Comparative genomic analysis of symbiotic and free-living Fluviibacter phosphoraccumulans strains provides insights into the evolutionary origins of obligate Euplotes–bacterial endosymbioses

ABSTRACT

Endosymbiosis is a widespread and important phenomenon requiring diverse model systems. Ciliates are a widespread group of protists that often form symbioses with diverse microorganisms. Endosymbioses between the ciliate Euplotes and heritable bacterial symbionts are common in nature, and four essential symbionts were described: Polynucleobacter necessarius, “Candidatus Protistobacter heckmanni,” “Ca. Devosia symbiotica,” and “Ca. Devosia euplotis.” Among them, only the genus Polynucleobacter comprises very close free-living and symbiotic representatives, which makes it an excellent model for investigating symbiont replacements and recent symbioses. In this article, we characterized a novel endosymbiont inhabiting the cytoplasm of Euplotes octocarinatus and found that it is a close relative of the free-living bacterium Fluviibacter phosphoraccumulans (Betaproteobacteria and Rhodocyclales). We present the complete genome sequence and annotation of the symbiotic Fluviibacter. Comparative analyses indicate that the genome of symbiotic Fluviibacter is small in size and rich in pseudogenes when compared with free-living strains, which seems to fit the prediction for recently established endosymbionts undergoing genome erosion. Further comparative analysis revealed reduced metabolic capacities in symbiotic Fluviibacter, which implies that the symbiont relies on the host Euplotes for carbon sources, organic nitrogen and sulfur, and some cofactors. We also estimated substitution rates between symbiotic and free-living Fluviibacter pairs for 233 genes; the results showed that symbiotic Fluviibacter displays higher dN/dS mean value than free-living relatives, which suggested that genetic drift is the main driving force behind molecular evolution in endosymbionts.

IMPORTANCE

In the long history of symbiosis research, most studies focused mainly on organelles or bacteria within multicellular hosts. The single-celled protists receive little attention despite harboring an immense diversity of symbiotic associations with bacteria and archaea. One subgroup of the ciliate Euplotes species is strictly dependent on essential symbionts for survival and has emerged as a valuable model for understanding symbiont replacements and recent symbioses. However, almost all of our knowledge about the evolution and functions of Euplotes symbioses comes from the Euplotes–Polynucleobacter system. In this article, we report a novel essential symbiont, which also has very close free-living relatives. Genome analysis indicated that it is a recently established endosymbiont undergoing genome erosion and relies on the Euplotes host for many essential molecules. Our results provide support for the notion that essential symbionts of the ciliate Euplotes evolve from free-living progenitors in the natural water environment.

INTRODUCTION

Endosymbiosis, defined as a close relationship between two organisms belonging to different species, one of which (the endosymbiont) lives inside the other (the host), is a ubiquitous and important mechanism in ecology and evolution (1–3). Ciliates are a diverse group of single-celled eukaryotes that occur in almost all aquatic environments. They are important grazers of algae, bacteria, and other microorganisms and are very important components of the microbial food web. Ecological and trophic preferences of ciliates seem to favor the formation of diverse symbiotic associations with bacteria, archaea, and algae (4, 5). The ability to gain and retain symbionts varies among different ciliate groups; Euplotes is certainly one of the most studied groups for its proneness to establish symbiotic relationships with different bacteria (6). The speciose genus Euplotes has been emerging as a model system for the investigation of prokaryotic symbioses over recent years.

Bacterial symbionts of Euplotes were first reported by Fauré-Fremiet in the cytoplasm of Euplotes patella and Euplotes eurystomus in 1952 (7). Since then, an increasing number of endosymbionts were described in Euplotes, with currently at least 17 genera and 22 species (6, 8, 9). According to the degree of dependence, symbionts harbored by Euplotes can be divided into essential symbionts and accessory symbionts (6). Essential symbionts are indispensable for host survival and reproduction (10–12), but accessory symbionts are probably not required for host survival since they always cooccur with known essential symbionts and usually belong to groups of specialized intracellular bacteria (9, 13–16).

Thus far, all Euplotes species in the “clade B” group (17) harbor essential symbionts (10–12). A similar close relationship was less observed in other Euplotes species. The sole report was the association between Euplotes magnicirratus (clade A) and its essential symbiont “Candidatus Devosia euplotis” (18). In Euplotes, the most common essential symbiont is Polynucleobacter necessarius (Betaproteobacteria and Burkholderiales) (10, 11, 19). If P. necessarius is absent, other less common essential symbionts, members of the genera “Ca. Protistobacter” (20) or Devosia (18, 21), can take its place. Further insights into the symbiotic relationship between Euplotes and essential symbionts came from detailed genomic analyses (22–24). Phylogenomic analyses of nine symbiotic and seven free-living Polynucleobacter showed at least eight independent origins of symbiosis among the nine symbionts, which suggested that the Euplotes–Polynucleobacter symbioses relationship originated multiple times independently (22). Recently, the genomes of different Ca. Protistobacter heckmanni and Ca. Devosia strains were characterized, and all were found to be large and enriched in pseudogenes and repetitive elements (23). The phylogenomic tree showed that Euplotes and symbionts in strains of Protistobacter-harboring Euplotes phylogenies are incongruent. Hence, just like Polynucleobacter, extant Ca. Protistobacter have not coevolved with Euplotes and are the descendants of independently established symbioses. In addition, single-cell microbiome suggested that Polynucleobacter and “Ca. Protistobacter” coexist in one Euplotes cell, which might represent an ongoing replacement event in nature (25). Overall, all known essential Euplotes symbionts are recently established and continuously replaced, making the mutually obligate symbiosis ancient for the host but not for any known essential symbionts (22, 23).

Closely related symbiotic and free-living pairs of organisms are valuable since comparing symbionts with their free-living relatives is a powerful approach to investigating the evolution of symbioses. Among the known essential symbionts of Euplotes, only the genus Polynucleobacter comprises close free-living and symbiotic representatives (19, 26–28). Thus, the Euplotes–Polynucleobacter symbiosis has been regarded as a promising model for bacteria–eukaryote endosymbioses and has provided some important insights into the endosymbiosis process (22, 24). Recently, another pair of closely related symbiotic/free-living bacteria, the bacteria “Ca. Nebulobacter yamunensis” and Fastidiosibacter lacustris, which are, respectively, known as an accessory symbiont of Euplotes aediculatus (14) and a free-living bacterium (29), were reported (30). Comparative genomic analysis revealed that the genome of Ca. Nebulobacter yamunensis is almost indistinguishable from that of the free-living strain. Ca. Nebulobacter and Fastidiosibacter may represent an extreme example, proving that a small number of factors might play a key role in the earliest stages of endosymbiosis establishment (30). Therefore, the detection of new close symbiont/free-living pairs may offer the chance to discover novel insights into the endosymbiosis process.

In the current study, we describe the discovery of a novel putatively essential symbiont inhabiting the cytoplasm of Euplotes octocarinatus, a clade B species that usually depends on Polynucleobacter or Ca. Protistobacter for survival. Phylogenetic analyses show that it is a close relative of the free-living bacterium Fluviibacter phosphoraccumulans (Betaproteobacteria and Rhodocyclales). We provide the complete genome sequence of the symbiotic Fluviibacter and present a comparative analysis of the genome sequences of symbiotic Fluviibacter and related free-living strains, addressing the possible biological basis of the Euplotes–Fluviibacter symbiosis. The discovery of the novel Euplotes–Fluviibacter symbiosis also provides support for the notion that essential symbionts of the ciliate Euplotes evolve from free-living progenitors.

RESULTS

Characterization of a novel betaproteobacterial symbiont from E. octocarinatus strains VTN7 and VTN8

The genus Euplotes, one of the most species-rich genera of ciliates, has been divided into five major clades (clades A–E; Fig. S1) (17). So far, all Euplotes species in the “clade B” group harbor essential symbionts (6). A preliminary fluorescence in situ hybridization (FISH) experiments using the probe Poly_862 that targets both Polynucleobacter and “Ca. Protistobacter” showed no positive signal in the cytoplasm of E. octocarinatus strains VTN7 and VTN8 (Fig. S2). As a positive control, a clearly positive signal can be observed in the cytoplasm of E. octocarinatus strain Zam5b-1, which has been reported for containing Polynucleobacter (9). The further metagenomic screening revealed three complete 16S rRNA gene sequences from both VTN7 and VTN8 libraries (Table S1). Blastn analysis showed that two of them affiliated to previously reported alphaproteobacterial accessory symbionts: one belonging to the genus Ca. Anadelfobacter (98.2% identity to Ca. Anadelfobacter veles, accession number: FN552695), and the other one affiliated to Ca. Megaira polyxenophila (99.0% sequence identity). Interestingly, the third one shares a high sequence identity (>99%) with a free-living bacterium F. phosphoraccumulans (Betaproteobacteria and Rhodocyclales; accession number: AP019011) (31).

The presence of the novel betaproteobacterial symbiont in the cytoplasm of Euplotes cells was confirmed using the species-specific oligonucleotide probe Fluvii_458 (Fig. 1). Most of the bacteria cells appeared as straight rods, but some are slightly curved. Cells measure about 0.4–0.5 × 1–3 µm. All attempts to grow the symbionts outside its host in solid MR2A failed at both 22°C and 27°C. However, a previous study reported that this culture medium is suitable for the free-living F. phosphoraccumulans (31).

Fig 1.

Detection of F. phosphoraccumulans in E. octocarinatus strain VTN7 and VTN8 using FISH. Microphotographs of fixed cells of E. octocarinatus after 4',6-diamidino-2-phenylindole (DAPI) staining (A and D), FISH using Cy3-labeled Fluvii_458 probe (B and E), and overlay of DAPI and FISH (C and F). DAPI staining indicated the typically reverse C-shaped macronucleus of the host E. octocarinatus. Gray outlines represent Euplotes cells and were drawn based on the corresponding bright field pictures. Bars represent 10 µm.

Complete sequences of the novel betaproteobacterial bacteria 16S rRNA gene from both EoVTN7 (accession number: OR016159) and EoVTN8 (accession number: OR018999) strains were obtained through PCR amplification and sequencing. The sequences from both strains differ by four nucleotide substitutions. Sequence analysis confirmed that these bacteria belong to the class Betaproteobacteria and order Rhodocyclales. The sequences had 99.53% identity to an uncultured proteobacterium (accession number: DQ450169) (32) and 99.07% identity to F. phosphoraccumulans (31). Phylogenetic trees, constructed with the Maximum Likelihood (ML) and Bayesian Inference (BI) methods, produced in general similar topologies (Fig. 2). Both methods assigned this newly detected betaproteobacterial symbiont to the recently established family “Fluviibacteraceae” with high confidence level (31). An identity value of 98.7% has been proposed as a threshold for species level (33). High similarity values (>99%) of the 16S rRNA gene sequences obtained from the symbionts in the Euplotes culture and free-living F. phosphoraccumulans indicate that these organisms belong to the same species. Considering the distinct differences in lifestyle, we propose to establish the novel subspecies “F. phosphoraccumulans subsp. symbioticus” for this novel endosymbiont.

Fig 2.

ML phylogenetic tree based on 16S rRNA genes showing the phylogenetic position of the bacterium detected in E. octocarinatus. Since BI approaches resulted in similar topologies, only ML is presented. The newly characterized sequences in this study are shown in bold. Species names, classification, and accession numbers of the employed sequences are shown. The numbers at the nodes represent the support values of ML/BI (values below 50/0.60 are not shown). Fully supported (100/1.00) branches are marked with solid circles. All branches are drawn to scale. The scale bar corresponds to 10 substitutions in 100 nucleotides.

Bacterial genome with features of obligate symbiosis

Comparing obligate endosymbionts with their free-living relatives is a common approach to investigating the evolution of symbioses. To further investigate the genomic traits associated with the establishment of symbiosis, we try to retrieve the complete genome of the symbiotic Fluviibacter from the metagenome assembly. The initial assembly of the Illumina NovaSeq paired-end reads from the strain EoVTN8 resulted in nine scaffolds (Fig. S3A). Manual inspection found two identical insertion sequences (762 bp) in Scaffold 2 and Scaffold 4 (Fig. S3B). Further analysis found that this insertion sequence is present at both ends of all nine scaffolds. Hence, the fragmentation of the assembly was probably caused by repetitive sequences derived from this insertion sequence. To obtain a complete closed genome, specific primers were designed according to the sequences flanking the insertion sequence (Table S2). The order and direction of the 11 fragments were determined through PCR amplifications and direct sequencing (Fig. S3C). Finally, we obtained the circular genome of the symbiotic Fluviibacter (Fig. 3A). The genome sequence of the symbiotic Fluviibacter shows a typical pattern of polarized nucleotide composition: an excess of G over C in the leading strand and an excess of C over G in the lagging strand (G + C skew, Fig. 3A).

Fig 3.

Comparative analyses of the symbiotic and free-living F. phosphoraccumulans genomes. (A) Features of the genome of “F. phosphoraccumulans subsp. symbiotica.” Circular genome plot of “F. phosphoraccumulans subsp. symbiotica,” showing (from inside to outside) guanine/cytosine (GC) skew (light green, positive; deep sky blue, negative), positions of insertion sequence elements (red bars), GC content, pseudogenes (black) and protein-coding genes (light blue) encoded on forward or reverse strand of the genome (in kb). (B) Protein-coding genes shared by symbiotic Fluviibacter and free-living strains. (C) Comparative analyses of symbiotic and free-living Fluviibacter genomes performed with Mauve software.

The genome of the symbiotic Fluviibacter consists of only one circular chromosome of 1,672,281 bases with an average GC content of 53.3% (Table 1). The plasmid observed in all three free-living strains (31) has been lost in the symbiont. There are a total of 1,380 protein-coding genes, 2 rRNA operons (16S, 23S, and 5S), and 43 transfer RNAs that decode all 20 standard amino acids encoded on the chromosome (Table 1). An apparently universal signature of the genomes of symbiotic bacteria is a reduction in genome size and gene number, which has also been previously reported in symbiotic P. necessarius genomes (22, 24). The symbiotic Fluviibacter had a smaller genome with reduced gene content compared with free-living strains (Table 1). The reduction in genome size is apparently caused by a decrease in the amount of coding DNA because of the huge number of pseudogenes in the symbiont. Of the 1,380 protein-coding genes, 1,219 genes are shared with all three free-living strains (Fig. 3B). Only 147 predicted genes in the symbiont are not shared with any free-living strains, but they are shorter than those shared with free-living strains (average: 545 bp vs 943 bp), and half of which code for hypothetical proteins with no assigned function. Thus, it seems that the gene inventory of the symbiotic Fluviibacter is largely a subset of that of those free-living relatives, indicating that symbiotic Fluviibacter is a bridged derivative of a strain free-living Fluviibacter-like ancestor.

TABLE 1.

Genomic features

Parameter	Symbiotic strain	Free-living strains
	EoVTN8	SHINM1	ICHIJ1	ICHIAU1
No. of contigs (chromosome, plasmid)	1, 0	1, 1	1, 1	1, 1
GC content of chromosome (%)	53.3	54.3	54.2	54.2
Genome size of chromosome (bp)	1,672,281	2,295,374	2,431,578	2,392,860
Coding size (bp)	1,264,381	2,136,761	2,262,419	2,229,187
Protein-coding genes	1,380	2,227	2,325	2,296
Pseudogenes	297	38	69	59
rRNA genes	6	6	6	6
tRNAs	43	46	45	45
Genome size of plasmid (bp)	/	9,965	16,356	16,356
Protein-coding genes	/	13	19	19
Coding size (bp)	/	8,319	14,190	14,190

Extensive expansions of insertion sequence elements and pseudogenes have been documented in a number of recently established symbionts (22, 34, 35), implying that they are a signature of the first stages of genome erosion (35, 36). A total of 297 pseudogenes are identified from the symbiotic Fluviibacter genome, of which 179 (60.3%) are predicted to be fragmented genes based on the presence of frameshift mutations or premature stop codons (Table S3). Genes considerably shorter than their top homologs are also relatively common (37%) among the pseudogenes. However, in the case of symbiotic Fluviibacter, there are only 11 identical insertion sequence elements, even less than free-living strains (Table S4). Furthermore, this insertion sequence is symbiont specific; none of the three free-living Fluviibacter genomes contain homologs of it. It is possible that the symbiotic Fluviibacter acquired this element independently. Classical insertion sequences are formed by a transposase coding region and short terminal inverted repeats (IRs). The insertion sequence discovered in symbiotic Fluviibacter is 762 bp long, flanked by 14 bp IRs (Fig. S3B). The transposase coding region contains two overlapped open reading frames (ORFs) in the relative translational reading frames 0 and +1, respectively. The first ORF encodes a 154-amino acid sequence containing part of the ISXO2-like transposase domain. The second ORF encodes a 98-amino acid sequence that includes the remainder of the ISXO2-like transposase domain. The fact that the two ORF products compose the complete ISXO2-like transposase domain implies that this gene undergoes a +1 programmed ribosomal frameshifting. However, the first ORF is disrupted by an in-frame stop codon TAA, which suggests that the transposase gene in the insertion sequence of symbiotic Fluviibacter is pseudogenized.

Whole-genome alignments performed with Mauve software revealed a high degree of genome-wide synteny with just a limited number of rearrangements between symbiotic Fluviibacter and free-living strains (Fig. 3C). It is notable that almost all insertion sequences are located in the bound of the rearranged regions in symbiotic Fluviibacter, which implies that the insertion sequences might participate in driving intragenomic rearrangements as reported in other endosymbionts and obligatory intracellular pathogens (35, 37).

Metabolic capacities comparison between symbiotic and free-living Fluviibacter

According to gene annotation, the metabolic capabilities of symbiotic Fluviibacter were reconstructed and compared with free-living strains. Compared to the free-living strains, the symbiotic Fluviibacter strain displays reduced metabolic capabilities (Fig. 4). More detailed information is listed in Table S5.

Fig 4.

Comparison of metabolic potential between the symbiotic and free-living Fluviibacter. Heatmaps of the predicted Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway completion separated by function and produced with ComplexHeatmap. High to low completeness is colored dark blue to light green. Free-living and symbiotic Fluviibacter strains are indicated in the legend with blue and orange colors, respectively. Pathways of interest are highlighted in red.

Concerning the central metabolism, free-living Fluviibacter possesses a virtually intact tricarboxylic acid cycle, glyoxylate cycle, gluconeogenesis, glycolytic pathway, and the nonoxidative part of the pentose phosphate pathway. While the symbiont possesses almost all of the aforementioned enzymatic paths, key enzymes for glycolysis and the glyoxylate cycle appeared to be absent. The phosphofructokinase gene and pyruvate kinase gene involved in the glycolytic pathway are pseudogenized in the symbiotic Fluviibacter, so they cannot exploit sugars as carbon or energy sources. In addition, the symbiotic Fluviibacter has lost the malate synthase gene specific to the glyoxylate cycle and thus cannot use acetyl-CoA for gluconeogenesis and for the generation of tricarboxylic acid cycle intermediates. Thus, the symbiotic Fluviibacter probably depends on its host for various carbon compounds.

The symbiotic Fluviibacter appears to retain complete pathways for energy metabolism, as well as most biosynthetic capabilities. As for the energy metabolism, both free-living and symbiotic Fluviibacter retained the ability to independently produce ATP through an electron transport chain. Interestingly, they do not possess the most widespread cytochrome c oxidase complex, replaced by the cytochrome bd-type oxidase and the cbb3-type cytochrome c oxidase, which are active under very low concentrations of oxygen. Furthermore, all genomes have two genes encoding for polyphosphate kinases (ppk1 and ppk2), which are related to the intracellular accumulation of polyphosphate. Regarding biosynthetic capabilities, the genome of the symbiotic Fluviibacter encoded complete metabolic pathways for the biosynthesis of all 20 L-amino acids, purines, pyrimidines, fatty acids, glycerophospholipids (except phosphatidyl-inositol and phosphatidylcholine), and cell envelope components (peptidoglycan and lipopolysaccharides). As for the biosynthesis of cofactors and vitamins, all invested Fluviibacter strains have the complete metabolic pathways for the biosynthesis of heme, riboflavin, pantothenate, pyridoxine, NAD(P), folate, ubiquinone, lipoic acid, and coenzyme A. However, genes for biotin, thiamine, and coenzyme B12 production were found only in free-living Fluviibacter, and the symbiont has lost these abilities. Like carbon, these cofactors and vitamins may also be provided by its host.

Besides, the symbiotic Fluviibacter genome exhibits extensive loss of the genes involved in nitrogen and sulfur metabolism. The free-living Fluviibacter can assimilate nitrate and sulfate from the environment, but the symbiont has lost this ability. The free-living strains can perform the reduction of nitrate, whereas the pathway is absent in the symbiont. Therefore, it is plausible that nitrogen and sulfur are acquired from the host as part of various compounds.

Another apparently universal signature of obligate endosymbionts with reduced genomes is the loss of DNA repair and recombination machinery. DNA repair systems are relatively underrepresented in both free-living and symbiotic Fluviibacter. Although base and nucleotide excision repair pathways are present and largely intact even in symbiotic Fluviibacter, the enzymes for the mismatch repair system are missing (Table S5). The homologous recombination pathway is present despite the recFOR pathway lacking the gene recF, similar to what is found in Polynucleobacter (22, 24). Of the DNA polymerases to perform translesion replication, only the DNA polymerase V is present in free-living strains and none in the symbiont.

Secretion systems are widely present in many obligate intracellular symbionts or pathogens, where they are postulated to be invoked as essential tools for host interactions. All investigated Fluviibacter genomes have conserved secretory (Sec) and twin-arginine translocation (Tat) secretion systems for the translocation of proteins to the periplasmic space. However, sets of genes coding for type IV secretion systems could be found only in free-living Fluviibacter and do not share an ortholog in the symbiotic strain’s genome (Fig. S4A). Interestingly, the best BLASTP hits of them are the orthologs that belong to Burkholderiales rather than Rhodocyclales. Furthermore, type IV pili biosynthetic genes were retrieved in all analyzed genomes (Fig. S4B). Free-living strains appear to harbor precisely one functional pil operon. Corresponding homologous genes could be found in the symbiotic Fluviibacter genome, but most of them are pseudogene (Fig. S4B). Besides this, the symbiotic Fluviibacter also possesses another pil operon and one tad operon. Both operons are incomplete and show high sequence identity with Burkholderiales. Again, almost all of them are pseudogene. Thus, we speculated that the symbiotic Fluviibacter cannot express a complete pili apparatus.

Comparison of the nucleotide substitution rates between free-living and symbiotic Fluviibacter

Compared with free-living relatives, endosymbionts usually exhibit higher rates of amino acid sequence evolution (38). Many studies in diverse symbiont systems suggested the importance of genetic drift in the evolution of obligate symbionts (22, 39, 40). To estimate the substitution rates in the Fluviibacter-Euplotes system, an environmental metagenome-assembled genome (GCA_010024635.1), previously characterized as unclassified Rhodocyclales (41), was introduced in this analysis. Although this metagenome-assembled genome lacks 16S rRNA sequence, core gene clusters were retained for phylogenetic placement. A phylogenomic analysis of 303 concatenated orthologous genes showed that the symbiotic Fluviibacter is most closely related to this unclassified Rhodocyclales bacterium (Fig. 5A). Both free-living and symbiotic Fluviibacter formed an apparently distinct lineage within the order Rhodocyclales, in accordance with the 16S rRNA gene-based phylogenetic analysis.

Fig 5.

Phylogenomic analysis and pairwise comparison of dN/dS values. (A) ML tree of 303 orthologous F. phosphoraccumulans and other Rhodocyclaceae genomes using RAxML. The genome of P. necessarius and Mycetohabitans rhizoxinica was used as an outgroup. The scale bar indicates a 10% estimated sequence divergence. (B) dN/dS values for 233 orthologous genes in the symbiotic strain EoVTN8 and the free-living strain Rho. The symbiotic strain displayed a significantly higher dN/dS mean value than its free-living relative (averages: 0.0527 vs 0.0437; P value = 0.001, Mann-Whitney U test). The blue circle represents the average position of 20 gene-windows projected onto the diagonal. The inset depicts the approach used to calculate the dN/dS values, comparing the same gene for each strain of the pair with the orthologous gene in the outgroup (strain SHINM1).

Knowing with confidence the phylogenetic relationships between the strains, substitution rates were estimated for 233 genes (Table S6). We performed pairwise comparison between symbiotic strain EoVTN8 and free-living strain SHINM1, as well as pairwise comparison between free-living strain Rho and strain SHINM1 as control (Fig. 5B). In Fluviibacter, the dS values showed no statistically significant difference between symbiotic strain and its free-living relative, although mean value for the symbiont was lower than the free-living strain (averages: 0.9995 vs 1.0429, P value = 0.312, Mann-Whitney U test). In contrast, symbiotic Fluviibacter displays a significantly higher dN/dS mean value than its free-living relative (averages: 0.0527 vs 0.0437; P value = 0.001, Mann-Whitney U test; Fig. 5B). This result demonstrates that protein-coding genes are evolving at a higher rate in symbiotic Fluviibacter relative to free-living strain. Moreover, it also confirmed that genetic drift is the main mechanism responsible for genomic evolution in symbiotic Fluviibacter.

DISCUSSION

Euplotes symbioses have emerged as an excellent model for understanding symbioses in unicellular eukaryotes and identifying universal features of recent symbioses (4, 6, 22). So far, a total of four different essential Euplotes symbionts have been described. P. necessarius is the most common essential symbiont of all Euplotes species in the “clade B” (6). If P. necessarius is not present, members of the genera Devosia and Ca. Protistobacter can take its place (20, 21). Here, we characterized a novel endosymbiont, “F. phosphoraccumulans subsp. symbioticus,” in strains VTN7 and VTN8 of E. octocarinatus, a species belonging to clade B that usually harbors Polynucleobacter (6, 9) or Ca. Protistobacter (20). This bacterium could not be grown outside its host with methods used for free-living Fluviibacter species (31). The metabolic profile also provides a clear explanation for the inability to grow symbiotic Fluviibacter outside its host (Table S5). The symbiont depends on the host Euplotes for many essential molecules, such as carbon sources, organic nitrogen and sulfur, and some cofactors. Therefore, F. phosphoraccumulans subsp. symbioticus probably represents an obligate symbiont, unable to grow outside the cytoplasm of its host. We were unable to test the degree of dependence of the Euplotes host on the bacterium because we have not found a Euplotes that solely harbor the F. phosphoraccumulans subsp. symbioticus. It is, however, known that E. octocarinatus as a species depends on essential symbionts for survival (6). All previously investigated strains harbored the betaproteobacterium Polynucleobacter or Ca. Protistobacter (6, 20). Moreover, the other two identified endosymbionts in the host E. octocarinatus strains VTN7 and VTN8 were all documented accessory symbionts (13, 16). Therefore, it is reasonable to regard the F. phosphoraccumulans subsp. symbioticus as a putatively essential symbiont. It is likely that the role of essential symbiont played by Polynucleobacter or Ca. Protistobacter in other E. octocarinatus strains is replaced by F. phosphoraccumulans in strains VTN7 and VTN8.

The betaproteobacterial order Rhodocyclales is an abundant bacterial order in wastewater treatment systems (42), and almost all species in this order are free living. The sole obligate endosymbiont reported so far is Ca. Dactylopiibacterium carminicum (Rhodocyclales and Rhodocyclaceae), a putative nitrogen-fixing symbiont from the carmine cochineal insects (43). Moreover, digestion-resistant bacteria belonging to the family Rhodocyclaceae have also been detected in the cytoplasm of the soil ciliate Metopus yantaiensis (44). Free-living F. phosphoraccumulans strains were reported recently as novel polyphosphate-accumulating bacteria from surface river water (Saitama Prefecture, Japan) (31). Phylogenetic analyses revealed that the strains formed a distinct phylogenetic lineage within the order Rhodocyclales. The endosymbiont reported here is closely related to the free-living Fluviibacter (Fig. 2) but is distinctly different in morphology (straight or slightly curved rods vs short rod or coccoids), cell size (0.4–0.5 × 1–3 µm vs 0.6–0.8 × 0.8–1.5 µm), and genome size (1.67 Mb vs 2.3–2.4 Mb). Furthermore, pure cultures of free-living Fluviibacter strains were established; however, the symbiotic Fluviibacter was an obligate intracellular symbiont. These differences indicated that the symbiotic and the free-living Fluviibacter represent strains fundamentally differing in lifestyle, rather than different stages of a facultative endosymbiotic lifestyle. It seems that a free-living ancestor of the symbiotic Fluviibacter successfully invaded a Euplotes cell and gradually adapted to the life strategy of an obligate endosymbiont. On the other hand, the identification of the symbiotic Fluviibacter from E. octocarinatus supported the hypothesis that Euplotes species constantly recruited potential symbionts from an available pool of natural environment (23).

Of the four previously described essential symbionts, only the species P. necessarius contains both symbiotic and free-living strains. Thus, most of our knowledge about the evolution and functions of Euplotes symbioses comes from the study of the Polynucleobacter-Euplotes system (12, 21, 22, 24). The phylogenetic analysis showed that the species F. phosphoraccumulans also contains both symbiotic and free-living strains, which provide another valuable system for studying the evolution of bacterial symbiosis in Euplotes. The universal signature of obligate symbionts is genome erosion (45). Ancient obligate symbionts possess small, compact genomes, while recently established symbionts have relatively large genomes, usually rich in pseudogenes and mobile genetic elements (22, 46, 47). In the current study, we show that the symbiotic Fluviibacter is a recently established endosymbiont as evidenced by the presence of a comparable number of pseudogenes in its genome. However, with only one available strain, the phylogeny of symbiotic Fluviibacter cannot be assessed. Overall, we conclude that the symbiotic Fluviibacter, like Polynucleobacter and other essential Euplotes symbionts, is the descendant of recently established symbiosis.

In a comparative sense, it is interesting to note that the symbiotic Fluviibacter shares some common features with the symbiotic Polynucleobacter. In terms of cell morphology, both Fluviibacter and Polynucleobacter seem to become longer in the process of transmission from free living to symbiosis (31, 48). In addition, as previously reported in Polynucleobacter, the genome of the symbiotic Fluviibacter is enriched in pseudogenes but contains very few insertion sequence elements (22, 24). Furthermore, multiple comparisons of Polynucleobacter genomes showed that symbiosis in this system has led to the accumulation of substitutions via enhanced genetic drift (22). In the Fluviibacter-Euplotes system, the symbiotic Fluviibacter also displays higher dN/dS values than the free-living relatives (Fig. 5). To sum up, it seems that these geographically isolated and genetically distinct species developed similar characteristics because they were adjusting to similar intracellular environments, which strongly supports the hypotheses of “essential Euplotes symbionts could have undergone convergent evolution.”

The metabolic capacities comparison between symbiotic and free-living Fluviibacter provides some clues for the physiological bases of the symbiont relies on the host Euplotes. Although the symbiont can perform its own basic anabolic processes and energy production, it needs the ciliate host to provide at least carbon sources, some cofactors and vitamins, and organic nitrogen and sulfur. Furthermore, reduced gene sets for sensing and stress resistance provide other probable reasons for the inability to grow symbiotic Fluviibacter outside the host (Fig. 4). These conditions are similar to that of P. necessarius (22, 24). By contrast, it is more difficult to understand why Euplotes depends on symbionts for survival. Nutritional supplementation is a common mechanism in many obligate symbionts of eukaryotes (49), but it is unlikely to be a particularly important driver of symbiosis in Euplotes, as they are an omnivorous predator. More likely, these Euplotes species have lost a universal and essential metabolic pathway conserved in both bacteria and eukaryotes. The essential symbionts might compensate for that loss, changing a lethal mutation into a neutral one. It thereby locks the host in a relationship with any symbiont that can provide the lost function (23, 24).

MATERIALS AND METHODS

Cell cultures and DNA isolation

E. octocarinatus strains VTN7 and VTN8 were collected in 2010 from Vietnam (50). Both strains were established from a single isolated cell and cultured in spring water at 22°C with the flagellate Chlorogonium elongatum as a food source. Euplotes cells (about 5 × 10⁵ cells) were starved for 1 week and then treated with 0.2 mg/mL chloramphenicol overnight to decrease bacterial contamination in the culture medium (22). The cells were concentrated (2,000 g and 5 min), and the DNA from both the host and the symbionts was extracted, using the MiniBEST Bacteria Genomic DNA Extraction Kit (TaKaRa, Japan) according to the manufacturer’s instructions.

Metagenome sequencing and screening

For metagenomic sequencing, DNA libraries were constructed as recommended by the Illumina TruSeq Nano DNA LT Library Preparation Kit (Illumina, USA). Sequencing-by-synthesis was performed on the Illumina NovaSeq platform with the 2 × 150 bp read mode. As for EoVTN8, the DNA library was additionally sequenced on the same platform with the 2 × 250 bp read mode. All raw reads were filtered by fastq_quality_filter (from the FASTX-Toolkit) with the parameters -q 20 p 80 and assembled using SPAdes v.3.15.3 (51) with default settings. To identify the 16S rRNA gene sequences of putative symbionts, the 16S rRNA gene sequence of the P. necessarius (LT606951) was searched against the assembly using BLASTN (E-value ≤ 0.01). All matched contigs were then extracted and BLASTN searched against the NCBI nucleotide collection (nr/nt) database. Only fully assembled 16S rRNA was considered.

Molecular characterization of the betaproteobacterial symbiont

The 16S rRNA sequences of the novel betaproteobacterial symbionts were obtained by PCR using extracted DNA. The forward primer Fluvii-F (5′-GCTCTTTCGGCTGGGAAGAAAT-3′) and the reverse primer Fluvii-R (5′-CCATACAGAGTATTAGCCTGTGCGA-3′) were used in two PCR amplifications, the first one with primers Fluvii-F and 1492R (5′-GGTTACCTTGTTACGACTT-3′; annealing temperature: 48°C and 30 cycles); the second one with primers 16S–8F (5′-AGAGTTTGATCCTGGCTCAG-3′) and Fluvii-R (annealing temperature: 51°C and 30 cycles). The amplified and purified PCR products were cloned in the pMD 18T Vector (TaKaRa) and sequenced with the BcaBEST Sequencing Primers RV-M.

For the detection of the novel betaproteobacterial symbiont, a new species-specific probe Fluvii_458 (5′-CATACAGAGTATTAGCCTG-3′) was designed based on the 16S rRNA sequence. The specificity of the new probe was evaluated in silico on both the SILVA (52) and RDP (53) databases. In silico, probe Fluvii_458 matched one and five sequences in the SILVA and RDP databases, respectively. All hits belong to the unclassified Rhodocyclales bacterium (EF019489, EF019494, EF176778, HQ190339, and HQ190430). Furthermore, probe Poly_862 (5′-GGCTGACTTCACGCGTTA-3′), targeting Polynucleobacter and Candidatus Protistobacter (11), was also used to verify the absence of these two symbionts. FISH assays were performed as described in previous studies (9, 54). The optimum condition for probe Fluvii_458 was assessed at 30% (vol/vol) of formamide concentration. As for probe Poly_862, the optimum experimental condition recommended by the author was used (11). DAPI was added to each slide to visualize the Euplotes nuclei. Negative controls (cells of Euplotes species harboring different kinds of essential symbionts not targeted by the employed probes) were also included. Microscopic observation was performed with a Delta Vision Elite deconvolution microscope system (Applied Precision/GE Healthcare, USA). FISH was repeatedly performed for about 4 years to check the presence of endosymbionts after metagenome sequencing.

Culture attempts of the betaproteobacterial symbiont

Starved E. octocarinatus cells were concentrated and treated with chloramphenicol (0.2 mg/mL) to remove external contaminants and then washed and homogenized. The homogenate was used as inoculum for the culture experiments. Aliquots (100 µL) of the homogenate were spread on modified Reasoner’s 2A (MR2A) agar plates and incubated at 22°C or 27°C for 1 week to inspect the growth. The MR2A agar plates contained 0.5 g/L yeast extract, 0.5 g/L proteose peptone, 0.5 g/L casamino acids, 0.3 g/L sodium pyruvate, 0.3 g/L K₂HPO₄, 0.05 g/L MgSO₄·7H₂O, and 15 g/L agar.

Phylogenetic analyses

Forty-two small subunit (SSU) rRNA gene sequences from the species of the order Rhodocyclales, plus three outgroup sequences from other essential symbionts of E. octocarinatus, were used in the phylogenetic analysis. Sequences were aligned with ClustalW (55) and trimmed to the same length at both ends. The character matrix included 45 sequences and 1,358 sites. jModelTest_2.1.7 (56) was used for the selection of the best evolutionary model for each phylogenetic analysis. An ML tree was constructed based on the GTR + G + I model with IQ-TREE v.1.6.12 software (57). The reliability of internal branches was assessed, using the nonparametric bootstrap method with 1,000 replicates. BI analyses were performed using MrBayes v.3.2.1 (58), applying the selected nucleotide substitution model, namely, GTR + G + I. Three different Markov Chain Monte Carlo runs with one cold and three heated chains were performed, running for 5,000,000 generations with a burn-in of 25%.

Genome assembly and annotation

Preliminary genome assembly was generated from metagenomic data (2 × 150 bp) using the software SPAdes v.3.15.3 (51) with default settings. The macronucleus of the host E. octocarinatus contains abundant gene-sized DNA molecules, each of which has telomeric repeats 5’-(C₄A₄)n-3’ at both ends. Therefore, all telomereless contigs were extracted and searched against the NCBI nonredundant protein sequence database using BLASTX (E-value ≤ 1e⁻⁵). Contigs assigned to betaproteobacteria were collected and further processed. All quality-trimmed reads were remapped to the collected contigs using Blat v.35 (-minIdentity = 99) (59). Mapped reads were then used together to generate the final assembly with SPAdes v.3.15.3. The final assembly was not circular, including nine scaffolds with the same insertion sequence at both ends. A total of 11 pairs of specific primers were designed according to the sequences flanking the insertion sequence. The order and direction of all scaffolds were determined by PCR using extracted DNA and direct sequencing of the amplicon. Finally, the genome of the symbiotic Fluviibacter from E. octocarinatus strain VTN8 was assembled manually into a circular chromosome.

Preliminary gene annotation was performed with RAST (60). To improve comparability, the previously sequenced genomes of the free-living strains (31, 61) SHINM1 (accession number: AP019011), ICHIJ1 (accession number: AP022347), and ICHIAU1 (accession number: AP022345) were annotated again with the same software. Detection of pseudogenes was performed as reported previously (22). A gene was considered a pseudogene if its longest open reading frame was less than 80% or more than 125% of the best-hit homolog. Insertion sequences were identified with ISfinder (62). A bidirectional best-hit blast method against prokaryotic references was performed on the annotated genomes using the KEGG Automatic Annotation Server (63) to predict the putative functions and metabolic pathways. Heatmaps of predicted KEGG pathway completion were separated by function and produced with ComplexHeatmap (64).

Phylogenomic analyses

Phylogenomic analyses were conducted for bacteria using a set of 303 conserved single-copy genes. Groups of orthologous genes were identified with OrthoMCL v2.0.9 (65). Each protein dataset was aligned using MUSCLE v3.8.31 (66) with the default parameters, and the ambiguously aligned regions were automatically selected using the BMGE software (67) for multiple-alignment trimming with the BLOSUM62 similarity matrix. Trimmed alignments were then concatenated into a supermatrix for phylogenetic analysis. The maximum-likelihood analysis was conducted using RAxML version 8.0.20 (68), with an LG amino acid substitution matrix and a Γ model of site heterogeneity with four categories (LG + Γ₄ + F). Bootstrap support for the ML analysis was evaluated with 100 replicates.

dN/dS analyses

The nonsynonymous (dN) and synonymous (dS) substitution rates and positive selection strength (dN/dS) were calculated by KaKs_Calculator (v2.0) (69). A total of 233 single-copy genes were used for the dS and dN/dS analyses. Each pairwise protein sequence was aligned by MUSCLE, and pairwise nucleotide sequence alignments were generated by transforming protein alignments into codon alignments with ParaAT (70). dN/dS ratios were calculated based on pairwise codon alignments using KaKs_Calculator. To perform a comparison between the symbiotic and free-living Fluviibacter, the free-living strain SHINM1 was chosen as the outgroup. dN/dS ratios were calculated as pairwise differences between target strains and the outgroup. The nonparametric Mann-Whitney U test was used to evaluate the significance of differences between the means of the two lineages.

DATA AVAILABILITY

The BioProject accession number for the genome is PRJNA971418. The genome sequence of “Fluviibacter phosphoraccumulans subsp. symbioticus” has been deposited at GenBank under the accession number CP126743. The raw sequences reads have been deposited in Sequence Read Archive (SRA) under accession numbers SRR26455685 and SRR26455686.

SUPPLEMENTAL MATERIAL

The following material is available online at https://doi.org/10.1128/aem.01900-23.

Supplemental figures. aem.01900-23-s0001.docx.

Fig. S1 to S4.

Table S1. aem.01900-23-s0002.docx.

List of detected bacterial symbionts in the Euplotes strains VTN7 and VTN8.

Table S2. aem.01900-23-s0003.docx.

Primers used for determination of the order and direction of the 9 scaffolds.

Table S3. aem.01900-23-s0004.docx.

Details on the pseudogenes.

Table S4. aem.01900-23-s0005.docx.

Details on the mobile elements.

Table S5. aem.01900-23-s0006.docx.

Main functional genomic analysis results.

Table S6. aem.01900-23-s0007.docx.

dN/dS values for 233 orthologous genes in the symbiotic strain EoVTN8 and the free-living strain Rho.

ASM does not own the copyrights to Supplemental Material that may be linked to, or accessed through, an article. The authors have granted ASM a non-exclusive, world-wide license to publish the Supplemental Material files. Please contact the corresponding author directly for reuse.