2025 Volume 40 Issue 2 Article ID: ME25005
Bacteria with endosymbiotic lifestyles often show marked genome reduction. While the shrinkage of genomes in intracellular symbionts of animals, including parasitic bacteria, has been extensively examined, less is known about symbiotic bacteria associated with single-celled eukaryotes. We herein report the genomes of two novel gammaproteobacterial lineages, RS3 and XS4, identified as putative parasitic endosymbionts of the dinoflagellate Citharistes regius. Phylogenetic analyses suggest that RS3 and XS4 belong to the family Fastidiosibacteraceae within the order Beggiatoales, forming independent lineages therein. The genomes of RS3 and XS4 are 529 and 436 kbp in size, respectively, revealing marked reductions from related bacterial genomes. XS4, which has a very reduced genome with a low GC content, uses a different genetic code, in which UGA assigned tryptophan. The small genomes of RS3 and XS4 encode a limited number of proteins, retaining only approximately 20% of the predicted ancestral proteome. Metabolic reconstruction suggests that RS3 and XS4 are parasitic symbionts that are heavily dependent on their host for essential metabolites. Furthermore, we found that the ancestor of both genomes likely acquired an ADP:ATP antiporter gene via horizontal gene transfer, an event that may have enabled their evolution as energy parasites by facilitating the acquisition of ATP from their host. These results on novel bacteria with highly reduced genomes expand our understanding of the phylogenetic and genomic diversities of endosymbiotic bacteria in protists.
In natural environments, many bacteria establish symbiotic relationships with various eukaryotic organisms, including mutualistic, commensalistic, and parasitic interactions. The intracellular bacterial symbionts of arthropods have been extensively examined. They provide metabolic complementation by supplying substances that the host cannot synthesize (e.g., specific amino acids) (McCutcheon and Moran, 2011), confer defensive capabilities against host predators (Oliver et al., 2003), or, in some cases, manipulate host reproductive systems to facilitate their own transmission to the next generation (Katsuma et al., 2022). These symbiotic relationships have a significant impact on the evolution of these bacteria, particularly their genomes. Bacteria that form intimate symbiotic relationships with eukaryotic hosts frequently undergo genome reduction, shedding genes that are unnecessary for their specialized lifestyle over their evolution. This may result in extremely small genomes, sometimes less than 1 Mbp in size (McCutcheon and Moran, 2011). Investigation of these minimal genomes is crucial, not only to elucidate the specific mechanisms underlying each symbiosis, but also to identify the essential gene sets required for bacterial survival (Koonin, 2003), which will provide valuable insights into fundamental biological processes.
To gain a comprehensive understanding of genome reduction in symbiotic bacteria, it is essential to investigate various symbiotic relationships hosted by a wide range of phylogenetic groups of eukaryotes. While previous studies revealed that phylogenetically diverse bacteria engage in symbiotic relationships with eukaryotes and analyzed their genomes (Moran et al., 2008; McCutcheon and Moran, 2011), most of these studies concentrated on symbiotic bacteria associated with animal hosts. However, advances in molecular phylogenetics have demonstrated that multicellular organisms, including animals, represent only a small fraction of overall eukaryotic diversity (Burki et al., 2020). The majority comprises single-celled eukaryotes, or protists, and these organisms frequently harbor symbiotic bacteria (Nowack and Melkonian, 2010; Husnik et al., 2021). Nevertheless, the phylogenetic and genomic diversities of symbiotic bacteria in protists have not yet been examined as extensively as those in animals (Husnik et al., 2021).
Dinoflagellates, belonging to Alveolata, represent a prominent group of protists thriving in various aquatic environments. These organisms are very biodiverse, with some lineages known to have symbiotic bacteria (Gavelis and Gile, 2018; Foster and Zehr, 2019; Husnik et al., 2021); however, the genomic-level characterization of these bacterial symbionts warrants further study. Genome analyses have recently been conducted on cyanobacteria symbiotic with marine Dinophysales dinoflagellates (Nakayama et al., 2019, 2024). Previous studies suggested the presence of additional, non-cyanobacterial symbionts within Dinophysales, including members of Gammaproteobacteria, Alphaproteobacteria, and Betaproteobacteria (Foster et al., 2006; Farnelid et al., 2010). However, only partial sequences of the nifH and 16S rRNA genes have been obtained, which has left their metabolic characteristics largely unknown.
A recent study focused on and sequenced the genome of a cyanobacterial episymbiont of the Dinophysales dinoflagellate Citharistes regius (Nakayama et al., 2024). The cyanobacterial genome was recovered from a genome assembly generated from the amplified DNA of a whole C. regius cell. Our re-analysis of this assembly also revealed additional genomic sequences representing putative endosymbiotic bacteria. We herein report on these genomic sequences and discuss the inferred evolutionary and metabolic characteristics of these bacteria. The results obtained herein revealed that these bacteria, designated as RS3 and XS4, have undergone extreme genome reduction, likely as an adaptation to their putative intracellular lifestyle and intimate association with the host dinoflagellate. They exhibit features of nutritional parasitism, relying heavily on their host for essential metabolites. Additionally, XS4 displays a unique genetic code alteration, further highlighting the evolutionary changes associated with its reduced genome and parasitic lifestyle.
C. regius sampling, whole-genome amplification, and genome assembly were performed as described by Nakayama et al. (2024). Briefly, a cell of C. regius (isolate M16) was found in a seawater sample collected with a plankton net off Shimoda, Shizuoka Prefecture, Japan (N34°29.222′, E139°06.200′). The cell was collected using a glass microcapillary and washed four times: three times by transferring it through droplets of sterilized seawater and then once with PCR-grade sterilized water, to remove residual seawater before placing it in sterilized fresh water for subsequent whole-genome amplification. Sterilized seawater was prepared by filtering natural seawater collected off the Ogasawara Islands, Japan, through a glass microfiber filter with a nominal pore size of 0.7 μm (Whatman GF/F), followed by autoclaving at 121°C for 20 min. Whole-genome amplification was performed using the REPLI-g Single Cell Kit (QIAGEN), and the amplified DNA sample was treated with S1 nuclease (TaKaRa) to reduce branching junctions. Amplified DNA was analyzed using Illumina NovaSeq 6000 and Oxford Nanopore MinION. The genome was assembled de novo in a hybrid manner, incorporating both Illumina short reads and nanopore long reads, using Unicycler (version 0.4.8) with the --no_pilon option as described by Nakayama et al. (2024). A preliminary annotation of the hybrid assembly of the amplified genomes was generated using the DFAST web service (Tanizawa et al., 2016, 2018). Predicted 16S rRNA gene sequences in the assembly were extracted and subsequently analyzed using the NCBI BLAST search service to examine their taxonomic origins and identify potential non-cyanobacterial genomes.
Estimation of genetic codes and genome annotationsThe preliminary genome annotation using the DFAST web service with default options revealed locations in the XS4 genome where open reading frames encoding proteins that are typically conserved in bacteria were interrupted by UGA, one of the stop codons in the standard genetic code. Since XS4 may utilize a different genetic code to that of typical bacteria, we estimated the genetic codes of RS3 and XS4 by focusing on highly conserved amino acid (HCAA) positions within protein sequences conserved across bacterial lineages. We examined the codons used to encode these amino acids in the RS3 and XS4 genomes as follows. We used the tBLASTn program to search the RS3 and XS4 genomes for regions likely to encode highly conserved, single-copy protein-coding genes. Sequence similarity searches of rRNA genes obtained from the preliminary annotation suggested that RS3 and XS4 are both closely related to Fastidiosibacter lacustris (genome accession number: GCF_003428155.1) among the bacterial genomes included in the Genome Taxonomy Database (GTDB) (Parks et al., 2022) r214. Therefore, we used the protein sequences of this genome as queries for tBLASTn. The target protein genes were 105 of the 120 genes encoding proteins used as phylogenetic markers in GTDB (bac120), detected from at least one of RS3 and XS4 (Supplementary Table S1). In parallel, we retrieved the aforementioned 105 proteins from 198 bacterial genome sequences representing diverse lineages included in GTDB r214 (Supplementary Table S1). Each set of 105 homologous protein sequences from the 198 bacteria was aligned using the L-INS-i method in MAFFT (version 7.490) (Katoh and Standley, 2013) to align evolutionarily homologous positions, resulting in 105 protein multiple alignments. Based on the amino acid positions in these alignments, we identified positions where 90% or more of the bacterial sequences had the same amino acids. We assumed that these HCAA were also utilized at the corresponding positions in the orthologous proteins of RS3 and XS4. Based on this assumption, we searched for sites corresponding to the HCAA positions within the homologous protein-coding sequences of the RS3 and XS4 genomes detected by tBLASTn, and investigated the codons used to encode these amino acids in each genome. The number of HCAA positions available for genetic code predictions in RS3 and XS4 were 10,707 and 10,214 sites, respectively. We counted the combinations of amino acid types at HCAA positions and the codons found in the RS3 and XS4 genomes, generating a 64×20 count matrix. This matrix was input into Weblogo3 (https://weblogo.threeplusone.com/create.cgi) (Crooks et al., 2004) to create sequence logos representing amino acid frequencies at HCAA positions corresponding to each codon. The Weblogo3 options used were as follows: Units, probability; Composition, No adjustment for composition.
The genome annotation of RS3 and XS4 was performed using the DFAST web service (version 1.6.0) (Tanizawa et al., 2016, 2018). Analyses of the genetic code suggested that the proteins of the RS3 genome are encoded by the standard bacterial genetic code (NCBI’s translation table 11), while the proteins of the XS4 genome are encoded using a similar genetic code to that of Mycoplasma (NCBI’s translation table 4). Therefore, DFAST was run with genetic code option 11 for RS3 and genetic code option 4 for XS4.
Phylogenetic analyses of 105 genesWe conducted two phylogenomic analyses: the first including the broad member of gammaproteobacterial species as operational taxonomic units (OTUs) and the second focusing specifically on species of the families Francisellaceae and Fastidiosibacteraceae as well as their close relatives. In both analyses, multiple sequence alignments were generated based on the set of 120 phylogenetically informative marker genes (bac120) provided in GTDB r214. In the first analysis, we selected one representative genome from each order within Gammaproteobacteria and one representative genome from each genus within ‘Francisellales’, the order employed within the GTDB. As outgroups, we included one representative genome each from Alphaproteobacteria, Zetaproteobacteria, and Magnetococcia. In cases where a lineage lacked a representative genome or had multiple representative genomes, the genome with the highest completeness was selected. The bac120 homologs of RS3 and XS4 were manually selected from the results of BLASTp searches against the putative proteomes of RS3 and XS4, using the bac120 dataset of the other genomes as a query. Among the bac120 dataset, 105 proteins were detected in at least one of RS3 and XS4 and were used for downstream analyses. The genomes and marker genes used in the analysis are listed in Supplementary Table S2. We aligned each of the 105 protein sequences using MAFFT (version 7.490) with the L-INS-i method and trimmed ambiguously aligned positions with BMGE (version 2.0) using default settings (Criscuolo and Gribaldo, 2010). The remaining sites of the 105 proteins were concatenated into a single alignment comprising 203 taxa and 33,898 amino acid positions. We inferred the maximum likelihood (ML) tree based on the alignment by using IQ-TREE (version 2.2.2.5) (Minh et al., 2020) under the LG+C10+F+I+G model, which was selected by ModelFinder of IQ-TREE (Kalyaanamoorthy et al., 2017) with the “-m TEST -mset LG+C10” option. Statistical support for each bipartition in the ML tree was calculated by 1000-replicate ultrafast bootstrap approximations (Hoang et al., 2018).
In the second phylogenomic analysis, we selected one genome from each species within Francisellaceae, Fastidiosibacteraceae, ‘Piscirickettsiales’, and related environmental clades that were closely related to RS3 and XS4 in the first phylogenomic tree, totaling 44 genomes. The same 105 proteins used in the first analysis were retrieved from the 44 genomes (Supplementary Table S3), as well as the RS3 and XS4 genomes, and the sequences were aligned and trimmed in the same manner as in the first analysis. The 105 genes were concatenated into a single alignment comprising 46 taxa and 36,688 amino acid positions. The ML tree was inferred using IQ-TREE (version 2.2.6) under the LG+C60+F+I+G model, which was selected by ModelFinder with the “-m TEST -mset LG+C60” option. Statistical support for each bipartition in the ML tree was calculated by a 100-replicate non-parametric bootstrap analysis with the ML tree used as the guide for estimating PMSF parameters (Wang et al., 2018).
Orthologous protein analysis and functional annotation of proteinsOrthologous relationships between the proteins encoded in the RS3 and XS4 genomes and those of other Fastidiosibacteraceae bacteria were estimated using OrthoFinder (version 2.5.3) (Emms and Kelly, 2019). The OrthoFinder analysis included all proteins from RS3 and XS4, as well as those from eight species belonging to five genera of Fastidiosibacteraceae bacteria (Supplementary Table S4). To infer the metabolic functions associated with each orthologous protein group (orthogroup), one representative protein sequence was extracted from each orthogroup, creating a single protein set comprising 2,935 sequences. This protein set was analyzed using the KEGG Automatic Annotation Server (KAAS) (Moriya et al., 2007) to estimate the KEGG Orthology (KO) ID corresponding to each orthogroup (Kanehisa et al., 2016). The search program used in the KAAS analysis was BLAST, and the assignment method was the bi-directional best hit (BBH). The metabolic function of each orthogroup was inferred based on the KEGG annotations associated with the assigned KO IDs.
Phylogenetic analysis of the ADP:ATP antiporterAn orthologous protein analysis revealed that the ADP:ATP antiporter found in RS3 and XS4 is not present in other closely related Fastidiosibacteraceae bacteria (Supplementary Table S5). To investigate the evolutionary origin of the ADP:ATP antiporter, we performed a phylogenetic analysis. We searched for the amino acid sequences of the ADP:ATP antiporter in the NCBI refseq database as of July 9, 2024 by BLASTp searches using the ADP:ATP antiporter sequences of RS3/XS4 as queries. We retrieved subject sequences with E-values less than 1e-10, and divided the sequences into gammaproteobacterial, alphaproteobacterial, chlamydial, eukaryotic, and other sequences based on the lineages to which each sequence belongs. Redundancy within the subject sequences of each phylogenetic group was removed by individual clustering analyses using CD-HIT (version 4.8.1) (Fu et al., 2012) with a 70% identity threshold. We aligned the representative sequences of all lineages and ADP:ATP antiporter sequences of RS3/XS4 using MAFFT (version 7.525) with the L-INS-i method. Ambiguously aligned positions were trimmed by trimAl (version 1.4) (Capella-Gutiérrez et al., 2009) with the “-gt 0.85” option. The final alignment comprised 229 sequences with 463 amino acid positions. We inferred the ML tree based on the alignment by using IQ-TREE (version 2.2.6) under the LG+C60+F+G model, which was selected by ModelFinder of IQ-TREE with the “-m TEST -mset LG+C60” option. Statistical support for each bipartition in the ML tree was calculated by 1000-replicate ultrafast bootstrap approximations.
Nakayama et al. (2024) aimed to amplify the genome sequence of a cyanobacterium symbiotic with C. regius by obtaining whole-genome amplification products from a single C. regius cell (Nakayama et al., 2024). Since the entire cell served as the template, the genomes of any organisms associated with this cell were theoretically amplified. Nakayama et al. (2024) focused on symbiotic cyanobacteria and extracted only the cyanobacterial genome. In the present study, we investigated whether the amplification product contained any microbes other than cyanobacteria. The results obtained revealed two circular chromosomes showing homology to Gammaproteobacteria from an assembly of the amplified genome using a single C. regius cell. The two gammaproteobacterial chromosomes have similar, but clearly distinguishable, rRNA gene sequences with 89.95% identity, suggesting that these chromosomes represent the genomes of two phylogenetically related, but separate bacteria (hereinafter referred to as RS3 and XS4). The genome of RS3 was 529 kbp in size with a GC content of 33%, whereas the genome of XS4 was 436 kbp in size with a GC content of 28%.
In addition to these bacterial genomes, we identified a potential complete archaeal genome sequence within the same assembly. This archaeal genome is exceptionally small and displays higher sequence divergence than any known archaeal genomes. We considered this genome to have originated from an archaeon representing a novel lineage. Therefore, discussing this archaeal genome alongside RS3 and XS4 within the same biological and evolutionary context is challenging and will be addressed in a separate publication.
Genetic code in XS4The preliminary annotation of the XS4 genome revealed locations where open reading frames were interrupted by UGA codons, which function as termination signals in the standard genetic code. Further analyses revealed that these UGA codons corresponded to sites at which tryptophan residues are typically conserved in other bacteria and, thus, are likely translated as tryptophan in XS4 (Supplementary Fig. S1, see Materials and Methods). This ‘UGA=W’ reassignment has also been observed in bacteria with a low GC content in their genomes, including Mycoplasma and some arthropod endosymbionts (Knight et al., 2001; McCutcheon and Moran, 2011). In the standard code, tryptophan is assigned by UGG, and the reassignment of UGA from translation termination to tryptophan may be attributed to selection pressure for a low GC content in the genome (Osawa et al., 1992; Knight et al., 2001). Therefore, the XS4 genome was annotated using the Mycoplasma genetic code (NCBI’s translation table 4).
Genomic features and phylogeny of RS3 and XS4The genome annotation of the two genomes revealed 495 protein-coding genes in the RS3 genome and 426 in the XS4 genome. Each genome contained one copy of the 16S, 23S, and 5S rRNA genes, as well as 32 tRNA genes on RS3 and 31 tRNA genes on XS4, along with one transfer-messenger RNA (tmRNA) gene on both genomes (Fig. 1). While the possibility of additional genomic elements, such as plasmids, cannot be excluded, the detection of rRNA genes and a complete set of tRNA genes corresponding to all 20 amino acids on both circular chromosomes strongly suggests that the genome sequences identified in this study represent the complete genomes of RS3 and XS4. Since typical bacterial genomes are several million base pairs in size, the genomes of RS3 and XS4, which are 529 and 436 kbp, respectively, are very compact.
Genome maps of RS3 and XS4. In each map, the outer two circles indicate the positions of protein-coding genes on the plus and minus strands. The third circle shows the positions of the tRNA (green bars) and rRNA (grey bars) genes. The second innermost circle shows the relative G+C content and the innermost circle plots the GC skew.
Previous studies detected nifH genes and 16S rRNA genes from bacteria other than cyanobacteria in Dinophysales dinoflagellates, including Citharistes sp. (Foster et al., 2006; Farnelid et al., 2010). The RS3 and XS4 genomes revealed in the present study lack nitrogen fixation-related genes, including nifH. Therefore, RS3 and XS4 may not have been detected in the survey of bacteria associated with Dinophysales cells using nifH sequences (Farnelid et al., 2010). Comparisons with the 16S rRNA gene sequences obtained from Dinophysales cells in a previous study (Foster et al., 2006) were not possible because the non-cyanobacterial sequences obtained were not deposited in public databases.
A multigene phylogenetic analysis was conducted to infer the phylogenetic positions of RS3 and XS4 among bacterial diversity. The preliminary sequence similarity search suggested that both of these bacteria are related to species in the families Fastidiosibacteraceae and Francisellaceae of Gammaproteobacteria. We reconstructed a phylogenetic tree using genome sequences covering the diversity of Gammaproteobacteria, including Fastidiosibacteraceae and Francisellaceae, based on 105 protein sequences, which are highly conserved across bacterial lineages (Supplementary Fig. S2). The topology of the tree suggested that RS3 and XS4 are both closely related to species within Fastidiosibacteraceae. To gain further insights into the precise phylogenetic positions of RS3 and XS4, an additional phylogenetic analysis was conducted using a dataset including a more comprehensive sampling of fastidiosibacteracean and francisellacean species. Fig. 2 shows the resulting ML phylogenetic tree, revealing that RS3 and XS4 constitute a monophyletic group within a cluster comprising Fastidiosibacteraceae bacteria. The lineage of RS3 and XS4 was shown to be a sister clade to the monophyletic group formed by the genera Cysteiniphilum, Caedibacter, and Fastidiosibacter. This monophyletic group, along with the clade comprising the genera Fangia and Facilibium, represents the monophyletic lineage of fastidiosibacteraceans. Additionally, a clade of Francisellaceae bacteria was inferred as a sister clade to the Fastidiosibacteraceae lineage. All the aforementioned phylogenetic relationships were robustly supported (bootstrap values of 100%).
Maximum likelihood phylogenetic tree based on 105 protein sequences. The tree was inferred using IQ-Tree under the LG+C60+F+I+G model. Non-parametric bootstrap values (100 replicates) were mapped on the tree. The nodes marked with filled circles were supported with 100% bootstrap values. The alignment included 46 taxa and 36,688 amino acid positions.
The branches leading to RS3 and XS4 from the node connecting them were longer than those of other fastidiosibacteraceans, indicating a rapid evolutionary rate for these bacterial genomes. This topology of the tree is reminiscent of the Long Branch Attraction (LBA) artifact, which is a methodological artifact where rapidly evolving lineages are incorrectly inferred to be closely related. However, since the present analysis employed a mixture model accounting for the site heterogeneity of amino acids frequencies, which is more robust to LBA (Quang et al., 2008), we considered this tree topology to represent the most plausible evolutionary hypothesis to date. The topology suggests that RS3 and XS4 are the closest relatives, diverging from a common ancestor shared with the genera Cysteiniphilum, Caedibacter, and Fastidiosibacter.
In the context of symbiotic relationships, including parasitism, Caedibacter, a member of the Fastidiosibacteraceae clade along with RS3 and XS4, is noteworthy. The representative species Caedibacter taeniospiralis is an obligate cytoplasmic endosymbiont found in Paramecium, known for imparting toxic traits, referred to as “killer traits”, to infected Paramecium against non-infected cells (Kusch et al., 2000; Grosser et al., 2018). Although most members of Fastidiosibacteraceae have been discovered in the past two decades, the research on Paramecium killer traits dates back to the mid-20th century (Sonneborn, 1943; Preer et al., 1974), making Caedibacter the most well-known genus within this clade. The dinoflagellate C. regius, the source of the genome amplification product from which the RS3 and XS4 genomes were identified, is a protist belonging to Alveolata, as is Paramecium, implying a similar symbiotic association. Nevertheless, our phylogenetic analysis revealed that the Caedibacter and RS3/XS4 lineages were not closely related. In addition to Caedibacter, the marine fastidiosibacteracean genus Cysteiniphilum, which has been reported to parasitize humans (Xu et al., 2021), is not the closest relative of the RS3/XS4 lineage, but is a sister clade to Caedibacter.
Genome reduction and codon reassignmentA comparative analysis of genome sizes based on the evolutionary relationships inferred by the phylogenomic tree revealed marked reductions in the genome sizes of RS3 and XS4. Fig. 3 shows a comparison of the genome size and GC content of each bacterial genome along with phylogenetic relationships; the genome sizes of Fangia and Facilibium, which represent the most basal branches of the Fastidiosibacteraceae lineage, are both slightly less than 3 Mbp, while those of Cysteiniphilum and Fastidiosibacter, which are members of the sister clade of the RS3/XS4 lineage, range from 2–3 Mbp. In consideration of the genome sizes of the bacteria surrounding RS3 and XS4 in the phylogeny, the ancestral bacteria of RS3 and XS4 appeared to possess a genome of at least 2 Mbp. In contrast, the genome sizes of RS3 and XS4 are 529 and 436 kbp, respectively, indicating that their genomes have undergone severe reductions to less than 25% of their ancestral size. Although the obligate cytoplasmic endosymbiont C. taeniospiralis is also known to possess a relatively compact genome, it is more than twice the size of the RS3/XS4 genomes.
Comparison of the size and GC content of genomes of RS3, XS4, and their related bacteria. The cladogram on the left side shows phylogenetic relationships as shown in Fig. 2. The bar graph on the right displays genome sizes, while the plot in the center indicates the GC content of each genome.
Another notable feature shared by RS3 and XS4 is their low GC content (Fig. 3). The GC content of the RS3 genome of 33.04% is lower than those of other bacterial genomes of Fastidiosibacteraceae, except for XS4. The XS4 genome exhibits an even lower GC content of 28.16%, which is less than those observed in the genomes of Francisellaceae bacteria or those of fastidiosibacteracean genomes. Previous genomic studies demonstrated a universal bias towards a lower GC content in reduced genomes (McCutcheon and Moran, 2011), and a similar evolutionary change appears to have occurred in the genomes of RS3 and XS4. Changes in the GC content in reduced genomes are consistent with the result showing that the XS4 genome, with its lower GC content, was smaller than the RS3 genome.
In bacterial genomes that underwent severe size reductions and received a strong bias towards a low GC content, UGA has often been reassigned from the termination signal to tryptophan (Ohama et al., 2008). XS4 most likely uses the deviant genetic code, the “UGA=W code”, which assigns UGA to tryptophan, while RS3 uses the standard code (i.e., UGA as a termination signal). In bacteria with the standard code, two class-I release factors (i.e., RF1 and RF2), which promote the release of the nascent peptide from the ribosome responding to a stop codon at the A site, bear distinct codon specificities: RF1 recognized UAA and UAG, and RF2 recognized UAA and UGA (Caskey et al., 1969). Therefore, there are two possibilities for RF2 in XS4. XS4 may have discarded RF2 during the UGA reassignment because RF1 is sufficient to cover UAA and UAG in this bacterium. Alternatively, RF2 persists, but may have lost codon specificity to UGA in XS4. The XS4 genome possesses the gene encoding RF2 and, thus, this protein is anticipated to be specific for UAA. We compared the putative amino acid sequence of RF2 in XS4 to that of RF2 in RS3 (note that the latter is assumed to recognize both UAA and UGA as termination signals). Unfortunately, the comparison provided no insights into the amino acid(s) that caused the change in codon specificity in RF2 in XS4 (Supplementary Fig. S3). The amino acids of RF2, which were shown to interact directly with the UGA and UAA triplets in the crystal of the translation termination complex comprising the ribosome, a triplet (UAA or UGA), and RF2 (Korostelev et al., 2008; Weixlbaumer et al., 2008), were conserved in both XS4 and RS3 proteins.
Responding to the UGA reassignment, XS4 requires a tryptophan tRNA (tRNATrp), of which the anticodon binds to UGA. We found only a single tRNATrp with the anticodon CCA (tRNA
Predicted secondary structures of tRNA
Reflecting their small genome sizes, the genomes of RS3 and XS4 encode a correspondingly limited number of proteins: 495 and 426 protein-coding genes, respectively. In contrast, other Fastidiosibacteraceae bacteria, including species of Cysteiniphilum, Facilibium, Fangia, and Fastidiosibacter, have approximately 2,000–3,000 protein-coding genes in their genomes. Even the relatively compact genome of C. taeniospiralis contains 1,218 proteins. We conducted a comprehensive analysis of orthologous relationships between the proteins of RS3/XS4 and those of related bacteria to infer the evolution of the RS3 and XS4 proteomes in the light of the organismal phylogenetic relationship. An orthologous protein analysis revealed that nearly all proteins observed in the RS3/XS4 genomes (RS3: 480 out of 495 proteins, XS4: 408 out of 426 proteins) possess orthologous counterparts in their closely related bacterial species. By examining the proteomes of the phylogenetically related taxa, we inferred the minimal proteome of the common ancestor of RS3 and XS4. This ancestral proteome was defined as the set of proteins that satisfies two criteria: the proteins found in 1) either Fangia or Facilibium, representing the most basal branch of the Fastidiosibacteraceae clade, and 2) at least one member of the sister lineage to RS3/XS4, comprising Cysteiniphilum, Fastidiosibacter, or Caedibacter. The minimal ancestral proteome was estimated to comprise 2,151 proteins, with RS3 and XS4 retaining 477 and 402 of these proteins, respectively, which corresponds to only 22.1 and 18.6%, respectively, of the ancestral proteome. Additionally, we assigned functional annotations using the KO database (Kanehisa et al., 2023) to each orthologous protein group in order to infer the metabolic capacities of RS3 and XS4 and the evolution of their proteomes. In comparisons with the ancestral proteome, the majority of the metabolic functions lost were shared between RS3 and XS4 (Supplementary Table S5), suggesting that these losses occurred at the stage of their common ancestor (Fig. 5). Among the lost metabolic functions, transporters and two-component regulatory systems were the most affected functional categories, which are typically lost in symbiotic bacteria (Shigenobu et al., 2000; McCutcheon and Moran, 2011). RS3 and XS4 both lacked biosynthetic pathways for essential metabolites that are indispensable for independent growth. This was particularly evident in the biosynthesis of amino acids, where both bacteria lost all major amino acid biosynthesis pathways. Additionally, both bacteria lacked genes for the synthesis of pyrimidine bases, heme, the majority of vitamins, most glycolytic enzymes, and 50% of TCA cycle enzyme genes. Moreover, XS4, with a more reductive genome than RS3, lost additional metabolic pathways, including biosynthetic pathways for purine bases, chorismic acid, and riboflavin, and the entire set of the remaining TCA cycle enzyme genes, which were conserved in RS3. This extensive loss of metabolic functions indicates that RS3 and XS4 rely on their host for a significant amount of the metabolites required for their growth. Moreover, a distinguishing feature of RS3 and XS4 is the absence of biosynthetic pathways for peptidoglycan and lipopolysaccharide (LPS). While the morphologies and symbiotic natures of these bacteria remain unclear, the loss of these key cell surface components suggests that both bacteria are intracellular symbionts (or parasites).
Predicted evolution of metabolic functions in the lineage of RS3 and XS4. Numbers above branches indicate the number of proteins derived from the minimal ancestral proteome (2,151 proteins), which was estimated based on the orthologous protein analysis. The percentages below the tips of the branches for RS3 and XS4 represent the percentage of proteins retained from the minimal ancestral proteome in each genome. Metabolic functions lost or gained during the evolutionary process of each bacterium are shown.
A number of examples in which intracellular symbiotic bacteria synthesize and supply metabolic products that the host organism is unable to produce have been reported (McCutcheon and Moran, 2011). In such cases, the host and symbiont both benefit from the relationship. Nevertheless, the relationship between RS3/XS4 and their host organisms is not necessarily analogous to the aforementioned cases. A lack of knowledge of the metabolic characteristics of the presumed host, C. regius, makes it challenging to discuss the relationship in detail. However, RS3/XS4 lost almost all pathways for synthesizing amino acids and vitamins, which makes it unlikely that they produce and supply the necessary substances for the host. On the other hand, it cannot be entirely ruled out that the residual biosynthetic pathways in RS3/XS4, such as those for folic acid and fatty acids, still confer benefits to the host. This possibility may be examined once the metabolic functions of C. regius are elucidated in future studies. Regardless, it is reasonable to assume that RS3/XS4 rely heavily on their host for the majority of the substances constituting their cells, characterizing these bacteria as primarily nutritional parasites. In this context, a protein-coding gene of interest has been identified in the genomes of these two bacteria. The ADP:ATP antiporter (RS3CREG01_2930 and XS4CREG01_3620; KO ID: K03301), which was detected from the proteomes of RS3 and XS4, was not observed in any proteomes of Fastidiosibacteraceae bacteria examined in the present study and, thus, was not included in the estimated ancestral proteome of RS3/XS4. A phylogenetic analysis of the antiporter from various bacteria as well as those of RS3 and XS4 (Supplementary Fig. S4) strongly supported the monophyletic relationship of the proteins from RS3 and XS4. Additionally, the RS3/XS4 clade was shown to be monophyletic with alphaproteobacterial sequences; however, the relationship was weakly supported. While homologous proteins were also identified from gammaproteobacterial species other than fastidiosibacteraceans and francisellaceans, none of these were suggested to be closely related to the RS3/XS4 sequences (Supplementary Fig. S4). The phylogenetic relationship indicates that the common ancestor of RS3/XS4 acquired the ADP:ATP antiporter gene through horizontal gene transfer; however, the donor has yet to be identified (Fig. 5). It is important to note that several notorious intracellular parasitic bacteria, such as Chlamydia and Rickettsia, are known to engage in energy parasitism. These parasites utilize the ADP:ATP antiporter as a key molecule for obtaining energy from their host organisms (Schmitz-Esser et al., 2004). Given these characteristics, the acquisition of this gene may enable RS3/XS4 to engage in energy parasitism. Similar genomic features, such as marked genome reductions and the presence of horizontally acquired ADP:ATP antiporters, have been reported in the denitrifying endosymbiont of the anaerobic ciliate, Candidatus Azoamicus ciliaticola (Graf et al., 2021), and related bacteria capable of both denitrification and aerobic respiration (Speth et al., 2024). However, in contrast to RS3/XS4, these bacteria are considered to function primarily as respiratory symbionts, generating ATP via their respiratory chains and providing it to their hosts. Although RS3/XS4 also encode genes for respiration (e.g., genes for NADH dehydrogenase and ATP synthase), we did not detect the metabolic pathways typically involved in generating sufficient reducing equivalents for the electron transport chain, including the denitrification pathway. This suggests that RS3/XS4 followed a different evolutionary trajectory from these respiratory symbionts.
We herein report the genomes of RS3 and XS4, which were identified using amplified DNA obtained from the dinoflagellate C. regius cell. RS3 and XS4 are regarded as novel symbiotic lineages belonging to Fastidiosibacteraceae, Gammaproteobacteria. These bacteria possess highly reduced genomes (529 kbp for RS3 and 436 kbp for XS4), which, to the best of our knowledge, are the smallest among the genomes sequenced at the chromosomal level within both Fastidiosibacteraceae and the encompassing order Beggiatoales (or Thiotrichales). Predicted proteomes of the two bacteria strongly suggest that RS3 and XS4 are adapted to an intracellular lifestyle. Furthermore, their markedly constrained biosynthetic capabilities indicate that these bacteria exist primarily as nutritional parasites within host cells. However, further research is needed to confirm their physical relationship with C. regius. Specifically, fluorescence in situ hybridization (FISH) or other microscopic techniques are required to visualize the relationship of RS3/XS4 with the C. regius cell. The genomic characteristics of RS3 and XS4 revealed in this analysis not only expand our understanding of genome diversity within Fastidiosibacteraceae, but also contribute to our knowledge of the phylogenetic and genomic diversities of symbiotic bacteria in protists. Further studies on bacteria symbiotic with protists are anticipated to yield even more insights.
Nakayama, T., Harada, R., Yabuki, A., Nomura, M., Shiba, K., Inaba, K., and Inagaki, Y. (2025) Marked Genome Reduction Driven by a Parasitic Lifestyle: Two Complete Genomes of Endosymbiotic Bacteria Possibly Hosted by a Dinoflagellate. Microbes Environ 40: ME25005.
https://doi.org/10.1264/jsme2.ME25005
This work was supported by JSPS KAKENHI Grant Numbers JP16H06280, 20H03305, 18KK0203, 21K15131, 23K27226, 24K09587, 22KJ0401, and BPI06050, as well as by the World Premier International Research Center Initiative (WPI Initiative), MEXT, Japan. R.H. was supported by JSPS Overseas Research Fellowships. Computations in this study were partially performed on the NIG supercomputer at ROIS National Institute of Genetics.
Data availabilityThe annotated genome sequences of RS3 and XS4 have been deposited in DDBJ/ENA/GenBank under accession numbers AP038918 and AP038919. Data related to phylogenetic and orthologous protein analyses in this study are available on Zenodo (DOI: 10.5281/zenodo.14597766).
