparts could give further insights into the nature of the “Ca. Finniella”-viridiraptorid symbiosis.

The axenic viridiraptorid cultures facilitated the molecular characterization of “Ca. Finniella,” since the amplified bacterial DNA could be directly subjected to sequencing. However, in some cultures of Orciraptor agilis, plastid DNA of the food algae contaminated the samples. Here, fluorescence-actived cell sorting (FACS) turned out to be a time-efficient method to enrich target cells from which DNA could be isolated. Cell sorting may also be a good strategy in the case of protists that depend on external bacteria (e.g., bacterivorous protists), either to try direct sequencing of sorted cells or to increase the success during cloning.

“Ca. Finniella” comprises two genotypes, which colonize closely related host species and for which specific FISH probes are available. Many rickettsial species either lack very close relatives which are characterized and in culture (and are therefore available for probe design and experiments) or have been used only to design multimismatch probes that often target broader phylogenetic assemblages (for example, see references 38, 51, 74, and 83). Therefore, the probes FININ731 and FINLU731 and the “Ca. Finniella” strains in permanent culture constitute an excellent experimental system for studies addressing host specificity.

The species-specific relationships between viridiraptorid genera and the members of “Ca. Finniella” that we know so far could be either explained by inherent host specificity of such Rickettsiales or by spatial separation of viridiraptorid populations, both preventing horizontal transmission of endosymbionts. Indeed, Orciraptor and Viridiraptor were isolated mostly from different habitat types (acidic bog ponds with extremely low conductivity versus mesotrophic/eutrophic freshwater ponds) and feed on different types of algal food (1). Future cross-infection experiments could circumvent these potential spatial barriers and elucidate the host specificity of “Ca. Finniella.” Together with the exploration of yet-unknown “Ca. Finniella” species and well-resolved host phylogenies, such data on host range specificity may provide insight into how these symbioses evolved.